{ "best_metric": null, "best_model_checkpoint": null, "epoch": 19.999615399407716, "global_step": 65000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0, "loss": 1.8412, "step": 1 }, { "epoch": 0.0, "learning_rate": 0.0, "loss": 2.4423, "step": 2 }, { "epoch": 0.0, "learning_rate": 2.0408163265306121e-07, "loss": 2.1726, "step": 3 }, { "epoch": 0.0, "learning_rate": 4.0816326530612243e-07, "loss": 2.6872, "step": 4 }, { "epoch": 0.0, "learning_rate": 6.122448979591837e-07, "loss": 2.1498, "step": 5 }, { "epoch": 0.0, "learning_rate": 8.163265306122449e-07, "loss": 2.6042, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.0204081632653063e-06, "loss": 2.3096, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.2244897959183673e-06, "loss": 2.4673, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.4285714285714286e-06, "loss": 2.2869, "step": 9 }, { "epoch": 0.0, "learning_rate": 1.6326530612244897e-06, "loss": 2.2953, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.8367346938775512e-06, "loss": 1.9851, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.0408163265306125e-06, "loss": 1.9965, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.244897959183674e-06, "loss": 1.7586, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.4489795918367347e-06, "loss": 2.4944, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.6530612244897964e-06, "loss": 2.0767, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.8571428571428573e-06, "loss": 2.6923, "step": 16 }, { "epoch": 0.01, "learning_rate": 3.0612244897959185e-06, "loss": 2.099, "step": 17 }, { "epoch": 0.01, "learning_rate": 3.2653061224489794e-06, "loss": 1.956, "step": 18 }, { "epoch": 0.01, "learning_rate": 3.469387755102041e-06, "loss": 1.7529, "step": 19 }, { "epoch": 0.01, "learning_rate": 3.6734693877551024e-06, "loss": 1.487, "step": 20 }, { "epoch": 0.01, "learning_rate": 3.877551020408164e-06, "loss": 1.518, "step": 21 }, { "epoch": 0.01, "learning_rate": 4.081632653061225e-06, "loss": 1.447, "step": 22 }, { "epoch": 0.01, "learning_rate": 4.2857142857142855e-06, "loss": 1.7908, "step": 23 }, { "epoch": 0.01, "learning_rate": 4.489795918367348e-06, "loss": 1.5428, "step": 24 }, { "epoch": 0.01, "learning_rate": 4.693877551020409e-06, "loss": 1.4308, "step": 25 }, { "epoch": 0.01, "learning_rate": 4.897959183673469e-06, "loss": 1.3911, "step": 26 }, { "epoch": 0.01, "learning_rate": 5.1020408163265315e-06, "loss": 1.6423, "step": 27 }, { "epoch": 0.01, "learning_rate": 5.306122448979593e-06, "loss": 1.6353, "step": 28 }, { "epoch": 0.01, "learning_rate": 5.510204081632653e-06, "loss": 1.9413, "step": 29 }, { "epoch": 0.01, "learning_rate": 5.7142857142857145e-06, "loss": 1.5212, "step": 30 }, { "epoch": 0.01, "learning_rate": 5.918367346938776e-06, "loss": 1.5408, "step": 31 }, { "epoch": 0.01, "learning_rate": 6.122448979591837e-06, "loss": 1.6166, "step": 32 }, { "epoch": 0.01, "learning_rate": 6.326530612244899e-06, "loss": 1.8031, "step": 33 }, { "epoch": 0.01, "learning_rate": 6.530612244897959e-06, "loss": 1.934, "step": 34 }, { "epoch": 0.01, "learning_rate": 6.734693877551021e-06, "loss": 1.4541, "step": 35 }, { "epoch": 0.01, "learning_rate": 6.938775510204082e-06, "loss": 1.5663, "step": 36 }, { "epoch": 0.01, "learning_rate": 7.1428571428571436e-06, "loss": 1.4314, "step": 37 }, { "epoch": 0.01, "learning_rate": 7.346938775510205e-06, "loss": 1.4307, "step": 38 }, { "epoch": 0.01, "learning_rate": 7.551020408163265e-06, "loss": 1.7589, "step": 39 }, { "epoch": 0.01, "learning_rate": 7.755102040816327e-06, "loss": 1.7896, "step": 40 }, { "epoch": 0.01, "learning_rate": 7.959183673469388e-06, "loss": 1.4308, "step": 41 }, { "epoch": 0.01, "learning_rate": 8.16326530612245e-06, "loss": 1.7288, "step": 42 }, { "epoch": 0.01, "learning_rate": 8.36734693877551e-06, "loss": 1.6473, "step": 43 }, { "epoch": 0.01, "learning_rate": 8.571428571428571e-06, "loss": 1.757, "step": 44 }, { "epoch": 0.01, "learning_rate": 8.775510204081633e-06, "loss": 1.369, "step": 45 }, { "epoch": 0.01, "learning_rate": 8.979591836734695e-06, "loss": 1.827, "step": 46 }, { "epoch": 0.01, "learning_rate": 9.183673469387756e-06, "loss": 2.2143, "step": 47 }, { "epoch": 0.01, "learning_rate": 9.387755102040818e-06, "loss": 1.781, "step": 48 }, { "epoch": 0.02, "learning_rate": 9.591836734693878e-06, "loss": 1.6623, "step": 49 }, { "epoch": 0.02, "learning_rate": 9.795918367346939e-06, "loss": 1.6225, "step": 50 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 1.7036, "step": 51 }, { "epoch": 0.02, "learning_rate": 1.0204081632653063e-05, "loss": 1.4179, "step": 52 }, { "epoch": 0.02, "learning_rate": 1.0408163265306123e-05, "loss": 1.4681, "step": 53 }, { "epoch": 0.02, "learning_rate": 1.0612244897959186e-05, "loss": 1.4616, "step": 54 }, { "epoch": 0.02, "learning_rate": 1.0816326530612246e-05, "loss": 1.6009, "step": 55 }, { "epoch": 0.02, "learning_rate": 1.1020408163265306e-05, "loss": 1.2498, "step": 56 }, { "epoch": 0.02, "learning_rate": 1.1224489795918367e-05, "loss": 1.7834, "step": 57 }, { "epoch": 0.02, "learning_rate": 1.1428571428571429e-05, "loss": 1.5594, "step": 58 }, { "epoch": 0.02, "learning_rate": 1.1632653061224491e-05, "loss": 1.7175, "step": 59 }, { "epoch": 0.02, "learning_rate": 1.1836734693877552e-05, "loss": 1.5544, "step": 60 }, { "epoch": 0.02, "learning_rate": 1.2040816326530614e-05, "loss": 2.25, "step": 61 }, { "epoch": 0.02, "learning_rate": 1.2244897959183674e-05, "loss": 1.5575, "step": 62 }, { "epoch": 0.02, "learning_rate": 1.2448979591836736e-05, "loss": 1.6405, "step": 63 }, { "epoch": 0.02, "learning_rate": 1.2653061224489798e-05, "loss": 1.5264, "step": 64 }, { "epoch": 0.02, "learning_rate": 1.2857142857142859e-05, "loss": 1.2849, "step": 65 }, { "epoch": 0.02, "learning_rate": 1.3061224489795918e-05, "loss": 1.5249, "step": 66 }, { "epoch": 0.02, "learning_rate": 1.326530612244898e-05, "loss": 1.5535, "step": 67 }, { "epoch": 0.02, "learning_rate": 1.3469387755102042e-05, "loss": 1.6305, "step": 68 }, { "epoch": 0.02, "learning_rate": 1.3673469387755102e-05, "loss": 1.4558, "step": 69 }, { "epoch": 0.02, "learning_rate": 1.3877551020408165e-05, "loss": 1.1902, "step": 70 }, { "epoch": 0.02, "learning_rate": 1.4081632653061225e-05, "loss": 1.3432, "step": 71 }, { "epoch": 0.02, "learning_rate": 1.4285714285714287e-05, "loss": 1.593, "step": 72 }, { "epoch": 0.02, "learning_rate": 1.448979591836735e-05, "loss": 1.5448, "step": 73 }, { "epoch": 0.02, "learning_rate": 1.469387755102041e-05, "loss": 1.5227, "step": 74 }, { "epoch": 0.02, "learning_rate": 1.4897959183673472e-05, "loss": 1.8562, "step": 75 }, { "epoch": 0.02, "learning_rate": 1.510204081632653e-05, "loss": 1.7151, "step": 76 }, { "epoch": 0.02, "learning_rate": 1.530612244897959e-05, "loss": 1.5046, "step": 77 }, { "epoch": 0.02, "learning_rate": 1.5510204081632655e-05, "loss": 1.6402, "step": 78 }, { "epoch": 0.02, "learning_rate": 1.5714285714285715e-05, "loss": 1.5564, "step": 79 }, { "epoch": 0.02, "learning_rate": 1.5918367346938776e-05, "loss": 1.8259, "step": 80 }, { "epoch": 0.02, "learning_rate": 1.612244897959184e-05, "loss": 1.6569, "step": 81 }, { "epoch": 0.03, "learning_rate": 1.63265306122449e-05, "loss": 1.5577, "step": 82 }, { "epoch": 0.03, "learning_rate": 1.653061224489796e-05, "loss": 1.7821, "step": 83 }, { "epoch": 0.03, "learning_rate": 1.673469387755102e-05, "loss": 1.4537, "step": 84 }, { "epoch": 0.03, "learning_rate": 1.6938775510204085e-05, "loss": 2.3595, "step": 85 }, { "epoch": 0.03, "learning_rate": 1.7142857142857142e-05, "loss": 1.7363, "step": 86 }, { "epoch": 0.03, "learning_rate": 1.7346938775510206e-05, "loss": 1.6054, "step": 87 }, { "epoch": 0.03, "learning_rate": 1.7551020408163266e-05, "loss": 1.343, "step": 88 }, { "epoch": 0.03, "learning_rate": 1.7755102040816327e-05, "loss": 1.822, "step": 89 }, { "epoch": 0.03, "learning_rate": 1.795918367346939e-05, "loss": 1.3541, "step": 90 }, { "epoch": 0.03, "learning_rate": 1.816326530612245e-05, "loss": 1.6902, "step": 91 }, { "epoch": 0.03, "learning_rate": 1.836734693877551e-05, "loss": 1.9583, "step": 92 }, { "epoch": 0.03, "learning_rate": 1.8571428571428575e-05, "loss": 1.9069, "step": 93 }, { "epoch": 0.03, "learning_rate": 1.8775510204081636e-05, "loss": 1.472, "step": 94 }, { "epoch": 0.03, "learning_rate": 1.8979591836734696e-05, "loss": 1.7204, "step": 95 }, { "epoch": 0.03, "learning_rate": 1.9183673469387756e-05, "loss": 1.6665, "step": 96 }, { "epoch": 0.03, "learning_rate": 1.9387755102040817e-05, "loss": 1.4465, "step": 97 }, { "epoch": 0.03, "learning_rate": 1.9591836734693877e-05, "loss": 1.8124, "step": 98 }, { "epoch": 0.03, "learning_rate": 1.979591836734694e-05, "loss": 1.7005, "step": 99 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 1.5046, "step": 100 }, { "epoch": 0.03, "learning_rate": 1.9999995032964135e-05, "loss": 1.3994, "step": 101 }, { "epoch": 0.03, "learning_rate": 1.999998013186146e-05, "loss": 2.3198, "step": 102 }, { "epoch": 0.03, "learning_rate": 1.9999955296706797e-05, "loss": 1.9909, "step": 103 }, { "epoch": 0.03, "learning_rate": 1.99999205275248e-05, "loss": 1.495, "step": 104 }, { "epoch": 0.03, "learning_rate": 1.9999875824350018e-05, "loss": 1.6305, "step": 105 }, { "epoch": 0.03, "learning_rate": 1.9999821187226853e-05, "loss": 1.594, "step": 106 }, { "epoch": 0.03, "learning_rate": 1.999975661620959e-05, "loss": 1.4535, "step": 107 }, { "epoch": 0.03, "learning_rate": 1.9999682111362368e-05, "loss": 1.8067, "step": 108 }, { "epoch": 0.03, "learning_rate": 1.9999597672759204e-05, "loss": 1.3976, "step": 109 }, { "epoch": 0.03, "learning_rate": 1.9999503300483978e-05, "loss": 1.42, "step": 110 }, { "epoch": 0.03, "learning_rate": 1.9999398994630442e-05, "loss": 1.5618, "step": 111 }, { "epoch": 0.03, "learning_rate": 1.9999284755302213e-05, "loss": 1.8627, "step": 112 }, { "epoch": 0.03, "learning_rate": 1.9999160582612775e-05, "loss": 1.8137, "step": 113 }, { "epoch": 0.04, "learning_rate": 1.9999026476685484e-05, "loss": 1.7382, "step": 114 }, { "epoch": 0.04, "learning_rate": 1.9998882437653565e-05, "loss": 1.7163, "step": 115 }, { "epoch": 0.04, "learning_rate": 1.9998728465660105e-05, "loss": 1.7374, "step": 116 }, { "epoch": 0.04, "learning_rate": 1.999856456085806e-05, "loss": 1.5717, "step": 117 }, { "epoch": 0.04, "learning_rate": 1.999839072341025e-05, "loss": 1.4653, "step": 118 }, { "epoch": 0.04, "learning_rate": 1.9998206953489374e-05, "loss": 1.5283, "step": 119 }, { "epoch": 0.04, "learning_rate": 1.9998013251277988e-05, "loss": 1.399, "step": 120 }, { "epoch": 0.04, "learning_rate": 1.9997809616968514e-05, "loss": 1.8169, "step": 121 }, { "epoch": 0.04, "learning_rate": 1.9997596050763253e-05, "loss": 1.7379, "step": 122 }, { "epoch": 0.04, "learning_rate": 1.999737255287435e-05, "loss": 1.2735, "step": 123 }, { "epoch": 0.04, "learning_rate": 1.999713912352384e-05, "loss": 1.6284, "step": 124 }, { "epoch": 0.04, "learning_rate": 1.9996895762943604e-05, "loss": 1.8106, "step": 125 }, { "epoch": 0.04, "learning_rate": 1.9996642471375406e-05, "loss": 1.628, "step": 126 }, { "epoch": 0.04, "learning_rate": 1.9996379249070865e-05, "loss": 2.1126, "step": 127 }, { "epoch": 0.04, "learning_rate": 1.999610609629147e-05, "loss": 1.5752, "step": 128 }, { "epoch": 0.04, "learning_rate": 1.9995823013308567e-05, "loss": 1.6581, "step": 129 }, { "epoch": 0.04, "learning_rate": 1.9995530000403378e-05, "loss": 2.675, "step": 130 }, { "epoch": 0.04, "learning_rate": 1.9995227057866985e-05, "loss": 1.3908, "step": 131 }, { "epoch": 0.04, "learning_rate": 1.999491418600033e-05, "loss": 1.7088, "step": 132 }, { "epoch": 0.04, "learning_rate": 1.9994591385114218e-05, "loss": 1.7882, "step": 133 }, { "epoch": 0.04, "learning_rate": 1.9994258655529336e-05, "loss": 1.5781, "step": 134 }, { "epoch": 0.04, "learning_rate": 1.9993915997576204e-05, "loss": 1.7004, "step": 135 }, { "epoch": 0.04, "learning_rate": 1.999356341159523e-05, "loss": 1.5753, "step": 136 }, { "epoch": 0.04, "learning_rate": 1.9993200897936673e-05, "loss": 1.7941, "step": 137 }, { "epoch": 0.04, "learning_rate": 1.9992828456960654e-05, "loss": 1.6074, "step": 138 }, { "epoch": 0.04, "learning_rate": 1.9992446089037164e-05, "loss": 1.3368, "step": 139 }, { "epoch": 0.04, "learning_rate": 1.999205379454605e-05, "loss": 1.6603, "step": 140 }, { "epoch": 0.04, "learning_rate": 1.9991651573877012e-05, "loss": 2.8197, "step": 141 }, { "epoch": 0.04, "learning_rate": 1.999123942742963e-05, "loss": 1.837, "step": 142 }, { "epoch": 0.04, "learning_rate": 1.9990817355613323e-05, "loss": 1.7987, "step": 143 }, { "epoch": 0.04, "learning_rate": 1.9990385358847385e-05, "loss": 1.4613, "step": 144 }, { "epoch": 0.04, "learning_rate": 1.998994343756097e-05, "loss": 2.5077, "step": 145 }, { "epoch": 0.04, "learning_rate": 1.9989491592193074e-05, "loss": 1.6775, "step": 146 }, { "epoch": 0.05, "learning_rate": 1.9989029823192575e-05, "loss": 1.6271, "step": 147 }, { "epoch": 0.05, "learning_rate": 1.9988558131018188e-05, "loss": 1.6006, "step": 148 }, { "epoch": 0.05, "learning_rate": 1.99880765161385e-05, "loss": 1.5418, "step": 149 }, { "epoch": 0.05, "learning_rate": 1.9987584979031954e-05, "loss": 1.4503, "step": 150 }, { "epoch": 0.05, "learning_rate": 1.998708352018684e-05, "loss": 1.7547, "step": 151 }, { "epoch": 0.05, "learning_rate": 1.9986572140101312e-05, "loss": 1.5231, "step": 152 }, { "epoch": 0.05, "learning_rate": 1.9986050839283383e-05, "loss": 1.536, "step": 153 }, { "epoch": 0.05, "learning_rate": 1.9985519618250912e-05, "loss": 1.578, "step": 154 }, { "epoch": 0.05, "learning_rate": 1.9984978477531625e-05, "loss": 1.4663, "step": 155 }, { "epoch": 0.05, "learning_rate": 1.9984427417663085e-05, "loss": 1.5846, "step": 156 }, { "epoch": 0.05, "learning_rate": 1.9983866439192726e-05, "loss": 1.4892, "step": 157 }, { "epoch": 0.05, "learning_rate": 1.998329554267783e-05, "loss": 1.4006, "step": 158 }, { "epoch": 0.05, "learning_rate": 1.998271472868552e-05, "loss": 1.5786, "step": 159 }, { "epoch": 0.05, "learning_rate": 1.998212399779279e-05, "loss": 1.6252, "step": 160 }, { "epoch": 0.05, "learning_rate": 1.998152335058647e-05, "loss": 1.5301, "step": 161 }, { "epoch": 0.05, "learning_rate": 1.9980912787663253e-05, "loss": 1.8721, "step": 162 }, { "epoch": 0.05, "learning_rate": 1.998029230962967e-05, "loss": 1.7937, "step": 163 }, { "epoch": 0.05, "learning_rate": 1.9979661917102116e-05, "loss": 1.685, "step": 164 }, { "epoch": 0.05, "learning_rate": 1.9979021610706822e-05, "loss": 2.0472, "step": 165 }, { "epoch": 0.05, "learning_rate": 1.9978371391079874e-05, "loss": 1.5465, "step": 166 }, { "epoch": 0.05, "learning_rate": 1.9978371391079874e-05, "loss": 1.3668, "step": 167 }, { "epoch": 0.05, "learning_rate": 1.9977711258867204e-05, "loss": 1.4535, "step": 168 }, { "epoch": 0.05, "learning_rate": 1.9977041214724594e-05, "loss": 1.6723, "step": 169 }, { "epoch": 0.05, "learning_rate": 1.9976361259317674e-05, "loss": 1.7369, "step": 170 }, { "epoch": 0.05, "learning_rate": 1.997567139332191e-05, "loss": 1.6297, "step": 171 }, { "epoch": 0.05, "learning_rate": 1.9974971617422622e-05, "loss": 1.5693, "step": 172 }, { "epoch": 0.05, "learning_rate": 1.997426193231497e-05, "loss": 1.7087, "step": 173 }, { "epoch": 0.05, "learning_rate": 1.9973542338703965e-05, "loss": 1.3286, "step": 174 }, { "epoch": 0.05, "learning_rate": 1.9972812837304454e-05, "loss": 1.7632, "step": 175 }, { "epoch": 0.05, "learning_rate": 1.9972073428841134e-05, "loss": 1.8314, "step": 176 }, { "epoch": 0.05, "learning_rate": 1.9971324114048533e-05, "loss": 1.4059, "step": 177 }, { "epoch": 0.05, "learning_rate": 1.997056489367102e-05, "loss": 1.3944, "step": 178 }, { "epoch": 0.06, "learning_rate": 1.9969795768462824e-05, "loss": 1.6226, "step": 179 }, { "epoch": 0.06, "learning_rate": 1.996901673918799e-05, "loss": 1.5514, "step": 180 }, { "epoch": 0.06, "learning_rate": 1.9968227806620413e-05, "loss": 1.7222, "step": 181 }, { "epoch": 0.06, "learning_rate": 1.9967428971543824e-05, "loss": 1.3574, "step": 182 }, { "epoch": 0.06, "learning_rate": 1.9966620234751794e-05, "loss": 1.396, "step": 183 }, { "epoch": 0.06, "learning_rate": 1.9965801597047724e-05, "loss": 1.5345, "step": 184 }, { "epoch": 0.06, "learning_rate": 1.996497305924486e-05, "loss": 1.5218, "step": 185 }, { "epoch": 0.06, "learning_rate": 1.9964134622166273e-05, "loss": 1.7672, "step": 186 }, { "epoch": 0.06, "learning_rate": 1.9963286286644875e-05, "loss": 1.8566, "step": 187 }, { "epoch": 0.06, "learning_rate": 1.9962428053523406e-05, "loss": 1.9358, "step": 188 }, { "epoch": 0.06, "learning_rate": 1.996155992365444e-05, "loss": 1.828, "step": 189 }, { "epoch": 0.06, "learning_rate": 1.9960681897900388e-05, "loss": 1.6402, "step": 190 }, { "epoch": 0.06, "learning_rate": 1.9959793977133483e-05, "loss": 1.5794, "step": 191 }, { "epoch": 0.06, "learning_rate": 1.9958896162235794e-05, "loss": 1.7885, "step": 192 }, { "epoch": 0.06, "learning_rate": 1.995798845409922e-05, "loss": 1.6629, "step": 193 }, { "epoch": 0.06, "learning_rate": 1.9957070853625476e-05, "loss": 1.6692, "step": 194 }, { "epoch": 0.06, "learning_rate": 1.9956143361726123e-05, "loss": 1.4805, "step": 195 }, { "epoch": 0.06, "learning_rate": 1.9955205979322528e-05, "loss": 1.6437, "step": 196 }, { "epoch": 0.06, "learning_rate": 1.9954258707345903e-05, "loss": 1.9134, "step": 197 }, { "epoch": 0.06, "learning_rate": 1.9953301546737267e-05, "loss": 1.4309, "step": 198 }, { "epoch": 0.06, "learning_rate": 1.9952334498447473e-05, "loss": 1.5045, "step": 199 }, { "epoch": 0.06, "learning_rate": 1.9951357563437198e-05, "loss": 1.5969, "step": 200 }, { "epoch": 0.06, "learning_rate": 1.995037074267693e-05, "loss": 1.2856, "step": 201 }, { "epoch": 0.06, "learning_rate": 1.9949374037146984e-05, "loss": 1.6973, "step": 202 }, { "epoch": 0.06, "learning_rate": 1.9948367447837503e-05, "loss": 1.8148, "step": 203 }, { "epoch": 0.06, "learning_rate": 1.9947350975748428e-05, "loss": 1.7634, "step": 204 }, { "epoch": 0.06, "learning_rate": 1.9946324621889538e-05, "loss": 1.6316, "step": 205 }, { "epoch": 0.06, "learning_rate": 1.994528838728042e-05, "loss": 1.5419, "step": 206 }, { "epoch": 0.06, "learning_rate": 1.9944242272950468e-05, "loss": 1.4158, "step": 207 }, { "epoch": 0.06, "learning_rate": 1.9943186279938908e-05, "loss": 1.7323, "step": 208 }, { "epoch": 0.06, "learning_rate": 1.9942120409294768e-05, "loss": 1.3954, "step": 209 }, { "epoch": 0.06, "learning_rate": 1.9941044662076898e-05, "loss": 1.7536, "step": 210 }, { "epoch": 0.06, "learning_rate": 1.9939959039353947e-05, "loss": 1.5485, "step": 211 }, { "epoch": 0.07, "learning_rate": 1.993886354220438e-05, "loss": 2.001, "step": 212 }, { "epoch": 0.07, "learning_rate": 1.9937758171716468e-05, "loss": 1.5152, "step": 213 }, { "epoch": 0.07, "learning_rate": 1.9936642928988304e-05, "loss": 2.101, "step": 214 }, { "epoch": 0.07, "learning_rate": 1.993551781512777e-05, "loss": 1.6773, "step": 215 }, { "epoch": 0.07, "learning_rate": 1.9934382831252565e-05, "loss": 2.0094, "step": 216 }, { "epoch": 0.07, "learning_rate": 1.9933237978490193e-05, "loss": 1.5861, "step": 217 }, { "epoch": 0.07, "learning_rate": 1.9932083257977954e-05, "loss": 1.6079, "step": 218 }, { "epoch": 0.07, "learning_rate": 1.993091867086296e-05, "loss": 1.3814, "step": 219 }, { "epoch": 0.07, "learning_rate": 1.9929744218302116e-05, "loss": 1.8781, "step": 220 }, { "epoch": 0.07, "learning_rate": 1.992855990146213e-05, "loss": 1.5523, "step": 221 }, { "epoch": 0.07, "learning_rate": 1.992736572151952e-05, "loss": 1.4769, "step": 222 }, { "epoch": 0.07, "learning_rate": 1.9926161679660588e-05, "loss": 1.6965, "step": 223 }, { "epoch": 0.07, "learning_rate": 1.9924947777081434e-05, "loss": 1.3362, "step": 224 }, { "epoch": 0.07, "learning_rate": 1.9923724014987963e-05, "loss": 1.5268, "step": 225 }, { "epoch": 0.07, "learning_rate": 1.9922490394595862e-05, "loss": 1.6003, "step": 226 }, { "epoch": 0.07, "learning_rate": 1.9921246917130627e-05, "loss": 1.3486, "step": 227 }, { "epoch": 0.07, "learning_rate": 1.991999358382753e-05, "loss": 1.4856, "step": 228 }, { "epoch": 0.07, "learning_rate": 1.9918730395931648e-05, "loss": 1.3157, "step": 229 }, { "epoch": 0.07, "learning_rate": 1.9917457354697838e-05, "loss": 1.5973, "step": 230 }, { "epoch": 0.07, "learning_rate": 1.991617446139075e-05, "loss": 1.4162, "step": 231 }, { "epoch": 0.07, "learning_rate": 1.9914881717284812e-05, "loss": 1.5909, "step": 232 }, { "epoch": 0.07, "learning_rate": 1.9913579123664253e-05, "loss": 1.4641, "step": 233 }, { "epoch": 0.07, "learning_rate": 1.9912266681823077e-05, "loss": 1.4476, "step": 234 }, { "epoch": 0.07, "learning_rate": 1.9910944393065073e-05, "loss": 1.4394, "step": 235 }, { "epoch": 0.07, "learning_rate": 1.990961225870381e-05, "loss": 1.7175, "step": 236 }, { "epoch": 0.07, "learning_rate": 1.9908270280062643e-05, "loss": 1.7933, "step": 237 }, { "epoch": 0.07, "learning_rate": 1.99069184584747e-05, "loss": 1.3454, "step": 238 }, { "epoch": 0.07, "learning_rate": 1.9905556795282894e-05, "loss": 1.2857, "step": 239 }, { "epoch": 0.07, "learning_rate": 1.990418529183991e-05, "loss": 1.7622, "step": 240 }, { "epoch": 0.07, "learning_rate": 1.9902803949508206e-05, "loss": 1.6376, "step": 241 }, { "epoch": 0.07, "learning_rate": 1.9901412769660023e-05, "loss": 1.5477, "step": 242 }, { "epoch": 0.07, "learning_rate": 1.9900011753677363e-05, "loss": 1.8393, "step": 243 }, { "epoch": 0.08, "learning_rate": 1.989860090295201e-05, "loss": 1.592, "step": 244 }, { "epoch": 0.08, "learning_rate": 1.989718021888551e-05, "loss": 1.3824, "step": 245 }, { "epoch": 0.08, "learning_rate": 1.989574970288918e-05, "loss": 1.7632, "step": 246 }, { "epoch": 0.08, "learning_rate": 1.989430935638411e-05, "loss": 1.5889, "step": 247 }, { "epoch": 0.08, "learning_rate": 1.9892859180801146e-05, "loss": 1.4855, "step": 248 }, { "epoch": 0.08, "learning_rate": 1.989139917758091e-05, "loss": 1.9986, "step": 249 }, { "epoch": 0.08, "learning_rate": 1.9889929348173767e-05, "loss": 1.7831, "step": 250 }, { "epoch": 0.08, "learning_rate": 1.9888449694039866e-05, "loss": 1.7295, "step": 251 }, { "epoch": 0.08, "learning_rate": 1.9886960216649102e-05, "loss": 2.1687, "step": 252 }, { "epoch": 0.08, "learning_rate": 1.9885460917481137e-05, "loss": 1.6867, "step": 253 }, { "epoch": 0.08, "learning_rate": 1.9883951798025377e-05, "loss": 1.6013, "step": 254 }, { "epoch": 0.08, "learning_rate": 1.9882432859781002e-05, "loss": 1.5046, "step": 255 }, { "epoch": 0.08, "learning_rate": 1.9880904104256927e-05, "loss": 1.5368, "step": 256 }, { "epoch": 0.08, "learning_rate": 1.987936553297184e-05, "loss": 1.3318, "step": 257 }, { "epoch": 0.08, "learning_rate": 1.9877817147454156e-05, "loss": 1.6516, "step": 258 }, { "epoch": 0.08, "learning_rate": 1.987625894924206e-05, "loss": 1.4754, "step": 259 }, { "epoch": 0.08, "learning_rate": 1.9874690939883473e-05, "loss": 1.8631, "step": 260 }, { "epoch": 0.08, "learning_rate": 1.9873113120936074e-05, "loss": 1.7939, "step": 261 }, { "epoch": 0.08, "learning_rate": 1.9871525493967273e-05, "loss": 1.715, "step": 262 }, { "epoch": 0.08, "learning_rate": 1.9869928060554234e-05, "loss": 1.7277, "step": 263 }, { "epoch": 0.08, "learning_rate": 1.9868320822283853e-05, "loss": 1.1737, "step": 264 }, { "epoch": 0.08, "learning_rate": 1.9866703780752777e-05, "loss": 1.6055, "step": 265 }, { "epoch": 0.08, "learning_rate": 1.9865076937567384e-05, "loss": 1.5523, "step": 266 }, { "epoch": 0.08, "learning_rate": 1.9863440294343793e-05, "loss": 1.4681, "step": 267 }, { "epoch": 0.08, "learning_rate": 1.986179385270786e-05, "loss": 1.8599, "step": 268 }, { "epoch": 0.08, "learning_rate": 1.986013761429517e-05, "loss": 1.7851, "step": 269 }, { "epoch": 0.08, "learning_rate": 1.985847158075104e-05, "loss": 1.663, "step": 270 }, { "epoch": 0.08, "learning_rate": 1.985679575373052e-05, "loss": 1.8522, "step": 271 }, { "epoch": 0.08, "learning_rate": 1.985511013489839e-05, "loss": 1.7483, "step": 272 }, { "epoch": 0.08, "learning_rate": 1.9853414725929157e-05, "loss": 1.6165, "step": 273 }, { "epoch": 0.08, "learning_rate": 1.9851709528507047e-05, "loss": 2.2397, "step": 274 }, { "epoch": 0.08, "learning_rate": 1.984999454432602e-05, "loss": 1.605, "step": 275 }, { "epoch": 0.08, "learning_rate": 1.9848269775089756e-05, "loss": 1.7188, "step": 276 }, { "epoch": 0.09, "learning_rate": 1.9846535222511648e-05, "loss": 1.7159, "step": 277 }, { "epoch": 0.09, "learning_rate": 1.9844790888314812e-05, "loss": 1.4937, "step": 278 }, { "epoch": 0.09, "learning_rate": 1.984303677423209e-05, "loss": 1.4489, "step": 279 }, { "epoch": 0.09, "learning_rate": 1.9841272882006026e-05, "loss": 1.4188, "step": 280 }, { "epoch": 0.09, "learning_rate": 1.983949921338888e-05, "loss": 1.6416, "step": 281 }, { "epoch": 0.09, "learning_rate": 1.9837715770142633e-05, "loss": 2.0655, "step": 282 }, { "epoch": 0.09, "learning_rate": 1.983592255403897e-05, "loss": 1.6055, "step": 283 }, { "epoch": 0.09, "learning_rate": 1.983411956685928e-05, "loss": 1.7311, "step": 284 }, { "epoch": 0.09, "learning_rate": 1.9832306810394665e-05, "loss": 1.8048, "step": 285 }, { "epoch": 0.09, "learning_rate": 1.9830484286445935e-05, "loss": 2.0122, "step": 286 }, { "epoch": 0.09, "learning_rate": 1.9828651996823593e-05, "loss": 1.6216, "step": 287 }, { "epoch": 0.09, "learning_rate": 1.982680994334785e-05, "loss": 1.6082, "step": 288 }, { "epoch": 0.09, "learning_rate": 1.9824958127848618e-05, "loss": 1.3639, "step": 289 }, { "epoch": 0.09, "learning_rate": 1.98230965521655e-05, "loss": 1.4936, "step": 290 }, { "epoch": 0.09, "learning_rate": 1.98212252181478e-05, "loss": 1.4603, "step": 291 }, { "epoch": 0.09, "learning_rate": 1.9819344127654515e-05, "loss": 1.6137, "step": 292 }, { "epoch": 0.09, "learning_rate": 1.9817453282554334e-05, "loss": 1.2439, "step": 293 }, { "epoch": 0.09, "learning_rate": 1.9815552684725635e-05, "loss": 1.8232, "step": 294 }, { "epoch": 0.09, "learning_rate": 1.9813642336056488e-05, "loss": 1.6075, "step": 295 }, { "epoch": 0.09, "learning_rate": 1.9811722238444642e-05, "loss": 1.7471, "step": 296 }, { "epoch": 0.09, "learning_rate": 1.9809792393797543e-05, "loss": 1.5875, "step": 297 }, { "epoch": 0.09, "learning_rate": 1.9807852804032306e-05, "loss": 1.6445, "step": 298 }, { "epoch": 0.09, "learning_rate": 1.9805903471075735e-05, "loss": 1.4694, "step": 299 }, { "epoch": 0.09, "learning_rate": 1.9803944396864313e-05, "loss": 2.347, "step": 300 }, { "epoch": 0.09, "learning_rate": 1.98019755833442e-05, "loss": 1.6418, "step": 301 }, { "epoch": 0.09, "learning_rate": 1.9799997032471223e-05, "loss": 1.6859, "step": 302 }, { "epoch": 0.09, "learning_rate": 1.9798008746210894e-05, "loss": 1.5495, "step": 303 }, { "epoch": 0.09, "learning_rate": 1.9796010726538393e-05, "loss": 1.5832, "step": 304 }, { "epoch": 0.09, "learning_rate": 1.979400297543856e-05, "loss": 1.6645, "step": 305 }, { "epoch": 0.09, "learning_rate": 1.9791985494905913e-05, "loss": 1.631, "step": 306 }, { "epoch": 0.09, "learning_rate": 1.9789958286944632e-05, "loss": 1.4156, "step": 307 }, { "epoch": 0.09, "learning_rate": 1.978792135356856e-05, "loss": 1.9984, "step": 308 }, { "epoch": 0.1, "learning_rate": 1.97858746968012e-05, "loss": 2.0683, "step": 309 }, { "epoch": 0.1, "learning_rate": 1.97858746968012e-05, "loss": 1.6231, "step": 310 }, { "epoch": 0.1, "learning_rate": 1.978381831867572e-05, "loss": 1.734, "step": 311 }, { "epoch": 0.1, "learning_rate": 1.9781752221234932e-05, "loss": 1.3379, "step": 312 }, { "epoch": 0.1, "learning_rate": 1.977967640653132e-05, "loss": 1.6322, "step": 313 }, { "epoch": 0.1, "learning_rate": 1.9777590876627012e-05, "loss": 1.4378, "step": 314 }, { "epoch": 0.1, "learning_rate": 1.977549563359378e-05, "loss": 1.6712, "step": 315 }, { "epoch": 0.1, "learning_rate": 1.9773390679513068e-05, "loss": 1.6888, "step": 316 }, { "epoch": 0.1, "learning_rate": 1.977127601647594e-05, "loss": 1.7305, "step": 317 }, { "epoch": 0.1, "learning_rate": 1.9769151646583122e-05, "loss": 1.5078, "step": 318 }, { "epoch": 0.1, "learning_rate": 1.9767017571944976e-05, "loss": 1.416, "step": 319 }, { "epoch": 0.1, "learning_rate": 1.9764873794681514e-05, "loss": 1.5982, "step": 320 }, { "epoch": 0.1, "learning_rate": 1.9762720316922372e-05, "loss": 1.6363, "step": 321 }, { "epoch": 0.1, "learning_rate": 1.9760557140806838e-05, "loss": 2.1026, "step": 322 }, { "epoch": 0.1, "learning_rate": 1.9758384268483816e-05, "loss": 1.3741, "step": 323 }, { "epoch": 0.1, "learning_rate": 1.9756201702111863e-05, "loss": 1.4933, "step": 324 }, { "epoch": 0.1, "learning_rate": 1.975400944385915e-05, "loss": 1.3125, "step": 325 }, { "epoch": 0.1, "learning_rate": 1.9751807495903484e-05, "loss": 1.6205, "step": 326 }, { "epoch": 0.1, "learning_rate": 1.9749595860432298e-05, "loss": 1.5026, "step": 327 }, { "epoch": 0.1, "learning_rate": 1.9747374539642643e-05, "loss": 1.5848, "step": 328 }, { "epoch": 0.1, "learning_rate": 1.97451435357412e-05, "loss": 1.852, "step": 329 }, { "epoch": 0.1, "learning_rate": 1.9742902850944257e-05, "loss": 1.5626, "step": 330 }, { "epoch": 0.1, "learning_rate": 1.974065248747773e-05, "loss": 1.6991, "step": 331 }, { "epoch": 0.1, "learning_rate": 1.9738392447577148e-05, "loss": 1.4958, "step": 332 }, { "epoch": 0.1, "learning_rate": 1.9736122733487648e-05, "loss": 1.5429, "step": 333 }, { "epoch": 0.1, "learning_rate": 1.9733843347463982e-05, "loss": 1.5443, "step": 334 }, { "epoch": 0.1, "learning_rate": 1.973155429177051e-05, "loss": 1.5045, "step": 335 }, { "epoch": 0.1, "learning_rate": 1.972925556868119e-05, "loss": 1.419, "step": 336 }, { "epoch": 0.1, "learning_rate": 1.9726947180479598e-05, "loss": 1.8166, "step": 337 }, { "epoch": 0.1, "learning_rate": 1.97246291294589e-05, "loss": 1.6045, "step": 338 }, { "epoch": 0.1, "learning_rate": 1.9722301417921864e-05, "loss": 1.526, "step": 339 }, { "epoch": 0.1, "learning_rate": 1.9719964048180853e-05, "loss": 1.4242, "step": 340 }, { "epoch": 0.1, "learning_rate": 1.9717617022557828e-05, "loss": 1.6578, "step": 341 }, { "epoch": 0.11, "learning_rate": 1.971526034338435e-05, "loss": 1.508, "step": 342 }, { "epoch": 0.11, "learning_rate": 1.971289401300155e-05, "loss": 1.4866, "step": 343 }, { "epoch": 0.11, "learning_rate": 1.9710518033760158e-05, "loss": 1.5905, "step": 344 }, { "epoch": 0.11, "learning_rate": 1.9708132408020493e-05, "loss": 1.6059, "step": 345 }, { "epoch": 0.11, "learning_rate": 1.970573713815245e-05, "loss": 1.5395, "step": 346 }, { "epoch": 0.11, "learning_rate": 1.9703332226535515e-05, "loss": 1.7586, "step": 347 }, { "epoch": 0.11, "learning_rate": 1.9700917675558733e-05, "loss": 1.3932, "step": 348 }, { "epoch": 0.11, "learning_rate": 1.9698493487620742e-05, "loss": 1.5638, "step": 349 }, { "epoch": 0.11, "learning_rate": 1.969605966512975e-05, "loss": 1.5222, "step": 350 }, { "epoch": 0.11, "learning_rate": 1.9693616210503528e-05, "loss": 1.322, "step": 351 }, { "epoch": 0.11, "learning_rate": 1.9691163126169428e-05, "loss": 1.636, "step": 352 }, { "epoch": 0.11, "learning_rate": 1.9688700414564355e-05, "loss": 1.6224, "step": 353 }, { "epoch": 0.11, "learning_rate": 1.968622807813479e-05, "loss": 1.4956, "step": 354 }, { "epoch": 0.11, "learning_rate": 1.9683746119336763e-05, "loss": 1.5849, "step": 355 }, { "epoch": 0.11, "learning_rate": 1.9681254540635877e-05, "loss": 1.4047, "step": 356 }, { "epoch": 0.11, "learning_rate": 1.967875334450728e-05, "loss": 1.4244, "step": 357 }, { "epoch": 0.11, "learning_rate": 1.967624253343568e-05, "loss": 1.8103, "step": 358 }, { "epoch": 0.11, "learning_rate": 1.9673722109915333e-05, "loss": 1.4233, "step": 359 }, { "epoch": 0.11, "learning_rate": 1.9671192076450047e-05, "loss": 1.5318, "step": 360 }, { "epoch": 0.11, "learning_rate": 1.9668652435553173e-05, "loss": 1.2261, "step": 361 }, { "epoch": 0.11, "learning_rate": 1.966610318974761e-05, "loss": 1.4746, "step": 362 }, { "epoch": 0.11, "learning_rate": 1.96635443415658e-05, "loss": 1.8807, "step": 363 }, { "epoch": 0.11, "learning_rate": 1.9660975893549722e-05, "loss": 1.5223, "step": 364 }, { "epoch": 0.11, "learning_rate": 1.9658397848250882e-05, "loss": 1.582, "step": 365 }, { "epoch": 0.11, "learning_rate": 1.9655810208230334e-05, "loss": 1.8018, "step": 366 }, { "epoch": 0.11, "learning_rate": 1.9653212976058664e-05, "loss": 1.5699, "step": 367 }, { "epoch": 0.11, "learning_rate": 1.9650606154315967e-05, "loss": 1.7314, "step": 368 }, { "epoch": 0.11, "learning_rate": 1.964798974559189e-05, "loss": 1.56, "step": 369 }, { "epoch": 0.11, "learning_rate": 1.9645363752485594e-05, "loss": 1.4964, "step": 370 }, { "epoch": 0.11, "learning_rate": 1.9642728177605747e-05, "loss": 1.7369, "step": 371 }, { "epoch": 0.11, "learning_rate": 1.964008302357056e-05, "loss": 1.2725, "step": 372 }, { "epoch": 0.11, "learning_rate": 1.9637428293007744e-05, "loss": 1.449, "step": 373 }, { "epoch": 0.12, "learning_rate": 1.963476398855452e-05, "loss": 1.6289, "step": 374 }, { "epoch": 0.12, "learning_rate": 1.963209011285764e-05, "loss": 0.988, "step": 375 }, { "epoch": 0.12, "learning_rate": 1.9629406668573346e-05, "loss": 1.6609, "step": 376 }, { "epoch": 0.12, "learning_rate": 1.9626713658367385e-05, "loss": 1.7869, "step": 377 }, { "epoch": 0.12, "learning_rate": 1.962401108491502e-05, "loss": 1.4871, "step": 378 }, { "epoch": 0.12, "learning_rate": 1.9621298950901005e-05, "loss": 1.3658, "step": 379 }, { "epoch": 0.12, "learning_rate": 1.9618577259019595e-05, "loss": 1.6384, "step": 380 }, { "epoch": 0.12, "learning_rate": 1.961584601197453e-05, "loss": 1.6409, "step": 381 }, { "epoch": 0.12, "learning_rate": 1.961310521247906e-05, "loss": 1.9667, "step": 382 }, { "epoch": 0.12, "learning_rate": 1.9610354863255914e-05, "loss": 1.3878, "step": 383 }, { "epoch": 0.12, "learning_rate": 1.96075949670373e-05, "loss": 1.5849, "step": 384 }, { "epoch": 0.12, "learning_rate": 1.960482552656493e-05, "loss": 1.4441, "step": 385 }, { "epoch": 0.12, "learning_rate": 1.960204654458998e-05, "loss": 1.8731, "step": 386 }, { "epoch": 0.12, "learning_rate": 1.9599258023873108e-05, "loss": 1.6578, "step": 387 }, { "epoch": 0.12, "learning_rate": 1.9596459967184458e-05, "loss": 1.4602, "step": 388 }, { "epoch": 0.12, "learning_rate": 1.959365237730363e-05, "loss": 1.8843, "step": 389 }, { "epoch": 0.12, "learning_rate": 1.9590835257019715e-05, "loss": 1.685, "step": 390 }, { "epoch": 0.12, "learning_rate": 1.9588008609131253e-05, "loss": 1.6099, "step": 391 }, { "epoch": 0.12, "learning_rate": 1.958517243644626e-05, "loss": 1.4345, "step": 392 }, { "epoch": 0.12, "learning_rate": 1.9582326741782204e-05, "loss": 1.5144, "step": 393 }, { "epoch": 0.12, "learning_rate": 1.9579471527966026e-05, "loss": 1.5208, "step": 394 }, { "epoch": 0.12, "learning_rate": 1.9576606797834114e-05, "loss": 1.493, "step": 395 }, { "epoch": 0.12, "learning_rate": 1.957373255423231e-05, "loss": 1.3919, "step": 396 }, { "epoch": 0.12, "learning_rate": 1.9570848800015904e-05, "loss": 1.518, "step": 397 }, { "epoch": 0.12, "learning_rate": 1.9567955538049643e-05, "loss": 1.4975, "step": 398 }, { "epoch": 0.12, "learning_rate": 1.956505277120772e-05, "loss": 1.6597, "step": 399 }, { "epoch": 0.12, "learning_rate": 1.9562140502373753e-05, "loss": 1.7864, "step": 400 }, { "epoch": 0.12, "learning_rate": 1.9559218734440818e-05, "loss": 1.6085, "step": 401 }, { "epoch": 0.12, "learning_rate": 1.9556287470311418e-05, "loss": 1.5447, "step": 402 }, { "epoch": 0.12, "learning_rate": 1.955334671289749e-05, "loss": 1.4193, "step": 403 }, { "epoch": 0.12, "learning_rate": 1.955039646512041e-05, "loss": 1.4759, "step": 404 }, { "epoch": 0.12, "learning_rate": 1.9547436729910968e-05, "loss": 1.4027, "step": 405 }, { "epoch": 0.12, "learning_rate": 1.954446751020939e-05, "loss": 1.5694, "step": 406 }, { "epoch": 0.13, "learning_rate": 1.9541488808965318e-05, "loss": 1.3939, "step": 407 }, { "epoch": 0.13, "learning_rate": 1.9538500629137814e-05, "loss": 1.3507, "step": 408 }, { "epoch": 0.13, "learning_rate": 1.9535502973695362e-05, "loss": 1.916, "step": 409 }, { "epoch": 0.13, "learning_rate": 1.9532495845615854e-05, "loss": 2.342, "step": 410 }, { "epoch": 0.13, "learning_rate": 1.952947924788659e-05, "loss": 1.3617, "step": 411 }, { "epoch": 0.13, "learning_rate": 1.952645318350428e-05, "loss": 1.4452, "step": 412 }, { "epoch": 0.13, "learning_rate": 1.952341765547504e-05, "loss": 1.5346, "step": 413 }, { "epoch": 0.13, "learning_rate": 1.952037266681438e-05, "loss": 1.6285, "step": 414 }, { "epoch": 0.13, "learning_rate": 1.951731822054722e-05, "loss": 1.8198, "step": 415 }, { "epoch": 0.13, "learning_rate": 1.9514254319707868e-05, "loss": 1.7929, "step": 416 }, { "epoch": 0.13, "learning_rate": 1.951118096734002e-05, "loss": 1.199, "step": 417 }, { "epoch": 0.13, "learning_rate": 1.950809816649677e-05, "loss": 1.7565, "step": 418 }, { "epoch": 0.13, "learning_rate": 1.9505005920240596e-05, "loss": 1.3858, "step": 419 }, { "epoch": 0.13, "learning_rate": 1.9501904231643353e-05, "loss": 1.974, "step": 420 }, { "epoch": 0.13, "learning_rate": 1.9498793103786288e-05, "loss": 1.4981, "step": 421 }, { "epoch": 0.13, "learning_rate": 1.949567253976001e-05, "loss": 1.9566, "step": 422 }, { "epoch": 0.13, "learning_rate": 1.949254254266451e-05, "loss": 1.9694, "step": 423 }, { "epoch": 0.13, "learning_rate": 1.9489403115609157e-05, "loss": 1.6272, "step": 424 }, { "epoch": 0.13, "learning_rate": 1.9486254261712677e-05, "loss": 1.465, "step": 425 }, { "epoch": 0.13, "learning_rate": 1.948309598410316e-05, "loss": 1.6619, "step": 426 }, { "epoch": 0.13, "learning_rate": 1.9479928285918065e-05, "loss": 1.3405, "step": 427 }, { "epoch": 0.13, "learning_rate": 1.9476751170304206e-05, "loss": 1.6428, "step": 428 }, { "epoch": 0.13, "learning_rate": 1.947356464041775e-05, "loss": 1.7289, "step": 429 }, { "epoch": 0.13, "learning_rate": 1.947036869942422e-05, "loss": 1.6445, "step": 430 }, { "epoch": 0.13, "learning_rate": 1.946716335049849e-05, "loss": 1.503, "step": 431 }, { "epoch": 0.13, "learning_rate": 1.9463948596824774e-05, "loss": 1.845, "step": 432 }, { "epoch": 0.13, "learning_rate": 1.946072444159663e-05, "loss": 1.7573, "step": 433 }, { "epoch": 0.13, "learning_rate": 1.9457490888016962e-05, "loss": 1.6144, "step": 434 }, { "epoch": 0.13, "learning_rate": 1.9454247939297996e-05, "loss": 1.7371, "step": 435 }, { "epoch": 0.13, "learning_rate": 1.945099559866131e-05, "loss": 1.5602, "step": 436 }, { "epoch": 0.13, "learning_rate": 1.94477338693378e-05, "loss": 1.5315, "step": 437 }, { "epoch": 0.13, "learning_rate": 1.9444462754567682e-05, "loss": 1.5651, "step": 438 }, { "epoch": 0.14, "learning_rate": 1.944118225760052e-05, "loss": 1.2868, "step": 439 }, { "epoch": 0.14, "learning_rate": 1.9437892381695174e-05, "loss": 1.589, "step": 440 }, { "epoch": 0.14, "learning_rate": 1.9434593130119835e-05, "loss": 1.5082, "step": 441 }, { "epoch": 0.14, "learning_rate": 1.9431284506151996e-05, "loss": 1.7603, "step": 442 }, { "epoch": 0.14, "learning_rate": 1.9427966513078476e-05, "loss": 1.5296, "step": 443 }, { "epoch": 0.14, "learning_rate": 1.9424639154195386e-05, "loss": 1.8436, "step": 444 }, { "epoch": 0.14, "learning_rate": 1.9421302432808153e-05, "loss": 1.5024, "step": 445 }, { "epoch": 0.14, "learning_rate": 1.94179563522315e-05, "loss": 1.4501, "step": 446 }, { "epoch": 0.14, "learning_rate": 1.9414600915789446e-05, "loss": 1.4316, "step": 447 }, { "epoch": 0.14, "learning_rate": 1.9411236126815306e-05, "loss": 1.4682, "step": 448 }, { "epoch": 0.14, "learning_rate": 1.9407861988651683e-05, "loss": 1.3392, "step": 449 }, { "epoch": 0.14, "learning_rate": 1.9404478504650473e-05, "loss": 1.8718, "step": 450 }, { "epoch": 0.14, "learning_rate": 1.9401085678172852e-05, "loss": 1.7281, "step": 451 }, { "epoch": 0.14, "learning_rate": 1.9397683512589282e-05, "loss": 1.5543, "step": 452 }, { "epoch": 0.14, "learning_rate": 1.9394272011279492e-05, "loss": 1.4684, "step": 453 }, { "epoch": 0.14, "learning_rate": 1.9390851177632496e-05, "loss": 1.5275, "step": 454 }, { "epoch": 0.14, "learning_rate": 1.9387421015046576e-05, "loss": 1.8237, "step": 455 }, { "epoch": 0.14, "learning_rate": 1.9383981526929277e-05, "loss": 1.5748, "step": 456 }, { "epoch": 0.14, "learning_rate": 1.9380532716697412e-05, "loss": 1.5865, "step": 457 }, { "epoch": 0.14, "learning_rate": 1.937707458777705e-05, "loss": 1.5173, "step": 458 }, { "epoch": 0.14, "learning_rate": 1.9373607143603533e-05, "loss": 1.5135, "step": 459 }, { "epoch": 0.14, "learning_rate": 1.937013038762143e-05, "loss": 1.5442, "step": 460 }, { "epoch": 0.14, "learning_rate": 1.9366644323284584e-05, "loss": 2.0023, "step": 461 }, { "epoch": 0.14, "learning_rate": 1.9363148954056077e-05, "loss": 1.4716, "step": 462 }, { "epoch": 0.14, "learning_rate": 1.9359644283408233e-05, "loss": 1.9954, "step": 463 }, { "epoch": 0.14, "learning_rate": 1.935613031482261e-05, "loss": 2.2567, "step": 464 }, { "epoch": 0.14, "learning_rate": 1.935260705179002e-05, "loss": 1.7179, "step": 465 }, { "epoch": 0.14, "learning_rate": 1.934907449781049e-05, "loss": 1.6664, "step": 466 }, { "epoch": 0.14, "learning_rate": 1.934553265639329e-05, "loss": 1.4733, "step": 467 }, { "epoch": 0.14, "learning_rate": 1.9341981531056905e-05, "loss": 1.563, "step": 468 }, { "epoch": 0.14, "learning_rate": 1.9338421125329053e-05, "loss": 1.4747, "step": 469 }, { "epoch": 0.14, "learning_rate": 1.9334851442746665e-05, "loss": 1.6416, "step": 470 }, { "epoch": 0.14, "learning_rate": 1.933127248685589e-05, "loss": 1.7337, "step": 471 }, { "epoch": 0.15, "learning_rate": 1.9327684261212084e-05, "loss": 1.5642, "step": 472 }, { "epoch": 0.15, "learning_rate": 1.9324086769379825e-05, "loss": 1.528, "step": 473 }, { "epoch": 0.15, "learning_rate": 1.932048001493288e-05, "loss": 1.876, "step": 474 }, { "epoch": 0.15, "learning_rate": 1.9316864001454226e-05, "loss": 1.4909, "step": 475 }, { "epoch": 0.15, "learning_rate": 1.931323873253604e-05, "loss": 1.6337, "step": 476 }, { "epoch": 0.15, "learning_rate": 1.9309604211779684e-05, "loss": 2.5694, "step": 477 }, { "epoch": 0.15, "learning_rate": 1.930596044279572e-05, "loss": 1.6391, "step": 478 }, { "epoch": 0.15, "learning_rate": 1.9302307429203902e-05, "loss": 1.4738, "step": 479 }, { "epoch": 0.15, "learning_rate": 1.9298645174633147e-05, "loss": 1.6618, "step": 480 }, { "epoch": 0.15, "learning_rate": 1.9294973682721573e-05, "loss": 1.7738, "step": 481 }, { "epoch": 0.15, "learning_rate": 1.929129295711646e-05, "loss": 1.6075, "step": 482 }, { "epoch": 0.15, "learning_rate": 1.9287603001474274e-05, "loss": 1.3121, "step": 483 }, { "epoch": 0.15, "learning_rate": 1.9283903819460644e-05, "loss": 1.4098, "step": 484 }, { "epoch": 0.15, "learning_rate": 1.928019541475036e-05, "loss": 1.8231, "step": 485 }, { "epoch": 0.15, "learning_rate": 1.9276477791027374e-05, "loss": 1.6571, "step": 486 }, { "epoch": 0.15, "learning_rate": 1.9272750951984808e-05, "loss": 1.5591, "step": 487 }, { "epoch": 0.15, "learning_rate": 1.9269014901324926e-05, "loss": 1.3717, "step": 488 }, { "epoch": 0.15, "learning_rate": 1.926526964275915e-05, "loss": 1.4766, "step": 489 }, { "epoch": 0.15, "learning_rate": 1.9261515180008047e-05, "loss": 1.4282, "step": 490 }, { "epoch": 0.15, "learning_rate": 1.9257751516801323e-05, "loss": 1.8362, "step": 491 }, { "epoch": 0.15, "learning_rate": 1.925397865687783e-05, "loss": 1.4829, "step": 492 }, { "epoch": 0.15, "learning_rate": 1.9250196603985557e-05, "loss": 1.4105, "step": 493 }, { "epoch": 0.15, "learning_rate": 1.9246405361881623e-05, "loss": 1.6449, "step": 494 }, { "epoch": 0.15, "learning_rate": 1.9242604934332268e-05, "loss": 2.0773, "step": 495 }, { "epoch": 0.15, "learning_rate": 1.9238795325112867e-05, "loss": 1.6207, "step": 496 }, { "epoch": 0.15, "learning_rate": 1.923497653800792e-05, "loss": 1.4271, "step": 497 }, { "epoch": 0.15, "learning_rate": 1.923114857681103e-05, "loss": 1.6192, "step": 498 }, { "epoch": 0.15, "learning_rate": 1.9227311445324924e-05, "loss": 1.2114, "step": 499 }, { "epoch": 0.15, "learning_rate": 1.9223465147361433e-05, "loss": 1.3277, "step": 500 }, { "epoch": 0.15, "learning_rate": 1.92196096867415e-05, "loss": 1.5036, "step": 501 }, { "epoch": 0.15, "learning_rate": 1.9215745067295168e-05, "loss": 1.5056, "step": 502 }, { "epoch": 0.15, "learning_rate": 1.9211871292861576e-05, "loss": 1.4721, "step": 503 }, { "epoch": 0.16, "learning_rate": 1.9207988367288956e-05, "loss": 1.8662, "step": 504 }, { "epoch": 0.16, "learning_rate": 1.920409629443464e-05, "loss": 1.231, "step": 505 }, { "epoch": 0.16, "learning_rate": 1.920019507816504e-05, "loss": 1.8711, "step": 506 }, { "epoch": 0.16, "learning_rate": 1.9196284722355648e-05, "loss": 1.7091, "step": 507 }, { "epoch": 0.16, "learning_rate": 1.9192365230891044e-05, "loss": 1.5687, "step": 508 }, { "epoch": 0.16, "learning_rate": 1.9188436607664878e-05, "loss": 1.4181, "step": 509 }, { "epoch": 0.16, "learning_rate": 1.918449885657987e-05, "loss": 1.4208, "step": 510 }, { "epoch": 0.16, "learning_rate": 1.9180551981547813e-05, "loss": 1.7203, "step": 511 }, { "epoch": 0.16, "learning_rate": 1.9176595986489558e-05, "loss": 1.6528, "step": 512 }, { "epoch": 0.16, "learning_rate": 1.9172630875335024e-05, "loss": 1.6635, "step": 513 }, { "epoch": 0.16, "learning_rate": 1.9168656652023176e-05, "loss": 1.5474, "step": 514 }, { "epoch": 0.16, "learning_rate": 1.916467332050204e-05, "loss": 1.4327, "step": 515 }, { "epoch": 0.16, "learning_rate": 1.916068088472868e-05, "loss": 1.2963, "step": 516 }, { "epoch": 0.16, "learning_rate": 1.9156679348669215e-05, "loss": 1.694, "step": 517 }, { "epoch": 0.16, "learning_rate": 1.91526687162988e-05, "loss": 1.4347, "step": 518 }, { "epoch": 0.16, "learning_rate": 1.9148648991601623e-05, "loss": 1.8322, "step": 519 }, { "epoch": 0.16, "learning_rate": 1.914462017857091e-05, "loss": 1.995, "step": 520 }, { "epoch": 0.16, "learning_rate": 1.914058228120891e-05, "loss": 1.3905, "step": 521 }, { "epoch": 0.16, "learning_rate": 1.9136535303526906e-05, "loss": 1.738, "step": 522 }, { "epoch": 0.16, "learning_rate": 1.9132479249545186e-05, "loss": 1.7038, "step": 523 }, { "epoch": 0.16, "learning_rate": 1.9128414123293065e-05, "loss": 1.387, "step": 524 }, { "epoch": 0.16, "learning_rate": 1.912433992880887e-05, "loss": 1.5046, "step": 525 }, { "epoch": 0.16, "learning_rate": 1.9120256670139942e-05, "loss": 1.6666, "step": 526 }, { "epoch": 0.16, "learning_rate": 1.911616435134261e-05, "loss": 1.5341, "step": 527 }, { "epoch": 0.16, "learning_rate": 1.9112062976482213e-05, "loss": 1.5286, "step": 528 }, { "epoch": 0.16, "learning_rate": 1.910795254963309e-05, "loss": 1.47, "step": 529 }, { "epoch": 0.16, "learning_rate": 1.9103833074878565e-05, "loss": 2.0051, "step": 530 }, { "epoch": 0.16, "learning_rate": 1.9099704556310962e-05, "loss": 1.4829, "step": 531 }, { "epoch": 0.16, "learning_rate": 1.909556699803157e-05, "loss": 1.4724, "step": 532 }, { "epoch": 0.16, "learning_rate": 1.9091420404150674e-05, "loss": 2.0598, "step": 533 }, { "epoch": 0.16, "learning_rate": 1.9087264778787534e-05, "loss": 1.6841, "step": 534 }, { "epoch": 0.16, "learning_rate": 1.9083100126070375e-05, "loss": 1.4274, "step": 535 }, { "epoch": 0.16, "learning_rate": 1.907892645013639e-05, "loss": 1.5352, "step": 536 }, { "epoch": 0.17, "learning_rate": 1.9074743755131743e-05, "loss": 1.7034, "step": 537 }, { "epoch": 0.17, "learning_rate": 1.9070552045211544e-05, "loss": 1.8087, "step": 538 }, { "epoch": 0.17, "learning_rate": 1.906635132453988e-05, "loss": 1.2675, "step": 539 }, { "epoch": 0.17, "learning_rate": 1.9062141597289773e-05, "loss": 1.3889, "step": 540 }, { "epoch": 0.17, "learning_rate": 1.9057922867643196e-05, "loss": 1.4754, "step": 541 }, { "epoch": 0.17, "learning_rate": 1.9053695139791058e-05, "loss": 1.6518, "step": 542 }, { "epoch": 0.17, "learning_rate": 1.9049458417933225e-05, "loss": 1.5358, "step": 543 }, { "epoch": 0.17, "learning_rate": 1.9045212706278485e-05, "loss": 1.6428, "step": 544 }, { "epoch": 0.17, "learning_rate": 1.904095800904455e-05, "loss": 1.408, "step": 545 }, { "epoch": 0.17, "learning_rate": 1.9036694330458077e-05, "loss": 2.1169, "step": 546 }, { "epoch": 0.17, "learning_rate": 1.903242167475463e-05, "loss": 2.1308, "step": 547 }, { "epoch": 0.17, "learning_rate": 1.9028140046178696e-05, "loss": 1.5411, "step": 548 }, { "epoch": 0.17, "learning_rate": 1.9023849448983674e-05, "loss": 1.5414, "step": 549 }, { "epoch": 0.17, "learning_rate": 1.901954988743188e-05, "loss": 1.3376, "step": 550 }, { "epoch": 0.17, "learning_rate": 1.901524136579452e-05, "loss": 1.2951, "step": 551 }, { "epoch": 0.17, "learning_rate": 1.9010923888351717e-05, "loss": 2.0171, "step": 552 }, { "epoch": 0.17, "learning_rate": 1.9006597459392483e-05, "loss": 1.5098, "step": 553 }, { "epoch": 0.17, "learning_rate": 1.9002262083214728e-05, "loss": 1.6109, "step": 554 }, { "epoch": 0.17, "learning_rate": 1.8997917764125233e-05, "loss": 1.3259, "step": 555 }, { "epoch": 0.17, "learning_rate": 1.8993564506439687e-05, "loss": 1.756, "step": 556 }, { "epoch": 0.17, "learning_rate": 1.8989202314482645e-05, "loss": 1.4023, "step": 557 }, { "epoch": 0.17, "learning_rate": 1.898483119258754e-05, "loss": 1.2758, "step": 558 }, { "epoch": 0.17, "learning_rate": 1.898045114509667e-05, "loss": 1.4951, "step": 559 }, { "epoch": 0.17, "learning_rate": 1.8976062176361213e-05, "loss": 1.7553, "step": 560 }, { "epoch": 0.17, "learning_rate": 1.89716642907412e-05, "loss": 1.6671, "step": 561 }, { "epoch": 0.17, "learning_rate": 1.8967257492605522e-05, "loss": 1.5865, "step": 562 }, { "epoch": 0.17, "learning_rate": 1.8962841786331923e-05, "loss": 1.3889, "step": 563 }, { "epoch": 0.17, "learning_rate": 1.8958417176307e-05, "loss": 1.5483, "step": 564 }, { "epoch": 0.17, "learning_rate": 1.8953983666926182e-05, "loss": 1.9848, "step": 565 }, { "epoch": 0.17, "learning_rate": 1.8949541262593764e-05, "loss": 1.4827, "step": 566 }, { "epoch": 0.17, "learning_rate": 1.8945089967722853e-05, "loss": 2.1598, "step": 567 }, { "epoch": 0.17, "learning_rate": 1.89406297867354e-05, "loss": 1.2254, "step": 568 }, { "epoch": 0.18, "learning_rate": 1.8936160724062182e-05, "loss": 1.6542, "step": 569 }, { "epoch": 0.18, "learning_rate": 1.8931682784142792e-05, "loss": 1.8275, "step": 570 }, { "epoch": 0.18, "learning_rate": 1.8927195971425655e-05, "loss": 1.6244, "step": 571 }, { "epoch": 0.18, "learning_rate": 1.8922700290367998e-05, "loss": 1.6699, "step": 572 }, { "epoch": 0.18, "learning_rate": 1.8918195745435866e-05, "loss": 1.4768, "step": 573 }, { "epoch": 0.18, "learning_rate": 1.8913682341104107e-05, "loss": 1.7655, "step": 574 }, { "epoch": 0.18, "learning_rate": 1.8909160081856365e-05, "loss": 1.4945, "step": 575 }, { "epoch": 0.18, "learning_rate": 1.8904628972185086e-05, "loss": 1.5845, "step": 576 }, { "epoch": 0.18, "learning_rate": 1.890008901659151e-05, "loss": 1.3189, "step": 577 }, { "epoch": 0.18, "learning_rate": 1.8895540219585656e-05, "loss": 2.177, "step": 578 }, { "epoch": 0.18, "learning_rate": 1.889098258568634e-05, "loss": 2.121, "step": 579 }, { "epoch": 0.18, "learning_rate": 1.888641611942114e-05, "loss": 1.397, "step": 580 }, { "epoch": 0.18, "learning_rate": 1.8881840825326416e-05, "loss": 1.6375, "step": 581 }, { "epoch": 0.18, "learning_rate": 1.8877256707947308e-05, "loss": 1.2365, "step": 582 }, { "epoch": 0.18, "learning_rate": 1.88726637718377e-05, "loss": 1.2064, "step": 583 }, { "epoch": 0.18, "learning_rate": 1.8868062021560254e-05, "loss": 1.3946, "step": 584 }, { "epoch": 0.18, "learning_rate": 1.8863451461686376e-05, "loss": 1.5162, "step": 585 }, { "epoch": 0.18, "learning_rate": 1.885883209679624e-05, "loss": 1.3973, "step": 586 }, { "epoch": 0.18, "learning_rate": 1.8854203931478742e-05, "loss": 1.8254, "step": 587 }, { "epoch": 0.18, "learning_rate": 1.884956697033155e-05, "loss": 1.6801, "step": 588 }, { "epoch": 0.18, "learning_rate": 1.884492121796104e-05, "loss": 1.3981, "step": 589 }, { "epoch": 0.18, "learning_rate": 1.8840266678982343e-05, "loss": 1.8412, "step": 590 }, { "epoch": 0.18, "learning_rate": 1.8835603358019313e-05, "loss": 1.675, "step": 591 }, { "epoch": 0.18, "learning_rate": 1.883093125970452e-05, "loss": 2.1703, "step": 592 }, { "epoch": 0.18, "learning_rate": 1.882625038867927e-05, "loss": 1.2109, "step": 593 }, { "epoch": 0.18, "learning_rate": 1.8821560749593564e-05, "loss": 1.5727, "step": 594 }, { "epoch": 0.18, "learning_rate": 1.8816862347106125e-05, "loss": 1.9582, "step": 595 }, { "epoch": 0.18, "learning_rate": 1.8812155185884385e-05, "loss": 1.484, "step": 596 }, { "epoch": 0.18, "learning_rate": 1.880743927060447e-05, "loss": 1.6283, "step": 597 }, { "epoch": 0.18, "learning_rate": 1.88027146059512e-05, "loss": 1.7687, "step": 598 }, { "epoch": 0.18, "learning_rate": 1.8797981196618095e-05, "loss": 1.586, "step": 599 }, { "epoch": 0.18, "learning_rate": 1.8793239047307358e-05, "loss": 1.5589, "step": 600 }, { "epoch": 0.18, "learning_rate": 1.8788488162729868e-05, "loss": 1.625, "step": 601 }, { "epoch": 0.19, "learning_rate": 1.8783728547605192e-05, "loss": 1.4686, "step": 602 }, { "epoch": 0.19, "learning_rate": 1.8778960206661567e-05, "loss": 1.3376, "step": 603 }, { "epoch": 0.19, "learning_rate": 1.8774183144635894e-05, "loss": 1.5638, "step": 604 }, { "epoch": 0.19, "learning_rate": 1.876939736627375e-05, "loss": 1.6563, "step": 605 }, { "epoch": 0.19, "learning_rate": 1.8764602876329346e-05, "loss": 1.8871, "step": 606 }, { "epoch": 0.19, "learning_rate": 1.8759799679565572e-05, "loss": 1.6298, "step": 607 }, { "epoch": 0.19, "learning_rate": 1.875498778075396e-05, "loss": 1.4083, "step": 608 }, { "epoch": 0.19, "learning_rate": 1.8750167184674682e-05, "loss": 1.3672, "step": 609 }, { "epoch": 0.19, "learning_rate": 1.874533789611655e-05, "loss": 1.859, "step": 610 }, { "epoch": 0.19, "learning_rate": 1.8740499919877017e-05, "loss": 1.3689, "step": 611 }, { "epoch": 0.19, "learning_rate": 1.8735653260762165e-05, "loss": 1.4764, "step": 612 }, { "epoch": 0.19, "learning_rate": 1.8730797923586696e-05, "loss": 1.2668, "step": 613 }, { "epoch": 0.19, "learning_rate": 1.872593391317394e-05, "loss": 1.7082, "step": 614 }, { "epoch": 0.19, "learning_rate": 1.872106123435584e-05, "loss": 1.4882, "step": 615 }, { "epoch": 0.19, "learning_rate": 1.8716179891972942e-05, "loss": 1.3451, "step": 616 }, { "epoch": 0.19, "learning_rate": 1.8711289890874415e-05, "loss": 1.742, "step": 617 }, { "epoch": 0.19, "learning_rate": 1.8706391235918022e-05, "loss": 1.7175, "step": 618 }, { "epoch": 0.19, "learning_rate": 1.870148393197012e-05, "loss": 1.3208, "step": 619 }, { "epoch": 0.19, "learning_rate": 1.8696567983905655e-05, "loss": 1.6317, "step": 620 }, { "epoch": 0.19, "learning_rate": 1.869164339660817e-05, "loss": 1.4781, "step": 621 }, { "epoch": 0.19, "learning_rate": 1.8686710174969786e-05, "loss": 1.383, "step": 622 }, { "epoch": 0.19, "learning_rate": 1.8681768323891196e-05, "loss": 1.5551, "step": 623 }, { "epoch": 0.19, "learning_rate": 1.8676817848281677e-05, "loss": 1.2818, "step": 624 }, { "epoch": 0.19, "learning_rate": 1.867185875305906e-05, "loss": 1.3895, "step": 625 }, { "epoch": 0.19, "learning_rate": 1.8666891043149757e-05, "loss": 1.9428, "step": 626 }, { "epoch": 0.19, "learning_rate": 1.866191472348871e-05, "loss": 1.4577, "step": 627 }, { "epoch": 0.19, "learning_rate": 1.8656929799019444e-05, "loss": 1.4175, "step": 628 }, { "epoch": 0.19, "learning_rate": 1.8651936274694013e-05, "loss": 1.3815, "step": 629 }, { "epoch": 0.19, "learning_rate": 1.8646934155473025e-05, "loss": 1.4734, "step": 630 }, { "epoch": 0.19, "learning_rate": 1.8641923446325615e-05, "loss": 1.6124, "step": 631 }, { "epoch": 0.19, "learning_rate": 1.8636904152229455e-05, "loss": 1.3308, "step": 632 }, { "epoch": 0.19, "learning_rate": 1.8631876278170758e-05, "loss": 1.5259, "step": 633 }, { "epoch": 0.2, "learning_rate": 1.8626839829144244e-05, "loss": 1.5456, "step": 634 }, { "epoch": 0.2, "learning_rate": 1.8621794810153155e-05, "loss": 1.691, "step": 635 }, { "epoch": 0.2, "learning_rate": 1.8616741226209256e-05, "loss": 1.3078, "step": 636 }, { "epoch": 0.2, "learning_rate": 1.8611679082332808e-05, "loss": 1.3482, "step": 637 }, { "epoch": 0.2, "learning_rate": 1.8606608383552583e-05, "loss": 1.8421, "step": 638 }, { "epoch": 0.2, "learning_rate": 1.860152913490585e-05, "loss": 1.2417, "step": 639 }, { "epoch": 0.2, "learning_rate": 1.859644134143837e-05, "loss": 1.6078, "step": 640 }, { "epoch": 0.2, "learning_rate": 1.8591345008204394e-05, "loss": 1.3818, "step": 641 }, { "epoch": 0.2, "learning_rate": 1.8586240140266655e-05, "loss": 2.0646, "step": 642 }, { "epoch": 0.2, "learning_rate": 1.858112674269637e-05, "loss": 1.7184, "step": 643 }, { "epoch": 0.2, "learning_rate": 1.8576004820573217e-05, "loss": 1.7486, "step": 644 }, { "epoch": 0.2, "learning_rate": 1.8570874378985357e-05, "loss": 1.6406, "step": 645 }, { "epoch": 0.2, "learning_rate": 1.8565735423029406e-05, "loss": 1.7162, "step": 646 }, { "epoch": 0.2, "learning_rate": 1.8560587957810434e-05, "loss": 1.7694, "step": 647 }, { "epoch": 0.2, "learning_rate": 1.8555431988441977e-05, "loss": 1.6327, "step": 648 }, { "epoch": 0.2, "learning_rate": 1.855026752004601e-05, "loss": 1.4946, "step": 649 }, { "epoch": 0.2, "learning_rate": 1.854509455775295e-05, "loss": 1.5323, "step": 650 }, { "epoch": 0.2, "learning_rate": 1.8539913106701663e-05, "loss": 1.3565, "step": 651 }, { "epoch": 0.2, "learning_rate": 1.853472317203943e-05, "loss": 1.6078, "step": 652 }, { "epoch": 0.2, "learning_rate": 1.8529524758921973e-05, "loss": 1.7001, "step": 653 }, { "epoch": 0.2, "learning_rate": 1.8524317872513434e-05, "loss": 1.809, "step": 654 }, { "epoch": 0.2, "learning_rate": 1.851910251798637e-05, "loss": 1.8589, "step": 655 }, { "epoch": 0.2, "learning_rate": 1.8513878700521752e-05, "loss": 1.46, "step": 656 }, { "epoch": 0.2, "learning_rate": 1.8508646425308958e-05, "loss": 1.6257, "step": 657 }, { "epoch": 0.2, "learning_rate": 1.8503405697545767e-05, "loss": 1.5721, "step": 658 }, { "epoch": 0.2, "learning_rate": 1.8498156522438356e-05, "loss": 1.4022, "step": 659 }, { "epoch": 0.2, "learning_rate": 1.8492898905201297e-05, "loss": 1.6251, "step": 660 }, { "epoch": 0.2, "learning_rate": 1.8487632851057533e-05, "loss": 1.4298, "step": 661 }, { "epoch": 0.2, "learning_rate": 1.8482358365238414e-05, "loss": 1.7755, "step": 662 }, { "epoch": 0.2, "learning_rate": 1.8477075452983643e-05, "loss": 1.6327, "step": 663 }, { "epoch": 0.2, "learning_rate": 1.847178411954131e-05, "loss": 1.754, "step": 664 }, { "epoch": 0.2, "learning_rate": 1.8466484370167855e-05, "loss": 1.5889, "step": 665 }, { "epoch": 0.2, "learning_rate": 1.8461176210128095e-05, "loss": 1.7053, "step": 666 }, { "epoch": 0.21, "learning_rate": 1.8455859644695186e-05, "loss": 1.5127, "step": 667 }, { "epoch": 0.21, "learning_rate": 1.845053467915065e-05, "loss": 1.4308, "step": 668 }, { "epoch": 0.21, "learning_rate": 1.8445201318784343e-05, "loss": 1.8029, "step": 669 }, { "epoch": 0.21, "learning_rate": 1.8439859568894464e-05, "loss": 1.8827, "step": 670 }, { "epoch": 0.21, "learning_rate": 1.8434509434787544e-05, "loss": 1.3758, "step": 671 }, { "epoch": 0.21, "learning_rate": 1.8429150921778448e-05, "loss": 1.404, "step": 672 }, { "epoch": 0.21, "learning_rate": 1.8423784035190354e-05, "loss": 1.485, "step": 673 }, { "epoch": 0.21, "learning_rate": 1.8418408780354776e-05, "loss": 1.54, "step": 674 }, { "epoch": 0.21, "learning_rate": 1.8413025162611525e-05, "loss": 1.6564, "step": 675 }, { "epoch": 0.21, "learning_rate": 1.8407633187308722e-05, "loss": 1.6289, "step": 676 }, { "epoch": 0.21, "learning_rate": 1.84022328598028e-05, "loss": 1.7564, "step": 677 }, { "epoch": 0.21, "learning_rate": 1.839682418545848e-05, "loss": 1.2386, "step": 678 }, { "epoch": 0.21, "learning_rate": 1.8391407169648777e-05, "loss": 1.7526, "step": 679 }, { "epoch": 0.21, "learning_rate": 1.8385981817754997e-05, "loss": 2.0724, "step": 680 }, { "epoch": 0.21, "learning_rate": 1.8380548135166725e-05, "loss": 1.9446, "step": 681 }, { "epoch": 0.21, "learning_rate": 1.837510612728181e-05, "loss": 1.5604, "step": 682 }, { "epoch": 0.21, "learning_rate": 1.8369655799506394e-05, "loss": 1.7632, "step": 683 }, { "epoch": 0.21, "learning_rate": 1.8364197157254863e-05, "loss": 1.3337, "step": 684 }, { "epoch": 0.21, "learning_rate": 1.835873020594988e-05, "loss": 1.5513, "step": 685 }, { "epoch": 0.21, "learning_rate": 1.8353254951022342e-05, "loss": 1.5122, "step": 686 }, { "epoch": 0.21, "learning_rate": 1.8347771397911414e-05, "loss": 1.9496, "step": 687 }, { "epoch": 0.21, "learning_rate": 1.83422795520645e-05, "loss": 1.3472, "step": 688 }, { "epoch": 0.21, "learning_rate": 1.833677941893723e-05, "loss": 1.2167, "step": 689 }, { "epoch": 0.21, "learning_rate": 1.833127100399348e-05, "loss": 2.1145, "step": 690 }, { "epoch": 0.21, "learning_rate": 1.8325754312705354e-05, "loss": 1.837, "step": 691 }, { "epoch": 0.21, "learning_rate": 1.8320229350553165e-05, "loss": 1.3419, "step": 692 }, { "epoch": 0.21, "learning_rate": 1.8314696123025456e-05, "loss": 1.6396, "step": 693 }, { "epoch": 0.21, "learning_rate": 1.8309154635618967e-05, "loss": 1.3863, "step": 694 }, { "epoch": 0.21, "learning_rate": 1.830360489383866e-05, "loss": 1.5589, "step": 695 }, { "epoch": 0.21, "learning_rate": 1.8298046903197685e-05, "loss": 1.1815, "step": 696 }, { "epoch": 0.21, "learning_rate": 1.8292480669217387e-05, "loss": 1.5832, "step": 697 }, { "epoch": 0.21, "learning_rate": 1.828690619742731e-05, "loss": 1.5901, "step": 698 }, { "epoch": 0.22, "learning_rate": 1.8281323493365165e-05, "loss": 1.657, "step": 699 }, { "epoch": 0.22, "learning_rate": 1.8275732562576857e-05, "loss": 1.3439, "step": 700 }, { "epoch": 0.22, "learning_rate": 1.8270133410616453e-05, "loss": 1.8008, "step": 701 }, { "epoch": 0.22, "learning_rate": 1.8264526043046193e-05, "loss": 1.3573, "step": 702 }, { "epoch": 0.22, "learning_rate": 1.8258910465436476e-05, "loss": 1.4711, "step": 703 }, { "epoch": 0.22, "learning_rate": 1.8253286683365855e-05, "loss": 1.2365, "step": 704 }, { "epoch": 0.22, "learning_rate": 1.8247654702421037e-05, "loss": 1.6896, "step": 705 }, { "epoch": 0.22, "learning_rate": 1.8242014528196875e-05, "loss": 1.6622, "step": 706 }, { "epoch": 0.22, "learning_rate": 1.8236366166296356e-05, "loss": 1.7212, "step": 707 }, { "epoch": 0.22, "learning_rate": 1.8230709622330596e-05, "loss": 1.7019, "step": 708 }, { "epoch": 0.22, "learning_rate": 1.8225044901918862e-05, "loss": 1.8005, "step": 709 }, { "epoch": 0.22, "learning_rate": 1.8219372010688516e-05, "loss": 1.586, "step": 710 }, { "epoch": 0.22, "learning_rate": 1.821369095427505e-05, "loss": 1.8295, "step": 711 }, { "epoch": 0.22, "learning_rate": 1.820800173832207e-05, "loss": 1.2133, "step": 712 }, { "epoch": 0.22, "learning_rate": 1.8202304368481285e-05, "loss": 1.6126, "step": 713 }, { "epoch": 0.22, "learning_rate": 1.8196598850412496e-05, "loss": 1.5056, "step": 714 }, { "epoch": 0.22, "learning_rate": 1.819088518978361e-05, "loss": 1.4861, "step": 715 }, { "epoch": 0.22, "learning_rate": 1.8185163392270623e-05, "loss": 1.5699, "step": 716 }, { "epoch": 0.22, "learning_rate": 1.8179433463557603e-05, "loss": 1.6072, "step": 717 }, { "epoch": 0.22, "learning_rate": 1.8173695409336703e-05, "loss": 1.5056, "step": 718 }, { "epoch": 0.22, "learning_rate": 1.8167949235308148e-05, "loss": 1.5162, "step": 719 }, { "epoch": 0.22, "learning_rate": 1.8162194947180228e-05, "loss": 1.6402, "step": 720 }, { "epoch": 0.22, "learning_rate": 1.8156432550669298e-05, "loss": 1.6683, "step": 721 }, { "epoch": 0.22, "learning_rate": 1.815066205149976e-05, "loss": 1.6646, "step": 722 }, { "epoch": 0.22, "learning_rate": 1.8144883455404072e-05, "loss": 1.959, "step": 723 }, { "epoch": 0.22, "learning_rate": 1.8139096768122727e-05, "loss": 1.3213, "step": 724 }, { "epoch": 0.22, "learning_rate": 1.813330199540427e-05, "loss": 1.3985, "step": 725 }, { "epoch": 0.22, "learning_rate": 1.8127499143005266e-05, "loss": 1.4718, "step": 726 }, { "epoch": 0.22, "learning_rate": 1.812168821669031e-05, "loss": 1.2316, "step": 727 }, { "epoch": 0.22, "learning_rate": 1.8115869222232015e-05, "loss": 2.1794, "step": 728 }, { "epoch": 0.22, "learning_rate": 1.8110042165411016e-05, "loss": 1.6209, "step": 729 }, { "epoch": 0.22, "learning_rate": 1.8104207052015952e-05, "loss": 1.7872, "step": 730 }, { "epoch": 0.22, "learning_rate": 1.8098363887843467e-05, "loss": 1.4011, "step": 731 }, { "epoch": 0.23, "learning_rate": 1.80925126786982e-05, "loss": 1.5551, "step": 732 }, { "epoch": 0.23, "learning_rate": 1.8086653430392784e-05, "loss": 1.7005, "step": 733 }, { "epoch": 0.23, "learning_rate": 1.8080786148747842e-05, "loss": 1.366, "step": 734 }, { "epoch": 0.23, "learning_rate": 1.8074910839591972e-05, "loss": 1.555, "step": 735 }, { "epoch": 0.23, "learning_rate": 1.8069027508761747e-05, "loss": 1.5576, "step": 736 }, { "epoch": 0.23, "learning_rate": 1.8063136162101715e-05, "loss": 1.5748, "step": 737 }, { "epoch": 0.23, "learning_rate": 1.805723680546437e-05, "loss": 1.8539, "step": 738 }, { "epoch": 0.23, "learning_rate": 1.8051329444710187e-05, "loss": 1.4886, "step": 739 }, { "epoch": 0.23, "learning_rate": 1.8045414085707578e-05, "loss": 1.3864, "step": 740 }, { "epoch": 0.23, "learning_rate": 1.80394907343329e-05, "loss": 1.4891, "step": 741 }, { "epoch": 0.23, "learning_rate": 1.8033559396470455e-05, "loss": 1.7299, "step": 742 }, { "epoch": 0.23, "learning_rate": 1.8027620078012474e-05, "loss": 1.4252, "step": 743 }, { "epoch": 0.23, "learning_rate": 1.802167278485912e-05, "loss": 1.5474, "step": 744 }, { "epoch": 0.23, "learning_rate": 1.801571752291848e-05, "loss": 1.4145, "step": 745 }, { "epoch": 0.23, "learning_rate": 1.8009754298106553e-05, "loss": 1.5132, "step": 746 }, { "epoch": 0.23, "learning_rate": 1.8003783116347245e-05, "loss": 2.0066, "step": 747 }, { "epoch": 0.23, "learning_rate": 1.7997803983572372e-05, "loss": 1.62, "step": 748 }, { "epoch": 0.23, "learning_rate": 1.7991816905721654e-05, "loss": 1.4211, "step": 749 }, { "epoch": 0.23, "learning_rate": 1.7985821888742687e-05, "loss": 1.5473, "step": 750 }, { "epoch": 0.23, "learning_rate": 1.7979818938590967e-05, "loss": 1.5114, "step": 751 }, { "epoch": 0.23, "learning_rate": 1.7973808061229873e-05, "loss": 1.3383, "step": 752 }, { "epoch": 0.23, "learning_rate": 1.7967789262630647e-05, "loss": 1.616, "step": 753 }, { "epoch": 0.23, "learning_rate": 1.7961762548772413e-05, "loss": 1.7893, "step": 754 }, { "epoch": 0.23, "learning_rate": 1.7955727925642147e-05, "loss": 1.4786, "step": 755 }, { "epoch": 0.23, "learning_rate": 1.794968539923469e-05, "loss": 1.6082, "step": 756 }, { "epoch": 0.23, "learning_rate": 1.794363497555273e-05, "loss": 1.5085, "step": 757 }, { "epoch": 0.23, "learning_rate": 1.79375766606068e-05, "loss": 1.2891, "step": 758 }, { "epoch": 0.23, "learning_rate": 1.7931510460415272e-05, "loss": 1.6728, "step": 759 }, { "epoch": 0.23, "learning_rate": 1.7925436381004356e-05, "loss": 1.4013, "step": 760 }, { "epoch": 0.23, "learning_rate": 1.7919354428408086e-05, "loss": 1.8139, "step": 761 }, { "epoch": 0.23, "learning_rate": 1.7913264608668316e-05, "loss": 1.4135, "step": 762 }, { "epoch": 0.23, "learning_rate": 1.7907166927834717e-05, "loss": 1.4739, "step": 763 }, { "epoch": 0.24, "learning_rate": 1.790106139196477e-05, "loss": 1.303, "step": 764 }, { "epoch": 0.24, "learning_rate": 1.789494800712376e-05, "loss": 1.6892, "step": 765 }, { "epoch": 0.24, "learning_rate": 1.788882677938476e-05, "loss": 1.3484, "step": 766 }, { "epoch": 0.24, "learning_rate": 1.7882697714828647e-05, "loss": 1.5391, "step": 767 }, { "epoch": 0.24, "learning_rate": 1.7876560819544077e-05, "loss": 1.3175, "step": 768 }, { "epoch": 0.24, "learning_rate": 1.7870416099627488e-05, "loss": 1.5041, "step": 769 }, { "epoch": 0.24, "learning_rate": 1.7864263561183085e-05, "loss": 1.7336, "step": 770 }, { "epoch": 0.24, "learning_rate": 1.785810321032284e-05, "loss": 1.6633, "step": 771 }, { "epoch": 0.24, "learning_rate": 1.7851935053166503e-05, "loss": 1.6733, "step": 772 }, { "epoch": 0.24, "learning_rate": 1.7845759095841553e-05, "loss": 1.9953, "step": 773 }, { "epoch": 0.24, "learning_rate": 1.7839575344483237e-05, "loss": 1.371, "step": 774 }, { "epoch": 0.24, "learning_rate": 1.7833383805234536e-05, "loss": 1.3293, "step": 775 }, { "epoch": 0.24, "learning_rate": 1.782718448424617e-05, "loss": 1.4687, "step": 776 }, { "epoch": 0.24, "learning_rate": 1.7820977387676583e-05, "loss": 1.587, "step": 777 }, { "epoch": 0.24, "learning_rate": 1.781476252169196e-05, "loss": 1.3963, "step": 778 }, { "epoch": 0.24, "learning_rate": 1.7808539892466186e-05, "loss": 1.3674, "step": 779 }, { "epoch": 0.24, "learning_rate": 1.7802309506180867e-05, "loss": 1.4119, "step": 780 }, { "epoch": 0.24, "learning_rate": 1.779607136902531e-05, "loss": 1.4019, "step": 781 }, { "epoch": 0.24, "learning_rate": 1.7789825487196538e-05, "loss": 1.4235, "step": 782 }, { "epoch": 0.24, "learning_rate": 1.7783571866899242e-05, "loss": 1.6729, "step": 783 }, { "epoch": 0.24, "learning_rate": 1.7777310514345816e-05, "loss": 2.3911, "step": 784 }, { "epoch": 0.24, "learning_rate": 1.7771041435756332e-05, "loss": 1.5293, "step": 785 }, { "epoch": 0.24, "learning_rate": 1.7764764637358545e-05, "loss": 1.5793, "step": 786 }, { "epoch": 0.24, "learning_rate": 1.775848012538786e-05, "loss": 1.7615, "step": 787 }, { "epoch": 0.24, "learning_rate": 1.7752187906087366e-05, "loss": 1.8085, "step": 788 }, { "epoch": 0.24, "learning_rate": 1.7745887985707796e-05, "loss": 1.6918, "step": 789 }, { "epoch": 0.24, "learning_rate": 1.7739580370507533e-05, "loss": 1.8163, "step": 790 }, { "epoch": 0.24, "learning_rate": 1.7733265066752606e-05, "loss": 1.7952, "step": 791 }, { "epoch": 0.24, "learning_rate": 1.7726942080716693e-05, "loss": 1.5377, "step": 792 }, { "epoch": 0.24, "learning_rate": 1.7720611418681082e-05, "loss": 1.4194, "step": 793 }, { "epoch": 0.24, "learning_rate": 1.7714273086934705e-05, "loss": 1.4771, "step": 794 }, { "epoch": 0.24, "learning_rate": 1.7707927091774103e-05, "loss": 1.5729, "step": 795 }, { "epoch": 0.24, "learning_rate": 1.7701573439503437e-05, "loss": 2.0678, "step": 796 }, { "epoch": 0.25, "learning_rate": 1.769521213643447e-05, "loss": 1.1961, "step": 797 }, { "epoch": 0.25, "learning_rate": 1.768884318888656e-05, "loss": 2.1071, "step": 798 }, { "epoch": 0.25, "learning_rate": 1.768246660318667e-05, "loss": 1.6993, "step": 799 }, { "epoch": 0.25, "learning_rate": 1.7676082385669343e-05, "loss": 1.6004, "step": 800 }, { "epoch": 0.25, "learning_rate": 1.7669690542676713e-05, "loss": 1.4411, "step": 801 }, { "epoch": 0.25, "learning_rate": 1.7663291080558477e-05, "loss": 1.7872, "step": 802 }, { "epoch": 0.25, "learning_rate": 1.7656884005671912e-05, "loss": 1.7064, "step": 803 }, { "epoch": 0.25, "learning_rate": 1.7650469324381843e-05, "loss": 1.4028, "step": 804 }, { "epoch": 0.25, "learning_rate": 1.764404704306067e-05, "loss": 1.5014, "step": 805 }, { "epoch": 0.25, "learning_rate": 1.7637617168088327e-05, "loss": 1.4999, "step": 806 }, { "epoch": 0.25, "learning_rate": 1.76311797058523e-05, "loss": 1.5959, "step": 807 }, { "epoch": 0.25, "learning_rate": 1.762473466274761e-05, "loss": 1.6539, "step": 808 }, { "epoch": 0.25, "learning_rate": 1.7618282045176812e-05, "loss": 1.3852, "step": 809 }, { "epoch": 0.25, "learning_rate": 1.7611821859549977e-05, "loss": 1.5617, "step": 810 }, { "epoch": 0.25, "learning_rate": 1.76053541122847e-05, "loss": 1.3193, "step": 811 }, { "epoch": 0.25, "learning_rate": 1.7598878809806095e-05, "loss": 1.3467, "step": 812 }, { "epoch": 0.25, "learning_rate": 1.7592395958546767e-05, "loss": 1.4551, "step": 813 }, { "epoch": 0.25, "learning_rate": 1.758590556494683e-05, "loss": 1.4005, "step": 814 }, { "epoch": 0.25, "learning_rate": 1.7579407635453884e-05, "loss": 1.4037, "step": 815 }, { "epoch": 0.25, "learning_rate": 1.7572902176523023e-05, "loss": 1.5949, "step": 816 }, { "epoch": 0.25, "learning_rate": 1.7566389194616816e-05, "loss": 1.4334, "step": 817 }, { "epoch": 0.25, "learning_rate": 1.7559868696205304e-05, "loss": 1.9066, "step": 818 }, { "epoch": 0.25, "learning_rate": 1.7553340687765996e-05, "loss": 1.6524, "step": 819 }, { "epoch": 0.25, "learning_rate": 1.7546805175783866e-05, "loss": 1.8763, "step": 820 }, { "epoch": 0.25, "learning_rate": 1.7540262166751336e-05, "loss": 1.458, "step": 821 }, { "epoch": 0.25, "learning_rate": 1.753371166716828e-05, "loss": 1.3904, "step": 822 }, { "epoch": 0.25, "learning_rate": 1.7527153683542007e-05, "loss": 1.4798, "step": 823 }, { "epoch": 0.25, "learning_rate": 1.752058822238727e-05, "loss": 1.3741, "step": 824 }, { "epoch": 0.25, "learning_rate": 1.7514015290226244e-05, "loss": 2.6011, "step": 825 }, { "epoch": 0.25, "learning_rate": 1.7507434893588525e-05, "loss": 1.5167, "step": 826 }, { "epoch": 0.25, "learning_rate": 1.7500847039011128e-05, "loss": 1.4155, "step": 827 }, { "epoch": 0.25, "learning_rate": 1.7494251733038473e-05, "loss": 1.4533, "step": 828 }, { "epoch": 0.26, "learning_rate": 1.748764898222239e-05, "loss": 1.5373, "step": 829 }, { "epoch": 0.26, "learning_rate": 1.748103879312209e-05, "loss": 1.5178, "step": 830 }, { "epoch": 0.26, "learning_rate": 1.7474421172304188e-05, "loss": 1.244, "step": 831 }, { "epoch": 0.26, "learning_rate": 1.7467796126342677e-05, "loss": 1.6269, "step": 832 }, { "epoch": 0.26, "learning_rate": 1.7461163661818917e-05, "loss": 1.4472, "step": 833 }, { "epoch": 0.26, "learning_rate": 1.745452378532166e-05, "loss": 1.652, "step": 834 }, { "epoch": 0.26, "learning_rate": 1.7447876503446996e-05, "loss": 1.8686, "step": 835 }, { "epoch": 0.26, "learning_rate": 1.7441221822798385e-05, "loss": 1.5792, "step": 836 }, { "epoch": 0.26, "learning_rate": 1.7434559749986633e-05, "loss": 1.3301, "step": 837 }, { "epoch": 0.26, "learning_rate": 1.7427890291629895e-05, "loss": 1.4879, "step": 838 }, { "epoch": 0.26, "learning_rate": 1.7421213454353657e-05, "loss": 1.7992, "step": 839 }, { "epoch": 0.26, "learning_rate": 1.7414529244790733e-05, "loss": 1.3126, "step": 840 }, { "epoch": 0.26, "learning_rate": 1.7407837669581274e-05, "loss": 1.6731, "step": 841 }, { "epoch": 0.26, "learning_rate": 1.740113873537273e-05, "loss": 1.5731, "step": 842 }, { "epoch": 0.26, "learning_rate": 1.7394432448819874e-05, "loss": 1.7039, "step": 843 }, { "epoch": 0.26, "learning_rate": 1.7387718816584775e-05, "loss": 1.5279, "step": 844 }, { "epoch": 0.26, "learning_rate": 1.738099784533681e-05, "loss": 1.4238, "step": 845 }, { "epoch": 0.26, "learning_rate": 1.737426954175264e-05, "loss": 1.3752, "step": 846 }, { "epoch": 0.26, "learning_rate": 1.7367533912516203e-05, "loss": 1.452, "step": 847 }, { "epoch": 0.26, "learning_rate": 1.7360790964318727e-05, "loss": 1.3888, "step": 848 }, { "epoch": 0.26, "learning_rate": 1.7354040703858707e-05, "loss": 1.7618, "step": 849 }, { "epoch": 0.26, "learning_rate": 1.734728313784189e-05, "loss": 1.9281, "step": 850 }, { "epoch": 0.26, "learning_rate": 1.7340518272981304e-05, "loss": 1.5503, "step": 851 }, { "epoch": 0.26, "learning_rate": 1.7333746115997204e-05, "loss": 1.5289, "step": 852 }, { "epoch": 0.26, "learning_rate": 1.7326966673617102e-05, "loss": 1.1858, "step": 853 }, { "epoch": 0.26, "learning_rate": 1.732017995257575e-05, "loss": 1.2547, "step": 854 }, { "epoch": 0.26, "learning_rate": 1.731338595961512e-05, "loss": 1.5849, "step": 855 }, { "epoch": 0.26, "learning_rate": 1.7306584701484414e-05, "loss": 1.832, "step": 856 }, { "epoch": 0.26, "learning_rate": 1.729977618494005e-05, "loss": 1.524, "step": 857 }, { "epoch": 0.26, "learning_rate": 1.7292960416745657e-05, "loss": 1.5202, "step": 858 }, { "epoch": 0.26, "learning_rate": 1.728613740367207e-05, "loss": 1.396, "step": 859 }, { "epoch": 0.26, "learning_rate": 1.7279307152497316e-05, "loss": 1.7042, "step": 860 }, { "epoch": 0.26, "learning_rate": 1.7272469670006625e-05, "loss": 1.5634, "step": 861 }, { "epoch": 0.27, "learning_rate": 1.726562496299239e-05, "loss": 2.8179, "step": 862 }, { "epoch": 0.27, "learning_rate": 1.7258773038254194e-05, "loss": 1.6097, "step": 863 }, { "epoch": 0.27, "learning_rate": 1.7251913902598795e-05, "loss": 1.5126, "step": 864 }, { "epoch": 0.27, "learning_rate": 1.72450475628401e-05, "loss": 1.2417, "step": 865 }, { "epoch": 0.27, "learning_rate": 1.7238174025799183e-05, "loss": 1.848, "step": 866 }, { "epoch": 0.27, "learning_rate": 1.7231293298304265e-05, "loss": 1.4678, "step": 867 }, { "epoch": 0.27, "learning_rate": 1.722440538719071e-05, "loss": 2.2675, "step": 868 }, { "epoch": 0.27, "learning_rate": 1.721751029930102e-05, "loss": 1.1174, "step": 869 }, { "epoch": 0.27, "learning_rate": 1.721060804148482e-05, "loss": 1.6167, "step": 870 }, { "epoch": 0.27, "learning_rate": 1.720369862059887e-05, "loss": 1.4021, "step": 871 }, { "epoch": 0.27, "learning_rate": 1.719678204350703e-05, "loss": 1.8564, "step": 872 }, { "epoch": 0.27, "learning_rate": 1.718985831708028e-05, "loss": 1.4963, "step": 873 }, { "epoch": 0.27, "learning_rate": 1.7182927448196705e-05, "loss": 1.3176, "step": 874 }, { "epoch": 0.27, "learning_rate": 1.7175989443741472e-05, "loss": 1.6269, "step": 875 }, { "epoch": 0.27, "learning_rate": 1.716904431060685e-05, "loss": 1.3289, "step": 876 }, { "epoch": 0.27, "learning_rate": 1.716209205569218e-05, "loss": 1.897, "step": 877 }, { "epoch": 0.27, "learning_rate": 1.7155132685903888e-05, "loss": 1.7436, "step": 878 }, { "epoch": 0.27, "learning_rate": 1.7148166208155453e-05, "loss": 1.5632, "step": 879 }, { "epoch": 0.27, "learning_rate": 1.714119262936743e-05, "loss": 1.454, "step": 880 }, { "epoch": 0.27, "learning_rate": 1.7134211956467425e-05, "loss": 1.6819, "step": 881 }, { "epoch": 0.27, "learning_rate": 1.712722419639008e-05, "loss": 1.491, "step": 882 }, { "epoch": 0.27, "learning_rate": 1.7120229356077095e-05, "loss": 1.9001, "step": 883 }, { "epoch": 0.27, "learning_rate": 1.7113227442477182e-05, "loss": 1.4566, "step": 884 }, { "epoch": 0.27, "learning_rate": 1.7106218462546107e-05, "loss": 1.6, "step": 885 }, { "epoch": 0.27, "learning_rate": 1.7099202423246632e-05, "loss": 1.4392, "step": 886 }, { "epoch": 0.27, "learning_rate": 1.7092179331548545e-05, "loss": 1.8749, "step": 887 }, { "epoch": 0.27, "learning_rate": 1.7085149194428628e-05, "loss": 1.3524, "step": 888 }, { "epoch": 0.27, "learning_rate": 1.707811201887068e-05, "loss": 1.5274, "step": 889 }, { "epoch": 0.27, "learning_rate": 1.7071067811865477e-05, "loss": 1.7023, "step": 890 }, { "epoch": 0.27, "learning_rate": 1.7064016580410782e-05, "loss": 1.4988, "step": 891 }, { "epoch": 0.27, "learning_rate": 1.7056958331511343e-05, "loss": 1.5491, "step": 892 }, { "epoch": 0.27, "learning_rate": 1.7049893072178874e-05, "loss": 1.6425, "step": 893 }, { "epoch": 0.28, "learning_rate": 1.7042820809432057e-05, "loss": 1.3709, "step": 894 }, { "epoch": 0.28, "learning_rate": 1.7035741550296522e-05, "loss": 1.5589, "step": 895 }, { "epoch": 0.28, "learning_rate": 1.7028655301804858e-05, "loss": 1.3701, "step": 896 }, { "epoch": 0.28, "learning_rate": 1.7021562070996597e-05, "loss": 1.4948, "step": 897 }, { "epoch": 0.28, "learning_rate": 1.7014461864918205e-05, "loss": 1.439, "step": 898 }, { "epoch": 0.28, "learning_rate": 1.7007354690623082e-05, "loss": 1.3605, "step": 899 }, { "epoch": 0.28, "learning_rate": 1.7000240555171536e-05, "loss": 1.7092, "step": 900 }, { "epoch": 0.28, "learning_rate": 1.6993119465630806e-05, "loss": 1.6787, "step": 901 }, { "epoch": 0.28, "learning_rate": 1.6985991429075038e-05, "loss": 1.4352, "step": 902 }, { "epoch": 0.28, "learning_rate": 1.6978856452585266e-05, "loss": 1.1501, "step": 903 }, { "epoch": 0.28, "learning_rate": 1.697171454324943e-05, "loss": 1.6532, "step": 904 }, { "epoch": 0.28, "learning_rate": 1.6964565708162354e-05, "loss": 1.9605, "step": 905 }, { "epoch": 0.28, "learning_rate": 1.6957409954425747e-05, "loss": 1.4744, "step": 906 }, { "epoch": 0.28, "learning_rate": 1.6950247289148176e-05, "loss": 1.6303, "step": 907 }, { "epoch": 0.28, "learning_rate": 1.694307771944509e-05, "loss": 1.3183, "step": 908 }, { "epoch": 0.28, "learning_rate": 1.6935901252438794e-05, "loss": 1.6299, "step": 909 }, { "epoch": 0.28, "learning_rate": 1.692871789525844e-05, "loss": 1.6264, "step": 910 }, { "epoch": 0.28, "learning_rate": 1.692152765504002e-05, "loss": 1.6717, "step": 911 }, { "epoch": 0.28, "learning_rate": 1.6914330538926377e-05, "loss": 1.7613, "step": 912 }, { "epoch": 0.28, "learning_rate": 1.6907126554067175e-05, "loss": 1.4282, "step": 913 }, { "epoch": 0.28, "learning_rate": 1.6899915707618905e-05, "loss": 1.2316, "step": 914 }, { "epoch": 0.28, "learning_rate": 1.6892698006744876e-05, "loss": 1.5535, "step": 915 }, { "epoch": 0.28, "learning_rate": 1.6885473458615204e-05, "loss": 1.514, "step": 916 }, { "epoch": 0.28, "learning_rate": 1.6878242070406797e-05, "loss": 1.4689, "step": 917 }, { "epoch": 0.28, "learning_rate": 1.687100384930338e-05, "loss": 1.58, "step": 918 }, { "epoch": 0.28, "learning_rate": 1.686375880249545e-05, "loss": 1.5177, "step": 919 }, { "epoch": 0.28, "learning_rate": 1.685650693718029e-05, "loss": 1.3925, "step": 920 }, { "epoch": 0.28, "learning_rate": 1.6849248260561947e-05, "loss": 1.3467, "step": 921 }, { "epoch": 0.28, "learning_rate": 1.6841982779851254e-05, "loss": 1.4362, "step": 922 }, { "epoch": 0.28, "learning_rate": 1.683471050226578e-05, "loss": 2.2228, "step": 923 }, { "epoch": 0.28, "learning_rate": 1.682743143502987e-05, "loss": 1.5136, "step": 924 }, { "epoch": 0.28, "learning_rate": 1.6820145585374588e-05, "loss": 1.7973, "step": 925 }, { "epoch": 0.28, "learning_rate": 1.6812852960537763e-05, "loss": 1.4953, "step": 926 }, { "epoch": 0.29, "learning_rate": 1.680555356776393e-05, "loss": 1.5916, "step": 927 }, { "epoch": 0.29, "learning_rate": 1.6798247414304363e-05, "loss": 2.7003, "step": 928 }, { "epoch": 0.29, "learning_rate": 1.679093450741705e-05, "loss": 1.3493, "step": 929 }, { "epoch": 0.29, "learning_rate": 1.678361485436668e-05, "loss": 1.7298, "step": 930 }, { "epoch": 0.29, "learning_rate": 1.677628846242465e-05, "loss": 1.5449, "step": 931 }, { "epoch": 0.29, "learning_rate": 1.676895533886905e-05, "loss": 1.4672, "step": 932 }, { "epoch": 0.29, "learning_rate": 1.676161549098466e-05, "loss": 1.5761, "step": 933 }, { "epoch": 0.29, "learning_rate": 1.6754268926062936e-05, "loss": 1.4477, "step": 934 }, { "epoch": 0.29, "learning_rate": 1.6746915651402006e-05, "loss": 1.5947, "step": 935 }, { "epoch": 0.29, "learning_rate": 1.6739555674306675e-05, "loss": 1.5135, "step": 936 }, { "epoch": 0.29, "learning_rate": 1.6732189002088386e-05, "loss": 1.698, "step": 937 }, { "epoch": 0.29, "learning_rate": 1.6724815642065245e-05, "loss": 1.3135, "step": 938 }, { "epoch": 0.29, "learning_rate": 1.671743560156201e-05, "loss": 1.4586, "step": 939 }, { "epoch": 0.29, "learning_rate": 1.6710048887910058e-05, "loss": 1.6851, "step": 940 }, { "epoch": 0.29, "learning_rate": 1.6702655508447402e-05, "loss": 1.4078, "step": 941 }, { "epoch": 0.29, "learning_rate": 1.6695255470518687e-05, "loss": 1.4554, "step": 942 }, { "epoch": 0.29, "learning_rate": 1.6687848781475156e-05, "loss": 1.6047, "step": 943 }, { "epoch": 0.29, "learning_rate": 1.6680435448674673e-05, "loss": 1.4915, "step": 944 }, { "epoch": 0.29, "learning_rate": 1.6673015479481688e-05, "loss": 1.7276, "step": 945 }, { "epoch": 0.29, "learning_rate": 1.666558888126726e-05, "loss": 1.425, "step": 946 }, { "epoch": 0.29, "learning_rate": 1.6658155661409017e-05, "loss": 1.4183, "step": 947 }, { "epoch": 0.29, "learning_rate": 1.6650715827291178e-05, "loss": 1.4812, "step": 948 }, { "epoch": 0.29, "learning_rate": 1.6643269386304526e-05, "loss": 1.382, "step": 949 }, { "epoch": 0.29, "learning_rate": 1.6635816345846413e-05, "loss": 1.5144, "step": 950 }, { "epoch": 0.29, "learning_rate": 1.6628356713320737e-05, "loss": 1.4955, "step": 951 }, { "epoch": 0.29, "learning_rate": 1.6620890496137953e-05, "loss": 1.5889, "step": 952 }, { "epoch": 0.29, "learning_rate": 1.6613417701715056e-05, "loss": 1.3152, "step": 953 }, { "epoch": 0.29, "learning_rate": 1.660593833747557e-05, "loss": 1.7075, "step": 954 }, { "epoch": 0.29, "learning_rate": 1.659845241084955e-05, "loss": 1.7521, "step": 955 }, { "epoch": 0.29, "learning_rate": 1.6590959929273574e-05, "loss": 1.7181, "step": 956 }, { "epoch": 0.29, "learning_rate": 1.658346090019072e-05, "loss": 1.6183, "step": 957 }, { "epoch": 0.29, "learning_rate": 1.6575955331050585e-05, "loss": 1.6797, "step": 958 }, { "epoch": 0.3, "learning_rate": 1.6568443229309247e-05, "loss": 1.66, "step": 959 }, { "epoch": 0.3, "learning_rate": 1.656092460242929e-05, "loss": 1.5973, "step": 960 }, { "epoch": 0.3, "learning_rate": 1.655339945787976e-05, "loss": 1.8312, "step": 961 }, { "epoch": 0.3, "learning_rate": 1.6545867803136203e-05, "loss": 1.4946, "step": 962 }, { "epoch": 0.3, "learning_rate": 1.6538329645680612e-05, "loss": 1.4063, "step": 963 }, { "epoch": 0.3, "learning_rate": 1.6530784993001448e-05, "loss": 1.3378, "step": 964 }, { "epoch": 0.3, "learning_rate": 1.6523233852593622e-05, "loss": 1.5978, "step": 965 }, { "epoch": 0.3, "learning_rate": 1.6515676231958488e-05, "loss": 1.4422, "step": 966 }, { "epoch": 0.3, "learning_rate": 1.650811213860385e-05, "loss": 2.9863, "step": 967 }, { "epoch": 0.3, "learning_rate": 1.650054158004392e-05, "loss": 1.4816, "step": 968 }, { "epoch": 0.3, "learning_rate": 1.649296456379936e-05, "loss": 1.3404, "step": 969 }, { "epoch": 0.3, "learning_rate": 1.6485381097397223e-05, "loss": 1.7107, "step": 970 }, { "epoch": 0.3, "learning_rate": 1.6477791188370978e-05, "loss": 1.4126, "step": 971 }, { "epoch": 0.3, "learning_rate": 1.6470194844260497e-05, "loss": 1.4862, "step": 972 }, { "epoch": 0.3, "learning_rate": 1.6462592072612044e-05, "loss": 1.4365, "step": 973 }, { "epoch": 0.3, "learning_rate": 1.645498288097827e-05, "loss": 1.7108, "step": 974 }, { "epoch": 0.3, "learning_rate": 1.644736727691819e-05, "loss": 1.4595, "step": 975 }, { "epoch": 0.3, "learning_rate": 1.643974526799721e-05, "loss": 1.6194, "step": 976 }, { "epoch": 0.3, "learning_rate": 1.643211686178708e-05, "loss": 1.5961, "step": 977 }, { "epoch": 0.3, "learning_rate": 1.6424482065865925e-05, "loss": 1.5277, "step": 978 }, { "epoch": 0.3, "learning_rate": 1.6416840887818195e-05, "loss": 1.5029, "step": 979 }, { "epoch": 0.3, "learning_rate": 1.6409193335234698e-05, "loss": 1.7965, "step": 980 }, { "epoch": 0.3, "learning_rate": 1.640153941571256e-05, "loss": 1.2555, "step": 981 }, { "epoch": 0.3, "learning_rate": 1.6393879136855247e-05, "loss": 1.7234, "step": 982 }, { "epoch": 0.3, "learning_rate": 1.6386212506272533e-05, "loss": 1.5772, "step": 983 }, { "epoch": 0.3, "learning_rate": 1.6378539531580502e-05, "loss": 1.4364, "step": 984 }, { "epoch": 0.3, "learning_rate": 1.6370860220401547e-05, "loss": 1.3704, "step": 985 }, { "epoch": 0.3, "learning_rate": 1.636317458036434e-05, "loss": 1.1916, "step": 986 }, { "epoch": 0.3, "learning_rate": 1.6355482619103866e-05, "loss": 1.4641, "step": 987 }, { "epoch": 0.3, "learning_rate": 1.6347784344261365e-05, "loss": 1.4288, "step": 988 }, { "epoch": 0.3, "learning_rate": 1.6340079763484357e-05, "loss": 1.7138, "step": 989 }, { "epoch": 0.3, "learning_rate": 1.633236888442663e-05, "loss": 1.9642, "step": 990 }, { "epoch": 0.3, "learning_rate": 1.6324651714748226e-05, "loss": 1.6564, "step": 991 }, { "epoch": 0.31, "learning_rate": 1.6316928262115443e-05, "loss": 1.6127, "step": 992 }, { "epoch": 0.31, "learning_rate": 1.6309198534200807e-05, "loss": 1.3114, "step": 993 }, { "epoch": 0.31, "learning_rate": 1.6301462538683087e-05, "loss": 1.9242, "step": 994 }, { "epoch": 0.31, "learning_rate": 1.6293720283247276e-05, "loss": 2.0184, "step": 995 }, { "epoch": 0.31, "learning_rate": 1.6285971775584587e-05, "loss": 1.5211, "step": 996 }, { "epoch": 0.31, "learning_rate": 1.6278217023392447e-05, "loss": 1.2897, "step": 997 }, { "epoch": 0.31, "learning_rate": 1.6270456034374477e-05, "loss": 1.6037, "step": 998 }, { "epoch": 0.31, "learning_rate": 1.62626888162405e-05, "loss": 1.7091, "step": 999 }, { "epoch": 0.31, "learning_rate": 1.6254915376706526e-05, "loss": 1.7262, "step": 1000 }, { "epoch": 0.31, "learning_rate": 1.624713572349475e-05, "loss": 1.5866, "step": 1001 }, { "epoch": 0.31, "learning_rate": 1.6239349864333527e-05, "loss": 1.532, "step": 1002 }, { "epoch": 0.31, "learning_rate": 1.623155780695739e-05, "loss": 1.3556, "step": 1003 }, { "epoch": 0.31, "learning_rate": 1.6223759559107028e-05, "loss": 1.5362, "step": 1004 }, { "epoch": 0.31, "learning_rate": 1.6215955128529273e-05, "loss": 1.68, "step": 1005 }, { "epoch": 0.31, "learning_rate": 1.62081445229771e-05, "loss": 1.285, "step": 1006 }, { "epoch": 0.31, "learning_rate": 1.620032775020963e-05, "loss": 1.5264, "step": 1007 }, { "epoch": 0.31, "learning_rate": 1.6192504817992086e-05, "loss": 1.4866, "step": 1008 }, { "epoch": 0.31, "learning_rate": 1.6184675734095838e-05, "loss": 1.6023, "step": 1009 }, { "epoch": 0.31, "learning_rate": 1.6176840506298345e-05, "loss": 1.5414, "step": 1010 }, { "epoch": 0.31, "learning_rate": 1.6168999142383187e-05, "loss": 1.6073, "step": 1011 }, { "epoch": 0.31, "learning_rate": 1.6161151650140027e-05, "loss": 1.4433, "step": 1012 }, { "epoch": 0.31, "learning_rate": 1.6153298037364617e-05, "loss": 1.6553, "step": 1013 }, { "epoch": 0.31, "learning_rate": 1.61454383118588e-05, "loss": 1.4277, "step": 1014 }, { "epoch": 0.31, "learning_rate": 1.6137572481430474e-05, "loss": 1.4453, "step": 1015 }, { "epoch": 0.31, "learning_rate": 1.612970055389362e-05, "loss": 1.3933, "step": 1016 }, { "epoch": 0.31, "learning_rate": 1.6121822537068263e-05, "loss": 1.6303, "step": 1017 }, { "epoch": 0.31, "learning_rate": 1.611393843878048e-05, "loss": 1.9715, "step": 1018 }, { "epoch": 0.31, "learning_rate": 1.6106048266862395e-05, "loss": 1.383, "step": 1019 }, { "epoch": 0.31, "learning_rate": 1.609815202915216e-05, "loss": 1.6517, "step": 1020 }, { "epoch": 0.31, "learning_rate": 1.6090249733493952e-05, "loss": 1.5899, "step": 1021 }, { "epoch": 0.31, "learning_rate": 1.608234138773797e-05, "loss": 1.3725, "step": 1022 }, { "epoch": 0.31, "learning_rate": 1.6074426999740422e-05, "loss": 1.651, "step": 1023 }, { "epoch": 0.32, "learning_rate": 1.6066506577363518e-05, "loss": 1.7959, "step": 1024 }, { "epoch": 0.32, "learning_rate": 1.6058580128475458e-05, "loss": 1.5919, "step": 1025 }, { "epoch": 0.32, "learning_rate": 1.605064766095044e-05, "loss": 1.3681, "step": 1026 }, { "epoch": 0.32, "learning_rate": 1.6042709182668628e-05, "loss": 2.5635, "step": 1027 }, { "epoch": 0.32, "learning_rate": 1.603476470151617e-05, "loss": 1.5505, "step": 1028 }, { "epoch": 0.32, "learning_rate": 1.6026814225385163e-05, "loss": 1.4891, "step": 1029 }, { "epoch": 0.32, "learning_rate": 1.6018857762173672e-05, "loss": 1.5213, "step": 1030 }, { "epoch": 0.32, "learning_rate": 1.60108953197857e-05, "loss": 1.7509, "step": 1031 }, { "epoch": 0.32, "learning_rate": 1.6002926906131202e-05, "loss": 1.4602, "step": 1032 }, { "epoch": 0.32, "learning_rate": 1.599495252912605e-05, "loss": 1.6008, "step": 1033 }, { "epoch": 0.32, "learning_rate": 1.5986972196692053e-05, "loss": 1.4041, "step": 1034 }, { "epoch": 0.32, "learning_rate": 1.5978985916756925e-05, "loss": 1.4373, "step": 1035 }, { "epoch": 0.32, "learning_rate": 1.5970993697254294e-05, "loss": 1.6824, "step": 1036 }, { "epoch": 0.32, "learning_rate": 1.5962995546123696e-05, "loss": 1.463, "step": 1037 }, { "epoch": 0.32, "learning_rate": 1.5954991471310542e-05, "loss": 1.6836, "step": 1038 }, { "epoch": 0.32, "learning_rate": 1.5946981480766146e-05, "loss": 1.3966, "step": 1039 }, { "epoch": 0.32, "learning_rate": 1.5938965582447683e-05, "loss": 1.0201, "step": 1040 }, { "epoch": 0.32, "learning_rate": 1.593094378431821e-05, "loss": 1.6136, "step": 1041 }, { "epoch": 0.32, "learning_rate": 1.5922916094346633e-05, "loss": 1.3993, "step": 1042 }, { "epoch": 0.32, "learning_rate": 1.5914882520507716e-05, "loss": 2.2084, "step": 1043 }, { "epoch": 0.32, "learning_rate": 1.5906843070782077e-05, "loss": 1.3329, "step": 1044 }, { "epoch": 0.32, "learning_rate": 1.5898797753156156e-05, "loss": 1.3613, "step": 1045 }, { "epoch": 0.32, "learning_rate": 1.589074657562223e-05, "loss": 1.5029, "step": 1046 }, { "epoch": 0.32, "learning_rate": 1.58826895461784e-05, "loss": 1.2901, "step": 1047 }, { "epoch": 0.32, "learning_rate": 1.5874626672828573e-05, "loss": 1.564, "step": 1048 }, { "epoch": 0.32, "learning_rate": 1.586655796358247e-05, "loss": 1.3343, "step": 1049 }, { "epoch": 0.32, "learning_rate": 1.58584834264556e-05, "loss": 1.7643, "step": 1050 }, { "epoch": 0.32, "learning_rate": 1.585040306946927e-05, "loss": 1.864, "step": 1051 }, { "epoch": 0.32, "learning_rate": 1.584231690065056e-05, "loss": 1.7747, "step": 1052 }, { "epoch": 0.32, "learning_rate": 1.583422492803233e-05, "loss": 1.9124, "step": 1053 }, { "epoch": 0.32, "learning_rate": 1.5826127159653203e-05, "loss": 1.5124, "step": 1054 }, { "epoch": 0.32, "learning_rate": 1.581802360355756e-05, "loss": 1.8209, "step": 1055 }, { "epoch": 0.32, "learning_rate": 1.5809914267795542e-05, "loss": 1.6177, "step": 1056 }, { "epoch": 0.33, "learning_rate": 1.5801799160423004e-05, "loss": 1.2172, "step": 1057 }, { "epoch": 0.33, "learning_rate": 1.5793678289501564e-05, "loss": 1.1287, "step": 1058 }, { "epoch": 0.33, "learning_rate": 1.5785551663098546e-05, "loss": 1.5319, "step": 1059 }, { "epoch": 0.33, "learning_rate": 1.5777419289287008e-05, "loss": 1.5833, "step": 1060 }, { "epoch": 0.33, "learning_rate": 1.57692811761457e-05, "loss": 1.5145, "step": 1061 }, { "epoch": 0.33, "learning_rate": 1.5761137331759084e-05, "loss": 1.556, "step": 1062 }, { "epoch": 0.33, "learning_rate": 1.5752987764217317e-05, "loss": 1.3611, "step": 1063 }, { "epoch": 0.33, "learning_rate": 1.5744832481616235e-05, "loss": 1.5687, "step": 1064 }, { "epoch": 0.33, "learning_rate": 1.5736671492057355e-05, "loss": 1.6609, "step": 1065 }, { "epoch": 0.33, "learning_rate": 1.5728504803647857e-05, "loss": 1.9621, "step": 1066 }, { "epoch": 0.33, "learning_rate": 1.57203324245006e-05, "loss": 1.5601, "step": 1067 }, { "epoch": 0.33, "learning_rate": 1.5712154362734073e-05, "loss": 1.3043, "step": 1068 }, { "epoch": 0.33, "learning_rate": 1.5703970626472424e-05, "loss": 1.3164, "step": 1069 }, { "epoch": 0.33, "learning_rate": 1.5695781223845442e-05, "loss": 1.3817, "step": 1070 }, { "epoch": 0.33, "learning_rate": 1.568758616298853e-05, "loss": 1.4874, "step": 1071 }, { "epoch": 0.33, "learning_rate": 1.5679385452042724e-05, "loss": 1.3455, "step": 1072 }, { "epoch": 0.33, "learning_rate": 1.567117909915467e-05, "loss": 1.8071, "step": 1073 }, { "epoch": 0.33, "learning_rate": 1.5662967112476613e-05, "loss": 1.3661, "step": 1074 }, { "epoch": 0.33, "learning_rate": 1.56547495001664e-05, "loss": 1.7409, "step": 1075 }, { "epoch": 0.33, "learning_rate": 1.5646526270387473e-05, "loss": 1.55, "step": 1076 }, { "epoch": 0.33, "learning_rate": 1.5638297431308846e-05, "loss": 1.31, "step": 1077 }, { "epoch": 0.33, "learning_rate": 1.56300629911051e-05, "loss": 1.6774, "step": 1078 }, { "epoch": 0.33, "learning_rate": 1.562182295795639e-05, "loss": 1.6859, "step": 1079 }, { "epoch": 0.33, "learning_rate": 1.5613577340048423e-05, "loss": 1.8374, "step": 1080 }, { "epoch": 0.33, "learning_rate": 1.560532614557246e-05, "loss": 2.1071, "step": 1081 }, { "epoch": 0.33, "learning_rate": 1.5597069382725293e-05, "loss": 1.1481, "step": 1082 }, { "epoch": 0.33, "learning_rate": 1.558880705970925e-05, "loss": 1.5379, "step": 1083 }, { "epoch": 0.33, "learning_rate": 1.5580539184732188e-05, "loss": 1.2301, "step": 1084 }, { "epoch": 0.33, "learning_rate": 1.557226576600746e-05, "loss": 1.4623, "step": 1085 }, { "epoch": 0.33, "learning_rate": 1.5563986811753948e-05, "loss": 1.4972, "step": 1086 }, { "epoch": 0.33, "learning_rate": 1.5555702330196024e-05, "loss": 1.512, "step": 1087 }, { "epoch": 0.33, "learning_rate": 1.554741232956355e-05, "loss": 1.4794, "step": 1088 }, { "epoch": 0.34, "learning_rate": 1.5539116818091873e-05, "loss": 1.6693, "step": 1089 }, { "epoch": 0.34, "learning_rate": 1.553081580402182e-05, "loss": 1.439, "step": 1090 }, { "epoch": 0.34, "learning_rate": 1.5522509295599662e-05, "loss": 1.4569, "step": 1091 }, { "epoch": 0.34, "learning_rate": 1.551419730107716e-05, "loss": 1.4267, "step": 1092 }, { "epoch": 0.34, "learning_rate": 1.5505879828711495e-05, "loss": 1.9064, "step": 1093 }, { "epoch": 0.34, "learning_rate": 1.5497556886765316e-05, "loss": 1.5052, "step": 1094 }, { "epoch": 0.34, "learning_rate": 1.548922848350669e-05, "loss": 1.5837, "step": 1095 }, { "epoch": 0.34, "learning_rate": 1.548089462720911e-05, "loss": 1.4804, "step": 1096 }, { "epoch": 0.34, "learning_rate": 1.5472555326151488e-05, "loss": 1.1238, "step": 1097 }, { "epoch": 0.34, "learning_rate": 1.5464210588618148e-05, "loss": 1.4484, "step": 1098 }, { "epoch": 0.34, "learning_rate": 1.5455860422898813e-05, "loss": 1.6398, "step": 1099 }, { "epoch": 0.34, "learning_rate": 1.544750483728859e-05, "loss": 1.9449, "step": 1100 }, { "epoch": 0.34, "learning_rate": 1.543914384008799e-05, "loss": 1.6514, "step": 1101 }, { "epoch": 0.34, "learning_rate": 1.5430777439602875e-05, "loss": 1.4899, "step": 1102 }, { "epoch": 0.34, "learning_rate": 1.54224056441445e-05, "loss": 1.3312, "step": 1103 }, { "epoch": 0.34, "learning_rate": 1.5414028462029454e-05, "loss": 1.6147, "step": 1104 }, { "epoch": 0.34, "learning_rate": 1.54056459015797e-05, "loss": 1.2962, "step": 1105 }, { "epoch": 0.34, "learning_rate": 1.5397257971122527e-05, "loss": 1.382, "step": 1106 }, { "epoch": 0.34, "learning_rate": 1.5388864678990568e-05, "loss": 2.0166, "step": 1107 }, { "epoch": 0.34, "learning_rate": 1.5380466033521783e-05, "loss": 1.2341, "step": 1108 }, { "epoch": 0.34, "learning_rate": 1.5372062043059436e-05, "loss": 1.6534, "step": 1109 }, { "epoch": 0.34, "learning_rate": 1.536365271595212e-05, "loss": 1.5405, "step": 1110 }, { "epoch": 0.34, "learning_rate": 1.5355238060553727e-05, "loss": 1.6008, "step": 1111 }, { "epoch": 0.34, "learning_rate": 1.5346818085223423e-05, "loss": 1.6252, "step": 1112 }, { "epoch": 0.34, "learning_rate": 1.5338392798325673e-05, "loss": 1.4996, "step": 1113 }, { "epoch": 0.34, "learning_rate": 1.532996220823023e-05, "loss": 1.4685, "step": 1114 }, { "epoch": 0.34, "learning_rate": 1.532152632331209e-05, "loss": 1.5744, "step": 1115 }, { "epoch": 0.34, "learning_rate": 1.5313085151951527e-05, "loss": 1.3161, "step": 1116 }, { "epoch": 0.34, "learning_rate": 1.5304638702534056e-05, "loss": 1.5155, "step": 1117 }, { "epoch": 0.34, "learning_rate": 1.529618698345045e-05, "loss": 1.4947, "step": 1118 }, { "epoch": 0.34, "learning_rate": 1.52877300030967e-05, "loss": 1.3002, "step": 1119 }, { "epoch": 0.34, "learning_rate": 1.5279267769874034e-05, "loss": 1.2234, "step": 1120 }, { "epoch": 0.34, "learning_rate": 1.527080029218889e-05, "loss": 1.342, "step": 1121 }, { "epoch": 0.35, "learning_rate": 1.5262327578452927e-05, "loss": 1.5391, "step": 1122 }, { "epoch": 0.35, "learning_rate": 1.5253849637082997e-05, "loss": 1.3691, "step": 1123 }, { "epoch": 0.35, "learning_rate": 1.5245366476501151e-05, "loss": 1.6297, "step": 1124 }, { "epoch": 0.35, "learning_rate": 1.5236878105134617e-05, "loss": 1.5292, "step": 1125 }, { "epoch": 0.35, "learning_rate": 1.5228384531415809e-05, "loss": 1.2432, "step": 1126 }, { "epoch": 0.35, "learning_rate": 1.5219885763782301e-05, "loss": 1.5855, "step": 1127 }, { "epoch": 0.35, "learning_rate": 1.5211381810676826e-05, "loss": 1.6148, "step": 1128 }, { "epoch": 0.35, "learning_rate": 1.5202872680547276e-05, "loss": 1.4231, "step": 1129 }, { "epoch": 0.35, "learning_rate": 1.5194358381846686e-05, "loss": 1.3389, "step": 1130 }, { "epoch": 0.35, "learning_rate": 1.5185838923033216e-05, "loss": 1.7355, "step": 1131 }, { "epoch": 0.35, "learning_rate": 1.5177314312570155e-05, "loss": 1.7297, "step": 1132 }, { "epoch": 0.35, "learning_rate": 1.516878455892592e-05, "loss": 1.5918, "step": 1133 }, { "epoch": 0.35, "learning_rate": 1.5160249670574026e-05, "loss": 1.6845, "step": 1134 }, { "epoch": 0.35, "learning_rate": 1.515170965599309e-05, "loss": 1.3584, "step": 1135 }, { "epoch": 0.35, "learning_rate": 1.5143164523666824e-05, "loss": 1.7132, "step": 1136 }, { "epoch": 0.35, "learning_rate": 1.5134614282084025e-05, "loss": 1.4241, "step": 1137 }, { "epoch": 0.35, "learning_rate": 1.5126058939738566e-05, "loss": 1.4952, "step": 1138 }, { "epoch": 0.35, "learning_rate": 1.5117498505129384e-05, "loss": 1.3988, "step": 1139 }, { "epoch": 0.35, "learning_rate": 1.510893298676048e-05, "loss": 1.4212, "step": 1140 }, { "epoch": 0.35, "learning_rate": 1.5100362393140891e-05, "loss": 1.5234, "step": 1141 }, { "epoch": 0.35, "learning_rate": 1.5091786732784717e-05, "loss": 2.0149, "step": 1142 }, { "epoch": 0.35, "learning_rate": 1.5083206014211071e-05, "loss": 1.5635, "step": 1143 }, { "epoch": 0.35, "learning_rate": 1.5074620245944112e-05, "loss": 1.5342, "step": 1144 }, { "epoch": 0.35, "learning_rate": 1.5066029436512992e-05, "loss": 1.4296, "step": 1145 }, { "epoch": 0.35, "learning_rate": 1.5057433594451891e-05, "loss": 1.6421, "step": 1146 }, { "epoch": 0.35, "learning_rate": 1.5048832728299978e-05, "loss": 1.2417, "step": 1147 }, { "epoch": 0.35, "learning_rate": 1.5040226846601413e-05, "loss": 1.6521, "step": 1148 }, { "epoch": 0.35, "learning_rate": 1.5031615957905343e-05, "loss": 1.5733, "step": 1149 }, { "epoch": 0.35, "learning_rate": 1.5023000070765886e-05, "loss": 1.4894, "step": 1150 }, { "epoch": 0.35, "learning_rate": 1.5014379193742123e-05, "loss": 1.6537, "step": 1151 }, { "epoch": 0.35, "learning_rate": 1.50057533353981e-05, "loss": 1.9206, "step": 1152 }, { "epoch": 0.35, "learning_rate": 1.4997122504302803e-05, "loss": 1.5018, "step": 1153 }, { "epoch": 0.36, "learning_rate": 1.4988486709030166e-05, "loss": 1.5288, "step": 1154 }, { "epoch": 0.36, "learning_rate": 1.4979845958159045e-05, "loss": 1.5589, "step": 1155 }, { "epoch": 0.36, "learning_rate": 1.4971200260273223e-05, "loss": 1.2415, "step": 1156 }, { "epoch": 0.36, "learning_rate": 1.4962549623961401e-05, "loss": 1.7756, "step": 1157 }, { "epoch": 0.36, "learning_rate": 1.495389405781719e-05, "loss": 1.7568, "step": 1158 }, { "epoch": 0.36, "learning_rate": 1.4945233570439079e-05, "loss": 1.4947, "step": 1159 }, { "epoch": 0.36, "learning_rate": 1.4936568170430465e-05, "loss": 1.6302, "step": 1160 }, { "epoch": 0.36, "learning_rate": 1.4927897866399619e-05, "loss": 1.6416, "step": 1161 }, { "epoch": 0.36, "learning_rate": 1.491922266695968e-05, "loss": 1.3903, "step": 1162 }, { "epoch": 0.36, "learning_rate": 1.4910542580728656e-05, "loss": 1.4154, "step": 1163 }, { "epoch": 0.36, "learning_rate": 1.4901857616329405e-05, "loss": 1.6626, "step": 1164 }, { "epoch": 0.36, "learning_rate": 1.4893167782389635e-05, "loss": 1.3699, "step": 1165 }, { "epoch": 0.36, "learning_rate": 1.488447308754189e-05, "loss": 1.2841, "step": 1166 }, { "epoch": 0.36, "learning_rate": 1.4875773540423538e-05, "loss": 1.8124, "step": 1167 }, { "epoch": 0.36, "learning_rate": 1.4867069149676774e-05, "loss": 1.6663, "step": 1168 }, { "epoch": 0.36, "learning_rate": 1.4858359923948604e-05, "loss": 1.1577, "step": 1169 }, { "epoch": 0.36, "learning_rate": 1.4849645871890832e-05, "loss": 1.7759, "step": 1170 }, { "epoch": 0.36, "learning_rate": 1.4840927002160062e-05, "loss": 1.2419, "step": 1171 }, { "epoch": 0.36, "learning_rate": 1.4832203323417679e-05, "loss": 1.3198, "step": 1172 }, { "epoch": 0.36, "learning_rate": 1.4823474844329852e-05, "loss": 1.5915, "step": 1173 }, { "epoch": 0.36, "learning_rate": 1.4814741573567514e-05, "loss": 1.5611, "step": 1174 }, { "epoch": 0.36, "learning_rate": 1.4806003519806363e-05, "loss": 1.842, "step": 1175 }, { "epoch": 0.36, "learning_rate": 1.4797260691726831e-05, "loss": 1.4088, "step": 1176 }, { "epoch": 0.36, "learning_rate": 1.4788513098014114e-05, "loss": 1.5827, "step": 1177 }, { "epoch": 0.36, "learning_rate": 1.4779760747358141e-05, "loss": 1.7068, "step": 1178 }, { "epoch": 0.36, "learning_rate": 1.4771003648453548e-05, "loss": 1.7779, "step": 1179 }, { "epoch": 0.36, "learning_rate": 1.4762241809999704e-05, "loss": 1.3589, "step": 1180 }, { "epoch": 0.36, "learning_rate": 1.475347524070069e-05, "loss": 1.7083, "step": 1181 }, { "epoch": 0.36, "learning_rate": 1.4744703949265268e-05, "loss": 1.4768, "step": 1182 }, { "epoch": 0.36, "learning_rate": 1.4735927944406909e-05, "loss": 1.5142, "step": 1183 }, { "epoch": 0.36, "learning_rate": 1.4727147234843752e-05, "loss": 1.8684, "step": 1184 }, { "epoch": 0.36, "learning_rate": 1.4718361829298626e-05, "loss": 1.4434, "step": 1185 }, { "epoch": 0.36, "learning_rate": 1.4709571736499008e-05, "loss": 1.508, "step": 1186 }, { "epoch": 0.37, "learning_rate": 1.4700776965177044e-05, "loss": 1.7898, "step": 1187 }, { "epoch": 0.37, "learning_rate": 1.4691977524069518e-05, "loss": 1.5498, "step": 1188 }, { "epoch": 0.37, "learning_rate": 1.4683173421917865e-05, "loss": 1.5826, "step": 1189 }, { "epoch": 0.37, "learning_rate": 1.467436466746814e-05, "loss": 1.973, "step": 1190 }, { "epoch": 0.37, "learning_rate": 1.4665551269471021e-05, "loss": 1.8257, "step": 1191 }, { "epoch": 0.37, "learning_rate": 1.4656733236681802e-05, "loss": 1.7267, "step": 1192 }, { "epoch": 0.37, "learning_rate": 1.4647910577860378e-05, "loss": 1.7979, "step": 1193 }, { "epoch": 0.37, "learning_rate": 1.4639083301771246e-05, "loss": 1.9642, "step": 1194 }, { "epoch": 0.37, "learning_rate": 1.4630251417183483e-05, "loss": 1.1606, "step": 1195 }, { "epoch": 0.37, "learning_rate": 1.4621414932870746e-05, "loss": 2.0086, "step": 1196 }, { "epoch": 0.37, "learning_rate": 1.4612573857611265e-05, "loss": 1.6921, "step": 1197 }, { "epoch": 0.37, "learning_rate": 1.4603728200187823e-05, "loss": 1.868, "step": 1198 }, { "epoch": 0.37, "learning_rate": 1.4594877969387761e-05, "loss": 1.6571, "step": 1199 }, { "epoch": 0.37, "learning_rate": 1.4586023174002961e-05, "loss": 1.5016, "step": 1200 }, { "epoch": 0.37, "learning_rate": 1.4577163822829845e-05, "loss": 1.5835, "step": 1201 }, { "epoch": 0.37, "learning_rate": 1.4568299924669351e-05, "loss": 1.893, "step": 1202 }, { "epoch": 0.37, "learning_rate": 1.4559431488326943e-05, "loss": 1.5338, "step": 1203 }, { "epoch": 0.37, "learning_rate": 1.4550558522612584e-05, "loss": 1.2956, "step": 1204 }, { "epoch": 0.37, "learning_rate": 1.4541681036340744e-05, "loss": 1.4081, "step": 1205 }, { "epoch": 0.37, "learning_rate": 1.4532799038330385e-05, "loss": 1.6298, "step": 1206 }, { "epoch": 0.37, "learning_rate": 1.4523912537404945e-05, "loss": 1.1129, "step": 1207 }, { "epoch": 0.37, "learning_rate": 1.4515021542392335e-05, "loss": 1.7905, "step": 1208 }, { "epoch": 0.37, "learning_rate": 1.4506126062124936e-05, "loss": 1.076, "step": 1209 }, { "epoch": 0.37, "learning_rate": 1.4497226105439586e-05, "loss": 1.2592, "step": 1210 }, { "epoch": 0.37, "learning_rate": 1.4488321681177558e-05, "loss": 1.4406, "step": 1211 }, { "epoch": 0.37, "learning_rate": 1.4479412798184576e-05, "loss": 1.6368, "step": 1212 }, { "epoch": 0.37, "learning_rate": 1.4470499465310786e-05, "loss": 1.5226, "step": 1213 }, { "epoch": 0.37, "learning_rate": 1.4461581691410757e-05, "loss": 1.418, "step": 1214 }, { "epoch": 0.37, "learning_rate": 1.4452659485343473e-05, "loss": 1.4452, "step": 1215 }, { "epoch": 0.37, "learning_rate": 1.4443732855972307e-05, "loss": 1.5885, "step": 1216 }, { "epoch": 0.37, "learning_rate": 1.4434801812165052e-05, "loss": 1.7341, "step": 1217 }, { "epoch": 0.37, "learning_rate": 1.4425866362793863e-05, "loss": 1.5881, "step": 1218 }, { "epoch": 0.38, "learning_rate": 1.4416926516735274e-05, "loss": 1.5061, "step": 1219 }, { "epoch": 0.38, "learning_rate": 1.4407982282870203e-05, "loss": 1.3805, "step": 1220 }, { "epoch": 0.38, "learning_rate": 1.4399033670083911e-05, "loss": 1.2547, "step": 1221 }, { "epoch": 0.38, "learning_rate": 1.4390080687266013e-05, "loss": 1.4346, "step": 1222 }, { "epoch": 0.38, "learning_rate": 1.4381123343310468e-05, "loss": 1.6429, "step": 1223 }, { "epoch": 0.38, "learning_rate": 1.4372161647115563e-05, "loss": 1.716, "step": 1224 }, { "epoch": 0.38, "learning_rate": 1.4363195607583919e-05, "loss": 1.6579, "step": 1225 }, { "epoch": 0.38, "learning_rate": 1.4354225233622457e-05, "loss": 1.4573, "step": 1226 }, { "epoch": 0.38, "learning_rate": 1.4345250534142408e-05, "loss": 1.443, "step": 1227 }, { "epoch": 0.38, "learning_rate": 1.433627151805931e-05, "loss": 1.7023, "step": 1228 }, { "epoch": 0.38, "learning_rate": 1.4327288194292978e-05, "loss": 1.1989, "step": 1229 }, { "epoch": 0.38, "learning_rate": 1.4318300571767514e-05, "loss": 1.4702, "step": 1230 }, { "epoch": 0.38, "learning_rate": 1.4309308659411282e-05, "loss": 2.2549, "step": 1231 }, { "epoch": 0.38, "learning_rate": 1.4300312466156917e-05, "loss": 1.369, "step": 1232 }, { "epoch": 0.38, "learning_rate": 1.42913120009413e-05, "loss": 1.2485, "step": 1233 }, { "epoch": 0.38, "learning_rate": 1.4282307272705555e-05, "loss": 1.8576, "step": 1234 }, { "epoch": 0.38, "learning_rate": 1.4273298290395043e-05, "loss": 1.5776, "step": 1235 }, { "epoch": 0.38, "learning_rate": 1.4264285062959358e-05, "loss": 1.5215, "step": 1236 }, { "epoch": 0.38, "learning_rate": 1.4255267599352301e-05, "loss": 1.4871, "step": 1237 }, { "epoch": 0.38, "learning_rate": 1.4246245908531883e-05, "loss": 1.4785, "step": 1238 }, { "epoch": 0.38, "learning_rate": 1.4237219999460323e-05, "loss": 1.5851, "step": 1239 }, { "epoch": 0.38, "learning_rate": 1.4228189881104016e-05, "loss": 1.5691, "step": 1240 }, { "epoch": 0.38, "learning_rate": 1.4219155562433551e-05, "loss": 1.4772, "step": 1241 }, { "epoch": 0.38, "learning_rate": 1.4210117052423683e-05, "loss": 1.6606, "step": 1242 }, { "epoch": 0.38, "learning_rate": 1.4201074360053332e-05, "loss": 1.7508, "step": 1243 }, { "epoch": 0.38, "learning_rate": 1.4192027494305578e-05, "loss": 1.7585, "step": 1244 }, { "epoch": 0.38, "learning_rate": 1.4182976464167637e-05, "loss": 1.712, "step": 1245 }, { "epoch": 0.38, "learning_rate": 1.4173921278630874e-05, "loss": 1.4973, "step": 1246 }, { "epoch": 0.38, "learning_rate": 1.4164861946690765e-05, "loss": 1.4005, "step": 1247 }, { "epoch": 0.38, "learning_rate": 1.4155798477346922e-05, "loss": 1.4932, "step": 1248 }, { "epoch": 0.38, "learning_rate": 1.4146730879603063e-05, "loss": 1.4863, "step": 1249 }, { "epoch": 0.38, "learning_rate": 1.4137659162466999e-05, "loss": 1.6575, "step": 1250 }, { "epoch": 0.38, "learning_rate": 1.4128583334950638e-05, "loss": 1.2822, "step": 1251 }, { "epoch": 0.39, "learning_rate": 1.411950340606998e-05, "loss": 1.8933, "step": 1252 }, { "epoch": 0.39, "learning_rate": 1.4110419384845086e-05, "loss": 1.678, "step": 1253 }, { "epoch": 0.39, "learning_rate": 1.410133128030009e-05, "loss": 1.7881, "step": 1254 }, { "epoch": 0.39, "learning_rate": 1.4092239101463174e-05, "loss": 1.5135, "step": 1255 }, { "epoch": 0.39, "learning_rate": 1.408314285736658e-05, "loss": 1.4438, "step": 1256 }, { "epoch": 0.39, "learning_rate": 1.4074042557046584e-05, "loss": 1.8626, "step": 1257 }, { "epoch": 0.39, "learning_rate": 1.4064938209543483e-05, "loss": 1.4916, "step": 1258 }, { "epoch": 0.39, "learning_rate": 1.4055829823901604e-05, "loss": 1.3214, "step": 1259 }, { "epoch": 0.39, "learning_rate": 1.4046717409169288e-05, "loss": 1.202, "step": 1260 }, { "epoch": 0.39, "learning_rate": 1.4037600974398863e-05, "loss": 1.3464, "step": 1261 }, { "epoch": 0.39, "learning_rate": 1.4028480528646669e-05, "loss": 1.493, "step": 1262 }, { "epoch": 0.39, "learning_rate": 1.4019356080973017e-05, "loss": 1.3624, "step": 1263 }, { "epoch": 0.39, "learning_rate": 1.4010227640442202e-05, "loss": 1.7877, "step": 1264 }, { "epoch": 0.39, "learning_rate": 1.400109521612248e-05, "loss": 1.2741, "step": 1265 }, { "epoch": 0.39, "learning_rate": 1.3991958817086068e-05, "loss": 1.6242, "step": 1266 }, { "epoch": 0.39, "learning_rate": 1.3982818452409132e-05, "loss": 1.5145, "step": 1267 }, { "epoch": 0.39, "learning_rate": 1.3973674131171775e-05, "loss": 1.5785, "step": 1268 }, { "epoch": 0.39, "learning_rate": 1.3964525862458029e-05, "loss": 1.5778, "step": 1269 }, { "epoch": 0.39, "learning_rate": 1.3955373655355852e-05, "loss": 1.1616, "step": 1270 }, { "epoch": 0.39, "learning_rate": 1.394621751895711e-05, "loss": 1.96, "step": 1271 }, { "epoch": 0.39, "learning_rate": 1.3937057462357578e-05, "loss": 1.4356, "step": 1272 }, { "epoch": 0.39, "learning_rate": 1.392789349465692e-05, "loss": 1.7084, "step": 1273 }, { "epoch": 0.39, "learning_rate": 1.3918725624958684e-05, "loss": 1.2878, "step": 1274 }, { "epoch": 0.39, "learning_rate": 1.3909553862370305e-05, "loss": 1.3199, "step": 1275 }, { "epoch": 0.39, "learning_rate": 1.3900378216003073e-05, "loss": 1.7056, "step": 1276 }, { "epoch": 0.39, "learning_rate": 1.3891198694972139e-05, "loss": 1.581, "step": 1277 }, { "epoch": 0.39, "learning_rate": 1.3882015308396508e-05, "loss": 1.5768, "step": 1278 }, { "epoch": 0.39, "learning_rate": 1.3872828065399023e-05, "loss": 1.198, "step": 1279 }, { "epoch": 0.39, "learning_rate": 1.3863636975106354e-05, "loss": 1.7849, "step": 1280 }, { "epoch": 0.39, "learning_rate": 1.3854442046649002e-05, "loss": 1.2813, "step": 1281 }, { "epoch": 0.39, "learning_rate": 1.3845243289161263e-05, "loss": 1.2747, "step": 1282 }, { "epoch": 0.39, "learning_rate": 1.3836040711781263e-05, "loss": 1.8253, "step": 1283 }, { "epoch": 0.4, "learning_rate": 1.3826834323650899e-05, "loss": 1.4281, "step": 1284 }, { "epoch": 0.4, "learning_rate": 1.3817624133915865e-05, "loss": 1.3276, "step": 1285 }, { "epoch": 0.4, "learning_rate": 1.3808410151725633e-05, "loss": 1.4582, "step": 1286 }, { "epoch": 0.4, "learning_rate": 1.3799192386233434e-05, "loss": 1.592, "step": 1287 }, { "epoch": 0.4, "learning_rate": 1.3789970846596268e-05, "loss": 1.2669, "step": 1288 }, { "epoch": 0.4, "learning_rate": 1.3780745541974871e-05, "loss": 1.3246, "step": 1289 }, { "epoch": 0.4, "learning_rate": 1.3771516481533733e-05, "loss": 1.9186, "step": 1290 }, { "epoch": 0.4, "learning_rate": 1.3762283674441066e-05, "loss": 1.5179, "step": 1291 }, { "epoch": 0.4, "learning_rate": 1.3753047129868808e-05, "loss": 1.4538, "step": 1292 }, { "epoch": 0.4, "learning_rate": 1.374380685699261e-05, "loss": 1.4632, "step": 1293 }, { "epoch": 0.4, "learning_rate": 1.3734562864991823e-05, "loss": 1.7178, "step": 1294 }, { "epoch": 0.4, "learning_rate": 1.3725315163049495e-05, "loss": 1.8489, "step": 1295 }, { "epoch": 0.4, "learning_rate": 1.3716063760352359e-05, "loss": 1.4246, "step": 1296 }, { "epoch": 0.4, "learning_rate": 1.3706808666090827e-05, "loss": 1.6744, "step": 1297 }, { "epoch": 0.4, "learning_rate": 1.3697549889458974e-05, "loss": 1.5797, "step": 1298 }, { "epoch": 0.4, "learning_rate": 1.3688287439654537e-05, "loss": 1.4504, "step": 1299 }, { "epoch": 0.4, "learning_rate": 1.3679021325878901e-05, "loss": 1.4808, "step": 1300 }, { "epoch": 0.4, "learning_rate": 1.3669751557337085e-05, "loss": 1.5905, "step": 1301 }, { "epoch": 0.4, "learning_rate": 1.3660478143237748e-05, "loss": 1.5264, "step": 1302 }, { "epoch": 0.4, "learning_rate": 1.3651201092793164e-05, "loss": 1.5836, "step": 1303 }, { "epoch": 0.4, "learning_rate": 1.3641920415219224e-05, "loss": 1.5105, "step": 1304 }, { "epoch": 0.4, "learning_rate": 1.3632636119735416e-05, "loss": 1.2502, "step": 1305 }, { "epoch": 0.4, "learning_rate": 1.3623348215564827e-05, "loss": 1.6512, "step": 1306 }, { "epoch": 0.4, "learning_rate": 1.361405671193413e-05, "loss": 1.4239, "step": 1307 }, { "epoch": 0.4, "learning_rate": 1.3604761618073567e-05, "loss": 1.3912, "step": 1308 }, { "epoch": 0.4, "learning_rate": 1.3595462943216959e-05, "loss": 1.613, "step": 1309 }, { "epoch": 0.4, "learning_rate": 1.3586160696601667e-05, "loss": 1.3493, "step": 1310 }, { "epoch": 0.4, "learning_rate": 1.3576854887468614e-05, "loss": 2.049, "step": 1311 }, { "epoch": 0.4, "learning_rate": 1.3567545525062256e-05, "loss": 1.7991, "step": 1312 }, { "epoch": 0.4, "learning_rate": 1.3558232618630582e-05, "loss": 1.1772, "step": 1313 }, { "epoch": 0.4, "learning_rate": 1.3548916177425095e-05, "loss": 1.5329, "step": 1314 }, { "epoch": 0.4, "learning_rate": 1.3539596210700825e-05, "loss": 1.4228, "step": 1315 }, { "epoch": 0.4, "learning_rate": 1.3530272727716286e-05, "loss": 1.2907, "step": 1316 }, { "epoch": 0.41, "learning_rate": 1.3520945737733496e-05, "loss": 1.4106, "step": 1317 }, { "epoch": 0.41, "learning_rate": 1.3511615250017948e-05, "loss": 1.6078, "step": 1318 }, { "epoch": 0.41, "learning_rate": 1.3502281273838622e-05, "loss": 1.1787, "step": 1319 }, { "epoch": 0.41, "learning_rate": 1.3492943818467956e-05, "loss": 1.8198, "step": 1320 }, { "epoch": 0.41, "learning_rate": 1.3483602893181841e-05, "loss": 1.4558, "step": 1321 }, { "epoch": 0.41, "learning_rate": 1.3474258507259623e-05, "loss": 1.4652, "step": 1322 }, { "epoch": 0.41, "learning_rate": 1.3464910669984084e-05, "loss": 1.9823, "step": 1323 }, { "epoch": 0.41, "learning_rate": 1.3455559390641426e-05, "loss": 1.1311, "step": 1324 }, { "epoch": 0.41, "learning_rate": 1.3446204678521281e-05, "loss": 1.453, "step": 1325 }, { "epoch": 0.41, "learning_rate": 1.3436846542916686e-05, "loss": 1.4574, "step": 1326 }, { "epoch": 0.41, "learning_rate": 1.3427484993124082e-05, "loss": 1.4486, "step": 1327 }, { "epoch": 0.41, "learning_rate": 1.34181200384433e-05, "loss": 1.3743, "step": 1328 }, { "epoch": 0.41, "learning_rate": 1.3408751688177551e-05, "loss": 1.6277, "step": 1329 }, { "epoch": 0.41, "learning_rate": 1.339937995163342e-05, "loss": 1.2064, "step": 1330 }, { "epoch": 0.41, "learning_rate": 1.3390004838120861e-05, "loss": 1.8015, "step": 1331 }, { "epoch": 0.41, "learning_rate": 1.3380626356953175e-05, "loss": 1.6031, "step": 1332 }, { "epoch": 0.41, "learning_rate": 1.3371244517447017e-05, "loss": 1.5908, "step": 1333 }, { "epoch": 0.41, "learning_rate": 1.3361859328922368e-05, "loss": 1.4192, "step": 1334 }, { "epoch": 0.41, "learning_rate": 1.3352470800702548e-05, "loss": 2.5847, "step": 1335 }, { "epoch": 0.41, "learning_rate": 1.3343078942114185e-05, "loss": 1.5099, "step": 1336 }, { "epoch": 0.41, "learning_rate": 1.333368376248722e-05, "loss": 1.6909, "step": 1337 }, { "epoch": 0.41, "learning_rate": 1.3324285271154889e-05, "loss": 1.8992, "step": 1338 }, { "epoch": 0.41, "learning_rate": 1.3314883477453725e-05, "loss": 1.4883, "step": 1339 }, { "epoch": 0.41, "learning_rate": 1.3305478390723532e-05, "loss": 1.5553, "step": 1340 }, { "epoch": 0.41, "learning_rate": 1.3296070020307394e-05, "loss": 1.5106, "step": 1341 }, { "epoch": 0.41, "learning_rate": 1.3286658375551654e-05, "loss": 1.7311, "step": 1342 }, { "epoch": 0.41, "learning_rate": 1.3277243465805905e-05, "loss": 1.3575, "step": 1343 }, { "epoch": 0.41, "learning_rate": 1.3267825300422991e-05, "loss": 1.3254, "step": 1344 }, { "epoch": 0.41, "learning_rate": 1.3258403888758978e-05, "loss": 1.3226, "step": 1345 }, { "epoch": 0.41, "learning_rate": 1.3248979240173167e-05, "loss": 1.4751, "step": 1346 }, { "epoch": 0.41, "learning_rate": 1.3239551364028076e-05, "loss": 1.4943, "step": 1347 }, { "epoch": 0.41, "learning_rate": 1.3230120269689417e-05, "loss": 1.3, "step": 1348 }, { "epoch": 0.42, "learning_rate": 1.3220685966526112e-05, "loss": 1.4543, "step": 1349 }, { "epoch": 0.42, "learning_rate": 1.3211248463910263e-05, "loss": 2.2739, "step": 1350 }, { "epoch": 0.42, "learning_rate": 1.3201807771217156e-05, "loss": 1.7298, "step": 1351 }, { "epoch": 0.42, "learning_rate": 1.319236389782524e-05, "loss": 1.3881, "step": 1352 }, { "epoch": 0.42, "learning_rate": 1.3182916853116124e-05, "loss": 1.432, "step": 1353 }, { "epoch": 0.42, "learning_rate": 1.3173466646474575e-05, "loss": 1.8247, "step": 1354 }, { "epoch": 0.42, "learning_rate": 1.3164013287288495e-05, "loss": 1.1211, "step": 1355 }, { "epoch": 0.42, "learning_rate": 1.3154556784948918e-05, "loss": 1.3836, "step": 1356 }, { "epoch": 0.42, "learning_rate": 1.3145097148850002e-05, "loss": 1.1664, "step": 1357 }, { "epoch": 0.42, "learning_rate": 1.3135634388389016e-05, "loss": 1.6141, "step": 1358 }, { "epoch": 0.42, "learning_rate": 1.3126168512966333e-05, "loss": 1.841, "step": 1359 }, { "epoch": 0.42, "learning_rate": 1.3116699531985426e-05, "loss": 1.7351, "step": 1360 }, { "epoch": 0.42, "learning_rate": 1.3107227454852841e-05, "loss": 1.7628, "step": 1361 }, { "epoch": 0.42, "learning_rate": 1.3097752290978219e-05, "loss": 1.3661, "step": 1362 }, { "epoch": 0.42, "learning_rate": 1.3088274049774245e-05, "loss": 1.5343, "step": 1363 }, { "epoch": 0.42, "learning_rate": 1.3078792740656679e-05, "loss": 1.6005, "step": 1364 }, { "epoch": 0.42, "learning_rate": 1.3069308373044315e-05, "loss": 1.43, "step": 1365 }, { "epoch": 0.42, "learning_rate": 1.3059820956358998e-05, "loss": 1.4833, "step": 1366 }, { "epoch": 0.42, "learning_rate": 1.3050330500025592e-05, "loss": 1.3415, "step": 1367 }, { "epoch": 0.42, "learning_rate": 1.3040837013471985e-05, "loss": 1.6342, "step": 1368 }, { "epoch": 0.42, "learning_rate": 1.3031340506129074e-05, "loss": 1.481, "step": 1369 }, { "epoch": 0.42, "learning_rate": 1.3021840987430761e-05, "loss": 1.4388, "step": 1370 }, { "epoch": 0.42, "learning_rate": 1.301233846681393e-05, "loss": 1.4746, "step": 1371 }, { "epoch": 0.42, "learning_rate": 1.3002832953718462e-05, "loss": 1.4783, "step": 1372 }, { "epoch": 0.42, "learning_rate": 1.2993324457587194e-05, "loss": 1.3375, "step": 1373 }, { "epoch": 0.42, "learning_rate": 1.2983812987865936e-05, "loss": 1.4069, "step": 1374 }, { "epoch": 0.42, "learning_rate": 1.2974298554003451e-05, "loss": 1.8389, "step": 1375 }, { "epoch": 0.42, "learning_rate": 1.2964781165451448e-05, "loss": 1.0991, "step": 1376 }, { "epoch": 0.42, "learning_rate": 1.2955260831664564e-05, "loss": 1.3493, "step": 1377 }, { "epoch": 0.42, "learning_rate": 1.2945737562100373e-05, "loss": 2.1661, "step": 1378 }, { "epoch": 0.42, "learning_rate": 1.2936211366219359e-05, "loss": 1.268, "step": 1379 }, { "epoch": 0.42, "learning_rate": 1.2926682253484906e-05, "loss": 1.2769, "step": 1380 }, { "epoch": 0.42, "learning_rate": 1.2917150233363307e-05, "loss": 1.7214, "step": 1381 }, { "epoch": 0.43, "learning_rate": 1.290761531532374e-05, "loss": 1.44, "step": 1382 }, { "epoch": 0.43, "learning_rate": 1.2898077508838263e-05, "loss": 1.596, "step": 1383 }, { "epoch": 0.43, "learning_rate": 1.2888536823381799e-05, "loss": 1.4986, "step": 1384 }, { "epoch": 0.43, "learning_rate": 1.287899326843213e-05, "loss": 1.4707, "step": 1385 }, { "epoch": 0.43, "learning_rate": 1.2869446853469898e-05, "loss": 1.6921, "step": 1386 }, { "epoch": 0.43, "learning_rate": 1.285989758797858e-05, "loss": 1.3802, "step": 1387 }, { "epoch": 0.43, "learning_rate": 1.285034548144448e-05, "loss": 1.5127, "step": 1388 }, { "epoch": 0.43, "learning_rate": 1.284079054335673e-05, "loss": 1.6045, "step": 1389 }, { "epoch": 0.43, "learning_rate": 1.2831232783207278e-05, "loss": 1.35, "step": 1390 }, { "epoch": 0.43, "learning_rate": 1.2821672210490865e-05, "loss": 1.4313, "step": 1391 }, { "epoch": 0.43, "learning_rate": 1.281210883470504e-05, "loss": 1.369, "step": 1392 }, { "epoch": 0.43, "learning_rate": 1.2802542665350124e-05, "loss": 1.2239, "step": 1393 }, { "epoch": 0.43, "learning_rate": 1.2792973711929222e-05, "loss": 1.5115, "step": 1394 }, { "epoch": 0.43, "learning_rate": 1.2783401983948198e-05, "loss": 1.5038, "step": 1395 }, { "epoch": 0.43, "learning_rate": 1.2773827490915678e-05, "loss": 1.3529, "step": 1396 }, { "epoch": 0.43, "learning_rate": 1.2764250242343025e-05, "loss": 1.3804, "step": 1397 }, { "epoch": 0.43, "learning_rate": 1.2754670247744353e-05, "loss": 1.35, "step": 1398 }, { "epoch": 0.43, "learning_rate": 1.2745087516636501e-05, "loss": 1.4144, "step": 1399 }, { "epoch": 0.43, "learning_rate": 1.2735502058539014e-05, "loss": 1.6021, "step": 1400 }, { "epoch": 0.43, "learning_rate": 1.272591388297416e-05, "loss": 1.1659, "step": 1401 }, { "epoch": 0.43, "learning_rate": 1.27163229994669e-05, "loss": 1.5967, "step": 1402 }, { "epoch": 0.43, "learning_rate": 1.2706729417544885e-05, "loss": 1.4683, "step": 1403 }, { "epoch": 0.43, "learning_rate": 1.2697133146738452e-05, "loss": 2.2997, "step": 1404 }, { "epoch": 0.43, "learning_rate": 1.2687534196580597e-05, "loss": 1.4074, "step": 1405 }, { "epoch": 0.43, "learning_rate": 1.2677932576606998e-05, "loss": 1.7564, "step": 1406 }, { "epoch": 0.43, "learning_rate": 1.2668328296355964e-05, "loss": 1.4242, "step": 1407 }, { "epoch": 0.43, "learning_rate": 1.2658721365368462e-05, "loss": 1.2521, "step": 1408 }, { "epoch": 0.43, "learning_rate": 1.264911179318808e-05, "loss": 1.7296, "step": 1409 }, { "epoch": 0.43, "learning_rate": 1.2639499589361041e-05, "loss": 1.6688, "step": 1410 }, { "epoch": 0.43, "learning_rate": 1.2629884763436176e-05, "loss": 1.8353, "step": 1411 }, { "epoch": 0.43, "learning_rate": 1.2620267324964918e-05, "loss": 1.4325, "step": 1412 }, { "epoch": 0.43, "learning_rate": 1.2610647283501303e-05, "loss": 1.4561, "step": 1413 }, { "epoch": 0.44, "learning_rate": 1.260102464860195e-05, "loss": 1.9443, "step": 1414 }, { "epoch": 0.44, "learning_rate": 1.2591399429826054e-05, "loss": 2.1286, "step": 1415 }, { "epoch": 0.44, "learning_rate": 1.258177163673537e-05, "loss": 1.1668, "step": 1416 }, { "epoch": 0.44, "learning_rate": 1.2572141278894226e-05, "loss": 1.3574, "step": 1417 }, { "epoch": 0.44, "learning_rate": 1.2562508365869484e-05, "loss": 1.3362, "step": 1418 }, { "epoch": 0.44, "learning_rate": 1.2552872907230547e-05, "loss": 1.8117, "step": 1419 }, { "epoch": 0.44, "learning_rate": 1.2543234912549352e-05, "loss": 1.7966, "step": 1420 }, { "epoch": 0.44, "learning_rate": 1.2533594391400347e-05, "loss": 1.3343, "step": 1421 }, { "epoch": 0.44, "learning_rate": 1.2523951353360504e-05, "loss": 1.7886, "step": 1422 }, { "epoch": 0.44, "learning_rate": 1.2514305808009279e-05, "loss": 1.3105, "step": 1423 }, { "epoch": 0.44, "learning_rate": 1.2504657764928626e-05, "loss": 1.3934, "step": 1424 }, { "epoch": 0.44, "learning_rate": 1.2495007233702982e-05, "loss": 1.7103, "step": 1425 }, { "epoch": 0.44, "learning_rate": 1.2485354223919255e-05, "loss": 1.5023, "step": 1426 }, { "epoch": 0.44, "learning_rate": 1.2475698745166813e-05, "loss": 1.4954, "step": 1427 }, { "epoch": 0.44, "learning_rate": 1.2466040807037478e-05, "loss": 1.4154, "step": 1428 }, { "epoch": 0.44, "learning_rate": 1.2456380419125516e-05, "loss": 1.617, "step": 1429 }, { "epoch": 0.44, "learning_rate": 1.2446717591027624e-05, "loss": 1.8304, "step": 1430 }, { "epoch": 0.44, "learning_rate": 1.2437052332342923e-05, "loss": 1.5548, "step": 1431 }, { "epoch": 0.44, "learning_rate": 1.2427384652672954e-05, "loss": 1.391, "step": 1432 }, { "epoch": 0.44, "learning_rate": 1.2417714561621656e-05, "loss": 1.6611, "step": 1433 }, { "epoch": 0.44, "learning_rate": 1.2408042068795373e-05, "loss": 1.4653, "step": 1434 }, { "epoch": 0.44, "learning_rate": 1.2398367183802822e-05, "loss": 1.233, "step": 1435 }, { "epoch": 0.44, "learning_rate": 1.2388689916255106e-05, "loss": 1.2863, "step": 1436 }, { "epoch": 0.44, "learning_rate": 1.2379010275765693e-05, "loss": 1.5886, "step": 1437 }, { "epoch": 0.44, "learning_rate": 1.2369328271950404e-05, "loss": 1.5959, "step": 1438 }, { "epoch": 0.44, "learning_rate": 1.2359643914427415e-05, "loss": 1.3312, "step": 1439 }, { "epoch": 0.44, "learning_rate": 1.2349957212817229e-05, "loss": 1.7169, "step": 1440 }, { "epoch": 0.44, "learning_rate": 1.2340268176742696e-05, "loss": 1.6262, "step": 1441 }, { "epoch": 0.44, "learning_rate": 1.2330576815828964e-05, "loss": 1.2585, "step": 1442 }, { "epoch": 0.44, "learning_rate": 1.2320883139703512e-05, "loss": 2.1559, "step": 1443 }, { "epoch": 0.44, "learning_rate": 1.2311187157996091e-05, "loss": 1.6384, "step": 1444 }, { "epoch": 0.44, "learning_rate": 1.2301488880338772e-05, "loss": 1.5301, "step": 1445 }, { "epoch": 0.44, "learning_rate": 1.229178831636589e-05, "loss": 1.4363, "step": 1446 }, { "epoch": 0.45, "learning_rate": 1.2282085475714052e-05, "loss": 1.5623, "step": 1447 }, { "epoch": 0.45, "learning_rate": 1.2272380368022133e-05, "loss": 1.4032, "step": 1448 }, { "epoch": 0.45, "learning_rate": 1.2262673002931254e-05, "loss": 1.4183, "step": 1449 }, { "epoch": 0.45, "learning_rate": 1.2252963390084784e-05, "loss": 1.5666, "step": 1450 }, { "epoch": 0.45, "learning_rate": 1.224325153912832e-05, "loss": 1.9528, "step": 1451 }, { "epoch": 0.45, "learning_rate": 1.2233537459709683e-05, "loss": 1.4767, "step": 1452 }, { "epoch": 0.45, "learning_rate": 1.2223821161478912e-05, "loss": 1.7994, "step": 1453 }, { "epoch": 0.45, "learning_rate": 1.2214102654088248e-05, "loss": 1.4598, "step": 1454 }, { "epoch": 0.45, "learning_rate": 1.2204381947192123e-05, "loss": 1.3264, "step": 1455 }, { "epoch": 0.45, "learning_rate": 1.2194659050447158e-05, "loss": 1.4388, "step": 1456 }, { "epoch": 0.45, "learning_rate": 1.2184933973512152e-05, "loss": 1.3978, "step": 1457 }, { "epoch": 0.45, "learning_rate": 1.2175206726048058e-05, "loss": 2.2029, "step": 1458 }, { "epoch": 0.45, "learning_rate": 1.2165477317718002e-05, "loss": 1.8447, "step": 1459 }, { "epoch": 0.45, "learning_rate": 1.215574575818724e-05, "loss": 1.3005, "step": 1460 }, { "epoch": 0.45, "learning_rate": 1.214601205712318e-05, "loss": 1.9965, "step": 1461 }, { "epoch": 0.45, "learning_rate": 1.2136276224195349e-05, "loss": 1.6336, "step": 1462 }, { "epoch": 0.45, "learning_rate": 1.2126538269075392e-05, "loss": 1.3078, "step": 1463 }, { "epoch": 0.45, "learning_rate": 1.2116798201437061e-05, "loss": 1.3533, "step": 1464 }, { "epoch": 0.45, "learning_rate": 1.2107056030956214e-05, "loss": 1.2577, "step": 1465 }, { "epoch": 0.45, "learning_rate": 1.2097311767310791e-05, "loss": 1.4717, "step": 1466 }, { "epoch": 0.45, "learning_rate": 1.208756542018081e-05, "loss": 1.5975, "step": 1467 }, { "epoch": 0.45, "learning_rate": 1.2077816999248366e-05, "loss": 1.3437, "step": 1468 }, { "epoch": 0.45, "learning_rate": 1.2068066514197611e-05, "loss": 1.4219, "step": 1469 }, { "epoch": 0.45, "learning_rate": 1.2058313974714746e-05, "loss": 1.3091, "step": 1470 }, { "epoch": 0.45, "learning_rate": 1.2048559390488013e-05, "loss": 1.5736, "step": 1471 }, { "epoch": 0.45, "learning_rate": 1.2038802771207686e-05, "loss": 1.2382, "step": 1472 }, { "epoch": 0.45, "learning_rate": 1.2029044126566058e-05, "loss": 1.3758, "step": 1473 }, { "epoch": 0.45, "learning_rate": 1.2019283466257441e-05, "loss": 1.4917, "step": 1474 }, { "epoch": 0.45, "learning_rate": 1.2009520799978143e-05, "loss": 1.4151, "step": 1475 }, { "epoch": 0.45, "learning_rate": 1.1999756137426465e-05, "loss": 1.6299, "step": 1476 }, { "epoch": 0.45, "learning_rate": 1.1989989488302695e-05, "loss": 1.4298, "step": 1477 }, { "epoch": 0.45, "learning_rate": 1.1980220862309097e-05, "loss": 1.5317, "step": 1478 }, { "epoch": 0.46, "learning_rate": 1.1970450269149882e-05, "loss": 1.9429, "step": 1479 }, { "epoch": 0.46, "learning_rate": 1.196067771853124e-05, "loss": 1.5051, "step": 1480 }, { "epoch": 0.46, "learning_rate": 1.1950903220161286e-05, "loss": 1.6433, "step": 1481 }, { "epoch": 0.46, "learning_rate": 1.1941126783750075e-05, "loss": 1.5357, "step": 1482 }, { "epoch": 0.46, "learning_rate": 1.1931348419009594e-05, "loss": 1.9704, "step": 1483 }, { "epoch": 0.46, "learning_rate": 1.1921568135653738e-05, "loss": 1.6505, "step": 1484 }, { "epoch": 0.46, "learning_rate": 1.1911785943398316e-05, "loss": 1.4378, "step": 1485 }, { "epoch": 0.46, "learning_rate": 1.1902001851961019e-05, "loss": 1.7249, "step": 1486 }, { "epoch": 0.46, "learning_rate": 1.1892215871061436e-05, "loss": 1.2966, "step": 1487 }, { "epoch": 0.46, "learning_rate": 1.1882428010421036e-05, "loss": 1.5901, "step": 1488 }, { "epoch": 0.46, "learning_rate": 1.1872638279763147e-05, "loss": 1.7083, "step": 1489 }, { "epoch": 0.46, "learning_rate": 1.1862846688812956e-05, "loss": 1.2796, "step": 1490 }, { "epoch": 0.46, "learning_rate": 1.1853053247297503e-05, "loss": 1.4548, "step": 1491 }, { "epoch": 0.46, "learning_rate": 1.184325796494566e-05, "loss": 1.6324, "step": 1492 }, { "epoch": 0.46, "learning_rate": 1.1833460851488132e-05, "loss": 2.0364, "step": 1493 }, { "epoch": 0.46, "learning_rate": 1.1823661916657441e-05, "loss": 1.5063, "step": 1494 }, { "epoch": 0.46, "learning_rate": 1.1813861170187919e-05, "loss": 2.2661, "step": 1495 }, { "epoch": 0.46, "learning_rate": 1.1804058621815699e-05, "loss": 1.6608, "step": 1496 }, { "epoch": 0.46, "learning_rate": 1.1794254281278704e-05, "loss": 1.8123, "step": 1497 }, { "epoch": 0.46, "learning_rate": 1.1784448158316633e-05, "loss": 1.5114, "step": 1498 }, { "epoch": 0.46, "learning_rate": 1.1774640262670962e-05, "loss": 1.5714, "step": 1499 }, { "epoch": 0.46, "learning_rate": 1.1764830604084923e-05, "loss": 1.2827, "step": 1500 }, { "epoch": 0.46, "learning_rate": 1.1755019192303503e-05, "loss": 1.3728, "step": 1501 }, { "epoch": 0.46, "learning_rate": 1.1745206037073426e-05, "loss": 1.451, "step": 1502 }, { "epoch": 0.46, "learning_rate": 1.1735391148143149e-05, "loss": 1.4311, "step": 1503 }, { "epoch": 0.46, "learning_rate": 1.1725574535262863e-05, "loss": 1.4488, "step": 1504 }, { "epoch": 0.46, "learning_rate": 1.171575620818445e-05, "loss": 1.577, "step": 1505 }, { "epoch": 0.46, "learning_rate": 1.1705936176661516e-05, "loss": 1.4405, "step": 1506 }, { "epoch": 0.46, "learning_rate": 1.1696114450449343e-05, "loss": 1.5442, "step": 1507 }, { "epoch": 0.46, "learning_rate": 1.168629103930491e-05, "loss": 1.5151, "step": 1508 }, { "epoch": 0.46, "learning_rate": 1.167646595298686e-05, "loss": 1.4906, "step": 1509 }, { "epoch": 0.46, "learning_rate": 1.1666639201255507e-05, "loss": 1.2289, "step": 1510 }, { "epoch": 0.46, "learning_rate": 1.1656810793872813e-05, "loss": 1.7102, "step": 1511 }, { "epoch": 0.47, "learning_rate": 1.1646980740602394e-05, "loss": 1.4379, "step": 1512 }, { "epoch": 0.47, "learning_rate": 1.1637149051209496e-05, "loss": 1.2132, "step": 1513 }, { "epoch": 0.47, "learning_rate": 1.162731573546098e-05, "loss": 1.4962, "step": 1514 }, { "epoch": 0.47, "learning_rate": 1.1617480803125338e-05, "loss": 1.503, "step": 1515 }, { "epoch": 0.47, "learning_rate": 1.1607644263972669e-05, "loss": 1.4258, "step": 1516 }, { "epoch": 0.47, "learning_rate": 1.1597806127774651e-05, "loss": 1.5131, "step": 1517 }, { "epoch": 0.47, "learning_rate": 1.1587966404304565e-05, "loss": 1.5341, "step": 1518 }, { "epoch": 0.47, "learning_rate": 1.157812510333726e-05, "loss": 1.3106, "step": 1519 }, { "epoch": 0.47, "learning_rate": 1.156828223464916e-05, "loss": 1.3401, "step": 1520 }, { "epoch": 0.47, "learning_rate": 1.1558437808018238e-05, "loss": 1.4217, "step": 1521 }, { "epoch": 0.47, "learning_rate": 1.1548591833224015e-05, "loss": 1.4464, "step": 1522 }, { "epoch": 0.47, "learning_rate": 1.1538744320047554e-05, "loss": 1.4141, "step": 1523 }, { "epoch": 0.47, "learning_rate": 1.152889527827145e-05, "loss": 1.3032, "step": 1524 }, { "epoch": 0.47, "learning_rate": 1.151904471767981e-05, "loss": 2.1309, "step": 1525 }, { "epoch": 0.47, "learning_rate": 1.150919264805825e-05, "loss": 1.8071, "step": 1526 }, { "epoch": 0.47, "learning_rate": 1.1499339079193886e-05, "loss": 1.4872, "step": 1527 }, { "epoch": 0.47, "learning_rate": 1.1489484020875325e-05, "loss": 1.7761, "step": 1528 }, { "epoch": 0.47, "learning_rate": 1.1479627482892655e-05, "loss": 1.1814, "step": 1529 }, { "epoch": 0.47, "learning_rate": 1.1469769475037427e-05, "loss": 1.8832, "step": 1530 }, { "epoch": 0.47, "learning_rate": 1.1459910007102657e-05, "loss": 1.4938, "step": 1531 }, { "epoch": 0.47, "learning_rate": 1.1450049088882816e-05, "loss": 1.6288, "step": 1532 }, { "epoch": 0.47, "learning_rate": 1.1440186730173808e-05, "loss": 1.5206, "step": 1533 }, { "epoch": 0.47, "learning_rate": 1.143032294077297e-05, "loss": 1.2526, "step": 1534 }, { "epoch": 0.47, "learning_rate": 1.1420457730479061e-05, "loss": 1.7994, "step": 1535 }, { "epoch": 0.47, "learning_rate": 1.1410591109092257e-05, "loss": 1.6462, "step": 1536 }, { "epoch": 0.47, "learning_rate": 1.1400723086414122e-05, "loss": 1.6014, "step": 1537 }, { "epoch": 0.47, "learning_rate": 1.1390853672247629e-05, "loss": 1.1764, "step": 1538 }, { "epoch": 0.47, "learning_rate": 1.1380982876397115e-05, "loss": 1.8971, "step": 1539 }, { "epoch": 0.47, "learning_rate": 1.1371110708668307e-05, "loss": 1.4256, "step": 1540 }, { "epoch": 0.47, "learning_rate": 1.1361237178868288e-05, "loss": 1.5668, "step": 1541 }, { "epoch": 0.47, "learning_rate": 1.1351362296805487e-05, "loss": 1.4795, "step": 1542 }, { "epoch": 0.47, "learning_rate": 1.1341486072289686e-05, "loss": 1.5851, "step": 1543 }, { "epoch": 0.48, "learning_rate": 1.1331608515132001e-05, "loss": 1.4929, "step": 1544 }, { "epoch": 0.48, "learning_rate": 1.1321729635144862e-05, "loss": 1.5407, "step": 1545 }, { "epoch": 0.48, "learning_rate": 1.1311849442142025e-05, "loss": 1.6635, "step": 1546 }, { "epoch": 0.48, "learning_rate": 1.1301967945938538e-05, "loss": 1.6153, "step": 1547 }, { "epoch": 0.48, "learning_rate": 1.1292085156350756e-05, "loss": 1.4714, "step": 1548 }, { "epoch": 0.48, "learning_rate": 1.128220108319631e-05, "loss": 1.4397, "step": 1549 }, { "epoch": 0.48, "learning_rate": 1.1272315736294108e-05, "loss": 1.4616, "step": 1550 }, { "epoch": 0.48, "learning_rate": 1.1262429125464328e-05, "loss": 1.5294, "step": 1551 }, { "epoch": 0.48, "learning_rate": 1.1252541260528398e-05, "loss": 1.8233, "step": 1552 }, { "epoch": 0.48, "learning_rate": 1.1242652151308997e-05, "loss": 1.5038, "step": 1553 }, { "epoch": 0.48, "learning_rate": 1.1232761807630032e-05, "loss": 1.4738, "step": 1554 }, { "epoch": 0.48, "learning_rate": 1.1222870239316645e-05, "loss": 1.3698, "step": 1555 }, { "epoch": 0.48, "learning_rate": 1.121297745619519e-05, "loss": 1.5445, "step": 1556 }, { "epoch": 0.48, "learning_rate": 1.120308346809323e-05, "loss": 1.667, "step": 1557 }, { "epoch": 0.48, "learning_rate": 1.1193188284839518e-05, "loss": 1.5548, "step": 1558 }, { "epoch": 0.48, "learning_rate": 1.1183291916264007e-05, "loss": 1.4341, "step": 1559 }, { "epoch": 0.48, "learning_rate": 1.117339437219782e-05, "loss": 1.3633, "step": 1560 }, { "epoch": 0.48, "learning_rate": 1.1163495662473242e-05, "loss": 1.254, "step": 1561 }, { "epoch": 0.48, "learning_rate": 1.115359579692373e-05, "loss": 2.2963, "step": 1562 }, { "epoch": 0.48, "learning_rate": 1.1143694785383877e-05, "loss": 1.367, "step": 1563 }, { "epoch": 0.48, "learning_rate": 1.113379263768942e-05, "loss": 1.5674, "step": 1564 }, { "epoch": 0.48, "learning_rate": 1.1123889363677223e-05, "loss": 1.5677, "step": 1565 }, { "epoch": 0.48, "learning_rate": 1.1113984973185268e-05, "loss": 1.4457, "step": 1566 }, { "epoch": 0.48, "learning_rate": 1.1104079476052649e-05, "loss": 1.3479, "step": 1567 }, { "epoch": 0.48, "learning_rate": 1.1094172882119562e-05, "loss": 1.5166, "step": 1568 }, { "epoch": 0.48, "learning_rate": 1.1084265201227284e-05, "loss": 1.3685, "step": 1569 }, { "epoch": 0.48, "learning_rate": 1.1074356443218175e-05, "loss": 1.5713, "step": 1570 }, { "epoch": 0.48, "learning_rate": 1.106444661793567e-05, "loss": 1.5827, "step": 1571 }, { "epoch": 0.48, "learning_rate": 1.1054535735224257e-05, "loss": 1.9172, "step": 1572 }, { "epoch": 0.48, "learning_rate": 1.1044623804929481e-05, "loss": 1.4493, "step": 1573 }, { "epoch": 0.48, "learning_rate": 1.1034710836897922e-05, "loss": 1.3679, "step": 1574 }, { "epoch": 0.48, "learning_rate": 1.1024796840977196e-05, "loss": 2.3108, "step": 1575 }, { "epoch": 0.48, "learning_rate": 1.1014881827015939e-05, "loss": 1.7419, "step": 1576 }, { "epoch": 0.49, "learning_rate": 1.1004965804863795e-05, "loss": 1.3969, "step": 1577 }, { "epoch": 0.49, "learning_rate": 1.0995048784371405e-05, "loss": 1.3786, "step": 1578 }, { "epoch": 0.49, "learning_rate": 1.098513077539042e-05, "loss": 1.8025, "step": 1579 }, { "epoch": 0.49, "learning_rate": 1.0975211787773455e-05, "loss": 1.0797, "step": 1580 }, { "epoch": 0.49, "learning_rate": 1.0965291831374104e-05, "loss": 1.4982, "step": 1581 }, { "epoch": 0.49, "learning_rate": 1.095537091604692e-05, "loss": 1.3096, "step": 1582 }, { "epoch": 0.49, "learning_rate": 1.0945449051647417e-05, "loss": 1.2556, "step": 1583 }, { "epoch": 0.49, "learning_rate": 1.0935526248032047e-05, "loss": 1.4918, "step": 1584 }, { "epoch": 0.49, "learning_rate": 1.0925602515058187e-05, "loss": 1.6028, "step": 1585 }, { "epoch": 0.49, "learning_rate": 1.0915677862584149e-05, "loss": 1.4136, "step": 1586 }, { "epoch": 0.49, "learning_rate": 1.0905752300469155e-05, "loss": 1.224, "step": 1587 }, { "epoch": 0.49, "learning_rate": 1.0895825838573326e-05, "loss": 1.4274, "step": 1588 }, { "epoch": 0.49, "learning_rate": 1.0885898486757686e-05, "loss": 1.5501, "step": 1589 }, { "epoch": 0.49, "learning_rate": 1.087597025488413e-05, "loss": 1.5916, "step": 1590 }, { "epoch": 0.49, "learning_rate": 1.0866041152815443e-05, "loss": 1.4807, "step": 1591 }, { "epoch": 0.49, "learning_rate": 1.0856111190415262e-05, "loss": 1.4673, "step": 1592 }, { "epoch": 0.49, "learning_rate": 1.0846180377548082e-05, "loss": 1.4942, "step": 1593 }, { "epoch": 0.49, "learning_rate": 1.0836248724079241e-05, "loss": 1.4996, "step": 1594 }, { "epoch": 0.49, "learning_rate": 1.0826316239874923e-05, "loss": 1.4299, "step": 1595 }, { "epoch": 0.49, "learning_rate": 1.0816382934802123e-05, "loss": 1.815, "step": 1596 }, { "epoch": 0.49, "learning_rate": 1.080644881872866e-05, "loss": 1.5384, "step": 1597 }, { "epoch": 0.49, "learning_rate": 1.0796513901523156e-05, "loss": 1.484, "step": 1598 }, { "epoch": 0.49, "learning_rate": 1.0786578193055029e-05, "loss": 1.4599, "step": 1599 }, { "epoch": 0.49, "learning_rate": 1.077664170319448e-05, "loss": 1.2585, "step": 1600 }, { "epoch": 0.49, "learning_rate": 1.0766704441812493e-05, "loss": 1.4028, "step": 1601 }, { "epoch": 0.49, "learning_rate": 1.0756766418780811e-05, "loss": 0.9765, "step": 1602 }, { "epoch": 0.49, "learning_rate": 1.0746827643971942e-05, "loss": 1.6094, "step": 1603 }, { "epoch": 0.49, "learning_rate": 1.0736888127259132e-05, "loss": 1.36, "step": 1604 }, { "epoch": 0.49, "learning_rate": 1.0726947878516373e-05, "loss": 1.4325, "step": 1605 }, { "epoch": 0.49, "learning_rate": 1.0717006907618377e-05, "loss": 1.6546, "step": 1606 }, { "epoch": 0.49, "learning_rate": 1.0707065224440575e-05, "loss": 1.5306, "step": 1607 }, { "epoch": 0.49, "learning_rate": 1.0697122838859106e-05, "loss": 1.5791, "step": 1608 }, { "epoch": 0.5, "learning_rate": 1.0687179760750808e-05, "loss": 1.9094, "step": 1609 }, { "epoch": 0.5, "learning_rate": 1.0677235999993205e-05, "loss": 1.4735, "step": 1610 }, { "epoch": 0.5, "learning_rate": 1.0667291566464504e-05, "loss": 1.4291, "step": 1611 }, { "epoch": 0.5, "learning_rate": 1.0657346470043575e-05, "loss": 1.892, "step": 1612 }, { "epoch": 0.5, "learning_rate": 1.0647400720609942e-05, "loss": 1.6397, "step": 1613 }, { "epoch": 0.5, "learning_rate": 1.0637454328043792e-05, "loss": 1.6173, "step": 1614 }, { "epoch": 0.5, "learning_rate": 1.0627507302225938e-05, "loss": 1.6717, "step": 1615 }, { "epoch": 0.5, "learning_rate": 1.0617559653037833e-05, "loss": 1.4193, "step": 1616 }, { "epoch": 0.5, "learning_rate": 1.0607611390361534e-05, "loss": 1.254, "step": 1617 }, { "epoch": 0.5, "learning_rate": 1.0597662524079721e-05, "loss": 1.3787, "step": 1618 }, { "epoch": 0.5, "learning_rate": 1.0587713064075673e-05, "loss": 1.7384, "step": 1619 }, { "epoch": 0.5, "learning_rate": 1.057776302023325e-05, "loss": 1.6016, "step": 1620 }, { "epoch": 0.5, "learning_rate": 1.0567812402436895e-05, "loss": 1.396, "step": 1621 }, { "epoch": 0.5, "learning_rate": 1.0557861220571626e-05, "loss": 1.406, "step": 1622 }, { "epoch": 0.5, "learning_rate": 1.0547909484523023e-05, "loss": 1.3675, "step": 1623 }, { "epoch": 0.5, "learning_rate": 1.0537957204177204e-05, "loss": 1.6323, "step": 1624 }, { "epoch": 0.5, "learning_rate": 1.052800438942084e-05, "loss": 1.5804, "step": 1625 }, { "epoch": 0.5, "learning_rate": 1.051805105014113e-05, "loss": 1.7015, "step": 1626 }, { "epoch": 0.5, "learning_rate": 1.0508097196225787e-05, "loss": 1.6947, "step": 1627 }, { "epoch": 0.5, "learning_rate": 1.0498142837563048e-05, "loss": 1.679, "step": 1628 }, { "epoch": 0.5, "learning_rate": 1.0488187984041636e-05, "loss": 1.6727, "step": 1629 }, { "epoch": 0.5, "learning_rate": 1.0478232645550784e-05, "loss": 1.7717, "step": 1630 }, { "epoch": 0.5, "learning_rate": 1.0468276831980186e-05, "loss": 2.1289, "step": 1631 }, { "epoch": 0.5, "learning_rate": 1.045832055322003e-05, "loss": 1.39, "step": 1632 }, { "epoch": 0.5, "learning_rate": 1.0448363819160946e-05, "loss": 1.4587, "step": 1633 }, { "epoch": 0.5, "learning_rate": 1.0438406639694025e-05, "loss": 1.8243, "step": 1634 }, { "epoch": 0.5, "learning_rate": 1.0428449024710805e-05, "loss": 1.3176, "step": 1635 }, { "epoch": 0.5, "learning_rate": 1.041849098410325e-05, "loss": 1.4693, "step": 1636 }, { "epoch": 0.5, "learning_rate": 1.0408532527763749e-05, "loss": 1.3979, "step": 1637 }, { "epoch": 0.5, "learning_rate": 1.0398573665585105e-05, "loss": 1.4386, "step": 1638 }, { "epoch": 0.5, "learning_rate": 1.0388614407460523e-05, "loss": 1.3095, "step": 1639 }, { "epoch": 0.5, "learning_rate": 1.0378654763283602e-05, "loss": 1.2712, "step": 1640 }, { "epoch": 0.5, "learning_rate": 1.036869474294832e-05, "loss": 1.3695, "step": 1641 }, { "epoch": 0.51, "learning_rate": 1.0358734356349035e-05, "loss": 1.5895, "step": 1642 }, { "epoch": 0.51, "learning_rate": 1.034877361338047e-05, "loss": 1.342, "step": 1643 }, { "epoch": 0.51, "learning_rate": 1.0338812523937694e-05, "loss": 1.7784, "step": 1644 }, { "epoch": 0.51, "learning_rate": 1.0328851097916124e-05, "loss": 1.3724, "step": 1645 }, { "epoch": 0.51, "learning_rate": 1.0318889345211519e-05, "loss": 1.4914, "step": 1646 }, { "epoch": 0.51, "learning_rate": 1.0308927275719949e-05, "loss": 1.323, "step": 1647 }, { "epoch": 0.51, "learning_rate": 1.0298964899337807e-05, "loss": 1.6419, "step": 1648 }, { "epoch": 0.51, "learning_rate": 1.0289002225961788e-05, "loss": 1.5935, "step": 1649 }, { "epoch": 0.51, "learning_rate": 1.0279039265488885e-05, "loss": 1.4547, "step": 1650 }, { "epoch": 0.51, "learning_rate": 1.0269076027816376e-05, "loss": 1.3306, "step": 1651 }, { "epoch": 0.51, "learning_rate": 1.0259112522841808e-05, "loss": 1.6222, "step": 1652 }, { "epoch": 0.51, "learning_rate": 1.0249148760463001e-05, "loss": 1.9939, "step": 1653 }, { "epoch": 0.51, "learning_rate": 1.023918475057803e-05, "loss": 1.4909, "step": 1654 }, { "epoch": 0.51, "learning_rate": 1.0229220503085214e-05, "loss": 1.4362, "step": 1655 }, { "epoch": 0.51, "learning_rate": 1.0219256027883102e-05, "loss": 1.4717, "step": 1656 }, { "epoch": 0.51, "learning_rate": 1.0209291334870479e-05, "loss": 1.3928, "step": 1657 }, { "epoch": 0.51, "learning_rate": 1.0199326433946345e-05, "loss": 1.5495, "step": 1658 }, { "epoch": 0.51, "learning_rate": 1.01893613350099e-05, "loss": 1.3461, "step": 1659 }, { "epoch": 0.51, "learning_rate": 1.0179396047960546e-05, "loss": 1.6597, "step": 1660 }, { "epoch": 0.51, "learning_rate": 1.0169430582697873e-05, "loss": 1.2357, "step": 1661 }, { "epoch": 0.51, "learning_rate": 1.0159464949121642e-05, "loss": 1.3447, "step": 1662 }, { "epoch": 0.51, "learning_rate": 1.0149499157131788e-05, "loss": 1.5661, "step": 1663 }, { "epoch": 0.51, "learning_rate": 1.0139533216628399e-05, "loss": 1.3869, "step": 1664 }, { "epoch": 0.51, "learning_rate": 1.012956713751171e-05, "loss": 1.34, "step": 1665 }, { "epoch": 0.51, "learning_rate": 1.0119600929682096e-05, "loss": 1.6007, "step": 1666 }, { "epoch": 0.51, "learning_rate": 1.0109634603040062e-05, "loss": 1.5608, "step": 1667 }, { "epoch": 0.51, "learning_rate": 1.0099668167486227e-05, "loss": 1.0849, "step": 1668 }, { "epoch": 0.51, "learning_rate": 1.0089701632921321e-05, "loss": 1.2905, "step": 1669 }, { "epoch": 0.51, "learning_rate": 1.0079735009246168e-05, "loss": 1.5287, "step": 1670 }, { "epoch": 0.51, "learning_rate": 1.0069768306361685e-05, "loss": 1.7288, "step": 1671 }, { "epoch": 0.51, "learning_rate": 1.0059801534168868e-05, "loss": 1.4597, "step": 1672 }, { "epoch": 0.51, "learning_rate": 1.0049834702568772e-05, "loss": 1.6025, "step": 1673 }, { "epoch": 0.52, "learning_rate": 1.0039867821462529e-05, "loss": 1.2934, "step": 1674 }, { "epoch": 0.52, "learning_rate": 1.0029900900751308e-05, "loss": 1.0656, "step": 1675 }, { "epoch": 0.52, "learning_rate": 1.0019933950336314e-05, "loss": 1.4362, "step": 1676 }, { "epoch": 0.52, "learning_rate": 1.0009966980118792e-05, "loss": 1.5876, "step": 1677 }, { "epoch": 0.52, "learning_rate": 1e-05, "loss": 1.4996, "step": 1678 }, { "epoch": 0.52, "learning_rate": 9.99003301988121e-06, "loss": 1.608, "step": 1679 }, { "epoch": 0.52, "learning_rate": 9.98006604966369e-06, "loss": 1.1425, "step": 1680 }, { "epoch": 0.52, "learning_rate": 9.970099099248694e-06, "loss": 1.4318, "step": 1681 }, { "epoch": 0.52, "learning_rate": 9.960132178537471e-06, "loss": 1.54, "step": 1682 }, { "epoch": 0.52, "learning_rate": 9.95016529743123e-06, "loss": 1.6719, "step": 1683 }, { "epoch": 0.52, "learning_rate": 9.940198465831137e-06, "loss": 1.1922, "step": 1684 }, { "epoch": 0.52, "learning_rate": 9.930231693638316e-06, "loss": 1.6652, "step": 1685 }, { "epoch": 0.52, "learning_rate": 9.920264990753837e-06, "loss": 1.341, "step": 1686 }, { "epoch": 0.52, "learning_rate": 9.910298367078684e-06, "loss": 2.004, "step": 1687 }, { "epoch": 0.52, "learning_rate": 9.900331832513775e-06, "loss": 1.7567, "step": 1688 }, { "epoch": 0.52, "learning_rate": 9.89036539695994e-06, "loss": 1.59, "step": 1689 }, { "epoch": 0.52, "learning_rate": 9.880399070317907e-06, "loss": 1.3059, "step": 1690 }, { "epoch": 0.52, "learning_rate": 9.870432862488295e-06, "loss": 1.317, "step": 1691 }, { "epoch": 0.52, "learning_rate": 9.860466783371605e-06, "loss": 1.228, "step": 1692 }, { "epoch": 0.52, "learning_rate": 9.850500842868215e-06, "loss": 1.2998, "step": 1693 }, { "epoch": 0.52, "learning_rate": 9.84053505087836e-06, "loss": 1.2889, "step": 1694 }, { "epoch": 0.52, "learning_rate": 9.83056941730213e-06, "loss": 1.5742, "step": 1695 }, { "epoch": 0.52, "learning_rate": 9.820603952039456e-06, "loss": 1.8903, "step": 1696 }, { "epoch": 0.52, "learning_rate": 9.810638664990101e-06, "loss": 1.5518, "step": 1697 }, { "epoch": 0.52, "learning_rate": 9.800673566053657e-06, "loss": 1.8494, "step": 1698 }, { "epoch": 0.52, "learning_rate": 9.790708665129521e-06, "loss": 1.511, "step": 1699 }, { "epoch": 0.52, "learning_rate": 9.780743972116903e-06, "loss": 1.6458, "step": 1700 }, { "epoch": 0.52, "learning_rate": 9.770779496914791e-06, "loss": 1.8114, "step": 1701 }, { "epoch": 0.52, "learning_rate": 9.760815249421973e-06, "loss": 1.2366, "step": 1702 }, { "epoch": 0.52, "learning_rate": 9.750851239537002e-06, "loss": 1.3515, "step": 1703 }, { "epoch": 0.52, "learning_rate": 9.740887477158194e-06, "loss": 1.5392, "step": 1704 }, { "epoch": 0.52, "learning_rate": 9.730923972183629e-06, "loss": 1.1979, "step": 1705 }, { "epoch": 0.52, "learning_rate": 9.720960734511118e-06, "loss": 1.6623, "step": 1706 }, { "epoch": 0.53, "learning_rate": 9.710997774038215e-06, "loss": 1.6344, "step": 1707 }, { "epoch": 0.53, "learning_rate": 9.701035100662195e-06, "loss": 1.4079, "step": 1708 }, { "epoch": 0.53, "learning_rate": 9.691072724280053e-06, "loss": 1.396, "step": 1709 }, { "epoch": 0.53, "learning_rate": 9.681110654788483e-06, "loss": 1.3516, "step": 1710 }, { "epoch": 0.53, "learning_rate": 9.671148902083876e-06, "loss": 1.6512, "step": 1711 }, { "epoch": 0.53, "learning_rate": 9.661187476062308e-06, "loss": 1.9343, "step": 1712 }, { "epoch": 0.53, "learning_rate": 9.65122638661953e-06, "loss": 1.5428, "step": 1713 }, { "epoch": 0.53, "learning_rate": 9.641265643650968e-06, "loss": 1.5049, "step": 1714 }, { "epoch": 0.53, "learning_rate": 9.631305257051683e-06, "loss": 1.0889, "step": 1715 }, { "epoch": 0.53, "learning_rate": 9.621345236716403e-06, "loss": 1.412, "step": 1716 }, { "epoch": 0.53, "learning_rate": 9.61138559253948e-06, "loss": 1.1771, "step": 1717 }, { "epoch": 0.53, "learning_rate": 9.601426334414898e-06, "loss": 1.3467, "step": 1718 }, { "epoch": 0.53, "learning_rate": 9.591467472236255e-06, "loss": 1.3944, "step": 1719 }, { "epoch": 0.53, "learning_rate": 9.581509015896751e-06, "loss": 1.5119, "step": 1720 }, { "epoch": 0.53, "learning_rate": 9.571550975289197e-06, "loss": 1.4827, "step": 1721 }, { "epoch": 0.53, "learning_rate": 9.561593360305977e-06, "loss": 1.427, "step": 1722 }, { "epoch": 0.53, "learning_rate": 9.55163618083906e-06, "loss": 1.3517, "step": 1723 }, { "epoch": 0.53, "learning_rate": 9.541679446779972e-06, "loss": 1.3801, "step": 1724 }, { "epoch": 0.53, "learning_rate": 9.531723168019814e-06, "loss": 1.3607, "step": 1725 }, { "epoch": 0.53, "learning_rate": 9.521767354449218e-06, "loss": 1.1911, "step": 1726 }, { "epoch": 0.53, "learning_rate": 9.511812015958364e-06, "loss": 1.1936, "step": 1727 }, { "epoch": 0.53, "learning_rate": 9.501857162436957e-06, "loss": 1.7146, "step": 1728 }, { "epoch": 0.53, "learning_rate": 9.491902803774216e-06, "loss": 1.4931, "step": 1729 }, { "epoch": 0.53, "learning_rate": 9.481948949858876e-06, "loss": 1.5108, "step": 1730 }, { "epoch": 0.53, "learning_rate": 9.471995610579161e-06, "loss": 1.4073, "step": 1731 }, { "epoch": 0.53, "learning_rate": 9.462042795822799e-06, "loss": 1.4956, "step": 1732 }, { "epoch": 0.53, "learning_rate": 9.452090515476982e-06, "loss": 1.5541, "step": 1733 }, { "epoch": 0.53, "learning_rate": 9.442138779428376e-06, "loss": 1.7594, "step": 1734 }, { "epoch": 0.53, "learning_rate": 9.43218759756311e-06, "loss": 1.2847, "step": 1735 }, { "epoch": 0.53, "learning_rate": 9.422236979766755e-06, "loss": 1.4494, "step": 1736 }, { "epoch": 0.53, "learning_rate": 9.41228693592433e-06, "loss": 1.4843, "step": 1737 }, { "epoch": 0.53, "learning_rate": 9.40233747592028e-06, "loss": 1.2377, "step": 1738 }, { "epoch": 0.54, "learning_rate": 9.39238860963847e-06, "loss": 1.7476, "step": 1739 }, { "epoch": 0.54, "learning_rate": 9.382440346962169e-06, "loss": 1.0276, "step": 1740 }, { "epoch": 0.54, "learning_rate": 9.372492697774062e-06, "loss": 1.4952, "step": 1741 }, { "epoch": 0.54, "learning_rate": 9.362545671956213e-06, "loss": 2.1685, "step": 1742 }, { "epoch": 0.54, "learning_rate": 9.352599279390062e-06, "loss": 1.3387, "step": 1743 }, { "epoch": 0.54, "learning_rate": 9.342653529956432e-06, "loss": 1.3846, "step": 1744 }, { "epoch": 0.54, "learning_rate": 9.3327084335355e-06, "loss": 1.4131, "step": 1745 }, { "epoch": 0.54, "learning_rate": 9.322764000006798e-06, "loss": 2.6065, "step": 1746 }, { "epoch": 0.54, "learning_rate": 9.312820239249194e-06, "loss": 1.3596, "step": 1747 }, { "epoch": 0.54, "learning_rate": 9.302877161140897e-06, "loss": 1.381, "step": 1748 }, { "epoch": 0.54, "learning_rate": 9.292934775559428e-06, "loss": 1.4341, "step": 1749 }, { "epoch": 0.54, "learning_rate": 9.282993092381626e-06, "loss": 1.3982, "step": 1750 }, { "epoch": 0.54, "learning_rate": 9.273052121483628e-06, "loss": 1.4022, "step": 1751 }, { "epoch": 0.54, "learning_rate": 9.263111872740866e-06, "loss": 1.4594, "step": 1752 }, { "epoch": 0.54, "learning_rate": 9.253172356028058e-06, "loss": 1.641, "step": 1753 }, { "epoch": 0.54, "learning_rate": 9.24323358121919e-06, "loss": 1.6621, "step": 1754 }, { "epoch": 0.54, "learning_rate": 9.23329555818751e-06, "loss": 1.4911, "step": 1755 }, { "epoch": 0.54, "learning_rate": 9.223358296805526e-06, "loss": 1.1983, "step": 1756 }, { "epoch": 0.54, "learning_rate": 9.213421806944978e-06, "loss": 1.1931, "step": 1757 }, { "epoch": 0.54, "learning_rate": 9.20348609847685e-06, "loss": 1.5871, "step": 1758 }, { "epoch": 0.54, "learning_rate": 9.193551181271341e-06, "loss": 1.3482, "step": 1759 }, { "epoch": 0.54, "learning_rate": 9.183617065197879e-06, "loss": 1.4159, "step": 1760 }, { "epoch": 0.54, "learning_rate": 9.17368376012508e-06, "loss": 1.627, "step": 1761 }, { "epoch": 0.54, "learning_rate": 9.163751275920762e-06, "loss": 1.6991, "step": 1762 }, { "epoch": 0.54, "learning_rate": 9.153819622451921e-06, "loss": 1.6957, "step": 1763 }, { "epoch": 0.54, "learning_rate": 9.143888809584741e-06, "loss": 1.3581, "step": 1764 }, { "epoch": 0.54, "learning_rate": 9.133958847184559e-06, "loss": 1.8417, "step": 1765 }, { "epoch": 0.54, "learning_rate": 9.124029745115872e-06, "loss": 1.6578, "step": 1766 }, { "epoch": 0.54, "learning_rate": 9.114101513242317e-06, "loss": 1.4522, "step": 1767 }, { "epoch": 0.54, "learning_rate": 9.104174161426674e-06, "loss": 1.6761, "step": 1768 }, { "epoch": 0.54, "learning_rate": 9.094247699530845e-06, "loss": 1.5325, "step": 1769 }, { "epoch": 0.54, "learning_rate": 9.084322137415855e-06, "loss": 1.3307, "step": 1770 }, { "epoch": 0.54, "learning_rate": 9.074397484941816e-06, "loss": 1.45, "step": 1771 }, { "epoch": 0.55, "learning_rate": 9.064473751967958e-06, "loss": 1.976, "step": 1772 }, { "epoch": 0.55, "learning_rate": 9.054550948352587e-06, "loss": 1.5559, "step": 1773 }, { "epoch": 0.55, "learning_rate": 9.044629083953082e-06, "loss": 1.2764, "step": 1774 }, { "epoch": 0.55, "learning_rate": 9.0347081686259e-06, "loss": 1.5768, "step": 1775 }, { "epoch": 0.55, "learning_rate": 9.024788212226549e-06, "loss": 1.3257, "step": 1776 }, { "epoch": 0.55, "learning_rate": 9.014869224609584e-06, "loss": 1.3006, "step": 1777 }, { "epoch": 0.55, "learning_rate": 9.004951215628597e-06, "loss": 1.2039, "step": 1778 }, { "epoch": 0.55, "learning_rate": 8.995034195136209e-06, "loss": 1.5623, "step": 1779 }, { "epoch": 0.55, "learning_rate": 8.985118172984063e-06, "loss": 1.5361, "step": 1780 }, { "epoch": 0.55, "learning_rate": 8.975203159022804e-06, "loss": 1.2436, "step": 1781 }, { "epoch": 0.55, "learning_rate": 8.96528916310208e-06, "loss": 1.5689, "step": 1782 }, { "epoch": 0.55, "learning_rate": 8.95537619507052e-06, "loss": 1.7321, "step": 1783 }, { "epoch": 0.55, "learning_rate": 8.945464264775748e-06, "loss": 1.8205, "step": 1784 }, { "epoch": 0.55, "learning_rate": 8.935553382064336e-06, "loss": 1.574, "step": 1785 }, { "epoch": 0.55, "learning_rate": 8.925643556781828e-06, "loss": 1.5428, "step": 1786 }, { "epoch": 0.55, "learning_rate": 8.915734798772721e-06, "loss": 2.2525, "step": 1787 }, { "epoch": 0.55, "learning_rate": 8.905827117880441e-06, "loss": 1.4458, "step": 1788 }, { "epoch": 0.55, "learning_rate": 8.895920523947353e-06, "loss": 1.4314, "step": 1789 }, { "epoch": 0.55, "learning_rate": 8.886015026814736e-06, "loss": 1.3173, "step": 1790 }, { "epoch": 0.55, "learning_rate": 8.87611063632278e-06, "loss": 1.3583, "step": 1791 }, { "epoch": 0.55, "learning_rate": 8.866207362310583e-06, "loss": 1.4705, "step": 1792 }, { "epoch": 0.55, "learning_rate": 8.856305214616125e-06, "loss": 1.2249, "step": 1793 }, { "epoch": 0.55, "learning_rate": 8.846404203076273e-06, "loss": 1.7282, "step": 1794 }, { "epoch": 0.55, "learning_rate": 8.836504337526758e-06, "loss": 1.3321, "step": 1795 }, { "epoch": 0.55, "learning_rate": 8.826605627802182e-06, "loss": 1.4131, "step": 1796 }, { "epoch": 0.55, "learning_rate": 8.816708083735993e-06, "loss": 1.4614, "step": 1797 }, { "epoch": 0.55, "learning_rate": 8.806811715160485e-06, "loss": 0.9892, "step": 1798 }, { "epoch": 0.55, "learning_rate": 8.796916531906776e-06, "loss": 1.2521, "step": 1799 }, { "epoch": 0.55, "learning_rate": 8.787022543804814e-06, "loss": 1.1811, "step": 1800 }, { "epoch": 0.55, "learning_rate": 8.77712976068336e-06, "loss": 1.4258, "step": 1801 }, { "epoch": 0.55, "learning_rate": 8.76723819236997e-06, "loss": 1.515, "step": 1802 }, { "epoch": 0.55, "learning_rate": 8.757347848691006e-06, "loss": 1.9053, "step": 1803 }, { "epoch": 0.56, "learning_rate": 8.747458739471603e-06, "loss": 1.2192, "step": 1804 }, { "epoch": 0.56, "learning_rate": 8.737570874535674e-06, "loss": 1.4117, "step": 1805 }, { "epoch": 0.56, "learning_rate": 8.727684263705896e-06, "loss": 1.6294, "step": 1806 }, { "epoch": 0.56, "learning_rate": 8.717798916803692e-06, "loss": 1.4099, "step": 1807 }, { "epoch": 0.56, "learning_rate": 8.707914843649247e-06, "loss": 1.6191, "step": 1808 }, { "epoch": 0.56, "learning_rate": 8.698032054061464e-06, "loss": 1.2496, "step": 1809 }, { "epoch": 0.56, "learning_rate": 8.688150557857979e-06, "loss": 1.3684, "step": 1810 }, { "epoch": 0.56, "learning_rate": 8.678270364855137e-06, "loss": 1.6995, "step": 1811 }, { "epoch": 0.56, "learning_rate": 8.668391484868004e-06, "loss": 1.8171, "step": 1812 }, { "epoch": 0.56, "learning_rate": 8.658513927710317e-06, "loss": 1.2929, "step": 1813 }, { "epoch": 0.56, "learning_rate": 8.648637703194515e-06, "loss": 1.7156, "step": 1814 }, { "epoch": 0.56, "learning_rate": 8.638762821131717e-06, "loss": 1.4532, "step": 1815 }, { "epoch": 0.56, "learning_rate": 8.628889291331696e-06, "loss": 1.4685, "step": 1816 }, { "epoch": 0.56, "learning_rate": 8.619017123602888e-06, "loss": 1.1877, "step": 1817 }, { "epoch": 0.56, "learning_rate": 8.609146327752376e-06, "loss": 1.4762, "step": 1818 }, { "epoch": 0.56, "learning_rate": 8.59927691358588e-06, "loss": 1.3398, "step": 1819 }, { "epoch": 0.56, "learning_rate": 8.589408890907746e-06, "loss": 1.5266, "step": 1820 }, { "epoch": 0.56, "learning_rate": 8.57954226952094e-06, "loss": 1.57, "step": 1821 }, { "epoch": 0.56, "learning_rate": 8.569677059227033e-06, "loss": 1.5417, "step": 1822 }, { "epoch": 0.56, "learning_rate": 8.559813269826194e-06, "loss": 1.5402, "step": 1823 }, { "epoch": 0.56, "learning_rate": 8.549950911117186e-06, "loss": 1.4313, "step": 1824 }, { "epoch": 0.56, "learning_rate": 8.540089992897343e-06, "loss": 1.3216, "step": 1825 }, { "epoch": 0.56, "learning_rate": 8.530230524962578e-06, "loss": 1.2815, "step": 1826 }, { "epoch": 0.56, "learning_rate": 8.520372517107352e-06, "loss": 1.4336, "step": 1827 }, { "epoch": 0.56, "learning_rate": 8.51051597912468e-06, "loss": 1.4191, "step": 1828 }, { "epoch": 0.56, "learning_rate": 8.500660920806119e-06, "loss": 1.3227, "step": 1829 }, { "epoch": 0.56, "learning_rate": 8.490807351941753e-06, "loss": 1.6915, "step": 1830 }, { "epoch": 0.56, "learning_rate": 8.480955282320192e-06, "loss": 1.6988, "step": 1831 }, { "epoch": 0.56, "learning_rate": 8.471104721728551e-06, "loss": 1.4046, "step": 1832 }, { "epoch": 0.56, "learning_rate": 8.461255679952448e-06, "loss": 1.2138, "step": 1833 }, { "epoch": 0.56, "learning_rate": 8.451408166775987e-06, "loss": 1.5134, "step": 1834 }, { "epoch": 0.56, "learning_rate": 8.441562191981766e-06, "loss": 1.2223, "step": 1835 }, { "epoch": 0.56, "learning_rate": 8.431717765350842e-06, "loss": 1.573, "step": 1836 }, { "epoch": 0.57, "learning_rate": 8.42187489666274e-06, "loss": 1.9228, "step": 1837 }, { "epoch": 0.57, "learning_rate": 8.412033595695436e-06, "loss": 1.3358, "step": 1838 }, { "epoch": 0.57, "learning_rate": 8.40219387222535e-06, "loss": 1.2591, "step": 1839 }, { "epoch": 0.57, "learning_rate": 8.392355736027338e-06, "loss": 1.372, "step": 1840 }, { "epoch": 0.57, "learning_rate": 8.382519196874664e-06, "loss": 1.5483, "step": 1841 }, { "epoch": 0.57, "learning_rate": 8.372684264539025e-06, "loss": 1.2932, "step": 1842 }, { "epoch": 0.57, "learning_rate": 8.36285094879051e-06, "loss": 1.6244, "step": 1843 }, { "epoch": 0.57, "learning_rate": 8.353019259397609e-06, "loss": 1.7016, "step": 1844 }, { "epoch": 0.57, "learning_rate": 8.343189206127189e-06, "loss": 1.5351, "step": 1845 }, { "epoch": 0.57, "learning_rate": 8.333360798744496e-06, "loss": 1.7508, "step": 1846 }, { "epoch": 0.57, "learning_rate": 8.323534047013141e-06, "loss": 1.2294, "step": 1847 }, { "epoch": 0.57, "learning_rate": 8.313708960695094e-06, "loss": 1.3116, "step": 1848 }, { "epoch": 0.57, "learning_rate": 8.30388554955066e-06, "loss": 1.6935, "step": 1849 }, { "epoch": 0.57, "learning_rate": 8.294063823338486e-06, "loss": 1.533, "step": 1850 }, { "epoch": 0.57, "learning_rate": 8.284243791815549e-06, "loss": 1.6415, "step": 1851 }, { "epoch": 0.57, "learning_rate": 8.274425464737139e-06, "loss": 1.1968, "step": 1852 }, { "epoch": 0.57, "learning_rate": 8.26460885185685e-06, "loss": 1.3317, "step": 1853 }, { "epoch": 0.57, "learning_rate": 8.254793962926579e-06, "loss": 1.4101, "step": 1854 }, { "epoch": 0.57, "learning_rate": 8.244980807696504e-06, "loss": 1.6742, "step": 1855 }, { "epoch": 0.57, "learning_rate": 8.23516939591508e-06, "loss": 1.0346, "step": 1856 }, { "epoch": 0.57, "learning_rate": 8.22535973732904e-06, "loss": 1.6658, "step": 1857 }, { "epoch": 0.57, "learning_rate": 8.215551841683369e-06, "loss": 1.5758, "step": 1858 }, { "epoch": 0.57, "learning_rate": 8.2057457187213e-06, "loss": 1.5242, "step": 1859 }, { "epoch": 0.57, "learning_rate": 8.195941378184303e-06, "loss": 1.87, "step": 1860 }, { "epoch": 0.57, "learning_rate": 8.186138829812085e-06, "loss": 1.1414, "step": 1861 }, { "epoch": 0.57, "learning_rate": 8.17633808334256e-06, "loss": 1.1334, "step": 1862 }, { "epoch": 0.57, "learning_rate": 8.166539148511871e-06, "loss": 1.8316, "step": 1863 }, { "epoch": 0.57, "learning_rate": 8.156742035054343e-06, "loss": 1.6207, "step": 1864 }, { "epoch": 0.57, "learning_rate": 8.1469467527025e-06, "loss": 1.2682, "step": 1865 }, { "epoch": 0.57, "learning_rate": 8.137153311187044e-06, "loss": 1.7005, "step": 1866 }, { "epoch": 0.57, "learning_rate": 8.127361720236854e-06, "loss": 1.6663, "step": 1867 }, { "epoch": 0.57, "learning_rate": 8.117571989578969e-06, "loss": 1.264, "step": 1868 }, { "epoch": 0.58, "learning_rate": 8.107784128938566e-06, "loss": 1.9644, "step": 1869 }, { "epoch": 0.58, "learning_rate": 8.097998148038986e-06, "loss": 1.2632, "step": 1870 }, { "epoch": 0.58, "learning_rate": 8.08821405660169e-06, "loss": 1.5973, "step": 1871 }, { "epoch": 0.58, "learning_rate": 8.078431864346265e-06, "loss": 1.5157, "step": 1872 }, { "epoch": 0.58, "learning_rate": 8.068651580990407e-06, "loss": 1.4383, "step": 1873 }, { "epoch": 0.58, "learning_rate": 8.058873216249927e-06, "loss": 1.3048, "step": 1874 }, { "epoch": 0.58, "learning_rate": 8.04909677983872e-06, "loss": 1.7001, "step": 1875 }, { "epoch": 0.58, "learning_rate": 8.039322281468765e-06, "loss": 1.6673, "step": 1876 }, { "epoch": 0.58, "learning_rate": 8.02954973085012e-06, "loss": 1.4748, "step": 1877 }, { "epoch": 0.58, "learning_rate": 8.019779137690906e-06, "loss": 1.6206, "step": 1878 }, { "epoch": 0.58, "learning_rate": 8.010010511697303e-06, "loss": 1.5354, "step": 1879 }, { "epoch": 0.58, "learning_rate": 8.000243862573536e-06, "loss": 1.4034, "step": 1880 }, { "epoch": 0.58, "learning_rate": 7.990479200021858e-06, "loss": 1.5352, "step": 1881 }, { "epoch": 0.58, "learning_rate": 7.980716533742558e-06, "loss": 1.5687, "step": 1882 }, { "epoch": 0.58, "learning_rate": 7.970955873433947e-06, "loss": 1.3477, "step": 1883 }, { "epoch": 0.58, "learning_rate": 7.96119722879232e-06, "loss": 1.1569, "step": 1884 }, { "epoch": 0.58, "learning_rate": 7.95144060951199e-06, "loss": 1.6488, "step": 1885 }, { "epoch": 0.58, "learning_rate": 7.941686025285257e-06, "loss": 1.3721, "step": 1886 }, { "epoch": 0.58, "learning_rate": 7.93193348580239e-06, "loss": 1.387, "step": 1887 }, { "epoch": 0.58, "learning_rate": 7.922183000751635e-06, "loss": 1.423, "step": 1888 }, { "epoch": 0.58, "learning_rate": 7.912434579819191e-06, "loss": 1.6236, "step": 1889 }, { "epoch": 0.58, "learning_rate": 7.902688232689212e-06, "loss": 1.1951, "step": 1890 }, { "epoch": 0.58, "learning_rate": 7.892943969043787e-06, "loss": 1.5847, "step": 1891 }, { "epoch": 0.58, "learning_rate": 7.88320179856294e-06, "loss": 1.4674, "step": 1892 }, { "epoch": 0.58, "learning_rate": 7.873461730924612e-06, "loss": 1.2043, "step": 1893 }, { "epoch": 0.58, "learning_rate": 7.863723775804651e-06, "loss": 1.506, "step": 1894 }, { "epoch": 0.58, "learning_rate": 7.85398794287682e-06, "loss": 1.4507, "step": 1895 }, { "epoch": 0.58, "learning_rate": 7.84425424181276e-06, "loss": 1.8841, "step": 1896 }, { "epoch": 0.58, "learning_rate": 7.834522682282003e-06, "loss": 1.422, "step": 1897 }, { "epoch": 0.58, "learning_rate": 7.824793273951945e-06, "loss": 1.138, "step": 1898 }, { "epoch": 0.58, "learning_rate": 7.815066026487855e-06, "loss": 1.423, "step": 1899 }, { "epoch": 0.58, "learning_rate": 7.805340949552846e-06, "loss": 1.2234, "step": 1900 }, { "epoch": 0.58, "learning_rate": 7.795618052807879e-06, "loss": 1.4055, "step": 1901 }, { "epoch": 0.59, "learning_rate": 7.785897345911755e-06, "loss": 1.5733, "step": 1902 }, { "epoch": 0.59, "learning_rate": 7.77617883852109e-06, "loss": 1.7592, "step": 1903 }, { "epoch": 0.59, "learning_rate": 7.76646254029032e-06, "loss": 1.4252, "step": 1904 }, { "epoch": 0.59, "learning_rate": 7.756748460871681e-06, "loss": 1.2516, "step": 1905 }, { "epoch": 0.59, "learning_rate": 7.747036609915217e-06, "loss": 1.6656, "step": 1906 }, { "epoch": 0.59, "learning_rate": 7.737326997068748e-06, "loss": 1.4984, "step": 1907 }, { "epoch": 0.59, "learning_rate": 7.727619631977869e-06, "loss": 1.5011, "step": 1908 }, { "epoch": 0.59, "learning_rate": 7.71791452428595e-06, "loss": 1.5291, "step": 1909 }, { "epoch": 0.59, "learning_rate": 7.708211683634112e-06, "loss": 1.1616, "step": 1910 }, { "epoch": 0.59, "learning_rate": 7.698511119661233e-06, "loss": 1.3613, "step": 1911 }, { "epoch": 0.59, "learning_rate": 7.68881284200391e-06, "loss": 1.1526, "step": 1912 }, { "epoch": 0.59, "learning_rate": 7.679116860296493e-06, "loss": 2.1535, "step": 1913 }, { "epoch": 0.59, "learning_rate": 7.669423184171037e-06, "loss": 1.5564, "step": 1914 }, { "epoch": 0.59, "learning_rate": 7.659731823257309e-06, "loss": 1.3777, "step": 1915 }, { "epoch": 0.59, "learning_rate": 7.650042787182773e-06, "loss": 1.7495, "step": 1916 }, { "epoch": 0.59, "learning_rate": 7.64035608557259e-06, "loss": 1.4611, "step": 1917 }, { "epoch": 0.59, "learning_rate": 7.6306717280496e-06, "loss": 2.4728, "step": 1918 }, { "epoch": 0.59, "learning_rate": 7.620989724234311e-06, "loss": 1.6802, "step": 1919 }, { "epoch": 0.59, "learning_rate": 7.611310083744897e-06, "loss": 1.1031, "step": 1920 }, { "epoch": 0.59, "learning_rate": 7.601632816197179e-06, "loss": 1.3549, "step": 1921 }, { "epoch": 0.59, "learning_rate": 7.591957931204629e-06, "loss": 1.6642, "step": 1922 }, { "epoch": 0.59, "learning_rate": 7.582285438378344e-06, "loss": 1.2271, "step": 1923 }, { "epoch": 0.59, "learning_rate": 7.572615347327048e-06, "loss": 1.5193, "step": 1924 }, { "epoch": 0.59, "learning_rate": 7.562947667657081e-06, "loss": 1.4017, "step": 1925 }, { "epoch": 0.59, "learning_rate": 7.553282408972382e-06, "loss": 1.5452, "step": 1926 }, { "epoch": 0.59, "learning_rate": 7.543619580874489e-06, "loss": 1.5713, "step": 1927 }, { "epoch": 0.59, "learning_rate": 7.533959192962525e-06, "loss": 1.5599, "step": 1928 }, { "epoch": 0.59, "learning_rate": 7.52430125483319e-06, "loss": 1.2469, "step": 1929 }, { "epoch": 0.59, "learning_rate": 7.514645776080747e-06, "loss": 1.4363, "step": 1930 }, { "epoch": 0.59, "learning_rate": 7.50499276629702e-06, "loss": 1.2099, "step": 1931 }, { "epoch": 0.59, "learning_rate": 7.495342235071377e-06, "loss": 1.7733, "step": 1932 }, { "epoch": 0.59, "learning_rate": 7.485694191990724e-06, "loss": 1.4292, "step": 1933 }, { "epoch": 0.6, "learning_rate": 7.476048646639498e-06, "loss": 1.6801, "step": 1934 }, { "epoch": 0.6, "learning_rate": 7.466405608599653e-06, "loss": 1.5798, "step": 1935 }, { "epoch": 0.6, "learning_rate": 7.456765087450652e-06, "loss": 1.2582, "step": 1936 }, { "epoch": 0.6, "learning_rate": 7.447127092769453e-06, "loss": 1.2624, "step": 1937 }, { "epoch": 0.6, "learning_rate": 7.437491634130518e-06, "loss": 1.563, "step": 1938 }, { "epoch": 0.6, "learning_rate": 7.4278587211057785e-06, "loss": 1.1712, "step": 1939 }, { "epoch": 0.6, "learning_rate": 7.418228363264631e-06, "loss": 1.3578, "step": 1940 }, { "epoch": 0.6, "learning_rate": 7.408600570173951e-06, "loss": 1.3691, "step": 1941 }, { "epoch": 0.6, "learning_rate": 7.398975351398054e-06, "loss": 1.6906, "step": 1942 }, { "epoch": 0.6, "learning_rate": 7.389352716498701e-06, "loss": 2.3196, "step": 1943 }, { "epoch": 0.6, "learning_rate": 7.379732675035085e-06, "loss": 1.7796, "step": 1944 }, { "epoch": 0.6, "learning_rate": 7.370115236563828e-06, "loss": 1.4547, "step": 1945 }, { "epoch": 0.6, "learning_rate": 7.360500410638962e-06, "loss": 1.5591, "step": 1946 }, { "epoch": 0.6, "learning_rate": 7.350888206811923e-06, "loss": 1.4449, "step": 1947 }, { "epoch": 0.6, "learning_rate": 7.3412786346315415e-06, "loss": 1.3225, "step": 1948 }, { "epoch": 0.6, "learning_rate": 7.3316717036440356e-06, "loss": 0.8577, "step": 1949 }, { "epoch": 0.6, "learning_rate": 7.322067423393002e-06, "loss": 1.4257, "step": 1950 }, { "epoch": 0.6, "learning_rate": 7.312465803419402e-06, "loss": 1.7571, "step": 1951 }, { "epoch": 0.6, "learning_rate": 7.302866853261552e-06, "loss": 1.6365, "step": 1952 }, { "epoch": 0.6, "learning_rate": 7.29327058245512e-06, "loss": 1.3949, "step": 1953 }, { "epoch": 0.6, "learning_rate": 7.283677000533106e-06, "loss": 1.2676, "step": 1954 }, { "epoch": 0.6, "learning_rate": 7.2740861170258445e-06, "loss": 1.2759, "step": 1955 }, { "epoch": 0.6, "learning_rate": 7.264497941460988e-06, "loss": 1.559, "step": 1956 }, { "epoch": 0.6, "learning_rate": 7.254912483363503e-06, "loss": 1.0111, "step": 1957 }, { "epoch": 0.6, "learning_rate": 7.2453297522556475e-06, "loss": 1.2686, "step": 1958 }, { "epoch": 0.6, "learning_rate": 7.235749757656978e-06, "loss": 1.3747, "step": 1959 }, { "epoch": 0.6, "learning_rate": 7.226172509084326e-06, "loss": 1.0332, "step": 1960 }, { "epoch": 0.6, "learning_rate": 7.216598016051804e-06, "loss": 1.1655, "step": 1961 }, { "epoch": 0.6, "learning_rate": 7.20702628807078e-06, "loss": 1.52, "step": 1962 }, { "epoch": 0.6, "learning_rate": 7.197457334649878e-06, "loss": 1.3386, "step": 1963 }, { "epoch": 0.6, "learning_rate": 7.1878911652949626e-06, "loss": 1.2893, "step": 1964 }, { "epoch": 0.6, "learning_rate": 7.178327789509134e-06, "loss": 1.3839, "step": 1965 }, { "epoch": 0.6, "learning_rate": 7.168767216792725e-06, "loss": 1.6415, "step": 1966 }, { "epoch": 0.61, "learning_rate": 7.1592094566432745e-06, "loss": 1.1777, "step": 1967 }, { "epoch": 0.61, "learning_rate": 7.149654518555526e-06, "loss": 1.2382, "step": 1968 }, { "epoch": 0.61, "learning_rate": 7.140102412021426e-06, "loss": 1.3861, "step": 1969 }, { "epoch": 0.61, "learning_rate": 7.130553146530105e-06, "loss": 1.4882, "step": 1970 }, { "epoch": 0.61, "learning_rate": 7.121006731567873e-06, "loss": 1.5152, "step": 1971 }, { "epoch": 0.61, "learning_rate": 7.111463176618205e-06, "loss": 1.5381, "step": 1972 }, { "epoch": 0.61, "learning_rate": 7.101922491161739e-06, "loss": 1.5232, "step": 1973 }, { "epoch": 0.61, "learning_rate": 7.092384684676263e-06, "loss": 1.5662, "step": 1974 }, { "epoch": 0.61, "learning_rate": 7.0828497666366965e-06, "loss": 1.2246, "step": 1975 }, { "epoch": 0.61, "learning_rate": 7.073317746515097e-06, "loss": 1.3451, "step": 1976 }, { "epoch": 0.61, "learning_rate": 7.063788633780646e-06, "loss": 1.4567, "step": 1977 }, { "epoch": 0.61, "learning_rate": 7.054262437899628e-06, "loss": 1.5154, "step": 1978 }, { "epoch": 0.61, "learning_rate": 7.044739168335436e-06, "loss": 1.3573, "step": 1979 }, { "epoch": 0.61, "learning_rate": 7.035218834548554e-06, "loss": 1.2748, "step": 1980 }, { "epoch": 0.61, "learning_rate": 7.025701445996553e-06, "loss": 1.7599, "step": 1981 }, { "epoch": 0.61, "learning_rate": 7.016187012134069e-06, "loss": 1.0552, "step": 1982 }, { "epoch": 0.61, "learning_rate": 7.006675542412811e-06, "loss": 1.2412, "step": 1983 }, { "epoch": 0.61, "learning_rate": 6.9971670462815425e-06, "loss": 1.5353, "step": 1984 }, { "epoch": 0.61, "learning_rate": 6.987661533186072e-06, "loss": 1.7036, "step": 1985 }, { "epoch": 0.61, "learning_rate": 6.978159012569243e-06, "loss": 1.9498, "step": 1986 }, { "epoch": 0.61, "learning_rate": 6.96865949387093e-06, "loss": 1.4701, "step": 1987 }, { "epoch": 0.61, "learning_rate": 6.959162986528019e-06, "loss": 1.6037, "step": 1988 }, { "epoch": 0.61, "learning_rate": 6.949669499974412e-06, "loss": 1.2395, "step": 1989 }, { "epoch": 0.61, "learning_rate": 6.940179043641005e-06, "loss": 1.9126, "step": 1990 }, { "epoch": 0.61, "learning_rate": 6.930691626955688e-06, "loss": 1.4737, "step": 1991 }, { "epoch": 0.61, "learning_rate": 6.921207259343323e-06, "loss": 1.3704, "step": 1992 }, { "epoch": 0.61, "learning_rate": 6.911725950225756e-06, "loss": 1.6618, "step": 1993 }, { "epoch": 0.61, "learning_rate": 6.902247709021783e-06, "loss": 1.3129, "step": 1994 }, { "epoch": 0.61, "learning_rate": 6.89277254514716e-06, "loss": 1.6496, "step": 1995 }, { "epoch": 0.61, "learning_rate": 6.88330046801458e-06, "loss": 1.6016, "step": 1996 }, { "epoch": 0.61, "learning_rate": 6.873831487033671e-06, "loss": 1.5081, "step": 1997 }, { "epoch": 0.61, "learning_rate": 6.86436561161099e-06, "loss": 1.5485, "step": 1998 }, { "epoch": 0.62, "learning_rate": 6.8549028511500016e-06, "loss": 1.3337, "step": 1999 }, { "epoch": 0.62, "learning_rate": 6.845443215051084e-06, "loss": 1.5207, "step": 2000 }, { "epoch": 0.62, "learning_rate": 6.835986712711507e-06, "loss": 1.6363, "step": 2001 }, { "epoch": 0.62, "learning_rate": 6.826533353525427e-06, "loss": 1.5806, "step": 2002 }, { "epoch": 0.62, "learning_rate": 6.81708314688388e-06, "loss": 2.0782, "step": 2003 }, { "epoch": 0.62, "learning_rate": 6.807636102174764e-06, "loss": 1.1988, "step": 2004 }, { "epoch": 0.62, "learning_rate": 6.798192228782847e-06, "loss": 1.31, "step": 2005 }, { "epoch": 0.62, "learning_rate": 6.788751536089739e-06, "loss": 1.7434, "step": 2006 }, { "epoch": 0.62, "learning_rate": 6.77931403347389e-06, "loss": 1.2421, "step": 2007 }, { "epoch": 0.62, "learning_rate": 6.769879730310583e-06, "loss": 1.4289, "step": 2008 }, { "epoch": 0.62, "learning_rate": 6.7604486359719305e-06, "loss": 1.6098, "step": 2009 }, { "epoch": 0.62, "learning_rate": 6.751020759826836e-06, "loss": 1.3215, "step": 2010 }, { "epoch": 0.62, "learning_rate": 6.741596111241026e-06, "loss": 1.2353, "step": 2011 }, { "epoch": 0.62, "learning_rate": 6.7321746995770144e-06, "loss": 1.2478, "step": 2012 }, { "epoch": 0.62, "learning_rate": 6.722756534194098e-06, "loss": 1.6369, "step": 2013 }, { "epoch": 0.62, "learning_rate": 6.713341624448351e-06, "loss": 1.5766, "step": 2014 }, { "epoch": 0.62, "learning_rate": 6.7039299796926086e-06, "loss": 1.3183, "step": 2015 }, { "epoch": 0.62, "learning_rate": 6.694521609276471e-06, "loss": 1.7239, "step": 2016 }, { "epoch": 0.62, "learning_rate": 6.68511652254628e-06, "loss": 1.3676, "step": 2017 }, { "epoch": 0.62, "learning_rate": 6.675714728845114e-06, "loss": 1.4914, "step": 2018 }, { "epoch": 0.62, "learning_rate": 6.666316237512784e-06, "loss": 1.3917, "step": 2019 }, { "epoch": 0.62, "learning_rate": 6.656921057885815e-06, "loss": 1.4206, "step": 2020 }, { "epoch": 0.62, "learning_rate": 6.6475291992974534e-06, "loss": 1.3812, "step": 2021 }, { "epoch": 0.62, "learning_rate": 6.638140671077633e-06, "loss": 1.59, "step": 2022 }, { "epoch": 0.62, "learning_rate": 6.628755482552989e-06, "loss": 1.0776, "step": 2023 }, { "epoch": 0.62, "learning_rate": 6.619373643046829e-06, "loss": 1.4771, "step": 2024 }, { "epoch": 0.62, "learning_rate": 6.609995161879146e-06, "loss": 1.3245, "step": 2025 }, { "epoch": 0.62, "learning_rate": 6.600620048366586e-06, "loss": 0.9801, "step": 2026 }, { "epoch": 0.62, "learning_rate": 6.591248311822454e-06, "loss": 1.7605, "step": 2027 }, { "epoch": 0.62, "learning_rate": 6.581879961556703e-06, "loss": 1.458, "step": 2028 }, { "epoch": 0.62, "learning_rate": 6.572515006875919e-06, "loss": 1.4061, "step": 2029 }, { "epoch": 0.62, "learning_rate": 6.563153457083315e-06, "loss": 1.4054, "step": 2030 }, { "epoch": 0.62, "learning_rate": 6.553795321478719e-06, "loss": 1.5378, "step": 2031 }, { "epoch": 0.63, "learning_rate": 6.544440609358575e-06, "loss": 1.4288, "step": 2032 }, { "epoch": 0.63, "learning_rate": 6.535089330015919e-06, "loss": 1.9646, "step": 2033 }, { "epoch": 0.63, "learning_rate": 6.525741492740378e-06, "loss": 1.3538, "step": 2034 }, { "epoch": 0.63, "learning_rate": 6.5163971068181596e-06, "loss": 1.519, "step": 2035 }, { "epoch": 0.63, "learning_rate": 6.507056181532045e-06, "loss": 1.4911, "step": 2036 }, { "epoch": 0.63, "learning_rate": 6.497718726161382e-06, "loss": 1.3094, "step": 2037 }, { "epoch": 0.63, "learning_rate": 6.488384749982054e-06, "loss": 1.5947, "step": 2038 }, { "epoch": 0.63, "learning_rate": 6.479054262266509e-06, "loss": 1.1979, "step": 2039 }, { "epoch": 0.63, "learning_rate": 6.469727272283717e-06, "loss": 1.1595, "step": 2040 }, { "epoch": 0.63, "learning_rate": 6.460403789299178e-06, "loss": 1.3664, "step": 2041 }, { "epoch": 0.63, "learning_rate": 6.451083822574907e-06, "loss": 1.454, "step": 2042 }, { "epoch": 0.63, "learning_rate": 6.441767381369422e-06, "loss": 1.1066, "step": 2043 }, { "epoch": 0.63, "learning_rate": 6.432454474937747e-06, "loss": 1.2649, "step": 2044 }, { "epoch": 0.63, "learning_rate": 6.42314511253139e-06, "loss": 1.2013, "step": 2045 }, { "epoch": 0.63, "learning_rate": 6.413839303398337e-06, "loss": 1.3546, "step": 2046 }, { "epoch": 0.63, "learning_rate": 6.404537056783042e-06, "loss": 1.7463, "step": 2047 }, { "epoch": 0.63, "learning_rate": 6.395238381926432e-06, "loss": 1.6325, "step": 2048 }, { "epoch": 0.63, "learning_rate": 6.385943288065871e-06, "loss": 1.3127, "step": 2049 }, { "epoch": 0.63, "learning_rate": 6.376651784435174e-06, "loss": 1.1262, "step": 2050 }, { "epoch": 0.63, "learning_rate": 6.367363880264587e-06, "loss": 1.6402, "step": 2051 }, { "epoch": 0.63, "learning_rate": 6.358079584780781e-06, "loss": 1.3605, "step": 2052 }, { "epoch": 0.63, "learning_rate": 6.34879890720684e-06, "loss": 1.372, "step": 2053 }, { "epoch": 0.63, "learning_rate": 6.339521856762254e-06, "loss": 1.2671, "step": 2054 }, { "epoch": 0.63, "learning_rate": 6.330248442662919e-06, "loss": 1.3776, "step": 2055 }, { "epoch": 0.63, "learning_rate": 6.320978674121103e-06, "loss": 1.6665, "step": 2056 }, { "epoch": 0.63, "learning_rate": 6.3117125603454665e-06, "loss": 2.2293, "step": 2057 }, { "epoch": 0.63, "learning_rate": 6.3024501105410295e-06, "loss": 1.6166, "step": 2058 }, { "epoch": 0.63, "learning_rate": 6.2931913339091755e-06, "loss": 1.5341, "step": 2059 }, { "epoch": 0.63, "learning_rate": 6.283936239647643e-06, "loss": 1.588, "step": 2060 }, { "epoch": 0.63, "learning_rate": 6.274684836950509e-06, "loss": 1.2892, "step": 2061 }, { "epoch": 0.63, "learning_rate": 6.265437135008181e-06, "loss": 1.5453, "step": 2062 }, { "epoch": 0.63, "learning_rate": 6.256193143007392e-06, "loss": 1.7253, "step": 2063 }, { "epoch": 0.64, "learning_rate": 6.2469528701311935e-06, "loss": 2.1678, "step": 2064 }, { "epoch": 0.64, "learning_rate": 6.237716325558939e-06, "loss": 1.3906, "step": 2065 }, { "epoch": 0.64, "learning_rate": 6.228483518466271e-06, "loss": 1.3488, "step": 2066 }, { "epoch": 0.64, "learning_rate": 6.219254458025132e-06, "loss": 1.6074, "step": 2067 }, { "epoch": 0.64, "learning_rate": 6.210029153403738e-06, "loss": 1.3285, "step": 2068 }, { "epoch": 0.64, "learning_rate": 6.2008076137665694e-06, "loss": 1.2593, "step": 2069 }, { "epoch": 0.64, "learning_rate": 6.191589848274369e-06, "loss": 1.4621, "step": 2070 }, { "epoch": 0.64, "learning_rate": 6.1823758660841356e-06, "loss": 1.3235, "step": 2071 }, { "epoch": 0.64, "learning_rate": 6.173165676349103e-06, "loss": 1.5771, "step": 2072 }, { "epoch": 0.64, "learning_rate": 6.163959288218741e-06, "loss": 1.3293, "step": 2073 }, { "epoch": 0.64, "learning_rate": 6.154756710838739e-06, "loss": 1.1208, "step": 2074 }, { "epoch": 0.64, "learning_rate": 6.145557953351002e-06, "loss": 1.5575, "step": 2075 }, { "epoch": 0.64, "learning_rate": 6.136363024893647e-06, "loss": 1.0037, "step": 2076 }, { "epoch": 0.64, "learning_rate": 6.12717193460098e-06, "loss": 1.3171, "step": 2077 }, { "epoch": 0.64, "learning_rate": 6.117984691603493e-06, "loss": 1.5036, "step": 2078 }, { "epoch": 0.64, "learning_rate": 6.108801305027863e-06, "loss": 1.5682, "step": 2079 }, { "epoch": 0.64, "learning_rate": 6.099621783996934e-06, "loss": 1.3691, "step": 2080 }, { "epoch": 0.64, "learning_rate": 6.0904461376297e-06, "loss": 1.6378, "step": 2081 }, { "epoch": 0.64, "learning_rate": 6.081274375041317e-06, "loss": 1.4463, "step": 2082 }, { "epoch": 0.64, "learning_rate": 6.072106505343085e-06, "loss": 1.3154, "step": 2083 }, { "epoch": 0.64, "learning_rate": 6.062942537642427e-06, "loss": 2.0535, "step": 2084 }, { "epoch": 0.64, "learning_rate": 6.053782481042894e-06, "loss": 1.464, "step": 2085 }, { "epoch": 0.64, "learning_rate": 6.044626344644151e-06, "loss": 1.4481, "step": 2086 }, { "epoch": 0.64, "learning_rate": 6.035474137541974e-06, "loss": 1.6842, "step": 2087 }, { "epoch": 0.64, "learning_rate": 6.026325868828228e-06, "loss": 1.5216, "step": 2088 }, { "epoch": 0.64, "learning_rate": 6.017181547590869e-06, "loss": 1.2619, "step": 2089 }, { "epoch": 0.64, "learning_rate": 6.008041182913933e-06, "loss": 1.5203, "step": 2090 }, { "epoch": 0.64, "learning_rate": 5.998904783877521e-06, "loss": 1.4824, "step": 2091 }, { "epoch": 0.64, "learning_rate": 5.989772359557799e-06, "loss": 1.4267, "step": 2092 }, { "epoch": 0.64, "learning_rate": 5.980643919026985e-06, "loss": 1.8765, "step": 2093 }, { "epoch": 0.64, "learning_rate": 5.971519471353335e-06, "loss": 1.3318, "step": 2094 }, { "epoch": 0.64, "learning_rate": 5.96239902560114e-06, "loss": 1.4291, "step": 2095 }, { "epoch": 0.64, "learning_rate": 5.953282590830718e-06, "loss": 1.4627, "step": 2096 }, { "epoch": 0.65, "learning_rate": 5.944170176098398e-06, "loss": 1.3275, "step": 2097 }, { "epoch": 0.65, "learning_rate": 5.935061790456519e-06, "loss": 1.3707, "step": 2098 }, { "epoch": 0.65, "learning_rate": 5.9259574429534195e-06, "loss": 1.4946, "step": 2099 }, { "epoch": 0.65, "learning_rate": 5.916857142633421e-06, "loss": 1.6703, "step": 2100 }, { "epoch": 0.65, "learning_rate": 5.907760898536828e-06, "loss": 1.4886, "step": 2101 }, { "epoch": 0.65, "learning_rate": 5.898668719699914e-06, "loss": 1.3178, "step": 2102 }, { "epoch": 0.65, "learning_rate": 5.889580615154915e-06, "loss": 1.3641, "step": 2103 }, { "epoch": 0.65, "learning_rate": 5.880496593930021e-06, "loss": 1.0599, "step": 2104 }, { "epoch": 0.65, "learning_rate": 5.871416665049361e-06, "loss": 1.5125, "step": 2105 }, { "epoch": 0.65, "learning_rate": 5.8623408375330035e-06, "loss": 1.1797, "step": 2106 }, { "epoch": 0.65, "learning_rate": 5.853269120396938e-06, "loss": 1.2804, "step": 2107 }, { "epoch": 0.65, "learning_rate": 5.844201522653081e-06, "loss": 1.4735, "step": 2108 }, { "epoch": 0.65, "learning_rate": 5.835138053309239e-06, "loss": 1.3389, "step": 2109 }, { "epoch": 0.65, "learning_rate": 5.82607872136913e-06, "loss": 1.2159, "step": 2110 }, { "epoch": 0.65, "learning_rate": 5.8170235358323665e-06, "loss": 1.3314, "step": 2111 }, { "epoch": 0.65, "learning_rate": 5.8079725056944245e-06, "loss": 1.4456, "step": 2112 }, { "epoch": 0.65, "learning_rate": 5.798925639946669e-06, "loss": 1.3746, "step": 2113 }, { "epoch": 0.65, "learning_rate": 5.789882947576321e-06, "loss": 1.6103, "step": 2114 }, { "epoch": 0.65, "learning_rate": 5.780844437566451e-06, "loss": 1.2023, "step": 2115 }, { "epoch": 0.65, "learning_rate": 5.771810118895984e-06, "loss": 1.123, "step": 2116 }, { "epoch": 0.65, "learning_rate": 5.762780000539681e-06, "loss": 1.7342, "step": 2117 }, { "epoch": 0.65, "learning_rate": 5.753754091468116e-06, "loss": 1.2948, "step": 2118 }, { "epoch": 0.65, "learning_rate": 5.7447324006477016e-06, "loss": 1.6341, "step": 2119 }, { "epoch": 0.65, "learning_rate": 5.735714937040642e-06, "loss": 1.6557, "step": 2120 }, { "epoch": 0.65, "learning_rate": 5.726701709604955e-06, "loss": 0.9759, "step": 2121 }, { "epoch": 0.65, "learning_rate": 5.717692727294453e-06, "loss": 1.3825, "step": 2122 }, { "epoch": 0.65, "learning_rate": 5.7086879990587064e-06, "loss": 1.4902, "step": 2123 }, { "epoch": 0.65, "learning_rate": 5.699687533843086e-06, "loss": 1.6783, "step": 2124 }, { "epoch": 0.65, "learning_rate": 5.690691340588721e-06, "loss": 1.4209, "step": 2125 }, { "epoch": 0.65, "learning_rate": 5.681699428232488e-06, "loss": 1.3473, "step": 2126 }, { "epoch": 0.65, "learning_rate": 5.6727118057070255e-06, "loss": 1.3774, "step": 2127 }, { "epoch": 0.65, "learning_rate": 5.6637284819406936e-06, "loss": 1.5165, "step": 2128 }, { "epoch": 0.66, "learning_rate": 5.654749465857594e-06, "loss": 1.3847, "step": 2129 }, { "epoch": 0.66, "learning_rate": 5.64577476637755e-06, "loss": 1.4085, "step": 2130 }, { "epoch": 0.66, "learning_rate": 5.636804392416084e-06, "loss": 1.5003, "step": 2131 }, { "epoch": 0.66, "learning_rate": 5.627838352884435e-06, "loss": 1.4249, "step": 2132 }, { "epoch": 0.66, "learning_rate": 5.618876656689534e-06, "loss": 1.4116, "step": 2133 }, { "epoch": 0.66, "learning_rate": 5.609919312733987e-06, "loss": 1.7253, "step": 2134 }, { "epoch": 0.66, "learning_rate": 5.600966329916091e-06, "loss": 1.4766, "step": 2135 }, { "epoch": 0.66, "learning_rate": 5.592017717129801e-06, "loss": 1.6466, "step": 2136 }, { "epoch": 0.66, "learning_rate": 5.583073483264727e-06, "loss": 1.7244, "step": 2137 }, { "epoch": 0.66, "learning_rate": 5.574133637206144e-06, "loss": 1.36, "step": 2138 }, { "epoch": 0.66, "learning_rate": 5.565198187834952e-06, "loss": 2.148, "step": 2139 }, { "epoch": 0.66, "learning_rate": 5.556267144027692e-06, "loss": 1.3596, "step": 2140 }, { "epoch": 0.66, "learning_rate": 5.547340514656534e-06, "loss": 1.3419, "step": 2141 }, { "epoch": 0.66, "learning_rate": 5.538418308589246e-06, "loss": 1.5568, "step": 2142 }, { "epoch": 0.66, "learning_rate": 5.529500534689216e-06, "loss": 1.6058, "step": 2143 }, { "epoch": 0.66, "learning_rate": 5.520587201815428e-06, "loss": 1.4502, "step": 2144 }, { "epoch": 0.66, "learning_rate": 5.511678318822443e-06, "loss": 1.4745, "step": 2145 }, { "epoch": 0.66, "learning_rate": 5.502773894560418e-06, "loss": 1.8227, "step": 2146 }, { "epoch": 0.66, "learning_rate": 5.493873937875063e-06, "loss": 1.4998, "step": 2147 }, { "epoch": 0.66, "learning_rate": 5.484978457607664e-06, "loss": 1.413, "step": 2148 }, { "epoch": 0.66, "learning_rate": 5.476087462595058e-06, "loss": 1.6073, "step": 2149 }, { "epoch": 0.66, "learning_rate": 5.467200961669619e-06, "loss": 1.3238, "step": 2150 }, { "epoch": 0.66, "learning_rate": 5.458318963659257e-06, "loss": 1.3572, "step": 2151 }, { "epoch": 0.66, "learning_rate": 5.4494414773874205e-06, "loss": 1.3029, "step": 2152 }, { "epoch": 0.66, "learning_rate": 5.440568511673061e-06, "loss": 1.2352, "step": 2153 }, { "epoch": 0.66, "learning_rate": 5.431700075330654e-06, "loss": 1.9073, "step": 2154 }, { "epoch": 0.66, "learning_rate": 5.422836177170159e-06, "loss": 1.6075, "step": 2155 }, { "epoch": 0.66, "learning_rate": 5.4139768259970385e-06, "loss": 1.2912, "step": 2156 }, { "epoch": 0.66, "learning_rate": 5.405122030612243e-06, "loss": 1.4296, "step": 2157 }, { "epoch": 0.66, "learning_rate": 5.39627179981218e-06, "loss": 1.1789, "step": 2158 }, { "epoch": 0.66, "learning_rate": 5.387426142388736e-06, "loss": 1.2714, "step": 2159 }, { "epoch": 0.66, "learning_rate": 5.378585067129256e-06, "loss": 1.6964, "step": 2160 }, { "epoch": 0.66, "learning_rate": 5.369748582816517e-06, "loss": 1.5475, "step": 2161 }, { "epoch": 0.67, "learning_rate": 5.360916698228756e-06, "loss": 1.6773, "step": 2162 }, { "epoch": 0.67, "learning_rate": 5.3520894221396214e-06, "loss": 1.3107, "step": 2163 }, { "epoch": 0.67, "learning_rate": 5.343266763318202e-06, "loss": 1.5463, "step": 2164 }, { "epoch": 0.67, "learning_rate": 5.334448730528985e-06, "loss": 1.3822, "step": 2165 }, { "epoch": 0.67, "learning_rate": 5.325635332531864e-06, "loss": 1.6067, "step": 2166 }, { "epoch": 0.67, "learning_rate": 5.316826578082136e-06, "loss": 1.4894, "step": 2167 }, { "epoch": 0.67, "learning_rate": 5.308022475930484e-06, "loss": 1.4188, "step": 2168 }, { "epoch": 0.67, "learning_rate": 5.299223034822959e-06, "loss": 1.558, "step": 2169 }, { "epoch": 0.67, "learning_rate": 5.290428263500996e-06, "loss": 1.3369, "step": 2170 }, { "epoch": 0.67, "learning_rate": 5.281638170701378e-06, "loss": 1.2599, "step": 2171 }, { "epoch": 0.67, "learning_rate": 5.2728527651562475e-06, "loss": 1.3512, "step": 2172 }, { "epoch": 0.67, "learning_rate": 5.264072055593094e-06, "loss": 1.5191, "step": 2173 }, { "epoch": 0.67, "learning_rate": 5.255296050734733e-06, "loss": 1.6249, "step": 2174 }, { "epoch": 0.67, "learning_rate": 5.2465247592993095e-06, "loss": 1.3065, "step": 2175 }, { "epoch": 0.67, "learning_rate": 5.237758190000295e-06, "loss": 1.6039, "step": 2176 }, { "epoch": 0.67, "learning_rate": 5.2289963515464525e-06, "loss": 1.4283, "step": 2177 }, { "epoch": 0.67, "learning_rate": 5.220239252641863e-06, "loss": 1.4831, "step": 2178 }, { "epoch": 0.67, "learning_rate": 5.211486901985889e-06, "loss": 1.2216, "step": 2179 }, { "epoch": 0.67, "learning_rate": 5.202739308273172e-06, "loss": 1.4167, "step": 2180 }, { "epoch": 0.67, "learning_rate": 5.193996480193646e-06, "loss": 0.9553, "step": 2181 }, { "epoch": 0.67, "learning_rate": 5.1852584264324864e-06, "loss": 1.4596, "step": 2182 }, { "epoch": 0.67, "learning_rate": 5.176525155670146e-06, "loss": 1.5661, "step": 2183 }, { "epoch": 0.67, "learning_rate": 5.1677966765823226e-06, "loss": 1.2616, "step": 2184 }, { "epoch": 0.67, "learning_rate": 5.15907299783994e-06, "loss": 1.3346, "step": 2185 }, { "epoch": 0.67, "learning_rate": 5.150354128109173e-06, "loss": 1.1332, "step": 2186 }, { "epoch": 0.67, "learning_rate": 5.1416400760514e-06, "loss": 1.2076, "step": 2187 }, { "epoch": 0.67, "learning_rate": 5.132930850323228e-06, "loss": 1.6124, "step": 2188 }, { "epoch": 0.67, "learning_rate": 5.1242264595764664e-06, "loss": 1.6916, "step": 2189 }, { "epoch": 0.67, "learning_rate": 5.115526912458113e-06, "loss": 1.2544, "step": 2190 }, { "epoch": 0.67, "learning_rate": 5.1068322176103646e-06, "loss": 1.1686, "step": 2191 }, { "epoch": 0.67, "learning_rate": 5.0981423836706e-06, "loss": 1.3855, "step": 2192 }, { "epoch": 0.67, "learning_rate": 5.089457419271349e-06, "loss": 1.4128, "step": 2193 }, { "epoch": 0.68, "learning_rate": 5.080777333040323e-06, "loss": 1.6512, "step": 2194 }, { "epoch": 0.68, "learning_rate": 5.072102133600386e-06, "loss": 1.181, "step": 2195 }, { "epoch": 0.68, "learning_rate": 5.063431829569537e-06, "loss": 1.3364, "step": 2196 }, { "epoch": 0.68, "learning_rate": 5.054766429560927e-06, "loss": 1.6372, "step": 2197 }, { "epoch": 0.68, "learning_rate": 5.046105942182815e-06, "loss": 1.4139, "step": 2198 }, { "epoch": 0.68, "learning_rate": 5.037450376038598e-06, "loss": 1.34, "step": 2199 }, { "epoch": 0.68, "learning_rate": 5.02879973972678e-06, "loss": 1.6916, "step": 2200 }, { "epoch": 0.68, "learning_rate": 5.020154041840959e-06, "loss": 1.5294, "step": 2201 }, { "epoch": 0.68, "learning_rate": 5.011513290969839e-06, "loss": 1.4231, "step": 2202 }, { "epoch": 0.68, "learning_rate": 5.002877495697198e-06, "loss": 1.2681, "step": 2203 }, { "epoch": 0.68, "learning_rate": 4.9942466646019e-06, "loss": 1.5514, "step": 2204 }, { "epoch": 0.68, "learning_rate": 4.98562080625788e-06, "loss": 1.7546, "step": 2205 }, { "epoch": 0.68, "learning_rate": 4.97699992923412e-06, "loss": 1.6773, "step": 2206 }, { "epoch": 0.68, "learning_rate": 4.96838404209466e-06, "loss": 1.2192, "step": 2207 }, { "epoch": 0.68, "learning_rate": 4.9597731533985915e-06, "loss": 1.4569, "step": 2208 }, { "epoch": 0.68, "learning_rate": 4.951167271700026e-06, "loss": 1.463, "step": 2209 }, { "epoch": 0.68, "learning_rate": 4.942566405548109e-06, "loss": 1.3425, "step": 2210 }, { "epoch": 0.68, "learning_rate": 4.933970563487011e-06, "loss": 1.2888, "step": 2211 }, { "epoch": 0.68, "learning_rate": 4.925379754055891e-06, "loss": 1.8507, "step": 2212 }, { "epoch": 0.68, "learning_rate": 4.916793985788931e-06, "loss": 1.406, "step": 2213 }, { "epoch": 0.68, "learning_rate": 4.908213267215287e-06, "loss": 1.4564, "step": 2214 }, { "epoch": 0.68, "learning_rate": 4.89963760685911e-06, "loss": 1.425, "step": 2215 }, { "epoch": 0.68, "learning_rate": 4.891067013239526e-06, "loss": 1.5078, "step": 2216 }, { "epoch": 0.68, "learning_rate": 4.882501494870616e-06, "loss": 1.624, "step": 2217 }, { "epoch": 0.68, "learning_rate": 4.873941060261437e-06, "loss": 1.5035, "step": 2218 }, { "epoch": 0.68, "learning_rate": 4.865385717915976e-06, "loss": 1.4894, "step": 2219 }, { "epoch": 0.68, "learning_rate": 4.856835476333181e-06, "loss": 1.4498, "step": 2220 }, { "epoch": 0.68, "learning_rate": 4.848290344006917e-06, "loss": 1.2851, "step": 2221 }, { "epoch": 0.68, "learning_rate": 4.839750329425979e-06, "loss": 1.6096, "step": 2222 }, { "epoch": 0.68, "learning_rate": 4.8312154410740815e-06, "loss": 1.3143, "step": 2223 }, { "epoch": 0.68, "learning_rate": 4.822685687429848e-06, "loss": 1.4599, "step": 2224 }, { "epoch": 0.68, "learning_rate": 4.8141610769667876e-06, "loss": 1.4214, "step": 2225 }, { "epoch": 0.68, "learning_rate": 4.805641618153316e-06, "loss": 1.3106, "step": 2226 }, { "epoch": 0.69, "learning_rate": 4.7971273194527254e-06, "loss": 1.3005, "step": 2227 }, { "epoch": 0.69, "learning_rate": 4.788618189323176e-06, "loss": 1.1305, "step": 2228 }, { "epoch": 0.69, "learning_rate": 4.780114236217706e-06, "loss": 1.1596, "step": 2229 }, { "epoch": 0.69, "learning_rate": 4.771615468584194e-06, "loss": 1.2986, "step": 2230 }, { "epoch": 0.69, "learning_rate": 4.763121894865382e-06, "loss": 1.7865, "step": 2231 }, { "epoch": 0.69, "learning_rate": 4.7546335234988515e-06, "loss": 1.404, "step": 2232 }, { "epoch": 0.69, "learning_rate": 4.746150362917003e-06, "loss": 1.6623, "step": 2233 }, { "epoch": 0.69, "learning_rate": 4.737672421547076e-06, "loss": 1.6717, "step": 2234 }, { "epoch": 0.69, "learning_rate": 4.7291997078111165e-06, "loss": 1.4524, "step": 2235 }, { "epoch": 0.69, "learning_rate": 4.720732230125973e-06, "loss": 1.3729, "step": 2236 }, { "epoch": 0.69, "learning_rate": 4.712269996903307e-06, "loss": 1.4386, "step": 2237 }, { "epoch": 0.69, "learning_rate": 4.703813016549555e-06, "loss": 1.5615, "step": 2238 }, { "epoch": 0.69, "learning_rate": 4.695361297465946e-06, "loss": 1.45, "step": 2239 }, { "epoch": 0.69, "learning_rate": 4.6869148480484795e-06, "loss": 1.3076, "step": 2240 }, { "epoch": 0.69, "learning_rate": 4.6784736766879136e-06, "loss": 1.5083, "step": 2241 }, { "epoch": 0.69, "learning_rate": 4.670037791769772e-06, "loss": 1.2165, "step": 2242 }, { "epoch": 0.69, "learning_rate": 4.661607201674328e-06, "loss": 1.1603, "step": 2243 }, { "epoch": 0.69, "learning_rate": 4.65318191477658e-06, "loss": 1.2037, "step": 2244 }, { "epoch": 0.69, "learning_rate": 4.644761939446277e-06, "loss": 1.6643, "step": 2245 }, { "epoch": 0.69, "learning_rate": 4.636347284047878e-06, "loss": 1.6561, "step": 2246 }, { "epoch": 0.69, "learning_rate": 4.6279379569405626e-06, "loss": 1.4538, "step": 2247 }, { "epoch": 0.69, "learning_rate": 4.619533966478225e-06, "loss": 1.5816, "step": 2248 }, { "epoch": 0.69, "learning_rate": 4.6111353210094346e-06, "loss": 1.3687, "step": 2249 }, { "epoch": 0.69, "learning_rate": 4.602742028877475e-06, "loss": 1.1537, "step": 2250 }, { "epoch": 0.69, "learning_rate": 4.594354098420305e-06, "loss": 1.7346, "step": 2251 }, { "epoch": 0.69, "learning_rate": 4.5859715379705475e-06, "loss": 1.4227, "step": 2252 }, { "epoch": 0.69, "learning_rate": 4.577594355855506e-06, "loss": 2.1561, "step": 2253 }, { "epoch": 0.69, "learning_rate": 4.569222560397126e-06, "loss": 1.4633, "step": 2254 }, { "epoch": 0.69, "learning_rate": 4.560856159912013e-06, "loss": 1.7257, "step": 2255 }, { "epoch": 0.69, "learning_rate": 4.552495162711412e-06, "loss": 1.3568, "step": 2256 }, { "epoch": 0.69, "learning_rate": 4.544139577101191e-06, "loss": 1.2608, "step": 2257 }, { "epoch": 0.69, "learning_rate": 4.535789411381852e-06, "loss": 1.4341, "step": 2258 }, { "epoch": 0.7, "learning_rate": 4.527444673848514e-06, "loss": 1.3042, "step": 2259 }, { "epoch": 0.7, "learning_rate": 4.519105372790892e-06, "loss": 1.2717, "step": 2260 }, { "epoch": 0.7, "learning_rate": 4.510771516493313e-06, "loss": 1.4883, "step": 2261 }, { "epoch": 0.7, "learning_rate": 4.502443113234688e-06, "loss": 2.1037, "step": 2262 }, { "epoch": 0.7, "learning_rate": 4.4941201712885065e-06, "loss": 1.3394, "step": 2263 }, { "epoch": 0.7, "learning_rate": 4.485802698922849e-06, "loss": 1.3954, "step": 2264 }, { "epoch": 0.7, "learning_rate": 4.4774907044003426e-06, "loss": 1.3984, "step": 2265 }, { "epoch": 0.7, "learning_rate": 4.4691841959781865e-06, "loss": 1.3015, "step": 2266 }, { "epoch": 0.7, "learning_rate": 4.46088318190813e-06, "loss": 1.231, "step": 2267 }, { "epoch": 0.7, "learning_rate": 4.452587670436451e-06, "loss": 1.3133, "step": 2268 }, { "epoch": 0.7, "learning_rate": 4.444297669803981e-06, "loss": 1.8337, "step": 2269 }, { "epoch": 0.7, "learning_rate": 4.436013188246056e-06, "loss": 1.4048, "step": 2270 }, { "epoch": 0.7, "learning_rate": 4.427734233992542e-06, "loss": 1.6339, "step": 2271 }, { "epoch": 0.7, "learning_rate": 4.419460815267818e-06, "loss": 1.5258, "step": 2272 }, { "epoch": 0.7, "learning_rate": 4.41119294029075e-06, "loss": 1.5853, "step": 2273 }, { "epoch": 0.7, "learning_rate": 4.402930617274704e-06, "loss": 1.2298, "step": 2274 }, { "epoch": 0.7, "learning_rate": 4.394673854427541e-06, "loss": 1.2447, "step": 2275 }, { "epoch": 0.7, "learning_rate": 4.386422659951575e-06, "loss": 1.4458, "step": 2276 }, { "epoch": 0.7, "learning_rate": 4.378177042043613e-06, "loss": 1.3232, "step": 2277 }, { "epoch": 0.7, "learning_rate": 4.369937008894906e-06, "loss": 1.3119, "step": 2278 }, { "epoch": 0.7, "learning_rate": 4.361702568691159e-06, "loss": 1.342, "step": 2279 }, { "epoch": 0.7, "learning_rate": 4.353473729612531e-06, "loss": 1.4981, "step": 2280 }, { "epoch": 0.7, "learning_rate": 4.3452504998336005e-06, "loss": 1.465, "step": 2281 }, { "epoch": 0.7, "learning_rate": 4.33703288752339e-06, "loss": 1.7319, "step": 2282 }, { "epoch": 0.7, "learning_rate": 4.328820900845336e-06, "loss": 1.513, "step": 2283 }, { "epoch": 0.7, "learning_rate": 4.320614547957278e-06, "loss": 1.512, "step": 2284 }, { "epoch": 0.7, "learning_rate": 4.31241383701147e-06, "loss": 1.4747, "step": 2285 }, { "epoch": 0.7, "learning_rate": 4.30421877615456e-06, "loss": 1.4082, "step": 2286 }, { "epoch": 0.7, "learning_rate": 4.296029373527573e-06, "loss": 1.6727, "step": 2287 }, { "epoch": 0.7, "learning_rate": 4.2878456372659295e-06, "loss": 1.6795, "step": 2288 }, { "epoch": 0.7, "learning_rate": 4.279667575499402e-06, "loss": 1.4107, "step": 2289 }, { "epoch": 0.7, "learning_rate": 4.271495196352141e-06, "loss": 1.4648, "step": 2290 }, { "epoch": 0.7, "learning_rate": 4.263328507942653e-06, "loss": 1.0956, "step": 2291 }, { "epoch": 0.71, "learning_rate": 4.255167518383771e-06, "loss": 1.1659, "step": 2292 }, { "epoch": 0.71, "learning_rate": 4.247012235782686e-06, "loss": 1.44, "step": 2293 }, { "epoch": 0.71, "learning_rate": 4.23886266824092e-06, "loss": 1.2127, "step": 2294 }, { "epoch": 0.71, "learning_rate": 4.230718823854304e-06, "loss": 1.3666, "step": 2295 }, { "epoch": 0.71, "learning_rate": 4.222580710712998e-06, "loss": 1.5255, "step": 2296 }, { "epoch": 0.71, "learning_rate": 4.2144483369014555e-06, "loss": 1.299, "step": 2297 }, { "epoch": 0.71, "learning_rate": 4.2063217104984374e-06, "loss": 1.1362, "step": 2298 }, { "epoch": 0.71, "learning_rate": 4.198200839576999e-06, "loss": 1.1963, "step": 2299 }, { "epoch": 0.71, "learning_rate": 4.190085732204462e-06, "loss": 1.5483, "step": 2300 }, { "epoch": 0.71, "learning_rate": 4.181976396442436e-06, "loss": 1.3075, "step": 2301 }, { "epoch": 0.71, "learning_rate": 4.173872840346798e-06, "loss": 1.069, "step": 2302 }, { "epoch": 0.71, "learning_rate": 4.1657750719676715e-06, "loss": 1.3717, "step": 2303 }, { "epoch": 0.71, "learning_rate": 4.157683099349444e-06, "loss": 1.5581, "step": 2304 }, { "epoch": 0.71, "learning_rate": 4.149596930530736e-06, "loss": 1.0952, "step": 2305 }, { "epoch": 0.71, "learning_rate": 4.1415165735444016e-06, "loss": 1.2826, "step": 2306 }, { "epoch": 0.71, "learning_rate": 4.133442036417534e-06, "loss": 1.3499, "step": 2307 }, { "epoch": 0.71, "learning_rate": 4.125373327171428e-06, "loss": 1.4579, "step": 2308 }, { "epoch": 0.71, "learning_rate": 4.117310453821602e-06, "loss": 1.2177, "step": 2309 }, { "epoch": 0.71, "learning_rate": 4.109253424377773e-06, "loss": 1.8603, "step": 2310 }, { "epoch": 0.71, "learning_rate": 4.101202246843846e-06, "loss": 1.3527, "step": 2311 }, { "epoch": 0.71, "learning_rate": 4.093156929217929e-06, "loss": 1.6952, "step": 2312 }, { "epoch": 0.71, "learning_rate": 4.085117479492285e-06, "loss": 1.3367, "step": 2313 }, { "epoch": 0.71, "learning_rate": 4.07708390565337e-06, "loss": 1.4601, "step": 2314 }, { "epoch": 0.71, "learning_rate": 4.069056215681794e-06, "loss": 1.5856, "step": 2315 }, { "epoch": 0.71, "learning_rate": 4.0610344175523175e-06, "loss": 1.3092, "step": 2316 }, { "epoch": 0.71, "learning_rate": 4.053018519233853e-06, "loss": 1.2615, "step": 2317 }, { "epoch": 0.71, "learning_rate": 4.045008528689457e-06, "loss": 1.7463, "step": 2318 }, { "epoch": 0.71, "learning_rate": 4.037004453876308e-06, "loss": 1.1891, "step": 2319 }, { "epoch": 0.71, "learning_rate": 4.029006302745706e-06, "loss": 1.5644, "step": 2320 }, { "epoch": 0.71, "learning_rate": 4.02101408324308e-06, "loss": 1.5895, "step": 2321 }, { "epoch": 0.71, "learning_rate": 4.01302780330795e-06, "loss": 1.4587, "step": 2322 }, { "epoch": 0.71, "learning_rate": 4.005047470873954e-06, "loss": 1.4265, "step": 2323 }, { "epoch": 0.72, "learning_rate": 3.997073093868802e-06, "loss": 1.6216, "step": 2324 }, { "epoch": 0.72, "learning_rate": 3.9891046802143e-06, "loss": 1.6914, "step": 2325 }, { "epoch": 0.72, "learning_rate": 3.981142237826332e-06, "loss": 1.4617, "step": 2326 }, { "epoch": 0.72, "learning_rate": 3.97318577461484e-06, "loss": 1.6414, "step": 2327 }, { "epoch": 0.72, "learning_rate": 3.965235298483835e-06, "loss": 1.636, "step": 2328 }, { "epoch": 0.72, "learning_rate": 3.957290817331373e-06, "loss": 1.6273, "step": 2329 }, { "epoch": 0.72, "learning_rate": 3.949352339049561e-06, "loss": 1.2289, "step": 2330 }, { "epoch": 0.72, "learning_rate": 3.941419871524542e-06, "loss": 1.7793, "step": 2331 }, { "epoch": 0.72, "learning_rate": 3.933493422636483e-06, "loss": 1.3266, "step": 2332 }, { "epoch": 0.72, "learning_rate": 3.92557300025958e-06, "loss": 1.2637, "step": 2333 }, { "epoch": 0.72, "learning_rate": 3.917658612262033e-06, "loss": 1.3153, "step": 2334 }, { "epoch": 0.72, "learning_rate": 3.9097502665060506e-06, "loss": 1.3495, "step": 2335 }, { "epoch": 0.72, "learning_rate": 3.901847970847842e-06, "loss": 2.0786, "step": 2336 }, { "epoch": 0.72, "learning_rate": 3.893951733137608e-06, "loss": 1.1121, "step": 2337 }, { "epoch": 0.72, "learning_rate": 3.886061561219521e-06, "loss": 1.5521, "step": 2338 }, { "epoch": 0.72, "learning_rate": 3.8781774629317415e-06, "loss": 1.3826, "step": 2339 }, { "epoch": 0.72, "learning_rate": 3.870299446106382e-06, "loss": 1.4512, "step": 2340 }, { "epoch": 0.72, "learning_rate": 3.862427518569526e-06, "loss": 1.2351, "step": 2341 }, { "epoch": 0.72, "learning_rate": 3.854561688141205e-06, "loss": 1.6359, "step": 2342 }, { "epoch": 0.72, "learning_rate": 3.846701962635383e-06, "loss": 1.4687, "step": 2343 }, { "epoch": 0.72, "learning_rate": 3.838848349859977e-06, "loss": 1.4153, "step": 2344 }, { "epoch": 0.72, "learning_rate": 3.831000857616815e-06, "loss": 1.3847, "step": 2345 }, { "epoch": 0.72, "learning_rate": 3.823159493701654e-06, "loss": 1.3268, "step": 2346 }, { "epoch": 0.72, "learning_rate": 3.815324265904169e-06, "loss": 1.3804, "step": 2347 }, { "epoch": 0.72, "learning_rate": 3.8074951820079186e-06, "loss": 1.5267, "step": 2348 }, { "epoch": 0.72, "learning_rate": 3.799672249790376e-06, "loss": 1.1626, "step": 2349 }, { "epoch": 0.72, "learning_rate": 3.791855477022903e-06, "loss": 1.5135, "step": 2350 }, { "epoch": 0.72, "learning_rate": 3.78404487147073e-06, "loss": 1.8566, "step": 2351 }, { "epoch": 0.72, "learning_rate": 3.776240440892973e-06, "loss": 1.637, "step": 2352 }, { "epoch": 0.72, "learning_rate": 3.768442193042612e-06, "loss": 1.6293, "step": 2353 }, { "epoch": 0.72, "learning_rate": 3.760650135666476e-06, "loss": 1.2378, "step": 2354 }, { "epoch": 0.72, "learning_rate": 3.752864276505257e-06, "loss": 1.2596, "step": 2355 }, { "epoch": 0.72, "learning_rate": 3.7450846232934768e-06, "loss": 1.5485, "step": 2356 }, { "epoch": 0.73, "learning_rate": 3.737311183759502e-06, "loss": 1.3137, "step": 2357 }, { "epoch": 0.73, "learning_rate": 3.729543965625526e-06, "loss": 1.3654, "step": 2358 }, { "epoch": 0.73, "learning_rate": 3.721782976607554e-06, "loss": 1.6538, "step": 2359 }, { "epoch": 0.73, "learning_rate": 3.7140282244154136e-06, "loss": 1.4991, "step": 2360 }, { "epoch": 0.73, "learning_rate": 3.7062797167527274e-06, "loss": 1.3429, "step": 2361 }, { "epoch": 0.73, "learning_rate": 3.6985374613169166e-06, "loss": 1.6156, "step": 2362 }, { "epoch": 0.73, "learning_rate": 3.6908014657991985e-06, "loss": 1.3656, "step": 2363 }, { "epoch": 0.73, "learning_rate": 3.6830717378845614e-06, "loss": 1.4568, "step": 2364 }, { "epoch": 0.73, "learning_rate": 3.675348285251774e-06, "loss": 1.5126, "step": 2365 }, { "epoch": 0.73, "learning_rate": 3.6676311155733745e-06, "loss": 1.6582, "step": 2366 }, { "epoch": 0.73, "learning_rate": 3.6599202365156483e-06, "loss": 1.1857, "step": 2367 }, { "epoch": 0.73, "learning_rate": 3.652215655738639e-06, "loss": 1.8582, "step": 2368 }, { "epoch": 0.73, "learning_rate": 3.6445173808961375e-06, "loss": 1.4223, "step": 2369 }, { "epoch": 0.73, "learning_rate": 3.6368254196356576e-06, "loss": 1.6014, "step": 2370 }, { "epoch": 0.73, "learning_rate": 3.6291397795984564e-06, "loss": 1.1939, "step": 2371 }, { "epoch": 0.73, "learning_rate": 3.6214604684194974e-06, "loss": 1.6545, "step": 2372 }, { "epoch": 0.73, "learning_rate": 3.6137874937274652e-06, "loss": 1.5132, "step": 2373 }, { "epoch": 0.73, "learning_rate": 3.606120863144753e-06, "loss": 1.2304, "step": 2374 }, { "epoch": 0.73, "learning_rate": 3.598460584287442e-06, "loss": 1.5432, "step": 2375 }, { "epoch": 0.73, "learning_rate": 3.590806664765305e-06, "loss": 1.6947, "step": 2376 }, { "epoch": 0.73, "learning_rate": 3.5831591121818086e-06, "loss": 2.0035, "step": 2377 }, { "epoch": 0.73, "learning_rate": 3.5755179341340772e-06, "loss": 0.7274, "step": 2378 }, { "epoch": 0.73, "learning_rate": 3.567883138212922e-06, "loss": 2.0919, "step": 2379 }, { "epoch": 0.73, "learning_rate": 3.5602547320027937e-06, "loss": 1.4689, "step": 2380 }, { "epoch": 0.73, "learning_rate": 3.5526327230818114e-06, "loss": 1.3267, "step": 2381 }, { "epoch": 0.73, "learning_rate": 3.5450171190217364e-06, "loss": 1.2935, "step": 2382 }, { "epoch": 0.73, "learning_rate": 3.537407927387957e-06, "loss": 1.2245, "step": 2383 }, { "epoch": 0.73, "learning_rate": 3.529805155739502e-06, "loss": 1.0391, "step": 2384 }, { "epoch": 0.73, "learning_rate": 3.5222088116290243e-06, "loss": 1.2109, "step": 2385 }, { "epoch": 0.73, "learning_rate": 3.5146189026027787e-06, "loss": 1.2736, "step": 2386 }, { "epoch": 0.73, "learning_rate": 3.507035436200642e-06, "loss": 1.3463, "step": 2387 }, { "epoch": 0.73, "learning_rate": 3.4994584199560767e-06, "loss": 1.3743, "step": 2388 }, { "epoch": 0.74, "learning_rate": 3.491887861396154e-06, "loss": 1.3068, "step": 2389 }, { "epoch": 0.74, "learning_rate": 3.4843237680415153e-06, "loss": 1.4922, "step": 2390 }, { "epoch": 0.74, "learning_rate": 3.476766147406384e-06, "loss": 1.2816, "step": 2391 }, { "epoch": 0.74, "learning_rate": 3.4692150069985563e-06, "loss": 1.4652, "step": 2392 }, { "epoch": 0.74, "learning_rate": 3.461670354319393e-06, "loss": 1.5029, "step": 2393 }, { "epoch": 0.74, "learning_rate": 3.4541321968637995e-06, "loss": 1.3108, "step": 2394 }, { "epoch": 0.74, "learning_rate": 3.446600542120243e-06, "loss": 1.208, "step": 2395 }, { "epoch": 0.74, "learning_rate": 3.439075397570716e-06, "loss": 1.3612, "step": 2396 }, { "epoch": 0.74, "learning_rate": 3.431556770690755e-06, "loss": 1.4533, "step": 2397 }, { "epoch": 0.74, "learning_rate": 3.42404466894942e-06, "loss": 1.3458, "step": 2398 }, { "epoch": 0.74, "learning_rate": 3.4165390998092806e-06, "loss": 1.3477, "step": 2399 }, { "epoch": 0.74, "learning_rate": 3.409040070726427e-06, "loss": 1.3303, "step": 2400 }, { "epoch": 0.74, "learning_rate": 3.4015475891504514e-06, "loss": 1.4927, "step": 2401 }, { "epoch": 0.74, "learning_rate": 3.3940616625244314e-06, "loss": 1.3388, "step": 2402 }, { "epoch": 0.74, "learning_rate": 3.3865822982849485e-06, "loss": 1.6998, "step": 2403 }, { "epoch": 0.74, "learning_rate": 3.3791095038620515e-06, "loss": 1.3615, "step": 2404 }, { "epoch": 0.74, "learning_rate": 3.371643286679266e-06, "loss": 1.1296, "step": 2405 }, { "epoch": 0.74, "learning_rate": 3.364183654153592e-06, "loss": 1.5067, "step": 2406 }, { "epoch": 0.74, "learning_rate": 3.356730613695476e-06, "loss": 1.4794, "step": 2407 }, { "epoch": 0.74, "learning_rate": 3.349284172708823e-06, "loss": 1.5227, "step": 2408 }, { "epoch": 0.74, "learning_rate": 3.341844338590987e-06, "loss": 1.4886, "step": 2409 }, { "epoch": 0.74, "learning_rate": 3.334411118732744e-06, "loss": 1.3697, "step": 2410 }, { "epoch": 0.74, "learning_rate": 3.326984520518313e-06, "loss": 1.4566, "step": 2411 }, { "epoch": 0.74, "learning_rate": 3.3195645513253316e-06, "loss": 1.8796, "step": 2412 }, { "epoch": 0.74, "learning_rate": 3.3121512185248437e-06, "loss": 1.4035, "step": 2413 }, { "epoch": 0.74, "learning_rate": 3.304744529481315e-06, "loss": 1.3376, "step": 2414 }, { "epoch": 0.74, "learning_rate": 3.2973444915525966e-06, "loss": 1.3358, "step": 2415 }, { "epoch": 0.74, "learning_rate": 3.289951112089943e-06, "loss": 0.9312, "step": 2416 }, { "epoch": 0.74, "learning_rate": 3.282564398437995e-06, "loss": 1.2696, "step": 2417 }, { "epoch": 0.74, "learning_rate": 3.2751843579347562e-06, "loss": 1.2902, "step": 2418 }, { "epoch": 0.74, "learning_rate": 3.267810997911617e-06, "loss": 1.448, "step": 2419 }, { "epoch": 0.74, "learning_rate": 3.2604443256933307e-06, "loss": 1.4241, "step": 2420 }, { "epoch": 0.74, "learning_rate": 3.2530843485979936e-06, "loss": 1.2873, "step": 2421 }, { "epoch": 0.75, "learning_rate": 3.2457310739370684e-06, "loss": 1.1503, "step": 2422 }, { "epoch": 0.75, "learning_rate": 3.2383845090153434e-06, "loss": 1.058, "step": 2423 }, { "epoch": 0.75, "learning_rate": 3.231044661130951e-06, "loss": 1.54, "step": 2424 }, { "epoch": 0.75, "learning_rate": 3.2237115375753547e-06, "loss": 1.2798, "step": 2425 }, { "epoch": 0.75, "learning_rate": 3.216385145633324e-06, "loss": 1.5906, "step": 2426 }, { "epoch": 0.75, "learning_rate": 3.209065492582951e-06, "loss": 1.3133, "step": 2427 }, { "epoch": 0.75, "learning_rate": 3.201752585695638e-06, "loss": 1.5513, "step": 2428 }, { "epoch": 0.75, "learning_rate": 3.1944464322360704e-06, "loss": 1.413, "step": 2429 }, { "epoch": 0.75, "learning_rate": 3.1871470394622407e-06, "loss": 1.3208, "step": 2430 }, { "epoch": 0.75, "learning_rate": 3.179854414625415e-06, "loss": 1.4812, "step": 2431 }, { "epoch": 0.75, "learning_rate": 3.1725685649701354e-06, "loss": 1.4869, "step": 2432 }, { "epoch": 0.75, "learning_rate": 3.165289497734223e-06, "loss": 1.2375, "step": 2433 }, { "epoch": 0.75, "learning_rate": 3.158017220148751e-06, "loss": 1.794, "step": 2434 }, { "epoch": 0.75, "learning_rate": 3.1507517394380536e-06, "loss": 1.7377, "step": 2435 }, { "epoch": 0.75, "learning_rate": 3.1434930628197145e-06, "loss": 1.3471, "step": 2436 }, { "epoch": 0.75, "learning_rate": 3.1362411975045504e-06, "loss": 1.3891, "step": 2437 }, { "epoch": 0.75, "learning_rate": 3.1289961506966217e-06, "loss": 1.2586, "step": 2438 }, { "epoch": 0.75, "learning_rate": 3.121757929593203e-06, "loss": 1.2273, "step": 2439 }, { "epoch": 0.75, "learning_rate": 3.114526541384799e-06, "loss": 1.3378, "step": 2440 }, { "epoch": 0.75, "learning_rate": 3.107301993255126e-06, "loss": 1.8218, "step": 2441 }, { "epoch": 0.75, "learning_rate": 3.100084292381095e-06, "loss": 1.6571, "step": 2442 }, { "epoch": 0.75, "learning_rate": 3.092873445932825e-06, "loss": 1.4724, "step": 2443 }, { "epoch": 0.75, "learning_rate": 3.085669461073626e-06, "loss": 1.0988, "step": 2444 }, { "epoch": 0.75, "learning_rate": 3.078472344959985e-06, "loss": 1.5087, "step": 2445 }, { "epoch": 0.75, "learning_rate": 3.0712821047415655e-06, "loss": 1.5175, "step": 2446 }, { "epoch": 0.75, "learning_rate": 3.0640987475612104e-06, "loss": 1.317, "step": 2447 }, { "epoch": 0.75, "learning_rate": 3.0569222805549114e-06, "loss": 1.6835, "step": 2448 }, { "epoch": 0.75, "learning_rate": 3.0497527108518286e-06, "loss": 1.3147, "step": 2449 }, { "epoch": 0.75, "learning_rate": 3.0425900455742584e-06, "loss": 1.5031, "step": 2450 }, { "epoch": 0.75, "learning_rate": 3.0354342918376467e-06, "loss": 1.4818, "step": 2451 }, { "epoch": 0.75, "learning_rate": 3.0282854567505737e-06, "loss": 1.6095, "step": 2452 }, { "epoch": 0.75, "learning_rate": 3.0211435474147364e-06, "loss": 1.3944, "step": 2453 }, { "epoch": 0.76, "learning_rate": 3.0140085709249666e-06, "loss": 1.2271, "step": 2454 }, { "epoch": 0.76, "learning_rate": 3.0068805343691944e-06, "loss": 1.3591, "step": 2455 }, { "epoch": 0.76, "learning_rate": 2.9997594448284638e-06, "loss": 1.3668, "step": 2456 }, { "epoch": 0.76, "learning_rate": 2.9926453093769215e-06, "loss": 1.3199, "step": 2457 }, { "epoch": 0.76, "learning_rate": 2.9855381350817937e-06, "loss": 1.335, "step": 2458 }, { "epoch": 0.76, "learning_rate": 2.9784379290034038e-06, "loss": 1.2559, "step": 2459 }, { "epoch": 0.76, "learning_rate": 2.9713446981951466e-06, "loss": 1.3688, "step": 2460 }, { "epoch": 0.76, "learning_rate": 2.964258449703482e-06, "loss": 1.7332, "step": 2461 }, { "epoch": 0.76, "learning_rate": 2.957179190567947e-06, "loss": 1.5644, "step": 2462 }, { "epoch": 0.76, "learning_rate": 2.950106927821128e-06, "loss": 1.426, "step": 2463 }, { "epoch": 0.76, "learning_rate": 2.9430416684886575e-06, "loss": 1.6684, "step": 2464 }, { "epoch": 0.76, "learning_rate": 2.93598341958922e-06, "loss": 1.4267, "step": 2465 }, { "epoch": 0.76, "learning_rate": 2.9289321881345257e-06, "loss": 1.4776, "step": 2466 }, { "epoch": 0.76, "learning_rate": 2.9218879811293198e-06, "loss": 1.4945, "step": 2467 }, { "epoch": 0.76, "learning_rate": 2.9148508055713733e-06, "loss": 1.3943, "step": 2468 }, { "epoch": 0.76, "learning_rate": 2.9078206684514575e-06, "loss": 1.3681, "step": 2469 }, { "epoch": 0.76, "learning_rate": 2.9007975767533714e-06, "loss": 1.5331, "step": 2470 }, { "epoch": 0.76, "learning_rate": 2.893781537453895e-06, "loss": 1.3095, "step": 2471 }, { "epoch": 0.76, "learning_rate": 2.8867725575228166e-06, "loss": 1.212, "step": 2472 }, { "epoch": 0.76, "learning_rate": 2.87977064392291e-06, "loss": 1.2286, "step": 2473 }, { "epoch": 0.76, "learning_rate": 2.872775803609924e-06, "loss": 1.3955, "step": 2474 }, { "epoch": 0.76, "learning_rate": 2.8657880435325793e-06, "loss": 1.7652, "step": 2475 }, { "epoch": 0.76, "learning_rate": 2.8588073706325726e-06, "loss": 1.3642, "step": 2476 }, { "epoch": 0.76, "learning_rate": 2.8518337918445484e-06, "loss": 1.3104, "step": 2477 }, { "epoch": 0.76, "learning_rate": 2.844867314096115e-06, "loss": 1.2454, "step": 2478 }, { "epoch": 0.76, "learning_rate": 2.8379079443078215e-06, "loss": 1.3106, "step": 2479 }, { "epoch": 0.76, "learning_rate": 2.8309556893931522e-06, "loss": 1.51, "step": 2480 }, { "epoch": 0.76, "learning_rate": 2.8240105562585307e-06, "loss": 1.3334, "step": 2481 }, { "epoch": 0.76, "learning_rate": 2.8170725518032983e-06, "loss": 1.0288, "step": 2482 }, { "epoch": 0.76, "learning_rate": 2.8101416829197204e-06, "loss": 1.6592, "step": 2483 }, { "epoch": 0.76, "learning_rate": 2.803217956492974e-06, "loss": 1.5159, "step": 2484 }, { "epoch": 0.76, "learning_rate": 2.796301379401134e-06, "loss": 1.8982, "step": 2485 }, { "epoch": 0.76, "learning_rate": 2.789391958515183e-06, "loss": 1.3649, "step": 2486 }, { "epoch": 0.77, "learning_rate": 2.782489700698984e-06, "loss": 1.392, "step": 2487 }, { "epoch": 0.77, "learning_rate": 2.7755946128092935e-06, "loss": 1.5098, "step": 2488 }, { "epoch": 0.77, "learning_rate": 2.7687067016957413e-06, "loss": 1.6033, "step": 2489 }, { "epoch": 0.77, "learning_rate": 2.7618259742008226e-06, "loss": 1.5718, "step": 2490 }, { "epoch": 0.77, "learning_rate": 2.7549524371599033e-06, "loss": 1.5207, "step": 2491 }, { "epoch": 0.77, "learning_rate": 2.74808609740121e-06, "loss": 1.5259, "step": 2492 }, { "epoch": 0.77, "learning_rate": 2.7412269617458074e-06, "loss": 1.2681, "step": 2493 }, { "epoch": 0.77, "learning_rate": 2.7343750370076127e-06, "loss": 1.6076, "step": 2494 }, { "epoch": 0.77, "learning_rate": 2.7275303299933786e-06, "loss": 1.6076, "step": 2495 }, { "epoch": 0.77, "learning_rate": 2.7206928475026828e-06, "loss": 1.2696, "step": 2496 }, { "epoch": 0.77, "learning_rate": 2.713862596327933e-06, "loss": 1.4948, "step": 2497 }, { "epoch": 0.77, "learning_rate": 2.707039583254346e-06, "loss": 1.5384, "step": 2498 }, { "epoch": 0.77, "learning_rate": 2.7002238150599527e-06, "loss": 1.3518, "step": 2499 }, { "epoch": 0.77, "learning_rate": 2.6934152985155904e-06, "loss": 1.5482, "step": 2500 }, { "epoch": 0.77, "learning_rate": 2.686614040384883e-06, "loss": 1.3428, "step": 2501 }, { "epoch": 0.77, "learning_rate": 2.679820047424253e-06, "loss": 1.3813, "step": 2502 }, { "epoch": 0.77, "learning_rate": 2.6730333263829e-06, "loss": 1.2026, "step": 2503 }, { "epoch": 0.77, "learning_rate": 2.6662538840027984e-06, "loss": 1.2863, "step": 2504 }, { "epoch": 0.77, "learning_rate": 2.6594817270187012e-06, "loss": 1.5945, "step": 2505 }, { "epoch": 0.77, "learning_rate": 2.6527168621581114e-06, "loss": 1.2914, "step": 2506 }, { "epoch": 0.77, "learning_rate": 2.6459592961412973e-06, "loss": 1.4734, "step": 2507 }, { "epoch": 0.77, "learning_rate": 2.639209035681276e-06, "loss": 1.5494, "step": 2508 }, { "epoch": 0.77, "learning_rate": 2.6324660874837982e-06, "loss": 1.4894, "step": 2509 }, { "epoch": 0.77, "learning_rate": 2.625730458247362e-06, "loss": 1.8397, "step": 2510 }, { "epoch": 0.77, "learning_rate": 2.6190021546631915e-06, "loss": 1.1767, "step": 2511 }, { "epoch": 0.77, "learning_rate": 2.6122811834152253e-06, "loss": 1.5503, "step": 2512 }, { "epoch": 0.77, "learning_rate": 2.6055675511801315e-06, "loss": 1.5331, "step": 2513 }, { "epoch": 0.77, "learning_rate": 2.5988612646272735e-06, "loss": 1.8818, "step": 2514 }, { "epoch": 0.77, "learning_rate": 2.592162330418728e-06, "loss": 1.351, "step": 2515 }, { "epoch": 0.77, "learning_rate": 2.5854707552092705e-06, "loss": 1.4186, "step": 2516 }, { "epoch": 0.77, "learning_rate": 2.578786545646348e-06, "loss": 1.3413, "step": 2517 }, { "epoch": 0.77, "learning_rate": 2.5721097083701085e-06, "loss": 1.9233, "step": 2518 }, { "epoch": 0.78, "learning_rate": 2.5654402500133714e-06, "loss": 1.3252, "step": 2519 }, { "epoch": 0.78, "learning_rate": 2.5587781772016196e-06, "loss": 1.5766, "step": 2520 }, { "epoch": 0.78, "learning_rate": 2.5521234965530094e-06, "loss": 1.5627, "step": 2521 }, { "epoch": 0.78, "learning_rate": 2.545476214678344e-06, "loss": 1.5373, "step": 2522 }, { "epoch": 0.78, "learning_rate": 2.538836338181082e-06, "loss": 1.2069, "step": 2523 }, { "epoch": 0.78, "learning_rate": 2.5322038736573273e-06, "loss": 1.4028, "step": 2524 }, { "epoch": 0.78, "learning_rate": 2.525578827695814e-06, "loss": 1.751, "step": 2525 }, { "epoch": 0.78, "learning_rate": 2.518961206877911e-06, "loss": 1.3764, "step": 2526 }, { "epoch": 0.78, "learning_rate": 2.5123510177776143e-06, "loss": 1.4346, "step": 2527 }, { "epoch": 0.78, "learning_rate": 2.5057482669615263e-06, "loss": 1.3396, "step": 2528 }, { "epoch": 0.78, "learning_rate": 2.4991529609888743e-06, "loss": 1.3525, "step": 2529 }, { "epoch": 0.78, "learning_rate": 2.4925651064114788e-06, "loss": 1.4713, "step": 2530 }, { "epoch": 0.78, "learning_rate": 2.485984709773759e-06, "loss": 1.2568, "step": 2531 }, { "epoch": 0.78, "learning_rate": 2.4794117776127337e-06, "loss": 1.8404, "step": 2532 }, { "epoch": 0.78, "learning_rate": 2.472846316457995e-06, "loss": 1.625, "step": 2533 }, { "epoch": 0.78, "learning_rate": 2.466288332831722e-06, "loss": 1.4472, "step": 2534 }, { "epoch": 0.78, "learning_rate": 2.4597378332486675e-06, "loss": 1.9582, "step": 2535 }, { "epoch": 0.78, "learning_rate": 2.4531948242161365e-06, "loss": 1.4819, "step": 2536 }, { "epoch": 0.78, "learning_rate": 2.4466593122340043e-06, "loss": 1.5532, "step": 2537 }, { "epoch": 0.78, "learning_rate": 2.4401313037946993e-06, "loss": 1.5089, "step": 2538 }, { "epoch": 0.78, "learning_rate": 2.4336108053831863e-06, "loss": 1.3625, "step": 2539 }, { "epoch": 0.78, "learning_rate": 2.4270978234769794e-06, "loss": 1.4871, "step": 2540 }, { "epoch": 0.78, "learning_rate": 2.4205923645461173e-06, "loss": 1.3577, "step": 2541 }, { "epoch": 0.78, "learning_rate": 2.4140944350531714e-06, "loss": 1.4938, "step": 2542 }, { "epoch": 0.78, "learning_rate": 2.4076040414532352e-06, "loss": 1.352, "step": 2543 }, { "epoch": 0.78, "learning_rate": 2.401121190193909e-06, "loss": 1.3946, "step": 2544 }, { "epoch": 0.78, "learning_rate": 2.394645887715301e-06, "loss": 1.0848, "step": 2545 }, { "epoch": 0.78, "learning_rate": 2.3881781404500284e-06, "loss": 1.4984, "step": 2546 }, { "epoch": 0.78, "learning_rate": 2.381717954823193e-06, "loss": 1.2543, "step": 2547 }, { "epoch": 0.78, "learning_rate": 2.3752653372523937e-06, "loss": 1.3804, "step": 2548 }, { "epoch": 0.78, "learning_rate": 2.3688202941477033e-06, "loss": 1.6574, "step": 2549 }, { "epoch": 0.78, "learning_rate": 2.362382831911675e-06, "loss": 1.0916, "step": 2550 }, { "epoch": 0.78, "learning_rate": 2.3559529569393337e-06, "loss": 1.474, "step": 2551 }, { "epoch": 0.79, "learning_rate": 2.349530675618158e-06, "loss": 1.5092, "step": 2552 }, { "epoch": 0.79, "learning_rate": 2.34311599432809e-06, "loss": 1.3123, "step": 2553 }, { "epoch": 0.79, "learning_rate": 2.3367089194415226e-06, "loss": 1.6283, "step": 2554 }, { "epoch": 0.79, "learning_rate": 2.3303094573232865e-06, "loss": 1.4334, "step": 2555 }, { "epoch": 0.79, "learning_rate": 2.3239176143306576e-06, "loss": 1.0077, "step": 2556 }, { "epoch": 0.79, "learning_rate": 2.317533396813332e-06, "loss": 1.1655, "step": 2557 }, { "epoch": 0.79, "learning_rate": 2.311156811113444e-06, "loss": 1.6687, "step": 2558 }, { "epoch": 0.79, "learning_rate": 2.3047878635655373e-06, "loss": 1.0468, "step": 2559 }, { "epoch": 0.79, "learning_rate": 2.298426560496566e-06, "loss": 1.4683, "step": 2560 }, { "epoch": 0.79, "learning_rate": 2.2920729082258974e-06, "loss": 1.8363, "step": 2561 }, { "epoch": 0.79, "learning_rate": 2.2857269130652983e-06, "loss": 1.3517, "step": 2562 }, { "epoch": 0.79, "learning_rate": 2.27938858131892e-06, "loss": 1.3948, "step": 2563 }, { "epoch": 0.79, "learning_rate": 2.2730579192833124e-06, "loss": 1.5454, "step": 2564 }, { "epoch": 0.79, "learning_rate": 2.2667349332473944e-06, "loss": 1.2572, "step": 2565 }, { "epoch": 0.79, "learning_rate": 2.2604196294924696e-06, "loss": 1.439, "step": 2566 }, { "epoch": 0.79, "learning_rate": 2.254112014292209e-06, "loss": 1.0789, "step": 2567 }, { "epoch": 0.79, "learning_rate": 2.254112014292209e-06, "loss": 1.4932, "step": 2568 }, { "epoch": 0.79, "learning_rate": 2.2478120939126357e-06, "loss": 1.4678, "step": 2569 }, { "epoch": 0.79, "learning_rate": 2.2415198746121392e-06, "loss": 1.4271, "step": 2570 }, { "epoch": 0.79, "learning_rate": 2.235235362641458e-06, "loss": 1.2138, "step": 2571 }, { "epoch": 0.79, "learning_rate": 2.2289585642436675e-06, "loss": 1.3228, "step": 2572 }, { "epoch": 0.79, "learning_rate": 2.222689485654187e-06, "loss": 1.2213, "step": 2573 }, { "epoch": 0.79, "learning_rate": 2.216428133100763e-06, "loss": 1.4364, "step": 2574 }, { "epoch": 0.79, "learning_rate": 2.2101745128034657e-06, "loss": 1.5871, "step": 2575 }, { "epoch": 0.79, "learning_rate": 2.203928630974691e-06, "loss": 1.494, "step": 2576 }, { "epoch": 0.79, "learning_rate": 2.1976904938191366e-06, "loss": 1.295, "step": 2577 }, { "epoch": 0.79, "learning_rate": 2.191460107533817e-06, "loss": 1.6491, "step": 2578 }, { "epoch": 0.79, "learning_rate": 2.1852374783080443e-06, "loss": 1.5261, "step": 2579 }, { "epoch": 0.79, "learning_rate": 2.179022612323418e-06, "loss": 1.4721, "step": 2580 }, { "epoch": 0.79, "learning_rate": 2.1728155157538367e-06, "loss": 1.2618, "step": 2581 }, { "epoch": 0.79, "learning_rate": 2.1666161947654673e-06, "loss": 1.5919, "step": 2582 }, { "epoch": 0.79, "learning_rate": 2.160424655516764e-06, "loss": 1.4995, "step": 2583 }, { "epoch": 0.8, "learning_rate": 2.1542409041584487e-06, "loss": 1.4501, "step": 2584 }, { "epoch": 0.8, "learning_rate": 2.148064946833498e-06, "loss": 1.3278, "step": 2585 }, { "epoch": 0.8, "learning_rate": 2.141896789677157e-06, "loss": 1.585, "step": 2586 }, { "epoch": 0.8, "learning_rate": 2.135736438816921e-06, "loss": 1.5089, "step": 2587 }, { "epoch": 0.8, "learning_rate": 2.129583900372516e-06, "loss": 1.514, "step": 2588 }, { "epoch": 0.8, "learning_rate": 2.123439180455924e-06, "loss": 1.3724, "step": 2589 }, { "epoch": 0.8, "learning_rate": 2.1173022851713554e-06, "loss": 1.3589, "step": 2590 }, { "epoch": 0.8, "learning_rate": 2.1111732206152424e-06, "loss": 1.7269, "step": 2591 }, { "epoch": 0.8, "learning_rate": 2.105051992876245e-06, "loss": 1.3547, "step": 2592 }, { "epoch": 0.8, "learning_rate": 2.0989386080352304e-06, "loss": 1.4995, "step": 2593 }, { "epoch": 0.8, "learning_rate": 2.092833072165282e-06, "loss": 1.309, "step": 2594 }, { "epoch": 0.8, "learning_rate": 2.086735391331686e-06, "loss": 1.0108, "step": 2595 }, { "epoch": 0.8, "learning_rate": 2.080645571591915e-06, "loss": 0.9986, "step": 2596 }, { "epoch": 0.8, "learning_rate": 2.0745636189956465e-06, "loss": 1.3846, "step": 2597 }, { "epoch": 0.8, "learning_rate": 2.0684895395847306e-06, "loss": 1.2547, "step": 2598 }, { "epoch": 0.8, "learning_rate": 2.0624233393932024e-06, "loss": 1.3957, "step": 2599 }, { "epoch": 0.8, "learning_rate": 2.0563650244472733e-06, "loss": 1.2948, "step": 2600 }, { "epoch": 0.8, "learning_rate": 2.050314600765313e-06, "loss": 1.2872, "step": 2601 }, { "epoch": 0.8, "learning_rate": 2.044272074357854e-06, "loss": 1.5883, "step": 2602 }, { "epoch": 0.8, "learning_rate": 2.038237451227589e-06, "loss": 1.4656, "step": 2603 }, { "epoch": 0.8, "learning_rate": 2.0322107373693535e-06, "loss": 1.6607, "step": 2604 }, { "epoch": 0.8, "learning_rate": 2.026191938770128e-06, "loss": 1.2261, "step": 2605 }, { "epoch": 0.8, "learning_rate": 2.0201810614090343e-06, "loss": 1.367, "step": 2606 }, { "epoch": 0.8, "learning_rate": 2.0141781112573155e-06, "loss": 1.2662, "step": 2607 }, { "epoch": 0.8, "learning_rate": 2.008183094278352e-06, "loss": 1.9735, "step": 2608 }, { "epoch": 0.8, "learning_rate": 2.002196016427629e-06, "loss": 1.4474, "step": 2609 }, { "epoch": 0.8, "learning_rate": 1.996216883652756e-06, "loss": 1.3568, "step": 2610 }, { "epoch": 0.8, "learning_rate": 1.9902457018934496e-06, "loss": 1.4749, "step": 2611 }, { "epoch": 0.8, "learning_rate": 1.9842824770815194e-06, "loss": 1.5196, "step": 2612 }, { "epoch": 0.8, "learning_rate": 1.9783272151408805e-06, "loss": 1.4516, "step": 2613 }, { "epoch": 0.8, "learning_rate": 1.972379921987527e-06, "loss": 1.1927, "step": 2614 }, { "epoch": 0.8, "learning_rate": 1.9664406035295493e-06, "loss": 1.1425, "step": 2615 }, { "epoch": 0.8, "learning_rate": 1.9605092656671055e-06, "loss": 2.1226, "step": 2616 }, { "epoch": 0.81, "learning_rate": 1.954585914292426e-06, "loss": 1.4244, "step": 2617 }, { "epoch": 0.81, "learning_rate": 1.9486705552898145e-06, "loss": 1.3919, "step": 2618 }, { "epoch": 0.81, "learning_rate": 1.942763194535633e-06, "loss": 1.3921, "step": 2619 }, { "epoch": 0.81, "learning_rate": 1.9368638378982908e-06, "loss": 1.4269, "step": 2620 }, { "epoch": 0.81, "learning_rate": 1.930972491238253e-06, "loss": 1.3205, "step": 2621 }, { "epoch": 0.81, "learning_rate": 1.9250891604080302e-06, "loss": 1.7154, "step": 2622 }, { "epoch": 0.81, "learning_rate": 1.9192138512521585e-06, "loss": 1.4195, "step": 2623 }, { "epoch": 0.81, "learning_rate": 1.9133465696072174e-06, "loss": 1.091, "step": 2624 }, { "epoch": 0.81, "learning_rate": 1.907487321301802e-06, "loss": 1.1435, "step": 2625 }, { "epoch": 0.81, "learning_rate": 1.901636112156534e-06, "loss": 1.3941, "step": 2626 }, { "epoch": 0.81, "learning_rate": 1.8957929479840497e-06, "loss": 1.2989, "step": 2627 }, { "epoch": 0.81, "learning_rate": 1.8899578345889846e-06, "loss": 1.3342, "step": 2628 }, { "epoch": 0.81, "learning_rate": 1.8841307777679874e-06, "loss": 1.5118, "step": 2629 }, { "epoch": 0.81, "learning_rate": 1.8783117833096953e-06, "loss": 1.4909, "step": 2630 }, { "epoch": 0.81, "learning_rate": 1.8725008569947366e-06, "loss": 1.6813, "step": 2631 }, { "epoch": 0.81, "learning_rate": 1.866698004595733e-06, "loss": 1.1792, "step": 2632 }, { "epoch": 0.81, "learning_rate": 1.8609032318772735e-06, "loss": 2.0535, "step": 2633 }, { "epoch": 0.81, "learning_rate": 1.8551165445959308e-06, "loss": 1.5891, "step": 2634 }, { "epoch": 0.81, "learning_rate": 1.8493379485002427e-06, "loss": 1.3032, "step": 2635 }, { "epoch": 0.81, "learning_rate": 1.8435674493307043e-06, "loss": 1.4067, "step": 2636 }, { "epoch": 0.81, "learning_rate": 1.8378050528197722e-06, "loss": 1.2746, "step": 2637 }, { "epoch": 0.81, "learning_rate": 1.8320507646918552e-06, "loss": 1.4263, "step": 2638 }, { "epoch": 0.81, "learning_rate": 1.8263045906633004e-06, "loss": 1.5586, "step": 2639 }, { "epoch": 0.81, "learning_rate": 1.8205665364424019e-06, "loss": 1.4499, "step": 2640 }, { "epoch": 0.81, "learning_rate": 1.81483660772938e-06, "loss": 0.9416, "step": 2641 }, { "epoch": 0.81, "learning_rate": 1.809114810216388e-06, "loss": 1.344, "step": 2642 }, { "epoch": 0.81, "learning_rate": 1.8034011495875082e-06, "loss": 1.193, "step": 2643 }, { "epoch": 0.81, "learning_rate": 1.79769563151872e-06, "loss": 1.4571, "step": 2644 }, { "epoch": 0.81, "learning_rate": 1.7919982616779308e-06, "loss": 1.0854, "step": 2645 }, { "epoch": 0.81, "learning_rate": 1.786309045724952e-06, "loss": 1.2303, "step": 2646 }, { "epoch": 0.81, "learning_rate": 1.7806279893114874e-06, "loss": 1.1636, "step": 2647 }, { "epoch": 0.81, "learning_rate": 1.7749550980811425e-06, "loss": 1.4557, "step": 2648 }, { "epoch": 0.82, "learning_rate": 1.7692903776694037e-06, "loss": 1.1746, "step": 2649 }, { "epoch": 0.82, "learning_rate": 1.7636338337036474e-06, "loss": 1.3724, "step": 2650 }, { "epoch": 0.82, "learning_rate": 1.7579854718031285e-06, "loss": 1.5699, "step": 2651 }, { "epoch": 0.82, "learning_rate": 1.7523452975789634e-06, "loss": 1.4874, "step": 2652 }, { "epoch": 0.82, "learning_rate": 1.7467133166341454e-06, "loss": 1.7284, "step": 2653 }, { "epoch": 0.82, "learning_rate": 1.741089534563527e-06, "loss": 1.2646, "step": 2654 }, { "epoch": 0.82, "learning_rate": 1.7354739569538081e-06, "loss": 1.3907, "step": 2655 }, { "epoch": 0.82, "learning_rate": 1.7298665893835498e-06, "loss": 1.1218, "step": 2656 }, { "epoch": 0.82, "learning_rate": 1.7242674374231483e-06, "loss": 1.4082, "step": 2657 }, { "epoch": 0.82, "learning_rate": 1.7186765066348388e-06, "loss": 1.5865, "step": 2658 }, { "epoch": 0.82, "learning_rate": 1.7130938025726963e-06, "loss": 1.1326, "step": 2659 }, { "epoch": 0.82, "learning_rate": 1.7075193307826144e-06, "loss": 1.1497, "step": 2660 }, { "epoch": 0.82, "learning_rate": 1.701953096802318e-06, "loss": 1.4058, "step": 2661 }, { "epoch": 0.82, "learning_rate": 1.6963951061613426e-06, "loss": 1.1423, "step": 2662 }, { "epoch": 0.82, "learning_rate": 1.6908453643810342e-06, "loss": 1.4075, "step": 2663 }, { "epoch": 0.82, "learning_rate": 1.6853038769745466e-06, "loss": 1.1628, "step": 2664 }, { "epoch": 0.82, "learning_rate": 1.679770649446837e-06, "loss": 1.4037, "step": 2665 }, { "epoch": 0.82, "learning_rate": 1.6742456872946478e-06, "loss": 1.7946, "step": 2666 }, { "epoch": 0.82, "learning_rate": 1.6687289960065213e-06, "loss": 1.777, "step": 2667 }, { "epoch": 0.82, "learning_rate": 1.6632205810627722e-06, "loss": 1.4469, "step": 2668 }, { "epoch": 0.82, "learning_rate": 1.6577204479355026e-06, "loss": 1.5557, "step": 2669 }, { "epoch": 0.82, "learning_rate": 1.652228602088587e-06, "loss": 1.7897, "step": 2670 }, { "epoch": 0.82, "learning_rate": 1.6467450489776581e-06, "loss": 1.2053, "step": 2671 }, { "epoch": 0.82, "learning_rate": 1.6412697940501244e-06, "loss": 1.6465, "step": 2672 }, { "epoch": 0.82, "learning_rate": 1.6358028427451379e-06, "loss": 1.6095, "step": 2673 }, { "epoch": 0.82, "learning_rate": 1.6303442004936077e-06, "loss": 1.1963, "step": 2674 }, { "epoch": 0.82, "learning_rate": 1.6248938727181917e-06, "loss": 1.2632, "step": 2675 }, { "epoch": 0.82, "learning_rate": 1.619451864833279e-06, "loss": 1.463, "step": 2676 }, { "epoch": 0.82, "learning_rate": 1.614018182245003e-06, "loss": 1.2237, "step": 2677 }, { "epoch": 0.82, "learning_rate": 1.6085928303512243e-06, "loss": 1.6416, "step": 2678 }, { "epoch": 0.82, "learning_rate": 1.6031758145415222e-06, "loss": 1.464, "step": 2679 }, { "epoch": 0.82, "learning_rate": 1.5977671401972018e-06, "loss": 1.207, "step": 2680 }, { "epoch": 0.82, "learning_rate": 1.5923668126912806e-06, "loss": 1.2699, "step": 2681 }, { "epoch": 0.83, "learning_rate": 1.5869748373884786e-06, "loss": 1.5458, "step": 2682 }, { "epoch": 0.83, "learning_rate": 1.5815912196452265e-06, "loss": 1.1712, "step": 2683 }, { "epoch": 0.83, "learning_rate": 1.5762159648096453e-06, "loss": 1.4899, "step": 2684 }, { "epoch": 0.83, "learning_rate": 1.5708490782215535e-06, "loss": 1.9001, "step": 2685 }, { "epoch": 0.83, "learning_rate": 1.5654905652124596e-06, "loss": 1.3993, "step": 2686 }, { "epoch": 0.83, "learning_rate": 1.560140431105539e-06, "loss": 1.4974, "step": 2687 }, { "epoch": 0.83, "learning_rate": 1.5547986812156579e-06, "loss": 1.5852, "step": 2688 }, { "epoch": 0.83, "learning_rate": 1.5494653208493527e-06, "loss": 1.4412, "step": 2689 }, { "epoch": 0.83, "learning_rate": 1.5441403553048162e-06, "loss": 1.5521, "step": 2690 }, { "epoch": 0.83, "learning_rate": 1.5388237898719105e-06, "loss": 1.2133, "step": 2691 }, { "epoch": 0.83, "learning_rate": 1.5335156298321485e-06, "loss": 1.7156, "step": 2692 }, { "epoch": 0.83, "learning_rate": 1.5282158804586934e-06, "loss": 0.9455, "step": 2693 }, { "epoch": 0.83, "learning_rate": 1.522924547016359e-06, "loss": 1.4694, "step": 2694 }, { "epoch": 0.83, "learning_rate": 1.5176416347615886e-06, "loss": 1.1641, "step": 2695 }, { "epoch": 0.83, "learning_rate": 1.5123671489424662e-06, "loss": 1.3567, "step": 2696 }, { "epoch": 0.83, "learning_rate": 1.5071010947987085e-06, "loss": 1.4311, "step": 2697 }, { "epoch": 0.83, "learning_rate": 1.5018434775616453e-06, "loss": 1.4747, "step": 2698 }, { "epoch": 0.83, "learning_rate": 1.4965943024542361e-06, "loss": 1.6376, "step": 2699 }, { "epoch": 0.83, "learning_rate": 1.4913535746910467e-06, "loss": 1.268, "step": 2700 }, { "epoch": 0.83, "learning_rate": 1.4861212994782514e-06, "loss": 1.3031, "step": 2701 }, { "epoch": 0.83, "learning_rate": 1.4808974820136345e-06, "loss": 1.2599, "step": 2702 }, { "epoch": 0.83, "learning_rate": 1.4756821274865696e-06, "loss": 1.4308, "step": 2703 }, { "epoch": 0.83, "learning_rate": 1.4704752410780287e-06, "loss": 1.2852, "step": 2704 }, { "epoch": 0.83, "learning_rate": 1.4652768279605734e-06, "loss": 1.2491, "step": 2705 }, { "epoch": 0.83, "learning_rate": 1.4600868932983402e-06, "loss": 1.3842, "step": 2706 }, { "epoch": 0.83, "learning_rate": 1.454905442247051e-06, "loss": 1.4247, "step": 2707 }, { "epoch": 0.83, "learning_rate": 1.4497324799539914e-06, "loss": 1.3358, "step": 2708 }, { "epoch": 0.83, "learning_rate": 1.444568011558023e-06, "loss": 1.3806, "step": 2709 }, { "epoch": 0.83, "learning_rate": 1.4394120421895674e-06, "loss": 1.4426, "step": 2710 }, { "epoch": 0.83, "learning_rate": 1.4342645769705977e-06, "loss": 1.153, "step": 2711 }, { "epoch": 0.83, "learning_rate": 1.4291256210146442e-06, "loss": 1.3113, "step": 2712 }, { "epoch": 0.83, "learning_rate": 1.4239951794267848e-06, "loss": 1.3176, "step": 2713 }, { "epoch": 0.84, "learning_rate": 1.4188732573036346e-06, "loss": 1.3246, "step": 2714 }, { "epoch": 0.84, "learning_rate": 1.4137598597333458e-06, "loss": 1.4593, "step": 2715 }, { "epoch": 0.84, "learning_rate": 1.4086549917956094e-06, "loss": 1.4904, "step": 2716 }, { "epoch": 0.84, "learning_rate": 1.4035586585616323e-06, "loss": 1.842, "step": 2717 }, { "epoch": 0.84, "learning_rate": 1.3984708650941536e-06, "loss": 1.319, "step": 2718 }, { "epoch": 0.84, "learning_rate": 1.3933916164474193e-06, "loss": 1.2174, "step": 2719 }, { "epoch": 0.84, "learning_rate": 1.3883209176671942e-06, "loss": 1.2161, "step": 2720 }, { "epoch": 0.84, "learning_rate": 1.3832587737907476e-06, "loss": 1.6838, "step": 2721 }, { "epoch": 0.84, "learning_rate": 1.3782051898468463e-06, "loss": 1.3245, "step": 2722 }, { "epoch": 0.84, "learning_rate": 1.3731601708557608e-06, "loss": 1.2519, "step": 2723 }, { "epoch": 0.84, "learning_rate": 1.3681237218292442e-06, "loss": 1.7184, "step": 2724 }, { "epoch": 0.84, "learning_rate": 1.3630958477705448e-06, "loss": 1.2253, "step": 2725 }, { "epoch": 0.84, "learning_rate": 1.35807655367439e-06, "loss": 1.6702, "step": 2726 }, { "epoch": 0.84, "learning_rate": 1.3530658445269784e-06, "loss": 1.3254, "step": 2727 }, { "epoch": 0.84, "learning_rate": 1.348063725305988e-06, "loss": 1.3084, "step": 2728 }, { "epoch": 0.84, "learning_rate": 1.3430702009805586e-06, "loss": 1.4824, "step": 2729 }, { "epoch": 0.84, "learning_rate": 1.3380852765112916e-06, "loss": 1.2608, "step": 2730 }, { "epoch": 0.84, "learning_rate": 1.3331089568502465e-06, "loss": 1.3589, "step": 2731 }, { "epoch": 0.84, "learning_rate": 1.3281412469409394e-06, "loss": 1.5002, "step": 2732 }, { "epoch": 0.84, "learning_rate": 1.323182151718324e-06, "loss": 1.6301, "step": 2733 }, { "epoch": 0.84, "learning_rate": 1.3182316761088055e-06, "loss": 1.8812, "step": 2734 }, { "epoch": 0.84, "learning_rate": 1.3132898250302173e-06, "loss": 1.7355, "step": 2735 }, { "epoch": 0.84, "learning_rate": 1.3083566033918316e-06, "loss": 1.3862, "step": 2736 }, { "epoch": 0.84, "learning_rate": 1.303432016094348e-06, "loss": 1.5021, "step": 2737 }, { "epoch": 0.84, "learning_rate": 1.2985160680298835e-06, "loss": 1.3212, "step": 2738 }, { "epoch": 0.84, "learning_rate": 1.2936087640819805e-06, "loss": 1.2141, "step": 2739 }, { "epoch": 0.84, "learning_rate": 1.2887101091255849e-06, "loss": 1.3545, "step": 2740 }, { "epoch": 0.84, "learning_rate": 1.2838201080270585e-06, "loss": 1.4226, "step": 2741 }, { "epoch": 0.84, "learning_rate": 1.2789387656441676e-06, "loss": 1.6109, "step": 2742 }, { "epoch": 0.84, "learning_rate": 1.2740660868260634e-06, "loss": 1.1917, "step": 2743 }, { "epoch": 0.84, "learning_rate": 1.269202076413305e-06, "loss": 1.3703, "step": 2744 }, { "epoch": 0.84, "learning_rate": 1.2643467392378383e-06, "loss": 1.4263, "step": 2745 }, { "epoch": 0.84, "learning_rate": 1.259500080122984e-06, "loss": 1.2747, "step": 2746 }, { "epoch": 0.85, "learning_rate": 1.2546621038834506e-06, "loss": 1.2362, "step": 2747 }, { "epoch": 0.85, "learning_rate": 1.2498328153253214e-06, "loss": 1.7445, "step": 2748 }, { "epoch": 0.85, "learning_rate": 1.2450122192460412e-06, "loss": 1.3055, "step": 2749 }, { "epoch": 0.85, "learning_rate": 1.2402003204344293e-06, "loss": 1.4602, "step": 2750 }, { "epoch": 0.85, "learning_rate": 1.2353971236706564e-06, "loss": 1.593, "step": 2751 }, { "epoch": 0.85, "learning_rate": 