{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9999322355492308, "global_step": 11067, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.504504504504505e-07, "loss": 7.2121, "step": 1 }, { "epoch": 0.0, "learning_rate": 9.00900900900901e-07, "loss": 7.3098, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.3513513513513515e-06, "loss": 6.7926, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.801801801801802e-06, "loss": 7.3778, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.2522522522522524e-06, "loss": 6.0732, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.702702702702703e-06, "loss": 5.6767, "step": 6 }, { "epoch": 0.0, "learning_rate": 3.153153153153153e-06, "loss": 5.5074, "step": 7 }, { "epoch": 0.0, "learning_rate": 3.603603603603604e-06, "loss": 5.2564, "step": 8 }, { "epoch": 0.0, "learning_rate": 4.0540540540540545e-06, "loss": 4.903, "step": 9 }, { "epoch": 0.0, "learning_rate": 4.504504504504505e-06, "loss": 4.1206, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.954954954954955e-06, "loss": 3.9064, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.405405405405406e-06, "loss": 3.3206, "step": 12 }, { "epoch": 0.0, "learning_rate": 5.855855855855856e-06, "loss": 2.9629, "step": 13 }, { "epoch": 0.0, "learning_rate": 6.306306306306306e-06, "loss": 2.6263, "step": 14 }, { "epoch": 0.0, "learning_rate": 6.7567567567567575e-06, "loss": 2.3534, "step": 15 }, { "epoch": 0.0, "learning_rate": 7.207207207207208e-06, "loss": 2.3459, "step": 16 }, { "epoch": 0.0, "learning_rate": 7.657657657657658e-06, "loss": 2.2729, "step": 17 }, { "epoch": 0.0, "learning_rate": 8.108108108108109e-06, "loss": 2.3194, "step": 18 }, { "epoch": 0.0, "learning_rate": 8.558558558558558e-06, "loss": 2.2588, "step": 19 }, { "epoch": 0.0, "learning_rate": 9.00900900900901e-06, "loss": 2.2146, "step": 20 }, { "epoch": 0.0, "learning_rate": 9.45945945945946e-06, "loss": 2.1797, "step": 21 }, { "epoch": 0.0, "learning_rate": 9.90990990990991e-06, "loss": 2.1729, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.0360360360360361e-05, "loss": 2.0443, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.0810810810810812e-05, "loss": 2.0647, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.1261261261261261e-05, "loss": 2.1158, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.1711711711711713e-05, "loss": 2.1053, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.2162162162162164e-05, "loss": 2.0793, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.2612612612612611e-05, "loss": 2.0453, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.3063063063063064e-05, "loss": 2.0288, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.3513513513513515e-05, "loss": 2.0081, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.3963963963963963e-05, "loss": 2.023, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.4414414414414416e-05, "loss": 2.1559, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.4864864864864867e-05, "loss": 2.0357, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.5315315315315316e-05, "loss": 1.9564, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.5765765765765765e-05, "loss": 2.0186, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.6216216216216218e-05, "loss": 1.9822, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.6666666666666667e-05, "loss": 2.0231, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.7117117117117117e-05, "loss": 2.0563, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.756756756756757e-05, "loss": 1.9816, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.801801801801802e-05, "loss": 1.9997, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.846846846846847e-05, "loss": 2.0906, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.891891891891892e-05, "loss": 2.0049, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.936936936936937e-05, "loss": 2.0028, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.981981981981982e-05, "loss": 2.0399, "step": 44 }, { "epoch": 0.0, "learning_rate": 2.0270270270270273e-05, "loss": 2.0554, "step": 45 }, { "epoch": 0.0, "learning_rate": 2.0720720720720722e-05, "loss": 1.934, "step": 46 }, { "epoch": 0.0, "learning_rate": 2.117117117117117e-05, "loss": 2.026, "step": 47 }, { "epoch": 0.0, "learning_rate": 2.1621621621621624e-05, "loss": 1.9811, "step": 48 }, { "epoch": 0.0, "learning_rate": 2.2072072072072073e-05, "loss": 1.9333, "step": 49 }, { "epoch": 0.0, "learning_rate": 2.2522522522522523e-05, "loss": 2.0218, "step": 50 }, { "epoch": 0.0, "learning_rate": 2.2972972972972976e-05, "loss": 1.976, "step": 51 }, { "epoch": 0.0, "learning_rate": 2.3423423423423425e-05, "loss": 1.9823, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.3873873873873874e-05, "loss": 1.9254, "step": 53 }, { "epoch": 0.0, "learning_rate": 2.4324324324324327e-05, "loss": 1.9724, "step": 54 }, { "epoch": 0.0, "learning_rate": 2.4774774774774777e-05, "loss": 2.0665, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.5225225225225222e-05, "loss": 1.9165, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.5675675675675675e-05, "loss": 1.9989, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.6126126126126128e-05, "loss": 1.9171, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.6576576576576577e-05, "loss": 1.9889, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.702702702702703e-05, "loss": 1.9165, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.7477477477477483e-05, "loss": 1.9161, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.7927927927927926e-05, "loss": 1.9632, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.8378378378378378e-05, "loss": 1.9144, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.882882882882883e-05, "loss": 1.9379, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.927927927927928e-05, "loss": 2.0306, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.9729729729729733e-05, "loss": 1.9367, "step": 66 }, { "epoch": 0.01, "learning_rate": 3.0180180180180183e-05, "loss": 1.9249, "step": 67 }, { "epoch": 0.01, "learning_rate": 3.063063063063063e-05, "loss": 1.978, "step": 68 }, { "epoch": 0.01, "learning_rate": 3.108108108108108e-05, "loss": 2.0517, "step": 69 }, { "epoch": 0.01, "learning_rate": 3.153153153153153e-05, "loss": 1.9659, "step": 70 }, { "epoch": 0.01, "learning_rate": 3.198198198198199e-05, "loss": 1.9414, "step": 71 }, { "epoch": 0.01, "learning_rate": 3.2432432432432436e-05, "loss": 1.9714, "step": 72 }, { "epoch": 0.01, "learning_rate": 3.2882882882882886e-05, "loss": 2.0321, "step": 73 }, { "epoch": 0.01, "learning_rate": 3.3333333333333335e-05, "loss": 2.0236, "step": 74 }, { "epoch": 0.01, "learning_rate": 3.3783783783783784e-05, "loss": 2.0571, "step": 75 }, { "epoch": 0.01, "learning_rate": 3.4234234234234234e-05, "loss": 2.0084, "step": 76 }, { "epoch": 0.01, "learning_rate": 3.468468468468469e-05, "loss": 2.0354, "step": 77 }, { "epoch": 0.01, "learning_rate": 3.513513513513514e-05, "loss": 1.9906, "step": 78 }, { "epoch": 0.01, "learning_rate": 3.558558558558558e-05, "loss": 1.9584, "step": 79 }, { "epoch": 0.01, "learning_rate": 3.603603603603604e-05, "loss": 1.9289, "step": 80 }, { "epoch": 0.01, "learning_rate": 3.648648648648649e-05, "loss": 1.9024, "step": 81 }, { "epoch": 0.01, "learning_rate": 3.693693693693694e-05, "loss": 1.9436, "step": 82 }, { "epoch": 0.01, "learning_rate": 3.738738738738739e-05, "loss": 1.9631, "step": 83 }, { "epoch": 0.01, "learning_rate": 3.783783783783784e-05, "loss": 1.9866, "step": 84 }, { "epoch": 0.01, "learning_rate": 3.8288288288288285e-05, "loss": 1.9061, "step": 85 }, { "epoch": 0.01, "learning_rate": 3.873873873873874e-05, "loss": 1.9148, "step": 86 }, { "epoch": 0.01, "learning_rate": 3.918918918918919e-05, "loss": 1.972, "step": 87 }, { "epoch": 0.01, "learning_rate": 3.963963963963964e-05, "loss": 1.9616, "step": 88 }, { "epoch": 0.01, "learning_rate": 4.0090090090090096e-05, "loss": 1.9962, "step": 89 }, { "epoch": 0.01, "learning_rate": 4.0540540540540545e-05, "loss": 1.9624, "step": 90 }, { "epoch": 0.01, "learning_rate": 4.099099099099099e-05, "loss": 1.944, "step": 91 }, { "epoch": 0.01, "learning_rate": 4.1441441441441444e-05, "loss": 1.9401, "step": 92 }, { "epoch": 0.01, "learning_rate": 4.189189189189189e-05, "loss": 1.9866, "step": 93 }, { "epoch": 0.01, "learning_rate": 4.234234234234234e-05, "loss": 1.9642, "step": 94 }, { "epoch": 0.01, "learning_rate": 4.27927927927928e-05, "loss": 1.9234, "step": 95 }, { "epoch": 0.01, "learning_rate": 4.324324324324325e-05, "loss": 1.9963, "step": 96 }, { "epoch": 0.01, "learning_rate": 4.369369369369369e-05, "loss": 1.8854, "step": 97 }, { "epoch": 0.01, "learning_rate": 4.414414414414415e-05, "loss": 1.9326, "step": 98 }, { "epoch": 0.01, "learning_rate": 4.4594594594594596e-05, "loss": 1.8964, "step": 99 }, { "epoch": 0.01, "learning_rate": 4.5045045045045046e-05, "loss": 2.0114, "step": 100 }, { "epoch": 0.01, "learning_rate": 4.54954954954955e-05, "loss": 1.8703, "step": 101 }, { "epoch": 0.01, "learning_rate": 4.594594594594595e-05, "loss": 1.9675, "step": 102 }, { "epoch": 0.01, "learning_rate": 4.6396396396396394e-05, "loss": 1.9123, "step": 103 }, { "epoch": 0.01, "learning_rate": 4.684684684684685e-05, "loss": 1.9869, "step": 104 }, { "epoch": 0.01, "learning_rate": 4.72972972972973e-05, "loss": 1.9879, "step": 105 }, { "epoch": 0.01, "learning_rate": 4.774774774774775e-05, "loss": 2.019, "step": 106 }, { "epoch": 0.01, "learning_rate": 4.8198198198198205e-05, "loss": 2.0097, "step": 107 }, { "epoch": 0.01, "learning_rate": 4.8648648648648654e-05, "loss": 1.9105, "step": 108 }, { "epoch": 0.01, "learning_rate": 4.90990990990991e-05, "loss": 1.9385, "step": 109 }, { "epoch": 0.01, "learning_rate": 4.954954954954955e-05, "loss": 1.9129, "step": 110 }, { "epoch": 0.01, "learning_rate": 5e-05, "loss": 1.9007, "step": 111 }, { "epoch": 0.01, "learning_rate": 4.999999897220688e-05, "loss": 1.9601, "step": 112 }, { "epoch": 0.01, "learning_rate": 4.999999588882759e-05, "loss": 1.9668, "step": 113 }, { "epoch": 0.01, "learning_rate": 4.9999990749862386e-05, "loss": 1.9602, "step": 114 }, { "epoch": 0.01, "learning_rate": 4.999998355531169e-05, "loss": 1.9331, "step": 115 }, { "epoch": 0.01, "learning_rate": 4.999997430517611e-05, "loss": 1.99, "step": 116 }, { "epoch": 0.01, "learning_rate": 4.999996299945638e-05, "loss": 1.9955, "step": 117 }, { "epoch": 0.01, "learning_rate": 4.9999949638153446e-05, "loss": 1.9119, "step": 118 }, { "epoch": 0.01, "learning_rate": 4.9999934221268406e-05, "loss": 1.9546, "step": 119 }, { "epoch": 0.01, "learning_rate": 4.999991674880252e-05, "loss": 1.9704, "step": 120 }, { "epoch": 0.01, "learning_rate": 4.9999897220757244e-05, "loss": 1.9467, "step": 121 }, { "epoch": 0.01, "learning_rate": 4.999987563713415e-05, "loss": 1.9429, "step": 122 }, { "epoch": 0.01, "learning_rate": 4.9999851997935046e-05, "loss": 1.9966, "step": 123 }, { "epoch": 0.01, "learning_rate": 4.999982630316185e-05, "loss": 1.9012, "step": 124 }, { "epoch": 0.01, "learning_rate": 4.999979855281669e-05, "loss": 1.8853, "step": 125 }, { "epoch": 0.01, "learning_rate": 4.999976874690185e-05, "loss": 1.8629, "step": 126 }, { "epoch": 0.01, "learning_rate": 4.999973688541977e-05, "loss": 1.8513, "step": 127 }, { "epoch": 0.01, "learning_rate": 4.9999702968373074e-05, "loss": 1.9323, "step": 128 }, { "epoch": 0.01, "learning_rate": 4.9999666995764556e-05, "loss": 1.8662, "step": 129 }, { "epoch": 0.01, "learning_rate": 4.999962896759717e-05, "loss": 1.9073, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.9999588883874024e-05, "loss": 1.9364, "step": 131 }, { "epoch": 0.01, "learning_rate": 4.9999546744598455e-05, "loss": 1.9557, "step": 132 }, { "epoch": 0.01, "learning_rate": 4.999950254977389e-05, "loss": 1.9359, "step": 133 }, { "epoch": 0.01, "learning_rate": 4.999945629940398e-05, "loss": 1.9777, "step": 134 }, { "epoch": 0.01, "learning_rate": 4.9999407993492533e-05, "loss": 1.9008, "step": 135 }, { "epoch": 0.01, "learning_rate": 4.999935763204351e-05, "loss": 1.8562, "step": 136 }, { "epoch": 0.01, "learning_rate": 4.999930521506105e-05, "loss": 1.8455, "step": 137 }, { "epoch": 0.01, "learning_rate": 4.999925074254947e-05, "loss": 1.8589, "step": 138 }, { "epoch": 0.01, "learning_rate": 4.9999194214513246e-05, "loss": 2.0503, "step": 139 }, { "epoch": 0.01, "learning_rate": 4.999913563095703e-05, "loss": 1.9121, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.999907499188563e-05, "loss": 2.0052, "step": 141 }, { "epoch": 0.01, "learning_rate": 4.9999012297304044e-05, "loss": 1.8833, "step": 142 }, { "epoch": 0.01, "learning_rate": 4.999894754721742e-05, "loss": 1.9253, "step": 143 }, { "epoch": 0.01, "learning_rate": 4.999888074163108e-05, "loss": 1.8383, "step": 144 }, { "epoch": 0.01, "learning_rate": 4.999881188055052e-05, "loss": 1.95, "step": 145 }, { "epoch": 0.01, "learning_rate": 4.99987409639814e-05, "loss": 1.8846, "step": 146 }, { "epoch": 0.01, "learning_rate": 4.999866799192956e-05, "loss": 1.9215, "step": 147 }, { "epoch": 0.01, "learning_rate": 4.999859296440098e-05, "loss": 1.9449, "step": 148 }, { "epoch": 0.01, "learning_rate": 4.999851588140185e-05, "loss": 1.8953, "step": 149 }, { "epoch": 0.01, "learning_rate": 4.99984367429385e-05, "loss": 1.875, "step": 150 }, { "epoch": 0.01, "learning_rate": 4.999835554901743e-05, "loss": 1.9905, "step": 151 }, { "epoch": 0.01, "learning_rate": 4.9998272299645334e-05, "loss": 1.8581, "step": 152 }, { "epoch": 0.01, "learning_rate": 4.9998186994829036e-05, "loss": 1.8699, "step": 153 }, { "epoch": 0.01, "learning_rate": 4.9998099634575565e-05, "loss": 1.8822, "step": 154 }, { "epoch": 0.01, "learning_rate": 4.99980102188921e-05, "loss": 1.93, "step": 155 }, { "epoch": 0.01, "learning_rate": 4.999791874778599e-05, "loss": 1.9427, "step": 156 }, { "epoch": 0.01, "learning_rate": 4.999782522126476e-05, "loss": 1.8032, "step": 157 }, { "epoch": 0.01, "learning_rate": 4.999772963933609e-05, "loss": 1.8562, "step": 158 }, { "epoch": 0.01, "learning_rate": 4.999763200200785e-05, "loss": 1.9623, "step": 159 }, { "epoch": 0.01, "learning_rate": 4.999753230928807e-05, "loss": 1.9133, "step": 160 }, { "epoch": 0.01, "learning_rate": 4.999743056118494e-05, "loss": 1.9268, "step": 161 }, { "epoch": 0.01, "learning_rate": 4.999732675770683e-05, "loss": 1.9666, "step": 162 }, { "epoch": 0.01, "learning_rate": 4.9997220898862274e-05, "loss": 1.88, "step": 163 }, { "epoch": 0.01, "learning_rate": 4.9997112984659974e-05, "loss": 1.9035, "step": 164 }, { "epoch": 0.01, "learning_rate": 4.99970030151088e-05, "loss": 1.9166, "step": 165 }, { "epoch": 0.01, "learning_rate": 4.9996890990217804e-05, "loss": 1.8882, "step": 166 }, { "epoch": 0.02, "learning_rate": 4.9996776909996196e-05, "loss": 1.8682, "step": 167 }, { "epoch": 0.02, "learning_rate": 4.999666077445335e-05, "loss": 1.9347, "step": 168 }, { "epoch": 0.02, "learning_rate": 4.999654258359881e-05, "loss": 1.8546, "step": 169 }, { "epoch": 0.02, "learning_rate": 4.999642233744231e-05, "loss": 1.8101, "step": 170 }, { "epoch": 0.02, "learning_rate": 4.999630003599372e-05, "loss": 1.9466, "step": 171 }, { "epoch": 0.02, "learning_rate": 4.99961756792631e-05, "loss": 1.8818, "step": 172 }, { "epoch": 0.02, "learning_rate": 4.999604926726069e-05, "loss": 1.8154, "step": 173 }, { "epoch": 0.02, "learning_rate": 4.999592079999688e-05, "loss": 1.8453, "step": 174 }, { "epoch": 0.02, "learning_rate": 4.9995790277482206e-05, "loss": 1.8827, "step": 175 }, { "epoch": 0.02, "learning_rate": 4.9995657699727435e-05, "loss": 1.8534, "step": 176 }, { "epoch": 0.02, "learning_rate": 4.999552306674344e-05, "loss": 1.8396, "step": 177 }, { "epoch": 0.02, "learning_rate": 4.999538637854132e-05, "loss": 1.8643, "step": 178 }, { "epoch": 0.02, "learning_rate": 4.9995247635132294e-05, "loss": 1.8572, "step": 179 }, { "epoch": 0.02, "learning_rate": 4.999510683652777e-05, "loss": 1.8368, "step": 180 }, { "epoch": 0.02, "learning_rate": 4.999496398273933e-05, "loss": 1.9118, "step": 181 }, { "epoch": 0.02, "learning_rate": 4.999481907377872e-05, "loss": 1.8214, "step": 182 }, { "epoch": 0.02, "learning_rate": 4.999467210965785e-05, "loss": 1.7981, "step": 183 }, { "epoch": 0.02, "learning_rate": 4.999452309038881e-05, "loss": 1.8707, "step": 184 }, { "epoch": 0.02, "learning_rate": 4.999437201598385e-05, "loss": 1.8954, "step": 185 }, { "epoch": 0.02, "learning_rate": 4.99942188864554e-05, "loss": 1.8505, "step": 186 }, { "epoch": 0.02, "learning_rate": 4.9994063701816035e-05, "loss": 1.8352, "step": 187 }, { "epoch": 0.02, "learning_rate": 4.999390646207852e-05, "loss": 1.8568, "step": 188 }, { "epoch": 0.02, "learning_rate": 4.99937471672558e-05, "loss": 1.828, "step": 189 }, { "epoch": 0.02, "learning_rate": 4.999358581736095e-05, "loss": 1.8336, "step": 190 }, { "epoch": 0.02, "learning_rate": 4.999342241240725e-05, "loss": 1.8502, "step": 191 }, { "epoch": 0.02, "learning_rate": 4.999325695240813e-05, "loss": 1.8512, "step": 192 }, { "epoch": 0.02, "learning_rate": 4.99930894373772e-05, "loss": 1.8194, "step": 193 }, { "epoch": 0.02, "learning_rate": 4.999291986732823e-05, "loss": 1.7967, "step": 194 }, { "epoch": 0.02, "learning_rate": 4.999274824227516e-05, "loss": 1.8533, "step": 195 }, { "epoch": 0.02, "learning_rate": 4.999257456223211e-05, "loss": 1.8663, "step": 196 }, { "epoch": 0.02, "learning_rate": 4.999239882721335e-05, "loss": 1.7805, "step": 197 }, { "epoch": 0.02, "learning_rate": 4.999222103723334e-05, "loss": 1.8628, "step": 198 }, { "epoch": 0.02, "learning_rate": 4.999204119230669e-05, "loss": 1.8552, "step": 199 }, { "epoch": 0.02, "learning_rate": 4.9991859292448194e-05, "loss": 1.7754, "step": 200 }, { "epoch": 0.02, "learning_rate": 4.9991675337672804e-05, "loss": 1.8343, "step": 201 }, { "epoch": 0.02, "learning_rate": 4.999148932799565e-05, "loss": 1.8607, "step": 202 }, { "epoch": 0.02, "learning_rate": 4.999130126343201e-05, "loss": 1.8479, "step": 203 }, { "epoch": 0.02, "learning_rate": 4.999111114399737e-05, "loss": 1.9056, "step": 204 }, { "epoch": 0.02, "learning_rate": 4.999091896970736e-05, "loss": 1.829, "step": 205 }, { "epoch": 0.02, "learning_rate": 4.9990724740577766e-05, "loss": 1.7785, "step": 206 }, { "epoch": 0.02, "learning_rate": 4.9990528456624564e-05, "loss": 1.8631, "step": 207 }, { "epoch": 0.02, "learning_rate": 4.99903301178639e-05, "loss": 1.8724, "step": 208 }, { "epoch": 0.02, "learning_rate": 4.9990129724312077e-05, "loss": 1.8224, "step": 209 }, { "epoch": 0.02, "learning_rate": 4.998992727598557e-05, "loss": 1.8784, "step": 210 }, { "epoch": 0.02, "learning_rate": 4.9989722772901024e-05, "loss": 1.8072, "step": 211 }, { "epoch": 0.02, "learning_rate": 4.998951621507526e-05, "loss": 1.7699, "step": 212 }, { "epoch": 0.02, "learning_rate": 4.998930760252526e-05, "loss": 1.7949, "step": 213 }, { "epoch": 0.02, "learning_rate": 4.9989096935268174e-05, "loss": 1.8763, "step": 214 }, { "epoch": 0.02, "learning_rate": 4.998888421332133e-05, "loss": 1.8293, "step": 215 }, { "epoch": 0.02, "learning_rate": 4.9988669436702215e-05, "loss": 1.8183, "step": 216 }, { "epoch": 0.02, "learning_rate": 4.9988452605428494e-05, "loss": 1.8249, "step": 217 }, { "epoch": 0.02, "learning_rate": 4.998823371951798e-05, "loss": 1.8039, "step": 218 }, { "epoch": 0.02, "learning_rate": 4.998801277898868e-05, "loss": 1.7498, "step": 219 }, { "epoch": 0.02, "learning_rate": 4.998778978385877e-05, "loss": 1.7853, "step": 220 }, { "epoch": 0.02, "learning_rate": 4.9987564734146566e-05, "loss": 1.8089, "step": 221 }, { "epoch": 0.02, "learning_rate": 4.998733762987059e-05, "loss": 1.7607, "step": 222 }, { "epoch": 0.02, "learning_rate": 4.9987108471049506e-05, "loss": 1.8264, "step": 223 }, { "epoch": 0.02, "learning_rate": 4.9986877257702155e-05, "loss": 1.7694, "step": 224 }, { "epoch": 0.02, "learning_rate": 4.998664398984756e-05, "loss": 1.766, "step": 225 }, { "epoch": 0.02, "learning_rate": 4.998640866750488e-05, "loss": 1.8086, "step": 226 }, { "epoch": 0.02, "learning_rate": 4.998617129069348e-05, "loss": 1.8136, "step": 227 }, { "epoch": 0.02, "learning_rate": 4.998593185943289e-05, "loss": 1.7507, "step": 228 }, { "epoch": 0.02, "learning_rate": 4.998569037374277e-05, "loss": 1.8594, "step": 229 }, { "epoch": 0.02, "learning_rate": 4.998544683364298e-05, "loss": 1.8483, "step": 230 }, { "epoch": 0.02, "learning_rate": 4.998520123915356e-05, "loss": 1.8194, "step": 231 }, { "epoch": 0.02, "learning_rate": 4.9984953590294694e-05, "loss": 1.7787, "step": 232 }, { "epoch": 0.02, "learning_rate": 4.998470388708675e-05, "loss": 1.7776, "step": 233 }, { "epoch": 0.02, "learning_rate": 4.9984452129550245e-05, "loss": 1.8227, "step": 234 }, { "epoch": 0.02, "learning_rate": 4.99841983177059e-05, "loss": 1.7869, "step": 235 }, { "epoch": 0.02, "learning_rate": 4.9983942451574565e-05, "loss": 1.7424, "step": 236 }, { "epoch": 0.02, "learning_rate": 4.99836845311773e-05, "loss": 1.7866, "step": 237 }, { "epoch": 0.02, "learning_rate": 4.9983424556535285e-05, "loss": 1.8076, "step": 238 }, { "epoch": 0.02, "learning_rate": 4.9983162527669924e-05, "loss": 1.6889, "step": 239 }, { "epoch": 0.02, "learning_rate": 4.9982898444602735e-05, "loss": 1.7016, "step": 240 }, { "epoch": 0.02, "learning_rate": 4.998263230735546e-05, "loss": 1.7898, "step": 241 }, { "epoch": 0.02, "learning_rate": 4.998236411594995e-05, "loss": 1.7682, "step": 242 }, { "epoch": 0.02, "learning_rate": 4.998209387040829e-05, "loss": 1.8042, "step": 243 }, { "epoch": 0.02, "learning_rate": 4.9981821570752674e-05, "loss": 1.8007, "step": 244 }, { "epoch": 0.02, "learning_rate": 4.998154721700551e-05, "loss": 1.755, "step": 245 }, { "epoch": 0.02, "learning_rate": 4.998127080918934e-05, "loss": 1.741, "step": 246 }, { "epoch": 0.02, "learning_rate": 4.9980992347326896e-05, "loss": 1.761, "step": 247 }, { "epoch": 0.02, "learning_rate": 4.998071183144109e-05, "loss": 1.7243, "step": 248 }, { "epoch": 0.02, "learning_rate": 4.9980429261554964e-05, "loss": 1.7676, "step": 249 }, { "epoch": 0.02, "learning_rate": 4.998014463769176e-05, "loss": 1.732, "step": 250 }, { "epoch": 0.02, "learning_rate": 4.997985795987489e-05, "loss": 1.8007, "step": 251 }, { "epoch": 0.02, "learning_rate": 4.9979569228127914e-05, "loss": 1.8557, "step": 252 }, { "epoch": 0.02, "learning_rate": 4.997927844247457e-05, "loss": 1.7485, "step": 253 }, { "epoch": 0.02, "learning_rate": 4.997898560293879e-05, "loss": 1.7388, "step": 254 }, { "epoch": 0.02, "learning_rate": 4.997869070954463e-05, "loss": 1.7989, "step": 255 }, { "epoch": 0.02, "learning_rate": 4.997839376231635e-05, "loss": 1.7279, "step": 256 }, { "epoch": 0.02, "learning_rate": 4.997809476127835e-05, "loss": 1.7216, "step": 257 }, { "epoch": 0.02, "learning_rate": 4.997779370645522e-05, "loss": 1.758, "step": 258 }, { "epoch": 0.02, "learning_rate": 4.997749059787173e-05, "loss": 1.7515, "step": 259 }, { "epoch": 0.02, "learning_rate": 4.9977185435552783e-05, "loss": 1.7357, "step": 260 }, { "epoch": 0.02, "learning_rate": 4.997687821952348e-05, "loss": 1.6863, "step": 261 }, { "epoch": 0.02, "learning_rate": 4.997656894980909e-05, "loss": 1.7276, "step": 262 }, { "epoch": 0.02, "learning_rate": 4.997625762643502e-05, "loss": 1.6692, "step": 263 }, { "epoch": 0.02, "learning_rate": 4.997594424942689e-05, "loss": 1.767, "step": 264 }, { "epoch": 0.02, "learning_rate": 4.9975628818810446e-05, "loss": 1.787, "step": 265 }, { "epoch": 0.02, "learning_rate": 4.9975311334611644e-05, "loss": 1.6857, "step": 266 }, { "epoch": 0.02, "learning_rate": 4.9974991796856576e-05, "loss": 1.783, "step": 267 }, { "epoch": 0.02, "learning_rate": 4.9974670205571514e-05, "loss": 1.7377, "step": 268 }, { "epoch": 0.02, "learning_rate": 4.9974346560782906e-05, "loss": 1.6845, "step": 269 }, { "epoch": 0.02, "learning_rate": 4.997402086251737e-05, "loss": 1.7257, "step": 270 }, { "epoch": 0.02, "learning_rate": 4.997369311080168e-05, "loss": 1.7225, "step": 271 }, { "epoch": 0.02, "learning_rate": 4.997336330566277e-05, "loss": 1.698, "step": 272 }, { "epoch": 0.02, "learning_rate": 4.9973031447127784e-05, "loss": 1.7966, "step": 273 }, { "epoch": 0.02, "learning_rate": 4.997269753522399e-05, "loss": 1.6773, "step": 274 }, { "epoch": 0.02, "learning_rate": 4.9972361569978854e-05, "loss": 1.7234, "step": 275 }, { "epoch": 0.02, "learning_rate": 4.997202355141999e-05, "loss": 1.7032, "step": 276 }, { "epoch": 0.03, "learning_rate": 4.99716834795752e-05, "loss": 1.7302, "step": 277 }, { "epoch": 0.03, "learning_rate": 4.9971341354472446e-05, "loss": 1.7238, "step": 278 }, { "epoch": 0.03, "learning_rate": 4.997099717613984e-05, "loss": 1.7269, "step": 279 }, { "epoch": 0.03, "learning_rate": 4.997065094460571e-05, "loss": 1.7111, "step": 280 }, { "epoch": 0.03, "learning_rate": 4.997030265989852e-05, "loss": 1.737, "step": 281 }, { "epoch": 0.03, "learning_rate": 4.996995232204687e-05, "loss": 1.6962, "step": 282 }, { "epoch": 0.03, "learning_rate": 4.9969599931079616e-05, "loss": 1.7051, "step": 283 }, { "epoch": 0.03, "learning_rate": 4.996924548702571e-05, "loss": 1.6014, "step": 284 }, { "epoch": 0.03, "learning_rate": 4.996888898991429e-05, "loss": 1.6186, "step": 285 }, { "epoch": 0.03, "learning_rate": 4.996853043977467e-05, "loss": 1.68, "step": 286 }, { "epoch": 0.03, "learning_rate": 4.9968169836636345e-05, "loss": 1.6845, "step": 287 }, { "epoch": 0.03, "learning_rate": 4.996780718052896e-05, "loss": 1.6488, "step": 288 }, { "epoch": 0.03, "learning_rate": 4.9967442471482317e-05, "loss": 1.7202, "step": 289 }, { "epoch": 0.03, "learning_rate": 4.996707570952642e-05, "loss": 1.6529, "step": 290 }, { "epoch": 0.03, "learning_rate": 4.9966706894691426e-05, "loss": 1.641, "step": 291 }, { "epoch": 0.03, "learning_rate": 4.996633602700766e-05, "loss": 1.7022, "step": 292 }, { "epoch": 0.03, "learning_rate": 4.996596310650561e-05, "loss": 1.6771, "step": 293 }, { "epoch": 0.03, "learning_rate": 4.9965588133215937e-05, "loss": 1.6726, "step": 294 }, { "epoch": 0.03, "learning_rate": 4.996521110716947e-05, "loss": 1.6967, "step": 295 }, { "epoch": 0.03, "learning_rate": 4.9964832028397234e-05, "loss": 1.6566, "step": 296 }, { "epoch": 0.03, "learning_rate": 4.996445089693037e-05, "loss": 1.7169, "step": 297 }, { "epoch": 0.03, "learning_rate": 4.996406771280023e-05, "loss": 1.7213, "step": 298 }, { "epoch": 0.03, "learning_rate": 4.996368247603831e-05, "loss": 1.6944, "step": 299 }, { "epoch": 0.03, "learning_rate": 4.996329518667629e-05, "loss": 1.6189, "step": 300 }, { "epoch": 0.03, "learning_rate": 4.9962905844746036e-05, "loss": 1.7225, "step": 301 }, { "epoch": 0.03, "learning_rate": 4.996251445027952e-05, "loss": 1.6343, "step": 302 }, { "epoch": 0.03, "learning_rate": 4.9962121003308956e-05, "loss": 1.646, "step": 303 }, { "epoch": 0.03, "learning_rate": 4.9961725503866685e-05, "loss": 1.6446, "step": 304 }, { "epoch": 0.03, "learning_rate": 4.996132795198523e-05, "loss": 1.6095, "step": 305 }, { "epoch": 0.03, "learning_rate": 4.996092834769726e-05, "loss": 1.6479, "step": 306 }, { "epoch": 0.03, "learning_rate": 4.996052669103566e-05, "loss": 1.5708, "step": 307 }, { "epoch": 0.03, "learning_rate": 4.996012298203343e-05, "loss": 1.6637, "step": 308 }, { "epoch": 0.03, "learning_rate": 4.9959717220723784e-05, "loss": 1.6763, "step": 309 }, { "epoch": 0.03, "learning_rate": 4.995930940714007e-05, "loss": 1.6187, "step": 310 }, { "epoch": 0.03, "learning_rate": 4.995889954131584e-05, "loss": 1.5799, "step": 311 }, { "epoch": 0.03, "learning_rate": 4.995848762328477e-05, "loss": 1.6125, "step": 312 }, { "epoch": 0.03, "learning_rate": 4.995807365308075e-05, "loss": 1.575, "step": 313 }, { "epoch": 0.03, "learning_rate": 4.9957657630737806e-05, "loss": 1.6532, "step": 314 }, { "epoch": 0.03, "learning_rate": 4.9957239556290145e-05, "loss": 1.5835, "step": 315 }, { "epoch": 0.03, "learning_rate": 4.995681942977215e-05, "loss": 1.6348, "step": 316 }, { "epoch": 0.03, "learning_rate": 4.995639725121836e-05, "loss": 1.6435, "step": 317 }, { "epoch": 0.03, "learning_rate": 4.9955973020663484e-05, "loss": 1.6393, "step": 318 }, { "epoch": 0.03, "learning_rate": 4.995554673814241e-05, "loss": 1.6145, "step": 319 }, { "epoch": 0.03, "learning_rate": 4.995511840369019e-05, "loss": 1.5813, "step": 320 }, { "epoch": 0.03, "learning_rate": 4.9954688017342036e-05, "loss": 1.5902, "step": 321 }, { "epoch": 0.03, "learning_rate": 4.995425557913335e-05, "loss": 1.5944, "step": 322 }, { "epoch": 0.03, "learning_rate": 4.995382108909967e-05, "loss": 1.6616, "step": 323 }, { "epoch": 0.03, "learning_rate": 4.995338454727673e-05, "loss": 1.6189, "step": 324 }, { "epoch": 0.03, "learning_rate": 4.995294595370043e-05, "loss": 1.6502, "step": 325 }, { "epoch": 0.03, "learning_rate": 4.995250530840681e-05, "loss": 1.6363, "step": 326 }, { "epoch": 0.03, "learning_rate": 4.9952062611432124e-05, "loss": 1.5529, "step": 327 }, { "epoch": 0.03, "learning_rate": 4.995161786281277e-05, "loss": 1.6252, "step": 328 }, { "epoch": 0.03, "learning_rate": 4.9951171062585324e-05, "loss": 1.5716, "step": 329 }, { "epoch": 0.03, "learning_rate": 4.9950722210786496e-05, "loss": 1.6273, "step": 330 }, { "epoch": 0.03, "learning_rate": 4.995027130745321e-05, "loss": 1.566, "step": 331 }, { "epoch": 0.03, "learning_rate": 4.994981835262254e-05, "loss": 1.6149, "step": 332 }, { "epoch": 0.03, "learning_rate": 4.994936334633173e-05, "loss": 1.6082, "step": 333 }, { "epoch": 0.03, "learning_rate": 4.994890628861819e-05, "loss": 1.575, "step": 334 }, { "epoch": 0.03, "learning_rate": 4.994844717951951e-05, "loss": 1.6021, "step": 335 }, { "epoch": 0.03, "learning_rate": 4.9947986019073425e-05, "loss": 1.5903, "step": 336 }, { "epoch": 0.03, "learning_rate": 4.994752280731785e-05, "loss": 1.6297, "step": 337 }, { "epoch": 0.03, "learning_rate": 4.9947057544290894e-05, "loss": 1.5692, "step": 338 }, { "epoch": 0.03, "learning_rate": 4.994659023003079e-05, "loss": 1.5788, "step": 339 }, { "epoch": 0.03, "learning_rate": 4.9946120864575976e-05, "loss": 1.5765, "step": 340 }, { "epoch": 0.03, "learning_rate": 4.994564944796505e-05, "loss": 1.6251, "step": 341 }, { "epoch": 0.03, "learning_rate": 4.994517598023675e-05, "loss": 1.5661, "step": 342 }, { "epoch": 0.03, "learning_rate": 4.994470046143003e-05, "loss": 1.5864, "step": 343 }, { "epoch": 0.03, "learning_rate": 4.994422289158397e-05, "loss": 1.5666, "step": 344 }, { "epoch": 0.03, "learning_rate": 4.994374327073785e-05, "loss": 1.5241, "step": 345 }, { "epoch": 0.03, "learning_rate": 4.994326159893111e-05, "loss": 1.5734, "step": 346 }, { "epoch": 0.03, "learning_rate": 4.994277787620334e-05, "loss": 1.6104, "step": 347 }, { "epoch": 0.03, "learning_rate": 4.994229210259432e-05, "loss": 1.5974, "step": 348 }, { "epoch": 0.03, "learning_rate": 4.9941804278144e-05, "loss": 1.6401, "step": 349 }, { "epoch": 0.03, "learning_rate": 4.9941314402892476e-05, "loss": 1.5379, "step": 350 }, { "epoch": 0.03, "learning_rate": 4.994082247688003e-05, "loss": 1.6041, "step": 351 }, { "epoch": 0.03, "learning_rate": 4.994032850014713e-05, "loss": 1.6047, "step": 352 }, { "epoch": 0.03, "learning_rate": 4.9939832472734367e-05, "loss": 1.5963, "step": 353 }, { "epoch": 0.03, "learning_rate": 4.9939334394682535e-05, "loss": 1.5879, "step": 354 }, { "epoch": 0.03, "learning_rate": 4.9938834266032595e-05, "loss": 1.607, "step": 355 }, { "epoch": 0.03, "learning_rate": 4.993833208682566e-05, "loss": 1.6454, "step": 356 }, { "epoch": 0.03, "learning_rate": 4.993782785710302e-05, "loss": 1.5233, "step": 357 }, { "epoch": 0.03, "learning_rate": 4.993732157690614e-05, "loss": 1.5886, "step": 358 }, { "epoch": 0.03, "learning_rate": 4.993681324627665e-05, "loss": 1.546, "step": 359 }, { "epoch": 0.03, "learning_rate": 4.993630286525634e-05, "loss": 1.6256, "step": 360 }, { "epoch": 0.03, "learning_rate": 4.993579043388718e-05, "loss": 1.5629, "step": 361 }, { "epoch": 0.03, "learning_rate": 4.99352759522113e-05, "loss": 1.6112, "step": 362 }, { "epoch": 0.03, "learning_rate": 4.993475942027101e-05, "loss": 1.6316, "step": 363 }, { "epoch": 0.03, "learning_rate": 4.993424083810877e-05, "loss": 1.5741, "step": 364 }, { "epoch": 0.03, "learning_rate": 4.9933720205767234e-05, "loss": 1.5686, "step": 365 }, { "epoch": 0.03, "learning_rate": 4.9933197523289186e-05, "loss": 1.588, "step": 366 }, { "epoch": 0.03, "learning_rate": 4.993267279071764e-05, "loss": 1.5511, "step": 367 }, { "epoch": 0.03, "learning_rate": 4.9932146008095705e-05, "loss": 1.5886, "step": 368 }, { "epoch": 0.03, "learning_rate": 4.993161717546671e-05, "loss": 1.566, "step": 369 }, { "epoch": 0.03, "learning_rate": 4.993108629287414e-05, "loss": 1.5143, "step": 370 }, { "epoch": 0.03, "learning_rate": 4.993055336036164e-05, "loss": 1.5442, "step": 371 }, { "epoch": 0.03, "learning_rate": 4.993001837797304e-05, "loss": 1.547, "step": 372 }, { "epoch": 0.03, "learning_rate": 4.9929481345752315e-05, "loss": 1.562, "step": 373 }, { "epoch": 0.03, "learning_rate": 4.992894226374363e-05, "loss": 1.5855, "step": 374 }, { "epoch": 0.03, "learning_rate": 4.9928401131991305e-05, "loss": 1.5284, "step": 375 }, { "epoch": 0.03, "learning_rate": 4.9927857950539834e-05, "loss": 1.6014, "step": 376 }, { "epoch": 0.03, "learning_rate": 4.992731271943388e-05, "loss": 1.5999, "step": 377 }, { "epoch": 0.03, "learning_rate": 4.9926765438718284e-05, "loss": 1.574, "step": 378 }, { "epoch": 0.03, "learning_rate": 4.992621610843803e-05, "loss": 1.5124, "step": 379 }, { "epoch": 0.03, "learning_rate": 4.99256647286383e-05, "loss": 1.516, "step": 380 }, { "epoch": 0.03, "learning_rate": 4.992511129936442e-05, "loss": 1.5894, "step": 381 }, { "epoch": 0.03, "learning_rate": 4.9924555820661894e-05, "loss": 1.5783, "step": 382 }, { "epoch": 0.03, "learning_rate": 4.99239982925764e-05, "loss": 1.5695, "step": 383 }, { "epoch": 0.03, "learning_rate": 4.992343871515378e-05, "loss": 1.5835, "step": 384 }, { "epoch": 0.03, "learning_rate": 4.9922877088440045e-05, "loss": 1.563, "step": 385 }, { "epoch": 0.03, "learning_rate": 4.992231341248137e-05, "loss": 1.5442, "step": 386 }, { "epoch": 0.03, "learning_rate": 4.992174768732411e-05, "loss": 1.5735, "step": 387 }, { "epoch": 0.04, "learning_rate": 4.992117991301476e-05, "loss": 1.5648, "step": 388 }, { "epoch": 0.04, "learning_rate": 4.992061008960003e-05, "loss": 1.5834, "step": 389 }, { "epoch": 0.04, "learning_rate": 4.9920038217126766e-05, "loss": 1.5088, "step": 390 }, { "epoch": 0.04, "learning_rate": 4.991946429564198e-05, "loss": 1.5502, "step": 391 }, { "epoch": 0.04, "learning_rate": 4.991888832519287e-05, "loss": 1.5344, "step": 392 }, { "epoch": 0.04, "learning_rate": 4.9918310305826796e-05, "loss": 1.5536, "step": 393 }, { "epoch": 0.04, "learning_rate": 4.991773023759128e-05, "loss": 1.571, "step": 394 }, { "epoch": 0.04, "learning_rate": 4.991714812053401e-05, "loss": 1.5432, "step": 395 }, { "epoch": 0.04, "learning_rate": 4.9916563954702864e-05, "loss": 1.5892, "step": 396 }, { "epoch": 0.04, "learning_rate": 4.9915977740145865e-05, "loss": 1.5808, "step": 397 }, { "epoch": 0.04, "learning_rate": 4.991538947691122e-05, "loss": 1.5576, "step": 398 }, { "epoch": 0.04, "learning_rate": 4.9914799165047286e-05, "loss": 1.5552, "step": 399 }, { "epoch": 0.04, "learning_rate": 4.991420680460262e-05, "loss": 1.6035, "step": 400 }, { "epoch": 0.04, "learning_rate": 4.991361239562591e-05, "loss": 1.4848, "step": 401 }, { "epoch": 0.04, "learning_rate": 4.9913015938166044e-05, "loss": 1.5731, "step": 402 }, { "epoch": 0.04, "learning_rate": 4.991241743227205e-05, "loss": 1.5882, "step": 403 }, { "epoch": 0.04, "learning_rate": 4.991181687799315e-05, "loss": 1.5203, "step": 404 }, { "epoch": 0.04, "learning_rate": 4.991121427537873e-05, "loss": 1.5649, "step": 405 }, { "epoch": 0.04, "learning_rate": 4.991060962447831e-05, "loss": 1.5415, "step": 406 }, { "epoch": 0.04, "learning_rate": 4.991000292534164e-05, "loss": 1.5422, "step": 407 }, { "epoch": 0.04, "learning_rate": 4.990939417801859e-05, "loss": 1.5356, "step": 408 }, { "epoch": 0.04, "learning_rate": 4.9908783382559204e-05, "loss": 1.5297, "step": 409 }, { "epoch": 0.04, "learning_rate": 4.990817053901372e-05, "loss": 1.5476, "step": 410 }, { "epoch": 0.04, "learning_rate": 4.990755564743252e-05, "loss": 1.5361, "step": 411 }, { "epoch": 0.04, "learning_rate": 4.9906938707866165e-05, "loss": 1.5243, "step": 412 }, { "epoch": 0.04, "learning_rate": 4.990631972036538e-05, "loss": 1.578, "step": 413 }, { "epoch": 0.04, "learning_rate": 4.9905698684981064e-05, "loss": 1.6054, "step": 414 }, { "epoch": 0.04, "learning_rate": 4.9905075601764274e-05, "loss": 1.571, "step": 415 }, { "epoch": 0.04, "learning_rate": 4.990445047076624e-05, "loss": 1.5409, "step": 416 }, { "epoch": 0.04, "learning_rate": 4.990382329203838e-05, "loss": 1.5009, "step": 417 }, { "epoch": 0.04, "learning_rate": 4.990319406563224e-05, "loss": 1.541, "step": 418 }, { "epoch": 0.04, "learning_rate": 4.9902562791599576e-05, "loss": 1.5341, "step": 419 }, { "epoch": 0.04, "learning_rate": 4.990192946999228e-05, "loss": 1.5157, "step": 420 }, { "epoch": 0.04, "learning_rate": 4.990129410086244e-05, "loss": 1.5914, "step": 421 }, { "epoch": 0.04, "learning_rate": 4.9900656684262286e-05, "loss": 1.5894, "step": 422 }, { "epoch": 0.04, "learning_rate": 4.990001722024423e-05, "loss": 1.5689, "step": 423 }, { "epoch": 0.04, "learning_rate": 4.989937570886086e-05, "loss": 1.502, "step": 424 }, { "epoch": 0.04, "learning_rate": 4.9898732150164906e-05, "loss": 1.5487, "step": 425 }, { "epoch": 0.04, "learning_rate": 4.989808654420931e-05, "loss": 1.5533, "step": 426 }, { "epoch": 0.04, "learning_rate": 4.989743889104712e-05, "loss": 1.5596, "step": 427 }, { "epoch": 0.04, "learning_rate": 4.989678919073162e-05, "loss": 1.5211, "step": 428 }, { "epoch": 0.04, "learning_rate": 4.989613744331622e-05, "loss": 1.5134, "step": 429 }, { "epoch": 0.04, "learning_rate": 4.989548364885451e-05, "loss": 1.4702, "step": 430 }, { "epoch": 0.04, "learning_rate": 4.989482780740024e-05, "loss": 1.5658, "step": 431 }, { "epoch": 0.04, "learning_rate": 4.989416991900734e-05, "loss": 1.5316, "step": 432 }, { "epoch": 0.04, "learning_rate": 4.989350998372991e-05, "loss": 1.5728, "step": 433 }, { "epoch": 0.04, "learning_rate": 4.9892848001622206e-05, "loss": 1.5559, "step": 434 }, { "epoch": 0.04, "learning_rate": 4.989218397273866e-05, "loss": 1.5953, "step": 435 }, { "epoch": 0.04, "learning_rate": 4.989151789713386e-05, "loss": 1.4981, "step": 436 }, { "epoch": 0.04, "learning_rate": 4.989084977486259e-05, "loss": 1.5358, "step": 437 }, { "epoch": 0.04, "learning_rate": 4.989017960597978e-05, "loss": 1.5213, "step": 438 }, { "epoch": 0.04, "learning_rate": 4.9889507390540535e-05, "loss": 1.5496, "step": 439 }, { "epoch": 0.04, "learning_rate": 4.988883312860012e-05, "loss": 1.5334, "step": 440 }, { "epoch": 0.04, "learning_rate": 4.9888156820213974e-05, "loss": 1.5816, "step": 441 }, { "epoch": 0.04, "learning_rate": 4.988747846543772e-05, "loss": 1.5202, "step": 442 }, { "epoch": 0.04, "learning_rate": 4.988679806432712e-05, "loss": 1.542, "step": 443 }, { "epoch": 0.04, "learning_rate": 4.9886115616938125e-05, "loss": 1.5274, "step": 444 }, { "epoch": 0.04, "learning_rate": 4.9885431123326845e-05, "loss": 1.515, "step": 445 }, { "epoch": 0.04, "learning_rate": 4.988474458354957e-05, "loss": 1.5974, "step": 446 }, { "epoch": 0.04, "learning_rate": 4.988405599766273e-05, "loss": 1.5294, "step": 447 }, { "epoch": 0.04, "learning_rate": 4.988336536572297e-05, "loss": 1.5442, "step": 448 }, { "epoch": 0.04, "learning_rate": 4.988267268778705e-05, "loss": 1.4943, "step": 449 }, { "epoch": 0.04, "learning_rate": 4.9881977963911945e-05, "loss": 1.4611, "step": 450 }, { "epoch": 0.04, "learning_rate": 4.988128119415477e-05, "loss": 1.5072, "step": 451 }, { "epoch": 0.04, "learning_rate": 4.988058237857281e-05, "loss": 1.5021, "step": 452 }, { "epoch": 0.04, "learning_rate": 4.987988151722354e-05, "loss": 1.4846, "step": 453 }, { "epoch": 0.04, "learning_rate": 4.987917861016456e-05, "loss": 1.5497, "step": 454 }, { "epoch": 0.04, "learning_rate": 4.987847365745369e-05, "loss": 1.5251, "step": 455 }, { "epoch": 0.04, "learning_rate": 4.987776665914889e-05, "loss": 1.5167, "step": 456 }, { "epoch": 0.04, "learning_rate": 4.987705761530828e-05, "loss": 1.4876, "step": 457 }, { "epoch": 0.04, "learning_rate": 4.987634652599017e-05, "loss": 1.5494, "step": 458 }, { "epoch": 0.04, "learning_rate": 4.987563339125303e-05, "loss": 1.5321, "step": 459 }, { "epoch": 0.04, "learning_rate": 4.987491821115548e-05, "loss": 1.4818, "step": 460 }, { "epoch": 0.04, "learning_rate": 4.987420098575635e-05, "loss": 1.5098, "step": 461 }, { "epoch": 0.04, "learning_rate": 4.98734817151146e-05, "loss": 1.4575, "step": 462 }, { "epoch": 0.04, "learning_rate": 4.987276039928936e-05, "loss": 1.5401, "step": 463 }, { "epoch": 0.04, "learning_rate": 4.987203703833995e-05, "loss": 1.5122, "step": 464 }, { "epoch": 0.04, "learning_rate": 4.987131163232585e-05, "loss": 1.5471, "step": 465 }, { "epoch": 0.04, "learning_rate": 4.9870584181306704e-05, "loss": 1.5116, "step": 466 }, { "epoch": 0.04, "learning_rate": 4.986985468534232e-05, "loss": 1.5096, "step": 467 }, { "epoch": 0.04, "learning_rate": 4.9869123144492684e-05, "loss": 1.4722, "step": 468 }, { "epoch": 0.04, "learning_rate": 4.986838955881795e-05, "loss": 1.5212, "step": 469 }, { "epoch": 0.04, "learning_rate": 4.986765392837842e-05, "loss": 1.5136, "step": 470 }, { "epoch": 0.04, "learning_rate": 4.98669162532346e-05, "loss": 1.5975, "step": 471 }, { "epoch": 0.04, "learning_rate": 4.986617653344713e-05, "loss": 1.5015, "step": 472 }, { "epoch": 0.04, "learning_rate": 4.9865434769076844e-05, "loss": 1.4936, "step": 473 }, { "epoch": 0.04, "learning_rate": 4.986469096018472e-05, "loss": 1.4793, "step": 474 }, { "epoch": 0.04, "learning_rate": 4.986394510683192e-05, "loss": 1.4799, "step": 475 }, { "epoch": 0.04, "learning_rate": 4.986319720907978e-05, "loss": 1.4832, "step": 476 }, { "epoch": 0.04, "learning_rate": 4.986244726698978e-05, "loss": 1.539, "step": 477 }, { "epoch": 0.04, "learning_rate": 4.98616952806236e-05, "loss": 1.5441, "step": 478 }, { "epoch": 0.04, "learning_rate": 4.986094125004305e-05, "loss": 1.5235, "step": 479 }, { "epoch": 0.04, "learning_rate": 4.9860185175310145e-05, "loss": 1.5083, "step": 480 }, { "epoch": 0.04, "learning_rate": 4.9859427056487055e-05, "loss": 1.4869, "step": 481 }, { "epoch": 0.04, "learning_rate": 4.98586668936361e-05, "loss": 1.5148, "step": 482 }, { "epoch": 0.04, "learning_rate": 4.9857904686819796e-05, "loss": 1.5284, "step": 483 }, { "epoch": 0.04, "learning_rate": 4.9857140436100805e-05, "loss": 1.5197, "step": 484 }, { "epoch": 0.04, "learning_rate": 4.985637414154197e-05, "loss": 1.5761, "step": 485 }, { "epoch": 0.04, "learning_rate": 4.985560580320631e-05, "loss": 1.4752, "step": 486 }, { "epoch": 0.04, "learning_rate": 4.985483542115698e-05, "loss": 1.5378, "step": 487 }, { "epoch": 0.04, "learning_rate": 4.985406299545733e-05, "loss": 1.4624, "step": 488 }, { "epoch": 0.04, "learning_rate": 4.985328852617087e-05, "loss": 1.5065, "step": 489 }, { "epoch": 0.04, "learning_rate": 4.98525120133613e-05, "loss": 1.5529, "step": 490 }, { "epoch": 0.04, "learning_rate": 4.985173345709244e-05, "loss": 1.5615, "step": 491 }, { "epoch": 0.04, "learning_rate": 4.9850952857428315e-05, "loss": 1.5059, "step": 492 }, { "epoch": 0.04, "learning_rate": 4.9850170214433117e-05, "loss": 1.5754, "step": 493 }, { "epoch": 0.04, "learning_rate": 4.9849385528171185e-05, "loss": 1.5317, "step": 494 }, { "epoch": 0.04, "learning_rate": 4.984859879870705e-05, "loss": 1.5717, "step": 495 }, { "epoch": 0.04, "learning_rate": 4.9847810026105394e-05, "loss": 1.5178, "step": 496 }, { "epoch": 0.04, "learning_rate": 4.9847019210431076e-05, "loss": 1.515, "step": 497 }, { "epoch": 0.04, "learning_rate": 4.9846226351749115e-05, "loss": 1.4848, "step": 498 }, { "epoch": 0.05, "learning_rate": 4.98454314501247e-05, "loss": 1.5227, "step": 499 }, { "epoch": 0.05, "learning_rate": 4.98446345056232e-05, "loss": 1.5197, "step": 500 }, { "epoch": 0.05, "learning_rate": 4.9843835518310134e-05, "loss": 1.5271, "step": 501 }, { "epoch": 0.05, "learning_rate": 4.9843034488251194e-05, "loss": 1.4997, "step": 502 }, { "epoch": 0.05, "learning_rate": 4.984223141551226e-05, "loss": 1.5008, "step": 503 }, { "epoch": 0.05, "learning_rate": 4.984142630015936e-05, "loss": 1.5229, "step": 504 }, { "epoch": 0.05, "learning_rate": 4.984061914225867e-05, "loss": 1.4982, "step": 505 }, { "epoch": 0.05, "learning_rate": 4.983980994187658e-05, "loss": 1.5451, "step": 506 }, { "epoch": 0.05, "learning_rate": 4.9838998699079625e-05, "loss": 1.478, "step": 507 }, { "epoch": 0.05, "learning_rate": 4.9838185413934504e-05, "loss": 1.4293, "step": 508 }, { "epoch": 0.05, "learning_rate": 4.983737008650808e-05, "loss": 1.4777, "step": 509 }, { "epoch": 0.05, "learning_rate": 4.98365527168674e-05, "loss": 1.5143, "step": 510 }, { "epoch": 0.05, "learning_rate": 4.983573330507967e-05, "loss": 1.4995, "step": 511 }, { "epoch": 0.05, "learning_rate": 4.9834911851212265e-05, "loss": 1.5279, "step": 512 }, { "epoch": 0.05, "learning_rate": 4.983408835533273e-05, "loss": 1.4842, "step": 513 }, { "epoch": 0.05, "learning_rate": 4.983326281750877e-05, "loss": 1.5031, "step": 514 }, { "epoch": 0.05, "learning_rate": 4.983243523780827e-05, "loss": 1.4495, "step": 515 }, { "epoch": 0.05, "learning_rate": 4.983160561629927e-05, "loss": 1.4315, "step": 516 }, { "epoch": 0.05, "learning_rate": 4.983077395304999e-05, "loss": 1.4992, "step": 517 }, { "epoch": 0.05, "learning_rate": 4.9829940248128806e-05, "loss": 1.4532, "step": 518 }, { "epoch": 0.05, "learning_rate": 4.9829104501604276e-05, "loss": 1.4905, "step": 519 }, { "epoch": 0.05, "learning_rate": 4.9828266713545114e-05, "loss": 1.5127, "step": 520 }, { "epoch": 0.05, "learning_rate": 4.98274268840202e-05, "loss": 1.4967, "step": 521 }, { "epoch": 0.05, "learning_rate": 4.9826585013098606e-05, "loss": 1.4932, "step": 522 }, { "epoch": 0.05, "learning_rate": 4.982574110084953e-05, "loss": 1.5019, "step": 523 }, { "epoch": 0.05, "learning_rate": 4.982489514734237e-05, "loss": 1.509, "step": 524 }, { "epoch": 0.05, "learning_rate": 4.982404715264669e-05, "loss": 1.5281, "step": 525 }, { "epoch": 0.05, "learning_rate": 4.982319711683221e-05, "loss": 1.4827, "step": 526 }, { "epoch": 0.05, "learning_rate": 4.9822345039968824e-05, "loss": 1.5482, "step": 527 }, { "epoch": 0.05, "learning_rate": 4.982149092212659e-05, "loss": 1.4783, "step": 528 }, { "epoch": 0.05, "learning_rate": 4.982063476337574e-05, "loss": 1.4727, "step": 529 }, { "epoch": 0.05, "learning_rate": 4.981977656378667e-05, "loss": 1.4993, "step": 530 }, { "epoch": 0.05, "learning_rate": 4.9818916323429945e-05, "loss": 1.4682, "step": 531 }, { "epoch": 0.05, "learning_rate": 4.981805404237628e-05, "loss": 1.4784, "step": 532 }, { "epoch": 0.05, "learning_rate": 4.981718972069661e-05, "loss": 1.5266, "step": 533 }, { "epoch": 0.05, "learning_rate": 4.981632335846197e-05, "loss": 1.447, "step": 534 }, { "epoch": 0.05, "learning_rate": 4.9815454955743614e-05, "loss": 1.528, "step": 535 }, { "epoch": 0.05, "learning_rate": 4.981458451261294e-05, "loss": 1.4613, "step": 536 }, { "epoch": 0.05, "learning_rate": 4.9813712029141504e-05, "loss": 1.4976, "step": 537 }, { "epoch": 0.05, "learning_rate": 4.9812837505401066e-05, "loss": 1.4969, "step": 538 }, { "epoch": 0.05, "learning_rate": 4.981196094146353e-05, "loss": 1.46, "step": 539 }, { "epoch": 0.05, "learning_rate": 4.9811082337400964e-05, "loss": 1.4773, "step": 540 }, { "epoch": 0.05, "learning_rate": 4.98102016932856e-05, "loss": 1.4999, "step": 541 }, { "epoch": 0.05, "learning_rate": 4.980931900918987e-05, "loss": 1.4797, "step": 542 }, { "epoch": 0.05, "learning_rate": 4.9808434285186326e-05, "loss": 1.5109, "step": 543 }, { "epoch": 0.05, "learning_rate": 4.980754752134774e-05, "loss": 1.4689, "step": 544 }, { "epoch": 0.05, "learning_rate": 4.9806658717747e-05, "loss": 1.5, "step": 545 }, { "epoch": 0.05, "learning_rate": 4.9805767874457204e-05, "loss": 1.4852, "step": 546 }, { "epoch": 0.05, "learning_rate": 4.980487499155159e-05, "loss": 1.5403, "step": 547 }, { "epoch": 0.05, "learning_rate": 4.980398006910358e-05, "loss": 1.4585, "step": 548 }, { "epoch": 0.05, "learning_rate": 4.980308310718675e-05, "loss": 1.4986, "step": 549 }, { "epoch": 0.05, "learning_rate": 4.980218410587487e-05, "loss": 1.5136, "step": 550 }, { "epoch": 0.05, "learning_rate": 4.980128306524183e-05, "loss": 1.4687, "step": 551 }, { "epoch": 0.05, "learning_rate": 4.980037998536174e-05, "loss": 1.4985, "step": 552 }, { "epoch": 0.05, "learning_rate": 4.979947486630885e-05, "loss": 1.501, "step": 553 }, { "epoch": 0.05, "learning_rate": 4.979856770815757e-05, "loss": 1.5031, "step": 554 }, { "epoch": 0.05, "learning_rate": 4.97976585109825e-05, "loss": 1.5229, "step": 555 }, { "epoch": 0.05, "learning_rate": 4.97967472748584e-05, "loss": 1.5083, "step": 556 }, { "epoch": 0.05, "learning_rate": 4.97958339998602e-05, "loss": 1.4408, "step": 557 }, { "epoch": 0.05, "learning_rate": 4.9794918686062964e-05, "loss": 1.4739, "step": 558 }, { "epoch": 0.05, "learning_rate": 4.9794001333541985e-05, "loss": 1.4846, "step": 559 }, { "epoch": 0.05, "learning_rate": 4.979308194237266e-05, "loss": 1.4938, "step": 560 }, { "epoch": 0.05, "learning_rate": 4.9792160512630626e-05, "loss": 1.445, "step": 561 }, { "epoch": 0.05, "learning_rate": 4.979123704439161e-05, "loss": 1.4636, "step": 562 }, { "epoch": 0.05, "learning_rate": 4.979031153773155e-05, "loss": 1.4949, "step": 563 }, { "epoch": 0.05, "learning_rate": 4.978938399272656e-05, "loss": 1.4644, "step": 564 }, { "epoch": 0.05, "learning_rate": 4.978845440945289e-05, "loss": 1.4561, "step": 565 }, { "epoch": 0.05, "learning_rate": 4.9787522787986976e-05, "loss": 1.4996, "step": 566 }, { "epoch": 0.05, "learning_rate": 4.978658912840542e-05, "loss": 1.4481, "step": 567 }, { "epoch": 0.05, "learning_rate": 4.9785653430785e-05, "loss": 1.491, "step": 568 }, { "epoch": 0.05, "learning_rate": 4.9784715695202636e-05, "loss": 1.5131, "step": 569 }, { "epoch": 0.05, "learning_rate": 4.978377592173544e-05, "loss": 1.5264, "step": 570 }, { "epoch": 0.05, "learning_rate": 4.97828341104607e-05, "loss": 1.4387, "step": 571 }, { "epoch": 0.05, "learning_rate": 4.978189026145582e-05, "loss": 1.5142, "step": 572 }, { "epoch": 0.05, "learning_rate": 4.9780944374798435e-05, "loss": 1.4788, "step": 573 }, { "epoch": 0.05, "learning_rate": 4.977999645056631e-05, "loss": 1.461, "step": 574 }, { "epoch": 0.05, "learning_rate": 4.977904648883738e-05, "loss": 1.4562, "step": 575 }, { "epoch": 0.05, "learning_rate": 4.977809448968977e-05, "loss": 1.5003, "step": 576 }, { "epoch": 0.05, "learning_rate": 4.9777140453201744e-05, "loss": 1.4673, "step": 577 }, { "epoch": 0.05, "learning_rate": 4.977618437945174e-05, "loss": 1.5131, "step": 578 }, { "epoch": 0.05, "learning_rate": 4.977522626851839e-05, "loss": 1.4649, "step": 579 }, { "epoch": 0.05, "learning_rate": 4.977426612048046e-05, "loss": 1.5017, "step": 580 }, { "epoch": 0.05, "learning_rate": 4.977330393541689e-05, "loss": 1.51, "step": 581 }, { "epoch": 0.05, "learning_rate": 4.977233971340681e-05, "loss": 1.4912, "step": 582 }, { "epoch": 0.05, "learning_rate": 4.977137345452949e-05, "loss": 1.4899, "step": 583 }, { "epoch": 0.05, "learning_rate": 4.977040515886439e-05, "loss": 1.4628, "step": 584 }, { "epoch": 0.05, "learning_rate": 4.9769434826491115e-05, "loss": 1.4246, "step": 585 }, { "epoch": 0.05, "learning_rate": 4.976846245748946e-05, "loss": 1.4391, "step": 586 }, { "epoch": 0.05, "learning_rate": 4.976748805193937e-05, "loss": 1.5025, "step": 587 }, { "epoch": 0.05, "learning_rate": 4.976651160992096e-05, "loss": 1.5138, "step": 588 }, { "epoch": 0.05, "learning_rate": 4.976553313151452e-05, "loss": 1.4342, "step": 589 }, { "epoch": 0.05, "learning_rate": 4.976455261680051e-05, "loss": 1.4329, "step": 590 }, { "epoch": 0.05, "learning_rate": 4.9763570065859546e-05, "loss": 1.4932, "step": 591 }, { "epoch": 0.05, "learning_rate": 4.976258547877242e-05, "loss": 1.4651, "step": 592 }, { "epoch": 0.05, "learning_rate": 4.976159885562008e-05, "loss": 1.5083, "step": 593 }, { "epoch": 0.05, "learning_rate": 4.976061019648366e-05, "loss": 1.4799, "step": 594 }, { "epoch": 0.05, "learning_rate": 4.975961950144444e-05, "loss": 1.4839, "step": 595 }, { "epoch": 0.05, "learning_rate": 4.9758626770583886e-05, "loss": 1.4777, "step": 596 }, { "epoch": 0.05, "learning_rate": 4.975763200398362e-05, "loss": 1.4743, "step": 597 }, { "epoch": 0.05, "learning_rate": 4.975663520172544e-05, "loss": 1.4865, "step": 598 }, { "epoch": 0.05, "learning_rate": 4.975563636389131e-05, "loss": 1.4628, "step": 599 }, { "epoch": 0.05, "learning_rate": 4.975463549056334e-05, "loss": 1.515, "step": 600 }, { "epoch": 0.05, "learning_rate": 4.9753632581823847e-05, "loss": 1.4083, "step": 601 }, { "epoch": 0.05, "learning_rate": 4.975262763775528e-05, "loss": 1.5497, "step": 602 }, { "epoch": 0.05, "learning_rate": 4.975162065844027e-05, "loss": 1.4598, "step": 603 }, { "epoch": 0.05, "learning_rate": 4.975061164396162e-05, "loss": 1.4143, "step": 604 }, { "epoch": 0.05, "learning_rate": 4.9749600594402276e-05, "loss": 1.44, "step": 605 }, { "epoch": 0.05, "learning_rate": 4.97485875098454e-05, "loss": 1.5054, "step": 606 }, { "epoch": 0.05, "learning_rate": 4.974757239037428e-05, "loss": 1.4678, "step": 607 }, { "epoch": 0.05, "learning_rate": 4.974655523607237e-05, "loss": 1.4993, "step": 608 }, { "epoch": 0.06, "learning_rate": 4.9745536047023324e-05, "loss": 1.467, "step": 609 }, { "epoch": 0.06, "learning_rate": 4.9744514823310925e-05, "loss": 1.4492, "step": 610 }, { "epoch": 0.06, "learning_rate": 4.9743491565019146e-05, "loss": 1.4556, "step": 611 }, { "epoch": 0.06, "learning_rate": 4.974246627223213e-05, "loss": 1.4873, "step": 612 }, { "epoch": 0.06, "learning_rate": 4.974143894503417e-05, "loss": 1.5242, "step": 613 }, { "epoch": 0.06, "learning_rate": 4.974040958350975e-05, "loss": 1.4793, "step": 614 }, { "epoch": 0.06, "learning_rate": 4.9739378187743504e-05, "loss": 1.4882, "step": 615 }, { "epoch": 0.06, "learning_rate": 4.9738344757820223e-05, "loss": 1.4556, "step": 616 }, { "epoch": 0.06, "learning_rate": 4.973730929382489e-05, "loss": 1.4385, "step": 617 }, { "epoch": 0.06, "learning_rate": 4.973627179584265e-05, "loss": 1.437, "step": 618 }, { "epoch": 0.06, "learning_rate": 4.9735232263958796e-05, "loss": 1.4301, "step": 619 }, { "epoch": 0.06, "learning_rate": 4.973419069825881e-05, "loss": 1.4616, "step": 620 }, { "epoch": 0.06, "learning_rate": 4.973314709882834e-05, "loss": 1.4826, "step": 621 }, { "epoch": 0.06, "learning_rate": 4.9732101465753176e-05, "loss": 1.3983, "step": 622 }, { "epoch": 0.06, "learning_rate": 4.973105379911931e-05, "loss": 1.4607, "step": 623 }, { "epoch": 0.06, "learning_rate": 4.973000409901288e-05, "loss": 1.4757, "step": 624 }, { "epoch": 0.06, "learning_rate": 4.972895236552019e-05, "loss": 1.4998, "step": 625 }, { "epoch": 0.06, "learning_rate": 4.972789859872772e-05, "loss": 1.4553, "step": 626 }, { "epoch": 0.06, "learning_rate": 4.9726842798722126e-05, "loss": 1.4687, "step": 627 }, { "epoch": 0.06, "learning_rate": 4.9725784965590205e-05, "loss": 1.4847, "step": 628 }, { "epoch": 0.06, "learning_rate": 4.972472509941894e-05, "loss": 1.4939, "step": 629 }, { "epoch": 0.06, "learning_rate": 4.972366320029548e-05, "loss": 1.4417, "step": 630 }, { "epoch": 0.06, "learning_rate": 4.9722599268307127e-05, "loss": 1.4813, "step": 631 }, { "epoch": 0.06, "learning_rate": 4.972153330354138e-05, "loss": 1.4813, "step": 632 }, { "epoch": 0.06, "learning_rate": 4.972046530608586e-05, "loss": 1.4243, "step": 633 }, { "epoch": 0.06, "learning_rate": 4.971939527602841e-05, "loss": 1.4854, "step": 634 }, { "epoch": 0.06, "learning_rate": 4.9718323213456996e-05, "loss": 1.4135, "step": 635 }, { "epoch": 0.06, "learning_rate": 4.971724911845977e-05, "loss": 1.5219, "step": 636 }, { "epoch": 0.06, "learning_rate": 4.9716172991125056e-05, "loss": 1.4915, "step": 637 }, { "epoch": 0.06, "learning_rate": 4.971509483154132e-05, "loss": 1.4659, "step": 638 }, { "epoch": 0.06, "learning_rate": 4.971401463979721e-05, "loss": 1.4857, "step": 639 }, { "epoch": 0.06, "learning_rate": 4.971293241598157e-05, "loss": 1.4946, "step": 640 }, { "epoch": 0.06, "learning_rate": 4.971184816018336e-05, "loss": 1.4887, "step": 641 }, { "epoch": 0.06, "learning_rate": 4.971076187249174e-05, "loss": 1.4593, "step": 642 }, { "epoch": 0.06, "learning_rate": 4.9709673552996026e-05, "loss": 1.3976, "step": 643 }, { "epoch": 0.06, "learning_rate": 4.9708583201785705e-05, "loss": 1.4629, "step": 644 }, { "epoch": 0.06, "learning_rate": 4.970749081895043e-05, "loss": 1.4534, "step": 645 }, { "epoch": 0.06, "learning_rate": 4.970639640458001e-05, "loss": 1.444, "step": 646 }, { "epoch": 0.06, "learning_rate": 4.970529995876446e-05, "loss": 1.4489, "step": 647 }, { "epoch": 0.06, "learning_rate": 4.97042014815939e-05, "loss": 1.4803, "step": 648 }, { "epoch": 0.06, "learning_rate": 4.970310097315867e-05, "loss": 1.4686, "step": 649 }, { "epoch": 0.06, "learning_rate": 4.970199843354925e-05, "loss": 1.4169, "step": 650 }, { "epoch": 0.06, "learning_rate": 4.970089386285629e-05, "loss": 1.491, "step": 651 }, { "epoch": 0.06, "learning_rate": 4.9699787261170635e-05, "loss": 1.4778, "step": 652 }, { "epoch": 0.06, "learning_rate": 4.969867862858325e-05, "loss": 1.414, "step": 653 }, { "epoch": 0.06, "learning_rate": 4.96975679651853e-05, "loss": 1.4099, "step": 654 }, { "epoch": 0.06, "learning_rate": 4.96964552710681e-05, "loss": 1.4748, "step": 655 }, { "epoch": 0.06, "learning_rate": 4.969534054632315e-05, "loss": 1.4658, "step": 656 }, { "epoch": 0.06, "learning_rate": 4.9694223791042096e-05, "loss": 1.4701, "step": 657 }, { "epoch": 0.06, "learning_rate": 4.969310500531677e-05, "loss": 1.5013, "step": 658 }, { "epoch": 0.06, "learning_rate": 4.969198418923916e-05, "loss": 1.4038, "step": 659 }, { "epoch": 0.06, "learning_rate": 4.969086134290143e-05, "loss": 1.502, "step": 660 }, { "epoch": 0.06, "learning_rate": 4.968973646639589e-05, "loss": 1.4137, "step": 661 }, { "epoch": 0.06, "learning_rate": 4.9688609559815034e-05, "loss": 1.5279, "step": 662 }, { "epoch": 0.06, "learning_rate": 4.968748062325153e-05, "loss": 1.4038, "step": 663 }, { "epoch": 0.06, "learning_rate": 4.96863496567982e-05, "loss": 1.5109, "step": 664 }, { "epoch": 0.06, "learning_rate": 4.9685216660548026e-05, "loss": 1.4594, "step": 665 }, { "epoch": 0.06, "learning_rate": 4.968408163459418e-05, "loss": 1.464, "step": 666 }, { "epoch": 0.06, "learning_rate": 4.968294457902999e-05, "loss": 1.416, "step": 667 }, { "epoch": 0.06, "learning_rate": 4.9681805493948926e-05, "loss": 1.4518, "step": 668 }, { "epoch": 0.06, "learning_rate": 4.9680664379444674e-05, "loss": 1.4988, "step": 669 }, { "epoch": 0.06, "learning_rate": 4.967952123561104e-05, "loss": 1.4429, "step": 670 }, { "epoch": 0.06, "learning_rate": 4.967837606254203e-05, "loss": 1.4664, "step": 671 }, { "epoch": 0.06, "learning_rate": 4.96772288603318e-05, "loss": 1.4594, "step": 672 }, { "epoch": 0.06, "learning_rate": 4.967607962907468e-05, "loss": 1.4264, "step": 673 }, { "epoch": 0.06, "learning_rate": 4.967492836886516e-05, "loss": 1.4598, "step": 674 }, { "epoch": 0.06, "learning_rate": 4.9673775079797893e-05, "loss": 1.4671, "step": 675 }, { "epoch": 0.06, "learning_rate": 4.967261976196772e-05, "loss": 1.3984, "step": 676 }, { "epoch": 0.06, "learning_rate": 4.967146241546963e-05, "loss": 1.4379, "step": 677 }, { "epoch": 0.06, "learning_rate": 4.967030304039879e-05, "loss": 1.4494, "step": 678 }, { "epoch": 0.06, "learning_rate": 4.966914163685051e-05, "loss": 1.4333, "step": 679 }, { "epoch": 0.06, "learning_rate": 4.966797820492031e-05, "loss": 1.4287, "step": 680 }, { "epoch": 0.06, "learning_rate": 4.966681274470383e-05, "loss": 1.482, "step": 681 }, { "epoch": 0.06, "learning_rate": 4.96656452562969e-05, "loss": 1.3918, "step": 682 }, { "epoch": 0.06, "learning_rate": 4.966447573979553e-05, "loss": 1.4656, "step": 683 }, { "epoch": 0.06, "learning_rate": 4.966330419529587e-05, "loss": 1.4822, "step": 684 }, { "epoch": 0.06, "learning_rate": 4.9662130622894255e-05, "loss": 1.4325, "step": 685 }, { "epoch": 0.06, "learning_rate": 4.966095502268717e-05, "loss": 1.4355, "step": 686 }, { "epoch": 0.06, "learning_rate": 4.9659777394771285e-05, "loss": 1.4388, "step": 687 }, { "epoch": 0.06, "learning_rate": 4.965859773924343e-05, "loss": 1.4295, "step": 688 }, { "epoch": 0.06, "learning_rate": 4.965741605620059e-05, "loss": 1.4798, "step": 689 }, { "epoch": 0.06, "learning_rate": 4.965623234573994e-05, "loss": 1.5025, "step": 690 }, { "epoch": 0.06, "learning_rate": 4.9655046607958796e-05, "loss": 1.4363, "step": 691 }, { "epoch": 0.06, "learning_rate": 4.965385884295467e-05, "loss": 1.3839, "step": 692 }, { "epoch": 0.06, "learning_rate": 4.9652669050825204e-05, "loss": 1.4865, "step": 693 }, { "epoch": 0.06, "learning_rate": 4.965147723166824e-05, "loss": 1.4727, "step": 694 }, { "epoch": 0.06, "learning_rate": 4.965028338558177e-05, "loss": 1.4714, "step": 695 }, { "epoch": 0.06, "learning_rate": 4.964908751266397e-05, "loss": 1.4837, "step": 696 }, { "epoch": 0.06, "learning_rate": 4.9647889613013146e-05, "loss": 1.4777, "step": 697 }, { "epoch": 0.06, "learning_rate": 4.96466896867278e-05, "loss": 1.4344, "step": 698 }, { "epoch": 0.06, "learning_rate": 4.9645487733906603e-05, "loss": 1.4867, "step": 699 }, { "epoch": 0.06, "learning_rate": 4.9644283754648373e-05, "loss": 1.4221, "step": 700 }, { "epoch": 0.06, "learning_rate": 4.9643077749052116e-05, "loss": 1.463, "step": 701 }, { "epoch": 0.06, "learning_rate": 4.964186971721698e-05, "loss": 1.46, "step": 702 }, { "epoch": 0.06, "learning_rate": 4.96406596592423e-05, "loss": 1.453, "step": 703 }, { "epoch": 0.06, "learning_rate": 4.963944757522759e-05, "loss": 1.4924, "step": 704 }, { "epoch": 0.06, "learning_rate": 4.963823346527248e-05, "loss": 1.4679, "step": 705 }, { "epoch": 0.06, "learning_rate": 4.963701732947682e-05, "loss": 1.473, "step": 706 }, { "epoch": 0.06, "learning_rate": 4.9635799167940586e-05, "loss": 1.4721, "step": 707 }, { "epoch": 0.06, "learning_rate": 4.9634578980763966e-05, "loss": 1.4621, "step": 708 }, { "epoch": 0.06, "learning_rate": 4.9633356768047265e-05, "loss": 1.444, "step": 709 }, { "epoch": 0.06, "learning_rate": 4.963213252989099e-05, "loss": 1.4616, "step": 710 }, { "epoch": 0.06, "learning_rate": 4.9630906266395795e-05, "loss": 1.5161, "step": 711 }, { "epoch": 0.06, "learning_rate": 4.962967797766251e-05, "loss": 1.4612, "step": 712 }, { "epoch": 0.06, "learning_rate": 4.962844766379213e-05, "loss": 1.3994, "step": 713 }, { "epoch": 0.06, "learning_rate": 4.9627215324885815e-05, "loss": 1.4506, "step": 714 }, { "epoch": 0.06, "learning_rate": 4.962598096104489e-05, "loss": 1.4754, "step": 715 }, { "epoch": 0.06, "learning_rate": 4.9624744572370865e-05, "loss": 1.4573, "step": 716 }, { "epoch": 0.06, "learning_rate": 4.962350615896537e-05, "loss": 1.4141, "step": 717 }, { "epoch": 0.06, "learning_rate": 4.962226572093025e-05, "loss": 1.4156, "step": 718 }, { "epoch": 0.06, "learning_rate": 4.9621023258367506e-05, "loss": 1.3867, "step": 719 }, { "epoch": 0.07, "learning_rate": 4.961977877137928e-05, "loss": 1.4706, "step": 720 }, { "epoch": 0.07, "learning_rate": 4.961853226006791e-05, "loss": 1.4402, "step": 721 }, { "epoch": 0.07, "learning_rate": 4.961728372453587e-05, "loss": 1.438, "step": 722 }, { "epoch": 0.07, "learning_rate": 4.9616033164885843e-05, "loss": 1.4355, "step": 723 }, { "epoch": 0.07, "learning_rate": 4.961478058122065e-05, "loss": 1.4815, "step": 724 }, { "epoch": 0.07, "learning_rate": 4.961352597364327e-05, "loss": 1.461, "step": 725 }, { "epoch": 0.07, "learning_rate": 4.961226934225687e-05, "loss": 1.4762, "step": 726 }, { "epoch": 0.07, "learning_rate": 4.961101068716477e-05, "loss": 1.4259, "step": 727 }, { "epoch": 0.07, "learning_rate": 4.960975000847047e-05, "loss": 1.4608, "step": 728 }, { "epoch": 0.07, "learning_rate": 4.9608487306277616e-05, "loss": 1.4236, "step": 729 }, { "epoch": 0.07, "learning_rate": 4.960722258069004e-05, "loss": 1.4777, "step": 730 }, { "epoch": 0.07, "learning_rate": 4.960595583181172e-05, "loss": 1.4431, "step": 731 }, { "epoch": 0.07, "learning_rate": 4.9604687059746825e-05, "loss": 1.4625, "step": 732 }, { "epoch": 0.07, "learning_rate": 4.960341626459968e-05, "loss": 1.474, "step": 733 }, { "epoch": 0.07, "learning_rate": 4.960214344647476e-05, "loss": 1.4113, "step": 734 }, { "epoch": 0.07, "learning_rate": 4.9600868605476736e-05, "loss": 1.4452, "step": 735 }, { "epoch": 0.07, "learning_rate": 4.9599591741710425e-05, "loss": 1.461, "step": 736 }, { "epoch": 0.07, "learning_rate": 4.959831285528081e-05, "loss": 1.4369, "step": 737 }, { "epoch": 0.07, "learning_rate": 4.9597031946293036e-05, "loss": 1.4305, "step": 738 }, { "epoch": 0.07, "learning_rate": 4.959574901485245e-05, "loss": 1.4352, "step": 739 }, { "epoch": 0.07, "learning_rate": 4.959446406106452e-05, "loss": 1.4151, "step": 740 }, { "epoch": 0.07, "learning_rate": 4.95931770850349e-05, "loss": 1.4637, "step": 741 }, { "epoch": 0.07, "learning_rate": 4.9591888086869425e-05, "loss": 1.4275, "step": 742 }, { "epoch": 0.07, "learning_rate": 4.959059706667406e-05, "loss": 1.4539, "step": 743 }, { "epoch": 0.07, "learning_rate": 4.958930402455497e-05, "loss": 1.4579, "step": 744 }, { "epoch": 0.07, "learning_rate": 4.9588008960618475e-05, "loss": 1.4403, "step": 745 }, { "epoch": 0.07, "learning_rate": 4.958671187497105e-05, "loss": 1.4215, "step": 746 }, { "epoch": 0.07, "learning_rate": 4.958541276771935e-05, "loss": 1.4183, "step": 747 }, { "epoch": 0.07, "learning_rate": 4.95841116389702e-05, "loss": 1.4439, "step": 748 }, { "epoch": 0.07, "learning_rate": 4.9582808488830565e-05, "loss": 1.4095, "step": 749 }, { "epoch": 0.07, "learning_rate": 4.958150331740762e-05, "loss": 1.5359, "step": 750 }, { "epoch": 0.07, "learning_rate": 4.958019612480865e-05, "loss": 1.4609, "step": 751 }, { "epoch": 0.07, "learning_rate": 4.957888691114117e-05, "loss": 1.4401, "step": 752 }, { "epoch": 0.07, "learning_rate": 4.95775756765128e-05, "loss": 1.4337, "step": 753 }, { "epoch": 0.07, "learning_rate": 4.957626242103136e-05, "loss": 1.4919, "step": 754 }, { "epoch": 0.07, "learning_rate": 4.957494714480485e-05, "loss": 1.4051, "step": 755 }, { "epoch": 0.07, "learning_rate": 4.9573629847941395e-05, "loss": 1.4216, "step": 756 }, { "epoch": 0.07, "learning_rate": 4.957231053054931e-05, "loss": 1.4145, "step": 757 }, { "epoch": 0.07, "learning_rate": 4.9570989192737096e-05, "loss": 1.4922, "step": 758 }, { "epoch": 0.07, "learning_rate": 4.956966583461337e-05, "loss": 1.499, "step": 759 }, { "epoch": 0.07, "learning_rate": 4.956834045628695e-05, "loss": 1.3605, "step": 760 }, { "epoch": 0.07, "learning_rate": 4.956701305786682e-05, "loss": 1.476, "step": 761 }, { "epoch": 0.07, "learning_rate": 4.956568363946212e-05, "loss": 1.4413, "step": 762 }, { "epoch": 0.07, "learning_rate": 4.956435220118217e-05, "loss": 1.4668, "step": 763 }, { "epoch": 0.07, "learning_rate": 4.9563018743136424e-05, "loss": 1.4033, "step": 764 }, { "epoch": 0.07, "learning_rate": 4.956168326543454e-05, "loss": 1.4449, "step": 765 }, { "epoch": 0.07, "learning_rate": 4.956034576818632e-05, "loss": 1.4391, "step": 766 }, { "epoch": 0.07, "learning_rate": 4.955900625150174e-05, "loss": 1.4449, "step": 767 }, { "epoch": 0.07, "learning_rate": 4.955766471549093e-05, "loss": 1.4512, "step": 768 }, { "epoch": 0.07, "learning_rate": 4.9556321160264206e-05, "loss": 1.4591, "step": 769 }, { "epoch": 0.07, "learning_rate": 4.955497558593204e-05, "loss": 1.3484, "step": 770 }, { "epoch": 0.07, "learning_rate": 4.9553627992605066e-05, "loss": 1.4593, "step": 771 }, { "epoch": 0.07, "learning_rate": 4.955227838039409e-05, "loss": 1.4228, "step": 772 }, { "epoch": 0.07, "learning_rate": 4.955092674941008e-05, "loss": 1.4842, "step": 773 }, { "epoch": 0.07, "learning_rate": 4.9549573099764174e-05, "loss": 1.4158, "step": 774 }, { "epoch": 0.07, "learning_rate": 4.9548217431567665e-05, "loss": 1.4009, "step": 775 }, { "epoch": 0.07, "learning_rate": 4.954685974493204e-05, "loss": 1.4643, "step": 776 }, { "epoch": 0.07, "learning_rate": 4.954550003996891e-05, "loss": 1.4338, "step": 777 }, { "epoch": 0.07, "learning_rate": 4.954413831679009e-05, "loss": 1.4511, "step": 778 }, { "epoch": 0.07, "learning_rate": 4.954277457550754e-05, "loss": 1.4405, "step": 779 }, { "epoch": 0.07, "learning_rate": 4.9541408816233394e-05, "loss": 1.4932, "step": 780 }, { "epoch": 0.07, "learning_rate": 4.954004103907994e-05, "loss": 1.4776, "step": 781 }, { "epoch": 0.07, "learning_rate": 4.9538671244159654e-05, "loss": 1.4215, "step": 782 }, { "epoch": 0.07, "learning_rate": 4.953729943158516e-05, "loss": 1.4081, "step": 783 }, { "epoch": 0.07, "learning_rate": 4.953592560146925e-05, "loss": 1.4835, "step": 784 }, { "epoch": 0.07, "learning_rate": 4.953454975392489e-05, "loss": 1.4263, "step": 785 }, { "epoch": 0.07, "learning_rate": 4.953317188906521e-05, "loss": 1.4302, "step": 786 }, { "epoch": 0.07, "learning_rate": 4.9531792007003496e-05, "loss": 1.4752, "step": 787 }, { "epoch": 0.07, "learning_rate": 4.953041010785321e-05, "loss": 1.4504, "step": 788 }, { "epoch": 0.07, "learning_rate": 4.9529026191727965e-05, "loss": 1.4248, "step": 789 }, { "epoch": 0.07, "learning_rate": 4.952764025874157e-05, "loss": 1.4983, "step": 790 }, { "epoch": 0.07, "learning_rate": 4.9526252309007976e-05, "loss": 1.4299, "step": 791 }, { "epoch": 0.07, "learning_rate": 4.95248623426413e-05, "loss": 1.4294, "step": 792 }, { "epoch": 0.07, "learning_rate": 4.952347035975583e-05, "loss": 1.4341, "step": 793 }, { "epoch": 0.07, "learning_rate": 4.952207636046602e-05, "loss": 1.4304, "step": 794 }, { "epoch": 0.07, "learning_rate": 4.9520680344886494e-05, "loss": 1.5109, "step": 795 }, { "epoch": 0.07, "learning_rate": 4.951928231313203e-05, "loss": 1.4223, "step": 796 }, { "epoch": 0.07, "learning_rate": 4.951788226531758e-05, "loss": 1.4557, "step": 797 }, { "epoch": 0.07, "learning_rate": 4.951648020155828e-05, "loss": 1.3503, "step": 798 }, { "epoch": 0.07, "learning_rate": 4.951507612196938e-05, "loss": 1.4136, "step": 799 }, { "epoch": 0.07, "learning_rate": 4.951367002666635e-05, "loss": 1.426, "step": 800 }, { "epoch": 0.07, "learning_rate": 4.95122619157648e-05, "loss": 1.447, "step": 801 }, { "epoch": 0.07, "learning_rate": 4.951085178938051e-05, "loss": 1.391, "step": 802 }, { "epoch": 0.07, "learning_rate": 4.950943964762942e-05, "loss": 1.4143, "step": 803 }, { "epoch": 0.07, "learning_rate": 4.950802549062764e-05, "loss": 1.4414, "step": 804 }, { "epoch": 0.07, "learning_rate": 4.950660931849146e-05, "loss": 1.4242, "step": 805 }, { "epoch": 0.07, "learning_rate": 4.9505191131337314e-05, "loss": 1.4539, "step": 806 }, { "epoch": 0.07, "learning_rate": 4.950377092928181e-05, "loss": 1.397, "step": 807 }, { "epoch": 0.07, "learning_rate": 4.9502348712441716e-05, "loss": 1.4288, "step": 808 }, { "epoch": 0.07, "learning_rate": 4.950092448093398e-05, "loss": 1.3883, "step": 809 }, { "epoch": 0.07, "learning_rate": 4.9499498234875715e-05, "loss": 1.4343, "step": 810 }, { "epoch": 0.07, "learning_rate": 4.9498069974384176e-05, "loss": 1.4443, "step": 811 }, { "epoch": 0.07, "learning_rate": 4.94966396995768e-05, "loss": 1.4339, "step": 812 }, { "epoch": 0.07, "learning_rate": 4.94952074105712e-05, "loss": 1.4211, "step": 813 }, { "epoch": 0.07, "learning_rate": 4.9493773107485144e-05, "loss": 1.3515, "step": 814 }, { "epoch": 0.07, "learning_rate": 4.949233679043655e-05, "loss": 1.4449, "step": 815 }, { "epoch": 0.07, "learning_rate": 4.9490898459543535e-05, "loss": 1.4327, "step": 816 }, { "epoch": 0.07, "learning_rate": 4.9489458114924346e-05, "loss": 1.4612, "step": 817 }, { "epoch": 0.07, "learning_rate": 4.948801575669743e-05, "loss": 1.4503, "step": 818 }, { "epoch": 0.07, "learning_rate": 4.9486571384981376e-05, "loss": 1.4511, "step": 819 }, { "epoch": 0.07, "learning_rate": 4.948512499989494e-05, "loss": 1.4979, "step": 820 }, { "epoch": 0.07, "learning_rate": 4.948367660155705e-05, "loss": 1.4138, "step": 821 }, { "epoch": 0.07, "learning_rate": 4.9482226190086814e-05, "loss": 1.4367, "step": 822 }, { "epoch": 0.07, "learning_rate": 4.9480773765603464e-05, "loss": 1.4238, "step": 823 }, { "epoch": 0.07, "learning_rate": 4.9479319328226444e-05, "loss": 1.4218, "step": 824 }, { "epoch": 0.07, "learning_rate": 4.9477862878075335e-05, "loss": 1.4257, "step": 825 }, { "epoch": 0.07, "learning_rate": 4.947640441526989e-05, "loss": 1.3928, "step": 826 }, { "epoch": 0.07, "learning_rate": 4.947494393993003e-05, "loss": 1.4296, "step": 827 }, { "epoch": 0.07, "learning_rate": 4.947348145217584e-05, "loss": 1.4241, "step": 828 }, { "epoch": 0.07, "learning_rate": 4.947201695212758e-05, "loss": 1.4401, "step": 829 }, { "epoch": 0.07, "learning_rate": 4.947055043990565e-05, "loss": 1.3636, "step": 830 }, { "epoch": 0.08, "learning_rate": 4.946908191563064e-05, "loss": 1.4019, "step": 831 }, { "epoch": 0.08, "learning_rate": 4.94676113794233e-05, "loss": 1.4131, "step": 832 }, { "epoch": 0.08, "learning_rate": 4.946613883140453e-05, "loss": 1.4359, "step": 833 }, { "epoch": 0.08, "learning_rate": 4.946466427169543e-05, "loss": 1.3895, "step": 834 }, { "epoch": 0.08, "learning_rate": 4.9463187700417225e-05, "loss": 1.3998, "step": 835 }, { "epoch": 0.08, "learning_rate": 4.946170911769134e-05, "loss": 1.4166, "step": 836 }, { "epoch": 0.08, "learning_rate": 4.946022852363932e-05, "loss": 1.438, "step": 837 }, { "epoch": 0.08, "learning_rate": 4.9458745918382935e-05, "loss": 1.4853, "step": 838 }, { "epoch": 0.08, "learning_rate": 4.945726130204408e-05, "loss": 1.4048, "step": 839 }, { "epoch": 0.08, "learning_rate": 4.945577467474482e-05, "loss": 1.3842, "step": 840 }, { "epoch": 0.08, "learning_rate": 4.945428603660739e-05, "loss": 1.4626, "step": 841 }, { "epoch": 0.08, "learning_rate": 4.9452795387754195e-05, "loss": 1.4218, "step": 842 }, { "epoch": 0.08, "learning_rate": 4.9451302728307806e-05, "loss": 1.4206, "step": 843 }, { "epoch": 0.08, "learning_rate": 4.944980805839095e-05, "loss": 1.4248, "step": 844 }, { "epoch": 0.08, "learning_rate": 4.944831137812652e-05, "loss": 1.3891, "step": 845 }, { "epoch": 0.08, "learning_rate": 4.944681268763759e-05, "loss": 1.3721, "step": 846 }, { "epoch": 0.08, "learning_rate": 4.9445311987047374e-05, "loss": 1.4341, "step": 847 }, { "epoch": 0.08, "learning_rate": 4.944380927647927e-05, "loss": 1.3905, "step": 848 }, { "epoch": 0.08, "learning_rate": 4.944230455605684e-05, "loss": 1.433, "step": 849 }, { "epoch": 0.08, "learning_rate": 4.94407978259038e-05, "loss": 1.4764, "step": 850 }, { "epoch": 0.08, "learning_rate": 4.9439289086144046e-05, "loss": 1.4347, "step": 851 }, { "epoch": 0.08, "learning_rate": 4.9437778336901625e-05, "loss": 1.4905, "step": 852 }, { "epoch": 0.08, "learning_rate": 4.943626557830076e-05, "loss": 1.3994, "step": 853 }, { "epoch": 0.08, "learning_rate": 4.9434750810465837e-05, "loss": 1.4264, "step": 854 }, { "epoch": 0.08, "learning_rate": 4.9433234033521396e-05, "loss": 1.4105, "step": 855 }, { "epoch": 0.08, "learning_rate": 4.943171524759216e-05, "loss": 1.4326, "step": 856 }, { "epoch": 0.08, "learning_rate": 4.9430194452803014e-05, "loss": 1.3894, "step": 857 }, { "epoch": 0.08, "learning_rate": 4.942867164927899e-05, "loss": 1.4654, "step": 858 }, { "epoch": 0.08, "learning_rate": 4.9427146837145314e-05, "loss": 1.392, "step": 859 }, { "epoch": 0.08, "learning_rate": 4.942562001652734e-05, "loss": 1.4157, "step": 860 }, { "epoch": 0.08, "learning_rate": 4.942409118755063e-05, "loss": 1.4014, "step": 861 }, { "epoch": 0.08, "learning_rate": 4.9422560350340864e-05, "loss": 1.4645, "step": 862 }, { "epoch": 0.08, "learning_rate": 4.942102750502394e-05, "loss": 1.3929, "step": 863 }, { "epoch": 0.08, "learning_rate": 4.941949265172588e-05, "loss": 1.4145, "step": 864 }, { "epoch": 0.08, "learning_rate": 4.941795579057289e-05, "loss": 1.4257, "step": 865 }, { "epoch": 0.08, "learning_rate": 4.941641692169133e-05, "loss": 1.4138, "step": 866 }, { "epoch": 0.08, "learning_rate": 4.9414876045207735e-05, "loss": 1.4503, "step": 867 }, { "epoch": 0.08, "learning_rate": 4.9413333161248795e-05, "loss": 1.4046, "step": 868 }, { "epoch": 0.08, "learning_rate": 4.941178826994139e-05, "loss": 1.4623, "step": 869 }, { "epoch": 0.08, "learning_rate": 4.9410241371412525e-05, "loss": 1.3934, "step": 870 }, { "epoch": 0.08, "learning_rate": 4.94086924657894e-05, "loss": 1.4551, "step": 871 }, { "epoch": 0.08, "learning_rate": 4.9407141553199375e-05, "loss": 1.4568, "step": 872 }, { "epoch": 0.08, "learning_rate": 4.940558863376996e-05, "loss": 1.3703, "step": 873 }, { "epoch": 0.08, "learning_rate": 4.9404033707628855e-05, "loss": 1.4376, "step": 874 }, { "epoch": 0.08, "learning_rate": 4.94024767749039e-05, "loss": 1.367, "step": 875 }, { "epoch": 0.08, "learning_rate": 4.9400917835723116e-05, "loss": 1.4429, "step": 876 }, { "epoch": 0.08, "learning_rate": 4.9399356890214687e-05, "loss": 1.4217, "step": 877 }, { "epoch": 0.08, "learning_rate": 4.939779393850697e-05, "loss": 1.4689, "step": 878 }, { "epoch": 0.08, "learning_rate": 4.939622898072844e-05, "loss": 1.3744, "step": 879 }, { "epoch": 0.08, "learning_rate": 4.939466201700781e-05, "loss": 1.4224, "step": 880 }, { "epoch": 0.08, "learning_rate": 4.939309304747391e-05, "loss": 1.4231, "step": 881 }, { "epoch": 0.08, "learning_rate": 4.939152207225573e-05, "loss": 1.4162, "step": 882 }, { "epoch": 0.08, "learning_rate": 4.9389949091482466e-05, "loss": 1.3988, "step": 883 }, { "epoch": 0.08, "learning_rate": 4.938837410528344e-05, "loss": 1.3847, "step": 884 }, { "epoch": 0.08, "learning_rate": 4.9386797113788154e-05, "loss": 1.437, "step": 885 }, { "epoch": 0.08, "learning_rate": 4.938521811712628e-05, "loss": 1.4385, "step": 886 }, { "epoch": 0.08, "learning_rate": 4.938363711542764e-05, "loss": 1.4529, "step": 887 }, { "epoch": 0.08, "learning_rate": 4.938205410882223e-05, "loss": 1.4474, "step": 888 }, { "epoch": 0.08, "learning_rate": 4.938046909744022e-05, "loss": 1.412, "step": 889 }, { "epoch": 0.08, "learning_rate": 4.937888208141192e-05, "loss": 1.4163, "step": 890 }, { "epoch": 0.08, "learning_rate": 4.9377293060867835e-05, "loss": 1.4158, "step": 891 }, { "epoch": 0.08, "learning_rate": 4.9375702035938606e-05, "loss": 1.4647, "step": 892 }, { "epoch": 0.08, "learning_rate": 4.937410900675507e-05, "loss": 1.4761, "step": 893 }, { "epoch": 0.08, "learning_rate": 4.93725139734482e-05, "loss": 1.5045, "step": 894 }, { "epoch": 0.08, "learning_rate": 4.9370916936149144e-05, "loss": 1.4546, "step": 895 }, { "epoch": 0.08, "learning_rate": 4.936931789498922e-05, "loss": 1.3641, "step": 896 }, { "epoch": 0.08, "learning_rate": 4.93677168500999e-05, "loss": 1.4336, "step": 897 }, { "epoch": 0.08, "learning_rate": 4.936611380161283e-05, "loss": 1.4544, "step": 898 }, { "epoch": 0.08, "learning_rate": 4.936450874965982e-05, "loss": 1.4337, "step": 899 }, { "epoch": 0.08, "learning_rate": 4.9362901694372854e-05, "loss": 1.4376, "step": 900 }, { "epoch": 0.08, "learning_rate": 4.936129263588405e-05, "loss": 1.3636, "step": 901 }, { "epoch": 0.08, "learning_rate": 4.9359681574325725e-05, "loss": 1.3634, "step": 902 }, { "epoch": 0.08, "learning_rate": 4.9358068509830334e-05, "loss": 1.4015, "step": 903 }, { "epoch": 0.08, "learning_rate": 4.9356453442530526e-05, "loss": 1.4403, "step": 904 }, { "epoch": 0.08, "learning_rate": 4.9354836372559076e-05, "loss": 1.4295, "step": 905 }, { "epoch": 0.08, "learning_rate": 4.935321730004896e-05, "loss": 1.4071, "step": 906 }, { "epoch": 0.08, "learning_rate": 4.935159622513329e-05, "loss": 1.3868, "step": 907 }, { "epoch": 0.08, "learning_rate": 4.934997314794538e-05, "loss": 1.3749, "step": 908 }, { "epoch": 0.08, "learning_rate": 4.934834806861867e-05, "loss": 1.4467, "step": 909 }, { "epoch": 0.08, "learning_rate": 4.934672098728677e-05, "loss": 1.3668, "step": 910 }, { "epoch": 0.08, "learning_rate": 4.934509190408348e-05, "loss": 1.4617, "step": 911 }, { "epoch": 0.08, "learning_rate": 4.934346081914274e-05, "loss": 1.4112, "step": 912 }, { "epoch": 0.08, "learning_rate": 4.934182773259867e-05, "loss": 1.4597, "step": 913 }, { "epoch": 0.08, "learning_rate": 4.934019264458555e-05, "loss": 1.3946, "step": 914 }, { "epoch": 0.08, "learning_rate": 4.933855555523781e-05, "loss": 1.4655, "step": 915 }, { "epoch": 0.08, "learning_rate": 4.933691646469006e-05, "loss": 1.4032, "step": 916 }, { "epoch": 0.08, "learning_rate": 4.9335275373077085e-05, "loss": 1.4808, "step": 917 }, { "epoch": 0.08, "learning_rate": 4.9333632280533815e-05, "loss": 1.4905, "step": 918 }, { "epoch": 0.08, "learning_rate": 4.933198718719534e-05, "loss": 1.4058, "step": 919 }, { "epoch": 0.08, "learning_rate": 4.9330340093196935e-05, "loss": 1.4293, "step": 920 }, { "epoch": 0.08, "learning_rate": 4.932869099867403e-05, "loss": 1.4572, "step": 921 }, { "epoch": 0.08, "learning_rate": 4.932703990376222e-05, "loss": 1.4612, "step": 922 }, { "epoch": 0.08, "learning_rate": 4.932538680859726e-05, "loss": 1.3845, "step": 923 }, { "epoch": 0.08, "learning_rate": 4.932373171331507e-05, "loss": 1.4486, "step": 924 }, { "epoch": 0.08, "learning_rate": 4.932207461805175e-05, "loss": 1.4405, "step": 925 }, { "epoch": 0.08, "learning_rate": 4.932041552294353e-05, "loss": 1.4655, "step": 926 }, { "epoch": 0.08, "learning_rate": 4.9318754428126854e-05, "loss": 1.3445, "step": 927 }, { "epoch": 0.08, "learning_rate": 4.9317091333738284e-05, "loss": 1.4923, "step": 928 }, { "epoch": 0.08, "learning_rate": 4.931542623991458e-05, "loss": 1.4592, "step": 929 }, { "epoch": 0.08, "learning_rate": 4.931375914679263e-05, "loss": 1.3938, "step": 930 }, { "epoch": 0.08, "learning_rate": 4.9312090054509515e-05, "loss": 1.4509, "step": 931 }, { "epoch": 0.08, "learning_rate": 4.93104189632025e-05, "loss": 1.4443, "step": 932 }, { "epoch": 0.08, "learning_rate": 4.9308745873008955e-05, "loss": 1.4841, "step": 933 }, { "epoch": 0.08, "learning_rate": 4.9307070784066455e-05, "loss": 1.3949, "step": 934 }, { "epoch": 0.08, "learning_rate": 4.9305393696512744e-05, "loss": 1.3878, "step": 935 }, { "epoch": 0.08, "learning_rate": 4.930371461048571e-05, "loss": 1.449, "step": 936 }, { "epoch": 0.08, "learning_rate": 4.9302033526123416e-05, "loss": 1.4192, "step": 937 }, { "epoch": 0.08, "learning_rate": 4.930035044356407e-05, "loss": 1.3565, "step": 938 }, { "epoch": 0.08, "learning_rate": 4.929866536294609e-05, "loss": 1.4123, "step": 939 }, { "epoch": 0.08, "learning_rate": 4.9296978284408e-05, "loss": 1.3909, "step": 940 }, { "epoch": 0.09, "learning_rate": 4.929528920808854e-05, "loss": 1.4409, "step": 941 }, { "epoch": 0.09, "learning_rate": 4.929359813412658e-05, "loss": 1.4089, "step": 942 }, { "epoch": 0.09, "learning_rate": 4.9291905062661175e-05, "loss": 1.4234, "step": 943 }, { "epoch": 0.09, "learning_rate": 4.9290209993831525e-05, "loss": 1.401, "step": 944 }, { "epoch": 0.09, "learning_rate": 4.928851292777701e-05, "loss": 1.3862, "step": 945 }, { "epoch": 0.09, "learning_rate": 4.928681386463716e-05, "loss": 1.3996, "step": 946 }, { "epoch": 0.09, "learning_rate": 4.928511280455169e-05, "loss": 1.4117, "step": 947 }, { "epoch": 0.09, "learning_rate": 4.9283409747660466e-05, "loss": 1.3376, "step": 948 }, { "epoch": 0.09, "learning_rate": 4.928170469410351e-05, "loss": 1.447, "step": 949 }, { "epoch": 0.09, "learning_rate": 4.9279997644021025e-05, "loss": 1.3996, "step": 950 }, { "epoch": 0.09, "learning_rate": 4.927828859755337e-05, "loss": 1.4737, "step": 951 }, { "epoch": 0.09, "learning_rate": 4.927657755484106e-05, "loss": 1.4252, "step": 952 }, { "epoch": 0.09, "learning_rate": 4.927486451602479e-05, "loss": 1.4631, "step": 953 }, { "epoch": 0.09, "learning_rate": 4.9273149481245416e-05, "loss": 1.3774, "step": 954 }, { "epoch": 0.09, "learning_rate": 4.927143245064394e-05, "loss": 1.4422, "step": 955 }, { "epoch": 0.09, "learning_rate": 4.9269713424361564e-05, "loss": 1.3545, "step": 956 }, { "epoch": 0.09, "learning_rate": 4.926799240253962e-05, "loss": 1.4186, "step": 957 }, { "epoch": 0.09, "learning_rate": 4.9266269385319606e-05, "loss": 1.4458, "step": 958 }, { "epoch": 0.09, "learning_rate": 4.926454437284321e-05, "loss": 1.3852, "step": 959 }, { "epoch": 0.09, "learning_rate": 4.926281736525227e-05, "loss": 1.3598, "step": 960 }, { "epoch": 0.09, "learning_rate": 4.926108836268877e-05, "loss": 1.442, "step": 961 }, { "epoch": 0.09, "learning_rate": 4.925935736529489e-05, "loss": 1.418, "step": 962 }, { "epoch": 0.09, "learning_rate": 4.925762437321295e-05, "loss": 1.3809, "step": 963 }, { "epoch": 0.09, "learning_rate": 4.9255889386585445e-05, "loss": 1.3615, "step": 964 }, { "epoch": 0.09, "learning_rate": 4.9254152405555034e-05, "loss": 1.3946, "step": 965 }, { "epoch": 0.09, "learning_rate": 4.925241343026454e-05, "loss": 1.3964, "step": 966 }, { "epoch": 0.09, "learning_rate": 4.925067246085694e-05, "loss": 1.4462, "step": 967 }, { "epoch": 0.09, "learning_rate": 4.924892949747538e-05, "loss": 1.3686, "step": 968 }, { "epoch": 0.09, "learning_rate": 4.924718454026318e-05, "loss": 1.3787, "step": 969 }, { "epoch": 0.09, "learning_rate": 4.924543758936382e-05, "loss": 1.4457, "step": 970 }, { "epoch": 0.09, "learning_rate": 4.924368864492093e-05, "loss": 1.4001, "step": 971 }, { "epoch": 0.09, "learning_rate": 4.9241937707078325e-05, "loss": 1.386, "step": 972 }, { "epoch": 0.09, "learning_rate": 4.924018477597996e-05, "loss": 1.412, "step": 973 }, { "epoch": 0.09, "learning_rate": 4.923842985176998e-05, "loss": 1.3575, "step": 974 }, { "epoch": 0.09, "learning_rate": 4.923667293459267e-05, "loss": 1.4078, "step": 975 }, { "epoch": 0.09, "learning_rate": 4.923491402459251e-05, "loss": 1.4034, "step": 976 }, { "epoch": 0.09, "learning_rate": 4.923315312191409e-05, "loss": 1.4452, "step": 977 }, { "epoch": 0.09, "learning_rate": 4.9231390226702226e-05, "loss": 1.3669, "step": 978 }, { "epoch": 0.09, "learning_rate": 4.922962533910186e-05, "loss": 1.4258, "step": 979 }, { "epoch": 0.09, "learning_rate": 4.92278584592581e-05, "loss": 1.4548, "step": 980 }, { "epoch": 0.09, "learning_rate": 4.9226089587316236e-05, "loss": 1.4192, "step": 981 }, { "epoch": 0.09, "learning_rate": 4.92243187234217e-05, "loss": 1.3915, "step": 982 }, { "epoch": 0.09, "learning_rate": 4.9222545867720116e-05, "loss": 1.4032, "step": 983 }, { "epoch": 0.09, "learning_rate": 4.922077102035724e-05, "loss": 1.4071, "step": 984 }, { "epoch": 0.09, "learning_rate": 4.9218994181479006e-05, "loss": 1.4277, "step": 985 }, { "epoch": 0.09, "learning_rate": 4.921721535123152e-05, "loss": 1.4127, "step": 986 }, { "epoch": 0.09, "learning_rate": 4.921543452976103e-05, "loss": 1.4215, "step": 987 }, { "epoch": 0.09, "learning_rate": 4.921365171721398e-05, "loss": 1.3644, "step": 988 }, { "epoch": 0.09, "learning_rate": 4.9211866913736944e-05, "loss": 1.3997, "step": 989 }, { "epoch": 0.09, "learning_rate": 4.921008011947668e-05, "loss": 1.3874, "step": 990 }, { "epoch": 0.09, "learning_rate": 4.9208291334580104e-05, "loss": 1.4137, "step": 991 }, { "epoch": 0.09, "learning_rate": 4.9206500559194296e-05, "loss": 1.407, "step": 992 }, { "epoch": 0.09, "learning_rate": 4.9204707793466495e-05, "loss": 1.4133, "step": 993 }, { "epoch": 0.09, "learning_rate": 4.9202913037544124e-05, "loss": 1.3837, "step": 994 }, { "epoch": 0.09, "learning_rate": 4.920111629157474e-05, "loss": 1.4465, "step": 995 }, { "epoch": 0.09, "learning_rate": 4.919931755570608e-05, "loss": 1.5064, "step": 996 }, { "epoch": 0.09, "learning_rate": 4.919751683008604e-05, "loss": 1.3613, "step": 997 }, { "epoch": 0.09, "learning_rate": 4.919571411486269e-05, "loss": 1.4393, "step": 998 }, { "epoch": 0.09, "learning_rate": 4.9193909410184244e-05, "loss": 1.3979, "step": 999 }, { "epoch": 0.09, "learning_rate": 4.9192102716199105e-05, "loss": 1.379, "step": 1000 }, { "epoch": 0.09, "learning_rate": 4.919029403305581e-05, "loss": 1.443, "step": 1001 }, { "epoch": 0.09, "learning_rate": 4.918848336090309e-05, "loss": 1.3686, "step": 1002 }, { "epoch": 0.09, "learning_rate": 4.918667069988981e-05, "loss": 1.4175, "step": 1003 }, { "epoch": 0.09, "learning_rate": 4.9184856050165035e-05, "loss": 1.4438, "step": 1004 }, { "epoch": 0.09, "learning_rate": 4.918303941187795e-05, "loss": 1.4319, "step": 1005 }, { "epoch": 0.09, "learning_rate": 4.918122078517794e-05, "loss": 1.4568, "step": 1006 }, { "epoch": 0.09, "learning_rate": 4.9179400170214515e-05, "loss": 1.3772, "step": 1007 }, { "epoch": 0.09, "learning_rate": 4.917757756713741e-05, "loss": 1.4339, "step": 1008 }, { "epoch": 0.09, "learning_rate": 4.9175752976096455e-05, "loss": 1.3692, "step": 1009 }, { "epoch": 0.09, "learning_rate": 4.917392639724168e-05, "loss": 1.4141, "step": 1010 }, { "epoch": 0.09, "learning_rate": 4.917209783072328e-05, "loss": 1.3442, "step": 1011 }, { "epoch": 0.09, "learning_rate": 4.917026727669161e-05, "loss": 1.4445, "step": 1012 }, { "epoch": 0.09, "learning_rate": 4.9168434735297173e-05, "loss": 1.4208, "step": 1013 }, { "epoch": 0.09, "learning_rate": 4.9166600206690647e-05, "loss": 1.3909, "step": 1014 }, { "epoch": 0.09, "learning_rate": 4.916476369102289e-05, "loss": 1.3884, "step": 1015 }, { "epoch": 0.09, "learning_rate": 4.916292518844487e-05, "loss": 1.3326, "step": 1016 }, { "epoch": 0.09, "learning_rate": 4.916108469910781e-05, "loss": 1.404, "step": 1017 }, { "epoch": 0.09, "learning_rate": 4.915924222316299e-05, "loss": 1.3599, "step": 1018 }, { "epoch": 0.09, "learning_rate": 4.9157397760761926e-05, "loss": 1.4075, "step": 1019 }, { "epoch": 0.09, "learning_rate": 4.915555131205628e-05, "loss": 1.4005, "step": 1020 }, { "epoch": 0.09, "learning_rate": 4.915370287719787e-05, "loss": 1.3846, "step": 1021 }, { "epoch": 0.09, "learning_rate": 4.9151852456338677e-05, "loss": 1.4335, "step": 1022 }, { "epoch": 0.09, "learning_rate": 4.9150000049630854e-05, "loss": 1.4105, "step": 1023 }, { "epoch": 0.09, "learning_rate": 4.914814565722671e-05, "loss": 1.4563, "step": 1024 }, { "epoch": 0.09, "learning_rate": 4.914628927927872e-05, "loss": 1.3737, "step": 1025 }, { "epoch": 0.09, "learning_rate": 4.914443091593952e-05, "loss": 1.4313, "step": 1026 }, { "epoch": 0.09, "learning_rate": 4.91425705673619e-05, "loss": 1.4186, "step": 1027 }, { "epoch": 0.09, "learning_rate": 4.9140708233698854e-05, "loss": 1.4275, "step": 1028 }, { "epoch": 0.09, "learning_rate": 4.9138843915103485e-05, "loss": 1.4233, "step": 1029 }, { "epoch": 0.09, "learning_rate": 4.913697761172909e-05, "loss": 1.38, "step": 1030 }, { "epoch": 0.09, "learning_rate": 4.913510932372912e-05, "loss": 1.4219, "step": 1031 }, { "epoch": 0.09, "learning_rate": 4.913323905125721e-05, "loss": 1.4202, "step": 1032 }, { "epoch": 0.09, "learning_rate": 4.913136679446711e-05, "loss": 1.4103, "step": 1033 }, { "epoch": 0.09, "learning_rate": 4.9129492553512784e-05, "loss": 1.3968, "step": 1034 }, { "epoch": 0.09, "learning_rate": 4.912761632854833e-05, "loss": 1.3798, "step": 1035 }, { "epoch": 0.09, "learning_rate": 4.912573811972803e-05, "loss": 1.3995, "step": 1036 }, { "epoch": 0.09, "learning_rate": 4.9123857927206304e-05, "loss": 1.4147, "step": 1037 }, { "epoch": 0.09, "learning_rate": 4.912197575113776e-05, "loss": 1.4457, "step": 1038 }, { "epoch": 0.09, "learning_rate": 4.912009159167714e-05, "loss": 1.4167, "step": 1039 }, { "epoch": 0.09, "learning_rate": 4.911820544897937e-05, "loss": 1.3484, "step": 1040 }, { "epoch": 0.09, "learning_rate": 4.911631732319955e-05, "loss": 1.4193, "step": 1041 }, { "epoch": 0.09, "learning_rate": 4.911442721449292e-05, "loss": 1.3883, "step": 1042 }, { "epoch": 0.09, "learning_rate": 4.911253512301488e-05, "loss": 1.3928, "step": 1043 }, { "epoch": 0.09, "learning_rate": 4.911064104892103e-05, "loss": 1.4398, "step": 1044 }, { "epoch": 0.09, "learning_rate": 4.910874499236707e-05, "loss": 1.3874, "step": 1045 }, { "epoch": 0.09, "learning_rate": 4.910684695350895e-05, "loss": 1.4292, "step": 1046 }, { "epoch": 0.09, "learning_rate": 4.910494693250268e-05, "loss": 1.3797, "step": 1047 }, { "epoch": 0.09, "learning_rate": 4.910304492950452e-05, "loss": 1.4308, "step": 1048 }, { "epoch": 0.09, "learning_rate": 4.9101140944670854e-05, "loss": 1.3973, "step": 1049 }, { "epoch": 0.09, "learning_rate": 4.909923497815823e-05, "loss": 1.3787, "step": 1050 }, { "epoch": 0.09, "learning_rate": 4.909732703012336e-05, "loss": 1.426, "step": 1051 }, { "epoch": 0.1, "learning_rate": 4.909541710072313e-05, "loss": 1.3922, "step": 1052 }, { "epoch": 0.1, "learning_rate": 4.9093505190114574e-05, "loss": 1.4458, "step": 1053 }, { "epoch": 0.1, "learning_rate": 4.9091591298454896e-05, "loss": 1.4, "step": 1054 }, { "epoch": 0.1, "learning_rate": 4.908967542590147e-05, "loss": 1.3751, "step": 1055 }, { "epoch": 0.1, "learning_rate": 4.9087757572611826e-05, "loss": 1.4089, "step": 1056 }, { "epoch": 0.1, "learning_rate": 4.9085837738743645e-05, "loss": 1.3392, "step": 1057 }, { "epoch": 0.1, "learning_rate": 4.9083915924454795e-05, "loss": 1.3792, "step": 1058 }, { "epoch": 0.1, "learning_rate": 4.9081992129903284e-05, "loss": 1.4001, "step": 1059 }, { "epoch": 0.1, "learning_rate": 4.9080066355247304e-05, "loss": 1.4513, "step": 1060 }, { "epoch": 0.1, "learning_rate": 4.907813860064518e-05, "loss": 1.3675, "step": 1061 }, { "epoch": 0.1, "learning_rate": 4.907620886625544e-05, "loss": 1.4414, "step": 1062 }, { "epoch": 0.1, "learning_rate": 4.907427715223674e-05, "loss": 1.3141, "step": 1063 }, { "epoch": 0.1, "learning_rate": 4.907234345874792e-05, "loss": 1.4323, "step": 1064 }, { "epoch": 0.1, "learning_rate": 4.907040778594797e-05, "loss": 1.3849, "step": 1065 }, { "epoch": 0.1, "learning_rate": 4.9068470133996056e-05, "loss": 1.3776, "step": 1066 }, { "epoch": 0.1, "learning_rate": 4.906653050305148e-05, "loss": 1.4113, "step": 1067 }, { "epoch": 0.1, "learning_rate": 4.906458889327374e-05, "loss": 1.462, "step": 1068 }, { "epoch": 0.1, "learning_rate": 4.906264530482249e-05, "loss": 1.4493, "step": 1069 }, { "epoch": 0.1, "learning_rate": 4.9060699737857515e-05, "loss": 1.4118, "step": 1070 }, { "epoch": 0.1, "learning_rate": 4.90587521925388e-05, "loss": 1.3526, "step": 1071 }, { "epoch": 0.1, "learning_rate": 4.905680266902648e-05, "loss": 1.3638, "step": 1072 }, { "epoch": 0.1, "learning_rate": 4.9054851167480846e-05, "loss": 1.3788, "step": 1073 }, { "epoch": 0.1, "learning_rate": 4.9052897688062364e-05, "loss": 1.3742, "step": 1074 }, { "epoch": 0.1, "learning_rate": 4.905094223093165e-05, "loss": 1.4061, "step": 1075 }, { "epoch": 0.1, "learning_rate": 4.9048984796249494e-05, "loss": 1.4219, "step": 1076 }, { "epoch": 0.1, "learning_rate": 4.904702538417683e-05, "loss": 1.3826, "step": 1077 }, { "epoch": 0.1, "learning_rate": 4.9045063994874784e-05, "loss": 1.4087, "step": 1078 }, { "epoch": 0.1, "learning_rate": 4.9043100628504625e-05, "loss": 1.4404, "step": 1079 }, { "epoch": 0.1, "learning_rate": 4.904113528522778e-05, "loss": 1.3977, "step": 1080 }, { "epoch": 0.1, "learning_rate": 4.903916796520585e-05, "loss": 1.4317, "step": 1081 }, { "epoch": 0.1, "learning_rate": 4.903719866860059e-05, "loss": 1.3754, "step": 1082 }, { "epoch": 0.1, "learning_rate": 4.903522739557393e-05, "loss": 1.3947, "step": 1083 }, { "epoch": 0.1, "learning_rate": 4.9033254146287955e-05, "loss": 1.3279, "step": 1084 }, { "epoch": 0.1, "learning_rate": 4.903127892090491e-05, "loss": 1.3112, "step": 1085 }, { "epoch": 0.1, "learning_rate": 4.90293017195872e-05, "loss": 1.4137, "step": 1086 }, { "epoch": 0.1, "learning_rate": 4.902732254249741e-05, "loss": 1.3766, "step": 1087 }, { "epoch": 0.1, "learning_rate": 4.902534138979826e-05, "loss": 1.4167, "step": 1088 }, { "epoch": 0.1, "learning_rate": 4.902335826165265e-05, "loss": 1.4538, "step": 1089 }, { "epoch": 0.1, "learning_rate": 4.902137315822365e-05, "loss": 1.4041, "step": 1090 }, { "epoch": 0.1, "learning_rate": 4.9019386079674476e-05, "loss": 1.383, "step": 1091 }, { "epoch": 0.1, "learning_rate": 4.901739702616851e-05, "loss": 1.4063, "step": 1092 }, { "epoch": 0.1, "learning_rate": 4.90154059978693e-05, "loss": 1.4092, "step": 1093 }, { "epoch": 0.1, "learning_rate": 4.901341299494055e-05, "loss": 1.3985, "step": 1094 }, { "epoch": 0.1, "learning_rate": 4.901141801754615e-05, "loss": 1.4162, "step": 1095 }, { "epoch": 0.1, "learning_rate": 4.900942106585012e-05, "loss": 1.3295, "step": 1096 }, { "epoch": 0.1, "learning_rate": 4.900742214001665e-05, "loss": 1.4512, "step": 1097 }, { "epoch": 0.1, "learning_rate": 4.9005421240210115e-05, "loss": 1.4555, "step": 1098 }, { "epoch": 0.1, "learning_rate": 4.900341836659502e-05, "loss": 1.4411, "step": 1099 }, { "epoch": 0.1, "learning_rate": 4.9001413519336056e-05, "loss": 1.3871, "step": 1100 }, { "epoch": 0.1, "learning_rate": 4.8999406698598074e-05, "loss": 1.3828, "step": 1101 }, { "epoch": 0.1, "learning_rate": 4.899739790454607e-05, "loss": 1.3737, "step": 1102 }, { "epoch": 0.1, "learning_rate": 4.899538713734522e-05, "loss": 1.4409, "step": 1103 }, { "epoch": 0.1, "learning_rate": 4.899337439716086e-05, "loss": 1.3442, "step": 1104 }, { "epoch": 0.1, "learning_rate": 4.899135968415848e-05, "loss": 1.4568, "step": 1105 }, { "epoch": 0.1, "learning_rate": 4.8989342998503726e-05, "loss": 1.3968, "step": 1106 }, { "epoch": 0.1, "learning_rate": 4.898732434036244e-05, "loss": 1.3673, "step": 1107 }, { "epoch": 0.1, "learning_rate": 4.8985303709900585e-05, "loss": 1.4369, "step": 1108 }, { "epoch": 0.1, "learning_rate": 4.898328110728431e-05, "loss": 1.4353, "step": 1109 }, { "epoch": 0.1, "learning_rate": 4.898125653267992e-05, "loss": 1.4026, "step": 1110 }, { "epoch": 0.1, "learning_rate": 4.8979229986253884e-05, "loss": 1.4433, "step": 1111 }, { "epoch": 0.1, "learning_rate": 4.8977201468172835e-05, "loss": 1.4093, "step": 1112 }, { "epoch": 0.1, "learning_rate": 4.897517097860356e-05, "loss": 1.3841, "step": 1113 }, { "epoch": 0.1, "learning_rate": 4.8973138517713e-05, "loss": 1.4352, "step": 1114 }, { "epoch": 0.1, "learning_rate": 4.8971104085668296e-05, "loss": 1.4143, "step": 1115 }, { "epoch": 0.1, "learning_rate": 4.8969067682636715e-05, "loss": 1.3908, "step": 1116 }, { "epoch": 0.1, "learning_rate": 4.896702930878569e-05, "loss": 1.4672, "step": 1117 }, { "epoch": 0.1, "learning_rate": 4.896498896428283e-05, "loss": 1.4179, "step": 1118 }, { "epoch": 0.1, "learning_rate": 4.89629466492959e-05, "loss": 1.3971, "step": 1119 }, { "epoch": 0.1, "learning_rate": 4.8960902363992825e-05, "loss": 1.4248, "step": 1120 }, { "epoch": 0.1, "learning_rate": 4.89588561085417e-05, "loss": 1.3778, "step": 1121 }, { "epoch": 0.1, "learning_rate": 4.8956807883110755e-05, "loss": 1.4363, "step": 1122 }, { "epoch": 0.1, "learning_rate": 4.8954757687868426e-05, "loss": 1.3464, "step": 1123 }, { "epoch": 0.1, "learning_rate": 4.8952705522983265e-05, "loss": 1.429, "step": 1124 }, { "epoch": 0.1, "learning_rate": 4.895065138862403e-05, "loss": 1.4344, "step": 1125 }, { "epoch": 0.1, "learning_rate": 4.8948595284959604e-05, "loss": 1.3663, "step": 1126 }, { "epoch": 0.1, "learning_rate": 4.894653721215905e-05, "loss": 1.383, "step": 1127 }, { "epoch": 0.1, "learning_rate": 4.894447717039159e-05, "loss": 1.3886, "step": 1128 }, { "epoch": 0.1, "learning_rate": 4.8942415159826615e-05, "loss": 1.3707, "step": 1129 }, { "epoch": 0.1, "learning_rate": 4.894035118063366e-05, "loss": 1.3865, "step": 1130 }, { "epoch": 0.1, "learning_rate": 4.893828523298244e-05, "loss": 1.3953, "step": 1131 }, { "epoch": 0.1, "learning_rate": 4.893621731704282e-05, "loss": 1.4297, "step": 1132 }, { "epoch": 0.1, "learning_rate": 4.8934147432984836e-05, "loss": 1.377, "step": 1133 }, { "epoch": 0.1, "learning_rate": 4.893207558097866e-05, "loss": 1.4312, "step": 1134 }, { "epoch": 0.1, "learning_rate": 4.8930001761194686e-05, "loss": 1.3722, "step": 1135 }, { "epoch": 0.1, "learning_rate": 4.89279259738034e-05, "loss": 1.4053, "step": 1136 }, { "epoch": 0.1, "learning_rate": 4.892584821897549e-05, "loss": 1.4014, "step": 1137 }, { "epoch": 0.1, "learning_rate": 4.892376849688179e-05, "loss": 1.4409, "step": 1138 }, { "epoch": 0.1, "learning_rate": 4.8921686807693327e-05, "loss": 1.3815, "step": 1139 }, { "epoch": 0.1, "learning_rate": 4.891960315158123e-05, "loss": 1.3645, "step": 1140 }, { "epoch": 0.1, "learning_rate": 4.891751752871684e-05, "loss": 1.3426, "step": 1141 }, { "epoch": 0.1, "learning_rate": 4.891542993927164e-05, "loss": 1.4513, "step": 1142 }, { "epoch": 0.1, "learning_rate": 4.8913340383417284e-05, "loss": 1.376, "step": 1143 }, { "epoch": 0.1, "learning_rate": 4.891124886132559e-05, "loss": 1.4213, "step": 1144 }, { "epoch": 0.1, "learning_rate": 4.8909155373168516e-05, "loss": 1.38, "step": 1145 }, { "epoch": 0.1, "learning_rate": 4.8907059919118195e-05, "loss": 1.4093, "step": 1146 }, { "epoch": 0.1, "learning_rate": 4.890496249934693e-05, "loss": 1.3905, "step": 1147 }, { "epoch": 0.1, "learning_rate": 4.8902863114027183e-05, "loss": 1.368, "step": 1148 }, { "epoch": 0.1, "learning_rate": 4.890076176333157e-05, "loss": 1.3606, "step": 1149 }, { "epoch": 0.1, "learning_rate": 4.8898658447432854e-05, "loss": 1.4432, "step": 1150 }, { "epoch": 0.1, "learning_rate": 4.8896553166504e-05, "loss": 1.3944, "step": 1151 }, { "epoch": 0.1, "learning_rate": 4.88944459207181e-05, "loss": 1.3722, "step": 1152 }, { "epoch": 0.1, "learning_rate": 4.889233671024842e-05, "loss": 1.3364, "step": 1153 }, { "epoch": 0.1, "learning_rate": 4.8890225535268387e-05, "loss": 1.3458, "step": 1154 }, { "epoch": 0.1, "learning_rate": 4.888811239595159e-05, "loss": 1.3367, "step": 1155 }, { "epoch": 0.1, "learning_rate": 4.8885997292471774e-05, "loss": 1.4398, "step": 1156 }, { "epoch": 0.1, "learning_rate": 4.8883880225002864e-05, "loss": 1.3629, "step": 1157 }, { "epoch": 0.1, "learning_rate": 4.888176119371891e-05, "loss": 1.4102, "step": 1158 }, { "epoch": 0.1, "learning_rate": 4.887964019879417e-05, "loss": 1.3809, "step": 1159 }, { "epoch": 0.1, "learning_rate": 4.887751724040303e-05, "loss": 1.412, "step": 1160 }, { "epoch": 0.1, "learning_rate": 4.8875392318720034e-05, "loss": 1.3982, "step": 1161 }, { "epoch": 0.1, "learning_rate": 4.887326543391992e-05, "loss": 1.3649, "step": 1162 }, { "epoch": 0.11, "learning_rate": 4.8871136586177556e-05, "loss": 1.3338, "step": 1163 }, { "epoch": 0.11, "learning_rate": 4.8869005775667985e-05, "loss": 1.4005, "step": 1164 }, { "epoch": 0.11, "learning_rate": 4.886687300256641e-05, "loss": 1.3729, "step": 1165 }, { "epoch": 0.11, "learning_rate": 4.886473826704821e-05, "loss": 1.3621, "step": 1166 }, { "epoch": 0.11, "learning_rate": 4.886260156928888e-05, "loss": 1.4211, "step": 1167 }, { "epoch": 0.11, "learning_rate": 4.8860462909464134e-05, "loss": 1.4152, "step": 1168 }, { "epoch": 0.11, "learning_rate": 4.8858322287749804e-05, "loss": 1.4478, "step": 1169 }, { "epoch": 0.11, "learning_rate": 4.8856179704321906e-05, "loss": 1.374, "step": 1170 }, { "epoch": 0.11, "learning_rate": 4.885403515935661e-05, "loss": 1.3871, "step": 1171 }, { "epoch": 0.11, "learning_rate": 4.885188865303025e-05, "loss": 1.3834, "step": 1172 }, { "epoch": 0.11, "learning_rate": 4.884974018551931e-05, "loss": 1.3909, "step": 1173 }, { "epoch": 0.11, "learning_rate": 4.884758975700045e-05, "loss": 1.3913, "step": 1174 }, { "epoch": 0.11, "learning_rate": 4.884543736765049e-05, "loss": 1.3873, "step": 1175 }, { "epoch": 0.11, "learning_rate": 4.88432830176464e-05, "loss": 1.3675, "step": 1176 }, { "epoch": 0.11, "learning_rate": 4.884112670716533e-05, "loss": 1.3771, "step": 1177 }, { "epoch": 0.11, "learning_rate": 4.883896843638456e-05, "loss": 1.4081, "step": 1178 }, { "epoch": 0.11, "learning_rate": 4.883680820548157e-05, "loss": 1.3558, "step": 1179 }, { "epoch": 0.11, "learning_rate": 4.8834646014633965e-05, "loss": 1.3937, "step": 1180 }, { "epoch": 0.11, "learning_rate": 4.883248186401954e-05, "loss": 1.3871, "step": 1181 }, { "epoch": 0.11, "learning_rate": 4.883031575381623e-05, "loss": 1.3905, "step": 1182 }, { "epoch": 0.11, "learning_rate": 4.8828147684202155e-05, "loss": 1.4037, "step": 1183 }, { "epoch": 0.11, "learning_rate": 4.882597765535556e-05, "loss": 1.3736, "step": 1184 }, { "epoch": 0.11, "learning_rate": 4.8823805667454893e-05, "loss": 1.3772, "step": 1185 }, { "epoch": 0.11, "learning_rate": 4.882163172067872e-05, "loss": 1.3641, "step": 1186 }, { "epoch": 0.11, "learning_rate": 4.8819455815205806e-05, "loss": 1.4027, "step": 1187 }, { "epoch": 0.11, "learning_rate": 4.881727795121506e-05, "loss": 1.3865, "step": 1188 }, { "epoch": 0.11, "learning_rate": 4.881509812888555e-05, "loss": 1.3425, "step": 1189 }, { "epoch": 0.11, "learning_rate": 4.8812916348396515e-05, "loss": 1.3766, "step": 1190 }, { "epoch": 0.11, "learning_rate": 4.881073260992734e-05, "loss": 1.3932, "step": 1191 }, { "epoch": 0.11, "learning_rate": 4.880854691365758e-05, "loss": 1.4014, "step": 1192 }, { "epoch": 0.11, "learning_rate": 4.880635925976696e-05, "loss": 1.349, "step": 1193 }, { "epoch": 0.11, "learning_rate": 4.880416964843535e-05, "loss": 1.3211, "step": 1194 }, { "epoch": 0.11, "learning_rate": 4.880197807984278e-05, "loss": 1.3962, "step": 1195 }, { "epoch": 0.11, "learning_rate": 4.879978455416946e-05, "loss": 1.3524, "step": 1196 }, { "epoch": 0.11, "learning_rate": 4.879758907159575e-05, "loss": 1.3396, "step": 1197 }, { "epoch": 0.11, "learning_rate": 4.879539163230216e-05, "loss": 1.4094, "step": 1198 }, { "epoch": 0.11, "learning_rate": 4.8793192236469377e-05, "loss": 1.445, "step": 1199 }, { "epoch": 0.11, "learning_rate": 4.879099088427824e-05, "loss": 1.3784, "step": 1200 }, { "epoch": 0.11, "learning_rate": 4.8788787575909756e-05, "loss": 1.3225, "step": 1201 }, { "epoch": 0.11, "learning_rate": 4.878658231154508e-05, "loss": 1.3828, "step": 1202 }, { "epoch": 0.11, "learning_rate": 4.878437509136555e-05, "loss": 1.395, "step": 1203 }, { "epoch": 0.11, "learning_rate": 4.878216591555265e-05, "loss": 1.3702, "step": 1204 }, { "epoch": 0.11, "learning_rate": 4.8779954784288004e-05, "loss": 1.3476, "step": 1205 }, { "epoch": 0.11, "learning_rate": 4.877774169775345e-05, "loss": 1.4038, "step": 1206 }, { "epoch": 0.11, "learning_rate": 4.877552665613093e-05, "loss": 1.3665, "step": 1207 }, { "epoch": 0.11, "learning_rate": 4.8773309659602595e-05, "loss": 1.3899, "step": 1208 }, { "epoch": 0.11, "learning_rate": 4.8771090708350707e-05, "loss": 1.4635, "step": 1209 }, { "epoch": 0.11, "learning_rate": 4.876886980255775e-05, "loss": 1.4673, "step": 1210 }, { "epoch": 0.11, "learning_rate": 4.87666469424063e-05, "loss": 1.3543, "step": 1211 }, { "epoch": 0.11, "learning_rate": 4.8764422128079144e-05, "loss": 1.3318, "step": 1212 }, { "epoch": 0.11, "learning_rate": 4.8762195359759225e-05, "loss": 1.4732, "step": 1213 }, { "epoch": 0.11, "learning_rate": 4.875996663762962e-05, "loss": 1.4335, "step": 1214 }, { "epoch": 0.11, "learning_rate": 4.875773596187358e-05, "loss": 1.3895, "step": 1215 }, { "epoch": 0.11, "learning_rate": 4.875550333267453e-05, "loss": 1.3866, "step": 1216 }, { "epoch": 0.11, "learning_rate": 4.875326875021604e-05, "loss": 1.3925, "step": 1217 }, { "epoch": 0.11, "learning_rate": 4.8751032214681844e-05, "loss": 1.4085, "step": 1218 }, { "epoch": 0.11, "learning_rate": 4.874879372625584e-05, "loss": 1.3375, "step": 1219 }, { "epoch": 0.11, "learning_rate": 4.8746553285122074e-05, "loss": 1.3616, "step": 1220 }, { "epoch": 0.11, "learning_rate": 4.8744310891464786e-05, "loss": 1.3578, "step": 1221 }, { "epoch": 0.11, "learning_rate": 4.874206654546833e-05, "loss": 1.4597, "step": 1222 }, { "epoch": 0.11, "learning_rate": 4.8739820247317254e-05, "loss": 1.3538, "step": 1223 }, { "epoch": 0.11, "learning_rate": 4.873757199719625e-05, "loss": 1.3939, "step": 1224 }, { "epoch": 0.11, "learning_rate": 4.873532179529019e-05, "loss": 1.3947, "step": 1225 }, { "epoch": 0.11, "learning_rate": 4.873306964178409e-05, "loss": 1.3801, "step": 1226 }, { "epoch": 0.11, "learning_rate": 4.8730815536863115e-05, "loss": 1.3157, "step": 1227 }, { "epoch": 0.11, "learning_rate": 4.872855948071262e-05, "loss": 1.3424, "step": 1228 }, { "epoch": 0.11, "learning_rate": 4.87263014735181e-05, "loss": 1.3936, "step": 1229 }, { "epoch": 0.11, "learning_rate": 4.8724041515465216e-05, "loss": 1.3922, "step": 1230 }, { "epoch": 0.11, "learning_rate": 4.872177960673979e-05, "loss": 1.3982, "step": 1231 }, { "epoch": 0.11, "learning_rate": 4.8719515747527814e-05, "loss": 1.4429, "step": 1232 }, { "epoch": 0.11, "learning_rate": 4.8717249938015415e-05, "loss": 1.4102, "step": 1233 }, { "epoch": 0.11, "learning_rate": 4.87149821783889e-05, "loss": 1.3941, "step": 1234 }, { "epoch": 0.11, "learning_rate": 4.871271246883474e-05, "loss": 1.3874, "step": 1235 }, { "epoch": 0.11, "learning_rate": 4.8710440809539545e-05, "loss": 1.3848, "step": 1236 }, { "epoch": 0.11, "learning_rate": 4.870816720069011e-05, "loss": 1.3649, "step": 1237 }, { "epoch": 0.11, "learning_rate": 4.870589164247337e-05, "loss": 1.3818, "step": 1238 }, { "epoch": 0.11, "learning_rate": 4.8703614135076436e-05, "loss": 1.386, "step": 1239 }, { "epoch": 0.11, "learning_rate": 4.870133467868658e-05, "loss": 1.3543, "step": 1240 }, { "epoch": 0.11, "learning_rate": 4.8699053273491205e-05, "loss": 1.3835, "step": 1241 }, { "epoch": 0.11, "learning_rate": 4.8696769919677914e-05, "loss": 1.4367, "step": 1242 }, { "epoch": 0.11, "learning_rate": 4.8694484617434444e-05, "loss": 1.369, "step": 1243 }, { "epoch": 0.11, "learning_rate": 4.869219736694871e-05, "loss": 1.4051, "step": 1244 }, { "epoch": 0.11, "learning_rate": 4.868990816840876e-05, "loss": 1.3957, "step": 1245 }, { "epoch": 0.11, "learning_rate": 4.868761702200284e-05, "loss": 1.4221, "step": 1246 }, { "epoch": 0.11, "learning_rate": 4.8685323927919324e-05, "loss": 1.3192, "step": 1247 }, { "epoch": 0.11, "learning_rate": 4.868302888634676e-05, "loss": 1.4049, "step": 1248 }, { "epoch": 0.11, "learning_rate": 4.8680731897473855e-05, "loss": 1.404, "step": 1249 }, { "epoch": 0.11, "learning_rate": 4.867843296148947e-05, "loss": 1.3604, "step": 1250 }, { "epoch": 0.11, "learning_rate": 4.867613207858265e-05, "loss": 1.3906, "step": 1251 }, { "epoch": 0.11, "learning_rate": 4.867382924894256e-05, "loss": 1.4189, "step": 1252 }, { "epoch": 0.11, "learning_rate": 4.8671524472758546e-05, "loss": 1.4391, "step": 1253 }, { "epoch": 0.11, "learning_rate": 4.866921775022013e-05, "loss": 1.3829, "step": 1254 }, { "epoch": 0.11, "learning_rate": 4.866690908151698e-05, "loss": 1.3995, "step": 1255 }, { "epoch": 0.11, "learning_rate": 4.8664598466838905e-05, "loss": 1.4131, "step": 1256 }, { "epoch": 0.11, "learning_rate": 4.86622859063759e-05, "loss": 1.3768, "step": 1257 }, { "epoch": 0.11, "learning_rate": 4.8659971400318124e-05, "loss": 1.3712, "step": 1258 }, { "epoch": 0.11, "learning_rate": 4.865765494885587e-05, "loss": 1.3069, "step": 1259 }, { "epoch": 0.11, "learning_rate": 4.86553365521796e-05, "loss": 1.4097, "step": 1260 }, { "epoch": 0.11, "learning_rate": 4.865301621047995e-05, "loss": 1.4007, "step": 1261 }, { "epoch": 0.11, "learning_rate": 4.865069392394771e-05, "loss": 1.4339, "step": 1262 }, { "epoch": 0.11, "learning_rate": 4.8648369692773816e-05, "loss": 1.3512, "step": 1263 }, { "epoch": 0.11, "learning_rate": 4.864604351714938e-05, "loss": 1.3582, "step": 1264 }, { "epoch": 0.11, "learning_rate": 4.864371539726567e-05, "loss": 1.3489, "step": 1265 }, { "epoch": 0.11, "learning_rate": 4.86413853333141e-05, "loss": 1.3293, "step": 1266 }, { "epoch": 0.11, "learning_rate": 4.8639053325486286e-05, "loss": 1.4035, "step": 1267 }, { "epoch": 0.11, "learning_rate": 4.863671937397394e-05, "loss": 1.3777, "step": 1268 }, { "epoch": 0.11, "learning_rate": 4.863438347896898e-05, "loss": 1.3221, "step": 1269 }, { "epoch": 0.11, "learning_rate": 4.863204564066347e-05, "loss": 1.3384, "step": 1270 }, { "epoch": 0.11, "learning_rate": 4.862970585924964e-05, "loss": 1.3445, "step": 1271 }, { "epoch": 0.11, "learning_rate": 4.862736413491987e-05, "loss": 1.3563, "step": 1272 }, { "epoch": 0.12, "learning_rate": 4.862502046786671e-05, "loss": 1.341, "step": 1273 }, { "epoch": 0.12, "learning_rate": 4.8622674858282866e-05, "loss": 1.3909, "step": 1274 }, { "epoch": 0.12, "learning_rate": 4.862032730636119e-05, "loss": 1.4342, "step": 1275 }, { "epoch": 0.12, "learning_rate": 4.8617977812294715e-05, "loss": 1.3585, "step": 1276 }, { "epoch": 0.12, "learning_rate": 4.8615626376276627e-05, "loss": 1.4378, "step": 1277 }, { "epoch": 0.12, "learning_rate": 4.861327299850026e-05, "loss": 1.3846, "step": 1278 }, { "epoch": 0.12, "learning_rate": 4.861091767915912e-05, "loss": 1.3695, "step": 1279 }, { "epoch": 0.12, "learning_rate": 4.860856041844687e-05, "loss": 1.3603, "step": 1280 }, { "epoch": 0.12, "learning_rate": 4.8606201216557334e-05, "loss": 1.3522, "step": 1281 }, { "epoch": 0.12, "learning_rate": 4.86038400736845e-05, "loss": 1.3775, "step": 1282 }, { "epoch": 0.12, "learning_rate": 4.8601476990022497e-05, "loss": 1.35, "step": 1283 }, { "epoch": 0.12, "learning_rate": 4.859911196576563e-05, "loss": 1.4071, "step": 1284 }, { "epoch": 0.12, "learning_rate": 4.859674500110836e-05, "loss": 1.3977, "step": 1285 }, { "epoch": 0.12, "learning_rate": 4.859437609624531e-05, "loss": 1.3867, "step": 1286 }, { "epoch": 0.12, "learning_rate": 4.8592005251371256e-05, "loss": 1.3557, "step": 1287 }, { "epoch": 0.12, "learning_rate": 4.858963246668113e-05, "loss": 1.3857, "step": 1288 }, { "epoch": 0.12, "learning_rate": 4.858725774237005e-05, "loss": 1.4043, "step": 1289 }, { "epoch": 0.12, "learning_rate": 4.8584881078633265e-05, "loss": 1.3327, "step": 1290 }, { "epoch": 0.12, "learning_rate": 4.858250247566618e-05, "loss": 1.3826, "step": 1291 }, { "epoch": 0.12, "learning_rate": 4.858012193366439e-05, "loss": 1.3643, "step": 1292 }, { "epoch": 0.12, "learning_rate": 4.8577739452823625e-05, "loss": 1.3758, "step": 1293 }, { "epoch": 0.12, "learning_rate": 4.857535503333978e-05, "loss": 1.4368, "step": 1294 }, { "epoch": 0.12, "learning_rate": 4.8572968675408896e-05, "loss": 1.4045, "step": 1295 }, { "epoch": 0.12, "learning_rate": 4.857058037922721e-05, "loss": 1.386, "step": 1296 }, { "epoch": 0.12, "learning_rate": 4.856819014499109e-05, "loss": 1.4094, "step": 1297 }, { "epoch": 0.12, "learning_rate": 4.8565797972897056e-05, "loss": 1.3825, "step": 1298 }, { "epoch": 0.12, "learning_rate": 4.856340386314182e-05, "loss": 1.3567, "step": 1299 }, { "epoch": 0.12, "learning_rate": 4.8561007815922224e-05, "loss": 1.4093, "step": 1300 }, { "epoch": 0.12, "learning_rate": 4.8558609831435273e-05, "loss": 1.3955, "step": 1301 }, { "epoch": 0.12, "learning_rate": 4.8556209909878156e-05, "loss": 1.3956, "step": 1302 }, { "epoch": 0.12, "learning_rate": 4.855380805144818e-05, "loss": 1.3935, "step": 1303 }, { "epoch": 0.12, "learning_rate": 4.855140425634286e-05, "loss": 1.3469, "step": 1304 }, { "epoch": 0.12, "learning_rate": 4.8548998524759815e-05, "loss": 1.3805, "step": 1305 }, { "epoch": 0.12, "learning_rate": 4.8546590856896876e-05, "loss": 1.3369, "step": 1306 }, { "epoch": 0.12, "learning_rate": 4.8544181252952e-05, "loss": 1.404, "step": 1307 }, { "epoch": 0.12, "learning_rate": 4.854176971312332e-05, "loss": 1.3681, "step": 1308 }, { "epoch": 0.12, "learning_rate": 4.8539356237609104e-05, "loss": 1.3669, "step": 1309 }, { "epoch": 0.12, "learning_rate": 4.8536940826607814e-05, "loss": 1.3908, "step": 1310 }, { "epoch": 0.12, "learning_rate": 4.8534523480318046e-05, "loss": 1.3773, "step": 1311 }, { "epoch": 0.12, "learning_rate": 4.853210419893857e-05, "loss": 1.3809, "step": 1312 }, { "epoch": 0.12, "learning_rate": 4.852968298266829e-05, "loss": 1.3703, "step": 1313 }, { "epoch": 0.12, "learning_rate": 4.85272598317063e-05, "loss": 1.3557, "step": 1314 }, { "epoch": 0.12, "learning_rate": 4.852483474625185e-05, "loss": 1.3638, "step": 1315 }, { "epoch": 0.12, "learning_rate": 4.8522407726504307e-05, "loss": 1.3544, "step": 1316 }, { "epoch": 0.12, "learning_rate": 4.851997877266327e-05, "loss": 1.3618, "step": 1317 }, { "epoch": 0.12, "learning_rate": 4.8517547884928415e-05, "loss": 1.3886, "step": 1318 }, { "epoch": 0.12, "learning_rate": 4.851511506349965e-05, "loss": 1.4485, "step": 1319 }, { "epoch": 0.12, "learning_rate": 4.8512680308576995e-05, "loss": 1.3521, "step": 1320 }, { "epoch": 0.12, "learning_rate": 4.851024362036064e-05, "loss": 1.3581, "step": 1321 }, { "epoch": 0.12, "learning_rate": 4.8507804999050954e-05, "loss": 1.3719, "step": 1322 }, { "epoch": 0.12, "learning_rate": 4.8505364444848436e-05, "loss": 1.3266, "step": 1323 }, { "epoch": 0.12, "learning_rate": 4.8502921957953765e-05, "loss": 1.4113, "step": 1324 }, { "epoch": 0.12, "learning_rate": 4.850047753856776e-05, "loss": 1.3764, "step": 1325 }, { "epoch": 0.12, "learning_rate": 4.849803118689141e-05, "loss": 1.385, "step": 1326 }, { "epoch": 0.12, "learning_rate": 4.849558290312588e-05, "loss": 1.3777, "step": 1327 }, { "epoch": 0.12, "learning_rate": 4.849313268747245e-05, "loss": 1.3597, "step": 1328 }, { "epoch": 0.12, "learning_rate": 4.8490680540132614e-05, "loss": 1.3732, "step": 1329 }, { "epoch": 0.12, "learning_rate": 4.848822646130797e-05, "loss": 1.4093, "step": 1330 }, { "epoch": 0.12, "learning_rate": 4.8485770451200316e-05, "loss": 1.4037, "step": 1331 }, { "epoch": 0.12, "learning_rate": 4.84833125100116e-05, "loss": 1.3083, "step": 1332 }, { "epoch": 0.12, "learning_rate": 4.84808526379439e-05, "loss": 1.4034, "step": 1333 }, { "epoch": 0.12, "learning_rate": 4.8478390835199496e-05, "loss": 1.4048, "step": 1334 }, { "epoch": 0.12, "learning_rate": 4.847592710198079e-05, "loss": 1.4221, "step": 1335 }, { "epoch": 0.12, "learning_rate": 4.8473461438490365e-05, "loss": 1.3329, "step": 1336 }, { "epoch": 0.12, "learning_rate": 4.847099384493097e-05, "loss": 1.3967, "step": 1337 }, { "epoch": 0.12, "learning_rate": 4.846852432150547e-05, "loss": 1.4114, "step": 1338 }, { "epoch": 0.12, "learning_rate": 4.846605286841695e-05, "loss": 1.3639, "step": 1339 }, { "epoch": 0.12, "learning_rate": 4.846357948586859e-05, "loss": 1.3885, "step": 1340 }, { "epoch": 0.12, "learning_rate": 4.846110417406378e-05, "loss": 1.3816, "step": 1341 }, { "epoch": 0.12, "learning_rate": 4.8458626933206056e-05, "loss": 1.3826, "step": 1342 }, { "epoch": 0.12, "learning_rate": 4.845614776349908e-05, "loss": 1.3544, "step": 1343 }, { "epoch": 0.12, "learning_rate": 4.845366666514671e-05, "loss": 1.3483, "step": 1344 }, { "epoch": 0.12, "learning_rate": 4.845118363835296e-05, "loss": 1.4223, "step": 1345 }, { "epoch": 0.12, "learning_rate": 4.844869868332199e-05, "loss": 1.3845, "step": 1346 }, { "epoch": 0.12, "learning_rate": 4.844621180025811e-05, "loss": 1.3528, "step": 1347 }, { "epoch": 0.12, "learning_rate": 4.84437229893658e-05, "loss": 1.3841, "step": 1348 }, { "epoch": 0.12, "learning_rate": 4.844123225084971e-05, "loss": 1.3699, "step": 1349 }, { "epoch": 0.12, "learning_rate": 4.843873958491464e-05, "loss": 1.3575, "step": 1350 }, { "epoch": 0.12, "learning_rate": 4.843624499176553e-05, "loss": 1.3577, "step": 1351 }, { "epoch": 0.12, "learning_rate": 4.84337484716075e-05, "loss": 1.2898, "step": 1352 }, { "epoch": 0.12, "learning_rate": 4.8431250024645835e-05, "loss": 1.3564, "step": 1353 }, { "epoch": 0.12, "learning_rate": 4.842874965108595e-05, "loss": 1.3242, "step": 1354 }, { "epoch": 0.12, "learning_rate": 4.842624735113344e-05, "loss": 1.3715, "step": 1355 }, { "epoch": 0.12, "learning_rate": 4.842374312499405e-05, "loss": 1.3448, "step": 1356 }, { "epoch": 0.12, "learning_rate": 4.8421236972873686e-05, "loss": 1.3542, "step": 1357 }, { "epoch": 0.12, "learning_rate": 4.8418728894978415e-05, "loss": 1.3522, "step": 1358 }, { "epoch": 0.12, "learning_rate": 4.8416218891514464e-05, "loss": 1.3373, "step": 1359 }, { "epoch": 0.12, "learning_rate": 4.841370696268821e-05, "loss": 1.4182, "step": 1360 }, { "epoch": 0.12, "learning_rate": 4.84111931087062e-05, "loss": 1.3622, "step": 1361 }, { "epoch": 0.12, "learning_rate": 4.840867732977511e-05, "loss": 1.3894, "step": 1362 }, { "epoch": 0.12, "learning_rate": 4.840615962610182e-05, "loss": 1.3387, "step": 1363 }, { "epoch": 0.12, "learning_rate": 4.8403639997893335e-05, "loss": 1.3865, "step": 1364 }, { "epoch": 0.12, "learning_rate": 4.840111844535682e-05, "loss": 1.3556, "step": 1365 }, { "epoch": 0.12, "learning_rate": 4.839859496869962e-05, "loss": 1.3516, "step": 1366 }, { "epoch": 0.12, "learning_rate": 4.839606956812921e-05, "loss": 1.3816, "step": 1367 }, { "epoch": 0.12, "learning_rate": 4.839354224385325e-05, "loss": 1.3775, "step": 1368 }, { "epoch": 0.12, "learning_rate": 4.839101299607954e-05, "loss": 1.453, "step": 1369 }, { "epoch": 0.12, "learning_rate": 4.8388481825016034e-05, "loss": 1.3632, "step": 1370 }, { "epoch": 0.12, "learning_rate": 4.838594873087087e-05, "loss": 1.3925, "step": 1371 }, { "epoch": 0.12, "learning_rate": 4.838341371385232e-05, "loss": 1.3996, "step": 1372 }, { "epoch": 0.12, "learning_rate": 4.838087677416881e-05, "loss": 1.3941, "step": 1373 }, { "epoch": 0.12, "learning_rate": 4.837833791202897e-05, "loss": 1.3519, "step": 1374 }, { "epoch": 0.12, "learning_rate": 4.837579712764152e-05, "loss": 1.3648, "step": 1375 }, { "epoch": 0.12, "learning_rate": 4.837325442121538e-05, "loss": 1.38, "step": 1376 }, { "epoch": 0.12, "learning_rate": 4.837070979295963e-05, "loss": 1.4414, "step": 1377 }, { "epoch": 0.12, "learning_rate": 4.836816324308349e-05, "loss": 1.3501, "step": 1378 }, { "epoch": 0.12, "learning_rate": 4.8365614771796345e-05, "loss": 1.358, "step": 1379 }, { "epoch": 0.12, "learning_rate": 4.836306437930775e-05, "loss": 1.3622, "step": 1380 }, { "epoch": 0.12, "learning_rate": 4.836051206582739e-05, "loss": 1.3127, "step": 1381 }, { "epoch": 0.12, "learning_rate": 4.8357957831565146e-05, "loss": 1.4018, "step": 1382 }, { "epoch": 0.12, "learning_rate": 4.8355401676731014e-05, "loss": 1.37, "step": 1383 }, { "epoch": 0.13, "learning_rate": 4.8352843601535186e-05, "loss": 1.3164, "step": 1384 }, { "epoch": 0.13, "learning_rate": 4.835028360618798e-05, "loss": 1.3929, "step": 1385 }, { "epoch": 0.13, "learning_rate": 4.8347721690899915e-05, "loss": 1.3747, "step": 1386 }, { "epoch": 0.13, "learning_rate": 4.834515785588161e-05, "loss": 1.3763, "step": 1387 }, { "epoch": 0.13, "learning_rate": 4.83425921013439e-05, "loss": 1.3953, "step": 1388 }, { "epoch": 0.13, "learning_rate": 4.834002442749772e-05, "loss": 1.3529, "step": 1389 }, { "epoch": 0.13, "learning_rate": 4.833745483455421e-05, "loss": 1.3154, "step": 1390 }, { "epoch": 0.13, "learning_rate": 4.833488332272465e-05, "loss": 1.3793, "step": 1391 }, { "epoch": 0.13, "learning_rate": 4.833230989222049e-05, "loss": 1.3635, "step": 1392 }, { "epoch": 0.13, "learning_rate": 4.832973454325331e-05, "loss": 1.3686, "step": 1393 }, { "epoch": 0.13, "learning_rate": 4.832715727603486e-05, "loss": 1.3653, "step": 1394 }, { "epoch": 0.13, "learning_rate": 4.832457809077707e-05, "loss": 1.3671, "step": 1395 }, { "epoch": 0.13, "learning_rate": 4.8321996987691996e-05, "loss": 