1.2306026337262544e-06, "loss": 1.3273, "step": 2752 }, { "epoch": 0.85, "learning_rate": 1.2258168553641058e-06, "loss": 1.5382, "step": 2753 }, { "epoch": 0.85, "learning_rate": 1.2210397933384344e-06, "loss": 1.8195, "step": 2754 }, { "epoch": 0.85, "learning_rate": 1.2162714523948104e-06, "loss": 1.542, "step": 2755 }, { "epoch": 0.85, "learning_rate": 1.2115118372701373e-06, "loss": 1.5063, "step": 2756 }, { "epoch": 0.85, "learning_rate": 1.206760952692647e-06, "loss": 1.1642, "step": 2757 }, { "epoch": 0.85, "learning_rate": 1.2020188033819092e-06, "loss": 1.1647, "step": 2758 }, { "epoch": 0.85, "learning_rate": 1.1972853940488017e-06, "loss": 1.2048, "step": 2759 }, { "epoch": 0.85, "learning_rate": 1.1925607293955322e-06, "loss": 0.9597, "step": 2760 }, { "epoch": 0.85, "learning_rate": 1.1878448141156162e-06, "loss": 1.2824, "step": 2761 }, { "epoch": 0.85, "learning_rate": 1.1831376528938742e-06, "loss": 1.0194, "step": 2762 }, { "epoch": 0.85, "learning_rate": 1.1784392504064378e-06, "loss": 1.1893, "step": 2763 }, { "epoch": 0.85, "learning_rate": 1.1737496113207336e-06, "loss": 1.2623, "step": 2764 }, { "epoch": 0.85, "learning_rate": 1.1690687402954792e-06, "loss": 1.0311, "step": 2765 }, { "epoch": 0.85, "learning_rate": 1.1643966419806907e-06, "loss": 1.508, "step": 2766 }, { "epoch": 0.85, "learning_rate": 1.1597333210176587e-06, "loss": 1.1493, "step": 2767 }, { "epoch": 0.85, "learning_rate": 1.1550787820389619e-06, "loss": 1.1789, "step": 2768 }, { "epoch": 0.85, "learning_rate": 1.1504330296684552e-06, "loss": 1.6571, "step": 2769 }, { "epoch": 0.85, "learning_rate": 1.1457960685212599e-06, "loss": 1.4267, "step": 2770 }, { "epoch": 0.85, "learning_rate": 1.1411679032037636e-06, "loss": 1.589, "step": 2771 }, { "epoch": 0.85, "learning_rate": 1.1365485383136255e-06, "loss": 1.3366, "step": 2772 }, { "epoch": 0.85, "learning_rate": 1.1319379784397499e-06, "loss": 1.5478, "step": 2773 }, { "epoch": 0.85, "learning_rate": 1.1273362281623045e-06, "loss": 1.5473, "step": 2774 }, { "epoch": 0.85, "learning_rate": 1.122743292052697e-06, "loss": 1.6997, "step": 2775 }, { "epoch": 0.85, "learning_rate": 1.1181591746735842e-06, "loss": 1.6264, "step": 2776 }, { "epoch": 0.85, "learning_rate": 1.1135838805788646e-06, "loss": 1.3311, "step": 2777 }, { "epoch": 0.85, "learning_rate": 1.1090174143136635e-06, "loss": 1.2912, "step": 2778 }, { "epoch": 0.86, "learning_rate": 1.1044597804143441e-06, "loss": 1.2467, "step": 2779 }, { "epoch": 0.86, "learning_rate": 1.099910983408493e-06, "loss": 1.4647, "step": 2780 }, { "epoch": 0.86, "learning_rate": 1.0953710278149155e-06, "loss": 1.3872, "step": 2781 }, { "epoch": 0.86, "learning_rate": 1.090839918143638e-06, "loss": 1.296, "step": 2782 }, { "epoch": 0.86, "learning_rate": 1.0863176588958957e-06, "loss": 1.3604, "step": 2783 }, { "epoch": 0.86, "learning_rate": 1.0818042545641349e-06, "loss": 1.4225, "step": 2784 }, { "epoch": 0.86, "learning_rate": 1.0772997096320049e-06, "loss": 1.3426, "step": 2785 }, { "epoch": 0.86, "learning_rate": 1.072804028574348e-06, "loss": 1.4228, "step": 2786 }, { "epoch": 0.86, "learning_rate": 1.0683172158572097e-06, "loss": 1.0379, "step": 2787 }, { "epoch": 0.86, "learning_rate": 1.0638392759378236e-06, "loss": 1.1639, "step": 2788 }, { "epoch": 0.86, "learning_rate": 1.0593702132646023e-06, "loss": 1.7137, "step": 2789 }, { "epoch": 0.86, "learning_rate": 1.0549100322771498e-06, "loss": 1.393, "step": 2790 }, { "epoch": 0.86, "learning_rate": 1.0504587374062392e-06, "loss": 1.4926, "step": 2791 }, { "epoch": 0.86, "learning_rate": 1.0460163330738182e-06, "loss": 1.5008, "step": 2792 }, { "epoch": 0.86, "learning_rate": 1.0415828236930059e-06, "loss": 1.5855, "step": 2793 }, { "epoch": 0.86, "learning_rate": 1.0371582136680792e-06, "loss": 1.285, "step": 2794 }, { "epoch": 0.86, "learning_rate": 1.0327425073944785e-06, "loss": 1.6029, "step": 2795 }, { "epoch": 0.86, "learning_rate": 1.028335709258801e-06, "loss": 1.0107, "step": 2796 }, { "epoch": 0.86, "learning_rate": 1.0239378236387864e-06, "loss": 1.4162, "step": 2797 }, { "epoch": 0.86, "learning_rate": 1.0195488549033305e-06, "loss": 1.5145, "step": 2798 }, { "epoch": 0.86, "learning_rate": 1.0151688074124645e-06, "loss": 1.6029, "step": 2799 }, { "epoch": 0.86, "learning_rate": 1.0107976855173574e-06, "loss": 1.3441, "step": 2800 }, { "epoch": 0.86, "learning_rate": 1.006435493560315e-06, "loss": 1.2033, "step": 2801 }, { "epoch": 0.86, "learning_rate": 1.0020822358747683e-06, "loss": 1.494, "step": 2802 }, { "epoch": 0.86, "learning_rate": 9.977379167852763e-07, "loss": 1.2489, "step": 2803 }, { "epoch": 0.86, "learning_rate": 9.934025406075176e-07, "loss": 1.6816, "step": 2804 }, { "epoch": 0.86, "learning_rate": 9.89076111648284e-07, "loss": 1.045, "step": 2805 }, { "epoch": 0.86, "learning_rate": 9.847586342054804e-07, "loss": 1.2116, "step": 2806 }, { "epoch": 0.86, "learning_rate": 9.804501125681243e-07, "loss": 1.435, "step": 2807 }, { "epoch": 0.86, "learning_rate": 9.76150551016326e-07, "loss": 1.3362, "step": 2808 }, { "epoch": 0.86, "learning_rate": 9.71859953821307e-07, "loss": 1.6681, "step": 2809 }, { "epoch": 0.86, "learning_rate": 9.675783252453718e-07, "loss": 1.2922, "step": 2810 }, { "epoch": 0.86, "learning_rate": 9.633056695419229e-07, "loss": 1.5096, "step": 2811 }, { "epoch": 0.87, "learning_rate": 9.590419909554516e-07, "loss": 1.3404, "step": 2812 }, { "epoch": 0.87, "learning_rate": 9.547872937215187e-07, "loss": 1.2368, "step": 2813 }, { "epoch": 0.87, "learning_rate": 9.505415820667751e-07, "loss": 1.2994, "step": 2814 }, { "epoch": 0.87, "learning_rate": 9.463048602089431e-07, "loss": 1.3717, "step": 2815 }, { "epoch": 0.87, "learning_rate": 9.420771323568079e-07, "loss": 1.2866, "step": 2816 }, { "epoch": 0.87, "learning_rate": 9.3785840271023e-07, "loss": 1.2937, "step": 2817 }, { "epoch": 0.87, "learning_rate": 9.336486754601204e-07, "loss": 1.3817, "step": 2818 }, { "epoch": 0.87, "learning_rate": 9.294479547884561e-07, "loss": 1.24, "step": 2819 }, { "epoch": 0.87, "learning_rate": 9.252562448682623e-07, "loss": 1.3247, "step": 2820 }, { "epoch": 0.87, "learning_rate": 9.210735498636125e-07, "loss": 1.4985, "step": 2821 }, { "epoch": 0.87, "learning_rate": 9.168998739296264e-07, "loss": 1.5737, "step": 2822 }, { "epoch": 0.87, "learning_rate": 9.127352212124663e-07, "loss": 1.2757, "step": 2823 }, { "epoch": 0.87, "learning_rate": 9.085795958493249e-07, "loss": 1.2902, "step": 2824 }, { "epoch": 0.87, "learning_rate": 9.044330019684311e-07, "loss": 1.1353, "step": 2825 }, { "epoch": 0.87, "learning_rate": 9.002954436890421e-07, "loss": 1.3566, "step": 2826 }, { "epoch": 0.87, "learning_rate": 8.961669251214356e-07, "loss": 1.3564, "step": 2827 }, { "epoch": 0.87, "learning_rate": 8.920474503669141e-07, "loss": 1.5027, "step": 2828 }, { "epoch": 0.87, "learning_rate": 8.879370235177909e-07, "loss": 1.6656, "step": 2829 }, { "epoch": 0.87, "learning_rate": 8.838356486573941e-07, "loss": 1.2436, "step": 2830 }, { "epoch": 0.87, "learning_rate": 8.797433298600622e-07, "loss": 1.6089, "step": 2831 }, { "epoch": 0.87, "learning_rate": 8.756600711911289e-07, "loss": 1.4268, "step": 2832 }, { "epoch": 0.87, "learning_rate": 8.715858767069374e-07, "loss": 1.304, "step": 2833 }, { "epoch": 0.87, "learning_rate": 8.675207504548178e-07, "loss": 1.3828, "step": 2834 }, { "epoch": 0.87, "learning_rate": 8.634646964730975e-07, "loss": 1.2387, "step": 2835 }, { "epoch": 0.87, "learning_rate": 8.594177187910913e-07, "loss": 1.4763, "step": 2836 }, { "epoch": 0.87, "learning_rate": 8.55379821429092e-07, "loss": 1.3914, "step": 2837 }, { "epoch": 0.87, "learning_rate": 8.513510083983778e-07, "loss": 1.1502, "step": 2838 }, { "epoch": 0.87, "learning_rate": 8.473312837012027e-07, "loss": 1.6727, "step": 2839 }, { "epoch": 0.87, "learning_rate": 8.433206513307878e-07, "loss": 1.0827, "step": 2840 }, { "epoch": 0.87, "learning_rate": 8.393191152713231e-07, "loss": 1.6531, "step": 2841 }, { "epoch": 0.87, "learning_rate": 8.353266794979653e-07, "loss": 1.4707, "step": 2842 }, { "epoch": 0.87, "learning_rate": 8.313433479768262e-07, "loss": 1.4687, "step": 2843 }, { "epoch": 0.88, "learning_rate": 8.273691246649795e-07, "loss": 1.4851, "step": 2844 }, { "epoch": 0.88, "learning_rate": 8.234040135104437e-07, "loss": 1.6681, "step": 2845 }, { "epoch": 0.88, "learning_rate": 8.194480184521891e-07, "loss": 1.5478, "step": 2846 }, { "epoch": 0.88, "learning_rate": 8.155011434201332e-07, "loss": 1.416, "step": 2847 }, { "epoch": 0.88, "learning_rate": 8.115633923351252e-07, "loss": 1.4117, "step": 2848 }, { "epoch": 0.88, "learning_rate": 8.076347691089592e-07, "loss": 1.3627, "step": 2849 }, { "epoch": 0.88, "learning_rate": 8.037152776443535e-07, "loss": 1.5196, "step": 2850 }, { "epoch": 0.88, "learning_rate": 7.998049218349624e-07, "loss": 1.433, "step": 2851 }, { "epoch": 0.88, "learning_rate": 7.959037055653617e-07, "loss": 1.1253, "step": 2852 }, { "epoch": 0.88, "learning_rate": 7.920116327110449e-07, "loss": 1.2896, "step": 2853 }, { "epoch": 0.88, "learning_rate": 7.88128707138428e-07, "loss": 1.0381, "step": 2854 }, { "epoch": 0.88, "learning_rate": 7.842549327048366e-07, "loss": 1.5005, "step": 2855 }, { "epoch": 0.88, "learning_rate": 7.803903132585023e-07, "loss": 1.5039, "step": 2856 }, { "epoch": 0.88, "learning_rate": 7.765348526385685e-07, "loss": 1.0744, "step": 2857 }, { "epoch": 0.88, "learning_rate": 7.726885546750796e-07, "loss": 1.219, "step": 2858 }, { "epoch": 0.88, "learning_rate": 7.688514231889721e-07, "loss": 1.2464, "step": 2859 }, { "epoch": 0.88, "learning_rate": 7.650234619920826e-07, "loss": 1.2983, "step": 2860 }, { "epoch": 0.88, "learning_rate": 7.612046748871327e-07, "loss": 1.2922, "step": 2861 }, { "epoch": 0.88, "learning_rate": 7.573950656677342e-07, "loss": 1.6027, "step": 2862 }, { "epoch": 0.88, "learning_rate": 7.535946381183812e-07, "loss": 1.1291, "step": 2863 }, { "epoch": 0.88, "learning_rate": 7.498033960144435e-07, "loss": 1.5354, "step": 2864 }, { "epoch": 0.88, "learning_rate": 7.460213431221719e-07, "loss": 1.1072, "step": 2865 }, { "epoch": 0.88, "learning_rate": 7.422484831986798e-07, "loss": 1.3677, "step": 2866 }, { "epoch": 0.88, "learning_rate": 7.384848199919558e-07, "loss": 1.9943, "step": 2867 }, { "epoch": 0.88, "learning_rate": 7.347303572408515e-07, "loss": 1.3345, "step": 2868 }, { "epoch": 0.88, "learning_rate": 7.309850986750766e-07, "loss": 1.1227, "step": 2869 }, { "epoch": 0.88, "learning_rate": 7.272490480151939e-07, "loss": 1.3734, "step": 2870 }, { "epoch": 0.88, "learning_rate": 7.23522208972628e-07, "loss": 1.6701, "step": 2871 }, { "epoch": 0.88, "learning_rate": 7.198045852496438e-07, "loss": 2.5635, "step": 2872 }, { "epoch": 0.88, "learning_rate": 7.160961805393574e-07, "loss": 1.7899, "step": 2873 }, { "epoch": 0.88, "learning_rate": 7.123969985257262e-07, "loss": 1.2196, "step": 2874 }, { "epoch": 0.88, "learning_rate": 7.087070428835397e-07, "loss": 1.4816, "step": 2875 }, { "epoch": 0.88, "learning_rate": 7.05026317278431e-07, "loss": 1.3796, "step": 2876 }, { "epoch": 0.89, "learning_rate": 7.013548253668556e-07, "loss": 1.7681, "step": 2877 }, { "epoch": 0.89, "learning_rate": 6.976925707961002e-07, "loss": 1.2791, "step": 2878 }, { "epoch": 0.89, "learning_rate": 6.940395572042791e-07, "loss": 1.3327, "step": 2879 }, { "epoch": 0.89, "learning_rate": 6.903957882203171e-07, "loss": 1.3877, "step": 2880 }, { "epoch": 0.89, "learning_rate": 6.867612674639634e-07, "loss": 1.1897, "step": 2881 }, { "epoch": 0.89, "learning_rate": 6.831359985457753e-07, "loss": 1.3656, "step": 2882 }, { "epoch": 0.89, "learning_rate": 6.79519985067123e-07, "loss": 1.3448, "step": 2883 }, { "epoch": 0.89, "learning_rate": 6.759132306201788e-07, "loss": 1.6175, "step": 2884 }, { "epoch": 0.89, "learning_rate": 6.723157387879164e-07, "loss": 1.2321, "step": 2885 }, { "epoch": 0.89, "learning_rate": 6.687275131441129e-07, "loss": 1.6447, "step": 2886 }, { "epoch": 0.89, "learning_rate": 6.651485572533379e-07, "loss": 1.2069, "step": 2887 }, { "epoch": 0.89, "learning_rate": 6.615788746709495e-07, "loss": 1.1779, "step": 2888 }, { "epoch": 0.89, "learning_rate": 6.580184689430958e-07, "loss": 1.3162, "step": 2889 }, { "epoch": 0.89, "learning_rate": 6.54467343606714e-07, "loss": 1.3883, "step": 2890 }, { "epoch": 0.89, "learning_rate": 6.509255021895111e-07, "loss": 1.4417, "step": 2891 }, { "epoch": 0.89, "learning_rate": 6.473929482099827e-07, "loss": 1.4512, "step": 2892 }, { "epoch": 0.89, "learning_rate": 6.438696851773906e-07, "loss": 1.3571, "step": 2893 }, { "epoch": 0.89, "learning_rate": 6.403557165917695e-07, "loss": 1.4242, "step": 2894 }, { "epoch": 0.89, "learning_rate": 6.368510459439248e-07, "loss": 1.3279, "step": 2895 }, { "epoch": 0.89, "learning_rate": 6.333556767154159e-07, "loss": 1.4234, "step": 2896 }, { "epoch": 0.89, "learning_rate": 6.298696123785719e-07, "loss": 1.3089, "step": 2897 }, { "epoch": 0.89, "learning_rate": 6.263928563964727e-07, "loss": 1.344, "step": 2898 }, { "epoch": 0.89, "learning_rate": 6.2292541222295e-07, "loss": 1.043, "step": 2899 }, { "epoch": 0.89, "learning_rate": 6.194672833025927e-07, "loss": 1.3056, "step": 2900 }, { "epoch": 0.89, "learning_rate": 6.160184730707264e-07, "loss": 1.5247, "step": 2901 }, { "epoch": 0.89, "learning_rate": 6.125789849534269e-07, "loss": 2.1458, "step": 2902 }, { "epoch": 0.89, "learning_rate": 6.091488223675058e-07, "loss": 1.2948, "step": 2903 }, { "epoch": 0.89, "learning_rate": 6.057279887205092e-07, "loss": 1.2128, "step": 2904 }, { "epoch": 0.89, "learning_rate": 6.023164874107202e-07, "loss": 1.5385, "step": 2905 }, { "epoch": 0.89, "learning_rate": 5.989143218271487e-07, "loss": 1.3961, "step": 2906 }, { "epoch": 0.89, "learning_rate": 5.95521495349527e-07, "loss": 1.3153, "step": 2907 }, { "epoch": 0.89, "learning_rate": 5.921380113483188e-07, "loss": 1.5059, "step": 2908 }, { "epoch": 0.9, "learning_rate": 5.887638731846967e-07, "loss": 1.2553, "step": 2909 }, { "epoch": 0.9, "learning_rate": 5.85399084210555e-07, "loss": 1.5093, "step": 2910 }, { "epoch": 0.9, "learning_rate": 5.820436477685021e-07, "loss": 1.5094, "step": 2911 }, { "epoch": 0.9, "learning_rate": 5.786975671918482e-07, "loss": 1.4329, "step": 2912 }, { "epoch": 0.9, "learning_rate": 5.753608458046145e-07, "loss": 1.3455, "step": 2913 }, { "epoch": 0.9, "learning_rate": 5.720334869215282e-07, "loss": 1.3079, "step": 2914 }, { "epoch": 0.9, "learning_rate": 5.687154938480055e-07, "loss": 1.3543, "step": 2915 }, { "epoch": 0.9, "learning_rate": 5.654068698801696e-07, "loss": 1.3558, "step": 2916 }, { "epoch": 0.9, "learning_rate": 5.62107618304828e-07, "loss": 1.3352, "step": 2917 }, { "epoch": 0.9, "learning_rate": 5.588177423994812e-07, "loss": 1.4507, "step": 2918 }, { "epoch": 0.9, "learning_rate": 5.555372454323182e-07, "loss": 1.3969, "step": 2919 }, { "epoch": 0.9, "learning_rate": 5.522661306622046e-07, "loss": 1.4068, "step": 2920 }, { "epoch": 0.9, "learning_rate": 5.490044013386919e-07, "loss": 1.2338, "step": 2921 }, { "epoch": 0.9, "learning_rate": 5.45752060702005e-07, "loss": 1.3456, "step": 2922 }, { "epoch": 0.9, "learning_rate": 5.425091119830417e-07, "loss": 1.644, "step": 2923 }, { "epoch": 0.9, "learning_rate": 5.392755584033715e-07, "loss": 2.0859, "step": 2924 }, { "epoch": 0.9, "learning_rate": 5.360514031752284e-07, "loss": 1.3264, "step": 2925 }, { "epoch": 0.9, "learning_rate": 5.328366495015113e-07, "loss": 1.3682, "step": 2926 }, { "epoch": 0.9, "learning_rate": 5.296313005757814e-07, "loss": 1.5484, "step": 2927 }, { "epoch": 0.9, "learning_rate": 5.264353595822535e-07, "loss": 1.6126, "step": 2928 }, { "epoch": 0.9, "learning_rate": 5.232488296957983e-07, "loss": 1.3289, "step": 2929 }, { "epoch": 0.9, "learning_rate": 5.200717140819389e-07, "loss": 1.3832, "step": 2930 }, { "epoch": 0.9, "learning_rate": 5.169040158968431e-07, "loss": 1.5822, "step": 2931 }, { "epoch": 0.9, "learning_rate": 5.137457382873257e-07, "loss": 1.3771, "step": 2932 }, { "epoch": 0.9, "learning_rate": 5.105968843908449e-07, "loss": 1.2385, "step": 2933 }, { "epoch": 0.9, "learning_rate": 5.074574573354896e-07, "loss": 1.606, "step": 2934 }, { "epoch": 0.9, "learning_rate": 5.043274602399939e-07, "loss": 1.4042, "step": 2935 }, { "epoch": 0.9, "learning_rate": 5.012068962137151e-07, "loss": 1.821, "step": 2936 }, { "epoch": 0.9, "learning_rate": 4.980957683566468e-07, "loss": 1.9755, "step": 2937 }, { "epoch": 0.9, "learning_rate": 4.949940797594066e-07, "loss": 1.5351, "step": 2938 }, { "epoch": 0.9, "learning_rate": 4.919018335032322e-07, "loss": 1.6128, "step": 2939 }, { "epoch": 0.9, "learning_rate": 4.888190326599829e-07, "loss": 1.2311, "step": 2940 }, { "epoch": 0.9, "learning_rate": 4.857456802921368e-07, "loss": 1.6116, "step": 2941 }, { "epoch": 0.91, "learning_rate": 4.826817794527828e-07, "loss": 1.2975, "step": 2942 }, { "epoch": 0.91, "learning_rate": 4.796273331856227e-07, "loss": 1.1241, "step": 2943 }, { "epoch": 0.91, "learning_rate": 4.76582344524964e-07, "loss": 1.291, "step": 2944 }, { "epoch": 0.91, "learning_rate": 4.735468164957213e-07, "loss": 1.3337, "step": 2945 }, { "epoch": 0.91, "learning_rate": 4.705207521134125e-07, "loss": 1.7244, "step": 2946 }, { "epoch": 0.91, "learning_rate": 4.6750415438414745e-07, "loss": 1.3703, "step": 2947 }, { "epoch": 0.91, "learning_rate": 4.644970263046367e-07, "loss": 1.4727, "step": 2948 }, { "epoch": 0.91, "learning_rate": 4.614993708621862e-07, "loss": 1.441, "step": 2949 }, { "epoch": 0.91, "learning_rate": 4.58511191034684e-07, "loss": 1.1016, "step": 2950 }, { "epoch": 0.91, "learning_rate": 4.555324897906133e-07, "loss": 1.6036, "step": 2951 }, { "epoch": 0.91, "learning_rate": 4.525632700890348e-07, "loss": 1.3771, "step": 2952 }, { "epoch": 0.91, "learning_rate": 4.4960353487959354e-07, "loss": 1.7475, "step": 2953 }, { "epoch": 0.91, "learning_rate": 4.46653287102512e-07, "loss": 1.4229, "step": 2954 }, { "epoch": 0.91, "learning_rate": 4.437125296885847e-07, "loss": 1.431, "step": 2955 }, { "epoch": 0.91, "learning_rate": 4.407812655591848e-07, "loss": 1.3778, "step": 2956 }, { "epoch": 0.91, "learning_rate": 4.378594976262496e-07, "loss": 1.3583, "step": 2957 }, { "epoch": 0.91, "learning_rate": 4.3494722879228404e-07, "loss": 1.4101, "step": 2958 }, { "epoch": 0.91, "learning_rate": 4.320444619503583e-07, "loss": 1.84, "step": 2959 }, { "epoch": 0.91, "learning_rate": 4.2915119998409914e-07, "loss": 1.3871, "step": 2960 }, { "epoch": 0.91, "learning_rate": 4.262674457676952e-07, "loss": 1.3767, "step": 2961 }, { "epoch": 0.91, "learning_rate": 4.233932021658904e-07, "loss": 1.7263, "step": 2962 }, { "epoch": 0.91, "learning_rate": 4.205284720339764e-07, "loss": 1.4118, "step": 2963 }, { "epoch": 0.91, "learning_rate": 4.176732582177967e-07, "loss": 1.3906, "step": 2964 }, { "epoch": 0.91, "learning_rate": 4.1482756355374463e-07, "loss": 1.4991, "step": 2965 }, { "epoch": 0.91, "learning_rate": 4.1199139086874895e-07, "loss": 1.3459, "step": 2966 }, { "epoch": 0.91, "learning_rate": 4.0916474298028694e-07, "loss": 1.1746, "step": 2967 }, { "epoch": 0.91, "learning_rate": 4.063476226963703e-07, "loss": 1.2817, "step": 2968 }, { "epoch": 0.91, "learning_rate": 4.0354003281554586e-07, "loss": 1.5201, "step": 2969 }, { "epoch": 0.91, "learning_rate": 4.007419761268938e-07, "loss": 1.0179, "step": 2970 }, { "epoch": 0.91, "learning_rate": 3.9795345541002395e-07, "loss": 1.7437, "step": 2971 }, { "epoch": 0.91, "learning_rate": 3.951744734350715e-07, "loss": 1.2231, "step": 2972 }, { "epoch": 0.91, "learning_rate": 3.924050329627005e-07, "loss": 1.5422, "step": 2973 }, { "epoch": 0.92, "learning_rate": 3.896451367440901e-07, "loss": 1.4199, "step": 2974 }, { "epoch": 0.92, "learning_rate": 3.8689478752094167e-07, "loss": 1.5682, "step": 2975 }, { "epoch": 0.92, "learning_rate": 3.841539880254708e-07, "loss": 1.7482, "step": 2976 }, { "epoch": 0.92, "learning_rate": 3.814227409804094e-07, "loss": 1.2629, "step": 2977 }, { "epoch": 0.92, "learning_rate": 3.787010490989973e-07, "loss": 1.4626, "step": 2978 }, { "epoch": 0.92, "learning_rate": 3.759889150849805e-07, "loss": 1.4805, "step": 2979 }, { "epoch": 0.92, "learning_rate": 3.73286341632616e-07, "loss": 1.18, "step": 2980 }, { "epoch": 0.92, "learning_rate": 3.7059333142665854e-07, "loss": 1.2404, "step": 2981 }, { "epoch": 0.92, "learning_rate": 3.6790988714236144e-07, "loss": 1.4339, "step": 2982 }, { "epoch": 0.92, "learning_rate": 3.6523601144548003e-07, "loss": 1.4648, "step": 2983 }, { "epoch": 0.92, "learning_rate": 3.6257170699226076e-07, "loss": 1.3987, "step": 2984 }, { "epoch": 0.92, "learning_rate": 3.599169764294419e-07, "loss": 1.2372, "step": 2985 }, { "epoch": 0.92, "learning_rate": 3.5727182239425393e-07, "loss": 1.3294, "step": 2986 }, { "epoch": 0.92, "learning_rate": 3.546362475144105e-07, "loss": 1.2964, "step": 2987 }, { "epoch": 0.92, "learning_rate": 3.5201025440810943e-07, "loss": 1.4169, "step": 2988 }, { "epoch": 0.92, "learning_rate": 3.49393845684034e-07, "loss": 1.2522, "step": 2989 }, { "epoch": 0.92, "learning_rate": 3.467870239413407e-07, "loss": 1.6116, "step": 2990 }, { "epoch": 0.92, "learning_rate": 3.441897917696679e-07, "loss": 1.6494, "step": 2991 }, { "epoch": 0.92, "learning_rate": 3.416021517491208e-07, "loss": 1.2379, "step": 2992 }, { "epoch": 0.92, "learning_rate": 3.3902410645028197e-07, "loss": 1.3141, "step": 2993 }, { "epoch": 0.92, "learning_rate": 3.364556584342005e-07, "loss": 1.4326, "step": 2994 }, { "epoch": 0.92, "learning_rate": 3.3389681025239095e-07, "loss": 1.4313, "step": 2995 }, { "epoch": 0.92, "learning_rate": 3.3134756444683e-07, "loss": 1.3092, "step": 2996 }, { "epoch": 0.92, "learning_rate": 3.288079235499575e-07, "loss": 1.5026, "step": 2997 }, { "epoch": 0.92, "learning_rate": 3.262778900846708e-07, "loss": 1.538, "step": 2998 }, { "epoch": 0.92, "learning_rate": 3.2375746656432284e-07, "loss": 1.1904, "step": 2999 }, { "epoch": 0.92, "learning_rate": 3.212466554927229e-07, "loss": 1.2712, "step": 3000 }, { "epoch": 0.92, "learning_rate": 1.979787329680057e-05, "loss": 1.2696, "step": 3001 }, { "epoch": 0.92, "learning_rate": 1.9797673898308793e-05, "loss": 1.2591, "step": 3002 }, { "epoch": 0.92, "learning_rate": 1.9797474402517084e-05, "loss": 1.5343, "step": 3003 }, { "epoch": 0.92, "learning_rate": 1.9797274809427425e-05, "loss": 1.137, "step": 3004 }, { "epoch": 0.92, "learning_rate": 1.97970751190418e-05, "loss": 1.1288, "step": 3005 }, { "epoch": 0.92, "learning_rate": 1.9796875331362188e-05, "loss": 1.2623, "step": 3006 }, { "epoch": 0.93, "learning_rate": 1.9796675446390573e-05, "loss": 1.4993, "step": 3007 }, { "epoch": 0.93, "learning_rate": 1.9796475464128943e-05, "loss": 1.7835, "step": 3008 }, { "epoch": 0.93, "learning_rate": 1.979627538457928e-05, "loss": 1.19, "step": 3009 }, { "epoch": 0.93, "learning_rate": 1.979607520774358e-05, "loss": 1.4252, "step": 3010 }, { "epoch": 0.93, "learning_rate": 1.979587493362382e-05, "loss": 1.3566, "step": 3011 }, { "epoch": 0.93, "learning_rate": 1.9795674562221995e-05, "loss": 1.3139, "step": 3012 }, { "epoch": 0.93, "learning_rate": 1.9795474093540097e-05, "loss": 1.7944, "step": 3013 }, { "epoch": 0.93, "learning_rate": 1.979527352758011e-05, "loss": 1.2389, "step": 3014 }, { "epoch": 0.93, "learning_rate": 1.979507286434403e-05, "loss": 1.2334, "step": 3015 }, { "epoch": 0.93, "learning_rate": 1.9794872103833848e-05, "loss": 1.3064, "step": 3016 }, { "epoch": 0.93, "learning_rate": 1.979467124605156e-05, "loss": 1.3675, "step": 3017 }, { "epoch": 0.93, "learning_rate": 1.9794470290999158e-05, "loss": 1.083, "step": 3018 }, { "epoch": 0.93, "learning_rate": 1.9794269238678644e-05, "loss": 1.3517, "step": 3019 }, { "epoch": 0.93, "learning_rate": 1.9794068089092006e-05, "loss": 1.2288, "step": 3020 }, { "epoch": 0.93, "learning_rate": 1.9793866842241245e-05, "loss": 1.8444, "step": 3021 }, { "epoch": 0.93, "learning_rate": 1.979366549812836e-05, "loss": 1.7898, "step": 3022 }, { "epoch": 0.93, "learning_rate": 1.9793464056755352e-05, "loss": 1.7039, "step": 3023 }, { "epoch": 0.93, "learning_rate": 1.979326251812422e-05, "loss": 1.3376, "step": 3024 }, { "epoch": 0.93, "learning_rate": 1.9793060882236964e-05, "loss": 1.4599, "step": 3025 }, { "epoch": 0.93, "learning_rate": 1.979285914909559e-05, "loss": 1.4507, "step": 3026 }, { "epoch": 0.93, "learning_rate": 1.9792657318702096e-05, "loss": 1.5273, "step": 3027 }, { "epoch": 0.93, "learning_rate": 1.979245539105849e-05, "loss": 1.7383, "step": 3028 }, { "epoch": 0.93, "learning_rate": 1.979225336616678e-05, "loss": 1.3046, "step": 3029 }, { "epoch": 0.93, "learning_rate": 1.9792051244028968e-05, "loss": 1.3494, "step": 3030 }, { "epoch": 0.93, "learning_rate": 1.9791849024647063e-05, "loss": 1.2686, "step": 3031 }, { "epoch": 0.93, "learning_rate": 1.979164670802307e-05, "loss": 1.4004, "step": 3032 }, { "epoch": 0.93, "learning_rate": 1.9791444294159007e-05, "loss": 1.6209, "step": 3033 }, { "epoch": 0.93, "learning_rate": 1.9791241783056874e-05, "loss": 1.3192, "step": 3034 }, { "epoch": 0.93, "learning_rate": 1.9791039174718686e-05, "loss": 1.5667, "step": 3035 }, { "epoch": 0.93, "learning_rate": 1.9790836469146452e-05, "loss": 2.0347, "step": 3036 }, { "epoch": 0.93, "learning_rate": 1.9790633666342195e-05, "loss": 1.6828, "step": 3037 }, { "epoch": 0.93, "learning_rate": 1.979043076630792e-05, "loss": 1.5033, "step": 3038 }, { "epoch": 0.94, "learning_rate": 1.979022776904564e-05, "loss": 1.6181, "step": 3039 }, { "epoch": 0.94, "learning_rate": 1.9790024674557382e-05, "loss": 1.5981, "step": 3040 }, { "epoch": 0.94, "learning_rate": 1.9789821482845154e-05, "loss": 1.5773, "step": 3041 }, { "epoch": 0.94, "learning_rate": 1.9789618193910972e-05, "loss": 1.2607, "step": 3042 }, { "epoch": 0.94, "learning_rate": 1.9789414807756865e-05, "loss": 1.0358, "step": 3043 }, { "epoch": 0.94, "learning_rate": 1.9789211324384844e-05, "loss": 1.7485, "step": 3044 }, { "epoch": 0.94, "learning_rate": 1.9789007743796933e-05, "loss": 1.5661, "step": 3045 }, { "epoch": 0.94, "learning_rate": 1.9788804065995153e-05, "loss": 1.5574, "step": 3046 }, { "epoch": 0.94, "learning_rate": 1.9788600290981525e-05, "loss": 1.681, "step": 3047 }, { "epoch": 0.94, "learning_rate": 1.9788396418758077e-05, "loss": 1.9243, "step": 3048 }, { "epoch": 0.94, "learning_rate": 1.978819244932683e-05, "loss": 1.4108, "step": 3049 }, { "epoch": 0.94, "learning_rate": 1.9787988382689812e-05, "loss": 1.8297, "step": 3050 }, { "epoch": 0.94, "learning_rate": 1.9787784218849048e-05, "loss": 1.4147, "step": 3051 }, { "epoch": 0.94, "learning_rate": 1.9787579957806567e-05, "loss": 2.2936, "step": 3052 }, { "epoch": 0.94, "learning_rate": 1.97873755995644e-05, "loss": 1.8234, "step": 3053 }, { "epoch": 0.94, "learning_rate": 1.9787171144124567e-05, "loss": 1.3654, "step": 3054 }, { "epoch": 0.94, "learning_rate": 1.9786966591489107e-05, "loss": 1.5912, "step": 3055 }, { "epoch": 0.94, "learning_rate": 1.9786761941660044e-05, "loss": 1.8558, "step": 3056 }, { "epoch": 0.94, "learning_rate": 1.9786557194639423e-05, "loss": 1.3472, "step": 3057 }, { "epoch": 0.94, "learning_rate": 1.9786352350429263e-05, "loss": 1.4876, "step": 3058 }, { "epoch": 0.94, "learning_rate": 1.978614740903161e-05, "loss": 1.4986, "step": 3059 }, { "epoch": 0.94, "learning_rate": 1.978594237044849e-05, "loss": 1.5191, "step": 3060 }, { "epoch": 0.94, "learning_rate": 1.9785737234681944e-05, "loss": 1.3431, "step": 3061 }, { "epoch": 0.94, "learning_rate": 1.978553200173401e-05, "loss": 1.5269, "step": 3062 }, { "epoch": 0.94, "learning_rate": 1.9785326671606725e-05, "loss": 1.5521, "step": 3063 }, { "epoch": 0.94, "learning_rate": 1.9785121244302126e-05, "loss": 1.6298, "step": 3064 }, { "epoch": 0.94, "learning_rate": 1.9784915719822255e-05, "loss": 1.6754, "step": 3065 }, { "epoch": 0.94, "learning_rate": 1.9784710098169155e-05, "loss": 1.647, "step": 3066 }, { "epoch": 0.94, "learning_rate": 1.9784504379344862e-05, "loss": 1.6559, "step": 3067 }, { "epoch": 0.94, "learning_rate": 1.9784298563351426e-05, "loss": 1.2047, "step": 3068 }, { "epoch": 0.94, "learning_rate": 1.9784092650190887e-05, "loss": 1.3296, "step": 3069 }, { "epoch": 0.94, "learning_rate": 1.9783886639865287e-05, "loss": 1.5587, "step": 3070 }, { "epoch": 0.94, "learning_rate": 1.9783680532376684e-05, "loss": 1.6363, "step": 3071 }, { "epoch": 0.95, "learning_rate": 1.978347432772711e-05, "loss": 2.1019, "step": 3072 }, { "epoch": 0.95, "learning_rate": 1.9783268025918622e-05, "loss": 1.3106, "step": 3073 }, { "epoch": 0.95, "learning_rate": 1.9783061626953263e-05, "loss": 1.852, "step": 3074 }, { "epoch": 0.95, "learning_rate": 1.978285513083309e-05, "loss": 1.5175, "step": 3075 }, { "epoch": 0.95, "learning_rate": 1.9782648537560143e-05, "loss": 1.5917, "step": 3076 }, { "epoch": 0.95, "learning_rate": 1.9782441847136483e-05, "loss": 1.6557, "step": 3077 }, { "epoch": 0.95, "learning_rate": 1.9782235059564164e-05, "loss": 1.2773, "step": 3078 }, { "epoch": 0.95, "learning_rate": 1.978202817484523e-05, "loss": 1.5901, "step": 3079 }, { "epoch": 0.95, "learning_rate": 1.9781821192981744e-05, "loss": 1.5741, "step": 3080 }, { "epoch": 0.95, "learning_rate": 1.9781614113975756e-05, "loss": 1.41, "step": 3081 }, { "epoch": 0.95, "learning_rate": 1.9781406937829326e-05, "loss": 1.3852, "step": 3082 }, { "epoch": 0.95, "learning_rate": 1.978119966454451e-05, "loss": 1.3436, "step": 3083 }, { "epoch": 0.95, "learning_rate": 1.9780992294123365e-05, "loss": 1.4239, "step": 3084 }, { "epoch": 0.95, "learning_rate": 1.9780784826567955e-05, "loss": 1.2628, "step": 3085 }, { "epoch": 0.95, "learning_rate": 1.9780577261880336e-05, "loss": 1.491, "step": 3086 }, { "epoch": 0.95, "learning_rate": 1.978036960006257e-05, "loss": 1.5771, "step": 3087 }, { "epoch": 0.95, "learning_rate": 1.978036960006257e-05, "loss": 1.9535, "step": 3088 }, { "epoch": 0.95, "learning_rate": 1.9780161841116724e-05, "loss": 1.5996, "step": 3089 }, { "epoch": 0.95, "learning_rate": 1.9779953985044855e-05, "loss": 1.483, "step": 3090 }, { "epoch": 0.95, "learning_rate": 1.9779746031849027e-05, "loss": 1.3525, "step": 3091 }, { "epoch": 0.95, "learning_rate": 1.977953798153131e-05, "loss": 1.6862, "step": 3092 }, { "epoch": 0.95, "learning_rate": 1.977932983409377e-05, "loss": 1.7067, "step": 3093 }, { "epoch": 0.95, "learning_rate": 1.9779121589538464e-05, "loss": 1.4338, "step": 3094 }, { "epoch": 0.95, "learning_rate": 1.9778913247867475e-05, "loss": 1.4013, "step": 3095 }, { "epoch": 0.95, "learning_rate": 1.9778704809082864e-05, "loss": 1.8228, "step": 3096 }, { "epoch": 0.95, "learning_rate": 1.97784962731867e-05, "loss": 1.4885, "step": 3097 }, { "epoch": 0.95, "learning_rate": 1.977828764018106e-05, "loss": 1.6163, "step": 3098 }, { "epoch": 0.95, "learning_rate": 1.9778078910068007e-05, "loss": 1.5627, "step": 3099 }, { "epoch": 0.95, "learning_rate": 1.977787008284962e-05, "loss": 1.3983, "step": 3100 }, { "epoch": 0.95, "learning_rate": 1.977766115852797e-05, "loss": 1.6671, "step": 3101 }, { "epoch": 0.95, "learning_rate": 1.9777452137105134e-05, "loss": 1.5304, "step": 3102 }, { "epoch": 0.95, "learning_rate": 1.9777243018583193e-05, "loss": 1.5117, "step": 3103 }, { "epoch": 0.96, "learning_rate": 1.9777033802964212e-05, "loss": 1.4577, "step": 3104 }, { "epoch": 0.96, "learning_rate": 1.9776824490250275e-05, "loss": 1.4492, "step": 3105 }, { "epoch": 0.96, "learning_rate": 1.9776615080443462e-05, "loss": 1.4854, "step": 3106 }, { "epoch": 0.96, "learning_rate": 1.977640557354585e-05, "loss": 1.2281, "step": 3107 }, { "epoch": 0.96, "learning_rate": 1.9776195969559518e-05, "loss": 1.5689, "step": 3108 }, { "epoch": 0.96, "learning_rate": 1.9775986268486554e-05, "loss": 1.1649, "step": 3109 }, { "epoch": 0.96, "learning_rate": 1.9775776470329033e-05, "loss": 1.3965, "step": 3110 }, { "epoch": 0.96, "learning_rate": 1.9775566575089043e-05, "loss": 1.7665, "step": 3111 }, { "epoch": 0.96, "learning_rate": 1.977535658276867e-05, "loss": 1.6678, "step": 3112 }, { "epoch": 0.96, "learning_rate": 1.9775146493369996e-05, "loss": 1.4842, "step": 3113 }, { "epoch": 0.96, "learning_rate": 1.9774936306895105e-05, "loss": 2.534, "step": 3114 }, { "epoch": 0.96, "learning_rate": 1.977472602334609e-05, "loss": 1.527, "step": 3115 }, { "epoch": 0.96, "learning_rate": 1.977451564272504e-05, "loss": 1.4043, "step": 3116 }, { "epoch": 0.96, "learning_rate": 1.9774305165034034e-05, "loss": 1.5708, "step": 3117 }, { "epoch": 0.96, "learning_rate": 1.977409459027518e-05, "loss": 1.3342, "step": 3118 }, { "epoch": 0.96, "learning_rate": 1.9773883918450548e-05, "loss": 1.5206, "step": 3119 }, { "epoch": 0.96, "learning_rate": 1.9773673149562244e-05, "loss": 1.5813, "step": 3120 }, { "epoch": 0.96, "learning_rate": 1.9773462283612363e-05, "loss": 2.7623, "step": 3121 }, { "epoch": 0.96, "learning_rate": 1.977325132060299e-05, "loss": 1.1793, "step": 3122 }, { "epoch": 0.96, "learning_rate": 1.9773040260536223e-05, "loss": 1.5447, "step": 3123 }, { "epoch": 0.96, "learning_rate": 1.977282910341416e-05, "loss": 1.2831, "step": 3124 }, { "epoch": 0.96, "learning_rate": 1.9772617849238896e-05, "loss": 1.4406, "step": 3125 }, { "epoch": 0.96, "learning_rate": 1.977240649801253e-05, "loss": 1.781, "step": 3126 }, { "epoch": 0.96, "learning_rate": 1.9772195049737165e-05, "loss": 1.7942, "step": 3127 }, { "epoch": 0.96, "learning_rate": 1.9771983504414893e-05, "loss": 1.5272, "step": 3128 }, { "epoch": 0.96, "learning_rate": 1.9771771862047823e-05, "loss": 1.4193, "step": 3129 }, { "epoch": 0.96, "learning_rate": 1.9771560122638048e-05, "loss": 1.2278, "step": 3130 }, { "epoch": 0.96, "learning_rate": 1.9771348286187675e-05, "loss": 1.5204, "step": 3131 }, { "epoch": 0.96, "learning_rate": 1.977113635269881e-05, "loss": 1.721, "step": 3132 }, { "epoch": 0.96, "learning_rate": 1.9770924322173555e-05, "loss": 1.5089, "step": 3133 }, { "epoch": 0.96, "learning_rate": 1.9770712194614015e-05, "loss": 1.4178, "step": 3134 }, { "epoch": 0.96, "learning_rate": 1.97704999700223e-05, "loss": 1.5056, "step": 3135 }, { "epoch": 0.96, "learning_rate": 1.9770287648400516e-05, "loss": 1.6385, "step": 3136 }, { "epoch": 0.97, "learning_rate": 1.9770075229750772e-05, "loss": 1.4993, "step": 3137 }, { "epoch": 0.97, "learning_rate": 1.9769862714075173e-05, "loss": 1.4258, "step": 3138 }, { "epoch": 0.97, "learning_rate": 1.9769650101375835e-05, "loss": 1.3308, "step": 3139 }, { "epoch": 0.97, "learning_rate": 1.9769437391654867e-05, "loss": 1.6151, "step": 3140 }, { "epoch": 0.97, "learning_rate": 1.9769224584914383e-05, "loss": 1.5369, "step": 3141 }, { "epoch": 0.97, "learning_rate": 1.9769011681156493e-05, "loss": 1.891, "step": 3142 }, { "epoch": 0.97, "learning_rate": 1.9768798680383315e-05, "loss": 1.7519, "step": 3143 }, { "epoch": 0.97, "learning_rate": 1.9768585582596964e-05, "loss": 1.1969, "step": 3144 }, { "epoch": 0.97, "learning_rate": 1.9768372387799556e-05, "loss": 1.7292, "step": 3145 }, { "epoch": 0.97, "learning_rate": 1.9768159095993207e-05, "loss": 1.2132, "step": 3146 }, { "epoch": 0.97, "learning_rate": 1.9767945707180033e-05, "loss": 1.5751, "step": 3147 }, { "epoch": 0.97, "learning_rate": 1.976773222136216e-05, "loss": 1.4277, "step": 3148 }, { "epoch": 0.97, "learning_rate": 1.97675186385417e-05, "loss": 1.505, "step": 3149 }, { "epoch": 0.97, "learning_rate": 1.9767304958720778e-05, "loss": 1.4893, "step": 3150 }, { "epoch": 0.97, "learning_rate": 1.9767091181901516e-05, "loss": 1.4271, "step": 3151 }, { "epoch": 0.97, "learning_rate": 1.9766877308086038e-05, "loss": 2.9101, "step": 3152 }, { "epoch": 0.97, "learning_rate": 1.9766663337276465e-05, "loss": 1.5278, "step": 3153 }, { "epoch": 0.97, "learning_rate": 1.9766449269474925e-05, "loss": 1.3757, "step": 3154 }, { "epoch": 0.97, "learning_rate": 1.9766235104683545e-05, "loss": 1.5958, "step": 3155 }, { "epoch": 0.97, "learning_rate": 1.976602084290445e-05, "loss": 1.8901, "step": 3156 }, { "epoch": 0.97, "learning_rate": 1.9765806484139763e-05, "loss": 1.6533, "step": 3157 }, { "epoch": 0.97, "learning_rate": 1.9765592028391617e-05, "loss": 1.3902, "step": 3158 }, { "epoch": 0.97, "learning_rate": 1.9765377475662145e-05, "loss": 1.5233, "step": 3159 }, { "epoch": 0.97, "learning_rate": 1.9765162825953473e-05, "loss": 1.3756, "step": 3160 }, { "epoch": 0.97, "learning_rate": 1.9764948079267738e-05, "loss": 1.3883, "step": 3161 }, { "epoch": 0.97, "learning_rate": 1.9764733235607064e-05, "loss": 1.5547, "step": 3162 }, { "epoch": 0.97, "learning_rate": 1.9764518294973592e-05, "loss": 1.8963, "step": 3163 }, { "epoch": 0.97, "learning_rate": 1.976430325736945e-05, "loss": 1.7707, "step": 3164 }, { "epoch": 0.97, "learning_rate": 1.9764088122796785e-05, "loss": 1.4518, "step": 3165 }, { "epoch": 0.97, "learning_rate": 1.9763872891257722e-05, "loss": 2.1467, "step": 3166 }, { "epoch": 0.97, "learning_rate": 1.9763657562754403e-05, "loss": 1.3611, "step": 3167 }, { "epoch": 0.97, "learning_rate": 1.9763442137288966e-05, "loss": 1.262, "step": 3168 }, { "epoch": 0.98, "learning_rate": 1.976322661486355e-05, "loss": 1.7065, "step": 3169 }, { "epoch": 0.98, "learning_rate": 1.9763010995480297e-05, "loss": 1.5953, "step": 3170 }, { "epoch": 0.98, "learning_rate": 1.9762795279141344e-05, "loss": 1.4536, "step": 3171 }, { "epoch": 0.98, "learning_rate": 1.9762579465848838e-05, "loss": 1.6124, "step": 3172 }, { "epoch": 0.98, "learning_rate": 1.976236355560492e-05, "loss": 1.4873, "step": 3173 }, { "epoch": 0.98, "learning_rate": 1.9762147548411736e-05, "loss": 1.4208, "step": 3174 }, { "epoch": 0.98, "learning_rate": 1.976193144427143e-05, "loss": 1.6065, "step": 3175 }, { "epoch": 0.98, "learning_rate": 1.9761715243186146e-05, "loss": 1.5742, "step": 3176 }, { "epoch": 0.98, "learning_rate": 1.9761498945158033e-05, "loss": 1.964, "step": 3177 }, { "epoch": 0.98, "learning_rate": 1.976128255018924e-05, "loss": 1.3077, "step": 3178 }, { "epoch": 0.98, "learning_rate": 1.9761066058281914e-05, "loss": 1.474, "step": 3179 }, { "epoch": 0.98, "learning_rate": 1.976084946943821e-05, "loss": 1.5432, "step": 3180 }, { "epoch": 0.98, "learning_rate": 1.976063278366027e-05, "loss": 1.8503, "step": 3181 }, { "epoch": 0.98, "learning_rate": 1.9760416000950254e-05, "loss": 1.5418, "step": 3182 }, { "epoch": 0.98, "learning_rate": 1.976019912131031e-05, "loss": 1.3869, "step": 3183 }, { "epoch": 0.98, "learning_rate": 1.975998214474259e-05, "loss": 1.3705, "step": 3184 }, { "epoch": 0.98, "learning_rate": 1.9759765071249258e-05, "loss": 1.3096, "step": 3185 }, { "epoch": 0.98, "learning_rate": 1.9759547900832463e-05, "loss": 1.3988, "step": 3186 }, { "epoch": 0.98, "learning_rate": 1.975933063349436e-05, "loss": 1.5076, "step": 3187 }, { "epoch": 0.98, "learning_rate": 1.975911326923711e-05, "loss": 1.7398, "step": 3188 }, { "epoch": 0.98, "learning_rate": 1.9758895808062873e-05, "loss": 1.5889, "step": 3189 }, { "epoch": 0.98, "learning_rate": 1.9758678249973803e-05, "loss": 1.2679, "step": 3190 }, { "epoch": 0.98, "learning_rate": 1.9758460594972068e-05, "loss": 1.4394, "step": 3191 }, { "epoch": 0.98, "learning_rate": 1.9758242843059826e-05, "loss": 1.4756, "step": 3192 }, { "epoch": 0.98, "learning_rate": 1.9758024994239235e-05, "loss": 1.6394, "step": 3193 }, { "epoch": 0.98, "learning_rate": 1.975780704851247e-05, "loss": 1.4522, "step": 3194 }, { "epoch": 0.98, "learning_rate": 1.975758900588168e-05, "loss": 1.5897, "step": 3195 }, { "epoch": 0.98, "learning_rate": 1.975737086634904e-05, "loss": 1.2209, "step": 3196 }, { "epoch": 0.98, "learning_rate": 1.9757152629916715e-05, "loss": 1.7673, "step": 3197 }, { "epoch": 0.98, "learning_rate": 1.9756934296586872e-05, "loss": 1.4036, "step": 3198 }, { "epoch": 0.98, "learning_rate": 1.975671586636168e-05, "loss": 1.2648, "step": 3199 }, { "epoch": 0.98, "learning_rate": 1.9756497339243307e-05, "loss": 1.7898, "step": 3200 }, { "epoch": 0.98, "learning_rate": 1.9756278715233925e-05, "loss": 1.3847, "step": 3201 }, { "epoch": 0.99, "learning_rate": 1.97560599943357e-05, "loss": 1.6079, "step": 3202 }, { "epoch": 0.99, "learning_rate": 1.9755841176550815e-05, "loss": 1.6971, "step": 3203 }, { "epoch": 0.99, "learning_rate": 1.975562226188143e-05, "loss": 1.6844, "step": 3204 }, { "epoch": 0.99, "learning_rate": 1.9755403250329724e-05, "loss": 1.4223, "step": 3205 }, { "epoch": 0.99, "learning_rate": 1.9755184141897876e-05, "loss": 1.4607, "step": 3206 }, { "epoch": 0.99, "learning_rate": 1.9754964936588058e-05, "loss": 1.6906, "step": 3207 }, { "epoch": 0.99, "learning_rate": 1.9754745634402445e-05, "loss": 1.949, "step": 3208 }, { "epoch": 0.99, "learning_rate": 1.9754526235343222e-05, "loss": 1.4385, "step": 3209 }, { "epoch": 0.99, "learning_rate": 1.975430673941256e-05, "loss": 1.7493, "step": 3210 }, { "epoch": 0.99, "learning_rate": 1.9754087146612644e-05, "loss": 1.5488, "step": 3211 }, { "epoch": 0.99, "learning_rate": 1.9753867456945653e-05, "loss": 1.2591, "step": 3212 }, { "epoch": 0.99, "learning_rate": 1.975364767041377e-05, "loss": 1.5994, "step": 3213 }, { "epoch": 0.99, "learning_rate": 1.9753427787019172e-05, "loss": 1.2894, "step": 3214 }, { "epoch": 0.99, "learning_rate": 1.975320780676405e-05, "loss": 1.7797, "step": 3215 }, { "epoch": 0.99, "learning_rate": 1.975298772965059e-05, "loss": 1.8774, "step": 3216 }, { "epoch": 0.99, "learning_rate": 1.9752767555680967e-05, "loss": 1.3997, "step": 3217 }, { "epoch": 0.99, "learning_rate": 1.9752547284857376e-05, "loss": 1.9061, "step": 3218 }, { "epoch": 0.99, "learning_rate": 1.9752326917182002e-05, "loss": 2.3092, "step": 3219 }, { "epoch": 0.99, "learning_rate": 1.9752106452657033e-05, "loss": 1.6286, "step": 3220 }, { "epoch": 0.99, "learning_rate": 1.9751885891284658e-05, "loss": 1.3421, "step": 3221 }, { "epoch": 0.99, "learning_rate": 1.9751665233067072e-05, "loss": 1.5724, "step": 3222 }, { "epoch": 0.99, "learning_rate": 1.975144447800646e-05, "loss": 1.7277, "step": 3223 }, { "epoch": 0.99, "learning_rate": 1.9751223626105018e-05, "loss": 1.7324, "step": 3224 }, { "epoch": 0.99, "learning_rate": 1.975100267736494e-05, "loss": 1.5843, "step": 3225 }, { "epoch": 0.99, "learning_rate": 1.9750781631788416e-05, "loss": 1.7229, "step": 3226 }, { "epoch": 0.99, "learning_rate": 1.9750560489377647e-05, "loss": 1.2159, "step": 3227 }, { "epoch": 0.99, "learning_rate": 1.9750339250134823e-05, "loss": 1.6146, "step": 3228 }, { "epoch": 0.99, "learning_rate": 1.9750117914062143e-05, "loss": 1.5981, "step": 3229 }, { "epoch": 0.99, "learning_rate": 1.9749896481161807e-05, "loss": 1.8601, "step": 3230 }, { "epoch": 0.99, "learning_rate": 1.9749674951436016e-05, "loss": 1.4024, "step": 3231 }, { "epoch": 0.99, "learning_rate": 1.9749453324886967e-05, "loss": 1.5905, "step": 3232 }, { "epoch": 0.99, "learning_rate": 1.9749231601516858e-05, "loss": 1.8678, "step": 3233 }, { "epoch": 1.0, "learning_rate": 1.9749009781327895e-05, "loss": 1.6668, "step": 3234 }, { "epoch": 1.0, "learning_rate": 1.9748787864322283e-05, "loss": 1.5366, "step": 3235 }, { "epoch": 1.0, "learning_rate": 1.9748565850502218e-05, "loss": 1.3881, "step": 3236 }, { "epoch": 1.0, "learning_rate": 1.9748343739869913e-05, "loss": 1.3458, "step": 3237 }, { "epoch": 1.0, "learning_rate": 1.974812153242757e-05, "loss": 1.5398, "step": 3238 }, { "epoch": 1.0, "learning_rate": 1.9747899228177393e-05, "loss": 1.2985, "step": 3239 }, { "epoch": 1.0, "learning_rate": 1.9747676827121596e-05, "loss": 1.2298, "step": 3240 }, { "epoch": 1.0, "learning_rate": 1.9747454329262382e-05, "loss": 1.7945, "step": 3241 }, { "epoch": 1.0, "learning_rate": 1.9747231734601966e-05, "loss": 1.5278, "step": 3242 }, { "epoch": 1.0, "learning_rate": 1.9747009043142556e-05, "loss": 1.5052, "step": 3243 }, { "epoch": 1.0, "learning_rate": 1.974678625488636e-05, "loss": 1.6162, "step": 3244 }, { "epoch": 1.0, "learning_rate": 1.9746563369835595e-05, "loss": 1.8171, "step": 3245 }, { "epoch": 1.0, "learning_rate": 1.974634038799247e-05, "loss": 1.9239, "step": 3246 }, { "epoch": 1.0, "learning_rate": 1.9746117309359208e-05, "loss": 1.6759, "step": 3247 }, { "epoch": 1.0, "learning_rate": 1.9745894133938014e-05, "loss": 1.7925, "step": 3248 }, { "epoch": 1.0, "learning_rate": 1.974567086173111e-05, "loss": 1.4051, "step": 3249 }, { "epoch": 1.0, "learning_rate": 1.9745447492740713e-05, "loss": 1.634, "step": 3250 }, { "epoch": 1.0, "learning_rate": 1.9745224026969042e-05, "loss": 1.2072, "step": 3251 }, { "epoch": 1.0, "learning_rate": 1.9745000464418317e-05, "loss": 1.1269, "step": 3252 }, { "epoch": 1.0, "learning_rate": 1.9744776805090754e-05, "loss": 1.2863, "step": 3253 }, { "epoch": 1.0, "learning_rate": 1.9744553048988573e-05, "loss": 1.2441, "step": 3254 }, { "epoch": 1.0, "learning_rate": 1.9744329196114002e-05, "loss": 1.0929, "step": 3255 }, { "epoch": 1.0, "learning_rate": 1.9744105246469264e-05, "loss": 1.4427, "step": 3256 }, { "epoch": 1.0, "learning_rate": 1.9743881200056575e-05, "loss": 0.9896, "step": 3257 }, { "epoch": 1.0, "learning_rate": 1.974365705687817e-05, "loss": 1.2782, "step": 3258 }, { "epoch": 1.0, "learning_rate": 1.9743432816936266e-05, "loss": 1.071, "step": 3259 }, { "epoch": 1.0, "learning_rate": 1.97432084802331e-05, "loss": 0.9296, "step": 3260 }, { "epoch": 1.0, "learning_rate": 1.974298404677089e-05, "loss": 1.2701, "step": 3261 }, { "epoch": 1.0, "learning_rate": 1.974275951655187e-05, "loss": 0.8088, "step": 3262 }, { "epoch": 1.0, "learning_rate": 1.974253488957827e-05, "loss": 1.1437, "step": 3263 }, { "epoch": 1.0, "learning_rate": 1.974231016585232e-05, "loss": 1.2829, "step": 3264 }, { "epoch": 1.0, "learning_rate": 1.974208534537625e-05, "loss": 1.1463, "step": 3265 }, { "epoch": 1.0, "learning_rate": 1.9741860428152294e-05, "loss": 1.26, "step": 3266 }, { "epoch": 1.01, "learning_rate": 1.9741635414182685e-05, "loss": 1.2474, "step": 3267 }, { "epoch": 1.01, "learning_rate": 1.974141030346966e-05, "loss": 1.1241, "step": 3268 }, { "epoch": 1.01, "learning_rate": 1.974118509601545e-05, "loss": 1.0507, "step": 3269 }, { "epoch": 1.01, "learning_rate": 1.9740959791822294e-05, "loss": 1.2482, "step": 3270 }, { "epoch": 1.01, "learning_rate": 1.9740734390892432e-05, "loss": 1.2787, "step": 3271 }, { "epoch": 1.01, "learning_rate": 1.97405088932281e-05, "loss": 1.2033, "step": 3272 }, { "epoch": 1.01, "learning_rate": 1.9740283298831538e-05, "loss": 1.2373, "step": 3273 }, { "epoch": 1.01, "learning_rate": 1.9740057607704984e-05, "loss": 0.8733, "step": 3274 }, { "epoch": 1.01, "learning_rate": 1.9739831819850683e-05, "loss": 1.4279, "step": 3275 }, { "epoch": 1.01, "learning_rate": 1.9739605935270877e-05, "loss": 1.4317, "step": 3276 }, { "epoch": 1.01, "learning_rate": 1.9739379953967803e-05, "loss": 1.159, "step": 3277 }, { "epoch": 1.01, "learning_rate": 1.9739153875943714e-05, "loss": 1.3155, "step": 3278 }, { "epoch": 1.01, "learning_rate": 1.973892770120085e-05, "loss": 1.3638, "step": 3279 }, { "epoch": 1.01, "learning_rate": 1.973870142974146e-05, "loss": 1.2148, "step": 3280 }, { "epoch": 1.01, "learning_rate": 1.9738475061567785e-05, "loss": 1.1785, "step": 3281 }, { "epoch": 1.01, "learning_rate": 1.9738248596682078e-05, "loss": 1.1945, "step": 3282 }, { "epoch": 1.01, "learning_rate": 1.973802203508659e-05, "loss": 1.0616, "step": 3283 }, { "epoch": 1.01, "learning_rate": 1.9737795376783568e-05, "loss": 1.0926, "step": 3284 }, { "epoch": 1.01, "learning_rate": 1.9737568621775263e-05, "loss": 0.9495, "step": 3285 }, { "epoch": 1.01, "learning_rate": 1.9737341770063928e-05, "loss": 1.306, "step": 3286 }, { "epoch": 1.01, "learning_rate": 1.973711482165181e-05, "loss": 1.1526, "step": 3287 }, { "epoch": 1.01, "learning_rate": 1.9736887776541173e-05, "loss": 1.0342, "step": 3288 }, { "epoch": 1.01, "learning_rate": 1.9736660634734267e-05, "loss": 1.0714, "step": 3289 }, { "epoch": 1.01, "learning_rate": 1.9736433396233346e-05, "loss": 1.402, "step": 3290 }, { "epoch": 1.01, "learning_rate": 1.973620606104067e-05, "loss": 1.2463, "step": 3291 }, { "epoch": 1.01, "learning_rate": 1.973597862915849e-05, "loss": 1.1419, "step": 3292 }, { "epoch": 1.01, "learning_rate": 1.9735751100589073e-05, "loss": 1.3318, "step": 3293 }, { "epoch": 1.01, "learning_rate": 1.9735523475334674e-05, "loss": 1.3604, "step": 3294 }, { "epoch": 1.01, "learning_rate": 1.973529575339755e-05, "loss": 0.9409, "step": 3295 }, { "epoch": 1.01, "learning_rate": 1.9735067934779973e-05, "loss": 0.9597, "step": 3296 }, { "epoch": 1.01, "learning_rate": 1.97348400194842e-05, "loss": 1.0699, "step": 3297 }, { "epoch": 1.01, "learning_rate": 1.973461200751249e-05, "loss": 0.9161, "step": 3298 }, { "epoch": 1.02, "learning_rate": 1.973438389886711e-05, "loss": 1.0139, "step": 3299 }, { "epoch": 1.02, "learning_rate": 1.9734155693550328e-05, "loss": 1.2045, "step": 3300 }, { "epoch": 1.02, "learning_rate": 1.973392739156441e-05, "loss": 0.7459, "step": 3301 }, { "epoch": 1.02, "learning_rate": 1.973369899291162e-05, "loss": 0.9083, "step": 3302 }, { "epoch": 1.02, "learning_rate": 1.973347049759423e-05, "loss": 1.1419, "step": 3303 }, { "epoch": 1.02, "learning_rate": 1.9733241905614506e-05, "loss": 0.9956, "step": 3304 }, { "epoch": 1.02, "learning_rate": 1.973301321697472e-05, "loss": 1.1228, "step": 3305 }, { "epoch": 1.02, "learning_rate": 1.9732784431677143e-05, "loss": 0.9387, "step": 3306 }, { "epoch": 1.02, "learning_rate": 1.9732555549724046e-05, "loss": 1.2131, "step": 3307 }, { "epoch": 1.02, "learning_rate": 1.9732326571117703e-05, "loss": 0.7602, "step": 3308 }, { "epoch": 1.02, "learning_rate": 1.9732097495860388e-05, "loss": 1.3224, "step": 3309 }, { "epoch": 1.02, "learning_rate": 1.9731868323954374e-05, "loss": 1.1431, "step": 3310 }, { "epoch": 1.02, "learning_rate": 1.9731639055401937e-05, "loss": 1.0695, "step": 3311 }, { "epoch": 1.02, "learning_rate": 1.973140969020536e-05, "loss": 1.0839, "step": 3312 }, { "epoch": 1.02, "learning_rate": 1.9731180228366913e-05, "loss": 1.1249, "step": 3313 }, { "epoch": 1.02, "learning_rate": 1.973095066988888e-05, "loss": 1.4787, "step": 3314 }, { "epoch": 1.02, "learning_rate": 1.9730721014773533e-05, "loss": 1.0815, "step": 3315 }, { "epoch": 1.02, "learning_rate": 1.9730491263023163e-05, "loss": 1.0252, "step": 3316 }, { "epoch": 1.02, "learning_rate": 1.9730261414640046e-05, "loss": 1.1696, "step": 3317 }, { "epoch": 1.02, "learning_rate": 1.9730031469626464e-05, "loss": 1.3357, "step": 3318 }, { "epoch": 1.02, "learning_rate": 1.9729801427984706e-05, "loss": 0.7267, "step": 3319 }, { "epoch": 1.02, "learning_rate": 1.9729571289717048e-05, "loss": 1.2913, "step": 3320 }, { "epoch": 1.02, "learning_rate": 1.9729341054825783e-05, "loss": 1.0639, "step": 3321 }, { "epoch": 1.02, "learning_rate": 1.9729110723313195e-05, "loss": 1.573, "step": 3322 }, { "epoch": 1.02, "learning_rate": 1.972888029518157e-05, "loss": 0.7333, "step": 3323 }, { "epoch": 1.02, "learning_rate": 1.9728649770433193e-05, "loss": 1.2015, "step": 3324 }, { "epoch": 1.02, "learning_rate": 1.9728419149070363e-05, "loss": 1.1688, "step": 3325 }, { "epoch": 1.02, "learning_rate": 1.9728188431095364e-05, "loss": 1.3605, "step": 3326 }, { "epoch": 1.02, "learning_rate": 1.972795761651049e-05, "loss": 1.2188, "step": 3327 }, { "epoch": 1.02, "learning_rate": 1.9727726705318025e-05, "loss": 1.3107, "step": 3328 }, { "epoch": 1.02, "learning_rate": 1.9727495697520275e-05, "loss": 1.3124, "step": 3329 }, { "epoch": 1.02, "learning_rate": 1.9727264593119522e-05, "loss": 0.845, "step": 3330 }, { "epoch": 1.02, "learning_rate": 1.972703339211807e-05, "loss": 0.9452, "step": 3331 }, { "epoch": 1.03, "learning_rate": 1.972680209451821e-05, "loss": 0.9196, "step": 3332 }, { "epoch": 1.03, "learning_rate": 1.9726570700322243e-05, "loss": 1.1852, "step": 3333 }, { "epoch": 1.03, "learning_rate": 1.9726339209532462e-05, "loss": 1.2955, "step": 3334 }, { "epoch": 1.03, "learning_rate": 1.972610762215117e-05, "loss": 0.8592, "step": 3335 }, { "epoch": 1.03, "learning_rate": 1.9725875938180667e-05, "loss": 0.9821, "step": 3336 }, { "epoch": 1.03, "learning_rate": 1.972564415762325e-05, "loss": 1.0646, "step": 3337 }, { "epoch": 1.03, "learning_rate": 1.9725412280481224e-05, "loss": 1.2463, "step": 3338 }, { "epoch": 1.03, "learning_rate": 1.9725180306756893e-05, "loss": 0.9201, "step": 3339 }, { "epoch": 1.03, "learning_rate": 1.9724948236452553e-05, "loss": 1.0508, "step": 3340 }, { "epoch": 1.03, "learning_rate": 1.972471606957052e-05, "loss": 1.2194, "step": 3341 }, { "epoch": 1.03, "learning_rate": 1.972448380611309e-05, "loss": 1.0021, "step": 3342 }, { "epoch": 1.03, "learning_rate": 1.9724251446082573e-05, "loss": 1.3657, "step": 3343 }, { "epoch": 1.03, "learning_rate": 1.9724018989481282e-05, "loss": 1.1688, "step": 3344 }, { "epoch": 1.03, "learning_rate": 1.9723786436311517e-05, "loss": 1.4165, "step": 3345 }, { "epoch": 1.03, "learning_rate": 1.972355378657559e-05, "loss": 1.339, "step": 3346 }, { "epoch": 1.03, "learning_rate": 1.9723321040275816e-05, "loss": 1.2437, "step": 3347 }, { "epoch": 1.03, "learning_rate": 1.97230881974145e-05, "loss": 1.2925, "step": 3348 }, { "epoch": 1.03, "learning_rate": 1.9722855257993955e-05, "loss": 0.9586, "step": 3349 }, { "epoch": 1.03, "learning_rate": 1.9722622222016504e-05, "loss": 0.9194, "step": 3350 }, { "epoch": 1.03, "learning_rate": 1.9722389089484446e-05, "loss": 1.2468, "step": 3351 }, { "epoch": 1.03, "learning_rate": 1.972215586040011e-05, "loss": 1.3679, "step": 3352 }, { "epoch": 1.03, "learning_rate": 1.97219225347658e-05, "loss": 1.0435, "step": 3353 }, { "epoch": 1.03, "learning_rate": 1.9721689112583843e-05, "loss": 1.1862, "step": 3354 }, { "epoch": 1.03, "learning_rate": 1.9721455593856552e-05, "loss": 1.1642, "step": 3355 }, { "epoch": 1.03, "learning_rate": 1.972122197858625e-05, "loss": 0.8688, "step": 3356 }, { "epoch": 1.03, "learning_rate": 1.9720988266775253e-05, "loss": 1.206, "step": 3357 }, { "epoch": 1.03, "learning_rate": 1.9720754458425883e-05, "loss": 1.2255, "step": 3358 }, { "epoch": 1.03, "learning_rate": 1.972052055354046e-05, "loss": 0.9269, "step": 3359 }, { "epoch": 1.03, "learning_rate": 1.972028655212131e-05, "loss": 1.0529, "step": 3360 }, { "epoch": 1.03, "learning_rate": 1.972005245417076e-05, "loss": 1.0241, "step": 3361 }, { "epoch": 1.03, "learning_rate": 1.9719818259691126e-05, "loss": 0.9048, "step": 3362 }, { "epoch": 1.03, "learning_rate": 1.971958396868474e-05, "loss": 1.0649, "step": 3363 }, { "epoch": 1.04, "learning_rate": 1.9719349581153927e-05, "loss": 1.7306, "step": 3364 }, { "epoch": 1.04, "learning_rate": 1.9719115097101018e-05, "loss": 1.0926, "step": 3365 }, { "epoch": 1.04, "learning_rate": 1.9718880516528335e-05, "loss": 1.0126, "step": 3366 }, { "epoch": 1.04, "learning_rate": 1.9718645839438214e-05, "loss": 1.4639, "step": 3367 }, { "epoch": 1.04, "learning_rate": 1.9718411065832984e-05, "loss": 1.2009, "step": 3368 }, { "epoch": 1.04, "learning_rate": 1.971817619571497e-05, "loss": 1.0, "step": 3369 }, { "epoch": 1.04, "learning_rate": 1.9717941229086513e-05, "loss": 1.406, "step": 3370 }, { "epoch": 1.04, "learning_rate": 1.9717706165949945e-05, "loss": 0.9699, "step": 3371 }, { "epoch": 1.04, "learning_rate": 1.9717471006307596e-05, "loss": 1.1299, "step": 3372 }, { "epoch": 1.04, "learning_rate": 1.9717235750161808e-05, "loss": 1.3403, "step": 3373 }, { "epoch": 1.04, "learning_rate": 1.9717000397514908e-05, "loss": 0.9812, "step": 3374 }, { "epoch": 1.04, "learning_rate": 1.9716764948369244e-05, "loss": 1.3274, "step": 3375 }, { "epoch": 1.04, "learning_rate": 1.9716529402727144e-05, "loss": 0.8273, "step": 3376 }, { "epoch": 1.04, "learning_rate": 1.9716293760590955e-05, "loss": 1.3321, "step": 3377 }, { "epoch": 1.04, "learning_rate": 1.9716058021963013e-05, "loss": 1.0313, "step": 3378 }, { "epoch": 1.04, "learning_rate": 1.9715822186845658e-05, "loss": 1.2239, "step": 3379 }, { "epoch": 1.04, "learning_rate": 1.971558625524124e-05, "loss": 1.1837, "step": 3380 }, { "epoch": 1.04, "learning_rate": 1.9715350227152093e-05, "loss": 1.1658, "step": 3381 }, { "epoch": 1.04, "learning_rate": 1.971511410258056e-05, "loss": 1.0313, "step": 3382 }, { "epoch": 1.04, "learning_rate": 1.9714877881528998e-05, "loss": 1.4523, "step": 3383 }, { "epoch": 1.04, "learning_rate": 1.9714641563999742e-05, "loss": 1.1566, "step": 3384 }, { "epoch": 1.04, "learning_rate": 1.971440514999514e-05, "loss": 0.9359, "step": 3385 }, { "epoch": 1.04, "learning_rate": 1.9714168639517543e-05, "loss": 1.0703, "step": 3386 }, { "epoch": 1.04, "learning_rate": 1.97139320325693e-05, "loss": 1.0945, "step": 3387 }, { "epoch": 1.04, "learning_rate": 1.9713695329152758e-05, "loss": 1.0843, "step": 3388 }, { "epoch": 1.04, "learning_rate": 1.971345852927027e-05, "loss": 1.1175, "step": 3389 }, { "epoch": 1.04, "learning_rate": 1.9713221632924188e-05, "loss": 1.0298, "step": 3390 }, { "epoch": 1.04, "learning_rate": 1.971298464011686e-05, "loss": 1.3198, "step": 3391 }, { "epoch": 1.04, "learning_rate": 1.9712747550850642e-05, "loss": 1.257, "step": 3392 }, { "epoch": 1.04, "learning_rate": 1.9712510365127893e-05, "loss": 1.0857, "step": 3393 }, { "epoch": 1.04, "learning_rate": 1.971227308295096e-05, "loss": 1.0481, "step": 3394 }, { "epoch": 1.04, "learning_rate": 1.971203570432221e-05, "loss": 1.1809, "step": 3395 }, { "epoch": 1.04, "learning_rate": 1.9711798229243993e-05, "loss": 0.8209, "step": 3396 }, { "epoch": 1.05, "learning_rate": 1.9711560657718668e-05, "loss": 1.0393, "step": 3397 }, { "epoch": 1.05, "learning_rate": 1.9711322989748592e-05, "loss": 1.1519, "step": 3398 }, { "epoch": 1.05, "learning_rate": 1.971108522533613e-05, "loss": 0.9928, "step": 3399 }, { "epoch": 1.05, "learning_rate": 1.9710847364483644e-05, "loss": 1.6466, "step": 3400 }, { "epoch": 1.05, "learning_rate": 1.9710609407193493e-05, "loss": 1.3545, "step": 3401 }, { "epoch": 1.05, "learning_rate": 1.9710371353468042e-05, "loss": 1.3567, "step": 3402 }, { "epoch": 1.05, "learning_rate": 1.9710133203309656e-05, "loss": 1.0606, "step": 3403 }, { "epoch": 1.05, "learning_rate": 1.9709894956720694e-05, "loss": 1.3238, "step": 3404 }, { "epoch": 1.05, "learning_rate": 1.9709656613703527e-05, "loss": 1.2565, "step": 3405 }, { "epoch": 1.05, "learning_rate": 1.9709418174260523e-05, "loss": 2.0029, "step": 3406 }, { "epoch": 1.05, "learning_rate": 1.9709179638394045e-05, "loss": 0.9917, "step": 3407 }, { "epoch": 1.05, "learning_rate": 1.9708941006106466e-05, "loss": 1.1724, "step": 3408 }, { "epoch": 1.05, "learning_rate": 1.9708702277400155e-05, "loss": 1.0275, "step": 3409 }, { "epoch": 1.05, "learning_rate": 1.9708463452277486e-05, "loss": 1.2286, "step": 3410 }, { "epoch": 1.05, "learning_rate": 1.9708224530740824e-05, "loss": 1.3685, "step": 3411 }, { "epoch": 1.05, "learning_rate": 1.9707985512792544e-05, "loss": 1.3016, "step": 3412 }, { "epoch": 1.05, "learning_rate": 1.9707746398435025e-05, "loss": 1.1246, "step": 3413 }, { "epoch": 1.05, "learning_rate": 1.9707507187670632e-05, "loss": 0.8998, "step": 3414 }, { "epoch": 1.05, "learning_rate": 1.970726788050175e-05, "loss": 1.0962, "step": 3415 }, { "epoch": 1.05, "learning_rate": 1.9707028476930747e-05, "loss": 1.0776, "step": 3416 }, { "epoch": 1.05, "learning_rate": 1.9706788976960008e-05, "loss": 1.1986, "step": 3417 }, { "epoch": 1.05, "learning_rate": 1.970654938059191e-05, "loss": 1.1021, "step": 3418 }, { "epoch": 1.05, "learning_rate": 1.9706309687828828e-05, "loss": 1.476, "step": 3419 }, { "epoch": 1.05, "learning_rate": 1.9706069898673147e-05, "loss": 1.1516, "step": 3420 }, { "epoch": 1.05, "learning_rate": 1.9705830013127248e-05, "loss": 0.9753, "step": 3421 }, { "epoch": 1.05, "learning_rate": 1.970559003119351e-05, "loss": 1.2712, "step": 3422 }, { "epoch": 1.05, "learning_rate": 1.9705349952874315e-05, "loss": 0.9532, "step": 3423 }, { "epoch": 1.05, "learning_rate": 1.9705109778172052e-05, "loss": 1.4504, "step": 3424 }, { "epoch": 1.05, "learning_rate": 1.9704869507089105e-05, "loss": 0.9567, "step": 3425 }, { "epoch": 1.05, "learning_rate": 1.970462913962786e-05, "loss": 1.067, "step": 3426 }, { "epoch": 1.05, "learning_rate": 1.9704388675790706e-05, "loss": 1.231, "step": 3427 }, { "epoch": 1.05, "learning_rate": 1.9704148115580025e-05, "loss": 0.7162, "step": 3428 }, { "epoch": 1.06, "learning_rate": 1.970390745899821e-05, "loss": 1.2428, "step": 3429 }, { "epoch": 1.06, "learning_rate": 1.970366670604765e-05, "loss": 1.0986, "step": 3430 }, { "epoch": 1.06, "learning_rate": 1.970342585673074e-05, "loss": 1.2188, "step": 3431 }, { "epoch": 1.06, "learning_rate": 1.9703184911049865e-05, "loss": 1.2386, "step": 3432 }, { "epoch": 1.06, "learning_rate": 1.9702943869007423e-05, "loss": 1.081, "step": 3433 }, { "epoch": 1.06, "learning_rate": 1.9702702730605805e-05, "loss": 1.303, "step": 3434 }, { "epoch": 1.06, "learning_rate": 1.970246149584741e-05, "loss": 1.5581, "step": 3435 }, { "epoch": 1.06, "learning_rate": 1.9702220164734624e-05, "loss": 0.9219, "step": 3436 }, { "epoch": 1.06, "learning_rate": 1.9701978737269853e-05, "loss": 1.2906, "step": 3437 }, { "epoch": 1.06, "learning_rate": 1.970173721345549e-05, "loss": 1.0124, "step": 3438 }, { "epoch": 1.06, "learning_rate": 1.970149559329394e-05, "loss": 2.0114, "step": 3439 }, { "epoch": 1.06, "learning_rate": 1.970125387678759e-05, "loss": 1.2143, "step": 3440 }, { "epoch": 1.06, "learning_rate": 1.9701012063938854e-05, "loss": 1.0908, "step": 3441 }, { "epoch": 1.06, "learning_rate": 1.9700770154750125e-05, "loss": 0.943, "step": 3442 }, { "epoch": 1.06, "learning_rate": 1.970052814922381e-05, "loss": 1.3333, "step": 3443 }, { "epoch": 1.06, "learning_rate": 1.9700286047362308e-05, "loss": 1.2137, "step": 3444 }, { "epoch": 1.06, "learning_rate": 1.9700043849168025e-05, "loss": 1.1514, "step": 3445 }, { "epoch": 1.06, "learning_rate": 1.9699801554643367e-05, "loss": 1.2551, "step": 3446 }, { "epoch": 1.06, "learning_rate": 1.9699559163790742e-05, "loss": 1.1238, "step": 3447 }, { "epoch": 1.06, "learning_rate": 1.9699316676612554e-05, "loss": 1.1423, "step": 3448 }, { "epoch": 1.06, "learning_rate": 1.9699074093111213e-05, "loss": 1.0602, "step": 3449 }, { "epoch": 1.06, "learning_rate": 1.9698831413289127e-05, "loss": 0.8845, "step": 3450 }, { "epoch": 1.06, "learning_rate": 1.9698588637148705e-05, "loss": 0.926, "step": 3451 }, { "epoch": 1.06, "learning_rate": 1.969834576469236e-05, "loss": 1.2968, "step": 3452 }, { "epoch": 1.06, "learning_rate": 1.9698102795922504e-05, "loss": 1.1512, "step": 3453 }, { "epoch": 1.06, "learning_rate": 1.969785973084155e-05, "loss": 0.9028, "step": 3454 }, { "epoch": 1.06, "learning_rate": 1.969761656945191e-05, "loss": 1.0343, "step": 3455 }, { "epoch": 1.06, "learning_rate": 1.9697373311755997e-05, "loss": 1.1716, "step": 3456 }, { "epoch": 1.06, "learning_rate": 1.9697129957756234e-05, "loss": 1.2731, "step": 3457 }, { "epoch": 1.06, "learning_rate": 1.969688650745503e-05, "loss": 1.0069, "step": 3458 }, { "epoch": 1.06, "learning_rate": 1.969664296085481e-05, "loss": 1.0554, "step": 3459 }, { "epoch": 1.06, "learning_rate": 1.9696399317957985e-05, "loss": 1.0336, "step": 3460 }, { "epoch": 1.06, "learning_rate": 1.969615557876698e-05, "loss": 1.2832, "step": 3461 }, { "epoch": 1.07, "learning_rate": 1.9695911743284214e-05, "loss": 1.2287, "step": 3462 }, { "epoch": 1.07, "learning_rate": 1.969566781151211e-05, "loss": 1.1581, "step": 3463 }, { "epoch": 1.07, "learning_rate": 1.9695423783453086e-05, "loss": 1.2016, "step": 3464 }, { "epoch": 1.07, "learning_rate": 1.969517965910957e-05, "loss": 0.8899, "step": 3465 }, { "epoch": 1.07, "learning_rate": 1.9694935438483985e-05, "loss": 1.2069, "step": 3466 }, { "epoch": 1.07, "learning_rate": 1.9694691121578754e-05, "loss": 1.2935, "step": 3467 }, { "epoch": 1.07, "learning_rate": 1.9694446708396305e-05, "loss": 1.5668, "step": 3468 }, { "epoch": 1.07, "learning_rate": 1.9694202198939068e-05, "loss": 1.1722, "step": 3469 }, { "epoch": 1.07, "learning_rate": 1.9693957593209466e-05, "loss": 1.2105, "step": 3470 }, { "epoch": 1.07, "learning_rate": 1.9693712891209934e-05, "loss": 1.1483, "step": 3471 }, { "epoch": 1.07, "learning_rate": 1.9693468092942892e-05, "loss": 1.3262, "step": 3472 }, { "epoch": 1.07, "learning_rate": 1.9693223198410786e-05, "loss": 1.0573, "step": 3473 }, { "epoch": 1.07, "learning_rate": 1.9692978207616038e-05, "loss": 1.2554, "step": 3474 }, { "epoch": 1.07, "learning_rate": 1.969273312056108e-05, "loss": 1.2364, "step": 3475 }, { "epoch": 1.07, "learning_rate": 1.9692487937248352e-05, "loss": 1.2996, "step": 3476 }, { "epoch": 1.07, "learning_rate": 1.9692242657680286e-05, "loss": 1.162, "step": 3477 }, { "epoch": 1.07, "learning_rate": 1.969199728185932e-05, "loss": 1.1229, "step": 3478 }, { "epoch": 1.07, "learning_rate": 1.9691751809787885e-05, "loss": 1.0984, "step": 3479 }, { "epoch": 1.07, "learning_rate": 1.9691506241468425e-05, "loss": 1.293, "step": 3480 }, { "epoch": 1.07, "learning_rate": 1.9691260576903372e-05, "loss": 1.1597, "step": 3481 }, { "epoch": 1.07, "learning_rate": 1.969101481609517e-05, "loss": 1.1057, "step": 3482 }, { "epoch": 1.07, "learning_rate": 1.9690768959046262e-05, "loss": 1.3079, "step": 3483 }, { "epoch": 1.07, "learning_rate": 1.969052300575909e-05, "loss": 1.2402, "step": 3484 }, { "epoch": 1.07, "learning_rate": 1.969027695623609e-05, "loss": 0.8369, "step": 3485 }, { "epoch": 1.07, "learning_rate": 1.9690030810479706e-05, "loss": 1.3034, "step": 3486 }, { "epoch": 1.07, "learning_rate": 1.968978456849239e-05, "loss": 1.1727, "step": 3487 }, { "epoch": 1.07, "learning_rate": 1.968953823027658e-05, "loss": 1.0663, "step": 3488 }, { "epoch": 1.07, "learning_rate": 1.9689291795834725e-05, "loss": 1.0896, "step": 3489 }, { "epoch": 1.07, "learning_rate": 1.9689045265169272e-05, "loss": 1.1273, "step": 3490 }, { "epoch": 1.07, "learning_rate": 1.9688798638282675e-05, "loss": 1.3608, "step": 3491 }, { "epoch": 1.07, "learning_rate": 1.9688551915177374e-05, "loss": 1.1332, "step": 3492 }, { "epoch": 1.07, "learning_rate": 1.9688305095855825e-05, "loss": 0.9422, "step": 3493 }, { "epoch": 1.08, "learning_rate": 1.9688058180320474e-05, "loss": 1.4469, "step": 3494 }, { "epoch": 1.08, "learning_rate": 1.968781116857378e-05, "loss": 1.1211, "step": 3495 }, { "epoch": 1.08, "learning_rate": 1.9687564060618193e-05, "loss": 1.0577, "step": 3496 }, { "epoch": 1.08, "learning_rate": 1.9687316856456166e-05, "loss": 1.159, "step": 3497 }, { "epoch": 1.08, "learning_rate": 1.9687069556090152e-05, "loss": 1.6112, "step": 3498 }, { "epoch": 1.08, "learning_rate": 1.968682215952261e-05, "loss": 1.1312, "step": 3499 }, { "epoch": 1.08, "learning_rate": 1.9686574666756e-05, "loss": 1.2512, "step": 3500 }, { "epoch": 1.08, "learning_rate": 1.9686327077792772e-05, "loss": 1.1057, "step": 3501 }, { "epoch": 1.08, "learning_rate": 1.968607939263539e-05, "loss": 1.1799, "step": 3502 }, { "epoch": 1.08, "learning_rate": 1.9685831611286312e-05, "loss": 1.356, "step": 3503 }, { "epoch": 1.08, "learning_rate": 1.9685583733748003e-05, "loss": 1.3459, "step": 3504 }, { "epoch": 1.08, "learning_rate": 1.9685335760022916e-05, "loss": 0.9402, "step": 3505 }, { "epoch": 1.08, "learning_rate": 1.968508769011352e-05, "loss": 1.1564, "step": 3506 }, { "epoch": 1.08, "learning_rate": 1.9684839524022278e-05, "loss": 0.8536, "step": 3507 }, { "epoch": 1.08, "learning_rate": 1.9684591261751656e-05, "loss": 1.303, "step": 3508 }, { "epoch": 1.08, "learning_rate": 1.9684342903304116e-05, "loss": 1.1378, "step": 3509 }, { "epoch": 1.08, "learning_rate": 1.9684094448682124e-05, "loss": 1.0405, "step": 3510 }, { "epoch": 1.08, "learning_rate": 1.968384589788815e-05, "loss": 1.348, "step": 3511 }, { "epoch": 1.08, "learning_rate": 1.968359725092466e-05, "loss": 1.0452, "step": 3512 }, { "epoch": 1.08, "learning_rate": 1.9683348507794127e-05, "loss": 1.4258, "step": 3513 }, { "epoch": 1.08, "learning_rate": 1.9683099668499017e-05, "loss": 0.9209, "step": 3514 }, { "epoch": 1.08, "learning_rate": 1.9682850733041805e-05, "loss": 1.1268, "step": 3515 }, { "epoch": 1.08, "learning_rate": 1.9682601701424958e-05, "loss": 1.1285, "step": 3516 }, { "epoch": 1.08, "learning_rate": 1.9682352573650955e-05, "loss": 1.3284, "step": 3517 }, { "epoch": 1.08, "learning_rate": 1.968210334972227e-05, "loss": 1.108, "step": 3518 }, { "epoch": 1.08, "learning_rate": 1.968185402964137e-05, "loss": 1.0042, "step": 3519 }, { "epoch": 1.08, "learning_rate": 1.968160461341074e-05, "loss": 0.9957, "step": 3520 }, { "epoch": 1.08, "learning_rate": 1.968135510103285e-05, "loss": 1.1817, "step": 3521 }, { "epoch": 1.08, "learning_rate": 1.9681105492510184e-05, "loss": 2.5868, "step": 3522 }, { "epoch": 1.08, "learning_rate": 1.9680855787845216e-05, "loss": 1.2031, "step": 3523 }, { "epoch": 1.08, "learning_rate": 1.9680605987040432e-05, "loss": 1.088, "step": 3524 }, { "epoch": 1.08, "learning_rate": 1.9680356090098303e-05, "loss": 1.4639, "step": 3525 }, { "epoch": 1.08, "learning_rate": 1.968010609702132e-05, "loss": 1.1407, "step": 3526 }, { "epoch": 1.09, "learning_rate": 1.9679856007811962e-05, "loss": 1.1203, "step": 3527 }, { "epoch": 1.09, "learning_rate": 1.9679605822472713e-05, "loss": 0.9806, "step": 3528 }, { "epoch": 1.09, "learning_rate": 1.9679355541006056e-05, "loss": 2.2399, "step": 3529 }, { "epoch": 1.09, "learning_rate": 1.9679105163414478e-05, "loss": 1.0146, "step": 3530 }, { "epoch": 1.09, "learning_rate": 1.9678854689700463e-05, "loss": 1.291, "step": 3531 }, { "epoch": 1.09, "learning_rate": 1.96786041198665e-05, "loss": 1.1111, "step": 3532 }, { "epoch": 1.09, "learning_rate": 1.9678353453915086e-05, "loss": 1.0139, "step": 3533 }, { "epoch": 1.09, "learning_rate": 1.9678102691848694e-05, "loss": 1.0113, "step": 3534 }, { "epoch": 1.09, "learning_rate": 1.9677851833669826e-05, "loss": 0.9218, "step": 3535 }, { "epoch": 1.09, "learning_rate": 1.9677600879380972e-05, "loss": 1.3471, "step": 3536 }, { "epoch": 1.09, "learning_rate": 1.967734982898462e-05, "loss": 0.9865, "step": 3537 }, { "epoch": 1.09, "learning_rate": 1.9677098682483265e-05, "loss": 1.5418, "step": 3538 }, { "epoch": 1.09, "learning_rate": 1.96768474398794e-05, "loss": 1.1506, "step": 3539 }, { "epoch": 1.09, "learning_rate": 1.9676596101175524e-05, "loss": 1.3688, "step": 3540 }, { "epoch": 1.09, "learning_rate": 1.967634466637413e-05, "loss": 1.2754, "step": 3541 }, { "epoch": 1.09, "learning_rate": 1.9676093135477713e-05, "loss": 1.2418, "step": 3542 }, { "epoch": 1.09, "learning_rate": 1.9675841508488777e-05, "loss": 1.0849, "step": 3543 }, { "epoch": 1.09, "learning_rate": 1.9675589785409815e-05, "loss": 1.1578, "step": 3544 }, { "epoch": 1.09, "learning_rate": 1.967533796624333e-05, "loss": 1.2, "step": 3545 }, { "epoch": 1.09, "learning_rate": 1.967508605099182e-05, "loss": 1.4154, "step": 3546 }, { "epoch": 1.09, "learning_rate": 1.967483403965779e-05, "loss": 1.1111, "step": 3547 }, { "epoch": 1.09, "learning_rate": 1.967458193224374e-05, "loss": 1.3463, "step": 3548 }, { "epoch": 1.09, "learning_rate": 1.967432972875218e-05, "loss": 1.1675, "step": 3549 }, { "epoch": 1.09, "learning_rate": 1.9674077429185607e-05, "loss": 1.192, "step": 3550 }, { "epoch": 1.09, "learning_rate": 1.9673825033546526e-05, "loss": 1.7574, "step": 3551 }, { "epoch": 1.09, "learning_rate": 1.967357254183745e-05, "loss": 1.5233, "step": 3552 }, { "epoch": 1.09, "learning_rate": 1.9673319954060885e-05, "loss": 1.0814, "step": 3553 }, { "epoch": 1.09, "learning_rate": 1.9673067270219336e-05, "loss": 1.2411, "step": 3554 }, { "epoch": 1.09, "learning_rate": 1.9672814490315312e-05, "loss": 1.2265, "step": 3555 }, { "epoch": 1.09, "learning_rate": 1.9672561614351325e-05, "loss": 1.2311, "step": 3556 }, { "epoch": 1.09, "learning_rate": 1.9672308642329892e-05, "loss": 1.3346, "step": 3557 }, { "epoch": 1.09, "learning_rate": 1.9672055574253516e-05, "loss": 1.4746, "step": 3558 }, { "epoch": 1.1, "learning_rate": 1.9671802410124712e-05, "loss": 1.1486, "step": 3559 }, { "epoch": 1.1, "learning_rate": 1.9671549149946e-05, "loss": 0.9424, "step": 3560 }, { "epoch": 1.1, "learning_rate": 1.9671295793719893e-05, "loss": 1.2862, "step": 3561 }, { "epoch": 1.1, "learning_rate": 1.9671042341448898e-05, "loss": 1.2099, "step": 3562 }, { "epoch": 1.1, "learning_rate": 1.9670788793135545e-05, "loss": 1.1699, "step": 3563 }, { "epoch": 1.1, "learning_rate": 1.9670535148782347e-05, "loss": 1.2906, "step": 3564 }, { "epoch": 1.1, "learning_rate": 1.967028140839182e-05, "loss": 1.2353, "step": 3565 }, { "epoch": 1.1, "learning_rate": 1.9670027571966486e-05, "loss": 1.2372, "step": 3566 }, { "epoch": 1.1, "learning_rate": 1.9669773639508868e-05, "loss": 1.3526, "step": 3567 }, { "epoch": 1.1, "learning_rate": 1.9669519611021485e-05, "loss": 1.0276, "step": 3568 }, { "epoch": 1.1, "learning_rate": 1.966926548650686e-05, "loss": 1.0551, "step": 3569 }, { "epoch": 1.1, "learning_rate": 1.966901126596752e-05, "loss": 0.8312, "step": 3570 }, { "epoch": 1.1, "learning_rate": 1.9668756949405986e-05, "loss": 1.133, "step": 3571 }, { "epoch": 1.1, "learning_rate": 1.966850253682478e-05, "loss": 1.2113, "step": 3572 }, { "epoch": 1.1, "learning_rate": 1.9668248028226435e-05, "loss": 1.2138, "step": 3573 }, { "epoch": 1.1, "learning_rate": 1.966799342361348e-05, "loss": 1.3367, "step": 3574 }, { "epoch": 1.1, "learning_rate": 1.966773872298844e-05, "loss": 1.2673, "step": 3575 }, { "epoch": 1.1, "learning_rate": 1.966748392635384e-05, "loss": 1.1079, "step": 3576 }, { "epoch": 1.1, "learning_rate": 1.966722903371222e-05, "loss": 1.1635, "step": 3577 }, { "epoch": 1.1, "learning_rate": 1.9666974045066102e-05, "loss": 1.1531, "step": 3578 }, { "epoch": 1.1, "learning_rate": 1.966671896041802e-05, "loss": 1.2018, "step": 3579 }, { "epoch": 1.1, "learning_rate": 1.9666463779770516e-05, "loss": 1.1466, "step": 3580 }, { "epoch": 1.1, "learning_rate": 1.9666208503126115e-05, "loss": 1.0401, "step": 3581 }, { "epoch": 1.1, "learning_rate": 1.9665953130487353e-05, "loss": 1.0433, "step": 3582 }, { "epoch": 1.1, "learning_rate": 1.966569766185677e-05, "loss": 1.415, "step": 3583 }, { "epoch": 1.1, "learning_rate": 1.96654420972369e-05, "loss": 0.968, "step": 3584 }, { "epoch": 1.1, "learning_rate": 1.9665186436630286e-05, "loss": 0.7763, "step": 3585 }, { "epoch": 1.1, "learning_rate": 1.9664930680039457e-05, "loss": 1.118, "step": 3586 }, { "epoch": 1.1, "learning_rate": 1.9664674827466964e-05, "loss": 0.8868, "step": 3587 }, { "epoch": 1.1, "learning_rate": 1.966441887891534e-05, "loss": 1.2153, "step": 3588 }, { "epoch": 1.1, "learning_rate": 1.966416283438713e-05, "loss": 1.3988, "step": 3589 }, { "epoch": 1.1, "learning_rate": 1.9663906693884877e-05, "loss": 1.2487, "step": 3590 }, { "epoch": 1.1, "learning_rate": 1.9663650457411123e-05, "loss": 1.0267, "step": 3591 }, { "epoch": 1.11, "learning_rate": 1.9663394124968414e-05, "loss": 1.6805, "step": 3592 }, { "epoch": 1.11, "learning_rate": 1.9663137696559295e-05, "loss": 1.4289, "step": 3593 }, { "epoch": 1.11, "learning_rate": 1.9662881172186313e-05, "loss": 1.3458, "step": 3594 }, { "epoch": 1.11, "learning_rate": 1.9662624551852017e-05, "loss": 1.2036, "step": 3595 }, { "epoch": 1.11, "learning_rate": 1.9662367835558954e-05, "loss": 1.1557, "step": 3596 }, { "epoch": 1.11, "learning_rate": 1.966211102330967e-05, "loss": 1.241, "step": 3597 }, { "epoch": 1.11, "learning_rate": 1.966185411510672e-05, "loss": 1.1132, "step": 3598 }, { "epoch": 1.11, "learning_rate": 1.9661597110952656e-05, "loss": 0.9886, "step": 3599 }, { "epoch": 1.11, "learning_rate": 1.9661340010850025e-05, "loss": 0.987, "step": 3600 }, { "epoch": 1.11, "learning_rate": 1.9661082814801387e-05, "loss": 1.0152, "step": 3601 }, { "epoch": 1.11, "learning_rate": 1.966082552280929e-05, "loss": 0.9004, "step": 3602 }, { "epoch": 1.11, "learning_rate": 1.9660568134876294e-05, "loss": 1.172, "step": 3603 }, { "epoch": 1.11, "learning_rate": 1.966031065100495e-05, "loss": 1.1553, "step": 3604 }, { "epoch": 1.11, "learning_rate": 1.9660053071197822e-05, "loss": 0.9801, "step": 3605 }, { "epoch": 1.11, "learning_rate": 1.965979539545746e-05, "loss": 0.9596, "step": 3606 }, { "epoch": 1.11, "learning_rate": 1.9659537623786428e-05, "loss": 1.0107, "step": 3607 }, { "epoch": 1.11, "learning_rate": 1.9659279756187287e-05, "loss": 1.0812, "step": 3608 }, { "epoch": 1.11, "learning_rate": 1.9659021792662597e-05, "loss": 1.4224, "step": 3609 }, { "epoch": 1.11, "learning_rate": 1.9658763733214915e-05, "loss": 1.2485, "step": 3610 }, { "epoch": 1.11, "learning_rate": 1.965850557784681e-05, "loss": 1.183, "step": 3611 }, { "epoch": 1.11, "learning_rate": 1.965824732656084e-05, "loss": 1.2817, "step": 3612 }, { "epoch": 1.11, "learning_rate": 1.9657988979359576e-05, "loss": 1.1635, "step": 3613 }, { "epoch": 1.11, "learning_rate": 1.965773053624558e-05, "loss": 1.3388, "step": 3614 }, { "epoch": 1.11, "learning_rate": 1.9657471997221416e-05, "loss": 0.9628, "step": 3615 }, { "epoch": 1.11, "learning_rate": 1.9657213362289657e-05, "loss": 1.1265, "step": 3616 }, { "epoch": 1.11, "learning_rate": 1.9656954631452863e-05, "loss": 1.1732, "step": 3617 }, { "epoch": 1.11, "learning_rate": 1.965669580471362e-05, "loss": 0.8342, "step": 3618 }, { "epoch": 1.11, "learning_rate": 1.9656436882074482e-05, "loss": 1.3293, "step": 3619 }, { "epoch": 1.11, "learning_rate": 1.9656177863538025e-05, "loss": 1.4151, "step": 3620 }, { "epoch": 1.11, "learning_rate": 1.9655918749106823e-05, "loss": 1.1614, "step": 3621 }, { "epoch": 1.11, "learning_rate": 1.965565953878345e-05, "loss": 1.3887, "step": 3622 }, { "epoch": 1.11, "learning_rate": 1.9655400232570478e-05, "loss": 1.1113, "step": 3623 }, { "epoch": 1.12, "learning_rate": 1.9655140830470482e-05, "loss": 1.1666, "step": 3624 }, { "epoch": 1.12, "learning_rate": 1.9654881332486045e-05, "loss": 1.4172, "step": 3625 }, { "epoch": 1.12, "learning_rate": 1.965462173861973e-05, "loss": 1.3266, "step": 3626 }, { "epoch": 1.12, "learning_rate": 1.965436204887413e-05, "loss": 1.1413, "step": 3627 }, { "epoch": 1.12, "learning_rate": 1.9654102263251813e-05, "loss": 1.3486, "step": 3628 }, { "epoch": 1.12, "learning_rate": 1.9653842381755364e-05, "loss": 1.1138, "step": 3629 }, { "epoch": 1.12, "learning_rate": 1.9653582404387365e-05, "loss": 1.1609, "step": 3630 }, { "epoch": 1.12, "learning_rate": 1.965332233115039e-05, "loss": 1.441, "step": 3631 }, { "epoch": 1.12, "learning_rate": 1.965306216204703e-05, "loss": 1.4644, "step": 3632 }, { "epoch": 1.12, "learning_rate": 1.965280189707987e-05, "loss": 1.4157, "step": 3633 }, { "epoch": 1.12, "learning_rate": 1.9652541536251488e-05, "loss": 1.0547, "step": 3634 }, { "epoch": 1.12, "learning_rate": 1.9652281079564474e-05, "loss": 1.3772, "step": 3635 }, { "epoch": 1.12, "learning_rate": 1.9652020527021412e-05, "loss": 1.231, "step": 3636 }, { "epoch": 1.12, "learning_rate": 1.965175987862489e-05, "loss": 1.0911, "step": 3637 }, { "epoch": 1.12, "learning_rate": 1.9651499134377495e-05, "loss": 1.4443, "step": 3638 }, { "epoch": 1.12, "learning_rate": 1.9651238294281824e-05, "loss": 1.0363, "step": 3639 }, { "epoch": 1.12, "learning_rate": 1.965097735834046e-05, "loss": 0.9316, "step": 3640 }, { "epoch": 1.12, "learning_rate": 1.965071632655599e-05, "loss": 1.1948, "step": 3641 }, { "epoch": 1.12, "learning_rate": 1.965045519893102e-05, "loss": 0.8148, "step": 3642 }, { "epoch": 1.12, "learning_rate": 1.9650193975468134e-05, "loss": 0.9167, "step": 3643 }, { "epoch": 1.12, "learning_rate": 1.9649932656169926e-05, "loss": 1.0751, "step": 3644 }, { "epoch": 1.12, "learning_rate": 1.9649671241038993e-05, "loss": 1.1008, "step": 3645 }, { "epoch": 1.12, "learning_rate": 1.9649409730077934e-05, "loss": 1.124, "step": 3646 }, { "epoch": 1.12, "learning_rate": 1.9649148123289345e-05, "loss": 0.9083, "step": 3647 }, { "epoch": 1.12, "learning_rate": 1.9648886420675816e-05, "loss": 0.8911, "step": 3648 }, { "epoch": 1.12, "learning_rate": 1.9648624622239957e-05, "loss": 1.5328, "step": 3649 }, { "epoch": 1.12, "learning_rate": 1.964836272798436e-05, "loss": 1.1821, "step": 3650 }, { "epoch": 1.12, "learning_rate": 1.964810073791163e-05, "loss": 1.0018, "step": 3651 }, { "epoch": 1.12, "learning_rate": 1.964783865202437e-05, "loss": 1.0126, "step": 3652 }, { "epoch": 1.12, "learning_rate": 1.9647576470325177e-05, "loss": 1.2669, "step": 3653 }, { "epoch": 1.12, "learning_rate": 1.964731419281666e-05, "loss": 1.0668, "step": 3654 }, { "epoch": 1.12, "learning_rate": 1.964705181950142e-05, "loss": 1.6984, "step": 3655 }, { "epoch": 1.12, "learning_rate": 1.9646789350382066e-05, "loss": 1.0059, "step": 3656 }, { "epoch": 1.13, "learning_rate": 1.9646526785461204e-05, "loss": 1.2097, "step": 3657 }, { "epoch": 1.13, "learning_rate": 1.9646264124741438e-05, "loss": 0.9715, "step": 3658 }, { "epoch": 1.13, "learning_rate": 1.9646001368225382e-05, "loss": 1.13, "step": 3659 }, { "epoch": 1.13, "learning_rate": 1.9645738515915643e-05, "loss": 0.9859, "step": 3660 }, { "epoch": 1.13, "learning_rate": 1.9645475567814826e-05, "loss": 1.1188, "step": 3661 }, { "epoch": 1.13, "learning_rate": 1.964521252392555e-05, "loss": 1.3289, "step": 3662 }, { "epoch": 1.13, "learning_rate": 1.9644949384250425e-05, "loss": 1.2563, "step": 3663 }, { "epoch": 1.13, "learning_rate": 1.964468614879206e-05, "loss": 1.2945, "step": 3664 }, { "epoch": 1.13, "learning_rate": 1.9644422817553078e-05, "loss": 1.252, "step": 3665 }, { "epoch": 1.13, "learning_rate": 1.9644159390536086e-05, "loss": 1.5527, "step": 3666 }, { "epoch": 1.13, "learning_rate": 1.9643895867743702e-05, "loss": 1.3083, "step": 3667 }, { "epoch": 1.13, "learning_rate": 1.9643632249178547e-05, "loss": 0.9033, "step": 3668 }, { "epoch": 1.13, "learning_rate": 1.9643368534843232e-05, "loss": 1.0251, "step": 3669 }, { "epoch": 1.13, "learning_rate": 1.964310472474038e-05, "loss": 1.3486, "step": 3670 }, { "epoch": 1.13, "learning_rate": 1.9642840818872616e-05, "loss": 1.5957, "step": 3671 }, { "epoch": 1.13, "learning_rate": 1.9642576817242553e-05, "loss": 1.2268, "step": 3672 }, { "epoch": 1.13, "learning_rate": 1.9642312719852813e-05, "loss": 1.2461, "step": 3673 }, { "epoch": 1.13, "learning_rate": 1.9642048526706022e-05, "loss": 1.1851, "step": 3674 }, { "epoch": 1.13, "learning_rate": 1.9641784237804803e-05, "loss": 1.1904, "step": 3675 }, { "epoch": 1.13, "learning_rate": 1.964151985315178e-05, "loss": 1.143, "step": 3676 }, { "epoch": 1.13, "learning_rate": 1.9641255372749582e-05, "loss": 1.1433, "step": 3677 }, { "epoch": 1.13, "learning_rate": 1.964099079660083e-05, "loss": 1.2484, "step": 3678 }, { "epoch": 1.13, "learning_rate": 1.9640726124708154e-05, "loss": 1.0573, "step": 3679 }, { "epoch": 1.13, "learning_rate": 1.9640461357074182e-05, "loss": 1.4601, "step": 3680 }, { "epoch": 1.13, "learning_rate": 1.9640196493701543e-05, "loss": 1.3013, "step": 3681 }, { "epoch": 1.13, "learning_rate": 1.963993153459287e-05, "loss": 1.2604, "step": 3682 }, { "epoch": 1.13, "learning_rate": 1.9639666479750792e-05, "loss": 1.3697, "step": 3683 }, { "epoch": 1.13, "learning_rate": 1.9639401329177942e-05, "loss": 1.1298, "step": 3684 }, { "epoch": 1.13, "learning_rate": 1.9639136082876954e-05, "loss": 1.5605, "step": 3685 }, { "epoch": 1.13, "learning_rate": 1.963887074085046e-05, "loss": 0.9264, "step": 3686 }, { "epoch": 1.13, "learning_rate": 1.9638605303101093e-05, "loss": 1.2899, "step": 3687 }, { "epoch": 1.13, "learning_rate": 1.9638339769631495e-05, "loss": 1.0388, "step": 3688 }, { "epoch": 1.14, "learning_rate": 1.96380741404443e-05, "loss": 1.0291, "step": 3689 }, { "epoch": 1.14, "learning_rate": 1.963780841554215e-05, "loss": 1.249, "step": 3690 }, { "epoch": 1.14, "learning_rate": 1.9637542594927673e-05, "loss": 1.3962, "step": 3691 }, { "epoch": 1.14, "learning_rate": 1.963727667860352e-05, "loss": 1.1962, "step": 3692 }, { "epoch": 1.14, "learning_rate": 1.9637010666572326e-05, "loss": 1.2074, "step": 3693 }, { "epoch": 1.14, "learning_rate": 1.9636744558836734e-05, "loss": 1.2556, "step": 3694 }, { "epoch": 1.14, "learning_rate": 1.9636478355399386e-05, "loss": 1.281, "step": 3695 }, { "epoch": 1.14, "learning_rate": 1.963621205626293e-05, "loss": 1.3002, "step": 3696 }, { "epoch": 1.14, "learning_rate": 1.9635945661430006e-05, "loss": 1.1487, "step": 3697 }, { "epoch": 1.14, "learning_rate": 1.9635679170903258e-05, "loss": 0.9776, "step": 3698 }, { "epoch": 1.14, "learning_rate": 1.963541258468534e-05, "loss": 1.4104, "step": 3699 }, { "epoch": 1.14, "learning_rate": 1.9635145902778894e-05, "loss": 0.9209, "step": 3700 }, { "epoch": 1.14, "learning_rate": 1.9634879125186567e-05, "loss": 1.1899, "step": 3701 }, { "epoch": 1.14, "learning_rate": 1.9634612251911013e-05, "loss": 1.17, "step": 3702 }, { "epoch": 1.14, "learning_rate": 1.9634345282954875e-05, "loss": 1.3822, "step": 3703 }, { "epoch": 1.14, "learning_rate": 1.9634078218320813e-05, "loss": 1.3046, "step": 3704 }, { "epoch": 1.14, "learning_rate": 1.9633811058011476e-05, "loss": 1.057, "step": 3705 }, { "epoch": 1.14, "learning_rate": 1.9633543802029513e-05, "loss": 0.9939, "step": 3706 }, { "epoch": 1.14, "learning_rate": 1.9633276450377584e-05, "loss": 1.1534, "step": 3707 }, { "epoch": 1.14, "learning_rate": 1.9633009003058342e-05, "loss": 0.922, "step": 3708 }, { "epoch": 1.14, "learning_rate": 1.9632741460074444e-05, "loss": 1.1254, "step": 3709 }, { "epoch": 1.14, "learning_rate": 1.963247382142854e-05, "loss": 1.218, "step": 3710 }, { "epoch": 1.14, "learning_rate": 1.9632206087123296e-05, "loss": 1.3705, "step": 3711 }, { "epoch": 1.14, "learning_rate": 1.963193825716137e-05, "loss": 1.4299, "step": 3712 }, { "epoch": 1.14, "learning_rate": 1.9631670331545418e-05, "loss": 1.1646, "step": 3713 }, { "epoch": 1.14, "learning_rate": 1.9631402310278105e-05, "loss": 1.6633, "step": 3714 }, { "epoch": 1.14, "learning_rate": 1.963113419336209e-05, "loss": 1.3264, "step": 3715 }, { "epoch": 1.14, "learning_rate": 1.9630865980800032e-05, "loss": 1.6507, "step": 3716 }, { "epoch": 1.14, "learning_rate": 1.9630597672594606e-05, "loss": 1.5162, "step": 3717 }, { "epoch": 1.14, "learning_rate": 1.9630329268748463e-05, "loss": 1.1715, "step": 3718 }, { "epoch": 1.14, "learning_rate": 1.9630060769264275e-05, "loss": 1.217, "step": 3719 }, { "epoch": 1.14, "learning_rate": 1.9629792174144713e-05, "loss": 1.2491, "step": 3720 }, { "epoch": 1.14, "learning_rate": 1.9629523483392435e-05, "loss": 1.4668, "step": 3721 }, { "epoch": 1.15, "learning_rate": 1.9629254697010117e-05, "loss": 1.6147, "step": 3722 }, { "epoch": 1.15, "learning_rate": 1.9628985815000425e-05, "loss": 1.2247, "step": 3723 }, { "epoch": 1.15, "learning_rate": 1.962871683736603e-05, "loss": 1.2789, "step": 3724 }, { "epoch": 1.15, "learning_rate": 1.96284477641096e-05, "loss": 0.9987, "step": 3725 }, { "epoch": 1.15, "learning_rate": 1.962817859523381e-05, "loss": 1.1824, "step": 3726 }, { "epoch": 1.15, "learning_rate": 1.9627909330741335e-05, "loss": 0.9055, "step": 3727 }, { "epoch": 1.15, "learning_rate": 1.9627639970634846e-05, "loss": 0.9502, "step": 3728 }, { "epoch": 1.15, "learning_rate": 1.962737051491702e-05, "loss": 0.9906, "step": 3729 }, { "epoch": 1.15, "learning_rate": 1.962710096359053e-05, "loss": 1.1647, "step": 3730 }, { "epoch": 1.15, "learning_rate": 1.9626831316658058e-05, "loss": 1.3378, "step": 3731 }, { "epoch": 1.15, "learning_rate": 1.9626561574122275e-05, "loss": 1.4095, "step": 3732 }, { "epoch": 1.15, "learning_rate": 1.9626291735985865e-05, "loss": 0.9295, "step": 3733 }, { "epoch": 1.15, "learning_rate": 1.9626021802251507e-05, "loss": 1.1032, "step": 3734 }, { "epoch": 1.15, "learning_rate": 1.9625751772921878e-05, "loss": 1.0482, "step": 3735 }, { "epoch": 1.15, "learning_rate": 1.9625481647999668e-05, "loss": 1.2916, "step": 3736 }, { "epoch": 1.15, "learning_rate": 1.962521142748755e-05, "loss": 1.4035, "step": 3737 }, { "epoch": 1.15, "learning_rate": 1.9624941111388215e-05, "loss": 1.3082, "step": 3738 }, { "epoch": 1.15, "learning_rate": 1.962467069970434e-05, "loss": 1.0557, "step": 3739 }, { "epoch": 1.15, "learning_rate": 1.9624400192438616e-05, "loss": 1.3305, "step": 3740 }, { "epoch": 1.15, "learning_rate": 1.962412958959373e-05, "loss": 1.2458, "step": 3741 }, { "epoch": 1.15, "learning_rate": 1.962385889117236e-05, "loss": 1.2214, "step": 3742 }, { "epoch": 1.15, "learning_rate": 1.962358809717721e-05, "loss": 1.8939, "step": 3743 }, { "epoch": 1.15, "learning_rate": 1.962331720761096e-05, "loss": 1.3082, "step": 3744 }, { "epoch": 1.15, "learning_rate": 1.9623046222476298e-05, "loss": 1.0573, "step": 3745 }, { "epoch": 1.15, "learning_rate": 1.962277514177592e-05, "loss": 1.3147, "step": 3746 }, { "epoch": 1.15, "learning_rate": 1.9622503965512516e-05, "loss": 1.2278, "step": 3747 }, { "epoch": 1.15, "learning_rate": 1.962223269368878e-05, "loss": 0.6946, "step": 3748 }, { "epoch": 1.15, "learning_rate": 1.96219613263074e-05, "loss": 1.1511, "step": 3749 }, { "epoch": 1.15, "learning_rate": 1.9621689863371083e-05, "loss": 1.1298, "step": 3750 }, { "epoch": 1.15, "learning_rate": 1.9621418304882514e-05, "loss": 1.3822, "step": 3751 }, { "epoch": 1.15, "learning_rate": 1.9621146650844392e-05, "loss": 1.1715, "step": 3752 }, { "epoch": 1.15, "learning_rate": 1.962087490125942e-05, "loss": 1.1127, "step": 3753 }, { "epoch": 1.16, "learning_rate": 1.9620603056130293e-05, "loss": 1.1582, "step": 3754 }, { "epoch": 1.16, "learning_rate": 1.9620331115459716e-05, "loss": 1.0159, "step": 3755 }, { "epoch": 1.16, "learning_rate": 1.9620059079250378e-05, "loss": 1.067, "step": 3756 }, { "epoch": 1.16, "learning_rate": 1.961978694750499e-05, "loss": 0.9342, "step": 3757 }, { "epoch": 1.16, "learning_rate": 1.961951472022625e-05, "loss": 1.345, "step": 3758 }, { "epoch": 1.16, "learning_rate": 1.9619242397416866e-05, "loss": 1.5409, "step": 3759 }, { "epoch": 1.16, "learning_rate": 1.9618969979079537e-05, "loss": 1.0804, "step": 3760 }, { "epoch": 1.16, "learning_rate": 1.961869746521697e-05, "loss": 1.1513, "step": 3761 }, { "epoch": 1.16, "learning_rate": 1.9618424855831875e-05, "loss": 1.0532, "step": 3762 }, { "epoch": 1.16, "learning_rate": 1.9618152150926953e-05, "loss": 1.1538, "step": 3763 }, { "epoch": 1.16, "learning_rate": 1.961787935050492e-05, "loss": 0.8475, "step": 3764 }, { "epoch": 1.16, "learning_rate": 1.961760645456848e-05, "loss": 1.3376, "step": 3765 }, { "epoch": 1.16, "learning_rate": 1.9617333463120342e-05, "loss": 1.0706, "step": 3766 }, { "epoch": 1.16, "learning_rate": 1.961706037616322e-05, "loss": 0.9491, "step": 3767 }, { "epoch": 1.16, "learning_rate": 1.9616787193699826e-05, "loss": 1.0695, "step": 3768 }, { "epoch": 1.16, "learning_rate": 1.9616513915732872e-05, "loss": 1.0269, "step": 3769 }, { "epoch": 1.16, "learning_rate": 1.961624054226507e-05, "loss": 1.0507, "step": 3770 }, { "epoch": 1.16, "learning_rate": 1.961596707329914e-05, "loss": 1.1743, "step": 3771 }, { "epoch": 1.16, "learning_rate": 1.9615693508837794e-05, "loss": 1.2736, "step": 3772 }, { "epoch": 1.16, "learning_rate": 1.961541984888375e-05, "loss": 1.2349, "step": 3773 }, { "epoch": 1.16, "learning_rate": 1.9615146093439722e-05, "loss": 1.048, "step": 3774 }, { "epoch": 1.16, "learning_rate": 1.9614872242508434e-05, "loss": 1.2539, "step": 3775 }, { "epoch": 1.16, "learning_rate": 1.9614598296092603e-05, "loss": 0.9278, "step": 3776 }, { "epoch": 1.16, "learning_rate": 1.961432425419495e-05, "loss": 1.1126, "step": 3777 }, { "epoch": 1.16, "learning_rate": 1.9614050116818192e-05, "loss": 1.1121, "step": 3778 }, { "epoch": 1.16, "learning_rate": 1.961377588396506e-05, "loss": 1.1613, "step": 3779 }, { "epoch": 1.16, "learning_rate": 1.9613501555638273e-05, "loss": 1.2085, "step": 3780 }, { "epoch": 1.16, "learning_rate": 1.9613227131840554e-05, "loss": 1.3282, "step": 3781 }, { "epoch": 1.16, "learning_rate": 1.9612952612574633e-05, "loss": 0.9037, "step": 3782 }, { "epoch": 1.16, "learning_rate": 1.9612677997843233e-05, "loss": 1.2207, "step": 3783 }, { "epoch": 1.16, "learning_rate": 1.9612403287649076e-05, "loss": 1.3642, "step": 3784 }, { "epoch": 1.16, "learning_rate": 1.9612128481994898e-05, "loss": 1.3799, "step": 3785 }, { "epoch": 1.16, "learning_rate": 1.9611853580883427e-05, "loss": 1.0669, "step": 3786 }, { "epoch": 1.17, "learning_rate": 1.961157858431739e-05, "loss": 1.0736, "step": 3787 }, { "epoch": 1.17, "learning_rate": 1.961130349229952e-05, "loss": 1.2313, "step": 3788 }, { "epoch": 1.17, "learning_rate": 1.9611028304832547e-05, "loss": 1.3756, "step": 3789 }, { "epoch": 1.17, "learning_rate": 1.9610753021919207e-05, "loss": 1.066, "step": 3790 }, { "epoch": 1.17, "learning_rate": 1.961047764356223e-05, "loss": 1.0074, "step": 3791 }, { "epoch": 1.17, "learning_rate": 1.9610202169764355e-05, "loss": 1.2459, "step": 3792 }, { "epoch": 1.17, "learning_rate": 1.9609926600528312e-05, "loss": 1.2632, "step": 3793 }, { "epoch": 1.17, "learning_rate": 1.9609650935856847e-05, "loss": 1.1684, "step": 3794 }, { "epoch": 1.17, "learning_rate": 1.9609375175752686e-05, "loss": 1.287, "step": 3795 }, { "epoch": 1.17, "learning_rate": 1.9609099320218575e-05, "loss": 1.0401, "step": 3796 }, { "epoch": 1.17, "learning_rate": 1.960882336925725e-05, "loss": 1.1742, "step": 3797 }, { "epoch": 1.17, "learning_rate": 1.9608547322871455e-05, "loss": 1.3155, "step": 3798 }, { "epoch": 1.17, "learning_rate": 1.9608271181063932e-05, "loss": 1.3163, "step": 3799 }, { "epoch": 1.17, "learning_rate": 1.960799494383742e-05, "loss": 1.2607, "step": 3800 }, { "epoch": 1.17, "learning_rate": 1.960771861119466e-05, "loss": 1.2047, "step": 3801 }, { "epoch": 1.17, "learning_rate": 1.9607442183138403e-05, "loss": 1.3513, "step": 3802 }, { "epoch": 1.17, "learning_rate": 1.960716565967139e-05, "loss": 1.0716, "step": 3803 }, { "epoch": 1.17, "learning_rate": 1.9606889040796363e-05, "loss": 1.2391, "step": 3804 }, { "epoch": 1.17, "learning_rate": 1.960661232651608e-05, "loss": 1.0884, "step": 3805 }, { "epoch": 1.17, "learning_rate": 1.9606335516833278e-05, "loss": 1.1033, "step": 3806 }, { "epoch": 1.17, "learning_rate": 1.9606058611750712e-05, "loss": 1.4092, "step": 3807 }, { "epoch": 1.17, "learning_rate": 1.9605781611271134e-05, "loss": 1.0033, "step": 3808 }, { "epoch": 1.17, "learning_rate": 1.960550451539729e-05, "loss": 1.171, "step": 3809 }, { "epoch": 1.17, "learning_rate": 1.9605227324131933e-05, "loss": 0.92, "step": 3810 }, { "epoch": 1.17, "learning_rate": 1.9604950037477816e-05, "loss": 1.2134, "step": 3811 }, { "epoch": 1.17, "learning_rate": 1.960467265543769e-05, "loss": 1.111, "step": 3812 }, { "epoch": 1.17, "learning_rate": 1.960439517801432e-05, "loss": 1.22, "step": 3813 }, { "epoch": 1.17, "learning_rate": 1.960411760521045e-05, "loss": 1.0866, "step": 3814 }, { "epoch": 1.17, "learning_rate": 1.960383993702884e-05, "loss": 1.5624, "step": 3815 }, { "epoch": 1.17, "learning_rate": 1.960356217347225e-05, "loss": 1.1257, "step": 3816 }, { "epoch": 1.17, "learning_rate": 1.9603284314543437e-05, "loss": 1.1642, "step": 3817 }, { "epoch": 1.17, "learning_rate": 1.9603006360245157e-05, "loss": 1.203, "step": 3818 }, { "epoch": 1.18, "learning_rate": 1.9602728310580176e-05, "loss": 1.1111, "step": 3819 }, { "epoch": 1.18, "learning_rate": 1.9602450165551252e-05, "loss": 1.1495, "step": 3820 }, { "epoch": 1.18, "learning_rate": 1.960217192516115e-05, "loss": 1.2175, "step": 3821 }, { "epoch": 1.18, "learning_rate": 1.9601893589412627e-05, "loss": 1.0811, "step": 3822 }, { "epoch": 1.18, "learning_rate": 1.9601615158308452e-05, "loss": 1.2775, "step": 3823 }, { "epoch": 1.18, "learning_rate": 1.9601336631851393e-05, "loss": 1.0611, "step": 3824 }, { "epoch": 1.18, "learning_rate": 1.960105801004421e-05, "loss": 1.0663, "step": 3825 }, { "epoch": 1.18, "learning_rate": 1.9600779292889675e-05, "loss": 1.1696, "step": 3826 }, { "epoch": 1.18, "learning_rate": 1.9600500480390548e-05, "loss": 0.9928, "step": 3827 }, { "epoch": 1.18, "learning_rate": 1.9600221572549607e-05, "loss": 1.2662, "step": 3828 }, { "epoch": 1.18, "learning_rate": 1.9599942569369618e-05, "loss": 0.9341, "step": 3829 }, { "epoch": 1.18, "learning_rate": 1.959966347085335e-05, "loss": 1.2443, "step": 3830 }, { "epoch": 1.18, "learning_rate": 1.9599384277003576e-05, "loss": 1.4058, "step": 3831 }, { "epoch": 1.18, "learning_rate": 1.9599104987823072e-05, "loss": 1.1419, "step": 3832 }, { "epoch": 1.18, "learning_rate": 1.959882560331461e-05, "loss": 1.318, "step": 3833 }, { "epoch": 1.18, "learning_rate": 1.9598546123480956e-05, "loss": 1.2065, "step": 3834 }, { "epoch": 1.18, "learning_rate": 1.9598266548324897e-05, "loss": 1.0261, "step": 3835 }, { "epoch": 1.18, "learning_rate": 1.9597986877849208e-05, "loss": 1.2458, "step": 3836 }, { "epoch": 1.18, "learning_rate": 1.9597707112056658e-05, "loss": 1.1278, "step": 3837 }, { "epoch": 1.18, "learning_rate": 1.9597427250950035e-05, "loss": 1.1203, "step": 3838 }, { "epoch": 1.18, "learning_rate": 1.9597147294532113e-05, "loss": 1.2184, "step": 3839 }, { "epoch": 1.18, "learning_rate": 1.9596867242805672e-05, "loss": 1.1402, "step": 3840 }, { "epoch": 1.18, "learning_rate": 1.9596587095773496e-05, "loss": 1.2598, "step": 3841 }, { "epoch": 1.18, "learning_rate": 1.9596306853438363e-05, "loss": 1.3333, "step": 3842 }, { "epoch": 1.18, "learning_rate": 1.9596026515803062e-05, "loss": 1.1889, "step": 3843 }, { "epoch": 1.18, "learning_rate": 1.959574608287037e-05, "loss": 1.1802, "step": 3844 }, { "epoch": 1.18, "learning_rate": 1.9595465554643083e-05, "loss": 1.0752, "step": 3845 }, { "epoch": 1.18, "learning_rate": 1.9595184931123974e-05, "loss": 1.5005, "step": 3846 }, { "epoch": 1.18, "learning_rate": 1.9594904212315837e-05, "loss": 1.2342, "step": 3847 }, { "epoch": 1.18, "learning_rate": 1.9594623398221458e-05, "loss": 1.1762, "step": 3848 }, { "epoch": 1.18, "learning_rate": 1.9594342488843627e-05, "loss": 1.124, "step": 3849 }, { "epoch": 1.18, "learning_rate": 1.959406148418513e-05, "loss": 1.1704, "step": 3850 }, { "epoch": 1.18, "learning_rate": 1.9593780384248762e-05, "loss": 1.4654, "step": 3851 }, { "epoch": 1.19, "learning_rate": 1.9593499189037312e-05, "loss": 1.5451, "step": 3852 }, { "epoch": 1.19, "learning_rate": 1.9593217898553574e-05, "loss": 0.9881, "step": 3853 }, { "epoch": 1.19, "learning_rate": 1.959293651280034e-05, "loss": 1.4531, "step": 3854 }, { "epoch": 1.19, "learning_rate": 1.959265503178041e-05, "loss": 1.072, "step": 3855 }, { "epoch": 1.19, "learning_rate": 1.959237345549657e-05, "loss": 1.1488, "step": 3856 }, { "epoch": 1.19, "learning_rate": 1.959209178395162e-05, "loss": 1.1251, "step": 3857 }, { "epoch": 1.19, "learning_rate": 1.959181001714836e-05, "loss": 1.2492, "step": 3858 }, { "epoch": 1.19, "learning_rate": 1.9591528155089588e-05, "loss": 1.0192, "step": 3859 }, { "epoch": 1.19, "learning_rate": 1.95912461977781e-05, "loss": 1.301, "step": 3860 }, { "epoch": 1.19, "learning_rate": 1.9590964145216698e-05, "loss": 1.0665, "step": 3861 }, { "epoch": 1.19, "learning_rate": 1.959068199740818e-05, "loss": 1.1809, "step": 3862 }, { "epoch": 1.19, "learning_rate": 1.9590399754355354e-05, "loss": 0.9856, "step": 3863 }, { "epoch": 1.19, "learning_rate": 1.9590117416061022e-05, "loss": 1.1293, "step": 3864 }, { "epoch": 1.19, "learning_rate": 1.9589834982527983e-05, "loss": 1.2562, "step": 3865 }, { "epoch": 1.19, "learning_rate": 1.958955245375904e-05, "loss": 0.9755, "step": 3866 }, { "epoch": 1.19, "learning_rate": 1.958926982975701e-05, "loss": 1.3374, "step": 3867 }, { "epoch": 1.19, "learning_rate": 1.958898711052469e-05, "loss": 1.1054, "step": 3868 }, { "epoch": 1.19, "learning_rate": 1.958870429606489e-05, "loss": 1.0455, "step": 3869 }, { "epoch": 1.19, "learning_rate": 1.958842138638042e-05, "loss": 1.7081, "step": 3870 }, { "epoch": 1.19, "learning_rate": 1.9588138381474085e-05, "loss": 1.2679, "step": 3871 }, { "epoch": 1.19, "learning_rate": 1.9587855281348705e-05, "loss": 1.1781, "step": 3872 }, { "epoch": 1.19, "learning_rate": 1.9587572086007084e-05, "loss": 1.0375, "step": 3873 }, { "epoch": 1.19, "learning_rate": 1.9587288795452034e-05, "loss": 1.2891, "step": 3874 }, { "epoch": 1.19, "learning_rate": 1.958700540968637e-05, "loss": 1.1709, "step": 3875 }, { "epoch": 1.19, "learning_rate": 1.958672192871291e-05, "loss": 1.1621, "step": 3876 }, { "epoch": 1.19, "learning_rate": 1.9586438352534466e-05, "loss": 0.9459, "step": 3877 }, { "epoch": 1.19, "learning_rate": 1.958615468115385e-05, "loss": 1.0947, "step": 3878 }, { "epoch": 1.19, "learning_rate": 1.9585870914573887e-05, "loss": 1.1352, "step": 3879 }, { "epoch": 1.19, "learning_rate": 1.958558705279739e-05, "loss": 1.1974, "step": 3880 }, { "epoch": 1.19, "learning_rate": 1.9585303095827174e-05, "loss": 1.0988, "step": 3881 }, { "epoch": 1.19, "learning_rate": 1.958501904366607e-05, "loss": 1.1989, "step": 3882 }, { "epoch": 1.19, "learning_rate": 1.9584734896316894e-05, "loss": 1.0324, "step": 3883 }, { "epoch": 1.2, "learning_rate": 1.9584450653782464e-05, "loss": 1.1823, "step": 3884 }, { "epoch": 1.2, "learning_rate": 1.9584166316065606e-05, "loss": 1.3419, "step": 3885 }, { "epoch": 1.2, "learning_rate": 1.9583881883169143e-05, "loss": 1.5887, "step": 3886 }, { "epoch": 1.2, "learning_rate": 1.9583597355095902e-05, "loss": 1.0906, "step": 3887 }, { "epoch": 1.2, "learning_rate": 1.9583312731848707e-05, "loss": 1.045, "step": 3888 }, { "epoch": 1.2, "learning_rate": 1.9583028013430384e-05, "loss": 1.1674, "step": 3889 }, { "epoch": 1.2, "learning_rate": 1.958274319984376e-05, "loss": 1.1818, "step": 3890 }, { "epoch": 1.2, "learning_rate": 1.9582458291091664e-05, "loss": 1.2656, "step": 3891 }, { "epoch": 1.2, "learning_rate": 1.9582173287176925e-05, "loss": 1.5492, "step": 3892 }, { "epoch": 1.2, "learning_rate": 1.9581888188102375e-05, "loss": 1.102, "step": 3893 }, { "epoch": 1.2, "learning_rate": 1.9581602993870846e-05, "loss": 1.1836, "step": 3894 }, { "epoch": 1.2, "learning_rate": 1.9581317704485164e-05, "loss": 1.1429, "step": 3895 }, { "epoch": 1.2, "learning_rate": 1.958103231994817e-05, "loss": 1.286, "step": 3896 }, { "epoch": 1.2, "learning_rate": 1.9580746840262697e-05, "loss": 1.3863, "step": 3897 }, { "epoch": 1.2, "learning_rate": 1.9580461265431577e-05, "loss": 1.162, "step": 3898 }, { "epoch": 1.2, "learning_rate": 1.9580175595457644e-05, "loss": 1.1886, "step": 3899 }, { "epoch": 1.2, "learning_rate": 1.957988983034374e-05, "loss": 1.0803, "step": 3900 }, { "epoch": 1.2, "learning_rate": 1.9579603970092702e-05, "loss": 1.5631, "step": 3901 }, { "epoch": 1.2, "learning_rate": 1.957931801470737e-05, "loss": 1.2531, "step": 3902 }, { "epoch": 1.2, "learning_rate": 1.9579031964190575e-05, "loss": 1.3188, "step": 3903 }, { "epoch": 1.2, "learning_rate": 1.9578745818545165e-05, "loss": 1.0848, "step": 3904 }, { "epoch": 1.2, "learning_rate": 1.9578459577773987e-05, "loss": 1.1167, "step": 3905 }, { "epoch": 1.2, "learning_rate": 1.957817324187987e-05, "loss": 1.0473, "step": 3906 }, { "epoch": 1.2, "learning_rate": 1.9577886810865673e-05, "loss": 1.211, "step": 3907 }, { "epoch": 1.2, "learning_rate": 1.957760028473423e-05, "loss": 1.8567, "step": 3908 }, { "epoch": 1.2, "learning_rate": 1.9577313663488387e-05, "loss": 1.2231, "step": 3909 }, { "epoch": 1.2, "learning_rate": 1.957702694713099e-05, "loss": 2.1259, "step": 3910 }, { "epoch": 1.2, "learning_rate": 1.9576740135664894e-05, "loss": 1.1726, "step": 3911 }, { "epoch": 1.2, "learning_rate": 1.9576453229092942e-05, "loss": 1.1114, "step": 3912 }, { "epoch": 1.2, "learning_rate": 1.9576166227417983e-05, "loss": 0.7391, "step": 3913 }, { "epoch": 1.2, "learning_rate": 1.9575879130642868e-05, "loss": 1.054, "step": 3914 }, { "epoch": 1.2, "learning_rate": 1.9575591938770446e-05, "loss": 1.1672, "step": 3915 }, { "epoch": 1.2, "learning_rate": 1.9575304651803572e-05, "loss": 1.7597, "step": 3916 }, { "epoch": 1.21, "learning_rate": 1.9575017269745097e-05, "loss": 1.1643, "step": 3917 }, { "epoch": 1.21, "learning_rate": 1.9574729792597877e-05, "loss": 1.176, "step": 3918 }, { "epoch": 1.21, "learning_rate": 1.9574442220364768e-05, "loss": 1.2244, "step": 3919 }, { "epoch": 1.21, "learning_rate": 1.9574154553048622e-05, "loss": 1.1195, "step": 3920 }, { "epoch": 1.21, "learning_rate": 1.9573866790652296e-05, "loss": 1.1706, "step": 3921 }, { "epoch": 1.21, "learning_rate": 1.9573578933178646e-05, "loss": 1.302, "step": 3922 }, { "epoch": 1.21, "learning_rate": 1.957329098063054e-05, "loss": 1.4469, "step": 3923 }, { "epoch": 1.21, "learning_rate": 1.9573002933010833e-05, "loss": 1.4378, "step": 3924 }, { "epoch": 1.21, "learning_rate": 1.957271479032238e-05, "loss": 1.3259, "step": 3925 }, { "epoch": 1.21, "learning_rate": 1.9572426552568045e-05, "loss": 1.0823, "step": 3926 }, { "epoch": 1.21, "learning_rate": 1.9572138219750693e-05, "loss": 1.4274, "step": 3927 }, { "epoch": 1.21, "learning_rate": 1.957184979187319e-05, "loss": 1.0751, "step": 3928 }, { "epoch": 1.21, "learning_rate": 1.9571561268938396e-05, "loss": 1.1251, "step": 3929 }, { "epoch": 1.21, "learning_rate": 1.9571272650949177e-05, "loss": 1.2586, "step": 3930 }, { "epoch": 1.21, "learning_rate": 1.9570983937908398e-05, "loss": 1.3095, "step": 3931 }, { "epoch": 1.21, "learning_rate": 1.9570695129818928e-05, "loss": 1.6412, "step": 3932 }, { "epoch": 1.21, "learning_rate": 1.9570406226683633e-05, "loss": 1.0748, "step": 3933 }, { "epoch": 1.21, "learning_rate": 1.9570117228505387e-05, "loss": 1.2791, "step": 3934 }, { "epoch": 1.21, "learning_rate": 1.9569828135287056e-05, "loss": 1.4553, "step": 3935 }, { "epoch": 1.21, "learning_rate": 1.956953894703151e-05, "loss": 1.2546, "step": 3936 }, { "epoch": 1.21, "learning_rate": 1.9569249663741624e-05, "loss": 1.3223, "step": 3937 }, { "epoch": 1.21, "learning_rate": 1.956896028542027e-05, "loss": 1.1634, "step": 3938 }, { "epoch": 1.21, "learning_rate": 1.956867081207032e-05, "loss": 1.5839, "step": 3939 }, { "epoch": 1.21, "learning_rate": 1.956838124369465e-05, "loss": 1.241, "step": 3940 }, { "epoch": 1.21, "learning_rate": 1.9568091580296138e-05, "loss": 1.3581, "step": 3941 }, { "epoch": 1.21, "learning_rate": 1.9567801821877656e-05, "loss": 1.2439, "step": 3942 }, { "epoch": 1.21, "learning_rate": 1.9567511968442088e-05, "loss": 1.2488, "step": 3943 }, { "epoch": 1.21, "learning_rate": 1.9567222019992303e-05, "loss": 1.2828, "step": 3944 }, { "epoch": 1.21, "learning_rate": 1.956693197653119e-05, "loss": 1.241, "step": 3945 }, { "epoch": 1.21, "learning_rate": 1.9566641838061624e-05, "loss": 1.0198, "step": 3946 }, { "epoch": 1.21, "learning_rate": 1.9566351604586485e-05, "loss": 1.2472, "step": 3947 }, { "epoch": 1.21, "learning_rate": 1.9566061276108664e-05, "loss": 0.966, "step": 3948 }, { "epoch": 1.22, "learning_rate": 1.9565770852631035e-05, "loss": 1.2021, "step": 3949 }, { "epoch": 1.22, "learning_rate": 1.9565480334156483e-05, "loss": 1.2339, "step": 3950 }, { "epoch": 1.22, "learning_rate": 1.95651897206879e-05, "loss": 1.5436, "step": 3951 }, { "epoch": 1.22, "learning_rate": 1.9564899012228164e-05, "loss": 1.2514, "step": 3952 }, { "epoch": 1.22, "learning_rate": 1.9564608208780168e-05, "loss": 1.2957, "step": 3953 }, { "epoch": 1.22, "learning_rate": 1.9564317310346795e-05, "loss": 1.1525, "step": 3954 }, { "epoch": 1.22, "learning_rate": 1.956402631693094e-05, "loss": 1.0429, "step": 3955 }, { "epoch": 1.22, "learning_rate": 1.9563735228535482e-05, "loss": 1.379, "step": 3956 }, { "epoch": 1.22, "learning_rate": 1.9563444045163326e-05, "loss": 1.2637, "step": 3957 }, { "epoch": 1.22, "learning_rate": 1.9563152766817356e-05, "loss": 1.1014, "step": 3958 }, { "epoch": 1.22, "learning_rate": 1.9562861393500464e-05, "loss": 0.8845, "step": 3959 }, { "epoch": 1.22, "learning_rate": 1.956256992521554e-05, "loss": 1.3921, "step": 3960 }, { "epoch": 1.22, "learning_rate": 1.9562278361965494e-05, "loss": 1.1854, "step": 3961 }, { "epoch": 1.22, "learning_rate": 1.9561986703753203e-05, "loss": 1.0733, "step": 3962 }, { "epoch": 1.22, "learning_rate": 1.9561694950581573e-05, "loss": 1.0452, "step": 3963 }, { "epoch": 1.22, "learning_rate": 1.95614031024535e-05, "loss": 0.82, "step": 3964 }, { "epoch": 1.22, "learning_rate": 1.9561111159371883e-05, "loss": 1.3063, "step": 3965 }, { "epoch": 1.22, "learning_rate": 1.956081912133962e-05, "loss": 1.422, "step": 3966 }, { "epoch": 1.22, "learning_rate": 1.9560526988359613e-05, "loss": 1.1806, "step": 3967 }, { "epoch": 1.22, "learning_rate": 1.956023476043476e-05, "loss": 1.0196, "step": 3968 }, { "epoch": 1.22, "learning_rate": 1.9559942437567965e-05, "loss": 0.9961, "step": 3969 }, { "epoch": 1.22, "learning_rate": 1.955965001976213e-05, "loss": 0.9373, "step": 3970 }, { "epoch": 1.22, "learning_rate": 1.9559357507020163e-05, "loss": 0.9352, "step": 3971 }, { "epoch": 1.22, "learning_rate": 1.9559064899344965e-05, "loss": 0.9301, "step": 3972 }, { "epoch": 1.22, "learning_rate": 1.9558772196739438e-05, "loss": 0.9881, "step": 3973 }, { "epoch": 1.22, "learning_rate": 1.9558479399206498e-05, "loss": 1.6275, "step": 3974 }, { "epoch": 1.22, "learning_rate": 1.9558186506749045e-05, "loss": 1.2618, "step": 3975 }, { "epoch": 1.22, "learning_rate": 1.9557893519369994e-05, "loss": 1.8165, "step": 3976 }, { "epoch": 1.22, "learning_rate": 1.9557600437072254e-05, "loss": 1.4299, "step": 3977 }, { "epoch": 1.22, "learning_rate": 1.955730725985873e-05, "loss": 1.3408, "step": 3978 }, { "epoch": 1.22, "learning_rate": 1.9557013987732335e-05, "loss": 1.2068, "step": 3979 }, { "epoch": 1.22, "learning_rate": 1.9556720620695983e-05, "loss": 1.3139, "step": 3980 }, { "epoch": 1.22, "learning_rate": 1.955642715875259e-05, "loss": 1.109, "step": 3981 }, { "epoch": 1.23, "learning_rate": 1.9556133601905066e-05, "loss": 1.1448, "step": 3982 }, { "epoch": 1.23, "learning_rate": 1.955583995015633e-05, "loss": 1.0304, "step": 3983 }, { "epoch": 1.23, "learning_rate": 1.9555546203509297e-05, "loss": 1.2958, "step": 3984 }, { "epoch": 1.23, "learning_rate": 1.955525236196688e-05, "loss": 1.4147, "step": 3985 }, { "epoch": 1.23, "learning_rate": 1.9554958425532002e-05, "loss": 1.1653, "step": 3986 }, { "epoch": 1.23, "learning_rate": 1.955466439420758e-05, "loss": 0.9219, "step": 3987 }, { "epoch": 1.23, "learning_rate": 1.9554370267996537e-05, "loss": 1.2637, "step": 3988 }, { "epoch": 1.23, "learning_rate": 1.955407604690179e-05, "loss": 1.2067, "step": 3989 }, { "epoch": 1.23, "learning_rate": 1.9553781730926266e-05, "loss": 1.6337, "step": 3990 }, { "epoch": 1.23, "learning_rate": 1.9553487320072883e-05, "loss": 1.428, "step": 3991 }, { "epoch": 1.23, "learning_rate": 1.9553192814344563e-05, "loss": 1.6993, "step": 3992 }, { "epoch": 1.23, "learning_rate": 1.9552898213744233e-05, "loss": 0.972, "step": 3993 }, { "epoch": 1.23, "learning_rate": 1.9552603518274823e-05, "loss": 1.2791, "step": 3994 }, { "epoch": 1.23, "learning_rate": 1.9552308727939255e-05, "loss": 1.0967, "step": 3995 }, { "epoch": 1.23, "learning_rate": 1.955201384274046e-05, "loss": 1.3703, "step": 3996 }, { "epoch": 1.23, "learning_rate": 1.9551718862681363e-05, "loss": 1.3188, "step": 3997 }, { "epoch": 1.23, "learning_rate": 1.9551423787764893e-05, "loss": 0.978, "step": 3998 }, { "epoch": 1.23, "learning_rate": 1.955112861799398e-05, "loss": 1.0958, "step": 3999 }, { "epoch": 1.23, "learning_rate": 1.955083335337156e-05, "loss": 1.0516, "step": 4000 }, { "epoch": 1.23, "learning_rate": 1.9550537993900566e-05, "loss": 1.1178, "step": 4001 }, { "epoch": 1.23, "learning_rate": 1.955024253958392e-05, "loss": 1.3536, "step": 4002 }, { "epoch": 1.23, "learning_rate": 1.954994699042457e-05, "loss": 1.0782, "step": 4003 }, { "epoch": 1.23, "learning_rate": 1.9549651346425443e-05, "loss": 1.1446, "step": 4004 }, { "epoch": 1.23, "learning_rate": 1.954935560758948e-05, "loss": 1.1421, "step": 4005 }, { "epoch": 1.23, "learning_rate": 1.954905977391961e-05, "loss": 1.1131, "step": 4006 }, { "epoch": 1.23, "learning_rate": 1.9548763845418777e-05, "loss": 0.8995, "step": 4007 }, { "epoch": 1.23, "learning_rate": 1.9548467822089918e-05, "loss": 1.0199, "step": 4008 }, { "epoch": 1.23, "learning_rate": 1.9548171703935975e-05, "loss": 1.1641, "step": 4009 }, { "epoch": 1.23, "learning_rate": 1.9547875490959884e-05, "loss": 1.2789, "step": 4010 }, { "epoch": 1.23, "learning_rate": 1.9547579183164595e-05, "loss": 1.1142, "step": 4011 }, { "epoch": 1.23, "learning_rate": 1.954728278055304e-05, "loss": 1.1547, "step": 4012 }, { "epoch": 1.23, "learning_rate": 1.9546986283128172e-05, "loss": 0.8144, "step": 4013 }, { "epoch": 1.24, "learning_rate": 1.954668969089293e-05, "loss": 1.2153, "step": 4014 }, { "epoch": 1.24, "learning_rate": 1.954639300385026e-05, "loss": 1.2215, "step": 4015 }, { "epoch": 1.24, "learning_rate": 1.9546096222003112e-05, "loss": 1.0817, "step": 4016 }, { "epoch": 1.24, "learning_rate": 1.9545799345354427e-05, "loss": 1.0208, "step": 4017 }, { "epoch": 1.24, "learning_rate": 1.954550237390716e-05, "loss": 1.1506, "step": 4018 }, { "epoch": 1.24, "learning_rate": 1.9545205307664258e-05, "loss": 0.9657, "step": 4019 }, { "epoch": 1.24, "learning_rate": 1.9544908146628666e-05, "loss": 1.1102, "step": 4020 }, { "epoch": 1.24, "learning_rate": 1.954461089080334e-05, "loss": 1.3018, "step": 4021 }, { "epoch": 1.24, "learning_rate": 1.9544313540191234e-05, "loss": 0.9121, "step": 4022 }, { "epoch": 1.24, "learning_rate": 1.9544016094795294e-05, "loss": 1.1408, "step": 4023 }, { "epoch": 1.24, "learning_rate": 1.954371855461848e-05, "loss": 1.0734, "step": 4024 }, { "epoch": 1.24, "learning_rate": 1.9543420919663747e-05, "loss": 1.3379, "step": 4025 }, { "epoch": 1.24, "learning_rate": 1.954312318993405e-05, "loss": 1.1203, "step": 4026 }, { "epoch": 1.24, "learning_rate": 1.954282536543234e-05, "loss": 1.1405, "step": 4027 }, { "epoch": 1.24, "learning_rate": 1.954252744616158e-05, "loss": 1.2647, "step": 4028 }, { "epoch": 1.24, "learning_rate": 1.9542229432124734e-05, "loss": 1.1198, "step": 4029 }, { "epoch": 1.24, "learning_rate": 1.954193132332475e-05, "loss": 1.0699, "step": 4030 }, { "epoch": 1.24, "learning_rate": 1.9541633119764594e-05, "loss": 1.0868, "step": 4031 }, { "epoch": 1.24, "learning_rate": 1.9541334821447228e-05, "loss": 1.2332, "step": 4032 }, { "epoch": 1.24, "learning_rate": 1.954103642837561e-05, "loss": 1.2252, "step": 4033 }, { "epoch": 1.24, "learning_rate": 1.9540737940552712e-05, "loss": 1.191, "step": 4034 }, { "epoch": 1.24, "learning_rate": 1.954043935798149e-05, "loss": 1.4326, "step": 4035 }, { "epoch": 1.24, "learning_rate": 1.9540140680664915e-05, "loss": 0.9453, "step": 4036 }, { "epoch": 1.24, "learning_rate": 1.953984190860595e-05, "loss": 1.4056, "step": 4037 }, { "epoch": 1.24, "learning_rate": 1.9539543041807563e-05, "loss": 1.359, "step": 4038 }, { "epoch": 1.24, "learning_rate": 1.9539244080272717e-05, "loss": 1.0052, "step": 4039 }, { "epoch": 1.24, "learning_rate": 1.9538945024004392e-05, "loss": 1.2007, "step": 4040 }, { "epoch": 1.24, "learning_rate": 1.953864587300555e-05, "loss": 1.006, "step": 4041 }, { "epoch": 1.24, "learning_rate": 1.953834662727916e-05, "loss": 1.2614, "step": 4042 }, { "epoch": 1.24, "learning_rate": 1.9538047286828202e-05, "loss": 1.3889, "step": 4043 }, { "epoch": 1.24, "learning_rate": 1.953774785165564e-05, "loss": 1.4437, "step": 4044 }, { "epoch": 1.24, "learning_rate": 1.9537448321764454e-05, "loss": 1.389, "step": 4045 }, { "epoch": 1.24, "learning_rate": 1.9537148697157614e-05, "loss": 1.1115, "step": 4046 }, { "epoch": 1.25, "learning_rate": 1.95368489778381e-05, "loss": 1.0551, "step": 4047 }, { "epoch": 1.25, "learning_rate": 1.9536549163808887e-05, "loss": 1.3444, "step": 4048 }, { "epoch": 1.25, "learning_rate": 1.953624925507295e-05, "loss": 1.518, "step": 4049 }, { "epoch": 1.25, "learning_rate": 1.953594925163327e-05, "loss": 1.0584, "step": 4050 }, { "epoch": 1.25, "learning_rate": 1.9535649153492823e-05, "loss": 1.179, "step": 4051 }, { "epoch": 1.25, "learning_rate": 1.953534896065459e-05, "loss": 1.5432, "step": 4052 }, { "epoch": 1.25, "learning_rate": 1.953504867312156e-05, "loss": 0.9331, "step": 4053 }, { "epoch": 1.25, "learning_rate": 1.9534748290896707e-05, "loss": 1.6695, "step": 4054 }, { "epoch": 1.25, "learning_rate": 1.953444781398301e-05, "loss": 1.0629, "step": 4055 }, { "epoch": 1.25, "learning_rate": 1.9534147242383467e-05, "loss": 1.2313, "step": 4056 }, { "epoch": 1.25, "learning_rate": 1.953384657610105e-05, "loss": 0.9746, "step": 4057 }, { "epoch": 1.25, "learning_rate": 1.9533545815138753e-05, "loss": 1.1282, "step": 4058 }, { "epoch": 1.25, "learning_rate": 1.9533244959499558e-05, "loss": 1.3581, "step": 4059 }, { "epoch": 1.25, "learning_rate": 1.9532944009186453e-05, "loss": 1.1111, "step": 4060 }, { "epoch": 1.25, "learning_rate": 1.953264296420243e-05, "loss": 1.1407, "step": 4061 }, { "epoch": 1.25, "learning_rate": 1.953234182455048e-05, "loss": 1.3823, "step": 4062 }, { "epoch": 1.25, "learning_rate": 1.9532040590233586e-05, "loss": 1.0899, "step": 4063 }, { "epoch": 1.25, "learning_rate": 1.9531739261254744e-05, "loss": 1.6966, "step": 4064 }, { "epoch": 1.25, "learning_rate": 1.9531437837616948e-05, "loss": 1.0062, "step": 4065 }, { "epoch": 1.25, "learning_rate": 1.953113631932319e-05, "loss": 1.2888, "step": 4066 }, { "epoch": 1.25, "learning_rate": 1.953083470637646e-05, "loss": 1.4964, "step": 4067 }, { "epoch": 1.25, "learning_rate": 1.9530532998779763e-05, "loss": 1.0498, "step": 4068 }, { "epoch": 1.25, "learning_rate": 1.953023119653609e-05, "loss": 1.1622, "step": 4069 }, { "epoch": 1.25, "learning_rate": 1.9529929299648432e-05, "loss": 1.2003, "step": 4070 }, { "epoch": 1.25, "learning_rate": 1.95296273081198e-05, "loss": 1.1307, "step": 4071 }, { "epoch": 1.25, "learning_rate": 1.952932522195318e-05, "loss": 1.0461, "step": 4072 }, { "epoch": 1.25, "learning_rate": 1.952902304115158e-05, "loss": 1.0059, "step": 4073 }, { "epoch": 1.25, "learning_rate": 1.9528720765718e-05, "loss": 1.1546, "step": 4074 }, { "epoch": 1.25, "learning_rate": 1.9528418395655443e-05, "loss": 1.1858, "step": 4075 }, { "epoch": 1.25, "learning_rate": 1.9528115930966907e-05, "loss": 1.2831, "step": 4076 }, { "epoch": 1.25, "learning_rate": 1.9527813371655396e-05, "loss": 1.243, "step": 4077 }, { "epoch": 1.25, "learning_rate": 1.9527510717723925e-05, "loss": 1.6684, "step": 4078 }, { "epoch": 1.26, "learning_rate": 1.9527207969175482e-05, "loss": 1.2344, "step": 4079 }, { "epoch": 1.26, "learning_rate": 1.952690512601309e-05, "loss": 1.2511, "step": 4080 }, { "epoch": 1.26, "learning_rate": 1.9526602188239746e-05, "loss": 1.116, "step": 4081 }, { "epoch": 1.26, "learning_rate": 1.9526299155858467e-05, "loss": 1.148, "step": 4082 }, { "epoch": 1.26, "learning_rate": 1.9525996028872255e-05, "loss": 1.15, "step": 4083 }, { "epoch": 1.26, "learning_rate": 1.9525692807284124e-05, "loss": 1.1381, "step": 4084 }, { "epoch": 1.26, "learning_rate": 1.952538949109708e-05, "loss": 1.0428, "step": 4085 }, { "epoch": 1.26, "learning_rate": 1.952508608031414e-05, "loss": 1.3571, "step": 4086 }, { "epoch": 1.26, "learning_rate": 1.952478257493832e-05, "loss": 1.477, "step": 4087 }, { "epoch": 1.26, "learning_rate": 1.952447897497263e-05, "loss": 1.1991, "step": 4088 }, { "epoch": 1.26, "learning_rate": 1.9524175280420086e-05, "loss": 1.2207, "step": 4089 }, { "epoch": 1.26, "learning_rate": 1.9523871491283706e-05, "loss": 1.8535, "step": 4090 }, { "epoch": 1.26, "learning_rate": 1.95235676075665e-05, "loss": 0.886, "step": 4091 }, { "epoch": 1.26, "learning_rate": 1.9523263629271492e-05, "loss": 1.1852, "step": 4092 }, { "epoch": 1.26, "learning_rate": 1.9522959556401703e-05, "loss": 1.0787, "step": 4093 }, { "epoch": 1.26, "learning_rate": 1.9522655388960144e-05, "loss": 1.27, "step": 4094 }, { "epoch": 1.26, "learning_rate": 1.952235112694984e-05, "loss": 1.2641, "step": 4095 }, { "epoch": 1.26, "learning_rate": 1.952204677037382e-05, "loss": 1.2831, "step": 4096 }, { "epoch": 1.26, "learning_rate": 1.9521742319235095e-05, "loss": 1.3654, "step": 4097 }, { "epoch": 1.26, "learning_rate": 1.952143777353669e-05, "loss": 0.9328, "step": 4098 }, { "epoch": 1.26, "learning_rate": 1.952113313328164e-05, "loss": 1.2987, "step": 4099 }, { "epoch": 1.26, "learning_rate": 1.9520828398472956e-05, "loss": 1.0626, "step": 4100 }, { "epoch": 1.26, "learning_rate": 1.952052356911368e-05, "loss": 1.0229, "step": 4101 }, { "epoch": 1.26, "learning_rate": 1.952021864520682e-05, "loss": 1.0622, "step": 4102 }, { "epoch": 1.26, "learning_rate": 1.9519913626755424e-05, "loss": 1.0263, "step": 4103 }, { "epoch": 1.26, "learning_rate": 1.951960851376251e-05, "loss": 0.9181, "step": 4104 }, { "epoch": 1.26, "learning_rate": 1.9519303306231104e-05, "loss": 1.2441, "step": 4105 }, { "epoch": 1.26, "learning_rate": 1.951899800416425e-05, "loss": 1.2092, "step": 4106 }, { "epoch": 1.26, "learning_rate": 1.9518692607564968e-05, "loss": 1.0078, "step": 4107 }, { "epoch": 1.26, "learning_rate": 1.9518387116436297e-05, "loss": 1.1928, "step": 4108 }, { "epoch": 1.26, "learning_rate": 1.951808153078127e-05, "loss": 1.1645, "step": 4109 }, { "epoch": 1.26, "learning_rate": 1.9517775850602923e-05, "loss": 1.2817, "step": 4110 }, { "epoch": 1.26, "learning_rate": 1.9517470075904285e-05, "loss": 1.5008, "step": 4111 }, { "epoch": 1.27, "learning_rate": 1.95171642066884e-05, "loss": 1.1963, "step": 4112 }, { "epoch": 1.27, "learning_rate": 1.9516858242958307e-05, "loss": 1.4265, "step": 4113 }, { "epoch": 1.27, "learning_rate": 1.9516552184717036e-05, "loss": 1.1247, "step": 4114 }, { "epoch": 1.27, "learning_rate": 1.9516246031967634e-05, "loss": 1.3566, "step": 4115 }, { "epoch": 1.27, "learning_rate": 1.951593978471314e-05, "loss": 1.373, "step": 4116 }, { "epoch": 1.27, "learning_rate": 1.9515633442956588e-05, "loss": 0.8791, "step": 4117 }, { "epoch": 1.27, "learning_rate": 1.9515327006701027e-05, "loss": 1.2939, "step": 4118 }, { "epoch": 1.27, "learning_rate": 1.9515020475949504e-05, "loss": 1.264, "step": 4119 }, { "epoch": 1.27, "learning_rate": 1.9514713850705056e-05, "loss": 1.1395, "step": 4120 }, { "epoch": 1.27, "learning_rate": 1.9514407130970728e-05, "loss": 1.2676, "step": 4121 }, { "epoch": 1.27, "learning_rate": 1.951410031674957e-05, "loss": 1.29, "step": 4122 }, { "epoch": 1.27, "learning_rate": 1.9513793408044628e-05, "loss": 1.2387, "step": 4123 }, { "epoch": 1.27, "learning_rate": 1.951348640485895e-05, "loss": 1.3318, "step": 4124 }, { "epoch": 1.27, "learning_rate": 1.9513179307195584e-05, "loss": 1.0685, "step": 4125 }, { "epoch": 1.27, "learning_rate": 1.951287211505758e-05, "loss": 1.0806, "step": 4126 }, { "epoch": 1.27, "learning_rate": 1.951256482844799e-05, "loss": 1.0082, "step": 4127 }, { "epoch": 1.27, "learning_rate": 1.951225744736986e-05, "loss": 1.1069, "step": 4128 }, { "epoch": 1.27, "learning_rate": 1.9511949971826246e-05, "loss": 1.3777, "step": 4129 }, { "epoch": 1.27, "learning_rate": 1.9511642401820205e-05, "loss": 1.2561, "step": 4130 }, { "epoch": 1.27, "learning_rate": 1.951133473735479e-05, "loss": 1.2438, "step": 4131 }, { "epoch": 1.27, "learning_rate": 1.9511026978433058e-05, "loss": 1.3006, "step": 4132 }, { "epoch": 1.27, "learning_rate": 1.9510719125058057e-05, "loss": 1.1278, "step": 4133 }, { "epoch": 1.27, "learning_rate": 1.951041117723285e-05, "loss": 0.8977, "step": 4134 }, { "epoch": 1.27, "learning_rate": 1.9510103134960497e-05, "loss": 1.2363, "step": 4135 }, { "epoch": 1.27, "learning_rate": 1.9509794998244054e-05, "loss": 1.3791, "step": 4136 }, { "epoch": 1.27, "learning_rate": 1.950948676708658e-05, "loss": 1.3518, "step": 4137 }, { "epoch": 1.27, "learning_rate": 1.9509178441491142e-05, "loss": 1.0874, "step": 4138 }, { "epoch": 1.27, "learning_rate": 1.9508870021460796e-05, "loss": 1.3147, "step": 4139 }, { "epoch": 1.27, "learning_rate": 1.9508561506998613e-05, "loss": 0.9539, "step": 4140 }, { "epoch": 1.27, "learning_rate": 1.950825289810764e-05, "loss": 0.8857, "step": 4141 }, { "epoch": 1.27, "learning_rate": 1.950794419479096e-05, "loss": 1.2756, "step": 4142 }, { "epoch": 1.27, "learning_rate": 1.9507635397051632e-05, "loss": 1.1628, "step": 4143 }, { "epoch": 1.28, "learning_rate": 1.9507326504892718e-05, "loss": 1.1219, "step": 4144 }, { "epoch": 1.28, "learning_rate": 1.950701751831729e-05, "loss": 1.0478, "step": 4145 }, { "epoch": 1.28, "learning_rate": 1.9506708437328423e-05, "loss": 1.132, "step": 4146 }, { "epoch": 1.28, "learning_rate": 1.950639926192917e-05, "loss": 1.1116, "step": 4147 }, { "epoch": 1.28, "learning_rate": 1.9506089992122613e-05, "loss": 1.4622, "step": 4148 }, { "epoch": 1.28, "learning_rate": 1.9505780627911827e-05, "loss": 0.8333, "step": 4149 }, { "epoch": 1.28, "learning_rate": 1.9505471169299876e-05, "loss": 1.0089, "step": 4150 }, { "epoch": 1.28, "learning_rate": 1.9505161616289833e-05, "loss": 0.9955, "step": 4151 }, { "epoch": 1.28, "learning_rate": 1.9504851968884776e-05, "loss": 1.0599, "step": 4152 }, { "epoch": 1.28, "learning_rate": 1.950454222708778e-05, "loss": 1.2332, "step": 4153 }, { "epoch": 1.28, "learning_rate": 1.950423239090192e-05, "loss": 1.3532, "step": 4154 }, { "epoch": 1.28, "learning_rate": 1.9503922460330272e-05, "loss": 1.1229, "step": 4155 }, { "epoch": 1.28, "learning_rate": 1.9503612435375914e-05, "loss": 1.148, "step": 4156 }, { "epoch": 1.28, "learning_rate": 1.950330231604193e-05, "loss": 1.2078, "step": 4157 }, { "epoch": 1.28, "learning_rate": 1.9502992102331393e-05, "loss": 1.4468, "step": 4158 }, { "epoch": 1.28, "learning_rate": 1.9502681794247386e-05, "loss": 1.1311, "step": 4159 }, { "epoch": 1.28, "learning_rate": 1.950237139179299e-05, "loss": 1.1802, "step": 4160 }, { "epoch": 1.28, "learning_rate": 1.950206089497129e-05, "loss": 1.0941, "step": 4161 }, { "epoch": 1.28, "learning_rate": 1.9501750303785368e-05, "loss": 1.4272, "step": 4162 }, { "epoch": 1.28, "learning_rate": 1.9501439618238306e-05, "loss": 1.0163, "step": 4163 }, { "epoch": 1.28, "learning_rate": 1.9501128838333196e-05, "loss": 1.1702, "step": 4164 }, { "epoch": 1.28, "learning_rate": 1.950081796407312e-05, "loss": 1.261, "step": 4165 }, { "epoch": 1.28, "learning_rate": 1.950050699546116e-05, "loss": 1.155, "step": 4166 }, { "epoch": 1.28, "learning_rate": 1.9500195932500415e-05, "loss": 1.1242, "step": 4167 }, { "epoch": 1.28, "learning_rate": 1.9499884775193964e-05, "loss": 0.7138, "step": 4168 }, { "epoch": 1.28, "learning_rate": 1.9499573523544907e-05, "loss": 0.92, "step": 4169 }, { "epoch": 1.28, "learning_rate": 1.9499262177556325e-05, "loss": 1.0517, "step": 4170 }, { "epoch": 1.28, "learning_rate": 1.9498950737231318e-05, "loss": 1.2575, "step": 4171 }, { "epoch": 1.28, "learning_rate": 1.9498639202572974e-05, "loss": 1.2189, "step": 4172 }, { "epoch": 1.28, "learning_rate": 1.949832757358439e-05, "loss": 1.1685, "step": 4173 }, { "epoch": 1.28, "learning_rate": 1.9498015850268657e-05, "loss": 1.0702, "step": 4174 }, { "epoch": 1.28, "learning_rate": 1.9497704032628874e-05, "loss": 1.0505, "step": 4175 }, { "epoch": 1.28, "learning_rate": 1.9497392120668136e-05, "loss": 1.0796, "step": 4176 }, { "epoch": 1.29, "learning_rate": 1.9497080114389545e-05, "loss": 1.4026, "step": 4177 }, { "epoch": 1.29, "learning_rate": 1.9496768013796193e-05, "loss": 1.1533, "step": 4178 }, { "epoch": 1.29, "learning_rate": 1.949645581889118e-05, "loss": 1.0526, "step": 4179 }, { "epoch": 1.29, "learning_rate": 1.9496143529677612e-05, "loss": 1.4011, "step": 4180 }, { "epoch": 1.29, "learning_rate": 1.9495831146158583e-05, "loss": 1.1013, "step": 4181 }, { "epoch": 1.29, "learning_rate": 1.9495518668337204e-05, "loss": 1.1431, "step": 4182 }, { "epoch": 1.29, "learning_rate": 1.949520609621657e-05, "loss": 1.1834, "step": 4183 }, { "epoch": 1.29, "learning_rate": 1.949489342979979e-05, "loss": 1.104, "step": 4184 }, { "epoch": 1.29, "learning_rate": 1.9494580669089966e-05, "loss": 1.2725, "step": 4185 }, { "epoch": 1.29, "learning_rate": 1.9494267814090207e-05, "loss": 1.227, "step": 4186 }, { "epoch": 1.29, "learning_rate": 1.9493954864803618e-05, "loss": 1.0972, "step": 4187 }, { "epoch": 1.29, "learning_rate": 1.9493641821233308e-05, "loss": 1.1119, "step": 4188 }, { "epoch": 1.29, "learning_rate": 1.9493328683382386e-05, "loss": 1.157, "step": 4189 }, { "epoch": 1.29, "learning_rate": 1.9493015451253956e-05, "loss": 1.2985, "step": 4190 }, { "epoch": 1.29, "learning_rate": 1.9492702124851135e-05, "loss": 1.3013, "step": 4191 }, { "epoch": 1.29, "learning_rate": 1.9492388704177036e-05, "loss": 1.338, "step": 4192 }, { "epoch": 1.29, "learning_rate": 1.9492075189234768e-05, "loss": 0.8632, "step": 4193 }, { "epoch": 1.29, "learning_rate": 1.9491761580027442e-05, "loss": 1.2315, "step": 4194 }, { "epoch": 1.29, "learning_rate": 1.949144787655818e-05, "loss": 1.7824, "step": 4195 }, { "epoch": 1.29, "learning_rate": 1.949113407883009e-05, "loss": 1.0877, "step": 4196 }, { "epoch": 1.29, "learning_rate": 1.9490820186846294e-05, "loss": 1.3306, "step": 4197 }, { "epoch": 1.29, "learning_rate": 1.9490506200609903e-05, "loss": 1.1937, "step": 4198 }, { "epoch": 1.29, "learning_rate": 1.9490192120124044e-05, "loss": 1.1985, "step": 4199 }, { "epoch": 1.29, "learning_rate": 1.9489877945391828e-05, "loss": 1.2679, "step": 4200 }, { "epoch": 1.29, "learning_rate": 1.9489563676416377e-05, "loss": 0.8765, "step": 4201 }, { "epoch": 1.29, "learning_rate": 1.9489249313200814e-05, "loss": 1.2242, "step": 4202 }, { "epoch": 1.29, "learning_rate": 1.9488934855748258e-05, "loss": 1.2771, "step": 4203 }, { "epoch": 1.29, "learning_rate": 1.9488620304061836e-05, "loss": 1.0309, "step": 4204 }, { "epoch": 1.29, "learning_rate": 1.9488305658144666e-05, "loss": 1.0711, "step": 4205 }, { "epoch": 1.29, "learning_rate": 1.9487990917999878e-05, "loss": 1.0029, "step": 4206 }, { "epoch": 1.29, "learning_rate": 1.94876760836306e-05, "loss": 1.2347, "step": 4207 }, { "epoch": 1.29, "learning_rate": 1.9487361155039948e-05, "loss": 1.3884, "step": 4208 }, { "epoch": 1.3, "learning_rate": 1.948704613223106e-05, "loss": 1.2636, "step": 4209 }, { "epoch": 1.3, "learning_rate": 1.9486731015207056e-05, "loss": 1.1674, "step": 4210 }, { "epoch": 1.3, "learning_rate": 1.9486415803971074e-05, "loss": 1.7372, "step": 4211 }, { "epoch": 1.3, "learning_rate": 1.9486100498526238e-05, "loss": 0.9865, "step": 4212 }, { "epoch": 1.3, "learning_rate": 1.9485785098875682e-05, "loss": 1.1844, "step": 4213 }, { "epoch": 1.3, "learning_rate": 1.948546960502254e-05, "loss": 1.0885, "step": 4214 }, { "epoch": 1.3, "learning_rate": 1.948515401696994e-05, "loss": 1.0414, "step": 4215 }, { "epoch": 1.3, "learning_rate": 1.9484838334721018e-05, "loss": 1.3285, "step": 4216 }, { "epoch": 1.3, "learning_rate": 1.948452255827891e-05, "loss": 1.2597, "step": 4217 }, { "epoch": 1.3, "learning_rate": 1.9484206687646753e-05, "loss": 1.6347, "step": 4218 }, { "epoch": 1.3, "learning_rate": 1.9483890722827686e-05, "loss": 1.0488, "step": 4219 }, { "epoch": 1.3, "learning_rate": 1.948357466382484e-05, "loss": 1.1532, "step": 4220 }, { "epoch": 1.3, "learning_rate": 1.9483258510641356e-05, "loss": 1.1728, "step": 4221 }, { "epoch": 1.3, "learning_rate": 1.9482942263280375e-05, "loss": 1.0737, "step": 4222 }, { "epoch": 1.3, "learning_rate": 1.9482625921745038e-05, "loss": 1.3895, "step": 4223 }, { "epoch": 1.3, "learning_rate": 1.9482309486038492e-05, "loss": 1.0094, "step": 4224 }, { "epoch": 1.3, "learning_rate": 1.9481992956163868e-05, "loss": 1.0802, "step": 4225 }, { "epoch": 1.3, "learning_rate": 1.9481676332124317e-05, "loss": 1.0374, "step": 4226 }, { "epoch": 1.3, "learning_rate": 1.9481359613922976e-05, "loss": 1.2141, "step": 4227 }, { "epoch": 1.3, "learning_rate": 1.9481042801563003e-05, "loss": 0.8592, "step": 4228 }, { "epoch": 1.3, "learning_rate": 1.9480725895047535e-05, "loss": 1.1408, "step": 4229 }, { "epoch": 1.3, "learning_rate": 1.948040889437972e-05, "loss": 0.9833, "step": 4230 }, { "epoch": 1.3, "learning_rate": 1.9480091799562706e-05, "loss": 1.0143, "step": 4231 }, { "epoch": 1.3, "learning_rate": 1.9479774610599643e-05, "loss": 1.1156, "step": 4232 }, { "epoch": 1.3, "learning_rate": 1.9479457327493686e-05, "loss": 1.1774, "step": 4233 }, { "epoch": 1.3, "learning_rate": 1.947913995024798e-05, "loss": 1.1653, "step": 4234 }, { "epoch": 1.3, "learning_rate": 1.9478822478865673e-05, "loss": 1.1345, "step": 4235 }, { "epoch": 1.3, "learning_rate": 1.9478504913349926e-05, "loss": 1.3812, "step": 4236 }, { "epoch": 1.3, "learning_rate": 1.947818725370389e-05, "loss": 1.1648, "step": 4237 }, { "epoch": 1.3, "learning_rate": 1.947786949993072e-05, "loss": 1.3013, "step": 4238 }, { "epoch": 1.3, "learning_rate": 1.947755165203357e-05, "loss": 1.2757, "step": 4239 }, { "epoch": 1.3, "learning_rate": 1.9477233710015597e-05, "loss": 1.3167, "step": 4240 }, { "epoch": 1.3, "learning_rate": 1.9476915673879962e-05, "loss": 1.4473, "step": 4241 }, { "epoch": 1.31, "learning_rate": 1.9476597543629812e-05, "loss": 1.093, "step": 4242 }, { "epoch": 1.31, "learning_rate": 1.947627931926832e-05, "loss": 1.628, "step": 4243 }, { "epoch": 1.31, "learning_rate": 1.9475961000798645e-05, "loss": 1.2254, "step": 4244 }, { "epoch": 1.31, "learning_rate": 1.9475642588223936e-05, "loss": 1.4906, "step": 4245 }, { "epoch": 1.31, "learning_rate": 1.9475324081547367e-05, "loss": 1.0991, "step": 4246 }, { "epoch": 1.31, "learning_rate": 1.94750054807721e-05, "loss": 1.0052, "step": 4247 }, { "epoch": 1.31, "learning_rate": 1.947468678590129e-05, "loss": 1.0112, "step": 4248 }, { "epoch": 1.31, "learning_rate": 1.947436799693811e-05, "loss": 1.3431, "step": 4249 }, { "epoch": 1.31, "learning_rate": 1.9474049113885725e-05, "loss": 1.2341, "step": 4250 }, { "epoch": 1.31, "learning_rate": 1.94737301367473e-05, "loss": 1.5271, "step": 4251 }, { "epoch": 1.31, "learning_rate": 1.9473411065526004e-05, "loss": 1.3602, "step": 4252 }, { "epoch": 1.31, "learning_rate": 1.9473091900225007e-05, "loss": 0.9379, "step": 4253 }, { "epoch": 1.31, "learning_rate": 1.9472772640847473e-05, "loss": 1.1761, "step": 4254 }, { "epoch": 1.31, "learning_rate": 1.947245328739658e-05, "loss": 1.2021, "step": 4255 }, { "epoch": 1.31, "learning_rate": 1.9472133839875495e-05, "loss": 1.032, "step": 4256 }, { "epoch": 1.31, "learning_rate": 1.947181429828739e-05, "loss": 1.1601, "step": 4257 }, { "epoch": 1.31, "learning_rate": 1.947149466263544e-05, "loss": 1.4748, "step": 4258 }, { "epoch": 1.31, "learning_rate": 1.947117493292282e-05, "loss": 1.0908, "step": 4259 }, { "epoch": 1.31, "learning_rate": 1.94708551091527e-05, "loss": 1.528, "step": 4260 }, { "epoch": 1.31, "learning_rate": 1.9470535191328265e-05, "loss": 0.9821, "step": 4261 }, { "epoch": 1.31, "learning_rate": 1.9470215179452685e-05, "loss": 1.2664, "step": 4262 }, { "epoch": 1.31, "learning_rate": 1.946989507352914e-05, "loss": 1.0716, "step": 4263 }, { "epoch": 1.31, "learning_rate": 1.946957487356081e-05, "loss": 1.0747, "step": 4264 }, { "epoch": 1.31, "learning_rate": 1.946925457955087e-05, "loss": 1.0833, "step": 4265 }, { "epoch": 1.31, "learning_rate": 1.946893419150251e-05, "loss": 1.5035, "step": 4266 }, { "epoch": 1.31, "learning_rate": 1.9468613709418907e-05, "loss": 1.0007, "step": 4267 }, { "epoch": 1.31, "learning_rate": 1.9468293133303242e-05, "loss": 1.1617, "step": 4268 }, { "epoch": 1.31, "learning_rate": 1.9467972463158697e-05, "loss": 1.2523, "step": 4269 }, { "epoch": 1.31, "learning_rate": 1.9467651698988464e-05, "loss": 1.0787, "step": 4270 }, { "epoch": 1.31, "learning_rate": 1.9467330840795723e-05, "loss": 1.0571, "step": 4271 }, { "epoch": 1.31, "learning_rate": 1.946700988858366e-05, "loss": 1.195, "step": 4272 }, { "epoch": 1.31, "learning_rate": 1.9466688842355464e-05, "loss": 0.9748, "step": 4273 }, { "epoch": 1.32, "learning_rate": 1.9466367702114324e-05, "loss": 1.2027, "step": 4274 }, { "epoch": 1.32, "learning_rate": 1.9466046467863428e-05, "loss": 1.1957, "step": 4275 }, { "epoch": 1.32, "learning_rate": 1.9465725139605965e-05, "loss": 0.9892, "step": 4276 }, { "epoch": 1.32, "learning_rate": 1.946540371734513e-05, "loss": 1.3507, "step": 4277 }, { "epoch": 1.32, "learning_rate": 1.946508220108411e-05, "loss": 0.9792, "step": 4278 }, { "epoch": 1.32, "learning_rate": 1.94647605908261e-05, "loss": 1.0006, "step": 4279 }, { "epoch": 1.32, "learning_rate": 1.9464438886574295e-05, "loss": 1.2218, "step": 4280 }, { "epoch": 1.32, "learning_rate": 1.946411708833189e-05, "loss": 0.9696, "step": 4281 }, { "epoch": 1.32, "learning_rate": 1.946379519610208e-05, "loss": 1.3757, "step": 4282 }, { "epoch": 1.32, "learning_rate": 1.9463473209888063e-05, "loss": 1.0545, "step": 4283 }, { "epoch": 1.32, "learning_rate": 1.9463151129693033e-05, "loss": 1.2694, "step": 4284 }, { "epoch": 1.32, "learning_rate": 1.946282895552019e-05, "loss": 1.2342, "step": 4285 }, { "epoch": 1.32, "learning_rate": 1.9462506687372738e-05, "loss": 1.5235, "step": 4286 }, { "epoch": 1.32, "learning_rate": 1.9462184325253866e-05, "loss": 1.0086, "step": 4287 }, { "epoch": 1.32, "learning_rate": 1.9461861869166793e-05, "loss": 1.2125, "step": 4288 }, { "epoch": 1.32, "learning_rate": 1.9461539319114706e-05, "loss": 1.2805, "step": 4289 }, { "epoch": 1.32, "learning_rate": 1.9461216675100814e-05, "loss": 0.8947, "step": 4290 }, { "epoch": 1.32, "learning_rate": 1.946089393712832e-05, "loss": 1.0315, "step": 4291 }, { "epoch": 1.32, "learning_rate": 1.9460571105200434e-05, "loss": 0.8642, "step": 4292 }, { "epoch": 1.32, "learning_rate": 1.9460248179320353e-05, "loss": 1.2734, "step": 4293 }, { "epoch": 1.32, "learning_rate": 1.9459925159491288e-05, "loss": 1.1414, "step": 4294 }, { "epoch": 1.32, "learning_rate": 1.945960204571645e-05, "loss": 1.0666, "step": 4295 }, { "epoch": 1.32, "learning_rate": 1.9459278837999048e-05, "loss": 1.3567, "step": 4296 }, { "epoch": 1.32, "learning_rate": 1.9458955536342287e-05, "loss": 1.4132, "step": 4297 }, { "epoch": 1.32, "learning_rate": 1.945863214074938e-05, "loss": 1.3273, "step": 4298 }, { "epoch": 1.32, "learning_rate": 1.9458308651223536e-05, "loss": 1.1698, "step": 4299 }, { "epoch": 1.32, "learning_rate": 1.9457985067767973e-05, "loss": 1.1836, "step": 4300 }, { "epoch": 1.32, "learning_rate": 1.94576613903859e-05, "loss": 1.2732, "step": 4301 }, { "epoch": 1.32, "learning_rate": 1.9457337619080534e-05, "loss": 1.3149, "step": 4302 }, { "epoch": 1.32, "learning_rate": 1.9457013753855087e-05, "loss": 1.0902, "step": 4303 }, { "epoch": 1.32, "learning_rate": 1.9456689794712782e-05, "loss": 1.1002, "step": 4304 }, { "epoch": 1.32, "learning_rate": 1.9456365741656827e-05, "loss": 1.4079, "step": 4305 }, { "epoch": 1.32, "learning_rate": 1.945604159469045e-05, "loss": 1.3952, "step": 4306 }, { "epoch": 1.33, "learning_rate": 1.945571735381686e-05, "loss": 1.0365, "step": 4307 }, { "epoch": 1.33, "learning_rate": 1.9455393019039286e-05, "loss": 1.3328, "step": 4308 }, { "epoch": 1.33, "learning_rate": 1.9455068590360943e-05, "loss": 1.098, "step": 4309 }, { "epoch": 1.33, "learning_rate": 1.9454744067785057e-05, "loss": 1.2083, "step": 4310 }, { "epoch": 1.33, "learning_rate": 1.9454419451314845e-05, "loss": 1.0375, "step": 4311 }, { "epoch": 1.33, "learning_rate": 1.9454094740953534e-05, "loss": 1.1865, "step": 4312 }, { "epoch": 1.33, "learning_rate": 1.9453769936704353e-05, "loss": 1.3586, "step": 4313 }, { "epoch": 1.33, "learning_rate": 1.945344503857052e-05, "loss": 0.8843, "step": 4314 }, { "epoch": 1.33, "learning_rate": 1.945312004655527e-05, "loss": 1.0659, "step": 4315 }, { "epoch": 1.33, "learning_rate": 1.9452794960661822e-05, "loss": 1.1713, "step": 4316 }, { "epoch": 1.33, "learning_rate": 1.945246978089341e-05, "loss": 1.2556, "step": 4317 }, { "epoch": 1.33, "learning_rate": 1.945214450725326e-05, "loss": 1.1992, "step": 4318 }, { "epoch": 1.33, "learning_rate": 1.9451819139744605e-05, "loss": 1.202, "step": 4319 }, { "epoch": 1.33, "learning_rate": 1.9451493678370672e-05, "loss": 1.2605, "step": 4320 }, { "epoch": 1.33, "learning_rate": 1.9451168123134698e-05, "loss": 1.2679, "step": 4321 }, { "epoch": 1.33, "learning_rate": 1.9450842474039914e-05, "loss": 1.2378, "step": 4322 }, { "epoch": 1.33, "learning_rate": 1.9450516731089556e-05, "loss": 1.264, "step": 4323 }, { "epoch": 1.33, "learning_rate": 1.9450190894286854e-05, "loss": 1.3123, "step": 4324 }, { "epoch": 1.33, "learning_rate": 1.944986496363505e-05, "loss": 1.4018, "step": 4325 }, { "epoch": 1.33, "learning_rate": 1.9449538939137373e-05, "loss": 1.0807, "step": 4326 }, { "epoch": 1.33, "learning_rate": 1.9449212820797067e-05, "loss": 0.8958, "step": 4327 }, { "epoch": 1.33, "learning_rate": 1.9448886608617372e-05, "loss": 1.1486, "step": 4328 }, { "epoch": 1.33, "learning_rate": 1.9448560302601524e-05, "loss": 1.2097, "step": 4329 }, { "epoch": 1.33, "learning_rate": 1.944823390275276e-05, "loss": 0.8203, "step": 4330 }, { "epoch": 1.33, "learning_rate": 1.9447907409074328e-05, "loss": 1.2178, "step": 4331 }, { "epoch": 1.33, "learning_rate": 1.944758082156947e-05, "loss": 1.0142, "step": 4332 }, { "epoch": 1.33, "learning_rate": 1.9447254140241426e-05, "loss": 1.1744, "step": 4333 }, { "epoch": 1.33, "learning_rate": 1.9446927365093442e-05, "loss": 1.5977, "step": 4334 }, { "epoch": 1.33, "learning_rate": 1.944660049612876e-05, "loss": 1.3109, "step": 4335 }, { "epoch": 1.33, "learning_rate": 1.9446273533350633e-05, "loss": 1.2913, "step": 4336 }, { "epoch": 1.33, "learning_rate": 1.9445946476762302e-05, "loss": 1.2228, "step": 4337 }, { "epoch": 1.33, "learning_rate": 1.9445619326367017e-05, "loss": 1.0704, "step": 4338 }, { "epoch": 1.34, "learning_rate": 1.9445292082168023e-05, "loss": 1.2099, "step": 4339 }, { "epoch": 1.34, "learning_rate": 1.944496474416858e-05, "loss": 1.0855, "step": 4340 }, { "epoch": 1.34, "learning_rate": 1.9444637312371925e-05, "loss": 1.4166, "step": 4341 }, { "epoch": 1.34, "learning_rate": 1.9444309786781323e-05, "loss": 1.1152, "step": 4342 }, { "epoch": 1.34, "learning_rate": 1.944398216740002e-05, "loss": 1.4148, "step": 4343 }, { "epoch": 1.34, "learning_rate": 1.944365445423127e-05, "loss": 1.0834, "step": 4344 }, { "epoch": 1.34, "learning_rate": 1.9443326647278328e-05, "loss": 1.1208, "step": 4345 }, { "epoch": 1.34, "learning_rate": 1.9442998746544446e-05, "loss": 0.8939, "step": 4346 }, { "epoch": 1.34, "learning_rate": 1.9442670752032887e-05, "loss": 1.8404, "step": 4347 }, { "epoch": 1.34, "learning_rate": 1.9442342663746903e-05, "loss": 1.1592, "step": 4348 }, { "epoch": 1.34, "learning_rate": 1.9442014481689754e-05, "loss": 1.5374, "step": 4349 }, { "epoch": 1.34, "learning_rate": 1.94416862058647e-05, "loss": 0.9297, "step": 4350 }, { "epoch": 1.34, "learning_rate": 1.9441357836274998e-05, "loss": 1.3649, "step": 4351 }, { "epoch": 1.34, "learning_rate": 1.9441029372923916e-05, "loss": 1.3864, "step": 4352 }, { "epoch": 1.34, "learning_rate": 1.9440700815814706e-05, "loss": 0.9649, "step": 4353 }, { "epoch": 1.34, "learning_rate": 1.944037216495064e-05, "loss": 1.2811, "step": 4354 }, { "epoch": 1.34, "learning_rate": 1.9440043420334975e-05, "loss": 1.021, "step": 4355 }, { "epoch": 1.34, "learning_rate": 1.9439714581970983e-05, "loss": 1.1926, "step": 4356 }, { "epoch": 1.34, "learning_rate": 1.943938564986192e-05, "loss": 1.1734, "step": 4357 }, { "epoch": 1.34, "learning_rate": 1.943905662401106e-05, "loss": 1.1904, "step": 4358 }, { "epoch": 1.34, "learning_rate": 1.9438727504421667e-05, "loss": 1.1926, "step": 4359 }, { "epoch": 1.34, "learning_rate": 1.9438398291097016e-05, "loss": 1.1209, "step": 4360 }, { "epoch": 1.34, "learning_rate": 1.9438068984040366e-05, "loss": 1.0193, "step": 4361 }, { "epoch": 1.34, "learning_rate": 1.9437739583254994e-05, "loss": 2.1199, "step": 4362 }, { "epoch": 1.34, "learning_rate": 1.9437410088744172e-05, "loss": 1.1858, "step": 4363 }, { "epoch": 1.34, "learning_rate": 1.9437080500511168e-05, "loss": 1.3717, "step": 4364 }, { "epoch": 1.34, "learning_rate": 1.9436750818559257e-05, "loss": 1.1696, "step": 4365 }, { "epoch": 1.34, "learning_rate": 1.9436421042891712e-05, "loss": 0.8787, "step": 4366 }, { "epoch": 1.34, "learning_rate": 1.9436091173511813e-05, "loss": 1.4672, "step": 4367 }, { "epoch": 1.34, "learning_rate": 1.943576121042283e-05, "loss": 1.1631, "step": 4368 }, { "epoch": 1.34, "learning_rate": 1.943543115362804e-05, "loss": 1.2517, "step": 4369 }, { "epoch": 1.34, "learning_rate": 1.9435101003130724e-05, "loss": 1.2994, "step": 4370 }, { "epoch": 1.34, "learning_rate": 1.9434770758934162e-05, "loss": 1.3321, "step": 4371 }, { "epoch": 1.35, "learning_rate": 1.943444042104163e-05, "loss": 1.4076, "step": 4372 }, { "epoch": 1.35, "learning_rate": 1.9434109989456406e-05, "loss": 0.8267, "step": 4373 }, { "epoch": 1.35, "learning_rate": 1.943377946418178e-05, "loss": 0.9139, "step": 4374 }, { "epoch": 1.35, "learning_rate": 1.9433448845221026e-05, "loss": 1.2662, "step": 4375 }, { "epoch": 1.35, "learning_rate": 1.9433118132577432e-05, "loss": 1.129, "step": 4376 }, { "epoch": 1.35, "learning_rate": 1.9432787326254285e-05, "loss": 0.8271, "step": 4377 }, { "epoch": 1.35, "learning_rate": 1.943245642625486e-05, "loss": 1.0219, "step": 4378 }, { "epoch": 1.35, "learning_rate": 1.9432125432582452e-05, "loss": 1.5189, "step": 4379 }, { "epoch": 1.35, "learning_rate": 1.9431794345240346e-05, "loss": 1.5472, "step": 4380 }, { "epoch": 1.35, "learning_rate": 1.943146316423183e-05, "loss": 0.9931, "step": 4381 }, { "epoch": 1.35, "learning_rate": 1.9431131889560187e-05, "loss": 1.2682, "step": 4382 }, { "epoch": 1.35, "learning_rate": 1.943080052122872e-05, "loss": 1.0882, "step": 4383 }, { "epoch": 1.35, "learning_rate": 1.943046905924071e-05, "loss": 0.9951, "step": 4384 }, { "epoch": 1.35, "learning_rate": 1.9430137503599448e-05, "loss": 1.3072, "step": 4385 }, { "epoch": 1.35, "learning_rate": 1.942980585430823e-05, "loss": 1.2437, "step": 4386 }, { "epoch": 1.35, "learning_rate": 1.942947411137035e-05, "loss": 1.0984, "step": 4387 }, { "epoch": 1.35, "learning_rate": 1.9429142274789104e-05, "loss": 1.189, "step": 4388 }, { "epoch": 1.35, "learning_rate": 1.9428810344567782e-05, "loss": 1.489, "step": 4389 }, { "epoch": 1.35, "learning_rate": 1.9428478320709688e-05, "loss": 1.1079, "step": 4390 }, { "epoch": 1.35, "learning_rate": 1.942814620321811e-05, "loss": 0.9442, "step": 4391 }, { "epoch": 1.35, "learning_rate": 1.9427813992096355e-05, "loss": 1.2571, "step": 4392 }, { "epoch": 1.35, "learning_rate": 1.9427481687347712e-05, "loss": 1.0045, "step": 4393 }, { "epoch": 1.35, "learning_rate": 1.9427149288975495e-05, "loss": 1.0182, "step": 4394 }, { "epoch": 1.35, "learning_rate": 1.942681679698299e-05, "loss": 1.3057, "step": 4395 }, { "epoch": 1.35, "learning_rate": 1.942648421137351e-05, "loss": 1.2606, "step": 4396 }, { "epoch": 1.35, "learning_rate": 1.942615153215035e-05, "loss": 1.131, "step": 4397 }, { "epoch": 1.35, "learning_rate": 1.9425818759316823e-05, "loss": 1.2812, "step": 4398 }, { "epoch": 1.35, "learning_rate": 1.942548589287623e-05, "loss": 0.9636, "step": 4399 }, { "epoch": 1.35, "learning_rate": 1.942515293283187e-05, "loss": 1.0905, "step": 4400 }, { "epoch": 1.35, "learning_rate": 1.9424819879187056e-05, "loss": 1.1988, "step": 4401 }, { "epoch": 1.35, "learning_rate": 1.9424486731945095e-05, "loss": 1.2042, "step": 4402 }, { "epoch": 1.35, "learning_rate": 1.9424153491109297e-05, "loss": 1.086, "step": 4403 }, { "epoch": 1.36, "learning_rate": 1.9423820156682965e-05, "loss": 1.0775, "step": 4404 }, { "epoch": 1.36, "learning_rate": 1.942348672866942e-05, "loss": 1.349, "step": 4405 }, { "epoch": 1.36, "learning_rate": 1.9423153207071958e-05, "loss": 1.0975, "step": 4406 }, { "epoch": 1.36, "learning_rate": 1.9422819591893906e-05, "loss": 1.1756, "step": 4407 }, { "epoch": 1.36, "learning_rate": 1.9422485883138572e-05, "loss": 1.0705, "step": 4408 }, { "epoch": 1.36, "learning_rate": 1.9422152080809268e-05, "loss": 0.844, "step": 4409 }, { "epoch": 1.36, "learning_rate": 1.942181818490931e-05, "loss": 1.1879, "step": 4410 }, { "epoch": 1.36, "learning_rate": 1.942148419544201e-05, "loss": 1.163, "step": 4411 }, { "epoch": 1.36, "learning_rate": 1.9421150112410694e-05, "loss": 0.8691, "step": 4412 }, { "epoch": 1.36, "learning_rate": 1.9420815935818673e-05, "loss": 0.8504, "step": 4413 }, { "epoch": 1.36, "learning_rate": 1.9420481665669268e-05, "loss": 1.4712, "step": 4414 }, { "epoch": 1.36, "learning_rate": 1.9420147301965795e-05, "loss": 1.0135, "step": 4415 }, { "epoch": 1.36, "learning_rate": 1.9419812844711582e-05, "loss": 1.3364, "step": 4416 }, { "epoch": 1.36, "learning_rate": 1.9419478293909942e-05, "loss": 1.3178, "step": 4417 }, { "epoch": 1.36, "learning_rate": 1.94191436495642e-05, "loss": 1.3421, "step": 4418 }, { "epoch": 1.36, "learning_rate": 1.9418808911677685e-05, "loss": 1.4225, "step": 4419 }, { "epoch": 1.36, "learning_rate": 1.9418474080253715e-05, "loss": 1.1911, "step": 4420 }, { "epoch": 1.36, "learning_rate": 1.941813915529562e-05, "loss": 1.6597, "step": 4421 }, { "epoch": 1.36, "learning_rate": 1.9417804136806717e-05, "loss": 1.2543, "step": 4422 }, { "epoch": 1.36, "learning_rate": 1.9417469024790344e-05, "loss": 1.1234, "step": 4423 }, { "epoch": 1.36, "learning_rate": 1.9417133819249824e-05, "loss": 1.1714, "step": 4424 }, { "epoch": 1.36, "learning_rate": 1.941679852018848e-05, "loss": 1.443, "step": 4425 }, { "epoch": 1.36, "learning_rate": 1.9416463127609655e-05, "loss": 1.3674, "step": 4426 }, { "epoch": 1.36, "learning_rate": 1.941612764151667e-05, "loss": 1.2558, "step": 4427 }, { "epoch": 1.36, "learning_rate": 1.9415792061912863e-05, "loss": 1.1262, "step": 4428 }, { "epoch": 1.36, "learning_rate": 1.9415456388801557e-05, "loss": 1.2867, "step": 4429 }, { "epoch": 1.36, "learning_rate": 1.9415120622186094e-05, "loss": 0.9914, "step": 4430 }, { "epoch": 1.36, "learning_rate": 1.941478476206981e-05, "loss": 1.042, "step": 4431 }, { "epoch": 1.36, "learning_rate": 1.941444880845603e-05, "loss": 1.238, "step": 4432 }, { "epoch": 1.36, "learning_rate": 1.94141127613481e-05, "loss": 1.0958, "step": 4433 }, { "epoch": 1.36, "learning_rate": 1.9413776620749354e-05, "loss": 1.3056, "step": 4434 }, { "epoch": 1.36, "learning_rate": 1.941344038666313e-05, "loss": 0.9431, "step": 4435 }, { "epoch": 1.36, "learning_rate": 1.9413104059092765e-05, "loss": 1.1889, "step": 4436 }, { "epoch": 1.37, "learning_rate": 1.9412767638041602e-05, "loss": 1.2732, "step": 4437 }, { "epoch": 1.37, "learning_rate": 1.9412431123512985e-05, "loss": 1.0416, "step": 4438 }, { "epoch": 1.37, "learning_rate": 1.941209451551025e-05, "loss": 1.1213, "step": 4439 }, { "epoch": 1.37, "learning_rate": 1.9411757814036743e-05, "loss": 1.0033, "step": 4440 }, { "epoch": 1.37, "learning_rate": 1.94114210190958e-05, "loss": 1.2216, "step": 4441 }, { "epoch": 1.37, "learning_rate": 1.941108413069078e-05, "loss": 1.5731, "step": 4442 }, { "epoch": 1.37, "learning_rate": 1.941074714882502e-05, "loss": 1.229, "step": 4443 }, { "epoch": 1.37, "learning_rate": 1.9410410073501867e-05, "loss": 1.5781, "step": 4444 }, { "epoch": 1.37, "learning_rate": 1.941007290472467e-05, "loss": 1.272, "step": 4445 }, { "epoch": 1.37, "learning_rate": 1.9409735642496775e-05, "loss": 1.2548, "step": 4446 }, { "epoch": 1.37, "learning_rate": 1.940939828682153e-05, "loss": 1.0553, "step": 4447 }, { "epoch": 1.37, "learning_rate": 1.940906083770229e-05, "loss": 1.065, "step": 4448 }, { "epoch": 1.37, "learning_rate": 1.9408723295142405e-05, "loss": 0.8937, "step": 4449 }, { "epoch": 1.37, "learning_rate": 1.9408385659145228e-05, "loss": 0.986, "step": 4450 }, { "epoch": 1.37, "learning_rate": 1.940804792971411e-05, "loss": 1.1133, "step": 4451 }, { "epoch": 1.37, "learning_rate": 1.9407710106852405e-05, "loss": 1.123, "step": 4452 }, { "epoch": 1.37, "learning_rate": 1.9407372190563467e-05, "loss": 1.4107, "step": 4453 }, { "epoch": 1.37, "learning_rate": 1.9407034180850654e-05, "loss": 1.2604, "step": 4454 }, { "epoch": 1.37, "learning_rate": 1.9406696077717324e-05, "loss": 1.2628, "step": 4455 }, { "epoch": 1.37, "learning_rate": 1.9406357881166827e-05, "loss": 0.8691, "step": 4456 }, { "epoch": 1.37, "learning_rate": 1.940601959120253e-05, "loss": 0.8136, "step": 4457 }, { "epoch": 1.37, "learning_rate": 1.9405681207827788e-05, "loss": 1.5383, "step": 4458 }, { "epoch": 1.37, "learning_rate": 1.9405342731045965e-05, "loss": 1.2442, "step": 4459 }, { "epoch": 1.37, "learning_rate": 1.9405004160860422e-05, "loss": 1.3629, "step": 4460 }, { "epoch": 1.37, "learning_rate": 1.940466549727452e-05, "loss": 0.9339, "step": 4461 }, { "epoch": 1.37, "learning_rate": 1.9404326740291618e-05, "loss": 1.1067, "step": 4462 }, { "epoch": 1.37, "learning_rate": 1.9403987889915085e-05, "loss": 1.289, "step": 4463 }, { "epoch": 1.37, "learning_rate": 1.9403648946148287e-05, "loss": 1.167, "step": 4464 }, { "epoch": 1.37, "learning_rate": 1.940330990899459e-05, "loss": 1.3192, "step": 4465 }, { "epoch": 1.37, "learning_rate": 1.9402970778457355e-05, "loss": 0.9804, "step": 4466 }, { "epoch": 1.37, "learning_rate": 1.9402631554539956e-05, "loss": 1.3001, "step": 4467 }, { "epoch": 1.37, "learning_rate": 1.940229223724576e-05, "loss": 1.2557, "step": 4468 }, { "epoch": 1.38, "learning_rate": 1.9401952826578135e-05, "loss": 1.0814, "step": 4469 }, { "epoch": 1.38, "learning_rate": 1.9401613322540456e-05, "loss": 1.2359, "step": 4470 }, { "epoch": 1.38, "learning_rate": 1.9401273725136093e-05, "loss": 1.4168, "step": 4471 }, { "epoch": 1.38, "learning_rate": 1.9400934034368415e-05, "loss": 1.2933, "step": 4472 }, { "epoch": 1.38, "learning_rate": 1.94005942502408e-05, "loss": 1.0279, "step": 4473 }, { "epoch": 1.38, "learning_rate": 1.9400254372756614e-05, "loss": 1.0294, "step": 4474 }, { "epoch": 1.38, "learning_rate": 1.9399914401919243e-05, "loss": 0.9518, "step": 4475 }, { "epoch": 1.38, "learning_rate": 1.9399574337732062e-05, "loss": 1.3417, "step": 4476 }, { "epoch": 1.38, "learning_rate": 1.939923418019844e-05, "loss": 1.7208, "step": 4477 }, { "epoch": 1.38, "learning_rate": 1.9398893929321763e-05, "loss": 1.0482, "step": 4478 }, { "epoch": 1.38, "learning_rate": 1.9398553585105405e-05, "loss": 1.1802, "step": 4479 }, { "epoch": 1.38, "learning_rate": 1.939821314755275e-05, "loss": 1.3315, "step": 4480 }, { "epoch": 1.38, "learning_rate": 1.9397872616667175e-05, "loss": 1.2316, "step": 4481 }, { "epoch": 1.38, "learning_rate": 1.9397531992452062e-05, "loss": 1.1249, "step": 4482 }, { "epoch": 1.38, "learning_rate": 1.9397191274910797e-05, "loss": 3.6207, "step": 4483 }, { "epoch": 1.38, "learning_rate": 1.9396850464046767e-05, "loss": 0.8512, "step": 4484 }, { "epoch": 1.38, "learning_rate": 1.9396509559863344e-05, "loss": 1.0561, "step": 4485 }, { "epoch": 1.38, "learning_rate": 1.9396168562363925e-05, "loss": 1.0724, "step": 4486 }, { "epoch": 1.38, "learning_rate": 1.939582747155189e-05, "loss": 0.9705, "step": 4487 }, { "epoch": 1.38, "learning_rate": 1.9395486287430628e-05, "loss": 1.3912, "step": 4488 }, { "epoch": 1.38, "learning_rate": 1.9395145010003533e-05, "loss": 2.9781, "step": 4489 }, { "epoch": 1.38, "learning_rate": 1.9394803639273988e-05, "loss": 1.5824, "step": 4490 }, { "epoch": 1.38, "learning_rate": 1.9394462175245382e-05, "loss": 1.2448, "step": 4491 }, { "epoch": 1.38, "learning_rate": 1.9394120617921105e-05, "loss": 1.3865, "step": 4492 }, { "epoch": 1.38, "learning_rate": 1.939377896730456e-05, "loss": 1.3789, "step": 4493 }, { "epoch": 1.38, "learning_rate": 1.939343722339913e-05, "loss": 1.0335, "step": 4494 }, { "epoch": 1.38, "learning_rate": 1.9393095386208208e-05, "loss": 1.1351, "step": 4495 }, { "epoch": 1.38, "learning_rate": 1.9392753455735194e-05, "loss": 1.2156, "step": 4496 }, { "epoch": 1.38, "learning_rate": 1.939241143198348e-05, "loss": 1.3077, "step": 4497 }, { "epoch": 1.38, "learning_rate": 1.9392069314956464e-05, "loss": 1.2435, "step": 4498 }, { "epoch": 1.38, "learning_rate": 1.9391727104657548e-05, "loss": 1.0224, "step": 4499 }, { "epoch": 1.38, "learning_rate": 1.9391384801090122e-05, "loss": 1.4452, "step": 4500 }, { "epoch": 1.38, "learning_rate": 1.9391042404257593e-05, "loss": 1.0328, "step": 4501 }, { "epoch": 1.39, "learning_rate": 1.9390699914163352e-05, "loss": 1.3525, "step": 4502 }, { "epoch": 1.39, "learning_rate": 1.9390357330810814e-05, "loss": 1.3745, "step": 4503 }, { "epoch": 1.39, "learning_rate": 1.939001465420337e-05, "loss": 0.9863, "step": 4504 }, { "epoch": 1.39, "learning_rate": 1.9389671884344422e-05, "loss": 1.1788, "step": 4505 }, { "epoch": 1.39, "learning_rate": 1.9389329021237383e-05, "loss": 1.2292, "step": 4506 }, { "epoch": 1.39, "learning_rate": 1.9388986064885653e-05, "loss": 1.5413, "step": 4507 }, { "epoch": 1.39, "learning_rate": 1.9388643015292638e-05, "loss": 1.2172, "step": 4508 }, { "epoch": 1.39, "learning_rate": 1.9388299872461743e-05, "loss": 1.1545, "step": 4509 }, { "epoch": 1.39, "learning_rate": 1.938795663639638e-05, "loss": 1.1933, "step": 4510 }, { "epoch": 1.39, "learning_rate": 1.9387613307099956e-05, "loss": 1.1566, "step": 4511 }, { "epoch": 1.39, "learning_rate": 1.938726988457588e-05, "loss": 0.8647, "step": 4512 }, { "epoch": 1.39, "learning_rate": 1.938692636882756e-05, "loss": 1.085, "step": 4513 }, { "epoch": 1.39, "learning_rate": 1.9386582759858413e-05, "loss": 1.3169, "step": 4514 }, { "epoch": 1.39, "learning_rate": 1.9386239057671845e-05, "loss": 1.1198, "step": 4515 }, { "epoch": 1.39, "learning_rate": 1.9385895262271273e-05, "loss": 1.0413, "step": 4516 }, { "epoch": 1.39, "learning_rate": 1.9385551373660113e-05, "loss": 1.6183, "step": 4517 }, { "epoch": 1.39, "learning_rate": 1.9385207391841776e-05, "loss": 1.2316, "step": 4518 }, { "epoch": 1.39, "learning_rate": 1.938486331681968e-05, "loss": 1.4488, "step": 4519 }, { "epoch": 1.39, "learning_rate": 1.9384519148597242e-05, "loss": 1.274, "step": 4520 }, { "epoch": 1.39, "learning_rate": 1.938417488717788e-05, "loss": 1.1096, "step": 4521 }, { "epoch": 1.39, "learning_rate": 1.9383830532565014e-05, "loss": 1.0805, "step": 4522 }, { "epoch": 1.39, "learning_rate": 1.9383486084762062e-05, "loss": 0.9051, "step": 4523 }, { "epoch": 1.39, "learning_rate": 1.9383141543772444e-05, "loss": 1.3032, "step": 4524 }, { "epoch": 1.39, "learning_rate": 1.9382796909599584e-05, "loss": 1.2296, "step": 4525 }, { "epoch": 1.39, "learning_rate": 1.9382452182246902e-05, "loss": 1.1672, "step": 4526 }, { "epoch": 1.39, "learning_rate": 1.9382107361717825e-05, "loss": 1.2608, "step": 4527 }, { "epoch": 1.39, "learning_rate": 1.9381762448015774e-05, "loss": 1.1487, "step": 4528 }, { "epoch": 1.39, "learning_rate": 1.938141744114417e-05, "loss": 1.035, "step": 4529 }, { "epoch": 1.39, "learning_rate": 1.9381072341106453e-05, "loss": 1.2141, "step": 4530 }, { "epoch": 1.39, "learning_rate": 1.9380727147906037e-05, "loss": 0.9633, "step": 4531 }, { "epoch": 1.39, "learning_rate": 1.9380381861546355e-05, "loss": 1.3636, "step": 4532 }, { "epoch": 1.39, "learning_rate": 1.9380036482030837e-05, "loss": 1.0667, "step": 4533 }, { "epoch": 1.4, "learning_rate": 1.9379691009362915e-05, "loss": 1.1127, "step": 4534 }, { "epoch": 1.4, "learning_rate": 1.937934544354601e-05, "loss": 1.1784, "step": 4535 }, { "epoch": 1.4, "learning_rate": 1.9378999784583565e-05, "loss": 1.3365, "step": 4536 }, { "epoch": 1.4, "learning_rate": 1.9378654032479006e-05, "loss": 1.1863, "step": 4537 }, { "epoch": 1.4, "learning_rate": 1.937830818723577e-05, "loss": 1.1467, "step": 4538 }, { "epoch": 1.4, "learning_rate": 1.9377962248857293e-05, "loss": 1.7257, "step": 4539 }, { "epoch": 1.4, "learning_rate": 1.9377616217347004e-05, "loss": 0.9723, "step": 4540 }, { "epoch": 1.4, "learning_rate": 1.9377270092708345e-05, "loss": 1.4796, "step": 4541 }, { "epoch": 1.4, "learning_rate": 1.937692387494475e-05, "loss": 1.2442, "step": 4542 }, { "epoch": 1.4, "learning_rate": 1.937657756405966e-05, "loss": 1.0296, "step": 4543 }, { "epoch": 1.4, "learning_rate": 1.9376231160056512e-05, "loss": 1.1614, "step": 4544 }, { "epoch": 1.4, "learning_rate": 1.937588466293875e-05, "loss": 1.2739, "step": 4545 }, { "epoch": 1.4, "learning_rate": 1.937553807270981e-05, "loss": 1.4576, "step": 4546 }, { "epoch": 1.4, "learning_rate": 1.9375191389373137e-05, "loss": 1.1306, "step": 4547 }, { "epoch": 1.4, "learning_rate": 1.9374844612932173e-05, "loss": 1.3795, "step": 4548 }, { "epoch": 1.4, "learning_rate": 1.9374497743390362e-05, "loss": 1.2962, "step": 4549 }, { "epoch": 1.4, "learning_rate": 1.937415078075115e-05, "loss": 1.2359, "step": 4550 }, { "epoch": 1.4, "learning_rate": 1.9373803725017975e-05, "loss": 1.0389, "step": 4551 }, { "epoch": 1.4, "learning_rate": 1.9373456576194298e-05, "loss": 1.3009, "step": 4552 }, { "epoch": 1.4, "learning_rate": 1.9373109334283552e-05, "loss": 1.4145, "step": 4553 }, { "epoch": 1.4, "learning_rate": 1.9372761999289195e-05, "loss": 1.3093, "step": 4554 }, { "epoch": 1.4, "learning_rate": 1.9372414571214675e-05, "loss": 1.2827, "step": 4555 }, { "epoch": 1.4, "learning_rate": 1.937206705006344e-05, "loss": 1.1773, "step": 4556 }, { "epoch": 1.4, "learning_rate": 1.9371719435838938e-05, "loss": 1.3472, "step": 4557 }, { "epoch": 1.4, "learning_rate": 1.9371371728544628e-05, "loss": 1.0389, "step": 4558 }, { "epoch": 1.4, "learning_rate": 1.9371023928183956e-05, "loss": 1.2689, "step": 4559 }, { "epoch": 1.4, "learning_rate": 1.9370676034760384e-05, "loss": 1.2133, "step": 4560 }, { "epoch": 1.4, "learning_rate": 1.937032804827736e-05, "loss": 1.3266, "step": 4561 }, { "epoch": 1.4, "learning_rate": 1.9369979968738343e-05, "loss": 1.0651, "step": 4562 }, { "epoch": 1.4, "learning_rate": 1.9369631796146792e-05, "loss": 1.0326, "step": 4563 }, { "epoch": 1.4, "learning_rate": 1.936928353050616e-05, "loss": 1.0335, "step": 4564 }, { "epoch": 1.4, "learning_rate": 1.9368935171819908e-05, "loss": 0.942, "step": 4565 }, { "epoch": 1.4, "learning_rate": 1.9368586720091492e-05, "loss": 1.1761, "step": 4566 }, { "epoch": 1.41, "learning_rate": 1.9368238175324378e-05, "loss": 1.2691, "step": 4567 }, { "epoch": 1.41, "learning_rate": 1.9367889537522024e-05, "loss": 1.0504, "step": 4568 }, { "epoch": 1.41, "learning_rate": 1.9367540806687894e-05, "loss": 1.1639, "step": 4569 }, { "epoch": 1.41, "learning_rate": 1.936719198282545e-05, "loss": 0.8441, "step": 4570 }, { "epoch": 1.41, "learning_rate": 1.936684306593816e-05, "loss": 0.9737, "step": 4571 }, { "epoch": 1.41, "learning_rate": 1.936649405602948e-05, "loss": 1.2283, "step": 4572 }, { "epoch": 1.41, "learning_rate": 1.9366144953102886e-05, "loss": 1.246, "step": 4573 }, { "epoch": 1.41, "learning_rate": 1.9365795757161837e-05, "loss": 1.2938, "step": 4574 }, { "epoch": 1.41, "learning_rate": 1.9365446468209806e-05, "loss": 1.091, "step": 4575 }, { "epoch": 1.41, "learning_rate": 1.936509708625026e-05, "loss": 1.3653, "step": 4576 }, { "epoch": 1.41, "learning_rate": 1.936474761128667e-05, "loss": 1.115, "step": 4577 }, { "epoch": 1.41, "learning_rate": 1.9364398043322505e-05, "loss": 1.3123, "step": 4578 }, { "epoch": 1.41, "learning_rate": 1.9364048382361234e-05, "loss": 1.0343, "step": 4579 }, { "epoch": 1.41, "learning_rate": 1.9363698628406337e-05, "loss": 1.1655, "step": 4580 }, { "epoch": 1.41, "learning_rate": 1.936334878146128e-05, "loss": 1.1561, "step": 4581 }, { "epoch": 1.41, "learning_rate": 1.9362998841529542e-05, "loss": 1.2338, "step": 4582 }, { "epoch": 1.41, "learning_rate": 1.9362648808614593e-05, "loss": 1.5557, "step": 4583 }, { "epoch": 1.41, "learning_rate": 1.9362298682719913e-05, "loss": 1.1928, "step": 4584 }, { "epoch": 1.41, "learning_rate": 1.9361948463848984e-05, "loss": 1.3308, "step": 4585 }, { "epoch": 1.41, "learning_rate": 1.9361598152005272e-05, "loss": 1.0606, "step": 4586 }, { "epoch": 1.41, "learning_rate": 1.9361247747192268e-05, "loss": 1.0559, "step": 4587 }, { "epoch": 1.41, "learning_rate": 1.936089724941344e-05, "loss": 1.4476, "step": 4588 }, { "epoch": 1.41, "learning_rate": 1.936054665867228e-05, "loss": 1.2856, "step": 4589 }, { "epoch": 1.41, "learning_rate": 1.936019597497226e-05, "loss": 1.3044, "step": 4590 }, { "epoch": 1.41, "learning_rate": 1.935984519831687e-05, "loss": 1.3659, "step": 4591 }, { "epoch": 1.41, "learning_rate": 1.935949432870959e-05, "loss": 1.4665, "step": 4592 }, { "epoch": 1.41, "learning_rate": 1.9359143366153908e-05, "loss": 1.1571, "step": 4593 }, { "epoch": 1.41, "learning_rate": 1.9358792310653308e-05, "loss": 1.0972, "step": 4594 }, { "epoch": 1.41, "learning_rate": 1.935844116221127e-05, "loss": 1.2085, "step": 4595 }, { "epoch": 1.41, "learning_rate": 1.935808992083129e-05, "loss": 1.1378, "step": 4596 }, { "epoch": 1.41, "learning_rate": 1.935773858651685e-05, "loss": 1.0707, "step": 4597 }, { "epoch": 1.41, "learning_rate": 1.9357387159271444e-05, "loss": 1.1182, "step": 4598 }, { "epoch": 1.42, "learning_rate": 1.935703563909856e-05, "loss": 1.1694, "step": 4599 }, { "epoch": 1.42, "learning_rate": 1.9356684026001686e-05, "loss": 1.2245, "step": 4600 }, { "epoch": 1.42, "learning_rate": 1.9356332319984315e-05, "loss": 1.2593, "step": 4601 }, { "epoch": 1.42, "learning_rate": 1.9355980521049943e-05, "loss": 1.305, "step": 4602 }, { "epoch": 1.42, "learning_rate": 1.935562862920206e-05, "loss": 1.3243, "step": 4603 }, { "epoch": 1.42, "learning_rate": 1.9355276644444166e-05, "loss": 1.1172, "step": 4604 }, { "epoch": 1.42, "learning_rate": 1.935492456677975e-05, "loss": 1.054, "step": 4605 }, { "epoch": 1.42, "learning_rate": 1.935457239621231e-05, "loss": 1.3189, "step": 4606 }, { "epoch": 1.42, "learning_rate": 1.9354220132745347e-05, "loss": 1.4782, "step": 4607 }, { "epoch": 1.42, "learning_rate": 1.9353867776382357e-05, "loss": 1.3226, "step": 4608 }, { "epoch": 1.42, "learning_rate": 1.9353515327126836e-05, "loss": 1.2175, "step": 4609 }, { "epoch": 1.42, "learning_rate": 1.9353162784982294e-05, "loss": 1.1233, "step": 4610 }, { "epoch": 1.42, "learning_rate": 1.935281014995222e-05, "loss": 1.2675, "step": 4611 }, { "epoch": 1.42, "learning_rate": 1.9352457422040123e-05, "loss": 1.0448, "step": 4612 }, { "epoch": 1.42, "learning_rate": 1.9352104601249504e-05, "loss": 1.023, "step": 4613 }, { "epoch": 1.42, "learning_rate": 1.9351751687583866e-05, "loss": 1.4691, "step": 4614 }, { "epoch": 1.42, "learning_rate": 1.9351398681046718e-05, "loss": 1.2438, "step": 4615 }, { "epoch": 1.42, "learning_rate": 1.935104558164156e-05, "loss": 0.9931, "step": 4616 }, { "epoch": 1.42, "learning_rate": 1.9350692389371904e-05, "loss": 1.0977, "step": 4617 }, { "epoch": 1.42, "learning_rate": 1.935033910424125e-05, "loss": 1.5207, "step": 4618 }, { "epoch": 1.42, "learning_rate": 1.9349985726253115e-05, "loss": 1.1286, "step": 4619 }, { "epoch": 1.42, "learning_rate": 1.9349632255411004e-05, "loss": 0.9843, "step": 4620 }, { "epoch": 1.42, "learning_rate": 1.9349278691718426e-05, "loss": 1.0525, "step": 4621 }, { "epoch": 1.42, "learning_rate": 1.9348925035178897e-05, "loss": 1.2014, "step": 4622 }, { "epoch": 1.42, "learning_rate": 1.9348571285795923e-05, "loss": 1.2891, "step": 4623 }, { "epoch": 1.42, "learning_rate": 1.9348217443573023e-05, "loss": 1.0459, "step": 4624 }, { "epoch": 1.42, "learning_rate": 1.9347863508513706e-05, "loss": 1.1395, "step": 4625 }, { "epoch": 1.42, "learning_rate": 1.9347509480621492e-05, "loss": 0.8998, "step": 4626 }, { "epoch": 1.42, "learning_rate": 1.934715535989989e-05, "loss": 1.5491, "step": 4627 }, { "epoch": 1.42, "learning_rate": 1.9346801146352425e-05, "loss": 1.1488, "step": 4628 }, { "epoch": 1.42, "learning_rate": 1.9346446839982605e-05, "loss": 0.9931, "step": 4629 }, { "epoch": 1.42, "learning_rate": 1.934609244079396e-05, "loss": 1.0289, "step": 4630 }, { "epoch": 1.42, "learning_rate": 1.934573794879e-05, "loss": 1.0704, "step": 4631 }, { "epoch": 1.43, "learning_rate": 1.9345383363974248e-05, "loss": 1.2242, "step": 4632 }, { "epoch": 1.43, "learning_rate": 1.934502868635023e-05, "loss": 1.0815, "step": 4633 }, { "epoch": 1.43, "learning_rate": 1.934467391592146e-05, "loss": 1.2222, "step": 4634 }, { "epoch": 1.43, "learning_rate": 1.934431905269147e-05, "loss": 1.2829, "step": 4635 }, { "epoch": 1.43, "learning_rate": 1.9343964096663774e-05, "loss": 1.2422, "step": 4636 }, { "epoch": 1.43, "learning_rate": 1.934360904784191e-05, "loss": 0.9855, "step": 4637 }, { "epoch": 1.43, "learning_rate": 1.9343253906229393e-05, "loss": 1.1189, "step": 4638 }, { "epoch": 1.43, "learning_rate": 1.9342898671829754e-05, "loss": 1.213, "step": 4639 }, { "epoch": 1.43, "learning_rate": 1.934254334464652e-05, "loss": 1.2657, "step": 4640 }, { "epoch": 1.43, "learning_rate": 1.934218792468322e-05, "loss": 1.3615, "step": 4641 }, { "epoch": 1.43, "learning_rate": 1.9341832411943384e-05, "loss": 1.0734, "step": 4642 }, { "epoch": 1.43, "learning_rate": 1.9341476806430544e-05, "loss": 1.1091, "step": 4643 }, { "epoch": 1.43, "learning_rate": 1.934112110814823e-05, "loss": 1.4369, "step": 4644 }, { "epoch": 1.43, "learning_rate": 1.934076531709997e-05, "loss": 1.2173, "step": 4645 }, { "epoch": 1.43, "learning_rate": 1.9340409433289306e-05, "loss": 1.132, "step": 4646 }, { "epoch": 1.43, "learning_rate": 1.9340053456719768e-05, "loss": 1.3852, "step": 4647 }, { "epoch": 1.43, "learning_rate": 1.933969738739489e-05, "loss": 1.4139, "step": 4648 }, { "epoch": 1.43, "learning_rate": 1.933934122531821e-05, "loss": 1.5365, "step": 4649 }, { "epoch": 1.43, "learning_rate": 1.9338984970493263e-05, "loss": 1.5207, "step": 4650 }, { "epoch": 1.43, "learning_rate": 1.933862862292359e-05, "loss": 1.2672, "step": 4651 }, { "epoch": 1.43, "learning_rate": 1.933827218261273e-05, "loss": 1.451, "step": 4652 }, { "epoch": 1.43, "learning_rate": 1.9337915649564215e-05, "loss": 1.1123, "step": 4653 }, { "epoch": 1.43, "learning_rate": 1.9337559023781596e-05, "loss": 1.4258, "step": 4654 }, { "epoch": 1.43, "learning_rate": 1.9337202305268405e-05, "loss": 0.9331, "step": 4655 }, { "epoch": 1.43, "learning_rate": 1.9336845494028193e-05, "loss": 1.1595, "step": 4656 }, { "epoch": 1.43, "learning_rate": 1.9336488590064504e-05, "loss": 1.1709, "step": 4657 }, { "epoch": 1.43, "learning_rate": 1.9336131593380875e-05, "loss": 1.2705, "step": 4658 }, { "epoch": 1.43, "learning_rate": 1.9335774503980854e-05, "loss": 1.2493, "step": 4659 }, { "epoch": 1.43, "learning_rate": 1.9335417321867988e-05, "loss": 1.2732, "step": 4660 }, { "epoch": 1.43, "learning_rate": 1.9335060047045827e-05, "loss": 1.0881, "step": 4661 }, { "epoch": 1.43, "learning_rate": 1.9334702679517913e-05, "loss": 0.9702, "step": 4662 }, { "epoch": 1.43, "learning_rate": 1.93343452192878e-05, "loss": 1.4056, "step": 4663 }, { "epoch": 1.44, "learning_rate": 1.9333987666359033e-05, "loss": 1.3206, "step": 4664 }, { "epoch": 1.44, "learning_rate": 1.933363002073517e-05, "loss": 1.1712, "step": 4665 }, { "epoch": 1.44, "learning_rate": 1.9333272282419758e-05, "loss": 1.0789, "step": 4666 }, { "epoch": 1.44, "learning_rate": 1.933291445141635e-05, "loss": 1.2305, "step": 4667 }, { "epoch": 1.44, "learning_rate": 1.93325565277285e-05, "loss": 1.3716, "step": 4668 }, { "epoch": 1.44, "learning_rate": 1.9332198511359762e-05, "loss": 1.5047, "step": 4669 }, { "epoch": 1.44, "learning_rate": 1.9331840402313694e-05, "loss": 1.0984, "step": 4670 }, { "epoch": 1.44, "learning_rate": 1.9331482200593848e-05, "loss": 1.0936, "step": 4671 }, { "epoch": 1.44, "learning_rate": 1.9331123906203784e-05, "loss": 1.2653, "step": 4672 }, { "epoch": 1.44, "learning_rate": 1.9330765519147058e-05, "loss": 1.1085, "step": 4673 }, { "epoch": 1.44, "learning_rate": 1.9330407039427235e-05, "loss": 0.9891, "step": 4674 }, { "epoch": 1.44, "learning_rate": 1.933004846704787e-05, "loss": 1.0307, "step": 4675 }, { "epoch": 1.44, "learning_rate": 1.9329689802012525e-05, "loss": 1.1701, "step": 4676 }, { "epoch": 1.44, "learning_rate": 1.932933104432476e-05, "loss": 0.9485, "step": 4677 }, { "epoch": 1.44, "learning_rate": 1.932897219398814e-05, "loss": 1.1421, "step": 4678 }, { "epoch": 1.44, "learning_rate": 1.932861325100623e-05, "loss": 1.2078, "step": 4679 }, { "epoch": 1.44, "learning_rate": 1.932825421538259e-05, "loss": 1.4705, "step": 4680 }, { "epoch": 1.44, "learning_rate": 1.932789508712079e-05, "loss": 1.4064, "step": 4681 }, { "epoch": 1.44, "learning_rate": 1.9327535866224393e-05, "loss": 1.5551, "step": 4682 }, { "epoch": 1.44, "learning_rate": 1.932717655269697e-05, "loss": 1.3065, "step": 4683 }, { "epoch": 1.44, "learning_rate": 1.932681714654209e-05, "loss": 1.0218, "step": 4684 }, { "epoch": 1.44, "learning_rate": 1.9326457647763318e-05, "loss": 1.6698, "step": 4685 }, { "epoch": 1.44, "learning_rate": 1.9326098056364224e-05, "loss": 0.9114, "step": 4686 }, { "epoch": 1.44, "learning_rate": 1.9325738372348383e-05, "loss": 1.4047, "step": 4687 }, { "epoch": 1.44, "learning_rate": 1.9325378595719366e-05, "loss": 1.2763, "step": 4688 }, { "epoch": 1.44, "learning_rate": 1.9325018726480743e-05, "loss": 1.2547, "step": 4689 }, { "epoch": 1.44, "learning_rate": 1.932465876463609e-05, "loss": 1.1243, "step": 4690 }, { "epoch": 1.44, "learning_rate": 1.9324298710188986e-05, "loss": 1.1519, "step": 4691 }, { "epoch": 1.44, "learning_rate": 1.9323938563142998e-05, "loss": 1.3331, "step": 4692 }, { "epoch": 1.44, "learning_rate": 1.9323578323501705e-05, "loss": 1.2158, "step": 4693 }, { "epoch": 1.44, "learning_rate": 1.932321799126869e-05, "loss": 1.4635, "step": 4694 }, { "epoch": 1.44, "learning_rate": 1.9322857566447527e-05, "loss": 1.0647, "step": 4695 }, { "epoch": 1.44, "learning_rate": 1.9322497049041797e-05, "loss": 1.1505, "step": 4696 }, { "epoch": 1.45, "learning_rate": 1.9322136439055077e-05, "loss": 1.1865, "step": 4697 }, { "epoch": 1.45, "learning_rate": 1.9321775736490953e-05, "loss": 1.2574, "step": 4698 }, { "epoch": 1.45, "learning_rate": 1.9321414941353006e-05, "loss": 1.2875, "step": 4699 }, { "epoch": 1.45, "learning_rate": 1.9321054053644814e-05, "loss": 1.2492, "step": 4700 }, { "epoch": 1.45, "learning_rate": 1.9320693073369967e-05, "loss": 1.7145, "step": 4701 }, { "epoch": 1.45, "learning_rate": 1.9320332000532045e-05, "loss": 1.2504, "step": 4702 }, { "epoch": 1.45, "learning_rate": 1.9319970835134637e-05, "loss": 0.9237, "step": 4703 }, { "epoch": 1.45, "learning_rate": 1.9319609577181334e-05, "loss": 1.3879, "step": 4704 }, { "epoch": 1.45, "learning_rate": 1.9319248226675714e-05, "loss": 1.3145, "step": 4705 }, { "epoch": 1.45, "learning_rate": 1.931888678362137e-05, "loss": 1.2287, "step": 4706 }, { "epoch": 1.45, "learning_rate": 1.9318525248021896e-05, "loss": 0.9778, "step": 4707 }, { "epoch": 1.45, "learning_rate": 1.9318163619880872e-05, "loss": 1.3596, "step": 4708 }, { "epoch": 1.45, "learning_rate": 1.93178018992019e-05, "loss": 1.2423, "step": 4709 }, { "epoch": 1.45, "learning_rate": 1.9317440085988567e-05, "loss": 1.6999, "step": 4710 }, { "epoch": 1.45, "learning_rate": 1.9317078180244466e-05, "loss": 1.1031, "step": 4711 }, { "epoch": 1.45, "learning_rate": 1.931671618197319e-05, "loss": 1.1376, "step": 4712 }, { "epoch": 1.45, "learning_rate": 1.9316354091178338e-05, "loss": 1.2538, "step": 4713 }, { "epoch": 1.45, "learning_rate": 1.9315991907863504e-05, "loss": 1.3629, "step": 4714 }, { "epoch": 1.45, "learning_rate": 1.931562963203228e-05, "loss": 1.357, "step": 4715 }, { "epoch": 1.45, "learning_rate": 1.9315267263688273e-05, "loss": 1.209, "step": 4716 }, { "epoch": 1.45, "learning_rate": 1.9314904802835074e-05, "loss": 1.3137, "step": 4717 }, { "epoch": 1.45, "learning_rate": 1.931454224947629e-05, "loss": 1.1187, "step": 4718 }, { "epoch": 1.45, "learning_rate": 1.9314179603615514e-05, "loss": 1.2079, "step": 4719 }, { "epoch": 1.45, "learning_rate": 1.9313816865256348e-05, "loss": 1.2712, "step": 4720 }, { "epoch": 1.45, "learning_rate": 1.93134540344024e-05, "loss": 1.2066, "step": 4721 }, { "epoch": 1.45, "learning_rate": 1.9313091111057266e-05, "loss": 1.2895, "step": 4722 }, { "epoch": 1.45, "learning_rate": 1.931272809522456e-05, "loss": 1.0855, "step": 4723 }, { "epoch": 1.45, "learning_rate": 1.9312364986907875e-05, "loss": 1.2083, "step": 4724 }, { "epoch": 1.45, "learning_rate": 1.931200178611083e-05, "loss": 1.162, "step": 4725 }, { "epoch": 1.45, "learning_rate": 1.931163849283702e-05, "loss": 1.1148, "step": 4726 }, { "epoch": 1.45, "learning_rate": 1.9311275107090058e-05, "loss": 1.0289, "step": 4727 }, { "epoch": 1.45, "learning_rate": 1.9310911628873556e-05, "loss": 1.2466, "step": 4728 }, { "epoch": 1.46, "learning_rate": 1.9310548058191115e-05, "loss": 0.9953, "step": 4729 }, { "epoch": 1.46, "learning_rate": 1.9310184395046358e-05, "loss": 1.3991, "step": 4730 }, { "epoch": 1.46, "learning_rate": 1.9309820639442886e-05, "loss": 1.4677, "step": 4731 }, { "epoch": 1.46, "learning_rate": 1.9309456791384314e-05, "loss": 1.2085, "step": 4732 }, { "epoch": 1.46, "learning_rate": 1.9309092850874254e-05, "loss": 1.3794, "step": 4733 }, { "epoch": 1.46, "learning_rate": 1.9308728817916325e-05, "loss": 1.0029, "step": 4734 }, { "epoch": 1.46, "learning_rate": 1.9308364692514142e-05, "loss": 1.197, "step": 4735 }, { "epoch": 1.46, "learning_rate": 1.9308000474671316e-05, "loss": 1.356, "step": 4736 }, { "epoch": 1.46, "learning_rate": 1.9307636164391466e-05, "loss": 1.142, "step": 4737 }, { "epoch": 1.46, "learning_rate": 1.9307271761678214e-05, "loss": 1.3134, "step": 4738 }, { "epoch": 1.46, "learning_rate": 1.9306907266535172e-05, "loss": 1.0578, "step": 4739 }, { "epoch": 1.46, "learning_rate": 1.9306542678965968e-05, "loss": 1.2111, "step": 4740 }, { "epoch": 1.46, "learning_rate": 1.930617799897421e-05, "loss": 1.1766, "step": 4741 }, { "epoch": 1.46, "learning_rate": 1.9305813226563535e-05, "loss": 1.0196, "step": 4742 }, { "epoch": 1.46, "learning_rate": 1.9305448361737554e-05, "loss": 1.4198, "step": 4743 }, { "epoch": 1.46, "learning_rate": 1.93050834044999e-05, "loss": 1.2848, "step": 4744 }, { "epoch": 1.46, "learning_rate": 1.9304718354854185e-05, "loss": 1.3858, "step": 4745 }, { "epoch": 1.46, "learning_rate": 1.9304353212804047e-05, "loss": 0.9239, "step": 4746 }, { "epoch": 1.46, "learning_rate": 1.93039879783531e-05, "loss": 0.977, "step": 4747 }, { "epoch": 1.46, "learning_rate": 1.9303622651504986e-05, "loss": 1.2169, "step": 4748 }, { "epoch": 1.46, "learning_rate": 1.9303257232263323e-05, "loss": 1.6657, "step": 4749 }, { "epoch": 1.46, "learning_rate": 1.9302891720631735e-05, "loss": 1.3172, "step": 4750 }, { "epoch": 1.46, "learning_rate": 1.9302526116613863e-05, "loss": 1.2613, "step": 4751 }, { "epoch": 1.46, "learning_rate": 1.9302160420213333e-05, "loss": 0.8685, "step": 4752 }, { "epoch": 1.46, "learning_rate": 1.9301794631433774e-05, "loss": 1.3352, "step": 4753 }, { "epoch": 1.46, "learning_rate": 1.9301428750278824e-05, "loss": 1.3726, "step": 4754 }, { "epoch": 1.46, "learning_rate": 1.9301062776752116e-05, "loss": 1.3523, "step": 4755 }, { "epoch": 1.46, "learning_rate": 1.9300696710857282e-05, "loss": 1.1591, "step": 4756 }, { "epoch": 1.46, "learning_rate": 1.9300330552597953e-05, "loss": 1.0816, "step": 4757 }, { "epoch": 1.46, "learning_rate": 1.9299964301977776e-05, "loss": 1.6117, "step": 4758 }, { "epoch": 1.46, "learning_rate": 1.929959795900038e-05, "loss": 1.0877, "step": 4759 }, { "epoch": 1.46, "learning_rate": 1.9299231523669407e-05, "loss": 1.0713, "step": 4760 }, { "epoch": 1.46, "learning_rate": 1.929886499598849e-05, "loss": 1.2126, "step": 4761 }, { "epoch": 1.47, "learning_rate": 1.9298498375961275e-05, "loss": 1.0795, "step": 4762 }, { "epoch": 1.47, "learning_rate": 1.9298131663591407e-05, "loss": 1.4427, "step": 4763 }, { "epoch": 1.47, "learning_rate": 1.9297764858882516e-05, "loss": 1.312, "step": 4764 }, { "epoch": 1.47, "learning_rate": 1.9297397961838252e-05, "loss": 1.3802, "step": 4765 }, { "epoch": 1.47, "learning_rate": 1.9297030972462257e-05, "loss": 1.2048, "step": 4766 }, { "epoch": 1.47, "learning_rate": 1.929666389075818e-05, "loss": 1.2378, "step": 4767 }, { "epoch": 1.47, "learning_rate": 1.929629671672966e-05, "loss": 1.0785, "step": 4768 }, { "epoch": 1.47, "learning_rate": 1.9295929450380346e-05, "loss": 1.2688, "step": 4769 }, { "epoch": 1.47, "learning_rate": 1.9295562091713883e-05, "loss": 1.6597, "step": 4770 }, { "epoch": 1.47, "learning_rate": 1.9295194640733924e-05, "loss": 1.3675, "step": 4771 }, { "epoch": 1.47, "learning_rate": 1.9294827097444114e-05, "loss": 1.6053, "step": 4772 }, { "epoch": 1.47, "learning_rate": 1.9294459461848107e-05, "loss": 0.9442, "step": 4773 }, { "epoch": 1.47, "learning_rate": 1.929409173394955e-05, "loss": 1.1209, "step": 4774 }, { "epoch": 1.47, "learning_rate": 1.9293723913752097e-05, "loss": 1.215, "step": 4775 }, { "epoch": 1.47, "learning_rate": 1.92933560012594e-05, "loss": 1.1213, "step": 4776 }, { "epoch": 1.47, "learning_rate": 1.9292987996475113e-05, "loss": 0.9779, "step": 4777 }, { "epoch": 1.47, "learning_rate": 1.9292619899402888e-05, "loss": 1.3766, "step": 4778 }, { "epoch": 1.47, "learning_rate": 1.9292251710046386e-05, "loss": 1.4494, "step": 4779 }, { "epoch": 1.47, "learning_rate": 1.929188342840926e-05, "loss": 1.4625, "step": 4780 }, { "epoch": 1.47, "learning_rate": 1.9291515054495167e-05, "loss": 1.183, "step": 4781 }, { "epoch": 1.47, "learning_rate": 1.9291146588307765e-05, "loss": 1.3836, "step": 4782 }, { "epoch": 1.47, "learning_rate": 1.929077802985072e-05, "loss": 1.2089, "step": 4783 }, { "epoch": 1.47, "learning_rate": 1.929040937912768e-05, "loss": 1.3718, "step": 4784 }, { "epoch": 1.47, "learning_rate": 1.9290040636142315e-05, "loss": 1.3192, "step": 4785 }, { "epoch": 1.47, "learning_rate": 1.9289671800898285e-05, "loss": 1.1724, "step": 4786 }, { "epoch": 1.47, "learning_rate": 1.928930287339925e-05, "loss": 1.3215, "step": 4787 }, { "epoch": 1.47, "learning_rate": 1.9288933853648882e-05, "loss": 1.1019, "step": 4788 }, { "epoch": 1.47, "learning_rate": 1.9288564741650836e-05, "loss": 1.1081, "step": 4789 }, { "epoch": 1.47, "learning_rate": 1.928819553740878e-05, "loss": 1.1076, "step": 4790 }, { "epoch": 1.47, "learning_rate": 1.9287826240926385e-05, "loss": 1.0323, "step": 4791 }, { "epoch": 1.47, "learning_rate": 1.9287456852207313e-05, "loss": 1.1659, "step": 4792 }, { "epoch": 1.47, "learning_rate": 1.9287087371255237e-05, "loss": 1.1695, "step": 4793 }, { "epoch": 1.48, "learning_rate": 1.9286717798073826e-05, "loss": 1.3717, "step": 4794 }, { "epoch": 1.48, "learning_rate": 1.9286348132666742e-05, "loss": 1.0784, "step": 4795 }, { "epoch": 1.48, "learning_rate": 1.9285978375037666e-05, "loss": 1.1033, "step": 4796 }, { "epoch": 1.48, "learning_rate": 1.9285608525190266e-05, "loss": 1.1712, "step": 4797 }, { "epoch": 1.48, "learning_rate": 1.9285238583128215e-05, "loss": 0.9174, "step": 4798 }, { "epoch": 1.48, "learning_rate": 1.9284868548855193e-05, "loss": 1.0051, "step": 4799 }, { "epoch": 1.48, "learning_rate": 1.9284498422374862e-05, "loss": 1.0688, "step": 4800 }, { "epoch": 1.48, "learning_rate": 1.928412820369091e-05, "loss": 0.7519, "step": 4801 }, { "epoch": 1.48, "learning_rate": 1.9283757892807003e-05, "loss": 1.2856, "step": 4802 }, { "epoch": 1.48, "learning_rate": 1.9283387489726827e-05, "loss": 1.2562, "step": 4803 }, { "epoch": 1.48, "learning_rate": 1.9283016994454057e-05, "loss": 1.1758, "step": 4804 }, { "epoch": 1.48, "learning_rate": 1.9282646406992375e-05, "loss": 1.2436, "step": 4805 }, { "epoch": 1.48, "learning_rate": 1.9282275727345455e-05, "loss": 1.1018, "step": 4806 }, { "epoch": 1.48, "learning_rate": 1.9281904955516987e-05, "loss": 1.0137, "step": 4807 }, { "epoch": 1.48, "learning_rate": 1.9281534091510646e-05, "loss": 1.2524, "step": 4808 }, { "epoch": 1.48, "learning_rate": 1.9281163135330115e-05, "loss": 0.9335, "step": 4809 }, { "epoch": 1.48, "learning_rate": 1.9280792086979084e-05, "loss": 1.1639, "step": 4810 }, { "epoch": 1.48, "learning_rate": 1.9280420946461233e-05, "loss": 1.2208, "step": 4811 }, { "epoch": 1.48, "learning_rate": 1.9280049713780247e-05, "loss": 1.1385, "step": 4812 }, { "epoch": 1.48, "learning_rate": 1.9279678388939817e-05, "loss": 1.6927, "step": 4813 }, { "epoch": 1.48, "learning_rate": 1.927930697194363e-05, "loss": 1.1381, "step": 4814 }, { "epoch": 1.48, "learning_rate": 1.9278935462795365e-05, "loss": 1.2657, "step": 4815 }, { "epoch": 1.48, "learning_rate": 1.9278563861498726e-05, "loss": 1.0153, "step": 4816 }, { "epoch": 1.48, "learning_rate": 1.9278192168057394e-05, "loss": 1.0354, "step": 4817 }, { "epoch": 1.48, "learning_rate": 1.9277820382475063e-05, "loss": 1.1719, "step": 4818 }, { "epoch": 1.48, "learning_rate": 1.9277448504755425e-05, "loss": 1.1201, "step": 4819 }, { "epoch": 1.48, "learning_rate": 1.9277076534902173e-05, "loss": 0.9686, "step": 4820 }, { "epoch": 1.48, "learning_rate": 1.9276704472919e-05, "loss": 1.1137, "step": 4821 }, { "epoch": 1.48, "learning_rate": 1.9276332318809606e-05, "loss": 1.1415, "step": 4822 }, { "epoch": 1.48, "learning_rate": 1.9275960072577676e-05, "loss": 1.0175, "step": 4823 }, { "epoch": 1.48, "learning_rate": 1.927558773422692e-05, "loss": 1.3438, "step": 4824 }, { "epoch": 1.48, "learning_rate": 1.9275215303761027e-05, "loss": 1.4589, "step": 4825 }, { "epoch": 1.48, "learning_rate": 1.9274842781183697e-05, "loss": 1.1379, "step": 4826 }, { "epoch": 1.49, "learning_rate": 1.9274470166498628e-05, "loss": 1.1946, "step": 4827 }, { "epoch": 1.49, "learning_rate": 1.9274097459709528e-05, "loss": 1.1729, "step": 4828 }, { "epoch": 1.49, "learning_rate": 1.9273724660820086e-05, "loss": 1.2224, "step": 4829 }, { "epoch": 1.49, "learning_rate": 1.927335176983402e-05, "loss": 0.971, "step": 4830 }, { "epoch": 1.49, "learning_rate": 1.9272978786755016e-05, "loss": 1.1971, "step": 4831 }, { "epoch": 1.49, "learning_rate": 1.927260571158679e-05, "loss": 1.4235, "step": 4832 }, { "epoch": 1.49, "learning_rate": 1.9272232544333044e-05, "loss": 1.2063, "step": 4833 }, { "epoch": 1.49, "learning_rate": 1.9271859284997485e-05, "loss": 0.9411, "step": 4834 }, { "epoch": 1.49, "learning_rate": 1.9271485933583816e-05, "loss": 1.1118, "step": 4835 }, { "epoch": 1.49, "learning_rate": 1.9271112490095745e-05, "loss": 1.1111, "step": 4836 }, { "epoch": 1.49, "learning_rate": 1.9270738954536985e-05, "loss": 1.2292, "step": 4837 }, { "epoch": 1.49, "learning_rate": 1.9270365326911243e-05, "loss": 1.0521, "step": 4838 }, { "epoch": 1.49, "learning_rate": 1.9269991607222228e-05, "loss": 1.1243, "step": 4839 }, { "epoch": 1.49, "learning_rate": 1.9269617795473654e-05, "loss": 1.0319, "step": 4840 }, { "epoch": 1.49, "learning_rate": 1.926924389166923e-05, "loss": 1.1158, "step": 4841 }, { "epoch": 1.49, "learning_rate": 1.9268869895812673e-05, "loss": 1.1298, "step": 4842 }, { "epoch": 1.49, "learning_rate": 1.9268495807907696e-05, "loss": 1.1934, "step": 4843 }, { "epoch": 1.49, "learning_rate": 1.9268121627958013e-05, "loss": 1.3336, "step": 4844 }, { "epoch": 1.49, "learning_rate": 1.926774735596734e-05, "loss": 1.3164, "step": 4845 }, { "epoch": 1.49, "learning_rate": 1.9267372991939398e-05, "loss": 1.1156, "step": 4846 }, { "epoch": 1.49, "learning_rate": 1.9266998535877896e-05, "loss": 1.2629, "step": 4847 }, { "epoch": 1.49, "learning_rate": 1.926662398778656e-05, "loss": 1.1671, "step": 4848 }, { "epoch": 1.49, "learning_rate": 1.9266249347669108e-05, "loss": 1.3264, "step": 4849 }, { "epoch": 1.49, "learning_rate": 1.9265874615529258e-05, "loss": 1.2929, "step": 4850 }, { "epoch": 1.49, "learning_rate": 1.9265499791370734e-05, "loss": 1.3219, "step": 4851 }, { "epoch": 1.49, "learning_rate": 1.926512487519726e-05, "loss": 1.2468, "step": 4852 }, { "epoch": 1.49, "learning_rate": 1.9264749867012557e-05, "loss": 1.3837, "step": 4853 }, { "epoch": 1.49, "learning_rate": 1.9264374766820344e-05, "loss": 0.9829, "step": 4854 }, { "epoch": 1.49, "learning_rate": 1.9263999574624357e-05, "loss": 1.0302, "step": 4855 }, { "epoch": 1.49, "learning_rate": 1.9263624290428314e-05, "loss": 1.7596, "step": 4856 }, { "epoch": 1.49, "learning_rate": 1.9263248914235945e-05, "loss": 1.1692, "step": 4857 }, { "epoch": 1.49, "learning_rate": 1.9262873446050973e-05, "loss": 1.0277, "step": 4858 }, { "epoch": 1.5, "learning_rate": 1.9262497885877135e-05, "loss": 1.1473, "step": 4859 }, { "epoch": 1.5, "learning_rate": 1.9262122233718155e-05, "loss": 1.2664, "step": 4860 }, { "epoch": 1.5, "learning_rate": 1.9261746489577767e-05, "loss": 1.0278, "step": 4861 }, { "epoch": 1.5, "learning_rate": 1.9261370653459698e-05, "loss": 1.4828, "step": 4862 }, { "epoch": 1.5, "learning_rate": 1.9260994725367684e-05, "loss": 0.9272, "step": 4863 }, { "epoch": 1.5, "learning_rate": 1.926061870530546e-05, "loss": 1.2624, "step": 4864 }, { "epoch": 1.5, "learning_rate": 1.926024259327675e-05, "loss": 1.5093, "step": 4865 }, { "epoch": 1.5, "learning_rate": 1.9259866389285303e-05, "loss": 1.3375, "step": 4866 }, { "epoch": 1.5, "learning_rate": 1.9259490093334845e-05, "loss": 1.8682, "step": 4867 }, { "epoch": 1.5, "learning_rate": 1.925911370542912e-05, "loss": 1.3162, "step": 4868 }, { "epoch": 1.5, "learning_rate": 1.925873722557186e-05, "loss": 1.3411, "step": 4869 }, { "epoch": 1.5, "learning_rate": 1.9258360653766804e-05, "loss": 1.1972, "step": 4870 }, { "epoch": 1.5, "learning_rate": 1.92579839900177e-05, "loss": 1.2392, "step": 4871 }, { "epoch": 1.5, "learning_rate": 1.925760723432828e-05, "loss": 1.4747, "step": 4872 }, { "epoch": 1.5, "learning_rate": 1.9257230386702285e-05, "loss": 1.0592, "step": 4873 }, { "epoch": 1.5, "learning_rate": 1.9256853447143465e-05, "loss": 1.3046, "step": 4874 }, { "epoch": 1.5, "learning_rate": 1.9256476415655556e-05, "loss": 1.029, "step": 4875 }, { "epoch": 1.5, "learning_rate": 1.925609929224231e-05, "loss": 1.0936, "step": 4876 }, { "epoch": 1.5, "learning_rate": 1.925572207690746e-05, "loss": 1.6637, "step": 4877 }, { "epoch": 1.5, "learning_rate": 1.9255344769654765e-05, "loss": 1.3267, "step": 4878 }, { "epoch": 1.5, "learning_rate": 1.9254967370487963e-05, "loss": 0.9412, "step": 4879 }, { "epoch": 1.5, "learning_rate": 1.9254589879410806e-05, "loss": 1.1468, "step": 4880 }, { "epoch": 1.5, "learning_rate": 1.9254212296427043e-05, "loss": 1.0595, "step": 4881 }, { "epoch": 1.5, "learning_rate": 1.9253834621540425e-05, "loss": 1.4156, "step": 4882 }, { "epoch": 1.5, "learning_rate": 1.9253456854754697e-05, "loss": 1.6613, "step": 4883 }, { "epoch": 1.5, "learning_rate": 1.925307899607362e-05, "loss": 1.1473, "step": 4884 }, { "epoch": 1.5, "learning_rate": 1.925270104550093e-05, "loss": 0.9508, "step": 4885 }, { "epoch": 1.5, "learning_rate": 1.92523230030404e-05, "loss": 1.1887, "step": 4886 }, { "epoch": 1.5, "learning_rate": 1.9251944868695772e-05, "loss": 1.2997, "step": 4887 }, { "epoch": 1.5, "learning_rate": 1.9251566642470805e-05, "loss": 1.2304, "step": 4888 }, { "epoch": 1.5, "learning_rate": 1.9251188324369258e-05, "loss": 1.1577, "step": 4889 }, { "epoch": 1.5, "learning_rate": 1.925080991439488e-05, "loss": 1.1746, "step": 4890 }, { "epoch": 1.5, "learning_rate": 1.9250431412551435e-05, "loss": 1.2397, "step": 4891 }, { "epoch": 1.51, "learning_rate": 1.925005281884268e-05, "loss": 1.2333, "step": 4892 }, { "epoch": 1.51, "learning_rate": 1.924967413327238e-05, "loss": 1.2395, "step": 4893 }, { "epoch": 1.51, "learning_rate": 1.9249295355844286e-05, "loss": 1.3478, "step": 4894 }, { "epoch": 1.51, "learning_rate": 1.9248916486562168e-05, "loss": 1.2121, "step": 4895 }, { "epoch": 1.51, "learning_rate": 1.9248537525429785e-05, "loss": 1.6817, "step": 4896 }, { "epoch": 1.51, "learning_rate": 1.92481584724509e-05, "loss": 1.4099, "step": 4897 }, { "epoch": 1.51, "learning_rate": 1.9247779327629276e-05, "loss": 1.1452, "step": 4898 }, { "epoch": 1.51, "learning_rate": 1.9247400090968684e-05, "loss": 1.0224, "step": 4899 }, { "epoch": 1.51, "learning_rate": 1.924702076247288e-05, "loss": 1.6646, "step": 4900 }, { "epoch": 1.51, "learning_rate": 1.9246641342145646e-05, "loss": 1.2968, "step": 4901 }, { "epoch": 1.51, "learning_rate": 1.9246261829990735e-05, "loss": 1.2305, "step": 4902 }, { "epoch": 1.51, "learning_rate": 1.9245882226011927e-05, "loss": 1.1574, "step": 4903 }, { "epoch": 1.51, "learning_rate": 1.924550253021298e-05, "loss": 1.3046, "step": 4904 }, { "epoch": 1.51, "learning_rate": 1.9245122742597678e-05, "loss": 1.4838, "step": 4905 }, { "epoch": 1.51, "learning_rate": 1.9244742863169786e-05, "loss": 1.2258, "step": 4906 }, { "epoch": 1.51, "learning_rate": 1.9244362891933077e-05, "loss": 1.3101, "step": 4907 }, { "epoch": 1.51, "learning_rate": 1.9243982828891326e-05, "loss": 0.9749, "step": 4908 }, { "epoch": 1.51, "learning_rate": 1.9243602674048306e-05, "loss": 1.3497, "step": 4909 }, { "epoch": 1.51, "learning_rate": 1.924322242740779e-05, "loss": 1.3445, "step": 4910 }, { "epoch": 1.51, "learning_rate": 1.9242842088973557e-05, "loss": 1.3174, "step": 4911 }, { "epoch": 1.51, "learning_rate": 1.9242461658749382e-05, "loss": 1.4442, "step": 4912 }, { "epoch": 1.51, "learning_rate": 1.9242081136739047e-05, "loss": 1.0752, "step": 4913 }, { "epoch": 1.51, "learning_rate": 1.924170052294633e-05, "loss": 1.8062, "step": 4914 }, { "epoch": 1.51, "learning_rate": 1.9241319817375007e-05, "loss": 1.0049, "step": 4915 }, { "epoch": 1.51, "learning_rate": 1.924093902002886e-05, "loss": 1.1412, "step": 4916 }, { "epoch": 1.51, "learning_rate": 1.924055813091167e-05, "loss": 1.0409, "step": 4917 }, { "epoch": 1.51, "learning_rate": 1.9240177150027227e-05, "loss": 1.1419, "step": 4918 }, { "epoch": 1.51, "learning_rate": 1.923979607737931e-05, "loss": 1.2464, "step": 4919 }, { "epoch": 1.51, "learning_rate": 1.9239414912971697e-05, "loss": 1.2362, "step": 4920 }, { "epoch": 1.51, "learning_rate": 1.923903365680818e-05, "loss": 1.0201, "step": 4921 }, { "epoch": 1.51, "learning_rate": 1.9238652308892544e-05, "loss": 1.0576, "step": 4922 }, { "epoch": 1.51, "learning_rate": 1.9238270869228578e-05, "loss": 1.0674, "step": 4923 }, { "epoch": 1.52, "learning_rate": 1.9237889337820067e-05, "loss": 1.3806, "step": 4924 }, { "epoch": 1.52, "learning_rate": 1.9237507714670797e-05, "loss": 1.1362, "step": 4925 }, { "epoch": 1.52, "learning_rate": 1.9237125999784566e-05, "loss": 1.5659, "step": 4926 }, { "epoch": 1.52, "learning_rate": 1.923674419316516e-05, "loss": 1.1574, "step": 4927 }, { "epoch": 1.52, "learning_rate": 1.9236362294816374e-05, "loss": 1.19, "step": 4928 }, { "epoch": 1.52, "learning_rate": 1.9235980304741996e-05, "loss": 1.2677, "step": 4929 }, { "epoch": 1.52, "learning_rate": 1.9235598222945818e-05, "loss": 1.1302, "step": 4930 }, { "epoch": 1.52, "learning_rate": 1.9235216049431642e-05, "loss": 1.4836, "step": 4931 }, { "epoch": 1.52, "learning_rate": 1.9234833784203256e-05, "loss": 1.1542, "step": 4932 }, { "epoch": 1.52, "learning_rate": 1.923445142726446e-05, "loss": 1.226, "step": 4933 }, { "epoch": 1.52, "learning_rate": 1.9234068978619053e-05, "loss": 1.3479, "step": 4934 }, { "epoch": 1.52, "learning_rate": 1.923368643827083e-05, "loss": 1.3283, "step": 4935 }, { "epoch": 1.52, "learning_rate": 1.923330380622359e-05, "loss": 1.5, "step": 4936 }, { "epoch": 1.52, "learning_rate": 1.9232921082481132e-05, "loss": 1.6591, "step": 4937 }, { "epoch": 1.52, "learning_rate": 1.9232538267047262e-05, "loss": 1.3956, "step": 4938 }, { "epoch": 1.52, "learning_rate": 1.9232155359925775e-05, "loss": 1.1288, "step": 4939 }, { "epoch": 1.52, "learning_rate": 1.9231772361120477e-05, "loss": 1.3345, "step": 4940 }, { "epoch": 1.52, "learning_rate": 1.9231389270635172e-05, "loss": 1.3053, "step": 4941 }, { "epoch": 1.52, "learning_rate": 1.923100608847366e-05, "loss": 1.3149, "step": 4942 }, { "epoch": 1.52, "learning_rate": 1.9230622814639756e-05, "loss": 1.2805, "step": 4943 }, { "epoch": 1.52, "learning_rate": 1.923023944913726e-05, "loss": 2.026, "step": 4944 }, { "epoch": 1.52, "learning_rate": 1.9229855991969975e-05, "loss": 1.3077, "step": 4945 }, { "epoch": 1.52, "learning_rate": 1.922947244314172e-05, "loss": 1.6123, "step": 4946 }, { "epoch": 1.52, "learning_rate": 1.922908880265629e-05, "loss": 1.0858, "step": 4947 }, { "epoch": 1.52, "learning_rate": 1.9228705070517507e-05, "loss": 1.0241, "step": 4948 }, { "epoch": 1.52, "learning_rate": 1.9228321246729175e-05, "loss": 1.1342, "step": 4949 }, { "epoch": 1.52, "learning_rate": 1.922793733129511e-05, "loss": 0.9863, "step": 4950 }, { "epoch": 1.52, "learning_rate": 1.922755332421912e-05, "loss": 1.1925, "step": 4951 }, { "epoch": 1.52, "learning_rate": 1.9227169225505022e-05, "loss": 1.1546, "step": 4952 }, { "epoch": 1.52, "learning_rate": 1.922678503515663e-05, "loss": 1.2752, "step": 4953 }, { "epoch": 1.52, "learning_rate": 1.922640075317776e-05, "loss": 0.7421, "step": 4954 }, { "epoch": 1.52, "learning_rate": 1.9226016379572227e-05, "loss": 1.3552, "step": 4955 }, { "epoch": 1.52, "learning_rate": 1.9225631914343846e-05, "loss": 1.1424, "step": 4956 }, { "epoch": 1.53, "learning_rate": 1.922524735749644e-05, "loss": 1.0307, "step": 4957 }, { "epoch": 1.53, "learning_rate": 1.9224862709033823e-05, "loss": 1.2386, "step": 4958 }, { "epoch": 1.53, "learning_rate": 1.922447796895982e-05, "loss": 1.2918, "step": 4959 }, { "epoch": 1.53, "learning_rate": 1.922409313727825e-05, "loss": 1.2353, "step": 4960 }, { "epoch": 1.53, "learning_rate": 1.9223708213992932e-05, "loss": 1.2993, "step": 4961 }, { "epoch": 1.53, "learning_rate": 1.922332319910769e-05, "loss": 1.3155, "step": 4962 }, { "epoch": 1.53, "learning_rate": 1.9222938092626352e-05, "loss": 0.9728, "step": 4963 }, { "epoch": 1.53, "learning_rate": 1.9222552894552732e-05, "loss": 1.4544, "step": 4964 }, { "epoch": 1.53, "learning_rate": 1.9222167604890666e-05, "loss": 1.0738, "step": 4965 }, { "epoch": 1.53, "learning_rate": 1.9221782223643976e-05, "loss": 1.2794, "step": 4966 }, { "epoch": 1.53, "learning_rate": 1.922139675081649e-05, "loss": 1.3051, "step": 4967 }, { "epoch": 1.53, "learning_rate": 1.9221011186412034e-05, "loss": 1.2253, "step": 4968 }, { "epoch": 1.53, "learning_rate": 1.922062553043444e-05, "loss": 1.3955, "step": 4969 }, { "epoch": 1.53, "learning_rate": 1.9220239782887535e-05, "loss": 1.3649, "step": 4970 }, { "epoch": 1.53, "learning_rate": 1.9219853943775155e-05, "loss": 1.2498, "step": 4971 }, { "epoch": 1.53, "learning_rate": 1.9219468013101123e-05, "loss": 1.1461, "step": 4972 }, { "epoch": 1.53, "learning_rate": 1.921908199086928e-05, "loss": 1.3588, "step": 4973 }, { "epoch": 1.53, "learning_rate": 1.9218695877083456e-05, "loss": 1.2372, "step": 4974 }, { "epoch": 1.53, "learning_rate": 1.9218309671747485e-05, "loss": 1.1128, "step": 4975 }, { "epoch": 1.53, "learning_rate": 1.9217923374865202e-05, "loss": 1.4043, "step": 4976 }, { "epoch": 1.53, "learning_rate": 1.9217536986440447e-05, "loss": 1.4197, "step": 4977 }, { "epoch": 1.53, "learning_rate": 1.9217150506477055e-05, "loss": 1.2696, "step": 4978 }, { "epoch": 1.53, "learning_rate": 1.921676393497886e-05, "loss": 1.2311, "step": 4979 }, { "epoch": 1.53, "learning_rate": 1.9216377271949707e-05, "loss": 1.5741, "step": 4980 }, { "epoch": 1.53, "learning_rate": 1.9215990517393436e-05, "loss": 1.2027, "step": 4981 }, { "epoch": 1.53, "learning_rate": 1.921560367131388e-05, "loss": 1.2696, "step": 4982 }, { "epoch": 1.53, "learning_rate": 1.9215216733714893e-05, "loss": 0.8851, "step": 4983 }, { "epoch": 1.53, "learning_rate": 1.9214829704600305e-05, "loss": 1.1411, "step": 4984 }, { "epoch": 1.53, "learning_rate": 1.9214442583973965e-05, "loss": 1.1878, "step": 4985 }, { "epoch": 1.53, "learning_rate": 1.9214055371839722e-05, "loss": 1.6377, "step": 4986 }, { "epoch": 1.53, "learning_rate": 1.9213668068201414e-05, "loss": 1.5263, "step": 4987 }, { "epoch": 1.53, "learning_rate": 1.9213280673062893e-05, "loss": 1.3854, "step": 4988 }, { "epoch": 1.54, "learning_rate": 1.9212893186428002e-05, "loss": 1.1567, "step": 4989 }, { "epoch": 1.54, "learning_rate": 1.921250560830059e-05, "loss": 0.8637, "step": 4990 }, { "epoch": 1.54, "learning_rate": 1.9212117938684513e-05, "loss": 0.9837, "step": 4991 }, { "epoch": 1.54, "learning_rate": 1.9211730177583608e-05, "loss": 1.0208, "step": 4992 }, { "epoch": 1.54, "learning_rate": 1.9211342325001732e-05, "loss": 1.2431, "step": 4993 }, { "epoch": 1.54, "learning_rate": 1.921095438094274e-05, "loss": 1.1782, "step": 4994 }, { "epoch": 1.54, "learning_rate": 1.9210566345410484e-05, "loss": 1.2828, "step": 4995 }, { "epoch": 1.54, "learning_rate": 1.9210178218408812e-05, "loss": 1.1324, "step": 4996 }, { "epoch": 1.54, "learning_rate": 1.9209789999941585e-05, "loss": 1.2796, "step": 4997 }, { "epoch": 1.54, "learning_rate": 1.920940169001265e-05, "loss": 1.175, "step": 4998 }, { "epoch": 1.54, "learning_rate": 1.9209013288625874e-05, "loss": 1.1391, "step": 4999 }, { "epoch": 1.54, "learning_rate": 1.9208624795785106e-05, "loss": 1.0778, "step": 5000 }, { "epoch": 1.54, "learning_rate": 1.920823621149421e-05, "loss": 1.5847, "step": 5001 }, { "epoch": 1.54, "learning_rate": 1.9207847535757038e-05, "loss": 1.6414, "step": 5002 }, { "epoch": 1.54, "learning_rate": 1.9207458768577453e-05, "loss": 1.225, "step": 5003 }, { "epoch": 1.54, "learning_rate": 1.9207069909959323e-05, "loss": 1.0622, "step": 5004 }, { "epoch": 1.54, "learning_rate": 1.92066809599065e-05, "loss": 1.1711, "step": 5005 }, { "epoch": 1.54, "learning_rate": 1.9206291918422847e-05, "loss": 1.0463, "step": 5006 }, { "epoch": 1.54, "learning_rate": 1.920590278551223e-05, "loss": 1.2001, "step": 5007 }, { "epoch": 1.54, "learning_rate": 1.9205513561178518e-05, "loss": 1.2801, "step": 5008 }, { "epoch": 1.54, "learning_rate": 1.9205124245425573e-05, "loss": 1.2092, "step": 5009 }, { "epoch": 1.54, "learning_rate": 1.9204734838257258e-05, "loss": 1.1014, "step": 5010 }, { "epoch": 1.54, "learning_rate": 1.9204345339677442e-05, "loss": 1.4138, "step": 5011 }, { "epoch": 1.54, "learning_rate": 1.9203955749689994e-05, "loss": 1.0551, "step": 5012 }, { "epoch": 1.54, "learning_rate": 1.9203566068298784e-05, "loss": 1.0981, "step": 5013 }, { "epoch": 1.54, "learning_rate": 1.9203176295507678e-05, "loss": 1.7858, "step": 5014 }, { "epoch": 1.54, "learning_rate": 1.9202786431320554e-05, "loss": 1.3094, "step": 5015 }, { "epoch": 1.54, "learning_rate": 1.920239647574128e-05, "loss": 1.3448, "step": 5016 }, { "epoch": 1.54, "learning_rate": 1.9202006428773722e-05, "loss": 1.3345, "step": 5017 }, { "epoch": 1.54, "learning_rate": 1.9201616290421763e-05, "loss": 1.2669, "step": 5018 }, { "epoch": 1.54, "learning_rate": 1.9201226060689273e-05, "loss": 1.1788, "step": 5019 }, { "epoch": 1.54, "learning_rate": 1.920083573958013e-05, "loss": 1.0563, "step": 5020 }, { "epoch": 1.54, "learning_rate": 1.9200445327098205e-05, "loss": 1.2118, "step": 5021 }, { "epoch": 1.55, "learning_rate": 1.9200054823247378e-05, "loss": 1.2287, "step": 5022 }, { "epoch": 1.55, "learning_rate": 1.9199664228031532e-05, "loss": 1.1065, "step": 5023 }, { "epoch": 1.55, "learning_rate": 1.919927354145454e-05, "loss": 1.2385, "step": 5024 }, { "epoch": 1.55, "learning_rate": 1.919888276352028e-05, "loss": 0.94, "step": 5025 }, { "epoch": 1.55, "learning_rate": 1.919849189423264e-05, "loss": 1.2623, "step": 5026 }, { "epoch": 1.55, "learning_rate": 1.9198100933595495e-05, "loss": 1.1202, "step": 5027 }, { "epoch": 1.55, "learning_rate": 1.919770988161273e-05, "loss": 1.5357, "step": 5028 }, { "epoch": 1.55, "learning_rate": 1.9197318738288234e-05, "loss": 1.1555, "step": 5029 }, { "epoch": 1.55, "learning_rate": 1.9196927503625885e-05, "loss": 1.1489, "step": 5030 }, { "epoch": 1.55, "learning_rate": 1.919653617762957e-05, "loss": 1.3459, "step": 5031 }, { "epoch": 1.55, "learning_rate": 1.919614476030317e-05, "loss": 1.2014, "step": 5032 }, { "epoch": 1.55, "learning_rate": 1.9195753251650583e-05, "loss": 0.978, "step": 5033 }, { "epoch": 1.55, "learning_rate": 1.9195361651675687e-05, "loss": 1.2328, "step": 5034 }, { "epoch": 1.55, "learning_rate": 1.9194969960382378e-05, "loss": 1.0971, "step": 5035 }, { "epoch": 1.55, "learning_rate": 1.919457817777454e-05, "loss": 1.1437, "step": 5036 }, { "epoch": 1.55, "learning_rate": 1.919418630385607e-05, "loss": 1.0184, "step": 5037 }, { "epoch": 1.55, "learning_rate": 1.9193794338630853e-05, "loss": 1.2836, "step": 5038 }, { "epoch": 1.55, "learning_rate": 1.919340228210279e-05, "loss": 1.2924, "step": 5039 }, { "epoch": 1.55, "learning_rate": 1.9193010134275764e-05, "loss": 1.4953, "step": 5040 }, { "epoch": 1.55, "learning_rate": 1.9192617895153678e-05, "loss": 1.6392, "step": 5041 }, { "epoch": 1.55, "learning_rate": 1.919222556474042e-05, "loss": 1.125, "step": 5042 }, { "epoch": 1.55, "learning_rate": 1.9191833143039896e-05, "loss": 1.1849, "step": 5043 }, { "epoch": 1.55, "learning_rate": 1.9191440630055995e-05, "loss": 1.0249, "step": 5044 }, { "epoch": 1.55, "learning_rate": 1.919104802579262e-05, "loss": 1.8414, "step": 5045 }, { "epoch": 1.55, "learning_rate": 1.9190655330253663e-05, "loss": 1.6052, "step": 5046 }, { "epoch": 1.55, "learning_rate": 1.9190262543443028e-05, "loss": 1.1932, "step": 5047 }, { "epoch": 1.55, "learning_rate": 1.918986966536462e-05, "loss": 1.2808, "step": 5048 }, { "epoch": 1.55, "learning_rate": 1.9189476696022335e-05, "loss": 1.08, "step": 5049 }, { "epoch": 1.55, "learning_rate": 1.9189083635420077e-05, "loss": 1.3755, "step": 5050 }, { "epoch": 1.55, "learning_rate": 1.9188690483561746e-05, "loss": 1.2142, "step": 5051 }, { "epoch": 1.55, "learning_rate": 1.9188297240451253e-05, "loss": 1.1846, "step": 5052 }, { "epoch": 1.55, "learning_rate": 1.9187903906092504e-05, "loss": 1.3151, "step": 5053 }, { "epoch": 1.56, "learning_rate": 1.9187510480489397e-05, "loss": 1.1922, "step": 5054 }, { "epoch": 1.56, "learning_rate": 1.9187116963645845e-05, "loss": 1.3968, "step": 5055 }, { "epoch": 1.56, "learning_rate": 1.9186723355565753e-05, "loss": 1.1891, "step": 5056 }, { "epoch": 1.56, "learning_rate": 1.9186329656253032e-05, "loss": 1.1046, "step": 5057 }, { "epoch": 1.56, "learning_rate": 1.918593586571159e-05, "loss": 1.5759, "step": 5058 }, { "epoch": 1.56, "learning_rate": 1.918554198394534e-05, "loss": 1.3057, "step": 5059 }, { "epoch": 1.56, "learning_rate": 1.9185148010958196e-05, "loss": 1.2914, "step": 5060 }, { "epoch": 1.56, "learning_rate": 1.9184753946754063e-05, "loss": 1.3649, "step": 5061 }, { "epoch": 1.56, "learning_rate": 1.9184359791336858e-05, "loss": 1.1655, "step": 5062 }, { "epoch": 1.56, "learning_rate": 1.9183965544710495e-05, "loss": 1.1028, "step": 5063 }, { "epoch": 1.56, "learning_rate": 1.9183571206878894e-05, "loss": 1.0975, "step": 5064 }, { "epoch": 1.56, "learning_rate": 1.9183176777845965e-05, "loss": 1.427, "step": 5065 }, { "epoch": 1.56, "learning_rate": 1.9182782257615624e-05, "loss": 1.1538, "step": 5066 }, { "epoch": 1.56, "learning_rate": 1.91823876461918e-05, "loss": 1.1343, "step": 5067 }, { "epoch": 1.56, "learning_rate": 1.9181992943578395e-05, "loss": 1.3897, "step": 5068 }, { "epoch": 1.56, "learning_rate": 1.9181598149779343e-05, "loss": 1.5157, "step": 5069 }, { "epoch": 1.56, "learning_rate": 1.918120326479856e-05, "loss": 1.5539, "step": 5070 }, { "epoch": 1.56, "learning_rate": 1.918080828863996e-05, "loss": 1.356, "step": 5071 }, { "epoch": 1.56, "learning_rate": 1.918041322130748e-05, "loss": 1.195, "step": 5072 }, { "epoch": 1.56, "learning_rate": 1.9180018062805032e-05, "loss": 1.2177, "step": 5073 }, { "epoch": 1.56, "learning_rate": 1.9179622813136544e-05, "loss": 1.193, "step": 5074 }, { "epoch": 1.56, "learning_rate": 1.9179227472305942e-05, "loss": 1.204, "step": 5075 }, { "epoch": 1.56, "learning_rate": 1.9178832040317153e-05, "loss": 1.2337, "step": 5076 }, { "epoch": 1.56, "learning_rate": 1.91784365171741e-05, "loss": 1.2511, "step": 5077 }, { "epoch": 1.56, "learning_rate": 1.9178040902880717e-05, "loss": 1.6362, "step": 5078 }, { "epoch": 1.56, "learning_rate": 1.9177645197440928e-05, "loss": 1.2975, "step": 5079 }, { "epoch": 1.56, "learning_rate": 1.9177249400858658e-05, "loss": 0.9116, "step": 5080 }, { "epoch": 1.56, "learning_rate": 1.917685351313785e-05, "loss": 1.2627, "step": 5081 }, { "epoch": 1.56, "learning_rate": 1.9176457534282427e-05, "loss": 0.9835, "step": 5082 }, { "epoch": 1.56, "learning_rate": 1.9176061464296324e-05, "loss": 1.257, "step": 5083 }, { "epoch": 1.56, "learning_rate": 1.9175665303183474e-05, "loss": 1.1436, "step": 5084 }, { "epoch": 1.56, "learning_rate": 1.9175269050947806e-05, "loss": 1.1249, "step": 5085 }, { "epoch": 1.56, "learning_rate": 1.9174872707593267e-05, "loss": 1.2918, "step": 5086 }, { "epoch": 1.57, "learning_rate": 1.9174476273123783e-05, "loss": 1.2964, "step": 5087 }, { "epoch": 1.57, "learning_rate": 1.9174079747543292e-05, "loss": 1.331, "step": 5088 }, { "epoch": 1.57, "learning_rate": 1.9173683130855737e-05, "loss": 1.883, "step": 5089 }, { "epoch": 1.57, "learning_rate": 1.9173286423065052e-05, "loss": 1.237, "step": 5090 }, { "epoch": 1.57, "learning_rate": 1.9172889624175185e-05, "loss": 1.5502, "step": 5091 }, { "epoch": 1.57, "learning_rate": 1.9172492734190064e-05, "loss": 0.9663, "step": 5092 }, { "epoch": 1.57, "learning_rate": 1.9172095753113638e-05, "loss": 1.1627, "step": 5093 }, { "epoch": 1.57, "learning_rate": 1.9171698680949843e-05, "loss": 1.3031, "step": 5094 }, { "epoch": 1.57, "learning_rate": 1.9171301517702634e-05, "loss": 1.0857, "step": 5095 }, { "epoch": 1.57, "learning_rate": 1.9170904263375948e-05, "loss": 0.9406, "step": 5096 }, { "epoch": 1.57, "learning_rate": 1.9170506917973727e-05, "loss": 1.1852, "step": 5097 }, { "epoch": 1.57, "learning_rate": 1.917010948149992e-05, "loss": 1.1981, "step": 5098 }, { "epoch": 1.57, "learning_rate": 1.9169711953958477e-05, "loss": 1.1851, "step": 5099 }, { "epoch": 1.57, "learning_rate": 1.9169314335353345e-05, "loss": 1.0946, "step": 5100 }, { "epoch": 1.57, "learning_rate": 1.916891662568847e-05, "loss": 1.6615, "step": 5101 }, { "epoch": 1.57, "learning_rate": 1.9168518824967797e-05, "loss": 1.1594, "step": 5102 }, { "epoch": 1.57, "learning_rate": 1.9168120933195285e-05, "loss": 1.1189, "step": 5103 }, { "epoch": 1.57, "learning_rate": 1.9167722950374887e-05, "loss": 1.5223, "step": 5104 }, { "epoch": 1.57, "learning_rate": 1.9167324876510545e-05, "loss": 1.1307, "step": 5105 }, { "epoch": 1.57, "learning_rate": 1.9166926711606222e-05, "loss": 1.0134, "step": 5106 }, { "epoch": 1.57, "learning_rate": 1.9166528455665867e-05, "loss": 1.0597, "step": 5107 }, { "epoch": 1.57, "learning_rate": 1.9166130108693438e-05, "loss": 0.9791, "step": 5108 }, { "epoch": 1.57, "learning_rate": 1.9165731670692885e-05, "loss": 1.2681, "step": 5109 }, { "epoch": 1.57, "learning_rate": 1.9165333141668173e-05, "loss": 1.1875, "step": 5110 }, { "epoch": 1.57, "learning_rate": 1.9164934521623255e-05, "loss": 1.432, "step": 5111 }, { "epoch": 1.57, "learning_rate": 1.9164535810562092e-05, "loss": 1.1682, "step": 5112 }, { "epoch": 1.57, "learning_rate": 1.916413700848864e-05, "loss": 0.9977, "step": 5113 }, { "epoch": 1.57, "learning_rate": 1.916373811540686e-05, "loss": 1.1091, "step": 5114 }, { "epoch": 1.57, "learning_rate": 1.916333913132072e-05, "loss": 1.2347, "step": 5115 }, { "epoch": 1.57, "learning_rate": 1.916294005623417e-05, "loss": 1.4328, "step": 5116 }, { "epoch": 1.57, "learning_rate": 1.9162540890151184e-05, "loss": 0.7994, "step": 5117 }, { "epoch": 1.57, "learning_rate": 1.9162141633075724e-05, "loss": 1.0388, "step": 5118 }, { "epoch": 1.58, "learning_rate": 1.916174228501175e-05, "loss": 1.2512, "step": 5119 }, { "epoch": 1.58, "learning_rate": 1.9161342845963237e-05, "loss": 0.997, "step": 5120 }, { "epoch": 1.58, "learning_rate": 1.9160943315934144e-05, "loss": 1.29, "step": 5121 }, { "epoch": 1.58, "learning_rate": 1.916054369492844e-05, "loss": 0.8347, "step": 5122 }, { "epoch": 1.58, "learning_rate": 1.9160143982950094e-05, "loss": 1.2257, "step": 5123 }, { "epoch": 1.58, "learning_rate": 1.9159744180003076e-05, "loss": 1.3781, "step": 5124 }, { "epoch": 1.58, "learning_rate": 1.915934428609136e-05, "loss": 1.3271, "step": 5125 }, { "epoch": 1.58, "learning_rate": 1.915894430121891e-05, "loss": 1.1212, "step": 5126 }, { "epoch": 1.58, "learning_rate": 1.91585442253897e-05, "loss": 1.0654, "step": 5127 }, { "epoch": 1.58, "learning_rate": 1.915814405860771e-05, "loss": 1.164, "step": 5128 }, { "epoch": 1.58, "learning_rate": 1.9157743800876908e-05, "loss": 1.0703, "step": 5129 }, { "epoch": 1.58, "learning_rate": 1.9157343452201266e-05, "loss": 1.3321, "step": 5130 }, { "epoch": 1.58, "learning_rate": 1.915694301258477e-05, "loss": 1.3613, "step": 5131 }, { "epoch": 1.58, "learning_rate": 1.9156542482031387e-05, "loss": 1.1329, "step": 5132 }, { "epoch": 1.58, "learning_rate": 1.9156141860545098e-05, "loss": 1.0202, "step": 5133 }, { "epoch": 1.58, "learning_rate": 1.915574114812988e-05, "loss": 1.2801, "step": 5134 }, { "epoch": 1.58, "learning_rate": 1.9155340344789717e-05, "loss": 1.372, "step": 5135 }, { "epoch": 1.58, "learning_rate": 1.9154939450528587e-05, "loss": 1.0815, "step": 5136 }, { "epoch": 1.58, "learning_rate": 1.915453846535047e-05, "loss": 0.9662, "step": 5137 }, { "epoch": 1.58, "learning_rate": 1.915413738925935e-05, "loss": 1.2824, "step": 5138 }, { "epoch": 1.58, "learning_rate": 1.9153736222259204e-05, "loss": 1.0507, "step": 5139 }, { "epoch": 1.58, "learning_rate": 1.9153334964354025e-05, "loss": 1.4908, "step": 5140 }, { "epoch": 1.58, "learning_rate": 1.91529336155478e-05, "loss": 1.075, "step": 5141 }, { "epoch": 1.58, "learning_rate": 1.91525321758445e-05, "loss": 1.3532, "step": 5142 }, { "epoch": 1.58, "learning_rate": 1.915213064524812e-05, "loss": 1.1369, "step": 5143 }, { "epoch": 1.58, "learning_rate": 1.915172902376265e-05, "loss": 1.2982, "step": 5144 }, { "epoch": 1.58, "learning_rate": 1.9151327311392078e-05, "loss": 1.4842, "step": 5145 }, { "epoch": 1.58, "learning_rate": 1.915092550814039e-05, "loss": 1.1298, "step": 5146 }, { "epoch": 1.58, "learning_rate": 1.9150523614011577e-05, "loss": 1.3256, "step": 5147 }, { "epoch": 1.58, "learning_rate": 1.9150121629009634e-05, "loss": 1.2534, "step": 5148 }, { "epoch": 1.58, "learning_rate": 1.9149719553138548e-05, "loss": 1.2896, "step": 5149 }, { "epoch": 1.58, "learning_rate": 1.914931738640232e-05, "loss": 1.5003, "step": 5150 }, { "epoch": 1.58, "learning_rate": 1.914891512880493e-05, "loss": 1.2615, "step": 5151 }, { "epoch": 1.59, "learning_rate": 1.9148512780350384e-05, "loss": 1.3798, "step": 5152 }, { "epoch": 1.59, "learning_rate": 1.9148110341042672e-05, "loss": 1.0621, "step": 5153 }, { "epoch": 1.59, "learning_rate": 1.9147707810885798e-05, "loss": 1.1365, "step": 5154 }, { "epoch": 1.59, "learning_rate": 1.9147305189883752e-05, "loss": 1.0323, "step": 5155 }, { "epoch": 1.59, "learning_rate": 1.9146902478040537e-05, "loss": 0.9904, "step": 5156 }, { "epoch": 1.59, "learning_rate": 1.914649967536015e-05, "loss": 1.2523, "step": 5157 }, { "epoch": 1.59, "learning_rate": 1.9146096781846588e-05, "loss": 1.4754, "step": 5158 }, { "epoch": 1.59, "learning_rate": 1.9145693797503857e-05, "loss": 1.1055, "step": 5159 }, { "epoch": 1.59, "learning_rate": 1.9145290722335962e-05, "loss": 1.0363, "step": 5160 }, { "epoch": 1.59, "learning_rate": 1.9144887556346898e-05, "loss": 1.2183, "step": 5161 }, { "epoch": 1.59, "learning_rate": 1.914448429954067e-05, "loss": 1.3656, "step": 5162 }, { "epoch": 1.59, "learning_rate": 1.914408095192129e-05, "loss": 1.4884, "step": 5163 }, { "epoch": 1.59, "learning_rate": 1.9143677513492757e-05, "loss": 0.9427, "step": 5164 }, { "epoch": 1.59, "learning_rate": 1.9143273984259078e-05, "loss": 1.1986, "step": 5165 }, { "epoch": 1.59, "learning_rate": 1.9142870364224263e-05, "loss": 1.0984, "step": 5166 }, { "epoch": 1.59, "learning_rate": 1.9142466653392317e-05, "loss": 1.2958, "step": 5167 }, { "epoch": 1.59, "learning_rate": 1.9142062851767252e-05, "loss": 1.3794, "step": 5168 }, { "epoch": 1.59, "learning_rate": 1.914165895935308e-05, "loss": 1.3547, "step": 5169 }, { "epoch": 1.59, "learning_rate": 1.9141254976153806e-05, "loss": 1.3174, "step": 5170 }, { "epoch": 1.59, "learning_rate": 1.9140850902173445e-05, "loss": 1.045, "step": 5171 }, { "epoch": 1.59, "learning_rate": 1.9140446737416014e-05, "loss": 1.1752, "step": 5172 }, { "epoch": 1.59, "learning_rate": 1.914004248188552e-05, "loss": 0.8303, "step": 5173 }, { "epoch": 1.59, "learning_rate": 1.913963813558598e-05, "loss": 1.4091, "step": 5174 }, { "epoch": 1.59, "learning_rate": 1.9139233698521412e-05, "loss": 1.426, "step": 5175 }, { "epoch": 1.59, "learning_rate": 1.9138829170695828e-05, "loss": 1.4177, "step": 5176 }, { "epoch": 1.59, "learning_rate": 1.9138424552113248e-05, "loss": 1.8262, "step": 5177 }, { "epoch": 1.59, "learning_rate": 1.913801984277769e-05, "loss": 1.5515, "step": 5178 }, { "epoch": 1.59, "learning_rate": 1.9137615042693177e-05, "loss": 1.3202, "step": 5179 }, { "epoch": 1.59, "learning_rate": 1.913721015186372e-05, "loss": 0.9016, "step": 5180 }, { "epoch": 1.59, "learning_rate": 1.9136805170293348e-05, "loss": 1.5153, "step": 5181 }, { "epoch": 1.59, "learning_rate": 1.9136400097986078e-05, "loss": 1.0907, "step": 5182 }, { "epoch": 1.59, "learning_rate": 1.913599493494594e-05, "loss": 1.3646, "step": 5183 }, { "epoch": 1.6, "learning_rate": 1.9135589681176947e-05, "loss": 1.4628, "step": 5184 }, { "epoch": 1.6, "learning_rate": 1.9135184336683127e-05, "loss": 1.1618, "step": 5185 }, { "epoch": 1.6, "learning_rate": 1.913477890146851e-05, "loss": 2.0321, "step": 5186 }, { "epoch": 1.6, "learning_rate": 1.9134373375537123e-05, "loss": 1.2052, "step": 5187 }, { "epoch": 1.6, "learning_rate": 1.9133967758892986e-05, "loss": 1.5422, "step": 5188 }, { "epoch": 1.6, "learning_rate": 1.9133562051540134e-05, "loss": 1.4002, "step": 5189 }, { "epoch": 1.6, "learning_rate": 1.913315625348259e-05, "loss": 1.9204, "step": 5190 }, { "epoch": 1.6, "learning_rate": 1.9132750364724385e-05, "loss": 1.1825, "step": 5191 }, { "epoch": 1.6, "learning_rate": 1.9132344385269558e-05, "loss": 1.1681, "step": 5192 }, { "epoch": 1.6, "learning_rate": 1.913193831512213e-05, "loss": 1.0903, "step": 5193 }, { "epoch": 1.6, "learning_rate": 1.9131532154286142e-05, "loss": 1.2153, "step": 5194 }, { "epoch": 1.6, "learning_rate": 1.913112590276562e-05, "loss": 1.227, "step": 5195 }, { "epoch": 1.6, "learning_rate": 1.9130719560564604e-05, "loss": 1.1419, "step": 5196 }, { "epoch": 1.6, "learning_rate": 1.9130313127687128e-05, "loss": 1.1595, "step": 5197 }, { "epoch": 1.6, "learning_rate": 1.9129906604137228e-05, "loss": 1.4409, "step": 5198 }, { "epoch": 1.6, "learning_rate": 1.912949998991894e-05, "loss": 1.3211, "step": 5199 }, { "epoch": 1.6, "learning_rate": 1.91290932850363e-05, "loss": 1.1723, "step": 5200 }, { "epoch": 1.6, "learning_rate": 1.9128686489493358e-05, "loss": 1.2635, "step": 5201 }, { "epoch": 1.6, "learning_rate": 1.9128279603294138e-05, "loss": 1.0206, "step": 5202 }, { "epoch": 1.6, "learning_rate": 1.9127872626442695e-05, "loss": 0.9009, "step": 5203 }, { "epoch": 1.6, "learning_rate": 1.9127465558943063e-05, "loss": 1.2458, "step": 5204 }, { "epoch": 1.6, "learning_rate": 1.9127058400799284e-05, "loss": 1.4555, "step": 5205 }, { "epoch": 1.6, "learning_rate": 1.9126651152015404e-05, "loss": 1.6545, "step": 5206 }, { "epoch": 1.6, "learning_rate": 1.9126243812595465e-05, "loss": 1.1959, "step": 5207 }, { "epoch": 1.6, "learning_rate": 1.9125836382543516e-05, "loss": 1.2675, "step": 5208 }, { "epoch": 1.6, "learning_rate": 1.91254288618636e-05, "loss": 1.208, "step": 5209 }, { "epoch": 1.6, "learning_rate": 1.9125021250559767e-05, "loss": 1.2609, "step": 5210 }, { "epoch": 1.6, "learning_rate": 1.9124613548636063e-05, "loss": 1.304, "step": 5211 }, { "epoch": 1.6, "learning_rate": 1.9124205756096533e-05, "loss": 0.7149, "step": 5212 }, { "epoch": 1.6, "learning_rate": 1.9123797872945237e-05, "loss": 1.5134, "step": 5213 }, { "epoch": 1.6, "learning_rate": 1.9123389899186216e-05, "loss": 1.0264, "step": 5214 }, { "epoch": 1.6, "learning_rate": 1.9122981834823524e-05, "loss": 1.2165, "step": 5215 }, { "epoch": 1.6, "learning_rate": 1.9122573679861215e-05, "loss": 1.2067, "step": 5216 }, { "epoch": 1.61, "learning_rate": 1.9122165434303342e-05, "loss": 0.9826, "step": 5217 }, { "epoch": 1.61, "learning_rate": 1.912175709815396e-05, "loss": 1.2624, "step": 5218 }, { "epoch": 1.61, "learning_rate": 1.912134867141712e-05, "loss": 1.3961, "step": 5219 }, { "epoch": 1.61, "learning_rate": 1.9120940154096884e-05, "loss": 1.4662, "step": 5220 }, { "epoch": 1.61, "learning_rate": 1.9120531546197306e-05, "loss": 1.2152, "step": 5221 }, { "epoch": 1.61, "learning_rate": 1.9120122847722446e-05, "loss": 1.3655, "step": 5222 }, { "epoch": 1.61, "learning_rate": 1.9119714058676355e-05, "loss": 0.9423, "step": 5223 }, { "epoch": 1.61, "learning_rate": 1.9119305179063103e-05, "loss": 1.0415, "step": 5224 }, { "epoch": 1.61, "learning_rate": 1.9118896208886744e-05, "loss": 1.4265, "step": 5225 }, { "epoch": 1.61, "learning_rate": 1.9118487148151342e-05, "loss": 1.2098, "step": 5226 }, { "epoch": 1.61, "learning_rate": 1.911807799686096e-05, "loss": 0.8389, "step": 5227 }, { "epoch": 1.61, "learning_rate": 1.9117668755019657e-05, "loss": 0.8733, "step": 5228 }, { "epoch": 1.61, "learning_rate": 1.9117259422631506e-05, "loss": 1.0421, "step": 5229 }, { "epoch": 1.61, "learning_rate": 1.911684999970056e-05, "loss": 1.1556, "step": 5230 }, { "epoch": 1.61, "learning_rate": 1.9116440486230893e-05, "loss": 0.9421, "step": 5231 }, { "epoch": 1.61, "learning_rate": 1.911603088222657e-05, "loss": 1.2163, "step": 5232 }, { "epoch": 1.61, "learning_rate": 1.911562118769166e-05, "loss": 1.0246, "step": 5233 }, { "epoch": 1.61, "learning_rate": 1.911521140263023e-05, "loss": 1.3843, "step": 5234 }, { "epoch": 1.61, "learning_rate": 1.911480152704635e-05, "loss": 1.4514, "step": 5235 }, { "epoch": 1.61, "learning_rate": 1.9114391560944086e-05, "loss": 1.508, "step": 5236 }, { "epoch": 1.61, "learning_rate": 1.911398150432752e-05, "loss": 1.4312, "step": 5237 }, { "epoch": 1.61, "learning_rate": 1.9113571357200716e-05, "loss": 1.2003, "step": 5238 }, { "epoch": 1.61, "learning_rate": 1.911316111956775e-05, "loss": 1.1165, "step": 5239 }, { "epoch": 1.61, "learning_rate": 1.9112750791432692e-05, "loss": 1.124, "step": 5240 }, { "epoch": 1.61, "learning_rate": 1.9112340372799623e-05, "loss": 1.504, "step": 5241 }, { "epoch": 1.61, "learning_rate": 1.9111929863672617e-05, "loss": 1.2892, "step": 5242 }, { "epoch": 1.61, "learning_rate": 1.9111519264055745e-05, "loss": 1.3553, "step": 5243 }, { "epoch": 1.61, "learning_rate": 1.9111108573953095e-05, "loss": 1.0253, "step": 5244 }, { "epoch": 1.61, "learning_rate": 1.9110697793368733e-05, "loss": 1.086, "step": 5245 }, { "epoch": 1.61, "learning_rate": 1.911028692230675e-05, "loss": 1.4472, "step": 5246 }, { "epoch": 1.61, "learning_rate": 1.910987596077122e-05, "loss": 1.0757, "step": 5247 }, { "epoch": 1.61, "learning_rate": 1.9109464908766225e-05, "loss": 1.3111, "step": 5248 }, { "epoch": 1.62, "learning_rate": 1.910905376629585e-05, "loss": 1.5533, "step": 5249 }, { "epoch": 1.62, "learning_rate": 1.910864253336418e-05, "loss": 1.3228, "step": 5250 }, { "epoch": 1.62, "learning_rate": 1.9108231209975286e-05, "loss": 1.724, "step": 5251 }, { "epoch": 1.62, "learning_rate": 1.9107819796133266e-05, "loss": 1.0847, "step": 5252 }, { "epoch": 1.62, "learning_rate": 1.91074082918422e-05, "loss": 1.1274, "step": 5253 }, { "epoch": 1.62, "learning_rate": 1.910699669710618e-05, "loss": 1.0879, "step": 5254 }, { "epoch": 1.62, "learning_rate": 1.9106585011929287e-05, "loss": 1.0651, "step": 5255 }, { "epoch": 1.62, "learning_rate": 1.9106173236315612e-05, "loss": 1.1969, "step": 5256 }, { "epoch": 1.62, "learning_rate": 1.9105761370269244e-05, "loss": 1.0422, "step": 5257 }, { "epoch": 1.62, "learning_rate": 1.9105349413794272e-05, "loss": 1.1916, "step": 5258 }, { "epoch": 1.62, "learning_rate": 1.9104937366894793e-05, "loss": 1.1709, "step": 5259 }, { "epoch": 1.62, "learning_rate": 1.910452522957489e-05, "loss": 1.273, "step": 5260 }, { "epoch": 1.62, "learning_rate": 1.9104113001838662e-05, "loss": 1.2464, "step": 5261 }, { "epoch": 1.62, "learning_rate": 1.9103700683690204e-05, "loss": 1.441, "step": 5262 }, { "epoch": 1.62, "learning_rate": 1.9103288275133602e-05, "loss": 1.0129, "step": 5263 }, { "epoch": 1.62, "learning_rate": 1.910287577617296e-05, "loss": 1.2394, "step": 5264 }, { "epoch": 1.62, "learning_rate": 1.9102463186812375e-05, "loss": 1.1477, "step": 5265 }, { "epoch": 1.62, "learning_rate": 1.910205050705594e-05, "loss": 1.4251, "step": 5266 }, { "epoch": 1.62, "learning_rate": 1.9101637736907756e-05, "loss": 0.9763, "step": 5267 }, { "epoch": 1.62, "learning_rate": 1.9101224876371917e-05, "loss": 1.1122, "step": 5268 }, { "epoch": 1.62, "learning_rate": 1.910081192545253e-05, "loss": 1.23, "step": 5269 }, { "epoch": 1.62, "learning_rate": 1.9100398884153694e-05, "loss": 1.4749, "step": 5270 }, { "epoch": 1.62, "learning_rate": 1.9099985752479505e-05, "loss": 1.1891, "step": 5271 }, { "epoch": 1.62, "learning_rate": 1.9099572530434077e-05, "loss": 1.2153, "step": 5272 }, { "epoch": 1.62, "learning_rate": 1.9099159218021503e-05, "loss": 1.1468, "step": 5273 }, { "epoch": 1.62, "learning_rate": 1.9098745815245895e-05, "loss": 0.9764, "step": 5274 }, { "epoch": 1.62, "learning_rate": 1.9098332322111353e-05, "loss": 1.475, "step": 5275 }, { "epoch": 1.62, "learning_rate": 1.9097918738621984e-05, "loss": 1.6849, "step": 5276 }, { "epoch": 1.62, "learning_rate": 1.90975050647819e-05, "loss": 1.2701, "step": 5277 }, { "epoch": 1.62, "learning_rate": 1.9097091300595204e-05, "loss": 1.375, "step": 5278 }, { "epoch": 1.62, "learning_rate": 1.9096677446066012e-05, "loss": 1.2172, "step": 5279 }, { "epoch": 1.62, "learning_rate": 1.9096263501198425e-05, "loss": 1.3928, "step": 5280 }, { "epoch": 1.62, "learning_rate": 1.909584946599656e-05, "loss": 1.1796, "step": 5281 }, { "epoch": 1.63, "learning_rate": 1.9095435340464526e-05, "loss": 1.3818, "step": 5282 }, { "epoch": 1.63, "learning_rate": 1.9095021124606434e-05, "loss": 1.3736, "step": 5283 }, { "epoch": 1.63, "learning_rate": 1.9094606818426403e-05, "loss": 1.1986, "step": 5284 }, { "epoch": 1.63, "learning_rate": 1.9094192421928547e-05, "loss": 1.2078, "step": 5285 }, { "epoch": 1.63, "learning_rate": 1.9093777935116975e-05, "loss": 1.2379, "step": 5286 }, { "epoch": 1.63, "learning_rate": 1.9093363357995807e-05, "loss": 1.0681, "step": 5287 }, { "epoch": 1.63, "learning_rate": 1.909294869056916e-05, "loss": 1.2264, "step": 5288 }, { "epoch": 1.63, "learning_rate": 1.9092533932841153e-05, "loss": 1.1184, "step": 5289 }, { "epoch": 1.63, "learning_rate": 1.9092119084815904e-05, "loss": 1.246, "step": 5290 }, { "epoch": 1.63, "learning_rate": 1.9091704146497533e-05, "loss": 1.2037, "step": 5291 }, { "epoch": 1.63, "learning_rate": 1.9091289117890165e-05, "loss": 1.1929, "step": 5292 }, { "epoch": 1.63, "learning_rate": 1.9090873998997912e-05, "loss": 1.3897, "step": 5293 }, { "epoch": 1.63, "learning_rate": 1.9090458789824902e-05, "loss": 1.1878, "step": 5294 }, { "epoch": 1.63, "learning_rate": 1.9090043490375257e-05, "loss": 1.2643, "step": 5295 }, { "epoch": 1.63, "learning_rate": 1.9089628100653106e-05, "loss": 1.0896, "step": 5296 }, { "epoch": 1.63, "learning_rate": 1.908921262066257e-05, "loss": 1.502, "step": 5297 }, { "epoch": 1.63, "learning_rate": 1.9088797050407775e-05, "loss": 0.8261, "step": 5298 }, { "epoch": 1.63, "learning_rate": 1.908838138989285e-05, "loss": 1.2527, "step": 5299 }, { "epoch": 1.63, "learning_rate": 1.908796563912192e-05, "loss": 1.0741, "step": 5300 }, { "epoch": 1.63, "learning_rate": 1.9087549798099116e-05, "loss": 1.2614, "step": 5301 }, { "epoch": 1.63, "learning_rate": 1.9087133866828567e-05, "loss": 0.9804, "step": 5302 }, { "epoch": 1.63, "learning_rate": 1.9086717845314403e-05, "loss": 1.2796, "step": 5303 }, { "epoch": 1.63, "learning_rate": 1.9086301733560758e-05, "loss": 1.271, "step": 5304 }, { "epoch": 1.63, "learning_rate": 1.908588553157176e-05, "loss": 1.1411, "step": 5305 }, { "epoch": 1.63, "learning_rate": 1.9085469239351552e-05, "loss": 0.9418, "step": 5306 }, { "epoch": 1.63, "learning_rate": 1.9085052856904252e-05, "loss": 1.4497, "step": 5307 }, { "epoch": 1.63, "learning_rate": 1.9084636384234008e-05, "loss": 1.4903, "step": 5308 }, { "epoch": 1.63, "learning_rate": 1.9084219821344953e-05, "loss": 1.8153, "step": 5309 }, { "epoch": 1.63, "learning_rate": 1.9083803168241225e-05, "loss": 1.2147, "step": 5310 }, { "epoch": 1.63, "learning_rate": 1.9083386424926952e-05, "loss": 1.2894, "step": 5311 }, { "epoch": 1.63, "learning_rate": 1.908296959140629e-05, "loss": 1.1502, "step": 5312 }, { "epoch": 1.63, "learning_rate": 1.908255266768336e-05, "loss": 0.8612, "step": 5313 }, { "epoch": 1.64, "learning_rate": 1.9082135653762316e-05, "loss": 1.1875, "step": 5314 }, { "epoch": 1.64, "learning_rate": 1.9081718549647292e-05, "loss": 1.2318, "step": 5315 }, { "epoch": 1.64, "learning_rate": 1.9081301355342433e-05, "loss": 1.161, "step": 5316 }, { "epoch": 1.64, "learning_rate": 1.9080884070851885e-05, "loss": 1.01, "step": 5317 }, { "epoch": 1.64, "learning_rate": 1.908046669617979e-05, "loss": 1.1245, "step": 5318 }, { "epoch": 1.64, "learning_rate": 1.908004923133029e-05, "loss": 1.2518, "step": 5319 }, { "epoch": 1.64, "learning_rate": 1.9079631676307532e-05, "loss": 1.4365, "step": 5320 }, { "epoch": 1.64, "learning_rate": 1.907921403111566e-05, "loss": 1.1752, "step": 5321 }, { "epoch": 1.64, "learning_rate": 1.907879629575883e-05, "loss": 1.306, "step": 5322 }, { "epoch": 1.64, "learning_rate": 1.9078378470241183e-05, "loss": 1.411, "step": 5323 }, { "epoch": 1.64, "learning_rate": 1.9077960554566873e-05, "loss": 1.1348, "step": 5324 }, { "epoch": 1.64, "learning_rate": 1.9077542548740046e-05, "loss": 1.2164, "step": 5325 }, { "epoch": 1.64, "learning_rate": 1.9077124452764856e-05, "loss": 1.522, "step": 5326 }, { "epoch": 1.64, "learning_rate": 1.9076706266645456e-05, "loss": 1.0307, "step": 5327 }, { "epoch": 1.64, "learning_rate": 1.9076287990386e-05, "loss": 0.7692, "step": 5328 }, { "epoch": 1.64, "learning_rate": 1.9075869623990632e-05, "loss": 1.1824, "step": 5329 }, { "epoch": 1.64, "learning_rate": 1.907545116746352e-05, "loss": 1.0002, "step": 5330 }, { "epoch": 1.64, "learning_rate": 1.9075032620808808e-05, "loss": 1.0598, "step": 5331 }, { "epoch": 1.64, "learning_rate": 1.9074613984030664e-05, "loss": 1.0832, "step": 5332 }, { "epoch": 1.64, "learning_rate": 1.9074195257133237e-05, "loss": 1.4813, "step": 5333 }, { "epoch": 1.64, "learning_rate": 1.907377644012069e-05, "loss": 0.9662, "step": 5334 }, { "epoch": 1.64, "learning_rate": 1.9073357532997178e-05, "loss": 1.2578, "step": 5335 }, { "epoch": 1.64, "learning_rate": 1.9072938535766864e-05, "loss": 0.9006, "step": 5336 }, { "epoch": 1.64, "learning_rate": 1.907251944843391e-05, "loss": 1.23, "step": 5337 }, { "epoch": 1.64, "learning_rate": 1.9072100271002472e-05, "loss": 0.968, "step": 5338 }, { "epoch": 1.64, "learning_rate": 1.907168100347672e-05, "loss": 1.0838, "step": 5339 }, { "epoch": 1.64, "learning_rate": 1.9071261645860818e-05, "loss": 1.5079, "step": 5340 }, { "epoch": 1.64, "learning_rate": 1.9070842198158925e-05, "loss": 1.0992, "step": 5341 }, { "epoch": 1.64, "learning_rate": 1.9070422660375207e-05, "loss": 1.0902, "step": 5342 }, { "epoch": 1.64, "learning_rate": 1.9070003032513838e-05, "loss": 0.9562, "step": 5343 }, { "epoch": 1.64, "learning_rate": 1.9069583314578975e-05, "loss": 1.9283, "step": 5344 }, { "epoch": 1.64, "learning_rate": 1.906916350657479e-05, "loss": 0.9199, "step": 5345 }, { "epoch": 1.64, "learning_rate": 1.9068743608505454e-05, "loss": 1.5508, "step": 5346 }, { "epoch": 1.65, "learning_rate": 1.906832362037514e-05, "loss": 1.1708, "step": 5347 }, { "epoch": 1.65, "learning_rate": 1.906790354218801e-05, "loss": 1.4278, "step": 5348 }, { "epoch": 1.65, "learning_rate": 1.9067483373948245e-05, "loss": 1.0736, "step": 5349 }, { "epoch": 1.65, "learning_rate": 1.906706311566001e-05, "loss": 1.4429, "step": 5350 }, { "epoch": 1.65, "learning_rate": 1.9066642767327482e-05, "loss": 1.1249, "step": 5351 }, { "epoch": 1.65, "learning_rate": 1.9066222328954834e-05, "loss": 1.1554, "step": 5352 }, { "epoch": 1.65, "learning_rate": 1.9065801800546247e-05, "loss": 1.8231, "step": 5353 }, { "epoch": 1.65, "learning_rate": 1.906538118210589e-05, "loss": 1.4839, "step": 5354 }, { "epoch": 1.65, "learning_rate": 1.9064960473637943e-05, "loss": 1.4494, "step": 5355 }, { "epoch": 1.65, "learning_rate": 1.9064539675146584e-05, "loss": 0.9884, "step": 5356 }, { "epoch": 1.65, "learning_rate": 1.9064118786635993e-05, "loss": 1.2292, "step": 5357 }, { "epoch": 1.65, "learning_rate": 1.9063697808110344e-05, "loss": 1.0625, "step": 5358 }, { "epoch": 1.65, "learning_rate": 1.9063276739573826e-05, "loss": 0.9923, "step": 5359 }, { "epoch": 1.65, "learning_rate": 1.9062855581030616e-05, "loss": 1.2512, "step": 5360 }, { "epoch": 1.65, "learning_rate": 1.90624343324849e-05, "loss": 1.0213, "step": 5361 }, { "epoch": 1.65, "learning_rate": 1.906201299394086e-05, "loss": 1.0373, "step": 5362 }, { "epoch": 1.65, "learning_rate": 1.9061591565402674e-05, "loss": 1.0304, "step": 5363 }, { "epoch": 1.65, "learning_rate": 1.9061170046874537e-05, "loss": 1.1587, "step": 5364 }, { "epoch": 1.65, "learning_rate": 1.9060748438360628e-05, "loss": 1.377, "step": 5365 }, { "epoch": 1.65, "learning_rate": 1.9060326739865135e-05, "loss": 0.957, "step": 5366 }, { "epoch": 1.65, "learning_rate": 1.905990495139225e-05, "loss": 1.2549, "step": 5367 }, { "epoch": 1.65, "learning_rate": 1.905948307294616e-05, "loss": 1.3719, "step": 5368 }, { "epoch": 1.65, "learning_rate": 1.905906110453105e-05, "loss": 1.2428, "step": 5369 }, { "epoch": 1.65, "learning_rate": 1.9058639046151116e-05, "loss": 1.1491, "step": 5370 }, { "epoch": 1.65, "learning_rate": 1.9058216897810547e-05, "loss": 0.9942, "step": 5371 }, { "epoch": 1.65, "learning_rate": 1.9057794659513537e-05, "loss": 1.2743, "step": 5372 }, { "epoch": 1.65, "learning_rate": 1.9057372331264277e-05, "loss": 1.0874, "step": 5373 }, { "epoch": 1.65, "learning_rate": 1.9056949913066963e-05, "loss": 1.3619, "step": 5374 }, { "epoch": 1.65, "learning_rate": 1.9056527404925788e-05, "loss": 1.314, "step": 5375 }, { "epoch": 1.65, "learning_rate": 1.905610480684495e-05, "loss": 1.4458, "step": 5376 }, { "epoch": 1.65, "learning_rate": 1.9055682118828645e-05, "loss": 1.1492, "step": 5377 }, { "epoch": 1.65, "learning_rate": 1.9055259340881072e-05, "loss": 1.3378, "step": 5378 }, { "epoch": 1.66, "learning_rate": 1.9054836473006424e-05, "loss": 1.058, "step": 5379 }, { "epoch": 1.66, "learning_rate": 1.905441351520891e-05, "loss": 1.1749, "step": 5380 }, { "epoch": 1.66, "learning_rate": 1.9053990467492724e-05, "loss": 1.3545, "step": 5381 }, { "epoch": 1.66, "learning_rate": 1.9053567329862063e-05, "loss": 1.2959, "step": 5382 }, { "epoch": 1.66, "learning_rate": 1.905314410232114e-05, "loss": 0.9524, "step": 5383 }, { "epoch": 1.66, "learning_rate": 1.905272078487415e-05, "loss": 1.0924, "step": 5384 }, { "epoch": 1.66, "learning_rate": 1.90522973775253e-05, "loss": 1.2703, "step": 5385 }, { "epoch": 1.66, "learning_rate": 1.9051873880278796e-05, "loss": 0.9249, "step": 5386 }, { "epoch": 1.66, "learning_rate": 1.9051450293138837e-05, "loss": 1.1478, "step": 5387 }, { "epoch": 1.66, "learning_rate": 1.9051026616109637e-05, "loss": 1.0944, "step": 5388 }, { "epoch": 1.66, "learning_rate": 1.9050602849195404e-05, "loss": 1.0239, "step": 5389 }, { "epoch": 1.66, "learning_rate": 1.905017899240034e-05, "loss": 1.2816, "step": 5390 }, { "epoch": 1.66, "learning_rate": 1.9049755045728658e-05, "loss": 1.3549, "step": 5391 }, { "epoch": 1.66, "learning_rate": 1.9049331009184567e-05, "loss": 1.046, "step": 5392 }, { "epoch": 1.66, "learning_rate": 1.9048906882772282e-05, "loss": 1.1462, "step": 5393 }, { "epoch": 1.66, "learning_rate": 1.9048482666496013e-05, "loss": 1.0857, "step": 5394 }, { "epoch": 1.66, "learning_rate": 1.9048058360359966e-05, "loss": 1.3278, "step": 5395 }, { "epoch": 1.66, "learning_rate": 1.9047633964368363e-05, "loss": 1.5102, "step": 5396 }, { "epoch": 1.66, "learning_rate": 1.9047209478525416e-05, "loss": 1.4596, "step": 5397 }, { "epoch": 1.66, "learning_rate": 1.9046784902835344e-05, "loss": 1.0463, "step": 5398 }, { "epoch": 1.66, "learning_rate": 1.9046360237302357e-05, "loss": 0.943, "step": 5399 }, { "epoch": 1.66, "learning_rate": 1.9045935481930677e-05, "loss": 1.2809, "step": 5400 }, { "epoch": 1.66, "learning_rate": 1.904551063672452e-05, "loss": 1.1784, "step": 5401 }, { "epoch": 1.66, "learning_rate": 1.9045085701688105e-05, "loss": 1.3148, "step": 5402 }, { "epoch": 1.66, "learning_rate": 1.9044660676825656e-05, "loss": 1.4399, "step": 5403 }, { "epoch": 1.66, "learning_rate": 1.9044235562141386e-05, "loss": 1.4366, "step": 5404 }, { "epoch": 1.66, "learning_rate": 1.9043810357639526e-05, "loss": 1.3724, "step": 5405 }, { "epoch": 1.66, "learning_rate": 1.9043385063324292e-05, "loss": 1.0423, "step": 5406 }, { "epoch": 1.66, "learning_rate": 1.904295967919991e-05, "loss": 1.1012, "step": 5407 }, { "epoch": 1.66, "learning_rate": 1.90425342052706e-05, "loss": 1.319, "step": 5408 }, { "epoch": 1.66, "learning_rate": 1.9042108641540596e-05, "loss": 1.2773, "step": 5409 }, { "epoch": 1.66, "learning_rate": 1.904168298801412e-05, "loss": 1.0636, "step": 5410 }, { "epoch": 1.66, "learning_rate": 1.9041257244695397e-05, "loss": 1.0985, "step": 5411 }, { "epoch": 1.67, "learning_rate": 1.9040831411588658e-05, "loss": 1.0101, "step": 5412 }, { "epoch": 1.67, "learning_rate": 1.904040548869813e-05, "loss": 0.7912, "step": 5413 }, { "epoch": 1.67, "learning_rate": 1.9039979476028044e-05, "loss": 1.0247, "step": 5414 }, { "epoch": 1.67, "learning_rate": 1.9039553373582632e-05, "loss": 1.0678, "step": 5415 }, { "epoch": 1.67, "learning_rate": 1.903912718136612e-05, "loss": 1.3779, "step": 5416 }, { "epoch": 1.67, "learning_rate": 1.9038700899382746e-05, "loss": 1.0926, "step": 5417 }, { "epoch": 1.67, "learning_rate": 1.9038274527636744e-05, "loss": 1.0292, "step": 5418 }, { "epoch": 1.67, "learning_rate": 1.9037848066132345e-05, "loss": 0.8897, "step": 5419 }, { "epoch": 1.67, "learning_rate": 1.9037421514873787e-05, "loss": 1.1055, "step": 5420 }, { "epoch": 1.67, "learning_rate": 1.90369948738653e-05, "loss": 1.4377, "step": 5421 }, { "epoch": 1.67, "learning_rate": 1.903656814311113e-05, "loss": 1.1535, "step": 5422 }, { "epoch": 1.67, "learning_rate": 1.9036141322615504e-05, "loss": 0.8926, "step": 5423 }, { "epoch": 1.67, "learning_rate": 1.903571441238267e-05, "loss": 1.2831, "step": 5424 }, { "epoch": 1.67, "learning_rate": 1.9035287412416864e-05, "loss": 1.2166, "step": 5425 }, { "epoch": 1.67, "learning_rate": 1.9034860322722325e-05, "loss": 1.3676, "step": 5426 }, { "epoch": 1.67, "learning_rate": 1.90344331433033e-05, "loss": 1.0668, "step": 5427 }, { "epoch": 1.67, "learning_rate": 1.9034005874164024e-05, "loss": 1.1417, "step": 5428 }, { "epoch": 1.67, "learning_rate": 1.9033578515308746e-05, "loss": 1.22, "step": 5429 }, { "epoch": 1.67, "learning_rate": 1.9033151066741706e-05, "loss": 1.381, "step": 5430 }, { "epoch": 1.67, "learning_rate": 1.9032723528467154e-05, "loss": 1.0651, "step": 5431 }, { "epoch": 1.67, "learning_rate": 1.903229590048933e-05, "loss": 1.0198, "step": 5432 }, { "epoch": 1.67, "learning_rate": 1.9031868182812484e-05, "loss": 1.1408, "step": 5433 }, { "epoch": 1.67, "learning_rate": 1.903144037544086e-05, "loss": 1.2572, "step": 5434 }, { "epoch": 1.67, "learning_rate": 1.9031012478378714e-05, "loss": 1.3719, "step": 5435 }, { "epoch": 1.67, "learning_rate": 1.903058449163029e-05, "loss": 1.0969, "step": 5436 }, { "epoch": 1.67, "learning_rate": 1.9030156415199836e-05, "loss": 1.1997, "step": 5437 }, { "epoch": 1.67, "learning_rate": 1.902972824909161e-05, "loss": 1.3828, "step": 5438 }, { "epoch": 1.67, "learning_rate": 1.9029299993309857e-05, "loss": 1.1637, "step": 5439 }, { "epoch": 1.67, "learning_rate": 1.9028871647858836e-05, "loss": 1.19, "step": 5440 }, { "epoch": 1.67, "learning_rate": 1.9028443212742797e-05, "loss": 1.6995, "step": 5441 }, { "epoch": 1.67, "learning_rate": 1.9028014687965997e-05, "loss": 1.1003, "step": 5442 }, { "epoch": 1.67, "learning_rate": 1.902758607353269e-05, "loss": 1.0467, "step": 5443 }, { "epoch": 1.68, "learning_rate": 1.9027157369447134e-05, "loss": 2.0516, "step": 5444 }, { "epoch": 1.68, "learning_rate": 1.9026728575713585e-05, "loss": 1.2166, "step": 5445 }, { "epoch": 1.68, "learning_rate": 1.90262996923363e-05, "loss": 1.1336, "step": 5446 }, { "epoch": 1.68, "learning_rate": 1.9025870719319544e-05, "loss": 1.3804, "step": 5447 }, { "epoch": 1.68, "learning_rate": 1.9025441656667572e-05, "loss": 1.0305, "step": 5448 }, { "epoch": 1.68, "learning_rate": 1.9025012504384645e-05, "loss": 1.2263, "step": 5449 }, { "epoch": 1.68, "learning_rate": 1.9024583262475027e-05, "loss": 0.7458, "step": 5450 }, { "epoch": 1.68, "learning_rate": 1.902415393094298e-05, "loss": 1.0802, "step": 5451 }, { "epoch": 1.68, "learning_rate": 1.902372450979277e-05, "loss": 1.0881, "step": 5452 }, { "epoch": 1.68, "learning_rate": 1.9023294999028654e-05, "loss": 1.0314, "step": 5453 }, { "epoch": 1.68, "learning_rate": 1.9022865398654903e-05, "loss": 1.2715, "step": 5454 }, { "epoch": 1.68, "learning_rate": 1.9022435708675785e-05, "loss": 1.1567, "step": 5455 }, { "epoch": 1.68, "learning_rate": 1.9022005929095568e-05, "loss": 0.8509, "step": 5456 }, { "epoch": 1.68, "learning_rate": 1.9021576059918513e-05, "loss": 1.0676, "step": 5457 }, { "epoch": 1.68, "learning_rate": 1.9021146101148896e-05, "loss": 1.0282, "step": 5458 }, { "epoch": 1.68, "learning_rate": 1.902071605279098e-05, "loss": 1.0879, "step": 5459 }, { "epoch": 1.68, "learning_rate": 1.9020285914849045e-05, "loss": 1.3586, "step": 5460 }, { "epoch": 1.68, "learning_rate": 1.9019855687327352e-05, "loss": 1.3574, "step": 5461 }, { "epoch": 1.68, "learning_rate": 1.901942537023018e-05, "loss": 1.1158, "step": 5462 }, { "epoch": 1.68, "learning_rate": 1.9018994963561806e-05, "loss": 1.0758, "step": 5463 }, { "epoch": 1.68, "learning_rate": 1.9018564467326498e-05, "loss": 1.2269, "step": 5464 }, { "epoch": 1.68, "learning_rate": 1.9018133881528534e-05, "loss": 1.215, "step": 5465 }, { "epoch": 1.68, "learning_rate": 1.9017703206172187e-05, "loss": 1.3076, "step": 5466 }, { "epoch": 1.68, "learning_rate": 1.9017272441261734e-05, "loss": 1.6913, "step": 5467 }, { "epoch": 1.68, "learning_rate": 1.901684158680146e-05, "loss": 0.9632, "step": 5468 }, { "epoch": 1.68, "learning_rate": 1.901641064279564e-05, "loss": 1.0992, "step": 5469 }, { "epoch": 1.68, "learning_rate": 1.9015979609248547e-05, "loss": 1.1868, "step": 5470 }, { "epoch": 1.68, "learning_rate": 1.901554848616447e-05, "loss": 1.1761, "step": 5471 }, { "epoch": 1.68, "learning_rate": 1.9015117273547684e-05, "loss": 1.3979, "step": 5472 }, { "epoch": 1.68, "learning_rate": 1.901468597140248e-05, "loss": 1.1191, "step": 5473 }, { "epoch": 1.68, "learning_rate": 1.9014254579733133e-05, "loss": 1.2304, "step": 5474 }, { "epoch": 1.68, "learning_rate": 1.9013823098543932e-05, "loss": 0.9204, "step": 5475 }, { "epoch": 1.68, "learning_rate": 1.901339152783916e-05, "loss": 1.2394, "step": 5476 }, { "epoch": 1.69, "learning_rate": 1.9012959867623102e-05, "loss": 1.0567, "step": 5477 }, { "epoch": 1.69, "learning_rate": 1.9012528117900045e-05, "loss": 1.3336, "step": 5478 }, { "epoch": 1.69, "learning_rate": 1.9012096278674283e-05, "loss": 1.0464, "step": 5479 }, { "epoch": 1.69, "learning_rate": 1.9011664349950095e-05, "loss": 1.5092, "step": 5480 }, { "epoch": 1.69, "learning_rate": 1.9011232331731773e-05, "loss": 1.6224, "step": 5481 }, { "epoch": 1.69, "learning_rate": 1.901080022402361e-05, "loss": 1.0523, "step": 5482 }, { "epoch": 1.69, "learning_rate": 1.90103680268299e-05, "loss": 1.2108, "step": 5483 }, { "epoch": 1.69, "learning_rate": 1.9009935740154928e-05, "loss": 1.3359, "step": 5484 }, { "epoch": 1.69, "learning_rate": 1.9009503364002986e-05, "loss": 1.4124, "step": 5485 }, { "epoch": 1.69, "learning_rate": 1.900907089837838e-05, "loss": 1.2792, "step": 5486 }, { "epoch": 1.69, "learning_rate": 1.9008638343285394e-05, "loss": 1.259, "step": 5487 }, { "epoch": 1.69, "learning_rate": 1.9008205698728326e-05, "loss": 1.4275, "step": 5488 }, { "epoch": 1.69, "learning_rate": 1.9007772964711476e-05, "loss": 1.4789, "step": 5489 }, { "epoch": 1.69, "learning_rate": 1.9007340141239135e-05, "loss": 1.249, "step": 5490 }, { "epoch": 1.69, "learning_rate": 1.9006907228315605e-05, "loss": 1.1428, "step": 5491 }, { "epoch": 1.69, "learning_rate": 1.900647422594519e-05, "loss": 1.5915, "step": 5492 }, { "epoch": 1.69, "learning_rate": 1.900604113413218e-05, "loss": 1.3612, "step": 5493 }, { "epoch": 1.69, "learning_rate": 1.9005607952880884e-05, "loss": 1.2284, "step": 5494 }, { "epoch": 1.69, "learning_rate": 1.90051746821956e-05, "loss": 1.2862, "step": 5495 }, { "epoch": 1.69, "learning_rate": 1.9004741322080633e-05, "loss": 1.1124, "step": 5496 }, { "epoch": 1.69, "learning_rate": 1.9004307872540283e-05, "loss": 1.3711, "step": 5497 }, { "epoch": 1.69, "learning_rate": 1.900387433357886e-05, "loss": 2.0831, "step": 5498 }, { "epoch": 1.69, "learning_rate": 1.9003440705200663e-05, "loss": 1.6991, "step": 5499 }, { "epoch": 1.69, "learning_rate": 1.9003006987410007e-05, "loss": 1.1269, "step": 5500 }, { "epoch": 1.69, "learning_rate": 1.9002573180211192e-05, "loss": 1.0742, "step": 5501 }, { "epoch": 1.69, "learning_rate": 1.900213928360853e-05, "loss": 1.0781, "step": 5502 }, { "epoch": 1.69, "learning_rate": 1.9001705297606325e-05, "loss": 1.2037, "step": 5503 }, { "epoch": 1.69, "learning_rate": 1.9001271222208893e-05, "loss": 1.4694, "step": 5504 }, { "epoch": 1.69, "learning_rate": 1.900083705742054e-05, "loss": 1.1861, "step": 5505 }, { "epoch": 1.69, "learning_rate": 1.900040280324558e-05, "loss": 1.2573, "step": 5506 }, { "epoch": 1.69, "learning_rate": 1.8999968459688328e-05, "loss": 0.9242, "step": 5507 }, { "epoch": 1.69, "learning_rate": 1.899953402675309e-05, "loss": 1.4251, "step": 5508 }, { "epoch": 1.7, "learning_rate": 1.8999099504444186e-05, "loss": 1.2347, "step": 5509 }, { "epoch": 1.7, "learning_rate": 1.8998664892765933e-05, "loss": 1.3887, "step": 5510 }, { "epoch": 1.7, "learning_rate": 1.8998230191722644e-05, "loss": 1.032, "step": 5511 }, { "epoch": 1.7, "learning_rate": 1.8997795401318633e-05, "loss": 1.0491, "step": 5512 }, { "epoch": 1.7, "learning_rate": 1.8997360521558223e-05, "loss": 1.2925, "step": 5513 }, { "epoch": 1.7, "learning_rate": 1.899692555244573e-05, "loss": 1.2208, "step": 5514 }, { "epoch": 1.7, "learning_rate": 1.8996490493985474e-05, "loss": 1.1319, "step": 5515 }, { "epoch": 1.7, "learning_rate": 1.8996055346181776e-05, "loss": 1.1709, "step": 5516 }, { "epoch": 1.7, "learning_rate": 1.899562010903896e-05, "loss": 1.18, "step": 5517 }, { "epoch": 1.7, "learning_rate": 1.8995184782561343e-05, "loss": 1.3453, "step": 5518 }, { "epoch": 1.7, "learning_rate": 1.8994749366753252e-05, "loss": 1.1343, "step": 5519 }, { "epoch": 1.7, "learning_rate": 1.899431386161901e-05, "loss": 1.1872, "step": 5520 }, { "epoch": 1.7, "learning_rate": 1.899387826716294e-05, "loss": 1.1344, "step": 5521 }, { "epoch": 1.7, "learning_rate": 1.8993442583389374e-05, "loss": 1.0894, "step": 5522 }, { "epoch": 1.7, "learning_rate": 1.8993006810302633e-05, "loss": 1.2043, "step": 5523 }, { "epoch": 1.7, "learning_rate": 1.8992570947907046e-05, "loss": 1.4119, "step": 5524 }, { "epoch": 1.7, "learning_rate": 1.899213499620694e-05, "loss": 1.2511, "step": 5525 }, { "epoch": 1.7, "learning_rate": 1.8991698955206645e-05, "loss": 1.1919, "step": 5526 }, { "epoch": 1.7, "learning_rate": 1.8991262824910498e-05, "loss": 1.4536, "step": 5527 }, { "epoch": 1.7, "learning_rate": 1.8990826605322818e-05, "loss": 1.2643, "step": 5528 }, { "epoch": 1.7, "learning_rate": 1.8990390296447948e-05, "loss": 1.091, "step": 5529 }, { "epoch": 1.7, "learning_rate": 1.8989953898290215e-05, "loss": 1.5458, "step": 5530 }, { "epoch": 1.7, "learning_rate": 1.8989517410853956e-05, "loss": 1.2833, "step": 5531 }, { "epoch": 1.7, "learning_rate": 1.8989080834143504e-05, "loss": 1.484, "step": 5532 }, { "epoch": 1.7, "learning_rate": 1.898864416816319e-05, "loss": 1.1092, "step": 5533 }, { "epoch": 1.7, "learning_rate": 1.898820741291736e-05, "loss": 0.9911, "step": 5534 }, { "epoch": 1.7, "learning_rate": 1.8987770568410346e-05, "loss": 0.9957, "step": 5535 }, { "epoch": 1.7, "learning_rate": 1.8987333634646487e-05, "loss": 1.217, "step": 5536 }, { "epoch": 1.7, "learning_rate": 1.898689661163012e-05, "loss": 1.4126, "step": 5537 }, { "epoch": 1.7, "learning_rate": 1.8986459499365587e-05, "loss": 1.194, "step": 5538 }, { "epoch": 1.7, "learning_rate": 1.8986022297857233e-05, "loss": 1.6199, "step": 5539 }, { "epoch": 1.7, "learning_rate": 1.898558500710939e-05, "loss": 1.2622, "step": 5540 }, { "epoch": 1.7, "learning_rate": 1.8985147627126414e-05, "loss": 1.1497, "step": 5541 }, { "epoch": 1.71, "learning_rate": 1.8984710157912632e-05, "loss": 1.0355, "step": 5542 }, { "epoch": 1.71, "learning_rate": 1.8984272599472404e-05, "loss": 1.1065, "step": 5543 }, { "epoch": 1.71, "learning_rate": 1.8983834951810068e-05, "loss": 0.9404, "step": 5544 }, { "epoch": 1.71, "learning_rate": 1.898339721492997e-05, "loss": 1.2294, "step": 5545 }, { "epoch": 1.71, "learning_rate": 1.898295938883646e-05, "loss": 1.2537, "step": 5546 }, { "epoch": 1.71, "learning_rate": 1.8982521473533882e-05, "loss": 1.2131, "step": 5547 }, { "epoch": 1.71, "learning_rate": 1.8982083469026587e-05, "loss": 1.1264, "step": 5548 }, { "epoch": 1.71, "learning_rate": 1.8981645375318928e-05, "loss": 1.2496, "step": 5549 }, { "epoch": 1.71, "learning_rate": 1.8981207192415253e-05, "loss": 1.3226, "step": 5550 }, { "epoch": 1.71, "learning_rate": 1.898076892031991e-05, "loss": 1.1554, "step": 5551 }, { "epoch": 1.71, "learning_rate": 1.8980330559037257e-05, "loss": 1.0462, "step": 5552 }, { "epoch": 1.71, "learning_rate": 1.8979892108571642e-05, "loss": 1.3695, "step": 5553 }, { "epoch": 1.71, "learning_rate": 1.8979453568927428e-05, "loss": 1.1727, "step": 5554 }, { "epoch": 1.71, "learning_rate": 1.8979014940108957e-05, "loss": 1.5765, "step": 5555 }, { "epoch": 1.71, "learning_rate": 1.8978576222120597e-05, "loss": 1.4364, "step": 5556 }, { "epoch": 1.71, "learning_rate": 1.89781374149667e-05, "loss": 1.0025, "step": 5557 }, { "epoch": 1.71, "learning_rate": 1.8977698518651624e-05, "loss": 1.203, "step": 5558 }, { "epoch": 1.71, "learning_rate": 1.8977259533179725e-05, "loss": 1.3586, "step": 5559 }, { "epoch": 1.71, "learning_rate": 1.8976820458555366e-05, "loss": 1.2311, "step": 5560 }, { "epoch": 1.71, "learning_rate": 1.8976381294782907e-05, "loss": 1.1518, "step": 5561 }, { "epoch": 1.71, "learning_rate": 1.897594204186671e-05, "loss": 1.0638, "step": 5562 }, { "epoch": 1.71, "learning_rate": 1.8975502699811137e-05, "loss": 1.2248, "step": 5563 }, { "epoch": 1.71, "learning_rate": 1.8975063268620547e-05, "loss": 1.3254, "step": 5564 }, { "epoch": 1.71, "learning_rate": 1.897462374829931e-05, "loss": 1.3484, "step": 5565 }, { "epoch": 1.71, "learning_rate": 1.897418413885179e-05, "loss": 1.3262, "step": 5566 }, { "epoch": 1.71, "learning_rate": 1.8973744440282344e-05, "loss": 1.4701, "step": 5567 }, { "epoch": 1.71, "learning_rate": 1.897330465259535e-05, "loss": 1.3526, "step": 5568 }, { "epoch": 1.71, "learning_rate": 1.8972864775795166e-05, "loss": 1.2633, "step": 5569 }, { "epoch": 1.71, "learning_rate": 1.897242480988617e-05, "loss": 1.3684, "step": 5570 }, { "epoch": 1.71, "learning_rate": 1.8971984754872724e-05, "loss": 1.4616, "step": 5571 }, { "epoch": 1.71, "learning_rate": 1.8971544610759202e-05, "loss": 1.1221, "step": 5572 }, { "epoch": 1.71, "learning_rate": 1.8971104377549974e-05, "loss": 1.2383, "step": 5573 }, { "epoch": 1.72, "learning_rate": 1.897066405524941e-05, "loss": 0.9827, "step": 5574 }, { "epoch": 1.72, "learning_rate": 1.8970223643861886e-05, "loss": 1.4238, "step": 5575 }, { "epoch": 1.72, "learning_rate": 1.896978314339177e-05, "loss": 1.1136, "step": 5576 }, { "epoch": 1.72, "learning_rate": 1.8969342553843443e-05, "loss": 0.8644, "step": 5577 }, { "epoch": 1.72, "learning_rate": 1.8968901875221282e-05, "loss": 1.3615, "step": 5578 }, { "epoch": 1.72, "learning_rate": 1.8968461107529653e-05, "loss": 0.9558, "step": 5579 }, { "epoch": 1.72, "learning_rate": 1.8968020250772944e-05, "loss": 1.3679, "step": 5580 }, { "epoch": 1.72, "learning_rate": 1.8967579304955527e-05, "loss": 0.9675, "step": 5581 }, { "epoch": 1.72, "learning_rate": 1.8967138270081782e-05, "loss": 1.0319, "step": 5582 }, { "epoch": 1.72, "learning_rate": 1.8966697146156092e-05, "loss": 1.3925, "step": 5583 }, { "epoch": 1.72, "learning_rate": 1.8966255933182834e-05, "loss": 1.2632, "step": 5584 }, { "epoch": 1.72, "learning_rate": 1.896581463116639e-05, "loss": 1.4041, "step": 5585 }, { "epoch": 1.72, "learning_rate": 1.8965373240111146e-05, "loss": 0.7759, "step": 5586 }, { "epoch": 1.72, "learning_rate": 1.896493176002148e-05, "loss": 0.979, "step": 5587 }, { "epoch": 1.72, "learning_rate": 1.8964490190901782e-05, "loss": 0.9705, "step": 5588 }, { "epoch": 1.72, "learning_rate": 1.896404853275643e-05, "loss": 1.1281, "step": 5589 }, { "epoch": 1.72, "learning_rate": 1.896360678558982e-05, "loss": 1.3731, "step": 5590 }, { "epoch": 1.72, "learning_rate": 1.896316494940633e-05, "loss": 1.2141, "step": 5591 }, { "epoch": 1.72, "learning_rate": 1.8962723024210357e-05, "loss": 1.1358, "step": 5592 }, { "epoch": 1.72, "learning_rate": 1.896228101000628e-05, "loss": 1.1985, "step": 5593 }, { "epoch": 1.72, "learning_rate": 1.896183890679849e-05, "loss": 1.005, "step": 5594 }, { "epoch": 1.72, "learning_rate": 1.8961396714591384e-05, "loss": 1.1082, "step": 5595 }, { "epoch": 1.72, "learning_rate": 1.896095443338935e-05, "loss": 1.2289, "step": 5596 }, { "epoch": 1.72, "learning_rate": 1.8960512063196774e-05, "loss": 1.3, "step": 5597 }, { "epoch": 1.72, "learning_rate": 1.896006960401806e-05, "loss": 1.8296, "step": 5598 }, { "epoch": 1.72, "learning_rate": 1.8959627055857596e-05, "loss": 1.4527, "step": 5599 }, { "epoch": 1.72, "learning_rate": 1.895918441871978e-05, "loss": 1.1792, "step": 5600 }, { "epoch": 1.72, "learning_rate": 1.8958741692609003e-05, "loss": 1.3502, "step": 5601 }, { "epoch": 1.72, "learning_rate": 1.8958298877529663e-05, "loss": 1.0676, "step": 5602 }, { "epoch": 1.72, "learning_rate": 1.895785597348616e-05, "loss": 1.4313, "step": 5603 }, { "epoch": 1.72, "learning_rate": 1.8957412980482895e-05, "loss": 1.4046, "step": 5604 }, { "epoch": 1.72, "learning_rate": 1.895696989852426e-05, "loss": 1.1929, "step": 5605 }, { "epoch": 1.72, "learning_rate": 1.8956526727614663e-05, "loss": 1.202, "step": 5606 }, { "epoch": 1.73, "learning_rate": 1.89560834677585e-05, "loss": 1.4496, "step": 5607 }, { "epoch": 1.73, "learning_rate": 1.8955640118960168e-05, "loss": 0.957, "step": 5608 }, { "epoch": 1.73, "learning_rate": 1.895519668122408e-05, "loss": 1.0511, "step": 5609 }, { "epoch": 1.73, "learning_rate": 1.8954753154554634e-05, "loss": 1.2144, "step": 5610 }, { "epoch": 1.73, "learning_rate": 1.895430953895624e-05, "loss": 1.2614, "step": 5611 }, { "epoch": 1.73, "learning_rate": 1.8953865834433297e-05, "loss": 0.9883, "step": 5612 }, { "epoch": 1.73, "learning_rate": 1.8953422040990212e-05, "loss": 1.1314, "step": 5613 }, { "epoch": 1.73, "learning_rate": 1.8952978158631398e-05, "loss": 0.9886, "step": 5614 }, { "epoch": 1.73, "learning_rate": 1.895253418736126e-05, "loss": 1.3012, "step": 5615 }, { "epoch": 1.73, "learning_rate": 1.8952090127184206e-05, "loss": 1.3758, "step": 5616 }, { "epoch": 1.73, "learning_rate": 1.8951645978104645e-05, "loss": 1.469, "step": 5617 }, { "epoch": 1.73, "learning_rate": 1.895120174012699e-05, "loss": 1.1743, "step": 5618 }, { "epoch": 1.73, "learning_rate": 1.8950757413255653e-05, "loss": 1.1235, "step": 5619 }, { "epoch": 1.73, "learning_rate": 1.8950312997495045e-05, "loss": 1.3884, "step": 5620 }, { "epoch": 1.73, "learning_rate": 1.894986849284958e-05, "loss": 1.3118, "step": 5621 }, { "epoch": 1.73, "learning_rate": 1.894942389932367e-05, "loss": 1.0397, "step": 5622 }, { "epoch": 1.73, "learning_rate": 1.8948979216921736e-05, "loss": 1.1979, "step": 5623 }, { "epoch": 1.73, "learning_rate": 1.8948534445648187e-05, "loss": 1.1458, "step": 5624 }, { "epoch": 1.73, "learning_rate": 1.894808958550745e-05, "loss": 1.1971, "step": 5625 }, { "epoch": 1.73, "learning_rate": 1.8947644636503933e-05, "loss": 1.0973, "step": 5626 }, { "epoch": 1.73, "learning_rate": 1.8947199598642058e-05, "loss": 1.0004, "step": 5627 }, { "epoch": 1.73, "learning_rate": 1.8946754471926243e-05, "loss": 1.137, "step": 5628 }, { "epoch": 1.73, "learning_rate": 1.8946309256360915e-05, "loss": 1.0146, "step": 5629 }, { "epoch": 1.73, "learning_rate": 1.8945863951950487e-05, "loss": 1.126, "step": 5630 }, { "epoch": 1.73, "learning_rate": 1.8945418558699387e-05, "loss": 0.8942, "step": 5631 }, { "epoch": 1.73, "learning_rate": 1.8944973076612038e-05, "loss": 1.3694, "step": 5632 }, { "epoch": 1.73, "learning_rate": 1.894452750569286e-05, "loss": 1.1193, "step": 5633 }, { "epoch": 1.73, "learning_rate": 1.894408184594628e-05, "loss": 1.0173, "step": 5634 }, { "epoch": 1.73, "learning_rate": 1.8943636097376728e-05, "loss": 1.3407, "step": 5635 }, { "epoch": 1.73, "learning_rate": 1.8943190259988623e-05, "loss": 1.1118, "step": 5636 }, { "epoch": 1.73, "learning_rate": 1.89427443337864e-05, "loss": 1.4162, "step": 5637 }, { "epoch": 1.73, "learning_rate": 1.894229831877448e-05, "loss": 1.2548, "step": 5638 }, { "epoch": 1.74, "learning_rate": 1.89418522149573e-05, "loss": 1.5327, "step": 5639 }, { "epoch": 1.74, "learning_rate": 1.8941406022339282e-05, "loss": 1.3282, "step": 5640 }, { "epoch": 1.74, "learning_rate": 1.8940959740924864e-05, "loss": 1.0271, "step": 5641 }, { "epoch": 1.74, "learning_rate": 1.8940513370718476e-05, "loss": 1.3634, "step": 5642 }, { "epoch": 1.74, "learning_rate": 1.894006691172455e-05, "loss": 1.3255, "step": 5643 }, { "epoch": 1.74, "learning_rate": 1.893962036394752e-05, "loss": 1.1152, "step": 5644 }, { "epoch": 1.74, "learning_rate": 1.893917372739182e-05, "loss": 1.3294, "step": 5645 }, { "epoch": 1.74, "learning_rate": 1.8938727002061888e-05, "loss": 1.1643, "step": 5646 }, { "epoch": 1.74, "learning_rate": 1.8938280187962154e-05, "loss": 1.0492, "step": 5647 }, { "epoch": 1.74, "learning_rate": 1.8937833285097067e-05, "loss": 1.3478, "step": 5648 }, { "epoch": 1.74, "learning_rate": 1.8937386293471052e-05, "loss": 1.3309, "step": 5649 }, { "epoch": 1.74, "learning_rate": 1.8936939213088558e-05, "loss": 1.0379, "step": 5650 }, { "epoch": 1.74, "learning_rate": 1.893649204395402e-05, "loss": 1.0785, "step": 5651 }, { "epoch": 1.74, "learning_rate": 1.893604478607188e-05, "loss": 1.1863, "step": 5652 }, { "epoch": 1.74, "learning_rate": 1.893559743944658e-05, "loss": 0.938, "step": 5653 }, { "epoch": 1.74, "learning_rate": 1.893515000408256e-05, "loss": 1.2678, "step": 5654 }, { "epoch": 1.74, "learning_rate": 1.893470247998427e-05, "loss": 1.0686, "step": 5655 }, { "epoch": 1.74, "learning_rate": 1.8934254867156146e-05, "loss": 1.0597, "step": 5656 }, { "epoch": 1.74, "learning_rate": 1.8933807165602636e-05, "loss": 1.0662, "step": 5657 }, { "epoch": 1.74, "learning_rate": 1.8933359375328192e-05, "loss": 1.1732, "step": 5658 }, { "epoch": 1.74, "learning_rate": 1.893291149633725e-05, "loss": 1.1579, "step": 5659 }, { "epoch": 1.74, "learning_rate": 1.893246352863427e-05, "loss": 0.9141, "step": 5660 }, { "epoch": 1.74, "learning_rate": 1.8932015472223692e-05, "loss": 1.3104, "step": 5661 }, { "epoch": 1.74, "learning_rate": 1.893156732710997e-05, "loss": 1.1048, "step": 5662 }, { "epoch": 1.74, "learning_rate": 1.893111909329755e-05, "loss": 1.3218, "step": 5663 }, { "epoch": 1.74, "learning_rate": 1.8930670770790888e-05, "loss": 1.2361, "step": 5664 }, { "epoch": 1.74, "learning_rate": 1.8930222359594435e-05, "loss": 0.944, "step": 5665 }, { "epoch": 1.74, "learning_rate": 1.8929773859712643e-05, "loss": 1.0178, "step": 5666 }, { "epoch": 1.74, "learning_rate": 1.8929325271149968e-05, "loss": 1.3979, "step": 5667 }, { "epoch": 1.74, "learning_rate": 1.892887659391086e-05, "loss": 1.3125, "step": 5668 }, { "epoch": 1.74, "learning_rate": 1.8928427827999782e-05, "loss": 1.3368, "step": 5669 }, { "epoch": 1.74, "learning_rate": 1.8927978973421185e-05, "loss": 1.013, "step": 5670 }, { "epoch": 1.74, "learning_rate": 1.892753003017953e-05, "loss": 1.1336, "step": 5671 }, { "epoch": 1.75, "learning_rate": 1.892708099827927e-05, "loss": 1.2953, "step": 5672 }, { "epoch": 1.75, "learning_rate": 1.8926631877724872e-05, "loss": 1.3589, "step": 5673 }, { "epoch": 1.75, "learning_rate": 1.8926182668520794e-05, "loss": 1.0823, "step": 5674 }, { "epoch": 1.75, "learning_rate": 1.8925733370671492e-05, "loss": 1.1865, "step": 5675 }, { "epoch": 1.75, "learning_rate": 1.8925283984181432e-05, "loss": 1.1582, "step": 5676 }, { "epoch": 1.75, "learning_rate": 1.8924834509055077e-05, "loss": 1.3544, "step": 5677 }, { "epoch": 1.75, "learning_rate": 1.892438494529689e-05, "loss": 1.2413, "step": 5678 }, { "epoch": 1.75, "learning_rate": 1.8923935292911335e-05, "loss": 1.0217, "step": 5679 }, { "epoch": 1.75, "learning_rate": 1.892348555190288e-05, "loss": 1.4613, "step": 5680 }, { "epoch": 1.75, "learning_rate": 1.892303572227599e-05, "loss": 1.0297, "step": 5681 }, { "epoch": 1.75, "learning_rate": 1.8922585804035125e-05, "loss": 1.1562, "step": 5682 }, { "epoch": 1.75, "learning_rate": 1.8922135797184766e-05, "loss": 1.1447, "step": 5683 }, { "epoch": 1.75, "learning_rate": 1.8921685701729373e-05, "loss": 1.2181, "step": 5684 }, { "epoch": 1.75, "learning_rate": 1.892123551767342e-05, "loss": 1.6407, "step": 5685 }, { "epoch": 1.75, "learning_rate": 1.8920785245021376e-05, "loss": 1.2482, "step": 5686 }, { "epoch": 1.75, "learning_rate": 1.892033488377771e-05, "loss": 1.2922, "step": 5687 }, { "epoch": 1.75, "learning_rate": 1.89198844339469e-05, "loss": 1.4085, "step": 5688 }, { "epoch": 1.75, "learning_rate": 1.8919433895533418e-05, "loss": 1.3497, "step": 5689 }, { "epoch": 1.75, "learning_rate": 1.8918983268541736e-05, "loss": 1.1493, "step": 5690 }, { "epoch": 1.75, "learning_rate": 1.8918532552976328e-05, "loss": 1.2189, "step": 5691 }, { "epoch": 1.75, "learning_rate": 1.8918081748841675e-05, "loss": 1.2046, "step": 5692 }, { "epoch": 1.75, "learning_rate": 1.891763085614225e-05, "loss": 1.2324, "step": 5693 }, { "epoch": 1.75, "learning_rate": 1.8917179874882528e-05, "loss": 1.4034, "step": 5694 }, { "epoch": 1.75, "learning_rate": 1.8916728805066996e-05, "loss": 1.295, "step": 5695 }, { "epoch": 1.75, "learning_rate": 1.891627764670013e-05, "loss": 1.34, "step": 5696 }, { "epoch": 1.75, "learning_rate": 1.8915826399786406e-05, "loss": 1.3665, "step": 5697 }, { "epoch": 1.75, "learning_rate": 1.8915375064330313e-05, "loss": 1.0947, "step": 5698 }, { "epoch": 1.75, "learning_rate": 1.8914923640336326e-05, "loss": 1.0893, "step": 5699 }, { "epoch": 1.75, "learning_rate": 1.891447212780893e-05, "loss": 1.189, "step": 5700 }, { "epoch": 1.75, "learning_rate": 1.8914020526752612e-05, "loss": 1.2564, "step": 5701 }, { "epoch": 1.75, "learning_rate": 1.8913568837171855e-05, "loss": 1.3463, "step": 5702 }, { "epoch": 1.75, "learning_rate": 1.8913117059071143e-05, "loss": 1.2773, "step": 5703 }, { "epoch": 1.76, "learning_rate": 1.8912665192454963e-05, "loss": 1.0019, "step": 5704 }, { "epoch": 1.76, "learning_rate": 1.8912213237327807e-05, "loss": 1.0782, "step": 5705 }, { "epoch": 1.76, "learning_rate": 1.891176119369416e-05, "loss": 1.2061, "step": 5706 }, { "epoch": 1.76, "learning_rate": 1.8911309061558507e-05, "loss": 1.1155, "step": 5707 }, { "epoch": 1.76, "learning_rate": 1.8910856840925343e-05, "loss": 1.3572, "step": 5708 }, { "epoch": 1.76, "learning_rate": 1.8910404531799162e-05, "loss": 1.2024, "step": 5709 }, { "epoch": 1.76, "learning_rate": 1.8909952134184448e-05, "loss": 1.0804, "step": 5710 }, { "epoch": 1.76, "learning_rate": 1.8909499648085697e-05, "loss": 1.0955, "step": 5711 }, { "epoch": 1.76, "learning_rate": 1.8909047073507406e-05, "loss": 1.297, "step": 5712 }, { "epoch": 1.76, "learning_rate": 1.8908594410454068e-05, "loss": 1.2556, "step": 5713 }, { "epoch": 1.76, "learning_rate": 1.8908141658930174e-05, "loss": 1.0759, "step": 5714 }, { "epoch": 1.76, "learning_rate": 1.8907688818940224e-05, "loss": 1.2264, "step": 5715 }, { "epoch": 1.76, "learning_rate": 1.8907235890488716e-05, "loss": 1.4589, "step": 5716 }, { "epoch": 1.76, "learning_rate": 1.8906782873580143e-05, "loss": 1.1195, "step": 5717 }, { "epoch": 1.76, "learning_rate": 1.890632976821901e-05, "loss": 1.0633, "step": 5718 }, { "epoch": 1.76, "learning_rate": 1.8905876574409815e-05, "loss": 1.2208, "step": 5719 }, { "epoch": 1.76, "learning_rate": 1.8905423292157056e-05, "loss": 1.3447, "step": 5720 }, { "epoch": 1.76, "learning_rate": 1.8904969921465234e-05, "loss": 1.3368, "step": 5721 }, { "epoch": 1.76, "learning_rate": 1.890451646233886e-05, "loss": 1.4269, "step": 5722 }, { "epoch": 1.76, "learning_rate": 1.8904062914782427e-05, "loss": 1.2346, "step": 5723 }, { "epoch": 1.76, "learning_rate": 1.890360927880044e-05, "loss": 1.3806, "step": 5724 }, { "epoch": 1.76, "learning_rate": 1.8903155554397412e-05, "loss": 0.9515, "step": 5725 }, { "epoch": 1.76, "learning_rate": 1.8902701741577844e-05, "loss": 1.2882, "step": 5726 }, { "epoch": 1.76, "learning_rate": 1.890224784034624e-05, "loss": 1.4086, "step": 5727 }, { "epoch": 1.76, "learning_rate": 1.8901793850707112e-05, "loss": 1.3845, "step": 5728 }, { "epoch": 1.76, "learning_rate": 1.8901339772664964e-05, "loss": 1.1041, "step": 5729 }, { "epoch": 1.76, "learning_rate": 1.8900885606224312e-05, "loss": 1.1908, "step": 5730 }, { "epoch": 1.76, "learning_rate": 1.8900431351389657e-05, "loss": 1.3916, "step": 5731 }, { "epoch": 1.76, "learning_rate": 1.8899977008165522e-05, "loss": 1.6902, "step": 5732 }, { "epoch": 1.76, "learning_rate": 1.889952257655641e-05, "loss": 1.356, "step": 5733 }, { "epoch": 1.76, "learning_rate": 1.889906805656684e-05, "loss": 1.2776, "step": 5734 }, { "epoch": 1.76, "learning_rate": 1.889861344820132e-05, "loss": 1.347, "step": 5735 }, { "epoch": 1.76, "learning_rate": 1.8898158751464368e-05, "loss": 1.1862, "step": 5736 }, { "epoch": 1.77, "learning_rate": 1.8897703966360498e-05, "loss": 1.2345, "step": 5737 }, { "epoch": 1.77, "learning_rate": 1.889724909289423e-05, "loss": 1.3768, "step": 5738 }, { "epoch": 1.77, "learning_rate": 1.8896794131070073e-05, "loss": 1.2098, "step": 5739 }, { "epoch": 1.77, "learning_rate": 1.8896339080892555e-05, "loss": 0.9148, "step": 5740 }, { "epoch": 1.77, "learning_rate": 1.889588394236619e-05, "loss": 1.0049, "step": 5741 }, { "epoch": 1.77, "learning_rate": 1.8895428715495495e-05, "loss": 1.6198, "step": 5742 }, { "epoch": 1.77, "learning_rate": 1.8894973400285e-05, "loss": 1.1361, "step": 5743 }, { "epoch": 1.77, "learning_rate": 1.8894517996739217e-05, "loss": 1.3155, "step": 5744 }, { "epoch": 1.77, "learning_rate": 1.8894062504862677e-05, "loss": 1.4628, "step": 5745 }, { "epoch": 1.77, "learning_rate": 1.8893606924659896e-05, "loss": 1.3148, "step": 5746 }, { "epoch": 1.77, "learning_rate": 1.8893151256135404e-05, "loss": 1.8015, "step": 5747 }, { "epoch": 1.77, "learning_rate": 1.8892695499293724e-05, "loss": 1.2297, "step": 5748 }, { "epoch": 1.77, "learning_rate": 1.889223965413938e-05, "loss": 1.482, "step": 5749 }, { "epoch": 1.77, "learning_rate": 1.8891783720676903e-05, "loss": 1.1685, "step": 5750 }, { "epoch": 1.77, "learning_rate": 1.889132769891082e-05, "loss": 1.1289, "step": 5751 }, { "epoch": 1.77, "learning_rate": 1.8890871588845653e-05, "loss": 1.376, "step": 5752 }, { "epoch": 1.77, "learning_rate": 1.8890415390485938e-05, "loss": 1.1207, "step": 5753 }, { "epoch": 1.77, "learning_rate": 1.8889959103836205e-05, "loss": 1.2448, "step": 5754 }, { "epoch": 1.77, "learning_rate": 1.888950272890099e-05, "loss": 1.4807, "step": 5755 }, { "epoch": 1.77, "learning_rate": 1.8889046265684817e-05, "loss": 1.1574, "step": 5756 }, { "epoch": 1.77, "learning_rate": 1.8888589714192222e-05, "loss": 1.3858, "step": 5757 }, { "epoch": 1.77, "learning_rate": 1.8888133074427738e-05, "loss": 1.4482, "step": 5758 }, { "epoch": 1.77, "learning_rate": 1.88876763463959e-05, "loss": 1.03, "step": 5759 }, { "epoch": 1.77, "learning_rate": 1.8887219530101245e-05, "loss": 1.1269, "step": 5760 }, { "epoch": 1.77, "learning_rate": 1.8886762625548315e-05, "loss": 1.2584, "step": 5761 }, { "epoch": 1.77, "learning_rate": 1.8886305632741638e-05, "loss": 1.2486, "step": 5762 }, { "epoch": 1.77, "learning_rate": 1.8885848551685755e-05, "loss": 1.3901, "step": 5763 }, { "epoch": 1.77, "learning_rate": 1.8885391382385207e-05, "loss": 0.9714, "step": 5764 }, { "epoch": 1.77, "learning_rate": 1.8884934124844534e-05, "loss": 1.3423, "step": 5765 }, { "epoch": 1.77, "learning_rate": 1.8884476779068277e-05, "loss": 1.1841, "step": 5766 }, { "epoch": 1.77, "learning_rate": 1.8884019345060978e-05, "loss": 1.0564, "step": 5767 }, { "epoch": 1.77, "learning_rate": 1.8883561822827178e-05, "loss": 1.1409, "step": 5768 }, { "epoch": 1.78, "learning_rate": 1.888310421237142e-05, "loss": 1.1158, "step": 5769 }, { "epoch": 1.78, "learning_rate": 1.8882646513698252e-05, "loss": 1.2759, "step": 5770 }, { "epoch": 1.78, "learning_rate": 1.8882188726812217e-05, "loss": 1.1919, "step": 5771 }, { "epoch": 1.78, "learning_rate": 1.8881730851717864e-05, "loss": 1.2857, "step": 5772 }, { "epoch": 1.78, "learning_rate": 1.8881272888419738e-05, "loss": 1.4416, "step": 5773 }, { "epoch": 1.78, "learning_rate": 1.8880814836922385e-05, "loss": 1.1659, "step": 5774 }, { "epoch": 1.78, "learning_rate": 1.888035669723036e-05, "loss": 1.7459, "step": 5775 }, { "epoch": 1.78, "learning_rate": 1.8879898469348203e-05, "loss": 1.9022, "step": 5776 }, { "epoch": 1.78, "learning_rate": 1.887944015328047e-05, "loss": 1.0215, "step": 5777 }, { "epoch": 1.78, "learning_rate": 1.8878981749031718e-05, "loss": 1.2745, "step": 5778 }, { "epoch": 1.78, "learning_rate": 1.887852325660649e-05, "loss": 1.4588, "step": 5779 }, { "epoch": 1.78, "learning_rate": 1.8878064676009344e-05, "loss": 1.2661, "step": 5780 }, { "epoch": 1.78, "learning_rate": 1.8877606007244835e-05, "loss": 1.0403, "step": 5781 }, { "epoch": 1.78, "learning_rate": 1.8877147250317516e-05, "loss": 0.989, "step": 5782 }, { "epoch": 1.78, "learning_rate": 1.8876688405231945e-05, "loss": 1.1391, "step": 5783 }, { "epoch": 1.78, "learning_rate": 1.8876229471992674e-05, "loss": 1.3808, "step": 5784 }, { "epoch": 1.78, "learning_rate": 1.8875770450604266e-05, "loss": 1.262, "step": 5785 }, { "epoch": 1.78, "learning_rate": 1.8875311341071273e-05, "loss": 1.1978, "step": 5786 }, { "epoch": 1.78, "learning_rate": 1.8874852143398263e-05, "loss": 1.1569, "step": 5787 }, { "epoch": 1.78, "learning_rate": 1.8874392857589788e-05, "loss": 1.1044, "step": 5788 }, { "epoch": 1.78, "learning_rate": 1.8873933483650418e-05, "loss": 1.2367, "step": 5789 }, { "epoch": 1.78, "learning_rate": 1.8873474021584707e-05, "loss": 1.2908, "step": 5790 }, { "epoch": 1.78, "learning_rate": 1.8873014471397225e-05, "loss": 1.5918, "step": 5791 }, { "epoch": 1.78, "learning_rate": 1.887255483309253e-05, "loss": 1.1515, "step": 5792 }, { "epoch": 1.78, "learning_rate": 1.8872095106675185e-05, "loss": 1.3371, "step": 5793 }, { "epoch": 1.78, "learning_rate": 1.887163529214976e-05, "loss": 1.1004, "step": 5794 }, { "epoch": 1.78, "learning_rate": 1.887117538952082e-05, "loss": 1.3325, "step": 5795 }, { "epoch": 1.78, "learning_rate": 1.8870715398792934e-05, "loss": 1.2156, "step": 5796 }, { "epoch": 1.78, "learning_rate": 1.8870255319970668e-05, "loss": 1.4961, "step": 5797 }, { "epoch": 1.78, "learning_rate": 1.8869795153058592e-05, "loss": 1.1313, "step": 5798 }, { "epoch": 1.78, "learning_rate": 1.8869334898061277e-05, "loss": 1.2681, "step": 5799 }, { "epoch": 1.78, "learning_rate": 1.886887455498329e-05, "loss": 1.1268, "step": 5800 }, { "epoch": 1.78, "learning_rate": 1.8868414123829206e-05, "loss": 1.346, "step": 5801 }, { "epoch": 1.79, "learning_rate": 1.8867953604603594e-05, "loss": 1.1131, "step": 5802 }, { "epoch": 1.79, "learning_rate": 1.886749299731103e-05, "loss": 1.1055, "step": 5803 }, { "epoch": 1.79, "learning_rate": 1.886703230195609e-05, "loss": 1.3157, "step": 5804 }, { "epoch": 1.79, "learning_rate": 1.8866571518543346e-05, "loss": 1.3374, "step": 5805 }, { "epoch": 1.79, "learning_rate": 1.8866110647077375e-05, "loss": 1.2593, "step": 5806 }, { "epoch": 1.79, "learning_rate": 1.8865649687562754e-05, "loss": 1.2124, "step": 5807 }, { "epoch": 1.79, "learning_rate": 1.8865188640004062e-05, "loss": 1.2001, "step": 5808 }, { "epoch": 1.79, "learning_rate": 1.8864727504405877e-05, "loss": 1.26, "step": 5809 }, { "epoch": 1.79, "learning_rate": 1.8864266280772776e-05, "loss": 1.1443, "step": 5810 }, { "epoch": 1.79, "learning_rate": 1.8863804969109343e-05, "loss": 1.4521, "step": 5811 }, { "epoch": 1.79, "learning_rate": 1.8863343569420156e-05, "loss": 1.1806, "step": 5812 }, { "epoch": 1.79, "learning_rate": 1.88628820817098e-05, "loss": 1.0273, "step": 5813 }, { "epoch": 1.79, "learning_rate": 1.8862420505982855e-05, "loss": 1.3477, "step": 5814 }, { "epoch": 1.79, "learning_rate": 1.8861958842243906e-05, "loss": 1.621, "step": 5815 }, { "epoch": 1.79, "learning_rate": 1.886149709049754e-05, "loss": 1.6195, "step": 5816 }, { "epoch": 1.79, "learning_rate": 1.8861035250748343e-05, "loss": 1.1612, "step": 5817 }, { "epoch": 1.79, "learning_rate": 1.8860573323000896e-05, "loss": 1.0728, "step": 5818 }, { "epoch": 1.79, "learning_rate": 1.8860111307259788e-05, "loss": 0.9489, "step": 5819 }, { "epoch": 1.79, "learning_rate": 1.8859649203529616e-05, "loss": 1.2121, "step": 5820 }, { "epoch": 1.79, "learning_rate": 1.8859187011814956e-05, "loss": 1.2022, "step": 5821 }, { "epoch": 1.79, "learning_rate": 1.8858724732120407e-05, "loss": 1.1058, "step": 5822 }, { "epoch": 1.79, "learning_rate": 1.8858262364450554e-05, "loss": 1.4858, "step": 5823 }, { "epoch": 1.79, "learning_rate": 1.8857799908809996e-05, "loss": 1.7402, "step": 5824 }, { "epoch": 1.79, "learning_rate": 1.8857337365203318e-05, "loss": 1.4368, "step": 5825 }, { "epoch": 1.79, "learning_rate": 1.8856874733635115e-05, "loss": 1.3462, "step": 5826 }, { "epoch": 1.79, "learning_rate": 1.8856412014109987e-05, "loss": 1.0406, "step": 5827 }, { "epoch": 1.79, "learning_rate": 1.8855949206632523e-05, "loss": 1.1073, "step": 5828 }, { "epoch": 1.79, "learning_rate": 1.8855486311207326e-05, "loss": 1.3543, "step": 5829 }, { "epoch": 1.79, "learning_rate": 1.8855023327838984e-05, "loss": 1.3314, "step": 5830 }, { "epoch": 1.79, "learning_rate": 1.8854560256532098e-05, "loss": 1.1792, "step": 5831 }, { "epoch": 1.79, "learning_rate": 1.8854097097291273e-05, "loss": 1.1188, "step": 5832 }, { "epoch": 1.79, "learning_rate": 1.8853633850121103e-05, "loss": 1.1029, "step": 5833 }, { "epoch": 1.8, "learning_rate": 1.8853170515026185e-05, "loss": 1.1273, "step": 5834 }, { "epoch": 1.8, "learning_rate": 1.8852707092011128e-05, "loss": 1.2635, "step": 5835 }, { "epoch": 1.8, "learning_rate": 1.885224358108053e-05, "loss": 1.2184, "step": 5836 }, { "epoch": 1.8, "learning_rate": 1.8851779982238993e-05, "loss": 1.4252, "step": 5837 }, { "epoch": 1.8, "learning_rate": 1.8851316295491124e-05, "loss": 1.5152, "step": 5838 }, { "epoch": 1.8, "learning_rate": 1.8850852520841526e-05, "loss": 1.2293, "step": 5839 }, { "epoch": 1.8, "learning_rate": 1.8850388658294808e-05, "loss": 0.9731, "step": 5840 }, { "epoch": 1.8, "learning_rate": 1.884992470785557e-05, "loss": 1.3046, "step": 5841 }, { "epoch": 1.8, "learning_rate": 1.8849460669528425e-05, "loss": 1.1865, "step": 5842 }, { "epoch": 1.8, "learning_rate": 1.8848996543317982e-05, "loss": 1.5167, "step": 5843 }, { "epoch": 1.8, "learning_rate": 1.8848532329228844e-05, "loss": 1.2405, "step": 5844 }, { "epoch": 1.8, "learning_rate": 1.8848068027265627e-05, "loss": 1.1002, "step": 5845 }, { "epoch": 1.8, "learning_rate": 1.8847603637432937e-05, "loss": 1.3105, "step": 5846 }, { "epoch": 1.8, "learning_rate": 1.8847139159735395e-05, "loss": 1.2938, "step": 5847 }, { "epoch": 1.8, "learning_rate": 1.8846674594177603e-05, "loss": 1.1335, "step": 5848 }, { "epoch": 1.8, "learning_rate": 1.8846209940764177e-05, "loss": 1.3542, "step": 5849 }, { "epoch": 1.8, "learning_rate": 1.8845745199499737e-05, "loss": 1.2121, "step": 5850 }, { "epoch": 1.8, "learning_rate": 1.8845280370388893e-05, "loss": 1.0904, "step": 5851 }, { "epoch": 1.8, "learning_rate": 1.8844815453436266e-05, "loss": 1.1116, "step": 5852 }, { "epoch": 1.8, "learning_rate": 1.8844350448646466e-05, "loss": 1.0829, "step": 5853 }, { "epoch": 1.8, "learning_rate": 1.8843885356024116e-05, "loss": 0.9803, "step": 5854 }, { "epoch": 1.8, "learning_rate": 1.884342017557383e-05, "loss": 1.3793, "step": 5855 }, { "epoch": 1.8, "learning_rate": 1.8842954907300236e-05, "loss": 1.3171, "step": 5856 }, { "epoch": 1.8, "learning_rate": 1.884248955120795e-05, "loss": 1.2279, "step": 5857 }, { "epoch": 1.8, "learning_rate": 1.8842024107301594e-05, "loss": 1.1557, "step": 5858 }, { "epoch": 1.8, "learning_rate": 1.8841558575585787e-05, "loss": 1.2331, "step": 5859 }, { "epoch": 1.8, "learning_rate": 1.8841092956065156e-05, "loss": 1.2517, "step": 5860 }, { "epoch": 1.8, "learning_rate": 1.8840627248744325e-05, "loss": 1.3514, "step": 5861 }, { "epoch": 1.8, "learning_rate": 1.8840161453627912e-05, "loss": 1.5597, "step": 5862 }, { "epoch": 1.8, "learning_rate": 1.8839695570720552e-05, "loss": 1.2498, "step": 5863 }, { "epoch": 1.8, "learning_rate": 1.8839229600026868e-05, "loss": 1.1637, "step": 5864 }, { "epoch": 1.8, "learning_rate": 1.883876354155149e-05, "loss": 1.2181, "step": 5865 }, { "epoch": 1.8, "learning_rate": 1.8838297395299042e-05, "loss": 1.2096, "step": 5866 }, { "epoch": 1.81, "learning_rate": 1.8837831161274158e-05, "loss": 1.6799, "step": 5867 }, { "epoch": 1.81, "learning_rate": 1.883736483948146e-05, "loss": 1.0675, "step": 5868 }, { "epoch": 1.81, "learning_rate": 1.8836898429925586e-05, "loss": 1.2894, "step": 5869 }, { "epoch": 1.81, "learning_rate": 1.883643193261117e-05, "loss": 1.3245, "step": 5870 }, { "epoch": 1.81, "learning_rate": 1.8835965347542834e-05, "loss": 1.3722, "step": 5871 }, { "epoch": 1.81, "learning_rate": 1.8835498674725227e-05, "loss": 1.3026, "step": 5872 }, { "epoch": 1.81, "learning_rate": 1.8835031914162968e-05, "loss": 1.2165, "step": 5873 }, { "epoch": 1.81, "learning_rate": 1.8834565065860703e-05, "loss": 1.3545, "step": 5874 }, { "epoch": 1.81, "learning_rate": 1.8834098129823064e-05, "loss": 1.1811, "step": 5875 }, { "epoch": 1.81, "learning_rate": 1.883363110605469e-05, "loss": 1.1224, "step": 5876 }, { "epoch": 1.81, "learning_rate": 1.8833163994560218e-05, "loss": 1.5397, "step": 5877 }, { "epoch": 1.81, "learning_rate": 1.8832696795344285e-05, "loss": 1.2625, "step": 5878 }, { "epoch": 1.81, "learning_rate": 1.883222950841153e-05, "loss": 1.3045, "step": 5879 }, { "epoch": 1.81, "learning_rate": 1.88317621337666e-05, "loss": 1.1038, "step": 5880 }, { "epoch": 1.81, "learning_rate": 1.8831294671414134e-05, "loss": 1.3045, "step": 5881 }, { "epoch": 1.81, "learning_rate": 1.883082712135877e-05, "loss": 1.2289, "step": 5882 }, { "epoch": 1.81, "learning_rate": 1.8830359483605153e-05, "loss": 1.0104, "step": 5883 }, { "epoch": 1.81, "learning_rate": 1.882989175815793e-05, "loss": 1.5085, "step": 5884 }, { "epoch": 1.81, "learning_rate": 1.882942394502174e-05, "loss": 1.3337, "step": 5885 }, { "epoch": 1.81, "learning_rate": 1.8828956044201237e-05, "loss": 1.4265, "step": 5886 }, { "epoch": 1.81, "learning_rate": 1.8828488055701064e-05, "loss": 1.0162, "step": 5887 }, { "epoch": 1.81, "learning_rate": 1.8828019979525867e-05, "loss": 1.283, "step": 5888 }, { "epoch": 1.81, "learning_rate": 1.8827551815680292e-05, "loss": 1.378, "step": 5889 }, { "epoch": 1.81, "learning_rate": 1.8827083564168996e-05, "loss": 1.1514, "step": 5890 }, { "epoch": 1.81, "learning_rate": 1.882661522499662e-05, "loss": 1.7426, "step": 5891 }, { "epoch": 1.81, "learning_rate": 1.882614679816783e-05, "loss": 1.2822, "step": 5892 }, { "epoch": 1.81, "learning_rate": 1.882567828368726e-05, "loss": 1.3073, "step": 5893 }, { "epoch": 1.81, "learning_rate": 1.8825209681559567e-05, "loss": 1.0657, "step": 5894 }, { "epoch": 1.81, "learning_rate": 1.8824740991789417e-05, "loss": 1.6406, "step": 5895 }, { "epoch": 1.81, "learning_rate": 1.8824272214381453e-05, "loss": 1.1522, "step": 5896 }, { "epoch": 1.81, "learning_rate": 1.882380334934033e-05, "loss": 1.1396, "step": 5897 }, { "epoch": 1.81, "learning_rate": 1.882333439667071e-05, "loss": 0.9923, "step": 5898 }, { "epoch": 1.82, "learning_rate": 1.8822865356377247e-05, "loss": 1.1394, "step": 5899 }, { "epoch": 1.82, "learning_rate": 1.88223962284646e-05, "loss": 1.1813, "step": 5900 }, { "epoch": 1.82, "learning_rate": 1.882192701293743e-05, "loss": 1.3706, "step": 5901 }, { "epoch": 1.82, "learning_rate": 1.882145770980039e-05, "loss": 1.0973, "step": 5902 }, { "epoch": 1.82, "learning_rate": 1.8820988319058147e-05, "loss": 1.2019, "step": 5903 }, { "epoch": 1.82, "learning_rate": 1.882051884071536e-05, "loss": 1.2544, "step": 5904 }, { "epoch": 1.82, "learning_rate": 1.8820049274776692e-05, "loss": 1.3162, "step": 5905 }, { "epoch": 1.82, "learning_rate": 1.8819579621246802e-05, "loss": 1.255, "step": 5906 }, { "epoch": 1.82, "learning_rate": 1.8819109880130365e-05, "loss": 1.7159, "step": 5907 }, { "epoch": 1.82, "learning_rate": 1.8818640051432036e-05, "loss": 1.1361, "step": 5908 }, { "epoch": 1.82, "learning_rate": 1.881817013515648e-05, "loss": 1.2042, "step": 5909 }, { "epoch": 1.82, "learning_rate": 1.8817700131308373e-05, "loss": 1.1527, "step": 5910 }, { "epoch": 1.82, "learning_rate": 1.8817230039892377e-05, "loss": 1.207, "step": 5911 }, { "epoch": 1.82, "learning_rate": 1.8816759860913158e-05, "loss": 1.208, "step": 5912 }, { "epoch": 1.82, "learning_rate": 1.8816289594375384e-05, "loss": 1.4967, "step": 5913 }, { "epoch": 1.82, "learning_rate": 1.8815819240283737e-05, "loss": 1.0583, "step": 5914 }, { "epoch": 1.82, "learning_rate": 1.8815348798642877e-05, "loss": 1.5494, "step": 5915 }, { "epoch": 1.82, "learning_rate": 1.8814878269457476e-05, "loss": 1.1214, "step": 5916 }, { "epoch": 1.82, "learning_rate": 1.8814407652732212e-05, "loss": 1.1997, "step": 5917 }, { "epoch": 1.82, "learning_rate": 1.8813936948471754e-05, "loss": 1.1964, "step": 5918 }, { "epoch": 1.82, "learning_rate": 1.8813466156680782e-05, "loss": 1.1966, "step": 5919 }, { "epoch": 1.82, "learning_rate": 1.8812995277363968e-05, "loss": 1.1495, "step": 5920 }, { "epoch": 1.82, "learning_rate": 1.881252431052599e-05, "loss": 1.3334, "step": 5921 }, { "epoch": 1.82, "learning_rate": 1.8812053256171522e-05, "loss": 1.8663, "step": 5922 }, { "epoch": 1.82, "learning_rate": 1.8811582114305245e-05, "loss": 1.2772, "step": 5923 }, { "epoch": 1.82, "learning_rate": 1.8811110884931835e-05, "loss": 0.9089, "step": 5924 }, { "epoch": 1.82, "learning_rate": 1.8810639568055974e-05, "loss": 1.3056, "step": 5925 }, { "epoch": 1.82, "learning_rate": 1.8810168163682338e-05, "loss": 1.2798, "step": 5926 }, { "epoch": 1.82, "learning_rate": 1.8809696671815616e-05, "loss": 1.0757, "step": 5927 }, { "epoch": 1.82, "learning_rate": 1.8809225092460488e-05, "loss": 1.1072, "step": 5928 }, { "epoch": 1.82, "learning_rate": 1.8808753425621635e-05, "loss": 1.1698, "step": 5929 }, { "epoch": 1.82, "learning_rate": 1.8808281671303744e-05, "loss": 1.0673, "step": 5930 }, { "epoch": 1.82, "learning_rate": 1.8807809829511496e-05, "loss": 1.1171, "step": 5931 }, { "epoch": 1.83, "learning_rate": 1.8807337900249578e-05, "loss": 1.1081, "step": 5932 }, { "epoch": 1.83, "learning_rate": 1.8806865883522678e-05, "loss": 1.2839, "step": 5933 }, { "epoch": 1.83, "learning_rate": 1.8806393779335483e-05, "loss": 1.1526, "step": 5934 }, { "epoch": 1.83, "learning_rate": 1.8805921587692682e-05, "loss": 1.174, "step": 5935 }, { "epoch": 1.83, "learning_rate": 1.8805449308598963e-05, "loss": 1.4581, "step": 5936 }, { "epoch": 1.83, "learning_rate": 1.880497694205902e-05, "loss": 1.2199, "step": 5937 }, { "epoch": 1.83, "learning_rate": 1.880450448807754e-05, "loss": 1.2219, "step": 5938 }, { "epoch": 1.83, "learning_rate": 1.8804031946659214e-05, "loss": 0.9132, "step": 5939 }, { "epoch": 1.83, "learning_rate": 1.8803559317808736e-05, "loss": 0.9531, "step": 5940 }, { "epoch": 1.83, "learning_rate": 1.8803086601530803e-05, "loss": 1.1006, "step": 5941 }, { "epoch": 1.83, "learning_rate": 1.8802613797830105e-05, "loss": 1.4266, "step": 5942 }, { "epoch": 1.83, "learning_rate": 1.880214090671134e-05, "loss": 1.3327, "step": 5943 }, { "epoch": 1.83, "learning_rate": 1.8801667928179205e-05, "loss": 1.1601, "step": 5944 }, { "epoch": 1.83, "learning_rate": 1.8801194862238394e-05, "loss": 1.1009, "step": 5945 }, { "epoch": 1.83, "learning_rate": 1.880072170889361e-05, "loss": 1.8478, "step": 5946 }, { "epoch": 1.83, "learning_rate": 1.8800248468149545e-05, "loss": 0.9635, "step": 5947 }, { "epoch": 1.83, "learning_rate": 1.87997751400109e-05, "loss": 1.282, "step": 5948 }, { "epoch": 1.83, "learning_rate": 1.879930172448238e-05, "loss": 1.106, "step": 5949 }, { "epoch": 1.83, "learning_rate": 1.8798828221568684e-05, "loss": 1.0259, "step": 5950 }, { "epoch": 1.83, "learning_rate": 1.8798354631274516e-05, "loss": 1.1757, "step": 5951 }, { "epoch": 1.83, "learning_rate": 1.879788095360458e-05, "loss": 1.0321, "step": 5952 }, { "epoch": 1.83, "learning_rate": 1.879740718856357e-05, "loss": 1.4116, "step": 5953 }, { "epoch": 1.83, "learning_rate": 1.8796933336156206e-05, "loss": 1.1292, "step": 5954 }, { "epoch": 1.83, "learning_rate": 1.8796459396387184e-05, "loss": 1.1254, "step": 5955 }, { "epoch": 1.83, "learning_rate": 1.879598536926121e-05, "loss": 1.0486, "step": 5956 }, { "epoch": 1.83, "learning_rate": 1.8795511254783e-05, "loss": 1.3186, "step": 5957 }, { "epoch": 1.83, "learning_rate": 1.8795037052957255e-05, "loss": 1.3214, "step": 5958 }, { "epoch": 1.83, "learning_rate": 1.8794562763788688e-05, "loss": 1.449, "step": 5959 }, { "epoch": 1.83, "learning_rate": 1.8794088387282e-05, "loss": 1.4245, "step": 5960 }, { "epoch": 1.83, "learning_rate": 1.8793613923441917e-05, "loss": 1.1856, "step": 5961 }, { "epoch": 1.83, "learning_rate": 1.879313937227314e-05, "loss": 1.1229, "step": 5962 }, { "epoch": 1.83, "learning_rate": 1.879266473378039e-05, "loss": 1.3943, "step": 5963 }, { "epoch": 1.84, "learning_rate": 1.879219000796837e-05, "loss": 1.0593, "step": 5964 }, { "epoch": 1.84, "learning_rate": 1.8791715194841802e-05, "loss": 1.1806, "step": 5965 }, { "epoch": 1.84, "learning_rate": 1.87912402944054e-05, "loss": 1.3716, "step": 5966 }, { "epoch": 1.84, "learning_rate": 1.8790765306663878e-05, "loss": 1.1382, "step": 5967 }, { "epoch": 1.84, "learning_rate": 1.8790290231621954e-05, "loss": 1.0203, "step": 5968 }, { "epoch": 1.84, "learning_rate": 1.878981506928435e-05, "loss": 1.2399, "step": 5969 }, { "epoch": 1.84, "learning_rate": 1.878933981965578e-05, "loss": 1.4157, "step": 5970 }, { "epoch": 1.84, "learning_rate": 1.8788864482740964e-05, "loss": 1.3412, "step": 5971 }, { "epoch": 1.84, "learning_rate": 1.8788389058544625e-05, "loss": 1.2215, "step": 5972 }, { "epoch": 1.84, "learning_rate": 1.8787913547071485e-05, "loss": 1.4236, "step": 5973 }, { "epoch": 1.84, "learning_rate": 1.878743794832626e-05, "loss": 1.0595, "step": 5974 }, { "epoch": 1.84, "learning_rate": 1.8786962262313677e-05, "loss": 1.0186, "step": 5975 }, { "epoch": 1.84, "learning_rate": 1.8786486489038466e-05, "loss": 1.3049, "step": 5976 }, { "epoch": 1.84, "learning_rate": 1.878601062850534e-05, "loss": 1.1868, "step": 5977 }, { "epoch": 1.84, "learning_rate": 1.8785534680719036e-05, "loss": 0.9719, "step": 5978 }, { "epoch": 1.84, "learning_rate": 1.8785058645684274e-05, "loss": 1.2152, "step": 5979 }, { "epoch": 1.84, "learning_rate": 1.8784582523405782e-05, "loss": 1.3481, "step": 5980 }, { "epoch": 1.84, "learning_rate": 1.878410631388829e-05, "loss": 1.106, "step": 5981 }, { "epoch": 1.84, "learning_rate": 1.8783630017136526e-05, "loss": 1.1508, "step": 5982 }, { "epoch": 1.84, "learning_rate": 1.878315363315522e-05, "loss": 1.1762, "step": 5983 }, { "epoch": 1.84, "learning_rate": 1.8782677161949107e-05, "loss": 1.3494, "step": 5984 }, { "epoch": 1.84, "learning_rate": 1.8782200603522913e-05, "loss": 1.0572, "step": 5985 }, { "epoch": 1.84, "learning_rate": 1.8781723957881374e-05, "loss": 1.0141, "step": 5986 }, { "epoch": 1.84, "learning_rate": 1.878124722502922e-05, "loss": 1.1646, "step": 5987 }, { "epoch": 1.84, "learning_rate": 1.878077040497119e-05, "loss": 1.2024, "step": 5988 }, { "epoch": 1.84, "learning_rate": 1.878029349771202e-05, "loss": 1.3746, "step": 5989 }, { "epoch": 1.84, "learning_rate": 1.877981650325644e-05, "loss": 1.1615, "step": 5990 }, { "epoch": 1.84, "learning_rate": 1.8779339421609194e-05, "loss": 0.9652, "step": 5991 }, { "epoch": 1.84, "learning_rate": 1.877886225277501e-05, "loss": 1.13, "step": 5992 }, { "epoch": 1.84, "learning_rate": 1.8778384996758637e-05, "loss": 1.4386, "step": 5993 }, { "epoch": 1.84, "learning_rate": 1.877790765356481e-05, "loss": 1.6565, "step": 5994 }, { "epoch": 1.84, "learning_rate": 1.8777430223198274e-05, "loss": 1.2639, "step": 5995 }, { "epoch": 1.84, "learning_rate": 1.877695270566376e-05, "loss": 0.8508, "step": 5996 }, { "epoch": 1.85, "learning_rate": 1.877647510096602e-05, "loss": 1.1837, "step": 5997 }, { "epoch": 1.85, "learning_rate": 1.877599740910979e-05, "loss": 1.0346, "step": 5998 }, { "epoch": 1.85, "learning_rate": 1.8775519630099822e-05, "loss": 1.3051, "step": 5999 }, { "epoch": 1.85, "learning_rate": 1.8775041763940853e-05, "loss": 1.4067, "step": 6000 }, { "epoch": 1.85, "learning_rate": 1.8774563810637636e-05, "loss": 1.0469, "step": 6001 }, { "epoch": 1.85, "learning_rate": 1.877408577019491e-05, "loss": 1.2621, "step": 6002 }, { "epoch": 1.85, "learning_rate": 1.8773607642617427e-05, "loss": 1.1981, "step": 6003 }, { "epoch": 1.85, "learning_rate": 1.8773129427909936e-05, "loss": 0.835, "step": 6004 }, { "epoch": 1.85, "learning_rate": 1.8772651126077178e-05, "loss": 1.2856, "step": 6005 }, { "epoch": 1.85, "learning_rate": 1.8772172737123913e-05, "loss": 1.1931, "step": 6006 }, { "epoch": 1.85, "learning_rate": 1.8771694261054892e-05, "loss": 1.132, "step": 6007 }, { "epoch": 1.85, "learning_rate": 1.8771215697874862e-05, "loss": 1.5031, "step": 6008 }, { "epoch": 1.85, "learning_rate": 1.877073704758857e-05, "loss": 0.9373, "step": 6009 }, { "epoch": 1.85, "learning_rate": 1.877025831020078e-05, "loss": 1.4372, "step": 6010 }, { "epoch": 1.85, "learning_rate": 1.876977948571624e-05, "loss": 1.3797, "step": 6011 }, { "epoch": 1.85, "learning_rate": 1.876930057413971e-05, "loss": 1.1841, "step": 6012 }, { "epoch": 1.85, "learning_rate": 1.8768821575475943e-05, "loss": 1.2869, "step": 6013 }, { "epoch": 1.85, "learning_rate": 1.8768342489729692e-05, "loss": 1.1641, "step": 6014 }, { "epoch": 1.85, "learning_rate": 1.8767863316905725e-05, "loss": 1.5404, "step": 6015 }, { "epoch": 1.85, "learning_rate": 1.8767384057008792e-05, "loss": 1.1628, "step": 6016 }, { "epoch": 1.85, "learning_rate": 1.8766904710043653e-05, "loss": 0.6712, "step": 6017 }, { "epoch": 1.85, "learning_rate": 1.8766425276015073e-05, "loss": 1.1542, "step": 6018 }, { "epoch": 1.85, "learning_rate": 1.876594575492781e-05, "loss": 1.1336, "step": 6019 }, { "epoch": 1.85, "learning_rate": 1.8765466146786628e-05, "loss": 1.3259, "step": 6020 }, { "epoch": 1.85, "learning_rate": 1.8764986451596286e-05, "loss": 1.1788, "step": 6021 }, { "epoch": 1.85, "learning_rate": 1.8764506669361552e-05, "loss": 1.5005, "step": 6022 }, { "epoch": 1.85, "learning_rate": 1.8764026800087186e-05, "loss": 1.0057, "step": 6023 }, { "epoch": 1.85, "learning_rate": 1.876354684377796e-05, "loss": 1.4307, "step": 6024 }, { "epoch": 1.85, "learning_rate": 1.8763066800438638e-05, "loss": 1.1305, "step": 6025 }, { "epoch": 1.85, "learning_rate": 1.8762586670073982e-05, "loss": 1.3708, "step": 6026 }, { "epoch": 1.85, "learning_rate": 1.876210645268877e-05, "loss": 1.2284, "step": 6027 }, { "epoch": 1.85, "learning_rate": 1.8761626148287762e-05, "loss": 1.1109, "step": 6028 }, { "epoch": 1.86, "learning_rate": 1.8761145756875735e-05, "loss": 1.2118, "step": 6029 }, { "epoch": 1.86, "learning_rate": 1.8760665278457452e-05, "loss": 1.1724, "step": 6030 }, { "epoch": 1.86, "learning_rate": 1.876018471303769e-05, "loss": 1.3702, "step": 6031 }, { "epoch": 1.86, "learning_rate": 1.875970406062122e-05, "loss": 1.1269, "step": 6032 }, { "epoch": 1.86, "learning_rate": 1.8759223321212818e-05, "loss": 1.3752, "step": 6033 }, { "epoch": 1.86, "learning_rate": 1.8758742494817253e-05, "loss": 1.067, "step": 6034 }, { "epoch": 1.86, "learning_rate": 1.87582615814393e-05, "loss": 1.0324, "step": 6035 }, { "epoch": 1.86, "learning_rate": 1.8757780581083742e-05, "loss": 1.1674, "step": 6036 }, { "epoch": 1.86, "learning_rate": 1.875729949375535e-05, "loss": 1.1796, "step": 6037 }, { "epoch": 1.86, "learning_rate": 1.875681831945891e-05, "loss": 1.5043, "step": 6038 }, { "epoch": 1.86, "learning_rate": 1.8756337058199183e-05, "loss": 1.3442, "step": 6039 }, { "epoch": 1.86, "learning_rate": 1.8755855709980965e-05, "loss": 0.9598, "step": 6040 }, { "epoch": 1.86, "learning_rate": 1.8755374274809027e-05, "loss": 1.7139, "step": 6041 }, { "epoch": 1.86, "learning_rate": 1.8754892752688155e-05, "loss": 1.5209, "step": 6042 }, { "epoch": 1.86, "learning_rate": 1.8754411143623125e-05, "loss": 1.1521, "step": 6043 }, { "epoch": 1.86, "learning_rate": 1.875392944761873e-05, "loss": 1.2314, "step": 6044 }, { "epoch": 1.86, "learning_rate": 1.8753447664679748e-05, "loss": 1.2894, "step": 6045 }, { "epoch": 1.86, "learning_rate": 1.8752965794810956e-05, "loss": 2.2167, "step": 6046 }, { "epoch": 1.86, "learning_rate": 1.8752483838017156e-05, "loss": 1.4829, "step": 6047 }, { "epoch": 1.86, "learning_rate": 1.8752001794303116e-05, "loss": 1.4072, "step": 6048 }, { "epoch": 1.86, "learning_rate": 1.875151966367364e-05, "loss": 1.1504, "step": 6049 }, { "epoch": 1.86, "learning_rate": 1.8751037446133502e-05, "loss": 1.1873, "step": 6050 }, { "epoch": 1.86, "learning_rate": 1.87505551416875e-05, "loss": 1.0045, "step": 6051 }, { "epoch": 1.86, "learning_rate": 1.875007275034042e-05, "loss": 1.3176, "step": 6052 }, { "epoch": 1.86, "learning_rate": 1.8749590272097052e-05, "loss": 1.4004, "step": 6053 }, { "epoch": 1.86, "learning_rate": 1.8749107706962194e-05, "loss": 1.1537, "step": 6054 }, { "epoch": 1.86, "learning_rate": 1.8748625054940624e-05, "loss": 1.0916, "step": 6055 }, { "epoch": 1.86, "learning_rate": 1.8748142316037153e-05, "loss": 1.3649, "step": 6056 }, { "epoch": 1.86, "learning_rate": 1.8747659490256562e-05, "loss": 1.0885, "step": 6057 }, { "epoch": 1.86, "learning_rate": 1.8747176577603652e-05, "loss": 1.0743, "step": 6058 }, { "epoch": 1.86, "learning_rate": 1.8746693578083217e-05, "loss": 1.1464, "step": 6059 }, { "epoch": 1.86, "learning_rate": 1.874621049170005e-05, "loss": 1.1459, "step": 6060 }, { "epoch": 1.86, "learning_rate": 1.8745727318458956e-05, "loss": 1.3947, "step": 6061 }, { "epoch": 1.87, "learning_rate": 1.8745244058364726e-05, "loss": 1.0067, "step": 6062 }, { "epoch": 1.87, "learning_rate": 1.8744760711422167e-05, "loss": 1.0945, "step": 6063 }, { "epoch": 1.87, "learning_rate": 1.874427727763607e-05, "loss": 1.0704, "step": 6064 }, { "epoch": 1.87, "learning_rate": 1.8743793757011244e-05, "loss": 1.3222, "step": 6065 }, { "epoch": 1.87, "learning_rate": 1.8743310149552487e-05, "loss": 1.154, "step": 6066 }, { "epoch": 1.87, "learning_rate": 1.87428264552646e-05, "loss": 1.4441, "step": 6067 }, { "epoch": 1.87, "learning_rate": 1.874234267415239e-05, "loss": 1.4023, "step": 6068 }, { "epoch": 1.87, "learning_rate": 1.8741858806220663e-05, "loss": 1.2405, "step": 6069 }, { "epoch": 1.87, "learning_rate": 1.8741374851474218e-05, "loss": 1.3357, "step": 6070 }, { "epoch": 1.87, "learning_rate": 1.8740890809917866e-05, "loss": 1.3455, "step": 6071 }, { "epoch": 1.87, "learning_rate": 1.8740406681556412e-05, "loss": 1.5587, "step": 6072 }, { "epoch": 1.87, "learning_rate": 1.873992246639466e-05, "loss": 1.5309, "step": 6073 }, { "epoch": 1.87, "learning_rate": 1.8739438164437433e-05, "loss": 1.2894, "step": 6074 }, { "epoch": 1.87, "learning_rate": 1.873895377568952e-05, "loss": 1.3334, "step": 6075 }, { "epoch": 1.87, "learning_rate": 1.873846930015575e-05, "loss": 1.2239, "step": 6076 }, { "epoch": 1.87, "learning_rate": 1.873798473784092e-05, "loss": 1.1775, "step": 6077 }, { "epoch": 1.87, "learning_rate": 1.873750008874985e-05, "loss": 1.0417, "step": 6078 }, { "epoch": 1.87, "learning_rate": 1.8737015352887354e-05, "loss": 1.3558, "step": 6079 }, { "epoch": 1.87, "learning_rate": 1.873653053025824e-05, "loss": 0.9908, "step": 6080 }, { "epoch": 1.87, "learning_rate": 1.873604562086733e-05, "loss": 1.0414, "step": 6081 }, { "epoch": 1.87, "learning_rate": 1.873556062471943e-05, "loss": 1.4357, "step": 6082 }, { "epoch": 1.87, "learning_rate": 1.8735075541819364e-05, "loss": 1.3635, "step": 6083 }, { "epoch": 1.87, "learning_rate": 1.8734590372171953e-05, "loss": 1.4315, "step": 6084 }, { "epoch": 1.87, "learning_rate": 1.8734105115782002e-05, "loss": 1.3719, "step": 6085 }, { "epoch": 1.87, "learning_rate": 1.8733619772654342e-05, "loss": 0.927, "step": 6086 }, { "epoch": 1.87, "learning_rate": 1.8733134342793788e-05, "loss": 0.9529, "step": 6087 }, { "epoch": 1.87, "learning_rate": 1.873264882620516e-05, "loss": 1.081, "step": 6088 }, { "epoch": 1.87, "learning_rate": 1.873216322289328e-05, "loss": 1.1299, "step": 6089 }, { "epoch": 1.87, "learning_rate": 1.8731677532862975e-05, "loss": 2.1402, "step": 6090 }, { "epoch": 1.87, "learning_rate": 1.873119175611906e-05, "loss": 1.1817, "step": 6091 }, { "epoch": 1.87, "learning_rate": 1.8730705892666367e-05, "loss": 1.2241, "step": 6092 }, { "epoch": 1.87, "learning_rate": 1.873021994250972e-05, "loss": 1.1897, "step": 6093 }, { "epoch": 1.88, "learning_rate": 1.872973390565394e-05, "loss": 1.0742, "step": 6094 }, { "epoch": 1.88, "learning_rate": 1.872924778210386e-05, "loss": 1.069, "step": 6095 }, { "epoch": 1.88, "learning_rate": 1.87287615718643e-05, "loss": 1.1691, "step": 6096 }, { "epoch": 1.88, "learning_rate": 1.8728275274940095e-05, "loss": 0.9923, "step": 6097 }, { "epoch": 1.88, "learning_rate": 1.872778889133607e-05, "loss": 0.8408, "step": 6098 }, { "epoch": 1.88, "learning_rate": 1.872730242105706e-05, "loss": 1.1825, "step": 6099 }, { "epoch": 1.88, "learning_rate": 1.8726815864107894e-05, "loss": 1.0143, "step": 6100 }, { "epoch": 1.88, "learning_rate": 1.8726329220493403e-05, "loss": 1.1306, "step": 6101 }, { "epoch": 1.88, "learning_rate": 1.872584249021842e-05, "loss": 1.4282, "step": 6102 }, { "epoch": 1.88, "learning_rate": 1.872535567328778e-05, "loss": 1.1651, "step": 6103 }, { "epoch": 1.88, "learning_rate": 1.8724868769706313e-05, "loss": 1.2825, "step": 6104 }, { "epoch": 1.88, "learning_rate": 1.8724381779478862e-05, "loss": 1.2024, "step": 6105 }, { "epoch": 1.88, "learning_rate": 1.8723894702610258e-05, "loss": 1.2551, "step": 6106 }, { "epoch": 1.88, "learning_rate": 1.872340753910534e-05, "loss": 1.3671, "step": 6107 }, { "epoch": 1.88, "learning_rate": 1.8722920288968947e-05, "loss": 1.6308, "step": 6108 }, { "epoch": 1.88, "learning_rate": 1.8722432952205914e-05, "loss": 1.3719, "step": 6109 }, { "epoch": 1.88, "learning_rate": 1.872194552882108e-05, "loss": 1.0633, "step": 6110 }, { "epoch": 1.88, "learning_rate": 1.8721458018819295e-05, "loss": 1.0118, "step": 6111 }, { "epoch": 1.88, "learning_rate": 1.8720970422205392e-05, "loss": 1.5955, "step": 6112 }, { "epoch": 1.88, "learning_rate": 1.8720482738984214e-05, "loss": 1.3909, "step": 6113 }, { "epoch": 1.88, "learning_rate": 1.8719994969160604e-05, "loss": 1.44, "step": 6114 }, { "epoch": 1.88, "learning_rate": 1.871950711273941e-05, "loss": 0.8714, "step": 6115 }, { "epoch": 1.88, "learning_rate": 1.871901916972547e-05, "loss": 1.2228, "step": 6116 }, { "epoch": 1.88, "learning_rate": 1.871853114012364e-05, "loss": 1.1317, "step": 6117 }, { "epoch": 1.88, "learning_rate": 1.8718043023938757e-05, "loss": 1.1401, "step": 6118 }, { "epoch": 1.88, "learning_rate": 1.8717554821175673e-05, "loss": 1.3029, "step": 6119 }, { "epoch": 1.88, "learning_rate": 1.8717066531839236e-05, "loss": 1.062, "step": 6120 }, { "epoch": 1.88, "learning_rate": 1.871657815593429e-05, "loss": 1.0814, "step": 6121 }, { "epoch": 1.88, "learning_rate": 1.8716089693465696e-05, "loss": 1.4719, "step": 6122 }, { "epoch": 1.88, "learning_rate": 1.8715601144438295e-05, "loss": 1.1842, "step": 6123 }, { "epoch": 1.88, "learning_rate": 1.871511250885694e-05, "loss": 1.0232, "step": 6124 }, { "epoch": 1.88, "learning_rate": 1.8714623786726487e-05, "loss": 1.0915, "step": 6125 }, { "epoch": 1.88, "learning_rate": 1.871413497805179e-05, "loss": 1.4538, "step": 6126 }, { "epoch": 1.89, "learning_rate": 1.87136460828377e-05, "loss": 1.1144, "step": 6127 }, { "epoch": 1.89, "learning_rate": 1.8713157101089074e-05, "loss": 1.0435, "step": 6128 }, { "epoch": 1.89, "learning_rate": 1.8712668032810767e-05, "loss": 0.965, "step": 6129 }, { "epoch": 1.89, "learning_rate": 1.8712178878007638e-05, "loss": 1.2034, "step": 6130 }, { "epoch": 1.89, "learning_rate": 1.8711689636684545e-05, "loss": 1.0622, "step": 6131 }, { "epoch": 1.89, "learning_rate": 1.871120030884634e-05, "loss": 0.9869, "step": 6132 }, { "epoch": 1.89, "learning_rate": 1.8710710894497893e-05, "loss": 1.2939, "step": 6133 }, { "epoch": 1.89, "learning_rate": 1.8710221393644054e-05, "loss": 1.1687, "step": 6134 }, { "epoch": 1.89, "learning_rate": 1.8709731806289693e-05, "loss": 1.2048, "step": 6135 }, { "epoch": 1.89, "learning_rate": 1.8709242132439665e-05, "loss": 1.3798, "step": 6136 }, { "epoch": 1.89, "learning_rate": 1.8708752372098838e-05, "loss": 1.3219, "step": 6137 }, { "epoch": 1.89, "learning_rate": 1.870826252527207e-05, "loss": 1.2545, "step": 6138 }, { "epoch": 1.89, "learning_rate": 1.8707772591964236e-05, "loss": 1.321, "step": 6139 }, { "epoch": 1.89, "learning_rate": 1.870728257218019e-05, "loss": 1.1698, "step": 6140 }, { "epoch": 1.89, "learning_rate": 1.8706792465924804e-05, "loss": 0.9488, "step": 6141 }, { "epoch": 1.89, "learning_rate": 1.870630227320294e-05, "loss": 1.3101, "step": 6142 }, { "epoch": 1.89, "learning_rate": 1.870581199401948e-05, "loss": 0.9985, "step": 6143 }, { "epoch": 1.89, "learning_rate": 1.8705321628379275e-05, "loss": 1.3676, "step": 6144 }, { "epoch": 1.89, "learning_rate": 1.8704831176287206e-05, "loss": 1.3072, "step": 6145 }, { "epoch": 1.89, "learning_rate": 1.8704340637748144e-05, "loss": 1.2828, "step": 6146 }, { "epoch": 1.89, "learning_rate": 1.8703850012766954e-05, "loss": 1.1684, "step": 6147 }, { "epoch": 1.89, "learning_rate": 1.8703359301348514e-05, "loss": 0.9803, "step": 6148 }, { "epoch": 1.89, "learning_rate": 1.870286850349769e-05, "loss": 1.1584, "step": 6149 }, { "epoch": 1.89, "learning_rate": 1.8702377619219366e-05, "loss": 1.1117, "step": 6150 }, { "epoch": 1.89, "learning_rate": 1.870188664851841e-05, "loss": 1.2507, "step": 6151 }, { "epoch": 1.89, "learning_rate": 1.87013955913997e-05, "loss": 1.0329, "step": 6152 }, { "epoch": 1.89, "learning_rate": 1.8700904447868112e-05, "loss": 1.192, "step": 6153 }, { "epoch": 1.89, "learning_rate": 1.8700413217928527e-05, "loss": 1.2365, "step": 6154 }, { "epoch": 1.89, "learning_rate": 1.8699921901585814e-05, "loss": 1.5139, "step": 6155 }, { "epoch": 1.89, "learning_rate": 1.8699430498844863e-05, "loss": 1.2603, "step": 6156 }, { "epoch": 1.89, "learning_rate": 1.8698939009710546e-05, "loss": 1.4836, "step": 6157 }, { "epoch": 1.89, "learning_rate": 1.8698447434187753e-05, "loss": 1.3196, "step": 6158 }, { "epoch": 1.9, "learning_rate": 1.8697955772281354e-05, "loss": 0.9478, "step": 6159 }, { "epoch": 1.9, "learning_rate": 1.869746402399624e-05, "loss": 0.7595, "step": 6160 }, { "epoch": 1.9, "learning_rate": 1.8696972189337296e-05, "loss": 1.0953, "step": 6161 }, { "epoch": 1.9, "learning_rate": 1.8696480268309397e-05, "loss": 1.2888, "step": 6162 }, { "epoch": 1.9, "learning_rate": 1.869598826091744e-05, "loss": 1.2963, "step": 6163 }, { "epoch": 1.9, "learning_rate": 1.86954961671663e-05, "loss": 1.0604, "step": 6164 }, { "epoch": 1.9, "learning_rate": 1.8695003987060874e-05, "loss": 1.5073, "step": 6165 }, { "epoch": 1.9, "learning_rate": 1.8694511720606038e-05, "loss": 1.0079, "step": 6166 }, { "epoch": 1.9, "learning_rate": 1.8694019367806696e-05, "loss": 1.5082, "step": 6167 }, { "epoch": 1.9, "learning_rate": 1.8693526928667724e-05, "loss": 1.4857, "step": 6168 }, { "epoch": 1.9, "learning_rate": 1.8693034403194017e-05, "loss": 1.5261, "step": 6169 }, { "epoch": 1.9, "learning_rate": 1.8692541791390465e-05, "loss": 1.1209, "step": 6170 }, { "epoch": 1.9, "learning_rate": 1.8692049093261967e-05, "loss": 1.0375, "step": 6171 }, { "epoch": 1.9, "learning_rate": 1.869155630881341e-05, "loss": 1.1637, "step": 6172 }, { "epoch": 1.9, "learning_rate": 1.8691063438049682e-05, "loss": 1.2004, "step": 6173 }, { "epoch": 1.9, "learning_rate": 1.869057048097569e-05, "loss": 1.3689, "step": 6174 }, { "epoch": 1.9, "learning_rate": 1.8690077437596322e-05, "loss": 1.166, "step": 6175 }, { "epoch": 1.9, "learning_rate": 1.8689584307916477e-05, "loss": 1.2447, "step": 6176 }, { "epoch": 1.9, "learning_rate": 1.8689091091941048e-05, "loss": 1.0398, "step": 6177 }, { "epoch": 1.9, "learning_rate": 1.8688597789674943e-05, "loss": 1.0298, "step": 6178 }, { "epoch": 1.9, "learning_rate": 1.868810440112305e-05, "loss": 0.9901, "step": 6179 }, { "epoch": 1.9, "learning_rate": 1.8687610926290272e-05, "loss": 1.168, "step": 6180 }, { "epoch": 1.9, "learning_rate": 1.8687117365181514e-05, "loss": 0.9922, "step": 6181 }, { "epoch": 1.9, "learning_rate": 1.868662371780167e-05, "loss": 1.1175, "step": 6182 }, { "epoch": 1.9, "learning_rate": 1.8686129984155648e-05, "loss": 1.4919, "step": 6183 }, { "epoch": 1.9, "learning_rate": 1.8685636164248353e-05, "loss": 1.3644, "step": 6184 }, { "epoch": 1.9, "learning_rate": 1.8685142258084684e-05, "loss": 1.2947, "step": 6185 }, { "epoch": 1.9, "learning_rate": 1.8684648265669548e-05, "loss": 1.0428, "step": 6186 }, { "epoch": 1.9, "learning_rate": 1.868415418700785e-05, "loss": 1.2124, "step": 6187 }, { "epoch": 1.9, "learning_rate": 1.8683660022104496e-05, "loss": 1.3051, "step": 6188 }, { "epoch": 1.9, "learning_rate": 1.8683165770964396e-05, "loss": 1.3293, "step": 6189 }, { "epoch": 1.9, "learning_rate": 1.868267143359246e-05, "loss": 1.4889, "step": 6190 }, { "epoch": 1.9, "learning_rate": 1.8682177009993592e-05, "loss": 1.1145, "step": 6191 }, { "epoch": 1.91, "learning_rate": 1.8681682500172703e-05, "loss": 1.5809, "step": 6192 }, { "epoch": 1.91, "learning_rate": 1.8681187904134712e-05, "loss": 1.6079, "step": 6193 }, { "epoch": 1.91, "learning_rate": 1.868069322188452e-05, "loss": 1.0679, "step": 6194 }, { "epoch": 1.91, "learning_rate": 1.8680198453427042e-05, "loss": 1.1076, "step": 6195 }, { "epoch": 1.91, "learning_rate": 1.8679703598767196e-05, "loss": 1.2864, "step": 6196 }, { "epoch": 1.91, "learning_rate": 1.867920865790989e-05, "loss": 0.8552, "step": 6197 }, { "epoch": 1.91, "learning_rate": 1.867871363086005e-05, "loss": 1.1333, "step": 6198 }, { "epoch": 1.91, "learning_rate": 1.867821851762258e-05, "loss": 1.1308, "step": 6199 }, { "epoch": 1.91, "learning_rate": 1.8677723318202403e-05, "loss": 1.3587, "step": 6200 }, { "epoch": 1.91, "learning_rate": 1.8677228032604437e-05, "loss": 1.403, "step": 6201 }, { "epoch": 1.91, "learning_rate": 1.86767326608336e-05, "loss": 1.0026, "step": 6202 }, { "epoch": 1.91, "learning_rate": 1.8676237202894807e-05, "loss": 1.1183, "step": 6203 }, { "epoch": 1.91, "learning_rate": 1.8675741658792984e-05, "loss": 2.007, "step": 6204 }, { "epoch": 1.91, "learning_rate": 1.8675246028533053e-05, "loss": 1.6377, "step": 6205 }, { "epoch": 1.91, "learning_rate": 1.867475031211993e-05, "loss": 2.8789, "step": 6206 }, { "epoch": 1.91, "learning_rate": 1.8674254509558544e-05, "loss": 1.9475, "step": 6207 }, { "epoch": 1.91, "learning_rate": 1.8673758620853818e-05, "loss": 1.091, "step": 6208 }, { "epoch": 1.91, "learning_rate": 1.8673262646010672e-05, "loss": 1.2191, "step": 6209 }, { "epoch": 1.91, "learning_rate": 1.8672766585034034e-05, "loss": 1.1453, "step": 6210 }, { "epoch": 1.91, "learning_rate": 1.8672270437928833e-05, "loss": 1.743, "step": 6211 }, { "epoch": 1.91, "learning_rate": 1.867177420469999e-05, "loss": 1.219, "step": 6212 }, { "epoch": 1.91, "learning_rate": 1.8671277885352442e-05, "loss": 1.3528, "step": 6213 }, { "epoch": 1.91, "learning_rate": 1.8670781479891113e-05, "loss": 1.4219, "step": 6214 }, { "epoch": 1.91, "learning_rate": 1.867028498832093e-05, "loss": 1.3894, "step": 6215 }, { "epoch": 1.91, "learning_rate": 1.8669788410646824e-05, "loss": 1.0281, "step": 6216 }, { "epoch": 1.91, "learning_rate": 1.8669291746873734e-05, "loss": 1.184, "step": 6217 }, { "epoch": 1.91, "learning_rate": 1.8668794997006582e-05, "loss": 1.1909, "step": 6218 }, { "epoch": 1.91, "learning_rate": 1.8668298161050308e-05, "loss": 1.3436, "step": 6219 }, { "epoch": 1.91, "learning_rate": 1.8667801239009845e-05, "loss": 1.2955, "step": 6220 }, { "epoch": 1.91, "learning_rate": 1.866730423089013e-05, "loss": 1.3049, "step": 6221 }, { "epoch": 1.91, "learning_rate": 1.8666807136696093e-05, "loss": 1.0393, "step": 6222 }, { "epoch": 1.91, "learning_rate": 1.8666309956432672e-05, "loss": 1.243, "step": 6223 }, { "epoch": 1.92, "learning_rate": 1.866581269010481e-05, "loss": 1.0693, "step": 6224 }, { "epoch": 1.92, "learning_rate": 1.866531533771744e-05, "loss": 1.7975, "step": 6225 }, { "epoch": 1.92, "learning_rate": 1.86648178992755e-05, "loss": 1.2894, "step": 6226 }, { "epoch": 1.92, "learning_rate": 1.8664320374783936e-05, "loss": 1.5022, "step": 6227 }, { "epoch": 1.92, "learning_rate": 1.8663822764247683e-05, "loss": 1.0991, "step": 6228 }, { "epoch": 1.92, "learning_rate": 1.8663325067671688e-05, "loss": 1.1788, "step": 6229 }, { "epoch": 1.92, "learning_rate": 1.8662827285060886e-05, "loss": 1.3557, "step": 6230 }, { "epoch": 1.92, "learning_rate": 1.866232941642023e-05, "loss": 1.3445, "step": 6231 }, { "epoch": 1.92, "learning_rate": 1.8661831461754658e-05, "loss": 1.1912, "step": 6232 }, { "epoch": 1.92, "learning_rate": 1.866133342106911e-05, "loss": 1.1641, "step": 6233 }, { "epoch": 1.92, "learning_rate": 1.8660835294368545e-05, "loss": 1.1892, "step": 6234 }, { "epoch": 1.92, "learning_rate": 1.8660337081657906e-05, "loss": 1.0504, "step": 6235 }, { "epoch": 1.92, "learning_rate": 1.8659838782942136e-05, "loss": 0.9675, "step": 6236 }, { "epoch": 1.92, "learning_rate": 1.8659340398226183e-05, "loss": 1.254, "step": 6237 }, { "epoch": 1.92, "learning_rate": 1.8658841927515e-05, "loss": 1.194, "step": 6238 }, { "epoch": 1.92, "learning_rate": 1.865834337081354e-05, "loss": 1.396, "step": 6239 }, { "epoch": 1.92, "learning_rate": 1.8657844728126747e-05, "loss": 1.5977, "step": 6240 }, { "epoch": 1.92, "learning_rate": 1.8657345999459577e-05, "loss": 1.2458, "step": 6241 }, { "epoch": 1.92, "learning_rate": 1.865684718481698e-05, "loss": 1.2064, "step": 6242 }, { "epoch": 1.92, "learning_rate": 1.865634828420392e-05, "loss": 1.2338, "step": 6243 }, { "epoch": 1.92, "learning_rate": 1.8655849297625337e-05, "loss": 1.2379, "step": 6244 }, { "epoch": 1.92, "learning_rate": 1.8655350225086196e-05, "loss": 1.4912, "step": 6245 }, { "epoch": 1.92, "learning_rate": 1.865485106659145e-05, "loss": 1.3138, "step": 6246 }, { "epoch": 1.92, "learning_rate": 1.865435182214605e-05, "loss": 1.1795, "step": 6247 }, { "epoch": 1.92, "learning_rate": 1.8653852491754966e-05, "loss": 1.2025, "step": 6248 }, { "epoch": 1.92, "learning_rate": 1.8653353075423152e-05, "loss": 1.5261, "step": 6249 }, { "epoch": 1.92, "learning_rate": 1.8652853573155564e-05, "loss": 1.0889, "step": 6250 }, { "epoch": 1.92, "learning_rate": 1.8652353984957168e-05, "loss": 1.1921, "step": 6251 }, { "epoch": 1.92, "learning_rate": 1.8651854310832924e-05, "loss": 1.305, "step": 6252 }, { "epoch": 1.92, "learning_rate": 1.865135455078779e-05, "loss": 1.2889, "step": 6253 }, { "epoch": 1.92, "learning_rate": 1.865085470482673e-05, "loss": 1.3401, "step": 6254 }, { "epoch": 1.92, "learning_rate": 1.8650354772954713e-05, "loss": 1.2016, "step": 6255 }, { "epoch": 1.92, "learning_rate": 1.86498547551767e-05, "loss": 1.1343, "step": 6256 }, { "epoch": 1.93, "learning_rate": 1.8649354651497657e-05, "loss": 1.3069, "step": 6257 }, { "epoch": 1.93, "learning_rate": 1.864885446192255e-05, "loss": 1.1775, "step": 6258 }, { "epoch": 1.93, "learning_rate": 1.864835418645635e-05, "loss": 1.2505, "step": 6259 }, { "epoch": 1.93, "learning_rate": 1.864785382510402e-05, "loss": 1.1091, "step": 6260 }, { "epoch": 1.93, "learning_rate": 1.864735337787053e-05, "loss": 1.3847, "step": 6261 }, { "epoch": 1.93, "learning_rate": 1.8646852844760857e-05, "loss": 1.2426, "step": 6262 }, { "epoch": 1.93, "learning_rate": 1.864635222577996e-05, "loss": 1.1603, "step": 6263 }, { "epoch": 1.93, "learning_rate": 1.864585152093282e-05, "loss": 1.261, "step": 6264 }, { "epoch": 1.93, "learning_rate": 1.8645350730224406e-05, "loss": 1.7334, "step": 6265 }, { "epoch": 1.93, "learning_rate": 1.864484985365969e-05, "loss": 1.2278, "step": 6266 }, { "epoch": 1.93, "learning_rate": 1.8644348891243647e-05, "loss": 1.3591, "step": 6267 }, { "epoch": 1.93, "learning_rate": 1.8643847842981254e-05, "loss": 1.1296, "step": 6268 }, { "epoch": 1.93, "learning_rate": 1.8643346708877487e-05, "loss": 1.2854, "step": 6269 }, { "epoch": 1.93, "learning_rate": 1.864284548893732e-05, "loss": 1.2737, "step": 6270 }, { "epoch": 1.93, "learning_rate": 1.864234418316573e-05, "loss": 1.0589, "step": 6271 }, { "epoch": 1.93, "learning_rate": 1.86418427915677e-05, "loss": 1.3824, "step": 6272 }, { "epoch": 1.93, "learning_rate": 1.8641341314148206e-05, "loss": 1.2414, "step": 6273 }, { "epoch": 1.93, "learning_rate": 1.864083975091223e-05, "loss": 1.2461, "step": 6274 }, { "epoch": 1.93, "learning_rate": 1.8640338101864747e-05, "loss": 1.3122, "step": 6275 }, { "epoch": 1.93, "learning_rate": 1.863983636701075e-05, "loss": 1.1551, "step": 6276 }, { "epoch": 1.93, "learning_rate": 1.8639334546355207e-05, "loss": 1.3417, "step": 6277 }, { "epoch": 1.93, "learning_rate": 1.8638832639903115e-05, "loss": 1.4944, "step": 6278 }, { "epoch": 1.93, "learning_rate": 1.8638330647659453e-05, "loss": 1.3161, "step": 6279 }, { "epoch": 1.93, "learning_rate": 1.8637828569629203e-05, "loss": 1.7466, "step": 6280 }, { "epoch": 1.93, "learning_rate": 1.863732640581736e-05, "loss": 1.2711, "step": 6281 }, { "epoch": 1.93, "learning_rate": 1.8636824156228905e-05, "loss": 1.1714, "step": 6282 }, { "epoch": 1.93, "learning_rate": 1.8636321820868822e-05, "loss": 1.4366, "step": 6283 }, { "epoch": 1.93, "learning_rate": 1.8635819399742103e-05, "loss": 1.3863, "step": 6284 }, { "epoch": 1.93, "learning_rate": 1.863531689285374e-05, "loss": 1.0436, "step": 6285 }, { "epoch": 1.93, "learning_rate": 1.8634814300208722e-05, "loss": 1.5671, "step": 6286 }, { "epoch": 1.93, "learning_rate": 1.8634311621812042e-05, "loss": 1.1413, "step": 6287 }, { "epoch": 1.93, "learning_rate": 1.8633808857668686e-05, "loss": 1.1636, "step": 6288 }, { "epoch": 1.94, "learning_rate": 1.8633306007783657e-05, "loss": 1.1356, "step": 6289 }, { "epoch": 1.94, "learning_rate": 1.863280307216194e-05, "loss": 0.9843, "step": 6290 }, { "epoch": 1.94, "learning_rate": 1.863230005080853e-05, "loss": 1.3341, "step": 6291 }, { "epoch": 1.94, "learning_rate": 1.8631796943728425e-05, "loss": 1.0566, "step": 6292 }, { "epoch": 1.94, "learning_rate": 1.8631293750926624e-05, "loss": 1.2812, "step": 6293 }, { "epoch": 1.94, "learning_rate": 1.8630790472408117e-05, "loss": 1.3345, "step": 6294 }, { "epoch": 1.94, "learning_rate": 1.863028710817791e-05, "loss": 1.1507, "step": 6295 }, { "epoch": 1.94, "learning_rate": 1.8629783658240995e-05, "loss": 1.1883, "step": 6296 }, { "epoch": 1.94, "learning_rate": 1.8629280122602374e-05, "loss": 1.5009, "step": 6297 }, { "epoch": 1.94, "learning_rate": 1.8628776501267052e-05, "loss": 1.2252, "step": 6298 }, { "epoch": 1.94, "learning_rate": 1.8628272794240026e-05, "loss": 0.9548, "step": 6299 }, { "epoch": 1.94, "learning_rate": 1.86277690015263e-05, "loss": 1.1128, "step": 6300 }, { "epoch": 1.94, "learning_rate": 1.8627265123130877e-05, "loss": 1.0562, "step": 6301 }, { "epoch": 1.94, "learning_rate": 1.8626761159058756e-05, "loss": 1.1999, "step": 6302 }, { "epoch": 1.94, "learning_rate": 1.8626257109314946e-05, "loss": 0.9332, "step": 6303 }, { "epoch": 1.94, "learning_rate": 1.8625752973904455e-05, "loss": 1.2425, "step": 6304 }, { "epoch": 1.94, "learning_rate": 1.862524875283229e-05, "loss": 1.0771, "step": 6305 }, { "epoch": 1.94, "learning_rate": 1.8624744446103453e-05, "loss": 1.2922, "step": 6306 }, { "epoch": 1.94, "learning_rate": 1.8624240053722953e-05, "loss": 1.9128, "step": 6307 }, { "epoch": 1.94, "learning_rate": 1.8623735575695804e-05, "loss": 1.412, "step": 6308 }, { "epoch": 1.94, "learning_rate": 1.862323101202701e-05, "loss": 1.1997, "step": 6309 }, { "epoch": 1.94, "learning_rate": 1.8622726362721587e-05, "loss": 1.2389, "step": 6310 }, { "epoch": 1.94, "learning_rate": 1.862222162778454e-05, "loss": 1.0568, "step": 6311 }, { "epoch": 1.94, "learning_rate": 1.862171680722089e-05, "loss": 1.2385, "step": 6312 }, { "epoch": 1.94, "learning_rate": 1.862121190103565e-05, "loss": 1.2663, "step": 6313 }, { "epoch": 1.94, "learning_rate": 1.862070690923382e-05, "loss": 1.2678, "step": 6314 }, { "epoch": 1.94, "learning_rate": 1.8620201831820433e-05, "loss": 1.3108, "step": 6315 }, { "epoch": 1.94, "learning_rate": 1.8619696668800494e-05, "loss": 1.503, "step": 6316 }, { "epoch": 1.94, "learning_rate": 1.8619191420179023e-05, "loss": 1.3892, "step": 6317 }, { "epoch": 1.94, "learning_rate": 1.861868608596104e-05, "loss": 1.8235, "step": 6318 }, { "epoch": 1.94, "learning_rate": 1.8618180666151558e-05, "loss": 1.6285, "step": 6319 }, { "epoch": 1.94, "learning_rate": 1.8617675160755603e-05, "loss": 1.4387, "step": 6320 }, { "epoch": 1.94, "learning_rate": 1.861716956977819e-05, "loss": 1.1077, "step": 6321 }, { "epoch": 1.95, "learning_rate": 1.8616663893224337e-05, "loss": 1.3553, "step": 6322 }, { "epoch": 1.95, "learning_rate": 1.8616158131099072e-05, "loss": 1.0607, "step": 6323 }, { "epoch": 1.95, "learning_rate": 1.861565228340742e-05, "loss": 1.1353, "step": 6324 }, { "epoch": 1.95, "learning_rate": 1.8615146350154398e-05, "loss": 1.0411, "step": 6325 }, { "epoch": 1.95, "learning_rate": 1.861464033134503e-05, "loss": 1.2034, "step": 6326 }, { "epoch": 1.95, "learning_rate": 1.8614134226984347e-05, "loss": 1.1045, "step": 6327 }, { "epoch": 1.95, "learning_rate": 1.861362803707737e-05, "loss": 1.681, "step": 6328 }, { "epoch": 1.95, "learning_rate": 1.8613121761629133e-05, "loss": 1.29, "step": 6329 }, { "epoch": 1.95, "learning_rate": 1.8612615400644656e-05, "loss": 1.337, "step": 6330 }, { "epoch": 1.95, "learning_rate": 1.861210895412897e-05, "loss": 1.387, "step": 6331 }, { "epoch": 1.95, "learning_rate": 1.8611602422087104e-05, "loss": 1.376, "step": 6332 }, { "epoch": 1.95, "learning_rate": 1.8611095804524092e-05, "loss": 1.3075, "step": 6333 }, { "epoch": 1.95, "learning_rate": 1.861058910144496e-05, "loss": 1.1277, "step": 6334 }, { "epoch": 1.95, "learning_rate": 1.8610082312854744e-05, "loss": 1.1256, "step": 6335 }, { "epoch": 1.95, "learning_rate": 1.8609575438758473e-05, "loss": 1.4351, "step": 6336 }, { "epoch": 1.95, "learning_rate": 1.8609068479161182e-05, "loss": 0.9759, "step": 6337 }, { "epoch": 1.95, "learning_rate": 1.8608561434067912e-05, "loss": 1.1365, "step": 6338 }, { "epoch": 1.95, "learning_rate": 1.860805430348369e-05, "loss": 1.5226, "step": 6339 }, { "epoch": 1.95, "learning_rate": 1.8607547087413556e-05, "loss": 0.9427, "step": 6340 }, { "epoch": 1.95, "learning_rate": 1.8607039785862545e-05, "loss": 1.5045, "step": 6341 }, { "epoch": 1.95, "learning_rate": 1.86065323988357e-05, "loss": 1.251, "step": 6342 }, { "epoch": 1.95, "learning_rate": 1.860602492633805e-05, "loss": 0.9935, "step": 6343 }, { "epoch": 1.95, "learning_rate": 1.8605517368374646e-05, "loss": 0.9891, "step": 6344 }, { "epoch": 1.95, "learning_rate": 1.8605009724950522e-05, "loss": 1.468, "step": 6345 }, { "epoch": 1.95, "learning_rate": 1.860450199607072e-05, "loss": 1.0235, "step": 6346 }, { "epoch": 1.95, "learning_rate": 1.8603994181740287e-05, "loss": 1.438, "step": 6347 }, { "epoch": 1.95, "learning_rate": 1.860348628196426e-05, "loss": 1.1493, "step": 6348 }, { "epoch": 1.95, "learning_rate": 1.8602978296747684e-05, "loss": 1.3455, "step": 6349 }, { "epoch": 1.95, "learning_rate": 1.8602470226095602e-05, "loss": 1.2091, "step": 6350 }, { "epoch": 1.95, "learning_rate": 1.8601962070013065e-05, "loss": 1.2086, "step": 6351 }, { "epoch": 1.95, "learning_rate": 1.8601453828505117e-05, "loss": 1.21, "step": 6352 }, { "epoch": 1.95, "learning_rate": 1.8600945501576807e-05, "loss": 1.4085, "step": 6353 }, { "epoch": 1.96, "learning_rate": 1.860043708923318e-05, "loss": 1.1391, "step": 6354 }, { "epoch": 1.96, "learning_rate": 1.8599928591479286e-05, "loss": 1.0764, "step": 6355 }, { "epoch": 1.96, "learning_rate": 1.8599420008320176e-05, "loss": 1.2909, "step": 6356 }, { "epoch": 1.96, "learning_rate": 1.8598911339760896e-05, "loss": 1.4003, "step": 6357 }, { "epoch": 1.96, "learning_rate": 1.8598402585806506e-05, "loss": 1.511, "step": 6358 }, { "epoch": 1.96, "learning_rate": 1.859789374646205e-05, "loss": 1.037, "step": 6359 }, { "epoch": 1.96, "learning_rate": 1.859738482173259e-05, "loss": 1.4354, "step": 6360 }, { "epoch": 1.96, "learning_rate": 1.859687581162317e-05, "loss": 0.9788, "step": 6361 }, { "epoch": 1.96, "learning_rate": 1.8596366716138853e-05, "loss": 1.0237, "step": 6362 }, { "epoch": 1.96, "learning_rate": 1.8595857535284692e-05, "loss": 1.2233, "step": 6363 }, { "epoch": 1.96, "learning_rate": 1.8595348269065744e-05, "loss": 1.3469, "step": 6364 }, { "epoch": 1.96, "learning_rate": 1.8594838917487065e-05, "loss": 1.0047, "step": 6365 }, { "epoch": 1.96, "learning_rate": 1.8594329480553712e-05, "loss": 1.4814, "step": 6366 }, { "epoch": 1.96, "learning_rate": 1.859381995827075e-05, "loss": 1.3296, "step": 6367 }, { "epoch": 1.96, "learning_rate": 1.8593310350643232e-05, "loss": 1.2401, "step": 6368 }, { "epoch": 1.96, "learning_rate": 1.8592800657676227e-05, "loss": 1.4951, "step": 6369 }, { "epoch": 1.96, "learning_rate": 1.8592290879374787e-05, "loss": 1.125, "step": 6370 }, { "epoch": 1.96, "learning_rate": 1.8591781015743983e-05, "loss": 1.8767, "step": 6371 }, { "epoch": 1.96, "learning_rate": 1.8591271066788877e-05, "loss": 1.06, "step": 6372 }, { "epoch": 1.96, "learning_rate": 1.8590761032514527e-05, "loss": 1.286, "step": 6373 }, { "epoch": 1.96, "learning_rate": 1.8590250912926002e-05, "loss": 1.2805, "step": 6374 }, { "epoch": 1.96, "learning_rate": 1.858974070802837e-05, "loss": 1.251, "step": 6375 }, { "epoch": 1.96, "learning_rate": 1.85892304178267e-05, "loss": 1.1862, "step": 6376 }, { "epoch": 1.96, "learning_rate": 1.8588720042326052e-05, "loss": 1.0989, "step": 6377 }, { "epoch": 1.96, "learning_rate": 1.85882095815315e-05, "loss": 1.322, "step": 6378 }, { "epoch": 1.96, "learning_rate": 1.8587699035448107e-05, "loss": 1.4631, "step": 6379 }, { "epoch": 1.96, "learning_rate": 1.8587188404080955e-05, "loss": 0.9686, "step": 6380 }, { "epoch": 1.96, "learning_rate": 1.8586677687435104e-05, "loss": 1.2665, "step": 6381 }, { "epoch": 1.96, "learning_rate": 1.858616688551563e-05, "loss": 1.2313, "step": 6382 }, { "epoch": 1.96, "learning_rate": 1.8585655998327605e-05, "loss": 1.3366, "step": 6383 }, { "epoch": 1.96, "learning_rate": 1.8585145025876107e-05, "loss": 1.2932, "step": 6384 }, { "epoch": 1.96, "learning_rate": 1.8584633968166203e-05, "loss": 1.2454, "step": 6385 }, { "epoch": 1.96, "learning_rate": 1.8584122825202974e-05, "loss": 1.1501, "step": 6386 }, { "epoch": 1.97, "learning_rate": 1.858361159699149e-05, "loss": 0.8284, "step": 6387 }, { "epoch": 1.97, "learning_rate": 1.8583100283536837e-05, "loss": 1.5539, "step": 6388 }, { "epoch": 1.97, "learning_rate": 1.8582588884844086e-05, "loss": 0.8372, "step": 6389 }, { "epoch": 1.97, "learning_rate": 1.8582077400918316e-05, "loss": 1.0311, "step": 6390 }, { "epoch": 1.97, "learning_rate": 1.858156583176461e-05, "loss": 1.2256, "step": 6391 }, { "epoch": 1.97, "learning_rate": 1.8581054177388044e-05, "loss": 1.3272, "step": 6392 }, { "epoch": 1.97, "learning_rate": 1.8580542437793703e-05, "loss": 2.186, "step": 6393 }, { "epoch": 1.97, "learning_rate": 1.8580030612986665e-05, "loss": 1.1764, "step": 6394 }, { "epoch": 1.97, "learning_rate": 1.857951870297202e-05, "loss": 1.079, "step": 6395 }, { "epoch": 1.97, "learning_rate": 1.8579006707754845e-05, "loss": 1.4945, "step": 6396 }, { "epoch": 1.97, "learning_rate": 1.8578494627340223e-05, "loss": 1.0618, "step": 6397 }, { "epoch": 1.97, "learning_rate": 1.857798246173325e-05, "loss": 0.9559, "step": 6398 }, { "epoch": 1.97, "learning_rate": 1.8577470210938998e-05, "loss": 1.1756, "step": 6399 }, { "epoch": 1.97, "learning_rate": 1.857695787496257e-05, "loss": 1.2188, "step": 6400 }, { "epoch": 1.97, "learning_rate": 1.857644545380904e-05, "loss": 1.0355, "step": 6401 }, { "epoch": 1.97, "learning_rate": 1.8575932947483503e-05, "loss": 1.0344, "step": 6402 }, { "epoch": 1.97, "learning_rate": 1.8575420355991047e-05, "loss": 1.0885, "step": 6403 }, { "epoch": 1.97, "learning_rate": 1.8574907679336768e-05, "loss": 1.4063, "step": 6404 }, { "epoch": 1.97, "learning_rate": 1.857439491752575e-05, "loss": 1.4525, "step": 6405 }, { "epoch": 1.97, "learning_rate": 1.857388207056309e-05, "loss": 1.4635, "step": 6406 }, { "epoch": 1.97, "learning_rate": 1.8573369138453874e-05, "loss": 0.9788, "step": 6407 }, { "epoch": 1.97, "learning_rate": 1.8572856121203204e-05, "loss": 1.0531, "step": 6408 }, { "epoch": 1.97, "learning_rate": 1.8572343018816172e-05, "loss": 1.2602, "step": 6409 }, { "epoch": 1.97, "learning_rate": 1.8571829831297874e-05, "loss": 1.3109, "step": 6410 }, { "epoch": 1.97, "learning_rate": 1.8571316558653404e-05, "loss": 1.1934, "step": 6411 }, { "epoch": 1.97, "learning_rate": 1.857080320088786e-05, "loss": 1.3762, "step": 6412 }, { "epoch": 1.97, "learning_rate": 1.8570289758006346e-05, "loss": 1.2031, "step": 6413 }, { "epoch": 1.97, "learning_rate": 1.8569776230013952e-05, "loss": 1.0772, "step": 6414 }, { "epoch": 1.97, "learning_rate": 1.8569262616915784e-05, "loss": 1.0231, "step": 6415 }, { "epoch": 1.97, "learning_rate": 1.8568748918716938e-05, "loss": 1.1101, "step": 6416 }, { "epoch": 1.97, "learning_rate": 1.856823513542252e-05, "loss": 1.8419, "step": 6417 }, { "epoch": 1.97, "learning_rate": 1.856772126703763e-05, "loss": 1.0359, "step": 6418 }, { "epoch": 1.98, "learning_rate": 1.8567207313567372e-05, "loss": 1.1589, "step": 6419 }, { "epoch": 1.98, "learning_rate": 1.856669327501685e-05, "loss": 1.1152, "step": 6420 }, { "epoch": 1.98, "learning_rate": 1.856617915139117e-05, "loss": 1.417, "step": 6421 }, { "epoch": 1.98, "learning_rate": 1.856566494269543e-05, "loss": 1.2729, "step": 6422 }, { "epoch": 1.98, "learning_rate": 1.856515064893475e-05, "loss": 1.1809, "step": 6423 }, { "epoch": 1.98, "learning_rate": 1.8564636270114226e-05, "loss": 1.0151, "step": 6424 }, { "epoch": 1.98, "learning_rate": 1.856412180623897e-05, "loss": 0.884, "step": 6425 }, { "epoch": 1.98, "learning_rate": 1.8563607257314093e-05, "loss": 1.0398, "step": 6426 }, { "epoch": 1.98, "learning_rate": 1.8563092623344702e-05, "loss": 1.4843, "step": 6427 }, { "epoch": 1.98, "learning_rate": 1.8562577904335913e-05, "loss": 1.2468, "step": 6428 }, { "epoch": 1.98, "learning_rate": 1.856206310029283e-05, "loss": 1.0299, "step": 6429 }, { "epoch": 1.98, "learning_rate": 1.8561548211220574e-05, "loss": 1.3006, "step": 6430 }, { "epoch": 1.98, "learning_rate": 1.856103323712425e-05, "loss": 1.3749, "step": 6431 }, { "epoch": 1.98, "learning_rate": 1.8560518178008974e-05, "loss": 1.2642, "step": 6432 }, { "epoch": 1.98, "learning_rate": 1.8560003033879868e-05, "loss": 1.4633, "step": 6433 }, { "epoch": 1.98, "learning_rate": 1.8559487804742037e-05, "loss": 0.9442, "step": 6434 }, { "epoch": 1.98, "learning_rate": 1.855897249060061e-05, "loss": 1.2721, "step": 6435 }, { "epoch": 1.98, "learning_rate": 1.8558457091460694e-05, "loss": 1.1578, "step": 6436 }, { "epoch": 1.98, "learning_rate": 1.855794160732741e-05, "loss": 1.57, "step": 6437 }, { "epoch": 1.98, "learning_rate": 1.8557426038205883e-05, "loss": 1.0041, "step": 6438 }, { "epoch": 1.98, "learning_rate": 1.8556910384101224e-05, "loss": 1.2846, "step": 6439 }, { "epoch": 1.98, "learning_rate": 1.8556394645018562e-05, "loss": 1.4396, "step": 6440 }, { "epoch": 1.98, "learning_rate": 1.8555878820963014e-05, "loss": 1.1006, "step": 6441 }, { "epoch": 1.98, "learning_rate": 1.8555362911939704e-05, "loss": 0.966, "step": 6442 }, { "epoch": 1.98, "learning_rate": 1.8554846917953755e-05, "loss": 1.0887, "step": 6443 }, { "epoch": 1.98, "learning_rate": 1.8554330839010293e-05, "loss": 1.063, "step": 6444 }, { "epoch": 1.98, "learning_rate": 1.855381467511444e-05, "loss": 2.2703, "step": 6445 }, { "epoch": 1.98, "learning_rate": 1.8553298426271328e-05, "loss": 0.8687, "step": 6446 }, { "epoch": 1.98, "learning_rate": 1.8552782092486074e-05, "loss": 1.3303, "step": 6447 }, { "epoch": 1.98, "learning_rate": 1.8552265673763816e-05, "loss": 1.577, "step": 6448 }, { "epoch": 1.98, "learning_rate": 1.8551749170109676e-05, "loss": 1.2445, "step": 6449 }, { "epoch": 1.98, "learning_rate": 1.8551232581528783e-05, "loss": 0.9764, "step": 6450 }, { "epoch": 1.98, "learning_rate": 1.8550715908026275e-05, "loss": 1.1028, "step": 6451 }, { "epoch": 1.99, "learning_rate": 1.8550199149607274e-05, "loss": 0.9804, "step": 6452 }, { "epoch": 1.99, "learning_rate": 1.8549682306276912e-05, "loss": 1.2268, "step": 6453 }, { "epoch": 1.99, "learning_rate": 1.8549165378040328e-05, "loss": 1.2428, "step": 6454 }, { "epoch": 1.99, "learning_rate": 1.8548648364902656e-05, "loss": 1.2157, "step": 6455 }, { "epoch": 1.99, "learning_rate": 1.8548131266869024e-05, "loss": 1.357, "step": 6456 }, { "epoch": 1.99, "learning_rate": 1.854761408394457e-05, "loss": 1.4042, "step": 6457 }, { "epoch": 1.99, "learning_rate": 1.8547096816134432e-05, "loss": 1.0632, "step": 6458 }, { "epoch": 1.99, "learning_rate": 1.8546579463443742e-05, "loss": 1.1182, "step": 6459 }, { "epoch": 1.99, "learning_rate": 1.8546062025877644e-05, "loss": 1.2198, "step": 6460 }, { "epoch": 1.99, "learning_rate": 1.8545544503441274e-05, "loss": 1.1667, "step": 6461 }, { "epoch": 1.99, "learning_rate": 1.854502689613977e-05, "loss": 1.0328, "step": 6462 }, { "epoch": 1.99, "learning_rate": 1.8544509203978273e-05, "loss": 1.3525, "step": 6463 }, { "epoch": 1.99, "learning_rate": 1.854399142696193e-05, "loss": 1.3158, "step": 6464 }, { "epoch": 1.99, "learning_rate": 1.854347356509587e-05, "loss": 1.3576, "step": 6465 }, { "epoch": 1.99, "learning_rate": 1.8542955618385248e-05, "loss": 1.2279, "step": 6466 }, { "epoch": 1.99, "learning_rate": 1.8542437586835202e-05, "loss": 1.2929, "step": 6467 }, { "epoch": 1.99, "learning_rate": 1.854191947045088e-05, "loss": 1.319, "step": 6468 }, { "epoch": 1.99, "learning_rate": 1.854191947045088e-05, "loss": 1.3244, "step": 6469 }, { "epoch": 1.99, "learning_rate": 1.8541401269237426e-05, "loss": 1.0706, "step": 6470 }, { "epoch": 1.99, "learning_rate": 1.854088298319998e-05, "loss": 1.4506, "step": 6471 }, { "epoch": 1.99, "learning_rate": 1.8540364612343697e-05, "loss": 1.2266, "step": 6472 }, { "epoch": 1.99, "learning_rate": 1.8539846156673723e-05, "loss": 1.0823, "step": 6473 }, { "epoch": 1.99, "learning_rate": 1.8539327616195206e-05, "loss": 1.2402, "step": 6474 }, { "epoch": 1.99, "learning_rate": 1.85388089909133e-05, "loss": 1.2059, "step": 6475 }, { "epoch": 1.99, "learning_rate": 1.853829028083314e-05, "loss": 1.5681, "step": 6476 }, { "epoch": 1.99, "learning_rate": 1.85377714859599e-05, "loss": 1.2445, "step": 6477 }, { "epoch": 1.99, "learning_rate": 1.8537252606298714e-05, "loss": 1.1561, "step": 6478 }, { "epoch": 1.99, "learning_rate": 1.8536733641854742e-05, "loss": 1.3023, "step": 6479 }, { "epoch": 1.99, "learning_rate": 1.8536214592633136e-05, "loss": 1.1537, "step": 6480 }, { "epoch": 1.99, "learning_rate": 1.8535695458639056e-05, "loss": 1.3162, "step": 6481 }, { "epoch": 1.99, "learning_rate": 1.8535176239877653e-05, "loss": 1.0362, "step": 6482 }, { "epoch": 1.99, "learning_rate": 1.8534656936354083e-05, "loss": 1.0655, "step": 6483 }, { "epoch": 2.0, "learning_rate": 1.8534137548073503e-05, "loss": 1.2901, "step": 6484 }, { "epoch": 2.0, "learning_rate": 1.8533618075041072e-05, "loss": 1.4886, "step": 6485 }, { "epoch": 2.0, "learning_rate": 1.853309851726195e-05, "loss": 1.2335, "step": 6486 }, { "epoch": 2.0, "learning_rate": 1.853257887474129e-05, "loss": 1.1866, "step": 6487 }, { "epoch": 2.0, "learning_rate": 1.8532059147484265e-05, "loss": 1.717, "step": 6488 }, { "epoch": 2.0, "learning_rate": 1.8531539335496025e-05, "loss": 1.3206, "step": 6489 }, { "epoch": 2.0, "learning_rate": 1.8531019438781737e-05, "loss": 1.1786, "step": 6490 }, { "epoch": 2.0, "learning_rate": 1.8530499457346565e-05, "loss": 1.338, "step": 6491 }, { "epoch": 2.0, "learning_rate": 1.8529979391195668e-05, "loss": 1.137, "step": 6492 }, { "epoch": 2.0, "learning_rate": 1.852945924033422e-05, "loss": 1.0479, "step": 6493 }, { "epoch": 2.0, "learning_rate": 1.8528939004767377e-05, "loss": 1.2455, "step": 6494 }, { "epoch": 2.0, "learning_rate": 1.852841868450031e-05, "loss": 1.3509, "step": 6495 }, { "epoch": 2.0, "learning_rate": 1.8527898279538186e-05, "loss": 1.0206, "step": 6496 }, { "epoch": 2.0, "learning_rate": 1.852737778988617e-05, "loss": 1.536, "step": 6497 }, { "epoch": 2.0, "learning_rate": 1.8526857215549437e-05, "loss": 1.1291, "step": 6498 }, { "epoch": 2.0, "learning_rate": 1.8526336556533153e-05, "loss": 1.0088, "step": 6499 }, { "epoch": 2.0, "learning_rate": 1.852581581284249e-05, "loss": 1.4339, "step": 6500 }, { "epoch": 2.0, "learning_rate": 1.852529498448261e-05, "loss": 1.0214, "step": 6501 }, { "epoch": 2.0, "learning_rate": 1.8524774071458703e-05, "loss": 0.7066, "step": 6502 }, { "epoch": 2.0, "learning_rate": 1.852425307377593e-05, "loss": 0.6084, "step": 6503 }, { "epoch": 2.0, "learning_rate": 1.8523731991439463e-05, "loss": 0.5267, "step": 6504 }, { "epoch": 2.0, "learning_rate": 1.8523210824454488e-05, "loss": 0.5512, "step": 6505 }, { "epoch": 2.0, "learning_rate": 1.8522689572826174e-05, "loss": 0.8208, "step": 6506 }, { "epoch": 2.0, "learning_rate": 1.8522168236559693e-05, "loss": 0.8283, "step": 6507 }, { "epoch": 2.0, "learning_rate": 1.852164681566023e-05, "loss": 0.7286, "step": 6508 }, { "epoch": 2.0, "learning_rate": 1.852112531013296e-05, "loss": 0.7372, "step": 6509 }, { "epoch": 2.0, "learning_rate": 1.8520603719983062e-05, "loss": 0.795, "step": 6510 }, { "epoch": 2.0, "learning_rate": 1.852008204521572e-05, "loss": 0.6021, "step": 6511 }, { "epoch": 2.0, "learning_rate": 1.8519560285836108e-05, "loss": 0.894, "step": 6512 }, { "epoch": 2.0, "learning_rate": 1.8519038441849408e-05, "loss": 0.7631, "step": 6513 }, { "epoch": 2.0, "learning_rate": 1.851851651326081e-05, "loss": 0.6773, "step": 6514 }, { "epoch": 2.0, "learning_rate": 1.851799450007549e-05, "loss": 0.7326, "step": 6515 }, { "epoch": 2.0, "learning_rate": 1.8517472402298634e-05, "loss": 0.8272, "step": 6516 }, { "epoch": 2.01, "learning_rate": 1.8516950219935424e-05, "loss": 0.6415, "step": 6517 }, { "epoch": 2.01, "learning_rate": 1.8516427952991055e-05, "loss": 0.788, "step": 6518 }, { "epoch": 2.01, "learning_rate": 1.85159056014707e-05, "loss": 0.5879, "step": 6519 }, { "epoch": 2.01, "learning_rate": 1.851538316537956e-05, "loss": 0.8011, "step": 6520 }, { "epoch": 2.01, "learning_rate": 1.8514860644722813e-05, "loss": 0.754, "step": 6521 }, { "epoch": 2.01, "learning_rate": 1.8514338039505654e-05, "loss": 0.7816, "step": 6522 }, { "epoch": 2.01, "learning_rate": 1.8513815349733277e-05, "loss": 0.7335, "step": 6523 }, { "epoch": 2.01, "learning_rate": 1.8513292575410856e-05, "loss": 0.7924, "step": 6524 }, { "epoch": 2.01, "learning_rate": 1.85127697165436e-05, "loss": 0.7176, "step": 6525 }, { "epoch": 2.01, "learning_rate": 1.8512246773136696e-05, "loss": 0.7208, "step": 6526 }, { "epoch": 2.01, "learning_rate": 1.8511723745195332e-05, "loss": 0.7862, "step": 6527 }, { "epoch": 2.01, "learning_rate": 1.8511200632724714e-05, "loss": 0.7874, "step": 6528 }, { "epoch": 2.01, "learning_rate": 1.8510677435730024e-05, "loss": 0.8285, "step": 6529 }, { "epoch": 2.01, "learning_rate": 1.8510154154216465e-05, "loss": 0.6213, "step": 6530 }, { "epoch": 2.01, "learning_rate": 1.850963078818923e-05, "loss": 0.8628, "step": 6531 }, { "epoch": 2.01, "learning_rate": 1.850910733765352e-05, "loss": 0.7856, "step": 6532 }, { "epoch": 2.01, "learning_rate": 1.8508583802614534e-05, "loss": 0.7868, "step": 6533 }, { "epoch": 2.01, "learning_rate": 1.8508060183077465e-05, "loss": 1.09, "step": 6534 }, { "epoch": 2.01, "learning_rate": 1.850753647904752e-05, "loss": 0.9081, "step": 6535 }, { "epoch": 2.01, "learning_rate": 1.8507012690529895e-05, "loss": 0.5339, "step": 6536 }, { "epoch": 2.01, "learning_rate": 1.8506488817529794e-05, "loss": 0.7542, "step": 6537 }, { "epoch": 2.01, "learning_rate": 1.850596486005242e-05, "loss": 0.686, "step": 6538 }, { "epoch": 2.01, "learning_rate": 1.8505440818102974e-05, "loss": 0.7824, "step": 6539 }, { "epoch": 2.01, "learning_rate": 1.8504916691686668e-05, "loss": 1.2265, "step": 6540 }, { "epoch": 2.01, "learning_rate": 1.8504392480808695e-05, "loss": 0.5811, "step": 6541 }, { "epoch": 2.01, "learning_rate": 1.850386818547427e-05, "loss": 0.7581, "step": 6542 }, { "epoch": 2.01, "learning_rate": 1.8503343805688595e-05, "loss": 0.7172, "step": 6543 }, { "epoch": 2.01, "learning_rate": 1.8502819341456874e-05, "loss": 0.8165, "step": 6544 }, { "epoch": 2.01, "learning_rate": 1.8502294792784327e-05, "loss": 0.6736, "step": 6545 }, { "epoch": 2.01, "learning_rate": 1.8501770159676157e-05, "loss": 0.664, "step": 6546 }, { "epoch": 2.01, "learning_rate": 1.850124544213757e-05, "loss": 0.7516, "step": 6547 }, { "epoch": 2.01, "learning_rate": 1.850072064017378e-05, "loss": 0.6684, "step": 6548 }, { "epoch": 2.01, "learning_rate": 1.8500195753790008e-05, "loss": 0.8071, "step": 6549 }, { "epoch": 2.02, "learning_rate": 1.849967078299145e-05, "loss": 0.5956, "step": 6550 }, { "epoch": 2.02, "learning_rate": 1.849914572778333e-05, "loss": 0.53, "step": 6551 }, { "epoch": 2.02, "learning_rate": 1.849862058817086e-05, "loss": 0.6523, "step": 6552 }, { "epoch": 2.02, "learning_rate": 1.8498095364159255e-05, "loss": 0.9209, "step": 6553 }, { "epoch": 2.02, "learning_rate": 1.849757005575373e-05, "loss": 0.7886, "step": 6554 }, { "epoch": 2.02, "learning_rate": 1.8497044662959504e-05, "loss": 0.8144, "step": 6555 }, { "epoch": 2.02, "learning_rate": 1.8496519185781793e-05, "loss": 0.8519, "step": 6556 }, { "epoch": 2.02, "learning_rate": 1.8495993624225814e-05, "loss": 0.9182, "step": 6557 }, { "epoch": 2.02, "learning_rate": 1.8495467978296792e-05, "loss": 0.9425, "step": 6558 }, { "epoch": 2.02, "learning_rate": 1.849494224799994e-05, "loss": 0.7802, "step": 6559 }, { "epoch": 2.02, "learning_rate": 1.8494416433340484e-05, "loss": 0.9283, "step": 6560 }, { "epoch": 2.02, "learning_rate": 1.8493890534323645e-05, "loss": 0.869, "step": 6561 }, { "epoch": 2.02, "learning_rate": 1.8493364550954644e-05, "loss": 0.6798, "step": 6562 }, { "epoch": 2.02, "learning_rate": 1.8492838483238704e-05, "loss": 0.7344, "step": 6563 }, { "epoch": 2.02, "learning_rate": 1.849231233118105e-05, "loss": 0.6497, "step": 6564 }, { "epoch": 2.02, "learning_rate": 1.849178609478691e-05, "loss": 1.0071, "step": 6565 }, { "epoch": 2.02, "learning_rate": 1.8491259774061508e-05, "loss": 0.8267, "step": 6566 }, { "epoch": 2.02, "learning_rate": 1.849073336901007e-05, "loss": 0.7808, "step": 6567 }, { "epoch": 2.02, "learning_rate": 1.8490206879637825e-05, "loss": 0.7163, "step": 6568 }, { "epoch": 2.02, "learning_rate": 1.8489680305949997e-05, "loss": 0.6491, "step": 6569 }, { "epoch": 2.02, "learning_rate": 1.8489153647951826e-05, "loss": 0.6503, "step": 6570 }, { "epoch": 2.02, "learning_rate": 1.848862690564853e-05, "loss": 0.6277, "step": 6571 }, { "epoch": 2.02, "learning_rate": 1.8488100079045345e-05, "loss": 0.8945, "step": 6572 }, { "epoch": 2.02, "learning_rate": 1.8487573168147507e-05, "loss": 0.6501, "step": 6573 }, { "epoch": 2.02, "learning_rate": 1.8487046172960242e-05, "loss": 0.7642, "step": 6574 }, { "epoch": 2.02, "learning_rate": 1.8486519093488787e-05, "loss": 0.8746, "step": 6575 }, { "epoch": 2.02, "learning_rate": 1.848599192973838e-05, "loss": 0.7121, "step": 6576 }, { "epoch": 2.02, "learning_rate": 1.848546468171425e-05, "loss": 0.4841, "step": 6577 }, { "epoch": 2.02, "learning_rate": 1.848493734942163e-05, "loss": 0.7301, "step": 6578 }, { "epoch": 2.02, "learning_rate": 1.848440993286577e-05, "loss": 0.8505, "step": 6579 }, { "epoch": 2.02, "learning_rate": 1.8483882432051898e-05, "loss": 1.0302, "step": 6580 }, { "epoch": 2.02, "learning_rate": 1.8483354846985253e-05, "loss": 0.838, "step": 6581 }, { "epoch": 2.03, "learning_rate": 1.8482827177671076e-05, "loss": 0.7934, "step": 6582 }, { "epoch": 2.03, "learning_rate": 1.8482299424114606e-05, "loss": 0.6197, "step": 6583 }, { "epoch": 2.03, "learning_rate": 1.8481771586321085e-05, "loss": 0.7376, "step": 6584 }, { "epoch": 2.03, "learning_rate": 1.848124366429576e-05, "loss": 0.652, "step": 6585 }, { "epoch": 2.03, "learning_rate": 1.8480715658043866e-05, "loss": 1.1795, "step": 6586 }, { "epoch": 2.03, "learning_rate": 1.848018756757065e-05, "loss": 0.6106, "step": 6587 }, { "epoch": 2.03, "learning_rate": 1.8479659392881352e-05, "loss": 0.7002, "step": 6588 }, { "epoch": 2.03, "learning_rate": 1.8479131133981224e-05, "loss": 0.8437, "step": 6589 }, { "epoch": 2.03, "learning_rate": 1.8478602790875508e-05, "loss": 0.9703, "step": 6590 }, { "epoch": 2.03, "learning_rate": 1.8478074363569452e-05, "loss": 0.8392, "step": 6591 }, { "epoch": 2.03, "learning_rate": 1.8477545852068308e-05, "loss": 0.7184, "step": 6592 }, { "epoch": 2.03, "learning_rate": 1.8477017256377316e-05, "loss": 0.6613, "step": 6593 }, { "epoch": 2.03, "learning_rate": 1.8476488576501732e-05, "loss": 0.6473, "step": 6594 }, { "epoch": 2.03, "learning_rate": 1.8475959812446805e-05, "loss": 0.8367, "step": 6595 }, { "epoch": 2.03, "learning_rate": 1.8475430964217785e-05, "loss": 0.7736, "step": 6596 }, { "epoch": 2.03, "learning_rate": 1.8474902031819925e-05, "loss": 0.7056, "step": 6597 }, { "epoch": 2.03, "learning_rate": 1.8474373015258472e-05, "loss": 0.6312, "step": 6598 }, { "epoch": 2.03, "learning_rate": 1.847384391453869e-05, "loss": 0.7687, "step": 6599 }, { "epoch": 2.03, "learning_rate": 1.847331472966583e-05, "loss": 0.9539, "step": 6600 }, { "epoch": 2.03, "learning_rate": 1.847278546064514e-05, "loss": 0.6622, "step": 6601 }, { "epoch": 2.03, "learning_rate": 1.8472256107481887e-05, "loss": 0.8903, "step": 6602 }, { "epoch": 2.03, "learning_rate": 1.8471726670181317e-05, "loss": 0.7565, "step": 6603 }, { "epoch": 2.03, "learning_rate": 1.8471197148748695e-05, "loss": 0.9755, "step": 6604 }, { "epoch": 2.03, "learning_rate": 1.847066754318928e-05, "loss": 0.5626, "step": 6605 }, { "epoch": 2.03, "learning_rate": 1.847013785350833e-05, "loss": 0.7963, "step": 6606 }, { "epoch": 2.03, "learning_rate": 1.8469608079711105e-05, "loss": 0.4982, "step": 6607 }, { "epoch": 2.03, "learning_rate": 1.846907822180286e-05, "loss": 0.9557, "step": 6608 }, { "epoch": 2.03, "learning_rate": 1.846854827978887e-05, "loss": 0.7711, "step": 6609 }, { "epoch": 2.03, "learning_rate": 1.8468018253674387e-05, "loss": 0.9696, "step": 6610 }, { "epoch": 2.03, "learning_rate": 1.846748814346468e-05, "loss": 0.6662, "step": 6611 }, { "epoch": 2.03, "learning_rate": 1.846695794916501e-05, "loss": 0.7167, "step": 6612 }, { "epoch": 2.03, "learning_rate": 1.8466427670780645e-05, "loss": 1.0886, "step": 6613 }, { "epoch": 2.03, "learning_rate": 1.846589730831685e-05, "loss": 0.5666, "step": 6614 }, { "epoch": 2.04, "learning_rate": 1.8465366861778894e-05, "loss": 1.1124, "step": 6615 }, { "epoch": 2.04, "learning_rate": 1.8464836331172042e-05, "loss": 0.8077, "step": 6616 }, { "epoch": 2.04, "learning_rate": 1.8464305716501564e-05, "loss": 0.7143, "step": 6617 }, { "epoch": 2.04, "learning_rate": 1.846377501777273e-05, "loss": 0.6982, "step": 6618 }, { "epoch": 2.04, "learning_rate": 1.846324423499081e-05, "loss": 0.7545, "step": 6619 }, { "epoch": 2.04, "learning_rate": 1.8462713368161072e-05, "loss": 0.8644, "step": 6620 }, { "epoch": 2.04, "learning_rate": 1.846218241728879e-05, "loss": 0.8246, "step": 6621 }, { "epoch": 2.04, "learning_rate": 1.8461651382379244e-05, "loss": 0.9608, "step": 6622 }, { "epoch": 2.04, "learning_rate": 1.84611202634377e-05, "loss": 0.6829, "step": 6623 }, { "epoch": 2.04, "learning_rate": 1.846058906046943e-05, "loss": 1.1897, "step": 6624 }, { "epoch": 2.04, "learning_rate": 1.8460057773479712e-05, "loss": 0.7728, "step": 6625 }, { "epoch": 2.04, "learning_rate": 1.8459526402473823e-05, "loss": 0.7386, "step": 6626 }, { "epoch": 2.04, "learning_rate": 1.8458994947457046e-05, "loss": 0.7766, "step": 6627 }, { "epoch": 2.04, "learning_rate": 1.8458463408434647e-05, "loss": 0.963, "step": 6628 }, { "epoch": 2.04, "learning_rate": 1.8457931785411917e-05, "loss": 0.7782, "step": 6629 }, { "epoch": 2.04, "learning_rate": 1.8457400078394126e-05, "loss": 0.7421, "step": 6630 }, { "epoch": 2.04, "learning_rate": 1.8456868287386558e-05, "loss": 0.8315, "step": 6631 }, { "epoch": 2.04, "learning_rate": 1.845633641239449e-05, "loss": 0.8838, "step": 6632 }, { "epoch": 2.04, "learning_rate": 1.8455804453423213e-05, "loss": 0.6108, "step": 6633 }, { "epoch": 2.04, "learning_rate": 1.8455272410478003e-05, "loss": 0.8106, "step": 6634 }, { "epoch": 2.04, "learning_rate": 1.8454740283564144e-05, "loss": 0.8472, "step": 6635 }, { "epoch": 2.04, "learning_rate": 1.845420807268693e-05, "loss": 0.6406, "step": 6636 }, { "epoch": 2.04, "learning_rate": 1.8453675777851627e-05, "loss": 0.6085, "step": 6637 }, { "epoch": 2.04, "learning_rate": 1.845314339906354e-05, "loss": 0.7772, "step": 6638 }, { "epoch": 2.04, "learning_rate": 1.8452610936327944e-05, "loss": 0.6667, "step": 6639 }, { "epoch": 2.04, "learning_rate": 1.845207838965013e-05, "loss": 1.1381, "step": 6640 }, { "epoch": 2.04, "learning_rate": 1.8451545759035393e-05, "loss": 0.8675, "step": 6641 }, { "epoch": 2.04, "learning_rate": 1.8451013044489015e-05, "loss": 0.5166, "step": 6642 }, { "epoch": 2.04, "learning_rate": 1.8450480246016285e-05, "loss": 1.0212, "step": 6643 }, { "epoch": 2.04, "learning_rate": 1.8449947363622506e-05, "loss": 0.6643, "step": 6644 }, { "epoch": 2.04, "learning_rate": 1.8449414397312955e-05, "loss": 0.6696, "step": 6645 }, { "epoch": 2.04, "learning_rate": 1.8448881347092936e-05, "loss": 0.7569, "step": 6646 }, { "epoch": 2.05, "learning_rate": 1.8448348212967734e-05, "loss": 1.1494, "step": 6647 }, { "epoch": 2.05, "learning_rate": 1.8447814994942655e-05, "loss": 0.7341, "step": 6648 }, { "epoch": 2.05, "learning_rate": 1.8447281693022984e-05, "loss": 0.5575, "step": 6649 }, { "epoch": 2.05, "learning_rate": 1.844674830721402e-05, "loss": 0.6947, "step": 6650 }, { "epoch": 2.05, "learning_rate": 1.844621483752106e-05, "loss": 0.5905, "step": 6651 }, { "epoch": 2.05, "learning_rate": 1.84456812839494e-05, "loss": 1.1885, "step": 6652 }, { "epoch": 2.05, "learning_rate": 1.8445147646504344e-05, "loss": 0.924, "step": 6653 }, { "epoch": 2.05, "learning_rate": 1.844461392519119e-05, "loss": 0.74, "step": 6654 }, { "epoch": 2.05, "learning_rate": 1.8444080120015235e-05, "loss": 0.87, "step": 6655 }, { "epoch": 2.05, "learning_rate": 1.8443546230981785e-05, "loss": 0.7266, "step": 6656 }, { "epoch": 2.05, "learning_rate": 1.8443012258096135e-05, "loss": 1.1133, "step": 6657 }, { "epoch": 2.05, "learning_rate": 1.8442478201363592e-05, "loss": 0.9751, "step": 6658 }, { "epoch": 2.05, "learning_rate": 1.844194406078946e-05, "loss": 0.6229, "step": 6659 }, { "epoch": 2.05, "learning_rate": 1.844140983637905e-05, "loss": 0.7754, "step": 6660 }, { "epoch": 2.05, "learning_rate": 1.8440875528137653e-05, "loss": 0.6834, "step": 6661 }, { "epoch": 2.05, "learning_rate": 1.8440341136070584e-05, "loss": 0.7182, "step": 6662 }, { "epoch": 2.05, "learning_rate": 1.843980666018315e-05, "loss": 0.7073, "step": 6663 }, { "epoch": 2.05, "learning_rate": 1.8439272100480657e-05, "loss": 0.7007, "step": 6664 }, { "epoch": 2.05, "learning_rate": 1.8438737456968415e-05, "loss": 0.782, "step": 6665 }, { "epoch": 2.05, "learning_rate": 1.843820272965173e-05, "loss": 0.7948, "step": 6666 }, { "epoch": 2.05, "learning_rate": 1.843766791853592e-05, "loss": 0.722, "step": 6667 }, { "epoch": 2.05, "learning_rate": 1.8437133023626287e-05, "loss": 0.5, "step": 6668 }, { "epoch": 2.05, "learning_rate": 1.8436598044928146e-05, "loss": 0.5571, "step": 6669 }, { "epoch": 2.05, "learning_rate": 1.8436062982446814e-05, "loss": 0.6777, "step": 6670 }, { "epoch": 2.05, "learning_rate": 1.8435527836187602e-05, "loss": 0.7959, "step": 6671 }, { "epoch": 2.05, "learning_rate": 1.8434992606155822e-05, "loss": 0.6537, "step": 6672 }, { "epoch": 2.05, "learning_rate": 1.843445729235679e-05, "loss": 0.63, "step": 6673 }, { "epoch": 2.05, "learning_rate": 1.843392189479583e-05, "loss": 0.7946, "step": 6674 }, { "epoch": 2.05, "learning_rate": 1.8433386413478246e-05, "loss": 0.8296, "step": 6675 }, { "epoch": 2.05, "learning_rate": 1.8432850848409367e-05, "loss": 0.4521, "step": 6676 }, { "epoch": 2.05, "learning_rate": 1.8432315199594506e-05, "loss": 0.614, "step": 6677 }, { "epoch": 2.05, "learning_rate": 1.8431779467038983e-05, "loss": 0.7444, "step": 6678 }, { "epoch": 2.05, "learning_rate": 1.8431243650748117e-05, "loss": 0.8337, "step": 6679 }, { "epoch": 2.06, "learning_rate": 1.8430707750727234e-05, "loss": 0.8471, "step": 6680 }, { "epoch": 2.06, "learning_rate": 1.843017176698165e-05, "loss": 0.7535, "step": 6681 }, { "epoch": 2.06, "learning_rate": 1.8429635699516694e-05, "loss": 0.7247, "step": 6682 }, { "epoch": 2.06, "learning_rate": 1.8429099548337685e-05, "loss": 0.6846, "step": 6683 }, { "epoch": 2.06, "learning_rate": 1.842856331344995e-05, "loss": 0.8258, "step": 6684 }, { "epoch": 2.06, "learning_rate": 1.8428026994858813e-05, "loss": 0.5891, "step": 6685 }, { "epoch": 2.06, "learning_rate": 1.84274905925696e-05, "loss": 0.7524, "step": 6686 }, { "epoch": 2.06, "learning_rate": 1.842695410658764e-05, "loss": 0.8618, "step": 6687 }, { "epoch": 2.06, "learning_rate": 1.8426417536918258e-05, "loss": 0.7749, "step": 6688 }, { "epoch": 2.06, "learning_rate": 1.8425880883566784e-05, "loss": 0.7, "step": 6689 }, { "epoch": 2.06, "learning_rate": 1.8425344146538545e-05, "loss": 0.5943, "step": 6690 }, { "epoch": 2.06, "learning_rate": 1.8424807325838876e-05, "loss": 0.626, "step": 6691 }, { "epoch": 2.06, "learning_rate": 1.842427042147311e-05, "loss": 0.7646, "step": 6692 }, { "epoch": 2.06, "learning_rate": 1.8423733433446568e-05, "loss": 0.9232, "step": 6693 }, { "epoch": 2.06, "learning_rate": 1.842319636176459e-05, "loss": 0.7307, "step": 6694 }, { "epoch": 2.06, "learning_rate": 1.842265920643251e-05, "loss": 0.6748, "step": 6695 }, { "epoch": 2.06, "learning_rate": 1.842212196745566e-05, "loss": 0.7431, "step": 6696 }, { "epoch": 2.06, "learning_rate": 1.8421584644839382e-05, "loss": 0.9977, "step": 6697 }, { "epoch": 2.06, "learning_rate": 1.8421047238589005e-05, "loss": 0.5871, "step": 6698 }, { "epoch": 2.06, "learning_rate": 1.8420509748709866e-05, "loss": 0.8701, "step": 6699 }, { "epoch": 2.06, "learning_rate": 1.8419972175207305e-05, "loss": 0.8691, "step": 6700 }, { "epoch": 2.06, "learning_rate": 1.8419434518086657e-05, "loss": 0.6434, "step": 6701 }, { "epoch": 2.06, "learning_rate": 1.8418896777353272e-05, "loss": 0.7109, "step": 6702 }, { "epoch": 2.06, "learning_rate": 1.8418358953012478e-05, "loss": 0.5356, "step": 6703 }, { "epoch": 2.06, "learning_rate": 1.8417821045069623e-05, "loss": 1.1604, "step": 6704 }, { "epoch": 2.06, "learning_rate": 1.8417283053530047e-05, "loss": 0.6038, "step": 6705 }, { "epoch": 2.06, "learning_rate": 1.841674497839909e-05, "loss": 0.709, "step": 6706 }, { "epoch": 2.06, "learning_rate": 1.8416206819682102e-05, "loss": 0.9568, "step": 6707 }, { "epoch": 2.06, "learning_rate": 1.841566857738442e-05, "loss": 0.444, "step": 6708 }, { "epoch": 2.06, "learning_rate": 1.8415130251511395e-05, "loss": 0.8473, "step": 6709 }, { "epoch": 2.06, "learning_rate": 1.841459184206837e-05, "loss": 0.8303, "step": 6710 }, { "epoch": 2.06, "learning_rate": 1.841405334906069e-05, "loss": 0.6667, "step": 6711 }, { "epoch": 2.07, "learning_rate": 1.8413514772493713e-05, "loss": 0.8951, "step": 6712 }, { "epoch": 2.07, "learning_rate": 1.8412976112372777e-05, "loss": 0.6756, "step": 6713 }, { "epoch": 2.07, "learning_rate": 1.8412437368703235e-05, "loss": 0.9854, "step": 6714 }, { "epoch": 2.07, "learning_rate": 1.8411898541490433e-05, "loss": 0.783, "step": 6715 }, { "epoch": 2.07, "learning_rate": 1.841135963073973e-05, "loss": 0.73, "step": 6716 }, { "epoch": 2.07, "learning_rate": 1.8410820636456472e-05, "loss": 0.5132, "step": 6717 }, { "epoch": 2.07, "learning_rate": 1.841028155864601e-05, "loss": 0.805, "step": 6718 }, { "epoch": 2.07, "learning_rate": 1.840974239731371e-05, "loss": 0.7316, "step": 6719 }, { "epoch": 2.07, "learning_rate": 1.840920315246491e-05, "loss": 0.8472, "step": 6720 }, { "epoch": 2.07, "learning_rate": 1.8408663824104972e-05, "loss": 0.9235, "step": 6721 }, { "epoch": 2.07, "learning_rate": 1.8408124412239256e-05, "loss": 0.9143, "step": 6722 }, { "epoch": 2.07, "learning_rate": 1.8407584916873117e-05, "loss": 0.8067, "step": 6723 }, { "epoch": 2.07, "learning_rate": 1.8407045338011908e-05, "loss": 0.7973, "step": 6724 }, { "epoch": 2.07, "learning_rate": 1.8406505675660992e-05, "loss": 0.5585, "step": 6725 }, { "epoch": 2.07, "learning_rate": 1.8405965929825726e-05, "loss": 0.8617, "step": 6726 }, { "epoch": 2.07, "learning_rate": 1.8405426100511473e-05, "loss": 0.8312, "step": 6727 }, { "epoch": 2.07, "learning_rate": 1.840488618772359e-05, "loss": 1.0305, "step": 6728 }, { "epoch": 2.07, "learning_rate": 1.840434619146744e-05, "loss": 0.6998, "step": 6729 }, { "epoch": 2.07, "learning_rate": 1.8403806111748387e-05, "loss": 0.5573, "step": 6730 }, { "epoch": 2.07, "learning_rate": 1.84032659485718e-05, "loss": 0.7375, "step": 6731 }, { "epoch": 2.07, "learning_rate": 1.8402725701943028e-05, "loss": 0.9201, "step": 6732 }, { "epoch": 2.07, "learning_rate": 1.840218537186745e-05, "loss": 1.0973, "step": 6733 }, { "epoch": 2.07, "learning_rate": 1.840164495835043e-05, "loss": 0.6436, "step": 6734 }, { "epoch": 2.07, "learning_rate": 1.840110446139733e-05, "loss": 0.8024, "step": 6735 }, { "epoch": 2.07, "learning_rate": 1.840056388101352e-05, "loss": 0.6444, "step": 6736 }, { "epoch": 2.07, "learning_rate": 1.8400023217204367e-05, "loss": 0.8041, "step": 6737 }, { "epoch": 2.07, "learning_rate": 1.8399482469975245e-05, "loss": 0.8277, "step": 6738 }, { "epoch": 2.07, "learning_rate": 1.839894163933152e-05, "loss": 0.8749, "step": 6739 }, { "epoch": 2.07, "learning_rate": 1.8398400725278564e-05, "loss": 0.7757, "step": 6740 }, { "epoch": 2.07, "learning_rate": 1.8397859727821747e-05, "loss": 0.8961, "step": 6741 }, { "epoch": 2.07, "learning_rate": 1.8397318646966445e-05, "loss": 0.7322, "step": 6742 }, { "epoch": 2.07, "learning_rate": 1.839677748271803e-05, "loss": 0.6915, "step": 6743 }, { "epoch": 2.07, "learning_rate": 1.8396236235081875e-05, "loss": 0.9916, "step": 6744 }, { "epoch": 2.08, "learning_rate": 1.839569490406335e-05, "loss": 0.676, "step": 6745 }, { "epoch": 2.08, "learning_rate": 1.8395153489667844e-05, "loss": 0.6575, "step": 6746 }, { "epoch": 2.08, "learning_rate": 1.8394611991900726e-05, "loss": 0.697, "step": 6747 }, { "epoch": 2.08, "learning_rate": 1.8394070410767376e-05, "loss": 0.9296, "step": 6748 }, { "epoch": 2.08, "learning_rate": 1.839352874627317e-05, "loss": 0.6719, "step": 6749 }, { "epoch": 2.08, "learning_rate": 1.8392986998423485e-05, "loss": 0.9944, "step": 6750 }, { "epoch": 2.08, "learning_rate": 1.83924451672237e-05, "loss": 0.6649, "step": 6751 }, { "epoch": 2.08, "learning_rate": 1.8391903252679206e-05, "loss": 1.0071, "step": 6752 }, { "epoch": 2.08, "learning_rate": 1.839136125479538e-05, "loss": 0.9503, "step": 6753 }, { "epoch": 2.08, "learning_rate": 1.83908191735776e-05, "loss": 0.6259, "step": 6754 }, { "epoch": 2.08, "learning_rate": 1.839027700903125e-05, "loss": 1.0312, "step": 6755 }, { "epoch": 2.08, "learning_rate": 1.838973476116172e-05, "loss": 0.8856, "step": 6756 }, { "epoch": 2.08, "learning_rate": 1.838919242997439e-05, "loss": 0.7189, "step": 6757 }, { "epoch": 2.08, "learning_rate": 1.838865001547465e-05, "loss": 0.8941, "step": 6758 }, { "epoch": 2.08, "learning_rate": 1.8388107517667882e-05, "loss": 0.9422, "step": 6759 }, { "epoch": 2.08, "learning_rate": 1.8387564936559476e-05, "loss": 0.6937, "step": 6760 }, { "epoch": 2.08, "learning_rate": 1.838702227215482e-05, "loss": 1.2152, "step": 6761 }, { "epoch": 2.08, "learning_rate": 1.83864795244593e-05, "loss": 0.6857, "step": 6762 }, { "epoch": 2.08, "learning_rate": 1.8385936693478313e-05, "loss": 0.437, "step": 6763 }, { "epoch": 2.08, "learning_rate": 1.8385393779217246e-05, "loss": 0.9079, "step": 6764 }, { "epoch": 2.08, "learning_rate": 1.838485078168149e-05, "loss": 0.617, "step": 6765 }, { "epoch": 2.08, "learning_rate": 1.838430770087644e-05, "loss": 0.8871, "step": 6766 }, { "epoch": 2.08, "learning_rate": 1.8383764536807486e-05, "loss": 0.825, "step": 6767 }, { "epoch": 2.08, "learning_rate": 1.838322128948002e-05, "loss": 0.7194, "step": 6768 }, { "epoch": 2.08, "learning_rate": 1.8382677958899443e-05, "loss": 0.9157, "step": 6769 }, { "epoch": 2.08, "learning_rate": 1.8382134545071147e-05, "loss": 0.6816, "step": 6770 }, { "epoch": 2.08, "learning_rate": 1.838159104800053e-05, "loss": 0.517, "step": 6771 }, { "epoch": 2.08, "learning_rate": 1.838104746769299e-05, "loss": 0.8699, "step": 6772 }, { "epoch": 2.08, "learning_rate": 1.8380503804153926e-05, "loss": 0.7831, "step": 6773 }, { "epoch": 2.08, "learning_rate": 1.8379960057388736e-05, "loss": 0.694, "step": 6774 }, { "epoch": 2.08, "learning_rate": 1.8379416227402816e-05, "loss": 0.7759, "step": 6775 }, { "epoch": 2.08, "learning_rate": 1.837887231420157e-05, "loss": 0.6893, "step": 6776 }, { "epoch": 2.09, "learning_rate": 1.83783283177904e-05, "loss": 0.9723, "step": 6777 }, { "epoch": 2.09, "learning_rate": 1.8377784238174713e-05, "loss": 0.8308, "step": 6778 }, { "epoch": 2.09, "learning_rate": 1.8377240075359905e-05, "loss": 0.9038, "step": 6779 }, { "epoch": 2.09, "learning_rate": 1.8376695829351378e-05, "loss": 0.6666, "step": 6780 }, { "epoch": 2.09, "learning_rate": 1.8376151500154546e-05, "loss": 0.8749, "step": 6781 }, { "epoch": 2.09, "learning_rate": 1.837560708777481e-05, "loss": 0.6895, "step": 6782 }, { "epoch": 2.09, "learning_rate": 1.837506259221757e-05, "loss": 0.6466, "step": 6783 }, { "epoch": 2.09, "learning_rate": 1.8374518013488245e-05, "loss": 0.7693, "step": 6784 }, { "epoch": 2.09, "learning_rate": 1.837397335159224e-05, "loss": 0.9653, "step": 6785 }, { "epoch": 2.09, "learning_rate": 1.8373428606534956e-05, "loss": 0.8657, "step": 6786 }, { "epoch": 2.09, "learning_rate": 1.8372883778321813e-05, "loss": 0.7185, "step": 6787 }, { "epoch": 2.09, "learning_rate": 1.8372338866958216e-05, "loss": 0.8559, "step": 6788 }, { "epoch": 2.09, "learning_rate": 1.837179387244958e-05, "loss": 0.6736, "step": 6789 }, { "epoch": 2.09, "learning_rate": 1.8371248794801314e-05, "loss": 0.7462, "step": 6790 }, { "epoch": 2.09, "learning_rate": 1.837070363401883e-05, "loss": 0.8935, "step": 6791 }, { "epoch": 2.09, "learning_rate": 1.8370158390107547e-05, "loss": 0.8971, "step": 6792 }, { "epoch": 2.09, "learning_rate": 1.8369613063072875e-05, "loss": 0.6558, "step": 6793 }, { "epoch": 2.09, "learning_rate": 1.8369067652920233e-05, "loss": 0.6596, "step": 6794 }, { "epoch": 2.09, "learning_rate": 1.8368522159655037e-05, "loss": 0.6463, "step": 6795 }, { "epoch": 2.09, "learning_rate": 1.8367976583282702e-05, "loss": 0.7981, "step": 6796 }, { "epoch": 2.09, "learning_rate": 1.8367430923808646e-05, "loss": 1.1223, "step": 6797 }, { "epoch": 2.09, "learning_rate": 1.836688518123829e-05, "loss": 0.5897, "step": 6798 }, { "epoch": 2.09, "learning_rate": 1.8366339355577053e-05, "loss": 0.6552, "step": 6799 }, { "epoch": 2.09, "learning_rate": 1.8365793446830358e-05, "loss": 1.1398, "step": 6800 }, { "epoch": 2.09, "learning_rate": 1.8365247455003623e-05, "loss": 0.8388, "step": 6801 }, { "epoch": 2.09, "learning_rate": 1.8364701380102267e-05, "loss": 0.6655, "step": 6802 }, { "epoch": 2.09, "learning_rate": 1.8364155222131724e-05, "loss": 0.7037, "step": 6803 }, { "epoch": 2.09, "learning_rate": 1.8363608981097406e-05, "loss": 0.7659, "step": 6804 }, { "epoch": 2.09, "learning_rate": 1.8363062657004744e-05, "loss": 0.8319, "step": 6805 }, { "epoch": 2.09, "learning_rate": 1.8362516249859164e-05, "loss": 0.8615, "step": 6806 }, { "epoch": 2.09, "learning_rate": 1.8361969759666086e-05, "loss": 0.4367, "step": 6807 }, { "epoch": 2.09, "learning_rate": 1.8361423186430945e-05, "loss": 0.9359, "step": 6808 }, { "epoch": 2.09, "learning_rate": 1.8360876530159165e-05, "loss": 0.7384, "step": 6809 }, { "epoch": 2.1, "learning_rate": 1.8360329790856178e-05, "loss": 1.0238, "step": 6810 }, { "epoch": 2.1, "learning_rate": 1.8359782968527407e-05, "loss": 0.8084, "step": 6811 }, { "epoch": 2.1, "learning_rate": 1.8359236063178287e-05, "loss": 0.5269, "step": 6812 }, { "epoch": 2.1, "learning_rate": 1.8358689074814254e-05, "loss": 0.5972, "step": 6813 }, { "epoch": 2.1, "learning_rate": 1.835814200344073e-05, "loss": 0.9184, "step": 6814 }, { "epoch": 2.1, "learning_rate": 1.8357594849063152e-05, "loss": 0.6505, "step": 6815 }, { "epoch": 2.1, "learning_rate": 1.8357047611686956e-05, "loss": 0.712, "step": 6816 }, { "epoch": 2.1, "learning_rate": 1.8356500291317574e-05, "loss": 0.8488, "step": 6817 }, { "epoch": 2.1, "learning_rate": 1.8355952887960448e-05, "loss": 0.9314, "step": 6818 }, { "epoch": 2.1, "learning_rate": 1.8355405401621e-05, "loss": 0.9754, "step": 6819 }, { "epoch": 2.1, "learning_rate": 1.8354857832304683e-05, "loss": 0.8576, "step": 6820 }, { "epoch": 2.1, "learning_rate": 1.8354310180016926e-05, "loss": 0.699, "step": 6821 }, { "epoch": 2.1, "learning_rate": 1.8353762444763167e-05, "loss": 1.0517, "step": 6822 }, { "epoch": 2.1, "learning_rate": 1.8353214626548852e-05, "loss": 0.6805, "step": 6823 }, { "epoch": 2.1, "learning_rate": 1.8352666725379414e-05, "loss": 0.7589, "step": 6824 }, { "epoch": 2.1, "learning_rate": 1.8352118741260297e-05, "loss": 0.8639, "step": 6825 }, { "epoch": 2.1, "learning_rate": 1.8351570674196942e-05, "loss": 0.9401, "step": 6826 }, { "epoch": 2.1, "learning_rate": 1.8351022524194797e-05, "loss": 0.6954, "step": 6827 }, { "epoch": 2.1, "learning_rate": 1.83504742912593e-05, "loss": 0.8128, "step": 6828 }, { "epoch": 2.1, "learning_rate": 1.8349925975395898e-05, "loss": 0.6683, "step": 6829 }, { "epoch": 2.1, "learning_rate": 1.8349377576610033e-05, "loss": 0.6693, "step": 6830 }, { "epoch": 2.1, "learning_rate": 1.8348829094907157e-05, "loss": 0.6725, "step": 6831 }, { "epoch": 2.1, "learning_rate": 1.8348280530292712e-05, "loss": 0.9511, "step": 6832 }, { "epoch": 2.1, "learning_rate": 1.834773188277215e-05, "loss": 0.9244, "step": 6833 }, { "epoch": 2.1, "learning_rate": 1.8347183152350913e-05, "loss": 0.6405, "step": 6834 }, { "epoch": 2.1, "learning_rate": 1.8346634339034456e-05, "loss": 0.7316, "step": 6835 }, { "epoch": 2.1, "learning_rate": 1.8346085442828228e-05, "loss": 0.8112, "step": 6836 }, { "epoch": 2.1, "learning_rate": 1.8345536463737678e-05, "loss": 0.6726, "step": 6837 }, { "epoch": 2.1, "learning_rate": 1.834498740176826e-05, "loss": 0.8892, "step": 6838 }, { "epoch": 2.1, "learning_rate": 1.834443825692543e-05, "loss": 0.7876, "step": 6839 }, { "epoch": 2.1, "learning_rate": 1.8343889029214633e-05, "loss": 0.7677, "step": 6840 }, { "epoch": 2.1, "learning_rate": 1.8343339718641327e-05, "loss": 0.7987, "step": 6841 }, { "epoch": 2.11, "learning_rate": 1.834279032521097e-05, "loss": 0.6441, "step": 6842 }, { "epoch": 2.11, "learning_rate": 1.834224084892902e-05, "loss": 0.6295, "step": 6843 }, { "epoch": 2.11, "learning_rate": 1.8341691289800925e-05, "loss": 0.712, "step": 6844 }, { "epoch": 2.11, "learning_rate": 1.834114164783215e-05, "loss": 0.8187, "step": 6845 }, { "epoch": 2.11, "learning_rate": 1.8340591923028148e-05, "loss": 0.8242, "step": 6846 }, { "epoch": 2.11, "learning_rate": 1.8340042115394384e-05, "loss": 0.6564, "step": 6847 }, { "epoch": 2.11, "learning_rate": 1.8339492224936315e-05, "loss": 0.7322, "step": 6848 }, { "epoch": 2.11, "learning_rate": 1.83389422516594e-05, "loss": 0.9826, "step": 6849 }, { "epoch": 2.11, "learning_rate": 1.8338392195569105e-05, "loss": 0.7975, "step": 6850 }, { "epoch": 2.11, "learning_rate": 1.8337842056670893e-05, "loss": 0.5778, "step": 6851 }, { "epoch": 2.11, "learning_rate": 1.8337291834970222e-05, "loss": 0.7946, "step": 6852 }, { "epoch": 2.11, "learning_rate": 1.833674153047256e-05, "loss": 0.6882, "step": 6853 }, { "epoch": 2.11, "learning_rate": 1.8336191143183372e-05, "loss": 0.8436, "step": 6854 }, { "epoch": 2.11, "learning_rate": 1.8335640673108122e-05, "loss": 0.7444, "step": 6855 }, { "epoch": 2.11, "learning_rate": 1.8335090120252274e-05, "loss": 0.6835, "step": 6856 }, { "epoch": 2.11, "learning_rate": 1.8334539484621304e-05, "loss": 0.7688, "step": 6857 }, { "epoch": 2.11, "learning_rate": 1.8333988766220676e-05, "loss": 0.4011, "step": 6858 }, { "epoch": 2.11, "learning_rate": 1.833343796505586e-05, "loss": 1.0992, "step": 6859 }, { "epoch": 2.11, "learning_rate": 1.833288708113232e-05, "loss": 0.916, "step": 6860 }, { "epoch": 2.11, "learning_rate": 1.8332336114455533e-05, "loss": 0.743, "step": 6861 }, { "epoch": 2.11, "learning_rate": 1.833178506503097e-05, "loss": 0.6344, "step": 6862 }, { "epoch": 2.11, "learning_rate": 1.8331233932864102e-05, "loss": 0.6564, "step": 6863 }, { "epoch": 2.11, "learning_rate": 1.8330682717960404e-05, "loss": 0.5528, "step": 6864 }, { "epoch": 2.11, "learning_rate": 1.8330131420325345e-05, "loss": 0.7095, "step": 6865 }, { "epoch": 2.11, "learning_rate": 1.8329580039964407e-05, "loss": 0.5265, "step": 6866 }, { "epoch": 2.11, "learning_rate": 1.8329028576883065e-05, "loss": 0.6878, "step": 6867 }, { "epoch": 2.11, "learning_rate": 1.832847703108679e-05, "loss": 0.7315, "step": 6868 }, { "epoch": 2.11, "learning_rate": 1.8327925402581058e-05, "loss": 0.6891, "step": 6869 }, { "epoch": 2.11, "learning_rate": 1.8327373691371357e-05, "loss": 0.6914, "step": 6870 }, { "epoch": 2.11, "learning_rate": 1.832682189746316e-05, "loss": 0.6808, "step": 6871 }, { "epoch": 2.11, "learning_rate": 1.8326270020861946e-05, "loss": 0.757, "step": 6872 }, { "epoch": 2.11, "learning_rate": 1.83257180615732e-05, "loss": 0.7823, "step": 6873 }, { "epoch": 2.11, "learning_rate": 1.8325166019602395e-05, "loss": 0.8077, "step": 6874 }, { "epoch": 2.12, "learning_rate": 1.8324613894955025e-05, "loss": 0.6586, "step": 6875 }, { "epoch": 2.12, "learning_rate": 1.8324061687636562e-05, "loss": 1.0608, "step": 6876 }, { "epoch": 2.12, "learning_rate": 1.83235093976525e-05, "loss": 0.8123, "step": 6877 }, { "epoch": 2.12, "learning_rate": 1.8322957025008312e-05, "loss": 0.7954, "step": 6878 }, { "epoch": 2.12, "learning_rate": 1.8322404569709492e-05, "loss": 0.7577, "step": 6879 }, { "epoch": 2.12, "learning_rate": 1.8321852031761523e-05, "loss": 0.8936, "step": 6880 }, { "epoch": 2.12, "learning_rate": 1.83212994111699e-05, "loss": 0.6425, "step": 6881 }, { "epoch": 2.12, "learning_rate": 1.8320746707940096e-05, "loss": 0.6437, "step": 6882 }, { "epoch": 2.12, "learning_rate": 1.8320193922077615e-05, "loss": 0.7928, "step": 6883 }, { "epoch": 2.12, "learning_rate": 1.831964105358794e-05, "loss": 0.7233, "step": 6884 }, { "epoch": 2.12, "learning_rate": 1.8319088102476563e-05, "loss": 0.9047, "step": 6885 }, { "epoch": 2.12, "learning_rate": 1.8318535068748972e-05, "loss": 0.7975, "step": 6886 }, { "epoch": 2.12, "learning_rate": 1.8317981952410657e-05, "loss": 0.8888, "step": 6887 }, { "epoch": 2.12, "learning_rate": 1.8317428753467118e-05, "loss": 0.6929, "step": 6888 }, { "epoch": 2.12, "learning_rate": 1.8316875471923847e-05, "loss": 0.6688, "step": 6889 }, { "epoch": 2.12, "learning_rate": 1.8316322107786338e-05, "loss": 0.6848, "step": 6890 }, { "epoch": 2.12, "learning_rate": 1.8315768661060085e-05, "loss": 1.0281, "step": 6891 }, { "epoch": 2.12, "learning_rate": 1.8315215131750586e-05, "loss": 0.8471, "step": 6892 }, { "epoch": 2.12, "learning_rate": 1.8314661519863333e-05, "loss": 0.8318, "step": 6893 }, { "epoch": 2.12, "learning_rate": 1.8314107825403833e-05, "loss": 0.7539, "step": 6894 }, { "epoch": 2.12, "learning_rate": 1.831355404837758e-05, "loss": 0.8029, "step": 6895 }, { "epoch": 2.12, "learning_rate": 1.831300018879007e-05, "loss": 1.0453, "step": 6896 }, { "epoch": 2.12, "learning_rate": 1.831244624664681e-05, "loss": 0.6266, "step": 6897 }, { "epoch": 2.12, "learning_rate": 1.8311892221953296e-05, "loss": 0.5873, "step": 6898 }, { "epoch": 2.12, "learning_rate": 1.831133811471503e-05, "loss": 1.3462, "step": 6899 }, { "epoch": 2.12, "learning_rate": 1.831078392493752e-05, "loss": 0.8284, "step": 6900 }, { "epoch": 2.12, "learning_rate": 1.8310229652626264e-05, "loss": 0.7298, "step": 6901 }, { "epoch": 2.12, "learning_rate": 1.830967529778677e-05, "loss": 0.942, "step": 6902 }, { "epoch": 2.12, "learning_rate": 1.830912086042454e-05, "loss": 0.8115, "step": 6903 }, { "epoch": 2.12, "learning_rate": 1.8308566340545083e-05, "loss": 0.7256, "step": 6904 }, { "epoch": 2.12, "learning_rate": 1.8308011738153904e-05, "loss": 0.7507, "step": 6905 }, { "epoch": 2.12, "learning_rate": 1.830745705325651e-05, "loss": 0.8836, "step": 6906 }, { "epoch": 2.13, "learning_rate": 1.8306902285858415e-05, "loss": 0.9546, "step": 6907 }, { "epoch": 2.13, "learning_rate": 1.8306347435965124e-05, "loss": 0.6503, "step": 6908 }, { "epoch": 2.13, "learning_rate": 1.8305792503582146e-05, "loss": 0.9493, "step": 6909 }, { "epoch": 2.13, "learning_rate": 1.8305237488714995e-05, "loss": 0.7569, "step": 6910 }, { "epoch": 2.13, "learning_rate": 1.830468239136918e-05, "loss": 0.8418, "step": 6911 }, { "epoch": 2.13, "learning_rate": 1.8304127211550216e-05, "loss": 0.82, "step": 6912 }, { "epoch": 2.13, "learning_rate": 1.8303571949263613e-05, "loss": 0.9233, "step": 6913 }, { "epoch": 2.13, "learning_rate": 1.8303016604514893e-05, "loss": 1.0447, "step": 6914 }, { "epoch": 2.13, "learning_rate": 1.830246117730956e-05, "loss": 0.7526, "step": 6915 }, { "epoch": 2.13, "learning_rate": 1.830190566765314e-05, "loss": 0.8354, "step": 6916 }, { "epoch": 2.13, "learning_rate": 1.830135007555114e-05, "loss": 0.5675, "step": 6917 }, { "epoch": 2.13, "learning_rate": 1.8300794401009087e-05, "loss": 0.8017, "step": 6918 }, { "epoch": 2.13, "learning_rate": 1.8300238644032498e-05, "loss": 0.8344, "step": 6919 }, { "epoch": 2.13, "learning_rate": 1.8299682804626884e-05, "loss": 0.6244, "step": 6920 }, { "epoch": 2.13, "learning_rate": 1.8299126882797775e-05, "loss": 0.902, "step": 6921 }, { "epoch": 2.13, "learning_rate": 1.8298570878550687e-05, "loss": 0.7488, "step": 6922 }, { "epoch": 2.13, "learning_rate": 1.8298014791891138e-05, "loss": 0.8065, "step": 6923 }, { "epoch": 2.13, "learning_rate": 1.829745862282466e-05, "loss": 0.6377, "step": 6924 }, { "epoch": 2.13, "learning_rate": 1.8296902371356764e-05, "loss": 0.6576, "step": 6925 }, { "epoch": 2.13, "learning_rate": 1.8296346037492984e-05, "loss": 0.8883, "step": 6926 }, { "epoch": 2.13, "learning_rate": 1.8295789621238844e-05, "loss": 0.7661, "step": 6927 }, { "epoch": 2.13, "learning_rate": 1.8295233122599865e-05, "loss": 0.8412, "step": 6928 }, { "epoch": 2.13, "learning_rate": 1.8294676541581578e-05, "loss": 0.8892, "step": 6929 }, { "epoch": 2.13, "learning_rate": 1.8294119878189506e-05, "loss": 0.6444, "step": 6930 }, { "epoch": 2.13, "learning_rate": 1.829356313242918e-05, "loss": 0.7883, "step": 6931 }, { "epoch": 2.13, "learning_rate": 1.829300630430613e-05, "loss": 0.9542, "step": 6932 }, { "epoch": 2.13, "learning_rate": 1.8292449393825884e-05, "loss": 0.8705, "step": 6933 }, { "epoch": 2.13, "learning_rate": 1.829189240099397e-05, "loss": 0.7715, "step": 6934 }, { "epoch": 2.13, "learning_rate": 1.8291335325815924e-05, "loss": 0.5418, "step": 6935 }, { "epoch": 2.13, "learning_rate": 1.829077816829728e-05, "loss": 0.5429, "step": 6936 }, { "epoch": 2.13, "learning_rate": 1.8290220928443565e-05, "loss": 0.6904, "step": 6937 }, { "epoch": 2.13, "learning_rate": 1.8289663606260316e-05, "loss": 1.2384, "step": 6938 }, { "epoch": 2.13, "learning_rate": 1.8289106201753067e-05, "loss": 0.8542, "step": 6939 }, { "epoch": 2.14, "learning_rate": 1.8288548714927352e-05, "loss": 0.6237, "step": 6940 }, { "epoch": 2.14, "learning_rate": 1.8287991145788713e-05, "loss": 0.6677, "step": 6941 }, { "epoch": 2.14, "learning_rate": 1.8287433494342683e-05, "loss": 0.8052, "step": 6942 }, { "epoch": 2.14, "learning_rate": 1.8286875760594798e-05, "loss": 0.8721, "step": 6943 }, { "epoch": 2.14, "learning_rate": 1.82863179445506e-05, "loss": 0.7979, "step": 6944 }, { "epoch": 2.14, "learning_rate": 1.828576004621563e-05, "loss": 0.8068, "step": 6945 }, { "epoch": 2.14, "learning_rate": 1.8285202065595425e-05, "loss": 0.946, "step": 6946 }, { "epoch": 2.14, "learning_rate": 1.8284644002695528e-05, "loss": 0.8133, "step": 6947 }, { "epoch": 2.14, "learning_rate": 1.828408585752148e-05, "loss": 0.8589, "step": 6948 }, { "epoch": 2.14, "learning_rate": 1.8283527630078827e-05, "loss": 0.738, "step": 6949 }, { "epoch": 2.14, "learning_rate": 1.8282969320373105e-05, "loss": 0.7513, "step": 6950 }, { "epoch": 2.14, "learning_rate": 1.828241092840987e-05, "loss": 0.611, "step": 6951 }, { "epoch": 2.14, "learning_rate": 1.8281852454194656e-05, "loss": 0.6468, "step": 6952 }, { "epoch": 2.14, "learning_rate": 1.828129389773302e-05, "loss": 0.6237, "step": 6953 }, { "epoch": 2.14, "learning_rate": 1.82807352590305e-05, "loss": 0.9198, "step": 6954 }, { "epoch": 2.14, "learning_rate": 1.8280176538092648e-05, "loss": 0.4305, "step": 6955 }, { "epoch": 2.14, "learning_rate": 1.8279617734925006e-05, "loss": 0.6985, "step": 6956 }, { "epoch": 2.14, "learning_rate": 1.8279058849533135e-05, "loss": 0.6305, "step": 6957 }, { "epoch": 2.14, "learning_rate": 1.827849988192258e-05, "loss": 0.6492, "step": 6958 }, { "epoch": 2.14, "learning_rate": 1.827794083209889e-05, "loss": 0.6774, "step": 6959 }, { "epoch": 2.14, "learning_rate": 1.827738170006762e-05, "loss": 0.7855, "step": 6960 }, { "epoch": 2.14, "learning_rate": 1.8276822485834315e-05, "loss": 0.7587, "step": 6961 }, { "epoch": 2.14, "learning_rate": 1.827626318940454e-05, "loss": 0.7605, "step": 6962 }, { "epoch": 2.14, "learning_rate": 1.8275703810783844e-05, "loss": 0.6322, "step": 6963 }, { "epoch": 2.14, "learning_rate": 1.827514434997778e-05, "loss": 0.9125, "step": 6964 }, { "epoch": 2.14, "learning_rate": 1.827458480699191e-05, "loss": 0.772, "step": 6965 }, { "epoch": 2.14, "learning_rate": 1.827402518183178e-05, "loss": 1.0732, "step": 6966 }, { "epoch": 2.14, "learning_rate": 1.827346547450296e-05, "loss": 0.6746, "step": 6967 }, { "epoch": 2.14, "learning_rate": 1.8272905685011e-05, "loss": 0.9267, "step": 6968 }, { "epoch": 2.14, "learning_rate": 1.8272345813361464e-05, "loss": 0.9506, "step": 6969 }, { "epoch": 2.14, "learning_rate": 1.8271785859559906e-05, "loss": 1.0169, "step": 6970 }, { "epoch": 2.14, "learning_rate": 1.8271225823611894e-05, "loss": 1.0495, "step": 6971 }, { "epoch": 2.15, "learning_rate": 1.8270665705522987e-05, "loss": 0.9201, "step": 6972 }, { "epoch": 2.15, "learning_rate": 1.8270105505298748e-05, "loss": 0.8616, "step": 6973 }, { "epoch": 2.15, "learning_rate": 1.8269545222944738e-05, "loss": 1.0745, "step": 6974 }, { "epoch": 2.15, "learning_rate": 1.8268984858466524e-05, "loss": 0.644, "step": 6975 }, { "epoch": 2.15, "learning_rate": 1.8268424411869667e-05, "loss": 0.5698, "step": 6976 }, { "epoch": 2.15, "learning_rate": 1.8267863883159736e-05, "loss": 0.9395, "step": 6977 }, { "epoch": 2.15, "learning_rate": 1.8267303272342297e-05, "loss": 0.7156, "step": 6978 }, { "epoch": 2.15, "learning_rate": 1.8266742579422914e-05, "loss": 0.7427, "step": 6979 }, { "epoch": 2.15, "learning_rate": 1.8266181804407162e-05, "loss": 0.9086, "step": 6980 }, { "epoch": 2.15, "learning_rate": 1.8265620947300606e-05, "loss": 1.0112, "step": 6981 }, { "epoch": 2.15, "learning_rate": 1.8265060008108815e-05, "loss": 0.7986, "step": 6982 }, { "epoch": 2.15, "learning_rate": 1.8264498986837364e-05, "loss": 0.8022, "step": 6983 }, { "epoch": 2.15, "learning_rate": 1.8263937883491814e-05, "loss": 0.695, "step": 6984 }, { "epoch": 2.15, "learning_rate": 1.826337669807775e-05, "loss": 1.3066, "step": 6985 }, { "epoch": 2.15, "learning_rate": 1.826281543060074e-05, "loss": 0.786, "step": 6986 }, { "epoch": 2.15, "learning_rate": 1.8262254081066356e-05, "loss": 0.7872, "step": 6987 }, { "epoch": 2.15, "learning_rate": 1.8261692649480174e-05, "loss": 0.7685, "step": 6988 }, { "epoch": 2.15, "learning_rate": 1.8261131135847773e-05, "loss": 1.0016, "step": 6989 }, { "epoch": 2.15, "learning_rate": 1.8260569540174724e-05, "loss": 0.9275, "step": 6990 }, { "epoch": 2.15, "learning_rate": 1.8260007862466606e-05, "loss": 0.6318, "step": 6991 }, { "epoch": 2.15, "learning_rate": 1.8259446102729e-05, "loss": 0.5705, "step": 6992 }, { "epoch": 2.15, "learning_rate": 1.825888426096748e-05, "loss": 0.8608, "step": 6993 }, { "epoch": 2.15, "learning_rate": 1.8258322337187622e-05, "loss": 0.7402, "step": 6994 }, { "epoch": 2.15, "learning_rate": 1.8257760331395017e-05, "loss": 0.6998, "step": 6995 }, { "epoch": 2.15, "learning_rate": 1.825719824359524e-05, "loss": 0.7943, "step": 6996 }, { "epoch": 2.15, "learning_rate": 1.825663607379388e-05, "loss": 0.7448, "step": 6997 }, { "epoch": 2.15, "learning_rate": 1.825607382199651e-05, "loss": 0.8581, "step": 6998 }, { "epoch": 2.15, "learning_rate": 1.825551148820872e-05, "loss": 0.7409, "step": 6999 }, { "epoch": 2.15, "learning_rate": 1.8254949072436087e-05, "loss": 0.9558, "step": 7000 }, { "epoch": 2.15, "learning_rate": 1.8254386574684205e-05, "loss": 0.6137, "step": 7001 }, { "epoch": 2.15, "learning_rate": 1.8253823994958657e-05, "loss": 1.0626, "step": 7002 }, { "epoch": 2.15, "learning_rate": 1.8253261333265032e-05, "loss": 0.796, "step": 7003 }, { "epoch": 2.15, "learning_rate": 1.825269858960891e-05, "loss": 0.6796, "step": 7004 }, { "epoch": 2.16, "learning_rate": 1.8252135763995892e-05, "loss": 0.7833, "step": 7005 }, { "epoch": 2.16, "learning_rate": 1.8251572856431554e-05, "loss": 0.7188, "step": 7006 }, { "epoch": 2.16, "learning_rate": 1.8251009866921498e-05, "loss": 0.7504, "step": 7007 }, { "epoch": 2.16, "learning_rate": 1.8250446795471307e-05, "loss": 0.8405, "step": 7008 }, { "epoch": 2.16, "learning_rate": 1.8249883642086573e-05, "loss": 0.8821, "step": 7009 }, { "epoch": 2.16, "learning_rate": 1.8249320406772895e-05, "loss": 0.8025, "step": 7010 }, { "epoch": 2.16, "learning_rate": 1.824875708953586e-05, "loss": 0.7022, "step": 7011 }, { "epoch": 2.16, "learning_rate": 1.8248193690381068e-05, "loss": 0.704, "step": 7012 }, { "epoch": 2.16, "learning_rate": 1.824763020931411e-05, "loss": 0.85, "step": 7013 }, { "epoch": 2.16, "learning_rate": 1.824706664634058e-05, "loss": 0.6007, "step": 7014 }, { "epoch": 2.16, "learning_rate": 1.8246503001466082e-05, "loss": 0.6123, "step": 7015 }, { "epoch": 2.16, "learning_rate": 1.8245939274696204e-05, "loss": 0.7984, "step": 7016 }, { "epoch": 2.16, "learning_rate": 1.8245375466036553e-05, "loss": 1.1378, "step": 7017 }, { "epoch": 2.16, "learning_rate": 1.8244811575492726e-05, "loss": 0.6004, "step": 7018 }, { "epoch": 2.16, "learning_rate": 1.8244247603070314e-05, "loss": 0.596, "step": 7019 }, { "epoch": 2.16, "learning_rate": 1.824368354877493e-05, "loss": 0.7971, "step": 7020 }, { "epoch": 2.16, "learning_rate": 1.8243119412612175e-05, "loss": 0.9439, "step": 7021 }, { "epoch": 2.16, "learning_rate": 1.824255519458764e-05, "loss": 0.7075, "step": 7022 }, { "epoch": 2.16, "learning_rate": 1.8241990894706938e-05, "loss": 0.8819, "step": 7023 }, { "epoch": 2.16, "learning_rate": 1.8241426512975668e-05, "loss": 0.7556, "step": 7024 }, { "epoch": 2.16, "learning_rate": 1.8240862049399442e-05, "loss": 0.8878, "step": 7025 }, { "epoch": 2.16, "learning_rate": 1.8240297503983855e-05, "loss": 0.543, "step": 7026 }, { "epoch": 2.16, "learning_rate": 1.8239732876734525e-05, "loss": 0.6756, "step": 7027 }, { "epoch": 2.16, "learning_rate": 1.8239168167657052e-05, "loss": 0.9694, "step": 7028 }, { "epoch": 2.16, "learning_rate": 1.8238603376757046e-05, "loss": 0.6935, "step": 7029 }, { "epoch": 2.16, "learning_rate": 1.8238038504040113e-05, "loss": 0.6826, "step": 7030 }, { "epoch": 2.16, "learning_rate": 1.8237473549511868e-05, "loss": 1.0961, "step": 7031 }, { "epoch": 2.16, "learning_rate": 1.8236908513177916e-05, "loss": 0.7854, "step": 7032 }, { "epoch": 2.16, "learning_rate": 1.8236343395043876e-05, "loss": 0.6273, "step": 7033 }, { "epoch": 2.16, "learning_rate": 1.823577819511535e-05, "loss": 0.6983, "step": 7034 }, { "epoch": 2.16, "learning_rate": 1.823521291339796e-05, "loss": 1.0477, "step": 7035 }, { "epoch": 2.16, "learning_rate": 1.823464754989731e-05, "loss": 0.738, "step": 7036 }, { "epoch": 2.17, "learning_rate": 1.8234082104619024e-05, "loss": 0.7668, "step": 7037 }, { "epoch": 2.17, "learning_rate": 1.8233516577568715e-05, "loss": 0.739, "step": 7038 }, { "epoch": 2.17, "learning_rate": 1.8232950968751998e-05, "loss": 0.5762, "step": 7039 }, { "epoch": 2.17, "learning_rate": 1.823238527817449e-05, "loss": 0.6579, "step": 7040 }, { "epoch": 2.17, "learning_rate": 1.823181950584181e-05, "loss": 0.6124, "step": 7041 }, { "epoch": 2.17, "learning_rate": 1.8231253651759575e-05, "loss": 0.5516, "step": 7042 }, { "epoch": 2.17, "learning_rate": 1.8230687715933405e-05, "loss": 0.6774, "step": 7043 }, { "epoch": 2.17, "learning_rate": 1.823012169836892e-05, "loss": 0.9235, "step": 7044 }, { "epoch": 2.17, "learning_rate": 1.822955559907174e-05, "loss": 0.6072, "step": 7045 }, { "epoch": 2.17, "learning_rate": 1.822898941804749e-05, "loss": 0.7549, "step": 7046 }, { "epoch": 2.17, "learning_rate": 1.822842315530179e-05, "loss": 0.9371, "step": 7047 }, { "epoch": 2.17, "learning_rate": 1.8227856810840268e-05, "loss": 0.6257, "step": 7048 }, { "epoch": 2.17, "learning_rate": 1.822729038466854e-05, "loss": 0.8062, "step": 7049 }, { "epoch": 2.17, "learning_rate": 1.822672387679224e-05, "loss": 0.8643, "step": 7050 }, { "epoch": 2.17, "learning_rate": 1.8226157287216988e-05, "loss": 0.7891, "step": 7051 }, { "epoch": 2.17, "learning_rate": 1.8225590615948413e-05, "loss": 0.7164, "step": 7052 }, { "epoch": 2.17, "learning_rate": 1.822502386299214e-05, "loss": 1.1233, "step": 7053 }, { "epoch": 2.17, "learning_rate": 1.8224457028353805e-05, "loss": 0.6425, "step": 7054 }, { "epoch": 2.17, "learning_rate": 1.8223890112039026e-05, "loss": 0.8771, "step": 7055 }, { "epoch": 2.17, "learning_rate": 1.8223323114053442e-05, "loss": 0.7014, "step": 7056 }, { "epoch": 2.17, "learning_rate": 1.822275603440268e-05, "loss": 0.6358, "step": 7057 }, { "epoch": 2.17, "learning_rate": 1.8222188873092374e-05, "loss": 0.8674, "step": 7058 }, { "epoch": 2.17, "learning_rate": 1.822162163012815e-05, "loss": 0.634, "step": 7059 }, { "epoch": 2.17, "learning_rate": 1.8221054305515648e-05, "loss": 0.7489, "step": 7060 }, { "epoch": 2.17, "learning_rate": 1.8220486899260502e-05, "loss": 0.6487, "step": 7061 }, { "epoch": 2.17, "learning_rate": 1.8219919411368344e-05, "loss": 0.8337, "step": 7062 }, { "epoch": 2.17, "learning_rate": 1.821935184184481e-05, "loss": 0.9345, "step": 7063 }, { "epoch": 2.17, "learning_rate": 1.8218784190695538e-05, "loss": 0.8351, "step": 7064 }, { "epoch": 2.17, "learning_rate": 1.8218216457926162e-05, "loss": 0.6805, "step": 7065 }, { "epoch": 2.17, "learning_rate": 1.8217648643542326e-05, "loss": 0.8382, "step": 7066 }, { "epoch": 2.17, "learning_rate": 1.821708074754966e-05, "loss": 0.5232, "step": 7067 }, { "epoch": 2.17, "learning_rate": 1.8216512769953817e-05, "loss": 0.833, "step": 7068 }, { "epoch": 2.17, "learning_rate": 1.8215944710760423e-05, "loss": 0.7921, "step": 7069 }, { "epoch": 2.18, "learning_rate": 1.8215376569975127e-05, "loss": 0.7444, "step": 7070 }, { "epoch": 2.18, "learning_rate": 1.821480834760357e-05, "loss": 0.7185, "step": 7071 }, { "epoch": 2.18, "learning_rate": 1.8214240043651398e-05, "loss": 0.7599, "step": 7072 }, { "epoch": 2.18, "learning_rate": 1.8213671658124253e-05, "loss": 0.9958, "step": 7073 }, { "epoch": 2.18, "learning_rate": 1.821310319102777e-05, "loss": 0.8195, "step": 7074 }, { "epoch": 2.18, "learning_rate": 1.821253464236761e-05, "loss": 0.7771, "step": 7075 }, { "epoch": 2.18, "learning_rate": 1.8211966012149415e-05, "loss": 0.7656, "step": 7076 }, { "epoch": 2.18, "learning_rate": 1.8211397300378823e-05, "loss": 0.5368, "step": 7077 }, { "epoch": 2.18, "learning_rate": 1.821082850706149e-05, "loss": 0.7542, "step": 7078 }, { "epoch": 2.18, "learning_rate": 1.8210259632203063e-05, "loss": 1.0235, "step": 7079 }, { "epoch": 2.18, "learning_rate": 1.820969067580919e-05, "loss": 0.8266, "step": 7080 }, { "epoch": 2.18, "learning_rate": 1.8209121637885522e-05, "loss": 0.6718, "step": 7081 }, { "epoch": 2.18, "learning_rate": 1.8208552518437712e-05, "loss": 0.7333, "step": 7082 }, { "epoch": 2.18, "learning_rate": 1.820798331747141e-05, "loss": 0.7638, "step": 7083 }, { "epoch": 2.18, "learning_rate": 1.820741403499227e-05, "loss": 0.8316, "step": 7084 }, { "epoch": 2.18, "learning_rate": 1.820684467100594e-05, "loss": 0.7865, "step": 7085 }, { "epoch": 2.18, "learning_rate": 1.820627522551808e-05, "loss": 0.5662, "step": 7086 }, { "epoch": 2.18, "learning_rate": 1.8205705698534348e-05, "loss": 0.5567, "step": 7087 }, { "epoch": 2.18, "learning_rate": 1.8205136090060393e-05, "loss": 0.6915, "step": 7088 }, { "epoch": 2.18, "learning_rate": 1.8204566400101876e-05, "loss": 0.764, "step": 7089 }, { "epoch": 2.18, "learning_rate": 1.820399662866445e-05, "loss": 0.762, "step": 7090 }, { "epoch": 2.18, "learning_rate": 1.820342677575378e-05, "loss": 0.8817, "step": 7091 }, { "epoch": 2.18, "learning_rate": 1.8202856841375517e-05, "loss": 0.9, "step": 7092 }, { "epoch": 2.18, "learning_rate": 1.820228682553533e-05, "loss": 0.7331, "step": 7093 }, { "epoch": 2.18, "learning_rate": 1.8201716728238876e-05, "loss": 0.5693, "step": 7094 }, { "epoch": 2.18, "learning_rate": 1.8201146549491815e-05, "loss": 0.6808, "step": 7095 }, { "epoch": 2.18, "learning_rate": 1.8200576289299808e-05, "loss": 0.8422, "step": 7096 }, { "epoch": 2.18, "learning_rate": 1.820000594766852e-05, "loss": 0.6415, "step": 7097 }, { "epoch": 2.18, "learning_rate": 1.8199435524603618e-05, "loss": 0.716, "step": 7098 }, { "epoch": 2.18, "learning_rate": 1.8198865020110765e-05, "loss": 0.7119, "step": 7099 }, { "epoch": 2.18, "learning_rate": 1.8198294434195626e-05, "loss": 0.7128, "step": 7100 }, { "epoch": 2.18, "learning_rate": 1.8197723766863867e-05, "loss": 0.7497, "step": 7101 }, { "epoch": 2.19, "learning_rate": 1.8197153018121156e-05, "loss": 0.6708, "step": 7102 }, { "epoch": 2.19, "learning_rate": 1.819658218797316e-05, "loss": 0.7381, "step": 7103 }, { "epoch": 2.19, "learning_rate": 1.8196011276425547e-05, "loss": 0.6765, "step": 7104 }, { "epoch": 2.19, "learning_rate": 1.819544028348399e-05, "loss": 0.8549, "step": 7105 }, { "epoch": 2.19, "learning_rate": 1.819486920915416e-05, "loss": 0.6616, "step": 7106 }, { "epoch": 2.19, "learning_rate": 1.8194298053441725e-05, "loss": 0.7197, "step": 7107 }, { "epoch": 2.19, "learning_rate": 1.8193726816352356e-05, "loss": 0.8884, "step": 7108 }, { "epoch": 2.19, "learning_rate": 1.8193155497891727e-05, "loss": 0.6075, "step": 7109 }, { "epoch": 2.19, "learning_rate": 1.8192584098065516e-05, "loss": 0.8727, "step": 7110 }, { "epoch": 2.19, "learning_rate": 1.8192012616879395e-05, "loss": 1.1906, "step": 7111 }, { "epoch": 2.19, "learning_rate": 1.8191441054339036e-05, "loss": 0.9233, "step": 7112 }, { "epoch": 2.19, "learning_rate": 1.8190869410450117e-05, "loss": 0.6831, "step": 7113 }, { "epoch": 2.19, "learning_rate": 1.819029768521832e-05, "loss": 0.9237, "step": 7114 }, { "epoch": 2.19, "learning_rate": 1.8189725878649317e-05, "loss": 0.7628, "step": 7115 }, { "epoch": 2.19, "learning_rate": 1.818915399074879e-05, "loss": 0.9673, "step": 7116 }, { "epoch": 2.19, "learning_rate": 1.8188582021522414e-05, "loss": 0.6236, "step": 7117 }, { "epoch": 2.19, "learning_rate": 1.818800997097587e-05, "loss": 0.9106, "step": 7118 }, { "epoch": 2.19, "learning_rate": 1.8187437839114842e-05, "loss": 0.8298, "step": 7119 }, { "epoch": 2.19, "learning_rate": 1.8186865625945013e-05, "loss": 0.911, "step": 7120 }, { "epoch": 2.19, "learning_rate": 1.8186293331472063e-05, "loss": 0.7721, "step": 7121 }, { "epoch": 2.19, "learning_rate": 1.818572095570167e-05, "loss": 0.7384, "step": 7122 }, { "epoch": 2.19, "learning_rate": 1.8185148498639532e-05, "loss": 0.8509, "step": 7123 }, { "epoch": 2.19, "learning_rate": 1.818457596029132e-05, "loss": 0.9811, "step": 7124 }, { "epoch": 2.19, "learning_rate": 1.8184003340662725e-05, "loss": 1.0471, "step": 7125 }, { "epoch": 2.19, "learning_rate": 1.818343063975944e-05, "loss": 0.6372, "step": 7126 }, { "epoch": 2.19, "learning_rate": 1.818285785758714e-05, "loss": 0.899, "step": 7127 }, { "epoch": 2.19, "learning_rate": 1.8182284994151525e-05, "loss": 0.6642, "step": 7128 }, { "epoch": 2.19, "learning_rate": 1.8181712049458276e-05, "loss": 0.6372, "step": 7129 }, { "epoch": 2.19, "learning_rate": 1.8181139023513087e-05, "loss": 0.7268, "step": 7130 }, { "epoch": 2.19, "learning_rate": 1.8180565916321646e-05, "loss": 0.664, "step": 7131 }, { "epoch": 2.19, "learning_rate": 1.817999272788965e-05, "loss": 0.8725, "step": 7132 }, { "epoch": 2.19, "learning_rate": 1.817941945822278e-05, "loss": 0.5681, "step": 7133 }, { "epoch": 2.19, "learning_rate": 1.8178846107326744e-05, "loss": 0.5861, "step": 7134 }, { "epoch": 2.2, "learning_rate": 1.8178272675207223e-05, "loss": 0.7287, "step": 7135 }, { "epoch": 2.2, "learning_rate": 1.8177699161869922e-05, "loss": 0.6352, "step": 7136 }, { "epoch": 2.2, "learning_rate": 1.8177125567320528e-05, "loss": 0.9479, "step": 7137 }, { "epoch": 2.2, "learning_rate": 1.8176551891564743e-05, "loss": 0.9058, "step": 7138 }, { "epoch": 2.2, "learning_rate": 1.817597813460826e-05, "loss": 0.5882, "step": 7139 }, { "epoch": 2.2, "learning_rate": 1.817540429645678e-05, "loss": 0.7367, "step": 7140 }, { "epoch": 2.2, "learning_rate": 1.8174830377115998e-05, "loss": 0.7064, "step": 7141 }, { "epoch": 2.2, "learning_rate": 1.817425637659162e-05, "loss": 0.8, "step": 7142 }, { "epoch": 2.2, "learning_rate": 1.817368229488934e-05, "loss": 0.8154, "step": 7143 }, { "epoch": 2.2, "learning_rate": 1.817310813201486e-05, "loss": 1.0181, "step": 7144 }, { "epoch": 2.2, "learning_rate": 1.817253388797389e-05, "loss": 0.8678, "step": 7145 }, { "epoch": 2.2, "learning_rate": 1.8171959562772125e-05, "loss": 0.7646, "step": 7146 }, { "epoch": 2.2, "learning_rate": 1.8171385156415265e-05, "loss": 0.6052, "step": 7147 }, { "epoch": 2.2, "learning_rate": 1.8170810668909023e-05, "loss": 0.833, "step": 7148 }, { "epoch": 2.2, "learning_rate": 1.8170236100259098e-05, "loss": 0.658, "step": 7149 }, { "epoch": 2.2, "learning_rate": 1.8169661450471203e-05, "loss": 0.7159, "step": 7150 }, { "epoch": 2.2, "learning_rate": 1.816908671955104e-05, "loss": 0.7677, "step": 7151 }, { "epoch": 2.2, "learning_rate": 1.8168511907504314e-05, "loss": 0.685, "step": 7152 }, { "epoch": 2.2, "learning_rate": 1.8167937014336735e-05, "loss": 0.804, "step": 7153 }, { "epoch": 2.2, "learning_rate": 1.8167362040054017e-05, "loss": 0.7507, "step": 7154 }, { "epoch": 2.2, "learning_rate": 1.8166786984661864e-05, "loss": 1.0087, "step": 7155 }, { "epoch": 2.2, "learning_rate": 1.8166211848165992e-05, "loss": 0.8547, "step": 7156 }, { "epoch": 2.2, "learning_rate": 1.816563663057211e-05, "loss": 1.0433, "step": 7157 }, { "epoch": 2.2, "learning_rate": 1.816506133188593e-05, "loss": 0.8392, "step": 7158 }, { "epoch": 2.2, "learning_rate": 1.8164485952113165e-05, "loss": 0.7981, "step": 7159 }, { "epoch": 2.2, "learning_rate": 1.816391049125953e-05, "loss": 0.7742, "step": 7160 }, { "epoch": 2.2, "learning_rate": 1.8163334949330745e-05, "loss": 0.8507, "step": 7161 }, { "epoch": 2.2, "learning_rate": 1.8162759326332514e-05, "loss": 0.4393, "step": 7162 }, { "epoch": 2.2, "learning_rate": 1.8162183622270563e-05, "loss": 0.6672, "step": 7163 }, { "epoch": 2.2, "learning_rate": 1.8161607837150605e-05, "loss": 0.704, "step": 7164 }, { "epoch": 2.2, "learning_rate": 1.816103197097836e-05, "loss": 0.7897, "step": 7165 }, { "epoch": 2.2, "learning_rate": 1.816045602375954e-05, "loss": 0.6568, "step": 7166 }, { "epoch": 2.21, "learning_rate": 1.815987999549988e-05, "loss": 0.7985, "step": 7167 }, { "epoch": 2.21, "learning_rate": 1.8159303886205086e-05, "loss": 0.5463, "step": 7168 }, { "epoch": 2.21, "learning_rate": 1.8158727695880887e-05, "loss": 0.6716, "step": 7169 }, { "epoch": 2.21, "learning_rate": 1.8158151424533002e-05, "loss": 0.6108, "step": 7170 }, { "epoch": 2.21, "learning_rate": 1.815757507216715e-05, "loss": 0.8502, "step": 7171 }, { "epoch": 2.21, "learning_rate": 1.8156998638789067e-05, "loss": 0.7338, "step": 7172 }, { "epoch": 2.21, "learning_rate": 1.8156422124404466e-05, "loss": 0.7796, "step": 7173 }, { "epoch": 2.21, "learning_rate": 1.8155845529019075e-05, "loss": 0.6821, "step": 7174 }, { "epoch": 2.21, "learning_rate": 1.8155268852638623e-05, "loss": 0.6311, "step": 7175 }, { "epoch": 2.21, "learning_rate": 1.8154692095268837e-05, "loss": 0.7585, "step": 7176 }, { "epoch": 2.21, "learning_rate": 1.8154115256915437e-05, "loss": 0.6935, "step": 7177 }, { "epoch": 2.21, "learning_rate": 1.8153538337584163e-05, "loss": 0.8441, "step": 7178 }, { "epoch": 2.21, "learning_rate": 1.8152961337280738e-05, "loss": 0.9635, "step": 7179 }, { "epoch": 2.21, "learning_rate": 1.8152384256010893e-05, "loss": 0.6631, "step": 7180 }, { "epoch": 2.21, "learning_rate": 1.8151807093780354e-05, "loss": 0.7375, "step": 7181 }, { "epoch": 2.21, "learning_rate": 1.8151229850594862e-05, "loss": 0.6332, "step": 7182 }, { "epoch": 2.21, "learning_rate": 1.8150652526460146e-05, "loss": 0.8408, "step": 7183 }, { "epoch": 2.21, "learning_rate": 1.8150075121381938e-05, "loss": 0.6239, "step": 7184 }, { "epoch": 2.21, "learning_rate": 1.814949763536597e-05, "loss": 0.9439, "step": 7185 }, { "epoch": 2.21, "learning_rate": 1.814892006841798e-05, "loss": 0.9377, "step": 7186 }, { "epoch": 2.21, "learning_rate": 1.814834242054371e-05, "loss": 0.9329, "step": 7187 }, { "epoch": 2.21, "learning_rate": 1.814776469174888e-05, "loss": 0.8138, "step": 7188 }, { "epoch": 2.21, "learning_rate": 1.8147186882039247e-05, "loss": 0.8912, "step": 7189 }, { "epoch": 2.21, "learning_rate": 1.8146608991420533e-05, "loss": 1.0117, "step": 7190 }, { "epoch": 2.21, "learning_rate": 1.8146031019898485e-05, "loss": 0.6485, "step": 7191 }, { "epoch": 2.21, "learning_rate": 1.814545296747884e-05, "loss": 0.7164, "step": 7192 }, { "epoch": 2.21, "learning_rate": 1.8144874834167342e-05, "loss": 0.8533, "step": 7193 }, { "epoch": 2.21, "learning_rate": 1.8144296619969734e-05, "loss": 0.7844, "step": 7194 }, { "epoch": 2.21, "learning_rate": 1.814371832489175e-05, "loss": 0.597, "step": 7195 }, { "epoch": 2.21, "learning_rate": 1.8143139948939138e-05, "loss": 1.0126, "step": 7196 }, { "epoch": 2.21, "learning_rate": 1.814256149211764e-05, "loss": 0.9233, "step": 7197 }, { "epoch": 2.21, "learning_rate": 1.8141982954433002e-05, "loss": 0.7298, "step": 7198 }, { "epoch": 2.21, "learning_rate": 1.8141404335890973e-05, "loss": 0.7452, "step": 7199 }, { "epoch": 2.22, "learning_rate": 1.8140825636497293e-05, "loss": 0.7684, "step": 7200 }, { "epoch": 2.22, "learning_rate": 1.8140246856257714e-05, "loss": 1.1157, "step": 7201 }, { "epoch": 2.22, "learning_rate": 1.813966799517798e-05, "loss": 0.6497, "step": 7202 }, { "epoch": 2.22, "learning_rate": 1.813908905326384e-05, "loss": 0.6751, "step": 7203 }, { "epoch": 2.22, "learning_rate": 1.8138510030521047e-05, "loss": 0.7357, "step": 7204 }, { "epoch": 2.22, "learning_rate": 1.8137930926955344e-05, "loss": 0.7822, "step": 7205 }, { "epoch": 2.22, "learning_rate": 1.813735174257249e-05, "loss": 0.8273, "step": 7206 }, { "epoch": 2.22, "learning_rate": 1.8136772477378234e-05, "loss": 1.146, "step": 7207 }, { "epoch": 2.22, "learning_rate": 1.813619313137833e-05, "loss": 0.7051, "step": 7208 }, { "epoch": 2.22, "learning_rate": 1.8135613704578525e-05, "loss": 0.8493, "step": 7209 }, { "epoch": 2.22, "learning_rate": 1.813503419698458e-05, "loss": 0.8412, "step": 7210 }, { "epoch": 2.22, "learning_rate": 1.813445460860225e-05, "loss": 0.9086, "step": 7211 }, { "epoch": 2.22, "learning_rate": 1.8133874939437288e-05, "loss": 0.968, "step": 7212 }, { "epoch": 2.22, "learning_rate": 1.8133295189495447e-05, "loss": 0.6168, "step": 7213 }, { "epoch": 2.22, "learning_rate": 1.8132715358782493e-05, "loss": 0.8494, "step": 7214 }, { "epoch": 2.22, "learning_rate": 1.813213544730418e-05, "loss": 0.8433, "step": 7215 }, { "epoch": 2.22, "learning_rate": 1.8131555455066267e-05, "loss": 0.9187, "step": 7216 }, { "epoch": 2.22, "learning_rate": 1.8130975382074514e-05, "loss": 0.7967, "step": 7217 }, { "epoch": 2.22, "learning_rate": 1.813039522833468e-05, "loss": 0.5511, "step": 7218 }, { "epoch": 2.22, "learning_rate": 1.812981499385253e-05, "loss": 0.805, "step": 7219 }, { "epoch": 2.22, "learning_rate": 1.8129234678633824e-05, "loss": 0.4901, "step": 7220 }, { "epoch": 2.22, "learning_rate": 1.812865428268433e-05, "loss": 0.8674, "step": 7221 }, { "epoch": 2.22, "learning_rate": 1.81280738060098e-05, "loss": 0.6675, "step": 7222 }, { "epoch": 2.22, "learning_rate": 1.812749324861601e-05, "loss": 0.7983, "step": 7223 }, { "epoch": 2.22, "learning_rate": 1.812691261050872e-05, "loss": 0.7311, "step": 7224 }, { "epoch": 2.22, "learning_rate": 1.81263318916937e-05, "loss": 0.8175, "step": 7225 }, { "epoch": 2.22, "learning_rate": 1.812575109217671e-05, "loss": 0.9628, "step": 7226 }, { "epoch": 2.22, "learning_rate": 1.812517021196353e-05, "loss": 0.704, "step": 7227 }, { "epoch": 2.22, "learning_rate": 1.8124589251059914e-05, "loss": 0.9627, "step": 7228 }, { "epoch": 2.22, "learning_rate": 1.8124008209471644e-05, "loss": 0.657, "step": 7229 }, { "epoch": 2.22, "learning_rate": 1.8123427087204485e-05, "loss": 0.6678, "step": 7230 }, { "epoch": 2.22, "learning_rate": 1.8122845884264208e-05, "loss": 0.6499, "step": 7231 }, { "epoch": 2.23, "learning_rate": 1.812226460065658e-05, "loss": 0.7637, "step": 7232 }, { "epoch": 2.23, "learning_rate": 1.8121683236387383e-05, "loss": 0.8663, "step": 7233 }, { "epoch": 2.23, "learning_rate": 1.8121101791462386e-05, "loss": 0.8344, "step": 7234 }, { "epoch": 2.23, "learning_rate": 1.8120520265887364e-05, "loss": 0.7332, "step": 7235 }, { "epoch": 2.23, "learning_rate": 1.8119938659668094e-05, "loss": 0.57, "step": 7236 }, { "epoch": 2.23, "learning_rate": 1.8119356972810344e-05, "loss": 0.7907, "step": 7237 }, { "epoch": 2.23, "learning_rate": 1.81187752053199e-05, "loss": 0.8604, "step": 7238 }, { "epoch": 2.23, "learning_rate": 1.8118193357202535e-05, "loss": 0.8648, "step": 7239 }, { "epoch": 2.23, "learning_rate": 1.8117611428464024e-05, "loss": 0.9807, "step": 7240 }, { "epoch": 2.23, "learning_rate": 1.8117029419110156e-05, "loss": 0.6664, "step": 7241 }, { "epoch": 2.23, "learning_rate": 1.8116447329146704e-05, "loss": 0.8033, "step": 7242 }, { "epoch": 2.23, "learning_rate": 1.811586515857945e-05, "loss": 0.9889, "step": 7243 }, { "epoch": 2.23, "learning_rate": 1.8115282907414172e-05, "loss": 0.7925, "step": 7244 }, { "epoch": 2.23, "learning_rate": 1.8114700575656657e-05, "loss": 0.6631, "step": 7245 }, { "epoch": 2.23, "learning_rate": 1.8114118163312685e-05, "loss": 0.9214, "step": 7246 }, { "epoch": 2.23, "learning_rate": 1.8113535670388046e-05, "loss": 0.7488, "step": 7247 }, { "epoch": 2.23, "learning_rate": 1.8112953096888517e-05, "loss": 0.6133, "step": 7248 }, { "epoch": 2.23, "learning_rate": 1.8112370442819888e-05, "loss": 0.6467, "step": 7249 }, { "epoch": 2.23, "learning_rate": 1.8111787708187943e-05, "loss": 0.5824, "step": 7250 }, { "epoch": 2.23, "learning_rate": 1.811120489299847e-05, "loss": 0.6692, "step": 7251 }, { "epoch": 2.23, "learning_rate": 1.8110621997257256e-05, "loss": 0.7545, "step": 7252 }, { "epoch": 2.23, "learning_rate": 1.811003902097009e-05, "loss": 0.7159, "step": 7253 }, { "epoch": 2.23, "learning_rate": 1.8109455964142764e-05, "loss": 0.9324, "step": 7254 }, { "epoch": 2.23, "learning_rate": 1.8108872826781066e-05, "loss": 0.8352, "step": 7255 }, { "epoch": 2.23, "learning_rate": 1.8108289608890787e-05, "loss": 0.8317, "step": 7256 }, { "epoch": 2.23, "learning_rate": 1.810770631047772e-05, "loss": 0.8389, "step": 7257 }, { "epoch": 2.23, "learning_rate": 1.8107122931547657e-05, "loss": 0.808, "step": 7258 }, { "epoch": 2.23, "learning_rate": 1.8106539472106394e-05, "loss": 0.7557, "step": 7259 }, { "epoch": 2.23, "learning_rate": 1.8105955932159717e-05, "loss": 0.9908, "step": 7260 }, { "epoch": 2.23, "learning_rate": 1.810537231171343e-05, "loss": 0.7913, "step": 7261 }, { "epoch": 2.23, "learning_rate": 1.8104788610773327e-05, "loss": 0.8648, "step": 7262 }, { "epoch": 2.23, "learning_rate": 1.81042048293452e-05, "loss": 0.5964, "step": 7263 }, { "epoch": 2.23, "learning_rate": 1.8103620967434854e-05, "loss": 0.7319, "step": 7264 }, { "epoch": 2.24, "learning_rate": 1.8103037025048083e-05, "loss": 0.6971, "step": 7265 }, { "epoch": 2.24, "learning_rate": 1.8102453002190684e-05, "loss": 0.7777, "step": 7266 }, { "epoch": 2.24, "learning_rate": 1.810186889886846e-05, "loss": 0.9221, "step": 7267 }, { "epoch": 2.24, "learning_rate": 1.8101284715087212e-05, "loss": 1.0125, "step": 7268 }, { "epoch": 2.24, "learning_rate": 1.810070045085274e-05, "loss": 0.8828, "step": 7269 }, { "epoch": 2.24, "learning_rate": 1.8100116106170847e-05, "loss": 0.6018, "step": 7270 }, { "epoch": 2.24, "learning_rate": 1.8099531681047334e-05, "loss": 0.6596, "step": 7271 }, { "epoch": 2.24, "learning_rate": 1.8098947175488007e-05, "loss": 0.6095, "step": 7272 }, { "epoch": 2.24, "learning_rate": 1.8098362589498674e-05, "loss": 0.8001, "step": 7273 }, { "epoch": 2.24, "learning_rate": 1.809777792308513e-05, "loss": 0.7557, "step": 7274 }, { "epoch": 2.24, "learning_rate": 1.8097193176253195e-05, "loss": 0.5348, "step": 7275 }, { "epoch": 2.24, "learning_rate": 1.8096608349008665e-05, "loss": 0.8302, "step": 7276 }, { "epoch": 2.24, "learning_rate": 1.8096023441357353e-05, "loss": 0.8905, "step": 7277 }, { "epoch": 2.24, "learning_rate": 1.8095438453305067e-05, "loss": 0.8301, "step": 7278 }, { "epoch": 2.24, "learning_rate": 1.8094853384857615e-05, "loss": 0.7946, "step": 7279 }, { "epoch": 2.24, "learning_rate": 1.809426823602081e-05, "loss": 0.831, "step": 7280 }, { "epoch": 2.24, "learning_rate": 1.809368300680046e-05, "loss": 0.7055, "step": 7281 }, { "epoch": 2.24, "learning_rate": 1.8093097697202383e-05, "loss": 0.8077, "step": 7282 }, { "epoch": 2.24, "learning_rate": 1.809251230723238e-05, "loss": 0.7405, "step": 7283 }, { "epoch": 2.24, "learning_rate": 1.8091926836896277e-05, "loss": 1.0411, "step": 7284 }, { "epoch": 2.24, "learning_rate": 1.809134128619988e-05, "loss": 1.1409, "step": 7285 }, { "epoch": 2.24, "learning_rate": 1.8090755655149007e-05, "loss": 0.7612, "step": 7286 }, { "epoch": 2.24, "learning_rate": 1.8090169943749477e-05, "loss": 0.8081, "step": 7287 }, { "epoch": 2.24, "learning_rate": 1.80895841520071e-05, "loss": 0.7363, "step": 7288 }, { "epoch": 2.24, "learning_rate": 1.8088998279927696e-05, "loss": 0.6164, "step": 7289 }, { "epoch": 2.24, "learning_rate": 1.8088412327517087e-05, "loss": 0.883, "step": 7290 }, { "epoch": 2.24, "learning_rate": 1.8087826294781087e-05, "loss": 0.7242, "step": 7291 }, { "epoch": 2.24, "learning_rate": 1.808724018172552e-05, "loss": 0.8812, "step": 7292 }, { "epoch": 2.24, "learning_rate": 1.8086653988356202e-05, "loss": 0.7608, "step": 7293 }, { "epoch": 2.24, "learning_rate": 1.8086067714678956e-05, "loss": 0.6673, "step": 7294 }, { "epoch": 2.24, "learning_rate": 1.808548136069961e-05, "loss": 0.8051, "step": 7295 }, { "epoch": 2.24, "learning_rate": 1.808489492642398e-05, "loss": 0.8366, "step": 7296 }, { "epoch": 2.25, "learning_rate": 1.8084308411857893e-05, "loss": 1.1724, "step": 7297 }, { "epoch": 2.25, "learning_rate": 1.808372181700717e-05, "loss": 0.6839, "step": 7298 }, { "epoch": 2.25, "learning_rate": 1.808313514187764e-05, "loss": 0.7709, "step": 7299 }, { "epoch": 2.25, "learning_rate": 1.808254838647513e-05, "loss": 0.8039, "step": 7300 }, { "epoch": 2.25, "learning_rate": 1.8081961550805465e-05, "loss": 0.7535, "step": 7301 }, { "epoch": 2.25, "learning_rate": 1.8081374634874472e-05, "loss": 1.0003, "step": 7302 }, { "epoch": 2.25, "learning_rate": 1.808078763868798e-05, "loss": 0.8467, "step": 7303 }, { "epoch": 2.25, "learning_rate": 1.808020056225182e-05, "loss": 0.7487, "step": 7304 }, { "epoch": 2.25, "learning_rate": 1.8079613405571824e-05, "loss": 0.6356, "step": 7305 }, { "epoch": 2.25, "learning_rate": 1.807902616865382e-05, "loss": 0.7485, "step": 7306 }, { "epoch": 2.25, "learning_rate": 1.8078438851503638e-05, "loss": 0.5627, "step": 7307 }, { "epoch": 2.25, "learning_rate": 1.8077851454127114e-05, "loss": 0.738, "step": 7308 }, { "epoch": 2.25, "learning_rate": 1.807726397653008e-05, "loss": 0.5945, "step": 7309 }, { "epoch": 2.25, "learning_rate": 1.807667641871837e-05, "loss": 0.8529, "step": 7310 }, { "epoch": 2.25, "learning_rate": 1.807608878069782e-05, "loss": 0.6804, "step": 7311 }, { "epoch": 2.25, "learning_rate": 1.8075501062474266e-05, "loss": 0.6105, "step": 7312 }, { "epoch": 2.25, "learning_rate": 1.8074913264053547e-05, "loss": 0.6076, "step": 7313 }, { "epoch": 2.25, "learning_rate": 1.807432538544149e-05, "loss": 0.7836, "step": 7314 }, { "epoch": 2.25, "learning_rate": 1.8073737426643947e-05, "loss": 0.8784, "step": 7315 }, { "epoch": 2.25, "learning_rate": 1.8073149387666745e-05, "loss": 0.496, "step": 7316 }, { "epoch": 2.25, "learning_rate": 1.8072561268515733e-05, "loss": 0.801, "step": 7317 }, { "epoch": 2.25, "learning_rate": 1.8071973069196746e-05, "loss": 0.8142, "step": 7318 }, { "epoch": 2.25, "learning_rate": 1.8071384789715628e-05, "loss": 0.6916, "step": 7319 }, { "epoch": 2.25, "learning_rate": 1.807079643007822e-05, "loss": 0.8846, "step": 7320 }, { "epoch": 2.25, "learning_rate": 1.8070207990290364e-05, "loss": 0.8096, "step": 7321 }, { "epoch": 2.25, "learning_rate": 1.8069619470357906e-05, "loss": 0.746, "step": 7322 }, { "epoch": 2.25, "learning_rate": 1.806903087028669e-05, "loss": 0.7472, "step": 7323 }, { "epoch": 2.25, "learning_rate": 1.806844219008256e-05, "loss": 0.7214, "step": 7324 }, { "epoch": 2.25, "learning_rate": 1.806785342975136e-05, "loss": 0.7649, "step": 7325 }, { "epoch": 2.25, "learning_rate": 1.8067264589298945e-05, "loss": 0.6798, "step": 7326 }, { "epoch": 2.25, "learning_rate": 1.8066675668731156e-05, "loss": 0.8708, "step": 7327 }, { "epoch": 2.25, "learning_rate": 1.806608666805384e-05, "loss": 0.5673, "step": 7328 }, { "epoch": 2.25, "learning_rate": 1.8065497587272855e-05, "loss": 0.6352, "step": 7329 }, { "epoch": 2.26, "learning_rate": 1.8064908426394043e-05, "loss": 0.7973, "step": 7330 }, { "epoch": 2.26, "learning_rate": 1.8064319185423255e-05, "loss": 0.8961, "step": 7331 }, { "epoch": 2.26, "learning_rate": 1.806372986436635e-05, "loss": 0.7466, "step": 7332 }, { "epoch": 2.26, "learning_rate": 1.8063140463229172e-05, "loss": 0.6748, "step": 7333 }, { "epoch": 2.26, "learning_rate": 1.8062550982017577e-05, "loss": 0.6834, "step": 7334 }, { "epoch": 2.26, "learning_rate": 1.806196142073742e-05, "loss": 0.6662, "step": 7335 }, { "epoch": 2.26, "learning_rate": 1.8061371779394565e-05, "loss": 0.6535, "step": 7336 }, { "epoch": 2.26, "learning_rate": 1.806078205799485e-05, "loss": 0.6614, "step": 7337 }, { "epoch": 2.26, "learning_rate": 1.806019225654414e-05, "loss": 0.7378, "step": 7338 }, { "epoch": 2.26, "learning_rate": 1.8059602375048294e-05, "loss": 0.7751, "step": 7339 }, { "epoch": 2.26, "learning_rate": 1.8059012413513168e-05, "loss": 0.8206, "step": 7340 }, { "epoch": 2.26, "learning_rate": 1.8058422371944625e-05, "loss": 0.7723, "step": 7341 }, { "epoch": 2.26, "learning_rate": 1.805783225034852e-05, "loss": 0.7648, "step": 7342 }, { "epoch": 2.26, "learning_rate": 1.805724204873071e-05, "loss": 0.8542, "step": 7343 }, { "epoch": 2.26, "learning_rate": 1.805665176709706e-05, "loss": 0.8499, "step": 7344 }, { "epoch": 2.26, "learning_rate": 1.8056061405453437e-05, "loss": 0.663, "step": 7345 }, { "epoch": 2.26, "learning_rate": 1.80554709638057e-05, "loss": 0.7407, "step": 7346 }, { "epoch": 2.26, "learning_rate": 1.805488044215971e-05, "loss": 1.1139, "step": 7347 }, { "epoch": 2.26, "learning_rate": 1.8054289840521335e-05, "loss": 0.5552, "step": 7348 }, { "epoch": 2.26, "learning_rate": 1.805369915889644e-05, "loss": 0.8393, "step": 7349 }, { "epoch": 2.26, "learning_rate": 1.8053108397290887e-05, "loss": 0.7215, "step": 7350 }, { "epoch": 2.26, "learning_rate": 1.8052517555710544e-05, "loss": 0.6803, "step": 7351 }, { "epoch": 2.26, "learning_rate": 1.8051926634161282e-05, "loss": 0.6321, "step": 7352 }, { "epoch": 2.26, "learning_rate": 1.8051335632648973e-05, "loss": 0.8987, "step": 7353 }, { "epoch": 2.26, "learning_rate": 1.8050744551179474e-05, "loss": 0.7776, "step": 7354 }, { "epoch": 2.26, "learning_rate": 1.8050153389758663e-05, "loss": 0.9218, "step": 7355 }, { "epoch": 2.26, "learning_rate": 1.8049562148392414e-05, "loss": 0.82, "step": 7356 }, { "epoch": 2.26, "learning_rate": 1.804897082708659e-05, "loss": 0.8111, "step": 7357 }, { "epoch": 2.26, "learning_rate": 1.804837942584707e-05, "loss": 0.7621, "step": 7358 }, { "epoch": 2.26, "learning_rate": 1.804778794467972e-05, "loss": 0.5494, "step": 7359 }, { "epoch": 2.26, "learning_rate": 1.804719638359042e-05, "loss": 0.7032, "step": 7360 }, { "epoch": 2.26, "learning_rate": 1.804660474258505e-05, "loss": 0.7392, "step": 7361 }, { "epoch": 2.27, "learning_rate": 1.804601302166947e-05, "loss": 0.9776, "step": 7362 }, { "epoch": 2.27, "learning_rate": 1.804542122084957e-05, "loss": 1.0311, "step": 7363 }, { "epoch": 2.27, "learning_rate": 1.8044829340131223e-05, "loss": 1.204, "step": 7364 }, { "epoch": 2.27, "learning_rate": 1.8044237379520305e-05, "loss": 0.8883, "step": 7365 }, { "epoch": 2.27, "learning_rate": 1.80436453390227e-05, "loss": 0.7961, "step": 7366 }, { "epoch": 2.27, "learning_rate": 1.8043053218644274e-05, "loss": 0.7557, "step": 7367 }, { "epoch": 2.27, "learning_rate": 1.8042461018390924e-05, "loss": 0.9109, "step": 7368 }, { "epoch": 2.27, "learning_rate": 1.8041868738268523e-05, "loss": 0.5272, "step": 7369 }, { "epoch": 2.27, "learning_rate": 1.8041276378282953e-05, "loss": 0.7747, "step": 7370 }, { "epoch": 2.27, "learning_rate": 1.80406839384401e-05, "loss": 0.6857, "step": 7371 }, { "epoch": 2.27, "learning_rate": 1.804009141874584e-05, "loss": 1.0294, "step": 7372 }, { "epoch": 2.27, "learning_rate": 1.803949881920607e-05, "loss": 0.8569, "step": 7373 }, { "epoch": 2.27, "learning_rate": 1.8038906139826658e-05, "loss": 0.8847, "step": 7374 }, { "epoch": 2.27, "learning_rate": 1.8038313380613505e-05, "loss": 0.6345, "step": 7375 }, { "epoch": 2.27, "learning_rate": 1.803772054157249e-05, "loss": 0.6596, "step": 7376 }, { "epoch": 2.27, "learning_rate": 1.80371276227095e-05, "loss": 0.653, "step": 7377 }, { "epoch": 2.27, "learning_rate": 1.8036534624030428e-05, "loss": 0.6944, "step": 7378 }, { "epoch": 2.27, "learning_rate": 1.803594154554116e-05, "loss": 1.0785, "step": 7379 }, { "epoch": 2.27, "learning_rate": 1.803534838724759e-05, "loss": 1.0569, "step": 7380 }, { "epoch": 2.27, "learning_rate": 1.8034755149155603e-05, "loss": 0.8256, "step": 7381 }, { "epoch": 2.27, "learning_rate": 1.8034161831271088e-05, "loss": 0.7558, "step": 7382 }, { "epoch": 2.27, "learning_rate": 1.8033568433599946e-05, "loss": 0.8077, "step": 7383 }, { "epoch": 2.27, "learning_rate": 1.8032974956148064e-05, "loss": 0.8112, "step": 7384 }, { "epoch": 2.27, "learning_rate": 1.8032381398921336e-05, "loss": 0.7343, "step": 7385 }, { "epoch": 2.27, "learning_rate": 1.803178776192566e-05, "loss": 0.6826, "step": 7386 }, { "epoch": 2.27, "learning_rate": 1.803119404516693e-05, "loss": 0.7242, "step": 7387 }, { "epoch": 2.27, "learning_rate": 1.8030600248651038e-05, "loss": 0.6, "step": 7388 }, { "epoch": 2.27, "learning_rate": 1.8030006372383888e-05, "loss": 0.7052, "step": 7389 }, { "epoch": 2.27, "learning_rate": 1.802941241637137e-05, "loss": 0.9514, "step": 7390 }, { "epoch": 2.27, "learning_rate": 1.802881838061939e-05, "loss": 0.8636, "step": 7391 }, { "epoch": 2.27, "learning_rate": 1.8028224265133842e-05, "loss": 0.7603, "step": 7392 }, { "epoch": 2.27, "learning_rate": 1.8027630069920627e-05, "loss": 0.9379, "step": 7393 }, { "epoch": 2.27, "learning_rate": 1.802703579498565e-05, "loss": 0.8426, "step": 7394 }, { "epoch": 2.28, "learning_rate": 1.8026441440334807e-05, "loss": 0.7037, "step": 7395 }, { "epoch": 2.28, "learning_rate": 1.8025847005974006e-05, "loss": 0.9547, "step": 7396 }, { "epoch": 2.28, "learning_rate": 1.8025252491909143e-05, "loss": 0.597, "step": 7397 }, { "epoch": 2.28, "learning_rate": 1.8024657898146133e-05, "loss": 0.9504, "step": 7398 }, { "epoch": 2.28, "learning_rate": 1.802406322469087e-05, "loss": 0.6865, "step": 7399 }, { "epoch": 2.28, "learning_rate": 1.8023468471549263e-05, "loss": 0.8067, "step": 7400 }, { "epoch": 2.28, "learning_rate": 1.802287363872722e-05, "loss": 0.7478, "step": 7401 }, { "epoch": 2.28, "learning_rate": 1.8022278726230648e-05, "loss": 0.6825, "step": 7402 }, { "epoch": 2.28, "learning_rate": 1.8021683734065458e-05, "loss": 0.8602, "step": 7403 }, { "epoch": 2.28, "learning_rate": 1.802108866223755e-05, "loss": 0.6878, "step": 7404 }, { "epoch": 2.28, "learning_rate": 1.8020493510752843e-05, "loss": 0.8865, "step": 7405 }, { "epoch": 2.28, "learning_rate": 1.8019898279617246e-05, "loss": 1.0075, "step": 7406 }, { "epoch": 2.28, "learning_rate": 1.8019302968836662e-05, "loss": 0.9945, "step": 7407 }, { "epoch": 2.28, "learning_rate": 1.801870757841701e-05, "loss": 0.805, "step": 7408 }, { "epoch": 2.28, "learning_rate": 1.8018112108364204e-05, "loss": 1.091, "step": 7409 }, { "epoch": 2.28, "learning_rate": 1.801751655868415e-05, "loss": 0.7398, "step": 7410 }, { "epoch": 2.28, "learning_rate": 1.8016920929382773e-05, "loss": 0.799, "step": 7411 }, { "epoch": 2.28, "learning_rate": 1.8016325220465984e-05, "loss": 0.9863, "step": 7412 }, { "epoch": 2.28, "learning_rate": 1.801572943193969e-05, "loss": 0.6292, "step": 7413 }, { "epoch": 2.28, "learning_rate": 1.8015133563809822e-05, "loss": 0.6008, "step": 7414 }, { "epoch": 2.28, "learning_rate": 1.8014537616082285e-05, "loss": 0.7768, "step": 7415 }, { "epoch": 2.28, "learning_rate": 1.8013941588763006e-05, "loss": 0.8644, "step": 7416 }, { "epoch": 2.28, "learning_rate": 1.8013345481857903e-05, "loss": 0.9042, "step": 7417 }, { "epoch": 2.28, "learning_rate": 1.801274929537289e-05, "loss": 0.8084, "step": 7418 }, { "epoch": 2.28, "learning_rate": 1.8012153029313894e-05, "loss": 0.7983, "step": 7419 }, { "epoch": 2.28, "learning_rate": 1.8011556683686836e-05, "loss": 0.8232, "step": 7420 }, { "epoch": 2.28, "learning_rate": 1.801096025849763e-05, "loss": 0.5308, "step": 7421 }, { "epoch": 2.28, "learning_rate": 1.8010363753752212e-05, "loss": 0.9259, "step": 7422 }, { "epoch": 2.28, "learning_rate": 1.80097671694565e-05, "loss": 0.5974, "step": 7423 }, { "epoch": 2.28, "learning_rate": 1.8009170505616414e-05, "loss": 0.8334, "step": 7424 }, { "epoch": 2.28, "learning_rate": 1.8008573762237883e-05, "loss": 1.1099, "step": 7425 }, { "epoch": 2.28, "learning_rate": 1.800797693932684e-05, "loss": 0.9253, "step": 7426 }, { "epoch": 2.29, "learning_rate": 1.8007380036889198e-05, "loss": 0.8412, "step": 7427 }, { "epoch": 2.29, "learning_rate": 1.8006783054930895e-05, "loss": 0.6782, "step": 7428 }, { "epoch": 2.29, "learning_rate": 1.800618599345786e-05, "loss": 0.7014, "step": 7429 }, { "epoch": 2.29, "learning_rate": 1.8005588852476018e-05, "loss": 0.7308, "step": 7430 }, { "epoch": 2.29, "learning_rate": 1.80049916319913e-05, "loss": 0.791, "step": 7431 }, { "epoch": 2.29, "learning_rate": 1.8004394332009634e-05, "loss": 0.7431, "step": 7432 }, { "epoch": 2.29, "learning_rate": 1.8003796952536956e-05, "loss": 0.8584, "step": 7433 }, { "epoch": 2.29, "learning_rate": 1.80031994935792e-05, "loss": 0.7695, "step": 7434 }, { "epoch": 2.29, "learning_rate": 1.80026019551423e-05, "loss": 0.8373, "step": 7435 }, { "epoch": 2.29, "learning_rate": 1.800200433723218e-05, "loss": 0.6359, "step": 7436 }, { "epoch": 2.29, "learning_rate": 1.800140663985479e-05, "loss": 0.8204, "step": 7437 }, { "epoch": 2.29, "learning_rate": 1.8000808863016053e-05, "loss": 0.802, "step": 7438 }, { "epoch": 2.29, "learning_rate": 1.8000211006721912e-05, "loss": 1.3114, "step": 7439 }, { "epoch": 2.29, "learning_rate": 1.79996130709783e-05, "loss": 0.6685, "step": 7440 }, { "epoch": 2.29, "learning_rate": 1.799901505579116e-05, "loss": 0.6867, "step": 7441 }, { "epoch": 2.29, "learning_rate": 1.7998416961166426e-05, "loss": 0.7713, "step": 7442 }, { "epoch": 2.29, "learning_rate": 1.7997818787110043e-05, "loss": 0.9505, "step": 7443 }, { "epoch": 2.29, "learning_rate": 1.7997220533627947e-05, "loss": 0.4522, "step": 7444 }, { "epoch": 2.29, "learning_rate": 1.799662220072608e-05, "loss": 0.6455, "step": 7445 }, { "epoch": 2.29, "learning_rate": 1.7996023788410385e-05, "loss": 0.8962, "step": 7446 }, { "epoch": 2.29, "learning_rate": 1.7995425296686804e-05, "loss": 0.8597, "step": 7447 }, { "epoch": 2.29, "learning_rate": 1.7994826725561284e-05, "loss": 0.6419, "step": 7448 }, { "epoch": 2.29, "learning_rate": 1.7994228075039765e-05, "loss": 0.6514, "step": 7449 }, { "epoch": 2.29, "learning_rate": 1.799362934512819e-05, "loss": 0.8091, "step": 7450 }, { "epoch": 2.29, "learning_rate": 1.799303053583251e-05, "loss": 0.6588, "step": 7451 }, { "epoch": 2.29, "learning_rate": 1.7992431647158673e-05, "loss": 0.7851, "step": 7452 }, { "epoch": 2.29, "learning_rate": 1.799183267911262e-05, "loss": 1.0261, "step": 7453 }, { "epoch": 2.29, "learning_rate": 1.7991233631700308e-05, "loss": 0.8879, "step": 7454 }, { "epoch": 2.29, "learning_rate": 1.7990634504927676e-05, "loss": 0.8682, "step": 7455 }, { "epoch": 2.29, "learning_rate": 1.7990035298800682e-05, "loss": 0.8238, "step": 7456 }, { "epoch": 2.29, "learning_rate": 1.798943601332527e-05, "loss": 0.7573, "step": 7457 }, { "epoch": 2.29, "learning_rate": 1.79888366485074e-05, "loss": 0.7782, "step": 7458 }, { "epoch": 2.29, "learning_rate": 1.7988237204353017e-05, "loss": 0.8577, "step": 7459 }, { "epoch": 2.3, "learning_rate": 1.7987637680868076e-05, "loss": 0.7185, "step": 7460 }, { "epoch": 2.3, "learning_rate": 1.798703807805853e-05, "loss": 0.8772, "step": 7461 }, { "epoch": 2.3, "learning_rate": 1.798643839593034e-05, "loss": 0.5682, "step": 7462 }, { "epoch": 2.3, "learning_rate": 1.798583863448945e-05, "loss": 0.7142, "step": 7463 }, { "epoch": 2.3, "learning_rate": 1.7985238793741824e-05, "loss": 0.664, "step": 7464 }, { "epoch": 2.3, "learning_rate": 1.7984638873693422e-05, "loss": 0.8935, "step": 7465 }, { "epoch": 2.3, "learning_rate": 1.7984038874350193e-05, "loss": 0.9261, "step": 7466 }, { "epoch": 2.3, "learning_rate": 1.79834387957181e-05, "loss": 0.7474, "step": 7467 }, { "epoch": 2.3, "learning_rate": 1.79828386378031e-05, "loss": 0.6834, "step": 7468 }, { "epoch": 2.3, "learning_rate": 1.798223840061116e-05, "loss": 0.5769, "step": 7469 }, { "epoch": 2.3, "learning_rate": 1.7981638084148235e-05, "loss": 0.589, "step": 7470 }, { "epoch": 2.3, "learning_rate": 1.7981037688420282e-05, "loss": 0.7134, "step": 7471 }, { "epoch": 2.3, "learning_rate": 1.7980437213433274e-05, "loss": 0.878, "step": 7472 }, { "epoch": 2.3, "learning_rate": 1.797983665919317e-05, "loss": 0.9508, "step": 7473 }, { "epoch": 2.3, "learning_rate": 1.7979236025705933e-05, "loss": 0.5555, "step": 7474 }, { "epoch": 2.3, "learning_rate": 1.7978635312977525e-05, "loss": 0.5679, "step": 7475 }, { "epoch": 2.3, "learning_rate": 1.797803452101392e-05, "loss": 0.7787, "step": 7476 }, { "epoch": 2.3, "learning_rate": 1.797743364982108e-05, "loss": 0.7716, "step": 7477 }, { "epoch": 2.3, "learning_rate": 1.7976832699404965e-05, "loss": 0.7197, "step": 7478 }, { "epoch": 2.3, "learning_rate": 1.7976231669771554e-05, "loss": 1.0479, "step": 7479 }, { "epoch": 2.3, "learning_rate": 1.797563056092681e-05, "loss": 0.8152, "step": 7480 }, { "epoch": 2.3, "learning_rate": 1.7975029372876706e-05, "loss": 0.7698, "step": 7481 }, { "epoch": 2.3, "learning_rate": 1.797442810562721e-05, "loss": 0.6259, "step": 7482 }, { "epoch": 2.3, "learning_rate": 1.7973826759184292e-05, "loss": 0.6247, "step": 7483 }, { "epoch": 2.3, "learning_rate": 1.797322533355393e-05, "loss": 0.9931, "step": 7484 }, { "epoch": 2.3, "learning_rate": 1.7972623828742085e-05, "loss": 1.0564, "step": 7485 }, { "epoch": 2.3, "learning_rate": 1.7972022244754738e-05, "loss": 0.8604, "step": 7486 }, { "epoch": 2.3, "learning_rate": 1.797142058159787e-05, "loss": 0.6254, "step": 7487 }, { "epoch": 2.3, "learning_rate": 1.7970818839277442e-05, "loss": 0.7842, "step": 7488 }, { "epoch": 2.3, "learning_rate": 1.797021701779944e-05, "loss": 0.7356, "step": 7489 }, { "epoch": 2.3, "learning_rate": 1.7969615117169836e-05, "loss": 0.8963, "step": 7490 }, { "epoch": 2.3, "learning_rate": 1.796901313739461e-05, "loss": 0.9774, "step": 7491 }, { "epoch": 2.31, "learning_rate": 1.796841107847974e-05, "loss": 1.1379, "step": 7492 }, { "epoch": 2.31, "learning_rate": 1.79678089404312e-05, "loss": 0.8256, "step": 7493 }, { "epoch": 2.31, "learning_rate": 1.796720672325498e-05, "loss": 0.5713, "step": 7494 }, { "epoch": 2.31, "learning_rate": 1.796660442695705e-05, "loss": 0.5911, "step": 7495 }, { "epoch": 2.31, "learning_rate": 1.7966002051543397e-05, "loss": 0.7988, "step": 7496 }, { "epoch": 2.31, "learning_rate": 1.7965399597020002e-05, "loss": 0.7472, "step": 7497 }, { "epoch": 2.31, "learning_rate": 1.7964797063392847e-05, "loss": 1.0597, "step": 7498 }, { "epoch": 2.31, "learning_rate": 1.796419445066792e-05, "loss": 0.5909, "step": 7499 }, { "epoch": 2.31, "learning_rate": 1.7963591758851197e-05, "loss": 0.8026, "step": 7500 }, { "epoch": 2.31, "learning_rate": 1.796298898794867e-05, "loss": 0.6596, "step": 7501 }, { "epoch": 2.31, "learning_rate": 1.7962386137966327e-05, "loss": 0.9338, "step": 7502 }, { "epoch": 2.31, "learning_rate": 1.7961783208910148e-05, "loss": 0.9689, "step": 7503 }, { "epoch": 2.31, "learning_rate": 1.7961180200786123e-05, "loss": 0.5962, "step": 7504 }, { "epoch": 2.31, "learning_rate": 1.7960577113600243e-05, "loss": 0.7148, "step": 7505 }, { "epoch": 2.31, "learning_rate": 1.7959973947358494e-05, "loss": 0.6611, "step": 7506 }, { "epoch": 2.31, "learning_rate": 1.795937070206687e-05, "loss": 0.7751, "step": 7507 }, { "epoch": 2.31, "learning_rate": 1.795876737773136e-05, "loss": 0.8726, "step": 7508 }, { "epoch": 2.31, "learning_rate": 1.7958163974357953e-05, "loss": 0.7548, "step": 7509 }, { "epoch": 2.31, "learning_rate": 1.7957560491952642e-05, "loss": 0.9902, "step": 7510 }, { "epoch": 2.31, "learning_rate": 1.7956956930521425e-05, "loss": 0.8485, "step": 7511 }, { "epoch": 2.31, "learning_rate": 1.795635329007029e-05, "loss": 0.7898, "step": 7512 }, { "epoch": 2.31, "learning_rate": 1.7955749570605235e-05, "loss": 0.7788, "step": 7513 }, { "epoch": 2.31, "learning_rate": 1.7955145772132253e-05, "loss": 0.9117, "step": 7514 }, { "epoch": 2.31, "learning_rate": 1.7954541894657342e-05, "loss": 0.8897, "step": 7515 }, { "epoch": 2.31, "learning_rate": 1.7953937938186502e-05, "loss": 0.7186, "step": 7516 }, { "epoch": 2.31, "learning_rate": 1.7953333902725726e-05, "loss": 0.9688, "step": 7517 }, { "epoch": 2.31, "learning_rate": 1.7952729788281012e-05, "loss": 0.8185, "step": 7518 }, { "epoch": 2.31, "learning_rate": 1.7952125594858363e-05, "loss": 0.9711, "step": 7519 }, { "epoch": 2.31, "learning_rate": 1.795152132246378e-05, "loss": 1.0678, "step": 7520 }, { "epoch": 2.31, "learning_rate": 1.795091697110326e-05, "loss": 0.7476, "step": 7521 }, { "epoch": 2.31, "learning_rate": 1.7950312540782808e-05, "loss": 0.7656, "step": 7522 }, { "epoch": 2.31, "learning_rate": 1.794970803150842e-05, "loss": 0.9867, "step": 7523 }, { "epoch": 2.31, "learning_rate": 1.794910344328611e-05, "loss": 0.5503, "step": 7524 }, { "epoch": 2.32, "learning_rate": 1.7948498776121877e-05, "loss": 0.8202, "step": 7525 }, { "epoch": 2.32, "learning_rate": 1.7947894030021722e-05, "loss": 0.6767, "step": 7526 }, { "epoch": 2.32, "learning_rate": 1.7947289204991657e-05, "loss": 1.2237, "step": 7527 }, { "epoch": 2.32, "learning_rate": 1.7946684301037683e-05, "loss": 0.6791, "step": 7528 }, { "epoch": 2.32, "learning_rate": 1.7946079318165816e-05, "loss": 0.7453, "step": 7529 }, { "epoch": 2.32, "learning_rate": 1.7945474256382055e-05, "loss": 1.0236, "step": 7530 }, { "epoch": 2.32, "learning_rate": 1.794486911569241e-05, "loss": 0.7896, "step": 7531 }, { "epoch": 2.32, "learning_rate": 1.7944263896102893e-05, "loss": 0.6092, "step": 7532 }, { "epoch": 2.32, "learning_rate": 1.7943658597619513e-05, "loss": 0.8537, "step": 7533 }, { "epoch": 2.32, "learning_rate": 1.7943053220248284e-05, "loss": 0.7427, "step": 7534 }, { "epoch": 2.32, "learning_rate": 1.7942447763995216e-05, "loss": 0.7758, "step": 7535 }, { "epoch": 2.32, "learning_rate": 1.794184222886632e-05, "loss": 0.687, "step": 7536 }, { "epoch": 2.32, "learning_rate": 1.7941236614867615e-05, "loss": 0.6193, "step": 7537 }, { "epoch": 2.32, "learning_rate": 1.7940630922005105e-05, "loss": 0.8431, "step": 7538 }, { "epoch": 2.32, "learning_rate": 1.7940025150284817e-05, "loss": 0.7236, "step": 7539 }, { "epoch": 2.32, "learning_rate": 1.793941929971276e-05, "loss": 1.1151, "step": 7540 }, { "epoch": 2.32, "learning_rate": 1.7938813370294955e-05, "loss": 0.8152, "step": 7541 }, { "epoch": 2.32, "learning_rate": 1.7938207362037412e-05, "loss": 0.8626, "step": 7542 }, { "epoch": 2.32, "learning_rate": 1.7937601274946158e-05, "loss": 0.8488, "step": 7543 }, { "epoch": 2.32, "learning_rate": 1.7936995109027205e-05, "loss": 0.5973, "step": 7544 }, { "epoch": 2.32, "learning_rate": 1.7936388864286575e-05, "loss": 0.9079, "step": 7545 }, { "epoch": 2.32, "learning_rate": 1.7935782540730288e-05, "loss": 0.7328, "step": 7546 }, { "epoch": 2.32, "learning_rate": 1.793517613836437e-05, "loss": 0.6522, "step": 7547 }, { "epoch": 2.32, "learning_rate": 1.793456965719484e-05, "loss": 0.7227, "step": 7548 }, { "epoch": 2.32, "learning_rate": 1.7933963097227717e-05, "loss": 0.7589, "step": 7549 }, { "epoch": 2.32, "learning_rate": 1.793335645846903e-05, "loss": 0.6234, "step": 7550 }, { "epoch": 2.32, "learning_rate": 1.7932749740924804e-05, "loss": 0.8964, "step": 7551 }, { "epoch": 2.32, "learning_rate": 1.7932142944601062e-05, "loss": 0.6225, "step": 7552 }, { "epoch": 2.32, "learning_rate": 1.793153606950383e-05, "loss": 0.9533, "step": 7553 }, { "epoch": 2.32, "learning_rate": 1.793092911563913e-05, "loss": 0.8114, "step": 7554 }, { "epoch": 2.32, "learning_rate": 1.7930322083012997e-05, "loss": 0.5534, "step": 7555 }, { "epoch": 2.32, "learning_rate": 1.7929714971631457e-05, "loss": 0.6717, "step": 7556 }, { "epoch": 2.33, "learning_rate": 1.792910778150054e-05, "loss": 0.5476, "step": 7557 }, { "epoch": 2.33, "learning_rate": 1.7928500512626277e-05, "loss": 0.7799, "step": 7558 }, { "epoch": 2.33, "learning_rate": 1.7927893165014692e-05, "loss": 0.7257, "step": 7559 }, { "epoch": 2.33, "learning_rate": 1.7927285738671825e-05, "loss": 0.7363, "step": 7560 }, { "epoch": 2.33, "learning_rate": 1.7926678233603703e-05, "loss": 0.8076, "step": 7561 }, { "epoch": 2.33, "learning_rate": 1.792607064981636e-05, "loss": 0.8653, "step": 7562 }, { "epoch": 2.33, "learning_rate": 1.7925462987315833e-05, "loss": 0.8623, "step": 7563 }, { "epoch": 2.33, "learning_rate": 1.7924855246108152e-05, "loss": 0.7772, "step": 7564 }, { "epoch": 2.33, "learning_rate": 1.7924247426199358e-05, "loss": 0.8669, "step": 7565 }, { "epoch": 2.33, "learning_rate": 1.7923639527595483e-05, "loss": 1.0121, "step": 7566 }, { "epoch": 2.33, "learning_rate": 1.792303155030256e-05, "loss": 0.8581, "step": 7567 }, { "epoch": 2.33, "learning_rate": 1.792242349432664e-05, "loss": 0.7876, "step": 7568 }, { "epoch": 2.33, "learning_rate": 1.7921815359673743e-05, "loss": 1.1887, "step": 7569 }, { "epoch": 2.33, "learning_rate": 1.7921207146349925e-05, "loss": 0.6545, "step": 7570 }, { "epoch": 2.33, "learning_rate": 1.792059885436122e-05, "loss": 0.6186, "step": 7571 }, { "epoch": 2.33, "learning_rate": 1.7919990483713666e-05, "loss": 0.7879, "step": 7572 }, { "epoch": 2.33, "learning_rate": 1.7919382034413306e-05, "loss": 0.833, "step": 7573 }, { "epoch": 2.33, "learning_rate": 1.7918773506466188e-05, "loss": 0.7686, "step": 7574 }, { "epoch": 2.33, "learning_rate": 1.791816489987835e-05, "loss": 0.7955, "step": 7575 }, { "epoch": 2.33, "learning_rate": 1.7917556214655834e-05, "loss": 0.5643, "step": 7576 }, { "epoch": 2.33, "learning_rate": 1.7916947450804687e-05, "loss": 0.762, "step": 7577 }, { "epoch": 2.33, "learning_rate": 1.791633860833096e-05, "loss": 0.7666, "step": 7578 }, { "epoch": 2.33, "learning_rate": 1.791572968724069e-05, "loss": 0.8925, "step": 7579 }, { "epoch": 2.33, "learning_rate": 1.7915120687539932e-05, "loss": 0.7608, "step": 7580 }, { "epoch": 2.33, "learning_rate": 1.7914511609234732e-05, "loss": 0.5699, "step": 7581 }, { "epoch": 2.33, "learning_rate": 1.7913902452331132e-05, "loss": 0.9335, "step": 7582 }, { "epoch": 2.33, "learning_rate": 1.7913293216835194e-05, "loss": 0.6754, "step": 7583 }, { "epoch": 2.33, "learning_rate": 1.7912683902752955e-05, "loss": 0.5795, "step": 7584 }, { "epoch": 2.33, "learning_rate": 1.7912074510090474e-05, "loss": 0.7489, "step": 7585 }, { "epoch": 2.33, "learning_rate": 1.7911465038853805e-05, "loss": 0.8569, "step": 7586 }, { "epoch": 2.33, "learning_rate": 1.7910855489048992e-05, "loss": 0.775, "step": 7587 }, { "epoch": 2.33, "learning_rate": 1.7910245860682097e-05, "loss": 0.5751, "step": 7588 }, { "epoch": 2.33, "learning_rate": 1.7909636153759167e-05, "loss": 0.8006, "step": 7589 }, { "epoch": 2.34, "learning_rate": 1.7909026368286265e-05, "loss": 0.7463, "step": 7590 }, { "epoch": 2.34, "learning_rate": 1.7908416504269437e-05, "loss": 0.755, "step": 7591 }, { "epoch": 2.34, "learning_rate": 1.7907806561714745e-05, "loss": 0.6119, "step": 7592 }, { "epoch": 2.34, "learning_rate": 1.7907196540628248e-05, "loss": 0.8986, "step": 7593 }, { "epoch": 2.34, "learning_rate": 1.7906586441016002e-05, "loss": 0.6036, "step": 7594 }, { "epoch": 2.34, "learning_rate": 1.790597626288406e-05, "loss": 0.8363, "step": 7595 }, { "epoch": 2.34, "learning_rate": 1.7905366006238493e-05, "loss": 1.0338, "step": 7596 }, { "epoch": 2.34, "learning_rate": 1.7904755671085358e-05, "loss": 0.9023, "step": 7597 }, { "epoch": 2.34, "learning_rate": 1.790414525743071e-05, "loss": 0.656, "step": 7598 }, { "epoch": 2.34, "learning_rate": 1.7903534765280616e-05, "loss": 0.7763, "step": 7599 }, { "epoch": 2.34, "learning_rate": 1.7902924194641137e-05, "loss": 0.7145, "step": 7600 }, { "epoch": 2.34, "learning_rate": 1.7902313545518338e-05, "loss": 0.6379, "step": 7601 }, { "epoch": 2.34, "learning_rate": 1.790170281791828e-05, "loss": 0.7649, "step": 7602 }, { "epoch": 2.34, "learning_rate": 1.790109201184703e-05, "loss": 0.6935, "step": 7603 }, { "epoch": 2.34, "learning_rate": 1.7900481127310662e-05, "loss": 0.8012, "step": 7604 }, { "epoch": 2.34, "learning_rate": 1.7899870164315226e-05, "loss": 1.2655, "step": 7605 }, { "epoch": 2.34, "learning_rate": 1.7899259122866802e-05, "loss": 0.8786, "step": 7606 }, { "epoch": 2.34, "learning_rate": 1.7898648002971456e-05, "loss": 0.6356, "step": 7607 }, { "epoch": 2.34, "learning_rate": 1.7898036804635257e-05, "loss": 0.9017, "step": 7608 }, { "epoch": 2.34, "learning_rate": 1.789742552786427e-05, "loss": 0.8179, "step": 7609 }, { "epoch": 2.34, "learning_rate": 1.7896814172664572e-05, "loss": 0.8017, "step": 7610 }, { "epoch": 2.34, "learning_rate": 1.789620273904223e-05, "loss": 0.7551, "step": 7611 }, { "epoch": 2.34, "learning_rate": 1.7895591227003316e-05, "loss": 1.1504, "step": 7612 }, { "epoch": 2.34, "learning_rate": 1.7894979636553906e-05, "loss": 0.5965, "step": 7613 }, { "epoch": 2.34, "learning_rate": 1.7894367967700074e-05, "loss": 0.7813, "step": 7614 }, { "epoch": 2.34, "learning_rate": 1.7893756220447888e-05, "loss": 0.8771, "step": 7615 }, { "epoch": 2.34, "learning_rate": 1.789314439480343e-05, "loss": 0.9132, "step": 7616 }, { "epoch": 2.34, "learning_rate": 1.7892532490772773e-05, "loss": 0.981, "step": 7617 }, { "epoch": 2.34, "learning_rate": 1.7891920508361994e-05, "loss": 0.6184, "step": 7618 }, { "epoch": 2.34, "learning_rate": 1.7891308447577174e-05, "loss": 0.781, "step": 7619 }, { "epoch": 2.34, "learning_rate": 1.7890696308424385e-05, "loss": 0.7116, "step": 7620 }, { "epoch": 2.34, "learning_rate": 1.7890084090909708e-05, "loss": 0.8528, "step": 7621 }, { "epoch": 2.35, "learning_rate": 1.788947179503923e-05, "loss": 0.6863, "step": 7622 }, { "epoch": 2.35, "learning_rate": 1.788885942081902e-05, "loss": 0.6899, "step": 7623 }, { "epoch": 2.35, "learning_rate": 1.7888246968255168e-05, "loss": 0.9806, "step": 7624 }, { "epoch": 2.35, "learning_rate": 1.7887634437353754e-05, "loss": 0.4553, "step": 7625 }, { "epoch": 2.35, "learning_rate": 1.7887021828120857e-05, "loss": 0.8241, "step": 7626 }, { "epoch": 2.35, "learning_rate": 1.788640914056257e-05, "loss": 0.7184, "step": 7627 }, { "epoch": 2.35, "learning_rate": 1.788579637468497e-05, "loss": 0.9594, "step": 7628 }, { "epoch": 2.35, "learning_rate": 1.7885183530494143e-05, "loss": 0.4524, "step": 7629 }, { "epoch": 2.35, "learning_rate": 1.7884570607996176e-05, "loss": 0.9809, "step": 7630 }, { "epoch": 2.35, "learning_rate": 1.7883957607197157e-05, "loss": 0.7975, "step": 7631 }, { "epoch": 2.35, "learning_rate": 1.788334452810317e-05, "loss": 0.6542, "step": 7632 }, { "epoch": 2.35, "learning_rate": 1.7882731370720313e-05, "loss": 0.6344, "step": 7633 }, { "epoch": 2.35, "learning_rate": 1.7882118135054665e-05, "loss": 0.7435, "step": 7634 }, { "epoch": 2.35, "learning_rate": 1.788150482111232e-05, "loss": 0.7624, "step": 7635 }, { "epoch": 2.35, "learning_rate": 1.7880891428899367e-05, "loss": 0.8455, "step": 7636 }, { "epoch": 2.35, "learning_rate": 1.78802779584219e-05, "loss": 0.7626, "step": 7637 }, { "epoch": 2.35, "learning_rate": 1.7879664409686007e-05, "loss": 0.7542, "step": 7638 }, { "epoch": 2.35, "learning_rate": 1.787905078269779e-05, "loss": 0.6508, "step": 7639 }, { "epoch": 2.35, "learning_rate": 1.7878437077463334e-05, "loss": 0.9807, "step": 7640 }, { "epoch": 2.35, "learning_rate": 1.7877823293988737e-05, "loss": 0.7286, "step": 7641 }, { "epoch": 2.35, "learning_rate": 1.7877209432280094e-05, "loss": 0.8558, "step": 7642 }, { "epoch": 2.35, "learning_rate": 1.7876595492343502e-05, "loss": 0.7013, "step": 7643 }, { "epoch": 2.35, "learning_rate": 1.7875981474185057e-05, "loss": 0.8257, "step": 7644 }, { "epoch": 2.35, "learning_rate": 1.787536737781086e-05, "loss": 1.0233, "step": 7645 }, { "epoch": 2.35, "learning_rate": 1.7874753203227e-05, "loss": 1.0101, "step": 7646 }, { "epoch": 2.35, "learning_rate": 1.787413895043959e-05, "loss": 1.0596, "step": 7647 }, { "epoch": 2.35, "learning_rate": 1.787352461945472e-05, "loss": 0.5429, "step": 7648 }, { "epoch": 2.35, "learning_rate": 1.7872910210278494e-05, "loss": 0.7958, "step": 7649 }, { "epoch": 2.35, "learning_rate": 1.7872295722917014e-05, "loss": 0.7783, "step": 7650 }, { "epoch": 2.35, "learning_rate": 1.7871681157376382e-05, "loss": 0.6119, "step": 7651 }, { "epoch": 2.35, "learning_rate": 1.7871066513662703e-05, "loss": 0.7957, "step": 7652 }, { "epoch": 2.35, "learning_rate": 1.787045179178208e-05, "loss": 0.8469, "step": 7653 }, { "epoch": 2.35, "learning_rate": 1.7869836991740613e-05, "loss": 1.0759, "step": 7654 }, { "epoch": 2.36, "learning_rate": 1.7869222113544416e-05, "loss": 0.8889, "step": 7655 }, { "epoch": 2.36, "learning_rate": 1.786860715719959e-05, "loss": 0.6934, "step": 7656 }, { "epoch": 2.36, "learning_rate": 1.7867992122712242e-05, "loss": 0.926, "step": 7657 }, { "epoch": 2.36, "learning_rate": 1.786737701008848e-05, "loss": 0.7686, "step": 7658 }, { "epoch": 2.36, "learning_rate": 1.7866761819334414e-05, "loss": 0.8578, "step": 7659 }, { "epoch": 2.36, "learning_rate": 1.7866146550456152e-05, "loss": 0.8393, "step": 7660 }, { "epoch": 2.36, "learning_rate": 1.7865531203459806e-05, "loss": 0.8708, "step": 7661 }, { "epoch": 2.36, "learning_rate": 1.7864915778351487e-05, "loss": 0.7523, "step": 7662 }, { "epoch": 2.36, "learning_rate": 1.7864300275137305e-05, "loss": 0.8405, "step": 7663 }, { "epoch": 2.36, "learning_rate": 1.7863684693823375e-05, "loss": 0.957, "step": 7664 }, { "epoch": 2.36, "learning_rate": 1.7863069034415806e-05, "loss": 0.845, "step": 7665 }, { "epoch": 2.36, "learning_rate": 1.7862453296920717e-05, "loss": 0.8244, "step": 7666 }, { "epoch": 2.36, "learning_rate": 1.786183748134422e-05, "loss": 0.8633, "step": 7667 }, { "epoch": 2.36, "learning_rate": 1.7861221587692426e-05, "loss": 0.6934, "step": 7668 }, { "epoch": 2.36, "learning_rate": 1.7860605615971463e-05, "loss": 0.7607, "step": 7669 }, { "epoch": 2.36, "learning_rate": 1.7859989566187442e-05, "loss": 0.8261, "step": 7670 }, { "epoch": 2.36, "learning_rate": 1.7859373438346475e-05, "loss": 0.8275, "step": 7671 }, { "epoch": 2.36, "learning_rate": 1.7858757232454692e-05, "loss": 0.8912, "step": 7672 }, { "epoch": 2.36, "learning_rate": 1.7858140948518205e-05, "loss": 0.8707, "step": 7673 }, { "epoch": 2.36, "learning_rate": 1.7857524586543135e-05, "loss": 0.8219, "step": 7674 }, { "epoch": 2.36, "learning_rate": 1.7856908146535602e-05, "loss": 0.851, "step": 7675 }, { "epoch": 2.36, "learning_rate": 1.7856291628501733e-05, "loss": 1.2929, "step": 7676 }, { "epoch": 2.36, "learning_rate": 1.7855675032447648e-05, "loss": 0.7896, "step": 7677 }, { "epoch": 2.36, "learning_rate": 1.785505835837947e-05, "loss": 0.9465, "step": 7678 }, { "epoch": 2.36, "learning_rate": 1.7854441606303323e-05, "loss": 0.8918, "step": 7679 }, { "epoch": 2.36, "learning_rate": 1.785382477622533e-05, "loss": 0.73, "step": 7680 }, { "epoch": 2.36, "learning_rate": 1.785320786815162e-05, "loss": 0.7215, "step": 7681 }, { "epoch": 2.36, "learning_rate": 1.785259088208832e-05, "loss": 0.9709, "step": 7682 }, { "epoch": 2.36, "learning_rate": 1.785197381804155e-05, "loss": 0.9199, "step": 7683 }, { "epoch": 2.36, "learning_rate": 1.785135667601745e-05, "loss": 0.6955, "step": 7684 }, { "epoch": 2.36, "learning_rate": 1.7850739456022137e-05, "loss": 0.8525, "step": 7685 }, { "epoch": 2.36, "learning_rate": 1.7850122158061746e-05, "loss": 0.7578, "step": 7686 }, { "epoch": 2.37, "learning_rate": 1.784950478214241e-05, "loss": 0.8818, "step": 7687 }, { "epoch": 2.37, "learning_rate": 1.7848887328270254e-05, "loss": 0.91, "step": 7688 }, { "epoch": 2.37, "learning_rate": 1.7848269796451418e-05, "loss": 0.7791, "step": 7689 }, { "epoch": 2.37, "learning_rate": 1.7847652186692025e-05, "loss": 1.0323, "step": 7690 }, { "epoch": 2.37, "learning_rate": 1.7847034498998216e-05, "loss": 0.8515, "step": 7691 }, { "epoch": 2.37, "learning_rate": 1.784641673337612e-05, "loss": 1.0529, "step": 7692 }, { "epoch": 2.37, "learning_rate": 1.7845798889831874e-05, "loss": 0.8497, "step": 7693 }, { "epoch": 2.37, "learning_rate": 1.7845180968371618e-05, "loss": 1.0092, "step": 7694 }, { "epoch": 2.37, "learning_rate": 1.784456296900148e-05, "loss": 0.827, "step": 7695 }, { "epoch": 2.37, "learning_rate": 1.784394489172761e-05, "loss": 0.6702, "step": 7696 }, { "epoch": 2.37, "learning_rate": 1.784332673655613e-05, "loss": 0.7727, "step": 7697 }, { "epoch": 2.37, "learning_rate": 1.784270850349319e-05, "loss": 0.6312, "step": 7698 }, { "epoch": 2.37, "learning_rate": 1.7842090192544927e-05, "loss": 0.7684, "step": 7699 }, { "epoch": 2.37, "learning_rate": 1.7841471803717477e-05, "loss": 0.7779, "step": 7700 }, { "epoch": 2.37, "learning_rate": 1.784085333701699e-05, "loss": 0.9267, "step": 7701 }, { "epoch": 2.37, "learning_rate": 1.78402347924496e-05, "loss": 0.8049, "step": 7702 }, { "epoch": 2.37, "learning_rate": 1.7839616170021452e-05, "loss": 0.699, "step": 7703 }, { "epoch": 2.37, "learning_rate": 1.7838997469738694e-05, "loss": 0.6669, "step": 7704 }, { "epoch": 2.37, "learning_rate": 1.783837869160746e-05, "loss": 0.8777, "step": 7705 }, { "epoch": 2.37, "learning_rate": 1.7837759835633908e-05, "loss": 0.9048, "step": 7706 }, { "epoch": 2.37, "learning_rate": 1.7837140901824177e-05, "loss": 0.7578, "step": 7707 }, { "epoch": 2.37, "learning_rate": 1.783652189018441e-05, "loss": 0.6954, "step": 7708 }, { "epoch": 2.37, "learning_rate": 1.7835902800720762e-05, "loss": 0.9096, "step": 7709 }, { "epoch": 2.37, "learning_rate": 1.783528363343938e-05, "loss": 0.7602, "step": 7710 }, { "epoch": 2.37, "learning_rate": 1.78346643883464e-05, "loss": 1.1734, "step": 7711 }, { "epoch": 2.37, "learning_rate": 1.783404506544799e-05, "loss": 0.6937, "step": 7712 }, { "epoch": 2.37, "learning_rate": 1.783342566475029e-05, "loss": 0.7471, "step": 7713 }, { "epoch": 2.37, "learning_rate": 1.7832806186259455e-05, "loss": 0.6756, "step": 7714 }, { "epoch": 2.37, "learning_rate": 1.7832186629981632e-05, "loss": 0.8853, "step": 7715 }, { "epoch": 2.37, "learning_rate": 1.7831566995922983e-05, "loss": 0.8223, "step": 7716 }, { "epoch": 2.37, "learning_rate": 1.7830947284089654e-05, "loss": 0.9409, "step": 7717 }, { "epoch": 2.37, "learning_rate": 1.78303274944878e-05, "loss": 0.8378, "step": 7718 }, { "epoch": 2.37, "learning_rate": 1.7829707627123576e-05, "loss": 0.7113, "step": 7719 }, { "epoch": 2.38, "learning_rate": 1.7829087682003143e-05, "loss": 0.8156, "step": 7720 }, { "epoch": 2.38, "learning_rate": 1.782846765913265e-05, "loss": 0.7256, "step": 7721 }, { "epoch": 2.38, "learning_rate": 1.782784755851826e-05, "loss": 0.6996, "step": 7722 }, { "epoch": 2.38, "learning_rate": 1.782722738016613e-05, "loss": 1.0819, "step": 7723 }, { "epoch": 2.38, "learning_rate": 1.7826607124082418e-05, "loss": 0.76, "step": 7724 }, { "epoch": 2.38, "learning_rate": 1.7825986790273283e-05, "loss": 0.9554, "step": 7725 }, { "epoch": 2.38, "learning_rate": 1.7825366378744888e-05, "loss": 0.8024, "step": 7726 }, { "epoch": 2.38, "learning_rate": 1.782474588950339e-05, "loss": 0.6481, "step": 7727 }, { "epoch": 2.38, "learning_rate": 1.7824125322554957e-05, "loss": 0.6962, "step": 7728 }, { "epoch": 2.38, "learning_rate": 1.782350467790575e-05, "loss": 0.8242, "step": 7729 }, { "epoch": 2.38, "learning_rate": 1.7822883955561925e-05, "loss": 0.8237, "step": 7730 }, { "epoch": 2.38, "learning_rate": 1.7822263155529658e-05, "loss": 0.7765, "step": 7731 }, { "epoch": 2.38, "learning_rate": 1.7821642277815108e-05, "loss": 1.1485, "step": 7732 }, { "epoch": 2.38, "learning_rate": 1.782102132242444e-05, "loss": 0.8579, "step": 7733 }, { "epoch": 2.38, "learning_rate": 1.7820400289363822e-05, "loss": 0.5565, "step": 7734 }, { "epoch": 2.38, "learning_rate": 1.7819779178639422e-05, "loss": 0.832, "step": 7735 }, { "epoch": 2.38, "learning_rate": 1.781915799025741e-05, "loss": 0.9872, "step": 7736 }, { "epoch": 2.38, "learning_rate": 1.781853672422395e-05, "loss": 0.8434, "step": 7737 }, { "epoch": 2.38, "learning_rate": 1.7817915380545217e-05, "loss": 0.6393, "step": 7738 }, { "epoch": 2.38, "learning_rate": 1.7817293959227378e-05, "loss": 0.716, "step": 7739 }, { "epoch": 2.38, "learning_rate": 1.7816672460276604e-05, "loss": 0.7543, "step": 7740 }, { "epoch": 2.38, "learning_rate": 1.781605088369907e-05, "loss": 0.6397, "step": 7741 }, { "epoch": 2.38, "learning_rate": 1.7815429229500946e-05, "loss": 0.9131, "step": 7742 }, { "epoch": 2.38, "learning_rate": 1.7814807497688408e-05, "loss": 0.8952, "step": 7743 }, { "epoch": 2.38, "learning_rate": 1.7814185688267628e-05, "loss": 0.909, "step": 7744 }, { "epoch": 2.38, "learning_rate": 1.7813563801244783e-05, "loss": 0.7926, "step": 7745 }, { "epoch": 2.38, "learning_rate": 1.781294183662605e-05, "loss": 0.8755, "step": 7746 }, { "epoch": 2.38, "learning_rate": 1.78123197944176e-05, "loss": 0.7988, "step": 7747 }, { "epoch": 2.38, "learning_rate": 1.7811697674625618e-05, "loss": 0.687, "step": 7748 }, { "epoch": 2.38, "learning_rate": 1.7811075477256277e-05, "loss": 0.8428, "step": 7749 }, { "epoch": 2.38, "learning_rate": 1.7810453202315757e-05, "loss": 0.8784, "step": 7750 }, { "epoch": 2.38, "learning_rate": 1.780983084981024e-05, "loss": 0.7452, "step": 7751 }, { "epoch": 2.39, "learning_rate": 1.7809208419745904e-05, "loss": 0.8429, "step": 7752 }, { "epoch": 2.39, "learning_rate": 1.780858591212893e-05, "loss": 0.7859, "step": 7753 }, { "epoch": 2.39, "learning_rate": 1.78079633269655e-05, "loss": 0.7542, "step": 7754 }, { "epoch": 2.39, "learning_rate": 1.78073406642618e-05, "loss": 0.5444, "step": 7755 }, { "epoch": 2.39, "learning_rate": 1.7806717924024014e-05, "loss": 0.9047, "step": 7756 }, { "epoch": 2.39, "learning_rate": 1.7806095106258323e-05, "loss": 0.7996, "step": 7757 }, { "epoch": 2.39, "learning_rate": 1.7805472210970914e-05, "loss": 1.109, "step": 7758 }, { "epoch": 2.39, "learning_rate": 1.780484923816797e-05, "loss": 0.7229, "step": 7759 }, { "epoch": 2.39, "learning_rate": 1.780422618785568e-05, "loss": 0.6895, "step": 7760 }, { "epoch": 2.39, "learning_rate": 1.7803603060040235e-05, "loss": 0.6497, "step": 7761 }, { "epoch": 2.39, "learning_rate": 1.7802979854727816e-05, "loss": 0.7649, "step": 7762 }, { "epoch": 2.39, "learning_rate": 1.780235657192462e-05, "loss": 0.6675, "step": 7763 }, { "epoch": 2.39, "learning_rate": 1.7801733211636826e-05, "loss": 1.3706, "step": 7764 }, { "epoch": 2.39, "learning_rate": 1.7801109773870635e-05, "loss": 0.7728, "step": 7765 }, { "epoch": 2.39, "learning_rate": 1.780048625863223e-05, "loss": 0.6036, "step": 7766 }, { "epoch": 2.39, "learning_rate": 1.779986266592781e-05, "loss": 0.7416, "step": 7767 }, { "epoch": 2.39, "learning_rate": 1.779923899576357e-05, "loss": 0.5234, "step": 7768 }, { "epoch": 2.39, "learning_rate": 1.7798615248145692e-05, "loss": 0.9835, "step": 7769 }, { "epoch": 2.39, "learning_rate": 1.779799142308038e-05, "loss": 1.0326, "step": 7770 }, { "epoch": 2.39, "learning_rate": 1.7797367520573825e-05, "loss": 0.7512, "step": 7771 }, { "epoch": 2.39, "learning_rate": 1.7796743540632226e-05, "loss": 0.8828, "step": 7772 }, { "epoch": 2.39, "learning_rate": 1.779611948326178e-05, "loss": 0.871, "step": 7773 }, { "epoch": 2.39, "learning_rate": 1.7795495348468676e-05, "loss": 0.9128, "step": 7774 }, { "epoch": 2.39, "learning_rate": 1.7794871136259123e-05, "loss": 0.925, "step": 7775 }, { "epoch": 2.39, "learning_rate": 1.7794246846639313e-05, "loss": 0.8591, "step": 7776 }, { "epoch": 2.39, "learning_rate": 1.779362247961545e-05, "loss": 0.8303, "step": 7777 }, { "epoch": 2.39, "learning_rate": 1.7792998035193733e-05, "loss": 0.9527, "step": 7778 }, { "epoch": 2.39, "learning_rate": 1.7792373513380363e-05, "loss": 0.8622, "step": 7779 }, { "epoch": 2.39, "learning_rate": 1.7791748914181544e-05, "loss": 0.5627, "step": 7780 }, { "epoch": 2.39, "learning_rate": 1.7791124237603477e-05, "loss": 0.8663, "step": 7781 }, { "epoch": 2.39, "learning_rate": 1.7790499483652363e-05, "loss": 0.9438, "step": 7782 }, { "epoch": 2.39, "learning_rate": 1.778987465233441e-05, "loss": 0.6801, "step": 7783 }, { "epoch": 2.39, "learning_rate": 1.7789249743655825e-05, "loss": 0.9578, "step": 7784 }, { "epoch": 2.4, "learning_rate": 1.778862475762281e-05, "loss": 0.8134, "step": 7785 }, { "epoch": 2.4, "learning_rate": 1.7787999694241574e-05, "loss": 0.7235, "step": 7786 }, { "epoch": 2.4, "learning_rate": 1.778737455351832e-05, "loss": 0.9217, "step": 7787 }, { "epoch": 2.4, "learning_rate": 1.778674933545926e-05, "loss": 0.6032, "step": 7788 }, { "epoch": 2.4, "learning_rate": 1.7786124040070606e-05, "loss": 0.7008, "step": 7789 }, { "epoch": 2.4, "learning_rate": 1.7785498667358566e-05, "loss": 0.9578, "step": 7790 }, { "epoch": 2.4, "learning_rate": 1.7784873217329345e-05, "loss": 0.6736, "step": 7791 }, { "epoch": 2.4, "learning_rate": 1.7784247689989156e-05, "loss": 0.7545, "step": 7792 }, { "epoch": 2.4, "learning_rate": 1.7783622085344218e-05, "loss": 0.7125, "step": 7793 }, { "epoch": 2.4, "learning_rate": 1.7782996403400737e-05, "loss": 0.6793, "step": 7794 }, { "epoch": 2.4, "learning_rate": 1.7782370644164927e-05, "loss": 0.8759, "step": 7795 }, { "epoch": 2.4, "learning_rate": 1.7781744807643005e-05, "loss": 0.5987, "step": 7796 }, { "epoch": 2.4, "learning_rate": 1.7781118893841185e-05, "loss": 0.8182, "step": 7797 }, { "epoch": 2.4, "learning_rate": 1.7780492902765684e-05, "loss": 0.7791, "step": 7798 }, { "epoch": 2.4, "learning_rate": 1.777986683442272e-05, "loss": 0.937, "step": 7799 }, { "epoch": 2.4, "learning_rate": 1.7779240688818506e-05, "loss": 1.5179, "step": 7800 }, { "epoch": 2.4, "learning_rate": 1.7778614465959263e-05, "loss": 0.6881, "step": 7801 }, { "epoch": 2.4, "learning_rate": 1.7777988165851208e-05, "loss": 0.5054, "step": 7802 }, { "epoch": 2.4, "learning_rate": 1.7777361788500562e-05, "loss": 0.7448, "step": 7803 }, { "epoch": 2.4, "learning_rate": 1.777673533391355e-05, "loss": 0.9083, "step": 7804 }, { "epoch": 2.4, "learning_rate": 1.7776108802096385e-05, "loss": 0.8083, "step": 7805 }, { "epoch": 2.4, "learning_rate": 1.777548219305529e-05, "loss": 0.9261, "step": 7806 }, { "epoch": 2.4, "learning_rate": 1.7774855506796497e-05, "loss": 0.7857, "step": 7807 }, { "epoch": 2.4, "learning_rate": 1.7774228743326218e-05, "loss": 0.7232, "step": 7808 }, { "epoch": 2.4, "learning_rate": 1.7773601902650687e-05, "loss": 1.0855, "step": 7809 }, { "epoch": 2.4, "learning_rate": 1.777297498477612e-05, "loss": 0.7954, "step": 7810 }, { "epoch": 2.4, "learning_rate": 1.777234798970875e-05, "loss": 0.7997, "step": 7811 }, { "epoch": 2.4, "learning_rate": 1.77717209174548e-05, "loss": 0.8185, "step": 7812 }, { "epoch": 2.4, "learning_rate": 1.7771093768020505e-05, "loss": 1.0121, "step": 7813 }, { "epoch": 2.4, "learning_rate": 1.7770466541412078e-05, "loss": 0.9293, "step": 7814 }, { "epoch": 2.4, "learning_rate": 1.7769839237635765e-05, "loss": 0.5756, "step": 7815 }, { "epoch": 2.4, "learning_rate": 1.7769211856697782e-05, "loss": 0.8085, "step": 7816 }, { "epoch": 2.41, "learning_rate": 1.7768584398604365e-05, "loss": 0.7668, "step": 7817 }, { "epoch": 2.41, "learning_rate": 1.7767956863361746e-05, "loss": 0.7873, "step": 7818 }, { "epoch": 2.41, "learning_rate": 1.7767329250976158e-05, "loss": 0.8142, "step": 7819 }, { "epoch": 2.41, "learning_rate": 1.776670156145383e-05, "loss": 0.7802, "step": 7820 }, { "epoch": 2.41, "learning_rate": 1.7766073794800998e-05, "loss": 0.8665, "step": 7821 }, { "epoch": 2.41, "learning_rate": 1.7765445951023898e-05, "loss": 0.8336, "step": 7822 }, { "epoch": 2.41, "learning_rate": 1.7764818030128758e-05, "loss": 0.5853, "step": 7823 }, { "epoch": 2.41, "learning_rate": 1.7764190032121825e-05, "loss": 0.7025, "step": 7824 }, { "epoch": 2.41, "learning_rate": 1.7763561957009324e-05, "loss": 0.9328, "step": 7825 }, { "epoch": 2.41, "learning_rate": 1.77629338047975e-05, "loss": 0.9153, "step": 7826 }, { "epoch": 2.41, "learning_rate": 1.7762305575492587e-05, "loss": 0.8793, "step": 7827 }, { "epoch": 2.41, "learning_rate": 1.776167726910083e-05, "loss": 0.8239, "step": 7828 }, { "epoch": 2.41, "learning_rate": 1.7761048885628457e-05, "loss": 1.2359, "step": 7829 }, { "epoch": 2.41, "learning_rate": 1.776042042508172e-05, "loss": 0.7697, "step": 7830 }, { "epoch": 2.41, "learning_rate": 1.775979188746686e-05, "loss": 0.7423, "step": 7831 }, { "epoch": 2.41, "learning_rate": 1.7759163272790108e-05, "loss": 0.9609, "step": 7832 }, { "epoch": 2.41, "learning_rate": 1.775853458105772e-05, "loss": 0.9316, "step": 7833 }, { "epoch": 2.41, "learning_rate": 1.7757905812275928e-05, "loss": 0.9517, "step": 7834 }, { "epoch": 2.41, "learning_rate": 1.775727696645098e-05, "loss": 0.7603, "step": 7835 }, { "epoch": 2.41, "learning_rate": 1.7756648043589125e-05, "loss": 1.0087, "step": 7836 }, { "epoch": 2.41, "learning_rate": 1.775601904369661e-05, "loss": 0.7399, "step": 7837 }, { "epoch": 2.41, "learning_rate": 1.775538996677967e-05, "loss": 0.8406, "step": 7838 }, { "epoch": 2.41, "learning_rate": 1.775476081284457e-05, "loss": 0.7947, "step": 7839 }, { "epoch": 2.41, "learning_rate": 1.7754131581897538e-05, "loss": 0.8397, "step": 7840 }, { "epoch": 2.41, "learning_rate": 1.775350227394484e-05, "loss": 1.0149, "step": 7841 }, { "epoch": 2.41, "learning_rate": 1.7752872888992714e-05, "loss": 0.6683, "step": 7842 }, { "epoch": 2.41, "learning_rate": 1.775224342704742e-05, "loss": 0.7913, "step": 7843 }, { "epoch": 2.41, "learning_rate": 1.7751613888115197e-05, "loss": 0.8207, "step": 7844 }, { "epoch": 2.41, "learning_rate": 1.775098427220231e-05, "loss": 0.6219, "step": 7845 }, { "epoch": 2.41, "learning_rate": 1.7750354579315004e-05, "loss": 0.886, "step": 7846 }, { "epoch": 2.41, "learning_rate": 1.7749724809459534e-05, "loss": 0.7897, "step": 7847 }, { "epoch": 2.41, "learning_rate": 1.7749094962642156e-05, "loss": 0.7308, "step": 7848 }, { "epoch": 2.41, "learning_rate": 1.774846503886912e-05, "loss": 0.9276, "step": 7849 }, { "epoch": 2.42, "learning_rate": 1.7747835038146687e-05, "loss": 0.9062, "step": 7850 }, { "epoch": 2.42, "learning_rate": 1.774720496048111e-05, "loss": 0.7604, "step": 7851 }, { "epoch": 2.42, "learning_rate": 1.7746574805878648e-05, "loss": 0.8588, "step": 7852 }, { "epoch": 2.42, "learning_rate": 1.7745944574345563e-05, "loss": 0.7797, "step": 7853 }, { "epoch": 2.42, "learning_rate": 1.7745314265888104e-05, "loss": 0.8927, "step": 7854 }, { "epoch": 2.42, "learning_rate": 1.7744683880512538e-05, "loss": 0.9194, "step": 7855 }, { "epoch": 2.42, "learning_rate": 1.7744053418225123e-05, "loss": 0.5864, "step": 7856 }, { "epoch": 2.42, "learning_rate": 1.774342287903212e-05, "loss": 0.7032, "step": 7857 }, { "epoch": 2.42, "learning_rate": 1.7742792262939793e-05, "loss": 0.932, "step": 7858 }, { "epoch": 2.42, "learning_rate": 1.77421615699544e-05, "loss": 0.775, "step": 7859 }, { "epoch": 2.42, "learning_rate": 1.774153080008221e-05, "loss": 0.8456, "step": 7860 }, { "epoch": 2.42, "learning_rate": 1.7740899953329482e-05, "loss": 0.5879, "step": 7861 }, { "epoch": 2.42, "learning_rate": 1.7740269029702486e-05, "loss": 0.7782, "step": 7862 }, { "epoch": 2.42, "learning_rate": 1.7739638029207484e-05, "loss": 0.5879, "step": 7863 }, { "epoch": 2.42, "learning_rate": 1.7739006951850744e-05, "loss": 0.705, "step": 7864 }, { "epoch": 2.42, "learning_rate": 1.773837579763853e-05, "loss": 0.8891, "step": 7865 }, { "epoch": 2.42, "learning_rate": 1.7737744566577115e-05, "loss": 0.6133, "step": 7866 }, { "epoch": 2.42, "learning_rate": 1.773711325867276e-05, "loss": 0.9807, "step": 7867 }, { "epoch": 2.42, "learning_rate": 1.7736481873931746e-05, "loss": 0.7292, "step": 7868 }, { "epoch": 2.42, "learning_rate": 1.7735850412360332e-05, "loss": 0.7475, "step": 7869 }, { "epoch": 2.42, "learning_rate": 1.7735218873964795e-05, "loss": 0.8804, "step": 7870 }, { "epoch": 2.42, "learning_rate": 1.7734587258751407e-05, "loss": 0.5503, "step": 7871 }, { "epoch": 2.42, "learning_rate": 1.7733955566726438e-05, "loss": 0.7242, "step": 7872 }, { "epoch": 2.42, "learning_rate": 1.7733323797896168e-05, "loss": 1.1179, "step": 7873 }, { "epoch": 2.42, "learning_rate": 1.773269195226686e-05, "loss": 0.7732, "step": 7874 }, { "epoch": 2.42, "learning_rate": 1.7732060029844792e-05, "loss": 0.6879, "step": 7875 }, { "epoch": 2.42, "learning_rate": 1.7731428030636248e-05, "loss": 0.8134, "step": 7876 }, { "epoch": 2.42, "learning_rate": 1.7730795954647493e-05, "loss": 0.8169, "step": 7877 }, { "epoch": 2.42, "learning_rate": 1.773016380188481e-05, "loss": 0.753, "step": 7878 }, { "epoch": 2.42, "learning_rate": 1.772953157235448e-05, "loss": 0.8867, "step": 7879 }, { "epoch": 2.42, "learning_rate": 1.7728899266062772e-05, "loss": 0.7706, "step": 7880 }, { "epoch": 2.42, "learning_rate": 1.7728266883015975e-05, "loss": 0.6562, "step": 7881 }, { "epoch": 2.43, "learning_rate": 1.7727634423220367e-05, "loss": 0.8178, "step": 7882 }, { "epoch": 2.43, "learning_rate": 1.772700188668222e-05, "loss": 0.824, "step": 7883 }, { "epoch": 2.43, "learning_rate": 1.772636927340783e-05, "loss": 0.978, "step": 7884 }, { "epoch": 2.43, "learning_rate": 1.772573658340347e-05, "loss": 0.5482, "step": 7885 }, { "epoch": 2.43, "learning_rate": 1.7725103816675428e-05, "loss": 0.9127, "step": 7886 }, { "epoch": 2.43, "learning_rate": 1.7724470973229978e-05, "loss": 0.8534, "step": 7887 }, { "epoch": 2.43, "learning_rate": 1.772383805307342e-05, "loss": 0.8969, "step": 7888 }, { "epoch": 2.43, "learning_rate": 1.7723205056212027e-05, "loss": 0.7605, "step": 7889 }, { "epoch": 2.43, "learning_rate": 1.772257198265209e-05, "loss": 0.7184, "step": 7890 }, { "epoch": 2.43, "learning_rate": 1.77219388323999e-05, "loss": 0.6385, "step": 7891 }, { "epoch": 2.43, "learning_rate": 1.7721305605461735e-05, "loss": 0.9162, "step": 7892 }, { "epoch": 2.43, "learning_rate": 1.7720672301843892e-05, "loss": 1.0636, "step": 7893 }, { "epoch": 2.43, "learning_rate": 1.772003892155266e-05, "loss": 0.9336, "step": 7894 }, { "epoch": 2.43, "learning_rate": 1.7719405464594325e-05, "loss": 0.6892, "step": 7895 }, { "epoch": 2.43, "learning_rate": 1.7718771930975177e-05, "loss": 0.6758, "step": 7896 }, { "epoch": 2.43, "learning_rate": 1.771813832070151e-05, "loss": 1.1167, "step": 7897 }, { "epoch": 2.43, "learning_rate": 1.7717504633779618e-05, "loss": 0.7677, "step": 7898 }, { "epoch": 2.43, "learning_rate": 1.7716870870215793e-05, "loss": 0.8929, "step": 7899 }, { "epoch": 2.43, "learning_rate": 1.7716237030016326e-05, "loss": 0.7818, "step": 7900 }, { "epoch": 2.43, "learning_rate": 1.7715603113187514e-05, "loss": 0.6883, "step": 7901 }, { "epoch": 2.43, "learning_rate": 1.7714969119735653e-05, "loss": 0.6679, "step": 7902 }, { "epoch": 2.43, "learning_rate": 1.7714335049667037e-05, "loss": 0.8364, "step": 7903 }, { "epoch": 2.43, "learning_rate": 1.7713700902987967e-05, "loss": 0.9257, "step": 7904 }, { "epoch": 2.43, "learning_rate": 1.7713066679704735e-05, "loss": 0.9166, "step": 7905 }, { "epoch": 2.43, "learning_rate": 1.7712432379823645e-05, "loss": 0.8542, "step": 7906 }, { "epoch": 2.43, "learning_rate": 1.7711798003350994e-05, "loss": 1.0501, "step": 7907 }, { "epoch": 2.43, "learning_rate": 1.7711163550293077e-05, "loss": 0.7264, "step": 7908 }, { "epoch": 2.43, "learning_rate": 1.77105290206562e-05, "loss": 1.1405, "step": 7909 }, { "epoch": 2.43, "learning_rate": 1.770989441444667e-05, "loss": 0.6805, "step": 7910 }, { "epoch": 2.43, "learning_rate": 1.7709259731670774e-05, "loss": 0.6341, "step": 7911 }, { "epoch": 2.43, "learning_rate": 1.770862497233483e-05, "loss": 0.6993, "step": 7912 }, { "epoch": 2.43, "learning_rate": 1.7707990136445136e-05, "loss": 0.8921, "step": 7913 }, { "epoch": 2.43, "learning_rate": 1.770735522400799e-05, "loss": 0.8605, "step": 7914 }, { "epoch": 2.44, "learning_rate": 1.7706720235029707e-05, "loss": 0.5966, "step": 7915 }, { "epoch": 2.44, "learning_rate": 1.770608516951659e-05, "loss": 0.8266, "step": 7916 }, { "epoch": 2.44, "learning_rate": 1.7705450027474945e-05, "loss": 0.7522, "step": 7917 }, { "epoch": 2.44, "learning_rate": 1.7704814808911076e-05, "loss": 0.7306, "step": 7918 }, { "epoch": 2.44, "learning_rate": 1.77041795138313e-05, "loss": 0.7078, "step": 7919 }, { "epoch": 2.44, "learning_rate": 1.770354414224192e-05, "loss": 0.7712, "step": 7920 }, { "epoch": 2.44, "learning_rate": 1.7702908694149245e-05, "loss": 0.7936, "step": 7921 }, { "epoch": 2.44, "learning_rate": 1.770227316955959e-05, "loss": 0.8736, "step": 7922 }, { "epoch": 2.44, "learning_rate": 1.770163756847926e-05, "loss": 0.9283, "step": 7923 }, { "epoch": 2.44, "learning_rate": 1.770100189091457e-05, "loss": 0.7369, "step": 7924 }, { "epoch": 2.44, "learning_rate": 1.770036613687184e-05, "loss": 0.7525, "step": 7925 }, { "epoch": 2.44, "learning_rate": 1.7699730306357375e-05, "loss": 0.7525, "step": 7926 }, { "epoch": 2.44, "learning_rate": 1.769909439937749e-05, "loss": 0.6047, "step": 7927 }, { "epoch": 2.44, "learning_rate": 1.7698458415938502e-05, "loss": 0.7841, "step": 7928 }, { "epoch": 2.44, "learning_rate": 1.769782235604673e-05, "loss": 0.4553, "step": 7929 }, { "epoch": 2.44, "learning_rate": 1.7697186219708484e-05, "loss": 0.6791, "step": 7930 }, { "epoch": 2.44, "learning_rate": 1.7696550006930087e-05, "loss": 0.8445, "step": 7931 }, { "epoch": 2.44, "learning_rate": 1.7695913717717854e-05, "loss": 0.7685, "step": 7932 }, { "epoch": 2.44, "learning_rate": 1.7695277352078105e-05, "loss": 0.778, "step": 7933 }, { "epoch": 2.44, "learning_rate": 1.769464091001716e-05, "loss": 0.9093, "step": 7934 }, { "epoch": 2.44, "learning_rate": 1.7694004391541337e-05, "loss": 0.7307, "step": 7935 }, { "epoch": 2.44, "learning_rate": 1.7693367796656963e-05, "loss": 0.8801, "step": 7936 }, { "epoch": 2.44, "learning_rate": 1.7692731125370355e-05, "loss": 0.8584, "step": 7937 }, { "epoch": 2.44, "learning_rate": 1.7692094377687837e-05, "loss": 0.7106, "step": 7938 }, { "epoch": 2.44, "learning_rate": 1.769145755361573e-05, "loss": 0.8105, "step": 7939 }, { "epoch": 2.44, "learning_rate": 1.7690820653160365e-05, "loss": 1.01, "step": 7940 }, { "epoch": 2.44, "learning_rate": 1.7690183676328062e-05, "loss": 0.8957, "step": 7941 }, { "epoch": 2.44, "learning_rate": 1.7689546623125147e-05, "loss": 0.8329, "step": 7942 }, { "epoch": 2.44, "learning_rate": 1.7688909493557944e-05, "loss": 0.8536, "step": 7943 }, { "epoch": 2.44, "learning_rate": 1.768827228763279e-05, "loss": 0.6889, "step": 7944 }, { "epoch": 2.44, "learning_rate": 1.7687635005356e-05, "loss": 0.6447, "step": 7945 }, { "epoch": 2.44, "learning_rate": 1.768699764673391e-05, "loss": 0.9173, "step": 7946 }, { "epoch": 2.45, "learning_rate": 1.7686360211772854e-05, "loss": 0.9209, "step": 7947 }, { "epoch": 2.45, "learning_rate": 1.7685722700479154e-05, "loss": 0.7249, "step": 7948 }, { "epoch": 2.45, "learning_rate": 1.7685085112859144e-05, "loss": 0.6938, "step": 7949 }, { "epoch": 2.45, "learning_rate": 1.7684447448919156e-05, "loss": 0.6653, "step": 7950 }, { "epoch": 2.45, "learning_rate": 1.7683809708665524e-05, "loss": 0.9699, "step": 7951 }, { "epoch": 2.45, "learning_rate": 1.7683171892104578e-05, "loss": 1.0996, "step": 7952 }, { "epoch": 2.45, "learning_rate": 1.7682533999242656e-05, "loss": 0.7986, "step": 7953 }, { "epoch": 2.45, "learning_rate": 1.768189603008609e-05, "loss": 0.5639, "step": 7954 }, { "epoch": 2.45, "learning_rate": 1.7681257984641217e-05, "loss": 0.9569, "step": 7955 }, { "epoch": 2.45, "learning_rate": 1.768061986291437e-05, "loss": 0.8188, "step": 7956 }, { "epoch": 2.45, "learning_rate": 1.7679981664911893e-05, "loss": 0.7489, "step": 7957 }, { "epoch": 2.45, "learning_rate": 1.767934339064012e-05, "loss": 0.8664, "step": 7958 }, { "epoch": 2.45, "learning_rate": 1.7678705040105385e-05, "loss": 1.006, "step": 7959 }, { "epoch": 2.45, "learning_rate": 1.767806661331404e-05, "loss": 0.5803, "step": 7960 }, { "epoch": 2.45, "learning_rate": 1.767742811027241e-05, "loss": 0.8319, "step": 7961 }, { "epoch": 2.45, "learning_rate": 1.7676789530986845e-05, "loss": 0.8315, "step": 7962 }, { "epoch": 2.45, "learning_rate": 1.7676150875463688e-05, "loss": 0.7745, "step": 7963 }, { "epoch": 2.45, "learning_rate": 1.7675512143709273e-05, "loss": 0.6499, "step": 7964 }, { "epoch": 2.45, "learning_rate": 1.767487333572995e-05, "loss": 0.9694, "step": 7965 }, { "epoch": 2.45, "learning_rate": 1.7674234451532065e-05, "loss": 0.8274, "step": 7966 }, { "epoch": 2.45, "learning_rate": 1.7673595491121957e-05, "loss": 0.6766, "step": 7967 }, { "epoch": 2.45, "learning_rate": 1.767295645450597e-05, "loss": 0.6486, "step": 7968 }, { "epoch": 2.45, "learning_rate": 1.767231734169046e-05, "loss": 0.9238, "step": 7969 }, { "epoch": 2.45, "learning_rate": 1.7671678152681763e-05, "loss": 0.9594, "step": 7970 }, { "epoch": 2.45, "learning_rate": 1.7671038887486233e-05, "loss": 0.8441, "step": 7971 }, { "epoch": 2.45, "learning_rate": 1.767039954611022e-05, "loss": 0.8899, "step": 7972 }, { "epoch": 2.45, "learning_rate": 1.7669760128560067e-05, "loss": 1.0769, "step": 7973 }, { "epoch": 2.45, "learning_rate": 1.766912063484213e-05, "loss": 0.8144, "step": 7974 }, { "epoch": 2.45, "learning_rate": 1.766848106496275e-05, "loss": 0.5444, "step": 7975 }, { "epoch": 2.45, "learning_rate": 1.7667841418928292e-05, "loss": 1.0192, "step": 7976 }, { "epoch": 2.45, "learning_rate": 1.76672016967451e-05, "loss": 0.6358, "step": 7977 }, { "epoch": 2.45, "learning_rate": 1.766656189841953e-05, "loss": 0.7567, "step": 7978 }, { "epoch": 2.45, "learning_rate": 1.7665922023957934e-05, "loss": 1.0261, "step": 7979 }, { "epoch": 2.46, "learning_rate": 1.7665282073366666e-05, "loss": 0.5656, "step": 7980 }, { "epoch": 2.46, "learning_rate": 1.7664642046652086e-05, "loss": 0.9683, "step": 7981 }, { "epoch": 2.46, "learning_rate": 1.766400194382054e-05, "loss": 0.7777, "step": 7982 }, { "epoch": 2.46, "learning_rate": 1.7663361764878396e-05, "loss": 0.719, "step": 7983 }, { "epoch": 2.46, "learning_rate": 1.7662721509832004e-05, "loss": 0.9005, "step": 7984 }, { "epoch": 2.46, "learning_rate": 1.7662081178687732e-05, "loss": 0.7774, "step": 7985 }, { "epoch": 2.46, "learning_rate": 1.766144077145193e-05, "loss": 0.6891, "step": 7986 }, { "epoch": 2.46, "learning_rate": 1.7660800288130954e-05, "loss": 0.8739, "step": 7987 }, { "epoch": 2.46, "learning_rate": 1.7660159728731178e-05, "loss": 0.9003, "step": 7988 }, { "epoch": 2.46, "learning_rate": 1.765951909325895e-05, "loss": 0.6852, "step": 7989 }, { "epoch": 2.46, "learning_rate": 1.765887838172064e-05, "loss": 0.5706, "step": 7990 }, { "epoch": 2.46, "learning_rate": 1.765823759412261e-05, "loss": 0.897, "step": 7991 }, { "epoch": 2.46, "learning_rate": 1.7657596730471223e-05, "loss": 0.866, "step": 7992 }, { "epoch": 2.46, "learning_rate": 1.7656955790772843e-05, "loss": 0.6348, "step": 7993 }, { "epoch": 2.46, "learning_rate": 1.7656314775033836e-05, "loss": 0.77, "step": 7994 }, { "epoch": 2.46, "learning_rate": 1.7655673683260564e-05, "loss": 0.6472, "step": 7995 }, { "epoch": 2.46, "learning_rate": 1.76550325154594e-05, "loss": 0.7837, "step": 7996 }, { "epoch": 2.46, "learning_rate": 1.7654391271636707e-05, "loss": 0.76, "step": 7997 }, { "epoch": 2.46, "learning_rate": 1.7653749951798852e-05, "loss": 0.8827, "step": 7998 }, { "epoch": 2.46, "learning_rate": 1.7653108555952207e-05, "loss": 0.6711, "step": 7999 }, { "epoch": 2.46, "learning_rate": 1.7652467084103142e-05, "loss": 0.9537, "step": 8000 }, { "epoch": 2.46, "learning_rate": 1.7651825536258024e-05, "loss": 1.0471, "step": 8001 }, { "epoch": 2.46, "learning_rate": 1.7651183912423228e-05, "loss": 0.6862, "step": 8002 }, { "epoch": 2.46, "learning_rate": 1.7650542212605125e-05, "loss": 0.8115, "step": 8003 }, { "epoch": 2.46, "learning_rate": 1.764990043681009e-05, "loss": 0.7331, "step": 8004 }, { "epoch": 2.46, "learning_rate": 1.7649258585044487e-05, "loss": 0.6803, "step": 8005 }, { "epoch": 2.46, "learning_rate": 1.76486166573147e-05, "loss": 0.8482, "step": 8006 }, { "epoch": 2.46, "learning_rate": 1.7647974653627095e-05, "loss": 0.8321, "step": 8007 }, { "epoch": 2.46, "learning_rate": 1.7647332573988062e-05, "loss": 0.7996, "step": 8008 }, { "epoch": 2.46, "learning_rate": 1.7646690418403963e-05, "loss": 0.9067, "step": 8009 }, { "epoch": 2.46, "learning_rate": 1.764604818688118e-05, "loss": 0.9017, "step": 8010 }, { "epoch": 2.46, "learning_rate": 1.7645405879426092e-05, "loss": 0.649, "step": 8011 }, { "epoch": 2.47, "learning_rate": 1.764476349604508e-05, "loss": 1.1969, "step": 8012 }, { "epoch": 2.47, "learning_rate": 1.764412103674452e-05, "loss": 0.8614, "step": 8013 }, { "epoch": 2.47, "learning_rate": 1.764347850153079e-05, "loss": 0.8881, "step": 8014 }, { "epoch": 2.47, "learning_rate": 1.764283589041028e-05, "loss": 0.811, "step": 8015 }, { "epoch": 2.47, "learning_rate": 1.764219320338936e-05, "loss": 0.7283, "step": 8016 }, { "epoch": 2.47, "learning_rate": 1.764155044047442e-05, "loss": 0.5439, "step": 8017 }, { "epoch": 2.47, "learning_rate": 1.7640907601671844e-05, "loss": 0.646, "step": 8018 }, { "epoch": 2.47, "learning_rate": 1.7640264686988013e-05, "loss": 0.7506, "step": 8019 }, { "epoch": 2.47, "learning_rate": 1.763962169642931e-05, "loss": 0.8629, "step": 8020 }, { "epoch": 2.47, "learning_rate": 1.7638978630002125e-05, "loss": 1.0622, "step": 8021 }, { "epoch": 2.47, "learning_rate": 1.7638335487712836e-05, "loss": 0.8742, "step": 8022 }, { "epoch": 2.47, "learning_rate": 1.763769226956784e-05, "loss": 0.7922, "step": 8023 }, { "epoch": 2.47, "learning_rate": 1.7637048975573523e-05, "loss": 0.6849, "step": 8024 }, { "epoch": 2.47, "learning_rate": 1.763640560573627e-05, "loss": 0.805, "step": 8025 }, { "epoch": 2.47, "learning_rate": 1.763576216006247e-05, "loss": 0.7389, "step": 8026 }, { "epoch": 2.47, "learning_rate": 1.7635118638558514e-05, "loss": 0.8108, "step": 8027 }, { "epoch": 2.47, "learning_rate": 1.7634475041230796e-05, "loss": 0.8033, "step": 8028 }, { "epoch": 2.47, "learning_rate": 1.76338313680857e-05, "loss": 0.6817, "step": 8029 }, { "epoch": 2.47, "learning_rate": 1.7633187619129626e-05, "loss": 0.8024, "step": 8030 }, { "epoch": 2.47, "learning_rate": 1.7632543794368967e-05, "loss": 0.7426, "step": 8031 }, { "epoch": 2.47, "learning_rate": 1.7631899893810108e-05, "loss": 0.8513, "step": 8032 }, { "epoch": 2.47, "learning_rate": 1.7631255917459453e-05, "loss": 0.7964, "step": 8033 }, { "epoch": 2.47, "learning_rate": 1.763061186532339e-05, "loss": 0.7595, "step": 8034 }, { "epoch": 2.47, "learning_rate": 1.7629967737408322e-05, "loss": 0.6747, "step": 8035 }, { "epoch": 2.47, "learning_rate": 1.762932353372064e-05, "loss": 0.5739, "step": 8036 }, { "epoch": 2.47, "learning_rate": 1.7628679254266747e-05, "loss": 1.0377, "step": 8037 }, { "epoch": 2.47, "learning_rate": 1.7628034899053036e-05, "loss": 1.0777, "step": 8038 }, { "epoch": 2.47, "learning_rate": 1.762739046808591e-05, "loss": 0.7593, "step": 8039 }, { "epoch": 2.47, "learning_rate": 1.7626745961371768e-05, "loss": 1.1224, "step": 8040 }, { "epoch": 2.47, "learning_rate": 1.7626101378917004e-05, "loss": 0.7048, "step": 8041 }, { "epoch": 2.47, "learning_rate": 1.7625456720728028e-05, "loss": 0.6495, "step": 8042 }, { "epoch": 2.47, "learning_rate": 1.762481198681124e-05, "loss": 1.1582, "step": 8043 }, { "epoch": 2.47, "learning_rate": 1.762416717717304e-05, "loss": 1.0568, "step": 8044 }, { "epoch": 2.48, "learning_rate": 1.7623522291819837e-05, "loss": 1.0483, "step": 8045 }, { "epoch": 2.48, "learning_rate": 1.7622877330758025e-05, "loss": 0.8545, "step": 8046 }, { "epoch": 2.48, "learning_rate": 1.7622232293994023e-05, "loss": 0.962, "step": 8047 }, { "epoch": 2.48, "learning_rate": 1.7621587181534226e-05, "loss": 0.5118, "step": 8048 }, { "epoch": 2.48, "learning_rate": 1.7620941993385043e-05, "loss": 0.83, "step": 8049 }, { "epoch": 2.48, "learning_rate": 1.7620296729552883e-05, "loss": 0.7565, "step": 8050 }, { "epoch": 2.48, "learning_rate": 1.7619651390044153e-05, "loss": 0.8112, "step": 8051 }, { "epoch": 2.48, "learning_rate": 1.7619005974865265e-05, "loss": 0.9035, "step": 8052 }, { "epoch": 2.48, "learning_rate": 1.761836048402262e-05, "loss": 0.8034, "step": 8053 }, { "epoch": 2.48, "learning_rate": 1.761771491752264e-05, "loss": 0.6988, "step": 8054 }, { "epoch": 2.48, "learning_rate": 1.761706927537173e-05, "loss": 0.7409, "step": 8055 }, { "epoch": 2.48, "learning_rate": 1.76164235575763e-05, "loss": 0.9198, "step": 8056 }, { "epoch": 2.48, "learning_rate": 1.7615777764142763e-05, "loss": 0.7682, "step": 8057 }, { "epoch": 2.48, "learning_rate": 1.7615131895077535e-05, "loss": 0.8091, "step": 8058 }, { "epoch": 2.48, "learning_rate": 1.761448595038703e-05, "loss": 0.9762, "step": 8059 }, { "epoch": 2.48, "learning_rate": 1.7613839930077657e-05, "loss": 0.8277, "step": 8060 }, { "epoch": 2.48, "learning_rate": 1.7613193834155844e-05, "loss": 0.7402, "step": 8061 }, { "epoch": 2.48, "learning_rate": 1.7612547662627996e-05, "loss": 0.6014, "step": 8062 }, { "epoch": 2.48, "learning_rate": 1.7611901415500536e-05, "loss": 0.8521, "step": 8063 }, { "epoch": 2.48, "learning_rate": 1.7611255092779878e-05, "loss": 0.8315, "step": 8064 }, { "epoch": 2.48, "learning_rate": 1.761060869447244e-05, "loss": 0.7785, "step": 8065 }, { "epoch": 2.48, "learning_rate": 1.7609962220584647e-05, "loss": 0.9628, "step": 8066 }, { "epoch": 2.48, "learning_rate": 1.7609315671122912e-05, "loss": 0.6601, "step": 8067 }, { "epoch": 2.48, "learning_rate": 1.7608669046093664e-05, "loss": 0.8623, "step": 8068 }, { "epoch": 2.48, "learning_rate": 1.7608022345503318e-05, "loss": 0.8532, "step": 8069 }, { "epoch": 2.48, "learning_rate": 1.7607375569358293e-05, "loss": 0.989, "step": 8070 }, { "epoch": 2.48, "learning_rate": 1.7606728717665024e-05, "loss": 0.7671, "step": 8071 }, { "epoch": 2.48, "learning_rate": 1.7606081790429925e-05, "loss": 0.9282, "step": 8072 }, { "epoch": 2.48, "learning_rate": 1.7605434787659426e-05, "loss": 0.7762, "step": 8073 }, { "epoch": 2.48, "learning_rate": 1.760478770935995e-05, "loss": 0.8844, "step": 8074 }, { "epoch": 2.48, "learning_rate": 1.760414055553792e-05, "loss": 1.0308, "step": 8075 }, { "epoch": 2.48, "learning_rate": 1.7603493326199773e-05, "loss": 0.5455, "step": 8076 }, { "epoch": 2.49, "learning_rate": 1.7602846021351925e-05, "loss": 0.7656, "step": 8077 }, { "epoch": 2.49, "learning_rate": 1.760219864100081e-05, "loss": 0.4907, "step": 8078 }, { "epoch": 2.49, "learning_rate": 1.7601551185152854e-05, "loss": 0.7994, "step": 8079 }, { "epoch": 2.49, "learning_rate": 1.760090365381449e-05, "loss": 0.7693, "step": 8080 }, { "epoch": 2.49, "learning_rate": 1.760025604699215e-05, "loss": 0.5523, "step": 8081 }, { "epoch": 2.49, "learning_rate": 1.759960836469226e-05, "loss": 0.5953, "step": 8082 }, { "epoch": 2.49, "learning_rate": 1.7598960606921258e-05, "loss": 0.8376, "step": 8083 }, { "epoch": 2.49, "learning_rate": 1.759831277368557e-05, "loss": 0.7981, "step": 8084 }, { "epoch": 2.49, "learning_rate": 1.7597664864991636e-05, "loss": 0.7327, "step": 8085 }, { "epoch": 2.49, "learning_rate": 1.7597016880845887e-05, "loss": 0.7963, "step": 8086 }, { "epoch": 2.49, "learning_rate": 1.759636882125476e-05, "loss": 0.8281, "step": 8087 }, { "epoch": 2.49, "learning_rate": 1.7595720686224687e-05, "loss": 0.8362, "step": 8088 }, { "epoch": 2.49, "learning_rate": 1.7595072475762107e-05, "loss": 0.7684, "step": 8089 }, { "epoch": 2.49, "learning_rate": 1.7594424189873458e-05, "loss": 0.8145, "step": 8090 }, { "epoch": 2.49, "learning_rate": 1.7593775828565182e-05, "loss": 0.9765, "step": 8091 }, { "epoch": 2.49, "learning_rate": 1.7593127391843706e-05, "loss": 0.9787, "step": 8092 }, { "epoch": 2.49, "learning_rate": 1.759247887971548e-05, "loss": 0.7655, "step": 8093 }, { "epoch": 2.49, "learning_rate": 1.7591830292186944e-05, "loss": 0.9697, "step": 8094 }, { "epoch": 2.49, "learning_rate": 1.7591181629264534e-05, "loss": 1.1025, "step": 8095 }, { "epoch": 2.49, "learning_rate": 1.7590532890954693e-05, "loss": 0.9035, "step": 8096 }, { "epoch": 2.49, "learning_rate": 1.7589884077263867e-05, "loss": 0.891, "step": 8097 }, { "epoch": 2.49, "learning_rate": 1.7589235188198496e-05, "loss": 0.737, "step": 8098 }, { "epoch": 2.49, "learning_rate": 1.758858622376503e-05, "loss": 0.7866, "step": 8099 }, { "epoch": 2.49, "learning_rate": 1.75879371839699e-05, "loss": 0.6448, "step": 8100 }, { "epoch": 2.49, "learning_rate": 1.7587288068819564e-05, "loss": 0.6084, "step": 8101 }, { "epoch": 2.49, "learning_rate": 1.7586638878320467e-05, "loss": 1.1869, "step": 8102 }, { "epoch": 2.49, "learning_rate": 1.7585989612479052e-05, "loss": 0.5605, "step": 8103 }, { "epoch": 2.49, "learning_rate": 1.7585340271301768e-05, "loss": 0.7024, "step": 8104 }, { "epoch": 2.49, "learning_rate": 1.7584690854795066e-05, "loss": 0.7912, "step": 8105 }, { "epoch": 2.49, "learning_rate": 1.7584041362965397e-05, "loss": 0.8458, "step": 8106 }, { "epoch": 2.49, "learning_rate": 1.75833917958192e-05, "loss": 0.8049, "step": 8107 }, { "epoch": 2.49, "learning_rate": 1.758274215336294e-05, "loss": 0.7104, "step": 8108 }, { "epoch": 2.49, "learning_rate": 1.7582092435603058e-05, "loss": 0.8075, "step": 8109 }, { "epoch": 2.5, "learning_rate": 1.7581442642546013e-05, "loss": 0.7346, "step": 8110 }, { "epoch": 2.5, "learning_rate": 1.7580792774198252e-05, "loss": 0.689, "step": 8111 }, { "epoch": 2.5, "learning_rate": 1.7580142830566233e-05, "loss": 0.6944, "step": 8112 }, { "epoch": 2.5, "learning_rate": 1.757949281165641e-05, "loss": 0.9077, "step": 8113 }, { "epoch": 2.5, "learning_rate": 1.757884271747524e-05, "loss": 0.5934, "step": 8114 }, { "epoch": 2.5, "learning_rate": 1.7578192548029174e-05, "loss": 0.5713, "step": 8115 }, { "epoch": 2.5, "learning_rate": 1.7577542303324674e-05, "loss": 0.8161, "step": 8116 }, { "epoch": 2.5, "learning_rate": 1.7576891983368192e-05, "loss": 1.0657, "step": 8117 }, { "epoch": 2.5, "learning_rate": 1.7576241588166196e-05, "loss": 0.6442, "step": 8118 }, { "epoch": 2.5, "learning_rate": 1.7575591117725132e-05, "loss": 0.8194, "step": 8119 }, { "epoch": 2.5, "learning_rate": 1.757494057205147e-05, "loss": 0.9786, "step": 8120 }, { "epoch": 2.5, "learning_rate": 1.7574289951151666e-05, "loss": 0.5245, "step": 8121 }, { "epoch": 2.5, "learning_rate": 1.757363925503218e-05, "loss": 0.63, "step": 8122 }, { "epoch": 2.5, "learning_rate": 1.757298848369948e-05, "loss": 0.868, "step": 8123 }, { "epoch": 2.5, "learning_rate": 1.757233763716002e-05, "loss": 0.7218, "step": 8124 }, { "epoch": 2.5, "learning_rate": 1.757168671542027e-05, "loss": 0.9273, "step": 8125 }, { "epoch": 2.5, "learning_rate": 1.7571035718486693e-05, "loss": 0.9598, "step": 8126 }, { "epoch": 2.5, "learning_rate": 1.7570384646365756e-05, "loss": 0.8489, "step": 8127 }, { "epoch": 2.5, "learning_rate": 1.756973349906392e-05, "loss": 0.7245, "step": 8128 }, { "epoch": 2.5, "learning_rate": 1.7569082276587653e-05, "loss": 0.924, "step": 8129 }, { "epoch": 2.5, "learning_rate": 1.7568430978943426e-05, "loss": 0.9123, "step": 8130 }, { "epoch": 2.5, "learning_rate": 1.7567779606137706e-05, "loss": 0.7864, "step": 8131 }, { "epoch": 2.5, "learning_rate": 1.7567128158176955e-05, "loss": 0.7441, "step": 8132 }, { "epoch": 2.5, "learning_rate": 1.7566476635067648e-05, "loss": 0.8088, "step": 8133 }, { "epoch": 2.5, "learning_rate": 1.7565825036816255e-05, "loss": 1.2382, "step": 8134 }, { "epoch": 2.5, "learning_rate": 1.7565173363429246e-05, "loss": 0.9387, "step": 8135 }, { "epoch": 2.5, "learning_rate": 1.7564521614913097e-05, "loss": 0.8366, "step": 8136 }, { "epoch": 2.5, "learning_rate": 1.7563869791274274e-05, "loss": 0.8204, "step": 8137 }, { "epoch": 2.5, "learning_rate": 1.756321789251925e-05, "loss": 0.8309, "step": 8138 }, { "epoch": 2.5, "learning_rate": 1.7562565918654507e-05, "loss": 0.7729, "step": 8139 }, { "epoch": 2.5, "learning_rate": 1.7561913869686512e-05, "loss": 0.9336, "step": 8140 }, { "epoch": 2.5, "learning_rate": 1.7561261745621742e-05, "loss": 1.0298, "step": 8141 }, { "epoch": 2.51, "learning_rate": 1.7560609546466674e-05, "loss": 0.7767, "step": 8142 }, { "epoch": 2.51, "learning_rate": 1.755995727222779e-05, "loss": 0.9419, "step": 8143 }, { "epoch": 2.51, "learning_rate": 1.755930492291156e-05, "loss": 0.8374, "step": 8144 }, { "epoch": 2.51, "learning_rate": 1.7558652498524464e-05, "loss": 1.0972, "step": 8145 }, { "epoch": 2.51, "learning_rate": 1.755799999907298e-05, "loss": 0.7224, "step": 8146 }, { "epoch": 2.51, "learning_rate": 1.7557347424563597e-05, "loss": 1.0593, "step": 8147 }, { "epoch": 2.51, "learning_rate": 1.7556694775002784e-05, "loss": 0.9906, "step": 8148 }, { "epoch": 2.51, "learning_rate": 1.755604205039703e-05, "loss": 0.7342, "step": 8149 }, { "epoch": 2.51, "learning_rate": 1.7555389250752813e-05, "loss": 0.9766, "step": 8150 }, { "epoch": 2.51, "learning_rate": 1.7554736376076616e-05, "loss": 0.818, "step": 8151 }, { "epoch": 2.51, "learning_rate": 1.7554083426374927e-05, "loss": 0.6976, "step": 8152 }, { "epoch": 2.51, "learning_rate": 1.7553430401654227e-05, "loss": 0.867, "step": 8153 }, { "epoch": 2.51, "learning_rate": 1.7552777301921004e-05, "loss": 0.6824, "step": 8154 }, { "epoch": 2.51, "learning_rate": 1.755212412718174e-05, "loss": 0.9647, "step": 8155 }, { "epoch": 2.51, "learning_rate": 1.755147087744292e-05, "loss": 0.7907, "step": 8156 }, { "epoch": 2.51, "learning_rate": 1.7550817552711036e-05, "loss": 0.6442, "step": 8157 }, { "epoch": 2.51, "learning_rate": 1.7550164152992573e-05, "loss": 0.9482, "step": 8158 }, { "epoch": 2.51, "learning_rate": 1.7549510678294022e-05, "loss": 0.8432, "step": 8159 }, { "epoch": 2.51, "learning_rate": 1.7548857128621878e-05, "loss": 0.9784, "step": 8160 }, { "epoch": 2.51, "learning_rate": 1.7548203503982622e-05, "loss": 0.8364, "step": 8161 }, { "epoch": 2.51, "learning_rate": 1.7547549804382746e-05, "loss": 0.6558, "step": 8162 }, { "epoch": 2.51, "learning_rate": 1.7546896029828745e-05, "loss": 1.0172, "step": 8163 }, { "epoch": 2.51, "learning_rate": 1.7546242180327114e-05, "loss": 0.7118, "step": 8164 }, { "epoch": 2.51, "learning_rate": 1.754558825588434e-05, "loss": 0.6889, "step": 8165 }, { "epoch": 2.51, "learning_rate": 1.754493425650692e-05, "loss": 0.7467, "step": 8166 }, { "epoch": 2.51, "learning_rate": 1.7544280182201353e-05, "loss": 0.927, "step": 8167 }, { "epoch": 2.51, "learning_rate": 1.754362603297413e-05, "loss": 0.8186, "step": 8168 }, { "epoch": 2.51, "learning_rate": 1.7542971808831746e-05, "loss": 0.7357, "step": 8169 }, { "epoch": 2.51, "learning_rate": 1.7542317509780704e-05, "loss": 0.7732, "step": 8170 }, { "epoch": 2.51, "learning_rate": 1.7541663135827493e-05, "loss": 0.7238, "step": 8171 }, { "epoch": 2.51, "learning_rate": 1.754100868697862e-05, "loss": 0.6994, "step": 8172 }, { "epoch": 2.51, "learning_rate": 1.754035416324058e-05, "loss": 1.1285, "step": 8173 }, { "epoch": 2.51, "learning_rate": 1.7539699564619872e-05, "loss": 0.6364, "step": 8174 }, { "epoch": 2.52, "learning_rate": 1.7539044891123e-05, "loss": 1.104, "step": 8175 }, { "epoch": 2.52, "learning_rate": 1.7538390142756464e-05, "loss": 0.7191, "step": 8176 }, { "epoch": 2.52, "learning_rate": 1.7537735319526766e-05, "loss": 0.6388, "step": 8177 }, { "epoch": 2.52, "learning_rate": 1.753708042144041e-05, "loss": 0.946, "step": 8178 }, { "epoch": 2.52, "learning_rate": 1.75364254485039e-05, "loss": 1.0621, "step": 8179 }, { "epoch": 2.52, "learning_rate": 1.7535770400723736e-05, "loss": 0.5594, "step": 8180 }, { "epoch": 2.52, "learning_rate": 1.7535115278106428e-05, "loss": 0.9261, "step": 8181 }, { "epoch": 2.52, "learning_rate": 1.7534460080658486e-05, "loss": 0.5086, "step": 8182 }, { "epoch": 2.52, "learning_rate": 1.7533804808386404e-05, "loss": 0.5542, "step": 8183 }, { "epoch": 2.52, "learning_rate": 1.75331494612967e-05, "loss": 0.6166, "step": 8184 }, { "epoch": 2.52, "learning_rate": 1.7532494039395877e-05, "loss": 0.9718, "step": 8185 }, { "epoch": 2.52, "learning_rate": 1.753183854269045e-05, "loss": 0.7349, "step": 8186 }, { "epoch": 2.52, "learning_rate": 1.7531182971186923e-05, "loss": 0.7866, "step": 8187 }, { "epoch": 2.52, "learning_rate": 1.7530527324891808e-05, "loss": 0.6405, "step": 8188 }, { "epoch": 2.52, "learning_rate": 1.7529871603811614e-05, "loss": 0.7262, "step": 8189 }, { "epoch": 2.52, "learning_rate": 1.752921580795286e-05, "loss": 0.7519, "step": 8190 }, { "epoch": 2.52, "learning_rate": 1.752855993732205e-05, "loss": 0.8484, "step": 8191 }, { "epoch": 2.52, "learning_rate": 1.7527903991925702e-05, "loss": 0.6633, "step": 8192 }, { "epoch": 2.52, "learning_rate": 1.752724797177033e-05, "loss": 0.8, "step": 8193 }, { "epoch": 2.52, "learning_rate": 1.752659187686245e-05, "loss": 0.8162, "step": 8194 }, { "epoch": 2.52, "learning_rate": 1.752593570720857e-05, "loss": 0.7859, "step": 8195 }, { "epoch": 2.52, "learning_rate": 1.752527946281522e-05, "loss": 0.782, "step": 8196 }, { "epoch": 2.52, "learning_rate": 1.7524623143688905e-05, "loss": 0.6971, "step": 8197 }, { "epoch": 2.52, "learning_rate": 1.7523966749836148e-05, "loss": 0.7045, "step": 8198 }, { "epoch": 2.52, "learning_rate": 1.7523310281263466e-05, "loss": 0.8843, "step": 8199 }, { "epoch": 2.52, "learning_rate": 1.7522653737977383e-05, "loss": 0.5809, "step": 8200 }, { "epoch": 2.52, "learning_rate": 1.752199711998441e-05, "loss": 0.6629, "step": 8201 }, { "epoch": 2.52, "learning_rate": 1.7521340427291077e-05, "loss": 0.6869, "step": 8202 }, { "epoch": 2.52, "learning_rate": 1.7520683659903898e-05, "loss": 0.8603, "step": 8203 }, { "epoch": 2.52, "learning_rate": 1.7520026817829402e-05, "loss": 0.8254, "step": 8204 }, { "epoch": 2.52, "learning_rate": 1.751936990107411e-05, "loss": 0.9915, "step": 8205 }, { "epoch": 2.52, "learning_rate": 1.7518712909644544e-05, "loss": 0.8356, "step": 8206 }, { "epoch": 2.53, "learning_rate": 1.7518055843547228e-05, "loss": 0.6167, "step": 8207 }, { "epoch": 2.53, "learning_rate": 1.751739870278869e-05, "loss": 0.8162, "step": 8208 }, { "epoch": 2.53, "learning_rate": 1.751674148737545e-05, "loss": 0.9135, "step": 8209 }, { "epoch": 2.53, "learning_rate": 1.7516084197314044e-05, "loss": 0.9275, "step": 8210 }, { "epoch": 2.53, "learning_rate": 1.7515426832610994e-05, "loss": 0.7795, "step": 8211 }, { "epoch": 2.53, "learning_rate": 1.7514769393272833e-05, "loss": 0.6328, "step": 8212 }, { "epoch": 2.53, "learning_rate": 1.751411187930608e-05, "loss": 0.8084, "step": 8213 }, { "epoch": 2.53, "learning_rate": 1.7513454290717274e-05, "loss": 0.6442, "step": 8214 }, { "epoch": 2.53, "learning_rate": 1.7512796627512943e-05, "loss": 1.0435, "step": 8215 }, { "epoch": 2.53, "learning_rate": 1.7512138889699615e-05, "loss": 0.9156, "step": 8216 }, { "epoch": 2.53, "learning_rate": 1.7511481077283826e-05, "loss": 0.6713, "step": 8217 }, { "epoch": 2.53, "learning_rate": 1.7510823190272106e-05, "loss": 0.831, "step": 8218 }, { "epoch": 2.53, "learning_rate": 1.751016522867099e-05, "loss": 0.9349, "step": 8219 }, { "epoch": 2.53, "learning_rate": 1.7509507192487012e-05, "loss": 0.5564, "step": 8220 }, { "epoch": 2.53, "learning_rate": 1.7508849081726707e-05, "loss": 0.7491, "step": 8221 }, { "epoch": 2.53, "learning_rate": 1.750819089639661e-05, "loss": 0.7735, "step": 8222 }, { "epoch": 2.53, "learning_rate": 1.7507532636503256e-05, "loss": 0.7582, "step": 8223 }, { "epoch": 2.53, "learning_rate": 1.7506874302053188e-05, "loss": 0.8317, "step": 8224 }, { "epoch": 2.53, "learning_rate": 1.7506215893052935e-05, "loss": 0.7092, "step": 8225 }, { "epoch": 2.53, "learning_rate": 1.7505557409509042e-05, "loss": 0.8577, "step": 8226 }, { "epoch": 2.53, "learning_rate": 1.7504898851428046e-05, "loss": 0.7228, "step": 8227 }, { "epoch": 2.53, "learning_rate": 1.7504240218816488e-05, "loss": 0.8741, "step": 8228 }, { "epoch": 2.53, "learning_rate": 1.750358151168091e-05, "loss": 1.1027, "step": 8229 }, { "epoch": 2.53, "learning_rate": 1.7502922730027853e-05, "loss": 0.9095, "step": 8230 }, { "epoch": 2.53, "learning_rate": 1.7502263873863853e-05, "loss": 0.8063, "step": 8231 }, { "epoch": 2.53, "learning_rate": 1.7501604943195466e-05, "loss": 0.8355, "step": 8232 }, { "epoch": 2.53, "learning_rate": 1.7500945938029223e-05, "loss": 0.9535, "step": 8233 }, { "epoch": 2.53, "learning_rate": 1.750028685837167e-05, "loss": 0.847, "step": 8234 }, { "epoch": 2.53, "learning_rate": 1.7499627704229364e-05, "loss": 0.8445, "step": 8235 }, { "epoch": 2.53, "learning_rate": 1.749896847560884e-05, "loss": 1.0344, "step": 8236 }, { "epoch": 2.53, "learning_rate": 1.7498309172516647e-05, "loss": 0.6909, "step": 8237 }, { "epoch": 2.53, "learning_rate": 1.7497649794959334e-05, "loss": 0.9083, "step": 8238 }, { "epoch": 2.53, "learning_rate": 1.749699034294345e-05, "loss": 0.7826, "step": 8239 }, { "epoch": 2.54, "learning_rate": 1.749633081647554e-05, "loss": 0.8532, "step": 8240 }, { "epoch": 2.54, "learning_rate": 1.7495671215562157e-05, "loss": 0.8443, "step": 8241 }, { "epoch": 2.54, "learning_rate": 1.749501154020985e-05, "loss": 0.8859, "step": 8242 }, { "epoch": 2.54, "learning_rate": 1.7494351790425172e-05, "loss": 0.849, "step": 8243 }, { "epoch": 2.54, "learning_rate": 1.749369196621467e-05, "loss": 0.9264, "step": 8244 }, { "epoch": 2.54, "learning_rate": 1.749303206758491e-05, "loss": 0.9193, "step": 8245 }, { "epoch": 2.54, "learning_rate": 1.7492372094542425e-05, "loss": 0.8808, "step": 8246 }, { "epoch": 2.54, "learning_rate": 1.7491712047093785e-05, "loss": 1.0471, "step": 8247 }, { "epoch": 2.54, "learning_rate": 1.749105192524554e-05, "loss": 0.7087, "step": 8248 }, { "epoch": 2.54, "learning_rate": 1.7490391729004242e-05, "loss": 0.5041, "step": 8249 }, { "epoch": 2.54, "learning_rate": 1.7489731458376454e-05, "loss": 0.6297, "step": 8250 }, { "epoch": 2.54, "learning_rate": 1.748907111336873e-05, "loss": 0.7692, "step": 8251 }, { "epoch": 2.54, "learning_rate": 1.7488410693987627e-05, "loss": 0.7657, "step": 8252 }, { "epoch": 2.54, "learning_rate": 1.7487750200239702e-05, "loss": 0.7653, "step": 8253 }, { "epoch": 2.54, "learning_rate": 1.748708963213152e-05, "loss": 0.8072, "step": 8254 }, { "epoch": 2.54, "learning_rate": 1.7486428989669635e-05, "loss": 0.9304, "step": 8255 }, { "epoch": 2.54, "learning_rate": 1.7485768272860616e-05, "loss": 0.695, "step": 8256 }, { "epoch": 2.54, "learning_rate": 1.7485107481711014e-05, "loss": 0.857, "step": 8257 }, { "epoch": 2.54, "learning_rate": 1.7484446616227397e-05, "loss": 0.8898, "step": 8258 }, { "epoch": 2.54, "learning_rate": 1.7483785676416326e-05, "loss": 0.7798, "step": 8259 }, { "epoch": 2.54, "learning_rate": 1.7483124662284366e-05, "loss": 0.82, "step": 8260 }, { "epoch": 2.54, "learning_rate": 1.7482463573838086e-05, "loss": 0.7882, "step": 8261 }, { "epoch": 2.54, "learning_rate": 1.748180241108404e-05, "loss": 1.0247, "step": 8262 }, { "epoch": 2.54, "learning_rate": 1.7481141174028807e-05, "loss": 0.9248, "step": 8263 }, { "epoch": 2.54, "learning_rate": 1.7480479862678942e-05, "loss": 1.0061, "step": 8264 }, { "epoch": 2.54, "learning_rate": 1.747981847704102e-05, "loss": 0.6437, "step": 8265 }, { "epoch": 2.54, "learning_rate": 1.7479157017121608e-05, "loss": 0.8757, "step": 8266 }, { "epoch": 2.54, "learning_rate": 1.7478495482927275e-05, "loss": 0.8602, "step": 8267 }, { "epoch": 2.54, "learning_rate": 1.7477833874464584e-05, "loss": 0.4631, "step": 8268 }, { "epoch": 2.54, "learning_rate": 1.7477172191740113e-05, "loss": 0.6706, "step": 8269 }, { "epoch": 2.54, "learning_rate": 1.7476510434760435e-05, "loss": 0.8109, "step": 8270 }, { "epoch": 2.54, "learning_rate": 1.7475848603532114e-05, "loss": 0.8295, "step": 8271 }, { "epoch": 2.55, "learning_rate": 1.747518669806173e-05, "loss": 0.7215, "step": 8272 }, { "epoch": 2.55, "learning_rate": 1.7474524718355848e-05, "loss": 0.9148, "step": 8273 }, { "epoch": 2.55, "learning_rate": 1.7473862664421048e-05, "loss": 0.6812, "step": 8274 }, { "epoch": 2.55, "learning_rate": 1.7473200536263905e-05, "loss": 0.6222, "step": 8275 }, { "epoch": 2.55, "learning_rate": 1.7472538333890992e-05, "loss": 0.6939, "step": 8276 }, { "epoch": 2.55, "learning_rate": 1.7471876057308886e-05, "loss": 0.8739, "step": 8277 }, { "epoch": 2.55, "learning_rate": 1.747121370652417e-05, "loss": 0.9333, "step": 8278 }, { "epoch": 2.55, "learning_rate": 1.7470551281543412e-05, "loss": 0.7114, "step": 8279 }, { "epoch": 2.55, "learning_rate": 1.7469888782373194e-05, "loss": 0.8899, "step": 8280 }, { "epoch": 2.55, "learning_rate": 1.7469226209020094e-05, "loss": 0.7366, "step": 8281 }, { "epoch": 2.55, "learning_rate": 1.7468563561490697e-05, "loss": 0.8572, "step": 8282 }, { "epoch": 2.55, "learning_rate": 1.7467900839791577e-05, "loss": 0.8169, "step": 8283 }, { "epoch": 2.55, "learning_rate": 1.7467238043929326e-05, "loss": 0.9699, "step": 8284 }, { "epoch": 2.55, "learning_rate": 1.7466575173910512e-05, "loss": 0.9374, "step": 8285 }, { "epoch": 2.55, "learning_rate": 1.746591222974173e-05, "loss": 0.8872, "step": 8286 }, { "epoch": 2.55, "learning_rate": 1.746524921142956e-05, "loss": 0.7828, "step": 8287 }, { "epoch": 2.55, "learning_rate": 1.746458611898058e-05, "loss": 0.7039, "step": 8288 }, { "epoch": 2.55, "learning_rate": 1.7463922952401384e-05, "loss": 0.6456, "step": 8289 }, { "epoch": 2.55, "learning_rate": 1.746325971169855e-05, "loss": 0.8696, "step": 8290 }, { "epoch": 2.55, "learning_rate": 1.7462596396878673e-05, "loss": 0.9101, "step": 8291 }, { "epoch": 2.55, "learning_rate": 1.7461933007948333e-05, "loss": 0.6054, "step": 8292 }, { "epoch": 2.55, "learning_rate": 1.7461269544914127e-05, "loss": 0.8957, "step": 8293 }, { "epoch": 2.55, "learning_rate": 1.746060600778263e-05, "loss": 0.8844, "step": 8294 }, { "epoch": 2.55, "learning_rate": 1.7459942396560443e-05, "loss": 0.7223, "step": 8295 }, { "epoch": 2.55, "learning_rate": 1.7459278711254152e-05, "loss": 0.6991, "step": 8296 }, { "epoch": 2.55, "learning_rate": 1.745861495187035e-05, "loss": 0.7564, "step": 8297 }, { "epoch": 2.55, "learning_rate": 1.7457951118415627e-05, "loss": 0.7186, "step": 8298 }, { "epoch": 2.55, "learning_rate": 1.7457287210896575e-05, "loss": 0.8266, "step": 8299 }, { "epoch": 2.55, "learning_rate": 1.745662322931979e-05, "loss": 0.9371, "step": 8300 }, { "epoch": 2.55, "learning_rate": 1.7455959173691863e-05, "loss": 0.7481, "step": 8301 }, { "epoch": 2.55, "learning_rate": 1.745529504401939e-05, "loss": 0.3898, "step": 8302 }, { "epoch": 2.55, "learning_rate": 1.7454630840308965e-05, "loss": 0.8487, "step": 8303 }, { "epoch": 2.55, "learning_rate": 1.7453966562567187e-05, "loss": 0.7222, "step": 8304 }, { "epoch": 2.56, "learning_rate": 1.7453302210800652e-05, "loss": 0.8414, "step": 8305 }, { "epoch": 2.56, "learning_rate": 1.7452637785015954e-05, "loss": 0.7586, "step": 8306 }, { "epoch": 2.56, "learning_rate": 1.7451973285219698e-05, "loss": 0.749, "step": 8307 }, { "epoch": 2.56, "learning_rate": 1.7451308711418475e-05, "loss": 0.8786, "step": 8308 }, { "epoch": 2.56, "learning_rate": 1.7450644063618892e-05, "loss": 1.0003, "step": 8309 }, { "epoch": 2.56, "learning_rate": 1.7449979341827545e-05, "loss": 0.8628, "step": 8310 }, { "epoch": 2.56, "learning_rate": 1.744931454605104e-05, "loss": 1.0747, "step": 8311 }, { "epoch": 2.56, "learning_rate": 1.7448649676295972e-05, "loss": 0.725, "step": 8312 }, { "epoch": 2.56, "learning_rate": 1.7447984732568953e-05, "loss": 0.5798, "step": 8313 }, { "epoch": 2.56, "learning_rate": 1.744731971487658e-05, "loss": 0.8779, "step": 8314 }, { "epoch": 2.56, "learning_rate": 1.744665462322545e-05, "loss": 0.8003, "step": 8315 }, { "epoch": 2.56, "learning_rate": 1.744598945762219e-05, "loss": 0.7395, "step": 8316 }, { "epoch": 2.56, "learning_rate": 1.7445324218073382e-05, "loss": 0.4272, "step": 8317 }, { "epoch": 2.56, "learning_rate": 1.744465890458565e-05, "loss": 0.7314, "step": 8318 }, { "epoch": 2.56, "learning_rate": 1.7443993517165587e-05, "loss": 0.5693, "step": 8319 }, { "epoch": 2.56, "learning_rate": 1.7443328055819813e-05, "loss": 0.707, "step": 8320 }, { "epoch": 2.56, "learning_rate": 1.7442662520554927e-05, "loss": 0.7521, "step": 8321 }, { "epoch": 2.56, "learning_rate": 1.7441996911377547e-05, "loss": 0.8849, "step": 8322 }, { "epoch": 2.56, "learning_rate": 1.7441331228294278e-05, "loss": 0.9998, "step": 8323 }, { "epoch": 2.56, "learning_rate": 1.7440665471311732e-05, "loss": 1.0532, "step": 8324 }, { "epoch": 2.56, "learning_rate": 1.743999964043652e-05, "loss": 0.8284, "step": 8325 }, { "epoch": 2.56, "learning_rate": 1.743933373567525e-05, "loss": 0.6921, "step": 8326 }, { "epoch": 2.56, "learning_rate": 1.7438667757034547e-05, "loss": 1.0128, "step": 8327 }, { "epoch": 2.56, "learning_rate": 1.7438001704521013e-05, "loss": 0.6209, "step": 8328 }, { "epoch": 2.56, "learning_rate": 1.7437335578141266e-05, "loss": 0.7152, "step": 8329 }, { "epoch": 2.56, "learning_rate": 1.7436669377901924e-05, "loss": 0.7463, "step": 8330 }, { "epoch": 2.56, "learning_rate": 1.7436003103809605e-05, "loss": 0.6611, "step": 8331 }, { "epoch": 2.56, "learning_rate": 1.743533675587092e-05, "loss": 0.7004, "step": 8332 }, { "epoch": 2.56, "learning_rate": 1.7434670334092484e-05, "loss": 0.9613, "step": 8333 }, { "epoch": 2.56, "learning_rate": 1.7434003838480924e-05, "loss": 1.0949, "step": 8334 }, { "epoch": 2.56, "learning_rate": 1.743333726904285e-05, "loss": 0.8227, "step": 8335 }, { "epoch": 2.56, "learning_rate": 1.743267062578489e-05, "loss": 0.6538, "step": 8336 }, { "epoch": 2.57, "learning_rate": 1.7432003908713663e-05, "loss": 0.8734, "step": 8337 }, { "epoch": 2.57, "learning_rate": 1.7431337117835783e-05, "loss": 0.563, "step": 8338 }, { "epoch": 2.57, "learning_rate": 1.7430670253157877e-05, "loss": 0.9603, "step": 8339 }, { "epoch": 2.57, "learning_rate": 1.743000331468657e-05, "loss": 0.7171, "step": 8340 }, { "epoch": 2.57, "learning_rate": 1.742933630242848e-05, "loss": 0.7567, "step": 8341 }, { "epoch": 2.57, "learning_rate": 1.7428669216390236e-05, "loss": 0.8791, "step": 8342 }, { "epoch": 2.57, "learning_rate": 1.7428002056578455e-05, "loss": 0.8357, "step": 8343 }, { "epoch": 2.57, "learning_rate": 1.7427334822999774e-05, "loss": 0.6626, "step": 8344 }, { "epoch": 2.57, "learning_rate": 1.742666751566081e-05, "loss": 0.7314, "step": 8345 }, { "epoch": 2.57, "learning_rate": 1.742600013456819e-05, "loss": 0.6724, "step": 8346 }, { "epoch": 2.57, "learning_rate": 1.742533267972855e-05, "loss": 0.7666, "step": 8347 }, { "epoch": 2.57, "learning_rate": 1.7424665151148507e-05, "loss": 0.7157, "step": 8348 }, { "epoch": 2.57, "learning_rate": 1.7423997548834703e-05, "loss": 0.705, "step": 8349 }, { "epoch": 2.57, "learning_rate": 1.7423329872793756e-05, "loss": 0.9304, "step": 8350 }, { "epoch": 2.57, "learning_rate": 1.7422662123032306e-05, "loss": 0.8462, "step": 8351 }, { "epoch": 2.57, "learning_rate": 1.7421994299556977e-05, "loss": 0.8267, "step": 8352 }, { "epoch": 2.57, "learning_rate": 1.7421326402374406e-05, "loss": 0.8265, "step": 8353 }, { "epoch": 2.57, "learning_rate": 1.7420658431491224e-05, "loss": 0.949, "step": 8354 }, { "epoch": 2.57, "learning_rate": 1.7419990386914064e-05, "loss": 0.8108, "step": 8355 }, { "epoch": 2.57, "learning_rate": 1.7419322268649562e-05, "loss": 0.7083, "step": 8356 }, { "epoch": 2.57, "learning_rate": 1.741865407670435e-05, "loss": 0.901, "step": 8357 }, { "epoch": 2.57, "learning_rate": 1.7417985811085066e-05, "loss": 0.5971, "step": 8358 }, { "epoch": 2.57, "learning_rate": 1.7417317471798348e-05, "loss": 0.677, "step": 8359 }, { "epoch": 2.57, "learning_rate": 1.7416649058850832e-05, "loss": 0.8101, "step": 8360 }, { "epoch": 2.57, "learning_rate": 1.7415980572249153e-05, "loss": 0.9136, "step": 8361 }, { "epoch": 2.57, "learning_rate": 1.741531201199995e-05, "loss": 0.8726, "step": 8362 }, { "epoch": 2.57, "learning_rate": 1.7414643378109872e-05, "loss": 0.9199, "step": 8363 }, { "epoch": 2.57, "learning_rate": 1.7413974670585547e-05, "loss": 0.7011, "step": 8364 }, { "epoch": 2.57, "learning_rate": 1.741330588943362e-05, "loss": 0.931, "step": 8365 }, { "epoch": 2.57, "learning_rate": 1.7412637034660735e-05, "loss": 0.87, "step": 8366 }, { "epoch": 2.57, "learning_rate": 1.7411968106273532e-05, "loss": 0.7447, "step": 8367 }, { "epoch": 2.57, "learning_rate": 1.7411299104278658e-05, "loss": 0.9875, "step": 8368 }, { "epoch": 2.57, "learning_rate": 1.7410630028682746e-05, "loss": 0.7905, "step": 8369 }, { "epoch": 2.58, "learning_rate": 1.7409960879492454e-05, "loss": 0.8029, "step": 8370 }, { "epoch": 2.58, "learning_rate": 1.740929165671442e-05, "loss": 0.8204, "step": 8371 }, { "epoch": 2.58, "learning_rate": 1.7408622360355288e-05, "loss": 0.9967, "step": 8372 }, { "epoch": 2.58, "learning_rate": 1.740795299042171e-05, "loss": 0.8697, "step": 8373 }, { "epoch": 2.58, "learning_rate": 1.7407283546920336e-05, "loss": 0.8, "step": 8374 }, { "epoch": 2.58, "learning_rate": 1.74066140298578e-05, "loss": 0.7074, "step": 8375 }, { "epoch": 2.58, "learning_rate": 1.740594443924077e-05, "loss": 1.0958, "step": 8376 }, { "epoch": 2.58, "learning_rate": 1.7405274775075882e-05, "loss": 0.8491, "step": 8377 }, { "epoch": 2.58, "learning_rate": 1.740460503736979e-05, "loss": 0.7295, "step": 8378 }, { "epoch": 2.58, "learning_rate": 1.740393522612915e-05, "loss": 0.6423, "step": 8379 }, { "epoch": 2.58, "learning_rate": 1.7403265341360604e-05, "loss": 0.9497, "step": 8380 }, { "epoch": 2.58, "learning_rate": 1.7402595383070814e-05, "loss": 0.6725, "step": 8381 }, { "epoch": 2.58, "learning_rate": 1.7401925351266425e-05, "loss": 0.8222, "step": 8382 }, { "epoch": 2.58, "learning_rate": 1.7401255245954098e-05, "loss": 0.8967, "step": 8383 }, { "epoch": 2.58, "learning_rate": 1.7400585067140486e-05, "loss": 0.7342, "step": 8384 }, { "epoch": 2.58, "learning_rate": 1.7399914814832245e-05, "loss": 0.7883, "step": 8385 }, { "epoch": 2.58, "learning_rate": 1.7399244489036025e-05, "loss": 0.8009, "step": 8386 }, { "epoch": 2.58, "learning_rate": 1.739857408975849e-05, "loss": 1.0579, "step": 8387 }, { "epoch": 2.58, "learning_rate": 1.7397903617006297e-05, "loss": 0.7596, "step": 8388 }, { "epoch": 2.58, "learning_rate": 1.7397233070786103e-05, "loss": 0.7826, "step": 8389 }, { "epoch": 2.58, "learning_rate": 1.7396562451104565e-05, "loss": 0.7571, "step": 8390 }, { "epoch": 2.58, "learning_rate": 1.739589175796835e-05, "loss": 0.6392, "step": 8391 }, { "epoch": 2.58, "learning_rate": 1.739522099138411e-05, "loss": 0.8327, "step": 8392 }, { "epoch": 2.58, "learning_rate": 1.739455015135851e-05, "loss": 0.8926, "step": 8393 }, { "epoch": 2.58, "learning_rate": 1.739387923789821e-05, "loss": 0.6581, "step": 8394 }, { "epoch": 2.58, "learning_rate": 1.739320825100988e-05, "loss": 1.0503, "step": 8395 }, { "epoch": 2.58, "learning_rate": 1.7392537190700172e-05, "loss": 0.7996, "step": 8396 }, { "epoch": 2.58, "learning_rate": 1.739186605697576e-05, "loss": 0.864, "step": 8397 }, { "epoch": 2.58, "learning_rate": 1.739119484984331e-05, "loss": 0.9178, "step": 8398 }, { "epoch": 2.58, "learning_rate": 1.7390523569309474e-05, "loss": 0.846, "step": 8399 }, { "epoch": 2.58, "learning_rate": 1.7389852215380932e-05, "loss": 0.9443, "step": 8400 }, { "epoch": 2.58, "learning_rate": 1.7389180788064345e-05, "loss": 0.8138, "step": 8401 }, { "epoch": 2.59, "learning_rate": 1.7388509287366386e-05, "loss": 0.6987, "step": 8402 }, { "epoch": 2.59, "learning_rate": 1.7387837713293717e-05, "loss": 0.741, "step": 8403 }, { "epoch": 2.59, "learning_rate": 1.7387166065853012e-05, "loss": 0.8839, "step": 8404 }, { "epoch": 2.59, "learning_rate": 1.7386494345050944e-05, "loss": 0.9125, "step": 8405 }, { "epoch": 2.59, "learning_rate": 1.7385822550894175e-05, "loss": 0.6668, "step": 8406 }, { "epoch": 2.59, "learning_rate": 1.738515068338938e-05, "loss": 0.718, "step": 8407 }, { "epoch": 2.59, "learning_rate": 1.7384478742543236e-05, "loss": 0.8447, "step": 8408 }, { "epoch": 2.59, "learning_rate": 1.738380672836241e-05, "loss": 0.8849, "step": 8409 }, { "epoch": 2.59, "learning_rate": 1.738313464085358e-05, "loss": 0.6374, "step": 8410 }, { "epoch": 2.59, "learning_rate": 1.7382462480023418e-05, "loss": 0.8133, "step": 8411 }, { "epoch": 2.59, "learning_rate": 1.7381790245878597e-05, "loss": 0.7824, "step": 8412 }, { "epoch": 2.59, "learning_rate": 1.73811179384258e-05, "loss": 0.7204, "step": 8413 }, { "epoch": 2.59, "learning_rate": 1.7380445557671693e-05, "loss": 0.789, "step": 8414 }, { "epoch": 2.59, "learning_rate": 1.7379773103622963e-05, "loss": 1.0044, "step": 8415 }, { "epoch": 2.59, "learning_rate": 1.7379100576286287e-05, "loss": 0.8394, "step": 8416 }, { "epoch": 2.59, "learning_rate": 1.737842797566834e-05, "loss": 0.7494, "step": 8417 }, { "epoch": 2.59, "learning_rate": 1.73777553017758e-05, "loss": 0.8198, "step": 8418 }, { "epoch": 2.59, "learning_rate": 1.7377082554615355e-05, "loss": 0.8495, "step": 8419 }, { "epoch": 2.59, "learning_rate": 1.7376409734193678e-05, "loss": 0.6172, "step": 8420 }, { "epoch": 2.59, "learning_rate": 1.7375736840517453e-05, "loss": 0.5639, "step": 8421 }, { "epoch": 2.59, "learning_rate": 1.737506387359337e-05, "loss": 0.8694, "step": 8422 }, { "epoch": 2.59, "learning_rate": 1.7374390833428097e-05, "loss": 0.965, "step": 8423 }, { "epoch": 2.59, "learning_rate": 1.7373717720028335e-05, "loss": 0.5609, "step": 8424 }, { "epoch": 2.59, "learning_rate": 1.7373044533400755e-05, "loss": 0.8282, "step": 8425 }, { "epoch": 2.59, "learning_rate": 1.7372371273552045e-05, "loss": 0.753, "step": 8426 }, { "epoch": 2.59, "learning_rate": 1.73716979404889e-05, "loss": 0.9544, "step": 8427 }, { "epoch": 2.59, "learning_rate": 1.7371024534218e-05, "loss": 1.0158, "step": 8428 }, { "epoch": 2.59, "learning_rate": 1.737035105474603e-05, "loss": 0.8841, "step": 8429 }, { "epoch": 2.59, "learning_rate": 1.7369677502079682e-05, "loss": 0.931, "step": 8430 }, { "epoch": 2.59, "learning_rate": 1.7369003876225644e-05, "loss": 0.8718, "step": 8431 }, { "epoch": 2.59, "learning_rate": 1.736833017719061e-05, "loss": 0.7224, "step": 8432 }, { "epoch": 2.59, "learning_rate": 1.7367656404981258e-05, "loss": 0.8828, "step": 8433 }, { "epoch": 2.59, "learning_rate": 1.7366982559604296e-05, "loss": 0.7366, "step": 8434 }, { "epoch": 2.6, "learning_rate": 1.7366308641066405e-05, "loss": 0.6745, "step": 8435 }, { "epoch": 2.6, "learning_rate": 1.7365634649374277e-05, "loss": 0.9832, "step": 8436 }, { "epoch": 2.6, "learning_rate": 1.7364960584534612e-05, "loss": 0.9523, "step": 8437 }, { "epoch": 2.6, "learning_rate": 1.73642864465541e-05, "loss": 0.8976, "step": 8438 }, { "epoch": 2.6, "learning_rate": 1.7363612235439435e-05, "loss": 0.9073, "step": 8439 }, { "epoch": 2.6, "learning_rate": 1.7362937951197316e-05, "loss": 0.8466, "step": 8440 }, { "epoch": 2.6, "learning_rate": 1.736226359383444e-05, "loss": 0.776, "step": 8441 }, { "epoch": 2.6, "learning_rate": 1.7361589163357497e-05, "loss": 0.9031, "step": 8442 }, { "epoch": 2.6, "learning_rate": 1.7360914659773187e-05, "loss": 0.3672, "step": 8443 }, { "epoch": 2.6, "learning_rate": 1.7360240083088213e-05, "loss": 1.0365, "step": 8444 }, { "epoch": 2.6, "learning_rate": 1.7359565433309272e-05, "loss": 0.846, "step": 8445 }, { "epoch": 2.6, "learning_rate": 1.7358890710443064e-05, "loss": 0.7468, "step": 8446 }, { "epoch": 2.6, "learning_rate": 1.735821591449629e-05, "loss": 0.7246, "step": 8447 }, { "epoch": 2.6, "learning_rate": 1.7357541045475647e-05, "loss": 0.7401, "step": 8448 }, { "epoch": 2.6, "learning_rate": 1.7356866103387842e-05, "loss": 1.1303, "step": 8449 }, { "epoch": 2.6, "learning_rate": 1.7356191088239576e-05, "loss": 0.6288, "step": 8450 }, { "epoch": 2.6, "learning_rate": 1.7355516000037555e-05, "loss": 0.6974, "step": 8451 }, { "epoch": 2.6, "learning_rate": 1.735484083878848e-05, "loss": 0.8731, "step": 8452 }, { "epoch": 2.6, "learning_rate": 1.7354165604499057e-05, "loss": 0.9803, "step": 8453 }, { "epoch": 2.6, "learning_rate": 1.735349029717599e-05, "loss": 0.859, "step": 8454 }, { "epoch": 2.6, "learning_rate": 1.7352814916825988e-05, "loss": 0.9302, "step": 8455 }, { "epoch": 2.6, "learning_rate": 1.7352139463455757e-05, "loss": 0.8463, "step": 8456 }, { "epoch": 2.6, "learning_rate": 1.7351463937072008e-05, "loss": 0.83, "step": 8457 }, { "epoch": 2.6, "learning_rate": 1.7350788337681442e-05, "loss": 0.8397, "step": 8458 }, { "epoch": 2.6, "learning_rate": 1.7350112665290778e-05, "loss": 0.8446, "step": 8459 }, { "epoch": 2.6, "learning_rate": 1.7349436919906717e-05, "loss": 0.7234, "step": 8460 }, { "epoch": 2.6, "learning_rate": 1.7348761101535976e-05, "loss": 0.7476, "step": 8461 }, { "epoch": 2.6, "learning_rate": 1.7348085210185263e-05, "loss": 0.8175, "step": 8462 }, { "epoch": 2.6, "learning_rate": 1.7347409245861295e-05, "loss": 0.6752, "step": 8463 }, { "epoch": 2.6, "learning_rate": 1.734673320857078e-05, "loss": 0.8592, "step": 8464 }, { "epoch": 2.6, "learning_rate": 1.7346057098320432e-05, "loss": 0.831, "step": 8465 }, { "epoch": 2.6, "learning_rate": 1.7345380915116968e-05, "loss": 0.7966, "step": 8466 }, { "epoch": 2.61, "learning_rate": 1.73447046589671e-05, "loss": 0.8386, "step": 8467 }, { "epoch": 2.61, "learning_rate": 1.7344028329877546e-05, "loss": 0.9478, "step": 8468 }, { "epoch": 2.61, "learning_rate": 1.7343351927855025e-05, "loss": 0.6966, "step": 8469 }, { "epoch": 2.61, "learning_rate": 1.734267545290625e-05, "loss": 0.7504, "step": 8470 }, { "epoch": 2.61, "learning_rate": 1.7341998905037942e-05, "loss": 0.8312, "step": 8471 }, { "epoch": 2.61, "learning_rate": 1.7341322284256816e-05, "loss": 0.8259, "step": 8472 }, { "epoch": 2.61, "learning_rate": 1.7340645590569593e-05, "loss": 0.8038, "step": 8473 }, { "epoch": 2.61, "learning_rate": 1.7339968823982995e-05, "loss": 0.6859, "step": 8474 }, { "epoch": 2.61, "learning_rate": 1.7339291984503742e-05, "loss": 0.8762, "step": 8475 }, { "epoch": 2.61, "learning_rate": 1.7338615072138555e-05, "loss": 0.6056, "step": 8476 }, { "epoch": 2.61, "learning_rate": 1.733793808689416e-05, "loss": 1.0081, "step": 8477 }, { "epoch": 2.61, "learning_rate": 1.7337261028777274e-05, "loss": 0.8823, "step": 8478 }, { "epoch": 2.61, "learning_rate": 1.733658389779462e-05, "loss": 0.8791, "step": 8479 }, { "epoch": 2.61, "learning_rate": 1.7335906693952936e-05, "loss": 0.8749, "step": 8480 }, { "epoch": 2.61, "learning_rate": 1.7335229417258932e-05, "loss": 0.9423, "step": 8481 }, { "epoch": 2.61, "learning_rate": 1.733455206771934e-05, "loss": 0.6259, "step": 8482 }, { "epoch": 2.61, "learning_rate": 1.7333874645340886e-05, "loss": 0.7685, "step": 8483 }, { "epoch": 2.61, "learning_rate": 1.7333197150130295e-05, "loss": 0.8946, "step": 8484 }, { "epoch": 2.61, "learning_rate": 1.73325195820943e-05, "loss": 0.7678, "step": 8485 }, { "epoch": 2.61, "learning_rate": 1.733184194123963e-05, "loss": 1.0548, "step": 8486 }, { "epoch": 2.61, "learning_rate": 1.7331164227573012e-05, "loss": 0.8125, "step": 8487 }, { "epoch": 2.61, "learning_rate": 1.7330486441101174e-05, "loss": 0.7232, "step": 8488 }, { "epoch": 2.61, "learning_rate": 1.732980858183085e-05, "loss": 1.0652, "step": 8489 }, { "epoch": 2.61, "learning_rate": 1.7329130649768774e-05, "loss": 0.8451, "step": 8490 }, { "epoch": 2.61, "learning_rate": 1.7328452644921674e-05, "loss": 0.7891, "step": 8491 }, { "epoch": 2.61, "learning_rate": 1.7327774567296284e-05, "loss": 0.5293, "step": 8492 }, { "epoch": 2.61, "learning_rate": 1.732709641689934e-05, "loss": 0.9854, "step": 8493 }, { "epoch": 2.61, "learning_rate": 1.7326418193737577e-05, "loss": 1.0144, "step": 8494 }, { "epoch": 2.61, "learning_rate": 1.732573989781773e-05, "loss": 0.6799, "step": 8495 }, { "epoch": 2.61, "learning_rate": 1.7325061529146528e-05, "loss": 0.7215, "step": 8496 }, { "epoch": 2.61, "learning_rate": 1.732438308773072e-05, "loss": 0.8515, "step": 8497 }, { "epoch": 2.61, "learning_rate": 1.732370457357704e-05, "loss": 0.7659, "step": 8498 }, { "epoch": 2.61, "learning_rate": 1.7323025986692216e-05, "loss": 1.3449, "step": 8499 }, { "epoch": 2.62, "learning_rate": 1.7322347327083e-05, "loss": 0.7131, "step": 8500 }, { "epoch": 2.62, "learning_rate": 1.7321668594756123e-05, "loss": 0.9022, "step": 8501 }, { "epoch": 2.62, "learning_rate": 1.7320989789718333e-05, "loss": 0.7305, "step": 8502 }, { "epoch": 2.62, "learning_rate": 1.7320310911976364e-05, "loss": 0.9273, "step": 8503 }, { "epoch": 2.62, "learning_rate": 1.7319631961536966e-05, "loss": 0.791, "step": 8504 }, { "epoch": 2.62, "learning_rate": 1.7318952938406874e-05, "loss": 0.6258, "step": 8505 }, { "epoch": 2.62, "learning_rate": 1.731827384259283e-05, "loss": 0.7476, "step": 8506 }, { "epoch": 2.62, "learning_rate": 1.7317594674101585e-05, "loss": 0.8605, "step": 8507 }, { "epoch": 2.62, "learning_rate": 1.7316915432939882e-05, "loss": 0.8457, "step": 8508 }, { "epoch": 2.62, "learning_rate": 1.7316236119114466e-05, "loss": 0.8485, "step": 8509 }, { "epoch": 2.62, "learning_rate": 1.7315556732632078e-05, "loss": 0.8141, "step": 8510 }, { "epoch": 2.62, "learning_rate": 1.7314877273499474e-05, "loss": 0.9422, "step": 8511 }, { "epoch": 2.62, "learning_rate": 1.7314197741723394e-05, "loss": 1.1321, "step": 8512 }, { "epoch": 2.62, "learning_rate": 1.7313518137310592e-05, "loss": 0.9605, "step": 8513 }, { "epoch": 2.62, "learning_rate": 1.7312838460267812e-05, "loss": 0.9549, "step": 8514 }, { "epoch": 2.62, "learning_rate": 1.7312158710601806e-05, "loss": 0.6622, "step": 8515 }, { "epoch": 2.62, "learning_rate": 1.7311478888319327e-05, "loss": 0.7912, "step": 8516 }, { "epoch": 2.62, "learning_rate": 1.7310798993427124e-05, "loss": 0.6198, "step": 8517 }, { "epoch": 2.62, "learning_rate": 1.7310119025931952e-05, "loss": 0.8292, "step": 8518 }, { "epoch": 2.62, "learning_rate": 1.7309438985840557e-05, "loss": 0.74, "step": 8519 }, { "epoch": 2.62, "learning_rate": 1.7308758873159693e-05, "loss": 0.7227, "step": 8520 }, { "epoch": 2.62, "learning_rate": 1.7308078687896125e-05, "loss": 0.8779, "step": 8521 }, { "epoch": 2.62, "learning_rate": 1.7307398430056595e-05, "loss": 0.708, "step": 8522 }, { "epoch": 2.62, "learning_rate": 1.7306718099647867e-05, "loss": 0.7004, "step": 8523 }, { "epoch": 2.62, "learning_rate": 1.7306037696676692e-05, "loss": 0.9393, "step": 8524 }, { "epoch": 2.62, "learning_rate": 1.730535722114983e-05, "loss": 1.1064, "step": 8525 }, { "epoch": 2.62, "learning_rate": 1.730467667307404e-05, "loss": 0.5407, "step": 8526 }, { "epoch": 2.62, "learning_rate": 1.7303996052456074e-05, "loss": 1.0347, "step": 8527 }, { "epoch": 2.62, "learning_rate": 1.73033153593027e-05, "loss": 0.9485, "step": 8528 }, { "epoch": 2.62, "learning_rate": 1.730263459362067e-05, "loss": 0.6096, "step": 8529 }, { "epoch": 2.62, "learning_rate": 1.7301953755416752e-05, "loss": 0.7434, "step": 8530 }, { "epoch": 2.62, "learning_rate": 1.73012728446977e-05, "loss": 0.9361, "step": 8531 }, { "epoch": 2.63, "learning_rate": 1.7300591861470282e-05, "loss": 0.9205, "step": 8532 }, { "epoch": 2.63, "learning_rate": 1.7299910805741257e-05, "loss": 1.0612, "step": 8533 }, { "epoch": 2.63, "learning_rate": 1.729922967751739e-05, "loss": 0.8884, "step": 8534 }, { "epoch": 2.63, "learning_rate": 1.7298548476805446e-05, "loss": 0.7488, "step": 8535 }, { "epoch": 2.63, "learning_rate": 1.7297867203612188e-05, "loss": 0.7817, "step": 8536 }, { "epoch": 2.63, "learning_rate": 1.7297185857944385e-05, "loss": 0.5603, "step": 8537 }, { "epoch": 2.63, "learning_rate": 1.72965044398088e-05, "loss": 0.8133, "step": 8538 }, { "epoch": 2.63, "learning_rate": 1.7295822949212197e-05, "loss": 0.6743, "step": 8539 }, { "epoch": 2.63, "learning_rate": 1.7295141386161357e-05, "loss": 1.0295, "step": 8540 }, { "epoch": 2.63, "learning_rate": 1.729445975066303e-05, "loss": 0.9873, "step": 8541 }, { "epoch": 2.63, "learning_rate": 1.7293778042724e-05, "loss": 0.6236, "step": 8542 }, { "epoch": 2.63, "learning_rate": 1.729309626235103e-05, "loss": 0.8722, "step": 8543 }, { "epoch": 2.63, "learning_rate": 1.7292414409550895e-05, "loss": 0.5885, "step": 8544 }, { "epoch": 2.63, "learning_rate": 1.7291732484330363e-05, "loss": 0.679, "step": 8545 }, { "epoch": 2.63, "learning_rate": 1.7291050486696208e-05, "loss": 0.5433, "step": 8546 }, { "epoch": 2.63, "learning_rate": 1.72903684166552e-05, "loss": 0.7498, "step": 8547 }, { "epoch": 2.63, "learning_rate": 1.7289686274214116e-05, "loss": 0.8761, "step": 8548 }, { "epoch": 2.63, "learning_rate": 1.728900405937973e-05, "loss": 0.7927, "step": 8549 }, { "epoch": 2.63, "learning_rate": 1.7288321772158815e-05, "loss": 0.935, "step": 8550 }, { "epoch": 2.63, "learning_rate": 1.7287639412558144e-05, "loss": 0.6511, "step": 8551 }, { "epoch": 2.63, "learning_rate": 1.7286956980584505e-05, "loss": 1.1942, "step": 8552 }, { "epoch": 2.63, "learning_rate": 1.728627447624466e-05, "loss": 0.7436, "step": 8553 }, { "epoch": 2.63, "learning_rate": 1.7285591899545397e-05, "loss": 0.8209, "step": 8554 }, { "epoch": 2.63, "learning_rate": 1.728490925049349e-05, "loss": 0.672, "step": 8555 }, { "epoch": 2.63, "learning_rate": 1.7284226529095726e-05, "loss": 0.9284, "step": 8556 }, { "epoch": 2.63, "learning_rate": 1.7283543735358872e-05, "loss": 0.7673, "step": 8557 }, { "epoch": 2.63, "learning_rate": 1.7282860869289718e-05, "loss": 1.1015, "step": 8558 }, { "epoch": 2.63, "learning_rate": 1.728217793089505e-05, "loss": 0.7675, "step": 8559 }, { "epoch": 2.63, "learning_rate": 1.7281494920181632e-05, "loss": 0.8901, "step": 8560 }, { "epoch": 2.63, "learning_rate": 1.7280811837156268e-05, "loss": 0.7879, "step": 8561 }, { "epoch": 2.63, "learning_rate": 1.728012868182573e-05, "loss": 0.8906, "step": 8562 }, { "epoch": 2.63, "learning_rate": 1.7279445454196802e-05, "loss": 0.7332, "step": 8563 }, { "epoch": 2.63, "learning_rate": 1.7278762154276278e-05, "loss": 1.0509, "step": 8564 }, { "epoch": 2.64, "learning_rate": 1.727807878207093e-05, "loss": 0.7951, "step": 8565 }, { "epoch": 2.64, "learning_rate": 1.7277395337587557e-05, "loss": 0.8263, "step": 8566 }, { "epoch": 2.64, "learning_rate": 1.7276711820832938e-05, "loss": 0.8725, "step": 8567 }, { "epoch": 2.64, "learning_rate": 1.7276028231813865e-05, "loss": 0.7283, "step": 8568 }, { "epoch": 2.64, "learning_rate": 1.7275344570537128e-05, "loss": 1.1262, "step": 8569 }, { "epoch": 2.64, "learning_rate": 1.727466083700951e-05, "loss": 0.6098, "step": 8570 }, { "epoch": 2.64, "learning_rate": 1.727397703123781e-05, "loss": 0.7274, "step": 8571 }, { "epoch": 2.64, "learning_rate": 1.7273293153228813e-05, "loss": 0.8424, "step": 8572 }, { "epoch": 2.64, "learning_rate": 1.727260920298931e-05, "loss": 1.2043, "step": 8573 }, { "epoch": 2.64, "learning_rate": 1.7271925180526094e-05, "loss": 0.677, "step": 8574 }, { "epoch": 2.64, "learning_rate": 1.7271241085845963e-05, "loss": 0.8367, "step": 8575 }, { "epoch": 2.64, "learning_rate": 1.7270556918955704e-05, "loss": 0.9337, "step": 8576 }, { "epoch": 2.64, "learning_rate": 1.7269872679862114e-05, "loss": 0.9344, "step": 8577 }, { "epoch": 2.64, "learning_rate": 1.726918836857199e-05, "loss": 0.9193, "step": 8578 }, { "epoch": 2.64, "learning_rate": 1.7268503985092125e-05, "loss": 0.6298, "step": 8579 }, { "epoch": 2.64, "learning_rate": 1.7267819529429317e-05, "loss": 1.0753, "step": 8580 }, { "epoch": 2.64, "learning_rate": 1.7267135001590364e-05, "loss": 0.8896, "step": 8581 }, { "epoch": 2.64, "learning_rate": 1.726645040158206e-05, "loss": 0.8876, "step": 8582 }, { "epoch": 2.64, "learning_rate": 1.7265765729411205e-05, "loss": 0.551, "step": 8583 }, { "epoch": 2.64, "learning_rate": 1.7265080985084603e-05, "loss": 0.7339, "step": 8584 }, { "epoch": 2.64, "learning_rate": 1.7264396168609053e-05, "loss": 0.6512, "step": 8585 }, { "epoch": 2.64, "learning_rate": 1.7263711279991347e-05, "loss": 0.6107, "step": 8586 }, { "epoch": 2.64, "learning_rate": 1.72630263192383e-05, "loss": 0.7668, "step": 8587 }, { "epoch": 2.64, "learning_rate": 1.7262341286356708e-05, "loss": 0.6156, "step": 8588 }, { "epoch": 2.64, "learning_rate": 1.7261656181353373e-05, "loss": 0.7501, "step": 8589 }, { "epoch": 2.64, "learning_rate": 1.7260971004235097e-05, "loss": 0.9267, "step": 8590 }, { "epoch": 2.64, "learning_rate": 1.7260285755008686e-05, "loss": 0.9665, "step": 8591 }, { "epoch": 2.64, "learning_rate": 1.7259600433680954e-05, "loss": 0.7435, "step": 8592 }, { "epoch": 2.64, "learning_rate": 1.725891504025869e-05, "loss": 0.7005, "step": 8593 }, { "epoch": 2.64, "learning_rate": 1.7258229574748714e-05, "loss": 0.7592, "step": 8594 }, { "epoch": 2.64, "learning_rate": 1.725754403715783e-05, "loss": 1.0113, "step": 8595 }, { "epoch": 2.64, "learning_rate": 1.7256858427492842e-05, "loss": 0.4641, "step": 8596 }, { "epoch": 2.65, "learning_rate": 1.7256172745760566e-05, "loss": 0.3853, "step": 8597 }, { "epoch": 2.65, "learning_rate": 1.72554869919678e-05, "loss": 0.7718, "step": 8598 }, { "epoch": 2.65, "learning_rate": 1.7254801166121368e-05, "loss": 0.5467, "step": 8599 }, { "epoch": 2.65, "learning_rate": 1.7254115268228073e-05, "loss": 0.959, "step": 8600 }, { "epoch": 2.65, "learning_rate": 1.7253429298294723e-05, "loss": 0.8339, "step": 8601 }, { "epoch": 2.65, "learning_rate": 1.725274325632814e-05, "loss": 0.6731, "step": 8602 }, { "epoch": 2.65, "learning_rate": 1.725205714233513e-05, "loss": 0.8825, "step": 8603 }, { "epoch": 2.65, "learning_rate": 1.725137095632251e-05, "loss": 0.9953, "step": 8604 }, { "epoch": 2.65, "learning_rate": 1.7250684698297093e-05, "loss": 0.7978, "step": 8605 }, { "epoch": 2.65, "learning_rate": 1.7249998368265694e-05, "loss": 0.7133, "step": 8606 }, { "epoch": 2.65, "learning_rate": 1.724931196623513e-05, "loss": 0.8725, "step": 8607 }, { "epoch": 2.65, "learning_rate": 1.724862549221222e-05, "loss": 0.6917, "step": 8608 }, { "epoch": 2.65, "learning_rate": 1.7247938946203773e-05, "loss": 0.7122, "step": 8609 }, { "epoch": 2.65, "learning_rate": 1.7247252328216616e-05, "loss": 0.6706, "step": 8610 }, { "epoch": 2.65, "learning_rate": 1.724656563825756e-05, "loss": 1.1412, "step": 8611 }, { "epoch": 2.65, "learning_rate": 1.724587887633343e-05, "loss": 0.6224, "step": 8612 }, { "epoch": 2.65, "learning_rate": 1.724519204245105e-05, "loss": 0.7482, "step": 8613 }, { "epoch": 2.65, "learning_rate": 1.724450513661723e-05, "loss": 0.8867, "step": 8614 }, { "epoch": 2.65, "learning_rate": 1.7243818158838798e-05, "loss": 0.7849, "step": 8615 }, { "epoch": 2.65, "learning_rate": 1.7243131109122576e-05, "loss": 0.6201, "step": 8616 }, { "epoch": 2.65, "learning_rate": 1.7242443987475388e-05, "loss": 0.849, "step": 8617 }, { "epoch": 2.65, "learning_rate": 1.7241756793904057e-05, "loss": 0.7573, "step": 8618 }, { "epoch": 2.65, "learning_rate": 1.7241069528415404e-05, "loss": 0.7148, "step": 8619 }, { "epoch": 2.65, "learning_rate": 1.724038219101626e-05, "loss": 0.5475, "step": 8620 }, { "epoch": 2.65, "learning_rate": 1.7239694781713446e-05, "loss": 0.9454, "step": 8621 }, { "epoch": 2.65, "learning_rate": 1.723900730051379e-05, "loss": 0.7476, "step": 8622 }, { "epoch": 2.65, "learning_rate": 1.7238319747424118e-05, "loss": 0.9814, "step": 8623 }, { "epoch": 2.65, "learning_rate": 1.7237632122451264e-05, "loss": 0.723, "step": 8624 }, { "epoch": 2.65, "learning_rate": 1.723694442560205e-05, "loss": 0.8637, "step": 8625 }, { "epoch": 2.65, "learning_rate": 1.723625665688331e-05, "loss": 0.8181, "step": 8626 }, { "epoch": 2.65, "learning_rate": 1.7235568816301874e-05, "loss": 0.8238, "step": 8627 }, { "epoch": 2.65, "learning_rate": 1.7234880903864567e-05, "loss": 0.792, "step": 8628 }, { "epoch": 2.65, "learning_rate": 1.723419291957823e-05, "loss": 0.6537, "step": 8629 }, { "epoch": 2.66, "learning_rate": 1.7233504863449684e-05, "loss": 0.7184, "step": 8630 }, { "epoch": 2.66, "learning_rate": 1.723281673548577e-05, "loss": 0.8232, "step": 8631 }, { "epoch": 2.66, "learning_rate": 1.723212853569332e-05, "loss": 0.9481, "step": 8632 }, { "epoch": 2.66, "learning_rate": 1.723144026407917e-05, "loss": 0.8333, "step": 8633 }, { "epoch": 2.66, "learning_rate": 1.7230751920650153e-05, "loss": 0.5423, "step": 8634 }, { "epoch": 2.66, "learning_rate": 1.7230063505413103e-05, "loss": 0.8457, "step": 8635 }, { "epoch": 2.66, "learning_rate": 1.722937501837486e-05, "loss": 0.831, "step": 8636 }, { "epoch": 2.66, "learning_rate": 1.7228686459542264e-05, "loss": 0.9122, "step": 8637 }, { "epoch": 2.66, "learning_rate": 1.7227997828922146e-05, "loss": 0.6981, "step": 8638 }, { "epoch": 2.66, "learning_rate": 1.7227309126521347e-05, "loss": 0.9402, "step": 8639 }, { "epoch": 2.66, "learning_rate": 1.722662035234671e-05, "loss": 1.0646, "step": 8640 }, { "epoch": 2.66, "learning_rate": 1.722593150640507e-05, "loss": 0.8095, "step": 8641 }, { "epoch": 2.66, "learning_rate": 1.722524258870327e-05, "loss": 0.88, "step": 8642 }, { "epoch": 2.66, "learning_rate": 1.7224553599248155e-05, "loss": 0.553, "step": 8643 }, { "epoch": 2.66, "learning_rate": 1.7223864538046564e-05, "loss": 0.7039, "step": 8644 }, { "epoch": 2.66, "learning_rate": 1.722317540510534e-05, "loss": 0.94, "step": 8645 }, { "epoch": 2.66, "learning_rate": 1.7222486200431328e-05, "loss": 0.6247, "step": 8646 }, { "epoch": 2.66, "learning_rate": 1.722179692403137e-05, "loss": 1.0105, "step": 8647 }, { "epoch": 2.66, "learning_rate": 1.7221107575912313e-05, "loss": 0.7928, "step": 8648 }, { "epoch": 2.66, "learning_rate": 1.7220418156081004e-05, "loss": 0.7696, "step": 8649 }, { "epoch": 2.66, "learning_rate": 1.7219728664544287e-05, "loss": 0.8354, "step": 8650 }, { "epoch": 2.66, "learning_rate": 1.721903910130901e-05, "loss": 0.8236, "step": 8651 }, { "epoch": 2.66, "learning_rate": 1.7218349466382024e-05, "loss": 0.8104, "step": 8652 }, { "epoch": 2.66, "learning_rate": 1.721765975977017e-05, "loss": 0.8202, "step": 8653 }, { "epoch": 2.66, "learning_rate": 1.7216969981480308e-05, "loss": 1.0819, "step": 8654 }, { "epoch": 2.66, "learning_rate": 1.7216280131519282e-05, "loss": 0.7328, "step": 8655 }, { "epoch": 2.66, "learning_rate": 1.721559020989394e-05, "loss": 1.1927, "step": 8656 }, { "epoch": 2.66, "learning_rate": 1.721490021661114e-05, "loss": 0.7417, "step": 8657 }, { "epoch": 2.66, "learning_rate": 1.7214210151677732e-05, "loss": 0.9235, "step": 8658 }, { "epoch": 2.66, "learning_rate": 1.7213520015100566e-05, "loss": 1.1134, "step": 8659 }, { "epoch": 2.66, "learning_rate": 1.72128298068865e-05, "loss": 0.8102, "step": 8660 }, { "epoch": 2.66, "learning_rate": 1.7212139527042383e-05, "loss": 0.7422, "step": 8661 }, { "epoch": 2.67, "learning_rate": 1.7211449175575076e-05, "loss": 0.655, "step": 8662 }, { "epoch": 2.67, "learning_rate": 1.7210758752491436e-05, "loss": 1.0748, "step": 8663 }, { "epoch": 2.67, "learning_rate": 1.721006825779831e-05, "loss": 1.0352, "step": 8664 }, { "epoch": 2.67, "learning_rate": 1.7209377691502565e-05, "loss": 0.8815, "step": 8665 }, { "epoch": 2.67, "learning_rate": 1.7208687053611054e-05, "loss": 0.7021, "step": 8666 }, { "epoch": 2.67, "learning_rate": 1.7207996344130636e-05, "loss": 0.8305, "step": 8667 }, { "epoch": 2.67, "learning_rate": 1.720730556306817e-05, "loss": 0.9937, "step": 8668 }, { "epoch": 2.67, "learning_rate": 1.720661471043052e-05, "loss": 0.8586, "step": 8669 }, { "epoch": 2.67, "learning_rate": 1.7205923786224544e-05, "loss": 1.0164, "step": 8670 }, { "epoch": 2.67, "learning_rate": 1.7205232790457103e-05, "loss": 0.5488, "step": 8671 }, { "epoch": 2.67, "learning_rate": 1.720454172313506e-05, "loss": 0.6398, "step": 8672 }, { "epoch": 2.67, "learning_rate": 1.7203850584265278e-05, "loss": 0.6939, "step": 8673 }, { "epoch": 2.67, "learning_rate": 1.720315937385462e-05, "loss": 0.8974, "step": 8674 }, { "epoch": 2.67, "learning_rate": 1.7202468091909952e-05, "loss": 0.8592, "step": 8675 }, { "epoch": 2.67, "learning_rate": 1.7201776738438136e-05, "loss": 0.6622, "step": 8676 }, { "epoch": 2.67, "learning_rate": 1.7201085313446042e-05, "loss": 0.7855, "step": 8677 }, { "epoch": 2.67, "learning_rate": 1.720039381694053e-05, "loss": 0.6654, "step": 8678 }, { "epoch": 2.67, "learning_rate": 1.7199702248928475e-05, "loss": 0.8228, "step": 8679 }, { "epoch": 2.67, "learning_rate": 1.719901060941674e-05, "loss": 0.7872, "step": 8680 }, { "epoch": 2.67, "learning_rate": 1.7198318898412194e-05, "loss": 0.5817, "step": 8681 }, { "epoch": 2.67, "learning_rate": 1.719762711592171e-05, "loss": 0.8327, "step": 8682 }, { "epoch": 2.67, "learning_rate": 1.7196935261952153e-05, "loss": 0.7584, "step": 8683 }, { "epoch": 2.67, "learning_rate": 1.7196243336510395e-05, "loss": 0.8328, "step": 8684 }, { "epoch": 2.67, "learning_rate": 1.7195551339603313e-05, "loss": 0.7475, "step": 8685 }, { "epoch": 2.67, "learning_rate": 1.7194859271237766e-05, "loss": 0.8522, "step": 8686 }, { "epoch": 2.67, "learning_rate": 1.7194167131420643e-05, "loss": 0.9163, "step": 8687 }, { "epoch": 2.67, "learning_rate": 1.719347492015881e-05, "loss": 0.7993, "step": 8688 }, { "epoch": 2.67, "learning_rate": 1.7192782637459137e-05, "loss": 1.6883, "step": 8689 }, { "epoch": 2.67, "learning_rate": 1.7192090283328506e-05, "loss": 0.9495, "step": 8690 }, { "epoch": 2.67, "learning_rate": 1.7191397857773787e-05, "loss": 0.7707, "step": 8691 }, { "epoch": 2.67, "learning_rate": 1.7190705360801864e-05, "loss": 0.9057, "step": 8692 }, { "epoch": 2.67, "learning_rate": 1.719001279241961e-05, "loss": 0.7805, "step": 8693 }, { "epoch": 2.67, "learning_rate": 1.71893201526339e-05, "loss": 0.6844, "step": 8694 }, { "epoch": 2.68, "learning_rate": 1.7188627441451614e-05, "loss": 0.9581, "step": 8695 }, { "epoch": 2.68, "learning_rate": 1.7187934658879633e-05, "loss": 0.4335, "step": 8696 }, { "epoch": 2.68, "learning_rate": 1.7187241804924838e-05, "loss": 0.6572, "step": 8697 }, { "epoch": 2.68, "learning_rate": 1.7186548879594103e-05, "loss": 0.8307, "step": 8698 }, { "epoch": 2.68, "learning_rate": 1.718585588289432e-05, "loss": 0.837, "step": 8699 }, { "epoch": 2.68, "learning_rate": 1.7185162814832362e-05, "loss": 0.9601, "step": 8700 }, { "epoch": 2.68, "learning_rate": 1.7184469675415117e-05, "loss": 0.6154, "step": 8701 }, { "epoch": 2.68, "learning_rate": 1.7183776464649465e-05, "loss": 0.9522, "step": 8702 }, { "epoch": 2.68, "learning_rate": 1.7183083182542293e-05, "loss": 0.9835, "step": 8703 }, { "epoch": 2.68, "learning_rate": 1.7182389829100484e-05, "loss": 1.1341, "step": 8704 }, { "epoch": 2.68, "learning_rate": 1.7181696404330928e-05, "loss": 1.0043, "step": 8705 }, { "epoch": 2.68, "learning_rate": 1.7181002908240506e-05, "loss": 1.0203, "step": 8706 }, { "epoch": 2.68, "learning_rate": 1.7180309340836107e-05, "loss": 0.9074, "step": 8707 }, { "epoch": 2.68, "learning_rate": 1.7179615702124616e-05, "loss": 0.8318, "step": 8708 }, { "epoch": 2.68, "learning_rate": 1.7178921992112923e-05, "loss": 0.5811, "step": 8709 }, { "epoch": 2.68, "learning_rate": 1.7178228210807924e-05, "loss": 0.7771, "step": 8710 }, { "epoch": 2.68, "learning_rate": 1.71775343582165e-05, "loss": 0.6365, "step": 8711 }, { "epoch": 2.68, "learning_rate": 1.7176840434345547e-05, "loss": 0.6178, "step": 8712 }, { "epoch": 2.68, "learning_rate": 1.7176146439201954e-05, "loss": 0.7746, "step": 8713 }, { "epoch": 2.68, "learning_rate": 1.717545237279261e-05, "loss": 0.9224, "step": 8714 }, { "epoch": 2.68, "learning_rate": 1.7174758235124412e-05, "loss": 0.7987, "step": 8715 }, { "epoch": 2.68, "learning_rate": 1.7174064026204252e-05, "loss": 0.7802, "step": 8716 }, { "epoch": 2.68, "learning_rate": 1.7173369746039026e-05, "loss": 0.8567, "step": 8717 }, { "epoch": 2.68, "learning_rate": 1.7172675394635626e-05, "loss": 0.7919, "step": 8718 }, { "epoch": 2.68, "learning_rate": 1.717198097200095e-05, "loss": 0.8942, "step": 8719 }, { "epoch": 2.68, "learning_rate": 1.7171286478141894e-05, "loss": 0.8773, "step": 8720 }, { "epoch": 2.68, "learning_rate": 1.7170591913065352e-05, "loss": 0.8205, "step": 8721 }, { "epoch": 2.68, "learning_rate": 1.7169897276778225e-05, "loss": 0.8916, "step": 8722 }, { "epoch": 2.68, "learning_rate": 1.7169202569287405e-05, "loss": 0.737, "step": 8723 }, { "epoch": 2.68, "learning_rate": 1.7168507790599803e-05, "loss": 1.0483, "step": 8724 }, { "epoch": 2.68, "learning_rate": 1.7167812940722307e-05, "loss": 0.8521, "step": 8725 }, { "epoch": 2.68, "learning_rate": 1.7167118019661826e-05, "loss": 0.777, "step": 8726 }, { "epoch": 2.69, "learning_rate": 1.7166423027425255e-05, "loss": 0.9282, "step": 8727 }, { "epoch": 2.69, "learning_rate": 1.71657279640195e-05, "loss": 0.8428, "step": 8728 }, { "epoch": 2.69, "learning_rate": 1.7165032829451464e-05, "loss": 0.7376, "step": 8729 }, { "epoch": 2.69, "learning_rate": 1.7164337623728044e-05, "loss": 0.693, "step": 8730 }, { "epoch": 2.69, "learning_rate": 1.7163642346856154e-05, "loss": 0.8861, "step": 8731 }, { "epoch": 2.69, "learning_rate": 1.716294699884269e-05, "loss": 0.9972, "step": 8732 }, { "epoch": 2.69, "learning_rate": 1.716225157969456e-05, "loss": 0.8707, "step": 8733 }, { "epoch": 2.69, "learning_rate": 1.7161556089418674e-05, "loss": 0.9301, "step": 8734 }, { "epoch": 2.69, "learning_rate": 1.7160860528021932e-05, "loss": 0.9522, "step": 8735 }, { "epoch": 2.69, "learning_rate": 1.716016489551125e-05, "loss": 0.9295, "step": 8736 }, { "epoch": 2.69, "learning_rate": 1.7159469191893524e-05, "loss": 1.1258, "step": 8737 }, { "epoch": 2.69, "learning_rate": 1.715877341717568e-05, "loss": 0.657, "step": 8738 }, { "epoch": 2.69, "learning_rate": 1.7158077571364612e-05, "loss": 1.0132, "step": 8739 }, { "epoch": 2.69, "learning_rate": 1.7157381654467238e-05, "loss": 0.4978, "step": 8740 }, { "epoch": 2.69, "learning_rate": 1.7156685666490467e-05, "loss": 0.8978, "step": 8741 }, { "epoch": 2.69, "learning_rate": 1.715598960744121e-05, "loss": 0.6487, "step": 8742 }, { "epoch": 2.69, "learning_rate": 1.7155293477326385e-05, "loss": 1.0065, "step": 8743 }, { "epoch": 2.69, "learning_rate": 1.71545972761529e-05, "loss": 0.8274, "step": 8744 }, { "epoch": 2.69, "learning_rate": 1.715390100392767e-05, "loss": 0.8102, "step": 8745 }, { "epoch": 2.69, "learning_rate": 1.715320466065761e-05, "loss": 0.7354, "step": 8746 }, { "epoch": 2.69, "learning_rate": 1.7152508246349637e-05, "loss": 1.0778, "step": 8747 }, { "epoch": 2.69, "learning_rate": 1.715181176101066e-05, "loss": 0.7602, "step": 8748 }, { "epoch": 2.69, "learning_rate": 1.7151115204647605e-05, "loss": 0.8913, "step": 8749 }, { "epoch": 2.69, "learning_rate": 1.7150418577267384e-05, "loss": 1.0467, "step": 8750 }, { "epoch": 2.69, "learning_rate": 1.7149721878876917e-05, "loss": 0.8004, "step": 8751 }, { "epoch": 2.69, "learning_rate": 1.714902510948312e-05, "loss": 0.8138, "step": 8752 }, { "epoch": 2.69, "learning_rate": 1.7148328269092918e-05, "loss": 0.7586, "step": 8753 }, { "epoch": 2.69, "learning_rate": 1.7147631357713228e-05, "loss": 0.9633, "step": 8754 }, { "epoch": 2.69, "learning_rate": 1.714693437535097e-05, "loss": 0.9526, "step": 8755 }, { "epoch": 2.69, "learning_rate": 1.714623732201307e-05, "loss": 0.9895, "step": 8756 }, { "epoch": 2.69, "learning_rate": 1.714554019770644e-05, "loss": 0.8237, "step": 8757 }, { "epoch": 2.69, "learning_rate": 1.7144843002438015e-05, "loss": 0.8843, "step": 8758 }, { "epoch": 2.69, "learning_rate": 1.7144145736214716e-05, "loss": 0.6849, "step": 8759 }, { "epoch": 2.7, "learning_rate": 1.7143448399043463e-05, "loss": 0.7361, "step": 8760 }, { "epoch": 2.7, "learning_rate": 1.7142750990931185e-05, "loss": 1.1667, "step": 8761 }, { "epoch": 2.7, "learning_rate": 1.7142053511884807e-05, "loss": 0.9755, "step": 8762 }, { "epoch": 2.7, "learning_rate": 1.7141355961911254e-05, "loss": 0.519, "step": 8763 }, { "epoch": 2.7, "learning_rate": 1.7140658341017452e-05, "loss": 0.795, "step": 8764 }, { "epoch": 2.7, "learning_rate": 1.713996064921034e-05, "loss": 1.0434, "step": 8765 }, { "epoch": 2.7, "learning_rate": 1.713926288649683e-05, "loss": 0.892, "step": 8766 }, { "epoch": 2.7, "learning_rate": 1.713856505288386e-05, "loss": 0.7461, "step": 8767 }, { "epoch": 2.7, "learning_rate": 1.7137867148378363e-05, "loss": 0.8632, "step": 8768 }, { "epoch": 2.7, "learning_rate": 1.713716917298727e-05, "loss": 1.0001, "step": 8769 }, { "epoch": 2.7, "learning_rate": 1.7136471126717506e-05, "loss": 0.7926, "step": 8770 }, { "epoch": 2.7, "learning_rate": 1.7135773009576005e-05, "loss": 0.9394, "step": 8771 }, { "epoch": 2.7, "learning_rate": 1.7135074821569704e-05, "loss": 0.8962, "step": 8772 }, { "epoch": 2.7, "learning_rate": 1.713437656270553e-05, "loss": 0.6757, "step": 8773 }, { "epoch": 2.7, "learning_rate": 1.7133678232990427e-05, "loss": 0.7156, "step": 8774 }, { "epoch": 2.7, "learning_rate": 1.713297983243132e-05, "loss": 0.7615, "step": 8775 }, { "epoch": 2.7, "learning_rate": 1.7132281361035153e-05, "loss": 0.8933, "step": 8776 }, { "epoch": 2.7, "learning_rate": 1.7131582818808855e-05, "loss": 0.6721, "step": 8777 }, { "epoch": 2.7, "learning_rate": 1.7130884205759372e-05, "loss": 0.7901, "step": 8778 }, { "epoch": 2.7, "learning_rate": 1.713018552189363e-05, "loss": 0.6869, "step": 8779 }, { "epoch": 2.7, "learning_rate": 1.7129486767218582e-05, "loss": 0.8247, "step": 8780 }, { "epoch": 2.7, "learning_rate": 1.7128787941741153e-05, "loss": 0.9474, "step": 8781 }, { "epoch": 2.7, "learning_rate": 1.7128089045468294e-05, "loss": 0.924, "step": 8782 }, { "epoch": 2.7, "learning_rate": 1.712739007840694e-05, "loss": 0.8539, "step": 8783 }, { "epoch": 2.7, "learning_rate": 1.7126691040564034e-05, "loss": 0.9086, "step": 8784 }, { "epoch": 2.7, "learning_rate": 1.7125991931946517e-05, "loss": 0.7896, "step": 8785 }, { "epoch": 2.7, "learning_rate": 1.7125292752561335e-05, "loss": 0.9508, "step": 8786 }, { "epoch": 2.7, "learning_rate": 1.7124593502415424e-05, "loss": 0.6837, "step": 8787 }, { "epoch": 2.7, "learning_rate": 1.712389418151574e-05, "loss": 0.8878, "step": 8788 }, { "epoch": 2.7, "learning_rate": 1.7123194789869217e-05, "loss": 0.8426, "step": 8789 }, { "epoch": 2.7, "learning_rate": 1.7122495327482806e-05, "loss": 0.74, "step": 8790 }, { "epoch": 2.7, "learning_rate": 1.7121795794363448e-05, "loss": 0.797, "step": 8791 }, { "epoch": 2.71, "learning_rate": 1.71210961905181e-05, "loss": 0.8286, "step": 8792 }, { "epoch": 2.71, "learning_rate": 1.71203965159537e-05, "loss": 1.0141, "step": 8793 }, { "epoch": 2.71, "learning_rate": 1.7119696770677203e-05, "loss": 0.7954, "step": 8794 }, { "epoch": 2.71, "learning_rate": 1.7118996954695553e-05, "loss": 0.7998, "step": 8795 }, { "epoch": 2.71, "learning_rate": 1.7118297068015706e-05, "loss": 0.8669, "step": 8796 }, { "epoch": 2.71, "learning_rate": 1.7117597110644606e-05, "loss": 0.8478, "step": 8797 }, { "epoch": 2.71, "learning_rate": 1.7116897082589206e-05, "loss": 0.8143, "step": 8798 }, { "epoch": 2.71, "learning_rate": 1.711619698385646e-05, "loss": 0.6492, "step": 8799 }, { "epoch": 2.71, "learning_rate": 1.7115496814453317e-05, "loss": 0.4813, "step": 8800 }, { "epoch": 2.71, "learning_rate": 1.7114796574386736e-05, "loss": 0.7261, "step": 8801 }, { "epoch": 2.71, "learning_rate": 1.7114096263663665e-05, "loss": 0.7256, "step": 8802 }, { "epoch": 2.71, "learning_rate": 1.7113395882291067e-05, "loss": 0.8026, "step": 8803 }, { "epoch": 2.71, "learning_rate": 1.7112695430275887e-05, "loss": 0.8759, "step": 8804 }, { "epoch": 2.71, "learning_rate": 1.711199490762509e-05, "loss": 0.9291, "step": 8805 }, { "epoch": 2.71, "learning_rate": 1.7111294314345626e-05, "loss": 0.6377, "step": 8806 }, { "epoch": 2.71, "learning_rate": 1.7110593650444457e-05, "loss": 0.7064, "step": 8807 }, { "epoch": 2.71, "learning_rate": 1.7109892915928535e-05, "loss": 0.7972, "step": 8808 }, { "epoch": 2.71, "learning_rate": 1.710919211080483e-05, "loss": 0.7165, "step": 8809 }, { "epoch": 2.71, "learning_rate": 1.7108491235080292e-05, "loss": 0.9105, "step": 8810 }, { "epoch": 2.71, "learning_rate": 1.7107790288761885e-05, "loss": 0.7018, "step": 8811 }, { "epoch": 2.71, "learning_rate": 1.710708927185657e-05, "loss": 0.7161, "step": 8812 }, { "epoch": 2.71, "learning_rate": 1.710638818437131e-05, "loss": 0.9839, "step": 8813 }, { "epoch": 2.71, "learning_rate": 1.7105687026313064e-05, "loss": 0.8302, "step": 8814 }, { "epoch": 2.71, "learning_rate": 1.7104985797688794e-05, "loss": 0.7343, "step": 8815 }, { "epoch": 2.71, "learning_rate": 1.7104284498505474e-05, "loss": 0.7688, "step": 8816 }, { "epoch": 2.71, "learning_rate": 1.7103583128770055e-05, "loss": 1.039, "step": 8817 }, { "epoch": 2.71, "learning_rate": 1.710288168848951e-05, "loss": 0.8862, "step": 8818 }, { "epoch": 2.71, "learning_rate": 1.7102180177670808e-05, "loss": 0.5955, "step": 8819 }, { "epoch": 2.71, "learning_rate": 1.7101478596320907e-05, "loss": 0.6382, "step": 8820 }, { "epoch": 2.71, "learning_rate": 1.7100776944446783e-05, "loss": 0.9495, "step": 8821 }, { "epoch": 2.71, "learning_rate": 1.7100075222055394e-05, "loss": 0.9651, "step": 8822 }, { "epoch": 2.71, "learning_rate": 1.709937342915372e-05, "loss": 0.8888, "step": 8823 }, { "epoch": 2.71, "learning_rate": 1.7098671565748718e-05, "loss": 0.6814, "step": 8824 }, { "epoch": 2.72, "learning_rate": 1.7097969631847374e-05, "loss": 0.8102, "step": 8825 }, { "epoch": 2.72, "learning_rate": 1.7097267627456643e-05, "loss": 0.8292, "step": 8826 }, { "epoch": 2.72, "learning_rate": 1.7096565552583504e-05, "loss": 0.9026, "step": 8827 }, { "epoch": 2.72, "learning_rate": 1.709586340723493e-05, "loss": 0.6712, "step": 8828 }, { "epoch": 2.72, "learning_rate": 1.709516119141789e-05, "loss": 0.6329, "step": 8829 }, { "epoch": 2.72, "learning_rate": 1.7094458905139366e-05, "loss": 0.706, "step": 8830 }, { "epoch": 2.72, "learning_rate": 1.7093756548406323e-05, "loss": 0.9904, "step": 8831 }, { "epoch": 2.72, "learning_rate": 1.709305412122574e-05, "loss": 0.9315, "step": 8832 }, { "epoch": 2.72, "learning_rate": 1.7092351623604594e-05, "loss": 1.0057, "step": 8833 }, { "epoch": 2.72, "learning_rate": 1.709164905554986e-05, "loss": 0.9634, "step": 8834 }, { "epoch": 2.72, "learning_rate": 1.709094641706851e-05, "loss": 0.8759, "step": 8835 }, { "epoch": 2.72, "learning_rate": 1.7090243708167532e-05, "loss": 0.9066, "step": 8836 }, { "epoch": 2.72, "learning_rate": 1.70895409288539e-05, "loss": 0.6533, "step": 8837 }, { "epoch": 2.72, "learning_rate": 1.7088838079134587e-05, "loss": 0.7209, "step": 8838 }, { "epoch": 2.72, "learning_rate": 1.7088135159016584e-05, "loss": 0.8073, "step": 8839 }, { "epoch": 2.72, "learning_rate": 1.7087432168506864e-05, "loss": 0.9995, "step": 8840 }, { "epoch": 2.72, "learning_rate": 1.7086729107612408e-05, "loss": 0.8971, "step": 8841 }, { "epoch": 2.72, "learning_rate": 1.7086025976340205e-05, "loss": 0.7119, "step": 8842 }, { "epoch": 2.72, "learning_rate": 1.7085322774697232e-05, "loss": 0.9569, "step": 8843 }, { "epoch": 2.72, "learning_rate": 1.7084619502690474e-05, "loss": 0.8882, "step": 8844 }, { "epoch": 2.72, "learning_rate": 1.7083916160326916e-05, "loss": 0.8107, "step": 8845 }, { "epoch": 2.72, "learning_rate": 1.708321274761354e-05, "loss": 0.8572, "step": 8846 }, { "epoch": 2.72, "learning_rate": 1.7082509264557333e-05, "loss": 0.8625, "step": 8847 }, { "epoch": 2.72, "learning_rate": 1.708180571116528e-05, "loss": 0.932, "step": 8848 }, { "epoch": 2.72, "learning_rate": 1.7081102087444375e-05, "loss": 0.8906, "step": 8849 }, { "epoch": 2.72, "learning_rate": 1.7080398393401597e-05, "loss": 0.9905, "step": 8850 }, { "epoch": 2.72, "learning_rate": 1.7079694629043938e-05, "loss": 0.5101, "step": 8851 }, { "epoch": 2.72, "learning_rate": 1.7078990794378384e-05, "loss": 0.5861, "step": 8852 }, { "epoch": 2.72, "learning_rate": 1.707828688941193e-05, "loss": 0.6447, "step": 8853 }, { "epoch": 2.72, "learning_rate": 1.707758291415156e-05, "loss": 0.7293, "step": 8854 }, { "epoch": 2.72, "learning_rate": 1.7076878868604273e-05, "loss": 0.9363, "step": 8855 }, { "epoch": 2.72, "learning_rate": 1.707617475277706e-05, "loss": 0.8751, "step": 8856 }, { "epoch": 2.73, "learning_rate": 1.70754705666769e-05, "loss": 0.8334, "step": 8857 }, { "epoch": 2.73, "learning_rate": 1.7074766310310803e-05, "loss": 0.8518, "step": 8858 }, { "epoch": 2.73, "learning_rate": 1.7074061983685758e-05, "loss": 0.8508, "step": 8859 }, { "epoch": 2.73, "learning_rate": 1.7073357586808753e-05, "loss": 0.8256, "step": 8860 }, { "epoch": 2.73, "learning_rate": 1.707265311968679e-05, "loss": 0.8998, "step": 8861 }, { "epoch": 2.73, "learning_rate": 1.7071948582326863e-05, "loss": 0.9152, "step": 8862 }, { "epoch": 2.73, "learning_rate": 1.707124397473597e-05, "loss": 0.9764, "step": 8863 }, { "epoch": 2.73, "learning_rate": 1.7070539296921103e-05, "loss": 0.9261, "step": 8864 }, { "epoch": 2.73, "learning_rate": 1.706983454888927e-05, "loss": 0.7119, "step": 8865 }, { "epoch": 2.73, "learning_rate": 1.706912973064746e-05, "loss": 0.6902, "step": 8866 }, { "epoch": 2.73, "learning_rate": 1.706842484220268e-05, "loss": 0.963, "step": 8867 }, { "epoch": 2.73, "learning_rate": 1.7067719883561925e-05, "loss": 1.0071, "step": 8868 }, { "epoch": 2.73, "learning_rate": 1.7067014854732198e-05, "loss": 0.3934, "step": 8869 }, { "epoch": 2.73, "learning_rate": 1.70663097557205e-05, "loss": 0.9161, "step": 8870 }, { "epoch": 2.73, "learning_rate": 1.7065604586533835e-05, "loss": 0.6135, "step": 8871 }, { "epoch": 2.73, "learning_rate": 1.7064899347179205e-05, "loss": 1.0176, "step": 8872 }, { "epoch": 2.73, "learning_rate": 1.706419403766361e-05, "loss": 0.7448, "step": 8873 }, { "epoch": 2.73, "learning_rate": 1.706348865799406e-05, "loss": 0.7421, "step": 8874 }, { "epoch": 2.73, "learning_rate": 1.7062783208177556e-05, "loss": 0.8166, "step": 8875 }, { "epoch": 2.73, "learning_rate": 1.706207768822111e-05, "loss": 1.2302, "step": 8876 }, { "epoch": 2.73, "learning_rate": 1.706137209813172e-05, "loss": 0.9821, "step": 8877 }, { "epoch": 2.73, "learning_rate": 1.7060666437916394e-05, "loss": 1.1562, "step": 8878 }, { "epoch": 2.73, "learning_rate": 1.705996070758215e-05, "loss": 0.7522, "step": 8879 }, { "epoch": 2.73, "learning_rate": 1.7059254907135988e-05, "loss": 0.7915, "step": 8880 }, { "epoch": 2.73, "learning_rate": 1.705854903658492e-05, "loss": 1.0789, "step": 8881 }, { "epoch": 2.73, "learning_rate": 1.705784309593595e-05, "loss": 0.8847, "step": 8882 }, { "epoch": 2.73, "learning_rate": 1.7057137085196093e-05, "loss": 1.1052, "step": 8883 }, { "epoch": 2.73, "learning_rate": 1.7056431004372363e-05, "loss": 0.7641, "step": 8884 }, { "epoch": 2.73, "learning_rate": 1.705572485347177e-05, "loss": 0.7888, "step": 8885 }, { "epoch": 2.73, "learning_rate": 1.7055018632501326e-05, "loss": 0.8418, "step": 8886 }, { "epoch": 2.73, "learning_rate": 1.7054312341468044e-05, "loss": 0.6932, "step": 8887 }, { "epoch": 2.73, "learning_rate": 1.705360598037894e-05, "loss": 0.5672, "step": 8888 }, { "epoch": 2.73, "learning_rate": 1.7052899549241026e-05, "loss": 0.9774, "step": 8889 }, { "epoch": 2.74, "learning_rate": 1.7052193048061324e-05, "loss": 0.9314, "step": 8890 }, { "epoch": 2.74, "learning_rate": 1.7051486476846843e-05, "loss": 0.8331, "step": 8891 }, { "epoch": 2.74, "learning_rate": 1.70507798356046e-05, "loss": 1.0102, "step": 8892 }, { "epoch": 2.74, "learning_rate": 1.705007312434162e-05, "loss": 0.755, "step": 8893 }, { "epoch": 2.74, "learning_rate": 1.7049366343064912e-05, "loss": 0.6383, "step": 8894 }, { "epoch": 2.74, "learning_rate": 1.7048659491781504e-05, "loss": 0.8099, "step": 8895 }, { "epoch": 2.74, "learning_rate": 1.7047952570498407e-05, "loss": 0.5578, "step": 8896 }, { "epoch": 2.74, "learning_rate": 1.7047245579222645e-05, "loss": 0.7622, "step": 8897 }, { "epoch": 2.74, "learning_rate": 1.704653851796124e-05, "loss": 0.9229, "step": 8898 }, { "epoch": 2.74, "learning_rate": 1.7045831386721213e-05, "loss": 0.7587, "step": 8899 }, { "epoch": 2.74, "learning_rate": 1.7045124185509586e-05, "loss": 0.8272, "step": 8900 }, { "epoch": 2.74, "learning_rate": 1.7044416914333387e-05, "loss": 0.8488, "step": 8901 }, { "epoch": 2.74, "learning_rate": 1.7043709573199635e-05, "loss": 0.7942, "step": 8902 }, { "epoch": 2.74, "learning_rate": 1.7043002162115348e-05, "loss": 0.9347, "step": 8903 }, { "epoch": 2.74, "learning_rate": 1.7042294681087566e-05, "loss": 1.1016, "step": 8904 }, { "epoch": 2.74, "learning_rate": 1.7041587130123306e-05, "loss": 0.622, "step": 8905 }, { "epoch": 2.74, "learning_rate": 1.7040879509229595e-05, "loss": 0.6742, "step": 8906 }, { "epoch": 2.74, "learning_rate": 1.7040171818413463e-05, "loss": 0.874, "step": 8907 }, { "epoch": 2.74, "learning_rate": 1.7039464057681937e-05, "loss": 0.9224, "step": 8908 }, { "epoch": 2.74, "learning_rate": 1.703875622704204e-05, "loss": 0.5873, "step": 8909 }, { "epoch": 2.74, "learning_rate": 1.7038048326500813e-05, "loss": 1.0789, "step": 8910 }, { "epoch": 2.74, "learning_rate": 1.7037340356065275e-05, "loss": 0.8092, "step": 8911 }, { "epoch": 2.74, "learning_rate": 1.7036632315742464e-05, "loss": 0.7906, "step": 8912 }, { "epoch": 2.74, "learning_rate": 1.703592420553941e-05, "loss": 0.8245, "step": 8913 }, { "epoch": 2.74, "learning_rate": 1.703521602546314e-05, "loss": 0.8885, "step": 8914 }, { "epoch": 2.74, "learning_rate": 1.7034507775520693e-05, "loss": 0.801, "step": 8915 }, { "epoch": 2.74, "learning_rate": 1.70337994557191e-05, "loss": 0.8613, "step": 8916 }, { "epoch": 2.74, "learning_rate": 1.70330910660654e-05, "loss": 0.7735, "step": 8917 }, { "epoch": 2.74, "learning_rate": 1.703238260656662e-05, "loss": 0.8767, "step": 8918 }, { "epoch": 2.74, "learning_rate": 1.7031674077229798e-05, "loss": 0.6508, "step": 8919 }, { "epoch": 2.74, "learning_rate": 1.7030965478061974e-05, "loss": 0.7771, "step": 8920 }, { "epoch": 2.74, "learning_rate": 1.7030256809070182e-05, "loss": 0.8238, "step": 8921 }, { "epoch": 2.75, "learning_rate": 1.7029548070261462e-05, "loss": 0.8048, "step": 8922 }, { "epoch": 2.75, "learning_rate": 1.7028839261642852e-05, "loss": 0.804, "step": 8923 }, { "epoch": 2.75, "learning_rate": 1.7028130383221388e-05, "loss": 0.7751, "step": 8924 }, { "epoch": 2.75, "learning_rate": 1.7027421435004114e-05, "loss": 0.8091, "step": 8925 }, { "epoch": 2.75, "learning_rate": 1.7026712416998064e-05, "loss": 0.693, "step": 8926 }, { "epoch": 2.75, "learning_rate": 1.702600332921029e-05, "loss": 0.7647, "step": 8927 }, { "epoch": 2.75, "learning_rate": 1.7025294171647823e-05, "loss": 0.7064, "step": 8928 }, { "epoch": 2.75, "learning_rate": 1.702458494431771e-05, "loss": 0.868, "step": 8929 }, { "epoch": 2.75, "learning_rate": 1.7023875647227e-05, "loss": 1.0766, "step": 8930 }, { "epoch": 2.75, "learning_rate": 1.7023166280382722e-05, "loss": 0.8566, "step": 8931 }, { "epoch": 2.75, "learning_rate": 1.702245684379194e-05, "loss": 1.0968, "step": 8932 }, { "epoch": 2.75, "learning_rate": 1.7021747337461682e-05, "loss": 1.0575, "step": 8933 }, { "epoch": 2.75, "learning_rate": 1.7021037761399007e-05, "loss": 0.709, "step": 8934 }, { "epoch": 2.75, "learning_rate": 1.702032811561095e-05, "loss": 0.9321, "step": 8935 }, { "epoch": 2.75, "learning_rate": 1.7019618400104572e-05, "loss": 0.96, "step": 8936 }, { "epoch": 2.75, "learning_rate": 1.7018908614886908e-05, "loss": 0.7947, "step": 8937 }, { "epoch": 2.75, "learning_rate": 1.7018198759965018e-05, "loss": 0.6787, "step": 8938 }, { "epoch": 2.75, "learning_rate": 1.701748883534594e-05, "loss": 0.8444, "step": 8939 }, { "epoch": 2.75, "learning_rate": 1.7016778841036733e-05, "loss": 0.7025, "step": 8940 }, { "epoch": 2.75, "learning_rate": 1.7016068777044448e-05, "loss": 0.887, "step": 8941 }, { "epoch": 2.75, "learning_rate": 1.7015358643376128e-05, "loss": 0.704, "step": 8942 }, { "epoch": 2.75, "learning_rate": 1.7014648440038837e-05, "loss": 0.7568, "step": 8943 }, { "epoch": 2.75, "learning_rate": 1.7013938167039617e-05, "loss": 0.7302, "step": 8944 }, { "epoch": 2.75, "learning_rate": 1.701322782438553e-05, "loss": 1.0081, "step": 8945 }, { "epoch": 2.75, "learning_rate": 1.701251741208363e-05, "loss": 0.7581, "step": 8946 }, { "epoch": 2.75, "learning_rate": 1.7011806930140962e-05, "loss": 0.7778, "step": 8947 }, { "epoch": 2.75, "learning_rate": 1.7011096378564594e-05, "loss": 0.8156, "step": 8948 }, { "epoch": 2.75, "learning_rate": 1.7010385757361575e-05, "loss": 0.8475, "step": 8949 }, { "epoch": 2.75, "learning_rate": 1.7009675066538967e-05, "loss": 0.814, "step": 8950 }, { "epoch": 2.75, "learning_rate": 1.7008964306103823e-05, "loss": 0.7471, "step": 8951 }, { "epoch": 2.75, "learning_rate": 1.7008253476063202e-05, "loss": 0.66, "step": 8952 }, { "epoch": 2.75, "learning_rate": 1.700754257642417e-05, "loss": 1.0581, "step": 8953 }, { "epoch": 2.75, "learning_rate": 1.700683160719378e-05, "loss": 1.2272, "step": 8954 }, { "epoch": 2.76, "learning_rate": 1.7006120568379092e-05, "loss": 0.6039, "step": 8955 }, { "epoch": 2.76, "learning_rate": 1.7005409459987174e-05, "loss": 0.9679, "step": 8956 }, { "epoch": 2.76, "learning_rate": 1.700469828202508e-05, "loss": 0.9624, "step": 8957 }, { "epoch": 2.76, "learning_rate": 1.700398703449988e-05, "loss": 0.7196, "step": 8958 }, { "epoch": 2.76, "learning_rate": 1.700327571741863e-05, "loss": 1.0251, "step": 8959 }, { "epoch": 2.76, "learning_rate": 1.70025643307884e-05, "loss": 0.895, "step": 8960 }, { "epoch": 2.76, "learning_rate": 1.700185287461625e-05, "loss": 0.7971, "step": 8961 }, { "epoch": 2.76, "learning_rate": 1.700114134890925e-05, "loss": 0.7275, "step": 8962 }, { "epoch": 2.76, "learning_rate": 1.7000429753674463e-05, "loss": 1.0445, "step": 8963 }, { "epoch": 2.76, "learning_rate": 1.6999718088918956e-05, "loss": 0.9615, "step": 8964 }, { "epoch": 2.76, "learning_rate": 1.69990063546498e-05, "loss": 0.9754, "step": 8965 }, { "epoch": 2.76, "learning_rate": 1.6998294550874057e-05, "loss": 0.8768, "step": 8966 }, { "epoch": 2.76, "learning_rate": 1.69975826775988e-05, "loss": 0.9324, "step": 8967 }, { "epoch": 2.76, "learning_rate": 1.69968707348311e-05, "loss": 0.8676, "step": 8968 }, { "epoch": 2.76, "learning_rate": 1.6996158722578024e-05, "loss": 0.8387, "step": 8969 }, { "epoch": 2.76, "learning_rate": 1.699544664084664e-05, "loss": 0.9743, "step": 8970 }, { "epoch": 2.76, "learning_rate": 1.6994734489644032e-05, "loss": 1.0737, "step": 8971 }, { "epoch": 2.76, "learning_rate": 1.699402226897726e-05, "loss": 0.6952, "step": 8972 }, { "epoch": 2.76, "learning_rate": 1.6993309978853402e-05, "loss": 0.8934, "step": 8973 }, { "epoch": 2.76, "learning_rate": 1.6992597619279526e-05, "loss": 0.7683, "step": 8974 }, { "epoch": 2.76, "learning_rate": 1.6991885190262714e-05, "loss": 0.5027, "step": 8975 }, { "epoch": 2.76, "learning_rate": 1.699117269181004e-05, "loss": 0.8347, "step": 8976 }, { "epoch": 2.76, "learning_rate": 1.6990460123928577e-05, "loss": 0.8026, "step": 8977 }, { "epoch": 2.76, "learning_rate": 1.69897474866254e-05, "loss": 0.598, "step": 8978 }, { "epoch": 2.76, "learning_rate": 1.698903477990759e-05, "loss": 0.7442, "step": 8979 }, { "epoch": 2.76, "learning_rate": 1.6988322003782224e-05, "loss": 0.9774, "step": 8980 }, { "epoch": 2.76, "learning_rate": 1.698760915825638e-05, "loss": 0.9387, "step": 8981 }, { "epoch": 2.76, "learning_rate": 1.6986896243337137e-05, "loss": 0.6452, "step": 8982 }, { "epoch": 2.76, "learning_rate": 1.6986183259031576e-05, "loss": 0.8804, "step": 8983 }, { "epoch": 2.76, "learning_rate": 1.6985470205346773e-05, "loss": 0.8172, "step": 8984 }, { "epoch": 2.76, "learning_rate": 1.6984757082289816e-05, "loss": 0.8449, "step": 8985 }, { "epoch": 2.76, "learning_rate": 1.6984043889867783e-05, "loss": 1.0844, "step": 8986 }, { "epoch": 2.77, "learning_rate": 1.6983330628087757e-05, "loss": 0.7201, "step": 8987 }, { "epoch": 2.77, "learning_rate": 1.698261729695682e-05, "loss": 0.8797, "step": 8988 }, { "epoch": 2.77, "learning_rate": 1.6981903896482062e-05, "loss": 1.0815, "step": 8989 }, { "epoch": 2.77, "learning_rate": 1.698119042667056e-05, "loss": 0.7622, "step": 8990 }, { "epoch": 2.77, "learning_rate": 1.6980476887529408e-05, "loss": 0.6064, "step": 8991 }, { "epoch": 2.77, "learning_rate": 1.6979763279065682e-05, "loss": 0.8145, "step": 8992 }, { "epoch": 2.77, "learning_rate": 1.697904960128648e-05, "loss": 0.7491, "step": 8993 }, { "epoch": 2.77, "learning_rate": 1.6978335854198875e-05, "loss": 0.6405, "step": 8994 }, { "epoch": 2.77, "learning_rate": 1.6977622037809967e-05, "loss": 0.7741, "step": 8995 }, { "epoch": 2.77, "learning_rate": 1.697690815212684e-05, "loss": 0.9339, "step": 8996 }, { "epoch": 2.77, "learning_rate": 1.6976194197156587e-05, "loss": 1.0016, "step": 8997 }, { "epoch": 2.77, "learning_rate": 1.6975480172906298e-05, "loss": 0.82, "step": 8998 }, { "epoch": 2.77, "learning_rate": 1.697476607938306e-05, "loss": 0.6912, "step": 8999 }, { "epoch": 2.77, "learning_rate": 1.6974051916593966e-05, "loss": 0.9349, "step": 9000 }, { "epoch": 2.77, "learning_rate": 1.6973337684546112e-05, "loss": 0.8074, "step": 9001 }, { "epoch": 2.77, "learning_rate": 1.6972623383246586e-05, "loss": 0.8215, "step": 9002 }, { "epoch": 2.77, "learning_rate": 1.6971909012702483e-05, "loss": 0.6792, "step": 9003 }, { "epoch": 2.77, "learning_rate": 1.69711945729209e-05, "loss": 1.0606, "step": 9004 }, { "epoch": 2.77, "learning_rate": 1.6970480063908925e-05, "loss": 0.6262, "step": 9005 }, { "epoch": 2.77, "learning_rate": 1.6969765485673665e-05, "loss": 0.6169, "step": 9006 }, { "epoch": 2.77, "learning_rate": 1.696905083822221e-05, "loss": 0.8192, "step": 9007 }, { "epoch": 2.77, "learning_rate": 1.6968336121561652e-05, "loss": 0.607, "step": 9008 }, { "epoch": 2.77, "learning_rate": 1.6967621335699097e-05, "loss": 0.6987, "step": 9009 }, { "epoch": 2.77, "learning_rate": 1.696690648064164e-05, "loss": 0.884, "step": 9010 }, { "epoch": 2.77, "learning_rate": 1.6966191556396382e-05, "loss": 0.807, "step": 9011 }, { "epoch": 2.77, "learning_rate": 1.696547656297042e-05, "loss": 0.5603, "step": 9012 }, { "epoch": 2.77, "learning_rate": 1.6964761500370857e-05, "loss": 0.6098, "step": 9013 }, { "epoch": 2.77, "learning_rate": 1.6964046368604793e-05, "loss": 0.9131, "step": 9014 }, { "epoch": 2.77, "learning_rate": 1.6963331167679332e-05, "loss": 0.4891, "step": 9015 }, { "epoch": 2.77, "learning_rate": 1.6962615897601573e-05, "loss": 0.7315, "step": 9016 }, { "epoch": 2.77, "learning_rate": 1.696190055837862e-05, "loss": 0.6395, "step": 9017 }, { "epoch": 2.77, "learning_rate": 1.6961185150017584e-05, "loss": 0.8173, "step": 9018 }, { "epoch": 2.77, "learning_rate": 1.696046967252556e-05, "loss": 0.6244, "step": 9019 }, { "epoch": 2.78, "learning_rate": 1.6959754125909656e-05, "loss": 0.6709, "step": 9020 }, { "epoch": 2.78, "learning_rate": 1.695903851017698e-05, "loss": 0.655, "step": 9021 }, { "epoch": 2.78, "learning_rate": 1.695832282533464e-05, "loss": 1.0468, "step": 9022 }, { "epoch": 2.78, "learning_rate": 1.6957607071389737e-05, "loss": 0.9914, "step": 9023 }, { "epoch": 2.78, "learning_rate": 1.6956891248349388e-05, "loss": 0.64, "step": 9024 }, { "epoch": 2.78, "learning_rate": 1.6956175356220696e-05, "loss": 0.7901, "step": 9025 }, { "epoch": 2.78, "learning_rate": 1.6955459395010767e-05, "loss": 1.0294, "step": 9026 }, { "epoch": 2.78, "learning_rate": 1.6954743364726722e-05, "loss": 0.9197, "step": 9027 }, { "epoch": 2.78, "learning_rate": 1.6954027265375664e-05, "loss": 1.7604, "step": 9028 }, { "epoch": 2.78, "learning_rate": 1.6953311096964706e-05, "loss": 0.5142, "step": 9029 }, { "epoch": 2.78, "learning_rate": 1.695259485950096e-05, "loss": 1.0155, "step": 9030 }, { "epoch": 2.78, "learning_rate": 1.695187855299154e-05, "loss": 0.8329, "step": 9031 }, { "epoch": 2.78, "learning_rate": 1.6951162177443557e-05, "loss": 0.9462, "step": 9032 }, { "epoch": 2.78, "learning_rate": 1.695044573286413e-05, "loss": 0.7455, "step": 9033 }, { "epoch": 2.78, "learning_rate": 1.6949729219260368e-05, "loss": 0.7072, "step": 9034 }, { "epoch": 2.78, "learning_rate": 1.6949012636639395e-05, "loss": 0.6787, "step": 9035 }, { "epoch": 2.78, "learning_rate": 1.6948295985008317e-05, "loss": 0.8643, "step": 9036 }, { "epoch": 2.78, "learning_rate": 1.694757926437426e-05, "loss": 0.559, "step": 9037 }, { "epoch": 2.78, "learning_rate": 1.6946862474744333e-05, "loss": 0.8549, "step": 9038 }, { "epoch": 2.78, "learning_rate": 1.6946145616125664e-05, "loss": 0.7348, "step": 9039 }, { "epoch": 2.78, "learning_rate": 1.6945428688525367e-05, "loss": 0.7591, "step": 9040 }, { "epoch": 2.78, "learning_rate": 1.694471169195056e-05, "loss": 0.7348, "step": 9041 }, { "epoch": 2.78, "learning_rate": 1.6943994626408365e-05, "loss": 1.0209, "step": 9042 }, { "epoch": 2.78, "learning_rate": 1.6943277491905905e-05, "loss": 0.9681, "step": 9043 }, { "epoch": 2.78, "learning_rate": 1.6942560288450297e-05, "loss": 0.9251, "step": 9044 }, { "epoch": 2.78, "learning_rate": 1.6941843016048667e-05, "loss": 0.8344, "step": 9045 }, { "epoch": 2.78, "learning_rate": 1.6941125674708144e-05, "loss": 0.7942, "step": 9046 }, { "epoch": 2.78, "learning_rate": 1.6940408264435842e-05, "loss": 0.932, "step": 9047 }, { "epoch": 2.78, "learning_rate": 1.693969078523889e-05, "loss": 0.7859, "step": 9048 }, { "epoch": 2.78, "learning_rate": 1.6938973237124415e-05, "loss": 0.8194, "step": 9049 }, { "epoch": 2.78, "learning_rate": 1.6938255620099536e-05, "loss": 0.8925, "step": 9050 }, { "epoch": 2.78, "learning_rate": 1.693753793417139e-05, "loss": 0.7141, "step": 9051 }, { "epoch": 2.79, "learning_rate": 1.6936820179347092e-05, "loss": 0.7165, "step": 9052 }, { "epoch": 2.79, "learning_rate": 1.693610235563378e-05, "loss": 0.7426, "step": 9053 }, { "epoch": 2.79, "learning_rate": 1.693538446303858e-05, "loss": 0.8216, "step": 9054 }, { "epoch": 2.79, "learning_rate": 1.6934666501568618e-05, "loss": 0.6451, "step": 9055 }, { "epoch": 2.79, "learning_rate": 1.6933948471231027e-05, "loss": 0.8383, "step": 9056 }, { "epoch": 2.79, "learning_rate": 1.6933230372032938e-05, "loss": 0.95, "step": 9057 }, { "epoch": 2.79, "learning_rate": 1.6932512203981485e-05, "loss": 0.8223, "step": 9058 }, { "epoch": 2.79, "learning_rate": 1.693179396708379e-05, "loss": 0.7394, "step": 9059 }, { "epoch": 2.79, "learning_rate": 1.6931075661346997e-05, "loss": 1.0606, "step": 9060 }, { "epoch": 2.79, "learning_rate": 1.6930357286778233e-05, "loss": 0.7838, "step": 9061 }, { "epoch": 2.79, "learning_rate": 1.6929638843384633e-05, "loss": 0.9072, "step": 9062 }, { "epoch": 2.79, "learning_rate": 1.6928920331173336e-05, "loss": 0.9381, "step": 9063 }, { "epoch": 2.79, "learning_rate": 1.6928201750151468e-05, "loss": 0.7523, "step": 9064 }, { "epoch": 2.79, "learning_rate": 1.6927483100326177e-05, "loss": 0.7488, "step": 9065 }, { "epoch": 2.79, "learning_rate": 1.6926764381704593e-05, "loss": 0.8542, "step": 9066 }, { "epoch": 2.79, "learning_rate": 1.6926045594293854e-05, "loss": 0.8996, "step": 9067 }, { "epoch": 2.79, "learning_rate": 1.69253267381011e-05, "loss": 0.8535, "step": 9068 }, { "epoch": 2.79, "learning_rate": 1.6924607813133465e-05, "loss": 0.9634, "step": 9069 }, { "epoch": 2.79, "learning_rate": 1.69238888193981e-05, "loss": 0.861, "step": 9070 }, { "epoch": 2.79, "learning_rate": 1.6923169756902132e-05, "loss": 0.6957, "step": 9071 }, { "epoch": 2.79, "learning_rate": 1.692245062565271e-05, "loss": 1.0829, "step": 9072 }, { "epoch": 2.79, "learning_rate": 1.6921731425656972e-05, "loss": 0.8945, "step": 9073 }, { "epoch": 2.79, "learning_rate": 1.6921012156922065e-05, "loss": 0.763, "step": 9074 }, { "epoch": 2.79, "learning_rate": 1.6920292819455123e-05, "loss": 0.7551, "step": 9075 }, { "epoch": 2.79, "learning_rate": 1.6919573413263298e-05, "loss": 0.7593, "step": 9076 }, { "epoch": 2.79, "learning_rate": 1.6918853938353734e-05, "loss": 0.7475, "step": 9077 }, { "epoch": 2.79, "learning_rate": 1.6918134394733572e-05, "loss": 0.7599, "step": 9078 }, { "epoch": 2.79, "learning_rate": 1.691741478240996e-05, "loss": 0.7242, "step": 9079 }, { "epoch": 2.79, "learning_rate": 1.6916695101390044e-05, "loss": 0.5296, "step": 9080 }, { "epoch": 2.79, "learning_rate": 1.6915975351680968e-05, "loss": 0.7929, "step": 9081 }, { "epoch": 2.79, "learning_rate": 1.6915255533289888e-05, "loss": 0.8954, "step": 9082 }, { "epoch": 2.79, "learning_rate": 1.6914535646223944e-05, "loss": 0.8419, "step": 9083 }, { "epoch": 2.79, "learning_rate": 1.691381569049029e-05, "loss": 0.8006, "step": 9084 }, { "epoch": 2.8, "learning_rate": 1.6913095666096073e-05, "loss": 0.9818, "step": 9085 }, { "epoch": 2.8, "learning_rate": 1.6912375573048444e-05, "loss": 0.7555, "step": 9086 }, { "epoch": 2.8, "learning_rate": 1.6911655411354557e-05, "loss": 0.5369, "step": 9087 }, { "epoch": 2.8, "learning_rate": 1.691093518102156e-05, "loss": 0.6997, "step": 9088 }, { "epoch": 2.8, "learning_rate": 1.691021488205661e-05, "loss": 0.8098, "step": 9089 }, { "epoch": 2.8, "learning_rate": 1.6909494514466855e-05, "loss": 0.7404, "step": 9090 }, { "epoch": 2.8, "learning_rate": 1.690877407825945e-05, "loss": 1.1954, "step": 9091 }, { "epoch": 2.8, "learning_rate": 1.6908053573441556e-05, "loss": 0.8266, "step": 9092 }, { "epoch": 2.8, "learning_rate": 1.6907333000020322e-05, "loss": 0.684, "step": 9093 }, { "epoch": 2.8, "learning_rate": 1.69066123580029e-05, "loss": 1.1403, "step": 9094 }, { "epoch": 2.8, "learning_rate": 1.690589164739646e-05, "loss": 1.0435, "step": 9095 }, { "epoch": 2.8, "learning_rate": 1.6905170868208144e-05, "loss": 0.6531, "step": 9096 }, { "epoch": 2.8, "learning_rate": 1.6904450020445123e-05, "loss": 0.7142, "step": 9097 }, { "epoch": 2.8, "learning_rate": 1.6903729104114547e-05, "loss": 0.6795, "step": 9098 }, { "epoch": 2.8, "learning_rate": 1.6903008119223576e-05, "loss": 0.7875, "step": 9099 }, { "epoch": 2.8, "learning_rate": 1.6902287065779376e-05, "loss": 0.6721, "step": 9100 }, { "epoch": 2.8, "learning_rate": 1.6901565943789103e-05, "loss": 0.7977, "step": 9101 }, { "epoch": 2.8, "learning_rate": 1.690084475325992e-05, "loss": 0.8397, "step": 9102 }, { "epoch": 2.8, "learning_rate": 1.690012349419899e-05, "loss": 0.7785, "step": 9103 }, { "epoch": 2.8, "learning_rate": 1.6899402166613473e-05, "loss": 0.7674, "step": 9104 }, { "epoch": 2.8, "learning_rate": 1.689868077051053e-05, "loss": 0.8677, "step": 9105 }, { "epoch": 2.8, "learning_rate": 1.6897959305897334e-05, "loss": 0.9335, "step": 9106 }, { "epoch": 2.8, "learning_rate": 1.6897237772781046e-05, "loss": 0.767, "step": 9107 }, { "epoch": 2.8, "learning_rate": 1.6896516171168825e-05, "loss": 0.6748, "step": 9108 }, { "epoch": 2.8, "learning_rate": 1.6895794501067844e-05, "loss": 0.9554, "step": 9109 }, { "epoch": 2.8, "learning_rate": 1.6895072762485273e-05, "loss": 0.7738, "step": 9110 }, { "epoch": 2.8, "learning_rate": 1.6894350955428267e-05, "loss": 0.6168, "step": 9111 }, { "epoch": 2.8, "learning_rate": 1.6893629079904006e-05, "loss": 0.8041, "step": 9112 }, { "epoch": 2.8, "learning_rate": 1.6892907135919655e-05, "loss": 0.7625, "step": 9113 }, { "epoch": 2.8, "learning_rate": 1.6892185123482384e-05, "loss": 0.8158, "step": 9114 }, { "epoch": 2.8, "learning_rate": 1.6891463042599362e-05, "loss": 0.9826, "step": 9115 }, { "epoch": 2.8, "learning_rate": 1.689074089327776e-05, "loss": 0.801, "step": 9116 }, { "epoch": 2.81, "learning_rate": 1.6890018675524752e-05, "loss": 0.9339, "step": 9117 }, { "epoch": 2.81, "learning_rate": 1.6889296389347507e-05, "loss": 0.9481, "step": 9118 }, { "epoch": 2.81, "learning_rate": 1.68885740347532e-05, "loss": 0.688, "step": 9119 }, { "epoch": 2.81, "learning_rate": 1.6887851611749005e-05, "loss": 0.6933, "step": 9120 }, { "epoch": 2.81, "learning_rate": 1.6887129120342093e-05, "loss": 1.0525, "step": 9121 }, { "epoch": 2.81, "learning_rate": 1.688640656053964e-05, "loss": 0.872, "step": 9122 }, { "epoch": 2.81, "learning_rate": 1.688568393234883e-05, "loss": 0.9563, "step": 9123 }, { "epoch": 2.81, "learning_rate": 1.6884961235776827e-05, "loss": 0.5043, "step": 9124 }, { "epoch": 2.81, "learning_rate": 1.6884238470830813e-05, "loss": 0.8616, "step": 9125 }, { "epoch": 2.81, "learning_rate": 1.688351563751797e-05, "loss": 0.9387, "step": 9126 }, { "epoch": 2.81, "learning_rate": 1.6882792735845472e-05, "loss": 0.8192, "step": 9127 }, { "epoch": 2.81, "learning_rate": 1.6882069765820496e-05, "loss": 0.5842, "step": 9128 }, { "epoch": 2.81, "learning_rate": 1.6881346727450228e-05, "loss": 0.7299, "step": 9129 }, { "epoch": 2.81, "learning_rate": 1.6880623620741843e-05, "loss": 0.7821, "step": 9130 }, { "epoch": 2.81, "learning_rate": 1.6879900445702524e-05, "loss": 0.8796, "step": 9131 }, { "epoch": 2.81, "learning_rate": 1.687917720233945e-05, "loss": 0.803, "step": 9132 }, { "epoch": 2.81, "learning_rate": 1.6878453890659815e-05, "loss": 0.6059, "step": 9133 }, { "epoch": 2.81, "learning_rate": 1.687773051067079e-05, "loss": 1.3515, "step": 9134 }, { "epoch": 2.81, "learning_rate": 1.687700706237956e-05, "loss": 0.8823, "step": 9135 }, { "epoch": 2.81, "learning_rate": 1.687628354579331e-05, "loss": 0.848, "step": 9136 }, { "epoch": 2.81, "learning_rate": 1.6875559960919227e-05, "loss": 0.7011, "step": 9137 }, { "epoch": 2.81, "learning_rate": 1.68748363077645e-05, "loss": 0.6588, "step": 9138 }, { "epoch": 2.81, "learning_rate": 1.6874112586336312e-05, "loss": 0.8178, "step": 9139 }, { "epoch": 2.81, "learning_rate": 1.687338879664185e-05, "loss": 1.0502, "step": 9140 }, { "epoch": 2.81, "learning_rate": 1.6872664938688302e-05, "loss": 0.8466, "step": 9141 }, { "epoch": 2.81, "learning_rate": 1.687194101248286e-05, "loss": 0.5418, "step": 9142 }, { "epoch": 2.81, "learning_rate": 1.687121701803271e-05, "loss": 0.7845, "step": 9143 }, { "epoch": 2.81, "learning_rate": 1.6870492955345038e-05, "loss": 0.7235, "step": 9144 }, { "epoch": 2.81, "learning_rate": 1.686976882442704e-05, "loss": 0.6755, "step": 9145 }, { "epoch": 2.81, "learning_rate": 1.686904462528591e-05, "loss": 0.8878, "step": 9146 }, { "epoch": 2.81, "learning_rate": 1.6868320357928833e-05, "loss": 0.7388, "step": 9147 }, { "epoch": 2.81, "learning_rate": 1.6867596022363008e-05, "loss": 0.7849, "step": 9148 }, { "epoch": 2.81, "learning_rate": 1.6866871618595622e-05, "loss": 0.7605, "step": 9149 }, { "epoch": 2.82, "learning_rate": 1.686614714663387e-05, "loss": 0.693, "step": 9150 }, { "epoch": 2.82, "learning_rate": 1.686542260648495e-05, "loss": 0.6036, "step": 9151 }, { "epoch": 2.82, "learning_rate": 1.6864697998156063e-05, "loss": 1.0689, "step": 9152 }, { "epoch": 2.82, "learning_rate": 1.686397332165439e-05, "loss": 0.8523, "step": 9153 }, { "epoch": 2.82, "learning_rate": 1.686324857698714e-05, "loss": 0.919, "step": 9154 }, { "epoch": 2.82, "learning_rate": 1.6862523764161503e-05, "loss": 0.8875, "step": 9155 }, { "epoch": 2.82, "learning_rate": 1.6861798883184686e-05, "loss": 0.8746, "step": 9156 }, { "epoch": 2.82, "learning_rate": 1.6861073934063878e-05, "loss": 0.9527, "step": 9157 }, { "epoch": 2.82, "learning_rate": 1.6860348916806283e-05, "loss": 0.7959, "step": 9158 }, { "epoch": 2.82, "learning_rate": 1.68596238314191e-05, "loss": 0.8297, "step": 9159 }, { "epoch": 2.82, "learning_rate": 1.6858898677909534e-05, "loss": 0.664, "step": 9160 }, { "epoch": 2.82, "learning_rate": 1.6858173456284777e-05, "loss": 0.9155, "step": 9161 }, { "epoch": 2.82, "learning_rate": 1.685744816655204e-05, "loss": 0.657, "step": 9162 }, { "epoch": 2.82, "learning_rate": 1.6856722808718524e-05, "loss": 0.8566, "step": 9163 }, { "epoch": 2.82, "learning_rate": 1.685599738279143e-05, "loss": 0.8138, "step": 9164 }, { "epoch": 2.82, "learning_rate": 1.685527188877796e-05, "loss": 0.7249, "step": 9165 }, { "epoch": 2.82, "learning_rate": 1.685454632668533e-05, "loss": 0.6442, "step": 9166 }, { "epoch": 2.82, "learning_rate": 1.6853820696520728e-05, "loss": 0.7809, "step": 9167 }, { "epoch": 2.82, "learning_rate": 1.6853094998291378e-05, "loss": 1.0208, "step": 9168 }, { "epoch": 2.82, "learning_rate": 1.6852369232004476e-05, "loss": 0.5413, "step": 9169 }, { "epoch": 2.82, "learning_rate": 1.6851643397667228e-05, "loss": 0.6598, "step": 9170 }, { "epoch": 2.82, "learning_rate": 1.685091749528685e-05, "loss": 0.7335, "step": 9171 }, { "epoch": 2.82, "learning_rate": 1.6850191524870548e-05, "loss": 0.804, "step": 9172 }, { "epoch": 2.82, "learning_rate": 1.6849465486425532e-05, "loss": 0.8533, "step": 9173 }, { "epoch": 2.82, "learning_rate": 1.6848739379959007e-05, "loss": 0.7667, "step": 9174 }, { "epoch": 2.82, "learning_rate": 1.684801320547819e-05, "loss": 0.5973, "step": 9175 }, { "epoch": 2.82, "learning_rate": 1.684728696299029e-05, "loss": 0.5041, "step": 9176 }, { "epoch": 2.82, "learning_rate": 1.684656065250252e-05, "loss": 0.9284, "step": 9177 }, { "epoch": 2.82, "learning_rate": 1.6845834274022094e-05, "loss": 0.9571, "step": 9178 }, { "epoch": 2.82, "learning_rate": 1.6845107827556226e-05, "loss": 0.7828, "step": 9179 }, { "epoch": 2.82, "learning_rate": 1.6844381313112125e-05, "loss": 0.9202, "step": 9180 }, { "epoch": 2.82, "learning_rate": 1.684365473069701e-05, "loss": 1.1028, "step": 9181 }, { "epoch": 2.83, "learning_rate": 1.6842928080318094e-05, "loss": 0.9044, "step": 9182 }, { "epoch": 2.83, "learning_rate": 1.6842201361982602e-05, "loss": 0.9569, "step": 9183 }, { "epoch": 2.83, "learning_rate": 1.684147457569774e-05, "loss": 0.7444, "step": 9184 }, { "epoch": 2.83, "learning_rate": 1.6840747721470733e-05, "loss": 0.7316, "step": 9185 }, { "epoch": 2.83, "learning_rate": 1.6840020799308793e-05, "loss": 0.9782, "step": 9186 }, { "epoch": 2.83, "learning_rate": 1.683929380921914e-05, "loss": 0.6628, "step": 9187 }, { "epoch": 2.83, "learning_rate": 1.6838566751209003e-05, "loss": 0.7487, "step": 9188 }, { "epoch": 2.83, "learning_rate": 1.6837839625285593e-05, "loss": 0.8048, "step": 9189 }, { "epoch": 2.83, "learning_rate": 1.683711243145613e-05, "loss": 0.8, "step": 9190 }, { "epoch": 2.83, "learning_rate": 1.6836385169727848e-05, "loss": 0.6835, "step": 9191 }, { "epoch": 2.83, "learning_rate": 1.6835657840107953e-05, "loss": 1.0364, "step": 9192 }, { "epoch": 2.83, "learning_rate": 1.683493044260368e-05, "loss": 1.0552, "step": 9193 }, { "epoch": 2.83, "learning_rate": 1.6834202977222248e-05, "loss": 0.8105, "step": 9194 }, { "epoch": 2.83, "learning_rate": 1.6833475443970878e-05, "loss": 0.8652, "step": 9195 }, { "epoch": 2.83, "learning_rate": 1.68327478428568e-05, "loss": 0.8349, "step": 9196 }, { "epoch": 2.83, "learning_rate": 1.683202017388724e-05, "loss": 0.7974, "step": 9197 }, { "epoch": 2.83, "learning_rate": 1.6831292437069425e-05, "loss": 0.8023, "step": 9198 }, { "epoch": 2.83, "learning_rate": 1.6830564632410576e-05, "loss": 0.821, "step": 9199 }, { "epoch": 2.83, "learning_rate": 1.682983675991793e-05, "loss": 0.8297, "step": 9200 }, { "epoch": 2.83, "learning_rate": 1.682910881959871e-05, "loss": 0.846, "step": 9201 }, { "epoch": 2.83, "learning_rate": 1.682838081146014e-05, "loss": 0.829, "step": 9202 }, { "epoch": 2.83, "learning_rate": 1.682765273550946e-05, "loss": 0.6526, "step": 9203 }, { "epoch": 2.83, "learning_rate": 1.6826924591753895e-05, "loss": 1.0127, "step": 9204 }, { "epoch": 2.83, "learning_rate": 1.6826196380200673e-05, "loss": 0.6854, "step": 9205 }, { "epoch": 2.83, "learning_rate": 1.6825468100857036e-05, "loss": 0.7865, "step": 9206 }, { "epoch": 2.83, "learning_rate": 1.6824739753730206e-05, "loss": 1.1228, "step": 9207 }, { "epoch": 2.83, "learning_rate": 1.6824011338827423e-05, "loss": 0.6722, "step": 9208 }, { "epoch": 2.83, "learning_rate": 1.6823282856155915e-05, "loss": 0.7061, "step": 9209 }, { "epoch": 2.83, "learning_rate": 1.6822554305722923e-05, "loss": 0.7966, "step": 9210 }, { "epoch": 2.83, "learning_rate": 1.6821825687535675e-05, "loss": 0.9511, "step": 9211 }, { "epoch": 2.83, "learning_rate": 1.6821097001601413e-05, "loss": 0.9665, "step": 9212 }, { "epoch": 2.83, "learning_rate": 1.6820368247927372e-05, "loss": 0.7932, "step": 9213 }, { "epoch": 2.83, "learning_rate": 1.681963942652079e-05, "loss": 0.6937, "step": 9214 }, { "epoch": 2.84, "learning_rate": 1.6818910537388903e-05, "loss": 0.7427, "step": 9215 }, { "epoch": 2.84, "learning_rate": 1.6818181580538946e-05, "loss": 0.6815, "step": 9216 }, { "epoch": 2.84, "learning_rate": 1.681745255597816e-05, "loss": 0.824, "step": 9217 }, { "epoch": 2.84, "learning_rate": 1.6816723463713795e-05, "loss": 0.8887, "step": 9218 }, { "epoch": 2.84, "learning_rate": 1.681599430375308e-05, "loss": 0.7692, "step": 9219 }, { "epoch": 2.84, "learning_rate": 1.6815265076103256e-05, "loss": 0.6755, "step": 9220 }, { "epoch": 2.84, "learning_rate": 1.6814535780771574e-05, "loss": 0.7902, "step": 9221 }, { "epoch": 2.84, "learning_rate": 1.681380641776527e-05, "loss": 0.956, "step": 9222 }, { "epoch": 2.84, "learning_rate": 1.6813076987091588e-05, "loss": 1.0903, "step": 9223 }, { "epoch": 2.84, "learning_rate": 1.6812347488757774e-05, "loss": 0.9328, "step": 9224 }, { "epoch": 2.84, "learning_rate": 1.681161792277107e-05, "loss": 0.7097, "step": 9225 }, { "epoch": 2.84, "learning_rate": 1.6810888289138717e-05, "loss": 0.9246, "step": 9226 }, { "epoch": 2.84, "learning_rate": 1.6810158587867973e-05, "loss": 0.7056, "step": 9227 }, { "epoch": 2.84, "learning_rate": 1.6809428818966072e-05, "loss": 0.9972, "step": 9228 }, { "epoch": 2.84, "learning_rate": 1.6808698982440276e-05, "loss": 0.6304, "step": 9229 }, { "epoch": 2.84, "learning_rate": 1.6807969078297816e-05, "loss": 1.0863, "step": 9230 }, { "epoch": 2.84, "learning_rate": 1.680723910654595e-05, "loss": 0.699, "step": 9231 }, { "epoch": 2.84, "learning_rate": 1.6806509067191926e-05, "loss": 0.6406, "step": 9232 }, { "epoch": 2.84, "learning_rate": 1.6805778960242998e-05, "loss": 0.7965, "step": 9233 }, { "epoch": 2.84, "learning_rate": 1.6805048785706407e-05, "loss": 0.7177, "step": 9234 }, { "epoch": 2.84, "learning_rate": 1.6804318543589415e-05, "loss": 0.6225, "step": 9235 }, { "epoch": 2.84, "learning_rate": 1.6803588233899265e-05, "loss": 0.8404, "step": 9236 }, { "epoch": 2.84, "learning_rate": 1.6802857856643214e-05, "loss": 0.7124, "step": 9237 }, { "epoch": 2.84, "learning_rate": 1.6802127411828517e-05, "loss": 0.8391, "step": 9238 }, { "epoch": 2.84, "learning_rate": 1.6801396899462427e-05, "loss": 1.0091, "step": 9239 }, { "epoch": 2.84, "learning_rate": 1.6800666319552192e-05, "loss": 0.8417, "step": 9240 }, { "epoch": 2.84, "learning_rate": 1.6799935672105076e-05, "loss": 0.7989, "step": 9241 }, { "epoch": 2.84, "learning_rate": 1.6799204957128332e-05, "loss": 0.865, "step": 9242 }, { "epoch": 2.84, "learning_rate": 1.6798474174629212e-05, "loss": 0.9389, "step": 9243 }, { "epoch": 2.84, "learning_rate": 1.6797743324614987e-05, "loss": 1.1124, "step": 9244 }, { "epoch": 2.84, "learning_rate": 1.6797012407092896e-05, "loss": 0.6327, "step": 9245 }, { "epoch": 2.84, "learning_rate": 1.6796281422070214e-05, "loss": 0.812, "step": 9246 }, { "epoch": 2.85, "learning_rate": 1.679555036955419e-05, "loss": 0.9711, "step": 9247 }, { "epoch": 2.85, "learning_rate": 1.679481924955209e-05, "loss": 0.7283, "step": 9248 }, { "epoch": 2.85, "learning_rate": 1.6794088062071174e-05, "loss": 0.6956, "step": 9249 }, { "epoch": 2.85, "learning_rate": 1.6793356807118695e-05, "loss": 0.6591, "step": 9250 }, { "epoch": 2.85, "learning_rate": 1.679262548470193e-05, "loss": 0.9175, "step": 9251 }, { "epoch": 2.85, "learning_rate": 1.679189409482813e-05, "loss": 0.5972, "step": 9252 }, { "epoch": 2.85, "learning_rate": 1.6791162637504564e-05, "loss": 0.7216, "step": 9253 }, { "epoch": 2.85, "learning_rate": 1.6790431112738495e-05, "loss": 0.6246, "step": 9254 }, { "epoch": 2.85, "learning_rate": 1.6789699520537183e-05, "loss": 0.9338, "step": 9255 }, { "epoch": 2.85, "learning_rate": 1.67889678609079e-05, "loss": 0.7762, "step": 9256 }, { "epoch": 2.85, "learning_rate": 1.6788236133857906e-05, "loss": 1.0656, "step": 9257 }, { "epoch": 2.85, "learning_rate": 1.6787504339394473e-05, "loss": 0.909, "step": 9258 }, { "epoch": 2.85, "learning_rate": 1.6786772477524866e-05, "loss": 0.6282, "step": 9259 }, { "epoch": 2.85, "learning_rate": 1.6786040548256354e-05, "loss": 0.859, "step": 9260 }, { "epoch": 2.85, "learning_rate": 1.6785308551596206e-05, "loss": 0.9886, "step": 9261 }, { "epoch": 2.85, "learning_rate": 1.6784576487551686e-05, "loss": 0.8822, "step": 9262 }, { "epoch": 2.85, "learning_rate": 1.6783844356130073e-05, "loss": 0.7124, "step": 9263 }, { "epoch": 2.85, "learning_rate": 1.678311215733863e-05, "loss": 1.044, "step": 9264 }, { "epoch": 2.85, "learning_rate": 1.6782379891184638e-05, "loss": 0.7494, "step": 9265 }, { "epoch": 2.85, "learning_rate": 1.6781647557675356e-05, "loss": 1.0608, "step": 9266 }, { "epoch": 2.85, "learning_rate": 1.678091515681807e-05, "loss": 0.6384, "step": 9267 }, { "epoch": 2.85, "learning_rate": 1.6780182688620042e-05, "loss": 0.8125, "step": 9268 }, { "epoch": 2.85, "learning_rate": 1.6779450153088555e-05, "loss": 1.0039, "step": 9269 }, { "epoch": 2.85, "learning_rate": 1.6778717550230877e-05, "loss": 0.7125, "step": 9270 }, { "epoch": 2.85, "learning_rate": 1.6777984880054287e-05, "loss": 0.9128, "step": 9271 }, { "epoch": 2.85, "learning_rate": 1.677725214256606e-05, "loss": 0.6925, "step": 9272 }, { "epoch": 2.85, "learning_rate": 1.6776519337773473e-05, "loss": 0.6814, "step": 9273 }, { "epoch": 2.85, "learning_rate": 1.6775786465683805e-05, "loss": 0.6882, "step": 9274 }, { "epoch": 2.85, "learning_rate": 1.677505352630433e-05, "loss": 0.7401, "step": 9275 }, { "epoch": 2.85, "learning_rate": 1.677432051964233e-05, "loss": 1.0367, "step": 9276 }, { "epoch": 2.85, "learning_rate": 1.6773587445705087e-05, "loss": 0.6205, "step": 9277 }, { "epoch": 2.85, "learning_rate": 1.6772854304499875e-05, "loss": 0.9036, "step": 9278 }, { "epoch": 2.85, "learning_rate": 1.677212109603398e-05, "loss": 0.8452, "step": 9279 }, { "epoch": 2.86, "learning_rate": 1.677138782031468e-05, "loss": 0.8202, "step": 9280 }, { "epoch": 2.86, "learning_rate": 1.6770654477349256e-05, "loss": 0.777, "step": 9281 }, { "epoch": 2.86, "learning_rate": 1.6769921067144994e-05, "loss": 0.755, "step": 9282 }, { "epoch": 2.86, "learning_rate": 1.676918758970918e-05, "loss": 0.6762, "step": 9283 }, { "epoch": 2.86, "learning_rate": 1.676845404504909e-05, "loss": 0.7241, "step": 9284 }, { "epoch": 2.86, "learning_rate": 1.6767720433172016e-05, "loss": 0.6781, "step": 9285 }, { "epoch": 2.86, "learning_rate": 1.676698675408524e-05, "loss": 0.6263, "step": 9286 }, { "epoch": 2.86, "learning_rate": 1.676625300779605e-05, "loss": 0.8511, "step": 9287 }, { "epoch": 2.86, "learning_rate": 1.676551919431173e-05, "loss": 0.8346, "step": 9288 }, { "epoch": 2.86, "learning_rate": 1.6764785313639568e-05, "loss": 0.8194, "step": 9289 }, { "epoch": 2.86, "learning_rate": 1.6764051365786855e-05, "loss": 1.0231, "step": 9290 }, { "epoch": 2.86, "learning_rate": 1.676331735076088e-05, "loss": 0.7598, "step": 9291 }, { "epoch": 2.86, "learning_rate": 1.6762583268568928e-05, "loss": 0.6759, "step": 9292 }, { "epoch": 2.86, "learning_rate": 1.6761849119218288e-05, "loss": 0.9438, "step": 9293 }, { "epoch": 2.86, "learning_rate": 1.6761114902716262e-05, "loss": 1.0766, "step": 9294 }, { "epoch": 2.86, "learning_rate": 1.6760380619070128e-05, "loss": 0.6772, "step": 9295 }, { "epoch": 2.86, "learning_rate": 1.675964626828719e-05, "loss": 0.6905, "step": 9296 }, { "epoch": 2.86, "learning_rate": 1.675891185037473e-05, "loss": 0.8625, "step": 9297 }, { "epoch": 2.86, "learning_rate": 1.6758177365340044e-05, "loss": 0.7431, "step": 9298 }, { "epoch": 2.86, "learning_rate": 1.675744281319043e-05, "loss": 0.9524, "step": 9299 }, { "epoch": 2.86, "learning_rate": 1.6756708193933184e-05, "loss": 0.7437, "step": 9300 }, { "epoch": 2.86, "learning_rate": 1.67559735075756e-05, "loss": 0.7171, "step": 9301 }, { "epoch": 2.86, "learning_rate": 1.6755238754124965e-05, "loss": 0.6721, "step": 9302 }, { "epoch": 2.86, "learning_rate": 1.675450393358859e-05, "loss": 0.8842, "step": 9303 }, { "epoch": 2.86, "learning_rate": 1.675376904597376e-05, "loss": 0.7219, "step": 9304 }, { "epoch": 2.86, "learning_rate": 1.6753034091287787e-05, "loss": 0.8571, "step": 9305 }, { "epoch": 2.86, "learning_rate": 1.6752299069537955e-05, "loss": 0.6699, "step": 9306 }, { "epoch": 2.86, "learning_rate": 1.675156398073157e-05, "loss": 0.8183, "step": 9307 }, { "epoch": 2.86, "learning_rate": 1.6750828824875936e-05, "loss": 0.7256, "step": 9308 }, { "epoch": 2.86, "learning_rate": 1.6750093601978345e-05, "loss": 0.6647, "step": 9309 }, { "epoch": 2.86, "learning_rate": 1.6749358312046112e-05, "loss": 0.9084, "step": 9310 }, { "epoch": 2.86, "learning_rate": 1.6748622955086526e-05, "loss": 0.7979, "step": 9311 }, { "epoch": 2.87, "learning_rate": 1.674788753110689e-05, "loss": 0.8523, "step": 9312 }, { "epoch": 2.87, "learning_rate": 1.6747152040114515e-05, "loss": 0.9793, "step": 9313 }, { "epoch": 2.87, "learning_rate": 1.6746416482116703e-05, "loss": 0.7022, "step": 9314 }, { "epoch": 2.87, "learning_rate": 1.6745680857120757e-05, "loss": 0.7277, "step": 9315 }, { "epoch": 2.87, "learning_rate": 1.6744945165133982e-05, "loss": 0.9769, "step": 9316 }, { "epoch": 2.87, "learning_rate": 1.674420940616369e-05, "loss": 0.5282, "step": 9317 }, { "epoch": 2.87, "learning_rate": 1.674347358021718e-05, "loss": 0.9747, "step": 9318 }, { "epoch": 2.87, "learning_rate": 1.6742737687301757e-05, "loss": 0.9101, "step": 9319 }, { "epoch": 2.87, "learning_rate": 1.6742001727424742e-05, "loss": 0.9555, "step": 9320 }, { "epoch": 2.87, "learning_rate": 1.6741265700593432e-05, "loss": 0.6879, "step": 9321 }, { "epoch": 2.87, "learning_rate": 1.674052960681514e-05, "loss": 0.6224, "step": 9322 }, { "epoch": 2.87, "learning_rate": 1.673979344609718e-05, "loss": 0.9166, "step": 9323 }, { "epoch": 2.87, "learning_rate": 1.673905721844686e-05, "loss": 1.1295, "step": 9324 }, { "epoch": 2.87, "learning_rate": 1.6738320923871487e-05, "loss": 0.9419, "step": 9325 }, { "epoch": 2.87, "learning_rate": 1.673758456237838e-05, "loss": 0.9524, "step": 9326 }, { "epoch": 2.87, "learning_rate": 1.673684813397485e-05, "loss": 0.7964, "step": 9327 }, { "epoch": 2.87, "learning_rate": 1.6736111638668203e-05, "loss": 0.8564, "step": 9328 }, { "epoch": 2.87, "learning_rate": 1.6735375076465764e-05, "loss": 1.0592, "step": 9329 }, { "epoch": 2.87, "learning_rate": 1.6734638447374845e-05, "loss": 0.9497, "step": 9330 }, { "epoch": 2.87, "learning_rate": 1.6733901751402757e-05, "loss": 1.1028, "step": 9331 }, { "epoch": 2.87, "learning_rate": 1.6733164988556817e-05, "loss": 1.0159, "step": 9332 }, { "epoch": 2.87, "learning_rate": 1.6732428158844345e-05, "loss": 0.9868, "step": 9333 }, { "epoch": 2.87, "learning_rate": 1.6731691262272652e-05, "loss": 0.6683, "step": 9334 }, { "epoch": 2.87, "learning_rate": 1.673095429884907e-05, "loss": 0.7295, "step": 9335 }, { "epoch": 2.87, "learning_rate": 1.67302172685809e-05, "loss": 1.1093, "step": 9336 }, { "epoch": 2.87, "learning_rate": 1.6729480171475472e-05, "loss": 1.0969, "step": 9337 }, { "epoch": 2.87, "learning_rate": 1.6728743007540103e-05, "loss": 1.1776, "step": 9338 }, { "epoch": 2.87, "learning_rate": 1.672800577678212e-05, "loss": 0.8253, "step": 9339 }, { "epoch": 2.87, "learning_rate": 1.6727268479208833e-05, "loss": 0.7696, "step": 9340 }, { "epoch": 2.87, "learning_rate": 1.6726531114827572e-05, "loss": 0.7644, "step": 9341 }, { "epoch": 2.87, "learning_rate": 1.6725793683645658e-05, "loss": 1.1054, "step": 9342 }, { "epoch": 2.87, "learning_rate": 1.6725056185670415e-05, "loss": 0.8121, "step": 9343 }, { "epoch": 2.87, "learning_rate": 1.6724318620909165e-05, "loss": 0.8366, "step": 9344 }, { "epoch": 2.88, "learning_rate": 1.6723580989369233e-05, "loss": 0.6349, "step": 9345 }, { "epoch": 2.88, "learning_rate": 1.6722843291057945e-05, "loss": 0.6909, "step": 9346 }, { "epoch": 2.88, "learning_rate": 1.672210552598263e-05, "loss": 0.9038, "step": 9347 }, { "epoch": 2.88, "learning_rate": 1.6721367694150607e-05, "loss": 0.6311, "step": 9348 }, { "epoch": 2.88, "learning_rate": 1.672062979556921e-05, "loss": 0.8455, "step": 9349 }, { "epoch": 2.88, "learning_rate": 1.671989183024577e-05, "loss": 0.7015, "step": 9350 }, { "epoch": 2.88, "learning_rate": 1.6719153798187605e-05, "loss": 0.8318, "step": 9351 }, { "epoch": 2.88, "learning_rate": 1.6718415699402054e-05, "loss": 1.1468, "step": 9352 }, { "epoch": 2.88, "learning_rate": 1.6717677533896442e-05, "loss": 0.6134, "step": 9353 }, { "epoch": 2.88, "learning_rate": 1.6716939301678098e-05, "loss": 0.9482, "step": 9354 }, { "epoch": 2.88, "learning_rate": 1.671620100275436e-05, "loss": 0.7411, "step": 9355 }, { "epoch": 2.88, "learning_rate": 1.6715462637132554e-05, "loss": 1.1548, "step": 9356 }, { "epoch": 2.88, "learning_rate": 1.6714724204820017e-05, "loss": 0.8551, "step": 9357 }, { "epoch": 2.88, "learning_rate": 1.671398570582408e-05, "loss": 1.0047, "step": 9358 }, { "epoch": 2.88, "learning_rate": 1.6713247140152074e-05, "loss": 0.5401, "step": 9359 }, { "epoch": 2.88, "learning_rate": 1.671250850781134e-05, "loss": 0.6278, "step": 9360 }, { "epoch": 2.88, "learning_rate": 1.671176980880921e-05, "loss": 0.8544, "step": 9361 }, { "epoch": 2.88, "learning_rate": 1.671103104315302e-05, "loss": 0.76, "step": 9362 }, { "epoch": 2.88, "learning_rate": 1.6710292210850102e-05, "loss": 0.9041, "step": 9363 }, { "epoch": 2.88, "learning_rate": 1.6709553311907802e-05, "loss": 0.8291, "step": 9364 }, { "epoch": 2.88, "learning_rate": 1.6708814346333456e-05, "loss": 0.9496, "step": 9365 }, { "epoch": 2.88, "learning_rate": 1.6708075314134395e-05, "loss": 0.8505, "step": 9366 }, { "epoch": 2.88, "learning_rate": 1.6707336215317968e-05, "loss": 0.6679, "step": 9367 }, { "epoch": 2.88, "learning_rate": 1.670659704989151e-05, "loss": 0.7448, "step": 9368 }, { "epoch": 2.88, "learning_rate": 1.670585781786236e-05, "loss": 0.8544, "step": 9369 }, { "epoch": 2.88, "learning_rate": 1.6705118519237867e-05, "loss": 0.9716, "step": 9370 }, { "epoch": 2.88, "learning_rate": 1.6704379154025364e-05, "loss": 1.3415, "step": 9371 }, { "epoch": 2.88, "learning_rate": 1.6703639722232194e-05, "loss": 0.8946, "step": 9372 }, { "epoch": 2.88, "learning_rate": 1.6702900223865704e-05, "loss": 0.6413, "step": 9373 }, { "epoch": 2.88, "learning_rate": 1.6702160658933243e-05, "loss": 0.8148, "step": 9374 }, { "epoch": 2.88, "learning_rate": 1.6701421027442146e-05, "loss": 0.6532, "step": 9375 }, { "epoch": 2.88, "learning_rate": 1.6700681329399762e-05, "loss": 0.817, "step": 9376 }, { "epoch": 2.89, "learning_rate": 1.669994156481344e-05, "loss": 0.8316, "step": 9377 }, { "epoch": 2.89, "learning_rate": 1.6699201733690517e-05, "loss": 0.9517, "step": 9378 }, { "epoch": 2.89, "learning_rate": 1.669846183603835e-05, "loss": 0.8553, "step": 9379 }, { "epoch": 2.89, "learning_rate": 1.6697721871864286e-05, "loss": 0.7632, "step": 9380 }, { "epoch": 2.89, "learning_rate": 1.6696981841175668e-05, "loss": 0.7406, "step": 9381 }, { "epoch": 2.89, "learning_rate": 1.669624174397985e-05, "loss": 0.9346, "step": 9382 }, { "epoch": 2.89, "learning_rate": 1.669550158028418e-05, "loss": 0.8812, "step": 9383 }, { "epoch": 2.89, "learning_rate": 1.6694761350096005e-05, "loss": 0.7552, "step": 9384 }, { "epoch": 2.89, "learning_rate": 1.6694021053422684e-05, "loss": 0.8788, "step": 9385 }, { "epoch": 2.89, "learning_rate": 1.669328069027156e-05, "loss": 1.1147, "step": 9386 }, { "epoch": 2.89, "learning_rate": 1.669254026064999e-05, "loss": 0.9345, "step": 9387 }, { "epoch": 2.89, "learning_rate": 1.6691799764565334e-05, "loss": 0.7871, "step": 9388 }, { "epoch": 2.89, "learning_rate": 1.6691059202024935e-05, "loss": 1.2248, "step": 9389 }, { "epoch": 2.89, "learning_rate": 1.669031857303615e-05, "loss": 0.8976, "step": 9390 }, { "epoch": 2.89, "learning_rate": 1.6689577877606338e-05, "loss": 0.6871, "step": 9391 }, { "epoch": 2.89, "learning_rate": 1.668883711574285e-05, "loss": 0.9797, "step": 9392 }, { "epoch": 2.89, "learning_rate": 1.6688096287453048e-05, "loss": 0.8418, "step": 9393 }, { "epoch": 2.89, "learning_rate": 1.668735539274428e-05, "loss": 0.7438, "step": 9394 }, { "epoch": 2.89, "learning_rate": 1.6686614431623916e-05, "loss": 0.8503, "step": 9395 }, { "epoch": 2.89, "learning_rate": 1.6685873404099306e-05, "loss": 0.8676, "step": 9396 }, { "epoch": 2.89, "learning_rate": 1.6685132310177812e-05, "loss": 0.7274, "step": 9397 }, { "epoch": 2.89, "learning_rate": 1.668439114986679e-05, "loss": 0.7117, "step": 9398 }, { "epoch": 2.89, "learning_rate": 1.6683649923173608e-05, "loss": 0.7258, "step": 9399 }, { "epoch": 2.89, "learning_rate": 1.6682908630105622e-05, "loss": 0.8187, "step": 9400 }, { "epoch": 2.89, "learning_rate": 1.668216727067019e-05, "loss": 0.7709, "step": 9401 }, { "epoch": 2.89, "learning_rate": 1.6681425844874678e-05, "loss": 0.8203, "step": 9402 }, { "epoch": 2.89, "learning_rate": 1.6680684352726455e-05, "loss": 1.0577, "step": 9403 }, { "epoch": 2.89, "learning_rate": 1.6679942794232875e-05, "loss": 0.9275, "step": 9404 }, { "epoch": 2.89, "learning_rate": 1.6679201169401308e-05, "loss": 0.8557, "step": 9405 }, { "epoch": 2.89, "learning_rate": 1.6678459478239116e-05, "loss": 0.9674, "step": 9406 }, { "epoch": 2.89, "learning_rate": 1.667771772075367e-05, "loss": 0.7252, "step": 9407 }, { "epoch": 2.89, "learning_rate": 1.667697589695233e-05, "loss": 0.9331, "step": 9408 }, { "epoch": 2.89, "learning_rate": 1.6676234006842465e-05, "loss": 1.0183, "step": 9409 }, { "epoch": 2.9, "learning_rate": 1.6675492050431445e-05, "loss": 0.7933, "step": 9410 }, { "epoch": 2.9, "learning_rate": 1.667475002772664e-05, "loss": 0.714, "step": 9411 }, { "epoch": 2.9, "learning_rate": 1.667400793873541e-05, "loss": 0.9634, "step": 9412 }, { "epoch": 2.9, "learning_rate": 1.667326578346513e-05, "loss": 1.0573, "step": 9413 }, { "epoch": 2.9, "learning_rate": 1.6672523561923175e-05, "loss": 0.7725, "step": 9414 }, { "epoch": 2.9, "learning_rate": 1.667178127411691e-05, "loss": 0.9727, "step": 9415 }, { "epoch": 2.9, "learning_rate": 1.6671038920053707e-05, "loss": 0.8259, "step": 9416 }, { "epoch": 2.9, "learning_rate": 1.667029649974094e-05, "loss": 1.0504, "step": 9417 }, { "epoch": 2.9, "learning_rate": 1.6669554013185977e-05, "loss": 0.7738, "step": 9418 }, { "epoch": 2.9, "learning_rate": 1.6668811460396202e-05, "loss": 0.7476, "step": 9419 }, { "epoch": 2.9, "learning_rate": 1.6668068841378978e-05, "loss": 0.7427, "step": 9420 }, { "epoch": 2.9, "learning_rate": 1.666732615614169e-05, "loss": 1.0425, "step": 9421 }, { "epoch": 2.9, "learning_rate": 1.6666583404691706e-05, "loss": 0.8021, "step": 9422 }, { "epoch": 2.9, "learning_rate": 1.6665840587036403e-05, "loss": 0.8363, "step": 9423 }, { "epoch": 2.9, "learning_rate": 1.666509770318316e-05, "loss": 0.6719, "step": 9424 }, { "epoch": 2.9, "learning_rate": 1.666435475313936e-05, "loss": 0.7058, "step": 9425 }, { "epoch": 2.9, "learning_rate": 1.666361173691237e-05, "loss": 0.9906, "step": 9426 }, { "epoch": 2.9, "learning_rate": 1.6662868654509573e-05, "loss": 1.0361, "step": 9427 }, { "epoch": 2.9, "learning_rate": 1.6662125505938352e-05, "loss": 0.6492, "step": 9428 }, { "epoch": 2.9, "learning_rate": 1.6661382291206082e-05, "loss": 0.7241, "step": 9429 }, { "epoch": 2.9, "learning_rate": 1.666063901032015e-05, "loss": 0.8449, "step": 9430 }, { "epoch": 2.9, "learning_rate": 1.6659895663287932e-05, "loss": 0.8231, "step": 9431 }, { "epoch": 2.9, "learning_rate": 1.665915225011681e-05, "loss": 0.8632, "step": 9432 }, { "epoch": 2.9, "learning_rate": 1.6658408770814173e-05, "loss": 0.7646, "step": 9433 }, { "epoch": 2.9, "learning_rate": 1.66576652253874e-05, "loss": 0.927, "step": 9434 }, { "epoch": 2.9, "learning_rate": 1.6656921613843873e-05, "loss": 0.6454, "step": 9435 }, { "epoch": 2.9, "learning_rate": 1.6656177936190983e-05, "loss": 0.7169, "step": 9436 }, { "epoch": 2.9, "learning_rate": 1.6655434192436107e-05, "loss": 0.8854, "step": 9437 }, { "epoch": 2.9, "learning_rate": 1.6654690382586636e-05, "loss": 0.6684, "step": 9438 }, { "epoch": 2.9, "learning_rate": 1.665394650664996e-05, "loss": 1.1533, "step": 9439 }, { "epoch": 2.9, "learning_rate": 1.665320256463346e-05, "loss": 0.9178, "step": 9440 }, { "epoch": 2.9, "learning_rate": 1.6652458556544528e-05, "loss": 0.9331, "step": 9441 }, { "epoch": 2.91, "learning_rate": 1.6651714482390552e-05, "loss": 0.9286, "step": 9442 }, { "epoch": 2.91, "learning_rate": 1.665097034217892e-05, "loss": 0.9965, "step": 9443 }, { "epoch": 2.91, "learning_rate": 1.665022613591702e-05, "loss": 0.9522, "step": 9444 }, { "epoch": 2.91, "learning_rate": 1.664948186361225e-05, "loss": 0.8432, "step": 9445 }, { "epoch": 2.91, "learning_rate": 1.6648737525271995e-05, "loss": 1.6491, "step": 9446 }, { "epoch": 2.91, "learning_rate": 1.6647993120903648e-05, "loss": 0.6237, "step": 9447 }, { "epoch": 2.91, "learning_rate": 1.6647248650514602e-05, "loss": 1.0573, "step": 9448 }, { "epoch": 2.91, "learning_rate": 1.6646504114112253e-05, "loss": 1.0023, "step": 9449 }, { "epoch": 2.91, "learning_rate": 1.6645759511703988e-05, "loss": 0.8771, "step": 9450 }, { "epoch": 2.91, "learning_rate": 1.6645014843297212e-05, "loss": 1.1197, "step": 9451 }, { "epoch": 2.91, "learning_rate": 1.6644270108899313e-05, "loss": 0.7851, "step": 9452 }, { "epoch": 2.91, "learning_rate": 1.6643525308517686e-05, "loss": 0.7831, "step": 9453 }, { "epoch": 2.91, "learning_rate": 1.6642780442159728e-05, "loss": 0.6091, "step": 9454 }, { "epoch": 2.91, "learning_rate": 1.664203550983284e-05, "loss": 0.6972, "step": 9455 }, { "epoch": 2.91, "learning_rate": 1.6641290511544422e-05, "loss": 0.8469, "step": 9456 }, { "epoch": 2.91, "learning_rate": 1.664054544730186e-05, "loss": 0.8949, "step": 9457 }, { "epoch": 2.91, "learning_rate": 1.663980031711257e-05, "loss": 0.7297, "step": 9458 }, { "epoch": 2.91, "learning_rate": 1.6639055120983937e-05, "loss": 1.1514, "step": 9459 }, { "epoch": 2.91, "learning_rate": 1.6638309858923373e-05, "loss": 0.9395, "step": 9460 }, { "epoch": 2.91, "learning_rate": 1.6637564530938272e-05, "loss": 0.8745, "step": 9461 }, { "epoch": 2.91, "learning_rate": 1.6636819137036038e-05, "loss": 0.6273, "step": 9462 }, { "epoch": 2.91, "learning_rate": 1.6636073677224072e-05, "loss": 1.1665, "step": 9463 }, { "epoch": 2.91, "learning_rate": 1.663532815150978e-05, "loss": 1.0412, "step": 9464 }, { "epoch": 2.91, "learning_rate": 1.6634582559900564e-05, "loss": 0.8284, "step": 9465 }, { "epoch": 2.91, "learning_rate": 1.6633836902403828e-05, "loss": 0.7829, "step": 9466 }, { "epoch": 2.91, "learning_rate": 1.663309117902698e-05, "loss": 0.827, "step": 9467 }, { "epoch": 2.91, "learning_rate": 1.6632345389777416e-05, "loss": 0.6597, "step": 9468 }, { "epoch": 2.91, "learning_rate": 1.6631599534662557e-05, "loss": 0.9424, "step": 9469 }, { "epoch": 2.91, "learning_rate": 1.6630853613689802e-05, "loss": 0.8434, "step": 9470 }, { "epoch": 2.91, "learning_rate": 1.6630107626866558e-05, "loss": 0.8386, "step": 9471 }, { "epoch": 2.91, "learning_rate": 1.6629361574200237e-05, "loss": 0.877, "step": 9472 }, { "epoch": 2.91, "learning_rate": 1.662861545569824e-05, "loss": 0.9615, "step": 9473 }, { "epoch": 2.91, "learning_rate": 1.6627869271367986e-05, "loss": 0.8644, "step": 9474 }, { "epoch": 2.92, "learning_rate": 1.6627123021216883e-05, "loss": 0.9223, "step": 9475 }, { "epoch": 2.92, "learning_rate": 1.662637670525234e-05, "loss": 0.81, "step": 9476 }, { "epoch": 2.92, "learning_rate": 1.6625630323481768e-05, "loss": 0.6388, "step": 9477 }, { "epoch": 2.92, "learning_rate": 1.662488387591258e-05, "loss": 0.9962, "step": 9478 }, { "epoch": 2.92, "learning_rate": 1.6624137362552192e-05, "loss": 0.7453, "step": 9479 }, { "epoch": 2.92, "learning_rate": 1.6623390783408013e-05, "loss": 0.812, "step": 9480 }, { "epoch": 2.92, "learning_rate": 1.662264413848746e-05, "loss": 0.7127, "step": 9481 }, { "epoch": 2.92, "learning_rate": 1.6621897427797947e-05, "loss": 1.0042, "step": 9482 }, { "epoch": 2.92, "learning_rate": 1.6621150651346892e-05, "loss": 0.9495, "step": 9483 }, { "epoch": 2.92, "learning_rate": 1.6620403809141707e-05, "loss": 0.7388, "step": 9484 }, { "epoch": 2.92, "learning_rate": 1.6619656901189807e-05, "loss": 0.7652, "step": 9485 }, { "epoch": 2.92, "learning_rate": 1.6618909927498616e-05, "loss": 0.6843, "step": 9486 }, { "epoch": 2.92, "learning_rate": 1.661816288807555e-05, "loss": 0.6879, "step": 9487 }, { "epoch": 2.92, "learning_rate": 1.661741578292803e-05, "loss": 0.6549, "step": 9488 }, { "epoch": 2.92, "learning_rate": 1.6616668612063465e-05, "loss": 0.8267, "step": 9489 }, { "epoch": 2.92, "learning_rate": 1.6615921375489284e-05, "loss": 0.8839, "step": 9490 }, { "epoch": 2.92, "learning_rate": 1.6615174073212915e-05, "loss": 0.8035, "step": 9491 }, { "epoch": 2.92, "learning_rate": 1.6614426705241766e-05, "loss": 0.9093, "step": 9492 }, { "epoch": 2.92, "learning_rate": 1.6613679271583258e-05, "loss": 0.8341, "step": 9493 }, { "epoch": 2.92, "learning_rate": 1.6612931772244823e-05, "loss": 0.5788, "step": 9494 }, { "epoch": 2.92, "learning_rate": 1.661218420723388e-05, "loss": 1.1103, "step": 9495 }, { "epoch": 2.92, "learning_rate": 1.661143657655786e-05, "loss": 0.9285, "step": 9496 }, { "epoch": 2.92, "learning_rate": 1.6610688880224178e-05, "loss": 0.7919, "step": 9497 }, { "epoch": 2.92, "learning_rate": 1.6609941118240257e-05, "loss": 0.8615, "step": 9498 }, { "epoch": 2.92, "learning_rate": 1.6609193290613536e-05, "loss": 0.767, "step": 9499 }, { "epoch": 2.92, "learning_rate": 1.6608445397351434e-05, "loss": 0.872, "step": 9500 }, { "epoch": 2.92, "learning_rate": 1.6607697438461376e-05, "loss": 0.8555, "step": 9501 }, { "epoch": 2.92, "learning_rate": 1.660694941395079e-05, "loss": 0.844, "step": 9502 }, { "epoch": 2.92, "learning_rate": 1.6606201323827113e-05, "loss": 0.7051, "step": 9503 }, { "epoch": 2.92, "learning_rate": 1.6605453168097768e-05, "loss": 0.7259, "step": 9504 }, { "epoch": 2.92, "learning_rate": 1.660470494677018e-05, "loss": 0.6718, "step": 9505 }, { "epoch": 2.92, "learning_rate": 1.660395665985179e-05, "loss": 0.6906, "step": 9506 }, { "epoch": 2.93, "learning_rate": 1.6603208307350024e-05, "loss": 0.8366, "step": 9507 }, { "epoch": 2.93, "learning_rate": 1.6602459889272312e-05, "loss": 0.6172, "step": 9508 }, { "epoch": 2.93, "learning_rate": 1.6601711405626085e-05, "loss": 0.9485, "step": 9509 }, { "epoch": 2.93, "learning_rate": 1.6600962856418782e-05, "loss": 0.7928, "step": 9510 }, { "epoch": 2.93, "learning_rate": 1.6600214241657835e-05, "loss": 0.9041, "step": 9511 }, { "epoch": 2.93, "learning_rate": 1.659946556135068e-05, "loss": 0.9496, "step": 9512 }, { "epoch": 2.93, "learning_rate": 1.6598716815504746e-05, "loss": 0.7819, "step": 9513 }, { "epoch": 2.93, "learning_rate": 1.6597968004127475e-05, "loss": 0.7392, "step": 9514 }, { "epoch": 2.93, "learning_rate": 1.65972191272263e-05, "loss": 0.9216, "step": 9515 }, { "epoch": 2.93, "learning_rate": 1.6596470184808654e-05, "loss": 0.7535, "step": 9516 }, { "epoch": 2.93, "learning_rate": 1.6595721176881986e-05, "loss": 0.7605, "step": 9517 }, { "epoch": 2.93, "learning_rate": 1.6594972103453727e-05, "loss": 0.8747, "step": 9518 }, { "epoch": 2.93, "learning_rate": 1.6594222964531316e-05, "loss": 0.8426, "step": 9519 }, { "epoch": 2.93, "learning_rate": 1.6593473760122192e-05, "loss": 0.8592, "step": 9520 }, { "epoch": 2.93, "learning_rate": 1.6592724490233796e-05, "loss": 0.7132, "step": 9521 }, { "epoch": 2.93, "learning_rate": 1.6591975154873576e-05, "loss": 0.6248, "step": 9522 }, { "epoch": 2.93, "learning_rate": 1.6591225754048963e-05, "loss": 0.7414, "step": 9523 }, { "epoch": 2.93, "learning_rate": 1.6590476287767403e-05, "loss": 0.6159, "step": 9524 }, { "epoch": 2.93, "learning_rate": 1.658972675603634e-05, "loss": 0.796, "step": 9525 }, { "epoch": 2.93, "learning_rate": 1.6588977158863215e-05, "loss": 0.8732, "step": 9526 }, { "epoch": 2.93, "learning_rate": 1.6588227496255477e-05, "loss": 0.7698, "step": 9527 }, { "epoch": 2.93, "learning_rate": 1.658747776822057e-05, "loss": 0.7409, "step": 9528 }, { "epoch": 2.93, "learning_rate": 1.6586727974765935e-05, "loss": 0.8756, "step": 9529 }, { "epoch": 2.93, "learning_rate": 1.658597811589902e-05, "loss": 0.9765, "step": 9530 }, { "epoch": 2.93, "learning_rate": 1.6585228191627274e-05, "loss": 0.8106, "step": 9531 }, { "epoch": 2.93, "learning_rate": 1.6584478201958144e-05, "loss": 0.9818, "step": 9532 }, { "epoch": 2.93, "learning_rate": 1.6583728146899076e-05, "loss": 0.8376, "step": 9533 }, { "epoch": 2.93, "learning_rate": 1.6582978026457518e-05, "loss": 0.8572, "step": 9534 }, { "epoch": 2.93, "learning_rate": 1.658222784064092e-05, "loss": 0.7882, "step": 9535 }, { "epoch": 2.93, "learning_rate": 1.6581477589456737e-05, "loss": 0.9742, "step": 9536 }, { "epoch": 2.93, "learning_rate": 1.6580727272912412e-05, "loss": 0.9197, "step": 9537 }, { "epoch": 2.93, "learning_rate": 1.6579976891015406e-05, "loss": 0.5138, "step": 9538 }, { "epoch": 2.93, "learning_rate": 1.657922644377316e-05, "loss": 0.8049, "step": 9539 }, { "epoch": 2.94, "learning_rate": 1.6578475931193133e-05, "loss": 0.795, "step": 9540 }, { "epoch": 2.94, "learning_rate": 1.657772535328278e-05, "loss": 0.9865, "step": 9541 }, { "epoch": 2.94, "learning_rate": 1.657697471004955e-05, "loss": 0.8959, "step": 9542 }, { "epoch": 2.94, "learning_rate": 1.65762240015009e-05, "loss": 0.962, "step": 9543 }, { "epoch": 2.94, "learning_rate": 1.6575473227644282e-05, "loss": 0.8929, "step": 9544 }, { "epoch": 2.94, "learning_rate": 1.6574722388487158e-05, "loss": 0.843, "step": 9545 }, { "epoch": 2.94, "learning_rate": 1.657397148403698e-05, "loss": 0.9874, "step": 9546 }, { "epoch": 2.94, "learning_rate": 1.6573220514301208e-05, "loss": 0.8105, "step": 9547 }, { "epoch": 2.94, "learning_rate": 1.6572469479287298e-05, "loss": 0.7736, "step": 9548 }, { "epoch": 2.94, "learning_rate": 1.6571718379002705e-05, "loss": 0.901, "step": 9549 }, { "epoch": 2.94, "learning_rate": 1.6570967213454895e-05, "loss": 0.9113, "step": 9550 }, { "epoch": 2.94, "learning_rate": 1.657021598265132e-05, "loss": 0.881, "step": 9551 }, { "epoch": 2.94, "learning_rate": 1.656946468659945e-05, "loss": 0.8515, "step": 9552 }, { "epoch": 2.94, "learning_rate": 1.6568713325306742e-05, "loss": 0.6346, "step": 9553 }, { "epoch": 2.94, "learning_rate": 1.6567961898780654e-05, "loss": 0.8106, "step": 9554 }, { "epoch": 2.94, "learning_rate": 1.6567210407028652e-05, "loss": 0.7027, "step": 9555 }, { "epoch": 2.94, "learning_rate": 1.65664588500582e-05, "loss": 0.5108, "step": 9556 }, { "epoch": 2.94, "learning_rate": 1.6565707227876758e-05, "loss": 0.9211, "step": 9557 }, { "epoch": 2.94, "learning_rate": 1.6564955540491793e-05, "loss": 0.7079, "step": 9558 }, { "epoch": 2.94, "learning_rate": 1.6564203787910767e-05, "loss": 0.7299, "step": 9559 }, { "epoch": 2.94, "learning_rate": 1.6563451970141147e-05, "loss": 0.7625, "step": 9560 }, { "epoch": 2.94, "learning_rate": 1.65627000871904e-05, "loss": 0.7924, "step": 9561 }, { "epoch": 2.94, "learning_rate": 1.6561948139065997e-05, "loss": 0.7647, "step": 9562 }, { "epoch": 2.94, "learning_rate": 1.65611961257754e-05, "loss": 0.5882, "step": 9563 }, { "epoch": 2.94, "learning_rate": 1.6560444047326077e-05, "loss": 0.8663, "step": 9564 }, { "epoch": 2.94, "learning_rate": 1.6559691903725498e-05, "loss": 0.9706, "step": 9565 }, { "epoch": 2.94, "learning_rate": 1.655893969498113e-05, "loss": 0.9001, "step": 9566 }, { "epoch": 2.94, "learning_rate": 1.6558187421100452e-05, "loss": 0.8752, "step": 9567 }, { "epoch": 2.94, "learning_rate": 1.6557435082090924e-05, "loss": 0.9974, "step": 9568 }, { "epoch": 2.94, "learning_rate": 1.6556682677960022e-05, "loss": 1.016, "step": 9569 }, { "epoch": 2.94, "learning_rate": 1.655593020871522e-05, "loss": 0.9993, "step": 9570 }, { "epoch": 2.94, "learning_rate": 1.655517767436399e-05, "loss": 0.5824, "step": 9571 }, { "epoch": 2.95, "learning_rate": 1.6554425074913803e-05, "loss": 1.0659, "step": 9572 }, { "epoch": 2.95, "learning_rate": 1.6553672410372136e-05, "loss": 0.9878, "step": 9573 }, { "epoch": 2.95, "learning_rate": 1.655291968074646e-05, "loss": 0.6985, "step": 9574 }, { "epoch": 2.95, "learning_rate": 1.6552166886044253e-05, "loss": 0.9082, "step": 9575 }, { "epoch": 2.95, "learning_rate": 1.655141402627299e-05, "loss": 1.034, "step": 9576 }, { "epoch": 2.95, "learning_rate": 1.6550661101440148e-05, "loss": 1.2934, "step": 9577 }, { "epoch": 2.95, "learning_rate": 1.6549908111553207e-05, "loss": 0.9704, "step": 9578 }, { "epoch": 2.95, "learning_rate": 1.6549155056619634e-05, "loss": 0.999, "step": 9579 }, { "epoch": 2.95, "learning_rate": 1.654840193664692e-05, "loss": 0.788, "step": 9580 }, { "epoch": 2.95, "learning_rate": 1.654764875164254e-05, "loss": 0.9124, "step": 9581 }, { "epoch": 2.95, "learning_rate": 1.6546895501613974e-05, "loss": 0.969, "step": 9582 }, { "epoch": 2.95, "learning_rate": 1.6546142186568702e-05, "loss": 0.7169, "step": 9583 }, { "epoch": 2.95, "learning_rate": 1.65453888065142e-05, "loss": 0.6497, "step": 9584 }, { "epoch": 2.95, "learning_rate": 1.654463536145796e-05, "loss": 0.6452, "step": 9585 }, { "epoch": 2.95, "learning_rate": 1.654388185140746e-05, "loss": 0.8116, "step": 9586 }, { "epoch": 2.95, "learning_rate": 1.654312827637018e-05, "loss": 0.9372, "step": 9587 }, { "epoch": 2.95, "learning_rate": 1.6542374636353605e-05, "loss": 1.1335, "step": 9588 }, { "epoch": 2.95, "learning_rate": 1.654162093136522e-05, "loss": 0.6381, "step": 9589 }, { "epoch": 2.95, "learning_rate": 1.6540867161412512e-05, "loss": 0.9104, "step": 9590 }, { "epoch": 2.95, "learning_rate": 1.6540113326502964e-05, "loss": 0.8242, "step": 9591 }, { "epoch": 2.95, "learning_rate": 1.6539359426644062e-05, "loss": 0.9812, "step": 9592 }, { "epoch": 2.95, "learning_rate": 1.6538605461843296e-05, "loss": 0.7098, "step": 9593 }, { "epoch": 2.95, "learning_rate": 1.653785143210815e-05, "loss": 0.6452, "step": 9594 }, { "epoch": 2.95, "learning_rate": 1.6537097337446115e-05, "loss": 0.6876, "step": 9595 }, { "epoch": 2.95, "learning_rate": 1.6536343177864678e-05, "loss": 0.8964, "step": 9596 }, { "epoch": 2.95, "learning_rate": 1.6535588953371332e-05, "loss": 0.8346, "step": 9597 }, { "epoch": 2.95, "learning_rate": 1.653483466397356e-05, "loss": 0.8169, "step": 9598 }, { "epoch": 2.95, "learning_rate": 1.653408030967886e-05, "loss": 0.8299, "step": 9599 }, { "epoch": 2.95, "learning_rate": 1.6533325890494717e-05, "loss": 0.8771, "step": 9600 }, { "epoch": 2.95, "learning_rate": 1.653257140642863e-05, "loss": 0.7034, "step": 9601 }, { "epoch": 2.95, "learning_rate": 1.653181685748809e-05, "loss": 0.645, "step": 9602 }, { "epoch": 2.95, "learning_rate": 1.6531062243680585e-05, "loss": 0.7578, "step": 9603 }, { "epoch": 2.95, "learning_rate": 1.6530307565013615e-05, "loss": 0.92, "step": 9604 }, { "epoch": 2.96, "learning_rate": 1.652955282149467e-05, "loss": 0.7689, "step": 9605 }, { "epoch": 2.96, "learning_rate": 1.6528798013131248e-05, "loss": 0.719, "step": 9606 }, { "epoch": 2.96, "learning_rate": 1.6528043139930845e-05, "loss": 0.5003, "step": 9607 }, { "epoch": 2.96, "learning_rate": 1.6527288201900956e-05, "loss": 0.8994, "step": 9608 }, { "epoch": 2.96, "learning_rate": 1.6526533199049085e-05, "loss": 0.6691, "step": 9609 }, { "epoch": 2.96, "learning_rate": 1.6525778131382718e-05, "loss": 1.032, "step": 9610 }, { "epoch": 2.96, "learning_rate": 1.6525022998909365e-05, "loss": 0.8805, "step": 9611 }, { "epoch": 2.96, "learning_rate": 1.6524267801636515e-05, "loss": 0.637, "step": 9612 }, { "epoch": 2.96, "learning_rate": 1.6523512539571675e-05, "loss": 0.7767, "step": 9613 }, { "epoch": 2.96, "learning_rate": 1.6522757212722346e-05, "loss": 0.6725, "step": 9614 }, { "epoch": 2.96, "learning_rate": 1.652200182109602e-05, "loss": 0.9157, "step": 9615 }, { "epoch": 2.96, "learning_rate": 1.652124636470021e-05, "loss": 0.6554, "step": 9616 }, { "epoch": 2.96, "learning_rate": 1.652049084354241e-05, "loss": 0.8414, "step": 9617 }, { "epoch": 2.96, "learning_rate": 1.6519735257630132e-05, "loss": 0.8814, "step": 9618 }, { "epoch": 2.96, "learning_rate": 1.651897960697087e-05, "loss": 0.8447, "step": 9619 }, { "epoch": 2.96, "learning_rate": 1.6518223891572135e-05, "loss": 0.8561, "step": 9620 }, { "epoch": 2.96, "learning_rate": 1.6517468111441423e-05, "loss": 0.9162, "step": 9621 }, { "epoch": 2.96, "learning_rate": 1.6516712266586253e-05, "loss": 1.0404, "step": 9622 }, { "epoch": 2.96, "learning_rate": 1.651595635701412e-05, "loss": 0.8926, "step": 9623 }, { "epoch": 2.96, "learning_rate": 1.6515200382732536e-05, "loss": 0.6386, "step": 9624 }, { "epoch": 2.96, "learning_rate": 1.651444434374901e-05, "loss": 0.9316, "step": 9625 }, { "epoch": 2.96, "learning_rate": 1.6513688240071046e-05, "loss": 1.1182, "step": 9626 }, { "epoch": 2.96, "learning_rate": 1.6512932071706153e-05, "loss": 0.8892, "step": 9627 }, { "epoch": 2.96, "learning_rate": 1.6512175838661845e-05, "loss": 1.0275, "step": 9628 }, { "epoch": 2.96, "learning_rate": 1.651141954094563e-05, "loss": 0.8335, "step": 9629 }, { "epoch": 2.96, "learning_rate": 1.6510663178565014e-05, "loss": 0.7934, "step": 9630 }, { "epoch": 2.96, "learning_rate": 1.6509906751527515e-05, "loss": 0.7581, "step": 9631 }, { "epoch": 2.96, "learning_rate": 1.650915025984064e-05, "loss": 0.9284, "step": 9632 }, { "epoch": 2.96, "learning_rate": 1.650839370351191e-05, "loss": 0.75, "step": 9633 }, { "epoch": 2.96, "learning_rate": 1.6507637082548825e-05, "loss": 0.765, "step": 9634 }, { "epoch": 2.96, "learning_rate": 1.6506880396958906e-05, "loss": 0.9165, "step": 9635 }, { "epoch": 2.96, "learning_rate": 1.650612364674967e-05, "loss": 0.918, "step": 9636 }, { "epoch": 2.97, "learning_rate": 1.6505366831928632e-05, "loss": 0.8902, "step": 9637 }, { "epoch": 2.97, "learning_rate": 1.6504609952503307e-05, "loss": 0.7593, "step": 9638 }, { "epoch": 2.97, "learning_rate": 1.6503853008481203e-05, "loss": 0.6107, "step": 9639 }, { "epoch": 2.97, "learning_rate": 1.650309599986985e-05, "loss": 0.899, "step": 9640 }, { "epoch": 2.97, "learning_rate": 1.650233892667676e-05, "loss": 0.8673, "step": 9641 }, { "epoch": 2.97, "learning_rate": 1.6501581788909448e-05, "loss": 0.8375, "step": 9642 }, { "epoch": 2.97, "learning_rate": 1.650082458657544e-05, "loss": 0.6672, "step": 9643 }, { "epoch": 2.97, "learning_rate": 1.6500067319682257e-05, "loss": 0.9799, "step": 9644 }, { "epoch": 2.97, "learning_rate": 1.6499309988237406e-05, "loss": 0.836, "step": 9645 }, { "epoch": 2.97, "learning_rate": 1.6498552592248422e-05, "loss": 0.8723, "step": 9646 }, { "epoch": 2.97, "learning_rate": 1.6497795131722818e-05, "loss": 0.8176, "step": 9647 }, { "epoch": 2.97, "learning_rate": 1.6497037606668124e-05, "loss": 1.0019, "step": 9648 }, { "epoch": 2.97, "learning_rate": 1.6496280017091856e-05, "loss": 0.9124, "step": 9649 }, { "epoch": 2.97, "learning_rate": 1.6495522363001545e-05, "loss": 0.9209, "step": 9650 }, { "epoch": 2.97, "learning_rate": 1.6494764644404707e-05, "loss": 0.871, "step": 9651 }, { "epoch": 2.97, "learning_rate": 1.649400686130887e-05, "loss": 0.969, "step": 9652 }, { "epoch": 2.97, "learning_rate": 1.6493249013721558e-05, "loss": 0.6634, "step": 9653 }, { "epoch": 2.97, "learning_rate": 1.64924911016503e-05, "loss": 0.9088, "step": 9654 }, { "epoch": 2.97, "learning_rate": 1.6491733125102623e-05, "loss": 0.9003, "step": 9655 }, { "epoch": 2.97, "learning_rate": 1.6490975084086055e-05, "loss": 0.9623, "step": 9656 }, { "epoch": 2.97, "learning_rate": 1.649021697860812e-05, "loss": 1.0032, "step": 9657 }, { "epoch": 2.97, "learning_rate": 1.6489458808676348e-05, "loss": 0.8802, "step": 9658 }, { "epoch": 2.97, "learning_rate": 1.648870057429827e-05, "loss": 0.9031, "step": 9659 }, { "epoch": 2.97, "learning_rate": 1.6487942275481413e-05, "loss": 0.9255, "step": 9660 }, { "epoch": 2.97, "learning_rate": 1.6487183912233312e-05, "loss": 0.9575, "step": 9661 }, { "epoch": 2.97, "learning_rate": 1.6486425484561498e-05, "loss": 0.5193, "step": 9662 }, { "epoch": 2.97, "learning_rate": 1.6485666992473496e-05, "loss": 0.7303, "step": 9663 }, { "epoch": 2.97, "learning_rate": 1.6484908435976847e-05, "loss": 0.7478, "step": 9664 }, { "epoch": 2.97, "learning_rate": 1.6484149815079078e-05, "loss": 0.6551, "step": 9665 }, { "epoch": 2.97, "learning_rate": 1.6483391129787725e-05, "loss": 0.5613, "step": 9666 }, { "epoch": 2.97, "learning_rate": 1.648263238011033e-05, "loss": 0.89, "step": 9667 }, { "epoch": 2.97, "learning_rate": 1.6481873566054414e-05, "loss": 0.7147, "step": 9668 }, { "epoch": 2.97, "learning_rate": 1.6481114687627522e-05, "loss": 0.9021, "step": 9669 }, { "epoch": 2.98, "learning_rate": 1.6480355744837188e-05, "loss": 0.8121, "step": 9670 }, { "epoch": 2.98, "learning_rate": 1.6479596737690948e-05, "loss": 0.7227, "step": 9671 }, { "epoch": 2.98, "learning_rate": 1.647883766619634e-05, "loss": 0.9018, "step": 9672 }, { "epoch": 2.98, "learning_rate": 1.6478078530360905e-05, "loss": 0.8449, "step": 9673 }, { "epoch": 2.98, "learning_rate": 1.6477319330192177e-05, "loss": 0.9567, "step": 9674 }, { "epoch": 2.98, "learning_rate": 1.64765600656977e-05, "loss": 0.547, "step": 9675 }, { "epoch": 2.98, "learning_rate": 1.6475800736885012e-05, "loss": 0.5766, "step": 9676 }, { "epoch": 2.98, "learning_rate": 1.6475041343761657e-05, "loss": 0.9228, "step": 9677 }, { "epoch": 2.98, "learning_rate": 1.647428188633517e-05, "loss": 0.8965, "step": 9678 }, { "epoch": 2.98, "learning_rate": 1.64735223646131e-05, "loss": 1.0057, "step": 9679 }, { "epoch": 2.98, "learning_rate": 1.6472762778602984e-05, "loss": 0.9876, "step": 9680 }, { "epoch": 2.98, "learning_rate": 1.647200312831237e-05, "loss": 0.6143, "step": 9681 }, { "epoch": 2.98, "learning_rate": 1.6471243413748798e-05, "loss": 0.6735, "step": 9682 }, { "epoch": 2.98, "learning_rate": 1.647048363491982e-05, "loss": 1.045, "step": 9683 }, { "epoch": 2.98, "learning_rate": 1.646972379183297e-05, "loss": 0.7429, "step": 9684 }, { "epoch": 2.98, "learning_rate": 1.6468963884495805e-05, "loss": 0.8232, "step": 9685 }, { "epoch": 2.98, "learning_rate": 1.6468203912915862e-05, "loss": 0.8062, "step": 9686 }, { "epoch": 2.98, "learning_rate": 1.6467443877100694e-05, "loss": 0.8644, "step": 9687 }, { "epoch": 2.98, "learning_rate": 1.646668377705785e-05, "loss": 0.8466, "step": 9688 }, { "epoch": 2.98, "learning_rate": 1.646592361279488e-05, "loss": 0.8648, "step": 9689 }, { "epoch": 2.98, "learning_rate": 1.646516338431932e-05, "loss": 1.013, "step": 9690 }, { "epoch": 2.98, "learning_rate": 1.6464403091638735e-05, "loss": 0.9161, "step": 9691 }, { "epoch": 2.98, "learning_rate": 1.646364273476067e-05, "loss": 0.8758, "step": 9692 }, { "epoch": 2.98, "learning_rate": 1.6462882313692674e-05, "loss": 0.8452, "step": 9693 }, { "epoch": 2.98, "learning_rate": 1.64621218284423e-05, "loss": 0.8646, "step": 9694 }, { "epoch": 2.98, "learning_rate": 1.64613612790171e-05, "loss": 0.9182, "step": 9695 }, { "epoch": 2.98, "learning_rate": 1.6460600665424628e-05, "loss": 0.5317, "step": 9696 }, { "epoch": 2.98, "learning_rate": 1.6459839987672436e-05, "loss": 0.9137, "step": 9697 }, { "epoch": 2.98, "learning_rate": 1.6459079245768082e-05, "loss": 0.7701, "step": 9698 }, { "epoch": 2.98, "learning_rate": 1.645831843971912e-05, "loss": 0.8989, "step": 9699 }, { "epoch": 2.98, "learning_rate": 1.6457557569533097e-05, "loss": 1.3289, "step": 9700 }, { "epoch": 2.98, "learning_rate": 1.645679663521758e-05, "loss": 1.1318, "step": 9701 }, { "epoch": 2.99, "learning_rate": 1.6456035636780124e-05, "loss": 0.8504, "step": 9702 }, { "epoch": 2.99, "learning_rate": 1.645527457422828e-05, "loss": 1.01, "step": 9703 }, { "epoch": 2.99, "learning_rate": 1.6454513447569613e-05, "loss": 0.9495, "step": 9704 }, { "epoch": 2.99, "learning_rate": 1.6453752256811676e-05, "loss": 0.9521, "step": 9705 }, { "epoch": 2.99, "learning_rate": 1.6452991001962032e-05, "loss": 0.5862, "step": 9706 }, { "epoch": 2.99, "learning_rate": 1.645222968302824e-05, "loss": 0.6999, "step": 9707 }, { "epoch": 2.99, "learning_rate": 1.645146830001786e-05, "loss": 0.99, "step": 9708 }, { "epoch": 2.99, "learning_rate": 1.645070685293846e-05, "loss": 0.8326, "step": 9709 }, { "epoch": 2.99, "learning_rate": 1.6449945341797587e-05, "loss": 0.8974, "step": 9710 }, { "epoch": 2.99, "learning_rate": 1.6449183766602817e-05, "loss": 0.9798, "step": 9711 }, { "epoch": 2.99, "learning_rate": 1.6448422127361707e-05, "loss": 1.1746, "step": 9712 }, { "epoch": 2.99, "learning_rate": 1.6447660424081823e-05, "loss": 0.8109, "step": 9713 }, { "epoch": 2.99, "learning_rate": 1.6446898656770727e-05, "loss": 0.5517, "step": 9714 }, { "epoch": 2.99, "learning_rate": 1.644613682543599e-05, "loss": 0.6114, "step": 9715 }, { "epoch": 2.99, "learning_rate": 1.644537493008517e-05, "loss": 0.8215, "step": 9716 }, { "epoch": 2.99, "learning_rate": 1.6444612970725837e-05, "loss": 0.7317, "step": 9717 }, { "epoch": 2.99, "learning_rate": 1.644385094736556e-05, "loss": 0.7797, "step": 9718 }, { "epoch": 2.99, "learning_rate": 1.6443088860011903e-05, "loss": 0.6352, "step": 9719 }, { "epoch": 2.99, "learning_rate": 1.6442326708672434e-05, "loss": 0.9105, "step": 9720 }, { "epoch": 2.99, "learning_rate": 1.6441564493354723e-05, "loss": 0.9907, "step": 9721 }, { "epoch": 2.99, "learning_rate": 1.644080221406634e-05, "loss": 0.7477, "step": 9722 }, { "epoch": 2.99, "learning_rate": 1.6440039870814856e-05, "loss": 1.0607, "step": 9723 }, { "epoch": 2.99, "learning_rate": 1.643927746360784e-05, "loss": 0.8627, "step": 9724 }, { "epoch": 2.99, "learning_rate": 1.6438514992452868e-05, "loss": 0.7527, "step": 9725 }, { "epoch": 2.99, "learning_rate": 1.6437752457357502e-05, "loss": 0.7572, "step": 9726 }, { "epoch": 2.99, "learning_rate": 1.6436989858329326e-05, "loss": 0.8536, "step": 9727 }, { "epoch": 2.99, "learning_rate": 1.6436227195375906e-05, "loss": 0.7752, "step": 9728 }, { "epoch": 2.99, "learning_rate": 1.643546446850482e-05, "loss": 0.8681, "step": 9729 }, { "epoch": 2.99, "learning_rate": 1.6434701677723638e-05, "loss": 0.6591, "step": 9730 }, { "epoch": 2.99, "learning_rate": 1.6433938823039942e-05, "loss": 0.936, "step": 9731 }, { "epoch": 2.99, "learning_rate": 1.64331759044613e-05, "loss": 0.7919, "step": 9732 }, { "epoch": 2.99, "learning_rate": 1.643241292199529e-05, "loss": 1.1225, "step": 9733 }, { "epoch": 2.99, "learning_rate": 1.6431649875649496e-05, "loss": 1.5297, "step": 9734 }, { "epoch": 3.0, "learning_rate": 1.6430886765431488e-05, "loss": 0.763, "step": 9735 }, { "epoch": 3.0, "learning_rate": 1.6430123591348852e-05, "loss": 0.9651, "step": 9736 }, { "epoch": 3.0, "learning_rate": 1.642936035340916e-05, "loss": 0.8813, "step": 9737 }, { "epoch": 3.0, "learning_rate": 1.6428597051619993e-05, "loss": 0.8936, "step": 9738 }, { "epoch": 3.0, "learning_rate": 1.6427833685988934e-05, "loss": 0.8807, "step": 9739 }, { "epoch": 3.0, "learning_rate": 1.642707025652356e-05, "loss": 0.7774, "step": 9740 }, { "epoch": 3.0, "learning_rate": 1.642630676323146e-05, "loss": 1.2593, "step": 9741 }, { "epoch": 3.0, "learning_rate": 1.6425543206120207e-05, "loss": 0.8139, "step": 9742 }, { "epoch": 3.0, "learning_rate": 1.642477958519739e-05, "loss": 0.8764, "step": 9743 }, { "epoch": 3.0, "learning_rate": 1.642401590047059e-05, "loss": 0.8206, "step": 9744 }, { "epoch": 3.0, "learning_rate": 1.642325215194739e-05, "loss": 0.5526, "step": 9745 }, { "epoch": 3.0, "learning_rate": 1.6422488339635378e-05, "loss": 0.6893, "step": 9746 }, { "epoch": 3.0, "learning_rate": 1.6421724463542136e-05, "loss": 0.931, "step": 9747 }, { "epoch": 3.0, "learning_rate": 1.6420960523675255e-05, "loss": 0.8938, "step": 9748 }, { "epoch": 3.0, "learning_rate": 1.6420196520042317e-05, "loss": 0.9639, "step": 9749 }, { "epoch": 3.0, "learning_rate": 1.641943245265091e-05, "loss": 0.8994, "step": 9750 }, { "epoch": 3.0, "learning_rate": 1.641866832150862e-05, "loss": 0.7818, "step": 9751 }, { "epoch": 3.0, "learning_rate": 1.641790412662304e-05, "loss": 0.4248, "step": 9752 }, { "epoch": 3.0, "learning_rate": 1.6417139868001755e-05, "loss": 0.7231, "step": 9753 }, { "epoch": 3.0, "learning_rate": 1.6416375545652363e-05, "loss": 0.6096, "step": 9754 }, { "epoch": 3.0, "learning_rate": 1.641561115958244e-05, "loss": 0.5145, "step": 9755 }, { "epoch": 3.0, "learning_rate": 1.641484670979959e-05, "loss": 0.5868, "step": 9756 }, { "epoch": 3.0, "learning_rate": 1.6414082196311402e-05, "loss": 0.6741, "step": 9757 }, { "epoch": 3.0, "learning_rate": 1.6413317619125462e-05, "loss": 0.6248, "step": 9758 }, { "epoch": 3.0, "learning_rate": 1.641255297824937e-05, "loss": 0.4502, "step": 9759 }, { "epoch": 3.0, "learning_rate": 1.6411788273690716e-05, "loss": 0.3451, "step": 9760 }, { "epoch": 3.0, "learning_rate": 1.6411023505457095e-05, "loss": 0.5428, "step": 9761 }, { "epoch": 3.0, "learning_rate": 1.6410258673556107e-05, "loss": 0.4905, "step": 9762 }, { "epoch": 3.0, "learning_rate": 1.6409493777995334e-05, "loss": 0.5248, "step": 9763 }, { "epoch": 3.0, "learning_rate": 1.6408728818782387e-05, "loss": 0.443, "step": 9764 }, { "epoch": 3.0, "learning_rate": 1.6407963795924854e-05, "loss": 0.6557, "step": 9765 }, { "epoch": 3.0, "learning_rate": 1.6407198709430333e-05, "loss": 0.6131, "step": 9766 }, { "epoch": 3.01, "learning_rate": 1.6406433559306426e-05, "loss": 0.5989, "step": 9767 }, { "epoch": 3.01, "learning_rate": 1.6405668345560732e-05, "loss": 0.5639, "step": 9768 }, { "epoch": 3.01, "learning_rate": 1.6404903068200844e-05, "loss": 0.4864, "step": 9769 }, { "epoch": 3.01, "learning_rate": 1.6404137727234366e-05, "loss": 0.5548, "step": 9770 }, { "epoch": 3.01, "learning_rate": 1.6403372322668897e-05, "loss": 0.3737, "step": 9771 }, { "epoch": 3.01, "learning_rate": 1.640260685451204e-05, "loss": 0.5841, "step": 9772 }, { "epoch": 3.01, "learning_rate": 1.6401841322771396e-05, "loss": 0.586, "step": 9773 }, { "epoch": 3.01, "learning_rate": 1.6401075727454568e-05, "loss": 0.6469, "step": 9774 }, { "epoch": 3.01, "learning_rate": 1.640031006856916e-05, "loss": 0.5283, "step": 9775 }, { "epoch": 3.01, "learning_rate": 1.6399544346122777e-05, "loss": 0.4864, "step": 9776 }, { "epoch": 3.01, "learning_rate": 1.6398778560123013e-05, "loss": 0.3699, "step": 9777 }, { "epoch": 3.01, "learning_rate": 1.6398012710577485e-05, "loss": 0.3956, "step": 9778 }, { "epoch": 3.01, "learning_rate": 1.6397246797493797e-05, "loss": 0.6092, "step": 9779 }, { "epoch": 3.01, "learning_rate": 1.6396480820879548e-05, "loss": 0.578, "step": 9780 }, { "epoch": 3.01, "learning_rate": 1.639571478074235e-05, "loss": 0.5095, "step": 9781 }, { "epoch": 3.01, "learning_rate": 1.639494867708981e-05, "loss": 0.4961, "step": 9782 }, { "epoch": 3.01, "learning_rate": 1.639418250992954e-05, "loss": 0.5718, "step": 9783 }, { "epoch": 3.01, "learning_rate": 1.639341627926914e-05, "loss": 0.29, "step": 9784 }, { "epoch": 3.01, "learning_rate": 1.639264998511622e-05, "loss": 0.4634, "step": 9785 }, { "epoch": 3.01, "learning_rate": 1.63918836274784e-05, "loss": 0.5081, "step": 9786 }, { "epoch": 3.01, "learning_rate": 1.6391117206363284e-05, "loss": 0.8156, "step": 9787 }, { "epoch": 3.01, "learning_rate": 1.6390350721778484e-05, "loss": 0.7057, "step": 9788 }, { "epoch": 3.01, "learning_rate": 1.6389584173731614e-05, "loss": 0.6615, "step": 9789 }, { "epoch": 3.01, "learning_rate": 1.638881756223028e-05, "loss": 0.3316, "step": 9790 }, { "epoch": 3.01, "learning_rate": 1.63880508872821e-05, "loss": 0.1825, "step": 9791 }, { "epoch": 3.01, "learning_rate": 1.6387284148894688e-05, "loss": 0.3964, "step": 9792 }, { "epoch": 3.01, "learning_rate": 1.638651734707566e-05, "loss": 0.346, "step": 9793 }, { "epoch": 3.01, "learning_rate": 1.6385750481832627e-05, "loss": 0.4288, "step": 9794 }, { "epoch": 3.01, "learning_rate": 1.6384983553173206e-05, "loss": 0.4059, "step": 9795 }, { "epoch": 3.01, "learning_rate": 1.6384216561105014e-05, "loss": 0.628, "step": 9796 }, { "epoch": 3.01, "learning_rate": 1.6383449505635668e-05, "loss": 0.6036, "step": 9797 }, { "epoch": 3.01, "learning_rate": 1.6382682386772783e-05, "loss": 0.5983, "step": 9798 }, { "epoch": 3.01, "learning_rate": 1.638191520452398e-05, "loss": 0.603, "step": 9799 }, { "epoch": 3.02, "learning_rate": 1.6381147958896882e-05, "loss": 0.5792, "step": 9800 }, { "epoch": 3.02, "learning_rate": 1.63803806498991e-05, "loss": 0.4558, "step": 9801 }, { "epoch": 3.02, "learning_rate": 1.6379613277538257e-05, "loss": 0.3932, "step": 9802 }, { "epoch": 3.02, "learning_rate": 1.6378845841821976e-05, "loss": 0.4201, "step": 9803 }, { "epoch": 3.02, "learning_rate": 1.6378078342757877e-05, "loss": 0.4912, "step": 9804 }, { "epoch": 3.02, "learning_rate": 1.6377310780353582e-05, "loss": 0.535, "step": 9805 }, { "epoch": 3.02, "learning_rate": 1.6376543154616712e-05, "loss": 0.7205, "step": 9806 }, { "epoch": 3.02, "learning_rate": 1.6375775465554893e-05, "loss": 0.4702, "step": 9807 }, { "epoch": 3.02, "learning_rate": 1.6375007713175745e-05, "loss": 0.3792, "step": 9808 }, { "epoch": 3.02, "learning_rate": 1.63742398974869e-05, "loss": 0.4169, "step": 9809 }, { "epoch": 3.02, "learning_rate": 1.6373472018495972e-05, "loss": 0.3548, "step": 9810 }, { "epoch": 3.02, "learning_rate": 1.6372704076210598e-05, "loss": 0.6064, "step": 9811 }, { "epoch": 3.02, "learning_rate": 1.6371936070638397e-05, "loss": 0.4488, "step": 9812 }, { "epoch": 3.02, "learning_rate": 1.6371168001786998e-05, "loss": 0.4325, "step": 9813 }, { "epoch": 3.02, "learning_rate": 1.637039986966403e-05, "loss": 0.5186, "step": 9814 }, { "epoch": 3.02, "learning_rate": 1.636963167427712e-05, "loss": 0.5172, "step": 9815 }, { "epoch": 3.02, "learning_rate": 1.6368863415633897e-05, "loss": 0.6583, "step": 9816 }, { "epoch": 3.02, "learning_rate": 1.636809509374199e-05, "loss": 0.4724, "step": 9817 }, { "epoch": 3.02, "learning_rate": 1.636732670860903e-05, "loss": 0.6265, "step": 9818 }, { "epoch": 3.02, "learning_rate": 1.636655826024265e-05, "loss": 0.4357, "step": 9819 }, { "epoch": 3.02, "learning_rate": 1.6365789748650475e-05, "loss": 0.3899, "step": 9820 }, { "epoch": 3.02, "learning_rate": 1.6365021173840144e-05, "loss": 0.4978, "step": 9821 }, { "epoch": 3.02, "learning_rate": 1.6364252535819284e-05, "loss": 0.629, "step": 9822 }, { "epoch": 3.02, "learning_rate": 1.636348383459553e-05, "loss": 0.5559, "step": 9823 }, { "epoch": 3.02, "learning_rate": 1.6362715070176516e-05, "loss": 0.5796, "step": 9824 }, { "epoch": 3.02, "learning_rate": 1.6361946242569883e-05, "loss": 0.2552, "step": 9825 }, { "epoch": 3.02, "learning_rate": 1.6361177351783258e-05, "loss": 0.441, "step": 9826 }, { "epoch": 3.02, "learning_rate": 1.6360408397824276e-05, "loss": 0.4693, "step": 9827 }, { "epoch": 3.02, "learning_rate": 1.635963938070058e-05, "loss": 0.3902, "step": 9828 }, { "epoch": 3.02, "learning_rate": 1.6358870300419806e-05, "loss": 0.6769, "step": 9829 }, { "epoch": 3.02, "learning_rate": 1.6358101156989586e-05, "loss": 0.809, "step": 9830 }, { "epoch": 3.02, "learning_rate": 1.6357331950417562e-05, "loss": 0.6245, "step": 9831 }, { "epoch": 3.03, "learning_rate": 1.635656268071137e-05, "loss": 0.4116, "step": 9832 }, { "epoch": 3.03, "learning_rate": 1.635579334787866e-05, "loss": 0.4662, "step": 9833 }, { "epoch": 3.03, "learning_rate": 1.6355023951927056e-05, "loss": 0.5886, "step": 9834 }, { "epoch": 3.03, "learning_rate": 1.635425449286421e-05, "loss": 0.4511, "step": 9835 }, { "epoch": 3.03, "learning_rate": 1.635348497069776e-05, "loss": 0.6446, "step": 9836 }, { "epoch": 3.03, "learning_rate": 1.635271538543535e-05, "loss": 0.3888, "step": 9837 }, { "epoch": 3.03, "learning_rate": 1.6351945737084623e-05, "loss": 0.5968, "step": 9838 }, { "epoch": 3.03, "learning_rate": 1.635117602565322e-05, "loss": 0.4916, "step": 9839 }, { "epoch": 3.03, "learning_rate": 1.6350406251148784e-05, "loss": 0.3657, "step": 9840 }, { "epoch": 3.03, "learning_rate": 1.6349636413578963e-05, "loss": 0.4787, "step": 9841 }, { "epoch": 3.03, "learning_rate": 1.63488665129514e-05, "loss": 0.7731, "step": 9842 }, { "epoch": 3.03, "learning_rate": 1.634809654927374e-05, "loss": 0.4487, "step": 9843 }, { "epoch": 3.03, "learning_rate": 1.634732652255363e-05, "loss": 0.3454, "step": 9844 }, { "epoch": 3.03, "learning_rate": 1.6346556432798718e-05, "loss": 0.4438, "step": 9845 }, { "epoch": 3.03, "learning_rate": 1.6345786280016655e-05, "loss": 0.5596, "step": 9846 }, { "epoch": 3.03, "learning_rate": 1.634501606421508e-05, "loss": 0.6934, "step": 9847 }, { "epoch": 3.03, "learning_rate": 1.6344245785401653e-05, "loss": 0.4258, "step": 9848 }, { "epoch": 3.03, "learning_rate": 1.6343475443584018e-05, "loss": 0.3978, "step": 9849 }, { "epoch": 3.03, "learning_rate": 1.6342705038769822e-05, "loss": 0.27, "step": 9850 }, { "epoch": 3.03, "learning_rate": 1.634193457096672e-05, "loss": 0.6212, "step": 9851 }, { "epoch": 3.03, "learning_rate": 1.6341164040182366e-05, "loss": 0.3341, "step": 9852 }, { "epoch": 3.03, "learning_rate": 1.6340393446424406e-05, "loss": 0.6909, "step": 9853 }, { "epoch": 3.03, "learning_rate": 1.6339622789700494e-05, "loss": 0.4091, "step": 9854 }, { "epoch": 3.03, "learning_rate": 1.633885207001829e-05, "loss": 0.5945, "step": 9855 }, { "epoch": 3.03, "learning_rate": 1.633808128738544e-05, "loss": 0.525, "step": 9856 }, { "epoch": 3.03, "learning_rate": 1.63373104418096e-05, "loss": 0.3985, "step": 9857 }, { "epoch": 3.03, "learning_rate": 1.633653953329843e-05, "loss": 0.4263, "step": 9858 }, { "epoch": 3.03, "learning_rate": 1.633576856185958e-05, "loss": 0.366, "step": 9859 }, { "epoch": 3.03, "learning_rate": 1.633499752750071e-05, "loss": 0.5035, "step": 9860 }, { "epoch": 3.03, "learning_rate": 1.6334226430229475e-05, "loss": 0.4125, "step": 9861 }, { "epoch": 3.03, "learning_rate": 1.6333455270053537e-05, "loss": 0.5457, "step": 9862 }, { "epoch": 3.03, "learning_rate": 1.6332684046980547e-05, "loss": 0.5803, "step": 9863 }, { "epoch": 3.03, "learning_rate": 1.633191276101817e-05, "loss": 0.2964, "step": 9864 }, { "epoch": 3.04, "learning_rate": 1.6331141412174065e-05, "loss": 0.7301, "step": 9865 }, { "epoch": 3.04, "learning_rate": 1.633037000045589e-05, "loss": 0.5987, "step": 9866 }, { "epoch": 3.04, "learning_rate": 1.6329598525871303e-05, "loss": 0.463, "step": 9867 }, { "epoch": 3.04, "learning_rate": 1.6328826988427974e-05, "loss": 0.4559, "step": 9868 }, { "epoch": 3.04, "learning_rate": 1.6328055388133557e-05, "loss": 0.3896, "step": 9869 }, { "epoch": 3.04, "learning_rate": 1.632728372499572e-05, "loss": 0.4418, "step": 9870 }, { "epoch": 3.04, "learning_rate": 1.6326511999022124e-05, "loss": 0.4956, "step": 9871 }, { "epoch": 3.04, "learning_rate": 1.632574021022043e-05, "loss": 0.5723, "step": 9872 }, { "epoch": 3.04, "learning_rate": 1.632496835859831e-05, "loss": 0.6736, "step": 9873 }, { "epoch": 3.04, "learning_rate": 1.632419644416342e-05, "loss": 0.4612, "step": 9874 }, { "epoch": 3.04, "learning_rate": 1.6323424466923435e-05, "loss": 0.4849, "step": 9875 }, { "epoch": 3.04, "learning_rate": 1.6322652426886016e-05, "loss": 0.577, "step": 9876 }, { "epoch": 3.04, "learning_rate": 1.6321880324058833e-05, "loss": 0.4301, "step": 9877 }, { "epoch": 3.04, "learning_rate": 1.6321108158449548e-05, "loss": 0.6939, "step": 9878 }, { "epoch": 3.04, "learning_rate": 1.6320335930065838e-05, "loss": 0.5325, "step": 9879 }, { "epoch": 3.04, "learning_rate": 1.6319563638915364e-05, "loss": 0.4935, "step": 9880 }, { "epoch": 3.04, "learning_rate": 1.6318791285005803e-05, "loss": 0.3906, "step": 9881 }, { "epoch": 3.04, "learning_rate": 1.6318018868344812e-05, "loss": 0.3981, "step": 9882 }, { "epoch": 3.04, "learning_rate": 1.6317246388940077e-05, "loss": 0.5798, "step": 9883 }, { "epoch": 3.04, "learning_rate": 1.6316473846799263e-05, "loss": 0.6492, "step": 9884 }, { "epoch": 3.04, "learning_rate": 1.6315701241930045e-05, "loss": 0.3647, "step": 9885 }, { "epoch": 3.04, "learning_rate": 1.631492857434009e-05, "loss": 0.5696, "step": 9886 }, { "epoch": 3.04, "learning_rate": 1.6314155844037074e-05, "loss": 0.5676, "step": 9887 }, { "epoch": 3.04, "learning_rate": 1.6313383051028672e-05, "loss": 0.704, "step": 9888 }, { "epoch": 3.04, "learning_rate": 1.6312610195322553e-05, "loss": 0.3768, "step": 9889 }, { "epoch": 3.04, "learning_rate": 1.63118372769264e-05, "loss": 0.3338, "step": 9890 }, { "epoch": 3.04, "learning_rate": 1.631106429584789e-05, "loss": 0.5185, "step": 9891 }, { "epoch": 3.04, "learning_rate": 1.631029125209469e-05, "loss": 0.5872, "step": 9892 }, { "epoch": 3.04, "learning_rate": 1.6309518145674484e-05, "loss": 0.4286, "step": 9893 }, { "epoch": 3.04, "learning_rate": 1.6308744976594947e-05, "loss": 0.5053, "step": 9894 }, { "epoch": 3.04, "learning_rate": 1.630797174486376e-05, "loss": 0.4254, "step": 9895 }, { "epoch": 3.04, "learning_rate": 1.6307198450488598e-05, "loss": 0.3922, "step": 9896 }, { "epoch": 3.05, "learning_rate": 1.6306425093477142e-05, "loss": 0.4944, "step": 9897 }, { "epoch": 3.05, "learning_rate": 1.6305651673837076e-05, "loss": 0.6882, "step": 9898 }, { "epoch": 3.05, "learning_rate": 1.630487819157607e-05, "loss": 0.3316, "step": 9899 }, { "epoch": 3.05, "learning_rate": 1.6304104646701818e-05, "loss": 0.3509, "step": 9900 }, { "epoch": 3.05, "learning_rate": 1.6303331039222e-05, "loss": 0.4312, "step": 9901 }, { "epoch": 3.05, "learning_rate": 1.630255736914429e-05, "loss": 0.6011, "step": 9902 }, { "epoch": 3.05, "learning_rate": 1.6301783636476383e-05, "loss": 0.3829, "step": 9903 }, { "epoch": 3.05, "learning_rate": 1.6301009841225954e-05, "loss": 0.4801, "step": 9904 }, { "epoch": 3.05, "learning_rate": 1.6300235983400684e-05, "loss": 0.342, "step": 9905 }, { "epoch": 3.05, "learning_rate": 1.6299462063008272e-05, "loss": 0.4842, "step": 9906 }, { "epoch": 3.05, "learning_rate": 1.629868808005639e-05, "loss": 0.5317, "step": 9907 }, { "epoch": 3.05, "learning_rate": 1.6297914034552734e-05, "loss": 0.3996, "step": 9908 }, { "epoch": 3.05, "learning_rate": 1.6297139926504987e-05, "loss": 0.3823, "step": 9909 }, { "epoch": 3.05, "learning_rate": 1.6296365755920835e-05, "loss": 0.4184, "step": 9910 }, { "epoch": 3.05, "learning_rate": 1.6295591522807972e-05, "loss": 0.6521, "step": 9911 }, { "epoch": 3.05, "learning_rate": 1.629481722717408e-05, "loss": 0.4593, "step": 9912 }, { "epoch": 3.05, "learning_rate": 1.629404286902685e-05, "loss": 0.3994, "step": 9913 }, { "epoch": 3.05, "learning_rate": 1.6293268448373975e-05, "loss": 0.6277, "step": 9914 }, { "epoch": 3.05, "learning_rate": 1.6292493965223145e-05, "loss": 0.8505, "step": 9915 }, { "epoch": 3.05, "learning_rate": 1.6291719419582054e-05, "loss": 0.4446, "step": 9916 }, { "epoch": 3.05, "learning_rate": 1.6290944811458387e-05, "loss": 0.4154, "step": 9917 }, { "epoch": 3.05, "learning_rate": 1.6290170140859838e-05, "loss": 0.6183, "step": 9918 }, { "epoch": 3.05, "learning_rate": 1.6289395407794105e-05, "loss": 0.3753, "step": 9919 }, { "epoch": 3.05, "learning_rate": 1.6288620612268875e-05, "loss": 0.3888, "step": 9920 }, { "epoch": 3.05, "learning_rate": 1.6287845754291852e-05, "loss": 0.3813, "step": 9921 }, { "epoch": 3.05, "learning_rate": 1.6287070833870723e-05, "loss": 0.3993, "step": 9922 }, { "epoch": 3.05, "learning_rate": 1.628629585101319e-05, "loss": 0.6096, "step": 9923 }, { "epoch": 3.05, "learning_rate": 1.6285520805726938e-05, "loss": 0.7234, "step": 9924 }, { "epoch": 3.05, "learning_rate": 1.628474569801968e-05, "loss": 0.4664, "step": 9925 }, { "epoch": 3.05, "learning_rate": 1.62839705278991e-05, "loss": 0.5699, "step": 9926 }, { "epoch": 3.05, "learning_rate": 1.62831952953729e-05, "loss": 0.3729, "step": 9927 }, { "epoch": 3.05, "learning_rate": 1.628242000044878e-05, "loss": 0.5589, "step": 9928 }, { "epoch": 3.05, "learning_rate": 1.6281644643134448e-05, "loss": 0.4072, "step": 9929 }, { "epoch": 3.06, "learning_rate": 1.6280869223437585e-05, "loss": 0.5574, "step": 9930 }, { "epoch": 3.06, "learning_rate": 1.6280093741365907e-05, "loss": 0.6138, "step": 9931 }, { "epoch": 3.06, "learning_rate": 1.627931819692711e-05, "loss": 0.4043, "step": 9932 }, { "epoch": 3.06, "learning_rate": 1.6278542590128893e-05, "loss": 0.3841, "step": 9933 }, { "epoch": 3.06, "learning_rate": 1.6277766920978963e-05, "loss": 0.5939, "step": 9934 }, { "epoch": 3.06, "learning_rate": 1.6276991189485027e-05, "loss": 0.5917, "step": 9935 }, { "epoch": 3.06, "learning_rate": 1.6276215395654778e-05, "loss": 0.5027, "step": 9936 }, { "epoch": 3.06, "learning_rate": 1.6275439539495925e-05, "loss": 0.4355, "step": 9937 }, { "epoch": 3.06, "learning_rate": 1.6274663621016175e-05, "loss": 0.6025, "step": 9938 }, { "epoch": 3.06, "learning_rate": 1.627388764022323e-05, "loss": 0.591, "step": 9939 }, { "epoch": 3.06, "learning_rate": 1.6273111597124806e-05, "loss": 0.5731, "step": 9940 }, { "epoch": 3.06, "learning_rate": 1.62723354917286e-05, "loss": 0.2202, "step": 9941 }, { "epoch": 3.06, "learning_rate": 1.6271559324042317e-05, "loss": 0.4489, "step": 9942 }, { "epoch": 3.06, "learning_rate": 1.627078309407367e-05, "loss": 0.6396, "step": 9943 }, { "epoch": 3.06, "learning_rate": 1.627000680183037e-05, "loss": 0.4094, "step": 9944 }, { "epoch": 3.06, "learning_rate": 1.6269230447320124e-05, "loss": 0.8877, "step": 9945 }, { "epoch": 3.06, "learning_rate": 1.6268454030550645e-05, "loss": 0.5232, "step": 9946 }, { "epoch": 3.06, "learning_rate": 1.6267677551529634e-05, "loss": 0.2832, "step": 9947 }, { "epoch": 3.06, "learning_rate": 1.6266901010264814e-05, "loss": 0.3942, "step": 9948 }, { "epoch": 3.06, "learning_rate": 1.6266124406763885e-05, "loss": 0.3046, "step": 9949 }, { "epoch": 3.06, "learning_rate": 1.626534774103457e-05, "loss": 0.4771, "step": 9950 }, { "epoch": 3.06, "learning_rate": 1.626457101308458e-05, "loss": 0.4151, "step": 9951 }, { "epoch": 3.06, "learning_rate": 1.626379422292162e-05, "loss": 0.6717, "step": 9952 }, { "epoch": 3.06, "learning_rate": 1.6263017370553414e-05, "loss": 0.4945, "step": 9953 }, { "epoch": 3.06, "learning_rate": 1.6262240455987672e-05, "loss": 0.2654, "step": 9954 }, { "epoch": 3.06, "learning_rate": 1.626146347923211e-05, "loss": 0.4175, "step": 9955 }, { "epoch": 3.06, "learning_rate": 1.626068644029445e-05, "loss": 0.5258, "step": 9956 }, { "epoch": 3.06, "learning_rate": 1.6259909339182402e-05, "loss": 0.4417, "step": 9957 }, { "epoch": 3.06, "learning_rate": 1.6259132175903684e-05, "loss": 0.6373, "step": 9958 }, { "epoch": 3.06, "learning_rate": 1.625835495046601e-05, "loss": 0.5828, "step": 9959 }, { "epoch": 3.06, "learning_rate": 1.6257577662877107e-05, "loss": 0.3584, "step": 9960 }, { "epoch": 3.06, "learning_rate": 1.6256800313144692e-05, "loss": 0.5896, "step": 9961 }, { "epoch": 3.07, "learning_rate": 1.6256022901276485e-05, "loss": 0.485, "step": 9962 }, { "epoch": 3.07, "learning_rate": 1.6255245427280203e-05, "loss": 0.5552, "step": 9963 }, { "epoch": 3.07, "learning_rate": 1.6254467891163565e-05, "loss": 0.4519, "step": 9964 }, { "epoch": 3.07, "learning_rate": 1.6253690292934303e-05, "loss": 0.7211, "step": 9965 }, { "epoch": 3.07, "learning_rate": 1.6252912632600127e-05, "loss": 0.5746, "step": 9966 }, { "epoch": 3.07, "learning_rate": 1.625213491016877e-05, "loss": 0.3796, "step": 9967 }, { "epoch": 3.07, "learning_rate": 1.625135712564795e-05, "loss": 0.4129, "step": 9968 }, { "epoch": 3.07, "learning_rate": 1.625057927904539e-05, "loss": 0.2859, "step": 9969 }, { "epoch": 3.07, "learning_rate": 1.624980137036882e-05, "loss": 0.4671, "step": 9970 }, { "epoch": 3.07, "learning_rate": 1.6249023399625963e-05, "loss": 0.4283, "step": 9971 }, { "epoch": 3.07, "learning_rate": 1.6248245366824543e-05, "loss": 0.4983, "step": 9972 }, { "epoch": 3.07, "learning_rate": 1.6247467271972287e-05, "loss": 0.425, "step": 9973 }, { "epoch": 3.07, "learning_rate": 1.6246689115076922e-05, "loss": 0.7279, "step": 9974 }, { "epoch": 3.07, "learning_rate": 1.624591089614618e-05, "loss": 0.534, "step": 9975 }, { "epoch": 3.07, "learning_rate": 1.6245132615187787e-05, "loss": 0.6155, "step": 9976 }, { "epoch": 3.07, "learning_rate": 1.6244354272209468e-05, "loss": 0.392, "step": 9977 }, { "epoch": 3.07, "learning_rate": 1.624357586721896e-05, "loss": 0.5256, "step": 9978 }, { "epoch": 3.07, "learning_rate": 1.6242797400223987e-05, "loss": 0.542, "step": 9979 }, { "epoch": 3.07, "learning_rate": 1.6242018871232283e-05, "loss": 0.5286, "step": 9980 }, { "epoch": 3.07, "learning_rate": 1.624124028025158e-05, "loss": 0.4356, "step": 9981 }, { "epoch": 3.07, "learning_rate": 1.6240461627289607e-05, "loss": 0.4801, "step": 9982 }, { "epoch": 3.07, "learning_rate": 1.62396829123541e-05, "loss": 0.5461, "step": 9983 }, { "epoch": 3.07, "learning_rate": 1.6238904135452788e-05, "loss": 0.3851, "step": 9984 }, { "epoch": 3.07, "learning_rate": 1.623812529659341e-05, "loss": 0.5669, "step": 9985 }, { "epoch": 3.07, "learning_rate": 1.6237346395783697e-05, "loss": 0.5315, "step": 9986 }, { "epoch": 3.07, "learning_rate": 1.6236567433031387e-05, "loss": 0.5102, "step": 9987 }, { "epoch": 3.07, "learning_rate": 1.6235788408344215e-05, "loss": 0.4332, "step": 9988 }, { "epoch": 3.07, "learning_rate": 1.6235009321729918e-05, "loss": 0.5123, "step": 9989 }, { "epoch": 3.07, "learning_rate": 1.6234230173196232e-05, "loss": 0.5552, "step": 9990 }, { "epoch": 3.07, "learning_rate": 1.6233450962750895e-05, "loss": 0.3555, "step": 9991 }, { "epoch": 3.07, "learning_rate": 1.6232671690401643e-05, "loss": 0.6026, "step": 9992 }, { "epoch": 3.07, "learning_rate": 1.6231892356156217e-05, "loss": 0.8226, "step": 9993 }, { "epoch": 3.07, "learning_rate": 1.6231112960022356e-05, "loss": 0.5143, "step": 9994 }, { "epoch": 3.08, "learning_rate": 1.62303335020078e-05, "loss": 0.4343, "step": 9995 }, { "epoch": 3.08, "learning_rate": 1.6229553982120295e-05, "loss": 0.4186, "step": 9996 }, { "epoch": 3.08, "learning_rate": 1.6228774400367573e-05, "loss": 0.4715, "step": 9997 }, { "epoch": 3.08, "learning_rate": 1.6227994756757383e-05, "loss": 0.4998, "step": 9998 }, { "epoch": 3.08, "learning_rate": 1.6227215051297464e-05, "loss": 0.4399, "step": 9999 }, { "epoch": 3.08, "learning_rate": 1.622643528399556e-05, "loss": 0.4529, "step": 10000 }, { "epoch": 3.08, "learning_rate": 1.6225655454859416e-05, "loss": 0.4252, "step": 10001 }, { "epoch": 3.08, "learning_rate": 1.6224875563896775e-05, "loss": 0.5635, "step": 10002 }, { "epoch": 3.08, "learning_rate": 1.6224095611115385e-05, "loss": 0.4227, "step": 10003 }, { "epoch": 3.08, "learning_rate": 1.622331559652299e-05, "loss": 0.2825, "step": 10004 }, { "epoch": 3.08, "learning_rate": 1.6222535520127332e-05, "loss": 0.4109, "step": 10005 }, { "epoch": 3.08, "learning_rate": 1.622175538193616e-05, "loss": 0.2786, "step": 10006 }, { "epoch": 3.08, "learning_rate": 1.6220975181957227e-05, "loss": 0.3904, "step": 10007 }, { "epoch": 3.08, "learning_rate": 1.6220194920198276e-05, "loss": 0.2375, "step": 10008 }, { "epoch": 3.08, "learning_rate": 1.6219414596667055e-05, "loss": 0.3741, "step": 10009 }, { "epoch": 3.08, "learning_rate": 1.621863421137132e-05, "loss": 0.449, "step": 10010 }, { "epoch": 3.08, "learning_rate": 1.621785376431881e-05, "loss": 0.4725, "step": 10011 }, { "epoch": 3.08, "learning_rate": 1.621707325551728e-05, "loss": 0.6257, "step": 10012 }, { "epoch": 3.08, "learning_rate": 1.6216292684974488e-05, "loss": 0.2795, "step": 10013 }, { "epoch": 3.08, "learning_rate": 1.621551205269818e-05, "loss": 0.66, "step": 10014 }, { "epoch": 3.08, "learning_rate": 1.621473135869611e-05, "loss": 0.4123, "step": 10015 }, { "epoch": 3.08, "learning_rate": 1.6213950602976027e-05, "loss": 0.398, "step": 10016 }, { "epoch": 3.08, "learning_rate": 1.6213169785545688e-05, "loss": 0.4669, "step": 10017 }, { "epoch": 3.08, "learning_rate": 1.6212388906412848e-05, "loss": 0.4576, "step": 10018 }, { "epoch": 3.08, "learning_rate": 1.621160796558526e-05, "loss": 0.5522, "step": 10019 }, { "epoch": 3.08, "learning_rate": 1.621082696307068e-05, "loss": 0.4825, "step": 10020 }, { "epoch": 3.08, "learning_rate": 1.621004589887686e-05, "loss": 0.477, "step": 10021 }, { "epoch": 3.08, "learning_rate": 1.6209264773011567e-05, "loss": 0.4432, "step": 10022 }, { "epoch": 3.08, "learning_rate": 1.6208483585482552e-05, "loss": 0.6384, "step": 10023 }, { "epoch": 3.08, "learning_rate": 1.6207702336297572e-05, "loss": 0.5942, "step": 10024 }, { "epoch": 3.08, "learning_rate": 1.6206921025464385e-05, "loss": 0.3731, "step": 10025 }, { "epoch": 3.08, "learning_rate": 1.620613965299075e-05, "loss": 0.5804, "step": 10026 }, { "epoch": 3.09, "learning_rate": 1.6205358218884434e-05, "loss": 0.6102, "step": 10027 }, { "epoch": 3.09, "learning_rate": 1.6204576723153188e-05, "loss": 0.4358, "step": 10028 }, { "epoch": 3.09, "learning_rate": 1.620379516580478e-05, "loss": 0.6777, "step": 10029 }, { "epoch": 3.09, "learning_rate": 1.6203013546846967e-05, "loss": 0.5444, "step": 10030 }, { "epoch": 3.09, "learning_rate": 1.620223186628751e-05, "loss": 0.5704, "step": 10031 }, { "epoch": 3.09, "learning_rate": 1.620145012413418e-05, "loss": 0.414, "step": 10032 }, { "epoch": 3.09, "learning_rate": 1.620066832039473e-05, "loss": 0.6241, "step": 10033 }, { "epoch": 3.09, "learning_rate": 1.619988645507693e-05, "loss": 0.5006, "step": 10034 }, { "epoch": 3.09, "learning_rate": 1.6199104528188543e-05, "loss": 0.3841, "step": 10035 }, { "epoch": 3.09, "learning_rate": 1.6198322539737333e-05, "loss": 0.3993, "step": 10036 }, { "epoch": 3.09, "learning_rate": 1.6197540489731073e-05, "loss": 0.2806, "step": 10037 }, { "epoch": 3.09, "learning_rate": 1.6196758378177523e-05, "loss": 0.3914, "step": 10038 }, { "epoch": 3.09, "learning_rate": 1.6195976205084447e-05, "loss": 0.5416, "step": 10039 }, { "epoch": 3.09, "learning_rate": 1.6195193970459618e-05, "loss": 0.3746, "step": 10040 }, { "epoch": 3.09, "learning_rate": 1.6194411674310805e-05, "loss": 0.5448, "step": 10041 }, { "epoch": 3.09, "learning_rate": 1.6193629316645777e-05, "loss": 0.5844, "step": 10042 }, { "epoch": 3.09, "learning_rate": 1.61928468974723e-05, "loss": 0.5905, "step": 10043 }, { "epoch": 3.09, "learning_rate": 1.6192064416798142e-05, "loss": 0.4297, "step": 10044 }, { "epoch": 3.09, "learning_rate": 1.6191281874631083e-05, "loss": 0.5663, "step": 10045 }, { "epoch": 3.09, "learning_rate": 1.6190499270978885e-05, "loss": 0.4399, "step": 10046 }, { "epoch": 3.09, "learning_rate": 1.6189716605849327e-05, "loss": 0.5007, "step": 10047 }, { "epoch": 3.09, "learning_rate": 1.6188933879250178e-05, "loss": 0.4316, "step": 10048 }, { "epoch": 3.09, "learning_rate": 1.6188151091189207e-05, "loss": 0.506, "step": 10049 }, { "epoch": 3.09, "learning_rate": 1.6187368241674196e-05, "loss": 0.6416, "step": 10050 }, { "epoch": 3.09, "learning_rate": 1.6186585330712916e-05, "loss": 0.5141, "step": 10051 }, { "epoch": 3.09, "learning_rate": 1.6185802358313147e-05, "loss": 0.4389, "step": 10052 }, { "epoch": 3.09, "learning_rate": 1.6185019324482652e-05, "loss": 0.6496, "step": 10053 }, { "epoch": 3.09, "learning_rate": 1.6184236229229218e-05, "loss": 0.5475, "step": 10054 }, { "epoch": 3.09, "learning_rate": 1.6183453072560617e-05, "loss": 0.502, "step": 10055 }, { "epoch": 3.09, "learning_rate": 1.618266985448463e-05, "loss": 0.5101, "step": 10056 }, { "epoch": 3.09, "learning_rate": 1.618188657500903e-05, "loss": 0.3981, "step": 10057 }, { "epoch": 3.09, "learning_rate": 1.6181103234141606e-05, "loss": 0.5687, "step": 10058 }, { "epoch": 3.09, "learning_rate": 1.6180319831890123e-05, "loss": 0.4345, "step": 10059 }, { "epoch": 3.1, "learning_rate": 1.617953636826237e-05, "loss": 0.4966, "step": 10060 }, { "epoch": 3.1, "learning_rate": 1.6178752843266128e-05, "loss": 0.554, "step": 10061 }, { "epoch": 3.1, "learning_rate": 1.6177969256909177e-05, "loss": 0.5059, "step": 10062 }, { "epoch": 3.1, "learning_rate": 1.6177185609199293e-05, "loss": 0.3639, "step": 10063 }, { "epoch": 3.1, "learning_rate": 1.6176401900144264e-05, "loss": 0.4837, "step": 10064 }, { "epoch": 3.1, "learning_rate": 1.6175618129751872e-05, "loss": 0.3732, "step": 10065 }, { "epoch": 3.1, "learning_rate": 1.61748342980299e-05, "loss": 0.5364, "step": 10066 }, { "epoch": 3.1, "learning_rate": 1.6174050404986135e-05, "loss": 0.5068, "step": 10067 }, { "epoch": 3.1, "learning_rate": 1.6173266450628355e-05, "loss": 0.4492, "step": 10068 }, { "epoch": 3.1, "learning_rate": 1.6172482434964353e-05, "loss": 0.392, "step": 10069 }, { "epoch": 3.1, "learning_rate": 1.617169835800191e-05, "loss": 0.4638, "step": 10070 }, { "epoch": 3.1, "learning_rate": 1.6170914219748817e-05, "loss": 0.6061, "step": 10071 }, { "epoch": 3.1, "learning_rate": 1.6170130020212854e-05, "loss": 0.7239, "step": 10072 }, { "epoch": 3.1, "learning_rate": 1.616934575940182e-05, "loss": 0.4612, "step": 10073 }, { "epoch": 3.1, "learning_rate": 1.616856143732349e-05, "loss": 0.3723, "step": 10074 }, { "epoch": 3.1, "learning_rate": 1.6167777053985664e-05, "loss": 0.6995, "step": 10075 }, { "epoch": 3.1, "learning_rate": 1.6166992609396124e-05, "loss": 0.5015, "step": 10076 }, { "epoch": 3.1, "learning_rate": 1.616620810356267e-05, "loss": 0.6687, "step": 10077 }, { "epoch": 3.1, "learning_rate": 1.616542353649308e-05, "loss": 0.3946, "step": 10078 }, { "epoch": 3.1, "learning_rate": 1.6164638908195155e-05, "loss": 0.5322, "step": 10079 }, { "epoch": 3.1, "learning_rate": 1.6163854218676684e-05, "loss": 0.3172, "step": 10080 }, { "epoch": 3.1, "learning_rate": 1.616306946794546e-05, "loss": 0.496, "step": 10081 }, { "epoch": 3.1, "learning_rate": 1.6162284656009276e-05, "loss": 0.4695, "step": 10082 }, { "epoch": 3.1, "learning_rate": 1.6161499782875926e-05, "loss": 0.6253, "step": 10083 }, { "epoch": 3.1, "learning_rate": 1.6160714848553203e-05, "loss": 0.6415, "step": 10084 }, { "epoch": 3.1, "learning_rate": 1.6159929853048904e-05, "loss": 0.4255, "step": 10085 }, { "epoch": 3.1, "learning_rate": 1.6159144796370825e-05, "loss": 0.397, "step": 10086 }, { "epoch": 3.1, "learning_rate": 1.6158359678526763e-05, "loss": 0.3562, "step": 10087 }, { "epoch": 3.1, "learning_rate": 1.6157574499524512e-05, "loss": 0.4438, "step": 10088 }, { "epoch": 3.1, "learning_rate": 1.615678925937187e-05, "loss": 0.5403, "step": 10089 }, { "epoch": 3.1, "learning_rate": 1.615600395807664e-05, "loss": 0.5487, "step": 10090 }, { "epoch": 3.1, "learning_rate": 1.6155218595646616e-05, "loss": 0.6572, "step": 10091 }, { "epoch": 3.11, "learning_rate": 1.61544331720896e-05, "loss": 0.3345, "step": 10092 }, { "epoch": 3.11, "learning_rate": 1.615364768741339e-05, "loss": 0.473, "step": 10093 }, { "epoch": 3.11, "learning_rate": 1.615286214162578e-05, "loss": 0.3896, "step": 10094 }, { "epoch": 3.11, "learning_rate": 1.6152076534734585e-05, "loss": 0.3366, "step": 10095 }, { "epoch": 3.11, "learning_rate": 1.61512908667476e-05, "loss": 0.592, "step": 10096 }, { "epoch": 3.11, "learning_rate": 1.6150505137672626e-05, "loss": 0.4909, "step": 10097 }, { "epoch": 3.11, "learning_rate": 1.614971934751747e-05, "loss": 0.7131, "step": 10098 }, { "epoch": 3.11, "learning_rate": 1.614893349628993e-05, "loss": 0.6835, "step": 10099 }, { "epoch": 3.11, "learning_rate": 1.6148147583997813e-05, "loss": 0.5334, "step": 10100 }, { "epoch": 3.11, "learning_rate": 1.6147361610648926e-05, "loss": 0.4705, "step": 10101 }, { "epoch": 3.11, "learning_rate": 1.614657557625107e-05, "loss": 0.6339, "step": 10102 }, { "epoch": 3.11, "learning_rate": 1.614578948081206e-05, "loss": 0.3783, "step": 10103 }, { "epoch": 3.11, "learning_rate": 1.614500332433969e-05, "loss": 0.5078, "step": 10104 }, { "epoch": 3.11, "learning_rate": 1.6144217106841775e-05, "loss": 0.489, "step": 10105 }, { "epoch": 3.11, "learning_rate": 1.6143430828326124e-05, "loss": 0.5791, "step": 10106 }, { "epoch": 3.11, "learning_rate": 1.614264448880054e-05, "loss": 0.5228, "step": 10107 }, { "epoch": 3.11, "learning_rate": 1.6141858088272838e-05, "loss": 0.5619, "step": 10108 }, { "epoch": 3.11, "learning_rate": 1.6141071626750822e-05, "loss": 0.6422, "step": 10109 }, { "epoch": 3.11, "learning_rate": 1.6140285104242306e-05, "loss": 0.5208, "step": 10110 }, { "epoch": 3.11, "learning_rate": 1.61394985207551e-05, "loss": 0.6151, "step": 10111 }, { "epoch": 3.11, "learning_rate": 1.6138711876297013e-05, "loss": 0.5529, "step": 10112 }, { "epoch": 3.11, "learning_rate": 1.6137925170875865e-05, "loss": 0.5907, "step": 10113 }, { "epoch": 3.11, "learning_rate": 1.6137138404499458e-05, "loss": 0.4099, "step": 10114 }, { "epoch": 3.11, "learning_rate": 1.6136351577175612e-05, "loss": 0.5031, "step": 10115 }, { "epoch": 3.11, "learning_rate": 1.613556468891214e-05, "loss": 0.4675, "step": 10116 }, { "epoch": 3.11, "learning_rate": 1.6134777739716855e-05, "loss": 0.4905, "step": 10117 }, { "epoch": 3.11, "learning_rate": 1.6133990729597576e-05, "loss": 0.5317, "step": 10118 }, { "epoch": 3.11, "learning_rate": 1.613320365856211e-05, "loss": 0.3878, "step": 10119 }, { "epoch": 3.11, "learning_rate": 1.6132416526618285e-05, "loss": 0.558, "step": 10120 }, { "epoch": 3.11, "learning_rate": 1.6131629333773908e-05, "loss": 0.4749, "step": 10121 }, { "epoch": 3.11, "learning_rate": 1.6130842080036804e-05, "loss": 0.6085, "step": 10122 }, { "epoch": 3.11, "learning_rate": 1.6130054765414788e-05, "loss": 0.4649, "step": 10123 }, { "epoch": 3.11, "learning_rate": 1.6129267389915674e-05, "loss": 0.5894, "step": 10124 }, { "epoch": 3.12, "learning_rate": 1.612847995354729e-05, "loss": 0.4812, "step": 10125 }, { "epoch": 3.12, "learning_rate": 1.612769245631745e-05, "loss": 0.3623, "step": 10126 }, { "epoch": 3.12, "learning_rate": 1.612690489823398e-05, "loss": 0.4434, "step": 10127 }, { "epoch": 3.12, "learning_rate": 1.612611727930469e-05, "loss": 0.3102, "step": 10128 }, { "epoch": 3.12, "learning_rate": 1.6125329599537414e-05, "loss": 0.4025, "step": 10129 }, { "epoch": 3.12, "learning_rate": 1.612454185893997e-05, "loss": 0.3765, "step": 10130 }, { "epoch": 3.12, "learning_rate": 1.612375405752018e-05, "loss": 0.4562, "step": 10131 }, { "epoch": 3.12, "learning_rate": 1.612296619528587e-05, "loss": 0.3857, "step": 10132 }, { "epoch": 3.12, "learning_rate": 1.6122178272244862e-05, "loss": 0.5135, "step": 10133 }, { "epoch": 3.12, "learning_rate": 1.612139028840498e-05, "loss": 0.6618, "step": 10134 }, { "epoch": 3.12, "learning_rate": 1.6120602243774054e-05, "loss": 0.4725, "step": 10135 }, { "epoch": 3.12, "learning_rate": 1.61198141383599e-05, "loss": 0.6847, "step": 10136 }, { "epoch": 3.12, "learning_rate": 1.6119025972170357e-05, "loss": 0.5448, "step": 10137 }, { "epoch": 3.12, "learning_rate": 1.6118237745213247e-05, "loss": 0.5866, "step": 10138 }, { "epoch": 3.12, "learning_rate": 1.6117449457496396e-05, "loss": 0.4159, "step": 10139 }, { "epoch": 3.12, "learning_rate": 1.6116661109027636e-05, "loss": 0.4707, "step": 10140 }, { "epoch": 3.12, "learning_rate": 1.6115872699814792e-05, "loss": 0.3423, "step": 10141 }, { "epoch": 3.12, "learning_rate": 1.611508422986569e-05, "loss": 0.71, "step": 10142 }, { "epoch": 3.12, "learning_rate": 1.6114295699188177e-05, "loss": 0.5015, "step": 10143 }, { "epoch": 3.12, "learning_rate": 1.6113507107790065e-05, "loss": 0.659, "step": 10144 }, { "epoch": 3.12, "learning_rate": 1.6112718455679194e-05, "loss": 0.5867, "step": 10145 }, { "epoch": 3.12, "learning_rate": 1.6111929742863395e-05, "loss": 0.4469, "step": 10146 }, { "epoch": 3.12, "learning_rate": 1.6111140969350504e-05, "loss": 0.1622, "step": 10147 }, { "epoch": 3.12, "learning_rate": 1.6110352135148348e-05, "loss": 0.4858, "step": 10148 }, { "epoch": 3.12, "learning_rate": 1.6109563240264766e-05, "loss": 0.4823, "step": 10149 }, { "epoch": 3.12, "learning_rate": 1.6108774284707587e-05, "loss": 0.366, "step": 10150 }, { "epoch": 3.12, "learning_rate": 1.6107985268484648e-05, "loss": 0.5228, "step": 10151 }, { "epoch": 3.12, "learning_rate": 1.6107196191603792e-05, "loss": 0.5199, "step": 10152 }, { "epoch": 3.12, "learning_rate": 1.6106407054072847e-05, "loss": 0.5414, "step": 10153 }, { "epoch": 3.12, "learning_rate": 1.610561785589965e-05, "loss": 0.3304, "step": 10154 }, { "epoch": 3.12, "learning_rate": 1.6104828597092043e-05, "loss": 0.3327, "step": 10155 }, { "epoch": 3.12, "learning_rate": 1.6104039277657858e-05, "loss": 0.3948, "step": 10156 }, { "epoch": 3.13, "learning_rate": 1.6103249897604944e-05, "loss": 0.409, "step": 10157 }, { "epoch": 3.13, "learning_rate": 1.610246045694113e-05, "loss": 0.3747, "step": 10158 }, { "epoch": 3.13, "learning_rate": 1.6101670955674257e-05, "loss": 0.5779, "step": 10159 }, { "epoch": 3.13, "learning_rate": 1.610088139381217e-05, "loss": 0.4067, "step": 10160 }, { "epoch": 3.13, "learning_rate": 1.610009177136271e-05, "loss": 0.4389, "step": 10161 }, { "epoch": 3.13, "learning_rate": 1.6099302088333714e-05, "loss": 0.4292, "step": 10162 }, { "epoch": 3.13, "learning_rate": 1.6098512344733026e-05, "loss": 0.4175, "step": 10163 }, { "epoch": 3.13, "learning_rate": 1.6097722540568496e-05, "loss": 0.311, "step": 10164 }, { "epoch": 3.13, "learning_rate": 1.6096932675847954e-05, "loss": 0.427, "step": 10165 }, { "epoch": 3.13, "learning_rate": 1.6096142750579257e-05, "loss": 0.335, "step": 10166 }, { "epoch": 3.13, "learning_rate": 1.609535276477024e-05, "loss": 0.4675, "step": 10167 }, { "epoch": 3.13, "learning_rate": 1.609456271842875e-05, "loss": 0.3716, "step": 10168 }, { "epoch": 3.13, "learning_rate": 1.609377261156264e-05, "loss": 0.6409, "step": 10169 }, { "epoch": 3.13, "learning_rate": 1.609298244417975e-05, "loss": 0.4324, "step": 10170 }, { "epoch": 3.13, "learning_rate": 1.609219221628793e-05, "loss": 0.4264, "step": 10171 }, { "epoch": 3.13, "learning_rate": 1.6091401927895028e-05, "loss": 0.4135, "step": 10172 }, { "epoch": 3.13, "learning_rate": 1.609061157900889e-05, "loss": 0.5333, "step": 10173 }, { "epoch": 3.13, "learning_rate": 1.6089821169637366e-05, "loss": 0.2092, "step": 10174 }, { "epoch": 3.13, "learning_rate": 1.6089030699788304e-05, "loss": 0.5121, "step": 10175 }, { "epoch": 3.13, "learning_rate": 1.6088240169469553e-05, "loss": 0.5217, "step": 10176 }, { "epoch": 3.13, "learning_rate": 1.608744957868897e-05, "loss": 0.359, "step": 10177 }, { "epoch": 3.13, "learning_rate": 1.60866589274544e-05, "loss": 0.4857, "step": 10178 }, { "epoch": 3.13, "learning_rate": 1.6085868215773702e-05, "loss": 0.4848, "step": 10179 }, { "epoch": 3.13, "learning_rate": 1.6085077443654716e-05, "loss": 0.5595, "step": 10180 }, { "epoch": 3.13, "learning_rate": 1.608428661110531e-05, "loss": 0.5704, "step": 10181 }, { "epoch": 3.13, "learning_rate": 1.6083495718133326e-05, "loss": 0.4334, "step": 10182 }, { "epoch": 3.13, "learning_rate": 1.6082704764746622e-05, "loss": 0.482, "step": 10183 }, { "epoch": 3.13, "learning_rate": 1.6081913750953056e-05, "loss": 0.5465, "step": 10184 }, { "epoch": 3.13, "learning_rate": 1.6081122676760485e-05, "loss": 0.406, "step": 10185 }, { "epoch": 3.13, "learning_rate": 1.6080331542176754e-05, "loss": 0.4598, "step": 10186 }, { "epoch": 3.13, "learning_rate": 1.607954034720973e-05, "loss": 1.107, "step": 10187 }, { "epoch": 3.13, "learning_rate": 1.607874909186727e-05, "loss": 0.3863, "step": 10188 }, { "epoch": 3.13, "learning_rate": 1.6077957776157225e-05, "loss": 0.6789, "step": 10189 }, { "epoch": 3.14, "learning_rate": 1.607716640008746e-05, "loss": 0.4911, "step": 10190 }, { "epoch": 3.14, "learning_rate": 1.607637496366583e-05, "loss": 0.5589, "step": 10191 }, { "epoch": 3.14, "learning_rate": 1.6075583466900196e-05, "loss": 0.3405, "step": 10192 }, { "epoch": 3.14, "learning_rate": 1.6074791909798423e-05, "loss": 0.3341, "step": 10193 }, { "epoch": 3.14, "learning_rate": 1.6074000292368362e-05, "loss": 1.1465, "step": 10194 }, { "epoch": 3.14, "learning_rate": 1.6073208614617883e-05, "loss": 0.5531, "step": 10195 }, { "epoch": 3.14, "learning_rate": 1.6072416876554847e-05, "loss": 0.5518, "step": 10196 }, { "epoch": 3.14, "learning_rate": 1.6071625078187113e-05, "loss": 0.4284, "step": 10197 }, { "epoch": 3.14, "learning_rate": 1.6070833219522546e-05, "loss": 0.4918, "step": 10198 }, { "epoch": 3.14, "learning_rate": 1.6070041300569014e-05, "loss": 0.5574, "step": 10199 }, { "epoch": 3.14, "learning_rate": 1.606924932133437e-05, "loss": 0.5764, "step": 10200 }, { "epoch": 3.14, "learning_rate": 1.6068457281826492e-05, "loss": 0.3223, "step": 10201 }, { "epoch": 3.14, "learning_rate": 1.606766518205324e-05, "loss": 0.4722, "step": 10202 }, { "epoch": 3.14, "learning_rate": 1.6066873022022477e-05, "loss": 0.5488, "step": 10203 }, { "epoch": 3.14, "learning_rate": 1.606608080174208e-05, "loss": 0.6932, "step": 10204 }, { "epoch": 3.14, "learning_rate": 1.6065288521219905e-05, "loss": 0.4852, "step": 10205 }, { "epoch": 3.14, "learning_rate": 1.6064496180463826e-05, "loss": 0.5245, "step": 10206 }, { "epoch": 3.14, "learning_rate": 1.6063703779481712e-05, "loss": 0.441, "step": 10207 }, { "epoch": 3.14, "learning_rate": 1.6062911318281427e-05, "loss": 0.6622, "step": 10208 }, { "epoch": 3.14, "learning_rate": 1.6062118796870852e-05, "loss": 0.6793, "step": 10209 }, { "epoch": 3.14, "learning_rate": 1.6061326215257844e-05, "loss": 0.4725, "step": 10210 }, { "epoch": 3.14, "learning_rate": 1.606053357345028e-05, "loss": 0.5166, "step": 10211 }, { "epoch": 3.14, "learning_rate": 1.6059740871456035e-05, "loss": 0.5557, "step": 10212 }, { "epoch": 3.14, "learning_rate": 1.605894810928298e-05, "loss": 0.5963, "step": 10213 }, { "epoch": 3.14, "learning_rate": 1.6058155286938982e-05, "loss": 0.7502, "step": 10214 }, { "epoch": 3.14, "learning_rate": 1.6057362404431923e-05, "loss": 0.4047, "step": 10215 }, { "epoch": 3.14, "learning_rate": 1.6056569461769667e-05, "loss": 0.4633, "step": 10216 }, { "epoch": 3.14, "learning_rate": 1.60557764589601e-05, "loss": 0.3003, "step": 10217 }, { "epoch": 3.14, "learning_rate": 1.605498339601109e-05, "loss": 0.6688, "step": 10218 }, { "epoch": 3.14, "learning_rate": 1.6054190272930515e-05, "loss": 0.4346, "step": 10219 }, { "epoch": 3.14, "learning_rate": 1.6053397089726247e-05, "loss": 0.4264, "step": 10220 }, { "epoch": 3.14, "learning_rate": 1.605260384640617e-05, "loss": 0.5211, "step": 10221 }, { "epoch": 3.15, "learning_rate": 1.6051810542978157e-05, "loss": 0.481, "step": 10222 }, { "epoch": 3.15, "learning_rate": 1.605101717945009e-05, "loss": 0.458, "step": 10223 }, { "epoch": 3.15, "learning_rate": 1.6050223755829844e-05, "loss": 0.5684, "step": 10224 }, { "epoch": 3.15, "learning_rate": 1.60494302721253e-05, "loss": 0.5678, "step": 10225 }, { "epoch": 3.15, "learning_rate": 1.604863672834434e-05, "loss": 0.7281, "step": 10226 }, { "epoch": 3.15, "learning_rate": 1.6047843124494843e-05, "loss": 0.5737, "step": 10227 }, { "epoch": 3.15, "learning_rate": 1.6047049460584687e-05, "loss": 0.3922, "step": 10228 }, { "epoch": 3.15, "learning_rate": 1.6046255736621758e-05, "loss": 0.4755, "step": 10229 }, { "epoch": 3.15, "learning_rate": 1.6045461952613938e-05, "loss": 0.4722, "step": 10230 }, { "epoch": 3.15, "learning_rate": 1.6044668108569107e-05, "loss": 0.5973, "step": 10231 }, { "epoch": 3.15, "learning_rate": 1.6043874204495152e-05, "loss": 0.4725, "step": 10232 }, { "epoch": 3.15, "learning_rate": 1.6043080240399955e-05, "loss": 0.2969, "step": 10233 }, { "epoch": 3.15, "learning_rate": 1.6042286216291407e-05, "loss": 0.6394, "step": 10234 }, { "epoch": 3.15, "learning_rate": 1.6041492132177384e-05, "loss": 0.7072, "step": 10235 }, { "epoch": 3.15, "learning_rate": 1.6040697988065775e-05, "loss": 0.3338, "step": 10236 }, { "epoch": 3.15, "learning_rate": 1.6039903783964467e-05, "loss": 0.5013, "step": 10237 }, { "epoch": 3.15, "learning_rate": 1.603910951988135e-05, "loss": 0.8711, "step": 10238 }, { "epoch": 3.15, "learning_rate": 1.603831519582431e-05, "loss": 0.672, "step": 10239 }, { "epoch": 3.15, "learning_rate": 1.6037520811801233e-05, "loss": 0.52, "step": 10240 }, { "epoch": 3.15, "learning_rate": 1.603672636782001e-05, "loss": 0.4744, "step": 10241 }, { "epoch": 3.15, "learning_rate": 1.6035931863888535e-05, "loss": 0.4481, "step": 10242 }, { "epoch": 3.15, "learning_rate": 1.603513730001469e-05, "loss": 0.427, "step": 10243 }, { "epoch": 3.15, "learning_rate": 1.603434267620637e-05, "loss": 0.6393, "step": 10244 }, { "epoch": 3.15, "learning_rate": 1.6033547992471466e-05, "loss": 0.4721, "step": 10245 }, { "epoch": 3.15, "learning_rate": 1.6032753248817868e-05, "loss": 0.597, "step": 10246 }, { "epoch": 3.15, "learning_rate": 1.6031958445253474e-05, "loss": 0.5418, "step": 10247 }, { "epoch": 3.15, "learning_rate": 1.603116358178617e-05, "loss": 0.4445, "step": 10248 }, { "epoch": 3.15, "learning_rate": 1.6030368658423857e-05, "loss": 0.5277, "step": 10249 }, { "epoch": 3.15, "learning_rate": 1.602957367517442e-05, "loss": 0.4172, "step": 10250 }, { "epoch": 3.15, "learning_rate": 1.602877863204576e-05, "loss": 0.48, "step": 10251 }, { "epoch": 3.15, "learning_rate": 1.6027983529045777e-05, "loss": 0.5261, "step": 10252 }, { "epoch": 3.15, "learning_rate": 1.602718836618236e-05, "loss": 0.5735, "step": 10253 }, { "epoch": 3.15, "learning_rate": 1.6026393143463407e-05, "loss": 0.539, "step": 10254 }, { "epoch": 3.16, "learning_rate": 1.6025597860896816e-05, "loss": 0.4749, "step": 10255 }, { "epoch": 3.16, "learning_rate": 1.602480251849048e-05, "loss": 0.721, "step": 10256 }, { "epoch": 3.16, "learning_rate": 1.602400711625231e-05, "loss": 0.5674, "step": 10257 }, { "epoch": 3.16, "learning_rate": 1.6023211654190196e-05, "loss": 0.4695, "step": 10258 }, { "epoch": 3.16, "learning_rate": 1.602241613231204e-05, "loss": 0.5206, "step": 10259 }, { "epoch": 3.16, "learning_rate": 1.6021620550625738e-05, "loss": 0.4111, "step": 10260 }, { "epoch": 3.16, "learning_rate": 1.6020824909139195e-05, "loss": 0.6146, "step": 10261 }, { "epoch": 3.16, "learning_rate": 1.602002920786031e-05, "loss": 0.458, "step": 10262 }, { "epoch": 3.16, "learning_rate": 1.6019233446796992e-05, "loss": 0.3712, "step": 10263 }, { "epoch": 3.16, "learning_rate": 1.6018437625957135e-05, "loss": 0.4198, "step": 10264 }, { "epoch": 3.16, "learning_rate": 1.6017641745348647e-05, "loss": 0.6001, "step": 10265 }, { "epoch": 3.16, "learning_rate": 1.6016845804979427e-05, "loss": 0.4877, "step": 10266 }, { "epoch": 3.16, "learning_rate": 1.6016049804857386e-05, "loss": 0.6177, "step": 10267 }, { "epoch": 3.16, "learning_rate": 1.6015253744990424e-05, "loss": 0.3395, "step": 10268 }, { "epoch": 3.16, "learning_rate": 1.601445762538645e-05, "loss": 0.5837, "step": 10269 }, { "epoch": 3.16, "learning_rate": 1.6013661446053366e-05, "loss": 0.5737, "step": 10270 }, { "epoch": 3.16, "learning_rate": 1.6012865206999085e-05, "loss": 0.4126, "step": 10271 }, { "epoch": 3.16, "learning_rate": 1.601206890823151e-05, "loss": 0.5198, "step": 10272 }, { "epoch": 3.16, "learning_rate": 1.601127254975855e-05, "loss": 0.7691, "step": 10273 }, { "epoch": 3.16, "learning_rate": 1.6010476131588113e-05, "loss": 0.4196, "step": 10274 }, { "epoch": 3.16, "learning_rate": 1.600967965372811e-05, "loss": 0.6124, "step": 10275 }, { "epoch": 3.16, "learning_rate": 1.6008883116186448e-05, "loss": 0.4048, "step": 10276 }, { "epoch": 3.16, "learning_rate": 1.6008086518971037e-05, "loss": 0.5896, "step": 10277 }, { "epoch": 3.16, "learning_rate": 1.6007289862089793e-05, "loss": 0.4896, "step": 10278 }, { "epoch": 3.16, "learning_rate": 1.6006493145550622e-05, "loss": 0.5109, "step": 10279 }, { "epoch": 3.16, "learning_rate": 1.6005696369361436e-05, "loss": 0.3486, "step": 10280 }, { "epoch": 3.16, "learning_rate": 1.6004899533530153e-05, "loss": 0.5773, "step": 10281 }, { "epoch": 3.16, "learning_rate": 1.6004102638064685e-05, "loss": 0.5454, "step": 10282 }, { "epoch": 3.16, "learning_rate": 1.600330568297294e-05, "loss": 0.4131, "step": 10283 }, { "epoch": 3.16, "learning_rate": 1.6002508668262842e-05, "loss": 0.4722, "step": 10284 }, { "epoch": 3.16, "learning_rate": 1.6001711593942298e-05, "loss": 0.4179, "step": 10285 }, { "epoch": 3.16, "learning_rate": 1.6000914460019226e-05, "loss": 0.3396, "step": 10286 }, { "epoch": 3.17, "learning_rate": 1.6000117266501542e-05, "loss": 0.4193, "step": 10287 }, { "epoch": 3.17, "learning_rate": 1.5999320013397164e-05, "loss": 0.3054, "step": 10288 }, { "epoch": 3.17, "learning_rate": 1.5998522700714012e-05, "loss": 0.386, "step": 10289 }, { "epoch": 3.17, "learning_rate": 1.599772532846e-05, "loss": 0.4815, "step": 10290 }, { "epoch": 3.17, "learning_rate": 1.5996927896643047e-05, "loss": 0.4127, "step": 10291 }, { "epoch": 3.17, "learning_rate": 1.5996130405271072e-05, "loss": 0.5606, "step": 10292 }, { "epoch": 3.17, "learning_rate": 1.5995332854351996e-05, "loss": 0.5726, "step": 10293 }, { "epoch": 3.17, "learning_rate": 1.5994535243893742e-05, "loss": 0.3595, "step": 10294 }, { "epoch": 3.17, "learning_rate": 1.5993737573904226e-05, "loss": 0.5002, "step": 10295 }, { "epoch": 3.17, "learning_rate": 1.5992939844391374e-05, "loss": 0.6308, "step": 10296 }, { "epoch": 3.17, "learning_rate": 1.5992142055363104e-05, "loss": 0.4723, "step": 10297 }, { "epoch": 3.17, "learning_rate": 1.599134420682734e-05, "loss": 0.4369, "step": 10298 }, { "epoch": 3.17, "learning_rate": 1.599054629879201e-05, "loss": 0.6914, "step": 10299 }, { "epoch": 3.17, "learning_rate": 1.5989748331265033e-05, "loss": 0.4719, "step": 10300 }, { "epoch": 3.17, "learning_rate": 1.598895030425433e-05, "loss": 0.4089, "step": 10301 }, { "epoch": 3.17, "learning_rate": 1.5988152217767837e-05, "loss": 0.4737, "step": 10302 }, { "epoch": 3.17, "learning_rate": 1.598735407181347e-05, "loss": 0.5479, "step": 10303 }, { "epoch": 3.17, "learning_rate": 1.5986555866399157e-05, "loss": 0.3179, "step": 10304 }, { "epoch": 3.17, "learning_rate": 1.5985757601532836e-05, "loss": 0.4772, "step": 10305 }, { "epoch": 3.17, "learning_rate": 1.5984959277222416e-05, "loss": 0.4569, "step": 10306 }, { "epoch": 3.17, "learning_rate": 1.598416089347584e-05, "loss": 0.6485, "step": 10307 }, { "epoch": 3.17, "learning_rate": 1.5983362450301027e-05, "loss": 0.4215, "step": 10308 }, { "epoch": 3.17, "learning_rate": 1.598256394770591e-05, "loss": 0.6529, "step": 10309 }, { "epoch": 3.17, "learning_rate": 1.5981765385698425e-05, "loss": 0.4905, "step": 10310 }, { "epoch": 3.17, "learning_rate": 1.5980966764286494e-05, "loss": 0.5158, "step": 10311 }, { "epoch": 3.17, "learning_rate": 1.598016808347805e-05, "loss": 0.5578, "step": 10312 }, { "epoch": 3.17, "learning_rate": 1.5979369343281027e-05, "loss": 0.3335, "step": 10313 }, { "epoch": 3.17, "learning_rate": 1.5978570543703352e-05, "loss": 0.442, "step": 10314 }, { "epoch": 3.17, "learning_rate": 1.5977771684752966e-05, "loss": 0.5645, "step": 10315 }, { "epoch": 3.17, "learning_rate": 1.5976972766437796e-05, "loss": 0.396, "step": 10316 }, { "epoch": 3.17, "learning_rate": 1.5976173788765775e-05, "loss": 0.4585, "step": 10317 }, { "epoch": 3.17, "learning_rate": 1.5975374751744843e-05, "loss": 0.5807, "step": 10318 }, { "epoch": 3.17, "learning_rate": 1.5974575655382937e-05, "loss": 0.4882, "step": 10319 }, { "epoch": 3.18, "learning_rate": 1.5973776499687982e-05, "loss": 0.5031, "step": 10320 }, { "epoch": 3.18, "learning_rate": 1.5972977284667924e-05, "loss": 0.4116, "step": 10321 }, { "epoch": 3.18, "learning_rate": 1.5972178010330694e-05, "loss": 0.3371, "step": 10322 }, { "epoch": 3.18, "learning_rate": 1.5971378676684234e-05, "loss": 0.3756, "step": 10323 }, { "epoch": 3.18, "learning_rate": 1.597057928373648e-05, "loss": 0.5964, "step": 10324 }, { "epoch": 3.18, "learning_rate": 1.5969779831495375e-05, "loss": 0.3668, "step": 10325 }, { "epoch": 3.18, "learning_rate": 1.596898031996885e-05, "loss": 0.5674, "step": 10326 }, { "epoch": 3.18, "learning_rate": 1.596818074916485e-05, "loss": 0.5948, "step": 10327 }, { "epoch": 3.18, "learning_rate": 1.596738111909131e-05, "loss": 0.6334, "step": 10328 }, { "epoch": 3.18, "learning_rate": 1.596658142975618e-05, "loss": 0.338, "step": 10329 }, { "epoch": 3.18, "learning_rate": 1.5965781681167403e-05, "loss": 0.6786, "step": 10330 }, { "epoch": 3.18, "learning_rate": 1.596498187333291e-05, "loss": 0.6404, "step": 10331 }, { "epoch": 3.18, "learning_rate": 1.596418200626065e-05, "loss": 0.6506, "step": 10332 }, { "epoch": 3.18, "learning_rate": 1.5963382079958566e-05, "loss": 0.2772, "step": 10333 }, { "epoch": 3.18, "learning_rate": 1.59625820944346e-05, "loss": 0.447, "step": 10334 }, { "epoch": 3.18, "learning_rate": 1.5961782049696705e-05, "loss": 0.5139, "step": 10335 }, { "epoch": 3.18, "learning_rate": 1.5960981945752814e-05, "loss": 0.685, "step": 10336 }, { "epoch": 3.18, "learning_rate": 1.596018178261088e-05, "loss": 0.4233, "step": 10337 }, { "epoch": 3.18, "learning_rate": 1.5959381560278847e-05, "loss": 0.4625, "step": 10338 }, { "epoch": 3.18, "learning_rate": 1.5958581278764663e-05, "loss": 0.5632, "step": 10339 }, { "epoch": 3.18, "learning_rate": 1.5957780938076277e-05, "loss": 0.438, "step": 10340 }, { "epoch": 3.18, "learning_rate": 1.5956980538221633e-05, "loss": 0.5255, "step": 10341 }, { "epoch": 3.18, "learning_rate": 1.5956180079208684e-05, "loss": 0.6195, "step": 10342 }, { "epoch": 3.18, "learning_rate": 1.5955379561045377e-05, "loss": 0.6183, "step": 10343 }, { "epoch": 3.18, "learning_rate": 1.595457898373966e-05, "loss": 0.4897, "step": 10344 }, { "epoch": 3.18, "learning_rate": 1.595377834729949e-05, "loss": 0.5214, "step": 10345 }, { "epoch": 3.18, "learning_rate": 1.595297765173281e-05, "loss": 0.4097, "step": 10346 }, { "epoch": 3.18, "learning_rate": 1.5952176897047577e-05, "loss": 0.3113, "step": 10347 }, { "epoch": 3.18, "learning_rate": 1.5951376083251743e-05, "loss": 0.6534, "step": 10348 }, { "epoch": 3.18, "learning_rate": 1.5950575210353258e-05, "loss": 0.6578, "step": 10349 }, { "epoch": 3.18, "learning_rate": 1.5949774278360078e-05, "loss": 0.502, "step": 10350 }, { "epoch": 3.18, "learning_rate": 1.5948973287280157e-05, "loss": 0.6628, "step": 10351 }, { "epoch": 3.19, "learning_rate": 1.5948172237121443e-05, "loss": 0.463, "step": 10352 }, { "epoch": 3.19, "learning_rate": 1.59473711278919e-05, "loss": 0.7165, "step": 10353 }, { "epoch": 3.19, "learning_rate": 1.594656995959948e-05, "loss": 0.4157, "step": 10354 }, { "epoch": 3.19, "learning_rate": 1.5945768732252144e-05, "loss": 0.4448, "step": 10355 }, { "epoch": 3.19, "learning_rate": 1.5944967445857842e-05, "loss": 0.3395, "step": 10356 }, { "epoch": 3.19, "learning_rate": 1.5944166100424534e-05, "loss": 0.4583, "step": 10357 }, { "epoch": 3.19, "learning_rate": 1.5943364695960175e-05, "loss": 0.5176, "step": 10358 }, { "epoch": 3.19, "learning_rate": 1.5942563232472733e-05, "loss": 0.5039, "step": 10359 }, { "epoch": 3.19, "learning_rate": 1.594176170997016e-05, "loss": 0.4251, "step": 10360 }, { "epoch": 3.19, "learning_rate": 1.5940960128460413e-05, "loss": 0.5587, "step": 10361 }, { "epoch": 3.19, "learning_rate": 1.5940158487951454e-05, "loss": 0.4489, "step": 10362 }, { "epoch": 3.19, "learning_rate": 1.5939356788451256e-05, "loss": 0.5449, "step": 10363 }, { "epoch": 3.19, "learning_rate": 1.5938555029967765e-05, "loss": 0.4968, "step": 10364 }, { "epoch": 3.19, "learning_rate": 1.5937753212508952e-05, "loss": 0.583, "step": 10365 }, { "epoch": 3.19, "learning_rate": 1.5936951336082778e-05, "loss": 0.4206, "step": 10366 }, { "epoch": 3.19, "learning_rate": 1.59361494006972e-05, "loss": 0.418, "step": 10367 }, { "epoch": 3.19, "learning_rate": 1.5935347406360192e-05, "loss": 0.4213, "step": 10368 }, { "epoch": 3.19, "learning_rate": 1.5934545353079715e-05, "loss": 0.5756, "step": 10369 }, { "epoch": 3.19, "learning_rate": 1.5933743240863733e-05, "loss": 0.548, "step": 10370 }, { "epoch": 3.19, "learning_rate": 1.593294106972021e-05, "loss": 0.5101, "step": 10371 }, { "epoch": 3.19, "learning_rate": 1.5932138839657113e-05, "loss": 0.5562, "step": 10372 }, { "epoch": 3.19, "learning_rate": 1.5931336550682414e-05, "loss": 0.4606, "step": 10373 }, { "epoch": 3.19, "learning_rate": 1.5930534202804074e-05, "loss": 0.419, "step": 10374 }, { "epoch": 3.19, "learning_rate": 1.5929731796030067e-05, "loss": 0.5129, "step": 10375 }, { "epoch": 3.19, "learning_rate": 1.5928929330368354e-05, "loss": 0.5489, "step": 10376 }, { "epoch": 3.19, "learning_rate": 1.592812680582691e-05, "loss": 0.5787, "step": 10377 }, { "epoch": 3.19, "learning_rate": 1.5927324222413703e-05, "loss": 0.4712, "step": 10378 }, { "epoch": 3.19, "learning_rate": 1.5926521580136706e-05, "loss": 0.6104, "step": 10379 }, { "epoch": 3.19, "learning_rate": 1.5925718879003886e-05, "loss": 0.3368, "step": 10380 }, { "epoch": 3.19, "learning_rate": 1.5924916119023214e-05, "loss": 0.5003, "step": 10381 }, { "epoch": 3.19, "learning_rate": 1.5924113300202667e-05, "loss": 0.5497, "step": 10382 }, { "epoch": 3.19, "learning_rate": 1.5923310422550214e-05, "loss": 0.4387, "step": 10383 }, { "epoch": 3.19, "learning_rate": 1.592250748607383e-05, "loss": 0.4095, "step": 10384 }, { "epoch": 3.2, "learning_rate": 1.592170449078149e-05, "loss": 0.591, "step": 10385 }, { "epoch": 3.2, "learning_rate": 1.5920901436681163e-05, "loss": 0.3205, "step": 10386 }, { "epoch": 3.2, "learning_rate": 1.592009832378083e-05, "loss": 0.3314, "step": 10387 }, { "epoch": 3.2, "learning_rate": 1.5919295152088464e-05, "loss": 0.7858, "step": 10388 }, { "epoch": 3.2, "learning_rate": 1.591849192161204e-05, "loss": 0.4168, "step": 10389 }, { "epoch": 3.2, "learning_rate": 1.591768863235954e-05, "loss": 0.6274, "step": 10390 }, { "epoch": 3.2, "learning_rate": 1.5916885284338937e-05, "loss": 0.4934, "step": 10391 }, { "epoch": 3.2, "learning_rate": 1.591608187755821e-05, "loss": 0.2551, "step": 10392 }, { "epoch": 3.2, "learning_rate": 1.5915278412025333e-05, "loss": 0.3552, "step": 10393 }, { "epoch": 3.2, "learning_rate": 1.5914474887748297e-05, "loss": 0.5041, "step": 10394 }, { "epoch": 3.2, "learning_rate": 1.591367130473507e-05, "loss": 0.5581, "step": 10395 }, { "epoch": 3.2, "learning_rate": 1.5912867662993638e-05, "loss": 0.4553, "step": 10396 }, { "epoch": 3.2, "learning_rate": 1.591206396253198e-05, "loss": 0.5497, "step": 10397 }, { "epoch": 3.2, "learning_rate": 1.5911260203358077e-05, "loss": 0.4611, "step": 10398 }, { "epoch": 3.2, "learning_rate": 1.5910456385479916e-05, "loss": 0.7225, "step": 10399 }, { "epoch": 3.2, "learning_rate": 1.5909652508905472e-05, "loss": 0.628, "step": 10400 }, { "epoch": 3.2, "learning_rate": 1.5908848573642737e-05, "loss": 0.4104, "step": 10401 }, { "epoch": 3.2, "learning_rate": 1.5908044579699687e-05, "loss": 0.4554, "step": 10402 }, { "epoch": 3.2, "learning_rate": 1.590724052708431e-05, "loss": 0.6415, "step": 10403 }, { "epoch": 3.2, "learning_rate": 1.5906436415804587e-05, "loss": 0.5033, "step": 10404 }, { "epoch": 3.2, "learning_rate": 1.590563224586851e-05, "loss": 0.5528, "step": 10405 }, { "epoch": 3.2, "learning_rate": 1.5904828017284064e-05, "loss": 0.6214, "step": 10406 }, { "epoch": 3.2, "learning_rate": 1.5904023730059227e-05, "loss": 0.5661, "step": 10407 }, { "epoch": 3.2, "learning_rate": 1.5903219384201998e-05, "loss": 0.465, "step": 10408 }, { "epoch": 3.2, "learning_rate": 1.590241497972036e-05, "loss": 0.6335, "step": 10409 }, { "epoch": 3.2, "learning_rate": 1.59016105166223e-05, "loss": 0.7371, "step": 10410 }, { "epoch": 3.2, "learning_rate": 1.590080599491581e-05, "loss": 0.6904, "step": 10411 }, { "epoch": 3.2, "learning_rate": 1.5900001414608875e-05, "loss": 0.5115, "step": 10412 }, { "epoch": 3.2, "learning_rate": 1.589919677570949e-05, "loss": 0.6892, "step": 10413 }, { "epoch": 3.2, "learning_rate": 1.5898392078225646e-05, "loss": 0.6302, "step": 10414 }, { "epoch": 3.2, "learning_rate": 1.5897587322165332e-05, "loss": 0.4819, "step": 10415 }, { "epoch": 3.2, "learning_rate": 1.589678250753654e-05, "loss": 0.544, "step": 10416 }, { "epoch": 3.21, "learning_rate": 1.5895977634347262e-05, "loss": 0.5967, "step": 10417 }, { "epoch": 3.21, "learning_rate": 1.589517270260549e-05, "loss": 0.5836, "step": 10418 }, { "epoch": 3.21, "learning_rate": 1.5894367712319226e-05, "loss": 0.3927, "step": 10419 }, { "epoch": 3.21, "learning_rate": 1.589356266349645e-05, "loss": 0.5474, "step": 10420 }, { "epoch": 3.21, "learning_rate": 1.5892757556145174e-05, "loss": 0.3663, "step": 10421 }, { "epoch": 3.21, "learning_rate": 1.589195239027338e-05, "loss": 0.4797, "step": 10422 }, { "epoch": 3.21, "learning_rate": 1.5891147165889068e-05, "loss": 0.5829, "step": 10423 }, { "epoch": 3.21, "learning_rate": 1.5890341883000238e-05, "loss": 0.3853, "step": 10424 }, { "epoch": 3.21, "learning_rate": 1.5889536541614882e-05, "loss": 0.3586, "step": 10425 }, { "epoch": 3.21, "learning_rate": 1.5888731141741e-05, "loss": 0.5289, "step": 10426 }, { "epoch": 3.21, "learning_rate": 1.5887925683386594e-05, "loss": 0.352, "step": 10427 }, { "epoch": 3.21, "learning_rate": 1.5887120166559655e-05, "loss": 0.5836, "step": 10428 }, { "epoch": 3.21, "learning_rate": 1.5886314591268187e-05, "loss": 0.5109, "step": 10429 }, { "epoch": 3.21, "learning_rate": 1.5885508957520193e-05, "loss": 0.4338, "step": 10430 }, { "epoch": 3.21, "learning_rate": 1.588470326532367e-05, "loss": 0.5176, "step": 10431 }, { "epoch": 3.21, "learning_rate": 1.5883897514686625e-05, "loss": 0.4546, "step": 10432 }, { "epoch": 3.21, "learning_rate": 1.5883091705617045e-05, "loss": 0.5127, "step": 10433 }, { "epoch": 3.21, "learning_rate": 1.5882285838122948e-05, "loss": 0.6414, "step": 10434 }, { "epoch": 3.21, "learning_rate": 1.588147991221233e-05, "loss": 0.403, "step": 10435 }, { "epoch": 3.21, "learning_rate": 1.58806739278932e-05, "loss": 0.4319, "step": 10436 }, { "epoch": 3.21, "learning_rate": 1.5879867885173553e-05, "loss": 0.286, "step": 10437 }, { "epoch": 3.21, "learning_rate": 1.58790617840614e-05, "loss": 0.4949, "step": 10438 }, { "epoch": 3.21, "learning_rate": 1.5878255624564745e-05, "loss": 0.5656, "step": 10439 }, { "epoch": 3.21, "learning_rate": 1.5877449406691593e-05, "loss": 0.7114, "step": 10440 }, { "epoch": 3.21, "learning_rate": 1.5876643130449956e-05, "loss": 0.4543, "step": 10441 }, { "epoch": 3.21, "learning_rate": 1.5875836795847834e-05, "loss": 0.3239, "step": 10442 }, { "epoch": 3.21, "learning_rate": 1.587503040289324e-05, "loss": 0.6124, "step": 10443 }, { "epoch": 3.21, "learning_rate": 1.5874223951594175e-05, "loss": 0.5615, "step": 10444 }, { "epoch": 3.21, "learning_rate": 1.5873417441958654e-05, "loss": 0.5685, "step": 10445 }, { "epoch": 3.21, "learning_rate": 1.5872610873994685e-05, "loss": 0.5527, "step": 10446 }, { "epoch": 3.21, "learning_rate": 1.587180424771028e-05, "loss": 0.8029, "step": 10447 }, { "epoch": 3.21, "learning_rate": 1.5870997563113442e-05, "loss": 0.4493, "step": 10448 }, { "epoch": 3.21, "learning_rate": 1.587019082021219e-05, "loss": 0.6906, "step": 10449 }, { "epoch": 3.22, "learning_rate": 1.5869384019014535e-05, "loss": 0.5391, "step": 10450 }, { "epoch": 3.22, "learning_rate": 1.5868577159528484e-05, "loss": 0.4931, "step": 10451 }, { "epoch": 3.22, "learning_rate": 1.5867770241762053e-05, "loss": 0.5098, "step": 10452 }, { "epoch": 3.22, "learning_rate": 1.586696326572326e-05, "loss": 0.4973, "step": 10453 }, { "epoch": 3.22, "learning_rate": 1.5866156231420108e-05, "loss": 0.6353, "step": 10454 }, { "epoch": 3.22, "learning_rate": 1.5865349138860622e-05, "loss": 0.6427, "step": 10455 }, { "epoch": 3.22, "learning_rate": 1.5864541988052816e-05, "loss": 0.6462, "step": 10456 }, { "epoch": 3.22, "learning_rate": 1.58637347790047e-05, "loss": 0.3158, "step": 10457 }, { "epoch": 3.22, "learning_rate": 1.5862927511724297e-05, "loss": 0.5071, "step": 10458 }, { "epoch": 3.22, "learning_rate": 1.5862120186219614e-05, "loss": 0.5443, "step": 10459 }, { "epoch": 3.22, "learning_rate": 1.586131280249868e-05, "loss": 0.5614, "step": 10460 }, { "epoch": 3.22, "learning_rate": 1.5860505360569503e-05, "loss": 0.4788, "step": 10461 }, { "epoch": 3.22, "learning_rate": 1.5859697860440112e-05, "loss": 0.2691, "step": 10462 }, { "epoch": 3.22, "learning_rate": 1.5858890302118517e-05, "loss": 0.6085, "step": 10463 }, { "epoch": 3.22, "learning_rate": 1.5858082685612744e-05, "loss": 0.4761, "step": 10464 }, { "epoch": 3.22, "learning_rate": 1.5857275010930807e-05, "loss": 0.5074, "step": 10465 }, { "epoch": 3.22, "learning_rate": 1.5856467278080736e-05, "loss": 0.4876, "step": 10466 }, { "epoch": 3.22, "learning_rate": 1.5855659487070547e-05, "loss": 0.5058, "step": 10467 }, { "epoch": 3.22, "learning_rate": 1.585485163790826e-05, "loss": 0.5511, "step": 10468 }, { "epoch": 3.22, "learning_rate": 1.58540437306019e-05, "loss": 0.4385, "step": 10469 }, { "epoch": 3.22, "learning_rate": 1.5853235765159493e-05, "loss": 0.3167, "step": 10470 }, { "epoch": 3.22, "learning_rate": 1.585242774158906e-05, "loss": 0.2736, "step": 10471 }, { "epoch": 3.22, "learning_rate": 1.5851619659898623e-05, "loss": 0.4248, "step": 10472 }, { "epoch": 3.22, "learning_rate": 1.585081152009621e-05, "loss": 0.5496, "step": 10473 }, { "epoch": 3.22, "learning_rate": 1.585000332218985e-05, "loss": 0.6405, "step": 10474 }, { "epoch": 3.22, "learning_rate": 1.5849195066187565e-05, "loss": 0.4734, "step": 10475 }, { "epoch": 3.22, "learning_rate": 1.584838675209738e-05, "loss": 0.5269, "step": 10476 }, { "epoch": 3.22, "learning_rate": 1.5847578379927325e-05, "loss": 0.5081, "step": 10477 }, { "epoch": 3.22, "learning_rate": 1.5846769949685426e-05, "loss": 0.6624, "step": 10478 }, { "epoch": 3.22, "learning_rate": 1.5845961461379714e-05, "loss": 0.4632, "step": 10479 }, { "epoch": 3.22, "learning_rate": 1.584515291501822e-05, "loss": 0.4662, "step": 10480 }, { "epoch": 3.22, "learning_rate": 1.5844344310608968e-05, "loss": 0.4963, "step": 10481 }, { "epoch": 3.23, "learning_rate": 1.584353564815999e-05, "loss": 0.6186, "step": 10482 }, { "epoch": 3.23, "learning_rate": 1.5842726927679318e-05, "loss": 0.5315, "step": 10483 }, { "epoch": 3.23, "learning_rate": 1.584191814917498e-05, "loss": 0.5326, "step": 10484 }, { "epoch": 3.23, "learning_rate": 1.5841109312655017e-05, "loss": 0.3706, "step": 10485 }, { "epoch": 3.23, "learning_rate": 1.584030041812745e-05, "loss": 0.4725, "step": 10486 }, { "epoch": 3.23, "learning_rate": 1.5839491465600318e-05, "loss": 0.4996, "step": 10487 }, { "epoch": 3.23, "learning_rate": 1.5838682455081657e-05, "loss": 0.6333, "step": 10488 }, { "epoch": 3.23, "learning_rate": 1.5837873386579494e-05, "loss": 0.4444, "step": 10489 }, { "epoch": 3.23, "learning_rate": 1.5837064260101873e-05, "loss": 0.6114, "step": 10490 }, { "epoch": 3.23, "learning_rate": 1.5836255075656817e-05, "loss": 0.6658, "step": 10491 }, { "epoch": 3.23, "learning_rate": 1.5835445833252376e-05, "loss": 0.4725, "step": 10492 }, { "epoch": 3.23, "learning_rate": 1.5834636532896577e-05, "loss": 0.573, "step": 10493 }, { "epoch": 3.23, "learning_rate": 1.583382717459746e-05, "loss": 0.4381, "step": 10494 }, { "epoch": 3.23, "learning_rate": 1.5833017758363064e-05, "loss": 0.46, "step": 10495 }, { "epoch": 3.23, "learning_rate": 1.5832208284201423e-05, "loss": 0.3618, "step": 10496 }, { "epoch": 3.23, "learning_rate": 1.583139875212058e-05, "loss": 0.5171, "step": 10497 }, { "epoch": 3.23, "learning_rate": 1.5830589162128574e-05, "loss": 0.3627, "step": 10498 }, { "epoch": 3.23, "learning_rate": 1.5829779514233443e-05, "loss": 0.5307, "step": 10499 }, { "epoch": 3.23, "learning_rate": 1.5828969808443226e-05, "loss": 0.5905, "step": 10500 }, { "epoch": 3.23, "learning_rate": 1.5828160044765972e-05, "loss": 0.6539, "step": 10501 }, { "epoch": 3.23, "learning_rate": 1.5827350223209716e-05, "loss": 0.5266, "step": 10502 }, { "epoch": 3.23, "learning_rate": 1.58265403437825e-05, "loss": 0.3941, "step": 10503 }, { "epoch": 3.23, "learning_rate": 1.5825730406492367e-05, "loss": 0.4345, "step": 10504 }, { "epoch": 3.23, "learning_rate": 1.5824920411347364e-05, "loss": 0.4871, "step": 10505 }, { "epoch": 3.23, "learning_rate": 1.5824110358355534e-05, "loss": 0.4662, "step": 10506 }, { "epoch": 3.23, "learning_rate": 1.582330024752492e-05, "loss": 0.4355, "step": 10507 }, { "epoch": 3.23, "learning_rate": 1.5822490078863567e-05, "loss": 0.8134, "step": 10508 }, { "epoch": 3.23, "learning_rate": 1.5821679852379518e-05, "loss": 0.4767, "step": 10509 }, { "epoch": 3.23, "learning_rate": 1.5820869568080825e-05, "loss": 0.4746, "step": 10510 }, { "epoch": 3.23, "learning_rate": 1.582005922597553e-05, "loss": 0.8396, "step": 10511 }, { "epoch": 3.23, "learning_rate": 1.581924882607169e-05, "loss": 0.6919, "step": 10512 }, { "epoch": 3.23, "learning_rate": 1.581843836837734e-05, "loss": 0.4262, "step": 10513 }, { "epoch": 3.23, "learning_rate": 1.5817627852900537e-05, "loss": 0.5257, "step": 10514 }, { "epoch": 3.24, "learning_rate": 1.581681727964933e-05, "loss": 0.5342, "step": 10515 }, { "epoch": 3.24, "learning_rate": 1.581600664863176e-05, "loss": 0.5786, "step": 10516 }, { "epoch": 3.24, "learning_rate": 1.5815195959855893e-05, "loss": 0.5749, "step": 10517 }, { "epoch": 3.24, "learning_rate": 1.5814385213329765e-05, "loss": 0.4945, "step": 10518 }, { "epoch": 3.24, "learning_rate": 1.5813574409061432e-05, "loss": 0.3833, "step": 10519 }, { "epoch": 3.24, "learning_rate": 1.581276354705895e-05, "loss": 0.6222, "step": 10520 }, { "epoch": 3.24, "learning_rate": 1.5811952627330364e-05, "loss": 0.6599, "step": 10521 }, { "epoch": 3.24, "learning_rate": 1.5811141649883736e-05, "loss": 0.5001, "step": 10522 }, { "epoch": 3.24, "learning_rate": 1.5810330614727116e-05, "loss": 0.3114, "step": 10523 }, { "epoch": 3.24, "learning_rate": 1.580951952186856e-05, "loss": 0.442, "step": 10524 }, { "epoch": 3.24, "learning_rate": 1.580870837131612e-05, "loss": 0.4992, "step": 10525 }, { "epoch": 3.24, "learning_rate": 1.5807897163077844e-05, "loss": 0.5581, "step": 10526 }, { "epoch": 3.24, "learning_rate": 1.5807085897161806e-05, "loss": 0.4628, "step": 10527 }, { "epoch": 3.24, "learning_rate": 1.580627457357605e-05, "loss": 0.6286, "step": 10528 }, { "epoch": 3.24, "learning_rate": 1.5805463192328638e-05, "loss": 0.6123, "step": 10529 }, { "epoch": 3.24, "learning_rate": 1.580465175342762e-05, "loss": 0.6159, "step": 10530 }, { "epoch": 3.24, "learning_rate": 1.5803840256881066e-05, "loss": 0.4903, "step": 10531 }, { "epoch": 3.24, "learning_rate": 1.580302870269703e-05, "loss": 0.6327, "step": 10532 }, { "epoch": 3.24, "learning_rate": 1.5802217090883568e-05, "loss": 0.8255, "step": 10533 }, { "epoch": 3.24, "learning_rate": 1.5801405421448742e-05, "loss": 0.4741, "step": 10534 }, { "epoch": 3.24, "learning_rate": 1.5800593694400615e-05, "loss": 0.5286, "step": 10535 }, { "epoch": 3.24, "learning_rate": 1.579978190974725e-05, "loss": 0.4428, "step": 10536 }, { "epoch": 3.24, "learning_rate": 1.57989700674967e-05, "loss": 0.4816, "step": 10537 }, { "epoch": 3.24, "learning_rate": 1.5798158167657035e-05, "loss": 0.5404, "step": 10538 }, { "epoch": 3.24, "learning_rate": 1.5797346210236316e-05, "loss": 0.4635, "step": 10539 }, { "epoch": 3.24, "learning_rate": 1.5796534195242602e-05, "loss": 0.3839, "step": 10540 }, { "epoch": 3.24, "learning_rate": 1.5795722122683966e-05, "loss": 0.4523, "step": 10541 }, { "epoch": 3.24, "learning_rate": 1.579490999256847e-05, "loss": 0.4605, "step": 10542 }, { "epoch": 3.24, "learning_rate": 1.579409780490417e-05, "loss": 0.4958, "step": 10543 }, { "epoch": 3.24, "learning_rate": 1.5793285559699147e-05, "loss": 0.3429, "step": 10544 }, { "epoch": 3.24, "learning_rate": 1.5792473256961452e-05, "loss": 0.3811, "step": 10545 }, { "epoch": 3.24, "learning_rate": 1.579166089669916e-05, "loss": 0.5111, "step": 10546 }, { "epoch": 3.25, "learning_rate": 1.579084847892034e-05, "loss": 0.4694, "step": 10547 }, { "epoch": 3.25, "learning_rate": 1.5790036003633058e-05, "loss": 0.5829, "step": 10548 }, { "epoch": 3.25, "learning_rate": 1.5789223470845382e-05, "loss": 0.4704, "step": 10549 }, { "epoch": 3.25, "learning_rate": 1.578841088056538e-05, "loss": 0.5221, "step": 10550 }, { "epoch": 3.25, "learning_rate": 1.5787598232801123e-05, "loss": 0.8326, "step": 10551 }, { "epoch": 3.25, "learning_rate": 1.5786785527560683e-05, "loss": 0.4311, "step": 10552 }, { "epoch": 3.25, "learning_rate": 1.5785972764852128e-05, "loss": 0.4519, "step": 10553 }, { "epoch": 3.25, "learning_rate": 1.578515994468353e-05, "loss": 0.4529, "step": 10554 }, { "epoch": 3.25, "learning_rate": 1.5784347067062965e-05, "loss": 0.3303, "step": 10555 }, { "epoch": 3.25, "learning_rate": 1.57835341319985e-05, "loss": 0.5121, "step": 10556 }, { "epoch": 3.25, "learning_rate": 1.5782721139498213e-05, "loss": 0.5324, "step": 10557 }, { "epoch": 3.25, "learning_rate": 1.5781908089570174e-05, "loss": 0.7009, "step": 10558 }, { "epoch": 3.25, "learning_rate": 1.578109498222246e-05, "loss": 0.4928, "step": 10559 }, { "epoch": 3.25, "learning_rate": 1.5780281817463142e-05, "loss": 0.3997, "step": 10560 }, { "epoch": 3.25, "learning_rate": 1.57794685953003e-05, "loss": 0.5457, "step": 10561 }, { "epoch": 3.25, "learning_rate": 1.577865531574201e-05, "loss": 0.4122, "step": 10562 }, { "epoch": 3.25, "learning_rate": 1.5777841978796348e-05, "loss": 0.6273, "step": 10563 }, { "epoch": 3.25, "learning_rate": 1.5777028584471387e-05, "loss": 0.5298, "step": 10564 }, { "epoch": 3.25, "learning_rate": 1.5776215132775207e-05, "loss": 0.5444, "step": 10565 }, { "epoch": 3.25, "learning_rate": 1.5775401623715893e-05, "loss": 0.3847, "step": 10566 }, { "epoch": 3.25, "learning_rate": 1.577458805730151e-05, "loss": 0.5168, "step": 10567 }, { "epoch": 3.25, "learning_rate": 1.5773774433540153e-05, "loss": 0.6641, "step": 10568 }, { "epoch": 3.25, "learning_rate": 1.577296075243989e-05, "loss": 0.4495, "step": 10569 }, { "epoch": 3.25, "learning_rate": 1.5772147014008806e-05, "loss": 0.6469, "step": 10570 }, { "epoch": 3.25, "learning_rate": 1.5771333218254986e-05, "loss": 0.4272, "step": 10571 }, { "epoch": 3.25, "learning_rate": 1.5770519365186508e-05, "loss": 0.3965, "step": 10572 }, { "epoch": 3.25, "learning_rate": 1.576970545481145e-05, "loss": 0.4932, "step": 10573 }, { "epoch": 3.25, "learning_rate": 1.5768891487137904e-05, "loss": 0.6365, "step": 10574 }, { "epoch": 3.25, "learning_rate": 1.5768077462173942e-05, "loss": 0.288, "step": 10575 }, { "epoch": 3.25, "learning_rate": 1.5767263379927663e-05, "loss": 0.2895, "step": 10576 }, { "epoch": 3.25, "learning_rate": 1.5766449240407137e-05, "loss": 0.6883, "step": 10577 }, { "epoch": 3.25, "learning_rate": 1.5765635043620458e-05, "loss": 0.6051, "step": 10578 }, { "epoch": 3.25, "learning_rate": 1.576482078957571e-05, "loss": 0.7077, "step": 10579 }, { "epoch": 3.26, "learning_rate": 1.576400647828098e-05, "loss": 0.5234, "step": 10580 }, { "epoch": 3.26, "learning_rate": 1.5763192109744353e-05, "loss": 0.4751, "step": 10581 }, { "epoch": 3.26, "learning_rate": 1.5762377683973915e-05, "loss": 0.6006, "step": 10582 }, { "epoch": 3.26, "learning_rate": 1.5761563200977754e-05, "loss": 0.4965, "step": 10583 }, { "epoch": 3.26, "learning_rate": 1.5760748660763962e-05, "loss": 0.4249, "step": 10584 }, { "epoch": 3.26, "learning_rate": 1.5759934063340627e-05, "loss": 0.521, "step": 10585 }, { "epoch": 3.26, "learning_rate": 1.5759119408715838e-05, "loss": 0.9293, "step": 10586 }, { "epoch": 3.26, "learning_rate": 1.5758304696897688e-05, "loss": 0.4843, "step": 10587 }, { "epoch": 3.26, "learning_rate": 1.5757489927894263e-05, "loss": 0.3476, "step": 10588 }, { "epoch": 3.26, "learning_rate": 1.5756675101713657e-05, "loss": 0.4784, "step": 10589 }, { "epoch": 3.26, "learning_rate": 1.5755860218363962e-05, "loss": 0.3417, "step": 10590 }, { "epoch": 3.26, "learning_rate": 1.575504527785327e-05, "loss": 0.373, "step": 10591 }, { "epoch": 3.26, "learning_rate": 1.5754230280189674e-05, "loss": 0.6131, "step": 10592 }, { "epoch": 3.26, "learning_rate": 1.575341522538127e-05, "loss": 0.8307, "step": 10593 }, { "epoch": 3.26, "learning_rate": 1.575260011343615e-05, "loss": 0.7156, "step": 10594 }, { "epoch": 3.26, "learning_rate": 1.5751784944362403e-05, "loss": 0.5928, "step": 10595 }, { "epoch": 3.26, "learning_rate": 1.5750969718168137e-05, "loss": 0.5868, "step": 10596 }, { "epoch": 3.26, "learning_rate": 1.575015443486144e-05, "loss": 0.3899, "step": 10597 }, { "epoch": 3.26, "learning_rate": 1.574933909445041e-05, "loss": 0.3804, "step": 10598 }, { "epoch": 3.26, "learning_rate": 1.5748523696943144e-05, "loss": 0.7054, "step": 10599 }, { "epoch": 3.26, "learning_rate": 1.5747708242347738e-05, "loss": 0.3458, "step": 10600 }, { "epoch": 3.26, "learning_rate": 1.5746892730672292e-05, "loss": 0.4027, "step": 10601 }, { "epoch": 3.26, "learning_rate": 1.5746077161924905e-05, "loss": 0.4907, "step": 10602 }, { "epoch": 3.26, "learning_rate": 1.5745261536113678e-05, "loss": 0.5275, "step": 10603 }, { "epoch": 3.26, "learning_rate": 1.5744445853246707e-05, "loss": 0.4397, "step": 10604 }, { "epoch": 3.26, "learning_rate": 1.57436301133321e-05, "loss": 0.6999, "step": 10605 }, { "epoch": 3.26, "learning_rate": 1.5742814316377945e-05, "loss": 0.7348, "step": 10606 }, { "epoch": 3.26, "learning_rate": 1.5741998462392352e-05, "loss": 0.7259, "step": 10607 }, { "epoch": 3.26, "learning_rate": 1.574118255138343e-05, "loss": 0.4102, "step": 10608 }, { "epoch": 3.26, "learning_rate": 1.5740366583359264e-05, "loss": 0.5326, "step": 10609 }, { "epoch": 3.26, "learning_rate": 1.5739550558327975e-05, "loss": 0.45, "step": 10610 }, { "epoch": 3.26, "learning_rate": 1.5738734476297653e-05, "loss": 0.4923, "step": 10611 }, { "epoch": 3.27, "learning_rate": 1.5737918337276414e-05, "loss": 0.5989, "step": 10612 }, { "epoch": 3.27, "learning_rate": 1.5737102141272357e-05, "loss": 0.4867, "step": 10613 }, { "epoch": 3.27, "learning_rate": 1.5736285888293585e-05, "loss": 0.4828, "step": 10614 }, { "epoch": 3.27, "learning_rate": 1.573546957834821e-05, "loss": 0.6385, "step": 10615 }, { "epoch": 3.27, "learning_rate": 1.573465321144433e-05, "loss": 0.4063, "step": 10616 }, { "epoch": 3.27, "learning_rate": 1.5733836787590066e-05, "loss": 0.6099, "step": 10617 }, { "epoch": 3.27, "learning_rate": 1.573302030679352e-05, "loss": 0.5647, "step": 10618 }, { "epoch": 3.27, "learning_rate": 1.5732203769062792e-05, "loss": 0.3825, "step": 10619 }, { "epoch": 3.27, "learning_rate": 1.5731387174406002e-05, "loss": 0.6679, "step": 10620 }, { "epoch": 3.27, "learning_rate": 1.573057052283125e-05, "loss": 0.5156, "step": 10621 }, { "epoch": 3.27, "learning_rate": 1.5729753814346655e-05, "loss": 0.4975, "step": 10622 }, { "epoch": 3.27, "learning_rate": 1.5728937048960323e-05, "loss": 0.4599, "step": 10623 }, { "epoch": 3.27, "learning_rate": 1.5728120226680366e-05, "loss": 0.7719, "step": 10624 }, { "epoch": 3.27, "learning_rate": 1.5727303347514897e-05, "loss": 0.4816, "step": 10625 }, { "epoch": 3.27, "learning_rate": 1.5726486411472025e-05, "loss": 0.3673, "step": 10626 }, { "epoch": 3.27, "learning_rate": 1.5725669418559867e-05, "loss": 0.3888, "step": 10627 }, { "epoch": 3.27, "learning_rate": 1.572485236878654e-05, "loss": 0.3806, "step": 10628 }, { "epoch": 3.27, "learning_rate": 1.5724035262160146e-05, "loss": 0.4357, "step": 10629 }, { "epoch": 3.27, "learning_rate": 1.5723218098688806e-05, "loss": 0.3995, "step": 10630 }, { "epoch": 3.27, "learning_rate": 1.5722400878380638e-05, "loss": 0.3221, "step": 10631 }, { "epoch": 3.27, "learning_rate": 1.5721583601243755e-05, "loss": 0.6531, "step": 10632 }, { "epoch": 3.27, "learning_rate": 1.5720766267286275e-05, "loss": 0.5007, "step": 10633 }, { "epoch": 3.27, "learning_rate": 1.571994887651631e-05, "loss": 0.3865, "step": 10634 }, { "epoch": 3.27, "learning_rate": 1.5719131428941985e-05, "loss": 0.4202, "step": 10635 }, { "epoch": 3.27, "learning_rate": 1.571831392457141e-05, "loss": 0.3561, "step": 10636 }, { "epoch": 3.27, "learning_rate": 1.5717496363412712e-05, "loss": 0.4522, "step": 10637 }, { "epoch": 3.27, "learning_rate": 1.5716678745474007e-05, "loss": 0.5126, "step": 10638 }, { "epoch": 3.27, "learning_rate": 1.571586107076341e-05, "loss": 0.4585, "step": 10639 }, { "epoch": 3.27, "learning_rate": 1.571504333928904e-05, "loss": 0.3938, "step": 10640 }, { "epoch": 3.27, "learning_rate": 1.5714225551059027e-05, "loss": 0.4145, "step": 10641 }, { "epoch": 3.27, "learning_rate": 1.5713407706081488e-05, "loss": 0.5111, "step": 10642 }, { "epoch": 3.27, "learning_rate": 1.5712589804364546e-05, "loss": 0.563, "step": 10643 }, { "epoch": 3.27, "learning_rate": 1.5711771845916322e-05, "loss": 0.5178, "step": 10644 }, { "epoch": 3.28, "learning_rate": 1.5710953830744938e-05, "loss": 0.4942, "step": 10645 }, { "epoch": 3.28, "learning_rate": 1.5710135758858518e-05, "loss": 0.6244, "step": 10646 }, { "epoch": 3.28, "learning_rate": 1.570931763026519e-05, "loss": 0.4709, "step": 10647 }, { "epoch": 3.28, "learning_rate": 1.5708499444973076e-05, "loss": 0.4786, "step": 10648 }, { "epoch": 3.28, "learning_rate": 1.57076812029903e-05, "loss": 0.4598, "step": 10649 }, { "epoch": 3.28, "learning_rate": 1.5706862904324992e-05, "loss": 0.431, "step": 10650 }, { "epoch": 3.28, "learning_rate": 1.5706044548985274e-05, "loss": 0.4367, "step": 10651 }, { "epoch": 3.28, "learning_rate": 1.5705226136979273e-05, "loss": 0.335, "step": 10652 }, { "epoch": 3.28, "learning_rate": 1.5704407668315122e-05, "loss": 0.6835, "step": 10653 }, { "epoch": 3.28, "learning_rate": 1.570358914300094e-05, "loss": 0.7259, "step": 10654 }, { "epoch": 3.28, "learning_rate": 1.5702770561044866e-05, "loss": 0.5483, "step": 10655 }, { "epoch": 3.28, "learning_rate": 1.5701951922455023e-05, "loss": 0.523, "step": 10656 }, { "epoch": 3.28, "learning_rate": 1.570113322723954e-05, "loss": 0.5334, "step": 10657 }, { "epoch": 3.28, "learning_rate": 1.5700314475406557e-05, "loss": 0.4104, "step": 10658 }, { "epoch": 3.28, "learning_rate": 1.5699495666964192e-05, "loss": 0.5254, "step": 10659 }, { "epoch": 3.28, "learning_rate": 1.569867680192058e-05, "loss": 0.7527, "step": 10660 }, { "epoch": 3.28, "learning_rate": 1.569785788028386e-05, "loss": 0.5042, "step": 10661 }, { "epoch": 3.28, "learning_rate": 1.5697038902062157e-05, "loss": 0.5418, "step": 10662 }, { "epoch": 3.28, "learning_rate": 1.569621986726361e-05, "loss": 0.4926, "step": 10663 }, { "epoch": 3.28, "learning_rate": 1.5695400775896348e-05, "loss": 0.5577, "step": 10664 }, { "epoch": 3.28, "learning_rate": 1.5694581627968504e-05, "loss": 0.5701, "step": 10665 }, { "epoch": 3.28, "learning_rate": 1.569376242348822e-05, "loss": 0.6001, "step": 10666 }, { "epoch": 3.28, "learning_rate": 1.5692943162463628e-05, "loss": 0.4425, "step": 10667 }, { "epoch": 3.28, "learning_rate": 1.5692123844902867e-05, "loss": 0.5797, "step": 10668 }, { "epoch": 3.28, "learning_rate": 1.5691304470814063e-05, "loss": 0.7533, "step": 10669 }, { "epoch": 3.28, "learning_rate": 1.5690485040205365e-05, "loss": 0.6156, "step": 10670 }, { "epoch": 3.28, "learning_rate": 1.5689665553084902e-05, "loss": 0.7857, "step": 10671 }, { "epoch": 3.28, "learning_rate": 1.568884600946082e-05, "loss": 0.6659, "step": 10672 }, { "epoch": 3.28, "learning_rate": 1.5688026409341253e-05, "loss": 0.3861, "step": 10673 }, { "epoch": 3.28, "learning_rate": 1.5687206752734342e-05, "loss": 0.6032, "step": 10674 }, { "epoch": 3.28, "learning_rate": 1.5686387039648227e-05, "loss": 0.4903, "step": 10675 }, { "epoch": 3.28, "learning_rate": 1.5685567270091045e-05, "loss": 0.3703, "step": 10676 }, { "epoch": 3.29, "learning_rate": 1.5684747444070942e-05, "loss": 0.4582, "step": 10677 }, { "epoch": 3.29, "learning_rate": 1.568392756159606e-05, "loss": 0.5136, "step": 10678 }, { "epoch": 3.29, "learning_rate": 1.5683107622674534e-05, "loss": 0.3901, "step": 10679 }, { "epoch": 3.29, "learning_rate": 1.5682287627314513e-05, "loss": 0.7318, "step": 10680 }, { "epoch": 3.29, "learning_rate": 1.5681467575524144e-05, "loss": 0.497, "step": 10681 }, { "epoch": 3.29, "learning_rate": 1.568064746731156e-05, "loss": 0.4779, "step": 10682 }, { "epoch": 3.29, "learning_rate": 1.5679827302684912e-05, "loss": 0.5343, "step": 10683 }, { "epoch": 3.29, "learning_rate": 1.5679007081652346e-05, "loss": 0.5308, "step": 10684 }, { "epoch": 3.29, "learning_rate": 1.5678186804222005e-05, "loss": 0.5109, "step": 10685 }, { "epoch": 3.29, "learning_rate": 1.5677366470402035e-05, "loss": 0.4448, "step": 10686 }, { "epoch": 3.29, "learning_rate": 1.5676546080200587e-05, "loss": 0.6379, "step": 10687 }, { "epoch": 3.29, "learning_rate": 1.56757256336258e-05, "loss": 0.5729, "step": 10688 }, { "epoch": 3.29, "learning_rate": 1.567490513068583e-05, "loss": 0.6142, "step": 10689 }, { "epoch": 3.29, "learning_rate": 1.5674084571388818e-05, "loss": 0.5561, "step": 10690 }, { "epoch": 3.29, "learning_rate": 1.567326395574292e-05, "loss": 0.3655, "step": 10691 }, { "epoch": 3.29, "learning_rate": 1.567244328375628e-05, "loss": 0.3602, "step": 10692 }, { "epoch": 3.29, "learning_rate": 1.5671622555437055e-05, "loss": 0.479, "step": 10693 }, { "epoch": 3.29, "learning_rate": 1.5670801770793387e-05, "loss": 0.3875, "step": 10694 }, { "epoch": 3.29, "learning_rate": 1.5669980929833433e-05, "loss": 0.4803, "step": 10695 }, { "epoch": 3.29, "learning_rate": 1.566916003256534e-05, "loss": 0.4408, "step": 10696 }, { "epoch": 3.29, "learning_rate": 1.5668339078997264e-05, "loss": 0.6356, "step": 10697 }, { "epoch": 3.29, "learning_rate": 1.5667518069137358e-05, "loss": 0.5456, "step": 10698 }, { "epoch": 3.29, "learning_rate": 1.5666697002993776e-05, "loss": 0.4207, "step": 10699 }, { "epoch": 3.29, "learning_rate": 1.5665875880574666e-05, "loss": 0.5401, "step": 10700 }, { "epoch": 3.29, "learning_rate": 1.5665054701888188e-05, "loss": 0.5953, "step": 10701 }, { "epoch": 3.29, "learning_rate": 1.56642334669425e-05, "loss": 1.0033, "step": 10702 }, { "epoch": 3.29, "learning_rate": 1.5663412175745746e-05, "loss": 0.5439, "step": 10703 }, { "epoch": 3.29, "learning_rate": 1.5662590828306096e-05, "loss": 0.6873, "step": 10704 }, { "epoch": 3.29, "learning_rate": 1.56617694246317e-05, "loss": 0.5566, "step": 10705 }, { "epoch": 3.29, "learning_rate": 1.566094796473071e-05, "loss": 0.3391, "step": 10706 }, { "epoch": 3.29, "learning_rate": 1.5660126448611292e-05, "loss": 0.5993, "step": 10707 }, { "epoch": 3.29, "learning_rate": 1.56593048762816e-05, "loss": 0.3799, "step": 10708 }, { "epoch": 3.29, "learning_rate": 1.56584832477498e-05, "loss": 0.5699, "step": 10709 }, { "epoch": 3.3, "learning_rate": 1.5657661563024044e-05, "loss": 0.6118, "step": 10710 }, { "epoch": 3.3, "learning_rate": 1.565683982211249e-05, "loss": 0.516, "step": 10711 }, { "epoch": 3.3, "learning_rate": 1.5656018025023308e-05, "loss": 0.4536, "step": 10712 }, { "epoch": 3.3, "learning_rate": 1.5655196171764656e-05, "loss": 0.4959, "step": 10713 }, { "epoch": 3.3, "learning_rate": 1.565437426234469e-05, "loss": 0.5802, "step": 10714 }, { "epoch": 3.3, "learning_rate": 1.565355229677158e-05, "loss": 0.5472, "step": 10715 }, { "epoch": 3.3, "learning_rate": 1.5652730275053484e-05, "loss": 0.4613, "step": 10716 }, { "epoch": 3.3, "learning_rate": 1.5651908197198564e-05, "loss": 0.5307, "step": 10717 }, { "epoch": 3.3, "learning_rate": 1.565108606321499e-05, "loss": 0.3665, "step": 10718 }, { "epoch": 3.3, "learning_rate": 1.565026387311092e-05, "loss": 0.7233, "step": 10719 }, { "epoch": 3.3, "learning_rate": 1.5649441626894524e-05, "loss": 0.4412, "step": 10720 }, { "epoch": 3.3, "learning_rate": 1.5648619324573965e-05, "loss": 0.4861, "step": 10721 }, { "epoch": 3.3, "learning_rate": 1.564779696615741e-05, "loss": 0.6412, "step": 10722 }, { "epoch": 3.3, "learning_rate": 1.5646974551653026e-05, "loss": 0.4797, "step": 10723 }, { "epoch": 3.3, "learning_rate": 1.5646152081068983e-05, "loss": 0.474, "step": 10724 }, { "epoch": 3.3, "learning_rate": 1.5645329554413445e-05, "loss": 0.7004, "step": 10725 }, { "epoch": 3.3, "learning_rate": 1.564450697169458e-05, "loss": 0.5451, "step": 10726 }, { "epoch": 3.3, "learning_rate": 1.564368433292056e-05, "loss": 0.679, "step": 10727 }, { "epoch": 3.3, "learning_rate": 1.564286163809955e-05, "loss": 0.5533, "step": 10728 }, { "epoch": 3.3, "learning_rate": 1.564203888723973e-05, "loss": 0.5026, "step": 10729 }, { "epoch": 3.3, "learning_rate": 1.564121608034926e-05, "loss": 0.6299, "step": 10730 }, { "epoch": 3.3, "learning_rate": 1.5640393217436316e-05, "loss": 0.4562, "step": 10731 }, { "epoch": 3.3, "learning_rate": 1.5639570298509067e-05, "loss": 0.555, "step": 10732 }, { "epoch": 3.3, "learning_rate": 1.5638747323575688e-05, "loss": 0.596, "step": 10733 }, { "epoch": 3.3, "learning_rate": 1.5637924292644354e-05, "loss": 0.3989, "step": 10734 }, { "epoch": 3.3, "learning_rate": 1.5637101205723235e-05, "loss": 0.5843, "step": 10735 }, { "epoch": 3.3, "learning_rate": 1.56362780628205e-05, "loss": 0.5515, "step": 10736 }, { "epoch": 3.3, "learning_rate": 1.5635454863944333e-05, "loss": 0.4532, "step": 10737 }, { "epoch": 3.3, "learning_rate": 1.5634631609102906e-05, "loss": 0.3742, "step": 10738 }, { "epoch": 3.3, "learning_rate": 1.5633808298304395e-05, "loss": 0.3261, "step": 10739 }, { "epoch": 3.3, "learning_rate": 1.5632984931556976e-05, "loss": 0.4511, "step": 10740 }, { "epoch": 3.3, "learning_rate": 1.5632161508868825e-05, "loss": 0.4562, "step": 10741 }, { "epoch": 3.31, "learning_rate": 1.5631338030248117e-05, "loss": 0.5286, "step": 10742 }, { "epoch": 3.31, "learning_rate": 1.5630514495703034e-05, "loss": 0.5019, "step": 10743 }, { "epoch": 3.31, "learning_rate": 1.5629690905241752e-05, "loss": 0.3397, "step": 10744 }, { "epoch": 3.31, "learning_rate": 1.562886725887245e-05, "loss": 0.5656, "step": 10745 }, { "epoch": 3.31, "learning_rate": 1.5628043556603308e-05, "loss": 0.565, "step": 10746 }, { "epoch": 3.31, "learning_rate": 1.562721979844251e-05, "loss": 0.4529, "step": 10747 }, { "epoch": 3.31, "learning_rate": 1.562639598439823e-05, "loss": 0.5408, "step": 10748 }, { "epoch": 3.31, "learning_rate": 1.5625572114478656e-05, "loss": 0.4677, "step": 10749 }, { "epoch": 3.31, "learning_rate": 1.5624748188691965e-05, "loss": 0.4963, "step": 10750 }, { "epoch": 3.31, "learning_rate": 1.562392420704634e-05, "loss": 0.5437, "step": 10751 }, { "epoch": 3.31, "learning_rate": 1.5623100169549965e-05, "loss": 0.4013, "step": 10752 }, { "epoch": 3.31, "learning_rate": 1.562227607621102e-05, "loss": 0.7354, "step": 10753 }, { "epoch": 3.31, "learning_rate": 1.562145192703769e-05, "loss": 0.5966, "step": 10754 }, { "epoch": 3.31, "learning_rate": 1.562062772203817e-05, "loss": 0.51, "step": 10755 }, { "epoch": 3.31, "learning_rate": 1.5619803461220628e-05, "loss": 0.5083, "step": 10756 }, { "epoch": 3.31, "learning_rate": 1.5618979144593264e-05, "loss": 0.2535, "step": 10757 }, { "epoch": 3.31, "learning_rate": 1.5618154772164257e-05, "loss": 0.5585, "step": 10758 }, { "epoch": 3.31, "learning_rate": 1.5617330343941793e-05, "loss": 0.5611, "step": 10759 }, { "epoch": 3.31, "learning_rate": 1.5616505859934064e-05, "loss": 0.5351, "step": 10760 }, { "epoch": 3.31, "learning_rate": 1.5615681320149257e-05, "loss": 0.4096, "step": 10761 }, { "epoch": 3.31, "learning_rate": 1.5614856724595556e-05, "loss": 0.3517, "step": 10762 }, { "epoch": 3.31, "learning_rate": 1.5614032073281154e-05, "loss": 0.5075, "step": 10763 }, { "epoch": 3.31, "learning_rate": 1.561320736621424e-05, "loss": 0.7694, "step": 10764 }, { "epoch": 3.31, "learning_rate": 1.5612382603403003e-05, "loss": 0.2987, "step": 10765 }, { "epoch": 3.31, "learning_rate": 1.5611557784855634e-05, "loss": 0.4955, "step": 10766 }, { "epoch": 3.31, "learning_rate": 1.5610732910580324e-05, "loss": 0.5194, "step": 10767 }, { "epoch": 3.31, "learning_rate": 1.5609907980585262e-05, "loss": 0.4474, "step": 10768 }, { "epoch": 3.31, "learning_rate": 1.5609082994878652e-05, "loss": 0.3504, "step": 10769 }, { "epoch": 3.31, "learning_rate": 1.560825795346867e-05, "loss": 0.5045, "step": 10770 }, { "epoch": 3.31, "learning_rate": 1.5607432856363523e-05, "loss": 0.3004, "step": 10771 }, { "epoch": 3.31, "learning_rate": 1.56066077035714e-05, "loss": 0.4575, "step": 10772 }, { "epoch": 3.31, "learning_rate": 1.560578249510049e-05, "loss": 0.6398, "step": 10773 }, { "epoch": 3.31, "learning_rate": 1.5604957230959003e-05, "loss": 0.5797, "step": 10774 }, { "epoch": 3.32, "learning_rate": 1.5604131911155117e-05, "loss": 0.3894, "step": 10775 }, { "epoch": 3.32, "learning_rate": 1.5603306535697036e-05, "loss": 0.6897, "step": 10776 }, { "epoch": 3.32, "learning_rate": 1.560248110459296e-05, "loss": 0.3929, "step": 10777 }, { "epoch": 3.32, "learning_rate": 1.560165561785108e-05, "loss": 0.4824, "step": 10778 }, { "epoch": 3.32, "learning_rate": 1.5600830075479604e-05, "loss": 0.435, "step": 10779 }, { "epoch": 3.32, "learning_rate": 1.5600004477486717e-05, "loss": 0.393, "step": 10780 }, { "epoch": 3.32, "learning_rate": 1.5599178823880624e-05, "loss": 0.5374, "step": 10781 }, { "epoch": 3.32, "learning_rate": 1.559835311466953e-05, "loss": 0.4476, "step": 10782 }, { "epoch": 3.32, "learning_rate": 1.5597527349861624e-05, "loss": 0.4758, "step": 10783 }, { "epoch": 3.32, "learning_rate": 1.559670152946512e-05, "loss": 0.2659, "step": 10784 }, { "epoch": 3.32, "learning_rate": 1.5595875653488205e-05, "loss": 0.3199, "step": 10785 }, { "epoch": 3.32, "learning_rate": 1.559504972193909e-05, "loss": 0.452, "step": 10786 }, { "epoch": 3.32, "learning_rate": 1.5594223734825974e-05, "loss": 0.7072, "step": 10787 }, { "epoch": 3.32, "learning_rate": 1.559339769215706e-05, "loss": 0.5387, "step": 10788 }, { "epoch": 3.32, "learning_rate": 1.5592571593940554e-05, "loss": 0.5167, "step": 10789 }, { "epoch": 3.32, "learning_rate": 1.5591745440184657e-05, "loss": 0.6109, "step": 10790 }, { "epoch": 3.32, "learning_rate": 1.5590919230897575e-05, "loss": 0.5714, "step": 10791 }, { "epoch": 3.32, "learning_rate": 1.559009296608751e-05, "loss": 0.4851, "step": 10792 }, { "epoch": 3.32, "learning_rate": 1.558926664576267e-05, "loss": 0.5779, "step": 10793 }, { "epoch": 3.32, "learning_rate": 1.558844026993126e-05, "loss": 0.5041, "step": 10794 }, { "epoch": 3.32, "learning_rate": 1.5587613838601495e-05, "loss": 0.4742, "step": 10795 }, { "epoch": 3.32, "learning_rate": 1.5586787351781567e-05, "loss": 0.6369, "step": 10796 }, { "epoch": 3.32, "learning_rate": 1.5585960809479698e-05, "loss": 0.5738, "step": 10797 }, { "epoch": 3.32, "learning_rate": 1.5585134211704085e-05, "loss": 0.4519, "step": 10798 }, { "epoch": 3.32, "learning_rate": 1.5584307558462942e-05, "loss": 0.4982, "step": 10799 }, { "epoch": 3.32, "learning_rate": 1.5583480849764482e-05, "loss": 0.5262, "step": 10800 }, { "epoch": 3.32, "learning_rate": 1.5582654085616908e-05, "loss": 0.4154, "step": 10801 }, { "epoch": 3.32, "learning_rate": 1.558182726602843e-05, "loss": 0.6119, "step": 10802 }, { "epoch": 3.32, "learning_rate": 1.5581000391007273e-05, "loss": 0.5415, "step": 10803 }, { "epoch": 3.32, "learning_rate": 1.5580173460561632e-05, "loss": 0.5037, "step": 10804 }, { "epoch": 3.32, "learning_rate": 1.5579346474699728e-05, "loss": 0.6872, "step": 10805 }, { "epoch": 3.32, "learning_rate": 1.557851943342977e-05, "loss": 0.5125, "step": 10806 }, { "epoch": 3.33, "learning_rate": 1.5577692336759974e-05, "loss": 0.602, "step": 10807 }, { "epoch": 3.33, "learning_rate": 1.5576865184698547e-05, "loss": 0.5165, "step": 10808 }, { "epoch": 3.33, "learning_rate": 1.5576037977253715e-05, "loss": 0.4191, "step": 10809 }, { "epoch": 3.33, "learning_rate": 1.5575210714433687e-05, "loss": 0.5358, "step": 10810 }, { "epoch": 3.33, "learning_rate": 1.5574383396246676e-05, "loss": 0.5579, "step": 10811 }, { "epoch": 3.33, "learning_rate": 1.5573556022700902e-05, "loss": 0.5292, "step": 10812 }, { "epoch": 3.33, "learning_rate": 1.5572728593804578e-05, "loss": 0.6175, "step": 10813 }, { "epoch": 3.33, "learning_rate": 1.5571901109565923e-05, "loss": 0.4541, "step": 10814 }, { "epoch": 3.33, "learning_rate": 1.557107356999316e-05, "loss": 0.581, "step": 10815 }, { "epoch": 3.33, "learning_rate": 1.5570245975094495e-05, "loss": 0.5027, "step": 10816 }, { "epoch": 3.33, "learning_rate": 1.5569418324878155e-05, "loss": 0.4288, "step": 10817 }, { "epoch": 3.33, "learning_rate": 1.556859061935236e-05, "loss": 0.5347, "step": 10818 }, { "epoch": 3.33, "learning_rate": 1.5567762858525327e-05, "loss": 0.5408, "step": 10819 }, { "epoch": 3.33, "learning_rate": 1.556693504240528e-05, "loss": 0.3837, "step": 10820 }, { "epoch": 3.33, "learning_rate": 1.5566107171000437e-05, "loss": 0.4927, "step": 10821 }, { "epoch": 3.33, "learning_rate": 1.5565279244319018e-05, "loss": 0.3851, "step": 10822 }, { "epoch": 3.33, "learning_rate": 1.5564451262369247e-05, "loss": 0.3971, "step": 10823 }, { "epoch": 3.33, "learning_rate": 1.5563623225159346e-05, "loss": 0.4185, "step": 10824 }, { "epoch": 3.33, "learning_rate": 1.556279513269754e-05, "loss": 0.5655, "step": 10825 }, { "epoch": 3.33, "learning_rate": 1.556196698499205e-05, "loss": 0.391, "step": 10826 }, { "epoch": 3.33, "learning_rate": 1.5561138782051103e-05, "loss": 0.5303, "step": 10827 }, { "epoch": 3.33, "learning_rate": 1.5560310523882924e-05, "loss": 0.2756, "step": 10828 }, { "epoch": 3.33, "learning_rate": 1.5559482210495734e-05, "loss": 0.454, "step": 10829 }, { "epoch": 3.33, "learning_rate": 1.5558653841897765e-05, "loss": 0.3493, "step": 10830 }, { "epoch": 3.33, "learning_rate": 1.555782541809724e-05, "loss": 0.421, "step": 10831 }, { "epoch": 3.33, "learning_rate": 1.5556996939102385e-05, "loss": 0.435, "step": 10832 }, { "epoch": 3.33, "learning_rate": 1.5556168404921428e-05, "loss": 0.3035, "step": 10833 }, { "epoch": 3.33, "learning_rate": 1.55553398155626e-05, "loss": 0.5513, "step": 10834 }, { "epoch": 3.33, "learning_rate": 1.5554511171034124e-05, "loss": 0.4243, "step": 10835 }, { "epoch": 3.33, "learning_rate": 1.5553682471344237e-05, "loss": 0.5864, "step": 10836 }, { "epoch": 3.33, "learning_rate": 1.5552853716501165e-05, "loss": 0.5816, "step": 10837 }, { "epoch": 3.33, "learning_rate": 1.5552024906513134e-05, "loss": 0.4163, "step": 10838 }, { "epoch": 3.33, "learning_rate": 1.555119604138838e-05, "loss": 0.6319, "step": 10839 }, { "epoch": 3.34, "learning_rate": 1.5550367121135136e-05, "loss": 0.533, "step": 10840 }, { "epoch": 3.34, "learning_rate": 1.554953814576163e-05, "loss": 0.3604, "step": 10841 }, { "epoch": 3.34, "learning_rate": 1.5548709115276095e-05, "loss": 0.4281, "step": 10842 }, { "epoch": 3.34, "learning_rate": 1.5547880029686763e-05, "loss": 0.5927, "step": 10843 }, { "epoch": 3.34, "learning_rate": 1.554705088900187e-05, "loss": 0.6722, "step": 10844 }, { "epoch": 3.34, "learning_rate": 1.554622169322965e-05, "loss": 0.7082, "step": 10845 }, { "epoch": 3.34, "learning_rate": 1.554539244237834e-05, "loss": 0.537, "step": 10846 }, { "epoch": 3.34, "learning_rate": 1.5544563136456167e-05, "loss": 0.6814, "step": 10847 }, { "epoch": 3.34, "learning_rate": 1.5543733775471377e-05, "loss": 0.557, "step": 10848 }, { "epoch": 3.34, "learning_rate": 1.5542904359432198e-05, "loss": 0.4151, "step": 10849 }, { "epoch": 3.34, "learning_rate": 1.554207488834687e-05, "loss": 0.7387, "step": 10850 }, { "epoch": 3.34, "learning_rate": 1.5541245362223634e-05, "loss": 0.6933, "step": 10851 }, { "epoch": 3.34, "learning_rate": 1.5540415781070723e-05, "loss": 0.6256, "step": 10852 }, { "epoch": 3.34, "learning_rate": 1.5539586144896374e-05, "loss": 0.4501, "step": 10853 }, { "epoch": 3.34, "learning_rate": 1.553875645370883e-05, "loss": 0.5006, "step": 10854 }, { "epoch": 3.34, "learning_rate": 1.5537926707516335e-05, "loss": 0.3784, "step": 10855 }, { "epoch": 3.34, "learning_rate": 1.553709690632712e-05, "loss": 0.6853, "step": 10856 }, { "epoch": 3.34, "learning_rate": 1.5536267050149428e-05, "loss": 0.7387, "step": 10857 }, { "epoch": 3.34, "learning_rate": 1.5535437138991505e-05, "loss": 0.4536, "step": 10858 }, { "epoch": 3.34, "learning_rate": 1.5534607172861587e-05, "loss": 0.5104, "step": 10859 }, { "epoch": 3.34, "learning_rate": 1.5533777151767922e-05, "loss": 0.5183, "step": 10860 }, { "epoch": 3.34, "learning_rate": 1.553294707571875e-05, "loss": 0.5083, "step": 10861 }, { "epoch": 3.34, "learning_rate": 1.5532116944722308e-05, "loss": 0.6006, "step": 10862 }, { "epoch": 3.34, "learning_rate": 1.5531286758786852e-05, "loss": 0.4828, "step": 10863 }, { "epoch": 3.34, "learning_rate": 1.553045651792062e-05, "loss": 0.5027, "step": 10864 }, { "epoch": 3.34, "learning_rate": 1.5529626222131855e-05, "loss": 0.4787, "step": 10865 }, { "epoch": 3.34, "learning_rate": 1.5528795871428806e-05, "loss": 0.2856, "step": 10866 }, { "epoch": 3.34, "learning_rate": 1.552796546581972e-05, "loss": 0.5428, "step": 10867 }, { "epoch": 3.34, "learning_rate": 1.5527135005312843e-05, "loss": 0.6103, "step": 10868 }, { "epoch": 3.34, "learning_rate": 1.5526304489916418e-05, "loss": 0.5396, "step": 10869 }, { "epoch": 3.34, "learning_rate": 1.55254739196387e-05, "loss": 0.5999, "step": 10870 }, { "epoch": 3.34, "learning_rate": 1.5524643294487933e-05, "loss": 0.4706, "step": 10871 }, { "epoch": 3.35, "learning_rate": 1.5523812614472365e-05, "loss": 0.4174, "step": 10872 }, { "epoch": 3.35, "learning_rate": 1.5522981879600248e-05, "loss": 0.4652, "step": 10873 }, { "epoch": 3.35, "learning_rate": 1.552215108987983e-05, "loss": 0.5461, "step": 10874 }, { "epoch": 3.35, "learning_rate": 1.5521320245319364e-05, "loss": 0.3178, "step": 10875 }, { "epoch": 3.35, "learning_rate": 1.5520489345927095e-05, "loss": 0.5574, "step": 10876 }, { "epoch": 3.35, "learning_rate": 1.5519658391711283e-05, "loss": 0.7356, "step": 10877 }, { "epoch": 3.35, "learning_rate": 1.5518827382680173e-05, "loss": 0.4845, "step": 10878 }, { "epoch": 3.35, "learning_rate": 1.5517996318842024e-05, "loss": 0.5889, "step": 10879 }, { "epoch": 3.35, "learning_rate": 1.5517165200205086e-05, "loss": 0.6024, "step": 10880 }, { "epoch": 3.35, "learning_rate": 1.551633402677761e-05, "loss": 0.4313, "step": 10881 }, { "epoch": 3.35, "learning_rate": 1.5515502798567856e-05, "loss": 0.4268, "step": 10882 }, { "epoch": 3.35, "learning_rate": 1.5514671515584076e-05, "loss": 0.3255, "step": 10883 }, { "epoch": 3.35, "learning_rate": 1.5513840177834523e-05, "loss": 0.5385, "step": 10884 }, { "epoch": 3.35, "learning_rate": 1.5513008785327455e-05, "loss": 0.6591, "step": 10885 }, { "epoch": 3.35, "learning_rate": 1.551217733807113e-05, "loss": 0.5398, "step": 10886 }, { "epoch": 3.35, "learning_rate": 1.5511345836073806e-05, "loss": 0.6503, "step": 10887 }, { "epoch": 3.35, "learning_rate": 1.5510514279343736e-05, "loss": 0.6521, "step": 10888 }, { "epoch": 3.35, "learning_rate": 1.5509682667889178e-05, "loss": 0.5382, "step": 10889 }, { "epoch": 3.35, "learning_rate": 1.55088510017184e-05, "loss": 0.6705, "step": 10890 }, { "epoch": 3.35, "learning_rate": 1.550801928083965e-05, "loss": 0.5178, "step": 10891 }, { "epoch": 3.35, "learning_rate": 1.550718750526119e-05, "loss": 0.3415, "step": 10892 }, { "epoch": 3.35, "learning_rate": 1.5506355674991287e-05, "loss": 0.5347, "step": 10893 }, { "epoch": 3.35, "learning_rate": 1.5505523790038196e-05, "loss": 0.4849, "step": 10894 }, { "epoch": 3.35, "learning_rate": 1.550469185041018e-05, "loss": 0.3656, "step": 10895 }, { "epoch": 3.35, "learning_rate": 1.55038598561155e-05, "loss": 0.4292, "step": 10896 }, { "epoch": 3.35, "learning_rate": 1.550302780716242e-05, "loss": 0.5216, "step": 10897 }, { "epoch": 3.35, "learning_rate": 1.5502195703559198e-05, "loss": 0.5764, "step": 10898 }, { "epoch": 3.35, "learning_rate": 1.550136354531411e-05, "loss": 0.6903, "step": 10899 }, { "epoch": 3.35, "learning_rate": 1.5500531332435404e-05, "loss": 0.4775, "step": 10900 }, { "epoch": 3.35, "learning_rate": 1.5499699064931354e-05, "loss": 0.6358, "step": 10901 }, { "epoch": 3.35, "learning_rate": 1.5498866742810224e-05, "loss": 0.3525, "step": 10902 }, { "epoch": 3.35, "learning_rate": 1.5498034366080284e-05, "loss": 0.7384, "step": 10903 }, { "epoch": 3.35, "learning_rate": 1.549720193474979e-05, "loss": 0.4622, "step": 10904 }, { "epoch": 3.36, "learning_rate": 1.5496369448827016e-05, "loss": 0.2665, "step": 10905 }, { "epoch": 3.36, "learning_rate": 1.549553690832023e-05, "loss": 0.4027, "step": 10906 }, { "epoch": 3.36, "learning_rate": 1.5494704313237693e-05, "loss": 0.3824, "step": 10907 }, { "epoch": 3.36, "learning_rate": 1.549387166358768e-05, "loss": 0.3574, "step": 10908 }, { "epoch": 3.36, "learning_rate": 1.549303895937846e-05, "loss": 0.5725, "step": 10909 }, { "epoch": 3.36, "learning_rate": 1.54922062006183e-05, "loss": 0.5746, "step": 10910 }, { "epoch": 3.36, "learning_rate": 1.549137338731547e-05, "loss": 0.3349, "step": 10911 }, { "epoch": 3.36, "learning_rate": 1.5490540519478242e-05, "loss": 0.6074, "step": 10912 }, { "epoch": 3.36, "learning_rate": 1.5489707597114884e-05, "loss": 0.4305, "step": 10913 }, { "epoch": 3.36, "learning_rate": 1.5488874620233674e-05, "loss": 0.4826, "step": 10914 }, { "epoch": 3.36, "learning_rate": 1.548804158884288e-05, "loss": 0.305, "step": 10915 }, { "epoch": 3.36, "learning_rate": 1.548720850295077e-05, "loss": 0.5057, "step": 10916 }, { "epoch": 3.36, "learning_rate": 1.5486375362565626e-05, "loss": 0.5258, "step": 10917 }, { "epoch": 3.36, "learning_rate": 1.548554216769572e-05, "loss": 0.4434, "step": 10918 }, { "epoch": 3.36, "learning_rate": 1.548470891834932e-05, "loss": 0.4488, "step": 10919 }, { "epoch": 3.36, "learning_rate": 1.548387561453471e-05, "loss": 0.4936, "step": 10920 }, { "epoch": 3.36, "learning_rate": 1.5483042256260157e-05, "loss": 0.6781, "step": 10921 }, { "epoch": 3.36, "learning_rate": 1.5482208843533946e-05, "loss": 0.545, "step": 10922 }, { "epoch": 3.36, "learning_rate": 1.5481375376364344e-05, "loss": 0.4326, "step": 10923 }, { "epoch": 3.36, "learning_rate": 1.5480541854759637e-05, "loss": 0.3537, "step": 10924 }, { "epoch": 3.36, "learning_rate": 1.5479708278728093e-05, "loss": 0.4139, "step": 10925 }, { "epoch": 3.36, "learning_rate": 1.5478874648278002e-05, "loss": 0.3052, "step": 10926 }, { "epoch": 3.36, "learning_rate": 1.547804096341763e-05, "loss": 0.4209, "step": 10927 }, { "epoch": 3.36, "learning_rate": 1.5477207224155266e-05, "loss": 0.6521, "step": 10928 }, { "epoch": 3.36, "learning_rate": 1.5476373430499185e-05, "loss": 0.4146, "step": 10929 }, { "epoch": 3.36, "learning_rate": 1.547553958245767e-05, "loss": 0.722, "step": 10930 }, { "epoch": 3.36, "learning_rate": 1.5474705680039002e-05, "loss": 0.6498, "step": 10931 }, { "epoch": 3.36, "learning_rate": 1.5473871723251457e-05, "loss": 0.5171, "step": 10932 }, { "epoch": 3.36, "learning_rate": 1.5473037712103323e-05, "loss": 0.248, "step": 10933 }, { "epoch": 3.36, "learning_rate": 1.547220364660288e-05, "loss": 0.4933, "step": 10934 }, { "epoch": 3.36, "learning_rate": 1.547136952675841e-05, "loss": 0.3826, "step": 10935 }, { "epoch": 3.36, "learning_rate": 1.5470535352578202e-05, "loss": 0.6288, "step": 10936 }, { "epoch": 3.37, "learning_rate": 1.5469701124070535e-05, "loss": 0.5752, "step": 10937 }, { "epoch": 3.37, "learning_rate": 1.546886684124369e-05, "loss": 0.5398, "step": 10938 }, { "epoch": 3.37, "learning_rate": 1.5468032504105962e-05, "loss": 0.4808, "step": 10939 }, { "epoch": 3.37, "learning_rate": 1.5467198112665632e-05, "loss": 0.5343, "step": 10940 }, { "epoch": 3.37, "learning_rate": 1.5466363666930985e-05, "loss": 0.4943, "step": 10941 }, { "epoch": 3.37, "learning_rate": 1.5465529166910308e-05, "loss": 0.6111, "step": 10942 }, { "epoch": 3.37, "learning_rate": 1.5464694612611888e-05, "loss": 0.5739, "step": 10943 }, { "epoch": 3.37, "learning_rate": 1.5463860004044017e-05, "loss": 0.4493, "step": 10944 }, { "epoch": 3.37, "learning_rate": 1.5463025341214977e-05, "loss": 0.4722, "step": 10945 }, { "epoch": 3.37, "learning_rate": 1.5462190624133064e-05, "loss": 0.453, "step": 10946 }, { "epoch": 3.37, "learning_rate": 1.5461355852806562e-05, "loss": 0.4927, "step": 10947 }, { "epoch": 3.37, "learning_rate": 1.5460521027243764e-05, "loss": 0.4965, "step": 10948 }, { "epoch": 3.37, "learning_rate": 1.5459686147452958e-05, "loss": 0.2839, "step": 10949 }, { "epoch": 