1.3552, "step": 1396 }, { "epoch": 0.13, "learning_rate": 4.8319413966991867e-05, "loss": 1.3782, "step": 1397 }, { "epoch": 0.13, "learning_rate": 4.831682902888907e-05, "loss": 1.3455, "step": 1398 }, { "epoch": 0.13, "learning_rate": 4.831424217359615e-05, "loss": 1.376, "step": 1399 }, { "epoch": 0.13, "learning_rate": 4.831165340132581e-05, "loss": 1.4107, "step": 1400 }, { "epoch": 0.13, "learning_rate": 4.8309062712290894e-05, "loss": 1.3856, "step": 1401 }, { "epoch": 0.13, "learning_rate": 4.830647010670443e-05, "loss": 1.3605, "step": 1402 }, { "epoch": 0.13, "learning_rate": 4.830387558477958e-05, "loss": 1.3576, "step": 1403 }, { "epoch": 0.13, "learning_rate": 4.830127914672968e-05, "loss": 1.2961, "step": 1404 }, { "epoch": 0.13, "learning_rate": 4.829868079276823e-05, "loss": 1.4078, "step": 1405 }, { "epoch": 0.13, "learning_rate": 4.8296080523108865e-05, "loss": 1.337, "step": 1406 }, { "epoch": 0.13, "learning_rate": 4.829347833796538e-05, "loss": 1.3962, "step": 1407 }, { "epoch": 0.13, "learning_rate": 4.8290874237551756e-05, "loss": 1.331, "step": 1408 }, { "epoch": 0.13, "learning_rate": 4.828826822208209e-05, "loss": 1.4144, "step": 1409 }, { "epoch": 0.13, "learning_rate": 4.8285660291770665e-05, "loss": 1.3906, "step": 1410 }, { "epoch": 0.13, "learning_rate": 4.828305044683191e-05, "loss": 1.4184, "step": 1411 }, { "epoch": 0.13, "learning_rate": 4.828043868748043e-05, "loss": 1.4085, "step": 1412 }, { "epoch": 0.13, "learning_rate": 4.827782501393096e-05, "loss": 1.4021, "step": 1413 }, { "epoch": 0.13, "learning_rate": 4.827520942639842e-05, "loss": 1.391, "step": 1414 }, { "epoch": 0.13, "learning_rate": 4.827259192509785e-05, "loss": 1.3827, "step": 1415 }, { "epoch": 0.13, "learning_rate": 4.826997251024449e-05, "loss": 1.3571, "step": 1416 }, { "epoch": 0.13, "learning_rate": 4.8267351182053703e-05, "loss": 1.343, "step": 1417 }, { "epoch": 0.13, "learning_rate": 4.8264727940741035e-05, "loss": 1.3439, "step": 1418 }, { "epoch": 0.13, "learning_rate": 4.8262102786522166e-05, "loss": 1.3809, "step": 1419 }, { "epoch": 0.13, "learning_rate": 4.8259475719612954e-05, "loss": 1.342, "step": 1420 }, { "epoch": 0.13, "learning_rate": 4.8256846740229405e-05, "loss": 1.4093, "step": 1421 }, { "epoch": 0.13, "learning_rate": 4.8254215848587684e-05, "loss": 1.3821, "step": 1422 }, { "epoch": 0.13, "learning_rate": 4.82515830449041e-05, "loss": 1.3411, "step": 1423 }, { "epoch": 0.13, "learning_rate": 4.824894832939515e-05, "loss": 1.3298, "step": 1424 }, { "epoch": 0.13, "learning_rate": 4.824631170227746e-05, "loss": 1.3627, "step": 1425 }, { "epoch": 0.13, "learning_rate": 4.824367316376781e-05, "loss": 1.3795, "step": 1426 }, { "epoch": 0.13, "learning_rate": 4.824103271408318e-05, "loss": 1.4234, "step": 1427 }, { "epoch": 0.13, "learning_rate": 4.8238390353440647e-05, "loss": 1.3179, "step": 1428 }, { "epoch": 0.13, "learning_rate": 4.8235746082057484e-05, "loss": 1.3875, "step": 1429 }, { "epoch": 0.13, "learning_rate": 4.823309990015112e-05, "loss": 1.3538, "step": 1430 }, { "epoch": 0.13, "learning_rate": 4.8230451807939135e-05, "loss": 1.3995, "step": 1431 }, { "epoch": 0.13, "learning_rate": 4.822780180563925e-05, "loss": 1.3314, "step": 1432 }, { "epoch": 0.13, "learning_rate": 4.822514989346937e-05, "loss": 1.4288, "step": 1433 }, { "epoch": 0.13, "learning_rate": 4.822249607164753e-05, "loss": 1.3101, "step": 1434 }, { "epoch": 0.13, "learning_rate": 4.821984034039195e-05, "loss": 1.3832, "step": 1435 }, { "epoch": 0.13, "learning_rate": 4.821718269992099e-05, "loss": 1.4141, "step": 1436 }, { "epoch": 0.13, "learning_rate": 4.821452315045317e-05, "loss": 1.3786, "step": 1437 }, { "epoch": 0.13, "learning_rate": 4.8211861692207174e-05, "loss": 1.3401, "step": 1438 }, { "epoch": 0.13, "learning_rate": 4.8209198325401815e-05, "loss": 1.3915, "step": 1439 }, { "epoch": 0.13, "learning_rate": 4.820653305025611e-05, "loss": 1.3608, "step": 1440 }, { "epoch": 0.13, "learning_rate": 4.8203865866989184e-05, "loss": 1.3514, "step": 1441 }, { "epoch": 0.13, "learning_rate": 4.8201196775820364e-05, "loss": 1.3305, "step": 1442 }, { "epoch": 0.13, "learning_rate": 4.81985257769691e-05, "loss": 1.3745, "step": 1443 }, { "epoch": 0.13, "learning_rate": 4.819585287065501e-05, "loss": 1.3672, "step": 1444 }, { "epoch": 0.13, "learning_rate": 4.8193178057097865e-05, "loss": 1.4145, "step": 1445 }, { "epoch": 0.13, "learning_rate": 4.819050133651761e-05, "loss": 1.4103, "step": 1446 }, { "epoch": 0.13, "learning_rate": 4.818782270913433e-05, "loss": 1.4122, "step": 1447 }, { "epoch": 0.13, "learning_rate": 4.8185142175168266e-05, "loss": 1.342, "step": 1448 }, { "epoch": 0.13, "learning_rate": 4.8182459734839824e-05, "loss": 1.4425, "step": 1449 }, { "epoch": 0.13, "learning_rate": 4.817977538836956e-05, "loss": 1.4249, "step": 1450 }, { "epoch": 0.13, "learning_rate": 4.8177089135978204e-05, "loss": 1.3588, "step": 1451 }, { "epoch": 0.13, "learning_rate": 4.817440097788661e-05, "loss": 1.4342, "step": 1452 }, { "epoch": 0.13, "learning_rate": 4.8171710914315816e-05, "loss": 1.4285, "step": 1453 }, { "epoch": 0.13, "learning_rate": 4.816901894548701e-05, "loss": 1.3494, "step": 1454 }, { "epoch": 0.13, "learning_rate": 4.8166325071621543e-05, "loss": 1.3372, "step": 1455 }, { "epoch": 0.13, "learning_rate": 4.816362929294089e-05, "loss": 1.3597, "step": 1456 }, { "epoch": 0.13, "learning_rate": 4.816093160966673e-05, "loss": 1.3713, "step": 1457 }, { "epoch": 0.13, "learning_rate": 4.815823202202087e-05, "loss": 1.3756, "step": 1458 }, { "epoch": 0.13, "learning_rate": 4.8155530530225276e-05, "loss": 1.4171, "step": 1459 }, { "epoch": 0.13, "learning_rate": 4.8152827134502074e-05, "loss": 1.3615, "step": 1460 }, { "epoch": 0.13, "learning_rate": 4.815012183507355e-05, "loss": 1.3777, "step": 1461 }, { "epoch": 0.13, "learning_rate": 4.814741463216213e-05, "loss": 1.367, "step": 1462 }, { "epoch": 0.13, "learning_rate": 4.814470552599044e-05, "loss": 1.3476, "step": 1463 }, { "epoch": 0.13, "learning_rate": 4.8141994516781196e-05, "loss": 1.4005, "step": 1464 }, { "epoch": 0.13, "learning_rate": 4.813928160475733e-05, "loss": 1.3439, "step": 1465 }, { "epoch": 0.13, "learning_rate": 4.81365667901419e-05, "loss": 1.3977, "step": 1466 }, { "epoch": 0.13, "learning_rate": 4.813385007315813e-05, "loss": 1.3557, "step": 1467 }, { "epoch": 0.13, "learning_rate": 4.8131131454029394e-05, "loss": 1.3311, "step": 1468 }, { "epoch": 0.13, "learning_rate": 4.812841093297923e-05, "loss": 1.3487, "step": 1469 }, { "epoch": 0.13, "learning_rate": 4.812568851023132e-05, "loss": 1.345, "step": 1470 }, { "epoch": 0.13, "learning_rate": 4.812296418600952e-05, "loss": 1.3474, "step": 1471 }, { "epoch": 0.13, "learning_rate": 4.812023796053783e-05, "loss": 1.3583, "step": 1472 }, { "epoch": 0.13, "learning_rate": 4.8117509834040417e-05, "loss": 1.3198, "step": 1473 }, { "epoch": 0.13, "learning_rate": 4.811477980674158e-05, "loss": 1.4126, "step": 1474 }, { "epoch": 0.13, "learning_rate": 4.8112047878865805e-05, "loss": 1.326, "step": 1475 }, { "epoch": 0.13, "learning_rate": 4.810931405063772e-05, "loss": 1.3886, "step": 1476 }, { "epoch": 0.13, "learning_rate": 4.810657832228211e-05, "loss": 1.4329, "step": 1477 }, { "epoch": 0.13, "learning_rate": 4.810384069402391e-05, "loss": 1.3266, "step": 1478 }, { "epoch": 0.13, "learning_rate": 4.8101101166088215e-05, "loss": 1.4008, "step": 1479 }, { "epoch": 0.13, "learning_rate": 4.809835973870029e-05, "loss": 1.3298, "step": 1480 }, { "epoch": 0.13, "learning_rate": 4.809561641208553e-05, "loss": 1.357, "step": 1481 }, { "epoch": 0.13, "learning_rate": 4.809287118646952e-05, "loss": 1.3558, "step": 1482 }, { "epoch": 0.13, "learning_rate": 4.809012406207797e-05, "loss": 1.3623, "step": 1483 }, { "epoch": 0.13, "learning_rate": 4.808737503913675e-05, "loss": 1.345, "step": 1484 }, { "epoch": 0.13, "learning_rate": 4.808462411787191e-05, "loss": 1.3739, "step": 1485 }, { "epoch": 0.13, "learning_rate": 4.808187129850963e-05, "loss": 1.4101, "step": 1486 }, { "epoch": 0.13, "learning_rate": 4.807911658127626e-05, "loss": 1.4016, "step": 1487 }, { "epoch": 0.13, "learning_rate": 4.80763599663983e-05, "loss": 1.4193, "step": 1488 }, { "epoch": 0.13, "learning_rate": 4.8073601454102416e-05, "loss": 1.3435, "step": 1489 }, { "epoch": 0.13, "learning_rate": 4.8070841044615413e-05, "loss": 1.3618, "step": 1490 }, { "epoch": 0.13, "learning_rate": 4.806807873816427e-05, "loss": 1.3169, "step": 1491 }, { "epoch": 0.13, "learning_rate": 4.80653145349761e-05, "loss": 1.3821, "step": 1492 }, { "epoch": 0.13, "learning_rate": 4.8062548435278204e-05, "loss": 1.3501, "step": 1493 }, { "epoch": 0.13, "learning_rate": 4.8059780439298e-05, "loss": 1.3445, "step": 1494 }, { "epoch": 0.14, "learning_rate": 4.8057010547263106e-05, "loss": 1.3403, "step": 1495 }, { "epoch": 0.14, "learning_rate": 4.805423875940125e-05, "loss": 1.3217, "step": 1496 }, { "epoch": 0.14, "learning_rate": 4.8051465075940336e-05, "loss": 1.3997, "step": 1497 }, { "epoch": 0.14, "learning_rate": 4.8048689497108454e-05, "loss": 1.4003, "step": 1498 }, { "epoch": 0.14, "learning_rate": 4.804591202313379e-05, "loss": 1.3186, "step": 1499 }, { "epoch": 0.14, "learning_rate": 4.804313265424474e-05, "loss": 1.3933, "step": 1500 }, { "epoch": 0.14, "learning_rate": 4.8040351390669824e-05, "loss": 1.4015, "step": 1501 }, { "epoch": 0.14, "learning_rate": 4.803756823263772e-05, "loss": 1.3098, "step": 1502 }, { "epoch": 0.14, "learning_rate": 4.803478318037728e-05, "loss": 1.372, "step": 1503 }, { "epoch": 0.14, "learning_rate": 4.803199623411751e-05, "loss": 1.3089, "step": 1504 }, { "epoch": 0.14, "learning_rate": 4.802920739408754e-05, "loss": 1.359, "step": 1505 }, { "epoch": 0.14, "learning_rate": 4.802641666051668e-05, "loss": 1.4152, "step": 1506 }, { "epoch": 0.14, "learning_rate": 4.8023624033634416e-05, "loss": 1.3559, "step": 1507 }, { "epoch": 0.14, "learning_rate": 4.802082951367034e-05, "loss": 1.3757, "step": 1508 }, { "epoch": 0.14, "learning_rate": 4.8018033100854245e-05, "loss": 1.336, "step": 1509 }, { "epoch": 0.14, "learning_rate": 4.801523479541606e-05, "loss": 1.3348, "step": 1510 }, { "epoch": 0.14, "learning_rate": 4.8012434597585865e-05, "loss": 1.347, "step": 1511 }, { "epoch": 0.14, "learning_rate": 4.80096325075939e-05, "loss": 1.3825, "step": 1512 }, { "epoch": 0.14, "learning_rate": 4.800682852567057e-05, "loss": 1.3468, "step": 1513 }, { "epoch": 0.14, "learning_rate": 4.8004022652046434e-05, "loss": 1.3305, "step": 1514 }, { "epoch": 0.14, "learning_rate": 4.800121488695218e-05, "loss": 1.3455, "step": 1515 }, { "epoch": 0.14, "learning_rate": 4.7998405230618684e-05, "loss": 1.3688, "step": 1516 }, { "epoch": 0.14, "learning_rate": 4.7995593683276975e-05, "loss": 1.2974, "step": 1517 }, { "epoch": 0.14, "learning_rate": 4.7992780245158214e-05, "loss": 1.3108, "step": 1518 }, { "epoch": 0.14, "learning_rate": 4.7989964916493736e-05, "loss": 1.397, "step": 1519 }, { "epoch": 0.14, "learning_rate": 4.7987147697515025e-05, "loss": 1.3564, "step": 1520 }, { "epoch": 0.14, "learning_rate": 4.798432858845373e-05, "loss": 1.3738, "step": 1521 }, { "epoch": 0.14, "learning_rate": 4.798150758954164e-05, "loss": 1.4045, "step": 1522 }, { "epoch": 0.14, "learning_rate": 4.797868470101071e-05, "loss": 1.4026, "step": 1523 }, { "epoch": 0.14, "learning_rate": 4.7975859923093054e-05, "loss": 1.3236, "step": 1524 }, { "epoch": 0.14, "learning_rate": 4.797303325602092e-05, "loss": 1.3438, "step": 1525 }, { "epoch": 0.14, "learning_rate": 4.797020470002674e-05, "loss": 1.3991, "step": 1526 }, { "epoch": 0.14, "learning_rate": 4.7967374255343076e-05, "loss": 1.3985, "step": 1527 }, { "epoch": 0.14, "learning_rate": 4.796454192220268e-05, "loss": 1.3334, "step": 1528 }, { "epoch": 0.14, "learning_rate": 4.79617077008384e-05, "loss": 1.3181, "step": 1529 }, { "epoch": 0.14, "learning_rate": 4.7958871591483303e-05, "loss": 1.2579, "step": 1530 }, { "epoch": 0.14, "learning_rate": 4.795603359437058e-05, "loss": 1.3996, "step": 1531 }, { "epoch": 0.14, "learning_rate": 4.7953193709733576e-05, "loss": 1.3676, "step": 1532 }, { "epoch": 0.14, "learning_rate": 4.7950351937805796e-05, "loss": 1.3282, "step": 1533 }, { "epoch": 0.14, "learning_rate": 4.7947508278820895e-05, "loss": 1.3291, "step": 1534 }, { "epoch": 0.14, "learning_rate": 4.79446627330127e-05, "loss": 1.4003, "step": 1535 }, { "epoch": 0.14, "learning_rate": 4.794181530061518e-05, "loss": 1.2957, "step": 1536 }, { "epoch": 0.14, "learning_rate": 4.7938965981862446e-05, "loss": 1.3857, "step": 1537 }, { "epoch": 0.14, "learning_rate": 4.79361147769888e-05, "loss": 1.3861, "step": 1538 }, { "epoch": 0.14, "learning_rate": 4.7933261686228666e-05, "loss": 1.3479, "step": 1539 }, { "epoch": 0.14, "learning_rate": 4.793040670981663e-05, "loss": 1.3371, "step": 1540 }, { "epoch": 0.14, "learning_rate": 4.792754984798745e-05, "loss": 1.3893, "step": 1541 }, { "epoch": 0.14, "learning_rate": 4.7924691100976025e-05, "loss": 1.3401, "step": 1542 }, { "epoch": 0.14, "learning_rate": 4.7921830469017404e-05, "loss": 1.3802, "step": 1543 }, { "epoch": 0.14, "learning_rate": 4.79189679523468e-05, "loss": 1.3765, "step": 1544 }, { "epoch": 0.14, "learning_rate": 4.7916103551199584e-05, "loss": 1.3881, "step": 1545 }, { "epoch": 0.14, "learning_rate": 4.7913237265811274e-05, "loss": 1.3742, "step": 1546 }, { "epoch": 0.14, "learning_rate": 4.791036909641754e-05, "loss": 1.3446, "step": 1547 }, { "epoch": 0.14, "learning_rate": 4.7907499043254224e-05, "loss": 1.3582, "step": 1548 }, { "epoch": 0.14, "learning_rate": 4.790462710655731e-05, "loss": 1.3809, "step": 1549 }, { "epoch": 0.14, "learning_rate": 4.790175328656292e-05, "loss": 1.3508, "step": 1550 }, { "epoch": 0.14, "learning_rate": 4.789887758350738e-05, "loss": 1.3742, "step": 1551 }, { "epoch": 0.14, "learning_rate": 4.789599999762712e-05, "loss": 1.3556, "step": 1552 }, { "epoch": 0.14, "learning_rate": 4.789312052915874e-05, "loss": 1.3268, "step": 1553 }, { "epoch": 0.14, "learning_rate": 4.789023917833902e-05, "loss": 1.3878, "step": 1554 }, { "epoch": 0.14, "learning_rate": 4.7887355945404855e-05, "loss": 1.3349, "step": 1555 }, { "epoch": 0.14, "learning_rate": 4.788447083059333e-05, "loss": 1.3216, "step": 1556 }, { "epoch": 0.14, "learning_rate": 4.788158383414166e-05, "loss": 1.3679, "step": 1557 }, { "epoch": 0.14, "learning_rate": 4.787869495628723e-05, "loss": 1.2985, "step": 1558 }, { "epoch": 0.14, "learning_rate": 4.787580419726756e-05, "loss": 1.294, "step": 1559 }, { "epoch": 0.14, "learning_rate": 4.7872911557320354e-05, "loss": 1.3715, "step": 1560 }, { "epoch": 0.14, "learning_rate": 4.787001703668345e-05, "loss": 1.3259, "step": 1561 }, { "epoch": 0.14, "learning_rate": 4.7867120635594844e-05, "loss": 1.4197, "step": 1562 }, { "epoch": 0.14, "learning_rate": 4.786422235429269e-05, "loss": 1.3815, "step": 1563 }, { "epoch": 0.14, "learning_rate": 4.786132219301528e-05, "loss": 1.3436, "step": 1564 }, { "epoch": 0.14, "learning_rate": 4.7858420152001096e-05, "loss": 1.4128, "step": 1565 }, { "epoch": 0.14, "learning_rate": 4.785551623148875e-05, "loss": 1.375, "step": 1566 }, { "epoch": 0.14, "learning_rate": 4.7852610431717e-05, "loss": 1.3475, "step": 1567 }, { "epoch": 0.14, "learning_rate": 4.7849702752924784e-05, "loss": 1.2898, "step": 1568 }, { "epoch": 0.14, "learning_rate": 4.784679319535118e-05, "loss": 1.3422, "step": 1569 }, { "epoch": 0.14, "learning_rate": 4.784388175923541e-05, "loss": 1.373, "step": 1570 }, { "epoch": 0.14, "learning_rate": 4.784096844481688e-05, "loss": 1.3265, "step": 1571 }, { "epoch": 0.14, "learning_rate": 4.7838053252335116e-05, "loss": 1.324, "step": 1572 }, { "epoch": 0.14, "learning_rate": 4.7835136182029824e-05, "loss": 1.3531, "step": 1573 }, { "epoch": 0.14, "learning_rate": 4.783221723414086e-05, "loss": 1.3598, "step": 1574 }, { "epoch": 0.14, "learning_rate": 4.782929640890822e-05, "loss": 1.3997, "step": 1575 }, { "epoch": 0.14, "learning_rate": 4.782637370657206e-05, "loss": 1.3564, "step": 1576 }, { "epoch": 0.14, "learning_rate": 4.7823449127372716e-05, "loss": 1.3559, "step": 1577 }, { "epoch": 0.14, "learning_rate": 4.782052267155064e-05, "loss": 1.3939, "step": 1578 }, { "epoch": 0.14, "learning_rate": 4.781759433934646e-05, "loss": 1.3758, "step": 1579 }, { "epoch": 0.14, "learning_rate": 4.781466413100095e-05, "loss": 1.3682, "step": 1580 }, { "epoch": 0.14, "learning_rate": 4.781173204675505e-05, "loss": 1.3476, "step": 1581 }, { "epoch": 0.14, "learning_rate": 4.7808798086849835e-05, "loss": 1.3886, "step": 1582 }, { "epoch": 0.14, "learning_rate": 4.7805862251526554e-05, "loss": 1.3642, "step": 1583 }, { "epoch": 0.14, "learning_rate": 4.7802924541026596e-05, "loss": 1.3052, "step": 1584 }, { "epoch": 0.14, "learning_rate": 4.7799984955591515e-05, "loss": 1.3584, "step": 1585 }, { "epoch": 0.14, "learning_rate": 4.779704349546301e-05, "loss": 1.335, "step": 1586 }, { "epoch": 0.14, "learning_rate": 4.779410016088294e-05, "loss": 1.3619, "step": 1587 }, { "epoch": 0.14, "learning_rate": 4.779115495209332e-05, "loss": 1.4738, "step": 1588 }, { "epoch": 0.14, "learning_rate": 4.7788207869336304e-05, "loss": 1.467, "step": 1589 }, { "epoch": 0.14, "learning_rate": 4.778525891285423e-05, "loss": 1.3868, "step": 1590 }, { "epoch": 0.14, "learning_rate": 4.778230808288954e-05, "loss": 1.3592, "step": 1591 }, { "epoch": 0.14, "learning_rate": 4.777935537968489e-05, "loss": 1.3242, "step": 1592 }, { "epoch": 0.14, "learning_rate": 4.777640080348306e-05, "loss": 1.3416, "step": 1593 }, { "epoch": 0.14, "learning_rate": 4.777344435452696e-05, "loss": 1.2884, "step": 1594 }, { "epoch": 0.14, "learning_rate": 4.777048603305971e-05, "loss": 1.3464, "step": 1595 }, { "epoch": 0.14, "learning_rate": 4.776752583932454e-05, "loss": 1.3584, "step": 1596 }, { "epoch": 0.14, "learning_rate": 4.7764563773564845e-05, "loss": 1.3661, "step": 1597 }, { "epoch": 0.14, "learning_rate": 4.7761599836024186e-05, "loss": 1.3288, "step": 1598 }, { "epoch": 0.14, "learning_rate": 4.7758634026946246e-05, "loss": 1.279, "step": 1599 }, { "epoch": 0.14, "learning_rate": 4.775566634657491e-05, "loss": 1.3508, "step": 1600 }, { "epoch": 0.14, "learning_rate": 4.775269679515418e-05, "loss": 1.3946, "step": 1601 }, { "epoch": 0.14, "learning_rate": 4.774972537292822e-05, "loss": 1.2784, "step": 1602 }, { "epoch": 0.14, "learning_rate": 4.774675208014136e-05, "loss": 1.3729, "step": 1603 }, { "epoch": 0.14, "learning_rate": 4.774377691703806e-05, "loss": 1.3834, "step": 1604 }, { "epoch": 0.15, "learning_rate": 4.774079988386296e-05, "loss": 1.3573, "step": 1605 }, { "epoch": 0.15, "learning_rate": 4.7737820980860836e-05, "loss": 1.3983, "step": 1606 }, { "epoch": 0.15, "learning_rate": 4.773484020827663e-05, "loss": 1.347, "step": 1607 }, { "epoch": 0.15, "learning_rate": 4.7731857566355426e-05, "loss": 1.3801, "step": 1608 }, { "epoch": 0.15, "learning_rate": 4.7728873055342465e-05, "loss": 1.325, "step": 1609 }, { "epoch": 0.15, "learning_rate": 4.7725886675483146e-05, "loss": 1.3784, "step": 1610 }, { "epoch": 0.15, "learning_rate": 4.7722898427023025e-05, "loss": 1.3261, "step": 1611 }, { "epoch": 0.15, "learning_rate": 4.77199083102078e-05, "loss": 1.3387, "step": 1612 }, { "epoch": 0.15, "learning_rate": 4.771691632528333e-05, "loss": 1.32, "step": 1613 }, { "epoch": 0.15, "learning_rate": 4.771392247249563e-05, "loss": 1.3624, "step": 1614 }, { "epoch": 0.15, "learning_rate": 4.7710926752090856e-05, "loss": 1.3043, "step": 1615 }, { "epoch": 0.15, "learning_rate": 4.7707929164315336e-05, "loss": 1.376, "step": 1616 }, { "epoch": 0.15, "learning_rate": 4.7704929709415535e-05, "loss": 1.3543, "step": 1617 }, { "epoch": 0.15, "learning_rate": 4.7701928387638084e-05, "loss": 1.3409, "step": 1618 }, { "epoch": 0.15, "learning_rate": 4.769892519922976e-05, "loss": 1.3493, "step": 1619 }, { "epoch": 0.15, "learning_rate": 4.769592014443749e-05, "loss": 1.3181, "step": 1620 }, { "epoch": 0.15, "learning_rate": 4.7692913223508374e-05, "loss": 1.3427, "step": 1621 }, { "epoch": 0.15, "learning_rate": 4.7689904436689635e-05, "loss": 1.3028, "step": 1622 }, { "epoch": 0.15, "learning_rate": 4.768689378422867e-05, "loss": 1.3666, "step": 1623 }, { "epoch": 0.15, "learning_rate": 4.7683881266373045e-05, "loss": 1.3539, "step": 1624 }, { "epoch": 0.15, "learning_rate": 4.768086688337042e-05, "loss": 1.3061, "step": 1625 }, { "epoch": 0.15, "learning_rate": 4.767785063546869e-05, "loss": 1.3764, "step": 1626 }, { "epoch": 0.15, "learning_rate": 4.767483252291584e-05, "loss": 1.3769, "step": 1627 }, { "epoch": 0.15, "learning_rate": 4.7671812545960025e-05, "loss": 1.3782, "step": 1628 }, { "epoch": 0.15, "learning_rate": 4.766879070484956e-05, "loss": 1.3307, "step": 1629 }, { "epoch": 0.15, "learning_rate": 4.7665766999832935e-05, "loss": 1.2972, "step": 1630 }, { "epoch": 0.15, "learning_rate": 4.766274143115874e-05, "loss": 1.3587, "step": 1631 }, { "epoch": 0.15, "learning_rate": 4.765971399907576e-05, "loss": 1.3478, "step": 1632 }, { "epoch": 0.15, "learning_rate": 4.765668470383291e-05, "loss": 1.354, "step": 1633 }, { "epoch": 0.15, "learning_rate": 4.765365354567929e-05, "loss": 1.3637, "step": 1634 }, { "epoch": 0.15, "learning_rate": 4.765062052486412e-05, "loss": 1.4227, "step": 1635 }, { "epoch": 0.15, "learning_rate": 4.764758564163678e-05, "loss": 1.4121, "step": 1636 }, { "epoch": 0.15, "learning_rate": 4.7644548896246814e-05, "loss": 1.3406, "step": 1637 }, { "epoch": 0.15, "learning_rate": 4.7641510288943926e-05, "loss": 1.3869, "step": 1638 }, { "epoch": 0.15, "learning_rate": 4.763846981997794e-05, "loss": 1.3527, "step": 1639 }, { "epoch": 0.15, "learning_rate": 4.763542748959886e-05, "loss": 1.327, "step": 1640 }, { "epoch": 0.15, "learning_rate": 4.763238329805684e-05, "loss": 1.321, "step": 1641 }, { "epoch": 0.15, "learning_rate": 4.762933724560219e-05, "loss": 1.2729, "step": 1642 }, { "epoch": 0.15, "learning_rate": 4.762628933248536e-05, "loss": 1.3602, "step": 1643 }, { "epoch": 0.15, "learning_rate": 4.7623239558956956e-05, "loss": 1.366, "step": 1644 }, { "epoch": 0.15, "learning_rate": 4.762018792526775e-05, "loss": 1.3457, "step": 1645 }, { "epoch": 0.15, "learning_rate": 4.761713443166865e-05, "loss": 1.3783, "step": 1646 }, { "epoch": 0.15, "learning_rate": 4.761407907841073e-05, "loss": 1.3822, "step": 1647 }, { "epoch": 0.15, "learning_rate": 4.761102186574521e-05, "loss": 1.3669, "step": 1648 }, { "epoch": 0.15, "learning_rate": 4.7607962793923465e-05, "loss": 1.3555, "step": 1649 }, { "epoch": 0.15, "learning_rate": 4.7604901863197024e-05, "loss": 1.3511, "step": 1650 }, { "epoch": 0.15, "learning_rate": 4.760183907381757e-05, "loss": 1.2991, "step": 1651 }, { "epoch": 0.15, "learning_rate": 4.7598774426036916e-05, "loss": 1.3153, "step": 1652 }, { "epoch": 0.15, "learning_rate": 4.759570792010708e-05, "loss": 1.3296, "step": 1653 }, { "epoch": 0.15, "learning_rate": 4.759263955628017e-05, "loss": 1.3953, "step": 1654 }, { "epoch": 0.15, "learning_rate": 4.758956933480849e-05, "loss": 1.3166, "step": 1655 }, { "epoch": 0.15, "learning_rate": 4.75864972559445e-05, "loss": 1.3307, "step": 1656 }, { "epoch": 0.15, "learning_rate": 4.758342331994077e-05, "loss": 1.326, "step": 1657 }, { "epoch": 0.15, "learning_rate": 4.758034752705007e-05, "loss": 1.3249, "step": 1658 }, { "epoch": 0.15, "learning_rate": 4.757726987752529e-05, "loss": 1.3759, "step": 1659 }, { "epoch": 0.15, "learning_rate": 4.7574190371619485e-05, "loss": 1.3572, "step": 1660 }, { "epoch": 0.15, "learning_rate": 4.757110900958588e-05, "loss": 1.3615, "step": 1661 }, { "epoch": 0.15, "learning_rate": 4.756802579167781e-05, "loss": 1.3924, "step": 1662 }, { "epoch": 0.15, "learning_rate": 4.7564940718148806e-05, "loss": 1.3574, "step": 1663 }, { "epoch": 0.15, "learning_rate": 4.756185378925253e-05, "loss": 1.3623, "step": 1664 }, { "epoch": 0.15, "learning_rate": 4.755876500524279e-05, "loss": 1.3992, "step": 1665 }, { "epoch": 0.15, "learning_rate": 4.755567436637356e-05, "loss": 1.3378, "step": 1666 }, { "epoch": 0.15, "learning_rate": 4.755258187289898e-05, "loss": 1.321, "step": 1667 }, { "epoch": 0.15, "learning_rate": 4.7549487525073305e-05, "loss": 1.371, "step": 1668 }, { "epoch": 0.15, "learning_rate": 4.7546391323150975e-05, "loss": 1.3115, "step": 1669 }, { "epoch": 0.15, "learning_rate": 4.754329326738656e-05, "loss": 1.3794, "step": 1670 }, { "epoch": 0.15, "learning_rate": 4.75401933580348e-05, "loss": 1.3284, "step": 1671 }, { "epoch": 0.15, "learning_rate": 4.753709159535058e-05, "loss": 1.3587, "step": 1672 }, { "epoch": 0.15, "learning_rate": 4.7533987979588935e-05, "loss": 1.3392, "step": 1673 }, { "epoch": 0.15, "learning_rate": 4.753088251100506e-05, "loss": 1.3096, "step": 1674 }, { "epoch": 0.15, "learning_rate": 4.7527775189854284e-05, "loss": 1.3211, "step": 1675 }, { "epoch": 0.15, "learning_rate": 4.752466601639212e-05, "loss": 1.2448, "step": 1676 }, { "epoch": 0.15, "learning_rate": 4.752155499087421e-05, "loss": 1.3623, "step": 1677 }, { "epoch": 0.15, "learning_rate": 4.751844211355634e-05, "loss": 1.3683, "step": 1678 }, { "epoch": 0.15, "learning_rate": 4.7515327384694476e-05, "loss": 1.3622, "step": 1679 }, { "epoch": 0.15, "learning_rate": 4.7512210804544714e-05, "loss": 1.2946, "step": 1680 }, { "epoch": 0.15, "learning_rate": 4.7509092373363314e-05, "loss": 1.3288, "step": 1681 }, { "epoch": 0.15, "learning_rate": 4.7505972091406685e-05, "loss": 1.3354, "step": 1682 }, { "epoch": 0.15, "learning_rate": 4.750284995893138e-05, "loss": 1.4073, "step": 1683 }, { "epoch": 0.15, "learning_rate": 4.7499725976194125e-05, "loss": 1.4047, "step": 1684 }, { "epoch": 0.15, "learning_rate": 4.7496600143451774e-05, "loss": 1.3785, "step": 1685 }, { "epoch": 0.15, "learning_rate": 4.749347246096134e-05, "loss": 1.3131, "step": 1686 }, { "epoch": 0.15, "learning_rate": 4.749034292898e-05, "loss": 1.3578, "step": 1687 }, { "epoch": 0.15, "learning_rate": 4.748721154776508e-05, "loss": 1.3299, "step": 1688 }, { "epoch": 0.15, "learning_rate": 4.7484078317574046e-05, "loss": 1.3193, "step": 1689 }, { "epoch": 0.15, "learning_rate": 4.7480943238664515e-05, "loss": 1.36, "step": 1690 }, { "epoch": 0.15, "learning_rate": 4.7477806311294276e-05, "loss": 1.3466, "step": 1691 }, { "epoch": 0.15, "learning_rate": 4.747466753572125e-05, "loss": 1.3548, "step": 1692 }, { "epoch": 0.15, "learning_rate": 4.7471526912203526e-05, "loss": 1.3438, "step": 1693 }, { "epoch": 0.15, "learning_rate": 4.746838444099934e-05, "loss": 1.2717, "step": 1694 }, { "epoch": 0.15, "learning_rate": 4.746524012236706e-05, "loss": 1.4008, "step": 1695 }, { "epoch": 0.15, "learning_rate": 4.746209395656523e-05, "loss": 1.3385, "step": 1696 }, { "epoch": 0.15, "learning_rate": 4.7458945943852544e-05, "loss": 1.3031, "step": 1697 }, { "epoch": 0.15, "learning_rate": 4.745579608448784e-05, "loss": 1.364, "step": 1698 }, { "epoch": 0.15, "learning_rate": 4.7452644378730115e-05, "loss": 1.3558, "step": 1699 }, { "epoch": 0.15, "learning_rate": 4.74494908268385e-05, "loss": 1.3444, "step": 1700 }, { "epoch": 0.15, "learning_rate": 4.74463354290723e-05, "loss": 1.3645, "step": 1701 }, { "epoch": 0.15, "learning_rate": 4.744317818569096e-05, "loss": 1.354, "step": 1702 }, { "epoch": 0.15, "learning_rate": 4.744001909695409e-05, "loss": 1.374, "step": 1703 }, { "epoch": 0.15, "learning_rate": 4.743685816312142e-05, "loss": 1.3806, "step": 1704 }, { "epoch": 0.15, "learning_rate": 4.743369538445288e-05, "loss": 1.3287, "step": 1705 }, { "epoch": 0.15, "learning_rate": 4.7430530761208494e-05, "loss": 1.3856, "step": 1706 }, { "epoch": 0.15, "learning_rate": 4.742736429364849e-05, "loss": 1.3614, "step": 1707 }, { "epoch": 0.15, "learning_rate": 4.7424195982033226e-05, "loss": 1.3263, "step": 1708 }, { "epoch": 0.15, "learning_rate": 4.74210258266232e-05, "loss": 1.3649, "step": 1709 }, { "epoch": 0.15, "learning_rate": 4.741785382767907e-05, "loss": 1.321, "step": 1710 }, { "epoch": 0.15, "learning_rate": 4.7414679985461666e-05, "loss": 1.3041, "step": 1711 }, { "epoch": 0.15, "learning_rate": 4.741150430023195e-05, "loss": 1.3573, "step": 1712 }, { "epoch": 0.15, "learning_rate": 4.740832677225102e-05, "loss": 1.3624, "step": 1713 }, { "epoch": 0.15, "learning_rate": 4.740514740178015e-05, "loss": 1.3091, "step": 1714 }, { "epoch": 0.15, "learning_rate": 4.740196618908078e-05, "loss": 1.3467, "step": 1715 }, { "epoch": 0.16, "learning_rate": 4.739878313441445e-05, "loss": 1.3568, "step": 1716 }, { "epoch": 0.16, "learning_rate": 4.73955982380429e-05, "loss": 1.3159, "step": 1717 }, { "epoch": 0.16, "learning_rate": 4.739241150022799e-05, "loss": 1.3159, "step": 1718 }, { "epoch": 0.16, "learning_rate": 4.738922292123177e-05, "loss": 1.3261, "step": 1719 }, { "epoch": 0.16, "learning_rate": 4.7386032501316386e-05, "loss": 1.3486, "step": 1720 }, { "epoch": 0.16, "learning_rate": 4.738284024074418e-05, "loss": 1.3057, "step": 1721 }, { "epoch": 0.16, "learning_rate": 4.7379646139777636e-05, "loss": 1.2799, "step": 1722 }, { "epoch": 0.16, "learning_rate": 4.737645019867937e-05, "loss": 1.3811, "step": 1723 }, { "epoch": 0.16, "learning_rate": 4.737325241771217e-05, "loss": 1.3698, "step": 1724 }, { "epoch": 0.16, "learning_rate": 4.7370052797138964e-05, "loss": 1.2833, "step": 1725 }, { "epoch": 0.16, "learning_rate": 4.736685133722285e-05, "loss": 1.3831, "step": 1726 }, { "epoch": 0.16, "learning_rate": 4.736364803822705e-05, "loss": 1.3531, "step": 1727 }, { "epoch": 0.16, "learning_rate": 4.736044290041495e-05, "loss": 1.3552, "step": 1728 }, { "epoch": 0.16, "learning_rate": 4.7357235924050106e-05, "loss": 1.3658, "step": 1729 }, { "epoch": 0.16, "learning_rate": 4.735402710939618e-05, "loss": 1.4148, "step": 1730 }, { "epoch": 0.16, "learning_rate": 4.735081645671702e-05, "loss": 1.3796, "step": 1731 }, { "epoch": 0.16, "learning_rate": 4.734760396627663e-05, "loss": 1.3522, "step": 1732 }, { "epoch": 0.16, "learning_rate": 4.734438963833914e-05, "loss": 1.365, "step": 1733 }, { "epoch": 0.16, "learning_rate": 4.734117347316884e-05, "loss": 1.3117, "step": 1734 }, { "epoch": 0.16, "learning_rate": 4.7337955471030194e-05, "loss": 1.3716, "step": 1735 }, { "epoch": 0.16, "learning_rate": 4.733473563218777e-05, "loss": 1.3422, "step": 1736 }, { "epoch": 0.16, "learning_rate": 4.7331513956906336e-05, "loss": 1.3061, "step": 1737 }, { "epoch": 0.16, "learning_rate": 4.7328290445450776e-05, "loss": 1.3177, "step": 1738 }, { "epoch": 0.16, "learning_rate": 4.732506509808615e-05, "loss": 1.3926, "step": 1739 }, { "epoch": 0.16, "learning_rate": 4.7321837915077644e-05, "loss": 1.3647, "step": 1740 }, { "epoch": 0.16, "learning_rate": 4.731860889669062e-05, "loss": 1.3851, "step": 1741 }, { "epoch": 0.16, "learning_rate": 4.7315378043190567e-05, "loss": 1.3085, "step": 1742 }, { "epoch": 0.16, "learning_rate": 4.731214535484315e-05, "loss": 1.3428, "step": 1743 }, { "epoch": 0.16, "learning_rate": 4.730891083191415e-05, "loss": 1.3595, "step": 1744 }, { "epoch": 0.16, "learning_rate": 4.730567447466956e-05, "loss": 1.3505, "step": 1745 }, { "epoch": 0.16, "learning_rate": 4.730243628337544e-05, "loss": 1.3466, "step": 1746 }, { "epoch": 0.16, "learning_rate": 4.729919625829808e-05, "loss": 1.3964, "step": 1747 }, { "epoch": 0.16, "learning_rate": 4.729595439970387e-05, "loss": 1.3892, "step": 1748 }, { "epoch": 0.16, "learning_rate": 4.729271070785936e-05, "loss": 1.3531, "step": 1749 }, { "epoch": 0.16, "learning_rate": 4.728946518303127e-05, "loss": 1.3098, "step": 1750 }, { "epoch": 0.16, "learning_rate": 4.7286217825486456e-05, "loss": 1.332, "step": 1751 }, { "epoch": 0.16, "learning_rate": 4.728296863549192e-05, "loss": 1.3259, "step": 1752 }, { "epoch": 0.16, "learning_rate": 4.727971761331483e-05, "loss": 1.3894, "step": 1753 }, { "epoch": 0.16, "learning_rate": 4.7276464759222494e-05, "loss": 1.3312, "step": 1754 }, { "epoch": 0.16, "learning_rate": 4.727321007348238e-05, "loss": 1.3119, "step": 1755 }, { "epoch": 0.16, "learning_rate": 4.7269953556362076e-05, "loss": 1.3423, "step": 1756 }, { "epoch": 0.16, "learning_rate": 4.726669520812937e-05, "loss": 1.2927, "step": 1757 }, { "epoch": 0.16, "learning_rate": 4.726343502905216e-05, "loss": 1.2913, "step": 1758 }, { "epoch": 0.16, "learning_rate": 4.726017301939851e-05, "loss": 1.3712, "step": 1759 }, { "epoch": 0.16, "learning_rate": 4.725690917943665e-05, "loss": 1.2841, "step": 1760 }, { "epoch": 0.16, "learning_rate": 4.725364350943492e-05, "loss": 1.3975, "step": 1761 }, { "epoch": 0.16, "learning_rate": 4.7250376009661846e-05, "loss": 1.3263, "step": 1762 }, { "epoch": 0.16, "learning_rate": 4.72471066803861e-05, "loss": 1.3684, "step": 1763 }, { "epoch": 0.16, "learning_rate": 4.724383552187648e-05, "loss": 1.3592, "step": 1764 }, { "epoch": 0.16, "learning_rate": 4.724056253440197e-05, "loss": 1.2965, "step": 1765 }, { "epoch": 0.16, "learning_rate": 4.7237287718231674e-05, "loss": 1.3624, "step": 1766 }, { "epoch": 0.16, "learning_rate": 4.723401107363486e-05, "loss": 1.3781, "step": 1767 }, { "epoch": 0.16, "learning_rate": 4.723073260088097e-05, "loss": 1.3488, "step": 1768 }, { "epoch": 0.16, "learning_rate": 4.7227452300239526e-05, "loss": 1.3954, "step": 1769 }, { "epoch": 0.16, "learning_rate": 4.722417017198028e-05, "loss": 1.3405, "step": 1770 }, { "epoch": 0.16, "learning_rate": 4.722088621637309e-05, "loss": 1.4261, "step": 1771 }, { "epoch": 0.16, "learning_rate": 4.721760043368797e-05, "loss": 1.366, "step": 1772 }, { "epoch": 0.16, "learning_rate": 4.7214312824195095e-05, "loss": 1.3365, "step": 1773 }, { "epoch": 0.16, "learning_rate": 4.7211023388164776e-05, "loss": 1.3318, "step": 1774 }, { "epoch": 0.16, "learning_rate": 4.720773212586749e-05, "loss": 1.352, "step": 1775 }, { "epoch": 0.16, "learning_rate": 4.720443903757385e-05, "loss": 1.4089, "step": 1776 }, { "epoch": 0.16, "learning_rate": 4.720114412355463e-05, "loss": 1.398, "step": 1777 }, { "epoch": 0.16, "learning_rate": 4.7197847384080747e-05, "loss": 1.338, "step": 1778 }, { "epoch": 0.16, "learning_rate": 4.7194548819423266e-05, "loss": 1.3587, "step": 1779 }, { "epoch": 0.16, "learning_rate": 4.7191248429853406e-05, "loss": 1.3945, "step": 1780 }, { "epoch": 0.16, "learning_rate": 4.7187946215642556e-05, "loss": 1.2749, "step": 1781 }, { "epoch": 0.16, "learning_rate": 4.7184642177062206e-05, "loss": 1.2959, "step": 1782 }, { "epoch": 0.16, "learning_rate": 4.718133631438404e-05, "loss": 1.3583, "step": 1783 }, { "epoch": 0.16, "learning_rate": 4.717802862787988e-05, "loss": 1.3245, "step": 1784 }, { "epoch": 0.16, "learning_rate": 4.71747191178217e-05, "loss": 1.3376, "step": 1785 }, { "epoch": 0.16, "learning_rate": 4.71714077844816e-05, "loss": 1.2907, "step": 1786 }, { "epoch": 0.16, "learning_rate": 4.716809462813187e-05, "loss": 1.3405, "step": 1787 }, { "epoch": 0.16, "learning_rate": 4.716477964904491e-05, "loss": 1.3352, "step": 1788 }, { "epoch": 0.16, "learning_rate": 4.716146284749331e-05, "loss": 1.3935, "step": 1789 }, { "epoch": 0.16, "learning_rate": 4.715814422374977e-05, "loss": 1.3547, "step": 1790 }, { "epoch": 0.16, "learning_rate": 4.715482377808717e-05, "loss": 1.3361, "step": 1791 }, { "epoch": 0.16, "learning_rate": 4.7151501510778526e-05, "loss": 1.38, "step": 1792 }, { "epoch": 0.16, "learning_rate": 4.7148177422097e-05, "loss": 1.3021, "step": 1793 }, { "epoch": 0.16, "learning_rate": 4.714485151231593e-05, "loss": 1.3952, "step": 1794 }, { "epoch": 0.16, "learning_rate": 4.714152378170875e-05, "loss": 1.3783, "step": 1795 }, { "epoch": 0.16, "learning_rate": 4.7138194230549106e-05, "loss": 1.3575, "step": 1796 }, { "epoch": 0.16, "learning_rate": 4.7134862859110754e-05, "loss": 1.3742, "step": 1797 }, { "epoch": 0.16, "learning_rate": 4.713152966766761e-05, "loss": 1.4106, "step": 1798 }, { "epoch": 0.16, "learning_rate": 4.712819465649375e-05, "loss": 1.3854, "step": 1799 }, { "epoch": 0.16, "learning_rate": 4.712485782586338e-05, "loss": 1.3383, "step": 1800 }, { "epoch": 0.16, "learning_rate": 4.712151917605086e-05, "loss": 1.3391, "step": 1801 }, { "epoch": 0.16, "learning_rate": 4.711817870733073e-05, "loss": 1.3528, "step": 1802 }, { "epoch": 0.16, "learning_rate": 4.7114836419977634e-05, "loss": 1.3137, "step": 1803 }, { "epoch": 0.16, "learning_rate": 4.711149231426639e-05, "loss": 1.4064, "step": 1804 }, { "epoch": 0.16, "learning_rate": 4.7108146390471966e-05, "loss": 1.3539, "step": 1805 }, { "epoch": 0.16, "learning_rate": 4.710479864886947e-05, "loss": 1.3123, "step": 1806 }, { "epoch": 0.16, "learning_rate": 4.710144908973417e-05, "loss": 1.3753, "step": 1807 }, { "epoch": 0.16, "learning_rate": 4.7098097713341476e-05, "loss": 1.312, "step": 1808 }, { "epoch": 0.16, "learning_rate": 4.709474451996695e-05, "loss": 1.3687, "step": 1809 }, { "epoch": 0.16, "learning_rate": 4.70913895098863e-05, "loss": 1.3283, "step": 1810 }, { "epoch": 0.16, "learning_rate": 4.7088032683375395e-05, "loss": 1.3194, "step": 1811 }, { "epoch": 0.16, "learning_rate": 4.708467404071023e-05, "loss": 1.3323, "step": 1812 }, { "epoch": 0.16, "learning_rate": 4.708131358216698e-05, "loss": 1.3755, "step": 1813 }, { "epoch": 0.16, "learning_rate": 4.707795130802195e-05, "loss": 1.3874, "step": 1814 }, { "epoch": 0.16, "learning_rate": 4.707458721855159e-05, "loss": 1.3422, "step": 1815 }, { "epoch": 0.16, "learning_rate": 4.707122131403251e-05, "loss": 1.3577, "step": 1816 }, { "epoch": 0.16, "learning_rate": 4.7067853594741475e-05, "loss": 1.3425, "step": 1817 }, { "epoch": 0.16, "learning_rate": 4.706448406095538e-05, "loss": 1.3266, "step": 1818 }, { "epoch": 0.16, "learning_rate": 4.7061112712951286e-05, "loss": 1.3764, "step": 1819 }, { "epoch": 0.16, "learning_rate": 4.7057739551006386e-05, "loss": 1.296, "step": 1820 }, { "epoch": 0.16, "learning_rate": 4.705436457539805e-05, "loss": 1.326, "step": 1821 }, { "epoch": 0.16, "learning_rate": 4.7050987786403764e-05, "loss": 1.3921, "step": 1822 }, { "epoch": 0.16, "learning_rate": 4.704760918430119e-05, "loss": 1.3542, "step": 1823 }, { "epoch": 0.16, "learning_rate": 4.7044228769368126e-05, "loss": 1.372, "step": 1824 }, { "epoch": 0.16, "learning_rate": 4.704084654188252e-05, "loss": 1.3421, "step": 1825 }, { "epoch": 0.16, "learning_rate": 4.703746250212247e-05, "loss": 1.3002, "step": 1826 }, { "epoch": 0.17, "learning_rate": 4.703407665036622e-05, "loss": 1.2762, "step": 1827 }, { "epoch": 0.17, "learning_rate": 4.703068898689218e-05, "loss": 1.3499, "step": 1828 }, { "epoch": 0.17, "learning_rate": 4.702729951197888e-05, "loss": 1.3094, "step": 1829 }, { "epoch": 0.17, "learning_rate": 4.702390822590502e-05, "loss": 1.3556, "step": 1830 }, { "epoch": 0.17, "learning_rate": 4.7020515128949446e-05, "loss": 1.2729, "step": 1831 }, { "epoch": 0.17, "learning_rate": 4.7017120221391145e-05, "loss": 1.3401, "step": 1832 }, { "epoch": 0.17, "learning_rate": 4.701372350350926e-05, "loss": 1.2902, "step": 1833 }, { "epoch": 0.17, "learning_rate": 4.7010324975583084e-05, "loss": 1.4109, "step": 1834 }, { "epoch": 0.17, "learning_rate": 4.7006924637892054e-05, "loss": 1.359, "step": 1835 }, { "epoch": 0.17, "learning_rate": 4.7003522490715756e-05, "loss": 1.3732, "step": 1836 }, { "epoch": 0.17, "learning_rate": 4.700011853433392e-05, "loss": 1.3435, "step": 1837 }, { "epoch": 0.17, "learning_rate": 4.699671276902645e-05, "loss": 1.3301, "step": 1838 }, { "epoch": 0.17, "learning_rate": 4.699330519507336e-05, "loss": 1.3291, "step": 1839 }, { "epoch": 0.17, "learning_rate": 4.6989895812754834e-05, "loss": 1.3554, "step": 1840 }, { "epoch": 0.17, "learning_rate": 4.698648462235122e-05, "loss": 1.3038, "step": 1841 }, { "epoch": 0.17, "learning_rate": 4.6983071624142985e-05, "loss": 1.3405, "step": 1842 }, { "epoch": 0.17, "learning_rate": 4.697965681841076e-05, "loss": 1.3563, "step": 1843 }, { "epoch": 0.17, "learning_rate": 4.697624020543532e-05, "loss": 1.357, "step": 1844 }, { "epoch": 0.17, "learning_rate": 4.69728217854976e-05, "loss": 1.3403, "step": 1845 }, { "epoch": 0.17, "learning_rate": 4.6969401558878655e-05, "loss": 1.3349, "step": 1846 }, { "epoch": 0.17, "learning_rate": 4.6965979525859724e-05, "loss": 1.3319, "step": 1847 }, { "epoch": 0.17, "learning_rate": 4.696255568672218e-05, "loss": 1.3457, "step": 1848 }, { "epoch": 0.17, "learning_rate": 4.6959130041747534e-05, "loss": 1.3562, "step": 1849 }, { "epoch": 0.17, "learning_rate": 4.695570259121745e-05, "loss": 1.3264, "step": 1850 }, { "epoch": 0.17, "learning_rate": 4.695227333541376e-05, "loss": 1.4315, "step": 1851 }, { "epoch": 0.17, "learning_rate": 4.6948842274618424e-05, "loss": 1.385, "step": 1852 }, { "epoch": 0.17, "learning_rate": 4.6945409409113536e-05, "loss": 1.3046, "step": 1853 }, { "epoch": 0.17, "learning_rate": 4.6941974739181395e-05, "loss": 1.3322, "step": 1854 }, { "epoch": 0.17, "learning_rate": 4.693853826510438e-05, "loss": 1.3705, "step": 1855 }, { "epoch": 0.17, "learning_rate": 4.693509998716506e-05, "loss": 1.2721, "step": 1856 }, { "epoch": 0.17, "learning_rate": 4.6931659905646155e-05, "loss": 1.3449, "step": 1857 }, { "epoch": 0.17, "learning_rate": 4.69282180208305e-05, "loss": 1.3291, "step": 1858 }, { "epoch": 0.17, "learning_rate": 4.692477433300111e-05, "loss": 1.3771, "step": 1859 }, { "epoch": 0.17, "learning_rate": 4.6921328842441125e-05, "loss": 1.365, "step": 1860 }, { "epoch": 0.17, "learning_rate": 4.6917881549433866e-05, "loss": 1.3192, "step": 1861 }, { "epoch": 0.17, "learning_rate": 4.691443245426277e-05, "loss": 1.3211, "step": 1862 }, { "epoch": 0.17, "learning_rate": 4.691098155721142e-05, "loss": 1.3357, "step": 1863 }, { "epoch": 0.17, "learning_rate": 4.690752885856359e-05, "loss": 1.3188, "step": 1864 }, { "epoch": 0.17, "learning_rate": 4.690407435860315e-05, "loss": 1.406, "step": 1865 }, { "epoch": 0.17, "learning_rate": 4.690061805761414e-05, "loss": 1.3395, "step": 1866 }, { "epoch": 0.17, "learning_rate": 4.6897159955880764e-05, "loss": 1.3657, "step": 1867 }, { "epoch": 0.17, "learning_rate": 4.689370005368735e-05, "loss": 1.3756, "step": 1868 }, { "epoch": 0.17, "learning_rate": 4.6890238351318385e-05, "loss": 1.3418, "step": 1869 }, { "epoch": 0.17, "learning_rate": 4.68867748490585e-05, "loss": 1.3616, "step": 1870 }, { "epoch": 0.17, "learning_rate": 4.6883309547192476e-05, "loss": 1.3601, "step": 1871 }, { "epoch": 0.17, "learning_rate": 4.687984244600525e-05, "loss": 1.3461, "step": 1872 }, { "epoch": 0.17, "learning_rate": 4.6876373545781895e-05, "loss": 1.4095, "step": 1873 }, { "epoch": 0.17, "learning_rate": 4.6872902846807625e-05, "loss": 1.358, "step": 1874 }, { "epoch": 0.17, "learning_rate": 4.6869430349367824e-05, "loss": 1.3341, "step": 1875 }, { "epoch": 0.17, "learning_rate": 4.6865956053748016e-05, "loss": 1.3701, "step": 1876 }, { "epoch": 0.17, "learning_rate": 4.686247996023386e-05, "loss": 1.3176, "step": 1877 }, { "epoch": 0.17, "learning_rate": 4.685900206911118e-05, "loss": 1.3221, "step": 1878 }, { "epoch": 0.17, "learning_rate": 4.685552238066593e-05, "loss": 1.3223, "step": 1879 }, { "epoch": 0.17, "learning_rate": 4.685204089518424e-05, "loss": 1.3592, "step": 1880 }, { "epoch": 0.17, "learning_rate": 4.6848557612952336e-05, "loss": 1.3702, "step": 1881 }, { "epoch": 0.17, "learning_rate": 4.6845072534256667e-05, "loss": 1.3991, "step": 1882 }, { "epoch": 0.17, "learning_rate": 4.684158565938376e-05, "loss": 1.3345, "step": 1883 }, { "epoch": 0.17, "learning_rate": 4.6838096988620336e-05, "loss": 1.3245, "step": 1884 }, { "epoch": 0.17, "learning_rate": 4.683460652225323e-05, "loss": 1.3356, "step": 1885 }, { "epoch": 0.17, "learning_rate": 4.683111426056945e-05, "loss": 1.3619, "step": 1886 }, { "epoch": 0.17, "learning_rate": 4.682762020385613e-05, "loss": 1.3699, "step": 1887 }, { "epoch": 0.17, "learning_rate": 4.682412435240058e-05, "loss": 1.3336, "step": 1888 }, { "epoch": 0.17, "learning_rate": 4.6820626706490233e-05, "loss": 1.3585, "step": 1889 }, { "epoch": 0.17, "learning_rate": 4.6817127266412675e-05, "loss": 1.3493, "step": 1890 }, { "epoch": 0.17, "learning_rate": 4.6813626032455646e-05, "loss": 1.3127, "step": 1891 }, { "epoch": 0.17, "learning_rate": 4.681012300490702e-05, "loss": 1.3358, "step": 1892 }, { "epoch": 0.17, "learning_rate": 4.680661818405485e-05, "loss": 1.3054, "step": 1893 }, { "epoch": 0.17, "learning_rate": 4.6803111570187285e-05, "loss": 1.3105, "step": 1894 }, { "epoch": 0.17, "learning_rate": 4.679960316359268e-05, "loss": 1.3519, "step": 1895 }, { "epoch": 0.17, "learning_rate": 4.679609296455949e-05, "loss": 1.3852, "step": 1896 }, { "epoch": 0.17, "learning_rate": 4.679258097337633e-05, "loss": 1.324, "step": 1897 }, { "epoch": 0.17, "learning_rate": 4.6789067190331995e-05, "loss": 1.3936, "step": 1898 }, { "epoch": 0.17, "learning_rate": 4.678555161571537e-05, "loss": 1.3419, "step": 1899 }, { "epoch": 0.17, "learning_rate": 4.678203424981554e-05, "loss": 1.3715, "step": 1900 }, { "epoch": 0.17, "learning_rate": 4.6778515092921704e-05, "loss": 1.3763, "step": 1901 }, { "epoch": 0.17, "learning_rate": 4.677499414532322e-05, "loss": 1.3944, "step": 1902 }, { "epoch": 0.17, "learning_rate": 4.6771471407309594e-05, "loss": 1.3393, "step": 1903 }, { "epoch": 0.17, "learning_rate": 4.676794687917048e-05, "loss": 1.2974, "step": 1904 }, { "epoch": 0.17, "learning_rate": 4.676442056119567e-05, "loss": 1.2159, "step": 1905 }, { "epoch": 0.17, "learning_rate": 4.676089245367512e-05, "loss": 1.3624, "step": 1906 }, { "epoch": 0.17, "learning_rate": 4.675736255689891e-05, "loss": 1.3546, "step": 1907 }, { "epoch": 0.17, "learning_rate": 4.6753830871157294e-05, "loss": 1.2833, "step": 1908 }, { "epoch": 0.17, "learning_rate": 4.675029739674065e-05, "loss": 1.3149, "step": 1909 }, { "epoch": 0.17, "learning_rate": 4.674676213393952e-05, "loss": 1.3193, "step": 1910 }, { "epoch": 0.17, "learning_rate": 4.6743225083044586e-05, "loss": 1.3487, "step": 1911 }, { "epoch": 0.17, "learning_rate": 4.673968624434666e-05, "loss": 1.381, "step": 1912 }, { "epoch": 0.17, "learning_rate": 4.673614561813673e-05, "loss": 1.3419, "step": 1913 }, { "epoch": 0.17, "learning_rate": 4.6732603204705924e-05, "loss": 1.3462, "step": 1914 }, { "epoch": 0.17, "learning_rate": 4.6729059004345497e-05, "loss": 1.3424, "step": 1915 }, { "epoch": 0.17, "learning_rate": 4.672551301734689e-05, "loss": 1.3355, "step": 1916 }, { "epoch": 0.17, "learning_rate": 4.672196524400164e-05, "loss": 1.3772, "step": 1917 }, { "epoch": 0.17, "learning_rate": 4.671841568460146e-05, "loss": 1.2813, "step": 1918 }, { "epoch": 0.17, "learning_rate": 4.6714864339438225e-05, "loss": 1.2772, "step": 1919 }, { "epoch": 0.17, "learning_rate": 4.671131120880392e-05, "loss": 1.328, "step": 1920 }, { "epoch": 0.17, "learning_rate": 4.670775629299071e-05, "loss": 1.3481, "step": 1921 }, { "epoch": 0.17, "learning_rate": 4.670419959229089e-05, "loss": 1.3414, "step": 1922 }, { "epoch": 0.17, "learning_rate": 4.670064110699689e-05, "loss": 1.4073, "step": 1923 }, { "epoch": 0.17, "learning_rate": 4.669708083740132e-05, "loss": 1.3668, "step": 1924 }, { "epoch": 0.17, "learning_rate": 4.6693518783796895e-05, "loss": 1.3029, "step": 1925 }, { "epoch": 0.17, "learning_rate": 4.668995494647653e-05, "loss": 1.3008, "step": 1926 }, { "epoch": 0.17, "learning_rate": 4.6686389325733226e-05, "loss": 1.3925, "step": 1927 }, { "epoch": 0.17, "learning_rate": 4.668282192186019e-05, "loss": 1.3495, "step": 1928 }, { "epoch": 0.17, "learning_rate": 4.667925273515071e-05, "loss": 1.3545, "step": 1929 }, { "epoch": 0.17, "learning_rate": 4.6675681765898296e-05, "loss": 1.3622, "step": 1930 }, { "epoch": 0.17, "learning_rate": 4.667210901439653e-05, "loss": 1.2961, "step": 1931 }, { "epoch": 0.17, "learning_rate": 4.6668534480939204e-05, "loss": 1.3739, "step": 1932 }, { "epoch": 0.17, "learning_rate": 4.666495816582021e-05, "loss": 1.3596, "step": 1933 }, { "epoch": 0.17, "learning_rate": 4.6661380069333616e-05, "loss": 1.3176, "step": 1934 }, { "epoch": 0.17, "learning_rate": 4.665780019177362e-05, "loss": 1.3654, "step": 1935 }, { "epoch": 0.17, "learning_rate": 4.665421853343457e-05, "loss": 1.3488, "step": 1936 }, { "epoch": 0.18, "learning_rate": 4.665063509461097e-05, "loss": 1.3122, "step": 1937 }, { "epoch": 0.18, "learning_rate": 4.6647049875597456e-05, "loss": 1.3327, "step": 1938 }, { "epoch": 0.18, "learning_rate": 4.6643462876688815e-05, "loss": 1.3581, "step": 1939 }, { "epoch": 0.18, "learning_rate": 4.663987409817999e-05, "loss": 1.3864, "step": 1940 }, { "epoch": 0.18, "learning_rate": 4.6636283540366064e-05, "loss": 1.383, "step": 1941 }, { "epoch": 0.18, "learning_rate": 4.6632691203542255e-05, "loss": 1.3282, "step": 1942 }, { "epoch": 0.18, "learning_rate": 4.662909708800395e-05, "loss": 1.3083, "step": 1943 }, { "epoch": 0.18, "learning_rate": 4.662550119404665e-05, "loss": 1.3024, "step": 1944 }, { "epoch": 0.18, "learning_rate": 4.6621903521966046e-05, "loss": 1.3672, "step": 1945 }, { "epoch": 0.18, "learning_rate": 4.661830407205793e-05, "loss": 1.3836, "step": 1946 }, { "epoch": 0.18, "learning_rate": 4.661470284461827e-05, "loss": 1.3159, "step": 1947 }, { "epoch": 0.18, "learning_rate": 4.661109983994318e-05, "loss": 1.303, "step": 1948 }, { "epoch": 0.18, "learning_rate": 4.66074950583289e-05, "loss": 1.3402, "step": 1949 }, { "epoch": 0.18, "learning_rate": 4.660388850007183e-05, "loss": 1.3301, "step": 1950 }, { "epoch": 0.18, "learning_rate": 4.660028016546851e-05, "loss": 1.3298, "step": 1951 }, { "epoch": 0.18, "learning_rate": 4.659667005481564e-05, "loss": 1.2856, "step": 1952 }, { "epoch": 0.18, "learning_rate": 4.6593058168410056e-05, "loss": 1.3276, "step": 1953 }, { "epoch": 0.18, "learning_rate": 4.658944450654872e-05, "loss": 1.3761, "step": 1954 }, { "epoch": 0.18, "learning_rate": 4.658582906952878e-05, "loss": 1.3252, "step": 1955 }, { "epoch": 0.18, "learning_rate": 4.65822118576475e-05, "loss": 1.353, "step": 1956 }, { "epoch": 0.18, "learning_rate": 4.657859287120231e-05, "loss": 1.3538, "step": 1957 }, { "epoch": 0.18, "learning_rate": 4.657497211049076e-05, "loss": 1.3368, "step": 1958 }, { "epoch": 0.18, "learning_rate": 4.657134957581057e-05, "loss": 1.3531, "step": 1959 }, { "epoch": 0.18, "learning_rate": 4.6567725267459604e-05, "loss": 1.3206, "step": 1960 }, { "epoch": 0.18, "learning_rate": 4.6564099185735854e-05, "loss": 1.3537, "step": 1961 }, { "epoch": 0.18, "learning_rate": 4.656047133093747e-05, "loss": 1.2873, "step": 1962 }, { "epoch": 0.18, "learning_rate": 4.655684170336275e-05, "loss": 1.3779, "step": 1963 }, { "epoch": 0.18, "learning_rate": 4.655321030331014e-05, "loss": 1.3551, "step": 1964 }, { "epoch": 0.18, "learning_rate": 4.654957713107822e-05, "loss": 1.3494, "step": 1965 }, { "epoch": 0.18, "learning_rate": 4.654594218696572e-05, "loss": 1.3179, "step": 1966 }, { "epoch": 0.18, "learning_rate": 4.654230547127152e-05, "loss": 1.3454, "step": 1967 }, { "epoch": 0.18, "learning_rate": 4.6538666984294646e-05, "loss": 1.3887, "step": 1968 }, { "epoch": 0.18, "learning_rate": 4.653502672633426e-05, "loss": 1.326, "step": 1969 }, { "epoch": 0.18, "learning_rate": 4.6531384697689684e-05, "loss": 1.3622, "step": 1970 }, { "epoch": 0.18, "learning_rate": 4.652774089866038e-05, "loss": 1.333, "step": 1971 }, { "epoch": 0.18, "learning_rate": 4.652409532954594e-05, "loss": 1.2957, "step": 1972 }, { "epoch": 0.18, "learning_rate": 4.652044799064613e-05, "loss": 1.3212, "step": 1973 }, { "epoch": 0.18, "learning_rate": 4.651679888226084e-05, "loss": 1.3169, "step": 1974 }, { "epoch": 0.18, "learning_rate": 4.651314800469011e-05, "loss": 1.2725, "step": 1975 }, { "epoch": 0.18, "learning_rate": 4.6509495358234125e-05, "loss": 1.348, "step": 1976 }, { "epoch": 0.18, "learning_rate": 4.650584094319324e-05, "loss": 1.3444, "step": 1977 }, { "epoch": 0.18, "learning_rate": 4.65021847598679e-05, "loss": 1.3145, "step": 1978 }, { "epoch": 0.18, "learning_rate": 4.649852680855876e-05, "loss": 1.3153, "step": 1979 }, { "epoch": 0.18, "learning_rate": 4.6494867089566576e-05, "loss": 1.2916, "step": 1980 }, { "epoch": 0.18, "learning_rate": 4.649120560319225e-05, "loss": 1.354, "step": 1981 }, { "epoch": 0.18, "learning_rate": 4.648754234973687e-05, "loss": 1.3288, "step": 1982 }, { "epoch": 0.18, "learning_rate": 4.6483877329501625e-05, "loss": 1.3354, "step": 1983 }, { "epoch": 0.18, "learning_rate": 4.6480210542787864e-05, "loss": 1.3592, "step": 1984 }, { "epoch": 0.18, "learning_rate": 4.647654198989708e-05, "loss": 1.3576, "step": 1985 }, { "epoch": 0.18, "learning_rate": 4.6472871671130924e-05, "loss": 1.3703, "step": 1986 }, { "epoch": 0.18, "learning_rate": 4.646919958679118e-05, "loss": 1.3256, "step": 1987 }, { "epoch": 0.18, "learning_rate": 4.646552573717977e-05, "loss": 1.3956, "step": 1988 }, { "epoch": 0.18, "learning_rate": 4.646185012259879e-05, "loss": 1.3173, "step": 1989 }, { "epoch": 0.18, "learning_rate": 4.6458172743350435e-05, "loss": 1.2775, "step": 1990 }, { "epoch": 0.18, "learning_rate": 4.6454493599737106e-05, "loss": 1.3437, "step": 1991 }, { "epoch": 0.18, "learning_rate": 4.645081269206128e-05, "loss": 1.3342, "step": 1992 }, { "epoch": 0.18, "learning_rate": 4.644713002062563e-05, "loss": 1.3122, "step": 1993 }, { "epoch": 0.18, "learning_rate": 4.6443445585732966e-05, "loss": 1.3794, "step": 1994 }, { "epoch": 0.18, "learning_rate": 4.6439759387686224e-05, "loss": 1.3468, "step": 1995 }, { "epoch": 0.18, "learning_rate": 4.6436071426788496e-05, "loss": 1.3173, "step": 1996 }, { "epoch": 0.18, "learning_rate": 4.643238170334302e-05, "loss": 1.345, "step": 1997 }, { "epoch": 0.18, "learning_rate": 4.6428690217653184e-05, "loss": 1.3522, "step": 1998 }, { "epoch": 0.18, "learning_rate": 4.6424996970022496e-05, "loss": 1.3596, "step": 1999 }, { "epoch": 0.18, "learning_rate": 4.6421301960754654e-05, "loss": 1.3038, "step": 2000 }, { "epoch": 0.18, "learning_rate": 4.6417605190153457e-05, "loss": 1.3244, "step": 2001 }, { "epoch": 0.18, "learning_rate": 4.6413906658522874e-05, "loss": 1.3665, "step": 2002 }, { "epoch": 0.18, "learning_rate": 4.641020636616701e-05, "loss": 1.3352, "step": 2003 }, { "epoch": 0.18, "learning_rate": 4.640650431339011e-05, "loss": 1.3498, "step": 2004 }, { "epoch": 0.18, "learning_rate": 4.6402800500496565e-05, "loss": 1.2894, "step": 2005 }, { "epoch": 0.18, "learning_rate": 4.639909492779093e-05, "loss": 1.3199, "step": 2006 }, { "epoch": 0.18, "learning_rate": 4.6395387595577885e-05, "loss": 1.2554, "step": 2007 }, { "epoch": 0.18, "learning_rate": 4.639167850416226e-05, "loss": 1.2832, "step": 2008 }, { "epoch": 0.18, "learning_rate": 4.638796765384903e-05, "loss": 1.326, "step": 2009 }, { "epoch": 0.18, "learning_rate": 4.63842550449433e-05, "loss": 1.3038, "step": 2010 }, { "epoch": 0.18, "learning_rate": 4.6380540677750354e-05, "loss": 1.3582, "step": 2011 }, { "epoch": 0.18, "learning_rate": 4.6376824552575596e-05, "loss": 1.2999, "step": 2012 }, { "epoch": 0.18, "learning_rate": 4.637310666972456e-05, "loss": 1.3637, "step": 2013 }, { "epoch": 0.18, "learning_rate": 4.6369387029502965e-05, "loss": 1.3293, "step": 2014 }, { "epoch": 0.18, "learning_rate": 4.636566563221664e-05, "loss": 1.3796, "step": 2015 }, { "epoch": 0.18, "learning_rate": 4.6361942478171575e-05, "loss": 1.332, "step": 2016 }, { "epoch": 0.18, "learning_rate": 4.63582175676739e-05, "loss": 1.3127, "step": 2017 }, { "epoch": 0.18, "learning_rate": 4.635449090102989e-05, "loss": 1.2751, "step": 2018 }, { "epoch": 0.18, "learning_rate": 4.635076247854598e-05, "loss": 1.3706, "step": 2019 }, { "epoch": 0.18, "learning_rate": 4.63470323005287e-05, "loss": 1.3122, "step": 2020 }, { "epoch": 0.18, "learning_rate": 4.634330036728479e-05, "loss": 1.3203, "step": 2021 }, { "epoch": 0.18, "learning_rate": 4.633956667912108e-05, "loss": 1.3586, "step": 2022 }, { "epoch": 0.18, "learning_rate": 4.633583123634459e-05, "loss": 1.3741, "step": 2023 }, { "epoch": 0.18, "learning_rate": 4.633209403926244e-05, "loss": 1.2839, "step": 2024 }, { "epoch": 0.18, "learning_rate": 4.6328355088181915e-05, "loss": 1.3018, "step": 2025 }, { "epoch": 0.18, "learning_rate": 4.6324614383410465e-05, "loss": 1.3179, "step": 2026 }, { "epoch": 0.18, "learning_rate": 4.632087192525564e-05, "loss": 1.2885, "step": 2027 }, { "epoch": 0.18, "learning_rate": 4.631712771402518e-05, "loss": 1.3777, "step": 2028 }, { "epoch": 0.18, "learning_rate": 4.631338175002693e-05, "loss": 1.3578, "step": 2029 }, { "epoch": 0.18, "learning_rate": 4.630963403356889e-05, "loss": 1.3445, "step": 2030 }, { "epoch": 0.18, "learning_rate": 4.630588456495924e-05, "loss": 1.3273, "step": 2031 }, { "epoch": 0.18, "learning_rate": 4.630213334450625e-05, "loss": 1.363, "step": 2032 }, { "epoch": 0.18, "learning_rate": 4.629838037251836e-05, "loss": 1.2944, "step": 2033 }, { "epoch": 0.18, "learning_rate": 4.6294625649304155e-05, "loss": 1.3018, "step": 2034 }, { "epoch": 0.18, "learning_rate": 4.629086917517237e-05, "loss": 1.2812, "step": 2035 }, { "epoch": 0.18, "learning_rate": 4.6287110950431865e-05, "loss": 1.3902, "step": 2036 }, { "epoch": 0.18, "learning_rate": 4.628335097539166e-05, "loss": 1.2859, "step": 2037 }, { "epoch": 0.18, "learning_rate": 4.627958925036091e-05, "loss": 1.3265, "step": 2038 }, { "epoch": 0.18, "learning_rate": 4.627582577564892e-05, "loss": 1.3319, "step": 2039 }, { "epoch": 0.18, "learning_rate": 4.627206055156512e-05, "loss": 1.3252, "step": 2040 }, { "epoch": 0.18, "learning_rate": 4.626829357841913e-05, "loss": 1.3544, "step": 2041 }, { "epoch": 0.18, "learning_rate": 4.6264524856520654e-05, "loss": 1.3225, "step": 2042 }, { "epoch": 0.18, "learning_rate": 4.626075438617959e-05, "loss": 1.3431, "step": 2043 }, { "epoch": 0.18, "learning_rate": 4.625698216770595e-05, "loss": 1.3152, "step": 2044 }, { "epoch": 0.18, "learning_rate": 4.6253208201409894e-05, "loss": 1.3082, "step": 2045 }, { "epoch": 0.18, "learning_rate": 4.624943248760174e-05, "loss": 1.3197, "step": 2046 }, { "epoch": 0.18, "learning_rate": 4.6245655026591936e-05, "loss": 1.3506, "step": 2047 }, { "epoch": 0.19, "learning_rate": 4.624187581869108e-05, "loss": 1.37, "step": 2048 }, { "epoch": 0.19, "learning_rate": 4.623809486420991e-05, "loss": 1.303, "step": 2049 }, { "epoch": 0.19, "learning_rate": 4.623431216345931e-05, "loss": 1.2685, "step": 2050 }, { "epoch": 0.19, "learning_rate": 4.6230527716750306e-05, "loss": 1.3406, "step": 2051 }, { "epoch": 0.19, "learning_rate": 4.622674152439407e-05, "loss": 1.2991, "step": 2052 }, { "epoch": 0.19, "learning_rate": 4.6222953586701903e-05, "loss": 1.2551, "step": 2053 }, { "epoch": 0.19, "learning_rate": 4.6219163903985286e-05, "loss": 1.3213, "step": 2054 }, { "epoch": 0.19, "learning_rate": 4.6215372476555794e-05, "loss": 1.403, "step": 2055 }, { "epoch": 0.19, "learning_rate": 4.62115793047252e-05, "loss": 1.3636, "step": 2056 }, { "epoch": 0.19, "learning_rate": 4.6207784388805366e-05, "loss": 1.3617, "step": 2057 }, { "epoch": 0.19, "learning_rate": 4.6203987729108334e-05, "loss": 1.3444, "step": 2058 }, { "epoch": 0.19, "learning_rate": 4.6200189325946276e-05, "loss": 1.338, "step": 2059 }, { "epoch": 0.19, "learning_rate": 4.619638917963152e-05, "loss": 1.3338, "step": 2060 }, { "epoch": 0.19, "learning_rate": 4.6192587290476505e-05, "loss": 1.3145, "step": 2061 }, { "epoch": 0.19, "learning_rate": 4.6188783658793855e-05, "loss": 1.3212, "step": 2062 }, { "epoch": 0.19, "learning_rate": 4.6184978284896305e-05, "loss": 1.3092, "step": 2063 }, { "epoch": 0.19, "learning_rate": 4.618117116909676e-05, "loss": 1.3254, "step": 2064 }, { "epoch": 0.19, "learning_rate": 4.617736231170825e-05, "loss": 1.3206, "step": 2065 }, { "epoch": 0.19, "learning_rate": 4.617355171304395e-05, "loss": 1.3196, "step": 2066 }, { "epoch": 0.19, "learning_rate": 4.616973937341718e-05, "loss": 1.3573, "step": 2067 }, { "epoch": 0.19, "learning_rate": 4.6165925293141386e-05, "loss": 1.3156, "step": 2068 }, { "epoch": 0.19, "learning_rate": 4.616210947253021e-05, "loss": 1.3267, "step": 2069 }, { "epoch": 0.19, "learning_rate": 4.615829191189738e-05, "loss": 1.3166, "step": 2070 }, { "epoch": 0.19, "learning_rate": 4.615447261155679e-05, "loss": 1.2886, "step": 2071 }, { "epoch": 0.19, "learning_rate": 4.615065157182248e-05, "loss": 1.285, "step": 2072 }, { "epoch": 0.19, "learning_rate": 4.614682879300864e-05, "loss": 1.3764, "step": 2073 }, { "epoch": 0.19, "learning_rate": 4.614300427542956e-05, "loss": 1.2454, "step": 2074 }, { "epoch": 0.19, "learning_rate": 4.613917801939974e-05, "loss": 1.301, "step": 2075 }, { "epoch": 0.19, "learning_rate": 4.6135350025233764e-05, "loss": 1.3154, "step": 2076 }, { "epoch": 0.19, "learning_rate": 4.61315202932464e-05, "loss": 1.3103, "step": 2077 }, { "epoch": 0.19, "learning_rate": 4.612768882375253e-05, "loss": 1.3064, "step": 2078 }, { "epoch": 0.19, "learning_rate": 4.612385561706719e-05, "loss": 1.3231, "step": 2079 }, { "epoch": 0.19, "learning_rate": 4.612002067350557e-05, "loss": 1.3133, "step": 2080 }, { "epoch": 0.19, "learning_rate": 4.6116183993382986e-05, "loss": 1.2588, "step": 2081 }, { "epoch": 0.19, "learning_rate": 4.61123455770149e-05, "loss": 1.3697, "step": 2082 }, { "epoch": 0.19, "learning_rate": 4.610850542471693e-05, "loss": 1.3442, "step": 2083 }, { "epoch": 0.19, "learning_rate": 4.610466353680482e-05, "loss": 1.2938, "step": 2084 }, { "epoch": 0.19, "learning_rate": 4.610081991359446e-05, "loss": 1.3152, "step": 2085 }, { "epoch": 0.19, "learning_rate": 4.609697455540189e-05, "loss": 1.3245, "step": 2086 }, { "epoch": 0.19, "learning_rate": 4.609312746254328e-05, "loss": 1.3079, "step": 2087 }, { "epoch": 0.19, "learning_rate": 4.608927863533497e-05, "loss": 1.3021, "step": 2088 }, { "epoch": 0.19, "learning_rate": 4.608542807409341e-05, "loss": 1.3548, "step": 2089 }, { "epoch": 0.19, "learning_rate": 4.6081575779135214e-05, "loss": 1.361, "step": 2090 }, { "epoch": 0.19, "learning_rate": 4.607772175077711e-05, "loss": 1.3586, "step": 2091 }, { "epoch": 0.19, "learning_rate": 4.607386598933602e-05, "loss": 1.3699, "step": 2092 }, { "epoch": 0.19, "learning_rate": 4.6070008495128955e-05, "loss": 1.3251, "step": 2093 }, { "epoch": 0.19, "learning_rate": 4.6066149268473107e-05, "loss": 1.2735, "step": 2094 }, { "epoch": 0.19, "learning_rate": 4.6062288309685784e-05, "loss": 1.3224, "step": 2095 }, { "epoch": 0.19, "learning_rate": 4.6058425619084456e-05, "loss": 1.2895, "step": 2096 }, { "epoch": 0.19, "learning_rate": 4.6054561196986725e-05, "loss": 1.3085, "step": 2097 }, { "epoch": 0.19, "learning_rate": 4.6050695043710325e-05, "loss": 1.2504, "step": 2098 }, { "epoch": 0.19, "learning_rate": 4.6046827159573155e-05, "loss": 1.3021, "step": 2099 }, { "epoch": 0.19, "learning_rate": 4.604295754489325e-05, "loss": 1.3306, "step": 2100 }, { "epoch": 0.19, "learning_rate": 4.603908619998878e-05, "loss": 1.292, "step": 2101 }, { "epoch": 0.19, "learning_rate": 4.603521312517806e-05, "loss": 1.2956, "step": 2102 }, { "epoch": 0.19, "learning_rate": 4.6031338320779534e-05, "loss": 1.3133, "step": 2103 }, { "epoch": 0.19, "learning_rate": 4.6027461787111817e-05, "loss": 1.2677, "step": 2104 }, { "epoch": 0.19, "learning_rate": 4.6023583524493654e-05, "loss": 1.34, "step": 2105 }, { "epoch": 0.19, "learning_rate": 4.601970353324391e-05, "loss": 1.354, "step": 2106 }, { "epoch": 0.19, "learning_rate": 4.601582181368164e-05, "loss": 1.3124, "step": 2107 }, { "epoch": 0.19, "learning_rate": 4.601193836612598e-05, "loss": 1.3342, "step": 2108 }, { "epoch": 0.19, "learning_rate": 4.6008053190896265e-05, "loss": 1.3131, "step": 2109 }, { "epoch": 0.19, "learning_rate": 4.600416628831194e-05, "loss": 1.3493, "step": 2110 }, { "epoch": 0.19, "learning_rate": 4.60002776586926e-05, "loss": 1.3844, "step": 2111 }, { "epoch": 0.19, "learning_rate": 4.5996387302357975e-05, "loss": 1.3154, "step": 2112 }, { "epoch": 0.19, "learning_rate": 4.599249521962795e-05, "loss": 1.3487, "step": 2113 }, { "epoch": 0.19, "learning_rate": 4.5988601410822545e-05, "loss": 1.413, "step": 2114 }, { "epoch": 0.19, "learning_rate": 4.598470587626193e-05, "loss": 1.3453, "step": 2115 }, { "epoch": 0.19, "learning_rate": 4.5980808616266384e-05, "loss": 1.3734, "step": 2116 }, { "epoch": 0.19, "learning_rate": 4.5976909631156384e-05, "loss": 1.3556, "step": 2117 }, { "epoch": 0.19, "learning_rate": 4.597300892125249e-05, "loss": 1.3334, "step": 2118 }, { "epoch": 0.19, "learning_rate": 4.5969106486875456e-05, "loss": 1.3294, "step": 2119 }, { "epoch": 0.19, "learning_rate": 4.5965202328346146e-05, "loss": 1.3253, "step": 2120 }, { "epoch": 0.19, "learning_rate": 4.5961296445985566e-05, "loss": 1.3664, "step": 2121 }, { "epoch": 0.19, "learning_rate": 4.595738884011488e-05, "loss": 1.3433, "step": 2122 }, { "epoch": 0.19, "learning_rate": 4.595347951105537e-05, "loss": 1.3014, "step": 2123 }, { "epoch": 0.19, "learning_rate": 4.59495684591285e-05, "loss": 1.3108, "step": 2124 }, { "epoch": 0.19, "learning_rate": 4.594565568465583e-05, "loss": 1.3054, "step": 2125 }, { "epoch": 0.19, "learning_rate": 4.5941741187959084e-05, "loss": 1.3303, "step": 2126 }, { "epoch": 0.19, "learning_rate": 4.5937824969360136e-05, "loss": 1.2643, "step": 2127 }, { "epoch": 0.19, "learning_rate": 4.593390702918098e-05, "loss": 1.3348, "step": 2128 }, { "epoch": 0.19, "learning_rate": 4.592998736774376e-05, "loss": 1.3074, "step": 2129 }, { "epoch": 0.19, "learning_rate": 4.592606598537077e-05, "loss": 1.3167, "step": 2130 }, { "epoch": 0.19, "learning_rate": 4.592214288238445e-05, "loss": 1.3061, "step": 2131 }, { "epoch": 0.19, "learning_rate": 4.5918218059107354e-05, "loss": 1.3797, "step": 2132 }, { "epoch": 0.19, "learning_rate": 4.5914291515862205e-05, "loss": 1.2759, "step": 2133 }, { "epoch": 0.19, "learning_rate": 4.591036325297186e-05, "loss": 1.423, "step": 2134 }, { "epoch": 0.19, "learning_rate": 4.5906433270759294e-05, "loss": 1.2795, "step": 2135 }, { "epoch": 0.19, "learning_rate": 4.590250156954767e-05, "loss": 1.2888, "step": 2136 }, { "epoch": 0.19, "learning_rate": 4.5898568149660246e-05, "loss": 1.3036, "step": 2137 }, { "epoch": 0.19, "learning_rate": 4.589463301142045e-05, "loss": 1.3074, "step": 2138 }, { "epoch": 0.19, "learning_rate": 4.5890696155151845e-05, "loss": 1.279, "step": 2139 }, { "epoch": 0.19, "learning_rate": 4.588675758117812e-05, "loss": 1.3354, "step": 2140 }, { "epoch": 0.19, "learning_rate": 4.588281728982314e-05, "loss": 1.3271, "step": 2141 }, { "epoch": 0.19, "learning_rate": 4.587887528141087e-05, "loss": 1.3605, "step": 2142 }, { "epoch": 0.19, "learning_rate": 4.5874931556265444e-05, "loss": 1.3172, "step": 2143 }, { "epoch": 0.19, "learning_rate": 4.587098611471112e-05, "loss": 1.3577, "step": 2144 }, { "epoch": 0.19, "learning_rate": 4.586703895707233e-05, "loss": 1.3182, "step": 2145 }, { "epoch": 0.19, "learning_rate": 4.586309008367359e-05, "loss": 1.3409, "step": 2146 }, { "epoch": 0.19, "learning_rate": 4.585913949483961e-05, "loss": 1.3633, "step": 2147 }, { "epoch": 0.19, "learning_rate": 4.5855187190895225e-05, "loss": 1.3346, "step": 2148 }, { "epoch": 0.19, "learning_rate": 4.58512331721654e-05, "loss": 1.3368, "step": 2149 }, { "epoch": 0.19, "learning_rate": 4.584727743897523e-05, "loss": 1.317, "step": 2150 }, { "epoch": 0.19, "learning_rate": 4.584331999165e-05, "loss": 1.3981, "step": 2151 }, { "epoch": 0.19, "learning_rate": 4.583936083051509e-05, "loss": 1.2616, "step": 2152 }, { "epoch": 0.19, "learning_rate": 4.583539995589603e-05, "loss": 1.3744, "step": 2153 }, { "epoch": 0.19, "learning_rate": 4.5831437368118504e-05, "loss": 1.3773, "step": 2154 }, { "epoch": 0.19, "learning_rate": 4.5827473067508344e-05, "loss": 1.3413, "step": 2155 }, { "epoch": 0.19, "learning_rate": 4.582350705439148e-05, "loss": 1.3513, "step": 2156 }, { "epoch": 0.19, "learning_rate": 4.581953932909403e-05, "loss": 1.262, "step": 2157 }, { "epoch": 0.19, "learning_rate": 4.5815569891942225e-05, "loss": 1.2905, "step": 2158 }, { "epoch": 0.2, "learning_rate": 4.5811598743262454e-05, "loss": 1.342, "step": 2159 }, { "epoch": 0.2, "learning_rate": 4.580762588338123e-05, "loss": 1.3316, "step": 2160 }, { "epoch": 0.2, "learning_rate": 4.580365131262523e-05, "loss": 1.3734, "step": 2161 }, { "epoch": 0.2, "learning_rate": 4.579967503132123e-05, "loss": 1.354, "step": 2162 }, { "epoch": 0.2, "learning_rate": 4.57956970397962e-05, "loss": 1.2776, "step": 2163 }, { "epoch": 0.2, "learning_rate": 4.579171733837722e-05, "loss": 1.3549, "step": 2164 }, { "epoch": 0.2, "learning_rate": 4.578773592739151e-05, "loss": 1.3538, "step": 2165 }, { "epoch": 0.2, "learning_rate": 4.578375280716642e-05, "loss": 1.3195, "step": 2166 }, { "epoch": 0.2, "learning_rate": 4.577976797802948e-05, "loss": 1.3113, "step": 2167 }, { "epoch": 0.2, "learning_rate": 4.577578144030833e-05, "loss": 1.2894, "step": 2168 }, { "epoch": 0.2, "learning_rate": 4.5771793194330746e-05, "loss": 1.3731, "step": 2169 }, { "epoch": 0.2, "learning_rate": 4.5767803240424665e-05, "loss": 1.3329, "step": 2170 }, { "epoch": 0.2, "learning_rate": 4.576381157891816e-05, "loss": 1.3131, "step": 2171 }, { "epoch": 0.2, "learning_rate": 4.5759818210139426e-05, "loss": 1.3462, "step": 2172 }, { "epoch": 0.2, "learning_rate": 4.575582313441682e-05, "loss": 1.2662, "step": 2173 }, { "epoch": 0.2, "learning_rate": 4.575182635207883e-05, "loss": 1.3427, "step": 2174 }, { "epoch": 0.2, "learning_rate": 4.574782786345407e-05, "loss": 1.3453, "step": 2175 }, { "epoch": 0.2, "learning_rate": 4.5743827668871345e-05, "loss": 1.3264, "step": 2176 }, { "epoch": 0.2, "learning_rate": 4.573982576865953e-05, "loss": 1.3386, "step": 2177 }, { "epoch": 0.2, "learning_rate": 4.573582216314769e-05, "loss": 1.3835, "step": 2178 }, { "epoch": 0.2, "learning_rate": 4.573181685266502e-05, "loss": 1.333, "step": 2179 }, { "epoch": 0.2, "learning_rate": 4.5727809837540836e-05, "loss": 1.302, "step": 2180 }, { "epoch": 0.2, "learning_rate": 4.5723801118104616e-05, "loss": 1.2668, "step": 2181 }, { "epoch": 0.2, "learning_rate": 4.571979069468597e-05, "loss": 1.3836, "step": 2182 }, { "epoch": 0.2, "learning_rate": 4.5715778567614654e-05, "loss": 1.313, "step": 2183 }, { "epoch": 0.2, "learning_rate": 4.571176473722055e-05, "loss": 1.3128, "step": 2184 }, { "epoch": 0.2, "learning_rate": 4.5707749203833705e-05, "loss": 1.3747, "step": 2185 }, { "epoch": 0.2, "learning_rate": 4.570373196778427e-05, "loss": 1.3421, "step": 2186 }, { "epoch": 0.2, "learning_rate": 4.569971302940257e-05, "loss": 1.3462, "step": 2187 }, { "epoch": 0.2, "learning_rate": 4.569569238901905e-05, "loss": 1.2911, "step": 2188 }, { "epoch": 0.2, "learning_rate": 4.56916700469643e-05, "loss": 1.3232, "step": 2189 }, { "epoch": 0.2, "learning_rate": 4.568764600356905e-05, "loss": 1.283, "step": 2190 }, { "epoch": 0.2, "learning_rate": 4.568362025916418e-05, "loss": 1.2924, "step": 2191 }, { "epoch": 0.2, "learning_rate": 4.567959281408069e-05, "loss": 1.3399, "step": 2192 }, { "epoch": 0.2, "learning_rate": 4.567556366864973e-05, "loss": 1.2864, "step": 2193 }, { "epoch": 0.2, "learning_rate": 4.567153282320261e-05, "loss": 1.3229, "step": 2194 }, { "epoch": 0.2, "learning_rate": 4.566750027807073e-05, "loss": 1.3507, "step": 2195 }, { "epoch": 0.2, "learning_rate": 4.566346603358568e-05, "loss": 1.2655, "step": 2196 }, { "epoch": 0.2, "learning_rate": 4.565943009007916e-05, "loss": 1.3212, "step": 2197 }, { "epoch": 0.2, "learning_rate": 4.5655392447883024e-05, "loss": 1.3225, "step": 2198 }, { "epoch": 0.2, "learning_rate": 4.565135310732927e-05, "loss": 1.3233, "step": 2199 }, { "epoch": 0.2, "learning_rate": 4.564731206875001e-05, "loss": 1.3171, "step": 2200 }, { "epoch": 0.2, "learning_rate": 4.564326933247752e-05, "loss": 1.3605, "step": 2201 }, { "epoch": 0.2, "learning_rate": 4.5639224898844205e-05, "loss": 1.2934, "step": 2202 }, { "epoch": 0.2, "learning_rate": 4.563517876818261e-05, "loss": 1.3437, "step": 2203 }, { "epoch": 0.2, "learning_rate": 4.563113094082543e-05, "loss": 1.32, "step": 2204 }, { "epoch": 0.2, "learning_rate": 4.562708141710549e-05, "loss": 1.3522, "step": 2205 }, { "epoch": 0.2, "learning_rate": 4.5623030197355745e-05, "loss": 1.2954, "step": 2206 }, { "epoch": 0.2, "learning_rate": 4.561897728190931e-05, "loss": 1.3034, "step": 2207 }, { "epoch": 0.2, "learning_rate": 4.561492267109943e-05, "loss": 1.2999, "step": 2208 }, { "epoch": 0.2, "learning_rate": 4.5610866365259487e-05, "loss": 1.3007, "step": 2209 }, { "epoch": 0.2, "learning_rate": 4.5606808364723e-05, "loss": 1.3243, "step": 2210 }, { "epoch": 0.2, "learning_rate": 4.5602748669823634e-05, "loss": 1.2867, "step": 2211 }, { "epoch": 0.2, "learning_rate": 4.55986872808952e-05, "loss": 1.3501, "step": 2212 }, { "epoch": 0.2, "learning_rate": 4.5594624198271634e-05, "loss": 1.3171, "step": 2213 }, { "epoch": 0.2, "learning_rate": 4.5590559422287013e-05, "loss": 1.346, "step": 2214 }, { "epoch": 0.2, "learning_rate": 4.5586492953275564e-05, "loss": 1.31, "step": 2215 }, { "epoch": 0.2, "learning_rate": 4.5582424791571634e-05, "loss": 1.3046, "step": 2216 }, { "epoch": 0.2, "learning_rate": 4.5578354937509725e-05, "loss": 1.3295, "step": 2217 }, { "epoch": 0.2, "learning_rate": 4.557428339142449e-05, "loss": 1.3002, "step": 2218 }, { "epoch": 0.2, "learning_rate": 4.5570210153650686e-05, "loss": 1.3294, "step": 2219 }, { "epoch": 0.2, "learning_rate": 4.556613522452323e-05, "loss": 1.3351, "step": 2220 }, { "epoch": 0.2, "learning_rate": 4.556205860437719e-05, "loss": 1.3577, "step": 2221 }, { "epoch": 0.2, "learning_rate": 4.555798029354775e-05, "loss": 1.3186, "step": 2222 }, { "epoch": 0.2, "learning_rate": 4.555390029237026e-05, "loss": 1.2862, "step": 2223 }, { "epoch": 0.2, "learning_rate": 4.554981860118016e-05, "loss": 1.2585, "step": 2224 }, { "epoch": 0.2, "learning_rate": 4.554573522031308e-05, "loss": 1.3372, "step": 2225 }, { "epoch": 0.2, "learning_rate": 4.554165015010478e-05, "loss": 1.3474, "step": 2226 }, { "epoch": 0.2, "learning_rate": 4.553756339089112e-05, "loss": 1.2725, "step": 2227 }, { "epoch": 0.2, "learning_rate": 4.5533474943008146e-05, "loss": 1.3232, "step": 2228 }, { "epoch": 0.2, "learning_rate": 4.552938480679203e-05, "loss": 1.3029, "step": 2229 }, { "epoch": 0.2, "learning_rate": 4.552529298257906e-05, "loss": 1.268, "step": 2230 }, { "epoch": 0.2, "learning_rate": 4.55211994707057e-05, "loss": 1.3625, "step": 2231 }, { "epoch": 0.2, "learning_rate": 4.551710427150851e-05, "loss": 1.3341, "step": 2232 }, { "epoch": 0.2, "learning_rate": 4.5513007385324234e-05, "loss": 1.3063, "step": 2233 }, { "epoch": 0.2, "learning_rate": 4.550890881248972e-05, "loss": 1.3155, "step": 2234 }, { "epoch": 0.2, "learning_rate": 4.5504808553341966e-05, "loss": 1.303, "step": 2235 }, { "epoch": 0.2, "learning_rate": 4.5500706608218106e-05, "loss": 1.3392, "step": 2236 }, { "epoch": 0.2, "learning_rate": 4.549660297745543e-05, "loss": 1.336, "step": 2237 }, { "epoch": 0.2, "learning_rate": 4.5492497661391335e-05, "loss": 1.3343, "step": 2238 }, { "epoch": 0.2, "learning_rate": 4.548839066036339e-05, "loss": 1.2885, "step": 2239 }, { "epoch": 0.2, "learning_rate": 4.5484281974709275e-05, "loss": 1.3208, "step": 2240 }, { "epoch": 0.2, "learning_rate": 4.548017160476683e-05, "loss": 1.3044, "step": 2241 }, { "epoch": 0.2, "learning_rate": 4.5476059550874014e-05, "loss": 1.3247, "step": 2242 }, { "epoch": 0.2, "learning_rate": 4.547194581336895e-05, "loss": 1.3142, "step": 2243 }, { "epoch": 0.2, "learning_rate": 4.5467830392589865e-05, "loss": 1.3105, "step": 2244 }, { "epoch": 0.2, "learning_rate": 4.546371328887515e-05, "loss": 1.295, "step": 2245 }, { "epoch": 0.2, "learning_rate": 4.545959450256334e-05, "loss": 1.2996, "step": 2246 }, { "epoch": 0.2, "learning_rate": 4.545547403399307e-05, "loss": 1.3418, "step": 2247 }, { "epoch": 0.2, "learning_rate": 4.5451351883503166e-05, "loss": 1.3781, "step": 2248 }, { "epoch": 0.2, "learning_rate": 4.544722805143254e-05, "loss": 1.3416, "step": 2249 }, { "epoch": 0.2, "learning_rate": 4.54431025381203e-05, "loss": 1.3606, "step": 2250 }, { "epoch": 0.2, "learning_rate": 4.5438975343905624e-05, "loss": 1.2714, "step": 2251 }, { "epoch": 0.2, "learning_rate": 4.543484646912789e-05, "loss": 1.3063, "step": 2252 }, { "epoch": 0.2, "learning_rate": 4.5430715914126576e-05, "loss": 1.3385, "step": 2253 }, { "epoch": 0.2, "learning_rate": 4.5426583679241314e-05, "loss": 1.2852, "step": 2254 }, { "epoch": 0.2, "learning_rate": 4.542244976481187e-05, "loss": 1.2999, "step": 2255 }, { "epoch": 0.2, "learning_rate": 4.541831417117815e-05, "loss": 1.2916, "step": 2256 }, { "epoch": 0.2, "learning_rate": 4.5414176898680196e-05, "loss": 1.3117, "step": 2257 }, { "epoch": 0.2, "learning_rate": 4.541003794765819e-05, "loss": 1.3227, "step": 2258 }, { "epoch": 0.2, "learning_rate": 4.540589731845244e-05, "loss": 1.3196, "step": 2259 }, { "epoch": 0.2, "learning_rate": 4.540175501140342e-05, "loss": 1.268, "step": 2260 }, { "epoch": 0.2, "learning_rate": 4.539761102685172e-05, "loss": 1.3326, "step": 2261 }, { "epoch": 0.2, "learning_rate": 4.5393465365138064e-05, "loss": 1.283, "step": 2262 }, { "epoch": 0.2, "learning_rate": 4.538931802660333e-05, "loss": 1.3318, "step": 2263 }, { "epoch": 0.2, "learning_rate": 4.538516901158854e-05, "loss": 1.3263, "step": 2264 }, { "epoch": 0.2, "learning_rate": 4.538101832043481e-05, "loss": 1.3266, "step": 2265 }, { "epoch": 0.2, "learning_rate": 4.537686595348344e-05, "loss": 1.2847, "step": 2266 }, { "epoch": 0.2, "learning_rate": 4.537271191107585e-05, "loss": 1.3052, "step": 2267 }, { "epoch": 0.2, "learning_rate": 4.53685561935536e-05, "loss": 1.2884, "step": 2268 }, { "epoch": 0.21, "learning_rate": 4.5364398801258396e-05, "loss": 1.2861, "step": 2269 }, { "epoch": 0.21, "learning_rate": 4.5360239734532064e-05, "loss": 1.3179, "step": 2270 }, { "epoch": 0.21, "learning_rate": 4.5356078993716574e-05, "loss": 1.2917, "step": 2271 }, { "epoch": 0.21, "learning_rate": 4.535191657915404e-05, "loss": 1.3478, "step": 2272 }, { "epoch": 0.21, "learning_rate": 4.534775249118672e-05, "loss": 1.3017, "step": 2273 }, { "epoch": 0.21, "learning_rate": 4.534358673015698e-05, "loss": 1.2866, "step": 2274 }, { "epoch": 0.21, "learning_rate": 4.533941929640736e-05, "loss": 1.3101, "step": 2275 }, { "epoch": 0.21, "learning_rate": 4.533525019028051e-05, "loss": 1.3744, "step": 2276 }, { "epoch": 0.21, "learning_rate": 4.5331079412119245e-05, "loss": 1.3173, "step": 2277 }, { "epoch": 0.21, "learning_rate": 4.532690696226648e-05, "loss": 1.3191, "step": 2278 }, { "epoch": 0.21, "learning_rate": 4.5322732841065295e-05, "loss": 1.3696, "step": 2279 }, { "epoch": 0.21, "learning_rate": 4.53185570488589e-05, "loss": 1.2824, "step": 2280 }, { "epoch": 0.21, "learning_rate": 4.531437958599066e-05, "loss": 1.2914, "step": 2281 }, { "epoch": 0.21, "learning_rate": 4.531020045280403e-05, "loss": 1.3498, "step": 2282 }, { "epoch": 0.21, "learning_rate": 4.530601964964266e-05, "loss": 1.3172, "step": 2283 }, { "epoch": 0.21, "learning_rate": 4.530183717685029e-05, "loss": 1.3635, "step": 2284 }, { "epoch": 0.21, "learning_rate": 4.5297653034770834e-05, "loss": 1.2958, "step": 2285 }, { "epoch": 0.21, "learning_rate": 4.529346722374832e-05, "loss": 1.3265, "step": 2286 }, { "epoch": 0.21, "learning_rate": 4.528927974412691e-05, "loss": 1.3041, "step": 2287 }, { "epoch": 0.21, "learning_rate": 4.5285090596250924e-05, "loss": 1.2851, "step": 2288 }, { "epoch": 0.21, "learning_rate": 4.528089978046481e-05, "loss": 1.339, "step": 2289 }, { "epoch": 0.21, "learning_rate": 4.527670729711315e-05, "loss": 1.2783, "step": 2290 }, { "epoch": 0.21, "learning_rate": 4.527251314654065e-05, "loss": 1.3531, "step": 2291 }, { "epoch": 0.21, "learning_rate": 4.526831732909219e-05, "loss": 1.3124, "step": 2292 }, { "epoch": 0.21, "learning_rate": 4.526411984511275e-05, "loss": 1.2661, "step": 2293 }, { "epoch": 0.21, "learning_rate": 4.525992069494747e-05, "loss": 1.3379, "step": 2294 }, { "epoch": 0.21, "learning_rate": 4.5255719878941606e-05, "loss": 1.3121, "step": 2295 }, { "epoch": 0.21, "learning_rate": 4.525151739744057e-05, "loss": 1.4082, "step": 2296 }, { "epoch": 0.21, "learning_rate": 4.524731325078991e-05, "loss": 1.2972, "step": 2297 }, { "epoch": 0.21, "learning_rate": 4.524310743933531e-05, "loss": 1.3339, "step": 2298 }, { "epoch": 0.21, "learning_rate": 4.523889996342257e-05, "loss": 1.3024, "step": 2299 }, { "epoch": 0.21, "learning_rate": 4.5234690823397647e-05, "loss": 1.3251, "step": 2300 }, { "epoch": 0.21, "learning_rate": 4.5230480019606636e-05, "loss": 1.2629, "step": 2301 }, { "epoch": 0.21, "learning_rate": 4.522626755239577e-05, "loss": 1.3299, "step": 2302 }, { "epoch": 0.21, "learning_rate": 4.52220534221114e-05, "loss": 1.3376, "step": 2303 }, { "epoch": 0.21, "learning_rate": 4.521783762910004e-05, "loss": 1.2745, "step": 2304 }, { "epoch": 0.21, "learning_rate": 4.52136201737083e-05, "loss": 1.2896, "step": 2305 }, { "epoch": 0.21, "learning_rate": 4.520940105628299e-05, "loss": 1.3548, "step": 2306 }, { "epoch": 0.21, "learning_rate": 4.5205180277171e-05, "loss": 1.3232, "step": 2307 }, { "epoch": 0.21, "learning_rate": 4.520095783671938e-05, "loss": 1.3272, "step": 2308 }, { "epoch": 0.21, "learning_rate": 4.519673373527531e-05, "loss": 1.2756, "step": 2309 }, { "epoch": 0.21, "learning_rate": 4.5192507973186115e-05, "loss": 1.3397, "step": 2310 }, { "epoch": 0.21, "learning_rate": 4.518828055079925e-05, "loss": 1.3006, "step": 2311 }, { "epoch": 0.21, "learning_rate": 4.5184051468462304e-05, "loss": 1.3745, "step": 2312 }, { "epoch": 0.21, "learning_rate": 4.517982072652303e-05, "loss": 1.3301, "step": 2313 }, { "epoch": 0.21, "learning_rate": 4.517558832532926e-05, "loss": 1.3427, "step": 2314 }, { "epoch": 0.21, "learning_rate": 4.517135426522901e-05, "loss": 1.333, "step": 2315 }, { "epoch": 0.21, "learning_rate": 4.516711854657043e-05, "loss": 1.3049, "step": 2316 }, { "epoch": 0.21, "learning_rate": 4.516288116970178e-05, "loss": 1.2945, "step": 2317 }, { "epoch": 0.21, "learning_rate": 4.515864213497149e-05, "loss": 1.2609, "step": 2318 }, { "epoch": 0.21, "learning_rate": 4.515440144272808e-05, "loss": 1.3009, "step": 2319 }, { "epoch": 0.21, "learning_rate": 4.5150159093320264e-05, "loss": 1.3072, "step": 2320 }, { "epoch": 0.21, "learning_rate": 4.5145915087096846e-05, "loss": 1.3887, "step": 2321 }, { "epoch": 0.21, "learning_rate": 4.5141669424406784e-05, "loss": 1.3458, "step": 2322 }, { "epoch": 0.21, "learning_rate": 4.513742210559918e-05, "loss": 1.2787, "step": 2323 }, { "epoch": 0.21, "learning_rate": 4.513317313102324e-05, "loss": 1.3478, "step": 2324 }, { "epoch": 0.21, "learning_rate": 4.512892250102836e-05, "loss": 1.3786, "step": 2325 }, { "epoch": 0.21, "learning_rate": 4.512467021596403e-05, "loss": 1.3292, "step": 2326 }, { "epoch": 0.21, "learning_rate": 4.5120416276179876e-05, "loss": 1.3027, "step": 2327 }, { "epoch": 0.21, "learning_rate": 4.511616068202569e-05, "loss": 1.286, "step": 2328 }, { "epoch": 0.21, "learning_rate": 4.5111903433851366e-05, "loss": 1.3191, "step": 2329 }, { "epoch": 0.21, "learning_rate": 4.5107644532006956e-05, "loss": 1.3167, "step": 2330 }, { "epoch": 0.21, "learning_rate": 4.5103383976842636e-05, "loss": 1.3263, "step": 2331 }, { "epoch": 0.21, "learning_rate": 4.509912176870874e-05, "loss": 1.3108, "step": 2332 }, { "epoch": 0.21, "learning_rate": 4.509485790795571e-05, "loss": 1.3033, "step": 2333 }, { "epoch": 0.21, "learning_rate": 4.509059239493413e-05, "loss": 1.3552, "step": 2334 }, { "epoch": 0.21, "learning_rate": 4.508632522999473e-05, "loss": 1.3314, "step": 2335 }, { "epoch": 0.21, "learning_rate": 4.508205641348837e-05, "loss": 1.3536, "step": 2336 }, { "epoch": 0.21, "learning_rate": 4.507778594576606e-05, "loss": 1.3031, "step": 2337 }, { "epoch": 0.21, "learning_rate": 4.507351382717891e-05, "loss": 1.3514, "step": 2338 }, { "epoch": 0.21, "learning_rate": 4.506924005807821e-05, "loss": 1.2995, "step": 2339 }, { "epoch": 0.21, "learning_rate": 4.506496463881535e-05, "loss": 1.2826, "step": 2340 }, { "epoch": 0.21, "learning_rate": 4.506068756974186e-05, "loss": 1.3408, "step": 2341 }, { "epoch": 0.21, "learning_rate": 4.505640885120945e-05, "loss": 1.3436, "step": 2342 }, { "epoch": 0.21, "learning_rate": 4.50521284835699e-05, "loss": 1.2895, "step": 2343 }, { "epoch": 0.21, "learning_rate": 4.504784646717517e-05, "loss": 1.306, "step": 2344 }, { "epoch": 0.21, "learning_rate": 4.504356280237733e-05, "loss": 1.3233, "step": 2345 }, { "epoch": 0.21, "learning_rate": 4.5039277489528616e-05, "loss": 1.2607, "step": 2346 }, { "epoch": 0.21, "learning_rate": 4.503499052898137e-05, "loss": 1.3339, "step": 2347 }, { "epoch": 0.21, "learning_rate": 4.5030701921088074e-05, "loss": 1.3081, "step": 2348 }, { "epoch": 0.21, "learning_rate": 4.502641166620137e-05, "loss": 1.3206, "step": 2349 }, { "epoch": 0.21, "learning_rate": 4.5022119764674005e-05, "loss": 1.3245, "step": 2350 }, { "epoch": 0.21, "learning_rate": 4.501782621685887e-05, "loss": 1.3167, "step": 2351 }, { "epoch": 0.21, "learning_rate": 4.5013531023109014e-05, "loss": 1.3424, "step": 2352 }, { "epoch": 0.21, "learning_rate": 4.500923418377758e-05, "loss": 1.2895, "step": 2353 }, { "epoch": 0.21, "learning_rate": 4.500493569921789e-05, "loss": 1.3523, "step": 2354 }, { "epoch": 0.21, "learning_rate": 4.500063556978337e-05, "loss": 1.3178, "step": 2355 }, { "epoch": 0.21, "learning_rate": 4.499633379582758e-05, "loss": 1.3157, "step": 2356 }, { "epoch": 0.21, "learning_rate": 4.4992030377704244e-05, "loss": 1.2899, "step": 2357 }, { "epoch": 0.21, "learning_rate": 4.4987725315767206e-05, "loss": 1.3682, "step": 2358 }, { "epoch": 0.21, "learning_rate": 4.498341861037042e-05, "loss": 1.3059, "step": 2359 }, { "epoch": 0.21, "learning_rate": 4.497911026186803e-05, "loss": 1.2731, "step": 2360 }, { "epoch": 0.21, "learning_rate": 4.4974800270614255e-05, "loss": 1.3051, "step": 2361 }, { "epoch": 0.21, "learning_rate": 4.497048863696349e-05, "loss": 1.2833, "step": 2362 }, { "epoch": 0.21, "learning_rate": 4.496617536127025e-05, "loss": 1.335, "step": 2363 }, { "epoch": 0.21, "learning_rate": 4.496186044388919e-05, "loss": 1.2976, "step": 2364 }, { "epoch": 0.21, "learning_rate": 4.49575438851751e-05, "loss": 1.2921, "step": 2365 }, { "epoch": 0.21, "learning_rate": 4.4953225685482904e-05, "loss": 1.3199, "step": 2366 }, { "epoch": 0.21, "learning_rate": 4.4948905845167645e-05, "loss": 1.3066, "step": 2367 }, { "epoch": 0.21, "learning_rate": 4.494458436458452e-05, "loss": 1.3408, "step": 2368 }, { "epoch": 0.21, "learning_rate": 4.494026124408887e-05, "loss": 1.2836, "step": 2369 }, { "epoch": 0.21, "learning_rate": 4.493593648403614e-05, "loss": 1.347, "step": 2370 }, { "epoch": 0.21, "learning_rate": 4.4931610084781936e-05, "loss": 1.3212, "step": 2371 }, { "epoch": 0.21, "learning_rate": 4.492728204668198e-05, "loss": 1.2551, "step": 2372 }, { "epoch": 0.21, "learning_rate": 4.492295237009216e-05, "loss": 1.2655, "step": 2373 }, { "epoch": 0.21, "learning_rate": 4.4918621055368455e-05, "loss": 1.3151, "step": 2374 }, { "epoch": 0.21, "learning_rate": 4.491428810286701e-05, "loss": 1.2651, "step": 2375 }, { "epoch": 0.21, "learning_rate": 4.4909953512944085e-05, "loss": 1.2834, "step": 2376 }, { "epoch": 0.21, "learning_rate": 4.49056172859561e-05, "loss": 1.2775, "step": 2377 }, { "epoch": 0.21, "learning_rate": 4.4901279422259584e-05, "loss": 1.307, "step": 2378 }, { "epoch": 0.21, "learning_rate": 4.4896939922211215e-05, "loss": 1.3327, "step": 2379 }, { "epoch": 0.22, "learning_rate": 4.489259878616781e-05, "loss": 1.3189, "step": 2380 }, { "epoch": 0.22, "learning_rate": 4.48882560144863e-05, "loss": 1.2983, "step": 2381 }, { "epoch": 0.22, "learning_rate": 4.488391160752375e-05, "loss": 1.2905, "step": 2382 }, { "epoch": 0.22, "learning_rate": 4.48795655656374e-05, "loss": 1.2976, "step": 2383 }, { "epoch": 0.22, "learning_rate": 4.487521788918459e-05, "loss": 1.3399, "step": 2384 }, { "epoch": 0.22, "learning_rate": 4.4870868578522795e-05, "loss": 1.3863, "step": 2385 }, { "epoch": 0.22, "learning_rate": 4.486651763400963e-05, "loss": 1.3098, "step": 2386 }, { "epoch": 0.22, "learning_rate": 4.486216505600284e-05, "loss": 1.354, "step": 2387 }, { "epoch": 0.22, "learning_rate": 4.485781084486032e-05, "loss": 1.3365, "step": 2388 }, { "epoch": 0.22, "learning_rate": 4.485345500094008e-05, "loss": 1.359, "step": 2389 }, { "epoch": 0.22, "learning_rate": 4.4849097524600283e-05, "loss": 1.3199, "step": 2390 }, { "epoch": 0.22, "learning_rate": 4.4844738416199205e-05, "loss": 1.2721, "step": 2391 }, { "epoch": 0.22, "learning_rate": 4.484037767609527e-05, "loss": 1.3055, "step": 2392 }, { "epoch": 0.22, "learning_rate": 4.483601530464703e-05, "loss": 1.269, "step": 2393 }, { "epoch": 0.22, "learning_rate": 4.4831651302213176e-05, "loss": 1.2938, "step": 2394 }, { "epoch": 0.22, "learning_rate": 4.482728566915254e-05, "loss": 1.3351, "step": 2395 }, { "epoch": 0.22, "learning_rate": 4.4822918405824066e-05, "loss": 1.3089, "step": 2396 }, { "epoch": 0.22, "learning_rate": 4.481854951258686e-05, "loss": 1.325, "step": 2397 }, { "epoch": 0.22, "learning_rate": 4.481417898980013e-05, "loss": 1.2789, "step": 2398 }, { "epoch": 0.22, "learning_rate": 4.4809806837823255e-05, "loss": 1.2757, "step": 2399 }, { "epoch": 0.22, "learning_rate": 4.480543305701571e-05, "loss": 1.3206, "step": 2400 }, { "epoch": 0.22, "learning_rate": 4.480105764773713e-05, "loss": 1.3226, "step": 2401 }, { "epoch": 0.22, "learning_rate": 4.479668061034728e-05, "loss": 1.342, "step": 2402 }, { "epoch": 0.22, "learning_rate": 4.479230194520605e-05, "loss": 1.3092, "step": 2403 }, { "epoch": 0.22, "learning_rate": 4.4787921652673475e-05, "loss": 1.3454, "step": 2404 }, { "epoch": 0.22, "learning_rate": 4.4783539733109715e-05, "loss": 1.2891, "step": 2405 }, { "epoch": 0.22, "learning_rate": 4.477915618687506e-05, "loss": 1.322, "step": 2406 }, { "epoch": 0.22, "learning_rate": 4.477477101432994e-05, "loss": 1.3482, "step": 2407 }, { "epoch": 0.22, "learning_rate": 4.4770384215834936e-05, "loss": 1.2895, "step": 2408 }, { "epoch": 0.22, "learning_rate": 4.476599579175072e-05, "loss": 1.328, "step": 2409 }, { "epoch": 0.22, "learning_rate": 4.476160574243816e-05, "loss": 1.3459, "step": 2410 }, { "epoch": 0.22, "learning_rate": 4.475721406825818e-05, "loss": 1.3169, "step": 2411 }, { "epoch": 0.22, "learning_rate": 4.4752820769571904e-05, "loss": 1.3222, "step": 2412 }, { "epoch": 0.22, "learning_rate": 4.474842584674056e-05, "loss": 1.2967, "step": 2413 }, { "epoch": 0.22, "learning_rate": 4.474402930012551e-05, "loss": 1.3306, "step": 2414 }, { "epoch": 0.22, "learning_rate": 4.473963113008825e-05, "loss": 1.3144, "step": 2415 }, { "epoch": 0.22, "learning_rate": 4.4735231336990414e-05, "loss": 1.3256, "step": 2416 }, { "epoch": 0.22, "learning_rate": 4.473082992119377e-05, "loss": 1.3411, "step": 2417 }, { "epoch": 0.22, "learning_rate": 4.472642688306024e-05, "loss": 1.3403, "step": 2418 }, { "epoch": 0.22, "learning_rate": 4.472202222295181e-05, "loss": 1.2842, "step": 2419 }, { "epoch": 0.22, "learning_rate": 4.471761594123068e-05, "loss": 1.3054, "step": 2420 }, { "epoch": 0.22, "learning_rate": 4.471320803825915e-05, "loss": 1.3347, "step": 2421 }, { "epoch": 0.22, "learning_rate": 4.470879851439964e-05, "loss": 1.2877, "step": 2422 }, { "epoch": 0.22, "learning_rate": 4.470438737001472e-05, "loss": 1.3769, "step": 2423 }, { "epoch": 0.22, "learning_rate": 4.469997460546709e-05, "loss": 1.2993, "step": 2424 }, { "epoch": 0.22, "learning_rate": 4.469556022111958e-05, "loss": 1.3082, "step": 2425 }, { "epoch": 0.22, "learning_rate": 4.4691144217335166e-05, "loss": 1.3111, "step": 2426 }, { "epoch": 0.22, "learning_rate": 4.468672659447693e-05, "loss": 1.3264, "step": 2427 }, { "epoch": 0.22, "learning_rate": 4.468230735290813e-05, "loss": 1.3061, "step": 2428 }, { "epoch": 0.22, "learning_rate": 4.4677886492992104e-05, "loss": 1.2729, "step": 2429 }, { "epoch": 0.22, "learning_rate": 4.4673464015092365e-05, "loss": 1.2807, "step": 2430 }, { "epoch": 0.22, "learning_rate": 4.466903991957254e-05, "loss": 1.3384, "step": 2431 }, { "epoch": 0.22, "learning_rate": 4.4664614206796403e-05, "loss": 1.3557, "step": 2432 }, { "epoch": 0.22, "learning_rate": 4.4660186877127836e-05, "loss": 1.3136, "step": 2433 }, { "epoch": 0.22, "learning_rate": 4.465575793093087e-05, "loss": 1.324, "step": 2434 }, { "epoch": 0.22, "learning_rate": 4.465132736856969e-05, "loss": 1.341, "step": 2435 }, { "epoch": 0.22, "learning_rate": 4.464689519040857e-05, "loss": 1.2685, "step": 2436 }, { "epoch": 0.22, "learning_rate": 4.464246139681195e-05, "loss": 1.3264, "step": 2437 }, { "epoch": 0.22, "learning_rate": 4.463802598814438e-05, "loss": 1.3252, "step": 2438 }, { "epoch": 0.22, "learning_rate": 4.463358896477056e-05, "loss": 1.322, "step": 2439 }, { "epoch": 0.22, "learning_rate": 4.4629150327055324e-05, "loss": 1.302, "step": 2440 }, { "epoch": 0.22, "learning_rate": 4.462471007536363e-05, "loss": 1.3135, "step": 2441 }, { "epoch": 0.22, "learning_rate": 4.462026821006057e-05, "loss": 1.3493, "step": 2442 }, { "epoch": 0.22, "learning_rate": 4.461582473151137e-05, "loss": 1.3326, "step": 2443 }, { "epoch": 0.22, "learning_rate": 4.461137964008137e-05, "loss": 1.323, "step": 2444 }, { "epoch": 0.22, "learning_rate": 4.460693293613609e-05, "loss": 1.3231, "step": 2445 }, { "epoch": 0.22, "learning_rate": 4.460248462004114e-05, "loss": 1.3658, "step": 2446 }, { "epoch": 0.22, "learning_rate": 4.4598034692162285e-05, "loss": 1.2967, "step": 2447 }, { "epoch": 0.22, "learning_rate": 4.459358315286539e-05, "loss": 1.3614, "step": 2448 }, { "epoch": 0.22, "learning_rate": 4.45891300025165e-05, "loss": 1.3121, "step": 2449 }, { "epoch": 0.22, "learning_rate": 4.458467524148175e-05, "loss": 1.3659, "step": 2450 }, { "epoch": 0.22, "learning_rate": 4.458021887012744e-05, "loss": 1.2767, "step": 2451 }, { "epoch": 0.22, "learning_rate": 4.457576088881998e-05, "loss": 1.2343, "step": 2452 }, { "epoch": 0.22, "learning_rate": 4.457130129792593e-05, "loss": 1.349, "step": 2453 }, { "epoch": 0.22, "learning_rate": 4.4566840097811954e-05, "loss": 1.3079, "step": 2454 }, { "epoch": 0.22, "learning_rate": 4.456237728884489e-05, "loss": 1.3407, "step": 2455 }, { "epoch": 0.22, "learning_rate": 4.455791287139168e-05, "loss": 1.2779, "step": 2456 }, { "epoch": 0.22, "learning_rate": 4.455344684581939e-05, "loss": 1.305, "step": 2457 }, { "epoch": 0.22, "learning_rate": 4.454897921249524e-05, "loss": 1.2788, "step": 2458 }, { "epoch": 0.22, "learning_rate": 4.454450997178659e-05, "loss": 1.3644, "step": 2459 }, { "epoch": 0.22, "learning_rate": 4.454003912406088e-05, "loss": 1.2995, "step": 2460 }, { "epoch": 0.22, "learning_rate": 4.453556666968576e-05, "loss": 1.326, "step": 2461 }, { "epoch": 0.22, "learning_rate": 4.453109260902894e-05, "loss": 1.3355, "step": 2462 }, { "epoch": 0.22, "learning_rate": 4.45266169424583e-05, "loss": 1.272, "step": 2463 }, { "epoch": 0.22, "learning_rate": 4.452213967034185e-05, "loss": 1.3111, "step": 2464 }, { "epoch": 0.22, "learning_rate": 4.4517660793047736e-05, "loss": 1.3136, "step": 2465 }, { "epoch": 0.22, "learning_rate": 4.4513180310944214e-05, "loss": 1.251, "step": 2466 }, { "epoch": 0.22, "learning_rate": 4.450869822439968e-05, "loss": 1.325, "step": 2467 }, { "epoch": 0.22, "learning_rate": 4.450421453378267e-05, "loss": 1.3047, "step": 2468 }, { "epoch": 0.22, "learning_rate": 4.449972923946186e-05, "loss": 1.359, "step": 2469 }, { "epoch": 0.22, "learning_rate": 4.449524234180604e-05, "loss": 1.2711, "step": 2470 }, { "epoch": 0.22, "learning_rate": 4.4490753841184126e-05, "loss": 1.3409, "step": 2471 }, { "epoch": 0.22, "learning_rate": 4.44862637379652e-05, "loss": 1.3437, "step": 2472 }, { "epoch": 0.22, "learning_rate": 4.448177203251843e-05, "loss": 1.2969, "step": 2473 }, { "epoch": 0.22, "learning_rate": 4.447727872521316e-05, "loss": 1.3052, "step": 2474 }, { "epoch": 0.22, "learning_rate": 4.447278381641884e-05, "loss": 1.2872, "step": 2475 }, { "epoch": 0.22, "learning_rate": 4.4468287306505045e-05, "loss": 1.2609, "step": 2476 }, { "epoch": 0.22, "learning_rate": 4.446378919584151e-05, "loss": 1.2448, "step": 2477 }, { "epoch": 0.22, "learning_rate": 4.4459289484798074e-05, "loss": 1.2676, "step": 2478 }, { "epoch": 0.22, "learning_rate": 4.445478817374472e-05, "loss": 1.3528, "step": 2479 }, { "epoch": 0.22, "learning_rate": 4.445028526305156e-05, "loss": 1.3264, "step": 2480 }, { "epoch": 0.22, "learning_rate": 4.444578075308885e-05, "loss": 1.2747, "step": 2481 }, { "epoch": 0.22, "learning_rate": 4.4441274644226954e-05, "loss": 1.345, "step": 2482 }, { "epoch": 0.22, "learning_rate": 4.443676693683639e-05, "loss": 1.3541, "step": 2483 }, { "epoch": 0.22, "learning_rate": 4.443225763128779e-05, "loss": 1.2834, "step": 2484 }, { "epoch": 0.22, "learning_rate": 4.4427746727951927e-05, "loss": 1.3229, "step": 2485 }, { "epoch": 0.22, "learning_rate": 4.442323422719969e-05, "loss": 1.3131, "step": 2486 }, { "epoch": 0.22, "learning_rate": 4.441872012940214e-05, "loss": 1.3192, "step": 2487 }, { "epoch": 0.22, "learning_rate": 4.441420443493042e-05, "loss": 1.2947, "step": 2488 }, { "epoch": 0.22, "learning_rate": 4.440968714415584e-05, "loss": 1.2793, "step": 2489 }, { "epoch": 0.22, "learning_rate": 4.4405168257449815e-05, "loss": 1.347, "step": 2490 }, { "epoch": 0.23, "learning_rate": 4.4400647775183904e-05, "loss": 1.3577, "step": 2491 }, { "epoch": 0.23, "learning_rate": 4.4396125697729806e-05, "loss": 1.3569, "step": 2492 }, { "epoch": 0.23, "learning_rate": 4.439160202545933e-05, "loss": 1.3118, "step": 2493 }, { "epoch": 0.23, "learning_rate": 4.438707675874445e-05, "loss": 1.289, "step": 2494 }, { "epoch": 0.23, "learning_rate": 4.438254989795722e-05, "loss": 1.2662, "step": 2495 }, { "epoch": 0.23, "learning_rate": 4.437802144346988e-05, "loss": 1.2947, "step": 2496 }, { "epoch": 0.23, "learning_rate": 4.437349139565475e-05, "loss": 1.3033, "step": 2497 }, { "epoch": 0.23, "learning_rate": 4.4368959754884324e-05, "loss": 1.3335, "step": 2498 }, { "epoch": 0.23, "learning_rate": 4.43644265215312e-05, "loss": 1.3058, "step": 2499 }, { "epoch": 0.23, "learning_rate": 4.435989169596813e-05, "loss": 1.2842, "step": 2500 }, { "epoch": 0.23, "learning_rate": 4.435535527856797e-05, "loss": 1.3034, "step": 2501 }, { "epoch": 0.23, "learning_rate": 4.4350817269703715e-05, "loss": 1.2973, "step": 2502 }, { "epoch": 0.23, "learning_rate": 4.4346277669748514e-05, "loss": 1.308, "step": 2503 }, { "epoch": 0.23, "learning_rate": 4.4341736479075616e-05, "loss": 1.2225, "step": 2504 }, { "epoch": 0.23, "learning_rate": 4.433719369805841e-05, "loss": 1.2964, "step": 2505 }, { "epoch": 0.23, "learning_rate": 4.433264932707043e-05, "loss": 1.2353, "step": 2506 }, { "epoch": 0.23, "learning_rate": 4.432810336648533e-05, "loss": 1.3163, "step": 2507 }, { "epoch": 0.23, "learning_rate": 4.432355581667688e-05, "loss": 1.2642, "step": 2508 }, { "epoch": 0.23, "learning_rate": 4.431900667801901e-05, "loss": 1.3478, "step": 2509 }, { "epoch": 0.23, "learning_rate": 4.431445595088576e-05, "loss": 1.2851, "step": 2510 }, { "epoch": 0.23, "learning_rate": 4.4309903635651304e-05, "loss": 1.3119, "step": 2511 }, { "epoch": 0.23, "learning_rate": 4.430534973268996e-05, "loss": 1.293, "step": 2512 }, { "epoch": 0.23, "learning_rate": 4.430079424237615e-05, "loss": 1.3419, "step": 2513 }, { "epoch": 0.23, "learning_rate": 4.429623716508445e-05, "loss": 1.3068, "step": 2514 }, { "epoch": 0.23, "learning_rate": 4.4291678501189566e-05, "loss": 1.3286, "step": 2515 }, { "epoch": 0.23, "learning_rate": 4.428711825106631e-05, "loss": 1.2754, "step": 2516 }, { "epoch": 0.23, "learning_rate": 4.428255641508966e-05, "loss": 1.2988, "step": 2517 }, { "epoch": 0.23, "learning_rate": 4.4277992993634696e-05, "loss": 1.2919, "step": 2518 }, { "epoch": 0.23, "learning_rate": 4.4273427987076634e-05, "loss": 1.3418, "step": 2519 }, { "epoch": 0.23, "learning_rate": 4.426886139579083e-05, "loss": 1.3129, "step": 2520 }, { "epoch": 0.23, "learning_rate": 4.4264293220152763e-05, "loss": 1.3332, "step": 2521 }, { "epoch": 0.23, "learning_rate": 4.425972346053805e-05, "loss": 1.2855, "step": 2522 }, { "epoch": 0.23, "learning_rate": 4.425515211732243e-05, "loss": 1.2843, "step": 2523 }, { "epoch": 0.23, "learning_rate": 4.425057919088177e-05, "loss": 1.3516, "step": 2524 }, { "epoch": 0.23, "learning_rate": 4.424600468159208e-05, "loss": 1.2742, "step": 2525 }, { "epoch": 0.23, "learning_rate": 4.424142858982948e-05, "loss": 1.3279, "step": 2526 }, { "epoch": 0.23, "learning_rate": 4.423685091597024e-05, "loss": 1.2879, "step": 2527 }, { "epoch": 0.23, "learning_rate": 4.4232271660390756e-05, "loss": 1.2974, "step": 2528 }, { "epoch": 0.23, "learning_rate": 4.422769082346754e-05, "loss": 1.3256, "step": 2529 }, { "epoch": 0.23, "learning_rate": 4.4223108405577246e-05, "loss": 1.3226, "step": 2530 }, { "epoch": 0.23, "learning_rate": 4.421852440709666e-05, "loss": 1.3422, "step": 2531 }, { "epoch": 0.23, "learning_rate": 4.42139388284027e-05, "loss": 1.3119, "step": 2532 }, { "epoch": 0.23, "learning_rate": 4.42093516698724e-05, "loss": 1.3203, "step": 2533 }, { "epoch": 0.23, "learning_rate": 4.4204762931882924e-05, "loss": 1.2913, "step": 2534 }, { "epoch": 0.23, "learning_rate": 4.4200172614811595e-05, "loss": 1.3736, "step": 2535 }, { "epoch": 0.23, "learning_rate": 4.419558071903583e-05, "loss": 1.312, "step": 2536 }, { "epoch": 0.23, "learning_rate": 4.41909872449332e-05, "loss": 1.2743, "step": 2537 }, { "epoch": 0.23, "learning_rate": 4.4186392192881375e-05, "loss": 1.3282, "step": 2538 }, { "epoch": 0.23, "learning_rate": 4.41817955632582e-05, "loss": 1.2698, "step": 2539 }, { "epoch": 0.23, "learning_rate": 4.417719735644162e-05, "loss": 1.2793, "step": 2540 }, { "epoch": 0.23, "learning_rate": 4.4172597572809706e-05, "loss": 1.3064, "step": 2541 }, { "epoch": 0.23, "learning_rate": 4.4167996212740684e-05, "loss": 1.2993, "step": 2542 }, { "epoch": 0.23, "learning_rate": 4.416339327661287e-05, "loss": 1.3279, "step": 2543 }, { "epoch": 0.23, "learning_rate": 4.415878876480476e-05, "loss": 1.2939, "step": 2544 }, { "epoch": 0.23, "learning_rate": 4.4154182677694936e-05, "loss": 1.3022, "step": 2545 }, { "epoch": 0.23, "learning_rate": 4.414957501566213e-05, "loss": 1.2981, "step": 2546 }, { "epoch": 0.23, "learning_rate": 4.414496577908519e-05, "loss": 1.3399, "step": 2547 }, { "epoch": 0.23, "learning_rate": 4.414035496834313e-05, "loss": 1.3364, "step": 2548 }, { "epoch": 0.23, "learning_rate": 4.413574258381504e-05, "loss": 1.3454, "step": 2549 }, { "epoch": 0.23, "learning_rate": 4.413112862588018e-05, "loss": 1.3048, "step": 2550 }, { "epoch": 0.23, "learning_rate": 4.412651309491792e-05, "loss": 1.3353, "step": 2551 }, { "epoch": 0.23, "learning_rate": 4.412189599130778e-05, "loss": 1.3622, "step": 2552 }, { "epoch": 0.23, "learning_rate": 4.4117277315429366e-05, "loss": 1.2949, "step": 2553 }, { "epoch": 0.23, "learning_rate": 4.411265706766247e-05, "loss": 1.2509, "step": 2554 }, { "epoch": 0.23, "learning_rate": 4.410803524838696e-05, "loss": 1.2907, "step": 2555 }, { "epoch": 0.23, "learning_rate": 4.4103411857982866e-05, "loss": 1.3073, "step": 2556 }, { "epoch": 0.23, "learning_rate": 4.409878689683036e-05, "loss": 1.2943, "step": 2557 }, { "epoch": 0.23, "learning_rate": 4.409416036530969e-05, "loss": 1.2923, "step": 2558 }, { "epoch": 0.23, "learning_rate": 4.408953226380128e-05, "loss": 1.3296, "step": 2559 }, { "epoch": 0.23, "learning_rate": 4.408490259268567e-05, "loss": 1.3075, "step": 2560 }, { "epoch": 0.23, "learning_rate": 4.408027135234354e-05, "loss": 1.2625, "step": 2561 }, { "epoch": 0.23, "learning_rate": 4.4075638543155655e-05, "loss": 1.3236, "step": 2562 }, { "epoch": 0.23, "learning_rate": 4.407100416550296e-05, "loss": 1.3484, "step": 2563 }, { "epoch": 0.23, "learning_rate": 4.406636821976652e-05, "loss": 1.3373, "step": 2564 }, { "epoch": 0.23, "learning_rate": 4.40617307063275e-05, "loss": 1.2592, "step": 2565 }, { "epoch": 0.23, "learning_rate": 4.4057091625567216e-05, "loss": 1.2888, "step": 2566 }, { "epoch": 0.23, "learning_rate": 4.4052450977867115e-05, "loss": 1.38, "step": 2567 }, { "epoch": 0.23, "learning_rate": 4.4047808763608764e-05, "loss": 1.2491, "step": 2568 }, { "epoch": 0.23, "learning_rate": 4.4043164983173856e-05, "loss": 1.2582, "step": 2569 }, { "epoch": 0.23, "learning_rate": 4.403851963694423e-05, "loss": 1.3305, "step": 2570 }, { "epoch": 0.23, "learning_rate": 4.403387272530184e-05, "loss": 1.3406, "step": 2571 }, { "epoch": 0.23, "learning_rate": 4.402922424862876e-05, "loss": 1.3092, "step": 2572 }, { "epoch": 0.23, "learning_rate": 4.4024574207307215e-05, "loss": 1.3356, "step": 2573 }, { "epoch": 0.23, "learning_rate": 4.401992260171954e-05, "loss": 1.3236, "step": 2574 }, { "epoch": 0.23, "learning_rate": 4.401526943224822e-05, "loss": 1.2732, "step": 2575 }, { "epoch": 0.23, "learning_rate": 4.401061469927584e-05, "loss": 1.293, "step": 2576 }, { "epoch": 0.23, "learning_rate": 4.400595840318513e-05, "loss": 1.2171, "step": 2577 }, { "epoch": 0.23, "learning_rate": 4.400130054435895e-05, "loss": 1.3138, "step": 2578 }, { "epoch": 0.23, "learning_rate": 4.399664112318029e-05, "loss": 1.3153, "step": 2579 }, { "epoch": 0.23, "learning_rate": 4.399198014003226e-05, "loss": 1.2724, "step": 2580 }, { "epoch": 0.23, "learning_rate": 4.3987317595298094e-05, "loss": 1.3018, "step": 2581 }, { "epoch": 0.23, "learning_rate": 4.398265348936117e-05, "loss": 1.293, "step": 2582 }, { "epoch": 0.23, "learning_rate": 4.397798782260499e-05, "loss": 1.3381, "step": 2583 }, { "epoch": 0.23, "learning_rate": 4.397332059541317e-05, "loss": 1.3108, "step": 2584 }, { "epoch": 0.23, "learning_rate": 4.396865180816948e-05, "loss": 1.3127, "step": 2585 }, { "epoch": 0.23, "learning_rate": 4.39639814612578e-05, "loss": 1.3099, "step": 2586 }, { "epoch": 0.23, "learning_rate": 4.3959309555062136e-05, "loss": 1.2974, "step": 2587 }, { "epoch": 0.23, "learning_rate": 4.395463608996663e-05, "loss": 1.3164, "step": 2588 }, { "epoch": 0.23, "learning_rate": 4.394996106635555e-05, "loss": 1.2611, "step": 2589 }, { "epoch": 0.23, "learning_rate": 4.394528448461329e-05, "loss": 1.3051, "step": 2590 }, { "epoch": 0.23, "learning_rate": 4.394060634512438e-05, "loss": 1.3123, "step": 2591 }, { "epoch": 0.23, "learning_rate": 4.393592664827347e-05, "loss": 1.2986, "step": 2592 }, { "epoch": 0.23, "learning_rate": 4.3931245394445354e-05, "loss": 1.2654, "step": 2593 }, { "epoch": 0.23, "learning_rate": 4.3926562584024924e-05, "loss": 1.2544, "step": 2594 }, { "epoch": 0.23, "learning_rate": 4.3921878217397217e-05, "loss": 1.2936, "step": 2595 }, { "epoch": 0.23, "learning_rate": 4.391719229494741e-05, "loss": 1.308, "step": 2596 }, { "epoch": 0.23, "learning_rate": 4.391250481706078e-05, "loss": 1.3297, "step": 2597 }, { "epoch": 0.23, "learning_rate": 4.390781578412276e-05, "loss": 1.3485, "step": 2598 }, { "epoch": 0.23, "learning_rate": 4.390312519651889e-05, "loss": 1.3344, "step": 2599 }, { "epoch": 0.23, "learning_rate": 4.3898433054634855e-05, "loss": 1.3213, "step": 2600 }, { "epoch": 0.24, "learning_rate": 4.389373935885646e-05, "loss": 1.2644, "step": 2601 }, { "epoch": 0.24, "learning_rate": 4.388904410956962e-05, "loss": 1.2751, "step": 2602 }, { "epoch": 0.24, "learning_rate": 4.388434730716042e-05, "loss": 1.2876, "step": 2603 }, { "epoch": 0.24, "learning_rate": 4.387964895201503e-05, "loss": 1.3127, "step": 2604 }, { "epoch": 0.24, "learning_rate": 4.3874949044519763e-05, "loss": 1.2917, "step": 2605 }, { "epoch": 0.24, "learning_rate": 4.387024758506107e-05, "loss": 1.3292, "step": 2606 }, { "epoch": 0.24, "learning_rate": 4.3865544574025517e-05, "loss": 1.2405, "step": 2607 }, { "epoch": 0.24, "learning_rate": 4.386084001179981e-05, "loss": 1.2561, "step": 2608 }, { "epoch": 0.24, "learning_rate": 4.3856133898770755e-05, "loss": 1.3833, "step": 2609 }, { "epoch": 0.24, "learning_rate": 4.3851426235325325e-05, "loss": 1.305, "step": 2610 }, { "epoch": 0.24, "learning_rate": 4.384671702185059e-05, "loss": 1.2418, "step": 2611 }, { "epoch": 0.24, "learning_rate": 4.384200625873377e-05, "loss": 1.2791, "step": 2612 }, { "epoch": 0.24, "learning_rate": 4.383729394636218e-05, "loss": 1.2859, "step": 2613 }, { "epoch": 0.24, "learning_rate": 4.38325800851233e-05, "loss": 1.2805, "step": 2614 }, { "epoch": 0.24, "learning_rate": 4.382786467540471e-05, "loss": 1.3133, "step": 2615 }, { "epoch": 0.24, "learning_rate": 4.382314771759413e-05, "loss": 1.2485, "step": 2616 }, { "epoch": 0.24, "learning_rate": 4.3818429212079404e-05, "loss": 1.3087, "step": 2617 }, { "epoch": 0.24, "learning_rate": 4.3813709159248504e-05, "loss": 1.3005, "step": 2618 }, { "epoch": 0.24, "learning_rate": 4.380898755948953e-05, "loss": 1.3278, "step": 2619 }, { "epoch": 0.24, "learning_rate": 4.380426441319071e-05, "loss": 1.2979, "step": 2620 }, { "epoch": 0.24, "learning_rate": 4.37995397207404e-05, "loss": 1.2423, "step": 2621 }, { "epoch": 0.24, "learning_rate": 4.3794813482527076e-05, "loss": 1.3099, "step": 2622 }, { "epoch": 0.24, "learning_rate": 4.3790085698939345e-05, "loss": 1.2621, "step": 2623 }, { "epoch": 0.24, "learning_rate": 4.378535637036593e-05, "loss": 1.2989, "step": 2624 }, { "epoch": 0.24, "learning_rate": 4.3780625497195716e-05, "loss": 1.2727, "step": 2625 }, { "epoch": 0.24, "learning_rate": 4.377589307981767e-05, "loss": 1.3071, "step": 2626 }, { "epoch": 0.24, "learning_rate": 4.377115911862093e-05, "loss": 1.2857, "step": 2627 }, { "epoch": 0.24, "learning_rate": 4.376642361399473e-05, "loss": 1.3119, "step": 2628 }, { "epoch": 0.24, "learning_rate": 4.376168656632843e-05, "loss": 1.2942, "step": 2629 }, { "epoch": 0.24, "learning_rate": 4.375694797601153e-05, "loss": 1.2749, "step": 2630 }, { "epoch": 0.24, "learning_rate": 4.3752207843433664e-05, "loss": 1.3353, "step": 2631 }, { "epoch": 0.24, "learning_rate": 4.374746616898457e-05, "loss": 1.2944, "step": 2632 }, { "epoch": 0.24, "learning_rate": 4.374272295305413e-05, "loss": 1.2891, "step": 2633 }, { "epoch": 0.24, "learning_rate": 4.3737978196032345e-05, "loss": 1.3259, "step": 2634 }, { "epoch": 0.24, "learning_rate": 4.373323189830935e-05, "loss": 1.3349, "step": 2635 }, { "epoch": 0.24, "learning_rate": 4.372848406027539e-05, "loss": 1.3019, "step": 2636 }, { "epoch": 0.24, "learning_rate": 4.372373468232086e-05, "loss": 1.314, "step": 2637 }, { "epoch": 0.24, "learning_rate": 4.371898376483627e-05, "loss": 1.2895, "step": 2638 }, { "epoch": 0.24, "learning_rate": 4.3714231308212264e-05, "loss": 1.2319, "step": 2639 }, { "epoch": 0.24, "learning_rate": 4.370947731283959e-05, "loss": 1.301, "step": 2640 }, { "epoch": 0.24, "learning_rate": 4.370472177910914e-05, "loss": 1.3343, "step": 2641 }, { "epoch": 0.24, "learning_rate": 4.369996470741193e-05, "loss": 1.3259, "step": 2642 }, { "epoch": 0.24, "learning_rate": 4.369520609813912e-05, "loss": 1.2562, "step": 2643 }, { "epoch": 0.24, "learning_rate": 4.369044595168196e-05, "loss": 1.3049, "step": 2644 }, { "epoch": 0.24, "learning_rate": 4.3685684268431846e-05, "loss": 1.3024, "step": 2645 }, { "epoch": 0.24, "learning_rate": 4.368092104878032e-05, "loss": 1.2919, "step": 2646 }, { "epoch": 0.24, "learning_rate": 4.367615629311901e-05, "loss": 1.3651, "step": 2647 }, { "epoch": 0.24, "learning_rate": 4.367139000183969e-05, "loss": 1.2747, "step": 2648 }, { "epoch": 0.24, "learning_rate": 4.366662217533427e-05, "loss": 1.3298, "step": 2649 }, { "epoch": 0.24, "learning_rate": 4.3661852813994784e-05, "loss": 1.3076, "step": 2650 }, { "epoch": 0.24, "learning_rate": 4.3657081918213375e-05, "loss": 1.2732, "step": 2651 }, { "epoch": 0.24, "learning_rate": 4.365230948838232e-05, "loss": 1.2826, "step": 2652 }, { "epoch": 0.24, "learning_rate": 4.3647535524894026e-05, "loss": 1.3111, "step": 2653 }, { "epoch": 0.24, "learning_rate": 4.364276002814103e-05, "loss": 1.275, "step": 2654 }, { "epoch": 0.24, "learning_rate": 4.363798299851598e-05, "loss": 1.2903, "step": 2655 }, { "epoch": 0.24, "learning_rate": 4.363320443641168e-05, "loss": 1.2689, "step": 2656 }, { "epoch": 0.24, "learning_rate": 4.362842434222102e-05, "loss": 1.356, "step": 2657 }, { "epoch": 0.24, "learning_rate": 4.3623642716337046e-05, "loss": 1.3171, "step": 2658 }, { "epoch": 0.24, "learning_rate": 4.3618859559152914e-05, "loss": 1.3213, "step": 2659 }, { "epoch": 0.24, "learning_rate": 4.361407487106192e-05, "loss": 1.2863, "step": 2660 }, { "epoch": 0.24, "learning_rate": 4.360928865245746e-05, "loss": 1.25, "step": 2661 }, { "epoch": 0.24, "learning_rate": 4.3604500903733095e-05, "loss": 1.2862, "step": 2662 }, { "epoch": 0.24, "learning_rate": 4.359971162528247e-05, "loss": 1.2959, "step": 2663 }, { "epoch": 0.24, "learning_rate": 4.3594920817499395e-05, "loss": 1.3112, "step": 2664 }, { "epoch": 0.24, "learning_rate": 4.359012848077778e-05, "loss": 1.3274, "step": 2665 }, { "epoch": 0.24, "learning_rate": 4.358533461551166e-05, "loss": 1.2566, "step": 2666 }, { "epoch": 0.24, "learning_rate": 4.358053922209521e-05, "loss": 1.2972, "step": 2667 }, { "epoch": 0.24, "learning_rate": 4.357574230092273e-05, "loss": 1.3408, "step": 2668 }, { "epoch": 0.24, "learning_rate": 4.3570943852388626e-05, "loss": 1.3076, "step": 2669 }, { "epoch": 0.24, "learning_rate": 4.3566143876887444e-05, "loss": 1.3452, "step": 2670 }, { "epoch": 0.24, "learning_rate": 4.356134237481386e-05, "loss": 1.2847, "step": 2671 }, { "epoch": 0.24, "learning_rate": 4.355653934656267e-05, "loss": 1.2663, "step": 2672 }, { "epoch": 0.24, "learning_rate": 4.355173479252881e-05, "loss": 1.2791, "step": 2673 }, { "epoch": 0.24, "learning_rate": 4.354692871310729e-05, "loss": 1.3079, "step": 2674 }, { "epoch": 0.24, "learning_rate": 4.354212110869332e-05, "loss": 1.3001, "step": 2675 }, { "epoch": 0.24, "learning_rate": 4.353731197968218e-05, "loss": 1.2879, "step": 2676 }, { "epoch": 0.24, "learning_rate": 4.353250132646929e-05, "loss": 1.2804, "step": 2677 }, { "epoch": 0.24, "learning_rate": 4.352768914945021e-05, "loss": 1.2769, "step": 2678 }, { "epoch": 0.24, "learning_rate": 4.35228754490206e-05, "loss": 1.2903, "step": 2679 }, { "epoch": 0.24, "learning_rate": 4.3518060225576275e-05, "loss": 1.3072, "step": 2680 }, { "epoch": 0.24, "learning_rate": 4.351324347951316e-05, "loss": 1.2973, "step": 2681 }, { "epoch": 0.24, "learning_rate": 4.3508425211227276e-05, "loss": 1.2978, "step": 2682 }, { "epoch": 0.24, "learning_rate": 4.3503605421114825e-05, "loss": 1.2941, "step": 2683 }, { "epoch": 0.24, "learning_rate": 4.34987841095721e-05, "loss": 1.2437, "step": 2684 }, { "epoch": 0.24, "learning_rate": 4.349396127699552e-05, "loss": 1.2975, "step": 2685 }, { "epoch": 0.24, "learning_rate": 4.348913692378165e-05, "loss": 1.2457, "step": 2686 }, { "epoch": 0.24, "learning_rate": 4.348431105032714e-05, "loss": 1.2828, "step": 2687 }, { "epoch": 0.24, "learning_rate": 4.3479483657028817e-05, "loss": 1.2385, "step": 2688 }, { "epoch": 0.24, "learning_rate": 4.3474654744283595e-05, "loss": 1.3085, "step": 2689 }, { "epoch": 0.24, "learning_rate": 4.3469824312488514e-05, "loss": 1.3015, "step": 2690 }, { "epoch": 0.24, "learning_rate": 4.346499236204076e-05, "loss": 1.2814, "step": 2691 }, { "epoch": 0.24, "learning_rate": 4.346015889333763e-05, "loss": 1.2967, "step": 2692 }, { "epoch": 0.24, "learning_rate": 4.345532390677655e-05, "loss": 1.3116, "step": 2693 }, { "epoch": 0.24, "learning_rate": 4.345048740275507e-05, "loss": 1.2852, "step": 2694 }, { "epoch": 0.24, "learning_rate": 4.344564938167086e-05, "loss": 1.3344, "step": 2695 }, { "epoch": 0.24, "learning_rate": 4.3440809843921725e-05, "loss": 1.3082, "step": 2696 }, { "epoch": 0.24, "learning_rate": 4.343596878990557e-05, "loss": 1.2928, "step": 2697 }, { "epoch": 0.24, "learning_rate": 4.343112622002047e-05, "loss": 1.317, "step": 2698 }, { "epoch": 0.24, "learning_rate": 4.342628213466459e-05, "loss": 1.2918, "step": 2699 }, { "epoch": 0.24, "learning_rate": 4.3421436534236205e-05, "loss": 1.3418, "step": 2700 }, { "epoch": 0.24, "learning_rate": 4.3416589419133766e-05, "loss": 1.305, "step": 2701 }, { "epoch": 0.24, "learning_rate": 4.3411740789755806e-05, "loss": 1.2798, "step": 2702 }, { "epoch": 0.24, "learning_rate": 4.3406890646501e-05, "loss": 1.2841, "step": 2703 }, { "epoch": 0.24, "learning_rate": 4.3402038989768136e-05, "loss": 1.2804, "step": 2704 }, { "epoch": 0.24, "learning_rate": 4.339718581995614e-05, "loss": 1.2646, "step": 2705 }, { "epoch": 0.24, "learning_rate": 4.339233113746406e-05, "loss": 1.2833, "step": 2706 }, { "epoch": 0.24, "learning_rate": 4.3387474942691054e-05, "loss": 1.2941, "step": 2707 }, { "epoch": 0.24, "learning_rate": 4.3382617236036424e-05, "loss": 1.3285, "step": 2708 }, { "epoch": 0.24, "learning_rate": 4.33777580178996e-05, "loss": 1.2714, "step": 2709 }, { "epoch": 0.24, "learning_rate": 4.3372897288680095e-05, "loss": 1.3072, "step": 2710 }, { "epoch": 0.24, "learning_rate": 4.336803504877758e-05, "loss": 1.3046, "step": 2711 }, { "epoch": 0.25, "learning_rate": 4.3363171298591866e-05, "loss": 1.2695, "step": 2712 }, { "epoch": 0.25, "learning_rate": 4.335830603852285e-05, "loss": 1.2989, "step": 2713 }, { "epoch": 0.25, "learning_rate": 4.335343926897058e-05, "loss": 1.3497, "step": 2714 }, { "epoch": 0.25, "learning_rate": 4.334857099033521e-05, "loss": 1.3167, "step": 2715 }, { "epoch": 0.25, "learning_rate": 4.334370120301704e-05, "loss": 1.3421, "step": 2716 }, { "epoch": 0.25, "learning_rate": 4.333882990741646e-05, "loss": 1.3383, "step": 2717 }, { "epoch": 0.25, "learning_rate": 4.333395710393402e-05, "loss": 1.2611, "step": 2718 }, { "epoch": 0.25, "learning_rate": 4.3329082792970374e-05, "loss": 1.2977, "step": 2719 }, { "epoch": 0.25, "learning_rate": 4.3324206974926314e-05, "loss": 1.3142, "step": 2720 }, { "epoch": 0.25, "learning_rate": 4.3319329650202734e-05, "loss": 1.3111, "step": 2721 }, { "epoch": 0.25, "learning_rate": 4.331445081920066e-05, "loss": 1.3205, "step": 2722 }, { "epoch": 0.25, "learning_rate": 4.330957048232126e-05, "loss": 1.2729, "step": 2723 }, { "epoch": 0.25, "learning_rate": 4.3304688639965816e-05, "loss": 1.2944, "step": 2724 }, { "epoch": 0.25, "learning_rate": 4.329980529253571e-05, "loss": 1.3317, "step": 2725 }, { "epoch": 0.25, "learning_rate": 4.3294920440432486e-05, "loss": 1.3161, "step": 2726 }, { "epoch": 0.25, "learning_rate": 4.329003408405778e-05, "loss": 1.3133, "step": 2727 }, { "epoch": 0.25, "learning_rate": 4.3285146223813375e-05, "loss": 1.3486, "step": 2728 }, { "epoch": 0.25, "learning_rate": 4.328025686010117e-05, "loss": 1.2803, "step": 2729 }, { "epoch": 0.25, "learning_rate": 4.327536599332317e-05, "loss": 1.3263, "step": 2730 }, { "epoch": 0.25, "learning_rate": 4.3270473623881526e-05, "loss": 1.3486, "step": 2731 }, { "epoch": 0.25, "learning_rate": 4.326557975217851e-05, "loss": 1.3222, "step": 2732 }, { "epoch": 0.25, "learning_rate": 4.326068437861651e-05, "loss": 1.3039, "step": 2733 }, { "epoch": 0.25, "learning_rate": 4.325578750359805e-05, "loss": 1.2528, "step": 2734 }, { "epoch": 0.25, "learning_rate": 4.325088912752575e-05, "loss": 1.2609, "step": 2735 }, { "epoch": 0.25, "learning_rate": 4.3245989250802385e-05, "loss": 1.3361, "step": 2736 }, { "epoch": 0.25, "learning_rate": 4.324108787383083e-05, "loss": 1.2571, "step": 2737 }, { "epoch": 0.25, "learning_rate": 4.3236184997014104e-05, "loss": 1.3206, "step": 2738 }, { "epoch": 0.25, "learning_rate": 4.323128062075534e-05, "loss": 1.2486, "step": 2739 }, { "epoch": 0.25, "learning_rate": 4.322637474545778e-05, "loss": 1.3464, "step": 2740 }, { "epoch": 0.25, "learning_rate": 4.32214673715248e-05, "loss": 1.2755, "step": 2741 }, { "epoch": 0.25, "learning_rate": 4.321655849935991e-05, "loss": 1.297, "step": 2742 }, { "epoch": 0.25, "learning_rate": 4.321164812936674e-05, "loss": 1.2934, "step": 2743 }, { "epoch": 0.25, "learning_rate": 4.320673626194903e-05, "loss": 1.2954, "step": 2744 }, { "epoch": 0.25, "learning_rate": 4.3201822897510646e-05, "loss": 1.2789, "step": 2745 }, { "epoch": 0.25, "learning_rate": 4.3196908036455595e-05, "loss": 1.2556, "step": 2746 }, { "epoch": 0.25, "learning_rate": 4.319199167918798e-05, "loss": 1.2597, "step": 2747 }, { "epoch": 0.25, "learning_rate": 4.318707382611205e-05, "loss": 1.326, "step": 2748 }, { "epoch": 0.25, "learning_rate": 4.3182154477632164e-05, "loss": 1.2825, "step": 2749 }, { "epoch": 0.25, "learning_rate": 4.317723363415281e-05, "loss": 1.3152, "step": 2750 }, { "epoch": 0.25, "learning_rate": 4.3172311296078595e-05, "loss": 1.2904, "step": 2751 }, { "epoch": 0.25, "learning_rate": 4.316738746381425e-05, "loss": 1.2506, "step": 2752 }, { "epoch": 0.25, "learning_rate": 4.316246213776463e-05, "loss": 1.2483, "step": 2753 }, { "epoch": 0.25, "learning_rate": 4.315753531833471e-05, "loss": 1.3183, "step": 2754 }, { "epoch": 0.25, "learning_rate": 4.31526070059296e-05, "loss": 1.3028, "step": 2755 }, { "epoch": 0.25, "learning_rate": 4.314767720095451e-05, "loss": 1.3124, "step": 2756 }, { "epoch": 0.25, "learning_rate": 4.3142745903814796e-05, "loss": 1.3289, "step": 2757 }, { "epoch": 0.25, "learning_rate": 4.313781311491592e-05, "loss": 1.3338, "step": 2758 }, { "epoch": 0.25, "learning_rate": 4.313287883466347e-05, "loss": 1.3023, "step": 2759 }, { "epoch": 0.25, "learning_rate": 4.312794306346317e-05, "loss": 1.3417, "step": 2760 }, { "epoch": 0.25, "learning_rate": 4.312300580172085e-05, "loss": 1.3077, "step": 2761 }, { "epoch": 0.25, "learning_rate": 4.311806704984246e-05, "loss": 1.2768, "step": 2762 }, { "epoch": 0.25, "learning_rate": 4.31131268082341e-05, "loss": 1.2929, "step": 2763 }, { "epoch": 0.25, "learning_rate": 4.310818507730196e-05, "loss": 1.2974, "step": 2764 }, { "epoch": 0.25, "learning_rate": 4.310324185745237e-05, "loss": 1.3003, "step": 2765 }, { "epoch": 0.25, "learning_rate": 4.309829714909178e-05, "loss": 1.2835, "step": 2766 }, { "epoch": 0.25, "learning_rate": 4.309335095262676e-05, "loss": 1.28, "step": 2767 }, { "epoch": 0.25, "learning_rate": 4.3088403268463996e-05, "loss": 1.3408, "step": 2768 }, { "epoch": 0.25, "learning_rate": 4.308345409701032e-05, "loss": 1.3675, "step": 2769 }, { "epoch": 0.25, "learning_rate": 4.307850343867266e-05, "loss": 1.3106, "step": 2770 }, { "epoch": 0.25, "learning_rate": 4.3073551293858064e-05, "loss": 1.3527, "step": 2771 }, { "epoch": 0.25, "learning_rate": 4.306859766297374e-05, "loss": 1.3429, "step": 2772 }, { "epoch": 0.25, "learning_rate": 4.306364254642697e-05, "loss": 1.3475, "step": 2773 }, { "epoch": 0.25, "learning_rate": 4.30586859446252e-05, "loss": 1.2552, "step": 2774 }, { "epoch": 0.25, "learning_rate": 4.305372785797596e-05, "loss": 1.3014, "step": 2775 }, { "epoch": 0.25, "learning_rate": 4.304876828688694e-05, "loss": 1.283, "step": 2776 }, { "epoch": 0.25, "learning_rate": 4.304380723176592e-05, "loss": 1.2425, "step": 2777 }, { "epoch": 0.25, "learning_rate": 4.303884469302082e-05, "loss": 1.3162, "step": 2778 }, { "epoch": 0.25, "learning_rate": 4.303388067105967e-05, "loss": 1.2791, "step": 2779 }, { "epoch": 0.25, "learning_rate": 4.302891516629064e-05, "loss": 1.2724, "step": 2780 }, { "epoch": 0.25, "learning_rate": 4.3023948179122e-05, "loss": 1.2706, "step": 2781 }, { "epoch": 0.25, "learning_rate": 4.301897970996216e-05, "loss": 1.2783, "step": 2782 }, { "epoch": 0.25, "learning_rate": 4.301400975921965e-05, "loss": 1.3426, "step": 2783 }, { "epoch": 0.25, "learning_rate": 4.30090383273031e-05, "loss": 1.2812, "step": 2784 }, { "epoch": 0.25, "learning_rate": 4.3004065414621294e-05, "loss": 1.277, "step": 2785 }, { "epoch": 0.25, "learning_rate": 4.299909102158311e-05, "loss": 1.2958, "step": 2786 }, { "epoch": 0.25, "learning_rate": 4.299411514859757e-05, "loss": 1.3076, "step": 2787 }, { "epoch": 0.25, "learning_rate": 4.2989137796073805e-05, "loss": 1.2808, "step": 2788 }, { "epoch": 0.25, "learning_rate": 4.298415896442106e-05, "loss": 1.3066, "step": 2789 }, { "epoch": 0.25, "learning_rate": 4.297917865404873e-05, "loss": 1.2583, "step": 2790 }, { "epoch": 0.25, "learning_rate": 4.29741968653663e-05, "loss": 1.2802, "step": 2791 }, { "epoch": 0.25, "learning_rate": 4.296921359878338e-05, "loss": 1.2923, "step": 2792 }, { "epoch": 0.25, "learning_rate": 4.2964228854709743e-05, "loss": 1.3216, "step": 2793 }, { "epoch": 0.25, "learning_rate": 4.295924263355523e-05, "loss": 1.2868, "step": 2794 }, { "epoch": 0.25, "learning_rate": 4.295425493572982e-05, "loss": 1.2889, "step": 2795 }, { "epoch": 0.25, "learning_rate": 4.2949265761643634e-05, "loss": 1.259, "step": 2796 }, { "epoch": 0.25, "learning_rate": 4.294427511170689e-05, "loss": 1.3143, "step": 2797 }, { "epoch": 0.25, "learning_rate": 4.2939282986329936e-05, "loss": 1.2863, "step": 2798 }, { "epoch": 0.25, "learning_rate": 4.293428938592325e-05, "loss": 1.2724, "step": 2799 }, { "epoch": 0.25, "learning_rate": 4.292929431089742e-05, "loss": 1.3239, "step": 2800 }, { "epoch": 0.25, "learning_rate": 4.292429776166314e-05, "loss": 1.3086, "step": 2801 }, { "epoch": 0.25, "learning_rate": 4.2919299738631275e-05, "loss": 1.2662, "step": 2802 }, { "epoch": 0.25, "learning_rate": 4.2914300242212754e-05, "loss": 1.325, "step": 2803 }, { "epoch": 0.25, "learning_rate": 4.290929927281867e-05, "loss": 1.2047, "step": 2804 }, { "epoch": 0.25, "learning_rate": 4.290429683086021e-05, "loss": 1.2531, "step": 2805 }, { "epoch": 0.25, "learning_rate": 4.28992929167487e-05, "loss": 1.2864, "step": 2806 }, { "epoch": 0.25, "learning_rate": 4.289428753089557e-05, "loss": 1.2586, "step": 2807 }, { "epoch": 0.25, "learning_rate": 4.288928067371238e-05, "loss": 1.3606, "step": 2808 }, { "epoch": 0.25, "learning_rate": 4.2884272345610824e-05, "loss": 1.2955, "step": 2809 }, { "epoch": 0.25, "learning_rate": 4.2879262547002694e-05, "loss": 1.2996, "step": 2810 }, { "epoch": 0.25, "learning_rate": 4.287425127829991e-05, "loss": 1.3205, "step": 2811 }, { "epoch": 0.25, "learning_rate": 4.2869238539914525e-05, "loss": 1.2869, "step": 2812 }, { "epoch": 0.25, "learning_rate": 4.286422433225869e-05, "loss": 1.3067, "step": 2813 }, { "epoch": 0.25, "learning_rate": 4.2859208655744705e-05, "loss": 1.2615, "step": 2814 }, { "epoch": 0.25, "learning_rate": 4.2854191510784966e-05, "loss": 1.2384, "step": 2815 }, { "epoch": 0.25, "learning_rate": 4.284917289779201e-05, "loss": 1.2914, "step": 2816 }, { "epoch": 0.25, "learning_rate": 4.284415281717847e-05, "loss": 1.2779, "step": 2817 }, { "epoch": 0.25, "learning_rate": 4.2839131269357134e-05, "loss": 1.3565, "step": 2818 }, { "epoch": 0.25, "learning_rate": 4.283410825474088e-05, "loss": 1.3326, "step": 2819 }, { "epoch": 0.25, "learning_rate": 4.282908377374271e-05, "loss": 1.2998, "step": 2820 }, { "epoch": 0.25, "learning_rate": 4.2824057826775766e-05, "loss": 1.2353, "step": 2821 }, { "epoch": 0.25, "learning_rate": 4.2819030414253294e-05, "loss": 1.2976, "step": 2822 }, { "epoch": 0.26, "learning_rate": 4.281400153658867e-05, "loss": 1.2545, "step": 2823 }, { "epoch": 0.26, "learning_rate": 4.280897119419537e-05, "loss": 1.2918, "step": 2824 }, { "epoch": 0.26, "learning_rate": 4.280393938748703e-05, "loss": 1.3122, "step": 2825 }, { "epoch": 0.26, "learning_rate": 4.279890611687737e-05, "loss": 1.2886, "step": 2826 }, { "epoch": 0.26, "learning_rate": 4.279387138278024e-05, "loss": 1.3302, "step": 2827 }, { "epoch": 0.26, "learning_rate": 4.278883518560961e-05, "loss": 1.2817, "step": 2828 }, { "epoch": 0.26, "learning_rate": 4.278379752577958e-05, "loss": 1.3227, "step": 2829 }, { "epoch": 0.26, "learning_rate": 4.277875840370437e-05, "loss": 1.2718, "step": 2830 }, { "epoch": 0.26, "learning_rate": 4.277371781979831e-05, "loss": 1.3012, "step": 2831 }, { "epoch": 0.26, "learning_rate": 4.276867577447584e-05, "loss": 1.2412, "step": 2832 }, { "epoch": 0.26, "learning_rate": 4.276363226815156e-05, "loss": 1.2802, "step": 2833 }, { "epoch": 0.26, "learning_rate": 4.2758587301240144e-05, "loss": 1.3164, "step": 2834 }, { "epoch": 0.26, "learning_rate": 4.275354087415642e-05, "loss": 1.2519, "step": 2835 }, { "epoch": 0.26, "learning_rate": 4.27484929873153e-05, "loss": 1.2567, "step": 2836 }, { "epoch": 0.26, "learning_rate": 4.274344364113186e-05, "loss": 1.3142, "step": 2837 }, { "epoch": 0.26, "learning_rate": 4.273839283602128e-05, "loss": 1.3075, "step": 2838 }, { "epoch": 0.26, "learning_rate": 4.273334057239884e-05, "loss": 1.3895, "step": 2839 }, { "epoch": 0.26, "learning_rate": 4.2728286850679957e-05, "loss": 1.2999, "step": 2840 }, { "epoch": 0.26, "learning_rate": 4.272323167128016e-05, "loss": 1.3149, "step": 2841 }, { "epoch": 0.26, "learning_rate": 4.271817503461512e-05, "loss": 1.2457, "step": 2842 }, { "epoch": 0.26, "learning_rate": 4.27131169411006e-05, "loss": 1.3007, "step": 2843 }, { "epoch": 0.26, "learning_rate": 4.270805739115248e-05, "loss": 1.3183, "step": 2844 }, { "epoch": 0.26, "learning_rate": 4.27029963851868e-05, "loss": 1.3436, "step": 2845 }, { "epoch": 0.26, "learning_rate": 4.269793392361968e-05, "loss": 1.2267, "step": 2846 }, { "epoch": 0.26, "learning_rate": 4.269287000686738e-05, "loss": 1.2957, "step": 2847 }, { "epoch": 0.26, "learning_rate": 4.268780463534627e-05, "loss": 1.2709, "step": 2848 }, { "epoch": 0.26, "learning_rate": 4.2682737809472835e-05, "loss": 1.2881, "step": 2849 }, { "epoch": 0.26, "learning_rate": 4.267766952966369e-05, "loss": 1.2983, "step": 2850 }, { "epoch": 0.26, "learning_rate": 4.267259979633557e-05, "loss": 1.3393, "step": 2851 }, { "epoch": 0.26, "learning_rate": 4.266752860990533e-05, "loss": 1.2601, "step": 2852 }, { "epoch": 0.26, "learning_rate": 4.2662455970789926e-05, "loss": 1.2919, "step": 2853 }, { "epoch": 0.26, "learning_rate": 4.265738187940647e-05, "loss": 1.3164, "step": 2854 }, { "epoch": 0.26, "learning_rate": 4.2652306336172145e-05, "loss": 1.2797, "step": 2855 }, { "epoch": 0.26, "learning_rate": 4.264722934150429e-05, "loss": 1.3301, "step": 2856 }, { "epoch": 0.26, "learning_rate": 4.2642150895820364e-05, "loss": 1.3467, "step": 2857 }, { "epoch": 0.26, "learning_rate": 4.263707099953793e-05, "loss": 1.2749, "step": 2858 }, { "epoch": 0.26, "learning_rate": 4.263198965307466e-05, "loss": 1.2731, "step": 2859 }, { "epoch": 0.26, "learning_rate": 4.262690685684838e-05, "loss": 1.3143, "step": 2860 }, { "epoch": 0.26, "learning_rate": 4.2621822611277e-05, "loss": 1.3049, "step": 2861 }, { "epoch": 0.26, "learning_rate": 4.2616736916778575e-05, "loss": 1.3049, "step": 2862 }, { "epoch": 0.26, "learning_rate": 4.2611649773771256e-05, "loss": 1.2657, "step": 2863 }, { "epoch": 0.26, "learning_rate": 4.260656118267334e-05, "loss": 1.291, "step": 2864 }, { "epoch": 0.26, "learning_rate": 4.260147114390322e-05, "loss": 1.3185, "step": 2865 }, { "epoch": 0.26, "learning_rate": 4.259637965787942e-05, "loss": 1.317, "step": 2866 }, { "epoch": 0.26, "learning_rate": 4.259128672502057e-05, "loss": 1.3525, "step": 2867 }, { "epoch": 0.26, "learning_rate": 4.258619234574543e-05, "loss": 1.3213, "step": 2868 }, { "epoch": 0.26, "learning_rate": 4.258109652047289e-05, "loss": 1.2444, "step": 2869 }, { "epoch": 0.26, "learning_rate": 4.257599924962195e-05, "loss": 1.3134, "step": 2870 }, { "epoch": 0.26, "learning_rate": 4.2570900533611705e-05, "loss": 1.3389, "step": 2871 }, { "epoch": 0.26, "learning_rate": 4.2565800372861394e-05, "loss": 1.3048, "step": 2872 }, { "epoch": 0.26, "learning_rate": 4.256069876779039e-05, "loss": 1.315, "step": 2873 }, { "epoch": 0.26, "learning_rate": 4.2555595718818135e-05, "loss": 1.3435, "step": 2874 }, { "epoch": 0.26, "learning_rate": 4.255049122636424e-05, "loss": 1.2982, "step": 2875 }, { "epoch": 0.26, "learning_rate": 4.254538529084841e-05, "loss": 1.2956, "step": 2876 }, { "epoch": 0.26, "learning_rate": 4.254027791269046e-05, "loss": 1.332, "step": 2877 }, { "epoch": 0.26, "learning_rate": 4.2535169092310356e-05, "loss": 1.3363, "step": 2878 }, { "epoch": 0.26, "learning_rate": 4.253005883012814e-05, "loss": 1.3382, "step": 2879 }, { "epoch": 0.26, "learning_rate": 4.252494712656403e-05, "loss": 1.2793, "step": 2880 }, { "epoch": 0.26, "learning_rate": 4.2519833982038286e-05, "loss": 1.3041, "step": 2881 }, { "epoch": 0.26, "learning_rate": 4.2514719396971356e-05, "loss": 1.2445, "step": 2882 }, { "epoch": 0.26, "learning_rate": 4.250960337178377e-05, "loss": 1.3427, "step": 2883 }, { "epoch": 0.26, "learning_rate": 4.250448590689619e-05, "loss": 1.2598, "step": 2884 }, { "epoch": 0.26, "learning_rate": 4.249936700272938e-05, "loss": 1.315, "step": 2885 }, { "epoch": 0.26, "learning_rate": 4.249424665970426e-05, "loss": 1.2561, "step": 2886 }, { "epoch": 0.26, "learning_rate": 4.2489124878241814e-05, "loss": 1.3286, "step": 2887 }, { "epoch": 0.26, "learning_rate": 4.248400165876318e-05, "loss": 1.2639, "step": 2888 }, { "epoch": 0.26, "learning_rate": 4.247887700168962e-05, "loss": 1.2889, "step": 2889 }, { "epoch": 0.26, "learning_rate": 4.247375090744248e-05, "loss": 1.271, "step": 2890 }, { "epoch": 0.26, "learning_rate": 4.246862337644326e-05, "loss": 1.2671, "step": 2891 }, { "epoch": 0.26, "learning_rate": 4.246349440911356e-05, "loss": 1.3263, "step": 2892 }, { "epoch": 0.26, "learning_rate": 4.24583640058751e-05, "loss": 1.2716, "step": 2893 }, { "epoch": 0.26, "learning_rate": 4.2453232167149724e-05, "loss": 1.2886, "step": 2894 }, { "epoch": 0.26, "learning_rate": 4.244809889335939e-05, "loss": 1.294, "step": 2895 }, { "epoch": 0.26, "learning_rate": 4.244296418492616e-05, "loss": 1.2757, "step": 2896 }, { "epoch": 0.26, "learning_rate": 4.2437828042272235e-05, "loss": 1.2814, "step": 2897 }, { "epoch": 0.26, "learning_rate": 4.243269046581993e-05, "loss": 1.2404, "step": 2898 }, { "epoch": 0.26, "learning_rate": 4.242755145599168e-05, "loss": 1.3464, "step": 2899 }, { "epoch": 0.26, "learning_rate": 4.242241101321002e-05, "loss": 1.3188, "step": 2900 }, { "epoch": 0.26, "learning_rate": 4.241726913789762e-05, "loss": 1.3064, "step": 2901 }, { "epoch": 0.26, "learning_rate": 4.241212583047726e-05, "loss": 1.3035, "step": 2902 }, { "epoch": 0.26, "learning_rate": 4.240698109137185e-05, "loss": 1.3421, "step": 2903 }, { "epoch": 0.26, "learning_rate": 4.240183492100439e-05, "loss": 1.3054, "step": 2904 }, { "epoch": 0.26, "learning_rate": 4.2396687319798026e-05, "loss": 1.3122, "step": 2905 }, { "epoch": 0.26, "learning_rate": 4.2391538288176024e-05, "loss": 1.2298, "step": 2906 }, { "epoch": 0.26, "learning_rate": 4.2386387826561735e-05, "loss": 1.3049, "step": 2907 }, { "epoch": 0.26, "learning_rate": 4.2381235935378656e-05, "loss": 1.2886, "step": 2908 }, { "epoch": 0.26, "learning_rate": 4.2376082615050396e-05, "loss": 1.2895, "step": 2909 }, { "epoch": 0.26, "learning_rate": 4.2370927866000676e-05, "loss": 1.3, "step": 2910 }, { "epoch": 0.26, "learning_rate": 4.236577168865333e-05, "loss": 1.2815, "step": 2911 }, { "epoch": 0.26, "learning_rate": 4.236061408343234e-05, "loss": 1.3156, "step": 2912 }, { "epoch": 0.26, "learning_rate": 4.235545505076175e-05, "loss": 1.2406, "step": 2913 }, { "epoch": 0.26, "learning_rate": 4.2350294591065774e-05, "loss": 1.2849, "step": 2914 }, { "epoch": 0.26, "learning_rate": 4.2345132704768725e-05, "loss": 1.3087, "step": 2915 }, { "epoch": 0.26, "learning_rate": 4.233996939229502e-05, "loss": 1.2786, "step": 2916 }, { "epoch": 0.26, "learning_rate": 4.233480465406921e-05, "loss": 1.3194, "step": 2917 }, { "epoch": 0.26, "learning_rate": 4.232963849051595e-05, "loss": 1.259, "step": 2918 }, { "epoch": 0.26, "learning_rate": 4.2324470902060045e-05, "loss": 1.3483, "step": 2919 }, { "epoch": 0.26, "learning_rate": 4.231930188912635e-05, "loss": 1.2814, "step": 2920 }, { "epoch": 0.26, "learning_rate": 4.231413145213991e-05, "loss": 1.2691, "step": 2921 }, { "epoch": 0.26, "learning_rate": 4.230895959152586e-05, "loss": 1.2939, "step": 2922 }, { "epoch": 0.26, "learning_rate": 4.2303786307709426e-05, "loss": 1.307, "step": 2923 }, { "epoch": 0.26, "learning_rate": 4.229861160111598e-05, "loss": 1.2838, "step": 2924 }, { "epoch": 0.26, "learning_rate": 4.2293435472171015e-05, "loss": 1.2677, "step": 2925 }, { "epoch": 0.26, "learning_rate": 4.228825792130012e-05, "loss": 1.2762, "step": 2926 }, { "epoch": 0.26, "learning_rate": 4.228307894892902e-05, "loss": 1.295, "step": 2927 }, { "epoch": 0.26, "learning_rate": 4.227789855548354e-05, "loss": 1.2507, "step": 2928 }, { "epoch": 0.26, "learning_rate": 4.2272716741389626e-05, "loss": 1.3314, "step": 2929 }, { "epoch": 0.26, "learning_rate": 4.2267533507073365e-05, "loss": 1.3179, "step": 2930 }, { "epoch": 0.26, "learning_rate": 4.226234885296091e-05, "loss": 1.2489, "step": 2931 }, { "epoch": 0.26, "learning_rate": 4.225716277947859e-05, "loss": 1.3006, "step": 2932 }, { "epoch": 0.27, "learning_rate": 4.2251975287052804e-05, "loss": 1.3286, "step": 2933 }, { "epoch": 0.27, "learning_rate": 4.224678637611009e-05, "loss": 1.2661, "step": 2934 }, { "epoch": 0.27, "learning_rate": 4.22415960470771e-05, "loss": 1.2442, "step": 2935 }, { "epoch": 0.27, "learning_rate": 4.22364043003806e-05, "loss": 1.2514, "step": 2936 }, { "epoch": 0.27, "learning_rate": 4.223121113644747e-05, "loss": 1.2487, "step": 2937 }, { "epoch": 0.27, "learning_rate": 4.2226016555704716e-05, "loss": 1.2635, "step": 2938 }, { "epoch": 0.27, "learning_rate": 4.222082055857945e-05, "loss": 1.2594, "step": 2939 }, { "epoch": 0.27, "learning_rate": 4.2215623145498914e-05, "loss": 1.3344, "step": 2940 }, { "epoch": 0.27, "learning_rate": 4.221042431689044e-05, "loss": 1.3301, "step": 2941 }, { "epoch": 0.27, "learning_rate": 4.2205224073181505e-05, "loss": 1.2596, "step": 2942 }, { "epoch": 0.27, "learning_rate": 4.220002241479969e-05, "loss": 1.3199, "step": 2943 }, { "epoch": 0.27, "learning_rate": 4.2194819342172684e-05, "loss": 1.2543, "step": 2944 }, { "epoch": 0.27, "learning_rate": 4.218961485572832e-05, "loss": 1.3305, "step": 2945 }, { "epoch": 0.27, "learning_rate": 4.2184408955894513e-05, "loss": 1.2799, "step": 2946 }, { "epoch": 0.27, "learning_rate": 4.2179201643099317e-05, "loss": 1.3232, "step": 2947 }, { "epoch": 0.27, "learning_rate": 4.2173992917770896e-05, "loss": 1.266, "step": 2948 }, { "epoch": 0.27, "learning_rate": 4.216878278033753e-05, "loss": 1.2558, "step": 2949 }, { "epoch": 0.27, "learning_rate": 4.2163571231227606e-05, "loss": 1.2647, "step": 2950 }, { "epoch": 0.27, "learning_rate": 4.215835827086965e-05, "loss": 1.2844, "step": 2951 }, { "epoch": 0.27, "learning_rate": 4.215314389969227e-05, "loss": 1.2966, "step": 2952 }, { "epoch": 0.27, "learning_rate": 4.2147928118124227e-05, "loss": 1.3592, "step": 2953 }, { "epoch": 0.27, "learning_rate": 4.2142710926594365e-05, "loss": 1.2641, "step": 2954 }, { "epoch": 0.27, "learning_rate": 4.213749232553168e-05, "loss": 1.2532, "step": 2955 }, { "epoch": 0.27, "learning_rate": 4.213227231536525e-05, "loss": 1.2793, "step": 2956 }, { "epoch": 0.27, "learning_rate": 4.212705089652428e-05, "loss": 1.2858, "step": 2957 }, { "epoch": 0.27, "learning_rate": 4.21218280694381e-05, "loss": 1.3014, "step": 2958 }, { "epoch": 0.27, "learning_rate": 4.211660383453614e-05, "loss": 1.2906, "step": 2959 }, { "epoch": 0.27, "learning_rate": 4.211137819224796e-05, "loss": 1.2748, "step": 2960 }, { "epoch": 0.27, "learning_rate": 4.2106151143003244e-05, "loss": 1.2961, "step": 2961 }, { "epoch": 0.27, "learning_rate": 4.2100922687231754e-05, "loss": 1.3038, "step": 2962 }, { "epoch": 0.27, "learning_rate": 4.20956928253634e-05, "loss": 1.316, "step": 2963 }, { "epoch": 0.27, "learning_rate": 4.209046155782821e-05, "loss": 1.2687, "step": 2964 }, { "epoch": 0.27, "learning_rate": 4.20852288850563e-05, "loss": 1.2823, "step": 2965 }, { "epoch": 0.27, "learning_rate": 4.207999480747794e-05, "loss": 1.2813, "step": 2966 }, { "epoch": 0.27, "learning_rate": 4.2074759325523476e-05, "loss": 1.3191, "step": 2967 }, { "epoch": 0.27, "learning_rate": 4.206952243962339e-05, "loss": 1.2435, "step": 2968 }, { "epoch": 0.27, "learning_rate": 4.206428415020828e-05, "loss": 1.2586, "step": 2969 }, { "epoch": 0.27, "learning_rate": 4.205904445770886e-05, "loss": 1.2538, "step": 2970 }, { "epoch": 0.27, "learning_rate": 4.205380336255594e-05, "loss": 1.2741, "step": 2971 }, { "epoch": 0.27, "learning_rate": 4.2048560865180486e-05, "loss": 1.3047, "step": 2972 }, { "epoch": 0.27, "learning_rate": 4.204331696601353e-05, "loss": 1.2348, "step": 2973 }, { "epoch": 0.27, "learning_rate": 4.203807166548627e-05, "loss": 1.3314, "step": 2974 }, { "epoch": 0.27, "learning_rate": 4.203282496402996e-05, "loss": 1.285, "step": 2975 }, { "epoch": 0.27, "learning_rate": 4.202757686207602e-05, "loss": 1.2966, "step": 2976 }, { "epoch": 0.27, "learning_rate": 4.2022327360055965e-05, "loss": 1.2697, "step": 2977 }, { "epoch": 0.27, "learning_rate": 4.201707645840144e-05, "loss": 1.2768, "step": 2978 }, { "epoch": 0.27, "learning_rate": 4.2011824157544166e-05, "loss": 1.3334, "step": 2979 }, { "epoch": 0.27, "learning_rate": 4.200657045791603e-05, "loss": 1.2787, "step": 2980 }, { "epoch": 0.27, "learning_rate": 4.200131535994899e-05, "loss": 1.3539, "step": 2981 }, { "epoch": 0.27, "learning_rate": 4.199605886407515e-05, "loss": 1.2827, "step": 2982 }, { "epoch": 0.27, "learning_rate": 4.1990800970726705e-05, "loss": 1.292, "step": 2983 }, { "epoch": 0.27, "learning_rate": 4.1985541680336e-05, "loss": 1.292, "step": 2984 }, { "epoch": 0.27, "learning_rate": 4.198028099333545e-05, "loss": 1.2573, "step": 2985 }, { "epoch": 0.27, "learning_rate": 4.197501891015762e-05, "loss": 1.2966, "step": 2986 }, { "epoch": 0.27, "learning_rate": 4.196975543123517e-05, "loss": 1.2795, "step": 2987 }, { "epoch": 0.27, "learning_rate": 4.196449055700088e-05, "loss": 1.2711, "step": 2988 }, { "epoch": 0.27, "learning_rate": 4.1959224287887656e-05, "loss": 1.2777, "step": 2989 }, { "epoch": 0.27, "learning_rate": 4.1953956624328495e-05, "loss": 1.3248, "step": 2990 }, { "epoch": 0.27, "learning_rate": 4.1948687566756526e-05, "loss": 1.2971, "step": 2991 }, { "epoch": 0.27, "learning_rate": 4.1943417115605004e-05, "loss": 1.2808, "step": 2992 }, { "epoch": 0.27, "learning_rate": 4.193814527130727e-05, "loss": 1.3022, "step": 2993 }, { "epoch": 0.27, "learning_rate": 4.193287203429679e-05, "loss": 1.2067, "step": 2994 }, { "epoch": 0.27, "learning_rate": 4.192759740500716e-05, "loss": 1.2583, "step": 2995 }, { "epoch": 0.27, "learning_rate": 4.192232138387206e-05, "loss": 1.337, "step": 2996 }, { "epoch": 0.27, "learning_rate": 4.191704397132532e-05, "loss": 1.2575, "step": 2997 }, { "epoch": 0.27, "learning_rate": 4.1911765167800864e-05, "loss": 1.3028, "step": 2998 }, { "epoch": 0.27, "learning_rate": 4.1906484973732724e-05, "loss": 1.3143, "step": 2999 }, { "epoch": 0.27, "learning_rate": 4.1901203389555067e-05, "loss": 1.222, "step": 3000 }, { "epoch": 0.27, "learning_rate": 4.189592041570216e-05, "loss": 1.2618, "step": 3001 }, { "epoch": 0.27, "learning_rate": 4.1890636052608376e-05, "loss": 1.3177, "step": 3002 }, { "epoch": 0.27, "learning_rate": 4.188535030070824e-05, "loss": 1.3334, "step": 3003 }, { "epoch": 0.27, "learning_rate": 4.188006316043633e-05, "loss": 1.2709, "step": 3004 }, { "epoch": 0.27, "learning_rate": 4.187477463222741e-05, "loss": 1.3207, "step": 3005 }, { "epoch": 0.27, "learning_rate": 4.186948471651628e-05, "loss": 1.2587, "step": 3006 }, { "epoch": 0.27, "learning_rate": 4.186419341373793e-05, "loss": 1.2822, "step": 3007 }, { "epoch": 0.27, "learning_rate": 4.1858900724327424e-05, "loss": 1.1966, "step": 3008 }, { "epoch": 0.27, "learning_rate": 4.185360664871992e-05, "loss": 1.2784, "step": 3009 }, { "epoch": 0.27, "learning_rate": 4.184831118735074e-05, "loss": 1.2394, "step": 3010 }, { "epoch": 0.27, "learning_rate": 4.18430143406553e-05, "loss": 1.35, "step": 3011 }, { "epoch": 0.27, "learning_rate": 4.183771610906909e-05, "loss": 1.2997, "step": 3012 }, { "epoch": 0.27, "learning_rate": 4.183241649302779e-05, "loss": 1.3132, "step": 3013 }, { "epoch": 0.27, "learning_rate": 4.182711549296713e-05, "loss": 1.2553, "step": 3014 }, { "epoch": 0.27, "learning_rate": 4.1821813109322974e-05, "loss": 1.2861, "step": 3015 }, { "epoch": 0.27, "learning_rate": 4.181650934253132e-05, "loss": 1.3064, "step": 3016 }, { "epoch": 0.27, "learning_rate": 4.181120419302824e-05, "loss": 1.3299, "step": 3017 }, { "epoch": 0.27, "learning_rate": 4.180589766124996e-05, "loss": 1.3361, "step": 3018 }, { "epoch": 0.27, "learning_rate": 4.180058974763279e-05, "loss": 1.2628, "step": 3019 }, { "epoch": 0.27, "learning_rate": 4.1795280452613175e-05, "loss": 1.2568, "step": 3020 }, { "epoch": 0.27, "learning_rate": 4.178996977662765e-05, "loss": 1.2865, "step": 3021 }, { "epoch": 0.27, "learning_rate": 4.178465772011288e-05, "loss": 1.3082, "step": 3022 }, { "epoch": 0.27, "learning_rate": 4.1779344283505656e-05, "loss": 1.2642, "step": 3023 }, { "epoch": 0.27, "learning_rate": 4.177402946724286e-05, "loss": 1.2637, "step": 3024 }, { "epoch": 0.27, "learning_rate": 4.176871327176148e-05, "loss": 1.2433, "step": 3025 }, { "epoch": 0.27, "learning_rate": 4.176339569749865e-05, "loss": 1.2784, "step": 3026 }, { "epoch": 0.27, "learning_rate": 4.175807674489159e-05, "loss": 1.2597, "step": 3027 }, { "epoch": 0.27, "learning_rate": 4.175275641437765e-05, "loss": 1.2787, "step": 3028 }, { "epoch": 0.27, "learning_rate": 4.174743470639427e-05, "loss": 1.3328, "step": 3029 }, { "epoch": 0.27, "learning_rate": 4.174211162137904e-05, "loss": 1.2951, "step": 3030 }, { "epoch": 0.27, "learning_rate": 4.173678715976963e-05, "loss": 1.2858, "step": 3031 }, { "epoch": 0.27, "learning_rate": 4.173146132200383e-05, "loss": 1.2158, "step": 3032 }, { "epoch": 0.27, "learning_rate": 4.1726134108519564e-05, "loss": 1.2988, "step": 3033 }, { "epoch": 0.27, "learning_rate": 4.172080551975485e-05, "loss": 1.3074, "step": 3034 }, { "epoch": 0.27, "learning_rate": 4.171547555614781e-05, "loss": 1.2982, "step": 3035 }, { "epoch": 0.27, "learning_rate": 4.171014421813671e-05, "loss": 1.281, "step": 3036 }, { "epoch": 0.27, "learning_rate": 4.1704811506159904e-05, "loss": 1.2772, "step": 3037 }, { "epoch": 0.27, "learning_rate": 4.1699477420655854e-05, "loss": 1.3037, "step": 3038 }, { "epoch": 0.27, "learning_rate": 4.169414196206317e-05, "loss": 1.308, "step": 3039 }, { "epoch": 0.27, "learning_rate": 4.168880513082053e-05, "loss": 1.2366, "step": 3040 }, { "epoch": 0.27, "learning_rate": 4.168346692736675e-05, "loss": 1.3061, "step": 3041 }, { "epoch": 0.27, "learning_rate": 4.167812735214078e-05, "loss": 1.225, "step": 3042 }, { "epoch": 0.27, "learning_rate": 4.167278640558163e-05, "loss": 1.2634, "step": 3043 }, { "epoch": 0.28, "learning_rate": 4.166744408812846e-05, "loss": 1.2894, "step": 3044 }, { "epoch": 0.28, "learning_rate": 4.1662100400220535e-05, "loss": 1.3436, "step": 3045 }, { "epoch": 0.28, "learning_rate": 4.165675534229723e-05, "loss": 1.2772, "step": 3046 }, { "epoch": 0.28, "learning_rate": 4.165140891479804e-05, "loss": 1.3304, "step": 3047 }, { "epoch": 0.28, "learning_rate": 4.1646061118162554e-05, "loss": 1.2786, "step": 3048 }, { "epoch": 0.28, "learning_rate": 4.16407119528305e-05, "loss": 1.323, "step": 3049 }, { "epoch": 0.28, "learning_rate": 4.1635361419241706e-05, "loss": 1.3265, "step": 3050 }, { "epoch": 0.28, "learning_rate": 4.163000951783609e-05, "loss": 1.3275, "step": 3051 }, { "epoch": 0.28, "learning_rate": 4.162465624905373e-05, "loss": 1.3252, "step": 3052 }, { "epoch": 0.28, "learning_rate": 4.161930161333477e-05, "loss": 1.2584, "step": 3053 }, { "epoch": 0.28, "learning_rate": 4.1613945611119494e-05, "loss": 1.3366, "step": 3054 }, { "epoch": 0.28, "learning_rate": 4.16085882428483e-05, "loss": 1.3433, "step": 3055 }, { "epoch": 0.28, "learning_rate": 4.160322950896167e-05, "loss": 1.2699, "step": 3056 }, { "epoch": 0.28, "learning_rate": 4.1597869409900235e-05, "loss": 1.274, "step": 3057 }, { "epoch": 0.28, "learning_rate": 4.159250794610472e-05, "loss": 1.2897, "step": 3058 }, { "epoch": 0.28, "learning_rate": 4.1587145118015945e-05, "loss": 1.2211, "step": 3059 }, { "epoch": 0.28, "learning_rate": 4.1581780926074885e-05, "loss": 1.3201, "step": 3060 }, { "epoch": 0.28, "learning_rate": 4.157641537072259e-05, "loss": 1.2933, "step": 3061 }, { "epoch": 0.28, "learning_rate": 4.157104845240023e-05, "loss": 1.3153, "step": 3062 }, { "epoch": 0.28, "learning_rate": 4.15656801715491e-05, "loss": 1.2666, "step": 3063 }, { "epoch": 0.28, "learning_rate": 4.156031052861059e-05, "loss": 1.282, "step": 3064 }, { "epoch": 0.28, "learning_rate": 4.155493952402622e-05, "loss": 1.2976, "step": 3065 }, { "epoch": 0.28, "learning_rate": 4.154956715823761e-05, "loss": 1.3017, "step": 3066 }, { "epoch": 0.28, "learning_rate": 4.154419343168649e-05, "loss": 1.3478, "step": 3067 }, { "epoch": 0.28, "learning_rate": 4.15388183448147e-05, "loss": 1.3047, "step": 3068 }, { "epoch": 0.28, "learning_rate": 4.153344189806422e-05, "loss": 1.2871, "step": 3069 }, { "epoch": 0.28, "learning_rate": 4.1528064091877096e-05, "loss": 1.2632, "step": 3070 }, { "epoch": 0.28, "learning_rate": 4.1522684926695526e-05, "loss": 1.2441, "step": 3071 }, { "epoch": 0.28, "learning_rate": 4.15173044029618e-05, "loss": 1.2772, "step": 3072 }, { "epoch": 0.28, "learning_rate": 4.1511922521118317e-05, "loss": 1.2555, "step": 3073 }, { "epoch": 0.28, "learning_rate": 4.15065392816076e-05, "loss": 1.3139, "step": 3074 }, { "epoch": 0.28, "learning_rate": 4.150115468487228e-05, "loss": 1.329, "step": 3075 }, { "epoch": 0.28, "learning_rate": 4.149576873135509e-05, "loss": 1.2949, "step": 3076 }, { "epoch": 0.28, "learning_rate": 4.149038142149888e-05, "loss": 1.2928, "step": 3077 }, { "epoch": 0.28, "learning_rate": 4.1484992755746626e-05, "loss": 1.2716, "step": 3078 }, { "epoch": 0.28, "learning_rate": 4.147960273454139e-05, "loss": 1.2364, "step": 3079 }, { "epoch": 0.28, "learning_rate": 4.1474211358326365e-05, "loss": 1.2669, "step": 3080 }, { "epoch": 0.28, "learning_rate": 4.1468818627544845e-05, "loss": 1.314, "step": 3081 }, { "epoch": 0.28, "learning_rate": 4.146342454264024e-05, "loss": 1.2935, "step": 3082 }, { "epoch": 0.28, "learning_rate": 4.1458029104056074e-05, "loss": 1.2864, "step": 3083 }, { "epoch": 0.28, "learning_rate": 4.145263231223597e-05, "loss": 1.3088, "step": 3084 }, { "epoch": 0.28, "learning_rate": 4.144723416762368e-05, "loss": 1.264, "step": 3085 }, { "epoch": 0.28, "learning_rate": 4.144183467066305e-05, "loss": 1.2582, "step": 3086 }, { "epoch": 0.28, "learning_rate": 4.143643382179805e-05, "loss": 1.2857, "step": 3087 }, { "epoch": 0.28, "learning_rate": 4.143103162147276e-05, "loss": 1.2898, "step": 3088 }, { "epoch": 0.28, "learning_rate": 4.142562807013136e-05, "loss": 1.2849, "step": 3089 }, { "epoch": 0.28, "learning_rate": 4.142022316821815e-05, "loss": 1.3122, "step": 3090 }, { "epoch": 0.28, "learning_rate": 4.141481691617755e-05, "loss": 1.2745, "step": 3091 }, { "epoch": 0.28, "learning_rate": 4.1409409314454065e-05, "loss": 1.2742, "step": 3092 }, { "epoch": 0.28, "learning_rate": 4.1404000363492336e-05, "loss": 1.2732, "step": 3093 }, { "epoch": 0.28, "learning_rate": 4.1398590063737094e-05, "loss": 1.2968, "step": 3094 }, { "epoch": 0.28, "learning_rate": 4.1393178415633214e-05, "loss": 1.2526, "step": 3095 }, { "epoch": 0.28, "learning_rate": 4.1387765419625646e-05, "loss": 1.252, "step": 3096 }, { "epoch": 0.28, "learning_rate": 4.138235107615947e-05, "loss": 1.3033, "step": 3097 }, { "epoch": 0.28, "learning_rate": 4.137693538567986e-05, "loss": 1.3263, "step": 3098 }, { "epoch": 0.28, "learning_rate": 4.137151834863213e-05, "loss": 1.2909, "step": 3099 }, { "epoch": 0.28, "learning_rate": 4.136609996546168e-05, "loss": 1.2925, "step": 3100 }, { "epoch": 0.28, "learning_rate": 4.136068023661402e-05, "loss": 1.3229, "step": 3101 }, { "epoch": 0.28, "learning_rate": 4.1355259162534796e-05, "loss": 1.2524, "step": 3102 }, { "epoch": 0.28, "learning_rate": 4.1349836743669734e-05, "loss": 1.2458, "step": 3103 }, { "epoch": 0.28, "learning_rate": 4.1344412980464695e-05, "loss": 1.3165, "step": 3104 }, { "epoch": 0.28, "learning_rate": 4.133898787336562e-05, "loss": 1.2925, "step": 3105 }, { "epoch": 0.28, "learning_rate": 4.13335614228186e-05, "loss": 1.3113, "step": 3106 }, { "epoch": 0.28, "learning_rate": 4.1328133629269815e-05, "loss": 1.2635, "step": 3107 }, { "epoch": 0.28, "learning_rate": 4.132270449316554e-05, "loss": 1.2949, "step": 3108 }, { "epoch": 0.28, "learning_rate": 4.131727401495219e-05, "loss": 1.2656, "step": 3109 }, { "epoch": 0.28, "learning_rate": 4.131184219507629e-05, "loss": 1.2928, "step": 3110 }, { "epoch": 0.28, "learning_rate": 4.130640903398443e-05, "loss": 1.307, "step": 3111 }, { "epoch": 0.28, "learning_rate": 4.1300974532123376e-05, "loss": 1.269, "step": 3112 }, { "epoch": 0.28, "learning_rate": 4.129553868993995e-05, "loss": 1.2634, "step": 3113 }, { "epoch": 0.28, "learning_rate": 4.129010150788112e-05, "loss": 1.2985, "step": 3114 }, { "epoch": 0.28, "learning_rate": 4.128466298639394e-05, "loss": 1.3156, "step": 3115 }, { "epoch": 0.28, "learning_rate": 4.1279223125925595e-05, "loss": 1.2375, "step": 3116 }, { "epoch": 0.28, "learning_rate": 4.127378192692335e-05, "loss": 1.2206, "step": 3117 }, { "epoch": 0.28, "learning_rate": 4.126833938983461e-05, "loss": 1.2864, "step": 3118 }, { "epoch": 0.28, "learning_rate": 4.126289551510688e-05, "loss": 1.3088, "step": 3119 }, { "epoch": 0.28, "learning_rate": 4.1257450303187785e-05, "loss": 1.3393, "step": 3120 }, { "epoch": 0.28, "learning_rate": 4.125200375452503e-05, "loss": 1.316, "step": 3121 }, { "epoch": 0.28, "learning_rate": 4.124655586956645e-05, "loss": 1.2982, "step": 3122 }, { "epoch": 0.28, "learning_rate": 4.1241106648760014e-05, "loss": 1.267, "step": 3123 }, { "epoch": 0.28, "learning_rate": 4.123565609255374e-05, "loss": 1.304, "step": 3124 }, { "epoch": 0.28, "learning_rate": 4.123020420139583e-05, "loss": 1.3348, "step": 3125 }, { "epoch": 0.28, "learning_rate": 4.122475097573452e-05, "loss": 1.3662, "step": 3126 }, { "epoch": 0.28, "learning_rate": 4.121929641601821e-05, "loss": 1.3453, "step": 3127 }, { "epoch": 0.28, "learning_rate": 4.12138405226954e-05, "loss": 1.2803, "step": 3128 }, { "epoch": 0.28, "learning_rate": 4.120838329621468e-05, "loss": 1.3026, "step": 3129 }, { "epoch": 0.28, "learning_rate": 4.1202924737024764e-05, "loss": 1.2676, "step": 3130 }, { "epoch": 0.28, "learning_rate": 4.119746484557449e-05, "loss": 1.2502, "step": 3131 }, { "epoch": 0.28, "learning_rate": 4.119200362231276e-05, "loss": 1.3157, "step": 3132 }, { "epoch": 0.28, "learning_rate": 4.1186541067688634e-05, "loss": 1.281, "step": 3133 }, { "epoch": 0.28, "learning_rate": 4.118107718215126e-05, "loss": 1.2751, "step": 3134 }, { "epoch": 0.28, "learning_rate": 4.117561196614989e-05, "loss": 1.2749, "step": 3135 }, { "epoch": 0.28, "learning_rate": 4.11701454201339e-05, "loss": 1.2511, "step": 3136 }, { "epoch": 0.28, "learning_rate": 4.116467754455278e-05, "loss": 1.32, "step": 3137 }, { "epoch": 0.28, "learning_rate": 4.115920833985609e-05, "loss": 1.2825, "step": 3138 }, { "epoch": 0.28, "learning_rate": 4.1153737806493544e-05, "loss": 1.2763, "step": 3139 }, { "epoch": 0.28, "learning_rate": 4.114826594491494e-05, "loss": 1.2891, "step": 3140 }, { "epoch": 0.28, "learning_rate": 4.1142792755570214e-05, "loss": 1.2657, "step": 3141 }, { "epoch": 0.28, "learning_rate": 4.113731823890936e-05, "loss": 1.3181, "step": 3142 }, { "epoch": 0.28, "learning_rate": 4.113184239538254e-05, "loss": 1.2445, "step": 3143 }, { "epoch": 0.28, "learning_rate": 4.112636522543998e-05, "loss": 1.3129, "step": 3144 }, { "epoch": 0.28, "learning_rate": 4.1120886729532025e-05, "loss": 1.2715, "step": 3145 }, { "epoch": 0.28, "learning_rate": 4.111540690810916e-05, "loss": 1.338, "step": 3146 }, { "epoch": 0.28, "learning_rate": 4.1109925761621925e-05, "loss": 1.3076, "step": 3147 }, { "epoch": 0.28, "learning_rate": 4.110444329052103e-05, "loss": 1.3515, "step": 3148 }, { "epoch": 0.28, "learning_rate": 4.109895949525724e-05, "loss": 1.242, "step": 3149 }, { "epoch": 0.28, "learning_rate": 4.109347437628146e-05, "loss": 1.3226, "step": 3150 }, { "epoch": 0.28, "learning_rate": 4.10879879340447e-05, "loss": 1.2517, "step": 3151 }, { "epoch": 0.28, "learning_rate": 4.108250016899806e-05, "loss": 1.2687, "step": 3152 }, { "epoch": 0.28, "learning_rate": 4.1077011081592776e-05, "loss": 1.3173, "step": 3153 }, { "epoch": 0.28, "learning_rate": 4.107152067228018e-05, "loss": 1.286, "step": 3154 }, { "epoch": 0.29, "learning_rate": 4.10660289415117e-05, "loss": 1.3217, "step": 3155 }, { "epoch": 0.29, "learning_rate": 4.106053588973889e-05, "loss": 1.2773, "step": 3156 }, { "epoch": 0.29, "learning_rate": 4.105504151741342e-05, "loss": 1.2939, "step": 3157 }, { "epoch": 0.29, "learning_rate": 4.104954582498705e-05, "loss": 1.2882, "step": 3158 }, { "epoch": 0.29, "learning_rate": 4.104404881291164e-05, "loss": 1.2834, "step": 3159 }, { "epoch": 0.29, "learning_rate": 4.103855048163919e-05, "loss": 1.2627, "step": 3160 }, { "epoch": 0.29, "learning_rate": 4.103305083162179e-05, "loss": 1.3573, "step": 3161 }, { "epoch": 0.29, "learning_rate": 4.1027549863311634e-05, "loss": 1.2329, "step": 3162 }, { "epoch": 0.29, "learning_rate": 4.102204757716103e-05, "loss": 1.3008, "step": 3163 }, { "epoch": 0.29, "learning_rate": 4.10165439736224e-05, "loss": 1.3134, "step": 3164 }, { "epoch": 0.29, "learning_rate": 4.101103905314827e-05, "loss": 1.2918, "step": 3165 }, { "epoch": 0.29, "learning_rate": 4.1005532816191265e-05, "loss": 1.2361, "step": 3166 }, { "epoch": 0.29, "learning_rate": 4.100002526320415e-05, "loss": 1.2643, "step": 3167 }, { "epoch": 0.29, "learning_rate": 4.099451639463974e-05, "loss": 1.2499, "step": 3168 }, { "epoch": 0.29, "learning_rate": 4.0989006210951014e-05, "loss": 1.3279, "step": 3169 }, { "epoch": 0.29, "learning_rate": 4.098349471259104e-05, "loss": 1.3017, "step": 3170 }, { "epoch": 0.29, "learning_rate": 4.097798190001298e-05, "loss": 1.2961, "step": 3171 }, { "epoch": 0.29, "learning_rate": 4.0972467773670134e-05, "loss": 1.3143, "step": 3172 }, { "epoch": 0.29, "learning_rate": 4.096695233401588e-05, "loss": 1.3355, "step": 3173 }, { "epoch": 0.29, "learning_rate": 4.096143558150372e-05, "loss": 1.2913, "step": 3174 }, { "epoch": 0.29, "learning_rate": 4.095591751658725e-05, "loss": 1.2646, "step": 3175 }, { "epoch": 0.29, "learning_rate": 4.095039813972021e-05, "loss": 1.2826, "step": 3176 }, { "epoch": 0.29, "learning_rate": 4.09448774513564e-05, "loss": 1.3083, "step": 3177 }, { "epoch": 0.29, "learning_rate": 4.093935545194975e-05, "loss": 1.2681, "step": 3178 }, { "epoch": 0.29, "learning_rate": 4.093383214195431e-05, "loss": 1.278, "step": 3179 }, { "epoch": 0.29, "learning_rate": 4.092830752182422e-05, "loss": 1.3097, "step": 3180 }, { "epoch": 0.29, "learning_rate": 4.092278159201373e-05, "loss": 1.2967, "step": 3181 }, { "epoch": 0.29, "learning_rate": 4.0917254352977206e-05, "loss": 1.2287, "step": 3182 }, { "epoch": 0.29, "learning_rate": 4.091172580516912e-05, "loss": 1.2753, "step": 3183 }, { "epoch": 0.29, "learning_rate": 4.090619594904404e-05, "loss": 1.2844, "step": 3184 }, { "epoch": 0.29, "learning_rate": 4.0900664785056645e-05, "loss": 1.23, "step": 3185 }, { "epoch": 0.29, "learning_rate": 4.0895132313661743e-05, "loss": 1.3583, "step": 3186 }, { "epoch": 0.29, "learning_rate": 4.0889598535314215e-05, "loss": 1.2929, "step": 3187 }, { "epoch": 0.29, "learning_rate": 4.088406345046909e-05, "loss": 1.2595, "step": 3188 }, { "epoch": 0.29, "learning_rate": 4.087852705958146e-05, "loss": 1.2231, "step": 3189 }, { "epoch": 0.29, "learning_rate": 4.0872989363106555e-05, "loss": 1.2934, "step": 3190 }, { "epoch": 0.29, "learning_rate": 4.08674503614997e-05, "loss": 1.2851, "step": 3191 }, { "epoch": 0.29, "learning_rate": 4.086191005521634e-05, "loss": 1.265, "step": 3192 }, { "epoch": 0.29, "learning_rate": 4.085636844471201e-05, "loss": 1.2974, "step": 3193 }, { "epoch": 0.29, "learning_rate": 4.085082553044236e-05, "loss": 1.2788, "step": 3194 }, { "epoch": 0.29, "learning_rate": 4.0845281312863154e-05, "loss": 1.2561, "step": 3195 }, { "epoch": 0.29, "learning_rate": 4.0839735792430246e-05, "loss": 1.2639, "step": 3196 }, { "epoch": 0.29, "learning_rate": 4.083418896959962e-05, "loss": 1.2501, "step": 3197 }, { "epoch": 0.29, "learning_rate": 4.082864084482735e-05, "loss": 1.3028, "step": 3198 }, { "epoch": 0.29, "learning_rate": 4.082309141856961e-05, "loss": 1.276, "step": 3199 }, { "epoch": 0.29, "learning_rate": 4.081754069128272e-05, "loss": 1.315, "step": 3200 }, { "epoch": 0.29, "learning_rate": 4.0811988663423054e-05, "loss": 1.3223, "step": 3201 }, { "epoch": 0.29, "learning_rate": 4.080643533544713e-05, "loss": 1.2543, "step": 3202 }, { "epoch": 0.29, "learning_rate": 4.080088070781156e-05, "loss": 1.2592, "step": 3203 }, { "epoch": 0.29, "learning_rate": 4.079532478097306e-05, "loss": 1.2788, "step": 3204 }, { "epoch": 0.29, "learning_rate": 4.078976755538848e-05, "loss": 1.2726, "step": 3205 }, { "epoch": 0.29, "learning_rate": 4.078420903151473e-05, "loss": 1.2001, "step": 3206 }, { "epoch": 0.29, "learning_rate": 4.077864920980885e-05, "loss": 1.2603, "step": 3207 }, { "epoch": 0.29, "learning_rate": 4.077308809072801e-05, "loss": 1.2595, "step": 3208 }, { "epoch": 0.29, "learning_rate": 4.076752567472943e-05, "loss": 1.3502, "step": 3209 }, { "epoch": 0.29, "learning_rate": 4.0761961962270506e-05, "loss": 1.2785, "step": 3210 }, { "epoch": 0.29, "learning_rate": 4.075639695380869e-05, "loss": 1.2495, "step": 3211 }, { "epoch": 0.29, "learning_rate": 4.0750830649801554e-05, "loss": 1.2906, "step": 3212 }, { "epoch": 0.29, "learning_rate": 4.0745263050706784e-05, "loss": 1.2906, "step": 3213 }, { "epoch": 0.29, "learning_rate": 4.0739694156982164e-05, "loss": 1.2877, "step": 3214 }, { "epoch": 0.29, "learning_rate": 4.073412396908559e-05, "loss": 1.3013, "step": 3215 }, { "epoch": 0.29, "learning_rate": 4.072855248747506e-05, "loss": 1.2893, "step": 3216 }, { "epoch": 0.29, "learning_rate": 4.0722979712608675e-05, "loss": 1.3069, "step": 3217 }, { "epoch": 0.29, "learning_rate": 4.071740564494466e-05, "loss": 1.3533, "step": 3218 }, { "epoch": 0.29, "learning_rate": 4.071183028494133e-05, "loss": 1.2976, "step": 3219 }, { "epoch": 0.29, "learning_rate": 4.07062536330571e-05, "loss": 1.2559, "step": 3220 }, { "epoch": 0.29, "learning_rate": 4.0700675689750516e-05, "loss": 1.2896, "step": 3221 }, { "epoch": 0.29, "learning_rate": 4.069509645548021e-05, "loss": 1.1729, "step": 3222 }, { "epoch": 0.29, "learning_rate": 4.068951593070493e-05, "loss": 1.2575, "step": 3223 }, { "epoch": 0.29, "learning_rate": 4.068393411588352e-05, "loss": 1.2719, "step": 3224 }, { "epoch": 0.29, "learning_rate": 4.0678351011474924e-05, "loss": 1.3071, "step": 3225 }, { "epoch": 0.29, "learning_rate": 4.067276661793823e-05, "loss": 1.2664, "step": 3226 }, { "epoch": 0.29, "learning_rate": 4.066718093573259e-05, "loss": 1.2777, "step": 3227 }, { "epoch": 0.29, "learning_rate": 4.066159396531728e-05, "loss": 1.3174, "step": 3228 }, { "epoch": 0.29, "learning_rate": 4.065600570715169e-05, "loss": 1.313, "step": 3229 }, { "epoch": 0.29, "learning_rate": 4.065041616169529e-05, "loss": 1.2386, "step": 3230 }, { "epoch": 0.29, "learning_rate": 4.064482532940768e-05, "loss": 1.2614, "step": 3231 }, { "epoch": 0.29, "learning_rate": 4.063923321074856e-05, "loss": 1.2957, "step": 3232 }, { "epoch": 0.29, "learning_rate": 4.063363980617773e-05, "loss": 1.3276, "step": 3233 }, { "epoch": 0.29, "learning_rate": 4.06280451161551e-05, "loss": 1.2738, "step": 3234 }, { "epoch": 0.29, "learning_rate": 4.0622449141140687e-05, "loss": 1.2886, "step": 3235 }, { "epoch": 0.29, "learning_rate": 4.0616851881594595e-05, "loss": 1.3145, "step": 3236 }, { "epoch": 0.29, "learning_rate": 4.0611253337977076e-05, "loss": 1.2863, "step": 3237 }, { "epoch": 0.29, "learning_rate": 4.060565351074844e-05, "loss": 1.2591, "step": 3238 }, { "epoch": 0.29, "learning_rate": 4.060005240036914e-05, "loss": 1.2606, "step": 3239 }, { "epoch": 0.29, "learning_rate": 4.059445000729971e-05, "loss": 1.2455, "step": 3240 }, { "epoch": 0.29, "learning_rate": 4.0588846332000804e-05, "loss": 1.2592, "step": 3241 }, { "epoch": 0.29, "learning_rate": 4.0583241374933164e-05, "loss": 1.254, "step": 3242 }, { "epoch": 0.29, "learning_rate": 4.057763513655767e-05, "loss": 1.2472, "step": 3243 }, { "epoch": 0.29, "learning_rate": 4.057202761733525e-05, "loss": 1.3682, "step": 3244 }, { "epoch": 0.29, "learning_rate": 4.056641881772701e-05, "loss": 1.2341, "step": 3245 }, { "epoch": 0.29, "learning_rate": 4.0560808738194114e-05, "loss": 1.3273, "step": 3246 }, { "epoch": 0.29, "learning_rate": 4.055519737919783e-05, "loss": 1.2982, "step": 3247 }, { "epoch": 0.29, "learning_rate": 4.0549584741199553e-05, "loss": 1.2834, "step": 3248 }, { "epoch": 0.29, "learning_rate": 4.054397082466078e-05, "loss": 1.2989, "step": 3249 }, { "epoch": 0.29, "learning_rate": 4.05383556300431e-05, "loss": 1.2954, "step": 3250 }, { "epoch": 0.29, "learning_rate": 4.05327391578082e-05, "loss": 1.3228, "step": 3251 }, { "epoch": 0.29, "learning_rate": 4.0527121408417906e-05, "loss": 1.2221, "step": 3252 }, { "epoch": 0.29, "learning_rate": 4.052150238233412e-05, "loss": 1.2782, "step": 3253 }, { "epoch": 0.29, "learning_rate": 4.0515882080018864e-05, "loss": 1.3149, "step": 3254 }, { "epoch": 0.29, "learning_rate": 4.051026050193425e-05, "loss": 1.3534, "step": 3255 }, { "epoch": 0.29, "learning_rate": 4.050463764854251e-05, "loss": 1.2488, "step": 3256 }, { "epoch": 0.29, "learning_rate": 4.0499013520305975e-05, "loss": 1.2443, "step": 3257 }, { "epoch": 0.29, "learning_rate": 4.049338811768707e-05, "loss": 1.3338, "step": 3258 }, { "epoch": 0.29, "learning_rate": 4.048776144114834e-05, "loss": 1.2692, "step": 3259 }, { "epoch": 0.29, "learning_rate": 4.048213349115243e-05, "loss": 1.3527, "step": 3260 }, { "epoch": 0.29, "learning_rate": 4.04765042681621e-05, "loss": 1.2385, "step": 3261 }, { "epoch": 0.29, "learning_rate": 4.04708737726402e-05, "loss": 1.3187, "step": 3262 }, { "epoch": 0.29, "learning_rate": 4.046524200504967e-05, "loss": 1.2429, "step": 3263 }, { "epoch": 0.29, "learning_rate": 4.045960896585359e-05, "loss": 1.2289, "step": 3264 }, { "epoch": 0.3, "learning_rate": 4.045397465551513e-05, "loss": 1.2371, "step": 3265 }, { "epoch": 0.3, "learning_rate": 4.044833907449756e-05, "loss": 1.3144, "step": 3266 }, { "epoch": 0.3, "learning_rate": 4.044270222326425e-05, "loss": 1.2781, "step": 3267 }, { "epoch": 0.3, "learning_rate": 4.043706410227869e-05, "loss": 1.2843, "step": 3268 }, { "epoch": 0.3, "learning_rate": 4.0431424712004463e-05, "loss": 1.2688, "step": 3269 }, { "epoch": 0.3, "learning_rate": 4.042578405290525e-05, "loss": 1.2595, "step": 3270 }, { "epoch": 0.3, "learning_rate": 4.042014212544486e-05, "loss": 1.2924, "step": 3271 }, { "epoch": 0.3, "learning_rate": 4.041449893008718e-05, "loss": 1.276, "step": 3272 }, { "epoch": 0.3, "learning_rate": 4.040885446729622e-05, "loss": 1.2749, "step": 3273 }, { "epoch": 0.3, "learning_rate": 4.040320873753609e-05, "loss": 1.332, "step": 3274 }, { "epoch": 0.3, "learning_rate": 4.039756174127099e-05, "loss": 1.2326, "step": 3275 }, { "epoch": 0.3, "learning_rate": 4.039191347896525e-05, "loss": 1.326, "step": 3276 }, { "epoch": 0.3, "learning_rate": 4.038626395108327e-05, "loss": 1.2763, "step": 3277 }, { "epoch": 0.3, "learning_rate": 4.03806131580896e-05, "loss": 1.3238, "step": 3278 }, { "epoch": 0.3, "learning_rate": 4.037496110044884e-05, "loss": 1.3024, "step": 3279 }, { "epoch": 0.3, "learning_rate": 4.0369307778625746e-05, "loss": 1.2706, "step": 3280 }, { "epoch": 0.3, "learning_rate": 4.036365319308514e-05, "loss": 1.2867, "step": 3281 }, { "epoch": 0.3, "learning_rate": 4.035799734429195e-05, "loss": 1.3026, "step": 3282 }, { "epoch": 0.3, "learning_rate": 4.035234023271125e-05, "loss": 1.304, "step": 3283 }, { "epoch": 0.3, "learning_rate": 4.034668185880817e-05, "loss": 1.3322, "step": 3284 }, { "epoch": 0.3, "learning_rate": 4.034102222304795e-05, "loss": 1.2614, "step": 3285 }, { "epoch": 0.3, "learning_rate": 4.033536132589597e-05, "loss": 1.2289, "step": 3286 }, { "epoch": 0.3, "learning_rate": 4.032969916781767e-05, "loss": 1.2459, "step": 3287 }, { "epoch": 0.3, "learning_rate": 4.032403574927861e-05, "loss": 1.3408, "step": 3288 }, { "epoch": 0.3, "learning_rate": 4.0318371070744474e-05, "loss": 1.2651, "step": 3289 }, { "epoch": 0.3, "learning_rate": 4.0312705132681014e-05, "loss": 1.3173, "step": 3290 }, { "epoch": 0.3, "learning_rate": 4.030703793555411e-05, "loss": 1.2779, "step": 3291 }, { "epoch": 0.3, "learning_rate": 4.030136947982974e-05, "loss": 1.3025, "step": 3292 }, { "epoch": 0.3, "learning_rate": 4.029569976597398e-05, "loss": 1.2769, "step": 3293 }, { "epoch": 0.3, "learning_rate": 4.029002879445302e-05, "loss": 1.3212, "step": 3294 }, { "epoch": 0.3, "learning_rate": 4.0284356565733126e-05, "loss": 1.2775, "step": 3295 }, { "epoch": 0.3, "learning_rate": 4.0278683080280724e-05, "loss": 1.2575, "step": 3296 }, { "epoch": 0.3, "learning_rate": 4.027300833856228e-05, "loss": 1.2793, "step": 3297 }, { "epoch": 0.3, "learning_rate": 4.026733234104439e-05, "loss": 1.2687, "step": 3298 }, { "epoch": 0.3, "learning_rate": 4.026165508819378e-05, "loss": 1.2796, "step": 3299 }, { "epoch": 0.3, "learning_rate": 4.0255976580477225e-05, "loss": 1.2935, "step": 3300 }, { "epoch": 0.3, "learning_rate": 4.0250296818361647e-05, "loss": 1.2856, "step": 3301 }, { "epoch": 0.3, "learning_rate": 4.024461580231405e-05, "loss": 1.2774, "step": 3302 }, { "epoch": 0.3, "learning_rate": 4.023893353280155e-05, "loss": 1.2622, "step": 3303 }, { "epoch": 0.3, "learning_rate": 4.023325001029136e-05, "loss": 1.274, "step": 3304 }, { "epoch": 0.3, "learning_rate": 4.0227565235250796e-05, "loss": 1.3084, "step": 3305 }, { "epoch": 0.3, "learning_rate": 4.0221879208147286e-05, "loss": 1.3256, "step": 3306 }, { "epoch": 0.3, "learning_rate": 4.021619192944835e-05, "loss": 1.3042, "step": 3307 }, { "epoch": 0.3, "learning_rate": 4.021050339962162e-05, "loss": 1.2368, "step": 3308 }, { "epoch": 0.3, "learning_rate": 4.020481361913482e-05, "loss": 1.3093, "step": 3309 }, { "epoch": 0.3, "learning_rate": 4.01991225884558e-05, "loss": 1.2715, "step": 3310 }, { "epoch": 0.3, "learning_rate": 4.019343030805247e-05, "loss": 1.2955, "step": 3311 }, { "epoch": 0.3, "learning_rate": 4.018773677839289e-05, "loss": 1.2756, "step": 3312 }, { "epoch": 0.3, "learning_rate": 4.018204199994519e-05, "loss": 1.3216, "step": 3313 }, { "epoch": 0.3, "learning_rate": 4.017634597317762e-05, "loss": 1.2866, "step": 3314 }, { "epoch": 0.3, "learning_rate": 4.017064869855853e-05, "loss": 1.2785, "step": 3315 }, { "epoch": 0.3, "learning_rate": 4.016495017655636e-05, "loss": 1.2358, "step": 3316 }, { "epoch": 0.3, "learning_rate": 4.015925040763968e-05, "loss": 1.3054, "step": 3317 }, { "epoch": 0.3, "learning_rate": 4.015354939227712e-05, "loss": 1.2767, "step": 3318 }, { "epoch": 0.3, "learning_rate": 4.0147847130937454e-05, "loss": 1.3355, "step": 3319 }, { "epoch": 0.3, "learning_rate": 4.014214362408953e-05, "loss": 1.2393, "step": 3320 }, { "epoch": 0.3, "learning_rate": 4.013643887220233e-05, "loss": 1.3031, "step": 3321 }, { "epoch": 0.3, "learning_rate": 4.013073287574489e-05, "loss": 1.3119, "step": 3322 }, { "epoch": 0.3, "learning_rate": 4.012502563518641e-05, "loss": 1.3073, "step": 3323 }, { "epoch": 0.3, "learning_rate": 4.0119317150996125e-05, "loss": 1.3, "step": 3324 }, { "epoch": 0.3, "learning_rate": 4.011360742364343e-05, "loss": 1.3231, "step": 3325 }, { "epoch": 0.3, "learning_rate": 4.010789645359779e-05, "loss": 1.3008, "step": 3326 }, { "epoch": 0.3, "learning_rate": 4.010218424132878e-05, "loss": 1.2541, "step": 3327 }, { "epoch": 0.3, "learning_rate": 4.009647078730608e-05, "loss": 1.2817, "step": 3328 }, { "epoch": 0.3, "learning_rate": 4.009075609199947e-05, "loss": 1.2534, "step": 3329 }, { "epoch": 0.3, "learning_rate": 4.008504015587883e-05, "loss": 1.2766, "step": 3330 }, { "epoch": 0.3, "learning_rate": 4.0079322979414156e-05, "loss": 1.2896, "step": 3331 }, { "epoch": 0.3, "learning_rate": 4.0073604563075505e-05, "loss": 1.31, "step": 3332 }, { "epoch": 0.3, "learning_rate": 4.00678849073331e-05, "loss": 1.3151, "step": 3333 }, { "epoch": 0.3, "learning_rate": 4.00621640126572e-05, "loss": 1.2359, "step": 3334 }, { "epoch": 0.3, "learning_rate": 4.0056441879518216e-05, "loss": 1.2659, "step": 3335 }, { "epoch": 0.3, "learning_rate": 4.005071850838664e-05, "loss": 1.3237, "step": 3336 }, { "epoch": 0.3, "learning_rate": 4.0044993899733055e-05, "loss": 1.2304, "step": 3337 }, { "epoch": 0.3, "learning_rate": 4.0039268054028166e-05, "loss": 1.3035, "step": 3338 }, { "epoch": 0.3, "learning_rate": 4.0033540971742785e-05, "loss": 1.2479, "step": 3339 }, { "epoch": 0.3, "learning_rate": 4.002781265334779e-05, "loss": 1.2651, "step": 3340 }, { "epoch": 0.3, "learning_rate": 4.00220830993142e-05, "loss": 1.3248, "step": 3341 }, { "epoch": 0.3, "learning_rate": 4.00163523101131e-05, "loss": 1.2874, "step": 3342 }, { "epoch": 0.3, "learning_rate": 4.0010620286215713e-05, "loss": 1.28, "step": 3343 }, { "epoch": 0.3, "learning_rate": 4.0004887028093343e-05, "loss": 1.2978, "step": 3344 }, { "epoch": 0.3, "learning_rate": 3.999915253621739e-05, "loss": 1.2705, "step": 3345 }, { "epoch": 0.3, "learning_rate": 3.9993416811059376e-05, "loss": 1.2323, "step": 3346 }, { "epoch": 0.3, "learning_rate": 3.99876798530909e-05, "loss": 1.2607, "step": 3347 }, { "epoch": 0.3, "learning_rate": 3.9981941662783674e-05, "loss": 1.323, "step": 3348 }, { "epoch": 0.3, "learning_rate": 3.997620224060953e-05, "loss": 1.2362, "step": 3349 }, { "epoch": 0.3, "learning_rate": 3.997046158704035e-05, "loss": 1.3162, "step": 3350 }, { "epoch": 0.3, "learning_rate": 3.996471970254818e-05, "loss": 1.3229, "step": 3351 }, { "epoch": 0.3, "learning_rate": 3.9958976587605135e-05, "loss": 1.2299, "step": 3352 }, { "epoch": 0.3, "learning_rate": 3.9953232242683426e-05, "loss": 1.2471, "step": 3353 }, { "epoch": 0.3, "learning_rate": 3.9947486668255366e-05, "loss": 1.2834, "step": 3354 }, { "epoch": 0.3, "learning_rate": 3.994173986479339e-05, "loss": 1.2622, "step": 3355 }, { "epoch": 0.3, "learning_rate": 3.993599183277001e-05, "loss": 1.2801, "step": 3356 }, { "epoch": 0.3, "learning_rate": 3.993024257265785e-05, "loss": 1.2773, "step": 3357 }, { "epoch": 0.3, "learning_rate": 3.992449208492964e-05, "loss": 1.2462, "step": 3358 }, { "epoch": 0.3, "learning_rate": 3.99187403700582e-05, "loss": 1.2963, "step": 3359 }, { "epoch": 0.3, "learning_rate": 3.9912987428516455e-05, "loss": 1.2966, "step": 3360 }, { "epoch": 0.3, "learning_rate": 3.990723326077744e-05, "loss": 1.2883, "step": 3361 }, { "epoch": 0.3, "learning_rate": 3.990147786731427e-05, "loss": 1.2408, "step": 3362 }, { "epoch": 0.3, "learning_rate": 3.989572124860019e-05, "loss": 1.3038, "step": 3363 }, { "epoch": 0.3, "learning_rate": 3.988996340510851e-05, "loss": 1.2817, "step": 3364 }, { "epoch": 0.3, "learning_rate": 3.988420433731267e-05, "loss": 1.2472, "step": 3365 }, { "epoch": 0.3, "learning_rate": 3.98784440456862e-05, "loss": 1.2746, "step": 3366 }, { "epoch": 0.3, "learning_rate": 3.987268253070273e-05, "loss": 1.26, "step": 3367 }, { "epoch": 0.3, "learning_rate": 3.986691979283599e-05, "loss": 1.2454, "step": 3368 }, { "epoch": 0.3, "learning_rate": 3.986115583255982e-05, "loss": 1.3144, "step": 3369 }, { "epoch": 0.3, "learning_rate": 3.985539065034815e-05, "loss": 1.2867, "step": 3370 }, { "epoch": 0.3, "learning_rate": 3.984962424667499e-05, "loss": 1.272, "step": 3371 }, { "epoch": 0.3, "learning_rate": 3.984385662201451e-05, "loss": 1.3608, "step": 3372 }, { "epoch": 0.3, "learning_rate": 3.9838087776840924e-05, "loss": 1.2608, "step": 3373 }, { "epoch": 0.3, "learning_rate": 3.9832317711628566e-05, "loss": 1.2711, "step": 3374 }, { "epoch": 0.3, "learning_rate": 3.982654642685188e-05, "loss": 1.2669, "step": 3375 }, { "epoch": 0.31, "learning_rate": 3.982077392298539e-05, "loss": 1.2164, "step": 3376 }, { "epoch": 0.31, "learning_rate": 3.981500020050374e-05, "loss": 1.3468, "step": 3377 }, { "epoch": 0.31, "learning_rate": 3.980922525988167e-05, "loss": 1.3525, "step": 3378 }, { "epoch": 0.31, "learning_rate": 3.980344910159399e-05, "loss": 1.3145, "step": 3379 }, { "epoch": 0.31, "learning_rate": 3.979767172611567e-05, "loss": 1.2915, "step": 3380 }, { "epoch": 0.31, "learning_rate": 3.979189313392172e-05, "loss": 1.2602, "step": 3381 }, { "epoch": 0.31, "learning_rate": 3.978611332548728e-05, "loss": 1.3081, "step": 3382 }, { "epoch": 0.31, "learning_rate": 3.97803323012876e-05, "loss": 1.3175, "step": 3383 }, { "epoch": 0.31, "learning_rate": 3.9774550061798e-05, "loss": 1.2985, "step": 3384 }, { "epoch": 0.31, "learning_rate": 3.9768766607493925e-05, "loss": 1.2053, "step": 3385 }, { "epoch": 0.31, "learning_rate": 3.9762981938850905e-05, "loss": 1.3074, "step": 3386 }, { "epoch": 0.31, "learning_rate": 3.975719605634458e-05, "loss": 1.2548, "step": 3387 }, { "epoch": 0.31, "learning_rate": 3.975140896045069e-05, "loss": 1.2969, "step": 3388 }, { "epoch": 0.31, "learning_rate": 3.974562065164504e-05, "loss": 1.2727, "step": 3389 }, { "epoch": 0.31, "learning_rate": 3.973983113040361e-05, "loss": 1.2874, "step": 3390 }, { "epoch": 0.31, "learning_rate": 3.973404039720241e-05, "loss": 1.242, "step": 3391 }, { "epoch": 0.31, "learning_rate": 3.972824845251757e-05, "loss": 1.2216, "step": 3392 }, { "epoch": 0.31, "learning_rate": 3.972245529682534e-05, "loss": 1.3254, "step": 3393 }, { "epoch": 0.31, "learning_rate": 3.971666093060203e-05, "loss": 1.3072, "step": 3394 }, { "epoch": 0.31, "learning_rate": 3.971086535432409e-05, "loss": 1.2529, "step": 3395 }, { "epoch": 0.31, "learning_rate": 3.970506856846805e-05, "loss": 1.2881, "step": 3396 }, { "epoch": 0.31, "learning_rate": 3.969927057351054e-05, "loss": 1.2644, "step": 3397 }, { "epoch": 0.31, "learning_rate": 3.9693471369928286e-05, "loss": 1.3219, "step": 3398 }, { "epoch": 0.31, "learning_rate": 3.968767095819813e-05, "loss": 1.2609, "step": 3399 }, { "epoch": 0.31, "learning_rate": 3.9681869338796996e-05, "loss": 1.2667, "step": 3400 }, { "epoch": 0.31, "learning_rate": 3.967606651220191e-05, "loss": 1.3065, "step": 3401 }, { "epoch": 0.31, "learning_rate": 3.967026247889e-05, "loss": 1.2835, "step": 3402 }, { "epoch": 0.31, "learning_rate": 3.9664457239338506e-05, "loss": 1.2839, "step": 3403 }, { "epoch": 0.31, "learning_rate": 3.9658650794024735e-05, "loss": 1.236, "step": 3404 }, { "epoch": 0.31, "learning_rate": 3.9652843143426136e-05, "loss": 1.2904, "step": 3405 }, { "epoch": 0.31, "learning_rate": 3.964703428802021e-05, "loss": 1.2692, "step": 3406 }, { "epoch": 0.31, "learning_rate": 3.96412242282846e-05, "loss": 1.2984, "step": 3407 }, { "epoch": 0.31, "learning_rate": 3.9635412964697014e-05, "loss": 1.3415, "step": 3408 }, { "epoch": 0.31, "learning_rate": 3.9629600497735294e-05, "loss": 1.2686, "step": 3409 }, { "epoch": 0.31, "learning_rate": 3.962378682787734e-05, "loss": 1.2681, "step": 3410 }, { "epoch": 0.31, "learning_rate": 3.961797195560118e-05, "loss": 1.2632, "step": 3411 }, { "epoch": 0.31, "learning_rate": 3.961215588138494e-05, "loss": 1.2779, "step": 3412 }, { "epoch": 0.31, "learning_rate": 3.960633860570683e-05, "loss": 1.274, "step": 3413 }, { "epoch": 0.31, "learning_rate": 3.960052012904517e-05, "loss": 1.3225, "step": 3414 }, { "epoch": 0.31, "learning_rate": 3.9594700451878364e-05, "loss": 1.2668, "step": 3415 }, { "epoch": 0.31, "learning_rate": 3.958887957468494e-05, "loss": 1.2991, "step": 3416 }, { "epoch": 0.31, "learning_rate": 3.958305749794351e-05, "loss": 1.221, "step": 3417 }, { "epoch": 0.31, "learning_rate": 3.9577234222132775e-05, "loss": 1.2953, "step": 3418 }, { "epoch": 0.31, "learning_rate": 3.957140974773156e-05, "loss": 1.2762, "step": 3419 }, { "epoch": 0.31, "learning_rate": 3.956558407521876e-05, "loss": 1.3691, "step": 3420 }, { "epoch": 0.31, "learning_rate": 3.955975720507338e-05, "loss": 1.2509, "step": 3421 }, { "epoch": 0.31, "learning_rate": 3.9553929137774544e-05, "loss": 1.2481, "step": 3422 }, { "epoch": 0.31, "learning_rate": 3.954809987380143e-05, "loss": 1.2189, "step": 3423 }, { "epoch": 0.31, "learning_rate": 3.954226941363336e-05, "loss": 1.3143, "step": 3424 }, { "epoch": 0.31, "learning_rate": 3.9536437757749736e-05, "loss": 1.2997, "step": 3425 }, { "epoch": 0.31, "learning_rate": 3.953060490663004e-05, "loss": 1.2607, "step": 3426 }, { "epoch": 0.31, "learning_rate": 3.952477086075388e-05, "loss": 1.243, "step": 3427 }, { "epoch": 0.31, "learning_rate": 3.951893562060095e-05, "loss": 1.2475, "step": 3428 }, { "epoch": 0.31, "learning_rate": 3.951309918665104e-05, "loss": 1.2144, "step": 3429 }, { "epoch": 0.31, "learning_rate": 3.950726155938405e-05, "loss": 1.2683, "step": 3430 }, { "epoch": 0.31, "learning_rate": 3.9501422739279956e-05, "loss": 1.2954, "step": 3431 }, { "epoch": 0.31, "learning_rate": 3.949558272681886e-05, "loss": 1.3057, "step": 3432 }, { "epoch": 0.31, "learning_rate": 3.9489741522480945e-05, "loss": 1.2856, "step": 3433 }, { "epoch": 0.31, "learning_rate": 3.9483899126746494e-05, "loss": 1.2524, "step": 3434 }, { "epoch": 0.31, "learning_rate": 3.947805554009588e-05, "loss": 1.2623, "step": 3435 }, { "epoch": 0.31, "learning_rate": 3.9472210763009595e-05, "loss": 1.3228, "step": 3436 }, { "epoch": 0.31, "learning_rate": 3.946636479596821e-05, "loss": 1.2901, "step": 3437 }, { "epoch": 0.31, "learning_rate": 3.946051763945241e-05, "loss": 1.2259, "step": 3438 }, { "epoch": 0.31, "learning_rate": 3.9454669293942946e-05, "loss": 1.2899, "step": 3439 }, { "epoch": 0.31, "learning_rate": 3.944881975992072e-05, "loss": 1.2178, "step": 3440 }, { "epoch": 0.31, "learning_rate": 3.944296903786668e-05, "loss": 1.2485, "step": 3441 }, { "epoch": 0.31, "learning_rate": 3.943711712826189e-05, "loss": 1.2894, "step": 3442 }, { "epoch": 0.31, "learning_rate": 3.943126403158754e-05, "loss": 1.3035, "step": 3443 }, { "epoch": 0.31, "learning_rate": 3.942540974832485e-05, "loss": 1.3058, "step": 3444 }, { "epoch": 0.31, "learning_rate": 3.941955427895522e-05, "loss": 1.2679, "step": 3445 }, { "epoch": 0.31, "learning_rate": 3.941369762396009e-05, "loss": 1.2864, "step": 3446 }, { "epoch": 0.31, "learning_rate": 3.9407839783821005e-05, "loss": 1.2415, "step": 3447 }, { "epoch": 0.31, "learning_rate": 3.940198075901963e-05, "loss": 1.3179, "step": 3448 }, { "epoch": 0.31, "learning_rate": 3.9396120550037706e-05, "loss": 1.2482, "step": 3449 }, { "epoch": 0.31, "learning_rate": 3.939025915735709e-05, "loss": 1.2594, "step": 3450 }, { "epoch": 0.31, "learning_rate": 3.938439658145972e-05, "loss": 1.2377, "step": 3451 }, { "epoch": 0.31, "learning_rate": 3.9378532822827636e-05, "loss": 1.3185, "step": 3452 }, { "epoch": 0.31, "learning_rate": 3.937266788194298e-05, "loss": 1.3021, "step": 3453 }, { "epoch": 0.31, "learning_rate": 3.936680175928798e-05, "loss": 1.2762, "step": 3454 }, { "epoch": 0.31, "learning_rate": 3.936093445534498e-05, "loss": 1.2362, "step": 3455 }, { "epoch": 0.31, "learning_rate": 3.93550659705964e-05, "loss": 1.2517, "step": 3456 }, { "epoch": 0.31, "learning_rate": 3.934919630552477e-05, "loss": 1.2734, "step": 3457 }, { "epoch": 0.31, "learning_rate": 3.934332546061272e-05, "loss": 1.3199, "step": 3458 }, { "epoch": 0.31, "learning_rate": 3.933745343634296e-05, "loss": 1.2212, "step": 3459 }, { "epoch": 0.31, "learning_rate": 3.9331580233198314e-05, "loss": 1.2362, "step": 3460 }, { "epoch": 0.31, "learning_rate": 3.932570585166171e-05, "loss": 1.2784, "step": 3461 }, { "epoch": 0.31, "learning_rate": 3.931983029221613e-05, "loss": 1.2336, "step": 3462 }, { "epoch": 0.31, "learning_rate": 3.9313953555344705e-05, "loss": 1.2951, "step": 3463 }, { "epoch": 0.31, "learning_rate": 3.930807564153065e-05, "loss": 1.2526, "step": 3464 }, { "epoch": 0.31, "learning_rate": 3.9302196551257235e-05, "loss": 1.2683, "step": 3465 }, { "epoch": 0.31, "learning_rate": 3.9296316285007887e-05, "loss": 1.2189, "step": 3466 }, { "epoch": 0.31, "learning_rate": 3.929043484326609e-05, "loss": 1.3102, "step": 3467 }, { "epoch": 0.31, "learning_rate": 3.9284552226515437e-05, "loss": 1.2565, "step": 3468 }, { "epoch": 0.31, "learning_rate": 3.927866843523962e-05, "loss": 1.2922, "step": 3469 }, { "epoch": 0.31, "learning_rate": 3.927278346992242e-05, "loss": 1.3117, "step": 3470 }, { "epoch": 0.31, "learning_rate": 3.926689733104772e-05, "loss": 1.2983, "step": 3471 }, { "epoch": 0.31, "learning_rate": 3.926101001909951e-05, "loss": 1.2558, "step": 3472 }, { "epoch": 0.31, "learning_rate": 3.925512153456185e-05, "loss": 1.2431, "step": 3473 }, { "epoch": 0.31, "learning_rate": 3.924923187791892e-05, "loss": 1.2194, "step": 3474 }, { "epoch": 0.31, "learning_rate": 3.9243341049654985e-05, "loss": 1.2864, "step": 3475 }, { "epoch": 0.31, "learning_rate": 3.92374490502544e-05, "loss": 1.2605, "step": 3476 }, { "epoch": 0.31, "learning_rate": 3.9231555880201655e-05, "loss": 1.2708, "step": 3477 }, { "epoch": 0.31, "learning_rate": 3.9225661539981274e-05, "loss": 1.2589, "step": 3478 }, { "epoch": 0.31, "learning_rate": 3.9219766030077924e-05, "loss": 1.2541, "step": 3479 }, { "epoch": 0.31, "learning_rate": 3.921386935097635e-05, "loss": 1.3086, "step": 3480 }, { "epoch": 0.31, "learning_rate": 3.92079715031614e-05, "loss": 1.317, "step": 3481 }, { "epoch": 0.31, "learning_rate": 3.920207248711802e-05, "loss": 1.2752, "step": 3482 }, { "epoch": 0.31, "learning_rate": 3.9196172303331246e-05, "loss": 1.2644, "step": 3483 }, { "epoch": 0.31, "learning_rate": 3.9190270952286204e-05, "loss": 1.2459, "step": 3484 }, { "epoch": 0.31, "learning_rate": 3.918436843446813e-05, "loss": 1.2608, "step": 3485 }, { "epoch": 0.31, "learning_rate": 3.917846475036234e-05, "loss": 1.2563, "step": 3486 }, { "epoch": 0.32, "learning_rate": 3.917255990045428e-05, "loss": 1.2688, "step": 3487 }, { "epoch": 0.32, "learning_rate": 3.9166653885229434e-05, "loss": 1.2603, "step": 3488 }, { "epoch": 0.32, "learning_rate": 3.916074670517343e-05, "loss": 1.2775, "step": 3489 }, { "epoch": 0.32, "learning_rate": 3.9154838360772e-05, "loss": 1.2716, "step": 3490 }, { "epoch": 0.32, "learning_rate": 3.91489288525109e-05, "loss": 1.256, "step": 3491 }, { "epoch": 0.32, "learning_rate": 3.914301818087607e-05, "loss": 1.2732, "step": 3492 }, { "epoch": 0.32, "learning_rate": 3.913710634635348e-05, "loss": 1.2747, "step": 3493 }, { "epoch": 0.32, "learning_rate": 3.913119334942924e-05, "loss": 1.2756, "step": 3494 }, { "epoch": 0.32, "learning_rate": 3.9125279190589534e-05, "loss": 1.2633, "step": 3495 }, { "epoch": 0.32, "learning_rate": 3.9119363870320637e-05, "loss": 1.2755, "step": 3496 }, { "epoch": 0.32, "learning_rate": 3.911344738910894e-05, "loss": 1.2693, "step": 3497 }, { "epoch": 0.32, "learning_rate": 3.91075297474409e-05, "loss": 1.3003, "step": 3498 }, { "epoch": 0.32, "learning_rate": 3.910161094580309e-05, "loss": 1.3665, "step": 3499 }, { "epoch": 0.32, "learning_rate": 3.9095690984682176e-05, "loss": 1.3103, "step": 3500 }, { "epoch": 0.32, "learning_rate": 3.9089769864564925e-05, "loss": 1.2921, "step": 3501 }, { "epoch": 0.32, "learning_rate": 3.908384758593818e-05, "loss": 1.2865, "step": 3502 }, { "epoch": 0.32, "learning_rate": 3.90779241492889e-05, "loss": 1.293, "step": 3503 }, { "epoch": 0.32, "learning_rate": 3.907199955510413e-05, "loss": 1.2972, "step": 3504 }, { "epoch": 0.32, "learning_rate": 3.9066073803871016e-05, "loss": 1.293, "step": 3505 }, { "epoch": 0.32, "learning_rate": 3.906014689607678e-05, "loss": 1.2996, "step": 3506 }, { "epoch": 0.32, "learning_rate": 3.905421883220876e-05, "loss": 1.2752, "step": 3507 }, { "epoch": 0.32, "learning_rate": 3.9048289612754374e-05, "loss": 1.2274, "step": 3508 }, { "epoch": 0.32, "learning_rate": 3.9042359238201154e-05, "loss": 1.2707, "step": 3509 }, { "epoch": 0.32, "learning_rate": 3.90364277090367e-05, "loss": 1.2471, "step": 3510 }, { "epoch": 0.32, "learning_rate": 3.9030495025748756e-05, "loss": 1.3395, "step": 3511 }, { "epoch": 0.32, "learning_rate": 3.902456118882509e-05, "loss": 1.2367, "step": 3512 }, { "epoch": 0.32, "learning_rate": 3.901862619875363e-05, "loss": 1.3076, "step": 3513 }, { "epoch": 0.32, "learning_rate": 3.901269005602235e-05, "loss": 1.2946, "step": 3514 }, { "epoch": 0.32, "learning_rate": 3.900675276111936e-05, "loss": 1.2391, "step": 3515 }, { "epoch": 0.32, "learning_rate": 3.900081431453283e-05, "loss": 1.2824, "step": 3516 }, { "epoch": 0.32, "learning_rate": 3.899487471675104e-05, "loss": 1.2821, "step": 3517 }, { "epoch": 0.32, "learning_rate": 3.898893396826237e-05, "loss": 1.2701, "step": 3518 }, { "epoch": 0.32, "learning_rate": 3.898299206955529e-05, "loss": 1.2291, "step": 3519 }, { "epoch": 0.32, "learning_rate": 3.8977049021118365e-05, "loss": 1.3175, "step": 3520 }, { "epoch": 0.32, "learning_rate": 3.897110482344024e-05, "loss": 1.2546, "step": 3521 }, { "epoch": 0.32, "learning_rate": 3.896515947700968e-05, "loss": 1.2976, "step": 3522 }, { "epoch": 0.32, "learning_rate": 3.8959212982315524e-05, "loss": 1.2177, "step": 3523 }, { "epoch": 0.32, "learning_rate": 3.8953265339846725e-05, "loss": 1.2853, "step": 3524 }, { "epoch": 0.32, "learning_rate": 3.89473165500923e-05, "loss": 1.3365, "step": 3525 }, { "epoch": 0.32, "learning_rate": 3.894136661354141e-05, "loss": 1.2766, "step": 3526 }, { "epoch": 0.32, "learning_rate": 3.8935415530683246e-05, "loss": 1.2773, "step": 3527 }, { "epoch": 0.32, "learning_rate": 3.8929463302007144e-05, "loss": 1.2364, "step": 3528 }, { "epoch": 0.32, "learning_rate": 3.892350992800251e-05, "loss": 1.2656, "step": 3529 }, { "epoch": 0.32, "learning_rate": 3.8917555409158843e-05, "loss": 1.2778, "step": 3530 }, { "epoch": 0.32, "learning_rate": 3.8911599745965775e-05, "loss": 1.2407, "step": 3531 }, { "epoch": 0.32, "learning_rate": 3.890564293891297e-05, "loss": 1.258, "step": 3532 }, { "epoch": 0.32, "learning_rate": 3.8899684988490234e-05, "loss": 1.3376, "step": 3533 }, { "epoch": 0.32, "learning_rate": 3.889372589518744e-05, "loss": 1.2918, "step": 3534 }, { "epoch": 0.32, "learning_rate": 3.888776565949457e-05, "loss": 1.2651, "step": 3535 }, { "epoch": 0.32, "learning_rate": 3.888180428190169e-05, "loss": 1.2966, "step": 3536 }, { "epoch": 0.32, "learning_rate": 3.8875841762898984e-05, "loss": 1.303, "step": 3537 }, { "epoch": 0.32, "learning_rate": 3.886987810297669e-05, "loss": 1.2702, "step": 3538 }, { "epoch": 0.32, "learning_rate": 3.886391330262517e-05, "loss": 1.262, "step": 3539 }, { "epoch": 0.32, "learning_rate": 3.885794736233486e-05, "loss": 1.3249, "step": 3540 }, { "epoch": 0.32, "learning_rate": 3.8851980282596306e-05, "loss": 1.2644, "step": 3541 }, { "epoch": 0.32, "learning_rate": 3.884601206390015e-05, "loss": 1.2796, "step": 3542 }, { "epoch": 0.32, "learning_rate": 3.8840042706737114e-05, "loss": 1.299, "step": 3543 }, { "epoch": 0.32, "learning_rate": 3.8834072211598024e-05, "loss": 1.2475, "step": 3544 }, { "epoch": 0.32, "learning_rate": 3.882810057897378e-05, "loss": 1.3129, "step": 3545 }, { "epoch": 0.32, "learning_rate": 3.88221278093554e-05, "loss": 1.2306, "step": 3546 }, { "epoch": 0.32, "learning_rate": 3.881615390323399e-05, "loss": 1.2561, "step": 3547 }, { "epoch": 0.32, "learning_rate": 3.8810178861100745e-05, "loss": 1.2345, "step": 3548 }, { "epoch": 0.32, "learning_rate": 3.8804202683446933e-05, "loss": 1.3371, "step": 3549 }, { "epoch": 0.32, "learning_rate": 3.8798225370763965e-05, "loss": 1.2854, "step": 3550 }, { "epoch": 0.32, "learning_rate": 3.87922469235433e-05, "loss": 1.276, "step": 3551 }, { "epoch": 0.32, "learning_rate": 3.8786267342276515e-05, "loss": 1.3222, "step": 3552 }, { "epoch": 0.32, "learning_rate": 3.878028662745526e-05, "loss": 1.2874, "step": 3553 }, { "epoch": 0.32, "learning_rate": 3.87743047795713e-05, "loss": 1.2963, "step": 3554 }, { "epoch": 0.32, "learning_rate": 3.876832179911648e-05, "loss": 1.2971, "step": 3555 }, { "epoch": 0.32, "learning_rate": 3.8762337686582736e-05, "loss": 1.247, "step": 3556 }, { "epoch": 0.32, "learning_rate": 3.8756352442462116e-05, "loss": 1.2826, "step": 3557 }, { "epoch": 0.32, "learning_rate": 3.875036606724673e-05, "loss": 1.3082, "step": 3558 }, { "epoch": 0.32, "learning_rate": 3.8744378561428814e-05, "loss": 1.2935, "step": 3559 }, { "epoch": 0.32, "learning_rate": 3.8738389925500675e-05, "loss": 1.264, "step": 3560 }, { "epoch": 0.32, "learning_rate": 3.8732400159954713e-05, "loss": 1.2559, "step": 3561 }, { "epoch": 0.32, "learning_rate": 3.8726409265283435e-05, "loss": 1.29, "step": 3562 }, { "epoch": 0.32, "learning_rate": 3.872041724197943e-05, "loss": 1.2588, "step": 3563 }, { "epoch": 0.32, "learning_rate": 3.8714424090535397e-05, "loss": 1.3144, "step": 3564 }, { "epoch": 0.32, "learning_rate": 3.870842981144409e-05, "loss": 1.2825, "step": 3565 }, { "epoch": 0.32, "learning_rate": 3.870243440519839e-05, "loss": 1.286, "step": 3566 }, { "epoch": 0.32, "learning_rate": 3.869643787229126e-05, "loss": 1.2519, "step": 3567 }, { "epoch": 0.32, "learning_rate": 3.869044021321575e-05, "loss": 1.2434, "step": 3568 }, { "epoch": 0.32, "learning_rate": 3.868444142846502e-05, "loss": 1.2975, "step": 3569 }, { "epoch": 0.32, "learning_rate": 3.8678441518532304e-05, "loss": 1.2892, "step": 3570 }, { "epoch": 0.32, "learning_rate": 3.867244048391094e-05, "loss": 1.27, "step": 3571 }, { "epoch": 0.32, "learning_rate": 3.8666438325094346e-05, "loss": 1.2346, "step": 3572 }, { "epoch": 0.32, "learning_rate": 3.8660435042576046e-05, "loss": 1.2746, "step": 3573 }, { "epoch": 0.32, "learning_rate": 3.8654430636849645e-05, "loss": 1.2423, "step": 3574 }, { "epoch": 0.32, "learning_rate": 3.864842510840885e-05, "loss": 1.2797, "step": 3575 }, { "epoch": 0.32, "learning_rate": 3.864241845774746e-05, "loss": 1.3341, "step": 3576 }, { "epoch": 0.32, "learning_rate": 3.863641068535935e-05, "loss": 1.3221, "step": 3577 }, { "epoch": 0.32, "learning_rate": 3.863040179173851e-05, "loss": 1.3111, "step": 3578 }, { "epoch": 0.32, "learning_rate": 3.862439177737902e-05, "loss": 1.2397, "step": 3579 }, { "epoch": 0.32, "learning_rate": 3.861838064277502e-05, "loss": 1.2931, "step": 3580 }, { "epoch": 0.32, "learning_rate": 3.86123683884208e-05, "loss": 1.2764, "step": 3581 }, { "epoch": 0.32, "learning_rate": 3.860635501481067e-05, "loss": 1.2835, "step": 3582 }, { "epoch": 0.32, "learning_rate": 3.860034052243909e-05, "loss": 1.3019, "step": 3583 }, { "epoch": 0.32, "learning_rate": 3.85943249118006e-05, "loss": 1.3288, "step": 3584 }, { "epoch": 0.32, "learning_rate": 3.858830818338981e-05, "loss": 1.2833, "step": 3585 }, { "epoch": 0.32, "learning_rate": 3.858229033770144e-05, "loss": 1.2603, "step": 3586 }, { "epoch": 0.32, "learning_rate": 3.857627137523031e-05, "loss": 1.2677, "step": 3587 }, { "epoch": 0.32, "learning_rate": 3.85702512964713e-05, "loss": 1.313, "step": 3588 }, { "epoch": 0.32, "learning_rate": 3.856423010191941e-05, "loss": 1.2821, "step": 3589 }, { "epoch": 0.32, "learning_rate": 3.855820779206974e-05, "loss": 1.2785, "step": 3590 }, { "epoch": 0.32, "learning_rate": 3.855218436741743e-05, "loss": 1.307, "step": 3591 }, { "epoch": 0.32, "learning_rate": 3.8546159828457774e-05, "loss": 1.2671, "step": 3592 }, { "epoch": 0.32, "learning_rate": 3.854013417568612e-05, "loss": 1.2261, "step": 3593 }, { "epoch": 0.32, "learning_rate": 3.8534107409597933e-05, "loss": 1.2813, "step": 3594 }, { "epoch": 0.32, "learning_rate": 3.852807953068873e-05, "loss": 1.2734, "step": 3595 }, { "epoch": 0.32, "learning_rate": 3.852205053945416e-05, "loss": 1.2933, "step": 3596 }, { "epoch": 0.32, "learning_rate": 3.851602043638994e-05, "loss": 1.2502, "step": 3597 }, { "epoch": 0.33, "learning_rate": 3.850998922199189e-05, "loss": 1.2863, "step": 3598 }, { "epoch": 0.33, "learning_rate": 3.8503956896755925e-05, "loss": 1.2559, "step": 3599 }, { "epoch": 0.33, "learning_rate": 3.849792346117803e-05, "loss": 1.3353, "step": 3600 }, { "epoch": 0.33, "learning_rate": 3.8491888915754306e-05, "loss": 1.2935, "step": 3601 }, { "epoch": 0.33, "learning_rate": 3.848585326098092e-05, "loss": 1.293, "step": 3602 }, { "epoch": 0.33, "learning_rate": 3.847981649735415e-05, "loss": 1.2795, "step": 3603 }, { "epoch": 0.33, "learning_rate": 3.847377862537037e-05, "loss": 1.2572, "step": 3604 }, { "epoch": 0.33, "learning_rate": 3.846773964552602e-05, "loss": 1.2576, "step": 3605 }, { "epoch": 0.33, "learning_rate": 3.846169955831766e-05, "loss": 1.3206, "step": 3606 }, { "epoch": 0.33, "learning_rate": 3.845565836424193e-05, "loss": 1.2748, "step": 3607 }, { "epoch": 0.33, "learning_rate": 3.844961606379553e-05, "loss": 1.2505, "step": 3608 }, { "epoch": 0.33, "learning_rate": 3.8443572657475304e-05, "loss": 1.2763, "step": 3609 }, { "epoch": 0.33, "learning_rate": 3.8437528145778154e-05, "loss": 1.2935, "step": 3610 }, { "epoch": 0.33, "learning_rate": 3.8431482529201076e-05, "loss": 1.2938, "step": 3611 }, { "epoch": 0.33, "learning_rate": 3.8425435808241165e-05, "loss": 1.328, "step": 3612 }, { "epoch": 0.33, "learning_rate": 3.841938798339561e-05, "loss": 1.2729, "step": 3613 }, { "epoch": 0.33, "learning_rate": 3.841333905516168e-05, "loss": 1.2894, "step": 3614 }, { "epoch": 0.33, "learning_rate": 3.840728902403673e-05, "loss": 1.2646, "step": 3615 }, { "epoch": 0.33, "learning_rate": 3.840123789051822e-05, "loss": 1.2553, "step": 3616 }, { "epoch": 0.33, "learning_rate": 3.8395185655103705e-05, "loss": 1.2418, "step": 3617 }, { "epoch": 0.33, "learning_rate": 3.838913231829081e-05, "loss": 1.2289, "step": 3618 }, { "epoch": 0.33, "learning_rate": 3.838307788057725e-05, "loss": 1.2655, "step": 3619 }, { "epoch": 0.33, "learning_rate": 3.837702234246088e-05, "loss": 1.2565, "step": 3620 }, { "epoch": 0.33, "learning_rate": 3.8370965704439565e-05, "loss": 1.2937, "step": 3621 }, { "epoch": 0.33, "learning_rate": 3.8364907967011324e-05, "loss": 1.3105, "step": 3622 }, { "epoch": 0.33, "learning_rate": 3.835884913067424e-05, "loss": 1.2823, "step": 3623 }, { "epoch": 0.33, "learning_rate": 3.835278919592649e-05, "loss": 1.2365, "step": 3624 }, { "epoch": 0.33, "learning_rate": 3.834672816326635e-05, "loss": 1.2459, "step": 3625 }, { "epoch": 0.33, "learning_rate": 3.8340666033192176e-05, "loss": 1.2698, "step": 3626 }, { "epoch": 0.33, "learning_rate": 3.833460280620241e-05, "loss": 1.2567, "step": 3627 }, { "epoch": 0.33, "learning_rate": 3.832853848279561e-05, "loss": 1.2216, "step": 3628 }, { "epoch": 0.33, "learning_rate": 3.832247306347038e-05, "loss": 1.2768, "step": 3629 }, { "epoch": 0.33, "learning_rate": 3.8316406548725454e-05, "loss": 1.312, "step": 3630 }, { "epoch": 0.33, "learning_rate": 3.8310338939059644e-05, "loss": 1.2851, "step": 3631 }, { "epoch": 0.33, "learning_rate": 3.830427023497184e-05, "loss": 1.2569, "step": 3632 }, { "epoch": 0.33, "learning_rate": 3.829820043696104e-05, "loss": 1.2748, "step": 3633 }, { "epoch": 0.33, "learning_rate": 3.829212954552632e-05, "loss": 1.2941, "step": 3634 }, { "epoch": 0.33, "learning_rate": 3.828605756116686e-05, "loss": 1.287, "step": 3635 }, { "epoch": 0.33, "learning_rate": 3.8279984484381905e-05, "loss": 1.2755, "step": 3636 }, { "epoch": 0.33, "learning_rate": 3.827391031567081e-05, "loss": 1.2838, "step": 3637 }, { "epoch": 0.33, "learning_rate": 3.826783505553301e-05, "loss": 1.2662, "step": 3638 }, { "epoch": 0.33, "learning_rate": 3.8261758704468044e-05, "loss": 1.2872, "step": 3639 }, { "epoch": 0.33, "learning_rate": 3.8255681262975515e-05, "loss": 1.3394, "step": 3640 }, { "epoch": 0.33, "learning_rate": 3.8249602731555156e-05, "loss": 1.2516, "step": 3641 }, { "epoch": 0.33, "learning_rate": 3.8243523110706736e-05, "loss": 1.2599, "step": 3642 }, { "epoch": 0.33, "learning_rate": 3.823744240093016e-05, "loss": 1.2467, "step": 3643 }, { "epoch": 0.33, "learning_rate": 3.823136060272541e-05, "loss": 1.2833, "step": 3644 }, { "epoch": 0.33, "learning_rate": 3.822527771659253e-05, "loss": 1.2873, "step": 3645 }, { "epoch": 0.33, "learning_rate": 3.821919374303169e-05, "loss": 1.2803, "step": 3646 }, { "epoch": 0.33, "learning_rate": 3.821310868254314e-05, "loss": 1.2672, "step": 3647 }, { "epoch": 0.33, "learning_rate": 3.820702253562721e-05, "loss": 1.2995, "step": 3648 }, { "epoch": 0.33, "learning_rate": 3.820093530278432e-05, "loss": 1.288, "step": 3649 }, { "epoch": 0.33, "learning_rate": 3.819484698451498e-05, "loss": 1.2661, "step": 3650 }, { "epoch": 0.33, "learning_rate": 3.8188757581319815e-05, "loss": 1.3028, "step": 3651 }, { "epoch": 0.33, "learning_rate": 3.818266709369949e-05, "loss": 1.2777, "step": 3652 }, { "epoch": 0.33, "learning_rate": 3.817657552215479e-05, "loss": 1.2778, "step": 3653 }, { "epoch": 0.33, "learning_rate": 3.81704828671866e-05, "loss": 1.2693, "step": 3654 }, { "epoch": 0.33, "learning_rate": 3.8164389129295866e-05, "loss": 1.3084, "step": 3655 }, { "epoch": 0.33, "learning_rate": 3.815829430898365e-05, "loss": 1.3243, "step": 3656 }, { "epoch": 0.33, "learning_rate": 3.8152198406751074e-05, "loss": 1.2234, "step": 3657 }, { "epoch": 0.33, "learning_rate": 3.814610142309937e-05, "loss": 1.2471, "step": 3658 }, { "epoch": 0.33, "learning_rate": 3.8140003358529855e-05, "loss": 1.2772, "step": 3659 }, { "epoch": 0.33, "learning_rate": 3.813390421354393e-05, "loss": 1.3198, "step": 3660 }, { "epoch": 0.33, "learning_rate": 3.8127803988643086e-05, "loss": 1.2942, "step": 3661 }, { "epoch": 0.33, "learning_rate": 3.812170268432891e-05, "loss": 1.3, "step": 3662 }, { "epoch": 0.33, "learning_rate": 3.8115600301103074e-05, "loss": 1.299, "step": 3663 }, { "epoch": 0.33, "learning_rate": 3.810949683946733e-05, "loss": 1.2158, "step": 3664 }, { "epoch": 0.33, "learning_rate": 3.810339229992352e-05, "loss": 1.2724, "step": 3665 }, { "epoch": 0.33, "learning_rate": 3.8097286682973596e-05, "loss": 1.2963, "step": 3666 }, { "epoch": 0.33, "learning_rate": 3.8091179989119575e-05, "loss": 1.3348, "step": 3667 }, { "epoch": 0.33, "learning_rate": 3.808507221886357e-05, "loss": 1.285, "step": 3668 }, { "epoch": 0.33, "learning_rate": 3.807896337270779e-05, "loss": 1.3061, "step": 3669 }, { "epoch": 0.33, "learning_rate": 3.807285345115451e-05, "loss": 1.2272, "step": 3670 }, { "epoch": 0.33, "learning_rate": 3.806674245470612e-05, "loss": 1.2893, "step": 3671 }, { "epoch": 0.33, "learning_rate": 3.806063038386509e-05, "loss": 1.2442, "step": 3672 }, { "epoch": 0.33, "learning_rate": 3.805451723913397e-05, "loss": 1.2953, "step": 3673 }, { "epoch": 0.33, "learning_rate": 3.80484030210154e-05, "loss": 1.2127, "step": 3674 }, { "epoch": 0.33, "learning_rate": 3.804228773001212e-05, "loss": 1.2483, "step": 3675 }, { "epoch": 0.33, "learning_rate": 3.8036171366626936e-05, "loss": 1.3074, "step": 3676 }, { "epoch": 0.33, "learning_rate": 3.803005393136279e-05, "loss": 1.2454, "step": 3677 }, { "epoch": 0.33, "learning_rate": 3.802393542472263e-05, "loss": 1.2152, "step": 3678 }, { "epoch": 0.33, "learning_rate": 3.8017815847209586e-05, "loss": 1.2556, "step": 3679 }, { "epoch": 0.33, "learning_rate": 3.801169519932681e-05, "loss": 1.336, "step": 3680 }, { "epoch": 0.33, "learning_rate": 3.800557348157756e-05, "loss": 1.2782, "step": 3681 }, { "epoch": 0.33, "learning_rate": 3.7999450694465194e-05, "loss": 1.2389, "step": 3682 }, { "epoch": 0.33, "learning_rate": 3.799332683849314e-05, "loss": 1.2694, "step": 3683 }, { "epoch": 0.33, "learning_rate": 3.798720191416492e-05, "loss": 1.2353, "step": 3684 }, { "epoch": 0.33, "learning_rate": 3.798107592198417e-05, "loss": 1.226, "step": 3685 }, { "epoch": 0.33, "learning_rate": 3.797494886245456e-05, "loss": 1.268, "step": 3686 }, { "epoch": 0.33, "learning_rate": 3.796882073607989e-05, "loss": 1.2947, "step": 3687 }, { "epoch": 0.33, "learning_rate": 3.796269154336404e-05, "loss": 1.2582, "step": 3688 }, { "epoch": 0.33, "learning_rate": 3.795656128481097e-05, "loss": 1.2799, "step": 3689 }, { "epoch": 0.33, "learning_rate": 3.795042996092474e-05, "loss": 1.2372, "step": 3690 }, { "epoch": 0.33, "learning_rate": 3.794429757220946e-05, "loss": 1.2722, "step": 3691 }, { "epoch": 0.33, "learning_rate": 3.7938164119169394e-05, "loss": 1.2586, "step": 3692 }, { "epoch": 0.33, "learning_rate": 3.793202960230883e-05, "loss": 1.2985, "step": 3693 }, { "epoch": 0.33, "learning_rate": 3.792589402213218e-05, "loss": 1.286, "step": 3694 }, { "epoch": 0.33, "learning_rate": 3.791975737914393e-05, "loss": 1.2813, "step": 3695 }, { "epoch": 0.33, "learning_rate": 3.791361967384865e-05, "loss": 1.258, "step": 3696 }, { "epoch": 0.33, "learning_rate": 3.7907480906751014e-05, "loss": 1.2842, "step": 3697 }, { "epoch": 0.33, "learning_rate": 3.790134107835577e-05, "loss": 1.2584, "step": 3698 }, { "epoch": 0.33, "learning_rate": 3.789520018916775e-05, "loss": 1.2522, "step": 3699 }, { "epoch": 0.33, "learning_rate": 3.7889058239691874e-05, "loss": 1.2508, "step": 3700 }, { "epoch": 0.33, "learning_rate": 3.788291523043317e-05, "loss": 1.3083, "step": 3701 }, { "epoch": 0.33, "learning_rate": 3.787677116189673e-05, "loss": 1.231, "step": 3702 }, { "epoch": 0.33, "learning_rate": 3.7870626034587744e-05, "loss": 1.2952, "step": 3703 }, { "epoch": 0.33, "learning_rate": 3.7864479849011466e-05, "loss": 1.271, "step": 3704 }, { "epoch": 0.33, "learning_rate": 3.7858332605673285e-05, "loss": 1.3628, "step": 3705 }, { "epoch": 0.33, "learning_rate": 3.785218430507863e-05, "loss": 1.2708, "step": 3706 }, { "epoch": 0.33, "learning_rate": 3.784603494773305e-05, "loss": 1.3019, "step": 3707 }, { "epoch": 0.34, "learning_rate": 3.7839884534142154e-05, "loss": 1.2582, "step": 3708 }, { "epoch": 0.34, "learning_rate": 3.7833733064811655e-05, "loss": 1.2398, "step": 3709 }, { "epoch": 0.34, "learning_rate": 3.782758054024734e-05, "loss": 1.2791, "step": 3710 }, { "epoch": 0.34, "learning_rate": 3.7821426960955106e-05, "loss": 1.2353, "step": 3711 }, { "epoch": 0.34, "learning_rate": 3.7815272327440915e-05, "loss": 1.34, "step": 3712 }, { "epoch": 0.34, "learning_rate": 3.780911664021082e-05, "loss": 1.2797, "step": 3713 }, { "epoch": 0.34, "learning_rate": 3.780295989977096e-05, "loss": 1.288, "step": 3714 }, { "epoch": 0.34, "learning_rate": 3.779680210662757e-05, "loss": 1.2721, "step": 3715 }, { "epoch": 0.34, "learning_rate": 3.7790643261286965e-05, "loss": 1.2555, "step": 3716 }, { "epoch": 0.34, "learning_rate": 3.778448336425553e-05, "loss": 1.2879, "step": 3717 }, { "epoch": 0.34, "learning_rate": 3.7778322416039765e-05, "loss": 1.2375, "step": 3718 }, { "epoch": 0.34, "learning_rate": 3.7772160417146265e-05, "loss": 1.3071, "step": 3719 }, { "epoch": 0.34, "learning_rate": 3.776599736808166e-05, "loss": 1.2345, "step": 3720 }, { "epoch": 0.34, "learning_rate": 3.775983326935272e-05, "loss": 1.2921, "step": 3721 }, { "epoch": 0.34, "learning_rate": 3.7753668121466256e-05, "loss": 1.2989, "step": 3722 }, { "epoch": 0.34, "learning_rate": 3.77475019249292e-05, "loss": 1.1779, "step": 3723 }, { "epoch": 0.34, "learning_rate": 3.7741334680248565e-05, "loss": 1.2392, "step": 3724 }, { "epoch": 0.34, "learning_rate": 3.773516638793142e-05, "loss": 1.2965, "step": 3725 }, { "epoch": 0.34, "learning_rate": 3.772899704848497e-05, "loss": 1.2272, "step": 3726 }, { "epoch": 0.34, "learning_rate": 3.772282666241647e-05, "loss": 1.2963, "step": 3727 }, { "epoch": 0.34, "learning_rate": 3.771665523023325e-05, "loss": 1.1983, "step": 3728 }, { "epoch": 0.34, "learning_rate": 3.771048275244279e-05, "loss": 1.2472, "step": 3729 }, { "epoch": 0.34, "learning_rate": 3.770430922955257e-05, "loss": 1.234, "step": 3730 }, { "epoch": 0.34, "learning_rate": 3.7698134662070215e-05, "loss": 1.2416, "step": 3731 }, { "epoch": 0.34, "learning_rate": 3.769195905050344e-05, "loss": 1.3292, "step": 3732 }, { "epoch": 0.34, "learning_rate": 3.768578239535998e-05, "loss": 1.282, "step": 3733 }, { "epoch": 0.34, "learning_rate": 3.767960469714774e-05, "loss": 1.2912, "step": 3734 }, { "epoch": 0.34, "learning_rate": 3.7673425956374654e-05, "loss": 1.2619, "step": 3735 }, { "epoch": 0.34, "learning_rate": 3.766724617354876e-05, "loss": 1.2862, "step": 3736 }, { "epoch": 0.34, "learning_rate": 3.766106534917819e-05, "loss": 1.3144, "step": 3737 }, { "epoch": 0.34, "learning_rate": 3.7654883483771136e-05, "loss": 1.2597, "step": 3738 }, { "epoch": 0.34, "learning_rate": 3.7648700577835906e-05, "loss": 1.295, "step": 3739 }, { "epoch": 0.34, "learning_rate": 3.764251663188088e-05, "loss": 1.2557, "step": 3740 }, { "epoch": 0.34, "learning_rate": 3.7636331646414524e-05, "loss": 1.3394, "step": 3741 }, { "epoch": 0.34, "learning_rate": 3.763014562194538e-05, "loss": 1.2706, "step": 3742 }, { "epoch": 0.34, "learning_rate": 3.762395855898209e-05, "loss": 1.2625, "step": 3743 }, { "epoch": 0.34, "learning_rate": 3.761777045803337e-05, "loss": 1.3078, "step": 3744 }, { "epoch": 0.34, "learning_rate": 3.761158131960804e-05, "loss": 1.2825, "step": 3745 }, { "epoch": 0.34, "learning_rate": 3.760539114421498e-05, "loss": 1.2344, "step": 3746 }, { "epoch": 0.34, "learning_rate": 3.759919993236317e-05, "loss": 1.3099, "step": 3747 }, { "epoch": 0.34, "learning_rate": 3.759300768456168e-05, "loss": 1.2273, "step": 3748 }, { "epoch": 0.34, "learning_rate": 3.7586814401319646e-05, "loss": 1.2558, "step": 3749 }, { "epoch": 0.34, "learning_rate": 3.758062008314631e-05, "loss": 1.2117, "step": 3750 }, { "epoch": 0.34, "learning_rate": 3.757442473055099e-05, "loss": 1.2585, "step": 3751 }, { "epoch": 0.34, "learning_rate": 3.756822834404308e-05, "loss": 1.282, "step": 3752 }, { "epoch": 0.34, "learning_rate": 3.756203092413209e-05, "loss": 1.2792, "step": 3753 }, { "epoch": 0.34, "learning_rate": 3.7555832471327565e-05, "loss": 1.2909, "step": 3754 }, { "epoch": 0.34, "learning_rate": 3.754963298613918e-05, "loss": 1.2474, "step": 3755 }, { "epoch": 0.34, "learning_rate": 3.754343246907667e-05, "loss": 1.2677, "step": 3756 }, { "epoch": 0.34, "learning_rate": 3.753723092064988e-05, "loss": 1.2062, "step": 3757 }, { "epoch": 0.34, "learning_rate": 3.7531028341368705e-05, "loss": 1.3092, "step": 3758 }, { "epoch": 0.34, "learning_rate": 3.7524824731743146e-05, "loss": 1.2546, "step": 3759 }, { "epoch": 0.34, "learning_rate": 3.751862009228328e-05, "loss": 1.2722, "step": 3760 }, { "epoch": 0.34, "learning_rate": 3.75124144234993e-05, "loss": 1.2545, "step": 3761 }, { "epoch": 0.34, "learning_rate": 3.7506207725901424e-05, "loss": 1.2567, "step": 3762 }, { "epoch": 0.34, "learning_rate": 3.7500000000000003e-05, "loss": 1.2713, "step": 3763 }, { "epoch": 0.34, "learning_rate": 3.749379124630546e-05, "loss": 1.2888, "step": 3764 }, { "epoch": 0.34, "learning_rate": 3.7487581465328295e-05, "loss": 1.2836, "step": 3765 }, { "epoch": 0.34, "learning_rate": 3.748137065757911e-05, "loss": 1.2908, "step": 3766 }, { "epoch": 0.34, "learning_rate": 3.747515882356856e-05, "loss": 1.2077, "step": 3767 }, { "epoch": 0.34, "learning_rate": 3.746894596380741e-05, "loss": 1.2838, "step": 3768 }, { "epoch": 0.34, "learning_rate": 3.7462732078806506e-05, "loss": 1.2344, "step": 3769 }, { "epoch": 0.34, "learning_rate": 3.745651716907678e-05, "loss": 1.2489, "step": 3770 }, { "epoch": 0.34, "learning_rate": 3.745030123512923e-05, "loss": 1.2668, "step": 3771 }, { "epoch": 0.34, "learning_rate": 3.744408427747497e-05, "loss": 1.3117, "step": 3772 }, { "epoch": 0.34, "learning_rate": 3.7437866296625154e-05, "loss": 1.2269, "step": 3773 }, { "epoch": 0.34, "learning_rate": 3.743164729309108e-05, "loss": 1.2767, "step": 3774 }, { "epoch": 0.34, "learning_rate": 3.742542726738405e-05, "loss": 1.3033, "step": 3775 }, { "epoch": 0.34, "learning_rate": 3.7419206220015547e-05, "loss": 1.2653, "step": 3776 }, { "epoch": 0.34, "learning_rate": 3.741298415149705e-05, "loss": 1.2837, "step": 3777 }, { "epoch": 0.34, "learning_rate": 3.740676106234017e-05, "loss": 1.2719, "step": 3778 }, { "epoch": 0.34, "learning_rate": 3.7400536953056604e-05, "loss": 1.2805, "step": 3779 }, { "epoch": 0.34, "learning_rate": 3.73943118241581e-05, "loss": 1.2955, "step": 3780 }, { "epoch": 0.34, "learning_rate": 3.738808567615652e-05, "loss": 1.3001, "step": 3781 }, { "epoch": 0.34, "learning_rate": 3.738185850956379e-05, "loss": 1.2786, "step": 3782 }, { "epoch": 0.34, "learning_rate": 3.737563032489194e-05, "loss": 1.2783, "step": 3783 }, { "epoch": 0.34, "learning_rate": 3.736940112265307e-05, "loss": 1.2694, "step": 3784 }, { "epoch": 0.34, "learning_rate": 3.736317090335937e-05, "loss": 1.2047, "step": 3785 }, { "epoch": 0.34, "learning_rate": 3.7356939667523106e-05, "loss": 1.2403, "step": 3786 }, { "epoch": 0.34, "learning_rate": 3.7350707415656625e-05, "loss": 1.2579, "step": 3787 }, { "epoch": 0.34, "learning_rate": 3.734447414827237e-05, "loss": 1.2872, "step": 3788 }, { "epoch": 0.34, "learning_rate": 3.733823986588287e-05, "loss": 1.2718, "step": 3789 }, { "epoch": 0.34, "learning_rate": 3.733200456900071e-05, "loss": 1.2669, "step": 3790 }, { "epoch": 0.34, "learning_rate": 3.7325768258138597e-05, "loss": 1.2784, "step": 3791 }, { "epoch": 0.34, "learning_rate": 3.7319530933809294e-05, "loss": 1.2863, "step": 3792 }, { "epoch": 0.34, "learning_rate": 3.731329259652565e-05, "loss": 1.2707, "step": 3793 }, { "epoch": 0.34, "learning_rate": 3.730705324680062e-05, "loss": 1.2306, "step": 3794 }, { "epoch": 0.34, "learning_rate": 3.7300812885147194e-05, "loss": 1.2516, "step": 3795 }, { "epoch": 0.34, "learning_rate": 3.7294571512078506e-05, "loss": 1.2725, "step": 3796 }, { "epoch": 0.34, "learning_rate": 3.7288329128107736e-05, "loss": 1.2972, "step": 3797 }, { "epoch": 0.34, "learning_rate": 3.728208573374814e-05, "loss": 1.2505, "step": 3798 }, { "epoch": 0.34, "learning_rate": 3.7275841329513095e-05, "loss": 1.3031, "step": 3799 }, { "epoch": 0.34, "learning_rate": 3.726959591591602e-05, "loss": 1.2633, "step": 3800 }, { "epoch": 0.34, "learning_rate": 3.726334949347043e-05, "loss": 1.2105, "step": 3801 }, { "epoch": 0.34, "learning_rate": 3.7257102062689943e-05, "loss": 1.2991, "step": 3802 }, { "epoch": 0.34, "learning_rate": 3.725085362408824e-05, "loss": 1.3161, "step": 3803 }, { "epoch": 0.34, "learning_rate": 3.7244604178179076e-05, "loss": 1.3015, "step": 3804 }, { "epoch": 0.34, "learning_rate": 3.723835372547633e-05, "loss": 1.2704, "step": 3805 }, { "epoch": 0.34, "learning_rate": 3.72321022664939e-05, "loss": 1.2624, "step": 3806 }, { "epoch": 0.34, "learning_rate": 3.7225849801745835e-05, "loss": 1.2558, "step": 3807 }, { "epoch": 0.34, "learning_rate": 3.7219596331746215e-05, "loss": 1.3038, "step": 3808 }, { "epoch": 0.34, "learning_rate": 3.7213341857009224e-05, "loss": 1.2801, "step": 3809 }, { "epoch": 0.34, "learning_rate": 3.720708637804914e-05, "loss": 1.2626, "step": 3810 }, { "epoch": 0.34, "learning_rate": 3.7200829895380284e-05, "loss": 1.2675, "step": 3811 }, { "epoch": 0.34, "learning_rate": 3.7194572409517114e-05, "loss": 1.24, "step": 3812 }, { "epoch": 0.34, "learning_rate": 3.7188313920974126e-05, "loss": 1.2257, "step": 3813 }, { "epoch": 0.34, "learning_rate": 3.7182054430265915e-05, "loss": 1.2639, "step": 3814 }, { "epoch": 0.34, "learning_rate": 3.7175793937907167e-05, "loss": 1.2547, "step": 3815 }, { "epoch": 0.34, "learning_rate": 3.716953244441264e-05, "loss": 1.249, "step": 3816 }, { "epoch": 0.34, "learning_rate": 3.716326995029715e-05, "loss": 1.2672, "step": 3817 }, { "epoch": 0.34, "learning_rate": 3.7157006456075654e-05, "loss": 1.2168, "step": 3818 }, { "epoch": 0.35, "learning_rate": 3.715074196226315e-05, "loss": 1.2918, "step": 3819 }, { "epoch": 0.35, "learning_rate": 3.714447646937471e-05, "loss": 1.25, "step": 3820 }, { "epoch": 0.35, "learning_rate": 3.713820997792553e-05, "loss": 1.2648, "step": 3821 }, { "epoch": 0.35, "learning_rate": 3.713194248843084e-05, "loss": 1.25, "step": 3822 }, { "epoch": 0.35, "learning_rate": 3.7125674001405985e-05, "loss": 1.3443, "step": 3823 }, { "epoch": 0.35, "learning_rate": 3.711940451736638e-05, "loss": 1.2433, "step": 3824 }, { "epoch": 0.35, "learning_rate": 3.7113134036827524e-05, "loss": 1.2546, "step": 3825 }, { "epoch": 0.35, "learning_rate": 3.7106862560305e-05, "loss": 1.261, "step": 3826 }, { "epoch": 0.35, "learning_rate": 3.710059008831447e-05, "loss": 1.2439, "step": 3827 }, { "epoch": 0.35, "learning_rate": 3.709431662137167e-05, "loss": 1.3012, "step": 3828 }, { "epoch": 0.35, "learning_rate": 3.708804215999244e-05, "loss": 1.2628, "step": 3829 }, { "epoch": 0.35, "learning_rate": 3.708176670469267e-05, "loss": 1.2474, "step": 3830 }, { "epoch": 0.35, "learning_rate": 3.7075490255988366e-05, "loss": 1.2883, "step": 3831 }, { "epoch": 0.35, "learning_rate": 3.706921281439559e-05, "loss": 1.2152, "step": 3832 }, { "epoch": 0.35, "learning_rate": 3.70629343804305e-05, "loss": 1.2968, "step": 3833 }, { "epoch": 0.35, "learning_rate": 3.7056654954609324e-05, "loss": 1.1965, "step": 3834 }, { "epoch": 0.35, "learning_rate": 3.7050374537448384e-05, "loss": 1.232, "step": 3835 }, { "epoch": 0.35, "learning_rate": 3.7044093129464074e-05, "loss": 1.2415, "step": 3836 }, { "epoch": 0.35, "learning_rate": 3.703781073117288e-05, "loss": 1.2401, "step": 3837 }, { "epoch": 0.35, "learning_rate": 3.703152734309135e-05, "loss": 1.2796, "step": 3838 }, { "epoch": 0.35, "learning_rate": 3.7025242965736145e-05, "loss": 1.2868, "step": 3839 }, { "epoch": 0.35, "learning_rate": 3.701895759962397e-05, "loss": 1.2275, "step": 3840 }, { "epoch": 0.35, "learning_rate": 3.701267124527164e-05, "loss": 1.2193, "step": 3841 }, { "epoch": 0.35, "learning_rate": 3.700638390319603e-05, "loss": 1.2576, "step": 3842 }, { "epoch": 0.35, "learning_rate": 3.700009557391412e-05, "loss": 1.2648, "step": 3843 }, { "epoch": 0.35, "learning_rate": 3.699380625794294e-05, "loss": 1.2631, "step": 3844 }, { "epoch": 0.35, "learning_rate": 3.698751595579965e-05, "loss": 1.2496, "step": 3845 }, { "epoch": 0.35, "learning_rate": 3.6981224668001424e-05, "loss": 1.2546, "step": 3846 }, { "epoch": 0.35, "learning_rate": 3.697493239506558e-05, "loss": 1.2581, "step": 3847 }, { "epoch": 0.35, "learning_rate": 3.6968639137509474e-05, "loss": 1.2861, "step": 3848 }, { "epoch": 0.35, "learning_rate": 3.696234489585057e-05, "loss": 1.2625, "step": 3849 }, { "epoch": 0.35, "learning_rate": 3.6956049670606403e-05, "loss": 1.3074, "step": 3850 }, { "epoch": 0.35, "learning_rate": 3.694975346229458e-05, "loss": 1.2272, "step": 3851 }, { "epoch": 0.35, "learning_rate": 3.6943456271432795e-05, "loss": 1.2276, "step": 3852 }, { "epoch": 0.35, "learning_rate": 3.693715809853884e-05, "loss": 1.2614, "step": 3853 }, { "epoch": 0.35, "learning_rate": 3.6930858944130564e-05, "loss": 1.2439, "step": 3854 }, { "epoch": 0.35, "learning_rate": 3.6924558808725896e-05, "loss": 1.2807, "step": 3855 }, { "epoch": 0.35, "learning_rate": 3.691825769284287e-05, "loss": 1.3274, "step": 3856 }, { "epoch": 0.35, "learning_rate": 3.691195559699957e-05, "loss": 1.2618, "step": 3857 }, { "epoch": 0.35, "learning_rate": 3.690565252171419e-05, "loss": 1.224, "step": 3858 }, { "epoch": 0.35, "learning_rate": 3.689934846750499e-05, "loss": 1.2614, "step": 3859 }, { "epoch": 0.35, "learning_rate": 3.68930434348903e-05, "loss": 1.2958, "step": 3860 }, { "epoch": 0.35, "learning_rate": 3.688673742438855e-05, "loss": 1.2548, "step": 3861 }, { "epoch": 0.35, "learning_rate": 3.688043043651824e-05, "loss": 1.2728, "step": 3862 }, { "epoch": 0.35, "learning_rate": 3.6874122471797955e-05, "loss": 1.2713, "step": 3863 }, { "epoch": 0.35, "learning_rate": 3.686781353074635e-05, "loss": 1.2228, "step": 3864 }, { "epoch": 0.35, "learning_rate": 3.6861503613882176e-05, "loss": 1.2427, "step": 3865 }, { "epoch": 0.35, "learning_rate": 3.685519272172425e-05, "loss": 1.2428, "step": 3866 }, { "epoch": 0.35, "learning_rate": 3.684888085479148e-05, "loss": 1.2723, "step": 3867 }, { "epoch": 0.35, "learning_rate": 3.684256801360285e-05, "loss": 1.2362, "step": 3868 }, { "epoch": 0.35, "learning_rate": 3.683625419867742e-05, "loss": 1.2217, "step": 3869 }, { "epoch": 0.35, "learning_rate": 3.682993941053433e-05, "loss": 1.2542, "step": 3870 }, { "epoch": 0.35, "learning_rate": 3.682362364969282e-05, "loss": 1.2685, "step": 3871 }, { "epoch": 0.35, "learning_rate": 3.681730691667217e-05, "loss": 1.2726, "step": 3872 }, { "epoch": 0.35, "learning_rate": 3.6810989211991774e-05, "loss": 1.2414, "step": 3873 }, { "epoch": 0.35, "learning_rate": 3.6804670536171105e-05, "loss": 1.2558, "step": 3874 }, { "epoch": 0.35, "learning_rate": 3.67983508897297e-05, "loss": 1.2773, "step": 3875 }, { "epoch": 0.35, "learning_rate": 3.679203027318718e-05, "loss": 1.2629, "step": 3876 }, { "epoch": 0.35, "learning_rate": 3.678570868706324e-05, "loss": 1.2741, "step": 3877 }, { "epoch": 0.35, "learning_rate": 3.677938613187768e-05, "loss": 1.2316, "step": 3878 }, { "epoch": 0.35, "learning_rate": 3.677306260815035e-05, "loss": 1.2729, "step": 3879 }, { "epoch": 0.35, "learning_rate": 3.67667381164012e-05, "loss": 1.2914, "step": 3880 }, { "epoch": 0.35, "learning_rate": 3.676041265715024e-05, "loss": 1.2746, "step": 3881 }, { "epoch": 0.35, "learning_rate": 3.6754086230917576e-05, "loss": 1.33, "step": 3882 }, { "epoch": 0.35, "learning_rate": 3.67477588382234e-05, "loss": 1.2615, "step": 3883 }, { "epoch": 0.35, "learning_rate": 3.6741430479587956e-05, "loss": 1.3361, "step": 3884 }, { "epoch": 0.35, "learning_rate": 3.6735101155531596e-05, "loss": 1.2698, "step": 3885 }, { "epoch": 0.35, "learning_rate": 3.672877086657473e-05, "loss": 1.2737, "step": 3886 }, { "epoch": 0.35, "learning_rate": 3.672243961323786e-05, "loss": 1.3293, "step": 3887 }, { "epoch": 0.35, "learning_rate": 3.671610739604157e-05, "loss": 1.278, "step": 3888 }, { "epoch": 0.35, "learning_rate": 3.6709774215506494e-05, "loss": 1.2976, "step": 3889 }, { "epoch": 0.35, "learning_rate": 3.6703440072153385e-05, "loss": 1.2636, "step": 3890 }, { "epoch": 0.35, "learning_rate": 3.6697104966503064e-05, "loss": 1.2588, "step": 3891 }, { "epoch": 0.35, "learning_rate": 3.669076889907642e-05, "loss": 1.2701, "step": 3892 }, { "epoch": 0.35, "learning_rate": 3.668443187039442e-05, "loss": 1.252, "step": 3893 }, { "epoch": 0.35, "learning_rate": 3.6678093880978116e-05, "loss": 1.2815, "step": 3894 }, { "epoch": 0.35, "learning_rate": 3.6671754931348644e-05, "loss": 1.282, "step": 3895 }, { "epoch": 0.35, "learning_rate": 3.6665415022027225e-05, "loss": 1.2688, "step": 3896 }, { "epoch": 0.35, "learning_rate": 3.6659074153535135e-05, "loss": 1.3119, "step": 3897 }, { "epoch": 0.35, "learning_rate": 3.6652732326393734e-05, "loss": 1.2574, "step": 3898 }, { "epoch": 0.35, "learning_rate": 3.664638954112449e-05, "loss": 1.2962, "step": 3899 }, { "epoch": 0.35, "learning_rate": 3.664004579824891e-05, "loss": 1.3139, "step": 3900 }, { "epoch": 0.35, "learning_rate": 3.6633701098288616e-05, "loss": 1.249, "step": 3901 }, { "epoch": 0.35, "learning_rate": 3.662735544176527e-05, "loss": 1.3303, "step": 3902 }, { "epoch": 0.35, "learning_rate": 3.662100882920065e-05, "loss": 1.2297, "step": 3903 }, { "epoch": 0.35, "learning_rate": 3.661466126111659e-05, "loss": 1.2125, "step": 3904 }, { "epoch": 0.35, "learning_rate": 3.660831273803502e-05, "loss": 1.2893, "step": 3905 }, { "epoch": 0.35, "learning_rate": 3.6601963260477924e-05, "loss": 1.294, "step": 3906 }, { "epoch": 0.35, "learning_rate": 3.659561282896737e-05, "loss": 1.256, "step": 3907 }, { "epoch": 0.35, "learning_rate": 3.658926144402554e-05, "loss": 1.2885, "step": 3908 }, { "epoch": 0.35, "learning_rate": 3.658290910617465e-05, "loss": 1.2343, "step": 3909 }, { "epoch": 0.35, "learning_rate": 3.6576555815937e-05, "loss": 1.2277, "step": 3910 }, { "epoch": 0.35, "learning_rate": 3.6570201573835007e-05, "loss": 1.2375, "step": 3911 }, { "epoch": 0.35, "learning_rate": 3.656384638039111e-05, "loss": 1.2823, "step": 3912 }, { "epoch": 0.35, "learning_rate": 3.655749023612788e-05, "loss": 1.2621, "step": 3913 }, { "epoch": 0.35, "learning_rate": 3.655113314156792e-05, "loss": 1.3077, "step": 3914 }, { "epoch": 0.35, "learning_rate": 3.654477509723395e-05, "loss": 1.3569, "step": 3915 }, { "epoch": 0.35, "learning_rate": 3.653841610364873e-05, "loss": 1.2482, "step": 3916 }, { "epoch": 0.35, "learning_rate": 3.6532056161335147e-05, "loss": 1.2665, "step": 3917 }, { "epoch": 0.35, "learning_rate": 3.652569527081611e-05, "loss": 1.2522, "step": 3918 }, { "epoch": 0.35, "learning_rate": 3.651933343261465e-05, "loss": 1.1898, "step": 3919 }, { "epoch": 0.35, "learning_rate": 3.651297064725385e-05, "loss": 1.2775, "step": 3920 }, { "epoch": 0.35, "learning_rate": 3.650660691525689e-05, "loss": 1.2957, "step": 3921 }, { "epoch": 0.35, "learning_rate": 3.6500242237147e-05, "loss": 1.2606, "step": 3922 }, { "epoch": 0.35, "learning_rate": 3.6493876613447526e-05, "loss": 1.2411, "step": 3923 }, { "epoch": 0.35, "learning_rate": 3.648751004468186e-05, "loss": 1.3088, "step": 3924 }, { "epoch": 0.35, "learning_rate": 3.648114253137349e-05, "loss": 1.2729, "step": 3925 }, { "epoch": 0.35, "learning_rate": 3.647477407404597e-05, "loss": 1.1947, "step": 3926 }, { "epoch": 0.35, "learning_rate": 3.6468404673222934e-05, "loss": 1.271, "step": 3927 }, { "epoch": 0.35, "learning_rate": 3.64620343294281e-05, "loss": 1.2229, "step": 3928 }, { "epoch": 0.35, "learning_rate": 3.645566304318526e-05, "loss": 1.2449, "step": 3929 }, { "epoch": 0.36, "learning_rate": 3.644929081501829e-05, "loss": 1.1899, "step": 3930 }, { "epoch": 0.36, "learning_rate": 3.6442917645451116e-05, "loss": 1.253, "step": 3931 }, { "epoch": 0.36, "learning_rate": 3.643654353500779e-05, "loss": 1.2498, "step": 3932 }, { "epoch": 0.36, "learning_rate": 3.643016848421239e-05, "loss": 1.2579, "step": 3933 }, { "epoch": 0.36, "learning_rate": 3.6423792493589104e-05, "loss": 1.2758, "step": 3934 }, { "epoch": 0.36, "learning_rate": 3.64174155636622e-05, "loss": 1.2693, "step": 3935 }, { "epoch": 0.36, "learning_rate": 3.6411037694955976e-05, "loss": 1.2971, "step": 3936 }, { "epoch": 0.36, "learning_rate": 3.640465888799488e-05, "loss": 1.2447, "step": 3937 }, { "epoch": 0.36, "learning_rate": 3.639827914330338e-05, "loss": 1.2822, "step": 3938 }, { "epoch": 0.36, "learning_rate": 3.639189846140604e-05, "loss": 1.2279, "step": 3939 }, { "epoch": 0.36, "learning_rate": 3.638551684282751e-05, "loss": 1.2012, "step": 3940 }, { "epoch": 0.36, "learning_rate": 3.637913428809251e-05, "loss": 1.2691, "step": 3941 }, { "epoch": 0.36, "learning_rate": 3.637275079772583e-05, "loss": 1.2174, "step": 3942 }, { "epoch": 0.36, "learning_rate": 3.6366366372252344e-05, "loss": 1.2832, "step": 3943 }, { "epoch": 0.36, "learning_rate": 3.635998101219699e-05, "loss": 1.3142, "step": 3944 }, { "epoch": 0.36, "learning_rate": 3.635359471808481e-05, "loss": 1.3243, "step": 3945 }, { "epoch": 0.36, "learning_rate": 3.634720749044091e-05, "loss": 1.2651, "step": 3946 }, { "epoch": 0.36, "learning_rate": 3.634081932979046e-05, "loss": 1.294, "step": 3947 }, { "epoch": 0.36, "learning_rate": 3.6334430236658726e-05, "loss": 1.2819, "step": 3948 }, { "epoch": 0.36, "learning_rate": 3.632804021157102e-05, "loss": 1.2253, "step": 3949 }, { "epoch": 0.36, "learning_rate": 3.632164925505277e-05, "loss": 1.2406, "step": 3950 }, { "epoch": 0.36, "learning_rate": 3.631525736762946e-05, "loss": 1.2818, "step": 3951 }, { "epoch": 0.36, "learning_rate": 3.630886454982665e-05, "loss": 1.3249, "step": 3952 }, { "epoch": 0.36, "learning_rate": 3.6302470802169994e-05, "loss": 1.2623, "step": 3953 }, { "epoch": 0.36, "learning_rate": 3.629607612518518e-05, "loss": 1.2407, "step": 3954 }, { "epoch": 0.36, "learning_rate": 3.6289680519398026e-05, "loss": 1.1996, "step": 3955 }, { "epoch": 0.36, "learning_rate": 3.6283283985334385e-05, "loss": 1.2588, "step": 3956 }, { "epoch": 0.36, "learning_rate": 3.6276886523520206e-05, "loss": 1.3012, "step": 3957 }, { "epoch": 0.36, "learning_rate": 3.627048813448151e-05, "loss": 1.262, "step": 3958 }, { "epoch": 0.36, "learning_rate": 3.62640888187444e-05, "loss": 1.2174, "step": 3959 }, { "epoch": 0.36, "learning_rate": 3.6257688576835045e-05, "loss": 1.199, "step": 3960 }, { "epoch": 0.36, "learning_rate": 3.625128740927971e-05, "loss": 1.2656, "step": 3961 }, { "epoch": 0.36, "learning_rate": 3.624488531660468e-05, "loss": 1.2215, "step": 3962 }, { "epoch": 0.36, "learning_rate": 3.623848229933641e-05, "loss": 1.2277, "step": 3963 }, { "epoch": 0.36, "learning_rate": 3.623207835800134e-05, "loss": 1.2583, "step": 3964 }, { "epoch": 0.36, "learning_rate": 3.622567349312604e-05, "loss": 1.2752, "step": 3965 }, { "epoch": 0.36, "learning_rate": 3.6219267705237135e-05, "loss": 1.2306, "step": 3966 }, { "epoch": 0.36, "learning_rate": 3.621286099486133e-05, "loss": 1.3046, "step": 3967 }, { "epoch": 0.36, "learning_rate": 3.620645336252542e-05, "loss": 1.3059, "step": 3968 }, { "epoch": 0.36, "learning_rate": 3.620004480875625e-05, "loss": 1.3474, "step": 3969 }, { "epoch": 0.36, "learning_rate": 3.6193635334080746e-05, "loss": 1.3568, "step": 3970 }, { "epoch": 0.36, "learning_rate": 3.618722493902594e-05, "loss": 1.2856, "step": 3971 }, { "epoch": 0.36, "learning_rate": 3.61808136241189e-05, "loss": 1.254, "step": 3972 }, { "epoch": 0.36, "learning_rate": 3.617440138988678e-05, "loss": 1.2546, "step": 3973 }, { "epoch": 0.36, "learning_rate": 3.6167988236856837e-05, "loss": 1.3317, "step": 3974 }, { "epoch": 0.36, "learning_rate": 3.616157416555637e-05, "loss": 1.2482, "step": 3975 }, { "epoch": 0.36, "learning_rate": 3.6155159176512766e-05, "loss": 1.2945, "step": 3976 }, { "epoch": 0.36, "learning_rate": 3.614874327025349e-05, "loss": 1.209, "step": 3977 }, { "epoch": 0.36, "learning_rate": 3.614232644730608e-05, "loss": 1.2467, "step": 3978 }, { "epoch": 0.36, "learning_rate": 3.613590870819814e-05, "loss": 1.234, "step": 3979 }, { "epoch": 0.36, "learning_rate": 3.6129490053457374e-05, "loss": 1.2996, "step": 3980 }, { "epoch": 0.36, "learning_rate": 3.612307048361154e-05, "loss": 1.2583, "step": 3981 }, { "epoch": 0.36, "learning_rate": 3.611664999918848e-05, "loss": 1.2726, "step": 3982 }, { "epoch": 0.36, "learning_rate": 3.61102286007161e-05, "loss": 1.2674, "step": 3983 }, { "epoch": 0.36, "learning_rate": 3.6103806288722396e-05, "loss": 1.2672, "step": 3984 }, { "epoch": 0.36, "learning_rate": 3.609738306373543e-05, "loss": 1.2539, "step": 3985 }, { "epoch": 0.36, "learning_rate": 3.6090958926283344e-05, "loss": 1.2441, "step": 3986 }, { "epoch": 0.36, "learning_rate": 3.608453387689435e-05, "loss": 1.3098, "step": 3987 }, { "epoch": 0.36, "learning_rate": 3.6078107916096736e-05, "loss": 1.2289, "step": 3988 }, { "epoch": 0.36, "learning_rate": 3.607168104441888e-05, "loss": 1.2686, "step": 3989 }, { "epoch": 0.36, "learning_rate": 3.60652532623892e-05, "loss": 1.3147, "step": 3990 }, { "epoch": 0.36, "learning_rate": 3.6058824570536226e-05, "loss": 1.2788, "step": 3991 }, { "epoch": 0.36, "learning_rate": 3.605239496938854e-05, "loss": 1.2965, "step": 3992 }, { "epoch": 0.36, "learning_rate": 3.604596445947482e-05, "loss": 1.2891, "step": 3993 }, { "epoch": 0.36, "learning_rate": 3.603953304132378e-05, "loss": 1.2087, "step": 3994 }, { "epoch": 0.36, "learning_rate": 3.603310071546426e-05, "loss": 1.2366, "step": 3995 }, { "epoch": 0.36, "learning_rate": 3.6026667482425116e-05, "loss": 1.2624, "step": 3996 }, { "epoch": 0.36, "learning_rate": 3.6020233342735356e-05, "loss": 1.2456, "step": 3997 }, { "epoch": 0.36, "learning_rate": 3.6013798296923976e-05, "loss": 1.2615, "step": 3998 }, { "epoch": 0.36, "learning_rate": 3.6007362345520094e-05, "loss": 1.2609, "step": 3999 }, { "epoch": 0.36, "learning_rate": 3.6000925489052915e-05, "loss": 1.2433, "step": 4000 }, { "epoch": 0.36, "learning_rate": 3.599448772805169e-05, "loss": 1.2951, "step": 4001 }, { "epoch": 0.36, "learning_rate": 3.5988049063045756e-05, "loss": 1.2772, "step": 4002 }, { "epoch": 0.36, "learning_rate": 3.598160949456452e-05, "loss": 1.2312, "step": 4003 }, { "epoch": 0.36, "learning_rate": 3.597516902313745e-05, "loss": 1.2246, "step": 4004 }, { "epoch": 0.36, "learning_rate": 3.596872764929413e-05, "loss": 1.2521, "step": 4005 }, { "epoch": 0.36, "learning_rate": 3.596228537356418e-05, "loss": 1.3146, "step": 4006 }, { "epoch": 0.36, "learning_rate": 3.595584219647731e-05, "loss": 1.2843, "step": 4007 }, { "epoch": 0.36, "learning_rate": 3.5949398118563294e-05, "loss": 1.2859, "step": 4008 }, { "epoch": 0.36, "learning_rate": 3.5942953140351987e-05, "loss": 1.25, "step": 4009 }, { "epoch": 0.36, "learning_rate": 3.593650726237332e-05, "loss": 1.2627, "step": 4010 }, { "epoch": 0.36, "learning_rate": 3.59300604851573e-05, "loss": 1.2795, "step": 4011 }, { "epoch": 0.36, "learning_rate": 3.592361280923399e-05, "loss": 1.2591, "step": 4012 }, { "epoch": 0.36, "learning_rate": 3.591716423513355e-05, "loss": 1.2718, "step": 4013 }, { "epoch": 0.36, "learning_rate": 3.591071476338621e-05, "loss": 1.26, "step": 4014 }, { "epoch": 0.36, "learning_rate": 3.590426439452226e-05, "loss": 1.2483, "step": 4015 }, { "epoch": 0.36, "learning_rate": 3.589781312907207e-05, "loss": 1.2901, "step": 4016 }, { "epoch": 0.36, "learning_rate": 3.589136096756608e-05, "loss": 1.2767, "step": 4017 }, { "epoch": 0.36, "learning_rate": 3.588490791053483e-05, "loss": 1.2748, "step": 4018 }, { "epoch": 0.36, "learning_rate": 3.5878453958508884e-05, "loss": 1.2275, "step": 4019 }, { "epoch": 0.36, "learning_rate": 3.587199911201893e-05, "loss": 1.2996, "step": 4020 }, { "epoch": 0.36, "learning_rate": 3.58655433715957e-05, "loss": 1.3066, "step": 4021 }, { "epoch": 0.36, "learning_rate": 3.585908673777001e-05, "loss": 1.2629, "step": 4022 }, { "epoch": 0.36, "learning_rate": 3.585262921107274e-05, "loss": 1.2945, "step": 4023 }, { "epoch": 0.36, "learning_rate": 3.5846170792034854e-05, "loss": 1.2583, "step": 4024 }, { "epoch": 0.36, "learning_rate": 3.583971148118739e-05, "loss": 1.2693, "step": 4025 }, { "epoch": 0.36, "learning_rate": 3.5833251279061456e-05, "loss": 1.2844, "step": 4026 }, { "epoch": 0.36, "learning_rate": 3.5826790186188216e-05, "loss": 1.2831, "step": 4027 }, { "epoch": 0.36, "learning_rate": 3.582032820309894e-05, "loss": 1.2878, "step": 4028 }, { "epoch": 0.36, "learning_rate": 3.581386533032496e-05, "loss": 1.2546, "step": 4029 }, { "epoch": 0.36, "learning_rate": 3.580740156839765e-05, "loss": 1.2359, "step": 4030 }, { "epoch": 0.36, "learning_rate": 3.58009369178485e-05, "loss": 1.2174, "step": 4031 }, { "epoch": 0.36, "learning_rate": 3.579447137920906e-05, "loss": 1.291, "step": 4032 }, { "epoch": 0.36, "learning_rate": 3.578800495301093e-05, "loss": 1.2685, "step": 4033 }, { "epoch": 0.36, "learning_rate": 3.578153763978583e-05, "loss": 1.2971, "step": 4034 }, { "epoch": 0.36, "learning_rate": 3.57750694400655e-05, "loss": 1.2878, "step": 4035 }, { "epoch": 0.36, "learning_rate": 3.5768600354381795e-05, "loss": 1.2533, "step": 4036 }, { "epoch": 0.36, "learning_rate": 3.576213038326661e-05, "loss": 1.2711, "step": 4037 }, { "epoch": 0.36, "learning_rate": 3.575565952725193e-05, "loss": 1.2225, "step": 4038 }, { "epoch": 0.36, "learning_rate": 3.574918778686983e-05, "loss": 1.2704, "step": 4039 }, { "epoch": 0.37, "learning_rate": 3.574271516265242e-05, "loss": 1.2861, "step": 4040 }, { "epoch": 0.37, "learning_rate": 3.5736241655131904e-05, "loss": 1.2728, "step": 4041 }, { "epoch": 0.37, "learning_rate": 3.5729767264840566e-05, "loss": 1.258, "step": 4042 }, { "epoch": 0.37, "learning_rate": 3.572329199231074e-05, "loss": 1.2827, "step": 4043 }, { "epoch": 0.37, "learning_rate": 3.571681583807485e-05, "loss": 1.201, "step": 4044 }, { "epoch": 0.37, "learning_rate": 3.571033880266539e-05, "loss": 1.2578, "step": 4045 }, { "epoch": 0.37, "learning_rate": 3.570386088661492e-05, "loss": 1.2439, "step": 4046 }, { "epoch": 0.37, "learning_rate": 3.569738209045609e-05, "loss": 1.2215, "step": 4047 }, { "epoch": 0.37, "learning_rate": 3.5690902414721595e-05, "loss": 1.2885, "step": 4048 }, { "epoch": 0.37, "learning_rate": 3.5684421859944214e-05, "loss": 1.3018, "step": 4049 }, { "epoch": 0.37, "learning_rate": 3.5677940426656806e-05, "loss": 1.2704, "step": 4050 }, { "epoch": 0.37, "learning_rate": 3.567145811539231e-05, "loss": 1.2385, "step": 4051 }, { "epoch": 0.37, "learning_rate": 3.5664974926683704e-05, "loss": 1.2302, "step": 4052 }, { "epoch": 0.37, "learning_rate": 3.5658490861064065e-05, "loss": 1.2621, "step": 4053 }, { "epoch": 0.37, "learning_rate": 3.565200591906653e-05, "loss": 1.2306, "step": 4054 }, { "epoch": 0.37, "learning_rate": 3.564552010122433e-05, "loss": 1.2492, "step": 4055 }, { "epoch": 0.37, "learning_rate": 3.563903340807073e-05, "loss": 1.2435, "step": 4056 }, { "epoch": 0.37, "learning_rate": 3.5632545840139107e-05, "loss": 1.2808, "step": 4057 }, { "epoch": 0.37, "learning_rate": 3.5626057397962875e-05, "loss": 1.2699, "step": 4058 }, { "epoch": 0.37, "learning_rate": 3.561956808207555e-05, "loss": 1.192, "step": 4059 }, { "epoch": 0.37, "learning_rate": 3.5613077893010705e-05, "loss": 1.2523, "step": 4060 }, { "epoch": 0.37, "learning_rate": 3.560658683130197e-05, "loss": 1.232, "step": 4061 }, { "epoch": 0.37, "learning_rate": 3.560009489748308e-05, "loss": 1.3041, "step": 4062 }, { "epoch": 0.37, "learning_rate": 3.5593602092087806e-05, "loss": 1.246, "step": 4063 }, { "epoch": 0.37, "learning_rate": 3.558710841565004e-05, "loss": 1.2648, "step": 4064 }, { "epoch": 0.37, "learning_rate": 3.558061386870368e-05, "loss": 1.252, "step": 4065 }, { "epoch": 0.37, "learning_rate": 3.557411845178275e-05, "loss": 1.2986, "step": 4066 }, { "epoch": 0.37, "learning_rate": 3.556762216542131e-05, "loss": 1.2484, "step": 4067 }, { "epoch": 0.37, "learning_rate": 3.556112501015353e-05, "loss": 1.2452, "step": 4068 }, { "epoch": 0.37, "learning_rate": 3.555462698651361e-05, "loss": 1.2513, "step": 4069 }, { "epoch": 0.37, "learning_rate": 3.5548128095035846e-05, "loss": 1.217, "step": 4070 }, { "epoch": 0.37, "learning_rate": 3.55416283362546e-05, "loss": 1.3245, "step": 4071 }, { "epoch": 0.37, "learning_rate": 3.553512771070431e-05, "loss": 1.242, "step": 4072 }, { "epoch": 0.37, "learning_rate": 3.552862621891947e-05, "loss": 1.2827, "step": 4073 }, { "epoch": 0.37, "learning_rate": 3.5522123861434655e-05, "loss": 1.2465, "step": 4074 }, { "epoch": 0.37, "learning_rate": 3.5515620638784514e-05, "loss": 1.2848, "step": 4075 }, { "epoch": 0.37, "learning_rate": 3.550911655150376e-05, "loss": 1.2817, "step": 4076 }, { "epoch": 0.37, "learning_rate": 3.55026116001272e-05, "loss": 1.2335, "step": 4077 }, { "epoch": 0.37, "learning_rate": 3.5496105785189684e-05, "loss": 1.2669, "step": 4078 }, { "epoch": 0.37, "learning_rate": 3.548959910722612e-05, "loss": 1.2609, "step": 4079 }, { "epoch": 0.37, "learning_rate": 3.548309156677154e-05, "loss": 1.2595, "step": 4080 }, { "epoch": 0.37, "learning_rate": 3.547658316436101e-05, "loss": 1.2782, "step": 4081 }, { "epoch": 0.37, "learning_rate": 3.5470073900529655e-05, "loss": 1.2382, "step": 4082 }, { "epoch": 0.37, "learning_rate": 3.546356377581271e-05, "loss": 1.3043, "step": 4083 }, { "epoch": 0.37, "learning_rate": 3.545705279074545e-05, "loss": 1.2526, "step": 4084 }, { "epoch": 0.37, "learning_rate": 3.545054094586324e-05, "loss": 1.2704, "step": 4085 }, { "epoch": 0.37, "learning_rate": 3.5444028241701486e-05, "loss": 1.2258, "step": 4086 }, { "epoch": 0.37, "learning_rate": 3.54375146787957e-05, "loss": 1.2781, "step": 4087 }, { "epoch": 0.37, "learning_rate": 3.5431000257681456e-05, "loss": 1.321, "step": 4088 }, { "epoch": 0.37, "learning_rate": 3.542448497889438e-05, "loss": 1.2345, "step": 4089 }, { "epoch": 0.37, "learning_rate": 3.541796884297019e-05, "loss": 1.23, "step": 4090 }, { "epoch": 0.37, "learning_rate": 3.541145185044466e-05, "loss": 1.2052, "step": 4091 }, { "epoch": 0.37, "learning_rate": 3.540493400185363e-05, "loss": 1.2748, "step": 4092 }, { "epoch": 0.37, "learning_rate": 3.5398415297733034e-05, "loss": 1.2317, "step": 4093 }, { "epoch": 0.37, "learning_rate": 3.5391895738618865e-05, "loss": 1.3018, "step": 4094 }, { "epoch": 0.37, "learning_rate": 3.5385375325047166e-05, "loss": 1.2473, "step": 4095 }, { "epoch": 0.37, "learning_rate": 3.537885405755408e-05, "loss": 1.2867, "step": 4096 }, { "epoch": 0.37, "learning_rate": 3.5372331936675804e-05, "loss": 1.2321, "step": 4097 }, { "epoch": 0.37, "learning_rate": 3.536580896294862e-05, "loss": 1.3013, "step": 4098 }, { "epoch": 0.37, "learning_rate": 3.535928513690886e-05, "loss": 1.2729, "step": 4099 }, { "epoch": 0.37, "learning_rate": 3.535276045909293e-05, "loss": 1.2551, "step": 4100 }, { "epoch": 0.37, "learning_rate": 3.534623493003732e-05, "loss": 1.237, "step": 4101 }, { "epoch": 0.37, "learning_rate": 3.533970855027859e-05, "loss": 1.2978, "step": 4102 }, { "epoch": 0.37, "learning_rate": 3.533318132035334e-05, "loss": 1.234, "step": 4103 }, { "epoch": 0.37, "learning_rate": 3.532665324079828e-05, "loss": 1.2898, "step": 4104 }, { "epoch": 0.37, "learning_rate": 3.5320124312150154e-05, "loss": 1.2276, "step": 4105 }, { "epoch": 0.37, "learning_rate": 3.531359453494581e-05, "loss": 1.2947, "step": 4106 }, { "epoch": 0.37, "learning_rate": 3.530706390972214e-05, "loss": 1.2756, "step": 4107 }, { "epoch": 0.37, "learning_rate": 3.530053243701611e-05, "loss": 1.2707, "step": 4108 }, { "epoch": 0.37, "learning_rate": 3.529400011736478e-05, "loss": 1.2684, "step": 4109 }, { "epoch": 0.37, "learning_rate": 3.528746695130523e-05, "loss": 1.2645, "step": 4110 }, { "epoch": 0.37, "learning_rate": 3.528093293937467e-05, "loss": 1.2833, "step": 4111 }, { "epoch": 0.37, "learning_rate": 3.527439808211033e-05, "loss": 1.2602, "step": 4112 }, { "epoch": 0.37, "learning_rate": 3.5267862380049525e-05, "loss": 1.2901, "step": 4113 }, { "epoch": 0.37, "learning_rate": 3.526132583372967e-05, "loss": 1.282, "step": 4114 }, { "epoch": 0.37, "learning_rate": 3.5254788443688184e-05, "loss": 1.2527, "step": 4115 }, { "epoch": 0.37, "learning_rate": 3.524825021046262e-05, "loss": 1.2495, "step": 4116 }, { "epoch": 0.37, "learning_rate": 3.524171113459057e-05, "loss": 1.2472, "step": 4117 }, { "epoch": 0.37, "learning_rate": 3.523517121660969e-05, "loss": 1.2564, "step": 4118 }, { "epoch": 0.37, "learning_rate": 3.5228630457057724e-05, "loss": 1.2659, "step": 4119 }, { "epoch": 0.37, "learning_rate": 3.522208885647248e-05, "loss": 1.2973, "step": 4120 }, { "epoch": 0.37, "learning_rate": 3.5215546415391816e-05, "loss": 1.2191, "step": 4121 }, { "epoch": 0.37, "learning_rate": 3.520900313435369e-05, "loss": 1.2741, "step": 4122 }, { "epoch": 0.37, "learning_rate": 3.52024590138961e-05, "loss": 1.2763, "step": 4123 }, { "epoch": 0.37, "learning_rate": 3.519591405455713e-05, "loss": 1.3389, "step": 4124 }, { "epoch": 0.37, "learning_rate": 3.518936825687493e-05, "loss": 1.2198, "step": 4125 }, { "epoch": 0.37, "learning_rate": 3.518282162138772e-05, "loss": 1.2499, "step": 4126 }, { "epoch": 0.37, "learning_rate": 3.517627414863379e-05, "loss": 1.2544, "step": 4127 }, { "epoch": 0.37, "learning_rate": 3.5169725839151494e-05, "loss": 1.258, "step": 4128 }, { "epoch": 0.37, "learning_rate": 3.516317669347925e-05, "loss": 1.2354, "step": 4129 }, { "epoch": 0.37, "learning_rate": 3.515662671215555e-05, "loss": 1.2164, "step": 4130 }, { "epoch": 0.37, "learning_rate": 3.5150075895718964e-05, "loss": 1.3167, "step": 4131 }, { "epoch": 0.37, "learning_rate": 3.514352424470813e-05, "loss": 1.2462, "step": 4132 }, { "epoch": 0.37, "learning_rate": 3.513697175966173e-05, "loss": 1.2386, "step": 4133 }, { "epoch": 0.37, "learning_rate": 3.513041844111854e-05, "loss": 1.2498, "step": 4134 }, { "epoch": 0.37, "learning_rate": 3.512386428961739e-05, "loss": 1.2732, "step": 4135 }, { "epoch": 0.37, "learning_rate": 3.51173093056972e-05, "loss": 1.1738, "step": 4136 }, { "epoch": 0.37, "learning_rate": 3.511075348989692e-05, "loss": 1.2886, "step": 4137 }, { "epoch": 0.37, "learning_rate": 3.510419684275562e-05, "loss": 1.2569, "step": 4138 }, { "epoch": 0.37, "learning_rate": 3.5097639364812395e-05, "loss": 1.2442, "step": 4139 }, { "epoch": 0.37, "learning_rate": 3.509108105660642e-05, "loss": 1.2966, "step": 4140 }, { "epoch": 0.37, "learning_rate": 3.5084521918676956e-05, "loss": 1.2863, "step": 4141 }, { "epoch": 0.37, "learning_rate": 3.5077961951563296e-05, "loss": 1.2667, "step": 4142 }, { "epoch": 0.37, "learning_rate": 3.507140115580485e-05, "loss": 1.2414, "step": 4143 }, { "epoch": 0.37, "learning_rate": 3.506483953194104e-05, "loss": 1.2642, "step": 4144 }, { "epoch": 0.37, "learning_rate": 3.50582770805114e-05, "loss": 1.2634, "step": 4145 }, { "epoch": 0.37, "learning_rate": 3.505171380205553e-05, "loss": 1.2703, "step": 4146 }, { "epoch": 0.37, "learning_rate": 3.5045149697113064e-05, "loss": 1.3077, "step": 4147 }, { "epoch": 0.37, "learning_rate": 3.503858476622374e-05, "loss": 1.2796, "step": 4148 }, { "epoch": 0.37, "learning_rate": 3.503201900992734e-05, "loss": 1.299, "step": 4149 }, { "epoch": 0.37, "learning_rate": 3.502545242876372e-05, "loss": 1.2408, "step": 4150 }, { "epoch": 0.38, "learning_rate": 3.501888502327282e-05, "loss": 1.2301, "step": 4151 }, { "epoch": 0.38, "learning_rate": 3.5012316793994625e-05, "loss": 1.2552, "step": 4152 }, { "epoch": 0.38, "learning_rate": 3.500574774146921e-05, "loss": 1.2923, "step": 4153 }, { "epoch": 0.38, "learning_rate": 3.4999177866236694e-05, "loss": 1.2191, "step": 4154 }, { "epoch": 0.38, "learning_rate": 3.499260716883727e-05, "loss": 1.26, "step": 4155 }, { "epoch": 0.38, "learning_rate": 3.498603564981121e-05, "loss": 1.2722, "step": 4156 }, { "epoch": 0.38, "learning_rate": 3.497946330969886e-05, "loss": 1.1987, "step": 4157 }, { "epoch": 0.38, "learning_rate": 3.4972890149040595e-05, "loss": 1.3019, "step": 4158 }, { "epoch": 0.38, "learning_rate": 3.49663161683769e-05, "loss": 1.2679, "step": 4159 }, { "epoch": 0.38, "learning_rate": 3.49597413682483e-05, "loss": 1.2698, "step": 4160 }, { "epoch": 0.38, "learning_rate": 3.495316574919541e-05, "loss": 1.2574, "step": 4161 }, { "epoch": 0.38, "learning_rate": 3.494658931175889e-05, "loss": 1.2256, "step": 4162 }, { "epoch": 0.38, "learning_rate": 3.494001205647948e-05, "loss": 1.2647, "step": 4163 }, { "epoch": 0.38, "learning_rate": 3.4933433983898e-05, "loss": 1.2554, "step": 4164 }, { "epoch": 0.38, "learning_rate": 3.4926855094555294e-05, "loss": 1.2881, "step": 4165 }, { "epoch": 0.38, "learning_rate": 3.4920275388992315e-05, "loss": 1.2492, "step": 4166 }, { "epoch": 0.38, "learning_rate": 3.491369486775007e-05, "loss": 1.2918, "step": 4167 }, { "epoch": 