3.37, "learning_rate": 1.5458851213442438e-05, "loss": 0.4951, "step": 10950 }, { "epoch": 3.37, "learning_rate": 1.5458016225220494e-05, "loss": 0.4397, "step": 10951 }, { "epoch": 3.37, "learning_rate": 1.5457181182795422e-05, "loss": 0.421, "step": 10952 }, { "epoch": 3.37, "learning_rate": 1.5456346086175508e-05, "loss": 0.4996, "step": 10953 }, { "epoch": 3.37, "learning_rate": 1.5455510935369046e-05, "loss": 0.4591, "step": 10954 }, { "epoch": 3.37, "learning_rate": 1.5454675730384338e-05, "loss": 0.7781, "step": 10955 }, { "epoch": 3.37, "learning_rate": 1.545384047122967e-05, "loss": 0.4877, "step": 10956 }, { "epoch": 3.37, "learning_rate": 1.5453005157913338e-05, "loss": 0.6404, "step": 10957 }, { "epoch": 3.37, "learning_rate": 1.5452169790443644e-05, "loss": 0.4387, "step": 10958 }, { "epoch": 3.37, "learning_rate": 1.5451334368828878e-05, "loss": 0.6099, "step": 10959 }, { "epoch": 3.37, "learning_rate": 1.5450498893077335e-05, "loss": 0.488, "step": 10960 }, { "epoch": 3.37, "learning_rate": 1.5449663363197317e-05, "loss": 0.5783, "step": 10961 }, { "epoch": 3.37, "learning_rate": 1.5448827779197115e-05, "loss": 0.6836, "step": 10962 }, { "epoch": 3.37, "learning_rate": 1.5447992141085033e-05, "loss": 0.5109, "step": 10963 }, { "epoch": 3.37, "learning_rate": 1.5447156448869372e-05, "loss": 0.4932, "step": 10964 }, { "epoch": 3.37, "learning_rate": 1.5446320702558423e-05, "loss": 0.3728, "step": 10965 }, { "epoch": 3.37, "learning_rate": 1.5445484902160494e-05, "loss": 0.5602, "step": 10966 }, { "epoch": 3.37, "learning_rate": 1.5444649047683878e-05, "loss": 0.342, "step": 10967 }, { "epoch": 3.37, "learning_rate": 1.544381313913688e-05, "loss": 0.4072, "step": 10968 }, { "epoch": 3.37, "learning_rate": 1.5442977176527797e-05, "loss": 0.5484, "step": 10969 }, { "epoch": 3.38, "learning_rate": 1.5442141159864937e-05, "loss": 0.7208, "step": 10970 }, { "epoch": 3.38, "learning_rate": 1.5441305089156602e-05, "loss": 0.5493, "step": 10971 }, { "epoch": 3.38, "learning_rate": 1.544046896441109e-05, "loss": 0.5485, "step": 10972 }, { "epoch": 3.38, "learning_rate": 1.5439632785636707e-05, "loss": 0.5261, "step": 10973 }, { "epoch": 3.38, "learning_rate": 1.543879655284176e-05, "loss": 0.4264, "step": 10974 }, { "epoch": 3.38, "learning_rate": 1.5437960266034547e-05, "loss": 0.3525, "step": 10975 }, { "epoch": 3.38, "learning_rate": 1.543712392522338e-05, "loss": 0.6006, "step": 10976 }, { "epoch": 3.38, "learning_rate": 1.543628753041656e-05, "loss": 0.5066, "step": 10977 }, { "epoch": 3.38, "learning_rate": 1.5435451081622393e-05, "loss": 0.5032, "step": 10978 }, { "epoch": 3.38, "learning_rate": 1.543461457884919e-05, "loss": 0.4725, "step": 10979 }, { "epoch": 3.38, "learning_rate": 1.5433778022105254e-05, "loss": 0.5384, "step": 10980 }, { "epoch": 3.38, "learning_rate": 1.5432941411398893e-05, "loss": 0.4105, "step": 10981 }, { "epoch": 3.38, "learning_rate": 1.5432104746738417e-05, "loss": 0.6037, "step": 10982 }, { "epoch": 3.38, "learning_rate": 1.5431268028132134e-05, "loss": 0.5366, "step": 10983 }, { "epoch": 3.38, "learning_rate": 1.543043125558836e-05, "loss": 0.46, "step": 10984 }, { "epoch": 3.38, "learning_rate": 1.542959442911539e-05, "loss": 0.453, "step": 10985 }, { "epoch": 3.38, "learning_rate": 1.5428757548721546e-05, "loss": 0.815, "step": 10986 }, { "epoch": 3.38, "learning_rate": 1.542792061441514e-05, "loss": 0.8057, "step": 10987 }, { "epoch": 3.38, "learning_rate": 1.542708362620448e-05, "loss": 0.2683, "step": 10988 }, { "epoch": 3.38, "learning_rate": 1.542624658409787e-05, "loss": 0.6056, "step": 10989 }, { "epoch": 3.38, "learning_rate": 1.5425409488103638e-05, "loss": 0.434, "step": 10990 }, { "epoch": 3.38, "learning_rate": 1.542457233823009e-05, "loss": 0.4962, "step": 10991 }, { "epoch": 3.38, "learning_rate": 1.5423735134485537e-05, "loss": 0.4303, "step": 10992 }, { "epoch": 3.38, "learning_rate": 1.5422897876878295e-05, "loss": 0.6126, "step": 10993 }, { "epoch": 3.38, "learning_rate": 1.5422060565416678e-05, "loss": 0.5274, "step": 10994 }, { "epoch": 3.38, "learning_rate": 1.5421223200109003e-05, "loss": 0.4652, "step": 10995 }, { "epoch": 3.38, "learning_rate": 1.5420385780963586e-05, "loss": 0.4247, "step": 10996 }, { "epoch": 3.38, "learning_rate": 1.5419548307988744e-05, "loss": 0.5634, "step": 10997 }, { "epoch": 3.38, "learning_rate": 1.541871078119279e-05, "loss": 0.571, "step": 10998 }, { "epoch": 3.38, "learning_rate": 1.5417873200584046e-05, "loss": 0.8925, "step": 10999 }, { "epoch": 3.38, "learning_rate": 1.5417035566170822e-05, "loss": 0.8073, "step": 11000 }, { "epoch": 3.38, "learning_rate": 1.5416197877961452e-05, "loss": 0.6664, "step": 11001 }, { "epoch": 3.39, "learning_rate": 1.541536013596424e-05, "loss": 0.561, "step": 11002 }, { "epoch": 3.39, "learning_rate": 1.5414522340187513e-05, "loss": 0.5618, "step": 11003 }, { "epoch": 3.39, "learning_rate": 1.5413684490639583e-05, "loss": 0.6238, "step": 11004 }, { "epoch": 3.39, "learning_rate": 1.541284658732878e-05, "loss": 0.4294, "step": 11005 }, { "epoch": 3.39, "learning_rate": 1.5412008630263426e-05, "loss": 0.5339, "step": 11006 }, { "epoch": 3.39, "learning_rate": 1.5411170619451837e-05, "loss": 0.6027, "step": 11007 }, { "epoch": 3.39, "learning_rate": 1.5410332554902334e-05, "loss": 0.4377, "step": 11008 }, { "epoch": 3.39, "learning_rate": 1.5409494436623244e-05, "loss": 0.4991, "step": 11009 }, { "epoch": 3.39, "learning_rate": 1.5408656264622884e-05, "loss": 0.6249, "step": 11010 }, { "epoch": 3.39, "learning_rate": 1.5407818038909588e-05, "loss": 0.4076, "step": 11011 }, { "epoch": 3.39, "learning_rate": 1.5406979759491675e-05, "loss": 0.3381, "step": 11012 }, { "epoch": 3.39, "learning_rate": 1.5406141426377467e-05, "loss": 0.4507, "step": 11013 }, { "epoch": 3.39, "learning_rate": 1.5405303039575295e-05, "loss": 0.4758, "step": 11014 }, { "epoch": 3.39, "learning_rate": 1.540446459909348e-05, "loss": 0.6095, "step": 11015 }, { "epoch": 3.39, "learning_rate": 1.540362610494035e-05, "loss": 0.364, "step": 11016 }, { "epoch": 3.39, "learning_rate": 1.5402787557124238e-05, "loss": 0.7331, "step": 11017 }, { "epoch": 3.39, "learning_rate": 1.540194895565346e-05, "loss": 0.5201, "step": 11018 }, { "epoch": 3.39, "learning_rate": 1.5401110300536354e-05, "loss": 0.6542, "step": 11019 }, { "epoch": 3.39, "learning_rate": 1.540027159178124e-05, "loss": 0.4776, "step": 11020 }, { "epoch": 3.39, "learning_rate": 1.5399432829396455e-05, "loss": 0.2968, "step": 11021 }, { "epoch": 3.39, "learning_rate": 1.5398594013390326e-05, "loss": 0.4816, "step": 11022 }, { "epoch": 3.39, "learning_rate": 1.5397755143771185e-05, "loss": 0.6395, "step": 11023 }, { "epoch": 3.39, "learning_rate": 1.539691622054736e-05, "loss": 0.5638, "step": 11024 }, { "epoch": 3.39, "learning_rate": 1.539607724372718e-05, "loss": 0.6462, "step": 11025 }, { "epoch": 3.39, "learning_rate": 1.539523821331898e-05, "loss": 0.4581, "step": 11026 }, { "epoch": 3.39, "learning_rate": 1.5394399129331096e-05, "loss": 0.465, "step": 11027 }, { "epoch": 3.39, "learning_rate": 1.5393559991771856e-05, "loss": 0.4093, "step": 11028 }, { "epoch": 3.39, "learning_rate": 1.5392720800649593e-05, "loss": 0.417, "step": 11029 }, { "epoch": 3.39, "learning_rate": 1.5391881555972642e-05, "loss": 0.4894, "step": 11030 }, { "epoch": 3.39, "learning_rate": 1.5391042257749338e-05, "loss": 0.4563, "step": 11031 }, { "epoch": 3.39, "learning_rate": 1.5390202905988015e-05, "loss": 0.4304, "step": 11032 }, { "epoch": 3.39, "learning_rate": 1.5389363500697012e-05, "loss": 0.528, "step": 11033 }, { "epoch": 3.39, "learning_rate": 1.538852404188466e-05, "loss": 0.5595, "step": 11034 }, { "epoch": 3.4, "learning_rate": 1.5387684529559298e-05, "loss": 0.3853, "step": 11035 }, { "epoch": 3.4, "learning_rate": 1.5386844963729266e-05, "loss": 0.5145, "step": 11036 }, { "epoch": 3.4, "learning_rate": 1.5386005344402897e-05, "loss": 0.661, "step": 11037 }, { "epoch": 3.4, "learning_rate": 1.5385165671588535e-05, "loss": 0.7036, "step": 11038 }, { "epoch": 3.4, "learning_rate": 1.5384325945294508e-05, "loss": 0.6674, "step": 11039 }, { "epoch": 3.4, "learning_rate": 1.5383486165529165e-05, "loss": 0.4459, "step": 11040 }, { "epoch": 3.4, "learning_rate": 1.5382646332300844e-05, "loss": 0.7578, "step": 11041 }, { "epoch": 3.4, "learning_rate": 1.5381806445617883e-05, "loss": 0.4298, "step": 11042 }, { "epoch": 3.4, "learning_rate": 1.5380966505488623e-05, "loss": 0.7516, "step": 11043 }, { "epoch": 3.4, "learning_rate": 1.5380126511921404e-05, "loss": 0.4483, "step": 11044 }, { "epoch": 3.4, "learning_rate": 1.5379286464924576e-05, "loss": 0.583, "step": 11045 }, { "epoch": 3.4, "learning_rate": 1.537844636450647e-05, "loss": 0.5636, "step": 11046 }, { "epoch": 3.4, "learning_rate": 1.537760621067544e-05, "loss": 0.5208, "step": 11047 }, { "epoch": 3.4, "learning_rate": 1.5376766003439823e-05, "loss": 0.4863, "step": 11048 }, { "epoch": 3.4, "learning_rate": 1.5375925742807964e-05, "loss": 0.4108, "step": 11049 }, { "epoch": 3.4, "learning_rate": 1.5375085428788207e-05, "loss": 0.5808, "step": 11050 }, { "epoch": 3.4, "learning_rate": 1.5374245061388898e-05, "loss": 0.6352, "step": 11051 }, { "epoch": 3.4, "learning_rate": 1.5373404640618382e-05, "loss": 0.5949, "step": 11052 }, { "epoch": 3.4, "learning_rate": 1.5372564166485004e-05, "loss": 0.3181, "step": 11053 }, { "epoch": 3.4, "learning_rate": 1.5371723638997115e-05, "loss": 0.4899, "step": 11054 }, { "epoch": 3.4, "learning_rate": 1.537088305816306e-05, "loss": 0.4412, "step": 11055 }, { "epoch": 3.4, "learning_rate": 1.5370042423991183e-05, "loss": 0.4394, "step": 11056 }, { "epoch": 3.4, "learning_rate": 1.536920173648984e-05, "loss": 0.5546, "step": 11057 }, { "epoch": 3.4, "learning_rate": 1.5368360995667372e-05, "loss": 0.5134, "step": 11058 }, { "epoch": 3.4, "learning_rate": 1.5367520201532133e-05, "loss": 0.5903, "step": 11059 }, { "epoch": 3.4, "learning_rate": 1.536667935409247e-05, "loss": 0.475, "step": 11060 }, { "epoch": 3.4, "learning_rate": 1.5365838453356738e-05, "loss": 0.3115, "step": 11061 }, { "epoch": 3.4, "learning_rate": 1.5364997499333286e-05, "loss": 0.2665, "step": 11062 }, { "epoch": 3.4, "learning_rate": 1.536415649203046e-05, "loss": 0.6159, "step": 11063 }, { "epoch": 3.4, "learning_rate": 1.5363315431456615e-05, "loss": 0.3874, "step": 11064 }, { "epoch": 3.4, "learning_rate": 1.5362474317620106e-05, "loss": 0.5572, "step": 11065 }, { "epoch": 3.4, "learning_rate": 1.536163315052929e-05, "loss": 0.451, "step": 11066 }, { "epoch": 3.41, "learning_rate": 1.536079193019251e-05, "loss": 0.5241, "step": 11067 }, { "epoch": 3.41, "learning_rate": 1.5359950656618127e-05, "loss": 0.4708, "step": 11068 }, { "epoch": 3.41, "learning_rate": 1.5359109329814495e-05, "loss": 0.3134, "step": 11069 }, { "epoch": 3.41, "learning_rate": 1.5358267949789968e-05, "loss": 0.657, "step": 11070 }, { "epoch": 3.41, "learning_rate": 1.5357426516552903e-05, "loss": 0.3828, "step": 11071 }, { "epoch": 3.41, "learning_rate": 1.535658503011165e-05, "loss": 0.292, "step": 11072 }, { "epoch": 3.41, "learning_rate": 1.5355743490474578e-05, "loss": 0.3505, "step": 11073 }, { "epoch": 3.41, "learning_rate": 1.535490189765003e-05, "loss": 0.4813, "step": 11074 }, { "epoch": 3.41, "learning_rate": 1.5354060251646377e-05, "loss": 0.6072, "step": 11075 }, { "epoch": 3.41, "learning_rate": 1.5353218552471968e-05, "loss": 0.5716, "step": 11076 }, { "epoch": 3.41, "learning_rate": 1.5352376800135166e-05, "loss": 0.4844, "step": 11077 }, { "epoch": 3.41, "learning_rate": 1.5351534994644332e-05, "loss": 0.4915, "step": 11078 }, { "epoch": 3.41, "learning_rate": 1.535069313600782e-05, "loss": 0.4447, "step": 11079 }, { "epoch": 3.41, "learning_rate": 1.5349851224233995e-05, "loss": 0.4754, "step": 11080 }, { "epoch": 3.41, "learning_rate": 1.5349009259331215e-05, "loss": 0.6252, "step": 11081 }, { "epoch": 3.41, "learning_rate": 1.5348167241307845e-05, "loss": 0.5428, "step": 11082 }, { "epoch": 3.41, "learning_rate": 1.5347325170172246e-05, "loss": 0.4741, "step": 11083 }, { "epoch": 3.41, "learning_rate": 1.5346483045932777e-05, "loss": 0.409, "step": 11084 }, { "epoch": 3.41, "learning_rate": 1.5345640868597803e-05, "loss": 0.5039, "step": 11085 }, { "epoch": 3.41, "learning_rate": 1.5344798638175694e-05, "loss": 0.6055, "step": 11086 }, { "epoch": 3.41, "learning_rate": 1.5343956354674806e-05, "loss": 0.4196, "step": 11087 }, { "epoch": 3.41, "learning_rate": 1.5343114018103507e-05, "loss": 0.5554, "step": 11088 }, { "epoch": 3.41, "learning_rate": 1.534227162847016e-05, "loss": 0.7116, "step": 11089 }, { "epoch": 3.41, "learning_rate": 1.5341429185783132e-05, "loss": 0.3925, "step": 11090 }, { "epoch": 3.41, "learning_rate": 1.5340586690050788e-05, "loss": 0.4253, "step": 11091 }, { "epoch": 3.41, "learning_rate": 1.53397441412815e-05, "loss": 0.5058, "step": 11092 }, { "epoch": 3.41, "learning_rate": 1.533890153948363e-05, "loss": 0.3945, "step": 11093 }, { "epoch": 3.41, "learning_rate": 1.5338058884665547e-05, "loss": 0.543, "step": 11094 }, { "epoch": 3.41, "learning_rate": 1.5337216176835617e-05, "loss": 0.5473, "step": 11095 }, { "epoch": 3.41, "learning_rate": 1.533637341600221e-05, "loss": 0.4153, "step": 11096 }, { "epoch": 3.41, "learning_rate": 1.5335530602173704e-05, "loss": 0.5198, "step": 11097 }, { "epoch": 3.41, "learning_rate": 1.533468773535846e-05, "loss": 0.4435, "step": 11098 }, { "epoch": 3.41, "learning_rate": 1.533384481556485e-05, "loss": 0.6018, "step": 11099 }, { "epoch": 3.42, "learning_rate": 1.533300184280124e-05, "loss": 0.7996, "step": 11100 }, { "epoch": 3.42, "learning_rate": 1.533215881707601e-05, "loss": 0.569, "step": 11101 }, { "epoch": 3.42, "learning_rate": 1.533131573839753e-05, "loss": 0.539, "step": 11102 }, { "epoch": 3.42, "learning_rate": 1.533047260677417e-05, "loss": 0.5518, "step": 11103 }, { "epoch": 3.42, "learning_rate": 1.532962942221431e-05, "loss": 0.5912, "step": 11104 }, { "epoch": 3.42, "learning_rate": 1.5328786184726308e-05, "loss": 0.465, "step": 11105 }, { "epoch": 3.42, "learning_rate": 1.5327942894318552e-05, "loss": 0.5838, "step": 11106 }, { "epoch": 3.42, "learning_rate": 1.5327099550999416e-05, "loss": 0.3823, "step": 11107 }, { "epoch": 3.42, "learning_rate": 1.532625615477727e-05, "loss": 0.429, "step": 11108 }, { "epoch": 3.42, "learning_rate": 1.532541270566049e-05, "loss": 0.4716, "step": 11109 }, { "epoch": 3.42, "learning_rate": 1.5324569203657452e-05, "loss": 0.5068, "step": 11110 }, { "epoch": 3.42, "learning_rate": 1.5323725648776536e-05, "loss": 0.3423, "step": 11111 }, { "epoch": 3.42, "learning_rate": 1.532288204102612e-05, "loss": 0.5941, "step": 11112 }, { "epoch": 3.42, "learning_rate": 1.532203838041458e-05, "loss": 0.4891, "step": 11113 }, { "epoch": 3.42, "learning_rate": 1.5321194666950292e-05, "loss": 0.4651, "step": 11114 }, { "epoch": 3.42, "learning_rate": 1.5320350900641635e-05, "loss": 0.509, "step": 11115 }, { "epoch": 3.42, "learning_rate": 1.5319507081496994e-05, "loss": 0.5243, "step": 11116 }, { "epoch": 3.42, "learning_rate": 1.5318663209524745e-05, "loss": 0.4265, "step": 11117 }, { "epoch": 3.42, "learning_rate": 1.5317819284733266e-05, "loss": 0.5973, "step": 11118 }, { "epoch": 3.42, "learning_rate": 1.5316975307130942e-05, "loss": 0.4233, "step": 11119 }, { "epoch": 3.42, "learning_rate": 1.531613127672615e-05, "loss": 0.5057, "step": 11120 }, { "epoch": 3.42, "learning_rate": 1.531528719352728e-05, "loss": 0.3948, "step": 11121 }, { "epoch": 3.42, "learning_rate": 1.5314443057542703e-05, "loss": 0.6926, "step": 11122 }, { "epoch": 3.42, "learning_rate": 1.5313598868780816e-05, "loss": 0.7226, "step": 11123 }, { "epoch": 3.42, "learning_rate": 1.531275462724999e-05, "loss": 0.4102, "step": 11124 }, { "epoch": 3.42, "learning_rate": 1.5311910332958615e-05, "loss": 0.611, "step": 11125 }, { "epoch": 3.42, "learning_rate": 1.5311065985915076e-05, "loss": 0.3645, "step": 11126 }, { "epoch": 3.42, "learning_rate": 1.5310221586127753e-05, "loss": 0.5084, "step": 11127 }, { "epoch": 3.42, "learning_rate": 1.5309377133605044e-05, "loss": 0.5052, "step": 11128 }, { "epoch": 3.42, "learning_rate": 1.5308532628355317e-05, "loss": 0.3997, "step": 11129 }, { "epoch": 3.42, "learning_rate": 1.5307688070386974e-05, "loss": 0.5311, "step": 11130 }, { "epoch": 3.42, "learning_rate": 1.5306843459708397e-05, "loss": 0.4345, "step": 11131 }, { "epoch": 3.43, "learning_rate": 1.530599879632797e-05, "loss": 0.4376, "step": 11132 }, { "epoch": 3.43, "learning_rate": 1.530515408025409e-05, "loss": 0.4096, "step": 11133 }, { "epoch": 3.43, "learning_rate": 1.5304309311495136e-05, "loss": 0.434, "step": 11134 }, { "epoch": 3.43, "learning_rate": 1.5303464490059506e-05, "loss": 0.6454, "step": 11135 }, { "epoch": 3.43, "learning_rate": 1.5302619615955583e-05, "loss": 0.4454, "step": 11136 }, { "epoch": 3.43, "learning_rate": 1.5301774689191758e-05, "loss": 0.4045, "step": 11137 }, { "epoch": 3.43, "learning_rate": 1.5300929709776432e-05, "loss": 0.4382, "step": 11138 }, { "epoch": 3.43, "learning_rate": 1.530008467771798e-05, "loss": 0.4715, "step": 11139 }, { "epoch": 3.43, "learning_rate": 1.5299239593024806e-05, "loss": 0.5954, "step": 11140 }, { "epoch": 3.43, "learning_rate": 1.52983944557053e-05, "loss": 0.4249, "step": 11141 }, { "epoch": 3.43, "learning_rate": 1.5297549265767847e-05, "loss": 0.6011, "step": 11142 }, { "epoch": 3.43, "learning_rate": 1.529670402322086e-05, "loss": 0.5075, "step": 11143 }, { "epoch": 3.43, "learning_rate": 1.529585872807271e-05, "loss": 0.4892, "step": 11144 }, { "epoch": 3.43, "learning_rate": 1.5295013380331804e-05, "loss": 0.4114, "step": 11145 }, { "epoch": 3.43, "learning_rate": 1.5294167980006535e-05, "loss": 0.5177, "step": 11146 }, { "epoch": 3.43, "learning_rate": 1.52933225271053e-05, "loss": 0.6077, "step": 11147 }, { "epoch": 3.43, "learning_rate": 1.5292477021636498e-05, "loss": 0.6621, "step": 11148 }, { "epoch": 3.43, "learning_rate": 1.5291631463608518e-05, "loss": 0.5849, "step": 11149 }, { "epoch": 3.43, "learning_rate": 1.529078585302976e-05, "loss": 0.477, "step": 11150 }, { "epoch": 3.43, "learning_rate": 1.5289940189908622e-05, "loss": 0.5014, "step": 11151 }, { "epoch": 3.43, "learning_rate": 1.5289094474253502e-05, "loss": 0.5595, "step": 11152 }, { "epoch": 3.43, "learning_rate": 1.5288248706072802e-05, "loss": 0.4671, "step": 11153 }, { "epoch": 3.43, "learning_rate": 1.5287402885374918e-05, "loss": 0.6373, "step": 11154 }, { "epoch": 3.43, "learning_rate": 1.5286557012168248e-05, "loss": 0.3947, "step": 11155 }, { "epoch": 3.43, "learning_rate": 1.5285711086461195e-05, "loss": 0.4808, "step": 11156 }, { "epoch": 3.43, "learning_rate": 1.5284865108262162e-05, "loss": 0.5029, "step": 11157 }, { "epoch": 3.43, "learning_rate": 1.5284019077579544e-05, "loss": 0.5289, "step": 11158 }, { "epoch": 3.43, "learning_rate": 1.5283172994421752e-05, "loss": 0.5179, "step": 11159 }, { "epoch": 3.43, "learning_rate": 1.528232685879718e-05, "loss": 0.6334, "step": 11160 }, { "epoch": 3.43, "learning_rate": 1.528148067071423e-05, "loss": 0.5031, "step": 11161 }, { "epoch": 3.43, "learning_rate": 1.5280634430181315e-05, "loss": 0.7227, "step": 11162 }, { "epoch": 3.43, "learning_rate": 1.527978813720683e-05, "loss": 0.4238, "step": 11163 }, { "epoch": 3.43, "learning_rate": 1.527894179179919e-05, "loss": 0.6691, "step": 11164 }, { "epoch": 3.44, "learning_rate": 1.5278095393966782e-05, "loss": 0.6726, "step": 11165 }, { "epoch": 3.44, "learning_rate": 1.527724894371803e-05, "loss": 0.5143, "step": 11166 }, { "epoch": 3.44, "learning_rate": 1.527640244106133e-05, "loss": 0.5431, "step": 11167 }, { "epoch": 3.44, "learning_rate": 1.527555588600509e-05, "loss": 0.6738, "step": 11168 }, { "epoch": 3.44, "learning_rate": 1.527470927855772e-05, "loss": 0.5958, "step": 11169 }, { "epoch": 3.44, "learning_rate": 1.5273862618727623e-05, "loss": 0.5801, "step": 11170 }, { "epoch": 3.44, "learning_rate": 1.527301590652321e-05, "loss": 0.4949, "step": 11171 }, { "epoch": 3.44, "learning_rate": 1.527216914195289e-05, "loss": 0.5139, "step": 11172 }, { "epoch": 3.44, "learning_rate": 1.5271322325025072e-05, "loss": 0.5731, "step": 11173 }, { "epoch": 3.44, "learning_rate": 1.5270475455748165e-05, "loss": 0.4668, "step": 11174 }, { "epoch": 3.44, "learning_rate": 1.5269628534130578e-05, "loss": 0.6492, "step": 11175 }, { "epoch": 3.44, "learning_rate": 1.5268781560180727e-05, "loss": 0.3946, "step": 11176 }, { "epoch": 3.44, "learning_rate": 1.526793453390701e-05, "loss": 0.388, "step": 11177 }, { "epoch": 3.44, "learning_rate": 1.526708745531786e-05, "loss": 0.5304, "step": 11178 }, { "epoch": 3.44, "learning_rate": 1.526624032442167e-05, "loss": 0.6711, "step": 11179 }, { "epoch": 3.44, "learning_rate": 1.5265393141226863e-05, "loss": 0.4468, "step": 11180 }, { "epoch": 3.44, "learning_rate": 1.5264545905741847e-05, "loss": 0.6692, "step": 11181 }, { "epoch": 3.44, "learning_rate": 1.5263698617975037e-05, "loss": 0.4522, "step": 11182 }, { "epoch": 3.44, "learning_rate": 1.5262851277934853e-05, "loss": 0.4829, "step": 11183 }, { "epoch": 3.44, "learning_rate": 1.5262003885629703e-05, "loss": 0.5595, "step": 11184 }, { "epoch": 3.44, "learning_rate": 1.5261156441068004e-05, "loss": 0.5412, "step": 11185 }, { "epoch": 3.44, "learning_rate": 1.5260308944258176e-05, "loss": 0.4603, "step": 11186 }, { "epoch": 3.44, "learning_rate": 1.5259461395208628e-05, "loss": 0.5015, "step": 11187 }, { "epoch": 3.44, "learning_rate": 1.5258613793927783e-05, "loss": 0.4473, "step": 11188 }, { "epoch": 3.44, "learning_rate": 1.5257766140424058e-05, "loss": 0.4438, "step": 11189 }, { "epoch": 3.44, "learning_rate": 1.5256918434705869e-05, "loss": 0.4348, "step": 11190 }, { "epoch": 3.44, "learning_rate": 1.5256070676781633e-05, "loss": 0.3828, "step": 11191 }, { "epoch": 3.44, "learning_rate": 1.525522286665977e-05, "loss": 0.5932, "step": 11192 }, { "epoch": 3.44, "learning_rate": 1.5254375004348705e-05, "loss": 0.5643, "step": 11193 }, { "epoch": 3.44, "learning_rate": 1.5253527089856853e-05, "loss": 0.3558, "step": 11194 }, { "epoch": 3.44, "learning_rate": 1.5252679123192633e-05, "loss": 0.2693, "step": 11195 }, { "epoch": 3.44, "learning_rate": 1.5251831104364468e-05, "loss": 0.383, "step": 11196 }, { "epoch": 3.45, "learning_rate": 1.525098303338078e-05, "loss": 0.7715, "step": 11197 }, { "epoch": 3.45, "learning_rate": 1.5250134910249992e-05, "loss": 0.4862, "step": 11198 }, { "epoch": 3.45, "learning_rate": 1.5249286734980526e-05, "loss": 0.535, "step": 11199 }, { "epoch": 3.45, "learning_rate": 1.5248438507580806e-05, "loss": 0.6078, "step": 11200 }, { "epoch": 3.45, "learning_rate": 1.5247590228059251e-05, "loss": 0.4544, "step": 11201 }, { "epoch": 3.45, "learning_rate": 1.524674189642429e-05, "loss": 0.6851, "step": 11202 }, { "epoch": 3.45, "learning_rate": 1.5245893512684347e-05, "loss": 0.6101, "step": 11203 }, { "epoch": 3.45, "learning_rate": 1.5245045076847849e-05, "loss": 0.5033, "step": 11204 }, { "epoch": 3.45, "learning_rate": 1.5244196588923217e-05, "loss": 0.4263, "step": 11205 }, { "epoch": 3.45, "learning_rate": 1.5243348048918875e-05, "loss": 0.759, "step": 11206 }, { "epoch": 3.45, "learning_rate": 1.5242499456843264e-05, "loss": 0.4859, "step": 11207 }, { "epoch": 3.45, "learning_rate": 1.5241650812704794e-05, "loss": 0.5206, "step": 11208 }, { "epoch": 3.45, "learning_rate": 1.5240802116511906e-05, "loss": 0.57, "step": 11209 }, { "epoch": 3.45, "learning_rate": 1.523995336827302e-05, "loss": 0.563, "step": 11210 }, { "epoch": 3.45, "learning_rate": 1.5239104567996568e-05, "loss": 0.3817, "step": 11211 }, { "epoch": 3.45, "learning_rate": 1.523825571569098e-05, "loss": 0.4947, "step": 11212 }, { "epoch": 3.45, "learning_rate": 1.5237406811364682e-05, "loss": 0.5433, "step": 11213 }, { "epoch": 3.45, "learning_rate": 1.5236557855026112e-05, "loss": 0.584, "step": 11214 }, { "epoch": 3.45, "learning_rate": 1.5235708846683693e-05, "loss": 0.5569, "step": 11215 }, { "epoch": 3.45, "learning_rate": 1.5234859786345859e-05, "loss": 0.645, "step": 11216 }, { "epoch": 3.45, "learning_rate": 1.5234010674021048e-05, "loss": 0.4032, "step": 11217 }, { "epoch": 3.45, "learning_rate": 1.523316150971768e-05, "loss": 0.7179, "step": 11218 }, { "epoch": 3.45, "learning_rate": 1.5232312293444203e-05, "loss": 0.3442, "step": 11219 }, { "epoch": 3.45, "learning_rate": 1.5231463025209037e-05, "loss": 0.4952, "step": 11220 }, { "epoch": 3.45, "learning_rate": 1.5230613705020623e-05, "loss": 0.5908, "step": 11221 }, { "epoch": 3.45, "learning_rate": 1.5229764332887395e-05, "loss": 0.5062, "step": 11222 }, { "epoch": 3.45, "learning_rate": 1.5228914908817784e-05, "loss": 0.543, "step": 11223 }, { "epoch": 3.45, "learning_rate": 1.5228065432820236e-05, "loss": 0.4142, "step": 11224 }, { "epoch": 3.45, "learning_rate": 1.5227215904903175e-05, "loss": 0.4457, "step": 11225 }, { "epoch": 3.45, "learning_rate": 1.5226366325075042e-05, "loss": 0.4979, "step": 11226 }, { "epoch": 3.45, "learning_rate": 1.5225516693344276e-05, "loss": 0.5674, "step": 11227 }, { "epoch": 3.45, "learning_rate": 1.5224667009719313e-05, "loss": 0.7015, "step": 11228 }, { "epoch": 3.45, "learning_rate": 1.5223817274208595e-05, "loss": 0.5222, "step": 11229 }, { "epoch": 3.46, "learning_rate": 1.522296748682055e-05, "loss": 0.5674, "step": 11230 }, { "epoch": 3.46, "learning_rate": 1.5222117647563628e-05, "loss": 0.5055, "step": 11231 }, { "epoch": 3.46, "learning_rate": 1.5221267756446266e-05, "loss": 0.7215, "step": 11232 }, { "epoch": 3.46, "learning_rate": 1.52204178134769e-05, "loss": 0.4954, "step": 11233 }, { "epoch": 3.46, "learning_rate": 1.5219567818663976e-05, "loss": 0.5428, "step": 11234 }, { "epoch": 3.46, "learning_rate": 1.5218717772015934e-05, "loss": 0.417, "step": 11235 }, { "epoch": 3.46, "learning_rate": 1.5217867673541213e-05, "loss": 0.6511, "step": 11236 }, { "epoch": 3.46, "learning_rate": 1.5217017523248257e-05, "loss": 0.4283, "step": 11237 }, { "epoch": 3.46, "learning_rate": 1.521616732114551e-05, "loss": 0.5132, "step": 11238 }, { "epoch": 3.46, "learning_rate": 1.5215317067241415e-05, "loss": 0.6106, "step": 11239 }, { "epoch": 3.46, "learning_rate": 1.5214466761544412e-05, "loss": 0.4703, "step": 11240 }, { "epoch": 3.46, "learning_rate": 1.5213616404062949e-05, "loss": 0.4158, "step": 11241 }, { "epoch": 3.46, "learning_rate": 1.5212765994805472e-05, "loss": 0.5106, "step": 11242 }, { "epoch": 3.46, "learning_rate": 1.5211915533780424e-05, "loss": 0.3977, "step": 11243 }, { "epoch": 3.46, "learning_rate": 1.5211065020996251e-05, "loss": 0.4302, "step": 11244 }, { "epoch": 3.46, "learning_rate": 1.5210214456461399e-05, "loss": 0.4879, "step": 11245 }, { "epoch": 3.46, "learning_rate": 1.5209363840184314e-05, "loss": 0.5555, "step": 11246 }, { "epoch": 3.46, "learning_rate": 1.5208513172173449e-05, "loss": 0.778, "step": 11247 }, { "epoch": 3.46, "learning_rate": 1.5207662452437246e-05, "loss": 0.4994, "step": 11248 }, { "epoch": 3.46, "learning_rate": 1.5206811680984156e-05, "loss": 0.5605, "step": 11249 }, { "epoch": 3.46, "learning_rate": 1.5205960857822626e-05, "loss": 0.5049, "step": 11250 }, { "epoch": 3.46, "learning_rate": 1.5205109982961106e-05, "loss": 0.243, "step": 11251 }, { "epoch": 3.46, "learning_rate": 1.5204259056408046e-05, "loss": 0.4578, "step": 11252 }, { "epoch": 3.46, "learning_rate": 1.5203408078171901e-05, "loss": 0.6193, "step": 11253 }, { "epoch": 3.46, "learning_rate": 1.5202557048261115e-05, "loss": 0.4145, "step": 11254 }, { "epoch": 3.46, "learning_rate": 1.5201705966684146e-05, "loss": 0.6173, "step": 11255 }, { "epoch": 3.46, "learning_rate": 1.5200854833449439e-05, "loss": 0.5439, "step": 11256 }, { "epoch": 3.46, "learning_rate": 1.5200003648565452e-05, "loss": 0.6496, "step": 11257 }, { "epoch": 3.46, "learning_rate": 1.5199152412040636e-05, "loss": 0.2042, "step": 11258 }, { "epoch": 3.46, "learning_rate": 1.5198301123883445e-05, "loss": 0.648, "step": 11259 }, { "epoch": 3.46, "learning_rate": 1.5197449784102333e-05, "loss": 0.7, "step": 11260 }, { "epoch": 3.46, "learning_rate": 1.5196598392705753e-05, "loss": 0.5564, "step": 11261 }, { "epoch": 3.47, "learning_rate": 1.5195746949702167e-05, "loss": 0.4458, "step": 11262 }, { "epoch": 3.47, "learning_rate": 1.5194895455100019e-05, "loss": 0.528, "step": 11263 }, { "epoch": 3.47, "learning_rate": 1.5194043908907774e-05, "loss": 0.3378, "step": 11264 }, { "epoch": 3.47, "learning_rate": 1.5193192311133884e-05, "loss": 0.5893, "step": 11265 }, { "epoch": 3.47, "learning_rate": 1.519234066178681e-05, "loss": 0.469, "step": 11266 }, { "epoch": 3.47, "learning_rate": 1.519148896087501e-05, "loss": 0.5938, "step": 11267 }, { "epoch": 3.47, "learning_rate": 1.5190637208406935e-05, "loss": 0.4079, "step": 11268 }, { "epoch": 3.47, "learning_rate": 1.5189785404391053e-05, "loss": 0.4157, "step": 11269 }, { "epoch": 3.47, "learning_rate": 1.5188933548835816e-05, "loss": 0.3554, "step": 11270 }, { "epoch": 3.47, "learning_rate": 1.5188081641749689e-05, "loss": 0.3224, "step": 11271 }, { "epoch": 3.47, "learning_rate": 1.5187229683141129e-05, "loss": 0.5707, "step": 11272 }, { "epoch": 3.47, "learning_rate": 1.5186377673018595e-05, "loss": 0.5798, "step": 11273 }, { "epoch": 3.47, "learning_rate": 1.5185525611390557e-05, "loss": 0.6831, "step": 11274 }, { "epoch": 3.47, "learning_rate": 1.5184673498265466e-05, "loss": 0.5053, "step": 11275 }, { "epoch": 3.47, "learning_rate": 1.518382133365179e-05, "loss": 0.497, "step": 11276 }, { "epoch": 3.47, "learning_rate": 1.5182969117557989e-05, "loss": 0.6416, "step": 11277 }, { "epoch": 3.47, "learning_rate": 1.5182116849992528e-05, "loss": 0.4876, "step": 11278 }, { "epoch": 3.47, "learning_rate": 1.5181264530963873e-05, "loss": 0.5299, "step": 11279 }, { "epoch": 3.47, "learning_rate": 1.5180412160480484e-05, "loss": 0.3964, "step": 11280 }, { "epoch": 3.47, "learning_rate": 1.5179559738550831e-05, "loss": 0.6269, "step": 11281 }, { "epoch": 3.47, "learning_rate": 1.517870726518337e-05, "loss": 0.4298, "step": 11282 }, { "epoch": 3.47, "learning_rate": 1.5177854740386578e-05, "loss": 0.5691, "step": 11283 }, { "epoch": 3.47, "learning_rate": 1.5177002164168915e-05, "loss": 0.4241, "step": 11284 }, { "epoch": 3.47, "learning_rate": 1.5176149536538849e-05, "loss": 0.7014, "step": 11285 }, { "epoch": 3.47, "learning_rate": 1.517529685750485e-05, "loss": 0.5937, "step": 11286 }, { "epoch": 3.47, "learning_rate": 1.5174444127075378e-05, "loss": 0.3885, "step": 11287 }, { "epoch": 3.47, "learning_rate": 1.5173591345258914e-05, "loss": 0.4395, "step": 11288 }, { "epoch": 3.47, "learning_rate": 1.5172738512063916e-05, "loss": 0.5642, "step": 11289 }, { "epoch": 3.47, "learning_rate": 1.5171885627498857e-05, "loss": 0.4361, "step": 11290 }, { "epoch": 3.47, "learning_rate": 1.5171032691572207e-05, "loss": 0.4525, "step": 11291 }, { "epoch": 3.47, "learning_rate": 1.5170179704292436e-05, "loss": 0.534, "step": 11292 }, { "epoch": 3.47, "learning_rate": 1.5169326665668016e-05, "loss": 0.6179, "step": 11293 }, { "epoch": 3.47, "learning_rate": 1.5168473575707418e-05, "loss": 0.5238, "step": 11294 }, { "epoch": 3.48, "learning_rate": 1.516762043441912e-05, "loss": 0.6935, "step": 11295 }, { "epoch": 3.48, "learning_rate": 1.5166767241811579e-05, "loss": 0.421, "step": 11296 }, { "epoch": 3.48, "learning_rate": 1.5165913997893285e-05, "loss": 0.4814, "step": 11297 }, { "epoch": 3.48, "learning_rate": 1.51650607026727e-05, "loss": 0.4583, "step": 11298 }, { "epoch": 3.48, "learning_rate": 1.51642073561583e-05, "loss": 0.481, "step": 11299 }, { "epoch": 3.48, "learning_rate": 1.5163353958358566e-05, "loss": 0.6172, "step": 11300 }, { "epoch": 3.48, "learning_rate": 1.5162500509281966e-05, "loss": 0.458, "step": 11301 }, { "epoch": 3.48, "learning_rate": 1.516164700893698e-05, "loss": 0.4703, "step": 11302 }, { "epoch": 3.48, "learning_rate": 1.5160793457332078e-05, "loss": 0.3398, "step": 11303 }, { "epoch": 3.48, "learning_rate": 1.5159939854475743e-05, "loss": 0.3895, "step": 11304 }, { "epoch": 3.48, "learning_rate": 1.5159086200376454e-05, "loss": 0.3504, "step": 11305 }, { "epoch": 3.48, "learning_rate": 1.5158232495042678e-05, "loss": 0.3373, "step": 11306 }, { "epoch": 3.48, "learning_rate": 1.5157378738482902e-05, "loss": 0.4104, "step": 11307 }, { "epoch": 3.48, "learning_rate": 1.5156524930705603e-05, "loss": 0.6089, "step": 11308 }, { "epoch": 3.48, "learning_rate": 1.5155671071719256e-05, "loss": 0.4655, "step": 11309 }, { "epoch": 3.48, "learning_rate": 1.5154817161532347e-05, "loss": 0.7987, "step": 11310 }, { "epoch": 3.48, "learning_rate": 1.515396320015335e-05, "loss": 0.4145, "step": 11311 }, { "epoch": 3.48, "learning_rate": 1.515310918759075e-05, "loss": 0.4794, "step": 11312 }, { "epoch": 3.48, "learning_rate": 1.5152255123853023e-05, "loss": 0.3665, "step": 11313 }, { "epoch": 3.48, "learning_rate": 1.515140100894866e-05, "loss": 0.4557, "step": 11314 }, { "epoch": 3.48, "learning_rate": 1.5150546842886136e-05, "loss": 0.6145, "step": 11315 }, { "epoch": 3.48, "learning_rate": 1.514969262567393e-05, "loss": 0.5058, "step": 11316 }, { "epoch": 3.48, "learning_rate": 1.5148838357320537e-05, "loss": 0.4052, "step": 11317 }, { "epoch": 3.48, "learning_rate": 1.5147984037834427e-05, "loss": 0.5175, "step": 11318 }, { "epoch": 3.48, "learning_rate": 1.5147129667224096e-05, "loss": 0.5335, "step": 11319 }, { "epoch": 3.48, "learning_rate": 1.5146275245498023e-05, "loss": 0.7808, "step": 11320 }, { "epoch": 3.48, "learning_rate": 1.5145420772664693e-05, "loss": 0.3671, "step": 11321 }, { "epoch": 3.48, "learning_rate": 1.5144566248732593e-05, "loss": 0.4716, "step": 11322 }, { "epoch": 3.48, "learning_rate": 1.5143711673710206e-05, "loss": 0.5312, "step": 11323 }, { "epoch": 3.48, "learning_rate": 1.5142857047606026e-05, "loss": 0.7151, "step": 11324 }, { "epoch": 3.48, "learning_rate": 1.5142002370428532e-05, "loss": 0.4468, "step": 11325 }, { "epoch": 3.48, "learning_rate": 1.5141147642186218e-05, "loss": 0.4319, "step": 11326 }, { "epoch": 3.49, "learning_rate": 1.5140292862887567e-05, "loss": 0.4481, "step": 11327 }, { "epoch": 3.49, "learning_rate": 1.5139438032541071e-05, "loss": 0.4573, "step": 11328 }, { "epoch": 3.49, "learning_rate": 1.5138583151155221e-05, "loss": 0.6261, "step": 11329 }, { "epoch": 3.49, "learning_rate": 1.5137728218738504e-05, "loss": 0.3252, "step": 11330 }, { "epoch": 3.49, "learning_rate": 1.5136873235299412e-05, "loss": 0.4548, "step": 11331 }, { "epoch": 3.49, "learning_rate": 1.513601820084643e-05, "loss": 0.5215, "step": 11332 }, { "epoch": 3.49, "learning_rate": 1.5135163115388057e-05, "loss": 0.6219, "step": 11333 }, { "epoch": 3.49, "learning_rate": 1.5134307978932782e-05, "loss": 0.4763, "step": 11334 }, { "epoch": 3.49, "learning_rate": 1.5133452791489096e-05, "loss": 0.4806, "step": 11335 }, { "epoch": 3.49, "learning_rate": 1.513259755306549e-05, "loss": 0.7154, "step": 11336 }, { "epoch": 3.49, "learning_rate": 1.5131742263670463e-05, "loss": 0.4729, "step": 11337 }, { "epoch": 3.49, "learning_rate": 1.5130886923312504e-05, "loss": 0.2888, "step": 11338 }, { "epoch": 3.49, "learning_rate": 1.513003153200011e-05, "loss": 0.4511, "step": 11339 }, { "epoch": 3.49, "learning_rate": 1.5129176089741777e-05, "loss": 0.738, "step": 11340 }, { "epoch": 3.49, "learning_rate": 1.5128320596545996e-05, "loss": 0.5019, "step": 11341 }, { "epoch": 3.49, "learning_rate": 1.5127465052421265e-05, "loss": 0.4929, "step": 11342 }, { "epoch": 3.49, "learning_rate": 1.512660945737608e-05, "loss": 0.4101, "step": 11343 }, { "epoch": 3.49, "learning_rate": 1.5125753811418937e-05, "loss": 0.537, "step": 11344 }, { "epoch": 3.49, "learning_rate": 1.5124898114558339e-05, "loss": 0.4386, "step": 11345 }, { "epoch": 3.49, "learning_rate": 1.5124042366802776e-05, "loss": 0.4346, "step": 11346 }, { "epoch": 3.49, "learning_rate": 1.512318656816075e-05, "loss": 0.4827, "step": 11347 }, { "epoch": 3.49, "learning_rate": 1.5122330718640763e-05, "loss": 0.5225, "step": 11348 }, { "epoch": 3.49, "learning_rate": 1.5121474818251308e-05, "loss": 0.4231, "step": 11349 }, { "epoch": 3.49, "learning_rate": 1.5120618867000892e-05, "loss": 0.278, "step": 11350 }, { "epoch": 3.49, "learning_rate": 1.5119762864898004e-05, "loss": 0.4498, "step": 11351 }, { "epoch": 3.49, "learning_rate": 1.5118906811951157e-05, "loss": 0.4205, "step": 11352 }, { "epoch": 3.49, "learning_rate": 1.5118050708168848e-05, "loss": 0.4718, "step": 11353 }, { "epoch": 3.49, "learning_rate": 1.5117194553559575e-05, "loss": 0.6498, "step": 11354 }, { "epoch": 3.49, "learning_rate": 1.511633834813185e-05, "loss": 0.5465, "step": 11355 }, { "epoch": 3.49, "learning_rate": 1.5115482091894164e-05, "loss": 0.4062, "step": 11356 }, { "epoch": 3.49, "learning_rate": 1.5114625784855028e-05, "loss": 0.4889, "step": 11357 }, { "epoch": 3.49, "learning_rate": 1.5113769427022945e-05, "loss": 0.6029, "step": 11358 }, { "epoch": 3.49, "learning_rate": 1.5112913018406418e-05, "loss": 0.3941, "step": 11359 }, { "epoch": 3.5, "learning_rate": 1.5112056559013954e-05, "loss": 0.5217, "step": 11360 }, { "epoch": 3.5, "learning_rate": 1.5111200048854055e-05, "loss": 0.397, "step": 11361 }, { "epoch": 3.5, "learning_rate": 1.5110343487935231e-05, "loss": 0.3738, "step": 11362 }, { "epoch": 3.5, "learning_rate": 1.5109486876265984e-05, "loss": 0.326, "step": 11363 }, { "epoch": 3.5, "learning_rate": 1.5108630213854825e-05, "loss": 0.5612, "step": 11364 }, { "epoch": 3.5, "learning_rate": 1.5107773500710256e-05, "loss": 0.5067, "step": 11365 }, { "epoch": 3.5, "learning_rate": 1.5106916736840793e-05, "loss": 0.6239, "step": 11366 }, { "epoch": 3.5, "learning_rate": 1.5106059922254938e-05, "loss": 0.5142, "step": 11367 }, { "epoch": 3.5, "learning_rate": 1.5105203056961201e-05, "loss": 0.5974, "step": 11368 }, { "epoch": 3.5, "learning_rate": 1.5104346140968096e-05, "loss": 0.6488, "step": 11369 }, { "epoch": 3.5, "learning_rate": 1.5103489174284129e-05, "loss": 0.3581, "step": 11370 }, { "epoch": 3.5, "learning_rate": 1.5102632156917811e-05, "loss": 0.5839, "step": 11371 }, { "epoch": 3.5, "learning_rate": 1.5101775088877653e-05, "loss": 0.5928, "step": 11372 }, { "epoch": 3.5, "learning_rate": 1.5100917970172164e-05, "loss": 0.4955, "step": 11373 }, { "epoch": 3.5, "learning_rate": 1.510006080080986e-05, "loss": 0.4539, "step": 11374 }, { "epoch": 3.5, "learning_rate": 1.5099203580799251e-05, "loss": 0.6276, "step": 11375 }, { "epoch": 3.5, "learning_rate": 1.5098346310148852e-05, "loss": 0.5185, "step": 11376 }, { "epoch": 3.5, "learning_rate": 1.5097488988867176e-05, "loss": 0.4485, "step": 11377 }, { "epoch": 3.5, "learning_rate": 1.5096631616962734e-05, "loss": 0.3584, "step": 11378 }, { "epoch": 3.5, "learning_rate": 1.5095774194444044e-05, "loss": 0.3981, "step": 11379 }, { "epoch": 3.5, "learning_rate": 1.509491672131962e-05, "loss": 0.4961, "step": 11380 }, { "epoch": 3.5, "learning_rate": 1.5094059197597978e-05, "loss": 0.4829, "step": 11381 }, { "epoch": 3.5, "learning_rate": 1.5093201623287631e-05, "loss": 0.4814, "step": 11382 }, { "epoch": 3.5, "learning_rate": 1.5092343998397102e-05, "loss": 0.6444, "step": 11383 }, { "epoch": 3.5, "learning_rate": 1.5091486322934899e-05, "loss": 0.6839, "step": 11384 }, { "epoch": 3.5, "learning_rate": 1.5090628596909546e-05, "loss": 0.4899, "step": 11385 }, { "epoch": 3.5, "learning_rate": 1.5089770820329563e-05, "loss": 0.5506, "step": 11386 }, { "epoch": 3.5, "learning_rate": 1.508891299320346e-05, "loss": 0.3413, "step": 11387 }, { "epoch": 3.5, "learning_rate": 1.5088055115539764e-05, "loss": 0.6518, "step": 11388 }, { "epoch": 3.5, "learning_rate": 1.508719718734699e-05, "loss": 0.4793, "step": 11389 }, { "epoch": 3.5, "learning_rate": 1.5086339208633658e-05, "loss": 0.6842, "step": 11390 }, { "epoch": 3.5, "learning_rate": 1.5085481179408295e-05, "loss": 0.6367, "step": 11391 }, { "epoch": 3.51, "learning_rate": 1.5084623099679409e-05, "loss": 0.4098, "step": 11392 }, { "epoch": 3.51, "learning_rate": 1.5083764969455537e-05, "loss": 0.5076, "step": 11393 }, { "epoch": 3.51, "learning_rate": 1.5082906788745189e-05, "loss": 0.35, "step": 11394 }, { "epoch": 3.51, "learning_rate": 1.5082048557556892e-05, "loss": 0.5386, "step": 11395 }, { "epoch": 3.51, "learning_rate": 1.5081190275899173e-05, "loss": 0.3868, "step": 11396 }, { "epoch": 3.51, "learning_rate": 1.5080331943780549e-05, "loss": 0.7135, "step": 11397 }, { "epoch": 3.51, "learning_rate": 1.5079473561209548e-05, "loss": 0.4614, "step": 11398 }, { "epoch": 3.51, "learning_rate": 1.5078615128194689e-05, "loss": 0.5864, "step": 11399 }, { "epoch": 3.51, "learning_rate": 1.5077756644744506e-05, "loss": 0.4314, "step": 11400 }, { "epoch": 3.51, "learning_rate": 1.5076898110867516e-05, "loss": 1.1308, "step": 11401 }, { "epoch": 3.51, "learning_rate": 1.5076039526572253e-05, "loss": 0.647, "step": 11402 }, { "epoch": 3.51, "learning_rate": 1.5075180891867238e-05, "loss": 0.5963, "step": 11403 }, { "epoch": 3.51, "learning_rate": 1.5074322206760999e-05, "loss": 0.5353, "step": 11404 }, { "epoch": 3.51, "learning_rate": 1.5073463471262062e-05, "loss": 0.5922, "step": 11405 }, { "epoch": 3.51, "learning_rate": 1.5072604685378957e-05, "loss": 0.4684, "step": 11406 }, { "epoch": 3.51, "learning_rate": 1.5071745849120216e-05, "loss": 0.558, "step": 11407 }, { "epoch": 3.51, "learning_rate": 1.507088696249436e-05, "loss": 0.458, "step": 11408 }, { "epoch": 3.51, "learning_rate": 1.5070028025509925e-05, "loss": 0.5238, "step": 11409 }, { "epoch": 3.51, "learning_rate": 1.506916903817544e-05, "loss": 0.4419, "step": 11410 }, { "epoch": 3.51, "learning_rate": 1.5068310000499433e-05, "loss": 0.6021, "step": 11411 }, { "epoch": 3.51, "learning_rate": 1.5067450912490438e-05, "loss": 0.4478, "step": 11412 }, { "epoch": 3.51, "learning_rate": 1.5066591774156983e-05, "loss": 0.5607, "step": 11413 }, { "epoch": 3.51, "learning_rate": 1.5065732585507604e-05, "loss": 0.3749, "step": 11414 }, { "epoch": 3.51, "learning_rate": 1.5064873346550832e-05, "loss": 0.4541, "step": 11415 }, { "epoch": 3.51, "learning_rate": 1.5064014057295198e-05, "loss": 0.4761, "step": 11416 }, { "epoch": 3.51, "learning_rate": 1.506315471774924e-05, "loss": 0.4513, "step": 11417 }, { "epoch": 3.51, "learning_rate": 1.5062295327921485e-05, "loss": 0.7049, "step": 11418 }, { "epoch": 3.51, "learning_rate": 1.5061435887820474e-05, "loss": 0.4966, "step": 11419 }, { "epoch": 3.51, "learning_rate": 1.5060576397454742e-05, "loss": 0.4887, "step": 11420 }, { "epoch": 3.51, "learning_rate": 1.505971685683282e-05, "loss": 0.4612, "step": 11421 }, { "epoch": 3.51, "learning_rate": 1.5058857265963248e-05, "loss": 0.5355, "step": 11422 }, { "epoch": 3.51, "learning_rate": 1.505799762485456e-05, "loss": 0.6332, "step": 11423 }, { "epoch": 3.51, "learning_rate": 1.5057137933515294e-05, "loss": 0.3741, "step": 11424 }, { "epoch": 3.52, "learning_rate": 1.5056278191953986e-05, "loss": 0.3943, "step": 11425 }, { "epoch": 3.52, "learning_rate": 1.505541840017918e-05, "loss": 0.6399, "step": 11426 }, { "epoch": 3.52, "learning_rate": 1.5054558558199408e-05, "loss": 0.6757, "step": 11427 }, { "epoch": 3.52, "learning_rate": 1.5053698666023207e-05, "loss": 0.6553, "step": 11428 }, { "epoch": 3.52, "learning_rate": 1.5052838723659124e-05, "loss": 0.5864, "step": 11429 }, { "epoch": 3.52, "learning_rate": 1.5051978731115695e-05, "loss": 0.4339, "step": 11430 }, { "epoch": 3.52, "learning_rate": 1.5051118688401465e-05, "loss": 0.4437, "step": 11431 }, { "epoch": 3.52, "learning_rate": 1.5050258595524968e-05, "loss": 0.5083, "step": 11432 }, { "epoch": 3.52, "learning_rate": 1.5049398452494748e-05, "loss": 0.6444, "step": 11433 }, { "epoch": 3.52, "learning_rate": 1.5048538259319347e-05, "loss": 0.5106, "step": 11434 }, { "epoch": 3.52, "learning_rate": 1.504767801600731e-05, "loss": 0.4018, "step": 11435 }, { "epoch": 3.52, "learning_rate": 1.504681772256718e-05, "loss": 0.396, "step": 11436 }, { "epoch": 3.52, "learning_rate": 1.5045957379007491e-05, "loss": 0.3878, "step": 11437 }, { "epoch": 3.52, "learning_rate": 1.5045096985336804e-05, "loss": 0.5711, "step": 11438 }, { "epoch": 3.52, "learning_rate": 1.504423654156365e-05, "loss": 0.4964, "step": 11439 }, { "epoch": 3.52, "learning_rate": 1.5043376047696576e-05, "loss": 0.4799, "step": 11440 }, { "epoch": 3.52, "learning_rate": 1.5042515503744137e-05, "loss": 0.4808, "step": 11441 }, { "epoch": 3.52, "learning_rate": 1.5041654909714864e-05, "loss": 0.6399, "step": 11442 }, { "epoch": 3.52, "learning_rate": 1.5040794265617315e-05, "loss": 0.6957, "step": 11443 }, { "epoch": 3.52, "learning_rate": 1.5039933571460031e-05, "loss": 0.5468, "step": 11444 }, { "epoch": 3.52, "learning_rate": 1.5039072827251563e-05, "loss": 0.5617, "step": 11445 }, { "epoch": 3.52, "learning_rate": 1.5038212033000462e-05, "loss": 0.6117, "step": 11446 }, { "epoch": 3.52, "learning_rate": 1.5037351188715265e-05, "loss": 0.6071, "step": 11447 }, { "epoch": 3.52, "learning_rate": 1.5036490294404532e-05, "loss": 0.5626, "step": 11448 }, { "epoch": 3.52, "learning_rate": 1.5035629350076808e-05, "loss": 0.5086, "step": 11449 }, { "epoch": 3.52, "learning_rate": 1.5034768355740644e-05, "loss": 0.5282, "step": 11450 }, { "epoch": 3.52, "learning_rate": 1.503390731140459e-05, "loss": 0.2045, "step": 11451 }, { "epoch": 3.52, "learning_rate": 1.5033046217077196e-05, "loss": 0.4794, "step": 11452 }, { "epoch": 3.52, "learning_rate": 1.5032185072767015e-05, "loss": 0.6873, "step": 11453 }, { "epoch": 3.52, "learning_rate": 1.5031323878482597e-05, "loss": 0.5428, "step": 11454 }, { "epoch": 3.52, "learning_rate": 1.50304626342325e-05, "loss": 0.479, "step": 11455 }, { "epoch": 3.52, "learning_rate": 1.5029601340025268e-05, "loss": 0.6249, "step": 11456 }, { "epoch": 3.53, "learning_rate": 1.5028739995869464e-05, "loss": 0.3756, "step": 11457 }, { "epoch": 3.53, "learning_rate": 1.5027878601773633e-05, "loss": 0.4531, "step": 11458 }, { "epoch": 3.53, "learning_rate": 1.5027017157746336e-05, "loss": 0.4214, "step": 11459 }, { "epoch": 3.53, "learning_rate": 1.5026155663796123e-05, "loss": 0.5929, "step": 11460 }, { "epoch": 3.53, "learning_rate": 1.5025294119931554e-05, "loss": 0.4898, "step": 11461 }, { "epoch": 3.53, "learning_rate": 1.5024432526161184e-05, "loss": 0.4649, "step": 11462 }, { "epoch": 3.53, "learning_rate": 1.5023570882493566e-05, "loss": 0.5847, "step": 11463 }, { "epoch": 3.53, "learning_rate": 1.5022709188937259e-05, "loss": 0.5645, "step": 11464 }, { "epoch": 3.53, "learning_rate": 1.5021847445500823e-05, "loss": 0.5369, "step": 11465 }, { "epoch": 3.53, "learning_rate": 1.5020985652192814e-05, "loss": 0.5493, "step": 11466 }, { "epoch": 3.53, "learning_rate": 1.502012380902179e-05, "loss": 0.4584, "step": 11467 }, { "epoch": 3.53, "learning_rate": 1.5019261915996307e-05, "loss": 0.4762, "step": 11468 }, { "epoch": 3.53, "learning_rate": 1.5018399973124929e-05, "loss": 0.3073, "step": 11469 }, { "epoch": 3.53, "learning_rate": 1.5017537980416214e-05, "loss": 0.464, "step": 11470 }, { "epoch": 3.53, "learning_rate": 1.5016675937878721e-05, "loss": 0.6789, "step": 11471 }, { "epoch": 3.53, "learning_rate": 1.5015813845521017e-05, "loss": 0.6227, "step": 11472 }, { "epoch": 3.53, "learning_rate": 1.5014951703351655e-05, "loss": 0.5087, "step": 11473 }, { "epoch": 3.53, "learning_rate": 1.5014089511379202e-05, "loss": 0.3679, "step": 11474 }, { "epoch": 3.53, "learning_rate": 1.5013227269612216e-05, "loss": 0.7281, "step": 11475 }, { "epoch": 3.53, "learning_rate": 1.5012364978059266e-05, "loss": 0.4637, "step": 11476 }, { "epoch": 3.53, "learning_rate": 1.5011502636728914e-05, "loss": 0.4555, "step": 11477 }, { "epoch": 3.53, "learning_rate": 1.5010640245629717e-05, "loss": 0.477, "step": 11478 }, { "epoch": 3.53, "learning_rate": 1.5009777804770249e-05, "loss": 0.3565, "step": 11479 }, { "epoch": 3.53, "learning_rate": 1.5008915314159068e-05, "loss": 0.5706, "step": 11480 }, { "epoch": 3.53, "learning_rate": 1.5008052773804741e-05, "loss": 0.7587, "step": 11481 }, { "epoch": 3.53, "learning_rate": 1.5007190183715836e-05, "loss": 0.6378, "step": 11482 }, { "epoch": 3.53, "learning_rate": 1.5006327543900919e-05, "loss": 0.6411, "step": 11483 }, { "epoch": 3.53, "learning_rate": 1.5005464854368553e-05, "loss": 0.603, "step": 11484 }, { "epoch": 3.53, "learning_rate": 1.5004602115127306e-05, "loss": 0.2781, "step": 11485 }, { "epoch": 3.53, "learning_rate": 1.500373932618575e-05, "loss": 0.4593, "step": 11486 }, { "epoch": 3.53, "learning_rate": 1.5002876487552453e-05, "loss": 0.6391, "step": 11487 }, { "epoch": 3.53, "learning_rate": 1.5002013599235981e-05, "loss": 0.5014, "step": 11488 }, { "epoch": 3.53, "learning_rate": 1.5001150661244901e-05, "loss": 0.6588, "step": 11489 }, { "epoch": 3.54, "learning_rate": 1.5000287673587788e-05, "loss": 0.5558, "step": 11490 }, { "epoch": 3.54, "learning_rate": 1.499942463627321e-05, "loss": 0.3956, "step": 11491 }, { "epoch": 3.54, "learning_rate": 1.4998561549309739e-05, "loss": 0.3908, "step": 11492 }, { "epoch": 3.54, "learning_rate": 1.4997698412705941e-05, "loss": 0.6758, "step": 11493 }, { "epoch": 3.54, "learning_rate": 1.4996835226470392e-05, "loss": 0.6036, "step": 11494 }, { "epoch": 3.54, "learning_rate": 1.4995971990611668e-05, "loss": 0.5976, "step": 11495 }, { "epoch": 3.54, "learning_rate": 1.4995108705138335e-05, "loss": 0.4943, "step": 11496 }, { "epoch": 3.54, "learning_rate": 1.499424537005897e-05, "loss": 0.4448, "step": 11497 }, { "epoch": 3.54, "learning_rate": 1.4993381985382148e-05, "loss": 0.4668, "step": 11498 }, { "epoch": 3.54, "learning_rate": 1.4992518551116436e-05, "loss": 0.4639, "step": 11499 }, { "epoch": 3.54, "learning_rate": 1.4991655067270414e-05, "loss": 0.6139, "step": 11500 }, { "epoch": 3.54, "learning_rate": 1.499079153385266e-05, "loss": 0.5959, "step": 11501 }, { "epoch": 3.54, "learning_rate": 1.4989927950871744e-05, "loss": 0.3844, "step": 11502 }, { "epoch": 3.54, "learning_rate": 1.4989064318336246e-05, "loss": 0.3121, "step": 11503 }, { "epoch": 3.54, "learning_rate": 1.498820063625474e-05, "loss": 0.3596, "step": 11504 }, { "epoch": 3.54, "learning_rate": 1.4987336904635805e-05, "loss": 0.5072, "step": 11505 }, { "epoch": 3.54, "learning_rate": 1.498647312348802e-05, "loss": 0.4439, "step": 11506 }, { "epoch": 3.54, "learning_rate": 1.4985609292819958e-05, "loss": 0.6005, "step": 11507 }, { "epoch": 3.54, "learning_rate": 1.4984745412640202e-05, "loss": 0.6032, "step": 11508 }, { "epoch": 3.54, "learning_rate": 1.498388148295733e-05, "loss": 0.3485, "step": 11509 }, { "epoch": 3.54, "learning_rate": 1.4983017503779922e-05, "loss": 0.713, "step": 11510 }, { "epoch": 3.54, "learning_rate": 1.4982153475116557e-05, "loss": 0.6354, "step": 11511 }, { "epoch": 3.54, "learning_rate": 1.4981289396975818e-05, "loss": 0.5856, "step": 11512 }, { "epoch": 3.54, "learning_rate": 1.498042526936628e-05, "loss": 0.5732, "step": 11513 }, { "epoch": 3.54, "learning_rate": 1.4979561092296533e-05, "loss": 0.5182, "step": 11514 }, { "epoch": 3.54, "learning_rate": 1.4978696865775156e-05, "loss": 0.4509, "step": 11515 }, { "epoch": 3.54, "learning_rate": 1.4977832589810725e-05, "loss": 0.5359, "step": 11516 }, { "epoch": 3.54, "learning_rate": 1.4976968264411836e-05, "loss": 0.5093, "step": 11517 }, { "epoch": 3.54, "learning_rate": 1.4976103889587061e-05, "loss": 0.5272, "step": 11518 }, { "epoch": 3.54, "learning_rate": 1.4975239465344989e-05, "loss": 0.5502, "step": 11519 }, { "epoch": 3.54, "learning_rate": 1.4974374991694203e-05, "loss": 0.4776, "step": 11520 }, { "epoch": 3.54, "learning_rate": 1.4973510468643288e-05, "loss": 0.5735, "step": 11521 }, { "epoch": 3.55, "learning_rate": 1.4972645896200834e-05, "loss": 0.5233, "step": 11522 }, { "epoch": 3.55, "learning_rate": 1.4971781274375421e-05, "loss": 0.5831, "step": 11523 }, { "epoch": 3.55, "learning_rate": 1.4970916603175638e-05, "loss": 0.5238, "step": 11524 }, { "epoch": 3.55, "learning_rate": 1.4970051882610073e-05, "loss": 0.4582, "step": 11525 }, { "epoch": 3.55, "learning_rate": 1.4969187112687308e-05, "loss": 0.4692, "step": 11526 }, { "epoch": 3.55, "learning_rate": 1.4968322293415942e-05, "loss": 0.4311, "step": 11527 }, { "epoch": 3.55, "learning_rate": 1.4967457424804553e-05, "loss": 0.5491, "step": 11528 }, { "epoch": 3.55, "learning_rate": 1.4966592506861732e-05, "loss": 0.6461, "step": 11529 }, { "epoch": 3.55, "learning_rate": 1.4965727539596071e-05, "loss": 0.6158, "step": 11530 }, { "epoch": 3.55, "learning_rate": 1.4964862523016159e-05, "loss": 0.4292, "step": 11531 }, { "epoch": 3.55, "learning_rate": 1.4963997457130588e-05, "loss": 0.5017, "step": 11532 }, { "epoch": 3.55, "learning_rate": 1.4963132341947948e-05, "loss": 0.4528, "step": 11533 }, { "epoch": 3.55, "learning_rate": 1.4962267177476828e-05, "loss": 0.5723, "step": 11534 }, { "epoch": 3.55, "learning_rate": 1.4961401963725818e-05, "loss": 0.4242, "step": 11535 }, { "epoch": 3.55, "learning_rate": 1.496053670070352e-05, "loss": 0.3664, "step": 11536 }, { "epoch": 3.55, "learning_rate": 1.4959671388418517e-05, "loss": 0.3355, "step": 11537 }, { "epoch": 3.55, "learning_rate": 1.4958806026879411e-05, "loss": 0.4562, "step": 11538 }, { "epoch": 3.55, "learning_rate": 1.4957940616094787e-05, "loss": 0.5892, "step": 11539 }, { "epoch": 3.55, "learning_rate": 1.4957075156073242e-05, "loss": 0.4502, "step": 11540 }, { "epoch": 3.55, "learning_rate": 1.4956209646823375e-05, "loss": 0.5636, "step": 11541 }, { "epoch": 3.55, "learning_rate": 1.4955344088353777e-05, "loss": 0.5217, "step": 11542 }, { "epoch": 3.55, "learning_rate": 1.4954478480673044e-05, "loss": 0.5719, "step": 11543 }, { "epoch": 3.55, "learning_rate": 1.4953612823789777e-05, "loss": 0.4654, "step": 11544 }, { "epoch": 3.55, "learning_rate": 1.4952747117712567e-05, "loss": 0.6816, "step": 11545 }, { "epoch": 3.55, "learning_rate": 1.4951881362450014e-05, "loss": 0.5524, "step": 11546 }, { "epoch": 3.55, "learning_rate": 1.4951015558010715e-05, "loss": 0.6436, "step": 11547 }, { "epoch": 3.55, "learning_rate": 1.495014970440327e-05, "loss": 0.4084, "step": 11548 }, { "epoch": 3.55, "learning_rate": 1.4949283801636272e-05, "loss": 0.4908, "step": 11549 }, { "epoch": 3.55, "learning_rate": 1.4948417849718329e-05, "loss": 0.4961, "step": 11550 }, { "epoch": 3.55, "learning_rate": 1.4947551848658036e-05, "loss": 0.3704, "step": 11551 }, { "epoch": 3.55, "learning_rate": 1.4946685798463991e-05, "loss": 0.4449, "step": 11552 }, { "epoch": 3.55, "learning_rate": 1.4945819699144798e-05, "loss": 0.43, "step": 11553 }, { "epoch": 3.55, "learning_rate": 1.4944953550709057e-05, "loss": 0.5624, "step": 11554 }, { "epoch": 3.56, "learning_rate": 1.494408735316537e-05, "loss": 0.4891, "step": 11555 }, { "epoch": 3.56, "learning_rate": 1.4943221106522339e-05, "loss": 0.5339, "step": 11556 }, { "epoch": 3.56, "learning_rate": 1.4942354810788567e-05, "loss": 0.7761, "step": 11557 }, { "epoch": 3.56, "learning_rate": 1.4941488465972658e-05, "loss": 0.4404, "step": 11558 }, { "epoch": 3.56, "learning_rate": 1.494062207208321e-05, "loss": 0.4881, "step": 11559 }, { "epoch": 3.56, "learning_rate": 1.4939755629128833e-05, "loss": 0.5287, "step": 11560 }, { "epoch": 3.56, "learning_rate": 1.493888913711813e-05, "loss": 0.4473, "step": 11561 }, { "epoch": 3.56, "learning_rate": 1.493802259605971e-05, "loss": 0.4448, "step": 11562 }, { "epoch": 3.56, "learning_rate": 1.493715600596217e-05, "loss": 0.6045, "step": 11563 }, { "epoch": 3.56, "learning_rate": 1.4936289366834123e-05, "loss": 0.5839, "step": 11564 }, { "epoch": 3.56, "learning_rate": 1.4935422678684173e-05, "loss": 0.5278, "step": 11565 }, { "epoch": 3.56, "learning_rate": 1.4934555941520926e-05, "loss": 0.4037, "step": 11566 }, { "epoch": 3.56, "learning_rate": 1.4933689155352992e-05, "loss": 0.4052, "step": 11567 }, { "epoch": 3.56, "learning_rate": 1.4932822320188975e-05, "loss": 0.3902, "step": 11568 }, { "epoch": 3.56, "learning_rate": 1.4931955436037488e-05, "loss": 0.5594, "step": 11569 }, { "epoch": 3.56, "learning_rate": 1.4931088502907141e-05, "loss": 0.4224, "step": 11570 }, { "epoch": 3.56, "learning_rate": 1.4930221520806536e-05, "loss": 0.6647, "step": 11571 }, { "epoch": 3.56, "learning_rate": 1.492935448974429e-05, "loss": 0.4152, "step": 11572 }, { "epoch": 3.56, "learning_rate": 1.4928487409729009e-05, "loss": 0.6145, "step": 11573 }, { "epoch": 3.56, "learning_rate": 1.4927620280769308e-05, "loss": 0.6107, "step": 11574 }, { "epoch": 3.56, "learning_rate": 1.4926753102873797e-05, "loss": 0.5734, "step": 11575 }, { "epoch": 3.56, "learning_rate": 1.4925885876051083e-05, "loss": 0.3864, "step": 11576 }, { "epoch": 3.56, "learning_rate": 1.4925018600309784e-05, "loss": 0.5358, "step": 11577 }, { "epoch": 3.56, "learning_rate": 1.4924151275658512e-05, "loss": 0.5088, "step": 11578 }, { "epoch": 3.56, "learning_rate": 1.4923283902105881e-05, "loss": 0.4756, "step": 11579 }, { "epoch": 3.56, "learning_rate": 1.4922416479660499e-05, "loss": 0.3433, "step": 11580 }, { "epoch": 3.56, "learning_rate": 1.492154900833099e-05, "loss": 0.4909, "step": 11581 }, { "epoch": 3.56, "learning_rate": 1.4920681488125959e-05, "loss": 0.4248, "step": 11582 }, { "epoch": 3.56, "learning_rate": 1.4919813919054029e-05, "loss": 0.4369, "step": 11583 }, { "epoch": 3.56, "learning_rate": 1.4918946301123814e-05, "loss": 0.6485, "step": 11584 }, { "epoch": 3.56, "learning_rate": 1.4918078634343921e-05, "loss": 0.6794, "step": 11585 }, { "epoch": 3.56, "learning_rate": 1.4917210918722983e-05, "loss": 0.4447, "step": 11586 }, { "epoch": 3.57, "learning_rate": 1.4916343154269605e-05, "loss": 0.5066, "step": 11587 }, { "epoch": 3.57, "learning_rate": 1.491547534099241e-05, "loss": 0.4943, "step": 11588 }, { "epoch": 3.57, "learning_rate": 1.4914607478900015e-05, "loss": 0.2823, "step": 11589 }, { "epoch": 3.57, "learning_rate": 1.4913739568001034e-05, "loss": 0.3925, "step": 11590 }, { "epoch": 3.57, "learning_rate": 1.4912871608304094e-05, "loss": 0.3207, "step": 11591 }, { "epoch": 3.57, "learning_rate": 1.4912003599817808e-05, "loss": 0.3699, "step": 11592 }, { "epoch": 3.57, "learning_rate": 1.4911135542550806e-05, "loss": 0.4918, "step": 11593 }, { "epoch": 3.57, "learning_rate": 1.4910267436511694e-05, "loss": 0.6417, "step": 11594 }, { "epoch": 3.57, "learning_rate": 1.4909399281709105e-05, "loss": 0.4512, "step": 11595 }, { "epoch": 3.57, "learning_rate": 1.4908531078151655e-05, "loss": 0.4667, "step": 11596 }, { "epoch": 3.57, "learning_rate": 1.4907662825847968e-05, "loss": 0.4661, "step": 11597 }, { "epoch": 3.57, "learning_rate": 1.4906794524806669e-05, "loss": 0.486, "step": 11598 }, { "epoch": 3.57, "learning_rate": 1.4905926175036374e-05, "loss": 0.55, "step": 11599 }, { "epoch": 3.57, "learning_rate": 1.4905057776545713e-05, "loss": 0.4145, "step": 11600 }, { "epoch": 3.57, "learning_rate": 1.4904189329343308e-05, "loss": 0.3964, "step": 11601 }, { "epoch": 3.57, "learning_rate": 1.490332083343778e-05, "loss": 0.5848, "step": 11602 }, { "epoch": 3.57, "learning_rate": 1.4902452288837761e-05, "loss": 0.5684, "step": 11603 }, { "epoch": 3.57, "learning_rate": 1.490158369555187e-05, "loss": 0.5754, "step": 11604 }, { "epoch": 3.57, "learning_rate": 1.4900715053588738e-05, "loss": 0.5869, "step": 11605 }, { "epoch": 3.57, "learning_rate": 1.4899846362956987e-05, "loss": 0.637, "step": 11606 }, { "epoch": 3.57, "learning_rate": 1.4898977623665243e-05, "loss": 0.4036, "step": 11607 }, { "epoch": 3.57, "learning_rate": 1.4898108835722143e-05, "loss": 0.3973, "step": 11608 }, { "epoch": 3.57, "learning_rate": 1.4897239999136302e-05, "loss": 0.5021, "step": 11609 }, { "epoch": 3.57, "learning_rate": 1.4896371113916357e-05, "loss": 0.4899, "step": 11610 }, { "epoch": 3.57, "learning_rate": 1.4895502180070931e-05, "loss": 0.5886, "step": 11611 }, { "epoch": 3.57, "learning_rate": 1.4894633197608661e-05, "loss": 0.4231, "step": 11612 }, { "epoch": 3.57, "learning_rate": 1.489376416653817e-05, "loss": 0.4509, "step": 11613 }, { "epoch": 3.57, "learning_rate": 1.4892895086868091e-05, "loss": 0.6924, "step": 11614 }, { "epoch": 3.57, "learning_rate": 1.4892025958607055e-05, "loss": 0.4765, "step": 11615 }, { "epoch": 3.57, "learning_rate": 1.4891156781763692e-05, "loss": 0.3992, "step": 11616 }, { "epoch": 3.57, "learning_rate": 1.4890287556346634e-05, "loss": 0.6723, "step": 11617 }, { "epoch": 3.57, "learning_rate": 1.4889418282364514e-05, "loss": 0.5692, "step": 11618 }, { "epoch": 3.57, "learning_rate": 1.4888548959825966e-05, "loss": 0.4982, "step": 11619 }, { "epoch": 3.58, "learning_rate": 1.4887679588739619e-05, "loss": 0.5493, "step": 11620 }, { "epoch": 3.58, "learning_rate": 1.4886810169114107e-05, "loss": 0.41, "step": 11621 }, { "epoch": 3.58, "learning_rate": 1.488594070095807e-05, "loss": 0.6613, "step": 11622 }, { "epoch": 3.58, "learning_rate": 1.4885071184280137e-05, "loss": 0.4857, "step": 11623 }, { "epoch": 3.58, "learning_rate": 1.4884201619088948e-05, "loss": 0.2953, "step": 11624 }, { "epoch": 3.58, "learning_rate": 1.4883332005393131e-05, "loss": 0.4335, "step": 11625 }, { "epoch": 3.58, "learning_rate": 1.488246234320133e-05, "loss": 0.6494, "step": 11626 }, { "epoch": 3.58, "learning_rate": 1.4881592632522175e-05, "loss": 0.3331, "step": 11627 }, { "epoch": 3.58, "learning_rate": 1.4880722873364311e-05, "loss": 0.4692, "step": 11628 }, { "epoch": 3.58, "learning_rate": 1.4879853065736366e-05, "loss": 0.4886, "step": 11629 }, { "epoch": 3.58, "learning_rate": 1.4878983209646985e-05, "loss": 0.551, "step": 11630 }, { "epoch": 3.58, "learning_rate": 1.4878113305104803e-05, "loss": 0.4374, "step": 11631 }, { "epoch": 3.58, "learning_rate": 1.487724335211846e-05, "loss": 0.5594, "step": 11632 }, { "epoch": 3.58, "learning_rate": 1.4876373350696595e-05, "loss": 0.5192, "step": 11633 }, { "epoch": 3.58, "learning_rate": 1.4875503300847851e-05, "loss": 0.7599, "step": 11634 }, { "epoch": 3.58, "learning_rate": 1.4874633202580864e-05, "loss": 0.4615, "step": 11635 }, { "epoch": 3.58, "learning_rate": 1.4873763055904276e-05, "loss": 0.803, "step": 11636 }, { "epoch": 3.58, "learning_rate": 1.487289286082673e-05, "loss": 0.5607, "step": 11637 }, { "epoch": 3.58, "learning_rate": 1.487202261735687e-05, "loss": 0.4423, "step": 11638 }, { "epoch": 3.58, "learning_rate": 1.4871152325503331e-05, "loss": 0.5663, "step": 11639 }, { "epoch": 3.58, "learning_rate": 1.4870281985274759e-05, "loss": 0.3893, "step": 11640 }, { "epoch": 3.58, "learning_rate": 1.4869411596679802e-05, "loss": 0.4792, "step": 11641 }, { "epoch": 3.58, "learning_rate": 1.4868541159727097e-05, "loss": 0.3843, "step": 11642 }, { "epoch": 3.58, "learning_rate": 1.4867670674425294e-05, "loss": 0.7481, "step": 11643 }, { "epoch": 3.58, "learning_rate": 1.4866800140783035e-05, "loss": 0.3099, "step": 11644 }, { "epoch": 3.58, "learning_rate": 1.4865929558808962e-05, "loss": 0.5479, "step": 11645 }, { "epoch": 3.58, "learning_rate": 1.4865058928511726e-05, "loss": 0.5244, "step": 11646 }, { "epoch": 3.58, "learning_rate": 1.486418824989997e-05, "loss": 0.5642, "step": 11647 }, { "epoch": 3.58, "learning_rate": 1.4863317522982345e-05, "loss": 0.5101, "step": 11648 }, { "epoch": 3.58, "learning_rate": 1.4862446747767492e-05, "loss": 0.4447, "step": 11649 }, { "epoch": 3.58, "learning_rate": 1.4861575924264063e-05, "loss": 0.4871, "step": 11650 }, { "epoch": 3.58, "learning_rate": 1.4860705052480705e-05, "loss": 0.5944, "step": 11651 }, { "epoch": 3.59, "learning_rate": 1.485983413242606e-05, "loss": 0.3688, "step": 11652 }, { "epoch": 3.59, "learning_rate": 1.4858963164108793e-05, "loss": 0.4166, "step": 11653 }, { "epoch": 3.59, "learning_rate": 1.4858092147537534e-05, "loss": 0.5797, "step": 11654 }, { "epoch": 3.59, "learning_rate": 1.485722108272095e-05, "loss": 0.5115, "step": 11655 }, { "epoch": 3.59, "learning_rate": 1.4856349969667681e-05, "loss": 0.444, "step": 11656 }, { "epoch": 3.59, "learning_rate": 1.4855478808386378e-05, "loss": 0.4115, "step": 11657 }, { "epoch": 3.59, "learning_rate": 1.4854607598885702e-05, "loss": 0.438, "step": 11658 }, { "epoch": 3.59, "learning_rate": 1.4853736341174293e-05, "loss": 0.4947, "step": 11659 }, { "epoch": 3.59, "learning_rate": 1.4852865035260813e-05, "loss": 0.7656, "step": 11660 }, { "epoch": 3.59, "learning_rate": 1.4851993681153905e-05, "loss": 0.426, "step": 11661 }, { "epoch": 3.59, "learning_rate": 1.4851122278862233e-05, "loss": 0.6796, "step": 11662 }, { "epoch": 3.59, "learning_rate": 1.4850250828394445e-05, "loss": 0.5426, "step": 11663 }, { "epoch": 3.59, "learning_rate": 1.4849379329759194e-05, "loss": 0.4992, "step": 11664 }, { "epoch": 3.59, "learning_rate": 1.484850778296514e-05, "loss": 0.3456, "step": 11665 }, { "epoch": 3.59, "learning_rate": 1.4847636188020933e-05, "loss": 0.441, "step": 11666 }, { "epoch": 3.59, "learning_rate": 1.4846764544935233e-05, "loss": 0.5032, "step": 11667 }, { "epoch": 3.59, "learning_rate": 1.4845892853716692e-05, "loss": 0.6234, "step": 11668 }, { "epoch": 3.59, "learning_rate": 1.484502111437397e-05, "loss": 0.3338, "step": 11669 }, { "epoch": 3.59, "learning_rate": 1.4844149326915722e-05, "loss": 0.6214, "step": 11670 }, { "epoch": 3.59, "learning_rate": 1.4843277491350607e-05, "loss": 0.7091, "step": 11671 }, { "epoch": 3.59, "learning_rate": 1.4842405607687282e-05, "loss": 0.3218, "step": 11672 }, { "epoch": 3.59, "learning_rate": 1.4841533675934408e-05, "loss": 0.3216, "step": 11673 }, { "epoch": 3.59, "learning_rate": 1.4840661696100643e-05, "loss": 0.6165, "step": 11674 }, { "epoch": 3.59, "learning_rate": 1.4839789668194646e-05, "loss": 0.6563, "step": 11675 }, { "epoch": 3.59, "learning_rate": 1.4838917592225075e-05, "loss": 0.56, "step": 11676 }, { "epoch": 3.59, "learning_rate": 1.4838045468200595e-05, "loss": 0.5254, "step": 11677 }, { "epoch": 3.59, "learning_rate": 1.4837173296129861e-05, "loss": 0.4143, "step": 11678 }, { "epoch": 3.59, "learning_rate": 1.4836301076021542e-05, "loss": 0.448, "step": 11679 }, { "epoch": 3.59, "learning_rate": 1.4835428807884294e-05, "loss": 0.8213, "step": 11680 }, { "epoch": 3.59, "learning_rate": 1.4834556491726781e-05, "loss": 0.5575, "step": 11681 }, { "epoch": 3.59, "learning_rate": 1.4833684127557671e-05, "loss": 0.6573, "step": 11682 }, { "epoch": 3.59, "learning_rate": 1.4832811715385616e-05, "loss": 0.3948, "step": 11683 }, { "epoch": 3.59, "learning_rate": 1.4831939255219292e-05, "loss": 0.4302, "step": 11684 }, { "epoch": 3.6, "learning_rate": 1.4831066747067355e-05, "loss": 0.6667, "step": 11685 }, { "epoch": 3.6, "learning_rate": 1.4830194190938475e-05, "loss": 0.433, "step": 11686 }, { "epoch": 3.6, "learning_rate": 1.4829321586841312e-05, "loss": 0.5332, "step": 11687 }, { "epoch": 3.6, "learning_rate": 1.4828448934784536e-05, "loss": 0.5954, "step": 11688 }, { "epoch": 3.6, "learning_rate": 1.4827576234776814e-05, "loss": 0.3788, "step": 11689 }, { "epoch": 3.6, "learning_rate": 1.4826703486826808e-05, "loss": 0.5923, "step": 11690 }, { "epoch": 3.6, "learning_rate": 1.482583069094319e-05, "loss": 0.421, "step": 11691 }, { "epoch": 3.6, "learning_rate": 1.4824957847134625e-05, "loss": 0.5799, "step": 11692 }, { "epoch": 3.6, "learning_rate": 1.4824084955409782e-05, "loss": 0.3149, "step": 11693 }, { "epoch": 3.6, "learning_rate": 1.482321201577733e-05, "loss": 0.6974, "step": 11694 }, { "epoch": 3.6, "learning_rate": 1.4822339028245935e-05, "loss": 0.529, "step": 11695 }, { "epoch": 3.6, "learning_rate": 1.4821465992824271e-05, "loss": 0.305, "step": 11696 }, { "epoch": 3.6, "learning_rate": 1.4820592909521004e-05, "loss": 0.622, "step": 11697 }, { "epoch": 3.6, "learning_rate": 1.4819719778344808e-05, "loss": 0.5136, "step": 11698 }, { "epoch": 3.6, "learning_rate": 1.4818846599304352e-05, "loss": 0.5174, "step": 11699 }, { "epoch": 3.6, "learning_rate": 1.4817973372408312e-05, "loss": 0.3166, "step": 11700 }, { "epoch": 3.6, "learning_rate": 1.4817100097665353e-05, "loss": 0.4823, "step": 11701 }, { "epoch": 3.6, "learning_rate": 1.4816226775084148e-05, "loss": 0.3885, "step": 11702 }, { "epoch": 3.6, "learning_rate": 1.4815353404673376e-05, "loss": 0.607, "step": 11703 }, { "epoch": 3.6, "learning_rate": 1.4814479986441705e-05, "loss": 0.4411, "step": 11704 }, { "epoch": 3.6, "learning_rate": 1.4813606520397811e-05, "loss": 0.53, "step": 11705 }, { "epoch": 3.6, "learning_rate": 1.481273300655037e-05, "loss": 0.488, "step": 11706 }, { "epoch": 3.6, "learning_rate": 1.4811859444908053e-05, "loss": 0.7187, "step": 11707 }, { "epoch": 3.6, "learning_rate": 1.4810985835479535e-05, "loss": 0.4357, "step": 11708 }, { "epoch": 3.6, "learning_rate": 1.4810112178273497e-05, "loss": 0.4152, "step": 11709 }, { "epoch": 3.6, "learning_rate": 1.480923847329861e-05, "loss": 0.5571, "step": 11710 }, { "epoch": 3.6, "learning_rate": 1.480836472056355e-05, "loss": 0.3626, "step": 11711 }, { "epoch": 3.6, "learning_rate": 1.4807490920077e-05, "loss": 0.487, "step": 11712 }, { "epoch": 3.6, "learning_rate": 1.4806617071847638e-05, "loss": 0.5615, "step": 11713 }, { "epoch": 3.6, "learning_rate": 1.4805743175884134e-05, "loss": 0.5063, "step": 11714 }, { "epoch": 3.6, "learning_rate": 1.4804869232195171e-05, "loss": 0.3225, "step": 11715 }, { "epoch": 3.6, "learning_rate": 1.4803995240789428e-05, "loss": 0.5547, "step": 11716 }, { "epoch": 3.61, "learning_rate": 1.4803121201675588e-05, "loss": 0.3232, "step": 11717 }, { "epoch": 3.61, "learning_rate": 1.4802247114862325e-05, "loss": 0.6311, "step": 11718 }, { "epoch": 3.61, "learning_rate": 1.4801372980358324e-05, "loss": 0.5056, "step": 11719 }, { "epoch": 3.61, "learning_rate": 1.4800498798172263e-05, "loss": 0.588, "step": 11720 }, { "epoch": 3.61, "learning_rate": 1.4799624568312823e-05, "loss": 0.5172, "step": 11721 }, { "epoch": 3.61, "learning_rate": 1.479875029078869e-05, "loss": 0.5862, "step": 11722 }, { "epoch": 3.61, "learning_rate": 1.4797875965608543e-05, "loss": 0.6625, "step": 11723 }, { "epoch": 3.61, "learning_rate": 1.4797001592781068e-05, "loss": 0.4899, "step": 11724 }, { "epoch": 3.61, "learning_rate": 1.4796127172314944e-05, "loss": 0.6225, "step": 11725 }, { "epoch": 3.61, "learning_rate": 1.4795252704218856e-05, "loss": 0.5104, "step": 11726 }, { "epoch": 3.61, "learning_rate": 1.479437818850149e-05, "loss": 0.4987, "step": 11727 }, { "epoch": 3.61, "learning_rate": 1.479350362517153e-05, "loss": 0.6956, "step": 11728 }, { "epoch": 3.61, "learning_rate": 1.4792629014237666e-05, "loss": 0.7219, "step": 11729 }, { "epoch": 3.61, "learning_rate": 1.4791754355708571e-05, "loss": 0.5941, "step": 11730 }, { "epoch": 3.61, "learning_rate": 1.4790879649592944e-05, "loss": 0.6536, "step": 11731 }, { "epoch": 3.61, "learning_rate": 1.4790004895899465e-05, "loss": 0.451, "step": 11732 }, { "epoch": 3.61, "learning_rate": 1.478913009463682e-05, "loss": 0.5485, "step": 11733 }, { "epoch": 3.61, "learning_rate": 1.4788255245813705e-05, "loss": 0.5166, "step": 11734 }, { "epoch": 3.61, "learning_rate": 1.4787380349438801e-05, "loss": 0.5426, "step": 11735 }, { "epoch": 3.61, "learning_rate": 1.4786505405520796e-05, "loss": 0.6194, "step": 11736 }, { "epoch": 3.61, "learning_rate": 1.4785630414068382e-05, "loss": 0.6479, "step": 11737 }, { "epoch": 3.61, "learning_rate": 1.4784755375090244e-05, "loss": 0.6421, "step": 11738 }, { "epoch": 3.61, "learning_rate": 1.4783880288595084e-05, "loss": 0.5701, "step": 11739 }, { "epoch": 3.61, "learning_rate": 1.4783005154591576e-05, "loss": 0.5027, "step": 11740 }, { "epoch": 3.61, "learning_rate": 1.4782129973088422e-05, "loss": 0.9063, "step": 11741 }, { "epoch": 3.61, "learning_rate": 1.4781254744094307e-05, "loss": 0.5133, "step": 11742 }, { "epoch": 3.61, "learning_rate": 1.4780379467617928e-05, "loss": 0.5691, "step": 11743 }, { "epoch": 3.61, "learning_rate": 1.4779504143667978e-05, "loss": 0.7736, "step": 11744 }, { "epoch": 3.61, "learning_rate": 1.4778628772253141e-05, "loss": 0.4625, "step": 11745 }, { "epoch": 3.61, "learning_rate": 1.4777753353382121e-05, "loss": 0.5419, "step": 11746 }, { "epoch": 3.61, "learning_rate": 1.4776877887063604e-05, "loss": 0.3913, "step": 11747 }, { "epoch": 3.61, "learning_rate": 1.4776002373306286e-05, "loss": 0.4884, "step": 11748 }, { "epoch": 3.61, "learning_rate": 1.4775126812118865e-05, "loss": 0.6505, "step": 11749 }, { "epoch": 3.62, "learning_rate": 1.4774251203510034e-05, "loss": 0.4323, "step": 11750 }, { "epoch": 3.62, "learning_rate": 1.4773375547488492e-05, "loss": 0.4386, "step": 11751 }, { "epoch": 3.62, "learning_rate": 1.4772499844062925e-05, "loss": 0.7983, "step": 11752 }, { "epoch": 3.62, "learning_rate": 1.4771624093242039e-05, "loss": 0.5505, "step": 11753 }, { "epoch": 3.62, "learning_rate": 1.4770748295034527e-05, "loss": 0.4873, "step": 11754 }, { "epoch": 3.62, "learning_rate": 1.476987244944909e-05, "loss": 0.4013, "step": 11755 }, { "epoch": 3.62, "learning_rate": 1.4768996556494422e-05, "loss": 0.5108, "step": 11756 }, { "epoch": 3.62, "learning_rate": 1.4768120616179222e-05, "loss": 0.468, "step": 11757 }, { "epoch": 3.62, "learning_rate": 1.4767244628512191e-05, "loss": 0.5337, "step": 11758 }, { "epoch": 3.62, "learning_rate": 1.4766368593502028e-05, "loss": 0.4734, "step": 11759 }, { "epoch": 3.62, "learning_rate": 1.476549251115743e-05, "loss": 0.6152, "step": 11760 }, { "epoch": 3.62, "learning_rate": 1.47646163814871e-05, "loss": 0.4688, "step": 11761 }, { "epoch": 3.62, "learning_rate": 1.4763740204499739e-05, "loss": 0.6991, "step": 11762 }, { "epoch": 3.62, "learning_rate": 1.4762863980204048e-05, "loss": 0.5981, "step": 11763 }, { "epoch": 3.62, "learning_rate": 1.4761987708608727e-05, "loss": 0.4917, "step": 11764 }, { "epoch": 3.62, "learning_rate": 1.476111138972248e-05, "loss": 0.3985, "step": 11765 }, { "epoch": 3.62, "learning_rate": 1.4760235023554006e-05, "loss": 0.4126, "step": 11766 }, { "epoch": 3.62, "learning_rate": 1.4759358610112014e-05, "loss": 0.512, "step": 11767 }, { "epoch": 3.62, "learning_rate": 1.4758482149405204e-05, "loss": 0.5941, "step": 11768 }, { "epoch": 3.62, "learning_rate": 1.475760564144228e-05, "loss": 0.5861, "step": 11769 }, { "epoch": 3.62, "learning_rate": 1.4756729086231948e-05, "loss": 0.4258, "step": 11770 }, { "epoch": 3.62, "learning_rate": 1.4755852483782911e-05, "loss": 0.4425, "step": 11771 }, { "epoch": 3.62, "learning_rate": 1.4754975834103877e-05, "loss": 0.5843, "step": 11772 }, { "epoch": 3.62, "learning_rate": 1.4754099137203553e-05, "loss": 0.4721, "step": 11773 }, { "epoch": 3.62, "learning_rate": 1.4753222393090638e-05, "loss": 0.341, "step": 11774 }, { "epoch": 3.62, "learning_rate": 1.475234560177385e-05, "loss": 0.5302, "step": 11775 }, { "epoch": 3.62, "learning_rate": 1.4751468763261885e-05, "loss": 0.5146, "step": 11776 }, { "epoch": 3.62, "learning_rate": 1.4750591877563456e-05, "loss": 0.5233, "step": 11777 }, { "epoch": 3.62, "learning_rate": 1.4749714944687273e-05, "loss": 0.3887, "step": 11778 }, { "epoch": 3.62, "learning_rate": 1.4748837964642044e-05, "loss": 0.5431, "step": 11779 }, { "epoch": 3.62, "learning_rate": 1.4747960937436476e-05, "loss": 0.4182, "step": 11780 }, { "epoch": 3.62, "learning_rate": 1.4747083863079282e-05, "loss": 0.501, "step": 11781 }, { "epoch": 3.63, "learning_rate": 1.4746206741579169e-05, "loss": 0.5884, "step": 11782 }, { "epoch": 3.63, "learning_rate": 1.474532957294485e-05, "loss": 0.5614, "step": 11783 }, { "epoch": 3.63, "learning_rate": 1.4744452357185033e-05, "loss": 0.415, "step": 11784 }, { "epoch": 3.63, "learning_rate": 1.474357509430843e-05, "loss": 0.5756, "step": 11785 }, { "epoch": 3.63, "learning_rate": 1.4742697784323758e-05, "loss": 0.624, "step": 11786 }, { "epoch": 3.63, "learning_rate": 1.4741820427239725e-05, "loss": 0.3882, "step": 11787 }, { "epoch": 3.63, "learning_rate": 1.4740943023065043e-05, "loss": 0.6623, "step": 11788 }, { "epoch": 3.63, "learning_rate": 1.4740065571808431e-05, "loss": 0.5561, "step": 11789 }, { "epoch": 3.63, "learning_rate": 1.4739188073478597e-05, "loss": 0.4691, "step": 11790 }, { "epoch": 3.63, "learning_rate": 1.4738310528084258e-05, "loss": 0.3788, "step": 11791 }, { "epoch": 3.63, "learning_rate": 1.4737432935634128e-05, "loss": 0.6766, "step": 11792 }, { "epoch": 3.63, "learning_rate": 1.4736555296136922e-05, "loss": 0.2701, "step": 11793 }, { "epoch": 3.63, "learning_rate": 1.473567760960136e-05, "loss": 0.5147, "step": 11794 }, { "epoch": 3.63, "learning_rate": 1.4734799876036152e-05, "loss": 0.627, "step": 11795 }, { "epoch": 3.63, "learning_rate": 1.473392209545002e-05, "loss": 0.5448, "step": 11796 }, { "epoch": 3.63, "learning_rate": 1.4733044267851676e-05, "loss": 0.6131, "step": 11797 }, { "epoch": 3.63, "learning_rate": 1.473216639324984e-05, "loss": 0.582, "step": 11798 }, { "epoch": 3.63, "learning_rate": 1.4731288471653234e-05, "loss": 0.5734, "step": 11799 }, { "epoch": 3.63, "learning_rate": 1.4730410503070572e-05, "loss": 0.3498, "step": 11800 }, { "epoch": 3.63, "learning_rate": 1.4729532487510571e-05, "loss": 0.5647, "step": 11801 }, { "epoch": 3.63, "learning_rate": 1.4728654424981953e-05, "loss": 0.4873, "step": 11802 }, { "epoch": 3.63, "learning_rate": 1.4727776315493443e-05, "loss": 0.5277, "step": 11803 }, { "epoch": 3.63, "learning_rate": 1.4726898159053754e-05, "loss": 0.489, "step": 11804 }, { "epoch": 3.63, "learning_rate": 1.472601995567161e-05, "loss": 0.3859, "step": 11805 }, { "epoch": 3.63, "learning_rate": 1.4725141705355734e-05, "loss": 0.4349, "step": 11806 }, { "epoch": 3.63, "learning_rate": 1.4724263408114842e-05, "loss": 0.4935, "step": 11807 }, { "epoch": 3.63, "learning_rate": 1.4723385063957665e-05, "loss": 0.6404, "step": 11808 }, { "epoch": 3.63, "learning_rate": 1.4722506672892917e-05, "loss": 0.444, "step": 11809 }, { "epoch": 3.63, "learning_rate": 1.4721628234929331e-05, "loss": 0.4996, "step": 11810 }, { "epoch": 3.63, "learning_rate": 1.472074975007562e-05, "loss": 0.4199, "step": 11811 }, { "epoch": 3.63, "learning_rate": 1.4719871218340516e-05, "loss": 0.4386, "step": 11812 }, { "epoch": 3.63, "learning_rate": 1.4718992639732739e-05, "loss": 0.598, "step": 11813 }, { "epoch": 3.63, "learning_rate": 1.4718114014261017e-05, "loss": 0.4986, "step": 11814 }, { "epoch": 3.64, "learning_rate": 1.4717235341934075e-05, "loss": 0.4, "step": 11815 }, { "epoch": 3.64, "learning_rate": 1.4716356622760635e-05, "loss": 0.4373, "step": 11816 }, { "epoch": 3.64, "learning_rate": 1.4715477856749432e-05, "loss": 0.4907, "step": 11817 }, { "epoch": 3.64, "learning_rate": 1.4714599043909185e-05, "loss": 0.5346, "step": 11818 }, { "epoch": 3.64, "learning_rate": 1.4713720184248623e-05, "loss": 0.6739, "step": 11819 }, { "epoch": 3.64, "learning_rate": 1.4712841277776479e-05, "loss": 0.4992, "step": 11820 }, { "epoch": 3.64, "learning_rate": 1.4711962324501477e-05, "loss": 0.3794, "step": 11821 }, { "epoch": 3.64, "learning_rate": 1.4711083324432344e-05, "loss": 0.4372, "step": 11822 }, { "epoch": 3.64, "learning_rate": 1.4710204277577812e-05, "loss": 0.5769, "step": 11823 }, { "epoch": 3.64, "learning_rate": 1.4709325183946613e-05, "loss": 0.7016, "step": 11824 }, { "epoch": 3.64, "learning_rate": 1.4708446043547474e-05, "loss": 0.561, "step": 11825 }, { "epoch": 3.64, "learning_rate": 1.4707566856389124e-05, "loss": 0.3303, "step": 11826 }, { "epoch": 3.64, "learning_rate": 1.47066876224803e-05, "loss": 0.5486, "step": 11827 }, { "epoch": 3.64, "learning_rate": 1.4705808341829727e-05, "loss": 0.5297, "step": 11828 }, { "epoch": 3.64, "learning_rate": 1.4704929014446142e-05, "loss": 0.5564, "step": 11829 }, { "epoch": 3.64, "learning_rate": 1.4704049640338274e-05, "loss": 0.6211, "step": 11830 }, { "epoch": 3.64, "learning_rate": 1.470317021951486e-05, "loss": 0.4893, "step": 11831 }, { "epoch": 3.64, "learning_rate": 1.470229075198463e-05, "loss": 0.4819, "step": 11832 }, { "epoch": 3.64, "learning_rate": 1.4701411237756315e-05, "loss": 0.5976, "step": 11833 }, { "epoch": 3.64, "learning_rate": 1.4700531676838659e-05, "loss": 0.3319, "step": 11834 }, { "epoch": 3.64, "learning_rate": 1.4699652069240391e-05, "loss": 0.4523, "step": 11835 }, { "epoch": 3.64, "learning_rate": 1.4698772414970246e-05, "loss": 0.3465, "step": 11836 }, { "epoch": 3.64, "learning_rate": 1.4697892714036959e-05, "loss": 0.4414, "step": 11837 }, { "epoch": 3.64, "learning_rate": 1.4697012966449266e-05, "loss": 0.5146, "step": 11838 }, { "epoch": 3.64, "learning_rate": 1.469613317221591e-05, "loss": 0.5361, "step": 11839 }, { "epoch": 3.64, "learning_rate": 1.4695253331345622e-05, "loss": 0.6132, "step": 11840 }, { "epoch": 3.64, "learning_rate": 1.469437344384714e-05, "loss": 0.3884, "step": 11841 }, { "epoch": 3.64, "learning_rate": 1.4693493509729205e-05, "loss": 0.3877, "step": 11842 }, { "epoch": 3.64, "learning_rate": 1.469261352900055e-05, "loss": 0.3796, "step": 11843 }, { "epoch": 3.64, "learning_rate": 1.469173350166992e-05, "loss": 0.3302, "step": 11844 }, { "epoch": 3.64, "learning_rate": 1.4690853427746055e-05, "loss": 0.4913, "step": 11845 }, { "epoch": 3.64, "learning_rate": 1.4689973307237687e-05, "loss": 0.6316, "step": 11846 }, { "epoch": 3.65, "learning_rate": 1.4689093140153565e-05, "loss": 0.4622, "step": 11847 }, { "epoch": 3.65, "learning_rate": 1.4688212926502426e-05, "loss": 0.6773, "step": 11848 }, { "epoch": 3.65, "learning_rate": 1.4687332666293013e-05, "loss": 0.3384, "step": 11849 }, { "epoch": 3.65, "learning_rate": 1.4686452359534067e-05, "loss": 0.4045, "step": 11850 }, { "epoch": 3.65, "learning_rate": 1.4685572006234327e-05, "loss": 0.5534, "step": 11851 }, { "epoch": 3.65, "learning_rate": 1.468469160640254e-05, "loss": 0.5584, "step": 11852 }, { "epoch": 3.65, "learning_rate": 1.4683811160047448e-05, "loss": 0.4847, "step": 11853 }, { "epoch": 3.65, "learning_rate": 1.4682930667177795e-05, "loss": 0.5153, "step": 11854 }, { "epoch": 3.65, "learning_rate": 1.4682050127802323e-05, "loss": 0.4004, "step": 11855 }, { "epoch": 3.65, "learning_rate": 1.4681169541929778e-05, "loss": 0.3838, "step": 11856 }, { "epoch": 3.65, "learning_rate": 1.4680288909568905e-05, "loss": 0.568, "step": 11857 }, { "epoch": 3.65, "learning_rate": 1.4679408230728452e-05, "loss": 0.6073, "step": 11858 }, { "epoch": 3.65, "learning_rate": 1.4678527505417159e-05, "loss": 0.5172, "step": 11859 }, { "epoch": 3.65, "learning_rate": 1.4677646733643783e-05, "loss": 0.4162, "step": 11860 }, { "epoch": 3.65, "learning_rate": 1.4676765915417057e-05, "loss": 0.4879, "step": 11861 }, { "epoch": 3.65, "learning_rate": 1.4675885050745737e-05, "loss": 0.4417, "step": 11862 }, { "epoch": 3.65, "learning_rate": 1.467500413963857e-05, "loss": 0.6222, "step": 11863 }, { "epoch": 3.65, "learning_rate": 1.4674123182104301e-05, "loss": 0.479, "step": 11864 }, { "epoch": 3.65, "learning_rate": 1.4673242178151685e-05, "loss": 0.5399, "step": 11865 }, { "epoch": 3.65, "learning_rate": 1.4672361127789466e-05, "loss": 0.7207, "step": 11866 }, { "epoch": 3.65, "learning_rate": 1.4671480031026392e-05, "loss": 0.6236, "step": 11867 }, { "epoch": 3.65, "learning_rate": 1.4670598887871219e-05, "loss": 0.5289, "step": 11868 }, { "epoch": 3.65, "learning_rate": 1.4669717698332692e-05, "loss": 0.5583, "step": 11869 }, { "epoch": 3.65, "learning_rate": 1.4668836462419566e-05, "loss": 0.4827, "step": 11870 }, { "epoch": 3.65, "learning_rate": 1.4667955180140592e-05, "loss": 0.4422, "step": 11871 }, { "epoch": 3.65, "learning_rate": 1.4667073851504519e-05, "loss": 0.451, "step": 11872 }, { "epoch": 3.65, "learning_rate": 1.4666192476520102e-05, "loss": 0.3366, "step": 11873 }, { "epoch": 3.65, "learning_rate": 1.4665311055196092e-05, "loss": 0.4465, "step": 11874 }, { "epoch": 3.65, "learning_rate": 1.4664429587541245e-05, "loss": 0.5378, "step": 11875 }, { "epoch": 3.65, "learning_rate": 1.4663548073564316e-05, "loss": 0.4545, "step": 11876 }, { "epoch": 3.65, "learning_rate": 1.4662666513274053e-05, "loss": 0.5549, "step": 11877 }, { "epoch": 3.65, "learning_rate": 1.4661784906679212e-05, "loss": 0.4086, "step": 11878 }, { "epoch": 3.65, "learning_rate": 1.4660903253788554e-05, "loss": 0.4597, "step": 11879 }, { "epoch": 3.66, "learning_rate": 1.4660021554610832e-05, "loss": 0.6748, "step": 11880 }, { "epoch": 3.66, "learning_rate": 1.4659139809154801e-05, "loss": 0.6684, "step": 11881 }, { "epoch": 3.66, "learning_rate": 1.4658258017429217e-05, "loss": 0.5484, "step": 11882 }, { "epoch": 3.66, "learning_rate": 1.4657376179442834e-05, "loss": 0.445, "step": 11883 }, { "epoch": 3.66, "learning_rate": 1.4656494295204417e-05, "loss": 0.5565, "step": 11884 }, { "epoch": 3.66, "learning_rate": 1.465561236472272e-05, "loss": 0.4527, "step": 11885 }, { "epoch": 3.66, "learning_rate": 1.4654730388006502e-05, "loss": 0.5854, "step": 11886 }, { "epoch": 3.66, "learning_rate": 1.465384836506452e-05, "loss": 0.5497, "step": 11887 }, { "epoch": 3.66, "learning_rate": 1.4652966295905533e-05, "loss": 0.4332, "step": 11888 }, { "epoch": 3.66, "learning_rate": 1.4652084180538304e-05, "loss": 0.4893, "step": 11889 }, { "epoch": 3.66, "learning_rate": 1.4651202018971592e-05, "loss": 0.863, "step": 11890 }, { "epoch": 3.66, "learning_rate": 1.465031981121416e-05, "loss": 0.4401, "step": 11891 }, { "epoch": 3.66, "learning_rate": 1.4649437557274762e-05, "loss": 0.4605, "step": 11892 }, { "epoch": 3.66, "learning_rate": 1.4648555257162165e-05, "loss": 0.5303, "step": 11893 }, { "epoch": 3.66, "learning_rate": 1.464767291088513e-05, "loss": 0.448, "step": 11894 }, { "epoch": 3.66, "learning_rate": 1.4646790518452419e-05, "loss": 0.5513, "step": 11895 }, { "epoch": 3.66, "learning_rate": 1.46459080798728e-05, "loss": 0.3833, "step": 11896 }, { "epoch": 3.66, "learning_rate": 1.4645025595155027e-05, "loss": 0.484, "step": 11897 }, { "epoch": 3.66, "learning_rate": 1.464414306430787e-05, "loss": 0.6635, "step": 11898 }, { "epoch": 3.66, "learning_rate": 1.4643260487340093e-05, "loss": 0.5307, "step": 11899 }, { "epoch": 3.66, "learning_rate": 1.4642377864260457e-05, "loss": 0.4866, "step": 11900 }, { "epoch": 3.66, "learning_rate": 1.4641495195077737e-05, "loss": 0.5401, "step": 11901 }, { "epoch": 3.66, "learning_rate": 1.4640612479800686e-05, "loss": 0.3662, "step": 11902 }, { "epoch": 3.66, "learning_rate": 1.4639729718438078e-05, "loss": 0.4982, "step": 11903 }, { "epoch": 3.66, "learning_rate": 1.4638846910998679e-05, "loss": 0.4829, "step": 11904 }, { "epoch": 3.66, "learning_rate": 1.4637964057491251e-05, "loss": 0.5988, "step": 11905 }, { "epoch": 3.66, "learning_rate": 1.463708115792457e-05, "loss": 0.4495, "step": 11906 }, { "epoch": 3.66, "learning_rate": 1.4636198212307397e-05, "loss": 0.5361, "step": 11907 }, { "epoch": 3.66, "learning_rate": 1.4635315220648505e-05, "loss": 0.4048, "step": 11908 }, { "epoch": 3.66, "learning_rate": 1.463443218295666e-05, "loss": 0.5856, "step": 11909 }, { "epoch": 3.66, "learning_rate": 1.4633549099240629e-05, "loss": 0.6915, "step": 11910 }, { "epoch": 3.66, "learning_rate": 1.463266596950919e-05, "loss": 0.6025, "step": 11911 }, { "epoch": 3.67, "learning_rate": 1.4631782793771106e-05, "loss": 0.4906, "step": 11912 }, { "epoch": 3.67, "learning_rate": 1.4630899572035149e-05, "loss": 0.4839, "step": 11913 }, { "epoch": 3.67, "learning_rate": 1.463001630431009e-05, "loss": 0.5594, "step": 11914 }, { "epoch": 3.67, "learning_rate": 1.4629132990604706e-05, "loss": 0.4466, "step": 11915 }, { "epoch": 3.67, "learning_rate": 1.4628249630927765e-05, "loss": 0.6081, "step": 11916 }, { "epoch": 3.67, "learning_rate": 1.4627366225288038e-05, "loss": 0.428, "step": 11917 }, { "epoch": 3.67, "learning_rate": 1.4626482773694299e-05, "loss": 0.5759, "step": 11918 }, { "epoch": 3.67, "learning_rate": 1.4625599276155323e-05, "loss": 0.5832, "step": 11919 }, { "epoch": 3.67, "learning_rate": 1.4624715732679881e-05, "loss": 0.4697, "step": 11920 }, { "epoch": 3.67, "learning_rate": 1.4623832143276753e-05, "loss": 0.6247, "step": 11921 }, { "epoch": 3.67, "learning_rate": 1.4622948507954711e-05, "loss": 0.6354, "step": 11922 }, { "epoch": 3.67, "learning_rate": 1.4622064826722527e-05, "loss": 0.476, "step": 11923 }, { "epoch": 3.67, "learning_rate": 1.4621181099588978e-05, "loss": 0.3233, "step": 11924 }, { "epoch": 3.67, "learning_rate": 1.4620297326562844e-05, "loss": 0.3038, "step": 11925 }, { "epoch": 3.67, "learning_rate": 1.46194135076529e-05, "loss": 0.4396, "step": 11926 }, { "epoch": 3.67, "learning_rate": 1.4618529642867922e-05, "loss": 0.6031, "step": 11927 }, { "epoch": 3.67, "learning_rate": 1.4617645732216686e-05, "loss": 0.5473, "step": 11928 }, { "epoch": 3.67, "learning_rate": 1.4616761775707974e-05, "loss": 0.598, "step": 11929 }, { "epoch": 3.67, "learning_rate": 1.4615877773350561e-05, "loss": 0.5921, "step": 11930 }, { "epoch": 3.67, "learning_rate": 1.461499372515323e-05, "loss": 0.6349, "step": 11931 }, { "epoch": 3.67, "learning_rate": 1.461410963112476e-05, "loss": 0.42, "step": 11932 }, { "epoch": 3.67, "learning_rate": 1.4613225491273923e-05, "loss": 0.4339, "step": 11933 }, { "epoch": 3.67, "learning_rate": 1.4612341305609507e-05, "loss": 0.3981, "step": 11934 }, { "epoch": 3.67, "learning_rate": 1.4611457074140294e-05, "loss": 0.4757, "step": 11935 }, { "epoch": 3.67, "learning_rate": 1.461057279687506e-05, "loss": 0.5277, "step": 11936 }, { "epoch": 3.67, "learning_rate": 1.460968847382259e-05, "loss": 0.6027, "step": 11937 }, { "epoch": 3.67, "learning_rate": 1.4608804104991662e-05, "loss": 0.277, "step": 11938 }, { "epoch": 3.67, "learning_rate": 1.4607919690391063e-05, "loss": 0.4403, "step": 11939 }, { "epoch": 3.67, "learning_rate": 1.4607035230029576e-05, "loss": 0.4957, "step": 11940 }, { "epoch": 3.67, "learning_rate": 1.4606150723915984e-05, "loss": 0.3998, "step": 11941 }, { "epoch": 3.67, "learning_rate": 1.4605266172059069e-05, "loss": 0.4731, "step": 11942 }, { "epoch": 3.67, "learning_rate": 1.4604381574467616e-05, "loss": 0.6144, "step": 11943 }, { "epoch": 3.67, "learning_rate": 1.4603496931150411e-05, "loss": 0.4813, "step": 11944 }, { "epoch": 3.68, "learning_rate": 1.4602612242116236e-05, "loss": 0.597, "step": 11945 }, { "epoch": 3.68, "learning_rate": 1.4601727507373886e-05, "loss": 0.4465, "step": 11946 }, { "epoch": 3.68, "learning_rate": 1.4600842726932136e-05, "loss": 0.4931, "step": 11947 }, { "epoch": 3.68, "learning_rate": 1.4599957900799778e-05, "loss": 0.3993, "step": 11948 }, { "epoch": 3.68, "learning_rate": 1.45990730289856e-05, "loss": 0.5547, "step": 11949 }, { "epoch": 3.68, "learning_rate": 1.4598188111498387e-05, "loss": 0.5546, "step": 11950 }, { "epoch": 3.68, "learning_rate": 1.4597303148346928e-05, "loss": 0.595, "step": 11951 }, { "epoch": 3.68, "learning_rate": 1.4596418139540012e-05, "loss": 0.5505, "step": 11952 }, { "epoch": 3.68, "learning_rate": 1.4595533085086428e-05, "loss": 0.4938, "step": 11953 }, { "epoch": 3.68, "learning_rate": 1.4594647984994966e-05, "loss": 0.5047, "step": 11954 }, { "epoch": 3.68, "learning_rate": 1.4593762839274412e-05, "loss": 0.4688, "step": 11955 }, { "epoch": 3.68, "learning_rate": 1.4592877647933563e-05, "loss": 0.4068, "step": 11956 }, { "epoch": 3.68, "learning_rate": 1.4591992410981201e-05, "loss": 0.705, "step": 11957 }, { "epoch": 3.68, "learning_rate": 1.4591107128426128e-05, "loss": 0.4835, "step": 11958 }, { "epoch": 3.68, "learning_rate": 1.4590221800277124e-05, "loss": 0.5054, "step": 11959 }, { "epoch": 3.68, "learning_rate": 1.4589336426542993e-05, "loss": 0.4205, "step": 11960 }, { "epoch": 3.68, "learning_rate": 1.458845100723252e-05, "loss": 0.5911, "step": 11961 }, { "epoch": 3.68, "learning_rate": 1.4587565542354498e-05, "loss": 0.4532, "step": 11962 }, { "epoch": 3.68, "learning_rate": 1.4586680031917723e-05, "loss": 0.669, "step": 11963 }, { "epoch": 3.68, "learning_rate": 1.4585794475930986e-05, "loss": 0.4676, "step": 11964 }, { "epoch": 3.68, "learning_rate": 1.4584908874403087e-05, "loss": 0.5709, "step": 11965 }, { "epoch": 3.68, "learning_rate": 1.4584023227342816e-05, "loss": 0.5836, "step": 11966 }, { "epoch": 3.68, "learning_rate": 1.4583137534758968e-05, "loss": 0.4486, "step": 11967 }, { "epoch": 3.68, "learning_rate": 1.4582251796660342e-05, "loss": 0.3893, "step": 11968 }, { "epoch": 3.68, "learning_rate": 1.4581366013055732e-05, "loss": 0.3559, "step": 11969 }, { "epoch": 3.68, "learning_rate": 1.4580480183953935e-05, "loss": 0.7981, "step": 11970 }, { "epoch": 3.68, "learning_rate": 1.457959430936375e-05, "loss": 0.5323, "step": 11971 }, { "epoch": 3.68, "learning_rate": 1.4578708389293973e-05, "loss": 0.4781, "step": 11972 }, { "epoch": 3.68, "learning_rate": 1.4577822423753399e-05, "loss": 0.7256, "step": 11973 }, { "epoch": 3.68, "learning_rate": 1.4576936412750828e-05, "loss": 0.4933, "step": 11974 }, { "epoch": 3.68, "learning_rate": 1.4576050356295066e-05, "loss": 0.5259, "step": 11975 }, { "epoch": 3.68, "learning_rate": 1.45751642543949e-05, "loss": 0.4903, "step": 11976 }, { "epoch": 3.69, "learning_rate": 1.4574278107059144e-05, "loss": 0.6444, "step": 11977 }, { "epoch": 3.69, "learning_rate": 1.4573391914296585e-05, "loss": 0.4908, "step": 11978 }, { "epoch": 3.69, "learning_rate": 1.4572505676116029e-05, "loss": 0.5629, "step": 11979 }, { "epoch": 3.69, "learning_rate": 1.4571619392526279e-05, "loss": 0.386, "step": 11980 }, { "epoch": 3.69, "learning_rate": 1.4570733063536136e-05, "loss": 0.8462, "step": 11981 }, { "epoch": 3.69, "learning_rate": 1.4569846689154401e-05, "loss": 0.6159, "step": 11982 }, { "epoch": 3.69, "learning_rate": 1.4568960269389875e-05, "loss": 0.5215, "step": 11983 }, { "epoch": 3.69, "learning_rate": 1.4568073804251364e-05, "loss": 0.5599, "step": 11984 }, { "epoch": 3.69, "learning_rate": 1.4567187293747671e-05, "loss": 0.4076, "step": 11985 }, { "epoch": 3.69, "learning_rate": 1.4566300737887595e-05, "loss": 0.5003, "step": 11986 }, { "epoch": 3.69, "learning_rate": 1.456541413667995e-05, "loss": 0.6935, "step": 11987 }, { "epoch": 3.69, "learning_rate": 1.4564527490133531e-05, "loss": 0.4216, "step": 11988 }, { "epoch": 3.69, "learning_rate": 1.4563640798257148e-05, "loss": 0.5658, "step": 11989 }, { "epoch": 3.69, "learning_rate": 1.4562754061059608e-05, "loss": 0.4839, "step": 11990 }, { "epoch": 3.69, "learning_rate": 1.456186727854971e-05, "loss": 0.4046, "step": 11991 }, { "epoch": 3.69, "learning_rate": 1.456098045073627e-05, "loss": 0.5051, "step": 11992 }, { "epoch": 3.69, "learning_rate": 1.456009357762809e-05, "loss": 0.5917, "step": 11993 }, { "epoch": 3.69, "learning_rate": 1.4559206659233977e-05, "loss": 0.3604, "step": 11994 }, { "epoch": 3.69, "learning_rate": 1.4558319695562739e-05, "loss": 0.4823, "step": 11995 }, { "epoch": 3.69, "learning_rate": 1.4557432686623187e-05, "loss": 0.5082, "step": 11996 }, { "epoch": 3.69, "learning_rate": 1.4556545632424128e-05, "loss": 0.5091, "step": 11997 }, { "epoch": 3.69, "learning_rate": 1.4555658532974373e-05, "loss": 0.4041, "step": 11998 }, { "epoch": 3.69, "learning_rate": 1.4554771388282728e-05, "loss": 0.5685, "step": 11999 }, { "epoch": 3.69, "learning_rate": 1.4553884198358004e-05, "loss": 0.3838, "step": 12000 }, { "epoch": 3.69, "learning_rate": 1.4552996963209013e-05, "loss": 0.4308, "step": 12001 }, { "epoch": 3.69, "learning_rate": 1.4552109682844569e-05, "loss": 0.4855, "step": 12002 }, { "epoch": 3.69, "learning_rate": 1.455122235727348e-05, "loss": 0.4668, "step": 12003 }, { "epoch": 3.69, "learning_rate": 1.4550334986504559e-05, "loss": 0.4445, "step": 12004 }, { "epoch": 3.69, "learning_rate": 1.4549447570546616e-05, "loss": 0.5822, "step": 12005 }, { "epoch": 3.69, "learning_rate": 1.4548560109408465e-05, "loss": 0.5237, "step": 12006 }, { "epoch": 3.69, "learning_rate": 1.4547672603098922e-05, "loss": 0.3939, "step": 12007 }, { "epoch": 3.69, "learning_rate": 1.4546785051626797e-05, "loss": 0.4885, "step": 12008 }, { "epoch": 3.69, "learning_rate": 1.4545897455000906e-05, "loss": 0.5084, "step": 12009 }, { "epoch": 3.7, "learning_rate": 1.4545009813230066e-05, "loss": 0.5778, "step": 12010 }, { "epoch": 3.7, "learning_rate": 1.4544122126323087e-05, "loss": 0.3944, "step": 12011 }, { "epoch": 3.7, "learning_rate": 1.4543234394288792e-05, "loss": 0.5704, "step": 12012 }, { "epoch": 3.7, "learning_rate": 1.4542346617135987e-05, "loss": 0.57, "step": 12013 }, { "epoch": 3.7, "learning_rate": 1.4541458794873495e-05, "loss": 0.3269, "step": 12014 }, { "epoch": 3.7, "learning_rate": 1.4540570927510134e-05, "loss": 0.5164, "step": 12015 }, { "epoch": 3.7, "learning_rate": 1.4539683015054717e-05, "loss": 0.5896, "step": 12016 }, { "epoch": 3.7, "learning_rate": 1.4538795057516066e-05, "loss": 0.3854, "step": 12017 }, { "epoch": 3.7, "learning_rate": 1.4537907054902996e-05, "loss": 0.702, "step": 12018 }, { "epoch": 3.7, "learning_rate": 1.4537019007224324e-05, "loss": 0.4081, "step": 12019 }, { "epoch": 3.7, "learning_rate": 1.4536130914488872e-05, "loss": 0.478, "step": 12020 }, { "epoch": 3.7, "learning_rate": 1.4535242776705463e-05, "loss": 0.5092, "step": 12021 }, { "epoch": 3.7, "learning_rate": 1.453435459388291e-05, "loss": 0.4914, "step": 12022 }, { "epoch": 3.7, "learning_rate": 1.4533466366030041e-05, "loss": 0.6901, "step": 12023 }, { "epoch": 3.7, "learning_rate": 1.4532578093155669e-05, "loss": 0.4749, "step": 12024 }, { "epoch": 3.7, "learning_rate": 1.453168977526862e-05, "loss": 0.5454, "step": 12025 }, { "epoch": 3.7, "learning_rate": 1.4530801412377713e-05, "loss": 0.427, "step": 12026 }, { "epoch": 3.7, "learning_rate": 1.4529913004491779e-05, "loss": 0.5408, "step": 12027 }, { "epoch": 3.7, "learning_rate": 1.4529024551619627e-05, "loss": 0.4355, "step": 12028 }, { "epoch": 3.7, "learning_rate": 1.452813605377009e-05, "loss": 0.5026, "step": 12029 }, { "epoch": 3.7, "learning_rate": 1.4527247510951989e-05, "loss": 0.4575, "step": 12030 }, { "epoch": 3.7, "learning_rate": 1.4526358923174147e-05, "loss": 0.6694, "step": 12031 }, { "epoch": 3.7, "learning_rate": 1.4525470290445392e-05, "loss": 0.3792, "step": 12032 }, { "epoch": 3.7, "learning_rate": 1.4524581612774542e-05, "loss": 0.4212, "step": 12033 }, { "epoch": 3.7, "learning_rate": 1.4523692890170429e-05, "loss": 0.4684, "step": 12034 }, { "epoch": 3.7, "learning_rate": 1.4522804122641875e-05, "loss": 0.6485, "step": 12035 }, { "epoch": 3.7, "learning_rate": 1.4521915310197707e-05, "loss": 0.6796, "step": 12036 }, { "epoch": 3.7, "learning_rate": 1.4521026452846756e-05, "loss": 0.4688, "step": 12037 }, { "epoch": 3.7, "learning_rate": 1.4520137550597844e-05, "loss": 0.3906, "step": 12038 }, { "epoch": 3.7, "learning_rate": 1.4519248603459798e-05, "loss": 0.394, "step": 12039 }, { "epoch": 3.7, "learning_rate": 1.4518359611441452e-05, "loss": 0.5304, "step": 12040 }, { "epoch": 3.7, "learning_rate": 1.4517470574551626e-05, "loss": 0.5386, "step": 12041 }, { "epoch": 3.71, "learning_rate": 1.451658149279916e-05, "loss": 0.4574, "step": 12042 }, { "epoch": 3.71, "learning_rate": 1.4515692366192872e-05, "loss": 0.3025, "step": 12043 }, { "epoch": 3.71, "learning_rate": 1.4514803194741601e-05, "loss": 0.5919, "step": 12044 }, { "epoch": 3.71, "learning_rate": 1.4513913978454169e-05, "loss": 0.2559, "step": 12045 }, { "epoch": 3.71, "learning_rate": 1.4513024717339415e-05, "loss": 0.7673, "step": 12046 }, { "epoch": 3.71, "learning_rate": 1.4512135411406166e-05, "loss": 0.7302, "step": 12047 }, { "epoch": 3.71, "learning_rate": 1.4511246060663251e-05, "loss": 0.5282, "step": 12048 }, { "epoch": 3.71, "learning_rate": 1.4510356665119507e-05, "loss": 0.4433, "step": 12049 }, { "epoch": 3.71, "learning_rate": 1.4509467224783762e-05, "loss": 0.4459, "step": 12050 }, { "epoch": 3.71, "learning_rate": 1.4508577739664852e-05, "loss": 0.4842, "step": 12051 }, { "epoch": 3.71, "learning_rate": 1.4507688209771613e-05, "loss": 0.7773, "step": 12052 }, { "epoch": 3.71, "learning_rate": 1.4506798635112876e-05, "loss": 0.4844, "step": 12053 }, { "epoch": 3.71, "learning_rate": 1.450590901569747e-05, "loss": 0.4928, "step": 12054 }, { "epoch": 3.71, "learning_rate": 1.4505019351534238e-05, "loss": 0.5802, "step": 12055 }, { "epoch": 3.71, "learning_rate": 1.450412964263201e-05, "loss": 0.5188, "step": 12056 }, { "epoch": 3.71, "learning_rate": 1.4503239888999625e-05, "loss": 0.4431, "step": 12057 }, { "epoch": 3.71, "learning_rate": 1.4502350090645919e-05, "loss": 0.4943, "step": 12058 }, { "epoch": 3.71, "learning_rate": 1.4501460247579723e-05, "loss": 0.5197, "step": 12059 }, { "epoch": 3.71, "learning_rate": 1.4500570359809882e-05, "loss": 0.7534, "step": 12060 }, { "epoch": 3.71, "learning_rate": 1.4499680427345228e-05, "loss": 0.4776, "step": 12061 }, { "epoch": 3.71, "learning_rate": 1.44987904501946e-05, "loss": 0.2823, "step": 12062 }, { "epoch": 3.71, "learning_rate": 1.4497900428366837e-05, "loss": 0.4945, "step": 12063 }, { "epoch": 3.71, "learning_rate": 1.4497010361870775e-05, "loss": 0.5145, "step": 12064 }, { "epoch": 3.71, "learning_rate": 1.449612025071526e-05, "loss": 0.4214, "step": 12065 }, { "epoch": 3.71, "learning_rate": 1.4495230094909125e-05, "loss": 0.4796, "step": 12066 }, { "epoch": 3.71, "learning_rate": 1.4494339894461212e-05, "loss": 0.4986, "step": 12067 }, { "epoch": 3.71, "learning_rate": 1.4493449649380365e-05, "loss": 0.4585, "step": 12068 }, { "epoch": 3.71, "learning_rate": 1.4492559359675416e-05, "loss": 0.5078, "step": 12069 }, { "epoch": 3.71, "learning_rate": 1.4491669025355215e-05, "loss": 0.5432, "step": 12070 }, { "epoch": 3.71, "learning_rate": 1.4490778646428601e-05, "loss": 0.4133, "step": 12071 }, { "epoch": 3.71, "learning_rate": 1.4489888222904415e-05, "loss": 0.484, "step": 12072 }, { "epoch": 3.71, "learning_rate": 1.4488997754791505e-05, "loss": 0.4673, "step": 12073 }, { "epoch": 3.71, "learning_rate": 1.4488107242098706e-05, "loss": 0.6132, "step": 12074 }, { "epoch": 3.72, "learning_rate": 1.4487216684834867e-05, "loss": 0.548, "step": 12075 }, { "epoch": 3.72, "learning_rate": 1.4486326083008827e-05, "loss": 0.5468, "step": 12076 }, { "epoch": 3.72, "learning_rate": 1.448543543662944e-05, "loss": 0.7891, "step": 12077 }, { "epoch": 3.72, "learning_rate": 1.4484544745705542e-05, "loss": 0.4289, "step": 12078 }, { "epoch": 3.72, "learning_rate": 1.4483654010245984e-05, "loss": 0.2609, "step": 12079 }, { "epoch": 3.72, "learning_rate": 1.4482763230259608e-05, "loss": 0.4566, "step": 12080 }, { "epoch": 3.72, "learning_rate": 1.4481872405755261e-05, "loss": 0.4776, "step": 12081 }, { "epoch": 3.72, "learning_rate": 1.4480981536741792e-05, "loss": 0.4383, "step": 12082 }, { "epoch": 3.72, "learning_rate": 1.4480090623228045e-05, "loss": 0.5125, "step": 12083 }, { "epoch": 3.72, "learning_rate": 1.4479199665222869e-05, "loss": 0.7849, "step": 12084 }, { "epoch": 3.72, "learning_rate": 1.4478308662735114e-05, "loss": 0.5537, "step": 12085 }, { "epoch": 3.72, "learning_rate": 1.4477417615773626e-05, "loss": 0.7077, "step": 12086 }, { "epoch": 3.72, "learning_rate": 1.4476526524347254e-05, "loss": 0.6103, "step": 12087 }, { "epoch": 3.72, "learning_rate": 1.4475635388464847e-05, "loss": 0.351, "step": 12088 }, { "epoch": 3.72, "learning_rate": 1.447474420813526e-05, "loss": 0.57, "step": 12089 }, { "epoch": 3.72, "learning_rate": 1.4473852983367334e-05, "loss": 0.4258, "step": 12090 }, { "epoch": 3.72, "learning_rate": 1.4472961714169924e-05, "loss": 0.412, "step": 12091 }, { "epoch": 3.72, "learning_rate": 1.4472070400551887e-05, "loss": 0.4894, "step": 12092 }, { "epoch": 3.72, "learning_rate": 1.4471179042522066e-05, "loss": 0.4539, "step": 12093 }, { "epoch": 3.72, "learning_rate": 1.4470287640089317e-05, "loss": 0.7609, "step": 12094 }, { "epoch": 3.72, "learning_rate": 1.4469396193262491e-05, "loss": 0.4466, "step": 12095 }, { "epoch": 3.72, "learning_rate": 1.4468504702050442e-05, "loss": 0.4546, "step": 12096 }, { "epoch": 3.72, "learning_rate": 1.4467613166462024e-05, "loss": 0.7575, "step": 12097 }, { "epoch": 3.72, "learning_rate": 1.446672158650609e-05, "loss": 0.745, "step": 12098 }, { "epoch": 3.72, "learning_rate": 1.4465829962191494e-05, "loss": 0.5373, "step": 12099 }, { "epoch": 3.72, "learning_rate": 1.4464938293527086e-05, "loss": 0.38, "step": 12100 }, { "epoch": 3.72, "learning_rate": 1.4464046580521732e-05, "loss": 0.6322, "step": 12101 }, { "epoch": 3.72, "learning_rate": 1.446315482318428e-05, "loss": 0.4431, "step": 12102 }, { "epoch": 3.72, "learning_rate": 1.4462263021523586e-05, "loss": 0.3966, "step": 12103 }, { "epoch": 3.72, "learning_rate": 1.446137117554851e-05, "loss": 0.3463, "step": 12104 }, { "epoch": 3.72, "learning_rate": 1.44604792852679e-05, "loss": 0.7103, "step": 12105 }, { "epoch": 3.72, "learning_rate": 1.4459587350690625e-05, "loss": 0.5081, "step": 12106 }, { "epoch": 3.73, "learning_rate": 1.4458695371825535e-05, "loss": 0.5556, "step": 12107 }, { "epoch": 3.73, "learning_rate": 1.4457803348681494e-05, "loss": 0.6, "step": 12108 }, { "epoch": 3.73, "learning_rate": 1.4456911281267357e-05, "loss": 0.3065, "step": 12109 }, { "epoch": 3.73, "learning_rate": 1.445601916959198e-05, "loss": 0.546, "step": 12110 }, { "epoch": 3.73, "learning_rate": 1.4455127013664228e-05, "loss": 0.4723, "step": 12111 }, { "epoch": 3.73, "learning_rate": 1.4454234813492958e-05, "loss": 0.4886, "step": 12112 }, { "epoch": 3.73, "learning_rate": 1.4453342569087035e-05, "loss": 0.4755, "step": 12113 }, { "epoch": 3.73, "learning_rate": 1.445245028045531e-05, "loss": 0.5111, "step": 12114 }, { "epoch": 3.73, "learning_rate": 1.4451557947606654e-05, "loss": 0.561, "step": 12115 }, { "epoch": 3.73, "learning_rate": 1.4450665570549924e-05, "loss": 0.3745, "step": 12116 }, { "epoch": 3.73, "learning_rate": 1.444977314929398e-05, "loss": 0.5407, "step": 12117 }, { "epoch": 3.73, "learning_rate": 1.4448880683847693e-05, "loss": 0.8641, "step": 12118 }, { "epoch": 3.73, "learning_rate": 1.4447988174219919e-05, "loss": 0.564, "step": 12119 }, { "epoch": 3.73, "learning_rate": 1.4447095620419522e-05, "loss": 0.8953, "step": 12120 }, { "epoch": 3.73, "learning_rate": 1.4446203022455367e-05, "loss": 0.7202, "step": 12121 }, { "epoch": 3.73, "learning_rate": 1.4445310380336316e-05, "loss": 0.4344, "step": 12122 }, { "epoch": 3.73, "learning_rate": 1.4444417694071242e-05, "loss": 0.4856, "step": 12123 }, { "epoch": 3.73, "learning_rate": 1.4443524963668998e-05, "loss": 0.5478, "step": 12124 }, { "epoch": 3.73, "learning_rate": 1.444263218913846e-05, "loss": 0.5586, "step": 12125 }, { "epoch": 3.73, "learning_rate": 1.4441739370488485e-05, "loss": 0.5712, "step": 12126 }, { "epoch": 3.73, "learning_rate": 1.4440846507727949e-05, "loss": 0.5484, "step": 12127 }, { "epoch": 3.73, "learning_rate": 1.4439953600865713e-05, "loss": 0.4357, "step": 12128 }, { "epoch": 3.73, "learning_rate": 1.4439060649910646e-05, "loss": 0.5084, "step": 12129 }, { "epoch": 3.73, "learning_rate": 1.4438167654871615e-05, "loss": 0.6311, "step": 12130 }, { "epoch": 3.73, "learning_rate": 1.4437274615757488e-05, "loss": 0.5186, "step": 12131 }, { "epoch": 3.73, "learning_rate": 1.4436381532577137e-05, "loss": 0.5706, "step": 12132 }, { "epoch": 3.73, "learning_rate": 1.4435488405339428e-05, "loss": 0.3506, "step": 12133 }, { "epoch": 3.73, "learning_rate": 1.443459523405323e-05, "loss": 0.6028, "step": 12134 }, { "epoch": 3.73, "learning_rate": 1.4433702018727417e-05, "loss": 0.5016, "step": 12135 }, { "epoch": 3.73, "learning_rate": 1.4432808759370853e-05, "loss": 0.5772, "step": 12136 }, { "epoch": 3.73, "learning_rate": 1.4431915455992416e-05, "loss": 0.4637, "step": 12137 }, { "epoch": 3.73, "learning_rate": 1.4431022108600972e-05, "loss": 0.5089, "step": 12138 }, { "epoch": 3.73, "learning_rate": 1.4430128717205394e-05, "loss": 0.4937, "step": 12139 }, { "epoch": 3.74, "learning_rate": 1.4429235281814558e-05, "loss": 0.5536, "step": 12140 }, { "epoch": 3.74, "learning_rate": 1.442834180243733e-05, "loss": 0.4637, "step": 12141 }, { "epoch": 3.74, "learning_rate": 1.4427448279082587e-05, "loss": 0.4955, "step": 12142 }, { "epoch": 3.74, "learning_rate": 1.4426554711759205e-05, "loss": 0.7764, "step": 12143 }, { "epoch": 3.74, "learning_rate": 1.4425661100476054e-05, "loss": 0.6397, "step": 12144 }, { "epoch": 3.74, "learning_rate": 1.4424767445242007e-05, "loss": 0.4891, "step": 12145 }, { "epoch": 3.74, "learning_rate": 1.4423873746065944e-05, "loss": 0.3363, "step": 12146 }, { "epoch": 3.74, "learning_rate": 1.4422980002956737e-05, "loss": 0.8011, "step": 12147 }, { "epoch": 3.74, "learning_rate": 1.4422086215923261e-05, "loss": 0.7323, "step": 12148 }, { "epoch": 3.74, "learning_rate": 1.4421192384974396e-05, "loss": 0.44, "step": 12149 }, { "epoch": 3.74, "learning_rate": 1.4420298510119014e-05, "loss": 0.4726, "step": 12150 }, { "epoch": 3.74, "learning_rate": 1.4419404591365993e-05, "loss": 0.506, "step": 12151 }, { "epoch": 3.74, "learning_rate": 1.4418510628724212e-05, "loss": 0.6982, "step": 12152 }, { "epoch": 3.74, "learning_rate": 1.4417616622202549e-05, "loss": 0.8321, "step": 12153 }, { "epoch": 3.74, "learning_rate": 1.4416722571809883e-05, "loss": 0.5365, "step": 12154 }, { "epoch": 3.74, "learning_rate": 1.4415828477555087e-05, "loss": 0.5421, "step": 12155 }, { "epoch": 3.74, "learning_rate": 1.4414934339447047e-05, "loss": 0.5431, "step": 12156 }, { "epoch": 3.74, "learning_rate": 1.441404015749464e-05, "loss": 0.5941, "step": 12157 }, { "epoch": 3.74, "learning_rate": 1.4413145931706747e-05, "loss": 0.6812, "step": 12158 }, { "epoch": 3.74, "learning_rate": 1.4412251662092246e-05, "loss": 0.5451, "step": 12159 }, { "epoch": 3.74, "learning_rate": 1.441135734866002e-05, "loss": 0.3931, "step": 12160 }, { "epoch": 3.74, "learning_rate": 1.4410462991418952e-05, "loss": 0.4035, "step": 12161 }, { "epoch": 3.74, "learning_rate": 1.4409568590377918e-05, "loss": 0.645, "step": 12162 }, { "epoch": 3.74, "learning_rate": 1.4408674145545809e-05, "loss": 0.6857, "step": 12163 }, { "epoch": 3.74, "learning_rate": 1.4407779656931497e-05, "loss": 0.4982, "step": 12164 }, { "epoch": 3.74, "learning_rate": 1.4406885124543872e-05, "loss": 0.5306, "step": 12165 }, { "epoch": 3.74, "learning_rate": 1.4405990548391819e-05, "loss": 0.5947, "step": 12166 }, { "epoch": 3.74, "learning_rate": 1.4405095928484217e-05, "loss": 0.5035, "step": 12167 }, { "epoch": 3.74, "learning_rate": 1.4404201264829953e-05, "loss": 0.4362, "step": 12168 }, { "epoch": 3.74, "learning_rate": 1.4403306557437913e-05, "loss": 0.5154, "step": 12169 }, { "epoch": 3.74, "learning_rate": 1.4402411806316978e-05, "loss": 0.4925, "step": 12170 }, { "epoch": 3.74, "learning_rate": 1.4401517011476039e-05, "loss": 0.476, "step": 12171 }, { "epoch": 3.75, "learning_rate": 1.4400622172923978e-05, "loss": 0.4031, "step": 12172 }, { "epoch": 3.75, "learning_rate": 1.4399727290669682e-05, "loss": 0.658, "step": 12173 }, { "epoch": 3.75, "learning_rate": 1.4398832364722039e-05, "loss": 0.5788, "step": 12174 }, { "epoch": 3.75, "learning_rate": 1.439793739508994e-05, "loss": 0.4467, "step": 12175 }, { "epoch": 3.75, "learning_rate": 1.4397042381782265e-05, "loss": 0.5404, "step": 12176 }, { "epoch": 3.75, "learning_rate": 1.439614732480791e-05, "loss": 0.5659, "step": 12177 }, { "epoch": 3.75, "learning_rate": 1.439525222417576e-05, "loss": 0.6331, "step": 12178 }, { "epoch": 3.75, "learning_rate": 1.4394357079894704e-05, "loss": 0.6772, "step": 12179 }, { "epoch": 3.75, "learning_rate": 1.4393461891973632e-05, "loss": 0.6164, "step": 12180 }, { "epoch": 3.75, "learning_rate": 1.4392566660421433e-05, "loss": 0.4976, "step": 12181 }, { "epoch": 3.75, "learning_rate": 1.4391671385247002e-05, "loss": 0.4196, "step": 12182 }, { "epoch": 3.75, "learning_rate": 1.4390776066459224e-05, "loss": 0.6146, "step": 12183 }, { "epoch": 3.75, "learning_rate": 1.4389880704066994e-05, "loss": 0.7536, "step": 12184 }, { "epoch": 3.75, "learning_rate": 1.4388985298079203e-05, "loss": 0.5616, "step": 12185 }, { "epoch": 3.75, "learning_rate": 1.4388089848504744e-05, "loss": 0.6787, "step": 12186 }, { "epoch": 3.75, "learning_rate": 1.4387194355352506e-05, "loss": 0.6227, "step": 12187 }, { "epoch": 3.75, "learning_rate": 1.4386298818631388e-05, "loss": 0.4718, "step": 12188 }, { "epoch": 3.75, "learning_rate": 1.4385403238350277e-05, "loss": 0.5243, "step": 12189 }, { "epoch": 3.75, "learning_rate": 1.4384507614518073e-05, "loss": 0.6789, "step": 12190 }, { "epoch": 3.75, "learning_rate": 1.4383611947143664e-05, "loss": 0.4771, "step": 12191 }, { "epoch": 3.75, "learning_rate": 1.438271623623595e-05, "loss": 0.6642, "step": 12192 }, { "epoch": 3.75, "learning_rate": 1.4381820481803823e-05, "loss": 0.5114, "step": 12193 }, { "epoch": 3.75, "learning_rate": 1.4380924683856186e-05, "loss": 0.5523, "step": 12194 }, { "epoch": 3.75, "learning_rate": 1.4380028842401923e-05, "loss": 0.7619, "step": 12195 }, { "epoch": 3.75, "learning_rate": 1.437913295744994e-05, "loss": 0.3127, "step": 12196 }, { "epoch": 3.75, "learning_rate": 1.437823702900913e-05, "loss": 0.4039, "step": 12197 }, { "epoch": 3.75, "learning_rate": 1.437734105708839e-05, "loss": 0.5416, "step": 12198 }, { "epoch": 3.75, "learning_rate": 1.4376445041696621e-05, "loss": 0.4006, "step": 12199 }, { "epoch": 3.75, "learning_rate": 1.4375548982842718e-05, "loss": 0.7549, "step": 12200 }, { "epoch": 3.75, "learning_rate": 1.437465288053558e-05, "loss": 0.6196, "step": 12201 }, { "epoch": 3.75, "learning_rate": 1.437375673478411e-05, "loss": 0.5582, "step": 12202 }, { "epoch": 3.75, "learning_rate": 1.4372860545597203e-05, "loss": 0.6015, "step": 12203 }, { "epoch": 3.75, "learning_rate": 1.4371964312983762e-05, "loss": 0.4983, "step": 12204 }, { "epoch": 3.76, "learning_rate": 1.4371068036952683e-05, "loss": 0.4779, "step": 12205 }, { "epoch": 3.76, "learning_rate": 1.437017171751287e-05, "loss": 0.521, "step": 12206 }, { "epoch": 3.76, "learning_rate": 1.4369275354673227e-05, "loss": 0.233, "step": 12207 }, { "epoch": 3.76, "learning_rate": 1.4368378948442652e-05, "loss": 0.315, "step": 12208 }, { "epoch": 3.76, "learning_rate": 1.436748249883005e-05, "loss": 0.6737, "step": 12209 }, { "epoch": 3.76, "learning_rate": 1.4366586005844316e-05, "loss": 0.53, "step": 12210 }, { "epoch": 3.76, "learning_rate": 1.4365689469494364e-05, "loss": 0.6579, "step": 12211 }, { "epoch": 3.76, "learning_rate": 1.4364792889789088e-05, "loss": 0.4887, "step": 12212 }, { "epoch": 3.76, "learning_rate": 1.4363896266737398e-05, "loss": 0.639, "step": 12213 }, { "epoch": 3.76, "learning_rate": 1.4362999600348198e-05, "loss": 0.7994, "step": 12214 }, { "epoch": 3.76, "learning_rate": 1.436210289063039e-05, "loss": 0.5168, "step": 12215 }, { "epoch": 3.76, "learning_rate": 1.4361206137592878e-05, "loss": 0.4836, "step": 12216 }, { "epoch": 3.76, "learning_rate": 1.436030934124457e-05, "loss": 0.4094, "step": 12217 }, { "epoch": 3.76, "learning_rate": 1.4359412501594371e-05, "loss": 0.3778, "step": 12218 }, { "epoch": 3.76, "learning_rate": 1.4358515618651192e-05, "loss": 0.4356, "step": 12219 }, { "epoch": 3.76, "learning_rate": 1.4357618692423934e-05, "loss": 0.6067, "step": 12220 }, { "epoch": 3.76, "learning_rate": 1.4356721722921506e-05, "loss": 0.4866, "step": 12221 }, { "epoch": 3.76, "learning_rate": 1.4355824710152812e-05, "loss": 0.6408, "step": 12222 }, { "epoch": 3.76, "learning_rate": 1.435492765412677e-05, "loss": 0.471, "step": 12223 }, { "epoch": 3.76, "learning_rate": 1.435403055485228e-05, "loss": 0.5507, "step": 12224 }, { "epoch": 3.76, "learning_rate": 1.4353133412338256e-05, "loss": 0.5054, "step": 12225 }, { "epoch": 3.76, "learning_rate": 1.4352236226593604e-05, "loss": 0.5754, "step": 12226 }, { "epoch": 3.76, "learning_rate": 1.4351338997627233e-05, "loss": 0.4688, "step": 12227 }, { "epoch": 3.76, "learning_rate": 1.4350441725448059e-05, "loss": 0.4297, "step": 12228 }, { "epoch": 3.76, "learning_rate": 1.4349544410064988e-05, "loss": 0.6586, "step": 12229 }, { "epoch": 3.76, "learning_rate": 1.4348647051486932e-05, "loss": 0.435, "step": 12230 }, { "epoch": 3.76, "learning_rate": 1.43477496497228e-05, "loss": 0.3313, "step": 12231 }, { "epoch": 3.76, "learning_rate": 1.4346852204781511e-05, "loss": 0.523, "step": 12232 }, { "epoch": 3.76, "learning_rate": 1.4345954716671973e-05, "loss": 0.442, "step": 12233 }, { "epoch": 3.76, "learning_rate": 1.43450571854031e-05, "loss": 0.599, "step": 12234 }, { "epoch": 3.76, "learning_rate": 1.4344159610983803e-05, "loss": 0.7069, "step": 12235 }, { "epoch": 3.76, "learning_rate": 1.4343261993422997e-05, "loss": 0.5932, "step": 12236 }, { "epoch": 3.77, "learning_rate": 1.4342364332729597e-05, "loss": 0.6067, "step": 12237 }, { "epoch": 3.77, "learning_rate": 1.4341466628912517e-05, "loss": 0.5851, "step": 12238 }, { "epoch": 3.77, "learning_rate": 1.4340568881980672e-05, "loss": 0.5968, "step": 12239 }, { "epoch": 3.77, "learning_rate": 1.433967109194298e-05, "loss": 0.5626, "step": 12240 }, { "epoch": 3.77, "learning_rate": 1.4338773258808349e-05, "loss": 0.6076, "step": 12241 }, { "epoch": 3.77, "learning_rate": 1.4337875382585706e-05, "loss": 0.568, "step": 12242 }, { "epoch": 3.77, "learning_rate": 1.4336977463283958e-05, "loss": 0.5465, "step": 12243 }, { "epoch": 3.77, "learning_rate": 1.4336079500912033e-05, "loss": 0.4472, "step": 12244 }, { "epoch": 3.77, "learning_rate": 1.4335181495478839e-05, "loss": 0.449, "step": 12245 }, { "epoch": 3.77, "learning_rate": 1.4334283446993296e-05, "loss": 0.5305, "step": 12246 }, { "epoch": 3.77, "learning_rate": 1.4333385355464327e-05, "loss": 0.6138, "step": 12247 }, { "epoch": 3.77, "learning_rate": 1.4332487220900841e-05, "loss": 0.553, "step": 12248 }, { "epoch": 3.77, "learning_rate": 1.4331589043311773e-05, "loss": 0.4605, "step": 12249 }, { "epoch": 3.77, "learning_rate": 1.4330690822706026e-05, "loss": 0.484, "step": 12250 }, { "epoch": 3.77, "learning_rate": 1.4329792559092533e-05, "loss": 0.5549, "step": 12251 }, { "epoch": 3.77, "learning_rate": 1.4328894252480208e-05, "loss": 0.8286, "step": 12252 }, { "epoch": 3.77, "learning_rate": 1.4327995902877972e-05, "loss": 0.6398, "step": 12253 }, { "epoch": 3.77, "learning_rate": 1.4327097510294749e-05, "loss": 0.2945, "step": 12254 }, { "epoch": 3.77, "learning_rate": 1.4326199074739459e-05, "loss": 0.5832, "step": 12255 }, { "epoch": 3.77, "learning_rate": 1.4325300596221028e-05, "loss": 0.4066, "step": 12256 }, { "epoch": 3.77, "learning_rate": 1.432440207474837e-05, "loss": 0.6255, "step": 12257 }, { "epoch": 3.77, "learning_rate": 1.432350351033042e-05, "loss": 0.3313, "step": 12258 }, { "epoch": 3.77, "learning_rate": 1.4322604902976094e-05, "loss": 0.6952, "step": 12259 }, { "epoch": 3.77, "learning_rate": 1.4321706252694314e-05, "loss": 0.7871, "step": 12260 }, { "epoch": 3.77, "learning_rate": 1.432080755949401e-05, "loss": 0.2807, "step": 12261 }, { "epoch": 3.77, "learning_rate": 1.4319908823384103e-05, "loss": 0.5588, "step": 12262 }, { "epoch": 3.77, "learning_rate": 1.4319010044373525e-05, "loss": 0.5608, "step": 12263 }, { "epoch": 3.77, "learning_rate": 1.4318111222471194e-05, "loss": 0.5594, "step": 12264 }, { "epoch": 3.77, "learning_rate": 1.4317212357686039e-05, "loss": 0.6017, "step": 12265 }, { "epoch": 3.77, "learning_rate": 1.4316313450026986e-05, "loss": 0.6273, "step": 12266 }, { "epoch": 3.77, "learning_rate": 1.431541449950296e-05, "loss": 0.3812, "step": 12267 }, { "epoch": 3.77, "learning_rate": 1.4314515506122895e-05, "loss": 0.4949, "step": 12268 }, { "epoch": 3.77, "learning_rate": 1.4313616469895712e-05, "loss": 0.4503, "step": 12269 }, { "epoch": 3.78, "learning_rate": 1.4312717390830344e-05, "loss": 0.6239, "step": 12270 }, { "epoch": 3.78, "learning_rate": 1.4311818268935717e-05, "loss": 0.3871, "step": 12271 }, { "epoch": 3.78, "learning_rate": 1.4310919104220759e-05, "loss": 0.483, "step": 12272 }, { "epoch": 3.78, "learning_rate": 1.4310019896694403e-05, "loss": 0.7047, "step": 12273 }, { "epoch": 3.78, "learning_rate": 1.4309120646365575e-05, "loss": 0.5755, "step": 12274 }, { "epoch": 3.78, "learning_rate": 1.4308221353243212e-05, "loss": 0.5463, "step": 12275 }, { "epoch": 3.78, "learning_rate": 1.4307322017336236e-05, "loss": 0.5159, "step": 12276 }, { "epoch": 3.78, "learning_rate": 1.4306422638653583e-05, "loss": 0.4264, "step": 12277 }, { "epoch": 3.78, "learning_rate": 1.4305523217204186e-05, "loss": 0.4885, "step": 12278 }, { "epoch": 3.78, "learning_rate": 1.4304623752996974e-05, "loss": 0.3273, "step": 12279 }, { "epoch": 3.78, "learning_rate": 1.4303724246040885e-05, "loss": 0.5148, "step": 12280 }, { "epoch": 3.78, "learning_rate": 1.4302824696344841e-05, "loss": 0.5883, "step": 12281 }, { "epoch": 3.78, "learning_rate": 1.4301925103917786e-05, "loss": 0.6851, "step": 12282 }, { "epoch": 3.78, "learning_rate": 1.430102546876865e-05, "loss": 0.4535, "step": 12283 }, { "epoch": 3.78, "learning_rate": 1.4300125790906364e-05, "loss": 0.4121, "step": 12284 }, { "epoch": 3.78, "learning_rate": 1.4299226070339874e-05, "loss": 0.769, "step": 12285 }, { "epoch": 3.78, "learning_rate": 1.4298326307078098e-05, "loss": 0.7393, "step": 12286 }, { "epoch": 3.78, "learning_rate": 1.4297426501129985e-05, "loss": 0.5281, "step": 12287 }, { "epoch": 3.78, "learning_rate": 1.4296526652504466e-05, "loss": 0.7774, "step": 12288 }, { "epoch": 3.78, "learning_rate": 1.4295626761210475e-05, "loss": 0.4251, "step": 12289 }, { "epoch": 3.78, "learning_rate": 1.4294726827256956e-05, "loss": 0.5142, "step": 12290 }, { "epoch": 3.78, "learning_rate": 1.4293826850652837e-05, "loss": 0.3552, "step": 12291 }, { "epoch": 3.78, "learning_rate": 1.429292683140706e-05, "loss": 0.4883, "step": 12292 }, { "epoch": 3.78, "learning_rate": 1.4292026769528563e-05, "loss": 0.5311, "step": 12293 }, { "epoch": 3.78, "learning_rate": 1.4291126665026288e-05, "loss": 0.5028, "step": 12294 }, { "epoch": 3.78, "learning_rate": 1.4290226517909167e-05, "loss": 0.2643, "step": 12295 }, { "epoch": 3.78, "learning_rate": 1.4289326328186146e-05, "loss": 0.7646, "step": 12296 }, { "epoch": 3.78, "learning_rate": 1.4288426095866159e-05, "loss": 0.5058, "step": 12297 }, { "epoch": 3.78, "learning_rate": 1.4287525820958146e-05, "loss": 0.5091, "step": 12298 }, { "epoch": 3.78, "learning_rate": 1.4286625503471054e-05, "loss": 0.6128, "step": 12299 }, { "epoch": 3.78, "learning_rate": 1.428572514341382e-05, "loss": 0.5325, "step": 12300 }, { "epoch": 3.78, "learning_rate": 1.4284824740795385e-05, "loss": 0.486, "step": 12301 }, { "epoch": 3.79, "learning_rate": 1.4283924295624691e-05, "loss": 0.7983, "step": 12302 }, { "epoch": 3.79, "learning_rate": 1.4283023807910679e-05, "loss": 0.325, "step": 12303 }, { "epoch": 3.79, "learning_rate": 1.4282123277662297e-05, "loss": 0.8938, "step": 12304 }, { "epoch": 3.79, "learning_rate": 1.428122270488848e-05, "loss": 0.7084, "step": 12305 }, { "epoch": 3.79, "learning_rate": 1.428032208959818e-05, "loss": 0.4408, "step": 12306 }, { "epoch": 3.79, "learning_rate": 1.4279421431800336e-05, "loss": 0.4623, "step": 12307 }, { "epoch": 3.79, "learning_rate": 1.427852073150389e-05, "loss": 0.3644, "step": 12308 }, { "epoch": 3.79, "learning_rate": 1.427761998871779e-05, "loss": 0.5277, "step": 12309 }, { "epoch": 3.79, "learning_rate": 1.4276719203450984e-05, "loss": 0.3964, "step": 12310 }, { "epoch": 3.79, "learning_rate": 1.4275818375712413e-05, "loss": 0.5602, "step": 12311 }, { "epoch": 3.79, "learning_rate": 1.4274917505511026e-05, "loss": 0.5846, "step": 12312 }, { "epoch": 3.79, "learning_rate": 1.4274016592855768e-05, "loss": 0.5085, "step": 12313 }, { "epoch": 3.79, "learning_rate": 1.4273115637755585e-05, "loss": 0.5924, "step": 12314 }, { "epoch": 3.79, "learning_rate": 1.4272214640219428e-05, "loss": 0.6418, "step": 12315 }, { "epoch": 3.79, "learning_rate": 1.4271313600256243e-05, "loss": 0.5646, "step": 12316 }, { "epoch": 3.79, "learning_rate": 1.4270412517874973e-05, "loss": 0.6235, "step": 12317 }, { "epoch": 3.79, "learning_rate": 1.4269511393084572e-05, "loss": 0.4056, "step": 12318 }, { "epoch": 3.79, "learning_rate": 1.426861022589399e-05, "loss": 0.682, "step": 12319 }, { "epoch": 3.79, "learning_rate": 1.4267709016312174e-05, "loss": 0.362, "step": 12320 }, { "epoch": 3.79, "learning_rate": 1.4266807764348074e-05, "loss": 0.4927, "step": 12321 }, { "epoch": 3.79, "learning_rate": 1.426590647001064e-05, "loss": 0.5925, "step": 12322 }, { "epoch": 3.79, "learning_rate": 1.4265005133308822e-05, "loss": 0.3197, "step": 12323 }, { "epoch": 3.79, "learning_rate": 1.4264103754251573e-05, "loss": 0.6284, "step": 12324 }, { "epoch": 3.79, "learning_rate": 1.4263202332847846e-05, "loss": 0.5802, "step": 12325 }, { "epoch": 3.79, "learning_rate": 1.4262300869106588e-05, "loss": 0.5185, "step": 12326 }, { "epoch": 3.79, "learning_rate": 1.4261399363036757e-05, "loss": 0.3718, "step": 12327 }, { "epoch": 3.79, "learning_rate": 1.42604978146473e-05, "loss": 0.4668, "step": 12328 }, { "epoch": 3.79, "learning_rate": 1.4259596223947174e-05, "loss": 0.8131, "step": 12329 }, { "epoch": 3.79, "learning_rate": 1.4258694590945332e-05, "loss": 0.5068, "step": 12330 }, { "epoch": 3.79, "learning_rate": 1.4257792915650728e-05, "loss": 0.4687, "step": 12331 }, { "epoch": 3.79, "learning_rate": 1.4256891198072315e-05, "loss": 0.4621, "step": 12332 }, { "epoch": 3.79, "learning_rate": 1.4255989438219052e-05, "loss": 0.5405, "step": 12333 }, { "epoch": 3.79, "learning_rate": 1.425508763609989e-05, "loss": 0.4306, "step": 12334 }, { "epoch": 3.8, "learning_rate": 1.4254185791723787e-05, "loss": 0.4781, "step": 12335 }, { "epoch": 3.8, "learning_rate": 1.4253283905099695e-05, "loss": 0.7454, "step": 12336 }, { "epoch": 3.8, "learning_rate": 1.4252381976236575e-05, "loss": 0.4548, "step": 12337 }, { "epoch": 3.8, "learning_rate": 1.4251480005143385e-05, "loss": 0.5733, "step": 12338 }, { "epoch": 3.8, "learning_rate": 1.4250577991829078e-05, "loss": 0.4994, "step": 12339 }, { "epoch": 3.8, "learning_rate": 1.4249675936302619e-05, "loss": 0.6874, "step": 12340 }, { "epoch": 3.8, "learning_rate": 1.4248773838572956e-05, "loss": 0.5632, "step": 12341 }, { "epoch": 3.8, "learning_rate": 1.4247871698649053e-05, "loss": 0.4977, "step": 12342 }, { "epoch": 3.8, "learning_rate": 1.4246969516539871e-05, "loss": 0.4863, "step": 12343 }, { "epoch": 3.8, "learning_rate": 1.4246067292254367e-05, "loss": 0.5932, "step": 12344 }, { "epoch": 3.8, "learning_rate": 1.4245165025801502e-05, "loss": 0.5266, "step": 12345 }, { "epoch": 3.8, "learning_rate": 1.4244262717190236e-05, "loss": 0.5195, "step": 12346 }, { "epoch": 3.8, "learning_rate": 1.424336036642953e-05, "loss": 0.5473, "step": 12347 }, { "epoch": 3.8, "learning_rate": 1.4242457973528341e-05, "loss": 0.5444, "step": 12348 }, { "epoch": 3.8, "learning_rate": 1.424155553849564e-05, "loss": 0.5131, "step": 12349 }, { "epoch": 3.8, "learning_rate": 1.424065306134038e-05, "loss": 0.4987, "step": 12350 }, { "epoch": 3.8, "learning_rate": 1.4239750542071527e-05, "loss": 0.3505, "step": 12351 }, { "epoch": 3.8, "learning_rate": 1.4238847980698045e-05, "loss": 0.4132, "step": 12352 }, { "epoch": 3.8, "learning_rate": 1.4237945377228894e-05, "loss": 0.3242, "step": 12353 }, { "epoch": 3.8, "learning_rate": 1.4237042731673041e-05, "loss": 0.6536, "step": 12354 }, { "epoch": 3.8, "learning_rate": 1.4236140044039449e-05, "loss": 0.5732, "step": 12355 }, { "epoch": 3.8, "learning_rate": 1.4235237314337083e-05, "loss": 0.5322, "step": 12356 }, { "epoch": 3.8, "learning_rate": 1.4234334542574906e-05, "loss": 0.5115, "step": 12357 }, { "epoch": 3.8, "learning_rate": 1.4233431728761882e-05, "loss": 0.6345, "step": 12358 }, { "epoch": 3.8, "learning_rate": 1.4232528872906981e-05, "loss": 0.7071, "step": 12359 }, { "epoch": 3.8, "learning_rate": 1.423162597501917e-05, "loss": 0.3671, "step": 12360 }, { "epoch": 3.8, "learning_rate": 1.423072303510741e-05, "loss": 0.3579, "step": 12361 }, { "epoch": 3.8, "learning_rate": 1.4229820053180672e-05, "loss": 0.5503, "step": 12362 }, { "epoch": 3.8, "learning_rate": 1.4228917029247922e-05, "loss": 0.4953, "step": 12363 }, { "epoch": 3.8, "learning_rate": 1.422801396331813e-05, "loss": 0.346, "step": 12364 }, { "epoch": 3.8, "learning_rate": 1.422711085540026e-05, "loss": 0.5433, "step": 12365 }, { "epoch": 3.8, "learning_rate": 1.4226207705503288e-05, "loss": 0.52, "step": 12366 }, { "epoch": 3.81, "learning_rate": 1.4225304513636173e-05, "loss": 0.4939, "step": 12367 }, { "epoch": 3.81, "learning_rate": 1.4224401279807893e-05, "loss": 0.4421, "step": 12368 }, { "epoch": 3.81, "learning_rate": 1.4223498004027416e-05, "loss": 0.3837, "step": 12369 }, { "epoch": 3.81, "learning_rate": 1.4222594686303707e-05, "loss": 0.6206, "step": 12370 }, { "epoch": 3.81, "learning_rate": 1.4221691326645747e-05, "loss": 0.6425, "step": 12371 }, { "epoch": 3.81, "learning_rate": 1.4220787925062499e-05, "loss": 0.3906, "step": 12372 }, { "epoch": 3.81, "learning_rate": 1.4219884481562935e-05, "loss": 0.522, "step": 12373 }, { "epoch": 3.81, "learning_rate": 1.421898099615603e-05, "loss": 0.4771, "step": 12374 }, { "epoch": 3.81, "learning_rate": 1.4218077468850755e-05, "loss": 0.4763, "step": 12375 }, { "epoch": 3.81, "learning_rate": 1.4217173899656085e-05, "loss": 0.6792, "step": 12376 }, { "epoch": 3.81, "learning_rate": 1.421627028858099e-05, "loss": 0.4945, "step": 12377 }, { "epoch": 3.81, "learning_rate": 1.4215366635634447e-05, "loss": 0.6574, "step": 12378 }, { "epoch": 3.81, "learning_rate": 1.4214462940825424e-05, "loss": 0.7078, "step": 12379 }, { "epoch": 3.81, "learning_rate": 1.4213559204162906e-05, "loss": 0.5911, "step": 12380 }, { "epoch": 3.81, "learning_rate": 1.4212655425655857e-05, "loss": 0.6085, "step": 12381 }, { "epoch": 3.81, "learning_rate": 1.421175160531326e-05, "loss": 0.3806, "step": 12382 }, { "epoch": 3.81, "learning_rate": 1.4210847743144087e-05, "loss": 0.505, "step": 12383 }, { "epoch": 3.81, "learning_rate": 1.4209943839157311e-05, "loss": 0.534, "step": 12384 }, { "epoch": 3.81, "learning_rate": 1.4209039893361918e-05, "loss": 0.631, "step": 12385 }, { "epoch": 3.81, "learning_rate": 1.420813590576688e-05, "loss": 0.6349, "step": 12386 }, { "epoch": 3.81, "learning_rate": 1.4207231876381172e-05, "loss": 0.6481, "step": 12387 }, { "epoch": 3.81, "learning_rate": 1.4206327805213774e-05, "loss": 0.7372, "step": 12388 }, { "epoch": 3.81, "learning_rate": 1.4205423692273664e-05, "loss": 0.5267, "step": 12389 }, { "epoch": 3.81, "learning_rate": 1.420451953756982e-05, "loss": 0.4375, "step": 12390 }, { "epoch": 3.81, "learning_rate": 1.4203615341111223e-05, "loss": 0.4775, "step": 12391 }, { "epoch": 3.81, "learning_rate": 1.4202711102906855e-05, "loss": 0.5093, "step": 12392 }, { "epoch": 3.81, "learning_rate": 1.4201806822965687e-05, "loss": 0.5004, "step": 12393 }, { "epoch": 3.81, "learning_rate": 1.4200902501296708e-05, "loss": 0.3155, "step": 12394 }, { "epoch": 3.81, "learning_rate": 1.4199998137908893e-05, "loss": 0.5157, "step": 12395 }, { "epoch": 3.81, "learning_rate": 1.4199093732811227e-05, "loss": 0.6755, "step": 12396 }, { "epoch": 3.81, "learning_rate": 1.4198189286012691e-05, "loss": 0.5041, "step": 12397 }, { "epoch": 3.81, "learning_rate": 1.4197284797522264e-05, "loss": 0.7567, "step": 12398 }, { "epoch": 3.81, "learning_rate": 1.4196380267348933e-05, "loss": 0.4156, "step": 12399 }, { "epoch": 3.82, "learning_rate": 1.4195475695501677e-05, "loss": 0.6095, "step": 12400 }, { "epoch": 3.82, "learning_rate": 1.4194571081989483e-05, "loss": 0.6749, "step": 12401 }, { "epoch": 3.82, "learning_rate": 1.4193666426821332e-05, "loss": 0.7595, "step": 12402 }, { "epoch": 3.82, "learning_rate": 1.4192761730006203e-05, "loss": 0.5054, "step": 12403 }, { "epoch": 3.82, "learning_rate": 1.4191856991553092e-05, "loss": 0.5852, "step": 12404 }, { "epoch": 3.82, "learning_rate": 1.4190952211470975e-05, "loss": 0.4489, "step": 12405 }, { "epoch": 3.82, "learning_rate": 1.4190047389768841e-05, "loss": 0.4611, "step": 12406 }, { "epoch": 3.82, "learning_rate": 1.4189142526455677e-05, "loss": 0.4436, "step": 12407 }, { "epoch": 3.82, "learning_rate": 1.4188237621540461e-05, "loss": 0.7161, "step": 12408 }, { "epoch": 3.82, "learning_rate": 1.4187332675032189e-05, "loss": 0.6142, "step": 12409 }, { "epoch": 3.82, "learning_rate": 1.4186427686939843e-05, "loss": 0.6788, "step": 12410 }, { "epoch": 3.82, "learning_rate": 1.4185522657272415e-05, "loss": 0.528, "step": 12411 }, { "epoch": 3.82, "learning_rate": 1.4184617586038887e-05, "loss": 0.5962, "step": 12412 }, { "epoch": 3.82, "learning_rate": 1.418371247324825e-05, "loss": 0.7031, "step": 12413 }, { "epoch": 3.82, "learning_rate": 1.4182807318909494e-05, "loss": 0.5245, "step": 12414 }, { "epoch": 3.82, "learning_rate": 1.4181902123031601e-05, "loss": 0.3861, "step": 12415 }, { "epoch": 3.82, "learning_rate": 1.4180996885623573e-05, "loss": 0.4697, "step": 12416 }, { "epoch": 3.82, "learning_rate": 1.4180091606694388e-05, "loss": 0.6154, "step": 12417 }, { "epoch": 3.82, "learning_rate": 1.4179186286253041e-05, "loss": 0.518, "step": 12418 }, { "epoch": 3.82, "learning_rate": 1.4178280924308525e-05, "loss": 0.4423, "step": 12419 }, { "epoch": 3.82, "learning_rate": 1.4177375520869825e-05, "loss": 0.5776, "step": 12420 }, { "epoch": 3.82, "learning_rate": 1.417647007594594e-05, "loss": 0.5295, "step": 12421 }, { "epoch": 3.82, "learning_rate": 1.4175564589545853e-05, "loss": 0.3529, "step": 12422 }, { "epoch": 3.82, "learning_rate": 1.4174659061678569e-05, "loss": 0.6777, "step": 12423 }, { "epoch": 3.82, "learning_rate": 1.4173753492353064e-05, "loss": 0.5613, "step": 12424 }, { "epoch": 3.82, "learning_rate": 1.4172847881578347e-05, "loss": 0.4721, "step": 12425 }, { "epoch": 3.82, "learning_rate": 1.4171942229363404e-05, "loss": 0.4504, "step": 12426 }, { "epoch": 3.82, "learning_rate": 1.4171036535717226e-05, "loss": 0.5773, "step": 12427 }, { "epoch": 3.82, "learning_rate": 1.4170130800648814e-05, "loss": 0.5801, "step": 12428 }, { "epoch": 3.82, "learning_rate": 1.4169225024167158e-05, "loss": 0.5414, "step": 12429 }, { "epoch": 3.82, "learning_rate": 1.4168319206281256e-05, "loss": 0.5553, "step": 12430 }, { "epoch": 3.82, "learning_rate": 1.4167413347000103e-05, "loss": 0.4842, "step": 12431 }, { "epoch": 3.83, "learning_rate": 1.4166507446332695e-05, "loss": 0.4315, "step": 12432 }, { "epoch": 3.83, "learning_rate": 1.416560150428803e-05, "loss": 0.3361, "step": 12433 }, { "epoch": 3.83, "learning_rate": 1.4164695520875098e-05, "loss": 0.409, "step": 12434 }, { "epoch": 3.83, "learning_rate": 1.4163789496102902e-05, "loss": 0.5107, "step": 12435 }, { "epoch": 3.83, "learning_rate": 1.4162883429980441e-05, "loss": 0.7001, "step": 12436 }, { "epoch": 3.83, "learning_rate": 1.4161977322516712e-05, "loss": 0.556, "step": 12437 }, { "epoch": 3.83, "learning_rate": 1.4161071173720711e-05, "loss": 0.6254, "step": 12438 }, { "epoch": 3.83, "learning_rate": 1.4160164983601437e-05, "loss": 0.5219, "step": 12439 }, { "epoch": 3.83, "learning_rate": 1.415925875216789e-05, "loss": 0.5088, "step": 12440 }, { "epoch": 3.83, "learning_rate": 1.4158352479429075e-05, "loss": 0.4801, "step": 12441 }, { "epoch": 3.83, "learning_rate": 1.4157446165393984e-05, "loss": 0.433, "step": 12442 }, { "epoch": 3.83, "learning_rate": 1.4156539810071618e-05, "loss": 0.4743, "step": 12443 }, { "epoch": 3.83, "learning_rate": 1.4155633413470985e-05, "loss": 0.6384, "step": 12444 }, { "epoch": 3.83, "learning_rate": 1.4154726975601083e-05, "loss": 0.3847, "step": 12445 }, { "epoch": 3.83, "learning_rate": 1.4153820496470911e-05, "loss": 0.6828, "step": 12446 }, { "epoch": 3.83, "learning_rate": 1.4152913976089476e-05, "loss": 0.7695, "step": 12447 }, { "epoch": 3.83, "learning_rate": 1.4152007414465771e-05, "loss": 0.4526, "step": 12448 }, { "epoch": 3.83, "learning_rate": 1.415110081160881e-05, "loss": 0.3832, "step": 12449 }, { "epoch": 3.83, "learning_rate": 1.4150194167527595e-05, "loss": 0.4156, "step": 12450 }, { "epoch": 3.83, "learning_rate": 1.4149287482231125e-05, "loss": 0.6367, "step": 12451 }, { "epoch": 3.83, "learning_rate": 1.4148380755728407e-05, "loss": 0.6183, "step": 12452 }, { "epoch": 3.83, "learning_rate": 1.414747398802844e-05, "loss": 0.4525, "step": 12453 }, { "epoch": 3.83, "learning_rate": 1.4146567179140239e-05, "loss": 0.5533, "step": 12454 }, { "epoch": 3.83, "learning_rate": 1.4145660329072803e-05, "loss": 0.4101, "step": 12455 }, { "epoch": 3.83, "learning_rate": 1.4144753437835139e-05, "loss": 0.39, "step": 12456 }, { "epoch": 3.83, "learning_rate": 1.4143846505436255e-05, "loss": 0.2531, "step": 12457 }, { "epoch": 3.83, "learning_rate": 1.4142939531885152e-05, "loss": 0.578, "step": 12458 }, { "epoch": 3.83, "learning_rate": 1.4142032517190842e-05, "loss": 0.6582, "step": 12459 }, { "epoch": 3.83, "learning_rate": 1.4141125461362334e-05, "loss": 0.6982, "step": 12460 }, { "epoch": 3.83, "learning_rate": 1.4140218364408634e-05, "loss": 0.5066, "step": 12461 }, { "epoch": 3.83, "learning_rate": 1.4139311226338747e-05, "loss": 0.4886, "step": 12462 }, { "epoch": 3.83, "learning_rate": 1.4138404047161688e-05, "loss": 0.7396, "step": 12463 }, { "epoch": 3.83, "learning_rate": 1.413749682688646e-05, "loss": 0.6558, "step": 12464 }, { "epoch": 3.84, "learning_rate": 1.4136589565522077e-05, "loss": 0.4456, "step": 12465 }, { "epoch": 3.84, "learning_rate": 1.4135682263077546e-05, "loss": 0.4331, "step": 12466 }, { "epoch": 3.84, "learning_rate": 1.413477491956188e-05, "loss": 0.5647, "step": 12467 }, { "epoch": 3.84, "learning_rate": 1.4133867534984087e-05, "loss": 0.3854, "step": 12468 }, { "epoch": 3.84, "learning_rate": 1.413296010935318e-05, "loss": 0.6143, "step": 12469 }, { "epoch": 3.84, "learning_rate": 1.4132052642678168e-05, "loss": 0.5879, "step": 12470 }, { "epoch": 3.84, "learning_rate": 1.413114513496807e-05, "loss": 0.6681, "step": 12471 }, { "epoch": 3.84, "learning_rate": 1.413023758623189e-05, "loss": 0.3844, "step": 12472 }, { "epoch": 3.84, "learning_rate": 1.4129329996478644e-05, "loss": 0.6103, "step": 12473 }, { "epoch": 3.84, "learning_rate": 1.4128422365717346e-05, "loss": 0.7308, "step": 12474 }, { "epoch": 3.84, "learning_rate": 1.4127514693957009e-05, "loss": 0.6173, "step": 12475 }, { "epoch": 3.84, "learning_rate": 1.4126606981206648e-05, "loss": 0.3533, "step": 12476 }, { "epoch": 3.84, "learning_rate": 1.4125699227475275e-05, "loss": 0.6097, "step": 12477 }, { "epoch": 3.84, "learning_rate": 1.4124791432771909e-05, "loss": 0.3981, "step": 12478 }, { "epoch": 3.84, "learning_rate": 1.4123883597105558e-05, "loss": 0.5183, "step": 12479 }, { "epoch": 3.84, "learning_rate": 1.4122975720485246e-05, "loss": 0.5386, "step": 12480 }, { "epoch": 3.84, "learning_rate": 1.4122067802919986e-05, "loss": 0.3976, "step": 12481 }, { "epoch": 3.84, "learning_rate": 1.4121159844418793e-05, "loss": 0.427, "step": 12482 }, { "epoch": 3.84, "learning_rate": 1.4120251844990682e-05, "loss": 0.5184, "step": 12483 }, { "epoch": 3.84, "learning_rate": 1.4119343804644672e-05, "loss": 0.3938, "step": 12484 }, { "epoch": 3.84, "learning_rate": 1.4118435723389785e-05, "loss": 0.5265, "step": 12485 }, { "epoch": 3.84, "learning_rate": 1.4117527601235035e-05, "loss": 0.6579, "step": 12486 }, { "epoch": 3.84, "learning_rate": 1.411661943818944e-05, "loss": 0.6052, "step": 12487 }, { "epoch": 3.84, "learning_rate": 1.4115711234262021e-05, "loss": 0.5788, "step": 12488 }, { "epoch": 3.84, "learning_rate": 1.4114802989461796e-05, "loss": 0.7465, "step": 12489 }, { "epoch": 3.84, "learning_rate": 1.4113894703797784e-05, "loss": 0.6426, "step": 12490 }, { "epoch": 3.84, "learning_rate": 1.4112986377279005e-05, "loss": 0.5002, "step": 12491 }, { "epoch": 3.84, "learning_rate": 1.4112078009914485e-05, "loss": 0.328, "step": 12492 }, { "epoch": 3.84, "learning_rate": 1.4111169601713237e-05, "loss": 0.56, "step": 12493 }, { "epoch": 3.84, "learning_rate": 1.4110261152684288e-05, "loss": 0.5102, "step": 12494 }, { "epoch": 3.84, "learning_rate": 1.4109352662836657e-05, "loss": 0.5427, "step": 12495 }, { "epoch": 3.84, "learning_rate": 1.4108444132179363e-05, "loss": 0.5124, "step": 12496 }, { "epoch": 3.85, "learning_rate": 1.410753556072144e-05, "loss": 0.4146, "step": 12497 }, { "epoch": 3.85, "learning_rate": 1.4106626948471896e-05, "loss": 0.5035, "step": 12498 }, { "epoch": 3.85, "learning_rate": 1.4105718295439765e-05, "loss": 0.5771, "step": 12499 }, { "epoch": 3.85, "learning_rate": 1.4104809601634069e-05, "loss": 0.66, "step": 12500 }, { "epoch": 3.85, "learning_rate": 1.4103900867063827e-05, "loss": 0.6104, "step": 12501 }, { "epoch": 3.85, "learning_rate": 1.410299209173807e-05, "loss": 0.5137, "step": 12502 }, { "epoch": 3.85, "learning_rate": 1.4102083275665816e-05, "loss": 0.5555, "step": 12503 }, { "epoch": 3.85, "learning_rate": 1.4101174418856099e-05, "loss": 0.5151, "step": 12504 }, { "epoch": 3.85, "learning_rate": 1.4100265521317939e-05, "loss": 0.4429, "step": 12505 }, { "epoch": 3.85, "learning_rate": 1.4099356583060361e-05, "loss": 0.666, "step": 12506 }, { "epoch": 3.85, "learning_rate": 1.4098447604092398e-05, "loss": 0.4351, "step": 12507 }, { "epoch": 3.85, "learning_rate": 1.409753858442307e-05, "loss": 0.6709, "step": 12508 }, { "epoch": 3.85, "learning_rate": 1.409662952406141e-05, "loss": 0.4778, "step": 12509 }, { "epoch": 3.85, "learning_rate": 1.4095720423016439e-05, "loss": 0.655, "step": 12510 }, { "epoch": 3.85, "learning_rate": 1.4094811281297193e-05, "loss": 0.4887, "step": 12511 }, { "epoch": 3.85, "learning_rate": 1.4093902098912695e-05, "loss": 0.8694, "step": 12512 }, { "epoch": 3.85, "learning_rate": 1.409299287587198e-05, "loss": 0.8913, "step": 12513 }, { "epoch": 3.85, "learning_rate": 1.409208361218407e-05, "loss": 0.4396, "step": 12514 }, { "epoch": 3.85, "learning_rate": 1.4091174307857998e-05, "loss": 0.4014, "step": 12515 }, { "epoch": 3.85, "learning_rate": 1.4090264962902795e-05, "loss": 0.5184, "step": 12516 }, { "epoch": 3.85, "learning_rate": 1.4089355577327495e-05, "loss": 0.5915, "step": 12517 }, { "epoch": 3.85, "learning_rate": 1.4088446151141122e-05, "loss": 0.4829, "step": 12518 }, { "epoch": 3.85, "learning_rate": 1.4087536684352711e-05, "loss": 0.7856, "step": 12519 }, { "epoch": 3.85, "learning_rate": 1.4086627176971292e-05, "loss": 0.3387, "step": 12520 }, { "epoch": 3.85, "learning_rate": 1.4085717629005902e-05, "loss": 0.534, "step": 12521 }, { "epoch": 3.85, "learning_rate": 1.4084808040465569e-05, "loss": 0.4714, "step": 12522 }, { "epoch": 3.85, "learning_rate": 1.4083898411359327e-05, "loss": 0.4704, "step": 12523 }, { "epoch": 3.85, "learning_rate": 1.4082988741696209e-05, "loss": 0.3154, "step": 12524 }, { "epoch": 3.85, "learning_rate": 1.4082079031485253e-05, "loss": 0.5153, "step": 12525 }, { "epoch": 3.85, "learning_rate": 1.4081169280735488e-05, "loss": 0.3025, "step": 12526 }, { "epoch": 3.85, "learning_rate": 1.408025948945595e-05, "loss": 0.5885, "step": 12527 }, { "epoch": 3.85, "learning_rate": 1.4079349657655677e-05, "loss": 0.4322, "step": 12528 }, { "epoch": 3.85, "learning_rate": 1.4078439785343698e-05, "loss": 0.3046, "step": 12529 }, { "epoch": 3.86, "learning_rate": 1.4077529872529057e-05, "loss": 0.5344, "step": 12530 }, { "epoch": 3.86, "learning_rate": 1.4076619919220787e-05, "loss": 0.4183, "step": 12531 }, { "epoch": 3.86, "learning_rate": 1.4075709925427923e-05, "loss": 0.5879, "step": 12532 }, { "epoch": 3.86, "learning_rate": 1.4074799891159501e-05, "loss": 0.5208, "step": 12533 }, { "epoch": 3.86, "learning_rate": 1.4073889816424562e-05, "loss": 0.4889, "step": 12534 }, { "epoch": 3.86, "learning_rate": 1.4072979701232143e-05, "loss": 0.5605, "step": 12535 }, { "epoch": 3.86, "learning_rate": 1.4072069545591283e-05, "loss": 0.4694, "step": 12536 }, { "epoch": 3.86, "learning_rate": 1.4071159349511017e-05, "loss": 0.5646, "step": 12537 }, { "epoch": 3.86, "learning_rate": 1.4070249113000389e-05, "loss": 0.6646, "step": 12538 }, { "epoch": 3.86, "learning_rate": 1.4069338836068434e-05, "loss": 0.4159, "step": 12539 }, { "epoch": 3.86, "learning_rate": 1.4068428518724194e-05, "loss": 0.6074, "step": 12540 }, { "epoch": 3.86, "learning_rate": 1.4067518160976707e-05, "loss": 0.5291, "step": 12541 }, { "epoch": 3.86, "learning_rate": 1.4066607762835022e-05, "loss": 0.602, "step": 12542 }, { "epoch": 3.86, "learning_rate": 1.406569732430817e-05, "loss": 0.5665, "step": 12543 }, { "epoch": 3.86, "learning_rate": 1.4064786845405199e-05, "loss": 0.3133, "step": 12544 }, { "epoch": 3.86, "learning_rate": 1.4063876326135146e-05, "loss": 0.7845, "step": 12545 }, { "epoch": 3.86, "learning_rate": 1.4062965766507055e-05, "loss": 0.7139, "step": 12546 }, { "epoch": 3.86, "learning_rate": 1.4062055166529973e-05, "loss": 0.5393, "step": 12547 }, { "epoch": 3.86, "learning_rate": 1.4061144526212935e-05, "loss": 0.4697, "step": 12548 }, { "epoch": 3.86, "learning_rate": 1.4060233845564992e-05, "loss": 0.5253, "step": 12549 }, { "epoch": 3.86, "learning_rate": 1.4059323124595186e-05, "loss": 0.3512, "step": 12550 }, { "epoch": 3.86, "learning_rate": 1.4058412363312557e-05, "loss": 0.3825, "step": 12551 }, { "epoch": 3.86, "learning_rate": 1.4057501561726157e-05, "loss": 0.5556, "step": 12552 }, { "epoch": 3.86, "learning_rate": 1.405659071984502e-05, "loss": 0.3312, "step": 12553 }, { "epoch": 3.86, "learning_rate": 1.4055679837678204e-05, "loss": 0.606, "step": 12554 }, { "epoch": 3.86, "learning_rate": 1.4054768915234748e-05, "loss": 0.5701, "step": 12555 }, { "epoch": 3.86, "learning_rate": 1.40538579525237e-05, "loss": 0.5766, "step": 12556 }, { "epoch": 3.86, "learning_rate": 1.4052946949554107e-05, "loss": 0.4456, "step": 12557 }, { "epoch": 3.86, "learning_rate": 1.4052035906335016e-05, "loss": 0.4769, "step": 12558 }, { "epoch": 3.86, "learning_rate": 1.4051124822875473e-05, "loss": 0.5093, "step": 12559 }, { "epoch": 3.86, "learning_rate": 1.4050213699184522e-05, "loss": 0.4537, "step": 12560 }, { "epoch": 3.86, "learning_rate": 1.4049302535271222e-05, "loss": 0.5814, "step": 12561 }, { "epoch": 3.87, "learning_rate": 1.4048391331144615e-05, "loss": 0.6207, "step": 12562 }, { "epoch": 3.87, "learning_rate": 1.4047480086813751e-05, "loss": 0.3367, "step": 12563 }, { "epoch": 3.87, "learning_rate": 1.404656880228768e-05, "loss": 0.4972, "step": 12564 }, { "epoch": 3.87, "learning_rate": 1.404565747757545e-05, "loss": 0.4698, "step": 12565 }, { "epoch": 3.87, "learning_rate": 1.4044746112686114e-05, "loss": 0.5237, "step": 12566 }, { "epoch": 3.87, "learning_rate": 1.4043834707628721e-05, "loss": 0.6368, "step": 12567 }, { "epoch": 3.87, "learning_rate": 1.4042923262412322e-05, "loss": 0.4978, "step": 12568 }, { "epoch": 3.87, "learning_rate": 1.404201177704597e-05, "loss": 0.4164, "step": 12569 }, { "epoch": 3.87, "learning_rate": 1.4041100251538712e-05, "loss": 0.5532, "step": 12570 }, { "epoch": 3.87, "learning_rate": 1.4040188685899607e-05, "loss": 0.513, "step": 12571 }, { "epoch": 3.87, "learning_rate": 1.4039277080137702e-05, "loss": 0.6213, "step": 12572 }, { "epoch": 3.87, "learning_rate": 1.4038365434262061e-05, "loss": 0.4779, "step": 12573 }, { "epoch": 3.87, "learning_rate": 1.4037453748281722e-05, "loss": 0.7556, "step": 12574 }, { "epoch": 3.87, "learning_rate": 1.4036542022205746e-05, "loss": 0.4491, "step": 12575 }, { "epoch": 3.87, "learning_rate": 1.4035630256043191e-05, "loss": 0.3702, "step": 12576 }, { "epoch": 3.87, "learning_rate": 1.4034718449803105e-05, "loss": 0.524, "step": 12577 }, { "epoch": 3.87, "learning_rate": 1.403380660349455e-05, "loss": 0.6074, "step": 12578 }, { "epoch": 3.87, "learning_rate": 1.4032894717126575e-05, "loss": 0.4437, "step": 12579 }, { "epoch": 3.87, "learning_rate": 1.403198279070824e-05, "loss": 0.5179, "step": 12580 }, { "epoch": 3.87, "learning_rate": 1.4031070824248601e-05, "loss": 0.4243, "step": 12581 }, { "epoch": 3.87, "learning_rate": 1.4030158817756709e-05, "loss": 0.6217, "step": 12582 }, { "epoch": 3.87, "learning_rate": 1.4029246771241632e-05, "loss": 0.3869, "step": 12583 }, { "epoch": 3.87, "learning_rate": 1.4028334684712417e-05, "loss": 0.7158, "step": 12584 }, { "epoch": 3.87, "learning_rate": 1.4027422558178125e-05, "loss": 0.5074, "step": 12585 }, { "epoch": 3.87, "learning_rate": 1.4026510391647819e-05, "loss": 0.3828, "step": 12586 }, { "epoch": 3.87, "learning_rate": 1.402559818513055e-05, "loss": 0.4571, "step": 12587 }, { "epoch": 3.87, "learning_rate": 1.4024685938635385e-05, "loss": 0.7358, "step": 12588 }, { "epoch": 3.87, "learning_rate": 1.4023773652171375e-05, "loss": 0.4001, "step": 12589 }, { "epoch": 3.87, "learning_rate": 1.4022861325747585e-05, "loss": 0.83, "step": 12590 }, { "epoch": 3.87, "learning_rate": 1.4021948959373075e-05, "loss": 0.7438, "step": 12591 }, { "epoch": 3.87, "learning_rate": 1.4021036553056905e-05, "loss": 0.2369, "step": 12592 }, { "epoch": 3.87, "learning_rate": 1.402012410680814e-05, "loss": 0.5514, "step": 12593 }, { "epoch": 3.87, "learning_rate": 1.4019211620635836e-05, "loss": 0.4832, "step": 12594 }, { "epoch": 3.88, "learning_rate": 1.4018299094549055e-05, "loss": 0.5501, "step": 12595 }, { "epoch": 3.88, "learning_rate": 1.4017386528556857e-05, "loss": 0.3494, "step": 12596 }, { "epoch": 3.88, "learning_rate": 1.4016473922668312e-05, "loss": 0.8007, "step": 12597 }, { "epoch": 3.88, "learning_rate": 1.401556127689248e-05, "loss": 0.4879, "step": 12598 }, { "epoch": 3.88, "learning_rate": 1.4014648591238424e-05, "loss": 0.6864, "step": 12599 }, { "epoch": 3.88, "learning_rate": 1.4013735865715208e-05, "loss": 0.5234, "step": 12600 }, { "epoch": 3.88, "learning_rate": 1.4012823100331891e-05, "loss": 0.4604, "step": 12601 }, { "epoch": 3.88, "learning_rate": 1.4011910295097545e-05, "loss": 0.5071, "step": 12602 }, { "epoch": 3.88, "learning_rate": 1.4010997450021233e-05, "loss": 0.4834, "step": 12603 }, { "epoch": 3.88, "learning_rate": 1.4010084565112018e-05, "loss": 0.5179, "step": 12604 }, { "epoch": 3.88, "learning_rate": 1.4009171640378972e-05, "loss": 0.381, "step": 12605 }, { "epoch": 3.88, "learning_rate": 1.400825867583115e-05, "loss": 0.6665, "step": 12606 }, { "epoch": 3.88, "learning_rate": 1.400734567147763e-05, "loss": 0.3911, "step": 12607 }, { "epoch": 3.88, "learning_rate": 1.4006432627327472e-05, "loss": 0.4925, "step": 12608 }, { "epoch": 3.88, "learning_rate": 1.4005519543389747e-05, "loss": 0.3494, "step": 12609 }, { "epoch": 3.88, "learning_rate": 1.4004606419673518e-05, "loss": 0.5093, "step": 12610 }, { "epoch": 3.88, "learning_rate": 1.400369325618786e-05, "loss": 0.5167, "step": 12611 }, { "epoch": 3.88, "learning_rate": 1.4002780052941837e-05, "loss": 0.4473, "step": 12612 }, { "epoch": 3.88, "learning_rate": 1.4001866809944521e-05, "loss": 0.3835, "step": 12613 }, { "epoch": 3.88, "learning_rate": 1.4000953527204977e-05, "loss": 0.5062, "step": 12614 }, { "epoch": 3.88, "learning_rate": 1.4000040204732277e-05, "loss": 0.7494, "step": 12615 }, { "epoch": 3.88, "learning_rate": 1.3999126842535491e-05, "loss": 0.8567, "step": 12616 }, { "epoch": 3.88, "learning_rate": 1.3998213440623691e-05, "loss": 0.3337, "step": 12617 }, { "epoch": 3.88, "learning_rate": 1.3997299999005946e-05, "loss": 0.5274, "step": 12618 }, { "epoch": 3.88, "learning_rate": 1.399638651769133e-05, "loss": 0.387, "step": 12619 }, { "epoch": 3.88, "learning_rate": 1.399547299668891e-05, "loss": 0.5339, "step": 12620 }, { "epoch": 3.88, "learning_rate": 1.3994559436007763e-05, "loss": 0.5212, "step": 12621 }, { "epoch": 3.88, "learning_rate": 1.3993645835656955e-05, "loss": 0.3801, "step": 12622 }, { "epoch": 3.88, "learning_rate": 1.399273219564557e-05, "loss": 0.5293, "step": 12623 }, { "epoch": 3.88, "learning_rate": 1.399181851598267e-05, "loss": 0.5854, "step": 12624 }, { "epoch": 3.88, "learning_rate": 1.3990904796677334e-05, "loss": 0.47, "step": 12625 }, { "epoch": 3.88, "learning_rate": 1.3989991037738635e-05, "loss": 0.4869, "step": 12626 }, { "epoch": 3.89, "learning_rate": 1.3989077239175646e-05, "loss": 0.7516, "step": 12627 }, { "epoch": 3.89, "learning_rate": 1.3988163400997447e-05, "loss": 0.46, "step": 12628 }, { "epoch": 3.89, "learning_rate": 1.3987249523213107e-05, "loss": 0.4093, "step": 12629 }, { "epoch": 3.89, "learning_rate": 1.3986335605831707e-05, "loss": 0.5036, "step": 12630 }, { "epoch": 3.89, "learning_rate": 1.3985421648862319e-05, "loss": 0.5897, "step": 12631 }, { "epoch": 3.89, "learning_rate": 1.3984507652314019e-05, "loss": 0.4776, "step": 12632 }, { "epoch": 3.89, "learning_rate": 1.3983593616195892e-05, "loss": 0.6124, "step": 12633 }, { "epoch": 3.89, "learning_rate": 1.3982679540517e-05, "loss": 0.5154, "step": 12634 }, { "epoch": 3.89, "learning_rate": 1.3981765425286437e-05, "loss": 0.4896, "step": 12635 }, { "epoch": 3.89, "learning_rate": 1.3980851270513271e-05, "loss": 0.5853, "step": 12636 }, { "epoch": 3.89, "learning_rate": 1.3979937076206583e-05, "loss": 0.5577, "step": 12637 }, { "epoch": 3.89, "learning_rate": 1.3979022842375453e-05, "loss": 0.4575, "step": 12638 }, { "epoch": 3.89, "learning_rate": 1.397810856902896e-05, "loss": 0.5564, "step": 12639 }, { "epoch": 3.89, "learning_rate": 1.3977194256176178e-05, "loss": 0.4824, "step": 12640 }, { "epoch": 3.89, "learning_rate": 1.3976279903826193e-05, "loss": 0.5117, "step": 12641 }, { "epoch": 3.89, "learning_rate": 1.3975365511988087e-05, "loss": 0.5295, "step": 12642 }, { "epoch": 3.89, "learning_rate": 1.3974451080670934e-05, "loss": 0.5575, "step": 12643 }, { "epoch": 3.89, "learning_rate": 1.3973536609883822e-05, "loss": 0.7031, "step": 12644 }, { "epoch": 3.89, "learning_rate": 1.397262209963583e-05, "loss": 0.9234, "step": 12645 }, { "epoch": 3.89, "learning_rate": 1.3971707549936034e-05, "loss": 0.4553, "step": 12646 }, { "epoch": 3.89, "learning_rate": 1.3970792960793524e-05, "loss": 0.5989, "step": 12647 }, { "epoch": 3.89, "learning_rate": 1.3969878332217384e-05, "loss": 0.4075, "step": 12648 }, { "epoch": 3.89, "learning_rate": 1.396896366421669e-05, "loss": 0.5975, "step": 12649 }, { "epoch": 3.89, "learning_rate": 1.3968048956800528e-05, "loss": 0.6633, "step": 12650 }, { "epoch": 3.89, "learning_rate": 1.3967134209977984e-05, "loss": 0.5894, "step": 12651 }, { "epoch": 3.89, "learning_rate": 1.3966219423758143e-05, "loss": 0.5144, "step": 12652 }, { "epoch": 3.89, "learning_rate": 1.3965304598150087e-05, "loss": 0.383, "step": 12653 }, { "epoch": 3.89, "learning_rate": 1.39643897331629e-05, "loss": 0.6527, "step": 12654 }, { "epoch": 3.89, "learning_rate": 1.396347482880567e-05, "loss": 0.6315, "step": 12655 }, { "epoch": 3.89, "learning_rate": 1.3962559885087482e-05, "loss": 0.4528, "step": 12656 }, { "epoch": 3.89, "learning_rate": 1.396164490201742e-05, "loss": 0.6085, "step": 12657 }, { "epoch": 3.89, "learning_rate": 1.3960729879604575e-05, "loss": 0.529, "step": 12658 }, { "epoch": 3.89, "learning_rate": 1.3959814817858035e-05, "loss": 0.6228, "step": 12659 }, { "epoch": 3.9, "learning_rate": 1.3958899716786881e-05, "loss": 0.5731, "step": 12660 }, { "epoch": 3.9, "learning_rate": 1.3957984576400205e-05, "loss": 0.6143, "step": 12661 }, { "epoch": 3.9, "learning_rate": 1.3957069396707094e-05, "loss": 0.4933, "step": 12662 }, { "epoch": 3.9, "learning_rate": 1.3956154177716636e-05, "loss": 0.7, "step": 12663 }, { "epoch": 3.9, "learning_rate": 1.3955238919437925e-05, "loss": 0.3457, "step": 12664 }, { "epoch": 3.9, "learning_rate": 1.3954323621880044e-05, "loss": 0.4016, "step": 12665 }, { "epoch": 3.9, "learning_rate": 1.3953408285052084e-05, "loss": 0.4215, "step": 12666 }, { "epoch": 3.9, "learning_rate": 1.3952492908963138e-05, "loss": 0.2734, "step": 12667 }, { "epoch": 3.9, "learning_rate": 1.395157749362229e-05, "loss": 0.5531, "step": 12668 }, { "epoch": 3.9, "learning_rate": 1.3950662039038643e-05, "loss": 0.4708, "step": 12669 }, { "epoch": 3.9, "learning_rate": 1.3949746545221275e-05, "loss": 0.5978, "step": 12670 }, { "epoch": 3.9, "learning_rate": 1.3948831012179284e-05, "loss": 0.4869, "step": 12671 }, { "epoch": 3.9, "learning_rate": 1.3947915439921764e-05, "loss": 0.4833, "step": 12672 }, { "epoch": 3.9, "learning_rate": 1.3946999828457805e-05, "loss": 0.4706, "step": 12673 }, { "epoch": 3.9, "learning_rate": 1.3946084177796501e-05, "loss": 0.5642, "step": 12674 }, { "epoch": 3.9, "learning_rate": 1.394516848794694e-05, "loss": 0.5475, "step": 12675 }, { "epoch": 3.9, "learning_rate": 1.3944252758918224e-05, "loss": 0.7223, "step": 12676 }, { "epoch": 3.9, "learning_rate": 1.394333699071944e-05, "loss": 0.5651, "step": 12677 }, { "epoch": 3.9, "learning_rate": 1.3942421183359688e-05, "loss": 0.5684, "step": 12678 }, { "epoch": 3.9, "learning_rate": 1.394150533684806e-05, "loss": 0.5614, "step": 12679 }, { "epoch": 3.9, "learning_rate": 1.3940589451193651e-05, "loss": 0.5219, "step": 12680 }, { "epoch": 3.9, "learning_rate": 1.3939673526405556e-05, "loss": 0.6404, "step": 12681 }, { "epoch": 3.9, "learning_rate": 1.3938757562492873e-05, "loss": 0.4251, "step": 12682 }, { "epoch": 3.9, "learning_rate": 1.3937841559464697e-05, "loss": 0.5668, "step": 12683 }, { "epoch": 3.9, "learning_rate": 1.3936925517330125e-05, "loss": 0.371, "step": 12684 }, { "epoch": 3.9, "learning_rate": 1.3936009436098255e-05, "loss": 0.533, "step": 12685 }, { "epoch": 3.9, "learning_rate": 1.3935093315778183e-05, "loss": 0.5667, "step": 12686 }, { "epoch": 3.9, "learning_rate": 1.3934177156379008e-05, "loss": 0.435, "step": 12687 }, { "epoch": 3.9, "learning_rate": 1.3933260957909826e-05, "loss": 0.7411, "step": 12688 }, { "epoch": 3.9, "learning_rate": 1.3932344720379742e-05, "loss": 0.6957, "step": 12689 }, { "epoch": 3.9, "learning_rate": 1.3931428443797848e-05, "loss": 0.4896, "step": 12690 }, { "epoch": 3.9, "learning_rate": 1.3930512128173245e-05, "loss": 0.4331, "step": 12691 }, { "epoch": 3.91, "learning_rate": 1.3929595773515035e-05, "loss": 0.5763, "step": 12692 }, { "epoch": 3.91, "learning_rate": 1.392867937983232e-05, "loss": 0.4939, "step": 12693 }, { "epoch": 3.91, "learning_rate": 1.3927762947134197e-05, "loss": 0.3872, "step": 12694 }, { "epoch": 3.91, "learning_rate": 1.3926846475429767e-05, "loss": 0.5269, "step": 12695 }, { "epoch": 3.91, "learning_rate": 1.3925929964728129e-05, "loss": 0.5852, "step": 12696 }, { "epoch": 3.91, "learning_rate": 1.3925013415038392e-05, "loss": 0.3299, "step": 12697 }, { "epoch": 3.91, "learning_rate": 1.3924096826369655e-05, "loss": 0.4351, "step": 12698 }, { "epoch": 3.91, "learning_rate": 1.3923180198731019e-05, "loss": 0.5611, "step": 12699 }, { "epoch": 3.91, "learning_rate": 1.3922263532131587e-05, "loss": 0.5544, "step": 12700 }, { "epoch": 3.91, "learning_rate": 1.3921346826580463e-05, "loss": 0.4523, "step": 12701 }, { "epoch": 3.91, "learning_rate": 1.3920430082086752e-05, "loss": 0.3705, "step": 12702 }, { "epoch": 3.91, "learning_rate": 1.391951329865956e-05, "loss": 0.5428, "step": 12703 }, { "epoch": 3.91, "learning_rate": 1.3918596476307987e-05, "loss": 0.4102, "step": 12704 }, { "epoch": 3.91, "learning_rate": 1.3917679615041138e-05, "loss": 0.3682, "step": 12705 }, { "epoch": 3.91, "learning_rate": 1.3916762714868118e-05, "loss": 0.4824, "step": 12706 }, { "epoch": 3.91, "learning_rate": 1.3915845775798037e-05, "loss": 0.406, "step": 12707 }, { "epoch": 3.91, "learning_rate": 1.3914928797839996e-05, "loss": 0.4158, "step": 12708 }, { "epoch": 3.91, "learning_rate": 1.3914011781003106e-05, "loss": 0.8193, "step": 12709 }, { "epoch": 3.91, "learning_rate": 1.3913094725296472e-05, "loss": 0.5879, "step": 12710 }, { "epoch": 3.91, "learning_rate": 1.3912177630729202e-05, "loss": 0.4839, "step": 12711 }, { "epoch": 3.91, "learning_rate": 1.3911260497310401e-05, "loss": 0.4352, "step": 12712 }, { "epoch": 3.91, "learning_rate": 1.3910343325049179e-05, "loss": 0.5166, "step": 12713 }, { "epoch": 3.91, "learning_rate": 1.3909426113954645e-05, "loss": 0.458, "step": 12714 }, { "epoch": 3.91, "learning_rate": 1.3908508864035903e-05, "loss": 0.3303, "step": 12715 }, { "epoch": 3.91, "learning_rate": 1.3907591575302068e-05, "loss": 0.6042, "step": 12716 }, { "epoch": 3.91, "learning_rate": 1.3906674247762249e-05, "loss": 0.4371, "step": 12717 }, { "epoch": 3.91, "learning_rate": 1.3905756881425552e-05, "loss": 0.5362, "step": 12718 }, { "epoch": 3.91, "learning_rate": 1.3904839476301091e-05, "loss": 0.6141, "step": 12719 }, { "epoch": 3.91, "learning_rate": 1.3903922032397974e-05, "loss": 0.6223, "step": 12720 }, { "epoch": 3.91, "learning_rate": 1.3903004549725313e-05, "loss": 0.4615, "step": 12721 }, { "epoch": 3.91, "learning_rate": 1.390208702829222e-05, "loss": 0.5815, "step": 12722 }, { "epoch": 3.91, "learning_rate": 1.3901169468107807e-05, "loss": 0.5205, "step": 12723 }, { "epoch": 3.91, "learning_rate": 1.3900251869181187e-05, "loss": 0.8002, "step": 12724 }, { "epoch": 3.92, "learning_rate": 1.389933423152147e-05, "loss": 0.5549, "step": 12725 }, { "epoch": 3.92, "learning_rate": 1.3898416555137768e-05, "loss": 0.6921, "step": 12726 }, { "epoch": 3.92, "learning_rate": 1.38974988400392e-05, "loss": 0.5607, "step": 12727 }, { "epoch": 3.92, "learning_rate": 1.3896581086234876e-05, "loss": 0.3853, "step": 12728 }, { "epoch": 3.92, "learning_rate": 1.389566329373391e-05, "loss": 0.676, "step": 12729 }, { "epoch": 3.92, "learning_rate": 1.3894745462545416e-05, "loss": 0.4282, "step": 12730 }, { "epoch": 3.92, "learning_rate": 1.3893827592678512e-05, "loss": 0.3741, "step": 12731 }, { "epoch": 3.92, "learning_rate": 1.3892909684142311e-05, "loss": 0.3981, "step": 12732 }, { "epoch": 3.92, "learning_rate": 1.3891991736945927e-05, "loss": 0.4694, "step": 12733 }, { "epoch": 3.92, "learning_rate": 1.3891073751098481e-05, "loss": 0.6159, "step": 12734 }, { "epoch": 3.92, "learning_rate": 1.3890155726609087e-05, "loss": 0.4821, "step": 12735 }, { "epoch": 3.92, "learning_rate": 1.388923766348686e-05, "loss": 0.3316, "step": 12736 }, { "epoch": 3.92, "learning_rate": 1.3888319561740916e-05, "loss": 0.7781, "step": 12737 }, { "epoch": 3.92, "learning_rate": 1.3887401421380378e-05, "loss": 0.4382, "step": 12738 }, { "epoch": 3.92, "learning_rate": 1.3886483242414362e-05, "loss": 0.525, "step": 12739 }, { "epoch": 3.92, "learning_rate": 1.3885565024851983e-05, "loss": 0.4951, "step": 12740 }, { "epoch": 3.92, "learning_rate": 1.3884646768702364e-05, "loss": 0.5193, "step": 12741 }, { "epoch": 3.92, "learning_rate": 1.388372847397462e-05, "loss": 0.5807, "step": 12742 }, { "epoch": 3.92, "learning_rate": 1.3882810140677876e-05, "loss": 0.7119, "step": 12743 }, { "epoch": 3.92, "learning_rate": 1.3881891768821247e-05, "loss": 0.3949, "step": 12744 }, { "epoch": 3.92, "learning_rate": 1.3880973358413856e-05, "loss": 0.5306, "step": 12745 }, { "epoch": 3.92, "learning_rate": 1.3880054909464822e-05, "loss": 0.6871, "step": 12746 }, { "epoch": 3.92, "learning_rate": 1.3879136421983265e-05, "loss": 0.6343, "step": 12747 }, { "epoch": 3.92, "learning_rate": 1.3878217895978313e-05, "loss": 0.4162, "step": 12748 }, { "epoch": 3.92, "learning_rate": 1.387729933145908e-05, "loss": 0.4589, "step": 12749 }, { "epoch": 3.92, "learning_rate": 1.387638072843469e-05, "loss": 0.6611, "step": 12750 }, { "epoch": 3.92, "learning_rate": 1.3875462086914268e-05, "loss": 0.7102, "step": 12751 }, { "epoch": 3.92, "learning_rate": 1.3874543406906937e-05, "loss": 0.5391, "step": 12752 }, { "epoch": 3.92, "learning_rate": 1.387362468842182e-05, "loss": 0.4411, "step": 12753 }, { "epoch": 3.92, "learning_rate": 1.3872705931468038e-05, "loss": 0.6102, "step": 12754 }, { "epoch": 3.92, "learning_rate": 1.3871787136054717e-05, "loss": 0.5276, "step": 12755 }, { "epoch": 3.92, "learning_rate": 1.3870868302190978e-05, "loss": 0.5971, "step": 12756 }, { "epoch": 3.93, "learning_rate": 1.3869949429885955e-05, "loss": 0.4179, "step": 12757 }, { "epoch": 3.93, "learning_rate": 1.3869030519148763e-05, "loss": 0.6078, "step": 12758 }, { "epoch": 3.93, "learning_rate": 1.3868111569988533e-05, "loss": 0.4268, "step": 12759 }, { "epoch": 3.93, "learning_rate": 1.3867192582414393e-05, "loss": 0.6088, "step": 12760 }, { "epoch": 3.93, "learning_rate": 1.3866273556435466e-05, "loss": 0.6768, "step": 12761 }, { "epoch": 3.93, "learning_rate": 1.3865354492060879e-05, "loss": 0.4389, "step": 12762 }, { "epoch": 3.93, "learning_rate": 1.3864435389299756e-05, "loss": 0.4976, "step": 12763 }, { "epoch": 3.93, "learning_rate": 1.386351624816123e-05, "loss": 0.753, "step": 12764 }, { "epoch": 3.93, "learning_rate": 1.3862597068654428e-05, "loss": 0.4782, "step": 12765 }, { "epoch": 3.93, "learning_rate": 1.3861677850788478e-05, "loss": 0.4406, "step": 12766 }, { "epoch": 3.93, "learning_rate": 1.3860758594572507e-05, "loss": 0.5786, "step": 12767 }, { "epoch": 3.93, "learning_rate": 1.3859839300015644e-05, "loss": 0.3852, "step": 12768 }, { "epoch": 3.93, "learning_rate": 1.385891996712702e-05, "loss": 0.5937, "step": 12769 }, { "epoch": 3.93, "learning_rate": 1.3858000595915764e-05, "loss": 0.6494, "step": 12770 }, { "epoch": 3.93, "learning_rate": 1.3857081186391007e-05, "loss": 0.5287, "step": 12771 }, { "epoch": 3.93, "learning_rate": 1.3856161738561879e-05, "loss": 0.4569, "step": 12772 }, { "epoch": 3.93, "learning_rate": 1.3855242252437511e-05, "loss": 0.4352, "step": 12773 }, { "epoch": 3.93, "learning_rate": 1.3854322728027035e-05, "loss": 0.4015, "step": 12774 }, { "epoch": 3.93, "learning_rate": 1.3853403165339581e-05, "loss": 0.46, "step": 12775 }, { "epoch": 3.93, "learning_rate": 1.3852483564384282e-05, "loss": 0.4317, "step": 12776 }, { "epoch": 3.93, "learning_rate": 1.385156392517027e-05, "loss": 0.6806, "step": 12777 }, { "epoch": 3.93, "learning_rate": 1.3850644247706679e-05, "loss": 0.5855, "step": 12778 }, { "epoch": 3.93, "learning_rate": 1.3849724532002644e-05, "loss": 0.5146, "step": 12779 }, { "epoch": 3.93, "learning_rate": 1.3848804778067296e-05, "loss": 0.4259, "step": 12780 }, { "epoch": 3.93, "learning_rate": 1.3847884985909767e-05, "loss": 0.3716, "step": 12781 }, { "epoch": 3.93, "learning_rate": 1.3846965155539194e-05, "loss": 0.3005, "step": 12782 }, { "epoch": 3.93, "learning_rate": 1.3846045286964714e-05, "loss": 0.3575, "step": 12783 }, { "epoch": 3.93, "learning_rate": 1.3845125380195459e-05, "loss": 0.4011, "step": 12784 }, { "epoch": 3.93, "learning_rate": 1.3844205435240562e-05, "loss": 0.4418, "step": 12785 }, { "epoch": 3.93, "learning_rate": 1.3843285452109166e-05, "loss": 0.577, "step": 12786 }, { "epoch": 3.93, "learning_rate": 1.38423654308104e-05, "loss": 0.5663, "step": 12787 }, { "epoch": 3.93, "learning_rate": 1.3841445371353407e-05, "loss": 0.3908, "step": 12788 }, { "epoch": 3.93, "learning_rate": 1.3840525273747319e-05, "loss": 0.6426, "step": 12789 }, { "epoch": 3.94, "learning_rate": 1.383960513800128e-05, "loss": 0.6307, "step": 12790 }, { "epoch": 3.94, "learning_rate": 1.3838684964124417e-05, "loss": 0.3729, "step": 12791 }, { "epoch": 3.94, "learning_rate": 1.3837764752125877e-05, "loss": 0.7501, "step": 12792 }, { "epoch": 3.94, "learning_rate": 1.3836844502014796e-05, "loss": 0.5195, "step": 12793 }, { "epoch": 3.94, "learning_rate": 1.3835924213800311e-05, "loss": 0.5705, "step": 12794 }, { "epoch": 3.94, "learning_rate": 1.383500388749157e-05, "loss": 0.4486, "step": 12795 }, { "epoch": 3.94, "learning_rate": 1.38340835230977e-05, "loss": 0.414, "step": 12796 }, { "epoch": 3.94, "learning_rate": 1.3833163120627848e-05, "loss": 0.5079, "step": 12797 }, { "epoch": 3.94, "learning_rate": 1.3832242680091155e-05, "loss": 0.5863, "step": 12798 }, { "epoch": 3.94, "learning_rate": 1.3831322201496757e-05, "loss": 0.4571, "step": 12799 }, { "epoch": 3.94, "learning_rate": 1.3830401684853802e-05, "loss": 0.4006, "step": 12800 }, { "epoch": 3.94, "learning_rate": 1.3829481130171427e-05, "loss": 0.6199, "step": 12801 }, { "epoch": 3.94, "learning_rate": 1.3828560537458775e-05, "loss": 0.5763, "step": 12802 }, { "epoch": 3.94, "learning_rate": 1.3827639906724988e-05, "loss": 0.5075, "step": 12803 }, { "epoch": 3.94, "learning_rate": 1.3826719237979207e-05, "loss": 0.3119, "step": 12804 }, { "epoch": 3.94, "learning_rate": 1.3825798531230582e-05, "loss": 0.5001, "step": 12805 }, { "epoch": 3.94, "learning_rate": 1.382487778648825e-05, "loss": 0.4383, "step": 12806 }, { "epoch": 3.94, "learning_rate": 1.3823957003761352e-05, "loss": 0.4517, "step": 12807 }, { "epoch": 3.94, "learning_rate": 1.382303618305904e-05, "loss": 0.4, "step": 12808 }, { "epoch": 3.94, "learning_rate": 1.3822115324390458e-05, "loss": 0.4387, "step": 12809 }, { "epoch": 3.94, "learning_rate": 1.3821194427764743e-05, "loss": 0.6066, "step": 12810 }, { "epoch": 3.94, "learning_rate": 1.3820273493191049e-05, "loss": 0.554, "step": 12811 }, { "epoch": 3.94, "learning_rate": 1.3819352520678519e-05, "loss": 0.3801, "step": 12812 }, { "epoch": 3.94, "learning_rate": 1.3818431510236296e-05, "loss": 0.5376, "step": 12813 }, { "epoch": 3.94, "learning_rate": 1.381751046187353e-05, "loss": 0.5254, "step": 12814 }, { "epoch": 3.94, "learning_rate": 1.3816589375599368e-05, "loss": 0.5196, "step": 12815 }, { "epoch": 3.94, "learning_rate": 1.3815668251422953e-05, "loss": 0.439, "step": 12816 }, { "epoch": 3.94, "learning_rate": 1.381474708935344e-05, "loss": 0.6318, "step": 12817 }, { "epoch": 3.94, "learning_rate": 1.3813825889399968e-05, "loss": 0.6069, "step": 12818 }, { "epoch": 3.94, "learning_rate": 1.3812904651571693e-05, "loss": 0.5284, "step": 12819 }, { "epoch": 3.94, "learning_rate": 1.381198337587776e-05, "loss": 0.5551, "step": 12820 }, { "epoch": 3.94, "learning_rate": 1.3811062062327319e-05, "loss": 0.5504, "step": 12821 }, { "epoch": 3.95, "learning_rate": 1.3810140710929521e-05, "loss": 0.5668, "step": 12822 }, { "epoch": 3.95, "learning_rate": 1.380921932169351e-05, "loss": 0.5201, "step": 12823 }, { "epoch": 3.95, "learning_rate": 1.3808297894628441e-05, "loss": 0.4849, "step": 12824 }, { "epoch": 3.95, "learning_rate": 1.3807376429743467e-05, "loss": 0.4243, "step": 12825 }, { "epoch": 3.95, "learning_rate": 1.3806454927047737e-05, "loss": 0.5747, "step": 12826 }, { "epoch": 3.95, "learning_rate": 1.3805533386550396e-05, "loss": 0.4595, "step": 12827 }, { "epoch": 3.95, "learning_rate": 1.3804611808260607e-05, "loss": 0.5389, "step": 12828 }, { "epoch": 3.95, "learning_rate": 1.3803690192187514e-05, "loss": 0.6434, "step": 12829 }, { "epoch": 3.95, "learning_rate": 1.380276853834027e-05, "loss": 0.6121, "step": 12830 }, { "epoch": 3.95, "learning_rate": 1.3801846846728034e-05, "loss": 0.5665, "step": 12831 }, { "epoch": 3.95, "learning_rate": 1.3800925117359949e-05, "loss": 0.4057, "step": 12832 }, { "epoch": 3.95, "learning_rate": 1.380000335024518e-05, "loss": 0.5569, "step": 12833 }, { "epoch": 3.95, "learning_rate": 1.3799081545392873e-05, "loss": 0.4486, "step": 12834 }, { "epoch": 3.95, "learning_rate": 1.3798159702812183e-05, "loss": 0.6017, "step": 12835 }, { "epoch": 3.95, "learning_rate": 1.379723782251227e-05, "loss": 0.6532, "step": 12836 }, { "epoch": 3.95, "learning_rate": 1.3796315904502284e-05, "loss": 0.4816, "step": 12837 }, { "epoch": 3.95, "learning_rate": 1.3795393948791382e-05, "loss": 0.5747, "step": 12838 }, { "epoch": 3.95, "learning_rate": 1.3794471955388721e-05, "loss": 0.479, "step": 12839 }, { "epoch": 3.95, "learning_rate": 1.3793549924303458e-05, "loss": 0.6051, "step": 12840 }, { "epoch": 3.95, "learning_rate": 1.3792627855544747e-05, "loss": 0.5423, "step": 12841 }, { "epoch": 3.95, "learning_rate": 1.3791705749121745e-05, "loss": 0.416, "step": 12842 }, { "epoch": 3.95, "learning_rate": 1.3790783605043612e-05, "loss": 0.6904, "step": 12843 }, { "epoch": 3.95, "learning_rate": 1.3789861423319501e-05, "loss": 0.5505, "step": 12844 }, { "epoch": 3.95, "learning_rate": 1.3788939203958577e-05, "loss": 0.5518, "step": 12845 }, { "epoch": 3.95, "learning_rate": 1.378801694696999e-05, "loss": 0.6415, "step": 12846 }, { "epoch": 3.95, "learning_rate": 1.378709465236291e-05, "loss": 0.6113, "step": 12847 }, { "epoch": 3.95, "learning_rate": 1.3786172320146484e-05, "loss": 0.4035, "step": 12848 }, { "epoch": 3.95, "learning_rate": 1.378524995032988e-05, "loss": 0.5519, "step": 12849 }, { "epoch": 3.95, "learning_rate": 1.3784327542922258e-05, "loss": 0.6104, "step": 12850 }, { "epoch": 3.95, "learning_rate": 1.3783405097932772e-05, "loss": 0.7691, "step": 12851 }, { "epoch": 3.95, "learning_rate": 1.3782482615370584e-05, "loss": 0.6588, "step": 12852 }, { "epoch": 3.95, "learning_rate": 1.3781560095244863e-05, "loss": 0.367, "step": 12853 }, { "epoch": 3.95, "learning_rate": 1.3780637537564763e-05, "loss": 0.4101, "step": 12854 }, { "epoch": 3.96, "learning_rate": 1.3779714942339446e-05, "loss": 0.5398, "step": 12855 }, { "epoch": 3.96, "learning_rate": 1.377879230957808e-05, "loss": 0.3625, "step": 12856 }, { "epoch": 3.96, "learning_rate": 1.3777869639289822e-05, "loss": 0.5705, "step": 12857 }, { "epoch": 3.96, "learning_rate": 1.3776946931483834e-05, "loss": 0.7352, "step": 12858 }, { "epoch": 3.96, "learning_rate": 1.3776024186169286e-05, "loss": 0.6971, "step": 12859 }, { "epoch": 3.96, "learning_rate": 1.3775101403355337e-05, "loss": 0.5171, "step": 12860 }, { "epoch": 3.96, "learning_rate": 1.3774178583051152e-05, "loss": 0.4474, "step": 12861 }, { "epoch": 3.96, "learning_rate": 1.3773255725265896e-05, "loss": 0.3886, "step": 12862 }, { "epoch": 3.96, "learning_rate": 1.3772332830008728e-05, "loss": 0.7143, "step": 12863 }, { "epoch": 3.96, "learning_rate": 1.3771409897288823e-05, "loss": 0.5192, "step": 12864 }, { "epoch": 3.96, "learning_rate": 1.3770486927115342e-05, "loss": 0.4483, "step": 12865 }, { "epoch": 3.96, "learning_rate": 1.376956391949745e-05, "loss": 0.5345, "step": 12866 }, { "epoch": 3.96, "learning_rate": 1.3768640874444312e-05, "loss": 0.579, "step": 12867 }, { "epoch": 3.96, "learning_rate": 1.3767717791965098e-05, "loss": 0.6501, "step": 12868 }, { "epoch": 3.96, "learning_rate": 1.3766794672068975e-05, "loss": 0.5085, "step": 12869 }, { "epoch": 3.96, "learning_rate": 1.3765871514765107e-05, "loss": 0.4182, "step": 12870 }, { "epoch": 3.96, "learning_rate": 1.3764948320062667e-05, "loss": 0.4666, "step": 12871 }, { "epoch": 3.96, "learning_rate": 1.376402508797082e-05, "loss": 0.4218, "step": 12872 }, { "epoch": 3.96, "learning_rate": 1.3763101818498735e-05, "loss": 0.4296, "step": 12873 }, { "epoch": 3.96, "learning_rate": 1.376217851165558e-05, "loss": 0.8346, "step": 12874 }, { "epoch": 3.96, "learning_rate": 1.3761255167450523e-05, "loss": 0.5309, "step": 12875 }, { "epoch": 3.96, "learning_rate": 1.3760331785892738e-05, "loss": 0.4698, "step": 12876 }, { "epoch": 3.96, "learning_rate": 1.3759408366991391e-05, "loss": 0.3088, "step": 12877 }, { "epoch": 3.96, "learning_rate": 1.3758484910755656e-05, "loss": 0.5651, "step": 12878 }, { "epoch": 3.96, "learning_rate": 1.3757561417194703e-05, "loss": 0.5247, "step": 12879 }, { "epoch": 3.96, "learning_rate": 1.37566378863177e-05, "loss": 0.5309, "step": 12880 }, { "epoch": 3.96, "learning_rate": 1.3755714318133821e-05, "loss": 0.5016, "step": 12881 }, { "epoch": 3.96, "learning_rate": 1.3754790712652238e-05, "loss": 0.4429, "step": 12882 }, { "epoch": 3.96, "learning_rate": 1.3753867069882121e-05, "loss": 0.5484, "step": 12883 }, { "epoch": 3.96, "learning_rate": 1.3752943389832648e-05, "loss": 0.6022, "step": 12884 }, { "epoch": 3.96, "learning_rate": 1.3752019672512981e-05, "loss": 0.5776, "step": 12885 }, { "epoch": 3.96, "learning_rate": 1.375109591793231e-05, "loss": 0.2989, "step": 12886 }, { "epoch": 3.97, "learning_rate": 1.3750172126099791e-05, "loss": 0.4817, "step": 12887 }, { "epoch": 3.97, "learning_rate": 1.3749248297024614e-05, "loss": 0.6834, "step": 12888 }, { "epoch": 3.97, "learning_rate": 1.3748324430715941e-05, "loss": 0.6485, "step": 12889 }, { "epoch": 3.97, "learning_rate": 1.3747400527182952e-05, "loss": 0.4998, "step": 12890 }, { "epoch": 3.97, "learning_rate": 1.3746476586434825e-05, "loss": 0.4439, "step": 12891 }, { "epoch": 3.97, "learning_rate": 1.374555260848073e-05, "loss": 0.4884, "step": 12892 }, { "epoch": 3.97, "learning_rate": 1.3744628593329848e-05, "loss": 0.7249, "step": 12893 }, { "epoch": 3.97, "learning_rate": 1.3743704540991349e-05, "loss": 0.5365, "step": 12894 }, { "epoch": 3.97, "learning_rate": 1.3742780451474418e-05, "loss": 0.5601, "step": 12895 }, { "epoch": 3.97, "learning_rate": 1.3741856324788225e-05, "loss": 0.4713, "step": 12896 }, { "epoch": 3.97, "learning_rate": 1.374093216094195e-05, "loss": 0.4545, "step": 12897 }, { "epoch": 3.97, "learning_rate": 1.3740007959944775e-05, "loss": 0.5317, "step": 12898 }, { "epoch": 3.97, "learning_rate": 1.3739083721805869e-05, "loss": 0.5003, "step": 12899 }, { "epoch": 3.97, "learning_rate": 1.3738159446534417e-05, "loss": 0.6844, "step": 12900 }, { "epoch": 3.97, "learning_rate": 1.3737235134139597e-05, "loss": 0.4132, "step": 12901 }, { "epoch": 3.97, "learning_rate": 1.3736310784630588e-05, "loss": 0.5118, "step": 12902 }, { "epoch": 3.97, "learning_rate": 1.373538639801657e-05, "loss": 0.6359, "step": 12903 }, { "epoch": 3.97, "learning_rate": 1.373446197430672e-05, "loss": 0.5971, "step": 12904 }, { "epoch": 3.97, "learning_rate": 1.3733537513510222e-05, "loss": 0.5752, "step": 12905 }, { "epoch": 3.97, "learning_rate": 1.3732613015636259e-05, "loss": 0.5005, "step": 12906 }, { "epoch": 3.97, "learning_rate": 1.3731688480694006e-05, "loss": 0.6425, "step": 12907 }, { "epoch": 3.97, "learning_rate": 1.3730763908692644e-05, "loss": 0.5278, "step": 12908 }, { "epoch": 3.97, "learning_rate": 1.3729839299641361e-05, "loss": 0.4266, "step": 12909 }, { "epoch": 3.97, "learning_rate": 1.3728914653549336e-05, "loss": 0.4988, "step": 12910 }, { "epoch": 3.97, "learning_rate": 1.3727989970425752e-05, "loss": 0.4152, "step": 12911 }, { "epoch": 3.97, "learning_rate": 1.3727065250279795e-05, "loss": 0.6594, "step": 12912 }, { "epoch": 3.97, "learning_rate": 1.3726140493120639e-05, "loss": 0.3451, "step": 12913 }, { "epoch": 3.97, "learning_rate": 1.3725215698957476e-05, "loss": 0.618, "step": 12914 }, { "epoch": 3.97, "learning_rate": 1.372429086779949e-05, "loss": 0.5231, "step": 12915 }, { "epoch": 3.97, "learning_rate": 1.3723365999655859e-05, "loss": 0.8543, "step": 12916 }, { "epoch": 3.97, "learning_rate": 1.3722441094535777e-05, "loss": 0.6478, "step": 12917 }, { "epoch": 3.97, "learning_rate": 1.372151615244842e-05, "loss": 0.4251, "step": 12918 }, { "epoch": 3.97, "learning_rate": 1.3720591173402981e-05, "loss": 0.688, "step": 12919 }, { "epoch": 3.98, "learning_rate": 1.3719666157408643e-05, "loss": 0.3071, "step": 12920 }, { "epoch": 3.98, "learning_rate": 1.3718741104474589e-05, "loss": 0.5755, "step": 12921 }, { "epoch": 3.98, "learning_rate": 1.371781601461001e-05, "loss": 0.5422, "step": 12922 }, { "epoch": 3.98, "learning_rate": 1.371689088782409e-05, "loss": 0.6768, "step": 12923 }, { "epoch": 3.98, "learning_rate": 1.3715965724126022e-05, "loss": 0.6094, "step": 12924 }, { "epoch": 3.98, "learning_rate": 1.3715040523524982e-05, "loss": 0.6568, "step": 12925 }, { "epoch": 3.98, "learning_rate": 1.3714115286030174e-05, "loss": 0.5307, "step": 12926 }, { "epoch": 3.98, "learning_rate": 1.3713190011650774e-05, "loss": 0.4236, "step": 12927 }, { "epoch": 3.98, "learning_rate": 1.3712264700395976e-05, "loss": 0.6766, "step": 12928 }, { "epoch": 3.98, "learning_rate": 1.3711339352274969e-05, "loss": 0.5397, "step": 12929 }, { "epoch": 3.98, "learning_rate": 1.3710413967296939e-05, "loss": 0.5302, "step": 12930 }, { "epoch": 3.98, "learning_rate": 1.3709488545471082e-05, "loss": 0.4814, "step": 12931 }, { "epoch": 3.98, "learning_rate": 1.3708563086806583e-05, "loss": 0.6343, "step": 12932 }, { "epoch": 3.98, "learning_rate": 1.3707637591312635e-05, "loss": 0.638, "step": 12933 }, { "epoch": 3.98, "learning_rate": 1.3706712058998428e-05, "loss": 0.6932, "step": 12934 }, { "epoch": 3.98, "learning_rate": 1.3705786489873153e-05, "loss": 0.6621, "step": 12935 }, { "epoch": 3.98, "learning_rate": 1.3704860883946007e-05, "loss": 0.6392, "step": 12936 }, { "epoch": 3.98, "learning_rate": 1.3703935241226174e-05, "loss": 0.726, "step": 12937 }, { "epoch": 3.98, "learning_rate": 1.3703009561722854e-05, "loss": 0.6029, "step": 12938 }, { "epoch": 3.98, "learning_rate": 1.3702083845445234e-05, "loss": 0.62, "step": 12939 }, { "epoch": 3.98, "learning_rate": 1.3701158092402506e-05, "loss": 0.4896, "step": 12940 }, { "epoch": 3.98, "learning_rate": 1.3700232302603875e-05, "loss": 0.4097, "step": 12941 }, { "epoch": 3.98, "learning_rate": 1.3699306476058523e-05, "loss": 0.5602, "step": 12942 }, { "epoch": 3.98, "learning_rate": 1.3698380612775646e-05, "loss": 0.6299, "step": 12943 }, { "epoch": 3.98, "learning_rate": 1.3697454712764442e-05, "loss": 0.3561, "step": 12944 }, { "epoch": 3.98, "learning_rate": 1.3696528776034105e-05, "loss": 0.486, "step": 12945 }, { "epoch": 3.98, "learning_rate": 1.3695602802593833e-05, "loss": 0.6125, "step": 12946 }, { "epoch": 3.98, "learning_rate": 1.369467679245282e-05, "loss": 0.3706, "step": 12947 }, { "epoch": 3.98, "learning_rate": 1.3693750745620258e-05, "loss": 0.6587, "step": 12948 }, { "epoch": 3.98, "learning_rate": 1.3692824662105346e-05, "loss": 0.3552, "step": 12949 }, { "epoch": 3.98, "learning_rate": 1.3691898541917283e-05, "loss": 0.5447, "step": 12950 }, { "epoch": 3.98, "learning_rate": 1.3690972385065267e-05, "loss": 0.5461, "step": 12951 }, { "epoch": 3.99, "learning_rate": 1.3690046191558492e-05, "loss": 0.4572, "step": 12952 }, { "epoch": 3.99, "learning_rate": 1.3689119961406159e-05, "loss": 0.3358, "step": 12953 }, { "epoch": 3.99, "learning_rate": 1.368819369461746e-05, "loss": 0.6458, "step": 12954 }, { "epoch": 3.99, "learning_rate": 1.3687267391201604e-05, "loss": 0.6518, "step": 12955 }, { "epoch": 3.99, "learning_rate": 1.368634105116778e-05, "loss": 0.4634, "step": 12956 }, { "epoch": 3.99, "learning_rate": 1.3685414674525197e-05, "loss": 0.5406, "step": 12957 }, { "epoch": 3.99, "learning_rate": 1.3684488261283045e-05, "loss": 0.6523, "step": 12958 }, { "epoch": 3.99, "learning_rate": 1.3683561811450532e-05, "loss": 0.4772, "step": 12959 }, { "epoch": 3.99, "learning_rate": 1.3682635325036853e-05, "loss": 0.401, "step": 12960 }, { "epoch": 3.99, "learning_rate": 1.368170880205121e-05, "loss": 0.4441, "step": 12961 }, { "epoch": 3.99, "learning_rate": 1.368078224250281e-05, "loss": 0.4682, "step": 12962 }, { "epoch": 3.99, "learning_rate": 1.3679855646400845e-05, "loss": 0.602, "step": 12963 }, { "epoch": 3.99, "learning_rate": 1.3678929013754527e-05, "loss": 0.5882, "step": 12964 }, { "epoch": 3.99, "learning_rate": 1.3678002344573051e-05, "loss": 0.4062, "step": 12965 }, { "epoch": 3.99, "learning_rate": 1.367707563886562e-05, "loss": 0.5379, "step": 12966 }, { "epoch": 3.99, "learning_rate": 1.3676148896641443e-05, "loss": 0.4701, "step": 12967 }, { "epoch": 3.99, "learning_rate": 1.3675222117909716e-05, "loss": 0.4071, "step": 12968 }, { "epoch": 3.99, "learning_rate": 1.3674295302679648e-05, "loss": 0.5059, "step": 12969 }, { "epoch": 3.99, "learning_rate": 1.3673368450960442e-05, "loss": 0.4499, "step": 12970 }, { "epoch": 3.99, "learning_rate": 1.3672441562761297e-05, "loss": 0.4025, "step": 12971 }, { "epoch": 3.99, "learning_rate": 1.367151463809143e-05, "loss": 0.3264, "step": 12972 }, { "epoch": 3.99, "learning_rate": 1.3670587676960035e-05, "loss": 0.6066, "step": 12973 }, { "epoch": 3.99, "learning_rate": 1.3669660679376322e-05, "loss": 0.5222, "step": 12974 }, { "epoch": 3.99, "learning_rate": 1.3668733645349495e-05, "loss": 0.4373, "step": 12975 }, { "epoch": 3.99, "learning_rate": 1.3667806574888762e-05, "loss": 0.6509, "step": 12976 }, { "epoch": 3.99, "learning_rate": 1.366687946800333e-05, "loss": 0.7483, "step": 12977 }, { "epoch": 3.99, "learning_rate": 1.3665952324702408e-05, "loss": 0.5712, "step": 12978 }, { "epoch": 3.99, "learning_rate": 1.3665025144995197e-05, "loss": 0.4891, "step": 12979 }, { "epoch": 3.99, "learning_rate": 1.366409792889091e-05, "loss": 0.4502, "step": 12980 }, { "epoch": 3.99, "learning_rate": 1.366317067639875e-05, "loss": 0.5393, "step": 12981 }, { "epoch": 3.99, "learning_rate": 1.3662243387527936e-05, "loss": 0.4733, "step": 12982 }, { "epoch": 3.99, "learning_rate": 1.3661316062287667e-05, "loss": 0.5377, "step": 12983 }, { "epoch": 3.99, "learning_rate": 1.3660388700687154e-05, "loss": 0.6184, "step": 12984 }, { "epoch": 4.0, "learning_rate": 1.3659461302735606e-05, "loss": 0.7913, "step": 12985 }, { "epoch": 4.0, "learning_rate": 1.3658533868442239e-05, "loss": 0.6055, "step": 12986 }, { "epoch": 4.0, "learning_rate": 1.3657606397816257e-05, "loss": 0.5064, "step": 12987 }, { "epoch": 4.0, "learning_rate": 1.3656678890866872e-05, "loss": 0.4947, "step": 12988 }, { "epoch": 4.0, "learning_rate": 1.3655751347603294e-05, "loss": 0.7325, "step": 12989 }, { "epoch": 4.0, "learning_rate": 1.365482376803474e-05, "loss": 0.5666, "step": 12990 }, { "epoch": 4.0, "learning_rate": 1.3653896152170413e-05, "loss": 0.5081, "step": 12991 }, { "epoch": 4.0, "learning_rate": 1.3652968500019534e-05, "loss": 0.4308, "step": 12992 }, { "epoch": 4.0, "learning_rate": 1.365204081159131e-05, "loss": 0.514, "step": 12993 }, { "epoch": 4.0, "learning_rate": 1.3651113086894951e-05, "loss": 0.5559, "step": 12994 }, { "epoch": 4.0, "learning_rate": 1.3650185325939678e-05, "loss": 0.5221, "step": 12995 }, { "epoch": 4.0, "learning_rate": 1.36492575287347e-05, "loss": 0.4673, "step": 12996 }, { "epoch": 4.0, "learning_rate": 1.3648329695289233e-05, "loss": 0.6641, "step": 12997 }, { "epoch": 4.0, "learning_rate": 1.3647401825612487e-05, "loss": 0.6723, "step": 12998 }, { "epoch": 4.0, "learning_rate": 1.3646473919713678e-05, "loss": 0.5391, "step": 12999 }, { "epoch": 4.0, "learning_rate": 1.3645545977602026e-05, "loss": 0.5467, "step": 13000 }, { "epoch": 4.0, "learning_rate": 1.3644617999286743e-05, "loss": 0.3947, "step": 13001 }, { "epoch": 4.0, "learning_rate": 1.3643689984777041e-05, "loss": 0.3178, "step": 13002 }, { "epoch": 4.0, "learning_rate": 1.3642761934082143e-05, "loss": 0.2337, "step": 13003 }, { "epoch": 4.0, "learning_rate": 1.3641833847211257e-05, "loss": 0.2187, "step": 13004 }, { "epoch": 4.0, "learning_rate": 1.3640905724173607e-05, "loss": 0.1926, "step": 13005 }, { "epoch": 4.0, "learning_rate": 1.3639977564978406e-05, "loss": 0.3616, "step": 13006 }, { "epoch": 4.0, "learning_rate": 1.3639049369634878e-05, "loss": 0.3491, "step": 13007 }, { "epoch": 4.0, "learning_rate": 1.3638121138152229e-05, "loss": 0.3945, "step": 13008 }, { "epoch": 4.0, "learning_rate": 1.3637192870539689e-05, "loss": 0.2514, "step": 13009 }, { "epoch": 4.0, "learning_rate": 1.3636264566806473e-05, "loss": 0.2713, "step": 13010 }, { "epoch": 4.0, "learning_rate": 1.3635336226961794e-05, "loss": 0.2823, "step": 13011 }, { "epoch": 4.0, "learning_rate": 1.363440785101488e-05, "loss": 0.2983, "step": 13012 }, { "epoch": 4.0, "learning_rate": 1.3633479438974942e-05, "loss": 0.3723, "step": 13013 }, { "epoch": 4.0, "learning_rate": 1.3632550990851207e-05, "loss": 0.3216, "step": 13014 }, { "epoch": 4.0, "learning_rate": 1.3631622506652893e-05, "loss": 0.3395, "step": 13015 }, { "epoch": 4.0, "learning_rate": 1.3630693986389219e-05, "loss": 0.2874, "step": 13016 }, { "epoch": 4.01, "learning_rate": 1.362976543006941e-05, "loss": 0.2195, "step": 13017 }, { "epoch": 4.01, "learning_rate": 1.3628836837702682e-05, "loss": 0.3161, "step": 13018 }, { "epoch": 4.01, "learning_rate": 1.3627908209298262e-05, "loss": 0.4471, "step": 13019 }, { "epoch": 4.01, "learning_rate": 1.3626979544865369e-05, "loss": 0.2827, "step": 13020 }, { "epoch": 4.01, "learning_rate": 1.3626050844413225e-05, "loss": 0.3907, "step": 13021 }, { "epoch": 4.01, "learning_rate": 1.3625122107951059e-05, "loss": 0.1423, "step": 13022 }, { "epoch": 4.01, "learning_rate": 1.3624193335488083e-05, "loss": 0.2409, "step": 13023 }, { "epoch": 4.01, "learning_rate": 1.3623264527033531e-05, "loss": 0.355, "step": 13024 }, { "epoch": 4.01, "learning_rate": 1.362233568259662e-05, "loss": 0.2971, "step": 13025 }, { "epoch": 4.01, "learning_rate": 1.362140680218658e-05, "loss": 0.1644, "step": 13026 }, { "epoch": 4.01, "learning_rate": 1.3620477885812635e-05, "loss": 0.2881, "step": 13027 }, { "epoch": 4.01, "learning_rate": 1.3619548933484007e-05, "loss": 0.2955, "step": 13028 }, { "epoch": 4.01, "learning_rate": 1.361861994520992e-05, "loss": 0.2847, "step": 13029 }, { "epoch": 4.01, "learning_rate": 1.36176909209996e-05, "loss": 0.1876, "step": 13030 }, { "epoch": 4.01, "learning_rate": 1.3616761860862278e-05, "loss": 0.2575, "step": 13031 }, { "epoch": 4.01, "learning_rate": 1.3615832764807178e-05, "loss": 0.4391, "step": 13032 }, { "epoch": 4.01, "learning_rate": 1.3614903632843523e-05, "loss": 0.2364, "step": 13033 }, { "epoch": 4.01, "learning_rate": 1.3613974464980548e-05, "loss": 0.4522, "step": 13034 }, { "epoch": 4.01, "learning_rate": 1.3613045261227471e-05, "loss": 0.2518, "step": 13035 }, { "epoch": 4.01, "learning_rate": 1.3612116021593526e-05, "loss": 0.2867, "step": 13036 }, { "epoch": 4.01, "learning_rate": 1.3611186746087941e-05, "loss": 0.2826, "step": 13037 }, { "epoch": 4.01, "learning_rate": 1.3610257434719944e-05, "loss": 0.2797, "step": 13038 }, { "epoch": 4.01, "learning_rate": 1.3609328087498763e-05, "loss": 0.1676, "step": 13039 }, { "epoch": 4.01, "learning_rate": 1.3608398704433625e-05, "loss": 0.2378, "step": 13040 }, { "epoch": 4.01, "learning_rate": 1.3607469285533765e-05, "loss": 0.2754, "step": 13041 }, { "epoch": 4.01, "learning_rate": 1.360653983080841e-05, "loss": 0.3913, "step": 13042 }, { "epoch": 4.01, "learning_rate": 1.360561034026679e-05, "loss": 0.3554, "step": 13043 }, { "epoch": 4.01, "learning_rate": 1.3604680813918136e-05, "loss": 0.3303, "step": 13044 }, { "epoch": 4.01, "learning_rate": 1.3603751251771681e-05, "loss": 0.2851, "step": 13045 }, { "epoch": 4.01, "learning_rate": 1.3602821653836654e-05, "loss": 0.1918, "step": 13046 }, { "epoch": 4.01, "learning_rate": 1.3601892020122287e-05, "loss": 0.3213, "step": 13047 }, { "epoch": 4.01, "learning_rate": 1.3600962350637815e-05, "loss": 0.1863, "step": 13048 }, { "epoch": 4.01, "learning_rate": 1.3600032645392463e-05, "loss": 0.3998, "step": 13049 }, { "epoch": 4.02, "learning_rate": 1.3599102904395474e-05, "loss": 0.4542, "step": 13050 }, { "epoch": 4.02, "learning_rate": 1.3598173127656074e-05, "loss": 0.2726, "step": 13051 }, { "epoch": 4.02, "learning_rate": 1.3597243315183498e-05, "loss": 0.2703, "step": 13052 }, { "epoch": 4.02, "learning_rate": 1.3596313466986982e-05, "loss": 0.2537, "step": 13053 }, { "epoch": 4.02, "learning_rate": 1.3595383583075757e-05, "loss": 0.2253, "step": 13054 }, { "epoch": 4.02, "learning_rate": 1.3594453663459059e-05, "loss": 0.3646, "step": 13055 }, { "epoch": 4.02, "learning_rate": 1.3593523708146123e-05, "loss": 0.2475, "step": 13056 }, { "epoch": 4.02, "learning_rate": 1.3592593717146188e-05, "loss": 0.2329, "step": 13057 }, { "epoch": 4.02, "learning_rate": 1.3591663690468486e-05, "loss": 0.2112, "step": 13058 }, { "epoch": 4.02, "learning_rate": 1.3590733628122253e-05, "loss": 0.347, "step": 13059 }, { "epoch": 4.02, "learning_rate": 1.3589803530116725e-05, "loss": 0.4971, "step": 13060 }, { "epoch": 4.02, "learning_rate": 1.3588873396461138e-05, "loss": 0.3146, "step": 13061 }, { "epoch": 4.02, "learning_rate": 1.3587943227164733e-05, "loss": 0.2781, "step": 13062 }, { "epoch": 4.02, "learning_rate": 1.3587013022236744e-05, "loss": 0.3566, "step": 13063 }, { "epoch": 4.02, "learning_rate": 1.358608278168641e-05, "loss": 0.2916, "step": 13064 }, { "epoch": 4.02, "learning_rate": 1.358515250552297e-05, "loss": 0.318, "step": 13065 }, { "epoch": 4.02, "learning_rate": 1.3584222193755659e-05, "loss": 0.275, "step": 13066 }, { "epoch": 4.02, "learning_rate": 1.3583291846393718e-05, "loss": 0.286, "step": 13067 }, { "epoch": 4.02, "learning_rate": 1.358236146344639e-05, "loss": 0.4723, "step": 13068 }, { "epoch": 4.02, "learning_rate": 1.3581431044922907e-05, "loss": 0.2333, "step": 13069 }, { "epoch": 4.02, "learning_rate": 1.3580500590832518e-05, "loss": 0.3334, "step": 13070 }, { "epoch": 4.02, "learning_rate": 1.3579570101184454e-05, "loss": 0.2573, "step": 13071 }, { "epoch": 4.02, "learning_rate": 1.357863957598796e-05, "loss": 0.2845, "step": 13072 }, { "epoch": 4.02, "learning_rate": 1.3577709015252278e-05, "loss": 0.2746, "step": 13073 }, { "epoch": 4.02, "learning_rate": 1.357677841898665e-05, "loss": 0.2189, "step": 13074 }, { "epoch": 4.02, "learning_rate": 1.3575847787200313e-05, "loss": 0.3117, "step": 13075 }, { "epoch": 4.02, "learning_rate": 1.3574917119902514e-05, "loss": 0.3626, "step": 13076 }, { "epoch": 4.02, "learning_rate": 1.3573986417102492e-05, "loss": 0.4281, "step": 13077 }, { "epoch": 4.02, "learning_rate": 1.3573055678809493e-05, "loss": 0.3055, "step": 13078 }, { "epoch": 4.02, "learning_rate": 1.3572124905032755e-05, "loss": 0.2845, "step": 13079 }, { "epoch": 4.02, "learning_rate": 1.3571194095781523e-05, "loss": 0.2632, "step": 13080 }, { "epoch": 4.02, "learning_rate": 1.3570263251065048e-05, "loss": 0.3092, "step": 13081 }, { "epoch": 4.03, "learning_rate": 1.3569332370892565e-05, "loss": 0.3199, "step": 13082 }, { "epoch": 4.03, "learning_rate": 1.3568401455273324e-05, "loss": 0.3287, "step": 13083 }, { "epoch": 4.03, "learning_rate": 1.3567470504216567e-05, "loss": 0.2935, "step": 13084 }, { "epoch": 4.03, "learning_rate": 1.3566539517731536e-05, "loss": 0.3414, "step": 13085 }, { "epoch": 4.03, "learning_rate": 1.3565608495827484e-05, "loss": 0.386, "step": 13086 }, { "epoch": 4.03, "learning_rate": 1.3564677438513654e-05, "loss": 0.2666, "step": 13087 }, { "epoch": 4.03, "learning_rate": 1.3563746345799294e-05, "loss": 0.2382, "step": 13088 }, { "epoch": 4.03, "learning_rate": 1.3562815217693643e-05, "loss": 0.1669, "step": 13089 }, { "epoch": 4.03, "learning_rate": 1.3561884054205956e-05, "loss": 0.2519, "step": 13090 }, { "epoch": 4.03, "learning_rate": 1.3560952855345478e-05, "loss": 0.3311, "step": 13091 }, { "epoch": 4.03, "learning_rate": 1.3560021621121454e-05, "loss": 0.2393, "step": 13092 }, { "epoch": 4.03, "learning_rate": 1.3559090351543138e-05, "loss": 0.275, "step": 13093 }, { "epoch": 4.03, "learning_rate": 1.3558159046619771e-05, "loss": 0.2807, "step": 13094 }, { "epoch": 4.03, "learning_rate": 1.3557227706360607e-05, "loss": 0.151, "step": 13095 }, { "epoch": 4.03, "learning_rate": 1.3556296330774894e-05, "loss": 0.163, "step": 13096 }, { "epoch": 4.03, "learning_rate": 1.3555364919871879e-05, "loss": 0.2132, "step": 13097 }, { "epoch": 4.03, "learning_rate": 1.3554433473660818e-05, "loss": 0.2719, "step": 13098 }, { "epoch": 4.03, "learning_rate": 1.3553501992150953e-05, "loss": 0.3121, "step": 13099 }, { "epoch": 4.03, "learning_rate": 1.3552570475351537e-05, "loss": 0.1192, "step": 13100 }, { "epoch": 4.03, "learning_rate": 1.3551638923271826e-05, "loss": 0.2124, "step": 13101 }, { "epoch": 4.03, "learning_rate": 1.3550707335921063e-05, "loss": 0.3908, "step": 13102 }, { "epoch": 4.03, "learning_rate": 1.3549775713308509e-05, "loss": 0.3538, "step": 13103 }, { "epoch": 4.03, "learning_rate": 1.3548844055443404e-05, "loss": 0.2547, "step": 13104 }, { "epoch": 4.03, "learning_rate": 1.3547912362335012e-05, "loss": 0.2239, "step": 13105 }, { "epoch": 4.03, "learning_rate": 1.3546980633992577e-05, "loss": 0.3669, "step": 13106 }, { "epoch": 4.03, "learning_rate": 1.3546048870425356e-05, "loss": 0.2678, "step": 13107 }, { "epoch": 4.03, "learning_rate": 1.3545117071642604e-05, "loss": 0.2521, "step": 13108 }, { "epoch": 4.03, "learning_rate": 1.3544185237653571e-05, "loss": 0.2614, "step": 13109 }, { "epoch": 4.03, "learning_rate": 1.3543253368467512e-05, "loss": 0.2911, "step": 13110 }, { "epoch": 4.03, "learning_rate": 1.354232146409368e-05, "loss": 0.1823, "step": 13111 }, { "epoch": 4.03, "learning_rate": 1.3541389524541333e-05, "loss": 0.3511, "step": 13112 }, { "epoch": 4.03, "learning_rate": 1.3540457549819724e-05, "loss": 0.2781, "step": 13113 }, { "epoch": 4.03, "learning_rate": 1.353952553993811e-05, "loss": 0.2838, "step": 13114 }, { "epoch": 4.04, "learning_rate": 1.3538593494905742e-05, "loss": 0.2688, "step": 13115 }, { "epoch": 4.04, "learning_rate": 1.3537661414731882e-05, "loss": 0.3635, "step": 13116 }, { "epoch": 4.04, "learning_rate": 1.3536729299425782e-05, "loss": 0.2467, "step": 13117 }, { "epoch": 4.04, "learning_rate": 1.35357971489967e-05, "loss": 0.2558, "step": 13118 }, { "epoch": 4.04, "learning_rate": 1.3534864963453895e-05, "loss": 0.2583, "step": 13119 }, { "epoch": 4.04, "learning_rate": 1.3533932742806626e-05, "loss": 0.6161, "step": 13120 }, { "epoch": 4.04, "learning_rate": 1.3533000487064144e-05, "loss": 0.275, "step": 13121 }, { "epoch": 4.04, "learning_rate": 1.353206819623571e-05, "loss": 0.4283, "step": 13122 }, { "epoch": 4.04, "learning_rate": 1.3531135870330587e-05, "loss": 0.2717, "step": 13123 }, { "epoch": 4.04, "learning_rate": 1.353020350935803e-05, "loss": 0.2196, "step": 13124 }, { "epoch": 4.04, "learning_rate": 1.3529271113327294e-05, "loss": 0.3289, "step": 13125 }, { "epoch": 4.04, "learning_rate": 1.3528338682247649e-05, "loss": 0.3929, "step": 13126 }, { "epoch": 4.04, "learning_rate": 1.3527406216128344e-05, "loss": 0.3424, "step": 13127 }, { "epoch": 4.04, "learning_rate": 1.352647371497865e-05, "loss": 0.2597, "step": 13128 }, { "epoch": 4.04, "learning_rate": 1.3525541178807817e-05, "loss": 0.2511, "step": 13129 }, { "epoch": 4.04, "learning_rate": 1.3524608607625113e-05, "loss": 0.2361, "step": 13130 }, { "epoch": 4.04, "learning_rate": 1.3523676001439797e-05, "loss": 0.2392, "step": 13131 }, { "epoch": 4.04, "learning_rate": 1.352274336026113e-05, "loss": 0.3233, "step": 13132 }, { "epoch": 4.04, "learning_rate": 1.3521810684098378e-05, "loss": 0.3859, "step": 13133 }, { "epoch": 4.04, "learning_rate": 1.3520877972960797e-05, "loss": 0.2314, "step": 13134 }, { "epoch": 4.04, "learning_rate": 1.3519945226857652e-05, "loss": 0.2846, "step": 13135 }, { "epoch": 4.04, "learning_rate": 1.3519012445798207e-05, "loss": 0.2436, "step": 13136 }, { "epoch": 4.04, "learning_rate": 1.3518079629791725e-05, "loss": 0.1954, "step": 13137 }, { "epoch": 4.04, "learning_rate": 1.3517146778847475e-05, "loss": 0.2682, "step": 13138 }, { "epoch": 4.04, "learning_rate": 1.3516213892974712e-05, "loss": 0.3248, "step": 13139 }, { "epoch": 4.04, "learning_rate": 1.3515280972182705e-05, "loss": 0.2002, "step": 13140 }, { "epoch": 4.04, "learning_rate": 1.3514348016480718e-05, "loss": 0.4658, "step": 13141 }, { "epoch": 4.04, "learning_rate": 1.3513415025878015e-05, "loss": 0.157, "step": 13142 }, { "epoch": 4.04, "learning_rate": 1.3512482000383864e-05, "loss": 0.2937, "step": 13143 }, { "epoch": 4.04, "learning_rate": 1.3511548940007528e-05, "loss": 0.2827, "step": 13144 }, { "epoch": 4.04, "learning_rate": 1.3510615844758278e-05, "loss": 0.1844, "step": 13145 }, { "epoch": 4.04, "learning_rate": 1.3509682714645375e-05, "loss": 0.2672, "step": 13146 }, { "epoch": 4.05, "learning_rate": 1.3508749549678086e-05, "loss": 0.2653, "step": 13147 }, { "epoch": 4.05, "learning_rate": 1.3507816349865685e-05, "loss": 0.2476, "step": 13148 }, { "epoch": 4.05, "learning_rate": 1.3506883115217431e-05, "loss": 0.2988, "step": 13149 }, { "epoch": 4.05, "learning_rate": 1.3505949845742599e-05, "loss": 0.2587, "step": 13150 }, { "epoch": 4.05, "learning_rate": 1.350501654145045e-05, "loss": 0.293, "step": 13151 }, { "epoch": 4.05, "learning_rate": 1.3504083202350257e-05, "loss": 0.3782, "step": 13152 }, { "epoch": 4.05, "learning_rate": 1.350314982845129e-05, "loss": 0.353, "step": 13153 }, { "epoch": 4.05, "learning_rate": 1.3502216419762816e-05, "loss": 0.3891, "step": 13154 }, { "epoch": 4.05, "learning_rate": 1.3501282976294108e-05, "loss": 0.2706, "step": 13155 }, { "epoch": 4.05, "learning_rate": 1.3500349498054429e-05, "loss": 0.207, "step": 13156 }, { "epoch": 4.05, "learning_rate": 1.3499415985053055e-05, "loss": 0.2822, "step": 13157 }, { "epoch": 4.05, "learning_rate": 1.3498482437299254e-05, "loss": 0.2611, "step": 13158 }, { "epoch": 4.05, "learning_rate": 1.34975488548023e-05, "loss": 0.3941, "step": 13159 }, { "epoch": 4.05, "learning_rate": 1.3496615237571464e-05, "loss": 0.2129, "step": 13160 }, { "epoch": 4.05, "learning_rate": 1.3495681585616011e-05, "loss": 0.3612, "step": 13161 }, { "epoch": 4.05, "learning_rate": 1.3494747898945221e-05, "loss": 0.2925, "step": 13162 }, { "epoch": 4.05, "learning_rate": 1.3493814177568365e-05, "loss": 0.4187, "step": 13163 }, { "epoch": 4.05, "learning_rate": 1.3492880421494712e-05, "loss": 0.2831, "step": 13164 }, { "epoch": 4.05, "learning_rate": 1.349194663073354e-05, "loss": 0.2357, "step": 13165 }, { "epoch": 4.05, "learning_rate": 1.3491012805294114e-05, "loss": 0.2164, "step": 13166 }, { "epoch": 4.05, "learning_rate": 1.3490078945185719e-05, "loss": 0.1827, "step": 13167 }, { "epoch": 4.05, "learning_rate": 1.3489145050417621e-05, "loss": 0.2367, "step": 13168 }, { "epoch": 4.05, "learning_rate": 1.3488211120999098e-05, "loss": 0.3448, "step": 13169 }, { "epoch": 4.05, "learning_rate": 1.3487277156939424e-05, "loss": 0.3925, "step": 13170 }, { "epoch": 4.05, "learning_rate": 1.348634315824787e-05, "loss": 0.3158, "step": 13171 }, { "epoch": 4.05, "learning_rate": 1.348540912493372e-05, "loss": 0.3461, "step": 13172 }, { "epoch": 4.05, "learning_rate": 1.3484475057006243e-05, "loss": 0.1738, "step": 13173 }, { "epoch": 4.05, "learning_rate": 1.348354095447472e-05, "loss": 0.2659, "step": 13174 }, { "epoch": 4.05, "learning_rate": 1.348260681734842e-05, "loss": 0.3832, "step": 13175 }, { "epoch": 4.05, "learning_rate": 1.3481672645636627e-05, "loss": 0.2403, "step": 13176 }, { "epoch": 4.05, "learning_rate": 1.3480738439348617e-05, "loss": 0.379, "step": 13177 }, { "epoch": 4.05, "learning_rate": 1.3479804198493664e-05, "loss": 0.2845, "step": 13178 }, { "epoch": 4.05, "learning_rate": 1.3478869923081054e-05, "loss": 0.4142, "step": 13179 }, { "epoch": 4.06, "learning_rate": 1.3477935613120053e-05, "loss": 0.2679, "step": 13180 }, { "epoch": 4.06, "learning_rate": 1.347700126861995e-05, "loss": 0.2298, "step": 13181 }, { "epoch": 4.06, "learning_rate": 1.3476066889590015e-05, "loss": 0.3886, "step": 13182 }, { "epoch": 4.06, "learning_rate": 1.3475132476039537e-05, "loss": 0.2322, "step": 13183 }, { "epoch": 4.06, "learning_rate": 1.347419802797779e-05, "loss": 0.2811, "step": 13184 }, { "epoch": 4.06, "learning_rate": 1.347326354541405e-05, "loss": 0.25, "step": 13185 }, { "epoch": 4.06, "learning_rate": 1.3472329028357607e-05, "loss": 0.2878, "step": 13186 }, { "epoch": 4.06, "learning_rate": 1.3471394476817734e-05, "loss": 0.3761, "step": 13187 }, { "epoch": 4.06, "learning_rate": 1.3470459890803716e-05, "loss": 0.2598, "step": 13188 }, { "epoch": 4.06, "learning_rate": 1.3469525270324835e-05, "loss": 0.3147, "step": 13189 }, { "epoch": 4.06, "learning_rate": 1.3468590615390367e-05, "loss": 0.29, "step": 13190 }, { "epoch": 4.06, "learning_rate": 1.34676559260096e-05, "loss": 0.269, "step": 13191 }, { "epoch": 4.06, "learning_rate": 1.3466721202191808e-05, "loss": 0.3136, "step": 13192 }, { "epoch": 4.06, "learning_rate": 1.3465786443946286e-05, "loss": 0.3507, "step": 13193 }, { "epoch": 4.06, "learning_rate": 1.3464851651282306e-05, "loss": 0.2349, "step": 13194 }, { "epoch": 4.06, "learning_rate": 1.346391682420916e-05, "loss": 0.3089, "step": 13195 }, { "epoch": 4.06, "learning_rate": 1.3462981962736126e-05, "loss": 0.2913, "step": 13196 }, { "epoch": 4.06, "learning_rate": 1.3462047066872485e-05, "loss": 0.3117, "step": 13197 }, { "epoch": 4.06, "learning_rate": 1.3461112136627533e-05, "loss": 0.4058, "step": 13198 }, { "epoch": 4.06, "learning_rate": 1.3460177172010544e-05, "loss": 0.3332, "step": 13199 }, { "epoch": 4.06, "learning_rate": 1.3459242173030806e-05, "loss": 0.3434, "step": 13200 }, { "epoch": 4.06, "learning_rate": 1.3458307139697605e-05, "loss": 0.3949, "step": 13201 }, { "epoch": 4.06, "learning_rate": 1.345737207202023e-05, "loss": 0.2153, "step": 13202 }, { "epoch": 4.06, "learning_rate": 1.3456436970007963e-05, "loss": 0.3987, "step": 13203 }, { "epoch": 4.06, "learning_rate": 1.3455501833670089e-05, "loss": 0.265, "step": 13204 }, { "epoch": 4.06, "learning_rate": 1.34545666630159e-05, "loss": 0.2993, "step": 13205 }, { "epoch": 4.06, "learning_rate": 1.3453631458054677e-05, "loss": 0.3626, "step": 13206 }, { "epoch": 4.06, "learning_rate": 1.3452696218795712e-05, "loss": 0.2476, "step": 13207 }, { "epoch": 4.06, "learning_rate": 1.345176094524829e-05, "loss": 0.4557, "step": 13208 }, { "epoch": 4.06, "learning_rate": 1.3450825637421705e-05, "loss": 0.3135, "step": 13209 }, { "epoch": 4.06, "learning_rate": 1.344989029532524e-05, "loss": 0.3059, "step": 13210 }, { "epoch": 4.06, "learning_rate": 1.344895491896818e-05, "loss": 0.3652, "step": 13211 }, { "epoch": 4.07, "learning_rate": 1.3448019508359821e-05, "loss": 0.2053, "step": 13212 }, { "epoch": 4.07, "learning_rate": 1.3447084063509453e-05, "loss": 0.2149, "step": 13213 }, { "epoch": 4.07, "learning_rate": 1.3446148584426362e-05, "loss": 0.1812, "step": 13214 }, { "epoch": 4.07, "learning_rate": 1.3445213071119841e-05, "loss": 0.3693, "step": 13215 }, { "epoch": 4.07, "learning_rate": 1.3444277523599176e-05, "loss": 0.2663, "step": 13216 }, { "epoch": 4.07, "learning_rate": 1.3443341941873662e-05, "loss": 0.2477, "step": 13217 }, { "epoch": 4.07, "learning_rate": 1.3442406325952591e-05, "loss": 0.3144, "step": 13218 }, { "epoch": 4.07, "learning_rate": 1.344147067584525e-05, "loss": 0.384, "step": 13219 }, { "epoch": 4.07, "learning_rate": 1.3440534991560936e-05, "loss": 0.2236, "step": 13220 }, { "epoch": 4.07, "learning_rate": 1.3439599273108935e-05, "loss": 0.2565, "step": 13221 }, { "epoch": 4.07, "learning_rate": 1.3438663520498545e-05, "loss": 0.1828, "step": 13222 }, { "epoch": 4.07, "learning_rate": 1.3437727733739054e-05, "loss": 0.3199, "step": 13223 }, { "epoch": 4.07, "learning_rate": 1.3436791912839766e-05, "loss": 0.2774, "step": 13224 }, { "epoch": 4.07, "learning_rate": 1.3435856057809962e-05, "loss": 0.2902, "step": 13225 }, { "epoch": 4.07, "learning_rate": 1.3434920168658941e-05, "loss": 0.2506, "step": 13226 }, { "epoch": 4.07, "learning_rate": 1.3433984245395997e-05, "loss": 0.395, "step": 13227 }, { "epoch": 4.07, "learning_rate": 1.3433048288030424e-05, "loss": 0.3262, "step": 13228 }, { "epoch": 4.07, "learning_rate": 1.343211229657152e-05, "loss": 0.1906, "step": 13229 }, { "epoch": 4.07, "learning_rate": 1.3431176271028575e-05, "loss": 0.3434, "step": 13230 }, { "epoch": 4.07, "learning_rate": 1.3430240211410887e-05, "loss": 0.3238, "step": 13231 }, { "epoch": 4.07, "learning_rate": 1.3429304117727755e-05, "loss": 0.4088, "step": 13232 }, { "epoch": 4.07, "learning_rate": 1.342836798998847e-05, "loss": 0.2516, "step": 13233 }, { "epoch": 4.07, "learning_rate": 1.3427431828202335e-05, "loss": 0.3427, "step": 13234 }, { "epoch": 4.07, "learning_rate": 1.3426495632378638e-05, "loss": 0.2886, "step": 13235 }, { "epoch": 4.07, "learning_rate": 1.3425559402526687e-05, "loss": 0.2754, "step": 13236 }, { "epoch": 4.07, "learning_rate": 1.342462313865577e-05, "loss": 0.2604, "step": 13237 }, { "epoch": 4.07, "learning_rate": 1.3423686840775188e-05, "loss": 0.2802, "step": 13238 }, { "epoch": 4.07, "learning_rate": 1.3422750508894247e-05, "loss": 0.3032, "step": 13239 }, { "epoch": 4.07, "learning_rate": 1.3421814143022232e-05, "loss": 0.2263, "step": 13240 }, { "epoch": 4.07, "learning_rate": 1.342087774316845e-05, "loss": 0.1989, "step": 13241 }, { "epoch": 4.07, "learning_rate": 1.34199413093422e-05, "loss": 0.293, "step": 13242 }, { "epoch": 4.07, "learning_rate": 1.341900484155278e-05, "loss": 0.2985, "step": 13243 }, { "epoch": 4.07, "learning_rate": 1.3418068339809493e-05, "loss": 0.2368, "step": 13244 }, { "epoch": 4.08, "learning_rate": 1.3417131804121639e-05, "loss": 0.3061, "step": 13245 }, { "epoch": 4.08, "learning_rate": 1.3416195234498515e-05, "loss": 0.3064, "step": 13246 }, { "epoch": 4.08, "learning_rate": 1.3415258630949422e-05, "loss": 0.3288, "step": 13247 }, { "epoch": 4.08, "learning_rate": 1.3414321993483664e-05, "loss": 0.2303, "step": 13248 }, { "epoch": 4.08, "learning_rate": 1.3413385322110541e-05, "loss": 0.2021, "step": 13249 }, { "epoch": 4.08, "learning_rate": 1.341244861683936e-05, "loss": 0.2225, "step": 13250 }, { "epoch": 4.08, "learning_rate": 1.3411511877679414e-05, "loss": 0.2891, "step": 13251 }, { "epoch": 4.08, "learning_rate": 1.3410575104640014e-05, "loss": 0.3062, "step": 13252 }, { "epoch": 4.08, "learning_rate": 1.3409638297730458e-05, "loss": 0.3205, "step": 13253 }, { "epoch": 4.08, "learning_rate": 1.3408701456960052e-05, "loss": 0.1447, "step": 13254 }, { "epoch": 4.08, "learning_rate": 1.3407764582338101e-05, "loss": 0.2261, "step": 13255 }, { "epoch": 4.08, "learning_rate": 1.3406827673873903e-05, "loss": 0.271, "step": 13256 }, { "epoch": 4.08, "learning_rate": 1.3405890731576772e-05, "loss": 0.311, "step": 13257 }, { "epoch": 4.08, "learning_rate": 1.3404953755456003e-05, "loss": 0.366, "step": 13258 }, { "epoch": 4.08, "learning_rate": 1.3404016745520905e-05, "loss": 0.2565, "step": 13259 }, { "epoch": 4.08, "learning_rate": 1.3403079701780786e-05, "loss": 0.3699, "step": 13260 }, { "epoch": 4.08, "learning_rate": 1.3402142624244947e-05, "loss": 0.2107, "step": 13261 }, { "epoch": 4.08, "learning_rate": 1.3401205512922697e-05, "loss": 0.2943, "step": 13262 }, { "epoch": 4.08, "learning_rate": 1.3400268367823341e-05, "loss": 0.1695, "step": 13263 }, { "epoch": 4.08, "learning_rate": 1.3399331188956187e-05, "loss": 0.3349, "step": 13264 }, { "epoch": 4.08, "learning_rate": 1.3398393976330544e-05, "loss": 0.2116, "step": 13265 }, { "epoch": 4.08, "learning_rate": 1.339745672995571e-05, "loss": 0.2074, "step": 13266 }, { "epoch": 4.08, "learning_rate": 1.3396519449841006e-05, "loss": 0.2714, "step": 13267 }, { "epoch": 4.08, "learning_rate": 1.339558213599573e-05, "loss": 0.3864, "step": 13268 }, { "epoch": 4.08, "learning_rate": 1.3394644788429195e-05, "loss": 0.3011, "step": 13269 }, { "epoch": 4.08, "learning_rate": 1.339370740715071e-05, "loss": 0.3328, "step": 13270 }, { "epoch": 4.08, "learning_rate": 1.3392769992169581e-05, "loss": 0.3513, "step": 13271 }, { "epoch": 4.08, "learning_rate": 1.339183254349512e-05, "loss": 0.2605, "step": 13272 }, { "epoch": 4.08, "learning_rate": 1.3390895061136633e-05, "loss": 0.3185, "step": 13273 }, { "epoch": 4.08, "learning_rate": 1.3389957545103436e-05, "loss": 0.287, "step": 13274 }, { "epoch": 4.08, "learning_rate": 1.3389019995404838e-05, "loss": 0.3333, "step": 13275 }, { "epoch": 4.08, "learning_rate": 1.3388082412050145e-05, "loss": 0.2896, "step": 13276 }, { "epoch": 4.09, "learning_rate": 1.338714479504867e-05, "loss": 0.2848, "step": 13277 }, { "epoch": 4.09, "learning_rate": 1.3386207144409727e-05, "loss": 0.286, "step": 13278 }, { "epoch": 4.09, "learning_rate": 1.3385269460142628e-05, "loss": 0.3462, "step": 13279 }, { "epoch": 4.09, "learning_rate": 1.338433174225668e-05, "loss": 0.1437, "step": 13280 }, { "epoch": 4.09, "learning_rate": 1.3383393990761201e-05, "loss": 0.2564, "step": 13281 }, { "epoch": 4.09, "learning_rate": 1.3382456205665502e-05, "loss": 0.3722, "step": 13282 }, { "epoch": 4.09, "learning_rate": 1.3381518386978894e-05, "loss": 0.2932, "step": 13283 }, { "epoch": 4.09, "learning_rate": 1.3380580534710692e-05, "loss": 0.1985, "step": 13284 }, { "epoch": 4.09, "learning_rate": 1.337964264887021e-05, "loss": 0.2241, "step": 13285 }, { "epoch": 4.09, "learning_rate": 1.3378704729466763e-05, "loss": 0.4194, "step": 13286 }, { "epoch": 4.09, "learning_rate": 1.3377766776509661e-05, "loss": 0.3686, "step": 13287 }, { "epoch": 4.09, "learning_rate": 1.3376828790008223e-05, "loss": 0.2565, "step": 13288 }, { "epoch": 4.09, "learning_rate": 1.3375890769971764e-05, "loss": 0.2764, "step": 13289 }, { "epoch": 4.09, "learning_rate": 1.3374952716409598e-05, "loss": 0.2892, "step": 13290 }, { "epoch": 4.09, "learning_rate": 1.3374014629331042e-05, "loss": 0.1197, "step": 13291 }, { "epoch": 4.09, "learning_rate": 1.3373076508745408e-05, "loss": 0.2411, "step": 13292 }, { "epoch": 4.09, "learning_rate": 1.3372138354662018e-05, "loss": 0.3459, "step": 13293 }, { "epoch": 4.09, "learning_rate": 1.3371200167090188e-05, "loss": 0.2274, "step": 13294 }, { "epoch": 4.09, "learning_rate": 1.3370261946039232e-05, "loss": 0.4122, "step": 13295 }, { "epoch": 4.09, "learning_rate": 1.3369323691518468e-05, "loss": 0.2495, "step": 13296 }, { "epoch": 4.09, "learning_rate": 1.3368385403537212e-05, "loss": 0.3016, "step": 13297 }, { "epoch": 4.09, "learning_rate": 1.3367447082104788e-05, "loss": 0.133, "step": 13298 }, { "epoch": 4.09, "learning_rate": 1.336650872723051e-05, "loss": 0.3329, "step": 13299 }, { "epoch": 4.09, "learning_rate": 1.3365570338923699e-05, "loss": 0.2382, "step": 13300 }, { "epoch": 4.09, "learning_rate": 1.3364631917193671e-05, "loss": 0.3234, "step": 13301 }, { "epoch": 4.09, "learning_rate": 1.3363693462049747e-05, "loss": 0.4993, "step": 13302 }, { "epoch": 4.09, "learning_rate": 1.3362754973501249e-05, "loss": 0.4278, "step": 13303 }, { "epoch": 4.09, "learning_rate": 1.336181645155749e-05, "loss": 0.3283, "step": 13304 }, { "epoch": 4.09, "learning_rate": 1.3360877896227802e-05, "loss": 0.3126, "step": 13305 }, { "epoch": 4.09, "learning_rate": 1.3359939307521494e-05, "loss": 0.3146, "step": 13306 }, { "epoch": 4.09, "learning_rate": 1.3359000685447894e-05, "loss": 0.259, "step": 13307 }, { "epoch": 4.09, "learning_rate": 1.3358062030016321e-05, "loss": 0.3241, "step": 13308 }, { "epoch": 4.09, "learning_rate": 1.3357123341236095e-05, "loss": 0.2829, "step": 13309 }, { "epoch": 4.1, "learning_rate": 1.3356184619116546e-05, "loss": 0.674, "step": 13310 }, { "epoch": 4.1, "learning_rate": 1.3355245863666986e-05, "loss": 0.1822, "step": 13311 }, { "epoch": 4.1, "learning_rate": 1.3354307074896745e-05, "loss": 0.3422, "step": 13312 }, { "epoch": 4.1, "learning_rate": 1.3353368252815139e-05, "loss": 0.2685, "step": 13313 }, { "epoch": 4.1, "learning_rate": 1.3352429397431498e-05, "loss": 0.3281, "step": 13314 }, { "epoch": 4.1, "learning_rate": 1.3351490508755148e-05, "loss": 0.2582, "step": 13315 }, { "epoch": 4.1, "learning_rate": 1.3350551586795401e-05, "loss": 0.2868, "step": 13316 }, { "epoch": 4.1, "learning_rate": 1.3349612631561592e-05, "loss": 0.2822, "step": 13317 }, { "epoch": 4.1, "learning_rate": 1.3348673643063044e-05, "loss": 0.3059, "step": 13318 }, { "epoch": 4.1, "learning_rate": 1.3347734621309076e-05, "loss": 0.283, "step": 13319 }, { "epoch": 4.1, "learning_rate": 1.3346795566309024e-05, "loss": 0.3337, "step": 13320 }, { "epoch": 4.1, "learning_rate": 1.3345856478072204e-05, "loss": 0.4261, "step": 13321 }, { "epoch": 4.1, "learning_rate": 1.3344917356607945e-05, "loss": 0.2574, "step": 13322 }, { "epoch": 4.1, "learning_rate": 1.3343978201925573e-05, "loss": 0.3094, "step": 13323 }, { "epoch": 4.1, "learning_rate": 1.3343039014034419e-05, "loss": 0.4394, "step": 13324 }, { "epoch": 4.1, "learning_rate": 1.3342099792943806e-05, "loss": 0.335, "step": 13325 }, { "epoch": 4.1, "learning_rate": 1.334116053866306e-05, "loss": 0.3711, "step": 13326 }, { "epoch": 4.1, "learning_rate": 1.3340221251201512e-05, "loss": 0.3133, "step": 13327 }, { "epoch": 4.1, "learning_rate": 1.3339281930568486e-05, "loss": 0.2318, "step": 13328 }, { "epoch": 4.1, "learning_rate": 1.3338342576773315e-05, "loss": 0.321, "step": 13329 }, { "epoch": 4.1, "learning_rate": 1.3337403189825327e-05, "loss": 0.4453, "step": 13330 }, { "epoch": 4.1, "learning_rate": 1.3336463769733848e-05, "loss": 0.1682, "step": 13331 }, { "epoch": 4.1, "learning_rate": 1.3335524316508208e-05, "loss": 0.3255, "step": 13332 }, { "epoch": 4.1, "learning_rate": 1.3334584830157737e-05, "loss": 0.1769, "step": 13333 }, { "epoch": 4.1, "learning_rate": 1.3333645310691767e-05, "loss": 0.2364, "step": 13334 }, { "epoch": 4.1, "learning_rate": 1.3332705758119627e-05, "loss": 0.2042, "step": 13335 }, { "epoch": 4.1, "learning_rate": 1.3331766172450647e-05, "loss": 0.2408, "step": 13336 }, { "epoch": 4.1, "learning_rate": 1.3330826553694156e-05, "loss": 0.2721, "step": 13337 }, { "epoch": 4.1, "learning_rate": 1.3329886901859488e-05, "loss": 0.4005, "step": 13338 }, { "epoch": 4.1, "learning_rate": 1.3328947216955977e-05, "loss": 0.2473, "step": 13339 }, { "epoch": 4.1, "learning_rate": 1.332800749899295e-05, "loss": 0.3563, "step": 13340 }, { "epoch": 4.1, "learning_rate": 1.3327067747979743e-05, "loss": 0.2828, "step": 13341 }, { "epoch": 4.11, "learning_rate": 1.3326127963925684e-05, "loss": 0.2942, "step": 13342 }, { "epoch": 4.11, "learning_rate": 1.3325188146840108e-05, "loss": 0.2171, "step": 13343 }, { "epoch": 4.11, "learning_rate": 1.3324248296732353e-05, "loss": 0.1474, "step": 13344 }, { "epoch": 4.11, "learning_rate": 1.3323308413611748e-05, "loss": 0.2416, "step": 13345 }, { "epoch": 4.11, "learning_rate": 1.3322368497487628e-05, "loss": 0.2007, "step": 13346 }, { "epoch": 4.11, "learning_rate": 1.3321428548369324e-05, "loss": 0.3282, "step": 13347 }, { "epoch": 4.11, "learning_rate": 1.3320488566266177e-05, "loss": 0.3148, "step": 13348 }, { "epoch": 4.11, "learning_rate": 1.3319548551187515e-05, "loss": 0.2844, "step": 13349 }, { "epoch": 4.11, "learning_rate": 1.3318608503142678e-05, "loss": 0.3278, "step": 13350 }, { "epoch": 4.11, "learning_rate": 1.3317668422141004e-05, "loss": 0.3031, "step": 13351 }, { "epoch": 4.11, "learning_rate": 1.3316728308191817e-05, "loss": 0.265, "step": 13352 }, { "epoch": 4.11, "learning_rate": 1.3315788161304466e-05, "loss": 0.3434, "step": 13353 }, { "epoch": 4.11, "learning_rate": 1.3314847981488281e-05, "loss": 0.3351, "step": 13354 }, { "epoch": 4.11, "learning_rate": 1.33139077687526e-05, "loss": 0.2614, "step": 13355 }, { "epoch": 4.11, "learning_rate": 1.3312967523106766e-05, "loss": 0.4458, "step": 13356 }, { "epoch": 4.11, "learning_rate": 1.3312027244560104e-05, "loss": 0.2985, "step": 13357 }, { "epoch": 4.11, "learning_rate": 1.3311086933121961e-05, "loss": 0.2297, "step": 13358 }, { "epoch": 4.11, "learning_rate": 1.3310146588801676e-05, "loss": 0.3491, "step": 13359 }, { "epoch": 4.11, "learning_rate": 1.3309206211608582e-05, "loss": 0.1192, "step": 13360 }, { "epoch": 4.11, "learning_rate": 1.3308265801552023e-05, "loss": 0.2621, "step": 13361 }, { "epoch": 4.11, "learning_rate": 1.3307325358641335e-05, "loss": 0.2793, "step": 13362 }, { "epoch": 4.11, "learning_rate": 1.3306384882885856e-05, "loss": 0.2929, "step": 13363 }, { "epoch": 4.11, "learning_rate": 1.3305444374294927e-05, "loss": 0.2854, "step": 13364 }, { "epoch": 4.11, "learning_rate": 1.3304503832877891e-05, "loss": 0.2342, "step": 13365 }, { "epoch": 4.11, "learning_rate": 1.3303563258644085e-05, "loss": 0.2144, "step": 13366 }, { "epoch": 4.11, "learning_rate": 1.3302622651602855e-05, "loss": 0.3777, "step": 13367 }, { "epoch": 4.11, "learning_rate": 1.3301682011763535e-05, "loss": 0.4432, "step": 13368 }, { "epoch": 4.11, "learning_rate": 1.330074133913547e-05, "loss": 0.3225, "step": 13369 }, { "epoch": 4.11, "learning_rate": 1.3299800633728002e-05, "loss": 0.2357, "step": 13370 }, { "epoch": 4.11, "learning_rate": 1.3298859895550473e-05, "loss": 0.4301, "step": 13371 }, { "epoch": 4.11, "learning_rate": 1.3297919124612225e-05, "loss": 0.41, "step": 13372 }, { "epoch": 4.11, "learning_rate": 1.3296978320922598e-05, "loss": 0.279, "step": 13373 }, { "epoch": 4.11, "learning_rate": 1.329603748449094e-05, "loss": 0.3902, "step": 13374 }, { "epoch": 4.12, "learning_rate": 1.3295096615326593e-05, "loss": 0.4489, "step": 13375 }, { "epoch": 4.12, "learning_rate": 1.3294155713438896e-05, "loss": 0.28, "step": 13376 }, { "epoch": 4.12, "learning_rate": 1.32932147788372e-05, "loss": 0.2126, "step": 13377 }, { "epoch": 4.12, "learning_rate": 1.3292273811530845e-05, "loss": 0.3785, "step": 13378 }, { "epoch": 4.12, "learning_rate": 1.3291332811529179e-05, "loss": 0.2471, "step": 13379 }, { "epoch": 4.12, "learning_rate": 1.3290391778841543e-05, "loss": 0.4551, "step": 13380 }, { "epoch": 4.12, "learning_rate": 1.3289450713477283e-05, "loss": 0.3126, "step": 13381 }, { "epoch": 4.12, "learning_rate": 1.3288509615445748e-05, "loss": 0.5302, "step": 13382 }, { "epoch": 4.12, "learning_rate": 1.3287568484756279e-05, "loss": 0.2736, "step": 13383 }, { "epoch": 4.12, "learning_rate": 1.3286627321418229e-05, "loss": 0.312, "step": 13384 }, { "epoch": 4.12, "learning_rate": 1.3285686125440938e-05, "loss": 0.3297, "step": 13385 }, { "epoch": 4.12, "learning_rate": 1.3284744896833755e-05, "loss": 0.4404, "step": 13386 }, { "epoch": 4.12, "learning_rate": 1.328380363560603e-05, "loss": 0.1744, "step": 13387 }, { "epoch": 4.12, "learning_rate": 1.3282862341767104e-05, "loss": 0.2611, "step": 13388 }, { "epoch": 4.12, "learning_rate": 1.3281921015326334e-05, "loss": 0.3395, "step": 13389 }, { "epoch": 4.12, "learning_rate": 1.328097965629306e-05, "loss": 0.1803, "step": 13390 }, { "epoch": 4.12, "learning_rate": 1.328003826467664e-05, "loss": 0.3705, "step": 13391 }, { "epoch": 4.12, "learning_rate": 1.3279096840486411e-05, "loss": 0.3584, "step": 13392 }, { "epoch": 4.12, "learning_rate": 1.327815538373173e-05, "loss": 0.1608, "step": 13393 }, { "epoch": 4.12, "learning_rate": 1.3277213894421948e-05, "loss": 0.6474, "step": 13394 }, { "epoch": 4.12, "learning_rate": 1.3276272372566407e-05, "loss": 0.258, "step": 13395 }, { "epoch": 4.12, "learning_rate": 1.3275330818174465e-05, "loss": 0.279, "step": 13396 }, { "epoch": 4.12, "learning_rate": 1.3274389231255466e-05, "loss": 0.2645, "step": 13397 }, { "epoch": 4.12, "learning_rate": 1.3273447611818768e-05, "loss": 0.2438, "step": 13398 }, { "epoch": 4.12, "learning_rate": 1.3272505959873718e-05, "loss": 0.2227, "step": 13399 }, { "epoch": 4.12, "learning_rate": 1.3271564275429664e-05, "loss": 0.4309, "step": 13400 }, { "epoch": 4.12, "learning_rate": 1.327062255849597e-05, "loss": 0.4187, "step": 13401 }, { "epoch": 4.12, "learning_rate": 1.3269680809081971e-05, "loss": 0.3318, "step": 13402 }, { "epoch": 4.12, "learning_rate": 1.326873902719703e-05, "loss": 0.2768, "step": 13403 }, { "epoch": 4.12, "learning_rate": 1.3267797212850498e-05, "loss": 0.3235, "step": 13404 }, { "epoch": 4.12, "learning_rate": 1.326685536605173e-05, "loss": 0.4476, "step": 13405 }, { "epoch": 4.12, "learning_rate": 1.3265913486810077e-05, "loss": 0.2312, "step": 13406 }, { "epoch": 4.13, "learning_rate": 1.3264971575134893e-05, "loss": 0.2668, "step": 13407 }, { "epoch": 4.13, "learning_rate": 1.326402963103553e-05, "loss": 0.43, "step": 13408 }, { "epoch": 4.13, "learning_rate": 1.3263087654521345e-05, "loss": 0.2026, "step": 13409 }, { "epoch": 4.13, "learning_rate": 1.3262145645601693e-05, "loss": 0.3224, "step": 13410 }, { "epoch": 4.13, "learning_rate": 1.326120360428593e-05, "loss": 0.2843, "step": 13411 }, { "epoch": 4.13, "learning_rate": 1.3260261530583409e-05, "loss": 0.5276, "step": 13412 }, { "epoch": 4.13, "learning_rate": 1.3259319424503488e-05, "loss": 0.2748, "step": 13413 }, { "epoch": 4.13, "learning_rate": 1.3258377286055515e-05, "loss": 0.2686, "step": 13414 }, { "epoch": 4.13, "learning_rate": 1.3257435115248858e-05, "loss": 0.2923, "step": 13415 }, { "epoch": 4.13, "learning_rate": 1.3256492912092866e-05, "loss": 0.2426, "step": 13416 }, { "epoch": 4.13, "learning_rate": 1.32555506765969e-05, "loss": 0.4496, "step": 13417 }, { "epoch": 4.13, "learning_rate": 1.3254608408770313e-05, "loss": 0.2453, "step": 13418 }, { "epoch": 4.13, "learning_rate": 1.3253666108622465e-05, "loss": 0.2495, "step": 13419 }, { "epoch": 4.13, "learning_rate": 1.3252723776162714e-05, "loss": 0.3445, "step": 13420 }, { "epoch": 4.13, "learning_rate": 1.325178141140042e-05, "loss": 0.317, "step": 13421 }, { "epoch": 4.13, "learning_rate": 1.3250839014344939e-05, "loss": 0.2743, "step": 13422 }, { "epoch": 4.13, "learning_rate": 1.3249896585005628e-05, "loss": 0.2108, "step": 13423 }, { "epoch": 4.13, "learning_rate": 1.3248954123391852e-05, "loss": 0.2761, "step": 13424 }, { "epoch": 4.13, "learning_rate": 1.3248011629512965e-05, "loss": 0.2617, "step": 13425 }, { "epoch": 4.13, "learning_rate": 1.3247069103378332e-05, "loss": 0.289, "step": 13426 }, { "epoch": 4.13, "learning_rate": 1.3246126544997308e-05, "loss": 0.3562, "step": 13427 }, { "epoch": 4.13, "learning_rate": 1.3245183954379255e-05, "loss": 0.297, "step": 13428 }, { "epoch": 4.13, "learning_rate": 1.3244241331533535e-05, "loss": 0.3243, "step": 13429 }, { "epoch": 4.13, "learning_rate": 1.324329867646951e-05, "loss": 0.1756, "step": 13430 }, { "epoch": 4.13, "learning_rate": 1.3242355989196539e-05, "loss": 0.1211, "step": 13431 }, { "epoch": 4.13, "learning_rate": 1.3241413269723987e-05, "loss": 0.3659, "step": 13432 }, { "epoch": 4.13, "learning_rate": 1.3240470518061208e-05, "loss": 0.3384, "step": 13433 }, { "epoch": 4.13, "learning_rate": 1.3239527734217576e-05, "loss": 0.2105, "step": 13434 }, { "epoch": 4.13, "learning_rate": 1.3238584918202446e-05, "loss": 0.1948, "step": 13435 }, { "epoch": 4.13, "learning_rate": 1.3237642070025183e-05, "loss": 0.3846, "step": 13436 }, { "epoch": 4.13, "learning_rate": 1.323669918969515e-05, "loss": 0.4459, "step": 13437 }, { "epoch": 4.13, "learning_rate": 1.323575627722171e-05, "loss": 0.2655, "step": 13438 }, { "epoch": 4.13, "learning_rate": 1.323481333261423e-05, "loss": 0.2926, "step": 13439 }, { "epoch": 4.14, "learning_rate": 1.323387035588207e-05, "loss": 0.3248, "step": 13440 }, { "epoch": 4.14, "learning_rate": 1.3232927347034601e-05, "loss": 0.2376, "step": 13441 }, { "epoch": 4.14, "learning_rate": 1.323198430608118e-05, "loss": 0.2139, "step": 13442 }, { "epoch": 4.14, "learning_rate": 1.323104123303118e-05, "loss": 0.3072, "step": 13443 }, { "epoch": 4.14, "learning_rate": 1.3230098127893957e-05, "loss": 0.4218, "step": 13444 }, { "epoch": 4.14, "learning_rate": 1.3229154990678886e-05, "loss": 0.4489, "step": 13445 }, { "epoch": 4.14, "learning_rate": 1.3228211821395332e-05, "loss": 0.2899, "step": 13446 }, { "epoch": 4.14, "learning_rate": 1.3227268620052656e-05, "loss": 0.2552, "step": 13447 }, { "epoch": 4.14, "learning_rate": 1.3226325386660231e-05, "loss": 0.3662, "step": 13448 }, { "epoch": 4.14, "learning_rate": 1.322538212122742e-05, "loss": 0.2706, "step": 13449 }, { "epoch": 4.14, "learning_rate": 1.322443882376359e-05, "loss": 0.2232, "step": 13450 }, { "epoch": 4.14, "learning_rate": 1.3223495494278114e-05, "loss": 0.3382, "step": 13451 }, { "epoch": 4.14, "learning_rate": 1.3222552132780356e-05, "loss": 0.3117, "step": 13452 }, { "epoch": 4.14, "learning_rate": 1.3221608739279686e-05, "loss": 0.2564, "step": 13453 }, { "epoch": 4.14, "learning_rate": 1.3220665313785467e-05, "loss": 0.2301, "step": 13454 }, { "epoch": 4.14, "learning_rate": 1.3219721856307078e-05, "loss": 0.2757, "step": 13455 }, { "epoch": 4.14, "learning_rate": 1.3218778366853884e-05, "loss": 0.2312, "step": 13456 }, { "epoch": 4.14, "learning_rate": 1.3217834845435254e-05, "loss": 0.2566, "step": 13457 }, { "epoch": 4.14, "learning_rate": 1.3216891292060556e-05, "loss": 0.2987, "step": 13458 }, { "epoch": 4.14, "learning_rate": 1.3215947706739163e-05, "loss": 0.2434, "step": 13459 }, { "epoch": 4.14, "learning_rate": 1.3215004089480448e-05, "loss": 0.1728, "step": 13460 }, { "epoch": 4.14, "learning_rate": 1.3214060440293777e-05, "loss": 0.2314, "step": 13461 }, { "epoch": 4.14, "learning_rate": 1.3213116759188525e-05, "loss": 0.2343, "step": 13462 }, { "epoch": 4.14, "learning_rate": 1.3212173046174061e-05, "loss": 0.3865, "step": 13463 }, { "epoch": 4.14, "learning_rate": 1.3211229301259759e-05, "loss": 0.2009, "step": 13464 }, { "epoch": 4.14, "learning_rate": 1.3210285524454991e-05, "loss": 0.1668, "step": 13465 }, { "epoch": 4.14, "learning_rate": 1.3209341715769128e-05, "loss": 0.2994, "step": 13466 }, { "epoch": 4.14, "learning_rate": 1.3208397875211545e-05, "loss": 0.246, "step": 13467 }, { "epoch": 4.14, "learning_rate": 1.3207454002791615e-05, "loss": 0.334, "step": 13468 }, { "epoch": 4.14, "learning_rate": 1.3206510098518707e-05, "loss": 0.2911, "step": 13469 }, { "epoch": 4.14, "learning_rate": 1.3205566162402201e-05, "loss": 0.2601, "step": 13470 }, { "epoch": 4.14, "learning_rate": 1.3204622194451466e-05, "loss": 0.4443, "step": 13471 }, { "epoch": 4.15, "learning_rate": 1.3203678194675886e-05, "loss": 0.3632, "step": 13472 }, { "epoch": 4.15, "learning_rate": 1.3202734163084823e-05, "loss": 0.3839, "step": 13473 }, { "epoch": 4.15, "learning_rate": 1.3201790099687659e-05, "loss": 0.2967, "step": 13474 }, { "epoch": 4.15, "learning_rate": 1.320084600449377e-05, "loss": 0.2702, "step": 13475 }, { "epoch": 4.15, "learning_rate": 1.3199901877512526e-05, "loss": 0.3102, "step": 13476 }, { "epoch": 4.15, "learning_rate": 1.3198957718753313e-05, "loss": 0.1654, "step": 13477 }, { "epoch": 4.15, "learning_rate": 1.31980135282255e-05, "loss": 0.2165, "step": 13478 }, { "epoch": 4.15, "learning_rate": 1.3197069305938464e-05, "loss": 0.2569, "step": 13479 }, { "epoch": 4.15, "learning_rate": 1.3196125051901582e-05, "loss": 0.2553, "step": 13480 }, { "epoch": 4.15, "learning_rate": 1.3195180766124234e-05, "loss": 0.394, "step": 13481 }, { "epoch": 4.15, "learning_rate": 1.3194236448615798e-05, "loss": 0.1342, "step": 13482 }, { "epoch": 4.15, "learning_rate": 1.3193292099385648e-05, "loss": 0.14, "step": 13483 }, { "epoch": 4.15, "learning_rate": 1.3192347718443165e-05, "loss": 0.2383, "step": 13484 }, { "epoch": 4.15, "learning_rate": 1.3191403305797727e-05, "loss": 0.3244, "step": 13485 }, { "epoch": 4.15, "learning_rate": 1.3190458861458713e-05, "loss": 0.3277, "step": 13486 }, { "epoch": 4.15, "learning_rate": 1.3189514385435502e-05, "loss": 0.2839, "step": 13487 }, { "epoch": 4.15, "learning_rate": 1.3188569877737474e-05, "loss": 0.3146, "step": 13488 }, { "epoch": 4.15, "learning_rate": 1.3187625338374008e-05, "loss": 0.4511, "step": 13489 }, { "epoch": 4.15, "learning_rate": 1.3186680767354483e-05, "loss": 0.3467, "step": 13490 }, { "epoch": 4.15, "learning_rate": 1.3185736164688284e-05, "loss": 0.3798, "step": 13491 }, { "epoch": 4.15, "learning_rate": 1.318479153038479e-05, "loss": 0.2656, "step": 13492 }, { "epoch": 4.15, "learning_rate": 1.3183846864453376e-05, "loss": 0.3032, "step": 13493 }, { "epoch": 4.15, "learning_rate": 1.3182902166903434e-05, "loss": 0.3972, "step": 13494 }, { "epoch": 4.15, "learning_rate": 1.3181957437744333e-05, "loss": 0.4941, "step": 13495 }, { "epoch": 4.15, "learning_rate": 1.3181012676985465e-05, "loss": 0.3289, "step": 13496 }, { "epoch": 4.15, "learning_rate": 1.3180067884636212e-05, "loss": 0.2963, "step": 13497 }, { "epoch": 4.15, "learning_rate": 1.3179123060705954e-05, "loss": 0.2428, "step": 13498 }, { "epoch": 4.15, "learning_rate": 1.317817820520407e-05, "loss": 0.1453, "step": 13499 }, { "epoch": 4.15, "learning_rate": 1.317723331813995e-05, "loss": 0.3722, "step": 13500 }, { "epoch": 4.15, "learning_rate": 1.3176288399522975e-05, "loss": 0.2579, "step": 13501 }, { "epoch": 4.15, "learning_rate": 1.3175343449362527e-05, "loss": 0.3991, "step": 13502 }, { "epoch": 4.15, "learning_rate": 1.3174398467667995e-05, "loss": 0.3186, "step": 13503 }, { "epoch": 4.15, "learning_rate": 1.317345345444876e-05, "loss": 0.2474, "step": 13504 }, { "epoch": 4.16, "learning_rate": 1.3172508409714207e-05, "loss": 0.4146, "step": 13505 }, { "epoch": 4.16, "learning_rate": 1.317156333347372e-05, "loss": 0.3543, "step": 13506 }, { "epoch": 4.16, "learning_rate": 1.3170618225736686e-05, "loss": 0.3917, "step": 13507 }, { "epoch": 4.16, "learning_rate": 1.3169673086512496e-05, "loss": 0.2605, "step": 13508 }, { "epoch": 4.16, "learning_rate": 1.3168727915810524e-05, "loss": 0.2258, "step": 13509 }, { "epoch": 4.16, "learning_rate": 1.3167782713640168e-05, "loss": 0.3536, "step": 13510 }, { "epoch": 4.16, "learning_rate": 1.3166837480010808e-05, "loss": 0.3322, "step": 13511 }, { "epoch": 4.16, "learning_rate": 1.3165892214931834e-05, "loss": 0.2473, "step": 13512 }, { "epoch": 4.16, "learning_rate": 1.3164946918412633e-05, "loss": 0.2777, "step": 13513 }, { "epoch": 4.16, "learning_rate": 1.3164001590462592e-05, "loss": 0.2714, "step": 13514 }, { "epoch": 4.16, "learning_rate": 1.3163056231091099e-05, "loss": 0.3809, "step": 13515 }, { "epoch": 4.16, "learning_rate": 1.316211084030754e-05, "loss": 0.2483, "step": 13516 }, { "epoch": 4.16, "learning_rate": 1.316116541812131e-05, "loss": 0.2201, "step": 13517 }, { "epoch": 4.16, "learning_rate": 1.3160219964541794e-05, "loss": 0.1822, "step": 13518 }, { "epoch": 4.16, "learning_rate": 1.3159274479578376e-05, "loss": 0.239, "step": 13519 }, { "epoch": 4.16, "learning_rate": 1.3158328963240456e-05, "loss": 0.2151, "step": 13520 }, { "epoch": 4.16, "learning_rate": 1.3157383415537415e-05, "loss": 0.2144, "step": 13521 }, { "epoch": 4.16, "learning_rate": 1.315643783647865e-05, "loss": 0.2824, "step": 13522 }, { "epoch": 4.16, "learning_rate": 1.3155492226073547e-05, "loss": 0.2862, "step": 13523 }, { "epoch": 4.16, "learning_rate": 1.3154546584331498e-05, "loss": 0.2541, "step": 13524 }, { "epoch": 4.16, "learning_rate": 1.3153600911261894e-05, "loss": 0.2713, "step": 13525 }, { "epoch": 4.16, "learning_rate": 1.3152655206874125e-05, "loss": 0.3434, "step": 13526 }, { "epoch": 4.16, "learning_rate": 1.3151709471177589e-05, "loss": 0.3186, "step": 13527 }, { "epoch": 4.16, "learning_rate": 1.3150763704181669e-05, "loss": 0.2101, "step": 13528 }, { "epoch": 4.16, "learning_rate": 1.3149817905895761e-05, "loss": 0.3967, "step": 13529 }, { "epoch": 4.16, "learning_rate": 1.3148872076329262e-05, "loss": 0.1807, "step": 13530 }, { "epoch": 4.16, "learning_rate": 1.3147926215491556e-05, "loss": 0.2973, "step": 13531 }, { "epoch": 4.16, "learning_rate": 1.3146980323392046e-05, "loss": 0.4708, "step": 13532 }, { "epoch": 4.16, "learning_rate": 1.3146034400040119e-05, "loss": 0.3643, "step": 13533 }, { "epoch": 4.16, "learning_rate": 1.3145088445445173e-05, "loss": 0.2014, "step": 13534 }, { "epoch": 4.16, "learning_rate": 1.3144142459616598e-05, "loss": 0.2317, "step": 13535 }, { "epoch": 4.16, "learning_rate": 1.314319644256379e-05, "loss": 0.2611, "step": 13536 }, { "epoch": 4.17, "learning_rate": 1.3142250394296147e-05, "loss": 0.4427, "step": 13537 }, { "epoch": 4.17, "learning_rate": 1.314130431482306e-05, "loss": 0.3501, "step": 13538 }, { "epoch": 4.17, "learning_rate": 1.3140358204153926e-05, "loss": 0.2596, "step": 13539 }, { "epoch": 4.17, "learning_rate": 1.3139412062298141e-05, "loss": 0.3283, "step": 13540 }, { "epoch": 4.17, "learning_rate": 1.3138465889265103e-05, "loss": 0.3223, "step": 13541 }, { "epoch": 4.17, "learning_rate": 1.3137519685064204e-05, "loss": 0.3011, "step": 13542 }, { "epoch": 4.17, "learning_rate": 1.3136573449704842e-05, "loss": 0.6327, "step": 13543 }, { "epoch": 4.17, "learning_rate": 1.3135627183196417e-05, "loss": 0.3225, "step": 13544 }, { "epoch": 4.17, "learning_rate": 1.3134680885548321e-05, "loss": 0.321, "step": 13545 }, { "epoch": 4.17, "learning_rate": 1.3133734556769958e-05, "loss": 0.335, "step": 13546 }, { "epoch": 4.17, "learning_rate": 1.3132788196870723e-05, "loss": 0.3286, "step": 13547 }, { "epoch": 4.17, "learning_rate": 1.3131841805860013e-05, "loss": 0.2318, "step": 13548 }, { "epoch": 4.17, "learning_rate": 1.3130895383747226e-05, "loss": 0.426, "step": 13549 }, { "epoch": 4.17, "learning_rate": 1.3129948930541762e-05, "loss": 0.2742, "step": 13550 }, { "epoch": 4.17, "learning_rate": 1.3129002446253023e-05, "loss": 0.3011, "step": 13551 }, { "epoch": 4.17, "learning_rate": 1.3128055930890402e-05, "loss": 0.3209, "step": 13552 }, { "epoch": 4.17, "learning_rate": 1.312710938446331e-05, "loss": 0.291, "step": 13553 }, { "epoch": 4.17, "learning_rate": 1.3126162806981136e-05, "loss": 0.3064, "step": 13554 }, { "epoch": 4.17, "learning_rate": 1.3125216198453282e-05, "loss": 0.3231, "step": 13555 }, { "epoch": 4.17, "learning_rate": 1.3124269558889151e-05, "loss": 0.33, "step": 13556 }, { "epoch": 4.17, "learning_rate": 1.3123322888298145e-05, "loss": 0.4634, "step": 13557 }, { "epoch": 4.17, "learning_rate": 1.3122376186689668e-05, "loss": 0.3129, "step": 13558 }, { "epoch": 4.17, "learning_rate": 1.3121429454073111e-05, "loss": 0.2003, "step": 13559 }, { "epoch": 4.17, "learning_rate": 1.3120482690457887e-05, "loss": 0.1244, "step": 13560 }, { "epoch": 4.17, "learning_rate": 1.3119535895853393e-05, "loss": 0.2256, "step": 13561 }, { "epoch": 4.17, "learning_rate": 1.311858907026903e-05, "loss": 0.3333, "step": 13562 }, { "epoch": 4.17, "learning_rate": 1.311764221371421e-05, "loss": 0.2475, "step": 13563 }, { "epoch": 4.17, "learning_rate": 1.3116695326198323e-05, "loss": 0.2535, "step": 13564 }, { "epoch": 4.17, "learning_rate": 1.3115748407730781e-05, "loss": 0.2774, "step": 13565 }, { "epoch": 4.17, "learning_rate": 1.3114801458320988e-05, "loss": 0.3528, "step": 13566 }, { "epoch": 4.17, "learning_rate": 1.311385447797834e-05, "loss": 0.3357, "step": 13567 }, { "epoch": 4.17, "learning_rate": 1.3112907466712254e-05, "loss": 0.3832, "step": 13568 }, { "epoch": 4.17, "learning_rate": 1.3111960424532123e-05, "loss": 0.268, "step": 13569 }, { "epoch": 4.18, "learning_rate": 1.311101335144736e-05, "loss": 0.2602, "step": 13570 }, { "epoch": 4.18, "learning_rate": 1.3110066247467366e-05, "loss": 0.294, "step": 13571 }, { "epoch": 4.18, "learning_rate": 1.3109119112601548e-05, "loss": 0.3379, "step": 13572 }, { "epoch": 4.18, "learning_rate": 1.3108171946859312e-05, "loss": 0.1793, "step": 13573 }, { "epoch": 4.18, "learning_rate": 1.3107224750250064e-05, "loss": 0.3284, "step": 13574 }, { "epoch": 4.18, "learning_rate": 1.3106277522783211e-05, "loss": 0.2633, "step": 13575 }, { "epoch": 4.18, "learning_rate": 1.3105330264468157e-05, "loss": 0.3095, "step": 13576 }, { "epoch": 4.18, "learning_rate": 1.3104382975314314e-05, "loss": 0.3531, "step": 13577 }, { "epoch": 4.18, "learning_rate": 1.3103435655331085e-05, "loss": 0.2869, "step": 13578 }, { "epoch": 4.18, "learning_rate": 1.3102488304527883e-05, "loss": 0.3145, "step": 13579 }, { "epoch": 4.18, "learning_rate": 1.3101540922914112e-05, "loss": 0.3362, "step": 13580 }, { "epoch": 4.18, "learning_rate": 1.3100593510499175e-05, "loss": 0.1957, "step": 13581 }, { "epoch": 4.18, "learning_rate": 1.3099646067292495e-05, "loss": 0.3817, "step": 13582 }, { "epoch": 4.18, "learning_rate": 1.3098698593303469e-05, "loss": 0.3581, "step": 13583 }, { "epoch": 4.18, "learning_rate": 1.309775108854151e-05, "loss": 0.2194, "step": 13584 }, { "epoch": 4.18, "learning_rate": 1.309680355301603e-05, "loss": 0.2732, "step": 13585 }, { "epoch": 4.18, "learning_rate": 1.3095855986736434e-05, "loss": 0.5151, "step": 13586 }, { "epoch": 4.18, "learning_rate": 1.3094908389712135e-05, "loss": 0.2262, "step": 13587 }, { "epoch": 4.18, "learning_rate": 1.3093960761952544e-05, "loss": 0.3371, "step": 13588 }, { "epoch": 4.18, "learning_rate": 1.3093013103467072e-05, "loss": 0.2413, "step": 13589 }, { "epoch": 4.18, "learning_rate": 1.3092065414265128e-05, "loss": 0.3616, "step": 13590 }, { "epoch": 4.18, "learning_rate": 1.3091117694356126e-05, "loss": 0.2449, "step": 13591 }, { "epoch": 4.18, "learning_rate": 1.3090169943749475e-05, "loss": 0.279, "step": 13592 }, { "epoch": 4.18, "learning_rate": 1.308922216245459e-05, "loss": 0.2443, "step": 13593 }, { "epoch": 4.18, "learning_rate": 1.3088274350480884e-05, "loss": 0.1584, "step": 13594 }, { "epoch": 4.18, "learning_rate": 1.308732650783776e-05, "loss": 0.3489, "step": 13595 }, { "epoch": 4.18, "learning_rate": 1.3086378634534645e-05, "loss": 0.2613, "step": 13596 }, { "epoch": 4.18, "learning_rate": 1.3085430730580944e-05, "loss": 0.2528, "step": 13597 }, { "epoch": 4.18, "learning_rate": 1.308448279598607e-05, "loss": 0.3385, "step": 13598 }, { "epoch": 4.18, "learning_rate": 1.3083534830759443e-05, "loss": 0.282, "step": 13599 }, { "epoch": 4.18, "learning_rate": 1.3082586834910468e-05, "loss": 0.4167, "step": 13600 }, { "epoch": 4.18, "learning_rate": 1.308163880844857e-05, "loss": 0.2967, "step": 13601 }, { "epoch": 4.19, "learning_rate": 1.3080690751383154e-05, "loss": 0.3721, "step": 13602 }, { "epoch": 4.19, "learning_rate": 1.3079742663723645e-05, "loss": 0.1706, "step": 13603 }, { "epoch": 4.19, "learning_rate": 1.307879454547945e-05, "loss": 0.392, "step": 13604 }, { "epoch": 4.19, "learning_rate": 1.3077846396659986e-05, "loss": 0.3947, "step": 13605 }, { "epoch": 4.19, "learning_rate": 1.3076898217274674e-05, "loss": 0.3489, "step": 13606 }, { "epoch": 4.19, "learning_rate": 1.3075950007332923e-05, "loss": 0.2766, "step": 13607 }, { "epoch": 4.19, "learning_rate": 1.3075001766844157e-05, "loss": 0.2167, "step": 13608 }, { "epoch": 4.19, "learning_rate": 1.3074053495817787e-05, "loss": 0.4858, "step": 13609 }, { "epoch": 4.19, "learning_rate": 1.3073105194263236e-05, "loss": 0.3954, "step": 13610 }, { "epoch": 4.19, "learning_rate": 1.3072156862189915e-05, "loss": 0.3157, "step": 13611 }, { "epoch": 4.19, "learning_rate": 1.3071208499607244e-05, "loss": 0.2141, "step": 13612 }, { "epoch": 4.19, "learning_rate": 1.3070260106524649e-05, "loss": 0.3178, "step": 13613 }, { "epoch": 4.19, "learning_rate": 1.3069311682951532e-05, "loss": 0.4158, "step": 13614 }, { "epoch": 4.19, "learning_rate": 1.3068363228897327e-05, "loss": 0.2117, "step": 13615 }, { "epoch": 4.19, "learning_rate": 1.3067414744371447e-05, "loss": 0.2353, "step": 13616 }, { "epoch": 4.19, "learning_rate": 1.306646622938331e-05, "loss": 0.3482, "step": 13617 }, { "epoch": 4.19, "learning_rate": 1.3065517683942339e-05, "loss": 0.3102, "step": 13618 }, { "epoch": 4.19, "learning_rate": 1.306456910805795e-05, "loss": 0.1728, "step": 13619 }, { "epoch": 4.19, "learning_rate": 1.3063620501739566e-05, "loss": 0.2748, "step": 13620 }, { "epoch": 4.19, "learning_rate": 1.3062671864996607e-05, "loss": 0.3794, "step": 13621 }, { "epoch": 4.19, "learning_rate": 1.3061723197838494e-05, "loss": 0.284, "step": 13622 }, { "epoch": 4.19, "learning_rate": 1.306077450027465e-05, "loss": 0.2891, "step": 13623 }, { "epoch": 4.19, "learning_rate": 1.3059825772314492e-05, "loss": 0.3324, "step": 13624 }, { "epoch": 4.19, "learning_rate": 1.3058877013967443e-05, "loss": 0.2702, "step": 13625 }, { "epoch": 4.19, "learning_rate": 1.3057928225242926e-05, "loss": 0.3309, "step": 13626 }, { "epoch": 4.19, "learning_rate": 1.3056979406150365e-05, "loss": 0.3454, "step": 13627 }, { "epoch": 4.19, "learning_rate": 1.3056030556699181e-05, "loss": 0.3155, "step": 13628 }, { "epoch": 4.19, "learning_rate": 1.30550816768988e-05, "loss": 0.2683, "step": 13629 }, { "epoch": 4.19, "learning_rate": 1.3054132766758639e-05, "loss": 0.2563, "step": 13630 }, { "epoch": 4.19, "learning_rate": 1.3053183826288124e-05, "loss": 0.412, "step": 13631 }, { "epoch": 4.19, "learning_rate": 1.305223485549668e-05, "loss": 0.2492, "step": 13632 }, { "epoch": 4.19, "learning_rate": 1.3051285854393734e-05, "loss": 0.3032, "step": 13633 }, { "epoch": 4.19, "learning_rate": 1.3050336822988704e-05, "loss": 0.3496, "step": 13634 }, { "epoch": 4.2, "learning_rate": 1.304938776129102e-05, "loss": 0.2374, "step": 13635 }, { "epoch": 4.2, "learning_rate": 1.3048438669310101e-05, "loss": 0.2505, "step": 13636 }, { "epoch": 4.2, "learning_rate": 1.304748954705538e-05, "loss": 0.3164, "step": 13637 }, { "epoch": 4.2, "learning_rate": 1.304654039453628e-05, "loss": 0.4115, "step": 13638 }, { "epoch": 4.2, "learning_rate": 1.3045591211762226e-05, "loss": 0.4589, "step": 13639 }, { "epoch": 4.2, "learning_rate": 1.3044641998742643e-05, "loss": 0.3989, "step": 13640 }, { "epoch": 4.2, "learning_rate": 1.3043692755486958e-05, "loss": 0.4362, "step": 13641 }, { "epoch": 4.2, "learning_rate": 1.30427434820046e-05, "loss": 0.3055, "step": 13642 }, { "epoch": 4.2, "learning_rate": 1.3041794178304994e-05, "loss": 0.2879, "step": 13643 }, { "epoch": 4.2, "learning_rate": 1.3040844844397573e-05, "loss": 0.3571, "step": 13644 }, { "epoch": 4.2, "learning_rate": 1.3039895480291755e-05, "loss": 0.2662, "step": 13645 }, { "epoch": 4.2, "learning_rate": 1.3038946085996976e-05, "loss": 0.3983, "step": 13646 }, { "epoch": 4.2, "learning_rate": 1.303799666152266e-05, "loss": 0.2668, "step": 13647 }, { "epoch": 4.2, "learning_rate": 1.3037047206878235e-05, "loss": 0.4151, "step": 13648 }, { "epoch": 4.2, "learning_rate": 1.3036097722073138e-05, "loss": 0.2918, "step": 13649 }, { "epoch": 4.2, "learning_rate": 1.3035148207116787e-05, "loss": 