0.38, "learning_rate": 3.490711353136963e-05, "loss": 1.2844, "step": 4168 }, { "epoch": 0.38, "learning_rate": 3.490053138039214e-05, "loss": 1.2358, "step": 4169 }, { "epoch": 0.38, "learning_rate": 3.48939484153588e-05, "loss": 1.3038, "step": 4170 }, { "epoch": 0.38, "learning_rate": 3.488736463681089e-05, "loss": 1.3153, "step": 4171 }, { "epoch": 0.38, "learning_rate": 3.488078004528975e-05, "loss": 1.2885, "step": 4172 }, { "epoch": 0.38, "learning_rate": 3.487419464133678e-05, "loss": 1.2343, "step": 4173 }, { "epoch": 0.38, "learning_rate": 3.486760842549347e-05, "loss": 1.2917, "step": 4174 }, { "epoch": 0.38, "learning_rate": 3.486102139830134e-05, "loss": 1.2703, "step": 4175 }, { "epoch": 0.38, "learning_rate": 3.4854433560302015e-05, "loss": 1.2294, "step": 4176 }, { "epoch": 0.38, "learning_rate": 3.484784491203717e-05, "loss": 1.2492, "step": 4177 }, { "epoch": 0.38, "learning_rate": 3.484125545404854e-05, "loss": 1.2593, "step": 4178 }, { "epoch": 0.38, "learning_rate": 3.483466518687792e-05, "loss": 1.275, "step": 4179 }, { "epoch": 0.38, "learning_rate": 3.4828074111067224e-05, "loss": 1.2841, "step": 4180 }, { "epoch": 0.38, "learning_rate": 3.482148222715835e-05, "loss": 1.2631, "step": 4181 }, { "epoch": 0.38, "learning_rate": 3.481488953569333e-05, "loss": 1.2386, "step": 4182 }, { "epoch": 0.38, "learning_rate": 3.480829603721423e-05, "loss": 1.2398, "step": 4183 }, { "epoch": 0.38, "learning_rate": 3.480170173226318e-05, "loss": 1.1799, "step": 4184 }, { "epoch": 0.38, "learning_rate": 3.479510662138241e-05, "loss": 1.2597, "step": 4185 }, { "epoch": 0.38, "learning_rate": 3.4788510705114184e-05, "loss": 1.2747, "step": 4186 }, { "epoch": 0.38, "learning_rate": 3.478191398400083e-05, "loss": 1.2551, "step": 4187 }, { "epoch": 0.38, "learning_rate": 3.4775316458584765e-05, "loss": 1.2831, "step": 4188 }, { "epoch": 0.38, "learning_rate": 3.476871812940845e-05, "loss": 1.3215, "step": 4189 }, { "epoch": 0.38, "learning_rate": 3.4762118997014435e-05, "loss": 1.2288, "step": 4190 }, { "epoch": 0.38, "learning_rate": 3.475551906194532e-05, "loss": 1.2329, "step": 4191 }, { "epoch": 0.38, "learning_rate": 3.474891832474376e-05, "loss": 1.2551, "step": 4192 }, { "epoch": 0.38, "learning_rate": 3.474231678595251e-05, "loss": 1.2642, "step": 4193 }, { "epoch": 0.38, "learning_rate": 3.4735714446114356e-05, "loss": 1.2133, "step": 4194 }, { "epoch": 0.38, "learning_rate": 3.472911130577218e-05, "loss": 1.2768, "step": 4195 }, { "epoch": 0.38, "learning_rate": 3.4722507365468903e-05, "loss": 1.3115, "step": 4196 }, { "epoch": 0.38, "learning_rate": 3.4715902625747524e-05, "loss": 1.245, "step": 4197 }, { "epoch": 0.38, "learning_rate": 3.470929708715112e-05, "loss": 1.293, "step": 4198 }, { "epoch": 0.38, "learning_rate": 3.470269075022281e-05, "loss": 1.2729, "step": 4199 }, { "epoch": 0.38, "learning_rate": 3.469608361550579e-05, "loss": 1.2908, "step": 4200 }, { "epoch": 0.38, "learning_rate": 3.468947568354333e-05, "loss": 1.2249, "step": 4201 }, { "epoch": 0.38, "learning_rate": 3.468286695487874e-05, "loss": 1.2397, "step": 4202 }, { "epoch": 0.38, "learning_rate": 3.4676257430055434e-05, "loss": 1.2187, "step": 4203 }, { "epoch": 0.38, "learning_rate": 3.466964710961685e-05, "loss": 1.2626, "step": 4204 }, { "epoch": 0.38, "learning_rate": 3.466303599410652e-05, "loss": 1.2651, "step": 4205 }, { "epoch": 0.38, "learning_rate": 3.465642408406804e-05, "loss": 1.2492, "step": 4206 }, { "epoch": 0.38, "learning_rate": 3.464981138004506e-05, "loss": 1.314, "step": 4207 }, { "epoch": 0.38, "learning_rate": 3.464319788258129e-05, "loss": 1.2468, "step": 4208 }, { "epoch": 0.38, "learning_rate": 3.463658359222052e-05, "loss": 1.282, "step": 4209 }, { "epoch": 0.38, "learning_rate": 3.46299685095066e-05, "loss": 1.2252, "step": 4210 }, { "epoch": 0.38, "learning_rate": 3.462335263498345e-05, "loss": 1.2247, "step": 4211 }, { "epoch": 0.38, "learning_rate": 3.4616735969195054e-05, "loss": 1.2745, "step": 4212 }, { "epoch": 0.38, "learning_rate": 3.461011851268544e-05, "loss": 1.2504, "step": 4213 }, { "epoch": 0.38, "learning_rate": 3.4603500265998725e-05, "loss": 1.2629, "step": 4214 }, { "epoch": 0.38, "learning_rate": 3.4596881229679084e-05, "loss": 1.2634, "step": 4215 }, { "epoch": 0.38, "learning_rate": 3.459026140427077e-05, "loss": 1.267, "step": 4216 }, { "epoch": 0.38, "learning_rate": 3.458364079031807e-05, "loss": 1.2552, "step": 4217 }, { "epoch": 0.38, "learning_rate": 3.4577019388365355e-05, "loss": 1.2176, "step": 4218 }, { "epoch": 0.38, "learning_rate": 3.457039719895707e-05, "loss": 1.2418, "step": 4219 }, { "epoch": 0.38, "learning_rate": 3.4563774222637716e-05, "loss": 1.2441, "step": 4220 }, { "epoch": 0.38, "learning_rate": 3.4557150459951845e-05, "loss": 1.2481, "step": 4221 }, { "epoch": 0.38, "learning_rate": 3.455052591144409e-05, "loss": 1.2986, "step": 4222 }, { "epoch": 0.38, "learning_rate": 3.454390057765915e-05, "loss": 1.268, "step": 4223 }, { "epoch": 0.38, "learning_rate": 3.453727445914178e-05, "loss": 1.3017, "step": 4224 }, { "epoch": 0.38, "learning_rate": 3.453064755643679e-05, "loss": 1.2356, "step": 4225 }, { "epoch": 0.38, "learning_rate": 3.4524019870089085e-05, "loss": 1.2631, "step": 4226 }, { "epoch": 0.38, "learning_rate": 3.451739140064361e-05, "loss": 1.2686, "step": 4227 }, { "epoch": 0.38, "learning_rate": 3.4510762148645376e-05, "loss": 1.2563, "step": 4228 }, { "epoch": 0.38, "learning_rate": 3.450413211463947e-05, "loss": 1.2179, "step": 4229 }, { "epoch": 0.38, "learning_rate": 3.449750129917102e-05, "loss": 1.288, "step": 4230 }, { "epoch": 0.38, "learning_rate": 3.4490869702785256e-05, "loss": 1.2449, "step": 4231 }, { "epoch": 0.38, "learning_rate": 3.4484237326027446e-05, "loss": 1.2724, "step": 4232 }, { "epoch": 0.38, "learning_rate": 3.447760416944292e-05, "loss": 1.3228, "step": 4233 }, { "epoch": 0.38, "learning_rate": 3.4470970233577074e-05, "loss": 1.3109, "step": 4234 }, { "epoch": 0.38, "learning_rate": 3.446433551897539e-05, "loss": 1.2744, "step": 4235 }, { "epoch": 0.38, "learning_rate": 3.4457700026183374e-05, "loss": 1.2689, "step": 4236 }, { "epoch": 0.38, "learning_rate": 3.445106375574665e-05, "loss": 1.2376, "step": 4237 }, { "epoch": 0.38, "learning_rate": 3.444442670821085e-05, "loss": 1.2412, "step": 4238 }, { "epoch": 0.38, "learning_rate": 3.44377888841217e-05, "loss": 1.2476, "step": 4239 }, { "epoch": 0.38, "learning_rate": 3.4431150284024994e-05, "loss": 1.2519, "step": 4240 }, { "epoch": 0.38, "learning_rate": 3.4424510908466576e-05, "loss": 1.258, "step": 4241 }, { "epoch": 0.38, "learning_rate": 3.4417870757992355e-05, "loss": 1.2543, "step": 4242 }, { "epoch": 0.38, "learning_rate": 3.44112298331483e-05, "loss": 1.2445, "step": 4243 }, { "epoch": 0.38, "learning_rate": 3.4404588134480466e-05, "loss": 1.2878, "step": 4244 }, { "epoch": 0.38, "learning_rate": 3.4397945662534956e-05, "loss": 1.2749, "step": 4245 }, { "epoch": 0.38, "learning_rate": 3.439130241785792e-05, "loss": 1.2434, "step": 4246 }, { "epoch": 0.38, "learning_rate": 3.4384658400995604e-05, "loss": 1.2148, "step": 4247 }, { "epoch": 0.38, "learning_rate": 3.437801361249429e-05, "loss": 1.2885, "step": 4248 }, { "epoch": 0.38, "learning_rate": 3.437136805290035e-05, "loss": 1.2737, "step": 4249 }, { "epoch": 0.38, "learning_rate": 3.436472172276021e-05, "loss": 1.3038, "step": 4250 }, { "epoch": 0.38, "learning_rate": 3.435807462262032e-05, "loss": 1.3004, "step": 4251 }, { "epoch": 0.38, "learning_rate": 3.4351426753027254e-05, "loss": 1.2486, "step": 4252 }, { "epoch": 0.38, "learning_rate": 3.4344778114527626e-05, "loss": 1.2766, "step": 4253 }, { "epoch": 0.38, "learning_rate": 3.43381287076681e-05, "loss": 1.2761, "step": 4254 }, { "epoch": 0.38, "learning_rate": 3.4331478532995416e-05, "loss": 1.2068, "step": 4255 }, { "epoch": 0.38, "learning_rate": 3.4324827591056366e-05, "loss": 1.2463, "step": 4256 }, { "epoch": 0.38, "learning_rate": 3.4318175882397824e-05, "loss": 1.2869, "step": 4257 }, { "epoch": 0.38, "learning_rate": 3.431152340756672e-05, "loss": 1.2659, "step": 4258 }, { "epoch": 0.38, "learning_rate": 3.4304870167110035e-05, "loss": 1.2385, "step": 4259 }, { "epoch": 0.38, "learning_rate": 3.429821616157482e-05, "loss": 1.281, "step": 4260 }, { "epoch": 0.38, "learning_rate": 3.4291561391508185e-05, "loss": 1.2921, "step": 4261 }, { "epoch": 0.39, "learning_rate": 3.4284905857457336e-05, "loss": 1.2708, "step": 4262 }, { "epoch": 0.39, "learning_rate": 3.4278249559969484e-05, "loss": 1.2013, "step": 4263 }, { "epoch": 0.39, "learning_rate": 3.4271592499591935e-05, "loss": 1.2718, "step": 4264 }, { "epoch": 0.39, "learning_rate": 3.4264934676872076e-05, "loss": 1.2784, "step": 4265 }, { "epoch": 0.39, "learning_rate": 3.425827609235732e-05, "loss": 1.269, "step": 4266 }, { "epoch": 0.39, "learning_rate": 3.425161674659516e-05, "loss": 1.245, "step": 4267 }, { "epoch": 0.39, "learning_rate": 3.4244956640133164e-05, "loss": 1.2169, "step": 4268 }, { "epoch": 0.39, "learning_rate": 3.4238295773518924e-05, "loss": 1.2775, "step": 4269 }, { "epoch": 0.39, "learning_rate": 3.423163414730014e-05, "loss": 1.2999, "step": 4270 }, { "epoch": 0.39, "learning_rate": 3.422497176202455e-05, "loss": 1.3154, "step": 4271 }, { "epoch": 0.39, "learning_rate": 3.421830861823995e-05, "loss": 1.3025, "step": 4272 }, { "epoch": 0.39, "learning_rate": 3.4211644716494216e-05, "loss": 1.2783, "step": 4273 }, { "epoch": 0.39, "learning_rate": 3.420498005733527e-05, "loss": 1.2312, "step": 4274 }, { "epoch": 0.39, "learning_rate": 3.4198314641311114e-05, "loss": 1.2545, "step": 4275 }, { "epoch": 0.39, "learning_rate": 3.419164846896979e-05, "loss": 1.2783, "step": 4276 }, { "epoch": 0.39, "learning_rate": 3.418498154085941e-05, "loss": 1.2228, "step": 4277 }, { "epoch": 0.39, "learning_rate": 3.417831385752817e-05, "loss": 1.2468, "step": 4278 }, { "epoch": 0.39, "learning_rate": 3.41716454195243e-05, "loss": 1.2596, "step": 4279 }, { "epoch": 0.39, "learning_rate": 3.4164976227396095e-05, "loss": 1.2333, "step": 4280 }, { "epoch": 0.39, "learning_rate": 3.415830628169193e-05, "loss": 1.2161, "step": 4281 }, { "epoch": 0.39, "learning_rate": 3.415163558296022e-05, "loss": 1.3071, "step": 4282 }, { "epoch": 0.39, "learning_rate": 3.414496413174947e-05, "loss": 1.2699, "step": 4283 }, { "epoch": 0.39, "learning_rate": 3.413829192860821e-05, "loss": 1.2652, "step": 4284 }, { "epoch": 0.39, "learning_rate": 3.413161897408506e-05, "loss": 1.2268, "step": 4285 }, { "epoch": 0.39, "learning_rate": 3.41249452687287e-05, "loss": 1.2523, "step": 4286 }, { "epoch": 0.39, "learning_rate": 3.411827081308786e-05, "loss": 1.2758, "step": 4287 }, { "epoch": 0.39, "learning_rate": 3.411159560771133e-05, "loss": 1.2347, "step": 4288 }, { "epoch": 0.39, "learning_rate": 3.4104919653147985e-05, "loss": 1.2547, "step": 4289 }, { "epoch": 0.39, "learning_rate": 3.409824294994672e-05, "loss": 1.2671, "step": 4290 }, { "epoch": 0.39, "learning_rate": 3.409156549865654e-05, "loss": 1.2565, "step": 4291 }, { "epoch": 0.39, "learning_rate": 3.4084887299826476e-05, "loss": 1.2289, "step": 4292 }, { "epoch": 0.39, "learning_rate": 3.407820835400563e-05, "loss": 1.3026, "step": 4293 }, { "epoch": 0.39, "learning_rate": 3.407152866174318e-05, "loss": 1.2528, "step": 4294 }, { "epoch": 0.39, "learning_rate": 3.4064848223588344e-05, "loss": 1.2576, "step": 4295 }, { "epoch": 0.39, "learning_rate": 3.405816704009042e-05, "loss": 1.2247, "step": 4296 }, { "epoch": 0.39, "learning_rate": 3.4051485111798743e-05, "loss": 1.2658, "step": 4297 }, { "epoch": 0.39, "learning_rate": 3.404480243926274e-05, "loss": 1.32, "step": 4298 }, { "epoch": 0.39, "learning_rate": 3.403811902303187e-05, "loss": 1.218, "step": 4299 }, { "epoch": 0.39, "learning_rate": 3.403143486365568e-05, "loss": 1.28, "step": 4300 }, { "epoch": 0.39, "learning_rate": 3.402474996168374e-05, "loss": 1.2415, "step": 4301 }, { "epoch": 0.39, "learning_rate": 3.401806431766574e-05, "loss": 1.1972, "step": 4302 }, { "epoch": 0.39, "learning_rate": 3.401137793215137e-05, "loss": 1.2744, "step": 4303 }, { "epoch": 0.39, "learning_rate": 3.4004690805690424e-05, "loss": 1.2574, "step": 4304 }, { "epoch": 0.39, "learning_rate": 3.399800293883273e-05, "loss": 1.2706, "step": 4305 }, { "epoch": 0.39, "learning_rate": 3.399131433212819e-05, "loss": 1.2461, "step": 4306 }, { "epoch": 0.39, "learning_rate": 3.3984624986126774e-05, "loss": 1.198, "step": 4307 }, { "epoch": 0.39, "learning_rate": 3.397793490137849e-05, "loss": 1.2764, "step": 4308 }, { "epoch": 0.39, "learning_rate": 3.3971244078433426e-05, "loss": 1.283, "step": 4309 }, { "epoch": 0.39, "learning_rate": 3.3964552517841716e-05, "loss": 1.2694, "step": 4310 }, { "epoch": 0.39, "learning_rate": 3.395786022015357e-05, "loss": 1.2091, "step": 4311 }, { "epoch": 0.39, "learning_rate": 3.3951167185919264e-05, "loss": 1.2331, "step": 4312 }, { "epoch": 0.39, "learning_rate": 3.394447341568911e-05, "loss": 1.2419, "step": 4313 }, { "epoch": 0.39, "learning_rate": 3.3937778910013484e-05, "loss": 1.2998, "step": 4314 }, { "epoch": 0.39, "learning_rate": 3.3931083669442845e-05, "loss": 1.2794, "step": 4315 }, { "epoch": 0.39, "learning_rate": 3.3924387694527695e-05, "loss": 1.2655, "step": 4316 }, { "epoch": 0.39, "learning_rate": 3.391769098581859e-05, "loss": 1.262, "step": 4317 }, { "epoch": 0.39, "learning_rate": 3.391099354386619e-05, "loss": 1.231, "step": 4318 }, { "epoch": 0.39, "learning_rate": 3.390429536922113e-05, "loss": 1.2409, "step": 4319 }, { "epoch": 0.39, "learning_rate": 3.38975964624342e-05, "loss": 1.2456, "step": 4320 }, { "epoch": 0.39, "learning_rate": 3.3890896824056196e-05, "loss": 1.254, "step": 4321 }, { "epoch": 0.39, "learning_rate": 3.3884196454637974e-05, "loss": 1.2518, "step": 4322 }, { "epoch": 0.39, "learning_rate": 3.387749535473047e-05, "loss": 1.27, "step": 4323 }, { "epoch": 0.39, "learning_rate": 3.387079352488467e-05, "loss": 1.2777, "step": 4324 }, { "epoch": 0.39, "learning_rate": 3.386409096565163e-05, "loss": 1.2711, "step": 4325 }, { "epoch": 0.39, "learning_rate": 3.3857387677582444e-05, "loss": 1.2812, "step": 4326 }, { "epoch": 0.39, "learning_rate": 3.385068366122828e-05, "loss": 1.2888, "step": 4327 }, { "epoch": 0.39, "learning_rate": 3.3843978917140377e-05, "loss": 1.2165, "step": 4328 }, { "epoch": 0.39, "learning_rate": 3.383727344587002e-05, "loss": 1.2455, "step": 4329 }, { "epoch": 0.39, "learning_rate": 3.383056724796854e-05, "loss": 1.2282, "step": 4330 }, { "epoch": 0.39, "learning_rate": 3.382386032398737e-05, "loss": 1.3413, "step": 4331 }, { "epoch": 0.39, "learning_rate": 3.381715267447795e-05, "loss": 1.2887, "step": 4332 }, { "epoch": 0.39, "learning_rate": 3.381044429999182e-05, "loss": 1.2029, "step": 4333 }, { "epoch": 0.39, "learning_rate": 3.3803735201080565e-05, "loss": 1.2048, "step": 4334 }, { "epoch": 0.39, "learning_rate": 3.379702537829583e-05, "loss": 1.2402, "step": 4335 }, { "epoch": 0.39, "learning_rate": 3.379031483218932e-05, "loss": 1.2441, "step": 4336 }, { "epoch": 0.39, "learning_rate": 3.378360356331279e-05, "loss": 1.2603, "step": 4337 }, { "epoch": 0.39, "learning_rate": 3.377689157221809e-05, "loss": 1.2491, "step": 4338 }, { "epoch": 0.39, "learning_rate": 3.377017885945707e-05, "loss": 1.2388, "step": 4339 }, { "epoch": 0.39, "learning_rate": 3.376346542558169e-05, "loss": 1.2348, "step": 4340 }, { "epoch": 0.39, "learning_rate": 3.375675127114394e-05, "loss": 1.1809, "step": 4341 }, { "epoch": 0.39, "learning_rate": 3.3750036396695906e-05, "loss": 1.2979, "step": 4342 }, { "epoch": 0.39, "learning_rate": 3.374332080278968e-05, "loss": 1.2967, "step": 4343 }, { "epoch": 0.39, "learning_rate": 3.3736604489977466e-05, "loss": 1.214, "step": 4344 }, { "epoch": 0.39, "learning_rate": 3.372988745881147e-05, "loss": 1.2973, "step": 4345 }, { "epoch": 0.39, "learning_rate": 3.3723169709844026e-05, "loss": 1.2431, "step": 4346 }, { "epoch": 0.39, "learning_rate": 3.371645124362747e-05, "loss": 1.2555, "step": 4347 }, { "epoch": 0.39, "learning_rate": 3.370973206071422e-05, "loss": 1.2108, "step": 4348 }, { "epoch": 0.39, "learning_rate": 3.370301216165675e-05, "loss": 1.238, "step": 4349 }, { "epoch": 0.39, "learning_rate": 3.3696291547007596e-05, "loss": 1.2571, "step": 4350 }, { "epoch": 0.39, "learning_rate": 3.3689570217319355e-05, "loss": 1.2313, "step": 4351 }, { "epoch": 0.39, "learning_rate": 3.368284817314466e-05, "loss": 1.2916, "step": 4352 }, { "epoch": 0.39, "learning_rate": 3.367612541503623e-05, "loss": 1.2859, "step": 4353 }, { "epoch": 0.39, "learning_rate": 3.366940194354685e-05, "loss": 1.2586, "step": 4354 }, { "epoch": 0.39, "learning_rate": 3.366267775922933e-05, "loss": 1.2581, "step": 4355 }, { "epoch": 0.39, "learning_rate": 3.3655952862636554e-05, "loss": 1.2659, "step": 4356 }, { "epoch": 0.39, "learning_rate": 3.364922725432147e-05, "loss": 1.2343, "step": 4357 }, { "epoch": 0.39, "learning_rate": 3.364250093483708e-05, "loss": 1.2989, "step": 4358 }, { "epoch": 0.39, "learning_rate": 3.363577390473646e-05, "loss": 1.2767, "step": 4359 }, { "epoch": 0.39, "learning_rate": 3.36290461645727e-05, "loss": 1.2545, "step": 4360 }, { "epoch": 0.39, "learning_rate": 3.3622317714899e-05, "loss": 1.2945, "step": 4361 }, { "epoch": 0.39, "learning_rate": 3.36155885562686e-05, "loss": 1.2616, "step": 4362 }, { "epoch": 0.39, "learning_rate": 3.360885868923477e-05, "loss": 1.3464, "step": 4363 }, { "epoch": 0.39, "learning_rate": 3.360212811435088e-05, "loss": 1.2068, "step": 4364 }, { "epoch": 0.39, "learning_rate": 3.359539683217035e-05, "loss": 1.2589, "step": 4365 }, { "epoch": 0.39, "learning_rate": 3.358866484324662e-05, "loss": 1.2071, "step": 4366 }, { "epoch": 0.39, "learning_rate": 3.3581932148133254e-05, "loss": 1.3074, "step": 4367 }, { "epoch": 0.39, "learning_rate": 3.3575198747383815e-05, "loss": 1.2522, "step": 4368 }, { "epoch": 0.39, "learning_rate": 3.356846464155195e-05, "loss": 1.3084, "step": 4369 }, { "epoch": 0.39, "learning_rate": 3.356172983119136e-05, "loss": 1.2573, "step": 4370 }, { "epoch": 0.39, "learning_rate": 3.355499431685581e-05, "loss": 1.2458, "step": 4371 }, { "epoch": 0.4, "learning_rate": 3.3548258099099105e-05, "loss": 1.243, "step": 4372 }, { "epoch": 0.4, "learning_rate": 3.3541521178475136e-05, "loss": 1.2083, "step": 4373 }, { "epoch": 0.4, "learning_rate": 3.353478355553781e-05, "loss": 1.2707, "step": 4374 }, { "epoch": 0.4, "learning_rate": 3.3528045230841155e-05, "loss": 1.2437, "step": 4375 }, { "epoch": 0.4, "learning_rate": 3.352130620493919e-05, "loss": 1.3321, "step": 4376 }, { "epoch": 0.4, "learning_rate": 3.351456647838603e-05, "loss": 1.2565, "step": 4377 }, { "epoch": 0.4, "learning_rate": 3.3507826051735847e-05, "loss": 1.296, "step": 4378 }, { "epoch": 0.4, "learning_rate": 3.350108492554284e-05, "loss": 1.2661, "step": 4379 }, { "epoch": 0.4, "learning_rate": 3.349434310036132e-05, "loss": 1.1899, "step": 4380 }, { "epoch": 0.4, "learning_rate": 3.3487600576745595e-05, "loss": 1.2597, "step": 4381 }, { "epoch": 0.4, "learning_rate": 3.348085735525006e-05, "loss": 1.214, "step": 4382 }, { "epoch": 0.4, "learning_rate": 3.347411343642919e-05, "loss": 1.2959, "step": 4383 }, { "epoch": 0.4, "learning_rate": 3.346736882083747e-05, "loss": 1.3081, "step": 4384 }, { "epoch": 0.4, "learning_rate": 3.346062350902948e-05, "loss": 1.2934, "step": 4385 }, { "epoch": 0.4, "learning_rate": 3.3453877501559825e-05, "loss": 1.3336, "step": 4386 }, { "epoch": 0.4, "learning_rate": 3.3447130798983196e-05, "loss": 1.2339, "step": 4387 }, { "epoch": 0.4, "learning_rate": 3.344038340185434e-05, "loss": 1.2942, "step": 4388 }, { "epoch": 0.4, "learning_rate": 3.343363531072803e-05, "loss": 1.2883, "step": 4389 }, { "epoch": 0.4, "learning_rate": 3.342688652615913e-05, "loss": 1.1759, "step": 4390 }, { "epoch": 0.4, "learning_rate": 3.342013704870256e-05, "loss": 1.269, "step": 4391 }, { "epoch": 0.4, "learning_rate": 3.3413386878913264e-05, "loss": 1.3022, "step": 4392 }, { "epoch": 0.4, "learning_rate": 3.3406636017346284e-05, "loss": 1.2598, "step": 4393 }, { "epoch": 0.4, "learning_rate": 3.339988446455667e-05, "loss": 1.2763, "step": 4394 }, { "epoch": 0.4, "learning_rate": 3.3393132221099585e-05, "loss": 1.2977, "step": 4395 }, { "epoch": 0.4, "learning_rate": 3.338637928753021e-05, "loss": 1.233, "step": 4396 }, { "epoch": 0.4, "learning_rate": 3.3379625664403804e-05, "loss": 1.2502, "step": 4397 }, { "epoch": 0.4, "learning_rate": 3.337287135227566e-05, "loss": 1.2592, "step": 4398 }, { "epoch": 0.4, "learning_rate": 3.336611635170114e-05, "loss": 1.3021, "step": 4399 }, { "epoch": 0.4, "learning_rate": 3.3359360663235676e-05, "loss": 1.2768, "step": 4400 }, { "epoch": 0.4, "learning_rate": 3.335260428743475e-05, "loss": 1.244, "step": 4401 }, { "epoch": 0.4, "learning_rate": 3.334584722485387e-05, "loss": 1.2493, "step": 4402 }, { "epoch": 0.4, "learning_rate": 3.333908947604864e-05, "loss": 1.1954, "step": 4403 }, { "epoch": 0.4, "learning_rate": 3.3332331041574706e-05, "loss": 1.2902, "step": 4404 }, { "epoch": 0.4, "learning_rate": 3.332557192198777e-05, "loss": 1.2388, "step": 4405 }, { "epoch": 0.4, "learning_rate": 3.331881211784359e-05, "loss": 1.241, "step": 4406 }, { "epoch": 0.4, "learning_rate": 3.331205162969796e-05, "loss": 1.2562, "step": 4407 }, { "epoch": 0.4, "learning_rate": 3.3305290458106786e-05, "loss": 1.2998, "step": 4408 }, { "epoch": 0.4, "learning_rate": 3.329852860362597e-05, "loss": 1.2474, "step": 4409 }, { "epoch": 0.4, "learning_rate": 3.329176606681151e-05, "loss": 1.2809, "step": 4410 }, { "epoch": 0.4, "learning_rate": 3.328500284821943e-05, "loss": 1.2516, "step": 4411 }, { "epoch": 0.4, "learning_rate": 3.327823894840583e-05, "loss": 1.2207, "step": 4412 }, { "epoch": 0.4, "learning_rate": 3.327147436792688e-05, "loss": 1.2187, "step": 4413 }, { "epoch": 0.4, "learning_rate": 3.326470910733875e-05, "loss": 1.2613, "step": 4414 }, { "epoch": 0.4, "learning_rate": 3.3257943167197734e-05, "loss": 1.2626, "step": 4415 }, { "epoch": 0.4, "learning_rate": 3.325117654806014e-05, "loss": 1.2506, "step": 4416 }, { "epoch": 0.4, "learning_rate": 3.324440925048234e-05, "loss": 1.2886, "step": 4417 }, { "epoch": 0.4, "learning_rate": 3.3237641275020784e-05, "loss": 1.2488, "step": 4418 }, { "epoch": 0.4, "learning_rate": 3.323087262223193e-05, "loss": 1.2807, "step": 4419 }, { "epoch": 0.4, "learning_rate": 3.3224103292672334e-05, "loss": 1.2229, "step": 4420 }, { "epoch": 0.4, "learning_rate": 3.3217333286898585e-05, "loss": 1.257, "step": 4421 }, { "epoch": 0.4, "learning_rate": 3.3210562605467356e-05, "loss": 1.2817, "step": 4422 }, { "epoch": 0.4, "learning_rate": 3.320379124893534e-05, "loss": 1.2483, "step": 4423 }, { "epoch": 0.4, "learning_rate": 3.31970192178593e-05, "loss": 1.2839, "step": 4424 }, { "epoch": 0.4, "learning_rate": 3.3190246512796056e-05, "loss": 1.244, "step": 4425 }, { "epoch": 0.4, "learning_rate": 3.31834731343025e-05, "loss": 1.2746, "step": 4426 }, { "epoch": 0.4, "learning_rate": 3.3176699082935545e-05, "loss": 1.2909, "step": 4427 }, { "epoch": 0.4, "learning_rate": 3.316992435925218e-05, "loss": 1.2628, "step": 4428 }, { "epoch": 0.4, "learning_rate": 3.316314896380944e-05, "loss": 1.3055, "step": 4429 }, { "epoch": 0.4, "learning_rate": 3.3156372897164447e-05, "loss": 1.1972, "step": 4430 }, { "epoch": 0.4, "learning_rate": 3.314959615987433e-05, "loss": 1.3143, "step": 4431 }, { "epoch": 0.4, "learning_rate": 3.31428187524963e-05, "loss": 1.2266, "step": 4432 }, { "epoch": 0.4, "learning_rate": 3.313604067558761e-05, "loss": 1.3938, "step": 4433 }, { "epoch": 0.4, "learning_rate": 3.3129261929705594e-05, "loss": 1.2118, "step": 4434 }, { "epoch": 0.4, "learning_rate": 3.312248251540761e-05, "loss": 1.2323, "step": 4435 }, { "epoch": 0.4, "learning_rate": 3.31157024332511e-05, "loss": 1.2889, "step": 4436 }, { "epoch": 0.4, "learning_rate": 3.310892168379354e-05, "loss": 1.2701, "step": 4437 }, { "epoch": 0.4, "learning_rate": 3.310214026759245e-05, "loss": 1.2812, "step": 4438 }, { "epoch": 0.4, "learning_rate": 3.309535818520544e-05, "loss": 1.2882, "step": 4439 }, { "epoch": 0.4, "learning_rate": 3.308857543719016e-05, "loss": 1.2711, "step": 4440 }, { "epoch": 0.4, "learning_rate": 3.308179202410429e-05, "loss": 1.1698, "step": 4441 }, { "epoch": 0.4, "learning_rate": 3.30750079465056e-05, "loss": 1.2786, "step": 4442 }, { "epoch": 0.4, "learning_rate": 3.3068223204951906e-05, "loss": 1.2883, "step": 4443 }, { "epoch": 0.4, "learning_rate": 3.306143780000106e-05, "loss": 1.2175, "step": 4444 }, { "epoch": 0.4, "learning_rate": 3.305465173221099e-05, "loss": 1.211, "step": 4445 }, { "epoch": 0.4, "learning_rate": 3.3047865002139656e-05, "loss": 1.2821, "step": 4446 }, { "epoch": 0.4, "learning_rate": 3.30410776103451e-05, "loss": 1.2356, "step": 4447 }, { "epoch": 0.4, "learning_rate": 3.303428955738541e-05, "loss": 1.2898, "step": 4448 }, { "epoch": 0.4, "learning_rate": 3.30275008438187e-05, "loss": 1.2384, "step": 4449 }, { "epoch": 0.4, "learning_rate": 3.302071147020319e-05, "loss": 1.3143, "step": 4450 }, { "epoch": 0.4, "learning_rate": 3.3013921437097105e-05, "loss": 1.2033, "step": 4451 }, { "epoch": 0.4, "learning_rate": 3.300713074505875e-05, "loss": 1.2931, "step": 4452 }, { "epoch": 0.4, "learning_rate": 3.3000339394646484e-05, "loss": 1.2134, "step": 4453 }, { "epoch": 0.4, "learning_rate": 3.299354738641871e-05, "loss": 1.2823, "step": 4454 }, { "epoch": 0.4, "learning_rate": 3.2986754720933894e-05, "loss": 1.1995, "step": 4455 }, { "epoch": 0.4, "learning_rate": 3.297996139875055e-05, "loss": 1.234, "step": 4456 }, { "epoch": 0.4, "learning_rate": 3.297316742042725e-05, "loss": 1.2775, "step": 4457 }, { "epoch": 0.4, "learning_rate": 3.2966372786522614e-05, "loss": 1.233, "step": 4458 }, { "epoch": 0.4, "learning_rate": 3.2959577497595325e-05, "loss": 1.2817, "step": 4459 }, { "epoch": 0.4, "learning_rate": 3.295278155420412e-05, "loss": 1.2648, "step": 4460 }, { "epoch": 0.4, "learning_rate": 3.294598495690777e-05, "loss": 1.2701, "step": 4461 }, { "epoch": 0.4, "learning_rate": 3.293918770626512e-05, "loss": 1.2701, "step": 4462 }, { "epoch": 0.4, "learning_rate": 3.293238980283508e-05, "loss": 1.29, "step": 4463 }, { "epoch": 0.4, "learning_rate": 3.292559124717658e-05, "loss": 1.2414, "step": 4464 }, { "epoch": 0.4, "learning_rate": 3.291879203984861e-05, "loss": 1.3331, "step": 4465 }, { "epoch": 0.4, "learning_rate": 3.291199218141026e-05, "loss": 1.2981, "step": 4466 }, { "epoch": 0.4, "learning_rate": 3.29051916724206e-05, "loss": 1.2508, "step": 4467 }, { "epoch": 0.4, "learning_rate": 3.2898390513438815e-05, "loss": 1.2659, "step": 4468 }, { "epoch": 0.4, "learning_rate": 3.2891588705024115e-05, "loss": 1.2609, "step": 4469 }, { "epoch": 0.4, "learning_rate": 3.288478624773576e-05, "loss": 1.2262, "step": 4470 }, { "epoch": 0.4, "learning_rate": 3.2877983142133083e-05, "loss": 1.3179, "step": 4471 }, { "epoch": 0.4, "learning_rate": 3.287117938877545e-05, "loss": 1.2635, "step": 4472 }, { "epoch": 0.4, "learning_rate": 3.286437498822229e-05, "loss": 1.2292, "step": 4473 }, { "epoch": 0.4, "learning_rate": 3.285756994103309e-05, "loss": 1.2977, "step": 4474 }, { "epoch": 0.4, "learning_rate": 3.285076424776737e-05, "loss": 1.2918, "step": 4475 }, { "epoch": 0.4, "learning_rate": 3.284395790898475e-05, "loss": 1.2916, "step": 4476 }, { "epoch": 0.4, "learning_rate": 3.2837150925244833e-05, "loss": 1.2341, "step": 4477 }, { "epoch": 0.4, "learning_rate": 3.283034329710734e-05, "loss": 1.1907, "step": 4478 }, { "epoch": 0.4, "learning_rate": 3.2823535025131994e-05, "loss": 1.2456, "step": 4479 }, { "epoch": 0.4, "learning_rate": 3.281672610987861e-05, "loss": 1.2477, "step": 4480 }, { "epoch": 0.4, "learning_rate": 3.280991655190705e-05, "loss": 1.2288, "step": 4481 }, { "epoch": 0.4, "learning_rate": 3.280310635177719e-05, "loss": 1.236, "step": 4482 }, { "epoch": 0.41, "learning_rate": 3.279629551004901e-05, "loss": 1.2661, "step": 4483 }, { "epoch": 0.41, "learning_rate": 3.278948402728252e-05, "loss": 1.2729, "step": 4484 }, { "epoch": 0.41, "learning_rate": 3.2782671904037785e-05, "loss": 1.2838, "step": 4485 }, { "epoch": 0.41, "learning_rate": 3.27758591408749e-05, "loss": 1.2803, "step": 4486 }, { "epoch": 0.41, "learning_rate": 3.276904573835407e-05, "loss": 1.2779, "step": 4487 }, { "epoch": 0.41, "learning_rate": 3.2762231697035475e-05, "loss": 1.2678, "step": 4488 }, { "epoch": 0.41, "learning_rate": 3.275541701747942e-05, "loss": 1.2599, "step": 4489 }, { "epoch": 0.41, "learning_rate": 3.2748601700246227e-05, "loss": 1.3218, "step": 4490 }, { "epoch": 0.41, "learning_rate": 3.274178574589626e-05, "loss": 1.1664, "step": 4491 }, { "epoch": 0.41, "learning_rate": 3.273496915498997e-05, "loss": 1.2667, "step": 4492 }, { "epoch": 0.41, "learning_rate": 3.272815192808782e-05, "loss": 1.2808, "step": 4493 }, { "epoch": 0.41, "learning_rate": 3.2721334065750364e-05, "loss": 1.2256, "step": 4494 }, { "epoch": 0.41, "learning_rate": 3.2714515568538185e-05, "loss": 1.2797, "step": 4495 }, { "epoch": 0.41, "learning_rate": 3.270769643701191e-05, "loss": 1.2451, "step": 4496 }, { "epoch": 0.41, "learning_rate": 3.2700876671732254e-05, "loss": 1.2851, "step": 4497 }, { "epoch": 0.41, "learning_rate": 3.2694056273259946e-05, "loss": 1.2103, "step": 4498 }, { "epoch": 0.41, "learning_rate": 3.2687235242155786e-05, "loss": 1.2493, "step": 4499 }, { "epoch": 0.41, "learning_rate": 3.2680413578980625e-05, "loss": 1.2023, "step": 4500 }, { "epoch": 0.41, "learning_rate": 3.267359128429536e-05, "loss": 1.2775, "step": 4501 }, { "epoch": 0.41, "learning_rate": 3.2666768358660946e-05, "loss": 1.3365, "step": 4502 }, { "epoch": 0.41, "learning_rate": 3.265994480263839e-05, "loss": 1.2914, "step": 4503 }, { "epoch": 0.41, "learning_rate": 3.265312061678874e-05, "loss": 1.2426, "step": 4504 }, { "epoch": 0.41, "learning_rate": 3.264629580167312e-05, "loss": 1.2647, "step": 4505 }, { "epoch": 0.41, "learning_rate": 3.2639470357852666e-05, "loss": 1.2871, "step": 4506 }, { "epoch": 0.41, "learning_rate": 3.2632644285888606e-05, "loss": 1.2152, "step": 4507 }, { "epoch": 0.41, "learning_rate": 3.262581758634221e-05, "loss": 1.2675, "step": 4508 }, { "epoch": 0.41, "learning_rate": 3.261899025977477e-05, "loss": 1.2365, "step": 4509 }, { "epoch": 0.41, "learning_rate": 3.261216230674768e-05, "loss": 1.2239, "step": 4510 }, { "epoch": 0.41, "learning_rate": 3.260533372782234e-05, "loss": 1.2726, "step": 4511 }, { "epoch": 0.41, "learning_rate": 3.259850452356021e-05, "loss": 1.2818, "step": 4512 }, { "epoch": 0.41, "learning_rate": 3.259167469452283e-05, "loss": 1.2615, "step": 4513 }, { "epoch": 0.41, "learning_rate": 3.258484424127177e-05, "loss": 1.2322, "step": 4514 }, { "epoch": 0.41, "learning_rate": 3.257801316436865e-05, "loss": 1.24, "step": 4515 }, { "epoch": 0.41, "learning_rate": 3.257118146437514e-05, "loss": 1.2844, "step": 4516 }, { "epoch": 0.41, "learning_rate": 3.256434914185296e-05, "loss": 1.2321, "step": 4517 }, { "epoch": 0.41, "learning_rate": 3.255751619736391e-05, "loss": 1.2528, "step": 4518 }, { "epoch": 0.41, "learning_rate": 3.25506826314698e-05, "loss": 1.2374, "step": 4519 }, { "epoch": 0.41, "learning_rate": 3.254384844473252e-05, "loss": 1.203, "step": 4520 }, { "epoch": 0.41, "learning_rate": 3.253701363771399e-05, "loss": 1.249, "step": 4521 }, { "epoch": 0.41, "learning_rate": 3.253017821097619e-05, "loss": 1.2351, "step": 4522 }, { "epoch": 0.41, "learning_rate": 3.2523342165081174e-05, "loss": 1.2552, "step": 4523 }, { "epoch": 0.41, "learning_rate": 3.2516505500590996e-05, "loss": 1.1917, "step": 4524 }, { "epoch": 0.41, "learning_rate": 3.25096682180678e-05, "loss": 1.2411, "step": 4525 }, { "epoch": 0.41, "learning_rate": 3.250283031807379e-05, "loss": 1.2449, "step": 4526 }, { "epoch": 0.41, "learning_rate": 3.249599180117118e-05, "loss": 1.2245, "step": 4527 }, { "epoch": 0.41, "learning_rate": 3.248915266792226e-05, "loss": 1.2351, "step": 4528 }, { "epoch": 0.41, "learning_rate": 3.248231291888937e-05, "loss": 1.2192, "step": 4529 }, { "epoch": 0.41, "learning_rate": 3.24754725546349e-05, "loss": 1.2865, "step": 4530 }, { "epoch": 0.41, "learning_rate": 3.246863157572129e-05, "loss": 1.1969, "step": 4531 }, { "epoch": 0.41, "learning_rate": 3.246178998271102e-05, "loss": 1.2764, "step": 4532 }, { "epoch": 0.41, "learning_rate": 3.2454947776166636e-05, "loss": 1.2296, "step": 4533 }, { "epoch": 0.41, "learning_rate": 3.244810495665073e-05, "loss": 1.2432, "step": 4534 }, { "epoch": 0.41, "learning_rate": 3.244126152472594e-05, "loss": 1.27, "step": 4535 }, { "epoch": 0.41, "learning_rate": 3.2434417480954946e-05, "loss": 1.3128, "step": 4536 }, { "epoch": 0.41, "learning_rate": 3.24275728259005e-05, "loss": 1.2394, "step": 4537 }, { "epoch": 0.41, "learning_rate": 3.24207275601254e-05, "loss": 1.2557, "step": 4538 }, { "epoch": 0.41, "learning_rate": 3.241388168419247e-05, "loss": 1.2586, "step": 4539 }, { "epoch": 0.41, "learning_rate": 3.240703519866462e-05, "loss": 1.2018, "step": 4540 }, { "epoch": 0.41, "learning_rate": 3.240018810410477e-05, "loss": 1.2589, "step": 4541 }, { "epoch": 0.41, "learning_rate": 3.2393340401075924e-05, "loss": 1.2296, "step": 4542 }, { "epoch": 0.41, "learning_rate": 3.238649209014112e-05, "loss": 1.2781, "step": 4543 }, { "epoch": 0.41, "learning_rate": 3.2379643171863475e-05, "loss": 1.2249, "step": 4544 }, { "epoch": 0.41, "learning_rate": 3.2372793646806085e-05, "loss": 1.2145, "step": 4545 }, { "epoch": 0.41, "learning_rate": 3.236594351553217e-05, "loss": 1.2488, "step": 4546 }, { "epoch": 0.41, "learning_rate": 3.2359092778604975e-05, "loss": 1.2624, "step": 4547 }, { "epoch": 0.41, "learning_rate": 3.235224143658778e-05, "loss": 1.3115, "step": 4548 }, { "epoch": 0.41, "learning_rate": 3.234538949004393e-05, "loss": 1.3088, "step": 4549 }, { "epoch": 0.41, "learning_rate": 3.233853693953681e-05, "loss": 1.2228, "step": 4550 }, { "epoch": 0.41, "learning_rate": 3.233168378562987e-05, "loss": 1.2695, "step": 4551 }, { "epoch": 0.41, "learning_rate": 3.232483002888659e-05, "loss": 1.2847, "step": 4552 }, { "epoch": 0.41, "learning_rate": 3.231797566987052e-05, "loss": 1.1843, "step": 4553 }, { "epoch": 0.41, "learning_rate": 3.231112070914524e-05, "loss": 1.2107, "step": 4554 }, { "epoch": 0.41, "learning_rate": 3.230426514727439e-05, "loss": 1.2562, "step": 4555 }, { "epoch": 0.41, "learning_rate": 3.229740898482166e-05, "loss": 1.2825, "step": 4556 }, { "epoch": 0.41, "learning_rate": 3.22905522223508e-05, "loss": 1.2594, "step": 4557 }, { "epoch": 0.41, "learning_rate": 3.228369486042557e-05, "loss": 1.2899, "step": 4558 }, { "epoch": 0.41, "learning_rate": 3.227683689960982e-05, "loss": 1.3117, "step": 4559 }, { "epoch": 0.41, "learning_rate": 3.2269978340467435e-05, "loss": 1.3274, "step": 4560 }, { "epoch": 0.41, "learning_rate": 3.2263119183562365e-05, "loss": 1.2052, "step": 4561 }, { "epoch": 0.41, "learning_rate": 3.2256259429458565e-05, "loss": 1.2454, "step": 4562 }, { "epoch": 0.41, "learning_rate": 3.224939907872008e-05, "loss": 1.2463, "step": 4563 }, { "epoch": 0.41, "learning_rate": 3.224253813191099e-05, "loss": 1.242, "step": 4564 }, { "epoch": 0.41, "learning_rate": 3.223567658959544e-05, "loss": 1.2848, "step": 4565 }, { "epoch": 0.41, "learning_rate": 3.222881445233759e-05, "loss": 1.2408, "step": 4566 }, { "epoch": 0.41, "learning_rate": 3.222195172070167e-05, "loss": 1.2338, "step": 4567 }, { "epoch": 0.41, "learning_rate": 3.221508839525197e-05, "loss": 1.3109, "step": 4568 }, { "epoch": 0.41, "learning_rate": 3.220822447655282e-05, "loss": 1.2804, "step": 4569 }, { "epoch": 0.41, "learning_rate": 3.220135996516858e-05, "loss": 1.2746, "step": 4570 }, { "epoch": 0.41, "learning_rate": 3.219449486166367e-05, "loss": 1.2508, "step": 4571 }, { "epoch": 0.41, "learning_rate": 3.218762916660257e-05, "loss": 1.2705, "step": 4572 }, { "epoch": 0.41, "learning_rate": 3.218076288054981e-05, "loss": 1.3124, "step": 4573 }, { "epoch": 0.41, "learning_rate": 3.2173896004069955e-05, "loss": 1.292, "step": 4574 }, { "epoch": 0.41, "learning_rate": 3.216702853772762e-05, "loss": 1.2579, "step": 4575 }, { "epoch": 0.41, "learning_rate": 3.216016048208746e-05, "loss": 1.242, "step": 4576 }, { "epoch": 0.41, "learning_rate": 3.2153291837714214e-05, "loss": 1.2941, "step": 4577 }, { "epoch": 0.41, "learning_rate": 3.2146422605172645e-05, "loss": 1.223, "step": 4578 }, { "epoch": 0.41, "learning_rate": 3.2139552785027535e-05, "loss": 1.2325, "step": 4579 }, { "epoch": 0.41, "learning_rate": 3.213268237784377e-05, "loss": 1.2966, "step": 4580 }, { "epoch": 0.41, "learning_rate": 3.2125811384186254e-05, "loss": 1.2174, "step": 4581 }, { "epoch": 0.41, "learning_rate": 3.211893980461994e-05, "loss": 1.2589, "step": 4582 }, { "epoch": 0.41, "learning_rate": 3.211206763970984e-05, "loss": 1.2375, "step": 4583 }, { "epoch": 0.41, "learning_rate": 3.2105194890021e-05, "loss": 1.2079, "step": 4584 }, { "epoch": 0.41, "learning_rate": 3.2098321556118514e-05, "loss": 1.2575, "step": 4585 }, { "epoch": 0.41, "learning_rate": 3.209144763856756e-05, "loss": 1.3464, "step": 4586 }, { "epoch": 0.41, "learning_rate": 3.208457313793329e-05, "loss": 1.2391, "step": 4587 }, { "epoch": 0.41, "learning_rate": 3.2077698054780996e-05, "loss": 1.2371, "step": 4588 }, { "epoch": 0.41, "learning_rate": 3.207082238967594e-05, "loss": 1.2778, "step": 4589 }, { "epoch": 0.41, "learning_rate": 3.2063946143183475e-05, "loss": 1.2332, "step": 4590 }, { "epoch": 0.41, "learning_rate": 3.205706931586899e-05, "loss": 1.2236, "step": 4591 }, { "epoch": 0.41, "learning_rate": 3.205019190829792e-05, "loss": 1.2724, "step": 4592 }, { "epoch": 0.41, "learning_rate": 3.2043313921035743e-05, "loss": 1.1979, "step": 4593 }, { "epoch": 0.42, "learning_rate": 3.2036435354648e-05, "loss": 1.2615, "step": 4594 }, { "epoch": 0.42, "learning_rate": 3.202955620970027e-05, "loss": 1.302, "step": 4595 }, { "epoch": 0.42, "learning_rate": 3.202267648675817e-05, "loss": 1.3111, "step": 4596 }, { "epoch": 0.42, "learning_rate": 3.201579618638738e-05, "loss": 1.2547, "step": 4597 }, { "epoch": 0.42, "learning_rate": 3.200891530915363e-05, "loss": 1.2691, "step": 4598 }, { "epoch": 0.42, "learning_rate": 3.200203385562268e-05, "loss": 1.2848, "step": 4599 }, { "epoch": 0.42, "learning_rate": 3.199515182636034e-05, "loss": 1.226, "step": 4600 }, { "epoch": 0.42, "learning_rate": 3.19882692219325e-05, "loss": 1.2489, "step": 4601 }, { "epoch": 0.42, "learning_rate": 3.1981386042905045e-05, "loss": 1.2723, "step": 4602 }, { "epoch": 0.42, "learning_rate": 3.197450228984394e-05, "loss": 1.3023, "step": 4603 }, { "epoch": 0.42, "learning_rate": 3.1967617963315204e-05, "loss": 1.2464, "step": 4604 }, { "epoch": 0.42, "learning_rate": 3.1960733063884876e-05, "loss": 1.2727, "step": 4605 }, { "epoch": 0.42, "learning_rate": 3.195384759211906e-05, "loss": 1.3309, "step": 4606 }, { "epoch": 0.42, "learning_rate": 3.1946961548583915e-05, "loss": 1.312, "step": 4607 }, { "epoch": 0.42, "learning_rate": 3.194007493384561e-05, "loss": 1.2396, "step": 4608 }, { "epoch": 0.42, "learning_rate": 3.193318774847041e-05, "loss": 1.2463, "step": 4609 }, { "epoch": 0.42, "learning_rate": 3.192629999302459e-05, "loss": 1.2409, "step": 4610 }, { "epoch": 0.42, "learning_rate": 3.19194116680745e-05, "loss": 1.2852, "step": 4611 }, { "epoch": 0.42, "learning_rate": 3.191252277418649e-05, "loss": 1.2725, "step": 4612 }, { "epoch": 0.42, "learning_rate": 3.190563331192702e-05, "loss": 1.2159, "step": 4613 }, { "epoch": 0.42, "learning_rate": 3.189874328186256e-05, "loss": 1.2478, "step": 4614 }, { "epoch": 0.42, "learning_rate": 3.1891852684559615e-05, "loss": 1.2557, "step": 4615 }, { "epoch": 0.42, "learning_rate": 3.188496152058478e-05, "loss": 1.277, "step": 4616 }, { "epoch": 0.42, "learning_rate": 3.187806979050464e-05, "loss": 1.2881, "step": 4617 }, { "epoch": 0.42, "learning_rate": 3.187117749488588e-05, "loss": 1.3132, "step": 4618 }, { "epoch": 0.42, "learning_rate": 3.1864284634295204e-05, "loss": 1.2761, "step": 4619 }, { "epoch": 0.42, "learning_rate": 3.185739120929937e-05, "loss": 1.2409, "step": 4620 }, { "epoch": 0.42, "learning_rate": 3.185049722046516e-05, "loss": 1.247, "step": 4621 }, { "epoch": 0.42, "learning_rate": 3.184360266835944e-05, "loss": 1.2432, "step": 4622 }, { "epoch": 0.42, "learning_rate": 3.183670755354909e-05, "loss": 1.2536, "step": 4623 }, { "epoch": 0.42, "learning_rate": 3.1829811876601066e-05, "loss": 1.2228, "step": 4624 }, { "epoch": 0.42, "learning_rate": 3.182291563808235e-05, "loss": 1.2588, "step": 4625 }, { "epoch": 0.42, "learning_rate": 3.1816018838559966e-05, "loss": 1.197, "step": 4626 }, { "epoch": 0.42, "learning_rate": 3.1809121478600996e-05, "loss": 1.2245, "step": 4627 }, { "epoch": 0.42, "learning_rate": 3.180222355877257e-05, "loss": 1.2282, "step": 4628 }, { "epoch": 0.42, "learning_rate": 3.179532507964185e-05, "loss": 1.2126, "step": 4629 }, { "epoch": 0.42, "learning_rate": 3.178842604177607e-05, "loss": 1.2885, "step": 4630 }, { "epoch": 0.42, "learning_rate": 3.1781526445742474e-05, "loss": 1.2325, "step": 4631 }, { "epoch": 0.42, "learning_rate": 3.1774626292108375e-05, "loss": 1.326, "step": 4632 }, { "epoch": 0.42, "learning_rate": 3.176772558144113e-05, "loss": 1.2661, "step": 4633 }, { "epoch": 0.42, "learning_rate": 3.176082431430814e-05, "loss": 1.276, "step": 4634 }, { "epoch": 0.42, "learning_rate": 3.1753922491276854e-05, "loss": 1.2541, "step": 4635 }, { "epoch": 0.42, "learning_rate": 3.174702011291476e-05, "loss": 1.2558, "step": 4636 }, { "epoch": 0.42, "learning_rate": 3.174011717978939e-05, "loss": 1.1899, "step": 4637 }, { "epoch": 0.42, "learning_rate": 3.173321369246834e-05, "loss": 1.2031, "step": 4638 }, { "epoch": 0.42, "learning_rate": 3.172630965151922e-05, "loss": 1.2584, "step": 4639 }, { "epoch": 0.42, "learning_rate": 3.171940505750972e-05, "loss": 1.3525, "step": 4640 }, { "epoch": 0.42, "learning_rate": 3.171249991100756e-05, "loss": 1.2851, "step": 4641 }, { "epoch": 0.42, "learning_rate": 3.170559421258049e-05, "loss": 1.2542, "step": 4642 }, { "epoch": 0.42, "learning_rate": 3.169868796279634e-05, "loss": 1.2731, "step": 4643 }, { "epoch": 0.42, "learning_rate": 3.169178116222294e-05, "loss": 1.2556, "step": 4644 }, { "epoch": 0.42, "learning_rate": 3.168487381142822e-05, "loss": 1.2513, "step": 4645 }, { "epoch": 0.42, "learning_rate": 3.167796591098011e-05, "loss": 1.2594, "step": 4646 }, { "epoch": 0.42, "learning_rate": 3.1671057461446594e-05, "loss": 1.2292, "step": 4647 }, { "epoch": 0.42, "learning_rate": 3.1664148463395725e-05, "loss": 1.2108, "step": 4648 }, { "epoch": 0.42, "learning_rate": 3.1657238917395584e-05, "loss": 1.249, "step": 4649 }, { "epoch": 0.42, "learning_rate": 3.1650328824014286e-05, "loss": 1.2389, "step": 4650 }, { "epoch": 0.42, "learning_rate": 3.1643418183820006e-05, "loss": 1.2525, "step": 4651 }, { "epoch": 0.42, "learning_rate": 3.163650699738097e-05, "loss": 1.2008, "step": 4652 }, { "epoch": 0.42, "learning_rate": 3.162959526526543e-05, "loss": 1.2242, "step": 4653 }, { "epoch": 0.42, "learning_rate": 3.1622682988041695e-05, "loss": 1.2517, "step": 4654 }, { "epoch": 0.42, "learning_rate": 3.161577016627811e-05, "loss": 1.2142, "step": 4655 }, { "epoch": 0.42, "learning_rate": 3.1608856800543084e-05, "loss": 1.2174, "step": 4656 }, { "epoch": 0.42, "learning_rate": 3.1601942891405054e-05, "loss": 1.2417, "step": 4657 }, { "epoch": 0.42, "learning_rate": 3.15950284394325e-05, "loss": 1.2326, "step": 4658 }, { "epoch": 0.42, "learning_rate": 3.1588113445193957e-05, "loss": 1.2009, "step": 4659 }, { "epoch": 0.42, "learning_rate": 3.1581197909257994e-05, "loss": 1.2284, "step": 4660 }, { "epoch": 0.42, "learning_rate": 3.1574281832193237e-05, "loss": 1.2369, "step": 4661 }, { "epoch": 0.42, "learning_rate": 3.156736521456834e-05, "loss": 1.2167, "step": 4662 }, { "epoch": 0.42, "learning_rate": 3.1560448056952024e-05, "loss": 1.2345, "step": 4663 }, { "epoch": 0.42, "learning_rate": 3.155353035991304e-05, "loss": 1.2668, "step": 4664 }, { "epoch": 0.42, "learning_rate": 3.154661212402017e-05, "loss": 1.2292, "step": 4665 }, { "epoch": 0.42, "learning_rate": 3.153969334984227e-05, "loss": 1.1838, "step": 4666 }, { "epoch": 0.42, "learning_rate": 3.153277403794822e-05, "loss": 1.2738, "step": 4667 }, { "epoch": 0.42, "learning_rate": 3.152585418890695e-05, "loss": 1.285, "step": 4668 }, { "epoch": 0.42, "learning_rate": 3.151893380328744e-05, "loss": 1.2764, "step": 4669 }, { "epoch": 0.42, "learning_rate": 3.15120128816587e-05, "loss": 1.2977, "step": 4670 }, { "epoch": 0.42, "learning_rate": 3.1505091424589786e-05, "loss": 1.2558, "step": 4671 }, { "epoch": 0.42, "learning_rate": 3.1498169432649825e-05, "loss": 1.2961, "step": 4672 }, { "epoch": 0.42, "learning_rate": 3.1491246906407944e-05, "loss": 1.287, "step": 4673 }, { "epoch": 0.42, "learning_rate": 3.148432384643336e-05, "loss": 1.2521, "step": 4674 }, { "epoch": 0.42, "learning_rate": 3.147740025329529e-05, "loss": 1.2355, "step": 4675 }, { "epoch": 0.42, "learning_rate": 3.147047612756302e-05, "loss": 1.2399, "step": 4676 }, { "epoch": 0.42, "learning_rate": 3.146355146980588e-05, "loss": 1.268, "step": 4677 }, { "epoch": 0.42, "learning_rate": 3.145662628059325e-05, "loss": 1.2379, "step": 4678 }, { "epoch": 0.42, "learning_rate": 3.144970056049453e-05, "loss": 1.2246, "step": 4679 }, { "epoch": 0.42, "learning_rate": 3.144277431007917e-05, "loss": 1.3216, "step": 4680 }, { "epoch": 0.42, "learning_rate": 3.143584752991668e-05, "loss": 1.2218, "step": 4681 }, { "epoch": 0.42, "learning_rate": 3.142892022057661e-05, "loss": 1.2488, "step": 4682 }, { "epoch": 0.42, "learning_rate": 3.142199238262854e-05, "loss": 1.213, "step": 4683 }, { "epoch": 0.42, "learning_rate": 3.1415064016642095e-05, "loss": 1.196, "step": 4684 }, { "epoch": 0.42, "learning_rate": 3.140813512318695e-05, "loss": 1.2783, "step": 4685 }, { "epoch": 0.42, "learning_rate": 3.140120570283283e-05, "loss": 1.2405, "step": 4686 }, { "epoch": 0.42, "learning_rate": 3.139427575614951e-05, "loss": 1.2414, "step": 4687 }, { "epoch": 0.42, "learning_rate": 3.1387345283706753e-05, "loss": 1.2153, "step": 4688 }, { "epoch": 0.42, "learning_rate": 3.138041428607444e-05, "loss": 1.2664, "step": 4689 }, { "epoch": 0.42, "learning_rate": 3.1373482763822454e-05, "loss": 1.245, "step": 4690 }, { "epoch": 0.42, "learning_rate": 3.1366550717520724e-05, "loss": 1.2362, "step": 4691 }, { "epoch": 0.42, "learning_rate": 3.1359618147739236e-05, "loss": 1.2905, "step": 4692 }, { "epoch": 0.42, "learning_rate": 3.1352685055047995e-05, "loss": 1.2797, "step": 4693 }, { "epoch": 0.42, "learning_rate": 3.1345751440017075e-05, "loss": 1.255, "step": 4694 }, { "epoch": 0.42, "learning_rate": 3.133881730321658e-05, "loss": 1.257, "step": 4695 }, { "epoch": 0.42, "learning_rate": 3.1331882645216656e-05, "loss": 1.296, "step": 4696 }, { "epoch": 0.42, "learning_rate": 3.132494746658749e-05, "loss": 1.2417, "step": 4697 }, { "epoch": 0.42, "learning_rate": 3.131801176789934e-05, "loss": 1.2198, "step": 4698 }, { "epoch": 0.42, "learning_rate": 3.131107554972245e-05, "loss": 1.2387, "step": 4699 }, { "epoch": 0.42, "learning_rate": 3.130413881262716e-05, "loss": 1.2734, "step": 4700 }, { "epoch": 0.42, "learning_rate": 3.129720155718383e-05, "loss": 1.2215, "step": 4701 }, { "epoch": 0.42, "learning_rate": 3.129026378396285e-05, "loss": 1.2763, "step": 4702 }, { "epoch": 0.42, "learning_rate": 3.128332549353469e-05, "loss": 1.3032, "step": 4703 }, { "epoch": 0.43, "learning_rate": 3.127638668646984e-05, "loss": 1.2712, "step": 4704 }, { "epoch": 0.43, "learning_rate": 3.126944736333881e-05, "loss": 1.2169, "step": 4705 }, { "epoch": 0.43, "learning_rate": 3.126250752471219e-05, "loss": 1.3356, "step": 4706 }, { "epoch": 0.43, "learning_rate": 3.1255567171160594e-05, "loss": 1.2766, "step": 4707 }, { "epoch": 0.43, "learning_rate": 3.12486263032547e-05, "loss": 1.2783, "step": 4708 }, { "epoch": 0.43, "learning_rate": 3.124168492156517e-05, "loss": 1.2756, "step": 4709 }, { "epoch": 0.43, "learning_rate": 3.123474302666278e-05, "loss": 1.2773, "step": 4710 }, { "epoch": 0.43, "learning_rate": 3.122780061911832e-05, "loss": 1.2373, "step": 4711 }, { "epoch": 0.43, "learning_rate": 3.12208576995026e-05, "loss": 1.2848, "step": 4712 }, { "epoch": 0.43, "learning_rate": 3.12139142683865e-05, "loss": 1.2298, "step": 4713 }, { "epoch": 0.43, "learning_rate": 3.120697032634092e-05, "loss": 1.1996, "step": 4714 }, { "epoch": 0.43, "learning_rate": 3.1200025873936836e-05, "loss": 1.2396, "step": 4715 }, { "epoch": 0.43, "learning_rate": 3.119308091174523e-05, "loss": 1.2318, "step": 4716 }, { "epoch": 0.43, "learning_rate": 3.118613544033715e-05, "loss": 1.2614, "step": 4717 }, { "epoch": 0.43, "learning_rate": 3.1179189460283664e-05, "loss": 1.2009, "step": 4718 }, { "epoch": 0.43, "learning_rate": 3.117224297215591e-05, "loss": 1.2338, "step": 4719 }, { "epoch": 0.43, "learning_rate": 3.116529597652505e-05, "loss": 1.1944, "step": 4720 }, { "epoch": 0.43, "learning_rate": 3.115834847396228e-05, "loss": 1.2819, "step": 4721 }, { "epoch": 0.43, "learning_rate": 3.115140046503884e-05, "loss": 1.2927, "step": 4722 }, { "epoch": 0.43, "learning_rate": 3.114445195032604e-05, "loss": 1.2686, "step": 4723 }, { "epoch": 0.43, "learning_rate": 3.113750293039521e-05, "loss": 1.2193, "step": 4724 }, { "epoch": 0.43, "learning_rate": 3.113055340581771e-05, "loss": 1.2438, "step": 4725 }, { "epoch": 0.43, "learning_rate": 3.112360337716496e-05, "loss": 1.2485, "step": 4726 }, { "epoch": 0.43, "learning_rate": 3.1116652845008405e-05, "loss": 1.2551, "step": 4727 }, { "epoch": 0.43, "learning_rate": 3.1109701809919556e-05, "loss": 1.2678, "step": 4728 }, { "epoch": 0.43, "learning_rate": 3.1102750272469946e-05, "loss": 1.3046, "step": 4729 }, { "epoch": 0.43, "learning_rate": 3.1095798233231155e-05, "loss": 1.2084, "step": 4730 }, { "epoch": 0.43, "learning_rate": 3.10888456927748e-05, "loss": 1.2672, "step": 4731 }, { "epoch": 0.43, "learning_rate": 3.108189265167255e-05, "loss": 1.2222, "step": 4732 }, { "epoch": 0.43, "learning_rate": 3.10749391104961e-05, "loss": 1.2583, "step": 4733 }, { "epoch": 0.43, "learning_rate": 3.1067985069817206e-05, "loss": 1.2289, "step": 4734 }, { "epoch": 0.43, "learning_rate": 3.1061030530207634e-05, "loss": 1.2672, "step": 4735 }, { "epoch": 0.43, "learning_rate": 3.105407549223923e-05, "loss": 1.2568, "step": 4736 }, { "epoch": 0.43, "learning_rate": 3.104711995648384e-05, "loss": 1.2289, "step": 4737 }, { "epoch": 0.43, "learning_rate": 3.10401639235134e-05, "loss": 1.2937, "step": 4738 }, { "epoch": 0.43, "learning_rate": 3.103320739389984e-05, "loss": 1.2215, "step": 4739 }, { "epoch": 0.43, "learning_rate": 3.102625036821515e-05, "loss": 1.3111, "step": 4740 }, { "epoch": 0.43, "learning_rate": 3.101929284703137e-05, "loss": 1.2563, "step": 4741 }, { "epoch": 0.43, "learning_rate": 3.101233483092056e-05, "loss": 1.2327, "step": 4742 }, { "epoch": 0.43, "learning_rate": 3.100537632045483e-05, "loss": 1.3189, "step": 4743 }, { "epoch": 0.43, "learning_rate": 3.099841731620635e-05, "loss": 1.2782, "step": 4744 }, { "epoch": 0.43, "learning_rate": 3.0991457818747304e-05, "loss": 1.2693, "step": 4745 }, { "epoch": 0.43, "learning_rate": 3.098449782864992e-05, "loss": 1.189, "step": 4746 }, { "epoch": 0.43, "learning_rate": 3.097753734648649e-05, "loss": 1.2919, "step": 4747 }, { "epoch": 0.43, "learning_rate": 3.097057637282931e-05, "loss": 1.2207, "step": 4748 }, { "epoch": 0.43, "learning_rate": 3.096361490825073e-05, "loss": 1.2421, "step": 4749 }, { "epoch": 0.43, "learning_rate": 3.095665295332318e-05, "loss": 1.2459, "step": 4750 }, { "epoch": 0.43, "learning_rate": 3.094969050861906e-05, "loss": 1.2993, "step": 4751 }, { "epoch": 0.43, "learning_rate": 3.0942727574710864e-05, "loss": 1.271, "step": 4752 }, { "epoch": 0.43, "learning_rate": 3.09357641521711e-05, "loss": 1.2857, "step": 4753 }, { "epoch": 0.43, "learning_rate": 3.0928800241572334e-05, "loss": 1.2168, "step": 4754 }, { "epoch": 0.43, "learning_rate": 3.092183584348716e-05, "loss": 1.2992, "step": 4755 }, { "epoch": 0.43, "learning_rate": 3.091487095848821e-05, "loss": 1.2695, "step": 4756 }, { "epoch": 0.43, "learning_rate": 3.090790558714816e-05, "loss": 1.1918, "step": 4757 }, { "epoch": 0.43, "learning_rate": 3.090093973003973e-05, "loss": 1.2354, "step": 4758 }, { "epoch": 0.43, "learning_rate": 3.0893973387735687e-05, "loss": 1.2014, "step": 4759 }, { "epoch": 0.43, "learning_rate": 3.088700656080881e-05, "loss": 1.2618, "step": 4760 }, { "epoch": 0.43, "learning_rate": 3.0880039249831943e-05, "loss": 1.2349, "step": 4761 }, { "epoch": 0.43, "learning_rate": 3.0873071455377965e-05, "loss": 1.2434, "step": 4762 }, { "epoch": 0.43, "learning_rate": 3.086610317801979e-05, "loss": 1.243, "step": 4763 }, { "epoch": 0.43, "learning_rate": 3.085913441833037e-05, "loss": 1.2511, "step": 4764 }, { "epoch": 0.43, "learning_rate": 3.085216517688271e-05, "loss": 1.287, "step": 4765 }, { "epoch": 0.43, "learning_rate": 3.084519545424984e-05, "loss": 1.2569, "step": 4766 }, { "epoch": 0.43, "learning_rate": 3.0838225251004825e-05, "loss": 1.2697, "step": 4767 }, { "epoch": 0.43, "learning_rate": 3.08312545677208e-05, "loss": 1.2351, "step": 4768 }, { "epoch": 0.43, "learning_rate": 3.082428340497089e-05, "loss": 1.2631, "step": 4769 }, { "epoch": 0.43, "learning_rate": 3.0817311763328315e-05, "loss": 1.2608, "step": 4770 }, { "epoch": 0.43, "learning_rate": 3.0810339643366307e-05, "loss": 1.2446, "step": 4771 }, { "epoch": 0.43, "learning_rate": 3.080336704565812e-05, "loss": 1.2414, "step": 4772 }, { "epoch": 0.43, "learning_rate": 3.079639397077707e-05, "loss": 1.1927, "step": 4773 }, { "epoch": 0.43, "learning_rate": 3.07894204192965e-05, "loss": 1.2475, "step": 4774 }, { "epoch": 0.43, "learning_rate": 3.078244639178983e-05, "loss": 1.2605, "step": 4775 }, { "epoch": 0.43, "learning_rate": 3.077547188883046e-05, "loss": 1.2218, "step": 4776 }, { "epoch": 0.43, "learning_rate": 3.076849691099186e-05, "loss": 1.2113, "step": 4777 }, { "epoch": 0.43, "learning_rate": 3.076152145884755e-05, "loss": 1.2445, "step": 4778 }, { "epoch": 0.43, "learning_rate": 3.075454553297107e-05, "loss": 1.1991, "step": 4779 }, { "epoch": 0.43, "learning_rate": 3.0747569133936e-05, "loss": 1.2534, "step": 4780 }, { "epoch": 0.43, "learning_rate": 3.074059226231597e-05, "loss": 1.2214, "step": 4781 }, { "epoch": 0.43, "learning_rate": 3.0733614918684635e-05, "loss": 1.3339, "step": 4782 }, { "epoch": 0.43, "learning_rate": 3.0726637103615703e-05, "loss": 1.2544, "step": 4783 }, { "epoch": 0.43, "learning_rate": 3.0719658817682905e-05, "loss": 1.239, "step": 4784 }, { "epoch": 0.43, "learning_rate": 3.071268006146003e-05, "loss": 1.2307, "step": 4785 }, { "epoch": 0.43, "learning_rate": 3.0705700835520895e-05, "loss": 1.3042, "step": 4786 }, { "epoch": 0.43, "learning_rate": 3.0698721140439346e-05, "loss": 1.2402, "step": 4787 }, { "epoch": 0.43, "learning_rate": 3.069174097678929e-05, "loss": 1.2828, "step": 4788 }, { "epoch": 0.43, "learning_rate": 3.068476034514466e-05, "loss": 1.2203, "step": 4789 }, { "epoch": 0.43, "learning_rate": 3.067777924607941e-05, "loss": 1.26, "step": 4790 }, { "epoch": 0.43, "learning_rate": 3.0670797680167575e-05, "loss": 1.2701, "step": 4791 }, { "epoch": 0.43, "learning_rate": 3.0663815647983176e-05, "loss": 1.3122, "step": 4792 }, { "epoch": 0.43, "learning_rate": 3.065683315010032e-05, "loss": 1.1999, "step": 4793 }, { "epoch": 0.43, "learning_rate": 3.0649850187093135e-05, "loss": 1.2304, "step": 4794 }, { "epoch": 0.43, "learning_rate": 3.064286675953576e-05, "loss": 1.2979, "step": 4795 }, { "epoch": 0.43, "learning_rate": 3.063588286800243e-05, "loss": 1.2887, "step": 4796 }, { "epoch": 0.43, "learning_rate": 3.0628898513067353e-05, "loss": 1.2934, "step": 4797 }, { "epoch": 0.43, "learning_rate": 3.0621913695304824e-05, "loss": 1.2965, "step": 4798 }, { "epoch": 0.43, "learning_rate": 3.0614928415289155e-05, "loss": 1.2425, "step": 4799 }, { "epoch": 0.43, "learning_rate": 3.06079426735947e-05, "loss": 1.2096, "step": 4800 }, { "epoch": 0.43, "learning_rate": 3.0600956470795854e-05, "loss": 1.2094, "step": 4801 }, { "epoch": 0.43, "learning_rate": 3.059396980746705e-05, "loss": 1.251, "step": 4802 }, { "epoch": 0.43, "learning_rate": 3.058698268418273e-05, "loss": 1.2239, "step": 4803 }, { "epoch": 0.43, "learning_rate": 3.0579995101517436e-05, "loss": 1.2783, "step": 4804 }, { "epoch": 0.43, "learning_rate": 3.05730070600457e-05, "loss": 1.255, "step": 4805 }, { "epoch": 0.43, "learning_rate": 3.056601856034208e-05, "loss": 1.2596, "step": 4806 }, { "epoch": 0.43, "learning_rate": 3.0559029602981216e-05, "loss": 1.2585, "step": 4807 }, { "epoch": 0.43, "learning_rate": 3.0552040188537755e-05, "loss": 1.3086, "step": 4808 }, { "epoch": 0.43, "learning_rate": 3.054505031758641e-05, "loss": 1.2549, "step": 4809 }, { "epoch": 0.43, "learning_rate": 3.0538059990701884e-05, "loss": 1.2401, "step": 4810 }, { "epoch": 0.43, "learning_rate": 3.0531069208458954e-05, "loss": 1.234, "step": 4811 }, { "epoch": 0.43, "learning_rate": 3.0524077971432446e-05, "loss": 1.2126, "step": 4812 }, { "epoch": 0.43, "learning_rate": 3.051708628019719e-05, "loss": 1.2357, "step": 4813 }, { "epoch": 0.43, "learning_rate": 3.0510094135328053e-05, "loss": 1.2862, "step": 4814 }, { "epoch": 0.44, "learning_rate": 3.0503101537399974e-05, "loss": 1.2765, "step": 4815 }, { "epoch": 0.44, "learning_rate": 3.0496108486987894e-05, "loss": 1.2271, "step": 4816 }, { "epoch": 0.44, "learning_rate": 3.0489114984666827e-05, "loss": 1.2213, "step": 4817 }, { "epoch": 0.44, "learning_rate": 3.048212103101178e-05, "loss": 1.2692, "step": 4818 }, { "epoch": 0.44, "learning_rate": 3.0475126626597827e-05, "loss": 1.2954, "step": 4819 }, { "epoch": 0.44, "learning_rate": 3.046813177200008e-05, "loss": 1.2722, "step": 4820 }, { "epoch": 0.44, "learning_rate": 3.0461136467793665e-05, "loss": 1.204, "step": 4821 }, { "epoch": 0.44, "learning_rate": 3.0454140714553775e-05, "loss": 1.2374, "step": 4822 }, { "epoch": 0.44, "learning_rate": 3.044714451285562e-05, "loss": 1.2956, "step": 4823 }, { "epoch": 0.44, "learning_rate": 3.0440147863274444e-05, "loss": 1.2318, "step": 4824 }, { "epoch": 0.44, "learning_rate": 3.043315076638555e-05, "loss": 1.1957, "step": 4825 }, { "epoch": 0.44, "learning_rate": 3.0426153222764253e-05, "loss": 1.2524, "step": 4826 }, { "epoch": 0.44, "learning_rate": 3.0419155232985914e-05, "loss": 1.2994, "step": 4827 }, { "epoch": 0.44, "learning_rate": 3.0412156797625935e-05, "loss": 1.246, "step": 4828 }, { "epoch": 0.44, "learning_rate": 3.040515791725975e-05, "loss": 1.3016, "step": 4829 }, { "epoch": 0.44, "learning_rate": 3.0398158592462845e-05, "loss": 1.2369, "step": 4830 }, { "epoch": 0.44, "learning_rate": 3.0391158823810712e-05, "loss": 1.285, "step": 4831 }, { "epoch": 0.44, "learning_rate": 3.03841586118789e-05, "loss": 1.2202, "step": 4832 }, { "epoch": 0.44, "learning_rate": 3.037715795724299e-05, "loss": 1.2001, "step": 4833 }, { "epoch": 0.44, "learning_rate": 3.0370156860478605e-05, "loss": 1.1809, "step": 4834 }, { "epoch": 0.44, "learning_rate": 3.036315532216139e-05, "loss": 1.25, "step": 4835 }, { "epoch": 0.44, "learning_rate": 3.035615334286705e-05, "loss": 1.2696, "step": 4836 }, { "epoch": 0.44, "learning_rate": 3.0349150923171293e-05, "loss": 1.2514, "step": 4837 }, { "epoch": 0.44, "learning_rate": 3.0342148063649905e-05, "loss": 1.2232, "step": 4838 }, { "epoch": 0.44, "learning_rate": 3.0335144764878663e-05, "loss": 1.271, "step": 4839 }, { "epoch": 0.44, "learning_rate": 3.0328141027433416e-05, "loss": 1.2438, "step": 4840 }, { "epoch": 0.44, "learning_rate": 3.0321136851890036e-05, "loss": 1.2452, "step": 4841 }, { "epoch": 0.44, "learning_rate": 3.0314132238824415e-05, "loss": 1.2452, "step": 4842 }, { "epoch": 0.44, "learning_rate": 3.0307127188812518e-05, "loss": 1.2145, "step": 4843 }, { "epoch": 0.44, "learning_rate": 3.030012170243031e-05, "loss": 1.2663, "step": 4844 }, { "epoch": 0.44, "learning_rate": 3.0293115780253812e-05, "loss": 1.2326, "step": 4845 }, { "epoch": 0.44, "learning_rate": 3.028610942285907e-05, "loss": 1.2622, "step": 4846 }, { "epoch": 0.44, "learning_rate": 3.0279102630822182e-05, "loss": 1.2238, "step": 4847 }, { "epoch": 0.44, "learning_rate": 3.027209540471925e-05, "loss": 1.2665, "step": 4848 }, { "epoch": 0.44, "learning_rate": 3.0265087745126458e-05, "loss": 1.284, "step": 4849 }, { "epoch": 0.44, "learning_rate": 3.0258079652619973e-05, "loss": 1.1986, "step": 4850 }, { "epoch": 0.44, "learning_rate": 3.0251071127776055e-05, "loss": 1.3132, "step": 4851 }, { "epoch": 0.44, "learning_rate": 3.0244062171170946e-05, "loss": 1.2235, "step": 4852 }, { "epoch": 0.44, "learning_rate": 3.0237052783380946e-05, "loss": 1.2702, "step": 4853 }, { "epoch": 0.44, "learning_rate": 3.023004296498241e-05, "loss": 1.2787, "step": 4854 }, { "epoch": 0.44, "learning_rate": 3.0223032716551692e-05, "loss": 1.2511, "step": 4855 }, { "epoch": 0.44, "learning_rate": 3.0216022038665204e-05, "loss": 1.2272, "step": 4856 }, { "epoch": 0.44, "learning_rate": 3.020901093189939e-05, "loss": 1.2327, "step": 4857 }, { "epoch": 0.44, "learning_rate": 3.020199939683072e-05, "loss": 1.2773, "step": 4858 }, { "epoch": 0.44, "learning_rate": 3.019498743403572e-05, "loss": 1.2823, "step": 4859 }, { "epoch": 0.44, "learning_rate": 3.0187975044090933e-05, "loss": 1.2721, "step": 4860 }, { "epoch": 0.44, "learning_rate": 3.0180962227572935e-05, "loss": 1.2784, "step": 4861 }, { "epoch": 0.44, "learning_rate": 3.017394898505835e-05, "loss": 1.2529, "step": 4862 }, { "epoch": 0.44, "learning_rate": 3.0166935317123823e-05, "loss": 1.2275, "step": 4863 }, { "epoch": 0.44, "learning_rate": 3.015992122434606e-05, "loss": 1.2142, "step": 4864 }, { "epoch": 0.44, "learning_rate": 3.0152906707301764e-05, "loss": 1.2518, "step": 4865 }, { "epoch": 0.44, "learning_rate": 3.01458917665677e-05, "loss": 1.2496, "step": 4866 }, { "epoch": 0.44, "learning_rate": 3.0138876402720668e-05, "loss": 1.2638, "step": 4867 }, { "epoch": 0.44, "learning_rate": 3.0131860616337486e-05, "loss": 1.2094, "step": 4868 }, { "epoch": 0.44, "learning_rate": 3.012484440799502e-05, "loss": 1.2435, "step": 4869 }, { "epoch": 0.44, "learning_rate": 3.0117827778270164e-05, "loss": 1.2556, "step": 4870 }, { "epoch": 0.44, "learning_rate": 3.0110810727739853e-05, "loss": 1.2379, "step": 4871 }, { "epoch": 0.44, "learning_rate": 3.0103793256981057e-05, "loss": 1.2464, "step": 4872 }, { "epoch": 0.44, "learning_rate": 3.009677536657076e-05, "loss": 1.2902, "step": 4873 }, { "epoch": 0.44, "learning_rate": 3.0089757057086016e-05, "loss": 1.1955, "step": 4874 }, { "epoch": 0.44, "learning_rate": 3.0082738329103887e-05, "loss": 1.2478, "step": 4875 }, { "epoch": 0.44, "learning_rate": 3.007571918320148e-05, "loss": 1.2911, "step": 4876 }, { "epoch": 0.44, "learning_rate": 3.0068699619955926e-05, "loss": 1.297, "step": 4877 }, { "epoch": 0.44, "learning_rate": 3.0061679639944403e-05, "loss": 1.2165, "step": 4878 }, { "epoch": 0.44, "learning_rate": 3.005465924374411e-05, "loss": 1.2559, "step": 4879 }, { "epoch": 0.44, "learning_rate": 3.0047638431932306e-05, "loss": 1.2195, "step": 4880 }, { "epoch": 0.44, "learning_rate": 3.004061720508625e-05, "loss": 1.2209, "step": 4881 }, { "epoch": 0.44, "learning_rate": 3.0033595563783262e-05, "loss": 1.2448, "step": 4882 }, { "epoch": 0.44, "learning_rate": 3.002657350860068e-05, "loss": 1.2609, "step": 4883 }, { "epoch": 0.44, "learning_rate": 3.001955104011588e-05, "loss": 1.2375, "step": 4884 }, { "epoch": 0.44, "learning_rate": 3.0012528158906282e-05, "loss": 1.1978, "step": 4885 }, { "epoch": 0.44, "learning_rate": 3.0005504865549318e-05, "loss": 1.2159, "step": 4886 }, { "epoch": 0.44, "learning_rate": 2.999848116062247e-05, "loss": 1.2103, "step": 4887 }, { "epoch": 0.44, "learning_rate": 2.999145704470327e-05, "loss": 1.2642, "step": 4888 }, { "epoch": 0.44, "learning_rate": 2.9984432518369247e-05, "loss": 1.3008, "step": 4889 }, { "epoch": 0.44, "learning_rate": 2.9977407582197987e-05, "loss": 1.2667, "step": 4890 }, { "epoch": 0.44, "learning_rate": 2.9970382236767096e-05, "loss": 1.2349, "step": 4891 }, { "epoch": 0.44, "learning_rate": 2.9963356482654235e-05, "loss": 1.2492, "step": 4892 }, { "epoch": 0.44, "learning_rate": 2.9956330320437086e-05, "loss": 1.2108, "step": 4893 }, { "epoch": 0.44, "learning_rate": 2.9949303750693354e-05, "loss": 1.2386, "step": 4894 }, { "epoch": 0.44, "learning_rate": 2.9942276774000788e-05, "loss": 1.2686, "step": 4895 }, { "epoch": 0.44, "learning_rate": 2.9935249390937183e-05, "loss": 1.221, "step": 4896 }, { "epoch": 0.44, "learning_rate": 2.9928221602080343e-05, "loss": 1.2177, "step": 4897 }, { "epoch": 0.44, "learning_rate": 2.992119340800812e-05, "loss": 1.2567, "step": 4898 }, { "epoch": 0.44, "learning_rate": 2.991416480929839e-05, "loss": 1.2236, "step": 4899 }, { "epoch": 0.44, "learning_rate": 2.9907135806529078e-05, "loss": 1.248, "step": 4900 }, { "epoch": 0.44, "learning_rate": 2.990010640027814e-05, "loss": 1.2716, "step": 4901 }, { "epoch": 0.44, "learning_rate": 2.9893076591123546e-05, "loss": 1.2135, "step": 4902 }, { "epoch": 0.44, "learning_rate": 2.9886046379643312e-05, "loss": 1.2846, "step": 4903 }, { "epoch": 0.44, "learning_rate": 2.9879015766415486e-05, "loss": 1.2503, "step": 4904 }, { "epoch": 0.44, "learning_rate": 2.9871984752018145e-05, "loss": 1.2221, "step": 4905 }, { "epoch": 0.44, "learning_rate": 2.9864953337029423e-05, "loss": 1.2925, "step": 4906 }, { "epoch": 0.44, "learning_rate": 2.9857921522027444e-05, "loss": 1.2095, "step": 4907 }, { "epoch": 0.44, "learning_rate": 2.9850889307590395e-05, "loss": 1.2388, "step": 4908 }, { "epoch": 0.44, "learning_rate": 2.9843856694296502e-05, "loss": 1.2809, "step": 4909 }, { "epoch": 0.44, "learning_rate": 2.9836823682723997e-05, "loss": 1.2039, "step": 4910 }, { "epoch": 0.44, "learning_rate": 2.9829790273451163e-05, "loss": 1.197, "step": 4911 }, { "epoch": 0.44, "learning_rate": 2.9822756467056306e-05, "loss": 1.235, "step": 4912 }, { "epoch": 0.44, "learning_rate": 2.9815722264117768e-05, "loss": 1.2203, "step": 4913 }, { "epoch": 0.44, "learning_rate": 2.9808687665213946e-05, "loss": 1.237, "step": 4914 }, { "epoch": 0.44, "learning_rate": 2.9801652670923225e-05, "loss": 1.3133, "step": 4915 }, { "epoch": 0.44, "learning_rate": 2.979461728182406e-05, "loss": 1.2712, "step": 4916 }, { "epoch": 0.44, "learning_rate": 2.9787581498494916e-05, "loss": 1.2285, "step": 4917 }, { "epoch": 0.44, "learning_rate": 2.9780545321514307e-05, "loss": 1.2752, "step": 4918 }, { "epoch": 0.44, "learning_rate": 2.977350875146077e-05, "loss": 1.2331, "step": 4919 }, { "epoch": 0.44, "learning_rate": 2.976647178891287e-05, "loss": 1.2307, "step": 4920 }, { "epoch": 0.44, "learning_rate": 2.9759434434449214e-05, "loss": 1.2424, "step": 4921 }, { "epoch": 0.44, "learning_rate": 2.9752396688648448e-05, "loss": 1.2998, "step": 4922 }, { "epoch": 0.44, "learning_rate": 2.9745358552089224e-05, "loss": 1.2445, "step": 4923 }, { "epoch": 0.44, "learning_rate": 2.9738320025350254e-05, "loss": 1.3226, "step": 4924 }, { "epoch": 0.44, "learning_rate": 2.9731281109010256e-05, "loss": 1.2841, "step": 4925 }, { "epoch": 0.45, "learning_rate": 2.9724241803648006e-05, "loss": 1.289, "step": 4926 }, { "epoch": 0.45, "learning_rate": 2.97172021098423e-05, "loss": 1.2499, "step": 4927 }, { "epoch": 0.45, "learning_rate": 2.9710162028171962e-05, "loss": 1.2201, "step": 4928 }, { "epoch": 0.45, "learning_rate": 2.9703121559215845e-05, "loss": 1.2429, "step": 4929 }, { "epoch": 0.45, "learning_rate": 2.9696080703552854e-05, "loss": 1.2892, "step": 4930 }, { "epoch": 0.45, "learning_rate": 2.96890394617619e-05, "loss": 1.2397, "step": 4931 }, { "epoch": 0.45, "learning_rate": 2.9681997834421955e-05, "loss": 1.2379, "step": 4932 }, { "epoch": 0.45, "learning_rate": 2.9674955822111984e-05, "loss": 1.2272, "step": 4933 }, { "epoch": 0.45, "learning_rate": 2.9667913425411022e-05, "loss": 1.2218, "step": 4934 }, { "epoch": 0.45, "learning_rate": 2.966087064489812e-05, "loss": 1.3013, "step": 4935 }, { "epoch": 0.45, "learning_rate": 2.965382748115234e-05, "loss": 1.2258, "step": 4936 }, { "epoch": 0.45, "learning_rate": 2.964678393475282e-05, "loss": 1.3023, "step": 4937 }, { "epoch": 0.45, "learning_rate": 2.963974000627869e-05, "loss": 1.2282, "step": 4938 }, { "epoch": 0.45, "learning_rate": 2.9632695696309128e-05, "loss": 1.2413, "step": 4939 }, { "epoch": 0.45, "learning_rate": 2.9625651005423348e-05, "loss": 1.2628, "step": 4940 }, { "epoch": 0.45, "learning_rate": 2.9618605934200578e-05, "loss": 1.2485, "step": 4941 }, { "epoch": 0.45, "learning_rate": 2.9611560483220092e-05, "loss": 1.2685, "step": 4942 }, { "epoch": 0.45, "learning_rate": 2.96045146530612e-05, "loss": 1.3182, "step": 4943 }, { "epoch": 0.45, "learning_rate": 2.9597468444303223e-05, "loss": 1.2367, "step": 4944 }, { "epoch": 0.45, "learning_rate": 2.9590421857525535e-05, "loss": 1.2944, "step": 4945 }, { "epoch": 0.45, "learning_rate": 2.958337489330752e-05, "loss": 1.2653, "step": 4946 }, { "epoch": 0.45, "learning_rate": 2.9576327552228613e-05, "loss": 1.2336, "step": 4947 }, { "epoch": 0.45, "learning_rate": 2.9569279834868264e-05, "loss": 1.2126, "step": 4948 }, { "epoch": 0.45, "learning_rate": 2.9562231741805957e-05, "loss": 1.2233, "step": 4949 }, { "epoch": 0.45, "learning_rate": 2.9555183273621222e-05, "loss": 1.2939, "step": 4950 }, { "epoch": 0.45, "learning_rate": 2.9548134430893604e-05, "loss": 1.1901, "step": 4951 }, { "epoch": 0.45, "learning_rate": 2.954108521420268e-05, "loss": 1.2495, "step": 4952 }, { "epoch": 0.45, "learning_rate": 2.9534035624128063e-05, "loss": 1.2742, "step": 4953 }, { "epoch": 0.45, "learning_rate": 2.9526985661249395e-05, "loss": 1.2591, "step": 4954 }, { "epoch": 0.45, "learning_rate": 2.9519935326146347e-05, "loss": 1.2457, "step": 4955 }, { "epoch": 0.45, "learning_rate": 2.951288461939863e-05, "loss": 1.1913, "step": 4956 }, { "epoch": 0.45, "learning_rate": 2.9505833541585963e-05, "loss": 1.218, "step": 4957 }, { "epoch": 0.45, "learning_rate": 2.949878209328812e-05, "loss": 1.2535, "step": 4958 }, { "epoch": 0.45, "learning_rate": 2.949173027508489e-05, "loss": 1.3047, "step": 4959 }, { "epoch": 0.45, "learning_rate": 2.94846780875561e-05, "loss": 1.2341, "step": 4960 }, { "epoch": 0.45, "learning_rate": 2.947762553128161e-05, "loss": 1.2648, "step": 4961 }, { "epoch": 0.45, "learning_rate": 2.9470572606841297e-05, "loss": 1.2651, "step": 4962 }, { "epoch": 0.45, "learning_rate": 2.946351931481508e-05, "loss": 1.2916, "step": 4963 }, { "epoch": 0.45, "learning_rate": 2.9456465655782905e-05, "loss": 1.2557, "step": 4964 }, { "epoch": 0.45, "learning_rate": 2.9449411630324758e-05, "loss": 1.2799, "step": 4965 }, { "epoch": 0.45, "learning_rate": 2.944235723902063e-05, "loss": 1.1962, "step": 4966 }, { "epoch": 0.45, "learning_rate": 2.943530248245056e-05, "loss": 1.2719, "step": 4967 }, { "epoch": 0.45, "learning_rate": 2.9428247361194628e-05, "loss": 1.2541, "step": 4968 }, { "epoch": 0.45, "learning_rate": 2.9421191875832915e-05, "loss": 1.2411, "step": 4969 }, { "epoch": 0.45, "learning_rate": 2.941413602694555e-05, "loss": 1.2593, "step": 4970 }, { "epoch": 0.45, "learning_rate": 2.940707981511269e-05, "loss": 1.272, "step": 4971 }, { "epoch": 0.45, "learning_rate": 2.9400023240914533e-05, "loss": 1.2541, "step": 4972 }, { "epoch": 0.45, "learning_rate": 2.939296630493128e-05, "loss": 1.2459, "step": 4973 }, { "epoch": 0.45, "learning_rate": 2.9385909007743183e-05, "loss": 1.2464, "step": 4974 }, { "epoch": 0.45, "learning_rate": 2.9378851349930514e-05, "loss": 1.2708, "step": 4975 }, { "epoch": 0.45, "learning_rate": 2.937179333207358e-05, "loss": 1.2456, "step": 4976 }, { "epoch": 0.45, "learning_rate": 2.936473495475272e-05, "loss": 1.231, "step": 4977 }, { "epoch": 0.45, "learning_rate": 2.9357676218548286e-05, "loss": 1.2043, "step": 4978 }, { "epoch": 0.45, "learning_rate": 2.9350617124040687e-05, "loss": 1.224, "step": 4979 }, { "epoch": 0.45, "learning_rate": 2.934355767181033e-05, "loss": 1.2744, "step": 4980 }, { "epoch": 0.45, "learning_rate": 2.9336497862437683e-05, "loss": 1.2444, "step": 4981 }, { "epoch": 0.45, "learning_rate": 2.932943769650322e-05, "loss": 1.2267, "step": 4982 }, { "epoch": 0.45, "learning_rate": 2.9322377174587452e-05, "loss": 1.2912, "step": 4983 }, { "epoch": 0.45, "learning_rate": 2.9315316297270923e-05, "loss": 1.277, "step": 4984 }, { "epoch": 0.45, "learning_rate": 2.9308255065134193e-05, "loss": 1.2928, "step": 4985 }, { "epoch": 0.45, "learning_rate": 2.9301193478757876e-05, "loss": 1.2375, "step": 4986 }, { "epoch": 0.45, "learning_rate": 2.929413153872259e-05, "loss": 1.2426, "step": 4987 }, { "epoch": 0.45, "learning_rate": 2.928706924560899e-05, "loss": 1.2334, "step": 4988 }, { "epoch": 0.45, "learning_rate": 2.928000659999777e-05, "loss": 1.2634, "step": 4989 }, { "epoch": 0.45, "learning_rate": 2.9272943602469643e-05, "loss": 1.2248, "step": 4990 }, { "epoch": 0.45, "learning_rate": 2.926588025360535e-05, "loss": 1.2381, "step": 4991 }, { "epoch": 0.45, "learning_rate": 2.9258816553985662e-05, "loss": 1.2537, "step": 4992 }, { "epoch": 0.45, "learning_rate": 2.9251752504191387e-05, "loss": 1.2054, "step": 4993 }, { "epoch": 0.45, "learning_rate": 2.924468810480335e-05, "loss": 1.2652, "step": 4994 }, { "epoch": 0.45, "learning_rate": 2.923762335640242e-05, "loss": 1.21, "step": 4995 }, { "epoch": 0.45, "learning_rate": 2.923055825956947e-05, "loss": 1.2543, "step": 4996 }, { "epoch": 0.45, "learning_rate": 2.922349281488543e-05, "loss": 1.2181, "step": 4997 }, { "epoch": 0.45, "learning_rate": 2.9216427022931243e-05, "loss": 1.2341, "step": 4998 }, { "epoch": 0.45, "learning_rate": 2.9209360884287868e-05, "loss": 1.2602, "step": 4999 }, { "epoch": 0.45, "learning_rate": 2.9202294399536327e-05, "loss": 1.3218, "step": 5000 }, { "epoch": 0.45, "learning_rate": 2.919522756925764e-05, "loss": 1.2681, "step": 5001 }, { "epoch": 0.45, "learning_rate": 2.918816039403287e-05, "loss": 1.2764, "step": 5002 }, { "epoch": 0.45, "learning_rate": 2.9181092874443105e-05, "loss": 1.3, "step": 5003 }, { "epoch": 0.45, "learning_rate": 2.917402501106945e-05, "loss": 1.2378, "step": 5004 }, { "epoch": 0.45, "learning_rate": 2.916695680449307e-05, "loss": 1.2542, "step": 5005 }, { "epoch": 0.45, "learning_rate": 2.9159888255295116e-05, "loss": 1.2635, "step": 5006 }, { "epoch": 0.45, "learning_rate": 2.9152819364056806e-05, "loss": 1.217, "step": 5007 }, { "epoch": 0.45, "learning_rate": 2.9145750131359355e-05, "loss": 1.2737, "step": 5008 }, { "epoch": 0.45, "learning_rate": 2.9138680557784025e-05, "loss": 1.232, "step": 5009 }, { "epoch": 0.45, "learning_rate": 2.913161064391211e-05, "loss": 1.2744, "step": 5010 }, { "epoch": 0.45, "learning_rate": 2.9124540390324906e-05, "loss": 1.2779, "step": 5011 }, { "epoch": 0.45, "learning_rate": 2.9117469797603762e-05, "loss": 1.2861, "step": 5012 }, { "epoch": 0.45, "learning_rate": 2.9110398866330048e-05, "loss": 1.2568, "step": 5013 }, { "epoch": 0.45, "learning_rate": 2.9103327597085156e-05, "loss": 1.2657, "step": 5014 }, { "epoch": 0.45, "learning_rate": 2.9096255990450518e-05, "loss": 1.2288, "step": 5015 }, { "epoch": 0.45, "learning_rate": 2.908918404700758e-05, "loss": 1.2677, "step": 5016 }, { "epoch": 0.45, "learning_rate": 2.908211176733781e-05, "loss": 1.2355, "step": 5017 }, { "epoch": 0.45, "learning_rate": 2.907503915202274e-05, "loss": 1.3011, "step": 5018 }, { "epoch": 0.45, "learning_rate": 2.906796620164389e-05, "loss": 1.2457, "step": 5019 }, { "epoch": 0.45, "learning_rate": 2.9060892916782817e-05, "loss": 1.2055, "step": 5020 }, { "epoch": 0.45, "learning_rate": 2.9053819298021127e-05, "loss": 1.2083, "step": 5021 }, { "epoch": 0.45, "learning_rate": 2.9046745345940423e-05, "loss": 1.2409, "step": 5022 }, { "epoch": 0.45, "learning_rate": 2.9039671061122358e-05, "loss": 1.2606, "step": 5023 }, { "epoch": 0.45, "learning_rate": 2.9032596444148607e-05, "loss": 1.2519, "step": 5024 }, { "epoch": 0.45, "learning_rate": 2.9025521495600854e-05, "loss": 1.2611, "step": 5025 }, { "epoch": 0.45, "learning_rate": 2.901844621606084e-05, "loss": 1.2606, "step": 5026 }, { "epoch": 0.45, "learning_rate": 2.9011370606110312e-05, "loss": 1.2005, "step": 5027 }, { "epoch": 0.45, "learning_rate": 2.9004294666331068e-05, "loss": 1.2593, "step": 5028 }, { "epoch": 0.45, "learning_rate": 2.8997218397304883e-05, "loss": 1.1969, "step": 5029 }, { "epoch": 0.45, "learning_rate": 2.899014179961361e-05, "loss": 1.2795, "step": 5030 }, { "epoch": 0.45, "learning_rate": 2.8983064873839126e-05, "loss": 1.2152, "step": 5031 }, { "epoch": 0.45, "learning_rate": 2.8975987620563306e-05, "loss": 1.2451, "step": 5032 }, { "epoch": 0.45, "learning_rate": 2.896891004036806e-05, "loss": 1.2492, "step": 5033 }, { "epoch": 0.45, "learning_rate": 2.8961832133835338e-05, "loss": 1.2247, "step": 5034 }, { "epoch": 0.45, "learning_rate": 2.895475390154711e-05, "loss": 1.2547, "step": 5035 }, { "epoch": 0.46, "learning_rate": 2.8947675344085372e-05, "loss": 1.2127, "step": 5036 }, { "epoch": 0.46, "learning_rate": 2.8940596462032154e-05, "loss": 1.2539, "step": 5037 }, { "epoch": 0.46, "learning_rate": 2.8933517255969487e-05, "loss": 1.2768, "step": 5038 }, { "epoch": 0.46, "learning_rate": 2.8926437726479477e-05, "loss": 1.1932, "step": 5039 }, { "epoch": 0.46, "learning_rate": 2.8919357874144204e-05, "loss": 1.227, "step": 5040 }, { "epoch": 0.46, "learning_rate": 2.8912277699545803e-05, "loss": 1.2648, "step": 5041 }, { "epoch": 0.46, "learning_rate": 2.890519720326643e-05, "loss": 1.294, "step": 5042 }, { "epoch": 0.46, "learning_rate": 2.889811638588827e-05, "loss": 1.2022, "step": 5043 }, { "epoch": 0.46, "learning_rate": 2.8891035247993538e-05, "loss": 1.2674, "step": 5044 }, { "epoch": 0.46, "learning_rate": 2.8883953790164464e-05, "loss": 1.2303, "step": 5045 }, { "epoch": 0.46, "learning_rate": 2.8876872012983302e-05, "loss": 1.2708, "step": 5046 }, { "epoch": 0.46, "learning_rate": 2.8869789917032352e-05, "loss": 1.206, "step": 5047 }, { "epoch": 0.46, "learning_rate": 2.8862707502893927e-05, "loss": 1.2242, "step": 5048 }, { "epoch": 0.46, "learning_rate": 2.885562477115037e-05, "loss": 1.2028, "step": 5049 }, { "epoch": 0.46, "learning_rate": 2.884854172238403e-05, "loss": 1.2445, "step": 5050 }, { "epoch": 0.46, "learning_rate": 2.8841458357177314e-05, "loss": 1.2887, "step": 5051 }, { "epoch": 0.46, "learning_rate": 2.8834374676112647e-05, "loss": 1.2335, "step": 5052 }, { "epoch": 0.46, "learning_rate": 2.8827290679772462e-05, "loss": 1.2524, "step": 5053 }, { "epoch": 0.46, "learning_rate": 2.8820206368739234e-05, "loss": 1.2484, "step": 5054 }, { "epoch": 0.46, "learning_rate": 2.881312174359545e-05, "loss": 1.2211, "step": 5055 }, { "epoch": 0.46, "learning_rate": 2.8806036804923646e-05, "loss": 1.2041, "step": 5056 }, { "epoch": 0.46, "learning_rate": 2.879895155330637e-05, "loss": 1.2124, "step": 5057 }, { "epoch": 0.46, "learning_rate": 2.879186598932619e-05, "loss": 1.2828, "step": 5058 }, { "epoch": 0.46, "learning_rate": 2.87847801135657e-05, "loss": 1.2455, "step": 5059 }, { "epoch": 0.46, "learning_rate": 2.877769392660754e-05, "loss": 1.2497, "step": 5060 }, { "epoch": 0.46, "learning_rate": 2.8770607429034352e-05, "loss": 1.2398, "step": 5061 }, { "epoch": 0.46, "learning_rate": 2.8763520621428806e-05, "loss": 1.2517, "step": 5062 }, { "epoch": 0.46, "learning_rate": 2.8756433504373614e-05, "loss": 1.2331, "step": 5063 }, { "epoch": 0.46, "learning_rate": 2.8749346078451495e-05, "loss": 1.2736, "step": 5064 }, { "epoch": 0.46, "learning_rate": 2.874225834424521e-05, "loss": 1.2587, "step": 5065 }, { "epoch": 0.46, "learning_rate": 2.8735170302337538e-05, "loss": 1.2455, "step": 5066 }, { "epoch": 0.46, "learning_rate": 2.8728081953311274e-05, "loss": 1.286, "step": 5067 }, { "epoch": 0.46, "learning_rate": 2.8720993297749248e-05, "loss": 1.2959, "step": 5068 }, { "epoch": 0.46, "learning_rate": 2.871390433623431e-05, "loss": 1.2061, "step": 5069 }, { "epoch": 0.46, "learning_rate": 2.8706815069349358e-05, "loss": 1.2696, "step": 5070 }, { "epoch": 0.46, "learning_rate": 2.8699725497677275e-05, "loss": 1.2627, "step": 5071 }, { "epoch": 0.46, "learning_rate": 2.8692635621800994e-05, "loss": 1.1882, "step": 5072 }, { "epoch": 0.46, "learning_rate": 2.868554544230348e-05, "loss": 1.2346, "step": 5073 }, { "epoch": 0.46, "learning_rate": 2.8678454959767704e-05, "loss": 1.244, "step": 5074 }, { "epoch": 0.46, "learning_rate": 2.8671364174776673e-05, "loss": 1.2856, "step": 5075 }, { "epoch": 0.46, "learning_rate": 2.8664273087913403e-05, "loss": 1.2285, "step": 5076 }, { "epoch": 0.46, "learning_rate": 2.865718169976096e-05, "loss": 1.281, "step": 5077 }, { "epoch": 0.46, "learning_rate": 2.8650090010902427e-05, "loss": 1.2844, "step": 5078 }, { "epoch": 0.46, "learning_rate": 2.8642998021920895e-05, "loss": 1.1998, "step": 5079 }, { "epoch": 0.46, "learning_rate": 2.8635905733399505e-05, "loss": 1.2113, "step": 5080 }, { "epoch": 0.46, "learning_rate": 2.8628813145921396e-05, "loss": 1.2441, "step": 5081 }, { "epoch": 0.46, "learning_rate": 2.8621720260069752e-05, "loss": 1.2051, "step": 5082 }, { "epoch": 0.46, "learning_rate": 2.8614627076427775e-05, "loss": 1.2812, "step": 5083 }, { "epoch": 0.46, "learning_rate": 2.8607533595578685e-05, "loss": 1.2505, "step": 5084 }, { "epoch": 0.46, "learning_rate": 2.8600439818105733e-05, "loss": 1.2148, "step": 5085 }, { "epoch": 0.46, "learning_rate": 2.8593345744592202e-05, "loss": 1.2364, "step": 5086 }, { "epoch": 0.46, "learning_rate": 2.8586251375621392e-05, "loss": 1.2412, "step": 5087 }, { "epoch": 0.46, "learning_rate": 2.857915671177662e-05, "loss": 1.2596, "step": 5088 }, { "epoch": 0.46, "learning_rate": 2.8572061753641233e-05, "loss": 1.1795, "step": 5089 }, { "epoch": 0.46, "learning_rate": 2.85649665017986e-05, "loss": 1.2726, "step": 5090 }, { "epoch": 0.46, "learning_rate": 2.8557870956832132e-05, "loss": 1.2535, "step": 5091 }, { "epoch": 0.46, "learning_rate": 2.8550775119325234e-05, "loss": 1.2667, "step": 5092 }, { "epoch": 0.46, "learning_rate": 2.8543678989861356e-05, "loss": 1.2622, "step": 5093 }, { "epoch": 0.46, "learning_rate": 2.8536582569023962e-05, "loss": 1.2758, "step": 5094 }, { "epoch": 0.46, "learning_rate": 2.8529485857396555e-05, "loss": 1.2255, "step": 5095 }, { "epoch": 0.46, "learning_rate": 2.8522388855562644e-05, "loss": 1.202, "step": 5096 }, { "epoch": 0.46, "learning_rate": 2.8515291564105762e-05, "loss": 1.2901, "step": 5097 }, { "epoch": 0.46, "learning_rate": 2.850819398360948e-05, "loss": 1.2434, "step": 5098 }, { "epoch": 0.46, "learning_rate": 2.8501096114657395e-05, "loss": 1.2756, "step": 5099 }, { "epoch": 0.46, "learning_rate": 2.84939979578331e-05, "loss": 1.2832, "step": 5100 }, { "epoch": 0.46, "learning_rate": 2.8486899513720245e-05, "loss": 1.2075, "step": 5101 }, { "epoch": 0.46, "learning_rate": 2.8479800782902478e-05, "loss": 1.2281, "step": 5102 }, { "epoch": 0.46, "learning_rate": 2.847270176596349e-05, "loss": 1.2988, "step": 5103 }, { "epoch": 0.46, "learning_rate": 2.846560246348698e-05, "loss": 1.245, "step": 5104 }, { "epoch": 0.46, "learning_rate": 2.8458502876056674e-05, "loss": 1.2419, "step": 5105 }, { "epoch": 0.46, "learning_rate": 2.845140300425633e-05, "loss": 1.2505, "step": 5106 }, { "epoch": 0.46, "learning_rate": 2.844430284866973e-05, "loss": 1.2363, "step": 5107 }, { "epoch": 0.46, "learning_rate": 2.8437202409880664e-05, "loss": 1.2223, "step": 5108 }, { "epoch": 0.46, "learning_rate": 2.843010168847296e-05, "loss": 1.187, "step": 5109 }, { "epoch": 0.46, "learning_rate": 2.8423000685030455e-05, "loss": 1.2648, "step": 5110 }, { "epoch": 0.46, "learning_rate": 2.8415899400137023e-05, "loss": 1.2703, "step": 5111 }, { "epoch": 0.46, "learning_rate": 2.840879783437657e-05, "loss": 1.2149, "step": 5112 }, { "epoch": 0.46, "learning_rate": 2.8401695988332987e-05, "loss": 1.2584, "step": 5113 }, { "epoch": 0.46, "learning_rate": 2.8394593862590234e-05, "loss": 1.1965, "step": 5114 }, { "epoch": 0.46, "learning_rate": 2.8387491457732252e-05, "loss": 1.2801, "step": 5115 }, { "epoch": 0.46, "learning_rate": 2.8380388774343047e-05, "loss": 1.2014, "step": 5116 }, { "epoch": 0.46, "learning_rate": 2.8373285813006604e-05, "loss": 1.2907, "step": 5117 }, { "epoch": 0.46, "learning_rate": 2.8366182574306967e-05, "loss": 1.2523, "step": 5118 }, { "epoch": 0.46, "learning_rate": 2.8359079058828187e-05, "loss": 1.2881, "step": 5119 }, { "epoch": 0.46, "learning_rate": 2.835197526715434e-05, "loss": 1.2622, "step": 5120 }, { "epoch": 0.46, "learning_rate": 2.8344871199869517e-05, "loss": 1.2528, "step": 5121 }, { "epoch": 0.46, "learning_rate": 2.8337766857557852e-05, "loss": 1.27, "step": 5122 }, { "epoch": 0.46, "learning_rate": 2.833066224080347e-05, "loss": 1.2587, "step": 5123 }, { "epoch": 0.46, "learning_rate": 2.8323557350190556e-05, "loss": 1.1966, "step": 5124 }, { "epoch": 0.46, "learning_rate": 2.831645218630329e-05, "loss": 1.2619, "step": 5125 }, { "epoch": 0.46, "learning_rate": 2.8309346749725873e-05, "loss": 1.2191, "step": 5126 }, { "epoch": 0.46, "learning_rate": 2.8302241041042565e-05, "loss": 1.2521, "step": 5127 }, { "epoch": 0.46, "learning_rate": 2.829513506083759e-05, "loss": 1.2407, "step": 5128 }, { "epoch": 0.46, "learning_rate": 2.8288028809695255e-05, "loss": 1.2143, "step": 5129 }, { "epoch": 0.46, "learning_rate": 2.8280922288199846e-05, "loss": 1.2373, "step": 5130 }, { "epoch": 0.46, "learning_rate": 2.8273815496935678e-05, "loss": 1.261, "step": 5131 }, { "epoch": 0.46, "learning_rate": 2.826670843648711e-05, "loss": 1.2575, "step": 5132 }, { "epoch": 0.46, "learning_rate": 2.8259601107438512e-05, "loss": 1.2637, "step": 5133 }, { "epoch": 0.46, "learning_rate": 2.8252493510374256e-05, "loss": 1.1957, "step": 5134 }, { "epoch": 0.46, "learning_rate": 2.8245385645878768e-05, "loss": 1.2107, "step": 5135 }, { "epoch": 0.46, "learning_rate": 2.823827751453647e-05, "loss": 1.3082, "step": 5136 }, { "epoch": 0.46, "learning_rate": 2.8231169116931828e-05, "loss": 1.227, "step": 5137 }, { "epoch": 0.46, "learning_rate": 2.8224060453649315e-05, "loss": 1.2464, "step": 5138 }, { "epoch": 0.46, "learning_rate": 2.8216951525273422e-05, "loss": 1.2174, "step": 5139 }, { "epoch": 0.46, "learning_rate": 2.8209842332388682e-05, "loss": 1.2184, "step": 5140 }, { "epoch": 0.46, "learning_rate": 2.820273287557963e-05, "loss": 1.2731, "step": 5141 }, { "epoch": 0.46, "learning_rate": 2.8195623155430828e-05, "loss": 1.2946, "step": 5142 }, { "epoch": 0.46, "learning_rate": 2.8188513172526864e-05, "loss": 1.2301, "step": 5143 }, { "epoch": 0.46, "learning_rate": 2.818140292745235e-05, "loss": 1.2489, "step": 5144 }, { "epoch": 0.46, "learning_rate": 2.8174292420791915e-05, "loss": 1.2204, "step": 5145 }, { "epoch": 0.46, "learning_rate": 2.8167181653130204e-05, "loss": 1.2502, "step": 5146 }, { "epoch": 0.47, "learning_rate": 2.8160070625051886e-05, "loss": 1.3065, "step": 5147 }, { "epoch": 0.47, "learning_rate": 2.8152959337141665e-05, "loss": 1.1991, "step": 5148 }, { "epoch": 0.47, "learning_rate": 2.814584778998424e-05, "loss": 1.2354, "step": 5149 }, { "epoch": 0.47, "learning_rate": 2.8138735984164367e-05, "loss": 1.2784, "step": 5150 }, { "epoch": 0.47, "learning_rate": 2.8131623920266792e-05, "loss": 1.2081, "step": 5151 }, { "epoch": 0.47, "learning_rate": 2.812451159887629e-05, "loss": 1.2575, "step": 5152 }, { "epoch": 0.47, "learning_rate": 2.8117399020577666e-05, "loss": 1.2906, "step": 5153 }, { "epoch": 0.47, "learning_rate": 2.8110286185955742e-05, "loss": 1.3051, "step": 5154 }, { "epoch": 0.47, "learning_rate": 2.8103173095595354e-05, "loss": 1.2386, "step": 5155 }, { "epoch": 0.47, "learning_rate": 2.809605975008137e-05, "loss": 1.2351, "step": 5156 }, { "epoch": 0.47, "learning_rate": 2.808894614999867e-05, "loss": 1.2959, "step": 5157 }, { "epoch": 0.47, "learning_rate": 2.8081832295932163e-05, "loss": 1.2311, "step": 5158 }, { "epoch": 0.47, "learning_rate": 2.8074718188466776e-05, "loss": 1.2407, "step": 5159 }, { "epoch": 0.47, "learning_rate": 2.8067603828187444e-05, "loss": 1.2554, "step": 5160 }, { "epoch": 0.47, "learning_rate": 2.806048921567915e-05, "loss": 1.2669, "step": 5161 }, { "epoch": 0.47, "learning_rate": 2.805337435152687e-05, "loss": 1.2862, "step": 5162 }, { "epoch": 0.47, "learning_rate": 2.804625923631561e-05, "loss": 1.2526, "step": 5163 }, { "epoch": 0.47, "learning_rate": 2.8039143870630413e-05, "loss": 1.2455, "step": 5164 }, { "epoch": 0.47, "learning_rate": 2.803202825505632e-05, "loss": 1.2599, "step": 5165 }, { "epoch": 0.47, "learning_rate": 2.8024912390178408e-05, "loss": 1.2128, "step": 5166 }, { "epoch": 0.47, "learning_rate": 2.801779627658176e-05, "loss": 1.2289, "step": 5167 }, { "epoch": 0.47, "learning_rate": 2.801067991485149e-05, "loss": 1.2603, "step": 5168 }, { "epoch": 0.47, "learning_rate": 2.8003563305572732e-05, "loss": 1.2695, "step": 5169 }, { "epoch": 0.47, "learning_rate": 2.7996446449330634e-05, "loss": 1.2387, "step": 5170 }, { "epoch": 0.47, "learning_rate": 2.7989329346710375e-05, "loss": 1.2332, "step": 5171 }, { "epoch": 0.47, "learning_rate": 2.7982211998297143e-05, "loss": 1.2358, "step": 5172 }, { "epoch": 0.47, "learning_rate": 2.797509440467615e-05, "loss": 1.2211, "step": 5173 }, { "epoch": 0.47, "learning_rate": 2.796797656643263e-05, "loss": 1.2954, "step": 5174 }, { "epoch": 0.47, "learning_rate": 2.7960858484151843e-05, "loss": 1.2527, "step": 5175 }, { "epoch": 0.47, "learning_rate": 2.795374015841905e-05, "loss": 1.2845, "step": 5176 }, { "epoch": 0.47, "learning_rate": 2.7946621589819557e-05, "loss": 1.2767, "step": 5177 }, { "epoch": 0.47, "learning_rate": 2.7939502778938665e-05, "loss": 1.3033, "step": 5178 }, { "epoch": 0.47, "learning_rate": 2.7932383726361717e-05, "loss": 1.304, "step": 5179 }, { "epoch": 0.47, "learning_rate": 2.7925264432674065e-05, "loss": 1.2447, "step": 5180 }, { "epoch": 0.47, "learning_rate": 2.7918144898461074e-05, "loss": 1.2829, "step": 5181 }, { "epoch": 0.47, "learning_rate": 2.7911025124308142e-05, "loss": 1.2897, "step": 5182 }, { "epoch": 0.47, "learning_rate": 2.7903905110800687e-05, "loss": 1.2343, "step": 5183 }, { "epoch": 0.47, "learning_rate": 2.7896784858524128e-05, "loss": 1.2324, "step": 5184 }, { "epoch": 0.47, "learning_rate": 2.7889664368063927e-05, "loss": 1.204, "step": 5185 }, { "epoch": 0.47, "learning_rate": 2.788254364000555e-05, "loss": 1.2173, "step": 5186 }, { "epoch": 0.47, "learning_rate": 2.787542267493449e-05, "loss": 1.2278, "step": 5187 }, { "epoch": 0.47, "learning_rate": 2.7868301473436264e-05, "loss": 1.2567, "step": 5188 }, { "epoch": 0.47, "learning_rate": 2.786118003609639e-05, "loss": 1.2694, "step": 5189 }, { "epoch": 0.47, "learning_rate": 2.785405836350043e-05, "loss": 1.2717, "step": 5190 }, { "epoch": 0.47, "learning_rate": 2.7846936456233936e-05, "loss": 1.2534, "step": 5191 }, { "epoch": 0.47, "learning_rate": 2.783981431488251e-05, "loss": 1.2521, "step": 5192 }, { "epoch": 0.47, "learning_rate": 2.783269194003175e-05, "loss": 1.237, "step": 5193 }, { "epoch": 0.47, "learning_rate": 2.7825569332267283e-05, "loss": 1.2705, "step": 5194 }, { "epoch": 0.47, "learning_rate": 2.7818446492174766e-05, "loss": 1.2016, "step": 5195 }, { "epoch": 0.47, "learning_rate": 2.781132342033985e-05, "loss": 1.222, "step": 5196 }, { "epoch": 0.47, "learning_rate": 2.7804200117348223e-05, "loss": 1.21, "step": 5197 }, { "epoch": 0.47, "learning_rate": 2.7797076583785586e-05, "loss": 1.2715, "step": 5198 }, { "epoch": 0.47, "learning_rate": 2.7789952820237665e-05, "loss": 1.1892, "step": 5199 }, { "epoch": 0.47, "learning_rate": 2.7782828827290196e-05, "loss": 1.2426, "step": 5200 }, { "epoch": 0.47, "learning_rate": 2.7775704605528946e-05, "loss": 1.2497, "step": 5201 }, { "epoch": 0.47, "learning_rate": 2.776858015553968e-05, "loss": 1.2842, "step": 5202 }, { "epoch": 0.47, "learning_rate": 2.7761455477908205e-05, "loss": 1.1876, "step": 5203 }, { "epoch": 0.47, "learning_rate": 2.775433057322034e-05, "loss": 1.2273, "step": 5204 }, { "epoch": 0.47, "learning_rate": 2.774720544206191e-05, "loss": 1.2727, "step": 5205 }, { "epoch": 0.47, "learning_rate": 2.774008008501876e-05, "loss": 1.2525, "step": 5206 }, { "epoch": 0.47, "learning_rate": 2.773295450267678e-05, "loss": 1.2552, "step": 5207 }, { "epoch": 0.47, "learning_rate": 2.7725828695621852e-05, "loss": 1.2662, "step": 5208 }, { "epoch": 0.47, "learning_rate": 2.7718702664439883e-05, "loss": 1.211, "step": 5209 }, { "epoch": 0.47, "learning_rate": 2.77115764097168e-05, "loss": 1.2585, "step": 5210 }, { "epoch": 0.47, "learning_rate": 2.7704449932038557e-05, "loss": 1.264, "step": 5211 }, { "epoch": 0.47, "learning_rate": 2.7697323231991102e-05, "loss": 1.184, "step": 5212 }, { "epoch": 0.47, "learning_rate": 2.7690196310160433e-05, "loss": 1.268, "step": 5213 }, { "epoch": 0.47, "learning_rate": 2.7683069167132538e-05, "loss": 1.2814, "step": 5214 }, { "epoch": 0.47, "learning_rate": 2.7675941803493434e-05, "loss": 1.2314, "step": 5215 }, { "epoch": 0.47, "learning_rate": 2.766881421982917e-05, "loss": 1.2053, "step": 5216 }, { "epoch": 0.47, "learning_rate": 2.7661686416725795e-05, "loss": 1.3016, "step": 5217 }, { "epoch": 0.47, "learning_rate": 2.765455839476938e-05, "loss": 1.2799, "step": 5218 }, { "epoch": 0.47, "learning_rate": 2.7647430154546007e-05, "loss": 1.3049, "step": 5219 }, { "epoch": 0.47, "learning_rate": 2.7640301696641797e-05, "loss": 1.2902, "step": 5220 }, { "epoch": 0.47, "learning_rate": 2.763317302164288e-05, "loss": 1.2242, "step": 5221 }, { "epoch": 0.47, "learning_rate": 2.7626044130135387e-05, "loss": 1.2809, "step": 5222 }, { "epoch": 0.47, "learning_rate": 2.7618915022705482e-05, "loss": 1.2749, "step": 5223 }, { "epoch": 0.47, "learning_rate": 2.761178569993935e-05, "loss": 1.2473, "step": 5224 }, { "epoch": 0.47, "learning_rate": 2.7604656162423192e-05, "loss": 1.2447, "step": 5225 }, { "epoch": 0.47, "learning_rate": 2.759752641074322e-05, "loss": 1.2155, "step": 5226 }, { "epoch": 0.47, "learning_rate": 2.7590396445485657e-05, "loss": 1.2739, "step": 5227 }, { "epoch": 0.47, "learning_rate": 2.758326626723676e-05, "loss": 1.2023, "step": 5228 }, { "epoch": 0.47, "learning_rate": 2.757613587658281e-05, "loss": 1.1792, "step": 5229 }, { "epoch": 0.47, "learning_rate": 2.756900527411007e-05, "loss": 1.2069, "step": 5230 }, { "epoch": 0.47, "learning_rate": 2.7561874460404857e-05, "loss": 1.2027, "step": 5231 }, { "epoch": 0.47, "learning_rate": 2.7554743436053486e-05, "loss": 1.2767, "step": 5232 }, { "epoch": 0.47, "learning_rate": 2.75476122016423e-05, "loss": 1.2151, "step": 5233 }, { "epoch": 0.47, "learning_rate": 2.7540480757757654e-05, "loss": 1.2003, "step": 5234 }, { "epoch": 0.47, "learning_rate": 2.7533349104985905e-05, "loss": 1.2482, "step": 5235 }, { "epoch": 0.47, "learning_rate": 2.7526217243913455e-05, "loss": 1.2664, "step": 5236 }, { "epoch": 0.47, "learning_rate": 2.751908517512671e-05, "loss": 1.2567, "step": 5237 }, { "epoch": 0.47, "learning_rate": 2.751195289921209e-05, "loss": 1.2343, "step": 5238 }, { "epoch": 0.47, "learning_rate": 2.750482041675604e-05, "loss": 1.2977, "step": 5239 }, { "epoch": 0.47, "learning_rate": 2.7497687728345007e-05, "loss": 1.197, "step": 5240 }, { "epoch": 0.47, "learning_rate": 2.7490554834565467e-05, "loss": 1.2721, "step": 5241 }, { "epoch": 0.47, "learning_rate": 2.748342173600393e-05, "loss": 1.2023, "step": 5242 }, { "epoch": 0.47, "learning_rate": 2.7476288433246887e-05, "loss": 1.2817, "step": 5243 }, { "epoch": 0.47, "learning_rate": 2.7469154926880858e-05, "loss": 1.2481, "step": 5244 }, { "epoch": 0.47, "learning_rate": 2.7462021217492402e-05, "loss": 1.2021, "step": 5245 }, { "epoch": 0.47, "learning_rate": 2.7454887305668063e-05, "loss": 1.2583, "step": 5246 }, { "epoch": 0.47, "learning_rate": 2.7447753191994424e-05, "loss": 1.1773, "step": 5247 }, { "epoch": 0.47, "learning_rate": 2.7440618877058065e-05, "loss": 1.2537, "step": 5248 }, { "epoch": 0.47, "learning_rate": 2.7433484361445605e-05, "loss": 1.2542, "step": 5249 }, { "epoch": 0.47, "learning_rate": 2.742634964574367e-05, "loss": 1.2759, "step": 5250 }, { "epoch": 0.47, "learning_rate": 2.7419214730538895e-05, "loss": 1.204, "step": 5251 }, { "epoch": 0.47, "learning_rate": 2.7412079616417935e-05, "loss": 1.315, "step": 5252 }, { "epoch": 0.47, "learning_rate": 2.7404944303967466e-05, "loss": 1.2898, "step": 5253 }, { "epoch": 0.47, "learning_rate": 2.739780879377418e-05, "loss": 1.3042, "step": 5254 }, { "epoch": 0.47, "learning_rate": 2.739067308642479e-05, "loss": 1.2609, "step": 5255 }, { "epoch": 0.47, "learning_rate": 2.7383537182506002e-05, "loss": 1.2494, "step": 5256 }, { "epoch": 0.47, "learning_rate": 2.7376401082604564e-05, "loss": 1.2571, "step": 5257 }, { "epoch": 0.48, "learning_rate": 2.7369264787307235e-05, "loss": 1.2159, "step": 5258 }, { "epoch": 0.48, "learning_rate": 2.7362128297200785e-05, "loss": 1.2405, "step": 5259 }, { "epoch": 0.48, "learning_rate": 2.7354991612871994e-05, "loss": 1.2608, "step": 5260 }, { "epoch": 0.48, "learning_rate": 2.7347854734907663e-05, "loss": 1.2244, "step": 5261 }, { "epoch": 0.48, "learning_rate": 2.7340717663894616e-05, "loss": 1.2381, "step": 5262 }, { "epoch": 0.48, "learning_rate": 2.7333580400419688e-05, "loss": 1.2637, "step": 5263 }, { "epoch": 0.48, "learning_rate": 2.7326442945069735e-05, "loss": 1.2477, "step": 5264 }, { "epoch": 0.48, "learning_rate": 2.7319305298431613e-05, "loss": 1.2461, "step": 5265 }, { "epoch": 0.48, "learning_rate": 2.731216746109221e-05, "loss": 1.2414, "step": 5266 }, { "epoch": 0.48, "learning_rate": 2.730502943363842e-05, "loss": 1.2478, "step": 5267 }, { "epoch": 0.48, "learning_rate": 2.7297891216657157e-05, "loss": 1.2356, "step": 5268 }, { "epoch": 0.48, "learning_rate": 2.7290752810735344e-05, "loss": 1.2389, "step": 5269 }, { "epoch": 0.48, "learning_rate": 2.7283614216459936e-05, "loss": 1.2339, "step": 5270 }, { "epoch": 0.48, "learning_rate": 2.7276475434417893e-05, "loss": 1.2461, "step": 5271 }, { "epoch": 0.48, "learning_rate": 2.7269336465196182e-05, "loss": 1.1942, "step": 5272 }, { "epoch": 0.48, "learning_rate": 2.7262197309381803e-05, "loss": 1.1998, "step": 5273 }, { "epoch": 0.48, "learning_rate": 2.725505796756175e-05, "loss": 1.2989, "step": 5274 }, { "epoch": 0.48, "learning_rate": 2.724791844032305e-05, "loss": 1.2466, "step": 5275 }, { "epoch": 0.48, "learning_rate": 2.7240778728252746e-05, "loss": 1.2726, "step": 5276 }, { "epoch": 0.48, "learning_rate": 2.723363883193788e-05, "loss": 1.2956, "step": 5277 }, { "epoch": 0.48, "learning_rate": 2.722649875196553e-05, "loss": 1.2356, "step": 5278 }, { "epoch": 0.48, "learning_rate": 2.7219358488922763e-05, "loss": 1.2924, "step": 5279 }, { "epoch": 0.48, "learning_rate": 2.7212218043396692e-05, "loss": 1.208, "step": 5280 }, { "epoch": 0.48, "learning_rate": 2.7205077415974416e-05, "loss": 1.2641, "step": 5281 }, { "epoch": 0.48, "learning_rate": 2.7197936607243068e-05, "loss": 1.2579, "step": 5282 }, { "epoch": 0.48, "learning_rate": 2.719079561778979e-05, "loss": 1.2437, "step": 5283 }, { "epoch": 0.48, "learning_rate": 2.7183654448201734e-05, "loss": 1.2334, "step": 5284 }, { "epoch": 0.48, "learning_rate": 2.7176513099066085e-05, "loss": 1.1932, "step": 5285 }, { "epoch": 0.48, "learning_rate": 2.7169371570970014e-05, "loss": 1.267, "step": 5286 }, { "epoch": 0.48, "learning_rate": 2.7162229864500727e-05, "loss": 1.2537, "step": 5287 }, { "epoch": 0.48, "learning_rate": 2.7155087980245448e-05, "loss": 1.2695, "step": 5288 }, { "epoch": 0.48, "learning_rate": 2.7147945918791394e-05, "loss": 1.2495, "step": 5289 }, { "epoch": 0.48, "learning_rate": 2.714080368072581e-05, "loss": 1.2867, "step": 5290 }, { "epoch": 0.48, "learning_rate": 2.7133661266635978e-05, "loss": 1.2015, "step": 5291 }, { "epoch": 0.48, "learning_rate": 2.712651867710914e-05, "loss": 1.2272, "step": 5292 }, { "epoch": 0.48, "learning_rate": 2.711937591273261e-05, "loss": 1.2629, "step": 5293 }, { "epoch": 0.48, "learning_rate": 2.7112232974093675e-05, "loss": 1.2494, "step": 5294 }, { "epoch": 0.48, "learning_rate": 2.7105089861779658e-05, "loss": 1.2463, "step": 5295 }, { "epoch": 0.48, "learning_rate": 2.7097946576377893e-05, "loss": 1.2655, "step": 5296 }, { "epoch": 0.48, "learning_rate": 2.7090803118475728e-05, "loss": 1.2499, "step": 5297 }, { "epoch": 0.48, "learning_rate": 2.7083659488660508e-05, "loss": 1.2489, "step": 5298 }, { "epoch": 0.48, "learning_rate": 2.707651568751962e-05, "loss": 1.2218, "step": 5299 }, { "epoch": 0.48, "learning_rate": 2.7069371715640447e-05, "loss": 1.2267, "step": 5300 }, { "epoch": 0.48, "learning_rate": 2.7062227573610387e-05, "loss": 1.265, "step": 5301 }, { "epoch": 0.48, "learning_rate": 2.7055083262016874e-05, "loss": 1.2836, "step": 5302 }, { "epoch": 0.48, "learning_rate": 2.7047938781447315e-05, "loss": 1.2232, "step": 5303 }, { "epoch": 0.48, "learning_rate": 2.7040794132489167e-05, "loss": 1.2913, "step": 5304 }, { "epoch": 0.48, "learning_rate": 2.703364931572988e-05, "loss": 1.2151, "step": 5305 }, { "epoch": 0.48, "learning_rate": 2.7026504331756934e-05, "loss": 1.2638, "step": 5306 }, { "epoch": 0.48, "learning_rate": 2.7019359181157815e-05, "loss": 1.2444, "step": 5307 }, { "epoch": 0.48, "learning_rate": 2.7012213864520015e-05, "loss": 1.2355, "step": 5308 }, { "epoch": 0.48, "learning_rate": 2.7005068382431048e-05, "loss": 1.2474, "step": 5309 }, { "epoch": 0.48, "learning_rate": 2.6997922735478442e-05, "loss": 1.2681, "step": 5310 }, { "epoch": 0.48, "learning_rate": 2.6990776924249728e-05, "loss": 1.2407, "step": 5311 }, { "epoch": 0.48, "learning_rate": 2.6983630949332476e-05, "loss": 1.2609, "step": 5312 }, { "epoch": 0.48, "learning_rate": 2.697648481131424e-05, "loss": 1.2408, "step": 5313 }, { "epoch": 0.48, "learning_rate": 2.696933851078261e-05, "loss": 1.2821, "step": 5314 }, { "epoch": 0.48, "learning_rate": 2.696219204832517e-05, "loss": 1.2789, "step": 5315 }, { "epoch": 0.48, "learning_rate": 2.6955045424529523e-05, "loss": 1.254, "step": 5316 }, { "epoch": 0.48, "learning_rate": 2.6947898639983303e-05, "loss": 1.2271, "step": 5317 }, { "epoch": 0.48, "learning_rate": 2.694075169527414e-05, "loss": 1.2258, "step": 5318 }, { "epoch": 0.48, "learning_rate": 2.693360459098967e-05, "loss": 1.2889, "step": 5319 }, { "epoch": 0.48, "learning_rate": 2.692645732771757e-05, "loss": 1.2564, "step": 5320 }, { "epoch": 0.48, "learning_rate": 2.6919309906045488e-05, "loss": 1.2254, "step": 5321 }, { "epoch": 0.48, "learning_rate": 2.691216232656113e-05, "loss": 1.2817, "step": 5322 }, { "epoch": 0.48, "learning_rate": 2.6905014589852195e-05, "loss": 1.2094, "step": 5323 }, { "epoch": 0.48, "learning_rate": 2.689786669650638e-05, "loss": 1.2506, "step": 5324 }, { "epoch": 0.48, "learning_rate": 2.6890718647111422e-05, "loss": 1.2591, "step": 5325 }, { "epoch": 0.48, "learning_rate": 2.6883570442255047e-05, "loss": 1.2556, "step": 5326 }, { "epoch": 0.48, "learning_rate": 2.687642208252502e-05, "loss": 1.218, "step": 5327 }, { "epoch": 0.48, "learning_rate": 2.6869273568509097e-05, "loss": 1.2437, "step": 5328 }, { "epoch": 0.48, "learning_rate": 2.6862124900795044e-05, "loss": 1.2188, "step": 5329 }, { "epoch": 0.48, "learning_rate": 2.685497607997066e-05, "loss": 1.2898, "step": 5330 }, { "epoch": 0.48, "learning_rate": 2.6847827106623746e-05, "loss": 1.2715, "step": 5331 }, { "epoch": 0.48, "learning_rate": 2.684067798134211e-05, "loss": 1.2999, "step": 5332 }, { "epoch": 0.48, "learning_rate": 2.6833528704713584e-05, "loss": 1.1892, "step": 5333 }, { "epoch": 0.48, "learning_rate": 2.6826379277325997e-05, "loss": 1.2341, "step": 5334 }, { "epoch": 0.48, "learning_rate": 2.6819229699767208e-05, "loss": 1.2553, "step": 5335 }, { "epoch": 0.48, "learning_rate": 2.6812079972625077e-05, "loss": 1.1953, "step": 5336 }, { "epoch": 0.48, "learning_rate": 2.6804930096487475e-05, "loss": 1.2037, "step": 5337 }, { "epoch": 0.48, "learning_rate": 2.6797780071942298e-05, "loss": 1.2273, "step": 5338 }, { "epoch": 0.48, "learning_rate": 2.679062989957744e-05, "loss": 1.25, "step": 5339 }, { "epoch": 0.48, "learning_rate": 2.6783479579980807e-05, "loss": 1.2062, "step": 5340 }, { "epoch": 0.48, "learning_rate": 2.677632911374034e-05, "loss": 1.251, "step": 5341 }, { "epoch": 0.48, "learning_rate": 2.676917850144396e-05, "loss": 1.2037, "step": 5342 }, { "epoch": 0.48, "learning_rate": 2.6762027743679623e-05, "loss": 1.3314, "step": 5343 }, { "epoch": 0.48, "learning_rate": 2.6754876841035288e-05, "loss": 1.2625, "step": 5344 }, { "epoch": 0.48, "learning_rate": 2.6747725794098917e-05, "loss": 1.2812, "step": 5345 }, { "epoch": 0.48, "learning_rate": 2.6740574603458508e-05, "loss": 1.3148, "step": 5346 }, { "epoch": 0.48, "learning_rate": 2.673342326970204e-05, "loss": 1.2269, "step": 5347 }, { "epoch": 0.48, "learning_rate": 2.6726271793417546e-05, "loss": 1.2449, "step": 5348 }, { "epoch": 0.48, "learning_rate": 2.6719120175193018e-05, "loss": 1.2829, "step": 5349 }, { "epoch": 0.48, "learning_rate": 2.67119684156165e-05, "loss": 1.2103, "step": 5350 }, { "epoch": 0.48, "learning_rate": 2.6704816515276037e-05, "loss": 1.2257, "step": 5351 }, { "epoch": 0.48, "learning_rate": 2.6697664474759682e-05, "loss": 1.2409, "step": 5352 }, { "epoch": 0.48, "learning_rate": 2.6690512294655485e-05, "loss": 1.2462, "step": 5353 }, { "epoch": 0.48, "learning_rate": 2.6683359975551542e-05, "loss": 1.2552, "step": 5354 }, { "epoch": 0.48, "learning_rate": 2.6676207518035928e-05, "loss": 1.2065, "step": 5355 }, { "epoch": 0.48, "learning_rate": 2.6669054922696763e-05, "loss": 1.2438, "step": 5356 }, { "epoch": 0.48, "learning_rate": 2.6661902190122134e-05, "loss": 1.2726, "step": 5357 }, { "epoch": 0.48, "learning_rate": 2.6654749320900168e-05, "loss": 1.2188, "step": 5358 }, { "epoch": 0.48, "learning_rate": 2.6647596315619016e-05, "loss": 1.2376, "step": 5359 }, { "epoch": 0.48, "learning_rate": 2.664044317486681e-05, "loss": 1.2605, "step": 5360 }, { "epoch": 0.48, "learning_rate": 2.6633289899231696e-05, "loss": 1.2646, "step": 5361 }, { "epoch": 0.48, "learning_rate": 2.6626136489301858e-05, "loss": 1.2653, "step": 5362 }, { "epoch": 0.48, "learning_rate": 2.6618982945665466e-05, "loss": 1.271, "step": 5363 }, { "epoch": 0.48, "learning_rate": 2.6611829268910715e-05, "loss": 1.2795, "step": 5364 }, { "epoch": 0.48, "learning_rate": 2.66046754596258e-05, "loss": 1.2003, "step": 5365 }, { "epoch": 0.48, "learning_rate": 2.6597521518398927e-05, "loss": 1.2782, "step": 5366 }, { "epoch": 0.48, "learning_rate": 2.6590367445818324e-05, "loss": 1.2721, "step": 5367 }, { "epoch": 0.49, "learning_rate": 2.6583213242472226e-05, "loss": 1.2538, "step": 5368 }, { "epoch": 0.49, "learning_rate": 2.6576058908948877e-05, "loss": 1.2582, "step": 5369 }, { "epoch": 0.49, "learning_rate": 2.6568904445836524e-05, "loss": 1.263, "step": 5370 }, { "epoch": 0.49, "learning_rate": 2.6561749853723428e-05, "loss": 1.2517, "step": 5371 }, { "epoch": 0.49, "learning_rate": 2.655459513319788e-05, "loss": 1.2602, "step": 5372 }, { "epoch": 0.49, "learning_rate": 2.6547440284848153e-05, "loss": 1.2497, "step": 5373 }, { "epoch": 0.49, "learning_rate": 2.654028530926254e-05, "loss": 1.2107, "step": 5374 }, { "epoch": 0.49, "learning_rate": 2.6533130207029368e-05, "loss": 1.2585, "step": 5375 }, { "epoch": 0.49, "learning_rate": 2.6525974978736933e-05, "loss": 1.2925, "step": 5376 }, { "epoch": 0.49, "learning_rate": 2.651881962497358e-05, "loss": 1.2416, "step": 5377 }, { "epoch": 0.49, "learning_rate": 2.651166414632763e-05, "loss": 1.2941, "step": 5378 }, { "epoch": 0.49, "learning_rate": 2.6504508543387437e-05, "loss": 1.2261, "step": 5379 }, { "epoch": 0.49, "learning_rate": 2.6497352816741362e-05, "loss": 1.2572, "step": 5380 }, { "epoch": 0.49, "learning_rate": 2.6490196966977774e-05, "loss": 1.2499, "step": 5381 }, { "epoch": 0.49, "learning_rate": 2.648304099468505e-05, "loss": 1.2535, "step": 5382 }, { "epoch": 0.49, "learning_rate": 2.647588490045158e-05, "loss": 1.2837, "step": 5383 }, { "epoch": 0.49, "learning_rate": 2.6468728684865758e-05, "loss": 1.2138, "step": 5384 }, { "epoch": 0.49, "learning_rate": 2.6461572348516e-05, "loss": 1.289, "step": 5385 }, { "epoch": 0.49, "learning_rate": 2.6454415891990725e-05, "loss": 1.2614, "step": 5386 }, { "epoch": 0.49, "learning_rate": 2.6447259315878347e-05, "loss": 1.2184, "step": 5387 }, { "epoch": 0.49, "learning_rate": 2.6440102620767326e-05, "loss": 1.2371, "step": 5388 }, { "epoch": 0.49, "learning_rate": 2.6432945807246094e-05, "loss": 1.2212, "step": 5389 }, { "epoch": 0.49, "learning_rate": 2.6425788875903117e-05, "loss": 1.2797, "step": 5390 }, { "epoch": 0.49, "learning_rate": 2.6418631827326857e-05, "loss": 1.2279, "step": 5391 }, { "epoch": 0.49, "learning_rate": 2.6411474662105785e-05, "loss": 1.2683, "step": 5392 }, { "epoch": 0.49, "learning_rate": 2.640431738082841e-05, "loss": 1.1979, "step": 5393 }, { "epoch": 0.49, "learning_rate": 2.6397159984083213e-05, "loss": 1.2606, "step": 5394 }, { "epoch": 0.49, "learning_rate": 2.6390002472458697e-05, "loss": 1.3018, "step": 5395 }, { "epoch": 0.49, "learning_rate": 2.6382844846543387e-05, "loss": 1.2196, "step": 5396 }, { "epoch": 0.49, "learning_rate": 2.6375687106925794e-05, "loss": 1.2257, "step": 5397 }, { "epoch": 0.49, "learning_rate": 2.636852925419448e-05, "loss": 1.2289, "step": 5398 }, { "epoch": 0.49, "learning_rate": 2.6361371288937953e-05, "loss": 1.2318, "step": 5399 }, { "epoch": 0.49, "learning_rate": 2.6354213211744784e-05, "loss": 1.2448, "step": 5400 }, { "epoch": 0.49, "learning_rate": 2.6347055023203533e-05, "loss": 1.2462, "step": 5401 }, { "epoch": 0.49, "learning_rate": 2.6339896723902774e-05, "loss": 1.2169, "step": 5402 }, { "epoch": 0.49, "learning_rate": 2.633273831443108e-05, "loss": 1.2114, "step": 5403 }, { "epoch": 0.49, "learning_rate": 2.6325579795377046e-05, "loss": 1.2601, "step": 5404 }, { "epoch": 0.49, "learning_rate": 2.6318421167329264e-05, "loss": 1.205, "step": 5405 }, { "epoch": 0.49, "learning_rate": 2.6311262430876356e-05, "loss": 1.253, "step": 5406 }, { "epoch": 0.49, "learning_rate": 2.630410358660692e-05, "loss": 1.276, "step": 5407 }, { "epoch": 0.49, "learning_rate": 2.6296944635109582e-05, "loss": 1.232, "step": 5408 }, { "epoch": 0.49, "learning_rate": 2.628978557697299e-05, "loss": 1.2697, "step": 5409 }, { "epoch": 0.49, "learning_rate": 2.628262641278578e-05, "loss": 1.2379, "step": 5410 }, { "epoch": 0.49, "learning_rate": 2.6275467143136594e-05, "loss": 1.2781, "step": 5411 }, { "epoch": 0.49, "learning_rate": 2.6268307768614105e-05, "loss": 1.2032, "step": 5412 }, { "epoch": 0.49, "learning_rate": 2.626114828980697e-05, "loss": 1.2002, "step": 5413 }, { "epoch": 0.49, "learning_rate": 2.6253988707303872e-05, "loss": 1.2518, "step": 5414 }, { "epoch": 0.49, "learning_rate": 2.6246829021693503e-05, "loss": 1.2651, "step": 5415 }, { "epoch": 0.49, "learning_rate": 2.6239669233564545e-05, "loss": 1.2338, "step": 5416 }, { "epoch": 0.49, "learning_rate": 2.6232509343505705e-05, "loss": 1.2795, "step": 5417 }, { "epoch": 0.49, "learning_rate": 2.6225349352105693e-05, "loss": 1.2686, "step": 5418 }, { "epoch": 0.49, "learning_rate": 2.621818925995324e-05, "loss": 1.2725, "step": 5419 }, { "epoch": 0.49, "learning_rate": 2.6211029067637054e-05, "loss": 1.2126, "step": 5420 }, { "epoch": 0.49, "learning_rate": 2.6203868775745883e-05, "loss": 1.2219, "step": 5421 }, { "epoch": 0.49, "learning_rate": 2.619670838486847e-05, "loss": 1.2147, "step": 5422 }, { "epoch": 0.49, "learning_rate": 2.6189547895593562e-05, "loss": 1.2284, "step": 5423 }, { "epoch": 0.49, "learning_rate": 2.6182387308509924e-05, "loss": 1.241, "step": 5424 }, { "epoch": 0.49, "learning_rate": 2.617522662420632e-05, "loss": 1.2204, "step": 5425 }, { "epoch": 0.49, "learning_rate": 2.6168065843271523e-05, "loss": 1.2602, "step": 5426 }, { "epoch": 0.49, "learning_rate": 2.6160904966294332e-05, "loss": 1.2772, "step": 5427 }, { "epoch": 0.49, "learning_rate": 2.6153743993863528e-05, "loss": 1.2361, "step": 5428 }, { "epoch": 0.49, "learning_rate": 2.6146582926567913e-05, "loss": 1.2459, "step": 5429 }, { "epoch": 0.49, "learning_rate": 2.6139421764996285e-05, "loss": 1.2366, "step": 5430 }, { "epoch": 0.49, "learning_rate": 2.6132260509737483e-05, "loss": 1.1965, "step": 5431 }, { "epoch": 0.49, "learning_rate": 2.6125099161380307e-05, "loss": 1.2776, "step": 5432 }, { "epoch": 0.49, "learning_rate": 2.61179377205136e-05, "loss": 1.2451, "step": 5433 }, { "epoch": 0.49, "learning_rate": 2.611077618772619e-05, "loss": 1.2443, "step": 5434 }, { "epoch": 0.49, "learning_rate": 2.610361456360693e-05, "loss": 1.2605, "step": 5435 }, { "epoch": 0.49, "learning_rate": 2.6096452848744677e-05, "loss": 1.2231, "step": 5436 }, { "epoch": 0.49, "learning_rate": 2.6089291043728293e-05, "loss": 1.2348, "step": 5437 }, { "epoch": 0.49, "learning_rate": 2.6082129149146634e-05, "loss": 1.1932, "step": 5438 }, { "epoch": 0.49, "learning_rate": 2.607496716558858e-05, "loss": 1.1926, "step": 5439 }, { "epoch": 0.49, "learning_rate": 2.606780509364303e-05, "loss": 1.1935, "step": 5440 }, { "epoch": 0.49, "learning_rate": 2.6060642933898853e-05, "loss": 1.2233, "step": 5441 }, { "epoch": 0.49, "learning_rate": 2.605348068694496e-05, "loss": 1.1993, "step": 5442 }, { "epoch": 0.49, "learning_rate": 2.604631835337024e-05, "loss": 1.2107, "step": 5443 }, { "epoch": 0.49, "learning_rate": 2.6039155933763622e-05, "loss": 1.1623, "step": 5444 }, { "epoch": 0.49, "learning_rate": 2.603199342871402e-05, "loss": 1.2387, "step": 5445 }, { "epoch": 0.49, "learning_rate": 2.602483083881035e-05, "loss": 1.2431, "step": 5446 }, { "epoch": 0.49, "learning_rate": 2.6017668164641557e-05, "loss": 1.2412, "step": 5447 }, { "epoch": 0.49, "learning_rate": 2.6010505406796577e-05, "loss": 1.2477, "step": 5448 }, { "epoch": 0.49, "learning_rate": 2.600334256586436e-05, "loss": 1.263, "step": 5449 }, { "epoch": 0.49, "learning_rate": 2.5996179642433856e-05, "loss": 1.2706, "step": 5450 }, { "epoch": 0.49, "learning_rate": 2.5989016637094017e-05, "loss": 1.2544, "step": 5451 }, { "epoch": 0.49, "learning_rate": 2.598185355043382e-05, "loss": 1.2152, "step": 5452 }, { "epoch": 0.49, "learning_rate": 2.5974690383042244e-05, "loss": 1.2338, "step": 5453 }, { "epoch": 0.49, "learning_rate": 2.5967527135508252e-05, "loss": 1.2303, "step": 5454 }, { "epoch": 0.49, "learning_rate": 2.596036380842085e-05, "loss": 1.2143, "step": 5455 }, { "epoch": 0.49, "learning_rate": 2.5953200402369016e-05, "loss": 1.1891, "step": 5456 }, { "epoch": 0.49, "learning_rate": 2.5946036917941762e-05, "loss": 1.2462, "step": 5457 }, { "epoch": 0.49, "learning_rate": 2.5938873355728087e-05, "loss": 1.2496, "step": 5458 }, { "epoch": 0.49, "learning_rate": 2.5931709716317e-05, "loss": 1.2446, "step": 5459 }, { "epoch": 0.49, "learning_rate": 2.5924546000297527e-05, "loss": 1.2262, "step": 5460 }, { "epoch": 0.49, "learning_rate": 2.59173822082587e-05, "loss": 1.2802, "step": 5461 }, { "epoch": 0.49, "learning_rate": 2.5910218340789538e-05, "loss": 1.2282, "step": 5462 }, { "epoch": 0.49, "learning_rate": 2.5903054398479088e-05, "loss": 1.2228, "step": 5463 }, { "epoch": 0.49, "learning_rate": 2.5895890381916378e-05, "loss": 1.2667, "step": 5464 }, { "epoch": 0.49, "learning_rate": 2.588872629169048e-05, "loss": 1.2188, "step": 5465 }, { "epoch": 0.49, "learning_rate": 2.588156212839044e-05, "loss": 1.2828, "step": 5466 }, { "epoch": 0.49, "learning_rate": 2.5874397892605318e-05, "loss": 1.2425, "step": 5467 }, { "epoch": 0.49, "learning_rate": 2.5867233584924184e-05, "loss": 1.2437, "step": 5468 }, { "epoch": 0.49, "learning_rate": 2.586006920593611e-05, "loss": 1.2402, "step": 5469 }, { "epoch": 0.49, "learning_rate": 2.5852904756230184e-05, "loss": 1.2695, "step": 5470 }, { "epoch": 0.49, "learning_rate": 2.584574023639549e-05, "loss": 1.2454, "step": 5471 }, { "epoch": 0.49, "learning_rate": 2.5838575647021108e-05, "loss": 1.2557, "step": 5472 }, { "epoch": 0.49, "learning_rate": 2.5831410988696147e-05, "loss": 1.254, "step": 5473 }, { "epoch": 0.49, "learning_rate": 2.5824246262009705e-05, "loss": 1.2699, "step": 5474 }, { "epoch": 0.49, "learning_rate": 2.5817081467550886e-05, "loss": 1.2544, "step": 5475 }, { "epoch": 0.49, "learning_rate": 2.580991660590882e-05, "loss": 1.2169, "step": 5476 }, { "epoch": 0.49, "learning_rate": 2.5802751677672608e-05, "loss": 1.2934, "step": 5477 }, { "epoch": 0.49, "learning_rate": 2.579558668343139e-05, "loss": 1.2557, "step": 5478 }, { "epoch": 0.5, "learning_rate": 2.578842162377429e-05, "loss": 1.2669, "step": 5479 }, { "epoch": 0.5, "learning_rate": 2.578125649929044e-05, "loss": 1.2625, "step": 5480 }, { "epoch": 0.5, "learning_rate": 2.5774091310568987e-05, "loss": 1.2645, "step": 5481 }, { "epoch": 0.5, "learning_rate": 2.5766926058199075e-05, "loss": 1.2078, "step": 5482 }, { "epoch": 0.5, "learning_rate": 2.5759760742769857e-05, "loss": 1.235, "step": 5483 }, { "epoch": 0.5, "learning_rate": 2.5752595364870492e-05, "loss": 1.2413, "step": 5484 }, { "epoch": 0.5, "learning_rate": 2.5745429925090135e-05, "loss": 1.2551, "step": 5485 }, { "epoch": 0.5, "learning_rate": 2.573826442401796e-05, "loss": 1.232, "step": 5486 }, { "epoch": 0.5, "learning_rate": 2.573109886224314e-05, "loss": 1.2288, "step": 5487 }, { "epoch": 0.5, "learning_rate": 2.5723933240354848e-05, "loss": 1.2905, "step": 5488 }, { "epoch": 0.5, "learning_rate": 2.5716767558942267e-05, "loss": 1.2463, "step": 5489 }, { "epoch": 0.5, "learning_rate": 2.570960181859458e-05, "loss": 1.2165, "step": 5490 }, { "epoch": 0.5, "learning_rate": 2.570243601990099e-05, "loss": 1.2273, "step": 5491 }, { "epoch": 0.5, "learning_rate": 2.5695270163450686e-05, "loss": 1.2437, "step": 5492 }, { "epoch": 0.5, "learning_rate": 2.5688104249832862e-05, "loss": 1.3076, "step": 5493 }, { "epoch": 0.5, "learning_rate": 2.5680938279636747e-05, "loss": 1.2732, "step": 5494 }, { "epoch": 0.5, "learning_rate": 2.5673772253451528e-05, "loss": 1.2593, "step": 5495 }, { "epoch": 0.5, "learning_rate": 2.5666606171866426e-05, "loss": 1.1998, "step": 5496 }, { "epoch": 0.5, "learning_rate": 2.5659440035470673e-05, "loss": 1.2822, "step": 5497 }, { "epoch": 0.5, "learning_rate": 2.5652273844853482e-05, "loss": 1.2553, "step": 5498 }, { "epoch": 0.5, "learning_rate": 2.5645107600604085e-05, "loss": 1.1912, "step": 5499 }, { "epoch": 0.5, "learning_rate": 2.563794130331172e-05, "loss": 1.2469, "step": 5500 }, { "epoch": 0.5, "learning_rate": 2.563077495356561e-05, "loss": 1.251, "step": 5501 }, { "epoch": 0.5, "learning_rate": 2.562360855195502e-05, "loss": 1.2363, "step": 5502 }, { "epoch": 0.5, "learning_rate": 2.5616442099069176e-05, "loss": 1.2514, "step": 5503 }, { "epoch": 0.5, "learning_rate": 2.5609275595497336e-05, "loss": 1.2674, "step": 5504 }, { "epoch": 0.5, "learning_rate": 2.5602109041828755e-05, "loss": 1.2559, "step": 5505 }, { "epoch": 0.5, "learning_rate": 2.5594942438652688e-05, "loss": 1.2439, "step": 5506 }, { "epoch": 0.5, "learning_rate": 2.558777578655841e-05, "loss": 1.2201, "step": 5507 }, { "epoch": 0.5, "learning_rate": 2.5580609086135172e-05, "loss": 1.202, "step": 5508 }, { "epoch": 0.5, "learning_rate": 2.5573442337972254e-05, "loss": 1.209, "step": 5509 }, { "epoch": 0.5, "learning_rate": 2.556627554265893e-05, "loss": 1.2333, "step": 5510 }, { "epoch": 0.5, "learning_rate": 2.5559108700784468e-05, "loss": 1.287, "step": 5511 }, { "epoch": 0.5, "learning_rate": 2.5551941812938167e-05, "loss": 1.2759, "step": 5512 }, { "epoch": 0.5, "learning_rate": 2.554477487970931e-05, "loss": 1.2187, "step": 5513 }, { "epoch": 0.5, "learning_rate": 2.5537607901687176e-05, "loss": 1.2305, "step": 5514 }, { "epoch": 0.5, "learning_rate": 2.553044087946107e-05, "loss": 1.2603, "step": 5515 }, { "epoch": 0.5, "learning_rate": 2.5523273813620285e-05, "loss": 1.2483, "step": 5516 }, { "epoch": 0.5, "learning_rate": 2.5516106704754117e-05, "loss": 1.2384, "step": 5517 }, { "epoch": 0.5, "learning_rate": 2.5508939553451882e-05, "loss": 1.2009, "step": 5518 }, { "epoch": 0.5, "learning_rate": 2.5501772360302873e-05, "loss": 1.29, "step": 5519 }, { "epoch": 0.5, "learning_rate": 2.549460512589642e-05, "loss": 1.2565, "step": 5520 }, { "epoch": 0.5, "learning_rate": 2.548743785082182e-05, "loss": 1.2392, "step": 5521 }, { "epoch": 0.5, "learning_rate": 2.5480270535668398e-05, "loss": 1.2517, "step": 5522 }, { "epoch": 0.5, "learning_rate": 2.5473103181025476e-05, "loss": 1.2762, "step": 5523 }, { "epoch": 0.5, "learning_rate": 2.5465935787482383e-05, "loss": 1.2733, "step": 5524 }, { "epoch": 0.5, "learning_rate": 2.5458768355628436e-05, "loss": 1.1939, "step": 5525 }, { "epoch": 0.5, "learning_rate": 2.5451600886052972e-05, "loss": 1.2334, "step": 5526 }, { "epoch": 0.5, "learning_rate": 2.544443337934533e-05, "loss": 1.2551, "step": 5527 }, { "epoch": 0.5, "learning_rate": 2.5437265836094837e-05, "loss": 1.1608, "step": 5528 }, { "epoch": 0.5, "learning_rate": 2.5430098256890843e-05, "loss": 1.2425, "step": 5529 }, { "epoch": 0.5, "learning_rate": 2.5422930642322684e-05, "loss": 1.249, "step": 5530 }, { "epoch": 0.5, "learning_rate": 2.5415762992979713e-05, "loss": 1.254, "step": 5531 }, { "epoch": 0.5, "learning_rate": 2.5408595309451265e-05, "loss": 1.2545, "step": 5532 }, { "epoch": 0.5, "learning_rate": 2.540142759232671e-05, "loss": 1.2685, "step": 5533 }, { "epoch": 0.5, "learning_rate": 2.539425984219539e-05, "loss": 1.1984, "step": 5534 }, { "epoch": 0.5, "learning_rate": 2.5387092059646662e-05, "loss": 1.2368, "step": 5535 }, { "epoch": 0.5, "learning_rate": 2.53799242452699e-05, "loss": 1.2274, "step": 5536 }, { "epoch": 0.5, "learning_rate": 2.5372756399654446e-05, "loss": 1.2083, "step": 5537 }, { "epoch": 0.5, "learning_rate": 2.5365588523389678e-05, "loss": 1.2336, "step": 5538 }, { "epoch": 0.5, "learning_rate": 2.5358420617064958e-05, "loss": 1.218, "step": 5539 }, { "epoch": 0.5, "learning_rate": 2.5351252681269654e-05, "loss": 1.2459, "step": 5540 }, { "epoch": 0.5, "learning_rate": 2.5344084716593153e-05, "loss": 1.2533, "step": 5541 }, { "epoch": 0.5, "learning_rate": 2.533691672362482e-05, "loss": 1.2358, "step": 5542 }, { "epoch": 0.5, "learning_rate": 2.532974870295402e-05, "loss": 1.2229, "step": 5543 }, { "epoch": 0.5, "learning_rate": 2.532258065517015e-05, "loss": 1.1912, "step": 5544 }, { "epoch": 0.5, "learning_rate": 2.5315412580862596e-05, "loss": 1.2684, "step": 5545 }, { "epoch": 0.5, "learning_rate": 2.5308244480620717e-05, "loss": 1.2634, "step": 5546 }, { "epoch": 0.5, "learning_rate": 2.530107635503392e-05, "loss": 1.2148, "step": 5547 }, { "epoch": 0.5, "learning_rate": 2.529390820469158e-05, "loss": 1.2145, "step": 5548 }, { "epoch": 0.5, "learning_rate": 2.5286740030183103e-05, "loss": 1.211, "step": 5549 }, { "epoch": 0.5, "learning_rate": 2.527957183209787e-05, "loss": 1.2032, "step": 5550 }, { "epoch": 0.5, "learning_rate": 2.527240361102527e-05, "loss": 1.2281, "step": 5551 }, { "epoch": 0.5, "learning_rate": 2.5265235367554707e-05, "loss": 1.2452, "step": 5552 }, { "epoch": 0.5, "learning_rate": 2.525806710227558e-05, "loss": 1.2993, "step": 5553 }, { "epoch": 0.5, "learning_rate": 2.5250898815777286e-05, "loss": 1.2555, "step": 5554 }, { "epoch": 0.5, "learning_rate": 2.5243730508649223e-05, "loss": 1.2709, "step": 5555 }, { "epoch": 0.5, "learning_rate": 2.5236562181480794e-05, "loss": 1.2582, "step": 5556 }, { "epoch": 0.5, "learning_rate": 2.522939383486141e-05, "loss": 1.2117, "step": 5557 }, { "epoch": 0.5, "learning_rate": 2.5222225469380478e-05, "loss": 1.2785, "step": 5558 }, { "epoch": 0.5, "learning_rate": 2.521505708562739e-05, "loss": 1.2224, "step": 5559 }, { "epoch": 0.5, "learning_rate": 2.5207888684191578e-05, "loss": 1.2558, "step": 5560 }, { "epoch": 0.5, "learning_rate": 2.520072026566243e-05, "loss": 1.3187, "step": 5561 }, { "epoch": 0.5, "learning_rate": 2.5193551830629387e-05, "loss": 1.2445, "step": 5562 }, { "epoch": 0.5, "learning_rate": 2.5186383379681828e-05, "loss": 1.2334, "step": 5563 }, { "epoch": 0.5, "learning_rate": 2.517921491340919e-05, "loss": 1.2479, "step": 5564 }, { "epoch": 0.5, "learning_rate": 2.517204643240088e-05, "loss": 1.2924, "step": 5565 }, { "epoch": 0.5, "learning_rate": 2.5164877937246324e-05, "loss": 1.22, "step": 5566 }, { "epoch": 0.5, "learning_rate": 2.5157709428534936e-05, "loss": 1.2512, "step": 5567 }, { "epoch": 0.5, "learning_rate": 2.515054090685613e-05, "loss": 1.1894, "step": 5568 }, { "epoch": 0.5, "learning_rate": 2.514337237279933e-05, "loss": 1.248, "step": 5569 }, { "epoch": 0.5, "learning_rate": 2.5136203826953963e-05, "loss": 1.2149, "step": 5570 }, { "epoch": 0.5, "learning_rate": 2.5129035269909453e-05, "loss": 1.195, "step": 5571 }, { "epoch": 0.5, "learning_rate": 2.5121866702255204e-05, "loss": 1.2777, "step": 5572 }, { "epoch": 0.5, "learning_rate": 2.5114698124580667e-05, "loss": 1.2376, "step": 5573 }, { "epoch": 0.5, "learning_rate": 2.5107529537475245e-05, "loss": 1.2393, "step": 5574 }, { "epoch": 0.5, "learning_rate": 2.5100360941528385e-05, "loss": 1.2449, "step": 5575 }, { "epoch": 0.5, "learning_rate": 2.509319233732949e-05, "loss": 1.2528, "step": 5576 }, { "epoch": 0.5, "learning_rate": 2.5086023725468e-05, "loss": 1.2386, "step": 5577 }, { "epoch": 0.5, "learning_rate": 2.507885510653335e-05, "loss": 1.1736, "step": 5578 }, { "epoch": 0.5, "learning_rate": 2.507168648111496e-05, "loss": 1.2552, "step": 5579 }, { "epoch": 0.5, "learning_rate": 2.506451784980226e-05, "loss": 1.2365, "step": 5580 }, { "epoch": 0.5, "learning_rate": 2.5057349213184673e-05, "loss": 1.2046, "step": 5581 }, { "epoch": 0.5, "learning_rate": 2.505018057185164e-05, "loss": 1.2259, "step": 5582 }, { "epoch": 0.5, "learning_rate": 2.5043011926392596e-05, "loss": 1.2248, "step": 5583 }, { "epoch": 0.5, "learning_rate": 2.5035843277396947e-05, "loss": 1.2285, "step": 5584 }, { "epoch": 0.5, "learning_rate": 2.5028674625454145e-05, "loss": 1.2244, "step": 5585 }, { "epoch": 0.5, "learning_rate": 2.5021505971153626e-05, "loss": 1.2375, "step": 5586 }, { "epoch": 0.5, "learning_rate": 2.50143373150848e-05, "loss": 1.2137, "step": 5587 }, { "epoch": 0.5, "learning_rate": 2.5007168657837123e-05, "loss": 1.2412, "step": 5588 }, { "epoch": 0.5, "learning_rate": 2.5e-05, "loss": 1.3486, "step": 5589 }, { "epoch": 0.51, "learning_rate": 2.4992831342162883e-05, "loss": 1.2868, "step": 5590 }, { "epoch": 0.51, "learning_rate": 2.49856626849152e-05, "loss": 1.2529, "step": 5591 }, { "epoch": 0.51, "learning_rate": 2.497849402884639e-05, "loss": 1.282, "step": 5592 }, { "epoch": 0.51, "learning_rate": 2.4971325374545858e-05, "loss": 1.2367, "step": 5593 }, { "epoch": 0.51, "learning_rate": 2.496415672260306e-05, "loss": 1.2581, "step": 5594 }, { "epoch": 0.51, "learning_rate": 2.4956988073607413e-05, "loss": 1.2006, "step": 5595 }, { "epoch": 0.51, "learning_rate": 2.4949819428148363e-05, "loss": 1.1646, "step": 5596 }, { "epoch": 0.51, "learning_rate": 2.494265078681534e-05, "loss": 1.2428, "step": 5597 }, { "epoch": 0.51, "learning_rate": 2.4935482150197752e-05, "loss": 1.2454, "step": 5598 }, { "epoch": 0.51, "learning_rate": 2.4928313518885047e-05, "loss": 1.2092, "step": 5599 }, { "epoch": 0.51, "learning_rate": 2.4921144893466658e-05, "loss": 1.1957, "step": 5600 }, { "epoch": 0.51, "learning_rate": 2.4913976274532e-05, "loss": 1.3304, "step": 5601 }, { "epoch": 0.51, "learning_rate": 2.4906807662670513e-05, "loss": 1.2793, "step": 5602 }, { "epoch": 0.51, "learning_rate": 2.489963905847163e-05, "loss": 1.2531, "step": 5603 }, { "epoch": 0.51, "learning_rate": 2.4892470462524764e-05, "loss": 1.2458, "step": 5604 }, { "epoch": 0.51, "learning_rate": 2.4885301875419342e-05, "loss": 1.2567, "step": 5605 }, { "epoch": 0.51, "learning_rate": 2.4878133297744795e-05, "loss": 1.2278, "step": 5606 }, { "epoch": 0.51, "learning_rate": 2.487096473009055e-05, "loss": 1.2745, "step": 5607 }, { "epoch": 0.51, "learning_rate": 2.4863796173046042e-05, "loss": 1.2641, "step": 5608 }, { "epoch": 0.51, "learning_rate": 2.4856627627200676e-05, "loss": 1.2581, "step": 5609 }, { "epoch": 0.51, "learning_rate": 2.4849459093143877e-05, "loss": 1.2342, "step": 5610 }, { "epoch": 0.51, "learning_rate": 2.484229057146507e-05, "loss": 1.1972, "step": 5611 }, { "epoch": 0.51, "learning_rate": 2.4835122062753678e-05, "loss": 1.2567, "step": 5612 }, { "epoch": 0.51, "learning_rate": 2.482795356759913e-05, "loss": 1.1924, "step": 5613 }, { "epoch": 0.51, "learning_rate": 2.482078508659082e-05, "loss": 1.2331, "step": 5614 }, { "epoch": 0.51, "learning_rate": 2.4813616620318178e-05, "loss": 1.268, "step": 5615 }, { "epoch": 0.51, "learning_rate": 2.4806448169370623e-05, "loss": 1.1967, "step": 5616 }, { "epoch": 0.51, "learning_rate": 2.479927973433757e-05, "loss": 1.2961, "step": 5617 }, { "epoch": 0.51, "learning_rate": 2.4792111315808434e-05, "loss": 1.2013, "step": 5618 }, { "epoch": 0.51, "learning_rate": 2.4784942914372614e-05, "loss": 1.2442, "step": 5619 }, { "epoch": 0.51, "learning_rate": 2.4777774530619528e-05, "loss": 1.2637, "step": 5620 }, { "epoch": 0.51, "learning_rate": 2.4770606165138593e-05, "loss": 1.2441, "step": 5621 }, { "epoch": 0.51, "learning_rate": 2.4763437818519205e-05, "loss": 1.2168, "step": 5622 }, { "epoch": 0.51, "learning_rate": 2.475626949135079e-05, "loss": 1.2246, "step": 5623 }, { "epoch": 0.51, "learning_rate": 2.4749101184222723e-05, "loss": 1.2386, "step": 5624 }, { "epoch": 0.51, "learning_rate": 2.474193289772443e-05, "loss": 1.1988, "step": 5625 }, { "epoch": 0.51, "learning_rate": 2.47347646324453e-05, "loss": 1.2709, "step": 5626 }, { "epoch": 0.51, "learning_rate": 2.4727596388974734e-05, "loss": 1.1978, "step": 5627 }, { "epoch": 0.51, "learning_rate": 2.472042816790213e-05, "loss": 1.1569, "step": 5628 }, { "epoch": 0.51, "learning_rate": 2.4713259969816906e-05, "loss": 1.2819, "step": 5629 }, { "epoch": 0.51, "learning_rate": 2.4706091795308422e-05, "loss": 1.2836, "step": 5630 }, { "epoch": 0.51, "learning_rate": 2.4698923644966087e-05, "loss": 1.1801, "step": 5631 }, { "epoch": 0.51, "learning_rate": 2.4691755519379285e-05, "loss": 1.2123, "step": 5632 }, { "epoch": 0.51, "learning_rate": 2.468458741913741e-05, "loss": 1.2509, "step": 5633 }, { "epoch": 0.51, "learning_rate": 2.4677419344829854e-05, "loss": 1.2339, "step": 5634 }, { "epoch": 0.51, "learning_rate": 2.4670251297045983e-05, "loss": 1.1972, "step": 5635 }, { "epoch": 0.51, "learning_rate": 2.4663083276375187e-05, "loss": 1.2613, "step": 5636 }, { "epoch": 0.51, "learning_rate": 2.4655915283406853e-05, "loss": 1.2242, "step": 5637 }, { "epoch": 0.51, "learning_rate": 2.464874731873034e-05, "loss": 1.2323, "step": 5638 }, { "epoch": 0.51, "learning_rate": 2.4641579382935055e-05, "loss": 1.2611, "step": 5639 }, { "epoch": 0.51, "learning_rate": 2.463441147661033e-05, "loss": 1.2053, "step": 5640 }, { "epoch": 0.51, "learning_rate": 2.462724360034556e-05, "loss": 1.2552, "step": 5641 }, { "epoch": 0.51, "learning_rate": 2.462007575473011e-05, "loss": 1.2272, "step": 5642 }, { "epoch": 0.51, "learning_rate": 2.4612907940353337e-05, "loss": 1.2625, "step": 5643 }, { "epoch": 0.51, "learning_rate": 2.460574015780462e-05, "loss": 1.2363, "step": 5644 }, { "epoch": 0.51, "learning_rate": 2.4598572407673297e-05, "loss": 1.2191, "step": 5645 }, { "epoch": 0.51, "learning_rate": 2.459140469054874e-05, "loss": 1.2824, "step": 5646 }, { "epoch": 0.51, "learning_rate": 2.4584237007020296e-05, "loss": 1.2678, "step": 5647 }, { "epoch": 0.51, "learning_rate": 2.4577069357677315e-05, "loss": 1.2181, "step": 5648 }, { "epoch": 0.51, "learning_rate": 2.4569901743109163e-05, "loss": 1.2437, "step": 5649 }, { "epoch": 0.51, "learning_rate": 2.4562734163905166e-05, "loss": 1.263, "step": 5650 }, { "epoch": 0.51, "learning_rate": 2.4555566620654677e-05, "loss": 1.2592, "step": 5651 }, { "epoch": 0.51, "learning_rate": 2.454839911394703e-05, "loss": 1.1791, "step": 5652 }, { "epoch": 0.51, "learning_rate": 2.4541231644371566e-05, "loss": 1.211, "step": 5653 }, { "epoch": 0.51, "learning_rate": 2.453406421251762e-05, "loss": 1.2409, "step": 5654 }, { "epoch": 0.51, "learning_rate": 2.4526896818974533e-05, "loss": 1.2372, "step": 5655 }, { "epoch": 0.51, "learning_rate": 2.4519729464331608e-05, "loss": 1.2445, "step": 5656 }, { "epoch": 0.51, "learning_rate": 2.4512562149178185e-05, "loss": 1.1933, "step": 5657 }, { "epoch": 0.51, "learning_rate": 2.4505394874103586e-05, "loss": 1.2504, "step": 5658 }, { "epoch": 0.51, "learning_rate": 2.4498227639697123e-05, "loss": 1.2254, "step": 5659 }, { "epoch": 0.51, "learning_rate": 2.4491060446548127e-05, "loss": 1.3033, "step": 5660 }, { "epoch": 0.51, "learning_rate": 2.4483893295245885e-05, "loss": 1.2523, "step": 5661 }, { "epoch": 0.51, "learning_rate": 2.4476726186379717e-05, "loss": 1.2485, "step": 5662 }, { "epoch": 0.51, "learning_rate": 2.4469559120538933e-05, "loss": 1.2833, "step": 5663 }, { "epoch": 0.51, "learning_rate": 2.4462392098312823e-05, "loss": 1.2743, "step": 5664 }, { "epoch": 0.51, "learning_rate": 2.4455225120290697e-05, "loss": 1.1803, "step": 5665 }, { "epoch": 0.51, "learning_rate": 2.4448058187061835e-05, "loss": 1.3245, "step": 5666 }, { "epoch": 0.51, "learning_rate": 2.4440891299215534e-05, "loss": 1.2392, "step": 5667 }, { "epoch": 0.51, "learning_rate": 2.443372445734108e-05, "loss": 1.2712, "step": 5668 }, { "epoch": 0.51, "learning_rate": 2.4426557662027748e-05, "loss": 1.2692, "step": 5669 }, { "epoch": 0.51, "learning_rate": 2.4419390913864837e-05, "loss": 1.2222, "step": 5670 }, { "epoch": 0.51, "learning_rate": 2.44122242134416e-05, "loss": 1.2391, "step": 5671 }, { "epoch": 0.51, "learning_rate": 2.4405057561347315e-05, "loss": 1.247, "step": 5672 }, { "epoch": 0.51, "learning_rate": 2.4397890958171247e-05, "loss": 1.1788, "step": 5673 }, { "epoch": 0.51, "learning_rate": 2.4390724404502667e-05, "loss": 1.236, "step": 5674 }, { "epoch": 0.51, "learning_rate": 2.4383557900930833e-05, "loss": 1.2028, "step": 5675 }, { "epoch": 0.51, "learning_rate": 2.4376391448044994e-05, "loss": 1.2744, "step": 5676 }, { "epoch": 0.51, "learning_rate": 2.4369225046434392e-05, "loss": 1.2881, "step": 5677 }, { "epoch": 0.51, "learning_rate": 2.4362058696688286e-05, "loss": 1.2529, "step": 5678 }, { "epoch": 0.51, "learning_rate": 2.435489239939592e-05, "loss": 1.2274, "step": 5679 }, { "epoch": 0.51, "learning_rate": 2.434772615514652e-05, "loss": 1.272, "step": 5680 }, { "epoch": 0.51, "learning_rate": 2.4340559964529336e-05, "loss": 1.2303, "step": 5681 }, { "epoch": 0.51, "learning_rate": 2.4333393828133576e-05, "loss": 1.2406, "step": 5682 }, { "epoch": 0.51, "learning_rate": 2.4326227746548478e-05, "loss": 1.2763, "step": 5683 }, { "epoch": 0.51, "learning_rate": 2.4319061720363262e-05, "loss": 1.2265, "step": 5684 }, { "epoch": 0.51, "learning_rate": 2.4311895750167137e-05, "loss": 1.1819, "step": 5685 }, { "epoch": 0.51, "learning_rate": 2.4304729836549326e-05, "loss": 1.2426, "step": 5686 }, { "epoch": 0.51, "learning_rate": 2.4297563980099016e-05, "loss": 1.2508, "step": 5687 }, { "epoch": 0.51, "learning_rate": 2.4290398181405426e-05, "loss": 1.2321, "step": 5688 }, { "epoch": 0.51, "learning_rate": 2.428323244105774e-05, "loss": 1.2101, "step": 5689 }, { "epoch": 0.51, "learning_rate": 2.4276066759645154e-05, "loss": 1.2148, "step": 5690 }, { "epoch": 0.51, "learning_rate": 2.4268901137756868e-05, "loss": 1.2479, "step": 5691 }, { "epoch": 0.51, "learning_rate": 2.4261735575982043e-05, "loss": 1.2452, "step": 5692 }, { "epoch": 0.51, "learning_rate": 2.425457007490987e-05, "loss": 1.2424, "step": 5693 }, { "epoch": 0.51, "learning_rate": 2.4247404635129514e-05, "loss": 1.2194, "step": 5694 }, { "epoch": 0.51, "learning_rate": 2.4240239257230142e-05, "loss": 1.2368, "step": 5695 }, { "epoch": 0.51, "learning_rate": 2.423307394180093e-05, "loss": 1.1663, "step": 5696 }, { "epoch": 0.51, "learning_rate": 2.4225908689431022e-05, "loss": 1.2465, "step": 5697 }, { "epoch": 0.51, "learning_rate": 2.4218743500709568e-05, "loss": 1.2449, "step": 5698 }, { "epoch": 0.51, "learning_rate": 2.4211578376225716e-05, "loss": 1.2465, "step": 5699 }, { "epoch": 0.52, "learning_rate": 2.4204413316568617e-05, "loss": 1.2714, "step": 5700 }, { "epoch": 0.52, "learning_rate": 2.4197248322327398e-05, "loss": 1.2032, "step": 5701 }, { "epoch": 0.52, "learning_rate": 2.4190083394091194e-05, "loss": 1.2875, "step": 5702 }, { "epoch": 0.52, "learning_rate": 2.4182918532449116e-05, "loss": 1.1996, "step": 5703 }, { "epoch": 0.52, "learning_rate": 2.41757537379903e-05, "loss": 1.2238, "step": 5704 }, { "epoch": 0.52, "learning_rate": 2.416858901130386e-05, "loss": 1.3189, "step": 5705 }, { "epoch": 0.52, "learning_rate": 2.4161424352978895e-05, "loss": 1.2097, "step": 5706 }, { "epoch": 0.52, "learning_rate": 2.4154259763604523e-05, "loss": 1.2485, "step": 5707 }, { "epoch": 0.52, "learning_rate": 2.4147095243769822e-05, "loss": 1.2564, "step": 5708 }, { "epoch": 0.52, "learning_rate": 2.4139930794063896e-05, "loss": 1.2158, "step": 5709 }, { "epoch": 0.52, "learning_rate": 2.413276641507582e-05, "loss": 1.2261, "step": 5710 }, { "epoch": 0.52, "learning_rate": 2.4125602107394684e-05, "loss": 1.2268, "step": 5711 }, { "epoch": 0.52, "learning_rate": 2.411843787160957e-05, "loss": 1.2499, "step": 5712 }, { "epoch": 0.52, "learning_rate": 2.4111273708309527e-05, "loss": 1.2482, "step": 5713 }, { "epoch": 0.52, "learning_rate": 2.4104109618083624e-05, "loss": 1.253, "step": 5714 }, { "epoch": 0.52, "learning_rate": 2.409694560152092e-05, "loss": 1.2261, "step": 5715 }, { "epoch": 0.52, "learning_rate": 2.4089781659210465e-05, "loss": 1.2704, "step": 5716 }, { "epoch": 0.52, "learning_rate": 2.4082617791741308e-05, "loss": 1.2152, "step": 5717 }, { "epoch": 0.52, "learning_rate": 2.407545399970248e-05, "loss": 1.2298, "step": 5718 }, { "epoch": 0.52, "learning_rate": 2.4068290283683007e-05, "loss": 1.2747, "step": 5719 }, { "epoch": 0.52, "learning_rate": 2.406112664427192e-05, "loss": 1.2232, "step": 5720 }, { "epoch": 0.52, "learning_rate": 2.4053963082058244e-05, "loss": 1.2267, "step": 5721 }, { "epoch": 0.52, "learning_rate": 2.404679959763099e-05, "loss": 1.2277, "step": 5722 }, { "epoch": 0.52, "learning_rate": 2.403963619157916e-05, "loss": 1.2512, "step": 5723 }, { "epoch": 0.52, "learning_rate": 2.403247286449175e-05, "loss": 1.3084, "step": 5724 }, { "epoch": 0.52, "learning_rate": 2.402530961695776e-05, "loss": 1.2053, "step": 5725 }, { "epoch": 0.52, "learning_rate": 2.401814644956618e-05, "loss": 1.2402, "step": 5726 }, { "epoch": 0.52, "learning_rate": 2.4010983362905995e-05, "loss": 1.2376, "step": 5727 }, { "epoch": 0.52, "learning_rate": 2.4003820357566156e-05, "loss": 1.1737, "step": 5728 }, { "epoch": 0.52, "learning_rate": 2.3996657434135647e-05, "loss": 1.2128, "step": 5729 }, { "epoch": 0.52, "learning_rate": 2.3989494593203426e-05, "loss": 1.242, "step": 5730 }, { "epoch": 0.52, "learning_rate": 2.3982331835358445e-05, "loss": 1.2646, "step": 5731 }, { "epoch": 0.52, "learning_rate": 2.3975169161189647e-05, "loss": 1.2733, "step": 5732 }, { "epoch": 0.52, "learning_rate": 2.396800657128599e-05, "loss": 1.2314, "step": 5733 }, { "epoch": 0.52, "learning_rate": 2.396084406623638e-05, "loss": 1.2268, "step": 5734 }, { "epoch": 0.52, "learning_rate": 2.3953681646629765e-05, "loss": 1.1726, "step": 5735 }, { "epoch": 0.52, "learning_rate": 2.394651931305505e-05, "loss": 1.2301, "step": 5736 }, { "epoch": 0.52, "learning_rate": 2.393935706610115e-05, "loss": 1.2339, "step": 5737 }, { "epoch": 0.52, "learning_rate": 2.393219490635698e-05, "loss": 1.2224, "step": 5738 }, { "epoch": 0.52, "learning_rate": 2.3925032834411424e-05, "loss": 1.2178, "step": 5739 }, { "epoch": 0.52, "learning_rate": 2.3917870850853372e-05, "loss": 1.2331, "step": 5740 }, { "epoch": 0.52, "learning_rate": 2.391070895627171e-05, "loss": 1.222, "step": 5741 }, { "epoch": 0.52, "learning_rate": 2.3903547151255325e-05, "loss": 1.3109, "step": 5742 }, { "epoch": 0.52, "learning_rate": 2.3896385436393075e-05, "loss": 1.2261, "step": 5743 }, { "epoch": 0.52, "learning_rate": 2.388922381227382e-05, "loss": 1.2407, "step": 5744 }, { "epoch": 0.52, "learning_rate": 2.388206227948641e-05, "loss": 1.2589, "step": 5745 }, { "epoch": 0.52, "learning_rate": 2.3874900838619695e-05, "loss": 1.2106, "step": 5746 }, { "epoch": 0.52, "learning_rate": 2.3867739490262526e-05, "loss": 1.2059, "step": 5747 }, { "epoch": 0.52, "learning_rate": 2.386057823500372e-05, "loss": 1.2453, "step": 5748 }, { "epoch": 0.52, "learning_rate": 2.38534170734321e-05, "loss": 1.2117, "step": 5749 }, { "epoch": 0.52, "learning_rate": 2.3846256006136478e-05, "loss": 1.2474, "step": 5750 }, { "epoch": 0.52, "learning_rate": 2.3839095033705674e-05, "loss": 1.2327, "step": 5751 }, { "epoch": 0.52, "learning_rate": 2.383193415672848e-05, "loss": 1.2827, "step": 5752 }, { "epoch": 0.52, "learning_rate": 2.3824773375793684e-05, "loss": 1.1777, "step": 5753 }, { "epoch": 0.52, "learning_rate": 2.3817612691490088e-05, "loss": 1.2525, "step": 5754 }, { "epoch": 0.52, "learning_rate": 2.3810452104406444e-05, "loss": 1.2184, "step": 5755 }, { "epoch": 0.52, "learning_rate": 2.380329161513154e-05, "loss": 1.234, "step": 5756 }, { "epoch": 0.52, "learning_rate": 2.379613122425412e-05, "loss": 1.2903, "step": 5757 }, { "epoch": 0.52, "learning_rate": 2.3788970932362945e-05, "loss": 1.2076, "step": 5758 }, { "epoch": 0.52, "learning_rate": 2.378181074004677e-05, "loss": 1.202, "step": 5759 }, { "epoch": 0.52, "learning_rate": 2.377465064789431e-05, "loss": 1.1906, "step": 5760 }, { "epoch": 0.52, "learning_rate": 2.3767490656494297e-05, "loss": 1.2298, "step": 5761 }, { "epoch": 0.52, "learning_rate": 2.3760330766435457e-05, "loss": 1.1966, "step": 5762 }, { "epoch": 0.52, "learning_rate": 2.3753170978306503e-05, "loss": 1.1875, "step": 5763 }, { "epoch": 0.52, "learning_rate": 2.3746011292696134e-05, "loss": 1.2201, "step": 5764 }, { "epoch": 0.52, "learning_rate": 2.373885171019304e-05, "loss": 1.2409, "step": 5765 }, { "epoch": 0.52, "learning_rate": 2.37316922313859e-05, "loss": 1.254, "step": 5766 }, { "epoch": 0.52, "learning_rate": 2.372453285686341e-05, "loss": 1.2318, "step": 5767 }, { "epoch": 0.52, "learning_rate": 2.3717373587214227e-05, "loss": 1.2516, "step": 5768 }, { "epoch": 0.52, "learning_rate": 2.371021442302702e-05, "loss": 1.2416, "step": 5769 }, { "epoch": 0.52, "learning_rate": 2.3703055364890423e-05, "loss": 1.2396, "step": 5770 }, { "epoch": 0.52, "learning_rate": 2.369589641339309e-05, "loss": 1.2569, "step": 5771 }, { "epoch": 0.52, "learning_rate": 2.3688737569123653e-05, "loss": 1.255, "step": 5772 }, { "epoch": 0.52, "learning_rate": 2.3681578832670735e-05, "loss": 1.2467, "step": 5773 }, { "epoch": 0.52, "learning_rate": 2.3674420204622966e-05, "loss": 1.2424, "step": 5774 }, { "epoch": 0.52, "learning_rate": 2.3667261685568924e-05, "loss": 1.24, "step": 5775 }, { "epoch": 0.52, "learning_rate": 2.3660103276097232e-05, "loss": 1.2596, "step": 5776 }, { "epoch": 0.52, "learning_rate": 2.3652944976796473e-05, "loss": 1.2407, "step": 5777 }, { "epoch": 0.52, "learning_rate": 2.364578678825522e-05, "loss": 1.2721, "step": 5778 }, { "epoch": 0.52, "learning_rate": 2.363862871106205e-05, "loss": 1.2482, "step": 5779 }, { "epoch": 0.52, "learning_rate": 2.3631470745805534e-05, "loss": 1.1664, "step": 5780 }, { "epoch": 0.52, "learning_rate": 2.362431289307421e-05, "loss": 1.2453, "step": 5781 }, { "epoch": 0.52, "learning_rate": 2.3617155153456622e-05, "loss": 1.2076, "step": 5782 }, { "epoch": 0.52, "learning_rate": 2.3609997527541306e-05, "loss": 1.1871, "step": 5783 }, { "epoch": 0.52, "learning_rate": 2.3602840015916792e-05, "loss": 1.1996, "step": 5784 }, { "epoch": 0.52, "learning_rate": 2.3595682619171597e-05, "loss": 1.202, "step": 5785 }, { "epoch": 0.52, "learning_rate": 2.3588525337894217e-05, "loss": 1.2434, "step": 5786 }, { "epoch": 0.52, "learning_rate": 2.3581368172673152e-05, "loss": 1.185, "step": 5787 }, { "epoch": 0.52, "learning_rate": 2.357421112409689e-05, "loss": 1.2625, "step": 5788 }, { "epoch": 0.52, "learning_rate": 2.3567054192753912e-05, "loss": 1.2434, "step": 5789 }, { "epoch": 0.52, "learning_rate": 2.3559897379232686e-05, "loss": 1.2078, "step": 5790 }, { "epoch": 0.52, "learning_rate": 2.355274068412166e-05, "loss": 1.235, "step": 5791 }, { "epoch": 0.52, "learning_rate": 2.3545584108009284e-05, "loss": 1.2854, "step": 5792 }, { "epoch": 0.52, "learning_rate": 2.3538427651484006e-05, "loss": 1.2319, "step": 5793 }, { "epoch": 0.52, "learning_rate": 2.3531271315134244e-05, "loss": 1.2331, "step": 5794 }, { "epoch": 0.52, "learning_rate": 2.352411509954843e-05, "loss": 1.1992, "step": 5795 }, { "epoch": 0.52, "learning_rate": 2.3516959005314955e-05, "loss": 1.2377, "step": 5796 }, { "epoch": 0.52, "learning_rate": 2.3509803033022228e-05, "loss": 1.2324, "step": 5797 }, { "epoch": 0.52, "learning_rate": 2.3502647183258644e-05, "loss": 1.2451, "step": 5798 }, { "epoch": 0.52, "learning_rate": 2.3495491456612566e-05, "loss": 1.2566, "step": 5799 }, { "epoch": 0.52, "learning_rate": 2.3488335853672384e-05, "loss": 1.2741, "step": 5800 }, { "epoch": 0.52, "learning_rate": 2.348118037502643e-05, "loss": 1.2519, "step": 5801 }, { "epoch": 0.52, "learning_rate": 2.3474025021263076e-05, "loss": 1.2164, "step": 5802 }, { "epoch": 0.52, "learning_rate": 2.3466869792970635e-05, "loss": 1.1915, "step": 5803 }, { "epoch": 0.52, "learning_rate": 2.3459714690737455e-05, "loss": 1.229, "step": 5804 }, { "epoch": 0.52, "learning_rate": 2.3452559715151853e-05, "loss": 1.2316, "step": 5805 }, { "epoch": 0.52, "learning_rate": 2.344540486680213e-05, "loss": 1.2111, "step": 5806 }, { "epoch": 0.52, "learning_rate": 2.3438250146276578e-05, "loss": 1.2732, "step": 5807 }, { "epoch": 0.52, "learning_rate": 2.3431095554163486e-05, "loss": 1.2447, "step": 5808 }, { "epoch": 0.52, "learning_rate": 2.3423941091051126e-05, "loss": 1.2224, "step": 5809 }, { "epoch": 0.52, "learning_rate": 2.3416786757527776e-05, "loss": 1.2232, "step": 5810 }, { "epoch": 0.53, "learning_rate": 2.3409632554181685e-05, "loss": 1.2438, "step": 5811 }, { "epoch": 0.53, "learning_rate": 2.3402478481601082e-05, "loss": 1.2653, "step": 5812 }, { "epoch": 0.53, "learning_rate": 2.3395324540374208e-05, "loss": 1.2123, "step": 5813 }, { "epoch": 0.53, "learning_rate": 2.3388170731089294e-05, "loss": 1.289, "step": 5814 }, { "epoch": 0.53, "learning_rate": 2.3381017054334536e-05, "loss": 1.2732, "step": 5815 }, { "epoch": 0.53, "learning_rate": 2.337386351069815e-05, "loss": 1.2459, "step": 5816 }, { "epoch": 0.53, "learning_rate": 2.336671010076831e-05, "loss": 1.2197, "step": 5817 }, { "epoch": 0.53, "learning_rate": 2.33595568251332e-05, "loss": 1.2076, "step": 5818 }, { "epoch": 0.53, "learning_rate": 2.335240368438099e-05, "loss": 1.2848, "step": 5819 }, { "epoch": 0.53, "learning_rate": 2.334525067909983e-05, "loss": 1.2742, "step": 5820 }, { "epoch": 0.53, "learning_rate": 2.333809780987788e-05, "loss": 1.1765, "step": 5821 }, { "epoch": 0.53, "learning_rate": 2.3330945077303246e-05, "loss": 1.285, "step": 5822 }, { "epoch": 0.53, "learning_rate": 2.3323792481964075e-05, "loss": 1.1874, "step": 5823 }, { "epoch": 0.53, "learning_rate": 2.3316640024448463e-05, "loss": 1.1888, "step": 5824 }, { "epoch": 0.53, "learning_rate": 2.3309487705344517e-05, "loss": 1.2195, "step": 5825 }, { "epoch": 0.53, "learning_rate": 2.330233552524033e-05, "loss": 1.2329, "step": 5826 }, { "epoch": 0.53, "learning_rate": 2.3295183484723966e-05, "loss": 1.2105, "step": 5827 }, { "epoch": 0.53, "learning_rate": 2.3288031584383503e-05, "loss": 1.2938, "step": 5828 }, { "epoch": 0.53, "learning_rate": 2.3280879824806984e-05, "loss": 1.2592, "step": 5829 }, { "epoch": 0.53, "learning_rate": 2.3273728206582456e-05, "loss": 1.1858, "step": 5830 }, { "epoch": 0.53, "learning_rate": 2.3266576730297956e-05, "loss": 1.2395, "step": 5831 }, { "epoch": 0.53, "learning_rate": 2.3259425396541505e-05, "loss": 1.2841, "step": 5832 }, { "epoch": 0.53, "learning_rate": 2.325227420590109e-05, "loss": 1.2648, "step": 5833 }, { "epoch": 0.53, "learning_rate": 2.3245123158964718e-05, "loss": 1.1671, "step": 5834 }, { "epoch": 0.53, "learning_rate": 2.3237972256320383e-05, "loss": 1.2192, "step": 5835 }, { "epoch": 0.53, "learning_rate": 2.323082149855604e-05, "loss": 1.2131, "step": 5836 }, { "epoch": 0.53, "learning_rate": 2.3223670886259667e-05, "loss": 1.1939, "step": 5837 }, { "epoch": 0.53, "learning_rate": 2.3216520420019195e-05, "loss": 1.2293, "step": 5838 }, { "epoch": 0.53, "learning_rate": 2.3209370100422566e-05, "loss": 1.2051, "step": 5839 }, { "epoch": 0.53, "learning_rate": 2.3202219928057708e-05, "loss": 1.2327, "step": 5840 }, { "epoch": 0.53, "learning_rate": 2.3195069903512524e-05, "loss": 1.2299, "step": 5841 }, { "epoch": 0.53, "learning_rate": 2.3187920027374936e-05, "loss": 1.1754, "step": 5842 }, { "epoch": 0.53, "learning_rate": 2.3180770300232798e-05, "loss": 1.2621, "step": 5843 }, { "epoch": 0.53, "learning_rate": 2.3173620722674012e-05, "loss": 1.2395, "step": 5844 }, { "epoch": 0.53, "learning_rate": 2.3166471295286422e-05, "loss": 1.2312, "step": 5845 }, { "epoch": 0.53, "learning_rate": 2.315932201865789e-05, "loss": 1.2368, "step": 5846 }, { "epoch": 0.53, "learning_rate": 2.3152172893376263e-05, "loss": 1.2457, "step": 5847 }, { "epoch": 0.53, "learning_rate": 2.3145023920029344e-05, "loss": 1.2495, "step": 5848 }, { "epoch": 0.53, "learning_rate": 2.3137875099204965e-05, "loss": 1.2475, "step": 5849 }, { "epoch": 0.53, "learning_rate": 2.313072643149091e-05, "loss": 1.1764, "step": 5850 }, { "epoch": 0.53, "learning_rate": 2.3123577917474982e-05, "loss": 1.2207, "step": 5851 }, { "epoch": 0.53, "learning_rate": 2.3116429557744955e-05, "loss": 1.2597, "step": 5852 }, { "epoch": 0.53, "learning_rate": 2.310928135288859e-05, "loss": 1.2455, "step": 5853 }, { "epoch": 0.53, "learning_rate": 2.3102133303493627e-05, "loss": 1.2193, "step": 5854 }, { "epoch": 0.53, "learning_rate": 2.309498541014781e-05, "loss": 1.2661, "step": 5855 }, { "epoch": 0.53, "learning_rate": 2.308783767343887e-05, "loss": 1.2534, "step": 5856 }, { "epoch": 0.53, "learning_rate": 2.3080690093954514e-05, "loss": 1.2317, "step": 5857 }, { "epoch": 0.53, "learning_rate": 2.3073542672282444e-05, "loss": 1.2821, "step": 5858 }, { "epoch": 0.53, "learning_rate": 2.3066395409010333e-05, "loss": 1.2372, "step": 5859 }, { "epoch": 0.53, "learning_rate": 2.3059248304725865e-05, "loss": 1.2571, "step": 5860 }, { "epoch": 0.53, "learning_rate": 2.30521013600167e-05, "loss": 1.218, "step": 5861 }, { "epoch": 0.53, "learning_rate": 2.3044954575470476e-05, "loss": 1.2166, "step": 5862 }, { "epoch": 0.53, "learning_rate": 2.3037807951674843e-05, "loss": 1.1985, "step": 5863 }, { "epoch": 0.53, "learning_rate": 2.3030661489217397e-05, "loss": 1.2506, "step": 5864 }, { "epoch": 0.53, "learning_rate": 2.3023515188685766e-05, "loss": 1.2699, "step": 5865 }, { "epoch": 0.53, "learning_rate": 2.3016369050667526e-05, "loss": 1.2291, "step": 5866 }, { "epoch": 0.53, "learning_rate": 2.3009223075750268e-05, "loss": 1.2047, "step": 5867 }, { "epoch": 0.53, "learning_rate": 2.300207726452157e-05, "loss": 1.1792, "step": 5868 }, { "epoch": 0.53, "learning_rate": 2.299493161756896e-05, "loss": 1.197, "step": 5869 }, { "epoch": 0.53, "learning_rate": 2.2987786135479995e-05, "loss": 1.1905, "step": 5870 }, { "epoch": 0.53, "learning_rate": 2.2980640818842187e-05, "loss": 1.2373, "step": 5871 }, { "epoch": 0.53, "learning_rate": 2.297349566824306e-05, "loss": 1.2563, "step": 5872 }, { "epoch": 0.53, "learning_rate": 2.2966350684270126e-05, "loss": 1.2496, "step": 5873 }, { "epoch": 0.53, "learning_rate": 2.2959205867510845e-05, "loss": 1.2201, "step": 5874 }, { "epoch": 0.53, "learning_rate": 2.295206121855269e-05, "loss": 1.2443, "step": 5875 }, { "epoch": 0.53, "learning_rate": 2.2944916737983132e-05, "loss": 1.233, "step": 5876 }, { "epoch": 0.53, "learning_rate": 2.2937772426389615e-05, "loss": 1.2328, "step": 5877 }, { "epoch": 0.53, "learning_rate": 2.2930628284359555e-05, "loss": 1.2378, "step": 5878 }, { "epoch": 0.53, "learning_rate": 2.2923484312480393e-05, "loss": 1.2425, "step": 5879 }, { "epoch": 0.53, "learning_rate": 2.2916340511339498e-05, "loss": 1.1838, "step": 5880 }, { "epoch": 0.53, "learning_rate": 2.2909196881524278e-05, "loss": 1.204, "step": 5881 }, { "epoch": 0.53, "learning_rate": 2.290205342362211e-05, "loss": 1.2777, "step": 5882 }, { "epoch": 0.53, "learning_rate": 2.2894910138220338e-05, "loss": 1.223, "step": 5883 }, { "epoch": 0.53, "learning_rate": 2.288776702590633e-05, "loss": 1.1909, "step": 5884 }, { "epoch": 0.53, "learning_rate": 2.2880624087267397e-05, "loss": 1.27, "step": 5885 }, { "epoch": 0.53, "learning_rate": 2.2873481322890862e-05, "loss": 1.1996, "step": 5886 }, { "epoch": 0.53, "learning_rate": 2.2866338733364028e-05, "loss": 1.2902, "step": 5887 }, { "epoch": 0.53, "learning_rate": 2.2859196319274184e-05, "loss": 1.1945, "step": 5888 }, { "epoch": 0.53, "learning_rate": 2.2852054081208615e-05, "loss": 1.2331, "step": 5889 }, { "epoch": 0.53, "learning_rate": 2.2844912019754558e-05, "loss": 1.2175, "step": 5890 }, { "epoch": 0.53, "learning_rate": 2.2837770135499276e-05, "loss": 1.2141, "step": 5891 }, { "epoch": 0.53, "learning_rate": 2.283062842902999e-05, "loss": 1.2313, "step": 5892 }, { "epoch": 0.53, "learning_rate": 2.2823486900933917e-05, "loss": 1.1804, "step": 5893 }, { "epoch": 0.53, "learning_rate": 2.281634555179827e-05, "loss": 1.2576, "step": 5894 }, { "epoch": 0.53, "learning_rate": 2.280920438221022e-05, "loss": 1.1915, "step": 5895 }, { "epoch": 0.53, "learning_rate": 2.2802063392756938e-05, "loss": 1.2511, "step": 5896 }, { "epoch": 0.53, "learning_rate": 2.279492258402559e-05, "loss": 1.2661, "step": 5897 }, { "epoch": 0.53, "learning_rate": 2.2787781956603317e-05, "loss": 1.2379, "step": 5898 }, { "epoch": 0.53, "learning_rate": 2.2780641511077243e-05, "loss": 1.2149, "step": 5899 }, { "epoch": 0.53, "learning_rate": 2.277350124803448e-05, "loss": 1.1807, "step": 5900 }, { "epoch": 0.53, "learning_rate": 2.2766361168062124e-05, "loss": 1.3336, "step": 5901 }, { "epoch": 0.53, "learning_rate": 2.2759221271747256e-05, "loss": 1.2266, "step": 5902 }, { "epoch": 0.53, "learning_rate": 2.2752081559676952e-05, "loss": 1.2368, "step": 5903 }, { "epoch": 0.53, "learning_rate": 2.2744942032438253e-05, "loss": 1.2123, "step": 5904 }, { "epoch": 0.53, "learning_rate": 2.273780269061821e-05, "loss": 1.1972, "step": 5905 }, { "epoch": 0.53, "learning_rate": 2.2730663534803824e-05, "loss": 1.2677, "step": 5906 }, { "epoch": 0.53, "learning_rate": 2.2723524565582116e-05, "loss": 1.2314, "step": 5907 }, { "epoch": 0.53, "learning_rate": 2.2716385783540066e-05, "loss": 1.2037, "step": 5908 }, { "epoch": 0.53, "learning_rate": 2.2709247189264655e-05, "loss": 1.2356, "step": 5909 }, { "epoch": 0.53, "learning_rate": 2.2702108783342855e-05, "loss": 1.2169, "step": 5910 }, { "epoch": 0.53, "learning_rate": 2.269497056636159e-05, "loss": 1.2778, "step": 5911 }, { "epoch": 0.53, "learning_rate": 2.26878325389078e-05, "loss": 1.1919, "step": 5912 }, { "epoch": 0.53, "learning_rate": 2.268069470156839e-05, "loss": 1.2912, "step": 5913 }, { "epoch": 0.53, "learning_rate": 2.2673557054930264e-05, "loss": 1.2624, "step": 5914 }, { "epoch": 0.53, "learning_rate": 2.2666419599580314e-05, "loss": 1.2405, "step": 5915 }, { "epoch": 0.53, "learning_rate": 2.265928233610539e-05, "loss": 1.2953, "step": 5916 }, { "epoch": 0.53, "learning_rate": 2.2652145265092343e-05, "loss": 1.2499, "step": 5917 }, { "epoch": 0.53, "learning_rate": 2.2645008387128012e-05, "loss": 1.2505, "step": 5918 }, { "epoch": 0.53, "learning_rate": 2.263787170279922e-05, "loss": 1.2591, "step": 5919 }, { "epoch": 0.53, "learning_rate": 2.2630735212692767e-05, "loss": 1.2522, "step": 5920 }, { "epoch": 0.53, "learning_rate": 2.2623598917395438e-05, "loss": 1.2763, "step": 5921 }, { "epoch": 0.54, "learning_rate": 2.2616462817494003e-05, "loss": 1.2229, "step": 5922 }, { "epoch": 0.54, "learning_rate": 2.2609326913575214e-05, "loss": 1.2841, "step": 5923 }, { "epoch": 0.54, "learning_rate": 2.260219120622582e-05, "loss": 1.2566, "step": 5924 }, { "epoch": 0.54, "learning_rate": 2.2595055696032543e-05, "loss": 1.2515, "step": 5925 }, { "epoch": 0.54, "learning_rate": 2.2587920383582074e-05, "loss": 1.2128, "step": 5926 }, { "epoch": 0.54, "learning_rate": 2.2580785269461114e-05, "loss": 1.2543, "step": 5927 }, { "epoch": 0.54, "learning_rate": 2.2573650354256338e-05, "loss": 1.2533, "step": 5928 }, { "epoch": 0.54, "learning_rate": 2.2566515638554394e-05, "loss": 1.2024, "step": 5929 }, { "epoch": 0.54, "learning_rate": 2.2559381122941934e-05, "loss": 1.1957, "step": 5930 }, { "epoch": 0.54, "learning_rate": 2.2552246808005586e-05, "loss": 1.2376, "step": 5931 }, { "epoch": 0.54, "learning_rate": 2.254511269433194e-05, "loss": 1.2688, "step": 5932 }, { "epoch": 0.54, "learning_rate": 2.2537978782507603e-05, "loss": 1.277, "step": 5933 }, { "epoch": 0.54, "learning_rate": 2.2530845073119138e-05, "loss": 1.2149, "step": 5934 }, { "epoch": 0.54, "learning_rate": 2.2523711566753115e-05, "loss": 1.174, "step": 5935 }, { "epoch": 0.54, "learning_rate": 2.2516578263996074e-05, "loss": 1.1805, "step": 5936 }, { "epoch": 0.54, "learning_rate": 2.2509445165434535e-05, "loss": 1.2191, "step": 5937 }, { "epoch": 0.54, "learning_rate": 2.2502312271655e-05, "loss": 1.3014, "step": 5938 }, { "epoch": 0.54, "learning_rate": 2.2495179583243967e-05, "loss": 1.2891, "step": 5939 }, { "epoch": 0.54, "learning_rate": 2.2488047100787916e-05, "loss": 1.2519, "step": 5940 }, { "epoch": 0.54, "learning_rate": 2.2480914824873297e-05, "loss": 1.2669, "step": 5941 }, { "epoch": 0.54, "learning_rate": 2.2473782756086554e-05, "loss": 1.1957, "step": 5942 }, { "epoch": 0.54, "learning_rate": 2.24666508950141e-05, "loss": 1.2559, "step": 5943 }, { "epoch": 0.54, "learning_rate": 2.2459519242242352e-05, "loss": 1.248, "step": 5944 }, { "epoch": 0.54, "learning_rate": 2.24523877983577e-05, "loss": 1.1572, "step": 5945 }, { "epoch": 0.54, "learning_rate": 2.244525656394652e-05, "loss": 1.2747, "step": 5946 }, { "epoch": 0.54, "learning_rate": 2.243812553959515e-05, "loss": 1.2733, "step": 5947 }, { "epoch": 0.54, "learning_rate": 2.2430994725889932e-05, "loss": 1.239, "step": 5948 }, { "epoch": 0.54, "learning_rate": 2.24238641234172e-05, "loss": 1.1819, "step": 5949 }, { "epoch": 0.54, "learning_rate": 2.241673373276324e-05, "loss": 1.1953, "step": 5950 }, { "epoch": 0.54, "learning_rate": 2.2409603554514355e-05, "loss": 1.2199, "step": 5951 }, { "epoch": 0.54, "learning_rate": 2.2402473589256793e-05, "loss": 1.2413, "step": 5952 }, { "epoch": 0.54, "learning_rate": 2.2395343837576814e-05, "loss": 1.2471, "step": 5953 }, { "epoch": 0.54, "learning_rate": 2.2388214300060654e-05, "loss": 1.2598, "step": 5954 }, { "epoch": 0.54, "learning_rate": 2.238108497729452e-05, "loss": 1.2204, "step": 5955 }, { "epoch": 0.54, "learning_rate": 2.2373955869864615e-05, "loss": 1.2711, "step": 5956 }, { "epoch": 0.54, "learning_rate": 2.236682697835713e-05, "loss": 1.2032, "step": 5957 }, { "epoch": 0.54, "learning_rate": 2.235969830335821e-05, "loss": 1.2031, "step": 5958 }, { "epoch": 0.54, "learning_rate": 2.2352569845453995e-05, "loss": 1.2122, "step": 5959 }, { "epoch": 0.54, "learning_rate": 2.2345441605230628e-05, "loss": 1.2473, "step": 5960 }, { "epoch": 0.54, "learning_rate": 2.233831358327421e-05, "loss": 1.2649, "step": 5961 }, { "epoch": 0.54, "learning_rate": 2.2331185780170836e-05, "loss": 1.236, "step": 5962 }, { "epoch": 0.54, "learning_rate": 2.232405819650657e-05, "loss": 1.2601, "step": 5963 }, { "epoch": 0.54, "learning_rate": 2.2316930832867468e-05, "loss": 1.2537, "step": 5964 }, { "epoch": 0.54, "learning_rate": 2.230980368983957e-05, "loss": 1.2221, "step": 5965 }, { "epoch": 0.54, "learning_rate": 2.2302676768008897e-05, "loss": 1.2373, "step": 5966 }, { "epoch": 0.54, "learning_rate": 2.2295550067961455e-05, "loss": 1.245, "step": 5967 }, { "epoch": 0.54, "learning_rate": 2.22884235902832e-05, "loss": 1.2484, "step": 5968 }, { "epoch": 0.54, "learning_rate": 2.228129733556012e-05, "loss": 1.3049, "step": 5969 }, { "epoch": 0.54, "learning_rate": 2.2274171304378154e-05, "loss": 1.2285, "step": 5970 }, { "epoch": 0.54, "learning_rate": 2.2267045497323223e-05, "loss": 1.2687, "step": 5971 }, { "epoch": 0.54, "learning_rate": 2.225991991498125e-05, "loss": 1.188, "step": 5972 }, { "epoch": 0.54, "learning_rate": 2.2252794557938104e-05, "loss": 1.2706, "step": 5973 }, { "epoch": 0.54, "learning_rate": 2.2245669426779666e-05, "loss": 1.1835, "step": 5974 }, { "epoch": 0.54, "learning_rate": 2.2238544522091797e-05, "loss": 1.2538, "step": 5975 }, { "epoch": 0.54, "learning_rate": 2.223141984446032e-05, "loss": 1.2108, "step": 5976 }, { "epoch": 0.54, "learning_rate": 2.2224295394471066e-05, "loss": 1.2869, "step": 5977 }, { "epoch": 0.54, "learning_rate": 2.2217171172709807e-05, "loss": 1.284, "step": 5978 }, { "epoch": 0.54, "learning_rate": 2.221004717976234e-05, "loss": 1.2322, "step": 5979 }, { "epoch": 0.54, "learning_rate": 2.2202923416214416e-05, "loss": 1.2266, "step": 5980 }, { "epoch": 0.54, "learning_rate": 2.219579988265178e-05, "loss": 1.2287, "step": 5981 }, { "epoch": 0.54, "learning_rate": 2.2188676579660154e-05, "loss": 1.2917, "step": 5982 }, { "epoch": 0.54, "learning_rate": 2.2181553507825247e-05, "loss": 1.2127, "step": 5983 }, { "epoch": 0.54, "learning_rate": 2.217443066773272e-05, "loss": 1.2817, "step": 5984 }, { "epoch": 0.54, "learning_rate": 2.2167308059968254e-05, "loss": 1.2644, "step": 5985 }, { "epoch": 0.54, "learning_rate": 2.2160185685117494e-05, "loss": 1.2383, "step": 5986 }, { "epoch": 0.54, "learning_rate": 2.2153063543766067e-05, "loss": 1.2915, "step": 5987 }, { "epoch": 0.54, "learning_rate": 2.2145941636499584e-05, "loss": 1.2427, "step": 5988 }, { "epoch": 0.54, "learning_rate": 2.2138819963903613e-05, "loss": 1.2154, "step": 5989 }, { "epoch": 0.54, "learning_rate": 2.213169852656374e-05, "loss": 1.241, "step": 5990 }, { "epoch": 0.54, "learning_rate": 2.212457732506551e-05, "loss": 1.2439, "step": 5991 }, { "epoch": 0.54, "learning_rate": 2.211745635999445e-05, "loss": 1.2583, "step": 5992 }, { "epoch": 0.54, "learning_rate": 2.2110335631936082e-05, "loss": 1.2117, "step": 5993 }, { "epoch": 0.54, "learning_rate": 2.2103215141475878e-05, "loss": 1.2424, "step": 5994 }, { "epoch": 0.54, "learning_rate": 2.2096094889199322e-05, "loss": 1.2381, "step": 5995 }, { "epoch": 0.54, "learning_rate": 2.2088974875691863e-05, "loss": 1.2541, "step": 5996 }, { "epoch": 0.54, "learning_rate": 2.2081855101538928e-05, "loss": 1.2776, "step": 5997 }, { "epoch": 0.54, "learning_rate": 2.2074735567325947e-05, "loss": 1.2661, "step": 5998 }, { "epoch": 0.54, "learning_rate": 2.206761627363829e-05, "loss": 1.2706, "step": 5999 }, { "epoch": 0.54, "learning_rate": 2.206049722106134e-05, "loss": 1.2632, "step": 6000 }, { "epoch": 0.54, "learning_rate": 2.2053378410180446e-05, "loss": 1.2143, "step": 6001 }, { "epoch": 0.54, "learning_rate": 2.204625984158095e-05, "loss": 1.2014, "step": 6002 }, { "epoch": 0.54, "learning_rate": 2.2039141515848167e-05, "loss": 1.2656, "step": 6003 }, { "epoch": 0.54, "learning_rate": 2.2032023433567378e-05, "loss": 1.2829, "step": 6004 }, { "epoch": 0.54, "learning_rate": 2.202490559532386e-05, "loss": 1.2163, "step": 6005 }, { "epoch": 0.54, "learning_rate": 2.201778800170286e-05, "loss": 1.2083, "step": 6006 }, { "epoch": 0.54, "learning_rate": 2.2010670653289624e-05, "loss": 1.2378, "step": 6007 }, { "epoch": 0.54, "learning_rate": 2.200355355066937e-05, "loss": 1.3061, "step": 6008 }, { "epoch": 0.54, "learning_rate": 2.199643669442728e-05, "loss": 1.1669, "step": 6009 }, { "epoch": 0.54, "learning_rate": 2.1989320085148517e-05, "loss": 1.2182, "step": 6010 }, { "epoch": 0.54, "learning_rate": 2.1982203723418242e-05, "loss": 1.2554, "step": 6011 }, { "epoch": 0.54, "learning_rate": 2.1975087609821598e-05, "loss": 1.243, "step": 6012 }, { "epoch": 0.54, "learning_rate": 2.196797174494368e-05, "loss": 1.2464, "step": 6013 }, { "epoch": 0.54, "learning_rate": 2.1960856129369593e-05, "loss": 1.2892, "step": 6014 }, { "epoch": 0.54, "learning_rate": 2.1953740763684393e-05, "loss": 1.2115, "step": 6015 }, { "epoch": 0.54, "learning_rate": 2.1946625648473136e-05, "loss": 1.2794, "step": 6016 }, { "epoch": 0.54, "learning_rate": 2.1939510784320857e-05, "loss": 1.2545, "step": 6017 }, { "epoch": 0.54, "learning_rate": 2.193239617181256e-05, "loss": 1.2474, "step": 6018 }, { "epoch": 0.54, "learning_rate": 2.1925281811533237e-05, "loss": 1.3244, "step": 6019 }, { "epoch": 0.54, "learning_rate": 2.191816770406784e-05, "loss": 1.2, "step": 6020 }, { "epoch": 0.54, "learning_rate": 2.1911053850001333e-05, "loss": 1.2414, "step": 6021 }, { "epoch": 0.54, "learning_rate": 2.1903940249918635e-05, "loss": 1.1871, "step": 6022 }, { "epoch": 0.54, "learning_rate": 2.1896826904404645e-05, "loss": 1.2519, "step": 6023 }, { "epoch": 0.54, "learning_rate": 2.1889713814044267e-05, "loss": 1.2176, "step": 6024 }, { "epoch": 0.54, "learning_rate": 2.1882600979422343e-05, "loss": 1.2041, "step": 6025 }, { "epoch": 0.54, "learning_rate": 2.1875488401123716e-05, "loss": 1.2215, "step": 6026 }, { "epoch": 0.54, "learning_rate": 2.186837607973321e-05, "loss": 1.2255, "step": 6027 }, { "epoch": 0.54, "learning_rate": 2.1861264015835632e-05, "loss": 1.2631, "step": 6028 }, { "epoch": 0.54, "learning_rate": 2.1854152210015758e-05, "loss": 1.2573, "step": 6029 }, { "epoch": 0.54, "learning_rate": 2.1847040662858347e-05, "loss": 1.2176, "step": 6030 }, { "epoch": 0.54, "learning_rate": 2.183992937494812e-05, "loss": 1.2813, "step": 6031 }, { "epoch": 0.55, "learning_rate": 2.1832818346869802e-05, "loss": 1.2037, "step": 6032 }, { "epoch": 0.55, "learning_rate": 2.182570757920809e-05, "loss": 1.2488, "step": 6033 }, { "epoch": 0.55, "learning_rate": 2.181859707254765e-05, "loss": 1.2484, "step": 6034 }, { "epoch": 0.55, "learning_rate": 2.181148682747314e-05, "loss": 1.2682, "step": 6035 }, { "epoch": 0.55, "learning_rate": 2.180437684456918e-05, "loss": 1.2144, "step": 6036 }, { "epoch": 0.55, "learning_rate": 2.179726712442038e-05, "loss": 1.2468, "step": 6037 }, { "epoch": 0.55, "learning_rate": 2.1790157667611327e-05, "loss": 1.2615, "step": 6038 }, { "epoch": 0.55, "learning_rate": 2.178304847472658e-05, "loss": 1.2397, "step": 6039 }, { "epoch": 0.55, "learning_rate": 2.1775939546350697e-05, "loss": 1.2547, "step": 6040 }, { "epoch": 0.55, "learning_rate": 2.1768830883068178e-05, "loss": 1.1945, "step": 6041 }, { "epoch": 0.55, "learning_rate": 2.1761722485463536e-05, "loss": 1.2772, "step": 6042 }, { "epoch": 0.55, "learning_rate": 2.1754614354121238e-05, "loss": 1.2697, "step": 6043 }, { "epoch": 0.55, "learning_rate": 2.1747506489625743e-05, "loss": 1.2107, "step": 6044 }, { "epoch": 0.55, "learning_rate": 2.1740398892561497e-05, "loss": 1.2471, "step": 6045 }, { "epoch": 0.55, "learning_rate": 2.1733291563512896e-05, "loss": 1.2581, "step": 6046 }, { "epoch": 0.55, "learning_rate": 2.1726184503064324e-05, "loss": 1.2114, "step": 6047 }, { "epoch": 0.55, "learning_rate": 2.1719077711800164e-05, "loss": 1.2718, "step": 6048 }, { "epoch": 0.55, "learning_rate": 2.1711971190304744e-05, "loss": 1.2325, "step": 6049 }, { "epoch": 0.55, "learning_rate": 2.1704864939162415e-05, "loss": 1.2347, "step": 6050 }, { "epoch": 0.55, "learning_rate": 2.1697758958957448e-05, "loss": 1.2527, "step": 6051 }, { "epoch": 0.55, "learning_rate": 2.169065325027413e-05, "loss": 1.24, "step": 6052 }, { "epoch": 0.55, "learning_rate": 2.1683547813696717e-05, "loss": 1.2574, "step": 6053 }, { "epoch": 0.55, "learning_rate": 2.167644264980945e-05, "loss": 1.2712, "step": 6054 }, { "epoch": 0.55, "learning_rate": 2.1669337759196532e-05, "loss": 1.2081, "step": 6055 }, { "epoch": 0.55, "learning_rate": 2.166223314244216e-05, "loss": 1.2902, "step": 6056 }, { "epoch": 0.55, "learning_rate": 2.1655128800130485e-05, "loss": 1.2479, "step": 6057 }, { "epoch": 0.55, "learning_rate": 2.1648024732845665e-05, "loss": 1.2231, "step": 6058 }, { "epoch": 0.55, "learning_rate": 2.1640920941171816e-05, "loss": 1.2523, "step": 6059 }, { "epoch": 0.55, "learning_rate": 2.1633817425693032e-05, "loss": 1.2754, "step": 6060 }, { "epoch": 0.55, "learning_rate": 2.16267141869934e-05, "loss": 1.2322, "step": 6061 }, { "epoch": 0.55, "learning_rate": 2.1619611225656962e-05, "loss": 1.2676, "step": 6062 }, { "epoch": 0.55, "learning_rate": 2.161250854226775e-05, "loss": 1.2178, "step": 6063 }, { "epoch": 0.55, "learning_rate": 2.1605406137409772e-05, "loss": 1.2268, "step": 6064 }, { "epoch": 0.55, "learning_rate": 2.159830401166701e-05, "loss": 1.1926, "step": 6065 }, { "epoch": 0.55, "learning_rate": 2.159120216562344e-05, "loss": 1.2013, "step": 6066 }, { "epoch": 0.55, "learning_rate": 2.158410059986298e-05, "loss": 1.229, "step": 6067 }, { "epoch": 0.55, "learning_rate": 2.157699931496955e-05, "loss": 1.2405, "step": 6068 }, { "epoch": 0.55, "learning_rate": 2.1569898311527045e-05, "loss": 1.2928, "step": 6069 }, { "epoch": 0.55, "learning_rate": 2.1562797590119335e-05, "loss": 1.2862, "step": 6070 }, { "epoch": 0.55, "learning_rate": 2.155569715133028e-05, "loss": 1.2376, "step": 6071 }, { "epoch": 0.55, "learning_rate": 2.1548596995743674e-05, "loss": 1.3053, "step": 6072 }, { "epoch": 0.55, "learning_rate": 2.154149712394333e-05, "loss": 1.2574, "step": 6073 }, { "epoch": 0.55, "learning_rate": 2.1534397536513027e-05, "loss": 1.2498, "step": 6074 }, { "epoch": 0.55, "learning_rate": 2.1527298234036515e-05, "loss": 1.2157, "step": 6075 }, { "epoch": 0.55, "learning_rate": 2.1520199217097534e-05, "loss": 1.266, "step": 6076 }, { "epoch": 0.55, "learning_rate": 2.1513100486279764e-05, "loss": 1.1629, "step": 6077 }, { "epoch": 0.55, "learning_rate": 2.1506002042166902e-05, "loss": 1.2162, "step": 6078 }, { "epoch": 0.55, "learning_rate": 2.1498903885342614e-05, "loss": 1.1513, "step": 6079 }, { "epoch": 0.55, "learning_rate": 2.149180601639052e-05, "loss": 1.2792, "step": 6080 }, { "epoch": 0.55, "learning_rate": 2.1484708435894234e-05, "loss": 1.2568, "step": 6081 }, { "epoch": 0.55, "learning_rate": 2.1477611144437365e-05, "loss": 1.2151, "step": 6082 }, { "epoch": 0.55, "learning_rate": 2.1470514142603447e-05, "loss": 1.2767, "step": 6083 }, { "epoch": 0.55, "learning_rate": 2.146341743097604e-05, "loss": 1.2534, "step": 6084 }, { "epoch": 0.55, "learning_rate": 2.1456321010138646e-05, "loss": 1.1839, "step": 6085 }, { "epoch": 0.55, "learning_rate": 2.144922488067477e-05, "loss": 1.2226, "step": 6086 }, { "epoch": 0.55, "learning_rate": 2.1442129043167874e-05, "loss": 1.2929, "step": 6087 }, { "epoch": 0.55, "learning_rate": 2.1435033498201406e-05, "loss": 1.2158, "step": 6088 }, { "epoch": 0.55, "learning_rate": 2.1427938246358776e-05, "loss": 1.2276, "step": 6089 }, { "epoch": 0.55, "learning_rate": 2.142084328822338e-05, "loss": 1.2152, "step": 6090 }, { "epoch": 0.55, "learning_rate": 2.1413748624378603e-05, "loss": 1.2275, "step": 6091 }, { "epoch": 0.55, "learning_rate": 2.14066542554078e-05, "loss": 1.2327, "step": 6092 }, { "epoch": 0.55, "learning_rate": 2.1399560181894272e-05, "loss": 1.2037, "step": 6093 }, { "epoch": 0.55, "learning_rate": 2.1392466404421325e-05, "loss": 1.1984, "step": 6094 }, { "epoch": 0.55, "learning_rate": 2.1385372923572234e-05, "loss": 1.2711, "step": 6095 }, { "epoch": 0.55, "learning_rate": 2.1378279739930253e-05, "loss": 1.2761, "step": 6096 }, { "epoch": 0.55, "learning_rate": 2.1371186854078616e-05, "loss": 1.2712, "step": 6097 }, { "epoch": 0.55, "learning_rate": 2.1364094266600508e-05, "loss": 1.2149, "step": 6098 }, { "epoch": 0.55, "learning_rate": 2.1357001978079107e-05, "loss": 1.2409, "step": 6099 }, { "epoch": 0.55, "learning_rate": 2.134990998909758e-05, "loss": 1.2457, "step": 6100 }, { "epoch": 0.55, "learning_rate": 2.1342818300239042e-05, "loss": 1.1794, "step": 6101 }, { "epoch": 0.55, "learning_rate": 2.133572691208661e-05, "loss": 1.2119, "step": 6102 }, { "epoch": 0.55, "learning_rate": 2.1328635825223343e-05, "loss": 1.2403, "step": 6103 }, { "epoch": 0.55, "learning_rate": 2.1321545040232302e-05, "loss": 1.2702, "step": 6104 }, { "epoch": 0.55, "learning_rate": 2.1314454557696526e-05, "loss": 1.2231, "step": 6105 }, { "epoch": 0.55, "learning_rate": 2.1307364378199005e-05, "loss": 1.2246, "step": 6106 }, { "epoch": 0.55, "learning_rate": 2.1300274502322727e-05, "loss": 1.2897, "step": 6107 }, { "epoch": 0.55, "learning_rate": 2.129318493065065e-05, "loss": 1.1776, "step": 6108 }, { "epoch": 0.55, "learning_rate": 2.1286095663765693e-05, "loss": 1.2441, "step": 6109 }, { "epoch": 0.55, "learning_rate": 2.1279006702250758e-05, "loss": 1.2263, "step": 6110 }, { "epoch": 0.55, "learning_rate": 2.127191804668873e-05, "loss": 1.1961, "step": 6111 }, { "epoch": 0.55, "learning_rate": 2.126482969766246e-05, "loss": 1.2942, "step": 6112 }, { "epoch": 0.55, "learning_rate": 2.1257741655754792e-05, "loss": 1.1964, "step": 6113 }, { "epoch": 0.55, "learning_rate": 2.125065392154851e-05, "loss": 1.2343, "step": 6114 }, { "epoch": 0.55, "learning_rate": 2.1243566495626392e-05, "loss": 1.2541, "step": 6115 }, { "epoch": 0.55, "learning_rate": 2.1236479378571196e-05, "loss": 1.2066, "step": 6116 }, { "epoch": 0.55, "learning_rate": 2.1229392570965657e-05, "loss": 1.242, "step": 6117 }, { "epoch": 0.55, "learning_rate": 2.1222306073392472e-05, "loss": 1.2257, "step": 6118 }, { "epoch": 0.55, "learning_rate": 2.1215219886434303e-05, "loss": 1.2515, "step": 6119 }, { "epoch": 0.55, "learning_rate": 2.1208134010673812e-05, "loss": 1.2274, "step": 6120 }, { "epoch": 0.55, "learning_rate": 2.1201048446693637e-05, "loss": 1.1887, "step": 6121 }, { "epoch": 0.55, "learning_rate": 2.1193963195076353e-05, "loss": 1.2449, "step": 6122 }, { "epoch": 0.55, "learning_rate": 2.1186878256404556e-05, "loss": 1.2327, "step": 6123 }, { "epoch": 0.55, "learning_rate": 2.1179793631260778e-05, "loss": 1.2114, "step": 6124 }, { "epoch": 0.55, "learning_rate": 2.1172709320227547e-05, "loss": 1.2642, "step": 6125 }, { "epoch": 0.55, "learning_rate": 2.1165625323887362e-05, "loss": 1.2624, "step": 6126 }, { "epoch": 0.55, "learning_rate": 2.1158541642822685e-05, "loss": 1.2042, "step": 6127 }, { "epoch": 0.55, "learning_rate": 2.115145827761598e-05, "loss": 1.2192, "step": 6128 }, { "epoch": 0.55, "learning_rate": 2.1144375228849646e-05, "loss": 1.1791, "step": 6129 }, { "epoch": 0.55, "learning_rate": 2.1137292497106083e-05, "loss": 1.2335, "step": 6130 }, { "epoch": 0.55, "learning_rate": 2.113021008296765e-05, "loss": 1.3045, "step": 6131 }, { "epoch": 0.55, "learning_rate": 2.1123127987016697e-05, "loss": 1.2085, "step": 6132 }, { "epoch": 0.55, "learning_rate": 2.1116046209835538e-05, "loss": 1.2649, "step": 6133 }, { "epoch": 0.55, "learning_rate": 2.1108964752006468e-05, "loss": 1.304, "step": 6134 }, { "epoch": 0.55, "learning_rate": 2.1101883614111733e-05, "loss": 1.2161, "step": 6135 }, { "epoch": 0.55, "learning_rate": 2.1094802796733575e-05, "loss": 1.2403, "step": 6136 }, { "epoch": 0.55, "learning_rate": 2.1087722300454202e-05, "loss": 1.2156, "step": 6137 }, { "epoch": 0.55, "learning_rate": 2.1080642125855805e-05, "loss": 1.2626, "step": 6138 }, { "epoch": 0.55, "learning_rate": 2.1073562273520536e-05, "loss": 1.2115, "step": 6139 }, { "epoch": 0.55, "learning_rate": 2.1066482744030515e-05, "loss": 1.2637, "step": 6140 }, { "epoch": 0.55, "learning_rate": 2.1059403537967852e-05, "loss": 1.2056, "step": 6141 }, { "epoch": 0.55, "learning_rate": 2.105232465591463e-05, "loss": 1.2888, "step": 6142 }, { "epoch": 0.56, "learning_rate": 2.104524609845289e-05, "loss": 1.2511, "step": 6143 }, { "epoch": 0.56, "learning_rate": 2.103816786616467e-05, "loss": 1.2431, "step": 6144 }, { "epoch": 0.56, "learning_rate": 2.1031089959631947e-05, "loss": 1.1926, "step": 6145 }, { "epoch": 0.56, "learning_rate": 2.10240123794367e-05, "loss": 1.2671, "step": 6146 }, { "epoch": 0.56, "learning_rate": 2.1016935126160876e-05, "loss": 1.2259, "step": 6147 }, { "epoch": 0.56, "learning_rate": 2.1009858200386384e-05, "loss": 1.2338, "step": 6148 }, { "epoch": 0.56, "learning_rate": 2.1002781602695126e-05, "loss": 1.2353, "step": 6149 }, { "epoch": 0.56, "learning_rate": 2.0995705333668948e-05, "loss": 1.2326, "step": 6150 }, { "epoch": 0.56, "learning_rate": 2.098862939388969e-05, "loss": 1.2109, "step": 6151 }, { "epoch": 0.56, "learning_rate": 2.098155378393916e-05, "loss": 1.2179, "step": 6152 }, { "epoch": 0.56, "learning_rate": 2.0974478504399145e-05, "loss": 1.2386, "step": 6153 }, { "epoch": 0.56, "learning_rate": 2.0967403555851406e-05, "loss": 1.2488, "step": 6154 }, { "epoch": 0.56, "learning_rate": 2.0960328938877644e-05, "loss": 1.2454, "step": 6155 }, { "epoch": 0.56, "learning_rate": 2.0953254654059583e-05, "loss": 1.2498, "step": 6156 }, { "epoch": 0.56, "learning_rate": 2.0946180701978875e-05, "loss": 1.2195, "step": 6157 }, { "epoch": 0.56, "learning_rate": 2.093910708321718e-05, "loss": 1.2582, "step": 6158 }, { "epoch": 0.56, "learning_rate": 2.0932033798356114e-05, "loss": 1.2356, "step": 6159 }, { "epoch": 0.56, "learning_rate": 2.092496084797727e-05, "loss": 1.2792, "step": 6160 }, { "epoch": 0.56, "learning_rate": 2.0917888232662196e-05, "loss": 1.2263, "step": 6161 }, { "epoch": 0.56, "learning_rate": 2.091081595299243e-05, "loss": 1.205, "step": 6162 }, { "epoch": 0.56, "learning_rate": 2.0903744009549488e-05, "loss": 1.2178, "step": 6163 }, { "epoch": 0.56, "learning_rate": 2.0896672402914843e-05, "loss": 1.2818, "step": 6164 }, { "epoch": 0.56, "learning_rate": 2.088960113366996e-05, "loss": 1.2608, "step": 6165 }, { "epoch": 0.56, "learning_rate": 2.0882530202396243e-05, "loss": 1.2392, "step": 6166 }, { "epoch": 0.56, "learning_rate": 2.0875459609675096e-05, "loss": 1.2785, "step": 6167 }, { "epoch": 0.56, "learning_rate": 2.08683893560879e-05, "loss": 1.2377, "step": 6168 }, { "epoch": 0.56, "learning_rate": 2.0861319442215974e-05, "loss": 1.2423, "step": 6169 }, { "epoch": 0.56, "learning_rate": 2.085424986864065e-05, "loss": 1.2539, "step": 6170 }, { "epoch": 0.56, "learning_rate": 2.08471806359432e-05, "loss": 1.2857, "step": 6171 }, { "epoch": 0.56, "learning_rate": 2.084011174470489e-05, "loss": 1.2346, "step": 6172 }, { "epoch": 0.56, "learning_rate": 2.0833043195506937e-05, "loss": 1.2237, "step": 6173 }, { "epoch": 0.56, "learning_rate": 2.082597498893055e-05, "loss": 1.2412, "step": 6174 }, { "epoch": 0.56, "learning_rate": 2.0818907125556907e-05, "loss": 1.2208, "step": 6175 }, { "epoch": 0.56, "learning_rate": 2.0811839605967136e-05, "loss": 1.1993, "step": 6176 }, { "epoch": 0.56, "learning_rate": 2.080477243074237e-05, "loss": 1.2043, "step": 6177 }, { "epoch": 0.56, "learning_rate": 2.079770560046368e-05, "loss": 1.2459, "step": 6178 }, { "epoch": 0.56, "learning_rate": 2.079063911571213e-05, "loss": 1.2815, "step": 6179 }, { "epoch": 0.56, "learning_rate": 2.0783572977068766e-05, "loss": 1.2548, "step": 6180 }, { "epoch": 0.56, "learning_rate": 2.077650718511458e-05, "loss": 1.2621, "step": 6181 }, { "epoch": 0.56, "learning_rate": 2.0769441740430534e-05, "loss": 1.2415, "step": 6182 }, { "epoch": 0.56, "learning_rate": 2.0762376643597582e-05, "loss": 1.2597, "step": 6183 }, { "epoch": 0.56, "learning_rate": 2.075531189519665e-05, "loss": 1.2112, "step": 6184 }, { "epoch": 0.56, "learning_rate": 2.0748247495808616e-05, "loss": 1.2357, "step": 6185 }, { "epoch": 0.56, "learning_rate": 2.0741183446014344e-05, "loss": 1.2289, "step": 6186 }, { "epoch": 0.56, "learning_rate": 2.073411974639466e-05, "loss": 1.2468, "step": 6187 }, { "epoch": 0.56, "learning_rate": 2.0727056397530363e-05, "loss": 1.2363, "step": 6188 }, { "epoch": 0.56, "learning_rate": 2.0719993400002235e-05, "loss": 1.2901, "step": 6189 }, { "epoch": 0.56, "learning_rate": 2.071293075439101e-05, "loss": 1.2359, "step": 6190 }, { "epoch": 0.56, "learning_rate": 2.0705868461277423e-05, "loss": 1.2291, "step": 6191 }, { "epoch": 0.56, "learning_rate": 2.069880652124213e-05, "loss": 1.2568, "step": 6192 }, { "epoch": 0.56, "learning_rate": 2.0691744934865813e-05, "loss": 1.2388, "step": 6193 }, { "epoch": 0.56, "learning_rate": 2.0684683702729086e-05, "loss": 1.2041, "step": 6194 }, { "epoch": 0.56, "learning_rate": 2.067762282541255e-05, "loss": 1.2005, "step": 6195 }, { "epoch": 0.56, "learning_rate": 2.0670562303496788e-05, "loss": 1.1828, "step": 6196 }, { "epoch": 0.56, "learning_rate": 2.066350213756232e-05, "loss": 1.2453, "step": 6197 }, { "epoch": 0.56, "learning_rate": 2.0656442328189674e-05, "loss": 1.2239, "step": 6198 }, { "epoch": 0.56, "learning_rate": 2.064938287595932e-05, "loss": 1.2177, "step": 6199 }, { "epoch": 0.56, "learning_rate": 2.0642323781451713e-05, "loss": 1.2158, "step": 6200 }, { "epoch": 0.56, "learning_rate": 2.0635265045247294e-05, "loss": 1.1836, "step": 6201 }, { "epoch": 0.56, "learning_rate": 2.062820666792643e-05, "loss": 1.2605, "step": 6202 }, { "epoch": 0.56, "learning_rate": 2.0621148650069496e-05, "loss": 1.2483, "step": 6203 }, { "epoch": 0.56, "learning_rate": 2.0614090992256823e-05, "loss": 1.2558, "step": 6204 }, { "epoch": 0.56, "learning_rate": 2.060703369506873e-05, "loss": 1.2512, "step": 6205 }, { "epoch": 0.56, "learning_rate": 2.059997675908547e-05, "loss": 1.2798, "step": 6206 }, { "epoch": 0.56, "learning_rate": 2.0592920184887315e-05, "loss": 1.2427, "step": 6207 }, { "epoch": 0.56, "learning_rate": 2.0585863973054456e-05, "loss": 1.2537, "step": 6208 }, { "epoch": 0.56, "learning_rate": 2.0578808124167094e-05, "loss": 1.305, "step": 6209 }, { "epoch": 0.56, "learning_rate": 2.057175263880538e-05, "loss": 1.2719, "step": 6210 }, { "epoch": 0.56, "learning_rate": 2.0564697517549438e-05, "loss": 1.2639, "step": 6211 }, { "epoch": 0.56, "learning_rate": 2.055764276097938e-05, "loss": 1.2146, "step": 6212 }, { "epoch": 0.56, "learning_rate": 2.055058836967525e-05, "loss": 1.2532, "step": 6213 }, { "epoch": 0.56, "learning_rate": 2.0543534344217097e-05, "loss": 1.2076, "step": 6214 }, { "epoch": 0.56, "learning_rate": 2.053648068518492e-05, "loss": 1.2299, "step": 6215 }, { "epoch": 0.56, "learning_rate": 2.0529427393158705e-05, "loss": 1.2374, "step": 6216 }, { "epoch": 0.56, "learning_rate": 2.05223744687184e-05, "loss": 1.251, "step": 6217 }, { "epoch": 0.56, "learning_rate": 2.0515321912443902e-05, "loss": 1.2481, "step": 6218 }, { "epoch": 0.56, "learning_rate": 2.0508269724915116e-05, "loss": 1.2213, "step": 6219 }, { "epoch": 0.56, "learning_rate": 2.0501217906711883e-05, "loss": 1.2263, "step": 6220 }, { "epoch": 0.56, "learning_rate": 2.049416645841404e-05, "loss": 1.2379, "step": 6221 }, { "epoch": 0.56, "learning_rate": 2.048711538060138e-05, "loss": 1.2525, "step": 6222 }, { "epoch": 0.56, "learning_rate": 2.048006467385366e-05, "loss": 1.2475, "step": 6223 }, { "epoch": 0.56, "learning_rate": 2.047301433875061e-05, "loss": 1.2116, "step": 6224 }, { "epoch": 0.56, "learning_rate": 2.046596437587194e-05, "loss": 1.2157, "step": 6225 }, { "epoch": 0.56, "learning_rate": 2.0458914785797324e-05, "loss": 1.2686, "step": 6226 }, { "epoch": 0.56, "learning_rate": 2.0451865569106405e-05, "loss": 1.1887, "step": 6227 }, { "epoch": 0.56, "learning_rate": 2.0444816726378784e-05, "loss": 1.2648, "step": 6228 }, { "epoch": 0.56, "learning_rate": 2.043776825819405e-05, "loss": 1.2022, "step": 6229 }, { "epoch": 0.56, "learning_rate": 2.043072016513174e-05, "loss": 1.2047, "step": 6230 }, { "epoch": 0.56, "learning_rate": 2.0423672447771397e-05, "loss": 1.2198, "step": 6231 }, { "epoch": 0.56, "learning_rate": 2.0416625106692482e-05, "loss": 1.1942, "step": 6232 }, { "epoch": 0.56, "learning_rate": 2.0409578142474474e-05, "loss": 1.2245, "step": 6233 }, { "epoch": 0.56, "learning_rate": 2.040253155569678e-05, "loss": 1.2726, "step": 6234 }, { "epoch": 0.56, "learning_rate": 2.0395485346938807e-05, "loss": 1.1898, "step": 6235 }, { "epoch": 0.56, "learning_rate": 2.038843951677991e-05, "loss": 1.2527, "step": 6236 }, { "epoch": 0.56, "learning_rate": 2.0381394065799425e-05, "loss": 1.2245, "step": 6237 }, { "epoch": 0.56, "learning_rate": 2.0374348994576665e-05, "loss": 1.1992, "step": 6238 }, { "epoch": 0.56, "learning_rate": 2.0367304303690875e-05, "loss": 1.214, "step": 6239 }, { "epoch": 0.56, "learning_rate": 2.036025999372132e-05, "loss": 1.1945, "step": 6240 }, { "epoch": 0.56, "learning_rate": 2.0353216065247186e-05, "loss": 1.2563, "step": 6241 }, { "epoch": 0.56, "learning_rate": 2.0346172518847655e-05, "loss": 1.2322, "step": 6242 }, { "epoch": 0.56, "learning_rate": 2.0339129355101893e-05, "loss": 1.2984, "step": 6243 }, { "epoch": 0.56, "learning_rate": 2.0332086574588984e-05, "loss": 1.2311, "step": 6244 }, { "epoch": 0.56, "learning_rate": 2.0325044177888018e-05, "loss": 1.1999, "step": 6245 }, { "epoch": 0.56, "learning_rate": 2.0318002165578048e-05, "loss": 1.2427, "step": 6246 }, { "epoch": 0.56, "learning_rate": 2.03109605382381e-05, "loss": 1.2334, "step": 6247 }, { "epoch": 0.56, "learning_rate": 2.0303919296447155e-05, "loss": 1.2512, "step": 6248 }, { "epoch": 0.56, "learning_rate": 2.029687844078416e-05, "loss": 1.2136, "step": 6249 }, { "epoch": 0.56, "learning_rate": 2.0289837971828047e-05, "loss": 1.2402, "step": 6250 }, { "epoch": 0.56, "learning_rate": 2.0282797890157702e-05, "loss": 1.2481, "step": 6251 }, { "epoch": 0.56, "learning_rate": 2.0275758196351997e-05, "loss": 1.2562, "step": 6252 }, { "epoch": 0.56, "learning_rate": 2.0268718890989753e-05, "loss": 1.1865, "step": 6253 }, { "epoch": 0.57, "learning_rate": 2.0261679974649755e-05, "loss": 1.2023, "step": 6254 }, { "epoch": 0.57, "learning_rate": 2.0254641447910778e-05, "loss": 1.204, "step": 6255 }, { "epoch": 0.57, "learning_rate": 2.0247603311351558e-05, "loss": 1.2226, "step": 6256 }, { "epoch": 0.57, "learning_rate": 2.0240565565550785e-05, "loss": 1.2369, "step": 6257 }, { "epoch": 0.57, "learning_rate": 2.023352821108713e-05, "loss": 1.2548, "step": 6258 }, { "epoch": 0.57, "learning_rate": 2.0226491248539242e-05, "loss": 1.2337, "step": 6259 }, { "epoch": 0.57, "learning_rate": 2.02194546784857e-05, "loss": 1.2133, "step": 6260 }, { "epoch": 0.57, "learning_rate": 2.021241850150509e-05, "loss": 1.2496, "step": 6261 }, { "epoch": 0.57, "learning_rate": 2.0205382718175948e-05, "loss": 1.1719, "step": 6262 }, { "epoch": 0.57, "learning_rate": 2.0198347329076774e-05, "loss": 1.203, "step": 6263 }, { "epoch": 0.57, "learning_rate": 2.019131233478606e-05, "loss": 1.2378, "step": 6264 }, { "epoch": 0.57, "learning_rate": 2.0184277735882234e-05, "loss": 1.1962, "step": 6265 }, { "epoch": 0.57, "learning_rate": 2.0177243532943703e-05, "loss": 1.2542, "step": 6266 }, { "epoch": 0.57, "learning_rate": 2.0170209726548843e-05, "loss": 1.2156, "step": 6267 }, { "epoch": 0.57, "learning_rate": 2.016317631727601e-05, "loss": 1.2366, "step": 6268 }, { "epoch": 0.57, "learning_rate": 2.0156143305703504e-05, "loss": 1.2363, "step": 6269 }, { "epoch": 0.57, "learning_rate": 2.0149110692409608e-05, "loss": 1.216, "step": 6270 }, { "epoch": 0.57, "learning_rate": 2.014207847797256e-05, "loss": 1.2147, "step": 6271 }, { "epoch": 0.57, "learning_rate": 2.013504666297058e-05, "loss": 1.2938, "step": 6272 }, { "epoch": 0.57, "learning_rate": 2.0128015247981854e-05, "loss": 1.171, "step": 6273 }, { "epoch": 0.57, "learning_rate": 2.0120984233584526e-05, "loss": 1.2349, "step": 6274 }, { "epoch": 0.57, "learning_rate": 2.0113953620356697e-05, "loss": 1.1879, "step": 6275 }, { "epoch": 0.57, "learning_rate": 2.010692340887646e-05, "loss": 1.2684, "step": 6276 }, { "epoch": 0.57, "learning_rate": 2.0099893599721866e-05, "loss": 1.2851, "step": 6277 }, { "epoch": 0.57, "learning_rate": 2.009286419347092e-05, "loss": 1.2335, "step": 6278 }, { "epoch": 0.57, "learning_rate": 2.0085835190701617e-05, "loss": 1.2412, "step": 6279 }, { "epoch": 0.57, "learning_rate": 2.007880659199189e-05, "loss": 1.1725, "step": 6280 }, { "epoch": 0.57, "learning_rate": 2.0071778397919666e-05, "loss": 1.2335, "step": 6281 }, { "epoch": 0.57, "learning_rate": 2.0064750609062827e-05, "loss": 1.2286, "step": 6282 }, { "epoch": 0.57, "learning_rate": 2.0057723225999214e-05, "loss": 1.217, "step": 6283 }, { "epoch": 0.57, "learning_rate": 2.005069624930665e-05, "loss": 1.1803, "step": 6284 }, { "epoch": 0.57, "learning_rate": 2.0043669679562924e-05, "loss": 1.1972, "step": 6285 }, { "epoch": 0.57, "learning_rate": 2.003664351734577e-05, "loss": 1.2559, "step": 6286 }, { "epoch": 0.57, "learning_rate": 2.0029617763232907e-05, "loss": 1.1859, "step": 6287 }, { "epoch": 0.57, "learning_rate": 2.0022592417802015e-05, "loss": 1.2006, "step": 6288 }, { "epoch": 0.57, "learning_rate": 2.001556748163076e-05, "loss": 1.228, "step": 6289 }, { "epoch": 0.57, "learning_rate": 2.0008542955296736e-05, "loss": 1.1985, "step": 6290 }, { "epoch": 0.57, "learning_rate": 2.0001518839377532e-05, "loss": 1.1726, "step": 6291 }, { "epoch": 0.57, "learning_rate": 1.9994495134450688e-05, "loss": 1.2874, "step": 6292 }, { "epoch": 0.57, "learning_rate": 1.9987471841093724e-05, "loss": 1.2244, "step": 6293 }, { "epoch": 0.57, "learning_rate": 1.9980448959884123e-05, "loss": 1.2935, "step": 6294 }, { "epoch": 0.57, "learning_rate": 1.9973426491399333e-05, "loss": 1.285, "step": 6295 }, { "epoch": 0.57, "learning_rate": 1.9966404436216747e-05, "loss": 1.2472, "step": 6296 }, { "epoch": 0.57, "learning_rate": 1.995938279491375e-05, "loss": 1.2727, "step": 6297 }, { "epoch": 0.57, "learning_rate": 1.99523615680677e-05, "loss": 1.2851, "step": 6298 }, { "epoch": 0.57, "learning_rate": 1.994534075625589e-05, "loss": 1.2321, "step": 6299 }, { "epoch": 0.57, "learning_rate": 1.9938320360055613e-05, "loss": 1.1708, "step": 6300 }, { "epoch": 0.57, "learning_rate": 1.9931300380044087e-05, "loss": 1.2132, "step": 6301 }, { "epoch": 0.57, "learning_rate": 1.992428081679853e-05, "loss": 1.1953, "step": 6302 }, { "epoch": 0.57, "learning_rate": 1.9917261670896122e-05, "loss": 1.231, "step": 6303 }, { "epoch": 0.57, "learning_rate": 1.9910242942913986e-05, "loss": 1.223, "step": 6304 }, { "epoch": 0.57, "learning_rate": 1.9903224633429248e-05, "loss": 1.267, "step": 6305 }, { "epoch": 0.57, "learning_rate": 1.9896206743018952e-05, "loss": 1.1741, "step": 6306 }, { "epoch": 0.57, "learning_rate": 1.9889189272260156e-05, "loss": 1.2404, "step": 6307 }, { "epoch": 0.57, "learning_rate": 1.9882172221729838e-05, "loss": 1.203, "step": 6308 }, { "epoch": 0.57, "learning_rate": 1.9875155592004984e-05, "loss": 1.3035, "step": 6309 }, { "epoch": 0.57, "learning_rate": 1.9868139383662517e-05, "loss": 1.2223, "step": 6310 }, { "epoch": 0.57, "learning_rate": 1.9861123597279338e-05, "loss": 1.2522, "step": 6311 }, { "epoch": 0.57, "learning_rate": 1.9854108233432304e-05, "loss": 1.2143, "step": 6312 }, { "epoch": 0.57, "learning_rate": 1.984709329269824e-05, "loss": 1.2703, "step": 6313 }, { "epoch": 0.57, "learning_rate": 1.984007877565394e-05, "loss": 1.201, "step": 6314 }, { "epoch": 0.57, "learning_rate": 1.9833064682876176e-05, "loss": 1.2665, "step": 6315 }, { "epoch": 0.57, "learning_rate": 1.982605101494166e-05, "loss": 1.2575, "step": 6316 }, { "epoch": 0.57, "learning_rate": 1.9819037772427074e-05, "loss": 1.1871, "step": 6317 }, { "epoch": 0.57, "learning_rate": 1.981202495590907e-05, "loss": 1.2316, "step": 6318 }, { "epoch": 0.57, "learning_rate": 1.980501256596428e-05, "loss": 1.1662, "step": 6319 }, { "epoch": 0.57, "learning_rate": 1.9798000603169278e-05, "loss": 1.2166, "step": 6320 }, { "epoch": 0.57, "learning_rate": 1.979098906810062e-05, "loss": 1.2412, "step": 6321 }, { "epoch": 0.57, "learning_rate": 1.9783977961334805e-05, "loss": 1.2323, "step": 6322 }, { "epoch": 0.57, "learning_rate": 1.9776967283448314e-05, "loss": 1.1792, "step": 6323 }, { "epoch": 0.57, "learning_rate": 1.9769957035017597e-05, "loss": 1.2158, "step": 6324 }, { "epoch": 0.57, "learning_rate": 1.9762947216619056e-05, "loss": 1.1994, "step": 6325 }, { "epoch": 0.57, "learning_rate": 1.9755937828829067e-05, "loss": 1.256, "step": 6326 }, { "epoch": 0.57, "learning_rate": 1.9748928872223954e-05, "loss": 1.2163, "step": 6327 }, { "epoch": 0.57, "learning_rate": 1.974192034738003e-05, "loss": 1.262, "step": 6328 }, { "epoch": 0.57, "learning_rate": 1.973491225487355e-05, "loss": 1.1936, "step": 6329 }, { "epoch": 0.57, "learning_rate": 1.972790459528075e-05, "loss": 1.2546, "step": 6330 }, { "epoch": 0.57, "learning_rate": 1.9720897369177827e-05, "loss": 1.2749, "step": 6331 }, { "epoch": 0.57, "learning_rate": 1.9713890577140932e-05, "loss": 1.2513, "step": 6332 }, { "epoch": 0.57, "learning_rate": 1.9706884219746197e-05, "loss": 1.2473, "step": 6333 }, { "epoch": 0.57, "learning_rate": 1.9699878297569692e-05, "loss": 1.2497, "step": 6334 }, { "epoch": 0.57, "learning_rate": 1.969287281118748e-05, "loss": 1.2544, "step": 6335 }, { "epoch": 0.57, "learning_rate": 1.9685867761175584e-05, "loss": 1.243, "step": 6336 }, { "epoch": 0.57, "learning_rate": 1.9678863148109977e-05, "loss": 1.1613, "step": 6337 }, { "epoch": 0.57, "learning_rate": 1.967185897256659e-05, "loss": 1.1898, "step": 6338 }, { "epoch": 0.57, "learning_rate": 1.966485523512134e-05, "loss": 1.2055, "step": 6339 }, { "epoch": 0.57, "learning_rate": 1.96578519363501e-05, "loss": 1.2496, "step": 6340 }, { "epoch": 0.57, "learning_rate": 1.9650849076828703e-05, "loss": 1.2181, "step": 6341 }, { "epoch": 0.57, "learning_rate": 1.964384665713296e-05, "loss": 1.2157, "step": 6342 }, { "epoch": 0.57, "learning_rate": 1.9636844677838613e-05, "loss": 1.2582, "step": 6343 }, { "epoch": 0.57, "learning_rate": 1.9629843139521397e-05, "loss": 1.187, "step": 6344 }, { "epoch": 0.57, "learning_rate": 1.9622842042757013e-05, "loss": 1.1986, "step": 6345 }, { "epoch": 0.57, "learning_rate": 1.9615841388121103e-05, "loss": 1.2257, "step": 6346 }, { "epoch": 0.57, "learning_rate": 1.9608841176189297e-05, "loss": 1.2341, "step": 6347 }, { "epoch": 0.57, "learning_rate": 1.9601841407537158e-05, "loss": 1.2092, "step": 6348 }, { "epoch": 0.57, "learning_rate": 1.959484208274025e-05, "loss": 1.2301, "step": 6349 }, { "epoch": 0.57, "learning_rate": 1.9587843202374067e-05, "loss": 1.2666, "step": 6350 }, { "epoch": 0.57, "learning_rate": 1.9580844767014088e-05, "loss": 1.238, "step": 6351 }, { "epoch": 0.57, "learning_rate": 1.957384677723576e-05, "loss": 1.1467, "step": 6352 }, { "epoch": 0.57, "learning_rate": 1.956684923361446e-05, "loss": 1.2025, "step": 6353 }, { "epoch": 0.57, "learning_rate": 1.9559852136725562e-05, "loss": 1.2064, "step": 6354 }, { "epoch": 0.57, "learning_rate": 1.9552855487144387e-05, "loss": 1.2361, "step": 6355 }, { "epoch": 0.57, "learning_rate": 1.9545859285446224e-05, "loss": 1.244, "step": 6356 }, { "epoch": 0.57, "learning_rate": 1.9538863532206338e-05, "loss": 1.2412, "step": 6357 }, { "epoch": 0.57, "learning_rate": 1.9531868227999932e-05, "loss": 1.2139, "step": 6358 }, { "epoch": 0.57, "learning_rate": 1.952487337340218e-05, "loss": 1.1937, "step": 6359 }, { "epoch": 0.57, "learning_rate": 1.9517878968988224e-05, "loss": 1.1797, "step": 6360 }, { "epoch": 0.57, "learning_rate": 1.951088501533318e-05, "loss": 1.2116, "step": 6361 }, { "epoch": 0.57, "learning_rate": 1.95038915130121e-05, "loss": 1.2118, "step": 6362 }, { "epoch": 0.57, "learning_rate": 1.9496898462600035e-05, "loss": 1.2365, "step": 6363 }, { "epoch": 0.58, "learning_rate": 1.9489905864671953e-05, "loss": 1.1965, "step": 6364 }, { "epoch": 0.58, "learning_rate": 1.948291371980282e-05, "loss": 1.2408, "step": 6365 }, { "epoch": 0.58, "learning_rate": 1.947592202856756e-05, "loss": 1.2983, "step": 6366 }, { "epoch": 0.58, "learning_rate": 1.9468930791541042e-05, "loss": 1.2142, "step": 6367 }, { "epoch": 0.58, "learning_rate": 1.9461940009298128e-05, "loss": 1.2426, "step": 6368 }, { "epoch": 0.58, "learning_rate": 1.9454949682413605e-05, "loss": 1.2139, "step": 6369 }, { "epoch": 0.58, "learning_rate": 1.944795981146225e-05, "loss": 1.1967, "step": 6370 }, { "epoch": 0.58, "learning_rate": 1.944097039701879e-05, "loss": 1.2353, "step": 6371 }, { "epoch": 0.58, "learning_rate": 1.9433981439657925e-05, "loss": 1.1613, "step": 6372 }, { "epoch": 0.58, "learning_rate": 1.9426992939954315e-05, "loss": 1.201, "step": 6373 }, { "epoch": 0.58, "learning_rate": 1.9420004898482567e-05, "loss": 1.2395, "step": 6374 }, { "epoch": 0.58, "learning_rate": 1.941301731581727e-05, "loss": 1.2465, "step": 6375 }, { "epoch": 0.58, "learning_rate": 1.940603019253296e-05, "loss": 1.1461, "step": 6376 }, { "epoch": 0.58, "learning_rate": 1.939904352920414e-05, "loss": 1.2818, "step": 6377 }, { "epoch": 0.58, "learning_rate": 1.93920573264053e-05, "loss": 1.2674, "step": 6378 }, { "epoch": 0.58, "learning_rate": 1.938507158471085e-05, "loss": 1.2626, "step": 6379 }, { "epoch": 0.58, "learning_rate": 1.937808630469518e-05, "loss": 1.2361, "step": 6380 }, { "epoch": 0.58, "learning_rate": 1.937110148693265e-05, "loss": 1.2367, "step": 6381 }, { "epoch": 0.58, "learning_rate": 1.9364117131997582e-05, "loss": 1.2514, "step": 6382 }, { "epoch": 0.58, "learning_rate": 1.9357133240464236e-05, "loss": 1.2414, "step": 6383 }, { "epoch": 0.58, "learning_rate": 1.9350149812906878e-05, "loss": 1.2268, "step": 6384 }, { "epoch": 0.58, "learning_rate": 1.934316684989968e-05, "loss": 1.2928, "step": 6385 }, { "epoch": 0.58, "learning_rate": 1.9336184352016826e-05, "loss": 1.2309, "step": 6386 }, { "epoch": 0.58, "learning_rate": 1.9329202319832434e-05, "loss": 1.2443, "step": 6387 }, { "epoch": 0.58, "learning_rate": 1.9322220753920588e-05, "loss": 1.257, "step": 6388 }, { "epoch": 0.58, "learning_rate": 1.9315239654855352e-05, "loss": 1.2186, "step": 6389 }, { "epoch": 0.58, "learning_rate": 1.9308259023210713e-05, "loss": 1.2242, "step": 6390 }, { "epoch": 0.58, "learning_rate": 1.9301278859560657e-05, "loss": 1.198, "step": 6391 }, { "epoch": 0.58, "learning_rate": 1.929429916447911e-05, "loss": 1.2215, "step": 6392 }, { "epoch": 0.58, "learning_rate": 1.928731993853997e-05, "loss": 1.2327, "step": 6393 }, { "epoch": 0.58, "learning_rate": 1.92803411823171e-05, "loss": 1.2693, "step": 6394 }, { "epoch": 0.58, "learning_rate": 1.9273362896384306e-05, "loss": 1.1773, "step": 6395 }, { "epoch": 0.58, "learning_rate": 1.9266385081315374e-05, "loss": 1.2567, "step": 6396 }, { "epoch": 0.58, "learning_rate": 1.9259407737684036e-05, "loss": 1.2126, "step": 6397 }, { "epoch": 0.58, "learning_rate": 1.9252430866064e-05, "loss": 1.2355, "step": 6398 }, { "epoch": 0.58, "learning_rate": 1.9245454467028937e-05, "loss": 1.2959, "step": 6399 }, { "epoch": 0.58, "learning_rate": 1.9238478541152455e-05, "loss": 1.2841, "step": 6400 }, { "epoch": 0.58, "learning_rate": 1.9231503089008142e-05, "loss": 1.2305, "step": 6401 }, { "epoch": 0.58, "learning_rate": 1.9224528111169542e-05, "loss": 1.2159, "step": 6402 }, { "epoch": 0.58, "learning_rate": 1.9217553608210177e-05, "loss": 1.2153, "step": 6403 }, { "epoch": 0.58, "learning_rate": 1.9210579580703496e-05, "loss": 1.2157, "step": 6404 }, { "epoch": 0.58, "learning_rate": 1.920360602922294e-05, "loss": 1.2191, "step": 6405 }, { "epoch": 0.58, "learning_rate": 1.919663295434189e-05, "loss": 1.2059, "step": 6406 }, { "epoch": 0.58, "learning_rate": 1.91896603566337e-05, "loss": 1.2544, "step": 6407 }, { "epoch": 0.58, "learning_rate": 1.918268823667168e-05, "loss": 1.249, "step": 6408 }, { "epoch": 0.58, "learning_rate": 1.9175716595029103e-05, "loss": 1.2379, "step": 6409 }, { "epoch": 0.58, "learning_rate": 1.916874543227921e-05, "loss": 1.2862, "step": 6410 }, { "epoch": 0.58, "learning_rate": 1.9161774748995174e-05, "loss": 1.2566, "step": 6411 }, { "epoch": 0.58, "learning_rate": 1.915480454575017e-05, "loss": 1.2175, "step": 6412 }, { "epoch": 0.58, "learning_rate": 1.9147834823117293e-05, "loss": 1.232, "step": 6413 }, { "epoch": 0.58, "learning_rate": 1.9140865581669627e-05, "loss": 1.2569, "step": 6414 }, { "epoch": 0.58, "learning_rate": 1.913389682198022e-05, "loss": 1.2308, "step": 6415 }, { "epoch": 0.58, "learning_rate": 1.912692854462204e-05, "loss": 1.204, "step": 6416 }, { "epoch": 0.58, "learning_rate": 1.9119960750168066e-05, "loss": 1.1669, "step": 6417 }, { "epoch": 0.58, "learning_rate": 1.9112993439191197e-05, "loss": 1.2438, "step": 6418 }, { "epoch": 0.58, "learning_rate": 1.9106026612264316e-05, "loss": 1.2396, "step": 6419 }, { "epoch": 0.58, "learning_rate": 1.909906026996027e-05, "loss": 1.1966, "step": 6420 }, { "epoch": 0.58, "learning_rate": 1.9092094412851848e-05, "loss": 1.2489, "step": 6421 }, { "epoch": 0.58, "learning_rate": 1.9085129041511798e-05, "loss": 1.1917, "step": 6422 }, { "epoch": 0.58, "learning_rate": 1.9078164156512845e-05, "loss": 1.2175, "step": 6423 }, { "epoch": 0.58, "learning_rate": 1.9071199758427672e-05, "loss": 1.2066, "step": 6424 }, { "epoch": 0.58, "learning_rate": 1.906423584782891e-05, "loss": 1.2758, "step": 6425 }, { "epoch": 0.58, "learning_rate": 1.9057272425289146e-05, "loss": 1.2195, "step": 6426 }, { "epoch": 0.58, "learning_rate": 1.9050309491380947e-05, "loss": 1.283, "step": 6427 }, { "epoch": 0.58, "learning_rate": 1.9043347046676828e-05, "loss": 1.2201, "step": 6428 }, { "epoch": 0.58, "learning_rate": 1.903638509174927e-05, "loss": 1.219, "step": 6429 }, { "epoch": 0.58, "learning_rate": 1.9029423627170706e-05, "loss": 1.3021, "step": 6430 }, { "epoch": 0.58, "learning_rate": 1.9022462653513524e-05, "loss": 1.2153, "step": 6431 }, { "epoch": 0.58, "learning_rate": 1.901550217135008e-05, "loss": 1.1762, "step": 6432 }, { "epoch": 0.58, "learning_rate": 1.9008542181252702e-05, "loss": 1.1951, "step": 6433 }, { "epoch": 0.58, "learning_rate": 1.9001582683793654e-05, "loss": 1.2142, "step": 6434 }, { "epoch": 0.58, "learning_rate": 1.8994623679545168e-05, "loss": 1.2165, "step": 6435 }, { "epoch": 0.58, "learning_rate": 1.8987665169079454e-05, "loss": 1.2292, "step": 6436 }, { "epoch": 0.58, "learning_rate": 1.898070715296864e-05, "loss": 1.2575, "step": 6437 }, { "epoch": 0.58, "learning_rate": 1.897374963178486e-05, "loss": 1.2201, "step": 6438 }, { "epoch": 0.58, "learning_rate": 1.8966792606100166e-05, "loss": 1.1996, "step": 6439 }, { "epoch": 0.58, "learning_rate": 1.8959836076486597e-05, "loss": 1.2327, "step": 6440 }, { "epoch": 0.58, "learning_rate": 1.895288004351616e-05, "loss": 1.2667, "step": 6441 }, { "epoch": 0.58, "learning_rate": 1.894592450776078e-05, "loss": 1.2584, "step": 6442 }, { "epoch": 0.58, "learning_rate": 1.893896946979237e-05, "loss": 1.1837, "step": 6443 }, { "epoch": 0.58, "learning_rate": 1.89320149301828e-05, "loss": 1.2464, "step": 6444 }, { "epoch": 0.58, "learning_rate": 1.89250608895039e-05, "loss": 1.22, "step": 6445 }, { "epoch": 0.58, "learning_rate": 1.891810734832746e-05, "loss": 1.1986, "step": 6446 }, { "epoch": 0.58, "learning_rate": 1.8911154307225203e-05, "loss": 1.25, "step": 6447 }, { "epoch": 0.58, "learning_rate": 1.890420176676885e-05, "loss": 1.265, "step": 6448 }, { "epoch": 0.58, "learning_rate": 1.8897249727530053e-05, "loss": 1.2335, "step": 6449 }, { "epoch": 0.58, "learning_rate": 1.8890298190080447e-05, "loss": 1.1965, "step": 6450 }, { "epoch": 0.58, "learning_rate": 1.8883347154991608e-05, "loss": 1.2366, "step": 6451 }, { "epoch": 0.58, "learning_rate": 1.8876396622835056e-05, "loss": 1.1491, "step": 6452 }, { "epoch": 0.58, "learning_rate": 1.8869446594182298e-05, "loss": 1.2264, "step": 6453 }, { "epoch": 0.58, "learning_rate": 1.8862497069604796e-05, "loss": 1.211, "step": 6454 }, { "epoch": 0.58, "learning_rate": 1.8855548049673956e-05, "loss": 1.2432, "step": 6455 }, { "epoch": 0.58, "learning_rate": 1.8848599534961156e-05, "loss": 1.2329, "step": 6456 }, { "epoch": 0.58, "learning_rate": 1.8841651526037735e-05, "loss": 1.2334, "step": 6457 }, { "epoch": 0.58, "learning_rate": 1.8834704023474958e-05, "loss": 1.2201, "step": 6458 }, { "epoch": 0.58, "learning_rate": 1.8827757027844093e-05, "loss": 1.2485, "step": 6459 }, { "epoch": 0.58, "learning_rate": 1.882081053971633e-05, "loss": 1.2373, "step": 6460 }, { "epoch": 0.58, "learning_rate": 1.881386455966285e-05, "loss": 1.2616, "step": 6461 }, { "epoch": 0.58, "learning_rate": 1.8806919088254775e-05, "loss": 1.2494, "step": 6462 }, { "epoch": 0.58, "learning_rate": 1.8799974126063173e-05, "loss": 1.1917, "step": 6463 }, { "epoch": 0.58, "learning_rate": 1.8793029673659083e-05, "loss": 1.278, "step": 6464 }, { "epoch": 0.58, "learning_rate": 1.8786085731613506e-05, "loss": 1.2502, "step": 6465 }, { "epoch": 0.58, "learning_rate": 1.8779142300497406e-05, "loss": 1.2427, "step": 6466 }, { "epoch": 0.58, "learning_rate": 1.8772199380881694e-05, "loss": 1.1619, "step": 6467 }, { "epoch": 0.58, "learning_rate": 1.876525697333722e-05, "loss": 1.2285, "step": 6468 }, { "epoch": 0.58, "learning_rate": 1.8758315078434828e-05, "loss": 1.1945, "step": 6469 }, { "epoch": 0.58, "learning_rate": 1.875137369674531e-05, "loss": 1.1958, "step": 6470 }, { "epoch": 0.58, "learning_rate": 1.8744432828839402e-05, "loss": 1.1854, "step": 6471 }, { "epoch": 0.58, "learning_rate": 1.873749247528782e-05, "loss": 1.2229, "step": 6472 }, { "epoch": 0.58, "learning_rate": 1.8730552636661197e-05, "loss": 1.2364, "step": 6473 }, { "epoch": 0.58, "learning_rate": 1.872361331353017e-05, "loss": 1.2886, "step": 6474 }, { "epoch": 0.59, "learning_rate": 1.8716674506465308e-05, "loss": 1.2528, "step": 6475 }, { "epoch": 0.59, "learning_rate": 1.870973621603715e-05, "loss": 1.2302, "step": 6476 }, { "epoch": 0.59, "learning_rate": 1.870279844281618e-05, "loss": 1.22, "step": 6477 }, { "epoch": 0.59, "learning_rate": 1.8695861187372848e-05, "loss": 1.2117, "step": 6478 }, { "epoch": 0.59, "learning_rate": 1.8688924450277557e-05, "loss": 1.2344, "step": 6479 }, { "epoch": 0.59, "learning_rate": 1.8681988232100674e-05, "loss": 1.2703, "step": 6480 }, { "epoch": 0.59, "learning_rate": 1.8675052533412506e-05, "loss": 1.2652, "step": 6481 }, { "epoch": 0.59, "learning_rate": 1.8668117354783343e-05, "loss": 1.2659, "step": 6482 }, { "epoch": 0.59, "learning_rate": 1.866118269678343e-05, "loss": 1.2627, "step": 6483 }, { "epoch": 0.59, "learning_rate": 1.8654248559982934e-05, "loss": 1.2068, "step": 6484 }, { "epoch": 0.59, "learning_rate": 1.864731494495201e-05, "loss": 1.215, "step": 6485 }, { "epoch": 0.59, "learning_rate": 1.864038185226077e-05, "loss": 1.2183, "step": 6486 }, { "epoch": 0.59, "learning_rate": 1.8633449282479275e-05, "loss": 1.1967, "step": 6487 }, { "epoch": 0.59, "learning_rate": 1.8626517236177558e-05, "loss": 1.2467, "step": 6488 }, { "epoch": 0.59, "learning_rate": 1.8619585713925568e-05, "loss": 1.211, "step": 6489 }, { "epoch": 0.59, "learning_rate": 1.861265471629325e-05, "loss": 1.2561, "step": 6490 }, { "epoch": 0.59, "learning_rate": 1.8605724243850502e-05, "loss": 1.2401, "step": 6491 }, { "epoch": 0.59, "learning_rate": 1.8598794297167167e-05, "loss": 1.2179, "step": 6492 }, { "epoch": 0.59, "learning_rate": 1.859186487681306e-05, "loss": 1.1946, "step": 6493 }, { "epoch": 0.59, "learning_rate": 1.8584935983357914e-05, "loss": 1.2321, "step": 6494 }, { "epoch": 0.59, "learning_rate": 1.857800761737147e-05, "loss": 1.26, "step": 6495 }, { "epoch": 0.59, "learning_rate": 1.8571079779423396e-05, "loss": 1.2743, "step": 6496 }, { "epoch": 0.59, "learning_rate": 1.856415247008332e-05, "loss": 1.2702, "step": 6497 }, { "epoch": 0.59, "learning_rate": 1.855722568992084e-05, "loss": 1.2531, "step": 6498 }, { "epoch": 0.59, "learning_rate": 1.8550299439505482e-05, "loss": 1.2201, "step": 6499 }, { "epoch": 0.59, "learning_rate": 1.8543373719406758e-05, "loss": 1.2235, "step": 6500 }, { "epoch": 0.59, "learning_rate": 1.8536448530194116e-05, "loss": 1.1793, "step": 6501 }, { "epoch": 0.59, "learning_rate": 1.852952387243698e-05, "loss": 1.2206, "step": 6502 }, { "epoch": 0.59, "learning_rate": 1.8522599746704722e-05, "loss": 1.2201, "step": 6503 }, { "epoch": 0.59, "learning_rate": 1.8515676153566654e-05, "loss": 1.1904, "step": 6504 }, { "epoch": 0.59, "learning_rate": 1.850875309359206e-05, "loss": 1.1558, "step": 6505 }, { "epoch": 0.59, "learning_rate": 1.850183056735018e-05, "loss": 1.2543, "step": 6506 }, { "epoch": 0.59, "learning_rate": 1.8494908575410213e-05, "loss": 1.2814, "step": 6507 }, { "epoch": 0.59, "learning_rate": 1.8487987118341305e-05, "loss": 1.2259, "step": 6508 }, { "epoch": 0.59, "learning_rate": 1.8481066196712572e-05, "loss": 1.2243, "step": 6509 }, { "epoch": 0.59, "learning_rate": 1.8474145811093054e-05, "loss": 1.214, "step": 6510 }, { "epoch": 0.59, "learning_rate": 1.846722596205178e-05, "loss": 1.1853, "step": 6511 }, { "epoch": 0.59, "learning_rate": 1.846030665015773e-05, "loss": 1.2097, "step": 6512 }, { "epoch": 0.59, "learning_rate": 1.8453387875979834e-05, "loss": 1.2187, "step": 6513 }, { "epoch": 0.59, "learning_rate": 1.844646964008697e-05, "loss": 1.1999, "step": 6514 }, { "epoch": 0.59, "learning_rate": 1.843955194304798e-05, "loss": 1.2285, "step": 6515 }, { "epoch": 0.59, "learning_rate": 1.8432634785431656e-05, "loss": 1.2325, "step": 6516 }, { "epoch": 0.59, "learning_rate": 1.842571816780677e-05, "loss": 1.2867, "step": 6517 }, { "epoch": 0.59, "learning_rate": 1.841880209074201e-05, "loss": 1.2826, "step": 6518 }, { "epoch": 0.59, "learning_rate": 1.8411886554806053e-05, "loss": 1.1742, "step": 6519 }, { "epoch": 0.59, "learning_rate": 1.8404971560567504e-05, "loss": 1.2343, "step": 6520 }, { "epoch": 0.59, "learning_rate": 1.839805710859495e-05, "loss": 1.1816, "step": 6521 }, { "epoch": 0.59, "learning_rate": 1.8391143199456918e-05, "loss": 1.2298, "step": 6522 }, { "epoch": 0.59, "learning_rate": 1.8384229833721887e-05, "loss": 1.1906, "step": 6523 }, { "epoch": 0.59, "learning_rate": 1.8377317011958317e-05, "loss": 1.2262, "step": 6524 }, { "epoch": 0.59, "learning_rate": 1.8370404734734577e-05, "loss": 1.1996, "step": 6525 }, { "epoch": 0.59, "learning_rate": 1.8363493002619038e-05, "loss": 1.2713, "step": 6526 }, { "epoch": 0.59, "learning_rate": 1.8356581816179997e-05, "loss": 1.2939, "step": 6527 }, { "epoch": 0.59, "learning_rate": 1.8349671175985717e-05, "loss": 1.2396, "step": 6528 }, { "epoch": 0.59, "learning_rate": 1.8342761082604425e-05, "loss": 1.1998, "step": 6529 }, { "epoch": 0.59, "learning_rate": 1.833585153660428e-05, "loss": 1.2531, "step": 6530 }, { "epoch": 0.59, "learning_rate": 1.8328942538553412e-05, "loss": 1.2436, "step": 6531 }, { "epoch": 0.59, "learning_rate": 1.8322034089019898e-05, "loss": 1.2066, "step": 6532 }, { "epoch": 0.59, "learning_rate": 1.8315126188571785e-05, "loss": 1.2165, "step": 6533 }, { "epoch": 0.59, "learning_rate": 1.830821883777706e-05, "loss": 1.2565, "step": 6534 }, { "epoch": 0.59, "learning_rate": 1.8301312037203676e-05, "loss": 1.2892, "step": 6535 }, { "epoch": 0.59, "learning_rate": 1.8294405787419516e-05, "loss": 1.2474, "step": 6536 }, { "epoch": 0.59, "learning_rate": 1.828750008899245e-05, "loss": 1.2408, "step": 6537 }, { "epoch": 0.59, "learning_rate": 1.8280594942490283e-05, "loss": 1.2485, "step": 6538 }, { "epoch": 0.59, "learning_rate": 1.8273690348480783e-05, "loss": 1.222, "step": 6539 }, { "epoch": 0.59, "learning_rate": 1.8266786307531674e-05, "loss": 1.2336, "step": 6540 }, { "epoch": 0.59, "learning_rate": 1.8259882820210618e-05, "loss": 1.23, "step": 6541 }, { "epoch": 0.59, "learning_rate": 1.8252979887085248e-05, "loss": 1.2697, "step": 6542 }, { "epoch": 0.59, "learning_rate": 1.8246077508723148e-05, "loss": 1.2009, "step": 6543 }, { "epoch": 0.59, "learning_rate": 1.8239175685691854e-05, "loss": 1.1988, "step": 6544 }, { "epoch": 0.59, "learning_rate": 1.8232274418558875e-05, "loss": 1.2206, "step": 6545 }, { "epoch": 0.59, "learning_rate": 1.822537370789163e-05, "loss": 1.2231, "step": 6546 }, { "epoch": 0.59, "learning_rate": 1.821847355425754e-05, "loss": 1.2074, "step": 6547 }, { "epoch": 0.59, "learning_rate": 1.8211573958223938e-05, "loss": 1.2105, "step": 6548 }, { "epoch": 0.59, "learning_rate": 1.8204674920358144e-05, "loss": 1.2578, "step": 6549 }, { "epoch": 0.59, "learning_rate": 1.8197776441227436e-05, "loss": 1.1977, "step": 6550 }, { "epoch": 0.59, "learning_rate": 1.8190878521399014e-05, "loss": 1.2003, "step": 6551 }, { "epoch": 0.59, "learning_rate": 1.8183981161440043e-05, "loss": 1.2646, "step": 6552 }, { "epoch": 0.59, "learning_rate": 1.8177084361917658e-05, "loss": 1.2336, "step": 6553 }, { "epoch": 0.59, "learning_rate": 1.8170188123398936e-05, "loss": 1.201, "step": 6554 }, { "epoch": 0.59, "learning_rate": 1.8163292446450915e-05, "loss": 1.2018, "step": 6555 }, { "epoch": 0.59, "learning_rate": 1.8156397331640575e-05, "loss": 1.2224, "step": 6556 }, { "epoch": 0.59, "learning_rate": 1.814950277953485e-05, "loss": 1.1977, "step": 6557 }, { "epoch": 0.59, "learning_rate": 1.814260879070064e-05, "loss": 1.1614, "step": 6558 }, { "epoch": 0.59, "learning_rate": 1.81357153657048e-05, "loss": 1.2955, "step": 6559 }, { "epoch": 0.59, "learning_rate": 1.8128822505114117e-05, "loss": 1.1951, "step": 6560 }, { "epoch": 0.59, "learning_rate": 1.8121930209495363e-05, "loss": 1.1961, "step": 6561 }, { "epoch": 0.59, "learning_rate": 1.811503847941523e-05, "loss": 1.2315, "step": 6562 }, { "epoch": 0.59, "learning_rate": 1.8108147315440388e-05, "loss": 1.2564, "step": 6563 }, { "epoch": 0.59, "learning_rate": 1.810125671813745e-05, "loss": 1.2852, "step": 6564 }, { "epoch": 0.59, "learning_rate": 1.809436668807298e-05, "loss": 1.2269, "step": 6565 }, { "epoch": 0.59, "learning_rate": 1.8087477225813515e-05, "loss": 1.2283, "step": 6566 }, { "epoch": 0.59, "learning_rate": 1.808058833192551e-05, "loss": 1.2366, "step": 6567 }, { "epoch": 0.59, "learning_rate": 1.8073700006975417e-05, "loss": 1.1875, "step": 6568 }, { "epoch": 0.59, "learning_rate": 1.8066812251529596e-05, "loss": 1.1762, "step": 6569 }, { "epoch": 0.59, "learning_rate": 1.8059925066154386e-05, "loss": 1.2235, "step": 6570 }, { "epoch": 0.59, "learning_rate": 1.8053038451416094e-05, "loss": 1.2281, "step": 6571 }, { "epoch": 0.59, "learning_rate": 1.8046152407880946e-05, "loss": 1.2504, "step": 6572 }, { "epoch": 0.59, "learning_rate": 1.803926693611513e-05, "loss": 1.206, "step": 6573 }, { "epoch": 0.59, "learning_rate": 1.8032382036684798e-05, "loss": 1.2572, "step": 6574 }, { "epoch": 0.59, "learning_rate": 1.8025497710156062e-05, "loss": 1.225, "step": 6575 }, { "epoch": 0.59, "learning_rate": 1.8018613957094965e-05, "loss": 1.2594, "step": 6576 }, { "epoch": 0.59, "learning_rate": 1.8011730778067516e-05, "loss": 1.2202, "step": 6577 }, { "epoch": 0.59, "learning_rate": 1.800484817363966e-05, "loss": 1.2084, "step": 6578 }, { "epoch": 0.59, "learning_rate": 1.7997966144377325e-05, "loss": 1.2267, "step": 6579 }, { "epoch": 0.59, "learning_rate": 1.7991084690846378e-05, "loss": 1.2838, "step": 6580 }, { "epoch": 0.59, "learning_rate": 1.7984203813612633e-05, "loss": 1.2948, "step": 6581 }, { "epoch": 0.59, "learning_rate": 1.7977323513241838e-05, "loss": 1.2918, "step": 6582 }, { "epoch": 0.59, "learning_rate": 1.7970443790299742e-05, "loss": 1.2618, "step": 6583 }, { "epoch": 0.59, "learning_rate": 1.7963564645352004e-05, "loss": 1.2606, "step": 6584 }, { "epoch": 0.59, "learning_rate": 1.795668607896426e-05, "loss": 1.2582, "step": 6585 }, { "epoch": 0.6, "learning_rate": 1.794980809170208e-05, "loss": 1.196, "step": 6586 }, { "epoch": 0.6, "learning_rate": 1.7942930684131015e-05, "loss": 1.2332, "step": 6587 }, { "epoch": 0.6, "learning_rate": 1.7936053856816527e-05, "loss": 1.1916, "step": 6588 }, { "epoch": 0.6, "learning_rate": 1.7929177610324062e-05, "loss": 1.2112, "step": 6589 }, { "epoch": 0.6, "learning_rate": 1.7922301945219006e-05, "loss": 1.2193, "step": 6590 }, { "epoch": 0.6, "learning_rate": 1.7915426862066704e-05, "loss": 1.2131, "step": 6591 }, { "epoch": 0.6, "learning_rate": 1.7908552361432454e-05, "loss": 1.2523, "step": 6592 }, { "epoch": 0.6, "learning_rate": 1.7901678443881492e-05, "loss": 1.2567, "step": 6593 }, { "epoch": 0.6, "learning_rate": 1.789480510997901e-05, "loss": 1.2429, "step": 6594 }, { "epoch": 0.6, "learning_rate": 1.7887932360290166e-05, "loss": 1.2417, "step": 6595 }, { "epoch": 0.6, "learning_rate": 1.7881060195380063e-05, "loss": 1.2504, "step": 6596 }, { "epoch": 0.6, "learning_rate": 1.7874188615813752e-05, "loss": 1.2096, "step": 6597 }, { "epoch": 0.6, "learning_rate": 1.7867317622156238e-05, "loss": 1.2541, "step": 6598 }, { "epoch": 0.6, "learning_rate": 1.786044721497247e-05, "loss": 1.244, "step": 6599 }, { "epoch": 0.6, "learning_rate": 1.7853577394827364e-05, "loss": 1.1907, "step": 6600 }, { "epoch": 0.6, "learning_rate": 1.7846708162285785e-05, "loss": 1.1935, "step": 6601 }, { "epoch": 0.6, "learning_rate": 1.7839839517912548e-05, "loss": 1.2292, "step": 6602 }, { "epoch": 0.6, "learning_rate": 1.7832971462272392e-05, "loss": 1.2016, "step": 6603 }, { "epoch": 0.6, "learning_rate": 1.782610399593005e-05, "loss": 1.1909, "step": 6604 }, { "epoch": 0.6, "learning_rate": 1.7819237119450194e-05, "loss": 1.1924, "step": 6605 }, { "epoch": 0.6, "learning_rate": 1.781237083339743e-05, "loss": 1.2268, "step": 6606 }, { "epoch": 0.6, "learning_rate": 1.7805505138336338e-05, "loss": 1.1988, "step": 6607 }, { "epoch": 0.6, "learning_rate": 1.7798640034831437e-05, "loss": 1.1963, "step": 6608 }, { "epoch": 0.6, "learning_rate": 1.779177552344719e-05, "loss": 1.1674, "step": 6609 }, { "epoch": 0.6, "learning_rate": 1.7784911604748033e-05, "loss": 1.2376, "step": 6610 }, { "epoch": 0.6, "learning_rate": 1.7778048279298333e-05, "loss": 1.2231, "step": 6611 }, { "epoch": 0.6, "learning_rate": 1.7771185547662416e-05, "loss": 1.227, "step": 6612 }, { "epoch": 0.6, "learning_rate": 1.776432341040457e-05, "loss": 1.2007, "step": 6613 }, { "epoch": 0.6, "learning_rate": 1.7757461868089018e-05, "loss": 1.2501, "step": 6614 }, { "epoch": 0.6, "learning_rate": 1.7750600921279928e-05, "loss": 1.2268, "step": 6615 }, { "epoch": 0.6, "learning_rate": 1.774374057054144e-05, "loss": 1.214, "step": 6616 }, { "epoch": 0.6, "learning_rate": 1.7736880816437644e-05, "loss": 1.196, "step": 6617 }, { "epoch": 0.6, "learning_rate": 1.7730021659532564e-05, "loss": 1.2047, "step": 6618 }, { "epoch": 0.6, "learning_rate": 1.7723163100390184e-05, "loss": 1.1789, "step": 6619 }, { "epoch": 0.6, "learning_rate": 1.7716305139574433e-05, "loss": 1.2328, "step": 6620 }, { "epoch": 0.6, "learning_rate": 1.7709447777649208e-05, "loss": 1.2782, "step": 6621 }, { "epoch": 0.6, "learning_rate": 1.7702591015178337e-05, "loss": 1.2452, "step": 6622 }, { "epoch": 0.6, "learning_rate": 1.7695734852725622e-05, "loss": 1.2196, "step": 6623 }, { "epoch": 0.6, "learning_rate": 1.7688879290854768e-05, "loss": 1.2235, "step": 6624 }, { "epoch": 0.6, "learning_rate": 1.768202433012949e-05, "loss": 1.1911, "step": 6625 }, { "epoch": 0.6, "learning_rate": 1.7675169971113413e-05, "loss": 1.2244, "step": 6626 }, { "epoch": 0.6, "learning_rate": 1.7668316214370133e-05, "loss": 1.239, "step": 6627 }, { "epoch": 0.6, "learning_rate": 1.76614630604632e-05, "loss": 1.2305, "step": 6628 }, { "epoch": 0.6, "learning_rate": 1.7654610509956077e-05, "loss": 1.2068, "step": 6629 }, { "epoch": 0.6, "learning_rate": 1.7647758563412224e-05, "loss": 1.2702, "step": 6630 }, { "epoch": 0.6, "learning_rate": 1.7640907221395027e-05, "loss": 1.1997, "step": 6631 }, { "epoch": 0.6, "learning_rate": 1.7634056484467827e-05, "loss": 1.2434, "step": 6632 }, { "epoch": 0.6, "learning_rate": 1.7627206353193914e-05, "loss": 1.2077, "step": 6633 }, { "epoch": 0.6, "learning_rate": 1.7620356828136538e-05, "loss": 1.2005, "step": 6634 }, { "epoch": 0.6, "learning_rate": 1.761350790985888e-05, "loss": 1.2376, "step": 6635 }, { "epoch": 0.6, "learning_rate": 1.760665959892408e-05, "loss": 1.2431, "step": 6636 }, { "epoch": 0.6, "learning_rate": 1.7599811895895236e-05, "loss": 1.2817, "step": 6637 }, { "epoch": 0.6, "learning_rate": 1.759296480133539e-05, "loss": 1.2505, "step": 6638 }, { "epoch": 0.6, "learning_rate": 1.7586118315807537e-05, "loss": 1.2367, "step": 6639 }, { "epoch": 0.6, "learning_rate": 1.757927243987461e-05, "loss": 1.3023, "step": 6640 }, { "epoch": 0.6, "learning_rate": 1.7572427174099503e-05, "loss": 1.256, "step": 6641 }, { "epoch": 0.6, "learning_rate": 1.7565582519045053e-05, "loss": 1.1935, "step": 6642 }, { "epoch": 0.6, "learning_rate": 1.755873847527407e-05, "loss": 1.2129, "step": 6643 }, { "epoch": 0.6, "learning_rate": 1.7551895043349283e-05, "loss": 1.2164, "step": 6644 }, { "epoch": 0.6, "learning_rate": 1.754505222383337e-05, "loss": 1.2249, "step": 6645 }, { "epoch": 0.6, "learning_rate": 1.753821001728898e-05, "loss": 1.2303, "step": 6646 }, { "epoch": 0.6, "learning_rate": 1.7531368424278717e-05, "loss": 1.1644, "step": 6647 }, { "epoch": 0.6, "learning_rate": 1.75245274453651e-05, "loss": 1.2547, "step": 6648 }, { "epoch": 0.6, "learning_rate": 1.7517687081110636e-05, "loss": 1.2428, "step": 6649 }, { "epoch": 0.6, "learning_rate": 1.7510847332077746e-05, "loss": 1.2567, "step": 6650 }, { "epoch": 0.6, "learning_rate": 1.7504008198828827e-05, "loss": 1.2174, "step": 6651 }, { "epoch": 0.6, "learning_rate": 1.7497169681926217e-05, "loss": 1.2599, "step": 6652 }, { "epoch": 0.6, "learning_rate": 1.7490331781932196e-05, "loss": 1.2232, "step": 6653 }, { "epoch": 0.6, "learning_rate": 1.7483494499409013e-05, "loss": 1.2063, "step": 6654 }, { "epoch": 0.6, "learning_rate": 1.747665783491884e-05, "loss": 1.2001, "step": 6655 }, { "epoch": 0.6, "learning_rate": 1.7469821789023815e-05, "loss": 1.2363, "step": 6656 }, { "epoch": 0.6, "learning_rate": 1.7462986362286014e-05, "loss": 1.2667, "step": 6657 }, { "epoch": 0.6, "learning_rate": 1.745615155526748e-05, "loss": 1.2089, "step": 6658 }, { "epoch": 0.6, "learning_rate": 1.74493173685302e-05, "loss": 1.2121, "step": 6659 }, { "epoch": 0.6, "learning_rate": 1.7442483802636096e-05, "loss": 1.2344, "step": 6660 }, { "epoch": 0.6, "learning_rate": 1.743565085814704e-05, "loss": 1.2648, "step": 6661 }, { "epoch": 0.6, "learning_rate": 1.742881853562487e-05, "loss": 1.2469, "step": 6662 }, { "epoch": 0.6, "learning_rate": 1.7421986835631354e-05, "loss": 1.286, "step": 6663 }, { "epoch": 0.6, "learning_rate": 1.741515575872823e-05, "loss": 1.2453, "step": 6664 }, { "epoch": 0.6, "learning_rate": 1.7408325305477173e-05, "loss": 1.2376, "step": 6665 }, { "epoch": 0.6, "learning_rate": 1.7401495476439792e-05, "loss": 1.2221, "step": 6666 }, { "epoch": 0.6, "learning_rate": 1.739466627217767e-05, "loss": 1.2335, "step": 6667 }, { "epoch": 0.6, "learning_rate": 1.7387837693252328e-05, "loss": 1.1807, "step": 6668 }, { "epoch": 0.6, "learning_rate": 1.7381009740225226e-05, "loss": 1.218, "step": 6669 }, { "epoch": 0.6, "learning_rate": 1.73741824136578e-05, "loss": 1.2249, "step": 6670 }, { "epoch": 0.6, "learning_rate": 1.7367355714111397e-05, "loss": 1.2011, "step": 6671 }, { "epoch": 0.6, "learning_rate": 1.736052964214734e-05, "loss": 1.216, "step": 6672 }, { "epoch": 0.6, "learning_rate": 1.7353704198326893e-05, "loss": 1.1955, "step": 6673 }, { "epoch": 0.6, "learning_rate": 1.7346879383211262e-05, "loss": 1.2614, "step": 6674 }, { "epoch": 0.6, "learning_rate": 1.734005519736162e-05, "loss": 1.2374, "step": 6675 }, { "epoch": 0.6, "learning_rate": 1.733323164133906e-05, "loss": 1.2787, "step": 6676 }, { "epoch": 0.6, "learning_rate": 1.7326408715704652e-05, "loss": 1.2326, "step": 6677 }, { "epoch": 0.6, "learning_rate": 1.731958642101938e-05, "loss": 1.2133, "step": 6678 }, { "epoch": 0.6, "learning_rate": 1.7312764757844216e-05, "loss": 1.2377, "step": 6679 }, { "epoch": 0.6, "learning_rate": 1.7305943726740066e-05, "loss": 1.2373, "step": 6680 }, { "epoch": 0.6, "learning_rate": 1.7299123328267752e-05, "loss": 1.2288, "step": 6681 }, { "epoch": 0.6, "learning_rate": 1.7292303562988094e-05, "loss": 1.2321, "step": 6682 }, { "epoch": 0.6, "learning_rate": 1.7285484431461817e-05, "loss": 1.2212, "step": 6683 }, { "epoch": 0.6, "learning_rate": 1.727866593424963e-05, "loss": 1.2388, "step": 6684 }, { "epoch": 0.6, "learning_rate": 1.7271848071912177e-05, "loss": 1.2526, "step": 6685 }, { "epoch": 0.6, "learning_rate": 1.726503084501004e-05, "loss": 1.2496, "step": 6686 }, { "epoch": 0.6, "learning_rate": 1.725821425410374e-05, "loss": 1.2057, "step": 6687 }, { "epoch": 0.6, "learning_rate": 1.7251398299753776e-05, "loss": 1.2734, "step": 6688 }, { "epoch": 0.6, "learning_rate": 1.724458298252058e-05, "loss": 1.2202, "step": 6689 }, { "epoch": 0.6, "learning_rate": 1.7237768302964524e-05, "loss": 1.2797, "step": 6690 }, { "epoch": 0.6, "learning_rate": 1.7230954261645945e-05, "loss": 1.221, "step": 6691 }, { "epoch": 0.6, "learning_rate": 1.72241408591251e-05, "loss": 1.2038, "step": 6692 }, { "epoch": 0.6, "learning_rate": 1.721732809596222e-05, "loss": 1.2217, "step": 6693 }, { "epoch": 0.6, "learning_rate": 1.7210515972717483e-05, "loss": 1.2322, "step": 6694 }, { "epoch": 0.6, "learning_rate": 1.720370448995099e-05, "loss": 1.193, "step": 6695 }, { "epoch": 0.61, "learning_rate": 1.7196893648222817e-05, "loss": 1.2166, "step": 6696 }, { "epoch": 0.61, "learning_rate": 1.7190083448092965e-05, "loss": 1.193, "step": 6697 }, { "epoch": 0.61, "learning_rate": 1.7183273890121396e-05, "loss": 1.219, "step": 6698 }, { "epoch": 0.61, "learning_rate": 1.7176464974868012e-05, "loss": 1.2835, "step": 6699 }, { "epoch": 0.61, "learning_rate": 1.716965670289267e-05, "loss": 1.2033, "step": 6700 }, { "epoch": 0.61, "learning_rate": 1.7162849074755176e-05, "loss": 1.2371, "step": 6701 }, { "epoch": 0.61, "learning_rate": 1.715604209101526e-05, "loss": 1.2744, "step": 6702 }, { "epoch": 0.61, "learning_rate": 1.7149235752232628e-05, "loss": 1.2242, "step": 6703 }, { "epoch": 0.61, "learning_rate": 1.7142430058966916e-05, "loss": 1.182, "step": 6704 }, { "epoch": 0.61, "learning_rate": 1.7135625011777713e-05, "loss": 1.2197, "step": 6705 }, { "epoch": 0.61, "learning_rate": 1.712882061122456e-05, "loss": 1.2146, "step": 6706 }, { "epoch": 0.61, "learning_rate": 1.712201685786693e-05, "loss": 1.2693, "step": 6707 }, { "epoch": 0.61, "learning_rate": 1.711521375226425e-05, "loss": 1.2937, "step": 6708 }, { "epoch": 0.61, "learning_rate": 1.710841129497589e-05, "loss": 1.211, "step": 6709 }, { "epoch": 0.61, "learning_rate": 1.710160948656119e-05, "loss": 1.2481, "step": 6710 }, { "epoch": 0.61, "learning_rate": 1.70948083275794e-05, "loss": 1.2325, "step": 6711 }, { "epoch": 0.61, "learning_rate": 1.7088007818589756e-05, "loss": 1.2393, "step": 6712 }, { "epoch": 0.61, "learning_rate": 1.708120796015139e-05, "loss": 1.2403, "step": 6713 }, { "epoch": 0.61, "learning_rate": 1.7074408752823428e-05, "loss": 1.2492, "step": 6714 }, { "epoch": 0.61, "learning_rate": 1.706761019716493e-05, "loss": 1.2472, "step": 6715 }, { "epoch": 0.61, "learning_rate": 1.7060812293734875e-05, "loss": 1.2484, "step": 6716 }, { "epoch": 0.61, "learning_rate": 1.7054015043092243e-05, "loss": 1.2301, "step": 6717 }, { "epoch": 0.61, "learning_rate": 1.704721844579589e-05, "loss": 1.2212, "step": 6718 }, { "epoch": 0.61, "learning_rate": 1.7040422502404678e-05, "loss": 1.2282, "step": 6719 }, { "epoch": 0.61, "learning_rate": 1.7033627213477388e-05, "loss": 1.2233, "step": 6720 }, { "epoch": 0.61, "learning_rate": 1.7026832579572752e-05, "loss": 1.2497, "step": 6721 }, { "epoch": 0.61, "learning_rate": 1.7020038601249458e-05, "loss": 1.2968, "step": 6722 }, { "epoch": 0.61, "learning_rate": 1.701324527906611e-05, "loss": 1.1897, "step": 6723 }, { "epoch": 0.61, "learning_rate": 1.7006452613581295e-05, "loss": 1.1931, "step": 6724 }, { "epoch": 0.61, "learning_rate": 1.699966060535352e-05, "loss": 1.2, "step": 6725 }, { "epoch": 0.61, "learning_rate": 1.6992869254941247e-05, "loss": 1.2198, "step": 6726 }, { "epoch": 0.61, "learning_rate": 1.69860785629029e-05, "loss": 1.2302, "step": 6727 }, { "epoch": 0.61, "learning_rate": 1.697928852979682e-05, "loss": 1.2462, "step": 6728 }, { "epoch": 0.61, "learning_rate": 1.69724991561813e-05, "loss": 1.1977, "step": 6729 }, { "epoch": 0.61, "learning_rate": 1.6965710442614597e-05, "loss": 1.2569, "step": 6730 }, { "epoch": 0.61, "learning_rate": 1.6958922389654903e-05, "loss": 1.2471, "step": 6731 }, { "epoch": 0.61, "learning_rate": 1.6952134997860346e-05, "loss": 1.247, "step": 6732 }, { "epoch": 0.61, "learning_rate": 1.6945348267789026e-05, "loss": 1.2274, "step": 6733 }, { "epoch": 0.61, "learning_rate": 1.6938562199998946e-05, "loss": 1.2136, "step": 6734 }, { "epoch": 0.61, "learning_rate": 1.6931776795048096e-05, "loss": 1.232, "step": 6735 }, { "epoch": 0.61, "learning_rate": 1.69249920534944e-05, "loss": 1.2291, "step": 6736 }, { "epoch": 0.61, "learning_rate": 1.6918207975895708e-05, "loss": 1.2542, "step": 6737 }, { "epoch": 0.61, "learning_rate": 1.6911424562809854e-05, "loss": 1.2203, "step": 6738 }, { "epoch": 0.61, "learning_rate": 1.6904641814794558e-05, "loss": 1.2717, "step": 6739 }, { "epoch": 0.61, "learning_rate": 1.6897859732407557e-05, "loss": 1.2131, "step": 6740 }, { "epoch": 0.61, "learning_rate": 1.6891078316206467e-05, "loss": 1.2679, "step": 6741 }, { "epoch": 0.61, "learning_rate": 1.6884297566748897e-05, "loss": 1.2819, "step": 6742 }, { "epoch": 0.61, "learning_rate": 1.6877517484592393e-05, "loss": 1.2557, "step": 6743 }, { "epoch": 0.61, "learning_rate": 1.6870738070294412e-05, "loss": 1.2194, "step": 6744 }, { "epoch": 0.61, "learning_rate": 1.6863959324412397e-05, "loss": 1.2009, "step": 6745 }, { "epoch": 0.61, "learning_rate": 1.6857181247503708e-05, "loss": 1.2502, "step": 6746 }, { "epoch": 0.61, "learning_rate": 1.6850403840125673e-05, "loss": 1.2314, "step": 6747 }, { "epoch": 0.61, "learning_rate": 1.684362710283556e-05, "loss": 1.2096, "step": 6748 }, { "epoch": 0.61, "learning_rate": 1.683685103619056e-05, "loss": 1.217, "step": 6749 }, { "epoch": 0.61, "learning_rate": 1.683007564074783e-05, "loss": 1.2245, "step": 6750 }, { "epoch": 0.61, "learning_rate": 1.682330091706446e-05, "loss": 1.2042, "step": 6751 }, { "epoch": 0.61, "learning_rate": 1.6816526865697506e-05, "loss": 1.258, "step": 6752 }, { "epoch": 0.61, "learning_rate": 1.6809753487203943e-05, "loss": 1.1985, "step": 6753 }, { "epoch": 0.61, "learning_rate": 1.680298078214071e-05, "loss": 1.2049, "step": 6754 }, { "epoch": 0.61, "learning_rate": 1.679620875106467e-05, "loss": 1.2235, "step": 6755 }, { "epoch": 0.61, "learning_rate": 1.6789437394532646e-05, "loss": 1.2069, "step": 6756 }, { "epoch": 0.61, "learning_rate": 1.6782666713101414e-05, "loss": 1.256, "step": 6757 }, { "epoch": 0.61, "learning_rate": 1.6775896707327672e-05, "loss": 1.2278, "step": 6758 }, { "epoch": 0.61, "learning_rate": 1.676912737776808e-05, "loss": 1.2558, "step": 6759 }, { "epoch": 0.61, "learning_rate": 1.6762358724979226e-05, "loss": 1.2131, "step": 6760 }, { "epoch": 0.61, "learning_rate": 1.675559074951766e-05, "loss": 1.2349, "step": 6761 }, { "epoch": 0.61, "learning_rate": 1.674882345193986e-05, "loss": 1.2192, "step": 6762 }, { "epoch": 0.61, "learning_rate": 1.6742056832802265e-05, "loss": 1.2231, "step": 6763 }, { "epoch": 0.61, "learning_rate": 1.6735290892661255e-05, "loss": 1.2452, "step": 6764 }, { "epoch": 0.61, "learning_rate": 1.6728525632073132e-05, "loss": 1.2407, "step": 6765 }, { "epoch": 0.61, "learning_rate": 1.672176105159417e-05, "loss": 1.2555, "step": 6766 }, { "epoch": 0.61, "learning_rate": 1.6714997151780573e-05, "loss": 1.2656, "step": 6767 }, { "epoch": 0.61, "learning_rate": 1.6708233933188494e-05, "loss": 1.2252, "step": 6768 }, { "epoch": 0.61, "learning_rate": 1.6701471396374035e-05, "loss": 1.2388, "step": 6769 }, { "epoch": 0.61, "learning_rate": 1.6694709541893224e-05, "loss": 1.2726, "step": 6770 }, { "epoch": 0.61, "learning_rate": 1.668794837030204e-05, "loss": 1.2342, "step": 6771 }, { "epoch": 0.61, "learning_rate": 1.668118788215642e-05, "loss": 1.2011, "step": 6772 }, { "epoch": 0.61, "learning_rate": 1.6674428078012237e-05, "loss": 1.1807, "step": 6773 }, { "epoch": 0.61, "learning_rate": 1.6667668958425297e-05, "loss": 1.1965, "step": 6774 }, { "epoch": 0.61, "learning_rate": 1.6660910523951368e-05, "loss": 1.295, "step": 6775 }, { "epoch": 0.61, "learning_rate": 1.6654152775146134e-05, "loss": 1.2668, "step": 6776 }, { "epoch": 0.61, "learning_rate": 1.6647395712565256e-05, "loss": 1.2096, "step": 6777 }, { "epoch": 0.61, "learning_rate": 1.6640639336764323e-05, "loss": 1.221, "step": 6778 }, { "epoch": 0.61, "learning_rate": 1.663388364829887e-05, "loss": 1.2201, "step": 6779 }, { "epoch": 0.61, "learning_rate": 1.6627128647724352e-05, "loss": 1.2872, "step": 6780 }, { "epoch": 0.61, "learning_rate": 1.6620374335596205e-05, "loss": 1.2397, "step": 6781 }, { "epoch": 0.61, "learning_rate": 1.6613620712469798e-05, "loss": 1.2376, "step": 6782 }, { "epoch": 0.61, "learning_rate": 1.660686777890042e-05, "loss": 1.3068, "step": 6783 }, { "epoch": 0.61, "learning_rate": 1.6600115535443328e-05, "loss": 1.1819, "step": 6784 }, { "epoch": 0.61, "learning_rate": 1.6593363982653732e-05, "loss": 1.28, "step": 6785 }, { "epoch": 0.61, "learning_rate": 1.6586613121086735e-05, "loss": 1.2168, "step": 6786 }, { "epoch": 0.61, "learning_rate": 1.657986295129744e-05, "loss": 1.1978, "step": 6787 }, { "epoch": 0.61, "learning_rate": 1.6573113473840864e-05, "loss": 1.2155, "step": 6788 }, { "epoch": 0.61, "learning_rate": 1.6566364689271966e-05, "loss": 1.1904, "step": 6789 }, { "epoch": 0.61, "learning_rate": 1.655961659814567e-05, "loss": 1.2725, "step": 6790 }, { "epoch": 0.61, "learning_rate": 1.655286920101681e-05, "loss": 1.2551, "step": 6791 }, { "epoch": 0.61, "learning_rate": 1.6546122498440184e-05, "loss": 1.2252, "step": 6792 }, { "epoch": 0.61, "learning_rate": 1.6539376490970533e-05, "loss": 1.2178, "step": 6793 }, { "epoch": 0.61, "learning_rate": 1.6532631179162535e-05, "loss": 1.2461, "step": 6794 }, { "epoch": 0.61, "learning_rate": 1.652588656357082e-05, "loss": 1.2054, "step": 6795 }, { "epoch": 0.61, "learning_rate": 1.6519142644749944e-05, "loss": 1.1616, "step": 6796 }, { "epoch": 0.61, "learning_rate": 1.6512399423254417e-05, "loss": 1.2385, "step": 6797 }, { "epoch": 0.61, "learning_rate": 1.6505656899638687e-05, "loss": 1.21, "step": 6798 }, { "epoch": 0.61, "learning_rate": 1.6498915074457157e-05, "loss": 1.2722, "step": 6799 }, { "epoch": 0.61, "learning_rate": 1.6492173948264166e-05, "loss": 1.2454, "step": 6800 }, { "epoch": 0.61, "learning_rate": 1.6485433521613975e-05, "loss": 1.2121, "step": 6801 }, { "epoch": 0.61, "learning_rate": 1.647869379506081e-05, "loss": 1.2059, "step": 6802 }, { "epoch": 0.61, "learning_rate": 1.6471954769158854e-05, "loss": 1.1591, "step": 6803 }, { "epoch": 0.61, "learning_rate": 1.646521644446219e-05, "loss": 1.2133, "step": 6804 }, { "epoch": 0.61, "learning_rate": 1.645847882152488e-05, "loss": 1.211, "step": 6805 }, { "epoch": 0.61, "learning_rate": 1.64517419009009e-05, "loss": 1.216, "step": 6806 }, { "epoch": 0.62, "learning_rate": 1.6445005683144198e-05, "loss": 1.2203, "step": 6807 }, { "epoch": 0.62, "learning_rate": 1.6438270168808645e-05, "loss": 1.1992, "step": 6808 }, { "epoch": 0.62, "learning_rate": 1.6431535358448052e-05, "loss": 1.2061, "step": 6809 }, { "epoch": 0.62, "learning_rate": 1.6424801252616184e-05, "loss": 1.2631, "step": 6810 }, { "epoch": 0.62, "learning_rate": 1.641806785186675e-05, "loss": 1.2769, "step": 6811 }, { "epoch": 0.62, "learning_rate": 1.6411335156753376e-05, "loss": 1.2513, "step": 6812 }, { "epoch": 0.62, "learning_rate": 1.640460316782966e-05, "loss": 1.1811, "step": 6813 }, { "epoch": 0.62, "learning_rate": 1.639787188564912e-05, "loss": 1.1826, "step": 6814 }, { "epoch": 0.62, "learning_rate": 1.6391141310765233e-05, "loss": 1.1951, "step": 6815 }, { "epoch": 0.62, "learning_rate": 1.638441144373141e-05, "loss": 1.253, "step": 6816 }, { "epoch": 0.62, "learning_rate": 1.6377682285101005e-05, "loss": 1.2249, "step": 6817 }, { "epoch": 0.62, "learning_rate": 1.6370953835427306e-05, "loss": 1.197, "step": 6818 }, { "epoch": 0.62, "learning_rate": 1.6364226095263545e-05, "loss": 1.2044, "step": 6819 }, { "epoch": 0.62, "learning_rate": 1.6357499065162917e-05, "loss": 1.2066, "step": 6820 }, { "epoch": 0.62, "learning_rate": 1.635077274567854e-05, "loss": 1.2566, "step": 6821 }, { "epoch": 0.62, "learning_rate": 1.6344047137363455e-05, "loss": 1.2417, "step": 6822 }, { "epoch": 0.62, "learning_rate": 1.6337322240770674e-05, "loss": 1.1795, "step": 6823 }, { "epoch": 0.62, "learning_rate": 1.6330598056453157e-05, "loss": 1.1583, "step": 6824 }, { "epoch": 0.62, "learning_rate": 1.6323874584963766e-05, "loss": 1.2246, "step": 6825 }, { "epoch": 0.62, "learning_rate": 1.631715182685535e-05, "loss": 1.2614, "step": 6826 }, { "epoch": 0.62, "learning_rate": 1.6310429782680657e-05, "loss": 1.2026, "step": 6827 }, { "epoch": 0.62, "learning_rate": 1.6303708452992407e-05, "loss": 1.2237, "step": 6828 }, { "epoch": 0.62, "learning_rate": 1.6296987838343254e-05, "loss": 1.2451, "step": 6829 }, { "epoch": 0.62, "learning_rate": 1.6290267939285784e-05, "loss": 1.2275, "step": 6830 }, { "epoch": 0.62, "learning_rate": 1.628354875637254e-05, "loss": 1.2376, "step": 6831 }, { "epoch": 0.62, "learning_rate": 1.627683029015598e-05, "loss": 1.2, "step": 6832 }, { "epoch": 0.62, "learning_rate": 1.627011254118853e-05, "loss": 1.237, "step": 6833 }, { "epoch": 0.62, "learning_rate": 1.6263395510022543e-05, "loss": 1.2237, "step": 6834 }, { "epoch": 0.62, "learning_rate": 1.6256679197210316e-05, "loss": 1.2191, "step": 6835 }, { "epoch": 0.62, "learning_rate": 1.62499636033041e-05, "loss": 1.2439, "step": 6836 }, { "epoch": 0.62, "learning_rate": 1.6243248728856058e-05, "loss": 1.2791, "step": 6837 }, { "epoch": 0.62, "learning_rate": 1.6236534574418323e-05, "loss": 1.2015, "step": 6838 }, { "epoch": 0.62, "learning_rate": 1.6229821140542938e-05, "loss": 1.2676, "step": 6839 }, { "epoch": 0.62, "learning_rate": 1.622310842778192e-05, "loss": 1.248, "step": 6840 }, { "epoch": 0.62, "learning_rate": 1.6216396436687208e-05, "loss": 1.2003, "step": 6841 }, { "epoch": 0.62, "learning_rate": 1.620968516781069e-05, "loss": 1.2284, "step": 6842 }, { "epoch": 0.62, "learning_rate": 1.6202974621704175e-05, "loss": 1.2008, "step": 6843 }, { "epoch": 0.62, "learning_rate": 1.6196264798919437e-05, "loss": 1.2568, "step": 6844 }, { "epoch": 0.62, "learning_rate": 1.6189555700008186e-05, "loss": 1.2527, "step": 6845 }, { "epoch": 0.62, "learning_rate": 1.6182847325522055e-05, "loss": 1.2217, "step": 6846 }, { "epoch": 0.62, "learning_rate": 1.6176139676012647e-05, "loss": 1.1548, "step": 6847 }, { "epoch": 0.62, "learning_rate": 1.6169432752031466e-05, "loss": 1.2417, "step": 6848 }, { "epoch": 0.62, "learning_rate": 1.6162726554129986e-05, "loss": 1.2499, "step": 6849 }, { "epoch": 0.62, "learning_rate": 1.6156021082859626e-05, "loss": 1.2091, "step": 6850 }, { "epoch": 0.62, "learning_rate": 1.614931633877172e-05, "loss": 1.2083, "step": 6851 }, { "epoch": 0.62, "learning_rate": 1.6142612322417565e-05, "loss": 1.2418, "step": 6852 }, { "epoch": 0.62, "learning_rate": 1.613590903434838e-05, "loss": 1.252, "step": 6853 }, { "epoch": 0.62, "learning_rate": 1.6129206475115333e-05, "loss": 1.2306, "step": 6854 }, { "epoch": 0.62, "learning_rate": 1.612250464526953e-05, "loss": 1.2262, "step": 6855 }, { "epoch": 0.62, "learning_rate": 1.6115803545362028e-05, "loss": 1.2249, "step": 6856 }, { "epoch": 0.62, "learning_rate": 1.6109103175943813e-05, "loss": 1.2228, "step": 6857 }, { "epoch": 0.62, "learning_rate": 1.6102403537565803e-05, "loss": 1.2633, "step": 6858 }, { "epoch": 0.62, "learning_rate": 1.609570463077887e-05, "loss": 1.2338, "step": 6859 }, { "epoch": 0.62, "learning_rate": 1.608900645613382e-05, "loss": 1.2407, "step": 6860 }, { "epoch": 0.62, "learning_rate": 1.60823090141814e-05, "loss": 1.1235, "step": 6861 }, { "epoch": 0.62, "learning_rate": 1.607561230547231e-05, "loss": 1.244, "step": 6862 }, { "epoch": 0.62, "learning_rate": 1.6068916330557164e-05, "loss": 1.234, "step": 6863 }, { "epoch": 0.62, "learning_rate": 1.6062221089986525e-05, "loss": 1.2355, "step": 6864 }, { "epoch": 0.62, "learning_rate": 1.60555265843109e-05, "loss": 1.2133, "step": 6865 }, { "epoch": 0.62, "learning_rate": 1.604883281408074e-05, "loss": 1.2449, "step": 6866 }, { "epoch": 0.62, "learning_rate": 1.6042139779846423e-05, "loss": 1.2655, "step": 6867 }, { "epoch": 0.62, "learning_rate": 1.603544748215829e-05, "loss": 1.2675, "step": 6868 }, { "epoch": 0.62, "learning_rate": 1.6028755921566583e-05, "loss": 1.1925, "step": 6869 }, { "epoch": 0.62, "learning_rate": 1.6022065098621514e-05, "loss": 1.2493, "step": 6870 }, { "epoch": 0.62, "learning_rate": 1.6015375013873228e-05, "loss": 1.2268, "step": 6871 }, { "epoch": 0.62, "learning_rate": 1.6008685667871805e-05, "loss": 1.2379, "step": 6872 }, { "epoch": 0.62, "learning_rate": 1.6001997061167277e-05, "loss": 1.2152, "step": 6873 }, { "epoch": 0.62, "learning_rate": 1.599530919430958e-05, "loss": 1.2102, "step": 6874 }, { "epoch": 0.62, "learning_rate": 1.598862206784863e-05, "loss": 1.2447, "step": 6875 }, { "epoch": 0.62, "learning_rate": 1.5981935682334264e-05, "loss": 1.2734, "step": 6876 }, { "epoch": 0.62, "learning_rate": 1.5975250038316256e-05, "loss": 1.2208, "step": 6877 }, { "epoch": 0.62, "learning_rate": 1.5968565136344335e-05, "loss": 1.2224, "step": 6878 }, { "epoch": 0.62, "learning_rate": 1.5961880976968136e-05, "loss": 1.25, "step": 6879 }, { "epoch": 0.62, "learning_rate": 1.595519756073727e-05, "loss": 1.1734, "step": 6880 }, { "epoch": 0.62, "learning_rate": 1.594851488820126e-05, "loss": 1.2316, "step": 6881 }, { "epoch": 0.62, "learning_rate": 1.5941832959909582e-05, "loss": 1.2384, "step": 6882 }, { "epoch": 0.62, "learning_rate": 1.5935151776411662e-05, "loss": 1.2666, "step": 6883 }, { "epoch": 0.62, "learning_rate": 1.592847133825683e-05, "loss": 1.2479, "step": 6884 }, { "epoch": 0.62, "learning_rate": 1.5921791645994376e-05, "loss": 1.2138, "step": 6885 }, { "epoch": 0.62, "learning_rate": 1.591511270017353e-05, "loss": 1.2092, "step": 6886 }, { "epoch": 0.62, "learning_rate": 1.5908434501343468e-05, "loss": 1.2144, "step": 6887 }, { "epoch": 0.62, "learning_rate": 1.5901757050053284e-05, "loss": 1.2, "step": 6888 }, { "epoch": 0.62, "learning_rate": 1.589508034685203e-05, "loss": 1.25, "step": 6889 }, { "epoch": 0.62, "learning_rate": 1.5888404392288676e-05, "loss": 1.2324, "step": 6890 }, { "epoch": 0.62, "learning_rate": 1.5881729186912145e-05, "loss": 1.2465, "step": 6891 }, { "epoch": 0.62, "learning_rate": 1.5875054731271306e-05, "loss": 1.2097, "step": 6892 }, { "epoch": 0.62, "learning_rate": 1.5868381025914936e-05, "loss": 1.1911, "step": 6893 }, { "epoch": 0.62, "learning_rate": 1.58617080713918e-05, "loss": 1.2171, "step": 6894 }, { "epoch": 0.62, "learning_rate": 1.585503586825054e-05, "loss": 1.2488, "step": 6895 }, { "epoch": 0.62, "learning_rate": 1.5848364417039786e-05, "loss": 1.2636, "step": 6896 }, { "epoch": 0.62, "learning_rate": 1.5841693718308075e-05, "loss": 1.2343, "step": 6897 }, { "epoch": 0.62, "learning_rate": 1.5835023772603908e-05, "loss": 1.2102, "step": 6898 }, { "epoch": 0.62, "learning_rate": 1.5828354580475714e-05, "loss": 1.244, "step": 6899 }, { "epoch": 0.62, "learning_rate": 1.5821686142471832e-05, "loss": 1.1991, "step": 6900 }, { "epoch": 0.62, "learning_rate": 1.5815018459140592e-05, "loss": 1.2293, "step": 6901 }, { "epoch": 0.62, "learning_rate": 1.5808351531030215e-05, "loss": 1.2397, "step": 6902 }, { "epoch": 0.62, "learning_rate": 1.580168535868889e-05, "loss": 1.306, "step": 6903 }, { "epoch": 0.62, "learning_rate": 1.5795019942664735e-05, "loss": 1.2501, "step": 6904 }, { "epoch": 0.62, "learning_rate": 1.5788355283505793e-05, "loss": 1.2223, "step": 6905 }, { "epoch": 0.62, "learning_rate": 1.5781691381760054e-05, "loss": 1.2402, "step": 6906 }, { "epoch": 0.62, "learning_rate": 1.5775028237975452e-05, "loss": 1.2003, "step": 6907 }, { "epoch": 0.62, "learning_rate": 1.5768365852699863e-05, "loss": 1.2585, "step": 6908 }, { "epoch": 0.62, "learning_rate": 1.576170422648108e-05, "loss": 1.2586, "step": 6909 }, { "epoch": 0.62, "learning_rate": 1.575504335986685e-05, "loss": 1.2327, "step": 6910 }, { "epoch": 0.62, "learning_rate": 1.574838325340484e-05, "loss": 1.2821, "step": 6911 }, { "epoch": 0.62, "learning_rate": 1.574172390764268e-05, "loss": 1.194, "step": 6912 }, { "epoch": 0.62, "learning_rate": 1.573506532312793e-05, "loss": 1.2032, "step": 6913 }, { "epoch": 0.62, "learning_rate": 1.5728407500408064e-05, "loss": 1.2653, "step": 6914 }, { "epoch": 0.62, "learning_rate": 1.5721750440030532e-05, "loss": 1.2446, "step": 6915 }, { "epoch": 0.62, "learning_rate": 1.5715094142542674e-05, "loss": 1.2545, "step": 6916 }, { "epoch": 0.62, "learning_rate": 1.5708438608491814e-05, "loss": 1.2084, "step": 6917 }, { "epoch": 0.63, "learning_rate": 1.5701783838425188e-05, "loss": 1.2346, "step": 6918 }, { "epoch": 0.63, "learning_rate": 1.569512983288997e-05, "loss": 1.2455, "step": 6919 }, { "epoch": 0.63, "learning_rate": 1.5688476592433287e-05, "loss": 1.2294, "step": 6920 }, { "epoch": 0.63, "learning_rate": 1.568182411760218e-05, "loss": 1.2132, "step": 6921 }, { "epoch": 0.63, "learning_rate": 1.567517240894364e-05, "loss": 1.3043, "step": 6922 }, { "epoch": 0.63, "learning_rate": 1.566852146700459e-05, "loss": 1.2269, "step": 6923 }, { "epoch": 0.63, "learning_rate": 1.56618712923319e-05, "loss": 1.2838, "step": 6924 }, { "epoch": 0.63, "learning_rate": 1.5655221885472383e-05, "loss": 1.2276, "step": 6925 }, { "epoch": 0.63, "learning_rate": 1.564857324697275e-05, "loss": 1.2332, "step": 6926 }, { "epoch": 0.63, "learning_rate": 1.5641925377379685e-05, "loss": 1.2093, "step": 6927 }, { "epoch": 0.63, "learning_rate": 1.5635278277239802e-05, "loss": 1.2616, "step": 6928 }, { "epoch": 0.63, "learning_rate": 1.562863194709965e-05, "loss": 1.2107, "step": 6929 }, { "epoch": 0.63, "learning_rate": 1.5621986387505717e-05, "loss": 1.2198, "step": 6930 }, { "epoch": 0.63, "learning_rate": 1.561534159900441e-05, "loss": 1.2751, "step": 6931 }, { "epoch": 