0.2696, "step": 13650 }, { "epoch": 4.2, "learning_rate": 1.303419866201862e-05, "loss": 0.2728, "step": 13651 }, { "epoch": 4.2, "learning_rate": 1.3033249086788063e-05, "loss": 0.4027, "step": 13652 }, { "epoch": 4.2, "learning_rate": 1.3032299481434546e-05, "loss": 0.2824, "step": 13653 }, { "epoch": 4.2, "learning_rate": 1.3031349845967504e-05, "loss": 0.3398, "step": 13654 }, { "epoch": 4.2, "learning_rate": 1.303040018039636e-05, "loss": 0.2493, "step": 13655 }, { "epoch": 4.2, "learning_rate": 1.3029450484730553e-05, "loss": 0.334, "step": 13656 }, { "epoch": 4.2, "learning_rate": 1.3028500758979507e-05, "loss": 0.2748, "step": 13657 }, { "epoch": 4.2, "learning_rate": 1.3027551003152661e-05, "loss": 0.3198, "step": 13658 }, { "epoch": 4.2, "learning_rate": 1.3026601217259442e-05, "loss": 0.3332, "step": 13659 }, { "epoch": 4.2, "learning_rate": 1.3025651401309283e-05, "loss": 0.168, "step": 13660 }, { "epoch": 4.2, "learning_rate": 1.3024701555311619e-05, "loss": 0.4084, "step": 13661 }, { "epoch": 4.2, "learning_rate": 1.3023751679275876e-05, "loss": 0.3048, "step": 13662 }, { "epoch": 4.2, "learning_rate": 1.3022801773211496e-05, "loss": 0.2068, "step": 13663 }, { "epoch": 4.2, "learning_rate": 1.3021851837127907e-05, "loss": 0.3485, "step": 13664 }, { "epoch": 4.2, "learning_rate": 1.3020901871034546e-05, "loss": 0.3817, "step": 13665 }, { "epoch": 4.2, "learning_rate": 1.3019951874940846e-05, "loss": 0.3777, "step": 13666 }, { "epoch": 4.21, "learning_rate": 1.3019001848856236e-05, "loss": 0.2568, "step": 13667 }, { "epoch": 4.21, "learning_rate": 1.3018051792790158e-05, "loss": 0.2966, "step": 13668 }, { "epoch": 4.21, "learning_rate": 1.3017101706752047e-05, "loss": 0.2858, "step": 13669 }, { "epoch": 4.21, "learning_rate": 1.3016151590751332e-05, "loss": 0.287, "step": 13670 }, { "epoch": 4.21, "learning_rate": 1.3015201444797451e-05, "loss": 0.449, "step": 13671 }, { "epoch": 4.21, "learning_rate": 1.3014251268899843e-05, "loss": 0.302, "step": 13672 }, { "epoch": 4.21, "learning_rate": 1.301330106306794e-05, "loss": 0.2121, "step": 13673 }, { "epoch": 4.21, "learning_rate": 1.3012350827311182e-05, "loss": 0.3124, "step": 13674 }, { "epoch": 4.21, "learning_rate": 1.3011400561639004e-05, "loss": 0.2816, "step": 13675 }, { "epoch": 4.21, "learning_rate": 1.3010450266060842e-05, "loss": 0.3683, "step": 13676 }, { "epoch": 4.21, "learning_rate": 1.3009499940586134e-05, "loss": 0.2569, "step": 13677 }, { "epoch": 4.21, "learning_rate": 1.300854958522432e-05, "loss": 0.258, "step": 13678 }, { "epoch": 4.21, "learning_rate": 1.3007599199984836e-05, "loss": 0.218, "step": 13679 }, { "epoch": 4.21, "learning_rate": 1.3006648784877119e-05, "loss": 0.3537, "step": 13680 }, { "epoch": 4.21, "learning_rate": 1.3005698339910606e-05, "loss": 0.3458, "step": 13681 }, { "epoch": 4.21, "learning_rate": 1.3004747865094742e-05, "loss": 0.2768, "step": 13682 }, { "epoch": 4.21, "learning_rate": 1.3003797360438961e-05, "loss": 0.2699, "step": 13683 }, { "epoch": 4.21, "learning_rate": 1.3002846825952704e-05, "loss": 0.2886, "step": 13684 }, { "epoch": 4.21, "learning_rate": 1.3001896261645412e-05, "loss": 0.4058, "step": 13685 }, { "epoch": 4.21, "learning_rate": 1.300094566752652e-05, "loss": 0.1588, "step": 13686 }, { "epoch": 4.21, "learning_rate": 1.2999995043605474e-05, "loss": 0.3636, "step": 13687 }, { "epoch": 4.21, "learning_rate": 1.2999044389891711e-05, "loss": 0.4479, "step": 13688 }, { "epoch": 4.21, "learning_rate": 1.2998093706394676e-05, "loss": 0.2987, "step": 13689 }, { "epoch": 4.21, "learning_rate": 1.2997142993123804e-05, "loss": 0.3596, "step": 13690 }, { "epoch": 4.21, "learning_rate": 1.2996192250088542e-05, "loss": 0.2996, "step": 13691 }, { "epoch": 4.21, "learning_rate": 1.299524147729833e-05, "loss": 0.2581, "step": 13692 }, { "epoch": 4.21, "learning_rate": 1.2994290674762607e-05, "loss": 0.2833, "step": 13693 }, { "epoch": 4.21, "learning_rate": 1.2993339842490823e-05, "loss": 0.3483, "step": 13694 }, { "epoch": 4.21, "learning_rate": 1.299238898049241e-05, "loss": 0.1215, "step": 13695 }, { "epoch": 4.21, "learning_rate": 1.2991438088776818e-05, "loss": 0.3449, "step": 13696 }, { "epoch": 4.21, "learning_rate": 1.2990487167353492e-05, "loss": 0.2854, "step": 13697 }, { "epoch": 4.21, "learning_rate": 1.2989536216231868e-05, "loss": 0.2763, "step": 13698 }, { "epoch": 4.21, "learning_rate": 1.2988585235421398e-05, "loss": 0.3625, "step": 13699 }, { "epoch": 4.22, "learning_rate": 1.2987634224931517e-05, "loss": 0.4741, "step": 13700 }, { "epoch": 4.22, "learning_rate": 1.298668318477168e-05, "loss": 0.2833, "step": 13701 }, { "epoch": 4.22, "learning_rate": 1.2985732114951323e-05, "loss": 0.2387, "step": 13702 }, { "epoch": 4.22, "learning_rate": 1.2984781015479894e-05, "loss": 0.2274, "step": 13703 }, { "epoch": 4.22, "learning_rate": 1.2983829886366845e-05, "loss": 0.3601, "step": 13704 }, { "epoch": 4.22, "learning_rate": 1.2982878727621609e-05, "loss": 0.2543, "step": 13705 }, { "epoch": 4.22, "learning_rate": 1.298192753925364e-05, "loss": 0.2727, "step": 13706 }, { "epoch": 4.22, "learning_rate": 1.2980976321272381e-05, "loss": 0.4676, "step": 13707 }, { "epoch": 4.22, "learning_rate": 1.2980025073687282e-05, "loss": 0.2379, "step": 13708 }, { "epoch": 4.22, "learning_rate": 1.2979073796507786e-05, "loss": 0.3131, "step": 13709 }, { "epoch": 4.22, "learning_rate": 1.2978122489743341e-05, "loss": 0.2834, "step": 13710 }, { "epoch": 4.22, "learning_rate": 1.2977171153403398e-05, "loss": 0.2989, "step": 13711 }, { "epoch": 4.22, "learning_rate": 1.2976219787497397e-05, "loss": 0.2269, "step": 13712 }, { "epoch": 4.22, "learning_rate": 1.2975268392034795e-05, "loss": 0.2896, "step": 13713 }, { "epoch": 4.22, "learning_rate": 1.2974316967025033e-05, "loss": 0.2189, "step": 13714 }, { "epoch": 4.22, "learning_rate": 1.2973365512477565e-05, "loss": 0.5512, "step": 13715 }, { "epoch": 4.22, "learning_rate": 1.2972414028401836e-05, "loss": 0.3414, "step": 13716 }, { "epoch": 4.22, "learning_rate": 1.2971462514807293e-05, "loss": 0.2983, "step": 13717 }, { "epoch": 4.22, "learning_rate": 1.2970510971703391e-05, "loss": 0.4083, "step": 13718 }, { "epoch": 4.22, "learning_rate": 1.2969559399099578e-05, "loss": 0.2362, "step": 13719 }, { "epoch": 4.22, "learning_rate": 1.2968607797005303e-05, "loss": 0.3148, "step": 13720 }, { "epoch": 4.22, "learning_rate": 1.2967656165430016e-05, "loss": 0.4041, "step": 13721 }, { "epoch": 4.22, "learning_rate": 1.296670450438317e-05, "loss": 0.3494, "step": 13722 }, { "epoch": 4.22, "learning_rate": 1.2965752813874214e-05, "loss": 0.3314, "step": 13723 }, { "epoch": 4.22, "learning_rate": 1.2964801093912595e-05, "loss": 0.2734, "step": 13724 }, { "epoch": 4.22, "learning_rate": 1.2963849344507775e-05, "loss": 0.3937, "step": 13725 }, { "epoch": 4.22, "learning_rate": 1.2962897565669193e-05, "loss": 0.2296, "step": 13726 }, { "epoch": 4.22, "learning_rate": 1.2961945757406311e-05, "loss": 0.402, "step": 13727 }, { "epoch": 4.22, "learning_rate": 1.296099391972858e-05, "loss": 0.2102, "step": 13728 }, { "epoch": 4.22, "learning_rate": 1.2960042052645447e-05, "loss": 0.3101, "step": 13729 }, { "epoch": 4.22, "learning_rate": 1.2959090156166369e-05, "loss": 0.1255, "step": 13730 }, { "epoch": 4.22, "learning_rate": 1.2958138230300799e-05, "loss": 0.2627, "step": 13731 }, { "epoch": 4.23, "learning_rate": 1.2957186275058188e-05, "loss": 0.2412, "step": 13732 }, { "epoch": 4.23, "learning_rate": 1.2956234290447995e-05, "loss": 0.219, "step": 13733 }, { "epoch": 4.23, "learning_rate": 1.295528227647967e-05, "loss": 0.4081, "step": 13734 }, { "epoch": 4.23, "learning_rate": 1.2954330233162669e-05, "loss": 0.3117, "step": 13735 }, { "epoch": 4.23, "learning_rate": 1.2953378160506445e-05, "loss": 0.2119, "step": 13736 }, { "epoch": 4.23, "learning_rate": 1.2952426058520454e-05, "loss": 0.2733, "step": 13737 }, { "epoch": 4.23, "learning_rate": 1.2951473927214149e-05, "loss": 0.3054, "step": 13738 }, { "epoch": 4.23, "learning_rate": 1.2950521766596993e-05, "loss": 0.2709, "step": 13739 }, { "epoch": 4.23, "learning_rate": 1.2949569576678432e-05, "loss": 0.155, "step": 13740 }, { "epoch": 4.23, "learning_rate": 1.2948617357467929e-05, "loss": 0.2881, "step": 13741 }, { "epoch": 4.23, "learning_rate": 1.2947665108974938e-05, "loss": 0.3398, "step": 13742 }, { "epoch": 4.23, "learning_rate": 1.2946712831208914e-05, "loss": 0.3304, "step": 13743 }, { "epoch": 4.23, "learning_rate": 1.2945760524179315e-05, "loss": 0.2572, "step": 13744 }, { "epoch": 4.23, "learning_rate": 1.2944808187895603e-05, "loss": 0.2733, "step": 13745 }, { "epoch": 4.23, "learning_rate": 1.2943855822367231e-05, "loss": 0.2928, "step": 13746 }, { "epoch": 4.23, "learning_rate": 1.2942903427603658e-05, "loss": 0.2381, "step": 13747 }, { "epoch": 4.23, "learning_rate": 1.2941951003614337e-05, "loss": 0.3394, "step": 13748 }, { "epoch": 4.23, "learning_rate": 1.2940998550408734e-05, "loss": 0.2294, "step": 13749 }, { "epoch": 4.23, "learning_rate": 1.2940046067996306e-05, "loss": 0.368, "step": 13750 }, { "epoch": 4.23, "learning_rate": 1.293909355638651e-05, "loss": 0.2629, "step": 13751 }, { "epoch": 4.23, "learning_rate": 1.2938141015588807e-05, "loss": 0.2034, "step": 13752 }, { "epoch": 4.23, "learning_rate": 1.2937188445612654e-05, "loss": 0.2287, "step": 13753 }, { "epoch": 4.23, "learning_rate": 1.2936235846467513e-05, "loss": 0.2827, "step": 13754 }, { "epoch": 4.23, "learning_rate": 1.2935283218162846e-05, "loss": 0.4256, "step": 13755 }, { "epoch": 4.23, "learning_rate": 1.2934330560708111e-05, "loss": 0.2672, "step": 13756 }, { "epoch": 4.23, "learning_rate": 1.2933377874112768e-05, "loss": 0.2467, "step": 13757 }, { "epoch": 4.23, "learning_rate": 1.2932425158386278e-05, "loss": 0.4094, "step": 13758 }, { "epoch": 4.23, "learning_rate": 1.2931472413538107e-05, "loss": 0.3728, "step": 13759 }, { "epoch": 4.23, "learning_rate": 1.293051963957771e-05, "loss": 0.4246, "step": 13760 }, { "epoch": 4.23, "learning_rate": 1.2929566836514556e-05, "loss": 0.2924, "step": 13761 }, { "epoch": 4.23, "learning_rate": 1.2928614004358098e-05, "loss": 0.4232, "step": 13762 }, { "epoch": 4.23, "learning_rate": 1.2927661143117807e-05, "loss": 0.3119, "step": 13763 }, { "epoch": 4.23, "learning_rate": 1.2926708252803142e-05, "loss": 0.2831, "step": 13764 }, { "epoch": 4.24, "learning_rate": 1.2925755333423569e-05, "loss": 0.2367, "step": 13765 }, { "epoch": 4.24, "learning_rate": 1.2924802384988546e-05, "loss": 0.2774, "step": 13766 }, { "epoch": 4.24, "learning_rate": 1.2923849407507541e-05, "loss": 0.2178, "step": 13767 }, { "epoch": 4.24, "learning_rate": 1.2922896400990017e-05, "loss": 0.3208, "step": 13768 }, { "epoch": 4.24, "learning_rate": 1.2921943365445435e-05, "loss": 0.2269, "step": 13769 }, { "epoch": 4.24, "learning_rate": 1.2920990300883267e-05, "loss": 0.3463, "step": 13770 }, { "epoch": 4.24, "learning_rate": 1.292003720731297e-05, "loss": 0.3355, "step": 13771 }, { "epoch": 4.24, "learning_rate": 1.2919084084744014e-05, "loss": 0.3667, "step": 13772 }, { "epoch": 4.24, "learning_rate": 1.2918130933185862e-05, "loss": 0.23, "step": 13773 }, { "epoch": 4.24, "learning_rate": 1.291717775264798e-05, "loss": 0.2548, "step": 13774 }, { "epoch": 4.24, "learning_rate": 1.2916224543139837e-05, "loss": 0.2797, "step": 13775 }, { "epoch": 4.24, "learning_rate": 1.2915271304670894e-05, "loss": 0.248, "step": 13776 }, { "epoch": 4.24, "learning_rate": 1.2914318037250619e-05, "loss": 0.4083, "step": 13777 }, { "epoch": 4.24, "learning_rate": 1.2913364740888483e-05, "loss": 0.2492, "step": 13778 }, { "epoch": 4.24, "learning_rate": 1.2912411415593948e-05, "loss": 0.2039, "step": 13779 }, { "epoch": 4.24, "learning_rate": 1.2911458061376487e-05, "loss": 0.3258, "step": 13780 }, { "epoch": 4.24, "learning_rate": 1.291050467824556e-05, "loss": 0.2078, "step": 13781 }, { "epoch": 4.24, "learning_rate": 1.2909551266210641e-05, "loss": 0.2505, "step": 13782 }, { "epoch": 4.24, "learning_rate": 1.2908597825281194e-05, "loss": 0.3249, "step": 13783 }, { "epoch": 4.24, "learning_rate": 1.2907644355466692e-05, "loss": 0.2773, "step": 13784 }, { "epoch": 4.24, "learning_rate": 1.2906690856776605e-05, "loss": 0.2063, "step": 13785 }, { "epoch": 4.24, "learning_rate": 1.2905737329220394e-05, "loss": 0.3143, "step": 13786 }, { "epoch": 4.24, "learning_rate": 1.2904783772807534e-05, "loss": 0.1883, "step": 13787 }, { "epoch": 4.24, "learning_rate": 1.2903830187547495e-05, "loss": 0.3504, "step": 13788 }, { "epoch": 4.24, "learning_rate": 1.2902876573449746e-05, "loss": 0.3768, "step": 13789 }, { "epoch": 4.24, "learning_rate": 1.2901922930523758e-05, "loss": 0.1668, "step": 13790 }, { "epoch": 4.24, "learning_rate": 1.2900969258779e-05, "loss": 0.2579, "step": 13791 }, { "epoch": 4.24, "learning_rate": 1.2900015558224946e-05, "loss": 0.2146, "step": 13792 }, { "epoch": 4.24, "learning_rate": 1.2899061828871058e-05, "loss": 0.3181, "step": 13793 }, { "epoch": 4.24, "learning_rate": 1.289810807072682e-05, "loss": 0.3076, "step": 13794 }, { "epoch": 4.24, "learning_rate": 1.2897154283801699e-05, "loss": 0.274, "step": 13795 }, { "epoch": 4.24, "learning_rate": 1.2896200468105162e-05, "loss": 0.1731, "step": 13796 }, { "epoch": 4.25, "learning_rate": 1.2895246623646687e-05, "loss": 0.4293, "step": 13797 }, { "epoch": 4.25, "learning_rate": 1.2894292750435744e-05, "loss": 0.2832, "step": 13798 }, { "epoch": 4.25, "learning_rate": 1.2893338848481807e-05, "loss": 0.2327, "step": 13799 }, { "epoch": 4.25, "learning_rate": 1.2892384917794347e-05, "loss": 0.2644, "step": 13800 }, { "epoch": 4.25, "learning_rate": 1.2891430958382842e-05, "loss": 0.2525, "step": 13801 }, { "epoch": 4.25, "learning_rate": 1.2890476970256762e-05, "loss": 0.3364, "step": 13802 }, { "epoch": 4.25, "learning_rate": 1.2889522953425579e-05, "loss": 0.245, "step": 13803 }, { "epoch": 4.25, "learning_rate": 1.2888568907898773e-05, "loss": 0.2484, "step": 13804 }, { "epoch": 4.25, "learning_rate": 1.2887614833685814e-05, "loss": 0.3405, "step": 13805 }, { "epoch": 4.25, "learning_rate": 1.2886660730796182e-05, "loss": 0.3176, "step": 13806 }, { "epoch": 4.25, "learning_rate": 1.2885706599239343e-05, "loss": 0.3469, "step": 13807 }, { "epoch": 4.25, "learning_rate": 1.2884752439024779e-05, "loss": 0.2553, "step": 13808 }, { "epoch": 4.25, "learning_rate": 1.2883798250161967e-05, "loss": 0.4181, "step": 13809 }, { "epoch": 4.25, "learning_rate": 1.288284403266038e-05, "loss": 0.2653, "step": 13810 }, { "epoch": 4.25, "learning_rate": 1.2881889786529495e-05, "loss": 0.3068, "step": 13811 }, { "epoch": 4.25, "learning_rate": 1.2880935511778785e-05, "loss": 0.2728, "step": 13812 }, { "epoch": 4.25, "learning_rate": 1.2879981208417735e-05, "loss": 0.3143, "step": 13813 }, { "epoch": 4.25, "learning_rate": 1.2879026876455815e-05, "loss": 0.4758, "step": 13814 }, { "epoch": 4.25, "learning_rate": 1.2878072515902506e-05, "loss": 0.1918, "step": 13815 }, { "epoch": 4.25, "learning_rate": 1.2877118126767285e-05, "loss": 0.2765, "step": 13816 }, { "epoch": 4.25, "learning_rate": 1.2876163709059626e-05, "loss": 0.369, "step": 13817 }, { "epoch": 4.25, "learning_rate": 1.2875209262789012e-05, "loss": 0.1117, "step": 13818 }, { "epoch": 4.25, "learning_rate": 1.2874254787964919e-05, "loss": 0.3974, "step": 13819 }, { "epoch": 4.25, "learning_rate": 1.287330028459683e-05, "loss": 0.2384, "step": 13820 }, { "epoch": 4.25, "learning_rate": 1.2872345752694223e-05, "loss": 0.4154, "step": 13821 }, { "epoch": 4.25, "learning_rate": 1.287139119226657e-05, "loss": 0.2393, "step": 13822 }, { "epoch": 4.25, "learning_rate": 1.2870436603323357e-05, "loss": 0.3175, "step": 13823 }, { "epoch": 4.25, "learning_rate": 1.2869481985874064e-05, "loss": 0.354, "step": 13824 }, { "epoch": 4.25, "learning_rate": 1.2868527339928173e-05, "loss": 0.2406, "step": 13825 }, { "epoch": 4.25, "learning_rate": 1.2867572665495156e-05, "loss": 0.2677, "step": 13826 }, { "epoch": 4.25, "learning_rate": 1.2866617962584504e-05, "loss": 0.2792, "step": 13827 }, { "epoch": 4.25, "learning_rate": 1.2865663231205693e-05, "loss": 0.3884, "step": 13828 }, { "epoch": 4.25, "learning_rate": 1.2864708471368203e-05, "loss": 0.2803, "step": 13829 }, { "epoch": 4.26, "learning_rate": 1.2863753683081523e-05, "loss": 0.3415, "step": 13830 }, { "epoch": 4.26, "learning_rate": 1.2862798866355123e-05, "loss": 0.3875, "step": 13831 }, { "epoch": 4.26, "learning_rate": 1.2861844021198497e-05, "loss": 0.1735, "step": 13832 }, { "epoch": 4.26, "learning_rate": 1.2860889147621121e-05, "loss": 0.3752, "step": 13833 }, { "epoch": 4.26, "learning_rate": 1.2859934245632477e-05, "loss": 0.3429, "step": 13834 }, { "epoch": 4.26, "learning_rate": 1.2858979315242052e-05, "loss": 0.3486, "step": 13835 }, { "epoch": 4.26, "learning_rate": 1.2858024356459327e-05, "loss": 0.2315, "step": 13836 }, { "epoch": 4.26, "learning_rate": 1.2857069369293784e-05, "loss": 0.1861, "step": 13837 }, { "epoch": 4.26, "learning_rate": 1.2856114353754912e-05, "loss": 0.3795, "step": 13838 }, { "epoch": 4.26, "learning_rate": 1.285515930985219e-05, "loss": 0.2345, "step": 13839 }, { "epoch": 4.26, "learning_rate": 1.2854204237595107e-05, "loss": 0.3105, "step": 13840 }, { "epoch": 4.26, "learning_rate": 1.2853249136993144e-05, "loss": 0.317, "step": 13841 }, { "epoch": 4.26, "learning_rate": 1.2852294008055787e-05, "loss": 0.2311, "step": 13842 }, { "epoch": 4.26, "learning_rate": 1.2851338850792522e-05, "loss": 0.2687, "step": 13843 }, { "epoch": 4.26, "learning_rate": 1.2850383665212836e-05, "loss": 0.3355, "step": 13844 }, { "epoch": 4.26, "learning_rate": 1.2849428451326213e-05, "loss": 0.405, "step": 13845 }, { "epoch": 4.26, "learning_rate": 1.2848473209142138e-05, "loss": 0.334, "step": 13846 }, { "epoch": 4.26, "learning_rate": 1.28475179386701e-05, "loss": 0.3359, "step": 13847 }, { "epoch": 4.26, "learning_rate": 1.2846562639919583e-05, "loss": 0.286, "step": 13848 }, { "epoch": 4.26, "learning_rate": 1.2845607312900075e-05, "loss": 0.3516, "step": 13849 }, { "epoch": 4.26, "learning_rate": 1.2844651957621066e-05, "loss": 0.2004, "step": 13850 }, { "epoch": 4.26, "learning_rate": 1.284369657409204e-05, "loss": 0.1792, "step": 13851 }, { "epoch": 4.26, "learning_rate": 1.2842741162322487e-05, "loss": 0.2472, "step": 13852 }, { "epoch": 4.26, "learning_rate": 1.2841785722321893e-05, "loss": 0.3488, "step": 13853 }, { "epoch": 4.26, "learning_rate": 1.284083025409975e-05, "loss": 0.4344, "step": 13854 }, { "epoch": 4.26, "learning_rate": 1.283987475766554e-05, "loss": 0.3862, "step": 13855 }, { "epoch": 4.26, "learning_rate": 1.283891923302876e-05, "loss": 0.239, "step": 13856 }, { "epoch": 4.26, "learning_rate": 1.2837963680198895e-05, "loss": 0.4234, "step": 13857 }, { "epoch": 4.26, "learning_rate": 1.2837008099185435e-05, "loss": 0.2783, "step": 13858 }, { "epoch": 4.26, "learning_rate": 1.2836052489997871e-05, "loss": 0.2007, "step": 13859 }, { "epoch": 4.26, "learning_rate": 1.283509685264569e-05, "loss": 0.2696, "step": 13860 }, { "epoch": 4.26, "learning_rate": 1.2834141187138387e-05, "loss": 0.3859, "step": 13861 }, { "epoch": 4.27, "learning_rate": 1.2833185493485447e-05, "loss": 0.2353, "step": 13862 }, { "epoch": 4.27, "learning_rate": 1.2832229771696365e-05, "loss": 0.2734, "step": 13863 }, { "epoch": 4.27, "learning_rate": 1.2831274021780633e-05, "loss": 0.2465, "step": 13864 }, { "epoch": 4.27, "learning_rate": 1.2830318243747736e-05, "loss": 0.3549, "step": 13865 }, { "epoch": 4.27, "learning_rate": 1.2829362437607176e-05, "loss": 0.5605, "step": 13866 }, { "epoch": 4.27, "learning_rate": 1.2828406603368435e-05, "loss": 0.4098, "step": 13867 }, { "epoch": 4.27, "learning_rate": 1.2827450741041013e-05, "loss": 0.2091, "step": 13868 }, { "epoch": 4.27, "learning_rate": 1.2826494850634394e-05, "loss": 0.2146, "step": 13869 }, { "epoch": 4.27, "learning_rate": 1.2825538932158081e-05, "loss": 0.4191, "step": 13870 }, { "epoch": 4.27, "learning_rate": 1.2824582985621563e-05, "loss": 0.317, "step": 13871 }, { "epoch": 4.27, "learning_rate": 1.282362701103433e-05, "loss": 0.3898, "step": 13872 }, { "epoch": 4.27, "learning_rate": 1.282267100840588e-05, "loss": 0.2535, "step": 13873 }, { "epoch": 4.27, "learning_rate": 1.2821714977745702e-05, "loss": 0.3419, "step": 13874 }, { "epoch": 4.27, "learning_rate": 1.2820758919063297e-05, "loss": 0.3236, "step": 13875 }, { "epoch": 4.27, "learning_rate": 1.2819802832368155e-05, "loss": 0.3061, "step": 13876 }, { "epoch": 4.27, "learning_rate": 1.2818846717669775e-05, "loss": 0.3362, "step": 13877 }, { "epoch": 4.27, "learning_rate": 1.2817890574977648e-05, "loss": 0.2646, "step": 13878 }, { "epoch": 4.27, "learning_rate": 1.2816934404301268e-05, "loss": 0.2826, "step": 13879 }, { "epoch": 4.27, "learning_rate": 1.2815978205650135e-05, "loss": 0.3403, "step": 13880 }, { "epoch": 4.27, "learning_rate": 1.2815021979033741e-05, "loss": 0.2854, "step": 13881 }, { "epoch": 4.27, "learning_rate": 1.2814065724461587e-05, "loss": 0.4431, "step": 13882 }, { "epoch": 4.27, "learning_rate": 1.2813109441943166e-05, "loss": 0.2448, "step": 13883 }, { "epoch": 4.27, "learning_rate": 1.2812153131487974e-05, "loss": 0.2902, "step": 13884 }, { "epoch": 4.27, "learning_rate": 1.2811196793105511e-05, "loss": 0.2437, "step": 13885 }, { "epoch": 4.27, "learning_rate": 1.2810240426805273e-05, "loss": 0.2523, "step": 13886 }, { "epoch": 4.27, "learning_rate": 1.2809284032596761e-05, "loss": 0.2031, "step": 13887 }, { "epoch": 4.27, "learning_rate": 1.2808327610489464e-05, "loss": 0.2342, "step": 13888 }, { "epoch": 4.27, "learning_rate": 1.2807371160492887e-05, "loss": 0.2843, "step": 13889 }, { "epoch": 4.27, "learning_rate": 1.2806414682616529e-05, "loss": 0.2782, "step": 13890 }, { "epoch": 4.27, "learning_rate": 1.2805458176869885e-05, "loss": 0.1511, "step": 13891 }, { "epoch": 4.27, "learning_rate": 1.2804501643262457e-05, "loss": 0.2773, "step": 13892 }, { "epoch": 4.27, "learning_rate": 1.2803545081803742e-05, "loss": 0.3262, "step": 13893 }, { "epoch": 4.27, "learning_rate": 1.280258849250324e-05, "loss": 0.2198, "step": 13894 }, { "epoch": 4.28, "learning_rate": 1.2801631875370452e-05, "loss": 0.2853, "step": 13895 }, { "epoch": 4.28, "learning_rate": 1.2800675230414879e-05, "loss": 0.38, "step": 13896 }, { "epoch": 4.28, "learning_rate": 1.2799718557646018e-05, "loss": 0.304, "step": 13897 }, { "epoch": 4.28, "learning_rate": 1.279876185707337e-05, "loss": 0.287, "step": 13898 }, { "epoch": 4.28, "learning_rate": 1.279780512870644e-05, "loss": 0.325, "step": 13899 }, { "epoch": 4.28, "learning_rate": 1.2796848372554727e-05, "loss": 0.412, "step": 13900 }, { "epoch": 4.28, "learning_rate": 1.279589158862773e-05, "loss": 0.2388, "step": 13901 }, { "epoch": 4.28, "learning_rate": 1.2794934776934953e-05, "loss": 0.3663, "step": 13902 }, { "epoch": 4.28, "learning_rate": 1.2793977937485897e-05, "loss": 0.3397, "step": 13903 }, { "epoch": 4.28, "learning_rate": 1.2793021070290065e-05, "loss": 0.3115, "step": 13904 }, { "epoch": 4.28, "learning_rate": 1.279206417535696e-05, "loss": 0.2675, "step": 13905 }, { "epoch": 4.28, "learning_rate": 1.2791107252696089e-05, "loss": 0.2182, "step": 13906 }, { "epoch": 4.28, "learning_rate": 1.2790150302316943e-05, "loss": 0.1838, "step": 13907 }, { "epoch": 4.28, "learning_rate": 1.2789193324229039e-05, "loss": 0.3722, "step": 13908 }, { "epoch": 4.28, "learning_rate": 1.2788236318441873e-05, "loss": 0.3259, "step": 13909 }, { "epoch": 4.28, "learning_rate": 1.2787279284964948e-05, "loss": 0.3611, "step": 13910 }, { "epoch": 4.28, "learning_rate": 1.2786322223807777e-05, "loss": 0.3351, "step": 13911 }, { "epoch": 4.28, "learning_rate": 1.2785365134979852e-05, "loss": 0.3033, "step": 13912 }, { "epoch": 4.28, "learning_rate": 1.2784408018490688e-05, "loss": 0.2126, "step": 13913 }, { "epoch": 4.28, "learning_rate": 1.2783450874349786e-05, "loss": 0.221, "step": 13914 }, { "epoch": 4.28, "learning_rate": 1.2782493702566648e-05, "loss": 0.1929, "step": 13915 }, { "epoch": 4.28, "learning_rate": 1.2781536503150789e-05, "loss": 0.2657, "step": 13916 }, { "epoch": 4.28, "learning_rate": 1.2780579276111702e-05, "loss": 0.3932, "step": 13917 }, { "epoch": 4.28, "learning_rate": 1.2779622021458904e-05, "loss": 0.2668, "step": 13918 }, { "epoch": 4.28, "learning_rate": 1.2778664739201896e-05, "loss": 0.331, "step": 13919 }, { "epoch": 4.28, "learning_rate": 1.2777707429350188e-05, "loss": 0.2394, "step": 13920 }, { "epoch": 4.28, "learning_rate": 1.2776750091913286e-05, "loss": 0.2866, "step": 13921 }, { "epoch": 4.28, "learning_rate": 1.2775792726900696e-05, "loss": 0.238, "step": 13922 }, { "epoch": 4.28, "learning_rate": 1.2774835334321925e-05, "loss": 0.2884, "step": 13923 }, { "epoch": 4.28, "learning_rate": 1.2773877914186478e-05, "loss": 0.2388, "step": 13924 }, { "epoch": 4.28, "learning_rate": 1.2772920466503872e-05, "loss": 0.2857, "step": 13925 }, { "epoch": 4.28, "learning_rate": 1.2771962991283609e-05, "loss": 0.3193, "step": 13926 }, { "epoch": 4.29, "learning_rate": 1.27710054885352e-05, "loss": 0.1926, "step": 13927 }, { "epoch": 4.29, "learning_rate": 1.277004795826815e-05, "loss": 0.2391, "step": 13928 }, { "epoch": 4.29, "learning_rate": 1.2769090400491972e-05, "loss": 0.2783, "step": 13929 }, { "epoch": 4.29, "learning_rate": 1.2768132815216174e-05, "loss": 0.2783, "step": 13930 }, { "epoch": 4.29, "learning_rate": 1.2767175202450268e-05, "loss": 0.2621, "step": 13931 }, { "epoch": 4.29, "learning_rate": 1.276621756220376e-05, "loss": 0.4658, "step": 13932 }, { "epoch": 4.29, "learning_rate": 1.2765259894486165e-05, "loss": 0.4998, "step": 13933 }, { "epoch": 4.29, "learning_rate": 1.2764302199306984e-05, "loss": 0.33, "step": 13934 }, { "epoch": 4.29, "learning_rate": 1.2763344476675739e-05, "loss": 0.3182, "step": 13935 }, { "epoch": 4.29, "learning_rate": 1.2762386726601937e-05, "loss": 0.1755, "step": 13936 }, { "epoch": 4.29, "learning_rate": 1.2761428949095089e-05, "loss": 0.1794, "step": 13937 }, { "epoch": 4.29, "learning_rate": 1.2760471144164703e-05, "loss": 0.2335, "step": 13938 }, { "epoch": 4.29, "learning_rate": 1.2759513311820298e-05, "loss": 0.2595, "step": 13939 }, { "epoch": 4.29, "learning_rate": 1.2758555452071381e-05, "loss": 0.387, "step": 13940 }, { "epoch": 4.29, "learning_rate": 1.2757597564927467e-05, "loss": 0.2544, "step": 13941 }, { "epoch": 4.29, "learning_rate": 1.2756639650398068e-05, "loss": 0.1991, "step": 13942 }, { "epoch": 4.29, "learning_rate": 1.2755681708492696e-05, "loss": 0.1978, "step": 13943 }, { "epoch": 4.29, "learning_rate": 1.2754723739220864e-05, "loss": 0.328, "step": 13944 }, { "epoch": 4.29, "learning_rate": 1.2753765742592088e-05, "loss": 0.2418, "step": 13945 }, { "epoch": 4.29, "learning_rate": 1.2752807718615878e-05, "loss": 0.3451, "step": 13946 }, { "epoch": 4.29, "learning_rate": 1.2751849667301754e-05, "loss": 0.3155, "step": 13947 }, { "epoch": 4.29, "learning_rate": 1.2750891588659222e-05, "loss": 0.2517, "step": 13948 }, { "epoch": 4.29, "learning_rate": 1.2749933482697805e-05, "loss": 0.1831, "step": 13949 }, { "epoch": 4.29, "learning_rate": 1.2748975349427014e-05, "loss": 0.2232, "step": 13950 }, { "epoch": 4.29, "learning_rate": 1.274801718885636e-05, "loss": 0.2704, "step": 13951 }, { "epoch": 4.29, "learning_rate": 1.274705900099537e-05, "loss": 0.2721, "step": 13952 }, { "epoch": 4.29, "learning_rate": 1.2746100785853546e-05, "loss": 0.2832, "step": 13953 }, { "epoch": 4.29, "learning_rate": 1.2745142543440414e-05, "loss": 0.184, "step": 13954 }, { "epoch": 4.29, "learning_rate": 1.2744184273765482e-05, "loss": 0.3232, "step": 13955 }, { "epoch": 4.29, "learning_rate": 1.2743225976838277e-05, "loss": 0.5012, "step": 13956 }, { "epoch": 4.29, "learning_rate": 1.2742267652668305e-05, "loss": 0.2694, "step": 13957 }, { "epoch": 4.29, "learning_rate": 1.274130930126509e-05, "loss": 0.2345, "step": 13958 }, { "epoch": 4.29, "learning_rate": 1.2740350922638147e-05, "loss": 0.2745, "step": 13959 }, { "epoch": 4.3, "learning_rate": 1.2739392516796994e-05, "loss": 0.2185, "step": 13960 }, { "epoch": 4.3, "learning_rate": 1.2738434083751148e-05, "loss": 0.3602, "step": 13961 }, { "epoch": 4.3, "learning_rate": 1.2737475623510127e-05, "loss": 0.331, "step": 13962 }, { "epoch": 4.3, "learning_rate": 1.2736517136083453e-05, "loss": 0.1885, "step": 13963 }, { "epoch": 4.3, "learning_rate": 1.2735558621480639e-05, "loss": 0.2076, "step": 13964 }, { "epoch": 4.3, "learning_rate": 1.2734600079711208e-05, "loss": 0.3051, "step": 13965 }, { "epoch": 4.3, "learning_rate": 1.2733641510784676e-05, "loss": 0.2498, "step": 13966 }, { "epoch": 4.3, "learning_rate": 1.2732682914710568e-05, "loss": 0.2748, "step": 13967 }, { "epoch": 4.3, "learning_rate": 1.2731724291498397e-05, "loss": 0.2967, "step": 13968 }, { "epoch": 4.3, "learning_rate": 1.2730765641157689e-05, "loss": 0.4155, "step": 13969 }, { "epoch": 4.3, "learning_rate": 1.2729806963697957e-05, "loss": 0.4279, "step": 13970 }, { "epoch": 4.3, "learning_rate": 1.272884825912873e-05, "loss": 0.3699, "step": 13971 }, { "epoch": 4.3, "learning_rate": 1.2727889527459523e-05, "loss": 0.2982, "step": 13972 }, { "epoch": 4.3, "learning_rate": 1.272693076869986e-05, "loss": 0.3926, "step": 13973 }, { "epoch": 4.3, "learning_rate": 1.272597198285926e-05, "loss": 0.3164, "step": 13974 }, { "epoch": 4.3, "learning_rate": 1.2725013169947247e-05, "loss": 0.2519, "step": 13975 }, { "epoch": 4.3, "learning_rate": 1.2724054329973341e-05, "loss": 0.2381, "step": 13976 }, { "epoch": 4.3, "learning_rate": 1.2723095462947067e-05, "loss": 0.1935, "step": 13977 }, { "epoch": 4.3, "learning_rate": 1.2722136568877942e-05, "loss": 0.3931, "step": 13978 }, { "epoch": 4.3, "learning_rate": 1.2721177647775494e-05, "loss": 0.2275, "step": 13979 }, { "epoch": 4.3, "learning_rate": 1.2720218699649243e-05, "loss": 0.2398, "step": 13980 }, { "epoch": 4.3, "learning_rate": 1.2719259724508715e-05, "loss": 0.3306, "step": 13981 }, { "epoch": 4.3, "learning_rate": 1.2718300722363431e-05, "loss": 0.1671, "step": 13982 }, { "epoch": 4.3, "learning_rate": 1.2717341693222916e-05, "loss": 0.3007, "step": 13983 }, { "epoch": 4.3, "learning_rate": 1.2716382637096692e-05, "loss": 0.2954, "step": 13984 }, { "epoch": 4.3, "learning_rate": 1.2715423553994284e-05, "loss": 0.3721, "step": 13985 }, { "epoch": 4.3, "learning_rate": 1.2714464443925219e-05, "loss": 0.2355, "step": 13986 }, { "epoch": 4.3, "learning_rate": 1.2713505306899024e-05, "loss": 0.1977, "step": 13987 }, { "epoch": 4.3, "learning_rate": 1.2712546142925215e-05, "loss": 0.2174, "step": 13988 }, { "epoch": 4.3, "learning_rate": 1.2711586952013323e-05, "loss": 0.2667, "step": 13989 }, { "epoch": 4.3, "learning_rate": 1.2710627734172879e-05, "loss": 0.467, "step": 13990 }, { "epoch": 4.3, "learning_rate": 1.2709668489413398e-05, "loss": 0.2886, "step": 13991 }, { "epoch": 4.31, "learning_rate": 1.2708709217744416e-05, "loss": 0.3621, "step": 13992 }, { "epoch": 4.31, "learning_rate": 1.2707749919175449e-05, "loss": 0.3548, "step": 13993 }, { "epoch": 4.31, "learning_rate": 1.2706790593716036e-05, "loss": 0.2862, "step": 13994 }, { "epoch": 4.31, "learning_rate": 1.2705831241375695e-05, "loss": 0.4427, "step": 13995 }, { "epoch": 4.31, "learning_rate": 1.2704871862163954e-05, "loss": 0.3216, "step": 13996 }, { "epoch": 4.31, "learning_rate": 1.2703912456090349e-05, "loss": 0.2555, "step": 13997 }, { "epoch": 4.31, "learning_rate": 1.2702953023164394e-05, "loss": 0.2187, "step": 13998 }, { "epoch": 4.31, "learning_rate": 1.2701993563395627e-05, "loss": 0.2171, "step": 13999 }, { "epoch": 4.31, "learning_rate": 1.2701034076793575e-05, "loss": 0.2891, "step": 14000 }, { "epoch": 4.31, "learning_rate": 1.2700074563367764e-05, "loss": 0.412, "step": 14001 }, { "epoch": 4.31, "learning_rate": 1.2699115023127726e-05, "loss": 0.329, "step": 14002 }, { "epoch": 4.31, "learning_rate": 1.2698155456082986e-05, "loss": 0.2844, "step": 14003 }, { "epoch": 4.31, "learning_rate": 1.2697195862243076e-05, "loss": 0.2496, "step": 14004 }, { "epoch": 4.31, "learning_rate": 1.2696236241617523e-05, "loss": 0.3058, "step": 14005 }, { "epoch": 4.31, "learning_rate": 1.2695276594215864e-05, "loss": 0.3371, "step": 14006 }, { "epoch": 4.31, "learning_rate": 1.269431692004762e-05, "loss": 0.2967, "step": 14007 }, { "epoch": 4.31, "learning_rate": 1.2693357219122331e-05, "loss": 0.2503, "step": 14008 }, { "epoch": 4.31, "learning_rate": 1.2692397491449518e-05, "loss": 0.3888, "step": 14009 }, { "epoch": 4.31, "learning_rate": 1.2691437737038717e-05, "loss": 0.2866, "step": 14010 }, { "epoch": 4.31, "learning_rate": 1.2690477955899461e-05, "loss": 0.4625, "step": 14011 }, { "epoch": 4.31, "learning_rate": 1.2689518148041278e-05, "loss": 0.3177, "step": 14012 }, { "epoch": 4.31, "learning_rate": 1.26885583134737e-05, "loss": 0.3081, "step": 14013 }, { "epoch": 4.31, "learning_rate": 1.2687598452206262e-05, "loss": 0.2209, "step": 14014 }, { "epoch": 4.31, "learning_rate": 1.2686638564248493e-05, "loss": 0.2226, "step": 14015 }, { "epoch": 4.31, "learning_rate": 1.2685678649609928e-05, "loss": 0.4053, "step": 14016 }, { "epoch": 4.31, "learning_rate": 1.2684718708300099e-05, "loss": 0.3704, "step": 14017 }, { "epoch": 4.31, "learning_rate": 1.268375874032854e-05, "loss": 0.2657, "step": 14018 }, { "epoch": 4.31, "learning_rate": 1.268279874570478e-05, "loss": 0.3617, "step": 14019 }, { "epoch": 4.31, "learning_rate": 1.2681838724438355e-05, "loss": 0.2722, "step": 14020 }, { "epoch": 4.31, "learning_rate": 1.2680878676538804e-05, "loss": 0.3508, "step": 14021 }, { "epoch": 4.31, "learning_rate": 1.2679918602015655e-05, "loss": 0.3776, "step": 14022 }, { "epoch": 4.31, "learning_rate": 1.2678958500878446e-05, "loss": 0.3407, "step": 14023 }, { "epoch": 4.31, "learning_rate": 1.2677998373136708e-05, "loss": 0.1871, "step": 14024 }, { "epoch": 4.32, "learning_rate": 1.2677038218799979e-05, "loss": 0.2685, "step": 14025 }, { "epoch": 4.32, "learning_rate": 1.2676078037877794e-05, "loss": 0.5573, "step": 14026 }, { "epoch": 4.32, "learning_rate": 1.2675117830379686e-05, "loss": 0.4248, "step": 14027 }, { "epoch": 4.32, "learning_rate": 1.2674157596315195e-05, "loss": 0.3556, "step": 14028 }, { "epoch": 4.32, "learning_rate": 1.2673197335693849e-05, "loss": 0.3045, "step": 14029 }, { "epoch": 4.32, "learning_rate": 1.2672237048525196e-05, "loss": 0.1944, "step": 14030 }, { "epoch": 4.32, "learning_rate": 1.2671276734818766e-05, "loss": 0.2894, "step": 14031 }, { "epoch": 4.32, "learning_rate": 1.2670316394584093e-05, "loss": 0.2902, "step": 14032 }, { "epoch": 4.32, "learning_rate": 1.2669356027830717e-05, "loss": 0.388, "step": 14033 }, { "epoch": 4.32, "learning_rate": 1.2668395634568175e-05, "loss": 0.2799, "step": 14034 }, { "epoch": 4.32, "learning_rate": 1.266743521480601e-05, "loss": 0.371, "step": 14035 }, { "epoch": 4.32, "learning_rate": 1.266647476855375e-05, "loss": 0.218, "step": 14036 }, { "epoch": 4.32, "learning_rate": 1.2665514295820938e-05, "loss": 0.2063, "step": 14037 }, { "epoch": 4.32, "learning_rate": 1.2664553796617116e-05, "loss": 0.2524, "step": 14038 }, { "epoch": 4.32, "learning_rate": 1.2663593270951817e-05, "loss": 0.3459, "step": 14039 }, { "epoch": 4.32, "learning_rate": 1.2662632718834582e-05, "loss": 0.3925, "step": 14040 }, { "epoch": 4.32, "learning_rate": 1.2661672140274949e-05, "loss": 0.3119, "step": 14041 }, { "epoch": 4.32, "learning_rate": 1.266071153528246e-05, "loss": 0.2474, "step": 14042 }, { "epoch": 4.32, "learning_rate": 1.2659750903866654e-05, "loss": 0.2088, "step": 14043 }, { "epoch": 4.32, "learning_rate": 1.2658790246037071e-05, "loss": 0.3592, "step": 14044 }, { "epoch": 4.32, "learning_rate": 1.2657829561803249e-05, "loss": 0.313, "step": 14045 }, { "epoch": 4.32, "learning_rate": 1.2656868851174728e-05, "loss": 0.3644, "step": 14046 }, { "epoch": 4.32, "learning_rate": 1.2655908114161053e-05, "loss": 0.2128, "step": 14047 }, { "epoch": 4.32, "learning_rate": 1.2654947350771763e-05, "loss": 0.2775, "step": 14048 }, { "epoch": 4.32, "learning_rate": 1.2653986561016399e-05, "loss": 0.355, "step": 14049 }, { "epoch": 4.32, "learning_rate": 1.2653025744904502e-05, "loss": 0.332, "step": 14050 }, { "epoch": 4.32, "learning_rate": 1.265206490244561e-05, "loss": 0.2752, "step": 14051 }, { "epoch": 4.32, "learning_rate": 1.2651104033649272e-05, "loss": 0.2836, "step": 14052 }, { "epoch": 4.32, "learning_rate": 1.265014313852503e-05, "loss": 0.2806, "step": 14053 }, { "epoch": 4.32, "learning_rate": 1.2649182217082425e-05, "loss": 0.2931, "step": 14054 }, { "epoch": 4.32, "learning_rate": 1.2648221269330996e-05, "loss": 0.352, "step": 14055 }, { "epoch": 4.32, "learning_rate": 1.264726029528029e-05, "loss": 0.3419, "step": 14056 }, { "epoch": 4.33, "learning_rate": 1.264629929493985e-05, "loss": 0.3636, "step": 14057 }, { "epoch": 4.33, "learning_rate": 1.2645338268319219e-05, "loss": 0.2715, "step": 14058 }, { "epoch": 4.33, "learning_rate": 1.264437721542794e-05, "loss": 0.3719, "step": 14059 }, { "epoch": 4.33, "learning_rate": 1.2643416136275557e-05, "loss": 0.3465, "step": 14060 }, { "epoch": 4.33, "learning_rate": 1.2642455030871618e-05, "loss": 0.3311, "step": 14061 }, { "epoch": 4.33, "learning_rate": 1.2641493899225666e-05, "loss": 0.2804, "step": 14062 }, { "epoch": 4.33, "learning_rate": 1.2640532741347245e-05, "loss": 0.4053, "step": 14063 }, { "epoch": 4.33, "learning_rate": 1.2639571557245897e-05, "loss": 0.3143, "step": 14064 }, { "epoch": 4.33, "learning_rate": 1.2638610346931173e-05, "loss": 0.2805, "step": 14065 }, { "epoch": 4.33, "learning_rate": 1.2637649110412615e-05, "loss": 0.4112, "step": 14066 }, { "epoch": 4.33, "learning_rate": 1.2636687847699769e-05, "loss": 0.2781, "step": 14067 }, { "epoch": 4.33, "learning_rate": 1.2635726558802188e-05, "loss": 0.3363, "step": 14068 }, { "epoch": 4.33, "learning_rate": 1.263476524372941e-05, "loss": 0.2882, "step": 14069 }, { "epoch": 4.33, "learning_rate": 1.2633803902490984e-05, "loss": 0.2291, "step": 14070 }, { "epoch": 4.33, "learning_rate": 1.263284253509646e-05, "loss": 0.2683, "step": 14071 }, { "epoch": 4.33, "learning_rate": 1.263188114155538e-05, "loss": 0.3, "step": 14072 }, { "epoch": 4.33, "learning_rate": 1.2630919721877299e-05, "loss": 0.3726, "step": 14073 }, { "epoch": 4.33, "learning_rate": 1.2629958276071757e-05, "loss": 0.3654, "step": 14074 }, { "epoch": 4.33, "learning_rate": 1.2628996804148307e-05, "loss": 0.3593, "step": 14075 }, { "epoch": 4.33, "learning_rate": 1.2628035306116495e-05, "loss": 0.4252, "step": 14076 }, { "epoch": 4.33, "learning_rate": 1.262707378198587e-05, "loss": 0.2761, "step": 14077 }, { "epoch": 4.33, "learning_rate": 1.2626112231765984e-05, "loss": 0.3604, "step": 14078 }, { "epoch": 4.33, "learning_rate": 1.262515065546638e-05, "loss": 0.439, "step": 14079 }, { "epoch": 4.33, "learning_rate": 1.2624189053096611e-05, "loss": 0.2862, "step": 14080 }, { "epoch": 4.33, "learning_rate": 1.2623227424666226e-05, "loss": 0.3994, "step": 14081 }, { "epoch": 4.33, "learning_rate": 1.2622265770184776e-05, "loss": 0.2612, "step": 14082 }, { "epoch": 4.33, "learning_rate": 1.2621304089661812e-05, "loss": 0.3027, "step": 14083 }, { "epoch": 4.33, "learning_rate": 1.2620342383106881e-05, "loss": 0.203, "step": 14084 }, { "epoch": 4.33, "learning_rate": 1.2619380650529533e-05, "loss": 0.3318, "step": 14085 }, { "epoch": 4.33, "learning_rate": 1.261841889193932e-05, "loss": 0.3303, "step": 14086 }, { "epoch": 4.33, "learning_rate": 1.2617457107345798e-05, "loss": 0.2168, "step": 14087 }, { "epoch": 4.33, "learning_rate": 1.2616495296758515e-05, "loss": 0.2946, "step": 14088 }, { "epoch": 4.33, "learning_rate": 1.2615533460187022e-05, "loss": 0.2932, "step": 14089 }, { "epoch": 4.34, "learning_rate": 1.261457159764087e-05, "loss": 0.2313, "step": 14090 }, { "epoch": 4.34, "learning_rate": 1.261360970912961e-05, "loss": 0.326, "step": 14091 }, { "epoch": 4.34, "learning_rate": 1.26126477946628e-05, "loss": 0.4083, "step": 14092 }, { "epoch": 4.34, "learning_rate": 1.2611685854249989e-05, "loss": 0.4683, "step": 14093 }, { "epoch": 4.34, "learning_rate": 1.261072388790073e-05, "loss": 0.3199, "step": 14094 }, { "epoch": 4.34, "learning_rate": 1.2609761895624579e-05, "loss": 0.3832, "step": 14095 }, { "epoch": 4.34, "learning_rate": 1.2608799877431084e-05, "loss": 0.2843, "step": 14096 }, { "epoch": 4.34, "learning_rate": 1.2607837833329803e-05, "loss": 0.3902, "step": 14097 }, { "epoch": 4.34, "learning_rate": 1.260687576333029e-05, "loss": 0.3256, "step": 14098 }, { "epoch": 4.34, "learning_rate": 1.2605913667442096e-05, "loss": 0.2613, "step": 14099 }, { "epoch": 4.34, "learning_rate": 1.2604951545674779e-05, "loss": 0.2409, "step": 14100 }, { "epoch": 4.34, "learning_rate": 1.260398939803789e-05, "loss": 0.3786, "step": 14101 }, { "epoch": 4.34, "learning_rate": 1.2603027224540989e-05, "loss": 0.3808, "step": 14102 }, { "epoch": 4.34, "learning_rate": 1.2602065025193628e-05, "loss": 0.3548, "step": 14103 }, { "epoch": 4.34, "learning_rate": 1.2601102800005362e-05, "loss": 0.2587, "step": 14104 }, { "epoch": 4.34, "learning_rate": 1.2600140548985749e-05, "loss": 0.5226, "step": 14105 }, { "epoch": 4.34, "learning_rate": 1.2599178272144343e-05, "loss": 0.2013, "step": 14106 }, { "epoch": 4.34, "learning_rate": 1.2598215969490703e-05, "loss": 0.2686, "step": 14107 }, { "epoch": 4.34, "learning_rate": 1.2597253641034384e-05, "loss": 0.2361, "step": 14108 }, { "epoch": 4.34, "learning_rate": 1.2596291286784943e-05, "loss": 0.3515, "step": 14109 }, { "epoch": 4.34, "learning_rate": 1.2595328906751934e-05, "loss": 0.3172, "step": 14110 }, { "epoch": 4.34, "learning_rate": 1.2594366500944917e-05, "loss": 0.3463, "step": 14111 }, { "epoch": 4.34, "learning_rate": 1.2593404069373452e-05, "loss": 0.1231, "step": 14112 }, { "epoch": 4.34, "learning_rate": 1.2592441612047096e-05, "loss": 0.2665, "step": 14113 }, { "epoch": 4.34, "learning_rate": 1.2591479128975402e-05, "loss": 0.3316, "step": 14114 }, { "epoch": 4.34, "learning_rate": 1.2590516620167932e-05, "loss": 0.2236, "step": 14115 }, { "epoch": 4.34, "learning_rate": 1.2589554085634246e-05, "loss": 0.3987, "step": 14116 }, { "epoch": 4.34, "learning_rate": 1.2588591525383897e-05, "loss": 0.408, "step": 14117 }, { "epoch": 4.34, "learning_rate": 1.2587628939426457e-05, "loss": 0.2601, "step": 14118 }, { "epoch": 4.34, "learning_rate": 1.258666632777147e-05, "loss": 0.3467, "step": 14119 }, { "epoch": 4.34, "learning_rate": 1.2585703690428504e-05, "loss": 0.4096, "step": 14120 }, { "epoch": 4.34, "learning_rate": 1.2584741027407117e-05, "loss": 0.2227, "step": 14121 }, { "epoch": 4.35, "learning_rate": 1.258377833871687e-05, "loss": 0.3206, "step": 14122 }, { "epoch": 4.35, "learning_rate": 1.2582815624367327e-05, "loss": 0.3743, "step": 14123 }, { "epoch": 4.35, "learning_rate": 1.2581852884368036e-05, "loss": 0.3135, "step": 14124 }, { "epoch": 4.35, "learning_rate": 1.2580890118728572e-05, "loss": 0.2554, "step": 14125 }, { "epoch": 4.35, "learning_rate": 1.257992732745849e-05, "loss": 0.3594, "step": 14126 }, { "epoch": 4.35, "learning_rate": 1.2578964510567348e-05, "loss": 0.347, "step": 14127 }, { "epoch": 4.35, "learning_rate": 1.257800166806472e-05, "loss": 0.2832, "step": 14128 }, { "epoch": 4.35, "learning_rate": 1.257703879996015e-05, "loss": 0.3399, "step": 14129 }, { "epoch": 4.35, "learning_rate": 1.2576075906263215e-05, "loss": 0.268, "step": 14130 }, { "epoch": 4.35, "learning_rate": 1.2575112986983472e-05, "loss": 0.3315, "step": 14131 }, { "epoch": 4.35, "learning_rate": 1.2574150042130479e-05, "loss": 0.327, "step": 14132 }, { "epoch": 4.35, "learning_rate": 1.257318707171381e-05, "loss": 0.3223, "step": 14133 }, { "epoch": 4.35, "learning_rate": 1.2572224075743018e-05, "loss": 0.3596, "step": 14134 }, { "epoch": 4.35, "learning_rate": 1.2571261054227672e-05, "loss": 0.3337, "step": 14135 }, { "epoch": 4.35, "learning_rate": 1.2570298007177332e-05, "loss": 0.654, "step": 14136 }, { "epoch": 4.35, "learning_rate": 1.2569334934601566e-05, "loss": 0.3427, "step": 14137 }, { "epoch": 4.35, "learning_rate": 1.2568371836509936e-05, "loss": 0.2395, "step": 14138 }, { "epoch": 4.35, "learning_rate": 1.2567408712912007e-05, "loss": 0.205, "step": 14139 }, { "epoch": 4.35, "learning_rate": 1.2566445563817343e-05, "loss": 0.4019, "step": 14140 }, { "epoch": 4.35, "learning_rate": 1.2565482389235506e-05, "loss": 0.3932, "step": 14141 }, { "epoch": 4.35, "learning_rate": 1.2564519189176067e-05, "loss": 0.2194, "step": 14142 }, { "epoch": 4.35, "learning_rate": 1.256355596364859e-05, "loss": 0.4443, "step": 14143 }, { "epoch": 4.35, "learning_rate": 1.256259271266264e-05, "loss": 0.2842, "step": 14144 }, { "epoch": 4.35, "learning_rate": 1.256162943622778e-05, "loss": 0.258, "step": 14145 }, { "epoch": 4.35, "learning_rate": 1.2560666134353578e-05, "loss": 0.3523, "step": 14146 }, { "epoch": 4.35, "learning_rate": 1.2559702807049606e-05, "loss": 0.2675, "step": 14147 }, { "epoch": 4.35, "learning_rate": 1.2558739454325424e-05, "loss": 0.4297, "step": 14148 }, { "epoch": 4.35, "learning_rate": 1.25577760761906e-05, "loss": 0.2361, "step": 14149 }, { "epoch": 4.35, "learning_rate": 1.2556812672654704e-05, "loss": 0.5234, "step": 14150 }, { "epoch": 4.35, "learning_rate": 1.2555849243727298e-05, "loss": 0.3735, "step": 14151 }, { "epoch": 4.35, "learning_rate": 1.2554885789417957e-05, "loss": 0.2916, "step": 14152 }, { "epoch": 4.35, "learning_rate": 1.2553922309736242e-05, "loss": 0.27, "step": 14153 }, { "epoch": 4.35, "learning_rate": 1.255295880469173e-05, "loss": 0.2975, "step": 14154 }, { "epoch": 4.36, "learning_rate": 1.255199527429398e-05, "loss": 0.2586, "step": 14155 }, { "epoch": 4.36, "learning_rate": 1.2551031718552566e-05, "loss": 0.3562, "step": 14156 }, { "epoch": 4.36, "learning_rate": 1.2550068137477057e-05, "loss": 0.4266, "step": 14157 }, { "epoch": 4.36, "learning_rate": 1.254910453107702e-05, "loss": 0.3523, "step": 14158 }, { "epoch": 4.36, "learning_rate": 1.2548140899362027e-05, "loss": 0.2787, "step": 14159 }, { "epoch": 4.36, "learning_rate": 1.2547177242341643e-05, "loss": 0.3318, "step": 14160 }, { "epoch": 4.36, "learning_rate": 1.2546213560025444e-05, "loss": 0.1737, "step": 14161 }, { "epoch": 4.36, "learning_rate": 1.2545249852422998e-05, "loss": 0.2909, "step": 14162 }, { "epoch": 4.36, "learning_rate": 1.2544286119543874e-05, "loss": 0.2547, "step": 14163 }, { "epoch": 4.36, "learning_rate": 1.2543322361397648e-05, "loss": 0.3018, "step": 14164 }, { "epoch": 4.36, "learning_rate": 1.2542358577993881e-05, "loss": 0.4055, "step": 14165 }, { "epoch": 4.36, "learning_rate": 1.2541394769342152e-05, "loss": 0.2089, "step": 14166 }, { "epoch": 4.36, "learning_rate": 1.2540430935452035e-05, "loss": 0.2438, "step": 14167 }, { "epoch": 4.36, "learning_rate": 1.253946707633309e-05, "loss": 0.2924, "step": 14168 }, { "epoch": 4.36, "learning_rate": 1.2538503191994903e-05, "loss": 0.1495, "step": 14169 }, { "epoch": 4.36, "learning_rate": 1.2537539282447036e-05, "loss": 0.3392, "step": 14170 }, { "epoch": 4.36, "learning_rate": 1.2536575347699067e-05, "loss": 0.1445, "step": 14171 }, { "epoch": 4.36, "learning_rate": 1.2535611387760564e-05, "loss": 0.271, "step": 14172 }, { "epoch": 4.36, "learning_rate": 1.2534647402641107e-05, "loss": 0.4278, "step": 14173 }, { "epoch": 4.36, "learning_rate": 1.2533683392350264e-05, "loss": 0.3296, "step": 14174 }, { "epoch": 4.36, "learning_rate": 1.2532719356897612e-05, "loss": 0.3479, "step": 14175 }, { "epoch": 4.36, "learning_rate": 1.253175529629272e-05, "loss": 0.1533, "step": 14176 }, { "epoch": 4.36, "learning_rate": 1.2530791210545163e-05, "loss": 0.3019, "step": 14177 }, { "epoch": 4.36, "learning_rate": 1.2529827099664518e-05, "loss": 0.3699, "step": 14178 }, { "epoch": 4.36, "learning_rate": 1.252886296366036e-05, "loss": 0.2576, "step": 14179 }, { "epoch": 4.36, "learning_rate": 1.2527898802542262e-05, "loss": 0.2595, "step": 14180 }, { "epoch": 4.36, "learning_rate": 1.2526934616319799e-05, "loss": 0.3847, "step": 14181 }, { "epoch": 4.36, "learning_rate": 1.2525970405002544e-05, "loss": 0.2753, "step": 14182 }, { "epoch": 4.36, "learning_rate": 1.2525006168600076e-05, "loss": 0.3736, "step": 14183 }, { "epoch": 4.36, "learning_rate": 1.252404190712197e-05, "loss": 0.3193, "step": 14184 }, { "epoch": 4.36, "learning_rate": 1.2523077620577804e-05, "loss": 0.3337, "step": 14185 }, { "epoch": 4.36, "learning_rate": 1.2522113308977147e-05, "loss": 0.2928, "step": 14186 }, { "epoch": 4.37, "learning_rate": 1.2521148972329584e-05, "loss": 0.2652, "step": 14187 }, { "epoch": 4.37, "learning_rate": 1.2520184610644688e-05, "loss": 0.2971, "step": 14188 }, { "epoch": 4.37, "learning_rate": 1.2519220223932037e-05, "loss": 0.2235, "step": 14189 }, { "epoch": 4.37, "learning_rate": 1.2518255812201203e-05, "loss": 0.284, "step": 14190 }, { "epoch": 4.37, "learning_rate": 1.251729137546177e-05, "loss": 0.1504, "step": 14191 }, { "epoch": 4.37, "learning_rate": 1.2516326913723316e-05, "loss": 0.3881, "step": 14192 }, { "epoch": 4.37, "learning_rate": 1.2515362426995414e-05, "loss": 0.3486, "step": 14193 }, { "epoch": 4.37, "learning_rate": 1.2514397915287647e-05, "loss": 0.2495, "step": 14194 }, { "epoch": 4.37, "learning_rate": 1.251343337860959e-05, "loss": 0.3936, "step": 14195 }, { "epoch": 4.37, "learning_rate": 1.2512468816970821e-05, "loss": 0.4877, "step": 14196 }, { "epoch": 4.37, "learning_rate": 1.2511504230380925e-05, "loss": 0.541, "step": 14197 }, { "epoch": 4.37, "learning_rate": 1.2510539618849476e-05, "loss": 0.3591, "step": 14198 }, { "epoch": 4.37, "learning_rate": 1.2509574982386054e-05, "loss": 0.3665, "step": 14199 }, { "epoch": 4.37, "learning_rate": 1.250861032100024e-05, "loss": 0.3827, "step": 14200 }, { "epoch": 4.37, "learning_rate": 1.2507645634701613e-05, "loss": 0.3074, "step": 14201 }, { "epoch": 4.37, "learning_rate": 1.2506680923499755e-05, "loss": 0.2783, "step": 14202 }, { "epoch": 4.37, "learning_rate": 1.2505716187404242e-05, "loss": 0.3344, "step": 14203 }, { "epoch": 4.37, "learning_rate": 1.2504751426424662e-05, "loss": 0.1538, "step": 14204 }, { "epoch": 4.37, "learning_rate": 1.250378664057059e-05, "loss": 0.2775, "step": 14205 }, { "epoch": 4.37, "learning_rate": 1.2502821829851608e-05, "loss": 0.3444, "step": 14206 }, { "epoch": 4.37, "learning_rate": 1.2501856994277302e-05, "loss": 0.4019, "step": 14207 }, { "epoch": 4.37, "learning_rate": 1.2500892133857246e-05, "loss": 0.3167, "step": 14208 }, { "epoch": 4.37, "learning_rate": 1.2499927248601032e-05, "loss": 0.2672, "step": 14209 }, { "epoch": 4.37, "learning_rate": 1.249896233851823e-05, "loss": 0.3831, "step": 14210 }, { "epoch": 4.37, "learning_rate": 1.2497997403618431e-05, "loss": 0.334, "step": 14211 }, { "epoch": 4.37, "learning_rate": 1.249703244391122e-05, "loss": 0.4115, "step": 14212 }, { "epoch": 4.37, "learning_rate": 1.2496067459406168e-05, "loss": 0.2879, "step": 14213 }, { "epoch": 4.37, "learning_rate": 1.2495102450112873e-05, "loss": 0.2331, "step": 14214 }, { "epoch": 4.37, "learning_rate": 1.2494137416040904e-05, "loss": 0.2203, "step": 14215 }, { "epoch": 4.37, "learning_rate": 1.2493172357199856e-05, "loss": 0.3119, "step": 14216 }, { "epoch": 4.37, "learning_rate": 1.2492207273599309e-05, "loss": 0.265, "step": 14217 }, { "epoch": 4.37, "learning_rate": 1.2491242165248845e-05, "loss": 0.321, "step": 14218 }, { "epoch": 4.37, "learning_rate": 1.2490277032158053e-05, "loss": 0.3495, "step": 14219 }, { "epoch": 4.38, "learning_rate": 1.2489311874336511e-05, "loss": 0.2766, "step": 14220 }, { "epoch": 4.38, "learning_rate": 1.2488346691793812e-05, "loss": 0.2319, "step": 14221 }, { "epoch": 4.38, "learning_rate": 1.248738148453953e-05, "loss": 0.252, "step": 14222 }, { "epoch": 4.38, "learning_rate": 1.2486416252583265e-05, "loss": 0.2611, "step": 14223 }, { "epoch": 4.38, "learning_rate": 1.2485450995934594e-05, "loss": 0.2985, "step": 14224 }, { "epoch": 4.38, "learning_rate": 1.2484485714603103e-05, "loss": 0.2024, "step": 14225 }, { "epoch": 4.38, "learning_rate": 1.2483520408598377e-05, "loss": 0.3835, "step": 14226 }, { "epoch": 4.38, "learning_rate": 1.2482555077930006e-05, "loss": 0.3507, "step": 14227 }, { "epoch": 4.38, "learning_rate": 1.2481589722607574e-05, "loss": 0.2791, "step": 14228 }, { "epoch": 4.38, "learning_rate": 1.2480624342640673e-05, "loss": 0.2983, "step": 14229 }, { "epoch": 4.38, "learning_rate": 1.2479658938038884e-05, "loss": 0.2856, "step": 14230 }, { "epoch": 4.38, "learning_rate": 1.2478693508811794e-05, "loss": 0.265, "step": 14231 }, { "epoch": 4.38, "learning_rate": 1.2477728054968993e-05, "loss": 0.2652, "step": 14232 }, { "epoch": 4.38, "learning_rate": 1.2476762576520072e-05, "loss": 0.3315, "step": 14233 }, { "epoch": 4.38, "learning_rate": 1.2475797073474611e-05, "loss": 0.2681, "step": 14234 }, { "epoch": 4.38, "learning_rate": 1.247483154584221e-05, "loss": 0.438, "step": 14235 }, { "epoch": 4.38, "learning_rate": 1.2473865993632446e-05, "loss": 0.3882, "step": 14236 }, { "epoch": 4.38, "learning_rate": 1.2472900416854916e-05, "loss": 0.3139, "step": 14237 }, { "epoch": 4.38, "learning_rate": 1.2471934815519205e-05, "loss": 0.2253, "step": 14238 }, { "epoch": 4.38, "learning_rate": 1.2470969189634898e-05, "loss": 0.1679, "step": 14239 }, { "epoch": 4.38, "learning_rate": 1.2470003539211597e-05, "loss": 0.459, "step": 14240 }, { "epoch": 4.38, "learning_rate": 1.246903786425888e-05, "loss": 0.3884, "step": 14241 }, { "epoch": 4.38, "learning_rate": 1.2468072164786342e-05, "loss": 0.2855, "step": 14242 }, { "epoch": 4.38, "learning_rate": 1.2467106440803575e-05, "loss": 0.3565, "step": 14243 }, { "epoch": 4.38, "learning_rate": 1.2466140692320164e-05, "loss": 0.4823, "step": 14244 }, { "epoch": 4.38, "learning_rate": 1.2465174919345708e-05, "loss": 0.252, "step": 14245 }, { "epoch": 4.38, "learning_rate": 1.2464209121889787e-05, "loss": 0.2682, "step": 14246 }, { "epoch": 4.38, "learning_rate": 1.2463243299962002e-05, "loss": 0.2708, "step": 14247 }, { "epoch": 4.38, "learning_rate": 1.246227745357194e-05, "loss": 0.2812, "step": 14248 }, { "epoch": 4.38, "learning_rate": 1.2461311582729195e-05, "loss": 0.1778, "step": 14249 }, { "epoch": 4.38, "learning_rate": 1.2460345687443356e-05, "loss": 0.4336, "step": 14250 }, { "epoch": 4.38, "learning_rate": 1.2459379767724016e-05, "loss": 0.3331, "step": 14251 }, { "epoch": 4.39, "learning_rate": 1.2458413823580771e-05, "loss": 0.3301, "step": 14252 }, { "epoch": 4.39, "learning_rate": 1.2457447855023208e-05, "loss": 0.4155, "step": 14253 }, { "epoch": 4.39, "learning_rate": 1.2456481862060923e-05, "loss": 0.2458, "step": 14254 }, { "epoch": 4.39, "learning_rate": 1.2455515844703512e-05, "loss": 0.2804, "step": 14255 }, { "epoch": 4.39, "learning_rate": 1.2454549802960564e-05, "loss": 0.3578, "step": 14256 }, { "epoch": 4.39, "learning_rate": 1.2453583736841675e-05, "loss": 0.226, "step": 14257 }, { "epoch": 4.39, "learning_rate": 1.2452617646356436e-05, "loss": 0.3204, "step": 14258 }, { "epoch": 4.39, "learning_rate": 1.2451651531514446e-05, "loss": 0.2757, "step": 14259 }, { "epoch": 4.39, "learning_rate": 1.2450685392325295e-05, "loss": 0.2974, "step": 14260 }, { "epoch": 4.39, "learning_rate": 1.244971922879858e-05, "loss": 0.1117, "step": 14261 }, { "epoch": 4.39, "learning_rate": 1.2448753040943897e-05, "loss": 0.301, "step": 14262 }, { "epoch": 4.39, "learning_rate": 1.2447786828770835e-05, "loss": 0.4223, "step": 14263 }, { "epoch": 4.39, "learning_rate": 1.2446820592288997e-05, "loss": 0.3747, "step": 14264 }, { "epoch": 4.39, "learning_rate": 1.2445854331507977e-05, "loss": 0.2472, "step": 14265 }, { "epoch": 4.39, "learning_rate": 1.2444888046437367e-05, "loss": 0.1906, "step": 14266 }, { "epoch": 4.39, "learning_rate": 1.2443921737086767e-05, "loss": 0.2709, "step": 14267 }, { "epoch": 4.39, "learning_rate": 1.2442955403465768e-05, "loss": 0.274, "step": 14268 }, { "epoch": 4.39, "learning_rate": 1.2441989045583975e-05, "loss": 0.3876, "step": 14269 }, { "epoch": 4.39, "learning_rate": 1.2441022663450979e-05, "loss": 0.2154, "step": 14270 }, { "epoch": 4.39, "learning_rate": 1.2440056257076376e-05, "loss": 0.2792, "step": 14271 }, { "epoch": 4.39, "learning_rate": 1.2439089826469766e-05, "loss": 0.2414, "step": 14272 }, { "epoch": 4.39, "learning_rate": 1.2438123371640747e-05, "loss": 0.3162, "step": 14273 }, { "epoch": 4.39, "learning_rate": 1.2437156892598917e-05, "loss": 0.1831, "step": 14274 }, { "epoch": 4.39, "learning_rate": 1.2436190389353872e-05, "loss": 0.2869, "step": 14275 }, { "epoch": 4.39, "learning_rate": 1.2435223861915211e-05, "loss": 0.3834, "step": 14276 }, { "epoch": 4.39, "learning_rate": 1.2434257310292532e-05, "loss": 0.3118, "step": 14277 }, { "epoch": 4.39, "learning_rate": 1.2433290734495436e-05, "loss": 0.3214, "step": 14278 }, { "epoch": 4.39, "learning_rate": 1.243232413453352e-05, "loss": 0.3544, "step": 14279 }, { "epoch": 4.39, "learning_rate": 1.2431357510416384e-05, "loss": 0.3182, "step": 14280 }, { "epoch": 4.39, "learning_rate": 1.2430390862153625e-05, "loss": 0.2209, "step": 14281 }, { "epoch": 4.39, "learning_rate": 1.2429424189754846e-05, "loss": 0.3286, "step": 14282 }, { "epoch": 4.39, "learning_rate": 1.2428457493229647e-05, "loss": 0.2696, "step": 14283 }, { "epoch": 4.39, "learning_rate": 1.2427490772587623e-05, "loss": 0.4191, "step": 14284 }, { "epoch": 4.4, "learning_rate": 1.2426524027838384e-05, "loss": 0.3062, "step": 14285 }, { "epoch": 4.4, "learning_rate": 1.2425557258991522e-05, "loss": 0.3583, "step": 14286 }, { "epoch": 4.4, "learning_rate": 1.242459046605664e-05, "loss": 0.3615, "step": 14287 }, { "epoch": 4.4, "learning_rate": 1.2423623649043341e-05, "loss": 0.215, "step": 14288 }, { "epoch": 4.4, "learning_rate": 1.2422656807961227e-05, "loss": 0.4001, "step": 14289 }, { "epoch": 4.4, "learning_rate": 1.2421689942819898e-05, "loss": 0.3574, "step": 14290 }, { "epoch": 4.4, "learning_rate": 1.2420723053628953e-05, "loss": 0.5226, "step": 14291 }, { "epoch": 4.4, "learning_rate": 1.2419756140398e-05, "loss": 0.264, "step": 14292 }, { "epoch": 4.4, "learning_rate": 1.241878920313664e-05, "loss": 0.2214, "step": 14293 }, { "epoch": 4.4, "learning_rate": 1.2417822241854466e-05, "loss": 0.2548, "step": 14294 }, { "epoch": 4.4, "learning_rate": 1.2416855256561097e-05, "loss": 0.2205, "step": 14295 }, { "epoch": 4.4, "learning_rate": 1.2415888247266125e-05, "loss": 0.3966, "step": 14296 }, { "epoch": 4.4, "learning_rate": 1.2414921213979157e-05, "loss": 0.2466, "step": 14297 }, { "epoch": 4.4, "learning_rate": 1.2413954156709795e-05, "loss": 0.3435, "step": 14298 }, { "epoch": 4.4, "learning_rate": 1.2412987075467641e-05, "loss": 0.1483, "step": 14299 }, { "epoch": 4.4, "learning_rate": 1.2412019970262308e-05, "loss": 0.2249, "step": 14300 }, { "epoch": 4.4, "learning_rate": 1.2411052841103389e-05, "loss": 0.3069, "step": 14301 }, { "epoch": 4.4, "learning_rate": 1.2410085688000493e-05, "loss": 0.4554, "step": 14302 }, { "epoch": 4.4, "learning_rate": 1.2409118510963222e-05, "loss": 0.3072, "step": 14303 }, { "epoch": 4.4, "learning_rate": 1.2408151310001189e-05, "loss": 0.2566, "step": 14304 }, { "epoch": 4.4, "learning_rate": 1.2407184085123994e-05, "loss": 0.2561, "step": 14305 }, { "epoch": 4.4, "learning_rate": 1.240621683634124e-05, "loss": 0.3168, "step": 14306 }, { "epoch": 4.4, "learning_rate": 1.2405249563662539e-05, "loss": 0.436, "step": 14307 }, { "epoch": 4.4, "learning_rate": 1.2404282267097488e-05, "loss": 0.3051, "step": 14308 }, { "epoch": 4.4, "learning_rate": 1.2403314946655703e-05, "loss": 0.3712, "step": 14309 }, { "epoch": 4.4, "learning_rate": 1.2402347602346782e-05, "loss": 0.3083, "step": 14310 }, { "epoch": 4.4, "learning_rate": 1.2401380234180337e-05, "loss": 0.2309, "step": 14311 }, { "epoch": 4.4, "learning_rate": 1.2400412842165972e-05, "loss": 0.3027, "step": 14312 }, { "epoch": 4.4, "learning_rate": 1.2399445426313295e-05, "loss": 0.2347, "step": 14313 }, { "epoch": 4.4, "learning_rate": 1.2398477986631914e-05, "loss": 0.3513, "step": 14314 }, { "epoch": 4.4, "learning_rate": 1.2397510523131438e-05, "loss": 0.444, "step": 14315 }, { "epoch": 4.4, "learning_rate": 1.239654303582147e-05, "loss": 0.2448, "step": 14316 }, { "epoch": 4.41, "learning_rate": 1.2395575524711626e-05, "loss": 0.3933, "step": 14317 }, { "epoch": 4.41, "learning_rate": 1.2394607989811502e-05, "loss": 0.3632, "step": 14318 }, { "epoch": 4.41, "learning_rate": 1.2393640431130718e-05, "loss": 0.2665, "step": 14319 }, { "epoch": 4.41, "learning_rate": 1.2392672848678877e-05, "loss": 0.3609, "step": 14320 }, { "epoch": 4.41, "learning_rate": 1.2391705242465592e-05, "loss": 0.1932, "step": 14321 }, { "epoch": 4.41, "learning_rate": 1.2390737612500465e-05, "loss": 0.3656, "step": 14322 }, { "epoch": 4.41, "learning_rate": 1.2389769958793113e-05, "loss": 0.2862, "step": 14323 }, { "epoch": 4.41, "learning_rate": 1.2388802281353145e-05, "loss": 0.4155, "step": 14324 }, { "epoch": 4.41, "learning_rate": 1.2387834580190166e-05, "loss": 0.4669, "step": 14325 }, { "epoch": 4.41, "learning_rate": 1.238686685531379e-05, "loss": 0.3479, "step": 14326 }, { "epoch": 4.41, "learning_rate": 1.2385899106733626e-05, "loss": 0.265, "step": 14327 }, { "epoch": 4.41, "learning_rate": 1.2384931334459284e-05, "loss": 0.2703, "step": 14328 }, { "epoch": 4.41, "learning_rate": 1.2383963538500377e-05, "loss": 0.3237, "step": 14329 }, { "epoch": 4.41, "learning_rate": 1.2382995718866514e-05, "loss": 0.2526, "step": 14330 }, { "epoch": 4.41, "learning_rate": 1.2382027875567309e-05, "loss": 0.5851, "step": 14331 }, { "epoch": 4.41, "learning_rate": 1.2381060008612368e-05, "loss": 0.3925, "step": 14332 }, { "epoch": 4.41, "learning_rate": 1.238009211801131e-05, "loss": 0.1906, "step": 14333 }, { "epoch": 4.41, "learning_rate": 1.2379124203773742e-05, "loss": 0.3237, "step": 14334 }, { "epoch": 4.41, "learning_rate": 1.2378156265909281e-05, "loss": 0.3039, "step": 14335 }, { "epoch": 4.41, "learning_rate": 1.2377188304427537e-05, "loss": 0.2036, "step": 14336 }, { "epoch": 4.41, "learning_rate": 1.2376220319338118e-05, "loss": 0.2707, "step": 14337 }, { "epoch": 4.41, "learning_rate": 1.2375252310650643e-05, "loss": 0.3824, "step": 14338 }, { "epoch": 4.41, "learning_rate": 1.2374284278374722e-05, "loss": 0.2243, "step": 14339 }, { "epoch": 4.41, "learning_rate": 1.2373316222519971e-05, "loss": 0.1567, "step": 14340 }, { "epoch": 4.41, "learning_rate": 1.2372348143096004e-05, "loss": 0.2966, "step": 14341 }, { "epoch": 4.41, "learning_rate": 1.2371380040112433e-05, "loss": 0.5443, "step": 14342 }, { "epoch": 4.41, "learning_rate": 1.2370411913578872e-05, "loss": 0.335, "step": 14343 }, { "epoch": 4.41, "learning_rate": 1.2369443763504936e-05, "loss": 0.2096, "step": 14344 }, { "epoch": 4.41, "learning_rate": 1.236847558990024e-05, "loss": 0.2127, "step": 14345 }, { "epoch": 4.41, "learning_rate": 1.2367507392774398e-05, "loss": 0.285, "step": 14346 }, { "epoch": 4.41, "learning_rate": 1.2366539172137027e-05, "loss": 0.3836, "step": 14347 }, { "epoch": 4.41, "learning_rate": 1.236557092799774e-05, "loss": 0.3138, "step": 14348 }, { "epoch": 4.41, "learning_rate": 1.2364602660366151e-05, "loss": 0.3921, "step": 14349 }, { "epoch": 4.42, "learning_rate": 1.2363634369251881e-05, "loss": 0.2531, "step": 14350 }, { "epoch": 4.42, "learning_rate": 1.2362666054664545e-05, "loss": 0.3461, "step": 14351 }, { "epoch": 4.42, "learning_rate": 1.2361697716613757e-05, "loss": 0.3996, "step": 14352 }, { "epoch": 4.42, "learning_rate": 1.2360729355109132e-05, "loss": 0.356, "step": 14353 }, { "epoch": 4.42, "learning_rate": 1.2359760970160288e-05, "loss": 0.4011, "step": 14354 }, { "epoch": 4.42, "learning_rate": 1.2358792561776848e-05, "loss": 0.244, "step": 14355 }, { "epoch": 4.42, "learning_rate": 1.2357824129968418e-05, "loss": 0.3506, "step": 14356 }, { "epoch": 4.42, "learning_rate": 1.2356855674744625e-05, "loss": 0.3052, "step": 14357 }, { "epoch": 4.42, "learning_rate": 1.2355887196115079e-05, "loss": 0.2418, "step": 14358 }, { "epoch": 4.42, "learning_rate": 1.2354918694089406e-05, "loss": 0.2578, "step": 14359 }, { "epoch": 4.42, "learning_rate": 1.235395016867722e-05, "loss": 0.3236, "step": 14360 }, { "epoch": 4.42, "learning_rate": 1.2352981619888138e-05, "loss": 0.2629, "step": 14361 }, { "epoch": 4.42, "learning_rate": 1.2352013047731779e-05, "loss": 0.2723, "step": 14362 }, { "epoch": 4.42, "learning_rate": 1.2351044452217763e-05, "loss": 0.3383, "step": 14363 }, { "epoch": 4.42, "learning_rate": 1.235007583335571e-05, "loss": 0.2116, "step": 14364 }, { "epoch": 4.42, "learning_rate": 1.2349107191155239e-05, "loss": 0.4324, "step": 14365 }, { "epoch": 4.42, "learning_rate": 1.2348138525625965e-05, "loss": 0.3511, "step": 14366 }, { "epoch": 4.42, "learning_rate": 1.2347169836777516e-05, "loss": 0.4184, "step": 14367 }, { "epoch": 4.42, "learning_rate": 1.2346201124619502e-05, "loss": 0.209, "step": 14368 }, { "epoch": 4.42, "learning_rate": 1.2345232389161552e-05, "loss": 0.2844, "step": 14369 }, { "epoch": 4.42, "learning_rate": 1.234426363041328e-05, "loss": 0.4021, "step": 14370 }, { "epoch": 4.42, "learning_rate": 1.2343294848384315e-05, "loss": 0.2497, "step": 14371 }, { "epoch": 4.42, "learning_rate": 1.2342326043084268e-05, "loss": 0.4265, "step": 14372 }, { "epoch": 4.42, "learning_rate": 1.2341357214522767e-05, "loss": 0.319, "step": 14373 }, { "epoch": 4.42, "learning_rate": 1.234038836270943e-05, "loss": 0.2968, "step": 14374 }, { "epoch": 4.42, "learning_rate": 1.2339419487653876e-05, "loss": 0.3222, "step": 14375 }, { "epoch": 4.42, "learning_rate": 1.2338450589365739e-05, "loss": 0.3539, "step": 14376 }, { "epoch": 4.42, "learning_rate": 1.2337481667854625e-05, "loss": 0.4219, "step": 14377 }, { "epoch": 4.42, "learning_rate": 1.2336512723130168e-05, "loss": 0.3339, "step": 14378 }, { "epoch": 4.42, "learning_rate": 1.2335543755201984e-05, "loss": 0.2534, "step": 14379 }, { "epoch": 4.42, "learning_rate": 1.2334574764079696e-05, "loss": 0.2053, "step": 14380 }, { "epoch": 4.42, "learning_rate": 1.2333605749772935e-05, "loss": 0.3198, "step": 14381 }, { "epoch": 4.43, "learning_rate": 1.2332636712291313e-05, "loss": 0.4178, "step": 14382 }, { "epoch": 4.43, "learning_rate": 1.233166765164446e-05, "loss": 0.2444, "step": 14383 }, { "epoch": 4.43, "learning_rate": 1.2330698567841997e-05, "loss": 0.4159, "step": 14384 }, { "epoch": 4.43, "learning_rate": 1.2329729460893552e-05, "loss": 0.3331, "step": 14385 }, { "epoch": 4.43, "learning_rate": 1.2328760330808747e-05, "loss": 0.3075, "step": 14386 }, { "epoch": 4.43, "learning_rate": 1.2327791177597201e-05, "loss": 0.2325, "step": 14387 }, { "epoch": 4.43, "learning_rate": 1.2326822001268548e-05, "loss": 0.1899, "step": 14388 }, { "epoch": 4.43, "learning_rate": 1.2325852801832403e-05, "loss": 0.2821, "step": 14389 }, { "epoch": 4.43, "learning_rate": 1.23248835792984e-05, "loss": 0.2569, "step": 14390 }, { "epoch": 4.43, "learning_rate": 1.2323914333676162e-05, "loss": 0.2942, "step": 14391 }, { "epoch": 4.43, "learning_rate": 1.232294506497531e-05, "loss": 0.2796, "step": 14392 }, { "epoch": 4.43, "learning_rate": 1.2321975773205472e-05, "loss": 0.3015, "step": 14393 }, { "epoch": 4.43, "learning_rate": 1.2321006458376275e-05, "loss": 0.2585, "step": 14394 }, { "epoch": 4.43, "learning_rate": 1.2320037120497345e-05, "loss": 0.4554, "step": 14395 }, { "epoch": 4.43, "learning_rate": 1.231906775957831e-05, "loss": 0.3179, "step": 14396 }, { "epoch": 4.43, "learning_rate": 1.2318098375628793e-05, "loss": 0.3857, "step": 14397 }, { "epoch": 4.43, "learning_rate": 1.2317128968658424e-05, "loss": 0.3632, "step": 14398 }, { "epoch": 4.43, "learning_rate": 1.2316159538676827e-05, "loss": 0.1835, "step": 14399 }, { "epoch": 4.43, "learning_rate": 1.2315190085693632e-05, "loss": 0.2568, "step": 14400 }, { "epoch": 4.43, "learning_rate": 1.2314220609718466e-05, "loss": 0.1993, "step": 14401 }, { "epoch": 4.43, "learning_rate": 1.2313251110760958e-05, "loss": 0.2322, "step": 14402 }, { "epoch": 4.43, "learning_rate": 1.231228158883073e-05, "loss": 0.3408, "step": 14403 }, { "epoch": 4.43, "learning_rate": 1.2311312043937418e-05, "loss": 0.4601, "step": 14404 }, { "epoch": 4.43, "learning_rate": 1.2310342476090646e-05, "loss": 0.3024, "step": 14405 }, { "epoch": 4.43, "learning_rate": 1.2309372885300045e-05, "loss": 0.1921, "step": 14406 }, { "epoch": 4.43, "learning_rate": 1.2308403271575245e-05, "loss": 0.315, "step": 14407 }, { "epoch": 4.43, "learning_rate": 1.2307433634925869e-05, "loss": 0.3671, "step": 14408 }, { "epoch": 4.43, "learning_rate": 1.2306463975361553e-05, "loss": 0.2558, "step": 14409 }, { "epoch": 4.43, "learning_rate": 1.2305494292891923e-05, "loss": 0.4033, "step": 14410 }, { "epoch": 4.43, "learning_rate": 1.2304524587526609e-05, "loss": 0.248, "step": 14411 }, { "epoch": 4.43, "learning_rate": 1.2303554859275244e-05, "loss": 0.1977, "step": 14412 }, { "epoch": 4.43, "learning_rate": 1.2302585108147453e-05, "loss": 0.3074, "step": 14413 }, { "epoch": 4.43, "learning_rate": 1.2301615334152875e-05, "loss": 0.2554, "step": 14414 }, { "epoch": 4.44, "learning_rate": 1.230064553730113e-05, "loss": 0.4166, "step": 14415 }, { "epoch": 4.44, "learning_rate": 1.2299675717601858e-05, "loss": 0.3681, "step": 14416 }, { "epoch": 4.44, "learning_rate": 1.2298705875064688e-05, "loss": 0.3827, "step": 14417 }, { "epoch": 4.44, "learning_rate": 1.2297736009699243e-05, "loss": 0.3668, "step": 14418 }, { "epoch": 4.44, "learning_rate": 1.2296766121515166e-05, "loss": 0.2468, "step": 14419 }, { "epoch": 4.44, "learning_rate": 1.2295796210522086e-05, "loss": 0.2041, "step": 14420 }, { "epoch": 4.44, "learning_rate": 1.2294826276729634e-05, "loss": 0.4926, "step": 14421 }, { "epoch": 4.44, "learning_rate": 1.229385632014744e-05, "loss": 0.284, "step": 14422 }, { "epoch": 4.44, "learning_rate": 1.229288634078514e-05, "loss": 0.4332, "step": 14423 }, { "epoch": 4.44, "learning_rate": 1.2291916338652365e-05, "loss": 0.1926, "step": 14424 }, { "epoch": 4.44, "learning_rate": 1.2290946313758748e-05, "loss": 0.2068, "step": 14425 }, { "epoch": 4.44, "learning_rate": 1.2289976266113925e-05, "loss": 0.3338, "step": 14426 }, { "epoch": 4.44, "learning_rate": 1.2289006195727523e-05, "loss": 0.3811, "step": 14427 }, { "epoch": 4.44, "learning_rate": 1.2288036102609185e-05, "loss": 0.221, "step": 14428 }, { "epoch": 4.44, "learning_rate": 1.2287065986768536e-05, "loss": 0.3463, "step": 14429 }, { "epoch": 4.44, "learning_rate": 1.2286095848215214e-05, "loss": 0.3246, "step": 14430 }, { "epoch": 4.44, "learning_rate": 1.2285125686958854e-05, "loss": 0.3579, "step": 14431 }, { "epoch": 4.44, "learning_rate": 1.228415550300909e-05, "loss": 0.3776, "step": 14432 }, { "epoch": 4.44, "learning_rate": 1.2283185296375558e-05, "loss": 0.3142, "step": 14433 }, { "epoch": 4.44, "learning_rate": 1.2282215067067891e-05, "loss": 0.3946, "step": 14434 }, { "epoch": 4.44, "learning_rate": 1.2281244815095723e-05, "loss": 0.2875, "step": 14435 }, { "epoch": 4.44, "learning_rate": 1.2280274540468693e-05, "loss": 0.2593, "step": 14436 }, { "epoch": 4.44, "learning_rate": 1.2279304243196438e-05, "loss": 0.3079, "step": 14437 }, { "epoch": 4.44, "learning_rate": 1.2278333923288589e-05, "loss": 0.3325, "step": 14438 }, { "epoch": 4.44, "learning_rate": 1.227736358075478e-05, "loss": 0.2543, "step": 14439 }, { "epoch": 4.44, "learning_rate": 1.2276393215604657e-05, "loss": 0.2402, "step": 14440 }, { "epoch": 4.44, "learning_rate": 1.2275422827847849e-05, "loss": 0.5318, "step": 14441 }, { "epoch": 4.44, "learning_rate": 1.2274452417493998e-05, "loss": 0.4205, "step": 14442 }, { "epoch": 4.44, "learning_rate": 1.2273481984552737e-05, "loss": 0.3916, "step": 14443 }, { "epoch": 4.44, "learning_rate": 1.2272511529033704e-05, "loss": 0.3166, "step": 14444 }, { "epoch": 4.44, "learning_rate": 1.2271541050946536e-05, "loss": 0.176, "step": 14445 }, { "epoch": 4.44, "learning_rate": 1.2270570550300875e-05, "loss": 0.2591, "step": 14446 }, { "epoch": 4.45, "learning_rate": 1.2269600027106354e-05, "loss": 0.2745, "step": 14447 }, { "epoch": 4.45, "learning_rate": 1.2268629481372614e-05, "loss": 0.2753, "step": 14448 }, { "epoch": 4.45, "learning_rate": 1.2267658913109289e-05, "loss": 0.3108, "step": 14449 }, { "epoch": 4.45, "learning_rate": 1.2266688322326024e-05, "loss": 0.1719, "step": 14450 }, { "epoch": 4.45, "learning_rate": 1.2265717709032452e-05, "loss": 0.2958, "step": 14451 }, { "epoch": 4.45, "learning_rate": 1.2264747073238221e-05, "loss": 0.324, "step": 14452 }, { "epoch": 4.45, "learning_rate": 1.2263776414952959e-05, "loss": 0.4244, "step": 14453 }, { "epoch": 4.45, "learning_rate": 1.2262805734186315e-05, "loss": 0.3294, "step": 14454 }, { "epoch": 4.45, "learning_rate": 1.226183503094792e-05, "loss": 0.3602, "step": 14455 }, { "epoch": 4.45, "learning_rate": 1.226086430524742e-05, "loss": 0.2707, "step": 14456 }, { "epoch": 4.45, "learning_rate": 1.2259893557094457e-05, "loss": 0.1734, "step": 14457 }, { "epoch": 4.45, "learning_rate": 1.2258922786498664e-05, "loss": 0.2195, "step": 14458 }, { "epoch": 4.45, "learning_rate": 1.2257951993469688e-05, "loss": 0.2687, "step": 14459 }, { "epoch": 4.45, "learning_rate": 1.2256981178017167e-05, "loss": 0.416, "step": 14460 }, { "epoch": 4.45, "learning_rate": 1.2256010340150742e-05, "loss": 0.3894, "step": 14461 }, { "epoch": 4.45, "learning_rate": 1.2255039479880057e-05, "loss": 0.2673, "step": 14462 }, { "epoch": 4.45, "learning_rate": 1.225406859721475e-05, "loss": 0.3203, "step": 14463 }, { "epoch": 4.45, "learning_rate": 1.2253097692164467e-05, "loss": 0.3449, "step": 14464 }, { "epoch": 4.45, "learning_rate": 1.2252126764738845e-05, "loss": 0.3902, "step": 14465 }, { "epoch": 4.45, "learning_rate": 1.2251155814947528e-05, "loss": 0.2036, "step": 14466 }, { "epoch": 4.45, "learning_rate": 1.2250184842800164e-05, "loss": 0.2225, "step": 14467 }, { "epoch": 4.45, "learning_rate": 1.2249213848306385e-05, "loss": 0.401, "step": 14468 }, { "epoch": 4.45, "learning_rate": 1.2248242831475841e-05, "loss": 0.3195, "step": 14469 }, { "epoch": 4.45, "learning_rate": 1.2247271792318172e-05, "loss": 0.2582, "step": 14470 }, { "epoch": 4.45, "learning_rate": 1.2246300730843025e-05, "loss": 0.3065, "step": 14471 }, { "epoch": 4.45, "learning_rate": 1.2245329647060044e-05, "loss": 0.4348, "step": 14472 }, { "epoch": 4.45, "learning_rate": 1.2244358540978869e-05, "loss": 0.3121, "step": 14473 }, { "epoch": 4.45, "learning_rate": 1.2243387412609144e-05, "loss": 0.4176, "step": 14474 }, { "epoch": 4.45, "learning_rate": 1.2242416261960513e-05, "loss": 0.4261, "step": 14475 }, { "epoch": 4.45, "learning_rate": 1.2241445089042623e-05, "loss": 0.2208, "step": 14476 }, { "epoch": 4.45, "learning_rate": 1.224047389386512e-05, "loss": 0.4211, "step": 14477 }, { "epoch": 4.45, "learning_rate": 1.2239502676437643e-05, "loss": 0.2164, "step": 14478 }, { "epoch": 4.45, "learning_rate": 1.2238531436769842e-05, "loss": 0.2799, "step": 14479 }, { "epoch": 4.46, "learning_rate": 1.223756017487136e-05, "loss": 0.3386, "step": 14480 }, { "epoch": 4.46, "learning_rate": 1.2236588890751844e-05, "loss": 0.2776, "step": 14481 }, { "epoch": 4.46, "learning_rate": 1.2235617584420938e-05, "loss": 0.3741, "step": 14482 }, { "epoch": 4.46, "learning_rate": 1.223464625588829e-05, "loss": 0.4507, "step": 14483 }, { "epoch": 4.46, "learning_rate": 1.2233674905163546e-05, "loss": 0.3287, "step": 14484 }, { "epoch": 4.46, "learning_rate": 1.223270353225635e-05, "loss": 0.3337, "step": 14485 }, { "epoch": 4.46, "learning_rate": 1.2231732137176348e-05, "loss": 0.4732, "step": 14486 }, { "epoch": 4.46, "learning_rate": 1.2230760719933194e-05, "loss": 0.1406, "step": 14487 }, { "epoch": 4.46, "learning_rate": 1.2229789280536529e-05, "loss": 0.2713, "step": 14488 }, { "epoch": 4.46, "learning_rate": 1.2228817818995998e-05, "loss": 0.2563, "step": 14489 }, { "epoch": 4.46, "learning_rate": 1.2227846335321254e-05, "loss": 0.5183, "step": 14490 }, { "epoch": 4.46, "learning_rate": 1.2226874829521944e-05, "loss": 0.3997, "step": 14491 }, { "epoch": 4.46, "learning_rate": 1.2225903301607713e-05, "loss": 0.3352, "step": 14492 }, { "epoch": 4.46, "learning_rate": 1.2224931751588211e-05, "loss": 0.3132, "step": 14493 }, { "epoch": 4.46, "learning_rate": 1.2223960179473084e-05, "loss": 0.2116, "step": 14494 }, { "epoch": 4.46, "learning_rate": 1.2222988585271986e-05, "loss": 0.3104, "step": 14495 }, { "epoch": 4.46, "learning_rate": 1.2222016968994563e-05, "loss": 0.2073, "step": 14496 }, { "epoch": 4.46, "learning_rate": 1.2221045330650463e-05, "loss": 0.2858, "step": 14497 }, { "epoch": 4.46, "learning_rate": 1.2220073670249336e-05, "loss": 0.3711, "step": 14498 }, { "epoch": 4.46, "learning_rate": 1.2219101987800828e-05, "loss": 0.317, "step": 14499 }, { "epoch": 4.46, "learning_rate": 1.2218130283314595e-05, "loss": 0.2218, "step": 14500 }, { "epoch": 4.46, "learning_rate": 1.2217158556800282e-05, "loss": 0.3306, "step": 14501 }, { "epoch": 4.46, "learning_rate": 1.2216186808267544e-05, "loss": 0.2232, "step": 14502 }, { "epoch": 4.46, "learning_rate": 1.2215215037726026e-05, "loss": 0.1959, "step": 14503 }, { "epoch": 4.46, "learning_rate": 1.2214243245185382e-05, "loss": 0.2331, "step": 14504 }, { "epoch": 4.46, "learning_rate": 1.2213271430655265e-05, "loss": 0.2503, "step": 14505 }, { "epoch": 4.46, "learning_rate": 1.2212299594145319e-05, "loss": 0.2479, "step": 14506 }, { "epoch": 4.46, "learning_rate": 1.2211327735665201e-05, "loss": 0.2073, "step": 14507 }, { "epoch": 4.46, "learning_rate": 1.2210355855224558e-05, "loss": 0.3009, "step": 14508 }, { "epoch": 4.46, "learning_rate": 1.2209383952833046e-05, "loss": 0.3731, "step": 14509 }, { "epoch": 4.46, "learning_rate": 1.2208412028500317e-05, "loss": 0.3502, "step": 14510 }, { "epoch": 4.46, "learning_rate": 1.2207440082236017e-05, "loss": 0.2722, "step": 14511 }, { "epoch": 4.47, "learning_rate": 1.2206468114049808e-05, "loss": 0.3243, "step": 14512 }, { "epoch": 4.47, "learning_rate": 1.220549612395133e-05, "loss": 0.3203, "step": 14513 }, { "epoch": 4.47, "learning_rate": 1.220452411195025e-05, "loss": 0.4061, "step": 14514 }, { "epoch": 4.47, "learning_rate": 1.2203552078056209e-05, "loss": 0.4221, "step": 14515 }, { "epoch": 4.47, "learning_rate": 1.2202580022278865e-05, "loss": 0.4239, "step": 14516 }, { "epoch": 4.47, "learning_rate": 1.2201607944627876e-05, "loss": 0.2439, "step": 14517 }, { "epoch": 4.47, "learning_rate": 1.2200635845112887e-05, "loss": 0.441, "step": 14518 }, { "epoch": 4.47, "learning_rate": 1.2199663723743558e-05, "loss": 0.3309, "step": 14519 }, { "epoch": 4.47, "learning_rate": 1.2198691580529535e-05, "loss": 0.4507, "step": 14520 }, { "epoch": 4.47, "learning_rate": 1.2197719415480486e-05, "loss": 0.3602, "step": 14521 }, { "epoch": 4.47, "learning_rate": 1.2196747228606056e-05, "loss": 0.4375, "step": 14522 }, { "epoch": 4.47, "learning_rate": 1.2195775019915899e-05, "loss": 0.465, "step": 14523 }, { "epoch": 4.47, "learning_rate": 1.2194802789419675e-05, "loss": 0.2278, "step": 14524 }, { "epoch": 4.47, "learning_rate": 1.2193830537127033e-05, "loss": 0.2594, "step": 14525 }, { "epoch": 4.47, "learning_rate": 1.2192858263047635e-05, "loss": 0.4702, "step": 14526 }, { "epoch": 4.47, "learning_rate": 1.2191885967191134e-05, "loss": 0.3661, "step": 14527 }, { "epoch": 4.47, "learning_rate": 1.2190913649567185e-05, "loss": 0.2952, "step": 14528 }, { "epoch": 4.47, "learning_rate": 1.2189941310185443e-05, "loss": 0.3446, "step": 14529 }, { "epoch": 4.47, "learning_rate": 1.2188968949055562e-05, "loss": 0.284, "step": 14530 }, { "epoch": 4.47, "learning_rate": 1.2187996566187205e-05, "loss": 0.325, "step": 14531 }, { "epoch": 4.47, "learning_rate": 1.2187024161590026e-05, "loss": 0.293, "step": 14532 }, { "epoch": 4.47, "learning_rate": 1.2186051735273683e-05, "loss": 0.3909, "step": 14533 }, { "epoch": 4.47, "learning_rate": 1.2185079287247831e-05, "loss": 0.2382, "step": 14534 }, { "epoch": 4.47, "learning_rate": 1.2184106817522125e-05, "loss": 0.41, "step": 14535 }, { "epoch": 4.47, "learning_rate": 1.2183134326106228e-05, "loss": 0.3077, "step": 14536 }, { "epoch": 4.47, "learning_rate": 1.2182161813009792e-05, "loss": 0.2527, "step": 14537 }, { "epoch": 4.47, "learning_rate": 1.2181189278242482e-05, "loss": 0.3029, "step": 14538 }, { "epoch": 4.47, "learning_rate": 1.2180216721813952e-05, "loss": 0.3745, "step": 14539 }, { "epoch": 4.47, "learning_rate": 1.2179244143733857e-05, "loss": 0.2459, "step": 14540 }, { "epoch": 4.47, "learning_rate": 1.2178271544011864e-05, "loss": 0.1922, "step": 14541 }, { "epoch": 4.47, "learning_rate": 1.2177298922657622e-05, "loss": 0.3347, "step": 14542 }, { "epoch": 4.47, "learning_rate": 1.21763262796808e-05, "loss": 0.4323, "step": 14543 }, { "epoch": 4.47, "learning_rate": 1.2175353615091047e-05, "loss": 0.2664, "step": 14544 }, { "epoch": 4.48, "learning_rate": 1.2174380928898031e-05, "loss": 0.3445, "step": 14545 }, { "epoch": 4.48, "learning_rate": 1.217340822111141e-05, "loss": 0.255, "step": 14546 }, { "epoch": 4.48, "learning_rate": 1.2172435491740837e-05, "loss": 0.3154, "step": 14547 }, { "epoch": 4.48, "learning_rate": 1.2171462740795982e-05, "loss": 0.2728, "step": 14548 }, { "epoch": 4.48, "learning_rate": 1.21704899682865e-05, "loss": 0.295, "step": 14549 }, { "epoch": 4.48, "learning_rate": 1.2169517174222052e-05, "loss": 0.3291, "step": 14550 }, { "epoch": 4.48, "learning_rate": 1.2168544358612295e-05, "loss": 0.2753, "step": 14551 }, { "epoch": 4.48, "learning_rate": 1.2167571521466899e-05, "loss": 0.3132, "step": 14552 }, { "epoch": 4.48, "learning_rate": 1.2166598662795518e-05, "loss": 0.2939, "step": 14553 }, { "epoch": 4.48, "learning_rate": 1.2165625782607817e-05, "loss": 0.3608, "step": 14554 }, { "epoch": 4.48, "learning_rate": 1.2164652880913456e-05, "loss": 0.445, "step": 14555 }, { "epoch": 4.48, "learning_rate": 1.2163679957722094e-05, "loss": 0.3655, "step": 14556 }, { "epoch": 4.48, "learning_rate": 1.2162707013043398e-05, "loss": 0.2327, "step": 14557 }, { "epoch": 4.48, "learning_rate": 1.2161734046887028e-05, "loss": 0.4384, "step": 14558 }, { "epoch": 4.48, "learning_rate": 1.2160761059262649e-05, "loss": 0.2471, "step": 14559 }, { "epoch": 4.48, "learning_rate": 1.215978805017992e-05, "loss": 0.5314, "step": 14560 }, { "epoch": 4.48, "learning_rate": 1.21588150196485e-05, "loss": 0.2566, "step": 14561 }, { "epoch": 4.48, "learning_rate": 1.2157841967678064e-05, "loss": 0.2559, "step": 14562 }, { "epoch": 4.48, "learning_rate": 1.2156868894278264e-05, "loss": 0.3127, "step": 14563 }, { "epoch": 4.48, "learning_rate": 1.215589579945877e-05, "loss": 0.2773, "step": 14564 }, { "epoch": 4.48, "learning_rate": 1.2154922683229245e-05, "loss": 0.4328, "step": 14565 }, { "epoch": 4.48, "learning_rate": 1.2153949545599349e-05, "loss": 0.2094, "step": 14566 }, { "epoch": 4.48, "learning_rate": 1.215297638657875e-05, "loss": 0.1578, "step": 14567 }, { "epoch": 4.48, "learning_rate": 1.2152003206177113e-05, "loss": 0.3052, "step": 14568 }, { "epoch": 4.48, "learning_rate": 1.2151030004404098e-05, "loss": 0.2141, "step": 14569 }, { "epoch": 4.48, "learning_rate": 1.2150056781269372e-05, "loss": 0.2783, "step": 14570 }, { "epoch": 4.48, "learning_rate": 1.2149083536782604e-05, "loss": 0.3956, "step": 14571 }, { "epoch": 4.48, "learning_rate": 1.2148110270953455e-05, "loss": 0.2858, "step": 14572 }, { "epoch": 4.48, "learning_rate": 1.2147136983791591e-05, "loss": 0.4324, "step": 14573 }, { "epoch": 4.48, "learning_rate": 1.2146163675306676e-05, "loss": 0.4194, "step": 14574 }, { "epoch": 4.48, "learning_rate": 1.214519034550838e-05, "loss": 0.2187, "step": 14575 }, { "epoch": 4.48, "learning_rate": 1.2144216994406367e-05, "loss": 0.3733, "step": 14576 }, { "epoch": 4.49, "learning_rate": 1.2143243622010302e-05, "loss": 0.2553, "step": 14577 }, { "epoch": 4.49, "learning_rate": 1.2142270228329854e-05, "loss": 0.4211, "step": 14578 }, { "epoch": 4.49, "learning_rate": 1.2141296813374686e-05, "loss": 0.3516, "step": 14579 }, { "epoch": 4.49, "learning_rate": 1.2140323377154467e-05, "loss": 0.2512, "step": 14580 }, { "epoch": 4.49, "learning_rate": 1.2139349919678865e-05, "loss": 0.2425, "step": 14581 }, { "epoch": 4.49, "learning_rate": 1.2138376440957544e-05, "loss": 0.33, "step": 14582 }, { "epoch": 4.49, "learning_rate": 1.2137402941000183e-05, "loss": 0.1924, "step": 14583 }, { "epoch": 4.49, "learning_rate": 1.2136429419816433e-05, "loss": 0.2962, "step": 14584 }, { "epoch": 4.49, "learning_rate": 1.2135455877415971e-05, "loss": 0.342, "step": 14585 }, { "epoch": 4.49, "learning_rate": 1.2134482313808466e-05, "loss": 0.4397, "step": 14586 }, { "epoch": 4.49, "learning_rate": 1.213350872900358e-05, "loss": 0.3384, "step": 14587 }, { "epoch": 4.49, "learning_rate": 1.2132535123010994e-05, "loss": 0.1715, "step": 14588 }, { "epoch": 4.49, "learning_rate": 1.2131561495840362e-05, "loss": 0.3556, "step": 14589 }, { "epoch": 4.49, "learning_rate": 1.2130587847501363e-05, "loss": 0.2932, "step": 14590 }, { "epoch": 4.49, "learning_rate": 1.2129614178003662e-05, "loss": 0.1727, "step": 14591 }, { "epoch": 4.49, "learning_rate": 1.2128640487356929e-05, "loss": 0.408, "step": 14592 }, { "epoch": 4.49, "learning_rate": 1.2127666775570837e-05, "loss": 0.3708, "step": 14593 }, { "epoch": 4.49, "learning_rate": 1.2126693042655049e-05, "loss": 0.2318, "step": 14594 }, { "epoch": 4.49, "learning_rate": 1.212571928861924e-05, "loss": 0.1608, "step": 14595 }, { "epoch": 4.49, "learning_rate": 1.2124745513473081e-05, "loss": 0.2105, "step": 14596 }, { "epoch": 4.49, "learning_rate": 1.2123771717226239e-05, "loss": 0.3265, "step": 14597 }, { "epoch": 4.49, "learning_rate": 1.212279789988839e-05, "loss": 0.2119, "step": 14598 }, { "epoch": 4.49, "learning_rate": 1.2121824061469197e-05, "loss": 0.4094, "step": 14599 }, { "epoch": 4.49, "learning_rate": 1.2120850201978338e-05, "loss": 0.2162, "step": 14600 }, { "epoch": 4.49, "learning_rate": 1.211987632142548e-05, "loss": 0.2997, "step": 14601 }, { "epoch": 4.49, "learning_rate": 1.2118902419820297e-05, "loss": 0.37, "step": 14602 }, { "epoch": 4.49, "learning_rate": 1.2117928497172464e-05, "loss": 0.3263, "step": 14603 }, { "epoch": 4.49, "learning_rate": 1.2116954553491645e-05, "loss": 0.4579, "step": 14604 }, { "epoch": 4.49, "learning_rate": 1.2115980588787517e-05, "loss": 0.33, "step": 14605 }, { "epoch": 4.49, "learning_rate": 1.211500660306975e-05, "loss": 0.2422, "step": 14606 }, { "epoch": 4.49, "learning_rate": 1.2114032596348023e-05, "loss": 0.4069, "step": 14607 }, { "epoch": 4.49, "learning_rate": 1.2113058568631999e-05, "loss": 0.3231, "step": 14608 }, { "epoch": 4.49, "learning_rate": 1.211208451993136e-05, "loss": 0.2357, "step": 14609 }, { "epoch": 4.5, "learning_rate": 1.2111110450255771e-05, "loss": 0.3575, "step": 14610 }, { "epoch": 4.5, "learning_rate": 1.2110136359614912e-05, "loss": 0.4217, "step": 14611 }, { "epoch": 4.5, "learning_rate": 1.2109162248018454e-05, "loss": 0.2617, "step": 14612 }, { "epoch": 4.5, "learning_rate": 1.2108188115476072e-05, "loss": 0.2605, "step": 14613 }, { "epoch": 4.5, "learning_rate": 1.2107213961997436e-05, "loss": 0.3629, "step": 14614 }, { "epoch": 4.5, "learning_rate": 1.2106239787592227e-05, "loss": 0.4016, "step": 14615 }, { "epoch": 4.5, "learning_rate": 1.2105265592270113e-05, "loss": 0.3898, "step": 14616 }, { "epoch": 4.5, "learning_rate": 1.2104291376040774e-05, "loss": 0.2214, "step": 14617 }, { "epoch": 4.5, "learning_rate": 1.2103317138913879e-05, "loss": 0.3184, "step": 14618 }, { "epoch": 4.5, "learning_rate": 1.210234288089911e-05, "loss": 0.2233, "step": 14619 }, { "epoch": 4.5, "learning_rate": 1.2101368602006134e-05, "loss": 0.1538, "step": 14620 }, { "epoch": 4.5, "learning_rate": 1.2100394302244634e-05, "loss": 0.4063, "step": 14621 }, { "epoch": 4.5, "learning_rate": 1.2099419981624284e-05, "loss": 0.5042, "step": 14622 }, { "epoch": 4.5, "learning_rate": 1.2098445640154756e-05, "loss": 0.2693, "step": 14623 }, { "epoch": 4.5, "learning_rate": 1.209747127784573e-05, "loss": 0.2631, "step": 14624 }, { "epoch": 4.5, "learning_rate": 1.209649689470688e-05, "loss": 0.261, "step": 14625 }, { "epoch": 4.5, "learning_rate": 1.2095522490747886e-05, "loss": 0.3267, "step": 14626 }, { "epoch": 4.5, "learning_rate": 1.2094548065978422e-05, "loss": 0.4151, "step": 14627 }, { "epoch": 4.5, "learning_rate": 1.2093573620408162e-05, "loss": 0.2892, "step": 14628 }, { "epoch": 4.5, "learning_rate": 1.2092599154046791e-05, "loss": 0.4123, "step": 14629 }, { "epoch": 4.5, "learning_rate": 1.2091624666903978e-05, "loss": 0.319, "step": 14630 }, { "epoch": 4.5, "learning_rate": 1.2090650158989408e-05, "loss": 0.4485, "step": 14631 }, { "epoch": 4.5, "learning_rate": 1.2089675630312755e-05, "loss": 0.2113, "step": 14632 }, { "epoch": 4.5, "learning_rate": 1.208870108088369e-05, "loss": 0.3997, "step": 14633 }, { "epoch": 4.5, "learning_rate": 1.2087726510711908e-05, "loss": 0.2934, "step": 14634 }, { "epoch": 4.5, "learning_rate": 1.2086751919807072e-05, "loss": 0.2716, "step": 14635 }, { "epoch": 4.5, "learning_rate": 1.2085777308178868e-05, "loss": 0.2947, "step": 14636 }, { "epoch": 4.5, "learning_rate": 1.208480267583697e-05, "loss": 0.3216, "step": 14637 }, { "epoch": 4.5, "learning_rate": 1.2083828022791062e-05, "loss": 0.3885, "step": 14638 }, { "epoch": 4.5, "learning_rate": 1.2082853349050823e-05, "loss": 0.2412, "step": 14639 }, { "epoch": 4.5, "learning_rate": 1.2081878654625929e-05, "loss": 0.3368, "step": 14640 }, { "epoch": 4.5, "learning_rate": 1.2080903939526063e-05, "loss": 0.2397, "step": 14641 }, { "epoch": 4.51, "learning_rate": 1.2079929203760901e-05, "loss": 0.3372, "step": 14642 }, { "epoch": 4.51, "learning_rate": 1.2078954447340126e-05, "loss": 0.388, "step": 14643 }, { "epoch": 4.51, "learning_rate": 1.2077979670273417e-05, "loss": 0.3026, "step": 14644 }, { "epoch": 4.51, "learning_rate": 1.2077004872570454e-05, "loss": 0.3987, "step": 14645 }, { "epoch": 4.51, "learning_rate": 1.2076030054240922e-05, "loss": 0.5686, "step": 14646 }, { "epoch": 4.51, "learning_rate": 1.2075055215294492e-05, "loss": 0.2973, "step": 14647 }, { "epoch": 4.51, "learning_rate": 1.2074080355740857e-05, "loss": 0.3538, "step": 14648 }, { "epoch": 4.51, "learning_rate": 1.2073105475589688e-05, "loss": 0.3386, "step": 14649 }, { "epoch": 4.51, "learning_rate": 1.2072130574850673e-05, "loss": 0.4226, "step": 14650 }, { "epoch": 4.51, "learning_rate": 1.207115565353349e-05, "loss": 0.3921, "step": 14651 }, { "epoch": 4.51, "learning_rate": 1.2070180711647828e-05, "loss": 0.2164, "step": 14652 }, { "epoch": 4.51, "learning_rate": 1.2069205749203359e-05, "loss": 0.3689, "step": 14653 }, { "epoch": 4.51, "learning_rate": 1.2068230766209773e-05, "loss": 0.4367, "step": 14654 }, { "epoch": 4.51, "learning_rate": 1.2067255762676745e-05, "loss": 0.3007, "step": 14655 }, { "epoch": 4.51, "learning_rate": 1.2066280738613965e-05, "loss": 0.2599, "step": 14656 }, { "epoch": 4.51, "learning_rate": 1.206530569403111e-05, "loss": 0.3676, "step": 14657 }, { "epoch": 4.51, "learning_rate": 1.206433062893787e-05, "loss": 0.2826, "step": 14658 }, { "epoch": 4.51, "learning_rate": 1.2063355543343925e-05, "loss": 0.2645, "step": 14659 }, { "epoch": 4.51, "learning_rate": 1.2062380437258953e-05, "loss": 0.3763, "step": 14660 }, { "epoch": 4.51, "learning_rate": 1.2061405310692644e-05, "loss": 0.3638, "step": 14661 }, { "epoch": 4.51, "learning_rate": 1.2060430163654681e-05, "loss": 0.3291, "step": 14662 }, { "epoch": 4.51, "learning_rate": 1.2059454996154749e-05, "loss": 0.4203, "step": 14663 }, { "epoch": 4.51, "learning_rate": 1.205847980820253e-05, "loss": 0.3339, "step": 14664 }, { "epoch": 4.51, "learning_rate": 1.205750459980771e-05, "loss": 0.3661, "step": 14665 }, { "epoch": 4.51, "learning_rate": 1.205652937097997e-05, "loss": 0.3105, "step": 14666 }, { "epoch": 4.51, "learning_rate": 1.2055554121729e-05, "loss": 0.2718, "step": 14667 }, { "epoch": 4.51, "learning_rate": 1.2054578852064482e-05, "loss": 0.1309, "step": 14668 }, { "epoch": 4.51, "learning_rate": 1.2053603561996105e-05, "loss": 0.4365, "step": 14669 }, { "epoch": 4.51, "learning_rate": 1.2052628251533549e-05, "loss": 0.2997, "step": 14670 }, { "epoch": 4.51, "learning_rate": 1.2051652920686505e-05, "loss": 0.414, "step": 14671 }, { "epoch": 4.51, "learning_rate": 1.2050677569464657e-05, "loss": 0.3324, "step": 14672 }, { "epoch": 4.51, "learning_rate": 1.2049702197877689e-05, "loss": 0.1814, "step": 14673 }, { "epoch": 4.51, "learning_rate": 1.2048726805935291e-05, "loss": 0.2673, "step": 14674 }, { "epoch": 4.52, "learning_rate": 1.2047751393647145e-05, "loss": 0.1597, "step": 14675 }, { "epoch": 4.52, "learning_rate": 1.2046775961022943e-05, "loss": 0.3633, "step": 14676 }, { "epoch": 4.52, "learning_rate": 1.2045800508072367e-05, "loss": 0.2662, "step": 14677 }, { "epoch": 4.52, "learning_rate": 1.2044825034805108e-05, "loss": 0.5092, "step": 14678 }, { "epoch": 4.52, "learning_rate": 1.2043849541230854e-05, "loss": 0.2775, "step": 14679 }, { "epoch": 4.52, "learning_rate": 1.2042874027359286e-05, "loss": 0.2319, "step": 14680 }, { "epoch": 4.52, "learning_rate": 1.2041898493200099e-05, "loss": 0.2592, "step": 14681 }, { "epoch": 4.52, "learning_rate": 1.2040922938762977e-05, "loss": 0.2902, "step": 14682 }, { "epoch": 4.52, "learning_rate": 1.2039947364057607e-05, "loss": 0.2676, "step": 14683 }, { "epoch": 4.52, "learning_rate": 1.2038971769093685e-05, "loss": 0.3046, "step": 14684 }, { "epoch": 4.52, "learning_rate": 1.203799615388089e-05, "loss": 0.3138, "step": 14685 }, { "epoch": 4.52, "learning_rate": 1.2037020518428917e-05, "loss": 0.1945, "step": 14686 }, { "epoch": 4.52, "learning_rate": 1.203604486274745e-05, "loss": 0.4375, "step": 14687 }, { "epoch": 4.52, "learning_rate": 1.2035069186846183e-05, "loss": 0.3776, "step": 14688 }, { "epoch": 4.52, "learning_rate": 1.2034093490734804e-05, "loss": 0.4081, "step": 14689 }, { "epoch": 4.52, "learning_rate": 1.2033117774423003e-05, "loss": 0.3242, "step": 14690 }, { "epoch": 4.52, "learning_rate": 1.2032142037920467e-05, "loss": 0.3439, "step": 14691 }, { "epoch": 4.52, "learning_rate": 1.2031166281236886e-05, "loss": 0.2245, "step": 14692 }, { "epoch": 4.52, "learning_rate": 1.2030190504381953e-05, "loss": 0.4294, "step": 14693 }, { "epoch": 4.52, "learning_rate": 1.202921470736536e-05, "loss": 0.1602, "step": 14694 }, { "epoch": 4.52, "learning_rate": 1.2028238890196793e-05, "loss": 0.3123, "step": 14695 }, { "epoch": 4.52, "learning_rate": 1.2027263052885942e-05, "loss": 0.3065, "step": 14696 }, { "epoch": 4.52, "learning_rate": 1.2026287195442503e-05, "loss": 0.3744, "step": 14697 }, { "epoch": 4.52, "learning_rate": 1.2025311317876164e-05, "loss": 0.2848, "step": 14698 }, { "epoch": 4.52, "learning_rate": 1.2024335420196618e-05, "loss": 0.5215, "step": 14699 }, { "epoch": 4.52, "learning_rate": 1.2023359502413556e-05, "loss": 0.2866, "step": 14700 }, { "epoch": 4.52, "learning_rate": 1.2022383564536665e-05, "loss": 0.2951, "step": 14701 }, { "epoch": 4.52, "learning_rate": 1.2021407606575646e-05, "loss": 0.2393, "step": 14702 }, { "epoch": 4.52, "learning_rate": 1.2020431628540186e-05, "loss": 0.3532, "step": 14703 }, { "epoch": 4.52, "learning_rate": 1.2019455630439978e-05, "loss": 0.4959, "step": 14704 }, { "epoch": 4.52, "learning_rate": 1.2018479612284713e-05, "loss": 0.4217, "step": 14705 }, { "epoch": 4.52, "learning_rate": 1.2017503574084084e-05, "loss": 0.3321, "step": 14706 }, { "epoch": 4.53, "learning_rate": 1.2016527515847786e-05, "loss": 0.3212, "step": 14707 }, { "epoch": 4.53, "learning_rate": 1.2015551437585513e-05, "loss": 0.3274, "step": 14708 }, { "epoch": 4.53, "learning_rate": 1.2014575339306957e-05, "loss": 0.4048, "step": 14709 }, { "epoch": 4.53, "learning_rate": 1.201359922102181e-05, "loss": 0.3462, "step": 14710 }, { "epoch": 4.53, "learning_rate": 1.2012623082739764e-05, "loss": 0.4218, "step": 14711 }, { "epoch": 4.53, "learning_rate": 1.201164692447052e-05, "loss": 0.3951, "step": 14712 }, { "epoch": 4.53, "learning_rate": 1.2010670746223766e-05, "loss": 0.2748, "step": 14713 }, { "epoch": 4.53, "learning_rate": 1.20096945480092e-05, "loss": 0.1817, "step": 14714 }, { "epoch": 4.53, "learning_rate": 1.2008718329836515e-05, "loss": 0.3605, "step": 14715 }, { "epoch": 4.53, "learning_rate": 1.2007742091715405e-05, "loss": 0.3448, "step": 14716 }, { "epoch": 4.53, "learning_rate": 1.2006765833655565e-05, "loss": 0.2535, "step": 14717 }, { "epoch": 4.53, "learning_rate": 1.200578955566669e-05, "loss": 0.2708, "step": 14718 }, { "epoch": 4.53, "learning_rate": 1.2004813257758478e-05, "loss": 0.3323, "step": 14719 }, { "epoch": 4.53, "learning_rate": 1.200383693994062e-05, "loss": 0.4655, "step": 14720 }, { "epoch": 4.53, "learning_rate": 1.2002860602222816e-05, "loss": 0.251, "step": 14721 }, { "epoch": 4.53, "learning_rate": 1.2001884244614762e-05, "loss": 0.4051, "step": 14722 }, { "epoch": 4.53, "learning_rate": 1.200090786712615e-05, "loss": 0.3566, "step": 14723 }, { "epoch": 4.53, "learning_rate": 1.1999931469766683e-05, "loss": 0.3205, "step": 14724 }, { "epoch": 4.53, "learning_rate": 1.1998955052546048e-05, "loss": 0.3025, "step": 14725 }, { "epoch": 4.53, "learning_rate": 1.199797861547395e-05, "loss": 0.3205, "step": 14726 }, { "epoch": 4.53, "learning_rate": 1.1997002158560083e-05, "loss": 0.1599, "step": 14727 }, { "epoch": 4.53, "learning_rate": 1.1996025681814141e-05, "loss": 0.3457, "step": 14728 }, { "epoch": 4.53, "learning_rate": 1.1995049185245826e-05, "loss": 0.206, "step": 14729 }, { "epoch": 4.53, "learning_rate": 1.1994072668864836e-05, "loss": 0.2212, "step": 14730 }, { "epoch": 4.53, "learning_rate": 1.1993096132680867e-05, "loss": 0.2415, "step": 14731 }, { "epoch": 4.53, "learning_rate": 1.1992119576703614e-05, "loss": 0.2916, "step": 14732 }, { "epoch": 4.53, "learning_rate": 1.1991143000942777e-05, "loss": 0.3633, "step": 14733 }, { "epoch": 4.53, "learning_rate": 1.1990166405408057e-05, "loss": 0.4685, "step": 14734 }, { "epoch": 4.53, "learning_rate": 1.198918979010915e-05, "loss": 0.3563, "step": 14735 }, { "epoch": 4.53, "learning_rate": 1.1988213155055754e-05, "loss": 0.5323, "step": 14736 }, { "epoch": 4.53, "learning_rate": 1.1987236500257568e-05, "loss": 0.2949, "step": 14737 }, { "epoch": 4.53, "learning_rate": 1.1986259825724296e-05, "loss": 0.1979, "step": 14738 }, { "epoch": 4.53, "learning_rate": 1.1985283131465631e-05, "loss": 0.1885, "step": 14739 }, { "epoch": 4.54, "learning_rate": 1.1984306417491274e-05, "loss": 0.4061, "step": 14740 }, { "epoch": 4.54, "learning_rate": 1.1983329683810927e-05, "loss": 0.2957, "step": 14741 }, { "epoch": 4.54, "learning_rate": 1.1982352930434287e-05, "loss": 0.2867, "step": 14742 }, { "epoch": 4.54, "learning_rate": 1.1981376157371055e-05, "loss": 0.3214, "step": 14743 }, { "epoch": 4.54, "learning_rate": 1.1980399364630935e-05, "loss": 0.3287, "step": 14744 }, { "epoch": 4.54, "learning_rate": 1.197942255222362e-05, "loss": 0.2795, "step": 14745 }, { "epoch": 4.54, "learning_rate": 1.1978445720158818e-05, "loss": 0.2008, "step": 14746 }, { "epoch": 4.54, "learning_rate": 1.1977468868446224e-05, "loss": 0.3835, "step": 14747 }, { "epoch": 4.54, "learning_rate": 1.1976491997095544e-05, "loss": 0.2819, "step": 14748 }, { "epoch": 4.54, "learning_rate": 1.1975515106116472e-05, "loss": 0.1511, "step": 14749 }, { "epoch": 4.54, "learning_rate": 1.1974538195518724e-05, "loss": 0.2952, "step": 14750 }, { "epoch": 4.54, "learning_rate": 1.1973561265311986e-05, "loss": 0.3407, "step": 14751 }, { "epoch": 4.54, "learning_rate": 1.1972584315505964e-05, "loss": 0.2774, "step": 14752 }, { "epoch": 4.54, "learning_rate": 1.1971607346110363e-05, "loss": 0.3704, "step": 14753 }, { "epoch": 4.54, "learning_rate": 1.1970630357134883e-05, "loss": 0.2934, "step": 14754 }, { "epoch": 4.54, "learning_rate": 1.1969653348589231e-05, "loss": 0.3128, "step": 14755 }, { "epoch": 4.54, "learning_rate": 1.1968676320483103e-05, "loss": 0.3038, "step": 14756 }, { "epoch": 4.54, "learning_rate": 1.1967699272826205e-05, "loss": 0.3034, "step": 14757 }, { "epoch": 4.54, "learning_rate": 1.1966722205628241e-05, "loss": 0.3305, "step": 14758 }, { "epoch": 4.54, "learning_rate": 1.196574511889891e-05, "loss": 0.2095, "step": 14759 }, { "epoch": 4.54, "learning_rate": 1.1964768012647925e-05, "loss": 0.2921, "step": 14760 }, { "epoch": 4.54, "learning_rate": 1.1963790886884977e-05, "loss": 0.3214, "step": 14761 }, { "epoch": 4.54, "learning_rate": 1.1962813741619777e-05, "loss": 0.2205, "step": 14762 }, { "epoch": 4.54, "learning_rate": 1.1961836576862027e-05, "loss": 0.2433, "step": 14763 }, { "epoch": 4.54, "learning_rate": 1.1960859392621432e-05, "loss": 0.3104, "step": 14764 }, { "epoch": 4.54, "learning_rate": 1.1959882188907698e-05, "loss": 0.3048, "step": 14765 }, { "epoch": 4.54, "learning_rate": 1.1958904965730524e-05, "loss": 0.2205, "step": 14766 }, { "epoch": 4.54, "learning_rate": 1.195792772309962e-05, "loss": 0.2421, "step": 14767 }, { "epoch": 4.54, "learning_rate": 1.1956950461024687e-05, "loss": 0.7019, "step": 14768 }, { "epoch": 4.54, "learning_rate": 1.1955973179515436e-05, "loss": 0.2446, "step": 14769 }, { "epoch": 4.54, "learning_rate": 1.195499587858157e-05, "loss": 0.5, "step": 14770 }, { "epoch": 4.54, "learning_rate": 1.195401855823279e-05, "loss": 0.2588, "step": 14771 }, { "epoch": 4.55, "learning_rate": 1.1953041218478807e-05, "loss": 0.3587, "step": 14772 }, { "epoch": 4.55, "learning_rate": 1.1952063859329322e-05, "loss": 0.3133, "step": 14773 }, { "epoch": 4.55, "learning_rate": 1.1951086480794042e-05, "loss": 0.312, "step": 14774 }, { "epoch": 4.55, "learning_rate": 1.1950109082882681e-05, "loss": 0.3078, "step": 14775 }, { "epoch": 4.55, "learning_rate": 1.1949131665604936e-05, "loss": 0.2972, "step": 14776 }, { "epoch": 4.55, "learning_rate": 1.1948154228970518e-05, "loss": 0.3529, "step": 14777 }, { "epoch": 4.55, "learning_rate": 1.1947176772989133e-05, "loss": 0.325, "step": 14778 }, { "epoch": 4.55, "learning_rate": 1.1946199297670486e-05, "loss": 0.3694, "step": 14779 }, { "epoch": 4.55, "learning_rate": 1.194522180302429e-05, "loss": 0.4809, "step": 14780 }, { "epoch": 4.55, "learning_rate": 1.1944244289060248e-05, "loss": 0.2683, "step": 14781 }, { "epoch": 4.55, "learning_rate": 1.1943266755788068e-05, "loss": 0.4495, "step": 14782 }, { "epoch": 4.55, "learning_rate": 1.1942289203217454e-05, "loss": 0.258, "step": 14783 }, { "epoch": 4.55, "learning_rate": 1.1941311631358124e-05, "loss": 0.2917, "step": 14784 }, { "epoch": 4.55, "learning_rate": 1.1940334040219777e-05, "loss": 0.3133, "step": 14785 }, { "epoch": 4.55, "learning_rate": 1.1939356429812125e-05, "loss": 0.2743, "step": 14786 }, { "epoch": 4.55, "learning_rate": 1.1938378800144877e-05, "loss": 0.3866, "step": 14787 }, { "epoch": 4.55, "learning_rate": 1.193740115122774e-05, "loss": 0.3635, "step": 14788 }, { "epoch": 4.55, "learning_rate": 1.1936423483070424e-05, "loss": 0.2543, "step": 14789 }, { "epoch": 4.55, "learning_rate": 1.1935445795682641e-05, "loss": 0.2897, "step": 14790 }, { "epoch": 4.55, "learning_rate": 1.1934468089074096e-05, "loss": 0.3615, "step": 14791 }, { "epoch": 4.55, "learning_rate": 1.1933490363254499e-05, "loss": 0.169, "step": 14792 }, { "epoch": 4.55, "learning_rate": 1.193251261823356e-05, "loss": 0.3882, "step": 14793 }, { "epoch": 4.55, "learning_rate": 1.1931534854020992e-05, "loss": 0.4016, "step": 14794 }, { "epoch": 4.55, "learning_rate": 1.1930557070626503e-05, "loss": 0.3411, "step": 14795 }, { "epoch": 4.55, "learning_rate": 1.1929579268059803e-05, "loss": 0.3699, "step": 14796 }, { "epoch": 4.55, "learning_rate": 1.1928601446330599e-05, "loss": 0.2606, "step": 14797 }, { "epoch": 4.55, "learning_rate": 1.1927623605448609e-05, "loss": 0.2129, "step": 14798 }, { "epoch": 4.55, "learning_rate": 1.1926645745423536e-05, "loss": 0.154, "step": 14799 }, { "epoch": 4.55, "learning_rate": 1.1925667866265101e-05, "loss": 0.2857, "step": 14800 }, { "epoch": 4.55, "learning_rate": 1.1924689967983006e-05, "loss": 0.2162, "step": 14801 }, { "epoch": 4.55, "learning_rate": 1.1923712050586964e-05, "loss": 0.2668, "step": 14802 }, { "epoch": 4.55, "learning_rate": 1.192273411408669e-05, "loss": 0.2896, "step": 14803 }, { "epoch": 4.55, "learning_rate": 1.1921756158491894e-05, "loss": 0.3581, "step": 14804 }, { "epoch": 4.56, "learning_rate": 1.192077818381229e-05, "loss": 0.3379, "step": 14805 }, { "epoch": 4.56, "learning_rate": 1.1919800190057584e-05, "loss": 0.3779, "step": 14806 }, { "epoch": 4.56, "learning_rate": 1.1918822177237497e-05, "loss": 0.3917, "step": 14807 }, { "epoch": 4.56, "learning_rate": 1.1917844145361736e-05, "loss": 0.3017, "step": 14808 }, { "epoch": 4.56, "learning_rate": 1.1916866094440014e-05, "loss": 0.2538, "step": 14809 }, { "epoch": 4.56, "learning_rate": 1.1915888024482047e-05, "loss": 0.2824, "step": 14810 }, { "epoch": 4.56, "learning_rate": 1.1914909935497546e-05, "loss": 0.4162, "step": 14811 }, { "epoch": 4.56, "learning_rate": 1.1913931827496223e-05, "loss": 0.3653, "step": 14812 }, { "epoch": 4.56, "learning_rate": 1.1912953700487795e-05, "loss": 0.3586, "step": 14813 }, { "epoch": 4.56, "learning_rate": 1.191197555448197e-05, "loss": 0.2668, "step": 14814 }, { "epoch": 4.56, "learning_rate": 1.1910997389488472e-05, "loss": 0.3903, "step": 14815 }, { "epoch": 4.56, "learning_rate": 1.1910019205517003e-05, "loss": 0.3908, "step": 14816 }, { "epoch": 4.56, "learning_rate": 1.1909041002577287e-05, "loss": 0.247, "step": 14817 }, { "epoch": 4.56, "learning_rate": 1.1908062780679029e-05, "loss": 0.3284, "step": 14818 }, { "epoch": 4.56, "learning_rate": 1.1907084539831952e-05, "loss": 0.4584, "step": 14819 }, { "epoch": 4.56, "learning_rate": 1.1906106280045768e-05, "loss": 0.3582, "step": 14820 }, { "epoch": 4.56, "learning_rate": 1.1905128001330195e-05, "loss": 0.2838, "step": 14821 }, { "epoch": 4.56, "learning_rate": 1.1904149703694943e-05, "loss": 0.2855, "step": 14822 }, { "epoch": 4.56, "learning_rate": 1.1903171387149726e-05, "loss": 0.3987, "step": 14823 }, { "epoch": 4.56, "learning_rate": 1.1902193051704267e-05, "loss": 0.3263, "step": 14824 }, { "epoch": 4.56, "learning_rate": 1.1901214697368277e-05, "loss": 0.2058, "step": 14825 }, { "epoch": 4.56, "learning_rate": 1.1900236324151472e-05, "loss": 0.4084, "step": 14826 }, { "epoch": 4.56, "learning_rate": 1.189925793206357e-05, "loss": 0.3324, "step": 14827 }, { "epoch": 4.56, "learning_rate": 1.1898279521114285e-05, "loss": 0.269, "step": 14828 }, { "epoch": 4.56, "learning_rate": 1.1897301091313336e-05, "loss": 0.2139, "step": 14829 }, { "epoch": 4.56, "learning_rate": 1.189632264267044e-05, "loss": 0.3623, "step": 14830 }, { "epoch": 4.56, "learning_rate": 1.189534417519531e-05, "loss": 0.2509, "step": 14831 }, { "epoch": 4.56, "learning_rate": 1.1894365688897666e-05, "loss": 0.3007, "step": 14832 }, { "epoch": 4.56, "learning_rate": 1.1893387183787227e-05, "loss": 0.2915, "step": 14833 }, { "epoch": 4.56, "learning_rate": 1.1892408659873704e-05, "loss": 0.2554, "step": 14834 }, { "epoch": 4.56, "learning_rate": 1.189143011716682e-05, "loss": 0.3495, "step": 14835 }, { "epoch": 4.56, "learning_rate": 1.1890451555676297e-05, "loss": 0.208, "step": 14836 }, { "epoch": 4.57, "learning_rate": 1.1889472975411843e-05, "loss": 0.2004, "step": 14837 }, { "epoch": 4.57, "learning_rate": 1.1888494376383181e-05, "loss": 0.4401, "step": 14838 }, { "epoch": 4.57, "learning_rate": 1.188751575860003e-05, "loss": 0.3295, "step": 14839 }, { "epoch": 4.57, "learning_rate": 1.1886537122072106e-05, "loss": 0.3376, "step": 14840 }, { "epoch": 4.57, "learning_rate": 1.1885558466809134e-05, "loss": 0.407, "step": 14841 }, { "epoch": 4.57, "learning_rate": 1.1884579792820827e-05, "loss": 0.368, "step": 14842 }, { "epoch": 4.57, "learning_rate": 1.1883601100116906e-05, "loss": 0.3821, "step": 14843 }, { "epoch": 4.57, "learning_rate": 1.188262238870709e-05, "loss": 0.3115, "step": 14844 }, { "epoch": 4.57, "learning_rate": 1.1881643658601094e-05, "loss": 0.4674, "step": 14845 }, { "epoch": 4.57, "learning_rate": 1.188066490980865e-05, "loss": 0.2314, "step": 14846 }, { "epoch": 4.57, "learning_rate": 1.1879686142339466e-05, "loss": 0.3486, "step": 14847 }, { "epoch": 4.57, "learning_rate": 1.1878707356203264e-05, "loss": 0.3959, "step": 14848 }, { "epoch": 4.57, "learning_rate": 1.1877728551409771e-05, "loss": 0.4069, "step": 14849 }, { "epoch": 4.57, "learning_rate": 1.1876749727968699e-05, "loss": 0.3487, "step": 14850 }, { "epoch": 4.57, "learning_rate": 1.187577088588978e-05, "loss": 0.3043, "step": 14851 }, { "epoch": 4.57, "learning_rate": 1.1874792025182718e-05, "loss": 0.4695, "step": 14852 }, { "epoch": 4.57, "learning_rate": 1.187381314585725e-05, "loss": 0.3421, "step": 14853 }, { "epoch": 4.57, "learning_rate": 1.1872834247923083e-05, "loss": 0.1947, "step": 14854 }, { "epoch": 4.57, "learning_rate": 1.1871855331389952e-05, "loss": 0.2839, "step": 14855 }, { "epoch": 4.57, "learning_rate": 1.1870876396267572e-05, "loss": 0.4115, "step": 14856 }, { "epoch": 4.57, "learning_rate": 1.1869897442565665e-05, "loss": 0.4077, "step": 14857 }, { "epoch": 4.57, "learning_rate": 1.1868918470293953e-05, "loss": 0.3857, "step": 14858 }, { "epoch": 4.57, "learning_rate": 1.1867939479462156e-05, "loss": 0.3842, "step": 14859 }, { "epoch": 4.57, "learning_rate": 1.186696047008e-05, "loss": 0.1603, "step": 14860 }, { "epoch": 4.57, "learning_rate": 1.1865981442157207e-05, "loss": 0.3258, "step": 14861 }, { "epoch": 4.57, "learning_rate": 1.1865002395703496e-05, "loss": 0.2625, "step": 14862 }, { "epoch": 4.57, "learning_rate": 1.1864023330728593e-05, "loss": 0.3158, "step": 14863 }, { "epoch": 4.57, "learning_rate": 1.186304424724222e-05, "loss": 0.3728, "step": 14864 }, { "epoch": 4.57, "learning_rate": 1.1862065145254103e-05, "loss": 0.2765, "step": 14865 }, { "epoch": 4.57, "learning_rate": 1.1861086024773963e-05, "loss": 0.3555, "step": 14866 }, { "epoch": 4.57, "learning_rate": 1.1860106885811522e-05, "loss": 0.2469, "step": 14867 }, { "epoch": 4.57, "learning_rate": 1.1859127728376503e-05, "loss": 0.3109, "step": 14868 }, { "epoch": 4.57, "learning_rate": 1.1858148552478635e-05, "loss": 0.279, "step": 14869 }, { "epoch": 4.58, "learning_rate": 1.185716935812764e-05, "loss": 0.2507, "step": 14870 }, { "epoch": 4.58, "learning_rate": 1.185619014533324e-05, "loss": 0.2568, "step": 14871 }, { "epoch": 4.58, "learning_rate": 1.1855210914105162e-05, "loss": 0.3422, "step": 14872 }, { "epoch": 4.58, "learning_rate": 1.185423166445313e-05, "loss": 0.3011, "step": 14873 }, { "epoch": 4.58, "learning_rate": 1.1853252396386868e-05, "loss": 0.3415, "step": 14874 }, { "epoch": 4.58, "learning_rate": 1.1852273109916102e-05, "loss": 0.3141, "step": 14875 }, { "epoch": 4.58, "learning_rate": 1.1851293805050556e-05, "loss": 0.348, "step": 14876 }, { "epoch": 4.58, "learning_rate": 1.1850314481799958e-05, "loss": 0.358, "step": 14877 }, { "epoch": 4.58, "learning_rate": 1.184933514017403e-05, "loss": 0.2465, "step": 14878 }, { "epoch": 4.58, "learning_rate": 1.1848355780182502e-05, "loss": 0.3725, "step": 14879 }, { "epoch": 4.58, "learning_rate": 1.1847376401835095e-05, "loss": 0.236, "step": 14880 }, { "epoch": 4.58, "learning_rate": 1.1846397005141542e-05, "loss": 0.2746, "step": 14881 }, { "epoch": 4.58, "learning_rate": 1.184541759011156e-05, "loss": 0.2251, "step": 14882 }, { "epoch": 4.58, "learning_rate": 1.1844438156754882e-05, "loss": 0.3017, "step": 14883 }, { "epoch": 4.58, "learning_rate": 1.1843458705081234e-05, "loss": 0.2639, "step": 14884 }, { "epoch": 4.58, "learning_rate": 1.1842479235100342e-05, "loss": 0.2497, "step": 14885 }, { "epoch": 4.58, "learning_rate": 1.1841499746821933e-05, "loss": 0.4303, "step": 14886 }, { "epoch": 4.58, "learning_rate": 1.1840520240255735e-05, "loss": 0.3401, "step": 14887 }, { "epoch": 4.58, "learning_rate": 1.1839540715411475e-05, "loss": 0.3903, "step": 14888 }, { "epoch": 4.58, "learning_rate": 1.1838561172298879e-05, "loss": 0.2978, "step": 14889 }, { "epoch": 4.58, "learning_rate": 1.1837581610927675e-05, "loss": 0.2707, "step": 14890 }, { "epoch": 4.58, "learning_rate": 1.1836602031307596e-05, "loss": 0.3209, "step": 14891 }, { "epoch": 4.58, "learning_rate": 1.1835622433448361e-05, "loss": 0.2721, "step": 14892 }, { "epoch": 4.58, "learning_rate": 1.183464281735971e-05, "loss": 0.4301, "step": 14893 }, { "epoch": 4.58, "learning_rate": 1.183366318305136e-05, "loss": 0.467, "step": 14894 }, { "epoch": 4.58, "learning_rate": 1.1832683530533045e-05, "loss": 0.2045, "step": 14895 }, { "epoch": 4.58, "learning_rate": 1.1831703859814497e-05, "loss": 0.2386, "step": 14896 }, { "epoch": 4.58, "learning_rate": 1.1830724170905437e-05, "loss": 0.3421, "step": 14897 }, { "epoch": 4.58, "learning_rate": 1.1829744463815601e-05, "loss": 0.3076, "step": 14898 }, { "epoch": 4.58, "learning_rate": 1.1828764738554714e-05, "loss": 0.282, "step": 14899 }, { "epoch": 4.58, "learning_rate": 1.182778499513251e-05, "loss": 0.3592, "step": 14900 }, { "epoch": 4.58, "learning_rate": 1.1826805233558719e-05, "loss": 0.3793, "step": 14901 }, { "epoch": 4.59, "learning_rate": 1.1825825453843063e-05, "loss": 0.342, "step": 14902 }, { "epoch": 4.59, "learning_rate": 1.1824845655995281e-05, "loss": 0.3343, "step": 14903 }, { "epoch": 4.59, "learning_rate": 1.1823865840025098e-05, "loss": 0.391, "step": 14904 }, { "epoch": 4.59, "learning_rate": 1.1822886005942244e-05, "loss": 0.3393, "step": 14905 }, { "epoch": 4.59, "learning_rate": 1.1821906153756457e-05, "loss": 0.2624, "step": 14906 }, { "epoch": 4.59, "learning_rate": 1.182092628347746e-05, "loss": 0.2138, "step": 14907 }, { "epoch": 4.59, "learning_rate": 1.1819946395114989e-05, "loss": 0.241, "step": 14908 }, { "epoch": 4.59, "learning_rate": 1.181896648867877e-05, "loss": 0.2776, "step": 14909 }, { "epoch": 4.59, "learning_rate": 1.1817986564178539e-05, "loss": 0.2954, "step": 14910 }, { "epoch": 4.59, "learning_rate": 1.1817006621624026e-05, "loss": 0.4023, "step": 14911 }, { "epoch": 4.59, "learning_rate": 1.1816026661024962e-05, "loss": 0.2909, "step": 14912 }, { "epoch": 4.59, "learning_rate": 1.1815046682391081e-05, "loss": 0.2747, "step": 14913 }, { "epoch": 4.59, "learning_rate": 1.181406668573211e-05, "loss": 0.3357, "step": 14914 }, { "epoch": 4.59, "learning_rate": 1.1813086671057789e-05, "loss": 0.2622, "step": 14915 }, { "epoch": 4.59, "learning_rate": 1.1812106638377844e-05, "loss": 0.2125, "step": 14916 }, { "epoch": 4.59, "learning_rate": 1.1811126587702014e-05, "loss": 0.4066, "step": 14917 }, { "epoch": 4.59, "learning_rate": 1.1810146519040023e-05, "loss": 0.2367, "step": 14918 }, { "epoch": 4.59, "learning_rate": 1.1809166432401611e-05, "loss": 0.2531, "step": 14919 }, { "epoch": 4.59, "learning_rate": 1.1808186327796507e-05, "loss": 0.3822, "step": 14920 }, { "epoch": 4.59, "learning_rate": 1.180720620523445e-05, "loss": 0.3344, "step": 14921 }, { "epoch": 4.59, "learning_rate": 1.1806226064725165e-05, "loss": 0.2708, "step": 14922 }, { "epoch": 4.59, "learning_rate": 1.1805245906278394e-05, "loss": 0.1985, "step": 14923 }, { "epoch": 4.59, "learning_rate": 1.1804265729903865e-05, "loss": 0.3757, "step": 14924 }, { "epoch": 4.59, "learning_rate": 1.1803285535611318e-05, "loss": 0.178, "step": 14925 }, { "epoch": 4.59, "learning_rate": 1.1802305323410479e-05, "loss": 0.2742, "step": 14926 }, { "epoch": 4.59, "learning_rate": 1.1801325093311091e-05, "loss": 0.6034, "step": 14927 }, { "epoch": 4.59, "learning_rate": 1.1800344845322882e-05, "loss": 0.3228, "step": 14928 }, { "epoch": 4.59, "learning_rate": 1.1799364579455592e-05, "loss": 0.1353, "step": 14929 }, { "epoch": 4.59, "learning_rate": 1.1798384295718951e-05, "loss": 0.2523, "step": 14930 }, { "epoch": 4.59, "learning_rate": 1.1797403994122698e-05, "loss": 0.3102, "step": 14931 }, { "epoch": 4.59, "learning_rate": 1.1796423674676566e-05, "loss": 0.4309, "step": 14932 }, { "epoch": 4.59, "learning_rate": 1.1795443337390294e-05, "loss": 0.2633, "step": 14933 }, { "epoch": 4.59, "learning_rate": 1.179446298227361e-05, "loss": 0.2285, "step": 14934 }, { "epoch": 4.6, "learning_rate": 1.1793482609336258e-05, "loss": 0.1908, "step": 14935 }, { "epoch": 4.6, "learning_rate": 1.1792502218587968e-05, "loss": 0.471, "step": 14936 }, { "epoch": 4.6, "learning_rate": 1.1791521810038481e-05, "loss": 0.2576, "step": 14937 }, { "epoch": 4.6, "learning_rate": 1.1790541383697533e-05, "loss": 0.4233, "step": 14938 }, { "epoch": 4.6, "learning_rate": 1.1789560939574858e-05, "loss": 0.2675, "step": 14939 }, { "epoch": 4.6, "learning_rate": 1.1788580477680191e-05, "loss": 0.3738, "step": 14940 }, { "epoch": 4.6, "learning_rate": 1.1787599998023273e-05, "loss": 0.1864, "step": 14941 }, { "epoch": 4.6, "learning_rate": 1.1786619500613841e-05, "loss": 0.2303, "step": 14942 }, { "epoch": 4.6, "learning_rate": 1.178563898546163e-05, "loss": 0.419, "step": 14943 }, { "epoch": 4.6, "learning_rate": 1.178465845257638e-05, "loss": 0.4526, "step": 14944 }, { "epoch": 4.6, "learning_rate": 1.178367790196782e-05, "loss": 0.4927, "step": 14945 }, { "epoch": 4.6, "learning_rate": 1.17826973336457e-05, "loss": 0.3646, "step": 14946 }, { "epoch": 4.6, "learning_rate": 1.1781716747619753e-05, "loss": 0.2645, "step": 14947 }, { "epoch": 4.6, "learning_rate": 1.1780736143899716e-05, "loss": 0.3533, "step": 14948 }, { "epoch": 4.6, "learning_rate": 1.1779755522495327e-05, "loss": 0.3196, "step": 14949 }, { "epoch": 4.6, "learning_rate": 1.1778774883416325e-05, "loss": 0.4162, "step": 14950 }, { "epoch": 4.6, "learning_rate": 1.1777794226672447e-05, "loss": 0.2806, "step": 14951 }, { "epoch": 4.6, "learning_rate": 1.1776813552273438e-05, "loss": 0.2469, "step": 14952 }, { "epoch": 4.6, "learning_rate": 1.1775832860229034e-05, "loss": 0.2982, "step": 14953 }, { "epoch": 4.6, "learning_rate": 1.1774852150548966e-05, "loss": 0.3465, "step": 14954 }, { "epoch": 4.6, "learning_rate": 1.1773871423242987e-05, "loss": 0.327, "step": 14955 }, { "epoch": 4.6, "learning_rate": 1.1772890678320828e-05, "loss": 0.3676, "step": 14956 }, { "epoch": 4.6, "learning_rate": 1.177190991579223e-05, "loss": 0.2901, "step": 14957 }, { "epoch": 4.6, "learning_rate": 1.1770929135666933e-05, "loss": 0.2392, "step": 14958 }, { "epoch": 4.6, "learning_rate": 1.1769948337954677e-05, "loss": 0.3905, "step": 14959 }, { "epoch": 4.6, "learning_rate": 1.1768967522665204e-05, "loss": 0.4163, "step": 14960 }, { "epoch": 4.6, "learning_rate": 1.1767986689808255e-05, "loss": 0.2963, "step": 14961 }, { "epoch": 4.6, "learning_rate": 1.1767005839393565e-05, "loss": 0.4585, "step": 14962 }, { "epoch": 4.6, "learning_rate": 1.1766024971430882e-05, "loss": 0.3648, "step": 14963 }, { "epoch": 4.6, "learning_rate": 1.176504408592994e-05, "loss": 0.2774, "step": 14964 }, { "epoch": 4.6, "learning_rate": 1.1764063182900487e-05, "loss": 0.2155, "step": 14965 }, { "epoch": 4.6, "learning_rate": 1.1763082262352257e-05, "loss": 0.3685, "step": 14966 }, { "epoch": 4.61, "learning_rate": 1.1762101324294998e-05, "loss": 0.3346, "step": 14967 }, { "epoch": 4.61, "learning_rate": 1.1761120368738448e-05, "loss": 0.2121, "step": 14968 }, { "epoch": 4.61, "learning_rate": 1.1760139395692352e-05, "loss": 0.3024, "step": 14969 }, { "epoch": 4.61, "learning_rate": 1.1759158405166446e-05, "loss": 0.283, "step": 14970 }, { "epoch": 4.61, "learning_rate": 1.1758177397170477e-05, "loss": 0.2948, "step": 14971 }, { "epoch": 4.61, "learning_rate": 1.1757196371714187e-05, "loss": 0.4393, "step": 14972 }, { "epoch": 4.61, "learning_rate": 1.1756215328807316e-05, "loss": 0.1998, "step": 14973 }, { "epoch": 4.61, "learning_rate": 1.175523426845961e-05, "loss": 0.3579, "step": 14974 }, { "epoch": 4.61, "learning_rate": 1.175425319068081e-05, "loss": 0.2253, "step": 14975 }, { "epoch": 4.61, "learning_rate": 1.1753272095480654e-05, "loss": 0.2473, "step": 14976 }, { "epoch": 4.61, "learning_rate": 1.1752290982868899e-05, "loss": 0.3377, "step": 14977 }, { "epoch": 4.61, "learning_rate": 1.1751309852855272e-05, "loss": 0.4538, "step": 14978 }, { "epoch": 4.61, "learning_rate": 1.1750328705449528e-05, "loss": 0.2957, "step": 14979 }, { "epoch": 4.61, "learning_rate": 1.1749347540661406e-05, "loss": 0.3221, "step": 14980 }, { "epoch": 4.61, "learning_rate": 1.1748366358500648e-05, "loss": 0.3342, "step": 14981 }, { "epoch": 4.61, "learning_rate": 1.1747385158977008e-05, "loss": 0.45, "step": 14982 }, { "epoch": 4.61, "learning_rate": 1.1746403942100215e-05, "loss": 0.3382, "step": 14983 }, { "epoch": 4.61, "learning_rate": 1.1745422707880025e-05, "loss": 0.3008, "step": 14984 }, { "epoch": 4.61, "learning_rate": 1.1744441456326177e-05, "loss": 0.3463, "step": 14985 }, { "epoch": 4.61, "learning_rate": 1.174346018744842e-05, "loss": 0.4982, "step": 14986 }, { "epoch": 4.61, "learning_rate": 1.1742478901256496e-05, "loss": 0.1605, "step": 14987 }, { "epoch": 4.61, "learning_rate": 1.174149759776015e-05, "loss": 0.411, "step": 14988 }, { "epoch": 4.61, "learning_rate": 1.174051627696913e-05, "loss": 0.3132, "step": 14989 }, { "epoch": 4.61, "learning_rate": 1.1739534938893174e-05, "loss": 0.302, "step": 14990 }, { "epoch": 4.61, "learning_rate": 1.1738553583542035e-05, "loss": 0.3047, "step": 14991 }, { "epoch": 4.61, "learning_rate": 1.1737572210925459e-05, "loss": 0.2442, "step": 14992 }, { "epoch": 4.61, "learning_rate": 1.173659082105319e-05, "loss": 0.3936, "step": 14993 }, { "epoch": 4.61, "learning_rate": 1.1735609413934972e-05, "loss": 0.2627, "step": 14994 }, { "epoch": 4.61, "learning_rate": 1.1734627989580552e-05, "loss": 0.2091, "step": 14995 }, { "epoch": 4.61, "learning_rate": 1.1733646547999678e-05, "loss": 0.2837, "step": 14996 }, { "epoch": 4.61, "learning_rate": 1.1732665089202095e-05, "loss": 0.5345, "step": 14997 }, { "epoch": 4.61, "learning_rate": 1.1731683613197554e-05, "loss": 0.3213, "step": 14998 }, { "epoch": 4.61, "learning_rate": 1.1730702119995798e-05, "loss": 0.3157, "step": 14999 }, { "epoch": 4.62, "learning_rate": 1.1729720609606572e-05, "loss": 0.4477, "step": 15000 }, { "epoch": 4.62, "learning_rate": 1.1728739082039627e-05, "loss": 0.2435, "step": 15001 }, { "epoch": 4.62, "learning_rate": 1.1727757537304712e-05, "loss": 0.2037, "step": 15002 }, { "epoch": 4.62, "learning_rate": 1.172677597541157e-05, "loss": 0.4418, "step": 15003 }, { "epoch": 4.62, "learning_rate": 1.1725794396369951e-05, "loss": 0.3717, "step": 15004 }, { "epoch": 4.62, "learning_rate": 1.1724812800189604e-05, "loss": 0.2888, "step": 15005 }, { "epoch": 4.62, "learning_rate": 1.1723831186880277e-05, "loss": 0.2477, "step": 15006 }, { "epoch": 4.62, "learning_rate": 1.1722849556451716e-05, "loss": 0.3117, "step": 15007 }, { "epoch": 4.62, "learning_rate": 1.1721867908913673e-05, "loss": 0.2164, "step": 15008 }, { "epoch": 4.62, "learning_rate": 1.1720886244275893e-05, "loss": 0.2502, "step": 15009 }, { "epoch": 4.62, "learning_rate": 1.1719904562548128e-05, "loss": 0.2317, "step": 15010 }, { "epoch": 4.62, "learning_rate": 1.1718922863740126e-05, "loss": 0.351, "step": 15011 }, { "epoch": 4.62, "learning_rate": 1.1717941147861633e-05, "loss": 0.3185, "step": 15012 }, { "epoch": 4.62, "learning_rate": 1.1716959414922404e-05, "loss": 0.2636, "step": 15013 }, { "epoch": 4.62, "learning_rate": 1.1715977664932183e-05, "loss": 0.2555, "step": 15014 }, { "epoch": 4.62, "learning_rate": 1.1714995897900723e-05, "loss": 0.3455, "step": 15015 }, { "epoch": 4.62, "learning_rate": 1.1714014113837772e-05, "loss": 0.2293, "step": 15016 }, { "epoch": 4.62, "learning_rate": 1.1713032312753086e-05, "loss": 0.2698, "step": 15017 }, { "epoch": 4.62, "learning_rate": 1.1712050494656407e-05, "loss": 0.3431, "step": 15018 }, { "epoch": 4.62, "learning_rate": 1.1711068659557488e-05, "loss": 0.2505, "step": 15019 }, { "epoch": 4.62, "learning_rate": 1.1710086807466082e-05, "loss": 0.2011, "step": 15020 }, { "epoch": 4.62, "learning_rate": 1.1709104938391936e-05, "loss": 0.4554, "step": 15021 }, { "epoch": 4.62, "learning_rate": 1.1708123052344803e-05, "loss": 0.2884, "step": 15022 }, { "epoch": 4.62, "learning_rate": 1.1707141149334435e-05, "loss": 0.2744, "step": 15023 }, { "epoch": 4.62, "learning_rate": 1.1707141149334435e-05, "loss": 0.3427, "step": 15024 }, { "epoch": 4.62, "learning_rate": 1.1706159229370584e-05, "loss": 0.3045, "step": 15025 }, { "epoch": 4.62, "learning_rate": 1.1705177292462997e-05, "loss": 0.3454, "step": 15026 }, { "epoch": 4.62, "learning_rate": 1.1704195338621426e-05, "loss": 0.2371, "step": 15027 }, { "epoch": 4.62, "learning_rate": 1.1703213367855629e-05, "loss": 0.3308, "step": 15028 }, { "epoch": 4.62, "learning_rate": 1.170223138017535e-05, "loss": 0.3701, "step": 15029 }, { "epoch": 4.62, "learning_rate": 1.1701249375590346e-05, "loss": 0.3972, "step": 15030 }, { "epoch": 4.62, "learning_rate": 1.170026735411037e-05, "loss": 0.2563, "step": 15031 }, { "epoch": 4.63, "learning_rate": 1.1699285315745167e-05, "loss": 0.1717, "step": 15032 }, { "epoch": 4.63, "learning_rate": 1.16983032605045e-05, "loss": 0.3289, "step": 15033 }, { "epoch": 4.63, "learning_rate": 1.1697321188398115e-05, "loss": 0.3486, "step": 15034 }, { "epoch": 4.63, "learning_rate": 1.1696339099435764e-05, "loss": 0.3108, "step": 15035 }, { "epoch": 4.63, "learning_rate": 1.1695356993627203e-05, "loss": 0.6576, "step": 15036 }, { "epoch": 4.63, "learning_rate": 1.1694374870982185e-05, "loss": 0.217, "step": 15037 }, { "epoch": 4.63, "learning_rate": 1.1693392731510464e-05, "loss": 0.2676, "step": 15038 }, { "epoch": 4.63, "learning_rate": 1.1692410575221791e-05, "loss": 0.5103, "step": 15039 }, { "epoch": 4.63, "learning_rate": 1.1691428402125923e-05, "loss": 0.2236, "step": 15040 }, { "epoch": 4.63, "learning_rate": 1.1690446212232609e-05, "loss": 0.1783, "step": 15041 }, { "epoch": 4.63, "learning_rate": 1.168946400555161e-05, "loss": 0.329, "step": 15042 }, { "epoch": 4.63, "learning_rate": 1.1688481782092676e-05, "loss": 0.3003, "step": 15043 }, { "epoch": 4.63, "learning_rate": 1.1687499541865559e-05, "loss": 0.3751, "step": 15044 }, { "epoch": 4.63, "learning_rate": 1.1686517284880018e-05, "loss": 0.2166, "step": 15045 }, { "epoch": 4.63, "learning_rate": 1.1685535011145803e-05, "loss": 0.493, "step": 15046 }, { "epoch": 4.63, "learning_rate": 1.1684552720672677e-05, "loss": 0.3485, "step": 15047 }, { "epoch": 4.63, "learning_rate": 1.1683570413470384e-05, "loss": 0.2742, "step": 15048 }, { "epoch": 4.63, "learning_rate": 1.1682588089548692e-05, "loss": 0.3915, "step": 15049 }, { "epoch": 4.63, "learning_rate": 1.1681605748917347e-05, "loss": 0.3419, "step": 15050 }, { "epoch": 4.63, "learning_rate": 1.1680623391586106e-05, "loss": 0.2526, "step": 15051 }, { "epoch": 4.63, "learning_rate": 1.1679641017564725e-05, "loss": 0.2902, "step": 15052 }, { "epoch": 4.63, "learning_rate": 1.1678658626862959e-05, "loss": 0.305, "step": 15053 }, { "epoch": 4.63, "learning_rate": 1.1677676219490573e-05, "loss": 0.1884, "step": 15054 }, { "epoch": 4.63, "learning_rate": 1.167669379545731e-05, "loss": 0.3438, "step": 15055 }, { "epoch": 4.63, "learning_rate": 1.1675711354772934e-05, "loss": 0.3954, "step": 15056 }, { "epoch": 4.63, "learning_rate": 1.16747288974472e-05, "loss": 0.3627, "step": 15057 }, { "epoch": 4.63, "learning_rate": 1.1673746423489862e-05, "loss": 0.306, "step": 15058 }, { "epoch": 4.63, "learning_rate": 1.1672763932910685e-05, "loss": 0.2897, "step": 15059 }, { "epoch": 4.63, "learning_rate": 1.1671781425719415e-05, "loss": 0.4529, "step": 15060 }, { "epoch": 4.63, "learning_rate": 1.1670798901925817e-05, "loss": 0.377, "step": 15061 }, { "epoch": 4.63, "learning_rate": 1.1669816361539647e-05, "loss": 0.2503, "step": 15062 }, { "epoch": 4.63, "learning_rate": 1.1668833804570658e-05, "loss": 0.3857, "step": 15063 }, { "epoch": 4.63, "learning_rate": 1.1667851231028617e-05, "loss": 0.342, "step": 15064 }, { "epoch": 4.64, "learning_rate": 1.1666868640923272e-05, "loss": 0.2134, "step": 15065 }, { "epoch": 4.64, "learning_rate": 1.1665886034264386e-05, "loss": 0.3296, "step": 15066 }, { "epoch": 4.64, "learning_rate": 1.1664903411061714e-05, "loss": 0.3161, "step": 15067 }, { "epoch": 4.64, "learning_rate": 1.1663920771325022e-05, "loss": 0.1375, "step": 15068 }, { "epoch": 4.64, "learning_rate": 1.1662938115064061e-05, "loss": 0.329, "step": 15069 }, { "epoch": 4.64, "learning_rate": 1.166195544228859e-05, "loss": 0.153, "step": 15070 }, { "epoch": 4.64, "learning_rate": 1.166097275300837e-05, "loss": 0.3333, "step": 15071 }, { "epoch": 4.64, "learning_rate": 1.165999004723316e-05, "loss": 0.3506, "step": 15072 }, { "epoch": 4.64, "learning_rate": 1.1659007324972722e-05, "loss": 0.2465, "step": 15073 }, { "epoch": 4.64, "learning_rate": 1.1658024586236808e-05, "loss": 0.4906, "step": 15074 }, { "epoch": 4.64, "learning_rate": 1.1657041831035186e-05, "loss": 0.2065, "step": 15075 }, { "epoch": 4.64, "learning_rate": 1.1656059059377609e-05, "loss": 0.3127, "step": 15076 }, { "epoch": 4.64, "learning_rate": 1.1655076271273837e-05, "loss": 0.4255, "step": 15077 }, { "epoch": 4.64, "learning_rate": 1.1654093466733636e-05, "loss": 0.3591, "step": 15078 }, { "epoch": 4.64, "learning_rate": 1.1653110645766763e-05, "loss": 0.2793, "step": 15079 }, { "epoch": 4.64, "learning_rate": 1.1652127808382976e-05, "loss": 0.26, "step": 15080 }, { "epoch": 4.64, "learning_rate": 1.1651144954592037e-05, "loss": 0.2756, "step": 15081 }, { "epoch": 4.64, "learning_rate": 1.1650162084403704e-05, "loss": 0.286, "step": 15082 }, { "epoch": 4.64, "learning_rate": 1.1649179197827744e-05, "loss": 0.2652, "step": 15083 }, { "epoch": 4.64, "learning_rate": 1.1648196294873916e-05, "loss": 0.3423, "step": 15084 }, { "epoch": 4.64, "learning_rate": 1.1647213375551981e-05, "loss": 0.2474, "step": 15085 }, { "epoch": 4.64, "learning_rate": 1.1646230439871693e-05, "loss": 0.3626, "step": 15086 }, { "epoch": 4.64, "learning_rate": 1.1645247487842824e-05, "loss": 0.1532, "step": 15087 }, { "epoch": 4.64, "learning_rate": 1.164426451947513e-05, "loss": 0.4919, "step": 15088 }, { "epoch": 4.64, "learning_rate": 1.1643281534778372e-05, "loss": 0.3995, "step": 15089 }, { "epoch": 4.64, "learning_rate": 1.1642298533762318e-05, "loss": 0.3701, "step": 15090 }, { "epoch": 4.64, "learning_rate": 1.1641315516436722e-05, "loss": 0.3013, "step": 15091 }, { "epoch": 4.64, "learning_rate": 1.1640332482811352e-05, "loss": 0.2407, "step": 15092 }, { "epoch": 4.64, "learning_rate": 1.1639349432895967e-05, "loss": 0.2976, "step": 15093 }, { "epoch": 4.64, "learning_rate": 1.1638366366700335e-05, "loss": 0.2981, "step": 15094 }, { "epoch": 4.64, "learning_rate": 1.1637383284234214e-05, "loss": 0.3844, "step": 15095 }, { "epoch": 4.64, "learning_rate": 1.1636400185507364e-05, "loss": 0.3441, "step": 15096 }, { "epoch": 4.65, "learning_rate": 1.1635417070529556e-05, "loss": 0.3702, "step": 15097 }, { "epoch": 4.65, "learning_rate": 1.1634433939310548e-05, "loss": 0.3963, "step": 15098 }, { "epoch": 4.65, "learning_rate": 1.1633450791860104e-05, "loss": 0.2924, "step": 15099 }, { "epoch": 4.65, "learning_rate": 1.163246762818799e-05, "loss": 0.3428, "step": 15100 }, { "epoch": 4.65, "learning_rate": 1.1631484448303964e-05, "loss": 0.3815, "step": 15101 }, { "epoch": 4.65, "learning_rate": 1.1630501252217798e-05, "loss": 0.2356, "step": 15102 }, { "epoch": 4.65, "learning_rate": 1.1629518039939249e-05, "loss": 0.3703, "step": 15103 }, { "epoch": 4.65, "learning_rate": 1.1628534811478087e-05, "loss": 0.2924, "step": 15104 }, { "epoch": 4.65, "learning_rate": 1.162755156684407e-05, "loss": 0.3029, "step": 15105 }, { "epoch": 4.65, "learning_rate": 1.1626568306046969e-05, "loss": 0.2913, "step": 15106 }, { "epoch": 4.65, "learning_rate": 1.1625585029096544e-05, "loss": 0.2485, "step": 15107 }, { "epoch": 4.65, "learning_rate": 1.162460173600256e-05, "loss": 0.1902, "step": 15108 }, { "epoch": 4.65, "learning_rate": 1.162361842677479e-05, "loss": 0.2483, "step": 15109 }, { "epoch": 4.65, "learning_rate": 1.1622635101422985e-05, "loss": 0.2526, "step": 15110 }, { "epoch": 4.65, "learning_rate": 1.1621651759956922e-05, "loss": 0.3708, "step": 15111 }, { "epoch": 4.65, "learning_rate": 1.1620668402386362e-05, "loss": 0.2539, "step": 15112 }, { "epoch": 4.65, "learning_rate": 1.1619685028721068e-05, "loss": 0.2298, "step": 15113 }, { "epoch": 4.65, "learning_rate": 1.1618701638970815e-05, "loss": 0.2228, "step": 15114 }, { "epoch": 4.65, "learning_rate": 1.161771823314536e-05, "loss": 0.2472, "step": 15115 }, { "epoch": 4.65, "learning_rate": 1.161673481125447e-05, "loss": 0.3076, "step": 15116 }, { "epoch": 4.65, "learning_rate": 1.1615751373307914e-05, "loss": 0.352, "step": 15117 }, { "epoch": 4.65, "learning_rate": 1.1614767919315457e-05, "loss": 0.2041, "step": 15118 }, { "epoch": 4.65, "learning_rate": 1.1613784449286872e-05, "loss": 0.24, "step": 15119 }, { "epoch": 4.65, "learning_rate": 1.1612800963231914e-05, "loss": 0.3414, "step": 15120 }, { "epoch": 4.65, "learning_rate": 1.1611817461160357e-05, "loss": 0.3092, "step": 15121 }, { "epoch": 4.65, "learning_rate": 1.1610833943081966e-05, "loss": 0.3243, "step": 15122 }, { "epoch": 4.65, "learning_rate": 1.1609850409006513e-05, "loss": 0.278, "step": 15123 }, { "epoch": 4.65, "learning_rate": 1.1608866858943762e-05, "loss": 0.3218, "step": 15124 }, { "epoch": 4.65, "learning_rate": 1.1607883292903476e-05, "loss": 0.3217, "step": 15125 }, { "epoch": 4.65, "learning_rate": 1.1606899710895431e-05, "loss": 0.4078, "step": 15126 }, { "epoch": 4.65, "learning_rate": 1.1605916112929388e-05, "loss": 0.2168, "step": 15127 }, { "epoch": 4.65, "learning_rate": 1.160493249901512e-05, "loss": 0.3352, "step": 15128 }, { "epoch": 4.65, "learning_rate": 1.1603948869162393e-05, "loss": 0.5141, "step": 15129 }, { "epoch": 4.66, "learning_rate": 1.1602965223380975e-05, "loss": 0.4301, "step": 15130 }, { "epoch": 4.66, "learning_rate": 1.1601981561680637e-05, "loss": 0.3226, "step": 15131 }, { "epoch": 4.66, "learning_rate": 1.1600997884071141e-05, "loss": 0.2654, "step": 15132 }, { "epoch": 4.66, "learning_rate": 1.1600014190562263e-05, "loss": 0.3308, "step": 15133 }, { "epoch": 4.66, "learning_rate": 1.1599030481163768e-05, "loss": 0.3312, "step": 15134 }, { "epoch": 4.66, "learning_rate": 1.1598046755885433e-05, "loss": 0.2739, "step": 15135 }, { "epoch": 4.66, "learning_rate": 1.1597063014737013e-05, "loss": 0.2645, "step": 15136 }, { "epoch": 4.66, "learning_rate": 1.159607925772829e-05, "loss": 0.4126, "step": 15137 }, { "epoch": 4.66, "learning_rate": 1.1595095484869028e-05, "loss": 0.3876, "step": 15138 }, { "epoch": 4.66, "learning_rate": 1.1594111696168995e-05, "loss": 0.3166, "step": 15139 }, { "epoch": 4.66, "learning_rate": 1.1593127891637968e-05, "loss": 0.3838, "step": 15140 }, { "epoch": 4.66, "learning_rate": 1.1592144071285709e-05, "loss": 0.2032, "step": 15141 }, { "epoch": 4.66, "learning_rate": 1.1591160235121995e-05, "loss": 0.3435, "step": 15142 }, { "epoch": 4.66, "learning_rate": 1.1590176383156593e-05, "loss": 0.4, "step": 15143 }, { "epoch": 4.66, "learning_rate": 1.1589192515399271e-05, "loss": 0.2809, "step": 15144 }, { "epoch": 4.66, "learning_rate": 1.1588208631859808e-05, "loss": 0.2635, "step": 15145 }, { "epoch": 4.66, "learning_rate": 1.1587224732547964e-05, "loss": 0.2501, "step": 15146 }, { "epoch": 4.66, "learning_rate": 1.1586240817473517e-05, "loss": 0.3142, "step": 15147 }, { "epoch": 4.66, "learning_rate": 1.1585256886646239e-05, "loss": 0.3515, "step": 15148 }, { "epoch": 4.66, "learning_rate": 1.1584272940075897e-05, "loss": 0.3323, "step": 15149 }, { "epoch": 4.66, "learning_rate": 1.1583288977772266e-05, "loss": 0.2798, "step": 15150 }, { "epoch": 4.66, "learning_rate": 1.1582304999745115e-05, "loss": 0.2546, "step": 15151 }, { "epoch": 4.66, "learning_rate": 1.1581321006004218e-05, "loss": 0.3805, "step": 15152 }, { "epoch": 4.66, "learning_rate": 1.1580336996559343e-05, "loss": 0.2295, "step": 15153 }, { "epoch": 4.66, "learning_rate": 1.1579352971420267e-05, "loss": 0.2932, "step": 15154 }, { "epoch": 4.66, "learning_rate": 1.1578368930596764e-05, "loss": 0.3365, "step": 15155 }, { "epoch": 4.66, "learning_rate": 1.1577384874098601e-05, "loss": 0.2691, "step": 15156 }, { "epoch": 4.66, "learning_rate": 1.1576400801935553e-05, "loss": 0.1305, "step": 15157 }, { "epoch": 4.66, "learning_rate": 1.1575416714117387e-05, "loss": 0.319, "step": 15158 }, { "epoch": 4.66, "learning_rate": 1.1574432610653885e-05, "loss": 0.492, "step": 15159 }, { "epoch": 4.66, "learning_rate": 1.1573448491554817e-05, "loss": 0.3702, "step": 15160 }, { "epoch": 4.66, "learning_rate": 1.1572464356829956e-05, "loss": 0.4333, "step": 15161 }, { "epoch": 4.67, "learning_rate": 1.1571480206489071e-05, "loss": 0.284, "step": 15162 }, { "epoch": 4.67, "learning_rate": 1.157049604054194e-05, "loss": 0.2209, "step": 15163 }, { "epoch": 4.67, "learning_rate": 1.156951185899834e-05, "loss": 0.2606, "step": 15164 }, { "epoch": 4.67, "learning_rate": 1.1568527661868037e-05, "loss": 0.3831, "step": 15165 }, { "epoch": 4.67, "learning_rate": 1.156754344916081e-05, "loss": 0.2184, "step": 15166 }, { "epoch": 4.67, "learning_rate": 1.1566559220886428e-05, "loss": 0.3639, "step": 15167 }, { "epoch": 4.67, "learning_rate": 1.1565574977054675e-05, "loss": 0.3425, "step": 15168 }, { "epoch": 4.67, "learning_rate": 1.1564590717675318e-05, "loss": 0.5302, "step": 15169 }, { "epoch": 4.67, "learning_rate": 1.156360644275813e-05, "loss": 0.2908, "step": 15170 }, { "epoch": 4.67, "learning_rate": 1.1562622152312894e-05, "loss": 0.2937, "step": 15171 }, { "epoch": 4.67, "learning_rate": 1.1561637846349374e-05, "loss": 0.2439, "step": 15172 }, { "epoch": 4.67, "learning_rate": 1.1560653524877356e-05, "loss": 0.3369, "step": 15173 }, { "epoch": 4.67, "learning_rate": 1.1559669187906608e-05, "loss": 0.2488, "step": 15174 }, { "epoch": 4.67, "learning_rate": 1.155868483544691e-05, "loss": 0.2935, "step": 15175 }, { "epoch": 4.67, "learning_rate": 1.1557700467508032e-05, "loss": 0.3918, "step": 15176 }, { "epoch": 4.67, "learning_rate": 1.1556716084099753e-05, "loss": 0.6809, "step": 15177 }, { "epoch": 4.67, "learning_rate": 1.1555731685231848e-05, "loss": 0.3914, "step": 15178 }, { "epoch": 4.67, "learning_rate": 1.1554747270914098e-05, "loss": 0.5092, "step": 15179 }, { "epoch": 4.67, "learning_rate": 1.155376284115627e-05, "loss": 0.2983, "step": 15180 }, { "epoch": 4.67, "learning_rate": 1.1552778395968147e-05, "loss": 0.2766, "step": 15181 }, { "epoch": 4.67, "learning_rate": 1.1551793935359501e-05, "loss": 0.2023, "step": 15182 }, { "epoch": 4.67, "learning_rate": 1.1550809459340114e-05, "loss": 0.346, "step": 15183 }, { "epoch": 4.67, "learning_rate": 1.1549824967919756e-05, "loss": 0.227, "step": 15184 }, { "epoch": 4.67, "learning_rate": 1.1548840461108216e-05, "loss": 0.2368, "step": 15185 }, { "epoch": 4.67, "learning_rate": 1.1547855938915253e-05, "loss": 0.3893, "step": 15186 }, { "epoch": 4.67, "learning_rate": 1.1546871401350658e-05, "loss": 0.2949, "step": 15187 }, { "epoch": 4.67, "learning_rate": 1.1545886848424206e-05, "loss": 0.3888, "step": 15188 }, { "epoch": 4.67, "learning_rate": 1.1544902280145668e-05, "loss": 0.447, "step": 15189 }, { "epoch": 4.67, "learning_rate": 1.1543917696524836e-05, "loss": 0.2752, "step": 15190 }, { "epoch": 4.67, "learning_rate": 1.154293309757147e-05, "loss": 0.2839, "step": 15191 }, { "epoch": 4.67, "learning_rate": 1.1541948483295358e-05, "loss": 0.359, "step": 15192 }, { "epoch": 4.67, "learning_rate": 1.1540963853706277e-05, "loss": 0.3157, "step": 15193 }, { "epoch": 4.67, "learning_rate": 1.1539979208814005e-05, "loss": 0.3379, "step": 15194 }, { "epoch": 4.68, "learning_rate": 1.1538994548628322e-05, "loss": 0.3917, "step": 15195 }, { "epoch": 4.68, "learning_rate": 1.1538009873159003e-05, "loss": 0.1615, "step": 15196 }, { "epoch": 4.68, "learning_rate": 1.1537025182415826e-05, "loss": 0.3343, "step": 15197 }, { "epoch": 4.68, "learning_rate": 1.1536040476408576e-05, "loss": 0.3043, "step": 15198 }, { "epoch": 4.68, "learning_rate": 1.1535055755147024e-05, "loss": 0.3771, "step": 15199 }, { "epoch": 4.68, "learning_rate": 1.153407101864096e-05, "loss": 0.3461, "step": 15200 }, { "epoch": 4.68, "learning_rate": 1.1533086266900151e-05, "loss": 0.2092, "step": 15201 }, { "epoch": 4.68, "learning_rate": 1.1532101499934387e-05, "loss": 0.248, "step": 15202 }, { "epoch": 4.68, "learning_rate": 1.1531116717753438e-05, "loss": 0.2626, "step": 15203 }, { "epoch": 4.68, "learning_rate": 1.1530131920367094e-05, "loss": 0.2982, "step": 15204 }, { "epoch": 4.68, "learning_rate": 1.1529147107785129e-05, "loss": 0.2519, "step": 15205 }, { "epoch": 4.68, "learning_rate": 1.1528162280017324e-05, "loss": 0.2857, "step": 15206 }, { "epoch": 4.68, "learning_rate": 1.1527177437073455e-05, "loss": 0.275, "step": 15207 }, { "epoch": 4.68, "learning_rate": 1.152619257896331e-05, "loss": 0.2795, "step": 15208 }, { "epoch": 4.68, "learning_rate": 1.1525207705696666e-05, "loss": 0.3347, "step": 15209 }, { "epoch": 4.68, "learning_rate": 1.1524222817283304e-05, "loss": 0.2568, "step": 15210 }, { "epoch": 4.68, "learning_rate": 1.1523237913733004e-05, "loss": 0.3736, "step": 15211 }, { "epoch": 4.68, "learning_rate": 1.1522252995055548e-05, "loss": 0.4206, "step": 15212 }, { "epoch": 4.68, "learning_rate": 1.1521268061260716e-05, "loss": 0.2073, "step": 15213 }, { "epoch": 4.68, "learning_rate": 1.1520283112358291e-05, "loss": 0.1766, "step": 15214 }, { "epoch": 4.68, "learning_rate": 1.151929814835805e-05, "loss": 0.2915, "step": 15215 }, { "epoch": 4.68, "learning_rate": 1.1518313169269784e-05, "loss": 0.3436, "step": 15216 }, { "epoch": 4.68, "learning_rate": 1.1517328175103266e-05, "loss": 0.3212, "step": 15217 }, { "epoch": 4.68, "learning_rate": 1.151634316586828e-05, "loss": 0.3067, "step": 15218 }, { "epoch": 4.68, "learning_rate": 1.1515358141574611e-05, "loss": 0.3615, "step": 15219 }, { "epoch": 4.68, "learning_rate": 1.1514373102232036e-05, "loss": 0.3226, "step": 15220 }, { "epoch": 4.68, "learning_rate": 1.1513388047850344e-05, "loss": 0.3489, "step": 15221 }, { "epoch": 4.68, "learning_rate": 1.1512402978439312e-05, "loss": 0.3431, "step": 15222 }, { "epoch": 4.68, "learning_rate": 1.1511417894008725e-05, "loss": 0.2194, "step": 15223 }, { "epoch": 4.68, "learning_rate": 1.1510432794568365e-05, "loss": 0.2289, "step": 15224 }, { "epoch": 4.68, "learning_rate": 1.1509447680128014e-05, "loss": 0.2034, "step": 15225 }, { "epoch": 4.68, "learning_rate": 1.150846255069746e-05, "loss": 0.1983, "step": 15226 }, { "epoch": 4.69, "learning_rate": 1.1507477406286478e-05, "loss": 0.1861, "step": 15227 }, { "epoch": 4.69, "learning_rate": 1.150649224690486e-05, "loss": 0.2413, "step": 15228 }, { "epoch": 4.69, "learning_rate": 1.1505507072562385e-05, "loss": 0.1716, "step": 15229 }, { "epoch": 4.69, "learning_rate": 1.1504521883268836e-05, "loss": 0.2666, "step": 15230 }, { "epoch": 4.69, "learning_rate": 1.1503536679034e-05, "loss": 0.2785, "step": 15231 }, { "epoch": 4.69, "learning_rate": 1.1502551459867655e-05, "loss": 0.2992, "step": 15232 }, { "epoch": 4.69, "learning_rate": 1.1501566225779592e-05, "loss": 0.2701, "step": 15233 }, { "epoch": 4.69, "learning_rate": 1.1500580976779592e-05, "loss": 0.2551, "step": 15234 }, { "epoch": 4.69, "learning_rate": 1.149959571287744e-05, "loss": 0.3535, "step": 15235 }, { "epoch": 4.69, "learning_rate": 1.1498610434082923e-05, "loss": 0.3258, "step": 15236 }, { "epoch": 4.69, "learning_rate": 1.149762514040582e-05, "loss": 0.4136, "step": 15237 }, { "epoch": 4.69, "learning_rate": 1.149663983185592e-05, "loss": 0.3137, "step": 15238 }, { "epoch": 4.69, "learning_rate": 1.1495654508443005e-05, "loss": 0.3662, "step": 15239 }, { "epoch": 4.69, "learning_rate": 1.1494669170176865e-05, "loss": 0.3959, "step": 15240 }, { "epoch": 4.69, "learning_rate": 1.1493683817067283e-05, "loss": 0.2104, "step": 15241 }, { "epoch": 4.69, "learning_rate": 1.1492698449124042e-05, "loss": 0.2539, "step": 15242 }, { "epoch": 4.69, "learning_rate": 1.1491713066356933e-05, "loss": 0.2364, "step": 15243 }, { "epoch": 4.69, "learning_rate": 1.1490727668775735e-05, "loss": 0.2866, "step": 15244 }, { "epoch": 4.69, "learning_rate": 1.1489742256390237e-05, "loss": 0.3361, "step": 15245 }, { "epoch": 4.69, "learning_rate": 1.1488756829210228e-05, "loss": 0.3924, "step": 15246 }, { "epoch": 4.69, "learning_rate": 1.1487771387245492e-05, "loss": 0.2894, "step": 15247 }, { "epoch": 4.69, "learning_rate": 1.1486785930505812e-05, "loss": 0.2572, "step": 15248 }, { "epoch": 4.69, "learning_rate": 1.1485800459000977e-05, "loss": 0.1937, "step": 15249 }, { "epoch": 4.69, "learning_rate": 1.1484814972740779e-05, "loss": 0.3276, "step": 15250 }, { "epoch": 4.69, "learning_rate": 1.1483829471734996e-05, "loss": 0.3813, "step": 15251 }, { "epoch": 4.69, "learning_rate": 1.1482843955993421e-05, "loss": 0.3296, "step": 15252 }, { "epoch": 4.69, "learning_rate": 1.1481858425525835e-05, "loss": 0.3048, "step": 15253 }, { "epoch": 4.69, "learning_rate": 1.1480872880342034e-05, "loss": 0.2454, "step": 15254 }, { "epoch": 4.69, "learning_rate": 1.1479887320451796e-05, "loss": 0.3281, "step": 15255 }, { "epoch": 4.69, "learning_rate": 1.1478901745864917e-05, "loss": 0.3759, "step": 15256 }, { "epoch": 4.69, "learning_rate": 1.147791615659118e-05, "loss": 0.1868, "step": 15257 }, { "epoch": 4.69, "learning_rate": 1.147693055264037e-05, "loss": 0.3118, "step": 15258 }, { "epoch": 4.69, "learning_rate": 1.147594493402228e-05, "loss": 0.3857, "step": 15259 }, { "epoch": 4.7, "learning_rate": 1.1474959300746699e-05, "loss": 0.2674, "step": 15260 }, { "epoch": 4.7, "learning_rate": 1.1473973652823412e-05, "loss": 0.3529, "step": 15261 }, { "epoch": 4.7, "learning_rate": 1.1472987990262208e-05, "loss": 0.2646, "step": 15262 }, { "epoch": 4.7, "learning_rate": 1.1472002313072874e-05, "loss": 0.4184, "step": 15263 }, { "epoch": 4.7, "learning_rate": 1.14710166212652e-05, "loss": 0.3814, "step": 15264 }, { "epoch": 4.7, "learning_rate": 1.1470030914848976e-05, "loss": 0.4423, "step": 15265 }, { "epoch": 4.7, "learning_rate": 1.1469045193833994e-05, "loss": 0.3673, "step": 15266 }, { "epoch": 4.7, "learning_rate": 1.1468059458230036e-05, "loss": 0.3826, "step": 15267 }, { "epoch": 4.7, "learning_rate": 1.1467073708046894e-05, "loss": 0.2366, "step": 15268 }, { "epoch": 4.7, "learning_rate": 1.1466087943294357e-05, "loss": 0.1954, "step": 15269 }, { "epoch": 4.7, "learning_rate": 1.1465102163982218e-05, "loss": 0.214, "step": 15270 }, { "epoch": 4.7, "learning_rate": 1.1464116370120263e-05, "loss": 0.3839, "step": 15271 }, { "epoch": 4.7, "learning_rate": 1.1463130561718284e-05, "loss": 0.3333, "step": 15272 }, { "epoch": 4.7, "learning_rate": 1.1462144738786069e-05, "loss": 0.1936, "step": 15273 }, { "epoch": 4.7, "learning_rate": 1.1461158901333412e-05, "loss": 0.3174, "step": 15274 }, { "epoch": 4.7, "learning_rate": 1.1460173049370097e-05, "loss": 0.3403, "step": 15275 }, { "epoch": 4.7, "learning_rate": 1.145918718290592e-05, "loss": 0.3188, "step": 15276 }, { "epoch": 4.7, "learning_rate": 1.1458201301950667e-05, "loss": 0.3294, "step": 15277 }, { "epoch": 4.7, "learning_rate": 1.1457215406514132e-05, "loss": 0.2341, "step": 15278 }, { "epoch": 4.7, "learning_rate": 1.1456229496606108e-05, "loss": 0.2849, "step": 15279 }, { "epoch": 4.7, "learning_rate": 1.1455243572236378e-05, "loss": 0.2527, "step": 15280 }, { "epoch": 4.7, "learning_rate": 1.1454257633414745e-05, "loss": 0.299, "step": 15281 }, { "epoch": 4.7, "learning_rate": 1.1453271680150989e-05, "loss": 0.3442, "step": 15282 }, { "epoch": 4.7, "learning_rate": 1.1452285712454905e-05, "loss": 0.2984, "step": 15283 }, { "epoch": 4.7, "learning_rate": 1.1451299730336285e-05, "loss": 0.3774, "step": 15284 }, { "epoch": 4.7, "learning_rate": 1.1450313733804924e-05, "loss": 0.3772, "step": 15285 }, { "epoch": 4.7, "learning_rate": 1.1449327722870609e-05, "loss": 0.459, "step": 15286 }, { "epoch": 4.7, "learning_rate": 1.1448341697543135e-05, "loss": 0.3318, "step": 15287 }, { "epoch": 4.7, "learning_rate": 1.1447355657832292e-05, "loss": 0.2166, "step": 15288 }, { "epoch": 4.7, "learning_rate": 1.1446369603747873e-05, "loss": 0.3113, "step": 15289 }, { "epoch": 4.7, "learning_rate": 1.1445383535299672e-05, "loss": 0.1851, "step": 15290 }, { "epoch": 4.7, "learning_rate": 1.1444397452497482e-05, "loss": 0.2423, "step": 15291 }, { "epoch": 4.71, "learning_rate": 1.1443411355351092e-05, "loss": 0.3008, "step": 15292 }, { "epoch": 4.71, "learning_rate": 1.1442425243870299e-05, "loss": 0.222, "step": 15293 }, { "epoch": 4.71, "learning_rate": 1.144143911806489e-05, "loss": 0.3677, "step": 15294 }, { "epoch": 4.71, "learning_rate": 1.1440452977944664e-05, "loss": 0.3309, "step": 15295 }, { "epoch": 4.71, "learning_rate": 1.1439466823519414e-05, "loss": 0.2494, "step": 15296 }, { "epoch": 4.71, "learning_rate": 1.1438480654798931e-05, "loss": 0.1886, "step": 15297 }, { "epoch": 4.71, "learning_rate": 1.1437494471793008e-05, "loss": 0.1488, "step": 15298 }, { "epoch": 4.71, "learning_rate": 1.1436508274511442e-05, "loss": 0.3692, "step": 15299 }, { "epoch": 4.71, "learning_rate": 1.1435522062964024e-05, "loss": 0.2921, "step": 15300 }, { "epoch": 4.71, "learning_rate": 1.1434535837160549e-05, "loss": 0.5219, "step": 15301 }, { "epoch": 4.71, "learning_rate": 1.1433549597110812e-05, "loss": 0.3105, "step": 15302 }, { "epoch": 4.71, "learning_rate": 1.1432563342824603e-05, "loss": 0.2762, "step": 15303 }, { "epoch": 4.71, "learning_rate": 1.1431577074311721e-05, "loss": 0.3502, "step": 15304 }, { "epoch": 4.71, "learning_rate": 1.1430590791581962e-05, "loss": 0.318, "step": 15305 }, { "epoch": 4.71, "learning_rate": 1.1429604494645117e-05, "loss": 0.2935, "step": 15306 }, { "epoch": 4.71, "learning_rate": 1.1428618183510979e-05, "loss": 0.3502, "step": 15307 }, { "epoch": 4.71, "learning_rate": 1.1427631858189347e-05, "loss": 0.2846, "step": 15308 }, { "epoch": 4.71, "learning_rate": 1.1426645518690015e-05, "loss": 0.3423, "step": 15309 }, { "epoch": 4.71, "learning_rate": 1.1425659165022782e-05, "loss": 0.302, "step": 15310 }, { "epoch": 4.71, "learning_rate": 1.1424672797197436e-05, "loss": 0.3273, "step": 15311 }, { "epoch": 4.71, "learning_rate": 1.1423686415223778e-05, "loss": 0.4469, "step": 15312 }, { "epoch": 4.71, "learning_rate": 1.1422700019111596e-05, "loss": 0.271, "step": 15313 }, { "epoch": 4.71, "learning_rate": 1.1421713608870698e-05, "loss": 0.359, "step": 15314 }, { "epoch": 4.71, "learning_rate": 1.1420727184510869e-05, "loss": 0.3568, "step": 15315 }, { "epoch": 4.71, "learning_rate": 1.1419740746041917e-05, "loss": 0.371, "step": 15316 }, { "epoch": 4.71, "learning_rate": 1.1418754293473624e-05, "loss": 0.3278, "step": 15317 }, { "epoch": 4.71, "learning_rate": 1.1417767826815793e-05, "loss": 0.2325, "step": 15318 }, { "epoch": 4.71, "learning_rate": 1.1416781346078225e-05, "loss": 0.3546, "step": 15319 }, { "epoch": 4.71, "learning_rate": 1.141579485127071e-05, "loss": 0.3399, "step": 15320 }, { "epoch": 4.71, "learning_rate": 1.1414808342403048e-05, "loss": 0.1933, "step": 15321 }, { "epoch": 4.71, "learning_rate": 1.1413821819485035e-05, "loss": 0.2455, "step": 15322 }, { "epoch": 4.71, "learning_rate": 1.1412835282526468e-05, "loss": 0.2015, "step": 15323 }, { "epoch": 4.71, "learning_rate": 1.1411848731537148e-05, "loss": 0.4283, "step": 15324 }, { "epoch": 4.72, "learning_rate": 1.1410862166526861e-05, "loss": 0.2994, "step": 15325 }, { "epoch": 4.72, "learning_rate": 1.1409875587505419e-05, "loss": 0.1624, "step": 15326 }, { "epoch": 4.72, "learning_rate": 1.1408888994482612e-05, "loss": 0.39, "step": 15327 }, { "epoch": 4.72, "learning_rate": 1.1407902387468239e-05, "loss": 0.2097, "step": 15328 }, { "epoch": 4.72, "learning_rate": 1.1406915766472098e-05, "loss": 0.2534, "step": 15329 }, { "epoch": 4.72, "learning_rate": 1.1405929131503984e-05, "loss": 0.167, "step": 15330 }, { "epoch": 4.72, "learning_rate": 1.14049424825737e-05, "loss": 0.2159, "step": 15331 }, { "epoch": 4.72, "learning_rate": 1.1403955819691042e-05, "loss": 0.2516, "step": 15332 }, { "epoch": 4.72, "learning_rate": 1.1402969142865809e-05, "loss": 0.1979, "step": 15333 }, { "epoch": 4.72, "learning_rate": 1.1401982452107797e-05, "loss": 0.371, "step": 15334 }, { "epoch": 4.72, "learning_rate": 1.140099574742681e-05, "loss": 0.2841, "step": 15335 }, { "epoch": 4.72, "learning_rate": 1.1400009028832643e-05, "loss": 0.2361, "step": 15336 }, { "epoch": 4.72, "learning_rate": 1.1399022296335099e-05, "loss": 0.249, "step": 15337 }, { "epoch": 4.72, "learning_rate": 1.1398035549943972e-05, "loss": 0.4276, "step": 15338 }, { "epoch": 4.72, "learning_rate": 1.1397048789669061e-05, "loss": 0.2107, "step": 15339 }, { "epoch": 4.72, "learning_rate": 1.1396062015520171e-05, "loss": 0.2045, "step": 15340 }, { "epoch": 4.72, "learning_rate": 1.1395075227507097e-05, "loss": 0.2503, "step": 15341 }, { "epoch": 4.72, "learning_rate": 1.1394088425639642e-05, "loss": 0.4158, "step": 15342 }, { "epoch": 4.72, "learning_rate": 1.1393101609927604e-05, "loss": 0.3653, "step": 15343 }, { "epoch": 4.72, "learning_rate": 1.1392114780380779e-05, "loss": 0.2375, "step": 15344 }, { "epoch": 4.72, "learning_rate": 1.1391127937008975e-05, "loss": 0.3135, "step": 15345 }, { "epoch": 4.72, "learning_rate": 1.1390141079821988e-05, "loss": 0.2856, "step": 15346 }, { "epoch": 4.72, "learning_rate": 1.138915420882962e-05, "loss": 0.3393, "step": 15347 }, { "epoch": 4.72, "learning_rate": 1.138816732404167e-05, "loss": 0.2303, "step": 15348 }, { "epoch": 4.72, "learning_rate": 1.1387180425467936e-05, "loss": 0.2841, "step": 15349 }, { "epoch": 4.72, "learning_rate": 1.1386193513118225e-05, "loss": 0.4906, "step": 15350 }, { "epoch": 4.72, "learning_rate": 1.138520658700233e-05, "loss": 0.1561, "step": 15351 }, { "epoch": 4.72, "learning_rate": 1.1384219647130065e-05, "loss": 0.2824, "step": 15352 }, { "epoch": 4.72, "learning_rate": 1.1383232693511216e-05, "loss": 0.3856, "step": 15353 }, { "epoch": 4.72, "learning_rate": 1.1382245726155597e-05, "loss": 0.4355, "step": 15354 }, { "epoch": 4.72, "learning_rate": 1.1381258745073e-05, "loss": 0.331, "step": 15355 }, { "epoch": 4.72, "learning_rate": 1.1380271750273234e-05, "loss": 0.3275, "step": 15356 }, { "epoch": 4.73, "learning_rate": 1.1379284741766097e-05, "loss": 0.4363, "step": 15357 }, { "epoch": 4.73, "learning_rate": 1.1378297719561388e-05, "loss": 0.4087, "step": 15358 }, { "epoch": 4.73, "learning_rate": 1.1377310683668914e-05, "loss": 0.3012, "step": 15359 }, { "epoch": 4.73, "learning_rate": 1.1376323634098476e-05, "loss": 0.2364, "step": 15360 }, { "epoch": 4.73, "learning_rate": 1.1375336570859877e-05, "loss": 0.2139, "step": 15361 }, { "epoch": 4.73, "learning_rate": 1.1374349493962918e-05, "loss": 0.2655, "step": 15362 }, { "epoch": 4.73, "learning_rate": 1.1373362403417398e-05, "loss": 0.2554, "step": 15363 }, { "epoch": 4.73, "learning_rate": 1.137237529923313e-05, "loss": 0.1886, "step": 15364 }, { "epoch": 4.73, "learning_rate": 1.1371388181419905e-05, "loss": 0.2376, "step": 15365 }, { "epoch": 4.73, "learning_rate": 1.1370401049987534e-05, "loss": 0.339, "step": 15366 }, { "epoch": 4.73, "learning_rate": 1.1369413904945819e-05, "loss": 0.265, "step": 15367 }, { "epoch": 4.73, "learning_rate": 1.1368426746304557e-05, "loss": 0.3385, "step": 15368 }, { "epoch": 4.73, "learning_rate": 1.136743957407356e-05, "loss": 0.3314, "step": 15369 }, { "epoch": 4.73, "learning_rate": 1.1366452388262626e-05, "loss": 0.3629, "step": 15370 }, { "epoch": 4.73, "learning_rate": 1.1365465188881562e-05, "loss": 0.3664, "step": 15371 }, { "epoch": 4.73, "learning_rate": 1.1364477975940169e-05, "loss": 0.5171, "step": 15372 }, { "epoch": 4.73, "learning_rate": 1.1363490749448253e-05, "loss": 0.4037, "step": 15373 }, { "epoch": 4.73, "learning_rate": 1.136250350941562e-05, "loss": 0.2237, "step": 15374 }, { "epoch": 4.73, "learning_rate": 1.1361516255852065e-05, "loss": 0.3892, "step": 15375 }, { "epoch": 4.73, "learning_rate": 1.1360528988767403e-05, "loss": 0.2959, "step": 15376 }, { "epoch": 4.73, "learning_rate": 1.1359541708171434e-05, "loss": 0.3342, "step": 15377 }, { "epoch": 4.73, "learning_rate": 1.1358554414073963e-05, "loss": 0.2307, "step": 15378 }, { "epoch": 4.73, "learning_rate": 1.1357567106484795e-05, "loss": 0.2351, "step": 15379 }, { "epoch": 4.73, "learning_rate": 1.1356579785413731e-05, "loss": 0.4155, "step": 15380 }, { "epoch": 4.73, "learning_rate": 1.1355592450870581e-05, "loss": 0.2923, "step": 15381 }, { "epoch": 4.73, "learning_rate": 1.135460510286515e-05, "loss": 0.3975, "step": 15382 }, { "epoch": 4.73, "learning_rate": 1.1353617741407241e-05, "loss": 0.4194, "step": 15383 }, { "epoch": 4.73, "learning_rate": 1.1352630366506661e-05, "loss": 0.591, "step": 15384 }, { "epoch": 4.73, "learning_rate": 1.1351642978173214e-05, "loss": 0.3649, "step": 15385 }, { "epoch": 4.73, "learning_rate": 1.1350655576416707e-05, "loss": 0.4127, "step": 15386 }, { "epoch": 4.73, "learning_rate": 1.1349668161246945e-05, "loss": 0.1993, "step": 15387 }, { "epoch": 4.73, "learning_rate": 1.1348680732673735e-05, "loss": 0.2191, "step": 15388 }, { "epoch": 4.73, "learning_rate": 1.134769329070688e-05, "loss": 0.2378, "step": 15389 }, { "epoch": 4.74, "learning_rate": 1.1346705835356191e-05, "loss": 0.3059, "step": 15390 }, { "epoch": 4.74, "learning_rate": 1.1345718366631474e-05, "loss": 0.4758, "step": 15391 }, { "epoch": 4.74, "learning_rate": 1.134473088454253e-05, "loss": 0.4291, "step": 15392 }, { "epoch": 4.74, "learning_rate": 1.1343743389099169e-05, "loss": 0.3414, "step": 15393 }, { "epoch": 4.74, "learning_rate": 1.1342755880311197e-05, "loss": 0.3622, "step": 15394 }, { "epoch": 4.74, "learning_rate": 1.1341768358188424e-05, "loss": 0.2589, "step": 15395 }, { "epoch": 4.74, "learning_rate": 1.1340780822740654e-05, "loss": 0.3595, "step": 15396 }, { "epoch": 4.74, "learning_rate": 1.1339793273977694e-05, "loss": 0.2498, "step": 15397 }, { "epoch": 4.74, "learning_rate": 1.1338805711909353e-05, "loss": 0.1463, "step": 15398 }, { "epoch": 4.74, "learning_rate": 1.1337818136545436e-05, "loss": 0.4013, "step": 15399 }, { "epoch": 4.74, "learning_rate": 1.1336830547895752e-05, "loss": 0.2834, "step": 15400 }, { "epoch": 4.74, "learning_rate": 1.1335842945970108e-05, "loss": 0.4183, "step": 15401 }, { "epoch": 4.74, "learning_rate": 1.1334855330778317e-05, "loss": 0.2743, "step": 15402 }, { "epoch": 4.74, "learning_rate": 1.1333867702330177e-05, "loss": 0.316, "step": 15403 }, { "epoch": 4.74, "learning_rate": 1.1332880060635502e-05, "loss": 0.3385, "step": 15404 }, { "epoch": 4.74, "learning_rate": 1.1331892405704103e-05, "loss": 0.2641, "step": 15405 }, { "epoch": 4.74, "learning_rate": 1.133090473754578e-05, "loss": 0.435, "step": 15406 }, { "epoch": 4.74, "learning_rate": 1.1329917056170352e-05, "loss": 0.291, "step": 15407 }, { "epoch": 4.74, "learning_rate": 1.1328929361587617e-05, "loss": 0.5472, "step": 15408 }, { "epoch": 4.74, "learning_rate": 1.1327941653807392e-05, "loss": 0.2504, "step": 15409 }, { "epoch": 4.74, "learning_rate": 1.1326953932839482e-05, "loss": 0.3286, "step": 15410 }, { "epoch": 4.74, "learning_rate": 1.1325966198693694e-05, "loss": 0.3774, "step": 15411 }, { "epoch": 4.74, "learning_rate": 1.1324978451379845e-05, "loss": 0.3094, "step": 15412 }, { "epoch": 4.74, "learning_rate": 1.1323990690907734e-05, "loss": 0.2541, "step": 15413 }, { "epoch": 4.74, "learning_rate": 1.1323002917287177e-05, "loss": 0.3489, "step": 15414 }, { "epoch": 4.74, "learning_rate": 1.1322015130527981e-05, "loss": 0.2877, "step": 15415 }, { "epoch": 4.74, "learning_rate": 1.1321027330639952e-05, "loss": 0.2956, "step": 15416 }, { "epoch": 4.74, "learning_rate": 1.132003951763291e-05, "loss": 0.3083, "step": 15417 }, { "epoch": 4.74, "learning_rate": 1.1319051691516659e-05, "loss": 0.3292, "step": 15418 }, { "epoch": 4.74, "learning_rate": 1.1318063852301009e-05, "loss": 0.3094, "step": 15419 }, { "epoch": 4.74, "learning_rate": 1.1317075999995766e-05, "loss": 0.2957, "step": 15420 }, { "epoch": 4.74, "learning_rate": 1.1316088134610748e-05, "loss": 0.3637, "step": 15421 }, { "epoch": 4.75, "learning_rate": 1.1315100256155763e-05, "loss": 0.2689, "step": 15422 }, { "epoch": 4.75, "learning_rate": 1.1314112364640616e-05, "loss": 0.3206, "step": 15423 }, { "epoch": 4.75, "learning_rate": 1.1313124460075126e-05, "loss": 0.2603, "step": 15424 }, { "epoch": 4.75, "learning_rate": 1.1312136542469098e-05, "loss": 0.2705, "step": 15425 }, { "epoch": 4.75, "learning_rate": 1.1311148611832346e-05, "loss": 0.3021, "step": 15426 }, { "epoch": 4.75, "learning_rate": 1.1310160668174679e-05, "loss": 0.4197, "step": 15427 }, { "epoch": 4.75, "learning_rate": 1.130917271150591e-05, "loss": 0.174, "step": 15428 }, { "epoch": 4.75, "learning_rate": 1.130818474183585e-05, "loss": 0.3103, "step": 15429 }, { "epoch": 4.75, "learning_rate": 1.1307196759174306e-05, "loss": 0.3013, "step": 15430 }, { "epoch": 4.75, "learning_rate": 1.1306208763531096e-05, "loss": 0.3455, "step": 15431 }, { "epoch": 4.75, "learning_rate": 1.1305220754916028e-05, "loss": 0.3414, "step": 15432 }, { "epoch": 4.75, "learning_rate": 1.130423273333892e-05, "loss": 0.3079, "step": 15433 }, { "epoch": 4.75, "learning_rate": 1.1303244698809572e-05, "loss": 0.3731, "step": 15434 }, { "epoch": 4.75, "learning_rate": 1.130225665133781e-05, "loss": 0.3028, "step": 15435 }, { "epoch": 4.75, "learning_rate": 1.1301268590933434e-05, "loss": 0.4066, "step": 15436 }, { "epoch": 4.75, "learning_rate": 1.1300280517606262e-05, "loss": 0.4359, "step": 15437 }, { "epoch": 4.75, "learning_rate": 1.129929243136611e-05, "loss": 0.4084, "step": 15438 }, { "epoch": 4.75, "learning_rate": 1.129830433222278e-05, "loss": 0.1926, "step": 15439 }, { "epoch": 4.75, "learning_rate": 1.1297316220186096e-05, "loss": 0.2914, "step": 15440 }, { "epoch": 4.75, "learning_rate": 1.1296328095265867e-05, "loss": 0.4618, "step": 15441 }, { "epoch": 4.75, "learning_rate": 1.1295339957471903e-05, "loss": 0.3962, "step": 15442 }, { "epoch": 4.75, "learning_rate": 1.1294351806814021e-05, "loss": 0.3362, "step": 15443 }, { "epoch": 4.75, "learning_rate": 1.1293363643302031e-05, "loss": 0.3264, "step": 15444 }, { "epoch": 4.75, "learning_rate": 1.1292375466945749e-05, "loss": 0.2921, "step": 15445 }, { "epoch": 4.75, "learning_rate": 1.1291387277754988e-05, "loss": 0.2823, "step": 15446 }, { "epoch": 4.75, "learning_rate": 1.129039907573956e-05, "loss": 0.2663, "step": 15447 }, { "epoch": 4.75, "learning_rate": 1.1289410860909283e-05, "loss": 0.3062, "step": 15448 }, { "epoch": 4.75, "learning_rate": 1.1288422633273963e-05, "loss": 0.3266, "step": 15449 }, { "epoch": 4.75, "learning_rate": 1.1287434392843421e-05, "loss": 0.3669, "step": 15450 }, { "epoch": 4.75, "learning_rate": 1.1286446139627467e-05, "loss": 0.3065, "step": 15451 }, { "epoch": 4.75, "learning_rate": 1.128545787363592e-05, "loss": 0.2531, "step": 15452 }, { "epoch": 4.75, "learning_rate": 1.1284469594878593e-05, "loss": 0.2784, "step": 15453 }, { "epoch": 4.75, "learning_rate": 1.1283481303365295e-05, "loss": 0.3054, "step": 15454 }, { "epoch": 4.76, "learning_rate": 1.128249299910585e-05, "loss": 0.3152, "step": 15455 }, { "epoch": 4.76, "learning_rate": 1.128150468211006e-05, "loss": 0.3679, "step": 15456 }, { "epoch": 4.76, "learning_rate": 1.1280516352387753e-05, "loss": 0.299, "step": 15457 }, { "epoch": 4.76, "learning_rate": 1.127952800994874e-05, "loss": 0.3695, "step": 15458 }, { "epoch": 4.76, "learning_rate": 1.1278539654802832e-05, "loss": 0.4918, "step": 15459 }, { "epoch": 4.76, "learning_rate": 1.1277551286959847e-05, "loss": 0.3899, "step": 15460 }, { "epoch": 4.76, "learning_rate": 1.1276562906429599e-05, "loss": 0.4625, "step": 15461 }, { "epoch": 4.76, "learning_rate": 1.1275574513221906e-05, "loss": 0.5906, "step": 15462 }, { "epoch": 4.76, "learning_rate": 1.1274586107346581e-05, "loss": 0.2715, "step": 15463 }, { "epoch": 4.76, "learning_rate": 1.1273597688813442e-05, "loss": 0.3016, "step": 15464 }, { "epoch": 4.76, "learning_rate": 1.1272609257632305e-05, "loss": 0.3731, "step": 15465 }, { "epoch": 4.76, "learning_rate": 1.1271620813812985e-05, "loss": 0.3396, "step": 15466 }, { "epoch": 4.76, "learning_rate": 1.1270632357365299e-05, "loss": 0.5244, "step": 15467 }, { "epoch": 4.76, "learning_rate": 1.1269643888299062e-05, "loss": 0.2746, "step": 15468 }, { "epoch": 4.76, "learning_rate": 1.1268655406624092e-05, "loss": 0.414, "step": 15469 }, { "epoch": 4.76, "learning_rate": 1.1267666912350201e-05, "loss": 0.4584, "step": 15470 }, { "epoch": 4.76, "learning_rate": 1.1266678405487213e-05, "loss": 0.4303, "step": 15471 }, { "epoch": 4.76, "learning_rate": 1.126568988604494e-05, "loss": 0.3984, "step": 15472 }, { "epoch": 4.76, "learning_rate": 1.12647013540332e-05, "loss": 0.5045, "step": 15473 }, { "epoch": 4.76, "learning_rate": 1.1263712809461811e-05, "loss": 0.2555, "step": 15474 }, { "epoch": 4.76, "learning_rate": 1.1262724252340585e-05, "loss": 0.3533, "step": 15475 }, { "epoch": 4.76, "learning_rate": 1.1261735682679346e-05, "loss": 0.2871, "step": 15476 }, { "epoch": 4.76, "learning_rate": 1.126074710048791e-05, "loss": 0.2552, "step": 15477 }, { "epoch": 4.76, "learning_rate": 1.1259758505776092e-05, "loss": 0.3625, "step": 15478 }, { "epoch": 4.76, "learning_rate": 1.1258769898553714e-05, "loss": 0.4747, "step": 15479 }, { "epoch": 4.76, "learning_rate": 1.1257781278830587e-05, "loss": 0.3511, "step": 15480 }, { "epoch": 4.76, "learning_rate": 1.1256792646616534e-05, "loss": 0.2541, "step": 15481 }, { "epoch": 4.76, "learning_rate": 1.1255804001921368e-05, "loss": 0.2732, "step": 15482 }, { "epoch": 4.76, "learning_rate": 1.1254815344754919e-05, "loss": 0.4065, "step": 15483 }, { "epoch": 4.76, "learning_rate": 1.125382667512699e-05, "loss": 0.3278, "step": 15484 }, { "epoch": 4.76, "learning_rate": 1.125283799304741e-05, "loss": 0.4077, "step": 15485 }, { "epoch": 4.76, "learning_rate": 1.1251849298525993e-05, "loss": 0.2442, "step": 15486 }, { "epoch": 4.77, "learning_rate": 1.1250860591572559e-05, "loss": 0.4087, "step": 15487 }, { "epoch": 4.77, "learning_rate": 1.1249871872196928e-05, "loss": 0.239, "step": 15488 }, { "epoch": 4.77, "learning_rate": 1.1248883140408915e-05, "loss": 0.3903, "step": 15489 }, { "epoch": 4.77, "learning_rate": 1.1247894396218341e-05, "loss": 0.4766, "step": 15490 }, { "epoch": 4.77, "learning_rate": 1.1246905639635029e-05, "loss": 0.2305, "step": 15491 }, { "epoch": 4.77, "learning_rate": 1.124591687066879e-05, "loss": 0.3437, "step": 15492 }, { "epoch": 4.77, "learning_rate": 1.1244928089329455e-05, "loss": 0.3296, "step": 15493 }, { "epoch": 4.77, "learning_rate": 1.124393929562683e-05, "loss": 0.3466, "step": 15494 }, { "epoch": 4.77, "learning_rate": 1.1242950489570746e-05, "loss": 0.3538, "step": 15495 }, { "epoch": 4.77, "learning_rate": 1.1241961671171015e-05, "loss": 0.3765, "step": 15496 }, { "epoch": 4.77, "learning_rate": 1.124097284043746e-05, "loss": 0.1819, "step": 15497 }, { "epoch": 4.77, "learning_rate": 1.1239983997379904e-05, "loss": 0.3726, "step": 15498 }, { "epoch": 4.77, "learning_rate": 1.1238995142008162e-05, "loss": 0.2403, "step": 15499 }, { "epoch": 4.77, "learning_rate": 1.1238006274332057e-05, "loss": 0.2685, "step": 15500 }, { "epoch": 4.77, "learning_rate": 1.1237017394361404e-05, "loss": 0.203, "step": 15501 }, { "epoch": 4.77, "learning_rate": 1.1236028502106036e-05, "loss": 0.2278, "step": 15502 }, { "epoch": 4.77, "learning_rate": 1.1235039597575761e-05, "loss": 0.3095, "step": 15503 }, { "epoch": 4.77, "learning_rate": 1.1234050680780407e-05, "loss": 0.6067, "step": 15504 }, { "epoch": 4.77, "learning_rate": 1.1233061751729792e-05, "loss": 0.3636, "step": 15505 }, { "epoch": 4.77, "learning_rate": 1.1232072810433733e-05, "loss": 0.3116, "step": 15506 }, { "epoch": 4.77, "learning_rate": 1.123108385690206e-05, "loss": 0.2967, "step": 15507 }, { "epoch": 4.77, "learning_rate": 1.123009489114459e-05, "loss": 0.3332, "step": 15508 }, { "epoch": 4.77, "learning_rate": 1.1229105913171143e-05, "loss": 0.3116, "step": 15509 }, { "epoch": 4.77, "learning_rate": 1.1228116922991544e-05, "loss": 0.3001, "step": 15510 }, { "epoch": 4.77, "learning_rate": 1.1227127920615606e-05, "loss": 0.355, "step": 15511 }, { "epoch": 4.77, "learning_rate": 1.1226138906053161e-05, "loss": 0.4747, "step": 15512 }, { "epoch": 4.77, "learning_rate": 1.1225149879314026e-05, "loss": 0.2869, "step": 15513 }, { "epoch": 4.77, "learning_rate": 1.1224160840408025e-05, "loss": 0.3263, "step": 15514 }, { "epoch": 4.77, "learning_rate": 1.1223171789344976e-05, "loss": 0.2441, "step": 15515 }, { "epoch": 4.77, "learning_rate": 1.1222182726134703e-05, "loss": 0.2913, "step": 15516 }, { "epoch": 4.77, "learning_rate": 1.1221193650787032e-05, "loss": 0.2279, "step": 15517 }, { "epoch": 4.77, "learning_rate": 1.1220204563311781e-05, "loss": 0.1812, "step": 15518 }, { "epoch": 4.77, "learning_rate": 1.1219215463718775e-05, "loss": 0.2566, "step": 15519 }, { "epoch": 4.78, "learning_rate": 1.1218226352017836e-05, "loss": 0.35, "step": 15520 }, { "epoch": 4.78, "learning_rate": 1.1217237228218785e-05, "loss": 0.2522, "step": 15521 }, { "epoch": 4.78, "learning_rate": 1.1216248092331447e-05, "loss": 0.3361, "step": 15522 }, { "epoch": 4.78, "learning_rate": 1.1215258944365646e-05, "loss": 0.3991, "step": 15523 }, { "epoch": 4.78, "learning_rate": 1.1214269784331203e-05, "loss": 0.2974, "step": 15524 }, { "epoch": 4.78, "learning_rate": 1.121328061223794e-05, "loss": 0.2897, "step": 15525 }, { "epoch": 4.78, "learning_rate": 1.1212291428095686e-05, "loss": 0.2547, "step": 15526 }, { "epoch": 4.78, "learning_rate": 1.1211302231914258e-05, "loss": 0.2775, "step": 15527 }, { "epoch": 4.78, "learning_rate": 1.1210313023703482e-05, "loss": 0.4842, "step": 15528 }, { "epoch": 4.78, "learning_rate": 1.1209323803473186e-05, "loss": 0.3992, "step": 15529 }, { "epoch": 4.78, "learning_rate": 1.1208334571233186e-05, "loss": 0.2342, "step": 15530 }, { "epoch": 4.78, "learning_rate": 1.1207345326993314e-05, "loss": 0.3116, "step": 15531 }, { "epoch": 4.78, "learning_rate": 1.1206356070763385e-05, "loss": 0.2145, "step": 15532 }, { "epoch": 4.78, "learning_rate": 1.1205366802553231e-05, "loss": 0.2132, "step": 15533 }, { "epoch": 4.78, "learning_rate": 1.1204377522372675e-05, "loss": 0.3728, "step": 15534 }, { "epoch": 4.78, "learning_rate": 1.120338823023154e-05, "loss": 0.1654, "step": 15535 }, { "epoch": 4.78, "learning_rate": 1.120239892613965e-05, "loss": 0.3256, "step": 15536 }, { "epoch": 4.78, "learning_rate": 1.1201409610106828e-05, "loss": 0.3324, "step": 15537 }, { "epoch": 4.78, "learning_rate": 1.1200420282142906e-05, "loss": 0.3809, "step": 15538 }, { "epoch": 4.78, "learning_rate": 1.1199430942257701e-05, "loss": 0.2244, "step": 15539 }, { "epoch": 4.78, "learning_rate": 1.1198441590461041e-05, "loss": 0.4288, "step": 15540 }, { "epoch": 4.78, "learning_rate": 1.1197452226762756e-05, "loss": 0.34, "step": 15541 }, { "epoch": 4.78, "learning_rate": 1.119646285117266e-05, "loss": 0.307, "step": 15542 }, { "epoch": 4.78, "learning_rate": 1.119547346370059e-05, "loss": 0.2866, "step": 15543 }, { "epoch": 4.78, "learning_rate": 1.1194484064356366e-05, "loss": 0.5015, "step": 15544 }, { "epoch": 4.78, "learning_rate": 1.1193494653149815e-05, "loss": 0.1623, "step": 15545 }, { "epoch": 4.78, "learning_rate": 1.1192505230090763e-05, "loss": 0.2517, "step": 15546 }, { "epoch": 4.78, "learning_rate": 1.1191515795189029e-05, "loss": 0.2917, "step": 15547 }, { "epoch": 4.78, "learning_rate": 1.1190526348454452e-05, "loss": 0.3596, "step": 15548 }, { "epoch": 4.78, "learning_rate": 1.1189536889896848e-05, "loss": 0.4686, "step": 15549 }, { "epoch": 4.78, "learning_rate": 1.1188547419526045e-05, "loss": 0.3038, "step": 15550 }, { "epoch": 4.78, "learning_rate": 1.1187557937351872e-05, "loss": 0.2341, "step": 15551 }, { "epoch": 4.79, "learning_rate": 1.1186568443384156e-05, "loss": 0.4157, "step": 15552 }, { "epoch": 4.79, "learning_rate": 1.118557893763272e-05, "loss": 0.338, "step": 15553 }, { "epoch": 4.79, "learning_rate": 1.1184589420107393e-05, "loss": 0.4803, "step": 15554 }, { "epoch": 4.79, "learning_rate": 1.1183599890818003e-05, "loss": 0.2733, "step": 15555 }, { "epoch": 4.79, "learning_rate": 1.118261034977437e-05, "loss": 0.2388, "step": 15556 }, { "epoch": 4.79, "learning_rate": 1.118162079698633e-05, "loss": 0.3871, "step": 15557 }, { "epoch": 4.79, "learning_rate": 1.1180631232463707e-05, "loss": 0.362, "step": 15558 }, { "epoch": 4.79, "learning_rate": 1.1179641656216327e-05, "loss": 0.503, "step": 15559 }, { "epoch": 4.79, "learning_rate": 1.1178652068254019e-05, "loss": 0.1818, "step": 15560 }, { "epoch": 4.79, "learning_rate": 1.1177662468586608e-05, "loss": 0.3394, "step": 15561 }, { "epoch": 4.79, "learning_rate": 1.1176672857223925e-05, "loss": 0.3501, "step": 15562 }, { "epoch": 4.79, "learning_rate": 1.1175683234175794e-05, "loss": 0.2713, "step": 15563 }, { "epoch": 4.79, "learning_rate": 1.1174693599452046e-05, "loss": 0.4561, "step": 15564 }, { "epoch": 4.79, "learning_rate": 1.117370395306251e-05, "loss": 0.4387, "step": 15565 }, { "epoch": 4.79, "learning_rate": 1.1172714295017009e-05, "loss": 0.1337, "step": 15566 }, { "epoch": 4.79, "learning_rate": 1.1171724625325376e-05, "loss": 0.2606, "step": 15567 }, { "epoch": 4.79, "learning_rate": 1.1170734943997435e-05, "loss": 0.2042, "step": 15568 }, { "epoch": 4.79, "learning_rate": 1.116974525104302e-05, "loss": 0.3279, "step": 15569 }, { "epoch": 4.79, "learning_rate": 1.1168755546471955e-05, "loss": 0.4829, "step": 15570 }, { "epoch": 4.79, "learning_rate": 1.1167765830294071e-05, "loss": 0.3455, "step": 15571 }, { "epoch": 4.79, "learning_rate": 1.1166776102519197e-05, "loss": 0.3331, "step": 15572 }, { "epoch": 4.79, "learning_rate": 1.1165786363157156e-05, "loss": 0.3163, "step": 15573 }, { "epoch": 4.79, "learning_rate": 1.116479661221779e-05, "loss": 0.2959, "step": 15574 }, { "epoch": 4.79, "learning_rate": 1.1163806849710914e-05, "loss": 0.3624, "step": 15575 }, { "epoch": 4.79, "learning_rate": 1.1162817075646367e-05, "loss": 0.3324, "step": 15576 }, { "epoch": 4.79, "learning_rate": 1.1161827290033974e-05, "loss": 0.3193, "step": 15577 }, { "epoch": 4.79, "learning_rate": 1.116083749288356e-05, "loss": 0.1969, "step": 15578 }, { "epoch": 4.79, "learning_rate": 1.1159847684204968e-05, "loss": 0.2325, "step": 15579 }, { "epoch": 4.79, "learning_rate": 1.1158857864008013e-05, "loss": 0.4755, "step": 15580 }, { "epoch": 4.79, "learning_rate": 1.1157868032302534e-05, "loss": 0.4396, "step": 15581 }, { "epoch": 4.79, "learning_rate": 1.1156878189098357e-05, "loss": 0.1955, "step": 15582 }, { "epoch": 4.79, "learning_rate": 1.1155888334405316e-05, "loss": 0.4647, "step": 15583 }, { "epoch": 4.79, "learning_rate": 1.115489846823324e-05, "loss": 0.3441, "step": 15584 }, { "epoch": 4.8, "learning_rate": 1.1153908590591952e-05, "loss": 0.3361, "step": 15585 }, { "epoch": 4.8, "learning_rate": 1.1152918701491293e-05, "loss": 0.3145, "step": 15586 }, { "epoch": 4.8, "learning_rate": 1.1151928800941086e-05, "loss": 0.3402, "step": 15587 }, { "epoch": 4.8, "learning_rate": 1.1150938888951165e-05, "loss": 0.3002, "step": 15588 }, { "epoch": 4.8, "learning_rate": 1.114994896553136e-05, "loss": 0.1962, "step": 15589 }, { "epoch": 4.8, "learning_rate": 1.1148959030691504e-05, "loss": 0.3833, "step": 15590 }, { "epoch": 4.8, "learning_rate": 1.1147969084441425e-05, "loss": 0.404, "step": 15591 }, { "epoch": 4.8, "learning_rate": 1.1146979126790953e-05, "loss": 0.2843, "step": 15592 }, { "epoch": 4.8, "learning_rate": 1.1145989157749923e-05, "loss": 0.2726, "step": 15593 }, { "epoch": 4.8, "learning_rate": 1.1144999177328163e-05, "loss": 0.4299, "step": 15594 }, { "epoch": 4.8, "learning_rate": 1.114400918553551e-05, "loss": 0.2702, "step": 15595 }, { "epoch": 4.8, "learning_rate": 1.1143019182381788e-05, "loss": 0.2031, "step": 15596 }, { "epoch": 4.8, "learning_rate": 1.1142029167876831e-05, "loss": 0.2081, "step": 15597 }, { "epoch": 4.8, "learning_rate": 1.1141039142030472e-05, "loss": 0.2763, "step": 15598 }, { "epoch": 4.8, "learning_rate": 1.1140049104852546e-05, "loss": 0.3221, "step": 15599 }, { "epoch": 4.8, "learning_rate": 1.113905905635288e-05, "loss": 0.3344, "step": 15600 }, { "epoch": 4.8, "learning_rate": 1.1138068996541305e-05, "loss": 0.2518, "step": 15601 }, { "epoch": 4.8, "learning_rate": 1.113707892542766e-05, "loss": 0.2833, "step": 15602 }, { "epoch": 4.8, "learning_rate": 1.113608884302177e-05, "loss": 0.2509, "step": 15603 }, { "epoch": 4.8, "learning_rate": 1.1135098749333474e-05, "loss": 0.247, "step": 15604 }, { "epoch": 4.8, "learning_rate": 1.11341086443726e-05, "loss": 0.2345, "step": 15605 }, { "epoch": 4.8, "learning_rate": 1.1133118528148981e-05, "loss": 0.2849, "step": 15606 }, { "epoch": 4.8, "learning_rate": 1.1132128400672452e-05, "loss": 0.3346, "step": 15607 }, { "epoch": 4.8, "learning_rate": 1.1131138261952845e-05, "loss": 0.3689, "step": 15608 }, { "epoch": 4.8, "learning_rate": 1.1130148111999992e-05, "loss": 0.4277, "step": 15609 }, { "epoch": 4.8, "learning_rate": 1.1129157950823725e-05, "loss": 0.2294, "step": 15610 }, { "epoch": 4.8, "learning_rate": 1.1128167778433881e-05, "loss": 0.2583, "step": 15611 }, { "epoch": 4.8, "learning_rate": 1.1127177594840291e-05, "loss": 0.2509, "step": 15612 }, { "epoch": 4.8, "learning_rate": 1.1126187400052788e-05, "loss": 0.3221, "step": 15613 }, { "epoch": 4.8, "learning_rate": 1.1125197194081203e-05, "loss": 0.2735, "step": 15614 }, { "epoch": 4.8, "learning_rate": 1.112420697693538e-05, "loss": 0.3432, "step": 15615 }, { "epoch": 4.8, "learning_rate": 1.1123216748625138e-05, "loss": 0.303, "step": 15616 }, { "epoch": 4.81, "learning_rate": 1.1122226509160323e-05, "loss": 0.2969, "step": 15617 }, { "epoch": 4.81, "learning_rate": 1.1122226509160323e-05, "loss": 0.404, "step": 15618 }, { "epoch": 4.81, "learning_rate": 1.1121236258550762e-05, "loss": 0.3248, "step": 15619 }, { "epoch": 4.81, "learning_rate": 1.1120245996806294e-05, "loss": 0.2651, "step": 15620 }, { "epoch": 4.81, "learning_rate": 1.111925572393675e-05, "loss": 0.3215, "step": 15621 }, { "epoch": 4.81, "learning_rate": 1.1118265439951968e-05, "loss": 0.278, "step": 15622 }, { "epoch": 4.81, "learning_rate": 1.1117275144861776e-05, "loss": 0.2005, "step": 15623 }, { "epoch": 4.81, "learning_rate": 1.111628483867601e-05, "loss": 0.1504, "step": 15624 }, { "epoch": 4.81, "learning_rate": 1.111529452140451e-05, "loss": 0.2901, "step": 15625 }, { "epoch": 4.81, "learning_rate": 1.1114304193057109e-05, "loss": 0.3137, "step": 15626 }, { "epoch": 4.81, "learning_rate": 1.111331385364364e-05, "loss": 0.3592, "step": 15627 }, { "epoch": 4.81, "learning_rate": 1.1112323503173937e-05, "loss": 0.2557, "step": 15628 }, { "epoch": 4.81, "learning_rate": 1.1111333141657834e-05, "loss": 0.1945, "step": 15629 }, { "epoch": 4.81, "learning_rate": 1.1110342769105172e-05, "loss": 0.277, "step": 15630 }, { "epoch": 4.81, "learning_rate": 1.1109352385525782e-05, "loss": 0.3801, "step": 15631 }, { "epoch": 4.81, "learning_rate": 1.1108361990929502e-05, "loss": 0.4246, "step": 15632 }, { "epoch": 4.81, "learning_rate": 1.110737158532616e-05, "loss": 0.2491, "step": 15633 }, { "epoch": 4.81, "learning_rate": 1.1106381168725604e-05, "loss": 0.2528, "step": 15634 }, { "epoch": 4.81, "learning_rate": 1.110539074113766e-05, "loss": 0.2967, "step": 15635 }, { "epoch": 4.81, "learning_rate": 1.110440030257217e-05, "loss": 0.4451, "step": 15636 }, { "epoch": 4.81, "learning_rate": 1.1103409853038967e-05, "loss": 0.4106, "step": 15637 }, { "epoch": 4.81, "learning_rate": 1.1102419392547883e-05, "loss": 0.2485, "step": 15638 }, { "epoch": 4.81, "learning_rate": 1.1101428921108762e-05, "loss": 0.3118, "step": 15639 }, { "epoch": 4.81, "learning_rate": 1.1100438438731435e-05, "loss": 0.3313, "step": 15640 }, { "epoch": 4.81, "learning_rate": 1.1099447945425744e-05, "loss": 0.3965, "step": 15641 }, { "epoch": 4.81, "learning_rate": 1.1098457441201518e-05, "loss": 0.5282, "step": 15642 }, { "epoch": 4.81, "learning_rate": 1.1097466926068596e-05, "loss": 0.3308, "step": 15643 }, { "epoch": 4.81, "learning_rate": 1.1096476400036817e-05, "loss": 0.2561, "step": 15644 }, { "epoch": 4.81, "learning_rate": 1.109548586311602e-05, "loss": 0.2669, "step": 15645 }, { "epoch": 4.81, "learning_rate": 1.1094495315316036e-05, "loss": 0.4192, "step": 15646 }, { "epoch": 4.81, "learning_rate": 1.1093504756646708e-05, "loss": 0.2391, "step": 15647 }, { "epoch": 4.81, "learning_rate": 1.1092514187117865e-05, "loss": 0.1941, "step": 15648 }, { "epoch": 4.81, "learning_rate": 1.109152360673935e-05, "loss": 0.2714, "step": 15649 }, { "epoch": 4.82, "learning_rate": 1.1090533015521e-05, "loss": 0.2554, "step": 15650 }, { "epoch": 4.82, "learning_rate": 1.1089542413472656e-05, "loss": 0.3188, "step": 15651 }, { "epoch": 4.82, "learning_rate": 1.1088551800604146e-05, "loss": 0.3637, "step": 15652 }, { "epoch": 4.82, "learning_rate": 1.1087561176925317e-05, "loss": 0.3788, "step": 15653 }, { "epoch": 4.82, "learning_rate": 1.1086570542446003e-05, "loss": 0.2796, "step": 15654 }, { "epoch": 4.82, "learning_rate": 1.108557989717604e-05, "loss": 0.327, "step": 15655 }, { "epoch": 4.82, "learning_rate": 1.1084589241125273e-05, "loss": 0.3489, "step": 15656 }, { "epoch": 4.82, "learning_rate": 1.1083598574303529e-05, "loss": 0.3606, "step": 15657 }, { "epoch": 4.82, "learning_rate": 1.1082607896720656e-05, "loss": 0.2331, "step": 15658 }, { "epoch": 4.82, "learning_rate": 1.1081617208386489e-05, "loss": 0.2864, "step": 15659 }, { "epoch": 4.82, "learning_rate": 1.1080626509310863e-05, "loss": 0.2909, "step": 15660 }, { "epoch": 4.82, "learning_rate": 1.1079635799503625e-05, "loss": 0.4171, "step": 15661 }, { "epoch": 4.82, "learning_rate": 1.1078645078974603e-05, "loss": 0.2503, "step": 15662 }, { "epoch": 4.82, "learning_rate": 1.1077654347733647e-05, "loss": 0.458, "step": 15663 }, { "epoch": 4.82, "learning_rate": 1.1076663605790585e-05, "loss": 0.5043, "step": 15664 }, { "epoch": 4.82, "learning_rate": 1.1075672853155262e-05, "loss": 0.2295, "step": 15665 }, { "epoch": 4.82, "learning_rate": 1.107468208983752e-05, "loss": 0.3681, "step": 15666 }, { "epoch": 4.82, "learning_rate": 1.107369131584719e-05, "loss": 0.3227, "step": 15667 }, { "epoch": 4.82, "learning_rate": 1.1072700531194119e-05, "loss": 0.2452, "step": 15668 }, { "epoch": 4.82, "learning_rate": 1.1071709735888139e-05, "loss": 0.3054, "step": 15669 }, { "epoch": 4.82, "learning_rate": 1.1070718929939097e-05, "loss": 0.5345, "step": 15670 }, { "epoch": 4.82, "learning_rate": 1.106972811335683e-05, "loss": 0.2686, "step": 15671 }, { "epoch": 4.82, "learning_rate": 1.1068737286151175e-05, "loss": 0.2259, "step": 15672 }, { "epoch": 4.82, "learning_rate": 1.1067746448331976e-05, "loss": 0.2103, "step": 15673 }, { "epoch": 4.82, "learning_rate": 1.1066755599909065e-05, "loss": 0.2538, "step": 15674 }, { "epoch": 4.82, "learning_rate": 1.1065764740892294e-05, "loss": 0.3371, "step": 15675 }, { "epoch": 4.82, "learning_rate": 1.1064773871291494e-05, "loss": 0.316, "step": 15676 }, { "epoch": 4.82, "learning_rate": 1.1063782991116509e-05, "loss": 0.2197, "step": 15677 }, { "epoch": 4.82, "learning_rate": 1.1062792100377177e-05, "loss": 0.3652, "step": 15678 }, { "epoch": 4.82, "learning_rate": 1.106180119908334e-05, "loss": 0.2679, "step": 15679 }, { "epoch": 4.82, "learning_rate": 1.106081028724484e-05, "loss": 0.3418, "step": 15680 }, { "epoch": 4.82, "learning_rate": 1.1059819364871518e-05, "loss": 0.2697, "step": 15681 }, { "epoch": 4.83, "learning_rate": 1.105882843197321e-05, "loss": 0.2833, "step": 15682 }, { "epoch": 4.83, "learning_rate": 1.1057837488559762e-05, "loss": 0.3197, "step": 15683 }, { "epoch": 4.83, "learning_rate": 1.1056846534641008e-05, "loss": 0.255, "step": 15684 }, { "epoch": 4.83, "learning_rate": 1.1055855570226799e-05, "loss": 0.4116, "step": 15685 }, { "epoch": 4.83, "learning_rate": 1.1054864595326968e-05, "loss": 0.4384, "step": 15686 }, { "epoch": 4.83, "learning_rate": 1.1053873609951362e-05, "loss": 0.3179, "step": 15687 }, { "epoch": 4.83, "learning_rate": 1.1052882614109816e-05, "loss": 0.3095, "step": 15688 }, { "epoch": 4.83, "learning_rate": 1.105189160781218e-05, "loss": 0.2912, "step": 15689 }, { "epoch": 4.83, "learning_rate": 1.1050900591068287e-05, "loss": 0.3818, "step": 15690 }, { "epoch": 4.83, "learning_rate": 1.1049909563887987e-05, "loss": 0.173, "step": 15691 }, { "epoch": 4.83, "learning_rate": 1.1048918526281114e-05, "loss": 0.2048, "step": 15692 }, { "epoch": 4.83, "learning_rate": 1.1047927478257511e-05, "loss": 0.2433, "step": 15693 }, { "epoch": 4.83, "learning_rate": 1.1046936419827026e-05, "loss": 0.2555, "step": 15694 }, { "epoch": 4.83, "learning_rate": 1.1045945350999498e-05, "loss": 0.4505, "step": 15695 }, { "epoch": 4.83, "learning_rate": 1.1044954271784768e-05, "loss": 0.3731, "step": 15696 }, { "epoch": 4.83, "learning_rate": 1.1043963182192679e-05, "loss": 0.3006, "step": 15697 }, { "epoch": 4.83, "learning_rate": 1.104297208223307e-05, "loss": 0.3449, "step": 15698 }, { "epoch": 4.83, "learning_rate": 1.1041980971915791e-05, "loss": 0.2827, "step": 15699 }, { "epoch": 4.83, "learning_rate": 1.1040989851250678e-05, "loss": 0.3363, "step": 15700 }, { "epoch": 4.83, "learning_rate": 1.103999872024758e-05, "loss": 0.2871, "step": 15701 }, { "epoch": 4.83, "learning_rate": 1.1039007578916333e-05, "loss": 0.2081, "step": 15702 }, { "epoch": 4.83, "learning_rate": 1.1038016427266784e-05, "loss": 0.1519, "step": 15703 }, { "epoch": 4.83, "learning_rate": 1.1037025265308776e-05, "loss": 0.3822, "step": 15704 }, { "epoch": 4.83, "learning_rate": 1.103603409305215e-05, "loss": 0.1915, "step": 15705 }, { "epoch": 4.83, "learning_rate": 1.1035042910506755e-05, "loss": 0.3044, "step": 15706 }, { "epoch": 4.83, "learning_rate": 1.1034051717682423e-05, "loss": 0.197, "step": 15707 }, { "epoch": 4.83, "learning_rate": 1.1033060514589009e-05, "loss": 0.2916, "step": 15708 }, { "epoch": 4.83, "learning_rate": 1.103206930123635e-05, "loss": 0.3921, "step": 15709 }, { "epoch": 4.83, "learning_rate": 1.1031078077634291e-05, "loss": 0.2683, "step": 15710 }, { "epoch": 4.83, "learning_rate": 1.1030086843792682e-05, "loss": 0.3978, "step": 15711 }, { "epoch": 4.83, "learning_rate": 1.1029095599721355e-05, "loss": 0.4192, "step": 15712 }, { "epoch": 4.83, "learning_rate": 1.1028104345430161e-05, "loss": 0.2568, "step": 15713 }, { "epoch": 4.83, "learning_rate": 1.1027113080928947e-05, "loss": 0.1347, "step": 15714 }, { "epoch": 4.84, "learning_rate": 1.102612180622755e-05, "loss": 0.3995, "step": 15715 }, { "epoch": 4.84, "learning_rate": 1.1025130521335819e-05, "loss": 0.4831, "step": 15716 }, { "epoch": 4.84, "learning_rate": 1.1024139226263596e-05, "loss": 0.4484, "step": 15717 }, { "epoch": 4.84, "learning_rate": 1.102314792102073e-05, "loss": 0.2882, "step": 15718 }, { "epoch": 4.84, "learning_rate": 1.1022156605617059e-05, "loss": 0.2315, "step": 15719 }, { "epoch": 4.84, "learning_rate": 1.1021165280062431e-05, "loss": 0.3942, "step": 15720 }, { "epoch": 4.84, "learning_rate": 1.1020173944366692e-05, "loss": 0.4038, "step": 15721 }, { "epoch": 4.84, "learning_rate": 1.1019182598539686e-05, "loss": 0.2708, "step": 15722 }, { "epoch": 4.84, "learning_rate": 1.1018191242591257e-05, "loss": 0.3199, "step": 15723 }, { "epoch": 4.84, "learning_rate": 1.1017199876531248e-05, "loss": 0.2332, "step": 15724 }, { "epoch": 4.84, "learning_rate": 1.101620850036951e-05, "loss": 0.1908, "step": 15725 }, { "epoch": 4.84, "learning_rate": 1.1015217114115884e-05, "loss": 0.4073, "step": 15726 }, { "epoch": 4.84, "learning_rate": 1.1014225717780216e-05, "loss": 0.4009, "step": 15727 }, { "epoch": 4.84, "learning_rate": 1.1013234311372353e-05, "loss": 0.3835, "step": 15728 }, { "epoch": 4.84, "learning_rate": 1.1012242894902136e-05, "loss": 0.4397, "step": 15729 }, { "epoch": 4.84, "learning_rate": 1.1011251468379418e-05, "loss": 0.2978, "step": 15730 }, { "epoch": 4.84, "learning_rate": 1.1010260031814037e-05, "loss": 0.3056, "step": 15731 }, { "epoch": 4.84, "learning_rate": 1.1009268585215848e-05, "loss": 0.4078, "step": 15732 }, { "epoch": 4.84, "learning_rate": 1.1008277128594692e-05, "loss": 0.2136, "step": 15733 }, { "epoch": 4.84, "learning_rate": 1.100728566196041e-05, "loss": 0.3506, "step": 15734 }, { "epoch": 4.84, "learning_rate": 1.1006294185322854e-05, "loss": 0.4022, "step": 15735 }, { "epoch": 4.84, "learning_rate": 1.1005302698691869e-05, "loss": 0.277, "step": 15736 }, { "epoch": 4.84, "learning_rate": 1.1004311202077307e-05, "loss": 0.377, "step": 15737 }, { "epoch": 4.84, "learning_rate": 1.1003319695489002e-05, "loss": 0.261, "step": 15738 }, { "epoch": 4.84, "learning_rate": 1.1002328178936813e-05, "loss": 0.3392, "step": 15739 }, { "epoch": 4.84, "learning_rate": 1.100133665243058e-05, "loss": 0.3587, "step": 15740 }, { "epoch": 4.84, "learning_rate": 1.1000345115980149e-05, "loss": 0.3201, "step": 15741 }, { "epoch": 4.84, "learning_rate": 1.0999353569595372e-05, "loss": 0.1948, "step": 15742 }, { "epoch": 4.84, "learning_rate": 1.099836201328609e-05, "loss": 0.213, "step": 15743 }, { "epoch": 4.84, "learning_rate": 1.0997370447062155e-05, "loss": 0.3553, "step": 15744 }, { "epoch": 4.84, "learning_rate": 1.0996378870933413e-05, "loss": 0.2528, "step": 15745 }, { "epoch": 4.84, "learning_rate": 1.0995387284909709e-05, "loss": 0.4233, "step": 15746 }, { "epoch": 4.85, "learning_rate": 1.0994395689000895e-05, "loss": 0.3292, "step": 15747 }, { "epoch": 4.85, "learning_rate": 1.099340408321681e-05, "loss": 0.274, "step": 15748 }, { "epoch": 4.85, "learning_rate": 1.0992412467567311e-05, "loss": 0.1744, "step": 15749 }, { "epoch": 4.85, "learning_rate": 1.0991420842062237e-05, "loss": 0.1543, "step": 15750 }, { "epoch": 4.85, "learning_rate": 1.0990429206711443e-05, "loss": 0.2144, "step": 15751 }, { "epoch": 4.85, "learning_rate": 1.0989437561524776e-05, "loss": 0.1663, "step": 15752 }, { "epoch": 4.85, "learning_rate": 1.0988445906512081e-05, "loss": 0.5366, "step": 15753 }, { "epoch": 4.85, "learning_rate": 1.0987454241683208e-05, "loss": 0.2563, "step": 15754 }, { "epoch": 4.85, "learning_rate": 1.0986462567048002e-05, "loss": 0.4515, "step": 15755 }, { "epoch": 4.85, "learning_rate": 1.0985470882616317e-05, "loss": 0.2527, "step": 15756 }, { "epoch": 4.85, "learning_rate": 1.0984479188397994e-05, "loss": 0.1418, "step": 15757 }, { "epoch": 4.85, "learning_rate": 1.0983487484402888e-05, "loss": 0.3167, "step": 15758 }, { "epoch": 4.85, "learning_rate": 1.0982495770640845e-05, "loss": 0.2775, "step": 15759 }, { "epoch": 4.85, "learning_rate": 1.098150404712171e-05, "loss": 0.2436, "step": 15760 }, { "epoch": 4.85, "learning_rate": 1.0980512313855338e-05, "loss": 0.4019, "step": 15761 }, { "epoch": 4.85, "learning_rate": 1.0979520570851577e-05, "loss": 0.2845, "step": 15762 }, { "epoch": 4.85, "learning_rate": 1.097852881812027e-05, "loss": 0.2681, "step": 15763 }, { "epoch": 4.85, "learning_rate": 1.0977537055671275e-05, "loss": 0.3756, "step": 15764 }, { "epoch": 4.85, "learning_rate": 1.097654528351443e-05, "loss": 0.3396, "step": 15765 }, { "epoch": 4.85, "learning_rate": 1.0975553501659593e-05, "loss": 0.3489, "step": 15766 }, { "epoch": 4.85, "learning_rate": 1.0974561710116612e-05, "loss": 0.2339, "step": 15767 }, { "epoch": 4.85, "learning_rate": 1.0973569908895334e-05, "loss": 0.2493, "step": 15768 }, { "epoch": 4.85, "learning_rate": 1.0972578098005609e-05, "loss": 0.3716, "step": 15769 }, { "epoch": 4.85, "learning_rate": 1.0971586277457287e-05, "loss": 0.4008, "step": 15770 }, { "epoch": 4.85, "learning_rate": 1.0970594447260218e-05, "loss": 0.3661, "step": 15771 }, { "epoch": 4.85, "learning_rate": 1.0969602607424251e-05, "loss": 0.3661, "step": 15772 }, { "epoch": 4.85, "learning_rate": 1.096861075795924e-05, "loss": 0.2516, "step": 15773 }, { "epoch": 4.85, "learning_rate": 1.0967618898875027e-05, "loss": 0.3201, "step": 15774 }, { "epoch": 4.85, "learning_rate": 1.096662703018147e-05, "loss": 0.3439, "step": 15775 }, { "epoch": 4.85, "learning_rate": 1.0965635151888415e-05, "loss": 0.4525, "step": 15776 }, { "epoch": 4.85, "learning_rate": 1.0964643264005712e-05, "loss": 0.1308, "step": 15777 }, { "epoch": 4.85, "learning_rate": 1.0963651366543214e-05, "loss": 0.2143, "step": 15778 }, { "epoch": 4.85, "learning_rate": 1.0962659459510765e-05, "loss": 0.2228, "step": 15779 }, { "epoch": 4.86, "learning_rate": 1.0961667542918224e-05, "loss": 0.3161, "step": 15780 }, { "epoch": 4.86, "learning_rate": 1.0960675616775436e-05, "loss": 0.3623, "step": 15781 }, { "epoch": 4.86, "learning_rate": 1.0959683681092259e-05, "loss": 0.2844, "step": 15782 }, { "epoch": 4.86, "learning_rate": 1.0958691735878536e-05, "loss": 0.3188, "step": 15783 }, { "epoch": 4.86, "learning_rate": 1.0957699781144119e-05, "loss": 0.3024, "step": 15784 }, { "epoch": 4.86, "learning_rate": 1.095670781689886e-05, "loss": 0.3427, "step": 15785 }, { "epoch": 4.86, "learning_rate": 1.095571584315261e-05, "loss": 0.1849, "step": 15786 }, { "epoch": 4.86, "learning_rate": 1.0954723859915224e-05, "loss": 0.3319, "step": 15787 }, { "epoch": 4.86, "learning_rate": 1.095373186719655e-05, "loss": 0.3789, "step": 15788 }, { "epoch": 4.86, "learning_rate": 1.0952739865006437e-05, "loss": 0.2292, "step": 15789 }, { "epoch": 4.86, "learning_rate": 1.095174785335474e-05, "loss": 0.3924, "step": 15790 }, { "epoch": 4.86, "learning_rate": 1.095075583225131e-05, "loss": 0.2183, "step": 15791 }, { "epoch": 4.86, "learning_rate": 1.0949763801705999e-05, "loss": 0.3854, "step": 15792 }, { "epoch": 4.86, "learning_rate": 1.0948771761728655e-05, "loss": 0.409, "step": 15793 }, { "epoch": 4.86, "learning_rate": 1.0947779712329138e-05, "loss": 0.2514, "step": 15794 }, { "epoch": 4.86, "learning_rate": 1.0946787653517292e-05, "loss": 0.4823, "step": 15795 }, { "epoch": 4.86, "learning_rate": 1.0945795585302973e-05, "loss": 0.2777, "step": 15796 }, { "epoch": 4.86, "learning_rate": 1.0944803507696033e-05, "loss": 0.3113, "step": 15797 }, { "epoch": 4.86, "learning_rate": 1.094381142070632e-05, "loss": 0.2481, "step": 15798 }, { "epoch": 4.86, "learning_rate": 1.0942819324343693e-05, "loss": 0.5006, "step": 15799 }, { "epoch": 4.86, "learning_rate": 1.0941827218617997e-05, "loss": 0.2351, "step": 15800 }, { "epoch": 4.86, "learning_rate": 1.0940835103539094e-05, "loss": 0.3089, "step": 15801 }, { "epoch": 4.86, "learning_rate": 1.093984297911683e-05, "loss": 0.5279, "step": 15802 }, { "epoch": 4.86, "learning_rate": 1.0938850845361058e-05, "loss": 0.3269, "step": 15803 }, { "epoch": 4.86, "learning_rate": 1.0937858702281631e-05, "loss": 0.315, "step": 15804 }, { "epoch": 4.86, "learning_rate": 1.0936866549888403e-05, "loss": 0.3078, "step": 15805 }, { "epoch": 4.86, "learning_rate": 1.0935874388191227e-05, "loss": 0.3585, "step": 15806 }, { "epoch": 4.86, "learning_rate": 1.0934882217199956e-05, "loss": 0.3678, "step": 15807 }, { "epoch": 4.86, "learning_rate": 1.0933890036924444e-05, "loss": 0.2744, "step": 15808 }, { "epoch": 4.86, "learning_rate": 1.0932897847374542e-05, "loss": 0.3527, "step": 15809 }, { "epoch": 4.86, "learning_rate": 1.0931905648560104e-05, "loss": 0.4147, "step": 15810 }, { "epoch": 4.86, "learning_rate": 1.0930913440490985e-05, "loss": 0.4412, "step": 15811 }, { "epoch": 4.87, "learning_rate": 1.0929921223177038e-05, "loss": 0.226, "step": 15812 }, { "epoch": 4.87, "learning_rate": 1.0928928996628117e-05, "loss": 0.4681, "step": 15813 }, { "epoch": 4.87, "learning_rate": 1.0927936760854075e-05, "loss": 0.2487, "step": 15814 }, { "epoch": 4.87, "learning_rate": 1.0926944515864762e-05, "loss": 0.2264, "step": 15815 }, { "epoch": 4.87, "learning_rate": 1.0925952261670038e-05, "loss": 0.2721, "step": 15816 }, { "epoch": 4.87, "learning_rate": 1.0924959998279754e-05, "loss": 0.2732, "step": 15817 }, { "epoch": 4.87, "learning_rate": 1.0923967725703767e-05, "loss": 0.2579, "step": 15818 }, { "epoch": 4.87, "learning_rate": 1.0922975443951926e-05, "loss": 0.2837, "step": 15819 }, { "epoch": 4.87, "learning_rate": 1.092198315303409e-05, "loss": 0.2929, "step": 15820 }, { "epoch": 4.87, "learning_rate": 1.0920990852960111e-05, "loss": 0.3344, "step": 15821 }, { "epoch": 4.87, "learning_rate": 1.0919998543739843e-05, "loss": 0.3065, "step": 15822 }, { "epoch": 4.87, "learning_rate": 1.0919006225383146e-05, "loss": 0.2486, "step": 15823 }, { "epoch": 4.87, "learning_rate": 1.0918013897899866e-05, "loss": 0.219, "step": 15824 }, { "epoch": 4.87, "learning_rate": 1.0917021561299864e-05, "loss": 0.1813, "step": 15825 }, { "epoch": 4.87, "learning_rate": 1.0916029215592992e-05, "loss": 0.3523, "step": 15826 }, { "epoch": 4.87, "learning_rate": 1.0915036860789103e-05, "loss": 0.2435, "step": 15827 }, { "epoch": 4.87, "learning_rate": 1.091404449689806e-05, "loss": 0.2419, "step": 15828 }, { "epoch": 4.87, "learning_rate": 1.0913052123929707e-05, "loss": 0.3657, "step": 15829 }, { "epoch": 4.87, "learning_rate": 1.0912059741893908e-05, "loss": 0.309, "step": 15830 }, { "epoch": 4.87, "learning_rate": 1.0911067350800512e-05, "loss": 0.2278, "step": 15831 }, { "epoch": 4.87, "learning_rate": 1.0910074950659381e-05, "loss": 0.368, "step": 15832 }, { "epoch": 4.87, "learning_rate": 1.0909082541480371e-05, "loss": 0.3923, "step": 15833 }, { "epoch": 4.87, "learning_rate": 1.0908090123273327e-05, "loss": 0.4148, "step": 15834 }, { "epoch": 4.87, "learning_rate": 1.0907097696048113e-05, "loss": 0.348, "step": 15835 }, { "epoch": 4.87, "learning_rate": 1.090610525981458e-05, "loss": 0.4585, "step": 15836 }, { "epoch": 4.87, "learning_rate": 1.090511281458259e-05, "loss": 0.3424, "step": 15837 }, { "epoch": 4.87, "learning_rate": 1.0904120360361996e-05, "loss": 0.2848, "step": 15838 }, { "epoch": 4.87, "learning_rate": 1.0903127897162654e-05, "loss": 0.2619, "step": 15839 }, { "epoch": 4.87, "learning_rate": 1.0902135424994418e-05, "loss": 0.4643, "step": 15840 }, { "epoch": 4.87, "learning_rate": 1.0901142943867144e-05, "loss": 0.5288, "step": 15841 }, { "epoch": 4.87, "learning_rate": 1.0900150453790694e-05, "loss": 0.2663, "step": 15842 }, { "epoch": 4.87, "learning_rate": 1.089915795477492e-05, "loss": 0.2667, "step": 15843 }, { "epoch": 4.87, "learning_rate": 1.0898165446829676e-05, "loss": 0.2438, "step": 15844 }, { "epoch": 4.88, "learning_rate": 1.0897172929964823e-05, "loss": 0.3409, "step": 15845 }, { "epoch": 4.88, "learning_rate": 1.0896180404190215e-05, "loss": 0.2591, "step": 15846 }, { "epoch": 4.88, "learning_rate": 1.089518786951571e-05, "loss": 0.2816, "step": 15847 }, { "epoch": 4.88, "learning_rate": 1.0894195325951166e-05, "loss": 0.235, "step": 15848 }, { "epoch": 4.88, "learning_rate": 1.0893202773506437e-05, "loss": 0.4407, "step": 15849 }, { "epoch": 4.88, "learning_rate": 1.089221021219138e-05, "loss": 0.2788, "step": 15850 }, { "epoch": 4.88, "learning_rate": 1.0891217642015854e-05, "loss": 0.3044, "step": 15851 }, { "epoch": 4.88, "learning_rate": 1.0890225062989718e-05, "loss": 0.4684, "step": 15852 }, { "epoch": 4.88, "learning_rate": 1.0889232475122827e-05, "loss": 0.2384, "step": 15853 }, { "epoch": 4.88, "learning_rate": 1.0888239878425035e-05, "loss": 0.2719, "step": 15854 }, { "epoch": 4.88, "learning_rate": 1.08872472729062e-05, "loss": 0.2965, "step": 15855 }, { "epoch": 4.88, "learning_rate": 1.0886254658576186e-05, "loss": 0.2474, "step": 15856 }, { "epoch": 4.88, "learning_rate": 1.0885262035444846e-05, "loss": 0.2721, "step": 15857 }, { "epoch": 4.88, "learning_rate": 1.088426940352204e-05, "loss": 0.3219, "step": 15858 }, { "epoch": 4.88, "learning_rate": 1.088327676281762e-05, "loss": 0.2696, "step": 15859 }, { "epoch": 4.88, "learning_rate": 1.0882284113341449e-05, "loss": 0.2946, "step": 15860 }, { "epoch": 4.88, "learning_rate": 1.0881291455103382e-05, "loss": 0.3103, "step": 15861 }, { "epoch": 4.88, "learning_rate": 1.0880298788113282e-05, "loss": 0.4802, "step": 15862 }, { "epoch": 4.88, "learning_rate": 1.0879306112381003e-05, "loss": 0.3222, "step": 15863 }, { "epoch": 4.88, "learning_rate": 1.0878313427916403e-05, "loss": 0.2579, "step": 15864 }, { "epoch": 4.88, "learning_rate": 1.087732073472934e-05, "loss": 0.17, "step": 15865 }, { "epoch": 4.88, "learning_rate": 1.0876328032829673e-05, "loss": 0.3472, "step": 15866 }, { "epoch": 4.88, "learning_rate": 1.087533532222726e-05, "loss": 0.3511, "step": 15867 }, { "epoch": 4.88, "learning_rate": 1.0874342602931967e-05, "loss": 0.2056, "step": 15868 }, { "epoch": 4.88, "learning_rate": 1.087334987495364e-05, "loss": 0.3458, "step": 15869 }, { "epoch": 4.88, "learning_rate": 1.0872357138302147e-05, "loss": 0.2589, "step": 15870 }, { "epoch": 4.88, "learning_rate": 1.0871364392987345e-05, "loss": 0.4089, "step": 15871 }, { "epoch": 4.88, "learning_rate": 1.0870371639019086e-05, "loss": 0.1636, "step": 15872 }, { "epoch": 4.88, "learning_rate": 1.0869378876407241e-05, "loss": 0.3526, "step": 15873 }, { "epoch": 4.88, "learning_rate": 1.0868386105161657e-05, "loss": 0.3175, "step": 15874 }, { "epoch": 4.88, "learning_rate": 1.0867393325292202e-05, "loss": 0.3159, "step": 15875 }, { "epoch": 4.88, "learning_rate": 1.0866400536808733e-05, "loss": 0.5461, "step": 15876 }, { "epoch": 4.89, "learning_rate": 1.0865407739721103e-05, "loss": 0.2059, "step": 15877 }, { "epoch": 4.89, "learning_rate": 1.0864414934039183e-05, "loss": 0.3471, "step": 15878 }, { "epoch": 4.89, "learning_rate": 1.0863422119772819e-05, "loss": 0.3271, "step": 15879 }, { "epoch": 4.89, "learning_rate": 1.0862429296931883e-05, "loss": 0.4607, "step": 15880 }, { "epoch": 4.89, "learning_rate": 1.0861436465526227e-05, "loss": 0.322, "step": 15881 }, { "epoch": 4.89, "learning_rate": 1.0860443625565712e-05, "loss": 0.3101, "step": 15882 }, { "epoch": 4.89, "learning_rate": 1.0859450777060202e-05, "loss": 0.4072, "step": 15883 }, { "epoch": 4.89, "learning_rate": 1.0858457920019552e-05, "loss": 0.6657, "step": 15884 }, { "epoch": 4.89, "learning_rate": 1.0857465054453622e-05, "loss": 0.2312, "step": 15885 }, { "epoch": 4.89, "learning_rate": 1.0856472180372275e-05, "loss": 0.2966, "step": 15886 }, { "epoch": 4.89, "learning_rate": 1.085547929778537e-05, "loss": 0.3053, "step": 15887 }, { "epoch": 4.89, "learning_rate": 1.0854486406702767e-05, "loss": 0.5039, "step": 15888 }, { "epoch": 4.89, "learning_rate": 1.0853493507134327e-05, "loss": 0.3981, "step": 15889 }, { "epoch": 4.89, "learning_rate": 1.085250059908991e-05, "loss": 0.4221, "step": 15890 }, { "epoch": 4.89, "learning_rate": 1.0851507682579376e-05, "loss": 0.285, "step": 15891 }, { "epoch": 4.89, "learning_rate": 1.0850514757612584e-05, "loss": 0.5096, "step": 15892 }, { "epoch": 4.89, "learning_rate": 1.0849521824199401e-05, "loss": 0.4699, "step": 15893 }, { "epoch": 4.89, "learning_rate": 1.084852888234968e-05, "loss": 0.256, "step": 15894 }, { "epoch": 4.89, "learning_rate": 1.0847535932073288e-05, "loss": 0.2296, "step": 15895 }, { "epoch": 4.89, "learning_rate": 1.0846542973380078e-05, "loss": 0.3485, "step": 15896 }, { "epoch": 4.89, "learning_rate": 1.084555000627992e-05, "loss": 0.2933, "step": 15897 }, { "epoch": 4.89, "learning_rate": 1.0844557030782672e-05, "loss": 0.3419, "step": 15898 }, { "epoch": 4.89, "learning_rate": 1.0843564046898192e-05, "loss": 0.4296, "step": 15899 }, { "epoch": 4.89, "learning_rate": 1.0842571054636343e-05, "loss": 0.3545, "step": 15900 }, { "epoch": 4.89, "learning_rate": 1.084157805400699e-05, "loss": 0.4477, "step": 15901 }, { "epoch": 4.89, "learning_rate": 1.0840585045019989e-05, "loss": 0.3167, "step": 15902 }, { "epoch": 4.89, "learning_rate": 1.0839592027685204e-05, "loss": 0.2161, "step": 15903 }, { "epoch": 4.89, "learning_rate": 1.0838599002012498e-05, "loss": 0.3407, "step": 15904 }, { "epoch": 4.89, "learning_rate": 1.0837605968011729e-05, "loss": 0.343, "step": 15905 }, { "epoch": 4.89, "learning_rate": 1.083661292569276e-05, "loss": 0.1991, "step": 15906 }, { "epoch": 4.89, "learning_rate": 1.0835619875065456e-05, "loss": 0.3433, "step": 15907 }, { "epoch": 4.89, "learning_rate": 1.0834626816139678e-05, "loss": 0.2618, "step": 15908 }, { "epoch": 4.89, "learning_rate": 1.0833633748925283e-05, "loss": 0.3732, "step": 15909 }, { "epoch": 4.9, "learning_rate": 1.0832640673432135e-05, "loss": 0.1827, "step": 15910 }, { "epoch": 4.9, "learning_rate": 1.0831647589670103e-05, "loss": 0.2686, "step": 15911 }, { "epoch": 4.9, "learning_rate": 1.083065449764904e-05, "loss": 0.2649, "step": 15912 }, { "epoch": 4.9, "learning_rate": 1.0829661397378813e-05, "loss": 0.2998, "step": 15913 }, { "epoch": 4.9, "learning_rate": 1.0828668288869287e-05, "loss": 0.2439, "step": 15914 }, { "epoch": 4.9, "learning_rate": 1.0827675172130317e-05, "loss": 0.4357, "step": 15915 }, { "epoch": 4.9, "learning_rate": 1.082668204717177e-05, "loss": 0.336, "step": 15916 }, { "epoch": 4.9, "learning_rate": 1.082568891400351e-05, "loss": 0.3219, "step": 15917 }, { "epoch": 4.9, "learning_rate": 1.0824695772635397e-05, "loss": 0.2788, "step": 15918 }, { "epoch": 4.9, "learning_rate": 1.0823702623077295e-05, "loss": 0.4302, "step": 15919 }, { "epoch": 4.9, "learning_rate": 1.0822709465339068e-05, "loss": 0.3569, "step": 15920 }, { "epoch": 4.9, "learning_rate": 1.0821716299430577e-05, "loss": 0.3744, "step": 15921 }, { "epoch": 4.9, "learning_rate": 1.0820723125361685e-05, "loss": 0.2339, "step": 15922 }, { "epoch": 4.9, "learning_rate": 1.0819729943142259e-05, "loss": 0.3208, "step": 15923 }, { "epoch": 4.9, "learning_rate": 1.0818736752782157e-05, "loss": 0.3229, "step": 15924 }, { "epoch": 4.9, "learning_rate": 1.0817743554291245e-05, "loss": 0.3794, "step": 15925 }, { "epoch": 4.9, "learning_rate": 1.0816750347679386e-05, "loss": 0.4051, "step": 15926 }, { "epoch": 4.9, "learning_rate": 1.0815757132956442e-05, "loss": 0.2943, "step": 15927 }, { "epoch": 4.9, "learning_rate": 1.0814763910132278e-05, "loss": 0.4388, "step": 15928 }, { "epoch": 4.9, "learning_rate": 1.0813770679216758e-05, "loss": 0.1637, "step": 15929 }, { "epoch": 4.9, "learning_rate": 1.0812777440219746e-05, "loss": 0.3369, "step": 15930 }, { "epoch": 4.9, "learning_rate": 1.0811784193151105e-05, "loss": 0.3374, "step": 15931 }, { "epoch": 4.9, "learning_rate": 1.0810790938020697e-05, "loss": 0.1802, "step": 15932 }, { "epoch": 4.9, "learning_rate": 1.0809797674838389e-05, "loss": 0.3361, "step": 15933 }, { "epoch": 4.9, "learning_rate": 1.0808804403614044e-05, "loss": 0.3593, "step": 15934 }, { "epoch": 4.9, "learning_rate": 1.0807811124357525e-05, "loss": 0.3406, "step": 15935 }, { "epoch": 4.9, "learning_rate": 1.0806817837078695e-05, "loss": 0.3312, "step": 15936 }, { "epoch": 4.9, "learning_rate": 1.0805824541787425e-05, "loss": 0.3299, "step": 15937 }, { "epoch": 4.9, "learning_rate": 1.0804831238493572e-05, "loss": 0.4453, "step": 15938 }, { "epoch": 4.9, "learning_rate": 1.0803837927207004e-05, "loss": 0.2187, "step": 15939 }, { "epoch": 4.9, "learning_rate": 1.0802844607937584e-05, "loss": 0.265, "step": 15940 }, { "epoch": 4.9, "learning_rate": 1.0801851280695175e-05, "loss": 0.2703, "step": 15941 }, { "epoch": 4.91, "learning_rate": 1.0800857945489647e-05, "loss": 0.2634, "step": 15942 }, { "epoch": 4.91, "learning_rate": 1.0799864602330861e-05, "loss": 0.2803, "step": 15943 }, { "epoch": 4.91, "learning_rate": 1.0798871251228681e-05, "loss": 0.4646, "step": 15944 }, { "epoch": 4.91, "learning_rate": 1.0797877892192974e-05, "loss": 0.4964, "step": 15945 }, { "epoch": 4.91, "learning_rate": 1.0796884525233602e-05, "loss": 0.4448, "step": 15946 }, { "epoch": 4.91, "learning_rate": 1.0795891150360435e-05, "loss": 0.2669, "step": 15947 }, { "epoch": 4.91, "learning_rate": 1.0794897767583334e-05, "loss": 0.2923, "step": 15948 }, { "epoch": 4.91, "learning_rate": 1.0793904376912168e-05, "loss": 0.2098, "step": 15949 }, { "epoch": 4.91, "learning_rate": 1.0792910978356797e-05, "loss": 0.2168, "step": 15950 }, { "epoch": 4.91, "learning_rate": 1.079191757192709e-05, "loss": 0.3059, "step": 15951 }, { "epoch": 4.91, "learning_rate": 1.0790924157632914e-05, "loss": 0.2767, "step": 15952 }, { "epoch": 4.91, "learning_rate": 1.0789930735484126e-05, "loss": 0.331, "step": 15953 }, { "epoch": 4.91, "learning_rate": 1.0788937305490605e-05, "loss": 0.1731, "step": 15954 }, { "epoch": 4.91, "learning_rate": 1.0787943867662206e-05, "loss": 0.2213, "step": 15955 }, { "epoch": 4.91, "learning_rate": 1.0786950422008799e-05, "loss": 0.1672, "step": 15956 }, { "epoch": 4.91, "learning_rate": 1.0785956968540248e-05, "loss": 0.3257, "step": 15957 }, { "epoch": 4.91, "learning_rate": 1.0784963507266419e-05, "loss": 0.3168, "step": 15958 }, { "epoch": 4.91, "learning_rate": 1.0783970038197183e-05, "loss": 0.4822, "step": 15959 }, { "epoch": 4.91, "learning_rate": 1.0782976561342398e-05, "loss": 0.2877, "step": 15960 }, { "epoch": 4.91, "learning_rate": 1.0781983076711937e-05, "loss": 0.3459, "step": 15961 }, { "epoch": 4.91, "learning_rate": 1.0780989584315662e-05, "loss": 0.2742, "step": 15962 }, { "epoch": 4.91, "learning_rate": 1.0779996084163437e-05, "loss": 0.242, "step": 15963 }, { "epoch": 4.91, "learning_rate": 1.077900257626514e-05, "loss": 0.2474, "step": 15964 }, { "epoch": 4.91, "learning_rate": 1.0778009060630624e-05, "loss": 0.3645, "step": 15965 }, { "epoch": 4.91, "learning_rate": 1.077701553726976e-05, "loss": 0.3073, "step": 15966 }, { "epoch": 4.91, "learning_rate": 1.0776022006192417e-05, "loss": 0.1855, "step": 15967 }, { "epoch": 4.91, "learning_rate": 1.0775028467408462e-05, "loss": 0.3362, "step": 15968 }, { "epoch": 4.91, "learning_rate": 1.0774034920927758e-05, "loss": 0.3776, "step": 15969 }, { "epoch": 4.91, "learning_rate": 1.0773041366760178e-05, "loss": 0.3026, "step": 15970 }, { "epoch": 4.91, "learning_rate": 1.077204780491558e-05, "loss": 0.3443, "step": 15971 }, { "epoch": 4.91, "learning_rate": 1.0771054235403838e-05, "loss": 0.2628, "step": 15972 }, { "epoch": 4.91, "learning_rate": 1.0770060658234815e-05, "loss": 0.345, "step": 15973 }, { "epoch": 4.91, "learning_rate": 1.0769067073418383e-05, "loss": 0.298, "step": 15974 }, { "epoch": 4.92, "learning_rate": 1.0768073480964404e-05, "loss": 0.301, "step": 15975 }, { "epoch": 4.92, "learning_rate": 1.0767079880882747e-05, "loss": 0.2672, "step": 15976 }, { "epoch": 4.92, "learning_rate": 1.0766086273183279e-05, "loss": 0.1446, "step": 15977 }, { "epoch": 4.92, "learning_rate": 1.0765092657875871e-05, "loss": 0.2106, "step": 15978 }, { "epoch": 4.92, "learning_rate": 1.0764099034970385e-05, "loss": 0.3409, "step": 15979 }, { "epoch": 4.92, "learning_rate": 1.0763105404476695e-05, "loss": 0.3101, "step": 15980 }, { "epoch": 4.92, "learning_rate": 1.0762111766404664e-05, "loss": 0.2414, "step": 15981 }, { "epoch": 4.92, "learning_rate": 1.0761118120764157e-05, "loss": 0.3392, "step": 15982 }, { "epoch": 4.92, "learning_rate": 1.076012446756505e-05, "loss": 0.3327, "step": 15983 }, { "epoch": 4.92, "learning_rate": 1.0759130806817204e-05, "loss": 0.276, "step": 15984 }, { "epoch": 4.92, "learning_rate": 1.0758137138530491e-05, "loss": 0.4435, "step": 15985 }, { "epoch": 4.92, "learning_rate": 1.0757143462714777e-05, "loss": 0.3776, "step": 15986 }, { "epoch": 4.92, "learning_rate": 1.075614977937993e-05, "loss": 0.3779, "step": 15987 }, { "epoch": 4.92, "learning_rate": 1.0755156088535819e-05, "loss": 0.2587, "step": 15988 }, { "epoch": 4.92, "learning_rate": 1.0754162390192313e-05, "loss": 0.2969, "step": 15989 }, { "epoch": 4.92, "learning_rate": 1.0753168684359278e-05, "loss": 0.2872, "step": 15990 }, { "epoch": 4.92, "learning_rate": 1.0752174971046583e-05, "loss": 0.4789, "step": 15991 }, { "epoch": 4.92, "learning_rate": 1.07511812502641e-05, "loss": 0.3163, "step": 15992 }, { "epoch": 4.92, "learning_rate": 1.0750187522021693e-05, "loss": 0.3238, "step": 15993 }, { "epoch": 4.92, "learning_rate": 1.0749193786329235e-05, "loss": 0.1908, "step": 15994 }, { "epoch": 4.92, "learning_rate": 1.074820004319659e-05, "loss": 0.3792, "step": 15995 }, { "epoch": 4.92, "learning_rate": 1.074720629263363e-05, "loss": 0.3016, "step": 15996 }, { "epoch": 4.92, "learning_rate": 1.0746212534650222e-05, "loss": 0.3507, "step": 15997 }, { "epoch": 4.92, "learning_rate": 1.0745218769256233e-05, "loss": 0.3771, "step": 15998 }, { "epoch": 4.92, "learning_rate": 1.0744224996461541e-05, "loss": 0.4868, "step": 15999 }, { "epoch": 4.92, "learning_rate": 1.0743231216276006e-05, "loss": 0.3356, "step": 16000 }, { "epoch": 4.92, "learning_rate": 1.0742237428709499e-05, "loss": 0.1973, "step": 16001 }, { "epoch": 4.92, "learning_rate": 1.0741243633771892e-05, "loss": 0.265, "step": 16002 }, { "epoch": 4.92, "learning_rate": 1.0740249831473049e-05, "loss": 0.2676, "step": 16003 }, { "epoch": 4.92, "learning_rate": 1.0739256021822849e-05, "loss": 0.2968, "step": 16004 }, { "epoch": 4.92, "learning_rate": 1.0738262204831148e-05, "loss": 0.3389, "step": 16005 }, { "epoch": 4.92, "learning_rate": 1.0737268380507827e-05, "loss": 0.2224, "step": 16006 }, { "epoch": 4.93, "learning_rate": 1.0736274548862755e-05, "loss": 0.1398, "step": 16007 }, { "epoch": 4.93, "learning_rate": 1.0735280709905791e-05, "loss": 0.39, "step": 16008 }, { "epoch": 4.93, "learning_rate": 1.0734286863646817e-05, "loss": 0.3084, "step": 16009 }, { "epoch": 4.93, "learning_rate": 1.0733293010095696e-05, "loss": 0.2468, "step": 16010 }, { "epoch": 4.93, "learning_rate": 1.0732299149262299e-05, "loss": 0.3878, "step": 16011 }, { "epoch": 4.93, "learning_rate": 1.0731305281156499e-05, "loss": 0.3776, "step": 16012 }, { "epoch": 4.93, "learning_rate": 1.073031140578816e-05, "loss": 0.2988, "step": 16013 }, { "epoch": 4.93, "learning_rate": 1.0729317523167158e-05, "loss": 0.2506, "step": 16014 }, { "epoch": 4.93, "learning_rate": 1.0728323633303361e-05, "loss": 0.4761, "step": 16015 }, { "epoch": 4.93, "learning_rate": 1.072732973620664e-05, "loss": 0.2747, "step": 16016 }, { "epoch": 4.93, "learning_rate": 1.072633583188686e-05, "loss": 0.3957, "step": 16017 }, { "epoch": 4.93, "learning_rate": 1.07253419203539e-05, "loss": 0.2529, "step": 16018 }, { "epoch": 4.93, "learning_rate": 1.0724348001617626e-05, "loss": 0.2101, "step": 16019 }, { "epoch": 4.93, "learning_rate": 1.0723354075687907e-05, "loss": 0.302, "step": 16020 }, { "epoch": 4.93, "learning_rate": 1.0722360142574617e-05, "loss": 0.3073, "step": 16021 }, { "epoch": 4.93, "learning_rate": 1.0721366202287621e-05, "loss": 0.2533, "step": 16022 }, { "epoch": 4.93, "learning_rate": 1.0720372254836798e-05, "loss": 0.4296, "step": 16023 }, { "epoch": 4.93, "learning_rate": 1.0719378300232013e-05, "loss": 0.2096, "step": 16024 }, { "epoch": 4.93, "learning_rate": 1.0718384338483141e-05, "loss": 0.3337, "step": 16025 }, { "epoch": 4.93, "learning_rate": 1.071739036960005e-05, "loss": 0.4211, "step": 16026 }, { "epoch": 4.93, "learning_rate": 1.0716396393592606e-05, "loss": 0.2326, "step": 16027 }, { "epoch": 4.93, "learning_rate": 1.071540241047069e-05, "loss": 0.3534, "step": 16028 }, { "epoch": 4.93, "learning_rate": 1.0714408420244169e-05, "loss": 0.3633, "step": 16029 }, { "epoch": 4.93, "learning_rate": 1.0713414422922912e-05, "loss": 0.3218, "step": 16030 }, { "epoch": 4.93, "learning_rate": 1.0712420418516796e-05, "loss": 0.1731, "step": 16031 }, { "epoch": 4.93, "learning_rate": 1.0711426407035684e-05, "loss": 0.4395, "step": 16032 }, { "epoch": 4.93, "learning_rate": 1.0710432388489453e-05, "loss": 0.1772, "step": 16033 }, { "epoch": 4.93, "learning_rate": 1.0709438362887975e-05, "loss": 0.2393, "step": 16034 }, { "epoch": 4.93, "learning_rate": 1.0708444330241122e-05, "loss": 0.379, "step": 16035 }, { "epoch": 4.93, "learning_rate": 1.0707450290558759e-05, "loss": 0.3213, "step": 16036 }, { "epoch": 4.93, "learning_rate": 1.0706456243850766e-05, "loss": 0.4308, "step": 16037 }, { "epoch": 4.93, "learning_rate": 1.0705462190127011e-05, "loss": 0.2884, "step": 16038 }, { "epoch": 4.93, "learning_rate": 1.0704468129397366e-05, "loss": 0.331, "step": 16039 }, { "epoch": 4.94, "learning_rate": 1.0703474061671706e-05, "loss": 0.3076, "step": 16040 }, { "epoch": 4.94, "learning_rate": 1.0702479986959895e-05, "loss": 0.256, "step": 16041 }, { "epoch": 4.94, "learning_rate": 1.0701485905271814e-05, "loss": 0.3438, "step": 16042 }, { "epoch": 4.94, "learning_rate": 1.0700491816617328e-05, "loss": 0.3856, "step": 16043 }, { "epoch": 4.94, "learning_rate": 1.0699497721006315e-05, "loss": 0.2974, "step": 16044 }, { "epoch": 4.94, "learning_rate": 1.0698503618448647e-05, "loss": 0.2363, "step": 16045 }, { "epoch": 4.94, "learning_rate": 1.0697509508954192e-05, "loss": 0.3833, "step": 16046 }, { "epoch": 4.94, "learning_rate": 1.0696515392532824e-05, "loss": 0.3634, "step": 16047 }, { "epoch": 4.94, "learning_rate": 1.0695521269194415e-05, "loss": 0.2252, "step": 16048 }, { "epoch": 4.94, "learning_rate": 1.0694527138948843e-05, "loss": 0.3342, "step": 16049 }, { "epoch": 4.94, "learning_rate": 1.0693533001805976e-05, "loss": 0.3931, "step": 16050 }, { "epoch": 4.94, "learning_rate": 1.0692538857775685e-05, "loss": 0.1503, "step": 16051 }, { "epoch": 4.94, "learning_rate": 1.0691544706867844e-05, "loss": 0.3161, "step": 16052 }, { "epoch": 4.94, "learning_rate": 1.0690550549092327e-05, "loss": 0.3755, "step": 16053 }, { "epoch": 4.94, "learning_rate": 1.0689556384459008e-05, "loss": 0.3529, "step": 16054 }, { "epoch": 4.94, "learning_rate": 1.068856221297776e-05, "loss": 0.3796, "step": 16055 }, { "epoch": 4.94, "learning_rate": 1.0687568034658454e-05, "loss": 0.3567, "step": 16056 }, { "epoch": 4.94, "learning_rate": 1.0686573849510962e-05, "loss": 0.3302, "step": 16057 }, { "epoch": 4.94, "learning_rate": 1.0685579657545157e-05, "loss": 0.2663, "step": 16058 }, { "epoch": 4.94, "learning_rate": 1.0684585458770918e-05, "loss": 0.3684, "step": 16059 }, { "epoch": 4.94, "learning_rate": 1.0683591253198113e-05, "loss": 0.2859, "step": 16060 }, { "epoch": 4.94, "learning_rate": 1.0682597040836617e-05, "loss": 0.3434, "step": 16061 }, { "epoch": 4.94, "learning_rate": 1.0681602821696305e-05, "loss": 0.3437, "step": 16062 }, { "epoch": 4.94, "learning_rate": 1.0680608595787045e-05, "loss": 0.2243, "step": 16063 }, { "epoch": 4.94, "learning_rate": 1.0679614363118718e-05, "loss": 0.2357, "step": 16064 }, { "epoch": 4.94, "learning_rate": 1.0678620123701192e-05, "loss": 0.2897, "step": 16065 }, { "epoch": 4.94, "learning_rate": 1.0677625877544344e-05, "loss": 0.262, "step": 16066 }, { "epoch": 4.94, "learning_rate": 1.0676631624658045e-05, "loss": 0.2896, "step": 16067 }, { "epoch": 4.94, "learning_rate": 1.0675637365052172e-05, "loss": 0.3465, "step": 16068 }, { "epoch": 4.94, "learning_rate": 1.0674643098736596e-05, "loss": 0.3664, "step": 16069 }, { "epoch": 4.94, "learning_rate": 1.0673648825721192e-05, "loss": 0.221, "step": 16070 }, { "epoch": 4.94, "learning_rate": 1.0672654546015837e-05, "loss": 0.2761, "step": 16071 }, { "epoch": 4.95, "learning_rate": 1.0671660259630398e-05, "loss": 0.3511, "step": 16072 }, { "epoch": 4.95, "learning_rate": 1.0670665966574756e-05, "loss": 0.3647, "step": 16073 }, { "epoch": 4.95, "learning_rate": 1.0669671666858783e-05, "loss": 0.246, "step": 16074 }, { "epoch": 4.95, "learning_rate": 1.0668677360492354e-05, "loss": 0.3611, "step": 16075 }, { "epoch": 4.95, "learning_rate": 1.0667683047485343e-05, "loss": 0.5122, "step": 16076 }, { "epoch": 4.95, "learning_rate": 1.066668872784762e-05, "loss": 0.2104, "step": 16077 }, { "epoch": 4.95, "learning_rate": 1.0665694401589066e-05, "loss": 0.4089, "step": 16078 }, { "epoch": 4.95, "learning_rate": 1.0664700068719554e-05, "loss": 0.2911, "step": 16079 }, { "epoch": 4.95, "learning_rate": 1.0663705729248957e-05, "loss": 0.3532, "step": 16080 }, { "epoch": 4.95, "learning_rate": 1.0662711383187152e-05, "loss": 0.3301, "step": 16081 }, { "epoch": 4.95, "learning_rate": 1.0661717030544008e-05, "loss": 0.2316, "step": 16082 }, { "epoch": 4.95, "learning_rate": 1.0660722671329406e-05, "loss": 0.2214, "step": 16083 }, { "epoch": 4.95, "learning_rate": 1.0659728305553218e-05, "loss": 0.2702, "step": 16084 }, { "epoch": 4.95, "learning_rate": 1.0658733933225325e-05, "loss": 0.2875, "step": 16085 }, { "epoch": 4.95, "learning_rate": 1.0657739554355591e-05, "loss": 0.3219, "step": 16086 }, { "epoch": 4.95, "learning_rate": 1.06567451689539e-05, "loss": 0.2622, "step": 16087 }, { "epoch": 4.95, "learning_rate": 1.0655750777030123e-05, "loss": 0.4325, "step": 16088 }, { "epoch": 4.95, "learning_rate": 1.0654756378594135e-05, "loss": 0.2393, "step": 16089 }, { "epoch": 4.95, "learning_rate": 1.0653761973655819e-05, "loss": 0.3477, "step": 16090 }, { "epoch": 4.95, "learning_rate": 1.0652767562225038e-05, "loss": 0.3474, "step": 16091 }, { "epoch": 4.95, "learning_rate": 1.0651773144311674e-05, "loss": 0.3866, "step": 16092 }, { "epoch": 4.95, "learning_rate": 1.0650778719925606e-05, "loss": 0.3846, "step": 16093 }, { "epoch": 4.95, "learning_rate": 1.0649784289076699e-05, "loss": 0.3769, "step": 16094 }, { "epoch": 4.95, "learning_rate": 1.0648789851774841e-05, "loss": 0.3798, "step": 16095 }, { "epoch": 4.95, "learning_rate": 1.0647795408029899e-05, "loss": 0.303, "step": 16096 }, { "epoch": 4.95, "learning_rate": 1.0646800957851753e-05, "loss": 0.3367, "step": 16097 }, { "epoch": 4.95, "learning_rate": 1.0645806501250275e-05, "loss": 0.1715, "step": 16098 }, { "epoch": 4.95, "learning_rate": 1.0644812038235343e-05, "loss": 0.2485, "step": 16099 }, { "epoch": 4.95, "learning_rate": 1.0643817568816839e-05, "loss": 0.3796, "step": 16100 }, { "epoch": 4.95, "learning_rate": 1.0642823093004628e-05, "loss": 0.3578, "step": 16101 }, { "epoch": 4.95, "learning_rate": 1.0641828610808593e-05, "loss": 0.5464, "step": 16102 }, { "epoch": 4.95, "learning_rate": 1.0640834122238606e-05, "loss": 0.3195, "step": 16103 }, { "epoch": 4.95, "learning_rate": 1.0639839627304548e-05, "loss": 0.3293, "step": 16104 }, { "epoch": 4.96, "learning_rate": 1.0638845126016294e-05, "loss": 0.4246, "step": 16105 }, { "epoch": 4.96, "learning_rate": 1.0637850618383717e-05, "loss": 0.3631, "step": 16106 }, { "epoch": 4.96, "learning_rate": 1.0636856104416695e-05, "loss": 0.2932, "step": 16107 }, { "epoch": 4.96, "learning_rate": 1.0635861584125104e-05, "loss": 0.2974, "step": 16108 }, { "epoch": 4.96, "learning_rate": 1.0634867057518824e-05, "loss": 0.3508, "step": 16109 }, { "epoch": 4.96, "learning_rate": 1.063387252460773e-05, "loss": 0.2866, "step": 16110 }, { "epoch": 4.96, "learning_rate": 1.0632877985401695e-05, "loss": 0.4951, "step": 16111 }, { "epoch": 4.96, "learning_rate": 1.0631883439910599e-05, "loss": 0.2274, "step": 16112 }, { "epoch": 4.96, "learning_rate": 1.0630888888144317e-05, "loss": 0.278, "step": 16113 }, { "epoch": 4.96, "learning_rate": 1.0629894330112729e-05, "loss": 0.427, "step": 16114 }, { "epoch": 4.96, "learning_rate": 1.0628899765825708e-05, "loss": 0.344, "step": 16115 }, { "epoch": 4.96, "learning_rate": 1.0627905195293135e-05, "loss": 0.2137, "step": 16116 }, { "epoch": 4.96, "learning_rate": 1.0626910618524882e-05, "loss": 0.2122, "step": 16117 }, { "epoch": 4.96, "learning_rate": 1.0625916035530831e-05, "loss": 0.2338, "step": 16118 }, { "epoch": 4.96, "learning_rate": 1.0624921446320857e-05, "loss": 0.2289, "step": 16119 }, { "epoch": 4.96, "learning_rate": 1.0623926850904836e-05, "loss": 0.4718, "step": 16120 }, { "epoch": 4.96, "learning_rate": 1.0622932249292647e-05, "loss": 0.3763, "step": 16121 }, { "epoch": 4.96, "learning_rate": 1.0621937641494164e-05, "loss": 0.2314, "step": 16122 }, { "epoch": 4.96, "learning_rate": 1.062094302751927e-05, "loss": 0.2627, "step": 16123 }, { "epoch": 4.96, "learning_rate": 1.061994840737784e-05, "loss": 0.227, "step": 16124 }, { "epoch": 4.96, "learning_rate": 1.0618953781079748e-05, "loss": 0.2873, "step": 16125 }, { "epoch": 4.96, "learning_rate": 1.0617959148634877e-05, "loss": 0.2932, "step": 16126 }, { "epoch": 4.96, "learning_rate": 1.06169645100531e-05, "loss": 0.4839, "step": 16127 }, { "epoch": 4.96, "learning_rate": 1.06159698653443e-05, "loss": 0.202, "step": 16128 }, { "epoch": 4.96, "learning_rate": 1.061497521451835e-05, "loss": 0.3755, "step": 16129 }, { "epoch": 4.96, "learning_rate": 1.0613980557585127e-05, "loss": 0.3048, "step": 16130 }, { "epoch": 4.96, "learning_rate": 1.0612985894554515e-05, "loss": 0.4197, "step": 16131 }, { "epoch": 4.96, "learning_rate": 1.0611991225436384e-05, "loss": 0.3861, "step": 16132 }, { "epoch": 4.96, "learning_rate": 1.061099655024062e-05, "loss": 0.3947, "step": 16133 }, { "epoch": 4.96, "learning_rate": 1.0610001868977094e-05, "loss": 0.3151, "step": 16134 }, { "epoch": 4.96, "learning_rate": 1.0609007181655689e-05, "loss": 0.3424, "step": 16135 }, { "epoch": 4.96, "learning_rate": 1.0608012488286281e-05, "loss": 0.33, "step": 16136 }, { "epoch": 4.97, "learning_rate": 1.060701778887875e-05, "loss": 0.4633, "step": 16137 }, { "epoch": 4.97, "learning_rate": 1.0606023083442972e-05, "loss": 0.3723, "step": 16138 }, { "epoch": 4.97, "learning_rate": 1.0605028371988824e-05, "loss": 0.1737, "step": 16139 }, { "epoch": 4.97, "learning_rate": 1.060403365452619e-05, "loss": 0.1614, "step": 16140 }, { "epoch": 4.97, "learning_rate": 1.0603038931064944e-05, "loss": 0.3956, "step": 16141 }, { "epoch": 4.97, "learning_rate": 1.0602044201614965e-05, "loss": 0.3037, "step": 16142 }, { "epoch": 4.97, "learning_rate": 1.0601049466186132e-05, "loss": 0.2479, "step": 16143 }, { "epoch": 4.97, "learning_rate": 1.0600054724788325e-05, "loss": 0.3228, "step": 16144 }, { "epoch": 4.97, "learning_rate": 1.059905997743142e-05, "loss": 0.2899, "step": 16145 }, { "epoch": 4.97, "learning_rate": 1.0598065224125298e-05, "loss": 0.3903, "step": 16146 }, { "epoch": 4.97, "learning_rate": 1.0597070464879839e-05, "loss": 0.3641, "step": 16147 }, { "epoch": 4.97, "learning_rate": 1.059607569970492e-05, "loss": 0.2906, "step": 16148 }, { "epoch": 4.97, "learning_rate": 1.0595080928610416e-05, "loss": 0.3099, "step": 16149 }, { "epoch": 4.97, "learning_rate": 1.0594086151606213e-05, "loss": 0.272, "step": 16150 }, { "epoch": 4.97, "learning_rate": 1.0593091368702186e-05, "loss": 0.4808, "step": 16151 }, { "epoch": 4.97, "learning_rate": 1.0592096579908216e-05, "loss": 0.2776, "step": 16152 }, { "epoch": 4.97, "learning_rate": 1.059110178523418e-05, "loss": 0.2639, "step": 16153 }, { "epoch": 4.97, "learning_rate": 1.0590106984689959e-05, "loss": 0.4978, "step": 16154 }, { "epoch": 4.97, "learning_rate": 1.0589112178285432e-05, "loss": 0.3664, "step": 16155 }, { "epoch": 4.97, "learning_rate": 1.058811736603048e-05, "loss": 0.4573, "step": 16156 }, { "epoch": 4.97, "learning_rate": 1.0587122547934977e-05, "loss": 0.4094, "step": 16157 }, { "epoch": 4.97, "learning_rate": 1.0586127724008805e-05, "loss": 0.4265, "step": 16158 }, { "epoch": 4.97, "learning_rate": 1.058513289426185e-05, "loss": 0.3264, "step": 16159 }, { "epoch": 4.97, "learning_rate": 1.0584138058703982e-05, "loss": 0.2473, "step": 16160 }, { "epoch": 4.97, "learning_rate": 1.0583143217345086e-05, "loss": 0.3675, "step": 16161 }, { "epoch": 4.97, "learning_rate": 1.0582148370195043e-05, "loss": 0.4348, "step": 16162 }, { "epoch": 4.97, "learning_rate": 1.0581153517263725e-05, "loss": 0.2277, "step": 16163 }, { "epoch": 4.97, "learning_rate": 1.058015865856102e-05, "loss": 0.356, "step": 16164 }, { "epoch": 4.97, "learning_rate": 1.0579163794096802e-05, "loss": 0.2422, "step": 16165 }, { "epoch": 4.97, "learning_rate": 1.0578168923880959e-05, "loss": 0.2893, "step": 16166 }, { "epoch": 4.97, "learning_rate": 1.057717404792336e-05, "loss": 0.2617, "step": 16167 }, { "epoch": 4.97, "learning_rate": 1.0576179166233895e-05, "loss": 0.2074, "step": 16168 }, { "epoch": 4.97, "learning_rate": 1.057518427882244e-05, "loss": 0.2095, "step": 16169 }, { "epoch": 4.98, "learning_rate": 1.057418938569887e-05, "loss": 0.3029, "step": 16170 }, { "epoch": 4.98, "learning_rate": 1.0573194486873077e-05, "loss": 0.3873, "step": 16171 }, { "epoch": 4.98, "learning_rate": 1.057219958235493e-05, "loss": 0.2905, "step": 16172 }, { "epoch": 4.98, "learning_rate": 1.0571204672154314e-05, "loss": 0.2383, "step": 16173 }, { "epoch": 4.98, "learning_rate": 1.0570209756281111e-05, "loss": 0.1905, "step": 16174 }, { "epoch": 4.98, "learning_rate": 1.0569214834745196e-05, "loss": 0.3954, "step": 16175 }, { "epoch": 4.98, "learning_rate": 1.0568219907556458e-05, "loss": 0.3366, "step": 16176 }, { "epoch": 4.98, "learning_rate": 1.056722497472477e-05, "loss": 0.3467, "step": 16177 }, { "epoch": 4.98, "learning_rate": 1.0566230036260016e-05, "loss": 0.2635, "step": 16178 }, { "epoch": 4.98, "learning_rate": 1.0565235092172075e-05, "loss": 0.2884, "step": 16179 }, { "epoch": 4.98, "learning_rate": 1.0564240142470827e-05, "loss": 0.2968, "step": 16180 }, { "epoch": 4.98, "learning_rate": 1.056324518716616e-05, "loss": 0.3843, "step": 16181 }, { "epoch": 4.98, "learning_rate": 1.0562250226267945e-05, "loss": 0.3921, "step": 16182 }, { "epoch": 4.98, "learning_rate": 1.0561255259786067e-05, "loss": 0.274, "step": 16183 }, { "epoch": 4.98, "learning_rate": 1.0560260287730406e-05, "loss": 0.4562, "step": 16184 }, { "epoch": 4.98, "learning_rate": 1.0559265310110847e-05, "loss": 0.2843, "step": 16185 }, { "epoch": 4.98, "learning_rate": 1.0558270326937269e-05, "loss": 0.2378, "step": 16186 }, { "epoch": 4.98, "learning_rate": 1.0557275338219547e-05, "loss": 0.3443, "step": 16187 }, { "epoch": 4.98, "learning_rate": 1.055628034396757e-05, "loss": 0.352, "step": 16188 }, { "epoch": 4.98, "learning_rate": 1.0555285344191217e-05, "loss": 0.1883, "step": 16189 }, { "epoch": 4.98, "learning_rate": 1.0554290338900368e-05, "loss": 0.3642, "step": 16190 }, { "epoch": 4.98, "learning_rate": 1.0553295328104906e-05, "loss": 0.3543, "step": 16191 }, { "epoch": 4.98, "learning_rate": 1.0552300311814712e-05, "loss": 0.352, "step": 16192 }, { "epoch": 4.98, "learning_rate": 1.0551305290039667e-05, "loss": 0.4353, "step": 16193 }, { "epoch": 4.98, "learning_rate": 1.055031026278965e-05, "loss": 0.3166, "step": 16194 }, { "epoch": 4.98, "learning_rate": 1.0549315230074547e-05, "loss": 0.3618, "step": 16195 }, { "epoch": 4.98, "learning_rate": 1.0548320191904235e-05, "loss": 0.3154, "step": 16196 }, { "epoch": 4.98, "learning_rate": 1.0547325148288601e-05, "loss": 0.3441, "step": 16197 }, { "epoch": 4.98, "learning_rate": 1.0546330099237523e-05, "loss": 0.2214, "step": 16198 }, { "epoch": 4.98, "learning_rate": 1.054533504476088e-05, "loss": 0.564, "step": 16199 }, { "epoch": 4.98, "learning_rate": 1.0544339984868564e-05, "loss": 0.3448, "step": 16200 }, { "epoch": 4.98, "learning_rate": 1.0543344919570447e-05, "loss": 0.4302, "step": 16201 }, { "epoch": 4.99, "learning_rate": 1.0542349848876414e-05, "loss": 0.2775, "step": 16202 }, { "epoch": 4.99, "learning_rate": 1.0541354772796347e-05, "loss": 0.2969, "step": 16203 }, { "epoch": 4.99, "learning_rate": 1.0540359691340127e-05, "loss": 0.3266, "step": 16204 }, { "epoch": 4.99, "learning_rate": 1.0539364604517639e-05, "loss": 0.275, "step": 16205 }, { "epoch": 4.99, "learning_rate": 1.0538369512338762e-05, "loss": 0.3041, "step": 16206 }, { "epoch": 4.99, "learning_rate": 1.0537374414813384e-05, "loss": 0.3129, "step": 16207 }, { "epoch": 4.99, "learning_rate": 1.0536379311951376e-05, "loss": 0.3626, "step": 16208 }, { "epoch": 4.99, "learning_rate": 1.053538420376263e-05, "loss": 0.279, "step": 16209 }, { "epoch": 4.99, "learning_rate": 1.0534389090257029e-05, "loss": 0.2896, "step": 16210 }, { "epoch": 4.99, "learning_rate": 1.0533393971444446e-05, "loss": 0.3897, "step": 16211 }, { "epoch": 4.99, "learning_rate": 1.0532398847334774e-05, "loss": 0.3183, "step": 16212 }, { "epoch": 4.99, "learning_rate": 1.0531403717937888e-05, "loss": 0.312, "step": 16213 }, { "epoch": 4.99, "learning_rate": 1.0530408583263674e-05, "loss": 0.3238, "step": 16214 }, { "epoch": 4.99, "learning_rate": 1.0529413443322012e-05, "loss": 0.3239, "step": 16215 }, { "epoch": 4.99, "learning_rate": 1.0528418298122792e-05, "loss": 0.3553, "step": 16216 }, { "epoch": 4.99, "learning_rate": 1.0527423147675886e-05, "loss": 0.3276, "step": 16217 }, { "epoch": 4.99, "learning_rate": 1.0526427991991186e-05, "loss": 0.3873, "step": 16218 }, { "epoch": 4.99, "learning_rate": 1.0525432831078568e-05, "loss": 0.4133, "step": 16219 }, { "epoch": 4.99, "learning_rate": 1.0524437664947918e-05, "loss": 0.3332, "step": 16220 }, { "epoch": 4.99, "learning_rate": 1.0523442493609119e-05, "loss": 0.3106, "step": 16221 }, { "epoch": 4.99, "learning_rate": 1.0522447317072055e-05, "loss": 0.3973, "step": 16222 }, { "epoch": 4.99, "learning_rate": 1.0521452135346606e-05, "loss": 0.1764, "step": 16223 }, { "epoch": 4.99, "learning_rate": 1.0520456948442659e-05, "loss": 0.3192, "step": 16224 }, { "epoch": 4.99, "learning_rate": 1.0519461756370091e-05, "loss": 0.2369, "step": 16225 }, { "epoch": 4.99, "learning_rate": 1.0518466559138791e-05, "loss": 0.4058, "step": 16226 }, { "epoch": 4.99, "learning_rate": 1.0517471356758642e-05, "loss": 0.39, "step": 16227 }, { "epoch": 4.99, "learning_rate": 1.0516476149239522e-05, "loss": 0.3562, "step": 16228 }, { "epoch": 4.99, "learning_rate": 1.0515480936591323e-05, "loss": 0.3705, "step": 16229 }, { "epoch": 4.99, "learning_rate": 1.0514485718823917e-05, "loss": 0.3233, "step": 16230 }, { "epoch": 4.99, "learning_rate": 1.0513490495947196e-05, "loss": 0.2538, "step": 16231 }, { "epoch": 4.99, "learning_rate": 1.0512495267971044e-05, "loss": 0.3027, "step": 16232 }, { "epoch": 4.99, "learning_rate": 1.051150003490534e-05, "loss": 0.2335, "step": 16233 }, { "epoch": 4.99, "learning_rate": 1.0510504796759966e-05, "loss": 0.3413, "step": 16234 }, { "epoch": 5.0, "learning_rate": 1.0509509553544813e-05, "loss": 0.3108, "step": 16235 }, { "epoch": 5.0, "learning_rate": 1.0508514305269759e-05, "loss": 0.2359, "step": 16236 }, { "epoch": 5.0, "learning_rate": 1.0507519051944688e-05, "loss": 0.2885, "step": 16237 }, { "epoch": 5.0, "learning_rate": 1.0506523793579488e-05, "loss": 0.3195, "step": 16238 }, { "epoch": 5.0, "learning_rate": 1.0505528530184037e-05, "loss": 0.3502, "step": 16239 }, { "epoch": 5.0, "learning_rate": 1.0504533261768222e-05, "loss": 0.3833, "step": 16240 }, { "epoch": 5.0, "learning_rate": 1.0503537988341929e-05, "loss": 0.3827, "step": 16241 }, { "epoch": 5.0, "learning_rate": 1.0502542709915038e-05, "loss": 0.2802, "step": 16242 }, { "epoch": 5.0, "learning_rate": 1.0501547426497436e-05, "loss": 0.345, "step": 16243 }, { "epoch": 5.0, "learning_rate": 1.0500552138099004e-05, "loss": 0.3344, "step": 16244 }, { "epoch": 5.0, "learning_rate": 1.049955684472963e-05, "loss": 0.273, "step": 16245 }, { "epoch": 5.0, "learning_rate": 1.0498561546399194e-05, "loss": 0.2892, "step": 16246 }, { "epoch": 5.0, "learning_rate": 1.0497566243117585e-05, "loss": 0.1709, "step": 16247 }, { "epoch": 5.0, "learning_rate": 1.0496570934894684e-05, "loss": 0.3281, "step": 16248 }, { "epoch": 5.0, "learning_rate": 1.0495575621740372e-05, "loss": 0.1831, "step": 16249 }, { "epoch": 5.0, "learning_rate": 1.049458030366454e-05, "loss": 0.2622, "step": 16250 }, { "epoch": 5.0, "learning_rate": 1.0493584980677069e-05, "loss": 0.188, "step": 16251 }, { "epoch": 5.0, "learning_rate": 1.049258965278785e-05, "loss": 0.1441, "step": 16252 }, { "epoch": 5.0, "learning_rate": 1.0491594320006755e-05, "loss": 0.1361, "step": 16253 }, { "epoch": 5.0, "learning_rate": 1.0490598982343678e-05, "loss": 0.11, "step": 16254 }, { "epoch": 5.0, "learning_rate": 1.04896036398085e-05, "loss": 0.2077, "step": 16255 }, { "epoch": 5.0, "learning_rate": 1.0488608292411106e-05, "loss": 0.1208, "step": 16256 }, { "epoch": 5.0, "learning_rate": 1.0487612940161384e-05, "loss": 0.1386, "step": 16257 }, { "epoch": 5.0, "learning_rate": 1.0486617583069214e-05, "loss": 0.201, "step": 16258 }, { "epoch": 5.0, "learning_rate": 1.0485622221144485e-05, "loss": 0.2631, "step": 16259 }, { "epoch": 5.0, "learning_rate": 1.0484626854397078e-05, "loss": 0.1658, "step": 16260 }, { "epoch": 5.0, "learning_rate": 1.0483631482836878e-05, "loss": 0.1646, "step": 16261 }, { "epoch": 5.0, "learning_rate": 1.0482636106473777e-05, "loss": 0.223, "step": 16262 }, { "epoch": 5.0, "learning_rate": 1.0481640725317647e-05, "loss": 0.254, "step": 16263 }, { "epoch": 5.0, "learning_rate": 1.0480645339378386e-05, "loss": 0.1986, "step": 16264 }, { "epoch": 5.0, "learning_rate": 1.047964994866587e-05, "loss": 0.1875, "step": 16265 }, { "epoch": 5.0, "learning_rate": 1.0478654553189994e-05, "loss": 0.1942, "step": 16266 }, { "epoch": 5.01, "learning_rate": 1.0477659152960636e-05, "loss": 0.1498, "step": 16267 }, { "epoch": 5.01, "learning_rate": 1.0476663747987678e-05, "loss": 0.1415, "step": 16268 }, { "epoch": 5.01, "learning_rate": 1.0475668338281013e-05, "loss": 0.2102, "step": 16269 }, { "epoch": 5.01, "learning_rate": 1.047467292385052e-05, "loss": 0.2107, "step": 16270 }, { "epoch": 5.01, "learning_rate": 1.047367750470609e-05, "loss": 0.1379, "step": 16271 }, { "epoch": 5.01, "learning_rate": 1.0472682080857606e-05, "loss": 0.1623, "step": 16272 }, { "epoch": 5.01, "learning_rate": 1.0471686652314952e-05, "loss": 0.159, "step": 16273 }, { "epoch": 5.01, "learning_rate": 1.0470691219088016e-05, "loss": 0.1558, "step": 16274 }, { "epoch": 5.01, "learning_rate": 1.046969578118668e-05, "loss": 0.1942, "step": 16275 }, { "epoch": 5.01, "learning_rate": 1.0468700338620835e-05, "loss": 0.1941, "step": 16276 }, { "epoch": 5.01, "learning_rate": 1.0467704891400361e-05, "loss": 0.1763, "step": 16277 }, { "epoch": 5.01, "learning_rate": 1.0466709439535148e-05, "loss": 0.1706, "step": 16278 }, { "epoch": 5.01, "learning_rate": 1.0465713983035082e-05, "loss": 0.1915, "step": 16279 }, { "epoch": 5.01, "learning_rate": 1.0464718521910043e-05, "loss": 0.144, "step": 16280 }, { "epoch": 5.01, "learning_rate": 1.0463723056169922e-05, "loss": 0.1278, "step": 16281 }, { "epoch": 5.01, "learning_rate": 1.0462727585824606e-05, "loss": 0.205, "step": 16282 }, { "epoch": 5.01, "learning_rate": 1.0461732110883975e-05, "loss": 0.2016, "step": 16283 }, { "epoch": 5.01, "learning_rate": 1.046073663135792e-05, "loss": 0.1903, "step": 16284 }, { "epoch": 5.01, "learning_rate": 1.0459741147256325e-05, "loss": 0.1034, "step": 16285 }, { "epoch": 5.01, "learning_rate": 1.045874565858908e-05, "loss": 0.1151, "step": 16286 }, { "epoch": 5.01, "learning_rate": 1.0457750165366064e-05, "loss": 0.0833, "step": 16287 }, { "epoch": 5.01, "learning_rate": 1.0456754667597171e-05, "loss": 0.1202, "step": 16288 }, { "epoch": 5.01, "learning_rate": 1.0455759165292278e-05, "loss": 0.1317, "step": 16289 }, { "epoch": 5.01, "learning_rate": 1.045476365846128e-05, "loss": 0.1322, "step": 16290 }, { "epoch": 5.01, "learning_rate": 1.045376814711406e-05, "loss": 0.108, "step": 16291 }, { "epoch": 5.01, "learning_rate": 1.0452772631260504e-05, "loss": 0.2025, "step": 16292 }, { "epoch": 5.01, "learning_rate": 1.0451777110910497e-05, "loss": 0.1738, "step": 16293 }, { "epoch": 5.01, "learning_rate": 1.0450781586073926e-05, "loss": 0.1866, "step": 16294 }, { "epoch": 5.01, "learning_rate": 1.044978605676068e-05, "loss": 0.1755, "step": 16295 }, { "epoch": 5.01, "learning_rate": 1.0448790522980642e-05, "loss": 0.1829, "step": 16296 }, { "epoch": 5.01, "learning_rate": 1.0447794984743705e-05, "loss": 0.1398, "step": 16297 }, { "epoch": 5.01, "learning_rate": 1.044679944205975e-05, "loss": 0.178, "step": 16298 }, { "epoch": 5.01, "learning_rate": 1.044580389493866e-05, "loss": 0.0884, "step": 16299 }, { "epoch": 5.02, "learning_rate": 1.0444808343390332e-05, "loss": 0.1829, "step": 16300 }, { "epoch": 5.02, "learning_rate": 1.0443812787424641e-05, "loss": 0.1949, "step": 16301 }, { "epoch": 5.02, "learning_rate": 1.0442817227051487e-05, "loss": 0.1764, "step": 16302 }, { "epoch": 5.02, "learning_rate": 1.0441821662280745e-05, "loss": 0.0757, "step": 16303 }, { "epoch": 5.02, "learning_rate": 1.0440826093122309e-05, "loss": 0.235, "step": 16304 }, { "epoch": 5.02, "learning_rate": 1.0439830519586063e-05, "loss": 0.2261, "step": 16305 }, { "epoch": 5.02, "learning_rate": 1.0438834941681893e-05, "loss": 0.1835, "step": 16306 }, { "epoch": 5.02, "learning_rate": 1.0437839359419688e-05, "loss": 0.1351, "step": 16307 }, { "epoch": 5.02, "learning_rate": 1.0436843772809338e-05, "loss": 0.1598, "step": 16308 }, { "epoch": 5.02, "learning_rate": 1.0435848181860724e-05, "loss": 0.1422, "step": 16309 }, { "epoch": 5.02, "learning_rate": 1.0434852586583737e-05, "loss": 0.1598, "step": 16310 }, { "epoch": 5.02, "learning_rate": 1.043385698698826e-05, "loss": 0.1932, "step": 16311 }, { "epoch": 5.02, "learning_rate": 1.0432861383084187e-05, "loss": 0.2411, "step": 16312 }, { "epoch": 5.02, "learning_rate": 1.04318657748814e-05, "loss": 0.2136, "step": 16313 }, { "epoch": 5.02, "learning_rate": 1.0430870162389787e-05, "loss": 0.1374, "step": 16314 }, { "epoch": 5.02, "learning_rate": 1.0429874545619236e-05, "loss": 0.1971, "step": 16315 }, { "epoch": 5.02, "learning_rate": 1.0428878924579638e-05, "loss": 0.1532, "step": 16316 }, { "epoch": 5.02, "learning_rate": 1.0427883299280875e-05, "loss": 0.1503, "step": 16317 }, { "epoch": 5.02, "learning_rate": 1.0426887669732837e-05, "loss": 0.1845, "step": 16318 }, { "epoch": 5.02, "learning_rate": 1.0425892035945411e-05, "loss": 0.1745, "step": 16319 }, { "epoch": 5.02, "learning_rate": 1.042489639792848e-05, "loss": 0.2592, "step": 16320 }, { "epoch": 5.02, "learning_rate": 1.0423900755691942e-05, "loss": 0.1768, "step": 16321 }, { "epoch": 5.02, "learning_rate": 1.0422905109245678e-05, "loss": 0.1186, "step": 16322 }, { "epoch": 5.02, "learning_rate": 1.0421909458599574e-05, "loss": 0.2537, "step": 16323 }, { "epoch": 5.02, "learning_rate": 1.0420913803763522e-05, "loss": 0.2181, "step": 16324 }, { "epoch": 5.02, "learning_rate": 1.0419918144747408e-05, "loss": 0.1347, "step": 16325 }, { "epoch": 5.02, "learning_rate": 1.0418922481561119e-05, "loss": 0.1604, "step": 16326 }, { "epoch": 5.02, "learning_rate": 1.0417926814214542e-05, "loss": 0.1432, "step": 16327 }, { "epoch": 5.02, "learning_rate": 1.041693114271757e-05, "loss": 0.132, "step": 16328 }, { "epoch": 5.02, "learning_rate": 1.0415935467080087e-05, "loss": 0.201, "step": 16329 }, { "epoch": 5.02, "learning_rate": 1.0414939787311979e-05, "loss": 0.1467, "step": 16330 }, { "epoch": 5.02, "learning_rate": 1.0413944103423139e-05, "loss": 0.1884, "step": 16331 }, { "epoch": 5.03, "learning_rate": 1.041294841542345e-05, "loss": 0.2135, "step": 16332 }, { "epoch": 5.03, "learning_rate": 1.0411952723322806e-05, "loss": 0.242, "step": 16333 }, { "epoch": 5.03, "learning_rate": 1.0410957027131089e-05, "loss": 0.2365, "step": 16334 }, { "epoch": 5.03, "learning_rate": 1.0409961326858192e-05, "loss": 0.152, "step": 16335 }, { "epoch": 5.03, "learning_rate": 1.0408965622514e-05, "loss": 0.2574, "step": 16336 }, { "epoch": 5.03, "learning_rate": 1.04079699141084e-05, "loss": 0.1205, "step": 16337 }, { "epoch": 5.03, "learning_rate": 1.040697420165129e-05, "loss": 0.1457, "step": 16338 }, { "epoch": 5.03, "learning_rate": 1.0405978485152544e-05, "loss": 0.177, "step": 16339 }, { "epoch": 5.03, "learning_rate": 1.0404982764622061e-05, "loss": 0.1016, "step": 16340 }, { "epoch": 5.03, "learning_rate": 1.0403987040069725e-05, "loss": 0.2286, "step": 16341 }, { "epoch": 5.03, "learning_rate": 1.0402991311505423e-05, "loss": 0.29, "step": 16342 }, { "epoch": 5.03, "learning_rate": 1.040199557893905e-05, "loss": 0.1956, "step": 16343 }, { "epoch": 5.03, "learning_rate": 1.0400999842380488e-05, "loss": 0.1389, "step": 16344 }, { "epoch": 5.03, "learning_rate": 1.040000410183963e-05, "loss": 0.1442, "step": 16345 }, { "epoch": 5.03, "learning_rate": 1.0399008357326361e-05, "loss": 0.1585, "step": 16346 }, { "epoch": 5.03, "learning_rate": 1.039801260885057e-05, "loss": 0.1235, "step": 16347 }, { "epoch": 5.03, "learning_rate": 1.0397016856422151e-05, "loss": 0.1322, "step": 16348 }, { "epoch": 5.03, "learning_rate": 1.0396021100050984e-05, "loss": 0.1123, "step": 16349 }, { "epoch": 5.03, "learning_rate": 1.0395025339746965e-05, "loss": 