0.63, "learning_rate": 1.5608697582142086e-05, "loss": 1.2555, "step": 6932 }, { "epoch": 0.63, "learning_rate": 1.5602054337465053e-05, "loss": 1.222, "step": 6933 }, { "epoch": 0.63, "learning_rate": 1.559541186551954e-05, "loss": 1.2479, "step": 6934 }, { "epoch": 0.63, "learning_rate": 1.55887701668517e-05, "loss": 1.1948, "step": 6935 }, { "epoch": 0.63, "learning_rate": 1.558212924200766e-05, "loss": 1.2284, "step": 6936 }, { "epoch": 0.63, "learning_rate": 1.557548909153343e-05, "loss": 1.1988, "step": 6937 }, { "epoch": 0.63, "learning_rate": 1.556884971597501e-05, "loss": 1.2523, "step": 6938 }, { "epoch": 0.63, "learning_rate": 1.55622111158783e-05, "loss": 1.2072, "step": 6939 }, { "epoch": 0.63, "learning_rate": 1.555557329178915e-05, "loss": 1.2267, "step": 6940 }, { "epoch": 0.63, "learning_rate": 1.5548936244253356e-05, "loss": 1.2119, "step": 6941 }, { "epoch": 0.63, "learning_rate": 1.5542299973816628e-05, "loss": 1.1919, "step": 6942 }, { "epoch": 0.63, "learning_rate": 1.553566448102462e-05, "loss": 1.2472, "step": 6943 }, { "epoch": 0.63, "learning_rate": 1.5529029766422925e-05, "loss": 1.2502, "step": 6944 }, { "epoch": 0.63, "learning_rate": 1.5522395830557084e-05, "loss": 1.2361, "step": 6945 }, { "epoch": 0.63, "learning_rate": 1.551576267397256e-05, "loss": 1.2384, "step": 6946 }, { "epoch": 0.63, "learning_rate": 1.550913029721475e-05, "loss": 1.2519, "step": 6947 }, { "epoch": 0.63, "learning_rate": 1.5502498700828983e-05, "loss": 1.1939, "step": 6948 }, { "epoch": 0.63, "learning_rate": 1.549586788536054e-05, "loss": 1.2265, "step": 6949 }, { "epoch": 0.63, "learning_rate": 1.548923785135463e-05, "loss": 1.183, "step": 6950 }, { "epoch": 0.63, "learning_rate": 1.5482608599356402e-05, "loss": 1.2395, "step": 6951 }, { "epoch": 0.63, "learning_rate": 1.547598012991092e-05, "loss": 1.2519, "step": 6952 }, { "epoch": 0.63, "learning_rate": 1.5469352443563213e-05, "loss": 1.1871, "step": 6953 }, { "epoch": 0.63, "learning_rate": 1.5462725540858227e-05, "loss": 1.2325, "step": 6954 }, { "epoch": 0.63, "learning_rate": 1.5456099422340853e-05, "loss": 1.2068, "step": 6955 }, { "epoch": 0.63, "learning_rate": 1.5449474088555916e-05, "loss": 1.2222, "step": 6956 }, { "epoch": 0.63, "learning_rate": 1.544284954004816e-05, "loss": 1.2196, "step": 6957 }, { "epoch": 0.63, "learning_rate": 1.5436225777362286e-05, "loss": 1.236, "step": 6958 }, { "epoch": 0.63, "learning_rate": 1.5429602801042932e-05, "loss": 1.1698, "step": 6959 }, { "epoch": 0.63, "learning_rate": 1.5422980611634644e-05, "loss": 1.1951, "step": 6960 }, { "epoch": 0.63, "learning_rate": 1.5416359209681934e-05, "loss": 1.2289, "step": 6961 }, { "epoch": 0.63, "learning_rate": 1.540973859572924e-05, "loss": 1.1956, "step": 6962 }, { "epoch": 0.63, "learning_rate": 1.5403118770320925e-05, "loss": 1.257, "step": 6963 }, { "epoch": 0.63, "learning_rate": 1.539649973400128e-05, "loss": 1.2035, "step": 6964 }, { "epoch": 0.63, "learning_rate": 1.5389881487314566e-05, "loss": 1.2035, "step": 6965 }, { "epoch": 0.63, "learning_rate": 1.538326403080495e-05, "loss": 1.2475, "step": 6966 }, { "epoch": 0.63, "learning_rate": 1.5376647365016553e-05, "loss": 1.2113, "step": 6967 }, { "epoch": 0.63, "learning_rate": 1.5370031490493403e-05, "loss": 1.259, "step": 6968 }, { "epoch": 0.63, "learning_rate": 1.5363416407779486e-05, "loss": 1.2112, "step": 6969 }, { "epoch": 0.63, "learning_rate": 1.5356802117418713e-05, "loss": 1.2178, "step": 6970 }, { "epoch": 0.63, "learning_rate": 1.535018861995495e-05, "loss": 1.2183, "step": 6971 }, { "epoch": 0.63, "learning_rate": 1.5343575915931967e-05, "loss": 1.2563, "step": 6972 }, { "epoch": 0.63, "learning_rate": 1.5336964005893483e-05, "loss": 1.224, "step": 6973 }, { "epoch": 0.63, "learning_rate": 1.5330352890383155e-05, "loss": 1.2149, "step": 6974 }, { "epoch": 0.63, "learning_rate": 1.5323742569944572e-05, "loss": 1.2327, "step": 6975 }, { "epoch": 0.63, "learning_rate": 1.5317133045121262e-05, "loss": 1.2798, "step": 6976 }, { "epoch": 0.63, "learning_rate": 1.5310524316456684e-05, "loss": 1.2184, "step": 6977 }, { "epoch": 0.63, "learning_rate": 1.5303916384494216e-05, "loss": 1.2075, "step": 6978 }, { "epoch": 0.63, "learning_rate": 1.5297309249777193e-05, "loss": 1.2507, "step": 6979 }, { "epoch": 0.63, "learning_rate": 1.5290702912848882e-05, "loss": 1.212, "step": 6980 }, { "epoch": 0.63, "learning_rate": 1.528409737425247e-05, "loss": 1.1914, "step": 6981 }, { "epoch": 0.63, "learning_rate": 1.5277492634531106e-05, "loss": 1.1816, "step": 6982 }, { "epoch": 0.63, "learning_rate": 1.5270888694227824e-05, "loss": 1.2432, "step": 6983 }, { "epoch": 0.63, "learning_rate": 1.526428555388565e-05, "loss": 1.2203, "step": 6984 }, { "epoch": 0.63, "learning_rate": 1.5257683214047497e-05, "loss": 1.2552, "step": 6985 }, { "epoch": 0.63, "learning_rate": 1.525108167525624e-05, "loss": 1.2577, "step": 6986 }, { "epoch": 0.63, "learning_rate": 1.5244480938054684e-05, "loss": 1.1808, "step": 6987 }, { "epoch": 0.63, "learning_rate": 1.5237881002985571e-05, "loss": 1.252, "step": 6988 }, { "epoch": 0.63, "learning_rate": 1.5231281870591555e-05, "loss": 1.2018, "step": 6989 }, { "epoch": 0.63, "learning_rate": 1.522468354141524e-05, "loss": 1.2378, "step": 6990 }, { "epoch": 0.63, "learning_rate": 1.521808601599917e-05, "loss": 1.1777, "step": 6991 }, { "epoch": 0.63, "learning_rate": 1.521148929488582e-05, "loss": 1.2047, "step": 6992 }, { "epoch": 0.63, "learning_rate": 1.5204893378617597e-05, "loss": 1.2861, "step": 6993 }, { "epoch": 0.63, "learning_rate": 1.519829826773682e-05, "loss": 1.2151, "step": 6994 }, { "epoch": 0.63, "learning_rate": 1.5191703962785778e-05, "loss": 1.2441, "step": 6995 }, { "epoch": 0.63, "learning_rate": 1.5185110464306674e-05, "loss": 1.2176, "step": 6996 }, { "epoch": 0.63, "learning_rate": 1.5178517772841655e-05, "loss": 1.1911, "step": 6997 }, { "epoch": 0.63, "learning_rate": 1.5171925888932792e-05, "loss": 1.2147, "step": 6998 }, { "epoch": 0.63, "learning_rate": 1.5165334813122078e-05, "loss": 1.2246, "step": 6999 }, { "epoch": 0.63, "learning_rate": 1.5158744545951467e-05, "loss": 1.2198, "step": 7000 }, { "epoch": 0.63, "learning_rate": 1.5152155087962839e-05, "loss": 1.2109, "step": 7001 }, { "epoch": 0.63, "learning_rate": 1.5145566439697984e-05, "loss": 1.1952, "step": 7002 }, { "epoch": 0.63, "learning_rate": 1.5138978601698667e-05, "loss": 1.2155, "step": 7003 }, { "epoch": 0.63, "learning_rate": 1.513239157450654e-05, "loss": 1.2122, "step": 7004 }, { "epoch": 0.63, "learning_rate": 1.5125805358663225e-05, "loss": 1.2094, "step": 7005 }, { "epoch": 0.63, "learning_rate": 1.5119219954710257e-05, "loss": 1.2348, "step": 7006 }, { "epoch": 0.63, "learning_rate": 1.5112635363189109e-05, "loss": 1.1942, "step": 7007 }, { "epoch": 0.63, "learning_rate": 1.5106051584641207e-05, "loss": 1.241, "step": 7008 }, { "epoch": 0.63, "learning_rate": 1.5099468619607865e-05, "loss": 1.213, "step": 7009 }, { "epoch": 0.63, "learning_rate": 1.5092886468630374e-05, "loss": 1.2185, "step": 7010 }, { "epoch": 0.63, "learning_rate": 1.5086305132249934e-05, "loss": 1.2609, "step": 7011 }, { "epoch": 0.63, "learning_rate": 1.5079724611007687e-05, "loss": 1.2399, "step": 7012 }, { "epoch": 0.63, "learning_rate": 1.507314490544471e-05, "loss": 1.2388, "step": 7013 }, { "epoch": 0.63, "learning_rate": 1.5066566016102016e-05, "loss": 1.2133, "step": 7014 }, { "epoch": 0.63, "learning_rate": 1.5059987943520521e-05, "loss": 1.2585, "step": 7015 }, { "epoch": 0.63, "learning_rate": 1.5053410688241112e-05, "loss": 1.2559, "step": 7016 }, { "epoch": 0.63, "learning_rate": 1.5046834250804598e-05, "loss": 1.2642, "step": 7017 }, { "epoch": 0.63, "learning_rate": 1.5040258631751703e-05, "loss": 1.1973, "step": 7018 }, { "epoch": 0.63, "learning_rate": 1.5033683831623114e-05, "loss": 1.2654, "step": 7019 }, { "epoch": 0.63, "learning_rate": 1.5027109850959414e-05, "loss": 1.2238, "step": 7020 }, { "epoch": 0.63, "learning_rate": 1.5020536690301148e-05, "loss": 1.2019, "step": 7021 }, { "epoch": 0.63, "learning_rate": 1.5013964350188791e-05, "loss": 1.1784, "step": 7022 }, { "epoch": 0.63, "learning_rate": 1.500739283116273e-05, "loss": 1.2412, "step": 7023 }, { "epoch": 0.63, "learning_rate": 1.5000822133763318e-05, "loss": 1.2618, "step": 7024 }, { "epoch": 0.63, "learning_rate": 1.4994252258530795e-05, "loss": 1.222, "step": 7025 }, { "epoch": 0.63, "learning_rate": 1.4987683206005376e-05, "loss": 1.1942, "step": 7026 }, { "epoch": 0.63, "learning_rate": 1.4981114976727184e-05, "loss": 1.2384, "step": 7027 }, { "epoch": 0.63, "learning_rate": 1.4974547571236281e-05, "loss": 1.2651, "step": 7028 }, { "epoch": 0.64, "learning_rate": 1.4967980990072672e-05, "loss": 1.2135, "step": 7029 }, { "epoch": 0.64, "learning_rate": 1.4961415233776272e-05, "loss": 1.2034, "step": 7030 }, { "epoch": 0.64, "learning_rate": 1.4954850302886944e-05, "loss": 1.2027, "step": 7031 }, { "epoch": 0.64, "learning_rate": 1.4948286197944478e-05, "loss": 1.2848, "step": 7032 }, { "epoch": 0.64, "learning_rate": 1.49417229194886e-05, "loss": 1.2353, "step": 7033 }, { "epoch": 0.64, "learning_rate": 1.4935160468058964e-05, "loss": 1.2351, "step": 7034 }, { "epoch": 0.64, "learning_rate": 1.492859884419517e-05, "loss": 1.2225, "step": 7035 }, { "epoch": 0.64, "learning_rate": 1.492203804843671e-05, "loss": 1.2253, "step": 7036 }, { "epoch": 0.64, "learning_rate": 1.491547808132305e-05, "loss": 1.2674, "step": 7037 }, { "epoch": 0.64, "learning_rate": 1.4908918943393582e-05, "loss": 1.2542, "step": 7038 }, { "epoch": 0.64, "learning_rate": 1.4902360635187607e-05, "loss": 1.2443, "step": 7039 }, { "epoch": 0.64, "learning_rate": 1.4895803157244386e-05, "loss": 1.2287, "step": 7040 }, { "epoch": 0.64, "learning_rate": 1.4889246510103077e-05, "loss": 1.1968, "step": 7041 }, { "epoch": 0.64, "learning_rate": 1.4882690694302807e-05, "loss": 1.1787, "step": 7042 }, { "epoch": 0.64, "learning_rate": 1.4876135710382617e-05, "loss": 1.1944, "step": 7043 }, { "epoch": 0.64, "learning_rate": 1.4869581558881468e-05, "loss": 1.2679, "step": 7044 }, { "epoch": 0.64, "learning_rate": 1.4863028240338284e-05, "loss": 1.2124, "step": 7045 }, { "epoch": 0.64, "learning_rate": 1.4856475755291878e-05, "loss": 1.2453, "step": 7046 }, { "epoch": 0.64, "learning_rate": 1.4849924104281037e-05, "loss": 1.196, "step": 7047 }, { "epoch": 0.64, "learning_rate": 1.4843373287844454e-05, "loss": 1.2155, "step": 7048 }, { "epoch": 0.64, "learning_rate": 1.4836823306520752e-05, "loss": 1.2052, "step": 7049 }, { "epoch": 0.64, "learning_rate": 1.4830274160848517e-05, "loss": 1.2095, "step": 7050 }, { "epoch": 0.64, "learning_rate": 1.482372585136621e-05, "loss": 1.2076, "step": 7051 }, { "epoch": 0.64, "learning_rate": 1.4817178378612281e-05, "loss": 1.2047, "step": 7052 }, { "epoch": 0.64, "learning_rate": 1.481063174312507e-05, "loss": 1.221, "step": 7053 }, { "epoch": 0.64, "learning_rate": 1.480408594544287e-05, "loss": 1.2037, "step": 7054 }, { "epoch": 0.64, "learning_rate": 1.4797540986103909e-05, "loss": 1.2009, "step": 7055 }, { "epoch": 0.64, "learning_rate": 1.4790996865646322e-05, "loss": 1.1657, "step": 7056 }, { "epoch": 0.64, "learning_rate": 1.4784453584608188e-05, "loss": 1.187, "step": 7057 }, { "epoch": 0.64, "learning_rate": 1.4777911143527524e-05, "loss": 1.1635, "step": 7058 }, { "epoch": 0.64, "learning_rate": 1.4771369542942275e-05, "loss": 1.2062, "step": 7059 }, { "epoch": 0.64, "learning_rate": 1.4764828783390311e-05, "loss": 1.2361, "step": 7060 }, { "epoch": 0.64, "learning_rate": 1.4758288865409442e-05, "loss": 1.221, "step": 7061 }, { "epoch": 0.64, "learning_rate": 1.475174978953739e-05, "loss": 1.2557, "step": 7062 }, { "epoch": 0.64, "learning_rate": 1.4745211556311821e-05, "loss": 1.296, "step": 7063 }, { "epoch": 0.64, "learning_rate": 1.4738674166270344e-05, "loss": 1.1503, "step": 7064 }, { "epoch": 0.64, "learning_rate": 1.4732137619950475e-05, "loss": 1.2152, "step": 7065 }, { "epoch": 0.64, "learning_rate": 1.472560191788968e-05, "loss": 1.2116, "step": 7066 }, { "epoch": 0.64, "learning_rate": 1.4719067060625335e-05, "loss": 1.2115, "step": 7067 }, { "epoch": 0.64, "learning_rate": 1.4712533048694774e-05, "loss": 1.1942, "step": 7068 }, { "epoch": 0.64, "learning_rate": 1.470599988263523e-05, "loss": 1.2193, "step": 7069 }, { "epoch": 0.64, "learning_rate": 1.4699467562983888e-05, "loss": 1.2893, "step": 7070 }, { "epoch": 0.64, "learning_rate": 1.469293609027787e-05, "loss": 1.2412, "step": 7071 }, { "epoch": 0.64, "learning_rate": 1.4686405465054198e-05, "loss": 1.228, "step": 7072 }, { "epoch": 0.64, "learning_rate": 1.4679875687849854e-05, "loss": 1.2481, "step": 7073 }, { "epoch": 0.64, "learning_rate": 1.4673346759201729e-05, "loss": 1.2179, "step": 7074 }, { "epoch": 0.64, "learning_rate": 1.466681867964666e-05, "loss": 1.2491, "step": 7075 }, { "epoch": 0.64, "learning_rate": 1.466029144972142e-05, "loss": 1.1901, "step": 7076 }, { "epoch": 0.64, "learning_rate": 1.4653765069962685e-05, "loss": 1.2477, "step": 7077 }, { "epoch": 0.64, "learning_rate": 1.4647239540907076e-05, "loss": 1.2359, "step": 7078 }, { "epoch": 0.64, "learning_rate": 1.4640714863091145e-05, "loss": 1.2078, "step": 7079 }, { "epoch": 0.64, "learning_rate": 1.4634191037051384e-05, "loss": 1.2459, "step": 7080 }, { "epoch": 0.64, "learning_rate": 1.4627668063324198e-05, "loss": 1.2435, "step": 7081 }, { "epoch": 0.64, "learning_rate": 1.462114594244593e-05, "loss": 1.2469, "step": 7082 }, { "epoch": 0.64, "learning_rate": 1.4614624674952842e-05, "loss": 1.1961, "step": 7083 }, { "epoch": 0.64, "learning_rate": 1.4608104261381145e-05, "loss": 1.2328, "step": 7084 }, { "epoch": 0.64, "learning_rate": 1.4601584702266969e-05, "loss": 1.1933, "step": 7085 }, { "epoch": 0.64, "learning_rate": 1.4595065998146376e-05, "loss": 1.2027, "step": 7086 }, { "epoch": 0.64, "learning_rate": 1.4588548149555353e-05, "loss": 1.1984, "step": 7087 }, { "epoch": 0.64, "learning_rate": 1.4582031157029822e-05, "loss": 1.2348, "step": 7088 }, { "epoch": 0.64, "learning_rate": 1.4575515021105617e-05, "loss": 1.2291, "step": 7089 }, { "epoch": 0.64, "learning_rate": 1.4568999742318546e-05, "loss": 1.1932, "step": 7090 }, { "epoch": 0.64, "learning_rate": 1.45624853212043e-05, "loss": 1.2379, "step": 7091 }, { "epoch": 0.64, "learning_rate": 1.455597175829852e-05, "loss": 1.181, "step": 7092 }, { "epoch": 0.64, "learning_rate": 1.4549459054136777e-05, "loss": 1.2102, "step": 7093 }, { "epoch": 0.64, "learning_rate": 1.4542947209254551e-05, "loss": 1.2541, "step": 7094 }, { "epoch": 0.64, "learning_rate": 1.4536436224187294e-05, "loss": 1.2123, "step": 7095 }, { "epoch": 0.64, "learning_rate": 1.4529926099470348e-05, "loss": 1.2036, "step": 7096 }, { "epoch": 0.64, "learning_rate": 1.4523416835639003e-05, "loss": 1.1867, "step": 7097 }, { "epoch": 0.64, "learning_rate": 1.4516908433228469e-05, "loss": 1.192, "step": 7098 }, { "epoch": 0.64, "learning_rate": 1.451040089277388e-05, "loss": 1.1986, "step": 7099 }, { "epoch": 0.64, "learning_rate": 1.4503894214810326e-05, "loss": 1.2888, "step": 7100 }, { "epoch": 0.64, "learning_rate": 1.4497388399872802e-05, "loss": 1.215, "step": 7101 }, { "epoch": 0.64, "learning_rate": 1.449088344849624e-05, "loss": 1.1801, "step": 7102 }, { "epoch": 0.64, "learning_rate": 1.4484379361215495e-05, "loss": 1.255, "step": 7103 }, { "epoch": 0.64, "learning_rate": 1.4477876138565361e-05, "loss": 1.2161, "step": 7104 }, { "epoch": 0.64, "learning_rate": 1.4471373781080539e-05, "loss": 1.2576, "step": 7105 }, { "epoch": 0.64, "learning_rate": 1.4464872289295697e-05, "loss": 1.2483, "step": 7106 }, { "epoch": 0.64, "learning_rate": 1.4458371663745401e-05, "loss": 1.2405, "step": 7107 }, { "epoch": 0.64, "learning_rate": 1.445187190496416e-05, "loss": 1.2717, "step": 7108 }, { "epoch": 0.64, "learning_rate": 1.44453730134864e-05, "loss": 1.2564, "step": 7109 }, { "epoch": 0.64, "learning_rate": 1.4438874989846474e-05, "loss": 1.2069, "step": 7110 }, { "epoch": 0.64, "learning_rate": 1.4432377834578691e-05, "loss": 1.1773, "step": 7111 }, { "epoch": 0.64, "learning_rate": 1.4425881548217262e-05, "loss": 1.2395, "step": 7112 }, { "epoch": 0.64, "learning_rate": 1.441938613129633e-05, "loss": 1.2267, "step": 7113 }, { "epoch": 0.64, "learning_rate": 1.4412891584349974e-05, "loss": 1.1983, "step": 7114 }, { "epoch": 0.64, "learning_rate": 1.4406397907912188e-05, "loss": 1.1867, "step": 7115 }, { "epoch": 0.64, "learning_rate": 1.4399905102516927e-05, "loss": 1.1933, "step": 7116 }, { "epoch": 0.64, "learning_rate": 1.4393413168698037e-05, "loss": 1.2784, "step": 7117 }, { "epoch": 0.64, "learning_rate": 1.4386922106989309e-05, "loss": 1.2107, "step": 7118 }, { "epoch": 0.64, "learning_rate": 1.4380431917924459e-05, "loss": 1.2038, "step": 7119 }, { "epoch": 0.64, "learning_rate": 1.4373942602037122e-05, "loss": 1.1912, "step": 7120 }, { "epoch": 0.64, "learning_rate": 1.43674541598609e-05, "loss": 1.2367, "step": 7121 }, { "epoch": 0.64, "learning_rate": 1.4360966591929274e-05, "loss": 1.1766, "step": 7122 }, { "epoch": 0.64, "learning_rate": 1.435447989877568e-05, "loss": 1.2188, "step": 7123 }, { "epoch": 0.64, "learning_rate": 1.4347994080933474e-05, "loss": 1.2487, "step": 7124 }, { "epoch": 0.64, "learning_rate": 1.4341509138935949e-05, "loss": 1.2316, "step": 7125 }, { "epoch": 0.64, "learning_rate": 1.43350250733163e-05, "loss": 1.1939, "step": 7126 }, { "epoch": 0.64, "learning_rate": 1.4328541884607697e-05, "loss": 1.2039, "step": 7127 }, { "epoch": 0.64, "learning_rate": 1.4322059573343194e-05, "loss": 1.2483, "step": 7128 }, { "epoch": 0.64, "learning_rate": 1.431557814005579e-05, "loss": 1.2603, "step": 7129 }, { "epoch": 0.64, "learning_rate": 1.430909758527842e-05, "loss": 1.2425, "step": 7130 }, { "epoch": 0.64, "learning_rate": 1.4302617909543913e-05, "loss": 1.2068, "step": 7131 }, { "epoch": 0.64, "learning_rate": 1.4296139113385079e-05, "loss": 1.1953, "step": 7132 }, { "epoch": 0.64, "learning_rate": 1.4289661197334614e-05, "loss": 1.2624, "step": 7133 }, { "epoch": 0.64, "learning_rate": 1.428318416192516e-05, "loss": 1.2321, "step": 7134 }, { "epoch": 0.64, "learning_rate": 1.427670800768927e-05, "loss": 1.1725, "step": 7135 }, { "epoch": 0.64, "learning_rate": 1.427023273515944e-05, "loss": 1.2375, "step": 7136 }, { "epoch": 0.64, "learning_rate": 1.42637583448681e-05, "loss": 1.1985, "step": 7137 }, { "epoch": 0.64, "learning_rate": 1.4257284837347587e-05, "loss": 1.2534, "step": 7138 }, { "epoch": 0.65, "learning_rate": 1.4250812213130176e-05, "loss": 1.2349, "step": 7139 }, { "epoch": 0.65, "learning_rate": 1.4244340472748075e-05, "loss": 1.2171, "step": 7140 }, { "epoch": 0.65, "learning_rate": 1.4237869616733393e-05, "loss": 1.2129, "step": 7141 }, { "epoch": 0.65, "learning_rate": 1.423139964561821e-05, "loss": 1.2144, "step": 7142 }, { "epoch": 0.65, "learning_rate": 1.4224930559934501e-05, "loss": 1.2067, "step": 7143 }, { "epoch": 0.65, "learning_rate": 1.4218462360214175e-05, "loss": 1.1592, "step": 7144 }, { "epoch": 0.65, "learning_rate": 1.4211995046989068e-05, "loss": 1.2504, "step": 7145 }, { "epoch": 0.65, "learning_rate": 1.420552862079095e-05, "loss": 1.2196, "step": 7146 }, { "epoch": 0.65, "learning_rate": 1.4199063082151498e-05, "loss": 1.2703, "step": 7147 }, { "epoch": 0.65, "learning_rate": 1.4192598431602353e-05, "loss": 1.2062, "step": 7148 }, { "epoch": 0.65, "learning_rate": 1.418613466967505e-05, "loss": 1.253, "step": 7149 }, { "epoch": 0.65, "learning_rate": 1.4179671796901061e-05, "loss": 1.258, "step": 7150 }, { "epoch": 0.65, "learning_rate": 1.4173209813811788e-05, "loss": 1.2048, "step": 7151 }, { "epoch": 0.65, "learning_rate": 1.4166748720938546e-05, "loss": 1.2126, "step": 7152 }, { "epoch": 0.65, "learning_rate": 1.416028851881261e-05, "loss": 1.2651, "step": 7153 }, { "epoch": 0.65, "learning_rate": 1.4153829207965146e-05, "loss": 1.272, "step": 7154 }, { "epoch": 0.65, "learning_rate": 1.4147370788927267e-05, "loss": 1.1838, "step": 7155 }, { "epoch": 0.65, "learning_rate": 1.4140913262230004e-05, "loss": 1.2474, "step": 7156 }, { "epoch": 0.65, "learning_rate": 1.4134456628404302e-05, "loss": 1.2003, "step": 7157 }, { "epoch": 0.65, "learning_rate": 1.4128000887981074e-05, "loss": 1.2661, "step": 7158 }, { "epoch": 0.65, "learning_rate": 1.4121546041491124e-05, "loss": 1.2251, "step": 7159 }, { "epoch": 0.65, "learning_rate": 1.4115092089465185e-05, "loss": 1.2131, "step": 7160 }, { "epoch": 0.65, "learning_rate": 1.4108639032433929e-05, "loss": 1.1706, "step": 7161 }, { "epoch": 0.65, "learning_rate": 1.4102186870927935e-05, "loss": 1.2303, "step": 7162 }, { "epoch": 0.65, "learning_rate": 1.4095735605477745e-05, "loss": 1.193, "step": 7163 }, { "epoch": 0.65, "learning_rate": 1.4089285236613792e-05, "loss": 1.2806, "step": 7164 }, { "epoch": 0.65, "learning_rate": 1.4082835764866451e-05, "loss": 1.1959, "step": 7165 }, { "epoch": 0.65, "learning_rate": 1.4076387190766017e-05, "loss": 1.2235, "step": 7166 }, { "epoch": 0.65, "learning_rate": 1.4069939514842711e-05, "loss": 1.2255, "step": 7167 }, { "epoch": 0.65, "learning_rate": 1.4063492737626682e-05, "loss": 1.2427, "step": 7168 }, { "epoch": 0.65, "learning_rate": 1.4057046859648018e-05, "loss": 1.2847, "step": 7169 }, { "epoch": 0.65, "learning_rate": 1.4050601881436715e-05, "loss": 1.1582, "step": 7170 }, { "epoch": 0.65, "learning_rate": 1.40441578035227e-05, "loss": 1.178, "step": 7171 }, { "epoch": 0.65, "learning_rate": 1.4037714626435828e-05, "loss": 1.2317, "step": 7172 }, { "epoch": 0.65, "learning_rate": 1.4031272350705871e-05, "loss": 1.2314, "step": 7173 }, { "epoch": 0.65, "learning_rate": 1.4024830976862551e-05, "loss": 1.2101, "step": 7174 }, { "epoch": 0.65, "learning_rate": 1.4018390505435492e-05, "loss": 1.1977, "step": 7175 }, { "epoch": 0.65, "learning_rate": 1.4011950936954255e-05, "loss": 1.2263, "step": 7176 }, { "epoch": 0.65, "learning_rate": 1.4005512271948318e-05, "loss": 1.2087, "step": 7177 }, { "epoch": 0.65, "learning_rate": 1.399907451094708e-05, "loss": 1.2338, "step": 7178 }, { "epoch": 0.65, "learning_rate": 1.3992637654479904e-05, "loss": 1.2005, "step": 7179 }, { "epoch": 0.65, "learning_rate": 1.3986201703076035e-05, "loss": 1.2199, "step": 7180 }, { "epoch": 0.65, "learning_rate": 1.3979766657264656e-05, "loss": 1.2429, "step": 7181 }, { "epoch": 0.65, "learning_rate": 1.3973332517574883e-05, "loss": 1.2026, "step": 7182 }, { "epoch": 0.65, "learning_rate": 1.3966899284535745e-05, "loss": 1.2288, "step": 7183 }, { "epoch": 0.65, "learning_rate": 1.3960466958676221e-05, "loss": 1.2362, "step": 7184 }, { "epoch": 0.65, "learning_rate": 1.3954035540525188e-05, "loss": 1.2141, "step": 7185 }, { "epoch": 0.65, "learning_rate": 1.394760503061146e-05, "loss": 1.2363, "step": 7186 }, { "epoch": 0.65, "learning_rate": 1.3941175429463781e-05, "loss": 1.2042, "step": 7187 }, { "epoch": 0.65, "learning_rate": 1.393474673761081e-05, "loss": 1.1955, "step": 7188 }, { "epoch": 0.65, "learning_rate": 1.3928318955581127e-05, "loss": 1.2594, "step": 7189 }, { "epoch": 0.65, "learning_rate": 1.3921892083903265e-05, "loss": 1.2637, "step": 7190 }, { "epoch": 0.65, "learning_rate": 1.3915466123105659e-05, "loss": 1.1985, "step": 7191 }, { "epoch": 0.65, "learning_rate": 1.3909041073716666e-05, "loss": 1.1565, "step": 7192 }, { "epoch": 0.65, "learning_rate": 1.3902616936264578e-05, "loss": 1.197, "step": 7193 }, { "epoch": 0.65, "learning_rate": 1.3896193711277605e-05, "loss": 1.2317, "step": 7194 }, { "epoch": 0.65, "learning_rate": 1.3889771399283902e-05, "loss": 1.2072, "step": 7195 }, { "epoch": 0.65, "learning_rate": 1.3883350000811527e-05, "loss": 1.3067, "step": 7196 }, { "epoch": 0.65, "learning_rate": 1.3876929516388465e-05, "loss": 1.2427, "step": 7197 }, { "epoch": 0.65, "learning_rate": 1.3870509946542632e-05, "loss": 1.2566, "step": 7198 }, { "epoch": 0.65, "learning_rate": 1.3864091291801856e-05, "loss": 1.1735, "step": 7199 }, { "epoch": 0.65, "learning_rate": 1.3857673552693928e-05, "loss": 1.2426, "step": 7200 }, { "epoch": 0.65, "learning_rate": 1.385125672974652e-05, "loss": 1.2095, "step": 7201 }, { "epoch": 0.65, "learning_rate": 1.3844840823487243e-05, "loss": 1.1711, "step": 7202 }, { "epoch": 0.65, "learning_rate": 1.3838425834443642e-05, "loss": 1.1777, "step": 7203 }, { "epoch": 0.65, "learning_rate": 1.3832011763143166e-05, "loss": 1.1834, "step": 7204 }, { "epoch": 0.65, "learning_rate": 1.3825598610113222e-05, "loss": 1.2089, "step": 7205 }, { "epoch": 0.65, "learning_rate": 1.381918637588112e-05, "loss": 1.2772, "step": 7206 }, { "epoch": 0.65, "learning_rate": 1.381277506097407e-05, "loss": 1.2617, "step": 7207 }, { "epoch": 0.65, "learning_rate": 1.3806364665919258e-05, "loss": 1.2455, "step": 7208 }, { "epoch": 0.65, "learning_rate": 1.3799955191243763e-05, "loss": 1.2188, "step": 7209 }, { "epoch": 0.65, "learning_rate": 1.379354663747458e-05, "loss": 1.2396, "step": 7210 }, { "epoch": 0.65, "learning_rate": 1.3787139005138667e-05, "loss": 1.3018, "step": 7211 }, { "epoch": 0.65, "learning_rate": 1.3780732294762869e-05, "loss": 1.2074, "step": 7212 }, { "epoch": 0.65, "learning_rate": 1.3774326506873966e-05, "loss": 1.2512, "step": 7213 }, { "epoch": 0.65, "learning_rate": 1.3767921641998668e-05, "loss": 1.2353, "step": 7214 }, { "epoch": 0.65, "learning_rate": 1.3761517700663592e-05, "loss": 1.2423, "step": 7215 }, { "epoch": 0.65, "learning_rate": 1.3755114683395314e-05, "loss": 1.1827, "step": 7216 }, { "epoch": 0.65, "learning_rate": 1.3748712590720304e-05, "loss": 1.2588, "step": 7217 }, { "epoch": 0.65, "learning_rate": 1.3742311423164956e-05, "loss": 1.1927, "step": 7218 }, { "epoch": 0.65, "learning_rate": 1.3735911181255606e-05, "loss": 1.2075, "step": 7219 }, { "epoch": 0.65, "learning_rate": 1.3729511865518487e-05, "loss": 1.1982, "step": 7220 }, { "epoch": 0.65, "learning_rate": 1.3723113476479799e-05, "loss": 1.1976, "step": 7221 }, { "epoch": 0.65, "learning_rate": 1.371671601466562e-05, "loss": 1.2369, "step": 7222 }, { "epoch": 0.65, "learning_rate": 1.3710319480601985e-05, "loss": 1.1999, "step": 7223 }, { "epoch": 0.65, "learning_rate": 1.3703923874814827e-05, "loss": 1.1281, "step": 7224 }, { "epoch": 0.65, "learning_rate": 1.3697529197830011e-05, "loss": 1.2058, "step": 7225 }, { "epoch": 0.65, "learning_rate": 1.3691135450173345e-05, "loss": 1.2852, "step": 7226 }, { "epoch": 0.65, "learning_rate": 1.368474263237055e-05, "loss": 1.2804, "step": 7227 }, { "epoch": 0.65, "learning_rate": 1.3678350744947233e-05, "loss": 1.219, "step": 7228 }, { "epoch": 0.65, "learning_rate": 1.3671959788428987e-05, "loss": 1.2333, "step": 7229 }, { "epoch": 0.65, "learning_rate": 1.366556976334129e-05, "loss": 1.2057, "step": 7230 }, { "epoch": 0.65, "learning_rate": 1.3659180670209537e-05, "loss": 1.2522, "step": 7231 }, { "epoch": 0.65, "learning_rate": 1.3652792509559097e-05, "loss": 1.2299, "step": 7232 }, { "epoch": 0.65, "learning_rate": 1.3646405281915186e-05, "loss": 1.2596, "step": 7233 }, { "epoch": 0.65, "learning_rate": 1.3640018987803014e-05, "loss": 1.2463, "step": 7234 }, { "epoch": 0.65, "learning_rate": 1.3633633627747669e-05, "loss": 1.2331, "step": 7235 }, { "epoch": 0.65, "learning_rate": 1.3627249202274173e-05, "loss": 1.2173, "step": 7236 }, { "epoch": 0.65, "learning_rate": 1.3620865711907493e-05, "loss": 1.1959, "step": 7237 }, { "epoch": 0.65, "learning_rate": 1.3614483157172491e-05, "loss": 1.2524, "step": 7238 }, { "epoch": 0.65, "learning_rate": 1.3608101538593965e-05, "loss": 1.2165, "step": 7239 }, { "epoch": 0.65, "learning_rate": 1.3601720856696631e-05, "loss": 1.2098, "step": 7240 }, { "epoch": 0.65, "learning_rate": 1.3595341112005122e-05, "loss": 1.2068, "step": 7241 }, { "epoch": 0.65, "learning_rate": 1.3588962305044023e-05, "loss": 1.2582, "step": 7242 }, { "epoch": 0.65, "learning_rate": 1.3582584436337814e-05, "loss": 1.2089, "step": 7243 }, { "epoch": 0.65, "learning_rate": 1.3576207506410899e-05, "loss": 1.2306, "step": 7244 }, { "epoch": 0.65, "learning_rate": 1.3569831515787618e-05, "loss": 1.2204, "step": 7245 }, { "epoch": 0.65, "learning_rate": 1.356345646499221e-05, "loss": 1.1726, "step": 7246 }, { "epoch": 0.65, "learning_rate": 1.3557082354548878e-05, "loss": 1.1904, "step": 7247 }, { "epoch": 0.65, "learning_rate": 1.3550709184981725e-05, "loss": 1.2485, "step": 7248 }, { "epoch": 0.65, "learning_rate": 1.354433695681474e-05, "loss": 1.2619, "step": 7249 }, { "epoch": 0.66, "learning_rate": 1.3537965670571906e-05, "loss": 1.1798, "step": 7250 }, { "epoch": 0.66, "learning_rate": 1.3531595326777075e-05, "loss": 1.1747, "step": 7251 }, { "epoch": 0.66, "learning_rate": 1.3525225925954035e-05, "loss": 1.2496, "step": 7252 }, { "epoch": 0.66, "learning_rate": 1.3518857468626525e-05, "loss": 1.2245, "step": 7253 }, { "epoch": 0.66, "learning_rate": 1.3512489955318144e-05, "loss": 1.2443, "step": 7254 }, { "epoch": 0.66, "learning_rate": 1.3506123386552483e-05, "loss": 1.2858, "step": 7255 }, { "epoch": 0.66, "learning_rate": 1.3499757762853007e-05, "loss": 1.2497, "step": 7256 }, { "epoch": 0.66, "learning_rate": 1.3493393084743117e-05, "loss": 1.2414, "step": 7257 }, { "epoch": 0.66, "learning_rate": 1.3487029352746162e-05, "loss": 1.2144, "step": 7258 }, { "epoch": 0.66, "learning_rate": 1.3480666567385355e-05, "loss": 1.2953, "step": 7259 }, { "epoch": 0.66, "learning_rate": 1.3474304729183895e-05, "loss": 1.2908, "step": 7260 }, { "epoch": 0.66, "learning_rate": 1.3467943838664863e-05, "loss": 1.2269, "step": 7261 }, { "epoch": 0.66, "learning_rate": 1.3461583896351265e-05, "loss": 1.199, "step": 7262 }, { "epoch": 0.66, "learning_rate": 1.3455224902766053e-05, "loss": 1.2291, "step": 7263 }, { "epoch": 0.66, "learning_rate": 1.3448866858432079e-05, "loss": 1.2179, "step": 7264 }, { "epoch": 0.66, "learning_rate": 1.344250976387213e-05, "loss": 1.2358, "step": 7265 }, { "epoch": 0.66, "learning_rate": 1.3436153619608894e-05, "loss": 1.2292, "step": 7266 }, { "epoch": 0.66, "learning_rate": 1.3429798426164998e-05, "loss": 1.2316, "step": 7267 }, { "epoch": 0.66, "learning_rate": 1.3423444184062997e-05, "loss": 1.2143, "step": 7268 }, { "epoch": 0.66, "learning_rate": 1.3417090893825368e-05, "loss": 1.2145, "step": 7269 }, { "epoch": 0.66, "learning_rate": 1.3410738555974464e-05, "loss": 1.2764, "step": 7270 }, { "epoch": 0.66, "learning_rate": 1.3404387171032631e-05, "loss": 1.2433, "step": 7271 }, { "epoch": 0.66, "learning_rate": 1.3398036739522087e-05, "loss": 1.2065, "step": 7272 }, { "epoch": 0.66, "learning_rate": 1.339168726196498e-05, "loss": 1.2208, "step": 7273 }, { "epoch": 0.66, "learning_rate": 1.3385338738883419e-05, "loss": 1.2022, "step": 7274 }, { "epoch": 0.66, "learning_rate": 1.3378991170799351e-05, "loss": 1.2292, "step": 7275 }, { "epoch": 0.66, "learning_rate": 1.3372644558234735e-05, "loss": 1.1897, "step": 7276 }, { "epoch": 0.66, "learning_rate": 1.3366298901711396e-05, "loss": 1.2931, "step": 7277 }, { "epoch": 0.66, "learning_rate": 1.335995420175109e-05, "loss": 1.2294, "step": 7278 }, { "epoch": 0.66, "learning_rate": 1.3353610458875523e-05, "loss": 1.1853, "step": 7279 }, { "epoch": 0.66, "learning_rate": 1.3347267673606268e-05, "loss": 1.2213, "step": 7280 }, { "epoch": 0.66, "learning_rate": 1.3340925846464874e-05, "loss": 1.2322, "step": 7281 }, { "epoch": 0.66, "learning_rate": 1.3334584977972783e-05, "loss": 1.2166, "step": 7282 }, { "epoch": 0.66, "learning_rate": 1.3328245068651348e-05, "loss": 1.2689, "step": 7283 }, { "epoch": 0.66, "learning_rate": 1.3321906119021893e-05, "loss": 1.2124, "step": 7284 }, { "epoch": 0.66, "learning_rate": 1.3315568129605588e-05, "loss": 1.2114, "step": 7285 }, { "epoch": 0.66, "learning_rate": 1.3309231100923585e-05, "loss": 1.1926, "step": 7286 }, { "epoch": 0.66, "learning_rate": 1.3302895033496943e-05, "loss": 1.2359, "step": 7287 }, { "epoch": 0.66, "learning_rate": 1.3296559927846611e-05, "loss": 1.2084, "step": 7288 }, { "epoch": 0.66, "learning_rate": 1.329022578449351e-05, "loss": 1.2101, "step": 7289 }, { "epoch": 0.66, "learning_rate": 1.3283892603958454e-05, "loss": 1.189, "step": 7290 }, { "epoch": 0.66, "learning_rate": 1.3277560386762144e-05, "loss": 1.2201, "step": 7291 }, { "epoch": 0.66, "learning_rate": 1.3271229133425277e-05, "loss": 1.1849, "step": 7292 }, { "epoch": 0.66, "learning_rate": 1.3264898844468413e-05, "loss": 1.21, "step": 7293 }, { "epoch": 0.66, "learning_rate": 1.3258569520412039e-05, "loss": 1.1572, "step": 7294 }, { "epoch": 0.66, "learning_rate": 1.3252241161776614e-05, "loss": 1.2648, "step": 7295 }, { "epoch": 0.66, "learning_rate": 1.3245913769082425e-05, "loss": 1.2393, "step": 7296 }, { "epoch": 0.66, "learning_rate": 1.3239587342849766e-05, "loss": 1.2208, "step": 7297 }, { "epoch": 0.66, "learning_rate": 1.3233261883598813e-05, "loss": 1.1969, "step": 7298 }, { "epoch": 0.66, "learning_rate": 1.3226937391849647e-05, "loss": 1.2095, "step": 7299 }, { "epoch": 0.66, "learning_rate": 1.3220613868122334e-05, "loss": 1.2343, "step": 7300 }, { "epoch": 0.66, "learning_rate": 1.3214291312936761e-05, "loss": 1.2265, "step": 7301 }, { "epoch": 0.66, "learning_rate": 1.320796972681283e-05, "loss": 1.2504, "step": 7302 }, { "epoch": 0.66, "learning_rate": 1.3201649110270308e-05, "loss": 1.2164, "step": 7303 }, { "epoch": 0.66, "learning_rate": 1.3195329463828892e-05, "loss": 1.2322, "step": 7304 }, { "epoch": 0.66, "learning_rate": 1.3189010788008233e-05, "loss": 1.1823, "step": 7305 }, { "epoch": 0.66, "learning_rate": 1.3182693083327834e-05, "loss": 1.3093, "step": 7306 }, { "epoch": 0.66, "learning_rate": 1.317637635030719e-05, "loss": 1.2077, "step": 7307 }, { "epoch": 0.66, "learning_rate": 1.3170060589465675e-05, "loss": 1.2314, "step": 7308 }, { "epoch": 0.66, "learning_rate": 1.316374580132258e-05, "loss": 1.2617, "step": 7309 }, { "epoch": 0.66, "learning_rate": 1.3157431986397162e-05, "loss": 1.2085, "step": 7310 }, { "epoch": 0.66, "learning_rate": 1.3151119145208529e-05, "loss": 1.2284, "step": 7311 }, { "epoch": 0.66, "learning_rate": 1.314480727827575e-05, "loss": 1.2147, "step": 7312 }, { "epoch": 0.66, "learning_rate": 1.313849638611783e-05, "loss": 1.1973, "step": 7313 }, { "epoch": 0.66, "learning_rate": 1.3132186469253655e-05, "loss": 1.2427, "step": 7314 }, { "epoch": 0.66, "learning_rate": 1.3125877528202046e-05, "loss": 1.2656, "step": 7315 }, { "epoch": 0.66, "learning_rate": 1.311956956348177e-05, "loss": 1.2406, "step": 7316 }, { "epoch": 0.66, "learning_rate": 1.311326257561145e-05, "loss": 1.2373, "step": 7317 }, { "epoch": 0.66, "learning_rate": 1.3106956565109701e-05, "loss": 1.244, "step": 7318 }, { "epoch": 0.66, "learning_rate": 1.3100651532495017e-05, "loss": 1.2067, "step": 7319 }, { "epoch": 0.66, "learning_rate": 1.3094347478285806e-05, "loss": 1.2476, "step": 7320 }, { "epoch": 0.66, "learning_rate": 1.3088044403000437e-05, "loss": 1.2176, "step": 7321 }, { "epoch": 0.66, "learning_rate": 1.3081742307157136e-05, "loss": 1.2024, "step": 7322 }, { "epoch": 0.66, "learning_rate": 1.3075441191274108e-05, "loss": 1.2221, "step": 7323 }, { "epoch": 0.66, "learning_rate": 1.3069141055869447e-05, "loss": 1.2378, "step": 7324 }, { "epoch": 0.66, "learning_rate": 1.3062841901461159e-05, "loss": 1.1761, "step": 7325 }, { "epoch": 0.66, "learning_rate": 1.3056543728567211e-05, "loss": 1.2052, "step": 7326 }, { "epoch": 0.66, "learning_rate": 1.3050246537705424e-05, "loss": 1.2643, "step": 7327 }, { "epoch": 0.66, "learning_rate": 1.3043950329393604e-05, "loss": 1.2801, "step": 7328 }, { "epoch": 0.66, "learning_rate": 1.3037655104149436e-05, "loss": 1.2098, "step": 7329 }, { "epoch": 0.66, "learning_rate": 1.3031360862490522e-05, "loss": 1.1823, "step": 7330 }, { "epoch": 0.66, "learning_rate": 1.3025067604934432e-05, "loss": 1.2366, "step": 7331 }, { "epoch": 0.66, "learning_rate": 1.301877533199859e-05, "loss": 1.2059, "step": 7332 }, { "epoch": 0.66, "learning_rate": 1.301248404420036e-05, "loss": 1.208, "step": 7333 }, { "epoch": 0.66, "learning_rate": 1.300619374205706e-05, "loss": 1.1854, "step": 7334 }, { "epoch": 0.66, "learning_rate": 1.299990442608589e-05, "loss": 1.1119, "step": 7335 }, { "epoch": 0.66, "learning_rate": 1.2993616096803968e-05, "loss": 1.208, "step": 7336 }, { "epoch": 0.66, "learning_rate": 1.2987328754728372e-05, "loss": 1.2343, "step": 7337 }, { "epoch": 0.66, "learning_rate": 1.2981042400376031e-05, "loss": 1.2686, "step": 7338 }, { "epoch": 0.66, "learning_rate": 1.2974757034263857e-05, "loss": 1.2065, "step": 7339 }, { "epoch": 0.66, "learning_rate": 1.2968472656908648e-05, "loss": 1.1762, "step": 7340 }, { "epoch": 0.66, "learning_rate": 1.2962189268827118e-05, "loss": 1.2477, "step": 7341 }, { "epoch": 0.66, "learning_rate": 1.2955906870535933e-05, "loss": 1.2763, "step": 7342 }, { "epoch": 0.66, "learning_rate": 1.2949625462551618e-05, "loss": 1.2206, "step": 7343 }, { "epoch": 0.66, "learning_rate": 1.2943345045390681e-05, "loss": 1.1975, "step": 7344 }, { "epoch": 0.66, "learning_rate": 1.2937065619569511e-05, "loss": 1.2649, "step": 7345 }, { "epoch": 0.66, "learning_rate": 1.293078718560441e-05, "loss": 1.2141, "step": 7346 }, { "epoch": 0.66, "learning_rate": 1.2924509744011648e-05, "loss": 1.2242, "step": 7347 }, { "epoch": 0.66, "learning_rate": 1.291823329530733e-05, "loss": 1.2257, "step": 7348 }, { "epoch": 0.66, "learning_rate": 1.2911957840007568e-05, "loss": 1.1853, "step": 7349 }, { "epoch": 0.66, "learning_rate": 1.2905683378628336e-05, "loss": 1.1858, "step": 7350 }, { "epoch": 0.66, "learning_rate": 1.2899409911685529e-05, "loss": 1.1853, "step": 7351 }, { "epoch": 0.66, "learning_rate": 1.2893137439695008e-05, "loss": 1.2216, "step": 7352 }, { "epoch": 0.66, "learning_rate": 1.2886865963172484e-05, "loss": 1.2287, "step": 7353 }, { "epoch": 0.66, "learning_rate": 1.288059548263362e-05, "loss": 1.179, "step": 7354 }, { "epoch": 0.66, "learning_rate": 1.2874325998594016e-05, "loss": 1.2306, "step": 7355 }, { "epoch": 0.66, "learning_rate": 1.2868057511569164e-05, "loss": 1.2424, "step": 7356 }, { "epoch": 0.66, "learning_rate": 1.2861790022074482e-05, "loss": 1.2687, "step": 7357 }, { "epoch": 0.66, "learning_rate": 1.2855523530625296e-05, "loss": 1.1451, "step": 7358 }, { "epoch": 0.66, "learning_rate": 1.2849258037736855e-05, "loss": 1.1829, "step": 7359 }, { "epoch": 0.66, "learning_rate": 1.2842993543924348e-05, "loss": 1.2048, "step": 7360 }, { "epoch": 0.67, "learning_rate": 1.2836730049702856e-05, "loss": 1.1725, "step": 7361 }, { "epoch": 0.67, "learning_rate": 1.2830467555587367e-05, "loss": 1.2078, "step": 7362 }, { "epoch": 0.67, "learning_rate": 1.2824206062092841e-05, "loss": 1.1882, "step": 7363 }, { "epoch": 0.67, "learning_rate": 1.2817945569734085e-05, "loss": 1.2678, "step": 7364 }, { "epoch": 0.67, "learning_rate": 1.2811686079025875e-05, "loss": 1.1993, "step": 7365 }, { "epoch": 0.67, "learning_rate": 1.2805427590482892e-05, "loss": 1.192, "step": 7366 }, { "epoch": 0.67, "learning_rate": 1.279917010461971e-05, "loss": 1.1678, "step": 7367 }, { "epoch": 0.67, "learning_rate": 1.2792913621950875e-05, "loss": 1.2343, "step": 7368 }, { "epoch": 0.67, "learning_rate": 1.2786658142990777e-05, "loss": 1.1865, "step": 7369 }, { "epoch": 0.67, "learning_rate": 1.2780403668253788e-05, "loss": 1.2846, "step": 7370 }, { "epoch": 0.67, "learning_rate": 1.277415019825417e-05, "loss": 1.1963, "step": 7371 }, { "epoch": 0.67, "learning_rate": 1.2767897733506092e-05, "loss": 1.2105, "step": 7372 }, { "epoch": 0.67, "learning_rate": 1.2761646274523684e-05, "loss": 1.2403, "step": 7373 }, { "epoch": 0.67, "learning_rate": 1.2755395821820931e-05, "loss": 1.2247, "step": 7374 }, { "epoch": 0.67, "learning_rate": 1.2749146375911766e-05, "loss": 1.2907, "step": 7375 }, { "epoch": 0.67, "learning_rate": 1.274289793731006e-05, "loss": 1.1962, "step": 7376 }, { "epoch": 0.67, "learning_rate": 1.2736650506529576e-05, "loss": 1.2253, "step": 7377 }, { "epoch": 0.67, "learning_rate": 1.2730404084083994e-05, "loss": 1.2555, "step": 7378 }, { "epoch": 0.67, "learning_rate": 1.2724158670486918e-05, "loss": 1.2325, "step": 7379 }, { "epoch": 0.67, "learning_rate": 1.2717914266251856e-05, "loss": 1.2256, "step": 7380 }, { "epoch": 0.67, "learning_rate": 1.271167087189227e-05, "loss": 1.2316, "step": 7381 }, { "epoch": 0.67, "learning_rate": 1.2705428487921498e-05, "loss": 1.1956, "step": 7382 }, { "epoch": 0.67, "learning_rate": 1.2699187114852812e-05, "loss": 1.2599, "step": 7383 }, { "epoch": 0.67, "learning_rate": 1.2692946753199397e-05, "loss": 1.2686, "step": 7384 }, { "epoch": 0.67, "learning_rate": 1.2686707403474352e-05, "loss": 1.2606, "step": 7385 }, { "epoch": 0.67, "learning_rate": 1.2680469066190712e-05, "loss": 1.2876, "step": 7386 }, { "epoch": 0.67, "learning_rate": 1.267423174186141e-05, "loss": 1.2596, "step": 7387 }, { "epoch": 0.67, "learning_rate": 1.2667995430999286e-05, "loss": 1.2393, "step": 7388 }, { "epoch": 0.67, "learning_rate": 1.2661760134117146e-05, "loss": 1.2106, "step": 7389 }, { "epoch": 0.67, "learning_rate": 1.265552585172764e-05, "loss": 1.2206, "step": 7390 }, { "epoch": 0.67, "learning_rate": 1.264929258434338e-05, "loss": 1.2178, "step": 7391 }, { "epoch": 0.67, "learning_rate": 1.2643060332476903e-05, "loss": 1.2161, "step": 7392 }, { "epoch": 0.67, "learning_rate": 1.2636829096640627e-05, "loss": 1.1846, "step": 7393 }, { "epoch": 0.67, "learning_rate": 1.2630598877346934e-05, "loss": 1.205, "step": 7394 }, { "epoch": 0.67, "learning_rate": 1.2624369675108067e-05, "loss": 1.2031, "step": 7395 }, { "epoch": 0.67, "learning_rate": 1.2618141490436207e-05, "loss": 1.2299, "step": 7396 }, { "epoch": 0.67, "learning_rate": 1.2611914323843488e-05, "loss": 1.1915, "step": 7397 }, { "epoch": 0.67, "learning_rate": 1.2605688175841906e-05, "loss": 1.1826, "step": 7398 }, { "epoch": 0.67, "learning_rate": 1.2599463046943405e-05, "loss": 1.216, "step": 7399 }, { "epoch": 0.67, "learning_rate": 1.2593238937659835e-05, "loss": 1.1902, "step": 7400 }, { "epoch": 0.67, "learning_rate": 1.2587015848502954e-05, "loss": 1.2562, "step": 7401 }, { "epoch": 0.67, "learning_rate": 1.258079377998446e-05, "loss": 1.2117, "step": 7402 }, { "epoch": 0.67, "learning_rate": 1.257457273261595e-05, "loss": 1.1926, "step": 7403 }, { "epoch": 0.67, "learning_rate": 1.2568352706908937e-05, "loss": 1.1883, "step": 7404 }, { "epoch": 0.67, "learning_rate": 1.2562133703374851e-05, "loss": 1.2718, "step": 7405 }, { "epoch": 0.67, "learning_rate": 1.2555915722525035e-05, "loss": 1.2644, "step": 7406 }, { "epoch": 0.67, "learning_rate": 1.254969876487077e-05, "loss": 1.2311, "step": 7407 }, { "epoch": 0.67, "learning_rate": 1.2543482830923228e-05, "loss": 1.2226, "step": 7408 }, { "epoch": 0.67, "learning_rate": 1.2537267921193502e-05, "loss": 1.1949, "step": 7409 }, { "epoch": 0.67, "learning_rate": 1.2531054036192602e-05, "loss": 1.1733, "step": 7410 }, { "epoch": 0.67, "learning_rate": 1.2524841176431457e-05, "loss": 1.2306, "step": 7411 }, { "epoch": 0.67, "learning_rate": 1.25186293424209e-05, "loss": 1.1972, "step": 7412 }, { "epoch": 0.67, "learning_rate": 1.2512418534671708e-05, "loss": 1.2571, "step": 7413 }, { "epoch": 0.67, "learning_rate": 1.2506208753694537e-05, "loss": 1.1691, "step": 7414 }, { "epoch": 0.67, "learning_rate": 1.2500000000000006e-05, "loss": 1.1966, "step": 7415 }, { "epoch": 0.67, "learning_rate": 1.249379227409859e-05, "loss": 1.2336, "step": 7416 }, { "epoch": 0.67, "learning_rate": 1.2487585576500706e-05, "loss": 1.2478, "step": 7417 }, { "epoch": 0.67, "learning_rate": 1.248137990771672e-05, "loss": 1.223, "step": 7418 }, { "epoch": 0.67, "learning_rate": 1.2475175268256861e-05, "loss": 1.2505, "step": 7419 }, { "epoch": 0.67, "learning_rate": 1.2468971658631306e-05, "loss": 1.2155, "step": 7420 }, { "epoch": 0.67, "learning_rate": 1.2462769079350132e-05, "loss": 1.1766, "step": 7421 }, { "epoch": 0.67, "learning_rate": 1.2456567530923328e-05, "loss": 1.2508, "step": 7422 }, { "epoch": 0.67, "learning_rate": 1.2450367013860825e-05, "loss": 1.2157, "step": 7423 }, { "epoch": 0.67, "learning_rate": 1.2444167528672443e-05, "loss": 1.2059, "step": 7424 }, { "epoch": 0.67, "learning_rate": 1.2437969075867922e-05, "loss": 1.2392, "step": 7425 }, { "epoch": 0.67, "learning_rate": 1.2431771655956925e-05, "loss": 1.2227, "step": 7426 }, { "epoch": 0.67, "learning_rate": 1.2425575269449014e-05, "loss": 1.2066, "step": 7427 }, { "epoch": 0.67, "learning_rate": 1.2419379916853693e-05, "loss": 1.1921, "step": 7428 }, { "epoch": 0.67, "learning_rate": 1.2413185598680357e-05, "loss": 1.254, "step": 7429 }, { "epoch": 0.67, "learning_rate": 1.2406992315438329e-05, "loss": 1.198, "step": 7430 }, { "epoch": 0.67, "learning_rate": 1.2400800067636836e-05, "loss": 1.2056, "step": 7431 }, { "epoch": 0.67, "learning_rate": 1.239460885578503e-05, "loss": 1.1993, "step": 7432 }, { "epoch": 0.67, "learning_rate": 1.2388418680391963e-05, "loss": 1.2276, "step": 7433 }, { "epoch": 0.67, "learning_rate": 1.238222954196663e-05, "loss": 1.2008, "step": 7434 }, { "epoch": 0.67, "learning_rate": 1.2376041441017916e-05, "loss": 1.1995, "step": 7435 }, { "epoch": 0.67, "learning_rate": 1.2369854378054629e-05, "loss": 1.2307, "step": 7436 }, { "epoch": 0.67, "learning_rate": 1.2363668353585487e-05, "loss": 1.2942, "step": 7437 }, { "epoch": 0.67, "learning_rate": 1.235748336811912e-05, "loss": 1.272, "step": 7438 }, { "epoch": 0.67, "learning_rate": 1.2351299422164094e-05, "loss": 1.2067, "step": 7439 }, { "epoch": 0.67, "learning_rate": 1.2345116516228868e-05, "loss": 1.2605, "step": 7440 }, { "epoch": 0.67, "learning_rate": 1.2338934650821823e-05, "loss": 1.1698, "step": 7441 }, { "epoch": 0.67, "learning_rate": 1.2332753826451249e-05, "loss": 1.2007, "step": 7442 }, { "epoch": 0.67, "learning_rate": 1.232657404362535e-05, "loss": 1.2007, "step": 7443 }, { "epoch": 0.67, "learning_rate": 1.2320395302852264e-05, "loss": 1.2468, "step": 7444 }, { "epoch": 0.67, "learning_rate": 1.2314217604640022e-05, "loss": 1.1811, "step": 7445 }, { "epoch": 0.67, "learning_rate": 1.2308040949496574e-05, "loss": 1.2619, "step": 7446 }, { "epoch": 0.67, "learning_rate": 1.2301865337929785e-05, "loss": 1.2018, "step": 7447 }, { "epoch": 0.67, "learning_rate": 1.2295690770447429e-05, "loss": 1.1575, "step": 7448 }, { "epoch": 0.67, "learning_rate": 1.2289517247557217e-05, "loss": 1.2316, "step": 7449 }, { "epoch": 0.67, "learning_rate": 1.2283344769766748e-05, "loss": 1.2297, "step": 7450 }, { "epoch": 0.67, "learning_rate": 1.2277173337583542e-05, "loss": 1.1643, "step": 7451 }, { "epoch": 0.67, "learning_rate": 1.2271002951515038e-05, "loss": 1.2777, "step": 7452 }, { "epoch": 0.67, "learning_rate": 1.2264833612068588e-05, "loss": 1.1804, "step": 7453 }, { "epoch": 0.67, "learning_rate": 1.2258665319751444e-05, "loss": 1.2337, "step": 7454 }, { "epoch": 0.67, "learning_rate": 1.2252498075070803e-05, "loss": 1.2473, "step": 7455 }, { "epoch": 0.67, "learning_rate": 1.2246331878533752e-05, "loss": 1.091, "step": 7456 }, { "epoch": 0.67, "learning_rate": 1.2240166730647294e-05, "loss": 1.194, "step": 7457 }, { "epoch": 0.67, "learning_rate": 1.2234002631918345e-05, "loss": 1.1932, "step": 7458 }, { "epoch": 0.67, "learning_rate": 1.2227839582853737e-05, "loss": 1.2287, "step": 7459 }, { "epoch": 0.67, "learning_rate": 1.2221677583960229e-05, "loss": 1.1949, "step": 7460 }, { "epoch": 0.67, "learning_rate": 1.2215516635744476e-05, "loss": 1.1813, "step": 7461 }, { "epoch": 0.67, "learning_rate": 1.2209356738713051e-05, "loss": 1.1675, "step": 7462 }, { "epoch": 0.67, "learning_rate": 1.2203197893372441e-05, "loss": 1.2397, "step": 7463 }, { "epoch": 0.67, "learning_rate": 1.2197040100229042e-05, "loss": 1.1939, "step": 7464 }, { "epoch": 0.67, "learning_rate": 1.2190883359789185e-05, "loss": 1.1955, "step": 7465 }, { "epoch": 0.67, "learning_rate": 1.2184727672559091e-05, "loss": 1.2295, "step": 7466 }, { "epoch": 0.67, "learning_rate": 1.2178573039044897e-05, "loss": 1.2239, "step": 7467 }, { "epoch": 0.67, "learning_rate": 1.2172419459752663e-05, "loss": 1.1666, "step": 7468 }, { "epoch": 0.67, "learning_rate": 1.2166266935188348e-05, "loss": 1.2289, "step": 7469 }, { "epoch": 0.67, "learning_rate": 1.216011546585785e-05, "loss": 1.2211, "step": 7470 }, { "epoch": 0.68, "learning_rate": 1.2153965052266956e-05, "loss": 1.259, "step": 7471 }, { "epoch": 0.68, "learning_rate": 1.2147815694921371e-05, "loss": 1.2255, "step": 7472 }, { "epoch": 0.68, "learning_rate": 1.2141667394326723e-05, "loss": 1.1819, "step": 7473 }, { "epoch": 0.68, "learning_rate": 1.2135520150988541e-05, "loss": 1.2019, "step": 7474 }, { "epoch": 0.68, "learning_rate": 1.2129373965412267e-05, "loss": 1.1337, "step": 7475 }, { "epoch": 0.68, "learning_rate": 1.212322883810328e-05, "loss": 1.1988, "step": 7476 }, { "epoch": 0.68, "learning_rate": 1.2117084769566838e-05, "loss": 1.2347, "step": 7477 }, { "epoch": 0.68, "learning_rate": 1.2110941760308134e-05, "loss": 1.2188, "step": 7478 }, { "epoch": 0.68, "learning_rate": 1.2104799810832265e-05, "loss": 1.1716, "step": 7479 }, { "epoch": 0.68, "learning_rate": 1.2098658921644237e-05, "loss": 1.1968, "step": 7480 }, { "epoch": 0.68, "learning_rate": 1.2092519093248988e-05, "loss": 1.2138, "step": 7481 }, { "epoch": 0.68, "learning_rate": 1.2086380326151353e-05, "loss": 1.2097, "step": 7482 }, { "epoch": 0.68, "learning_rate": 1.2080242620856078e-05, "loss": 1.268, "step": 7483 }, { "epoch": 0.68, "learning_rate": 1.2074105977867829e-05, "loss": 1.1918, "step": 7484 }, { "epoch": 0.68, "learning_rate": 1.2067970397691167e-05, "loss": 1.1905, "step": 7485 }, { "epoch": 0.68, "learning_rate": 1.2061835880830608e-05, "loss": 1.2239, "step": 7486 }, { "epoch": 0.68, "learning_rate": 1.2055702427790538e-05, "loss": 1.2478, "step": 7487 }, { "epoch": 0.68, "learning_rate": 1.2049570039075272e-05, "loss": 1.1926, "step": 7488 }, { "epoch": 0.68, "learning_rate": 1.2043438715189035e-05, "loss": 1.247, "step": 7489 }, { "epoch": 0.68, "learning_rate": 1.203730845663596e-05, "loss": 1.2078, "step": 7490 }, { "epoch": 0.68, "learning_rate": 1.2031179263920112e-05, "loss": 1.2238, "step": 7491 }, { "epoch": 0.68, "learning_rate": 1.2025051137545449e-05, "loss": 1.1693, "step": 7492 }, { "epoch": 0.68, "learning_rate": 1.2018924078015843e-05, "loss": 1.198, "step": 7493 }, { "epoch": 0.68, "learning_rate": 1.2012798085835084e-05, "loss": 1.1985, "step": 7494 }, { "epoch": 0.68, "learning_rate": 1.2006673161506873e-05, "loss": 1.2871, "step": 7495 }, { "epoch": 0.68, "learning_rate": 1.2000549305534812e-05, "loss": 1.1804, "step": 7496 }, { "epoch": 0.68, "learning_rate": 1.1994426518422444e-05, "loss": 1.2492, "step": 7497 }, { "epoch": 0.68, "learning_rate": 1.1988304800673197e-05, "loss": 1.2587, "step": 7498 }, { "epoch": 0.68, "learning_rate": 1.1982184152790418e-05, "loss": 1.1678, "step": 7499 }, { "epoch": 0.68, "learning_rate": 1.1976064575277371e-05, "loss": 1.2375, "step": 7500 }, { "epoch": 0.68, "learning_rate": 1.1969946068637217e-05, "loss": 1.2439, "step": 7501 }, { "epoch": 0.68, "learning_rate": 1.1963828633373061e-05, "loss": 1.2303, "step": 7502 }, { "epoch": 0.68, "learning_rate": 1.195771226998789e-05, "loss": 1.2079, "step": 7503 }, { "epoch": 0.68, "learning_rate": 1.195159697898461e-05, "loss": 1.2005, "step": 7504 }, { "epoch": 0.68, "learning_rate": 1.1945482760866044e-05, "loss": 1.2227, "step": 7505 }, { "epoch": 0.68, "learning_rate": 1.1939369616134913e-05, "loss": 1.2055, "step": 7506 }, { "epoch": 0.68, "learning_rate": 1.1933257545293883e-05, "loss": 1.186, "step": 7507 }, { "epoch": 0.68, "learning_rate": 1.1927146548845493e-05, "loss": 1.2298, "step": 7508 }, { "epoch": 0.68, "learning_rate": 1.192103662729222e-05, "loss": 1.2254, "step": 7509 }, { "epoch": 0.68, "learning_rate": 1.1914927781136437e-05, "loss": 1.1912, "step": 7510 }, { "epoch": 0.68, "learning_rate": 1.1908820010880423e-05, "loss": 1.2392, "step": 7511 }, { "epoch": 0.68, "learning_rate": 1.1902713317026403e-05, "loss": 1.1497, "step": 7512 }, { "epoch": 0.68, "learning_rate": 1.1896607700076481e-05, "loss": 1.231, "step": 7513 }, { "epoch": 0.68, "learning_rate": 1.1890503160532678e-05, "loss": 1.1956, "step": 7514 }, { "epoch": 0.68, "learning_rate": 1.1884399698896934e-05, "loss": 1.2761, "step": 7515 }, { "epoch": 0.68, "learning_rate": 1.1878297315671097e-05, "loss": 1.1974, "step": 7516 }, { "epoch": 0.68, "learning_rate": 1.1872196011356911e-05, "loss": 1.2096, "step": 7517 }, { "epoch": 0.68, "learning_rate": 1.1866095786456073e-05, "loss": 1.2054, "step": 7518 }, { "epoch": 0.68, "learning_rate": 1.185999664147015e-05, "loss": 1.1983, "step": 7519 }, { "epoch": 0.68, "learning_rate": 1.1853898576900635e-05, "loss": 1.3275, "step": 7520 }, { "epoch": 0.68, "learning_rate": 1.1847801593248933e-05, "loss": 1.1865, "step": 7521 }, { "epoch": 0.68, "learning_rate": 1.184170569101635e-05, "loss": 1.1823, "step": 7522 }, { "epoch": 0.68, "learning_rate": 1.1835610870704131e-05, "loss": 1.2531, "step": 7523 }, { "epoch": 0.68, "learning_rate": 1.1829517132813403e-05, "loss": 1.2091, "step": 7524 }, { "epoch": 0.68, "learning_rate": 1.1823424477845213e-05, "loss": 1.2219, "step": 7525 }, { "epoch": 0.68, "learning_rate": 1.181733290630052e-05, "loss": 1.2354, "step": 7526 }, { "epoch": 0.68, "learning_rate": 1.1811242418680191e-05, "loss": 1.2299, "step": 7527 }, { "epoch": 0.68, "learning_rate": 1.1805153015485017e-05, "loss": 1.219, "step": 7528 }, { "epoch": 0.68, "learning_rate": 1.1799064697215687e-05, "loss": 1.2005, "step": 7529 }, { "epoch": 0.68, "learning_rate": 1.1792977464372799e-05, "loss": 1.2619, "step": 7530 }, { "epoch": 0.68, "learning_rate": 1.1786891317456867e-05, "loss": 1.1711, "step": 7531 }, { "epoch": 0.68, "learning_rate": 1.1780806256968307e-05, "loss": 1.2338, "step": 7532 }, { "epoch": 0.68, "learning_rate": 1.1774722283407474e-05, "loss": 1.1886, "step": 7533 }, { "epoch": 0.68, "learning_rate": 1.1768639397274608e-05, "loss": 1.221, "step": 7534 }, { "epoch": 0.68, "learning_rate": 1.1762557599069843e-05, "loss": 1.2657, "step": 7535 }, { "epoch": 0.68, "learning_rate": 1.1756476889293269e-05, "loss": 1.2293, "step": 7536 }, { "epoch": 0.68, "learning_rate": 1.1750397268444857e-05, "loss": 1.2203, "step": 7537 }, { "epoch": 0.68, "learning_rate": 1.1744318737024482e-05, "loss": 1.1922, "step": 7538 }, { "epoch": 0.68, "learning_rate": 1.173824129553196e-05, "loss": 1.2294, "step": 7539 }, { "epoch": 0.68, "learning_rate": 1.1732164944466994e-05, "loss": 1.1781, "step": 7540 }, { "epoch": 0.68, "learning_rate": 1.1726089684329197e-05, "loss": 1.2374, "step": 7541 }, { "epoch": 0.68, "learning_rate": 1.1720015515618105e-05, "loss": 1.1955, "step": 7542 }, { "epoch": 0.68, "learning_rate": 1.1713942438833141e-05, "loss": 1.2053, "step": 7543 }, { "epoch": 0.68, "learning_rate": 1.1707870454473676e-05, "loss": 1.2002, "step": 7544 }, { "epoch": 0.68, "learning_rate": 1.1701799563038962e-05, "loss": 1.2089, "step": 7545 }, { "epoch": 0.68, "learning_rate": 1.1695729765028166e-05, "loss": 1.2027, "step": 7546 }, { "epoch": 0.68, "learning_rate": 1.1689661060940366e-05, "loss": 1.2383, "step": 7547 }, { "epoch": 0.68, "learning_rate": 1.1683593451274547e-05, "loss": 1.2274, "step": 7548 }, { "epoch": 0.68, "learning_rate": 1.1677526936529622e-05, "loss": 1.2296, "step": 7549 }, { "epoch": 0.68, "learning_rate": 1.16714615172044e-05, "loss": 1.2155, "step": 7550 }, { "epoch": 0.68, "learning_rate": 1.166539719379759e-05, "loss": 1.2079, "step": 7551 }, { "epoch": 0.68, "learning_rate": 1.1659333966807832e-05, "loss": 1.1786, "step": 7552 }, { "epoch": 0.68, "learning_rate": 1.1653271836733645e-05, "loss": 1.183, "step": 7553 }, { "epoch": 0.68, "learning_rate": 1.1647210804073508e-05, "loss": 1.2186, "step": 7554 }, { "epoch": 0.68, "learning_rate": 1.1641150869325773e-05, "loss": 1.2173, "step": 7555 }, { "epoch": 0.68, "learning_rate": 1.1635092032988682e-05, "loss": 1.2533, "step": 7556 }, { "epoch": 0.68, "learning_rate": 1.1629034295560442e-05, "loss": 1.2022, "step": 7557 }, { "epoch": 0.68, "learning_rate": 1.1622977657539133e-05, "loss": 1.2086, "step": 7558 }, { "epoch": 0.68, "learning_rate": 1.1616922119422744e-05, "loss": 1.1834, "step": 7559 }, { "epoch": 0.68, "learning_rate": 1.1610867681709206e-05, "loss": 1.2121, "step": 7560 }, { "epoch": 0.68, "learning_rate": 1.16048143448963e-05, "loss": 1.2199, "step": 7561 }, { "epoch": 0.68, "learning_rate": 1.1598762109481784e-05, "loss": 1.1924, "step": 7562 }, { "epoch": 0.68, "learning_rate": 1.159271097596328e-05, "loss": 1.2478, "step": 7563 }, { "epoch": 0.68, "learning_rate": 1.1586660944838326e-05, "loss": 1.1958, "step": 7564 }, { "epoch": 0.68, "learning_rate": 1.1580612016604392e-05, "loss": 1.238, "step": 7565 }, { "epoch": 0.68, "learning_rate": 1.1574564191758838e-05, "loss": 1.1576, "step": 7566 }, { "epoch": 0.68, "learning_rate": 1.1568517470798931e-05, "loss": 1.227, "step": 7567 }, { "epoch": 0.68, "learning_rate": 1.1562471854221857e-05, "loss": 1.2174, "step": 7568 }, { "epoch": 0.68, "learning_rate": 1.1556427342524698e-05, "loss": 1.2489, "step": 7569 }, { "epoch": 0.68, "learning_rate": 1.155038393620447e-05, "loss": 1.2237, "step": 7570 }, { "epoch": 0.68, "learning_rate": 1.1544341635758078e-05, "loss": 1.2434, "step": 7571 }, { "epoch": 0.68, "learning_rate": 1.1538300441682336e-05, "loss": 1.1384, "step": 7572 }, { "epoch": 0.68, "learning_rate": 1.1532260354473978e-05, "loss": 1.1903, "step": 7573 }, { "epoch": 0.68, "learning_rate": 1.1526221374629626e-05, "loss": 1.2232, "step": 7574 }, { "epoch": 0.68, "learning_rate": 1.1520183502645845e-05, "loss": 1.217, "step": 7575 }, { "epoch": 0.68, "learning_rate": 1.1514146739019094e-05, "loss": 1.2292, "step": 7576 }, { "epoch": 0.68, "learning_rate": 1.1508111084245705e-05, "loss": 1.2357, "step": 7577 }, { "epoch": 0.68, "learning_rate": 1.1502076538821977e-05, "loss": 1.2608, "step": 7578 }, { "epoch": 0.68, "learning_rate": 1.1496043103244084e-05, "loss": 1.2163, "step": 7579 }, { "epoch": 0.68, "learning_rate": 1.1490010778008106e-05, "loss": 1.2107, "step": 7580 }, { "epoch": 0.68, "learning_rate": 1.148397956361007e-05, "loss": 1.2219, "step": 7581 }, { "epoch": 0.69, "learning_rate": 1.1477949460545845e-05, "loss": 1.2714, "step": 7582 }, { "epoch": 0.69, "learning_rate": 1.1471920469311275e-05, "loss": 1.2288, "step": 7583 }, { "epoch": 0.69, "learning_rate": 1.146589259040208e-05, "loss": 1.1961, "step": 7584 }, { "epoch": 0.69, "learning_rate": 1.1459865824313876e-05, "loss": 1.2436, "step": 7585 }, { "epoch": 0.69, "learning_rate": 1.1453840171542235e-05, "loss": 1.2432, "step": 7586 }, { "epoch": 0.69, "learning_rate": 1.1447815632582574e-05, "loss": 1.2639, "step": 7587 }, { "epoch": 0.69, "learning_rate": 1.1441792207930274e-05, "loss": 1.2423, "step": 7588 }, { "epoch": 0.69, "learning_rate": 1.1435769898080592e-05, "loss": 1.2793, "step": 7589 }, { "epoch": 0.69, "learning_rate": 1.1429748703528701e-05, "loss": 1.2189, "step": 7590 }, { "epoch": 0.69, "learning_rate": 1.1423728624769695e-05, "loss": 1.2311, "step": 7591 }, { "epoch": 0.69, "learning_rate": 1.1417709662298558e-05, "loss": 1.1641, "step": 7592 }, { "epoch": 0.69, "learning_rate": 1.1411691816610194e-05, "loss": 1.1848, "step": 7593 }, { "epoch": 0.69, "learning_rate": 1.1405675088199408e-05, "loss": 1.1998, "step": 7594 }, { "epoch": 0.69, "learning_rate": 1.1399659477560906e-05, "loss": 1.1644, "step": 7595 }, { "epoch": 0.69, "learning_rate": 1.1393644985189333e-05, "loss": 1.2496, "step": 7596 }, { "epoch": 0.69, "learning_rate": 1.1387631611579219e-05, "loss": 1.201, "step": 7597 }, { "epoch": 0.69, "learning_rate": 1.138161935722498e-05, "loss": 1.2473, "step": 7598 }, { "epoch": 0.69, "learning_rate": 1.1375608222620987e-05, "loss": 1.2311, "step": 7599 }, { "epoch": 0.69, "learning_rate": 1.1369598208261492e-05, "loss": 1.1647, "step": 7600 }, { "epoch": 0.69, "learning_rate": 1.136358931464065e-05, "loss": 1.2136, "step": 7601 }, { "epoch": 0.69, "learning_rate": 1.1357581542252555e-05, "loss": 1.2576, "step": 7602 }, { "epoch": 0.69, "learning_rate": 1.1351574891591151e-05, "loss": 1.2468, "step": 7603 }, { "epoch": 0.69, "learning_rate": 1.134556936315036e-05, "loss": 1.2286, "step": 7604 }, { "epoch": 0.69, "learning_rate": 1.1339564957423962e-05, "loss": 1.2522, "step": 7605 }, { "epoch": 0.69, "learning_rate": 1.1333561674905655e-05, "loss": 1.2201, "step": 7606 }, { "epoch": 0.69, "learning_rate": 1.132755951608907e-05, "loss": 1.2147, "step": 7607 }, { "epoch": 0.69, "learning_rate": 1.1321558481467695e-05, "loss": 1.1832, "step": 7608 }, { "epoch": 0.69, "learning_rate": 1.1315558571534982e-05, "loss": 1.2249, "step": 7609 }, { "epoch": 0.69, "learning_rate": 1.1309559786784255e-05, "loss": 1.1735, "step": 7610 }, { "epoch": 0.69, "learning_rate": 1.1303562127708744e-05, "loss": 1.2387, "step": 7611 }, { "epoch": 0.69, "learning_rate": 1.1297565594801615e-05, "loss": 1.1956, "step": 7612 }, { "epoch": 0.69, "learning_rate": 1.1291570188555919e-05, "loss": 1.1862, "step": 7613 }, { "epoch": 0.69, "learning_rate": 1.1285575909464613e-05, "loss": 1.2356, "step": 7614 }, { "epoch": 0.69, "learning_rate": 1.1279582758020574e-05, "loss": 1.1746, "step": 7615 }, { "epoch": 0.69, "learning_rate": 1.1273590734716562e-05, "loss": 1.1934, "step": 7616 }, { "epoch": 0.69, "learning_rate": 1.1267599840045287e-05, "loss": 1.2196, "step": 7617 }, { "epoch": 0.69, "learning_rate": 1.1261610074499341e-05, "loss": 1.2365, "step": 7618 }, { "epoch": 0.69, "learning_rate": 1.125562143857119e-05, "loss": 1.2139, "step": 7619 }, { "epoch": 0.69, "learning_rate": 1.1249633932753276e-05, "loss": 1.1978, "step": 7620 }, { "epoch": 0.69, "learning_rate": 1.1243647557537895e-05, "loss": 1.2181, "step": 7621 }, { "epoch": 0.69, "learning_rate": 1.1237662313417263e-05, "loss": 1.2236, "step": 7622 }, { "epoch": 0.69, "learning_rate": 1.1231678200883533e-05, "loss": 1.2072, "step": 7623 }, { "epoch": 0.69, "learning_rate": 1.1225695220428703e-05, "loss": 1.182, "step": 7624 }, { "epoch": 0.69, "learning_rate": 1.1219713372544746e-05, "loss": 1.2023, "step": 7625 }, { "epoch": 0.69, "learning_rate": 1.1213732657723494e-05, "loss": 1.1716, "step": 7626 }, { "epoch": 0.69, "learning_rate": 1.1207753076456698e-05, "loss": 1.2595, "step": 7627 }, { "epoch": 0.69, "learning_rate": 1.1201774629236042e-05, "loss": 1.2062, "step": 7628 }, { "epoch": 0.69, "learning_rate": 1.1195797316553064e-05, "loss": 1.2201, "step": 7629 }, { "epoch": 0.69, "learning_rate": 1.1189821138899265e-05, "loss": 1.1826, "step": 7630 }, { "epoch": 0.69, "learning_rate": 1.1183846096766015e-05, "loss": 1.2145, "step": 7631 }, { "epoch": 0.69, "learning_rate": 1.1177872190644597e-05, "loss": 1.2023, "step": 7632 }, { "epoch": 0.69, "learning_rate": 1.117189942102623e-05, "loss": 1.2492, "step": 7633 }, { "epoch": 0.69, "learning_rate": 1.1165927788401984e-05, "loss": 1.1902, "step": 7634 }, { "epoch": 0.69, "learning_rate": 1.1159957293262888e-05, "loss": 1.2197, "step": 7635 }, { "epoch": 0.69, "learning_rate": 1.1153987936099856e-05, "loss": 1.2268, "step": 7636 }, { "epoch": 0.69, "learning_rate": 1.114801971740369e-05, "loss": 1.2, "step": 7637 }, { "epoch": 0.69, "learning_rate": 1.1142052637665142e-05, "loss": 1.2464, "step": 7638 }, { "epoch": 0.69, "learning_rate": 1.1136086697374848e-05, "loss": 1.2496, "step": 7639 }, { "epoch": 0.69, "learning_rate": 1.1130121897023318e-05, "loss": 1.2072, "step": 7640 }, { "epoch": 0.69, "learning_rate": 1.1124158237101023e-05, "loss": 1.2353, "step": 7641 }, { "epoch": 0.69, "learning_rate": 1.1118195718098311e-05, "loss": 1.2302, "step": 7642 }, { "epoch": 0.69, "learning_rate": 1.1112234340505429e-05, "loss": 1.2323, "step": 7643 }, { "epoch": 0.69, "learning_rate": 1.1106274104812573e-05, "loss": 1.2213, "step": 7644 }, { "epoch": 0.69, "learning_rate": 1.110031501150977e-05, "loss": 1.251, "step": 7645 }, { "epoch": 0.69, "learning_rate": 1.1094357061087033e-05, "loss": 1.2083, "step": 7646 }, { "epoch": 0.69, "learning_rate": 1.1088400254034231e-05, "loss": 1.2246, "step": 7647 }, { "epoch": 0.69, "learning_rate": 1.1082444590841149e-05, "loss": 1.1942, "step": 7648 }, { "epoch": 0.69, "learning_rate": 1.1076490071997503e-05, "loss": 1.2062, "step": 7649 }, { "epoch": 0.69, "learning_rate": 1.1070536697992864e-05, "loss": 1.174, "step": 7650 }, { "epoch": 0.69, "learning_rate": 1.106458446931676e-05, "loss": 1.1786, "step": 7651 }, { "epoch": 0.69, "learning_rate": 1.10586333864586e-05, "loss": 1.1788, "step": 7652 }, { "epoch": 0.69, "learning_rate": 1.1052683449907691e-05, "loss": 1.2197, "step": 7653 }, { "epoch": 0.69, "learning_rate": 1.1046734660153285e-05, "loss": 1.2, "step": 7654 }, { "epoch": 0.69, "learning_rate": 1.1040787017684476e-05, "loss": 1.1671, "step": 7655 }, { "epoch": 0.69, "learning_rate": 1.1034840522990325e-05, "loss": 1.2505, "step": 7656 }, { "epoch": 0.69, "learning_rate": 1.1028895176559767e-05, "loss": 1.2035, "step": 7657 }, { "epoch": 0.69, "learning_rate": 1.1022950978881638e-05, "loss": 1.2585, "step": 7658 }, { "epoch": 0.69, "learning_rate": 1.101700793044472e-05, "loss": 1.1875, "step": 7659 }, { "epoch": 0.69, "learning_rate": 1.101106603173764e-05, "loss": 1.2779, "step": 7660 }, { "epoch": 0.69, "learning_rate": 1.1005125283248965e-05, "loss": 1.2432, "step": 7661 }, { "epoch": 0.69, "learning_rate": 1.0999185685467181e-05, "loss": 1.259, "step": 7662 }, { "epoch": 0.69, "learning_rate": 1.0993247238880649e-05, "loss": 1.2584, "step": 7663 }, { "epoch": 0.69, "learning_rate": 1.0987309943977647e-05, "loss": 1.2517, "step": 7664 }, { "epoch": 0.69, "learning_rate": 1.098137380124638e-05, "loss": 1.2749, "step": 7665 }, { "epoch": 0.69, "learning_rate": 1.097543881117491e-05, "loss": 1.2021, "step": 7666 }, { "epoch": 0.69, "learning_rate": 1.096950497425125e-05, "loss": 1.177, "step": 7667 }, { "epoch": 0.69, "learning_rate": 1.0963572290963298e-05, "loss": 1.1821, "step": 7668 }, { "epoch": 0.69, "learning_rate": 1.0957640761798845e-05, "loss": 1.1968, "step": 7669 }, { "epoch": 0.69, "learning_rate": 1.0951710387245636e-05, "loss": 1.2526, "step": 7670 }, { "epoch": 0.69, "learning_rate": 1.0945781167791248e-05, "loss": 1.2237, "step": 7671 }, { "epoch": 0.69, "learning_rate": 1.0939853103923228e-05, "loss": 1.2639, "step": 7672 }, { "epoch": 0.69, "learning_rate": 1.0933926196128991e-05, "loss": 1.2453, "step": 7673 }, { "epoch": 0.69, "learning_rate": 1.092800044489586e-05, "loss": 1.2563, "step": 7674 }, { "epoch": 0.69, "learning_rate": 1.0922075850711103e-05, "loss": 1.1884, "step": 7675 }, { "epoch": 0.69, "learning_rate": 1.0916152414061817e-05, "loss": 1.2216, "step": 7676 }, { "epoch": 0.69, "learning_rate": 1.091023013543508e-05, "loss": 1.1873, "step": 7677 }, { "epoch": 0.69, "learning_rate": 1.090430901531783e-05, "loss": 1.189, "step": 7678 }, { "epoch": 0.69, "learning_rate": 1.089838905419691e-05, "loss": 1.1737, "step": 7679 }, { "epoch": 0.69, "learning_rate": 1.0892470252559117e-05, "loss": 1.1819, "step": 7680 }, { "epoch": 0.69, "learning_rate": 1.0886552610891077e-05, "loss": 1.1759, "step": 7681 }, { "epoch": 0.69, "learning_rate": 1.0880636129679364e-05, "loss": 1.2659, "step": 7682 }, { "epoch": 0.69, "learning_rate": 1.0874720809410472e-05, "loss": 1.1959, "step": 7683 }, { "epoch": 0.69, "learning_rate": 1.0868806650570762e-05, "loss": 1.2219, "step": 7684 }, { "epoch": 0.69, "learning_rate": 1.0862893653646525e-05, "loss": 1.1912, "step": 7685 }, { "epoch": 0.69, "learning_rate": 1.0856981819123943e-05, "loss": 1.2033, "step": 7686 }, { "epoch": 0.69, "learning_rate": 1.0851071147489103e-05, "loss": 1.2109, "step": 7687 }, { "epoch": 0.69, "learning_rate": 1.0845161639228014e-05, "loss": 1.2276, "step": 7688 }, { "epoch": 0.69, "learning_rate": 1.0839253294826568e-05, "loss": 1.2564, "step": 7689 }, { "epoch": 0.69, "learning_rate": 1.0833346114770561e-05, "loss": 1.2521, "step": 7690 }, { "epoch": 0.69, "learning_rate": 1.0827440099545733e-05, "loss": 1.1636, "step": 7691 }, { "epoch": 0.69, "learning_rate": 1.0821535249637654e-05, "loss": 1.1907, "step": 7692 }, { "epoch": 0.7, "learning_rate": 1.0815631565531873e-05, "loss": 1.1883, "step": 7693 }, { "epoch": 0.7, "learning_rate": 1.0809729047713798e-05, "loss": 1.1799, "step": 7694 }, { "epoch": 0.7, "learning_rate": 1.080382769666875e-05, "loss": 1.2271, "step": 7695 }, { "epoch": 0.7, "learning_rate": 1.0797927512881983e-05, "loss": 1.2094, "step": 7696 }, { "epoch": 0.7, "learning_rate": 1.0792028496838596e-05, "loss": 1.2416, "step": 7697 }, { "epoch": 0.7, "learning_rate": 1.0786130649023651e-05, "loss": 1.2317, "step": 7698 }, { "epoch": 0.7, "learning_rate": 1.0780233969922082e-05, "loss": 1.2319, "step": 7699 }, { "epoch": 0.7, "learning_rate": 1.0774338460018727e-05, "loss": 1.173, "step": 7700 }, { "epoch": 0.7, "learning_rate": 1.0768444119798357e-05, "loss": 1.1699, "step": 7701 }, { "epoch": 0.7, "learning_rate": 1.07625509497456e-05, "loss": 1.2701, "step": 7702 }, { "epoch": 0.7, "learning_rate": 1.0756658950345017e-05, "loss": 1.1895, "step": 7703 }, { "epoch": 0.7, "learning_rate": 1.0750768122081085e-05, "loss": 1.217, "step": 7704 }, { "epoch": 0.7, "learning_rate": 1.0744878465438154e-05, "loss": 1.259, "step": 7705 }, { "epoch": 0.7, "learning_rate": 1.0738989980900496e-05, "loss": 1.2152, "step": 7706 }, { "epoch": 0.7, "learning_rate": 1.0733102668952282e-05, "loss": 1.2442, "step": 7707 }, { "epoch": 0.7, "learning_rate": 1.072721653007758e-05, "loss": 1.2383, "step": 7708 }, { "epoch": 0.7, "learning_rate": 1.0721331564760385e-05, "loss": 1.1995, "step": 7709 }, { "epoch": 0.7, "learning_rate": 1.071544777348457e-05, "loss": 1.2252, "step": 7710 }, { "epoch": 0.7, "learning_rate": 1.0709565156733919e-05, "loss": 1.2, "step": 7711 }, { "epoch": 0.7, "learning_rate": 1.0703683714992121e-05, "loss": 1.202, "step": 7712 }, { "epoch": 0.7, "learning_rate": 1.0697803448742766e-05, "loss": 1.2064, "step": 7713 }, { "epoch": 0.7, "learning_rate": 1.0691924358469357e-05, "loss": 1.2407, "step": 7714 }, { "epoch": 0.7, "learning_rate": 1.0686046444655295e-05, "loss": 1.1684, "step": 7715 }, { "epoch": 0.7, "learning_rate": 1.0680169707783866e-05, "loss": 1.2077, "step": 7716 }, { "epoch": 0.7, "learning_rate": 1.0674294148338304e-05, "loss": 1.2431, "step": 7717 }, { "epoch": 0.7, "learning_rate": 1.0668419766801685e-05, "loss": 1.1489, "step": 7718 }, { "epoch": 0.7, "learning_rate": 1.0662546563657044e-05, "loss": 1.2604, "step": 7719 }, { "epoch": 0.7, "learning_rate": 1.0656674539387287e-05, "loss": 1.2397, "step": 7720 }, { "epoch": 0.7, "learning_rate": 1.0650803694475228e-05, "loss": 1.2447, "step": 7721 }, { "epoch": 0.7, "learning_rate": 1.064493402940361e-05, "loss": 1.2544, "step": 7722 }, { "epoch": 0.7, "learning_rate": 1.0639065544655033e-05, "loss": 1.2454, "step": 7723 }, { "epoch": 0.7, "learning_rate": 1.0633198240712021e-05, "loss": 1.1915, "step": 7724 }, { "epoch": 0.7, "learning_rate": 1.0627332118057026e-05, "loss": 1.2405, "step": 7725 }, { "epoch": 0.7, "learning_rate": 1.0621467177172367e-05, "loss": 1.2898, "step": 7726 }, { "epoch": 0.7, "learning_rate": 1.0615603418540284e-05, "loss": 1.2456, "step": 7727 }, { "epoch": 0.7, "learning_rate": 1.0609740842642915e-05, "loss": 1.2217, "step": 7728 }, { "epoch": 0.7, "learning_rate": 1.0603879449962292e-05, "loss": 1.1833, "step": 7729 }, { "epoch": 0.7, "learning_rate": 1.0598019240980373e-05, "loss": 1.166, "step": 7730 }, { "epoch": 0.7, "learning_rate": 1.0592160216179e-05, "loss": 1.1917, "step": 7731 }, { "epoch": 0.7, "learning_rate": 1.0586302376039922e-05, "loss": 1.1945, "step": 7732 }, { "epoch": 0.7, "learning_rate": 1.0580445721044788e-05, "loss": 1.1491, "step": 7733 }, { "epoch": 0.7, "learning_rate": 1.0574590251675145e-05, "loss": 1.1736, "step": 7734 }, { "epoch": 0.7, "learning_rate": 1.0568735968412469e-05, "loss": 1.2281, "step": 7735 }, { "epoch": 0.7, "learning_rate": 1.056288287173811e-05, "loss": 1.2047, "step": 7736 }, { "epoch": 0.7, "learning_rate": 1.0557030962133327e-05, "loss": 1.2655, "step": 7737 }, { "epoch": 0.7, "learning_rate": 1.0551180240079288e-05, "loss": 1.2113, "step": 7738 }, { "epoch": 0.7, "learning_rate": 1.054533070605705e-05, "loss": 1.1593, "step": 7739 }, { "epoch": 0.7, "learning_rate": 1.0539482360547596e-05, "loss": 1.2801, "step": 7740 }, { "epoch": 0.7, "learning_rate": 1.0533635204031791e-05, "loss": 1.2002, "step": 7741 }, { "epoch": 0.7, "learning_rate": 1.05277892369904e-05, "loss": 1.2253, "step": 7742 }, { "epoch": 0.7, "learning_rate": 1.0521944459904127e-05, "loss": 1.2727, "step": 7743 }, { "epoch": 0.7, "learning_rate": 1.051610087325352e-05, "loss": 1.2043, "step": 7744 }, { "epoch": 0.7, "learning_rate": 1.0510258477519058e-05, "loss": 1.2033, "step": 7745 }, { "epoch": 0.7, "learning_rate": 1.0504417273181141e-05, "loss": 1.1759, "step": 7746 }, { "epoch": 0.7, "learning_rate": 1.049857726072005e-05, "loss": 1.2435, "step": 7747 }, { "epoch": 0.7, "learning_rate": 1.0492738440615962e-05, "loss": 1.1433, "step": 7748 }, { "epoch": 0.7, "learning_rate": 1.048690081334897e-05, "loss": 1.2237, "step": 7749 }, { "epoch": 0.7, "learning_rate": 1.0481064379399056e-05, "loss": 1.2529, "step": 7750 }, { "epoch": 0.7, "learning_rate": 1.0475229139246127e-05, "loss": 1.2553, "step": 7751 }, { "epoch": 0.7, "learning_rate": 1.0469395093369968e-05, "loss": 1.168, "step": 7752 }, { "epoch": 0.7, "learning_rate": 1.0463562242250275e-05, "loss": 1.1334, "step": 7753 }, { "epoch": 0.7, "learning_rate": 1.0457730586366645e-05, "loss": 1.2884, "step": 7754 }, { "epoch": 0.7, "learning_rate": 1.0451900126198569e-05, "loss": 1.2167, "step": 7755 }, { "epoch": 0.7, "learning_rate": 1.0446070862225463e-05, "loss": 1.2536, "step": 7756 }, { "epoch": 0.7, "learning_rate": 1.0440242794926621e-05, "loss": 1.1916, "step": 7757 }, { "epoch": 0.7, "learning_rate": 1.0434415924781248e-05, "loss": 1.1585, "step": 7758 }, { "epoch": 0.7, "learning_rate": 1.042859025226845e-05, "loss": 1.2218, "step": 7759 }, { "epoch": 0.7, "learning_rate": 1.0422765777867222e-05, "loss": 1.249, "step": 7760 }, { "epoch": 0.7, "learning_rate": 1.0416942502056492e-05, "loss": 1.2073, "step": 7761 }, { "epoch": 0.7, "learning_rate": 1.0411120425315062e-05, "loss": 1.2313, "step": 7762 }, { "epoch": 0.7, "learning_rate": 1.0405299548121633e-05, "loss": 1.2185, "step": 7763 }, { "epoch": 0.7, "learning_rate": 1.0399479870954845e-05, "loss": 1.175, "step": 7764 }, { "epoch": 0.7, "learning_rate": 1.0393661394293183e-05, "loss": 1.1974, "step": 7765 }, { "epoch": 0.7, "learning_rate": 1.0387844118615065e-05, "loss": 1.2244, "step": 7766 }, { "epoch": 0.7, "learning_rate": 1.0382028044398822e-05, "loss": 1.2171, "step": 7767 }, { "epoch": 0.7, "learning_rate": 1.0376213172122668e-05, "loss": 1.2072, "step": 7768 }, { "epoch": 0.7, "learning_rate": 1.0370399502264719e-05, "loss": 1.2146, "step": 7769 }, { "epoch": 0.7, "learning_rate": 1.0364587035302995e-05, "loss": 1.2028, "step": 7770 }, { "epoch": 0.7, "learning_rate": 1.0358775771715404e-05, "loss": 1.231, "step": 7771 }, { "epoch": 0.7, "learning_rate": 1.0352965711979795e-05, "loss": 1.2375, "step": 7772 }, { "epoch": 0.7, "learning_rate": 1.0347156856573873e-05, "loss": 1.1624, "step": 7773 }, { "epoch": 0.7, "learning_rate": 1.034134920597527e-05, "loss": 1.1875, "step": 7774 }, { "epoch": 0.7, "learning_rate": 1.0335542760661504e-05, "loss": 1.1497, "step": 7775 }, { "epoch": 0.7, "learning_rate": 1.0329737521109997e-05, "loss": 1.1677, "step": 7776 }, { "epoch": 0.7, "learning_rate": 1.0323933487798093e-05, "loss": 1.2236, "step": 7777 }, { "epoch": 0.7, "learning_rate": 1.031813066120301e-05, "loss": 1.2437, "step": 7778 }, { "epoch": 0.7, "learning_rate": 1.0312329041801878e-05, "loss": 1.2014, "step": 7779 }, { "epoch": 0.7, "learning_rate": 1.0306528630071718e-05, "loss": 1.2595, "step": 7780 }, { "epoch": 0.7, "learning_rate": 1.0300729426489463e-05, "loss": 1.2105, "step": 7781 }, { "epoch": 0.7, "learning_rate": 1.0294931431531953e-05, "loss": 1.1624, "step": 7782 }, { "epoch": 0.7, "learning_rate": 1.0289134645675915e-05, "loss": 1.1903, "step": 7783 }, { "epoch": 0.7, "learning_rate": 1.0283339069397979e-05, "loss": 1.2119, "step": 7784 }, { "epoch": 0.7, "learning_rate": 1.0277544703174677e-05, "loss": 1.1495, "step": 7785 }, { "epoch": 0.7, "learning_rate": 1.0271751547482439e-05, "loss": 1.2383, "step": 7786 }, { "epoch": 0.7, "learning_rate": 1.0265959602797594e-05, "loss": 1.2379, "step": 7787 }, { "epoch": 0.7, "learning_rate": 1.0260168869596393e-05, "loss": 1.2025, "step": 7788 }, { "epoch": 0.7, "learning_rate": 1.0254379348354956e-05, "loss": 1.2199, "step": 7789 }, { "epoch": 0.7, "learning_rate": 1.0248591039549324e-05, "loss": 1.204, "step": 7790 }, { "epoch": 0.7, "learning_rate": 1.024280394365543e-05, "loss": 1.202, "step": 7791 }, { "epoch": 0.7, "learning_rate": 1.0237018061149095e-05, "loss": 1.2067, "step": 7792 }, { "epoch": 0.7, "learning_rate": 1.0231233392506079e-05, "loss": 1.1685, "step": 7793 }, { "epoch": 0.7, "learning_rate": 1.0225449938202006e-05, "loss": 1.2136, "step": 7794 }, { "epoch": 0.7, "learning_rate": 1.0219667698712409e-05, "loss": 1.2673, "step": 7795 }, { "epoch": 0.7, "learning_rate": 1.0213886674512726e-05, "loss": 1.2217, "step": 7796 }, { "epoch": 0.7, "learning_rate": 1.0208106866078284e-05, "loss": 1.2011, "step": 7797 }, { "epoch": 0.7, "learning_rate": 1.0202328273884338e-05, "loss": 1.2401, "step": 7798 }, { "epoch": 0.7, "learning_rate": 1.0196550898406012e-05, "loss": 1.221, "step": 7799 }, { "epoch": 0.7, "learning_rate": 1.0190774740118343e-05, "loss": 1.2503, "step": 7800 }, { "epoch": 0.7, "learning_rate": 1.0184999799496268e-05, "loss": 1.2041, "step": 7801 }, { "epoch": 0.7, "learning_rate": 1.0179226077014609e-05, "loss": 1.2084, "step": 7802 }, { "epoch": 0.71, "learning_rate": 1.0173453573148123e-05, "loss": 1.2351, "step": 7803 }, { "epoch": 0.71, "learning_rate": 1.0167682288371435e-05, "loss": 1.2284, "step": 7804 }, { "epoch": 0.71, "learning_rate": 1.0161912223159081e-05, "loss": 1.2117, "step": 7805 }, { "epoch": 0.71, "learning_rate": 1.0156143377985497e-05, "loss": 1.231, "step": 7806 }, { "epoch": 0.71, "learning_rate": 1.0150375753325015e-05, "loss": 1.2354, "step": 7807 }, { "epoch": 0.71, "learning_rate": 1.0144609349651859e-05, "loss": 1.1916, "step": 7808 }, { "epoch": 0.71, "learning_rate": 1.0138844167440184e-05, "loss": 1.2179, "step": 7809 }, { "epoch": 0.71, "learning_rate": 1.0133080207164011e-05, "loss": 1.1693, "step": 7810 }, { "epoch": 0.71, "learning_rate": 1.0127317469297277e-05, "loss": 1.1781, "step": 7811 }, { "epoch": 0.71, "learning_rate": 1.0121555954313807e-05, "loss": 1.1711, "step": 7812 }, { "epoch": 0.71, "learning_rate": 1.011579566268733e-05, "loss": 1.2397, "step": 7813 }, { "epoch": 0.71, "learning_rate": 1.0110036594891491e-05, "loss": 1.2581, "step": 7814 }, { "epoch": 0.71, "learning_rate": 1.0104278751399818e-05, "loss": 1.2535, "step": 7815 }, { "epoch": 0.71, "learning_rate": 1.0098522132685731e-05, "loss": 1.2008, "step": 7816 }, { "epoch": 0.71, "learning_rate": 1.0092766739222568e-05, "loss": 1.227, "step": 7817 }, { "epoch": 0.71, "learning_rate": 1.0087012571483542e-05, "loss": 1.2114, "step": 7818 }, { "epoch": 0.71, "learning_rate": 1.0081259629941804e-05, "loss": 1.1819, "step": 7819 }, { "epoch": 0.71, "learning_rate": 1.0075507915070365e-05, "loss": 1.2255, "step": 7820 }, { "epoch": 0.71, "learning_rate": 1.0069757427342156e-05, "loss": 1.169, "step": 7821 }, { "epoch": 0.71, "learning_rate": 1.0064008167229998e-05, "loss": 1.218, "step": 7822 }, { "epoch": 0.71, "learning_rate": 1.0058260135206613e-05, "loss": 1.2697, "step": 7823 }, { "epoch": 0.71, "learning_rate": 1.0052513331744634e-05, "loss": 1.1981, "step": 7824 }, { "epoch": 0.71, "learning_rate": 1.0046767757316581e-05, "loss": 1.193, "step": 7825 }, { "epoch": 0.71, "learning_rate": 1.0041023412394867e-05, "loss": 1.1802, "step": 7826 }, { "epoch": 0.71, "learning_rate": 1.003528029745182e-05, "loss": 1.1786, "step": 7827 }, { "epoch": 0.71, "learning_rate": 1.0029538412959655e-05, "loss": 1.1878, "step": 7828 }, { "epoch": 0.71, "learning_rate": 1.0023797759390478e-05, "loss": 1.1946, "step": 7829 }, { "epoch": 0.71, "learning_rate": 1.0018058337216327e-05, "loss": 1.1721, "step": 7830 }, { "epoch": 0.71, "learning_rate": 1.0012320146909107e-05, "loss": 1.2155, "step": 7831 }, { "epoch": 0.71, "learning_rate": 1.0006583188940633e-05, "loss": 1.2452, "step": 7832 }, { "epoch": 0.71, "learning_rate": 1.0000847463782615e-05, "loss": 1.2228, "step": 7833 }, { "epoch": 0.71, "learning_rate": 9.995112971906656e-06, "loss": 1.2015, "step": 7834 }, { "epoch": 0.71, "learning_rate": 9.989379713784286e-06, "loss": 1.2228, "step": 7835 }, { "epoch": 0.71, "learning_rate": 9.983647689886901e-06, "loss": 1.2549, "step": 7836 }, { "epoch": 0.71, "learning_rate": 9.97791690068581e-06, "loss": 1.1952, "step": 7837 }, { "epoch": 0.71, "learning_rate": 9.972187346652218e-06, "loss": 1.2481, "step": 7838 }, { "epoch": 0.71, "learning_rate": 9.966459028257217e-06, "loss": 1.268, "step": 7839 }, { "epoch": 0.71, "learning_rate": 9.96073194597183e-06, "loss": 1.2345, "step": 7840 }, { "epoch": 0.71, "learning_rate": 9.955006100266948e-06, "loss": 1.1957, "step": 7841 }, { "epoch": 0.71, "learning_rate": 9.949281491613368e-06, "loss": 1.2425, "step": 7842 }, { "epoch": 0.71, "learning_rate": 9.943558120481791e-06, "loss": 1.2373, "step": 7843 }, { "epoch": 0.71, "learning_rate": 9.9378359873428e-06, "loss": 1.254, "step": 7844 }, { "epoch": 0.71, "learning_rate": 9.932115092666907e-06, "loss": 1.2087, "step": 7845 }, { "epoch": 0.71, "learning_rate": 9.926395436924495e-06, "loss": 1.1965, "step": 7846 }, { "epoch": 0.71, "learning_rate": 9.920677020585854e-06, "loss": 1.2168, "step": 7847 }, { "epoch": 0.71, "learning_rate": 9.914959844121172e-06, "loss": 1.215, "step": 7848 }, { "epoch": 0.71, "learning_rate": 9.909243908000535e-06, "loss": 1.2339, "step": 7849 }, { "epoch": 0.71, "learning_rate": 9.903529212693919e-06, "loss": 1.2257, "step": 7850 }, { "epoch": 0.71, "learning_rate": 9.897815758671219e-06, "loss": 1.2146, "step": 7851 }, { "epoch": 0.71, "learning_rate": 9.892103546402212e-06, "loss": 1.2478, "step": 7852 }, { "epoch": 0.71, "learning_rate": 9.886392576356573e-06, "loss": 1.1921, "step": 7853 }, { "epoch": 0.71, "learning_rate": 9.88068284900388e-06, "loss": 1.2115, "step": 7854 }, { "epoch": 0.71, "learning_rate": 9.874974364813594e-06, "loss": 1.2644, "step": 7855 }, { "epoch": 0.71, "learning_rate": 9.869267124255107e-06, "loss": 1.259, "step": 7856 }, { "epoch": 0.71, "learning_rate": 9.863561127797676e-06, "loss": 1.2329, "step": 7857 }, { "epoch": 0.71, "learning_rate": 9.857856375910471e-06, "loss": 1.1994, "step": 7858 }, { "epoch": 0.71, "learning_rate": 9.852152869062555e-06, "loss": 1.2277, "step": 7859 }, { "epoch": 0.71, "learning_rate": 9.846450607722881e-06, "loss": 1.2452, "step": 7860 }, { "epoch": 0.71, "learning_rate": 9.840749592360326e-06, "loss": 1.2272, "step": 7861 }, { "epoch": 0.71, "learning_rate": 9.835049823443648e-06, "loss": 1.2335, "step": 7862 }, { "epoch": 0.71, "learning_rate": 9.829351301441476e-06, "loss": 1.2318, "step": 7863 }, { "epoch": 0.71, "learning_rate": 9.823654026822386e-06, "loss": 1.2065, "step": 7864 }, { "epoch": 0.71, "learning_rate": 9.817958000054808e-06, "loss": 1.1815, "step": 7865 }, { "epoch": 0.71, "learning_rate": 9.812263221607112e-06, "loss": 1.1828, "step": 7866 }, { "epoch": 0.71, "learning_rate": 9.806569691947534e-06, "loss": 1.1537, "step": 7867 }, { "epoch": 0.71, "learning_rate": 9.800877411544212e-06, "loss": 1.2021, "step": 7868 }, { "epoch": 0.71, "learning_rate": 9.795186380865184e-06, "loss": 1.1788, "step": 7869 }, { "epoch": 0.71, "learning_rate": 9.789496600378392e-06, "loss": 1.2039, "step": 7870 }, { "epoch": 0.71, "learning_rate": 9.783808070551654e-06, "loss": 1.1597, "step": 7871 }, { "epoch": 0.71, "learning_rate": 9.778120791852722e-06, "loss": 1.2418, "step": 7872 }, { "epoch": 0.71, "learning_rate": 9.77243476474921e-06, "loss": 1.2112, "step": 7873 }, { "epoch": 0.71, "learning_rate": 9.76674998970865e-06, "loss": 1.2176, "step": 7874 }, { "epoch": 0.71, "learning_rate": 9.761066467198463e-06, "loss": 1.2174, "step": 7875 }, { "epoch": 0.71, "learning_rate": 9.755384197685951e-06, "loss": 1.2123, "step": 7876 }, { "epoch": 0.71, "learning_rate": 9.749703181638356e-06, "loss": 1.2579, "step": 7877 }, { "epoch": 0.71, "learning_rate": 9.744023419522777e-06, "loss": 1.2553, "step": 7878 }, { "epoch": 0.71, "learning_rate": 9.738344911806228e-06, "loss": 1.2437, "step": 7879 }, { "epoch": 0.71, "learning_rate": 9.732667658955611e-06, "loss": 1.2083, "step": 7880 }, { "epoch": 0.71, "learning_rate": 9.726991661437722e-06, "loss": 1.2099, "step": 7881 }, { "epoch": 0.71, "learning_rate": 9.721316919719279e-06, "loss": 1.199, "step": 7882 }, { "epoch": 0.71, "learning_rate": 9.71564343426688e-06, "loss": 1.1855, "step": 7883 }, { "epoch": 0.71, "learning_rate": 9.70997120554699e-06, "loss": 1.1981, "step": 7884 }, { "epoch": 0.71, "learning_rate": 9.704300234026028e-06, "loss": 1.2293, "step": 7885 }, { "epoch": 0.71, "learning_rate": 9.698630520170262e-06, "loss": 1.2578, "step": 7886 }, { "epoch": 0.71, "learning_rate": 9.69296206444589e-06, "loss": 1.2041, "step": 7887 }, { "epoch": 0.71, "learning_rate": 9.687294867319e-06, "loss": 1.1807, "step": 7888 }, { "epoch": 0.71, "learning_rate": 9.681628929255532e-06, "loss": 1.2791, "step": 7889 }, { "epoch": 0.71, "learning_rate": 9.675964250721392e-06, "loss": 1.1891, "step": 7890 }, { "epoch": 0.71, "learning_rate": 9.670300832182342e-06, "loss": 1.2731, "step": 7891 }, { "epoch": 0.71, "learning_rate": 9.664638674104032e-06, "loss": 1.1896, "step": 7892 }, { "epoch": 0.71, "learning_rate": 9.658977776952049e-06, "loss": 1.2621, "step": 7893 }, { "epoch": 0.71, "learning_rate": 9.653318141191836e-06, "loss": 1.1995, "step": 7894 }, { "epoch": 0.71, "learning_rate": 9.647659767288755e-06, "loss": 1.1816, "step": 7895 }, { "epoch": 0.71, "learning_rate": 9.642002655708051e-06, "loss": 1.2469, "step": 7896 }, { "epoch": 0.71, "learning_rate": 9.636346806914865e-06, "loss": 1.2188, "step": 7897 }, { "epoch": 0.71, "learning_rate": 9.630692221374257e-06, "loss": 1.2324, "step": 7898 }, { "epoch": 0.71, "learning_rate": 9.625038899551161e-06, "loss": 1.2258, "step": 7899 }, { "epoch": 0.71, "learning_rate": 9.619386841910409e-06, "loss": 1.2444, "step": 7900 }, { "epoch": 0.71, "learning_rate": 9.613736048916733e-06, "loss": 1.1961, "step": 7901 }, { "epoch": 0.71, "learning_rate": 9.608086521034754e-06, "loss": 1.2061, "step": 7902 }, { "epoch": 0.71, "learning_rate": 9.60243825872901e-06, "loss": 1.2666, "step": 7903 }, { "epoch": 0.71, "learning_rate": 9.596791262463924e-06, "loss": 1.2285, "step": 7904 }, { "epoch": 0.71, "learning_rate": 9.591145532703783e-06, "loss": 1.1517, "step": 7905 }, { "epoch": 0.71, "learning_rate": 9.585501069912825e-06, "loss": 1.2025, "step": 7906 }, { "epoch": 0.71, "learning_rate": 9.57985787455515e-06, "loss": 1.2108, "step": 7907 }, { "epoch": 0.71, "learning_rate": 9.57421594709475e-06, "loss": 1.175, "step": 7908 }, { "epoch": 0.71, "learning_rate": 9.568575287995554e-06, "loss": 1.2041, "step": 7909 }, { "epoch": 0.71, "learning_rate": 9.562935897721315e-06, "loss": 1.1871, "step": 7910 }, { "epoch": 0.71, "learning_rate": 9.557297776735754e-06, "loss": 1.194, "step": 7911 }, { "epoch": 0.71, "learning_rate": 9.551660925502448e-06, "loss": 1.1727, "step": 7912 }, { "epoch": 0.71, "learning_rate": 9.546025344484869e-06, "loss": 1.1768, "step": 7913 }, { "epoch": 0.72, "learning_rate": 9.540391034146408e-06, "loss": 1.2323, "step": 7914 }, { "epoch": 0.72, "learning_rate": 9.534757994950332e-06, "loss": 1.1983, "step": 7915 }, { "epoch": 0.72, "learning_rate": 9.52912622735981e-06, "loss": 1.2828, "step": 7916 }, { "epoch": 0.72, "learning_rate": 9.523495731837906e-06, "loss": 1.2303, "step": 7917 }, { "epoch": 0.72, "learning_rate": 9.517866508847565e-06, "loss": 1.1831, "step": 7918 }, { "epoch": 0.72, "learning_rate": 9.51223855885166e-06, "loss": 1.1882, "step": 7919 }, { "epoch": 0.72, "learning_rate": 9.506611882312938e-06, "loss": 1.2013, "step": 7920 }, { "epoch": 0.72, "learning_rate": 9.500986479694036e-06, "loss": 1.217, "step": 7921 }, { "epoch": 0.72, "learning_rate": 9.495362351457495e-06, "loss": 1.2284, "step": 7922 }, { "epoch": 0.72, "learning_rate": 9.489739498065747e-06, "loss": 1.2326, "step": 7923 }, { "epoch": 0.72, "learning_rate": 9.484117919981137e-06, "loss": 1.2213, "step": 7924 }, { "epoch": 0.72, "learning_rate": 9.478497617665887e-06, "loss": 1.2273, "step": 7925 }, { "epoch": 0.72, "learning_rate": 9.472878591582097e-06, "loss": 1.2142, "step": 7926 }, { "epoch": 0.72, "learning_rate": 9.467260842191806e-06, "loss": 1.1836, "step": 7927 }, { "epoch": 0.72, "learning_rate": 9.461644369956915e-06, "loss": 1.2132, "step": 7928 }, { "epoch": 0.72, "learning_rate": 9.45602917533922e-06, "loss": 1.258, "step": 7929 }, { "epoch": 0.72, "learning_rate": 9.450415258800454e-06, "loss": 1.2087, "step": 7930 }, { "epoch": 0.72, "learning_rate": 9.444802620802174e-06, "loss": 1.2203, "step": 7931 }, { "epoch": 0.72, "learning_rate": 9.439191261805893e-06, "loss": 1.2499, "step": 7932 }, { "epoch": 0.72, "learning_rate": 9.433581182272993e-06, "loss": 1.2101, "step": 7933 }, { "epoch": 0.72, "learning_rate": 9.427972382664746e-06, "loss": 1.2066, "step": 7934 }, { "epoch": 0.72, "learning_rate": 9.422364863442348e-06, "loss": 1.1938, "step": 7935 }, { "epoch": 0.72, "learning_rate": 9.416758625066835e-06, "loss": 1.2203, "step": 7936 }, { "epoch": 0.72, "learning_rate": 9.4111536679992e-06, "loss": 1.2058, "step": 7937 }, { "epoch": 0.72, "learning_rate": 9.40554999270029e-06, "loss": 1.1939, "step": 7938 }, { "epoch": 0.72, "learning_rate": 9.399947599630854e-06, "loss": 1.2736, "step": 7939 }, { "epoch": 0.72, "learning_rate": 9.394346489251555e-06, "loss": 1.2142, "step": 7940 }, { "epoch": 0.72, "learning_rate": 9.388746662022926e-06, "loss": 1.216, "step": 7941 }, { "epoch": 0.72, "learning_rate": 9.383148118405405e-06, "loss": 1.2858, "step": 7942 }, { "epoch": 0.72, "learning_rate": 9.377550858859324e-06, "loss": 1.2338, "step": 7943 }, { "epoch": 0.72, "learning_rate": 9.3719548838449e-06, "loss": 1.2136, "step": 7944 }, { "epoch": 0.72, "learning_rate": 9.366360193822268e-06, "loss": 1.2819, "step": 7945 }, { "epoch": 0.72, "learning_rate": 9.360766789251448e-06, "loss": 1.2381, "step": 7946 }, { "epoch": 0.72, "learning_rate": 9.35517467059232e-06, "loss": 1.2095, "step": 7947 }, { "epoch": 0.72, "learning_rate": 9.349583838304714e-06, "loss": 1.1964, "step": 7948 }, { "epoch": 0.72, "learning_rate": 9.34399429284832e-06, "loss": 1.2026, "step": 7949 }, { "epoch": 0.72, "learning_rate": 9.338406034682718e-06, "loss": 1.1902, "step": 7950 }, { "epoch": 0.72, "learning_rate": 9.33281906426742e-06, "loss": 1.1653, "step": 7951 }, { "epoch": 0.72, "learning_rate": 9.327233382061776e-06, "loss": 1.1839, "step": 7952 }, { "epoch": 0.72, "learning_rate": 9.32164898852508e-06, "loss": 1.2452, "step": 7953 }, { "epoch": 0.72, "learning_rate": 9.316065884116495e-06, "loss": 1.1881, "step": 7954 }, { "epoch": 0.72, "learning_rate": 9.310484069295076e-06, "loss": 1.2153, "step": 7955 }, { "epoch": 0.72, "learning_rate": 9.3049035445198e-06, "loss": 1.2699, "step": 7956 }, { "epoch": 0.72, "learning_rate": 9.299324310249485e-06, "loss": 1.2542, "step": 7957 }, { "epoch": 0.72, "learning_rate": 9.293746366942904e-06, "loss": 1.2012, "step": 7958 }, { "epoch": 0.72, "learning_rate": 9.28816971505868e-06, "loss": 1.1903, "step": 7959 }, { "epoch": 0.72, "learning_rate": 9.28259435505534e-06, "loss": 1.2325, "step": 7960 }, { "epoch": 0.72, "learning_rate": 9.277020287391332e-06, "loss": 1.2355, "step": 7961 }, { "epoch": 0.72, "learning_rate": 9.271447512524948e-06, "loss": 1.2092, "step": 7962 }, { "epoch": 0.72, "learning_rate": 9.265876030914417e-06, "loss": 1.2345, "step": 7963 }, { "epoch": 0.72, "learning_rate": 9.260305843017844e-06, "loss": 1.2463, "step": 7964 }, { "epoch": 0.72, "learning_rate": 9.254736949293215e-06, "loss": 1.2518, "step": 7965 }, { "epoch": 0.72, "learning_rate": 9.249169350198447e-06, "loss": 1.1359, "step": 7966 }, { "epoch": 0.72, "learning_rate": 9.243603046191321e-06, "loss": 1.2076, "step": 7967 }, { "epoch": 0.72, "learning_rate": 9.238038037729496e-06, "loss": 1.2163, "step": 7968 }, { "epoch": 0.72, "learning_rate": 9.232474325270571e-06, "loss": 1.1699, "step": 7969 }, { "epoch": 0.72, "learning_rate": 9.226911909272005e-06, "loss": 1.1907, "step": 7970 }, { "epoch": 0.72, "learning_rate": 9.221350790191147e-06, "loss": 1.2633, "step": 7971 }, { "epoch": 0.72, "learning_rate": 9.215790968485283e-06, "loss": 1.1704, "step": 7972 }, { "epoch": 0.72, "learning_rate": 9.210232444611524e-06, "loss": 1.1606, "step": 7973 }, { "epoch": 0.72, "learning_rate": 9.204675219026935e-06, "loss": 1.1696, "step": 7974 }, { "epoch": 0.72, "learning_rate": 9.199119292188443e-06, "loss": 1.243, "step": 7975 }, { "epoch": 0.72, "learning_rate": 9.19356466455287e-06, "loss": 1.2194, "step": 7976 }, { "epoch": 0.72, "learning_rate": 9.188011336576957e-06, "loss": 1.217, "step": 7977 }, { "epoch": 0.72, "learning_rate": 9.182459308717287e-06, "loss": 1.202, "step": 7978 }, { "epoch": 0.72, "learning_rate": 9.17690858143039e-06, "loss": 1.2316, "step": 7979 }, { "epoch": 0.72, "learning_rate": 9.17135915517266e-06, "loss": 1.2355, "step": 7980 }, { "epoch": 0.72, "learning_rate": 9.165811030400378e-06, "loss": 1.2254, "step": 7981 }, { "epoch": 0.72, "learning_rate": 9.160264207569763e-06, "loss": 1.2177, "step": 7982 }, { "epoch": 0.72, "learning_rate": 9.154718687136852e-06, "loss": 1.1749, "step": 7983 }, { "epoch": 0.72, "learning_rate": 9.149174469557642e-06, "loss": 1.2276, "step": 7984 }, { "epoch": 0.72, "learning_rate": 9.143631555287997e-06, "loss": 1.2086, "step": 7985 }, { "epoch": 0.72, "learning_rate": 9.13808994478366e-06, "loss": 1.1684, "step": 7986 }, { "epoch": 0.72, "learning_rate": 9.132549638500307e-06, "loss": 1.2357, "step": 7987 }, { "epoch": 0.72, "learning_rate": 9.127010636893456e-06, "loss": 1.2242, "step": 7988 }, { "epoch": 0.72, "learning_rate": 9.121472940418546e-06, "loss": 1.2323, "step": 7989 }, { "epoch": 0.72, "learning_rate": 9.115936549530917e-06, "loss": 1.2126, "step": 7990 }, { "epoch": 0.72, "learning_rate": 9.110401464685786e-06, "loss": 1.2242, "step": 7991 }, { "epoch": 0.72, "learning_rate": 9.104867686338259e-06, "loss": 1.2278, "step": 7992 }, { "epoch": 0.72, "learning_rate": 9.099335214943364e-06, "loss": 1.2548, "step": 7993 }, { "epoch": 0.72, "learning_rate": 9.093804050955967e-06, "loss": 1.2275, "step": 7994 }, { "epoch": 0.72, "learning_rate": 9.088274194830888e-06, "loss": 1.1844, "step": 7995 }, { "epoch": 0.72, "learning_rate": 9.082745647022797e-06, "loss": 1.1794, "step": 7996 }, { "epoch": 0.72, "learning_rate": 9.077218407986268e-06, "loss": 1.2039, "step": 7997 }, { "epoch": 0.72, "learning_rate": 9.071692478175789e-06, "loss": 1.1768, "step": 7998 }, { "epoch": 0.72, "learning_rate": 9.066167858045691e-06, "loss": 1.2133, "step": 7999 }, { "epoch": 0.72, "learning_rate": 9.06064454805025e-06, "loss": 1.1628, "step": 8000 }, { "epoch": 0.72, "learning_rate": 9.05512254864361e-06, "loss": 1.2657, "step": 8001 }, { "epoch": 0.72, "learning_rate": 9.04960186027979e-06, "loss": 1.2091, "step": 8002 }, { "epoch": 0.72, "learning_rate": 9.044082483412753e-06, "loss": 1.2268, "step": 8003 }, { "epoch": 0.72, "learning_rate": 9.038564418496284e-06, "loss": 1.155, "step": 8004 }, { "epoch": 0.72, "learning_rate": 9.033047665984125e-06, "loss": 1.2236, "step": 8005 }, { "epoch": 0.72, "learning_rate": 9.02753222632987e-06, "loss": 1.1743, "step": 8006 }, { "epoch": 0.72, "learning_rate": 9.022018099987015e-06, "loss": 1.2402, "step": 8007 }, { "epoch": 0.72, "learning_rate": 9.016505287408971e-06, "loss": 1.1974, "step": 8008 }, { "epoch": 0.72, "learning_rate": 9.010993789048997e-06, "loss": 1.1605, "step": 8009 }, { "epoch": 0.72, "learning_rate": 9.005483605360269e-06, "loss": 1.2146, "step": 8010 }, { "epoch": 0.72, "learning_rate": 8.999974736795865e-06, "loss": 1.1922, "step": 8011 }, { "epoch": 0.72, "learning_rate": 8.994467183808738e-06, "loss": 1.2217, "step": 8012 }, { "epoch": 0.72, "learning_rate": 8.98896094685174e-06, "loss": 1.2251, "step": 8013 }, { "epoch": 0.72, "learning_rate": 8.98345602637761e-06, "loss": 1.2551, "step": 8014 }, { "epoch": 0.72, "learning_rate": 8.977952422838972e-06, "loss": 1.1956, "step": 8015 }, { "epoch": 0.72, "learning_rate": 8.972450136688374e-06, "loss": 1.1838, "step": 8016 }, { "epoch": 0.72, "learning_rate": 8.96694916837822e-06, "loss": 1.1747, "step": 8017 }, { "epoch": 0.72, "learning_rate": 8.961449518360809e-06, "loss": 1.1795, "step": 8018 }, { "epoch": 0.72, "learning_rate": 8.955951187088368e-06, "loss": 1.1941, "step": 8019 }, { "epoch": 0.72, "learning_rate": 8.950454175012957e-06, "loss": 1.2138, "step": 8020 }, { "epoch": 0.72, "learning_rate": 8.944958482586583e-06, "loss": 1.1841, "step": 8021 }, { "epoch": 0.72, "learning_rate": 8.939464110261109e-06, "loss": 1.1845, "step": 8022 }, { "epoch": 0.72, "learning_rate": 8.9339710584883e-06, "loss": 1.2324, "step": 8023 }, { "epoch": 0.72, "learning_rate": 8.928479327719833e-06, "loss": 1.2094, "step": 8024 }, { "epoch": 0.73, "learning_rate": 8.922988918407224e-06, "loss": 1.3204, "step": 8025 }, { "epoch": 0.73, "learning_rate": 8.91749983100194e-06, "loss": 1.2325, "step": 8026 }, { "epoch": 0.73, "learning_rate": 8.912012065955308e-06, "loss": 1.2391, "step": 8027 }, { "epoch": 0.73, "learning_rate": 8.906525623718537e-06, "loss": 1.2485, "step": 8028 }, { "epoch": 0.73, "learning_rate": 8.901040504742769e-06, "loss": 1.2357, "step": 8029 }, { "epoch": 0.73, "learning_rate": 8.895556709478984e-06, "loss": 1.2179, "step": 8030 }, { "epoch": 0.73, "learning_rate": 8.890074238378074e-06, "loss": 1.2391, "step": 8031 }, { "epoch": 0.73, "learning_rate": 8.88459309189085e-06, "loss": 1.2194, "step": 8032 }, { "epoch": 0.73, "learning_rate": 8.879113270467981e-06, "loss": 1.2137, "step": 8033 }, { "epoch": 0.73, "learning_rate": 8.873634774560036e-06, "loss": 1.2295, "step": 8034 }, { "epoch": 0.73, "learning_rate": 8.868157604617474e-06, "loss": 1.2212, "step": 8035 }, { "epoch": 0.73, "learning_rate": 8.86268176109064e-06, "loss": 1.1833, "step": 8036 }, { "epoch": 0.73, "learning_rate": 8.857207244429793e-06, "loss": 1.2813, "step": 8037 }, { "epoch": 0.73, "learning_rate": 8.851734055085061e-06, "loss": 1.2094, "step": 8038 }, { "epoch": 0.73, "learning_rate": 8.846262193506465e-06, "loss": 1.23, "step": 8039 }, { "epoch": 0.73, "learning_rate": 8.840791660143922e-06, "loss": 1.2068, "step": 8040 }, { "epoch": 0.73, "learning_rate": 8.835322455447229e-06, "loss": 1.2371, "step": 8041 }, { "epoch": 0.73, "learning_rate": 8.8298545798661e-06, "loss": 1.2134, "step": 8042 }, { "epoch": 0.73, "learning_rate": 8.824388033850115e-06, "loss": 1.2065, "step": 8043 }, { "epoch": 0.73, "learning_rate": 8.818922817848744e-06, "loss": 1.2538, "step": 8044 }, { "epoch": 0.73, "learning_rate": 8.813458932311377e-06, "loss": 1.1634, "step": 8045 }, { "epoch": 0.73, "learning_rate": 8.807996377687248e-06, "loss": 1.184, "step": 8046 }, { "epoch": 0.73, "learning_rate": 8.802535154425523e-06, "loss": 1.2109, "step": 8047 }, { "epoch": 0.73, "learning_rate": 8.797075262975238e-06, "loss": 1.1734, "step": 8048 }, { "epoch": 0.73, "learning_rate": 8.791616703785318e-06, "loss": 1.2144, "step": 8049 }, { "epoch": 0.73, "learning_rate": 8.78615947730461e-06, "loss": 1.2495, "step": 8050 }, { "epoch": 0.73, "learning_rate": 8.780703583981797e-06, "loss": 1.1581, "step": 8051 }, { "epoch": 0.73, "learning_rate": 8.775249024265484e-06, "loss": 1.176, "step": 8052 }, { "epoch": 0.73, "learning_rate": 8.76979579860418e-06, "loss": 1.1952, "step": 8053 }, { "epoch": 0.73, "learning_rate": 8.76434390744626e-06, "loss": 1.2217, "step": 8054 }, { "epoch": 0.73, "learning_rate": 8.758893351239993e-06, "loss": 1.1956, "step": 8055 }, { "epoch": 0.73, "learning_rate": 8.75344413043355e-06, "loss": 1.1553, "step": 8056 }, { "epoch": 0.73, "learning_rate": 8.747996245474973e-06, "loss": 1.2001, "step": 8057 }, { "epoch": 0.73, "learning_rate": 8.742549696812219e-06, "loss": 1.2426, "step": 8058 }, { "epoch": 0.73, "learning_rate": 8.737104484893119e-06, "loss": 1.1986, "step": 8059 }, { "epoch": 0.73, "learning_rate": 8.731660610165397e-06, "loss": 1.2551, "step": 8060 }, { "epoch": 0.73, "learning_rate": 8.726218073076662e-06, "loss": 1.1831, "step": 8061 }, { "epoch": 0.73, "learning_rate": 8.720776874074412e-06, "loss": 1.192, "step": 8062 }, { "epoch": 0.73, "learning_rate": 8.715337013606062e-06, "loss": 1.1961, "step": 8063 }, { "epoch": 0.73, "learning_rate": 8.709898492118884e-06, "loss": 1.166, "step": 8064 }, { "epoch": 0.73, "learning_rate": 8.704461310060044e-06, "loss": 1.2054, "step": 8065 }, { "epoch": 0.73, "learning_rate": 8.699025467876632e-06, "loss": 1.2365, "step": 8066 }, { "epoch": 0.73, "learning_rate": 8.693590966015566e-06, "loss": 1.2476, "step": 8067 }, { "epoch": 0.73, "learning_rate": 8.688157804923716e-06, "loss": 1.2324, "step": 8068 }, { "epoch": 0.73, "learning_rate": 8.682725985047807e-06, "loss": 1.1779, "step": 8069 }, { "epoch": 0.73, "learning_rate": 8.677295506834455e-06, "loss": 1.217, "step": 8070 }, { "epoch": 0.73, "learning_rate": 8.671866370730195e-06, "loss": 1.2151, "step": 8071 }, { "epoch": 0.73, "learning_rate": 8.666438577181408e-06, "loss": 1.1903, "step": 8072 }, { "epoch": 0.73, "learning_rate": 8.66101212663438e-06, "loss": 1.1963, "step": 8073 }, { "epoch": 0.73, "learning_rate": 8.655587019535314e-06, "loss": 1.1986, "step": 8074 }, { "epoch": 0.73, "learning_rate": 8.65016325633027e-06, "loss": 1.1964, "step": 8075 }, { "epoch": 0.73, "learning_rate": 8.644740837465212e-06, "loss": 1.1269, "step": 8076 }, { "epoch": 0.73, "learning_rate": 8.639319763385986e-06, "loss": 1.2115, "step": 8077 }, { "epoch": 0.73, "learning_rate": 8.633900034538325e-06, "loss": 1.1899, "step": 8078 }, { "epoch": 0.73, "learning_rate": 8.628481651367876e-06, "loss": 1.2282, "step": 8079 }, { "epoch": 0.73, "learning_rate": 8.623064614320146e-06, "loss": 1.2328, "step": 8080 }, { "epoch": 0.73, "learning_rate": 8.617648923840543e-06, "loss": 1.2044, "step": 8081 }, { "epoch": 0.73, "learning_rate": 8.612234580374365e-06, "loss": 1.2599, "step": 8082 }, { "epoch": 0.73, "learning_rate": 8.60682158436679e-06, "loss": 1.2103, "step": 8083 }, { "epoch": 0.73, "learning_rate": 8.601409936262905e-06, "loss": 1.1851, "step": 8084 }, { "epoch": 0.73, "learning_rate": 8.595999636507673e-06, "loss": 1.1306, "step": 8085 }, { "epoch": 0.73, "learning_rate": 8.590590685545946e-06, "loss": 1.218, "step": 8086 }, { "epoch": 0.73, "learning_rate": 8.585183083822462e-06, "loss": 1.1807, "step": 8087 }, { "epoch": 0.73, "learning_rate": 8.57977683178185e-06, "loss": 1.1923, "step": 8088 }, { "epoch": 0.73, "learning_rate": 8.574371929868644e-06, "loss": 1.2531, "step": 8089 }, { "epoch": 0.73, "learning_rate": 8.568968378527245e-06, "loss": 1.205, "step": 8090 }, { "epoch": 0.73, "learning_rate": 8.563566178201945e-06, "loss": 1.1511, "step": 8091 }, { "epoch": 0.73, "learning_rate": 8.558165329336957e-06, "loss": 1.1667, "step": 8092 }, { "epoch": 0.73, "learning_rate": 8.552765832376331e-06, "loss": 1.2145, "step": 8093 }, { "epoch": 0.73, "learning_rate": 8.547367687764033e-06, "loss": 1.2293, "step": 8094 }, { "epoch": 0.73, "learning_rate": 8.541970895943932e-06, "loss": 1.2314, "step": 8095 }, { "epoch": 0.73, "learning_rate": 8.536575457359764e-06, "loss": 1.2149, "step": 8096 }, { "epoch": 0.73, "learning_rate": 8.531181372455161e-06, "loss": 1.2427, "step": 8097 }, { "epoch": 0.73, "learning_rate": 8.525788641673644e-06, "loss": 1.1834, "step": 8098 }, { "epoch": 0.73, "learning_rate": 8.52039726545861e-06, "loss": 1.2455, "step": 8099 }, { "epoch": 0.73, "learning_rate": 8.515007244253378e-06, "loss": 1.2963, "step": 8100 }, { "epoch": 0.73, "learning_rate": 8.509618578501123e-06, "loss": 1.1562, "step": 8101 }, { "epoch": 0.73, "learning_rate": 8.50423126864492e-06, "loss": 1.192, "step": 8102 }, { "epoch": 0.73, "learning_rate": 8.49884531512773e-06, "loss": 1.1779, "step": 8103 }, { "epoch": 0.73, "learning_rate": 8.493460718392398e-06, "loss": 1.2013, "step": 8104 }, { "epoch": 0.73, "learning_rate": 8.488077478881684e-06, "loss": 1.2132, "step": 8105 }, { "epoch": 0.73, "learning_rate": 8.482695597038207e-06, "loss": 1.2317, "step": 8106 }, { "epoch": 0.73, "learning_rate": 8.477315073304481e-06, "loss": 1.1888, "step": 8107 }, { "epoch": 0.73, "learning_rate": 8.471935908122913e-06, "loss": 1.2636, "step": 8108 }, { "epoch": 0.73, "learning_rate": 8.466558101935787e-06, "loss": 1.2397, "step": 8109 }, { "epoch": 0.73, "learning_rate": 8.461181655185302e-06, "loss": 1.1946, "step": 8110 }, { "epoch": 0.73, "learning_rate": 8.455806568313521e-06, "loss": 1.2183, "step": 8111 }, { "epoch": 0.73, "learning_rate": 8.4504328417624e-06, "loss": 1.1858, "step": 8112 }, { "epoch": 0.73, "learning_rate": 8.445060475973787e-06, "loss": 1.2264, "step": 8113 }, { "epoch": 0.73, "learning_rate": 8.439689471389417e-06, "loss": 1.2652, "step": 8114 }, { "epoch": 0.73, "learning_rate": 8.434319828450905e-06, "loss": 1.1972, "step": 8115 }, { "epoch": 0.73, "learning_rate": 8.428951547599773e-06, "loss": 1.1464, "step": 8116 }, { "epoch": 0.73, "learning_rate": 8.423584629277415e-06, "loss": 1.1837, "step": 8117 }, { "epoch": 0.73, "learning_rate": 8.418219073925118e-06, "loss": 1.2474, "step": 8118 }, { "epoch": 0.73, "learning_rate": 8.412854881984055e-06, "loss": 1.1804, "step": 8119 }, { "epoch": 0.73, "learning_rate": 8.407492053895284e-06, "loss": 1.1864, "step": 8120 }, { "epoch": 0.73, "learning_rate": 8.402130590099766e-06, "loss": 1.241, "step": 8121 }, { "epoch": 0.73, "learning_rate": 8.396770491038334e-06, "loss": 1.1989, "step": 8122 }, { "epoch": 0.73, "learning_rate": 8.39141175715171e-06, "loss": 1.2206, "step": 8123 }, { "epoch": 0.73, "learning_rate": 8.386054388880512e-06, "loss": 1.1242, "step": 8124 }, { "epoch": 0.73, "learning_rate": 8.380698386665233e-06, "loss": 1.2724, "step": 8125 }, { "epoch": 0.73, "learning_rate": 8.375343750946279e-06, "loss": 1.1896, "step": 8126 }, { "epoch": 0.73, "learning_rate": 8.369990482163914e-06, "loss": 1.2262, "step": 8127 }, { "epoch": 0.73, "learning_rate": 8.364638580758308e-06, "loss": 1.2774, "step": 8128 }, { "epoch": 0.73, "learning_rate": 8.359288047169505e-06, "loss": 1.1748, "step": 8129 }, { "epoch": 0.73, "learning_rate": 8.353938881837445e-06, "loss": 1.2214, "step": 8130 }, { "epoch": 0.73, "learning_rate": 8.348591085201965e-06, "loss": 1.2001, "step": 8131 }, { "epoch": 0.73, "learning_rate": 8.343244657702773e-06, "loss": 1.201, "step": 8132 }, { "epoch": 0.73, "learning_rate": 8.337899599779472e-06, "loss": 1.2515, "step": 8133 }, { "epoch": 0.73, "learning_rate": 8.332555911871548e-06, "loss": 1.2173, "step": 8134 }, { "epoch": 0.74, "learning_rate": 8.327213594418382e-06, "loss": 1.1972, "step": 8135 }, { "epoch": 0.74, "learning_rate": 8.321872647859225e-06, "loss": 1.1952, "step": 8136 }, { "epoch": 0.74, "learning_rate": 8.316533072633247e-06, "loss": 1.1815, "step": 8137 }, { "epoch": 0.74, "learning_rate": 8.311194869179479e-06, "loss": 1.1694, "step": 8138 }, { "epoch": 0.74, "learning_rate": 8.305858037936842e-06, "loss": 1.154, "step": 8139 }, { "epoch": 0.74, "learning_rate": 8.300522579344152e-06, "loss": 1.2783, "step": 8140 }, { "epoch": 0.74, "learning_rate": 8.295188493840104e-06, "loss": 1.2006, "step": 8141 }, { "epoch": 0.74, "learning_rate": 8.289855781863293e-06, "loss": 1.2536, "step": 8142 }, { "epoch": 0.74, "learning_rate": 8.284524443852192e-06, "loss": 1.1643, "step": 8143 }, { "epoch": 0.74, "learning_rate": 8.279194480245159e-06, "loss": 1.2374, "step": 8144 }, { "epoch": 0.74, "learning_rate": 8.273865891480442e-06, "loss": 1.2758, "step": 8145 }, { "epoch": 0.74, "learning_rate": 8.268538677996168e-06, "loss": 1.1944, "step": 8146 }, { "epoch": 0.74, "learning_rate": 8.263212840230378e-06, "loss": 1.2244, "step": 8147 }, { "epoch": 0.74, "learning_rate": 8.257888378620968e-06, "loss": 1.2319, "step": 8148 }, { "epoch": 0.74, "learning_rate": 8.252565293605735e-06, "loss": 1.1913, "step": 8149 }, { "epoch": 0.74, "learning_rate": 8.247243585622361e-06, "loss": 1.2376, "step": 8150 }, { "epoch": 0.74, "learning_rate": 8.24192325510841e-06, "loss": 1.2173, "step": 8151 }, { "epoch": 0.74, "learning_rate": 8.236604302501352e-06, "loss": 1.1865, "step": 8152 }, { "epoch": 0.74, "learning_rate": 8.231286728238522e-06, "loss": 1.1874, "step": 8153 }, { "epoch": 0.74, "learning_rate": 8.225970532757149e-06, "loss": 1.184, "step": 8154 }, { "epoch": 0.74, "learning_rate": 8.220655716494346e-06, "loss": 1.1922, "step": 8155 }, { "epoch": 0.74, "learning_rate": 8.215342279887122e-06, "loss": 1.2296, "step": 8156 }, { "epoch": 0.74, "learning_rate": 8.210030223372353e-06, "loss": 1.1699, "step": 8157 }, { "epoch": 0.74, "learning_rate": 8.204719547386836e-06, "loss": 1.1898, "step": 8158 }, { "epoch": 0.74, "learning_rate": 8.199410252367214e-06, "loss": 1.2045, "step": 8159 }, { "epoch": 0.74, "learning_rate": 8.194102338750048e-06, "loss": 1.2607, "step": 8160 }, { "epoch": 0.74, "learning_rate": 8.188795806971768e-06, "loss": 1.2034, "step": 8161 }, { "epoch": 0.74, "learning_rate": 8.183490657468688e-06, "loss": 1.2423, "step": 8162 }, { "epoch": 0.74, "learning_rate": 8.178186890677029e-06, "loss": 1.1633, "step": 8163 }, { "epoch": 0.74, "learning_rate": 8.172884507032878e-06, "loss": 1.2688, "step": 8164 }, { "epoch": 0.74, "learning_rate": 8.167583506972216e-06, "loss": 1.2199, "step": 8165 }, { "epoch": 0.74, "learning_rate": 8.16228389093091e-06, "loss": 1.2535, "step": 8166 }, { "epoch": 0.74, "learning_rate": 8.156985659344707e-06, "loss": 1.2112, "step": 8167 }, { "epoch": 0.74, "learning_rate": 8.151688812649255e-06, "loss": 1.2042, "step": 8168 }, { "epoch": 0.74, "learning_rate": 8.146393351280079e-06, "loss": 1.2426, "step": 8169 }, { "epoch": 0.74, "learning_rate": 8.141099275672585e-06, "loss": 1.2372, "step": 8170 }, { "epoch": 0.74, "learning_rate": 8.135806586262071e-06, "loss": 1.1955, "step": 8171 }, { "epoch": 0.74, "learning_rate": 8.130515283483713e-06, "loss": 1.1415, "step": 8172 }, { "epoch": 0.74, "learning_rate": 8.125225367772596e-06, "loss": 1.2234, "step": 8173 }, { "epoch": 0.74, "learning_rate": 8.119936839563669e-06, "loss": 1.1782, "step": 8174 }, { "epoch": 0.74, "learning_rate": 8.114649699291767e-06, "loss": 1.159, "step": 8175 }, { "epoch": 0.74, "learning_rate": 8.109363947391624e-06, "loss": 1.1891, "step": 8176 }, { "epoch": 0.74, "learning_rate": 8.10407958429785e-06, "loss": 1.1536, "step": 8177 }, { "epoch": 0.74, "learning_rate": 8.098796610444934e-06, "loss": 1.1826, "step": 8178 }, { "epoch": 0.74, "learning_rate": 8.093515026267278e-06, "loss": 1.2461, "step": 8179 }, { "epoch": 0.74, "learning_rate": 8.088234832199143e-06, "loss": 1.1832, "step": 8180 }, { "epoch": 0.74, "learning_rate": 8.082956028674688e-06, "loss": 1.2208, "step": 8181 }, { "epoch": 0.74, "learning_rate": 8.077678616127948e-06, "loss": 1.2048, "step": 8182 }, { "epoch": 0.74, "learning_rate": 8.072402594992847e-06, "loss": 1.2388, "step": 8183 }, { "epoch": 0.74, "learning_rate": 8.067127965703213e-06, "loss": 1.2133, "step": 8184 }, { "epoch": 0.74, "learning_rate": 8.061854728692738e-06, "loss": 1.188, "step": 8185 }, { "epoch": 0.74, "learning_rate": 8.056582884395004e-06, "loss": 1.2526, "step": 8186 }, { "epoch": 0.74, "learning_rate": 8.051312433243476e-06, "loss": 1.2312, "step": 8187 }, { "epoch": 0.74, "learning_rate": 8.046043375671506e-06, "loss": 1.1711, "step": 8188 }, { "epoch": 0.74, "learning_rate": 8.04077571211235e-06, "loss": 1.1726, "step": 8189 }, { "epoch": 0.74, "learning_rate": 8.035509442999123e-06, "loss": 1.1615, "step": 8190 }, { "epoch": 0.74, "learning_rate": 8.030244568764837e-06, "loss": 1.2167, "step": 8191 }, { "epoch": 0.74, "learning_rate": 8.024981089842388e-06, "loss": 1.2425, "step": 8192 }, { "epoch": 0.74, "learning_rate": 8.019719006664548e-06, "loss": 1.1807, "step": 8193 }, { "epoch": 0.74, "learning_rate": 8.014458319664003e-06, "loss": 1.2269, "step": 8194 }, { "epoch": 0.74, "learning_rate": 8.009199029273293e-06, "loss": 1.2106, "step": 8195 }, { "epoch": 0.74, "learning_rate": 8.003941135924858e-06, "loss": 1.2652, "step": 8196 }, { "epoch": 0.74, "learning_rate": 7.998684640051019e-06, "loss": 1.226, "step": 8197 }, { "epoch": 0.74, "learning_rate": 7.993429542083983e-06, "loss": 1.2478, "step": 8198 }, { "epoch": 0.74, "learning_rate": 7.988175842455836e-06, "loss": 1.2598, "step": 8199 }, { "epoch": 0.74, "learning_rate": 7.982923541598566e-06, "loss": 1.1576, "step": 8200 }, { "epoch": 0.74, "learning_rate": 7.977672639944034e-06, "loss": 1.2695, "step": 8201 }, { "epoch": 0.74, "learning_rate": 7.972423137923984e-06, "loss": 1.216, "step": 8202 }, { "epoch": 0.74, "learning_rate": 7.967175035970047e-06, "loss": 1.1755, "step": 8203 }, { "epoch": 0.74, "learning_rate": 7.961928334513739e-06, "loss": 1.1706, "step": 8204 }, { "epoch": 0.74, "learning_rate": 7.956683033986467e-06, "loss": 1.1962, "step": 8205 }, { "epoch": 0.74, "learning_rate": 7.951439134819517e-06, "loss": 1.2101, "step": 8206 }, { "epoch": 0.74, "learning_rate": 7.94619663744406e-06, "loss": 1.2025, "step": 8207 }, { "epoch": 0.74, "learning_rate": 7.940955542291148e-06, "loss": 1.1751, "step": 8208 }, { "epoch": 0.74, "learning_rate": 7.935715849791717e-06, "loss": 1.1521, "step": 8209 }, { "epoch": 0.74, "learning_rate": 7.930477560376608e-06, "loss": 1.1646, "step": 8210 }, { "epoch": 0.74, "learning_rate": 7.925240674476525e-06, "loss": 1.1922, "step": 8211 }, { "epoch": 0.74, "learning_rate": 7.920005192522062e-06, "loss": 1.2584, "step": 8212 }, { "epoch": 0.74, "learning_rate": 7.914771114943697e-06, "loss": 1.3046, "step": 8213 }, { "epoch": 0.74, "learning_rate": 7.909538442171788e-06, "loss": 1.1987, "step": 8214 }, { "epoch": 0.74, "learning_rate": 7.904307174636597e-06, "loss": 1.2094, "step": 8215 }, { "epoch": 0.74, "learning_rate": 7.89907731276825e-06, "loss": 1.169, "step": 8216 }, { "epoch": 0.74, "learning_rate": 7.893848856996765e-06, "loss": 1.1764, "step": 8217 }, { "epoch": 0.74, "learning_rate": 7.888621807752042e-06, "loss": 1.2881, "step": 8218 }, { "epoch": 0.74, "learning_rate": 7.883396165463868e-06, "loss": 1.2113, "step": 8219 }, { "epoch": 0.74, "learning_rate": 7.878171930561906e-06, "loss": 1.1989, "step": 8220 }, { "epoch": 0.74, "learning_rate": 7.872949103475726e-06, "loss": 1.2184, "step": 8221 }, { "epoch": 0.74, "learning_rate": 7.867727684634759e-06, "loss": 1.2631, "step": 8222 }, { "epoch": 0.74, "learning_rate": 7.862507674468328e-06, "loss": 1.2091, "step": 8223 }, { "epoch": 0.74, "learning_rate": 7.857289073405639e-06, "loss": 1.1957, "step": 8224 }, { "epoch": 0.74, "learning_rate": 7.852071881875778e-06, "loss": 1.2321, "step": 8225 }, { "epoch": 0.74, "learning_rate": 7.846856100307734e-06, "loss": 1.1783, "step": 8226 }, { "epoch": 0.74, "learning_rate": 7.841641729130359e-06, "loss": 1.2021, "step": 8227 }, { "epoch": 0.74, "learning_rate": 7.836428768772397e-06, "loss": 1.2028, "step": 8228 }, { "epoch": 0.74, "learning_rate": 7.83121721966248e-06, "loss": 1.2277, "step": 8229 }, { "epoch": 0.74, "learning_rate": 7.8260070822291e-06, "loss": 1.1731, "step": 8230 }, { "epoch": 0.74, "learning_rate": 7.820798356900679e-06, "loss": 1.2156, "step": 8231 }, { "epoch": 0.74, "learning_rate": 7.815591044105489e-06, "loss": 1.2523, "step": 8232 }, { "epoch": 0.74, "learning_rate": 7.810385144271684e-06, "loss": 1.2524, "step": 8233 }, { "epoch": 0.74, "learning_rate": 7.805180657827318e-06, "loss": 1.1518, "step": 8234 }, { "epoch": 0.74, "learning_rate": 7.799977585200313e-06, "loss": 1.1948, "step": 8235 }, { "epoch": 0.74, "learning_rate": 7.794775926818498e-06, "loss": 1.2871, "step": 8236 }, { "epoch": 0.74, "learning_rate": 7.789575683109574e-06, "loss": 1.2309, "step": 8237 }, { "epoch": 0.74, "learning_rate": 7.784376854501097e-06, "loss": 1.2278, "step": 8238 }, { "epoch": 0.74, "learning_rate": 7.779179441420554e-06, "loss": 1.2414, "step": 8239 }, { "epoch": 0.74, "learning_rate": 7.77398344429529e-06, "loss": 1.1614, "step": 8240 }, { "epoch": 0.74, "learning_rate": 7.768788863552529e-06, "loss": 1.1401, "step": 8241 }, { "epoch": 0.74, "learning_rate": 7.763595699619403e-06, "loss": 1.237, "step": 8242 }, { "epoch": 0.74, "learning_rate": 7.758403952922904e-06, "loss": 1.1939, "step": 8243 }, { "epoch": 0.74, "learning_rate": 7.753213623889916e-06, "loss": 1.2544, "step": 8244 }, { "epoch": 0.74, "learning_rate": 7.748024712947205e-06, "loss": 1.237, "step": 8245 }, { "epoch": 0.75, "learning_rate": 7.742837220521412e-06, "loss": 1.1994, "step": 8246 }, { "epoch": 0.75, "learning_rate": 7.737651147039088e-06, "loss": 1.1939, "step": 8247 }, { "epoch": 0.75, "learning_rate": 7.732466492926644e-06, "loss": 1.1968, "step": 8248 }, { "epoch": 0.75, "learning_rate": 7.727283258610373e-06, "loss": 1.1968, "step": 8249 }, { "epoch": 0.75, "learning_rate": 7.722101444516466e-06, "loss": 1.2104, "step": 8250 }, { "epoch": 0.75, "learning_rate": 7.71692105107098e-06, "loss": 1.2101, "step": 8251 }, { "epoch": 0.75, "learning_rate": 7.711742078699877e-06, "loss": 1.2136, "step": 8252 }, { "epoch": 0.75, "learning_rate": 7.706564527828994e-06, "loss": 1.2387, "step": 8253 }, { "epoch": 0.75, "learning_rate": 7.701388398884021e-06, "loss": 1.1745, "step": 8254 }, { "epoch": 0.75, "learning_rate": 7.696213692290583e-06, "loss": 1.2254, "step": 8255 }, { "epoch": 0.75, "learning_rate": 7.691040408474143e-06, "loss": 1.2039, "step": 8256 }, { "epoch": 0.75, "learning_rate": 7.685868547860086e-06, "loss": 1.2494, "step": 8257 }, { "epoch": 0.75, "learning_rate": 7.680698110873657e-06, "loss": 1.1989, "step": 8258 }, { "epoch": 0.75, "learning_rate": 7.675529097939967e-06, "loss": 1.2566, "step": 8259 }, { "epoch": 0.75, "learning_rate": 7.67036150948405e-06, "loss": 1.2065, "step": 8260 }, { "epoch": 0.75, "learning_rate": 7.665195345930797e-06, "loss": 1.2474, "step": 8261 }, { "epoch": 0.75, "learning_rate": 7.660030607704979e-06, "loss": 1.175, "step": 8262 }, { "epoch": 0.75, "learning_rate": 7.654867295231286e-06, "loss": 1.2224, "step": 8263 }, { "epoch": 0.75, "learning_rate": 7.649705408934226e-06, "loss": 1.1345, "step": 8264 }, { "epoch": 0.75, "learning_rate": 7.644544949238253e-06, "loss": 1.2515, "step": 8265 }, { "epoch": 0.75, "learning_rate": 7.639385916567673e-06, "loss": 1.1807, "step": 8266 }, { "epoch": 0.75, "learning_rate": 7.634228311346667e-06, "loss": 1.2478, "step": 8267 }, { "epoch": 0.75, "learning_rate": 7.6290721339993294e-06, "loss": 1.1828, "step": 8268 }, { "epoch": 0.75, "learning_rate": 7.623917384949608e-06, "loss": 1.2573, "step": 8269 }, { "epoch": 0.75, "learning_rate": 7.61876406462135e-06, "loss": 1.2057, "step": 8270 }, { "epoch": 0.75, "learning_rate": 7.613612173438273e-06, "loss": 1.2346, "step": 8271 }, { "epoch": 0.75, "learning_rate": 7.608461711823981e-06, "loss": 1.1918, "step": 8272 }, { "epoch": 0.75, "learning_rate": 7.603312680201971e-06, "loss": 1.2106, "step": 8273 }, { "epoch": 0.75, "learning_rate": 7.598165078995623e-06, "loss": 1.2012, "step": 8274 }, { "epoch": 0.75, "learning_rate": 7.593018908628161e-06, "loss": 1.255, "step": 8275 }, { "epoch": 0.75, "learning_rate": 7.587874169522746e-06, "loss": 1.1704, "step": 8276 }, { "epoch": 0.75, "learning_rate": 7.58273086210238e-06, "loss": 1.3022, "step": 8277 }, { "epoch": 0.75, "learning_rate": 7.57758898678998e-06, "loss": 1.2057, "step": 8278 }, { "epoch": 0.75, "learning_rate": 7.5724485440083285e-06, "loss": 1.2165, "step": 8279 }, { "epoch": 0.75, "learning_rate": 7.567309534180067e-06, "loss": 1.2657, "step": 8280 }, { "epoch": 0.75, "learning_rate": 7.562171957727768e-06, "loss": 1.1608, "step": 8281 }, { "epoch": 0.75, "learning_rate": 7.557035815073848e-06, "loss": 1.1613, "step": 8282 }, { "epoch": 0.75, "learning_rate": 7.551901106640613e-06, "loss": 1.186, "step": 8283 }, { "epoch": 0.75, "learning_rate": 7.546767832850282e-06, "loss": 1.1339, "step": 8284 }, { "epoch": 0.75, "learning_rate": 7.541635994124899e-06, "loss": 1.2409, "step": 8285 }, { "epoch": 0.75, "learning_rate": 7.536505590886442e-06, "loss": 1.2385, "step": 8286 }, { "epoch": 0.75, "learning_rate": 7.531376623556744e-06, "loss": 1.214, "step": 8287 }, { "epoch": 0.75, "learning_rate": 7.5262490925575194e-06, "loss": 1.1993, "step": 8288 }, { "epoch": 0.75, "learning_rate": 7.521122998310392e-06, "loss": 1.1937, "step": 8289 }, { "epoch": 0.75, "learning_rate": 7.5159983412368214e-06, "loss": 1.1692, "step": 8290 }, { "epoch": 0.75, "learning_rate": 7.5108751217581904e-06, "loss": 1.1028, "step": 8291 }, { "epoch": 0.75, "learning_rate": 7.505753340295748e-06, "loss": 1.2533, "step": 8292 }, { "epoch": 0.75, "learning_rate": 7.50063299727061e-06, "loss": 1.1407, "step": 8293 }, { "epoch": 0.75, "learning_rate": 7.495514093103809e-06, "loss": 1.2154, "step": 8294 }, { "epoch": 0.75, "learning_rate": 7.490396628216237e-06, "loss": 1.1912, "step": 8295 }, { "epoch": 0.75, "learning_rate": 7.485280603028649e-06, "loss": 1.2139, "step": 8296 }, { "epoch": 0.75, "learning_rate": 7.48016601796172e-06, "loss": 1.2102, "step": 8297 }, { "epoch": 0.75, "learning_rate": 7.475052873435978e-06, "loss": 1.2396, "step": 8298 }, { "epoch": 0.75, "learning_rate": 7.469941169871855e-06, "loss": 1.2598, "step": 8299 }, { "epoch": 0.75, "learning_rate": 7.464830907689657e-06, "loss": 1.2353, "step": 8300 }, { "epoch": 0.75, "learning_rate": 7.459722087309542e-06, "loss": 1.2189, "step": 8301 }, { "epoch": 0.75, "learning_rate": 7.4546147091516e-06, "loss": 1.257, "step": 8302 }, { "epoch": 0.75, "learning_rate": 7.449508773635766e-06, "loss": 1.1911, "step": 8303 }, { "epoch": 0.75, "learning_rate": 7.444404281181863e-06, "loss": 1.2132, "step": 8304 }, { "epoch": 0.75, "learning_rate": 7.439301232209622e-06, "loss": 1.2648, "step": 8305 }, { "epoch": 0.75, "learning_rate": 7.434199627138602e-06, "loss": 1.171, "step": 8306 }, { "epoch": 0.75, "learning_rate": 7.429099466388298e-06, "loss": 1.2138, "step": 8307 }, { "epoch": 0.75, "learning_rate": 7.424000750378057e-06, "loss": 1.2257, "step": 8308 }, { "epoch": 0.75, "learning_rate": 7.418903479527101e-06, "loss": 1.2785, "step": 8309 }, { "epoch": 0.75, "learning_rate": 7.413807654254573e-06, "loss": 1.2371, "step": 8310 }, { "epoch": 0.75, "learning_rate": 7.408713274979434e-06, "loss": 1.1987, "step": 8311 }, { "epoch": 0.75, "learning_rate": 7.40362034212059e-06, "loss": 1.1971, "step": 8312 }, { "epoch": 0.75, "learning_rate": 7.398528856096787e-06, "loss": 1.2045, "step": 8313 }, { "epoch": 0.75, "learning_rate": 7.393438817326659e-06, "loss": 1.1881, "step": 8314 }, { "epoch": 0.75, "learning_rate": 7.388350226228749e-06, "loss": 1.247, "step": 8315 }, { "epoch": 0.75, "learning_rate": 7.383263083221434e-06, "loss": 1.1845, "step": 8316 }, { "epoch": 0.75, "learning_rate": 7.378177388723001e-06, "loss": 1.1869, "step": 8317 }, { "epoch": 0.75, "learning_rate": 7.373093143151624e-06, "loss": 1.1734, "step": 8318 }, { "epoch": 0.75, "learning_rate": 7.368010346925333e-06, "loss": 1.2226, "step": 8319 }, { "epoch": 0.75, "learning_rate": 7.362929000462071e-06, "loss": 1.218, "step": 8320 }, { "epoch": 0.75, "learning_rate": 7.3578491041796425e-06, "loss": 1.2439, "step": 8321 }, { "epoch": 0.75, "learning_rate": 7.3527706584957085e-06, "loss": 1.2404, "step": 8322 }, { "epoch": 0.75, "learning_rate": 7.3476936638278624e-06, "loss": 1.1813, "step": 8323 }, { "epoch": 0.75, "learning_rate": 7.342618120593544e-06, "loss": 1.2108, "step": 8324 }, { "epoch": 0.75, "learning_rate": 7.337544029210073e-06, "loss": 1.2001, "step": 8325 }, { "epoch": 0.75, "learning_rate": 7.332471390094683e-06, "loss": 1.2316, "step": 8326 }, { "epoch": 0.75, "learning_rate": 7.327400203664431e-06, "loss": 1.2468, "step": 8327 }, { "epoch": 0.75, "learning_rate": 7.3223304703363135e-06, "loss": 1.1515, "step": 8328 }, { "epoch": 0.75, "learning_rate": 7.317262190527172e-06, "loss": 1.1634, "step": 8329 }, { "epoch": 0.75, "learning_rate": 7.312195364653731e-06, "loss": 1.187, "step": 8330 }, { "epoch": 0.75, "learning_rate": 7.307129993132627e-06, "loss": 1.2058, "step": 8331 }, { "epoch": 0.75, "learning_rate": 7.302066076380318e-06, "loss": 1.2167, "step": 8332 }, { "epoch": 0.75, "learning_rate": 7.2970036148131994e-06, "loss": 1.2122, "step": 8333 }, { "epoch": 0.75, "learning_rate": 7.291942608847521e-06, "loss": 1.1865, "step": 8334 }, { "epoch": 0.75, "learning_rate": 7.286883058899407e-06, "loss": 1.2015, "step": 8335 }, { "epoch": 0.75, "learning_rate": 7.281824965384893e-06, "loss": 1.1969, "step": 8336 }, { "epoch": 0.75, "learning_rate": 7.276768328719849e-06, "loss": 1.1391, "step": 8337 }, { "epoch": 0.75, "learning_rate": 7.27171314932005e-06, "loss": 1.1856, "step": 8338 }, { "epoch": 0.75, "learning_rate": 7.266659427601169e-06, "loss": 1.2044, "step": 8339 }, { "epoch": 0.75, "learning_rate": 7.261607163978718e-06, "loss": 1.2266, "step": 8340 }, { "epoch": 0.75, "learning_rate": 7.256556358868133e-06, "loss": 1.2915, "step": 8341 }, { "epoch": 0.75, "learning_rate": 7.251507012684705e-06, "loss": 1.2157, "step": 8342 }, { "epoch": 0.75, "learning_rate": 7.246459125843588e-06, "loss": 1.2424, "step": 8343 }, { "epoch": 0.75, "learning_rate": 7.241412698759861e-06, "loss": 1.2492, "step": 8344 }, { "epoch": 0.75, "learning_rate": 7.236367731848446e-06, "loss": 1.2226, "step": 8345 }, { "epoch": 0.75, "learning_rate": 7.231324225524152e-06, "loss": 1.2326, "step": 8346 }, { "epoch": 0.75, "learning_rate": 7.226282180201701e-06, "loss": 1.2267, "step": 8347 }, { "epoch": 0.75, "learning_rate": 7.221241596295631e-06, "loss": 1.2639, "step": 8348 }, { "epoch": 0.75, "learning_rate": 7.216202474220421e-06, "loss": 1.2473, "step": 8349 }, { "epoch": 0.75, "learning_rate": 7.211164814390398e-06, "loss": 1.1734, "step": 8350 }, { "epoch": 0.75, "learning_rate": 7.206128617219765e-06, "loss": 1.2364, "step": 8351 }, { "epoch": 0.75, "learning_rate": 7.201093883122642e-06, "loss": 1.1889, "step": 8352 }, { "epoch": 0.75, "learning_rate": 7.196060612512973e-06, "loss": 1.1868, "step": 8353 }, { "epoch": 0.75, "learning_rate": 7.191028805804628e-06, "loss": 1.1871, "step": 8354 }, { "epoch": 0.75, "learning_rate": 7.185998463411339e-06, "loss": 1.2128, "step": 8355 }, { "epoch": 0.75, "learning_rate": 7.180969585746705e-06, "loss": 1.1998, "step": 8356 }, { "epoch": 0.76, "learning_rate": 7.175942173224242e-06, "loss": 1.2201, "step": 8357 }, { "epoch": 0.76, "learning_rate": 7.1709162262572986e-06, "loss": 1.254, "step": 8358 }, { "epoch": 0.76, "learning_rate": 7.165891745259126e-06, "loss": 1.2224, "step": 8359 }, { "epoch": 0.76, "learning_rate": 7.160868730642869e-06, "loss": 1.1511, "step": 8360 }, { "epoch": 0.76, "learning_rate": 7.155847182821523e-06, "loss": 1.1978, "step": 8361 }, { "epoch": 0.76, "learning_rate": 7.150827102208e-06, "loss": 1.2069, "step": 8362 }, { "epoch": 0.76, "learning_rate": 7.145808489215041e-06, "loss": 1.2034, "step": 8363 }, { "epoch": 0.76, "learning_rate": 7.1407913442552985e-06, "loss": 1.2084, "step": 8364 }, { "epoch": 0.76, "learning_rate": 7.135775667741315e-06, "loss": 1.2313, "step": 8365 }, { "epoch": 0.76, "learning_rate": 7.130761460085486e-06, "loss": 1.2137, "step": 8366 }, { "epoch": 0.76, "learning_rate": 7.12574872170009e-06, "loss": 1.2201, "step": 8367 }, { "epoch": 0.76, "learning_rate": 7.120737452997317e-06, "loss": 1.2051, "step": 8368 }, { "epoch": 0.76, "learning_rate": 7.115727654389176e-06, "loss": 1.2124, "step": 8369 }, { "epoch": 0.76, "learning_rate": 7.110719326287618e-06, "loss": 1.2064, "step": 8370 }, { "epoch": 0.76, "learning_rate": 7.105712469104434e-06, "loss": 1.1967, "step": 8371 }, { "epoch": 0.76, "learning_rate": 7.100707083251299e-06, "loss": 1.2032, "step": 8372 }, { "epoch": 0.76, "learning_rate": 7.095703169139795e-06, "loss": 1.2337, "step": 8373 }, { "epoch": 0.76, "learning_rate": 7.090700727181329e-06, "loss": 1.175, "step": 8374 }, { "epoch": 0.76, "learning_rate": 7.085699757787246e-06, "loss": 1.2009, "step": 8375 }, { "epoch": 0.76, "learning_rate": 7.080700261368731e-06, "loss": 1.2254, "step": 8376 }, { "epoch": 0.76, "learning_rate": 7.075702238336854e-06, "loss": 1.1646, "step": 8377 }, { "epoch": 0.76, "learning_rate": 7.070705689102594e-06, "loss": 1.2488, "step": 8378 }, { "epoch": 0.76, "learning_rate": 7.0657106140767595e-06, "loss": 1.218, "step": 8379 }, { "epoch": 0.76, "learning_rate": 7.0607170136700655e-06, "loss": 1.2411, "step": 8380 }, { "epoch": 0.76, "learning_rate": 7.055724888293116e-06, "loss": 1.2679, "step": 8381 }, { "epoch": 0.76, "learning_rate": 7.050734238356365e-06, "loss": 1.2031, "step": 8382 }, { "epoch": 0.76, "learning_rate": 7.0457450642701876e-06, "loss": 1.1723, "step": 8383 }, { "epoch": 0.76, "learning_rate": 7.0407573664447805e-06, "loss": 1.2295, "step": 8384 }, { "epoch": 0.76, "learning_rate": 7.03577114529026e-06, "loss": 1.2476, "step": 8385 }, { "epoch": 0.76, "learning_rate": 7.030786401216616e-06, "loss": 1.2259, "step": 8386 }, { "epoch": 0.76, "learning_rate": 7.025803134633707e-06, "loss": 1.2314, "step": 8387 }, { "epoch": 0.76, "learning_rate": 7.0208213459512786e-06, "loss": 1.1895, "step": 8388 }, { "epoch": 0.76, "learning_rate": 7.015841035578946e-06, "loss": 1.1961, "step": 8389 }, { "epoch": 0.76, "learning_rate": 7.0108622039261995e-06, "loss": 1.2045, "step": 8390 }, { "epoch": 0.76, "learning_rate": 7.005884851402433e-06, "loss": 1.2414, "step": 8391 }, { "epoch": 0.76, "learning_rate": 7.000908978416892e-06, "loss": 1.1954, "step": 8392 }, { "epoch": 0.76, "learning_rate": 6.995934585378708e-06, "loss": 1.2428, "step": 8393 }, { "epoch": 0.76, "learning_rate": 6.990961672696908e-06, "loss": 1.2339, "step": 8394 }, { "epoch": 0.76, "learning_rate": 6.985990240780354e-06, "loss": 1.1904, "step": 8395 }, { "epoch": 0.76, "learning_rate": 6.98102029003784e-06, "loss": 1.2209, "step": 8396 }, { "epoch": 0.76, "learning_rate": 6.976051820878005e-06, "loss": 1.1895, "step": 8397 }, { "epoch": 0.76, "learning_rate": 6.9710848337093626e-06, "loss": 1.2344, "step": 8398 }, { "epoch": 0.76, "learning_rate": 6.966119328940338e-06, "loss": 1.1917, "step": 8399 }, { "epoch": 0.76, "learning_rate": 6.961155306979192e-06, "loss": 1.1581, "step": 8400 }, { "epoch": 0.76, "learning_rate": 6.956192768234085e-06, "loss": 1.2419, "step": 8401 }, { "epoch": 0.76, "learning_rate": 6.951231713113065e-06, "loss": 1.1919, "step": 8402 }, { "epoch": 0.76, "learning_rate": 6.946272142024035e-06, "loss": 1.1803, "step": 8403 }, { "epoch": 0.76, "learning_rate": 6.941314055374809e-06, "loss": 1.2447, "step": 8404 }, { "epoch": 0.76, "learning_rate": 6.936357453573036e-06, "loss": 1.1825, "step": 8405 }, { "epoch": 0.76, "learning_rate": 6.931402337026266e-06, "loss": 1.2368, "step": 8406 }, { "epoch": 0.76, "learning_rate": 6.926448706141939e-06, "loss": 1.208, "step": 8407 }, { "epoch": 0.76, "learning_rate": 6.921496561327354e-06, "loss": 1.2991, "step": 8408 }, { "epoch": 0.76, "learning_rate": 6.916545902989691e-06, "loss": 1.206, "step": 8409 }, { "epoch": 0.76, "learning_rate": 6.91159673153601e-06, "loss": 1.2249, "step": 8410 }, { "epoch": 0.76, "learning_rate": 6.906649047373246e-06, "loss": 1.1866, "step": 8411 }, { "epoch": 0.76, "learning_rate": 6.901702850908226e-06, "loss": 1.1816, "step": 8412 }, { "epoch": 0.76, "learning_rate": 6.896758142547635e-06, "loss": 1.1711, "step": 8413 }, { "epoch": 0.76, "learning_rate": 6.891814922698048e-06, "loss": 1.227, "step": 8414 }, { "epoch": 0.76, "learning_rate": 6.886873191765908e-06, "loss": 1.2197, "step": 8415 }, { "epoch": 0.76, "learning_rate": 6.881932950157538e-06, "loss": 1.2322, "step": 8416 }, { "epoch": 0.76, "learning_rate": 6.876994198279157e-06, "loss": 1.2253, "step": 8417 }, { "epoch": 0.76, "learning_rate": 6.872056936536836e-06, "loss": 1.1704, "step": 8418 }, { "epoch": 0.76, "learning_rate": 6.8671211653365265e-06, "loss": 1.1903, "step": 8419 }, { "epoch": 0.76, "learning_rate": 6.862186885084087e-06, "loss": 1.193, "step": 8420 }, { "epoch": 0.76, "learning_rate": 6.857254096185211e-06, "loss": 1.201, "step": 8421 }, { "epoch": 0.76, "learning_rate": 6.8523227990454906e-06, "loss": 1.1734, "step": 8422 }, { "epoch": 0.76, "learning_rate": 6.847392994070403e-06, "loss": 1.2706, "step": 8423 }, { "epoch": 0.76, "learning_rate": 6.842464681665292e-06, "loss": 1.205, "step": 8424 }, { "epoch": 0.76, "learning_rate": 6.8375378622353766e-06, "loss": 1.2162, "step": 8425 }, { "epoch": 0.76, "learning_rate": 6.832612536185759e-06, "loss": 1.226, "step": 8426 }, { "epoch": 0.76, "learning_rate": 6.827688703921406e-06, "loss": 1.2197, "step": 8427 }, { "epoch": 0.76, "learning_rate": 6.822766365847191e-06, "loss": 1.2004, "step": 8428 }, { "epoch": 0.76, "learning_rate": 6.817845522367838e-06, "loss": 1.2258, "step": 8429 }, { "epoch": 0.76, "learning_rate": 6.812926173887954e-06, "loss": 1.2147, "step": 8430 }, { "epoch": 0.76, "learning_rate": 6.808008320812026e-06, "loss": 1.2019, "step": 8431 }, { "epoch": 0.76, "learning_rate": 6.803091963544406e-06, "loss": 1.2068, "step": 8432 }, { "epoch": 0.76, "learning_rate": 6.7981771024893535e-06, "loss": 1.2463, "step": 8433 }, { "epoch": 0.76, "learning_rate": 6.7932637380509755e-06, "loss": 1.2311, "step": 8434 }, { "epoch": 0.76, "learning_rate": 6.788351870633264e-06, "loss": 1.2713, "step": 8435 }, { "epoch": 0.76, "learning_rate": 6.783441500640092e-06, "loss": 1.1855, "step": 8436 }, { "epoch": 0.76, "learning_rate": 6.778532628475201e-06, "loss": 1.2441, "step": 8437 }, { "epoch": 0.76, "learning_rate": 6.773625254542227e-06, "loss": 1.2267, "step": 8438 }, { "epoch": 0.76, "learning_rate": 6.768719379244667e-06, "loss": 1.2251, "step": 8439 }, { "epoch": 0.76, "learning_rate": 6.763815002985896e-06, "loss": 1.2148, "step": 8440 }, { "epoch": 0.76, "learning_rate": 6.75891212616917e-06, "loss": 1.1978, "step": 8441 }, { "epoch": 0.76, "learning_rate": 6.754010749197623e-06, "loss": 1.2723, "step": 8442 }, { "epoch": 0.76, "learning_rate": 6.749110872474249e-06, "loss": 1.2125, "step": 8443 }, { "epoch": 0.76, "learning_rate": 6.744212496401955e-06, "loss": 1.1753, "step": 8444 }, { "epoch": 0.76, "learning_rate": 6.739315621383491e-06, "loss": 1.2302, "step": 8445 }, { "epoch": 0.76, "learning_rate": 6.734420247821494e-06, "loss": 1.2214, "step": 8446 }, { "epoch": 0.76, "learning_rate": 6.7295263761184835e-06, "loss": 1.222, "step": 8447 }, { "epoch": 0.76, "learning_rate": 6.724634006676836e-06, "loss": 1.1766, "step": 8448 }, { "epoch": 0.76, "learning_rate": 6.71974313989884e-06, "loss": 1.1852, "step": 8449 }, { "epoch": 0.76, "learning_rate": 6.71485377618663e-06, "loss": 1.2054, "step": 8450 }, { "epoch": 0.76, "learning_rate": 6.709965915942226e-06, "loss": 1.1611, "step": 8451 }, { "epoch": 0.76, "learning_rate": 6.705079559567523e-06, "loss": 1.1975, "step": 8452 }, { "epoch": 0.76, "learning_rate": 6.700194707464291e-06, "loss": 1.1976, "step": 8453 }, { "epoch": 0.76, "learning_rate": 6.69531136003419e-06, "loss": 1.2275, "step": 8454 }, { "epoch": 0.76, "learning_rate": 6.69042951767874e-06, "loss": 1.1949, "step": 8455 }, { "epoch": 0.76, "learning_rate": 6.685549180799344e-06, "loss": 1.177, "step": 8456 }, { "epoch": 0.76, "learning_rate": 6.6806703497972776e-06, "loss": 1.2111, "step": 8457 }, { "epoch": 0.76, "learning_rate": 6.67579302507369e-06, "loss": 1.1856, "step": 8458 }, { "epoch": 0.76, "learning_rate": 6.670917207029623e-06, "loss": 1.2287, "step": 8459 }, { "epoch": 0.76, "learning_rate": 6.666042896065983e-06, "loss": 1.2342, "step": 8460 }, { "epoch": 0.76, "learning_rate": 6.661170092583543e-06, "loss": 1.2557, "step": 8461 }, { "epoch": 0.76, "learning_rate": 6.656298796982971e-06, "loss": 1.2235, "step": 8462 }, { "epoch": 0.76, "learning_rate": 6.651429009664795e-06, "loss": 1.2122, "step": 8463 }, { "epoch": 0.76, "learning_rate": 6.646560731029422e-06, "loss": 1.2425, "step": 8464 }, { "epoch": 0.76, "learning_rate": 6.64169396147715e-06, "loss": 1.1918, "step": 8465 }, { "epoch": 0.76, "learning_rate": 6.636828701408137e-06, "loss": 1.2307, "step": 8466 }, { "epoch": 0.77, "learning_rate": 6.631964951222425e-06, "loss": 1.1976, "step": 8467 }, { "epoch": 0.77, "learning_rate": 6.627102711319918e-06, "loss": 1.2167, "step": 8468 }, { "epoch": 0.77, "learning_rate": 6.622241982100411e-06, "loss": 1.1792, "step": 8469 }, { "epoch": 0.77, "learning_rate": 6.617382763963575e-06, "loss": 1.1998, "step": 8470 }, { "epoch": 0.77, "learning_rate": 6.612525057308949e-06, "loss": 1.1942, "step": 8471 }, { "epoch": 0.77, "learning_rate": 6.6076688625359484e-06, "loss": 1.1866, "step": 8472 }, { "epoch": 0.77, "learning_rate": 6.602814180043868e-06, "loss": 1.2196, "step": 8473 }, { "epoch": 0.77, "learning_rate": 6.597961010231868e-06, "loss": 1.1772, "step": 8474 }, { "epoch": 0.77, "learning_rate": 6.593109353499008e-06, "loss": 1.1572, "step": 8475 }, { "epoch": 0.77, "learning_rate": 6.588259210244199e-06, "loss": 1.2087, "step": 8476 }, { "epoch": 0.77, "learning_rate": 6.583410580866242e-06, "loss": 1.1937, "step": 8477 }, { "epoch": 0.77, "learning_rate": 6.578563465763801e-06, "loss": 1.2114, "step": 8478 }, { "epoch": 0.77, "learning_rate": 6.5737178653354184e-06, "loss": 1.2025, "step": 8479 }, { "epoch": 0.77, "learning_rate": 6.5688737799795315e-06, "loss": 1.2058, "step": 8480 }, { "epoch": 0.77, "learning_rate": 6.564031210094429e-06, "loss": 1.2717, "step": 8481 }, { "epoch": 0.77, "learning_rate": 6.559190156078285e-06, "loss": 1.1883, "step": 8482 }, { "epoch": 0.77, "learning_rate": 6.554350618329147e-06, "loss": 1.256, "step": 8483 }, { "epoch": 0.77, "learning_rate": 6.549512597244936e-06, "loss": 1.1692, "step": 8484 }, { "epoch": 0.77, "learning_rate": 6.544676093223448e-06, "loss": 1.2049, "step": 8485 }, { "epoch": 0.77, "learning_rate": 6.53984110666237e-06, "loss": 1.2507, "step": 8486 }, { "epoch": 0.77, "learning_rate": 6.5350076379592425e-06, "loss": 1.2148, "step": 8487 }, { "epoch": 0.77, "learning_rate": 6.530175687511492e-06, "loss": 1.1898, "step": 8488 }, { "epoch": 0.77, "learning_rate": 6.525345255716414e-06, "loss": 1.1772, "step": 8489 }, { "epoch": 0.77, "learning_rate": 6.520516342971183e-06, "loss": 1.1856, "step": 8490 }, { "epoch": 0.77, "learning_rate": 6.515688949672857e-06, "loss": 1.2148, "step": 8491 }, { "epoch": 0.77, "learning_rate": 6.510863076218357e-06, "loss": 1.1646, "step": 8492 }, { "epoch": 0.77, "learning_rate": 6.506038723004484e-06, "loss": 1.2349, "step": 8493 }, { "epoch": 0.77, "learning_rate": 6.501215890427909e-06, "loss": 1.1943, "step": 8494 }, { "epoch": 0.77, "learning_rate": 6.496394578885176e-06, "loss": 1.2907, "step": 8495 }, { "epoch": 0.77, "learning_rate": 6.491574788772728e-06, "loss": 1.1522, "step": 8496 }, { "epoch": 0.77, "learning_rate": 6.486756520486853e-06, "loss": 1.2305, "step": 8497 }, { "epoch": 0.77, "learning_rate": 6.4819397744237285e-06, "loss": 1.2137, "step": 8498 }, { "epoch": 0.77, "learning_rate": 6.477124550979402e-06, "loss": 1.2044, "step": 8499 }, { "epoch": 0.77, "learning_rate": 6.47231085054979e-06, "loss": 1.1716, "step": 8500 }, { "epoch": 0.77, "learning_rate": 6.4674986735307086e-06, "loss": 1.2351, "step": 8501 }, { "epoch": 0.77, "learning_rate": 6.462688020317822e-06, "loss": 1.1337, "step": 8502 }, { "epoch": 0.77, "learning_rate": 6.457878891306682e-06, "loss": 1.1761, "step": 8503 }, { "epoch": 0.77, "learning_rate": 6.453071286892709e-06, "loss": 1.2083, "step": 8504 }, { "epoch": 0.77, "learning_rate": 6.448265207471202e-06, "loss": 1.1732, "step": 8505 }, { "epoch": 0.77, "learning_rate": 6.4434606534373245e-06, "loss": 1.1955, "step": 8506 }, { "epoch": 0.77, "learning_rate": 6.4386576251861395e-06, "loss": 1.1858, "step": 8507 }, { "epoch": 0.77, "learning_rate": 6.433856123112561e-06, "loss": 1.2086, "step": 8508 }, { "epoch": 0.77, "learning_rate": 6.429056147611384e-06, "loss": 1.2245, "step": 8509 }, { "epoch": 0.77, "learning_rate": 6.4242576990772795e-06, "loss": 1.2124, "step": 8510 }, { "epoch": 0.77, "learning_rate": 6.4194607779047875e-06, "loss": 1.2585, "step": 8511 }, { "epoch": 0.77, "learning_rate": 6.414665384488341e-06, "loss": 1.1826, "step": 8512 }, { "epoch": 0.77, "learning_rate": 6.409871519222224e-06, "loss": 1.1611, "step": 8513 }, { "epoch": 0.77, "learning_rate": 6.405079182500609e-06, "loss": 1.1911, "step": 8514 }, { "epoch": 0.77, "learning_rate": 6.400288374717533e-06, "loss": 1.1743, "step": 8515 }, { "epoch": 0.77, "learning_rate": 6.395499096266908e-06, "loss": 1.2086, "step": 8516 }, { "epoch": 0.77, "learning_rate": 6.390711347542541e-06, "loss": 1.2293, "step": 8517 }, { "epoch": 0.77, "learning_rate": 6.385925128938089e-06, "loss": 1.2059, "step": 8518 }, { "epoch": 0.77, "learning_rate": 6.381140440847091e-06, "loss": 1.2329, "step": 8519 }, { "epoch": 0.77, "learning_rate": 6.376357283662962e-06, "loss": 1.1839, "step": 8520 }, { "epoch": 0.77, "learning_rate": 6.37157565777898e-06, "loss": 1.2237, "step": 8521 }, { "epoch": 0.77, "learning_rate": 6.3667955635883235e-06, "loss": 1.2032, "step": 8522 }, { "epoch": 0.77, "learning_rate": 6.362017001484019e-06, "loss": 1.2447, "step": 8523 }, { "epoch": 0.77, "learning_rate": 6.357239971858978e-06, "loss": 1.2244, "step": 8524 }, { "epoch": 0.77, "learning_rate": 6.352464475105982e-06, "loss": 1.2117, "step": 8525 }, { "epoch": 0.77, "learning_rate": 6.347690511617693e-06, "loss": 1.1552, "step": 8526 }, { "epoch": 0.77, "learning_rate": 6.342918081786633e-06, "loss": 1.1964, "step": 8527 }, { "epoch": 0.77, "learning_rate": 6.33814718600522e-06, "loss": 1.1773, "step": 8528 }, { "epoch": 0.77, "learning_rate": 6.333377824665732e-06, "loss": 1.1867, "step": 8529 }, { "epoch": 0.77, "learning_rate": 6.3286099981603144e-06, "loss": 1.1861, "step": 8530 }, { "epoch": 0.77, "learning_rate": 6.323843706881003e-06, "loss": 1.1919, "step": 8531 }, { "epoch": 0.77, "learning_rate": 6.319078951219684e-06, "loss": 1.2123, "step": 8532 }, { "epoch": 0.77, "learning_rate": 6.314315731568152e-06, "loss": 1.1902, "step": 8533 }, { "epoch": 0.77, "learning_rate": 6.309554048318045e-06, "loss": 1.1876, "step": 8534 }, { "epoch": 0.77, "learning_rate": 6.304793901860886e-06, "loss": 1.2205, "step": 8535 }, { "epoch": 0.77, "learning_rate": 6.300035292588072e-06, "loss": 1.2124, "step": 8536 }, { "epoch": 0.77, "learning_rate": 6.2952782208908625e-06, "loss": 1.2343, "step": 8537 }, { "epoch": 0.77, "learning_rate": 6.290522687160416e-06, "loss": 1.1903, "step": 8538 }, { "epoch": 0.77, "learning_rate": 6.2857686917877415e-06, "loss": 1.2481, "step": 8539 }, { "epoch": 0.77, "learning_rate": 6.281016235163728e-06, "loss": 1.2162, "step": 8540 }, { "epoch": 0.77, "learning_rate": 6.276265317679142e-06, "loss": 1.1714, "step": 8541 }, { "epoch": 0.77, "learning_rate": 6.271515939724609e-06, "loss": 1.169, "step": 8542 }, { "epoch": 0.77, "learning_rate": 6.266768101690656e-06, "loss": 1.1675, "step": 8543 }, { "epoch": 0.77, "learning_rate": 6.2620218039676605e-06, "loss": 1.2822, "step": 8544 }, { "epoch": 0.77, "learning_rate": 6.257277046945878e-06, "loss": 1.2184, "step": 8545 }, { "epoch": 0.77, "learning_rate": 6.252533831015439e-06, "loss": 1.2178, "step": 8546 }, { "epoch": 0.77, "learning_rate": 6.247792156566345e-06, "loss": 1.1812, "step": 8547 }, { "epoch": 0.77, "learning_rate": 6.2430520239884685e-06, "loss": 1.2348, "step": 8548 }, { "epoch": 0.77, "learning_rate": 6.238313433671575e-06, "loss": 1.174, "step": 8549 }, { "epoch": 0.77, "learning_rate": 6.233576386005274e-06, "loss": 1.2342, "step": 8550 }, { "epoch": 0.77, "learning_rate": 6.228840881379069e-06, "loss": 1.2067, "step": 8551 }, { "epoch": 0.77, "learning_rate": 6.22410692018233e-06, "loss": 1.1391, "step": 8552 }, { "epoch": 0.77, "learning_rate": 6.219374502804284e-06, "loss": 1.2206, "step": 8553 }, { "epoch": 0.77, "learning_rate": 6.214643629634071e-06, "loss": 1.212, "step": 8554 }, { "epoch": 0.77, "learning_rate": 6.209914301060665e-06, "loss": 1.2238, "step": 8555 }, { "epoch": 0.77, "learning_rate": 6.205186517472933e-06, "loss": 1.2309, "step": 8556 }, { "epoch": 0.77, "learning_rate": 6.200460279259607e-06, "loss": 1.1799, "step": 8557 }, { "epoch": 0.77, "learning_rate": 6.195735586809287e-06, "loss": 1.203, "step": 8558 }, { "epoch": 0.77, "learning_rate": 6.1910124405104686e-06, "loss": 1.2106, "step": 8559 }, { "epoch": 0.77, "learning_rate": 6.186290840751496e-06, "loss": 1.2064, "step": 8560 }, { "epoch": 0.77, "learning_rate": 6.181570787920601e-06, "loss": 1.2143, "step": 8561 }, { "epoch": 0.77, "learning_rate": 6.176852282405879e-06, "loss": 1.2066, "step": 8562 }, { "epoch": 0.77, "learning_rate": 6.1721353245952935e-06, "loss": 1.1898, "step": 8563 }, { "epoch": 0.77, "learning_rate": 6.167419914876704e-06, "loss": 1.2082, "step": 8564 }, { "epoch": 0.77, "learning_rate": 6.162706053637829e-06, "loss": 1.238, "step": 8565 }, { "epoch": 0.77, "learning_rate": 6.157993741266238e-06, "loss": 1.2133, "step": 8566 }, { "epoch": 0.77, "learning_rate": 6.153282978149411e-06, "loss": 1.2189, "step": 8567 }, { "epoch": 0.77, "learning_rate": 6.148573764674681e-06, "loss": 1.1975, "step": 8568 }, { "epoch": 0.77, "learning_rate": 6.143866101229242e-06, "loss": 1.2011, "step": 8569 }, { "epoch": 0.77, "learning_rate": 6.139159988200197e-06, "loss": 1.1544, "step": 8570 }, { "epoch": 0.77, "learning_rate": 6.134455425974487e-06, "loss": 1.1798, "step": 8571 }, { "epoch": 0.77, "learning_rate": 6.129752414938936e-06, "loss": 1.2293, "step": 8572 }, { "epoch": 0.77, "learning_rate": 6.125050955480244e-06, "loss": 1.204, "step": 8573 }, { "epoch": 0.77, "learning_rate": 6.120351047984974e-06, "loss": 1.1869, "step": 8574 }, { "epoch": 0.77, "learning_rate": 6.115652692839582e-06, "loss": 1.2027, "step": 8575 }, { "epoch": 0.77, "learning_rate": 6.110955890430379e-06, "loss": 1.2283, "step": 8576 }, { "epoch": 0.77, "learning_rate": 6.106260641143546e-06, "loss": 1.1874, "step": 8577 }, { "epoch": 0.78, "learning_rate": 6.1015669453651474e-06, "loss": 1.2445, "step": 8578 }, { "epoch": 0.78, "learning_rate": 6.0968748034811064e-06, "loss": 1.2184, "step": 8579 }, { "epoch": 0.78, "learning_rate": 6.092184215877242e-06, "loss": 1.2071, "step": 8580 }, { "epoch": 0.78, "learning_rate": 6.0874951829392234e-06, "loss": 1.1884, "step": 8581 }, { "epoch": 0.78, "learning_rate": 6.082807705052598e-06, "loss": 1.201, "step": 8582 }, { "epoch": 0.78, "learning_rate": 6.078121782602788e-06, "loss": 1.2721, "step": 8583 }, { "epoch": 0.78, "learning_rate": 6.073437415975078e-06, "loss": 1.183, "step": 8584 }, { "epoch": 0.78, "learning_rate": 6.06875460555465e-06, "loss": 1.1922, "step": 8585 }, { "epoch": 0.78, "learning_rate": 6.0640733517265346e-06, "loss": 1.1908, "step": 8586 }, { "epoch": 0.78, "learning_rate": 6.059393654875623e-06, "loss": 1.1639, "step": 8587 }, { "epoch": 0.78, "learning_rate": 6.054715515386717e-06, "loss": 1.2038, "step": 8588 }, { "epoch": 0.78, "learning_rate": 6.050038933644461e-06, "loss": 1.193, "step": 8589 }, { "epoch": 0.78, "learning_rate": 6.045363910033378e-06, "loss": 1.1637, "step": 8590 }, { "epoch": 0.78, "learning_rate": 6.04069044493788e-06, "loss": 1.1981, "step": 8591 }, { "epoch": 0.78, "learning_rate": 6.036018538742208e-06, "loss": 1.2496, "step": 8592 }, { "epoch": 0.78, "learning_rate": 6.0313481918305245e-06, "loss": 1.2578, "step": 8593 }, { "epoch": 0.78, "learning_rate": 6.0266794045868326e-06, "loss": 1.204, "step": 8594 }, { "epoch": 0.78, "learning_rate": 6.022012177395012e-06, "loss": 1.2225, "step": 8595 }, { "epoch": 0.78, "learning_rate": 6.017346510638833e-06, "loss": 1.2506, "step": 8596 }, { "epoch": 0.78, "learning_rate": 6.012682404701911e-06, "loss": 1.2314, "step": 8597 }, { "epoch": 0.78, "learning_rate": 6.0080198599677504e-06, "loss": 1.2214, "step": 8598 }, { "epoch": 0.78, "learning_rate": 6.003358876819717e-06, "loss": 1.2362, "step": 8599 }, { "epoch": 0.78, "learning_rate": 5.9986994556410465e-06, "loss": 1.1855, "step": 8600 }, { "epoch": 0.78, "learning_rate": 5.99404159681487e-06, "loss": 1.1757, "step": 8601 }, { "epoch": 0.78, "learning_rate": 5.989385300724165e-06, "loss": 1.2252, "step": 8602 }, { "epoch": 0.78, "learning_rate": 5.984730567751787e-06, "loss": 1.1552, "step": 8603 }, { "epoch": 0.78, "learning_rate": 5.980077398280462e-06, "loss": 1.1605, "step": 8604 }, { "epoch": 0.78, "learning_rate": 5.975425792692785e-06, "loss": 1.2125, "step": 8605 }, { "epoch": 0.78, "learning_rate": 5.970775751371241e-06, "loss": 1.2557, "step": 8606 }, { "epoch": 0.78, "learning_rate": 5.9661272746981745e-06, "loss": 1.1438, "step": 8607 }, { "epoch": 0.78, "learning_rate": 5.961480363055774e-06, "loss": 1.2099, "step": 8608 }, { "epoch": 0.78, "learning_rate": 5.956835016826145e-06, "loss": 1.2511, "step": 8609 }, { "epoch": 0.78, "learning_rate": 5.952191236391244e-06, "loss": 1.2559, "step": 8610 }, { "epoch": 0.78, "learning_rate": 5.947549022132886e-06, "loss": 1.2184, "step": 8611 }, { "epoch": 0.78, "learning_rate": 5.942908374432793e-06, "loss": 1.2106, "step": 8612 }, { "epoch": 0.78, "learning_rate": 5.938269293672505e-06, "loss": 1.2339, "step": 8613 }, { "epoch": 0.78, "learning_rate": 5.933631780233484e-06, "loss": 1.2715, "step": 8614 }, { "epoch": 0.78, "learning_rate": 5.928995834497039e-06, "loss": 1.2258, "step": 8615 }, { "epoch": 0.78, "learning_rate": 5.924361456844343e-06, "loss": 1.2079, "step": 8616 }, { "epoch": 0.78, "learning_rate": 5.919728647656467e-06, "loss": 1.2096, "step": 8617 }, { "epoch": 0.78, "learning_rate": 5.915097407314327e-06, "loss": 1.2489, "step": 8618 }, { "epoch": 0.78, "learning_rate": 5.910467736198722e-06, "loss": 1.1672, "step": 8619 }, { "epoch": 0.78, "learning_rate": 5.905839634690319e-06, "loss": 1.1655, "step": 8620 }, { "epoch": 0.78, "learning_rate": 5.901213103169645e-06, "loss": 1.2221, "step": 8621 }, { "epoch": 0.78, "learning_rate": 5.896588142017131e-06, "loss": 1.229, "step": 8622 }, { "epoch": 0.78, "learning_rate": 5.891964751613046e-06, "loss": 1.1989, "step": 8623 }, { "epoch": 0.78, "learning_rate": 5.887342932337539e-06, "loss": 1.2307, "step": 8624 }, { "epoch": 0.78, "learning_rate": 5.882722684570638e-06, "loss": 1.1653, "step": 8625 }, { "epoch": 0.78, "learning_rate": 5.878104008692223e-06, "loss": 1.2145, "step": 8626 }, { "epoch": 0.78, "learning_rate": 5.873486905082073e-06, "loss": 1.1748, "step": 8627 }, { "epoch": 0.78, "learning_rate": 5.868871374119825e-06, "loss": 1.2764, "step": 8628 }, { "epoch": 0.78, "learning_rate": 5.864257416184959e-06, "loss": 1.2513, "step": 8629 }, { "epoch": 0.78, "learning_rate": 5.859645031656877e-06, "loss": 1.2284, "step": 8630 }, { "epoch": 0.78, "learning_rate": 5.855034220914812e-06, "loss": 1.1824, "step": 8631 }, { "epoch": 0.78, "learning_rate": 5.850424984337874e-06, "loss": 1.2514, "step": 8632 }, { "epoch": 0.78, "learning_rate": 5.8458173223050775e-06, "loss": 1.1745, "step": 8633 }, { "epoch": 0.78, "learning_rate": 5.841211235195246e-06, "loss": 1.2495, "step": 8634 }, { "epoch": 0.78, "learning_rate": 5.8366067233871306e-06, "loss": 1.1877, "step": 8635 }, { "epoch": 0.78, "learning_rate": 5.832003787259327e-06, "loss": 1.204, "step": 8636 }, { "epoch": 0.78, "learning_rate": 5.827402427190293e-06, "loss": 1.1891, "step": 8637 }, { "epoch": 0.78, "learning_rate": 5.8228026435583895e-06, "loss": 1.1651, "step": 8638 }, { "epoch": 0.78, "learning_rate": 5.8182044367418e-06, "loss": 1.1883, "step": 8639 }, { "epoch": 0.78, "learning_rate": 5.813607807118626e-06, "loss": 1.1574, "step": 8640 }, { "epoch": 0.78, "learning_rate": 5.809012755066812e-06, "loss": 1.2149, "step": 8641 }, { "epoch": 0.78, "learning_rate": 5.804419280964169e-06, "loss": 1.1641, "step": 8642 }, { "epoch": 0.78, "learning_rate": 5.799827385188405e-06, "loss": 1.1786, "step": 8643 }, { "epoch": 0.78, "learning_rate": 5.795237068117074e-06, "loss": 1.2591, "step": 8644 }, { "epoch": 0.78, "learning_rate": 5.790648330127607e-06, "loss": 1.2045, "step": 8645 }, { "epoch": 0.78, "learning_rate": 5.786061171597307e-06, "loss": 1.2391, "step": 8646 }, { "epoch": 0.78, "learning_rate": 5.781475592903338e-06, "loss": 1.159, "step": 8647 }, { "epoch": 0.78, "learning_rate": 5.776891594422756e-06, "loss": 1.2578, "step": 8648 }, { "epoch": 0.78, "learning_rate": 5.772309176532473e-06, "loss": 1.1785, "step": 8649 }, { "epoch": 0.78, "learning_rate": 5.767728339609252e-06, "loss": 1.1867, "step": 8650 }, { "epoch": 0.78, "learning_rate": 5.763149084029764e-06, "loss": 1.2266, "step": 8651 }, { "epoch": 0.78, "learning_rate": 5.758571410170527e-06, "loss": 1.1792, "step": 8652 }, { "epoch": 0.78, "learning_rate": 5.753995318407923e-06, "loss": 1.1858, "step": 8653 }, { "epoch": 0.78, "learning_rate": 5.7494208091182364e-06, "loss": 1.1955, "step": 8654 }, { "epoch": 0.78, "learning_rate": 5.744847882677573e-06, "loss": 1.2122, "step": 8655 }, { "epoch": 0.78, "learning_rate": 5.74027653946195e-06, "loss": 1.1826, "step": 8656 }, { "epoch": 0.78, "learning_rate": 5.735706779847239e-06, "loss": 1.1865, "step": 8657 }, { "epoch": 0.78, "learning_rate": 5.731138604209169e-06, "loss": 1.1931, "step": 8658 }, { "epoch": 0.78, "learning_rate": 5.726572012923376e-06, "loss": 1.216, "step": 8659 }, { "epoch": 0.78, "learning_rate": 5.72200700636531e-06, "loss": 1.1804, "step": 8660 }, { "epoch": 0.78, "learning_rate": 5.7174435849103455e-06, "loss": 1.2316, "step": 8661 }, { "epoch": 0.78, "learning_rate": 5.712881748933691e-06, "loss": 1.252, "step": 8662 }, { "epoch": 0.78, "learning_rate": 5.708321498810434e-06, "loss": 1.1757, "step": 8663 }, { "epoch": 0.78, "learning_rate": 5.7037628349155545e-06, "loss": 1.1579, "step": 8664 }, { "epoch": 0.78, "learning_rate": 5.699205757623852e-06, "loss": 1.2316, "step": 8665 }, { "epoch": 0.78, "learning_rate": 5.6946502673100445e-06, "loss": 1.179, "step": 8666 }, { "epoch": 0.78, "learning_rate": 5.690096364348699e-06, "loss": 1.2591, "step": 8667 }, { "epoch": 0.78, "learning_rate": 5.6855440491142385e-06, "loss": 1.2482, "step": 8668 }, { "epoch": 0.78, "learning_rate": 5.6809933219809905e-06, "loss": 1.1753, "step": 8669 }, { "epoch": 0.78, "learning_rate": 5.676444183323129e-06, "loss": 1.2267, "step": 8670 }, { "epoch": 0.78, "learning_rate": 5.671896633514676e-06, "loss": 1.213, "step": 8671 }, { "epoch": 0.78, "learning_rate": 5.667350672929572e-06, "loss": 1.2833, "step": 8672 }, { "epoch": 0.78, "learning_rate": 5.662806301941595e-06, "loss": 1.2137, "step": 8673 }, { "epoch": 0.78, "learning_rate": 5.658263520924387e-06, "loss": 1.2463, "step": 8674 }, { "epoch": 0.78, "learning_rate": 5.653722330251496e-06, "loss": 1.2032, "step": 8675 }, { "epoch": 0.78, "learning_rate": 5.649182730296288e-06, "loss": 1.2224, "step": 8676 }, { "epoch": 0.78, "learning_rate": 5.644644721432038e-06, "loss": 1.2371, "step": 8677 }, { "epoch": 0.78, "learning_rate": 5.64010830403188e-06, "loss": 1.2165, "step": 8678 }, { "epoch": 0.78, "learning_rate": 5.635573478468798e-06, "loss": 1.1919, "step": 8679 }, { "epoch": 0.78, "learning_rate": 5.631040245115687e-06, "loss": 1.2087, "step": 8680 }, { "epoch": 0.78, "learning_rate": 5.6265086043452546e-06, "loss": 1.2224, "step": 8681 }, { "epoch": 0.78, "learning_rate": 5.62197855653013e-06, "loss": 1.1885, "step": 8682 }, { "epoch": 0.78, "learning_rate": 5.617450102042784e-06, "loss": 1.2321, "step": 8683 }, { "epoch": 0.78, "learning_rate": 5.612923241255552e-06, "loss": 1.2768, "step": 8684 }, { "epoch": 0.78, "learning_rate": 5.608397974540671e-06, "loss": 1.1976, "step": 8685 }, { "epoch": 0.78, "learning_rate": 5.603874302270196e-06, "loss": 1.2014, "step": 8686 }, { "epoch": 0.78, "learning_rate": 5.599352224816098e-06, "loss": 1.221, "step": 8687 }, { "epoch": 0.78, "learning_rate": 5.5948317425501915e-06, "loss": 1.219, "step": 8688 }, { "epoch": 0.79, "learning_rate": 5.59031285584416e-06, "loss": 1.1688, "step": 8689 }, { "epoch": 0.79, "learning_rate": 5.5857955650695845e-06, "loss": 1.1987, "step": 8690 }, { "epoch": 0.79, "learning_rate": 5.581279870597867e-06, "loss": 1.1853, "step": 8691 }, { "epoch": 0.79, "learning_rate": 5.576765772800308e-06, "loss": 1.1766, "step": 8692 }, { "epoch": 0.79, "learning_rate": 5.572253272048083e-06, "loss": 1.2229, "step": 8693 }, { "epoch": 0.79, "learning_rate": 5.56774236871222e-06, "loss": 1.1977, "step": 8694 }, { "epoch": 0.79, "learning_rate": 5.563233063163612e-06, "loss": 1.1887, "step": 8695 }, { "epoch": 0.79, "learning_rate": 5.5587253557730535e-06, "loss": 1.2229, "step": 8696 }, { "epoch": 0.79, "learning_rate": 5.554219246911155e-06, "loss": 1.2435, "step": 8697 }, { "epoch": 0.79, "learning_rate": 5.549714736948441e-06, "loss": 1.251, "step": 8698 }, { "epoch": 0.79, "learning_rate": 5.545211826255289e-06, "loss": 1.2284, "step": 8699 }, { "epoch": 0.79, "learning_rate": 5.54071051520193e-06, "loss": 1.2478, "step": 8700 }, { "epoch": 0.79, "learning_rate": 5.536210804158501e-06, "loss": 1.2082, "step": 8701 }, { "epoch": 0.79, "learning_rate": 5.531712693494956e-06, "loss": 1.2345, "step": 8702 }, { "epoch": 0.79, "learning_rate": 5.5272161835811655e-06, "loss": 1.2092, "step": 8703 }, { "epoch": 0.79, "learning_rate": 5.522721274786843e-06, "loss": 1.1809, "step": 8704 }, { "epoch": 0.79, "learning_rate": 5.518227967481565e-06, "loss": 1.2083, "step": 8705 }, { "epoch": 0.79, "learning_rate": 5.513736262034808e-06, "loss": 1.1582, "step": 8706 }, { "epoch": 0.79, "learning_rate": 5.50924615881587e-06, "loss": 1.1868, "step": 8707 }, { "epoch": 0.79, "learning_rate": 5.504757658193963e-06, "loss": 1.2132, "step": 8708 }, { "epoch": 0.79, "learning_rate": 5.50027076053814e-06, "loss": 1.2375, "step": 8709 }, { "epoch": 0.79, "learning_rate": 5.495785466217321e-06, "loss": 1.1874, "step": 8710 }, { "epoch": 0.79, "learning_rate": 5.4913017756003274e-06, "loss": 1.1817, "step": 8711 }, { "epoch": 0.79, "learning_rate": 5.486819689055794e-06, "loss": 1.2258, "step": 8712 }, { "epoch": 0.79, "learning_rate": 5.482339206952264e-06, "loss": 1.2037, "step": 8713 }, { "epoch": 0.79, "learning_rate": 5.477860329658144e-06, "loss": 1.2141, "step": 8714 }, { "epoch": 0.79, "learning_rate": 5.4733830575417e-06, "loss": 1.2525, "step": 8715 }, { "epoch": 0.79, "learning_rate": 5.4689073909710676e-06, "loss": 1.1865, "step": 8716 }, { "epoch": 0.79, "learning_rate": 5.46443333031425e-06, "loss": 1.244, "step": 8717 }, { "epoch": 0.79, "learning_rate": 5.459960875939119e-06, "loss": 1.1939, "step": 8718 }, { "epoch": 0.79, "learning_rate": 5.45549002821342e-06, "loss": 1.1481, "step": 8719 }, { "epoch": 0.79, "learning_rate": 5.45102078750476e-06, "loss": 1.1913, "step": 8720 }, { "epoch": 0.79, "learning_rate": 5.446553154180608e-06, "loss": 1.1805, "step": 8721 }, { "epoch": 0.79, "learning_rate": 5.4420871286083304e-06, "loss": 1.2083, "step": 8722 }, { "epoch": 0.79, "learning_rate": 5.437622711155108e-06, "loss": 1.1608, "step": 8723 }, { "epoch": 0.79, "learning_rate": 5.4331599021880425e-06, "loss": 1.1982, "step": 8724 }, { "epoch": 0.79, "learning_rate": 5.4286987020740755e-06, "loss": 1.2103, "step": 8725 }, { "epoch": 0.79, "learning_rate": 5.4242391111800154e-06, "loss": 1.2111, "step": 8726 }, { "epoch": 0.79, "learning_rate": 5.419781129872567e-06, "loss": 1.1952, "step": 8727 }, { "epoch": 0.79, "learning_rate": 5.41532475851825e-06, "loss": 1.1709, "step": 8728 }, { "epoch": 0.79, "learning_rate": 5.4108699974835055e-06, "loss": 1.1968, "step": 8729 }, { "epoch": 0.79, "learning_rate": 5.4064168471346124e-06, "loss": 1.1934, "step": 8730 }, { "epoch": 0.79, "learning_rate": 5.401965307837717e-06, "loss": 1.211, "step": 8731 }, { "epoch": 0.79, "learning_rate": 5.397515379958862e-06, "loss": 1.1994, "step": 8732 }, { "epoch": 0.79, "learning_rate": 5.393067063863913e-06, "loss": 1.2196, "step": 8733 }, { "epoch": 0.79, "learning_rate": 5.388620359918625e-06, "loss": 1.2249, "step": 8734 }, { "epoch": 0.79, "learning_rate": 5.3841752684886395e-06, "loss": 1.1766, "step": 8735 }, { "epoch": 0.79, "learning_rate": 5.379731789939435e-06, "loss": 1.1924, "step": 8736 }, { "epoch": 0.79, "learning_rate": 5.375289924636373e-06, "loss": 1.2137, "step": 8737 }, { "epoch": 0.79, "learning_rate": 5.370849672944681e-06, "loss": 1.2279, "step": 8738 }, { "epoch": 0.79, "learning_rate": 5.36641103522944e-06, "loss": 1.181, "step": 8739 }, { "epoch": 0.79, "learning_rate": 5.361974011855628e-06, "loss": 1.2428, "step": 8740 }, { "epoch": 0.79, "learning_rate": 5.357538603188061e-06, "loss": 1.2189, "step": 8741 }, { "epoch": 0.79, "learning_rate": 5.353104809591439e-06, "loss": 1.1452, "step": 8742 }, { "epoch": 0.79, "learning_rate": 5.348672631430318e-06, "loss": 1.2153, "step": 8743 }, { "epoch": 0.79, "learning_rate": 5.344242069069125e-06, "loss": 1.1991, "step": 8744 }, { "epoch": 0.79, "learning_rate": 5.339813122872167e-06, "loss": 1.1703, "step": 8745 }, { "epoch": 0.79, "learning_rate": 5.335385793203604e-06, "loss": 1.1811, "step": 8746 }, { "epoch": 0.79, "learning_rate": 5.3309600804274555e-06, "loss": 1.2159, "step": 8747 }, { "epoch": 0.79, "learning_rate": 5.3265359849076405e-06, "loss": 1.225, "step": 8748 }, { "epoch": 0.79, "learning_rate": 5.322113507007898e-06, "loss": 1.2563, "step": 8749 }, { "epoch": 0.79, "learning_rate": 5.317692647091876e-06, "loss": 1.1632, "step": 8750 }, { "epoch": 0.79, "learning_rate": 5.313273405523067e-06, "loss": 1.2242, "step": 8751 }, { "epoch": 0.79, "learning_rate": 5.3088557826648345e-06, "loss": 1.2047, "step": 8752 }, { "epoch": 0.79, "learning_rate": 5.304439778880424e-06, "loss": 1.2559, "step": 8753 }, { "epoch": 0.79, "learning_rate": 5.300025394532921e-06, "loss": 1.2411, "step": 8754 }, { "epoch": 0.79, "learning_rate": 5.295612629985286e-06, "loss": 1.2737, "step": 8755 }, { "epoch": 0.79, "learning_rate": 5.291201485600366e-06, "loss": 1.2074, "step": 8756 }, { "epoch": 0.79, "learning_rate": 5.2867919617408556e-06, "loss": 1.2079, "step": 8757 }, { "epoch": 0.79, "learning_rate": 5.282384058769321e-06, "loss": 1.2129, "step": 8758 }, { "epoch": 0.79, "learning_rate": 5.277977777048196e-06, "loss": 1.2607, "step": 8759 }, { "epoch": 0.79, "learning_rate": 5.2735731169397704e-06, "loss": 1.1889, "step": 8760 }, { "epoch": 0.79, "learning_rate": 5.269170078806227e-06, "loss": 1.2363, "step": 8761 }, { "epoch": 0.79, "learning_rate": 5.26476866300959e-06, "loss": 1.2831, "step": 8762 }, { "epoch": 0.79, "learning_rate": 5.260368869911758e-06, "loss": 1.2154, "step": 8763 }, { "epoch": 0.79, "learning_rate": 5.255970699874502e-06, "loss": 1.2073, "step": 8764 }, { "epoch": 0.79, "learning_rate": 5.251574153259445e-06, "loss": 1.2141, "step": 8765 }, { "epoch": 0.79, "learning_rate": 5.247179230428098e-06, "loss": 1.1863, "step": 8766 }, { "epoch": 0.79, "learning_rate": 5.242785931741823e-06, "loss": 1.2107, "step": 8767 }, { "epoch": 0.79, "learning_rate": 5.238394257561843e-06, "loss": 1.2667, "step": 8768 }, { "epoch": 0.79, "learning_rate": 5.234004208249277e-06, "loss": 1.1867, "step": 8769 }, { "epoch": 0.79, "learning_rate": 5.2296157841650735e-06, "loss": 1.1601, "step": 8770 }, { "epoch": 0.79, "learning_rate": 5.225228985670058e-06, "loss": 1.1389, "step": 8771 }, { "epoch": 0.79, "learning_rate": 5.2208438131249455e-06, "loss": 1.2407, "step": 8772 }, { "epoch": 0.79, "learning_rate": 5.2164602668902854e-06, "loss": 1.2436, "step": 8773 }, { "epoch": 0.79, "learning_rate": 5.212078347326532e-06, "loss": 1.2221, "step": 8774 }, { "epoch": 0.79, "learning_rate": 5.207698054793955e-06, "loss": 1.2, "step": 8775 }, { "epoch": 0.79, "learning_rate": 5.2033193896527225e-06, "loss": 1.2138, "step": 8776 }, { "epoch": 0.79, "learning_rate": 5.198942352262873e-06, "loss": 1.19, "step": 8777 }, { "epoch": 0.79, "learning_rate": 5.194566942984297e-06, "loss": 1.2213, "step": 8778 }, { "epoch": 0.79, "learning_rate": 5.190193162176754e-06, "loss": 1.2533, "step": 8779 }, { "epoch": 0.79, "learning_rate": 5.185821010199876e-06, "loss": 1.1814, "step": 8780 }, { "epoch": 0.79, "learning_rate": 5.1814504874131445e-06, "loss": 1.2376, "step": 8781 }, { "epoch": 0.79, "learning_rate": 5.177081594175934e-06, "loss": 1.1662, "step": 8782 }, { "epoch": 0.79, "learning_rate": 5.172714330847467e-06, "loss": 1.2097, "step": 8783 }, { "epoch": 0.79, "learning_rate": 5.168348697786826e-06, "loss": 1.1966, "step": 8784 }, { "epoch": 0.79, "learning_rate": 5.163984695352977e-06, "loss": 1.204, "step": 8785 }, { "epoch": 0.79, "learning_rate": 5.1596223239047365e-06, "loss": 1.1416, "step": 8786 }, { "epoch": 0.79, "learning_rate": 5.1552615838008e-06, "loss": 1.1815, "step": 8787 }, { "epoch": 0.79, "learning_rate": 5.1509024753997215e-06, "loss": 1.2122, "step": 8788 }, { "epoch": 0.79, "learning_rate": 5.146544999059921e-06, "loss": 1.2005, "step": 8789 }, { "epoch": 0.79, "learning_rate": 5.1421891551396844e-06, "loss": 1.2018, "step": 8790 }, { "epoch": 0.79, "learning_rate": 5.1378349439971666e-06, "loss": 1.2481, "step": 8791 }, { "epoch": 0.79, "learning_rate": 5.133482365990375e-06, "loss": 1.1864, "step": 8792 }, { "epoch": 0.79, "learning_rate": 5.12913142147721e-06, "loss": 1.2423, "step": 8793 }, { "epoch": 0.79, "learning_rate": 5.1247821108154074e-06, "loss": 1.1809, "step": 8794 }, { "epoch": 0.79, "learning_rate": 5.120434434362601e-06, "loss": 1.1911, "step": 8795 }, { "epoch": 0.79, "learning_rate": 5.116088392476254e-06, "loss": 1.1549, "step": 8796 }, { "epoch": 0.79, "learning_rate": 5.111743985513709e-06, "loss": 1.1618, "step": 8797 }, { "epoch": 0.79, "learning_rate": 5.107401213832197e-06, "loss": 1.1853, "step": 8798 }, { "epoch": 0.8, "learning_rate": 5.103060077788788e-06, "loss": 1.2498, "step": 8799 }, { "epoch": 0.8, "learning_rate": 5.098720577740421e-06, "loss": 1.1707, "step": 8800 }, { "epoch": 0.8, "learning_rate": 5.094382714043907e-06, "loss": 1.2177, "step": 8801 }, { "epoch": 0.8, "learning_rate": 5.090046487055916e-06, "loss": 1.1552, "step": 8802 }, { "epoch": 0.8, "learning_rate": 5.085711897132997e-06, "loss": 1.1917, "step": 8803 }, { "epoch": 0.8, "learning_rate": 5.081378944631551e-06, "loss": 1.2297, "step": 8804 }, { "epoch": 0.8, "learning_rate": 5.077047629907847e-06, "loss": 1.1402, "step": 8805 }, { "epoch": 0.8, "learning_rate": 5.072717953318021e-06, "loss": 1.168, "step": 8806 }, { "epoch": 0.8, "learning_rate": 5.0683899152180665e-06, "loss": 1.1894, "step": 8807 }, { "epoch": 0.8, "learning_rate": 5.064063515963863e-06, "loss": 1.1907, "step": 8808 }, { "epoch": 0.8, "learning_rate": 5.0597387559111365e-06, "loss": 1.2136, "step": 8809 }, { "epoch": 0.8, "learning_rate": 5.055415635415483e-06, "loss": 1.12, "step": 8810 }, { "epoch": 0.8, "learning_rate": 5.0510941548323644e-06, "loss": 1.2062, "step": 8811 }, { "epoch": 0.8, "learning_rate": 5.046774314517108e-06, "loss": 1.1971, "step": 8812 }, { "epoch": 0.8, "learning_rate": 5.0424561148249e-06, "loss": 1.1904, "step": 8813 }, { "epoch": 0.8, "learning_rate": 5.0381395561108106e-06, "loss": 1.1993, "step": 8814 }, { "epoch": 0.8, "learning_rate": 5.033824638729753e-06, "loss": 1.2073, "step": 8815 }, { "epoch": 0.8, "learning_rate": 5.029511363036518e-06, "loss": 1.2024, "step": 8816 }, { "epoch": 0.8, "learning_rate": 5.025199729385755e-06, "loss": 1.2167, "step": 8817 }, { "epoch": 0.8, "learning_rate": 5.020889738131978e-06, "loss": 1.2113, "step": 8818 }, { "epoch": 0.8, "learning_rate": 5.016581389629579e-06, "loss": 1.2128, "step": 8819 }, { "epoch": 0.8, "learning_rate": 5.012274684232804e-06, "loss": 1.214, "step": 8820 }, { "epoch": 0.8, "learning_rate": 5.0079696222957605e-06, "loss": 1.2283, "step": 8821 }, { "epoch": 0.8, "learning_rate": 5.003666204172428e-06, "loss": 1.2497, "step": 8822 }, { "epoch": 0.8, "learning_rate": 4.999364430216638e-06, "loss": 1.2546, "step": 8823 }, { "epoch": 0.8, "learning_rate": 4.995064300782115e-06, "loss": 1.1988, "step": 8824 }, { "epoch": 0.8, "learning_rate": 4.99076581622242e-06, "loss": 1.2063, "step": 8825 }, { "epoch": 0.8, "learning_rate": 4.986468976890993e-06, "loss": 1.2065, "step": 8826 }, { "epoch": 0.8, "learning_rate": 4.982173783141133e-06, "loss": 1.2211, "step": 8827 }, { "epoch": 0.8, "learning_rate": 4.977880235325999e-06, "loss": 1.2488, "step": 8828 }, { "epoch": 0.8, "learning_rate": 4.973588333798634e-06, "loss": 1.2107, "step": 8829 }, { "epoch": 0.8, "learning_rate": 4.969298078911927e-06, "loss": 1.1641, "step": 8830 }, { "epoch": 0.8, "learning_rate": 4.965009471018639e-06, "loss": 1.1877, "step": 8831 }, { "epoch": 0.8, "learning_rate": 4.960722510471391e-06, "loss": 1.2032, "step": 8832 }, { "epoch": 0.8, "learning_rate": 4.956437197622677e-06, "loss": 1.2246, "step": 8833 }, { "epoch": 0.8, "learning_rate": 4.952153532824835e-06, "loss": 1.2506, "step": 8834 }, { "epoch": 0.8, "learning_rate": 4.947871516430106e-06, "loss": 1.2356, "step": 8835 }, { "epoch": 0.8, "learning_rate": 4.943591148790555e-06, "loss": 1.1546, "step": 8836 }, { "epoch": 0.8, "learning_rate": 4.939312430258139e-06, "loss": 1.2118, "step": 8837 }, { "epoch": 0.8, "learning_rate": 4.935035361184662e-06, "loss": 1.2361, "step": 8838 }, { "epoch": 0.8, "learning_rate": 4.930759941921795e-06, "loss": 1.2011, "step": 8839 }, { "epoch": 0.8, "learning_rate": 4.9264861728210904e-06, "loss": 1.2463, "step": 8840 }, { "epoch": 0.8, "learning_rate": 4.9222140542339476e-06, "loss": 1.1904, "step": 8841 }, { "epoch": 0.8, "learning_rate": 4.917943586511631e-06, "loss": 1.2603, "step": 8842 }, { "epoch": 0.8, "learning_rate": 4.913674770005278e-06, "loss": 1.2351, "step": 8843 }, { "epoch": 0.8, "learning_rate": 4.909407605065874e-06, "loss": 1.1568, "step": 8844 }, { "epoch": 0.8, "learning_rate": 4.905142092044296e-06, "loss": 1.2563, "step": 8845 }, { "epoch": 0.8, "learning_rate": 4.9008782312912644e-06, "loss": 1.1338, "step": 8846 }, { "epoch": 0.8, "learning_rate": 4.8966160231573635e-06, "loss": 1.1984, "step": 8847 }, { "epoch": 0.8, "learning_rate": 4.892355467993051e-06, "loss": 1.2129, "step": 8848 }, { "epoch": 0.8, "learning_rate": 4.888096566148637e-06, "loss": 1.2294, "step": 8849 }, { "epoch": 0.8, "learning_rate": 4.883839317974315e-06, "loss": 1.1728, "step": 8850 }, { "epoch": 0.8, "learning_rate": 4.879583723820125e-06, "loss": 1.2103, "step": 8851 }, { "epoch": 0.8, "learning_rate": 4.875329784035975e-06, "loss": 1.235, "step": 8852 }, { "epoch": 0.8, "learning_rate": 4.8710774989716435e-06, "loss": 1.1353, "step": 8853 }, { "epoch": 0.8, "learning_rate": 4.866826868976762e-06, "loss": 1.2077, "step": 8854 }, { "epoch": 0.8, "learning_rate": 4.862577894400827e-06, "loss": 1.2033, "step": 8855 }, { "epoch": 0.8, "learning_rate": 4.85833057559322e-06, "loss": 1.2002, "step": 8856 }, { "epoch": 0.8, "learning_rate": 4.854084912903162e-06, "loss": 1.271, "step": 8857 }, { "epoch": 0.8, "learning_rate": 4.849840906679745e-06, "loss": 1.2966, "step": 8858 }, { "epoch": 0.8, "learning_rate": 4.845598557271924e-06, "loss": 1.2245, "step": 8859 }, { "epoch": 0.8, "learning_rate": 4.841357865028518e-06, "loss": 1.2056, "step": 8860 }, { "epoch": 0.8, "learning_rate": 4.8371188302982205e-06, "loss": 1.1854, "step": 8861 }, { "epoch": 0.8, "learning_rate": 4.832881453429577e-06, "loss": 1.2578, "step": 8862 }, { "epoch": 0.8, "learning_rate": 4.8286457347709956e-06, "loss": 1.1996, "step": 8863 }, { "epoch": 0.8, "learning_rate": 4.824411674670751e-06, "loss": 1.1624, "step": 8864 }, { "epoch": 0.8, "learning_rate": 4.820179273476979e-06, "loss": 1.2028, "step": 8865 }, { "epoch": 0.8, "learning_rate": 4.815948531537692e-06, "loss": 1.2032, "step": 8866 }, { "epoch": 0.8, "learning_rate": 4.811719449200755e-06, "loss": 1.147, "step": 8867 }, { "epoch": 0.8, "learning_rate": 4.807492026813893e-06, "loss": 1.2163, "step": 8868 }, { "epoch": 0.8, "learning_rate": 4.803266264724698e-06, "loss": 1.1744, "step": 8869 }, { "epoch": 0.8, "learning_rate": 4.799042163280624e-06, "loss": 1.1805, "step": 8870 }, { "epoch": 0.8, "learning_rate": 4.7948197228290026e-06, "loss": 1.1735, "step": 8871 }, { "epoch": 0.8, "learning_rate": 4.790598943717012e-06, "loss": 1.2033, "step": 8872 }, { "epoch": 0.8, "learning_rate": 4.786379826291698e-06, "loss": 1.2829, "step": 8873 }, { "epoch": 0.8, "learning_rate": 4.782162370899973e-06, "loss": 1.2018, "step": 8874 }, { "epoch": 0.8, "learning_rate": 4.7779465778886065e-06, "loss": 1.1907, "step": 8875 }, { "epoch": 0.8, "learning_rate": 4.7737324476042345e-06, "loss": 1.2014, "step": 8876 }, { "epoch": 0.8, "learning_rate": 4.769519980393366e-06, "loss": 1.2225, "step": 8877 }, { "epoch": 0.8, "learning_rate": 4.765309176602359e-06, "loss": 1.1735, "step": 8878 }, { "epoch": 0.8, "learning_rate": 4.7611000365774424e-06, "loss": 1.2591, "step": 8879 }, { "epoch": 0.8, "learning_rate": 4.756892560664703e-06, "loss": 1.1571, "step": 8880 }, { "epoch": 0.8, "learning_rate": 4.752686749210089e-06, "loss": 1.1757, "step": 8881 }, { "epoch": 0.8, "learning_rate": 4.748482602559431e-06, "loss": 1.234, "step": 8882 }, { "epoch": 0.8, "learning_rate": 4.744280121058403e-06, "loss": 1.2212, "step": 8883 }, { "epoch": 0.8, "learning_rate": 4.740079305052542e-06, "loss": 1.2351, "step": 8884 }, { "epoch": 0.8, "learning_rate": 4.73588015488726e-06, "loss": 1.2643, "step": 8885 }, { "epoch": 0.8, "learning_rate": 4.7316826709078135e-06, "loss": 1.2072, "step": 8886 }, { "epoch": 0.8, "learning_rate": 4.727486853459351e-06, "loss": 1.1477, "step": 8887 }, { "epoch": 0.8, "learning_rate": 4.723292702886858e-06, "loss": 1.1959, "step": 8888 }, { "epoch": 0.8, "learning_rate": 4.719100219535194e-06, "loss": 1.2201, "step": 8889 }, { "epoch": 0.8, "learning_rate": 4.714909403749079e-06, "loss": 1.2357, "step": 8890 }, { "epoch": 0.8, "learning_rate": 4.710720255873091e-06, "loss": 1.258, "step": 8891 }, { "epoch": 0.8, "learning_rate": 4.7065327762516874e-06, "loss": 1.2368, "step": 8892 }, { "epoch": 0.8, "learning_rate": 4.7023469652291765e-06, "loss": 1.185, "step": 8893 }, { "epoch": 0.8, "learning_rate": 4.69816282314971e-06, "loss": 1.1823, "step": 8894 }, { "epoch": 0.8, "learning_rate": 4.693980350357347e-06, "loss": 1.2111, "step": 8895 }, { "epoch": 0.8, "learning_rate": 4.689799547195975e-06, "loss": 1.2048, "step": 8896 }, { "epoch": 0.8, "learning_rate": 4.6856204140093455e-06, "loss": 1.2711, "step": 8897 }, { "epoch": 0.8, "learning_rate": 4.681442951141099e-06, "loss": 1.1758, "step": 8898 }, { "epoch": 0.8, "learning_rate": 4.677267158934709e-06, "loss": 1.2811, "step": 8899 }, { "epoch": 0.8, "learning_rate": 4.673093037733528e-06, "loss": 1.2227, "step": 8900 }, { "epoch": 0.8, "learning_rate": 4.668920587880765e-06, "loss": 1.2078, "step": 8901 }, { "epoch": 0.8, "learning_rate": 4.664749809719484e-06, "loss": 1.2447, "step": 8902 }, { "epoch": 0.8, "learning_rate": 4.66058070359264e-06, "loss": 1.2144, "step": 8903 }, { "epoch": 0.8, "learning_rate": 4.656413269843021e-06, "loss": 1.2599, "step": 8904 }, { "epoch": 0.8, "learning_rate": 4.652247508813287e-06, "loss": 1.185, "step": 8905 }, { "epoch": 0.8, "learning_rate": 4.648083420845964e-06, "loss": 1.1539, "step": 8906 }, { "epoch": 0.8, "learning_rate": 4.643921006283427e-06, "loss": 1.2933, "step": 8907 }, { "epoch": 0.8, "learning_rate": 4.639760265467941e-06, "loss": 1.2585, "step": 8908 }, { "epoch": 0.8, "learning_rate": 4.6356011987416075e-06, "loss": 1.2517, "step": 8909 }, { "epoch": 0.81, "learning_rate": 4.631443806446401e-06, "loss": 1.2267, "step": 8910 }, { "epoch": 0.81, "learning_rate": 4.627288088924156e-06, "loss": 1.1585, "step": 8911 }, { "epoch": 0.81, "learning_rate": 4.623134046516564e-06, "loss": 1.1655, "step": 8912 }, { "epoch": 0.81, "learning_rate": 4.618981679565196e-06, "loss": 1.2012, "step": 8913 }, { "epoch": 0.81, "learning_rate": 4.6148309884114755e-06, "loss": 1.2188, "step": 8914 }, { "epoch": 0.81, "learning_rate": 4.610681973396669e-06, "loss": 1.2025, "step": 8915 }, { "epoch": 0.81, "learning_rate": 4.606534634861936e-06, "loss": 1.2005, "step": 8916 }, { "epoch": 0.81, "learning_rate": 4.602388973148286e-06, "loss": 1.2682, "step": 8917 }, { "epoch": 0.81, "learning_rate": 4.5982449885965784e-06, "loss": 1.2004, "step": 8918 }, { "epoch": 0.81, "learning_rate": 4.594102681547557e-06, "loss": 1.2027, "step": 8919 }, { "epoch": 0.81, "learning_rate": 4.589962052341818e-06, "loss": 1.2521, "step": 8920 }, { "epoch": 0.81, "learning_rate": 4.585823101319811e-06, "loss": 1.1798, "step": 8921 }, { "epoch": 0.81, "learning_rate": 4.581685828821858e-06, "loss": 1.2291, "step": 8922 }, { "epoch": 0.81, "learning_rate": 4.577550235188133e-06, "loss": 1.1514, "step": 8923 }, { "epoch": 0.81, "learning_rate": 4.573416320758689e-06, "loss": 1.1804, "step": 8924 }, { "epoch": 0.81, "learning_rate": 4.569284085873429e-06, "loss": 1.2107, "step": 8925 }, { "epoch": 0.81, "learning_rate": 4.565153530872116e-06, "loss": 1.2414, "step": 8926 }, { "epoch": 0.81, "learning_rate": 4.561024656094381e-06, "loss": 1.1874, "step": 8927 }, { "epoch": 0.81, "learning_rate": 4.556897461879706e-06, "loss": 1.2194, "step": 8928 }, { "epoch": 0.81, "learning_rate": 4.5527719485674555e-06, "loss": 1.1876, "step": 8929 }, { "epoch": 0.81, "learning_rate": 4.548648116496837e-06, "loss": 1.1991, "step": 8930 }, { "epoch": 0.81, "learning_rate": 4.544525966006929e-06, "loss": 1.188, "step": 8931 }, { "epoch": 0.81, "learning_rate": 4.540405497436667e-06, "loss": 1.1644, "step": 8932 }, { "epoch": 0.81, "learning_rate": 4.536286711124846e-06, "loss": 1.2124, "step": 8933 }, { "epoch": 0.81, "learning_rate": 4.532169607410133e-06, "loss": 1.196, "step": 8934 }, { "epoch": 0.81, "learning_rate": 4.5280541866310585e-06, "loss": 1.1946, "step": 8935 }, { "epoch": 0.81, "learning_rate": 4.523940449125982e-06, "loss": 1.2524, "step": 8936 }, { "epoch": 0.81, "learning_rate": 4.519828395233172e-06, "loss": 1.2117, "step": 8937 }, { "epoch": 0.81, "learning_rate": 4.515718025290727e-06, "loss": 1.312, "step": 8938 }, { "epoch": 0.81, "learning_rate": 4.511609339636611e-06, "loss": 1.2104, "step": 8939 }, { "epoch": 0.81, "learning_rate": 4.507502338608671e-06, "loss": 1.2541, "step": 8940 }, { "epoch": 0.81, "learning_rate": 4.503397022544576e-06, "loss": 1.1699, "step": 8941 }, { "epoch": 0.81, "learning_rate": 4.4992933917818975e-06, "loss": 1.2035, "step": 8942 }, { "epoch": 0.81, "learning_rate": 4.495191446658042e-06, "loss": 1.2097, "step": 8943 }, { "epoch": 0.81, "learning_rate": 4.491091187510283e-06, "loss": 1.2705, "step": 8944 }, { "epoch": 0.81, "learning_rate": 4.486992614675767e-06, "loss": 1.1649, "step": 8945 }, { "epoch": 0.81, "learning_rate": 4.482895728491487e-06, "loss": 1.2276, "step": 8946 }, { "epoch": 0.81, "learning_rate": 4.478800529294302e-06, "loss": 1.2273, "step": 8947 }, { "epoch": 0.81, "learning_rate": 4.47470701742094e-06, "loss": 1.2007, "step": 8948 }, { "epoch": 0.81, "learning_rate": 4.470615193207969e-06, "loss": 1.2386, "step": 8949 }, { "epoch": 0.81, "learning_rate": 4.4665250569918525e-06, "loss": 1.2263, "step": 8950 }, { "epoch": 0.81, "learning_rate": 4.4624366091088835e-06, "loss": 1.1853, "step": 8951 }, { "epoch": 0.81, "learning_rate": 4.4583498498952325e-06, "loss": 1.1118, "step": 8952 }, { "epoch": 0.81, "learning_rate": 4.454264779686926e-06, "loss": 1.238, "step": 8953 }, { "epoch": 0.81, "learning_rate": 4.450181398819841e-06, "loss": 1.2006, "step": 8954 }, { "epoch": 0.81, "learning_rate": 4.44609970762975e-06, "loss": 1.1959, "step": 8955 }, { "epoch": 0.81, "learning_rate": 4.442019706452255e-06, "loss": 1.21, "step": 8956 }, { "epoch": 0.81, "learning_rate": 4.4379413956228104e-06, "loss": 1.245, "step": 8957 }, { "epoch": 0.81, "learning_rate": 4.4338647754767735e-06, "loss": 1.2161, "step": 8958 }, { "epoch": 0.81, "learning_rate": 4.429789846349325e-06, "loss": 1.1135, "step": 8959 }, { "epoch": 0.81, "learning_rate": 4.425716608575517e-06, "loss": 1.1996, "step": 8960 }, { "epoch": 0.81, "learning_rate": 4.421645062490282e-06, "loss": 1.2064, "step": 8961 }, { "epoch": 0.81, "learning_rate": 4.417575208428373e-06, "loss": 1.1923, "step": 8962 }, { "epoch": 0.81, "learning_rate": 4.413507046724447e-06, "loss": 1.154, "step": 8963 }, { "epoch": 0.81, "learning_rate": 4.409440577712992e-06, "loss": 1.2162, "step": 8964 }, { "epoch": 0.81, "learning_rate": 4.4053758017283645e-06, "loss": 1.2314, "step": 8965 }, { "epoch": 0.81, "learning_rate": 4.401312719104802e-06, "loss": 1.1526, "step": 8966 }, { "epoch": 0.81, "learning_rate": 4.397251330176364e-06, "loss": 1.2109, "step": 8967 }, { "epoch": 0.81, "learning_rate": 4.393191635277003e-06, "loss": 1.2123, "step": 8968 }, { "epoch": 0.81, "learning_rate": 4.389133634740519e-06, "loss": 1.1683, "step": 8969 }, { "epoch": 0.81, "learning_rate": 4.385077328900572e-06, "loss": 1.156, "step": 8970 }, { "epoch": 0.81, "learning_rate": 4.381022718090691e-06, "loss": 1.1312, "step": 8971 }, { "epoch": 0.81, "learning_rate": 4.376969802644259e-06, "loss": 1.2171, "step": 8972 }, { "epoch": 0.81, "learning_rate": 4.372918582894519e-06, "loss": 1.1983, "step": 8973 }, { "epoch": 0.81, "learning_rate": 4.368869059174577e-06, "loss": 1.245, "step": 8974 }, { "epoch": 0.81, "learning_rate": 4.36482123181739e-06, "loss": 1.2002, "step": 8975 }, { "epoch": 0.81, "learning_rate": 4.360775101155798e-06, "loss": 1.1865, "step": 8976 }, { "epoch": 0.81, "learning_rate": 4.35673066752249e-06, "loss": 1.1757, "step": 8977 }, { "epoch": 0.81, "learning_rate": 4.3526879312499935e-06, "loss": 1.1911, "step": 8978 }, { "epoch": 0.81, "learning_rate": 4.348646892670735e-06, "loss": 1.2079, "step": 8979 }, { "epoch": 0.81, "learning_rate": 4.3446075521169746e-06, "loss": 1.234, "step": 8980 }, { "epoch": 0.81, "learning_rate": 4.340569909920836e-06, "loss": 1.2105, "step": 8981 }, { "epoch": 0.81, "learning_rate": 4.3365339664143275e-06, "loss": 1.2586, "step": 8982 }, { "epoch": 0.81, "learning_rate": 4.332499721929273e-06, "loss": 1.1577, "step": 8983 }, { "epoch": 0.81, "learning_rate": 4.328467176797399e-06, "loss": 1.1852, "step": 8984 }, { "epoch": 0.81, "learning_rate": 4.324436331350268e-06, "loss": 1.2178, "step": 8985 }, { "epoch": 0.81, "learning_rate": 4.320407185919309e-06, "loss": 1.1683, "step": 8986 }, { "epoch": 0.81, "learning_rate": 4.316379740835827e-06, "loss": 1.2212, "step": 8987 }, { "epoch": 0.81, "learning_rate": 4.312353996430948e-06, "loss": 1.1992, "step": 8988 }, { "epoch": 0.81, "learning_rate": 4.308329953035703e-06, "loss": 1.2572, "step": 8989 }, { "epoch": 0.81, "learning_rate": 4.304307610980956e-06, "loss": 1.266, "step": 8990 }, { "epoch": 0.81, "learning_rate": 4.300286970597428e-06, "loss": 1.231, "step": 8991 }, { "epoch": 0.81, "learning_rate": 4.296268032215733e-06, "loss": 1.2021, "step": 8992 }, { "epoch": 0.81, "learning_rate": 4.292250796166297e-06, "loss": 1.2458, "step": 8993 }, { "epoch": 0.81, "learning_rate": 4.288235262779447e-06, "loss": 1.1947, "step": 8994 }, { "epoch": 0.81, "learning_rate": 4.284221432385349e-06, "loss": 1.1831, "step": 8995 }, { "epoch": 0.81, "learning_rate": 4.2802093053140285e-06, "loss": 1.2466, "step": 8996 }, { "epoch": 0.81, "learning_rate": 4.276198881895388e-06, "loss": 1.1999, "step": 8997 }, { "epoch": 0.81, "learning_rate": 4.2721901624591774e-06, "loss": 1.1985, "step": 8998 }, { "epoch": 0.81, "learning_rate": 4.268183147334989e-06, "loss": 1.1876, "step": 8999 }, { "epoch": 0.81, "learning_rate": 4.264177836852315e-06, "loss": 1.1908, "step": 9000 }, { "epoch": 0.81, "learning_rate": 4.260174231340477e-06, "loss": 1.2216, "step": 9001 }, { "epoch": 0.81, "learning_rate": 4.256172331128661e-06, "loss": 1.2296, "step": 9002 }, { "epoch": 0.81, "learning_rate": 4.252172136545932e-06, "loss": 1.2134, "step": 9003 }, { "epoch": 0.81, "learning_rate": 4.248173647921177e-06, "loss": 1.1963, "step": 9004 }, { "epoch": 0.81, "learning_rate": 4.244176865583186e-06, "loss": 1.2077, "step": 9005 }, { "epoch": 0.81, "learning_rate": 4.240181789860578e-06, "loss": 1.2145, "step": 9006 }, { "epoch": 0.81, "learning_rate": 4.236188421081841e-06, "loss": 1.2418, "step": 9007 }, { "epoch": 0.81, "learning_rate": 4.232196759575338e-06, "loss": 1.1741, "step": 9008 }, { "epoch": 0.81, "learning_rate": 4.228206805669255e-06, "loss": 1.224, "step": 9009 }, { "epoch": 0.81, "learning_rate": 4.2242185596916755e-06, "loss": 1.1783, "step": 9010 }, { "epoch": 0.81, "learning_rate": 4.220232021970524e-06, "loss": 1.2439, "step": 9011 }, { "epoch": 0.81, "learning_rate": 4.2162471928335775e-06, "loss": 1.212, "step": 9012 }, { "epoch": 0.81, "learning_rate": 4.212264072608505e-06, "loss": 1.2326, "step": 9013 }, { "epoch": 0.81, "learning_rate": 4.208282661622784e-06, "loss": 1.1869, "step": 9014 }, { "epoch": 0.81, "learning_rate": 4.204302960203799e-06, "loss": 1.219, "step": 9015 }, { "epoch": 0.81, "learning_rate": 4.20032496867877e-06, "loss": 1.162, "step": 9016 }, { "epoch": 0.81, "learning_rate": 4.196348687374774e-06, "loss": 1.194, "step": 9017 }, { "epoch": 0.81, "learning_rate": 4.192374116618775e-06, "loss": 1.1909, "step": 9018 }, { "epoch": 0.81, "learning_rate": 4.188401256737556e-06, "loss": 1.1883, "step": 9019 }, { "epoch": 0.81, "learning_rate": 4.18443010805778e-06, "loss": 1.1472, "step": 9020 }, { "epoch": 0.82, "learning_rate": 4.180460670905978e-06, "loss": 1.1814, "step": 9021 }, { "epoch": 0.82, "learning_rate": 4.176492945608526e-06, "loss": 1.2046, "step": 9022 }, { "epoch": 0.82, "learning_rate": 4.172526932491663e-06, "loss": 1.1829, "step": 9023 }, { "epoch": 0.82, "learning_rate": 4.168562631881498e-06, "loss": 1.2019, "step": 9024 }, { "epoch": 0.82, "learning_rate": 4.164600044103972e-06, "loss": 1.2416, "step": 9025 }, { "epoch": 0.82, "learning_rate": 4.160639169484918e-06, "loss": 1.1986, "step": 9026 }, { "epoch": 0.82, "learning_rate": 4.156680008350006e-06, "loss": 1.1822, "step": 9027 }, { "epoch": 0.82, "learning_rate": 4.1527225610247665e-06, "loss": 1.1756, "step": 9028 }, { "epoch": 0.82, "learning_rate": 4.1487668278346155e-06, "loss": 1.1705, "step": 9029 }, { "epoch": 0.82, "learning_rate": 4.144812809104779e-06, "loss": 1.1578, "step": 9030 }, { "epoch": 0.82, "learning_rate": 4.140860505160388e-06, "loss": 1.2436, "step": 9031 }, { "epoch": 0.82, "learning_rate": 4.136909916326412e-06, "loss": 1.1882, "step": 9032 }, { "epoch": 0.82, "learning_rate": 4.132961042927671e-06, "loss": 1.2554, "step": 9033 }, { "epoch": 0.82, "learning_rate": 4.129013885288879e-06, "loss": 1.2064, "step": 9034 }, { "epoch": 0.82, "learning_rate": 4.1250684437345585e-06, "loss": 1.186, "step": 9035 }, { "epoch": 0.82, "learning_rate": 4.121124718589134e-06, "loss": 1.2274, "step": 9036 }, { "epoch": 0.82, "learning_rate": 4.117182710176868e-06, "loss": 1.2222, "step": 9037 }, { "epoch": 0.82, "learning_rate": 4.1132424188218746e-06, "loss": 1.1958, "step": 9038 }, { "epoch": 0.82, "learning_rate": 4.109303844848164e-06, "loss": 1.2312, "step": 9039 }, { "epoch": 0.82, "learning_rate": 4.105366988579559e-06, "loss": 1.1965, "step": 9040 }, { "epoch": 0.82, "learning_rate": 4.101431850339757e-06, "loss": 1.1739, "step": 9041 }, { "epoch": 0.82, "learning_rate": 4.097498430452337e-06, "loss": 1.2136, "step": 9042 }, { "epoch": 0.82, "learning_rate": 4.093566729240708e-06, "loss": 1.1991, "step": 9043 }, { "epoch": 0.82, "learning_rate": 4.08963674702815e-06, "loss": 1.185, "step": 9044 }, { "epoch": 0.82, "learning_rate": 4.085708484137798e-06, "loss": 1.1799, "step": 9045 }, { "epoch": 0.82, "learning_rate": 4.081781940892645e-06, "loss": 1.2551, "step": 9046 }, { "epoch": 0.82, "learning_rate": 4.077857117615552e-06, "loss": 1.1849, "step": 9047 }, { "epoch": 0.82, "learning_rate": 4.0739340146292294e-06, "loss": 1.1675, "step": 9048 }, { "epoch": 0.82, "learning_rate": 4.07001263225624e-06, "loss": 1.2034, "step": 9049 }, { "epoch": 0.82, "learning_rate": 4.066092970819033e-06, "loss": 1.1701, "step": 9050 }, { "epoch": 0.82, "learning_rate": 4.06217503063987e-06, "loss": 1.1715, "step": 9051 }, { "epoch": 0.82, "learning_rate": 4.058258812040918e-06, "loss": 1.1985, "step": 9052 }, { "epoch": 0.82, "learning_rate": 4.054344315344178e-06, "loss": 1.231, "step": 9053 }, { "epoch": 0.82, "learning_rate": 4.050431540871499e-06, "loss": 1.2096, "step": 9054 }, { "epoch": 0.82, "learning_rate": 4.04652048894463e-06, "loss": 1.1977, "step": 9055 }, { "epoch": 0.82, "learning_rate": 4.042611159885124e-06, "loss": 1.2048, "step": 9056 }, { "epoch": 0.82, "learning_rate": 4.038703554014436e-06, "loss": 1.2229, "step": 9057 }, { "epoch": 0.82, "learning_rate": 4.034797671653859e-06, "loss": 1.2418, "step": 9058 }, { "epoch": 0.82, "learning_rate": 4.0308935131245396e-06, "loss": 1.1921, "step": 9059 }, { "epoch": 0.82, "learning_rate": 4.0269910787475106e-06, "loss": 1.1837, "step": 9060 }, { "epoch": 0.82, "learning_rate": 4.0230903688436265e-06, "loss": 1.182, "step": 9061 }, { "epoch": 0.82, "learning_rate": 4.019191383733617e-06, "loss": 1.1905, "step": 9062 }, { "epoch": 0.82, "learning_rate": 4.015294123738081e-06, "loss": 1.211, "step": 9063 }, { "epoch": 0.82, "learning_rate": 4.0113985891774594e-06, "loss": 1.218, "step": 9064 }, { "epoch": 0.82, "learning_rate": 4.007504780372057e-06, "loss": 1.2461, "step": 9065 }, { "epoch": 0.82, "learning_rate": 4.003612697642034e-06, "loss": 1.1738, "step": 9066 }, { "epoch": 0.82, "learning_rate": 3.999722341307405e-06, "loss": 1.1844, "step": 9067 }, { "epoch": 0.82, "learning_rate": 3.9958337116880626e-06, "loss": 1.1136, "step": 9068 }, { "epoch": 0.82, "learning_rate": 3.991946809103738e-06, "loss": 1.1728, "step": 9069 }, { "epoch": 0.82, "learning_rate": 3.988061633874019e-06, "loss": 1.1993, "step": 9070 }, { "epoch": 0.82, "learning_rate": 3.9841781863183745e-06, "loss": 1.1718, "step": 9071 }, { "epoch": 0.82, "learning_rate": 3.980296466756089e-06, "loss": 1.1649, "step": 9072 }, { "epoch": 0.82, "learning_rate": 3.976416475506353e-06, "loss": 1.2483, "step": 9073 }, { "epoch": 0.82, "learning_rate": 3.972538212888188e-06, "loss": 1.2133, "step": 9074 }, { "epoch": 0.82, "learning_rate": 3.968661679220468e-06, "loss": 1.1365, "step": 9075 }, { "epoch": 0.82, "learning_rate": 3.964786874821955e-06, "loss": 1.2206, "step": 9076 }, { "epoch": 0.82, "learning_rate": 3.960913800011224e-06, "loss": 1.1594, "step": 9077 }, { "epoch": 0.82, "learning_rate": 3.957042455106749e-06, "loss": 1.1587, "step": 9078 }, { "epoch": 0.82, "learning_rate": 3.953172840426844e-06, "loss": 1.2353, "step": 9079 }, { "epoch": 0.82, "learning_rate": 3.949304956289674e-06, "loss": 1.2966, "step": 9080 }, { "epoch": 0.82, "learning_rate": 3.945438803013285e-06, "loss": 1.2057, "step": 9081 }, { "epoch": 0.82, "learning_rate": 3.94157438091555e-06, "loss": 1.1969, "step": 9082 }, { "epoch": 0.82, "learning_rate": 3.937711690314214e-06, "loss": 1.1567, "step": 9083 }, { "epoch": 0.82, "learning_rate": 3.933850731526895e-06, "loss": 1.1795, "step": 9084 }, { "epoch": 0.82, "learning_rate": 3.929991504871047e-06, "loss": 1.1999, "step": 9085 }, { "epoch": 0.82, "learning_rate": 3.926134010663987e-06, "loss": 1.202, "step": 9086 }, { "epoch": 0.82, "learning_rate": 3.922278249222894e-06, "loss": 1.2098, "step": 9087 }, { "epoch": 0.82, "learning_rate": 3.918424220864797e-06, "loss": 1.2506, "step": 9088 }, { "epoch": 0.82, "learning_rate": 3.914571925906596e-06, "loss": 1.2314, "step": 9089 }, { "epoch": 0.82, "learning_rate": 3.910721364665035e-06, "loss": 1.2479, "step": 9090 }, { "epoch": 0.82, "learning_rate": 3.9068725374567215e-06, "loss": 1.1366, "step": 9091 }, { "epoch": 0.82, "learning_rate": 3.903025444598119e-06, "loss": 1.2464, "step": 9092 }, { "epoch": 0.82, "learning_rate": 3.899180086405543e-06, "loss": 1.1933, "step": 9093 }, { "epoch": 0.82, "learning_rate": 3.895336463195185e-06, "loss": 1.1369, "step": 9094 }, { "epoch": 0.82, "learning_rate": 3.891494575283072e-06, "loss": 1.1978, "step": 9095 }, { "epoch": 0.82, "learning_rate": 3.8876544229850935e-06, "loss": 1.2962, "step": 9096 }, { "epoch": 0.82, "learning_rate": 3.883816006617019e-06, "loss": 1.1941, "step": 9097 }, { "epoch": 0.82, "learning_rate": 3.879979326494429e-06, "loss": 1.1687, "step": 9098 }, { "epoch": 0.82, "learning_rate": 3.8761443829328114e-06, "loss": 1.1986, "step": 9099 }, { "epoch": 0.82, "learning_rate": 3.872311176247476e-06, "loss": 1.2104, "step": 9100 }, { "epoch": 0.82, "learning_rate": 3.868479706753603e-06, "loss": 1.1507, "step": 9101 }, { "epoch": 0.82, "learning_rate": 3.864649974766243e-06, "loss": 1.1846, "step": 9102 }, { "epoch": 0.82, "learning_rate": 3.860821980600271e-06, "loss": 1.235, "step": 9103 }, { "epoch": 0.82, "learning_rate": 3.856995724570442e-06, "loss": 1.2268, "step": 9104 }, { "epoch": 0.82, "learning_rate": 3.853171206991374e-06, "loss": 1.225, "step": 9105 }, { "epoch": 0.82, "learning_rate": 3.849348428177524e-06, "loss": 1.2152, "step": 9106 }, { "epoch": 0.82, "learning_rate": 3.845527388443216e-06, "loss": 1.1792, "step": 9107 }, { "epoch": 0.82, "learning_rate": 3.8417080881026304e-06, "loss": 1.2378, "step": 9108 }, { "epoch": 0.82, "learning_rate": 3.837890527469793e-06, "loss": 1.2534, "step": 9109 }, { "epoch": 0.82, "learning_rate": 3.834074706858612e-06, "loss": 1.1645, "step": 9110 }, { "epoch": 0.82, "learning_rate": 3.830260626582832e-06, "loss": 1.2181, "step": 9111 }, { "epoch": 0.82, "learning_rate": 3.826448286956058e-06, "loss": 1.2222, "step": 9112 }, { "epoch": 0.82, "learning_rate": 3.822637688291755e-06, "loss": 1.2074, "step": 9113 }, { "epoch": 0.82, "learning_rate": 3.818828830903235e-06, "loss": 1.2207, "step": 9114 }, { "epoch": 0.82, "learning_rate": 3.815021715103692e-06, "loss": 1.2146, "step": 9115 }, { "epoch": 0.82, "learning_rate": 3.8112163412061504e-06, "loss": 1.2876, "step": 9116 }, { "epoch": 0.82, "learning_rate": 3.8074127095235002e-06, "loss": 1.1946, "step": 9117 }, { "epoch": 0.82, "learning_rate": 3.8036108203684947e-06, "loss": 1.1959, "step": 9118 }, { "epoch": 0.82, "learning_rate": 3.7998106740537248e-06, "loss": 1.1835, "step": 9119 }, { "epoch": 0.82, "learning_rate": 3.7960122708916716e-06, "loss": 1.1423, "step": 9120 }, { "epoch": 0.82, "learning_rate": 3.7922156111946413e-06, "loss": 1.2198, "step": 9121 }, { "epoch": 0.82, "learning_rate": 3.7884206952748023e-06, "loss": 1.1601, "step": 9122 }, { "epoch": 0.82, "learning_rate": 3.7846275234442092e-06, "loss": 1.2253, "step": 9123 }, { "epoch": 0.82, "learning_rate": 3.7808360960147256e-06, "loss": 1.2091, "step": 9124 }, { "epoch": 0.82, "learning_rate": 3.7770464132980964e-06, "loss": 1.2675, "step": 9125 }, { "epoch": 0.82, "learning_rate": 3.7732584756059396e-06, "loss": 1.2305, "step": 9126 }, { "epoch": 0.82, "learning_rate": 3.7694722832497002e-06, "loss": 1.1596, "step": 9127 }, { "epoch": 0.82, "learning_rate": 3.765687836540696e-06, "loss": 1.2246, "step": 9128 }, { "epoch": 0.82, "learning_rate": 3.7619051357900946e-06, "loss": 1.1524, "step": 9129 }, { "epoch": 0.82, "learning_rate": 3.7581241813089197e-06, "loss": 1.2188, "step": 9130 }, { "epoch": 0.83, "learning_rate": 3.754344973408064e-06, "loss": 1.2219, "step": 9131 }, { "epoch": 0.83, "learning_rate": 3.750567512398262e-06, "loss": 1.1914, "step": 9132 }, { "epoch": 0.83, "learning_rate": 3.74679179859011e-06, "loss": 1.27, "step": 9133 }, { "epoch": 0.83, "learning_rate": 3.74301783229406e-06, "loss": 1.2168, "step": 9134 }, { "epoch": 0.83, "learning_rate": 3.7392456138204136e-06, "loss": 1.2306, "step": 9135 }, { "epoch": 0.83, "learning_rate": 3.735475143479347e-06, "loss": 1.227, "step": 9136 }, { "epoch": 0.83, "learning_rate": 3.731706421580877e-06, "loss": 1.2297, "step": 9137 }, { "epoch": 0.83, "learning_rate": 3.7279394484348806e-06, "loss": 1.2162, "step": 9138 }, { "epoch": 0.83, "learning_rate": 3.72417422435109e-06, "loss": 1.2286, "step": 9139 }, { "epoch": 0.83, "learning_rate": 3.7204107496390918e-06, "loss": 1.1744, "step": 9140 }, { "epoch": 0.83, "learning_rate": 3.716649024608343e-06, "loss": 1.1999, "step": 9141 }, { "epoch": 0.83, "learning_rate": 3.7128890495681355e-06, "loss": 1.2269, "step": 9142 }, { "epoch": 0.83, "learning_rate": 3.7091308248276336e-06, "loss": 1.2192, "step": 9143 }, { "epoch": 0.83, "learning_rate": 3.705374350695845e-06, "loss": 1.2485, "step": 9144 }, { "epoch": 0.83, "learning_rate": 3.701619627481645e-06, "loss": 1.2406, "step": 9145 }, { "epoch": 0.83, "learning_rate": 3.6978666554937533e-06, "loss": 1.2303, "step": 9146 }, { "epoch": 0.83, "learning_rate": 3.6941154350407627e-06, "loss": 1.2036, "step": 9147 }, { "epoch": 0.83, "learning_rate": 3.6903659664311087e-06, "loss": 1.1936, "step": 9148 }, { "epoch": 0.83, "learning_rate": 3.6866182499730782e-06, "loss": 1.2374, "step": 9149 }, { "epoch": 0.83, "learning_rate": 3.6828722859748283e-06, "loss": 1.2262, "step": 9150 }, { "epoch": 0.83, "learning_rate": 3.679128074744359e-06, "loss": 1.244, "step": 9151 }, { "epoch": 0.83, "learning_rate": 3.6753856165895382e-06, "loss": 1.2268, "step": 9152 }, { "epoch": 0.83, "learning_rate": 3.671644911818084e-06, "loss": 1.2286, "step": 9153 }, { "epoch": 0.83, "learning_rate": 3.6679059607375667e-06, "loss": 1.1673, "step": 9154 }, { "epoch": 0.83, "learning_rate": 3.664168763655418e-06, "loss": 1.1794, "step": 9155 }, { "epoch": 0.83, "learning_rate": 3.6604333208789143e-06, "loss": 1.2083, "step": 9156 }, { "epoch": 0.83, "learning_rate": 3.656699632715213e-06, "loss": 1.206, "step": 9157 }, { "epoch": 0.83, "learning_rate": 3.652967699471299e-06, "loss": 1.2229, "step": 9158 }, { "epoch": 0.83, "learning_rate": 3.6492375214540298e-06, "loss": 1.2711, "step": 9159 }, { "epoch": 0.83, "learning_rate": 3.6455090989701096e-06, "loss": 1.2424, "step": 9160 }, { "epoch": 0.83, "learning_rate": 3.641782432326099e-06, "loss": 1.243, "step": 9161 }, { "epoch": 0.83, "learning_rate": 3.6380575218284278e-06, "loss": 1.2189, "step": 9162 }, { "epoch": 0.83, "learning_rate": 3.634334367783365e-06, "loss": 1.2601, "step": 9163 }, { "epoch": 0.83, "learning_rate": 3.630612970497044e-06, "loss": 1.2198, "step": 9164 }, { "epoch": 0.83, "learning_rate": 3.6268933302754463e-06, "loss": 1.1554, "step": 9165 }, { "epoch": 0.83, "learning_rate": 3.623175447424418e-06, "loss": 1.2225, "step": 9166 }, { "epoch": 0.83, "learning_rate": 3.6194593222496463e-06, "loss": 1.1928, "step": 9167 }, { "epoch": 0.83, "learning_rate": 3.615744955056699e-06, "loss": 1.1686, "step": 9168 }, { "epoch": 0.83, "learning_rate": 3.6120323461509757e-06, "loss": 1.2275, "step": 9169 }, { "epoch": 0.83, "learning_rate": 3.6083214958377435e-06, "loss": 1.1586, "step": 9170 }, { "epoch": 0.83, "learning_rate": 3.604612404422117e-06, "loss": 1.1703, "step": 9171 }, { "epoch": 0.83, "learning_rate": 3.600905072209068e-06, "loss": 1.2305, "step": 9172 }, { "epoch": 0.83, "learning_rate": 3.597199499503434e-06, "loss": 1.2286, "step": 9173 }, { "epoch": 0.83, "learning_rate": 3.593495686609899e-06, "loss": 1.1573, "step": 9174 }, { "epoch": 0.83, "learning_rate": 3.5897936338329997e-06, "loss": 1.2494, "step": 9175 }, { "epoch": 0.83, "learning_rate": 3.5860933414771313e-06, "loss": 1.262, "step": 9176 }, { "epoch": 0.83, "learning_rate": 3.5823948098465424e-06, "loss": 1.1941, "step": 9177 }, { "epoch": 0.83, "learning_rate": 3.5786980392453485e-06, "loss": 1.2086, "step": 9178 }, { "epoch": 0.83, "learning_rate": 3.5750030299775026e-06, "loss": 1.199, "step": 9179 }, { "epoch": 0.83, "learning_rate": 3.5713097823468267e-06, "loss": 1.2081, "step": 9180 }, { "epoch": 0.83, "learning_rate": 3.5676182966569882e-06, "loss": 1.2148, "step": 9181 }, { "epoch": 0.83, "learning_rate": 3.5639285732115084e-06, "loss": 1.1592, "step": 9182 }, { "epoch": 0.83, "learning_rate": 3.560240612313781e-06, "loss": 1.228, "step": 9183 }, { "epoch": 0.83, "learning_rate": 3.556554414267038e-06, "loss": 1.1438, "step": 9184 }, { "epoch": 0.83, "learning_rate": 3.5528699793743707e-06, "loss": 1.2698, "step": 9185 }, { "epoch": 0.83, "learning_rate": 3.5491873079387256e-06, "loss": 1.2653, "step": 9186 }, { "epoch": 0.83, "learning_rate": 3.5455064002629047e-06, "loss": 1.1895, "step": 9187 }, { "epoch": 0.83, "learning_rate": 3.5418272566495602e-06, "loss": 1.1445, "step": 9188 }, { "epoch": 0.83, "learning_rate": 3.5381498774012146e-06, "loss": 1.2049, "step": 9189 }, { "epoch": 0.83, "learning_rate": 3.534474262820231e-06, "loss": 1.1761, "step": 9190 }, { "epoch": 0.83, "learning_rate": 3.530800413208826e-06, "loss": 1.2293, "step": 9191 }, { "epoch": 0.83, "learning_rate": 3.527128328869081e-06, "loss": 1.2037, "step": 9192 }, { "epoch": 0.83, "learning_rate": 3.52345801010292e-06, "loss": 1.1631, "step": 9193 }, { "epoch": 0.83, "learning_rate": 3.5197894572121414e-06, "loss": 1.1404, "step": 9194 }, { "epoch": 0.83, "learning_rate": 3.5161226704983814e-06, "loss": 1.2092, "step": 9195 }, { "epoch": 0.83, "learning_rate": 3.5124576502631318e-06, "loss": 1.2184, "step": 9196 }, { "epoch": 0.83, "learning_rate": 3.508794396807749e-06, "loss": 1.2617, "step": 9197 }, { "epoch": 0.83, "learning_rate": 3.5051329104334258e-06, "loss": 1.2072, "step": 9198 }, { "epoch": 0.83, "learning_rate": 3.5014731914412403e-06, "loss": 1.1934, "step": 9199 }, { "epoch": 0.83, "learning_rate": 3.497815240132099e-06, "loss": 1.2098, "step": 9200 }, { "epoch": 0.83, "learning_rate": 3.49415905680677e-06, "loss": 1.2154, "step": 9201 }, { "epoch": 0.83, "learning_rate": 3.4905046417658767e-06, "loss": 1.2194, "step": 9202 }, { "epoch": 0.83, "learning_rate": 3.486851995309895e-06, "loss": 1.2358, "step": 9203 }, { "epoch": 0.83, "learning_rate": 3.483201117739168e-06, "loss": 1.2058, "step": 9204 }, { "epoch": 0.83, "learning_rate": 3.479552009353876e-06, "loss": 1.2317, "step": 9205 }, { "epoch": 0.83, "learning_rate": 3.4759046704540667e-06, "loss": 1.2079, "step": 9206 }, { "epoch": 0.83, "learning_rate": 3.472259101339631e-06, "loss": 1.1636, "step": 9207 }, { "epoch": 0.83, "learning_rate": 3.4686153023103215e-06, "loss": 1.1847, "step": 9208 }, { "epoch": 0.83, "learning_rate": 3.4649732736657393e-06, "loss": 1.1938, "step": 9209 }, { "epoch": 0.83, "learning_rate": 3.4613330157053593e-06, "loss": 1.1777, "step": 9210 }, { "epoch": 0.83, "learning_rate": 3.4576945287284835e-06, "loss": 1.1992, "step": 9211 }, { "epoch": 0.83, "learning_rate": 3.454057813034284e-06, "loss": 1.2131, "step": 9212 }, { "epoch": 0.83, "learning_rate": 3.450422868921785e-06, "loss": 1.2163, "step": 9213 }, { "epoch": 0.83, "learning_rate": 3.446789696689859e-06, "loss": 1.2022, "step": 9214 }, { "epoch": 0.83, "learning_rate": 3.4431582966372473e-06, "loss": 1.18, "step": 9215 }, { "epoch": 0.83, "learning_rate": 3.4395286690625305e-06, "loss": 1.1908, "step": 9216 }, { "epoch": 0.83, "learning_rate": 3.435900814264151e-06, "loss": 1.1708, "step": 9217 }, { "epoch": 0.83, "learning_rate": 3.4322747325404028e-06, "loss": 1.1731, "step": 9218 }, { "epoch": 0.83, "learning_rate": 3.428650424189428e-06, "loss": 1.1701, "step": 9219 }, { "epoch": 0.83, "learning_rate": 3.4250278895092414e-06, "loss": 1.2273, "step": 9220 }, { "epoch": 0.83, "learning_rate": 3.4214071287976962e-06, "loss": 1.2108, "step": 9221 }, { "epoch": 0.83, "learning_rate": 3.4177881423525018e-06, "loss": 1.2122, "step": 9222 }, { "epoch": 0.83, "learning_rate": 3.414170930471225e-06, "loss": 1.1932, "step": 9223 }, { "epoch": 0.83, "learning_rate": 3.4105554934512786e-06, "loss": 1.2274, "step": 9224 }, { "epoch": 0.83, "learning_rate": 3.4069418315899493e-06, "loss": 1.166, "step": 9225 }, { "epoch": 0.83, "learning_rate": 3.4033299451843607e-06, "loss": 1.147, "step": 9226 }, { "epoch": 0.83, "learning_rate": 3.3997198345314894e-06, "loss": 1.1647, "step": 9227 }, { "epoch": 0.83, "learning_rate": 3.3961114999281756e-06, "loss": 1.1694, "step": 9228 }, { "epoch": 0.83, "learning_rate": 3.392504941671104e-06, "loss": 1.2051, "step": 9229 }, { "epoch": 0.83, "learning_rate": 3.388900160056821e-06, "loss": 1.1864, "step": 9230 }, { "epoch": 0.83, "learning_rate": 3.385297155381728e-06, "loss": 1.1785, "step": 9231 }, { "epoch": 0.83, "learning_rate": 3.3816959279420716e-06, "loss": 1.2421, "step": 9232 }, { "epoch": 0.83, "learning_rate": 3.3780964780339622e-06, "loss": 1.2058, "step": 9233 }, { "epoch": 0.83, "learning_rate": 3.3744988059533524e-06, "loss": 1.1833, "step": 9234 }, { "epoch": 0.83, "learning_rate": 3.3709029119960543e-06, "loss": 1.1993, "step": 9235 }, { "epoch": 0.83, "learning_rate": 3.3673087964577464e-06, "loss": 1.1967, "step": 9236 }, { "epoch": 0.83, "learning_rate": 3.3637164596339395e-06, "loss": 1.2123, "step": 9237 }, { "epoch": 0.83, "learning_rate": 3.3601259018200105e-06, "loss": 1.2018, "step": 9238 }, { "epoch": 0.83, "learning_rate": 3.3565371233111877e-06, "loss": 1.1932, "step": 9239 }, { "epoch": 0.83, "learning_rate": 3.352950124402546e-06, "loss": 1.1764, "step": 9240 }, { "epoch": 0.83, "learning_rate": 3.3493649053890326e-06, "loss": 1.2167, "step": 9241 }, { "epoch": 0.84, "learning_rate": 3.345781466565437e-06, "loss": 1.1918, "step": 9242 }, { "epoch": 0.84, "learning_rate": 3.342199808226387e-06, "loss": 1.2157, "step": 9243 }, { "epoch": 0.84, "learning_rate": 3.3386199306663894e-06, "loss": 1.2357, "step": 9244 }, { "epoch": 0.84, "learning_rate": 3.335041834179789e-06, "loss": 1.2368, "step": 9245 }, { "epoch": 0.84, "learning_rate": 3.3314655190608e-06, "loss": 1.1789, "step": 9246 }, { "epoch": 0.84, "learning_rate": 3.3278909856034713e-06, "loss": 1.2109, "step": 9247 }, { "epoch": 0.84, "learning_rate": 3.324318234101714e-06, "loss": 1.2063, "step": 9248 }, { "epoch": 0.84, "learning_rate": 3.3207472648492917e-06, "loss": 1.2057, "step": 9249 }, { "epoch": 0.84, "learning_rate": 3.317178078139821e-06, "loss": 1.2138, "step": 9250 }, { "epoch": 0.84, "learning_rate": 3.3136106742667704e-06, "loss": 1.2102, "step": 9251 }, { "epoch": 0.84, "learning_rate": 3.310045053523475e-06, "loss": 1.1958, "step": 9252 }, { "epoch": 0.84, "learning_rate": 3.306481216203103e-06, "loss": 1.1547, "step": 9253 }, { "epoch": 0.84, "learning_rate": 3.302919162598689e-06, "loss": 1.2324, "step": 9254 }, { "epoch": 0.84, "learning_rate": 3.2993588930031154e-06, "loss": 1.1674, "step": 9255 }, { "epoch": 0.84, "learning_rate": 3.2958004077091154e-06, "loss": 1.2789, "step": 9256 }, { "epoch": 0.84, "learning_rate": 3.2922437070092905e-06, "loss": 1.1943, "step": 9257 }, { "epoch": 0.84, "learning_rate": 3.288688791196079e-06, "loss": 1.1676, "step": 9258 }, { "epoch": 0.84, "learning_rate": 3.2851356605617805e-06, "loss": 1.2282, "step": 9259 }, { "epoch": 0.84, "learning_rate": 3.281584315398542e-06, "loss": 1.2541, "step": 9260 }, { "epoch": 0.84, "learning_rate": 3.278034755998363e-06, "loss": 1.1644, "step": 9261 }, { "epoch": 0.84, "learning_rate": 3.274486982653116e-06, "loss": 1.1823, "step": 9262 }, { "epoch": 0.84, "learning_rate": 3.2709409956545062e-06, "loss": 1.2016, "step": 9263 }, { "epoch": 0.84, "learning_rate": 3.267396795294081e-06, "loss": 1.19, "step": 9264 }, { "epoch": 0.84, "learning_rate": 3.2638543818632737e-06, "loss": 1.2011, "step": 9265 }, { "epoch": 0.84, "learning_rate": 3.2603137556533436e-06, "loss": 1.2291, "step": 9266 }, { "epoch": 0.84, "learning_rate": 3.2567749169554214e-06, "loss": 1.2348, "step": 9267 }, { "epoch": 0.84, "learning_rate": 3.253237866060488e-06, "loss": 1.1798, "step": 9268 }, { "epoch": 0.84, "learning_rate": 3.2497026032593524e-06, "loss": 1.1932, "step": 9269 }, { "epoch": 0.84, "learning_rate": 3.2461691288427103e-06, "loss": 1.1681, "step": 9270 }, { "epoch": 0.84, "learning_rate": 3.242637443101093e-06, "loss": 1.1907, "step": 9271 }, { "epoch": 0.84, "learning_rate": 3.239107546324885e-06, "loss": 1.2352, "step": 9272 }, { "epoch": 0.84, "learning_rate": 3.2355794388043316e-06, "loss": 1.1961, "step": 9273 }, { "epoch": 0.84, "learning_rate": 3.2320531208295258e-06, "loss": 1.1728, "step": 9274 }, { "epoch": 0.84, "learning_rate": 3.2285285926904107e-06, "loss": 1.1522, "step": 9275 }, { "epoch": 0.84, "learning_rate": 3.225005854676785e-06, "loss": 1.1525, "step": 9276 }, { "epoch": 0.84, "learning_rate": 3.221484907078298e-06, "loss": 1.1549, "step": 9277 }, { "epoch": 0.84, "learning_rate": 3.217965750184462e-06, "loss": 1.1835, "step": 9278 }, { "epoch": 0.84, "learning_rate": 3.2144483842846296e-06, "loss": 1.2196, "step": 9279 }, { "epoch": 0.84, "learning_rate": 3.2109328096680104e-06, "loss": 1.2654, "step": 9280 }, { "epoch": 0.84, "learning_rate": 3.2074190266236705e-06, "loss": 1.2486, "step": 9281 }, { "epoch": 0.84, "learning_rate": 3.203907035440515e-06, "loss": 1.2245, "step": 9282 }, { "epoch": 0.84, "learning_rate": 3.200396836407324e-06, "loss": 1.1963, "step": 9283 }, { "epoch": 0.84, "learning_rate": 3.1968884298127193e-06, "loss": 1.213, "step": 9284 }, { "epoch": 0.84, "learning_rate": 3.1933818159451566e-06, "loss": 1.1788, "step": 9285 }, { "epoch": 0.84, "learning_rate": 3.1898769950929803e-06, "loss": 1.2736, "step": 9286 }, { "epoch": 0.84, "learning_rate": 3.186373967544362e-06, "loss": 1.215, "step": 9287 }, { "epoch": 0.84, "learning_rate": 3.1828727335873277e-06, "loss": 1.1903, "step": 9288 }, { "epoch": 0.84, "learning_rate": 3.179373293509774e-06, "loss": 1.1736, "step": 9289 }, { "epoch": 0.84, "learning_rate": 3.1758756475994217e-06, "loss": 1.2405, "step": 9290 }, { "epoch": 0.84, "learning_rate": 3.1723797961438705e-06, "loss": 1.2503, "step": 9291 }, { "epoch": 0.84, "learning_rate": 3.1688857394305577e-06, "loss": 1.1699, "step": 9292 }, { "epoch": 0.84, "learning_rate": 3.1653934777467724e-06, "loss": 1.2303, "step": 9293 }, { "epoch": 0.84, "learning_rate": 3.1619030113796742e-06, "loss": 1.1853, "step": 9294 }, { "epoch": 0.84, "learning_rate": 3.1584143406162394e-06, "loss": 1.2116, "step": 9295 }, { "epoch": 0.84, "learning_rate": 3.1549274657433375e-06, "loss": 1.2334, "step": 9296 }, { "epoch": 0.84, "learning_rate": 3.1514423870476652e-06, "loss": 1.2067, "step": 9297 }, { "epoch": 0.84, "learning_rate": 3.1479591048157703e-06, "loss": 1.2484, "step": 9298 }, { "epoch": 0.84, "learning_rate": 3.1444776193340714e-06, "loss": 1.1358, "step": 9299 }, { "epoch": 0.84, "learning_rate": 3.140997930888828e-06, "loss": 1.1937, "step": 9300 }, { "epoch": 0.84, "learning_rate": 3.137520039766145e-06, "loss": 1.2655, "step": 9301 }, { "epoch": 0.84, "learning_rate": 3.1340439462519884e-06, "loss": 1.1924, "step": 9302 }, { "epoch": 0.84, "learning_rate": 3.130569650632173e-06, "loss": 1.2192, "step": 9303 }, { "epoch": 0.84, "learning_rate": 3.1270971531923737e-06, "loss": 1.1952, "step": 9304 }, { "epoch": 0.84, "learning_rate": 3.123626454218115e-06, "loss": 1.2345, "step": 9305 }, { "epoch": 0.84, "learning_rate": 3.1201575539947515e-06, "loss": 1.2185, "step": 9306 }, { "epoch": 0.84, "learning_rate": 3.1166904528075246e-06, "loss": 1.1589, "step": 9307 }, { "epoch": 0.84, "learning_rate": 3.113225150941504e-06, "loss": 1.1986, "step": 9308 }, { "epoch": 0.84, "learning_rate": 3.109761648681617e-06, "loss": 1.2097, "step": 9309 }, { "epoch": 0.84, "learning_rate": 3.1062999463126575e-06, "loss": 1.1998, "step": 9310 }, { "epoch": 0.84, "learning_rate": 3.10284004411924e-06, "loss": 1.1981, "step": 9311 }, { "epoch": 0.84, "learning_rate": 3.0993819423858623e-06, "loss": 1.199, "step": 9312 }, { "epoch": 0.84, "learning_rate": 3.09592564139686e-06, "loss": 1.2054, "step": 9313 }, { "epoch": 0.84, "learning_rate": 3.092471141436415e-06, "loss": 1.2438, "step": 9314 }, { "epoch": 0.84, "learning_rate": 3.0890184427885828e-06, "loss": 1.1981, "step": 9315 }, { "epoch": 0.84, "learning_rate": 3.0855675457372364e-06, "loss": 1.2159, "step": 9316 }, { "epoch": 0.84, "learning_rate": 3.0821184505661377e-06, "loss": 1.2327, "step": 9317 }, { "epoch": 0.84, "learning_rate": 3.0786711575588767e-06, "loss": 1.1598, "step": 9318 }, { "epoch": 0.84, "learning_rate": 3.075225666998893e-06, "loss": 1.2053, "step": 9319 }, { "epoch": 0.84, "learning_rate": 3.0717819791695077e-06, "loss": 1.1907, "step": 9320 }, { "epoch": 0.84, "learning_rate": 3.06834009435385e-06, "loss": 1.1757, "step": 9321 }, { "epoch": 0.84, "learning_rate": 3.06490001283494e-06, "loss": 1.1845, "step": 9322 }, { "epoch": 0.84, "learning_rate": 3.0614617348956236e-06, "loss": 1.2146, "step": 9323 }, { "epoch": 0.84, "learning_rate": 3.058025260818609e-06, "loss": 1.1374, "step": 9324 }, { "epoch": 0.84, "learning_rate": 3.054590590886458e-06, "loss": 1.2312, "step": 9325 }, { "epoch": 0.84, "learning_rate": 3.05115772538159e-06, "loss": 1.2111, "step": 9326 }, { "epoch": 0.84, "learning_rate": 3.0477266645862423e-06, "loss": 1.1521, "step": 9327 }, { "epoch": 0.84, "learning_rate": 3.044297408782551e-06, "loss": 1.1985, "step": 9328 }, { "epoch": 0.84, "learning_rate": 3.0408699582524735e-06, "loss": 1.2129, "step": 9329 }, { "epoch": 0.84, "learning_rate": 3.0374443132778235e-06, "loss": 1.1717, "step": 9330 }, { "epoch": 0.84, "learning_rate": 3.034020474140281e-06, "loss": 1.1559, "step": 9331 }, { "epoch": 0.84, "learning_rate": 3.0305984411213464e-06, "loss": 1.1967, "step": 9332 }, { "epoch": 0.84, "learning_rate": 3.0271782145024074e-06, "loss": 1.198, "step": 9333 }, { "epoch": 0.84, "learning_rate": 3.0237597945646815e-06, "loss": 1.202, "step": 9334 }, { "epoch": 0.84, "learning_rate": 3.0203431815892384e-06, "loss": 1.2298, "step": 9335 }, { "epoch": 0.84, "learning_rate": 3.01692837585702e-06, "loss": 1.2343, "step": 9336 }, { "epoch": 0.84, "learning_rate": 3.013515377648779e-06, "loss": 1.1344, "step": 9337 }, { "epoch": 0.84, "learning_rate": 3.0101041872451633e-06, "loss": 1.172, "step": 9338 }, { "epoch": 0.84, "learning_rate": 3.006694804926646e-06, "loss": 1.193, "step": 9339 }, { "epoch": 0.84, "learning_rate": 3.003287230973553e-06, "loss": 1.212, "step": 9340 }, { "epoch": 0.84, "learning_rate": 2.999881465666085e-06, "loss": 1.2282, "step": 9341 }, { "epoch": 0.84, "learning_rate": 2.9964775092842482e-06, "loss": 1.1715, "step": 9342 }, { "epoch": 0.84, "learning_rate": 2.9930753621079526e-06, "loss": 1.2055, "step": 9343 }, { "epoch": 0.84, "learning_rate": 2.9896750244169214e-06, "loss": 1.1975, "step": 9344 }, { "epoch": 0.84, "learning_rate": 2.986276496490739e-06, "loss": 1.1719, "step": 9345 }, { "epoch": 0.84, "learning_rate": 2.9828797786088574e-06, "loss": 1.2394, "step": 9346 }, { "epoch": 0.84, "learning_rate": 2.979484871050564e-06, "loss": 1.1798, "step": 9347 }, { "epoch": 0.84, "learning_rate": 2.9760917740949827e-06, "loss": 1.1714, "step": 9348 }, { "epoch": 0.84, "learning_rate": 2.972700488021124e-06, "loss": 1.1439, "step": 9349 }, { "epoch": 0.84, "learning_rate": 2.9693110131078257e-06, "loss": 1.2062, "step": 9350 }, { "epoch": 0.84, "learning_rate": 2.9659233496337786e-06, "loss": 1.1775, "step": 9351 }, { "epoch": 0.84, "learning_rate": 2.962537497877538e-06, "loss": 1.1813, "step": 9352 }, { "epoch": 0.85, "learning_rate": 2.9591534581174834e-06, "loss": 1.1954, "step": 9353 }, { "epoch": 0.85, "learning_rate": 2.955771230631879e-06, "loss": 1.2159, "step": 9354 }, { "epoch": 0.85, "learning_rate": 2.9523908156988133e-06, "loss": 1.1836, "step": 9355 }, { "epoch": 0.85, "learning_rate": 2.949012213596236e-06, "loss": 1.1967, "step": 9356 }, { "epoch": 0.85, "learning_rate": 2.9456354246019613e-06, "loss": 1.2194, "step": 9357 }, { "epoch": 0.85, "learning_rate": 2.942260448993617e-06, "loss": 1.2113, "step": 9358 }, { "epoch": 0.85, "learning_rate": 2.9388872870487223e-06, "loss": 1.2365, "step": 9359 }, { "epoch": 0.85, "learning_rate": 2.935515939044628e-06, "loss": 1.1907, "step": 9360 }, { "epoch": 0.85, "learning_rate": 2.9321464052585263e-06, "loss": 1.2099, "step": 9361 }, { "epoch": 0.85, "learning_rate": 2.9287786859674953e-06, "loss": 1.219, "step": 9362 }, { "epoch": 0.85, "learning_rate": 2.925412781448414e-06, "loss": 1.1398, "step": 9363 }, { "epoch": 0.85, "learning_rate": 2.922048691978055e-06, "loss": 1.1579, "step": 9364 }, { "epoch": 0.85, "learning_rate": 2.918686417833022e-06, "loss": 1.2328, "step": 9365 }, { "epoch": 0.85, "learning_rate": 2.9153259592897663e-06, "loss": 1.1341, "step": 9366 }, { "epoch": 0.85, "learning_rate": 2.9119673166246142e-06, "loss": 1.197, "step": 9367 }, { "epoch": 0.85, "learning_rate": 2.908610490113706e-06, "loss": 1.2114, "step": 9368 }, { "epoch": 0.85, "learning_rate": 2.905255480033053e-06, "loss": 1.2013, "step": 9369 }, { "epoch": 0.85, "learning_rate": 2.9019022866585285e-06, "loss": 1.2002, "step": 9370 }, { "epoch": 0.85, "learning_rate": 2.8985509102658346e-06, "loss": 1.2428, "step": 9371 }, { "epoch": 0.85, "learning_rate": 2.8952013511305305e-06, "loss": 1.1677, "step": 9372 }, { "epoch": 0.85, "learning_rate": 2.891853609528039e-06, "loss": 1.2025, "step": 9373 }, { "epoch": 0.85, "learning_rate": 2.8885076857336104e-06, "loss": 1.2155, "step": 9374 }, { "epoch": 0.85, "learning_rate": 2.885163580022368e-06, "loss": 1.2098, "step": 9375 }, { "epoch": 0.85, "learning_rate": 2.8818212926692706e-06, "loss": 1.2013, "step": 9376 }, { "epoch": 0.85, "learning_rate": 2.8784808239491314e-06, "loss": 1.2297, "step": 9377 }, { "epoch": 0.85, "learning_rate": 2.8751421741366274e-06, "loss": 1.1773, "step": 9378 }, { "epoch": 0.85, "learning_rate": 2.8718053435062536e-06, "loss": 1.1899, "step": 9379 }, { "epoch": 0.85, "learning_rate": 2.8684703323323902e-06, "loss": 1.2471, "step": 9380 }, { "epoch": 0.85, "learning_rate": 2.865137140889251e-06, "loss": 1.1731, "step": 9381 }, { "epoch": 0.85, "learning_rate": 2.8618057694508955e-06, "loss": 1.1954, "step": 9382 }, { "epoch": 0.85, "learning_rate": 2.8584762182912563e-06, "loss": 1.2271, "step": 9383 }, { "epoch": 0.85, "learning_rate": 2.855148487684081e-06, "loss": 1.1925, "step": 9384 }, { "epoch": 0.85, "learning_rate": 2.8518225779030015e-06, "loss": 1.2437, "step": 9385 }, { "epoch": 0.85, "learning_rate": 2.8484984892214795e-06, "loss": 1.2249, "step": 9386 }, { "epoch": 0.85, "learning_rate": 2.845176221912829e-06, "loss": 1.1934, "step": 9387 }, { "epoch": 0.85, "learning_rate": 2.841855776250235e-06, "loss": 1.1644, "step": 9388 }, { "epoch": 0.85, "learning_rate": 2.838537152506698e-06, "loss": 1.2251, "step": 9389 }, { "epoch": 0.85, "learning_rate": 2.8352203509550897e-06, "loss": 1.2554, "step": 9390 }, { "epoch": 0.85, "learning_rate": 2.8319053718681377e-06, "loss": 1.1736, "step": 9391 }, { "epoch": 0.85, "learning_rate": 2.8285922155184025e-06, "loss": 1.1934, "step": 9392 }, { "epoch": 0.85, "learning_rate": 2.825280882178308e-06, "loss": 1.1826, "step": 9393 }, { "epoch": 0.85, "learning_rate": 2.8219713721201217e-06, "loss": 1.2565, "step": 9394 }, { "epoch": 0.85, "learning_rate": 2.8186636856159603e-06, "loss": 1.2199, "step": 9395 }, { "epoch": 0.85, "learning_rate": 2.8153578229377975e-06, "loss": 1.1847, "step": 9396 }, { "epoch": 0.85, "learning_rate": 2.8120537843574545e-06, "loss": 1.1438, "step": 9397 }, { "epoch": 0.85, "learning_rate": 2.808751570146589e-06, "loss": 1.1927, "step": 9398 }, { "epoch": 0.85, "learning_rate": 2.805451180576743e-06, "loss": 1.1824, "step": 9399 }, { "epoch": 0.85, "learning_rate": 2.8021526159192597e-06, "loss": 1.2283, "step": 9400 }, { "epoch": 0.85, "learning_rate": 2.798855876445375e-06, "loss": 1.2377, "step": 9401 }, { "epoch": 0.85, "learning_rate": 2.7955609624261525e-06, "loss": 1.2224, "step": 9402 }, { "epoch": 0.85, "learning_rate": 2.792267874132509e-06, "loss": 1.2057, "step": 9403 }, { "epoch": 0.85, "learning_rate": 2.7889766118352286e-06, "loss": 1.2092, "step": 9404 }, { "epoch": 0.85, "learning_rate": 2.7856871758049075e-06, "loss": 1.1691, "step": 9405 }, { "epoch": 0.85, "learning_rate": 2.7823995663120327e-06, "loss": 1.1846, "step": 9406 }, { "epoch": 0.85, "learning_rate": 2.779113783626916e-06, "loss": 1.1843, "step": 9407 }, { "epoch": 0.85, "learning_rate": 2.775829828019719e-06, "loss": 1.1535, "step": 9408 }, { "epoch": 0.85, "learning_rate": 2.7725476997604777e-06, "loss": 1.2248, "step": 9409 }, { "epoch": 0.85, "learning_rate": 2.7692673991190436e-06, "loss": 1.2029, "step": 9410 }, { "epoch": 0.85, "learning_rate": 2.7659889263651344e-06, "loss": 1.1977, "step": 9411 }, { "epoch": 0.85, "learning_rate": 2.762712281768326e-06, "loss": 1.1763, "step": 9412 }, { "epoch": 0.85, "learning_rate": 2.759437465598036e-06, "loss": 1.1921, "step": 9413 }, { "epoch": 0.85, "learning_rate": 2.756164478123524e-06, "loss": 1.1893, "step": 9414 }, { "epoch": 0.85, "learning_rate": 2.7528933196139112e-06, "loss": 1.1615, "step": 9415 }, { "epoch": 0.85, "learning_rate": 2.749623990338157e-06, "loss": 1.2043, "step": 9416 }, { "epoch": 0.85, "learning_rate": 2.7463564905650858e-06, "loss": 1.1906, "step": 9417 }, { "epoch": 0.85, "learning_rate": 2.7430908205633597e-06, "loss": 1.1712, "step": 9418 }, { "epoch": 0.85, "learning_rate": 2.7398269806014925e-06, "loss": 1.2139, "step": 9419 }, { "epoch": 0.85, "learning_rate": 2.7365649709478468e-06, "loss": 1.1943, "step": 9420 }, { "epoch": 0.85, "learning_rate": 2.7333047918706333e-06, "loss": 1.1837, "step": 9421 }, { "epoch": 0.85, "learning_rate": 2.7300464436379265e-06, "loss": 1.1485, "step": 9422 }, { "epoch": 0.85, "learning_rate": 2.726789926517631e-06, "loss": 1.1846, "step": 9423 }, { "epoch": 0.85, "learning_rate": 2.7235352407775028e-06, "loss": 1.1991, "step": 9424 }, { "epoch": 0.85, "learning_rate": 2.7202823866851746e-06, "loss": 1.2529, "step": 9425 }, { "epoch": 0.85, "learning_rate": 2.71703136450808e-06, "loss": 1.2121, "step": 9426 }, { "epoch": 0.85, "learning_rate": 2.713782174513549e-06, "loss": 1.1854, "step": 9427 }, { "epoch": 0.85, "learning_rate": 2.710534816968735e-06, "loss": 1.1926, "step": 9428 }, { "epoch": 0.85, "learning_rate": 2.7072892921406414e-06, "loss": 1.2228, "step": 9429 }, { "epoch": 0.85, "learning_rate": 2.70404560029614e-06, "loss": 1.199, "step": 9430 }, { "epoch": 0.85, "learning_rate": 2.7008037417019264e-06, "loss": 1.24, "step": 9431 }, { "epoch": 0.85, "learning_rate": 2.697563716624557e-06, "loss": 1.2232, "step": 9432 }, { "epoch": 0.85, "learning_rate": 2.694325525330446e-06, "loss": 1.2265, "step": 9433 }, { "epoch": 0.85, "learning_rate": 2.6910891680858473e-06, "loss": 1.1674, "step": 9434 }, { "epoch": 0.85, "learning_rate": 2.6878546451568593e-06, "loss": 1.2437, "step": 9435 }, { "epoch": 0.85, "learning_rate": 2.684621956809441e-06, "loss": 1.1784, "step": 9436 }, { "epoch": 0.85, "learning_rate": 2.681391103309386e-06, "loss": 1.2595, "step": 9437 }, { "epoch": 0.85, "learning_rate": 2.6781620849223614e-06, "loss": 1.2235, "step": 9438 }, { "epoch": 0.85, "learning_rate": 2.6749349019138586e-06, "loss": 1.2144, "step": 9439 }, { "epoch": 0.85, "learning_rate": 2.6717095545492288e-06, "loss": 1.2024, "step": 9440 }, { "epoch": 0.85, "learning_rate": 2.668486043093671e-06, "loss": 1.2134, "step": 9441 }, { "epoch": 0.85, "learning_rate": 2.6652643678122315e-06, "loss": 1.237, "step": 9442 }, { "epoch": 0.85, "learning_rate": 2.662044528969812e-06, "loss": 1.1858, "step": 9443 }, { "epoch": 0.85, "learning_rate": 2.6588265268311597e-06, "loss": 1.2762, "step": 9444 }, { "epoch": 0.85, "learning_rate": 2.655610361660868e-06, "loss": 1.1703, "step": 9445 }, { "epoch": 0.85, "learning_rate": 2.652396033723378e-06, "loss": 1.227, "step": 9446 }, { "epoch": 0.85, "learning_rate": 2.649183543282982e-06, "loss": 1.1967, "step": 9447 }, { "epoch": 0.85, "learning_rate": 2.645972890603829e-06, "loss": 1.156, "step": 9448 }, { "epoch": 0.85, "learning_rate": 2.6427640759499053e-06, "loss": 1.1785, "step": 9449 }, { "epoch": 0.85, "learning_rate": 2.6395570995850465e-06, "loss": 1.1904, "step": 9450 }, { "epoch": 0.85, "learning_rate": 2.6363519617729565e-06, "loss": 1.2774, "step": 9451 }, { "epoch": 0.85, "learning_rate": 2.6331486627771573e-06, "loss": 1.1966, "step": 9452 }, { "epoch": 0.85, "learning_rate": 2.6299472028610356e-06, "loss": 1.2036, "step": 9453 }, { "epoch": 0.85, "learning_rate": 2.6267475822878364e-06, "loss": 1.2296, "step": 9454 }, { "epoch": 0.85, "learning_rate": 2.6235498013206406e-06, "loss": 1.2263, "step": 9455 }, { "epoch": 0.85, "learning_rate": 2.6203538602223744e-06, "loss": 1.211, "step": 9456 }, { "epoch": 0.85, "learning_rate": 2.617159759255827e-06, "loss": 1.2641, "step": 9457 }, { "epoch": 0.85, "learning_rate": 2.613967498683617e-06, "loss": 1.2285, "step": 9458 }, { "epoch": 0.85, "learning_rate": 2.6107770787682366e-06, "loss": 1.2448, "step": 9459 }, { "epoch": 0.85, "learning_rate": 2.6075884997720097e-06, "loss": 1.1578, "step": 9460 }, { "epoch": 0.85, "learning_rate": 2.6044017619571065e-06, "loss": 1.2039, "step": 9461 }, { "epoch": 0.85, "learning_rate": 2.6012168655855567e-06, "loss": 1.1874, "step": 9462 }, { "epoch": 0.86, "learning_rate": 2.598033810919226e-06, "loss": 1.2327, "step": 9463 }, { "epoch": 0.86, "learning_rate": 2.5948525982198492e-06, "loss": 1.241, "step": 9464 }, { "epoch": 0.86, "learning_rate": 2.5916732277489864e-06, "loss": 1.2239, "step": 9465 }, { "epoch": 0.86, "learning_rate": 2.5884956997680594e-06, "loss": 1.1773, "step": 9466 }, { "epoch": 0.86, "learning_rate": 2.5853200145383365e-06, "loss": 1.218, "step": 9467 }, { "epoch": 0.86, "learning_rate": 2.582146172320926e-06, "loss": 1.2142, "step": 9468 }, { "epoch": 0.86, "learning_rate": 2.578974173376805e-06, "loss": 1.1672, "step": 9469 }, { "epoch": 0.86, "learning_rate": 2.5758040179667793e-06, "loss": 1.2237, "step": 9470 }, { "epoch": 0.86, "learning_rate": 2.5726357063515118e-06, "loss": 1.191, "step": 9471 }, { "epoch": 0.86, "learning_rate": 2.569469238791508e-06, "loss": 1.18, "step": 9472 }, { "epoch": 0.86, "learning_rate": 2.5663046155471297e-06, "loss": 1.1609, "step": 9473 }, { "epoch": 0.86, "learning_rate": 2.563141836878577e-06, "loss": 1.1873, "step": 9474 }, { "epoch": 0.86, "learning_rate": 2.5599809030459138e-06, "loss": 1.2097, "step": 9475 }, { "epoch": 0.86, "learning_rate": 2.556821814309038e-06, "loss": 1.2476, "step": 9476 }, { "epoch": 0.86, "learning_rate": 2.5536645709277025e-06, "loss": 1.1402, "step": 9477 }, { "epoch": 0.86, "learning_rate": 2.5505091731615056e-06, "loss": 1.1924, "step": 9478 }, { "epoch": 0.86, "learning_rate": 2.5473556212698895e-06, "loss": 1.2196, "step": 9479 }, { "epoch": 0.86, "learning_rate": 2.544203915512161e-06, "loss": 1.2633, "step": 9480 }, { "epoch": 0.86, "learning_rate": 2.5410540561474573e-06, "loss": 1.1762, "step": 9481 }, { "epoch": 0.86, "learning_rate": 2.5379060434347734e-06, "loss": 1.2019, "step": 9482 }, { "epoch": 0.86, "learning_rate": 2.53475987763295e-06, "loss": 1.1903, "step": 9483 }, { "epoch": 0.86, "learning_rate": 2.5316155590006664e-06, "loss": 1.2034, "step": 9484 }, { "epoch": 0.86, "learning_rate": 2.5284730877964735e-06, "loss": 1.2153, "step": 9485 }, { "epoch": 0.86, "learning_rate": 2.525332464278751e-06, "loss": 1.1692, "step": 9486 }, { "epoch": 0.86, "learning_rate": 2.522193688705729e-06, "loss": 1.2133, "step": 9487 }, { "epoch": 0.86, "learning_rate": 2.5190567613354893e-06, "loss": 1.2319, "step": 9488 }, { "epoch": 0.86, "learning_rate": 2.515921682425959e-06, "loss": 1.2334, "step": 9489 }, { "epoch": 0.86, "learning_rate": 2.512788452234921e-06, "loss": 1.2147, "step": 9490 }, { "epoch": 0.86, "learning_rate": 2.509657071019997e-06, "loss": 1.1323, "step": 9491 }, { "epoch": 0.86, "learning_rate": 2.5065275390386612e-06, "loss": 1.1867, "step": 9492 }, { "epoch": 0.86, "learning_rate": 2.503399856548233e-06, "loss": 1.1546, "step": 9493 }, { "epoch": 0.86, "learning_rate": 2.500274023805882e-06, "loss": 1.1939, "step": 9494 }, { "epoch": 0.86, "learning_rate": 2.497150041068619e-06, "loss": 1.2323, "step": 9495 }, { "epoch": 0.86, "learning_rate": 2.494027908593319e-06, "loss": 1.1428, "step": 9496 }, { "epoch": 0.86, "learning_rate": 2.4909076266366876e-06, "loss": 1.1576, "step": 9497 }, { "epoch": 0.86, "learning_rate": 2.4877891954552895e-06, "loss": 1.2096, "step": 9498 }, { "epoch": 0.86, "learning_rate": 2.4846726153055307e-06, "loss": 1.2195, "step": 9499 }, { "epoch": 0.86, "learning_rate": 2.481557886443661e-06, "loss": 1.1953, "step": 9500 }, { "epoch": 0.86, "learning_rate": 2.4784450091257962e-06, "loss": 1.2207, "step": 9501 }, { "epoch": 0.86, "learning_rate": 2.4753339836078833e-06, "loss": 1.1915, "step": 9502 }, { "epoch": 0.86, "learning_rate": 2.472224810145718e-06, "loss": 1.2565, "step": 9503 }, { "epoch": 0.86, "learning_rate": 2.4691174889949487e-06, "loss": 1.2036, "step": 9504 }, { "epoch": 0.86, "learning_rate": 2.4660120204110653e-06, "loss": 1.2496, "step": 9505 }, { "epoch": 0.86, "learning_rate": 2.462908404649422e-06, "loss": 1.1991, "step": 9506 }, { "epoch": 0.86, "learning_rate": 2.4598066419652038e-06, "loss": 1.2071, "step": 9507 }, { "epoch": 0.86, "learning_rate": 2.4567067326134454e-06, "loss": 1.2293, "step": 9508 }, { "epoch": 0.86, "learning_rate": 2.453608676849034e-06, "loss": 1.213, "step": 9509 }, { "epoch": 0.86, "learning_rate": 2.4505124749266945e-06, "loss": 1.1713, "step": 9510 }, { "epoch": 0.86, "learning_rate": 2.4474181271010225e-06, "loss": 1.1128, "step": 9511 }, { "epoch": 0.86, "learning_rate": 2.4443256336264374e-06, "loss": 1.1666, "step": 9512 }, { "epoch": 0.86, "learning_rate": 2.4412349947572162e-06, "loss": 1.2218, "step": 9513 }, { "epoch": 0.86, "learning_rate": 2.43814621074748e-06, "loss": 1.197, "step": 9514 }, { "epoch": 0.86, "learning_rate": 2.4350592818512014e-06, "loss": 1.181, "step": 9515 }, { "epoch": 0.86, "learning_rate": 2.431974208322191e-06, "loss": 1.2298, "step": 9516 }, { "epoch": 0.86, "learning_rate": 2.428890990414126e-06, "loss": 1.1644, "step": 9517 }, { "epoch": 0.86, "learning_rate": 2.4258096283805155e-06, "loss": 1.1973, "step": 9518 }, { "epoch": 0.86, "learning_rate": 2.4227301224747174e-06, "loss": 1.2132, "step": 9519 }, { "epoch": 0.86, "learning_rate": 2.4196524729499376e-06, "loss": 1.1989, "step": 9520 }, { "epoch": 0.86, "learning_rate": 2.416576680059232e-06, "loss": 1.2234, "step": 9521 }, { "epoch": 0.86, "learning_rate": 2.413502744055507e-06, "loss": 1.2372, "step": 9522 }, { "epoch": 0.86, "learning_rate": 2.41043066519151e-06, "loss": 1.1765, "step": 9523 }, { "epoch": 0.86, "learning_rate": 2.407360443719836e-06, "loss": 1.14, "step": 9524 }, { "epoch": 0.86, "learning_rate": 2.4042920798929334e-06, "loss": 1.2219, "step": 9525 }, { "epoch": 0.86, "learning_rate": 2.401225573963084e-06, "loss": 1.2012, "step": 9526 }, { "epoch": 0.86, "learning_rate": 2.3981609261824383e-06, "loss": 1.1761, "step": 9527 }, { "epoch": 0.86, "learning_rate": 2.3950981368029785e-06, "loss": 1.1734, "step": 9528 }, { "epoch": 0.86, "learning_rate": 2.392037206076536e-06, "loss": 1.1812, "step": 9529 }, { "epoch": 0.86, "learning_rate": 2.3889781342547905e-06, "loss": 1.2292, "step": 9530 }, { "epoch": 0.86, "learning_rate": 2.3859209215892684e-06, "loss": 1.1868, "step": 9531 }, { "epoch": 0.86, "learning_rate": 2.3828655683313492e-06, "loss": 1.207, "step": 9532 }, { "epoch": 0.86, "learning_rate": 2.3798120747322515e-06, "loss": 1.1796, "step": 9533 }, { "epoch": 0.86, "learning_rate": 2.3767604410430464e-06, "loss": 1.2765, "step": 9534 }, { "epoch": 0.86, "learning_rate": 2.3737106675146446e-06, "loss": 1.208, "step": 9535 }, { "epoch": 0.86, "learning_rate": 2.3706627543978148e-06, "loss": 1.1795, "step": 9536 }, { "epoch": 0.86, "learning_rate": 2.3676167019431592e-06, "loss": 1.2394, "step": 9537 }, { "epoch": 0.86, "learning_rate": 2.364572510401142e-06, "loss": 1.2132, "step": 9538 }, { "epoch": 0.86, "learning_rate": 2.3615301800220673e-06, "loss": 1.2385, "step": 9539 }, { "epoch": 0.86, "learning_rate": 2.3584897110560833e-06, "loss": 1.2234, "step": 9540 }, { "epoch": 0.86, "learning_rate": 2.3554511037531897e-06, "loss": 1.1774, "step": 9541 }, { "epoch": 0.86, "learning_rate": 2.3524143583632225e-06, "loss": 1.1981, "step": 9542 }, { "epoch": 0.86, "learning_rate": 2.349379475135885e-06, "loss": 1.2054, "step": 9543 }, { "epoch": 0.86, "learning_rate": 2.346346454320714e-06, "loss": 1.2094, "step": 9544 }, { "epoch": 0.86, "learning_rate": 2.3433152961670896e-06, "loss": 1.1936, "step": 9545 }, { "epoch": 0.86, "learning_rate": 2.3402860009242496e-06, "loss": 1.2447, "step": 9546 }, { "epoch": 0.86, "learning_rate": 2.3372585688412635e-06, "loss": 1.1823, "step": 9547 }, { "epoch": 0.86, "learning_rate": 2.3342330001670687e-06, "loss": 1.1495, "step": 9548 }, { "epoch": 0.86, "learning_rate": 2.3312092951504353e-06, "loss": 1.1885, "step": 9549 }, { "epoch": 0.86, "learning_rate": 2.328187454039979e-06, "loss": 1.1777, "step": 9550 }, { "epoch": 0.86, "learning_rate": 2.325167477084167e-06, "loss": 1.2307, "step": 9551 }, { "epoch": 0.86, "learning_rate": 2.322149364531309e-06, "loss": 1.2312, "step": 9552 }, { "epoch": 0.86, "learning_rate": 2.319133116629574e-06, "loss": 1.2395, "step": 9553 }, { "epoch": 0.86, "learning_rate": 2.3161187336269626e-06, "loss": 1.204, "step": 9554 }, { "epoch": 0.86, "learning_rate": 2.3131062157713295e-06, "loss": 1.2526, "step": 9555 }, { "epoch": 0.86, "learning_rate": 2.3100955633103688e-06, "loss": 1.2083, "step": 9556 }, { "epoch": 0.86, "learning_rate": 2.3070867764916347e-06, "loss": 1.1779, "step": 9557 }, { "epoch": 0.86, "learning_rate": 2.3040798555625103e-06, "loss": 1.2165, "step": 9558 }, { "epoch": 0.86, "learning_rate": 2.3010748007702447e-06, "loss": 1.1679, "step": 9559 }, { "epoch": 0.86, "learning_rate": 2.298071612361921e-06, "loss": 1.1671, "step": 9560 }, { "epoch": 0.86, "learning_rate": 2.2950702905844695e-06, "loss": 1.1931, "step": 9561 }, { "epoch": 0.86, "learning_rate": 2.2920708356846705e-06, "loss": 1.2007, "step": 9562 }, { "epoch": 0.86, "learning_rate": 2.2890732479091457e-06, "loss": 1.1736, "step": 9563 }, { "epoch": 0.86, "learning_rate": 2.2860775275043737e-06, "loss": 1.1574, "step": 9564 }, { "epoch": 0.86, "learning_rate": 2.2830836747166713e-06, "loss": 1.2341, "step": 9565 }, { "epoch": 0.86, "learning_rate": 2.280091689792202e-06, "loss": 1.1813, "step": 9566 }, { "epoch": 0.86, "learning_rate": 2.277101572976978e-06, "loss": 1.2265, "step": 9567 }, { "epoch": 0.86, "learning_rate": 2.2741133245168523e-06, "loss": 1.1748, "step": 9568 }, { "epoch": 0.86, "learning_rate": 2.271126944657537e-06, "loss": 1.2052, "step": 9569 }, { "epoch": 0.86, "learning_rate": 2.2681424336445794e-06, "loss": 1.2059, "step": 9570 }, { "epoch": 0.86, "learning_rate": 2.265159791723373e-06, "loss": 1.1878, "step": 9571 }, { "epoch": 0.86, "learning_rate": 2.2621790191391658e-06, "loss": 1.216, "step": 9572 }, { "epoch": 0.86, "learning_rate": 2.2592001161370392e-06, "loss": 1.1459, "step": 9573 }, { "epoch": 0.87, "learning_rate": 2.256223082961939e-06, "loss": 1.1367, "step": 9574 }, { "epoch": 0.87, "learning_rate": 2.2532479198586447e-06, "loss": 1.2283, "step": 9575 }, { "epoch": 0.87, "learning_rate": 2.25027462707178e-06, "loss": 1.2502, "step": 9576 }, { "epoch": 0.87, "learning_rate": 2.2473032048458215e-06, "loss": 1.1827, "step": 9577 }, { "epoch": 0.87, "learning_rate": 2.244333653425093e-06, "loss": 1.2785, "step": 9578 }, { "epoch": 0.87, "learning_rate": 2.2413659730537497e-06, "loss": 1.2086, "step": 9579 }, { "epoch": 0.87, "learning_rate": 2.238400163975821e-06, "loss": 1.1935, "step": 9580 }, { "epoch": 0.87, "learning_rate": 2.2354362264351557e-06, "loss": 1.1716, "step": 9581 }, { "epoch": 0.87, "learning_rate": 2.2324741606754627e-06, "loss": 1.2102, "step": 9582 }, { "epoch": 0.87, "learning_rate": 2.2295139669402913e-06, "loss": 1.1893, "step": 9583 }, { "epoch": 0.87, "learning_rate": 2.2265556454730356e-06, "loss": 1.1943, "step": 9584 }, { "epoch": 0.87, "learning_rate": 2.2235991965169483e-06, "loss": 1.1691, "step": 9585 }, { "epoch": 0.87, "learning_rate": 2.22064462031511e-06, "loss": 1.2337, "step": 9586 }, { "epoch": 0.87, "learning_rate": 2.217691917110462e-06, "loss": 1.2014, "step": 9587 }, { "epoch": 0.87, "learning_rate": 2.214741087145783e-06, "loss": 1.2208, "step": 9588 }, { "epoch": 0.87, "learning_rate": 2.211792130663695e-06, "loss": 1.1496, "step": 9589 }, { "epoch": 0.87, "learning_rate": 2.2088450479066822e-06, "loss": 1.1315, "step": 9590 }, { "epoch": 0.87, "learning_rate": 2.2058998391170594e-06, "loss": 1.228, "step": 9591 }, { "epoch": 0.87, "learning_rate": 2.2029565045369906e-06, "loss": 1.2104, "step": 9592 }, { "epoch": 0.87, "learning_rate": 2.2000150444084876e-06, "loss": 1.2044, "step": 9593 }, { "epoch": 0.87, "learning_rate": 2.197075458973402e-06, "loss": 1.1649, "step": 9594 }, { "epoch": 0.87, "learning_rate": 2.1941377484734487e-06, "loss": 1.2215, "step": 9595 }, { "epoch": 0.87, "learning_rate": 2.1912019131501734e-06, "loss": 1.2044, "step": 9596 }, { "epoch": 0.87, "learning_rate": 2.1882679532449572e-06, "loss": 1.2069, "step": 9597 }, { "epoch": 0.87, "learning_rate": 2.1853358689990556e-06, "loss": 1.2186, "step": 9598 }, { "epoch": 0.87, "learning_rate": 2.182405660653547e-06, "loss": 1.1663, "step": 9599 }, { "epoch": 0.87, "learning_rate": 2.1794773284493615e-06, "loss": 1.211, "step": 9600 }, { "epoch": 0.87, "learning_rate": 2.1765508726272865e-06, "loss": 1.2413, "step": 9601 }, { "epoch": 0.87, "learning_rate": 2.1736262934279382e-06, "loss": 1.1803, "step": 9602 }, { "epoch": 0.87, "learning_rate": 2.1707035910917856e-06, "loss": 1.2252, "step": 9603 }, { "epoch": 0.87, "learning_rate": 2.167782765859147e-06, "loss": 1.1992, "step": 9604 }, { "epoch": 0.87, "learning_rate": 2.164863817970175e-06, "loss": 1.2212, "step": 9605 }, { "epoch": 0.87, "learning_rate": 2.1619467476648834e-06, "loss": 1.1964, "step": 9606 }, { "epoch": 0.87, "learning_rate": 2.159031555183125e-06, "loss": 1.187, "step": 9607 }, { "epoch": 0.87, "learning_rate": 2.15611824076459e-06, "loss": 1.2052, "step": 9608 }, { "epoch": 0.87, "learning_rate": 2.153206804648825e-06, "loss": 1.1933, "step": 9609 }, { "epoch": 0.87, "learning_rate": 2.1502972470752142e-06, "loss": 1.1552, "step": 9610 }, { "epoch": 0.87, "learning_rate": 2.147389568282998e-06, "loss": 1.2326, "step": 9611 }, { "epoch": 0.87, "learning_rate": 2.1444837685112546e-06, "loss": 1.215, "step": 9612 }, { "epoch": 0.87, "learning_rate": 2.1415798479989036e-06, "loss": 1.2125, "step": 9613 }, { "epoch": 0.87, "learning_rate": 2.1386778069847234e-06, "loss": 1.2313, "step": 9614 }, { "epoch": 0.87, "learning_rate": 2.135777645707318e-06, "loss": 1.219, "step": 9615 }, { "epoch": 0.87, "learning_rate": 2.1328793644051605e-06, "loss": 1.1782, "step": 9616 }, { "epoch": 0.87, "learning_rate": 2.129982963316557e-06, "loss": 1.1861, "step": 9617 }, { "epoch": 0.87, "learning_rate": 2.127088442679648e-06, "loss": 1.1552, "step": 9618 }, { "epoch": 0.87, "learning_rate": 2.1241958027324397e-06, "loss": 1.2194, "step": 9619 }, { "epoch": 0.87, "learning_rate": 2.1213050437127762e-06, "loss": 1.2476, "step": 9620 }, { "epoch": 0.87, "learning_rate": 2.118416165858339e-06, "loss": 1.2261, "step": 9621 }, { "epoch": 0.87, "learning_rate": 2.1155291694066742e-06, "loss": 1.2404, "step": 9622 }, { "epoch": 0.87, "learning_rate": 2.1126440545951447e-06, "loss": 1.1777, "step": 9623 }, { "epoch": 0.87, "learning_rate": 2.109760821660986e-06, "loss": 1.1781, "step": 9624 }, { "epoch": 0.87, "learning_rate": 2.106879470841261e-06, "loss": 1.2308, "step": 9625 }, { "epoch": 0.87, "learning_rate": 2.104000002372886e-06, "loss": 1.211, "step": 9626 }, { "epoch": 0.87, "learning_rate": 2.1011224164926234e-06, "loss": 1.233, "step": 9627 }, { "epoch": 0.87, "learning_rate": 2.098246713437077e-06, "loss": 1.1495, "step": 9628 }, { "epoch": 0.87, "learning_rate": 2.0953728934426985e-06, "loss": 1.1808, "step": 9629 }, { "epoch": 0.87, "learning_rate": 2.09250095674578e-06, "loss": 1.1986, "step": 9630 }, { "epoch": 0.87, "learning_rate": 2.0896309035824595e-06, "loss": 1.2164, "step": 9631 }, { "epoch": 0.87, "learning_rate": 2.0867627341887297e-06, "loss": 1.231, "step": 9632 }, { "epoch": 0.87, "learning_rate": 2.0838964488004236e-06, "loss": 1.2445, "step": 9633 }, { "epoch": 0.87, "learning_rate": 2.081032047653203e-06, "loss": 1.1638, "step": 9634 }, { "epoch": 0.87, "learning_rate": 2.0781695309826015e-06, "loss": 1.2084, "step": 9635 }, { "epoch": 0.87, "learning_rate": 2.0753088990239783e-06, "loss": 1.1468, "step": 9636 }, { "epoch": 0.87, "learning_rate": 2.0724501520125506e-06, "loss": 1.2184, "step": 9637 }, { "epoch": 0.87, "learning_rate": 2.0695932901833747e-06, "loss": 1.2635, "step": 9638 }, { "epoch": 0.87, "learning_rate": 2.066738313771341e-06, "loss": 1.2383, "step": 9639 }, { "epoch": 0.87, "learning_rate": 2.0638852230112055e-06, "loss": 1.1878, "step": 9640 }, { "epoch": 0.87, "learning_rate": 2.061034018137556e-06, "loss": 1.1979, "step": 9641 }, { "epoch": 0.87, "learning_rate": 2.058184699384827e-06, "loss": 1.2365, "step": 9642 }, { "epoch": 0.87, "learning_rate": 2.055337266987306e-06, "loss": 1.1798, "step": 9643 }, { "epoch": 0.87, "learning_rate": 2.052491721179109e-06, "loss": 1.2034, "step": 9644 }, { "epoch": 0.87, "learning_rate": 2.0496480621942123e-06, "loss": 1.2117, "step": 9645 }, { "epoch": 0.87, "learning_rate": 2.0468062902664314e-06, "loss": 1.1833, "step": 9646 }, { "epoch": 0.87, "learning_rate": 2.0439664056294215e-06, "loss": 1.1894, "step": 9647 }, { "epoch": 0.87, "learning_rate": 2.041128408516696e-06, "loss": 1.2144, "step": 9648 }, { "epoch": 0.87, "learning_rate": 2.038292299161601e-06, "loss": 1.1887, "step": 9649 }, { "epoch": 0.87, "learning_rate": 2.035458077797331e-06, "loss": 1.2027, "step": 9650 }, { "epoch": 0.87, "learning_rate": 2.032625744656924e-06, "loss": 1.1716, "step": 9651 }, { "epoch": 0.87, "learning_rate": 2.029795299973261e-06, "loss": 1.2047, "step": 9652 }, { "epoch": 0.87, "learning_rate": 2.0269667439790807e-06, "loss": 1.2056, "step": 9653 }, { "epoch": 0.87, "learning_rate": 2.0241400769069554e-06, "loss": 1.1778, "step": 9654 }, { "epoch": 0.87, "learning_rate": 2.0213152989892903e-06, "loss": 1.1727, "step": 9655 }, { "epoch": 0.87, "learning_rate": 2.0184924104583613e-06, "loss": 1.2315, "step": 9656 }, { "epoch": 0.87, "learning_rate": 2.0156714115462687e-06, "loss": 1.2178, "step": 9657 }, { "epoch": 0.87, "learning_rate": 2.0128523024849744e-06, "loss": 1.1902, "step": 9658 }, { "epoch": 0.87, "learning_rate": 2.0100350835062705e-06, "loss": 1.199, "step": 9659 }, { "epoch": 0.87, "learning_rate": 2.007219754841791e-06, "loss": 1.1962, "step": 9660 }, { "epoch": 0.87, "learning_rate": 2.004406316723029e-06, "loss": 1.1878, "step": 9661 }, { "epoch": 0.87, "learning_rate": 2.001594769381318e-06, "loss": 1.2757, "step": 9662 }, { "epoch": 0.87, "learning_rate": 1.9987851130478213e-06, "loss": 1.1854, "step": 9663 }, { "epoch": 0.87, "learning_rate": 1.995977347953579e-06, "loss": 1.1853, "step": 9664 }, { "epoch": 0.87, "learning_rate": 1.9931714743294305e-06, "loss": 1.2057, "step": 9665 }, { "epoch": 0.87, "learning_rate": 1.9903674924061034e-06, "loss": 1.2322, "step": 9666 }, { "epoch": 0.87, "learning_rate": 1.987565402414143e-06, "loss": 1.1935, "step": 9667 }, { "epoch": 0.87, "learning_rate": 1.9847652045839437e-06, "loss": 1.2202, "step": 9668 }, { "epoch": 0.87, "learning_rate": 1.9819668991457597e-06, "loss": 1.1355, "step": 9669 }, { "epoch": 0.87, "learning_rate": 1.979170486329662e-06, "loss": 1.1874, "step": 9670 }, { "epoch": 0.87, "learning_rate": 1.97637596636559e-06, "loss": 1.1851, "step": 9671 }, { "epoch": 0.87, "learning_rate": 1.9735833394833202e-06, "loss": 1.2235, "step": 9672 }, { "epoch": 0.87, "learning_rate": 1.9707926059124637e-06, "loss": 1.2257, "step": 9673 }, { "epoch": 0.87, "learning_rate": 1.9680037658824953e-06, "loss": 1.1832, "step": 9674 }, { "epoch": 0.87, "learning_rate": 1.96521681962272e-06, "loss": 1.2063, "step": 9675 }, { "epoch": 0.87, "learning_rate": 1.962431767362277e-06, "loss": 1.1967, "step": 9676 }, { "epoch": 0.87, "learning_rate": 1.959648609330181e-06, "loss": 1.1836, "step": 9677 }, { "epoch": 0.87, "learning_rate": 1.956867345755259e-06, "loss": 1.1919, "step": 9678 }, { "epoch": 0.87, "learning_rate": 1.954087976866209e-06, "loss": 1.1906, "step": 9679 }, { "epoch": 0.87, "learning_rate": 1.951310502891554e-06, "loss": 1.1918, "step": 9680 }, { "epoch": 0.87, "learning_rate": 1.9485349240596613e-06, "loss": 1.1733, "step": 9681 }, { "epoch": 0.87, "learning_rate": 1.945761240598759e-06, "loss": 1.2498, "step": 9682 }, { "epoch": 0.87, "learning_rate": 1.942989452736904e-06, "loss": 1.152, "step": 9683 }, { "epoch": 0.87, "learning_rate": 1.9402195607019975e-06, "loss": 1.2365, "step": 9684 }, { "epoch": 0.88, "learning_rate": 1.937451564721804e-06, "loss": 1.2155, "step": 9685 }, { "epoch": 0.88, "learning_rate": 1.9346854650239e-06, "loss": 1.2173, "step": 9686 }, { "epoch": 0.88, "learning_rate": 1.931921261835734e-06, "loss": 1.2328, "step": 9687 }, { "epoch": 0.88, "learning_rate": 1.9291589553845885e-06, "loss": 1.2469, "step": 9688 }, { "epoch": 0.88, "learning_rate": 1.9263985458975843e-06, "loss": 1.2394, "step": 9689 }, { "epoch": 0.88, "learning_rate": 1.9236400336017034e-06, "loss": 1.1946, "step": 9690 }, { "epoch": 0.88, "learning_rate": 1.920883418723743e-06, "loss": 1.1851, "step": 9691 }, { "epoch": 0.88, "learning_rate": 1.9181287014903738e-06, "loss": 1.1818, "step": 9692 }, { "epoch": 0.88, "learning_rate": 1.9153758821280954e-06, "loss": 1.2189, "step": 9693 }, { "epoch": 0.88, "learning_rate": 1.912624960863249e-06, "loss": 1.1989, "step": 9694 }, { "epoch": 0.88, "learning_rate": 1.90987593792204e-06, "loss": 1.1839, "step": 9695 }, { "epoch": 0.88, "learning_rate": 1.9071288135304871e-06, "loss": 1.2092, "step": 9696 }, { "epoch": 0.88, "learning_rate": 1.9043835879144684e-06, "loss": 1.227, "step": 9697 }, { "epoch": 0.88, "learning_rate": 1.901640261299714e-06, "loss": 1.1968, "step": 9698 }, { "epoch": 