0.214, "step": 16350 }, { "epoch": 5.03, "learning_rate": 1.0394029575519977e-05, "loss": 0.2011, "step": 16351 }, { "epoch": 5.03, "learning_rate": 1.0393033807379917e-05, "loss": 0.2367, "step": 16352 }, { "epoch": 5.03, "learning_rate": 1.0392038035336667e-05, "loss": 0.1864, "step": 16353 }, { "epoch": 5.03, "learning_rate": 1.0391042259400119e-05, "loss": 0.2213, "step": 16354 }, { "epoch": 5.03, "learning_rate": 1.0390046479580161e-05, "loss": 0.2111, "step": 16355 }, { "epoch": 5.03, "learning_rate": 1.0389050695886677e-05, "loss": 0.2614, "step": 16356 }, { "epoch": 5.03, "learning_rate": 1.0388054908329568e-05, "loss": 0.1225, "step": 16357 }, { "epoch": 5.03, "learning_rate": 1.0387059116918714e-05, "loss": 0.1492, "step": 16358 }, { "epoch": 5.03, "learning_rate": 1.0386063321664007e-05, "loss": 0.2056, "step": 16359 }, { "epoch": 5.03, "learning_rate": 1.0385067522575331e-05, "loss": 0.0898, "step": 16360 }, { "epoch": 5.03, "learning_rate": 1.0384071719662582e-05, "loss": 0.1284, "step": 16361 }, { "epoch": 5.03, "learning_rate": 1.0383075912935647e-05, "loss": 0.1545, "step": 16362 }, { "epoch": 5.03, "learning_rate": 1.0382080102404417e-05, "loss": 0.1769, "step": 16363 }, { "epoch": 5.03, "learning_rate": 1.0381084288078778e-05, "loss": 0.1663, "step": 16364 }, { "epoch": 5.04, "learning_rate": 1.0380088469968619e-05, "loss": 0.1602, "step": 16365 }, { "epoch": 5.04, "learning_rate": 1.0379092648083832e-05, "loss": 0.178, "step": 16366 }, { "epoch": 5.04, "learning_rate": 1.037809682243431e-05, "loss": 0.1548, "step": 16367 }, { "epoch": 5.04, "learning_rate": 1.0377100993029934e-05, "loss": 0.1927, "step": 16368 }, { "epoch": 5.04, "learning_rate": 1.0376105159880599e-05, "loss": 0.1658, "step": 16369 }, { "epoch": 5.04, "learning_rate": 1.037510932299619e-05, "loss": 0.1337, "step": 16370 }, { "epoch": 5.04, "learning_rate": 1.0374113482386604e-05, "loss": 0.2176, "step": 16371 }, { "epoch": 5.04, "learning_rate": 1.0373117638061725e-05, "loss": 0.2049, "step": 16372 }, { "epoch": 5.04, "learning_rate": 1.0372121790031443e-05, "loss": 0.1599, "step": 16373 }, { "epoch": 5.04, "learning_rate": 1.0371125938305648e-05, "loss": 0.1172, "step": 16374 }, { "epoch": 5.04, "learning_rate": 1.0370130082894232e-05, "loss": 0.1697, "step": 16375 }, { "epoch": 5.04, "learning_rate": 1.0369134223807082e-05, "loss": 0.164, "step": 16376 }, { "epoch": 5.04, "learning_rate": 1.036813836105409e-05, "loss": 0.1173, "step": 16377 }, { "epoch": 5.04, "learning_rate": 1.0367142494645146e-05, "loss": 0.1643, "step": 16378 }, { "epoch": 5.04, "learning_rate": 1.0366146624590137e-05, "loss": 0.2005, "step": 16379 }, { "epoch": 5.04, "learning_rate": 1.0365150750898952e-05, "loss": 0.1581, "step": 16380 }, { "epoch": 5.04, "learning_rate": 1.0364154873581486e-05, "loss": 0.2079, "step": 16381 }, { "epoch": 5.04, "learning_rate": 1.0363158992647626e-05, "loss": 0.1254, "step": 16382 }, { "epoch": 5.04, "learning_rate": 1.0362163108107267e-05, "loss": 0.1652, "step": 16383 }, { "epoch": 5.04, "learning_rate": 1.0361167219970287e-05, "loss": 0.1826, "step": 16384 }, { "epoch": 5.04, "learning_rate": 1.0360171328246589e-05, "loss": 0.1919, "step": 16385 }, { "epoch": 5.04, "learning_rate": 1.0359175432946057e-05, "loss": 0.147, "step": 16386 }, { "epoch": 5.04, "learning_rate": 1.0358179534078579e-05, "loss": 0.1845, "step": 16387 }, { "epoch": 5.04, "learning_rate": 1.0357183631654051e-05, "loss": 0.1677, "step": 16388 }, { "epoch": 5.04, "learning_rate": 1.0356187725682359e-05, "loss": 0.1749, "step": 16389 }, { "epoch": 5.04, "learning_rate": 1.0355191816173395e-05, "loss": 0.2228, "step": 16390 }, { "epoch": 5.04, "learning_rate": 1.035419590313705e-05, "loss": 0.1594, "step": 16391 }, { "epoch": 5.04, "learning_rate": 1.035319998658321e-05, "loss": 0.1035, "step": 16392 }, { "epoch": 5.04, "learning_rate": 1.0352204066521773e-05, "loss": 0.1375, "step": 16393 }, { "epoch": 5.04, "learning_rate": 1.0351208142962621e-05, "loss": 0.181, "step": 16394 }, { "epoch": 5.04, "learning_rate": 1.0350212215915651e-05, "loss": 0.1467, "step": 16395 }, { "epoch": 5.04, "learning_rate": 1.0349216285390751e-05, "loss": 0.1322, "step": 16396 }, { "epoch": 5.05, "learning_rate": 1.0348220351397808e-05, "loss": 0.1283, "step": 16397 }, { "epoch": 5.05, "learning_rate": 1.0347224413946721e-05, "loss": 0.2278, "step": 16398 }, { "epoch": 5.05, "learning_rate": 1.0346228473047373e-05, "loss": 0.1947, "step": 16399 }, { "epoch": 5.05, "learning_rate": 1.0345232528709656e-05, "loss": 0.2055, "step": 16400 }, { "epoch": 5.05, "learning_rate": 1.0344236580943463e-05, "loss": 0.1349, "step": 16401 }, { "epoch": 5.05, "learning_rate": 1.0343240629758683e-05, "loss": 0.1596, "step": 16402 }, { "epoch": 5.05, "learning_rate": 1.034224467516521e-05, "loss": 0.1206, "step": 16403 }, { "epoch": 5.05, "learning_rate": 1.034124871717293e-05, "loss": 0.1501, "step": 16404 }, { "epoch": 5.05, "learning_rate": 1.0340252755791736e-05, "loss": 0.2122, "step": 16405 }, { "epoch": 5.05, "learning_rate": 1.0339256791031515e-05, "loss": 0.07, "step": 16406 }, { "epoch": 5.05, "learning_rate": 1.0338260822902166e-05, "loss": 0.0863, "step": 16407 }, { "epoch": 5.05, "learning_rate": 1.0337264851413575e-05, "loss": 0.2242, "step": 16408 }, { "epoch": 5.05, "learning_rate": 1.0336268876575632e-05, "loss": 0.1786, "step": 16409 }, { "epoch": 5.05, "learning_rate": 1.033527289839823e-05, "loss": 0.1297, "step": 16410 }, { "epoch": 5.05, "learning_rate": 1.0334276916891257e-05, "loss": 0.1733, "step": 16411 }, { "epoch": 5.05, "learning_rate": 1.0333280932064606e-05, "loss": 0.1922, "step": 16412 }, { "epoch": 5.05, "learning_rate": 1.0332284943928169e-05, "loss": 0.1427, "step": 16413 }, { "epoch": 5.05, "learning_rate": 1.033128895249184e-05, "loss": 0.247, "step": 16414 }, { "epoch": 5.05, "learning_rate": 1.0330292957765502e-05, "loss": 0.2122, "step": 16415 }, { "epoch": 5.05, "learning_rate": 1.0329296959759052e-05, "loss": 0.1226, "step": 16416 }, { "epoch": 5.05, "learning_rate": 1.032830095848238e-05, "loss": 0.1048, "step": 16417 }, { "epoch": 5.05, "learning_rate": 1.0327304953945374e-05, "loss": 0.1383, "step": 16418 }, { "epoch": 5.05, "learning_rate": 1.0326308946157933e-05, "loss": 0.2533, "step": 16419 }, { "epoch": 5.05, "learning_rate": 1.032531293512994e-05, "loss": 0.1197, "step": 16420 }, { "epoch": 5.05, "learning_rate": 1.0324316920871292e-05, "loss": 0.1399, "step": 16421 }, { "epoch": 5.05, "learning_rate": 1.0323320903391876e-05, "loss": 0.076, "step": 16422 }, { "epoch": 5.05, "learning_rate": 1.0322324882701584e-05, "loss": 0.1389, "step": 16423 }, { "epoch": 5.05, "learning_rate": 1.0321328858810314e-05, "loss": 0.1608, "step": 16424 }, { "epoch": 5.05, "learning_rate": 1.0320332831727949e-05, "loss": 0.2409, "step": 16425 }, { "epoch": 5.05, "learning_rate": 1.0319336801464384e-05, "loss": 0.1541, "step": 16426 }, { "epoch": 5.05, "learning_rate": 1.0318340768029509e-05, "loss": 0.1715, "step": 16427 }, { "epoch": 5.05, "learning_rate": 1.0317344731433217e-05, "loss": 0.1989, "step": 16428 }, { "epoch": 5.05, "learning_rate": 1.0316348691685403e-05, "loss": 0.1561, "step": 16429 }, { "epoch": 5.06, "learning_rate": 1.0315352648795951e-05, "loss": 0.2853, "step": 16430 }, { "epoch": 5.06, "learning_rate": 1.0314356602774758e-05, "loss": 0.2119, "step": 16431 }, { "epoch": 5.06, "learning_rate": 1.0313360553631713e-05, "loss": 0.216, "step": 16432 }, { "epoch": 5.06, "learning_rate": 1.031236450137671e-05, "loss": 0.1737, "step": 16433 }, { "epoch": 5.06, "learning_rate": 1.031136844601964e-05, "loss": 0.1413, "step": 16434 }, { "epoch": 5.06, "learning_rate": 1.0310372387570395e-05, "loss": 0.2095, "step": 16435 }, { "epoch": 5.06, "learning_rate": 1.0309376326038868e-05, "loss": 0.1906, "step": 16436 }, { "epoch": 5.06, "learning_rate": 1.0308380261434942e-05, "loss": 0.155, "step": 16437 }, { "epoch": 5.06, "learning_rate": 1.0307384193768521e-05, "loss": 0.174, "step": 16438 }, { "epoch": 5.06, "learning_rate": 1.0306388123049492e-05, "loss": 0.2059, "step": 16439 }, { "epoch": 5.06, "learning_rate": 1.0305392049287744e-05, "loss": 0.2429, "step": 16440 }, { "epoch": 5.06, "learning_rate": 1.0304395972493172e-05, "loss": 0.1119, "step": 16441 }, { "epoch": 5.06, "learning_rate": 1.0303399892675668e-05, "loss": 0.2141, "step": 16442 }, { "epoch": 5.06, "learning_rate": 1.0302403809845126e-05, "loss": 0.194, "step": 16443 }, { "epoch": 5.06, "learning_rate": 1.0301407724011432e-05, "loss": 0.1321, "step": 16444 }, { "epoch": 5.06, "learning_rate": 1.0300411635184482e-05, "loss": 0.1505, "step": 16445 }, { "epoch": 5.06, "learning_rate": 1.029941554337417e-05, "loss": 0.1652, "step": 16446 }, { "epoch": 5.06, "learning_rate": 1.0298419448590382e-05, "loss": 0.2183, "step": 16447 }, { "epoch": 5.06, "learning_rate": 1.0297423350843017e-05, "loss": 0.1389, "step": 16448 }, { "epoch": 5.06, "learning_rate": 1.0296427250141962e-05, "loss": 0.1864, "step": 16449 }, { "epoch": 5.06, "learning_rate": 1.0295431146497112e-05, "loss": 0.1713, "step": 16450 }, { "epoch": 5.06, "learning_rate": 1.0294435039918357e-05, "loss": 0.1524, "step": 16451 }, { "epoch": 5.06, "learning_rate": 1.0293438930415592e-05, "loss": 0.2026, "step": 16452 }, { "epoch": 5.06, "learning_rate": 1.0292442817998709e-05, "loss": 0.1847, "step": 16453 }, { "epoch": 5.06, "learning_rate": 1.0291446702677598e-05, "loss": 0.1774, "step": 16454 }, { "epoch": 5.06, "learning_rate": 1.0290450584462155e-05, "loss": 0.2023, "step": 16455 }, { "epoch": 5.06, "learning_rate": 1.0289454463362265e-05, "loss": 0.1973, "step": 16456 }, { "epoch": 5.06, "learning_rate": 1.028845833938783e-05, "loss": 0.0673, "step": 16457 }, { "epoch": 5.06, "learning_rate": 1.0287462212548736e-05, "loss": 0.3608, "step": 16458 }, { "epoch": 5.06, "learning_rate": 1.0286466082854878e-05, "loss": 0.0813, "step": 16459 }, { "epoch": 5.06, "learning_rate": 1.0285469950316147e-05, "loss": 0.1831, "step": 16460 }, { "epoch": 5.06, "learning_rate": 1.0284473814942436e-05, "loss": 0.2081, "step": 16461 }, { "epoch": 5.07, "learning_rate": 1.0283477676743639e-05, "loss": 0.2137, "step": 16462 }, { "epoch": 5.07, "learning_rate": 1.0282481535729645e-05, "loss": 0.2175, "step": 16463 }, { "epoch": 5.07, "learning_rate": 1.0281485391910353e-05, "loss": 0.0968, "step": 16464 }, { "epoch": 5.07, "learning_rate": 1.028048924529565e-05, "loss": 0.1815, "step": 16465 }, { "epoch": 5.07, "learning_rate": 1.0279493095895427e-05, "loss": 0.143, "step": 16466 }, { "epoch": 5.07, "learning_rate": 1.0278496943719585e-05, "loss": 0.1298, "step": 16467 }, { "epoch": 5.07, "learning_rate": 1.0277500788778008e-05, "loss": 0.1731, "step": 16468 }, { "epoch": 5.07, "learning_rate": 1.0276504631080595e-05, "loss": 0.2681, "step": 16469 }, { "epoch": 5.07, "learning_rate": 1.0275508470637234e-05, "loss": 0.1642, "step": 16470 }, { "epoch": 5.07, "learning_rate": 1.0274512307457822e-05, "loss": 0.1497, "step": 16471 }, { "epoch": 5.07, "learning_rate": 1.027351614155225e-05, "loss": 0.1887, "step": 16472 }, { "epoch": 5.07, "learning_rate": 1.0272519972930406e-05, "loss": 0.1536, "step": 16473 }, { "epoch": 5.07, "learning_rate": 1.0271523801602194e-05, "loss": 0.076, "step": 16474 }, { "epoch": 5.07, "learning_rate": 1.0270527627577495e-05, "loss": 0.1467, "step": 16475 }, { "epoch": 5.07, "learning_rate": 1.026953145086621e-05, "loss": 0.2491, "step": 16476 }, { "epoch": 5.07, "learning_rate": 1.0268535271478231e-05, "loss": 0.1306, "step": 16477 }, { "epoch": 5.07, "learning_rate": 1.0267539089423445e-05, "loss": 0.1882, "step": 16478 }, { "epoch": 5.07, "learning_rate": 1.0266542904711753e-05, "loss": 0.1281, "step": 16479 }, { "epoch": 5.07, "learning_rate": 1.0265546717353041e-05, "loss": 0.1561, "step": 16480 }, { "epoch": 5.07, "learning_rate": 1.0264550527357209e-05, "loss": 0.182, "step": 16481 }, { "epoch": 5.07, "learning_rate": 1.0263554334734143e-05, "loss": 0.1538, "step": 16482 }, { "epoch": 5.07, "learning_rate": 1.0262558139493742e-05, "loss": 0.0791, "step": 16483 }, { "epoch": 5.07, "learning_rate": 1.0261561941645894e-05, "loss": 0.185, "step": 16484 }, { "epoch": 5.07, "learning_rate": 1.02605657412005e-05, "loss": 0.23, "step": 16485 }, { "epoch": 5.07, "learning_rate": 1.0259569538167444e-05, "loss": 0.1938, "step": 16486 }, { "epoch": 5.07, "learning_rate": 1.025857333255662e-05, "loss": 0.1034, "step": 16487 }, { "epoch": 5.07, "learning_rate": 1.025757712437793e-05, "loss": 0.1741, "step": 16488 }, { "epoch": 5.07, "learning_rate": 1.0256580913641261e-05, "loss": 0.2349, "step": 16489 }, { "epoch": 5.07, "learning_rate": 1.0255584700356507e-05, "loss": 0.1931, "step": 16490 }, { "epoch": 5.07, "learning_rate": 1.0254588484533561e-05, "loss": 0.1851, "step": 16491 }, { "epoch": 5.07, "learning_rate": 1.0253592266182315e-05, "loss": 0.1555, "step": 16492 }, { "epoch": 5.07, "learning_rate": 1.0252596045312666e-05, "loss": 0.1092, "step": 16493 }, { "epoch": 5.07, "learning_rate": 1.0251599821934506e-05, "loss": 0.1679, "step": 16494 }, { "epoch": 5.08, "learning_rate": 1.0250603596057727e-05, "loss": 0.185, "step": 16495 }, { "epoch": 5.08, "learning_rate": 1.0249607367692222e-05, "loss": 0.1182, "step": 16496 }, { "epoch": 5.08, "learning_rate": 1.0248611136847885e-05, "loss": 0.1077, "step": 16497 }, { "epoch": 5.08, "learning_rate": 1.0247614903534615e-05, "loss": 0.3022, "step": 16498 }, { "epoch": 5.08, "learning_rate": 1.0246618667762299e-05, "loss": 0.244, "step": 16499 }, { "epoch": 5.08, "learning_rate": 1.0245622429540832e-05, "loss": 0.2347, "step": 16500 }, { "epoch": 5.08, "learning_rate": 1.0244626188880105e-05, "loss": 0.1391, "step": 16501 }, { "epoch": 5.08, "learning_rate": 1.0243629945790017e-05, "loss": 0.147, "step": 16502 }, { "epoch": 5.08, "learning_rate": 1.024263370028046e-05, "loss": 0.1132, "step": 16503 }, { "epoch": 5.08, "learning_rate": 1.0241637452361323e-05, "loss": 0.0878, "step": 16504 }, { "epoch": 5.08, "learning_rate": 1.024064120204251e-05, "loss": 0.1624, "step": 16505 }, { "epoch": 5.08, "learning_rate": 1.02396449493339e-05, "loss": 0.1575, "step": 16506 }, { "epoch": 5.08, "learning_rate": 1.0238648694245402e-05, "loss": 0.2673, "step": 16507 }, { "epoch": 5.08, "learning_rate": 1.0237652436786897e-05, "loss": 0.1188, "step": 16508 }, { "epoch": 5.08, "learning_rate": 1.0236656176968287e-05, "loss": 0.1016, "step": 16509 }, { "epoch": 5.08, "learning_rate": 1.0235659914799465e-05, "loss": 0.2073, "step": 16510 }, { "epoch": 5.08, "learning_rate": 1.0234663650290317e-05, "loss": 0.1894, "step": 16511 }, { "epoch": 5.08, "learning_rate": 1.0233667383450748e-05, "loss": 0.1425, "step": 16512 }, { "epoch": 5.08, "learning_rate": 1.0232671114290644e-05, "loss": 0.1163, "step": 16513 }, { "epoch": 5.08, "learning_rate": 1.0231674842819901e-05, "loss": 0.1539, "step": 16514 }, { "epoch": 5.08, "learning_rate": 1.0230678569048418e-05, "loss": 0.132, "step": 16515 }, { "epoch": 5.08, "learning_rate": 1.0229682292986078e-05, "loss": 0.2287, "step": 16516 }, { "epoch": 5.08, "learning_rate": 1.0228686014642786e-05, "loss": 0.1181, "step": 16517 }, { "epoch": 5.08, "learning_rate": 1.0227689734028427e-05, "loss": 0.1733, "step": 16518 }, { "epoch": 5.08, "learning_rate": 1.02266934511529e-05, "loss": 0.1674, "step": 16519 }, { "epoch": 5.08, "learning_rate": 1.02256971660261e-05, "loss": 0.1379, "step": 16520 }, { "epoch": 5.08, "learning_rate": 1.022470087865792e-05, "loss": 0.1539, "step": 16521 }, { "epoch": 5.08, "learning_rate": 1.0223704589058253e-05, "loss": 0.2463, "step": 16522 }, { "epoch": 5.08, "learning_rate": 1.022270829723699e-05, "loss": 0.1961, "step": 16523 }, { "epoch": 5.08, "learning_rate": 1.022171200320403e-05, "loss": 0.1326, "step": 16524 }, { "epoch": 5.08, "learning_rate": 1.0220715706969264e-05, "loss": 0.2499, "step": 16525 }, { "epoch": 5.08, "learning_rate": 1.021971940854259e-05, "loss": 0.1121, "step": 16526 }, { "epoch": 5.09, "learning_rate": 1.02187231079339e-05, "loss": 0.2333, "step": 16527 }, { "epoch": 5.09, "learning_rate": 1.0217726805153085e-05, "loss": 0.1879, "step": 16528 }, { "epoch": 5.09, "learning_rate": 1.0216730500210046e-05, "loss": 0.1731, "step": 16529 }, { "epoch": 5.09, "learning_rate": 1.021573419311467e-05, "loss": 0.199, "step": 16530 }, { "epoch": 5.09, "learning_rate": 1.0214737883876858e-05, "loss": 0.1311, "step": 16531 }, { "epoch": 5.09, "learning_rate": 1.0213741572506497e-05, "loss": 0.1573, "step": 16532 }, { "epoch": 5.09, "learning_rate": 1.0212745259013488e-05, "loss": 0.1107, "step": 16533 }, { "epoch": 5.09, "learning_rate": 1.0211748943407723e-05, "loss": 0.1717, "step": 16534 }, { "epoch": 5.09, "learning_rate": 1.0210752625699094e-05, "loss": 0.2455, "step": 16535 }, { "epoch": 5.09, "learning_rate": 1.0209756305897499e-05, "loss": 0.1986, "step": 16536 }, { "epoch": 5.09, "learning_rate": 1.0208759984012826e-05, "loss": 0.1374, "step": 16537 }, { "epoch": 5.09, "learning_rate": 1.0207763660054979e-05, "loss": 0.2218, "step": 16538 }, { "epoch": 5.09, "learning_rate": 1.0206767334033845e-05, "loss": 0.1743, "step": 16539 }, { "epoch": 5.09, "learning_rate": 1.0205771005959321e-05, "loss": 0.1189, "step": 16540 }, { "epoch": 5.09, "learning_rate": 1.0204774675841302e-05, "loss": 0.1115, "step": 16541 }, { "epoch": 5.09, "learning_rate": 1.0203778343689682e-05, "loss": 0.2112, "step": 16542 }, { "epoch": 5.09, "learning_rate": 1.0202782009514354e-05, "loss": 0.2137, "step": 16543 }, { "epoch": 5.09, "learning_rate": 1.0201785673325214e-05, "loss": 0.1889, "step": 16544 }, { "epoch": 5.09, "learning_rate": 1.0200789335132157e-05, "loss": 0.1905, "step": 16545 }, { "epoch": 5.09, "learning_rate": 1.019979299494508e-05, "loss": 0.13, "step": 16546 }, { "epoch": 5.09, "learning_rate": 1.0198796652773868e-05, "loss": 0.1677, "step": 16547 }, { "epoch": 5.09, "learning_rate": 1.0197800308628425e-05, "loss": 0.2285, "step": 16548 }, { "epoch": 5.09, "learning_rate": 1.0196803962518643e-05, "loss": 0.2289, "step": 16549 }, { "epoch": 5.09, "learning_rate": 1.0195807614454416e-05, "loss": 0.0918, "step": 16550 }, { "epoch": 5.09, "learning_rate": 1.0194811264445639e-05, "loss": 0.1959, "step": 16551 }, { "epoch": 5.09, "learning_rate": 1.0193814912502206e-05, "loss": 0.1426, "step": 16552 }, { "epoch": 5.09, "learning_rate": 1.0192818558634012e-05, "loss": 0.2064, "step": 16553 }, { "epoch": 5.09, "learning_rate": 1.0191822202850952e-05, "loss": 0.1884, "step": 16554 }, { "epoch": 5.09, "learning_rate": 1.0190825845162925e-05, "loss": 0.1488, "step": 16555 }, { "epoch": 5.09, "learning_rate": 1.0189829485579816e-05, "loss": 0.0868, "step": 16556 }, { "epoch": 5.09, "learning_rate": 1.0188833124111528e-05, "loss": 0.1159, "step": 16557 }, { "epoch": 5.09, "learning_rate": 1.0187836760767954e-05, "loss": 0.1278, "step": 16558 }, { "epoch": 5.09, "learning_rate": 1.0186840395558984e-05, "loss": 0.135, "step": 16559 }, { "epoch": 5.1, "learning_rate": 1.018584402849452e-05, "loss": 0.1912, "step": 16560 }, { "epoch": 5.1, "learning_rate": 1.018484765958445e-05, "loss": 0.1931, "step": 16561 }, { "epoch": 5.1, "learning_rate": 1.0183851288838676e-05, "loss": 0.2424, "step": 16562 }, { "epoch": 5.1, "learning_rate": 1.0182854916267087e-05, "loss": 0.1638, "step": 16563 }, { "epoch": 5.1, "learning_rate": 1.0181858541879582e-05, "loss": 0.138, "step": 16564 }, { "epoch": 5.1, "learning_rate": 1.0180862165686054e-05, "loss": 0.2127, "step": 16565 }, { "epoch": 5.1, "learning_rate": 1.0179865787696397e-05, "loss": 0.1437, "step": 16566 }, { "epoch": 5.1, "learning_rate": 1.0178869407920507e-05, "loss": 0.0362, "step": 16567 }, { "epoch": 5.1, "learning_rate": 1.0177873026368277e-05, "loss": 0.115, "step": 16568 }, { "epoch": 5.1, "learning_rate": 1.0176876643049607e-05, "loss": 0.1262, "step": 16569 }, { "epoch": 5.1, "learning_rate": 1.017588025797439e-05, "loss": 0.2615, "step": 16570 }, { "epoch": 5.1, "learning_rate": 1.0174883871152517e-05, "loss": 0.2124, "step": 16571 }, { "epoch": 5.1, "learning_rate": 1.0173887482593886e-05, "loss": 0.1377, "step": 16572 }, { "epoch": 5.1, "learning_rate": 1.0172891092308393e-05, "loss": 0.1711, "step": 16573 }, { "epoch": 5.1, "learning_rate": 1.0171894700305931e-05, "loss": 0.1323, "step": 16574 }, { "epoch": 5.1, "learning_rate": 1.0170898306596397e-05, "loss": 0.187, "step": 16575 }, { "epoch": 5.1, "learning_rate": 1.0169901911189688e-05, "loss": 0.1906, "step": 16576 }, { "epoch": 5.1, "learning_rate": 1.0168905514095692e-05, "loss": 0.0998, "step": 16577 }, { "epoch": 5.1, "learning_rate": 1.0167909115324311e-05, "loss": 0.1755, "step": 16578 }, { "epoch": 5.1, "learning_rate": 1.0166912714885437e-05, "loss": 0.208, "step": 16579 }, { "epoch": 5.1, "learning_rate": 1.0165916312788967e-05, "loss": 0.1479, "step": 16580 }, { "epoch": 5.1, "learning_rate": 1.0164919909044795e-05, "loss": 0.2863, "step": 16581 }, { "epoch": 5.1, "learning_rate": 1.0163923503662813e-05, "loss": 0.1854, "step": 16582 }, { "epoch": 5.1, "learning_rate": 1.0162927096652923e-05, "loss": 0.2235, "step": 16583 }, { "epoch": 5.1, "learning_rate": 1.0161930688025018e-05, "loss": 0.2302, "step": 16584 }, { "epoch": 5.1, "learning_rate": 1.016093427778899e-05, "loss": 0.1407, "step": 16585 }, { "epoch": 5.1, "learning_rate": 1.0159937865954737e-05, "loss": 0.2033, "step": 16586 }, { "epoch": 5.1, "learning_rate": 1.0158941452532149e-05, "loss": 0.1964, "step": 16587 }, { "epoch": 5.1, "learning_rate": 1.0157945037531132e-05, "loss": 0.2363, "step": 16588 }, { "epoch": 5.1, "learning_rate": 1.0156948620961574e-05, "loss": 0.2467, "step": 16589 }, { "epoch": 5.1, "learning_rate": 1.015595220283337e-05, "loss": 0.2106, "step": 16590 }, { "epoch": 5.1, "learning_rate": 1.0154955783156418e-05, "loss": 0.1837, "step": 16591 }, { "epoch": 5.11, "learning_rate": 1.0153959361940608e-05, "loss": 0.0902, "step": 16592 }, { "epoch": 5.11, "learning_rate": 1.0152962939195845e-05, "loss": 0.1768, "step": 16593 }, { "epoch": 5.11, "learning_rate": 1.0151966514932017e-05, "loss": 0.2075, "step": 16594 }, { "epoch": 5.11, "learning_rate": 1.0150970089159023e-05, "loss": 0.1179, "step": 16595 }, { "epoch": 5.11, "learning_rate": 1.0149973661886757e-05, "loss": 0.1178, "step": 16596 }, { "epoch": 5.11, "learning_rate": 1.014897723312511e-05, "loss": 0.2127, "step": 16597 }, { "epoch": 5.11, "learning_rate": 1.0147980802883986e-05, "loss": 0.1302, "step": 16598 }, { "epoch": 5.11, "learning_rate": 1.0146984371173272e-05, "loss": 0.2089, "step": 16599 }, { "epoch": 5.11, "learning_rate": 1.0145987938002873e-05, "loss": 0.1196, "step": 16600 }, { "epoch": 5.11, "learning_rate": 1.0144991503382676e-05, "loss": 0.2628, "step": 16601 }, { "epoch": 5.11, "learning_rate": 1.0143995067322581e-05, "loss": 0.1032, "step": 16602 }, { "epoch": 5.11, "learning_rate": 1.0142998629832484e-05, "loss": 0.2006, "step": 16603 }, { "epoch": 5.11, "learning_rate": 1.0142002190922273e-05, "loss": 0.1754, "step": 16604 }, { "epoch": 5.11, "learning_rate": 1.0141005750601853e-05, "loss": 0.0894, "step": 16605 }, { "epoch": 5.11, "learning_rate": 1.0140009308881117e-05, "loss": 0.1926, "step": 16606 }, { "epoch": 5.11, "learning_rate": 1.013901286576996e-05, "loss": 0.2001, "step": 16607 }, { "epoch": 5.11, "learning_rate": 1.0138016421278276e-05, "loss": 0.0925, "step": 16608 }, { "epoch": 5.11, "learning_rate": 1.013701997541596e-05, "loss": 0.1394, "step": 16609 }, { "epoch": 5.11, "learning_rate": 1.013602352819291e-05, "loss": 0.1682, "step": 16610 }, { "epoch": 5.11, "learning_rate": 1.0135027079619024e-05, "loss": 0.0915, "step": 16611 }, { "epoch": 5.11, "learning_rate": 1.0134030629704192e-05, "loss": 0.1504, "step": 16612 }, { "epoch": 5.11, "learning_rate": 1.0133034178458315e-05, "loss": 0.0783, "step": 16613 }, { "epoch": 5.11, "learning_rate": 1.0132037725891281e-05, "loss": 0.2286, "step": 16614 }, { "epoch": 5.11, "learning_rate": 1.0131041272012993e-05, "loss": 0.3156, "step": 16615 }, { "epoch": 5.11, "learning_rate": 1.0130044816833346e-05, "loss": 0.2552, "step": 16616 }, { "epoch": 5.11, "learning_rate": 1.0129048360362233e-05, "loss": 0.2192, "step": 16617 }, { "epoch": 5.11, "learning_rate": 1.012805190260955e-05, "loss": 0.1366, "step": 16618 }, { "epoch": 5.11, "learning_rate": 1.0127055443585195e-05, "loss": 0.1977, "step": 16619 }, { "epoch": 5.11, "learning_rate": 1.012605898329906e-05, "loss": 0.2238, "step": 16620 }, { "epoch": 5.11, "learning_rate": 1.0125062521761047e-05, "loss": 0.1371, "step": 16621 }, { "epoch": 5.11, "learning_rate": 1.0124066058981046e-05, "loss": 0.1744, "step": 16622 }, { "epoch": 5.11, "learning_rate": 1.0123069594968952e-05, "loss": 0.1586, "step": 16623 }, { "epoch": 5.11, "learning_rate": 1.0122073129734668e-05, "loss": 0.2008, "step": 16624 }, { "epoch": 5.12, "learning_rate": 1.0121076663288084e-05, "loss": 0.1605, "step": 16625 }, { "epoch": 5.12, "learning_rate": 1.0120080195639097e-05, "loss": 0.1508, "step": 16626 }, { "epoch": 5.12, "learning_rate": 1.0119083726797603e-05, "loss": 0.117, "step": 16627 }, { "epoch": 5.12, "learning_rate": 1.0118087256773494e-05, "loss": 0.1261, "step": 16628 }, { "epoch": 5.12, "learning_rate": 1.0117090785576676e-05, "loss": 0.1842, "step": 16629 }, { "epoch": 5.12, "learning_rate": 1.0116094313217032e-05, "loss": 0.1895, "step": 16630 }, { "epoch": 5.12, "learning_rate": 1.0115097839704472e-05, "loss": 0.1781, "step": 16631 }, { "epoch": 5.12, "learning_rate": 1.0114101365048879e-05, "loss": 0.1629, "step": 16632 }, { "epoch": 5.12, "learning_rate": 1.0113104889260154e-05, "loss": 0.2044, "step": 16633 }, { "epoch": 5.12, "learning_rate": 1.0112108412348196e-05, "loss": 0.186, "step": 16634 }, { "epoch": 5.12, "learning_rate": 1.0111111934322895e-05, "loss": 0.1511, "step": 16635 }, { "epoch": 5.12, "learning_rate": 1.0110115455194157e-05, "loss": 0.1765, "step": 16636 }, { "epoch": 5.12, "learning_rate": 1.0109118974971863e-05, "loss": 0.2108, "step": 16637 }, { "epoch": 5.12, "learning_rate": 1.0108122493665922e-05, "loss": 0.099, "step": 16638 }, { "epoch": 5.12, "learning_rate": 1.010712601128622e-05, "loss": 0.1034, "step": 16639 }, { "epoch": 5.12, "learning_rate": 1.0106129527842661e-05, "loss": 0.1925, "step": 16640 }, { "epoch": 5.12, "learning_rate": 1.0105133043345142e-05, "loss": 0.1784, "step": 16641 }, { "epoch": 5.12, "learning_rate": 1.010413655780355e-05, "loss": 0.1475, "step": 16642 }, { "epoch": 5.12, "learning_rate": 1.0103140071227786e-05, "loss": 0.1234, "step": 16643 }, { "epoch": 5.12, "learning_rate": 1.0102143583627748e-05, "loss": 0.2171, "step": 16644 }, { "epoch": 5.12, "learning_rate": 1.0101147095013328e-05, "loss": 0.3476, "step": 16645 }, { "epoch": 5.12, "learning_rate": 1.0100150605394427e-05, "loss": 0.262, "step": 16646 }, { "epoch": 5.12, "learning_rate": 1.0099154114780936e-05, "loss": 0.2306, "step": 16647 }, { "epoch": 5.12, "learning_rate": 1.0098157623182754e-05, "loss": 0.1679, "step": 16648 }, { "epoch": 5.12, "learning_rate": 1.0097161130609774e-05, "loss": 0.1544, "step": 16649 }, { "epoch": 5.12, "learning_rate": 1.0096164637071897e-05, "loss": 0.1597, "step": 16650 }, { "epoch": 5.12, "learning_rate": 1.0095168142579017e-05, "loss": 0.1789, "step": 16651 }, { "epoch": 5.12, "learning_rate": 1.0094171647141028e-05, "loss": 0.1166, "step": 16652 }, { "epoch": 5.12, "learning_rate": 1.0093175150767831e-05, "loss": 0.1468, "step": 16653 }, { "epoch": 5.12, "learning_rate": 1.0092178653469313e-05, "loss": 0.0933, "step": 16654 }, { "epoch": 5.12, "learning_rate": 1.0091182155255381e-05, "loss": 0.1267, "step": 16655 }, { "epoch": 5.12, "learning_rate": 1.0090185656135925e-05, "loss": 0.1647, "step": 16656 }, { "epoch": 5.13, "learning_rate": 1.0089189156120842e-05, "loss": 0.162, "step": 16657 }, { "epoch": 5.13, "learning_rate": 1.0088192655220028e-05, "loss": 0.13, "step": 16658 }, { "epoch": 5.13, "learning_rate": 1.0087196153443379e-05, "loss": 0.1053, "step": 16659 }, { "epoch": 5.13, "learning_rate": 1.0086199650800795e-05, "loss": 0.2228, "step": 16660 }, { "epoch": 5.13, "learning_rate": 1.0085203147302166e-05, "loss": 0.1502, "step": 16661 }, { "epoch": 5.13, "learning_rate": 1.0084206642957393e-05, "loss": 0.2065, "step": 16662 }, { "epoch": 5.13, "learning_rate": 1.008321013777637e-05, "loss": 0.1259, "step": 16663 }, { "epoch": 5.13, "learning_rate": 1.008221363176899e-05, "loss": 0.1606, "step": 16664 }, { "epoch": 5.13, "learning_rate": 1.0081217124945157e-05, "loss": 0.3147, "step": 16665 }, { "epoch": 5.13, "learning_rate": 1.0080220617314764e-05, "loss": 0.2323, "step": 16666 }, { "epoch": 5.13, "learning_rate": 1.0079224108887704e-05, "loss": 0.2088, "step": 16667 }, { "epoch": 5.13, "learning_rate": 1.0078227599673875e-05, "loss": 0.1149, "step": 16668 }, { "epoch": 5.13, "learning_rate": 1.0077231089683175e-05, "loss": 0.194, "step": 16669 }, { "epoch": 5.13, "learning_rate": 1.00762345789255e-05, "loss": 0.186, "step": 16670 }, { "epoch": 5.13, "learning_rate": 1.0075238067410745e-05, "loss": 0.1628, "step": 16671 }, { "epoch": 5.13, "learning_rate": 1.0074241555148805e-05, "loss": 0.2404, "step": 16672 }, { "epoch": 5.13, "learning_rate": 1.0073245042149578e-05, "loss": 0.1677, "step": 16673 }, { "epoch": 5.13, "learning_rate": 1.007224852842296e-05, "loss": 0.1691, "step": 16674 }, { "epoch": 5.13, "learning_rate": 1.0071252013978852e-05, "loss": 0.2539, "step": 16675 }, { "epoch": 5.13, "learning_rate": 1.0070255498827141e-05, "loss": 0.2047, "step": 16676 }, { "epoch": 5.13, "learning_rate": 1.006925898297773e-05, "loss": 0.1514, "step": 16677 }, { "epoch": 5.13, "learning_rate": 1.0068262466440513e-05, "loss": 0.255, "step": 16678 }, { "epoch": 5.13, "learning_rate": 1.0067265949225387e-05, "loss": 0.1933, "step": 16679 }, { "epoch": 5.13, "learning_rate": 1.0066269431342246e-05, "loss": 0.2177, "step": 16680 }, { "epoch": 5.13, "learning_rate": 1.0065272912800993e-05, "loss": 0.1218, "step": 16681 }, { "epoch": 5.13, "learning_rate": 1.0064276393611517e-05, "loss": 0.1553, "step": 16682 }, { "epoch": 5.13, "learning_rate": 1.0063279873783718e-05, "loss": 0.1885, "step": 16683 }, { "epoch": 5.13, "learning_rate": 1.006228335332749e-05, "loss": 0.1621, "step": 16684 }, { "epoch": 5.13, "learning_rate": 1.006128683225273e-05, "loss": 0.1864, "step": 16685 }, { "epoch": 5.13, "learning_rate": 1.006029031056934e-05, "loss": 0.1967, "step": 16686 }, { "epoch": 5.13, "learning_rate": 1.0059293788287207e-05, "loss": 0.2676, "step": 16687 }, { "epoch": 5.13, "learning_rate": 1.0058297265416234e-05, "loss": 0.1806, "step": 16688 }, { "epoch": 5.13, "learning_rate": 1.0057300741966316e-05, "loss": 0.1547, "step": 16689 }, { "epoch": 5.14, "learning_rate": 1.0056304217947347e-05, "loss": 0.1925, "step": 16690 }, { "epoch": 5.14, "learning_rate": 1.005530769336923e-05, "loss": 0.1606, "step": 16691 }, { "epoch": 5.14, "learning_rate": 1.0054311168241851e-05, "loss": 0.2025, "step": 16692 }, { "epoch": 5.14, "learning_rate": 1.0053314642575114e-05, "loss": 0.2703, "step": 16693 }, { "epoch": 5.14, "learning_rate": 1.0052318116378915e-05, "loss": 0.0786, "step": 16694 }, { "epoch": 5.14, "learning_rate": 1.0051321589663146e-05, "loss": 0.2181, "step": 16695 }, { "epoch": 5.14, "learning_rate": 1.0050325062437708e-05, "loss": 0.1672, "step": 16696 }, { "epoch": 5.14, "learning_rate": 1.0049328534712495e-05, "loss": 0.2148, "step": 16697 }, { "epoch": 5.14, "learning_rate": 1.0048332006497406e-05, "loss": 0.2142, "step": 16698 }, { "epoch": 5.14, "learning_rate": 1.0047335477802331e-05, "loss": 0.2221, "step": 16699 }, { "epoch": 5.14, "learning_rate": 1.0046338948637175e-05, "loss": 0.1914, "step": 16700 }, { "epoch": 5.14, "learning_rate": 1.0045342419011832e-05, "loss": 0.1542, "step": 16701 }, { "epoch": 5.14, "learning_rate": 1.0044345888936195e-05, "loss": 0.1834, "step": 16702 }, { "epoch": 5.14, "learning_rate": 1.0043349358420162e-05, "loss": 0.2042, "step": 16703 }, { "epoch": 5.14, "learning_rate": 1.004235282747363e-05, "loss": 0.2999, "step": 16704 }, { "epoch": 5.14, "learning_rate": 1.0041356296106494e-05, "loss": 0.1335, "step": 16705 }, { "epoch": 5.14, "learning_rate": 1.0040359764328656e-05, "loss": 0.2185, "step": 16706 }, { "epoch": 5.14, "learning_rate": 1.0039363232150006e-05, "loss": 0.1519, "step": 16707 }, { "epoch": 5.14, "learning_rate": 1.0038366699580441e-05, "loss": 0.2078, "step": 16708 }, { "epoch": 5.14, "learning_rate": 1.003737016662986e-05, "loss": 0.2043, "step": 16709 }, { "epoch": 5.14, "learning_rate": 1.0036373633308161e-05, "loss": 0.1346, "step": 16710 }, { "epoch": 5.14, "learning_rate": 1.0035377099625236e-05, "loss": 0.1939, "step": 16711 }, { "epoch": 5.14, "learning_rate": 1.0034380565590987e-05, "loss": 0.1023, "step": 16712 }, { "epoch": 5.14, "learning_rate": 1.0033384031215304e-05, "loss": 0.1652, "step": 16713 }, { "epoch": 5.14, "learning_rate": 1.003238749650809e-05, "loss": 0.1635, "step": 16714 }, { "epoch": 5.14, "learning_rate": 1.0031390961479234e-05, "loss": 0.1955, "step": 16715 }, { "epoch": 5.14, "learning_rate": 1.0030394426138639e-05, "loss": 0.1527, "step": 16716 }, { "epoch": 5.14, "learning_rate": 1.0029397890496203e-05, "loss": 0.1721, "step": 16717 }, { "epoch": 5.14, "learning_rate": 1.0028401354561812e-05, "loss": 0.1185, "step": 16718 }, { "epoch": 5.14, "learning_rate": 1.0027404818345374e-05, "loss": 0.2177, "step": 16719 }, { "epoch": 5.14, "learning_rate": 1.0026408281856779e-05, "loss": 0.2021, "step": 16720 }, { "epoch": 5.14, "learning_rate": 1.0025411745105924e-05, "loss": 0.165, "step": 16721 }, { "epoch": 5.15, "learning_rate": 1.0024415208102713e-05, "loss": 0.2083, "step": 16722 }, { "epoch": 5.15, "learning_rate": 1.002341867085703e-05, "loss": 0.1064, "step": 16723 }, { "epoch": 5.15, "learning_rate": 1.0022422133378783e-05, "loss": 0.1558, "step": 16724 }, { "epoch": 5.15, "learning_rate": 1.002142559567786e-05, "loss": 0.0979, "step": 16725 }, { "epoch": 5.15, "learning_rate": 1.002042905776416e-05, "loss": 0.1825, "step": 16726 }, { "epoch": 5.15, "learning_rate": 1.0019432519647585e-05, "loss": 0.1676, "step": 16727 }, { "epoch": 5.15, "learning_rate": 1.0018435981338023e-05, "loss": 0.1372, "step": 16728 }, { "epoch": 5.15, "learning_rate": 1.0017439442845378e-05, "loss": 0.2282, "step": 16729 }, { "epoch": 5.15, "learning_rate": 1.0016442904179539e-05, "loss": 0.0922, "step": 16730 }, { "epoch": 5.15, "learning_rate": 1.0015446365350413e-05, "loss": 0.1971, "step": 16731 }, { "epoch": 5.15, "learning_rate": 1.0014449826367889e-05, "loss": 0.2346, "step": 16732 }, { "epoch": 5.15, "learning_rate": 1.001345328724186e-05, "loss": 0.1868, "step": 16733 }, { "epoch": 5.15, "learning_rate": 1.0012456747982231e-05, "loss": 0.1026, "step": 16734 }, { "epoch": 5.15, "learning_rate": 1.0011460208598893e-05, "loss": 0.1232, "step": 16735 }, { "epoch": 5.15, "learning_rate": 1.0010463669101746e-05, "loss": 0.156, "step": 16736 }, { "epoch": 5.15, "learning_rate": 1.0009467129500685e-05, "loss": 0.1226, "step": 16737 }, { "epoch": 5.15, "learning_rate": 1.0008470589805608e-05, "loss": 0.1485, "step": 16738 }, { "epoch": 5.15, "learning_rate": 1.0007474050026409e-05, "loss": 0.1168, "step": 16739 }, { "epoch": 5.15, "learning_rate": 1.0006477510172984e-05, "loss": 0.0979, "step": 16740 }, { "epoch": 5.15, "learning_rate": 1.0005480970255234e-05, "loss": 0.238, "step": 16741 }, { "epoch": 5.15, "learning_rate": 1.0004484430283052e-05, "loss": 0.1471, "step": 16742 }, { "epoch": 5.15, "learning_rate": 1.0003487890266337e-05, "loss": 0.2474, "step": 16743 }, { "epoch": 5.15, "learning_rate": 1.0002491350214983e-05, "loss": 0.1676, "step": 16744 }, { "epoch": 5.15, "learning_rate": 1.0001494810138885e-05, "loss": 0.2077, "step": 16745 }, { "epoch": 5.15, "learning_rate": 1.0000498270047946e-05, "loss": 0.1163, "step": 16746 }, { "epoch": 5.15, "learning_rate": 9.999501729952056e-06, "loss": 0.1375, "step": 16747 }, { "epoch": 5.15, "learning_rate": 9.998505189861115e-06, "loss": 0.1289, "step": 16748 }, { "epoch": 5.15, "learning_rate": 9.997508649785019e-06, "loss": 0.2214, "step": 16749 }, { "epoch": 5.15, "learning_rate": 9.996512109733668e-06, "loss": 0.1741, "step": 16750 }, { "epoch": 5.15, "learning_rate": 9.99551556971695e-06, "loss": 0.2284, "step": 16751 }, { "epoch": 5.15, "learning_rate": 9.994519029744767e-06, "loss": 0.1216, "step": 16752 }, { "epoch": 5.15, "learning_rate": 9.993522489827016e-06, "loss": 0.2105, "step": 16753 }, { "epoch": 5.15, "learning_rate": 9.992525949973593e-06, "loss": 0.1572, "step": 16754 }, { "epoch": 5.16, "learning_rate": 9.991529410194397e-06, "loss": 0.2655, "step": 16755 }, { "epoch": 5.16, "learning_rate": 9.990532870499317e-06, "loss": 0.0924, "step": 16756 }, { "epoch": 5.16, "learning_rate": 9.989536330898256e-06, "loss": 0.1013, "step": 16757 }, { "epoch": 5.16, "learning_rate": 9.988539791401107e-06, "loss": 0.1888, "step": 16758 }, { "epoch": 5.16, "learning_rate": 9.987543252017772e-06, "loss": 0.1572, "step": 16759 }, { "epoch": 5.16, "learning_rate": 9.986546712758145e-06, "loss": 0.1577, "step": 16760 }, { "epoch": 5.16, "learning_rate": 9.985550173632115e-06, "loss": 0.1176, "step": 16761 }, { "epoch": 5.16, "learning_rate": 9.984553634649592e-06, "loss": 0.1559, "step": 16762 }, { "epoch": 5.16, "learning_rate": 9.983557095820461e-06, "loss": 0.1417, "step": 16763 }, { "epoch": 5.16, "learning_rate": 9.982560557154624e-06, "loss": 0.2137, "step": 16764 }, { "epoch": 5.16, "learning_rate": 9.981564018661982e-06, "loss": 0.2373, "step": 16765 }, { "epoch": 5.16, "learning_rate": 9.980567480352417e-06, "loss": 0.1695, "step": 16766 }, { "epoch": 5.16, "learning_rate": 9.979570942235842e-06, "loss": 0.1669, "step": 16767 }, { "epoch": 5.16, "learning_rate": 9.978574404322143e-06, "loss": 0.1925, "step": 16768 }, { "epoch": 5.16, "learning_rate": 9.977577866621222e-06, "loss": 0.1851, "step": 16769 }, { "epoch": 5.16, "learning_rate": 9.976581329142973e-06, "loss": 0.1477, "step": 16770 }, { "epoch": 5.16, "learning_rate": 9.97558479189729e-06, "loss": 0.2018, "step": 16771 }, { "epoch": 5.16, "learning_rate": 9.974588254894077e-06, "loss": 0.1071, "step": 16772 }, { "epoch": 5.16, "learning_rate": 9.973591718143223e-06, "loss": 0.1569, "step": 16773 }, { "epoch": 5.16, "learning_rate": 9.97259518165463e-06, "loss": 0.1403, "step": 16774 }, { "epoch": 5.16, "learning_rate": 9.97159864543819e-06, "loss": 0.171, "step": 16775 }, { "epoch": 5.16, "learning_rate": 9.9706021095038e-06, "loss": 0.1257, "step": 16776 }, { "epoch": 5.16, "learning_rate": 9.969605573861365e-06, "loss": 0.2018, "step": 16777 }, { "epoch": 5.16, "learning_rate": 9.968609038520766e-06, "loss": 0.2366, "step": 16778 }, { "epoch": 5.16, "learning_rate": 9.967612503491915e-06, "loss": 0.1595, "step": 16779 }, { "epoch": 5.16, "learning_rate": 9.966615968784699e-06, "loss": 0.1642, "step": 16780 }, { "epoch": 5.16, "learning_rate": 9.965619434409014e-06, "loss": 0.1974, "step": 16781 }, { "epoch": 5.16, "learning_rate": 9.964622900374767e-06, "loss": 0.1134, "step": 16782 }, { "epoch": 5.16, "learning_rate": 9.963626366691839e-06, "loss": 0.204, "step": 16783 }, { "epoch": 5.16, "learning_rate": 9.962629833370143e-06, "loss": 0.1589, "step": 16784 }, { "epoch": 5.16, "learning_rate": 9.961633300419562e-06, "loss": 0.2111, "step": 16785 }, { "epoch": 5.16, "learning_rate": 9.960636767849998e-06, "loss": 0.1915, "step": 16786 }, { "epoch": 5.17, "learning_rate": 9.959640235671347e-06, "loss": 0.286, "step": 16787 }, { "epoch": 5.17, "learning_rate": 9.958643703893506e-06, "loss": 0.1968, "step": 16788 }, { "epoch": 5.17, "learning_rate": 9.957647172526374e-06, "loss": 0.2283, "step": 16789 }, { "epoch": 5.17, "learning_rate": 9.956650641579843e-06, "loss": 0.0887, "step": 16790 }, { "epoch": 5.17, "learning_rate": 9.955654111063808e-06, "loss": 0.0856, "step": 16791 }, { "epoch": 5.17, "learning_rate": 9.954657580988171e-06, "loss": 0.1566, "step": 16792 }, { "epoch": 5.17, "learning_rate": 9.953661051362825e-06, "loss": 0.1456, "step": 16793 }, { "epoch": 5.17, "learning_rate": 9.95266452219767e-06, "loss": 0.2023, "step": 16794 }, { "epoch": 5.17, "learning_rate": 9.951667993502599e-06, "loss": 0.1936, "step": 16795 }, { "epoch": 5.17, "learning_rate": 9.950671465287509e-06, "loss": 0.1716, "step": 16796 }, { "epoch": 5.17, "learning_rate": 9.949674937562295e-06, "loss": 0.1925, "step": 16797 }, { "epoch": 5.17, "learning_rate": 9.948678410336854e-06, "loss": 0.1557, "step": 16798 }, { "epoch": 5.17, "learning_rate": 9.947681883621089e-06, "loss": 0.159, "step": 16799 }, { "epoch": 5.17, "learning_rate": 9.946685357424891e-06, "loss": 0.1657, "step": 16800 }, { "epoch": 5.17, "learning_rate": 9.945688831758152e-06, "loss": 0.171, "step": 16801 }, { "epoch": 5.17, "learning_rate": 9.944692306630775e-06, "loss": 0.1833, "step": 16802 }, { "epoch": 5.17, "learning_rate": 9.943695782052654e-06, "loss": 0.1492, "step": 16803 }, { "epoch": 5.17, "learning_rate": 9.942699258033686e-06, "loss": 0.1782, "step": 16804 }, { "epoch": 5.17, "learning_rate": 9.941702734583771e-06, "loss": 0.1507, "step": 16805 }, { "epoch": 5.17, "learning_rate": 9.940706211712794e-06, "loss": 0.1776, "step": 16806 }, { "epoch": 5.17, "learning_rate": 9.939709689430665e-06, "loss": 0.2856, "step": 16807 }, { "epoch": 5.17, "learning_rate": 9.93871316774727e-06, "loss": 0.1557, "step": 16808 }, { "epoch": 5.17, "learning_rate": 9.937716646672513e-06, "loss": 0.2039, "step": 16809 }, { "epoch": 5.17, "learning_rate": 9.936720126216289e-06, "loss": 0.1882, "step": 16810 }, { "epoch": 5.17, "learning_rate": 9.935723606388486e-06, "loss": 0.2263, "step": 16811 }, { "epoch": 5.17, "learning_rate": 9.934727087199012e-06, "loss": 0.2368, "step": 16812 }, { "epoch": 5.17, "learning_rate": 9.933730568657754e-06, "loss": 0.1472, "step": 16813 }, { "epoch": 5.17, "learning_rate": 9.932734050774616e-06, "loss": 0.1467, "step": 16814 }, { "epoch": 5.17, "learning_rate": 9.931737533559492e-06, "loss": 0.2024, "step": 16815 }, { "epoch": 5.17, "learning_rate": 9.930741017022272e-06, "loss": 0.1807, "step": 16816 }, { "epoch": 5.17, "learning_rate": 9.929744501172864e-06, "loss": 0.1297, "step": 16817 }, { "epoch": 5.17, "learning_rate": 9.928747986021153e-06, "loss": 0.1452, "step": 16818 }, { "epoch": 5.17, "learning_rate": 9.927751471577041e-06, "loss": 0.1322, "step": 16819 }, { "epoch": 5.18, "learning_rate": 9.926754957850426e-06, "loss": 0.1449, "step": 16820 }, { "epoch": 5.18, "learning_rate": 9.925758444851196e-06, "loss": 0.1007, "step": 16821 }, { "epoch": 5.18, "learning_rate": 9.92476193258926e-06, "loss": 0.1528, "step": 16822 }, { "epoch": 5.18, "learning_rate": 9.923765421074504e-06, "loss": 0.136, "step": 16823 }, { "epoch": 5.18, "learning_rate": 9.922768910316829e-06, "loss": 0.1394, "step": 16824 }, { "epoch": 5.18, "learning_rate": 9.921772400326128e-06, "loss": 0.1789, "step": 16825 }, { "epoch": 5.18, "learning_rate": 9.9207758911123e-06, "loss": 0.1477, "step": 16826 }, { "epoch": 5.18, "learning_rate": 9.919779382685242e-06, "loss": 0.1852, "step": 16827 }, { "epoch": 5.18, "learning_rate": 9.918782875054843e-06, "loss": 0.1589, "step": 16828 }, { "epoch": 5.18, "learning_rate": 9.917786368231011e-06, "loss": 0.194, "step": 16829 }, { "epoch": 5.18, "learning_rate": 9.916789862223636e-06, "loss": 0.1896, "step": 16830 }, { "epoch": 5.18, "learning_rate": 9.91579335704261e-06, "loss": 0.192, "step": 16831 }, { "epoch": 5.18, "learning_rate": 9.914796852697839e-06, "loss": 0.1083, "step": 16832 }, { "epoch": 5.18, "learning_rate": 9.913800349199207e-06, "loss": 0.1498, "step": 16833 }, { "epoch": 5.18, "learning_rate": 9.912803846556623e-06, "loss": 0.204, "step": 16834 }, { "epoch": 5.18, "learning_rate": 9.911807344779975e-06, "loss": 0.1634, "step": 16835 }, { "epoch": 5.18, "learning_rate": 9.910810843879161e-06, "loss": 0.262, "step": 16836 }, { "epoch": 5.18, "learning_rate": 9.909814343864078e-06, "loss": 0.2179, "step": 16837 }, { "epoch": 5.18, "learning_rate": 9.908817844744619e-06, "loss": 0.198, "step": 16838 }, { "epoch": 5.18, "learning_rate": 9.907821346530689e-06, "loss": 0.202, "step": 16839 }, { "epoch": 5.18, "learning_rate": 9.906824849232176e-06, "loss": 0.2096, "step": 16840 }, { "epoch": 5.18, "learning_rate": 9.905828352858973e-06, "loss": 0.2036, "step": 16841 }, { "epoch": 5.18, "learning_rate": 9.904831857420985e-06, "loss": 0.2083, "step": 16842 }, { "epoch": 5.18, "learning_rate": 9.903835362928103e-06, "loss": 0.1063, "step": 16843 }, { "epoch": 5.18, "learning_rate": 9.90283886939023e-06, "loss": 0.1681, "step": 16844 }, { "epoch": 5.18, "learning_rate": 9.90184237681725e-06, "loss": 0.1995, "step": 16845 }, { "epoch": 5.18, "learning_rate": 9.900845885219068e-06, "loss": 0.1464, "step": 16846 }, { "epoch": 5.18, "learning_rate": 9.899849394605578e-06, "loss": 0.0996, "step": 16847 }, { "epoch": 5.18, "learning_rate": 9.898852904986674e-06, "loss": 0.1736, "step": 16848 }, { "epoch": 5.18, "learning_rate": 9.897856416372254e-06, "loss": 0.1354, "step": 16849 }, { "epoch": 5.18, "learning_rate": 9.896859928772214e-06, "loss": 0.147, "step": 16850 }, { "epoch": 5.18, "learning_rate": 9.895863442196454e-06, "loss": 0.1925, "step": 16851 }, { "epoch": 5.19, "learning_rate": 9.894866956654863e-06, "loss": 0.1937, "step": 16852 }, { "epoch": 5.19, "learning_rate": 9.893870472157339e-06, "loss": 0.2153, "step": 16853 }, { "epoch": 5.19, "learning_rate": 9.892873988713781e-06, "loss": 0.1468, "step": 16854 }, { "epoch": 5.19, "learning_rate": 9.89187750633408e-06, "loss": 0.18, "step": 16855 }, { "epoch": 5.19, "learning_rate": 9.89088102502814e-06, "loss": 0.0931, "step": 16856 }, { "epoch": 5.19, "learning_rate": 9.88988454480585e-06, "loss": 0.1557, "step": 16857 }, { "epoch": 5.19, "learning_rate": 9.888888065677105e-06, "loss": 0.1058, "step": 16858 }, { "epoch": 5.19, "learning_rate": 9.887891587651807e-06, "loss": 0.1728, "step": 16859 }, { "epoch": 5.19, "learning_rate": 9.886895110739846e-06, "loss": 0.1994, "step": 16860 }, { "epoch": 5.19, "learning_rate": 9.885898634951126e-06, "loss": 0.3056, "step": 16861 }, { "epoch": 5.19, "learning_rate": 9.884902160295533e-06, "loss": 0.261, "step": 16862 }, { "epoch": 5.19, "learning_rate": 9.88390568678297e-06, "loss": 0.1416, "step": 16863 }, { "epoch": 5.19, "learning_rate": 9.882909214423329e-06, "loss": 0.2354, "step": 16864 }, { "epoch": 5.19, "learning_rate": 9.881912743226505e-06, "loss": 0.1205, "step": 16865 }, { "epoch": 5.19, "learning_rate": 9.8809162732024e-06, "loss": 0.2406, "step": 16866 }, { "epoch": 5.19, "learning_rate": 9.879919804360908e-06, "loss": 0.1841, "step": 16867 }, { "epoch": 5.19, "learning_rate": 9.87892333671192e-06, "loss": 0.1954, "step": 16868 }, { "epoch": 5.19, "learning_rate": 9.877926870265335e-06, "loss": 0.2121, "step": 16869 }, { "epoch": 5.19, "learning_rate": 9.876930405031047e-06, "loss": 0.149, "step": 16870 }, { "epoch": 5.19, "learning_rate": 9.875933941018956e-06, "loss": 0.1721, "step": 16871 }, { "epoch": 5.19, "learning_rate": 9.874937478238958e-06, "loss": 0.1317, "step": 16872 }, { "epoch": 5.19, "learning_rate": 9.873941016700941e-06, "loss": 0.1289, "step": 16873 }, { "epoch": 5.19, "learning_rate": 9.872944556414808e-06, "loss": 0.1864, "step": 16874 }, { "epoch": 5.19, "learning_rate": 9.871948097390451e-06, "loss": 0.073, "step": 16875 }, { "epoch": 5.19, "learning_rate": 9.870951639637769e-06, "loss": 0.175, "step": 16876 }, { "epoch": 5.19, "learning_rate": 9.869955183166659e-06, "loss": 0.1388, "step": 16877 }, { "epoch": 5.19, "learning_rate": 9.868958727987007e-06, "loss": 0.192, "step": 16878 }, { "epoch": 5.19, "learning_rate": 9.867962274108722e-06, "loss": 0.2692, "step": 16879 }, { "epoch": 5.19, "learning_rate": 9.86696582154169e-06, "loss": 0.1236, "step": 16880 }, { "epoch": 5.19, "learning_rate": 9.86596937029581e-06, "loss": 0.1861, "step": 16881 }, { "epoch": 5.19, "learning_rate": 9.864972920380982e-06, "loss": 0.1867, "step": 16882 }, { "epoch": 5.19, "learning_rate": 9.86397647180709e-06, "loss": 0.1743, "step": 16883 }, { "epoch": 5.19, "learning_rate": 9.862980024584044e-06, "loss": 0.1433, "step": 16884 }, { "epoch": 5.2, "learning_rate": 9.861983578721727e-06, "loss": 0.1685, "step": 16885 }, { "epoch": 5.2, "learning_rate": 9.860987134230043e-06, "loss": 0.1353, "step": 16886 }, { "epoch": 5.2, "learning_rate": 9.859990691118884e-06, "loss": 0.1358, "step": 16887 }, { "epoch": 5.2, "learning_rate": 9.858994249398147e-06, "loss": 0.2274, "step": 16888 }, { "epoch": 5.2, "learning_rate": 9.857997809077729e-06, "loss": 0.2027, "step": 16889 }, { "epoch": 5.2, "learning_rate": 9.85700137016752e-06, "loss": 0.2311, "step": 16890 }, { "epoch": 5.2, "learning_rate": 9.856004932677422e-06, "loss": 0.1889, "step": 16891 }, { "epoch": 5.2, "learning_rate": 9.855008496617326e-06, "loss": 0.1938, "step": 16892 }, { "epoch": 5.2, "learning_rate": 9.854012061997128e-06, "loss": 0.1777, "step": 16893 }, { "epoch": 5.2, "learning_rate": 9.85301562882673e-06, "loss": 0.1984, "step": 16894 }, { "epoch": 5.2, "learning_rate": 9.852019197116015e-06, "loss": 0.2396, "step": 16895 }, { "epoch": 5.2, "learning_rate": 9.851022766874892e-06, "loss": 0.2096, "step": 16896 }, { "epoch": 5.2, "learning_rate": 9.850026338113248e-06, "loss": 0.1774, "step": 16897 }, { "epoch": 5.2, "learning_rate": 9.84902991084098e-06, "loss": 0.138, "step": 16898 }, { "epoch": 5.2, "learning_rate": 9.848033485067984e-06, "loss": 0.1678, "step": 16899 }, { "epoch": 5.2, "learning_rate": 9.847037060804155e-06, "loss": 0.1669, "step": 16900 }, { "epoch": 5.2, "learning_rate": 9.846040638059394e-06, "loss": 0.2297, "step": 16901 }, { "epoch": 5.2, "learning_rate": 9.845044216843587e-06, "loss": 0.2248, "step": 16902 }, { "epoch": 5.2, "learning_rate": 9.844047797166633e-06, "loss": 0.1531, "step": 16903 }, { "epoch": 5.2, "learning_rate": 9.843051379038431e-06, "loss": 0.0903, "step": 16904 }, { "epoch": 5.2, "learning_rate": 9.84205496246887e-06, "loss": 0.1436, "step": 16905 }, { "epoch": 5.2, "learning_rate": 9.841058547467853e-06, "loss": 0.2126, "step": 16906 }, { "epoch": 5.2, "learning_rate": 9.84006213404527e-06, "loss": 0.1757, "step": 16907 }, { "epoch": 5.2, "learning_rate": 9.839065722211014e-06, "loss": 0.1992, "step": 16908 }, { "epoch": 5.2, "learning_rate": 9.838069311974986e-06, "loss": 0.1972, "step": 16909 }, { "epoch": 5.2, "learning_rate": 9.837072903347078e-06, "loss": 0.1733, "step": 16910 }, { "epoch": 5.2, "learning_rate": 9.836076496337189e-06, "loss": 0.193, "step": 16911 }, { "epoch": 5.2, "learning_rate": 9.83508009095521e-06, "loss": 0.243, "step": 16912 }, { "epoch": 5.2, "learning_rate": 9.834083687211035e-06, "loss": 0.1254, "step": 16913 }, { "epoch": 5.2, "learning_rate": 9.833087285114565e-06, "loss": 0.132, "step": 16914 }, { "epoch": 5.2, "learning_rate": 9.83209088467569e-06, "loss": 0.1643, "step": 16915 }, { "epoch": 5.2, "learning_rate": 9.83109448590431e-06, "loss": 0.1243, "step": 16916 }, { "epoch": 5.21, "learning_rate": 9.830098088810317e-06, "loss": 0.2016, "step": 16917 }, { "epoch": 5.21, "learning_rate": 9.829101693403605e-06, "loss": 0.2301, "step": 16918 }, { "epoch": 5.21, "learning_rate": 9.828105299694072e-06, "loss": 0.2247, "step": 16919 }, { "epoch": 5.21, "learning_rate": 9.827108907691609e-06, "loss": 0.1737, "step": 16920 }, { "epoch": 5.21, "learning_rate": 9.826112517406115e-06, "loss": 0.1708, "step": 16921 }, { "epoch": 5.21, "learning_rate": 9.825116128847488e-06, "loss": 0.2526, "step": 16922 }, { "epoch": 5.21, "learning_rate": 9.824119742025615e-06, "loss": 0.085, "step": 16923 }, { "epoch": 5.21, "learning_rate": 9.823123356950396e-06, "loss": 0.2522, "step": 16924 }, { "epoch": 5.21, "learning_rate": 9.822126973631723e-06, "loss": 0.221, "step": 16925 }, { "epoch": 5.21, "learning_rate": 9.821130592079496e-06, "loss": 0.1388, "step": 16926 }, { "epoch": 5.21, "learning_rate": 9.82013421230361e-06, "loss": 0.1893, "step": 16927 }, { "epoch": 5.21, "learning_rate": 9.819137834313948e-06, "loss": 0.1583, "step": 16928 }, { "epoch": 5.21, "learning_rate": 9.818141458120423e-06, "loss": 0.1396, "step": 16929 }, { "epoch": 5.21, "learning_rate": 9.817145083732914e-06, "loss": 0.139, "step": 16930 }, { "epoch": 5.21, "learning_rate": 9.816148711161328e-06, "loss": 0.2698, "step": 16931 }, { "epoch": 5.21, "learning_rate": 9.815152340415554e-06, "loss": 0.132, "step": 16932 }, { "epoch": 5.21, "learning_rate": 9.814155971505483e-06, "loss": 0.278, "step": 16933 }, { "epoch": 5.21, "learning_rate": 9.813159604441021e-06, "loss": 0.1032, "step": 16934 }, { "epoch": 5.21, "learning_rate": 9.812163239232051e-06, "loss": 0.3162, "step": 16935 }, { "epoch": 5.21, "learning_rate": 9.811166875888475e-06, "loss": 0.1817, "step": 16936 }, { "epoch": 5.21, "learning_rate": 9.810170514420188e-06, "loss": 0.1326, "step": 16937 }, { "epoch": 5.21, "learning_rate": 9.809174154837079e-06, "loss": 0.2805, "step": 16938 }, { "epoch": 5.21, "learning_rate": 9.808177797149052e-06, "loss": 0.108, "step": 16939 }, { "epoch": 5.21, "learning_rate": 9.80718144136599e-06, "loss": 0.1789, "step": 16940 }, { "epoch": 5.21, "learning_rate": 9.806185087497797e-06, "loss": 0.1789, "step": 16941 }, { "epoch": 5.21, "learning_rate": 9.805188735554364e-06, "loss": 0.1984, "step": 16942 }, { "epoch": 5.21, "learning_rate": 9.804192385545586e-06, "loss": 0.1378, "step": 16943 }, { "epoch": 5.21, "learning_rate": 9.803196037481362e-06, "loss": 0.1598, "step": 16944 }, { "epoch": 5.21, "learning_rate": 9.802199691371576e-06, "loss": 0.149, "step": 16945 }, { "epoch": 5.21, "learning_rate": 9.801203347226136e-06, "loss": 0.1346, "step": 16946 }, { "epoch": 5.21, "learning_rate": 9.800207005054927e-06, "loss": 0.1855, "step": 16947 }, { "epoch": 5.21, "learning_rate": 9.799210664867844e-06, "loss": 0.1597, "step": 16948 }, { "epoch": 5.21, "learning_rate": 9.798214326674787e-06, "loss": 0.126, "step": 16949 }, { "epoch": 5.22, "learning_rate": 9.797217990485647e-06, "loss": 0.2343, "step": 16950 }, { "epoch": 5.22, "learning_rate": 9.796221656310322e-06, "loss": 0.1632, "step": 16951 }, { "epoch": 5.22, "learning_rate": 9.795225324158702e-06, "loss": 0.1267, "step": 16952 }, { "epoch": 5.22, "learning_rate": 9.79422899404068e-06, "loss": 0.3474, "step": 16953 }, { "epoch": 5.22, "learning_rate": 9.793232665966159e-06, "loss": 0.1887, "step": 16954 }, { "epoch": 5.22, "learning_rate": 9.792236339945023e-06, "loss": 0.2196, "step": 16955 }, { "epoch": 5.22, "learning_rate": 9.791240015987177e-06, "loss": 0.1684, "step": 16956 }, { "epoch": 5.22, "learning_rate": 9.790243694102506e-06, "loss": 0.2068, "step": 16957 }, { "epoch": 5.22, "learning_rate": 9.78924737430091e-06, "loss": 0.4341, "step": 16958 }, { "epoch": 5.22, "learning_rate": 9.788251056592282e-06, "loss": 0.1424, "step": 16959 }, { "epoch": 5.22, "learning_rate": 9.787254740986514e-06, "loss": 0.1352, "step": 16960 }, { "epoch": 5.22, "learning_rate": 9.786258427493505e-06, "loss": 0.1245, "step": 16961 }, { "epoch": 5.22, "learning_rate": 9.785262116123147e-06, "loss": 0.2457, "step": 16962 }, { "epoch": 5.22, "learning_rate": 9.784265806885332e-06, "loss": 0.2059, "step": 16963 }, { "epoch": 5.22, "learning_rate": 9.783269499789958e-06, "loss": 0.1113, "step": 16964 }, { "epoch": 5.22, "learning_rate": 9.782273194846915e-06, "loss": 0.1224, "step": 16965 }, { "epoch": 5.22, "learning_rate": 9.781276892066103e-06, "loss": 0.1404, "step": 16966 }, { "epoch": 5.22, "learning_rate": 9.780280591457414e-06, "loss": 0.2316, "step": 16967 }, { "epoch": 5.22, "learning_rate": 9.779284293030737e-06, "loss": 0.1644, "step": 16968 }, { "epoch": 5.22, "learning_rate": 9.778287996795974e-06, "loss": 0.1853, "step": 16969 }, { "epoch": 5.22, "learning_rate": 9.777291702763012e-06, "loss": 0.1977, "step": 16970 }, { "epoch": 5.22, "learning_rate": 9.776295410941752e-06, "loss": 0.1023, "step": 16971 }, { "epoch": 5.22, "learning_rate": 9.775299121342086e-06, "loss": 0.2302, "step": 16972 }, { "epoch": 5.22, "learning_rate": 9.774302833973903e-06, "loss": 0.1735, "step": 16973 }, { "epoch": 5.22, "learning_rate": 9.773306548847102e-06, "loss": 0.1809, "step": 16974 }, { "epoch": 5.22, "learning_rate": 9.772310265971575e-06, "loss": 0.1324, "step": 16975 }, { "epoch": 5.22, "learning_rate": 9.771313985357217e-06, "loss": 0.1533, "step": 16976 }, { "epoch": 5.22, "learning_rate": 9.770317707013927e-06, "loss": 0.1693, "step": 16977 }, { "epoch": 5.22, "learning_rate": 9.769321430951585e-06, "loss": 0.2246, "step": 16978 }, { "epoch": 5.22, "learning_rate": 9.7683251571801e-06, "loss": 0.1083, "step": 16979 }, { "epoch": 5.22, "learning_rate": 9.767328885709358e-06, "loss": 0.2043, "step": 16980 }, { "epoch": 5.22, "learning_rate": 9.766332616549254e-06, "loss": 0.1686, "step": 16981 }, { "epoch": 5.23, "learning_rate": 9.765336349709687e-06, "loss": 0.2324, "step": 16982 }, { "epoch": 5.23, "learning_rate": 9.764340085200539e-06, "loss": 0.239, "step": 16983 }, { "epoch": 5.23, "learning_rate": 9.763343823031718e-06, "loss": 0.1773, "step": 16984 }, { "epoch": 5.23, "learning_rate": 9.762347563213105e-06, "loss": 0.1743, "step": 16985 }, { "epoch": 5.23, "learning_rate": 9.761351305754603e-06, "loss": 0.0852, "step": 16986 }, { "epoch": 5.23, "learning_rate": 9.760355050666102e-06, "loss": 0.1594, "step": 16987 }, { "epoch": 5.23, "learning_rate": 9.759358797957495e-06, "loss": 0.1369, "step": 16988 }, { "epoch": 5.23, "learning_rate": 9.75836254763868e-06, "loss": 0.2254, "step": 16989 }, { "epoch": 5.23, "learning_rate": 9.757366299719543e-06, "loss": 0.1287, "step": 16990 }, { "epoch": 5.23, "learning_rate": 9.756370054209987e-06, "loss": 0.1417, "step": 16991 }, { "epoch": 5.23, "learning_rate": 9.755373811119898e-06, "loss": 0.1695, "step": 16992 }, { "epoch": 5.23, "learning_rate": 9.754377570459173e-06, "loss": 0.1824, "step": 16993 }, { "epoch": 5.23, "learning_rate": 9.753381332237706e-06, "loss": 0.1548, "step": 16994 }, { "epoch": 5.23, "learning_rate": 9.752385096465386e-06, "loss": 0.2217, "step": 16995 }, { "epoch": 5.23, "learning_rate": 9.751388863152116e-06, "loss": 0.1367, "step": 16996 }, { "epoch": 5.23, "learning_rate": 9.750392632307782e-06, "loss": 0.0969, "step": 16997 }, { "epoch": 5.23, "learning_rate": 9.749396403942277e-06, "loss": 0.1311, "step": 16998 }, { "epoch": 5.23, "learning_rate": 9.748400178065499e-06, "loss": 0.1965, "step": 16999 }, { "epoch": 5.23, "learning_rate": 9.747403954687334e-06, "loss": 0.1319, "step": 17000 }, { "epoch": 5.23, "learning_rate": 9.746407733817688e-06, "loss": 0.2074, "step": 17001 }, { "epoch": 5.23, "learning_rate": 9.745411515466444e-06, "loss": 0.2166, "step": 17002 }, { "epoch": 5.23, "learning_rate": 9.744415299643496e-06, "loss": 0.1507, "step": 17003 }, { "epoch": 5.23, "learning_rate": 9.743419086358742e-06, "loss": 0.1008, "step": 17004 }, { "epoch": 5.23, "learning_rate": 9.74242287562207e-06, "loss": 0.2004, "step": 17005 }, { "epoch": 5.23, "learning_rate": 9.741426667443381e-06, "loss": 0.1565, "step": 17006 }, { "epoch": 5.23, "learning_rate": 9.740430461832562e-06, "loss": 0.1824, "step": 17007 }, { "epoch": 5.23, "learning_rate": 9.739434258799506e-06, "loss": 0.186, "step": 17008 }, { "epoch": 5.23, "learning_rate": 9.738438058354108e-06, "loss": 0.2377, "step": 17009 }, { "epoch": 5.23, "learning_rate": 9.73744186050626e-06, "loss": 0.1116, "step": 17010 }, { "epoch": 5.23, "learning_rate": 9.73644566526586e-06, "loss": 0.2709, "step": 17011 }, { "epoch": 5.23, "learning_rate": 9.735449472642798e-06, "loss": 0.2093, "step": 17012 }, { "epoch": 5.23, "learning_rate": 9.734453282646962e-06, "loss": 0.306, "step": 17013 }, { "epoch": 5.23, "learning_rate": 9.733457095288252e-06, "loss": 0.1946, "step": 17014 }, { "epoch": 5.24, "learning_rate": 9.732460910576557e-06, "loss": 0.2764, "step": 17015 }, { "epoch": 5.24, "learning_rate": 9.731464728521774e-06, "loss": 0.2165, "step": 17016 }, { "epoch": 5.24, "learning_rate": 9.730468549133794e-06, "loss": 0.1168, "step": 17017 }, { "epoch": 5.24, "learning_rate": 9.729472372422507e-06, "loss": 0.1794, "step": 17018 }, { "epoch": 5.24, "learning_rate": 9.728476198397811e-06, "loss": 0.1465, "step": 17019 }, { "epoch": 5.24, "learning_rate": 9.727480027069594e-06, "loss": 0.1403, "step": 17020 }, { "epoch": 5.24, "learning_rate": 9.726483858447754e-06, "loss": 0.0951, "step": 17021 }, { "epoch": 5.24, "learning_rate": 9.725487692542184e-06, "loss": 0.2202, "step": 17022 }, { "epoch": 5.24, "learning_rate": 9.72449152936277e-06, "loss": 0.1654, "step": 17023 }, { "epoch": 5.24, "learning_rate": 9.72349536891941e-06, "loss": 0.1888, "step": 17024 }, { "epoch": 5.24, "learning_rate": 9.722499211221993e-06, "loss": 0.1987, "step": 17025 }, { "epoch": 5.24, "learning_rate": 9.721503056280418e-06, "loss": 0.1701, "step": 17026 }, { "epoch": 5.24, "learning_rate": 9.720506904104578e-06, "loss": 0.156, "step": 17027 }, { "epoch": 5.24, "learning_rate": 9.719510754704353e-06, "loss": 0.1835, "step": 17028 }, { "epoch": 5.24, "learning_rate": 9.718514608089652e-06, "loss": 0.1237, "step": 17029 }, { "epoch": 5.24, "learning_rate": 9.717518464270355e-06, "loss": 0.1142, "step": 17030 }, { "epoch": 5.24, "learning_rate": 9.716522323256365e-06, "loss": 0.0828, "step": 17031 }, { "epoch": 5.24, "learning_rate": 9.71552618505757e-06, "loss": 0.1769, "step": 17032 }, { "epoch": 5.24, "learning_rate": 9.714530049683855e-06, "loss": 0.2208, "step": 17033 }, { "epoch": 5.24, "learning_rate": 9.713533917145127e-06, "loss": 0.0973, "step": 17034 }, { "epoch": 5.24, "learning_rate": 9.712537787451268e-06, "loss": 0.126, "step": 17035 }, { "epoch": 5.24, "learning_rate": 9.711541660612174e-06, "loss": 0.1552, "step": 17036 }, { "epoch": 5.24, "learning_rate": 9.710545536637738e-06, "loss": 0.1761, "step": 17037 }, { "epoch": 5.24, "learning_rate": 9.709549415537848e-06, "loss": 0.2088, "step": 17038 }, { "epoch": 5.24, "learning_rate": 9.708553297322407e-06, "loss": 0.1239, "step": 17039 }, { "epoch": 5.24, "learning_rate": 9.707557182001294e-06, "loss": 0.3534, "step": 17040 }, { "epoch": 5.24, "learning_rate": 9.70656106958441e-06, "loss": 0.2342, "step": 17041 }, { "epoch": 5.24, "learning_rate": 9.705564960081646e-06, "loss": 0.1839, "step": 17042 }, { "epoch": 5.24, "learning_rate": 9.704568853502891e-06, "loss": 0.161, "step": 17043 }, { "epoch": 5.24, "learning_rate": 9.703572749858044e-06, "loss": 0.1316, "step": 17044 }, { "epoch": 5.24, "learning_rate": 9.702576649156984e-06, "loss": 0.1635, "step": 17045 }, { "epoch": 5.24, "learning_rate": 9.701580551409622e-06, "loss": 0.107, "step": 17046 }, { "epoch": 5.25, "learning_rate": 9.700584456625834e-06, "loss": 0.1847, "step": 17047 }, { "epoch": 5.25, "learning_rate": 9.69958836481552e-06, "loss": 0.1352, "step": 17048 }, { "epoch": 5.25, "learning_rate": 9.698592275988573e-06, "loss": 0.2372, "step": 17049 }, { "epoch": 5.25, "learning_rate": 9.697596190154876e-06, "loss": 0.2541, "step": 17050 }, { "epoch": 5.25, "learning_rate": 9.696600107324334e-06, "loss": 0.16, "step": 17051 }, { "epoch": 5.25, "learning_rate": 9.69560402750683e-06, "loss": 0.1552, "step": 17052 }, { "epoch": 5.25, "learning_rate": 9.694607950712258e-06, "loss": 0.1457, "step": 17053 }, { "epoch": 5.25, "learning_rate": 9.693611876950512e-06, "loss": 0.291, "step": 17054 }, { "epoch": 5.25, "learning_rate": 9.69261580623148e-06, "loss": 0.1274, "step": 17055 }, { "epoch": 5.25, "learning_rate": 9.69161973856506e-06, "loss": 0.1557, "step": 17056 }, { "epoch": 5.25, "learning_rate": 9.690623673961134e-06, "loss": 0.1256, "step": 17057 }, { "epoch": 5.25, "learning_rate": 9.689627612429607e-06, "loss": 0.225, "step": 17058 }, { "epoch": 5.25, "learning_rate": 9.688631553980361e-06, "loss": 0.164, "step": 17059 }, { "epoch": 5.25, "learning_rate": 9.687635498623291e-06, "loss": 0.1685, "step": 17060 }, { "epoch": 5.25, "learning_rate": 9.68663944636829e-06, "loss": 0.1058, "step": 17061 }, { "epoch": 5.25, "learning_rate": 9.685643397225242e-06, "loss": 0.2037, "step": 17062 }, { "epoch": 5.25, "learning_rate": 9.684647351204052e-06, "loss": 0.1796, "step": 17063 }, { "epoch": 5.25, "learning_rate": 9.683651308314602e-06, "loss": 0.1936, "step": 17064 }, { "epoch": 5.25, "learning_rate": 9.682655268566783e-06, "loss": 0.196, "step": 17065 }, { "epoch": 5.25, "learning_rate": 9.681659231970493e-06, "loss": 0.1771, "step": 17066 }, { "epoch": 5.25, "learning_rate": 9.680663198535618e-06, "loss": 0.208, "step": 17067 }, { "epoch": 5.25, "learning_rate": 9.679667168272056e-06, "loss": 0.1983, "step": 17068 }, { "epoch": 5.25, "learning_rate": 9.678671141189691e-06, "loss": 0.271, "step": 17069 }, { "epoch": 5.25, "learning_rate": 9.677675117298416e-06, "loss": 0.2073, "step": 17070 }, { "epoch": 5.25, "learning_rate": 9.676679096608127e-06, "loss": 0.0653, "step": 17071 }, { "epoch": 5.25, "learning_rate": 9.67568307912871e-06, "loss": 0.1008, "step": 17072 }, { "epoch": 5.25, "learning_rate": 9.674687064870062e-06, "loss": 0.2465, "step": 17073 }, { "epoch": 5.25, "learning_rate": 9.673691053842072e-06, "loss": 0.1196, "step": 17074 }, { "epoch": 5.25, "learning_rate": 9.672695046054626e-06, "loss": 0.1585, "step": 17075 }, { "epoch": 5.25, "learning_rate": 9.671699041517623e-06, "loss": 0.1733, "step": 17076 }, { "epoch": 5.25, "learning_rate": 9.67070304024095e-06, "loss": 0.1317, "step": 17077 }, { "epoch": 5.25, "learning_rate": 9.669707042234502e-06, "loss": 0.1746, "step": 17078 }, { "epoch": 5.25, "learning_rate": 9.668711047508167e-06, "loss": 0.1373, "step": 17079 }, { "epoch": 5.26, "learning_rate": 9.667715056071831e-06, "loss": 0.3431, "step": 17080 }, { "epoch": 5.26, "learning_rate": 9.666719067935395e-06, "loss": 0.1279, "step": 17081 }, { "epoch": 5.26, "learning_rate": 9.665723083108744e-06, "loss": 0.0893, "step": 17082 }, { "epoch": 5.26, "learning_rate": 9.664727101601773e-06, "loss": 0.2718, "step": 17083 }, { "epoch": 5.26, "learning_rate": 9.663731123424373e-06, "loss": 0.1209, "step": 17084 }, { "epoch": 5.26, "learning_rate": 9.662735148586429e-06, "loss": 0.1545, "step": 17085 }, { "epoch": 5.26, "learning_rate": 9.661739177097836e-06, "loss": 0.1534, "step": 17086 }, { "epoch": 5.26, "learning_rate": 9.660743208968484e-06, "loss": 0.2101, "step": 17087 }, { "epoch": 5.26, "learning_rate": 9.659747244208267e-06, "loss": 0.1552, "step": 17088 }, { "epoch": 5.26, "learning_rate": 9.658751282827075e-06, "loss": 0.2522, "step": 17089 }, { "epoch": 5.26, "learning_rate": 9.657755324834794e-06, "loss": 0.1719, "step": 17090 }, { "epoch": 5.26, "learning_rate": 9.656759370241318e-06, "loss": 0.141, "step": 17091 }, { "epoch": 5.26, "learning_rate": 9.655763419056537e-06, "loss": 0.1625, "step": 17092 }, { "epoch": 5.26, "learning_rate": 9.654767471290346e-06, "loss": 0.1367, "step": 17093 }, { "epoch": 5.26, "learning_rate": 9.653771526952632e-06, "loss": 0.151, "step": 17094 }, { "epoch": 5.26, "learning_rate": 9.65277558605328e-06, "loss": 0.1435, "step": 17095 }, { "epoch": 5.26, "learning_rate": 9.651779648602193e-06, "loss": 0.1415, "step": 17096 }, { "epoch": 5.26, "learning_rate": 9.650783714609252e-06, "loss": 0.2191, "step": 17097 }, { "epoch": 5.26, "learning_rate": 9.649787784084352e-06, "loss": 0.1858, "step": 17098 }, { "epoch": 5.26, "learning_rate": 9.648791857037384e-06, "loss": 0.2587, "step": 17099 }, { "epoch": 5.26, "learning_rate": 9.647795933478229e-06, "loss": 0.2579, "step": 17100 }, { "epoch": 5.26, "learning_rate": 9.646800013416792e-06, "loss": 0.1315, "step": 17101 }, { "epoch": 5.26, "learning_rate": 9.645804096862953e-06, "loss": 0.1544, "step": 17102 }, { "epoch": 5.26, "learning_rate": 9.644808183826609e-06, "loss": 0.1536, "step": 17103 }, { "epoch": 5.26, "learning_rate": 9.643812274317644e-06, "loss": 0.1269, "step": 17104 }, { "epoch": 5.26, "learning_rate": 9.64281636834595e-06, "loss": 0.2246, "step": 17105 }, { "epoch": 5.26, "learning_rate": 9.641820465921426e-06, "loss": 0.1361, "step": 17106 }, { "epoch": 5.26, "learning_rate": 9.640824567053945e-06, "loss": 0.1577, "step": 17107 }, { "epoch": 5.26, "learning_rate": 9.639828671753415e-06, "loss": 0.1099, "step": 17108 }, { "epoch": 5.26, "learning_rate": 9.638832780029716e-06, "loss": 0.1436, "step": 17109 }, { "epoch": 5.26, "learning_rate": 9.637836891892738e-06, "loss": 0.115, "step": 17110 }, { "epoch": 5.26, "learning_rate": 9.636841007352378e-06, "loss": 0.2018, "step": 17111 }, { "epoch": 5.27, "learning_rate": 9.635845126418514e-06, "loss": 0.1541, "step": 17112 }, { "epoch": 5.27, "learning_rate": 9.63484924910105e-06, "loss": 0.1873, "step": 17113 }, { "epoch": 5.27, "learning_rate": 9.633853375409868e-06, "loss": 0.1878, "step": 17114 }, { "epoch": 5.27, "learning_rate": 9.632857505354859e-06, "loss": 0.1203, "step": 17115 }, { "epoch": 5.27, "learning_rate": 9.631861638945912e-06, "loss": 0.2178, "step": 17116 }, { "epoch": 5.27, "learning_rate": 9.630865776192918e-06, "loss": 0.1472, "step": 17117 }, { "epoch": 5.27, "learning_rate": 9.629869917105771e-06, "loss": 0.0559, "step": 17118 }, { "epoch": 5.27, "learning_rate": 9.628874061694355e-06, "loss": 0.1158, "step": 17119 }, { "epoch": 5.27, "learning_rate": 9.627878209968559e-06, "loss": 0.2817, "step": 17120 }, { "epoch": 5.27, "learning_rate": 9.62688236193828e-06, "loss": 0.1349, "step": 17121 }, { "epoch": 5.27, "learning_rate": 9.625886517613398e-06, "loss": 0.182, "step": 17122 }, { "epoch": 5.27, "learning_rate": 9.624890677003812e-06, "loss": 0.2058, "step": 17123 }, { "epoch": 5.27, "learning_rate": 9.623894840119406e-06, "loss": 0.1756, "step": 17124 }, { "epoch": 5.27, "learning_rate": 9.62289900697007e-06, "loss": 0.2263, "step": 17125 }, { "epoch": 5.27, "learning_rate": 9.621903177565694e-06, "loss": 0.2765, "step": 17126 }, { "epoch": 5.27, "learning_rate": 9.620907351916166e-06, "loss": 0.2053, "step": 17127 }, { "epoch": 5.27, "learning_rate": 9.619911530031384e-06, "loss": 0.1551, "step": 17128 }, { "epoch": 5.27, "learning_rate": 9.618915711921226e-06, "loss": 0.1556, "step": 17129 }, { "epoch": 5.27, "learning_rate": 9.617919897595586e-06, "loss": 0.1417, "step": 17130 }, { "epoch": 5.27, "learning_rate": 9.616924087064355e-06, "loss": 0.095, "step": 17131 }, { "epoch": 5.27, "learning_rate": 9.61592828033742e-06, "loss": 0.2567, "step": 17132 }, { "epoch": 5.27, "learning_rate": 9.61493247742467e-06, "loss": 0.2394, "step": 17133 }, { "epoch": 5.27, "learning_rate": 9.613936678336e-06, "loss": 0.2971, "step": 17134 }, { "epoch": 5.27, "learning_rate": 9.612940883081289e-06, "loss": 0.2819, "step": 17135 }, { "epoch": 5.27, "learning_rate": 9.611945091670435e-06, "loss": 0.1591, "step": 17136 }, { "epoch": 5.27, "learning_rate": 9.610949304113321e-06, "loss": 0.2017, "step": 17137 }, { "epoch": 5.27, "learning_rate": 9.609953520419842e-06, "loss": 0.1502, "step": 17138 }, { "epoch": 5.27, "learning_rate": 9.608957740599888e-06, "loss": 0.1621, "step": 17139 }, { "epoch": 5.27, "learning_rate": 9.607961964663337e-06, "loss": 0.2028, "step": 17140 }, { "epoch": 5.27, "learning_rate": 9.606966192620086e-06, "loss": 0.1288, "step": 17141 }, { "epoch": 5.27, "learning_rate": 9.605970424480021e-06, "loss": 0.2367, "step": 17142 }, { "epoch": 5.27, "learning_rate": 9.604974660253039e-06, "loss": 0.1593, "step": 17143 }, { "epoch": 5.27, "learning_rate": 9.603978899949021e-06, "loss": 0.1789, "step": 17144 }, { "epoch": 5.28, "learning_rate": 9.602983143577852e-06, "loss": 0.2278, "step": 17145 }, { "epoch": 5.28, "learning_rate": 9.601987391149434e-06, "loss": 0.1486, "step": 17146 }, { "epoch": 5.28, "learning_rate": 9.600991642673642e-06, "loss": 0.1497, "step": 17147 }, { "epoch": 5.28, "learning_rate": 9.599995898160374e-06, "loss": 0.1959, "step": 17148 }, { "epoch": 5.28, "learning_rate": 9.599000157619517e-06, "loss": 0.263, "step": 17149 }, { "epoch": 5.28, "learning_rate": 9.598004421060952e-06, "loss": 0.2402, "step": 17150 }, { "epoch": 5.28, "learning_rate": 9.59700868849458e-06, "loss": 0.1294, "step": 17151 }, { "epoch": 5.28, "learning_rate": 9.596012959930278e-06, "loss": 0.1385, "step": 17152 }, { "epoch": 5.28, "learning_rate": 9.595017235377942e-06, "loss": 0.1408, "step": 17153 }, { "epoch": 5.28, "learning_rate": 9.59402151484746e-06, "loss": 0.1323, "step": 17154 }, { "epoch": 5.28, "learning_rate": 9.593025798348713e-06, "loss": 0.1743, "step": 17155 }, { "epoch": 5.28, "learning_rate": 9.592030085891602e-06, "loss": 0.1882, "step": 17156 }, { "epoch": 5.28, "learning_rate": 9.591034377486003e-06, "loss": 0.1446, "step": 17157 }, { "epoch": 5.28, "learning_rate": 9.590038673141812e-06, "loss": 0.2724, "step": 17158 }, { "epoch": 5.28, "learning_rate": 9.589042972868913e-06, "loss": 0.173, "step": 17159 }, { "epoch": 5.28, "learning_rate": 9.588047276677196e-06, "loss": 0.1623, "step": 17160 }, { "epoch": 5.28, "learning_rate": 9.587051584576554e-06, "loss": 0.1114, "step": 17161 }, { "epoch": 5.28, "learning_rate": 9.586055896576863e-06, "loss": 0.1897, "step": 17162 }, { "epoch": 5.28, "learning_rate": 9.585060212688023e-06, "loss": 0.0965, "step": 17163 }, { "epoch": 5.28, "learning_rate": 9.584064532919918e-06, "loss": 0.1183, "step": 17164 }, { "epoch": 5.28, "learning_rate": 9.583068857282431e-06, "loss": 0.167, "step": 17165 }, { "epoch": 5.28, "learning_rate": 9.58207318578546e-06, "loss": 0.2042, "step": 17166 }, { "epoch": 5.28, "learning_rate": 9.581077518438883e-06, "loss": 0.211, "step": 17167 }, { "epoch": 5.28, "learning_rate": 9.580081855252597e-06, "loss": 0.1367, "step": 17168 }, { "epoch": 5.28, "learning_rate": 9.579086196236483e-06, "loss": 0.2299, "step": 17169 }, { "epoch": 5.28, "learning_rate": 9.578090541400429e-06, "loss": 0.1753, "step": 17170 }, { "epoch": 5.28, "learning_rate": 9.577094890754325e-06, "loss": 0.1394, "step": 17171 }, { "epoch": 5.28, "learning_rate": 9.57609924430806e-06, "loss": 0.1837, "step": 17172 }, { "epoch": 5.28, "learning_rate": 9.575103602071521e-06, "loss": 0.1989, "step": 17173 }, { "epoch": 5.28, "learning_rate": 9.574107964054596e-06, "loss": 0.1097, "step": 17174 }, { "epoch": 5.28, "learning_rate": 9.573112330267166e-06, "loss": 0.1539, "step": 17175 }, { "epoch": 5.28, "learning_rate": 9.572116700719129e-06, "loss": 0.1643, "step": 17176 }, { "epoch": 5.29, "learning_rate": 9.571121075420364e-06, "loss": 0.1701, "step": 17177 }, { "epoch": 5.29, "learning_rate": 9.570125454380765e-06, "loss": 0.1757, "step": 17178 }, { "epoch": 5.29, "learning_rate": 9.569129837610216e-06, "loss": 0.1714, "step": 17179 }, { "epoch": 5.29, "learning_rate": 9.568134225118603e-06, "loss": 0.2046, "step": 17180 }, { "epoch": 5.29, "learning_rate": 9.567138616915816e-06, "loss": 0.1519, "step": 17181 }, { "epoch": 5.29, "learning_rate": 9.56614301301174e-06, "loss": 0.1265, "step": 17182 }, { "epoch": 5.29, "learning_rate": 9.565147413416266e-06, "loss": 0.2951, "step": 17183 }, { "epoch": 5.29, "learning_rate": 9.564151818139281e-06, "loss": 0.1295, "step": 17184 }, { "epoch": 5.29, "learning_rate": 9.563156227190665e-06, "loss": 0.196, "step": 17185 }, { "epoch": 5.29, "learning_rate": 9.562160640580313e-06, "loss": 0.1311, "step": 17186 }, { "epoch": 5.29, "learning_rate": 9.561165058318108e-06, "loss": 0.2409, "step": 17187 }, { "epoch": 5.29, "learning_rate": 9.56016948041394e-06, "loss": 0.0892, "step": 17188 }, { "epoch": 5.29, "learning_rate": 9.559173906877696e-06, "loss": 0.1781, "step": 17189 }, { "epoch": 5.29, "learning_rate": 9.558178337719259e-06, "loss": 0.1314, "step": 17190 }, { "epoch": 5.29, "learning_rate": 9.557182772948518e-06, "loss": 0.2247, "step": 17191 }, { "epoch": 5.29, "learning_rate": 9.556187212575359e-06, "loss": 0.2167, "step": 17192 }, { "epoch": 5.29, "learning_rate": 9.555191656609673e-06, "loss": 0.1345, "step": 17193 }, { "epoch": 5.29, "learning_rate": 9.554196105061345e-06, "loss": 0.1472, "step": 17194 }, { "epoch": 5.29, "learning_rate": 9.553200557940254e-06, "loss": 0.1786, "step": 17195 }, { "epoch": 5.29, "learning_rate": 9.5522050152563e-06, "loss": 0.1064, "step": 17196 }, { "epoch": 5.29, "learning_rate": 9.55120947701936e-06, "loss": 0.198, "step": 17197 }, { "epoch": 5.29, "learning_rate": 9.550213943239323e-06, "loss": 0.277, "step": 17198 }, { "epoch": 5.29, "learning_rate": 9.54921841392608e-06, "loss": 0.2374, "step": 17199 }, { "epoch": 5.29, "learning_rate": 9.548222889089505e-06, "loss": 0.1385, "step": 17200 }, { "epoch": 5.29, "learning_rate": 9.547227368739503e-06, "loss": 0.1757, "step": 17201 }, { "epoch": 5.29, "learning_rate": 9.546231852885943e-06, "loss": 0.1314, "step": 17202 }, { "epoch": 5.29, "learning_rate": 9.545236341538722e-06, "loss": 0.1979, "step": 17203 }, { "epoch": 5.29, "learning_rate": 9.544240834707724e-06, "loss": 0.1241, "step": 17204 }, { "epoch": 5.29, "learning_rate": 9.543245332402832e-06, "loss": 0.1529, "step": 17205 }, { "epoch": 5.29, "learning_rate": 9.54224983463394e-06, "loss": 0.1892, "step": 17206 }, { "epoch": 5.29, "learning_rate": 9.541254341410924e-06, "loss": 0.1613, "step": 17207 }, { "epoch": 5.29, "learning_rate": 9.540258852743676e-06, "loss": 0.1492, "step": 17208 }, { "epoch": 5.29, "learning_rate": 9.539263368642084e-06, "loss": 0.2036, "step": 17209 }, { "epoch": 5.3, "learning_rate": 9.538267889116027e-06, "loss": 0.2239, "step": 17210 }, { "epoch": 5.3, "learning_rate": 9.5372724141754e-06, "loss": 0.1634, "step": 17211 }, { "epoch": 5.3, "learning_rate": 9.536276943830078e-06, "loss": 0.1935, "step": 17212 }, { "epoch": 5.3, "learning_rate": 9.53528147808996e-06, "loss": 0.1938, "step": 17213 }, { "epoch": 5.3, "learning_rate": 9.534286016964924e-06, "loss": 0.1687, "step": 17214 }, { "epoch": 5.3, "learning_rate": 9.533290560464854e-06, "loss": 0.2279, "step": 17215 }, { "epoch": 5.3, "learning_rate": 9.532295108599642e-06, "loss": 0.2558, "step": 17216 }, { "epoch": 5.3, "learning_rate": 9.531299661379167e-06, "loss": 0.1774, "step": 17217 }, { "epoch": 5.3, "learning_rate": 9.530304218813323e-06, "loss": 0.1367, "step": 17218 }, { "epoch": 5.3, "learning_rate": 9.52930878091199e-06, "loss": 0.1315, "step": 17219 }, { "epoch": 5.3, "learning_rate": 9.528313347685052e-06, "loss": 0.1423, "step": 17220 }, { "epoch": 5.3, "learning_rate": 9.527317919142398e-06, "loss": 0.2814, "step": 17221 }, { "epoch": 5.3, "learning_rate": 9.52632249529391e-06, "loss": 0.2234, "step": 17222 }, { "epoch": 5.3, "learning_rate": 9.525327076149482e-06, "loss": 0.1294, "step": 17223 }, { "epoch": 5.3, "learning_rate": 9.524331661718994e-06, "loss": 0.0925, "step": 17224 }, { "epoch": 5.3, "learning_rate": 9.523336252012325e-06, "loss": 0.1381, "step": 17225 }, { "epoch": 5.3, "learning_rate": 9.52234084703937e-06, "loss": 0.1369, "step": 17226 }, { "epoch": 5.3, "learning_rate": 9.521345446810008e-06, "loss": 0.1206, "step": 17227 }, { "epoch": 5.3, "learning_rate": 9.520350051334131e-06, "loss": 0.1031, "step": 17228 }, { "epoch": 5.3, "learning_rate": 9.519354660621617e-06, "loss": 0.1392, "step": 17229 }, { "epoch": 5.3, "learning_rate": 9.518359274682354e-06, "loss": 0.1678, "step": 17230 }, { "epoch": 5.3, "learning_rate": 9.517363893526228e-06, "loss": 0.1509, "step": 17231 }, { "epoch": 5.3, "learning_rate": 9.516368517163123e-06, "loss": 0.2795, "step": 17232 }, { "epoch": 5.3, "learning_rate": 9.515373145602926e-06, "loss": 0.2057, "step": 17233 }, { "epoch": 5.3, "learning_rate": 9.514377778855521e-06, "loss": 0.1648, "step": 17234 }, { "epoch": 5.3, "learning_rate": 9.51338241693079e-06, "loss": 0.2223, "step": 17235 }, { "epoch": 5.3, "learning_rate": 9.51238705983862e-06, "loss": 0.1509, "step": 17236 }, { "epoch": 5.3, "learning_rate": 9.511391707588895e-06, "loss": 0.1581, "step": 17237 }, { "epoch": 5.3, "learning_rate": 9.510396360191501e-06, "loss": 0.1641, "step": 17238 }, { "epoch": 5.3, "learning_rate": 9.509401017656327e-06, "loss": 0.1798, "step": 17239 }, { "epoch": 5.3, "learning_rate": 9.508405679993248e-06, "loss": 0.1893, "step": 17240 }, { "epoch": 5.3, "learning_rate": 9.507410347212157e-06, "loss": 0.1071, "step": 17241 }, { "epoch": 5.31, "learning_rate": 9.50641501932293e-06, "loss": 0.1542, "step": 17242 }, { "epoch": 5.31, "learning_rate": 9.505419696335461e-06, "loss": 0.175, "step": 17243 }, { "epoch": 5.31, "learning_rate": 9.504424378259628e-06, "loss": 0.1817, "step": 17244 }, { "epoch": 5.31, "learning_rate": 9.50342906510532e-06, "loss": 0.1942, "step": 17245 }, { "epoch": 5.31, "learning_rate": 9.50243375688242e-06, "loss": 0.1191, "step": 17246 }, { "epoch": 5.31, "learning_rate": 9.501438453600808e-06, "loss": 0.2771, "step": 17247 }, { "epoch": 5.31, "learning_rate": 9.500443155270374e-06, "loss": 0.1431, "step": 17248 }, { "epoch": 5.31, "learning_rate": 9.499447861900997e-06, "loss": 0.1923, "step": 17249 }, { "epoch": 5.31, "learning_rate": 9.498452573502565e-06, "loss": 0.1451, "step": 17250 }, { "epoch": 5.31, "learning_rate": 9.497457290084967e-06, "loss": 0.1306, "step": 17251 }, { "epoch": 5.31, "learning_rate": 9.496462011658074e-06, "loss": 0.1069, "step": 17252 }, { "epoch": 5.31, "learning_rate": 9.49546673823178e-06, "loss": 0.1912, "step": 17253 }, { "epoch": 5.31, "learning_rate": 9.494471469815965e-06, "loss": 0.0978, "step": 17254 }, { "epoch": 5.31, "learning_rate": 9.493476206420516e-06, "loss": 0.1802, "step": 17255 }, { "epoch": 5.31, "learning_rate": 9.492480948055316e-06, "loss": 0.2368, "step": 17256 }, { "epoch": 5.31, "learning_rate": 9.491485694730245e-06, "loss": 0.1904, "step": 17257 }, { "epoch": 5.31, "learning_rate": 9.490490446455192e-06, "loss": 0.2199, "step": 17258 }, { "epoch": 5.31, "learning_rate": 9.489495203240036e-06, "loss": 0.2014, "step": 17259 }, { "epoch": 5.31, "learning_rate": 9.488499965094664e-06, "loss": 0.2365, "step": 17260 }, { "epoch": 5.31, "learning_rate": 9.487504732028963e-06, "loss": 0.1785, "step": 17261 }, { "epoch": 5.31, "learning_rate": 9.486509504052804e-06, "loss": 0.162, "step": 17262 }, { "epoch": 5.31, "learning_rate": 9.485514281176086e-06, "loss": 0.1239, "step": 17263 }, { "epoch": 5.31, "learning_rate": 9.484519063408682e-06, "loss": 0.1359, "step": 17264 }, { "epoch": 5.31, "learning_rate": 9.48352385076048e-06, "loss": 0.161, "step": 17265 }, { "epoch": 5.31, "learning_rate": 9.482528643241363e-06, "loss": 0.2205, "step": 17266 }, { "epoch": 5.31, "learning_rate": 9.481533440861209e-06, "loss": 0.1743, "step": 17267 }, { "epoch": 5.31, "learning_rate": 9.480538243629912e-06, "loss": 0.2333, "step": 17268 }, { "epoch": 5.31, "learning_rate": 9.479543051557344e-06, "loss": 0.1418, "step": 17269 }, { "epoch": 5.31, "learning_rate": 9.478547864653395e-06, "loss": 0.1847, "step": 17270 }, { "epoch": 5.31, "learning_rate": 9.477552682927948e-06, "loss": 0.185, "step": 17271 }, { "epoch": 5.31, "learning_rate": 9.476557506390881e-06, "loss": 0.304, "step": 17272 }, { "epoch": 5.31, "learning_rate": 9.475562335052086e-06, "loss": 0.1642, "step": 17273 }, { "epoch": 5.31, "learning_rate": 9.474567168921433e-06, "loss": 0.1788, "step": 17274 }, { "epoch": 5.32, "learning_rate": 9.473572008008817e-06, "loss": 0.1508, "step": 17275 }, { "epoch": 5.32, "learning_rate": 9.472576852324116e-06, "loss": 0.2837, "step": 17276 }, { "epoch": 5.32, "learning_rate": 9.47158170187721e-06, "loss": 0.1985, "step": 17277 }, { "epoch": 5.32, "learning_rate": 9.47058655667799e-06, "loss": 0.2828, "step": 17278 }, { "epoch": 5.32, "learning_rate": 9.469591416736326e-06, "loss": 0.1537, "step": 17279 }, { "epoch": 5.32, "learning_rate": 9.468596282062114e-06, "loss": 0.179, "step": 17280 }, { "epoch": 5.32, "learning_rate": 9.46760115266523e-06, "loss": 0.1511, "step": 17281 }, { "epoch": 5.32, "learning_rate": 9.466606028555556e-06, "loss": 0.1879, "step": 17282 }, { "epoch": 5.32, "learning_rate": 9.465610909742974e-06, "loss": 0.2234, "step": 17283 }, { "epoch": 5.32, "learning_rate": 9.46461579623737e-06, "loss": 0.2357, "step": 17284 }, { "epoch": 5.32, "learning_rate": 9.463620688048626e-06, "loss": 0.3041, "step": 17285 }, { "epoch": 5.32, "learning_rate": 9.462625585186621e-06, "loss": 0.1638, "step": 17286 }, { "epoch": 5.32, "learning_rate": 9.46163048766124e-06, "loss": 0.1148, "step": 17287 }, { "epoch": 5.32, "learning_rate": 9.460635395482364e-06, "loss": 0.1731, "step": 17288 }, { "epoch": 5.32, "learning_rate": 9.459640308659873e-06, "loss": 0.1016, "step": 17289 }, { "epoch": 5.32, "learning_rate": 9.458645227203658e-06, "loss": 0.2234, "step": 17290 }, { "epoch": 5.32, "learning_rate": 9.457650151123592e-06, "loss": 0.1396, "step": 17291 }, { "epoch": 5.32, "learning_rate": 9.456655080429557e-06, "loss": 0.2311, "step": 17292 }, { "epoch": 5.32, "learning_rate": 9.455660015131441e-06, "loss": 0.1315, "step": 17293 }, { "epoch": 5.32, "learning_rate": 9.454664955239119e-06, "loss": 0.1965, "step": 17294 }, { "epoch": 5.32, "learning_rate": 9.45366990076248e-06, "loss": 0.127, "step": 17295 }, { "epoch": 5.32, "learning_rate": 9.452674851711404e-06, "loss": 0.2039, "step": 17296 }, { "epoch": 5.32, "learning_rate": 9.451679808095766e-06, "loss": 0.1659, "step": 17297 }, { "epoch": 5.32, "learning_rate": 9.450684769925458e-06, "loss": 0.1629, "step": 17298 }, { "epoch": 5.32, "learning_rate": 9.449689737210352e-06, "loss": 0.0879, "step": 17299 }, { "epoch": 5.32, "learning_rate": 9.448694709960336e-06, "loss": 0.098, "step": 17300 }, { "epoch": 5.32, "learning_rate": 9.447699688185293e-06, "loss": 0.135, "step": 17301 }, { "epoch": 5.32, "learning_rate": 9.446704671895096e-06, "loss": 0.216, "step": 17302 }, { "epoch": 5.32, "learning_rate": 9.445709661099635e-06, "loss": 0.1134, "step": 17303 }, { "epoch": 5.32, "learning_rate": 9.444714655808784e-06, "loss": 0.3431, "step": 17304 }, { "epoch": 5.32, "learning_rate": 9.443719656032431e-06, "loss": 0.1517, "step": 17305 }, { "epoch": 5.32, "learning_rate": 9.442724661780458e-06, "loss": 0.1455, "step": 17306 }, { "epoch": 5.33, "learning_rate": 9.441729673062736e-06, "loss": 0.21, "step": 17307 }, { "epoch": 5.33, "learning_rate": 9.440734689889156e-06, "loss": 0.1974, "step": 17308 }, { "epoch": 5.33, "learning_rate": 9.439739712269594e-06, "loss": 0.1682, "step": 17309 }, { "epoch": 5.33, "learning_rate": 9.438744740213936e-06, "loss": 0.1271, "step": 17310 }, { "epoch": 5.33, "learning_rate": 9.437749773732061e-06, "loss": 0.1093, "step": 17311 }, { "epoch": 5.33, "learning_rate": 9.436754812833843e-06, "loss": 0.0807, "step": 17312 }, { "epoch": 5.33, "learning_rate": 9.435759857529174e-06, "loss": 0.1429, "step": 17313 }, { "epoch": 5.33, "learning_rate": 9.434764907827927e-06, "loss": 0.1749, "step": 17314 }, { "epoch": 5.33, "learning_rate": 9.433769963739987e-06, "loss": 0.1401, "step": 17315 }, { "epoch": 5.33, "learning_rate": 9.432775025275233e-06, "loss": 0.1507, "step": 17316 }, { "epoch": 5.33, "learning_rate": 9.431780092443543e-06, "loss": 0.1711, "step": 17317 }, { "epoch": 5.33, "learning_rate": 9.430785165254805e-06, "loss": 0.1998, "step": 17318 }, { "epoch": 5.33, "learning_rate": 9.429790243718892e-06, "loss": 0.1727, "step": 17319 }, { "epoch": 5.33, "learning_rate": 9.428795327845688e-06, "loss": 0.1422, "step": 17320 }, { "epoch": 5.33, "learning_rate": 9.427800417645073e-06, "loss": 0.1454, "step": 17321 }, { "epoch": 5.33, "learning_rate": 9.426805513126926e-06, "loss": 0.1141, "step": 17322 }, { "epoch": 5.33, "learning_rate": 9.425810614301132e-06, "loss": 0.0838, "step": 17323 }, { "epoch": 5.33, "learning_rate": 9.42481572117756e-06, "loss": 0.1427, "step": 17324 }, { "epoch": 5.33, "learning_rate": 9.423820833766108e-06, "loss": 0.1565, "step": 17325 }, { "epoch": 5.33, "learning_rate": 9.422825952076641e-06, "loss": 0.1274, "step": 17326 }, { "epoch": 5.33, "learning_rate": 9.421831076119043e-06, "loss": 0.1246, "step": 17327 }, { "epoch": 5.33, "learning_rate": 9.420836205903201e-06, "loss": 0.2059, "step": 17328 }, { "epoch": 5.33, "learning_rate": 9.41984134143898e-06, "loss": 0.2535, "step": 17329 }, { "epoch": 5.33, "learning_rate": 9.418846482736279e-06, "loss": 0.168, "step": 17330 }, { "epoch": 5.33, "learning_rate": 9.417851629804962e-06, "loss": 0.1789, "step": 17331 }, { "epoch": 5.33, "learning_rate": 9.416856782654915e-06, "loss": 0.1663, "step": 17332 }, { "epoch": 5.33, "learning_rate": 9.415861941296021e-06, "loss": 0.2136, "step": 17333 }, { "epoch": 5.33, "learning_rate": 9.414867105738152e-06, "loss": 0.2367, "step": 17334 }, { "epoch": 5.33, "learning_rate": 9.413872275991196e-06, "loss": 0.1434, "step": 17335 }, { "epoch": 5.33, "learning_rate": 9.412877452065028e-06, "loss": 0.17, "step": 17336 }, { "epoch": 5.33, "learning_rate": 9.411882633969525e-06, "loss": 0.1695, "step": 17337 }, { "epoch": 5.33, "learning_rate": 9.410887821714571e-06, "loss": 0.138, "step": 17338 }, { "epoch": 5.33, "learning_rate": 9.409893015310041e-06, "loss": 0.1365, "step": 17339 }, { "epoch": 5.34, "learning_rate": 9.408898214765824e-06, "loss": 0.1568, "step": 17340 }, { "epoch": 5.34, "learning_rate": 9.407903420091789e-06, "loss": 0.1555, "step": 17341 }, { "epoch": 5.34, "learning_rate": 9.406908631297816e-06, "loss": 0.1752, "step": 17342 }, { "epoch": 5.34, "learning_rate": 9.40591384839379e-06, "loss": 0.2179, "step": 17343 }, { "epoch": 5.34, "learning_rate": 9.404919071389586e-06, "loss": 0.1695, "step": 17344 }, { "epoch": 5.34, "learning_rate": 9.403924300295085e-06, "loss": 0.1301, "step": 17345 }, { "epoch": 5.34, "learning_rate": 9.402929535120166e-06, "loss": 0.2152, "step": 17346 }, { "epoch": 5.34, "learning_rate": 9.401934775874704e-06, "loss": 0.1045, "step": 17347 }, { "epoch": 5.34, "learning_rate": 9.400940022568582e-06, "loss": 0.0794, "step": 17348 }, { "epoch": 5.34, "learning_rate": 9.399945275211677e-06, "loss": 0.1412, "step": 17349 }, { "epoch": 5.34, "learning_rate": 9.39895053381387e-06, "loss": 0.2288, "step": 17350 }, { "epoch": 5.34, "learning_rate": 9.39795579838504e-06, "loss": 0.2352, "step": 17351 }, { "epoch": 5.34, "learning_rate": 9.39696106893506e-06, "loss": 0.3186, "step": 17352 }, { "epoch": 5.34, "learning_rate": 9.395966345473814e-06, "loss": 0.1037, "step": 17353 }, { "epoch": 5.34, "learning_rate": 9.394971628011177e-06, "loss": 0.1927, "step": 17354 }, { "epoch": 5.34, "learning_rate": 9.393976916557032e-06, "loss": 0.2699, "step": 17355 }, { "epoch": 5.34, "learning_rate": 9.392982211121255e-06, "loss": 0.2025, "step": 17356 }, { "epoch": 5.34, "learning_rate": 9.391987511713722e-06, "loss": 0.1976, "step": 17357 }, { "epoch": 5.34, "learning_rate": 9.390992818344314e-06, "loss": 0.0967, "step": 17358 }, { "epoch": 5.34, "learning_rate": 9.389998131022906e-06, "loss": 0.1175, "step": 17359 }, { "epoch": 5.34, "learning_rate": 9.389003449759383e-06, "loss": 0.1752, "step": 17360 }, { "epoch": 5.34, "learning_rate": 9.388008774563621e-06, "loss": 0.2775, "step": 17361 }, { "epoch": 5.34, "learning_rate": 9.387014105445488e-06, "loss": 0.2188, "step": 17362 }, { "epoch": 5.34, "learning_rate": 9.386019442414876e-06, "loss": 0.1769, "step": 17363 }, { "epoch": 5.34, "learning_rate": 9.385024785481653e-06, "loss": 0.1217, "step": 17364 }, { "epoch": 5.34, "learning_rate": 9.384030134655703e-06, "loss": 0.19, "step": 17365 }, { "epoch": 5.34, "learning_rate": 9.383035489946905e-06, "loss": 0.0975, "step": 17366 }, { "epoch": 5.34, "learning_rate": 9.382040851365125e-06, "loss": 0.1613, "step": 17367 }, { "epoch": 5.34, "learning_rate": 9.381046218920256e-06, "loss": 0.199, "step": 17368 }, { "epoch": 5.34, "learning_rate": 9.380051592622164e-06, "loss": 0.2244, "step": 17369 }, { "epoch": 5.34, "learning_rate": 9.379056972480732e-06, "loss": 0.1063, "step": 17370 }, { "epoch": 5.34, "learning_rate": 9.378062358505838e-06, "loss": 0.1293, "step": 17371 }, { "epoch": 5.35, "learning_rate": 9.377067750707354e-06, "loss": 0.0881, "step": 17372 }, { "epoch": 5.35, "learning_rate": 9.376073149095169e-06, "loss": 0.1174, "step": 17373 }, { "epoch": 5.35, "learning_rate": 9.375078553679143e-06, "loss": 0.1976, "step": 17374 }, { "epoch": 5.35, "learning_rate": 9.374083964469172e-06, "loss": 0.1051, "step": 17375 }, { "epoch": 5.35, "learning_rate": 9.37308938147512e-06, "loss": 0.2051, "step": 17376 }, { "epoch": 5.35, "learning_rate": 9.372094804706867e-06, "loss": 0.1712, "step": 17377 }, { "epoch": 5.35, "learning_rate": 9.371100234174295e-06, "loss": 0.1657, "step": 17378 }, { "epoch": 5.35, "learning_rate": 9.37010566988727e-06, "loss": 0.0879, "step": 17379 }, { "epoch": 5.35, "learning_rate": 9.369111111855686e-06, "loss": 0.1172, "step": 17380 }, { "epoch": 5.35, "learning_rate": 9.368116560089404e-06, "loss": 0.1778, "step": 17381 }, { "epoch": 5.35, "learning_rate": 9.367122014598307e-06, "loss": 0.1353, "step": 17382 }, { "epoch": 5.35, "learning_rate": 9.366127475392273e-06, "loss": 0.1218, "step": 17383 }, { "epoch": 5.35, "learning_rate": 9.365132942481177e-06, "loss": 0.1776, "step": 17384 }, { "epoch": 5.35, "learning_rate": 9.364138415874898e-06, "loss": 0.1351, "step": 17385 }, { "epoch": 5.35, "learning_rate": 9.363143895583308e-06, "loss": 0.3295, "step": 17386 }, { "epoch": 5.35, "learning_rate": 9.362149381616287e-06, "loss": 0.2126, "step": 17387 }, { "epoch": 5.35, "learning_rate": 9.36115487398371e-06, "loss": 0.0601, "step": 17388 }, { "epoch": 5.35, "learning_rate": 9.360160372695453e-06, "loss": 0.0842, "step": 17389 }, { "epoch": 5.35, "learning_rate": 9.359165877761396e-06, "loss": 0.1548, "step": 17390 }, { "epoch": 5.35, "learning_rate": 9.358171389191412e-06, "loss": 0.1274, "step": 17391 }, { "epoch": 5.35, "learning_rate": 9.357176906995374e-06, "loss": 0.1334, "step": 17392 }, { "epoch": 5.35, "learning_rate": 9.356182431183165e-06, "loss": 0.179, "step": 17393 }, { "epoch": 5.35, "learning_rate": 9.355187961764655e-06, "loss": 0.2521, "step": 17394 }, { "epoch": 5.35, "learning_rate": 9.354193498749728e-06, "loss": 0.1409, "step": 17395 }, { "epoch": 5.35, "learning_rate": 9.353199042148252e-06, "loss": 0.1959, "step": 17396 }, { "epoch": 5.35, "learning_rate": 9.352204591970105e-06, "loss": 0.0933, "step": 17397 }, { "epoch": 5.35, "learning_rate": 9.351210148225162e-06, "loss": 0.105, "step": 17398 }, { "epoch": 5.35, "learning_rate": 9.350215710923301e-06, "loss": 0.2096, "step": 17399 }, { "epoch": 5.35, "learning_rate": 9.3492212800744e-06, "loss": 0.1396, "step": 17400 }, { "epoch": 5.35, "learning_rate": 9.34822685568833e-06, "loss": 0.2328, "step": 17401 }, { "epoch": 5.35, "learning_rate": 9.347232437774966e-06, "loss": 0.1521, "step": 17402 }, { "epoch": 5.35, "learning_rate": 9.346238026344186e-06, "loss": 0.1896, "step": 17403 }, { "epoch": 5.35, "learning_rate": 9.345243621405865e-06, "loss": 0.2717, "step": 17404 }, { "epoch": 5.36, "learning_rate": 9.344249222969878e-06, "loss": 0.1043, "step": 17405 }, { "epoch": 5.36, "learning_rate": 9.343254831046105e-06, "loss": 0.2264, "step": 17406 }, { "epoch": 5.36, "learning_rate": 9.342260445644412e-06, "loss": 0.141, "step": 17407 }, { "epoch": 5.36, "learning_rate": 9.34126606677468e-06, "loss": 0.1109, "step": 17408 }, { "epoch": 5.36, "learning_rate": 9.340271694446782e-06, "loss": 0.2203, "step": 17409 }, { "epoch": 5.36, "learning_rate": 9.339277328670596e-06, "loss": 0.114, "step": 17410 }, { "epoch": 5.36, "learning_rate": 9.338282969455997e-06, "loss": 0.1786, "step": 17411 }, { "epoch": 5.36, "learning_rate": 9.337288616812852e-06, "loss": 0.2428, "step": 17412 }, { "epoch": 5.36, "learning_rate": 9.336294270751047e-06, "loss": 0.2096, "step": 17413 }, { "epoch": 5.36, "learning_rate": 9.335299931280448e-06, "loss": 0.1245, "step": 17414 }, { "epoch": 5.36, "learning_rate": 9.334305598410937e-06, "loss": 0.1976, "step": 17415 }, { "epoch": 5.36, "learning_rate": 9.333311272152385e-06, "loss": 0.2566, "step": 17416 }, { "epoch": 5.36, "learning_rate": 9.33231695251466e-06, "loss": 0.1631, "step": 17417 }, { "epoch": 5.36, "learning_rate": 9.331322639507651e-06, "loss": 0.2815, "step": 17418 }, { "epoch": 5.36, "learning_rate": 9.330328333141219e-06, "loss": 0.0699, "step": 17419 }, { "epoch": 5.36, "learning_rate": 9.329334033425245e-06, "loss": 0.2084, "step": 17420 }, { "epoch": 5.36, "learning_rate": 9.328339740369605e-06, "loss": 0.1405, "step": 17421 }, { "epoch": 5.36, "learning_rate": 9.327345453984167e-06, "loss": 0.206, "step": 17422 }, { "epoch": 5.36, "learning_rate": 9.326351174278813e-06, "loss": 0.1484, "step": 17423 }, { "epoch": 5.36, "learning_rate": 9.325356901263407e-06, "loss": 0.2455, "step": 17424 }, { "epoch": 5.36, "learning_rate": 9.324362634947833e-06, "loss": 0.1498, "step": 17425 }, { "epoch": 5.36, "learning_rate": 9.32336837534196e-06, "loss": 0.1809, "step": 17426 }, { "epoch": 5.36, "learning_rate": 9.322374122455658e-06, "loss": 0.1396, "step": 17427 }, { "epoch": 5.36, "learning_rate": 9.321379876298813e-06, "loss": 0.1832, "step": 17428 }, { "epoch": 5.36, "learning_rate": 9.320385636881283e-06, "loss": 0.2136, "step": 17429 }, { "epoch": 5.36, "learning_rate": 9.319391404212957e-06, "loss": 0.1125, "step": 17430 }, { "epoch": 5.36, "learning_rate": 9.3183971783037e-06, "loss": 0.1713, "step": 17431 }, { "epoch": 5.36, "learning_rate": 9.317402959163384e-06, "loss": 0.2065, "step": 17432 }, { "epoch": 5.36, "learning_rate": 9.31640874680189e-06, "loss": 0.127, "step": 17433 }, { "epoch": 5.36, "learning_rate": 9.315414541229082e-06, "loss": 0.1614, "step": 17434 }, { "epoch": 5.36, "learning_rate": 9.314420342454844e-06, "loss": 0.1632, "step": 17435 }, { "epoch": 5.36, "learning_rate": 9.313426150489043e-06, "loss": 0.1384, "step": 17436 }, { "epoch": 5.37, "learning_rate": 9.31243196534155e-06, "loss": 0.2236, "step": 17437 }, { "epoch": 5.37, "learning_rate": 9.311437787022243e-06, "loss": 0.204, "step": 17438 }, { "epoch": 5.37, "learning_rate": 9.310443615540992e-06, "loss": 0.1618, "step": 17439 }, { "epoch": 5.37, "learning_rate": 9.309449450907674e-06, "loss": 0.1166, "step": 17440 }, { "epoch": 5.37, "learning_rate": 9.308455293132156e-06, "loss": 0.132, "step": 17441 }, { "epoch": 5.37, "learning_rate": 9.307461142224318e-06, "loss": 0.1963, "step": 17442 }, { "epoch": 5.37, "learning_rate": 9.306466998194029e-06, "loss": 0.1575, "step": 17443 }, { "epoch": 5.37, "learning_rate": 9.305472861051157e-06, "loss": 0.2131, "step": 17444 }, { "epoch": 5.37, "learning_rate": 9.304478730805587e-06, "loss": 0.1571, "step": 17445 }, { "epoch": 5.37, "learning_rate": 9.303484607467176e-06, "loss": 0.2116, "step": 17446 }, { "epoch": 5.37, "learning_rate": 9.302490491045812e-06, "loss": 0.1608, "step": 17447 }, { "epoch": 5.37, "learning_rate": 9.301496381551358e-06, "loss": 0.1433, "step": 17448 }, { "epoch": 5.37, "learning_rate": 9.300502278993687e-06, "loss": 0.2061, "step": 17449 }, { "epoch": 5.37, "learning_rate": 9.299508183382674e-06, "loss": 0.2467, "step": 17450 }, { "epoch": 5.37, "learning_rate": 9.298514094728188e-06, "loss": 0.1672, "step": 17451 }, { "epoch": 5.37, "learning_rate": 9.29752001304011e-06, "loss": 0.0976, "step": 17452 }, { "epoch": 5.37, "learning_rate": 9.2965259383283e-06, "loss": 0.1865, "step": 17453 }, { "epoch": 5.37, "learning_rate": 9.295531870602636e-06, "loss": 0.1317, "step": 17454 }, { "epoch": 5.37, "learning_rate": 9.29453780987299e-06, "loss": 0.1884, "step": 17455 }, { "epoch": 5.37, "learning_rate": 9.293543756149235e-06, "loss": 0.1752, "step": 17456 }, { "epoch": 5.37, "learning_rate": 9.292549709441243e-06, "loss": 0.2092, "step": 17457 }, { "epoch": 5.37, "learning_rate": 9.291555669758883e-06, "loss": 0.2418, "step": 17458 }, { "epoch": 5.37, "learning_rate": 9.290561637112027e-06, "loss": 0.145, "step": 17459 }, { "epoch": 5.37, "learning_rate": 9.289567611510549e-06, "loss": 0.0758, "step": 17460 }, { "epoch": 5.37, "learning_rate": 9.288573592964318e-06, "loss": 0.1166, "step": 17461 }, { "epoch": 5.37, "learning_rate": 9.287579581483209e-06, "loss": 0.1437, "step": 17462 }, { "epoch": 5.37, "learning_rate": 9.286585577077092e-06, "loss": 0.1423, "step": 17463 }, { "epoch": 5.37, "learning_rate": 9.285591579755834e-06, "loss": 0.0805, "step": 17464 }, { "epoch": 5.37, "learning_rate": 9.284597589529313e-06, "loss": 0.1115, "step": 17465 }, { "epoch": 5.37, "learning_rate": 9.283603606407394e-06, "loss": 0.109, "step": 17466 }, { "epoch": 5.37, "learning_rate": 9.282609630399954e-06, "loss": 0.2548, "step": 17467 }, { "epoch": 5.37, "learning_rate": 9.281615661516866e-06, "loss": 0.1862, "step": 17468 }, { "epoch": 5.37, "learning_rate": 9.280621699767988e-06, "loss": 0.1999, "step": 17469 }, { "epoch": 5.38, "learning_rate": 9.279627745163204e-06, "loss": 0.1871, "step": 17470 }, { "epoch": 5.38, "learning_rate": 9.278633797712379e-06, "loss": 0.2067, "step": 17471 }, { "epoch": 5.38, "learning_rate": 9.277639857425387e-06, "loss": 0.1997, "step": 17472 }, { "epoch": 5.38, "learning_rate": 9.276645924312098e-06, "loss": 0.1823, "step": 17473 }, { "epoch": 5.38, "learning_rate": 9.275651998382377e-06, "loss": 0.1566, "step": 17474 }, { "epoch": 5.38, "learning_rate": 9.274658079646104e-06, "loss": 0.2449, "step": 17475 }, { "epoch": 5.38, "learning_rate": 9.27366416811314e-06, "loss": 0.221, "step": 17476 }, { "epoch": 5.38, "learning_rate": 9.272670263793363e-06, "loss": 0.1274, "step": 17477 }, { "epoch": 5.38, "learning_rate": 9.271676366696644e-06, "loss": 0.1923, "step": 17478 }, { "epoch": 5.38, "learning_rate": 9.270682476832842e-06, "loss": 0.1282, "step": 17479 }, { "epoch": 5.38, "learning_rate": 9.269688594211842e-06, "loss": 0.2364, "step": 17480 }, { "epoch": 5.38, "learning_rate": 9.268694718843503e-06, "loss": 0.2117, "step": 17481 }, { "epoch": 5.38, "learning_rate": 9.267700850737703e-06, "loss": 0.2022, "step": 17482 }, { "epoch": 5.38, "learning_rate": 9.266706989904305e-06, "loss": 0.151, "step": 17483 }, { "epoch": 5.38, "learning_rate": 9.265713136353184e-06, "loss": 0.2545, "step": 17484 }, { "epoch": 5.38, "learning_rate": 9.26471929009421e-06, "loss": 0.1237, "step": 17485 }, { "epoch": 5.38, "learning_rate": 9.263725451137248e-06, "loss": 0.1985, "step": 17486 }, { "epoch": 5.38, "learning_rate": 9.262731619492174e-06, "loss": 0.0905, "step": 17487 }, { "epoch": 5.38, "learning_rate": 9.261737795168853e-06, "loss": 0.1524, "step": 17488 }, { "epoch": 5.38, "learning_rate": 9.260743978177155e-06, "loss": 0.1614, "step": 17489 }, { "epoch": 5.38, "learning_rate": 9.259750168526953e-06, "loss": 0.1876, "step": 17490 }, { "epoch": 5.38, "learning_rate": 9.25875636622811e-06, "loss": 0.1239, "step": 17491 }, { "epoch": 5.38, "learning_rate": 9.257762571290505e-06, "loss": 0.2052, "step": 17492 }, { "epoch": 5.38, "learning_rate": 9.256768783723999e-06, "loss": 0.2531, "step": 17493 }, { "epoch": 5.38, "learning_rate": 9.255775003538462e-06, "loss": 0.0643, "step": 17494 }, { "epoch": 5.38, "learning_rate": 9.254781230743768e-06, "loss": 0.1429, "step": 17495 }, { "epoch": 5.38, "learning_rate": 9.25378746534978e-06, "loss": 0.1681, "step": 17496 }, { "epoch": 5.38, "learning_rate": 9.252793707366374e-06, "loss": 0.1629, "step": 17497 }, { "epoch": 5.38, "learning_rate": 9.251799956803414e-06, "loss": 0.0978, "step": 17498 }, { "epoch": 5.38, "learning_rate": 9.250806213670767e-06, "loss": 0.0931, "step": 17499 }, { "epoch": 5.38, "learning_rate": 9.249812477978309e-06, "loss": 0.1614, "step": 17500 }, { "epoch": 5.38, "learning_rate": 9.2488187497359e-06, "loss": 0.1997, "step": 17501 }, { "epoch": 5.39, "learning_rate": 9.247825028953419e-06, "loss": 0.1383, "step": 17502 }, { "epoch": 5.39, "learning_rate": 9.246831315640727e-06, "loss": 0.2097, "step": 17503 }, { "epoch": 5.39, "learning_rate": 9.24583760980769e-06, "loss": 0.2192, "step": 17504 }, { "epoch": 5.39, "learning_rate": 9.244843911464185e-06, "loss": 0.188, "step": 17505 }, { "epoch": 5.39, "learning_rate": 9.243850220620071e-06, "loss": 0.2176, "step": 17506 }, { "epoch": 5.39, "learning_rate": 9.242856537285227e-06, "loss": 0.1766, "step": 17507 }, { "epoch": 5.39, "learning_rate": 9.241862861469514e-06, "loss": 0.1846, "step": 17508 }, { "epoch": 5.39, "learning_rate": 9.2408691931828e-06, "loss": 0.1921, "step": 17509 }, { "epoch": 5.39, "learning_rate": 9.239875532434954e-06, "loss": 0.1561, "step": 17510 }, { "epoch": 5.39, "learning_rate": 9.238881879235844e-06, "loss": 0.0745, "step": 17511 }, { "epoch": 5.39, "learning_rate": 9.23788823359534e-06, "loss": 0.2009, "step": 17512 }, { "epoch": 5.39, "learning_rate": 9.23689459552331e-06, "loss": 0.2881, "step": 17513 }, { "epoch": 5.39, "learning_rate": 9.235900965029616e-06, "loss": 0.1816, "step": 17514 }, { "epoch": 5.39, "learning_rate": 9.234907342124134e-06, "loss": 0.1621, "step": 17515 }, { "epoch": 5.39, "learning_rate": 9.233913726816721e-06, "loss": 0.1712, "step": 17516 }, { "epoch": 5.39, "learning_rate": 9.232920119117256e-06, "loss": 0.1327, "step": 17517 }, { "epoch": 5.39, "learning_rate": 9.231926519035601e-06, "loss": 0.1233, "step": 17518 }, { "epoch": 5.39, "learning_rate": 9.230932926581622e-06, "loss": 0.221, "step": 17519 }, { "epoch": 5.39, "learning_rate": 9.229939341765188e-06, "loss": 0.111, "step": 17520 }, { "epoch": 5.39, "learning_rate": 9.228945764596164e-06, "loss": 0.0985, "step": 17521 }, { "epoch": 5.39, "learning_rate": 9.227952195084421e-06, "loss": 0.1388, "step": 17522 }, { "epoch": 5.39, "learning_rate": 9.226958633239829e-06, "loss": 0.1047, "step": 17523 }, { "epoch": 5.39, "learning_rate": 9.225965079072244e-06, "loss": 0.1717, "step": 17524 }, { "epoch": 5.39, "learning_rate": 9.224971532591542e-06, "loss": 0.1395, "step": 17525 }, { "epoch": 5.39, "learning_rate": 9.223977993807584e-06, "loss": 0.1299, "step": 17526 }, { "epoch": 5.39, "learning_rate": 9.222984462730241e-06, "loss": 0.2026, "step": 17527 }, { "epoch": 5.39, "learning_rate": 9.221990939369381e-06, "loss": 0.1519, "step": 17528 }, { "epoch": 5.39, "learning_rate": 9.220997423734863e-06, "loss": 0.1619, "step": 17529 }, { "epoch": 5.39, "learning_rate": 9.220003915836564e-06, "loss": 0.1489, "step": 17530 }, { "epoch": 5.39, "learning_rate": 9.219010415684341e-06, "loss": 0.0974, "step": 17531 }, { "epoch": 5.39, "learning_rate": 9.218016923288067e-06, "loss": 0.203, "step": 17532 }, { "epoch": 5.39, "learning_rate": 9.217023438657606e-06, "loss": 0.2557, "step": 17533 }, { "epoch": 5.39, "learning_rate": 9.216029961802819e-06, "loss": 0.1046, "step": 17534 }, { "epoch": 5.4, "learning_rate": 9.215036492733583e-06, "loss": 0.0972, "step": 17535 }, { "epoch": 5.4, "learning_rate": 9.214043031459754e-06, "loss": 0.2027, "step": 17536 }, { "epoch": 5.4, "learning_rate": 9.213049577991204e-06, "loss": 0.1557, "step": 17537 }, { "epoch": 5.4, "learning_rate": 9.212056132337798e-06, "loss": 0.18, "step": 17538 }, { "epoch": 5.4, "learning_rate": 9.211062694509396e-06, "loss": 0.1202, "step": 17539 }, { "epoch": 5.4, "learning_rate": 9.210069264515876e-06, "loss": 0.2837, "step": 17540 }, { "epoch": 5.4, "learning_rate": 9.209075842367088e-06, "loss": 0.233, "step": 17541 }, { "epoch": 5.4, "learning_rate": 9.208082428072913e-06, "loss": 0.1476, "step": 17542 }, { "epoch": 5.4, "learning_rate": 9.207089021643207e-06, "loss": 0.1085, "step": 17543 }, { "epoch": 5.4, "learning_rate": 9.206095623087834e-06, "loss": 0.152, "step": 17544 }, { "epoch": 5.4, "learning_rate": 9.205102232416669e-06, "loss": 0.1683, "step": 17545 }, { "epoch": 5.4, "learning_rate": 9.204108849639565e-06, "loss": 0.1615, "step": 17546 }, { "epoch": 5.4, "learning_rate": 9.2031154747664e-06, "loss": 0.2161, "step": 17547 }, { "epoch": 5.4, "learning_rate": 9.20212210780703e-06, "loss": 0.213, "step": 17548 }, { "epoch": 5.4, "learning_rate": 9.201128748771322e-06, "loss": 0.1602, "step": 17549 }, { "epoch": 5.4, "learning_rate": 9.200135397669144e-06, "loss": 0.2073, "step": 17550 }, { "epoch": 5.4, "learning_rate": 9.199142054510354e-06, "loss": 0.1728, "step": 17551 }, { "epoch": 5.4, "learning_rate": 9.198148719304828e-06, "loss": 0.2417, "step": 17552 }, { "epoch": 5.4, "learning_rate": 9.197155392062421e-06, "loss": 0.1631, "step": 17553 }, { "epoch": 5.4, "learning_rate": 9.196162072793e-06, "loss": 0.1542, "step": 17554 }, { "epoch": 5.4, "learning_rate": 9.195168761506432e-06, "loss": 0.1235, "step": 17555 }, { "epoch": 5.4, "learning_rate": 9.194175458212577e-06, "loss": 0.1497, "step": 17556 }, { "epoch": 5.4, "learning_rate": 9.193182162921307e-06, "loss": 0.1069, "step": 17557 }, { "epoch": 5.4, "learning_rate": 9.19218887564248e-06, "loss": 0.1509, "step": 17558 }, { "epoch": 5.4, "learning_rate": 9.19119559638596e-06, "loss": 0.1747, "step": 17559 }, { "epoch": 5.4, "learning_rate": 9.190202325161614e-06, "loss": 0.1226, "step": 17560 }, { "epoch": 5.4, "learning_rate": 9.189209061979305e-06, "loss": 0.1689, "step": 17561 }, { "epoch": 5.4, "learning_rate": 9.188215806848898e-06, "loss": 0.1671, "step": 17562 }, { "epoch": 5.4, "learning_rate": 9.187222559780259e-06, "loss": 0.1721, "step": 17563 }, { "epoch": 5.4, "learning_rate": 9.186229320783244e-06, "loss": 0.1179, "step": 17564 }, { "epoch": 5.4, "learning_rate": 9.185236089867726e-06, "loss": 0.2001, "step": 17565 }, { "epoch": 5.4, "learning_rate": 9.18424286704356e-06, "loss": 0.2408, "step": 17566 }, { "epoch": 5.41, "learning_rate": 9.183249652320618e-06, "loss": 0.1494, "step": 17567 }, { "epoch": 5.41, "learning_rate": 9.18225644570876e-06, "loss": 0.2137, "step": 17568 }, { "epoch": 5.41, "learning_rate": 9.181263247217847e-06, "loss": 0.1825, "step": 17569 }, { "epoch": 5.41, "learning_rate": 9.180270056857746e-06, "loss": 0.1506, "step": 17570 }, { "epoch": 5.41, "learning_rate": 9.179276874638315e-06, "loss": 0.1307, "step": 17571 }, { "epoch": 5.41, "learning_rate": 9.178283700569424e-06, "loss": 0.1427, "step": 17572 }, { "epoch": 5.41, "learning_rate": 9.177290534660937e-06, "loss": 0.1877, "step": 17573 }, { "epoch": 5.41, "learning_rate": 9.176297376922706e-06, "loss": 0.1465, "step": 17574 }, { "epoch": 5.41, "learning_rate": 9.175304227364606e-06, "loss": 0.1374, "step": 17575 }, { "epoch": 5.41, "learning_rate": 9.174311085996492e-06, "loss": 0.1367, "step": 17576 }, { "epoch": 5.41, "learning_rate": 9.173317952828232e-06, "loss": 0.1616, "step": 17577 }, { "epoch": 5.41, "learning_rate": 9.17232482786969e-06, "loss": 0.1497, "step": 17578 }, { "epoch": 5.41, "learning_rate": 9.171331711130717e-06, "loss": 0.1474, "step": 17579 }, { "epoch": 5.41, "learning_rate": 9.17033860262119e-06, "loss": 0.1773, "step": 17580 }, { "epoch": 5.41, "learning_rate": 9.169345502350962e-06, "loss": 0.0974, "step": 17581 }, { "epoch": 5.41, "learning_rate": 9.1683524103299e-06, "loss": 0.1327, "step": 17582 }, { "epoch": 5.41, "learning_rate": 9.167359326567867e-06, "loss": 0.2325, "step": 17583 }, { "epoch": 5.41, "learning_rate": 9.166366251074719e-06, "loss": 0.2568, "step": 17584 }, { "epoch": 5.41, "learning_rate": 9.165373183860329e-06, "loss": 0.2756, "step": 17585 }, { "epoch": 5.41, "learning_rate": 9.164380124934546e-06, "loss": 0.2445, "step": 17586 }, { "epoch": 5.41, "learning_rate": 9.163387074307241e-06, "loss": 0.2421, "step": 17587 }, { "epoch": 5.41, "learning_rate": 9.162394031988275e-06, "loss": 0.1172, "step": 17588 }, { "epoch": 5.41, "learning_rate": 9.161400997987505e-06, "loss": 0.1967, "step": 17589 }, { "epoch": 5.41, "learning_rate": 9.160407972314799e-06, "loss": 0.2373, "step": 17590 }, { "epoch": 5.41, "learning_rate": 9.159414954980011e-06, "loss": 0.1752, "step": 17591 }, { "epoch": 5.41, "learning_rate": 9.158421945993014e-06, "loss": 0.1744, "step": 17592 }, { "epoch": 5.41, "learning_rate": 9.15742894536366e-06, "loss": 0.1362, "step": 17593 }, { "epoch": 5.41, "learning_rate": 9.15643595310181e-06, "loss": 0.3315, "step": 17594 }, { "epoch": 5.41, "learning_rate": 9.155442969217333e-06, "loss": 0.1654, "step": 17595 }, { "epoch": 5.41, "learning_rate": 9.154449993720081e-06, "loss": 0.1104, "step": 17596 }, { "epoch": 5.41, "learning_rate": 9.153457026619923e-06, "loss": 0.18, "step": 17597 }, { "epoch": 5.41, "learning_rate": 9.152464067926717e-06, "loss": 0.1279, "step": 17598 }, { "epoch": 5.41, "learning_rate": 9.151471117650321e-06, "loss": 0.1761, "step": 17599 }, { "epoch": 5.42, "learning_rate": 9.150478175800602e-06, "loss": 0.1795, "step": 17600 }, { "epoch": 5.42, "learning_rate": 9.149485242387414e-06, "loss": 0.2299, "step": 17601 }, { "epoch": 5.42, "learning_rate": 9.148492317420627e-06, "loss": 0.1587, "step": 17602 }, { "epoch": 5.42, "learning_rate": 9.147499400910093e-06, "loss": 0.1844, "step": 17603 }, { "epoch": 5.42, "learning_rate": 9.146506492865674e-06, "loss": 0.2095, "step": 17604 }, { "epoch": 5.42, "learning_rate": 9.145513593297234e-06, "loss": 0.0855, "step": 17605 }, { "epoch": 5.42, "learning_rate": 9.144520702214632e-06, "loss": 0.1683, "step": 17606 }, { "epoch": 5.42, "learning_rate": 9.143527819627729e-06, "loss": 0.1144, "step": 17607 }, { "epoch": 5.42, "learning_rate": 9.142534945546381e-06, "loss": 0.2301, "step": 17608 }, { "epoch": 5.42, "learning_rate": 9.141542079980453e-06, "loss": 0.1508, "step": 17609 }, { "epoch": 5.42, "learning_rate": 9.140549222939803e-06, "loss": 0.1078, "step": 17610 }, { "epoch": 5.42, "learning_rate": 9.139556374434288e-06, "loss": 0.1285, "step": 17611 }, { "epoch": 5.42, "learning_rate": 9.138563534473776e-06, "loss": 0.1362, "step": 17612 }, { "epoch": 5.42, "learning_rate": 9.137570703068122e-06, "loss": 0.1525, "step": 17613 }, { "epoch": 5.42, "learning_rate": 9.136577880227183e-06, "loss": 0.2801, "step": 17614 }, { "epoch": 5.42, "learning_rate": 9.135585065960824e-06, "loss": 0.1231, "step": 17615 }, { "epoch": 5.42, "learning_rate": 9.134592260278898e-06, "loss": 0.2113, "step": 17616 }, { "epoch": 5.42, "learning_rate": 9.133599463191272e-06, "loss": 0.1872, "step": 17617 }, { "epoch": 5.42, "learning_rate": 9.132606674707803e-06, "loss": 0.1421, "step": 17618 }, { "epoch": 5.42, "learning_rate": 9.131613894838345e-06, "loss": 0.2138, "step": 17619 }, { "epoch": 5.42, "learning_rate": 9.130621123592764e-06, "loss": 0.0893, "step": 17620 }, { "epoch": 5.42, "learning_rate": 9.129628360980915e-06, "loss": 0.1793, "step": 17621 }, { "epoch": 5.42, "learning_rate": 9.12863560701266e-06, "loss": 0.2147, "step": 17622 }, { "epoch": 5.42, "learning_rate": 9.127642861697858e-06, "loss": 0.1175, "step": 17623 }, { "epoch": 5.42, "learning_rate": 9.126650125046361e-06, "loss": 0.1987, "step": 17624 }, { "epoch": 5.42, "learning_rate": 9.125657397068038e-06, "loss": 0.1229, "step": 17625 }, { "epoch": 5.42, "learning_rate": 9.12466467777274e-06, "loss": 0.1445, "step": 17626 }, { "epoch": 5.42, "learning_rate": 9.123671967170328e-06, "loss": 0.1797, "step": 17627 }, { "epoch": 5.42, "learning_rate": 9.122679265270666e-06, "loss": 0.1137, "step": 17628 }, { "epoch": 5.42, "learning_rate": 9.1216865720836e-06, "loss": 0.2001, "step": 17629 }, { "epoch": 5.42, "learning_rate": 9.120693887619003e-06, "loss": 0.1273, "step": 17630 }, { "epoch": 5.42, "learning_rate": 9.119701211886723e-06, "loss": 0.1422, "step": 17631 }, { "epoch": 5.43, "learning_rate": 9.11870854489662e-06, "loss": 0.1388, "step": 17632 }, { "epoch": 5.43, "learning_rate": 9.117715886658558e-06, "loss": 0.1255, "step": 17633 }, { "epoch": 5.43, "learning_rate": 9.116723237182383e-06, "loss": 0.1546, "step": 17634 }, { "epoch": 5.43, "learning_rate": 9.115730596477967e-06, "loss": 0.1989, "step": 17635 }, { "epoch": 5.43, "learning_rate": 9.114737964555157e-06, "loss": 0.1738, "step": 17636 }, { "epoch": 5.43, "learning_rate": 9.113745341423816e-06, "loss": 0.1501, "step": 17637 }, { "epoch": 5.43, "learning_rate": 9.112752727093799e-06, "loss": 0.1151, "step": 17638 }, { "epoch": 5.43, "learning_rate": 9.111760121574968e-06, "loss": 0.1659, "step": 17639 }, { "epoch": 5.43, "learning_rate": 9.11076752487718e-06, "loss": 0.1701, "step": 17640 }, { "epoch": 5.43, "learning_rate": 9.109774937010283e-06, "loss": 0.1289, "step": 17641 }, { "epoch": 5.43, "learning_rate": 9.108782357984147e-06, "loss": 0.1621, "step": 17642 }, { "epoch": 5.43, "learning_rate": 9.107789787808621e-06, "loss": 0.1389, "step": 17643 }, { "epoch": 5.43, "learning_rate": 9.106797226493565e-06, "loss": 0.1383, "step": 17644 }, { "epoch": 5.43, "learning_rate": 9.105804674048839e-06, "loss": 0.1185, "step": 17645 }, { "epoch": 5.43, "learning_rate": 9.10481213048429e-06, "loss": 0.2496, "step": 17646 }, { "epoch": 5.43, "learning_rate": 9.103819595809788e-06, "loss": 0.1539, "step": 17647 }, { "epoch": 5.43, "learning_rate": 9.102827070035179e-06, "loss": 0.1283, "step": 17648 }, { "epoch": 5.43, "learning_rate": 9.101834553170326e-06, "loss": 0.1858, "step": 17649 }, { "epoch": 5.43, "learning_rate": 9.100842045225084e-06, "loss": 0.226, "step": 17650 }, { "epoch": 5.43, "learning_rate": 9.099849546209307e-06, "loss": 0.1333, "step": 17651 }, { "epoch": 5.43, "learning_rate": 9.098857056132857e-06, "loss": 0.2653, "step": 17652 }, { "epoch": 5.43, "learning_rate": 9.097864575005583e-06, "loss": 0.084, "step": 17653 }, { "epoch": 5.43, "learning_rate": 9.096872102837348e-06, "loss": 0.1033, "step": 17654 }, { "epoch": 5.43, "learning_rate": 9.095879639638006e-06, "loss": 0.2649, "step": 17655 }, { "epoch": 5.43, "learning_rate": 9.094887185417409e-06, "loss": 0.167, "step": 17656 }, { "epoch": 5.43, "learning_rate": 9.093894740185421e-06, "loss": 0.204, "step": 17657 }, { "epoch": 5.43, "learning_rate": 9.092902303951889e-06, "loss": 0.2099, "step": 17658 }, { "epoch": 5.43, "learning_rate": 9.091909876726676e-06, "loss": 0.2639, "step": 17659 }, { "epoch": 5.43, "learning_rate": 9.090917458519635e-06, "loss": 0.0957, "step": 17660 }, { "epoch": 5.43, "learning_rate": 9.089925049340619e-06, "loss": 0.1536, "step": 17661 }, { "epoch": 5.43, "learning_rate": 9.08893264919949e-06, "loss": 0.166, "step": 17662 }, { "epoch": 5.43, "learning_rate": 9.087940258106093e-06, "loss": 0.2053, "step": 17663 }, { "epoch": 5.43, "learning_rate": 9.086947876070295e-06, "loss": 0.1815, "step": 17664 }, { "epoch": 5.44, "learning_rate": 9.085955503101947e-06, "loss": 0.2322, "step": 17665 }, { "epoch": 5.44, "learning_rate": 9.084963139210899e-06, "loss": 0.2835, "step": 17666 }, { "epoch": 5.44, "learning_rate": 9.083970784407012e-06, "loss": 0.1716, "step": 17667 }, { "epoch": 5.44, "learning_rate": 9.082978438700138e-06, "loss": 0.1227, "step": 17668 }, { "epoch": 5.44, "learning_rate": 9.081986102100137e-06, "loss": 0.2011, "step": 17669 }, { "epoch": 5.44, "learning_rate": 9.080993774616859e-06, "loss": 0.2086, "step": 17670 }, { "epoch": 5.44, "learning_rate": 9.080001456260157e-06, "loss": 0.2525, "step": 17671 }, { "epoch": 5.44, "learning_rate": 9.079009147039892e-06, "loss": 0.1387, "step": 17672 }, { "epoch": 5.44, "learning_rate": 9.07801684696591e-06, "loss": 0.1159, "step": 17673 }, { "epoch": 5.44, "learning_rate": 9.077024556048077e-06, "loss": 0.1603, "step": 17674 }, { "epoch": 5.44, "learning_rate": 9.076032274296238e-06, "loss": 0.151, "step": 17675 }, { "epoch": 5.44, "learning_rate": 9.075040001720247e-06, "loss": 0.1835, "step": 17676 }, { "epoch": 5.44, "learning_rate": 9.074047738329965e-06, "loss": 0.1101, "step": 17677 }, { "epoch": 5.44, "learning_rate": 9.073055484135238e-06, "loss": 0.195, "step": 17678 }, { "epoch": 5.44, "learning_rate": 9.072063239145929e-06, "loss": 0.1801, "step": 17679 }, { "epoch": 5.44, "learning_rate": 9.071071003371888e-06, "loss": 0.1727, "step": 17680 }, { "epoch": 5.44, "learning_rate": 9.070078776822965e-06, "loss": 0.2113, "step": 17681 }, { "epoch": 5.44, "learning_rate": 9.069086559509017e-06, "loss": 0.2242, "step": 17682 }, { "epoch": 5.44, "learning_rate": 9.068094351439896e-06, "loss": 0.1745, "step": 17683 }, { "epoch": 5.44, "learning_rate": 9.06710215262546e-06, "loss": 0.1026, "step": 17684 }, { "epoch": 5.44, "learning_rate": 9.06610996307556e-06, "loss": 0.239, "step": 17685 }, { "epoch": 5.44, "learning_rate": 9.065117782800046e-06, "loss": 0.1203, "step": 17686 }, { "epoch": 5.44, "learning_rate": 9.064125611808776e-06, "loss": 0.2117, "step": 17687 }, { "epoch": 5.44, "learning_rate": 9.063133450111599e-06, "loss": 0.1437, "step": 17688 }, { "epoch": 5.44, "learning_rate": 9.062141297718372e-06, "loss": 0.1238, "step": 17689 }, { "epoch": 5.44, "learning_rate": 9.061149154638949e-06, "loss": 0.2052, "step": 17690 }, { "epoch": 5.44, "learning_rate": 9.060157020883173e-06, "loss": 0.1638, "step": 17691 }, { "epoch": 5.44, "learning_rate": 9.05916489646091e-06, "loss": 0.1876, "step": 17692 }, { "epoch": 5.44, "learning_rate": 9.058172781382003e-06, "loss": 0.1527, "step": 17693 }, { "epoch": 5.44, "learning_rate": 9.05718067565631e-06, "loss": 0.2523, "step": 17694 }, { "epoch": 5.44, "learning_rate": 9.056188579293685e-06, "loss": 0.1284, "step": 17695 }, { "epoch": 5.44, "learning_rate": 9.05519649230397e-06, "loss": 0.1755, "step": 17696 }, { "epoch": 5.45, "learning_rate": 9.054204414697033e-06, "loss": 0.1188, "step": 17697 }, { "epoch": 5.45, "learning_rate": 9.053212346482711e-06, "loss": 0.1945, "step": 17698 }, { "epoch": 5.45, "learning_rate": 9.052220287670865e-06, "loss": 0.1229, "step": 17699 }, { "epoch": 5.45, "learning_rate": 9.051228238271346e-06, "loss": 0.1438, "step": 17700 }, { "epoch": 5.45, "learning_rate": 9.050236198294003e-06, "loss": 0.1907, "step": 17701 }, { "epoch": 5.45, "learning_rate": 9.049244167748694e-06, "loss": 0.2304, "step": 17702 }, { "epoch": 5.45, "learning_rate": 9.048252146645262e-06, "loss": 0.1384, "step": 17703 }, { "epoch": 5.45, "learning_rate": 9.047260134993566e-06, "loss": 0.1191, "step": 17704 }, { "epoch": 5.45, "learning_rate": 9.046268132803456e-06, "loss": 0.1998, "step": 17705 }, { "epoch": 5.45, "learning_rate": 9.045276140084778e-06, "loss": 0.1491, "step": 17706 }, { "epoch": 5.45, "learning_rate": 9.044284156847393e-06, "loss": 0.2163, "step": 17707 }, { "epoch": 5.45, "learning_rate": 9.04329218310114e-06, "loss": 0.2626, "step": 17708 }, { "epoch": 5.45, "learning_rate": 9.042300218855887e-06, "loss": 0.1582, "step": 17709 }, { "epoch": 5.45, "learning_rate": 9.04130826412147e-06, "loss": 0.1724, "step": 17710 }, { "epoch": 5.45, "learning_rate": 9.040316318907745e-06, "loss": 0.1789, "step": 17711 }, { "epoch": 5.45, "learning_rate": 9.039324383224565e-06, "loss": 0.1999, "step": 17712 }, { "epoch": 5.45, "learning_rate": 9.038332457081776e-06, "loss": 0.1158, "step": 17713 }, { "epoch": 5.45, "learning_rate": 9.037340540489239e-06, "loss": 0.1493, "step": 17714 }, { "epoch": 5.45, "learning_rate": 9.036348633456791e-06, "loss": 0.1877, "step": 17715 }, { "epoch": 5.45, "learning_rate": 9.03535673599429e-06, "loss": 0.1285, "step": 17716 }, { "epoch": 5.45, "learning_rate": 9.034364848111589e-06, "loss": 0.1798, "step": 17717 }, { "epoch": 5.45, "learning_rate": 9.03337296981853e-06, "loss": 0.1967, "step": 17718 }, { "epoch": 5.45, "learning_rate": 9.032381101124975e-06, "loss": 0.2124, "step": 17719 }, { "epoch": 5.45, "learning_rate": 9.031389242040764e-06, "loss": 0.1569, "step": 17720 }, { "epoch": 5.45, "learning_rate": 9.03039739257575e-06, "loss": 0.2081, "step": 17721 }, { "epoch": 5.45, "learning_rate": 9.029405552739783e-06, "loss": 0.0952, "step": 17722 }, { "epoch": 5.45, "learning_rate": 9.028413722542715e-06, "loss": 0.2481, "step": 17723 }, { "epoch": 5.45, "learning_rate": 9.027421901994395e-06, "loss": 0.1133, "step": 17724 }, { "epoch": 5.45, "learning_rate": 9.026430091104671e-06, "loss": 0.1971, "step": 17725 }, { "epoch": 5.45, "learning_rate": 9.025438289883391e-06, "loss": 0.2008, "step": 17726 }, { "epoch": 5.45, "learning_rate": 9.024446498340408e-06, "loss": 0.1722, "step": 17727 }, { "epoch": 5.45, "learning_rate": 9.023454716485572e-06, "loss": 0.1387, "step": 17728 }, { "epoch": 5.45, "learning_rate": 9.02246294432873e-06, "loss": 0.1919, "step": 17729 }, { "epoch": 5.46, "learning_rate": 9.021471181879735e-06, "loss": 0.1604, "step": 17730 }, { "epoch": 5.46, "learning_rate": 9.020479429148427e-06, "loss": 0.3499, "step": 17731 }, { "epoch": 5.46, "learning_rate": 9.019487686144663e-06, "loss": 0.2886, "step": 17732 }, { "epoch": 5.46, "learning_rate": 9.018495952878291e-06, "loss": 0.1661, "step": 17733 }, { "epoch": 5.46, "learning_rate": 9.017504229359157e-06, "loss": 0.211, "step": 17734 }, { "epoch": 5.46, "learning_rate": 9.016512515597117e-06, "loss": 0.0855, "step": 17735 }, { "epoch": 5.46, "learning_rate": 9.015520811602008e-06, "loss": 0.1318, "step": 17736 }, { "epoch": 5.46, "learning_rate": 9.014529117383687e-06, "loss": 0.3089, "step": 17737 }, { "epoch": 5.46, "learning_rate": 9.013537432952e-06, "loss": 0.1557, "step": 17738 }, { "epoch": 5.46, "learning_rate": 9.012545758316794e-06, "loss": 0.2231, "step": 17739 }, { "epoch": 5.46, "learning_rate": 9.011554093487924e-06, "loss": 0.219, "step": 17740 }, { "epoch": 5.46, "learning_rate": 9.010562438475225e-06, "loss": 0.1997, "step": 17741 }, { "epoch": 5.46, "learning_rate": 9.009570793288558e-06, "loss": 0.1557, "step": 17742 }, { "epoch": 5.46, "learning_rate": 9.008579157937763e-06, "loss": 0.1486, "step": 17743 }, { "epoch": 5.46, "learning_rate": 9.007587532432692e-06, "loss": 0.1726, "step": 17744 }, { "epoch": 5.46, "learning_rate": 9.006595916783196e-06, "loss": 0.0884, "step": 17745 }, { "epoch": 5.46, "learning_rate": 9.005604310999109e-06, "loss": 0.2822, "step": 17746 }, { "epoch": 5.46, "learning_rate": 9.004612715090294e-06, "loss": 0.206, "step": 17747 }, { "epoch": 5.46, "learning_rate": 9.00362112906659e-06, "loss": 0.1943, "step": 17748 }, { "epoch": 5.46, "learning_rate": 9.002629552937846e-06, "loss": 0.1592, "step": 17749 }, { "epoch": 5.46, "learning_rate": 9.001637986713913e-06, "loss": 0.1454, "step": 17750 }, { "epoch": 5.46, "learning_rate": 9.000646430404631e-06, "loss": 0.1328, "step": 17751 }, { "epoch": 5.46, "learning_rate": 8.999654884019854e-06, "loss": 0.1125, "step": 17752 }, { "epoch": 5.46, "learning_rate": 8.998663347569424e-06, "loss": 0.1752, "step": 17753 }, { "epoch": 5.46, "learning_rate": 8.99767182106319e-06, "loss": 0.1151, "step": 17754 }, { "epoch": 5.46, "learning_rate": 8.996680304511e-06, "loss": 0.1396, "step": 17755 }, { "epoch": 5.46, "learning_rate": 8.995688797922697e-06, "loss": 0.1876, "step": 17756 }, { "epoch": 5.46, "learning_rate": 8.994697301308133e-06, "loss": 0.0729, "step": 17757 }, { "epoch": 5.46, "learning_rate": 8.993705814677146e-06, "loss": 0.2242, "step": 17758 }, { "epoch": 5.46, "learning_rate": 8.992714338039595e-06, "loss": 0.2423, "step": 17759 }, { "epoch": 5.46, "learning_rate": 8.991722871405313e-06, "loss": 0.2874, "step": 17760 }, { "epoch": 5.46, "learning_rate": 8.990731414784153e-06, "loss": 0.15, "step": 17761 }, { "epoch": 5.47, "learning_rate": 8.989739968185965e-06, "loss": 0.1233, "step": 17762 }, { "epoch": 5.47, "learning_rate": 8.988748531620582e-06, "loss": 0.1886, "step": 17763 }, { "epoch": 5.47, "learning_rate": 8.987757105097866e-06, "loss": 0.256, "step": 17764 }, { "epoch": 5.47, "learning_rate": 8.986765688627652e-06, "loss": 0.1121, "step": 17765 }, { "epoch": 5.47, "learning_rate": 8.985774282219786e-06, "loss": 0.1947, "step": 17766 }, { "epoch": 5.47, "learning_rate": 8.984782885884119e-06, "loss": 0.1677, "step": 17767 }, { "epoch": 5.47, "learning_rate": 8.98379149963049e-06, "loss": 0.1284, "step": 17768 }, { "epoch": 5.47, "learning_rate": 8.982800123468755e-06, "loss": 0.1632, "step": 17769 }, { "epoch": 5.47, "learning_rate": 8.981808757408748e-06, "loss": 0.1537, "step": 17770 }, { "epoch": 5.47, "learning_rate": 8.980817401460317e-06, "loss": 0.1785, "step": 17771 }, { "epoch": 5.47, "learning_rate": 8.97982605563331e-06, "loss": 0.1977, "step": 17772 }, { "epoch": 5.47, "learning_rate": 8.978834719937569e-06, "loss": 0.2777, "step": 17773 }, { "epoch": 5.47, "learning_rate": 8.977843394382945e-06, "loss": 0.1541, "step": 17774 }, { "epoch": 5.47, "learning_rate": 8.976852078979274e-06, "loss": 0.113, "step": 17775 }, { "epoch": 5.47, "learning_rate": 8.975860773736405e-06, "loss": 0.1086, "step": 17776 }, { "epoch": 5.47, "learning_rate": 8.974869478664183e-06, "loss": 0.1859, "step": 17777 }, { "epoch": 5.47, "learning_rate": 8.973878193772451e-06, "loss": 0.1814, "step": 17778 }, { "epoch": 5.47, "learning_rate": 8.972886919071055e-06, "loss": 0.187, "step": 17779 }, { "epoch": 5.47, "learning_rate": 8.971895654569842e-06, "loss": 0.1862, "step": 17780 }, { "epoch": 5.47, "learning_rate": 8.970904400278648e-06, "loss": 0.2645, "step": 17781 }, { "epoch": 5.47, "learning_rate": 8.969913156207323e-06, "loss": 0.2058, "step": 17782 }, { "epoch": 5.47, "learning_rate": 8.968921922365709e-06, "loss": 0.1124, "step": 17783 }, { "epoch": 5.47, "learning_rate": 8.967930698763651e-06, "loss": 0.1731, "step": 17784 }, { "epoch": 5.47, "learning_rate": 8.966939485410996e-06, "loss": 0.1487, "step": 17785 }, { "epoch": 5.47, "learning_rate": 8.965948282317579e-06, "loss": 0.2025, "step": 17786 }, { "epoch": 5.47, "learning_rate": 8.964957089493251e-06, "loss": 0.1737, "step": 17787 }, { "epoch": 5.47, "learning_rate": 8.963965906947852e-06, "loss": 0.1904, "step": 17788 }, { "epoch": 5.47, "learning_rate": 8.962974734691226e-06, "loss": 0.1387, "step": 17789 }, { "epoch": 5.47, "learning_rate": 8.96198357273322e-06, "loss": 0.2672, "step": 17790 }, { "epoch": 5.47, "learning_rate": 8.960992421083669e-06, "loss": 0.193, "step": 17791 }, { "epoch": 5.47, "learning_rate": 8.960001279752424e-06, "loss": 0.1656, "step": 17792 }, { "epoch": 5.47, "learning_rate": 8.959010148749324e-06, "loss": 0.2457, "step": 17793 }, { "epoch": 5.47, "learning_rate": 8.958019028084212e-06, "loss": 0.1419, "step": 17794 }, { "epoch": 5.48, "learning_rate": 8.957027917766934e-06, "loss": 0.1101, "step": 17795 }, { "epoch": 5.48, "learning_rate": 8.956036817807324e-06, "loss": 0.1243, "step": 17796 }, { "epoch": 5.48, "learning_rate": 8.955045728215238e-06, "loss": 0.1632, "step": 17797 }, { "epoch": 5.48, "learning_rate": 8.954054649000505e-06, "loss": 0.1673, "step": 17798 }, { "epoch": 5.48, "learning_rate": 8.953063580172975e-06, "loss": 0.1531, "step": 17799 }, { "epoch": 5.48, "learning_rate": 8.95207252174249e-06, "loss": 0.1311, "step": 17800 }, { "epoch": 5.48, "learning_rate": 8.95108147371889e-06, "loss": 0.1117, "step": 17801 }, { "epoch": 5.48, "learning_rate": 8.95009043611202e-06, "loss": 0.1976, "step": 17802 }, { "epoch": 5.48, "learning_rate": 8.949099408931716e-06, "loss": 0.2115, "step": 17803 }, { "epoch": 5.48, "learning_rate": 8.948108392187824e-06, "loss": 0.1786, "step": 17804 }, { "epoch": 5.48, "learning_rate": 8.947117385890187e-06, "loss": 0.2204, "step": 17805 }, { "epoch": 5.48, "learning_rate": 8.94612639004864e-06, "loss": 0.2675, "step": 17806 }, { "epoch": 5.48, "learning_rate": 8.945135404673037e-06, "loss": 0.0999, "step": 17807 }, { "epoch": 5.48, "learning_rate": 8.944144429773203e-06, "loss": 0.1378, "step": 17808 }, { "epoch": 5.48, "learning_rate": 8.943153465358994e-06, "loss": 0.1716, "step": 17809 }, { "epoch": 5.48, "learning_rate": 8.942162511440245e-06, "loss": 0.1289, "step": 17810 }, { "epoch": 5.48, "learning_rate": 8.941171568026794e-06, "loss": 0.1694, "step": 17811 }, { "epoch": 5.48, "learning_rate": 8.940180635128489e-06, "loss": 0.314, "step": 17812 }, { "epoch": 5.48, "learning_rate": 8.93918971275516e-06, "loss": 0.1667, "step": 17813 }, { "epoch": 5.48, "learning_rate": 8.938198800916662e-06, "loss": 0.1481, "step": 17814 }, { "epoch": 5.48, "learning_rate": 8.937207899622826e-06, "loss": 0.2002, "step": 17815 }, { "epoch": 5.48, "learning_rate": 8.936217008883494e-06, "loss": 0.1347, "step": 17816 }, { "epoch": 5.48, "learning_rate": 8.93522612870851e-06, "loss": 0.1486, "step": 17817 }, { "epoch": 5.48, "learning_rate": 8.934235259107708e-06, "loss": 0.1805, "step": 17818 }, { "epoch": 5.48, "learning_rate": 8.933244400090937e-06, "loss": 0.163, "step": 17819 }, { "epoch": 5.48, "learning_rate": 8.932253551668031e-06, "loss": 0.1841, "step": 17820 }, { "epoch": 5.48, "learning_rate": 8.931262713848829e-06, "loss": 0.1995, "step": 17821 }, { "epoch": 5.48, "learning_rate": 8.930271886643173e-06, "loss": 0.2419, "step": 17822 }, { "epoch": 5.48, "learning_rate": 8.929281070060903e-06, "loss": 0.1645, "step": 17823 }, { "epoch": 5.48, "learning_rate": 8.928290264111863e-06, "loss": 0.1341, "step": 17824 }, { "epoch": 5.48, "learning_rate": 8.927299468805886e-06, "loss": 0.1765, "step": 17825 }, { "epoch": 5.48, "learning_rate": 8.926308684152813e-06, "loss": 0.1953, "step": 17826 }, { "epoch": 5.49, "learning_rate": 8.925317910162485e-06, "loss": 0.2722, "step": 17827 }, { "epoch": 5.49, "learning_rate": 8.924327146844738e-06, "loss": 0.2479, "step": 17828 }, { "epoch": 5.49, "learning_rate": 8.923336394209416e-06, "loss": 0.1298, "step": 17829 }, { "epoch": 5.49, "learning_rate": 8.92234565226636e-06, "loss": 0.1017, "step": 17830 }, { "epoch": 5.49, "learning_rate": 8.921354921025398e-06, "loss": 0.2166, "step": 17831 }, { "epoch": 5.49, "learning_rate": 8.92036420049638e-06, "loss": 0.1884, "step": 17832 }, { "epoch": 5.49, "learning_rate": 8.919373490689137e-06, "loss": 0.127, "step": 17833 }, { "epoch": 5.49, "learning_rate": 8.918382791613514e-06, "loss": 0.116, "step": 17834 }, { "epoch": 5.49, "learning_rate": 8.917392103279346e-06, "loss": 0.1435, "step": 17835 }, { "epoch": 5.49, "learning_rate": 8.916401425696473e-06, "loss": 0.2558, "step": 17836 }, { "epoch": 5.49, "learning_rate": 8.915410758874732e-06, "loss": 0.1951, "step": 17837 }, { "epoch": 5.49, "learning_rate": 8.91442010282396e-06, "loss": 0.1622, "step": 17838 }, { "epoch": 5.49, "learning_rate": 8.913429457553999e-06, "loss": 0.2072, "step": 17839 }, { "epoch": 5.49, "learning_rate": 8.912438823074683e-06, "loss": 0.1302, "step": 17840 }, { "epoch": 5.49, "learning_rate": 8.911448199395856e-06, "loss": 0.1794, "step": 17841 }, { "epoch": 5.49, "learning_rate": 8.91045758652735e-06, "loss": 0.1809, "step": 17842 }, { "epoch": 5.49, "learning_rate": 8.909466984479e-06, "loss": 0.1209, "step": 17843 }, { "epoch": 5.49, "learning_rate": 8.908476393260652e-06, "loss": 0.1219, "step": 17844 }, { "epoch": 5.49, "learning_rate": 8.907485812882137e-06, "loss": 0.2416, "step": 17845 }, { "epoch": 5.49, "learning_rate": 8.906495243353297e-06, "loss": 0.1544, "step": 17846 }, { "epoch": 5.49, "learning_rate": 8.905504684683969e-06, "loss": 0.0977, "step": 17847 }, { "epoch": 5.49, "learning_rate": 8.904514136883984e-06, "loss": 0.1251, "step": 17848 }, { "epoch": 5.49, "learning_rate": 8.903523599963184e-06, "loss": 0.1818, "step": 17849 }, { "epoch": 5.49, "learning_rate": 8.902533073931404e-06, "loss": 0.1294, "step": 17850 }, { "epoch": 5.49, "learning_rate": 8.901542558798485e-06, "loss": 0.2101, "step": 17851 }, { "epoch": 5.49, "learning_rate": 8.900552054574261e-06, "loss": 0.1678, "step": 17852 }, { "epoch": 5.49, "learning_rate": 8.899561561268566e-06, "loss": 0.2532, "step": 17853 }, { "epoch": 5.49, "learning_rate": 8.89857107889124e-06, "loss": 0.2024, "step": 17854 }, { "epoch": 5.49, "learning_rate": 8.897580607452117e-06, "loss": 0.1293, "step": 17855 }, { "epoch": 5.49, "learning_rate": 8.896590146961038e-06, "loss": 0.1369, "step": 17856 }, { "epoch": 5.49, "learning_rate": 8.895599697427834e-06, "loss": 0.2924, "step": 17857 }, { "epoch": 5.49, "learning_rate": 8.89460925886234e-06, "loss": 0.2217, "step": 17858 }, { "epoch": 5.49, "learning_rate": 8.893618831274399e-06, "loss": 0.2505, "step": 17859 }, { "epoch": 5.5, "learning_rate": 8.89262841467384e-06, "loss": 0.1158, "step": 17860 }, { "epoch": 5.5, "learning_rate": 8.891638009070502e-06, "loss": 0.1692, "step": 17861 }, { "epoch": 5.5, "learning_rate": 8.890647614474223e-06, "loss": 0.1456, "step": 17862 }, { "epoch": 5.5, "learning_rate": 8.88965723089483e-06, "loss": 0.1388, "step": 17863 }, { "epoch": 5.5, "learning_rate": 8.888666858342169e-06, "loss": 0.1369, "step": 17864 }, { "epoch": 5.5, "learning_rate": 8.887676496826066e-06, "loss": 0.1858, "step": 17865 }, { "epoch": 5.5, "learning_rate": 8.886686146356364e-06, "loss": 0.183, "step": 17866 }, { "epoch": 5.5, "learning_rate": 8.885695806942893e-06, "loss": 0.1068, "step": 17867 }, { "epoch": 5.5, "learning_rate": 8.884705478595489e-06, "loss": 0.1921, "step": 17868 }, { "epoch": 5.5, "learning_rate": 8.883715161323991e-06, "loss": 0.1301, "step": 17869 }, { "epoch": 5.5, "learning_rate": 8.882724855138227e-06, "loss": 0.2231, "step": 17870 }, { "epoch": 5.5, "learning_rate": 8.881734560048037e-06, "loss": 0.1379, "step": 17871 }, { "epoch": 5.5, "learning_rate": 8.880744276063252e-06, "loss": 0.1555, "step": 17872 }, { "epoch": 5.5, "learning_rate": 8.879754003193706e-06, "loss": 0.1574, "step": 17873 }, { "epoch": 5.5, "learning_rate": 8.87876374144924e-06, "loss": 0.2748, "step": 17874 }, { "epoch": 5.5, "learning_rate": 8.877773490839677e-06, "loss": 0.1132, "step": 17875 }, { "epoch": 5.5, "learning_rate": 8.876783251374864e-06, "loss": 0.1778, "step": 17876 }, { "epoch": 5.5, "learning_rate": 8.875793023064626e-06, "loss": 0.1529, "step": 17877 }, { "epoch": 5.5, "learning_rate": 8.874802805918797e-06, "loss": 0.1508, "step": 17878 }, { "epoch": 5.5, "learning_rate": 8.873812599947215e-06, "loss": 0.3114, "step": 17879 }, { "epoch": 5.5, "learning_rate": 8.87282240515971e-06, "loss": 0.1358, "step": 17880 }, { "epoch": 5.5, "learning_rate": 8.871832221566124e-06, "loss": 0.1566, "step": 17881 }, { "epoch": 5.5, "learning_rate": 8.870842049176278e-06, "loss": 0.131, "step": 17882 }, { "epoch": 5.5, "learning_rate": 8.869851888000011e-06, "loss": 0.1547, "step": 17883 }, { "epoch": 5.5, "learning_rate": 8.868861738047158e-06, "loss": 0.1292, "step": 17884 }, { "epoch": 5.5, "learning_rate": 8.86787159932755e-06, "loss": 0.1971, "step": 17885 }, { "epoch": 5.5, "learning_rate": 8.866881471851022e-06, "loss": 0.1826, "step": 17886 }, { "epoch": 5.5, "learning_rate": 8.865891355627404e-06, "loss": 0.3243, "step": 17887 }, { "epoch": 5.5, "learning_rate": 8.864901250666527e-06, "loss": 0.147, "step": 17888 }, { "epoch": 5.5, "learning_rate": 8.863911156978231e-06, "loss": 0.2083, "step": 17889 }, { "epoch": 5.5, "learning_rate": 8.862921074572342e-06, "loss": 0.1269, "step": 17890 }, { "epoch": 5.5, "learning_rate": 8.861931003458697e-06, "loss": 0.2501, "step": 17891 }, { "epoch": 5.51, "learning_rate": 8.860940943647126e-06, "loss": 0.1115, "step": 17892 }, { "epoch": 5.51, "learning_rate": 8.859950895147458e-06, "loss": 0.1193, "step": 17893 }, { "epoch": 5.51, "learning_rate": 8.85896085796953e-06, "loss": 0.2063, "step": 17894 }, { "epoch": 5.51, "learning_rate": 8.85797083212317e-06, "loss": 0.1471, "step": 17895 }, { "epoch": 5.51, "learning_rate": 8.856980817618214e-06, "loss": 0.2136, "step": 17896 }, { "epoch": 5.51, "learning_rate": 8.855990814464497e-06, "loss": 0.1319, "step": 17897 }, { "epoch": 5.51, "learning_rate": 8.855000822671838e-06, "loss": 0.1829, "step": 17898 }, { "epoch": 5.51, "learning_rate": 8.85401084225008e-06, "loss": 0.1631, "step": 17899 }, { "epoch": 5.51, "learning_rate": 8.853020873209048e-06, "loss": 0.2553, "step": 17900 }, { "epoch": 5.51, "learning_rate": 8.852030915558578e-06, "loss": 0.1342, "step": 17901 }, { "epoch": 5.51, "learning_rate": 8.8510409693085e-06, "loss": 0.1297, "step": 17902 }, { "epoch": 5.51, "learning_rate": 8.850051034468642e-06, "loss": 0.2578, "step": 17903 }, { "epoch": 5.51, "learning_rate": 8.849061111048838e-06, "loss": 0.1832, "step": 17904 }, { "epoch": 5.51, "learning_rate": 8.848071199058916e-06, "loss": 0.1794, "step": 17905 }, { "epoch": 5.51, "learning_rate": 8.84708129850871e-06, "loss": 0.1219, "step": 17906 }, { "epoch": 5.51, "learning_rate": 8.846091409408053e-06, "loss": 0.191, "step": 17907 }, { "epoch": 5.51, "learning_rate": 8.845101531766763e-06, "loss": 0.1463, "step": 17908 }, { "epoch": 5.51, "learning_rate": 8.844111665594687e-06, "loss": 0.1619, "step": 17909 }, { "epoch": 5.51, "learning_rate": 8.843121810901643e-06, "loss": 0.2398, "step": 17910 }, { "epoch": 5.51, "learning_rate": 8.842131967697469e-06, "loss": 0.1552, "step": 17911 }, { "epoch": 5.51, "learning_rate": 8.84114213599199e-06, "loss": 0.1032, "step": 17912 }, { "epoch": 5.51, "learning_rate": 8.840152315795036e-06, "loss": 0.2383, "step": 17913 }, { "epoch": 5.51, "learning_rate": 8.839162507116443e-06, "loss": 0.1547, "step": 17914 }, { "epoch": 5.51, "learning_rate": 8.838172709966031e-06, "loss": 0.2537, "step": 17915 }, { "epoch": 5.51, "learning_rate": 8.837182924353638e-06, "loss": 0.1608, "step": 17916 }, { "epoch": 5.51, "learning_rate": 8.836193150289089e-06, "loss": 0.183, "step": 17917 }, { "epoch": 5.51, "learning_rate": 8.835203387782213e-06, "loss": 0.1211, "step": 17918 }, { "epoch": 5.51, "learning_rate": 8.834213636842846e-06, "loss": 0.1852, "step": 17919 }, { "epoch": 5.51, "learning_rate": 8.833223897480807e-06, "loss": 0.1985, "step": 17920 }, { "epoch": 5.51, "learning_rate": 8.832234169705932e-06, "loss": 0.1887, "step": 17921 }, { "epoch": 5.51, "learning_rate": 8.831244453528048e-06, "loss": 0.1971, "step": 17922 }, { "epoch": 5.51, "learning_rate": 8.830254748956983e-06, "loss": 0.0935, "step": 17923 }, { "epoch": 5.51, "learning_rate": 8.829265056002569e-06, "loss": 0.2689, "step": 17924 }, { "epoch": 5.52, "learning_rate": 8.828275374674626e-06, "loss": 0.2129, "step": 17925 }, { "epoch": 5.52, "learning_rate": 8.827285704982994e-06, "loss": 0.2285, "step": 17926 }, { "epoch": 5.52, "learning_rate": 8.826296046937495e-06, "loss": 0.1252, "step": 17927 }, { "epoch": 5.52, "learning_rate": 8.825306400547955e-06, "loss": 0.2133, "step": 17928 }, { "epoch": 5.52, "learning_rate": 8.824316765824207e-06, "loss": 0.1478, "step": 17929 }, { "epoch": 5.52, "learning_rate": 8.823327142776076e-06, "loss": 0.1462, "step": 17930 }, { "epoch": 5.52, "learning_rate": 8.822337531413395e-06, "loss": 0.1709, "step": 17931 }, { "epoch": 5.52, "learning_rate": 8.821347931745986e-06, "loss": 0.1228, "step": 17932 }, { "epoch": 5.52, "learning_rate": 8.820358343783675e-06, "loss": 0.1886, "step": 17933 }, { "epoch": 5.52, "learning_rate": 8.819368767536296e-06, "loss": 0.1661, "step": 17934 }, { "epoch": 5.52, "learning_rate": 8.81837920301367e-06, "loss": 0.2815, "step": 17935 }, { "epoch": 5.52, "learning_rate": 8.817389650225631e-06, "loss": 0.1999, "step": 17936 }, { "epoch": 5.52, "learning_rate": 8.816400109182002e-06, "loss": 0.2195, "step": 17937 }, { "epoch": 5.52, "learning_rate": 8.815410579892609e-06, "loss": 0.1302, "step": 17938 }, { "epoch": 5.52, "learning_rate": 8.814421062367284e-06, "loss": 0.1249, "step": 17939 }, { "epoch": 5.52, "learning_rate": 8.813431556615845e-06, "loss": 0.1689, "step": 17940 }, { "epoch": 5.52, "learning_rate": 8.812442062648131e-06, "loss": 0.1889, "step": 17941 }, { "epoch": 5.52, "learning_rate": 8.811452580473958e-06, "loss": 0.0979, "step": 17942 }, { "epoch": 5.52, "learning_rate": 8.810463110103155e-06, "loss": 0.1739, "step": 17943 }, { "epoch": 5.52, "learning_rate": 8.809473651545553e-06, "loss": 0.1338, "step": 17944 }, { "epoch": 5.52, "learning_rate": 8.80848420481097e-06, "loss": 0.2267, "step": 17945 }, { "epoch": 5.52, "learning_rate": 8.807494769909242e-06, "loss": 0.2226, "step": 17946 }, { "epoch": 5.52, "learning_rate": 8.80650534685019e-06, "loss": 0.0958, "step": 17947 }, { "epoch": 5.52, "learning_rate": 8.805515935643635e-06, "loss": 0.1817, "step": 17948 }, { "epoch": 5.52, "learning_rate": 8.804526536299413e-06, "loss": 0.1207, "step": 17949 }, { "epoch": 5.52, "learning_rate": 8.803537148827339e-06, "loss": 0.0934, "step": 17950 }, { "epoch": 5.52, "learning_rate": 8.802547773237248e-06, "loss": 0.1862, "step": 17951 }, { "epoch": 5.52, "learning_rate": 8.801558409538962e-06, "loss": 0.3268, "step": 17952 }, { "epoch": 5.52, "learning_rate": 8.800569057742302e-06, "loss": 0.1427, "step": 17953 }, { "epoch": 5.52, "learning_rate": 8.799579717857098e-06, "loss": 0.1105, "step": 17954 }, { "epoch": 5.52, "learning_rate": 8.798590389893172e-06, "loss": 0.1401, "step": 17955 }, { "epoch": 5.52, "learning_rate": 8.797601073860353e-06, "loss": 0.1556, "step": 17956 }, { "epoch": 5.53, "learning_rate": 8.796611769768467e-06, "loss": 0.2951, "step": 17957 }, { "epoch": 5.53, "learning_rate": 8.795622477627326e-06, "loss": 0.1598, "step": 17958 }, { "epoch": 5.53, "learning_rate": 8.79463319744677e-06, "loss": 0.2157, "step": 17959 }, { "epoch": 5.53, "learning_rate": 8.793643929236617e-06, "loss": 0.1245, "step": 17960 }, { "epoch": 5.53, "learning_rate": 8.792654673006691e-06, "loss": 0.2236, "step": 17961 }, { "epoch": 5.53, "learning_rate": 8.79166542876682e-06, "loss": 0.1703, "step": 17962 }, { "epoch": 5.53, "learning_rate": 8.790676196526816e-06, "loss": 0.1579, "step": 17963 }, { "epoch": 5.53, "learning_rate": 8.789686976296521e-06, "loss": 0.202, "step": 17964 }, { "epoch": 5.53, "learning_rate": 8.788697768085744e-06, "loss": 0.1268, "step": 17965 }, { "epoch": 5.53, "learning_rate": 8.787708571904317e-06, "loss": 0.1475, "step": 17966 }, { "epoch": 5.53, "learning_rate": 8.786719387762061e-06, "loss": 0.0904, "step": 17967 }, { "epoch": 5.53, "learning_rate": 8.785730215668798e-06, "loss": 0.171, "step": 17968 }, { "epoch": 5.53, "learning_rate": 8.784741055634359e-06, "loss": 0.1346, "step": 17969 }, { "epoch": 5.53, "learning_rate": 8.783751907668554e-06, "loss": 0.2143, "step": 17970 }, { "epoch": 5.53, "learning_rate": 8.782762771781217e-06, "loss": 0.2021, "step": 17971 }, { "epoch": 5.53, "learning_rate": 8.781773647982169e-06, "loss": 0.141, "step": 17972 }, { "epoch": 5.53, "learning_rate": 8.780784536281226e-06, "loss": 0.252, "step": 17973 }, { "epoch": 5.53, "learning_rate": 8.779795436688224e-06, "loss": 0.2105, "step": 17974 }, { "epoch": 5.53, "learning_rate": 8.778806349212968e-06, "loss": 0.2195, "step": 17975 }, { "epoch": 5.53, "learning_rate": 8.777817273865299e-06, "loss": 0.1259, "step": 17976 }, { "epoch": 5.53, "learning_rate": 8.77682821065503e-06, "loss": 0.1534, "step": 17977 }, { "epoch": 5.53, "learning_rate": 8.77583915959198e-06, "loss": 0.1328, "step": 17978 }, { "epoch": 5.53, "learning_rate": 8.774850120685979e-06, "loss": 0.1337, "step": 17979 }, { "epoch": 5.53, "learning_rate": 8.77386109394684e-06, "loss": 0.1605, "step": 17980 }, { "epoch": 5.53, "learning_rate": 8.772872079384397e-06, "loss": 0.1939, "step": 17981 }, { "epoch": 5.53, "learning_rate": 8.771883077008463e-06, "loss": 0.2031, "step": 17982 }, { "epoch": 5.53, "learning_rate": 8.770894086828859e-06, "loss": 0.1786, "step": 17983 }, { "epoch": 5.53, "learning_rate": 8.769905108855412e-06, "loss": 0.163, "step": 17984 }, { "epoch": 5.53, "learning_rate": 8.76891614309794e-06, "loss": 0.1778, "step": 17985 }, { "epoch": 5.53, "learning_rate": 8.767927189566268e-06, "loss": 0.2483, "step": 17986 }, { "epoch": 5.53, "learning_rate": 8.766938248270213e-06, "loss": 0.2403, "step": 17987 }, { "epoch": 5.53, "learning_rate": 8.765949319219595e-06, "loss": 0.1209, "step": 17988 }, { "epoch": 5.53, "learning_rate": 8.76496040242424e-06, "loss": 0.1761, "step": 17989 }, { "epoch": 5.54, "learning_rate": 8.763971497893966e-06, "loss": 0.195, "step": 17990 }, { "epoch": 5.54, "learning_rate": 8.762982605638598e-06, "loss": 0.1986, "step": 17991 }, { "epoch": 5.54, "learning_rate": 8.761993725667948e-06, "loss": 0.1952, "step": 17992 }, { "epoch": 5.54, "learning_rate": 8.761004857991841e-06, "loss": 0.1509, "step": 17993 }, { "epoch": 5.54, "learning_rate": 8.7600160026201e-06, "loss": 0.1318, "step": 17994 }, { "epoch": 5.54, "learning_rate": 8.759027159562541e-06, "loss": 0.1475, "step": 17995 }, { "epoch": 5.54, "learning_rate": 8.758038328828987e-06, "loss": 0.1833, "step": 17996 }, { "epoch": 5.54, "learning_rate": 8.757049510429259e-06, "loss": 0.2006, "step": 17997 }, { "epoch": 5.54, "learning_rate": 8.756060704373173e-06, "loss": 0.2271, "step": 17998 }, { "epoch": 5.54, "learning_rate": 8.75507191067055e-06, "loss": 0.1675, "step": 17999 }, { "epoch": 5.54, "learning_rate": 8.754083129331211e-06, "loss": 0.1773, "step": 18000 }, { "epoch": 5.54, "learning_rate": 8.753094360364973e-06, "loss": 0.2569, "step": 18001 }, { "epoch": 5.54, "learning_rate": 8.752105603781662e-06, "loss": 0.1718, "step": 18002 }, { "epoch": 5.54, "learning_rate": 8.751116859591089e-06, "loss": 0.1275, "step": 18003 }, { "epoch": 5.54, "learning_rate": 8.750128127803077e-06, "loss": 0.1257, "step": 18004 }, { "epoch": 5.54, "learning_rate": 8.749139408427443e-06, "loss": 0.3016, "step": 18005 }, { "epoch": 5.54, "learning_rate": 8.748150701474008e-06, "loss": 0.122, "step": 18006 }, { "epoch": 5.54, "learning_rate": 8.747162006952595e-06, "loss": 0.1834, "step": 18007 }, { "epoch": 5.54, "learning_rate": 8.746173324873013e-06, "loss": 0.2159, "step": 18008 }, { "epoch": 5.54, "learning_rate": 8.745184655245088e-06, "loss": 0.1442, "step": 18009 }, { "epoch": 5.54, "learning_rate": 8.744195998078632e-06, "loss": 0.1828, "step": 18010 }, { "epoch": 5.54, "learning_rate": 8.743207353383469e-06, "loss": 0.1218, "step": 18011 }, { "epoch": 5.54, "learning_rate": 8.74221872116942e-06, "loss": 0.1906, "step": 18012 }, { "epoch": 5.54, "learning_rate": 8.741230101446289e-06, "loss": 0.1532, "step": 18013 }, { "epoch": 5.54, "learning_rate": 8.740241494223911e-06, "loss": 0.1705, "step": 18014 }, { "epoch": 5.54, "learning_rate": 8.739252899512091e-06, "loss": 0.1758, "step": 18015 }, { "epoch": 5.54, "learning_rate": 8.738264317320655e-06, "loss": 0.2614, "step": 18016 }, { "epoch": 5.54, "learning_rate": 8.737275747659416e-06, "loss": 0.1493, "step": 18017 }, { "epoch": 5.54, "learning_rate": 8.736287190538194e-06, "loss": 0.1298, "step": 18018 }, { "epoch": 5.54, "learning_rate": 8.735298645966806e-06, "loss": 0.1581, "step": 18019 }, { "epoch": 5.54, "learning_rate": 8.734310113955064e-06, "loss": 0.1991, "step": 18020 }, { "epoch": 5.54, "learning_rate": 8.73332159451279e-06, "loss": 0.1662, "step": 18021 }, { "epoch": 5.55, "learning_rate": 8.732333087649802e-06, "loss": 0.2014, "step": 18022 }, { "epoch": 5.55, "learning_rate": 8.731344593375912e-06, "loss": 0.0913, "step": 18023 }, { "epoch": 5.55, "learning_rate": 8.730356111700944e-06, "loss": 0.114, "step": 18024 }, { "epoch": 5.55, "learning_rate": 8.729367642634703e-06, "loss": 0.1603, "step": 18025 }, { "epoch": 5.55, "learning_rate": 8.728379186187019e-06, "loss": 0.1465, "step": 18026 }, { "epoch": 5.55, "learning_rate": 8.727390742367698e-06, "loss": 0.1575, "step": 18027 }, { "epoch": 5.55, "learning_rate": 8.72640231118656e-06, "loss": 0.1684, "step": 18028 }, { "epoch": 5.55, "learning_rate": 8.725413892653424e-06, "loss": 0.1243, "step": 18029 }, { "epoch": 5.55, "learning_rate": 8.724425486778096e-06, "loss": 0.2519, "step": 18030 }, { "epoch": 5.55, "learning_rate": 8.723437093570406e-06, "loss": 0.2059, "step": 18031 }, { "epoch": 5.55, "learning_rate": 8.722448713040158e-06, "loss": 0.0763, "step": 18032 }, { "epoch": 5.55, "learning_rate": 8.721460345197173e-06, "loss": 0.1735, "step": 18033 }, { "epoch": 5.55, "learning_rate": 8.720471990051265e-06, "loss": 0.1523, "step": 18034 }, { "epoch": 5.55, "learning_rate": 8.719483647612247e-06, "loss": 0.145, "step": 18035 }, { "epoch": 5.55, "learning_rate": 8.718495317889941e-06, "loss": 0.1987, "step": 18036 }, { "epoch": 5.55, "learning_rate": 8.717507000894152e-06, "loss": 0.0673, "step": 18037 }, { "epoch": 5.55, "learning_rate": 8.716518696634707e-06, "loss": 0.1412, "step": 18038 }, { "epoch": 5.55, "learning_rate": 8.71553040512141e-06, "loss": 0.2145, "step": 18039 }, { "epoch": 5.55, "learning_rate": 8.71454212636408e-06, "loss": 0.1913, "step": 18040 }, { "epoch": 5.55, "learning_rate": 8.713553860372534e-06, "loss": 0.1116, "step": 18041 }, { "epoch": 5.55, "learning_rate": 8.712565607156579e-06, "loss": 0.1267, "step": 18042 }, { "epoch": 5.55, "learning_rate": 8.71157736672604e-06, "loss": 0.1733, "step": 18043 }, { "epoch": 5.55, "learning_rate": 8.710589139090722e-06, "loss": 0.1106, "step": 18044 }, { "epoch": 5.55, "learning_rate": 8.709600924260441e-06, "loss": 0.1621, "step": 18045 }, { "epoch": 5.55, "learning_rate": 8.708612722245014e-06, "loss": 0.1482, "step": 18046 }, { "epoch": 5.55, "learning_rate": 8.707624533054251e-06, "loss": 0.1225, "step": 18047 }, { "epoch": 5.55, "learning_rate": 8.706636356697972e-06, "loss": 0.137, "step": 18048 }, { "epoch": 5.55, "learning_rate": 8.705648193185982e-06, "loss": 0.1853, "step": 18049 }, { "epoch": 5.55, "learning_rate": 8.704660042528098e-06, "loss": 0.1787, "step": 18050 }, { "epoch": 5.55, "learning_rate": 8.703671904734136e-06, "loss": 0.1515, "step": 18051 }, { "epoch": 5.55, "learning_rate": 8.702683779813904e-06, "loss": 0.1399, "step": 18052 }, { "epoch": 5.55, "learning_rate": 8.701695667777221e-06, "loss": 0.1935, "step": 18053 }, { "epoch": 5.55, "learning_rate": 8.700707568633896e-06, "loss": 0.2137, "step": 18054 }, { "epoch": 5.56, "learning_rate": 8.69971948239374e-06, "loss": 0.1378, "step": 18055 }, { "epoch": 5.56, "learning_rate": 8.69873140906657e-06, "loss": 0.2338, "step": 18056 }, { "epoch": 5.56, "learning_rate": 8.697743348662192e-06, "loss": 0.1223, "step": 18057 }, { "epoch": 5.56, "learning_rate": 8.69675530119043e-06, "loss": 0.2372, "step": 18058 }, { "epoch": 5.56, "learning_rate": 8.695767266661085e-06, "loss": 0.2382, "step": 18059 }, { "epoch": 5.56, "learning_rate": 8.694779245083972e-06, "loss": 0.1411, "step": 18060 }, { "epoch": 5.56, "learning_rate": 8.693791236468905e-06, "loss": 0.1821, "step": 18061 }, { "epoch": 5.56, "learning_rate": 8.692803240825693e-06, "loss": 0.1412, "step": 18062 }, { "epoch": 5.56, "learning_rate": 8.691815258164153e-06, "loss": 0.1175, "step": 18063 }, { "epoch": 5.56, "learning_rate": 8.690827288494094e-06, "loss": 0.1577, "step": 18064 }, { "epoch": 5.56, "learning_rate": 8.689839331825322e-06, "loss": 0.1572, "step": 18065 }, { "epoch": 5.56, "learning_rate": 8.688851388167658e-06, "loss": 0.2145, "step": 18066 }, { "epoch": 5.56, "learning_rate": 8.687863457530903e-06, "loss": 0.2199, "step": 18067 }, { "epoch": 5.56, "learning_rate": 8.686875539924875e-06, "loss": 0.261, "step": 18068 }, { "epoch": 5.56, "learning_rate": 8.685887635359387e-06, "loss": 0.1072, "step": 18069 }, { "epoch": 5.56, "learning_rate": 8.68489974384424e-06, "loss": 0.1826, "step": 18070 }, { "epoch": 5.56, "learning_rate": 8.683911865389254e-06, "loss": 0.3874, "step": 18071 }, { "epoch": 5.56, "learning_rate": 8.682924000004234e-06, "loss": 0.275, "step": 18072 }, { "epoch": 5.56, "learning_rate": 8.681936147698994e-06, "loss": 0.1755, "step": 18073 }, { "epoch": 5.56, "learning_rate": 8.680948308483346e-06, "loss": 0.1912, "step": 18074 }, { "epoch": 5.56, "learning_rate": 8.679960482367088e-06, "loss": 0.1753, "step": 18075 }, { "epoch": 5.56, "learning_rate": 8.67897266936005e-06, "loss": 0.196, "step": 18076 }, { "epoch": 5.56, "learning_rate": 8.677984869472022e-06, "loss": 0.1886, "step": 18077 }, { "epoch": 5.56, "learning_rate": 8.676997082712827e-06, "loss": 0.1713, "step": 18078 }, { "epoch": 5.56, "learning_rate": 8.676009309092273e-06, "loss": 0.239, "step": 18079 }, { "epoch": 5.56, "learning_rate": 8.675021548620158e-06, "loss": 0.1329, "step": 18080 }, { "epoch": 5.56, "learning_rate": 8.674033801306308e-06, "loss": 0.1117, "step": 18081 }, { "epoch": 5.56, "learning_rate": 8.673046067160521e-06, "loss": 0.1462, "step": 18082 }, { "epoch": 5.56, "learning_rate": 8.672058346192611e-06, "loss": 0.1164, "step": 18083 }, { "epoch": 5.56, "learning_rate": 8.671070638412385e-06, "loss": 0.188, "step": 18084 }, { "epoch": 5.56, "learning_rate": 8.670082943829649e-06, "loss": 0.099, "step": 18085 }, { "epoch": 5.56, "learning_rate": 8.669095262454222e-06, "loss": 0.162, "step": 18086 }, { "epoch": 5.57, "learning_rate": 8.6681075942959e-06, "loss": 0.1027, "step": 18087 }, { "epoch": 5.57, "learning_rate": 8.6671199393645e-06, "loss": 0.1491, "step": 18088 }, { "epoch": 5.57, "learning_rate": 8.666132297669827e-06, "loss": 0.1746, "step": 18089 }, { "epoch": 5.57, "learning_rate": 8.665144669221687e-06, "loss": 0.2036, "step": 18090 }, { "epoch": 5.57, "learning_rate": 8.664157054029895e-06, "loss": 0.2487, "step": 18091 }, { "epoch": 5.57, "learning_rate": 8.663169452104248e-06, "loss": 0.1306, "step": 18092 }, { "epoch": 5.57, "learning_rate": 8.662181863454568e-06, "loss": 0.1324, "step": 18093 }, { "epoch": 5.57, "learning_rate": 8.66119428809065e-06, "loss": 0.2345, "step": 18094 }, { "epoch": 5.57, "learning_rate": 8.660206726022307e-06, "loss": 0.2009, "step": 18095 }, { "epoch": 5.57, "learning_rate": 8.659219177259348e-06, "loss": 0.2204, "step": 18096 }, { "epoch": 5.57, "learning_rate": 8.658231641811576e-06, "loss": 0.1014, "step": 18097 }, { "epoch": 5.57, "learning_rate": 8.657244119688806e-06, "loss": 0.1748, "step": 18098 }, { "epoch": 5.57, "learning_rate": 8.656256610900834e-06, "loss": 0.156, "step": 18099 }, { "epoch": 5.57, "learning_rate": 8.655269115457474e-06, "loss": 0.1812, "step": 18100 }, { "epoch": 5.57, "learning_rate": 8.65428163336853e-06, "loss": 0.2493, "step": 18101 }, { "epoch": 5.57, "learning_rate": 8.653294164643809e-06, "loss": 0.0737, "step": 18102 }, { "epoch": 5.57, "learning_rate": 8.652306709293121e-06, "loss": 0.1641, "step": 18103 }, { "epoch": 5.57, "learning_rate": 8.65131926732627e-06, "loss": 0.1696, "step": 18104 }, { "epoch": 5.57, "learning_rate": 8.650331838753057e-06, "loss": 0.202, "step": 18105 }, { "epoch": 5.57, "learning_rate": 8.649344423583295e-06, "loss": 0.2405, "step": 18106 }, { "epoch": 5.57, "learning_rate": 8.648357021826786e-06, "loss": 0.1686, "step": 18107 }, { "epoch": 5.57, "learning_rate": 8.647369633493344e-06, "loss": 0.1528, "step": 18108 }, { "epoch": 5.57, "learning_rate": 8.646382258592762e-06, "loss": 0.2154, "step": 18109 }, { "epoch": 5.57, "learning_rate": 8.645394897134852e-06, "loss": 0.2155, "step": 18110 }, { "epoch": 5.57, "learning_rate": 8.64440754912942e-06, "loss": 0.2208, "step": 18111 }, { "epoch": 5.57, "learning_rate": 8.64342021458627e-06, "loss": 0.1784, "step": 18112 }, { "epoch": 5.57, "learning_rate": 8.64243289351521e-06, "loss": 0.2897, "step": 18113 }, { "epoch": 5.57, "learning_rate": 8.641445585926042e-06, "loss": 0.2175, "step": 18114 }, { "epoch": 5.57, "learning_rate": 8.64045829182857e-06, "loss": 0.2297, "step": 18115 }, { "epoch": 5.57, "learning_rate": 8.6394710112326e-06, "loss": 0.1775, "step": 18116 }, { "epoch": 5.57, "learning_rate": 8.638483744147935e-06, "loss": 0.2282, "step": 18117 }, { "epoch": 5.57, "learning_rate": 8.637496490584385e-06, "loss": 0.1684, "step": 18118 }, { "epoch": 5.57, "learning_rate": 8.636509250551752e-06, "loss": 0.1941, "step": 18119 }, { "epoch": 5.58, "learning_rate": 8.635522024059833e-06, "loss": 0.1767, "step": 18120 }, { "epoch": 5.58, "learning_rate": 8.634534811118442e-06, "loss": 0.2373, "step": 18121 }, { "epoch": 5.58, "learning_rate": 8.633547611737376e-06, "loss": 0.1155, "step": 18122 }, { "epoch": 5.58, "learning_rate": 8.632560425926442e-06, "loss": 0.1417, "step": 18123 }, { "epoch": 5.58, "learning_rate": 8.631573253695446e-06, "loss": 0.0961, "step": 18124 }, { "epoch": 5.58, "learning_rate": 8.630586095054185e-06, "loss": 0.1825, "step": 18125 }, { "epoch": 5.58, "learning_rate": 8.62959895001247e-06, "loss": 0.2188, "step": 18126 }, { "epoch": 5.58, "learning_rate": 8.628611818580097e-06, "loss": 0.0707, "step": 18127 }, { "epoch": 5.58, "learning_rate": 8.627624700766875e-06, "loss": 0.1883, "step": 18128 }, { "epoch": 5.58, "learning_rate": 8.626637596582605e-06, "loss": 0.1664, "step": 18129 }, { "epoch": 5.58, "learning_rate": 8.625650506037084e-06, "loss": 0.1703, "step": 18130 }, { "epoch": 5.58, "learning_rate": 8.624663429140128e-06, "loss": 0.1287, "step": 18131 }, { "epoch": 5.58, "learning_rate": 8.623676365901526e-06, "loss": 0.208, "step": 18132 }, { "epoch": 5.58, "learning_rate": 8.622689316331088e-06, "loss": 0.2017, "step": 18133 }, { "epoch": 5.58, "learning_rate": 8.621702280438615e-06, "loss": 0.1361, "step": 18134 }, { "epoch": 5.58, "learning_rate": 8.620715258233906e-06, "loss": 0.1519, "step": 18135 }, { "epoch": 5.58, "learning_rate": 8.619728249726771e-06, "loss": 0.1515, "step": 18136 }, { "epoch": 5.58, "learning_rate": 8.618741254927001e-06, "loss": 0.1438, "step": 18137 }, { "epoch": 5.58, "learning_rate": 8.617754273844407e-06, "loss": 0.1968, "step": 18138 }, { "epoch": 5.58, "learning_rate": 8.616767306488785e-06, "loss": 0.1372, "step": 18139 }, { "epoch": 5.58, "learning_rate": 8.615780352869938e-06, "loss": 0.1102, "step": 18140 }, { "epoch": 5.58, "learning_rate": 8.614793412997671e-06, "loss": 0.1801, "step": 18141 }, { "epoch": 5.58, "learning_rate": 8.613806486881775e-06, "loss": 0.1936, "step": 18142 }, { "epoch": 5.58, "learning_rate": 8.612819574532067e-06, "loss": 0.1767, "step": 18143 }, { "epoch": 5.58, "learning_rate": 8.611832675958335e-06, "loss": 0.1758, "step": 18144 }, { "epoch": 5.58, "learning_rate": 8.610845791170383e-06, "loss": 0.2136, "step": 18145 }, { "epoch": 5.58, "learning_rate": 8.609858920178015e-06, "loss": 0.1521, "step": 18146 }, { "epoch": 5.58, "learning_rate": 8.608872062991025e-06, "loss": 0.1254, "step": 18147 }, { "epoch": 5.58, "learning_rate": 8.607885219619223e-06, "loss": 0.1671, "step": 18148 }, { "epoch": 5.58, "learning_rate": 8.606898390072403e-06, "loss": 0.229, "step": 18149 }, { "epoch": 5.58, "learning_rate": 8.605911574360362e-06, "loss": 0.1922, "step": 18150 }, { "epoch": 5.58, "learning_rate": 8.604924772492905e-06, "loss": 0.1078, "step": 18151 }, { "epoch": 5.59, "learning_rate": 8.60393798447983e-06, "loss": 0.1769, "step": 18152 }, { "epoch": 5.59, "learning_rate": 8.602951210330942e-06, "loss": 0.2164, "step": 18153 }, { "epoch": 5.59, "learning_rate": 8.601964450056033e-06, "loss": 0.1642, "step": 18154 }, { "epoch": 5.59, "learning_rate": 8.600977703664904e-06, "loss": 0.1866, "step": 18155 }, { "epoch": 5.59, "learning_rate": 8.599990971167359e-06, "loss": 0.2273, "step": 18156 }, { "epoch": 5.59, "learning_rate": 8.599004252573191e-06, "loss": 0.0821, "step": 18157 }, { "epoch": 5.59, "learning_rate": 8.598017547892206e-06, "loss": 0.063, "step": 18158 }, { "epoch": 5.59, "learning_rate": 8.597030857134196e-06, "loss": 0.1543, "step": 18159 }, { "epoch": 5.59, "learning_rate": 8.596044180308961e-06, "loss": 0.1871, "step": 18160 }, { "epoch": 5.59, "learning_rate": 8.595057517426304e-06, "loss": 0.2408, "step": 18161 }, { "epoch": 5.59, "learning_rate": 8.594070868496018e-06, "loss": 0.1838, "step": 18162 }, { "epoch": 5.59, "learning_rate": 8.593084233527905e-06, "loss": 0.1156, "step": 18163 }, { "epoch": 5.59, "learning_rate": 8.592097612531766e-06, "loss": 0.1691, "step": 18164 }, { "epoch": 5.59, "learning_rate": 8.591111005517391e-06, "loss": 0.1281, "step": 18165 }, { "epoch": 5.59, "learning_rate": 8.590124412494584e-06, "loss": 0.1091, "step": 18166 }, { "epoch": 5.59, "learning_rate": 8.589137833473137e-06, "loss": 0.1767, "step": 18167 }, { "epoch": 5.59, "learning_rate": 8.588151268462857e-06, "loss": 0.2033, "step": 18168 }, { "epoch": 5.59, "learning_rate": 8.587164717473537e-06, "loss": 0.1759, "step": 18169 }, { "epoch": 5.59, "learning_rate": 8.586178180514968e-06, "loss": 0.1812, "step": 18170 }, { "epoch": 5.59, "learning_rate": 8.585191657596955e-06, "loss": 0.2547, "step": 18171 }, { "epoch": 5.59, "learning_rate": 8.584205148729292e-06, "loss": 0.17, "step": 18172 }, { "epoch": 5.59, "learning_rate": 8.583218653921777e-06, "loss": 0.2026, "step": 18173 }, { "epoch": 5.59, "learning_rate": 8.58223217318421e-06, "loss": 0.193, "step": 18174 }, { "epoch": 5.59, "learning_rate": 8.581245706526378e-06, "loss": 0.1966, "step": 18175 }, { "epoch": 5.59, "learning_rate": 8.58025925395809e-06, "loss": 0.227, "step": 18176 }, { "epoch": 5.59, "learning_rate": 8.579272815489131e-06, "loss": 0.1459, "step": 18177 }, { "epoch": 5.59, "learning_rate": 8.578286391129304e-06, "loss": 0.1414, "step": 18178 }, { "epoch": 5.59, "learning_rate": 8.577299980888407e-06, "loss": 0.2181, "step": 18179 }, { "epoch": 5.59, "learning_rate": 8.576313584776226e-06, "loss": 0.182, "step": 18180 }, { "epoch": 5.59, "learning_rate": 8.575327202802569e-06, "loss": 0.1086, "step": 18181 }, { "epoch": 5.59, "learning_rate": 8.574340834977222e-06, "loss": 0.1171, "step": 18182 }, { "epoch": 5.59, "learning_rate": 8.573354481309986e-06, "loss": 0.1731, "step": 18183 }, { "epoch": 5.59, "learning_rate": 8.572368141810656e-06, "loss": 0.1781, "step": 18184 }, { "epoch": 5.6, "learning_rate": 8.571381816489023e-06, "loss": 0.1753, "step": 18185 }, { "epoch": 5.6, "learning_rate": 8.570395505354888e-06, "loss": 0.2167, "step": 18186 }, { "epoch": 5.6, "learning_rate": 8.56940920841804e-06, "loss": 0.1348, "step": 18187 }, { "epoch": 5.6, "learning_rate": 8.56842292568828e-06, "loss": 0.1361, "step": 18188 }, { "epoch": 5.6, "learning_rate": 8.5674366571754e-06, "loss": 0.2584, "step": 18189 }, { "epoch": 5.6, "learning_rate": 8.566450402889192e-06, "loss": 0.182, "step": 18190 }, { "epoch": 5.6, "learning_rate": 8.565464162839456e-06, "loss": 0.2316, "step": 18191 }, { "epoch": 5.6, "learning_rate": 8.564477937035976e-06, "loss": 0.1347, "step": 18192 }, { "epoch": 5.6, "learning_rate": 8.563491725488562e-06, "loss": 0.1795, "step": 18193 }, { "epoch": 5.6, "learning_rate": 8.562505528206995e-06, "loss": 0.1876, "step": 18194 }, { "epoch": 5.6, "learning_rate": 8.56151934520107e-06, "loss": 0.1903, "step": 18195 }, { "epoch": 5.6, "learning_rate": 8.560533176480588e-06, "loss": 0.1951, "step": 18196 }, { "epoch": 5.6, "learning_rate": 8.559547022055335e-06, "loss": 0.2178, "step": 18197 }, { "epoch": 5.6, "learning_rate": 8.558560881935111e-06, "loss": 0.1317, "step": 18198 }, { "epoch": 5.6, "learning_rate": 8.558560881935111e-06, "loss": 0.179, "step": 18199 }, { "epoch": 5.6, "learning_rate": 8.557574756129707e-06, "loss": 0.1575, "step": 18200 }, { "epoch": 5.6, "learning_rate": 8.55658864464891e-06, "loss": 0.1742, "step": 18201 }, { "epoch": 5.6, "learning_rate": 8.55560254750252e-06, "loss": 0.2007, "step": 18202 }, { "epoch": 5.6, "learning_rate": 8.554616464700327e-06, "loss": 0.1009, "step": 18203 }, { "epoch": 5.6, "learning_rate": 8.553630396252129e-06, "loss": 0.2286, "step": 18204 }, { "epoch": 5.6, "learning_rate": 8.552644342167713e-06, "loss": 0.2239, "step": 18205 }, { "epoch": 5.6, "learning_rate": 8.551658302456868e-06, "loss": 0.2518, "step": 18206 }, { "epoch": 5.6, "learning_rate": 8.550672277129393e-06, "loss": 0.2049, "step": 18207 }, { "epoch": 5.6, "learning_rate": 8.549686266195078e-06, "loss": 0.1919, "step": 18208 }, { "epoch": 5.6, "learning_rate": 8.548700269663718e-06, "loss": 0.2029, "step": 18209 }, { "epoch": 5.6, "learning_rate": 8.5477142875451e-06, "loss": 0.1106, "step": 18210 }, { "epoch": 5.6, "learning_rate": 8.546728319849016e-06, "loss": 0.2024, "step": 18211 }, { "epoch": 5.6, "learning_rate": 8.54574236658526e-06, "loss": 0.1971, "step": 18212 }, { "epoch": 5.6, "learning_rate": 8.544756427763622e-06, "loss": 0.2183, "step": 18213 }, { "epoch": 5.6, "learning_rate": 8.543770503393895e-06, "loss": 0.1206, "step": 18214 }, { "epoch": 5.6, "learning_rate": 8.542784593485871e-06, "loss": 0.171, "step": 18215 }, { "epoch": 5.6, "learning_rate": 8.541798698049335e-06, "loss": 0.1894, "step": 18216 }, { "epoch": 5.61, "learning_rate": 8.540812817094084e-06, "loss": 0.1572, "step": 18217 }, { "epoch": 5.61, "learning_rate": 8.539826950629905e-06, "loss": 0.1115, "step": 18218 }, { "epoch": 5.61, "learning_rate": 8.538841098666591e-06, "loss": 0.1681, "step": 18219 }, { "epoch": 5.61, "learning_rate": 8.537855261213936e-06, "loss": 0.1943, "step": 18220 }, { "epoch": 5.61, "learning_rate": 8.536869438281718e-06, "loss": 0.2615, "step": 18221 }, { "epoch": 5.61, "learning_rate": 8.53588362987974e-06, "loss": 0.1587, "step": 18222 }, { "epoch": 5.61, "learning_rate": 8.534897836017784e-06, "loss": 0.2003, "step": 18223 }, { "epoch": 5.61, "learning_rate": 8.533912056705645e-06, "loss": 0.1975, "step": 18224 }, { "epoch": 5.61, "learning_rate": 8.532926291953113e-06, "loss": 0.1298, "step": 18225 }, { "epoch": 5.61, "learning_rate": 8.531940541769967e-06, "loss": 0.2148, "step": 18226 }, { "epoch": 5.61, "learning_rate": 8.530954806166012e-06, "loss": 0.1855, "step": 18227 }, { "epoch": 5.61, "learning_rate": 8.529969085151024e-06, "loss": 0.1851, "step": 18228 }, { "epoch": 5.61, "learning_rate": 8.528983378734802e-06, "loss": 0.3045, "step": 18229 }, { "epoch": 5.61, "learning_rate": 8.527997686927127e-06, "loss": 0.196, "step": 18230 }, { "epoch": 5.61, "learning_rate": 8.527012009737795e-06, "loss": 0.1863, "step": 18231 }, { "epoch": 5.61, "learning_rate": 8.526026347176593e-06, "loss": 0.1433, "step": 18232 }, { "epoch": 5.61, "learning_rate": 8.525040699253304e-06, "loss": 0.12, "step": 18233 }, { "epoch": 5.61, "learning_rate": 8.524055065977721e-06, "loss": 0.1161, "step": 18234 }, { "epoch": 5.61, "learning_rate": 8.523069447359631e-06, "loss": 0.1564, "step": 18235 }, { "epoch": 5.61, "learning_rate": 8.522083843408823e-06, "loss": 0.1107, "step": 18236 }, { "epoch": 5.61, "learning_rate": 8.521098254135087e-06, "loss": 0.2078, "step": 18237 }, { "epoch": 5.61, "learning_rate": 8.520112679548206e-06, "loss": 0.242, "step": 18238 }, { "epoch": 5.61, "learning_rate": 8.519127119657971e-06, "loss": 0.2089, "step": 18239 }, { "epoch": 5.61, "learning_rate": 8.518141574474167e-06, "loss": 0.1651, "step": 18240 }, { "epoch": 5.61, "learning_rate": 8.517156044006582e-06, "loss": 0.1353, "step": 18241 }, { "epoch": 5.61, "learning_rate": 8.51617052826501e-06, "loss": 0.1854, "step": 18242 }, { "epoch": 5.61, "learning_rate": 8.515185027259223e-06, "loss": 0.2463, "step": 18243 }, { "epoch": 5.61, "learning_rate": 8.514199540999025e-06, "loss": 0.1284, "step": 18244 }, { "epoch": 5.61, "learning_rate": 8.51321406949419e-06, "loss": 0.1409, "step": 18245 }, { "epoch": 5.61, "learning_rate": 8.512228612754511e-06, "loss": 0.1854, "step": 18246 }, { "epoch": 5.61, "learning_rate": 8.511243170789774e-06, "loss": 0.1385, "step": 18247 }, { "epoch": 5.61, "learning_rate": 8.510257743609763e-06, "loss": 0.2225, "step": 18248 }, { "epoch": 5.61, "learning_rate": 8.50927233122427e-06, "loss": 0.1401, "step": 18249 }, { "epoch": 5.62, "learning_rate": 8.50828693364307e-06, "loss": 0.1466, "step": 18250 }, { "epoch": 5.62, "learning_rate": 8.50730155087596e-06, "loss": 0.2197, "step": 18251 }, { "epoch": 5.62, "learning_rate": 8.506316182932722e-06, "loss": 0.2099, "step": 18252 }, { "epoch": 5.62, "learning_rate": 8.505330829823137e-06, "loss": 0.198, "step": 18253 }, { "epoch": 5.62, "learning_rate": 8.504345491556998e-06, "loss": 0.1774, "step": 18254 }, { "epoch": 5.62, "learning_rate": 8.503360168144081e-06, "loss": 0.245, "step": 18255 }, { "epoch": 5.62, "learning_rate": 8.502374859594184e-06, "loss": 0.2087, "step": 18256 }, { "epoch": 5.62, "learning_rate": 8.501389565917082e-06, "loss": 0.1014, "step": 18257 }, { "epoch": 5.62, "learning_rate": 8.500404287122561e-06, "loss": 0.1493, "step": 18258 }, { "epoch": 5.62, "learning_rate": 8.499419023220413e-06, "loss": 0.2404, "step": 18259 }, { "epoch": 5.62, "learning_rate": 8.498433774220408e-06, "loss": 0.1392, "step": 18260 }, { "epoch": 5.62, "learning_rate": 8.497448540132347e-06, "loss": 0.1927, "step": 18261 }, { "epoch": 5.62, "learning_rate": 8.496463320966004e-06, "loss": 0.2166, "step": 18262 }, { "epoch": 5.62, "learning_rate": 8.495478116731166e-06, "loss": 0.1518, "step": 18263 }, { "epoch": 5.62, "learning_rate": 8.494492927437619e-06, "loss": 0.0875, "step": 18264 }, { "epoch": 5.62, "learning_rate": 8.49350775309514e-06, "loss": 0.1105, "step": 18265 }, { "epoch": 5.62, "learning_rate": 8.492522593713523e-06, "loss": 0.235, "step": 18266 }, { "epoch": 5.62, "learning_rate": 8.491537449302545e-06, "loss": 0.1694, "step": 18267 }, { "epoch": 5.62, "learning_rate": 8.490552319871986e-06, "loss": 0.1029, "step": 18268 }, { "epoch": 5.62, "learning_rate": 8.489567205431638e-06, "loss": 0.1595, "step": 18269 }, { "epoch": 5.62, "learning_rate": 8.488582105991277e-06, "loss": 0.1982, "step": 18270 }, { "epoch": 5.62, "learning_rate": 8.487597021560692e-06, "loss": 0.1746, "step": 18271 }, { "epoch": 5.62, "learning_rate": 8.486611952149661e-06, "loss": 0.1636, "step": 18272 }, { "epoch": 5.62, "learning_rate": 8.485626897767964e-06, "loss": 0.1827, "step": 18273 }, { "epoch": 5.62, "learning_rate": 8.484641858425394e-06, "loss": 0.2119, "step": 18274 }, { "epoch": 5.62, "learning_rate": 8.48365683413172e-06, "loss": 0.1544, "step": 18275 }, { "epoch": 5.62, "learning_rate": 8.482671824896737e-06, "loss": 0.2485, "step": 18276 }, { "epoch": 5.62, "learning_rate": 8.481686830730222e-06, "loss": 0.1236, "step": 18277 }, { "epoch": 5.62, "learning_rate": 8.480701851641951e-06, "loss": 0.2038, "step": 18278 }, { "epoch": 5.62, "learning_rate": 8.479716887641714e-06, "loss": 0.1298, "step": 18279 }, { "epoch": 5.62, "learning_rate": 8.478731938739286e-06, "loss": 0.1456, "step": 18280 }, { "epoch": 5.62, "learning_rate": 8.477747004944455e-06, "loss": 0.2401, "step": 18281 }, { "epoch": 5.63, "learning_rate": 8.476762086267002e-06, "loss": 0.187, "step": 18282 }, { "epoch": 5.63, "learning_rate": 8.4757771827167e-06, "loss": 0.1715, "step": 18283 }, { "epoch": 5.63, "learning_rate": 8.474792294303338e-06, "loss": 0.1754, "step": 18284 }, { "epoch": 5.63, "learning_rate": 8.473807421036692e-06, "loss": 0.2077, "step": 18285 }, { "epoch": 5.63, "learning_rate": 8.472822562926546e-06, "loss": 0.1825, "step": 18286 }, { "epoch": 5.63, "learning_rate": 8.471837719982683e-06, "loss": 0.293, "step": 18287 }, { "epoch": 5.63, "learning_rate": 8.470852892214875e-06, "loss": 0.1965, "step": 18288 }, { "epoch": 5.63, "learning_rate": 8.469868079632909e-06, "loss": 0.1945, "step": 18289 }, { "epoch": 5.63, "learning_rate": 8.468883282246562e-06, "loss": 0.1858, "step": 18290 }, { "epoch": 5.63, "learning_rate": 8.467898500065616e-06, "loss": 0.1502, "step": 18291 }, { "epoch": 5.63, "learning_rate": 8.466913733099852e-06, "loss": 0.3081, "step": 18292 }, { "epoch": 5.63, "learning_rate": 8.465928981359043e-06, "loss": 0.1458, "step": 18293 }, { "epoch": 5.63, "learning_rate": 8.464944244852977e-06, "loss": 0.1358, "step": 18294 }, { "epoch": 5.63, "learning_rate": 8.463959523591428e-06, "loss": 0.2165, "step": 18295 }, { "epoch": 5.63, "learning_rate": 8.462974817584177e-06, "loss": 0.1582, "step": 18296 }, { "epoch": 5.63, "learning_rate": 8.461990126841004e-06, "loss": 0.2137, "step": 18297 }, { "epoch": 5.63, "learning_rate": 8.461005451371682e-06, "loss": 0.2348, "step": 18298 }, { "epoch": 5.63, "learning_rate": 8.460020791185998e-06, "loss": 0.1739, "step": 18299 }, { "epoch": 5.63, "learning_rate": 8.459036146293725e-06, "loss": 0.1387, "step": 18300 }, { "epoch": 5.63, "learning_rate": 8.458051516704644e-06, "loss": 0.2092, "step": 18301 }, { "epoch": 5.63, "learning_rate": 8.457066902428534e-06, "loss": 0.1425, "step": 18302 }, { "epoch": 5.63, "learning_rate": 8.456082303475169e-06, "loss": 0.2106, "step": 18303 }, { "epoch": 5.63, "learning_rate": 8.455097719854333e-06, "loss": 0.0939, "step": 18304 }, { "epoch": 5.63, "learning_rate": 8.454113151575795e-06, "loss": 0.2575, "step": 18305 }, { "epoch": 5.63, "learning_rate": 8.453128598649343e-06, "loss": 0.2198, "step": 18306 }, { "epoch": 5.63, "learning_rate": 8.452144061084749e-06, "loss": 0.1811, "step": 18307 }, { "epoch": 5.63, "learning_rate": 8.45115953889179e-06, "loss": 0.1596, "step": 18308 }, { "epoch": 5.63, "learning_rate": 8.450175032080245e-06, "loss": 0.172, "step": 18309 }, { "epoch": 5.63, "learning_rate": 8.449190540659888e-06, "loss": 0.1948, "step": 18310 }, { "epoch": 5.63, "learning_rate": 8.448206064640502e-06, "loss": 0.2045, "step": 18311 }, { "epoch": 5.63, "learning_rate": 8.447221604031858e-06, "loss": 0.1499, "step": 18312 }, { "epoch": 5.63, "learning_rate": 8.446237158843733e-06, "loss": 0.1366, "step": 18313 }, { "epoch": 5.63, "learning_rate": 8.445252729085907e-06, "loss": 0.2742, "step": 18314 }, { "epoch": 5.64, "learning_rate": 8.444268314768152e-06, "loss": 0.1592, "step": 18315 }, { "epoch": 5.64, "learning_rate": 8.44328391590025e-06, "loss": 0.0993, "step": 18316 }, { "epoch": 5.64, "learning_rate": 8.442299532491973e-06, "loss": 0.159, "step": 18317 }, { "epoch": 5.64, "learning_rate": 8.441315164553095e-06, "loss": 0.1469, "step": 18318 }, { "epoch": 5.64, "learning_rate": 8.440330812093395e-06, "loss": 0.1772, "step": 18319 }, { "epoch": 5.64, "learning_rate": 8.439346475122645e-06, "loss": 0.1631, "step": 18320 }, { "epoch": 5.64, "learning_rate": 8.438362153650628e-06, "loss": 0.1406, "step": 18321 }, { "epoch": 5.64, "learning_rate": 8.437377847687111e-06, "loss": 0.0952, "step": 18322 }, { "epoch": 5.64, "learning_rate": 8.436393557241872e-06, "loss": 0.1385, "step": 18323 }, { "epoch": 5.64, "learning_rate": 8.435409282324686e-06, "loss": 0.1146, "step": 18324 }, { "epoch": 5.64, "learning_rate": 8.434425022945326e-06, "loss": 0.1485, "step": 18325 }, { "epoch": 5.64, "learning_rate": 8.433440779113574e-06, "loss": 0.2029, "step": 18326 }, { "epoch": 5.64, "learning_rate": 8.432456550839196e-06, "loss": 0.2079, "step": 18327 }, { "epoch": 5.64, "learning_rate": 8.431472338131967e-06, "loss": 0.0897, "step": 18328 }, { "epoch": 5.64, "learning_rate": 8.430488141001665e-06, "loss": 0.2319, "step": 18329 }, { "epoch": 5.64, "learning_rate": 8.429503959458061e-06, "loss": 0.1241, "step": 18330 }, { "epoch": 5.64, "learning_rate": 8.42851979351093e-06, "loss": 0.1715, "step": 18331 }, { "epoch": 5.64, "learning_rate": 8.427535643170051e-06, "loss": 0.2027, "step": 18332 }, { "epoch": 5.64, "learning_rate": 8.426551508445187e-06, "loss": 0.17, "step": 18333 }, { "epoch": 5.64, "learning_rate": 8.425567389346118e-06, "loss": 0.1977, "step": 18334 }, { "epoch": 5.64, "learning_rate": 8.424583285882614e-06, "loss": 0.1694, "step": 18335 }, { "epoch": 5.64, "learning_rate": 8.423599198064452e-06, "loss": 0.174, "step": 18336 }, { "epoch": 5.64, "learning_rate": 8.422615125901404e-06, "loss": 0.1746, "step": 18337 }, { "epoch": 5.64, "learning_rate": 8.42163106940324e-06, "loss": 0.2094, "step": 18338 }, { "epoch": 5.64, "learning_rate": 8.420647028579734e-06, "loss": 0.1296, "step": 18339 }, { "epoch": 5.64, "learning_rate": 8.419663003440657e-06, "loss": 0.2016, "step": 18340 }, { "epoch": 5.64, "learning_rate": 8.418678993995785e-06, "loss": 0.2348, "step": 18341 }, { "epoch": 5.64, "learning_rate": 8.417695000254891e-06, "loss": 0.16, "step": 18342 }, { "epoch": 5.64, "learning_rate": 8.416711022227736e-06, "loss": 0.181, "step": 18343 }, { "epoch": 5.64, "learning_rate": 8.415727059924107e-06, "loss": 0.1238, "step": 18344 }, { "epoch": 5.64, "learning_rate": 8.414743113353764e-06, "loss": 0.1891, "step": 18345 }, { "epoch": 5.64, "learning_rate": 8.413759182526485e-06, "loss": 0.1001, "step": 18346 }, { "epoch": 5.65, "learning_rate": 8.412775267452042e-06, "loss": 0.0903, "step": 18347 }, { "epoch": 5.65, "learning_rate": 8.411791368140197e-06, "loss": 0.2488, "step": 18348 }, { "epoch": 5.65, "learning_rate": 8.410807484600732e-06, "loss": 0.1613, "step": 18349 }, { "epoch": 5.65, "learning_rate": 8.40982361684341e-06, "loss": 0.2708, "step": 18350 }, { "epoch": 5.65, "learning_rate": 8.408839764878009e-06, "loss": 0.1883, "step": 18351 }, { "epoch": 5.65, "learning_rate": 8.407855928714294e-06, "loss": 0.2452, "step": 18352 }, { "epoch": 5.65, "learning_rate": 8.406872108362034e-06, "loss": 0.1457, "step": 18353 }, { "epoch": 5.65, "learning_rate": 8.405888303831009e-06, "loss": 0.199, "step": 18354 }, { "epoch": 5.65, "learning_rate": 8.404904515130977e-06, "loss": 0.1876, "step": 18355 }, { "epoch": 5.65, "learning_rate": 8.403920742271713e-06, "loss": 0.1385, "step": 18356 }, { "epoch": 5.65, "learning_rate": 8.402936985262989e-06, "loss": 0.1524, "step": 18357 }, { "epoch": 5.65, "learning_rate": 8.401953244114572e-06, "loss": 0.1161, "step": 18358 }, { "epoch": 5.65, "learning_rate": 8.400969518836234e-06, "loss": 0.074, "step": 18359 }, { "epoch": 5.65, "learning_rate": 8.399985809437737e-06, "loss": 0.1428, "step": 18360 }, { "epoch": 5.65, "learning_rate": 8.399002115928862e-06, "loss": 0.1444, "step": 18361 }, { "epoch": 5.65, "learning_rate": 8.398018438319368e-06, "loss": 0.2285, "step": 18362 }, { "epoch": 5.65, "learning_rate": 8.397034776619028e-06, "loss": 0.1952, "step": 18363 }, { "epoch": 5.65, "learning_rate": 8.396051130837608e-06, "loss": 0.1207, "step": 18364 }, { "epoch": 5.65, "learning_rate": 8.39506750098488e-06, "loss": 0.1775, "step": 18365 }, { "epoch": 5.65, "learning_rate": 8.394083887070614e-06, "loss": 0.1215, "step": 18366 }, { "epoch": 5.65, "learning_rate": 8.393100289104574e-06, "loss": 0.1763, "step": 18367 }, { "epoch": 5.65, "learning_rate": 8.392116707096526e-06, "loss": 0.0838, "step": 18368 }, { "epoch": 5.65, "learning_rate": 8.391133141056243e-06, "loss": 0.1682, "step": 18369 }, { "epoch": 5.65, "learning_rate": 8.390149590993488e-06, "loss": 0.197, "step": 18370 }, { "epoch": 5.65, "learning_rate": 8.389166056918035e-06, "loss": 0.1459, "step": 18371 }, { "epoch": 5.65, "learning_rate": 8.388182538839648e-06, "loss": 0.11, "step": 18372 }, { "epoch": 5.65, "learning_rate": 8.38719903676809e-06, "loss": 0.2263, "step": 18373 }, { "epoch": 5.65, "learning_rate": 8.386215550713133e-06, "loss": 0.1689, "step": 18374 }, { "epoch": 5.65, "learning_rate": 8.385232080684543e-06, "loss": 0.2252, "step": 18375 }, { "epoch": 5.65, "learning_rate": 8.38424862669209e-06, "loss": 0.1795, "step": 18376 }, { "epoch": 5.65, "learning_rate": 8.383265188745535e-06, "loss": 0.1232, "step": 18377 }, { "epoch": 5.65, "learning_rate": 8.382281766854646e-06, "loss": 0.1468, "step": 18378 }, { "epoch": 5.65, "learning_rate": 8.38129836102919e-06, "loss": 0.225, "step": 18379 }, { "epoch": 5.66, "learning_rate": 8.380314971278932e-06, "loss": 0.3094, "step": 18380 }, { "epoch": 5.66, "learning_rate": 8.37933159761364e-06, "loss": 0.121, "step": 18381 }, { "epoch": 5.66, "learning_rate": 8.378348240043083e-06, "loss": 0.1822, "step": 18382 }, { "epoch": 5.66, "learning_rate": 8.377364898577017e-06, "loss": 0.2882, "step": 18383 }, { "epoch": 5.66, "learning_rate": 8.376381573225217e-06, "loss": 0.1538, "step": 18384 }, { "epoch": 5.66, "learning_rate": 8.37539826399744e-06, "loss": 0.2435, "step": 18385 }, { "epoch": 5.66, "learning_rate": 8.374414970903458e-06, "loss": 0.1418, "step": 18386 }, { "epoch": 5.66, "learning_rate": 8.373431693953036e-06, "loss": 0.0928, "step": 18387 }, { "epoch": 5.66, "learning_rate": 8.372448433155931e-06, "loss": 0.2332, "step": 18388 }, { "epoch": 5.66, "learning_rate": 8.371465188521916e-06, "loss": 0.2859, "step": 18389 }, { "epoch": 5.66, "learning_rate": 8.370481960060753e-06, "loss": 0.0818, "step": 18390 }, { "epoch": 5.66, "learning_rate": 8.369498747782205e-06, "loss": 0.1324, "step": 18391 }, { "epoch": 5.66, "learning_rate": 8.36851555169604e-06, "loss": 0.1915, "step": 18392 }, { "epoch": 5.66, "learning_rate": 8.367532371812012e-06, "loss": 0.137, "step": 18393 }, { "epoch": 5.66, "learning_rate": 8.3665492081399e-06, "loss": 0.1424, "step": 18394 }, { "epoch": 5.66, "learning_rate": 8.365566060689455e-06, "loss": 0.1314, "step": 18395 }, { "epoch": 5.66, "learning_rate": 8.364582929470448e-06, "loss": 0.1205, "step": 18396 }, { "epoch": 5.66, "learning_rate": 8.363599814492641e-06, "loss": 0.2252, "step": 18397 }, { "epoch": 5.66, "learning_rate": 8.36261671576579e-06, "loss": 0.1532, "step": 18398 }, { "epoch": 5.66, "learning_rate": 8.36163363329967e-06, "loss": 0.1577, "step": 18399 }, { "epoch": 5.66, "learning_rate": 8.360650567104034e-06, "loss": 0.2295, "step": 18400 }, { "epoch": 5.66, "learning_rate": 8.359667517188652e-06, "loss": 0.1502, "step": 18401 }, { "epoch": 5.66, "learning_rate": 8.358684483563283e-06, "loss": 0.1776, "step": 18402 }, { "epoch": 5.66, "learning_rate": 8.357701466237686e-06, "loss": 0.1465, "step": 18403 }, { "epoch": 5.66, "learning_rate": 8.356718465221632e-06, "loss": 0.1638, "step": 18404 }, { "epoch": 5.66, "learning_rate": 8.355735480524874e-06, "loss": 0.1667, "step": 18405 }, { "epoch": 5.66, "learning_rate": 8.354752512157181e-06, "loss": 0.0879, "step": 18406 }, { "epoch": 5.66, "learning_rate": 8.35376956012831e-06, "loss": 0.1738, "step": 18407 }, { "epoch": 5.66, "learning_rate": 8.352786624448024e-06, "loss": 0.1152, "step": 18408 }, { "epoch": 5.66, "learning_rate": 8.351803705126087e-06, "loss": 0.2621, "step": 18409 }, { "epoch": 5.66, "learning_rate": 8.350820802172256e-06, "loss": 0.2177, "step": 18410 }, { "epoch": 5.66, "learning_rate": 8.349837915596297e-06, "loss": 0.1493, "step": 18411 }, { "epoch": 5.67, "learning_rate": 8.348855045407967e-06, "loss": 0.2779, "step": 18412 }, { "epoch": 5.67, "learning_rate": 8.347872191617027e-06, "loss": 0.1823, "step": 18413 }, { "epoch": 5.67, "learning_rate": 8.34688935423324e-06, "loss": 0.2239, "step": 18414 }, { "epoch": 5.67, "learning_rate": 8.345906533266364e-06, "loss": 0.1644, "step": 18415 }, { "epoch": 5.67, "learning_rate": 8.344923728726165e-06, "loss": 0.1557, "step": 18416 }, { "epoch": 5.67, "learning_rate": 8.343940940622394e-06, "loss": 0.1776, "step": 18417 }, { "epoch": 5.67, "learning_rate": 8.342958168964816e-06, "loss": 0.2862, "step": 18418 }, { "epoch": 5.67, "learning_rate": 8.341975413763193e-06, "loss": 0.1794, "step": 18419 }, { "epoch": 5.67, "learning_rate": 8.34099267502728e-06, "loss": 0.2164, "step": 18420 }, { "epoch": 5.67, "learning_rate": 8.340009952766843e-06, "loss": 0.1105, "step": 18421 }, { "epoch": 5.67, "learning_rate": 8.339027246991633e-06, "loss": 0.2036, "step": 18422 }, { "epoch": 5.67, "learning_rate": 8.338044557711413e-06, "loss": 0.145, "step": 18423 }, { "epoch": 5.67, "learning_rate": 8.337061884935944e-06, "loss": 0.1399, "step": 18424 }, { "epoch": 5.67, "learning_rate": 8.33607922867498e-06, "loss": 0.1989, "step": 18425 }, { "epoch": 5.67, "learning_rate": 8.335096588938288e-06, "loss": 0.1906, "step": 18426 }, { "epoch": 5.67, "learning_rate": 8.334113965735614e-06, "loss": 0.3032, "step": 18427 }, { "epoch": 5.67, "learning_rate": 8.33313135907673e-06, "loss": 0.2096, "step": 18428 }, { "epoch": 5.67, "learning_rate": 8.332148768971388e-06, "loss": 0.2268, "step": 18429 }, { "epoch": 5.67, "learning_rate": 8.331166195429342e-06, "loss": 0.1978, "step": 18430 }, { "epoch": 5.67, "learning_rate": 8.330183638460356e-06, "loss": 0.1765, "step": 18431 }, { "epoch": 5.67, "learning_rate": 8.329201098074182e-06, "loss": 0.209, "step": 18432 }, { "epoch": 5.67, "learning_rate": 8.328218574280588e-06, "loss": 0.2622, "step": 18433 }, { "epoch": 5.67, "learning_rate": 8.32723606708932e-06, "loss": 0.2318, "step": 18434 }, { "epoch": 5.67, "learning_rate": 8.326253576510138e-06, "loss": 0.1942, "step": 18435 }, { "epoch": 5.67, "learning_rate": 8.325271102552803e-06, "loss": 0.2101, "step": 18436 }, { "epoch": 5.67, "learning_rate": 8.324288645227066e-06, "loss": 0.1697, "step": 18437 }, { "epoch": 5.67, "learning_rate": 8.323306204542694e-06, "loss": 0.2531, "step": 18438 }, { "epoch": 5.67, "learning_rate": 8.322323780509432e-06, "loss": 0.2146, "step": 18439 }, { "epoch": 5.67, "learning_rate": 8.32134137313704e-06, "loss": 0.1435, "step": 18440 }, { "epoch": 5.67, "learning_rate": 8.320358982435276e-06, "loss": 0.1244, "step": 18441 }, { "epoch": 5.67, "learning_rate": 8.319376608413896e-06, "loss": 0.1798, "step": 18442 }, { "epoch": 5.67, "learning_rate": 8.318394251082655e-06, "loss": 0.1967, "step": 18443 }, { "epoch": 5.67, "learning_rate": 8.317411910451313e-06, "loss": 0.144, "step": 18444 }, { "epoch": 5.68, "learning_rate": 8.316429586529616e-06, "loss": 0.2705, "step": 18445 }, { "epoch": 5.68, "learning_rate": 8.315447279327326e-06, "loss": 0.1535, "step": 18446 }, { "epoch": 5.68, "learning_rate": 8.314464988854196e-06, "loss": 0.1068, "step": 18447 }, { "epoch": 5.68, "learning_rate": 8.313482715119985e-06, "loss": 0.1211, "step": 18448 }, { "epoch": 5.68, "learning_rate": 8.312500458134446e-06, "loss": 0.1865, "step": 18449 }, { "epoch": 5.68, "learning_rate": 8.311518217907328e-06, "loss": 0.1047, "step": 18450 }, { "epoch": 5.68, "learning_rate": 8.310535994448393e-06, "loss": 0.0712, "step": 18451 }, { "epoch": 5.68, "learning_rate": 8.309553787767391e-06, "loss": 0.1603, "step": 18452 }, { "epoch": 5.68, "learning_rate": 8.30857159787408e-06, "loss": 0.1879, "step": 18453 }, { "epoch": 5.68, "learning_rate": 8.307589424778214e-06, "loss": 0.1852, "step": 18454 }, { "epoch": 5.68, "learning_rate": 8.30660726848954e-06, "loss": 0.2077, "step": 18455 }, { "epoch": 5.68, "learning_rate": 8.305625129017818e-06, "loss": 0.1288, "step": 18456 }, { "epoch": 5.68, "learning_rate": 8.304643006372797e-06, "loss": 0.1869, "step": 18457 }, { "epoch": 5.68, "learning_rate": 8.303660900564239e-06, "loss": 0.2614, "step": 18458 }, { "epoch": 5.68, "learning_rate": 8.302678811601892e-06, "loss": 0.155, "step": 18459 }, { "epoch": 5.68, "learning_rate": 8.301696739495502e-06, "loss": 0.1893, "step": 18460 }, { "epoch": 5.68, "learning_rate": 8.300714684254834e-06, "loss": 0.1732, "step": 18461 }, { "epoch": 5.68, "learning_rate": 8.299732645889633e-06, "loss": 0.2546, "step": 18462 }, { "epoch": 5.68, "learning_rate": 8.298750624409655e-06, "loss": 0.1635, "step": 18463 }, { "epoch": 5.68, "learning_rate": 8.297768619824653e-06, "loss": 0.0927, "step": 18464 }, { "epoch": 5.68, "learning_rate": 8.296786632144373e-06, "loss": 0.1134, "step": 18465 }, { "epoch": 5.68, "learning_rate": 8.295804661378576e-06, "loss": 0.2397, "step": 18466 }, { "epoch": 5.68, "learning_rate": 8.294822707537006e-06, "loss": 0.1841, "step": 18467 }, { "epoch": 5.68, "learning_rate": 8.29384077062942e-06, "loss": 0.2941, "step": 18468 }, { "epoch": 5.68, "learning_rate": 8.292858850665567e-06, "loss": 0.2252, "step": 18469 }, { "epoch": 5.68, "learning_rate": 8.291876947655197e-06, "loss": 0.1439, "step": 18470 }, { "epoch": 5.68, "learning_rate": 8.290895061608067e-06, "loss": 0.1531, "step": 18471 }, { "epoch": 5.68, "learning_rate": 8.289913192533918e-06, "loss": 0.1691, "step": 18472 }, { "epoch": 5.68, "learning_rate": 8.288931340442514e-06, "loss": 0.1714, "step": 18473 }, { "epoch": 5.68, "learning_rate": 8.287949505343598e-06, "loss": 0.2302, "step": 18474 }, { "epoch": 5.68, "learning_rate": 8.286967687246916e-06, "loss": 0.1672, "step": 18475 }, { "epoch": 5.68, "learning_rate": 8.28598588616223e-06, "loss": 0.1356, "step": 18476 }, { "epoch": 5.69, "learning_rate": 8.285004102099277e-06, "loss": 0.1603, "step": 18477 }, { "epoch": 5.69, "learning_rate": 8.28402233506782e-06, "loss": 0.255, "step": 18478 }, { "epoch": 5.69, "learning_rate": 8.283040585077601e-06, "loss": 0.1277, "step": 18479 }, { "epoch": 5.69, "learning_rate": 8.282058852138369e-06, "loss": 0.2365, "step": 18480 }, { "epoch": 5.69, "learning_rate": 8.28107713625988e-06, "loss": 0.2303, "step": 18481 }, { "epoch": 5.69, "learning_rate": 8.280095437451873e-06, "loss": 0.195, "step": 18482 }, { "epoch": 5.69, "learning_rate": 8.27911375572411e-06, "loss": 0.1199, "step": 18483 }, { "epoch": 5.69, "learning_rate": 8.278132091086332e-06, "loss": 0.0948, "step": 18484 }, { "epoch": 5.69, "learning_rate": 8.277150443548285e-06, "loss": 0.1245, "step": 18485 }, { "epoch": 5.69, "learning_rate": 8.276168813119726e-06, "loss": 0.2202, "step": 18486 }, { "epoch": 5.69, "learning_rate": 8.275187199810396e-06, "loss": 0.1549, "step": 18487 }, { "epoch": 5.69, "learning_rate": 8.274205603630052e-06, "loss": 0.3927, "step": 18488 }, { "epoch": 5.69, "learning_rate": 8.273224024588434e-06, "loss": 0.1778, "step": 18489 }, { "epoch": 5.69, "learning_rate": 8.272242462695291e-06, "loss": 0.1453, "step": 18490 }, { "epoch": 5.69, "learning_rate": 8.271260917960375e-06, "loss": 0.1857, "step": 18491 }, { "epoch": 5.69, "learning_rate": 8.27027939039343e-06, "loss": 0.1685, "step": 18492 }, { "epoch": 5.69, "learning_rate": 8.269297880004205e-06, "loss": 0.1235, "step": 18493 }, { "epoch": 5.69, "learning_rate": 8.268316386802451e-06, "loss": 0.1374, "step": 18494 }, { "epoch": 5.69, "learning_rate": 8.267334910797907e-06, "loss": 0.1686, "step": 18495 }, { "epoch": 5.69, "learning_rate": 8.266353452000326e-06, "loss": 0.191, "step": 18496 }, { "epoch": 5.69, "learning_rate": 8.26537201041945e-06, "loss": 0.2845, "step": 18497 }, { "epoch": 5.69, "learning_rate": 8.264390586065031e-06, "loss": 0.1815, "step": 18498 }, { "epoch": 5.69, "learning_rate": 8.263409178946816e-06, "loss": 0.2211, "step": 18499 }, { "epoch": 5.69, "learning_rate": 8.262427789074543e-06, "loss": 0.1842, "step": 18500 }, { "epoch": 5.69, "learning_rate": 8.261446416457966e-06, "loss": 0.0756, "step": 18501 }, { "epoch": 5.69, "learning_rate": 8.260465061106826e-06, "loss": 0.1857, "step": 18502 }, { "epoch": 5.69, "learning_rate": 8.259483723030874e-06, "loss": 0.2309, "step": 18503 }, { "epoch": 5.69, "learning_rate": 8.258502402239854e-06, "loss": 0.1967, "step": 18504 }, { "epoch": 5.69, "learning_rate": 8.257521098743507e-06, "loss": 0.1683, "step": 18505 }, { "epoch": 5.69, "learning_rate": 8.256539812551582e-06, "loss": 0.2401, "step": 18506 }, { "epoch": 5.69, "learning_rate": 8.255558543673823e-06, "loss": 0.1662, "step": 18507 }, { "epoch": 5.69, "learning_rate": 8.254577292119977e-06, "loss": 0.1908, "step": 18508 }, { "epoch": 5.69, "learning_rate": 8.253596057899788e-06, "loss": 0.2488, "step": 18509 }, { "epoch": 5.7, "learning_rate": 8.252614841022996e-06, "loss": 0.1611, "step": 18510 }, { "epoch": 5.7, "learning_rate": 8.251633641499353e-06, "loss": 0.1971, "step": 18511 }, { "epoch": 5.7, "learning_rate": 8.250652459338596e-06, "loss": 0.1628, "step": 18512 }, { "epoch": 5.7, "learning_rate": 8.249671294550473e-06, "loss": 0.2051, "step": 18513 }, { "epoch": 5.7, "learning_rate": 8.24869014714473e-06, "loss": 0.19, "step": 18514 }, { "epoch": 5.7, "learning_rate": 8.247709017131104e-06, "loss": 0.2709, "step": 18515 }, { "epoch": 5.7, "learning_rate": 8.246727904519348e-06, "loss": 0.0898, "step": 18516 }, { "epoch": 5.7, "learning_rate": 8.245746809319194e-06, "loss": 0.2438, "step": 18517 }, { "epoch": 5.7, "learning_rate": 8.244765731540393e-06, "loss": 0.1683, "step": 18518 }, { "epoch": 5.7, "learning_rate": 8.243784671192687e-06, "loss": 0.1562, "step": 18519 }, { "epoch": 5.7, "learning_rate": 8.242803628285814e-06, "loss": 0.2003, "step": 18520 }, { "epoch": 5.7, "learning_rate": 8.241822602829527e-06, "loss": 0.0782, "step": 18521 }, { "epoch": 5.7, "learning_rate": 8.240841594833554e-06, "loss": 0.1259, "step": 18522 }, { "epoch": 5.7, "learning_rate": 8.239860604307653e-06, "loss": 0.1477, "step": 18523 }, { "epoch": 5.7, "learning_rate": 8.238879631261556e-06, "loss": 0.276, "step": 18524 }, { "epoch": 5.7, "learning_rate": 8.237898675705003e-06, "loss": 0.1942, "step": 18525 }, { "epoch": 5.7, "learning_rate": 8.236917737647747e-06, "loss": 0.1199, "step": 18526 }, { "epoch": 5.7, "learning_rate": 8.235936817099515e-06, "loss": 0.2014, "step": 18527 }, { "epoch": 5.7, "learning_rate": 8.234955914070062e-06, "loss": 0.1737, "step": 18528 }, { "epoch": 5.7, "learning_rate": 8.233975028569123e-06, "loss": 0.1162, "step": 18529 }, { "epoch": 5.7, "learning_rate": 8.232994160606436e-06, "loss": 0.1146, "step": 18530 }, { "epoch": 5.7, "learning_rate": 8.232013310191748e-06, "loss": 0.2887, "step": 18531 }, { "epoch": 5.7, "learning_rate": 8.231032477334796e-06, "loss": 0.0899, "step": 18532 }, { "epoch": 5.7, "learning_rate": 8.230051662045326e-06, "loss": 0.2046, "step": 18533 }, { "epoch": 5.7, "learning_rate": 8.229070864333072e-06, "loss": 0.1632, "step": 18534 }, { "epoch": 5.7, "learning_rate": 8.228090084207773e-06, "loss": 0.0765, "step": 18535 }, { "epoch": 5.7, "learning_rate": 8.227109321679177e-06, "loss": 0.1653, "step": 18536 }, { "epoch": 5.7, "learning_rate": 8.226128576757015e-06, "loss": 0.1378, "step": 18537 }, { "epoch": 5.7, "learning_rate": 8.225147849451035e-06, "loss": 0.1874, "step": 18538 }, { "epoch": 5.7, "learning_rate": 8.224167139770973e-06, "loss": 0.2353, "step": 18539 }, { "epoch": 5.7, "learning_rate": 8.223186447726564e-06, "loss": 0.177, "step": 18540 }, { "epoch": 5.7, "learning_rate": 8.222205773327555e-06, "loss": 0.1517, "step": 18541 }, { "epoch": 5.71, "learning_rate": 8.221225116583677e-06, "loss": 0.2082, "step": 18542 }, { "epoch": 5.71, "learning_rate": 8.220244477504677e-06, "loss": 0.2097, "step": 18543 }, { "epoch": 5.71, "learning_rate": 8.21926385610029e-06, "loss": 0.1696, "step": 18544 }, { "epoch": 5.71, "learning_rate": 8.21828325238025e-06, "loss": 0.1645, "step": 18545 }, { "epoch": 5.71, "learning_rate": 8.217302666354301e-06, "loss": 0.0993, "step": 18546 }, { "epoch": 5.71, "learning_rate": 8.21632209803218e-06, "loss": 0.1519, "step": 18547 }, { "epoch": 5.71, "learning_rate": 8.215341547423624e-06, "loss": 0.1517, "step": 18548 }, { "epoch": 5.71, "learning_rate": 8.214361014538374e-06, "loss": 0.1453, "step": 18549 }, { "epoch": 5.71, "learning_rate": 8.21338049938616e-06, "loss": 0.1231, "step": 18550 }, { "epoch": 5.71, "learning_rate": 8.212400001976729e-06, "loss": 0.2136, "step": 18551 }, { "epoch": 5.71, "learning_rate": 8.211419522319808e-06, "loss": 0.1885, "step": 18552 }, { "epoch": 5.71, "learning_rate": 8.210439060425144e-06, "loss": 0.104, "step": 18553 }, { "epoch": 5.71, "learning_rate": 8.209458616302472e-06, "loss": 0.138, "step": 18554 }, { "epoch": 5.71, "learning_rate": 8.20847818996152e-06, "loss": 0.2151, "step": 18555 }, { "epoch": 5.71, "learning_rate": 8.207497781412033e-06, "loss": 0.0915, "step": 18556 }, { "epoch": 5.71, "learning_rate": 8.206517390663743e-06, "loss": 0.153, "step": 18557 }, { "epoch": 5.71, "learning_rate": 8.205537017726391e-06, "loss": 0.1425, "step": 18558 }, { "epoch": 5.71, "learning_rate": 8.204556662609713e-06, "loss": 0.0989, "step": 18559 }, { "epoch": 5.71, "learning_rate": 8.203576325323435e-06, "loss": 0.2309, "step": 18560 }, { "epoch": 5.71, "learning_rate": 8.202596005877307e-06, "loss": 0.1724, "step": 18561 }, { "epoch": 5.71, "learning_rate": 8.20161570428105e-06, "loss": 0.1597, "step": 18562 }, { "epoch": 5.71, "learning_rate": 8.200635420544411e-06, "loss": 0.1968, "step": 18563 }, { "epoch": 5.71, "learning_rate": 8.199655154677123e-06, "loss": 0.1415, "step": 18564 }, { "epoch": 5.71, "learning_rate": 8.19867490668891e-06, "loss": 0.1727, "step": 18565 }, { "epoch": 5.71, "learning_rate": 8.197694676589523e-06, "loss": 0.1553, "step": 18566 }, { "epoch": 5.71, "learning_rate": 8.196714464388687e-06, "loss": 0.1512, "step": 18567 }, { "epoch": 5.71, "learning_rate": 8.195734270096137e-06, "loss": 0.1999, "step": 18568 }, { "epoch": 5.71, "learning_rate": 8.19475409372161e-06, "loss": 0.1424, "step": 18569 }, { "epoch": 5.71, "learning_rate": 8.193773935274836e-06, "loss": 0.2944, "step": 18570 }, { "epoch": 5.71, "learning_rate": 8.192793794765556e-06, "loss": 0.184, "step": 18571 }, { "epoch": 5.71, "learning_rate": 8.191813672203493e-06, "loss": 0.1796, "step": 18572 }, { "epoch": 5.71, "learning_rate": 8.190833567598392e-06, "loss": 0.2182, "step": 18573 }, { "epoch": 5.71, "learning_rate": 8.189853480959982e-06, "loss": 0.195, "step": 18574 }, { "epoch": 5.72, "learning_rate": 8.18887341229799e-06, "loss": 0.0984, "step": 18575 }, { "epoch": 5.72, "learning_rate": 8.18789336162216e-06, "loss": 0.1213, "step": 18576 }, { "epoch": 5.72, "learning_rate": 8.186913328942213e-06, "loss": 0.2133, "step": 18577 }, { "epoch": 5.72, "learning_rate": 8.185933314267891e-06, "loss": 0.1604, "step": 18578 }, { "epoch": 5.72, "learning_rate": 8.184953317608924e-06, "loss": 0.1714, "step": 18579 }, { "epoch": 5.72, "learning_rate": 8.18397333897504e-06, "loss": 0.1513, "step": 18580 }, { "epoch": 5.72, "learning_rate": 8.182993378375976e-06, "loss": 0.1388, "step": 18581 }, { "epoch": 5.72, "learning_rate": 8.182013435821461e-06, "loss": 0.1085, "step": 18582 }, { "epoch": 5.72, "learning_rate": 8.181033511321232e-06, "loss": 0.2412, "step": 18583 }, { "epoch": 5.72, "learning_rate": 8.180053604885016e-06, "loss": 0.1544, "step": 18584 }, { "epoch": 5.72, "learning_rate": 8.179073716522542e-06, "loss": 0.1841, "step": 18585 }, { "epoch": 5.72, "learning_rate": 8.178093846243546e-06, "loss": 0.1086, "step": 18586 }, { "epoch": 5.72, "learning_rate": 8.177113994057756e-06, "loss": 0.1585, "step": 18587 }, { "epoch": 5.72, "learning_rate": 8.176134159974907e-06, "loss": 0.0884, "step": 18588 }, { "epoch": 5.72, "learning_rate": 8.175154344004724e-06, "loss": 0.1324, "step": 18589 }, { "epoch": 5.72, "learning_rate": 8.17417454615694e-06, "loss": 0.1739, "step": 18590 }, { "epoch": 5.72, "learning_rate": 8.173194766441286e-06, "loss": 0.1814, "step": 18591 }, { "epoch": 5.72, "learning_rate": 8.17221500486749e-06, "loss": 0.2009, "step": 18592 }, { "epoch": 5.72, "learning_rate": 8.171235261445289e-06, "loss": 0.1452, "step": 18593 }, { "epoch": 5.72, "learning_rate": 8.170255536184404e-06, "loss": 0.1081, "step": 18594 }, { "epoch": 5.72, "learning_rate": 8.169275829094566e-06, "loss": 0.0783, "step": 18595 }, { "epoch": 5.72, "learning_rate": 8.168296140185508e-06, "loss": 0.213, "step": 18596 }, { "epoch": 5.72, "learning_rate": 8.167316469466957e-06, "loss": 0.2154, "step": 18597 }, { "epoch": 5.72, "learning_rate": 8.166336816948643e-06, "loss": 0.2025, "step": 18598 }, { "epoch": 5.72, "learning_rate": 8.165357182640297e-06, "loss": 0.1218, "step": 18599 }, { "epoch": 5.72, "learning_rate": 8.16437756655164e-06, "loss": 0.1573, "step": 18600 }, { "epoch": 5.72, "learning_rate": 8.163397968692409e-06, "loss": 0.2636, "step": 18601 }, { "epoch": 5.72, "learning_rate": 8.162418389072325e-06, "loss": 0.085, "step": 18602 }, { "epoch": 5.72, "learning_rate": 8.161438827701123e-06, "loss": 0.1394, "step": 18603 }, { "epoch": 5.72, "learning_rate": 8.16045928458853e-06, "loss": 0.2102, "step": 18604 }, { "epoch": 5.72, "learning_rate": 8.159479759744267e-06, "loss": 0.1268, "step": 18605 }, { "epoch": 5.72, "learning_rate": 8.15850025317807e-06, "loss": 0.1828, "step": 18606 }, { "epoch": 5.73, "learning_rate": 8.15752076489966e-06, "loss": 0.132, "step": 18607 }, { "epoch": 5.73, "learning_rate": 8.156541294918767e-06, "loss": 0.1369, "step": 18608 }, { "epoch": 5.73, "learning_rate": 8.155561843245123e-06, "loss": 0.1274, "step": 18609 }, { "epoch": 5.73, "learning_rate": 8.154582409888441e-06, "loss": 0.2115, "step": 18610 }, { "epoch": 5.73, "learning_rate": 8.153602994858463e-06, "loss": 0.1888, "step": 18611 }, { "epoch": 5.73, "learning_rate": 8.152623598164906e-06, "loss": 0.2774, "step": 18612 }, { "epoch": 5.73, "learning_rate": 8.1516442198175e-06, "loss": 0.1932, "step": 18613 }, { "epoch": 5.73, "learning_rate": 8.150664859825973e-06, "loss": 0.1578, "step": 18614 }, { "epoch": 5.73, "learning_rate": 8.149685518200044e-06, "loss": 0.1362, "step": 18615 }, { "epoch": 5.73, "learning_rate": 8.148706194949448e-06, "loss": 0.1252, "step": 18616 }, { "epoch": 5.73, "learning_rate": 8.1477268900839e-06, "loss": 0.1301, "step": 18617 }, { "epoch": 5.73, "learning_rate": 8.146747603613135e-06, "loss": 0.2256, "step": 18618 }, { "epoch": 5.73, "learning_rate": 8.145768335546874e-06, "loss": 0.119, "step": 18619 }, { "epoch": 5.73, "learning_rate": 8.14478908589484e-06, "loss": 0.1638, "step": 18620 }, { "epoch": 5.73, "learning_rate": 8.143809854666765e-06, "loss": 0.2896, "step": 18621 }, { "epoch": 5.73, "learning_rate": 8.142830641872362e-06, "loss": 0.1632, "step": 18622 }, { "epoch": 5.73, "learning_rate": 8.141851447521367e-06, "loss": 0.1735, "step": 18623 }, { "epoch": 5.73, "learning_rate": 8.140872271623499e-06, "loss": 0.1919, "step": 18624 }, { "epoch": 5.73, "learning_rate": 8.139893114188482e-06, "loss": 0.2616, "step": 18625 }, { "epoch": 5.73, "learning_rate": 8.138913975226044e-06, "loss": 0.21, "step": 18626 }, { "epoch": 5.73, "learning_rate": 8.137934854745897e-06, "loss": 0.1099, "step": 18627 }, { "epoch": 5.73, "learning_rate": 8.136955752757783e-06, "loss": 0.2306, "step": 18628 }, { "epoch": 5.73, "learning_rate": 8.135976669271408e-06, "loss": 0.1943, "step": 18629 }, { "epoch": 5.73, "learning_rate": 8.134997604296506e-06, "loss": 0.1939, "step": 18630 }, { "epoch": 5.73, "learning_rate": 8.134018557842798e-06, "loss": 0.1535, "step": 18631 }, { "epoch": 5.73, "learning_rate": 8.133039529920002e-06, "loss": 0.112, "step": 18632 }, { "epoch": 5.73, "learning_rate": 8.132060520537849e-06, "loss": 0.1107, "step": 18633 }, { "epoch": 5.73, "learning_rate": 8.13108152970605e-06, "loss": 0.1287, "step": 18634 }, { "epoch": 5.73, "learning_rate": 8.130102557434338e-06, "loss": 0.1758, "step": 18635 }, { "epoch": 5.73, "learning_rate": 8.129123603732431e-06, "loss": 0.1972, "step": 18636 }, { "epoch": 5.73, "learning_rate": 8.12814466861005e-06, "loss": 0.1687, "step": 18637 }, { "epoch": 5.73, "learning_rate": 8.127165752076918e-06, "loss": 0.1142, "step": 18638 }, { "epoch": 5.73, "learning_rate": 8.126186854142752e-06, "loss": 0.2349, "step": 18639 }, { "epoch": 5.74, "learning_rate": 8.125207974817284e-06, "loss": 0.1971, "step": 18640 }, { "epoch": 5.74, "learning_rate": 8.124229114110227e-06, "loss": 0.2552, "step": 18641 }, { "epoch": 5.74, "learning_rate": 8.1232502720313e-06, "loss": 0.2229, "step": 18642 }, { "epoch": 5.74, "learning_rate": 8.122271448590234e-06, "loss": 0.1292, "step": 18643 }, { "epoch": 5.74, "learning_rate": 8.121292643796734e-06, "loss": 0.2019, "step": 18644 }, { "epoch": 5.74, "learning_rate": 8.120313857660537e-06, "loss": 0.1975, "step": 18645 }, { "epoch": 5.74, "learning_rate": 8.119335090191354e-06, "loss": 0.1541, "step": 18646 }, { "epoch": 5.74, "learning_rate": 8.118356341398906e-06, "loss": 0.2328, "step": 18647 }, { "epoch": 5.74, "learning_rate": 8.117377611292914e-06, "loss": 0.1618, "step": 18648 }, { "epoch": 5.74, "learning_rate": 8.116398899883096e-06, "loss": 0.1589, "step": 18649 }, { "epoch": 5.74, "learning_rate": 8.115420207179177e-06, "loss": 0.1112, "step": 18650 }, { "epoch": 5.74, "learning_rate": 8.114441533190869e-06, "loss": 0.0585, "step": 18651 }, { "epoch": 5.74, "learning_rate": 8.113462877927893e-06, "loss": 0.1319, "step": 18652 }, { "epoch": 5.74, "learning_rate": 8.112484241399972e-06, "loss": 0.0757, "step": 18653 }, { "epoch": 5.74, "learning_rate": 8.11150562361682e-06, "loss": 0.1883, "step": 18654 }, { "epoch": 5.74, "learning_rate": 8.110527024588162e-06, "loss": 0.2181, "step": 18655 }, { "epoch": 5.74, "learning_rate": 8.109548444323708e-06, "loss": 0.268, "step": 18656 }, { "epoch": 5.74, "learning_rate": 8.10856988283318e-06, "loss": 0.1122, "step": 18657 }, { "epoch": 5.74, "learning_rate": 8.107591340126297e-06, "loss": 0.1815, "step": 18658 }, { "epoch": 5.74, "learning_rate": 8.106612816212775e-06, "loss": 0.1351, "step": 18659 }, { "epoch": 5.74, "learning_rate": 8.105634311102336e-06, "loss": 0.17, "step": 18660 }, { "epoch": 5.74, "learning_rate": 8.104655824804694e-06, "loss": 0.1585, "step": 18661 }, { "epoch": 5.74, "learning_rate": 8.103677357329564e-06, "loss": 0.2044, "step": 18662 }, { "epoch": 5.74, "learning_rate": 8.102698908686665e-06, "loss": 0.1159, "step": 18663 }, { "epoch": 5.74, "learning_rate": 8.101720478885715e-06, "loss": 0.288, "step": 18664 }, { "epoch": 5.74, "learning_rate": 8.100742067936432e-06, "loss": 0.1615, "step": 18665 }, { "epoch": 5.74, "learning_rate": 8.099763675848533e-06, "loss": 0.2184, "step": 18666 }, { "epoch": 5.74, "learning_rate": 8.098785302631726e-06, "loss": 0.148, "step": 18667 }, { "epoch": 5.74, "learning_rate": 8.097806948295736e-06, "loss": 0.2493, "step": 18668 }, { "epoch": 5.74, "learning_rate": 8.096828612850276e-06, "loss": 0.2288, "step": 18669 }, { "epoch": 5.74, "learning_rate": 8.09585029630506e-06, "loss": 0.1532, "step": 18670 }, { "epoch": 5.74, "learning_rate": 8.09487199866981e-06, "loss": 0.1791, "step": 18671 }, { "epoch": 5.75, "learning_rate": 8.093893719954234e-06, "loss": 0.1565, "step": 18672 }, { "epoch": 5.75, "learning_rate": 8.092915460168051e-06, "loss": 0.1556, "step": 18673 }, { "epoch": 5.75, "learning_rate": 8.091937219320973e-06, "loss": 0.1401, "step": 18674 }, { "epoch": 5.75, "learning_rate": 8.090958997422718e-06, "loss": 0.1848, "step": 18675 }, { "epoch": 5.75, "learning_rate": 8.089980794483002e-06, "loss": 0.1591, "step": 18676 }, { "epoch": 5.75, "learning_rate": 8.089002610511532e-06, "loss": 0.1558, "step": 18677 }, { "epoch": 5.75, "learning_rate": 8.088024445518033e-06, "loss": 0.0901, "step": 18678 }, { "epoch": 5.75, "learning_rate": 8.087046299512209e-06, "loss": 0.2162, "step": 18679 }, { "epoch": 5.75, "learning_rate": 8.086068172503778e-06, "loss": 0.2307, "step": 18680 }, { "epoch": 5.75, "learning_rate": 8.085090064502458e-06, "loss": 0.1868, "step": 18681 }, { "epoch": 5.75, "learning_rate": 8.084111975517955e-06, "loss": 0.1968, "step": 18682 }, { "epoch": 5.75, "learning_rate": 8.083133905559989e-06, "loss": 0.1761, "step": 18683 }, { "epoch": 5.75, "learning_rate": 8.082155854638267e-06, "loss": 0.1767, "step": 18684 }, { "epoch": 5.75, "learning_rate": 8.081177822762506e-06, "loss": 0.2208, "step": 18685 }, { "epoch": 5.75, "learning_rate": 8.080199809942418e-06, "loss": 0.1902, "step": 18686 }, { "epoch": 5.75, "learning_rate": 8.079221816187714e-06, "loss": 0.2469, "step": 18687 }, { "epoch": 5.75, "learning_rate": 8.078243841508111e-06, "loss": 0.193, "step": 18688 }, { "epoch": 5.75, "learning_rate": 8.077265885913312e-06, "loss": 0.1622, "step": 18689 }, { "epoch": 5.75, "learning_rate": 8.076287949413039e-06, "loss": 0.2676, "step": 18690 }, { "epoch": 5.75, "learning_rate": 8.075310032017e-06, "loss": 0.2121, "step": 18691 }, { "epoch": 5.75, "learning_rate": 8.074332133734904e-06, "loss": 0.1864, "step": 18692 }, { "epoch": 5.75, "learning_rate": 8.073354254576468e-06, "loss": 0.1972, "step": 18693 }, { "epoch": 5.75, "learning_rate": 8.072376394551393e-06, "loss": 0.1235, "step": 18694 }, { "epoch": 5.75, "learning_rate": 8.071398553669405e-06, "loss": 0.1552, "step": 18695 }, { "epoch": 5.75, "learning_rate": 8.070420731940202e-06, "loss": 0.2416, "step": 18696 }, { "epoch": 5.75, "learning_rate": 8.0694429293735e-06, "loss": 0.1114, "step": 18697 }, { "epoch": 5.75, "learning_rate": 8.06846514597901e-06, "loss": 0.2275, "step": 18698 }, { "epoch": 5.75, "learning_rate": 8.06748738176644e-06, "loss": 0.1709, "step": 18699 }, { "epoch": 5.75, "learning_rate": 8.066509636745505e-06, "loss": 0.1268, "step": 18700 }, { "epoch": 5.75, "learning_rate": 8.065531910925907e-06, "loss": 0.3071, "step": 18701 }, { "epoch": 5.75, "learning_rate": 8.06455420431736e-06, "loss": 0.162, "step": 18702 }, { "epoch": 5.75, "learning_rate": 8.063576516929577e-06, "loss": 0.1981, "step": 18703 }, { "epoch": 5.75, "learning_rate": 8.062598848772261e-06, "loss": 0.1531, "step": 18704 }, { "epoch": 5.76, "learning_rate": 8.061621199855127e-06, "loss": 0.2275, "step": 18705 }, { "epoch": 5.76, "learning_rate": 8.060643570187879e-06, "loss": 0.167, "step": 18706 }, { "epoch": 5.76, "learning_rate": 8.059665959780225e-06, "loss": 0.2137, "step": 18707 }, { "epoch": 5.76, "learning_rate": 8.05868836864188e-06, "loss": 0.1741, "step": 18708 }, { "epoch": 5.76, "learning_rate": 8.057710796782546e-06, "loss": 0.1028, "step": 18709 }, { "epoch": 5.76, "learning_rate": 8.056733244211935e-06, "loss": 0.0886, "step": 18710 }, { "epoch": 5.76, "learning_rate": 8.055755710939757e-06, "loss": 0.1591, "step": 18711 }, { "epoch": 5.76, "learning_rate": 8.054778196975714e-06, "loss": 0.182, "step": 18712 }, { "epoch": 5.76, "learning_rate": 8.053800702329515e-06, "loss": 0.1426, "step": 18713 }, { "epoch": 5.76, "learning_rate": 8.052823227010869e-06, "loss": 0.1581, "step": 18714 }, { "epoch": 5.76, "learning_rate": 8.051845771029483e-06, "loss": 0.2215, "step": 18715 }, { "epoch": 5.76, "learning_rate": 8.050868334395067e-06, "loss": 0.1468, "step": 18716 }, { "epoch": 5.76, "learning_rate": 8.049890917117322e-06, "loss": 0.1713, "step": 18717 }, { "epoch": 5.76, "learning_rate": 8.04891351920596e-06, "loss": 0.1528, "step": 18718 }, { "epoch": 5.76, "learning_rate": 8.04793614067068e-06, "loss": 0.1602, "step": 18719 }, { "epoch": 5.76, "learning_rate": 8.046958781521198e-06, "loss": 0.1944, "step": 18720 }, { "epoch": 5.76, "learning_rate": 8.045981441767215e-06, "loss": 0.1372, "step": 18721 }, { "epoch": 5.76, "learning_rate": 8.045004121418434e-06, "loss": 0.1761, "step": 18722 }, { "epoch": 5.76, "learning_rate": 8.044026820484565e-06, "loss": 0.1725, "step": 18723 }, { "epoch": 5.76, "learning_rate": 8.04304953897531e-06, "loss": 0.134, "step": 18724 }, { "epoch": 5.76, "learning_rate": 8.042072276900382e-06, "loss": 0.2628, "step": 18725 }, { "epoch": 5.76, "learning_rate": 8.04109503426948e-06, "loss": 0.1412, "step": 18726 }, { "epoch": 5.76, "learning_rate": 8.040117811092305e-06, "loss": 0.1423, "step": 18727 }, { "epoch": 5.76, "learning_rate": 8.039140607378571e-06, "loss": 0.1665, "step": 18728 }, { "epoch": 5.76, "learning_rate": 8.038163423137974e-06, "loss": 0.1673, "step": 18729 }, { "epoch": 5.76, "learning_rate": 8.037186258380226e-06, "loss": 0.1628, "step": 18730 }, { "epoch": 5.76, "learning_rate": 8.036209113115027e-06, "loss": 0.3049, "step": 18731 }, { "epoch": 5.76, "learning_rate": 8.035231987352078e-06, "loss": 0.1451, "step": 18732 }, { "epoch": 5.76, "learning_rate": 8.034254881101091e-06, "loss": 0.1367, "step": 18733 }, { "epoch": 5.76, "learning_rate": 8.03327779437176e-06, "loss": 0.1414, "step": 18734 }, { "epoch": 5.76, "learning_rate": 8.032300727173797e-06, "loss": 0.1361, "step": 18735 }, { "epoch": 5.76, "learning_rate": 8.0313236795169e-06, "loss": 0.1762, "step": 18736 }, { "epoch": 5.77, "learning_rate": 8.03034665141077e-06, "loss": 0.1472, "step": 18737 }, { "epoch": 5.77, "learning_rate": 8.029369642865118e-06, "loss": 0.2551, "step": 18738 }, { "epoch": 5.77, "learning_rate": 8.028392653889637e-06, "loss": 0.0743, "step": 18739 }, { "epoch": 5.77, "learning_rate": 8.02741568449404e-06, "loss": 0.2534, "step": 18740 }, { "epoch": 5.77, "learning_rate": 8.02643873468802e-06, "loss": 0.1544, "step": 18741 }, { "epoch": 5.77, "learning_rate": 8.025461804481281e-06, "loss": 0.227, "step": 18742 }, { "epoch": 5.77, "learning_rate": 8.02448489388353e-06, "loss": 0.1212, "step": 18743 }, { "epoch": 5.77, "learning_rate": 8.023508002904456e-06, "loss": 0.1819, "step": 18744 }, { "epoch": 5.77, "learning_rate": 8.02253113155378e-06, "loss": 0.1755, "step": 18745 }, { "epoch": 5.77, "learning_rate": 8.021554279841185e-06, "loss": 0.1587, "step": 18746 }, { "epoch": 5.77, "learning_rate": 8.020577447776381e-06, "loss": 0.2505, "step": 18747 }, { "epoch": 5.77, "learning_rate": 8.019600635369068e-06, "loss": 0.1292, "step": 18748 }, { "epoch": 5.77, "learning_rate": 8.018623842628946e-06, "loss": 0.1645, "step": 18749 }, { "epoch": 5.77, "learning_rate": 8.017647069565718e-06, "loss": 0.2088, "step": 18750 }, { "epoch": 5.77, "learning_rate": 8.016670316189078e-06, "loss": 0.1468, "step": 18751 }, { "epoch": 5.77, "learning_rate": 8.015693582508728e-06, "loss": 0.1356, "step": 18752 }, { "epoch": 5.77, "learning_rate": 8.014716868534372e-06, "loss": 0.0911, "step": 18753 }, { "epoch": 5.77, "learning_rate": 8.013740174275706e-06, "loss": 0.1839, "step": 18754 }, { "epoch": 5.77, "learning_rate": 8.012763499742435e-06, "loss": 0.1676, "step": 18755 }, { "epoch": 5.77, "learning_rate": 8.01178684494425e-06, "loss": 0.1652, "step": 18756 }, { "epoch": 5.77, "learning_rate": 8.010810209890853e-06, "loss": 0.1693, "step": 18757 }, { "epoch": 5.77, "learning_rate": 8.009833594591948e-06, "loss": 0.1568, "step": 18758 }, { "epoch": 5.77, "learning_rate": 8.008856999057225e-06, "loss": 0.1707, "step": 18759 }, { "epoch": 5.77, "learning_rate": 8.00788042329639e-06, "loss": 0.2124, "step": 18760 }, { "epoch": 5.77, "learning_rate": 8.006903867319138e-06, "loss": 0.1564, "step": 18761 }, { "epoch": 5.77, "learning_rate": 8.005927331135167e-06, "loss": 0.0978, "step": 18762 }, { "epoch": 5.77, "learning_rate": 8.004950814754175e-06, "loss": 0.1391, "step": 18763 }, { "epoch": 5.77, "learning_rate": 8.00397431818586e-06, "loss": 0.1138, "step": 18764 }, { "epoch": 5.77, "learning_rate": 8.00299784143992e-06, "loss": 0.107, "step": 18765 }, { "epoch": 5.77, "learning_rate": 8.002021384526056e-06, "loss": 0.1182, "step": 18766 }, { "epoch": 5.77, "learning_rate": 8.001044947453955e-06, "loss": 0.1306, "step": 18767 }, { "epoch": 5.77, "learning_rate": 8.000068530233322e-06, "loss": 0.1463, "step": 18768 }, { "epoch": 5.77, "learning_rate": 7.999092132873851e-06, "loss": 0.1937, "step": 18769 }, { "epoch": 5.78, "learning_rate": 7.998115755385241e-06, "loss": 0.1698, "step": 18770 }, { "epoch": 5.78, "learning_rate": 7.997139397777187e-06, "loss": 0.164, "step": 18771 }, { "epoch": 5.78, "learning_rate": 7.996163060059381e-06, "loss": 0.1902, "step": 18772 }, { "epoch": 5.78, "learning_rate": 7.995186742241527e-06, "loss": 0.0994, "step": 18773 }, { "epoch": 5.78, "learning_rate": 7.994210444333312e-06, "loss": 0.221, "step": 18774 }, { "epoch": 5.78, "learning_rate": 7.99323416634444e-06, "loss": 0.1957, "step": 18775 }, { "epoch": 5.78, "learning_rate": 7.992257908284602e-06, "loss": 0.1114, "step": 18776 }, { "epoch": 5.78, "learning_rate": 7.991281670163488e-06, "loss": 0.3038, "step": 18777 }, { "epoch": 5.78, "learning_rate": 7.990305451990805e-06, "loss": 0.1769, "step": 18778 }, { "epoch": 5.78, "learning_rate": 7.989329253776235e-06, "loss": 0.1112, "step": 18779 }, { "epoch": 5.78, "learning_rate": 7.988353075529482e-06, "loss": 0.2054, "step": 18780 }, { "epoch": 5.78, "learning_rate": 7.987376917260237e-06, "loss": 0.1497, "step": 18781 }, { "epoch": 5.78, "learning_rate": 7.986400778978192e-06, "loss": 0.1112, "step": 18782 }, { "epoch": 5.78, "learning_rate": 7.985424660693048e-06, "loss": 0.2173, "step": 18783 }, { "epoch": 5.78, "learning_rate": 7.984448562414489e-06, "loss": 0.1604, "step": 18784 }, { "epoch": 5.78, "learning_rate": 7.983472484152216e-06, "loss": 0.1641, "step": 18785 }, { "epoch": 5.78, "learning_rate": 7.982496425915919e-06, "loss": 0.1353, "step": 18786 }, { "epoch": 5.78, "learning_rate": 7.981520387715289e-06, "loss": 0.1037, "step": 18787 }, { "epoch": 5.78, "learning_rate": 7.980544369560029e-06, "loss": 0.0967, "step": 18788 }, { "epoch": 5.78, "learning_rate": 7.979568371459816e-06, "loss": 0.147, "step": 18789 }, { "epoch": 5.78, "learning_rate": 7.978592393424356e-06, "loss": 0.1921, "step": 18790 }, { "epoch": 5.78, "learning_rate": 7.977616435463336e-06, "loss": 0.2794, "step": 18791 }, { "epoch": 5.78, "learning_rate": 7.976640497586447e-06, "loss": 0.2353, "step": 18792 }, { "epoch": 5.78, "learning_rate": 7.975664579803387e-06, "loss": 0.1369, "step": 18793 }, { "epoch": 5.78, "learning_rate": 7.974688682123837e-06, "loss": 0.2127, "step": 18794 }, { "epoch": 5.78, "learning_rate": 7.9737128045575e-06, "loss": 0.1662, "step": 18795 }, { "epoch": 5.78, "learning_rate": 7.972736947114062e-06, "loss": 0.1298, "step": 18796 }, { "epoch": 5.78, "learning_rate": 7.97176110980321e-06, "loss": 0.1322, "step": 18797 }, { "epoch": 5.78, "learning_rate": 7.970785292634644e-06, "loss": 0.0568, "step": 18798 }, { "epoch": 5.78, "learning_rate": 7.969809495618047e-06, "loss": 0.1924, "step": 18799 }, { "epoch": 5.78, "learning_rate": 7.968833718763117e-06, "loss": 0.1653, "step": 18800 }, { "epoch": 5.78, "learning_rate": 7.967857962079538e-06, "loss": 0.1129, "step": 18801 }, { "epoch": 5.79, "learning_rate": 7.966882225577e-06, "loss": 0.2, "step": 18802 }, { "epoch": 5.79, "learning_rate": 7.9659065092652e-06, "loss": 0.1523, "step": 18803 }, { "epoch": 5.79, "learning_rate": 7.964930813153817e-06, "loss": 0.1571, "step": 18804 }, { "epoch": 5.79, "learning_rate": 7.963955137252552e-06, "loss": 0.166, "step": 18805 }, { "epoch": 5.79, "learning_rate": 7.96297948157109e-06, "loss": 0.1518, "step": 18806 }, { "epoch": 5.79, "learning_rate": 7.962003846119113e-06, "loss": 0.2592, "step": 18807 }, { "epoch": 5.79, "learning_rate": 7.96102823090632e-06, "loss": 0.1352, "step": 18808 }, { "epoch": 5.79, "learning_rate": 7.960052635942393e-06, "loss": 0.1502, "step": 18809 }, { "epoch": 5.79, "learning_rate": 7.959077061237026e-06, "loss": 0.1237, "step": 18810 }, { "epoch": 5.79, "learning_rate": 7.958101506799906e-06, "loss": 0.1275, "step": 18811 }, { "epoch": 5.79, "learning_rate": 7.957125972640717e-06, "loss": 0.2126, "step": 18812 }, { "epoch": 5.79, "learning_rate": 7.956150458769151e-06, "loss": 0.1637, "step": 18813 }, { "epoch": 5.79, "learning_rate": 7.955174965194892e-06, "loss": 0.2175, "step": 18814 }, { "epoch": 5.79, "learning_rate": 7.954199491927635e-06, "loss": 0.1142, "step": 18815 }, { "epoch": 5.79, "learning_rate": 7.953224038977062e-06, "loss": 0.1612, "step": 18816 }, { "epoch": 5.79, "learning_rate": 7.952248606352856e-06, "loss": 0.1943, "step": 18817 }, { "epoch": 5.79, "learning_rate": 7.951273194064714e-06, "loss": 0.1051, "step": 18818 }, { "epoch": 5.79, "learning_rate": 7.950297802122313e-06, "loss": 0.2668, "step": 18819 }, { "epoch": 5.79, "learning_rate": 7.949322430535345e-06, "loss": 0.2111, "step": 18820 }, { "epoch": 5.79, "learning_rate": 7.948347079313494e-06, "loss": 0.1595, "step": 18821 }, { "epoch": 5.79, "learning_rate": 7.947371748466453e-06, "loss": 0.1297, "step": 18822 }, { "epoch": 5.79, "learning_rate": 7.946396438003898e-06, "loss": 0.1164, "step": 18823 }, { "epoch": 5.79, "learning_rate": 7.94542114793552e-06, "loss": 0.1739, "step": 18824 }, { "epoch": 5.79, "learning_rate": 7.944445878271002e-06, "loss": 0.1404, "step": 18825 }, { "epoch": 5.79, "learning_rate": 7.943470629020032e-06, "loss": 0.1798, "step": 18826 }, { "epoch": 5.79, "learning_rate": 7.942495400192293e-06, "loss": 0.128, "step": 18827 }, { "epoch": 5.79, "learning_rate": 7.941520191797475e-06, "loss": 0.2385, "step": 18828 }, { "epoch": 5.79, "learning_rate": 7.940545003845253e-06, "loss": 0.2313, "step": 18829 }, { "epoch": 5.79, "learning_rate": 7.93956983634532e-06, "loss": 0.1368, "step": 18830 }, { "epoch": 5.79, "learning_rate": 7.938594689307358e-06, "loss": 0.1425, "step": 18831 }, { "epoch": 5.79, "learning_rate": 7.937619562741048e-06, "loss": 0.1782, "step": 18832 }, { "epoch": 5.79, "learning_rate": 7.936644456656082e-06, "loss": 0.2746, "step": 18833 }, { "epoch": 5.79, "learning_rate": 7.935669371062132e-06, "loss": 0.133, "step": 18834 }, { "epoch": 5.8, "learning_rate": 7.934694305968893e-06, "loss": 0.1094, "step": 18835 }, { "epoch": 5.8, "learning_rate": 7.933719261386036e-06, "loss": 0.1773, "step": 18836 }, { "epoch": 5.8, "learning_rate": 7.932744237323256e-06, "loss": 0.1534, "step": 18837 }, { "epoch": 5.8, "learning_rate": 7.931769233790234e-06, "loss": 0.216, "step": 18838 }, { "epoch": 5.8, "learning_rate": 7.930794250796643e-06, "loss": 0.1525, "step": 18839 }, { "epoch": 5.8, "learning_rate": 7.929819288352176e-06, "loss": 0.1473, "step": 18840 }, { "epoch": 5.8, "learning_rate": 7.92884434646651e-06, "loss": 0.1602, "step": 18841 }, { "epoch": 5.8, "learning_rate": 7.927869425149329e-06, "loss": 0.1782, "step": 18842 }, { "epoch": 5.8, "learning_rate": 7.926894524410316e-06, "loss": 0.1482, "step": 18843 }, { "epoch": 5.8, "learning_rate": 7.925919644259145e-06, "loss": 0.1565, "step": 18844 }, { "epoch": 5.8, "learning_rate": 7.92494478470551e-06, "loss": 0.1909, "step": 18845 }, { "epoch": 5.8, "learning_rate": 7.923969945759082e-06, "loss": 0.1419, "step": 18846 }, { "epoch": 5.8, "learning_rate": 7.922995127429547e-06, "loss": 0.183, "step": 18847 }, { "epoch": 5.8, "learning_rate": 7.922020329726587e-06, "loss": 0.184, "step": 18848 }, { "epoch": 5.8, "learning_rate": 7.921045552659876e-06, "loss": 0.2283, "step": 18849 }, { "epoch": 5.8, "learning_rate": 7.920070796239104e-06, "loss": 0.2305, "step": 18850 }, { "epoch": 5.8, "learning_rate": 7.91909606047394e-06, "loss": 0.1506, "step": 18851 }, { "epoch": 5.8, "learning_rate": 7.918121345374073e-06, "loss": 0.2067, "step": 18852 }, { "epoch": 5.8, "learning_rate": 7.91714665094918e-06, "loss": 0.203, "step": 18853 }, { "epoch": 5.8, "learning_rate": 7.916171977208938e-06, "loss": 0.186, "step": 18854 }, { "epoch": 5.8, "learning_rate": 7.915197324163033e-06, "loss": 0.1614, "step": 18855 }, { "epoch": 5.8, "learning_rate": 7.914222691821134e-06, "loss": 0.2666, "step": 18856 }, { "epoch": 5.8, "learning_rate": 7.913248080192931e-06, "loss": 0.1069, "step": 18857 }, { "epoch": 5.8, "learning_rate": 7.912273489288097e-06, "loss": 0.2233, "step": 18858 }, { "epoch": 5.8, "learning_rate": 7.91129891911631e-06, "loss": 0.1744, "step": 18859 }, { "epoch": 5.8, "learning_rate": 7.91032436968725e-06, "loss": 0.2459, "step": 18860 }, { "epoch": 5.8, "learning_rate": 7.909349841010594e-06, "loss": 0.1573, "step": 18861 }, { "epoch": 5.8, "learning_rate": 7.908375333096024e-06, "loss": 0.1829, "step": 18862 }, { "epoch": 5.8, "learning_rate": 7.907400845953214e-06, "loss": 0.191, "step": 18863 }, { "epoch": 5.8, "learning_rate": 7.90642637959184e-06, "loss": 0.1743, "step": 18864 }, { "epoch": 5.8, "learning_rate": 7.905451934021581e-06, "loss": 0.3239, "step": 18865 }, { "epoch": 5.8, "learning_rate": 7.904477509252116e-06, "loss": 0.2158, "step": 18866 }, { "epoch": 5.81, "learning_rate": 7.903503105293122e-06, "loss": 0.1621, "step": 18867 }, { "epoch": 5.81, "learning_rate": 7.902528722154274e-06, "loss": 0.1782, "step": 18868 }, { "epoch": 5.81, "learning_rate": 7.901554359845246e-06, "loss": 0.183, "step": 18869 }, { "epoch": 5.81, "learning_rate": 7.90058001837572e-06, "loss": 0.1277, "step": 18870 }, { "epoch": 5.81, "learning_rate": 7.899605697755367e-06, "loss": 0.2524, "step": 18871 }, { "epoch": 5.81, "learning_rate": 7.898631397993869e-06, "loss": 0.253, "step": 18872 }, { "epoch": 5.81, "learning_rate": 7.897657119100896e-06, "loss": 0.1229, "step": 18873 }, { "epoch": 5.81, "learning_rate": 7.896682861086124e-06, "loss": 0.2537, "step": 18874 }, { "epoch": 5.81, "learning_rate": 7.895708623959231e-06, "loss": 0.1794, "step": 18875 }, { "epoch": 5.81, "learning_rate": 7.894734407729887e-06, "loss": 0.1093, "step": 18876 }, { "epoch": 5.81, "learning_rate": 7.893760212407775e-06, "loss": 0.073, "step": 18877 }, { "epoch": 5.81, "learning_rate": 7.892786038002567e-06, "loss": 0.0405, "step": 18878 }, { "epoch": 5.81, "learning_rate": 7.891811884523933e-06, "loss": 0.2479, "step": 18879 }, { "epoch": 5.81, "learning_rate": 7.89083775198155e-06, "loss": 0.2284, "step": 18880 }, { "epoch": 5.81, "learning_rate": 7.88986364038509e-06, "loss": 0.1871, "step": 18881 }, { "epoch": 5.81, "learning_rate": 7.88888954974423e-06, "loss": 0.1672, "step": 18882 }, { "epoch": 5.81, "learning_rate": 7.887915480068646e-06, "loss": 0.1546, "step": 18883 }, { "epoch": 5.81, "learning_rate": 7.886941431368003e-06, "loss": 0.1284, "step": 18884 }, { "epoch": 5.81, "learning_rate": 7.885967403651982e-06, "loss": 0.1227, "step": 18885 }, { "epoch": 5.81, "learning_rate": 7.88499339693025e-06, "loss": 0.1946, "step": 18886 }, { "epoch": 5.81, "learning_rate": 7.884019411212486e-06, "loss": 0.3113, "step": 18887 }, { "epoch": 5.81, "learning_rate": 7.88304544650836e-06, "loss": 0.1995, "step": 18888 }, { "epoch": 5.81, "learning_rate": 7.882071502827539e-06, "loss": 0.2302, "step": 18889 }, { "epoch": 5.81, "learning_rate": 7.881097580179705e-06, "loss": 0.2178, "step": 18890 }, { "epoch": 5.81, "learning_rate": 7.88012367857452e-06, "loss": 0.1756, "step": 18891 }, { "epoch": 5.81, "learning_rate": 7.879149798021666e-06, "loss": 0.1921, "step": 18892 }, { "epoch": 5.81, "learning_rate": 7.878175938530808e-06, "loss": 0.0776, "step": 18893 }, { "epoch": 5.81, "learning_rate": 7.877202100111612e-06, "loss": 0.15, "step": 18894 }, { "epoch": 5.81, "learning_rate": 7.876228282773765e-06, "loss": 0.1647, "step": 18895 }, { "epoch": 5.81, "learning_rate": 7.875254486526922e-06, "loss": 0.1938, "step": 18896 }, { "epoch": 5.81, "learning_rate": 7.874280711380762e-06, "loss": 0.133, "step": 18897 }, { "epoch": 5.81, "learning_rate": 7.873306957344954e-06, "loss": 0.1041, "step": 18898 }, { "epoch": 5.81, "learning_rate": 7.872333224429166e-06, "loss": 0.264, "step": 18899 }, { "epoch": 5.82, "learning_rate": 7.871359512643076e-06, "loss": 0.1892, "step": 18900 }, { "epoch": 5.82, "learning_rate": 7.87038582199634e-06, "loss": 0.102, "step": 18901 }, { "epoch": 5.82, "learning_rate": 7.86941215249864e-06, "loss": 0.1685, "step": 18902 }, { "epoch": 5.82, "learning_rate": 7.868438504159641e-06, "loss": 0.1527, "step": 18903 }, { "epoch": 5.82, "learning_rate": 7.86746487698901e-06, "loss": 0.1424, "step": 18904 }, { "epoch": 5.82, "learning_rate": 7.866491270996421e-06, "loss": 0.1487, "step": 18905 }, { "epoch": 5.82, "learning_rate": 7.865517686191536e-06, "loss": 0.1968, "step": 18906 }, { "epoch": 5.82, "learning_rate": 7.864544122584032e-06, "loss": 0.1525, "step": 18907 }, { "epoch": 5.82, "learning_rate": 7.863570580183572e-06, "loss": 0.1473, "step": 18908 }, { "epoch": 5.82, "learning_rate": 7.86259705899982e-06, "loss": 0.1107, "step": 18909 }, { "epoch": 5.82, "learning_rate": 7.861623559042457e-06, "loss": 0.1349, "step": 18910 }, { "epoch": 5.82, "learning_rate": 7.860650080321135e-06, "loss": 0.1221, "step": 18911 }, { "epoch": 5.82, "learning_rate": 7.859676622845535e-06, "loss": 0.1131, "step": 18912 }, { "epoch": 5.82, "learning_rate": 7.858703186625319e-06, "loss": 0.1041, "step": 18913 }, { "epoch": 5.82, "learning_rate": 7.857729771670151e-06, "loss": 0.0769, "step": 18914 }, { "epoch": 5.82, "learning_rate": 7.856756377989702e-06, "loss": 0.1209, "step": 18915 }, { "epoch": 5.82, "learning_rate": 7.855783005593635e-06, "loss": 0.1429, "step": 18916 }, { "epoch": 5.82, "learning_rate": 7.854809654491622e-06, "loss": 0.2002, "step": 18917 }, { "epoch": 5.82, "learning_rate": 7.853836324693327e-06, "loss": 0.1608, "step": 18918 }, { "epoch": 5.82, "learning_rate": 7.852863016208412e-06, "loss": 0.1826, "step": 18919 }, { "epoch": 5.82, "learning_rate": 7.851889729046548e-06, "loss": 0.1585, "step": 18920 }, { "epoch": 5.82, "learning_rate": 7.850916463217396e-06, "loss": 0.1994, "step": 18921 }, { "epoch": 5.82, "learning_rate": 7.84994321873063e-06, "loss": 0.2122, "step": 18922 }, { "epoch": 5.82, "learning_rate": 7.848969995595905e-06, "loss": 0.17, "step": 18923 }, { "epoch": 5.82, "learning_rate": 7.84799679382289e-06, "loss": 0.223, "step": 18924 }, { "epoch": 5.82, "learning_rate": 7.847023613421251e-06, "loss": 0.1419, "step": 18925 }, { "epoch": 5.82, "learning_rate": 7.846050454400653e-06, "loss": 0.1802, "step": 18926 }, { "epoch": 5.82, "learning_rate": 7.845077316770757e-06, "loss": 0.2022, "step": 18927 }, { "epoch": 5.82, "learning_rate": 7.844104200541235e-06, "loss": 0.1712, "step": 18928 }, { "epoch": 5.82, "learning_rate": 7.843131105721739e-06, "loss": 0.147, "step": 18929 }, { "epoch": 5.82, "learning_rate": 7.84215803232194e-06, "loss": 0.0873, "step": 18930 }, { "epoch": 5.82, "learning_rate": 7.8411849803515e-06, "loss": 0.1705, "step": 18931 }, { "epoch": 5.83, "learning_rate": 7.840211949820085e-06, "loss": 0.2342, "step": 18932 }, { "epoch": 5.83, "learning_rate": 7.839238940737356e-06, "loss": 0.1795, "step": 18933 }, { "epoch": 5.83, "learning_rate": 7.838265953112973e-06, "loss": 0.1489, "step": 18934 }, { "epoch": 5.83, "learning_rate": 7.837292986956604e-06, "loss": 0.1206, "step": 18935 }, { "epoch": 5.83, "learning_rate": 7.836320042277906e-06, "loss": 0.212, "step": 18936 }, { "epoch": 5.83, "learning_rate": 7.835347119086547e-06, "loss": 0.1411, "step": 18937 }, { "epoch": 5.83, "learning_rate": 7.834374217392188e-06, "loss": 0.1926, "step": 18938 }, { "epoch": 5.83, "learning_rate": 7.833401337204483e-06, "loss": 0.2671, "step": 18939 }, { "epoch": 5.83, "learning_rate": 7.832428478533105e-06, "loss": 0.1369, "step": 18940 }, { "epoch": 5.83, "learning_rate": 7.831455641387705e-06, "loss": 0.2046, "step": 18941 }, { "epoch": 5.83, "learning_rate": 7.830482825777951e-06, "loss": 0.1086, "step": 18942 }, { "epoch": 5.83, "learning_rate": 7.829510031713506e-06, "loss": 0.2192, "step": 18943 }, { "epoch": 5.83, "learning_rate": 7.828537259204019e-06, "loss": 0.0873, "step": 18944 }, { "epoch": 5.83, "learning_rate": 7.827564508259165e-06, "loss": 0.1601, "step": 18945 }, { "epoch": 5.83, "learning_rate": 7.826591778888594e-06, "loss": 0.1392, "step": 18946 }, { "epoch": 5.83, "learning_rate": 7.82561907110197e-06, "loss": 0.1788, "step": 18947 }, { "epoch": 5.83, "learning_rate": 7.824646384908955e-06, "loss": 0.2712, "step": 18948 }, { "epoch": 5.83, "learning_rate": 7.823673720319202e-06, "loss": 0.2005, "step": 18949 }, { "epoch": 5.83, "learning_rate": 7.82270107734238e-06, "loss": 0.1452, "step": 18950 }, { "epoch": 5.83, "learning_rate": 7.82172845598814e-06, "loss": 0.1564, "step": 18951 }, { "epoch": 5.83, "learning_rate": 7.820755856266145e-06, "loss": 0.0393, "step": 18952 }, { "epoch": 5.83, "learning_rate": 7.819783278186053e-06, "loss": 0.3849, "step": 18953 }, { "epoch": 5.83, "learning_rate": 7.81881072175752e-06, "loss": 0.1192, "step": 18954 }, { "epoch": 5.83, "learning_rate": 7.81783818699021e-06, "loss": 0.2345, "step": 18955 }, { "epoch": 5.83, "learning_rate": 7.816865673893774e-06, "loss": 0.1484, "step": 18956 }, { "epoch": 5.83, "learning_rate": 7.815893182477878e-06, "loss": 0.1905, "step": 18957 }, { "epoch": 5.83, "learning_rate": 7.814920712752174e-06, "loss": 0.322, "step": 18958 }, { "epoch": 5.83, "learning_rate": 7.813948264726318e-06, "loss": 0.154, "step": 18959 }, { "epoch": 5.83, "learning_rate": 7.812975838409977e-06, "loss": 0.1699, "step": 18960 }, { "epoch": 5.83, "learning_rate": 7.812003433812795e-06, "loss": 0.0983, "step": 18961 }, { "epoch": 5.83, "learning_rate": 7.81103105094444e-06, "loss": 0.1925, "step": 18962 }, { "epoch": 5.83, "learning_rate": 7.810058689814564e-06, "loss": 0.1617, "step": 18963 }, { "epoch": 5.83, "learning_rate": 7.80908635043282e-06, "loss": 0.214, "step": 18964 }, { "epoch": 5.84, "learning_rate": 7.80811403280887e-06, "loss": 0.1369, "step": 18965 }, { "epoch": 5.84, "learning_rate": 7.807141736952367e-06, "loss": 0.2474, "step": 18966 }, { "epoch": 5.84, "learning_rate": 7.80616946287297e-06, "loss": 0.1289, "step": 18967 }, { "epoch": 5.84, "learning_rate": 7.80519721058033e-06, "loss": 0.137, "step": 18968 }, { "epoch": 5.84, "learning_rate": 7.804224980084103e-06, "loss": 0.0834, "step": 18969 }, { "epoch": 5.84, "learning_rate": 7.803252771393948e-06, "loss": 0.135, "step": 18970 }, { "epoch": 5.84, "learning_rate": 7.802280584519515e-06, "loss": 0.1122, "step": 18971 }, { "epoch": 5.84, "learning_rate": 7.801308419470466e-06, "loss": 0.2239, "step": 18972 }, { "epoch": 5.84, "learning_rate": 7.800336276256449e-06, "loss": 0.1599, "step": 18973 }, { "epoch": 5.84, "learning_rate": 7.799364154887117e-06, "loss": 0.2452, "step": 18974 }, { "epoch": 5.84, "learning_rate": 7.798392055372129e-06, "loss": 0.1471, "step": 18975 }, { "epoch": 5.84, "learning_rate": 7.797419977721135e-06, "loss": 0.19, "step": 18976 }, { "epoch": 5.84, "learning_rate": 7.796447921943793e-06, "loss": 0.1174, "step": 18977 }, { "epoch": 5.84, "learning_rate": 7.795475888049757e-06, "loss": 0.1474, "step": 18978 }, { "epoch": 5.84, "learning_rate": 7.794503876048672e-06, "loss": 0.0636, "step": 18979 }, { "epoch": 5.84, "learning_rate": 7.793531885950199e-06, "loss": 0.1625, "step": 18980 }, { "epoch": 5.84, "learning_rate": 7.792559917763985e-06, "loss": 0.1903, "step": 18981 }, { "epoch": 5.84, "learning_rate": 7.791587971499687e-06, "loss": 0.19, "step": 18982 }, { "epoch": 5.84, "learning_rate": 7.790616047166959e-06, "loss": 0.1778, "step": 18983 }, { "epoch": 5.84, "learning_rate": 7.789644144775445e-06, "loss": 0.2425, "step": 18984 }, { "epoch": 5.84, "learning_rate": 7.788672264334804e-06, "loss": 0.1526, "step": 18985 }, { "epoch": 5.84, "learning_rate": 7.787700405854683e-06, "loss": 0.161, "step": 18986 }, { "epoch": 5.84, "learning_rate": 7.78672856934474e-06, "loss": 0.2185, "step": 18987 }, { "epoch": 5.84, "learning_rate": 7.785756754814623e-06, "loss": 0.1661, "step": 18988 }, { "epoch": 5.84, "learning_rate": 7.784784962273977e-06, "loss": 0.1237, "step": 18989 }, { "epoch": 5.84, "learning_rate": 7.78381319173246e-06, "loss": 0.1822, "step": 18990 }, { "epoch": 5.84, "learning_rate": 7.782841443199718e-06, "loss": 0.2028, "step": 18991 }, { "epoch": 5.84, "learning_rate": 7.781869716685408e-06, "loss": 0.1847, "step": 18992 }, { "epoch": 5.84, "learning_rate": 7.780898012199177e-06, "loss": 0.3369, "step": 18993 }, { "epoch": 5.84, "learning_rate": 7.779926329750668e-06, "loss": 0.2307, "step": 18994 }, { "epoch": 5.84, "learning_rate": 7.778954669349544e-06, "loss": 0.159, "step": 18995 }, { "epoch": 5.84, "learning_rate": 7.77798303100544e-06, "loss": 0.1365, "step": 18996 }, { "epoch": 5.85, "learning_rate": 7.777011414728016e-06, "loss": 0.1367, "step": 18997 }, { "epoch": 5.85, "learning_rate": 7.776039820526918e-06, "loss": 0.1869, "step": 18998 }, { "epoch": 5.85, "learning_rate": 7.77506824841179e-06, "loss": 0.1082, "step": 18999 }, { "epoch": 5.85, "learning_rate": 7.774096698392292e-06, "loss": 0.1385, "step": 19000 }, { "epoch": 5.85, "learning_rate": 7.77312517047806e-06, "loss": 0.2335, "step": 19001 }, { "epoch": 5.85, "learning_rate": 7.772153664678749e-06, "loss": 0.2012, "step": 19002 }, { "epoch": 5.85, "learning_rate": 7.771182181004005e-06, "loss": 0.1504, "step": 19003 }, { "epoch": 5.85, "learning_rate": 7.770210719463475e-06, "loss": 0.1798, "step": 19004 }, { "epoch": 5.85, "learning_rate": 7.76923928006681e-06, "loss": 0.2259, "step": 19005 }, { "epoch": 5.85, "learning_rate": 7.76826786282365e-06, "loss": 0.1652, "step": 19006 }, { "epoch": 5.85, "learning_rate": 7.767296467743655e-06, "loss": 0.1925, "step": 19007 }, { "epoch": 5.85, "learning_rate": 7.76632509483646e-06, "loss": 0.122, "step": 19008 }, { "epoch": 5.85, "learning_rate": 7.765353744111712e-06, "loss": 0.1654, "step": 19009 }, { "epoch": 5.85, "learning_rate": 7.764382415579067e-06, "loss": 0.2274, "step": 19010 }, { "epoch": 5.85, "learning_rate": 7.763411109248157e-06, "loss": 0.2423, "step": 19011 }, { "epoch": 5.85, "learning_rate": 7.762439825128643e-06, "loss": 0.1977, "step": 19012 }, { "epoch": 5.85, "learning_rate": 7.761468563230163e-06, "loss": 0.1492, "step": 19013 }, { "epoch": 5.85, "learning_rate": 7.760497323562359e-06, "loss": 0.1009, "step": 19014 }, { "epoch": 5.85, "learning_rate": 7.759526106134885e-06, "loss": 0.1913, "step": 19015 }, { "epoch": 5.85, "learning_rate": 7.758554910957378e-06, "loss": 0.1702, "step": 19016 }, { "epoch": 5.85, "learning_rate": 7.75758373803949e-06, "loss": 0.0833, "step": 19017 }, { "epoch": 5.85, "learning_rate": 7.75661258739086e-06, "loss": 0.1644, "step": 19018 }, { "epoch": 5.85, "learning_rate": 7.755641459021135e-06, "loss": 0.2136, "step": 19019 }, { "epoch": 5.85, "learning_rate": 7.754670352939959e-06, "loss": 0.1659, "step": 19020 }, { "epoch": 5.85, "learning_rate": 7.753699269156974e-06, "loss": 0.1187, "step": 19021 }, { "epoch": 5.85, "learning_rate": 7.75272820768183e-06, "loss": 0.1686, "step": 19022 }, { "epoch": 5.85, "learning_rate": 7.751757168524159e-06, "loss": 0.2144, "step": 19023 }, { "epoch": 5.85, "learning_rate": 7.750786151693618e-06, "loss": 0.1656, "step": 19024 }, { "epoch": 5.85, "learning_rate": 7.749815157199843e-06, "loss": 0.2291, "step": 19025 }, { "epoch": 5.85, "learning_rate": 7.748844185052474e-06, "loss": 0.2026, "step": 19026 }, { "epoch": 5.85, "learning_rate": 7.747873235261157e-06, "loss": 0.255, "step": 19027 }, { "epoch": 5.85, "learning_rate": 7.746902307835535e-06, "loss": 0.2376, "step": 19028 }, { "epoch": 5.85, "learning_rate": 7.745931402785252e-06, "loss": 0.2172, "step": 19029 }, { "epoch": 5.86, "learning_rate": 7.744960520119946e-06, "loss": 0.1712, "step": 19030 }, { "epoch": 5.86, "learning_rate": 7.74398965984926e-06, "loss": 0.2017, "step": 19031 }, { "epoch": 5.86, "learning_rate": 7.743018821982835e-06, "loss": 0.0678, "step": 19032 }, { "epoch": 5.86, "learning_rate": 7.742048006530313e-06, "loss": 0.1305, "step": 19033 }, { "epoch": 5.86, "learning_rate": 7.741077213501339e-06, "loss": 0.0845, "step": 19034 }, { "epoch": 5.86, "learning_rate": 7.740106442905548e-06, "loss": 0.1706, "step": 19035 }, { "epoch": 5.86, "learning_rate": 7.739135694752582e-06, "loss": 0.2232, "step": 19036 }, { "epoch": 5.86, "learning_rate": 7.738164969052082e-06, "loss": 0.1506, "step": 19037 }, { "epoch": 5.86, "learning_rate": 7.737194265813687e-06, "loss": 0.2936, "step": 19038 }, { "epoch": 5.86, "learning_rate": 7.736223585047044e-06, "loss": 0.1656, "step": 19039 }, { "epoch": 5.86, "learning_rate": 7.735252926761784e-06, "loss": 0.0848, "step": 19040 }, { "epoch": 5.86, "learning_rate": 7.734282290967548e-06, "loss": 0.1883, "step": 19041 }, { "epoch": 5.86, "learning_rate": 7.733311677673979e-06, "loss": 0.1527, "step": 19042 }, { "epoch": 5.86, "learning_rate": 7.732341086890711e-06, "loss": 0.1798, "step": 19043 }, { "epoch": 5.86, "learning_rate": 7.73137051862739e-06, "loss": 0.1735, "step": 19044 }, { "epoch": 5.86, "learning_rate": 7.730399972893651e-06, "loss": 0.167, "step": 19045 }, { "epoch": 5.86, "learning_rate": 7.729429449699128e-06, "loss": 0.1681, "step": 19046 }, { "epoch": 5.86, "learning_rate": 7.728458949053465e-06, "loss": 0.1451, "step": 19047 }, { "epoch": 5.86, "learning_rate": 7.727488470966298e-06, "loss": 0.1824, "step": 19048 }, { "epoch": 5.86, "learning_rate": 7.726518015447266e-06, "loss": 0.1178, "step": 19049 }, { "epoch": 5.86, "learning_rate": 7.725547582506007e-06, "loss": 0.1008, "step": 19050 }, { "epoch": 5.86, "learning_rate": 7.724577172152153e-06, "loss": 0.1767, "step": 19051 }, { "epoch": 5.86, "learning_rate": 7.723606784395345e-06, "loss": 0.2036, "step": 19052 }, { "epoch": 5.86, "learning_rate": 7.72263641924522e-06, "loss": 0.1251, "step": 19053 }, { "epoch": 5.86, "learning_rate": 7.721666076711416e-06, "loss": 0.178, "step": 19054 }, { "epoch": 5.86, "learning_rate": 7.720695756803569e-06, "loss": 0.1788, "step": 19055 }, { "epoch": 5.86, "learning_rate": 7.719725459531308e-06, "loss": 0.1411, "step": 19056 }, { "epoch": 5.86, "learning_rate": 7.71875518490428e-06, "loss": 0.1126, "step": 19057 }, { "epoch": 5.86, "learning_rate": 7.717784932932112e-06, "loss": 0.2221, "step": 19058 }, { "epoch": 5.86, "learning_rate": 7.716814703624445e-06, "loss": 0.1803, "step": 19059 }, { "epoch": 5.86, "learning_rate": 7.715844496990914e-06, "loss": 0.2037, "step": 19060 }, { "epoch": 5.86, "learning_rate": 7.714874313041148e-06, "loss": 0.2674, "step": 19061 }, { "epoch": 5.87, "learning_rate": 7.71390415178479e-06, "loss": 0.1671, "step": 19062 }, { "epoch": 5.87, "learning_rate": 7.712934013231466e-06, "loss": 0.204, "step": 19063 }, { "epoch": 5.87, "learning_rate": 7.71196389739082e-06, "loss": 0.1905, "step": 19064 }, { "epoch": 5.87, "learning_rate": 7.71099380427248e-06, "loss": 0.2, "step": 19065 }, { "epoch": 5.87, "learning_rate": 7.710023733886078e-06, "loss": 0.1207, "step": 19066 }, { "epoch": 5.87, "learning_rate": 7.709053686241256e-06, "loss": 0.1788, "step": 19067 }, { "epoch": 5.87, "learning_rate": 7.708083661347637e-06, "loss": 0.2764, "step": 19068 }, { "epoch": 5.87, "learning_rate": 7.707113659214863e-06, "loss": 0.2594, "step": 19069 }, { "epoch": 5.87, "learning_rate": 7.706143679852562e-06, "loss": 0.2368, "step": 19070 }, { "epoch": 5.87, "learning_rate": 7.705173723270368e-06, "loss": 0.2078, "step": 19071 }, { "epoch": 5.87, "learning_rate": 7.704203789477917e-06, "loss": 0.1494, "step": 19072 }, { "epoch": 5.87, "learning_rate": 7.703233878484832e-06, "loss": 0.1552, "step": 19073 }, { "epoch": 5.87, "learning_rate": 7.702263990300758e-06, "loss": 0.3042, "step": 19074 }, { "epoch": 5.87, "learning_rate": 7.701294124935319e-06, "loss": 0.1476, "step": 