0.88, "learning_rate": 1.898898833911783e-06, "loss": 1.2179, "step": 9699 }, { "epoch": 0.88, "learning_rate": 1.8961593059760945e-06, "loss": 1.2323, "step": 9700 }, { "epoch": 0.88, "learning_rate": 1.8934216777178965e-06, "loss": 1.2527, "step": 9701 }, { "epoch": 0.88, "learning_rate": 1.8906859493622808e-06, "loss": 1.2177, "step": 9702 }, { "epoch": 0.88, "learning_rate": 1.8879521211341955e-06, "loss": 1.1918, "step": 9703 }, { "epoch": 0.88, "learning_rate": 1.8852201932584218e-06, "loss": 1.2409, "step": 9704 }, { "epoch": 0.88, "learning_rate": 1.8824901659595884e-06, "loss": 1.1402, "step": 9705 }, { "epoch": 0.88, "learning_rate": 1.879762039462174e-06, "loss": 1.1861, "step": 9706 }, { "epoch": 0.88, "learning_rate": 1.8770358139904826e-06, "loss": 1.1607, "step": 9707 }, { "epoch": 0.88, "learning_rate": 1.8743114897686848e-06, "loss": 1.2041, "step": 9708 }, { "epoch": 0.88, "learning_rate": 1.8715890670207792e-06, "loss": 1.1966, "step": 9709 }, { "epoch": 0.88, "learning_rate": 1.8688685459706063e-06, "loss": 1.2217, "step": 9710 }, { "epoch": 0.88, "learning_rate": 1.8661499268418759e-06, "loss": 1.1795, "step": 9711 }, { "epoch": 0.88, "learning_rate": 1.8634332098581008e-06, "loss": 1.1421, "step": 9712 }, { "epoch": 0.88, "learning_rate": 1.8607183952426721e-06, "loss": 1.2211, "step": 9713 }, { "epoch": 0.88, "learning_rate": 1.8580054832188055e-06, "loss": 1.1787, "step": 9714 }, { "epoch": 0.88, "learning_rate": 1.8552944740095667e-06, "loss": 1.2409, "step": 9715 }, { "epoch": 0.88, "learning_rate": 1.8525853678378696e-06, "loss": 1.2564, "step": 9716 }, { "epoch": 0.88, "learning_rate": 1.8498781649264608e-06, "loss": 1.2322, "step": 9717 }, { "epoch": 0.88, "learning_rate": 1.8471728654979315e-06, "loss": 1.2295, "step": 9718 }, { "epoch": 0.88, "learning_rate": 1.8444694697747317e-06, "loss": 1.1887, "step": 9719 }, { "epoch": 0.88, "learning_rate": 1.8417679779791308e-06, "loss": 1.168, "step": 9720 }, { "epoch": 0.88, "learning_rate": 1.8390683903332734e-06, "loss": 1.1571, "step": 9721 }, { "epoch": 0.88, "learning_rate": 1.8363707070591124e-06, "loss": 1.1945, "step": 9722 }, { "epoch": 0.88, "learning_rate": 1.833674928378465e-06, "loss": 1.2344, "step": 9723 }, { "epoch": 0.88, "learning_rate": 1.8309810545129896e-06, "loss": 1.1727, "step": 9724 }, { "epoch": 0.88, "learning_rate": 1.8282890856841845e-06, "loss": 1.2257, "step": 9725 }, { "epoch": 0.88, "learning_rate": 1.8255990221133917e-06, "loss": 1.2668, "step": 9726 }, { "epoch": 0.88, "learning_rate": 1.822910864021804e-06, "loss": 1.1997, "step": 9727 }, { "epoch": 0.88, "learning_rate": 1.820224611630439e-06, "loss": 1.2022, "step": 9728 }, { "epoch": 0.88, "learning_rate": 1.817540265160178e-06, "loss": 1.1235, "step": 9729 }, { "epoch": 0.88, "learning_rate": 1.8148578248317394e-06, "loss": 1.2161, "step": 9730 }, { "epoch": 0.88, "learning_rate": 1.8121772908656715e-06, "loss": 1.2568, "step": 9731 }, { "epoch": 0.88, "learning_rate": 1.8094986634823951e-06, "loss": 1.2235, "step": 9732 }, { "epoch": 0.88, "learning_rate": 1.806821942902137e-06, "loss": 1.1934, "step": 9733 }, { "epoch": 0.88, "learning_rate": 1.8041471293449963e-06, "loss": 1.2166, "step": 9734 }, { "epoch": 0.88, "learning_rate": 1.8014742230309078e-06, "loss": 1.2359, "step": 9735 }, { "epoch": 0.88, "learning_rate": 1.7988032241796376e-06, "loss": 1.1916, "step": 9736 }, { "epoch": 0.88, "learning_rate": 1.7961341330108184e-06, "loss": 1.1983, "step": 9737 }, { "epoch": 0.88, "learning_rate": 1.7934669497438994e-06, "loss": 1.1922, "step": 9738 }, { "epoch": 0.88, "learning_rate": 1.790801674598186e-06, "loss": 1.1899, "step": 9739 }, { "epoch": 0.88, "learning_rate": 1.7881383077928332e-06, "loss": 1.2034, "step": 9740 }, { "epoch": 0.88, "learning_rate": 1.7854768495468272e-06, "loss": 1.2138, "step": 9741 }, { "epoch": 0.88, "learning_rate": 1.782817300079012e-06, "loss": 1.2144, "step": 9742 }, { "epoch": 0.88, "learning_rate": 1.7801596596080521e-06, "loss": 1.1629, "step": 9743 }, { "epoch": 0.88, "learning_rate": 1.7775039283524724e-06, "loss": 1.2111, "step": 9744 }, { "epoch": 0.88, "learning_rate": 1.7748501065306371e-06, "loss": 1.2211, "step": 9745 }, { "epoch": 0.88, "learning_rate": 1.772198194360755e-06, "loss": 1.2088, "step": 9746 }, { "epoch": 0.88, "learning_rate": 1.7695481920608715e-06, "loss": 1.2567, "step": 9747 }, { "epoch": 0.88, "learning_rate": 1.766900099848881e-06, "loss": 1.1958, "step": 9748 }, { "epoch": 0.88, "learning_rate": 1.7642539179425156e-06, "loss": 1.1908, "step": 9749 }, { "epoch": 0.88, "learning_rate": 1.761609646559359e-06, "loss": 1.2289, "step": 9750 }, { "epoch": 0.88, "learning_rate": 1.7589672859168293e-06, "loss": 1.1942, "step": 9751 }, { "epoch": 0.88, "learning_rate": 1.756326836232186e-06, "loss": 1.1891, "step": 9752 }, { "epoch": 0.88, "learning_rate": 1.7536882977225499e-06, "loss": 1.142, "step": 9753 }, { "epoch": 0.88, "learning_rate": 1.751051670604853e-06, "loss": 1.2023, "step": 9754 }, { "epoch": 0.88, "learning_rate": 1.7484169550958994e-06, "loss": 1.2083, "step": 9755 }, { "epoch": 0.88, "learning_rate": 1.745784151412322e-06, "loss": 1.2086, "step": 9756 }, { "epoch": 0.88, "learning_rate": 1.743153259770594e-06, "loss": 1.2297, "step": 9757 }, { "epoch": 0.88, "learning_rate": 1.7405242803870486e-06, "loss": 1.1854, "step": 9758 }, { "epoch": 0.88, "learning_rate": 1.7378972134778404e-06, "loss": 1.1697, "step": 9759 }, { "epoch": 0.88, "learning_rate": 1.7352720592589715e-06, "loss": 1.1655, "step": 9760 }, { "epoch": 0.88, "learning_rate": 1.7326488179463002e-06, "loss": 1.1662, "step": 9761 }, { "epoch": 0.88, "learning_rate": 1.730027489755512e-06, "loss": 1.2296, "step": 9762 }, { "epoch": 0.88, "learning_rate": 1.727408074902151e-06, "loss": 1.2313, "step": 9763 }, { "epoch": 0.88, "learning_rate": 1.7247905736015867e-06, "loss": 1.1807, "step": 9764 }, { "epoch": 0.88, "learning_rate": 1.7221749860690363e-06, "loss": 1.2402, "step": 9765 }, { "epoch": 0.88, "learning_rate": 1.7195613125195686e-06, "loss": 1.2578, "step": 9766 }, { "epoch": 0.88, "learning_rate": 1.7169495531680874e-06, "loss": 1.1442, "step": 9767 }, { "epoch": 0.88, "learning_rate": 1.7143397082293427e-06, "loss": 1.2143, "step": 9768 }, { "epoch": 0.88, "learning_rate": 1.7117317779179216e-06, "loss": 1.183, "step": 9769 }, { "epoch": 0.88, "learning_rate": 1.709125762448252e-06, "loss": 1.2256, "step": 9770 }, { "epoch": 0.88, "learning_rate": 1.7065216620346213e-06, "loss": 1.177, "step": 9771 }, { "epoch": 0.88, "learning_rate": 1.7039194768911416e-06, "loss": 1.1962, "step": 9772 }, { "epoch": 0.88, "learning_rate": 1.701319207231769e-06, "loss": 1.2263, "step": 9773 }, { "epoch": 0.88, "learning_rate": 1.6987208532703192e-06, "loss": 1.1968, "step": 9774 }, { "epoch": 0.88, "learning_rate": 1.696124415220421e-06, "loss": 1.1711, "step": 9775 }, { "epoch": 0.88, "learning_rate": 1.6935298932955784e-06, "loss": 1.2884, "step": 9776 }, { "epoch": 0.88, "learning_rate": 1.690937287709113e-06, "loss": 1.2049, "step": 9777 }, { "epoch": 0.88, "learning_rate": 1.6883465986741953e-06, "loss": 1.1792, "step": 9778 }, { "epoch": 0.88, "learning_rate": 1.6857578264038525e-06, "loss": 1.1701, "step": 9779 }, { "epoch": 0.88, "learning_rate": 1.6831709711109339e-06, "loss": 1.2252, "step": 9780 }, { "epoch": 0.88, "learning_rate": 1.6805860330081359e-06, "loss": 1.1812, "step": 9781 }, { "epoch": 0.88, "learning_rate": 1.6780030123080076e-06, "loss": 1.2211, "step": 9782 }, { "epoch": 0.88, "learning_rate": 1.675421909222935e-06, "loss": 1.2139, "step": 9783 }, { "epoch": 0.88, "learning_rate": 1.6728427239651428e-06, "loss": 1.225, "step": 9784 }, { "epoch": 0.88, "learning_rate": 1.6702654567466974e-06, "loss": 1.2133, "step": 9785 }, { "epoch": 0.88, "learning_rate": 1.6676901077795127e-06, "loss": 1.1286, "step": 9786 }, { "epoch": 0.88, "learning_rate": 1.6651166772753441e-06, "loss": 1.174, "step": 9787 }, { "epoch": 0.88, "learning_rate": 1.662545165445789e-06, "loss": 1.2163, "step": 9788 }, { "epoch": 0.88, "learning_rate": 1.6599755725022843e-06, "loss": 1.1956, "step": 9789 }, { "epoch": 0.88, "learning_rate": 1.657407898656113e-06, "loss": 1.1745, "step": 9790 }, { "epoch": 0.88, "learning_rate": 1.6548421441183875e-06, "loss": 1.1904, "step": 9791 }, { "epoch": 0.88, "learning_rate": 1.6522783091000888e-06, "loss": 1.2194, "step": 9792 }, { "epoch": 0.88, "learning_rate": 1.6497163938120146e-06, "loss": 1.2204, "step": 9793 }, { "epoch": 0.88, "learning_rate": 1.6471563984648192e-06, "loss": 1.176, "step": 9794 }, { "epoch": 0.89, "learning_rate": 1.6445983232689893e-06, "loss": 1.2027, "step": 9795 }, { "epoch": 0.89, "learning_rate": 1.6420421684348597e-06, "loss": 1.1907, "step": 9796 }, { "epoch": 0.89, "learning_rate": 1.6394879341726094e-06, "loss": 1.2111, "step": 9797 }, { "epoch": 0.89, "learning_rate": 1.6369356206922565e-06, "loss": 1.1647, "step": 9798 }, { "epoch": 0.89, "learning_rate": 1.6343852282036525e-06, "loss": 1.2255, "step": 9799 }, { "epoch": 0.89, "learning_rate": 1.631836756916516e-06, "loss": 1.1808, "step": 9800 }, { "epoch": 0.89, "learning_rate": 1.6292902070403765e-06, "loss": 1.1509, "step": 9801 }, { "epoch": 0.89, "learning_rate": 1.626745578784622e-06, "loss": 1.1581, "step": 9802 }, { "epoch": 0.89, "learning_rate": 1.624202872358488e-06, "loss": 1.2037, "step": 9803 }, { "epoch": 0.89, "learning_rate": 1.6216620879710376e-06, "loss": 1.175, "step": 9804 }, { "epoch": 0.89, "learning_rate": 1.6191232258311872e-06, "loss": 1.1892, "step": 9805 }, { "epoch": 0.89, "learning_rate": 1.6165862861476867e-06, "loss": 1.2087, "step": 9806 }, { "epoch": 0.89, "learning_rate": 1.6140512691291326e-06, "loss": 1.1859, "step": 9807 }, { "epoch": 0.89, "learning_rate": 1.611518174983967e-06, "loss": 1.1959, "step": 9808 }, { "epoch": 0.89, "learning_rate": 1.6089870039204675e-06, "loss": 1.2179, "step": 9809 }, { "epoch": 0.89, "learning_rate": 1.6064577561467564e-06, "loss": 1.2021, "step": 9810 }, { "epoch": 0.89, "learning_rate": 1.6039304318707926e-06, "loss": 1.2292, "step": 9811 }, { "epoch": 0.89, "learning_rate": 1.6014050313003847e-06, "loss": 1.1542, "step": 9812 }, { "epoch": 0.89, "learning_rate": 1.5988815546431808e-06, "loss": 1.2043, "step": 9813 }, { "epoch": 0.89, "learning_rate": 1.5963600021066699e-06, "loss": 1.1893, "step": 9814 }, { "epoch": 0.89, "learning_rate": 1.593840373898184e-06, "loss": 1.218, "step": 9815 }, { "epoch": 0.89, "learning_rate": 1.591322670224893e-06, "loss": 1.1947, "step": 9816 }, { "epoch": 0.89, "learning_rate": 1.5888068912938063e-06, "loss": 1.2446, "step": 9817 }, { "epoch": 0.89, "learning_rate": 1.5862930373117897e-06, "loss": 1.2384, "step": 9818 }, { "epoch": 0.89, "learning_rate": 1.5837811084855352e-06, "loss": 1.1951, "step": 9819 }, { "epoch": 0.89, "learning_rate": 1.581271105021584e-06, "loss": 1.1959, "step": 9820 }, { "epoch": 0.89, "learning_rate": 1.5787630271263204e-06, "loss": 1.2497, "step": 9821 }, { "epoch": 0.89, "learning_rate": 1.5762568750059604e-06, "loss": 1.1399, "step": 9822 }, { "epoch": 0.89, "learning_rate": 1.5737526488665666e-06, "loss": 1.133, "step": 9823 }, { "epoch": 0.89, "learning_rate": 1.5712503489140578e-06, "loss": 1.1612, "step": 9824 }, { "epoch": 0.89, "learning_rate": 1.5687499753541722e-06, "loss": 1.1974, "step": 9825 }, { "epoch": 0.89, "learning_rate": 1.5662515283925012e-06, "loss": 1.1995, "step": 9826 }, { "epoch": 0.89, "learning_rate": 1.5637550082344772e-06, "loss": 1.2458, "step": 9827 }, { "epoch": 0.89, "learning_rate": 1.5612604150853638e-06, "loss": 1.224, "step": 9828 }, { "epoch": 0.89, "learning_rate": 1.5587677491502889e-06, "loss": 1.1983, "step": 9829 }, { "epoch": 0.89, "learning_rate": 1.5562770106342022e-06, "loss": 1.2604, "step": 9830 }, { "epoch": 0.89, "learning_rate": 1.5537881997418985e-06, "loss": 1.1915, "step": 9831 }, { "epoch": 0.89, "learning_rate": 1.5513013166780194e-06, "loss": 1.192, "step": 9832 }, { "epoch": 0.89, "learning_rate": 1.5488163616470403e-06, "loss": 1.1947, "step": 9833 }, { "epoch": 0.89, "learning_rate": 1.5463333348532893e-06, "loss": 1.2094, "step": 9834 }, { "epoch": 0.89, "learning_rate": 1.5438522365009257e-06, "loss": 1.177, "step": 9835 }, { "epoch": 0.89, "learning_rate": 1.5413730667939552e-06, "loss": 1.2098, "step": 9836 }, { "epoch": 0.89, "learning_rate": 1.5388958259362236e-06, "loss": 1.2089, "step": 9837 }, { "epoch": 0.89, "learning_rate": 1.536420514131412e-06, "loss": 1.1754, "step": 9838 }, { "epoch": 0.89, "learning_rate": 1.5339471315830607e-06, "loss": 1.1839, "step": 9839 }, { "epoch": 0.89, "learning_rate": 1.5314756784945317e-06, "loss": 1.1745, "step": 9840 }, { "epoch": 0.89, "learning_rate": 1.5290061550690405e-06, "loss": 1.1695, "step": 9841 }, { "epoch": 0.89, "learning_rate": 1.5265385615096383e-06, "loss": 1.2573, "step": 9842 }, { "epoch": 0.89, "learning_rate": 1.5240728980192159e-06, "loss": 1.1845, "step": 9843 }, { "epoch": 0.89, "learning_rate": 1.521609164800511e-06, "loss": 1.2134, "step": 9844 }, { "epoch": 0.89, "learning_rate": 1.519147362056103e-06, "loss": 1.152, "step": 9845 }, { "epoch": 0.89, "learning_rate": 1.5166874899884053e-06, "loss": 1.2151, "step": 9846 }, { "epoch": 0.89, "learning_rate": 1.5142295487996839e-06, "loss": 1.1782, "step": 9847 }, { "epoch": 0.89, "learning_rate": 1.5117735386920323e-06, "loss": 1.1756, "step": 9848 }, { "epoch": 0.89, "learning_rate": 1.5093194598673893e-06, "loss": 1.2814, "step": 9849 }, { "epoch": 0.89, "learning_rate": 1.506867312527549e-06, "loss": 1.2245, "step": 9850 }, { "epoch": 0.89, "learning_rate": 1.5044170968741278e-06, "loss": 1.1877, "step": 9851 }, { "epoch": 0.89, "learning_rate": 1.5019688131085924e-06, "loss": 1.1923, "step": 9852 }, { "epoch": 0.89, "learning_rate": 1.4995224614322484e-06, "loss": 1.1559, "step": 9853 }, { "epoch": 0.89, "learning_rate": 1.4970780420462404e-06, "loss": 1.207, "step": 9854 }, { "epoch": 0.89, "learning_rate": 1.4946355551515661e-06, "loss": 1.1667, "step": 9855 }, { "epoch": 0.89, "learning_rate": 1.4921950009490483e-06, "loss": 1.2154, "step": 9856 }, { "epoch": 0.89, "learning_rate": 1.4897563796393593e-06, "loss": 1.1849, "step": 9857 }, { "epoch": 0.89, "learning_rate": 1.4873196914230113e-06, "loss": 1.2517, "step": 9858 }, { "epoch": 0.89, "learning_rate": 1.4848849365003525e-06, "loss": 1.132, "step": 9859 }, { "epoch": 0.89, "learning_rate": 1.4824521150715864e-06, "loss": 1.189, "step": 9860 }, { "epoch": 0.89, "learning_rate": 1.4800212273367393e-06, "loss": 1.1949, "step": 9861 }, { "epoch": 0.89, "learning_rate": 1.4775922734956927e-06, "loss": 1.2283, "step": 9862 }, { "epoch": 0.89, "learning_rate": 1.4751652537481619e-06, "loss": 1.1914, "step": 9863 }, { "epoch": 0.89, "learning_rate": 1.4727401682937041e-06, "loss": 1.1833, "step": 9864 }, { "epoch": 0.89, "learning_rate": 1.4703170173317126e-06, "loss": 1.2117, "step": 9865 }, { "epoch": 0.89, "learning_rate": 1.4678958010614392e-06, "loss": 1.2125, "step": 9866 }, { "epoch": 0.89, "learning_rate": 1.4654765196819575e-06, "loss": 1.1901, "step": 9867 }, { "epoch": 0.89, "learning_rate": 1.4630591733921894e-06, "loss": 1.1689, "step": 9868 }, { "epoch": 0.89, "learning_rate": 1.4606437623909008e-06, "loss": 1.1698, "step": 9869 }, { "epoch": 0.89, "learning_rate": 1.4582302868766857e-06, "loss": 1.1196, "step": 9870 }, { "epoch": 0.89, "learning_rate": 1.4558187470480016e-06, "loss": 1.2254, "step": 9871 }, { "epoch": 0.89, "learning_rate": 1.4534091431031266e-06, "loss": 1.1456, "step": 9872 }, { "epoch": 0.89, "learning_rate": 1.4510014752401852e-06, "loss": 1.178, "step": 9873 }, { "epoch": 0.89, "learning_rate": 1.4485957436571467e-06, "loss": 1.1653, "step": 9874 }, { "epoch": 0.89, "learning_rate": 1.4461919485518172e-06, "loss": 1.1956, "step": 9875 }, { "epoch": 0.89, "learning_rate": 1.4437900901218464e-06, "loss": 1.2039, "step": 9876 }, { "epoch": 0.89, "learning_rate": 1.4413901685647268e-06, "loss": 1.2364, "step": 9877 }, { "epoch": 0.89, "learning_rate": 1.4389921840777808e-06, "loss": 1.1764, "step": 9878 }, { "epoch": 0.89, "learning_rate": 1.4365961368581842e-06, "loss": 1.1788, "step": 9879 }, { "epoch": 0.89, "learning_rate": 1.4342020271029433e-06, "loss": 1.2547, "step": 9880 }, { "epoch": 0.89, "learning_rate": 1.431809855008917e-06, "loss": 1.2081, "step": 9881 }, { "epoch": 0.89, "learning_rate": 1.4294196207727929e-06, "loss": 1.2449, "step": 9882 }, { "epoch": 0.89, "learning_rate": 1.4270313245911076e-06, "loss": 1.1235, "step": 9883 }, { "epoch": 0.89, "learning_rate": 1.4246449666602325e-06, "loss": 1.1857, "step": 9884 }, { "epoch": 0.89, "learning_rate": 1.4222605471763822e-06, "loss": 1.1749, "step": 9885 }, { "epoch": 0.89, "learning_rate": 1.4198780663356114e-06, "loss": 1.2069, "step": 9886 }, { "epoch": 0.89, "learning_rate": 1.417497524333819e-06, "loss": 1.2257, "step": 9887 }, { "epoch": 0.89, "learning_rate": 1.4151189213667398e-06, "loss": 1.2226, "step": 9888 }, { "epoch": 0.89, "learning_rate": 1.4127422576299483e-06, "loss": 1.2353, "step": 9889 }, { "epoch": 0.89, "learning_rate": 1.4103675333188682e-06, "loss": 1.1991, "step": 9890 }, { "epoch": 0.89, "learning_rate": 1.4079947486287465e-06, "loss": 1.2033, "step": 9891 }, { "epoch": 0.89, "learning_rate": 1.4056239037546936e-06, "loss": 1.2303, "step": 9892 }, { "epoch": 0.89, "learning_rate": 1.4032549988916428e-06, "loss": 1.217, "step": 9893 }, { "epoch": 0.89, "learning_rate": 1.4008880342343766e-06, "loss": 1.2156, "step": 9894 }, { "epoch": 0.89, "learning_rate": 1.3985230099775092e-06, "loss": 1.2496, "step": 9895 }, { "epoch": 0.89, "learning_rate": 1.3961599263155045e-06, "loss": 1.2743, "step": 9896 }, { "epoch": 0.89, "learning_rate": 1.3937987834426652e-06, "loss": 1.1931, "step": 9897 }, { "epoch": 0.89, "learning_rate": 1.3914395815531333e-06, "loss": 1.2176, "step": 9898 }, { "epoch": 0.89, "learning_rate": 1.389082320840887e-06, "loss": 1.225, "step": 9899 }, { "epoch": 0.89, "learning_rate": 1.386727001499749e-06, "loss": 1.1551, "step": 9900 }, { "epoch": 0.89, "learning_rate": 1.3843736237233784e-06, "loss": 1.1463, "step": 9901 }, { "epoch": 0.89, "learning_rate": 1.3820221877052868e-06, "loss": 1.1996, "step": 9902 }, { "epoch": 0.89, "learning_rate": 1.3796726936388115e-06, "loss": 1.2521, "step": 9903 }, { "epoch": 0.89, "learning_rate": 1.3773251417171396e-06, "loss": 1.2349, "step": 9904 }, { "epoch": 0.89, "learning_rate": 1.3749795321332887e-06, "loss": 1.1655, "step": 9905 }, { "epoch": 0.9, "learning_rate": 1.3726358650801297e-06, "loss": 1.1345, "step": 9906 }, { "epoch": 0.9, "learning_rate": 1.3702941407503583e-06, "loss": 1.2133, "step": 9907 }, { "epoch": 0.9, "learning_rate": 1.367954359336532e-06, "loss": 1.2439, "step": 9908 }, { "epoch": 0.9, "learning_rate": 1.365616521031024e-06, "loss": 1.1857, "step": 9909 }, { "epoch": 0.9, "learning_rate": 1.3632806260260672e-06, "loss": 1.1307, "step": 9910 }, { "epoch": 0.9, "learning_rate": 1.3609466745137217e-06, "loss": 1.2439, "step": 9911 }, { "epoch": 0.9, "learning_rate": 1.3586146666858924e-06, "loss": 1.2066, "step": 9912 }, { "epoch": 0.9, "learning_rate": 1.3562846027343313e-06, "loss": 1.2402, "step": 9913 }, { "epoch": 0.9, "learning_rate": 1.3539564828506212e-06, "loss": 1.2247, "step": 9914 }, { "epoch": 0.9, "learning_rate": 1.351630307226187e-06, "loss": 1.2412, "step": 9915 }, { "epoch": 0.9, "learning_rate": 1.3493060760522951e-06, "loss": 1.208, "step": 9916 }, { "epoch": 0.9, "learning_rate": 1.346983789520051e-06, "loss": 1.1261, "step": 9917 }, { "epoch": 0.9, "learning_rate": 1.3446634478204044e-06, "loss": 1.2201, "step": 9918 }, { "epoch": 0.9, "learning_rate": 1.3423450511441393e-06, "loss": 1.2198, "step": 9919 }, { "epoch": 0.9, "learning_rate": 1.3400285996818808e-06, "loss": 1.1419, "step": 9920 }, { "epoch": 0.9, "learning_rate": 1.3377140936241012e-06, "loss": 1.2064, "step": 9921 }, { "epoch": 0.9, "learning_rate": 1.3354015331610958e-06, "loss": 1.2215, "step": 9922 }, { "epoch": 0.9, "learning_rate": 1.3330909184830265e-06, "loss": 1.2258, "step": 9923 }, { "epoch": 0.9, "learning_rate": 1.3307822497798689e-06, "loss": 1.19, "step": 9924 }, { "epoch": 0.9, "learning_rate": 1.3284755272414572e-06, "loss": 1.1776, "step": 9925 }, { "epoch": 0.9, "learning_rate": 1.326170751057451e-06, "loss": 1.1777, "step": 9926 }, { "epoch": 0.9, "learning_rate": 1.32386792141736e-06, "loss": 1.1583, "step": 9927 }, { "epoch": 0.9, "learning_rate": 1.3215670385105295e-06, "loss": 1.1691, "step": 9928 }, { "epoch": 0.9, "learning_rate": 1.31926810252615e-06, "loss": 1.2017, "step": 9929 }, { "epoch": 0.9, "learning_rate": 1.3169711136532453e-06, "loss": 1.2218, "step": 9930 }, { "epoch": 0.9, "learning_rate": 1.314676072080681e-06, "loss": 1.1706, "step": 9931 }, { "epoch": 0.9, "learning_rate": 1.3123829779971648e-06, "loss": 1.2108, "step": 9932 }, { "epoch": 0.9, "learning_rate": 1.310091831591237e-06, "loss": 1.2172, "step": 9933 }, { "epoch": 0.9, "learning_rate": 1.3078026330512943e-06, "loss": 1.2061, "step": 9934 }, { "epoch": 0.9, "learning_rate": 1.3055153825655558e-06, "loss": 1.1569, "step": 9935 }, { "epoch": 0.9, "learning_rate": 1.3032300803220876e-06, "loss": 1.2008, "step": 9936 }, { "epoch": 0.9, "learning_rate": 1.3009467265087977e-06, "loss": 1.2631, "step": 9937 }, { "epoch": 0.9, "learning_rate": 1.2986653213134247e-06, "loss": 1.1723, "step": 9938 }, { "epoch": 0.9, "learning_rate": 1.2963858649235634e-06, "loss": 1.2552, "step": 9939 }, { "epoch": 0.9, "learning_rate": 1.29410835752663e-06, "loss": 1.1619, "step": 9940 }, { "epoch": 0.9, "learning_rate": 1.2918327993098943e-06, "loss": 1.1743, "step": 9941 }, { "epoch": 0.9, "learning_rate": 1.28955919046046e-06, "loss": 1.1629, "step": 9942 }, { "epoch": 0.9, "learning_rate": 1.2872875311652654e-06, "loss": 1.1823, "step": 9943 }, { "epoch": 0.9, "learning_rate": 1.2850178216111007e-06, "loss": 1.2284, "step": 9944 }, { "epoch": 0.9, "learning_rate": 1.2827500619845918e-06, "loss": 1.1891, "step": 9945 }, { "epoch": 0.9, "learning_rate": 1.2804842524721915e-06, "loss": 1.1737, "step": 9946 }, { "epoch": 0.9, "learning_rate": 1.2782203932602099e-06, "loss": 1.2243, "step": 9947 }, { "epoch": 0.9, "learning_rate": 1.2759584845347866e-06, "loss": 1.1874, "step": 9948 }, { "epoch": 0.9, "learning_rate": 1.2736985264819007e-06, "loss": 1.2282, "step": 9949 }, { "epoch": 0.9, "learning_rate": 1.271440519287384e-06, "loss": 1.1892, "step": 9950 }, { "epoch": 0.9, "learning_rate": 1.2691844631368883e-06, "loss": 1.1585, "step": 9951 }, { "epoch": 0.9, "learning_rate": 1.2669303582159175e-06, "loss": 1.1931, "step": 9952 }, { "epoch": 0.9, "learning_rate": 1.2646782047098126e-06, "loss": 1.203, "step": 9953 }, { "epoch": 0.9, "learning_rate": 1.2624280028037478e-06, "loss": 1.191, "step": 9954 }, { "epoch": 0.9, "learning_rate": 1.2601797526827497e-06, "loss": 1.2187, "step": 9955 }, { "epoch": 0.9, "learning_rate": 1.2579334545316733e-06, "loss": 1.1732, "step": 9956 }, { "epoch": 0.9, "learning_rate": 1.2556891085352213e-06, "loss": 1.1909, "step": 9957 }, { "epoch": 0.9, "learning_rate": 1.2534467148779261e-06, "loss": 1.227, "step": 9958 }, { "epoch": 0.9, "learning_rate": 1.2512062737441654e-06, "loss": 1.1622, "step": 9959 }, { "epoch": 0.9, "learning_rate": 1.2489677853181614e-06, "loss": 1.245, "step": 9960 }, { "epoch": 0.9, "learning_rate": 1.246731249783964e-06, "loss": 1.2684, "step": 9961 }, { "epoch": 0.9, "learning_rate": 1.244496667325476e-06, "loss": 1.1534, "step": 9962 }, { "epoch": 0.9, "learning_rate": 1.2422640381264255e-06, "loss": 1.1339, "step": 9963 }, { "epoch": 0.9, "learning_rate": 1.2400333623703852e-06, "loss": 1.2002, "step": 9964 }, { "epoch": 0.9, "learning_rate": 1.2378046402407805e-06, "loss": 1.1807, "step": 9965 }, { "epoch": 0.9, "learning_rate": 1.2355778719208567e-06, "loss": 1.1596, "step": 9966 }, { "epoch": 0.9, "learning_rate": 1.233353057593703e-06, "loss": 1.228, "step": 9967 }, { "epoch": 0.9, "learning_rate": 1.2311301974422595e-06, "loss": 1.1804, "step": 9968 }, { "epoch": 0.9, "learning_rate": 1.2289092916492911e-06, "loss": 1.2913, "step": 9969 }, { "epoch": 0.9, "learning_rate": 1.22669034039741e-06, "loss": 1.1677, "step": 9970 }, { "epoch": 0.9, "learning_rate": 1.224473343869073e-06, "loss": 1.1924, "step": 9971 }, { "epoch": 0.9, "learning_rate": 1.222258302246554e-06, "loss": 1.217, "step": 9972 }, { "epoch": 0.9, "learning_rate": 1.2200452157119958e-06, "loss": 1.1861, "step": 9973 }, { "epoch": 0.9, "learning_rate": 1.2178340844473585e-06, "loss": 1.2123, "step": 9974 }, { "epoch": 0.9, "learning_rate": 1.2156249086344495e-06, "loss": 1.1477, "step": 9975 }, { "epoch": 0.9, "learning_rate": 1.213417688454918e-06, "loss": 1.2102, "step": 9976 }, { "epoch": 0.9, "learning_rate": 1.2112124240902496e-06, "loss": 1.1688, "step": 9977 }, { "epoch": 0.9, "learning_rate": 1.2090091157217654e-06, "loss": 1.212, "step": 9978 }, { "epoch": 0.9, "learning_rate": 1.2068077635306318e-06, "loss": 1.2007, "step": 9979 }, { "epoch": 0.9, "learning_rate": 1.2046083676978432e-06, "loss": 1.1666, "step": 9980 }, { "epoch": 0.9, "learning_rate": 1.2024109284042545e-06, "loss": 1.1562, "step": 9981 }, { "epoch": 0.9, "learning_rate": 1.200215445830541e-06, "loss": 1.1991, "step": 9982 }, { "epoch": 0.9, "learning_rate": 1.1980219201572217e-06, "loss": 1.2808, "step": 9983 }, { "epoch": 0.9, "learning_rate": 1.1958303515646558e-06, "loss": 1.2136, "step": 9984 }, { "epoch": 0.9, "learning_rate": 1.1936407402330402e-06, "loss": 1.2418, "step": 9985 }, { "epoch": 0.9, "learning_rate": 1.1914530863424173e-06, "loss": 1.1813, "step": 9986 }, { "epoch": 0.9, "learning_rate": 1.1892673900726653e-06, "loss": 1.1998, "step": 9987 }, { "epoch": 0.9, "learning_rate": 1.1870836516034878e-06, "loss": 1.2232, "step": 9988 }, { "epoch": 0.9, "learning_rate": 1.1849018711144494e-06, "loss": 1.182, "step": 9989 }, { "epoch": 0.9, "learning_rate": 1.1827220487849427e-06, "loss": 1.1675, "step": 9990 }, { "epoch": 0.9, "learning_rate": 1.180544184794194e-06, "loss": 1.1681, "step": 9991 }, { "epoch": 0.9, "learning_rate": 1.1783682793212847e-06, "loss": 1.2037, "step": 9992 }, { "epoch": 0.9, "learning_rate": 1.1761943325451164e-06, "loss": 1.1679, "step": 9993 }, { "epoch": 0.9, "learning_rate": 1.1740223446444433e-06, "loss": 1.1884, "step": 9994 }, { "epoch": 0.9, "learning_rate": 1.17185231579785e-06, "loss": 1.1764, "step": 9995 }, { "epoch": 0.9, "learning_rate": 1.1696842461837666e-06, "loss": 1.1898, "step": 9996 }, { "epoch": 0.9, "learning_rate": 1.1675181359804638e-06, "loss": 1.1779, "step": 9997 }, { "epoch": 0.9, "learning_rate": 1.1653539853660327e-06, "loss": 1.1632, "step": 9998 }, { "epoch": 0.9, "learning_rate": 1.163191794518434e-06, "loss": 1.2123, "step": 9999 }, { "epoch": 0.9, "learning_rate": 1.161031563615439e-06, "loss": 1.1628, "step": 10000 }, { "epoch": 0.9, "learning_rate": 1.1588732928346723e-06, "loss": 1.1922, "step": 10001 }, { "epoch": 0.9, "learning_rate": 1.156716982353595e-06, "loss": 1.1962, "step": 10002 }, { "epoch": 0.9, "learning_rate": 1.1545626323495096e-06, "loss": 1.1737, "step": 10003 }, { "epoch": 0.9, "learning_rate": 1.1524102429995493e-06, "loss": 1.2215, "step": 10004 }, { "epoch": 0.9, "learning_rate": 1.1502598144806947e-06, "loss": 1.2529, "step": 10005 }, { "epoch": 0.9, "learning_rate": 1.1481113469697546e-06, "loss": 1.1596, "step": 10006 }, { "epoch": 0.9, "learning_rate": 1.1459648406433904e-06, "loss": 1.1835, "step": 10007 }, { "epoch": 0.9, "learning_rate": 1.1438202956780996e-06, "loss": 1.2099, "step": 10008 }, { "epoch": 0.9, "learning_rate": 1.1416777122501998e-06, "loss": 1.1794, "step": 10009 }, { "epoch": 0.9, "learning_rate": 1.1395370905358721e-06, "loss": 1.1969, "step": 10010 }, { "epoch": 0.9, "learning_rate": 1.137398430711123e-06, "loss": 1.2184, "step": 10011 }, { "epoch": 0.9, "learning_rate": 1.135261732951795e-06, "loss": 1.2312, "step": 10012 }, { "epoch": 0.9, "learning_rate": 1.1331269974335896e-06, "loss": 1.2647, "step": 10013 }, { "epoch": 0.9, "learning_rate": 1.130994224332016e-06, "loss": 1.1821, "step": 10014 }, { "epoch": 0.9, "learning_rate": 1.128863413822448e-06, "loss": 1.1607, "step": 10015 }, { "epoch": 0.9, "learning_rate": 1.1267345660800871e-06, "loss": 1.1898, "step": 10016 }, { "epoch": 0.91, "learning_rate": 1.1246076812799683e-06, "loss": 1.173, "step": 10017 }, { "epoch": 0.91, "learning_rate": 1.1224827595969823e-06, "loss": 1.249, "step": 10018 }, { "epoch": 0.91, "learning_rate": 1.1203598012058335e-06, "loss": 1.186, "step": 10019 }, { "epoch": 0.91, "learning_rate": 1.1182388062810884e-06, "loss": 1.1844, "step": 10020 }, { "epoch": 0.91, "learning_rate": 1.116119774997143e-06, "loss": 1.2192, "step": 10021 }, { "epoch": 0.91, "learning_rate": 1.114002707528225e-06, "loss": 1.2839, "step": 10022 }, { "epoch": 0.91, "learning_rate": 1.1118876040484173e-06, "loss": 1.2033, "step": 10023 }, { "epoch": 0.91, "learning_rate": 1.1097744647316165e-06, "loss": 1.1902, "step": 10024 }, { "epoch": 0.91, "learning_rate": 1.1076632897515843e-06, "loss": 1.1365, "step": 10025 }, { "epoch": 0.91, "learning_rate": 1.1055540792819036e-06, "loss": 1.1589, "step": 10026 }, { "epoch": 0.91, "learning_rate": 1.1034468334960024e-06, "loss": 1.1767, "step": 10027 }, { "epoch": 0.91, "learning_rate": 1.101341552567145e-06, "loss": 1.1994, "step": 10028 }, { "epoch": 0.91, "learning_rate": 1.09923823666844e-06, "loss": 1.2764, "step": 10029 }, { "epoch": 0.91, "learning_rate": 1.0971368859728187e-06, "loss": 1.1566, "step": 10030 }, { "epoch": 0.91, "learning_rate": 1.0950375006530677e-06, "loss": 1.1991, "step": 10031 }, { "epoch": 0.91, "learning_rate": 1.0929400808818075e-06, "loss": 1.1831, "step": 10032 }, { "epoch": 0.91, "learning_rate": 1.0908446268314887e-06, "loss": 1.1937, "step": 10033 }, { "epoch": 0.91, "learning_rate": 1.088751138674418e-06, "loss": 1.2254, "step": 10034 }, { "epoch": 0.91, "learning_rate": 1.0866596165827165e-06, "loss": 1.1961, "step": 10035 }, { "epoch": 0.91, "learning_rate": 1.0845700607283626e-06, "loss": 1.2071, "step": 10036 }, { "epoch": 0.91, "learning_rate": 1.0824824712831693e-06, "loss": 1.2288, "step": 10037 }, { "epoch": 0.91, "learning_rate": 1.0803968484187772e-06, "loss": 1.1974, "step": 10038 }, { "epoch": 0.91, "learning_rate": 1.0783131923066847e-06, "loss": 1.1881, "step": 10039 }, { "epoch": 0.91, "learning_rate": 1.0762315031182051e-06, "loss": 1.2447, "step": 10040 }, { "epoch": 0.91, "learning_rate": 1.0741517810245128e-06, "loss": 1.2429, "step": 10041 }, { "epoch": 0.91, "learning_rate": 1.0720740261966038e-06, "loss": 1.2029, "step": 10042 }, { "epoch": 0.91, "learning_rate": 1.069998238805317e-06, "loss": 1.1801, "step": 10043 }, { "epoch": 0.91, "learning_rate": 1.0679244190213378e-06, "loss": 1.1745, "step": 10044 }, { "epoch": 0.91, "learning_rate": 1.0658525670151714e-06, "loss": 1.1736, "step": 10045 }, { "epoch": 0.91, "learning_rate": 1.0637826829571844e-06, "loss": 1.1616, "step": 10046 }, { "epoch": 0.91, "learning_rate": 1.0617147670175658e-06, "loss": 1.1787, "step": 10047 }, { "epoch": 0.91, "learning_rate": 1.0596488193663406e-06, "loss": 1.1696, "step": 10048 }, { "epoch": 0.91, "learning_rate": 1.0575848401733894e-06, "loss": 1.1866, "step": 10049 }, { "epoch": 0.91, "learning_rate": 1.055522829608413e-06, "loss": 1.2465, "step": 10050 }, { "epoch": 0.91, "learning_rate": 1.0534627878409531e-06, "loss": 1.1719, "step": 10051 }, { "epoch": 0.91, "learning_rate": 1.0514047150403993e-06, "loss": 1.1641, "step": 10052 }, { "epoch": 0.91, "learning_rate": 1.0493486113759747e-06, "loss": 1.2195, "step": 10053 }, { "epoch": 0.91, "learning_rate": 1.047294477016733e-06, "loss": 1.2163, "step": 10054 }, { "epoch": 0.91, "learning_rate": 1.0452423121315835e-06, "loss": 1.239, "step": 10055 }, { "epoch": 0.91, "learning_rate": 1.0431921168892468e-06, "loss": 1.2506, "step": 10056 }, { "epoch": 0.91, "learning_rate": 1.0411438914583072e-06, "loss": 1.2506, "step": 10057 }, { "epoch": 0.91, "learning_rate": 1.0390976360071746e-06, "loss": 1.1786, "step": 10058 }, { "epoch": 0.91, "learning_rate": 1.037053350704098e-06, "loss": 1.2217, "step": 10059 }, { "epoch": 0.91, "learning_rate": 1.0350110357171732e-06, "loss": 1.1993, "step": 10060 }, { "epoch": 0.91, "learning_rate": 1.0329706912143133e-06, "loss": 1.1873, "step": 10061 }, { "epoch": 0.91, "learning_rate": 1.0309323173632896e-06, "loss": 1.1929, "step": 10062 }, { "epoch": 0.91, "learning_rate": 1.0288959143317068e-06, "loss": 1.19, "step": 10063 }, { "epoch": 0.91, "learning_rate": 1.0268614822869977e-06, "loss": 1.2013, "step": 10064 }, { "epoch": 0.91, "learning_rate": 1.0248290213964506e-06, "loss": 1.2315, "step": 10065 }, { "epoch": 0.91, "learning_rate": 1.0227985318271682e-06, "loss": 1.1532, "step": 10066 }, { "epoch": 0.91, "learning_rate": 1.0207700137461168e-06, "loss": 1.2172, "step": 10067 }, { "epoch": 0.91, "learning_rate": 1.0187434673200824e-06, "loss": 1.1684, "step": 10068 }, { "epoch": 0.91, "learning_rate": 1.0167188927156902e-06, "loss": 1.1924, "step": 10069 }, { "epoch": 0.91, "learning_rate": 1.0146962900994206e-06, "loss": 1.1725, "step": 10070 }, { "epoch": 0.91, "learning_rate": 1.0126756596375686e-06, "loss": 1.1997, "step": 10071 }, { "epoch": 0.91, "learning_rate": 1.0106570014962763e-06, "loss": 1.1838, "step": 10072 }, { "epoch": 0.91, "learning_rate": 1.0086403158415302e-06, "loss": 1.1591, "step": 10073 }, { "epoch": 0.91, "learning_rate": 1.006625602839148e-06, "loss": 1.152, "step": 10074 }, { "epoch": 0.91, "learning_rate": 1.0046128626547801e-06, "loss": 1.1369, "step": 10075 }, { "epoch": 0.91, "learning_rate": 1.0026020954539362e-06, "loss": 1.192, "step": 10076 }, { "epoch": 0.91, "learning_rate": 1.0005933014019308e-06, "loss": 1.183, "step": 10077 }, { "epoch": 0.91, "learning_rate": 9.98586480663946e-07, "loss": 1.2363, "step": 10078 }, { "epoch": 0.91, "learning_rate": 9.965816334049833e-07, "loss": 1.2003, "step": 10079 }, { "epoch": 0.91, "learning_rate": 9.945787597898882e-07, "loss": 1.2358, "step": 10080 }, { "epoch": 0.91, "learning_rate": 9.925778599833513e-07, "loss": 1.1758, "step": 10081 }, { "epoch": 0.91, "learning_rate": 9.905789341498828e-07, "loss": 1.2354, "step": 10082 }, { "epoch": 0.91, "learning_rate": 9.88581982453851e-07, "loss": 1.1926, "step": 10083 }, { "epoch": 0.91, "learning_rate": 9.865870050594473e-07, "loss": 1.148, "step": 10084 }, { "epoch": 0.91, "learning_rate": 9.84594002130701e-07, "loss": 1.193, "step": 10085 }, { "epoch": 0.91, "learning_rate": 9.826029738314952e-07, "loss": 1.1468, "step": 10086 }, { "epoch": 0.91, "learning_rate": 9.806139203255265e-07, "loss": 1.1848, "step": 10087 }, { "epoch": 0.91, "learning_rate": 9.786268417763505e-07, "loss": 1.1968, "step": 10088 }, { "epoch": 0.91, "learning_rate": 9.766417383473503e-07, "loss": 1.1993, "step": 10089 }, { "epoch": 0.91, "learning_rate": 9.746586102017452e-07, "loss": 1.2049, "step": 10090 }, { "epoch": 0.91, "learning_rate": 9.726774575025994e-07, "loss": 1.1971, "step": 10091 }, { "epoch": 0.91, "learning_rate": 9.706982804128045e-07, "loss": 1.2369, "step": 10092 }, { "epoch": 0.91, "learning_rate": 9.687210790950945e-07, "loss": 1.1791, "step": 10093 }, { "epoch": 0.91, "learning_rate": 9.667458537120506e-07, "loss": 1.2275, "step": 10094 }, { "epoch": 0.91, "learning_rate": 9.64772604426073e-07, "loss": 1.2331, "step": 10095 }, { "epoch": 0.91, "learning_rate": 9.628013313994156e-07, "loss": 1.2437, "step": 10096 }, { "epoch": 0.91, "learning_rate": 9.608320347941596e-07, "loss": 1.222, "step": 10097 }, { "epoch": 0.91, "learning_rate": 9.588647147722284e-07, "loss": 1.1686, "step": 10098 }, { "epoch": 0.91, "learning_rate": 9.568993714953818e-07, "loss": 1.2262, "step": 10099 }, { "epoch": 0.91, "learning_rate": 9.549360051252177e-07, "loss": 1.1701, "step": 10100 }, { "epoch": 0.91, "learning_rate": 9.529746158231684e-07, "loss": 1.1969, "step": 10101 }, { "epoch": 0.91, "learning_rate": 9.510152037505132e-07, "loss": 1.1867, "step": 10102 }, { "epoch": 0.91, "learning_rate": 9.490577690683511e-07, "loss": 1.1899, "step": 10103 }, { "epoch": 0.91, "learning_rate": 9.471023119376393e-07, "loss": 1.2131, "step": 10104 }, { "epoch": 0.91, "learning_rate": 9.451488325191548e-07, "loss": 1.2259, "step": 10105 }, { "epoch": 0.91, "learning_rate": 9.431973309735193e-07, "loss": 1.1783, "step": 10106 }, { "epoch": 0.91, "learning_rate": 9.412478074612041e-07, "loss": 1.2144, "step": 10107 }, { "epoch": 0.91, "learning_rate": 9.393002621424868e-07, "loss": 1.299, "step": 10108 }, { "epoch": 0.91, "learning_rate": 9.373546951775169e-07, "loss": 1.2367, "step": 10109 }, { "epoch": 0.91, "learning_rate": 9.354111067262584e-07, "loss": 1.2322, "step": 10110 }, { "epoch": 0.91, "learning_rate": 9.334694969485192e-07, "loss": 1.1284, "step": 10111 }, { "epoch": 0.91, "learning_rate": 9.315298660039523e-07, "loss": 1.1813, "step": 10112 }, { "epoch": 0.91, "learning_rate": 9.29592214052033e-07, "loss": 1.1618, "step": 10113 }, { "epoch": 0.91, "learning_rate": 9.276565412520811e-07, "loss": 1.2484, "step": 10114 }, { "epoch": 0.91, "learning_rate": 9.257228477632607e-07, "loss": 1.1509, "step": 10115 }, { "epoch": 0.91, "learning_rate": 9.237911337445643e-07, "loss": 1.1427, "step": 10116 }, { "epoch": 0.91, "learning_rate": 9.218613993548231e-07, "loss": 1.2121, "step": 10117 }, { "epoch": 0.91, "learning_rate": 9.199336447527074e-07, "loss": 1.1802, "step": 10118 }, { "epoch": 0.91, "learning_rate": 9.180078700967181e-07, "loss": 1.1666, "step": 10119 }, { "epoch": 0.91, "learning_rate": 9.160840755452093e-07, "loss": 1.2038, "step": 10120 }, { "epoch": 0.91, "learning_rate": 9.141622612563571e-07, "loss": 1.2047, "step": 10121 }, { "epoch": 0.91, "learning_rate": 9.122424273881796e-07, "loss": 1.2299, "step": 10122 }, { "epoch": 0.91, "learning_rate": 9.103245740985311e-07, "loss": 1.2347, "step": 10123 }, { "epoch": 0.91, "learning_rate": 9.084087015451021e-07, "loss": 1.2209, "step": 10124 }, { "epoch": 0.91, "learning_rate": 9.064948098854304e-07, "loss": 1.1354, "step": 10125 }, { "epoch": 0.91, "learning_rate": 9.045828992768762e-07, "loss": 1.1451, "step": 10126 }, { "epoch": 0.92, "learning_rate": 9.026729698766418e-07, "loss": 1.1607, "step": 10127 }, { "epoch": 0.92, "learning_rate": 9.00765021841779e-07, "loss": 1.2326, "step": 10128 }, { "epoch": 0.92, "learning_rate": 8.988590553291515e-07, "loss": 1.2381, "step": 10129 }, { "epoch": 0.92, "learning_rate": 8.96955070495481e-07, "loss": 1.1519, "step": 10130 }, { "epoch": 0.92, "learning_rate": 8.950530674973229e-07, "loss": 1.208, "step": 10131 }, { "epoch": 0.92, "learning_rate": 8.931530464910604e-07, "loss": 1.2372, "step": 10132 }, { "epoch": 0.92, "learning_rate": 8.912550076329268e-07, "loss": 1.1908, "step": 10133 }, { "epoch": 0.92, "learning_rate": 8.893589510789807e-07, "loss": 1.1745, "step": 10134 }, { "epoch": 0.92, "learning_rate": 8.874648769851196e-07, "loss": 1.1719, "step": 10135 }, { "epoch": 0.92, "learning_rate": 8.855727855070856e-07, "loss": 1.2286, "step": 10136 }, { "epoch": 0.92, "learning_rate": 8.836826768004514e-07, "loss": 1.2306, "step": 10137 }, { "epoch": 0.92, "learning_rate": 8.817945510206316e-07, "loss": 1.19, "step": 10138 }, { "epoch": 0.92, "learning_rate": 8.799084083228687e-07, "loss": 1.176, "step": 10139 }, { "epoch": 0.92, "learning_rate": 8.78024248862247e-07, "loss": 1.1861, "step": 10140 }, { "epoch": 0.92, "learning_rate": 8.76142072793698e-07, "loss": 1.2092, "step": 10141 }, { "epoch": 0.92, "learning_rate": 8.742618802719732e-07, "loss": 1.1632, "step": 10142 }, { "epoch": 0.92, "learning_rate": 8.723836714516681e-07, "loss": 1.2305, "step": 10143 }, { "epoch": 0.92, "learning_rate": 8.705074464872204e-07, "loss": 1.2197, "step": 10144 }, { "epoch": 0.92, "learning_rate": 8.686332055328928e-07, "loss": 1.2083, "step": 10145 }, { "epoch": 0.92, "learning_rate": 8.66760948742798e-07, "loss": 1.2075, "step": 10146 }, { "epoch": 0.92, "learning_rate": 8.648906762708797e-07, "loss": 1.2012, "step": 10147 }, { "epoch": 0.92, "learning_rate": 8.630223882709121e-07, "loss": 1.188, "step": 10148 }, { "epoch": 0.92, "learning_rate": 8.611560848965194e-07, "loss": 1.2131, "step": 10149 }, { "epoch": 0.92, "learning_rate": 8.592917663011513e-07, "loss": 1.2292, "step": 10150 }, { "epoch": 0.92, "learning_rate": 8.57429432638096e-07, "loss": 1.2222, "step": 10151 }, { "epoch": 0.92, "learning_rate": 8.555690840604868e-07, "loss": 1.2017, "step": 10152 }, { "epoch": 0.92, "learning_rate": 8.537107207212847e-07, "loss": 1.2451, "step": 10153 }, { "epoch": 0.92, "learning_rate": 8.51854342773295e-07, "loss": 1.1719, "step": 10154 }, { "epoch": 0.92, "learning_rate": 8.499999503691485e-07, "loss": 1.1665, "step": 10155 }, { "epoch": 0.92, "learning_rate": 8.481475436613235e-07, "loss": 1.1909, "step": 10156 }, { "epoch": 0.92, "learning_rate": 8.46297122802131e-07, "loss": 1.1832, "step": 10157 }, { "epoch": 0.92, "learning_rate": 8.444486879437219e-07, "loss": 1.2172, "step": 10158 }, { "epoch": 0.92, "learning_rate": 8.426022392380744e-07, "loss": 1.1568, "step": 10159 }, { "epoch": 0.92, "learning_rate": 8.407577768370173e-07, "loss": 1.1769, "step": 10160 }, { "epoch": 0.92, "learning_rate": 8.389153008921985e-07, "loss": 1.2046, "step": 10161 }, { "epoch": 0.92, "learning_rate": 8.370748115551246e-07, "loss": 1.1264, "step": 10162 }, { "epoch": 0.92, "learning_rate": 8.352363089771193e-07, "loss": 1.1858, "step": 10163 }, { "epoch": 0.92, "learning_rate": 8.333997933093557e-07, "loss": 1.2051, "step": 10164 }, { "epoch": 0.92, "learning_rate": 8.315652647028327e-07, "loss": 1.1528, "step": 10165 }, { "epoch": 0.92, "learning_rate": 8.297327233083935e-07, "loss": 1.2486, "step": 10166 }, { "epoch": 0.92, "learning_rate": 8.279021692767175e-07, "loss": 1.2124, "step": 10167 }, { "epoch": 0.92, "learning_rate": 8.260736027583205e-07, "loss": 1.1647, "step": 10168 }, { "epoch": 0.92, "learning_rate": 8.242470239035515e-07, "loss": 1.2165, "step": 10169 }, { "epoch": 0.92, "learning_rate": 8.224224328625962e-07, "loss": 1.2247, "step": 10170 }, { "epoch": 0.92, "learning_rate": 8.205998297854844e-07, "loss": 1.2056, "step": 10171 }, { "epoch": 0.92, "learning_rate": 8.187792148220685e-07, "loss": 1.2053, "step": 10172 }, { "epoch": 0.92, "learning_rate": 8.169605881220538e-07, "loss": 1.2188, "step": 10173 }, { "epoch": 0.92, "learning_rate": 8.151439498349678e-07, "loss": 1.1751, "step": 10174 }, { "epoch": 0.92, "learning_rate": 8.133293001101883e-07, "loss": 1.1512, "step": 10175 }, { "epoch": 0.92, "learning_rate": 8.115166390969125e-07, "loss": 1.2132, "step": 10176 }, { "epoch": 0.92, "learning_rate": 8.097059669441909e-07, "loss": 1.1956, "step": 10177 }, { "epoch": 0.92, "learning_rate": 8.078972838009013e-07, "loss": 1.2028, "step": 10178 }, { "epoch": 0.92, "learning_rate": 8.060905898157583e-07, "loss": 1.2132, "step": 10179 }, { "epoch": 0.92, "learning_rate": 8.04285885137318e-07, "loss": 1.183, "step": 10180 }, { "epoch": 0.92, "learning_rate": 8.024831699139645e-07, "loss": 1.2194, "step": 10181 }, { "epoch": 0.92, "learning_rate": 8.006824442939237e-07, "loss": 1.2661, "step": 10182 }, { "epoch": 0.92, "learning_rate": 7.98883708425266e-07, "loss": 1.212, "step": 10183 }, { "epoch": 0.92, "learning_rate": 7.970869624558785e-07, "loss": 1.1968, "step": 10184 }, { "epoch": 0.92, "learning_rate": 7.952922065335045e-07, "loss": 1.1785, "step": 10185 }, { "epoch": 0.92, "learning_rate": 7.93499440805709e-07, "loss": 1.1955, "step": 10186 }, { "epoch": 0.92, "learning_rate": 7.917086654198991e-07, "loss": 1.14, "step": 10187 }, { "epoch": 0.92, "learning_rate": 7.899198805233238e-07, "loss": 1.2127, "step": 10188 }, { "epoch": 0.92, "learning_rate": 7.881330862630598e-07, "loss": 1.1766, "step": 10189 }, { "epoch": 0.92, "learning_rate": 7.863482827860258e-07, "loss": 1.2358, "step": 10190 }, { "epoch": 0.92, "learning_rate": 7.845654702389709e-07, "loss": 1.2036, "step": 10191 }, { "epoch": 0.92, "learning_rate": 7.827846487684892e-07, "loss": 1.1361, "step": 10192 }, { "epoch": 0.92, "learning_rate": 7.810058185209968e-07, "loss": 1.2095, "step": 10193 }, { "epoch": 0.92, "learning_rate": 7.792289796427654e-07, "loss": 1.1633, "step": 10194 }, { "epoch": 0.92, "learning_rate": 7.774541322798867e-07, "loss": 1.236, "step": 10195 }, { "epoch": 0.92, "learning_rate": 7.756812765782995e-07, "loss": 1.1897, "step": 10196 }, { "epoch": 0.92, "learning_rate": 7.739104126837704e-07, "loss": 1.1773, "step": 10197 }, { "epoch": 0.92, "learning_rate": 7.721415407419025e-07, "loss": 1.1502, "step": 10198 }, { "epoch": 0.92, "learning_rate": 7.703746608981488e-07, "loss": 1.1736, "step": 10199 }, { "epoch": 0.92, "learning_rate": 7.686097732977793e-07, "loss": 1.1748, "step": 10200 }, { "epoch": 0.92, "learning_rate": 7.668468780859139e-07, "loss": 1.1983, "step": 10201 }, { "epoch": 0.92, "learning_rate": 7.650859754075007e-07, "loss": 1.1718, "step": 10202 }, { "epoch": 0.92, "learning_rate": 7.633270654073266e-07, "loss": 1.24, "step": 10203 }, { "epoch": 0.92, "learning_rate": 7.615701482300231e-07, "loss": 1.1658, "step": 10204 }, { "epoch": 0.92, "learning_rate": 7.598152240200412e-07, "loss": 1.1975, "step": 10205 }, { "epoch": 0.92, "learning_rate": 7.580622929216796e-07, "loss": 1.1632, "step": 10206 }, { "epoch": 0.92, "learning_rate": 7.563113550790729e-07, "loss": 1.1766, "step": 10207 }, { "epoch": 0.92, "learning_rate": 7.545624106361809e-07, "loss": 1.1744, "step": 10208 }, { "epoch": 0.92, "learning_rate": 7.528154597368192e-07, "loss": 1.1827, "step": 10209 }, { "epoch": 0.92, "learning_rate": 7.510705025246228e-07, "loss": 1.2415, "step": 10210 }, { "epoch": 0.92, "learning_rate": 7.493275391430688e-07, "loss": 1.2205, "step": 10211 }, { "epoch": 0.92, "learning_rate": 7.475865697354673e-07, "loss": 1.1851, "step": 10212 }, { "epoch": 0.92, "learning_rate": 7.458475944449705e-07, "loss": 1.1709, "step": 10213 }, { "epoch": 0.92, "learning_rate": 7.441106134145559e-07, "loss": 1.1951, "step": 10214 }, { "epoch": 0.92, "learning_rate": 7.423756267870536e-07, "loss": 1.1809, "step": 10215 }, { "epoch": 0.92, "learning_rate": 7.40642634705116e-07, "loss": 1.222, "step": 10216 }, { "epoch": 0.92, "learning_rate": 7.38911637311232e-07, "loss": 1.2016, "step": 10217 }, { "epoch": 0.92, "learning_rate": 7.371826347477378e-07, "loss": 1.2009, "step": 10218 }, { "epoch": 0.92, "learning_rate": 7.354556271567891e-07, "loss": 1.1602, "step": 10219 }, { "epoch": 0.92, "learning_rate": 7.337306146803946e-07, "loss": 1.206, "step": 10220 }, { "epoch": 0.92, "learning_rate": 7.320075974603879e-07, "loss": 1.2152, "step": 10221 }, { "epoch": 0.92, "learning_rate": 7.302865756384391e-07, "loss": 1.1443, "step": 10222 }, { "epoch": 0.92, "learning_rate": 7.285675493560601e-07, "loss": 1.2113, "step": 10223 }, { "epoch": 0.92, "learning_rate": 7.268505187545876e-07, "loss": 1.2191, "step": 10224 }, { "epoch": 0.92, "learning_rate": 7.251354839752117e-07, "loss": 1.2012, "step": 10225 }, { "epoch": 0.92, "learning_rate": 7.234224451589444e-07, "loss": 1.2121, "step": 10226 }, { "epoch": 0.92, "learning_rate": 7.217114024466398e-07, "loss": 1.1886, "step": 10227 }, { "epoch": 0.92, "learning_rate": 7.200023559789798e-07, "loss": 1.1851, "step": 10228 }, { "epoch": 0.92, "learning_rate": 7.182953058964909e-07, "loss": 1.1905, "step": 10229 }, { "epoch": 0.92, "learning_rate": 7.165902523395356e-07, "loss": 1.2069, "step": 10230 }, { "epoch": 0.92, "learning_rate": 7.148871954483105e-07, "loss": 1.2091, "step": 10231 }, { "epoch": 0.92, "learning_rate": 7.131861353628394e-07, "loss": 1.183, "step": 10232 }, { "epoch": 0.92, "learning_rate": 7.114870722229938e-07, "loss": 1.2114, "step": 10233 }, { "epoch": 0.92, "learning_rate": 7.097900061684787e-07, "loss": 1.1981, "step": 10234 }, { "epoch": 0.92, "learning_rate": 7.080949373388268e-07, "loss": 1.2668, "step": 10235 }, { "epoch": 0.92, "learning_rate": 7.064018658734156e-07, "loss": 1.1675, "step": 10236 }, { "epoch": 0.92, "learning_rate": 7.047107919114588e-07, "loss": 1.1921, "step": 10237 }, { "epoch": 0.93, "learning_rate": 7.030217155919977e-07, "loss": 1.169, "step": 10238 }, { "epoch": 0.93, "learning_rate": 7.013346370539159e-07, "loss": 1.1936, "step": 10239 }, { "epoch": 0.93, "learning_rate": 6.996495564359273e-07, "loss": 1.2054, "step": 10240 }, { "epoch": 0.93, "learning_rate": 6.979664738765907e-07, "loss": 1.1976, "step": 10241 }, { "epoch": 0.93, "learning_rate": 6.962853895142923e-07, "loss": 1.1917, "step": 10242 }, { "epoch": 0.93, "learning_rate": 6.94606303487258e-07, "loss": 1.2109, "step": 10243 }, { "epoch": 0.93, "learning_rate": 6.929292159335465e-07, "loss": 1.2135, "step": 10244 }, { "epoch": 0.93, "learning_rate": 6.912541269910506e-07, "loss": 1.2081, "step": 10245 }, { "epoch": 0.93, "learning_rate": 6.89581036797507e-07, "loss": 1.2176, "step": 10246 }, { "epoch": 0.93, "learning_rate": 6.879099454904809e-07, "loss": 1.2055, "step": 10247 }, { "epoch": 0.93, "learning_rate": 6.862408532073788e-07, "loss": 1.1773, "step": 10248 }, { "epoch": 0.93, "learning_rate": 6.84573760085433e-07, "loss": 1.162, "step": 10249 }, { "epoch": 0.93, "learning_rate": 6.829086662617168e-07, "loss": 1.2183, "step": 10250 }, { "epoch": 0.93, "learning_rate": 6.812455718731486e-07, "loss": 1.231, "step": 10251 }, { "epoch": 0.93, "learning_rate": 6.795844770564691e-07, "loss": 1.1978, "step": 10252 }, { "epoch": 0.93, "learning_rate": 6.779253819482578e-07, "loss": 1.185, "step": 10253 }, { "epoch": 0.93, "learning_rate": 6.762682866849334e-07, "loss": 1.2361, "step": 10254 }, { "epoch": 0.93, "learning_rate": 6.746131914027481e-07, "loss": 1.1308, "step": 10255 }, { "epoch": 0.93, "learning_rate": 6.729600962377819e-07, "loss": 1.2044, "step": 10256 }, { "epoch": 0.93, "learning_rate": 6.713090013259704e-07, "loss": 1.1806, "step": 10257 }, { "epoch": 0.93, "learning_rate": 6.696599068030663e-07, "loss": 1.1888, "step": 10258 }, { "epoch": 0.93, "learning_rate": 6.680128128046636e-07, "loss": 1.2206, "step": 10259 }, { "epoch": 0.93, "learning_rate": 6.663677194661933e-07, "loss": 1.1837, "step": 10260 }, { "epoch": 0.93, "learning_rate": 6.647246269229162e-07, "loss": 1.2501, "step": 10261 }, { "epoch": 0.93, "learning_rate": 6.630835353099385e-07, "loss": 1.2062, "step": 10262 }, { "epoch": 0.93, "learning_rate": 6.614444447621965e-07, "loss": 1.2133, "step": 10263 }, { "epoch": 0.93, "learning_rate": 6.598073554144574e-07, "loss": 1.1908, "step": 10264 }, { "epoch": 0.93, "learning_rate": 6.58172267401333e-07, "loss": 1.215, "step": 10265 }, { "epoch": 0.93, "learning_rate": 6.565391808572601e-07, "loss": 1.2073, "step": 10266 }, { "epoch": 0.93, "learning_rate": 6.549080959165232e-07, "loss": 1.1711, "step": 10267 }, { "epoch": 0.93, "learning_rate": 6.532790127132343e-07, "loss": 1.2113, "step": 10268 }, { "epoch": 0.93, "learning_rate": 6.516519313813391e-07, "loss": 1.2074, "step": 10269 }, { "epoch": 0.93, "learning_rate": 6.500268520546249e-07, "loss": 1.2581, "step": 10270 }, { "epoch": 0.93, "learning_rate": 6.484037748667044e-07, "loss": 1.2519, "step": 10271 }, { "epoch": 0.93, "learning_rate": 6.467826999510429e-07, "loss": 1.2105, "step": 10272 }, { "epoch": 0.93, "learning_rate": 6.451636274409312e-07, "loss": 1.2182, "step": 10273 }, { "epoch": 0.93, "learning_rate": 6.435465574694822e-07, "loss": 1.2108, "step": 10274 }, { "epoch": 0.93, "learning_rate": 6.419314901696671e-07, "loss": 1.2348, "step": 10275 }, { "epoch": 0.93, "learning_rate": 6.4031842567428e-07, "loss": 1.2094, "step": 10276 }, { "epoch": 0.93, "learning_rate": 6.387073641159508e-07, "loss": 1.2061, "step": 10277 }, { "epoch": 0.93, "learning_rate": 6.370983056271485e-07, "loss": 1.1999, "step": 10278 }, { "epoch": 0.93, "learning_rate": 6.354912503401783e-07, "loss": 1.1871, "step": 10279 }, { "epoch": 0.93, "learning_rate": 6.338861983871736e-07, "loss": 1.2122, "step": 10280 }, { "epoch": 0.93, "learning_rate": 6.322831499001091e-07, "loss": 1.1597, "step": 10281 }, { "epoch": 0.93, "learning_rate": 6.306821050107881e-07, "loss": 1.2218, "step": 10282 }, { "epoch": 0.93, "learning_rate": 6.290830638508605e-07, "loss": 1.1526, "step": 10283 }, { "epoch": 0.93, "learning_rate": 6.274860265518045e-07, "loss": 1.1496, "step": 10284 }, { "epoch": 0.93, "learning_rate": 6.258909932449319e-07, "loss": 1.1405, "step": 10285 }, { "epoch": 0.93, "learning_rate": 6.242979640613933e-07, "loss": 1.2073, "step": 10286 }, { "epoch": 0.93, "learning_rate": 6.227069391321672e-07, "loss": 1.1859, "step": 10287 }, { "epoch": 0.93, "learning_rate": 6.211179185880822e-07, "loss": 1.1869, "step": 10288 }, { "epoch": 0.93, "learning_rate": 6.195309025597867e-07, "loss": 1.1861, "step": 10289 }, { "epoch": 0.93, "learning_rate": 6.179458911777735e-07, "loss": 1.2396, "step": 10290 }, { "epoch": 0.93, "learning_rate": 6.16362884572369e-07, "loss": 1.208, "step": 10291 }, { "epoch": 0.93, "learning_rate": 6.147818828737272e-07, "loss": 1.2335, "step": 10292 }, { "epoch": 0.93, "learning_rate": 6.132028862118499e-07, "loss": 1.2261, "step": 10293 }, { "epoch": 0.93, "learning_rate": 6.116258947165665e-07, "loss": 1.1791, "step": 10294 }, { "epoch": 0.93, "learning_rate": 6.100509085175399e-07, "loss": 1.164, "step": 10295 }, { "epoch": 0.93, "learning_rate": 6.08477927744272e-07, "loss": 1.17, "step": 10296 }, { "epoch": 0.93, "learning_rate": 6.069069525261011e-07, "loss": 1.2031, "step": 10297 }, { "epoch": 0.93, "learning_rate": 6.053379829921935e-07, "loss": 1.2266, "step": 10298 }, { "epoch": 0.93, "learning_rate": 6.037710192715651e-07, "loss": 1.1763, "step": 10299 }, { "epoch": 0.93, "learning_rate": 6.022060614930436e-07, "loss": 1.2097, "step": 10300 }, { "epoch": 0.93, "learning_rate": 6.006431097853121e-07, "loss": 1.1898, "step": 10301 }, { "epoch": 0.93, "learning_rate": 5.990821642768846e-07, "loss": 1.1948, "step": 10302 }, { "epoch": 0.93, "learning_rate": 5.975232250960999e-07, "loss": 1.1835, "step": 10303 }, { "epoch": 0.93, "learning_rate": 5.959662923711473e-07, "loss": 1.184, "step": 10304 }, { "epoch": 0.93, "learning_rate": 5.94411366230041e-07, "loss": 1.1871, "step": 10305 }, { "epoch": 0.93, "learning_rate": 5.928584468006288e-07, "loss": 1.1724, "step": 10306 }, { "epoch": 0.93, "learning_rate": 5.913075342106029e-07, "loss": 1.1829, "step": 10307 }, { "epoch": 0.93, "learning_rate": 5.897586285874751e-07, "loss": 1.1905, "step": 10308 }, { "epoch": 0.93, "learning_rate": 5.882117300586132e-07, "loss": 1.182, "step": 10309 }, { "epoch": 0.93, "learning_rate": 5.866668387512015e-07, "loss": 1.1543, "step": 10310 }, { "epoch": 0.93, "learning_rate": 5.85123954792266e-07, "loss": 1.1911, "step": 10311 }, { "epoch": 0.93, "learning_rate": 5.835830783086721e-07, "loss": 1.205, "step": 10312 }, { "epoch": 0.93, "learning_rate": 5.820442094271128e-07, "loss": 1.1754, "step": 10313 }, { "epoch": 0.93, "learning_rate": 5.805073482741203e-07, "loss": 1.2015, "step": 10314 }, { "epoch": 0.93, "learning_rate": 5.789724949760655e-07, "loss": 1.1696, "step": 10315 }, { "epoch": 0.93, "learning_rate": 5.774396496591367e-07, "loss": 1.21, "step": 10316 }, { "epoch": 0.93, "learning_rate": 5.759088124493828e-07, "loss": 1.1564, "step": 10317 }, { "epoch": 0.93, "learning_rate": 5.743799834726671e-07, "loss": 1.1989, "step": 10318 }, { "epoch": 0.93, "learning_rate": 5.728531628546946e-07, "loss": 1.2119, "step": 10319 }, { "epoch": 0.93, "learning_rate": 5.713283507210148e-07, "loss": 1.2256, "step": 10320 }, { "epoch": 0.93, "learning_rate": 5.698055471969915e-07, "loss": 1.1646, "step": 10321 }, { "epoch": 0.93, "learning_rate": 5.682847524078411e-07, "loss": 1.1915, "step": 10322 }, { "epoch": 0.93, "learning_rate": 5.667659664786079e-07, "loss": 1.2044, "step": 10323 }, { "epoch": 0.93, "learning_rate": 5.652491895341699e-07, "loss": 1.1723, "step": 10324 }, { "epoch": 0.93, "learning_rate": 5.637344216992468e-07, "loss": 1.1476, "step": 10325 }, { "epoch": 0.93, "learning_rate": 5.622216630983779e-07, "loss": 1.1717, "step": 10326 }, { "epoch": 0.93, "learning_rate": 5.607109138559608e-07, "loss": 1.2291, "step": 10327 }, { "epoch": 0.93, "learning_rate": 5.592021740962044e-07, "loss": 1.1781, "step": 10328 }, { "epoch": 0.93, "learning_rate": 5.576954439431626e-07, "loss": 1.198, "step": 10329 }, { "epoch": 0.93, "learning_rate": 5.561907235207303e-07, "loss": 1.2073, "step": 10330 }, { "epoch": 0.93, "learning_rate": 5.546880129526283e-07, "loss": 1.1738, "step": 10331 }, { "epoch": 0.93, "learning_rate": 5.531873123624132e-07, "loss": 1.1339, "step": 10332 }, { "epoch": 0.93, "learning_rate": 5.516886218734807e-07, "loss": 1.19, "step": 10333 }, { "epoch": 0.93, "learning_rate": 5.501919416090518e-07, "loss": 1.212, "step": 10334 }, { "epoch": 0.93, "learning_rate": 5.486972716921946e-07, "loss": 1.1625, "step": 10335 }, { "epoch": 0.93, "learning_rate": 5.47204612245808e-07, "loss": 1.183, "step": 10336 }, { "epoch": 0.93, "learning_rate": 5.457139633926161e-07, "loss": 1.2402, "step": 10337 }, { "epoch": 0.93, "learning_rate": 5.442253252551876e-07, "loss": 1.1736, "step": 10338 }, { "epoch": 0.93, "learning_rate": 5.427386979559301e-07, "loss": 1.1282, "step": 10339 }, { "epoch": 0.93, "learning_rate": 5.41254081617068e-07, "loss": 1.1777, "step": 10340 }, { "epoch": 0.93, "learning_rate": 5.397714763606843e-07, "loss": 1.2146, "step": 10341 }, { "epoch": 0.93, "learning_rate": 5.382908823086702e-07, "loss": 1.2383, "step": 10342 }, { "epoch": 0.93, "learning_rate": 5.368122995827785e-07, "loss": 1.1754, "step": 10343 }, { "epoch": 0.93, "learning_rate": 5.353357283045729e-07, "loss": 1.1702, "step": 10344 }, { "epoch": 0.93, "learning_rate": 5.338611685954675e-07, "loss": 1.1858, "step": 10345 }, { "epoch": 0.93, "learning_rate": 5.32388620576707e-07, "loss": 1.1935, "step": 10346 }, { "epoch": 0.93, "learning_rate": 5.30918084369364e-07, "loss": 1.1973, "step": 10347 }, { "epoch": 0.93, "learning_rate": 5.294495600943583e-07, "loss": 1.1524, "step": 10348 }, { "epoch": 0.94, "learning_rate": 5.279830478724296e-07, "loss": 1.2101, "step": 10349 }, { "epoch": 0.94, "learning_rate": 5.265185478241619e-07, "loss": 1.1897, "step": 10350 }, { "epoch": 0.94, "learning_rate": 5.250560600699728e-07, "loss": 1.2006, "step": 10351 }, { "epoch": 0.94, "learning_rate": 5.235955847301161e-07, "loss": 1.2254, "step": 10352 }, { "epoch": 0.94, "learning_rate": 5.221371219246706e-07, "loss": 1.1711, "step": 10353 }, { "epoch": 0.94, "learning_rate": 5.206806717735602e-07, "loss": 1.1693, "step": 10354 }, { "epoch": 0.94, "learning_rate": 5.19226234396536e-07, "loss": 1.2572, "step": 10355 }, { "epoch": 0.94, "learning_rate": 5.177738099131941e-07, "loss": 1.1896, "step": 10356 }, { "epoch": 0.94, "learning_rate": 5.163233984429527e-07, "loss": 1.2134, "step": 10357 }, { "epoch": 0.94, "learning_rate": 5.148750001050639e-07, "loss": 1.2149, "step": 10358 }, { "epoch": 0.94, "learning_rate": 5.134286150186296e-07, "loss": 1.1664, "step": 10359 }, { "epoch": 0.94, "learning_rate": 5.11984243302574e-07, "loss": 1.2265, "step": 10360 }, { "epoch": 0.94, "learning_rate": 5.105418850756549e-07, "loss": 1.1777, "step": 10361 }, { "epoch": 0.94, "learning_rate": 5.09101540456472e-07, "loss": 1.1866, "step": 10362 }, { "epoch": 0.94, "learning_rate": 5.076632095634525e-07, "loss": 1.1791, "step": 10363 }, { "epoch": 0.94, "learning_rate": 5.062268925148633e-07, "loss": 1.1869, "step": 10364 }, { "epoch": 0.94, "learning_rate": 5.047925894288014e-07, "loss": 1.1914, "step": 10365 }, { "epoch": 0.94, "learning_rate": 5.033603004232001e-07, "loss": 1.2132, "step": 10366 }, { "epoch": 0.94, "learning_rate": 5.019300256158321e-07, "loss": 1.171, "step": 10367 }, { "epoch": 0.94, "learning_rate": 5.005017651242893e-07, "loss": 1.2412, "step": 10368 }, { "epoch": 0.94, "learning_rate": 4.990755190660196e-07, "loss": 1.2074, "step": 10369 }, { "epoch": 0.94, "learning_rate": 4.976512875582845e-07, "loss": 1.1852, "step": 10370 }, { "epoch": 0.94, "learning_rate": 4.962290707181961e-07, "loss": 1.1614, "step": 10371 }, { "epoch": 0.94, "learning_rate": 4.948088686626911e-07, "loss": 1.1876, "step": 10372 }, { "epoch": 0.94, "learning_rate": 4.933906815085404e-07, "loss": 1.137, "step": 10373 }, { "epoch": 0.94, "learning_rate": 4.919745093723588e-07, "loss": 1.1631, "step": 10374 }, { "epoch": 0.94, "learning_rate": 4.905603523705865e-07, "loss": 1.1654, "step": 10375 }, { "epoch": 0.94, "learning_rate": 4.891482106194945e-07, "loss": 1.2589, "step": 10376 }, { "epoch": 0.94, "learning_rate": 4.877380842352008e-07, "loss": 1.2235, "step": 10377 }, { "epoch": 0.94, "learning_rate": 4.863299733336518e-07, "loss": 1.1671, "step": 10378 }, { "epoch": 0.94, "learning_rate": 4.849238780306214e-07, "loss": 1.198, "step": 10379 }, { "epoch": 0.94, "learning_rate": 4.835197984417256e-07, "loss": 1.1662, "step": 10380 }, { "epoch": 0.94, "learning_rate": 4.821177346824163e-07, "loss": 1.1768, "step": 10381 }, { "epoch": 0.94, "learning_rate": 4.807176868679708e-07, "loss": 1.1795, "step": 10382 }, { "epoch": 0.94, "learning_rate": 4.793196551135109e-07, "loss": 1.1444, "step": 10383 }, { "epoch": 0.94, "learning_rate": 4.779236395339809e-07, "loss": 1.1477, "step": 10384 }, { "epoch": 0.94, "learning_rate": 4.765296402441749e-07, "loss": 1.2145, "step": 10385 }, { "epoch": 0.94, "learning_rate": 4.751376573587041e-07, "loss": 1.1671, "step": 10386 }, { "epoch": 0.94, "learning_rate": 4.737476909920241e-07, "loss": 1.1442, "step": 10387 }, { "epoch": 0.94, "learning_rate": 4.7235974125842964e-07, "loss": 1.186, "step": 10388 }, { "epoch": 0.94, "learning_rate": 4.709738082720322e-07, "loss": 1.1841, "step": 10389 }, { "epoch": 0.94, "learning_rate": 4.6958989214679615e-07, "loss": 1.1683, "step": 10390 }, { "epoch": 0.94, "learning_rate": 4.682079929965083e-07, "loss": 1.1704, "step": 10391 }, { "epoch": 0.94, "learning_rate": 4.6682811093479163e-07, "loss": 1.2228, "step": 10392 }, { "epoch": 0.94, "learning_rate": 4.654502460751109e-07, "loss": 1.2089, "step": 10393 }, { "epoch": 0.94, "learning_rate": 4.6407439853075055e-07, "loss": 1.1618, "step": 10394 }, { "epoch": 0.94, "learning_rate": 4.62700568414845e-07, "loss": 1.1756, "step": 10395 }, { "epoch": 0.94, "learning_rate": 4.613287558403512e-07, "loss": 1.1536, "step": 10396 }, { "epoch": 0.94, "learning_rate": 4.599589609200622e-07, "loss": 1.2126, "step": 10397 }, { "epoch": 0.94, "learning_rate": 4.585911837666157e-07, "loss": 1.2048, "step": 10398 }, { "epoch": 0.94, "learning_rate": 4.5722542449246617e-07, "loss": 1.2133, "step": 10399 }, { "epoch": 0.94, "learning_rate": 4.5586168320991273e-07, "loss": 1.2532, "step": 10400 }, { "epoch": 0.94, "learning_rate": 4.5449996003109343e-07, "loss": 1.1272, "step": 10401 }, { "epoch": 0.94, "learning_rate": 4.531402550679659e-07, "loss": 1.1832, "step": 10402 }, { "epoch": 0.94, "learning_rate": 4.517825684323324e-07, "loss": 1.1763, "step": 10403 }, { "epoch": 0.94, "learning_rate": 4.5042690023583136e-07, "loss": 1.2092, "step": 10404 }, { "epoch": 0.94, "learning_rate": 4.490732505899209e-07, "loss": 1.2184, "step": 10405 }, { "epoch": 0.94, "learning_rate": 4.477216196059092e-07, "loss": 1.216, "step": 10406 }, { "epoch": 0.94, "learning_rate": 4.463720073949351e-07, "loss": 1.2426, "step": 10407 }, { "epoch": 0.94, "learning_rate": 4.450244140679599e-07, "loss": 1.1686, "step": 10408 }, { "epoch": 0.94, "learning_rate": 4.4367883973579495e-07, "loss": 1.2278, "step": 10409 }, { "epoch": 0.94, "learning_rate": 4.423352845090711e-07, "loss": 1.1785, "step": 10410 }, { "epoch": 0.94, "learning_rate": 4.409937484982668e-07, "loss": 1.1375, "step": 10411 }, { "epoch": 0.94, "learning_rate": 4.396542318136826e-07, "loss": 1.2061, "step": 10412 }, { "epoch": 0.94, "learning_rate": 4.3831673456546097e-07, "loss": 1.2526, "step": 10413 }, { "epoch": 0.94, "learning_rate": 4.369812568635778e-07, "loss": 1.155, "step": 10414 }, { "epoch": 0.94, "learning_rate": 4.3564779881783426e-07, "loss": 1.1782, "step": 10415 }, { "epoch": 0.94, "learning_rate": 4.3431636053787584e-07, "loss": 1.2112, "step": 10416 }, { "epoch": 0.94, "learning_rate": 4.329869421331789e-07, "loss": 1.2072, "step": 10417 }, { "epoch": 0.94, "learning_rate": 4.316595437130505e-07, "loss": 1.1944, "step": 10418 }, { "epoch": 0.94, "learning_rate": 4.3033416538663664e-07, "loss": 1.2342, "step": 10419 }, { "epoch": 0.94, "learning_rate": 4.290108072629112e-07, "loss": 1.1938, "step": 10420 }, { "epoch": 0.94, "learning_rate": 4.2768946945068444e-07, "loss": 1.22, "step": 10421 }, { "epoch": 0.94, "learning_rate": 4.263701520586083e-07, "loss": 1.183, "step": 10422 }, { "epoch": 0.94, "learning_rate": 4.2505285519515437e-07, "loss": 1.2041, "step": 10423 }, { "epoch": 0.94, "learning_rate": 4.237375789686387e-07, "loss": 1.2144, "step": 10424 }, { "epoch": 0.94, "learning_rate": 4.224243234872083e-07, "loss": 1.1858, "step": 10425 }, { "epoch": 0.94, "learning_rate": 4.211130888588377e-07, "loss": 1.1859, "step": 10426 }, { "epoch": 0.94, "learning_rate": 4.198038751913491e-07, "loss": 1.2708, "step": 10427 }, { "epoch": 0.94, "learning_rate": 4.184966825923869e-07, "loss": 1.1759, "step": 10428 }, { "epoch": 0.94, "learning_rate": 4.171915111694319e-07, "loss": 1.218, "step": 10429 }, { "epoch": 0.94, "learning_rate": 4.1588836102980644e-07, "loss": 1.175, "step": 10430 }, { "epoch": 0.94, "learning_rate": 4.145872322806499e-07, "loss": 1.2467, "step": 10431 }, { "epoch": 0.94, "learning_rate": 4.1328812502895163e-07, "loss": 1.2281, "step": 10432 }, { "epoch": 0.94, "learning_rate": 4.1199103938152903e-07, "loss": 1.237, "step": 10433 }, { "epoch": 0.94, "learning_rate": 4.1069597544502735e-07, "loss": 1.1717, "step": 10434 }, { "epoch": 0.94, "learning_rate": 4.0940293332594193e-07, "loss": 1.1234, "step": 10435 }, { "epoch": 0.94, "learning_rate": 4.081119131305766e-07, "loss": 1.2271, "step": 10436 }, { "epoch": 0.94, "learning_rate": 4.068229149650965e-07, "loss": 1.2022, "step": 10437 }, { "epoch": 0.94, "learning_rate": 4.0553593893548346e-07, "loss": 1.1561, "step": 10438 }, { "epoch": 0.94, "learning_rate": 4.042509851475529e-07, "loss": 1.1913, "step": 10439 }, { "epoch": 0.94, "learning_rate": 4.0296805370696466e-07, "loss": 1.1418, "step": 10440 }, { "epoch": 0.94, "learning_rate": 4.016871447192011e-07, "loss": 1.259, "step": 10441 }, { "epoch": 0.94, "learning_rate": 4.004082582895807e-07, "loss": 1.1842, "step": 10442 }, { "epoch": 0.94, "learning_rate": 3.9913139452326665e-07, "loss": 1.2171, "step": 10443 }, { "epoch": 0.94, "learning_rate": 3.9785655352523876e-07, "loss": 1.25, "step": 10444 }, { "epoch": 0.94, "learning_rate": 3.9658373540032436e-07, "loss": 1.2591, "step": 10445 }, { "epoch": 0.94, "learning_rate": 3.953129402531758e-07, "loss": 1.2101, "step": 10446 }, { "epoch": 0.94, "learning_rate": 3.940441681882817e-07, "loss": 1.2363, "step": 10447 }, { "epoch": 0.94, "learning_rate": 3.927774193099698e-07, "loss": 1.1806, "step": 10448 }, { "epoch": 0.94, "learning_rate": 3.9151269372239006e-07, "loss": 1.2041, "step": 10449 }, { "epoch": 0.94, "learning_rate": 3.902499915295371e-07, "loss": 1.201, "step": 10450 }, { "epoch": 0.94, "learning_rate": 3.8898931283523344e-07, "loss": 1.2139, "step": 10451 }, { "epoch": 0.94, "learning_rate": 3.877306577431322e-07, "loss": 1.1924, "step": 10452 }, { "epoch": 0.94, "learning_rate": 3.86474026356734e-07, "loss": 1.2268, "step": 10453 }, { "epoch": 0.94, "learning_rate": 3.8521941877935344e-07, "loss": 1.1705, "step": 10454 }, { "epoch": 0.94, "learning_rate": 3.839668351141551e-07, "loss": 1.2812, "step": 10455 }, { "epoch": 0.94, "learning_rate": 3.827162754641289e-07, "loss": 1.1862, "step": 10456 }, { "epoch": 0.94, "learning_rate": 3.814677399320982e-07, "loss": 1.1901, "step": 10457 }, { "epoch": 0.94, "learning_rate": 3.802212286207252e-07, "loss": 1.2163, "step": 10458 }, { "epoch": 0.94, "learning_rate": 3.789767416325002e-07, "loss": 1.1759, "step": 10459 }, { "epoch": 0.95, "learning_rate": 3.777342790697469e-07, "loss": 1.1652, "step": 10460 }, { "epoch": 0.95, "learning_rate": 3.764938410346336e-07, "loss": 1.2524, "step": 10461 }, { "epoch": 0.95, "learning_rate": 3.752554276291426e-07, "loss": 1.1845, "step": 10462 }, { "epoch": 0.95, "learning_rate": 3.7401903895510627e-07, "loss": 1.2246, "step": 10463 }, { "epoch": 0.95, "learning_rate": 3.72784675114185e-07, "loss": 1.1501, "step": 10464 }, { "epoch": 0.95, "learning_rate": 3.7155233620787265e-07, "loss": 1.1937, "step": 10465 }, { "epoch": 0.95, "learning_rate": 3.7032202233749093e-07, "loss": 1.1662, "step": 10466 }, { "epoch": 0.95, "learning_rate": 3.6909373360420884e-07, "loss": 1.188, "step": 10467 }, { "epoch": 0.95, "learning_rate": 3.6786747010901235e-07, "loss": 1.2008, "step": 10468 }, { "epoch": 0.95, "learning_rate": 3.6664323195273464e-07, "loss": 1.1715, "step": 10469 }, { "epoch": 0.95, "learning_rate": 3.654210192360369e-07, "loss": 1.1713, "step": 10470 }, { "epoch": 0.95, "learning_rate": 3.6420083205941094e-07, "loss": 1.1977, "step": 10471 }, { "epoch": 0.95, "learning_rate": 3.629826705231848e-07, "loss": 1.1689, "step": 10472 }, { "epoch": 0.95, "learning_rate": 3.617665347275201e-07, "loss": 1.1934, "step": 10473 }, { "epoch": 0.95, "learning_rate": 3.6055242477241457e-07, "loss": 1.1786, "step": 10474 }, { "epoch": 0.95, "learning_rate": 3.5934034075769395e-07, "loss": 1.1992, "step": 10475 }, { "epoch": 0.95, "learning_rate": 3.581302827830202e-07, "loss": 1.1743, "step": 10476 }, { "epoch": 0.95, "learning_rate": 3.569222509478887e-07, "loss": 1.1914, "step": 10477 }, { "epoch": 0.95, "learning_rate": 3.557162453516283e-07, "loss": 1.1961, "step": 10478 }, { "epoch": 0.95, "learning_rate": 3.545122660934014e-07, "loss": 1.2205, "step": 10479 }, { "epoch": 0.95, "learning_rate": 3.53310313272201e-07, "loss": 1.2263, "step": 10480 }, { "epoch": 0.95, "learning_rate": 3.521103869868564e-07, "loss": 1.1297, "step": 10481 }, { "epoch": 0.95, "learning_rate": 3.509124873360359e-07, "loss": 1.1881, "step": 10482 }, { "epoch": 0.95, "learning_rate": 3.4971661441822457e-07, "loss": 1.1697, "step": 10483 }, { "epoch": 0.95, "learning_rate": 3.485227683317577e-07, "loss": 1.1989, "step": 10484 }, { "epoch": 0.95, "learning_rate": 3.473309491747956e-07, "loss": 1.1784, "step": 10485 }, { "epoch": 0.95, "learning_rate": 3.4614115704533767e-07, "loss": 1.1831, "step": 10486 }, { "epoch": 0.95, "learning_rate": 3.449533920412057e-07, "loss": 1.192, "step": 10487 }, { "epoch": 0.95, "learning_rate": 3.43767654260066e-07, "loss": 1.1989, "step": 10488 }, { "epoch": 0.95, "learning_rate": 3.425839437994127e-07, "loss": 1.237, "step": 10489 }, { "epoch": 0.95, "learning_rate": 3.4140226075657646e-07, "loss": 1.3116, "step": 10490 }, { "epoch": 0.95, "learning_rate": 3.402226052287183e-07, "loss": 1.1832, "step": 10491 }, { "epoch": 0.95, "learning_rate": 3.390449773128357e-07, "loss": 1.2094, "step": 10492 }, { "epoch": 0.95, "learning_rate": 3.378693771057512e-07, "loss": 1.2152, "step": 10493 }, { "epoch": 0.95, "learning_rate": 3.366958047041319e-07, "loss": 1.1951, "step": 10494 }, { "epoch": 0.95, "learning_rate": 3.355242602044728e-07, "loss": 1.2592, "step": 10495 }, { "epoch": 0.95, "learning_rate": 3.3435474370309973e-07, "loss": 1.189, "step": 10496 }, { "epoch": 0.95, "learning_rate": 3.3318725529617733e-07, "loss": 1.1596, "step": 10497 }, { "epoch": 0.95, "learning_rate": 3.320217950796983e-07, "loss": 1.2654, "step": 10498 }, { "epoch": 0.95, "learning_rate": 3.308583631494888e-07, "loss": 1.1452, "step": 10499 }, { "epoch": 0.95, "learning_rate": 3.2969695960121395e-07, "loss": 1.1355, "step": 10500 }, { "epoch": 0.95, "learning_rate": 3.2853758453036955e-07, "loss": 1.2151, "step": 10501 }, { "epoch": 0.95, "learning_rate": 3.2738023803227934e-07, "loss": 1.1738, "step": 10502 }, { "epoch": 0.95, "learning_rate": 3.262249202021089e-07, "loss": 1.1853, "step": 10503 }, { "epoch": 0.95, "learning_rate": 3.2507163113484894e-07, "loss": 1.208, "step": 10504 }, { "epoch": 0.95, "learning_rate": 3.239203709253236e-07, "loss": 1.1704, "step": 10505 }, { "epoch": 0.95, "learning_rate": 3.227711396682015e-07, "loss": 1.1855, "step": 10506 }, { "epoch": 0.95, "learning_rate": 3.2162393745797105e-07, "loss": 1.2155, "step": 10507 }, { "epoch": 0.95, "learning_rate": 3.204787643889623e-07, "loss": 1.1857, "step": 10508 }, { "epoch": 0.95, "learning_rate": 3.1933562055533327e-07, "loss": 1.1747, "step": 10509 }, { "epoch": 0.95, "learning_rate": 3.181945060510727e-07, "loss": 1.1956, "step": 10510 }, { "epoch": 0.95, "learning_rate": 3.170554209700166e-07, "loss": 1.2032, "step": 10511 }, { "epoch": 0.95, "learning_rate": 3.1591836540581786e-07, "loss": 1.1821, "step": 10512 }, { "epoch": 0.95, "learning_rate": 3.1478333945197393e-07, "loss": 1.2123, "step": 10513 }, { "epoch": 0.95, "learning_rate": 3.136503432018045e-07, "loss": 1.1928, "step": 10514 }, { "epoch": 0.95, "learning_rate": 3.125193767484713e-07, "loss": 1.2352, "step": 10515 }, { "epoch": 0.95, "learning_rate": 3.113904401849665e-07, "loss": 1.2448, "step": 10516 }, { "epoch": 0.95, "learning_rate": 3.10263533604116e-07, "loss": 1.139, "step": 10517 }, { "epoch": 0.95, "learning_rate": 3.0913865709857883e-07, "loss": 1.2256, "step": 10518 }, { "epoch": 0.95, "learning_rate": 3.080158107608422e-07, "loss": 1.1938, "step": 10519 }, { "epoch": 0.95, "learning_rate": 3.0689499468322946e-07, "loss": 1.1782, "step": 10520 }, { "epoch": 0.95, "learning_rate": 3.0577620895790584e-07, "loss": 1.1949, "step": 10521 }, { "epoch": 0.95, "learning_rate": 3.04659453676856e-07, "loss": 1.2244, "step": 10522 }, { "epoch": 0.95, "learning_rate": 3.0354472893190377e-07, "loss": 1.2316, "step": 10523 }, { "epoch": 0.95, "learning_rate": 3.0243203481470917e-07, "loss": 1.2034, "step": 10524 }, { "epoch": 0.95, "learning_rate": 3.013213714167573e-07, "loss": 1.1746, "step": 10525 }, { "epoch": 0.95, "learning_rate": 3.002127388293696e-07, "loss": 1.2103, "step": 10526 }, { "epoch": 0.95, "learning_rate": 2.991061371437065e-07, "loss": 1.1918, "step": 10527 }, { "epoch": 0.95, "learning_rate": 2.980015664507535e-07, "loss": 1.1895, "step": 10528 }, { "epoch": 0.95, "learning_rate": 2.968990268413352e-07, "loss": 1.1632, "step": 10529 }, { "epoch": 0.95, "learning_rate": 2.957985184061041e-07, "loss": 1.2578, "step": 10530 }, { "epoch": 0.95, "learning_rate": 2.9470004123554605e-07, "loss": 1.2228, "step": 10531 }, { "epoch": 0.95, "learning_rate": 2.9360359541998615e-07, "loss": 1.1668, "step": 10532 }, { "epoch": 0.95, "learning_rate": 2.9250918104957447e-07, "loss": 1.2194, "step": 10533 }, { "epoch": 0.95, "learning_rate": 2.9141679821429735e-07, "loss": 1.2236, "step": 10534 }, { "epoch": 0.95, "learning_rate": 2.9032644700397736e-07, "loss": 1.1831, "step": 10535 }, { "epoch": 0.95, "learning_rate": 2.892381275082623e-07, "loss": 1.1628, "step": 10536 }, { "epoch": 0.95, "learning_rate": 2.881518398166416e-07, "loss": 1.2117, "step": 10537 }, { "epoch": 0.95, "learning_rate": 2.870675840184328e-07, "loss": 1.1377, "step": 10538 }, { "epoch": 0.95, "learning_rate": 2.8598536020278677e-07, "loss": 1.1655, "step": 10539 }, { "epoch": 0.95, "learning_rate": 2.8490516845868785e-07, "loss": 1.1816, "step": 10540 }, { "epoch": 0.95, "learning_rate": 2.838270088749484e-07, "loss": 1.2001, "step": 10541 }, { "epoch": 0.95, "learning_rate": 2.82750881540228e-07, "loss": 1.1502, "step": 10542 }, { "epoch": 0.95, "learning_rate": 2.8167678654300313e-07, "loss": 1.1711, "step": 10543 }, { "epoch": 0.95, "learning_rate": 2.806047239715892e-07, "loss": 1.2042, "step": 10544 }, { "epoch": 0.95, "learning_rate": 2.79534693914138e-07, "loss": 1.2225, "step": 10545 }, { "epoch": 0.95, "learning_rate": 2.784666964586291e-07, "loss": 1.1917, "step": 10546 }, { "epoch": 0.95, "learning_rate": 2.7740073169287564e-07, "loss": 1.2401, "step": 10547 }, { "epoch": 0.95, "learning_rate": 2.763367997045269e-07, "loss": 1.2322, "step": 10548 }, { "epoch": 0.95, "learning_rate": 2.7527490058106565e-07, "loss": 1.1972, "step": 10549 }, { "epoch": 0.95, "learning_rate": 2.742150344097999e-07, "loss": 1.1341, "step": 10550 }, { "epoch": 0.95, "learning_rate": 2.731572012778794e-07, "loss": 1.2451, "step": 10551 }, { "epoch": 0.95, "learning_rate": 2.721014012722789e-07, "loss": 1.2136, "step": 10552 }, { "epoch": 0.95, "learning_rate": 2.7104763447981243e-07, "loss": 1.1755, "step": 10553 }, { "epoch": 0.95, "learning_rate": 2.699959009871245e-07, "loss": 1.2196, "step": 10554 }, { "epoch": 0.95, "learning_rate": 2.689462008806931e-07, "loss": 1.25, "step": 10555 }, { "epoch": 0.95, "learning_rate": 2.678985342468271e-07, "loss": 1.2284, "step": 10556 }, { "epoch": 0.95, "learning_rate": 2.668529011716658e-07, "loss": 1.1593, "step": 10557 }, { "epoch": 0.95, "learning_rate": 2.6580930174119054e-07, "loss": 1.246, "step": 10558 }, { "epoch": 0.95, "learning_rate": 2.6476773604120487e-07, "loss": 1.1549, "step": 10559 }, { "epoch": 0.95, "learning_rate": 2.637282041573541e-07, "loss": 1.1603, "step": 10560 }, { "epoch": 0.95, "learning_rate": 2.626907061751116e-07, "loss": 1.1447, "step": 10561 }, { "epoch": 0.95, "learning_rate": 2.6165524217977853e-07, "loss": 1.2146, "step": 10562 }, { "epoch": 0.95, "learning_rate": 2.606218122565007e-07, "loss": 1.2274, "step": 10563 }, { "epoch": 0.95, "learning_rate": 2.5959041649024905e-07, "loss": 1.2036, "step": 10564 }, { "epoch": 0.95, "learning_rate": 2.5856105496582796e-07, "loss": 1.1924, "step": 10565 }, { "epoch": 0.95, "learning_rate": 2.5753372776787245e-07, "loss": 1.2077, "step": 10566 }, { "epoch": 0.95, "learning_rate": 2.5650843498085663e-07, "loss": 1.1489, "step": 10567 }, { "epoch": 0.95, "learning_rate": 2.5548517668907977e-07, "loss": 1.2018, "step": 10568 }, { "epoch": 0.95, "learning_rate": 2.544639529766829e-07, "loss": 1.1348, "step": 10569 }, { "epoch": 0.96, "learning_rate": 2.534447639276294e-07, "loss": 1.1353, "step": 10570 }, { "epoch": 0.96, "learning_rate": 2.524276096257244e-07, "loss": 1.2121, "step": 10571 }, { "epoch": 0.96, "learning_rate": 2.514124901545983e-07, "loss": 1.1907, "step": 10572 }, { "epoch": 0.96, "learning_rate": 2.503994055977205e-07, "loss": 1.2339, "step": 10573 }, { "epoch": 0.96, "learning_rate": 2.4938835603838807e-07, "loss": 1.2135, "step": 10574 }, { "epoch": 0.96, "learning_rate": 2.483793415597346e-07, "loss": 1.1816, "step": 10575 }, { "epoch": 0.96, "learning_rate": 2.4737236224472716e-07, "loss": 1.1824, "step": 10576 }, { "epoch": 0.96, "learning_rate": 2.463674181761577e-07, "loss": 1.1756, "step": 10577 }, { "epoch": 0.96, "learning_rate": 2.453645094366602e-07, "loss": 1.2472, "step": 10578 }, { "epoch": 0.96, "learning_rate": 2.443636361086937e-07, "loss": 1.1912, "step": 10579 }, { "epoch": 0.96, "learning_rate": 2.43364798274559e-07, "loss": 1.1979, "step": 10580 }, { "epoch": 0.96, "learning_rate": 2.423679960163794e-07, "loss": 1.2213, "step": 10581 }, { "epoch": 0.96, "learning_rate": 2.4137322941611696e-07, "loss": 1.1796, "step": 10582 }, { "epoch": 0.96, "learning_rate": 2.4038049855556186e-07, "loss": 1.2291, "step": 10583 }, { "epoch": 0.96, "learning_rate": 2.3938980351634335e-07, "loss": 1.1861, "step": 10584 }, { "epoch": 0.96, "learning_rate": 2.3840114437992113e-07, "loss": 1.2133, "step": 10585 }, { "epoch": 0.96, "learning_rate": 2.37414521227583e-07, "loss": 1.1731, "step": 10586 }, { "epoch": 0.96, "learning_rate": 2.36429934140453e-07, "loss": 1.154, "step": 10587 }, { "epoch": 0.96, "learning_rate": 2.3544738319948856e-07, "loss": 1.189, "step": 10588 }, { "epoch": 0.96, "learning_rate": 2.3446686848547784e-07, "loss": 1.1976, "step": 10589 }, { "epoch": 0.96, "learning_rate": 2.3348839007904243e-07, "loss": 1.2111, "step": 10590 }, { "epoch": 0.96, "learning_rate": 2.325119480606347e-07, "loss": 1.1995, "step": 10591 }, { "epoch": 0.96, "learning_rate": 2.3153754251054315e-07, "loss": 1.2522, "step": 10592 }, { "epoch": 0.96, "learning_rate": 2.3056517350888428e-07, "loss": 1.1686, "step": 10593 }, { "epoch": 0.96, "learning_rate": 2.295948411356108e-07, "loss": 1.1812, "step": 10594 }, { "epoch": 0.96, "learning_rate": 2.2862654547050887e-07, "loss": 1.2206, "step": 10595 }, { "epoch": 0.96, "learning_rate": 2.276602865931926e-07, "loss": 1.195, "step": 10596 }, { "epoch": 0.96, "learning_rate": 2.2669606458310955e-07, "loss": 1.1998, "step": 10597 }, { "epoch": 0.96, "learning_rate": 2.2573387951954628e-07, "loss": 1.1177, "step": 10598 }, { "epoch": 0.96, "learning_rate": 2.2477373148161452e-07, "loss": 1.2065, "step": 10599 }, { "epoch": 0.96, "learning_rate": 2.2381562054825943e-07, "loss": 1.1627, "step": 10600 }, { "epoch": 0.96, "learning_rate": 2.2285954679826248e-07, "loss": 1.203, "step": 10601 }, { "epoch": 0.96, "learning_rate": 2.21905510310233e-07, "loss": 1.1666, "step": 10602 }, { "epoch": 0.96, "learning_rate": 2.2095351116261932e-07, "loss": 1.2338, "step": 10603 }, { "epoch": 0.96, "learning_rate": 2.2000354943369217e-07, "loss": 1.1828, "step": 10604 }, { "epoch": 0.96, "learning_rate": 2.1905562520156686e-07, "loss": 1.1617, "step": 10605 }, { "epoch": 0.96, "learning_rate": 2.1810973854417826e-07, "loss": 1.2665, "step": 10606 }, { "epoch": 0.96, "learning_rate": 2.171658895393086e-07, "loss": 1.1796, "step": 10607 }, { "epoch": 0.96, "learning_rate": 2.1622407826455693e-07, "loss": 1.2539, "step": 10608 }, { "epoch": 0.96, "learning_rate": 2.1528430479736683e-07, "loss": 1.2082, "step": 10609 }, { "epoch": 0.96, "learning_rate": 2.1434656921500706e-07, "loss": 1.25, "step": 10610 }, { "epoch": 0.96, "learning_rate": 2.1341087159458263e-07, "loss": 1.1525, "step": 10611 }, { "epoch": 0.96, "learning_rate": 2.1247721201302917e-07, "loss": 1.1902, "step": 10612 }, { "epoch": 0.96, "learning_rate": 2.1154559054711865e-07, "loss": 1.1956, "step": 10613 }, { "epoch": 0.96, "learning_rate": 2.106160072734481e-07, "loss": 1.1895, "step": 10614 }, { "epoch": 0.96, "learning_rate": 2.0968846226845084e-07, "loss": 1.1869, "step": 10615 }, { "epoch": 0.96, "learning_rate": 2.0876295560839364e-07, "loss": 1.1899, "step": 10616 }, { "epoch": 0.96, "learning_rate": 2.0783948736937953e-07, "loss": 1.1833, "step": 10617 }, { "epoch": 0.96, "learning_rate": 2.0691805762733385e-07, "loss": 1.1227, "step": 10618 }, { "epoch": 0.96, "learning_rate": 2.0599866645802102e-07, "loss": 1.2403, "step": 10619 }, { "epoch": 0.96, "learning_rate": 2.050813139370361e-07, "loss": 1.2115, "step": 10620 }, { "epoch": 0.96, "learning_rate": 2.041660001398077e-07, "loss": 1.1598, "step": 10621 }, { "epoch": 0.96, "learning_rate": 2.0325272514159777e-07, "loss": 1.253, "step": 10622 }, { "epoch": 0.96, "learning_rate": 2.0234148901749628e-07, "loss": 1.2329, "step": 10623 }, { "epoch": 0.96, "learning_rate": 2.014322918424294e-07, "loss": 1.1917, "step": 10624 }, { "epoch": 0.96, "learning_rate": 2.00525133691154e-07, "loss": 1.1546, "step": 10625 }, { "epoch": 0.96, "learning_rate": 1.9962001463825762e-07, "loss": 1.1924, "step": 10626 }, { "epoch": 0.96, "learning_rate": 1.9871693475816965e-07, "loss": 1.1098, "step": 10627 }, { "epoch": 0.96, "learning_rate": 1.9781589412513624e-07, "loss": 1.1834, "step": 10628 }, { "epoch": 0.96, "learning_rate": 1.969168928132481e-07, "loss": 1.222, "step": 10629 }, { "epoch": 0.96, "learning_rate": 1.960199308964239e-07, "loss": 1.229, "step": 10630 }, { "epoch": 0.96, "learning_rate": 1.951250084484102e-07, "loss": 1.2182, "step": 10631 }, { "epoch": 0.96, "learning_rate": 1.9423212554279813e-07, "loss": 1.1412, "step": 10632 }, { "epoch": 0.96, "learning_rate": 1.933412822530012e-07, "loss": 1.1787, "step": 10633 }, { "epoch": 0.96, "learning_rate": 1.9245247865226635e-07, "loss": 1.2377, "step": 10634 }, { "epoch": 0.96, "learning_rate": 1.9156571481367401e-07, "loss": 1.1705, "step": 10635 }, { "epoch": 0.96, "learning_rate": 1.9068099081013535e-07, "loss": 1.2434, "step": 10636 }, { "epoch": 0.96, "learning_rate": 1.8979830671440047e-07, "loss": 1.1739, "step": 10637 }, { "epoch": 0.96, "learning_rate": 1.8891766259904186e-07, "loss": 1.2245, "step": 10638 }, { "epoch": 0.96, "learning_rate": 1.8803905853647386e-07, "loss": 1.2736, "step": 10639 }, { "epoch": 0.96, "learning_rate": 1.871624945989331e-07, "loss": 1.1614, "step": 10640 }, { "epoch": 0.96, "learning_rate": 1.8628797085849527e-07, "loss": 1.1895, "step": 10641 }, { "epoch": 0.96, "learning_rate": 1.854154873870667e-07, "loss": 1.172, "step": 10642 }, { "epoch": 0.96, "learning_rate": 1.845450442563901e-07, "loss": 1.1667, "step": 10643 }, { "epoch": 0.96, "learning_rate": 1.8367664153803033e-07, "loss": 1.1959, "step": 10644 }, { "epoch": 0.96, "learning_rate": 1.8281027930339422e-07, "loss": 1.1399, "step": 10645 }, { "epoch": 0.96, "learning_rate": 1.8194595762371648e-07, "loss": 1.2319, "step": 10646 }, { "epoch": 0.96, "learning_rate": 1.810836765700624e-07, "loss": 1.1475, "step": 10647 }, { "epoch": 0.96, "learning_rate": 1.8022343621333648e-07, "loss": 1.1943, "step": 10648 }, { "epoch": 0.96, "learning_rate": 1.7936523662426262e-07, "loss": 1.1896, "step": 10649 }, { "epoch": 0.96, "learning_rate": 1.785090778734122e-07, "loss": 1.173, "step": 10650 }, { "epoch": 0.96, "learning_rate": 1.7765496003118164e-07, "loss": 1.2442, "step": 10651 }, { "epoch": 0.96, "learning_rate": 1.768028831677926e-07, "loss": 1.1874, "step": 10652 }, { "epoch": 0.96, "learning_rate": 1.7595284735331397e-07, "loss": 1.1899, "step": 10653 }, { "epoch": 0.96, "learning_rate": 1.7510485265763431e-07, "loss": 1.2167, "step": 10654 }, { "epoch": 0.96, "learning_rate": 1.7425889915047845e-07, "loss": 1.1854, "step": 10655 }, { "epoch": 0.96, "learning_rate": 1.7341498690140457e-07, "loss": 1.167, "step": 10656 }, { "epoch": 0.96, "learning_rate": 1.7257311597979886e-07, "loss": 1.2545, "step": 10657 }, { "epoch": 0.96, "learning_rate": 1.7173328645488928e-07, "loss": 1.1527, "step": 10658 }, { "epoch": 0.96, "learning_rate": 1.7089549839572617e-07, "loss": 1.1934, "step": 10659 }, { "epoch": 0.96, "learning_rate": 1.7005975187119326e-07, "loss": 1.189, "step": 10660 }, { "epoch": 0.96, "learning_rate": 1.6922604695001343e-07, "loss": 1.1757, "step": 10661 }, { "epoch": 0.96, "learning_rate": 1.683943837007318e-07, "loss": 1.1605, "step": 10662 }, { "epoch": 0.96, "learning_rate": 1.6756476219173257e-07, "loss": 1.1945, "step": 10663 }, { "epoch": 0.96, "learning_rate": 1.667371824912334e-07, "loss": 1.1794, "step": 10664 }, { "epoch": 0.96, "learning_rate": 1.6591164466727428e-07, "loss": 1.2398, "step": 10665 }, { "epoch": 0.96, "learning_rate": 1.6508814878773705e-07, "loss": 1.17, "step": 10666 }, { "epoch": 0.96, "learning_rate": 1.6426669492033142e-07, "loss": 1.25, "step": 10667 }, { "epoch": 0.96, "learning_rate": 1.6344728313260337e-07, "loss": 1.1859, "step": 10668 }, { "epoch": 0.96, "learning_rate": 1.626299134919268e-07, "loss": 1.2292, "step": 10669 }, { "epoch": 0.96, "learning_rate": 1.6181458606550347e-07, "loss": 1.2097, "step": 10670 }, { "epoch": 0.96, "learning_rate": 1.6100130092037703e-07, "loss": 1.1919, "step": 10671 }, { "epoch": 0.96, "learning_rate": 1.6019005812341893e-07, "loss": 1.2272, "step": 10672 }, { "epoch": 0.96, "learning_rate": 1.593808577413314e-07, "loss": 1.2084, "step": 10673 }, { "epoch": 0.96, "learning_rate": 1.5857369984065007e-07, "loss": 1.1428, "step": 10674 }, { "epoch": 0.96, "learning_rate": 1.5776858448774133e-07, "loss": 1.1872, "step": 10675 }, { "epoch": 0.96, "learning_rate": 1.5696551174880503e-07, "loss": 1.1548, "step": 10676 }, { "epoch": 0.96, "learning_rate": 1.5616448168987164e-07, "loss": 1.2621, "step": 10677 }, { "epoch": 0.96, "learning_rate": 1.5536549437680515e-07, "loss": 1.1422, "step": 10678 }, { "epoch": 0.96, "learning_rate": 1.54568549875303e-07, "loss": 1.18, "step": 10679 }, { "epoch": 0.96, "learning_rate": 1.5377364825089058e-07, "loss": 1.2081, "step": 10680 }, { "epoch": 0.97, "learning_rate": 1.5298078956892669e-07, "loss": 1.2222, "step": 10681 }, { "epoch": 0.97, "learning_rate": 1.5218997389460644e-07, "loss": 1.2506, "step": 10682 }, { "epoch": 0.97, "learning_rate": 1.5140120129294998e-07, "loss": 1.204, "step": 10683 }, { "epoch": 0.97, "learning_rate": 1.5061447182881384e-07, "loss": 1.2058, "step": 10684 }, { "epoch": 0.97, "learning_rate": 1.4982978556688786e-07, "loss": 1.162, "step": 10685 }, { "epoch": 0.97, "learning_rate": 1.4904714257168716e-07, "loss": 1.2107, "step": 10686 }, { "epoch": 0.97, "learning_rate": 1.482665429075658e-07, "loss": 1.143, "step": 10687 }, { "epoch": 0.97, "learning_rate": 1.4748798663870856e-07, "loss": 1.1878, "step": 10688 }, { "epoch": 0.97, "learning_rate": 1.467114738291281e-07, "loss": 1.1756, "step": 10689 }, { "epoch": 0.97, "learning_rate": 1.4593700454267612e-07, "loss": 1.1991, "step": 10690 }, { "epoch": 0.97, "learning_rate": 1.4516457884302947e-07, "loss": 1.2191, "step": 10691 }, { "epoch": 0.97, "learning_rate": 1.4439419679369847e-07, "loss": 1.199, "step": 10692 }, { "epoch": 0.97, "learning_rate": 1.4362585845802968e-07, "loss": 1.1427, "step": 10693 }, { "epoch": 0.97, "learning_rate": 1.4285956389919752e-07, "loss": 1.133, "step": 10694 }, { "epoch": 0.97, "learning_rate": 1.4209531318020719e-07, "loss": 1.1373, "step": 10695 }, { "epoch": 0.97, "learning_rate": 1.4133310636390007e-07, "loss": 1.1865, "step": 10696 }, { "epoch": 0.97, "learning_rate": 1.4057294351294825e-07, "loss": 1.2119, "step": 10697 }, { "epoch": 0.97, "learning_rate": 1.3981482468985175e-07, "loss": 1.1706, "step": 10698 }, { "epoch": 0.97, "learning_rate": 1.3905874995694957e-07, "loss": 1.2246, "step": 10699 }, { "epoch": 0.97, "learning_rate": 1.3830471937640588e-07, "loss": 1.201, "step": 10700 }, { "epoch": 0.97, "learning_rate": 1.375527330102211e-07, "loss": 1.1169, "step": 10701 }, { "epoch": 0.97, "learning_rate": 1.368027909202263e-07, "loss": 1.1754, "step": 10702 }, { "epoch": 0.97, "learning_rate": 1.3605489316808328e-07, "loss": 1.1732, "step": 10703 }, { "epoch": 0.97, "learning_rate": 1.3530903981528453e-07, "loss": 1.2227, "step": 10704 }, { "epoch": 0.97, "learning_rate": 1.3456523092316154e-07, "loss": 1.1546, "step": 10705 }, { "epoch": 0.97, "learning_rate": 1.338234665528737e-07, "loss": 1.1606, "step": 10706 }, { "epoch": 0.97, "learning_rate": 1.330837467654028e-07, "loss": 1.2102, "step": 10707 }, { "epoch": 0.97, "learning_rate": 1.3234607162158073e-07, "loss": 1.1665, "step": 10708 }, { "epoch": 0.97, "learning_rate": 1.3161044118205624e-07, "loss": 1.2076, "step": 10709 }, { "epoch": 0.97, "learning_rate": 1.3087685550731698e-07, "loss": 1.1919, "step": 10710 }, { "epoch": 0.97, "learning_rate": 1.3014531465768144e-07, "loss": 1.1795, "step": 10711 }, { "epoch": 0.97, "learning_rate": 1.2941581869329866e-07, "loss": 1.1598, "step": 10712 }, { "epoch": 0.97, "learning_rate": 1.2868836767415127e-07, "loss": 1.2147, "step": 10713 }, { "epoch": 0.97, "learning_rate": 1.2796296166004972e-07, "loss": 1.1764, "step": 10714 }, { "epoch": 0.97, "learning_rate": 1.2723960071064357e-07, "loss": 1.1776, "step": 10715 }, { "epoch": 0.97, "learning_rate": 1.265182848854074e-07, "loss": 1.1855, "step": 10716 }, { "epoch": 0.97, "learning_rate": 1.2579901424365215e-07, "loss": 1.1311, "step": 10717 }, { "epoch": 0.97, "learning_rate": 1.250817888445166e-07, "loss": 1.1951, "step": 10718 }, { "epoch": 0.97, "learning_rate": 1.2436660874697302e-07, "loss": 1.2209, "step": 10719 }, { "epoch": 0.97, "learning_rate": 1.2365347400982995e-07, "loss": 1.2534, "step": 10720 }, { "epoch": 0.97, "learning_rate": 1.2294238469172105e-07, "loss": 1.1987, "step": 10721 }, { "epoch": 0.97, "learning_rate": 1.2223334085111337e-07, "loss": 1.1883, "step": 10722 }, { "epoch": 0.97, "learning_rate": 1.2152634254631033e-07, "loss": 1.1548, "step": 10723 }, { "epoch": 0.97, "learning_rate": 1.2082138983544046e-07, "loss": 1.2529, "step": 10724 }, { "epoch": 0.97, "learning_rate": 1.2011848277646842e-07, "loss": 1.1983, "step": 10725 }, { "epoch": 0.97, "learning_rate": 1.194176214271897e-07, "loss": 1.1689, "step": 10726 }, { "epoch": 0.97, "learning_rate": 1.187188058452332e-07, "loss": 1.1681, "step": 10727 }, { "epoch": 0.97, "learning_rate": 1.1802203608805573e-07, "loss": 1.2089, "step": 10728 }, { "epoch": 0.97, "learning_rate": 1.1732731221295035e-07, "loss": 1.2155, "step": 10729 }, { "epoch": 0.97, "learning_rate": 1.1663463427703803e-07, "loss": 1.207, "step": 10730 }, { "epoch": 0.97, "learning_rate": 1.1594400233727043e-07, "loss": 1.2479, "step": 10731 }, { "epoch": 0.97, "learning_rate": 1.1525541645044103e-07, "loss": 1.149, "step": 10732 }, { "epoch": 0.97, "learning_rate": 1.1456887667316007e-07, "loss": 1.2054, "step": 10733 }, { "epoch": 0.97, "learning_rate": 1.1388438306187965e-07, "loss": 1.196, "step": 10734 }, { "epoch": 0.97, "learning_rate": 1.1320193567288529e-07, "loss": 1.1915, "step": 10735 }, { "epoch": 0.97, "learning_rate": 1.1252153456228487e-07, "loss": 1.2098, "step": 10736 }, { "epoch": 0.97, "learning_rate": 1.1184317978602809e-07, "loss": 1.1955, "step": 10737 }, { "epoch": 0.97, "learning_rate": 1.1116687139988424e-07, "loss": 1.1935, "step": 10738 }, { "epoch": 0.97, "learning_rate": 1.1049260945946993e-07, "loss": 1.2011, "step": 10739 }, { "epoch": 0.97, "learning_rate": 1.0982039402022138e-07, "loss": 1.1971, "step": 10740 }, { "epoch": 0.97, "learning_rate": 1.0915022513740824e-07, "loss": 1.1986, "step": 10741 }, { "epoch": 0.97, "learning_rate": 1.0848210286614201e-07, "loss": 1.2227, "step": 10742 }, { "epoch": 0.97, "learning_rate": 1.0781602726134821e-07, "loss": 1.2155, "step": 10743 }, { "epoch": 0.97, "learning_rate": 1.0715199837779966e-07, "loss": 1.2006, "step": 10744 }, { "epoch": 0.97, "learning_rate": 1.0649001627009158e-07, "loss": 1.1969, "step": 10745 }, { "epoch": 0.97, "learning_rate": 1.0583008099265823e-07, "loss": 1.2272, "step": 10746 }, { "epoch": 0.97, "learning_rate": 1.0517219259976175e-07, "loss": 1.1915, "step": 10747 }, { "epoch": 0.97, "learning_rate": 1.0451635114549497e-07, "loss": 1.2437, "step": 10748 }, { "epoch": 0.97, "learning_rate": 1.0386255668378141e-07, "loss": 1.2184, "step": 10749 }, { "epoch": 0.97, "learning_rate": 1.0321080926837812e-07, "loss": 1.1931, "step": 10750 }, { "epoch": 0.97, "learning_rate": 1.025611089528783e-07, "loss": 1.1989, "step": 10751 }, { "epoch": 0.97, "learning_rate": 1.0191345579070032e-07, "loss": 1.1694, "step": 10752 }, { "epoch": 0.97, "learning_rate": 1.012678498350933e-07, "loss": 1.1973, "step": 10753 }, { "epoch": 0.97, "learning_rate": 1.0062429113914529e-07, "loss": 1.1906, "step": 10754 }, { "epoch": 0.97, "learning_rate": 9.998277975576952e-08, "loss": 1.1332, "step": 10755 }, { "epoch": 0.97, "learning_rate": 9.934331573771549e-08, "loss": 1.2344, "step": 10756 }, { "epoch": 0.97, "learning_rate": 9.870589913756056e-08, "loss": 1.1856, "step": 10757 }, { "epoch": 0.97, "learning_rate": 9.807053000771837e-08, "loss": 1.1793, "step": 10758 }, { "epoch": 0.97, "learning_rate": 9.743720840042491e-08, "loss": 1.1652, "step": 10759 }, { "epoch": 0.97, "learning_rate": 9.680593436776075e-08, "loss": 1.1714, "step": 10760 }, { "epoch": 0.97, "learning_rate": 9.617670796162603e-08, "loss": 1.2008, "step": 10761 }, { "epoch": 0.97, "learning_rate": 9.554952923375992e-08, "loss": 1.21, "step": 10762 }, { "epoch": 0.97, "learning_rate": 9.492439823573506e-08, "loss": 1.198, "step": 10763 }, { "epoch": 0.97, "learning_rate": 9.430131501894368e-08, "loss": 1.2145, "step": 10764 }, { "epoch": 0.97, "learning_rate": 9.368027963462534e-08, "loss": 1.2282, "step": 10765 }, { "epoch": 0.97, "learning_rate": 9.306129213383919e-08, "loss": 1.1771, "step": 10766 }, { "epoch": 0.97, "learning_rate": 9.24443525674834e-08, "loss": 1.1616, "step": 10767 }, { "epoch": 0.97, "learning_rate": 9.182946098628409e-08, "loss": 1.2295, "step": 10768 }, { "epoch": 0.97, "learning_rate": 9.121661744080079e-08, "loss": 1.2076, "step": 10769 }, { "epoch": 0.97, "learning_rate": 9.06058219814182e-08, "loss": 1.2016, "step": 10770 }, { "epoch": 0.97, "learning_rate": 8.99970746583656e-08, "loss": 1.2221, "step": 10771 }, { "epoch": 0.97, "learning_rate": 8.939037552169183e-08, "loss": 1.196, "step": 10772 }, { "epoch": 0.97, "learning_rate": 8.878572462128198e-08, "loss": 1.2534, "step": 10773 }, { "epoch": 0.97, "learning_rate": 8.818312200685186e-08, "loss": 1.211, "step": 10774 }, { "epoch": 0.97, "learning_rate": 8.758256772795348e-08, "loss": 1.2254, "step": 10775 }, { "epoch": 0.97, "learning_rate": 8.698406183396123e-08, "loss": 1.174, "step": 10776 }, { "epoch": 0.97, "learning_rate": 8.638760437409132e-08, "loss": 1.1451, "step": 10777 }, { "epoch": 0.97, "learning_rate": 8.579319539738506e-08, "loss": 1.2264, "step": 10778 }, { "epoch": 0.97, "learning_rate": 8.520083495271447e-08, "loss": 1.2294, "step": 10779 }, { "epoch": 0.97, "learning_rate": 8.461052308878503e-08, "loss": 1.2202, "step": 10780 }, { "epoch": 0.97, "learning_rate": 8.402225985413848e-08, "loss": 1.1602, "step": 10781 }, { "epoch": 0.97, "learning_rate": 8.343604529714166e-08, "loss": 1.172, "step": 10782 }, { "epoch": 0.97, "learning_rate": 8.285187946599215e-08, "loss": 1.1928, "step": 10783 }, { "epoch": 0.97, "learning_rate": 8.22697624087293e-08, "loss": 1.2338, "step": 10784 }, { "epoch": 0.97, "learning_rate": 8.168969417320927e-08, "loss": 1.1974, "step": 10785 }, { "epoch": 0.97, "learning_rate": 8.111167480713277e-08, "loss": 1.1444, "step": 10786 }, { "epoch": 0.97, "learning_rate": 8.053570435802293e-08, "loss": 1.1715, "step": 10787 }, { "epoch": 0.97, "learning_rate": 7.99617828732363e-08, "loss": 1.2417, "step": 10788 }, { "epoch": 0.97, "learning_rate": 7.938991039997124e-08, "loss": 1.1898, "step": 10789 }, { "epoch": 0.97, "learning_rate": 7.882008698524013e-08, "loss": 1.1254, "step": 10790 }, { "epoch": 0.97, "learning_rate": 7.825231267589994e-08, "loss": 1.1859, "step": 10791 }, { "epoch": 0.98, "learning_rate": 7.768658751863556e-08, "loss": 1.159, "step": 10792 }, { "epoch": 0.98, "learning_rate": 7.712291155995976e-08, "loss": 1.2058, "step": 10793 }, { "epoch": 0.98, "learning_rate": 7.656128484622438e-08, "loss": 1.1357, "step": 10794 }, { "epoch": 0.98, "learning_rate": 7.600170742360358e-08, "loss": 1.2033, "step": 10795 }, { "epoch": 0.98, "learning_rate": 7.544417933810777e-08, "loss": 1.1793, "step": 10796 }, { "epoch": 0.98, "learning_rate": 7.488870063558363e-08, "loss": 1.1782, "step": 10797 }, { "epoch": 0.98, "learning_rate": 7.433527136170293e-08, "loss": 1.1587, "step": 10798 }, { "epoch": 0.98, "learning_rate": 7.378389156196819e-08, "loss": 1.1999, "step": 10799 }, { "epoch": 0.98, "learning_rate": 7.323456128171535e-08, "loss": 1.1872, "step": 10800 }, { "epoch": 0.98, "learning_rate": 7.268728056611662e-08, "loss": 1.1976, "step": 10801 }, { "epoch": 0.98, "learning_rate": 7.214204946016656e-08, "loss": 1.218, "step": 10802 }, { "epoch": 0.98, "learning_rate": 7.159886800869875e-08, "loss": 1.2435, "step": 10803 }, { "epoch": 0.98, "learning_rate": 7.105773625637469e-08, "loss": 1.2181, "step": 10804 }, { "epoch": 0.98, "learning_rate": 7.051865424768933e-08, "loss": 1.1967, "step": 10805 }, { "epoch": 0.98, "learning_rate": 6.998162202696556e-08, "loss": 1.2535, "step": 10806 }, { "epoch": 0.98, "learning_rate": 6.944663963835974e-08, "loss": 1.2338, "step": 10807 }, { "epoch": 0.98, "learning_rate": 6.891370712586442e-08, "loss": 1.1501, "step": 10808 }, { "epoch": 0.98, "learning_rate": 6.83828245332918e-08, "loss": 1.2001, "step": 10809 }, { "epoch": 0.98, "learning_rate": 6.785399190430141e-08, "loss": 1.2026, "step": 10810 }, { "epoch": 0.98, "learning_rate": 6.732720928236957e-08, "loss": 1.1728, "step": 10811 }, { "epoch": 0.98, "learning_rate": 6.680247671081164e-08, "loss": 1.2267, "step": 10812 }, { "epoch": 0.98, "learning_rate": 6.627979423277364e-08, "loss": 1.1489, "step": 10813 }, { "epoch": 0.98, "learning_rate": 6.575916189122955e-08, "loss": 1.2036, "step": 10814 }, { "epoch": 0.98, "learning_rate": 6.524057972899233e-08, "loss": 1.1951, "step": 10815 }, { "epoch": 0.98, "learning_rate": 6.472404778870012e-08, "loss": 1.2457, "step": 10816 }, { "epoch": 0.98, "learning_rate": 6.42095661128217e-08, "loss": 1.1197, "step": 10817 }, { "epoch": 0.98, "learning_rate": 6.369713474366212e-08, "loss": 1.1436, "step": 10818 }, { "epoch": 0.98, "learning_rate": 6.318675372335159e-08, "loss": 1.1663, "step": 10819 }, { "epoch": 0.98, "learning_rate": 6.267842309386208e-08, "loss": 1.1678, "step": 10820 }, { "epoch": 0.98, "learning_rate": 6.217214289698237e-08, "loss": 1.2201, "step": 10821 }, { "epoch": 0.98, "learning_rate": 6.166791317434583e-08, "loss": 1.1984, "step": 10822 }, { "epoch": 0.98, "learning_rate": 6.116573396740822e-08, "loss": 1.1769, "step": 10823 }, { "epoch": 0.98, "learning_rate": 6.066560531746423e-08, "loss": 1.2445, "step": 10824 }, { "epoch": 0.98, "learning_rate": 6.016752726563656e-08, "loss": 1.1584, "step": 10825 }, { "epoch": 0.98, "learning_rate": 5.96714998528758e-08, "loss": 1.1681, "step": 10826 }, { "epoch": 0.98, "learning_rate": 5.917752311996594e-08, "loss": 1.1966, "step": 10827 }, { "epoch": 0.98, "learning_rate": 5.86855971075273e-08, "loss": 1.1836, "step": 10828 }, { "epoch": 0.98, "learning_rate": 5.8195721856005303e-08, "loss": 1.1813, "step": 10829 }, { "epoch": 0.98, "learning_rate": 5.770789740567883e-08, "loss": 1.2242, "step": 10830 }, { "epoch": 0.98, "learning_rate": 5.722212379666303e-08, "loss": 1.1925, "step": 10831 }, { "epoch": 0.98, "learning_rate": 5.6738401068895384e-08, "loss": 1.2082, "step": 10832 }, { "epoch": 0.98, "learning_rate": 5.6256729262149645e-08, "loss": 1.2069, "step": 10833 }, { "epoch": 0.98, "learning_rate": 5.577710841603023e-08, "loss": 1.2114, "step": 10834 }, { "epoch": 0.98, "learning_rate": 5.529953856997505e-08, "loss": 1.1338, "step": 10835 }, { "epoch": 0.98, "learning_rate": 5.4824019763252685e-08, "loss": 1.224, "step": 10836 }, { "epoch": 0.98, "learning_rate": 5.435055203495687e-08, "loss": 1.1512, "step": 10837 }, { "epoch": 0.98, "learning_rate": 5.387913542402312e-08, "loss": 1.207, "step": 10838 }, { "epoch": 0.98, "learning_rate": 5.340976996920932e-08, "loss": 1.1751, "step": 10839 }, { "epoch": 0.98, "learning_rate": 5.294245570910961e-08, "loss": 1.2161, "step": 10840 }, { "epoch": 0.98, "learning_rate": 5.247719268214879e-08, "loss": 1.2141, "step": 10841 }, { "epoch": 0.98, "learning_rate": 5.201398092658238e-08, "loss": 1.1455, "step": 10842 }, { "epoch": 0.98, "learning_rate": 5.155282048049381e-08, "loss": 1.1883, "step": 10843 }, { "epoch": 0.98, "learning_rate": 5.10937113818083e-08, "loss": 1.1454, "step": 10844 }, { "epoch": 0.98, "learning_rate": 5.0636653668267865e-08, "loss": 1.1443, "step": 10845 }, { "epoch": 0.98, "learning_rate": 5.018164737745912e-08, "loss": 1.1792, "step": 10846 }, { "epoch": 0.98, "learning_rate": 4.9728692546791024e-08, "loss": 1.1566, "step": 10847 }, { "epoch": 0.98, "learning_rate": 4.927778921350601e-08, "loss": 1.166, "step": 10848 }, { "epoch": 0.98, "learning_rate": 4.882893741468275e-08, "loss": 1.1841, "step": 10849 }, { "epoch": 0.98, "learning_rate": 4.8382137187227835e-08, "loss": 1.1241, "step": 10850 }, { "epoch": 0.98, "learning_rate": 4.7937388567872996e-08, "loss": 1.2259, "step": 10851 }, { "epoch": 0.98, "learning_rate": 4.749469159319175e-08, "loss": 1.19, "step": 10852 }, { "epoch": 0.98, "learning_rate": 4.705404629958277e-08, "loss": 1.2275, "step": 10853 }, { "epoch": 0.98, "learning_rate": 4.661545272327539e-08, "loss": 1.1251, "step": 10854 }, { "epoch": 0.98, "learning_rate": 4.617891090033521e-08, "loss": 1.2404, "step": 10855 }, { "epoch": 0.98, "learning_rate": 4.574442086665853e-08, "loss": 1.1621, "step": 10856 }, { "epoch": 0.98, "learning_rate": 4.5311982657963993e-08, "loss": 1.1848, "step": 10857 }, { "epoch": 0.98, "learning_rate": 4.4881596309812036e-08, "loss": 1.1915, "step": 10858 }, { "epoch": 0.98, "learning_rate": 4.445326185759102e-08, "loss": 1.194, "step": 10859 }, { "epoch": 0.98, "learning_rate": 4.402697933651723e-08, "loss": 1.2313, "step": 10860 }, { "epoch": 0.98, "learning_rate": 4.3602748781645944e-08, "loss": 1.2011, "step": 10861 }, { "epoch": 0.98, "learning_rate": 4.3180570227854824e-08, "loss": 1.2284, "step": 10862 }, { "epoch": 0.98, "learning_rate": 4.276044370985777e-08, "loss": 1.1566, "step": 10863 }, { "epoch": 0.98, "learning_rate": 4.234236926219659e-08, "loss": 1.1901, "step": 10864 }, { "epoch": 0.98, "learning_rate": 4.1926346919252127e-08, "loss": 1.1686, "step": 10865 }, { "epoch": 0.98, "learning_rate": 4.151237671523034e-08, "loss": 1.183, "step": 10866 }, { "epoch": 0.98, "learning_rate": 4.110045868416512e-08, "loss": 1.2272, "step": 10867 }, { "epoch": 0.98, "learning_rate": 4.06905928599266e-08, "loss": 1.14, "step": 10868 }, { "epoch": 0.98, "learning_rate": 4.0282779276218374e-08, "loss": 1.2253, "step": 10869 }, { "epoch": 0.98, "learning_rate": 3.987701796657195e-08, "loss": 1.1914, "step": 10870 }, { "epoch": 0.98, "learning_rate": 3.947330896434676e-08, "loss": 1.2485, "step": 10871 }, { "epoch": 0.98, "learning_rate": 3.907165230274124e-08, "loss": 1.2267, "step": 10872 }, { "epoch": 0.98, "learning_rate": 3.867204801477897e-08, "loss": 1.1701, "step": 10873 }, { "epoch": 0.98, "learning_rate": 3.827449613331702e-08, "loss": 1.1941, "step": 10874 }, { "epoch": 0.98, "learning_rate": 3.787899669104311e-08, "loss": 1.1625, "step": 10875 }, { "epoch": 0.98, "learning_rate": 3.7485549720478463e-08, "loss": 1.1714, "step": 10876 }, { "epoch": 0.98, "learning_rate": 3.7094155253972196e-08, "loss": 1.1649, "step": 10877 }, { "epoch": 0.98, "learning_rate": 3.670481332370412e-08, "loss": 1.1889, "step": 10878 }, { "epoch": 0.98, "learning_rate": 3.6317523961693076e-08, "loss": 1.1975, "step": 10879 }, { "epoch": 0.98, "learning_rate": 3.593228719977471e-08, "loss": 1.2513, "step": 10880 }, { "epoch": 0.98, "learning_rate": 3.5549103069632e-08, "loss": 1.2119, "step": 10881 }, { "epoch": 0.98, "learning_rate": 3.51679716027703e-08, "loss": 1.1979, "step": 10882 }, { "epoch": 0.98, "learning_rate": 3.4788892830525665e-08, "loss": 1.1846, "step": 10883 }, { "epoch": 0.98, "learning_rate": 3.44118667840676e-08, "loss": 1.2264, "step": 10884 }, { "epoch": 0.98, "learning_rate": 3.4036893494396314e-08, "loss": 1.2077, "step": 10885 }, { "epoch": 0.98, "learning_rate": 3.3663972992342694e-08, "loss": 1.1564, "step": 10886 }, { "epoch": 0.98, "learning_rate": 3.329310530857388e-08, "loss": 1.1963, "step": 10887 }, { "epoch": 0.98, "learning_rate": 3.292429047357937e-08, "loss": 1.2034, "step": 10888 }, { "epoch": 0.98, "learning_rate": 3.25575285176849e-08, "loss": 1.2422, "step": 10889 }, { "epoch": 0.98, "learning_rate": 3.219281947104691e-08, "loss": 1.1618, "step": 10890 }, { "epoch": 0.98, "learning_rate": 3.1830163363655296e-08, "loss": 1.2376, "step": 10891 }, { "epoch": 0.98, "learning_rate": 3.146956022532788e-08, "loss": 1.1857, "step": 10892 }, { "epoch": 0.98, "learning_rate": 3.111101008571593e-08, "loss": 1.1914, "step": 10893 }, { "epoch": 0.98, "learning_rate": 3.075451297429588e-08, "loss": 1.1433, "step": 10894 }, { "epoch": 0.98, "learning_rate": 3.040006892038594e-08, "loss": 1.2001, "step": 10895 }, { "epoch": 0.98, "learning_rate": 3.004767795312391e-08, "loss": 1.1839, "step": 10896 }, { "epoch": 0.98, "learning_rate": 2.969734010148939e-08, "loss": 1.2388, "step": 10897 }, { "epoch": 0.98, "learning_rate": 2.9349055394287118e-08, "loss": 1.2474, "step": 10898 }, { "epoch": 0.98, "learning_rate": 2.9002823860155292e-08, "loss": 1.2027, "step": 10899 }, { "epoch": 0.98, "learning_rate": 2.8658645527560037e-08, "loss": 1.2172, "step": 10900 }, { "epoch": 0.98, "learning_rate": 2.831652042480093e-08, "loss": 1.1917, "step": 10901 }, { "epoch": 0.99, "learning_rate": 2.797644858000825e-08, "loss": 1.1292, "step": 10902 }, { "epoch": 0.99, "learning_rate": 2.763843002114852e-08, "loss": 1.1807, "step": 10903 }, { "epoch": 0.99, "learning_rate": 2.7302464776010615e-08, "loss": 1.1845, "step": 10904 }, { "epoch": 0.99, "learning_rate": 2.696855287221689e-08, "loss": 1.2614, "step": 10905 }, { "epoch": 0.99, "learning_rate": 2.663669433722871e-08, "loss": 1.23, "step": 10906 }, { "epoch": 0.99, "learning_rate": 2.6306889198327024e-08, "loss": 1.2215, "step": 10907 }, { "epoch": 0.99, "learning_rate": 2.597913748263181e-08, "loss": 1.1563, "step": 10908 }, { "epoch": 0.99, "learning_rate": 2.5653439217093733e-08, "loss": 1.1348, "step": 10909 }, { "epoch": 0.99, "learning_rate": 2.5329794428488596e-08, "loss": 1.1927, "step": 10910 }, { "epoch": 0.99, "learning_rate": 2.500820314343122e-08, "loss": 1.177, "step": 10911 }, { "epoch": 0.99, "learning_rate": 2.468866538836434e-08, "loss": 1.2071, "step": 10912 }, { "epoch": 0.99, "learning_rate": 2.437118118955861e-08, "loss": 1.155, "step": 10913 }, { "epoch": 0.99, "learning_rate": 2.4055750573118152e-08, "loss": 1.1779, "step": 10914 }, { "epoch": 0.99, "learning_rate": 2.3742373564983322e-08, "loss": 1.1505, "step": 10915 }, { "epoch": 0.99, "learning_rate": 2.3431050190916848e-08, "loss": 1.2232, "step": 10916 }, { "epoch": 0.99, "learning_rate": 2.3121780476520473e-08, "loss": 1.2247, "step": 10917 }, { "epoch": 0.99, "learning_rate": 2.28145644472183e-08, "loss": 1.1966, "step": 10918 }, { "epoch": 0.99, "learning_rate": 2.250940212827346e-08, "loss": 1.2029, "step": 10919 }, { "epoch": 0.99, "learning_rate": 2.2206293544779766e-08, "loss": 1.1619, "step": 10920 }, { "epoch": 0.99, "learning_rate": 2.1905238721656172e-08, "loss": 1.206, "step": 10921 }, { "epoch": 0.99, "learning_rate": 2.1606237683660658e-08, "loss": 1.2256, "step": 10922 }, { "epoch": 0.99, "learning_rate": 2.1309290455373554e-08, "loss": 1.2068, "step": 10923 }, { "epoch": 0.99, "learning_rate": 2.1014397061211444e-08, "loss": 1.2364, "step": 10924 }, { "epoch": 0.99, "learning_rate": 2.0721557525424374e-08, "loss": 1.1895, "step": 10925 }, { "epoch": 0.99, "learning_rate": 2.043077187208753e-08, "loss": 1.2214, "step": 10926 }, { "epoch": 0.99, "learning_rate": 2.0142040125112337e-08, "loss": 1.2018, "step": 10927 }, { "epoch": 0.99, "learning_rate": 1.9855362308240922e-08, "loss": 1.1854, "step": 10928 }, { "epoch": 0.99, "learning_rate": 1.9570738445040537e-08, "loss": 1.2018, "step": 10929 }, { "epoch": 0.99, "learning_rate": 1.9288168558917464e-08, "loss": 1.2034, "step": 10930 }, { "epoch": 0.99, "learning_rate": 1.9007652673103115e-08, "loss": 1.1904, "step": 10931 }, { "epoch": 0.99, "learning_rate": 1.8729190810665154e-08, "loss": 1.2236, "step": 10932 }, { "epoch": 0.99, "learning_rate": 1.8452782994496377e-08, "loss": 1.1944, "step": 10933 }, { "epoch": 0.99, "learning_rate": 1.8178429247325822e-08, "loss": 1.1739, "step": 10934 }, { "epoch": 0.99, "learning_rate": 1.790612959171323e-08, "loss": 1.1747, "step": 10935 }, { "epoch": 0.99, "learning_rate": 1.7635884050046236e-08, "loss": 1.1836, "step": 10936 }, { "epoch": 0.99, "learning_rate": 1.7367692644545964e-08, "loss": 1.2124, "step": 10937 }, { "epoch": 0.99, "learning_rate": 1.7101555397264213e-08, "loss": 1.1867, "step": 10938 }, { "epoch": 0.99, "learning_rate": 1.6837472330080705e-08, "loss": 1.1494, "step": 10939 }, { "epoch": 0.99, "learning_rate": 1.657544346471418e-08, "loss": 1.1545, "step": 10940 }, { "epoch": 0.99, "learning_rate": 1.631546882270574e-08, "loss": 1.2117, "step": 10941 }, { "epoch": 0.99, "learning_rate": 1.6057548425435498e-08, "loss": 1.2425, "step": 10942 }, { "epoch": 0.99, "learning_rate": 1.580168229410317e-08, "loss": 1.1583, "step": 10943 }, { "epoch": 0.99, "learning_rate": 1.5547870449755807e-08, "loss": 1.177, "step": 10944 }, { "epoch": 0.99, "learning_rate": 1.5296112913257278e-08, "loss": 1.1875, "step": 10945 }, { "epoch": 0.99, "learning_rate": 1.504640970531046e-08, "loss": 1.168, "step": 10946 }, { "epoch": 0.99, "learning_rate": 1.4798760846443383e-08, "loss": 1.2515, "step": 10947 }, { "epoch": 0.99, "learning_rate": 1.4553166357020309e-08, "loss": 1.126, "step": 10948 }, { "epoch": 0.99, "learning_rate": 1.4309626257238973e-08, "loss": 1.1868, "step": 10949 }, { "epoch": 0.99, "learning_rate": 1.4068140567116695e-08, "loss": 1.1573, "step": 10950 }, { "epoch": 0.99, "learning_rate": 1.3828709306515363e-08, "loss": 1.2346, "step": 10951 }, { "epoch": 0.99, "learning_rate": 1.3591332495119235e-08, "loss": 1.1902, "step": 10952 }, { "epoch": 0.99, "learning_rate": 1.335601015244603e-08, "loss": 1.2035, "step": 10953 }, { "epoch": 0.99, "learning_rate": 1.3122742297844159e-08, "loss": 1.1881, "step": 10954 }, { "epoch": 0.99, "learning_rate": 1.2891528950498277e-08, "loss": 1.1793, "step": 10955 }, { "epoch": 0.99, "learning_rate": 1.2662370129412627e-08, "loss": 1.1581, "step": 10956 }, { "epoch": 0.99, "learning_rate": 1.2435265853436017e-08, "loss": 1.1962, "step": 10957 }, { "epoch": 0.99, "learning_rate": 1.2210216141236852e-08, "loss": 1.1854, "step": 10958 }, { "epoch": 0.99, "learning_rate": 1.1987221011319771e-08, "loss": 1.1855, "step": 10959 }, { "epoch": 0.99, "learning_rate": 1.1766280482022885e-08, "loss": 1.2117, "step": 10960 }, { "epoch": 0.99, "learning_rate": 1.1547394571512215e-08, "loss": 1.1834, "step": 10961 }, { "epoch": 0.99, "learning_rate": 1.133056329778448e-08, "loss": 1.2258, "step": 10962 }, { "epoch": 0.99, "learning_rate": 1.1115786678667085e-08, "loss": 1.1748, "step": 10963 }, { "epoch": 0.99, "learning_rate": 1.09030647318209e-08, "loss": 1.169, "step": 10964 }, { "epoch": 0.99, "learning_rate": 1.0692397474740268e-08, "loss": 1.2572, "step": 10965 }, { "epoch": 0.99, "learning_rate": 1.0483784924739115e-08, "loss": 1.2032, "step": 10966 }, { "epoch": 0.99, "learning_rate": 1.0277227098978714e-08, "loss": 1.1455, "step": 10967 }, { "epoch": 0.99, "learning_rate": 1.0072724014437152e-08, "loss": 1.1888, "step": 10968 }, { "epoch": 0.99, "learning_rate": 9.870275687928753e-09, "loss": 1.123, "step": 10969 }, { "epoch": 0.99, "learning_rate": 9.669882136104092e-09, "loss": 1.2348, "step": 10970 }, { "epoch": 0.99, "learning_rate": 9.471543375438874e-09, "loss": 1.2253, "step": 10971 }, { "epoch": 0.99, "learning_rate": 9.275259422239502e-09, "loss": 1.1888, "step": 10972 }, { "epoch": 0.99, "learning_rate": 9.081030292645842e-09, "loss": 1.2371, "step": 10973 }, { "epoch": 0.99, "learning_rate": 8.888856002628454e-09, "loss": 1.23, "step": 10974 }, { "epoch": 0.99, "learning_rate": 8.698736567988586e-09, "loss": 1.1955, "step": 10975 }, { "epoch": 0.99, "learning_rate": 8.510672004358177e-09, "loss": 1.1908, "step": 10976 }, { "epoch": 0.99, "learning_rate": 8.32466232719986e-09, "loss": 1.2184, "step": 10977 }, { "epoch": 0.99, "learning_rate": 8.140707551809734e-09, "loss": 1.1524, "step": 10978 }, { "epoch": 0.99, "learning_rate": 7.95880769331181e-09, "loss": 1.2087, "step": 10979 }, { "epoch": 0.99, "learning_rate": 7.778962766663567e-09, "loss": 1.1365, "step": 10980 }, { "epoch": 0.99, "learning_rate": 7.601172786650402e-09, "loss": 1.2093, "step": 10981 }, { "epoch": 0.99, "learning_rate": 7.42543776789395e-09, "loss": 1.1414, "step": 10982 }, { "epoch": 0.99, "learning_rate": 7.2517577248409905e-09, "loss": 1.1403, "step": 10983 }, { "epoch": 0.99, "learning_rate": 7.080132671774542e-09, "loss": 1.1857, "step": 10984 }, { "epoch": 0.99, "learning_rate": 6.910562622802763e-09, "loss": 1.1864, "step": 10985 }, { "epoch": 0.99, "learning_rate": 6.743047591870055e-09, "loss": 1.1858, "step": 10986 }, { "epoch": 0.99, "learning_rate": 6.577587592751511e-09, "loss": 1.1362, "step": 10987 }, { "epoch": 0.99, "learning_rate": 6.414182639052912e-09, "loss": 1.1768, "step": 10988 }, { "epoch": 0.99, "learning_rate": 6.252832744205184e-09, "loss": 1.188, "step": 10989 }, { "epoch": 0.99, "learning_rate": 6.093537921478265e-09, "loss": 1.1566, "step": 10990 }, { "epoch": 0.99, "learning_rate": 5.936298183967237e-09, "loss": 1.164, "step": 10991 }, { "epoch": 0.99, "learning_rate": 5.781113544606198e-09, "loss": 1.15, "step": 10992 }, { "epoch": 0.99, "learning_rate": 5.62798401615161e-09, "loss": 1.1821, "step": 10993 }, { "epoch": 0.99, "learning_rate": 5.476909611193404e-09, "loss": 1.1498, "step": 10994 }, { "epoch": 0.99, "learning_rate": 5.327890342152197e-09, "loss": 1.1919, "step": 10995 }, { "epoch": 0.99, "learning_rate": 5.180926221284854e-09, "loss": 1.1664, "step": 10996 }, { "epoch": 0.99, "learning_rate": 5.0360172606761514e-09, "loss": 1.2009, "step": 10997 }, { "epoch": 0.99, "learning_rate": 4.8931634722360065e-09, "loss": 1.2271, "step": 10998 }, { "epoch": 0.99, "learning_rate": 4.752364867713355e-09, "loss": 1.1902, "step": 10999 }, { "epoch": 0.99, "learning_rate": 4.6136214586822714e-09, "loss": 1.1841, "step": 11000 }, { "epoch": 0.99, "learning_rate": 4.4769332565558485e-09, "loss": 1.1735, "step": 11001 }, { "epoch": 0.99, "learning_rate": 4.3423002725695435e-09, "loss": 1.2336, "step": 11002 }, { "epoch": 0.99, "learning_rate": 4.2097225177950565e-09, "loss": 1.2087, "step": 11003 }, { "epoch": 0.99, "learning_rate": 4.079200003129224e-09, "loss": 1.2112, "step": 11004 }, { "epoch": 0.99, "learning_rate": 3.950732739310681e-09, "loss": 1.1482, "step": 11005 }, { "epoch": 0.99, "learning_rate": 3.8243207368976466e-09, "loss": 1.2347, "step": 11006 }, { "epoch": 0.99, "learning_rate": 3.6999640062845843e-09, "loss": 1.1346, "step": 11007 }, { "epoch": 0.99, "learning_rate": 3.5776625576966482e-09, "loss": 1.126, "step": 11008 }, { "epoch": 0.99, "learning_rate": 3.4574164011924594e-09, "loss": 1.2216, "step": 11009 }, { "epoch": 0.99, "learning_rate": 3.3392255466557775e-09, "loss": 1.1946, "step": 11010 }, { "epoch": 0.99, "learning_rate": 3.223090003806606e-09, "loss": 1.1603, "step": 11011 }, { "epoch": 0.99, "learning_rate": 3.1090097821956377e-09, "loss": 1.1428, "step": 11012 }, { "epoch": 1.0, "learning_rate": 2.9969848911987065e-09, "loss": 1.1792, "step": 11013 }, { "epoch": 1.0, "learning_rate": 2.8870153400306633e-09, "loss": 1.2201, "step": 11014 }, { "epoch": 1.0, "learning_rate": 2.7791011377287235e-09, "loss": 1.2223, "step": 11015 }, { "epoch": 1.0, "learning_rate": 2.673242293171896e-09, "loss": 1.1427, "step": 11016 }, { "epoch": 1.0, "learning_rate": 2.5694388150615534e-09, "loss": 1.1979, "step": 11017 }, { "epoch": 1.0, "learning_rate": 2.4676907119325353e-09, "loss": 1.1848, "step": 11018 }, { "epoch": 1.0, "learning_rate": 2.3679979921503727e-09, "loss": 1.1858, "step": 11019 }, { "epoch": 1.0, "learning_rate": 2.2703606639112864e-09, "loss": 1.2189, "step": 11020 }, { "epoch": 1.0, "learning_rate": 2.17477873524774e-09, "loss": 1.1964, "step": 11021 }, { "epoch": 1.0, "learning_rate": 2.081252214014562e-09, "loss": 1.1844, "step": 11022 }, { "epoch": 1.0, "learning_rate": 1.9897811079055972e-09, "loss": 1.1421, "step": 11023 }, { "epoch": 1.0, "learning_rate": 1.900365424437056e-09, "loss": 1.1588, "step": 11024 }, { "epoch": 1.0, "learning_rate": 1.813005170964166e-09, "loss": 1.1725, "step": 11025 }, { "epoch": 1.0, "learning_rate": 1.7277003546700699e-09, "loss": 1.1895, "step": 11026 }, { "epoch": 1.0, "learning_rate": 1.6444509825686016e-09, "loss": 1.1874, "step": 11027 }, { "epoch": 1.0, "learning_rate": 1.5632570615042862e-09, "loss": 1.1423, "step": 11028 }, { "epoch": 1.0, "learning_rate": 1.4841185981523397e-09, "loss": 1.1372, "step": 11029 }, { "epoch": 1.0, "learning_rate": 1.4070355990214445e-09, "loss": 1.2094, "step": 11030 }, { "epoch": 1.0, "learning_rate": 1.3320080704481986e-09, "loss": 1.2233, "step": 11031 }, { "epoch": 1.0, "learning_rate": 1.2590360186026662e-09, "loss": 1.1436, "step": 11032 }, { "epoch": 1.0, "learning_rate": 1.188119449485603e-09, "loss": 1.1839, "step": 11033 }, { "epoch": 1.0, "learning_rate": 1.1192583689256796e-09, "loss": 1.1546, "step": 11034 }, { "epoch": 1.0, "learning_rate": 1.052452782585034e-09, "loss": 1.2353, "step": 11035 }, { "epoch": 1.0, "learning_rate": 9.877026959592695e-10, "loss": 1.1949, "step": 11036 }, { "epoch": 1.0, "learning_rate": 9.250081143719058e-10, "loss": 1.2243, "step": 11037 }, { "epoch": 1.0, "learning_rate": 8.643690429743779e-10, "loss": 1.2141, "step": 11038 }, { "epoch": 1.0, "learning_rate": 8.057854867571379e-10, "loss": 1.1602, "step": 11039 }, { "epoch": 1.0, "learning_rate": 7.49257450533003e-10, "loss": 1.1779, "step": 11040 }, { "epoch": 1.0, "learning_rate": 6.947849389538074e-10, "loss": 1.2321, "step": 11041 }, { "epoch": 1.0, "learning_rate": 6.423679564965257e-10, "loss": 1.1646, "step": 11042 }, { "epoch": 1.0, "learning_rate": 5.920065074715986e-10, "loss": 1.1734, "step": 11043 }, { "epoch": 1.0, "learning_rate": 5.43700596020158e-10, "loss": 1.2082, "step": 11044 }, { "epoch": 1.0, "learning_rate": 4.974502261112512e-10, "loss": 1.2052, "step": 11045 }, { "epoch": 1.0, "learning_rate": 4.5325540155016777e-10, "loss": 1.1354, "step": 11046 }, { "epoch": 1.0, "learning_rate": 4.1111612597288794e-10, "loss": 1.227, "step": 11047 }, { "epoch": 1.0, "learning_rate": 3.710324028377565e-10, "loss": 1.1945, "step": 11048 }, { "epoch": 1.0, "learning_rate": 3.33004235447687e-10, "loss": 1.1623, "step": 11049 }, { "epoch": 1.0, "learning_rate": 2.9703162692795717e-10, "loss": 1.2172, "step": 11050 }, { "epoch": 1.0, "learning_rate": 2.6311458023176026e-10, "loss": 1.1786, "step": 11051 }, { "epoch": 1.0, "learning_rate": 2.312530981540828e-10, "loss": 1.2336, "step": 11052 }, { "epoch": 1.0, "learning_rate": 2.0144718330949996e-10, "loss": 1.1388, "step": 11053 }, { "epoch": 1.0, "learning_rate": 1.7369683815160466e-10, "loss": 1.18, "step": 11054 }, { "epoch": 1.0, "learning_rate": 1.480020649591296e-10, "loss": 1.2175, "step": 11055 }, { "epoch": 1.0, "learning_rate": 1.2436286584982527e-10, "loss": 1.1674, "step": 11056 }, { "epoch": 1.0, "learning_rate": 1.0277924276380635e-10, "loss": 1.1433, "step": 11057 }, { "epoch": 1.0, "learning_rate": 8.325119747742971e-11, "loss": 1.228, "step": 11058 }, { "epoch": 1.0, "learning_rate": 6.577873159496761e-11, "loss": 1.178, "step": 11059 }, { "epoch": 1.0, "learning_rate": 5.036184655415887e-11, "loss": 1.2088, "step": 11060 }, { "epoch": 1.0, "learning_rate": 3.700054362065775e-11, "loss": 1.2027, "step": 11061 }, { "epoch": 1.0, "learning_rate": 2.5694823896360577e-11, "loss": 1.1579, "step": 11062 }, { "epoch": 1.0, "learning_rate": 1.644468830830359e-11, "loss": 1.1927, "step": 11063 }, { "epoch": 1.0, "learning_rate": 9.250137616989563e-12, "loss": 1.2267, "step": 11064 }, { "epoch": 1.0, "learning_rate": 4.111172413612252e-12, "loss": 1.1807, "step": 11065 }, { "epoch": 1.0, "learning_rate": 1.0277931256075236e-12, "loss": 1.2327, "step": 11066 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.1903, "step": 11067 }, { "epoch": 1.0, "step": 11067, "total_flos": 6.249348806081511e+18, "train_loss": 1.2857411102603593, "train_runtime": 86507.6926, "train_samples_per_second": 49.129, "train_steps_per_second": 0.128 } ], "max_steps": 11067, "num_train_epochs": 1, "total_flos": 6.249348806081511e+18, "trial_name": null, "trial_params": null }