19075 }, { "epoch": 5.87, "learning_rate": 7.700324282398146e-06, "loss": 0.2683, "step": 19076 }, { "epoch": 5.87, "learning_rate": 7.699354462698871e-06, "loss": 0.1346, "step": 19077 }, { "epoch": 5.87, "learning_rate": 7.698384665847127e-06, "loss": 0.268, "step": 19078 }, { "epoch": 5.87, "learning_rate": 7.697414891852549e-06, "loss": 0.2415, "step": 19079 }, { "epoch": 5.87, "learning_rate": 7.69644514072476e-06, "loss": 0.1267, "step": 19080 }, { "epoch": 5.87, "learning_rate": 7.695475412473393e-06, "loss": 0.1617, "step": 19081 }, { "epoch": 5.87, "learning_rate": 7.694505707108079e-06, "loss": 0.1273, "step": 19082 }, { "epoch": 5.87, "learning_rate": 7.693536024638449e-06, "loss": 0.2074, "step": 19083 }, { "epoch": 5.87, "learning_rate": 7.692566365074135e-06, "loss": 0.2522, "step": 19084 }, { "epoch": 5.87, "learning_rate": 7.69159672842476e-06, "loss": 0.1962, "step": 19085 }, { "epoch": 5.87, "learning_rate": 7.690627114699956e-06, "loss": 0.2397, "step": 19086 }, { "epoch": 5.87, "learning_rate": 7.689657523909356e-06, "loss": 0.1778, "step": 19087 }, { "epoch": 5.87, "learning_rate": 7.688687956062582e-06, "loss": 0.1902, "step": 19088 }, { "epoch": 5.87, "learning_rate": 7.687718411169273e-06, "loss": 0.1934, "step": 19089 }, { "epoch": 5.87, "learning_rate": 7.686748889239049e-06, "loss": 0.1919, "step": 19090 }, { "epoch": 5.87, "learning_rate": 7.685779390281536e-06, "loss": 0.0828, "step": 19091 }, { "epoch": 5.87, "learning_rate": 7.68480991430637e-06, "loss": 0.2523, "step": 19092 }, { "epoch": 5.87, "learning_rate": 7.683840461323175e-06, "loss": 0.1024, "step": 19093 }, { "epoch": 5.87, "learning_rate": 7.682871031341579e-06, "loss": 0.189, "step": 19094 }, { "epoch": 5.88, "learning_rate": 7.68190162437121e-06, "loss": 0.1994, "step": 19095 }, { "epoch": 5.88, "learning_rate": 7.680932240421692e-06, "loss": 0.2068, "step": 19096 }, { "epoch": 5.88, "learning_rate": 7.679962879502656e-06, "loss": 0.3422, "step": 19097 }, { "epoch": 5.88, "learning_rate": 7.678993541623725e-06, "loss": 0.177, "step": 19098 }, { "epoch": 5.88, "learning_rate": 7.67802422679453e-06, "loss": 0.0985, "step": 19099 }, { "epoch": 5.88, "learning_rate": 7.677054935024696e-06, "loss": 0.1735, "step": 19100 }, { "epoch": 5.88, "learning_rate": 7.676085666323843e-06, "loss": 0.1142, "step": 19101 }, { "epoch": 5.88, "learning_rate": 7.675116420701603e-06, "loss": 0.1404, "step": 19102 }, { "epoch": 5.88, "learning_rate": 7.674147198167597e-06, "loss": 0.154, "step": 19103 }, { "epoch": 5.88, "learning_rate": 7.673177998731456e-06, "loss": 0.1876, "step": 19104 }, { "epoch": 5.88, "learning_rate": 7.672208822402802e-06, "loss": 0.1901, "step": 19105 }, { "epoch": 5.88, "learning_rate": 7.671239669191257e-06, "loss": 0.1122, "step": 19106 }, { "epoch": 5.88, "learning_rate": 7.670270539106452e-06, "loss": 0.224, "step": 19107 }, { "epoch": 5.88, "learning_rate": 7.669301432158004e-06, "loss": 0.2174, "step": 19108 }, { "epoch": 5.88, "learning_rate": 7.668332348355542e-06, "loss": 0.2005, "step": 19109 }, { "epoch": 5.88, "learning_rate": 7.667363287708692e-06, "loss": 0.2339, "step": 19110 }, { "epoch": 5.88, "learning_rate": 7.666394250227068e-06, "loss": 0.1322, "step": 19111 }, { "epoch": 5.88, "learning_rate": 7.665425235920307e-06, "loss": 0.1174, "step": 19112 }, { "epoch": 5.88, "learning_rate": 7.66445624479802e-06, "loss": 0.1057, "step": 19113 }, { "epoch": 5.88, "learning_rate": 7.663487276869837e-06, "loss": 0.2438, "step": 19114 }, { "epoch": 5.88, "learning_rate": 7.662518332145377e-06, "loss": 0.2653, "step": 19115 }, { "epoch": 5.88, "learning_rate": 7.661549410634265e-06, "loss": 0.2047, "step": 19116 }, { "epoch": 5.88, "learning_rate": 7.660580512346126e-06, "loss": 0.2312, "step": 19117 }, { "epoch": 5.88, "learning_rate": 7.659611637290573e-06, "loss": 0.0896, "step": 19118 }, { "epoch": 5.88, "learning_rate": 7.658642785477236e-06, "loss": 0.2733, "step": 19119 }, { "epoch": 5.88, "learning_rate": 7.657673956915735e-06, "loss": 0.144, "step": 19120 }, { "epoch": 5.88, "learning_rate": 7.656705151615688e-06, "loss": 0.2044, "step": 19121 }, { "epoch": 5.88, "learning_rate": 7.655736369586723e-06, "loss": 0.2167, "step": 19122 }, { "epoch": 5.88, "learning_rate": 7.65476761083845e-06, "loss": 0.2392, "step": 19123 }, { "epoch": 5.88, "learning_rate": 7.6537988753805e-06, "loss": 0.1403, "step": 19124 }, { "epoch": 5.88, "learning_rate": 7.65283016322249e-06, "loss": 0.1416, "step": 19125 }, { "epoch": 5.88, "learning_rate": 7.651861474374036e-06, "loss": 0.2148, "step": 19126 }, { "epoch": 5.89, "learning_rate": 7.650892808844765e-06, "loss": 0.1707, "step": 19127 }, { "epoch": 5.89, "learning_rate": 7.649924166644291e-06, "loss": 0.2009, "step": 19128 }, { "epoch": 5.89, "learning_rate": 7.648955547782239e-06, "loss": 0.202, "step": 19129 }, { "epoch": 5.89, "learning_rate": 7.647986952268224e-06, "loss": 0.16, "step": 19130 }, { "epoch": 5.89, "learning_rate": 7.647018380111865e-06, "loss": 0.2125, "step": 19131 }, { "epoch": 5.89, "learning_rate": 7.646049831322783e-06, "loss": 0.251, "step": 19132 }, { "epoch": 5.89, "learning_rate": 7.645081305910596e-06, "loss": 0.1657, "step": 19133 }, { "epoch": 5.89, "learning_rate": 7.644112803884923e-06, "loss": 0.1282, "step": 19134 }, { "epoch": 5.89, "learning_rate": 7.64314432525538e-06, "loss": 0.1998, "step": 19135 }, { "epoch": 5.89, "learning_rate": 7.642175870031584e-06, "loss": 0.4001, "step": 19136 }, { "epoch": 5.89, "learning_rate": 7.641207438223157e-06, "loss": 0.1497, "step": 19137 }, { "epoch": 5.89, "learning_rate": 7.64023902983971e-06, "loss": 0.2052, "step": 19138 }, { "epoch": 5.89, "learning_rate": 7.639270644890873e-06, "loss": 0.2408, "step": 19139 }, { "epoch": 5.89, "learning_rate": 7.638302283386248e-06, "loss": 0.1979, "step": 19140 }, { "epoch": 5.89, "learning_rate": 7.637333945335458e-06, "loss": 0.2513, "step": 19141 }, { "epoch": 5.89, "learning_rate": 7.63636563074812e-06, "loss": 0.2381, "step": 19142 }, { "epoch": 5.89, "learning_rate": 7.635397339633847e-06, "loss": 0.239, "step": 19143 }, { "epoch": 5.89, "learning_rate": 7.634429072002263e-06, "loss": 0.233, "step": 19144 }, { "epoch": 5.89, "learning_rate": 7.633460827862977e-06, "loss": 0.1451, "step": 19145 }, { "epoch": 5.89, "learning_rate": 7.632492607225604e-06, "loss": 0.165, "step": 19146 }, { "epoch": 5.89, "learning_rate": 7.631524410099763e-06, "loss": 0.2178, "step": 19147 }, { "epoch": 5.89, "learning_rate": 7.630556236495066e-06, "loss": 0.1594, "step": 19148 }, { "epoch": 5.89, "learning_rate": 7.62958808642113e-06, "loss": 0.1766, "step": 19149 }, { "epoch": 5.89, "learning_rate": 7.628619959887572e-06, "loss": 0.1587, "step": 19150 }, { "epoch": 5.89, "learning_rate": 7.627651856903999e-06, "loss": 0.1205, "step": 19151 }, { "epoch": 5.89, "learning_rate": 7.6266837774800315e-06, "loss": 0.1855, "step": 19152 }, { "epoch": 5.89, "learning_rate": 7.62571572162528e-06, "loss": 0.1015, "step": 19153 }, { "epoch": 5.89, "learning_rate": 7.62474768934936e-06, "loss": 0.1839, "step": 19154 }, { "epoch": 5.89, "learning_rate": 7.623779680661888e-06, "loss": 0.2716, "step": 19155 }, { "epoch": 5.89, "learning_rate": 7.622811695572467e-06, "loss": 0.1721, "step": 19156 }, { "epoch": 5.89, "learning_rate": 7.621843734090724e-06, "loss": 0.1553, "step": 19157 }, { "epoch": 5.89, "learning_rate": 7.620875796226258e-06, "loss": 0.2061, "step": 19158 }, { "epoch": 5.89, "learning_rate": 7.619907881988692e-06, "loss": 0.2239, "step": 19159 }, { "epoch": 5.9, "learning_rate": 7.618939991387636e-06, "loss": 0.2931, "step": 19160 }, { "epoch": 5.9, "learning_rate": 7.617972124432693e-06, "loss": 0.1695, "step": 19161 }, { "epoch": 5.9, "learning_rate": 7.61700428113349e-06, "loss": 0.1293, "step": 19162 }, { "epoch": 5.9, "learning_rate": 7.616036461499626e-06, "loss": 0.1261, "step": 19163 }, { "epoch": 5.9, "learning_rate": 7.615068665540719e-06, "loss": 0.1351, "step": 19164 }, { "epoch": 5.9, "learning_rate": 7.6141008932663786e-06, "loss": 0.1604, "step": 19165 }, { "epoch": 5.9, "learning_rate": 7.613133144686212e-06, "loss": 0.1845, "step": 19166 }, { "epoch": 5.9, "learning_rate": 7.612165419809839e-06, "loss": 0.1581, "step": 19167 }, { "epoch": 5.9, "learning_rate": 7.611197718646859e-06, "loss": 0.1681, "step": 19168 }, { "epoch": 5.9, "learning_rate": 7.610230041206888e-06, "loss": 0.1635, "step": 19169 }, { "epoch": 5.9, "learning_rate": 7.609262387499536e-06, "loss": 0.3031, "step": 19170 }, { "epoch": 5.9, "learning_rate": 7.608294757534411e-06, "loss": 0.1926, "step": 19171 }, { "epoch": 5.9, "learning_rate": 7.607327151321127e-06, "loss": 0.171, "step": 19172 }, { "epoch": 5.9, "learning_rate": 7.606359568869282e-06, "loss": 0.2227, "step": 19173 }, { "epoch": 5.9, "learning_rate": 7.6053920101885e-06, "loss": 0.2198, "step": 19174 }, { "epoch": 5.9, "learning_rate": 7.604424475288381e-06, "loss": 0.2781, "step": 19175 }, { "epoch": 5.9, "learning_rate": 7.60345696417853e-06, "loss": 0.1874, "step": 19176 }, { "epoch": 5.9, "learning_rate": 7.602489476868565e-06, "loss": 0.0783, "step": 19177 }, { "epoch": 5.9, "learning_rate": 7.601522013368086e-06, "loss": 0.1635, "step": 19178 }, { "epoch": 5.9, "learning_rate": 7.600554573686707e-06, "loss": 0.1466, "step": 19179 }, { "epoch": 5.9, "learning_rate": 7.599587157834032e-06, "loss": 0.224, "step": 19180 }, { "epoch": 5.9, "learning_rate": 7.598619765819666e-06, "loss": 0.1394, "step": 19181 }, { "epoch": 5.9, "learning_rate": 7.597652397653221e-06, "loss": 0.1803, "step": 19182 }, { "epoch": 5.9, "learning_rate": 7.596685053344299e-06, "loss": 0.2704, "step": 19183 }, { "epoch": 5.9, "learning_rate": 7.595717732902515e-06, "loss": 0.18, "step": 19184 }, { "epoch": 5.9, "learning_rate": 7.594750436337467e-06, "loss": 0.1501, "step": 19185 }, { "epoch": 5.9, "learning_rate": 7.593783163658761e-06, "loss": 0.1045, "step": 19186 }, { "epoch": 5.9, "learning_rate": 7.592815914876009e-06, "loss": 0.1841, "step": 19187 }, { "epoch": 5.9, "learning_rate": 7.591848689998811e-06, "loss": 0.2214, "step": 19188 }, { "epoch": 5.9, "learning_rate": 7.590881489036779e-06, "loss": 0.1709, "step": 19189 }, { "epoch": 5.9, "learning_rate": 7.589914311999512e-06, "loss": 0.1514, "step": 19190 }, { "epoch": 5.9, "learning_rate": 7.588947158896615e-06, "loss": 0.2673, "step": 19191 }, { "epoch": 5.91, "learning_rate": 7.587980029737698e-06, "loss": 0.1359, "step": 19192 }, { "epoch": 5.91, "learning_rate": 7.587012924532359e-06, "loss": 0.21, "step": 19193 }, { "epoch": 5.91, "learning_rate": 7.586045843290207e-06, "loss": 0.1087, "step": 19194 }, { "epoch": 5.91, "learning_rate": 7.585078786020848e-06, "loss": 0.1765, "step": 19195 }, { "epoch": 5.91, "learning_rate": 7.584111752733878e-06, "loss": 0.1509, "step": 19196 }, { "epoch": 5.91, "learning_rate": 7.583144743438906e-06, "loss": 0.1248, "step": 19197 }, { "epoch": 5.91, "learning_rate": 7.582177758145532e-06, "loss": 0.18, "step": 19198 }, { "epoch": 5.91, "learning_rate": 7.5812107968633656e-06, "loss": 0.2264, "step": 19199 }, { "epoch": 5.91, "learning_rate": 7.580243859602006e-06, "loss": 0.1605, "step": 19200 }, { "epoch": 5.91, "learning_rate": 7.579276946371049e-06, "loss": 0.2002, "step": 19201 }, { "epoch": 5.91, "learning_rate": 7.5783100571801075e-06, "loss": 0.2053, "step": 19202 }, { "epoch": 5.91, "learning_rate": 7.577343192038775e-06, "loss": 0.0957, "step": 19203 }, { "epoch": 5.91, "learning_rate": 7.57637635095666e-06, "loss": 0.1538, "step": 19204 }, { "epoch": 5.91, "learning_rate": 7.575409533943364e-06, "loss": 0.0998, "step": 19205 }, { "epoch": 5.91, "learning_rate": 7.5744427410084794e-06, "loss": 0.2242, "step": 19206 }, { "epoch": 5.91, "learning_rate": 7.573475972161622e-06, "loss": 0.1739, "step": 19207 }, { "epoch": 5.91, "learning_rate": 7.572509227412378e-06, "loss": 0.1771, "step": 19208 }, { "epoch": 5.91, "learning_rate": 7.571542506770357e-06, "loss": 0.219, "step": 19209 }, { "epoch": 5.91, "learning_rate": 7.57057581024516e-06, "loss": 0.1944, "step": 19210 }, { "epoch": 5.91, "learning_rate": 7.569609137846376e-06, "loss": 0.1858, "step": 19211 }, { "epoch": 5.91, "learning_rate": 7.568642489583622e-06, "loss": 0.1449, "step": 19212 }, { "epoch": 5.91, "learning_rate": 7.567675865466484e-06, "loss": 0.2511, "step": 19213 }, { "epoch": 5.91, "learning_rate": 7.566709265504567e-06, "loss": 0.1108, "step": 19214 }, { "epoch": 5.91, "learning_rate": 7.5657426897074694e-06, "loss": 0.1578, "step": 19215 }, { "epoch": 5.91, "learning_rate": 7.564776138084791e-06, "loss": 0.1567, "step": 19216 }, { "epoch": 5.91, "learning_rate": 7.563809610646132e-06, "loss": 0.1442, "step": 19217 }, { "epoch": 5.91, "learning_rate": 7.562843107401086e-06, "loss": 0.2609, "step": 19218 }, { "epoch": 5.91, "learning_rate": 7.5618766283592545e-06, "loss": 0.325, "step": 19219 }, { "epoch": 5.91, "learning_rate": 7.5609101735302344e-06, "loss": 0.2176, "step": 19220 }, { "epoch": 5.91, "learning_rate": 7.559943742923626e-06, "loss": 0.1804, "step": 19221 }, { "epoch": 5.91, "learning_rate": 7.558977336549027e-06, "loss": 0.1408, "step": 19222 }, { "epoch": 5.91, "learning_rate": 7.558010954416026e-06, "loss": 0.1636, "step": 19223 }, { "epoch": 5.91, "learning_rate": 7.557044596534234e-06, "loss": 0.1311, "step": 19224 }, { "epoch": 5.92, "learning_rate": 7.556078262913236e-06, "loss": 0.1201, "step": 19225 }, { "epoch": 5.92, "learning_rate": 7.555111953562635e-06, "loss": 0.2186, "step": 19226 }, { "epoch": 5.92, "learning_rate": 7.554145668492029e-06, "loss": 0.1991, "step": 19227 }, { "epoch": 5.92, "learning_rate": 7.553179407711003e-06, "loss": 0.1784, "step": 19228 }, { "epoch": 5.92, "learning_rate": 7.552213171229168e-06, "loss": 0.1441, "step": 19229 }, { "epoch": 5.92, "learning_rate": 7.551246959056106e-06, "loss": 0.2021, "step": 19230 }, { "epoch": 5.92, "learning_rate": 7.550280771201423e-06, "loss": 0.2014, "step": 19231 }, { "epoch": 5.92, "learning_rate": 7.549314607674708e-06, "loss": 0.1929, "step": 19232 }, { "epoch": 5.92, "learning_rate": 7.548348468485556e-06, "loss": 0.1696, "step": 19233 }, { "epoch": 5.92, "learning_rate": 7.547382353643568e-06, "loss": 0.1395, "step": 19234 }, { "epoch": 5.92, "learning_rate": 7.546416263158326e-06, "loss": 0.1338, "step": 19235 }, { "epoch": 5.92, "learning_rate": 7.545450197039439e-06, "loss": 0.2172, "step": 19236 }, { "epoch": 5.92, "learning_rate": 7.544484155296492e-06, "loss": 0.2002, "step": 19237 }, { "epoch": 5.92, "learning_rate": 7.5435181379390765e-06, "loss": 0.1415, "step": 19238 }, { "epoch": 5.92, "learning_rate": 7.5425521449767955e-06, "loss": 0.2386, "step": 19239 }, { "epoch": 5.92, "learning_rate": 7.541586176419231e-06, "loss": 0.1317, "step": 19240 }, { "epoch": 5.92, "learning_rate": 7.540620232275986e-06, "loss": 0.1369, "step": 19241 }, { "epoch": 5.92, "learning_rate": 7.539654312556648e-06, "loss": 0.1397, "step": 19242 }, { "epoch": 5.92, "learning_rate": 7.538688417270808e-06, "loss": 0.142, "step": 19243 }, { "epoch": 5.92, "learning_rate": 7.537722546428061e-06, "loss": 0.1316, "step": 19244 }, { "epoch": 5.92, "learning_rate": 7.536756700037998e-06, "loss": 0.1242, "step": 19245 }, { "epoch": 5.92, "learning_rate": 7.535790878110215e-06, "loss": 0.1848, "step": 19246 }, { "epoch": 5.92, "learning_rate": 7.534825080654298e-06, "loss": 0.2166, "step": 19247 }, { "epoch": 5.92, "learning_rate": 7.533859307679836e-06, "loss": 0.2595, "step": 19248 }, { "epoch": 5.92, "learning_rate": 7.532893559196428e-06, "loss": 0.1624, "step": 19249 }, { "epoch": 5.92, "learning_rate": 7.531927835213657e-06, "loss": 0.1582, "step": 19250 }, { "epoch": 5.92, "learning_rate": 7.5309621357411224e-06, "loss": 0.1284, "step": 19251 }, { "epoch": 5.92, "learning_rate": 7.529996460788407e-06, "loss": 0.1981, "step": 19252 }, { "epoch": 5.92, "learning_rate": 7.529030810365101e-06, "loss": 0.1566, "step": 19253 }, { "epoch": 5.92, "learning_rate": 7.5280651844807995e-06, "loss": 0.1086, "step": 19254 }, { "epoch": 5.92, "learning_rate": 7.527099583145085e-06, "loss": 0.1929, "step": 19255 }, { "epoch": 5.92, "learning_rate": 7.526134006367556e-06, "loss": 0.2232, "step": 19256 }, { "epoch": 5.93, "learning_rate": 7.525168454157794e-06, "loss": 0.2447, "step": 19257 }, { "epoch": 5.93, "learning_rate": 7.524202926525388e-06, "loss": 0.1825, "step": 19258 }, { "epoch": 5.93, "learning_rate": 7.523237423479932e-06, "loss": 0.2264, "step": 19259 }, { "epoch": 5.93, "learning_rate": 7.5222719450310075e-06, "loss": 0.1667, "step": 19260 }, { "epoch": 5.93, "learning_rate": 7.521306491188208e-06, "loss": 0.2521, "step": 19261 }, { "epoch": 5.93, "learning_rate": 7.520341061961123e-06, "loss": 0.11, "step": 19262 }, { "epoch": 5.93, "learning_rate": 7.519375657359331e-06, "loss": 0.1893, "step": 19263 }, { "epoch": 5.93, "learning_rate": 7.518410277392428e-06, "loss": 0.1403, "step": 19264 }, { "epoch": 5.93, "learning_rate": 7.517444922069995e-06, "loss": 0.1285, "step": 19265 }, { "epoch": 5.93, "learning_rate": 7.516479591401624e-06, "loss": 0.199, "step": 19266 }, { "epoch": 5.93, "learning_rate": 7.515514285396903e-06, "loss": 0.1163, "step": 19267 }, { "epoch": 5.93, "learning_rate": 7.51454900406541e-06, "loss": 0.1422, "step": 19268 }, { "epoch": 5.93, "learning_rate": 7.513583747416738e-06, "loss": 0.1298, "step": 19269 }, { "epoch": 5.93, "learning_rate": 7.512618515460468e-06, "loss": 0.1603, "step": 19270 }, { "epoch": 5.93, "learning_rate": 7.511653308206192e-06, "loss": 0.1381, "step": 19271 }, { "epoch": 5.93, "learning_rate": 7.5106881256634936e-06, "loss": 0.2063, "step": 19272 }, { "epoch": 5.93, "learning_rate": 7.50972296784195e-06, "loss": 0.1157, "step": 19273 }, { "epoch": 5.93, "learning_rate": 7.508757834751159e-06, "loss": 0.133, "step": 19274 }, { "epoch": 5.93, "learning_rate": 7.507792726400694e-06, "loss": 0.174, "step": 19275 }, { "epoch": 5.93, "learning_rate": 7.506827642800146e-06, "loss": 0.204, "step": 19276 }, { "epoch": 5.93, "learning_rate": 7.5058625839591e-06, "loss": 0.2181, "step": 19277 }, { "epoch": 5.93, "learning_rate": 7.504897549887131e-06, "loss": 0.1399, "step": 19278 }, { "epoch": 5.93, "learning_rate": 7.503932540593835e-06, "loss": 0.1494, "step": 19279 }, { "epoch": 5.93, "learning_rate": 7.502967556088785e-06, "loss": 0.1477, "step": 19280 }, { "epoch": 5.93, "learning_rate": 7.50200259638157e-06, "loss": 0.1843, "step": 19281 }, { "epoch": 5.93, "learning_rate": 7.501037661481773e-06, "loss": 0.1836, "step": 19282 }, { "epoch": 5.93, "learning_rate": 7.500072751398972e-06, "loss": 0.1672, "step": 19283 }, { "epoch": 5.93, "learning_rate": 7.499107866142757e-06, "loss": 0.2756, "step": 19284 }, { "epoch": 5.93, "learning_rate": 7.4981430057227015e-06, "loss": 0.201, "step": 19285 }, { "epoch": 5.93, "learning_rate": 7.497178170148394e-06, "loss": 0.1112, "step": 19286 }, { "epoch": 5.93, "learning_rate": 7.496213359429414e-06, "loss": 0.1726, "step": 19287 }, { "epoch": 5.93, "learning_rate": 7.49524857357534e-06, "loss": 0.1289, "step": 19288 }, { "epoch": 5.93, "learning_rate": 7.49428381259576e-06, "loss": 0.2444, "step": 19289 }, { "epoch": 5.94, "learning_rate": 7.4933190765002476e-06, "loss": 0.0938, "step": 19290 }, { "epoch": 5.94, "learning_rate": 7.492354365298391e-06, "loss": 0.1727, "step": 19291 }, { "epoch": 5.94, "learning_rate": 7.491389678999764e-06, "loss": 0.1667, "step": 19292 }, { "epoch": 5.94, "learning_rate": 7.490425017613949e-06, "loss": 0.1713, "step": 19293 }, { "epoch": 5.94, "learning_rate": 7.489460381150527e-06, "loss": 0.2829, "step": 19294 }, { "epoch": 5.94, "learning_rate": 7.488495769619077e-06, "loss": 0.1733, "step": 19295 }, { "epoch": 5.94, "learning_rate": 7.487531183029182e-06, "loss": 0.1692, "step": 19296 }, { "epoch": 5.94, "learning_rate": 7.486566621390414e-06, "loss": 0.1242, "step": 19297 }, { "epoch": 5.94, "learning_rate": 7.485602084712356e-06, "loss": 0.1447, "step": 19298 }, { "epoch": 5.94, "learning_rate": 7.4846375730045885e-06, "loss": 0.2317, "step": 19299 }, { "epoch": 5.94, "learning_rate": 7.483673086276686e-06, "loss": 0.2521, "step": 19300 }, { "epoch": 5.94, "learning_rate": 7.482708624538233e-06, "loss": 0.2186, "step": 19301 }, { "epoch": 5.94, "learning_rate": 7.4817441877988005e-06, "loss": 0.2073, "step": 19302 }, { "epoch": 5.94, "learning_rate": 7.4807797760679675e-06, "loss": 0.1267, "step": 19303 }, { "epoch": 5.94, "learning_rate": 7.479815389355316e-06, "loss": 0.1235, "step": 19304 }, { "epoch": 5.94, "learning_rate": 7.478851027670417e-06, "loss": 0.1502, "step": 19305 }, { "epoch": 5.94, "learning_rate": 7.477886691022856e-06, "loss": 0.0765, "step": 19306 }, { "epoch": 5.94, "learning_rate": 7.476922379422202e-06, "loss": 0.141, "step": 19307 }, { "epoch": 5.94, "learning_rate": 7.475958092878032e-06, "loss": 0.2151, "step": 19308 }, { "epoch": 5.94, "learning_rate": 7.474993831399926e-06, "loss": 0.2208, "step": 19309 }, { "epoch": 5.94, "learning_rate": 7.474029594997456e-06, "loss": 0.1294, "step": 19310 }, { "epoch": 5.94, "learning_rate": 7.473065383680205e-06, "loss": 0.1473, "step": 19311 }, { "epoch": 5.94, "learning_rate": 7.4721011974577425e-06, "loss": 0.1963, "step": 19312 }, { "epoch": 5.94, "learning_rate": 7.471137036339642e-06, "loss": 0.2132, "step": 19313 }, { "epoch": 5.94, "learning_rate": 7.470172900335484e-06, "loss": 0.1443, "step": 19314 }, { "epoch": 5.94, "learning_rate": 7.469208789454838e-06, "loss": 0.1843, "step": 19315 }, { "epoch": 5.94, "learning_rate": 7.4682447037072835e-06, "loss": 0.2256, "step": 19316 }, { "epoch": 5.94, "learning_rate": 7.467280643102393e-06, "loss": 0.2129, "step": 19317 }, { "epoch": 5.94, "learning_rate": 7.4663166076497376e-06, "loss": 0.1577, "step": 19318 }, { "epoch": 5.94, "learning_rate": 7.465352597358895e-06, "loss": 0.1923, "step": 19319 }, { "epoch": 5.94, "learning_rate": 7.464388612239435e-06, "loss": 0.1423, "step": 19320 }, { "epoch": 5.94, "learning_rate": 7.463424652300934e-06, "loss": 0.0983, "step": 19321 }, { "epoch": 5.95, "learning_rate": 7.462460717552968e-06, "loss": 0.2402, "step": 19322 }, { "epoch": 5.95, "learning_rate": 7.461496808005099e-06, "loss": 0.2357, "step": 19323 }, { "epoch": 5.95, "learning_rate": 7.460532923666912e-06, "loss": 0.1906, "step": 19324 }, { "epoch": 5.95, "learning_rate": 7.459569064547969e-06, "loss": 0.1585, "step": 19325 }, { "epoch": 5.95, "learning_rate": 7.45860523065785e-06, "loss": 0.2762, "step": 19326 }, { "epoch": 5.95, "learning_rate": 7.457641422006123e-06, "loss": 0.1631, "step": 19327 }, { "epoch": 5.95, "learning_rate": 7.456677638602355e-06, "loss": 0.247, "step": 19328 }, { "epoch": 5.95, "learning_rate": 7.455713880456129e-06, "loss": 0.2248, "step": 19329 }, { "epoch": 5.95, "learning_rate": 7.454750147577004e-06, "loss": 0.2187, "step": 19330 }, { "epoch": 5.95, "learning_rate": 7.453786439974557e-06, "loss": 0.1614, "step": 19331 }, { "epoch": 5.95, "learning_rate": 7.45282275765836e-06, "loss": 0.2158, "step": 19332 }, { "epoch": 5.95, "learning_rate": 7.451859100637976e-06, "loss": 0.1501, "step": 19333 }, { "epoch": 5.95, "learning_rate": 7.450895468922985e-06, "loss": 0.2574, "step": 19334 }, { "epoch": 5.95, "learning_rate": 7.449931862522946e-06, "loss": 0.0748, "step": 19335 }, { "epoch": 5.95, "learning_rate": 7.448968281447437e-06, "loss": 0.1964, "step": 19336 }, { "epoch": 5.95, "learning_rate": 7.448004725706024e-06, "loss": 0.161, "step": 19337 }, { "epoch": 5.95, "learning_rate": 7.447041195308273e-06, "loss": 0.2068, "step": 19338 }, { "epoch": 5.95, "learning_rate": 7.44607769026376e-06, "loss": 0.1736, "step": 19339 }, { "epoch": 5.95, "learning_rate": 7.445114210582044e-06, "loss": 0.1218, "step": 19340 }, { "epoch": 5.95, "learning_rate": 7.444150756272704e-06, "loss": 0.1424, "step": 19341 }, { "epoch": 5.95, "learning_rate": 7.443187327345301e-06, "loss": 0.1766, "step": 19342 }, { "epoch": 5.95, "learning_rate": 7.442223923809402e-06, "loss": 0.2261, "step": 19343 }, { "epoch": 5.95, "learning_rate": 7.44126054567458e-06, "loss": 0.1345, "step": 19344 }, { "epoch": 5.95, "learning_rate": 7.440297192950395e-06, "loss": 0.0754, "step": 19345 }, { "epoch": 5.95, "learning_rate": 7.439333865646423e-06, "loss": 0.2329, "step": 19346 }, { "epoch": 5.95, "learning_rate": 7.438370563772224e-06, "loss": 0.1258, "step": 19347 }, { "epoch": 5.95, "learning_rate": 7.437407287337364e-06, "loss": 0.073, "step": 19348 }, { "epoch": 5.95, "learning_rate": 7.4364440363514125e-06, "loss": 0.2119, "step": 19349 }, { "epoch": 5.95, "learning_rate": 7.4354808108239325e-06, "loss": 0.2301, "step": 19350 }, { "epoch": 5.95, "learning_rate": 7.434517610764495e-06, "loss": 0.1878, "step": 19351 }, { "epoch": 5.95, "learning_rate": 7.4335544361826616e-06, "loss": 0.2581, "step": 19352 }, { "epoch": 5.95, "learning_rate": 7.432591287087996e-06, "loss": 0.1331, "step": 19353 }, { "epoch": 5.95, "learning_rate": 7.431628163490067e-06, "loss": 0.1122, "step": 19354 }, { "epoch": 5.96, "learning_rate": 7.430665065398435e-06, "loss": 0.2238, "step": 19355 }, { "epoch": 5.96, "learning_rate": 7.429701992822671e-06, "loss": 0.149, "step": 19356 }, { "epoch": 5.96, "learning_rate": 7.428738945772332e-06, "loss": 0.2141, "step": 19357 }, { "epoch": 5.96, "learning_rate": 7.427775924256984e-06, "loss": 0.1944, "step": 19358 }, { "epoch": 5.96, "learning_rate": 7.426812928286193e-06, "loss": 0.1756, "step": 19359 }, { "epoch": 5.96, "learning_rate": 7.42584995786952e-06, "loss": 0.1343, "step": 19360 }, { "epoch": 5.96, "learning_rate": 7.424887013016531e-06, "loss": 0.1148, "step": 19361 }, { "epoch": 5.96, "learning_rate": 7.423924093736789e-06, "loss": 0.1288, "step": 19362 }, { "epoch": 5.96, "learning_rate": 7.422961200039852e-06, "loss": 0.0981, "step": 19363 }, { "epoch": 5.96, "learning_rate": 7.421998331935287e-06, "loss": 0.2526, "step": 19364 }, { "epoch": 5.96, "learning_rate": 7.421035489432652e-06, "loss": 0.1315, "step": 19365 }, { "epoch": 5.96, "learning_rate": 7.420072672541514e-06, "loss": 0.1386, "step": 19366 }, { "epoch": 5.96, "learning_rate": 7.419109881271434e-06, "loss": 0.2044, "step": 19367 }, { "epoch": 5.96, "learning_rate": 7.418147115631966e-06, "loss": 0.1577, "step": 19368 }, { "epoch": 5.96, "learning_rate": 7.41718437563268e-06, "loss": 0.1845, "step": 19369 }, { "epoch": 5.96, "learning_rate": 7.416221661283132e-06, "loss": 0.2448, "step": 19370 }, { "epoch": 5.96, "learning_rate": 7.415258972592885e-06, "loss": 0.1124, "step": 19371 }, { "epoch": 5.96, "learning_rate": 7.4142963095715e-06, "loss": 0.1658, "step": 19372 }, { "epoch": 5.96, "learning_rate": 7.413333672228531e-06, "loss": 0.1031, "step": 19373 }, { "epoch": 5.96, "learning_rate": 7.412371060573549e-06, "loss": 0.1375, "step": 19374 }, { "epoch": 5.96, "learning_rate": 7.411408474616102e-06, "loss": 0.1691, "step": 19375 }, { "epoch": 5.96, "learning_rate": 7.410445914365758e-06, "loss": 0.2545, "step": 19376 }, { "epoch": 5.96, "learning_rate": 7.409483379832074e-06, "loss": 0.1786, "step": 19377 }, { "epoch": 5.96, "learning_rate": 7.408520871024601e-06, "loss": 0.1295, "step": 19378 }, { "epoch": 5.96, "learning_rate": 7.407558387952911e-06, "loss": 0.1771, "step": 19379 }, { "epoch": 5.96, "learning_rate": 7.40659593062655e-06, "loss": 0.176, "step": 19380 }, { "epoch": 5.96, "learning_rate": 7.405633499055085e-06, "loss": 0.1289, "step": 19381 }, { "epoch": 5.96, "learning_rate": 7.40467109324807e-06, "loss": 0.2338, "step": 19382 }, { "epoch": 5.96, "learning_rate": 7.4037087132150605e-06, "loss": 0.177, "step": 19383 }, { "epoch": 5.96, "learning_rate": 7.40274635896562e-06, "loss": 0.1447, "step": 19384 }, { "epoch": 5.96, "learning_rate": 7.401784030509299e-06, "loss": 0.125, "step": 19385 }, { "epoch": 5.96, "learning_rate": 7.400821727855658e-06, "loss": 0.1162, "step": 19386 }, { "epoch": 5.97, "learning_rate": 7.399859451014254e-06, "loss": 0.1762, "step": 19387 }, { "epoch": 5.97, "learning_rate": 7.398897199994639e-06, "loss": 0.2382, "step": 19388 }, { "epoch": 5.97, "learning_rate": 7.397934974806378e-06, "loss": 0.1514, "step": 19389 }, { "epoch": 5.97, "learning_rate": 7.396972775459013e-06, "loss": 0.2498, "step": 19390 }, { "epoch": 5.97, "learning_rate": 7.396010601962114e-06, "loss": 0.17, "step": 19391 }, { "epoch": 5.97, "learning_rate": 7.395048454325227e-06, "loss": 0.143, "step": 19392 }, { "epoch": 5.97, "learning_rate": 7.394086332557907e-06, "loss": 0.1431, "step": 19393 }, { "epoch": 5.97, "learning_rate": 7.393124236669714e-06, "loss": 0.1778, "step": 19394 }, { "epoch": 5.97, "learning_rate": 7.392162166670198e-06, "loss": 0.1917, "step": 19395 }, { "epoch": 5.97, "learning_rate": 7.39120012256892e-06, "loss": 0.236, "step": 19396 }, { "epoch": 5.97, "learning_rate": 7.390238104375427e-06, "loss": 0.3008, "step": 19397 }, { "epoch": 5.97, "learning_rate": 7.389276112099272e-06, "loss": 0.1571, "step": 19398 }, { "epoch": 5.97, "learning_rate": 7.388314145750013e-06, "loss": 0.1594, "step": 19399 }, { "epoch": 5.97, "learning_rate": 7.3873522053372e-06, "loss": 0.0795, "step": 19400 }, { "epoch": 5.97, "learning_rate": 7.386390290870392e-06, "loss": 0.2447, "step": 19401 }, { "epoch": 5.97, "learning_rate": 7.385428402359135e-06, "loss": 0.1094, "step": 19402 }, { "epoch": 5.97, "learning_rate": 7.384466539812981e-06, "loss": 0.1301, "step": 19403 }, { "epoch": 5.97, "learning_rate": 7.383504703241488e-06, "loss": 0.1649, "step": 19404 }, { "epoch": 5.97, "learning_rate": 7.382542892654201e-06, "loss": 0.1864, "step": 19405 }, { "epoch": 5.97, "learning_rate": 7.38158110806068e-06, "loss": 0.1197, "step": 19406 }, { "epoch": 5.97, "learning_rate": 7.380619349470472e-06, "loss": 0.1041, "step": 19407 }, { "epoch": 5.97, "learning_rate": 7.379657616893124e-06, "loss": 0.146, "step": 19408 }, { "epoch": 5.97, "learning_rate": 7.378695910338191e-06, "loss": 0.1052, "step": 19409 }, { "epoch": 5.97, "learning_rate": 7.377734229815225e-06, "loss": 0.1669, "step": 19410 }, { "epoch": 5.97, "learning_rate": 7.376772575333775e-06, "loss": 0.2198, "step": 19411 }, { "epoch": 5.97, "learning_rate": 7.37581094690339e-06, "loss": 0.1504, "step": 19412 }, { "epoch": 5.97, "learning_rate": 7.3748493445336235e-06, "loss": 0.1248, "step": 19413 }, { "epoch": 5.97, "learning_rate": 7.37388776823402e-06, "loss": 0.2179, "step": 19414 }, { "epoch": 5.97, "learning_rate": 7.372926218014131e-06, "loss": 0.2158, "step": 19415 }, { "epoch": 5.97, "learning_rate": 7.371964693883507e-06, "loss": 0.1815, "step": 19416 }, { "epoch": 5.97, "learning_rate": 7.371003195851693e-06, "loss": 0.1106, "step": 19417 }, { "epoch": 5.97, "learning_rate": 7.370041723928246e-06, "loss": 0.1393, "step": 19418 }, { "epoch": 5.97, "learning_rate": 7.3690802781227056e-06, "loss": 0.2019, "step": 19419 }, { "epoch": 5.98, "learning_rate": 7.3681188584446205e-06, "loss": 0.2049, "step": 19420 }, { "epoch": 5.98, "learning_rate": 7.367157464903543e-06, "loss": 0.1979, "step": 19421 }, { "epoch": 5.98, "learning_rate": 7.366196097509016e-06, "loss": 0.2452, "step": 19422 }, { "epoch": 5.98, "learning_rate": 7.365234756270592e-06, "loss": 0.167, "step": 19423 }, { "epoch": 5.98, "learning_rate": 7.3642734411978166e-06, "loss": 0.0954, "step": 19424 }, { "epoch": 5.98, "learning_rate": 7.36331215230023e-06, "loss": 0.1955, "step": 19425 }, { "epoch": 5.98, "learning_rate": 7.362350889587389e-06, "loss": 0.2022, "step": 19426 }, { "epoch": 5.98, "learning_rate": 7.361389653068829e-06, "loss": 0.1856, "step": 19427 }, { "epoch": 5.98, "learning_rate": 7.360428442754105e-06, "loss": 0.167, "step": 19428 }, { "epoch": 5.98, "learning_rate": 7.359467258652762e-06, "loss": 0.1327, "step": 19429 }, { "epoch": 5.98, "learning_rate": 7.358506100774338e-06, "loss": 0.1636, "step": 19430 }, { "epoch": 5.98, "learning_rate": 7.357544969128385e-06, "loss": 0.1496, "step": 19431 }, { "epoch": 5.98, "learning_rate": 7.356583863724442e-06, "loss": 0.1691, "step": 19432 }, { "epoch": 5.98, "learning_rate": 7.355622784572063e-06, "loss": 0.2494, "step": 19433 }, { "epoch": 5.98, "learning_rate": 7.354661731680785e-06, "loss": 0.1299, "step": 19434 }, { "epoch": 5.98, "learning_rate": 7.353700705060154e-06, "loss": 0.1575, "step": 19435 }, { "epoch": 5.98, "learning_rate": 7.352739704719714e-06, "loss": 0.1347, "step": 19436 }, { "epoch": 5.98, "learning_rate": 7.3517787306690055e-06, "loss": 0.0868, "step": 19437 }, { "epoch": 5.98, "learning_rate": 7.350817782917578e-06, "loss": 0.1357, "step": 19438 }, { "epoch": 5.98, "learning_rate": 7.3498568614749735e-06, "loss": 0.1583, "step": 19439 }, { "epoch": 5.98, "learning_rate": 7.348895966350727e-06, "loss": 0.1456, "step": 19440 }, { "epoch": 5.98, "learning_rate": 7.3479350975543925e-06, "loss": 0.1924, "step": 19441 }, { "epoch": 5.98, "learning_rate": 7.3469742550955026e-06, "loss": 0.1569, "step": 19442 }, { "epoch": 5.98, "learning_rate": 7.3460134389836055e-06, "loss": 0.1486, "step": 19443 }, { "epoch": 5.98, "learning_rate": 7.345052649228241e-06, "loss": 0.1976, "step": 19444 }, { "epoch": 5.98, "learning_rate": 7.344091885838949e-06, "loss": 0.2286, "step": 19445 }, { "epoch": 5.98, "learning_rate": 7.343131148825275e-06, "loss": 0.1779, "step": 19446 }, { "epoch": 5.98, "learning_rate": 7.342170438196754e-06, "loss": 0.1859, "step": 19447 }, { "epoch": 5.98, "learning_rate": 7.341209753962933e-06, "loss": 0.1399, "step": 19448 }, { "epoch": 5.98, "learning_rate": 7.340249096133348e-06, "loss": 0.1859, "step": 19449 }, { "epoch": 5.98, "learning_rate": 7.33928846471754e-06, "loss": 0.1263, "step": 19450 }, { "epoch": 5.98, "learning_rate": 7.338327859725054e-06, "loss": 0.1432, "step": 19451 }, { "epoch": 5.99, "learning_rate": 7.3373672811654175e-06, "loss": 0.2227, "step": 19452 }, { "epoch": 5.99, "learning_rate": 7.336406729048187e-06, "loss": 0.2218, "step": 19453 }, { "epoch": 5.99, "learning_rate": 7.335446203382888e-06, "loss": 0.1281, "step": 19454 }, { "epoch": 5.99, "learning_rate": 7.3344857041790616e-06, "loss": 0.2282, "step": 19455 }, { "epoch": 5.99, "learning_rate": 7.333525231446255e-06, "loss": 0.1901, "step": 19456 }, { "epoch": 5.99, "learning_rate": 7.332564785193992e-06, "loss": 0.1723, "step": 19457 }, { "epoch": 5.99, "learning_rate": 7.331604365431826e-06, "loss": 0.3221, "step": 19458 }, { "epoch": 5.99, "learning_rate": 7.330643972169287e-06, "loss": 0.1392, "step": 19459 }, { "epoch": 5.99, "learning_rate": 7.32968360541591e-06, "loss": 0.27, "step": 19460 }, { "epoch": 5.99, "learning_rate": 7.328723265181239e-06, "loss": 0.1809, "step": 19461 }, { "epoch": 5.99, "learning_rate": 7.327762951474805e-06, "loss": 0.1436, "step": 19462 }, { "epoch": 5.99, "learning_rate": 7.326802664306152e-06, "loss": 0.1483, "step": 19463 }, { "epoch": 5.99, "learning_rate": 7.325842403684811e-06, "loss": 0.1253, "step": 19464 }, { "epoch": 5.99, "learning_rate": 7.324882169620315e-06, "loss": 0.2249, "step": 19465 }, { "epoch": 5.99, "learning_rate": 7.323921962122209e-06, "loss": 0.1939, "step": 19466 }, { "epoch": 5.99, "learning_rate": 7.322961781200022e-06, "loss": 0.1796, "step": 19467 }, { "epoch": 5.99, "learning_rate": 7.322001626863295e-06, "loss": 0.2173, "step": 19468 }, { "epoch": 5.99, "learning_rate": 7.321041499121558e-06, "loss": 0.2891, "step": 19469 }, { "epoch": 5.99, "learning_rate": 7.320081397984347e-06, "loss": 0.1693, "step": 19470 }, { "epoch": 5.99, "learning_rate": 7.319121323461198e-06, "loss": 0.2043, "step": 19471 }, { "epoch": 5.99, "learning_rate": 7.318161275561644e-06, "loss": 0.1626, "step": 19472 }, { "epoch": 5.99, "learning_rate": 7.317201254295222e-06, "loss": 0.0844, "step": 19473 }, { "epoch": 5.99, "learning_rate": 7.316241259671466e-06, "loss": 0.0613, "step": 19474 }, { "epoch": 5.99, "learning_rate": 7.315281291699904e-06, "loss": 0.2087, "step": 19475 }, { "epoch": 5.99, "learning_rate": 7.314321350390075e-06, "loss": 0.3039, "step": 19476 }, { "epoch": 5.99, "learning_rate": 7.313361435751507e-06, "loss": 0.1046, "step": 19477 }, { "epoch": 5.99, "learning_rate": 7.31240154779374e-06, "loss": 0.1438, "step": 19478 }, { "epoch": 5.99, "learning_rate": 7.311441686526304e-06, "loss": 0.2379, "step": 19479 }, { "epoch": 5.99, "learning_rate": 7.310481851958725e-06, "loss": 0.2362, "step": 19480 }, { "epoch": 5.99, "learning_rate": 7.309522044100544e-06, "loss": 0.1816, "step": 19481 }, { "epoch": 5.99, "learning_rate": 7.308562262961283e-06, "loss": 0.084, "step": 19482 }, { "epoch": 5.99, "learning_rate": 7.3076025085504845e-06, "loss": 0.1893, "step": 19483 }, { "epoch": 5.99, "learning_rate": 7.3066427808776754e-06, "loss": 0.2017, "step": 19484 }, { "epoch": 6.0, "learning_rate": 7.305683079952381e-06, "loss": 0.1571, "step": 19485 }, { "epoch": 6.0, "learning_rate": 7.30472340578414e-06, "loss": 0.1559, "step": 19486 }, { "epoch": 6.0, "learning_rate": 7.303763758382477e-06, "loss": 0.2252, "step": 19487 }, { "epoch": 6.0, "learning_rate": 7.302804137756927e-06, "loss": 0.1529, "step": 19488 }, { "epoch": 6.0, "learning_rate": 7.30184454391702e-06, "loss": 0.1382, "step": 19489 }, { "epoch": 6.0, "learning_rate": 7.300884976872277e-06, "loss": 0.1964, "step": 19490 }, { "epoch": 6.0, "learning_rate": 7.29992543663224e-06, "loss": 0.1438, "step": 19491 }, { "epoch": 6.0, "learning_rate": 7.298965923206427e-06, "loss": 0.1862, "step": 19492 }, { "epoch": 6.0, "learning_rate": 7.298006436604374e-06, "loss": 0.2138, "step": 19493 }, { "epoch": 6.0, "learning_rate": 7.2970469768356115e-06, "loss": 0.1844, "step": 19494 }, { "epoch": 6.0, "learning_rate": 7.2960875439096555e-06, "loss": 0.1275, "step": 19495 }, { "epoch": 6.0, "learning_rate": 7.295128137836048e-06, "loss": 0.2045, "step": 19496 }, { "epoch": 6.0, "learning_rate": 7.294168758624307e-06, "loss": 0.1723, "step": 19497 }, { "epoch": 6.0, "learning_rate": 7.293209406283967e-06, "loss": 0.1863, "step": 19498 }, { "epoch": 6.0, "learning_rate": 7.2922500808245525e-06, "loss": 0.2875, "step": 19499 }, { "epoch": 6.0, "learning_rate": 7.291290782255587e-06, "loss": 0.2119, "step": 19500 }, { "epoch": 6.0, "learning_rate": 7.2903315105866055e-06, "loss": 0.1691, "step": 19501 }, { "epoch": 6.0, "learning_rate": 7.289372265827122e-06, "loss": 0.1331, "step": 19502 }, { "epoch": 6.0, "learning_rate": 7.288413047986677e-06, "loss": 0.1041, "step": 19503 }, { "epoch": 6.0, "learning_rate": 7.287453857074789e-06, "loss": 0.0586, "step": 19504 }, { "epoch": 6.0, "learning_rate": 7.28649469310098e-06, "loss": 0.068, "step": 19505 }, { "epoch": 6.0, "learning_rate": 7.2855355560747835e-06, "loss": 0.0697, "step": 19506 }, { "epoch": 6.0, "learning_rate": 7.284576446005716e-06, "loss": 0.0944, "step": 19507 }, { "epoch": 6.0, "learning_rate": 7.2836173629033125e-06, "loss": 0.078, "step": 19508 }, { "epoch": 6.0, "learning_rate": 7.282658306777089e-06, "loss": 0.0461, "step": 19509 }, { "epoch": 6.0, "learning_rate": 7.2816992776365714e-06, "loss": 0.1112, "step": 19510 }, { "epoch": 6.0, "learning_rate": 7.2807402754912885e-06, "loss": 0.0927, "step": 19511 }, { "epoch": 6.0, "learning_rate": 7.279781300350758e-06, "loss": 0.1355, "step": 19512 }, { "epoch": 6.0, "learning_rate": 7.278822352224511e-06, "loss": 0.0727, "step": 19513 }, { "epoch": 6.0, "learning_rate": 7.277863431122062e-06, "loss": 0.0664, "step": 19514 }, { "epoch": 6.0, "learning_rate": 7.276904537052938e-06, "loss": 0.1016, "step": 19515 }, { "epoch": 6.0, "learning_rate": 7.275945670026662e-06, "loss": 0.0646, "step": 19516 }, { "epoch": 6.0, "learning_rate": 7.274986830052755e-06, "loss": 0.0653, "step": 19517 }, { "epoch": 6.01, "learning_rate": 7.274028017140744e-06, "loss": 0.1187, "step": 19518 }, { "epoch": 6.01, "learning_rate": 7.273069231300145e-06, "loss": 0.0672, "step": 19519 }, { "epoch": 6.01, "learning_rate": 7.272110472540479e-06, "loss": 0.096, "step": 19520 }, { "epoch": 6.01, "learning_rate": 7.271151740871274e-06, "loss": 0.0893, "step": 19521 }, { "epoch": 6.01, "learning_rate": 7.2701930363020435e-06, "loss": 0.0913, "step": 19522 }, { "epoch": 6.01, "learning_rate": 7.269234358842314e-06, "loss": 0.1042, "step": 19523 }, { "epoch": 6.01, "learning_rate": 7.268275708501606e-06, "loss": 0.0714, "step": 19524 }, { "epoch": 6.01, "learning_rate": 7.267317085289436e-06, "loss": 0.1204, "step": 19525 }, { "epoch": 6.01, "learning_rate": 7.266358489215326e-06, "loss": 0.1326, "step": 19526 }, { "epoch": 6.01, "learning_rate": 7.265399920288794e-06, "loss": 0.105, "step": 19527 }, { "epoch": 6.01, "learning_rate": 7.264441378519363e-06, "loss": 0.1062, "step": 19528 }, { "epoch": 6.01, "learning_rate": 7.2634828639165525e-06, "loss": 0.068, "step": 19529 }, { "epoch": 6.01, "learning_rate": 7.2625243764898746e-06, "loss": 0.0492, "step": 19530 }, { "epoch": 6.01, "learning_rate": 7.261565916248856e-06, "loss": 0.053, "step": 19531 }, { "epoch": 6.01, "learning_rate": 7.260607483203008e-06, "loss": 0.0526, "step": 19532 }, { "epoch": 6.01, "learning_rate": 7.259649077361854e-06, "loss": 0.1128, "step": 19533 }, { "epoch": 6.01, "learning_rate": 7.258690698734913e-06, "loss": 0.0935, "step": 19534 }, { "epoch": 6.01, "learning_rate": 7.257732347331698e-06, "loss": 0.0977, "step": 19535 }, { "epoch": 6.01, "learning_rate": 7.256774023161728e-06, "loss": 0.0482, "step": 19536 }, { "epoch": 6.01, "learning_rate": 7.2558157262345186e-06, "loss": 0.0666, "step": 19537 }, { "epoch": 6.01, "learning_rate": 7.254857456559589e-06, "loss": 0.1364, "step": 19538 }, { "epoch": 6.01, "learning_rate": 7.253899214146458e-06, "loss": 0.0757, "step": 19539 }, { "epoch": 6.01, "learning_rate": 7.252940999004633e-06, "loss": 0.0582, "step": 19540 }, { "epoch": 6.01, "learning_rate": 7.2519828111436405e-06, "loss": 0.1093, "step": 19541 }, { "epoch": 6.01, "learning_rate": 7.251024650572989e-06, "loss": 0.0969, "step": 19542 }, { "epoch": 6.01, "learning_rate": 7.250066517302197e-06, "loss": 0.1053, "step": 19543 }, { "epoch": 6.01, "learning_rate": 7.2491084113407816e-06, "loss": 0.1177, "step": 19544 }, { "epoch": 6.01, "learning_rate": 7.248150332698248e-06, "loss": 0.0445, "step": 19545 }, { "epoch": 6.01, "learning_rate": 7.2471922813841245e-06, "loss": 0.063, "step": 19546 }, { "epoch": 6.01, "learning_rate": 7.246234257407914e-06, "loss": 0.125, "step": 19547 }, { "epoch": 6.01, "learning_rate": 7.245276260779138e-06, "loss": 0.1179, "step": 19548 }, { "epoch": 6.01, "learning_rate": 7.244318291507308e-06, "loss": 0.1019, "step": 19549 }, { "epoch": 6.02, "learning_rate": 7.243360349601933e-06, "loss": 0.0854, "step": 19550 }, { "epoch": 6.02, "learning_rate": 7.242402435072537e-06, "loss": 0.0493, "step": 19551 }, { "epoch": 6.02, "learning_rate": 7.241444547928619e-06, "loss": 0.0583, "step": 19552 }, { "epoch": 6.02, "learning_rate": 7.240486688179704e-06, "loss": 0.1225, "step": 19553 }, { "epoch": 6.02, "learning_rate": 7.239528855835299e-06, "loss": 0.141, "step": 19554 }, { "epoch": 6.02, "learning_rate": 7.238571050904914e-06, "loss": 0.0689, "step": 19555 }, { "epoch": 6.02, "learning_rate": 7.2376132733980675e-06, "loss": 0.0582, "step": 19556 }, { "epoch": 6.02, "learning_rate": 7.2366555233242605e-06, "loss": 0.1396, "step": 19557 }, { "epoch": 6.02, "learning_rate": 7.235697800693019e-06, "loss": 0.0874, "step": 19558 }, { "epoch": 6.02, "learning_rate": 7.234740105513842e-06, "loss": 0.1257, "step": 19559 }, { "epoch": 6.02, "learning_rate": 7.233782437796243e-06, "loss": 0.1138, "step": 19560 }, { "epoch": 6.02, "learning_rate": 7.232824797549736e-06, "loss": 0.0727, "step": 19561 }, { "epoch": 6.02, "learning_rate": 7.231867184783826e-06, "loss": 0.1292, "step": 19562 }, { "epoch": 6.02, "learning_rate": 7.230909599508031e-06, "loss": 0.0877, "step": 19563 }, { "epoch": 6.02, "learning_rate": 7.229952041731854e-06, "loss": 0.0552, "step": 19564 }, { "epoch": 6.02, "learning_rate": 7.228994511464804e-06, "loss": 0.1012, "step": 19565 }, { "epoch": 6.02, "learning_rate": 7.228037008716393e-06, "loss": 0.1159, "step": 19566 }, { "epoch": 6.02, "learning_rate": 7.227079533496128e-06, "loss": 0.1206, "step": 19567 }, { "epoch": 6.02, "learning_rate": 7.226122085813524e-06, "loss": 0.0945, "step": 19568 }, { "epoch": 6.02, "learning_rate": 7.225164665678082e-06, "loss": 0.1042, "step": 19569 }, { "epoch": 6.02, "learning_rate": 7.2242072730993085e-06, "loss": 0.0945, "step": 19570 }, { "epoch": 6.02, "learning_rate": 7.223249908086718e-06, "loss": 0.0941, "step": 19571 }, { "epoch": 6.02, "learning_rate": 7.222292570649813e-06, "loss": 0.0733, "step": 19572 }, { "epoch": 6.02, "learning_rate": 7.221335260798107e-06, "loss": 0.0461, "step": 19573 }, { "epoch": 6.02, "learning_rate": 7.2203779785411e-06, "loss": 0.0772, "step": 19574 }, { "epoch": 6.02, "learning_rate": 7.219420723888301e-06, "loss": 0.0497, "step": 19575 }, { "epoch": 6.02, "learning_rate": 7.218463496849218e-06, "loss": 0.0692, "step": 19576 }, { "epoch": 6.02, "learning_rate": 7.217506297433353e-06, "loss": 0.1073, "step": 19577 }, { "epoch": 6.02, "learning_rate": 7.2165491256502186e-06, "loss": 0.116, "step": 19578 }, { "epoch": 6.02, "learning_rate": 7.215591981509318e-06, "loss": 0.0843, "step": 19579 }, { "epoch": 6.02, "learning_rate": 7.21463486502015e-06, "loss": 0.0671, "step": 19580 }, { "epoch": 6.02, "learning_rate": 7.213677776192228e-06, "loss": 0.1154, "step": 19581 }, { "epoch": 6.02, "learning_rate": 7.212720715035052e-06, "loss": 0.0547, "step": 19582 }, { "epoch": 6.03, "learning_rate": 7.21176368155813e-06, "loss": 0.0935, "step": 19583 }, { "epoch": 6.03, "learning_rate": 7.210806675770966e-06, "loss": 0.0867, "step": 19584 }, { "epoch": 6.03, "learning_rate": 7.209849697683058e-06, "loss": 0.0999, "step": 19585 }, { "epoch": 6.03, "learning_rate": 7.208892747303917e-06, "loss": 0.071, "step": 19586 }, { "epoch": 6.03, "learning_rate": 7.20793582464304e-06, "loss": 0.1345, "step": 19587 }, { "epoch": 6.03, "learning_rate": 7.2069789297099355e-06, "loss": 0.0948, "step": 19588 }, { "epoch": 6.03, "learning_rate": 7.206022062514108e-06, "loss": 0.0566, "step": 19589 }, { "epoch": 6.03, "learning_rate": 7.20506522306505e-06, "loss": 0.0885, "step": 19590 }, { "epoch": 6.03, "learning_rate": 7.204108411372275e-06, "loss": 0.0878, "step": 19591 }, { "epoch": 6.03, "learning_rate": 7.2031516274452775e-06, "loss": 0.1355, "step": 19592 }, { "epoch": 6.03, "learning_rate": 7.202194871293564e-06, "loss": 0.1137, "step": 19593 }, { "epoch": 6.03, "learning_rate": 7.201238142926634e-06, "loss": 0.09, "step": 19594 }, { "epoch": 6.03, "learning_rate": 7.200281442353985e-06, "loss": 0.0653, "step": 19595 }, { "epoch": 6.03, "learning_rate": 7.199324769585126e-06, "loss": 0.085, "step": 19596 }, { "epoch": 6.03, "learning_rate": 7.198368124629551e-06, "loss": 0.0875, "step": 19597 }, { "epoch": 6.03, "learning_rate": 7.197411507496763e-06, "loss": 0.0848, "step": 19598 }, { "epoch": 6.03, "learning_rate": 7.1964549181962615e-06, "loss": 0.0932, "step": 19599 }, { "epoch": 6.03, "learning_rate": 7.195498356737545e-06, "loss": 0.1279, "step": 19600 }, { "epoch": 6.03, "learning_rate": 7.194541823130119e-06, "loss": 0.0813, "step": 19601 }, { "epoch": 6.03, "learning_rate": 7.193585317383474e-06, "loss": 0.1045, "step": 19602 }, { "epoch": 6.03, "learning_rate": 7.192628839507116e-06, "loss": 0.0791, "step": 19603 }, { "epoch": 6.03, "learning_rate": 7.1916723895105404e-06, "loss": 0.1286, "step": 19604 }, { "epoch": 6.03, "learning_rate": 7.190715967403243e-06, "loss": 0.1159, "step": 19605 }, { "epoch": 6.03, "learning_rate": 7.18975957319473e-06, "loss": 0.0779, "step": 19606 }, { "epoch": 6.03, "learning_rate": 7.188803206894488e-06, "loss": 0.0694, "step": 19607 }, { "epoch": 6.03, "learning_rate": 7.187846868512028e-06, "loss": 0.0994, "step": 19608 }, { "epoch": 6.03, "learning_rate": 7.186890558056836e-06, "loss": 0.075, "step": 19609 }, { "epoch": 6.03, "learning_rate": 7.185934275538416e-06, "loss": 0.0853, "step": 19610 }, { "epoch": 6.03, "learning_rate": 7.1849780209662614e-06, "loss": 0.0834, "step": 19611 }, { "epoch": 6.03, "learning_rate": 7.184021794349867e-06, "loss": 0.1271, "step": 19612 }, { "epoch": 6.03, "learning_rate": 7.1830655956987375e-06, "loss": 0.0814, "step": 19613 }, { "epoch": 6.03, "learning_rate": 7.182109425022357e-06, "loss": 0.2295, "step": 19614 }, { "epoch": 6.04, "learning_rate": 7.181153282330229e-06, "loss": 0.117, "step": 19615 }, { "epoch": 6.04, "learning_rate": 7.180197167631846e-06, "loss": 0.0918, "step": 19616 }, { "epoch": 6.04, "learning_rate": 7.179241080936703e-06, "loss": 0.1501, "step": 19617 }, { "epoch": 6.04, "learning_rate": 7.178285022254301e-06, "loss": 0.1205, "step": 19618 }, { "epoch": 6.04, "learning_rate": 7.177328991594121e-06, "loss": 0.0825, "step": 19619 }, { "epoch": 6.04, "learning_rate": 7.176372988965673e-06, "loss": 0.0944, "step": 19620 }, { "epoch": 6.04, "learning_rate": 7.175417014378441e-06, "loss": 0.1006, "step": 19621 }, { "epoch": 6.04, "learning_rate": 7.17446106784192e-06, "loss": 0.0903, "step": 19622 }, { "epoch": 6.04, "learning_rate": 7.173505149365607e-06, "loss": 0.0544, "step": 19623 }, { "epoch": 6.04, "learning_rate": 7.172549258958989e-06, "loss": 0.1126, "step": 19624 }, { "epoch": 6.04, "learning_rate": 7.1715933966315675e-06, "loss": 0.1026, "step": 19625 }, { "epoch": 6.04, "learning_rate": 7.1706375623928286e-06, "loss": 0.047, "step": 19626 }, { "epoch": 6.04, "learning_rate": 7.169681756252265e-06, "loss": 0.0972, "step": 19627 }, { "epoch": 6.04, "learning_rate": 7.168725978219371e-06, "loss": 0.0867, "step": 19628 }, { "epoch": 6.04, "learning_rate": 7.1677702283036344e-06, "loss": 0.1004, "step": 19629 }, { "epoch": 6.04, "learning_rate": 7.166814506514555e-06, "loss": 0.0784, "step": 19630 }, { "epoch": 6.04, "learning_rate": 7.165858812861617e-06, "loss": 0.0679, "step": 19631 }, { "epoch": 6.04, "learning_rate": 7.164903147354312e-06, "loss": 0.0736, "step": 19632 }, { "epoch": 6.04, "learning_rate": 7.163947510002132e-06, "loss": 0.0853, "step": 19633 }, { "epoch": 6.04, "learning_rate": 7.162991900814564e-06, "loss": 0.0547, "step": 19634 }, { "epoch": 6.04, "learning_rate": 7.162036319801106e-06, "loss": 0.1286, "step": 19635 }, { "epoch": 6.04, "learning_rate": 7.1610807669712425e-06, "loss": 0.0828, "step": 19636 }, { "epoch": 6.04, "learning_rate": 7.1601252423344604e-06, "loss": 0.0597, "step": 19637 }, { "epoch": 6.04, "learning_rate": 7.159169745900254e-06, "loss": 0.0823, "step": 19638 }, { "epoch": 6.04, "learning_rate": 7.158214277678109e-06, "loss": 0.1069, "step": 19639 }, { "epoch": 6.04, "learning_rate": 7.157258837677514e-06, "loss": 0.0902, "step": 19640 }, { "epoch": 6.04, "learning_rate": 7.156303425907965e-06, "loss": 0.1078, "step": 19641 }, { "epoch": 6.04, "learning_rate": 7.155348042378937e-06, "loss": 0.1101, "step": 19642 }, { "epoch": 6.04, "learning_rate": 7.154392687099927e-06, "loss": 0.1188, "step": 19643 }, { "epoch": 6.04, "learning_rate": 7.15343736008042e-06, "loss": 0.0879, "step": 19644 }, { "epoch": 6.04, "learning_rate": 7.152482061329903e-06, "loss": 0.0591, "step": 19645 }, { "epoch": 6.04, "learning_rate": 7.1515267908578665e-06, "loss": 0.1032, "step": 19646 }, { "epoch": 6.04, "learning_rate": 7.1505715486737905e-06, "loss": 0.0777, "step": 19647 }, { "epoch": 6.05, "learning_rate": 7.1496163347871675e-06, "loss": 0.1057, "step": 19648 }, { "epoch": 6.05, "learning_rate": 7.1486611492074784e-06, "loss": 0.0803, "step": 19649 }, { "epoch": 6.05, "learning_rate": 7.1477059919442135e-06, "loss": 0.1082, "step": 19650 }, { "epoch": 6.05, "learning_rate": 7.14675086300686e-06, "loss": 0.0736, "step": 19651 }, { "epoch": 6.05, "learning_rate": 7.145795762404896e-06, "loss": 0.0867, "step": 19652 }, { "epoch": 6.05, "learning_rate": 7.144840690147812e-06, "loss": 0.0832, "step": 19653 }, { "epoch": 6.05, "learning_rate": 7.14388564624509e-06, "loss": 0.1396, "step": 19654 }, { "epoch": 6.05, "learning_rate": 7.142930630706217e-06, "loss": 0.122, "step": 19655 }, { "epoch": 6.05, "learning_rate": 7.141975643540679e-06, "loss": 0.0693, "step": 19656 }, { "epoch": 6.05, "learning_rate": 7.14102068475795e-06, "loss": 0.075, "step": 19657 }, { "epoch": 6.05, "learning_rate": 7.1400657543675264e-06, "loss": 0.0826, "step": 19658 }, { "epoch": 6.05, "learning_rate": 7.139110852378883e-06, "loss": 0.0594, "step": 19659 }, { "epoch": 6.05, "learning_rate": 7.1381559788015065e-06, "loss": 0.0935, "step": 19660 }, { "epoch": 6.05, "learning_rate": 7.137201133644877e-06, "loss": 0.0752, "step": 19661 }, { "epoch": 6.05, "learning_rate": 7.1362463169184804e-06, "loss": 0.054, "step": 19662 }, { "epoch": 6.05, "learning_rate": 7.135291528631798e-06, "loss": 0.1633, "step": 19663 }, { "epoch": 6.05, "learning_rate": 7.1343367687943085e-06, "loss": 0.0846, "step": 19664 }, { "epoch": 6.05, "learning_rate": 7.133382037415497e-06, "loss": 0.041, "step": 19665 }, { "epoch": 6.05, "learning_rate": 7.132427334504846e-06, "loss": 0.075, "step": 19666 }, { "epoch": 6.05, "learning_rate": 7.13147266007183e-06, "loss": 0.0863, "step": 19667 }, { "epoch": 6.05, "learning_rate": 7.13051801412594e-06, "loss": 0.1137, "step": 19668 }, { "epoch": 6.05, "learning_rate": 7.129563396676643e-06, "loss": 0.077, "step": 19669 }, { "epoch": 6.05, "learning_rate": 7.128608807733434e-06, "loss": 0.0922, "step": 19670 }, { "epoch": 6.05, "learning_rate": 7.127654247305783e-06, "loss": 0.0767, "step": 19671 }, { "epoch": 6.05, "learning_rate": 7.126699715403171e-06, "loss": 0.0253, "step": 19672 }, { "epoch": 6.05, "learning_rate": 7.125745212035084e-06, "loss": 0.1118, "step": 19673 }, { "epoch": 6.05, "learning_rate": 7.124790737210989e-06, "loss": 0.1141, "step": 19674 }, { "epoch": 6.05, "learning_rate": 7.123836290940378e-06, "loss": 0.0883, "step": 19675 }, { "epoch": 6.05, "learning_rate": 7.122881873232721e-06, "loss": 0.0936, "step": 19676 }, { "epoch": 6.05, "learning_rate": 7.121927484097497e-06, "loss": 0.0976, "step": 19677 }, { "epoch": 6.05, "learning_rate": 7.120973123544188e-06, "loss": 0.0315, "step": 19678 }, { "epoch": 6.05, "learning_rate": 7.120018791582266e-06, "loss": 0.1621, "step": 19679 }, { "epoch": 6.06, "learning_rate": 7.119064488221217e-06, "loss": 0.1001, "step": 19680 }, { "epoch": 6.06, "learning_rate": 7.11811021347051e-06, "loss": 0.1221, "step": 19681 }, { "epoch": 6.06, "learning_rate": 7.117155967339623e-06, "loss": 0.1162, "step": 19682 }, { "epoch": 6.06, "learning_rate": 7.116201749838037e-06, "loss": 0.0743, "step": 19683 }, { "epoch": 6.06, "learning_rate": 7.115247560975221e-06, "loss": 0.121, "step": 19684 }, { "epoch": 6.06, "learning_rate": 7.114293400760661e-06, "loss": 0.0984, "step": 19685 }, { "epoch": 6.06, "learning_rate": 7.113339269203825e-06, "loss": 0.0831, "step": 19686 }, { "epoch": 6.06, "learning_rate": 7.112385166314189e-06, "loss": 0.1103, "step": 19687 }, { "epoch": 6.06, "learning_rate": 7.1114310921012305e-06, "loss": 0.0813, "step": 19688 }, { "epoch": 6.06, "learning_rate": 7.110477046574422e-06, "loss": 0.0809, "step": 19689 }, { "epoch": 6.06, "learning_rate": 7.109523029743242e-06, "loss": 0.0827, "step": 19690 }, { "epoch": 6.06, "learning_rate": 7.108569041617164e-06, "loss": 0.1342, "step": 19691 }, { "epoch": 6.06, "learning_rate": 7.107615082205654e-06, "loss": 0.115, "step": 19692 }, { "epoch": 6.06, "learning_rate": 7.106661151518197e-06, "loss": 0.0951, "step": 19693 }, { "epoch": 6.06, "learning_rate": 7.105707249564258e-06, "loss": 0.089, "step": 19694 }, { "epoch": 6.06, "learning_rate": 7.104753376353315e-06, "loss": 0.1399, "step": 19695 }, { "epoch": 6.06, "learning_rate": 7.103799531894843e-06, "loss": 0.0726, "step": 19696 }, { "epoch": 6.06, "learning_rate": 7.102845716198306e-06, "loss": 0.0727, "step": 19697 }, { "epoch": 6.06, "learning_rate": 7.101891929273182e-06, "loss": 0.0835, "step": 19698 }, { "epoch": 6.06, "learning_rate": 7.100938171128942e-06, "loss": 0.1052, "step": 19699 }, { "epoch": 6.06, "learning_rate": 7.0999844417750584e-06, "loss": 0.1476, "step": 19700 }, { "epoch": 6.06, "learning_rate": 7.099030741221005e-06, "loss": 0.1032, "step": 19701 }, { "epoch": 6.06, "learning_rate": 7.0980770694762455e-06, "loss": 0.0596, "step": 19702 }, { "epoch": 6.06, "learning_rate": 7.097123426550257e-06, "loss": 0.0668, "step": 19703 }, { "epoch": 6.06, "learning_rate": 7.096169812452507e-06, "loss": 0.0801, "step": 19704 }, { "epoch": 6.06, "learning_rate": 7.095216227192467e-06, "loss": 0.0895, "step": 19705 }, { "epoch": 6.06, "learning_rate": 7.094262670779611e-06, "loss": 0.0987, "step": 19706 }, { "epoch": 6.06, "learning_rate": 7.093309143223398e-06, "loss": 0.0879, "step": 19707 }, { "epoch": 6.06, "learning_rate": 7.092355644533311e-06, "loss": 0.1065, "step": 19708 }, { "epoch": 6.06, "learning_rate": 7.091402174718807e-06, "loss": 0.086, "step": 19709 }, { "epoch": 6.06, "learning_rate": 7.090448733789362e-06, "loss": 0.1194, "step": 19710 }, { "epoch": 6.06, "learning_rate": 7.089495321754444e-06, "loss": 0.0631, "step": 19711 }, { "epoch": 6.06, "learning_rate": 7.0885419386235165e-06, "loss": 0.0764, "step": 19712 }, { "epoch": 6.07, "learning_rate": 7.0875885844060555e-06, "loss": 0.0292, "step": 19713 }, { "epoch": 6.07, "learning_rate": 7.08663525911152e-06, "loss": 0.1057, "step": 19714 }, { "epoch": 6.07, "learning_rate": 7.085681962749382e-06, "loss": 0.0735, "step": 19715 }, { "epoch": 6.07, "learning_rate": 7.084728695329111e-06, "loss": 0.0947, "step": 19716 }, { "epoch": 6.07, "learning_rate": 7.083775456860166e-06, "loss": 0.068, "step": 19717 }, { "epoch": 6.07, "learning_rate": 7.082822247352024e-06, "loss": 0.0821, "step": 19718 }, { "epoch": 6.07, "learning_rate": 7.081869066814138e-06, "loss": 0.0979, "step": 19719 }, { "epoch": 6.07, "learning_rate": 7.080915915255989e-06, "loss": 0.1176, "step": 19720 }, { "epoch": 6.07, "learning_rate": 7.079962792687033e-06, "loss": 0.0973, "step": 19721 }, { "epoch": 6.07, "learning_rate": 7.079009699116736e-06, "loss": 0.0704, "step": 19722 }, { "epoch": 6.07, "learning_rate": 7.078056634554569e-06, "loss": 0.124, "step": 19723 }, { "epoch": 6.07, "learning_rate": 7.077103599009984e-06, "loss": 0.056, "step": 19724 }, { "epoch": 6.07, "learning_rate": 7.076150592492464e-06, "loss": 0.0751, "step": 19725 }, { "epoch": 6.07, "learning_rate": 7.075197615011458e-06, "loss": 0.0784, "step": 19726 }, { "epoch": 6.07, "learning_rate": 7.0742446665764356e-06, "loss": 0.0664, "step": 19727 }, { "epoch": 6.07, "learning_rate": 7.07329174719686e-06, "loss": 0.0725, "step": 19728 }, { "epoch": 6.07, "learning_rate": 7.072338856882194e-06, "loss": 0.0809, "step": 19729 }, { "epoch": 6.07, "learning_rate": 7.071385995641905e-06, "loss": 0.0735, "step": 19730 }, { "epoch": 6.07, "learning_rate": 7.070433163485451e-06, "loss": 0.0591, "step": 19731 }, { "epoch": 6.07, "learning_rate": 7.069480360422291e-06, "loss": 0.1103, "step": 19732 }, { "epoch": 6.07, "learning_rate": 7.068527586461897e-06, "loss": 0.1191, "step": 19733 }, { "epoch": 6.07, "learning_rate": 7.067574841613722e-06, "loss": 0.0474, "step": 19734 }, { "epoch": 6.07, "learning_rate": 7.066622125887237e-06, "loss": 0.1042, "step": 19735 }, { "epoch": 6.07, "learning_rate": 7.065669439291895e-06, "loss": 0.0743, "step": 19736 }, { "epoch": 6.07, "learning_rate": 7.0647167818371575e-06, "loss": 0.0803, "step": 19737 }, { "epoch": 6.07, "learning_rate": 7.063764153532489e-06, "loss": 0.0668, "step": 19738 }, { "epoch": 6.07, "learning_rate": 7.062811554387348e-06, "loss": 0.0692, "step": 19739 }, { "epoch": 6.07, "learning_rate": 7.061858984411196e-06, "loss": 0.0815, "step": 19740 }, { "epoch": 6.07, "learning_rate": 7.060906443613495e-06, "loss": 0.0782, "step": 19741 }, { "epoch": 6.07, "learning_rate": 7.059953932003697e-06, "loss": 0.0575, "step": 19742 }, { "epoch": 6.07, "learning_rate": 7.059001449591268e-06, "loss": 0.0813, "step": 19743 }, { "epoch": 6.07, "learning_rate": 7.0580489963856646e-06, "loss": 0.1116, "step": 19744 }, { "epoch": 6.08, "learning_rate": 7.057096572396347e-06, "loss": 0.149, "step": 19745 }, { "epoch": 6.08, "learning_rate": 7.056144177632774e-06, "loss": 0.1563, "step": 19746 }, { "epoch": 6.08, "learning_rate": 7.0551918121044e-06, "loss": 0.0928, "step": 19747 }, { "epoch": 6.08, "learning_rate": 7.0542394758206854e-06, "loss": 0.0516, "step": 19748 }, { "epoch": 6.08, "learning_rate": 7.053287168791087e-06, "loss": 0.0715, "step": 19749 }, { "epoch": 6.08, "learning_rate": 7.052334891025065e-06, "loss": 0.0635, "step": 19750 }, { "epoch": 6.08, "learning_rate": 7.051382642532077e-06, "loss": 0.0304, "step": 19751 }, { "epoch": 6.08, "learning_rate": 7.05043042332157e-06, "loss": 0.0869, "step": 19752 }, { "epoch": 6.08, "learning_rate": 7.0494782334030125e-06, "loss": 0.1149, "step": 19753 }, { "epoch": 6.08, "learning_rate": 7.048526072785851e-06, "loss": 0.0919, "step": 19754 }, { "epoch": 6.08, "learning_rate": 7.047573941479549e-06, "loss": 0.1126, "step": 19755 }, { "epoch": 6.08, "learning_rate": 7.046621839493562e-06, "loss": 0.105, "step": 19756 }, { "epoch": 6.08, "learning_rate": 7.045669766837333e-06, "loss": 0.0851, "step": 19757 }, { "epoch": 6.08, "learning_rate": 7.044717723520334e-06, "loss": 0.0717, "step": 19758 }, { "epoch": 6.08, "learning_rate": 7.0437657095520065e-06, "loss": 0.0448, "step": 19759 }, { "epoch": 6.08, "learning_rate": 7.0428137249418125e-06, "loss": 0.1328, "step": 19760 }, { "epoch": 6.08, "learning_rate": 7.041861769699205e-06, "loss": 0.0649, "step": 19761 }, { "epoch": 6.08, "learning_rate": 7.040909843833632e-06, "loss": 0.0949, "step": 19762 }, { "epoch": 6.08, "learning_rate": 7.039957947354557e-06, "loss": 0.0792, "step": 19763 }, { "epoch": 6.08, "learning_rate": 7.039006080271425e-06, "loss": 0.0791, "step": 19764 }, { "epoch": 6.08, "learning_rate": 7.038054242593691e-06, "loss": 0.0995, "step": 19765 }, { "epoch": 6.08, "learning_rate": 7.037102434330809e-06, "loss": 0.0571, "step": 19766 }, { "epoch": 6.08, "learning_rate": 7.0361506554922285e-06, "loss": 0.0653, "step": 19767 }, { "epoch": 6.08, "learning_rate": 7.035198906087408e-06, "loss": 0.0822, "step": 19768 }, { "epoch": 6.08, "learning_rate": 7.034247186125789e-06, "loss": 0.0807, "step": 19769 }, { "epoch": 6.08, "learning_rate": 7.033295495616834e-06, "loss": 0.0539, "step": 19770 }, { "epoch": 6.08, "learning_rate": 7.032343834569987e-06, "loss": 0.1137, "step": 19771 }, { "epoch": 6.08, "learning_rate": 7.031392202994698e-06, "loss": 0.0901, "step": 19772 }, { "epoch": 6.08, "learning_rate": 7.030440600900425e-06, "loss": 0.0868, "step": 19773 }, { "epoch": 6.08, "learning_rate": 7.029489028296609e-06, "loss": 0.0913, "step": 19774 }, { "epoch": 6.08, "learning_rate": 7.02853748519271e-06, "loss": 0.1226, "step": 19775 }, { "epoch": 6.08, "learning_rate": 7.02758597159817e-06, "loss": 0.0761, "step": 19776 }, { "epoch": 6.08, "learning_rate": 7.026634487522439e-06, "loss": 0.0436, "step": 19777 }, { "epoch": 6.09, "learning_rate": 7.0256830329749694e-06, "loss": 0.0705, "step": 19778 }, { "epoch": 6.09, "learning_rate": 7.0247316079652074e-06, "loss": 0.1311, "step": 19779 }, { "epoch": 6.09, "learning_rate": 7.023780212502604e-06, "loss": 0.075, "step": 19780 }, { "epoch": 6.09, "learning_rate": 7.022828846596607e-06, "loss": 0.123, "step": 19781 }, { "epoch": 6.09, "learning_rate": 7.021877510256661e-06, "loss": 0.0983, "step": 19782 }, { "epoch": 6.09, "learning_rate": 7.020926203492218e-06, "loss": 0.1145, "step": 19783 }, { "epoch": 6.09, "learning_rate": 7.019974926312721e-06, "loss": 0.0902, "step": 19784 }, { "epoch": 6.09, "learning_rate": 7.019023678727622e-06, "loss": 0.0618, "step": 19785 }, { "epoch": 6.09, "learning_rate": 7.018072460746365e-06, "loss": 0.1092, "step": 19786 }, { "epoch": 6.09, "learning_rate": 7.0171212723783935e-06, "loss": 0.1393, "step": 19787 }, { "epoch": 6.09, "learning_rate": 7.016170113633161e-06, "loss": 0.0817, "step": 19788 }, { "epoch": 6.09, "learning_rate": 7.015218984520105e-06, "loss": 0.0676, "step": 19789 }, { "epoch": 6.09, "learning_rate": 7.0142678850486775e-06, "loss": 0.0557, "step": 19790 }, { "epoch": 6.09, "learning_rate": 7.0133168152283245e-06, "loss": 0.0983, "step": 19791 }, { "epoch": 6.09, "learning_rate": 7.0123657750684835e-06, "loss": 0.0771, "step": 19792 }, { "epoch": 6.09, "learning_rate": 7.011414764578607e-06, "loss": 0.0491, "step": 19793 }, { "epoch": 6.09, "learning_rate": 7.010463783768133e-06, "loss": 0.0572, "step": 19794 }, { "epoch": 6.09, "learning_rate": 7.009512832646513e-06, "loss": 0.1409, "step": 19795 }, { "epoch": 6.09, "learning_rate": 7.008561911223186e-06, "loss": 0.0847, "step": 19796 }, { "epoch": 6.09, "learning_rate": 7.007611019507593e-06, "loss": 0.0679, "step": 19797 }, { "epoch": 6.09, "learning_rate": 7.006660157509183e-06, "loss": 0.0645, "step": 19798 }, { "epoch": 6.09, "learning_rate": 7.005709325237394e-06, "loss": 0.08, "step": 19799 }, { "epoch": 6.09, "learning_rate": 7.004758522701673e-06, "loss": 0.0883, "step": 19800 }, { "epoch": 6.09, "learning_rate": 7.003807749911462e-06, "loss": 0.1139, "step": 19801 }, { "epoch": 6.09, "learning_rate": 7.002857006876198e-06, "loss": 0.0767, "step": 19802 }, { "epoch": 6.09, "learning_rate": 7.001906293605329e-06, "loss": 0.1447, "step": 19803 }, { "epoch": 6.09, "learning_rate": 7.000955610108291e-06, "loss": 0.0842, "step": 19804 }, { "epoch": 6.09, "learning_rate": 7.000004956394528e-06, "loss": 0.1187, "step": 19805 }, { "epoch": 6.09, "learning_rate": 6.999054332473481e-06, "loss": 0.0476, "step": 19806 }, { "epoch": 6.09, "learning_rate": 6.9981037383545915e-06, "loss": 0.0784, "step": 19807 }, { "epoch": 6.09, "learning_rate": 6.9971531740472995e-06, "loss": 0.0931, "step": 19808 }, { "epoch": 6.09, "learning_rate": 6.9962026395610416e-06, "loss": 0.1078, "step": 19809 }, { "epoch": 6.1, "learning_rate": 6.995252134905261e-06, "loss": 0.0694, "step": 19810 }, { "epoch": 6.1, "learning_rate": 6.994301660089394e-06, "loss": 0.0779, "step": 19811 }, { "epoch": 6.1, "learning_rate": 6.993351215122883e-06, "loss": 0.1297, "step": 19812 }, { "epoch": 6.1, "learning_rate": 6.992400800015169e-06, "loss": 0.0835, "step": 19813 }, { "epoch": 6.1, "learning_rate": 6.991450414775682e-06, "loss": 0.0769, "step": 19814 }, { "epoch": 6.1, "learning_rate": 6.990500059413868e-06, "loss": 0.0684, "step": 19815 }, { "epoch": 6.1, "learning_rate": 6.989549733939158e-06, "loss": 0.1387, "step": 19816 }, { "epoch": 6.1, "learning_rate": 6.988599438360998e-06, "loss": 0.0952, "step": 19817 }, { "epoch": 6.1, "learning_rate": 6.9876491726888215e-06, "loss": 0.1098, "step": 19818 }, { "epoch": 6.1, "learning_rate": 6.986698936932059e-06, "loss": 0.1075, "step": 19819 }, { "epoch": 6.1, "learning_rate": 6.9857487311001605e-06, "loss": 0.0924, "step": 19820 }, { "epoch": 6.1, "learning_rate": 6.98479855520255e-06, "loss": 0.061, "step": 19821 }, { "epoch": 6.1, "learning_rate": 6.983848409248672e-06, "loss": 0.1161, "step": 19822 }, { "epoch": 6.1, "learning_rate": 6.98289829324796e-06, "loss": 0.1603, "step": 19823 }, { "epoch": 6.1, "learning_rate": 6.981948207209842e-06, "loss": 0.107, "step": 19824 }, { "epoch": 6.1, "learning_rate": 6.980998151143766e-06, "loss": 0.0485, "step": 19825 }, { "epoch": 6.1, "learning_rate": 6.980048125059158e-06, "loss": 0.079, "step": 19826 }, { "epoch": 6.1, "learning_rate": 6.9790981289654556e-06, "loss": 0.0743, "step": 19827 }, { "epoch": 6.1, "learning_rate": 6.978148162872095e-06, "loss": 0.1053, "step": 19828 }, { "epoch": 6.1, "learning_rate": 6.977198226788504e-06, "loss": 0.1087, "step": 19829 }, { "epoch": 6.1, "learning_rate": 6.976248320724126e-06, "loss": 0.0619, "step": 19830 }, { "epoch": 6.1, "learning_rate": 6.975298444688385e-06, "loss": 0.0826, "step": 19831 }, { "epoch": 6.1, "learning_rate": 6.974348598690719e-06, "loss": 0.0572, "step": 19832 }, { "epoch": 6.1, "learning_rate": 6.973398782740561e-06, "loss": 0.0721, "step": 19833 }, { "epoch": 6.1, "learning_rate": 6.972448996847341e-06, "loss": 0.0921, "step": 19834 }, { "epoch": 6.1, "learning_rate": 6.971499241020495e-06, "loss": 0.0913, "step": 19835 }, { "epoch": 6.1, "learning_rate": 6.970549515269447e-06, "loss": 0.0735, "step": 19836 }, { "epoch": 6.1, "learning_rate": 6.969599819603641e-06, "loss": 0.0692, "step": 19837 }, { "epoch": 6.1, "learning_rate": 6.968650154032499e-06, "loss": 0.0799, "step": 19838 }, { "epoch": 6.1, "learning_rate": 6.967700518565454e-06, "loss": 0.0677, "step": 19839 }, { "epoch": 6.1, "learning_rate": 6.966750913211941e-06, "loss": 0.1339, "step": 19840 }, { "epoch": 6.1, "learning_rate": 6.96580133798138e-06, "loss": 0.0795, "step": 19841 }, { "epoch": 6.1, "learning_rate": 6.964851792883216e-06, "loss": 0.1097, "step": 19842 }, { "epoch": 6.11, "learning_rate": 6.9639022779268675e-06, "loss": 0.0585, "step": 19843 }, { "epoch": 6.11, "learning_rate": 6.962952793121765e-06, "loss": 0.072, "step": 19844 }, { "epoch": 6.11, "learning_rate": 6.962003338477343e-06, "loss": 0.0955, "step": 19845 }, { "epoch": 6.11, "learning_rate": 6.961053914003026e-06, "loss": 0.076, "step": 19846 }, { "epoch": 6.11, "learning_rate": 6.960104519708249e-06, "loss": 0.1101, "step": 19847 }, { "epoch": 6.11, "learning_rate": 6.959155155602433e-06, "loss": 0.0613, "step": 19848 }, { "epoch": 6.11, "learning_rate": 6.958205821695006e-06, "loss": 0.1372, "step": 19849 }, { "epoch": 6.11, "learning_rate": 6.957256517995401e-06, "loss": 0.1071, "step": 19850 }, { "epoch": 6.11, "learning_rate": 6.956307244513043e-06, "loss": 0.0634, "step": 19851 }, { "epoch": 6.11, "learning_rate": 6.955358001257362e-06, "loss": 0.1366, "step": 19852 }, { "epoch": 6.11, "learning_rate": 6.954408788237779e-06, "loss": 0.0646, "step": 19853 }, { "epoch": 6.11, "learning_rate": 6.953459605463721e-06, "loss": 0.1338, "step": 19854 }, { "epoch": 6.11, "learning_rate": 6.952510452944621e-06, "loss": 0.0834, "step": 19855 }, { "epoch": 6.11, "learning_rate": 6.951561330689899e-06, "loss": 0.0869, "step": 19856 }, { "epoch": 6.11, "learning_rate": 6.950612238708983e-06, "loss": 0.1267, "step": 19857 }, { "epoch": 6.11, "learning_rate": 6.949663177011301e-06, "loss": 0.0653, "step": 19858 }, { "epoch": 6.11, "learning_rate": 6.94871414560627e-06, "loss": 0.0965, "step": 19859 }, { "epoch": 6.11, "learning_rate": 6.947765144503323e-06, "loss": 0.1157, "step": 19860 }, { "epoch": 6.11, "learning_rate": 6.946816173711878e-06, "loss": 0.1036, "step": 19861 }, { "epoch": 6.11, "learning_rate": 6.9458672332413645e-06, "loss": 0.0679, "step": 19862 }, { "epoch": 6.11, "learning_rate": 6.944918323101206e-06, "loss": 0.0781, "step": 19863 }, { "epoch": 6.11, "learning_rate": 6.94396944330082e-06, "loss": 0.0874, "step": 19864 }, { "epoch": 6.11, "learning_rate": 6.943020593849637e-06, "loss": 0.0487, "step": 19865 }, { "epoch": 6.11, "learning_rate": 6.942071774757074e-06, "loss": 0.0657, "step": 19866 }, { "epoch": 6.11, "learning_rate": 6.941122986032559e-06, "loss": 0.0647, "step": 19867 }, { "epoch": 6.11, "learning_rate": 6.940174227685515e-06, "loss": 0.1034, "step": 19868 }, { "epoch": 6.11, "learning_rate": 6.939225499725355e-06, "loss": 0.1167, "step": 19869 }, { "epoch": 6.11, "learning_rate": 6.9382768021615085e-06, "loss": 0.089, "step": 19870 }, { "epoch": 6.11, "learning_rate": 6.937328135003394e-06, "loss": 0.1904, "step": 19871 }, { "epoch": 6.11, "learning_rate": 6.936379498260436e-06, "loss": 0.0476, "step": 19872 }, { "epoch": 6.11, "learning_rate": 6.935430891942054e-06, "loss": 0.08, "step": 19873 }, { "epoch": 6.11, "learning_rate": 6.934482316057663e-06, "loss": 0.1449, "step": 19874 }, { "epoch": 6.12, "learning_rate": 6.933533770616694e-06, "loss": 0.1185, "step": 19875 }, { "epoch": 6.12, "learning_rate": 6.932585255628556e-06, "loss": 0.0757, "step": 19876 }, { "epoch": 6.12, "learning_rate": 6.931636771102674e-06, "loss": 0.046, "step": 19877 }, { "epoch": 6.12, "learning_rate": 6.930688317048469e-06, "loss": 0.1131, "step": 19878 }, { "epoch": 6.12, "learning_rate": 6.929739893475355e-06, "loss": 0.0773, "step": 19879 }, { "epoch": 6.12, "learning_rate": 6.928791500392758e-06, "loss": 0.0299, "step": 19880 }, { "epoch": 6.12, "learning_rate": 6.927843137810087e-06, "loss": 0.1044, "step": 19881 }, { "epoch": 6.12, "learning_rate": 6.926894805736768e-06, "loss": 0.1298, "step": 19882 }, { "epoch": 6.12, "learning_rate": 6.925946504182215e-06, "loss": 0.114, "step": 19883 }, { "epoch": 6.12, "learning_rate": 6.924998233155844e-06, "loss": 0.0403, "step": 19884 }, { "epoch": 6.12, "learning_rate": 6.92404999266708e-06, "loss": 0.0989, "step": 19885 }, { "epoch": 6.12, "learning_rate": 6.923101782725328e-06, "loss": 0.118, "step": 19886 }, { "epoch": 6.12, "learning_rate": 6.922153603340016e-06, "loss": 0.112, "step": 19887 }, { "epoch": 6.12, "learning_rate": 6.921205454520554e-06, "loss": 0.1096, "step": 19888 }, { "epoch": 6.12, "learning_rate": 6.920257336276358e-06, "loss": 0.0825, "step": 19889 }, { "epoch": 6.12, "learning_rate": 6.919309248616848e-06, "loss": 0.0821, "step": 19890 }, { "epoch": 6.12, "learning_rate": 6.91836119155143e-06, "loss": 0.068, "step": 19891 }, { "epoch": 6.12, "learning_rate": 6.917413165089533e-06, "loss": 0.0912, "step": 19892 }, { "epoch": 6.12, "learning_rate": 6.916465169240562e-06, "loss": 0.0702, "step": 19893 }, { "epoch": 6.12, "learning_rate": 6.9155172040139306e-06, "loss": 0.1038, "step": 19894 }, { "epoch": 6.12, "learning_rate": 6.91456926941906e-06, "loss": 0.1397, "step": 19895 }, { "epoch": 6.12, "learning_rate": 6.913621365465357e-06, "loss": 0.0488, "step": 19896 }, { "epoch": 6.12, "learning_rate": 6.912673492162241e-06, "loss": 0.0973, "step": 19897 }, { "epoch": 6.12, "learning_rate": 6.911725649519123e-06, "loss": 0.1186, "step": 19898 }, { "epoch": 6.12, "learning_rate": 6.910777837545413e-06, "loss": 0.0554, "step": 19899 }, { "epoch": 6.12, "learning_rate": 6.909830056250527e-06, "loss": 0.045, "step": 19900 }, { "epoch": 6.12, "learning_rate": 6.908882305643876e-06, "loss": 0.0319, "step": 19901 }, { "epoch": 6.12, "learning_rate": 6.907934585734875e-06, "loss": 0.0812, "step": 19902 }, { "epoch": 6.12, "learning_rate": 6.906986896532932e-06, "loss": 0.0901, "step": 19903 }, { "epoch": 6.12, "learning_rate": 6.906039238047457e-06, "loss": 0.1351, "step": 19904 }, { "epoch": 6.12, "learning_rate": 6.9050916102878676e-06, "loss": 0.1409, "step": 19905 }, { "epoch": 6.12, "learning_rate": 6.904144013263567e-06, "loss": 0.0746, "step": 19906 }, { "epoch": 6.12, "learning_rate": 6.903196446983973e-06, "loss": 0.1006, "step": 19907 }, { "epoch": 6.13, "learning_rate": 6.902248911458494e-06, "loss": 0.0909, "step": 19908 }, { "epoch": 6.13, "learning_rate": 6.901301406696534e-06, "loss": 0.1067, "step": 19909 }, { "epoch": 6.13, "learning_rate": 6.900353932707509e-06, "loss": 0.1066, "step": 19910 }, { "epoch": 6.13, "learning_rate": 6.899406489500824e-06, "loss": 0.0752, "step": 19911 }, { "epoch": 6.13, "learning_rate": 6.898459077085893e-06, "loss": 0.1034, "step": 19912 }, { "epoch": 6.13, "learning_rate": 6.897511695472124e-06, "loss": 0.0877, "step": 19913 }, { "epoch": 6.13, "learning_rate": 6.896564344668917e-06, "loss": 0.1545, "step": 19914 }, { "epoch": 6.13, "learning_rate": 6.895617024685689e-06, "loss": 0.1038, "step": 19915 }, { "epoch": 6.13, "learning_rate": 6.894669735531844e-06, "loss": 0.0774, "step": 19916 }, { "epoch": 6.13, "learning_rate": 6.893722477216792e-06, "loss": 0.1205, "step": 19917 }, { "epoch": 6.13, "learning_rate": 6.892775249749942e-06, "loss": 0.0754, "step": 19918 }, { "epoch": 6.13, "learning_rate": 6.891828053140691e-06, "loss": 0.1044, "step": 19919 }, { "epoch": 6.13, "learning_rate": 6.890880887398455e-06, "loss": 0.1313, "step": 19920 }, { "epoch": 6.13, "learning_rate": 6.889933752532636e-06, "loss": 0.0601, "step": 19921 }, { "epoch": 6.13, "learning_rate": 6.888986648552643e-06, "loss": 0.1023, "step": 19922 }, { "epoch": 6.13, "learning_rate": 6.888039575467881e-06, "loss": 0.112, "step": 19923 }, { "epoch": 6.13, "learning_rate": 6.887092533287748e-06, "loss": 0.0991, "step": 19924 }, { "epoch": 6.13, "learning_rate": 6.886145522021661e-06, "loss": 0.055, "step": 19925 }, { "epoch": 6.13, "learning_rate": 6.885198541679016e-06, "loss": 0.105, "step": 19926 }, { "epoch": 6.13, "learning_rate": 6.884251592269221e-06, "loss": 0.0855, "step": 19927 }, { "epoch": 6.13, "learning_rate": 6.88330467380168e-06, "loss": 0.0905, "step": 19928 }, { "epoch": 6.13, "learning_rate": 6.8823577862857935e-06, "loss": 0.0647, "step": 19929 }, { "epoch": 6.13, "learning_rate": 6.881410929730972e-06, "loss": 0.127, "step": 19930 }, { "epoch": 6.13, "learning_rate": 6.880464104146609e-06, "loss": 0.0463, "step": 19931 }, { "epoch": 6.13, "learning_rate": 6.879517309542116e-06, "loss": 0.104, "step": 19932 }, { "epoch": 6.13, "learning_rate": 6.878570545926893e-06, "loss": 0.0631, "step": 19933 }, { "epoch": 6.13, "learning_rate": 6.877623813310337e-06, "loss": 0.0541, "step": 19934 }, { "epoch": 6.13, "learning_rate": 6.8766771117018585e-06, "loss": 0.088, "step": 19935 }, { "epoch": 6.13, "learning_rate": 6.8757304411108485e-06, "loss": 0.1227, "step": 19936 }, { "epoch": 6.13, "learning_rate": 6.8747838015467215e-06, "loss": 0.0874, "step": 19937 }, { "epoch": 6.13, "learning_rate": 6.87383719301887e-06, "loss": 0.0993, "step": 19938 }, { "epoch": 6.13, "learning_rate": 6.872890615536694e-06, "loss": 0.0428, "step": 19939 }, { "epoch": 6.14, "learning_rate": 6.871944069109599e-06, "loss": 0.0818, "step": 19940 }, { "epoch": 6.14, "learning_rate": 6.870997553746979e-06, "loss": 0.1, "step": 19941 }, { "epoch": 6.14, "learning_rate": 6.87005106945824e-06, "loss": 0.0791, "step": 19942 }, { "epoch": 6.14, "learning_rate": 6.869104616252777e-06, "loss": 0.114, "step": 19943 }, { "epoch": 6.14, "learning_rate": 6.868158194139991e-06, "loss": 0.1282, "step": 19944 }, { "epoch": 6.14, "learning_rate": 6.867211803129281e-06, "loss": 0.0803, "step": 19945 }, { "epoch": 6.14, "learning_rate": 6.866265443230042e-06, "loss": 0.1125, "step": 19946 }, { "epoch": 6.14, "learning_rate": 6.865319114451682e-06, "loss": 0.1321, "step": 19947 }, { "epoch": 6.14, "learning_rate": 6.864372816803588e-06, "loss": 0.1184, "step": 19948 }, { "epoch": 6.14, "learning_rate": 6.863426550295161e-06, "loss": 0.133, "step": 19949 }, { "epoch": 6.14, "learning_rate": 6.8624803149358e-06, "loss": 0.0962, "step": 19950 }, { "epoch": 6.14, "learning_rate": 6.8615341107349e-06, "loss": 0.1532, "step": 19951 }, { "epoch": 6.14, "learning_rate": 6.860587937701862e-06, "loss": 0.1181, "step": 19952 }, { "epoch": 6.14, "learning_rate": 6.859641795846078e-06, "loss": 0.0619, "step": 19953 }, { "epoch": 6.14, "learning_rate": 6.858695685176942e-06, "loss": 0.1104, "step": 19954 }, { "epoch": 6.14, "learning_rate": 6.857749605703856e-06, "loss": 0.1057, "step": 19955 }, { "epoch": 6.14, "learning_rate": 6.856803557436211e-06, "loss": 0.0502, "step": 19956 }, { "epoch": 6.14, "learning_rate": 6.855857540383404e-06, "loss": 0.1496, "step": 19957 }, { "epoch": 6.14, "learning_rate": 6.854911554554832e-06, "loss": 0.1479, "step": 19958 }, { "epoch": 6.14, "learning_rate": 6.853965599959883e-06, "loss": 0.0763, "step": 19959 }, { "epoch": 6.14, "learning_rate": 6.853019676607957e-06, "loss": 0.0778, "step": 19960 }, { "epoch": 6.14, "learning_rate": 6.852073784508444e-06, "loss": 0.0732, "step": 19961 }, { "epoch": 6.14, "learning_rate": 6.851127923670742e-06, "loss": 0.0768, "step": 19962 }, { "epoch": 6.14, "learning_rate": 6.850182094104243e-06, "loss": 0.0698, "step": 19963 }, { "epoch": 6.14, "learning_rate": 6.849236295818334e-06, "loss": 0.1028, "step": 19964 }, { "epoch": 6.14, "learning_rate": 6.848290528822417e-06, "loss": 0.0846, "step": 19965 }, { "epoch": 6.14, "learning_rate": 6.847344793125875e-06, "loss": 0.0995, "step": 19966 }, { "epoch": 6.14, "learning_rate": 6.846399088738108e-06, "loss": 0.1077, "step": 19967 }, { "epoch": 6.14, "learning_rate": 6.845453415668507e-06, "loss": 0.1233, "step": 19968 }, { "epoch": 6.14, "learning_rate": 6.844507773926456e-06, "loss": 0.0841, "step": 19969 }, { "epoch": 6.14, "learning_rate": 6.843562163521353e-06, "loss": 0.059, "step": 19970 }, { "epoch": 6.14, "learning_rate": 6.8426165844625845e-06, "loss": 0.1042, "step": 19971 }, { "epoch": 6.14, "learning_rate": 6.841671036759548e-06, "loss": 0.104, "step": 19972 }, { "epoch": 6.15, "learning_rate": 6.840725520421627e-06, "loss": 0.0938, "step": 19973 }, { "epoch": 6.15, "learning_rate": 6.83978003545821e-06, "loss": 0.0743, "step": 19974 }, { "epoch": 6.15, "learning_rate": 6.838834581878695e-06, "loss": 0.099, "step": 19975 }, { "epoch": 6.15, "learning_rate": 6.837889159692462e-06, "loss": 0.1127, "step": 19976 }, { "epoch": 6.15, "learning_rate": 6.836943768908905e-06, "loss": 0.1379, "step": 19977 }, { "epoch": 6.15, "learning_rate": 6.835998409537412e-06, "loss": 0.1002, "step": 19978 }, { "epoch": 6.15, "learning_rate": 6.8350530815873694e-06, "loss": 0.0729, "step": 19979 }, { "epoch": 6.15, "learning_rate": 6.83410778506817e-06, "loss": 0.0827, "step": 19980 }, { "epoch": 6.15, "learning_rate": 6.833162519989195e-06, "loss": 0.1023, "step": 19981 }, { "epoch": 6.15, "learning_rate": 6.8322172863598355e-06, "loss": 0.0964, "step": 19982 }, { "epoch": 6.15, "learning_rate": 6.83127208418948e-06, "loss": 0.0869, "step": 19983 }, { "epoch": 6.15, "learning_rate": 6.830326913487508e-06, "loss": 0.0997, "step": 19984 }, { "epoch": 6.15, "learning_rate": 6.8293817742633175e-06, "loss": 0.1591, "step": 19985 }, { "epoch": 6.15, "learning_rate": 6.828436666526281e-06, "loss": 0.0698, "step": 19986 }, { "epoch": 6.15, "learning_rate": 6.827491590285798e-06, "loss": 0.0585, "step": 19987 }, { "epoch": 6.15, "learning_rate": 6.826546545551244e-06, "loss": 0.1013, "step": 19988 }, { "epoch": 6.15, "learning_rate": 6.825601532332007e-06, "loss": 0.0976, "step": 19989 }, { "epoch": 6.15, "learning_rate": 6.824656550637476e-06, "loss": 0.1168, "step": 19990 }, { "epoch": 6.15, "learning_rate": 6.823711600477025e-06, "loss": 0.1381, "step": 19991 }, { "epoch": 6.15, "learning_rate": 6.822766681860053e-06, "loss": 0.0737, "step": 19992 }, { "epoch": 6.15, "learning_rate": 6.821821794795933e-06, "loss": 0.089, "step": 19993 }, { "epoch": 6.15, "learning_rate": 6.82087693929405e-06, "loss": 0.1077, "step": 19994 }, { "epoch": 6.15, "learning_rate": 6.819932115363791e-06, "loss": 0.1048, "step": 19995 }, { "epoch": 6.15, "learning_rate": 6.818987323014534e-06, "loss": 0.1182, "step": 19996 }, { "epoch": 6.15, "learning_rate": 6.81804256225567e-06, "loss": 0.0785, "step": 19997 }, { "epoch": 6.15, "learning_rate": 6.817097833096573e-06, "loss": 0.0988, "step": 19998 }, { "epoch": 6.15, "learning_rate": 6.816153135546626e-06, "loss": 0.0816, "step": 19999 }, { "epoch": 6.15, "learning_rate": 6.815208469615216e-06, "loss": 0.0642, "step": 20000 }, { "epoch": 6.15, "learning_rate": 6.814263835311718e-06, "loss": 0.0466, "step": 20001 }, { "epoch": 6.15, "learning_rate": 6.813319232645519e-06, "loss": 0.1873, "step": 20002 }, { "epoch": 6.15, "learning_rate": 6.812374661625993e-06, "loss": 0.1281, "step": 20003 }, { "epoch": 6.15, "learning_rate": 6.811430122262529e-06, "loss": 0.1097, "step": 20004 }, { "epoch": 6.16, "learning_rate": 6.810485614564502e-06, "loss": 0.0837, "step": 20005 }, { "epoch": 6.16, "learning_rate": 6.809541138541289e-06, "loss": 0.059, "step": 20006 }, { "epoch": 6.16, "learning_rate": 6.808596694202275e-06, "loss": 0.0373, "step": 20007 }, { "epoch": 6.16, "learning_rate": 6.807652281556836e-06, "loss": 0.0777, "step": 20008 }, { "epoch": 6.16, "learning_rate": 6.806707900614355e-06, "loss": 0.1141, "step": 20009 }, { "epoch": 6.16, "learning_rate": 6.805763551384205e-06, "loss": 0.0504, "step": 20010 }, { "epoch": 6.16, "learning_rate": 6.804819233875767e-06, "loss": 0.0909, "step": 20011 }, { "epoch": 6.16, "learning_rate": 6.803874948098419e-06, "loss": 0.0718, "step": 20012 }, { "epoch": 6.16, "learning_rate": 6.802930694061538e-06, "loss": 0.0823, "step": 20013 }, { "epoch": 6.16, "learning_rate": 6.801986471774504e-06, "loss": 0.1111, "step": 20014 }, { "epoch": 6.16, "learning_rate": 6.80104228124669e-06, "loss": 0.1459, "step": 20015 }, { "epoch": 6.16, "learning_rate": 6.800098122487473e-06, "loss": 0.0603, "step": 20016 }, { "epoch": 6.16, "learning_rate": 6.799153995506234e-06, "loss": 0.0863, "step": 20017 }, { "epoch": 6.16, "learning_rate": 6.798209900312343e-06, "loss": 0.0796, "step": 20018 }, { "epoch": 6.16, "learning_rate": 6.797265836915181e-06, "loss": 0.0893, "step": 20019 }, { "epoch": 6.16, "learning_rate": 6.7963218053241195e-06, "loss": 0.1119, "step": 20020 }, { "epoch": 6.16, "learning_rate": 6.7953778055485335e-06, "loss": 0.0785, "step": 20021 }, { "epoch": 6.16, "learning_rate": 6.794433837597802e-06, "loss": 0.0658, "step": 20022 }, { "epoch": 6.16, "learning_rate": 6.793489901481294e-06, "loss": 0.1479, "step": 20023 }, { "epoch": 6.16, "learning_rate": 6.792545997208388e-06, "loss": 0.0427, "step": 20024 }, { "epoch": 6.16, "learning_rate": 6.791602124788459e-06, "loss": 0.0659, "step": 20025 }, { "epoch": 6.16, "learning_rate": 6.790658284230874e-06, "loss": 0.0974, "step": 20026 }, { "epoch": 6.16, "learning_rate": 6.789714475545013e-06, "loss": 0.0533, "step": 20027 }, { "epoch": 6.16, "learning_rate": 6.788770698740243e-06, "loss": 0.1378, "step": 20028 }, { "epoch": 6.16, "learning_rate": 6.7878269538259404e-06, "loss": 0.1055, "step": 20029 }, { "epoch": 6.16, "learning_rate": 6.786883240811479e-06, "loss": 0.0883, "step": 20030 }, { "epoch": 6.16, "learning_rate": 6.785939559706225e-06, "loss": 0.0514, "step": 20031 }, { "epoch": 6.16, "learning_rate": 6.7849959105195565e-06, "loss": 0.159, "step": 20032 }, { "epoch": 6.16, "learning_rate": 6.7840522932608375e-06, "loss": 0.0739, "step": 20033 }, { "epoch": 6.16, "learning_rate": 6.783108707939447e-06, "loss": 0.0959, "step": 20034 }, { "epoch": 6.16, "learning_rate": 6.782165154564753e-06, "loss": 0.1042, "step": 20035 }, { "epoch": 6.16, "learning_rate": 6.781221633146118e-06, "loss": 0.1378, "step": 20036 }, { "epoch": 6.16, "learning_rate": 6.780278143692925e-06, "loss": 0.119, "step": 20037 }, { "epoch": 6.17, "learning_rate": 6.779334686214533e-06, "loss": 0.1224, "step": 20038 }, { "epoch": 6.17, "learning_rate": 6.778391260720319e-06, "loss": 0.0425, "step": 20039 }, { "epoch": 6.17, "learning_rate": 6.777447867219651e-06, "loss": 0.0882, "step": 20040 }, { "epoch": 6.17, "learning_rate": 6.776504505721888e-06, "loss": 0.1261, "step": 20041 }, { "epoch": 6.17, "learning_rate": 6.775561176236413e-06, "loss": 0.0999, "step": 20042 }, { "epoch": 6.17, "learning_rate": 6.774617878772582e-06, "loss": 0.1272, "step": 20043 }, { "epoch": 6.17, "learning_rate": 6.773674613339774e-06, "loss": 0.0845, "step": 20044 }, { "epoch": 6.17, "learning_rate": 6.772731379947346e-06, "loss": 0.0696, "step": 20045 }, { "epoch": 6.17, "learning_rate": 6.77178817860467e-06, "loss": 0.0627, "step": 20046 }, { "epoch": 6.17, "learning_rate": 6.770845009321117e-06, "loss": 0.1137, "step": 20047 }, { "epoch": 6.17, "learning_rate": 6.769901872106044e-06, "loss": 0.0953, "step": 20048 }, { "epoch": 6.17, "learning_rate": 6.768958766968825e-06, "loss": 0.1029, "step": 20049 }, { "epoch": 6.17, "learning_rate": 6.7680156939188235e-06, "loss": 0.1142, "step": 20050 }, { "epoch": 6.17, "learning_rate": 6.767072652965402e-06, "loss": 0.1117, "step": 20051 }, { "epoch": 6.17, "learning_rate": 6.766129644117932e-06, "loss": 0.1145, "step": 20052 }, { "epoch": 6.17, "learning_rate": 6.76518666738577e-06, "loss": 0.106, "step": 20053 }, { "epoch": 6.17, "learning_rate": 6.764243722778292e-06, "loss": 0.1582, "step": 20054 }, { "epoch": 6.17, "learning_rate": 6.763300810304853e-06, "loss": 0.1414, "step": 20055 }, { "epoch": 6.17, "learning_rate": 6.76235792997482e-06, "loss": 0.1063, "step": 20056 }, { "epoch": 6.17, "learning_rate": 6.761415081797558e-06, "loss": 0.0702, "step": 20057 }, { "epoch": 6.17, "learning_rate": 6.760472265782425e-06, "loss": 0.1008, "step": 20058 }, { "epoch": 6.17, "learning_rate": 6.759529481938794e-06, "loss": 0.0725, "step": 20059 }, { "epoch": 6.17, "learning_rate": 6.75858673027602e-06, "loss": 0.1277, "step": 20060 }, { "epoch": 6.17, "learning_rate": 6.757644010803464e-06, "loss": 0.0758, "step": 20061 }, { "epoch": 6.17, "learning_rate": 6.756701323530492e-06, "loss": 0.1049, "step": 20062 }, { "epoch": 6.17, "learning_rate": 6.7557586684664655e-06, "loss": 0.0299, "step": 20063 }, { "epoch": 6.17, "learning_rate": 6.754816045620749e-06, "loss": 0.1339, "step": 20064 }, { "epoch": 6.17, "learning_rate": 6.753873455002698e-06, "loss": 0.0734, "step": 20065 }, { "epoch": 6.17, "learning_rate": 6.752930896621671e-06, "loss": 0.1109, "step": 20066 }, { "epoch": 6.17, "learning_rate": 6.751988370487037e-06, "loss": 0.082, "step": 20067 }, { "epoch": 6.17, "learning_rate": 6.751045876608149e-06, "loss": 0.1277, "step": 20068 }, { "epoch": 6.17, "learning_rate": 6.750103414994374e-06, "loss": 0.085, "step": 20069 }, { "epoch": 6.18, "learning_rate": 6.7491609856550656e-06, "loss": 0.0404, "step": 20070 }, { "epoch": 6.18, "learning_rate": 6.748218588599582e-06, "loss": 0.0519, "step": 20071 }, { "epoch": 6.18, "learning_rate": 6.747276223837289e-06, "loss": 0.0669, "step": 20072 }, { "epoch": 6.18, "learning_rate": 6.7463338913775365e-06, "loss": 0.1147, "step": 20073 }, { "epoch": 6.18, "learning_rate": 6.745391591229689e-06, "loss": 0.1054, "step": 20074 }, { "epoch": 6.18, "learning_rate": 6.744449323403106e-06, "loss": 0.0872, "step": 20075 }, { "epoch": 6.18, "learning_rate": 6.7435070879071374e-06, "loss": 0.1306, "step": 20076 }, { "epoch": 6.18, "learning_rate": 6.742564884751146e-06, "loss": 0.1131, "step": 20077 }, { "epoch": 6.18, "learning_rate": 6.7416227139444854e-06, "loss": 0.0699, "step": 20078 }, { "epoch": 6.18, "learning_rate": 6.740680575496516e-06, "loss": 0.0528, "step": 20079 }, { "epoch": 6.18, "learning_rate": 6.739738469416596e-06, "loss": 0.1033, "step": 20080 }, { "epoch": 6.18, "learning_rate": 6.738796395714072e-06, "loss": 0.154, "step": 20081 }, { "epoch": 6.18, "learning_rate": 6.737854354398308e-06, "loss": 0.0971, "step": 20082 }, { "epoch": 6.18, "learning_rate": 6.736912345478654e-06, "loss": 0.0608, "step": 20083 }, { "epoch": 6.18, "learning_rate": 6.735970368964472e-06, "loss": 0.1417, "step": 20084 }, { "epoch": 6.18, "learning_rate": 6.735028424865113e-06, "loss": 0.1071, "step": 20085 }, { "epoch": 6.18, "learning_rate": 6.734086513189924e-06, "loss": 0.095, "step": 20086 }, { "epoch": 6.18, "learning_rate": 6.733144633948274e-06, "loss": 0.1012, "step": 20087 }, { "epoch": 6.18, "learning_rate": 6.732202787149502e-06, "loss": 0.1324, "step": 20088 }, { "epoch": 6.18, "learning_rate": 6.731260972802972e-06, "loss": 0.0518, "step": 20089 }, { "epoch": 6.18, "learning_rate": 6.730319190918036e-06, "loss": 0.1349, "step": 20090 }, { "epoch": 6.18, "learning_rate": 6.729377441504035e-06, "loss": 0.1228, "step": 20091 }, { "epoch": 6.18, "learning_rate": 6.728435724570337e-06, "loss": 0.0748, "step": 20092 }, { "epoch": 6.18, "learning_rate": 6.727494040126284e-06, "loss": 0.1196, "step": 20093 }, { "epoch": 6.18, "learning_rate": 6.7265523881812335e-06, "loss": 0.1053, "step": 20094 }, { "epoch": 6.18, "learning_rate": 6.725610768744535e-06, "loss": 0.1191, "step": 20095 }, { "epoch": 6.18, "learning_rate": 6.724669181825536e-06, "loss": 0.1036, "step": 20096 }, { "epoch": 6.18, "learning_rate": 6.723727627433596e-06, "loss": 0.102, "step": 20097 }, { "epoch": 6.18, "learning_rate": 6.722786105578056e-06, "loss": 0.1417, "step": 20098 }, { "epoch": 6.18, "learning_rate": 6.721844616268271e-06, "loss": 0.1288, "step": 20099 }, { "epoch": 6.18, "learning_rate": 6.720903159513592e-06, "loss": 0.0601, "step": 20100 }, { "epoch": 6.18, "learning_rate": 6.719961735323364e-06, "loss": 0.0824, "step": 20101 }, { "epoch": 6.18, "learning_rate": 6.719020343706942e-06, "loss": 0.0595, "step": 20102 }, { "epoch": 6.19, "learning_rate": 6.718078984673666e-06, "loss": 0.0753, "step": 20103 }, { "epoch": 6.19, "learning_rate": 6.717137658232897e-06, "loss": 0.118, "step": 20104 }, { "epoch": 6.19, "learning_rate": 6.7161963643939755e-06, "loss": 0.1145, "step": 20105 }, { "epoch": 6.19, "learning_rate": 6.715255103166247e-06, "loss": 0.0765, "step": 20106 }, { "epoch": 6.19, "learning_rate": 6.714313874559067e-06, "loss": 0.112, "step": 20107 }, { "epoch": 6.19, "learning_rate": 6.713372678581773e-06, "loss": 0.0983, "step": 20108 }, { "epoch": 6.19, "learning_rate": 6.712431515243724e-06, "loss": 0.0684, "step": 20109 }, { "epoch": 6.19, "learning_rate": 6.711490384554257e-06, "loss": 0.0888, "step": 20110 }, { "epoch": 6.19, "learning_rate": 6.710549286522718e-06, "loss": 0.0782, "step": 20111 }, { "epoch": 6.19, "learning_rate": 6.709608221158461e-06, "loss": 0.1206, "step": 20112 }, { "epoch": 6.19, "learning_rate": 6.708667188470823e-06, "loss": 0.0832, "step": 20113 }, { "epoch": 6.19, "learning_rate": 6.707726188469157e-06, "loss": 0.1433, "step": 20114 }, { "epoch": 6.19, "learning_rate": 6.7067852211628035e-06, "loss": 0.1193, "step": 20115 }, { "epoch": 6.19, "learning_rate": 6.705844286561105e-06, "loss": 0.0997, "step": 20116 }, { "epoch": 6.19, "learning_rate": 6.704903384673411e-06, "loss": 0.1074, "step": 20117 }, { "epoch": 6.19, "learning_rate": 6.703962515509062e-06, "loss": 0.109, "step": 20118 }, { "epoch": 6.19, "learning_rate": 6.703021679077405e-06, "loss": 0.1168, "step": 20119 }, { "epoch": 6.19, "learning_rate": 6.702080875387781e-06, "loss": 0.0963, "step": 20120 }, { "epoch": 6.19, "learning_rate": 6.7011401044495304e-06, "loss": 0.1336, "step": 20121 }, { "epoch": 6.19, "learning_rate": 6.700199366272001e-06, "loss": 0.0522, "step": 20122 }, { "epoch": 6.19, "learning_rate": 6.699258660864532e-06, "loss": 0.1056, "step": 20123 }, { "epoch": 6.19, "learning_rate": 6.6983179882364665e-06, "loss": 0.1504, "step": 20124 }, { "epoch": 6.19, "learning_rate": 6.697377348397151e-06, "loss": 0.0746, "step": 20125 }, { "epoch": 6.19, "learning_rate": 6.696436741355916e-06, "loss": 0.0759, "step": 20126 }, { "epoch": 6.19, "learning_rate": 6.6954961671221115e-06, "loss": 0.1113, "step": 20127 }, { "epoch": 6.19, "learning_rate": 6.694555625705075e-06, "loss": 0.1266, "step": 20128 }, { "epoch": 6.19, "learning_rate": 6.693615117114147e-06, "loss": 0.0841, "step": 20129 }, { "epoch": 6.19, "learning_rate": 6.692674641358672e-06, "loss": 0.0899, "step": 20130 }, { "epoch": 6.19, "learning_rate": 6.691734198447981e-06, "loss": 0.1181, "step": 20131 }, { "epoch": 6.19, "learning_rate": 6.690793788391421e-06, "loss": 0.1088, "step": 20132 }, { "epoch": 6.19, "learning_rate": 6.689853411198326e-06, "loss": 0.1128, "step": 20133 }, { "epoch": 6.19, "learning_rate": 6.68891306687804e-06, "loss": 0.1203, "step": 20134 }, { "epoch": 6.2, "learning_rate": 6.6879727554398995e-06, "loss": 0.0631, "step": 20135 }, { "epoch": 6.2, "learning_rate": 6.687032476893238e-06, "loss": 0.0591, "step": 20136 }, { "epoch": 6.2, "learning_rate": 6.686092231247402e-06, "loss": 0.073, "step": 20137 }, { "epoch": 6.2, "learning_rate": 6.68515201851172e-06, "loss": 0.0741, "step": 20138 }, { "epoch": 6.2, "learning_rate": 6.684211838695537e-06, "loss": 0.0939, "step": 20139 }, { "epoch": 6.2, "learning_rate": 6.683271691808187e-06, "loss": 0.1052, "step": 20140 }, { "epoch": 6.2, "learning_rate": 6.682331577859001e-06, "loss": 0.1076, "step": 20141 }, { "epoch": 6.2, "learning_rate": 6.681391496857326e-06, "loss": 0.0718, "step": 20142 }, { "epoch": 6.2, "learning_rate": 6.680451448812486e-06, "loss": 0.0581, "step": 20143 }, { "epoch": 6.2, "learning_rate": 6.679511433733827e-06, "loss": 0.0598, "step": 20144 }, { "epoch": 6.2, "learning_rate": 6.678571451630678e-06, "loss": 0.1097, "step": 20145 }, { "epoch": 6.2, "learning_rate": 6.677631502512374e-06, "loss": 0.0869, "step": 20146 }, { "epoch": 6.2, "learning_rate": 6.676691586388255e-06, "loss": 0.1048, "step": 20147 }, { "epoch": 6.2, "learning_rate": 6.6757517032676475e-06, "loss": 0.0937, "step": 20148 }, { "epoch": 6.2, "learning_rate": 6.6748118531598924e-06, "loss": 0.094, "step": 20149 }, { "epoch": 6.2, "learning_rate": 6.673872036074321e-06, "loss": 0.0712, "step": 20150 }, { "epoch": 6.2, "learning_rate": 6.672932252020261e-06, "loss": 0.0893, "step": 20151 }, { "epoch": 6.2, "learning_rate": 6.671992501007055e-06, "loss": 0.1207, "step": 20152 }, { "epoch": 6.2, "learning_rate": 6.671052783044025e-06, "loss": 0.0645, "step": 20153 }, { "epoch": 6.2, "learning_rate": 6.670113098140514e-06, "loss": 0.1169, "step": 20154 }, { "epoch": 6.2, "learning_rate": 6.6691734463058476e-06, "loss": 0.1033, "step": 20155 }, { "epoch": 6.2, "learning_rate": 6.6682338275493576e-06, "loss": 0.1472, "step": 20156 }, { "epoch": 6.2, "learning_rate": 6.667294241880378e-06, "loss": 0.0993, "step": 20157 }, { "epoch": 6.2, "learning_rate": 6.666354689308234e-06, "loss": 0.1017, "step": 20158 }, { "epoch": 6.2, "learning_rate": 6.665415169842267e-06, "loss": 0.1035, "step": 20159 }, { "epoch": 6.2, "learning_rate": 6.664475683491797e-06, "loss": 0.0643, "step": 20160 }, { "epoch": 6.2, "learning_rate": 6.663536230266156e-06, "loss": 0.1203, "step": 20161 }, { "epoch": 6.2, "learning_rate": 6.662596810174677e-06, "loss": 0.12, "step": 20162 }, { "epoch": 6.2, "learning_rate": 6.661657423226685e-06, "loss": 0.1205, "step": 20163 }, { "epoch": 6.2, "learning_rate": 6.6607180694315165e-06, "loss": 0.1366, "step": 20164 }, { "epoch": 6.2, "learning_rate": 6.6597787487984935e-06, "loss": 0.1035, "step": 20165 }, { "epoch": 6.2, "learning_rate": 6.658839461336943e-06, "loss": 0.1089, "step": 20166 }, { "epoch": 6.2, "learning_rate": 6.657900207056198e-06, "loss": 0.0578, "step": 20167 }, { "epoch": 6.21, "learning_rate": 6.656960985965581e-06, "loss": 0.064, "step": 20168 }, { "epoch": 6.21, "learning_rate": 6.656021798074429e-06, "loss": 0.0706, "step": 20169 }, { "epoch": 6.21, "learning_rate": 6.655082643392059e-06, "loss": 0.1, "step": 20170 }, { "epoch": 6.21, "learning_rate": 6.654143521927799e-06, "loss": 0.1184, "step": 20171 }, { "epoch": 6.21, "learning_rate": 6.653204433690981e-06, "loss": 0.0699, "step": 20172 }, { "epoch": 6.21, "learning_rate": 6.652265378690923e-06, "loss": 0.0793, "step": 20173 }, { "epoch": 6.21, "learning_rate": 6.651326356936959e-06, "loss": 0.068, "step": 20174 }, { "epoch": 6.21, "learning_rate": 6.650387368438412e-06, "loss": 0.0489, "step": 20175 }, { "epoch": 6.21, "learning_rate": 6.649448413204602e-06, "loss": 0.0629, "step": 20176 }, { "epoch": 6.21, "learning_rate": 6.648509491244859e-06, "loss": 0.0571, "step": 20177 }, { "epoch": 6.21, "learning_rate": 6.647570602568503e-06, "loss": 0.1046, "step": 20178 }, { "epoch": 6.21, "learning_rate": 6.646631747184862e-06, "loss": 0.1425, "step": 20179 }, { "epoch": 6.21, "learning_rate": 6.645692925103262e-06, "loss": 0.1169, "step": 20180 }, { "epoch": 6.21, "learning_rate": 6.644754136333018e-06, "loss": 0.0873, "step": 20181 }, { "epoch": 6.21, "learning_rate": 6.643815380883458e-06, "loss": 0.0972, "step": 20182 }, { "epoch": 6.21, "learning_rate": 6.6428766587639045e-06, "loss": 0.1199, "step": 20183 }, { "epoch": 6.21, "learning_rate": 6.641937969983682e-06, "loss": 0.0377, "step": 20184 }, { "epoch": 6.21, "learning_rate": 6.640999314552112e-06, "loss": 0.0742, "step": 20185 }, { "epoch": 6.21, "learning_rate": 6.6400606924785095e-06, "loss": 0.0894, "step": 20186 }, { "epoch": 6.21, "learning_rate": 6.639122103772204e-06, "loss": 0.07, "step": 20187 }, { "epoch": 6.21, "learning_rate": 6.638183548442511e-06, "loss": 0.1016, "step": 20188 }, { "epoch": 6.21, "learning_rate": 6.637245026498755e-06, "loss": 0.1247, "step": 20189 }, { "epoch": 6.21, "learning_rate": 6.636306537950259e-06, "loss": 0.087, "step": 20190 }, { "epoch": 6.21, "learning_rate": 6.6353680828063306e-06, "loss": 0.098, "step": 20191 }, { "epoch": 6.21, "learning_rate": 6.634429661076307e-06, "loss": 0.0881, "step": 20192 }, { "epoch": 6.21, "learning_rate": 6.633491272769492e-06, "loss": 0.1226, "step": 20193 }, { "epoch": 6.21, "learning_rate": 6.632552917895214e-06, "loss": 0.0886, "step": 20194 }, { "epoch": 6.21, "learning_rate": 6.63161459646279e-06, "loss": 0.0993, "step": 20195 }, { "epoch": 6.21, "learning_rate": 6.6306763084815365e-06, "loss": 0.0568, "step": 20196 }, { "epoch": 6.21, "learning_rate": 6.629738053960774e-06, "loss": 0.075, "step": 20197 }, { "epoch": 6.21, "learning_rate": 6.628799832909816e-06, "loss": 0.1221, "step": 20198 }, { "epoch": 6.21, "learning_rate": 6.627861645337984e-06, "loss": 0.1514, "step": 20199 }, { "epoch": 6.22, "learning_rate": 6.6269234912545925e-06, "loss": 0.0462, "step": 20200 }, { "epoch": 6.22, "learning_rate": 6.625985370668961e-06, "loss": 0.0972, "step": 20201 }, { "epoch": 6.22, "learning_rate": 6.625047283590406e-06, "loss": 0.1265, "step": 20202 }, { "epoch": 6.22, "learning_rate": 6.624109230028237e-06, "loss": 0.0977, "step": 20203 }, { "epoch": 6.22, "learning_rate": 6.62317120999178e-06, "loss": 0.1158, "step": 20204 }, { "epoch": 6.22, "learning_rate": 6.62223322349034e-06, "loss": 0.1005, "step": 20205 }, { "epoch": 6.22, "learning_rate": 6.6212952705332416e-06, "loss": 0.1247, "step": 20206 }, { "epoch": 6.22, "learning_rate": 6.620357351129796e-06, "loss": 0.13, "step": 20207 }, { "epoch": 6.22, "learning_rate": 6.6194194652893095e-06, "loss": 0.119, "step": 20208 }, { "epoch": 6.22, "learning_rate": 6.61848161302111e-06, "loss": 0.1062, "step": 20209 }, { "epoch": 6.22, "learning_rate": 6.617543794334501e-06, "loss": 0.1144, "step": 20210 }, { "epoch": 6.22, "learning_rate": 6.616606009238802e-06, "loss": 0.0922, "step": 20211 }, { "epoch": 6.22, "learning_rate": 6.615668257743322e-06, "loss": 0.1549, "step": 20212 }, { "epoch": 6.22, "learning_rate": 6.614730539857375e-06, "loss": 0.0941, "step": 20213 }, { "epoch": 6.22, "learning_rate": 6.613792855590276e-06, "loss": 0.097, "step": 20214 }, { "epoch": 6.22, "learning_rate": 6.612855204951332e-06, "loss": 0.0485, "step": 20215 }, { "epoch": 6.22, "learning_rate": 6.611917587949859e-06, "loss": 0.1354, "step": 20216 }, { "epoch": 6.22, "learning_rate": 6.610980004595167e-06, "loss": 0.1233, "step": 20217 }, { "epoch": 6.22, "learning_rate": 6.610042454896564e-06, "loss": 0.1292, "step": 20218 }, { "epoch": 6.22, "learning_rate": 6.609104938863369e-06, "loss": 0.0665, "step": 20219 }, { "epoch": 6.22, "learning_rate": 6.608167456504882e-06, "loss": 0.0907, "step": 20220 }, { "epoch": 6.22, "learning_rate": 6.607230007830422e-06, "loss": 0.1154, "step": 20221 }, { "epoch": 6.22, "learning_rate": 6.606292592849294e-06, "loss": 0.0479, "step": 20222 }, { "epoch": 6.22, "learning_rate": 6.605355211570805e-06, "loss": 0.0751, "step": 20223 }, { "epoch": 6.22, "learning_rate": 6.604417864004271e-06, "loss": 0.0608, "step": 20224 }, { "epoch": 6.22, "learning_rate": 6.603480550158995e-06, "loss": 0.0999, "step": 20225 }, { "epoch": 6.22, "learning_rate": 6.60254327004429e-06, "loss": 0.125, "step": 20226 }, { "epoch": 6.22, "learning_rate": 6.601606023669461e-06, "loss": 0.1237, "step": 20227 }, { "epoch": 6.22, "learning_rate": 6.6006688110438134e-06, "loss": 0.1352, "step": 20228 }, { "epoch": 6.22, "learning_rate": 6.599731632176661e-06, "loss": 0.1776, "step": 20229 }, { "epoch": 6.22, "learning_rate": 6.598794487077304e-06, "loss": 0.0994, "step": 20230 }, { "epoch": 6.22, "learning_rate": 6.597857375755055e-06, "loss": 0.1596, "step": 20231 }, { "epoch": 6.22, "learning_rate": 6.596920298219218e-06, "loss": 0.1018, "step": 20232 }, { "epoch": 6.23, "learning_rate": 6.595983254479096e-06, "loss": 0.0702, "step": 20233 }, { "epoch": 6.23, "learning_rate": 6.595046244543999e-06, "loss": 0.1003, "step": 20234 }, { "epoch": 6.23, "learning_rate": 6.59410926842323e-06, "loss": 0.1198, "step": 20235 }, { "epoch": 6.23, "learning_rate": 6.593172326126098e-06, "loss": 0.0887, "step": 20236 }, { "epoch": 6.23, "learning_rate": 6.592235417661903e-06, "loss": 0.0815, "step": 20237 }, { "epoch": 6.23, "learning_rate": 6.591298543039949e-06, "loss": 0.1422, "step": 20238 }, { "epoch": 6.23, "learning_rate": 6.5903617022695435e-06, "loss": 0.1121, "step": 20239 }, { "epoch": 6.23, "learning_rate": 6.589424895359988e-06, "loss": 0.0518, "step": 20240 }, { "epoch": 6.23, "learning_rate": 6.588488122320586e-06, "loss": 0.0817, "step": 20241 }, { "epoch": 6.23, "learning_rate": 6.587551383160646e-06, "loss": 0.1096, "step": 20242 }, { "epoch": 6.23, "learning_rate": 6.58661467788946e-06, "loss": 0.0467, "step": 20243 }, { "epoch": 6.23, "learning_rate": 6.58567800651634e-06, "loss": 0.0625, "step": 20244 }, { "epoch": 6.23, "learning_rate": 6.58474136905058e-06, "loss": 0.1142, "step": 20245 }, { "epoch": 6.23, "learning_rate": 6.583804765501489e-06, "loss": 0.0524, "step": 20246 }, { "epoch": 6.23, "learning_rate": 6.582868195878367e-06, "loss": 0.0696, "step": 20247 }, { "epoch": 6.23, "learning_rate": 6.581931660190508e-06, "loss": 0.0899, "step": 20248 }, { "epoch": 6.23, "learning_rate": 6.580995158447221e-06, "loss": 0.0918, "step": 20249 }, { "epoch": 6.23, "learning_rate": 6.580058690657801e-06, "loss": 0.0577, "step": 20250 }, { "epoch": 6.23, "learning_rate": 6.579122256831551e-06, "loss": 0.0918, "step": 20251 }, { "epoch": 6.23, "learning_rate": 6.5781858569777745e-06, "loss": 0.1018, "step": 20252 }, { "epoch": 6.23, "learning_rate": 6.5772494911057575e-06, "loss": 0.1518, "step": 20253 }, { "epoch": 6.23, "learning_rate": 6.5763131592248144e-06, "loss": 0.0947, "step": 20254 }, { "epoch": 6.23, "learning_rate": 6.575376861344233e-06, "loss": 0.1303, "step": 20255 }, { "epoch": 6.23, "learning_rate": 6.574440597473317e-06, "loss": 0.0815, "step": 20256 }, { "epoch": 6.23, "learning_rate": 6.5735043676213665e-06, "loss": 0.0723, "step": 20257 }, { "epoch": 6.23, "learning_rate": 6.572568171797669e-06, "loss": 0.115, "step": 20258 }, { "epoch": 6.23, "learning_rate": 6.571632010011533e-06, "loss": 0.1041, "step": 20259 }, { "epoch": 6.23, "learning_rate": 6.570695882272248e-06, "loss": 0.0377, "step": 20260 }, { "epoch": 6.23, "learning_rate": 6.569759788589114e-06, "loss": 0.0614, "step": 20261 }, { "epoch": 6.23, "learning_rate": 6.568823728971428e-06, "loss": 0.1388, "step": 20262 }, { "epoch": 6.23, "learning_rate": 6.567887703428484e-06, "loss": 0.1205, "step": 20263 }, { "epoch": 6.23, "learning_rate": 6.566951711969581e-06, "loss": 0.1062, "step": 20264 }, { "epoch": 6.24, "learning_rate": 6.566015754604006e-06, "loss": 0.0945, "step": 20265 }, { "epoch": 6.24, "learning_rate": 6.565079831341063e-06, "loss": 0.1288, "step": 20266 }, { "epoch": 6.24, "learning_rate": 6.564143942190042e-06, "loss": 0.0893, "step": 20267 }, { "epoch": 6.24, "learning_rate": 6.563208087160238e-06, "loss": 0.1238, "step": 20268 }, { "epoch": 6.24, "learning_rate": 6.562272266260947e-06, "loss": 0.0954, "step": 20269 }, { "epoch": 6.24, "learning_rate": 6.561336479501455e-06, "loss": 0.0944, "step": 20270 }, { "epoch": 6.24, "learning_rate": 6.560400726891069e-06, "loss": 0.1558, "step": 20271 }, { "epoch": 6.24, "learning_rate": 6.5594650084390694e-06, "loss": 0.0989, "step": 20272 }, { "epoch": 6.24, "learning_rate": 6.558529324154752e-06, "loss": 0.0768, "step": 20273 }, { "epoch": 6.24, "learning_rate": 6.557593674047413e-06, "loss": 0.0708, "step": 20274 }, { "epoch": 6.24, "learning_rate": 6.5566580581263396e-06, "loss": 0.0744, "step": 20275 }, { "epoch": 6.24, "learning_rate": 6.555722476400827e-06, "loss": 0.0849, "step": 20276 }, { "epoch": 6.24, "learning_rate": 6.554786928880165e-06, "loss": 0.0699, "step": 20277 }, { "epoch": 6.24, "learning_rate": 6.55385141557364e-06, "loss": 0.1253, "step": 20278 }, { "epoch": 6.24, "learning_rate": 6.5529159364905496e-06, "loss": 0.0609, "step": 20279 }, { "epoch": 6.24, "learning_rate": 6.551980491640179e-06, "loss": 0.0855, "step": 20280 }, { "epoch": 6.24, "learning_rate": 6.551045081031824e-06, "loss": 0.1008, "step": 20281 }, { "epoch": 6.24, "learning_rate": 6.5501097046747674e-06, "loss": 0.1397, "step": 20282 }, { "epoch": 6.24, "learning_rate": 6.549174362578299e-06, "loss": 0.0589, "step": 20283 }, { "epoch": 6.24, "learning_rate": 6.548239054751712e-06, "loss": 0.106, "step": 20284 }, { "epoch": 6.24, "learning_rate": 6.547303781204289e-06, "loss": 0.1544, "step": 20285 }, { "epoch": 6.24, "learning_rate": 6.546368541945327e-06, "loss": 0.0766, "step": 20286 }, { "epoch": 6.24, "learning_rate": 6.5454333369841064e-06, "loss": 0.1252, "step": 20287 }, { "epoch": 6.24, "learning_rate": 6.5444981663299135e-06, "loss": 0.1362, "step": 20288 }, { "epoch": 6.24, "learning_rate": 6.5435630299920415e-06, "loss": 0.075, "step": 20289 }, { "epoch": 6.24, "learning_rate": 6.542627927979772e-06, "loss": 0.0682, "step": 20290 }, { "epoch": 6.24, "learning_rate": 6.541692860302395e-06, "loss": 0.0723, "step": 20291 }, { "epoch": 6.24, "learning_rate": 6.540757826969197e-06, "loss": 0.1236, "step": 20292 }, { "epoch": 6.24, "learning_rate": 6.539822827989459e-06, "loss": 0.1495, "step": 20293 }, { "epoch": 6.24, "learning_rate": 6.538887863372472e-06, "loss": 0.0297, "step": 20294 }, { "epoch": 6.24, "learning_rate": 6.537952933127513e-06, "loss": 0.1076, "step": 20295 }, { "epoch": 6.24, "learning_rate": 6.537018037263878e-06, "loss": 0.0339, "step": 20296 }, { "epoch": 6.24, "learning_rate": 6.536083175790846e-06, "loss": 0.0571, "step": 20297 }, { "epoch": 6.25, "learning_rate": 6.535148348717694e-06, "loss": 0.1502, "step": 20298 }, { "epoch": 6.25, "learning_rate": 6.534213556053718e-06, "loss": 0.1447, "step": 20299 }, { "epoch": 6.25, "learning_rate": 6.533278797808191e-06, "loss": 0.0986, "step": 20300 }, { "epoch": 6.25, "learning_rate": 6.5323440739904045e-06, "loss": 0.0372, "step": 20301 }, { "epoch": 6.25, "learning_rate": 6.531409384609638e-06, "loss": 0.0735, "step": 20302 }, { "epoch": 6.25, "learning_rate": 6.530474729675167e-06, "loss": 0.1005, "step": 20303 }, { "epoch": 6.25, "learning_rate": 6.529540109196286e-06, "loss": 0.0923, "step": 20304 }, { "epoch": 6.25, "learning_rate": 6.5286055231822665e-06, "loss": 0.0868, "step": 20305 }, { "epoch": 6.25, "learning_rate": 6.527670971642395e-06, "loss": 0.1495, "step": 20306 }, { "epoch": 6.25, "learning_rate": 6.526736454585954e-06, "loss": 0.0686, "step": 20307 }, { "epoch": 6.25, "learning_rate": 6.525801972022213e-06, "loss": 0.131, "step": 20308 }, { "epoch": 6.25, "learning_rate": 6.524867523960468e-06, "loss": 0.0784, "step": 20309 }, { "epoch": 6.25, "learning_rate": 6.5239331104099855e-06, "loss": 0.0463, "step": 20310 }, { "epoch": 6.25, "learning_rate": 6.522998731380055e-06, "loss": 0.0949, "step": 20311 }, { "epoch": 6.25, "learning_rate": 6.52206438687995e-06, "loss": 0.0467, "step": 20312 }, { "epoch": 6.25, "learning_rate": 6.521130076918949e-06, "loss": 0.1638, "step": 20313 }, { "epoch": 6.25, "learning_rate": 6.520195801506338e-06, "loss": 0.0512, "step": 20314 }, { "epoch": 6.25, "learning_rate": 6.519261560651384e-06, "loss": 0.1207, "step": 20315 }, { "epoch": 6.25, "learning_rate": 6.518327354363374e-06, "loss": 0.1048, "step": 20316 }, { "epoch": 6.25, "learning_rate": 6.517393182651582e-06, "loss": 0.1135, "step": 20317 }, { "epoch": 6.25, "learning_rate": 6.516459045525283e-06, "loss": 0.0943, "step": 20318 }, { "epoch": 6.25, "learning_rate": 6.515524942993761e-06, "loss": 0.0953, "step": 20319 }, { "epoch": 6.25, "learning_rate": 6.51459087506628e-06, "loss": 0.0777, "step": 20320 }, { "epoch": 6.25, "learning_rate": 6.513656841752132e-06, "loss": 0.0982, "step": 20321 }, { "epoch": 6.25, "learning_rate": 6.512722843060582e-06, "loss": 0.1002, "step": 20322 }, { "epoch": 6.25, "learning_rate": 6.511788879000905e-06, "loss": 0.0978, "step": 20323 }, { "epoch": 6.25, "learning_rate": 6.5108549495823815e-06, "loss": 0.1625, "step": 20324 }, { "epoch": 6.25, "learning_rate": 6.509921054814283e-06, "loss": 0.0486, "step": 20325 }, { "epoch": 6.25, "learning_rate": 6.508987194705888e-06, "loss": 0.086, "step": 20326 }, { "epoch": 6.25, "learning_rate": 6.508053369266466e-06, "loss": 0.0744, "step": 20327 }, { "epoch": 6.25, "learning_rate": 6.5071195785052895e-06, "loss": 0.0593, "step": 20328 }, { "epoch": 6.25, "learning_rate": 6.506185822431639e-06, "loss": 0.0941, "step": 20329 }, { "epoch": 6.26, "learning_rate": 6.50525210105478e-06, "loss": 0.0931, "step": 20330 }, { "epoch": 6.26, "learning_rate": 6.504318414383992e-06, "loss": 0.1343, "step": 20331 }, { "epoch": 6.26, "learning_rate": 6.503384762428542e-06, "loss": 0.0477, "step": 20332 }, { "epoch": 6.26, "learning_rate": 6.502451145197701e-06, "loss": 0.0676, "step": 20333 }, { "epoch": 6.26, "learning_rate": 6.501517562700747e-06, "loss": 0.1236, "step": 20334 }, { "epoch": 6.26, "learning_rate": 6.500584014946947e-06, "loss": 0.1162, "step": 20335 }, { "epoch": 6.26, "learning_rate": 6.499650501945575e-06, "loss": 0.0974, "step": 20336 }, { "epoch": 6.26, "learning_rate": 6.498717023705898e-06, "loss": 0.1062, "step": 20337 }, { "epoch": 6.26, "learning_rate": 6.4977835802371845e-06, "loss": 0.066, "step": 20338 }, { "epoch": 6.26, "learning_rate": 6.496850171548712e-06, "loss": 0.0641, "step": 20339 }, { "epoch": 6.26, "learning_rate": 6.495916797649743e-06, "loss": 0.1053, "step": 20340 }, { "epoch": 6.26, "learning_rate": 6.494983458549551e-06, "loss": 0.1568, "step": 20341 }, { "epoch": 6.26, "learning_rate": 6.494050154257408e-06, "loss": 0.1049, "step": 20342 }, { "epoch": 6.26, "learning_rate": 6.493116884782571e-06, "loss": 0.1131, "step": 20343 }, { "epoch": 6.26, "learning_rate": 6.492183650134318e-06, "loss": 0.0872, "step": 20344 }, { "epoch": 6.26, "learning_rate": 6.491250450321913e-06, "loss": 0.1016, "step": 20345 }, { "epoch": 6.26, "learning_rate": 6.490317285354628e-06, "loss": 0.083, "step": 20346 }, { "epoch": 6.26, "learning_rate": 6.4893841552417266e-06, "loss": 0.0556, "step": 20347 }, { "epoch": 6.26, "learning_rate": 6.488451059992473e-06, "loss": 0.0921, "step": 20348 }, { "epoch": 6.26, "learning_rate": 6.487517999616139e-06, "loss": 0.0842, "step": 20349 }, { "epoch": 6.26, "learning_rate": 6.486584974121988e-06, "loss": 0.0982, "step": 20350 }, { "epoch": 6.26, "learning_rate": 6.485651983519286e-06, "loss": 0.1045, "step": 20351 }, { "epoch": 6.26, "learning_rate": 6.484719027817301e-06, "loss": 0.1213, "step": 20352 }, { "epoch": 6.26, "learning_rate": 6.483786107025289e-06, "loss": 0.0898, "step": 20353 }, { "epoch": 6.26, "learning_rate": 6.482853221152529e-06, "loss": 0.0709, "step": 20354 }, { "epoch": 6.26, "learning_rate": 6.481920370208274e-06, "loss": 0.0967, "step": 20355 }, { "epoch": 6.26, "learning_rate": 6.480987554201794e-06, "loss": 0.0638, "step": 20356 }, { "epoch": 6.26, "learning_rate": 6.4800547731423525e-06, "loss": 0.1304, "step": 20357 }, { "epoch": 6.26, "learning_rate": 6.4791220270392065e-06, "loss": 0.1061, "step": 20358 }, { "epoch": 6.26, "learning_rate": 6.478189315901629e-06, "loss": 0.0716, "step": 20359 }, { "epoch": 6.26, "learning_rate": 6.477256639738872e-06, "loss": 0.1205, "step": 20360 }, { "epoch": 6.26, "learning_rate": 6.476323998560206e-06, "loss": 0.1138, "step": 20361 }, { "epoch": 6.26, "learning_rate": 6.47539139237489e-06, "loss": 0.0977, "step": 20362 }, { "epoch": 6.27, "learning_rate": 6.474458821192184e-06, "loss": 0.1255, "step": 20363 }, { "epoch": 6.27, "learning_rate": 6.4735262850213566e-06, "loss": 0.1772, "step": 20364 }, { "epoch": 6.27, "learning_rate": 6.472593783871657e-06, "loss": 0.0764, "step": 20365 }, { "epoch": 6.27, "learning_rate": 6.471661317752355e-06, "loss": 0.1138, "step": 20366 }, { "epoch": 6.27, "learning_rate": 6.470728886672708e-06, "loss": 0.0493, "step": 20367 }, { "epoch": 6.27, "learning_rate": 6.469796490641974e-06, "loss": 0.0572, "step": 20368 }, { "epoch": 6.27, "learning_rate": 6.468864129669418e-06, "loss": 0.1283, "step": 20369 }, { "epoch": 6.27, "learning_rate": 6.46793180376429e-06, "loss": 0.1034, "step": 20370 }, { "epoch": 6.27, "learning_rate": 6.466999512935861e-06, "loss": 0.0411, "step": 20371 }, { "epoch": 6.27, "learning_rate": 6.46606725719338e-06, "loss": 0.0682, "step": 20372 }, { "epoch": 6.27, "learning_rate": 6.4651350365461065e-06, "loss": 0.0319, "step": 20373 }, { "epoch": 6.27, "learning_rate": 6.4642028510033e-06, "loss": 0.0505, "step": 20374 }, { "epoch": 6.27, "learning_rate": 6.463270700574219e-06, "loss": 0.0639, "step": 20375 }, { "epoch": 6.27, "learning_rate": 6.462338585268122e-06, "loss": 0.1699, "step": 20376 }, { "epoch": 6.27, "learning_rate": 6.461406505094261e-06, "loss": 0.0713, "step": 20377 }, { "epoch": 6.27, "learning_rate": 6.460474460061894e-06, "loss": 0.061, "step": 20378 }, { "epoch": 6.27, "learning_rate": 6.4595424501802785e-06, "loss": 0.0799, "step": 20379 }, { "epoch": 6.27, "learning_rate": 6.458610475458666e-06, "loss": 0.0625, "step": 20380 }, { "epoch": 6.27, "learning_rate": 6.4576785359063225e-06, "loss": 0.1027, "step": 20381 }, { "epoch": 6.27, "learning_rate": 6.456746631532492e-06, "loss": 0.0714, "step": 20382 }, { "epoch": 6.27, "learning_rate": 6.4558147623464315e-06, "loss": 0.1392, "step": 20383 }, { "epoch": 6.27, "learning_rate": 6.4548829283573985e-06, "loss": 0.1105, "step": 20384 }, { "epoch": 6.27, "learning_rate": 6.453951129574644e-06, "loss": 0.1272, "step": 20385 }, { "epoch": 6.27, "learning_rate": 6.453019366007427e-06, "loss": 0.071, "step": 20386 }, { "epoch": 6.27, "learning_rate": 6.452087637664993e-06, "loss": 0.0803, "step": 20387 }, { "epoch": 6.27, "learning_rate": 6.451155944556598e-06, "loss": 0.1533, "step": 20388 }, { "epoch": 6.27, "learning_rate": 6.450224286691496e-06, "loss": 0.0858, "step": 20389 }, { "epoch": 6.27, "learning_rate": 6.449292664078939e-06, "loss": 0.0646, "step": 20390 }, { "epoch": 6.27, "learning_rate": 6.448361076728178e-06, "loss": 0.1335, "step": 20391 }, { "epoch": 6.27, "learning_rate": 6.4474295246484675e-06, "loss": 0.0646, "step": 20392 }, { "epoch": 6.27, "learning_rate": 6.4464980078490515e-06, "loss": 0.1074, "step": 20393 }, { "epoch": 6.27, "learning_rate": 6.445566526339187e-06, "loss": 0.1006, "step": 20394 }, { "epoch": 6.28, "learning_rate": 6.444635080128121e-06, "loss": 0.0443, "step": 20395 }, { "epoch": 6.28, "learning_rate": 6.443703669225109e-06, "loss": 0.092, "step": 20396 }, { "epoch": 6.28, "learning_rate": 6.4427722936393936e-06, "loss": 0.1319, "step": 20397 }, { "epoch": 6.28, "learning_rate": 6.441840953380232e-06, "loss": 0.0576, "step": 20398 }, { "epoch": 6.28, "learning_rate": 6.4409096484568674e-06, "loss": 0.0645, "step": 20399 }, { "epoch": 6.28, "learning_rate": 6.439978378878547e-06, "loss": 0.086, "step": 20400 }, { "epoch": 6.28, "learning_rate": 6.439047144654526e-06, "loss": 0.092, "step": 20401 }, { "epoch": 6.28, "learning_rate": 6.438115945794046e-06, "loss": 0.0725, "step": 20402 }, { "epoch": 6.28, "learning_rate": 6.437184782306358e-06, "loss": 0.0641, "step": 20403 }, { "epoch": 6.28, "learning_rate": 6.436253654200712e-06, "loss": 0.0747, "step": 20404 }, { "epoch": 6.28, "learning_rate": 6.435322561486348e-06, "loss": 0.0855, "step": 20405 }, { "epoch": 6.28, "learning_rate": 6.434391504172518e-06, "loss": 0.0277, "step": 20406 }, { "epoch": 6.28, "learning_rate": 6.4334604822684645e-06, "loss": 0.0708, "step": 20407 }, { "epoch": 6.28, "learning_rate": 6.4325294957834375e-06, "loss": 0.0632, "step": 20408 }, { "epoch": 6.28, "learning_rate": 6.4315985447266825e-06, "loss": 0.075, "step": 20409 }, { "epoch": 6.28, "learning_rate": 6.430667629107439e-06, "loss": 0.1004, "step": 20410 }, { "epoch": 6.28, "learning_rate": 6.429736748934956e-06, "loss": 0.0876, "step": 20411 }, { "epoch": 6.28, "learning_rate": 6.428805904218476e-06, "loss": 0.0928, "step": 20412 }, { "epoch": 6.28, "learning_rate": 6.4278750949672485e-06, "loss": 0.2553, "step": 20413 }, { "epoch": 6.28, "learning_rate": 6.426944321190513e-06, "loss": 0.0819, "step": 20414 }, { "epoch": 6.28, "learning_rate": 6.426013582897511e-06, "loss": 0.1416, "step": 20415 }, { "epoch": 6.28, "learning_rate": 6.42508288009749e-06, "loss": 0.1112, "step": 20416 }, { "epoch": 6.28, "learning_rate": 6.424152212799688e-06, "loss": 0.131, "step": 20417 }, { "epoch": 6.28, "learning_rate": 6.423221581013353e-06, "loss": 0.1057, "step": 20418 }, { "epoch": 6.28, "learning_rate": 6.422290984747726e-06, "loss": 0.0522, "step": 20419 }, { "epoch": 6.28, "learning_rate": 6.421360424012039e-06, "loss": 0.0555, "step": 20420 }, { "epoch": 6.28, "learning_rate": 6.42042989881555e-06, "loss": 0.0941, "step": 20421 }, { "epoch": 6.28, "learning_rate": 6.4194994091674855e-06, "loss": 0.1044, "step": 20422 }, { "epoch": 6.28, "learning_rate": 6.4185689550770935e-06, "loss": 0.1303, "step": 20423 }, { "epoch": 6.28, "learning_rate": 6.417638536553617e-06, "loss": 0.2622, "step": 20424 }, { "epoch": 6.28, "learning_rate": 6.416708153606282e-06, "loss": 0.0776, "step": 20425 }, { "epoch": 6.28, "learning_rate": 6.415777806244345e-06, "loss": 0.1535, "step": 20426 }, { "epoch": 6.28, "learning_rate": 6.414847494477034e-06, "loss": 0.0909, "step": 20427 }, { "epoch": 6.29, "learning_rate": 6.413917218313592e-06, "loss": 0.0714, "step": 20428 }, { "epoch": 6.29, "learning_rate": 6.412986977763259e-06, "loss": 0.0982, "step": 20429 }, { "epoch": 6.29, "learning_rate": 6.412056772835269e-06, "loss": 0.0807, "step": 20430 }, { "epoch": 6.29, "learning_rate": 6.411126603538865e-06, "loss": 0.1027, "step": 20431 }, { "epoch": 6.29, "learning_rate": 6.4101964698832765e-06, "loss": 0.1175, "step": 20432 }, { "epoch": 6.29, "learning_rate": 6.409266371877751e-06, "loss": 0.0689, "step": 20433 }, { "epoch": 6.29, "learning_rate": 6.408336309531518e-06, "loss": 0.1228, "step": 20434 }, { "epoch": 6.29, "learning_rate": 6.407406282853812e-06, "loss": 0.0783, "step": 20435 }, { "epoch": 6.29, "learning_rate": 6.406476291853878e-06, "loss": 0.1097, "step": 20436 }, { "epoch": 6.29, "learning_rate": 6.405546336540941e-06, "loss": 0.0657, "step": 20437 }, { "epoch": 6.29, "learning_rate": 6.4046164169242475e-06, "loss": 0.1297, "step": 20438 }, { "epoch": 6.29, "learning_rate": 6.403686533013023e-06, "loss": 0.0737, "step": 20439 }, { "epoch": 6.29, "learning_rate": 6.402756684816505e-06, "loss": 0.0778, "step": 20440 }, { "epoch": 6.29, "learning_rate": 6.4018268723439304e-06, "loss": 0.0392, "step": 20441 }, { "epoch": 6.29, "learning_rate": 6.400897095604528e-06, "loss": 0.0942, "step": 20442 }, { "epoch": 6.29, "learning_rate": 6.399967354607539e-06, "loss": 0.1108, "step": 20443 }, { "epoch": 6.29, "learning_rate": 6.399037649362191e-06, "loss": 0.0976, "step": 20444 }, { "epoch": 6.29, "learning_rate": 6.398107979877714e-06, "loss": 0.0865, "step": 20445 }, { "epoch": 6.29, "learning_rate": 6.397178346163348e-06, "loss": 0.069, "step": 20446 }, { "epoch": 6.29, "learning_rate": 6.3962487482283195e-06, "loss": 0.0759, "step": 20447 }, { "epoch": 6.29, "learning_rate": 6.395319186081866e-06, "loss": 0.0996, "step": 20448 }, { "epoch": 6.29, "learning_rate": 6.394389659733214e-06, "loss": 0.0661, "step": 20449 }, { "epoch": 6.29, "learning_rate": 6.393460169191592e-06, "loss": 0.0676, "step": 20450 }, { "epoch": 6.29, "learning_rate": 6.392530714466236e-06, "loss": 0.0787, "step": 20451 }, { "epoch": 6.29, "learning_rate": 6.391601295566375e-06, "loss": 0.1061, "step": 20452 }, { "epoch": 6.29, "learning_rate": 6.39067191250124e-06, "loss": 0.1226, "step": 20453 }, { "epoch": 6.29, "learning_rate": 6.3897425652800606e-06, "loss": 0.0872, "step": 20454 }, { "epoch": 6.29, "learning_rate": 6.388813253912061e-06, "loss": 0.1016, "step": 20455 }, { "epoch": 6.29, "learning_rate": 6.387883978406476e-06, "loss": 0.1225, "step": 20456 }, { "epoch": 6.29, "learning_rate": 6.38695473877253e-06, "loss": 0.1109, "step": 20457 }, { "epoch": 6.29, "learning_rate": 6.386025535019456e-06, "loss": 0.0611, "step": 20458 }, { "epoch": 6.29, "learning_rate": 6.38509636715648e-06, "loss": 0.0734, "step": 20459 }, { "epoch": 6.3, "learning_rate": 6.384167235192827e-06, "loss": 0.0556, "step": 20460 }, { "epoch": 6.3, "learning_rate": 6.3832381391377255e-06, "loss": 0.0679, "step": 20461 }, { "epoch": 6.3, "learning_rate": 6.382309079000402e-06, "loss": 0.096, "step": 20462 }, { "epoch": 6.3, "learning_rate": 6.381380054790084e-06, "loss": 0.1275, "step": 20463 }, { "epoch": 6.3, "learning_rate": 6.380451066516e-06, "loss": 0.1097, "step": 20464 }, { "epoch": 6.3, "learning_rate": 6.379522114187369e-06, "loss": 0.0859, "step": 20465 }, { "epoch": 6.3, "learning_rate": 6.378593197813421e-06, "loss": 0.1207, "step": 20466 }, { "epoch": 6.3, "learning_rate": 6.377664317403379e-06, "loss": 0.0635, "step": 20467 }, { "epoch": 6.3, "learning_rate": 6.376735472966471e-06, "loss": 0.0972, "step": 20468 }, { "epoch": 6.3, "learning_rate": 6.375806664511921e-06, "loss": 0.1029, "step": 20469 }, { "epoch": 6.3, "learning_rate": 6.374877892048944e-06, "loss": 0.0672, "step": 20470 }, { "epoch": 6.3, "learning_rate": 6.3739491555867785e-06, "loss": 0.0977, "step": 20471 }, { "epoch": 6.3, "learning_rate": 6.373020455134633e-06, "loss": 0.1234, "step": 20472 }, { "epoch": 6.3, "learning_rate": 6.372091790701742e-06, "loss": 0.049, "step": 20473 }, { "epoch": 6.3, "learning_rate": 6.371163162297322e-06, "loss": 0.0879, "step": 20474 }, { "epoch": 6.3, "learning_rate": 6.3702345699305935e-06, "loss": 0.1328, "step": 20475 }, { "epoch": 6.3, "learning_rate": 6.369306013610785e-06, "loss": 0.1005, "step": 20476 }, { "epoch": 6.3, "learning_rate": 6.368377493347111e-06, "loss": 0.0855, "step": 20477 }, { "epoch": 6.3, "learning_rate": 6.367449009148796e-06, "loss": 0.2239, "step": 20478 }, { "epoch": 6.3, "learning_rate": 6.3665205610250615e-06, "loss": 0.0652, "step": 20479 }, { "epoch": 6.3, "learning_rate": 6.365592148985124e-06, "loss": 0.0547, "step": 20480 }, { "epoch": 6.3, "learning_rate": 6.364663773038211e-06, "loss": 0.0935, "step": 20481 }, { "epoch": 6.3, "learning_rate": 6.36373543319353e-06, "loss": 0.0744, "step": 20482 }, { "epoch": 6.3, "learning_rate": 6.3628071294603135e-06, "loss": 0.0618, "step": 20483 }, { "epoch": 6.3, "learning_rate": 6.361878861847772e-06, "loss": 0.0565, "step": 20484 }, { "epoch": 6.3, "learning_rate": 6.360950630365126e-06, "loss": 0.0948, "step": 20485 }, { "epoch": 6.3, "learning_rate": 6.360022435021597e-06, "loss": 0.0811, "step": 20486 }, { "epoch": 6.3, "learning_rate": 6.3590942758263945e-06, "loss": 0.0812, "step": 20487 }, { "epoch": 6.3, "learning_rate": 6.358166152788745e-06, "loss": 0.0875, "step": 20488 }, { "epoch": 6.3, "learning_rate": 6.357238065917864e-06, "loss": 0.0953, "step": 20489 }, { "epoch": 6.3, "learning_rate": 6.35631001522296e-06, "loss": 0.0895, "step": 20490 }, { "epoch": 6.3, "learning_rate": 6.35538200071326e-06, "loss": 0.1213, "step": 20491 }, { "epoch": 6.3, "learning_rate": 6.354454022397974e-06, "loss": 0.1513, "step": 20492 }, { "epoch": 6.31, "learning_rate": 6.353526080286323e-06, "loss": 0.1088, "step": 20493 }, { "epoch": 6.31, "learning_rate": 6.352598174387516e-06, "loss": 0.0993, "step": 20494 }, { "epoch": 6.31, "learning_rate": 6.351670304710771e-06, "loss": 0.0601, "step": 20495 }, { "epoch": 6.31, "learning_rate": 6.350742471265302e-06, "loss": 0.1027, "step": 20496 }, { "epoch": 6.31, "learning_rate": 6.349814674060322e-06, "loss": 0.1449, "step": 20497 }, { "epoch": 6.31, "learning_rate": 6.3488869131050505e-06, "loss": 0.1018, "step": 20498 }, { "epoch": 6.31, "learning_rate": 6.347959188408694e-06, "loss": 0.1073, "step": 20499 }, { "epoch": 6.31, "learning_rate": 6.347031499980468e-06, "loss": 0.1047, "step": 20500 }, { "epoch": 6.31, "learning_rate": 6.346103847829588e-06, "loss": 0.1538, "step": 20501 }, { "epoch": 6.31, "learning_rate": 6.345176231965262e-06, "loss": 0.0894, "step": 20502 }, { "epoch": 6.31, "learning_rate": 6.344248652396707e-06, "loss": 0.0831, "step": 20503 }, { "epoch": 6.31, "learning_rate": 6.343321109133132e-06, "loss": 0.0892, "step": 20504 }, { "epoch": 6.31, "learning_rate": 6.342393602183745e-06, "loss": 0.0956, "step": 20505 }, { "epoch": 6.31, "learning_rate": 6.341466131557764e-06, "loss": 0.0579, "step": 20506 }, { "epoch": 6.31, "learning_rate": 6.3405386972643935e-06, "loss": 0.0616, "step": 20507 }, { "epoch": 6.31, "learning_rate": 6.339611299312849e-06, "loss": 0.0977, "step": 20508 }, { "epoch": 6.31, "learning_rate": 6.338683937712337e-06, "loss": 0.0941, "step": 20509 }, { "epoch": 6.31, "learning_rate": 6.337756612472067e-06, "loss": 0.0915, "step": 20510 }, { "epoch": 6.31, "learning_rate": 6.33682932360125e-06, "loss": 0.0505, "step": 20511 }, { "epoch": 6.31, "learning_rate": 6.335902071109092e-06, "loss": 0.1047, "step": 20512 }, { "epoch": 6.31, "learning_rate": 6.3349748550048055e-06, "loss": 0.1133, "step": 20513 }, { "epoch": 6.31, "learning_rate": 6.3340476752975975e-06, "loss": 0.031, "step": 20514 }, { "epoch": 6.31, "learning_rate": 6.333120531996672e-06, "loss": 0.0311, "step": 20515 }, { "epoch": 6.31, "learning_rate": 6.33219342511124e-06, "loss": 0.0888, "step": 20516 }, { "epoch": 6.31, "learning_rate": 6.331266354650507e-06, "loss": 0.1135, "step": 20517 }, { "epoch": 6.31, "learning_rate": 6.330339320623681e-06, "loss": 0.0851, "step": 20518 }, { "epoch": 6.31, "learning_rate": 6.32941232303997e-06, "loss": 0.0936, "step": 20519 }, { "epoch": 6.31, "learning_rate": 6.328485361908572e-06, "loss": 0.0976, "step": 20520 }, { "epoch": 6.31, "learning_rate": 6.327558437238703e-06, "loss": 0.0436, "step": 20521 }, { "epoch": 6.31, "learning_rate": 6.32663154903956e-06, "loss": 0.0659, "step": 20522 }, { "epoch": 6.31, "learning_rate": 6.325704697320354e-06, "loss": 0.1396, "step": 20523 }, { "epoch": 6.31, "learning_rate": 6.324777882090287e-06, "loss": 0.1088, "step": 20524 }, { "epoch": 6.32, "learning_rate": 6.323851103358558e-06, "loss": 0.1208, "step": 20525 }, { "epoch": 6.32, "learning_rate": 6.322924361134382e-06, "loss": 0.1203, "step": 20526 }, { "epoch": 6.32, "learning_rate": 6.3219976554269515e-06, "loss": 0.1114, "step": 20527 }, { "epoch": 6.32, "learning_rate": 6.321070986245475e-06, "loss": 0.1077, "step": 20528 }, { "epoch": 6.32, "learning_rate": 6.320144353599156e-06, "loss": 0.0616, "step": 20529 }, { "epoch": 6.32, "learning_rate": 6.319217757497192e-06, "loss": 0.1029, "step": 20530 }, { "epoch": 6.32, "learning_rate": 6.3182911979487915e-06, "loss": 0.1196, "step": 20531 }, { "epoch": 6.32, "learning_rate": 6.317364674963149e-06, "loss": 0.0883, "step": 20532 }, { "epoch": 6.32, "learning_rate": 6.316438188549472e-06, "loss": 0.0905, "step": 20533 }, { "epoch": 6.32, "learning_rate": 6.315511738716958e-06, "loss": 0.1215, "step": 20534 }, { "epoch": 6.32, "learning_rate": 6.3145853254748065e-06, "loss": 0.1334, "step": 20535 }, { "epoch": 6.32, "learning_rate": 6.313658948832223e-06, "loss": 0.0815, "step": 20536 }, { "epoch": 6.32, "learning_rate": 6.3127326087983974e-06, "loss": 0.0647, "step": 20537 }, { "epoch": 6.32, "learning_rate": 6.311806305382542e-06, "loss": 0.1384, "step": 20538 }, { "epoch": 6.32, "learning_rate": 6.310880038593846e-06, "loss": 0.0705, "step": 20539 }, { "epoch": 6.32, "learning_rate": 6.30995380844151e-06, "loss": 0.1065, "step": 20540 }, { "epoch": 6.32, "learning_rate": 6.309027614934735e-06, "loss": 0.0955, "step": 20541 }, { "epoch": 6.32, "learning_rate": 6.308101458082716e-06, "loss": 0.0793, "step": 20542 }, { "epoch": 6.32, "learning_rate": 6.307175337894657e-06, "loss": 0.059, "step": 20543 }, { "epoch": 6.32, "learning_rate": 6.306249254379746e-06, "loss": 0.0814, "step": 20544 }, { "epoch": 6.32, "learning_rate": 6.3053232075471846e-06, "loss": 0.0982, "step": 20545 }, { "epoch": 6.32, "learning_rate": 6.304397197406169e-06, "loss": 0.0813, "step": 20546 }, { "epoch": 6.32, "learning_rate": 6.303471223965894e-06, "loss": 0.071, "step": 20547 }, { "epoch": 6.32, "learning_rate": 6.3025452872355606e-06, "loss": 0.0881, "step": 20548 }, { "epoch": 6.32, "learning_rate": 6.301619387224358e-06, "loss": 0.0925, "step": 20549 }, { "epoch": 6.32, "learning_rate": 6.300693523941481e-06, "loss": 0.0724, "step": 20550 }, { "epoch": 6.32, "learning_rate": 6.29976769739613e-06, "loss": 0.0886, "step": 20551 }, { "epoch": 6.32, "learning_rate": 6.298841907597493e-06, "loss": 0.085, "step": 20552 }, { "epoch": 6.32, "learning_rate": 6.29791615455477e-06, "loss": 0.1016, "step": 20553 }, { "epoch": 6.32, "learning_rate": 6.296990438277151e-06, "loss": 0.0911, "step": 20554 }, { "epoch": 6.32, "learning_rate": 6.296064758773828e-06, "loss": 0.0531, "step": 20555 }, { "epoch": 6.32, "learning_rate": 6.295139116053997e-06, "loss": 0.1003, "step": 20556 }, { "epoch": 6.32, "learning_rate": 6.2942135101268465e-06, "loss": 0.0938, "step": 20557 }, { "epoch": 6.33, "learning_rate": 6.2932879410015735e-06, "loss": 0.09, "step": 20558 }, { "epoch": 6.33, "learning_rate": 6.292362408687369e-06, "loss": 0.0297, "step": 20559 }, { "epoch": 6.33, "learning_rate": 6.291436913193421e-06, "loss": 0.0494, "step": 20560 }, { "epoch": 6.33, "learning_rate": 6.2905114545289225e-06, "loss": 0.1582, "step": 20561 }, { "epoch": 6.33, "learning_rate": 6.289586032703062e-06, "loss": 0.1036, "step": 20562 }, { "epoch": 6.33, "learning_rate": 6.2886606477250345e-06, "loss": 0.0561, "step": 20563 }, { "epoch": 6.33, "learning_rate": 6.287735299604029e-06, "loss": 0.1298, "step": 20564 }, { "epoch": 6.33, "learning_rate": 6.28680998834923e-06, "loss": 0.1441, "step": 20565 }, { "epoch": 6.33, "learning_rate": 6.28588471396983e-06, "loss": 0.063, "step": 20566 }, { "epoch": 6.33, "learning_rate": 6.284959476475017e-06, "loss": 0.0995, "step": 20567 }, { "epoch": 6.33, "learning_rate": 6.284034275873982e-06, "loss": 0.0765, "step": 20568 }, { "epoch": 6.33, "learning_rate": 6.2831091121759135e-06, "loss": 0.1117, "step": 20569 }, { "epoch": 6.33, "learning_rate": 6.282183985389991e-06, "loss": 0.1308, "step": 20570 }, { "epoch": 6.33, "learning_rate": 6.281258895525415e-06, "loss": 0.0563, "step": 20571 }, { "epoch": 6.33, "learning_rate": 6.2803338425913615e-06, "loss": 0.0985, "step": 20572 }, { "epoch": 6.33, "learning_rate": 6.279408826597022e-06, "loss": 0.1016, "step": 20573 }, { "epoch": 6.33, "learning_rate": 6.278483847551584e-06, "loss": 0.0745, "step": 20574 }, { "epoch": 6.33, "learning_rate": 6.277558905464225e-06, "loss": 0.072, "step": 20575 }, { "epoch": 6.33, "learning_rate": 6.276634000344144e-06, "loss": 0.0856, "step": 20576 }, { "epoch": 6.33, "learning_rate": 6.2757091322005134e-06, "loss": 0.0641, "step": 20577 }, { "epoch": 6.33, "learning_rate": 6.274784301042527e-06, "loss": 0.0957, "step": 20578 }, { "epoch": 6.33, "learning_rate": 6.273859506879365e-06, "loss": 0.0899, "step": 20579 }, { "epoch": 6.33, "learning_rate": 6.2729347497202095e-06, "loss": 0.0497, "step": 20580 }, { "epoch": 6.33, "learning_rate": 6.272010029574251e-06, "loss": 0.0634, "step": 20581 }, { "epoch": 6.33, "learning_rate": 6.271085346450666e-06, "loss": 0.0664, "step": 20582 }, { "epoch": 6.33, "learning_rate": 6.270160700358642e-06, "loss": 0.0738, "step": 20583 }, { "epoch": 6.33, "learning_rate": 6.269236091307359e-06, "loss": 0.1685, "step": 20584 }, { "epoch": 6.33, "learning_rate": 6.268311519305999e-06, "loss": 0.1204, "step": 20585 }, { "epoch": 6.33, "learning_rate": 6.267386984363748e-06, "loss": 0.1051, "step": 20586 }, { "epoch": 6.33, "learning_rate": 6.266462486489777e-06, "loss": 0.0845, "step": 20587 }, { "epoch": 6.33, "learning_rate": 6.265538025693283e-06, "loss": 0.0949, "step": 20588 }, { "epoch": 6.33, "learning_rate": 6.264613601983435e-06, "loss": 0.0826, "step": 20589 }, { "epoch": 6.34, "learning_rate": 6.263689215369414e-06, "loss": 0.1014, "step": 20590 }, { "epoch": 6.34, "learning_rate": 6.2627648658604045e-06, "loss": 0.1001, "step": 20591 }, { "epoch": 6.34, "learning_rate": 6.2618405534655835e-06, "loss": 0.1116, "step": 20592 }, { "epoch": 6.34, "learning_rate": 6.2609162781941345e-06, "loss": 0.0738, "step": 20593 }, { "epoch": 6.34, "learning_rate": 6.259992040055229e-06, "loss": 0.154, "step": 20594 }, { "epoch": 6.34, "learning_rate": 6.25906783905805e-06, "loss": 0.1274, "step": 20595 }, { "epoch": 6.34, "learning_rate": 6.258143675211777e-06, "loss": 0.077, "step": 20596 }, { "epoch": 6.34, "learning_rate": 6.257219548525584e-06, "loss": 0.1773, "step": 20597 }, { "epoch": 6.34, "learning_rate": 6.256295459008654e-06, "loss": 0.1093, "step": 20598 }, { "epoch": 6.34, "learning_rate": 6.255371406670153e-06, "loss": 0.0841, "step": 20599 }, { "epoch": 6.34, "learning_rate": 6.254447391519271e-06, "loss": 0.1136, "step": 20600 }, { "epoch": 6.34, "learning_rate": 6.253523413565179e-06, "loss": 0.116, "step": 20601 }, { "epoch": 6.34, "learning_rate": 6.2525994728170495e-06, "loss": 0.0645, "step": 20602 }, { "epoch": 6.34, "learning_rate": 6.251675569284064e-06, "loss": 0.0819, "step": 20603 }, { "epoch": 6.34, "learning_rate": 6.250751702975388e-06, "loss": 0.082, "step": 20604 }, { "epoch": 6.34, "learning_rate": 6.249827873900209e-06, "loss": 0.1489, "step": 20605 }, { "epoch": 6.34, "learning_rate": 6.248904082067695e-06, "loss": 0.098, "step": 20606 }, { "epoch": 6.34, "learning_rate": 6.247980327487017e-06, "loss": 0.1235, "step": 20607 }, { "epoch": 6.34, "learning_rate": 6.247056610167357e-06, "loss": 0.1058, "step": 20608 }, { "epoch": 6.34, "learning_rate": 6.24613293011788e-06, "loss": 0.0533, "step": 20609 }, { "epoch": 6.34, "learning_rate": 6.245209287347767e-06, "loss": 0.0616, "step": 20610 }, { "epoch": 6.34, "learning_rate": 6.244285681866182e-06, "loss": 0.101, "step": 20611 }, { "epoch": 6.34, "learning_rate": 6.243362113682303e-06, "loss": 0.12, "step": 20612 }, { "epoch": 6.34, "learning_rate": 6.242438582805301e-06, "loss": 0.1103, "step": 20613 }, { "epoch": 6.34, "learning_rate": 6.241515089244343e-06, "loss": 0.0608, "step": 20614 }, { "epoch": 6.34, "learning_rate": 6.2405916330086106e-06, "loss": 0.0515, "step": 20615 }, { "epoch": 6.34, "learning_rate": 6.239668214107266e-06, "loss": 0.1204, "step": 20616 }, { "epoch": 6.34, "learning_rate": 6.23874483254948e-06, "loss": 0.077, "step": 20617 }, { "epoch": 6.34, "learning_rate": 6.237821488344425e-06, "loss": 0.0875, "step": 20618 }, { "epoch": 6.34, "learning_rate": 6.236898181501268e-06, "loss": 0.0852, "step": 20619 }, { "epoch": 6.34, "learning_rate": 6.235974912029181e-06, "loss": 0.0606, "step": 20620 }, { "epoch": 6.34, "learning_rate": 6.235051679937336e-06, "loss": 0.0816, "step": 20621 }, { "epoch": 6.34, "learning_rate": 6.234128485234894e-06, "loss": 0.1331, "step": 20622 }, { "epoch": 6.35, "learning_rate": 6.2332053279310275e-06, "loss": 0.0884, "step": 20623 }, { "epoch": 6.35, "learning_rate": 6.232282208034903e-06, "loss": 0.1226, "step": 20624 }, { "epoch": 6.35, "learning_rate": 6.231359125555689e-06, "loss": 0.078, "step": 20625 }, { "epoch": 6.35, "learning_rate": 6.230436080502556e-06, "loss": 0.0892, "step": 20626 }, { "epoch": 6.35, "learning_rate": 6.229513072884662e-06, "loss": 0.0736, "step": 20627 }, { "epoch": 6.35, "learning_rate": 6.2285901027111806e-06, "loss": 0.0914, "step": 20628 }, { "epoch": 6.35, "learning_rate": 6.227667169991272e-06, "loss": 0.1439, "step": 20629 }, { "epoch": 6.35, "learning_rate": 6.2267442747341085e-06, "loss": 0.0815, "step": 20630 }, { "epoch": 6.35, "learning_rate": 6.225821416948854e-06, "loss": 0.1504, "step": 20631 }, { "epoch": 6.35, "learning_rate": 6.2248985966446665e-06, "loss": 0.0911, "step": 20632 }, { "epoch": 6.35, "learning_rate": 6.223975813830716e-06, "loss": 0.0791, "step": 20633 }, { "epoch": 6.35, "learning_rate": 6.223053068516165e-06, "loss": 0.1247, "step": 20634 }, { "epoch": 6.35, "learning_rate": 6.222130360710181e-06, "loss": 0.1493, "step": 20635 }, { "epoch": 6.35, "learning_rate": 6.2212076904219245e-06, "loss": 0.0842, "step": 20636 }, { "epoch": 6.35, "learning_rate": 6.2202850576605534e-06, "loss": 0.0778, "step": 20637 }, { "epoch": 6.35, "learning_rate": 6.219362462435241e-06, "loss": 0.0871, "step": 20638 }, { "epoch": 6.35, "learning_rate": 6.2184399047551396e-06, "loss": 0.0897, "step": 20639 }, { "epoch": 6.35, "learning_rate": 6.217517384629417e-06, "loss": 0.1006, "step": 20640 }, { "epoch": 6.35, "learning_rate": 6.216594902067233e-06, "loss": 0.0778, "step": 20641 }, { "epoch": 6.35, "learning_rate": 6.215672457077746e-06, "loss": 0.126, "step": 20642 }, { "epoch": 6.35, "learning_rate": 6.214750049670124e-06, "loss": 0.0599, "step": 20643 }, { "epoch": 6.35, "learning_rate": 6.213827679853517e-06, "loss": 0.0808, "step": 20644 }, { "epoch": 6.35, "learning_rate": 6.212905347637095e-06, "loss": 0.0588, "step": 20645 }, { "epoch": 6.35, "learning_rate": 6.211983053030011e-06, "loss": 0.0997, "step": 20646 }, { "epoch": 6.35, "learning_rate": 6.211060796041426e-06, "loss": 0.1222, "step": 20647 }, { "epoch": 6.35, "learning_rate": 6.210138576680502e-06, "loss": 0.1563, "step": 20648 }, { "epoch": 6.35, "learning_rate": 6.20921639495639e-06, "loss": 0.0738, "step": 20649 }, { "epoch": 6.35, "learning_rate": 6.2082942508782576e-06, "loss": 0.1083, "step": 20650 }, { "epoch": 6.35, "learning_rate": 6.207372144455257e-06, "loss": 0.1747, "step": 20651 }, { "epoch": 6.35, "learning_rate": 6.206450075696544e-06, "loss": 0.0856, "step": 20652 }, { "epoch": 6.35, "learning_rate": 6.205528044611282e-06, "loss": 0.0703, "step": 20653 }, { "epoch": 6.35, "learning_rate": 6.204606051208617e-06, "loss": 0.1128, "step": 20654 }, { "epoch": 6.36, "learning_rate": 6.20368409549772e-06, "loss": 0.0365, "step": 20655 }, { "epoch": 6.36, "learning_rate": 6.202762177487733e-06, "loss": 0.1044, "step": 20656 }, { "epoch": 6.36, "learning_rate": 6.201840297187818e-06, "loss": 0.0268, "step": 20657 }, { "epoch": 6.36, "learning_rate": 6.200918454607131e-06, "loss": 0.0696, "step": 20658 }, { "epoch": 6.36, "learning_rate": 6.199996649754822e-06, "loss": 0.1213, "step": 20659 }, { "epoch": 6.36, "learning_rate": 6.1990748826400535e-06, "loss": 0.0692, "step": 20660 }, { "epoch": 6.36, "learning_rate": 6.198153153271973e-06, "loss": 0.0645, "step": 20661 }, { "epoch": 6.36, "learning_rate": 6.197231461659732e-06, "loss": 0.1173, "step": 20662 }, { "epoch": 6.36, "learning_rate": 6.196309807812491e-06, "loss": 0.0742, "step": 20663 }, { "epoch": 6.36, "learning_rate": 6.1953881917393954e-06, "loss": 0.1635, "step": 20664 }, { "epoch": 6.36, "learning_rate": 6.1944666134496056e-06, "loss": 0.1135, "step": 20665 }, { "epoch": 6.36, "learning_rate": 6.193545072952269e-06, "loss": 0.0982, "step": 20666 }, { "epoch": 6.36, "learning_rate": 6.192623570256535e-06, "loss": 0.1102, "step": 20667 }, { "epoch": 6.36, "learning_rate": 6.19170210537156e-06, "loss": 0.1333, "step": 20668 }, { "epoch": 6.36, "learning_rate": 6.190780678306492e-06, "loss": 0.1013, "step": 20669 }, { "epoch": 6.36, "learning_rate": 6.189859289070483e-06, "loss": 0.0877, "step": 20670 }, { "epoch": 6.36, "learning_rate": 6.188937937672686e-06, "loss": 0.0974, "step": 20671 }, { "epoch": 6.36, "learning_rate": 6.188016624122243e-06, "loss": 0.1307, "step": 20672 }, { "epoch": 6.36, "learning_rate": 6.18709534842831e-06, "loss": 0.0792, "step": 20673 }, { "epoch": 6.36, "learning_rate": 6.186174110600033e-06, "loss": 0.0912, "step": 20674 }, { "epoch": 6.36, "learning_rate": 6.1852529106465635e-06, "loss": 0.0908, "step": 20675 }, { "epoch": 6.36, "learning_rate": 6.18433174857705e-06, "loss": 0.0566, "step": 20676 }, { "epoch": 6.36, "learning_rate": 6.183410624400635e-06, "loss": 0.0975, "step": 20677 }, { "epoch": 6.36, "learning_rate": 6.182489538126473e-06, "loss": 0.0864, "step": 20678 }, { "epoch": 6.36, "learning_rate": 6.181568489763705e-06, "loss": 0.0521, "step": 20679 }, { "epoch": 6.36, "learning_rate": 6.180647479321484e-06, "loss": 0.1276, "step": 20680 }, { "epoch": 6.36, "learning_rate": 6.179726506808954e-06, "loss": 0.1305, "step": 20681 }, { "epoch": 6.36, "learning_rate": 6.178805572235258e-06, "loss": 0.1416, "step": 20682 }, { "epoch": 6.36, "learning_rate": 6.177884675609547e-06, "loss": 0.1695, "step": 20683 }, { "epoch": 6.36, "learning_rate": 6.176963816940961e-06, "loss": 0.0922, "step": 20684 }, { "epoch": 6.36, "learning_rate": 6.176042996238649e-06, "loss": 0.0271, "step": 20685 }, { "epoch": 6.36, "learning_rate": 6.175122213511757e-06, "loss": 0.1201, "step": 20686 }, { "epoch": 6.36, "learning_rate": 6.174201468769421e-06, "loss": 0.0789, "step": 20687 }, { "epoch": 6.37, "learning_rate": 6.173280762020795e-06, "loss": 0.0577, "step": 20688 }, { "epoch": 6.37, "learning_rate": 6.172360093275015e-06, "loss": 0.0902, "step": 20689 }, { "epoch": 6.37, "learning_rate": 6.171439462541229e-06, "loss": 0.096, "step": 20690 }, { "epoch": 6.37, "learning_rate": 6.170518869828577e-06, "loss": 0.1117, "step": 20691 }, { "epoch": 6.37, "learning_rate": 6.169598315146199e-06, "loss": 0.1562, "step": 20692 }, { "epoch": 6.37, "learning_rate": 6.168677798503246e-06, "loss": 0.1624, "step": 20693 }, { "epoch": 6.37, "learning_rate": 6.167757319908848e-06, "loss": 0.0863, "step": 20694 }, { "epoch": 6.37, "learning_rate": 6.1668368793721545e-06, "loss": 0.0706, "step": 20695 }, { "epoch": 6.37, "learning_rate": 6.1659164769023024e-06, "loss": 0.1052, "step": 20696 }, { "epoch": 6.37, "learning_rate": 6.164996112508434e-06, "loss": 0.0641, "step": 20697 }, { "epoch": 6.37, "learning_rate": 6.164075786199691e-06, "loss": 0.0698, "step": 20698 }, { "epoch": 6.37, "learning_rate": 6.163155497985204e-06, "loss": 0.067, "step": 20699 }, { "epoch": 6.37, "learning_rate": 6.162235247874126e-06, "loss": 0.0836, "step": 20700 }, { "epoch": 6.37, "learning_rate": 6.161315035875586e-06, "loss": 0.0719, "step": 20701 }, { "epoch": 6.37, "learning_rate": 6.160394861998724e-06, "loss": 0.0771, "step": 20702 }, { "epoch": 6.37, "learning_rate": 6.159474726252684e-06, "loss": 0.1041, "step": 20703 }, { "epoch": 6.37, "learning_rate": 6.1585546286465936e-06, "loss": 0.1003, "step": 20704 }, { "epoch": 6.37, "learning_rate": 6.157634569189601e-06, "loss": 0.108, "step": 20705 }, { "epoch": 6.37, "learning_rate": 6.156714547890838e-06, "loss": 0.0687, "step": 20706 }, { "epoch": 6.37, "learning_rate": 6.1557945647594385e-06, "loss": 0.1625, "step": 20707 }, { "epoch": 6.37, "learning_rate": 6.1548746198045454e-06, "loss": 0.0395, "step": 20708 }, { "epoch": 6.37, "learning_rate": 6.153954713035288e-06, "loss": 0.1094, "step": 20709 }, { "epoch": 6.37, "learning_rate": 6.153034844460808e-06, "loss": 0.1201, "step": 20710 }, { "epoch": 6.37, "learning_rate": 6.1521150140902365e-06, "loss": 0.0783, "step": 20711 }, { "epoch": 6.37, "learning_rate": 6.151195221932706e-06, "loss": 0.0975, "step": 20712 }, { "epoch": 6.37, "learning_rate": 6.150275467997358e-06, "loss": 0.0989, "step": 20713 }, { "epoch": 6.37, "learning_rate": 6.14935575229332e-06, "loss": 0.1763, "step": 20714 }, { "epoch": 6.37, "learning_rate": 6.148436074829732e-06, "loss": 0.0776, "step": 20715 }, { "epoch": 6.37, "learning_rate": 6.147516435615721e-06, "loss": 0.0802, "step": 20716 }, { "epoch": 6.37, "learning_rate": 6.146596834660421e-06, "loss": 0.0968, "step": 20717 }, { "epoch": 6.37, "learning_rate": 6.145677271972969e-06, "loss": 0.0849, "step": 20718 }, { "epoch": 6.37, "learning_rate": 6.14475774756249e-06, "loss": 0.053, "step": 20719 }, { "epoch": 6.38, "learning_rate": 6.143838261438122e-06, "loss": 0.081, "step": 20720 }, { "epoch": 6.38, "learning_rate": 6.142918813608997e-06, "loss": 0.0446, "step": 20721 }, { "epoch": 6.38, "learning_rate": 6.141999404084238e-06, "loss": 0.1317, "step": 20722 }, { "epoch": 6.38, "learning_rate": 6.141080032872983e-06, "loss": 0.086, "step": 20723 }, { "epoch": 6.38, "learning_rate": 6.140160699984356e-06, "loss": 0.0959, "step": 20724 }, { "epoch": 6.38, "learning_rate": 6.139241405427495e-06, "loss": 0.1081, "step": 20725 }, { "epoch": 6.38, "learning_rate": 6.138322149211526e-06, "loss": 0.0622, "step": 20726 }, { "epoch": 6.38, "learning_rate": 6.137402931345573e-06, "loss": 0.0775, "step": 20727 }, { "epoch": 6.38, "learning_rate": 6.136483751838772e-06, "loss": 0.0923, "step": 20728 }, { "epoch": 6.38, "learning_rate": 6.135564610700244e-06, "loss": 0.0938, "step": 20729 }, { "epoch": 6.38, "learning_rate": 6.1346455079391254e-06, "loss": 0.0834, "step": 20730 }, { "epoch": 6.38, "learning_rate": 6.13372644356454e-06, "loss": 0.059, "step": 20731 }, { "epoch": 6.38, "learning_rate": 6.13280741758561e-06, "loss": 0.0908, "step": 20732 }, { "epoch": 6.38, "learning_rate": 6.131888430011469e-06, "loss": 0.0977, "step": 20733 }, { "epoch": 6.38, "learning_rate": 6.130969480851239e-06, "loss": 0.0913, "step": 20734 }, { "epoch": 6.38, "learning_rate": 6.130050570114049e-06, "loss": 0.0779, "step": 20735 }, { "epoch": 6.38, "learning_rate": 6.129131697809025e-06, "loss": 0.0822, "step": 20736 }, { "epoch": 6.38, "learning_rate": 6.128212863945287e-06, "loss": 0.1215, "step": 20737 }, { "epoch": 6.38, "learning_rate": 6.127294068531968e-06, "loss": 0.0775, "step": 20738 }, { "epoch": 6.38, "learning_rate": 6.126375311578183e-06, "loss": 0.0634, "step": 20739 }, { "epoch": 6.38, "learning_rate": 6.125456593093066e-06, "loss": 0.1239, "step": 20740 }, { "epoch": 6.38, "learning_rate": 6.124537913085736e-06, "loss": 0.1626, "step": 20741 }, { "epoch": 6.38, "learning_rate": 6.123619271565311e-06, "loss": 0.0961, "step": 20742 }, { "epoch": 6.38, "learning_rate": 6.122700668540924e-06, "loss": 0.0765, "step": 20743 }, { "epoch": 6.38, "learning_rate": 6.121782104021691e-06, "loss": 0.0802, "step": 20744 }, { "epoch": 6.38, "learning_rate": 6.120863578016736e-06, "loss": 0.1068, "step": 20745 }, { "epoch": 6.38, "learning_rate": 6.119945090535182e-06, "loss": 0.0821, "step": 20746 }, { "epoch": 6.38, "learning_rate": 6.1190266415861465e-06, "loss": 0.1077, "step": 20747 }, { "epoch": 6.38, "learning_rate": 6.118108231178758e-06, "loss": 0.0947, "step": 20748 }, { "epoch": 6.38, "learning_rate": 6.117189859322126e-06, "loss": 0.0734, "step": 20749 }, { "epoch": 6.38, "learning_rate": 6.116271526025383e-06, "loss": 0.0925, "step": 20750 }, { "epoch": 6.38, "learning_rate": 6.115353231297641e-06, "loss": 0.1728, "step": 20751 }, { "epoch": 6.38, "learning_rate": 6.114434975148019e-06, "loss": 0.1083, "step": 20752 }, { "epoch": 6.39, "learning_rate": 6.113516757585645e-06, "loss": 0.1481, "step": 20753 }, { "epoch": 6.39, "learning_rate": 6.112598578619623e-06, "loss": 0.1082, "step": 20754 }, { "epoch": 6.39, "learning_rate": 6.111680438259087e-06, "loss": 0.096, "step": 20755 }, { "epoch": 6.39, "learning_rate": 6.110762336513145e-06, "loss": 0.1037, "step": 20756 }, { "epoch": 6.39, "learning_rate": 6.109844273390917e-06, "loss": 0.0491, "step": 20757 }, { "epoch": 6.39, "learning_rate": 6.108926248901521e-06, "loss": 0.0837, "step": 20758 }, { "epoch": 6.39, "learning_rate": 6.108008263054073e-06, "loss": 0.1045, "step": 20759 }, { "epoch": 6.39, "learning_rate": 6.107090315857693e-06, "loss": 0.1049, "step": 20760 }, { "epoch": 6.39, "learning_rate": 6.106172407321491e-06, "loss": 0.0756, "step": 20761 }, { "epoch": 6.39, "learning_rate": 6.105254537454586e-06, "loss": 0.0793, "step": 20762 }, { "epoch": 6.39, "learning_rate": 6.104336706266094e-06, "loss": 0.0612, "step": 20763 }, { "epoch": 6.39, "learning_rate": 6.103418913765127e-06, "loss": 0.1403, "step": 20764 }, { "epoch": 6.39, "learning_rate": 6.102501159960804e-06, "loss": 0.0903, "step": 20765 }, { "epoch": 6.39, "learning_rate": 6.101583444862236e-06, "loss": 0.1012, "step": 20766 }, { "epoch": 6.39, "learning_rate": 6.100665768478534e-06, "loss": 0.0722, "step": 20767 }, { "epoch": 6.39, "learning_rate": 6.0997481308188176e-06, "loss": 0.149, "step": 20768 }, { "epoch": 6.39, "learning_rate": 6.098830531892195e-06, "loss": 0.0996, "step": 20769 }, { "epoch": 6.39, "learning_rate": 6.097912971707784e-06, "loss": 0.0766, "step": 20770 }, { "epoch": 6.39, "learning_rate": 6.0969954502746916e-06, "loss": 0.091, "step": 20771 }, { "epoch": 6.39, "learning_rate": 6.09607796760203e-06, "loss": 0.1744, "step": 20772 }, { "epoch": 6.39, "learning_rate": 6.095160523698913e-06, "loss": 0.0972, "step": 20773 }, { "epoch": 6.39, "learning_rate": 6.09424311857445e-06, "loss": 0.0661, "step": 20774 }, { "epoch": 6.39, "learning_rate": 6.0933257522377545e-06, "loss": 0.0743, "step": 20775 }, { "epoch": 6.39, "learning_rate": 6.092408424697935e-06, "loss": 0.0601, "step": 20776 }, { "epoch": 6.39, "learning_rate": 6.0914911359640994e-06, "loss": 0.1052, "step": 20777 }, { "epoch": 6.39, "learning_rate": 6.090573886045361e-06, "loss": 0.0502, "step": 20778 }, { "epoch": 6.39, "learning_rate": 6.089656674950824e-06, "loss": 0.1119, "step": 20779 }, { "epoch": 6.39, "learning_rate": 6.088739502689603e-06, "loss": 0.1217, "step": 20780 }, { "epoch": 6.39, "learning_rate": 6.087822369270804e-06, "loss": 0.1032, "step": 20781 }, { "epoch": 6.39, "learning_rate": 6.086905274703531e-06, "loss": 0.0377, "step": 20782 }, { "epoch": 6.39, "learning_rate": 6.085988218996897e-06, "loss": 0.0673, "step": 20783 }, { "epoch": 6.39, "learning_rate": 6.0850712021600044e-06, "loss": 0.0747, "step": 20784 }, { "epoch": 6.4, "learning_rate": 6.084154224201966e-06, "loss": 0.2186, "step": 20785 }, { "epoch": 6.4, "learning_rate": 6.0832372851318865e-06, "loss": 0.0894, "step": 20786 }, { "epoch": 6.4, "learning_rate": 6.082320384958866e-06, "loss": 0.064, "step": 20787 }, { "epoch": 6.4, "learning_rate": 6.08140352369202e-06, "loss": 0.1482, "step": 20788 }, { "epoch": 6.4, "learning_rate": 6.080486701340444e-06, "loss": 0.1902, "step": 20789 }, { "epoch": 6.4, "learning_rate": 6.079569917913249e-06, "loss": 0.0926, "step": 20790 }, { "epoch": 6.4, "learning_rate": 6.078653173419537e-06, "loss": 0.0715, "step": 20791 }, { "epoch": 6.4, "learning_rate": 6.0777364678684134e-06, "loss": 0.1032, "step": 20792 }, { "epoch": 6.4, "learning_rate": 6.076819801268985e-06, "loss": 0.0523, "step": 20793 }, { "epoch": 6.4, "learning_rate": 6.0759031736303485e-06, "loss": 0.0524, "step": 20794 }, { "epoch": 6.4, "learning_rate": 6.074986584961611e-06, "loss": 0.0716, "step": 20795 }, { "epoch": 6.4, "learning_rate": 6.07407003527187e-06, "loss": 0.0786, "step": 20796 }, { "epoch": 6.4, "learning_rate": 6.073153524570236e-06, "loss": 0.1196, "step": 20797 }, { "epoch": 6.4, "learning_rate": 6.072237052865809e-06, "loss": 0.0847, "step": 20798 }, { "epoch": 6.4, "learning_rate": 6.0713206201676825e-06, "loss": 0.0647, "step": 20799 }, { "epoch": 6.4, "learning_rate": 6.070404226484967e-06, "loss": 0.1133, "step": 20800 }, { "epoch": 6.4, "learning_rate": 6.069487871826756e-06, "loss": 0.0524, "step": 20801 }, { "epoch": 6.4, "learning_rate": 6.068571556202154e-06, "loss": 0.1089, "step": 20802 }, { "epoch": 6.4, "learning_rate": 6.067655279620264e-06, "loss": 0.1089, "step": 20803 }, { "epoch": 6.4, "learning_rate": 6.066739042090172e-06, "loss": 0.0395, "step": 20804 }, { "epoch": 6.4, "learning_rate": 6.065822843620997e-06, "loss": 0.1168, "step": 20805 }, { "epoch": 6.4, "learning_rate": 6.064906684221819e-06, "loss": 0.0954, "step": 20806 }, { "epoch": 6.4, "learning_rate": 6.0639905639017475e-06, "loss": 0.0946, "step": 20807 }, { "epoch": 6.4, "learning_rate": 6.063074482669878e-06, "loss": 0.1552, "step": 20808 }, { "epoch": 6.4, "learning_rate": 6.062158440535304e-06, "loss": 0.1845, "step": 20809 }, { "epoch": 6.4, "learning_rate": 6.061242437507131e-06, "loss": 0.0781, "step": 20810 }, { "epoch": 6.4, "learning_rate": 6.060326473594445e-06, "loss": 0.1357, "step": 20811 }, { "epoch": 6.4, "learning_rate": 6.059410548806351e-06, "loss": 0.0642, "step": 20812 }, { "epoch": 6.4, "learning_rate": 6.058494663151943e-06, "loss": 0.0488, "step": 20813 }, { "epoch": 6.4, "learning_rate": 6.0575788166403125e-06, "loss": 0.1402, "step": 20814 }, { "epoch": 6.4, "learning_rate": 6.056663009280561e-06, "loss": 0.0895, "step": 20815 }, { "epoch": 6.4, "learning_rate": 6.055747241081775e-06, "loss": 0.0791, "step": 20816 }, { "epoch": 6.4, "learning_rate": 6.054831512053061e-06, "loss": 0.0819, "step": 20817 }, { "epoch": 6.41, "learning_rate": 6.053915822203504e-06, "loss": 0.0756, "step": 20818 }, { "epoch": 6.41, "learning_rate": 6.053000171542196e-06, "loss": 0.1413, "step": 20819 }, { "epoch": 6.41, "learning_rate": 6.0520845600782395e-06, "loss": 0.0905, "step": 20820 }, { "epoch": 6.41, "learning_rate": 6.051168987820715e-06, "loss": 0.1142, "step": 20821 }, { "epoch": 6.41, "learning_rate": 6.050253454778728e-06, "loss": 0.1114, "step": 20822 }, { "epoch": 6.41, "learning_rate": 6.049337960961362e-06, "loss": 0.0758, "step": 20823 }, { "epoch": 6.41, "learning_rate": 6.0484225063777094e-06, "loss": 0.1379, "step": 20824 }, { "epoch": 6.41, "learning_rate": 6.047507091036866e-06, "loss": 0.1377, "step": 20825 }, { "epoch": 6.41, "learning_rate": 6.046591714947917e-06, "loss": 0.1189, "step": 20826 }, { "epoch": 6.41, "learning_rate": 6.04567637811996e-06, "loss": 0.0898, "step": 20827 }, { "epoch": 6.41, "learning_rate": 6.044761080562078e-06, "loss": 0.1296, "step": 20828 }, { "epoch": 6.41, "learning_rate": 6.043845822283363e-06, "loss": 0.0982, "step": 20829 }, { "epoch": 6.41, "learning_rate": 6.042930603292907e-06, "loss": 0.1087, "step": 20830 }, { "epoch": 6.41, "learning_rate": 6.0420154235997955e-06, "loss": 0.0433, "step": 20831 }, { "epoch": 6.41, "learning_rate": 6.041100283213122e-06, "loss": 0.1274, "step": 20832 }, { "epoch": 6.41, "learning_rate": 6.040185182141969e-06, "loss": 0.0521, "step": 20833 }, { "epoch": 6.41, "learning_rate": 6.039270120395424e-06, "loss": 0.0917, "step": 20834 }, { "epoch": 6.41, "learning_rate": 6.03835509798258e-06, "loss": 0.0693, "step": 20835 }, { "epoch": 6.41, "learning_rate": 6.037440114912521e-06, "loss": 0.0631, "step": 20836 }, { "epoch": 6.41, "learning_rate": 6.036525171194332e-06, "loss": 0.0988, "step": 20837 }, { "epoch": 6.41, "learning_rate": 6.035610266837104e-06, "loss": 0.1374, "step": 20838 }, { "epoch": 6.41, "learning_rate": 6.034695401849917e-06, "loss": 0.088, "step": 20839 }, { "epoch": 6.41, "learning_rate": 6.033780576241861e-06, "loss": 0.061, "step": 20840 }, { "epoch": 6.41, "learning_rate": 6.032865790022015e-06, "loss": 0.0774, "step": 20841 }, { "epoch": 6.41, "learning_rate": 6.031951043199473e-06, "loss": 0.1162, "step": 20842 }, { "epoch": 6.41, "learning_rate": 6.0310363357833155e-06, "loss": 0.1431, "step": 20843 }, { "epoch": 6.41, "learning_rate": 6.03012166778262e-06, "loss": 0.1244, "step": 20844 }, { "epoch": 6.41, "learning_rate": 6.0292070392064774e-06, "loss": 0.049, "step": 20845 }, { "epoch": 6.41, "learning_rate": 6.028292450063966e-06, "loss": 0.0542, "step": 20846 }, { "epoch": 6.41, "learning_rate": 6.027377900364174e-06, "loss": 0.1116, "step": 20847 }, { "epoch": 6.41, "learning_rate": 6.026463390116182e-06, "loss": 0.1406, "step": 20848 }, { "epoch": 6.41, "learning_rate": 6.0255489193290675e-06, "loss": 0.0502, "step": 20849 }, { "epoch": 6.42, "learning_rate": 6.0246344880119165e-06, "loss": 0.148, "step": 20850 }, { "epoch": 6.42, "learning_rate": 6.0237200961738074e-06, "loss": 0.0909, "step": 20851 }, { "epoch": 6.42, "learning_rate": 6.022805743823824e-06, "loss": 0.1087, "step": 20852 }, { "epoch": 6.42, "learning_rate": 6.021891430971047e-06, "loss": 0.0884, "step": 20853 }, { "epoch": 6.42, "learning_rate": 6.0209771576245475e-06, "loss": 0.1135, "step": 20854 }, { "epoch": 6.42, "learning_rate": 6.02006292379342e-06, "loss": 0.1028, "step": 20855 }, { "epoch": 6.42, "learning_rate": 6.01914872948673e-06, "loss": 0.0528, "step": 20856 }, { "epoch": 6.42, "learning_rate": 6.018234574713565e-06, "loss": 0.114, "step": 20857 }, { "epoch": 6.42, "learning_rate": 6.017320459483e-06, "loss": 0.105, "step": 20858 }, { "epoch": 6.42, "learning_rate": 6.016406383804112e-06, "loss": 0.0922, "step": 20859 }, { "epoch": 6.42, "learning_rate": 6.015492347685983e-06, "loss": 0.079, "step": 20860 }, { "epoch": 6.42, "learning_rate": 6.014578351137683e-06, "loss": 0.0889, "step": 20861 }, { "epoch": 6.42, "learning_rate": 6.013664394168297e-06, "loss": 0.0745, "step": 20862 }, { "epoch": 6.42, "learning_rate": 6.012750476786895e-06, "loss": 0.1252, "step": 20863 }, { "epoch": 6.42, "learning_rate": 6.011836599002555e-06, "loss": 0.0745, "step": 20864 }, { "epoch": 6.42, "learning_rate": 6.010922760824357e-06, "loss": 0.0824, "step": 20865 }, { "epoch": 6.42, "learning_rate": 6.010008962261367e-06, "loss": 0.0949, "step": 20866 }, { "epoch": 6.42, "learning_rate": 6.00909520332267e-06, "loss": 0.0734, "step": 20867 }, { "epoch": 6.42, "learning_rate": 6.008181484017335e-06, "loss": 0.1068, "step": 20868 }, { "epoch": 6.42, "learning_rate": 6.007267804354433e-06, "loss": 0.0654, "step": 20869 }, { "epoch": 6.42, "learning_rate": 6.006354164343047e-06, "loss": 0.0667, "step": 20870 }, { "epoch": 6.42, "learning_rate": 6.005440563992239e-06, "loss": 0.0723, "step": 20871 }, { "epoch": 6.42, "learning_rate": 6.004527003311094e-06, "loss": 0.0672, "step": 20872 }, { "epoch": 6.42, "learning_rate": 6.003613482308675e-06, "loss": 0.0993, "step": 20873 }, { "epoch": 6.42, "learning_rate": 6.002700000994055e-06, "loss": 0.1083, "step": 20874 }, { "epoch": 6.42, "learning_rate": 6.00178655937631e-06, "loss": 0.0908, "step": 20875 }, { "epoch": 6.42, "learning_rate": 6.000873157464508e-06, "loss": 0.0864, "step": 20876 }, { "epoch": 6.42, "learning_rate": 5.999959795267726e-06, "loss": 0.1014, "step": 20877 }, { "epoch": 6.42, "learning_rate": 5.999046472795027e-06, "loss": 0.128, "step": 20878 }, { "epoch": 6.42, "learning_rate": 5.998133190055482e-06, "loss": 0.1184, "step": 20879 }, { "epoch": 6.42, "learning_rate": 5.997219947058164e-06, "loss": 0.0712, "step": 20880 }, { "epoch": 6.42, "learning_rate": 5.9963067438121395e-06, "loss": 0.1592, "step": 20881 }, { "epoch": 6.42, "learning_rate": 5.995393580326483e-06, "loss": 0.1592, "step": 20882 }, { "epoch": 6.43, "learning_rate": 5.994480456610258e-06, "loss": 0.1073, "step": 20883 }, { "epoch": 6.43, "learning_rate": 5.993567372672531e-06, "loss": 0.0995, "step": 20884 }, { "epoch": 6.43, "learning_rate": 5.9926543285223734e-06, "loss": 0.117, "step": 20885 }, { "epoch": 6.43, "learning_rate": 5.991741324168849e-06, "loss": 0.134, "step": 20886 }, { "epoch": 6.43, "learning_rate": 5.990828359621032e-06, "loss": 0.1205, "step": 20887 }, { "epoch": 6.43, "learning_rate": 5.989915434887985e-06, "loss": 0.0731, "step": 20888 }, { "epoch": 6.43, "learning_rate": 5.98900254997877e-06, "loss": 0.1437, "step": 20889 }, { "epoch": 6.43, "learning_rate": 5.988089704902458e-06, "loss": 0.0547, "step": 20890 }, { "epoch": 6.43, "learning_rate": 5.98717689966811e-06, "loss": 0.0924, "step": 20891 }, { "epoch": 6.43, "learning_rate": 5.986264134284796e-06, "loss": 0.0654, "step": 20892 }, { "epoch": 6.43, "learning_rate": 5.985351408761582e-06, "loss": 0.0608, "step": 20893 }, { "epoch": 6.43, "learning_rate": 5.984438723107522e-06, "loss": 0.1058, "step": 20894 }, { "epoch": 6.43, "learning_rate": 5.98352607733169e-06, "loss": 0.0557, "step": 20895 }, { "epoch": 6.43, "learning_rate": 5.982613471443143e-06, "loss": 0.0723, "step": 20896 }, { "epoch": 6.43, "learning_rate": 5.9817009054509485e-06, "loss": 0.1027, "step": 20897 }, { "epoch": 6.43, "learning_rate": 5.980788379364171e-06, "loss": 0.083, "step": 20898 }, { "epoch": 6.43, "learning_rate": 5.979875893191863e-06, "loss": 0.0805, "step": 20899 }, { "epoch": 6.43, "learning_rate": 5.978963446943095e-06, "loss": 0.0801, "step": 20900 }, { "epoch": 6.43, "learning_rate": 5.9780510406269245e-06, "loss": 0.0226, "step": 20901 }, { "epoch": 6.43, "learning_rate": 5.977138674252416e-06, "loss": 0.0841, "step": 20902 }, { "epoch": 6.43, "learning_rate": 5.97622634782863e-06, "loss": 0.056, "step": 20903 }, { "epoch": 6.43, "learning_rate": 5.975314061364619e-06, "loss": 0.1235, "step": 20904 }, { "epoch": 6.43, "learning_rate": 5.974401814869453e-06, "loss": 0.095, "step": 20905 }, { "epoch": 6.43, "learning_rate": 5.973489608352185e-06, "loss": 0.1235, "step": 20906 }, { "epoch": 6.43, "learning_rate": 5.972577441821878e-06, "loss": 0.0802, "step": 20907 }, { "epoch": 6.43, "learning_rate": 5.971665315287587e-06, "loss": 0.1401, "step": 20908 }, { "epoch": 6.43, "learning_rate": 5.970753228758371e-06, "loss": 0.0541, "step": 20909 }, { "epoch": 6.43, "learning_rate": 5.969841182243293e-06, "loss": 0.0559, "step": 20910 }, { "epoch": 6.43, "learning_rate": 5.968929175751403e-06, "loss": 0.1092, "step": 20911 }, { "epoch": 6.43, "learning_rate": 5.968017209291763e-06, "loss": 0.0764, "step": 20912 }, { "epoch": 6.43, "learning_rate": 5.967105282873428e-06, "loss": 0.0734, "step": 20913 }, { "epoch": 6.43, "learning_rate": 5.966193396505452e-06, "loss": 0.0946, "step": 20914 }, { "epoch": 6.44, "learning_rate": 5.9652815501968975e-06, "loss": 0.1466, "step": 20915 }, { "epoch": 6.44, "learning_rate": 5.96436974395681e-06, "loss": 0.0912, "step": 20916 }, { "epoch": 6.44, "learning_rate": 5.963457977794256e-06, "loss": 0.1311, "step": 20917 }, { "epoch": 6.44, "learning_rate": 5.962546251718282e-06, "loss": 0.0937, "step": 20918 }, { "epoch": 6.44, "learning_rate": 5.961634565737944e-06, "loss": 0.1104, "step": 20919 }, { "epoch": 6.44, "learning_rate": 5.960722919862301e-06, "loss": 0.1093, "step": 20920 }, { "epoch": 6.44, "learning_rate": 5.959811314100394e-06, "loss": 0.087, "step": 20921 }, { "epoch": 6.44, "learning_rate": 5.958899748461291e-06, "loss": 0.1096, "step": 20922 }, { "epoch": 6.44, "learning_rate": 5.957988222954036e-06, "loss": 0.096, "step": 20923 }, { "epoch": 6.44, "learning_rate": 5.9570767375876815e-06, "loss": 0.067, "step": 20924 }, { "epoch": 6.44, "learning_rate": 5.956165292371283e-06, "loss": 0.2038, "step": 20925 }, { "epoch": 6.44, "learning_rate": 5.955253887313888e-06, "loss": 0.1403, "step": 20926 }, { "epoch": 6.44, "learning_rate": 5.954342522424553e-06, "loss": 0.0814, "step": 20927 }, { "epoch": 6.44, "learning_rate": 5.953431197712325e-06, "loss": 0.1132, "step": 20928 }, { "epoch": 6.44, "learning_rate": 5.952519913186251e-06, "loss": 0.0696, "step": 20929 }, { "epoch": 6.44, "learning_rate": 5.951608668855387e-06, "loss": 0.1762, "step": 20930 }, { "epoch": 6.44, "learning_rate": 5.950697464728778e-06, "loss": 0.1063, "step": 20931 }, { "epoch": 6.44, "learning_rate": 5.9497863008154785e-06, "loss": 0.0669, "step": 20932 }, { "epoch": 6.44, "learning_rate": 5.948875177124533e-06, "loss": 0.0927, "step": 20933 }, { "epoch": 6.44, "learning_rate": 5.947964093664988e-06, "loss": 0.075, "step": 20934 }, { "epoch": 6.44, "learning_rate": 5.947053050445897e-06, "loss": 0.0764, "step": 20935 }, { "epoch": 6.44, "learning_rate": 5.946142047476302e-06, "loss": 0.0655, "step": 20936 }, { "epoch": 6.44, "learning_rate": 5.9452310847652536e-06, "loss": 0.0655, "step": 20937 }, { "epoch": 6.44, "learning_rate": 5.944320162321801e-06, "loss": 0.1056, "step": 20938 }, { "epoch": 6.44, "learning_rate": 5.943409280154981e-06, "loss": 0.0722, "step": 20939 }, { "epoch": 6.44, "learning_rate": 5.942498438273849e-06, "loss": 0.1051, "step": 20940 }, { "epoch": 6.44, "learning_rate": 5.941587636687445e-06, "loss": 0.0882, "step": 20941 }, { "epoch": 6.44, "learning_rate": 5.940676875404818e-06, "loss": 0.0651, "step": 20942 }, { "epoch": 6.44, "learning_rate": 5.9397661544350125e-06, "loss": 0.0551, "step": 20943 }, { "epoch": 6.44, "learning_rate": 5.938855473787067e-06, "loss": 0.0782, "step": 20944 }, { "epoch": 6.44, "learning_rate": 5.937944833470032e-06, "loss": 0.0461, "step": 20945 }, { "epoch": 6.44, "learning_rate": 5.937034233492946e-06, "loss": 0.0416, "step": 20946 }, { "epoch": 6.44, "learning_rate": 5.9361236738648575e-06, "loss": 0.0653, "step": 20947 }, { "epoch": 6.45, "learning_rate": 5.935213154594808e-06, "loss": 0.0799, "step": 20948 }, { "epoch": 6.45, "learning_rate": 5.934302675691833e-06, "loss": 0.0572, "step": 20949 }, { "epoch": 6.45, "learning_rate": 5.933392237164983e-06, "loss": 0.1022, "step": 20950 }, { "epoch": 6.45, "learning_rate": 5.932481839023292e-06, "loss": 0.069, "step": 20951 }, { "epoch": 6.45, "learning_rate": 5.931571481275809e-06, "loss": 0.0668, "step": 20952 }, { "epoch": 6.45, "learning_rate": 5.930661163931572e-06, "loss": 0.1047, "step": 20953 }, { "epoch": 6.45, "learning_rate": 5.929750886999614e-06, "loss": 0.0694, "step": 20954 }, { "epoch": 6.45, "learning_rate": 5.928840650488988e-06, "loss": 0.1264, "step": 20955 }, { "epoch": 6.45, "learning_rate": 5.9279304544087215e-06, "loss": 0.1415, "step": 20956 }, { "epoch": 6.45, "learning_rate": 5.927020298767861e-06, "loss": 0.0917, "step": 20957 }, { "epoch": 6.45, "learning_rate": 5.926110183575441e-06, "loss": 0.1035, "step": 20958 }, { "epoch": 6.45, "learning_rate": 5.925200108840501e-06, "loss": 0.1316, "step": 20959 }, { "epoch": 6.45, "learning_rate": 5.924290074572083e-06, "loss": 0.0715, "step": 20960 }, { "epoch": 6.45, "learning_rate": 5.9233800807792174e-06, "loss": 0.0991, "step": 20961 }, { "epoch": 6.45, "learning_rate": 5.9224701274709454e-06, "loss": 0.0912, "step": 20962 }, { "epoch": 6.45, "learning_rate": 5.921560214656303e-06, "loss": 0.111, "step": 20963 }, { "epoch": 6.45, "learning_rate": 5.920650342344326e-06, "loss": 0.0936, "step": 20964 }, { "epoch": 6.45, "learning_rate": 5.919740510544054e-06, "loss": 0.0844, "step": 20965 }, { "epoch": 6.45, "learning_rate": 5.918830719264514e-06, "loss": 0.0715, "step": 20966 }, { "epoch": 6.45, "learning_rate": 5.9179209685147525e-06, "loss": 0.1052, "step": 20967 }, { "epoch": 6.45, "learning_rate": 5.917011258303793e-06, "loss": 0.1027, "step": 20968 }, { "epoch": 6.45, "learning_rate": 5.916101588640675e-06, "loss": 0.0726, "step": 20969 }, { "epoch": 6.45, "learning_rate": 5.915191959534436e-06, "loss": 0.0975, "step": 20970 }, { "epoch": 6.45, "learning_rate": 5.9142823709941e-06, "loss": 0.097, "step": 20971 }, { "epoch": 6.45, "learning_rate": 5.91337282302871e-06, "loss": 0.0779, "step": 20972 }, { "epoch": 6.45, "learning_rate": 5.912463315647293e-06, "loss": 0.1161, "step": 20973 }, { "epoch": 6.45, "learning_rate": 5.911553848858881e-06, "loss": 0.0484, "step": 20974 }, { "epoch": 6.45, "learning_rate": 5.910644422672509e-06, "loss": 0.128, "step": 20975 }, { "epoch": 6.45, "learning_rate": 5.909735037097204e-06, "loss": 0.057, "step": 20976 }, { "epoch": 6.45, "learning_rate": 5.908825692142004e-06, "loss": 0.1256, "step": 20977 }, { "epoch": 6.45, "learning_rate": 5.907916387815934e-06, "loss": 0.0697, "step": 20978 }, { "epoch": 6.45, "learning_rate": 5.9070071241280235e-06, "loss": 0.0847, "step": 20979 }, { "epoch": 6.46, "learning_rate": 5.906097901087306e-06, "loss": 0.0591, "step": 20980 }, { "epoch": 6.46, "learning_rate": 5.9051887187028076e-06, "loss": 0.0624, "step": 20981 }, { "epoch": 6.46, "learning_rate": 5.904279576983564e-06, "loss": 0.0888, "step": 20982 }, { "epoch": 6.46, "learning_rate": 5.903370475938595e-06, "loss": 0.0755, "step": 20983 }, { "epoch": 6.46, "learning_rate": 5.902461415576932e-06, "loss": 0.2042, "step": 20984 }, { "epoch": 6.46, "learning_rate": 5.901552395907606e-06, "loss": 0.0394, "step": 20985 }, { "epoch": 6.46, "learning_rate": 5.90064341693964e-06, "loss": 0.075, "step": 20986 }, { "epoch": 6.46, "learning_rate": 5.899734478682064e-06, "loss": 0.1456, "step": 20987 }, { "epoch": 6.46, "learning_rate": 5.898825581143902e-06, "loss": 0.1171, "step": 20988 }, { "epoch": 6.46, "learning_rate": 5.897916724334185e-06, "loss": 0.0618, "step": 20989 }, { "epoch": 6.46, "learning_rate": 5.897007908261935e-06, "loss": 0.0922, "step": 20990 }, { "epoch": 6.46, "learning_rate": 5.896099132936175e-06, "loss": 0.1179, "step": 20991 }, { "epoch": 6.46, "learning_rate": 5.895190398365935e-06, "loss": 0.0762, "step": 20992 }, { "epoch": 6.46, "learning_rate": 5.894281704560235e-06, "loss": 0.0497, "step": 20993 }, { "epoch": 6.46, "learning_rate": 5.893373051528106e-06, "loss": 0.117, "step": 20994 }, { "epoch": 6.46, "learning_rate": 5.892464439278566e-06, "loss": 0.0615, "step": 20995 }, { "epoch": 6.46, "learning_rate": 5.891555867820637e-06, "loss": 0.07, "step": 20996 }, { "epoch": 6.46, "learning_rate": 5.890647337163347e-06, "loss": 0.0866, "step": 20997 }, { "epoch": 6.46, "learning_rate": 5.8897388473157134e-06, "loss": 0.1645, "step": 20998 }, { "epoch": 6.46, "learning_rate": 5.888830398286765e-06, "loss": 0.0685, "step": 20999 }, { "epoch": 6.46, "learning_rate": 5.8879219900855185e-06, "loss": 0.1132, "step": 21000 }, { "epoch": 6.46, "learning_rate": 5.8870136227209955e-06, "loss": 0.1355, "step": 21001 }, { "epoch": 6.46, "learning_rate": 5.886105296202218e-06, "loss": 0.1246, "step": 21002 }, { "epoch": 6.46, "learning_rate": 5.885197010538206e-06, "loss": 0.1, "step": 21003 }, { "epoch": 6.46, "learning_rate": 5.884288765737981e-06, "loss": 0.0662, "step": 21004 }, { "epoch": 6.46, "learning_rate": 5.8833805618105635e-06, "loss": 0.1269, "step": 21005 }, { "epoch": 6.46, "learning_rate": 5.882472398764968e-06, "loss": 0.1023, "step": 21006 }, { "epoch": 6.46, "learning_rate": 5.881564276610218e-06, "loss": 0.0928, "step": 21007 }, { "epoch": 6.46, "learning_rate": 5.880656195355328e-06, "loss": 0.0933, "step": 21008 }, { "epoch": 6.46, "learning_rate": 5.879748155009321e-06, "loss": 0.0795, "step": 21009 }, { "epoch": 6.46, "learning_rate": 5.878840155581213e-06, "loss": 0.1293, "step": 21010 }, { "epoch": 6.46, "learning_rate": 5.877932197080017e-06, "loss": 0.0956, "step": 21011 }, { "epoch": 6.46, "learning_rate": 5.877024279514756e-06, "loss": 0.0793, "step": 21012 }, { "epoch": 6.47, "learning_rate": 5.876116402894442e-06, "loss": 0.1627, "step": 21013 }, { "epoch": 6.47, "learning_rate": 5.875208567228094e-06, "loss": 0.1005, "step": 21014 }, { "epoch": 6.47, "learning_rate": 5.874300772524728e-06, "loss": 0.0773, "step": 21015 }, { "epoch": 6.47, "learning_rate": 5.873393018793352e-06, "loss": 0.0793, "step": 21016 }, { "epoch": 6.47, "learning_rate": 5.872485306042993e-06, "loss": 0.0674, "step": 21017 }, { "epoch": 6.47, "learning_rate": 5.871577634282655e-06, "loss": 0.0839, "step": 21018 }, { "epoch": 6.47, "learning_rate": 5.870670003521357e-06, "loss": 0.1261, "step": 21019 }, { "epoch": 6.47, "learning_rate": 5.869762413768114e-06, "loss": 0.0863, "step": 21020 }, { "epoch": 6.47, "learning_rate": 5.868854865031932e-06, "loss": 0.0793, "step": 21021 }, { "epoch": 6.47, "learning_rate": 5.8679473573218335e-06, "loss": 0.1335, "step": 21022 }, { "epoch": 6.47, "learning_rate": 5.867039890646822e-06, "loss": 0.0982, "step": 21023 }, { "epoch": 6.47, "learning_rate": 5.866132465015916e-06, "loss": 0.0423, "step": 21024 }, { "epoch": 6.47, "learning_rate": 5.865225080438124e-06, "loss": 0.1167, "step": 21025 }, { "epoch": 6.47, "learning_rate": 5.864317736922456e-06, "loss": 0.058, "step": 21026 }, { "epoch": 6.47, "learning_rate": 5.863410434477927e-06, "loss": 0.1009, "step": 21027 }, { "epoch": 6.47, "learning_rate": 5.862503173113542e-06, "loss": 0.1295, "step": 21028 }, { "epoch": 6.47, "learning_rate": 5.861595952838315e-06, "loss": 0.0677, "step": 21029 }, { "epoch": 6.47, "learning_rate": 5.860688773661256e-06, "loss": 0.1041, "step": 21030 }, { "epoch": 6.47, "learning_rate": 5.8597816355913685e-06, "loss": 0.0542, "step": 21031 }, { "epoch": 6.47, "learning_rate": 5.85887453863767e-06, "loss": 0.0218, "step": 21032 }, { "epoch": 6.47, "learning_rate": 5.857967482809157e-06, "loss": 0.0898, "step": 21033 }, { "epoch": 6.47, "learning_rate": 5.857060468114852e-06, "loss": 0.06, "step": 21034 }, { "epoch": 6.47, "learning_rate": 5.856153494563752e-06, "loss": 0.0929, "step": 21035 }, { "epoch": 6.47, "learning_rate": 5.855246562164864e-06, "loss": 0.0641, "step": 21036 }, { "epoch": 6.47, "learning_rate": 5.8543396709272005e-06, "loss": 0.1481, "step": 21037 }, { "epoch": 6.47, "learning_rate": 5.853432820859763e-06, "loss": 0.1078, "step": 21038 }, { "epoch": 6.47, "learning_rate": 5.852526011971562e-06, "loss": 0.1306, "step": 21039 }, { "epoch": 6.47, "learning_rate": 5.851619244271599e-06, "loss": 0.0738, "step": 21040 }, { "epoch": 6.47, "learning_rate": 5.850712517768877e-06, "loss": 0.064, "step": 21041 }, { "epoch": 6.47, "learning_rate": 5.849805832472408e-06, "loss": 0.0885, "step": 21042 }, { "epoch": 6.47, "learning_rate": 5.848899188391189e-06, "loss": 0.1265, "step": 21043 }, { "epoch": 6.47, "learning_rate": 5.84799258553423e-06, "loss": 0.0958, "step": 21044 }, { "epoch": 6.48, "learning_rate": 5.847086023910531e-06, "loss": 0.0764, "step": 21045 }, { "epoch": 6.48, "learning_rate": 5.846179503529091e-06, "loss": 0.1255, "step": 21046 }, { "epoch": 6.48, "learning_rate": 5.8452730243989205e-06, "loss": 0.1146, "step": 21047 }, { "epoch": 6.48, "learning_rate": 5.844366586529015e-06, "loss": 0.0868, "step": 21048 }, { "epoch": 6.48, "learning_rate": 5.843460189928384e-06, "loss": 0.0776, "step": 21049 }, { "epoch": 6.48, "learning_rate": 5.842553834606021e-06, "loss": 0.1236, "step": 21050 }, { "epoch": 6.48, "learning_rate": 5.84164752057093e-06, "loss": 0.1058, "step": 21051 }, { "epoch": 6.48, "learning_rate": 5.840741247832112e-06, "loss": 0.0815, "step": 21052 }, { "epoch": 6.48, "learning_rate": 5.839835016398565e-06, "loss": 0.0668, "step": 21053 }, { "epoch": 6.48, "learning_rate": 5.838928826279292e-06, "loss": 0.1136, "step": 21054 }, { "epoch": 6.48, "learning_rate": 5.838022677483291e-06, "loss": 0.1262, "step": 21055 }, { "epoch": 6.48, "learning_rate": 5.837116570019561e-06, "loss": 0.0449, "step": 21056 }, { "epoch": 6.48, "learning_rate": 5.836210503897099e-06, "loss": 0.0822, "step": 21057 }, { "epoch": 6.48, "learning_rate": 5.8353044791249015e-06, "loss": 0.0413, "step": 21058 }, { "epoch": 6.48, "learning_rate": 5.834398495711977e-06, "loss": 0.0468, "step": 21059 }, { "epoch": 6.48, "learning_rate": 5.83349255366731e-06, "loss": 0.0967, "step": 21060 }, { "epoch": 6.48, "learning_rate": 5.832586652999898e-06, "loss": 0.1268, "step": 21061 }, { "epoch": 6.48, "learning_rate": 5.831680793718748e-06, "loss": 0.0612, "step": 21062 }, { "epoch": 6.48, "learning_rate": 5.830774975832842e-06, "loss": 0.0387, "step": 21063 }, { "epoch": 6.48, "learning_rate": 5.829869199351188e-06, "loss": 0.1134, "step": 21064 }, { "epoch": 6.48, "learning_rate": 5.828963464282776e-06, "loss": 0.0721, "step": 21065 }, { "epoch": 6.48, "learning_rate": 5.828057770636601e-06, "loss": 0.0868, "step": 21066 }, { "epoch": 6.48, "learning_rate": 5.827152118421657e-06, "loss": 0.1184, "step": 21067 }, { "epoch": 6.48, "learning_rate": 5.826246507646933e-06, "loss": 0.0928, "step": 21068 }, { "epoch": 6.48, "learning_rate": 5.825340938321438e-06, "loss": 0.1401, "step": 21069 }, { "epoch": 6.48, "learning_rate": 5.82443541045415e-06, "loss": 0.0874, "step": 21070 }, { "epoch": 6.48, "learning_rate": 5.823529924054062e-06, "loss": 0.109, "step": 21071 }, { "epoch": 6.48, "learning_rate": 5.822624479130176e-06, "loss": 0.067, "step": 21072 }, { "epoch": 6.48, "learning_rate": 5.821719075691479e-06, "loss": 0.0692, "step": 21073 }, { "epoch": 6.48, "learning_rate": 5.820813713746961e-06, "loss": 0.1127, "step": 21074 }, { "epoch": 6.48, "learning_rate": 5.819908393305616e-06, "loss": 0.0732, "step": 21075 }, { "epoch": 6.48, "learning_rate": 5.819003114376431e-06, "loss": 0.108, "step": 21076 }, { "epoch": 6.48, "learning_rate": 5.818097876968401e-06, "loss": 0.0837, "step": 21077 }, { "epoch": 6.49, "learning_rate": 5.817192681090509e-06, "loss": 0.0516, "step": 21078 }, { "epoch": 6.49, "learning_rate": 5.8162875267517515e-06, "loss": 0.1385, "step": 21079 }, { "epoch": 6.49, "learning_rate": 5.815382413961119e-06, "loss": 0.0662, "step": 21080 }, { "epoch": 6.49, "learning_rate": 5.814477342727587e-06, "loss": 0.065, "step": 21081 }, { "epoch": 6.49, "learning_rate": 5.8135723130601575e-06, "loss": 0.1579, "step": 21082 }, { "epoch": 6.49, "learning_rate": 5.812667324967813e-06, "loss": 0.0831, "step": 21083 }, { "epoch": 6.49, "learning_rate": 5.811762378459541e-06, "loss": 0.0958, "step": 21084 }, { "epoch": 6.49, "learning_rate": 5.810857473544329e-06, "loss": 0.0889, "step": 21085 }, { "epoch": 6.49, "learning_rate": 5.809952610231162e-06, "loss": 0.0931, "step": 21086 }, { "epoch": 6.49, "learning_rate": 5.80904778852903e-06, "loss": 0.0957, "step": 21087 }, { "epoch": 6.49, "learning_rate": 5.8081430084469095e-06, "loss": 0.0463, "step": 21088 }, { "epoch": 6.49, "learning_rate": 5.807238269993796e-06, "loss": 0.1279, "step": 21089 }, { "epoch": 6.49, "learning_rate": 5.806333573178675e-06, "loss": 0.1048, "step": 21090 }, { "epoch": 6.49, "learning_rate": 5.80542891801052e-06, "loss": 0.1168, "step": 21091 }, { "epoch": 6.49, "learning_rate": 5.804524304498325e-06, "loss": 0.0344, "step": 21092 }, { "epoch": 6.49, "learning_rate": 5.803619732651069e-06, "loss": 0.0685, "step": 21093 }, { "epoch": 6.49, "learning_rate": 5.802715202477738e-06, "loss": 0.0598, "step": 21094 }, { "epoch": 6.49, "learning_rate": 5.801810713987313e-06, "loss": 0.14, "step": 21095 }, { "epoch": 6.49, "learning_rate": 5.800906267188773e-06, "loss": 0.1174, "step": 21096 }, { "epoch": 6.49, "learning_rate": 5.800001862091112e-06, "loss": 0.0907, "step": 21097 }, { "epoch": 6.49, "learning_rate": 5.799097498703294e-06, "loss": 0.1032, "step": 21098 }, { "epoch": 6.49, "learning_rate": 5.798193177034315e-06, "loss": 0.1102, "step": 21099 }, { "epoch": 6.49, "learning_rate": 5.797288897093149e-06, "loss": 0.0968, "step": 21100 }, { "epoch": 6.49, "learning_rate": 5.796384658888778e-06, "loss": 0.1, "step": 21101 }, { "epoch": 6.49, "learning_rate": 5.795480462430182e-06, "loss": 0.1281, "step": 21102 }, { "epoch": 6.49, "learning_rate": 5.794576307726339e-06, "loss": 0.05, "step": 21103 }, { "epoch": 6.49, "learning_rate": 5.793672194786231e-06, "loss": 0.0455, "step": 21104 }, { "epoch": 6.49, "learning_rate": 5.7927681236188325e-06, "loss": 0.084, "step": 21105 }, { "epoch": 6.49, "learning_rate": 5.791864094233122e-06, "loss": 0.0748, "step": 21106 }, { "epoch": 6.49, "learning_rate": 5.790960106638086e-06, "loss": 0.0657, "step": 21107 }, { "epoch": 6.49, "learning_rate": 5.790056160842688e-06, "loss": 0.1529, "step": 21108 }, { "epoch": 6.49, "learning_rate": 5.789152256855917e-06, "loss": 0.1008, "step": 21109 }, { "epoch": 6.5, "learning_rate": 5.788248394686743e-06, "loss": 0.0581, "step": 21110 }, { "epoch": 6.5, "learning_rate": 5.787344574344145e-06, "loss": 0.1274, "step": 21111 }, { "epoch": 6.5, "learning_rate": 5.786440795837098e-06, "loss": 0.1173, "step": 21112 }, { "epoch": 6.5, "learning_rate": 5.785537059174573e-06, "loss": 0.1084, "step": 21113 }, { "epoch": 6.5, "learning_rate": 5.784633364365557e-06, "loss": 0.0853, "step": 21114 }, { "epoch": 6.5, "learning_rate": 5.783729711419014e-06, "loss": 0.0399, "step": 21115 }, { "epoch": 6.5, "learning_rate": 5.782826100343916e-06, "loss": 0.0887, "step": 21116 }, { "epoch": 6.5, "learning_rate": 5.781922531149244e-06, "loss": 0.0914, "step": 21117 }, { "epoch": 6.5, "learning_rate": 5.781019003843971e-06, "loss": 0.1062, "step": 21118 }, { "epoch": 6.5, "learning_rate": 5.780115518437067e-06, "loss": 0.0378, "step": 21119 }, { "epoch": 6.5, "learning_rate": 5.779212074937505e-06, "loss": 0.0717, "step": 21120 }, { "epoch": 6.5, "learning_rate": 5.778308673354256e-06, "loss": 0.1234, "step": 21121 }, { "epoch": 6.5, "learning_rate": 5.777405313696294e-06, "loss": 0.0766, "step": 21122 }, { "epoch": 6.5, "learning_rate": 5.7765019959725855e-06, "loss": 0.1239, "step": 21123 }, { "epoch": 6.5, "learning_rate": 5.77559872019211e-06, "loss": 0.104, "step": 21124 }, { "epoch": 6.5, "learning_rate": 5.77469548636383e-06, "loss": 0.111, "step": 21125 }, { "epoch": 6.5, "learning_rate": 5.773792294496715e-06, "loss": 0.0959, "step": 21126 }, { "epoch": 6.5, "learning_rate": 5.772889144599741e-06, "loss": 0.0943, "step": 21127 }, { "epoch": 6.5, "learning_rate": 5.771986036681874e-06, "loss": 0.1093, "step": 21128 }, { "epoch": 6.5, "learning_rate": 5.771082970752081e-06, "loss": 0.1106, "step": 21129 }, { "epoch": 6.5, "learning_rate": 5.770179946819332e-06, "loss": 0.0512, "step": 21130 }, { "epoch": 6.5, "learning_rate": 5.769276964892594e-06, "loss": 0.0715, "step": 21131 }, { "epoch": 6.5, "learning_rate": 5.7683740249808365e-06, "loss": 0.1042, "step": 21132 }, { "epoch": 6.5, "learning_rate": 5.76747112709302e-06, "loss": 0.0756, "step": 21133 }, { "epoch": 6.5, "learning_rate": 5.76656827123812e-06, "loss": 0.0857, "step": 21134 }, { "epoch": 6.5, "learning_rate": 5.765665457425102e-06, "loss": 0.0665, "step": 21135 }, { "epoch": 6.5, "learning_rate": 5.764762685662921e-06, "loss": 0.0493, "step": 21136 }, { "epoch": 6.5, "learning_rate": 5.763859955960554e-06, "loss": 0.1195, "step": 21137 }, { "epoch": 6.5, "learning_rate": 5.762957268326962e-06, "loss": 0.0858, "step": 21138 }, { "epoch": 6.5, "learning_rate": 5.762054622771109e-06, "loss": 0.0992, "step": 21139 }, { "epoch": 6.5, "learning_rate": 5.7611520193019585e-06, "loss": 0.0685, "step": 21140 }, { "epoch": 6.5, "learning_rate": 5.760249457928477e-06, "loss": 0.0965, "step": 21141 }, { "epoch": 6.5, "learning_rate": 5.759346938659624e-06, "loss": 0.0945, "step": 21142 }, { "epoch": 6.51, "learning_rate": 5.758444461504362e-06, "loss": 0.1479, "step": 21143 }, { "epoch": 6.51, "learning_rate": 5.7575420264716586e-06, "loss": 0.0773, "step": 21144 }, { "epoch": 6.51, "learning_rate": 5.756639633570478e-06, "loss": 0.1379, "step": 21145 }, { "epoch": 6.51, "learning_rate": 5.755737282809767e-06, "loss": 0.095, "step": 21146 }, { "epoch": 6.51, "learning_rate": 5.754834974198501e-06, "loss": 0.0787, "step": 21147 }, { "epoch": 6.51, "learning_rate": 5.753932707745635e-06, "loss": 0.1194, "step": 21148 }, { "epoch": 6.51, "learning_rate": 5.753030483460132e-06, "loss": 0.0972, "step": 21149 }, { "epoch": 6.51, "learning_rate": 5.752128301350951e-06, "loss": 0.0628, "step": 21150 }, { "epoch": 6.51, "learning_rate": 5.751226161427046e-06, "loss": 0.0742, "step": 21151 }, { "epoch": 6.51, "learning_rate": 5.750324063697389e-06, "loss": 0.1224, "step": 21152 }, { "epoch": 6.51, "learning_rate": 5.749422008170922e-06, "loss": 0.0978, "step": 21153 }, { "epoch": 6.51, "learning_rate": 5.7485199948566175e-06, "loss": 0.1236, "step": 21154 }, { "epoch": 6.51, "learning_rate": 5.747618023763426e-06, "loss": 0.1318, "step": 21155 }, { "epoch": 6.51, "learning_rate": 5.7467160949003085e-06, "loss": 0.1038, "step": 21156 }, { "epoch": 6.51, "learning_rate": 5.745814208276219e-06, "loss": 0.0898, "step": 21157 }, { "epoch": 6.51, "learning_rate": 5.7449123639001105e-06, "loss": 0.0717, "step": 21158 }, { "epoch": 6.51, "learning_rate": 5.744010561780953e-06, "loss": 0.0559, "step": 21159 }, { "epoch": 6.51, "learning_rate": 5.7431088019276884e-06, "loss": 0.0898, "step": 21160 }, { "epoch": 6.51, "learning_rate": 5.742207084349274e-06, "loss": 0.0632, "step": 21161 }, { "epoch": 6.51, "learning_rate": 5.7413054090546735e-06, "loss": 0.0857, "step": 21162 }, { "epoch": 6.51, "learning_rate": 5.740403776052827e-06, "loss": 0.0643, "step": 21163 }, { "epoch": 6.51, "learning_rate": 5.739502185352703e-06, "loss": 0.0503, "step": 21164 }, { "epoch": 6.51, "learning_rate": 5.738600636963247e-06, "loss": 0.0938, "step": 21165 }, { "epoch": 6.51, "learning_rate": 5.737699130893415e-06, "loss": 0.0588, "step": 21166 }, { "epoch": 6.51, "learning_rate": 5.7367976671521584e-06, "loss": 0.1001, "step": 21167 }, { "epoch": 6.51, "learning_rate": 5.735896245748427e-06, "loss": 0.0958, "step": 21168 }, { "epoch": 6.51, "learning_rate": 5.734994866691182e-06, "loss": 0.0536, "step": 21169 }, { "epoch": 6.51, "learning_rate": 5.734093529989366e-06, "loss": 0.0928, "step": 21170 }, { "epoch": 6.51, "learning_rate": 5.7331922356519286e-06, "loss": 0.0937, "step": 21171 }, { "epoch": 6.51, "learning_rate": 5.732290983687829e-06, "loss": 0.0943, "step": 21172 }, { "epoch": 6.51, "learning_rate": 5.731389774106013e-06, "loss": 0.1146, "step": 21173 }, { "epoch": 6.51, "learning_rate": 5.73048860691543e-06, "loss": 0.1679, "step": 21174 }, { "epoch": 6.52, "learning_rate": 5.729587482125031e-06, "loss": 0.0903, "step": 21175 }, { "epoch": 6.52, "learning_rate": 5.728686399743764e-06, "loss": 0.1549, "step": 21176 }, { "epoch": 6.52, "learning_rate": 5.727785359780577e-06, "loss": 0.0964, "step": 21177 }, { "epoch": 6.52, "learning_rate": 5.726884362244414e-06, "loss": 0.1157, "step": 21178 }, { "epoch": 6.52, "learning_rate": 5.725983407144233e-06, "loss": 0.0931, "step": 21179 }, { "epoch": 6.52, "learning_rate": 5.725082494488979e-06, "loss": 0.0958, "step": 21180 }, { "epoch": 6.52, "learning_rate": 5.7241816242875885e-06, "loss": 0.0714, "step": 21181 }, { "epoch": 6.52, "learning_rate": 5.723280796549018e-06, "loss": 0.0915, "step": 21182 }, { "epoch": 6.52, "learning_rate": 5.722380011282211e-06, "loss": 0.1148, "step": 21183 }, { "epoch": 6.52, "learning_rate": 5.721479268496114e-06, "loss": 0.0439, "step": 21184 }, { "epoch": 6.52, "learning_rate": 5.720578568199666e-06, "loss": 0.1533, "step": 21185 }, { "epoch": 6.52, "learning_rate": 5.719677910401825e-06, "loss": 0.0326, "step": 21186 }, { "epoch": 6.52, "learning_rate": 5.718777295111524e-06, "loss": 0.0902, "step": 21187 }, { "epoch": 6.52, "learning_rate": 5.717876722337705e-06, "loss": 0.1136, "step": 21188 }, { "epoch": 6.52, "learning_rate": 5.716976192089321e-06, "loss": 0.0626, "step": 21189 }, { "epoch": 6.52, "learning_rate": 5.716075704375311e-06, "loss": 0.1071, "step": 21190 }, { "epoch": 6.52, "learning_rate": 5.715175259204617e-06, "loss": 0.085, "step": 21191 }, { "epoch": 6.52, "learning_rate": 5.714274856586183e-06, "loss": 0.1486, "step": 21192 }, { "epoch": 6.52, "learning_rate": 5.713374496528948e-06, "loss": 0.125, "step": 21193 }, { "epoch": 6.52, "learning_rate": 5.712474179041856e-06, "loss": 0.0825, "step": 21194 }, { "epoch": 6.52, "learning_rate": 5.711573904133842e-06, "loss": 0.0618, "step": 21195 }, { "epoch": 6.52, "learning_rate": 5.710673671813856e-06, "loss": 0.1415, "step": 21196 }, { "epoch": 6.52, "learning_rate": 5.709773482090837e-06, "loss": 0.1216, "step": 21197 }, { "epoch": 6.52, "learning_rate": 5.708873334973715e-06, "loss": 0.1383, "step": 21198 }, { "epoch": 6.52, "learning_rate": 5.707973230471437e-06, "loss": 0.0876, "step": 21199 }, { "epoch": 6.52, "learning_rate": 5.707073168592943e-06, "loss": 0.0938, "step": 21200 }, { "epoch": 6.52, "learning_rate": 5.706173149347167e-06, "loss": 0.0919, "step": 21201 }, { "epoch": 6.52, "learning_rate": 5.70527317274305e-06, "loss": 0.0916, "step": 21202 }, { "epoch": 6.52, "learning_rate": 5.7043732387895275e-06, "loss": 0.1566, "step": 21203 }, { "epoch": 6.52, "learning_rate": 5.7034733474955385e-06, "loss": 0.1119, "step": 21204 }, { "epoch": 6.52, "learning_rate": 5.702573498870015e-06, "loss": 0.0708, "step": 21205 }, { "epoch": 6.52, "learning_rate": 5.701673692921901e-06, "loss": 0.0861, "step": 21206 }, { "epoch": 6.52, "learning_rate": 5.700773929660133e-06, "loss": 0.058, "step": 21207 }, { "epoch": 6.53, "learning_rate": 5.699874209093635e-06, "loss": 0.0724, "step": 21208 }, { "epoch": 6.53, "learning_rate": 5.698974531231352e-06, "loss": 0.102, "step": 21209 }, { "epoch": 6.53, "learning_rate": 5.698074896082215e-06, "loss": 0.1441, "step": 21210 }, { "epoch": 6.53, "learning_rate": 5.697175303655161e-06, "loss": 0.0635, "step": 21211 }, { "epoch": 6.53, "learning_rate": 5.69627575395912e-06, "loss": 0.0916, "step": 21212 }, { "epoch": 6.53, "learning_rate": 5.695376247003025e-06, "loss": 0.0914, "step": 21213 }, { "epoch": 6.53, "learning_rate": 5.694476782795818e-06, "loss": 0.0804, "step": 21214 }, { "epoch": 6.53, "learning_rate": 5.6935773613464165e-06, "loss": 0.0874, "step": 21215 }, { "epoch": 6.53, "learning_rate": 5.692677982663766e-06, "loss": 0.0722, "step": 21216 }, { "epoch": 6.53, "learning_rate": 5.691778646756792e-06, "loss": 0.1041, "step": 21217 }, { "epoch": 6.53, "learning_rate": 5.690879353634427e-06, "loss": 0.1046, "step": 21218 }, { "epoch": 6.53, "learning_rate": 5.6899801033056e-06, "loss": 0.0891, "step": 21219 }, { "epoch": 6.53, "learning_rate": 5.689080895779244e-06, "loss": 0.0703, "step": 21220 }, { "epoch": 6.53, "learning_rate": 5.688181731064288e-06, "loss": 0.0821, "step": 21221 }, { "epoch": 6.53, "learning_rate": 5.6872826091696595e-06, "loss": 0.0694, "step": 21222 }, { "epoch": 6.53, "learning_rate": 5.686383530104288e-06, "loss": 0.0823, "step": 21223 }, { "epoch": 6.53, "learning_rate": 5.685484493877111e-06, "loss": 0.0713, "step": 21224 }, { "epoch": 6.53, "learning_rate": 5.68458550049704e-06, "loss": 0.0822, "step": 21225 }, { "epoch": 6.53, "learning_rate": 5.683686549973018e-06, "loss": 0.1223, "step": 21226 }, { "epoch": 6.53, "learning_rate": 5.682787642313965e-06, "loss": 0.0625, "step": 21227 }, { "epoch": 6.53, "learning_rate": 5.68188877752881e-06, "loss": 0.0671, "step": 21228 }, { "epoch": 6.53, "learning_rate": 5.6809899556264795e-06, "loss": 0.1002, "step": 21229 }, { "epoch": 6.53, "learning_rate": 5.680091176615894e-06, "loss": 0.1269, "step": 21230 }, { "epoch": 6.53, "learning_rate": 5.679192440505994e-06, "loss": 0.0449, "step": 21231 }, { "epoch": 6.53, "learning_rate": 5.67829374730569e-06, "loss": 0.0634, "step": 21232 }, { "epoch": 6.53, "learning_rate": 5.677395097023909e-06, "loss": 0.0734, "step": 21233 }, { "epoch": 6.53, "learning_rate": 5.676496489669583e-06, "loss": 0.0691, "step": 21234 }, { "epoch": 6.53, "learning_rate": 5.67559792525163e-06, "loss": 0.1069, "step": 21235 }, { "epoch": 6.53, "learning_rate": 5.674699403778976e-06, "loss": 0.0858, "step": 21236 }, { "epoch": 6.53, "learning_rate": 5.673800925260543e-06, "loss": 0.111, "step": 21237 }, { "epoch": 6.53, "learning_rate": 5.672902489705254e-06, "loss": 0.1207, "step": 21238 }, { "epoch": 6.53, "learning_rate": 5.672004097122033e-06, "loss": 0.071, "step": 21239 }, { "epoch": 6.54, "learning_rate": 5.671105747519794e-06, "loss": 0.0884, "step": 21240 }, { "epoch": 6.54, "learning_rate": 5.6702074409074715e-06, "loss": 0.0733, "step": 21241 }, { "epoch": 6.54, "learning_rate": 5.669309177293978e-06, "loss": 0.2079, "step": 21242 }, { "epoch": 6.54, "learning_rate": 5.6684109566882305e-06, "loss": 0.0712, "step": 21243 }, { "epoch": 6.54, "learning_rate": 5.667512779099158e-06, "loss": 0.1025, "step": 21244 }, { "epoch": 6.54, "learning_rate": 5.6666146445356775e-06, "loss": 0.1078, "step": 21245 }, { "epoch": 6.54, "learning_rate": 5.665716553006707e-06, "loss": 0.0849, "step": 21246 }, { "epoch": 6.54, "learning_rate": 5.6648185045211656e-06, "loss": 0.1138, "step": 21247 }, { "epoch": 6.54, "learning_rate": 5.663920499087971e-06, "loss": 0.1013, "step": 21248 }, { "epoch": 6.54, "learning_rate": 5.663022536716044e-06, "loss": 0.1651, "step": 21249 }, { "epoch": 6.54, "learning_rate": 5.662124617414295e-06, "loss": 0.0585, "step": 21250 }, { "epoch": 6.54, "learning_rate": 5.661226741191651e-06, "loss": 0.0904, "step": 21251 }, { "epoch": 6.54, "learning_rate": 5.6603289080570274e-06, "loss": 0.0972, "step": 21252 }, { "epoch": 6.54, "learning_rate": 5.659431118019329e-06, "loss": 0.0671, "step": 21253 }, { "epoch": 6.54, "learning_rate": 5.658533371087487e-06, "loss": 0.0659, "step": 21254 }, { "epoch": 6.54, "learning_rate": 5.6576356672704065e-06, "loss": 0.0869, "step": 21255 }, { "epoch": 6.54, "learning_rate": 5.656738006577007e-06, "loss": 0.0781, "step": 21256 }, { "epoch": 6.54, "learning_rate": 5.655840389016203e-06, "loss": 0.0863, "step": 21257 }, { "epoch": 6.54, "learning_rate": 5.654942814596902e-06, "loss": 0.1446, "step": 21258 }, { "epoch": 6.54, "learning_rate": 5.654045283328031e-06, "loss": 0.0532, "step": 21259 }, { "epoch": 6.54, "learning_rate": 5.65314779521849e-06, "loss": 0.1152, "step": 21260 }, { "epoch": 6.54, "learning_rate": 5.6522503502772e-06, "loss": 0.0909, "step": 21261 }, { "epoch": 6.54, "learning_rate": 5.651352948513075e-06, "loss": 0.0876, "step": 21262 }, { "epoch": 6.54, "learning_rate": 5.650455589935015e-06, "loss": 0.0977, "step": 21263 }, { "epoch": 6.54, "learning_rate": 5.6495582745519454e-06, "loss": 0.1009, "step": 21264 }, { "epoch": 6.54, "learning_rate": 5.648661002372769e-06, "loss": 0.1329, "step": 21265 }, { "epoch": 6.54, "learning_rate": 5.647763773406401e-06, "loss": 0.095, "step": 21266 }, { "epoch": 6.54, "learning_rate": 5.646866587661748e-06, "loss": 0.1358, "step": 21267 }, { "epoch": 6.54, "learning_rate": 5.64596944514772e-06, "loss": 0.0707, "step": 21268 }, { "epoch": 6.54, "learning_rate": 5.6450723458732345e-06, "loss": 0.0562, "step": 21269 }, { "epoch": 6.54, "learning_rate": 5.644175289847187e-06, "loss": 0.1402, "step": 21270 }, { "epoch": 6.54, "learning_rate": 5.643278277078497e-06, "loss": 0.078, "step": 21271 }, { "epoch": 6.54, "learning_rate": 5.64238130757607e-06, "loss": 0.1246, "step": 21272 }, { "epoch": 6.55, "learning_rate": 5.641484381348812e-06, "loss": 0.0878, "step": 21273 }, { "epoch": 6.55, "learning_rate": 5.64058749840563e-06, "loss": 0.0601, "step": 21274 }, { "epoch": 6.55, "learning_rate": 5.639690658755429e-06, "loss": 0.0756, "step": 21275 }, { "epoch": 6.55, "learning_rate": 5.638793862407126e-06, "loss": 0.0785, "step": 21276 }, { "epoch": 6.55, "learning_rate": 5.637897109369617e-06, "loss": 0.1125, "step": 21277 }, { "epoch": 6.55, "learning_rate": 5.637000399651804e-06, "loss": 0.0755, "step": 21278 }, { "epoch": 6.55, "learning_rate": 5.636103733262601e-06, "loss": 0.1032, "step": 21279 }, { "epoch": 6.55, "learning_rate": 5.635207110210912e-06, "loss": 0.0776, "step": 21280 }, { "epoch": 6.55, "learning_rate": 5.634310530505639e-06, "loss": 0.0934, "step": 21281 }, { "epoch": 6.55, "learning_rate": 5.633413994155685e-06, "loss": 0.053, "step": 21282 }, { "epoch": 6.55, "learning_rate": 5.632517501169955e-06, "loss": 0.0744, "step": 21283 }, { "epoch": 6.55, "learning_rate": 5.631621051557351e-06, "loss": 0.128, "step": 21284 }, { "epoch": 6.55, "learning_rate": 5.630724645326773e-06, "loss": 0.0649, "step": 21285 }, { "epoch": 6.55, "learning_rate": 5.6298282824871335e-06, "loss": 0.101, "step": 21286 }, { "epoch": 6.55, "learning_rate": 5.628931963047322e-06, "loss": 0.0246, "step": 21287 }, { "epoch": 6.55, "learning_rate": 5.6280356870162416e-06, "loss": 0.0606, "step": 21288 }, { "epoch": 6.55, "learning_rate": 5.627139454402799e-06, "loss": 0.1241, "step": 21289 }, { "epoch": 6.55, "learning_rate": 5.626243265215893e-06, "loss": 0.0745, "step": 21290 }, { "epoch": 6.55, "learning_rate": 5.625347119464422e-06, "loss": 0.096, "step": 21291 }, { "epoch": 6.55, "learning_rate": 5.624451017157286e-06, "loss": 0.0965, "step": 21292 }, { "epoch": 6.55, "learning_rate": 5.623554958303384e-06, "loss": 0.0828, "step": 21293 }, { "epoch": 6.55, "learning_rate": 5.622658942911614e-06, "loss": 0.129, "step": 21294 }, { "epoch": 6.55, "learning_rate": 5.621762970990871e-06, "loss": 0.0978, "step": 21295 }, { "epoch": 6.55, "learning_rate": 5.620867042550062e-06, "loss": 0.0885, "step": 21296 }, { "epoch": 6.55, "learning_rate": 5.619971157598083e-06, "loss": 0.1022, "step": 21297 }, { "epoch": 6.55, "learning_rate": 5.6190753161438185e-06, "loss": 0.0872, "step": 21298 }, { "epoch": 6.55, "learning_rate": 5.618179518196177e-06, "loss": 0.0846, "step": 21299 }, { "epoch": 6.55, "learning_rate": 5.617283763764052e-06, "loss": 0.0815, "step": 21300 }, { "epoch": 6.55, "learning_rate": 5.616388052856339e-06, "loss": 0.0983, "step": 21301 }, { "epoch": 6.55, "learning_rate": 5.6154923854819334e-06, "loss": 0.0489, "step": 21302 }, { "epoch": 6.55, "learning_rate": 5.614596761649728e-06, "loss": 0.0616, "step": 21303 }, { "epoch": 6.55, "learning_rate": 5.613701181368618e-06, "loss": 0.0707, "step": 21304 }, { "epoch": 6.56, "learning_rate": 5.6128056446474944e-06, "loss": 0.0689, "step": 21305 }, { "epoch": 6.56, "learning_rate": 5.611910151495259e-06, "loss": 0.0928, "step": 21306 }, { "epoch": 6.56, "learning_rate": 5.611014701920803e-06, "loss": 0.1328, "step": 21307 }, { "epoch": 6.56, "learning_rate": 5.610119295933008e-06, "loss": 0.1098, "step": 21308 }, { "epoch": 6.56, "learning_rate": 5.609223933540778e-06, "loss": 0.0585, "step": 21309 }, { "epoch": 6.56, "learning_rate": 5.608328614753001e-06, "loss": 0.1078, "step": 21310 }, { "epoch": 6.56, "learning_rate": 5.60743333957857e-06, "loss": 0.0755, "step": 21311 }, { "epoch": 6.56, "learning_rate": 5.606538108026372e-06, "loss": 0.069, "step": 21312 }, { "epoch": 6.56, "learning_rate": 5.605642920105296e-06, "loss": 0.1148, "step": 21313 }, { "epoch": 6.56, "learning_rate": 5.604747775824245e-06, "loss": 0.0995, "step": 21314 }, { "epoch": 6.56, "learning_rate": 5.603852675192092e-06, "loss": 0.1485, "step": 21315 }, { "epoch": 6.56, "learning_rate": 5.602957618217735e-06, "loss": 0.1183, "step": 21316 }, { "epoch": 6.56, "learning_rate": 5.602062604910064e-06, "loss": 0.1256, "step": 21317 }, { "epoch": 6.56, "learning_rate": 5.601167635277962e-06, "loss": 0.0437, "step": 21318 }, { "epoch": 6.56, "learning_rate": 5.600272709330321e-06, "loss": 0.1185, "step": 21319 }, { "epoch": 6.56, "learning_rate": 5.5993778270760266e-06, "loss": 0.0452, "step": 21320 }, { "epoch": 6.56, "learning_rate": 5.5984829885239665e-06, "loss": 0.0747, "step": 21321 }, { "epoch": 6.56, "learning_rate": 5.597588193683026e-06, "loss": 0.1264, "step": 21322 }, { "epoch": 6.56, "learning_rate": 5.596693442562089e-06, "loss": 0.1043, "step": 21323 }, { "epoch": 6.56, "learning_rate": 5.595798735170053e-06, "loss": 0.0884, "step": 21324 }, { "epoch": 6.56, "learning_rate": 5.594904071515784e-06, "loss": 0.0542, "step": 21325 }, { "epoch": 6.56, "learning_rate": 5.594009451608183e-06, "loss": 0.0871, "step": 21326 }, { "epoch": 6.56, "learning_rate": 5.5931148754561295e-06, "loss": 0.1374, "step": 21327 }, { "epoch": 6.56, "learning_rate": 5.5922203430685065e-06, "loss": 0.102, "step": 21328 }, { "epoch": 6.56, "learning_rate": 5.591325854454197e-06, "loss": 0.0664, "step": 21329 }, { "epoch": 6.56, "learning_rate": 5.590431409622081e-06, "loss": 0.1143, "step": 21330 }, { "epoch": 6.56, "learning_rate": 5.5895370085810544e-06, "loss": 0.1564, "step": 21331 }, { "epoch": 6.56, "learning_rate": 5.588642651339984e-06, "loss": 0.0706, "step": 21332 }, { "epoch": 6.56, "learning_rate": 5.587748337907754e-06, "loss": 0.1479, "step": 21333 }, { "epoch": 6.56, "learning_rate": 5.586854068293255e-06, "loss": 0.0832, "step": 21334 }, { "epoch": 6.56, "learning_rate": 5.585959842505362e-06, "loss": 0.1128, "step": 21335 }, { "epoch": 6.56, "learning_rate": 5.5850656605529554e-06, "loss": 0.0767, "step": 21336 }, { "epoch": 6.56, "learning_rate": 5.584171522444916e-06, "loss": 0.0754, "step": 21337 }, { "epoch": 6.57, "learning_rate": 5.583277428190124e-06, "loss": 0.1301, "step": 21338 }, { "epoch": 6.57, "learning_rate": 5.5823833777974555e-06, "loss": 0.0998, "step": 21339 }, { "epoch": 6.57, "learning_rate": 5.581489371275789e-06, "loss": 0.0823, "step": 21340 }, { "epoch": 6.57, "learning_rate": 5.580595408634013e-06, "loss": 0.1598, "step": 21341 }, { "epoch": 6.57, "learning_rate": 5.579701489880993e-06, "loss": 0.1242, "step": 21342 }, { "epoch": 6.57, "learning_rate": 5.5788076150256075e-06, "loss": 0.0542, "step": 21343 }, { "epoch": 6.57, "learning_rate": 5.577913784076741e-06, "loss": 0.1145, "step": 21344 }, { "epoch": 6.57, "learning_rate": 5.577019997043267e-06, "loss": 0.0908, "step": 21345 }, { "epoch": 6.57, "learning_rate": 5.57612625393406e-06, "loss": 0.1235, "step": 21346 }, { "epoch": 6.57, "learning_rate": 5.575232554757997e-06, "loss": 0.058, "step": 21347 }, { "epoch": 6.57, "learning_rate": 5.5743388995239525e-06, "loss": 0.0453, "step": 21348 }, { "epoch": 6.57, "learning_rate": 5.5734452882407994e-06, "loss": 0.0673, "step": 21349 }, { "epoch": 6.57, "learning_rate": 5.572551720917413e-06, "loss": 0.0527, "step": 21350 }, { "epoch": 6.57, "learning_rate": 5.5716581975626715e-06, "loss": 0.0873, "step": 21351 }, { "epoch": 6.57, "learning_rate": 5.57076471818545e-06, "loss": 0.0555, "step": 21352 }, { "epoch": 6.57, "learning_rate": 5.569871282794608e-06, "loss": 0.0921, "step": 21353 }, { "epoch": 6.57, "learning_rate": 5.568977891399032e-06, "loss": 0.0965, "step": 21354 }, { "epoch": 6.57, "learning_rate": 5.5680845440075885e-06, "loss": 0.0682, "step": 21355 }, { "epoch": 6.57, "learning_rate": 5.567191240629151e-06, "loss": 0.1575, "step": 21356 }, { "epoch": 6.57, "learning_rate": 5.5662979812725885e-06, "loss": 0.1053, "step": 21357 }, { "epoch": 6.57, "learning_rate": 5.565404765946769e-06, "loss": 0.1649, "step": 21358 }, { "epoch": 6.57, "learning_rate": 5.5645115946605775e-06, "loss": 0.1174, "step": 21359 }, { "epoch": 6.57, "learning_rate": 5.563618467422864e-06, "loss": 0.1201, "step": 21360 }, { "epoch": 6.57, "learning_rate": 5.562725384242512e-06, "loss": 0.1135, "step": 21361 }, { "epoch": 6.57, "learning_rate": 5.5618323451283905e-06, "loss": 0.0956, "step": 21362 }, { "epoch": 6.57, "learning_rate": 5.560939350089356e-06, "loss": 0.1594, "step": 21363 }, { "epoch": 6.57, "learning_rate": 5.56004639913429e-06, "loss": 0.1024, "step": 21364 }, { "epoch": 6.57, "learning_rate": 5.559153492272054e-06, "loss": 0.1079, "step": 21365 }, { "epoch": 6.57, "learning_rate": 5.558260629511518e-06, "loss": 0.0919, "step": 21366 }, { "epoch": 6.57, "learning_rate": 5.557367810861547e-06, "loss": 0.049, "step": 21367 }, { "epoch": 6.57, "learning_rate": 5.556475036331002e-06, "loss": 0.0394, "step": 21368 }, { "epoch": 6.57, "learning_rate": 5.555582305928766e-06, "loss": 0.1393, "step": 21369 }, { "epoch": 6.58, "learning_rate": 5.554689619663684e-06, "loss": 0.0839, "step": 21370 }, { "epoch": 6.58, "learning_rate": 5.553796977544637e-06, "loss": 0.0437, "step": 21371 }, { "epoch": 6.58, "learning_rate": 5.5529043795804815e-06, "loss": 0.0385, "step": 21372 }, { "epoch": 6.58, "learning_rate": 5.5520118257800855e-06, "loss": 0.0474, "step": 21373 }, { "epoch": 6.58, "learning_rate": 5.551119316152311e-06, "loss": 0.0907, "step": 21374 }, { "epoch": 6.58, "learning_rate": 5.550226850706018e-06, "loss": 0.0815, "step": 21375 }, { "epoch": 6.58, "learning_rate": 5.549334429450082e-06, "loss": 0.0827, "step": 21376 }, { "epoch": 6.58, "learning_rate": 5.548442052393352e-06, "loss": 0.0977, "step": 21377 }, { "epoch": 6.58, "learning_rate": 5.547549719544692e-06, "loss": 0.0678, "step": 21378 }, { "epoch": 6.58, "learning_rate": 5.5466574309129725e-06, "loss": 0.0554, "step": 21379 }, { "epoch": 6.58, "learning_rate": 5.545765186507044e-06, "loss": 0.0771, "step": 21380 }, { "epoch": 6.58, "learning_rate": 5.544872986335774e-06, "loss": 0.0495, "step": 21381 }, { "epoch": 6.58, "learning_rate": 5.5439808304080225e-06, "loss": 0.0743, "step": 21382 }, { "epoch": 6.58, "learning_rate": 5.543088718732647e-06, "loss": 0.1086, "step": 21383 }, { "epoch": 6.58, "learning_rate": 5.54219665131851e-06, "loss": 0.1076, "step": 21384 }, { "epoch": 6.58, "learning_rate": 5.541304628174464e-06, "loss": 0.0721, "step": 21385 }, { "epoch": 6.58, "learning_rate": 5.540412649309379e-06, "loss": 0.1068, "step": 21386 }, { "epoch": 6.58, "learning_rate": 5.5395207147321e-06, "loss": 0.0874, "step": 21387 }, { "epoch": 6.58, "learning_rate": 5.538628824451495e-06, "loss": 0.0775, "step": 21388 }, { "epoch": 6.58, "learning_rate": 5.537736978476417e-06, "loss": 0.0964, "step": 21389 }, { "epoch": 6.58, "learning_rate": 5.536845176815724e-06, "loss": 0.0932, "step": 21390 }, { "epoch": 6.58, "learning_rate": 5.53595341947827e-06, "loss": 0.0709, "step": 21391 }, { "epoch": 6.58, "learning_rate": 5.535061706472911e-06, "loss": 0.0951, "step": 21392 }, { "epoch": 6.58, "learning_rate": 5.534170037808511e-06, "loss": 0.0987, "step": 21393 }, { "epoch": 6.58, "learning_rate": 5.533278413493915e-06, "loss": 0.1065, "step": 21394 }, { "epoch": 6.58, "learning_rate": 5.5323868335379775e-06, "loss": 0.0852, "step": 21395 }, { "epoch": 6.58, "learning_rate": 5.531495297949558e-06, "loss": 0.0635, "step": 21396 }, { "epoch": 6.58, "learning_rate": 5.53060380673751e-06, "loss": 0.1161, "step": 21397 }, { "epoch": 6.58, "learning_rate": 5.529712359910685e-06, "loss": 0.0847, "step": 21398 }, { "epoch": 6.58, "learning_rate": 5.528820957477936e-06, "loss": 0.0454, "step": 21399 }, { "epoch": 6.58, "learning_rate": 5.527929599448116e-06, "loss": 0.0417, "step": 21400 }, { "epoch": 6.58, "learning_rate": 5.5270382858300775e-06, "loss": 0.0909, "step": 21401 }, { "epoch": 6.58, "learning_rate": 5.526147016632666e-06, "loss": 0.1058, "step": 21402 }, { "epoch": 6.59, "learning_rate": 5.525255791864747e-06, "loss": 0.151, "step": 21403 }, { "epoch": 6.59, "learning_rate": 5.524364611535157e-06, "loss": 0.0778, "step": 21404 }, { "epoch": 6.59, "learning_rate": 5.5234734756527475e-06, "loss": 0.0673, "step": 21405 }, { "epoch": 6.59, "learning_rate": 5.522582384226376e-06, "loss": 0.1242, "step": 21406 }, { "epoch": 6.59, "learning_rate": 5.521691337264887e-06, "loss": 0.1348, "step": 21407 }, { "epoch": 6.59, "learning_rate": 5.520800334777132e-06, "loss": 0.1424, "step": 21408 }, { "epoch": 6.59, "learning_rate": 5.519909376771957e-06, "loss": 0.1675, "step": 21409 }, { "epoch": 6.59, "learning_rate": 5.519018463258212e-06, "loss": 0.1212, "step": 21410 }, { "epoch": 6.59, "learning_rate": 5.518127594244742e-06, "loss": 0.0539, "step": 21411 }, { "epoch": 6.59, "learning_rate": 5.5172367697403925e-06, "loss": 0.1104, "step": 21412 }, { "epoch": 6.59, "learning_rate": 5.516345989754018e-06, "loss": 0.1134, "step": 21413 }, { "epoch": 6.59, "learning_rate": 5.515455254294461e-06, "loss": 0.0956, "step": 21414 }, { "epoch": 6.59, "learning_rate": 5.514564563370562e-06, "loss": 0.1247, "step": 21415 }, { "epoch": 6.59, "learning_rate": 5.513673916991173e-06, "loss": 0.0821, "step": 21416 }, { "epoch": 6.59, "learning_rate": 5.512783315165136e-06, "loss": 0.066, "step": 21417 }, { "epoch": 6.59, "learning_rate": 5.5118927579012985e-06, "loss": 0.0551, "step": 21418 }, { "epoch": 6.59, "learning_rate": 5.511002245208501e-06, "loss": 0.068, "step": 21419 }, { "epoch": 6.59, "learning_rate": 5.510111777095588e-06, "loss": 0.0591, "step": 21420 }, { "epoch": 6.59, "learning_rate": 5.509221353571404e-06, "loss": 0.0834, "step": 21421 }, { "epoch": 6.59, "learning_rate": 5.5083309746447845e-06, "loss": 0.1256, "step": 21422 }, { "epoch": 6.59, "learning_rate": 5.507440640324584e-06, "loss": 0.0733, "step": 21423 }, { "epoch": 6.59, "learning_rate": 5.506550350619642e-06, "loss": 0.0756, "step": 21424 }, { "epoch": 6.59, "learning_rate": 5.505660105538789e-06, "loss": 0.0939, "step": 21425 }, { "epoch": 6.59, "learning_rate": 5.504769905090876e-06, "loss": 0.074, "step": 21426 }, { "epoch": 6.59, "learning_rate": 5.503879749284742e-06, "loss": 0.1019, "step": 21427 }, { "epoch": 6.59, "learning_rate": 5.5029896381292254e-06, "loss": 0.0805, "step": 21428 }, { "epoch": 6.59, "learning_rate": 5.502099571633166e-06, "loss": 0.1029, "step": 21429 }, { "epoch": 6.59, "learning_rate": 5.501209549805399e-06, "loss": 0.0981, "step": 21430 }, { "epoch": 6.59, "learning_rate": 5.500319572654777e-06, "loss": 0.143, "step": 21431 }, { "epoch": 6.59, "learning_rate": 5.499429640190119e-06, "loss": 0.088, "step": 21432 }, { "epoch": 6.59, "learning_rate": 5.498539752420278e-06, "loss": 0.1047, "step": 21433 }, { "epoch": 6.59, "learning_rate": 5.497649909354084e-06, "loss": 0.0712, "step": 21434 }, { "epoch": 6.6, "learning_rate": 5.496760111000376e-06, "loss": 0.0603, "step": 21435 }, { "epoch": 6.6, "learning_rate": 5.495870357367992e-06, "loss": 0.0674, "step": 21436 }, { "epoch": 6.6, "learning_rate": 5.494980648465762e-06, "loss": 0.0723, "step": 21437 }, { "epoch": 6.6, "learning_rate": 5.494090984302534e-06, "loss": 0.0784, "step": 21438 }, { "epoch": 6.6, "learning_rate": 5.493201364887131e-06, "loss": 0.0998, "step": 21439 }, { "epoch": 6.6, "learning_rate": 5.492311790228389e-06, "loss": 0.1096, "step": 21440 }, { "epoch": 6.6, "learning_rate": 5.491422260335151e-06, "loss": 0.0609, "step": 21441 }, { "epoch": 6.6, "learning_rate": 5.490532775216239e-06, "loss": 0.0878, "step": 21442 }, { "epoch": 6.6, "learning_rate": 5.4896433348804965e-06, "loss": 0.0618, "step": 21443 }, { "epoch": 6.6, "learning_rate": 5.488753939336752e-06, "loss": 0.0959, "step": 21444 }, { "epoch": 6.6, "learning_rate": 5.487864588593839e-06, "loss": 0.0855, "step": 21445 }, { "epoch": 6.6, "learning_rate": 5.486975282660589e-06, "loss": 0.0587, "step": 21446 }, { "epoch": 6.6, "learning_rate": 5.486086021545829e-06, "loss": 0.0649, "step": 21447 }, { "epoch": 6.6, "learning_rate": 5.485196805258404e-06, "loss": 0.1748, "step": 21448 }, { "epoch": 6.6, "learning_rate": 5.484307633807131e-06, "loss": 0.0798, "step": 21449 }, { "epoch": 6.6, "learning_rate": 5.483418507200842e-06, "loss": 0.0968, "step": 21450 }, { "epoch": 6.6, "learning_rate": 5.482529425448373e-06, "loss": 0.1176, "step": 21451 }, { "epoch": 6.6, "learning_rate": 5.481640388558551e-06, "loss": 0.1105, "step": 21452 }, { "epoch": 6.6, "learning_rate": 5.4807513965402024e-06, "loss": 0.1226, "step": 21453 }, { "epoch": 6.6, "learning_rate": 5.4798624494021605e-06, "loss": 0.1312, "step": 21454 }, { "epoch": 6.6, "learning_rate": 5.478973547153248e-06, "loss": 0.0637, "step": 21455 }, { "epoch": 6.6, "learning_rate": 5.478084689802295e-06, "loss": 0.0981, "step": 21456 }, { "epoch": 6.6, "learning_rate": 5.4771958773581256e-06, "loss": 0.0553, "step": 21457 }, { "epoch": 6.6, "learning_rate": 5.476307109829577e-06, "loss": 0.1319, "step": 21458 }, { "epoch": 6.6, "learning_rate": 5.475418387225464e-06, "loss": 0.1402, "step": 21459 }, { "epoch": 6.6, "learning_rate": 5.4745297095546125e-06, "loss": 0.0824, "step": 21460 }, { "epoch": 6.6, "learning_rate": 5.473641076825855e-06, "loss": 0.1115, "step": 21461 }, { "epoch": 6.6, "learning_rate": 5.4727524890480135e-06, "loss": 0.0754, "step": 21462 }, { "epoch": 6.6, "learning_rate": 5.471863946229911e-06, "loss": 0.0444, "step": 21463 }, { "epoch": 6.6, "learning_rate": 5.470975448380376e-06, "loss": 0.0937, "step": 21464 }, { "epoch": 6.6, "learning_rate": 5.470086995508226e-06, "loss": 0.1028, "step": 21465 }, { "epoch": 6.6, "learning_rate": 5.469198587622289e-06, "loss": 0.0811, "step": 21466 }, { "epoch": 6.6, "learning_rate": 5.468310224731381e-06, "loss": 0.0875, "step": 21467 }, { "epoch": 6.61, "learning_rate": 5.467421906844333e-06, "loss": 0.065, "step": 21468 }, { "epoch": 6.61, "learning_rate": 5.466533633969966e-06, "loss": 0.1084, "step": 21469 }, { "epoch": 6.61, "learning_rate": 5.465645406117091e-06, "loss": 0.1323, "step": 21470 }, { "epoch": 6.61, "learning_rate": 5.46475722329454e-06, "loss": 0.1196, "step": 21471 }, { "epoch": 6.61, "learning_rate": 5.463869085511128e-06, "loss": 0.0781, "step": 21472 }, { "epoch": 6.61, "learning_rate": 5.4629809927756794e-06, "loss": 0.1315, "step": 21473 }, { "epoch": 6.61, "learning_rate": 5.462092945097011e-06, "loss": 0.0512, "step": 21474 }, { "epoch": 6.61, "learning_rate": 5.461204942483935e-06, "loss": 0.1042, "step": 21475 }, { "epoch": 6.61, "learning_rate": 5.460316984945286e-06, "loss": 0.0864, "step": 21476 }, { "epoch": 6.61, "learning_rate": 5.4594290724898665e-06, "loss": 0.0512, "step": 21477 }, { "epoch": 6.61, "learning_rate": 5.458541205126504e-06, "loss": 0.0685, "step": 21478 }, { "epoch": 6.61, "learning_rate": 5.4576533828640184e-06, "loss": 0.08, "step": 21479 }, { "epoch": 6.61, "learning_rate": 5.456765605711212e-06, "loss": 0.1033, "step": 21480 }, { "epoch": 6.61, "learning_rate": 5.455877873676913e-06, "loss": 0.1416, "step": 21481 }, { "epoch": 6.61, "learning_rate": 5.454990186769937e-06, "loss": 0.0573, "step": 21482 }, { "epoch": 6.61, "learning_rate": 5.454102544999096e-06, "loss": 0.0738, "step": 21483 }, { "epoch": 6.61, "learning_rate": 5.453214948373206e-06, "loss": 0.0526, "step": 21484 }, { "epoch": 6.61, "learning_rate": 5.452327396901078e-06, "loss": 0.1269, "step": 21485 }, { "epoch": 6.61, "learning_rate": 5.451439890591539e-06, "loss": 0.1074, "step": 21486 }, { "epoch": 6.61, "learning_rate": 5.450552429453386e-06, "loss": 0.0722, "step": 21487 }, { "epoch": 6.61, "learning_rate": 5.4496650134954445e-06, "loss": 0.0763, "step": 21488 }, { "epoch": 6.61, "learning_rate": 5.448777642726522e-06, "loss": 0.0684, "step": 21489 }, { "epoch": 6.61, "learning_rate": 5.447890317155433e-06, "loss": 0.0597, "step": 21490 }, { "epoch": 6.61, "learning_rate": 5.4470030367909874e-06, "loss": 0.08, "step": 21491 }, { "epoch": 6.61, "learning_rate": 5.446115801641995e-06, "loss": 0.2047, "step": 21492 }, { "epoch": 6.61, "learning_rate": 5.445228611717277e-06, "loss": 0.0642, "step": 21493 }, { "epoch": 6.61, "learning_rate": 5.444341467025633e-06, "loss": 0.08, "step": 21494 }, { "epoch": 6.61, "learning_rate": 5.443454367575873e-06, "loss": 0.1055, "step": 21495 }, { "epoch": 6.61, "learning_rate": 5.442567313376814e-06, "loss": 0.1266, "step": 21496 }, { "epoch": 6.61, "learning_rate": 5.441680304437263e-06, "loss": 0.1701, "step": 21497 }, { "epoch": 6.61, "learning_rate": 5.440793340766027e-06, "loss": 0.0478, "step": 21498 }, { "epoch": 6.61, "learning_rate": 5.439906422371914e-06, "loss": 0.1162, "step": 21499 }, { "epoch": 6.62, "learning_rate": 5.4390195492637335e-06, "loss": 0.1092, "step": 21500 }, { "epoch": 6.62, "learning_rate": 5.438132721450293e-06, "loss": 0.0352, "step": 21501 }, { "epoch": 6.62, "learning_rate": 5.437245938940394e-06, "loss": 0.0518, "step": 21502 }, { "epoch": 6.62, "learning_rate": 5.436359201742857e-06, "loss": 0.0673, "step": 21503 }, { "epoch": 6.62, "learning_rate": 5.435472509866475e-06, "loss": 0.0916, "step": 21504 }, { "epoch": 6.62, "learning_rate": 5.434585863320052e-06, "loss": 0.0947, "step": 21505 }, { "epoch": 6.62, "learning_rate": 5.433699262112405e-06, "loss": 0.1009, "step": 21506 }, { "epoch": 6.62, "learning_rate": 5.432812706252333e-06, "loss": 0.1287, "step": 21507 }, { "epoch": 6.62, "learning_rate": 5.431926195748638e-06, "loss": 0.069, "step": 21508 }, { "epoch": 6.62, "learning_rate": 5.431039730610128e-06, "loss": 0.0462, "step": 21509 }, { "epoch": 6.62, "learning_rate": 5.430153310845603e-06, "loss": 0.1078, "step": 21510 }, { "epoch": 6.62, "learning_rate": 5.4292669364638685e-06, "loss": 0.0866, "step": 21511 }, { "epoch": 6.62, "learning_rate": 5.42838060747372e-06, "loss": 0.0704, "step": 21512 }, { "epoch": 6.62, "learning_rate": 5.42749432388397e-06, "loss": 0.0882, "step": 21513 }, { "epoch": 6.62, "learning_rate": 5.426608085703422e-06, "loss": 0.0734, "step": 21514 }, { "epoch": 6.62, "learning_rate": 5.42572189294086e-06, "loss": 0.0763, "step": 21515 }, { "epoch": 6.62, "learning_rate": 5.4248357456051e-06, "loss": 0.0471, "step": 21516 }, { "epoch": 6.62, "learning_rate": 5.423949643704938e-06, "loss": 0.0936, "step": 21517 }, { "epoch": 6.62, "learning_rate": 5.423063587249173e-06, "loss": 0.0873, "step": 21518 }, { "epoch": 6.62, "learning_rate": 5.4221775762466055e-06, "loss": 0.143, "step": 21519 }, { "epoch": 6.62, "learning_rate": 5.421291610706032e-06, "loss": 0.1109, "step": 21520 }, { "epoch": 6.62, "learning_rate": 5.420405690636255e-06, "loss": 0.0586, "step": 21521 }, { "epoch": 6.62, "learning_rate": 5.419519816046065e-06, "loss": 0.1324, "step": 21522 }, { "epoch": 6.62, "learning_rate": 5.4186339869442684e-06, "loss": 0.0977, "step": 21523 }, { "epoch": 6.62, "learning_rate": 5.417748203339663e-06, "loss": 0.195, "step": 21524 }, { "epoch": 6.62, "learning_rate": 5.416862465241033e-06, "loss": 0.123, "step": 21525 }, { "epoch": 6.62, "learning_rate": 5.415976772657188e-06, "loss": 0.0918, "step": 21526 }, { "epoch": 6.62, "learning_rate": 5.415091125596916e-06, "loss": 0.0552, "step": 21527 }, { "epoch": 6.62, "learning_rate": 5.414205524069016e-06, "loss": 0.1103, "step": 21528 }, { "epoch": 6.62, "learning_rate": 5.4133199680822815e-06, "loss": 0.1195, "step": 21529 }, { "epoch": 6.62, "learning_rate": 5.412434457645503e-06, "loss": 0.0803, "step": 21530 }, { "epoch": 6.62, "learning_rate": 5.411548992767486e-06, "loss": 0.1204, "step": 21531 }, { "epoch": 6.62, "learning_rate": 5.410663573457009e-06, "loss": 0.1408, "step": 21532 }, { "epoch": 6.63, "learning_rate": 5.4097781997228745e-06, "loss": 0.0899, "step": 21533 }, { "epoch": 6.63, "learning_rate": 5.408892871573874e-06, "loss": 0.0955, "step": 21534 }, { "epoch": 6.63, "learning_rate": 5.4080075890188e-06, "loss": 0.094, "step": 21535 }, { "epoch": 6.63, "learning_rate": 5.407122352066442e-06, "loss": 0.0888, "step": 21536 }, { "epoch": 6.63, "learning_rate": 5.406237160725591e-06, "loss": 0.1434, "step": 21537 }, { "epoch": 6.63, "learning_rate": 5.405352015005039e-06, "loss": 0.1132, "step": 21538 }, { "epoch": 6.63, "learning_rate": 5.404466914913577e-06, "loss": 0.0626, "step": 21539 }, { "epoch": 6.63, "learning_rate": 5.40358186045999e-06, "loss": 0.1212, "step": 21540 }, { "epoch": 6.63, "learning_rate": 5.402696851653077e-06, "loss": 0.1025, "step": 21541 }, { "epoch": 6.63, "learning_rate": 5.401811888501616e-06, "loss": 0.0829, "step": 21542 }, { "epoch": 6.63, "learning_rate": 5.400926971014403e-06, "loss": 0.0548, "step": 21543 }, { "epoch": 6.63, "learning_rate": 5.4000420992002245e-06, "loss": 0.0925, "step": 21544 }, { "epoch": 6.63, "learning_rate": 5.399157273067868e-06, "loss": 0.0832, "step": 21545 }, { "epoch": 6.63, "learning_rate": 5.3982724926261185e-06, "loss": 0.1792, "step": 21546 }, { "epoch": 6.63, "learning_rate": 5.397387757883761e-06, "loss": 0.0901, "step": 21547 }, { "epoch": 6.63, "learning_rate": 5.396503068849593e-06, "loss": 0.1141, "step": 21548 }, { "epoch": 6.63, "learning_rate": 5.39561842553239e-06, "loss": 0.1276, "step": 21549 }, { "epoch": 6.63, "learning_rate": 5.394733827940933e-06, "loss": 0.0961, "step": 21550 }, { "epoch": 6.63, "learning_rate": 5.3938492760840176e-06, "loss": 0.0662, "step": 21551 }, { "epoch": 6.63, "learning_rate": 5.392964769970426e-06, "loss": 0.0705, "step": 21552 }, { "epoch": 6.63, "learning_rate": 5.392080309608938e-06, "loss": 0.0722, "step": 21553 }, { "epoch": 6.63, "learning_rate": 5.391195895008341e-06, "loss": 0.0813, "step": 21554 }, { "epoch": 6.63, "learning_rate": 5.390311526177414e-06, "loss": 0.115, "step": 21555 }, { "epoch": 6.63, "learning_rate": 5.389427203124944e-06, "loss": 0.0824, "step": 21556 }, { "epoch": 6.63, "learning_rate": 5.388542925859708e-06, "loss": 0.1184, "step": 21557 }, { "epoch": 6.63, "learning_rate": 5.387658694390496e-06, "loss": 0.1211, "step": 21558 }, { "epoch": 6.63, "learning_rate": 5.386774508726083e-06, "loss": 0.1406, "step": 21559 }, { "epoch": 6.63, "learning_rate": 5.3858903688752436e-06, "loss": 0.0845, "step": 21560 }, { "epoch": 6.63, "learning_rate": 5.385006274846771e-06, "loss": 0.0782, "step": 21561 }, { "epoch": 6.63, "learning_rate": 5.384122226649439e-06, "loss": 0.1191, "step": 21562 }, { "epoch": 6.63, "learning_rate": 5.383238224292029e-06, "loss": 0.0864, "step": 21563 }, { "epoch": 6.63, "learning_rate": 5.382354267783316e-06, "loss": 0.1565, "step": 21564 }, { "epoch": 6.64, "learning_rate": 5.381470357132084e-06, "loss": 0.106, "step": 21565 }, { "epoch": 6.64, "learning_rate": 5.380586492347104e-06, "loss": 0.1847, "step": 21566 }, { "epoch": 6.64, "learning_rate": 5.379702673437157e-06, "loss": 0.1662, "step": 21567 }, { "epoch": 6.64, "learning_rate": 5.378818900411022e-06, "loss": 0.0861, "step": 21568 }, { "epoch": 6.64, "learning_rate": 5.37793517327748e-06, "loss": 0.1629, "step": 21569 }, { "epoch": 6.64, "learning_rate": 5.377051492045293e-06, "loss": 0.105, "step": 21570 }, { "epoch": 6.64, "learning_rate": 5.376167856723249e-06, "loss": 0.1012, "step": 21571 }, { "epoch": 6.64, "learning_rate": 5.37528426732012e-06, "loss": 0.106, "step": 21572 }, { "epoch": 6.64, "learning_rate": 5.374400723844681e-06, "loss": 0.0711, "step": 21573 }, { "epoch": 6.64, "learning_rate": 5.373517226305704e-06, "loss": 0.0942, "step": 21574 }, { "epoch": 6.64, "learning_rate": 5.3726337747119625e-06, "loss": 0.0911, "step": 21575 }, { "epoch": 6.64, "learning_rate": 5.371750369072241e-06, "loss": 0.0843, "step": 21576 }, { "epoch": 6.64, "learning_rate": 5.370867009395294e-06, "loss": 0.0594, "step": 21577 }, { "epoch": 6.64, "learning_rate": 5.369983695689908e-06, "loss": 0.1345, "step": 21578 }, { "epoch": 6.64, "learning_rate": 5.369100427964853e-06, "loss": 0.0759, "step": 21579 }, { "epoch": 6.64, "learning_rate": 5.368217206228898e-06, "loss": 0.0853, "step": 21580 }, { "epoch": 6.64, "learning_rate": 5.367334030490814e-06, "loss": 0.141, "step": 21581 }, { "epoch": 6.64, "learning_rate": 5.366450900759372e-06, "loss": 0.0852, "step": 21582 }, { "epoch": 6.64, "learning_rate": 5.365567817043345e-06, "loss": 0.0543, "step": 21583 }, { "epoch": 6.64, "learning_rate": 5.364684779351495e-06, "loss": 0.0756, "step": 21584 }, { "epoch": 6.64, "learning_rate": 5.363801787692603e-06, "loss": 0.068, "step": 21585 }, { "epoch": 6.64, "learning_rate": 5.362918842075435e-06, "loss": 0.127, "step": 21586 }, { "epoch": 6.64, "learning_rate": 5.362035942508749e-06, "loss": 0.0636, "step": 21587 }, { "epoch": 6.64, "learning_rate": 5.361153089001323e-06, "loss": 0.0632, "step": 21588 }, { "epoch": 6.64, "learning_rate": 5.360270281561924e-06, "loss": 0.1129, "step": 21589 }, { "epoch": 6.64, "learning_rate": 5.359387520199317e-06, "loss": 0.1015, "step": 21590 }, { "epoch": 6.64, "learning_rate": 5.358504804922269e-06, "loss": 0.0899, "step": 21591 }, { "epoch": 6.64, "learning_rate": 5.35762213573954e-06, "loss": 0.123, "step": 21592 }, { "epoch": 6.64, "learning_rate": 5.356739512659912e-06, "loss": 0.1025, "step": 21593 }, { "epoch": 6.64, "learning_rate": 5.355856935692131e-06, "loss": 0.0694, "step": 21594 }, { "epoch": 6.64, "learning_rate": 5.354974404844975e-06, "loss": 0.0914, "step": 21595 }, { "epoch": 6.64, "learning_rate": 5.3540919201272045e-06, "loss": 0.0752, "step": 21596 }, { "epoch": 6.64, "learning_rate": 5.353209481547582e-06, "loss": 0.0857, "step": 21597 }, { "epoch": 6.65, "learning_rate": 5.352327089114873e-06, "loss": 0.0865, "step": 21598 }, { "epoch": 6.65, "learning_rate": 5.351444742837839e-06, "loss": 0.1348, "step": 21599 }, { "epoch": 6.65, "learning_rate": 5.350562442725242e-06, "loss": 0.1136, "step": 21600 }, { "epoch": 6.65, "learning_rate": 5.349680188785846e-06, "loss": 0.092, "step": 21601 }, { "epoch": 6.65, "learning_rate": 5.348797981028408e-06, "loss": 0.0978, "step": 21602 }, { "epoch": 6.65, "learning_rate": 5.3479158194617e-06, "loss": 0.099, "step": 21603 }, { "epoch": 6.65, "learning_rate": 5.347033704094467e-06, "loss": 0.0758, "step": 21604 }, { "epoch": 6.65, "learning_rate": 5.346151634935484e-06, "loss": 0.1254, "step": 21605 }, { "epoch": 6.65, "learning_rate": 5.345269611993502e-06, "loss": 0.0711, "step": 21606 }, { "epoch": 6.65, "learning_rate": 5.344387635277283e-06, "loss": 0.1042, "step": 21607 }, { "epoch": 6.65, "learning_rate": 5.343505704795586e-06, "loss": 0.0748, "step": 21608 }, { "epoch": 6.65, "learning_rate": 5.342623820557165e-06, "loss": 0.0767, "step": 21609 }, { "epoch": 6.65, "learning_rate": 5.341741982570789e-06, "loss": 0.0748, "step": 21610 }, { "epoch": 6.65, "learning_rate": 5.340860190845205e-06, "loss": 0.105, "step": 21611 }, { "epoch": 6.65, "learning_rate": 5.339978445389169e-06, "loss": 0.0435, "step": 21612 }, { "epoch": 6.65, "learning_rate": 5.339096746211445e-06, "loss": 0.1174, "step": 21613 }, { "epoch": 6.65, "learning_rate": 5.338215093320788e-06, "loss": 0.0865, "step": 21614 }, { "epoch": 6.65, "learning_rate": 5.337333486725951e-06, "loss": 0.0952, "step": 21615 }, { "epoch": 6.65, "learning_rate": 5.336451926435688e-06, "loss": 0.1136, "step": 21616 }, { "epoch": 6.65, "learning_rate": 5.335570412458757e-06, "loss": 0.112, "step": 21617 }, { "epoch": 6.65, "learning_rate": 5.334688944803909e-06, "loss": 0.1144, "step": 21618 }, { "epoch": 6.65, "learning_rate": 5.333807523479898e-06, "loss": 0.0589, "step": 21619 }, { "epoch": 6.65, "learning_rate": 5.332926148495485e-06, "loss": 0.0762, "step": 21620 }, { "epoch": 6.65, "learning_rate": 5.332044819859414e-06, "loss": 0.0721, "step": 21621 }, { "epoch": 6.65, "learning_rate": 5.331163537580435e-06, "loss": 0.0819, "step": 21622 }, { "epoch": 6.65, "learning_rate": 5.330282301667309e-06, "loss": 0.1489, "step": 21623 }, { "epoch": 6.65, "learning_rate": 5.329401112128784e-06, "loss": 0.114, "step": 21624 }, { "epoch": 6.65, "learning_rate": 5.32851996897361e-06, "loss": 0.0744, "step": 21625 }, { "epoch": 6.65, "learning_rate": 5.3276388722105385e-06, "loss": 0.1062, "step": 21626 }, { "epoch": 6.65, "learning_rate": 5.3267578218483185e-06, "loss": 0.0709, "step": 21627 }, { "epoch": 6.65, "learning_rate": 5.325876817895701e-06, "loss": 0.0968, "step": 21628 }, { "epoch": 6.65, "learning_rate": 5.32499586036143e-06, "loss": 0.0825, "step": 21629 }, { "epoch": 6.66, "learning_rate": 5.3241149492542635e-06, "loss": 0.0787, "step": 21630 }, { "epoch": 6.66, "learning_rate": 5.323234084582948e-06, "loss": 0.0619, "step": 21631 }, { "epoch": 6.66, "learning_rate": 5.32235326635622e-06, "loss": 0.0551, "step": 21632 }, { "epoch": 6.66, "learning_rate": 5.32147249458284e-06, "loss": 0.0653, "step": 21633 }, { "epoch": 6.66, "learning_rate": 5.32059176927155e-06, "loss": 0.0846, "step": 21634 }, { "epoch": 6.66, "learning_rate": 5.319711090431097e-06, "loss": 0.0712, "step": 21635 }, { "epoch": 6.66, "learning_rate": 5.318830458070226e-06, "loss": 0.0667, "step": 21636 }, { "epoch": 6.66, "learning_rate": 5.3179498721976806e-06, "loss": 0.1026, "step": 21637 }, { "epoch": 6.66, "learning_rate": 5.317069332822211e-06, "loss": 0.0711, "step": 21638 }, { "epoch": 6.66, "learning_rate": 5.3161888399525545e-06, "loss": 0.1252, "step": 21639 }, { "epoch": 6.66, "learning_rate": 5.315308393597462e-06, "loss": 0.1136, "step": 21640 }, { "epoch": 6.66, "learning_rate": 5.314427993765679e-06, "loss": 0.0531, "step": 21641 }, { "epoch": 6.66, "learning_rate": 5.313547640465937e-06, "loss": 0.1098, "step": 21642 }, { "epoch": 6.66, "learning_rate": 5.31266733370699e-06, "loss": 0.0923, "step": 21643 }, { "epoch": 6.66, "learning_rate": 5.3117870734975765e-06, "loss": 0.0948, "step": 21644 }, { "epoch": 6.66, "learning_rate": 5.310906859846439e-06, "loss": 0.131, "step": 21645 }, { "epoch": 6.66, "learning_rate": 5.310026692762316e-06, "loss": 0.0397, "step": 21646 }, { "epoch": 6.66, "learning_rate": 5.309146572253947e-06, "loss": 0.0917, "step": 21647 }, { "epoch": 6.66, "learning_rate": 5.3082664983300835e-06, "loss": 0.1021, "step": 21648 }, { "epoch": 6.66, "learning_rate": 5.30738647099945e-06, "loss": 0.136, "step": 21649 }, { "epoch": 6.66, "learning_rate": 5.3065064902707985e-06, "loss": 0.0571, "step": 21650 }, { "epoch": 6.66, "learning_rate": 5.305626556152863e-06, "loss": 0.0704, "step": 21651 }, { "epoch": 6.66, "learning_rate": 5.304746668654382e-06, "loss": 0.096, "step": 21652 }, { "epoch": 6.66, "learning_rate": 5.3038668277840935e-06, "loss": 0.1246, "step": 21653 }, { "epoch": 6.66, "learning_rate": 5.302987033550731e-06, "loss": 0.0482, "step": 21654 }, { "epoch": 6.66, "learning_rate": 5.302107285963045e-06, "loss": 0.0759, "step": 21655 }, { "epoch": 6.66, "learning_rate": 5.30122758502976e-06, "loss": 0.1682, "step": 21656 }, { "epoch": 6.66, "learning_rate": 5.300347930759611e-06, "loss": 0.0534, "step": 21657 }, { "epoch": 6.66, "learning_rate": 5.299468323161346e-06, "loss": 0.0437, "step": 21658 }, { "epoch": 6.66, "learning_rate": 5.298588762243684e-06, "loss": 0.0945, "step": 21659 }, { "epoch": 6.66, "learning_rate": 5.2977092480153734e-06, "loss": 0.103, "step": 21660 }, { "epoch": 6.66, "learning_rate": 5.296829780485144e-06, "loss": 0.1179, "step": 21661 }, { "epoch": 6.66, "learning_rate": 5.295950359661729e-06, "loss": 0.091, "step": 21662 }, { "epoch": 6.67, "learning_rate": 5.295070985553863e-06, "loss": 0.1184, "step": 21663 }, { "epoch": 6.67, "learning_rate": 5.294191658170273e-06, "loss": 0.1133, "step": 21664 }, { "epoch": 6.67, "learning_rate": 5.293312377519706e-06, "loss": 0.1148, "step": 21665 }, { "epoch": 6.67, "learning_rate": 5.29243314361088e-06, "loss": 0.0652, "step": 21666 }, { "epoch": 6.67, "learning_rate": 5.291553956452529e-06, "loss": 0.1102, "step": 21667 }, { "epoch": 6.67, "learning_rate": 5.2906748160533895e-06, "loss": 0.053, "step": 21668 }, { "epoch": 6.67, "learning_rate": 5.2897957224221884e-06, "loss": 0.0988, "step": 21669 }, { "epoch": 6.67, "learning_rate": 5.288916675567659e-06, "loss": 0.0692, "step": 21670 }, { "epoch": 6.67, "learning_rate": 5.288037675498528e-06, "loss": 0.0937, "step": 21671 }, { "epoch": 6.67, "learning_rate": 5.287158722223524e-06, "loss": 0.043, "step": 21672 }, { "epoch": 6.67, "learning_rate": 5.28627981575138e-06, "loss": 0.0743, "step": 21673 }, { "epoch": 6.67, "learning_rate": 5.285400956090816e-06, "loss": 0.0793, "step": 21674 }, { "epoch": 6.67, "learning_rate": 5.28452214325057e-06, "loss": 0.0339, "step": 21675 }, { "epoch": 6.67, "learning_rate": 5.283643377239369e-06, "loss": 0.1327, "step": 21676 }, { "epoch": 6.67, "learning_rate": 5.282764658065929e-06, "loss": 0.0594, "step": 21677 }, { "epoch": 6.67, "learning_rate": 5.281885985738986e-06, "loss": 0.0806, "step": 21678 }, { "epoch": 6.67, "learning_rate": 5.281007360267263e-06, "loss": 0.0918, "step": 21679 }, { "epoch": 6.67, "learning_rate": 5.280128781659488e-06, "loss": 0.107, "step": 21680 }, { "epoch": 6.67, "learning_rate": 5.279250249924384e-06, "loss": 0.0683, "step": 21681 }, { "epoch": 6.67, "learning_rate": 5.278371765070673e-06, "loss": 0.086, "step": 21682 }, { "epoch": 6.67, "learning_rate": 5.277493327107085e-06, "loss": 0.075, "step": 21683 }, { "epoch": 6.67, "learning_rate": 5.276614936042336e-06, "loss": 0.1667, "step": 21684 }, { "epoch": 6.67, "learning_rate": 5.275736591885157e-06, "loss": 0.1382, "step": 21685 }, { "epoch": 6.67, "learning_rate": 5.274858294644272e-06, "loss": 0.0795, "step": 21686 }, { "epoch": 6.67, "learning_rate": 5.273980044328392e-06, "loss": 0.0627, "step": 21687 }, { "epoch": 6.67, "learning_rate": 5.273101840946249e-06, "loss": 0.119, "step": 21688 }, { "epoch": 6.67, "learning_rate": 5.272223684506561e-06, "loss": 0.0991, "step": 21689 }, { "epoch": 6.67, "learning_rate": 5.271345575018049e-06, "loss": 0.0701, "step": 21690 }, { "epoch": 6.67, "learning_rate": 5.270467512489433e-06, "loss": 0.0629, "step": 21691 }, { "epoch": 6.67, "learning_rate": 5.26958949692943e-06, "loss": 0.1233, "step": 21692 }, { "epoch": 6.67, "learning_rate": 5.268711528346771e-06, "loss": 0.0489, "step": 21693 }, { "epoch": 6.67, "learning_rate": 5.26783360675016e-06, "loss": 0.087, "step": 21694 }, { "epoch": 6.68, "learning_rate": 5.266955732148325e-06, "loss": 0.1201, "step": 21695 }, { "epoch": 6.68, "learning_rate": 5.266077904549987e-06, "loss": 0.0953, "step": 21696 }, { "epoch": 6.68, "learning_rate": 5.2652001239638495e-06, "loss": 0.0636, "step": 21697 }, { "epoch": 6.68, "learning_rate": 5.264322390398643e-06, "loss": 0.1146, "step": 21698 }, { "epoch": 6.68, "learning_rate": 5.263444703863079e-06, "loss": 0.0703, "step": 21699 }, { "epoch": 6.68, "learning_rate": 5.262567064365875e-06, "loss": 0.023, "step": 21700 }, { "epoch": 6.68, "learning_rate": 5.2616894719157455e-06, "loss": 0.1031, "step": 21701 }, { "epoch": 6.68, "learning_rate": 5.260811926521403e-06, "loss": 0.0377, "step": 21702 }, { "epoch": 6.68, "learning_rate": 5.2599344281915745e-06, "loss": 0.0813, "step": 21703 }, { "epoch": 6.68, "learning_rate": 5.259056976934957e-06, "loss": 0.0848, "step": 21704 }, { "epoch": 6.68, "learning_rate": 5.258179572760277e-06, "loss": 0.0906, "step": 21705 }, { "epoch": 6.68, "learning_rate": 5.257302215676243e-06, "loss": 0.0804, "step": 21706 }, { "epoch": 6.68, "learning_rate": 5.2564249056915704e-06, "loss": 0.1041, "step": 21707 }, { "epoch": 6.68, "learning_rate": 5.255547642814971e-06, "loss": 0.0786, "step": 21708 }, { "epoch": 6.68, "learning_rate": 5.2546704270551506e-06, "loss": 0.0815, "step": 21709 }, { "epoch": 6.68, "learning_rate": 5.253793258420835e-06, "loss": 0.0892, "step": 21710 }, { "epoch": 6.68, "learning_rate": 5.2529161369207225e-06, "loss": 0.0526, "step": 21711 }, { "epoch": 6.68, "learning_rate": 5.252039062563523e-06, "loss": 0.0681, "step": 21712 }, { "epoch": 6.68, "learning_rate": 5.251162035357957e-06, "loss": 0.0737, "step": 21713 }, { "epoch": 6.68, "learning_rate": 5.250285055312727e-06, "loss": 0.1101, "step": 21714 }, { "epoch": 6.68, "learning_rate": 5.249408122436545e-06, "loss": 0.0904, "step": 21715 }, { "epoch": 6.68, "learning_rate": 5.248531236738119e-06, "loss": 0.0675, "step": 21716 }, { "epoch": 6.68, "learning_rate": 5.247654398226157e-06, "loss": 0.0997, "step": 21717 }, { "epoch": 6.68, "learning_rate": 5.246777606909366e-06, "loss": 0.0907, "step": 21718 }, { "epoch": 6.68, "learning_rate": 5.245900862796449e-06, "loss": 0.1057, "step": 21719 }, { "epoch": 6.68, "learning_rate": 5.245024165896126e-06, "loss": 0.071, "step": 21720 }, { "epoch": 6.68, "learning_rate": 5.244147516217093e-06, "loss": 0.0651, "step": 21721 }, { "epoch": 6.68, "learning_rate": 5.243270913768054e-06, "loss": 0.0996, "step": 21722 }, { "epoch": 6.68, "learning_rate": 5.242394358557722e-06, "loss": 0.0935, "step": 21723 }, { "epoch": 6.68, "learning_rate": 5.241517850594799e-06, "loss": 0.1037, "step": 21724 }, { "epoch": 6.68, "learning_rate": 5.240641389887989e-06, "loss": 0.0692, "step": 21725 }, { "epoch": 6.68, "learning_rate": 5.239764976445998e-06, "loss": 0.0852, "step": 21726 }, { "epoch": 6.68, "learning_rate": 5.238888610277526e-06, "loss": 0.099, "step": 21727 }, { "epoch": 6.69, "learning_rate": 5.238012291391279e-06, "loss": 0.1214, "step": 21728 }, { "epoch": 6.69, "learning_rate": 5.237136019795954e-06, "loss": 0.0975, "step": 21729 }, { "epoch": 6.69, "learning_rate": 5.236259795500263e-06, "loss": 0.0832, "step": 21730 }, { "epoch": 6.69, "learning_rate": 5.235383618512906e-06, "loss": 0.1114, "step": 21731 }, { "epoch": 6.69, "learning_rate": 5.234507488842573e-06, "loss": 0.0395, "step": 21732 }, { "epoch": 6.69, "learning_rate": 5.2336314064979766e-06, "loss": 0.1241, "step": 21733 }, { "epoch": 6.69, "learning_rate": 5.232755371487812e-06, "loss": 0.0565, "step": 21734 }, { "epoch": 6.69, "learning_rate": 5.231879383820782e-06, "loss": 0.0806, "step": 21735 }, { "epoch": 6.69, "learning_rate": 5.231003443505582e-06, "loss": 0.0759, "step": 21736 }, { "epoch": 6.69, "learning_rate": 5.230127550550916e-06, "loss": 0.1472, "step": 21737 }, { "epoch": 6.69, "learning_rate": 5.229251704965477e-06, "loss": 0.0433, "step": 21738 }, { "epoch": 6.69, "learning_rate": 5.2283759067579635e-06, "loss": 0.0833, "step": 21739 }, { "epoch": 6.69, "learning_rate": 5.227500155937078e-06, "loss": 0.0865, "step": 21740 }, { "epoch": 6.69, "learning_rate": 5.2266244525115175e-06, "loss": 0.1235, "step": 21741 }, { "epoch": 6.69, "learning_rate": 5.225748796489966e-06, "loss": 0.1036, "step": 21742 }, { "epoch": 6.69, "learning_rate": 5.224873187881136e-06, "loss": 0.0657, "step": 21743 }, { "epoch": 6.69, "learning_rate": 5.223997626693716e-06, "loss": 0.0793, "step": 21744 }, { "epoch": 6.69, "learning_rate": 5.223122112936401e-06, "loss": 0.0889, "step": 21745 }, { "epoch": 6.69, "learning_rate": 5.222246646617886e-06, "loss": 0.088, "step": 21746 }, { "epoch": 6.69, "learning_rate": 5.221371227746859e-06, "loss": 0.0962, "step": 21747 }, { "epoch": 6.69, "learning_rate": 5.22049585633203e-06, "loss": 0.0917, "step": 21748 }, { "epoch": 6.69, "learning_rate": 5.219620532382072e-06, "loss": 0.1127, "step": 21749 }, { "epoch": 6.69, "learning_rate": 5.218745255905695e-06, "loss": 0.0819, "step": 21750 }, { "epoch": 6.69, "learning_rate": 5.217870026911582e-06, "loss": 0.0796, "step": 21751 }, { "epoch": 6.69, "learning_rate": 5.2169948454084275e-06, "loss": 0.0834, "step": 21752 }, { "epoch": 6.69, "learning_rate": 5.216119711404922e-06, "loss": 0.1206, "step": 21753 }, { "epoch": 6.69, "learning_rate": 5.215244624909756e-06, "loss": 0.0889, "step": 21754 }, { "epoch": 6.69, "learning_rate": 5.214369585931622e-06, "loss": 0.0563, "step": 21755 }, { "epoch": 6.69, "learning_rate": 5.2134945944792095e-06, "loss": 0.0223, "step": 21756 }, { "epoch": 6.69, "learning_rate": 5.2126196505612015e-06, "loss": 0.0671, "step": 21757 }, { "epoch": 6.69, "learning_rate": 5.211744754186301e-06, "loss": 0.0976, "step": 21758 }, { "epoch": 6.69, "learning_rate": 5.210869905363178e-06, "loss": 0.0977, "step": 21759 }, { "epoch": 6.7, "learning_rate": 5.209995104100538e-06, "loss": 0.0704, "step": 21760 }, { "epoch": 6.7, "learning_rate": 5.20912035040706e-06, "loss": 0.114, "step": 21761 }, { "epoch": 6.7, "learning_rate": 5.2082456442914305e-06, "loss": 0.0762, "step": 21762 }, { "epoch": 6.7, "learning_rate": 5.20737098576234e-06, "loss": 0.1174, "step": 21763 }, { "epoch": 6.7, "learning_rate": 5.206496374828468e-06, "loss": 0.0438, "step": 21764 }, { "epoch": 6.7, "learning_rate": 5.205621811498514e-06, "loss": 0.1166, "step": 21765 }, { "epoch": 6.7, "learning_rate": 5.204747295781147e-06, "loss": 0.0667, "step": 21766 }, { "epoch": 6.7, "learning_rate": 5.203872827685058e-06, "loss": 0.1325, "step": 21767 }, { "epoch": 6.7, "learning_rate": 5.202998407218936e-06, "loss": 0.1003, "step": 21768 }, { "epoch": 6.7, "learning_rate": 5.202124034391458e-06, "loss": 0.0863, "step": 21769 }, { "epoch": 6.7, "learning_rate": 5.201249709211312e-06, "loss": 0.1087, "step": 21770 }, { "epoch": 6.7, "learning_rate": 5.200375431687179e-06, "loss": 0.0761, "step": 21771 }, { "epoch": 6.7, "learning_rate": 5.199501201827741e-06, "loss": 0.1578, "step": 21772 }, { "epoch": 6.7, "learning_rate": 5.198627019641681e-06, "loss": 0.117, "step": 21773 }, { "epoch": 6.7, "learning_rate": 5.1977528851376756e-06, "loss": 0.1199, "step": 21774 }, { "epoch": 6.7, "learning_rate": 5.196878798324418e-06, "loss": 0.1189, "step": 21775 }, { "epoch": 6.7, "learning_rate": 5.196004759210572e-06, "loss": 0.0801, "step": 21776 }, { "epoch": 6.7, "learning_rate": 5.19513076780483e-06, "loss": 0.1109, "step": 21777 }, { "epoch": 6.7, "learning_rate": 5.194256824115869e-06, "loss": 0.0895, "step": 21778 }, { "epoch": 6.7, "learning_rate": 5.193382928152366e-06, "loss": 0.1049, "step": 21779 }, { "epoch": 6.7, "learning_rate": 5.192509079923e-06, "loss": 0.0729, "step": 21780 }, { "epoch": 6.7, "learning_rate": 5.191635279436447e-06, "loss": 0.1121, "step": 21781 }, { "epoch": 6.7, "learning_rate": 5.190761526701395e-06, "loss": 0.1249, "step": 21782 }, { "epoch": 6.7, "learning_rate": 5.189887821726509e-06, "loss": 0.1757, "step": 21783 }, { "epoch": 6.7, "learning_rate": 5.189014164520465e-06, "loss": 0.1133, "step": 21784 }, { "epoch": 6.7, "learning_rate": 5.18814055509195e-06, "loss": 0.1067, "step": 21785 }, { "epoch": 6.7, "learning_rate": 5.187266993449633e-06, "loss": 0.0926, "step": 21786 }, { "epoch": 6.7, "learning_rate": 5.186393479602189e-06, "loss": 0.0764, "step": 21787 }, { "epoch": 6.7, "learning_rate": 5.185520013558296e-06, "loss": 0.1496, "step": 21788 }, { "epoch": 6.7, "learning_rate": 5.184646595326627e-06, "loss": 0.0509, "step": 21789 }, { "epoch": 6.7, "learning_rate": 5.183773224915853e-06, "loss": 0.1073, "step": 21790 }, { "epoch": 6.7, "learning_rate": 5.182899902334648e-06, "loss": 0.0781, "step": 21791 }, { "epoch": 6.7, "learning_rate": 5.18202662759169e-06, "loss": 0.1627, "step": 21792 }, { "epoch": 6.71, "learning_rate": 5.181153400695651e-06, "loss": 0.0687, "step": 21793 }, { "epoch": 6.71, "learning_rate": 5.180280221655191e-06, "loss": 0.1163, "step": 21794 }, { "epoch": 6.71, "learning_rate": 5.179407090478996e-06, "loss": 0.0528, "step": 21795 }, { "epoch": 6.71, "learning_rate": 5.178534007175732e-06, "loss": 0.0977, "step": 21796 }, { "epoch": 6.71, "learning_rate": 5.177660971754068e-06, "loss": 0.117, "step": 21797 }, { "epoch": 6.71, "learning_rate": 5.1767879842226745e-06, "loss": 0.0612, "step": 21798 }, { "epoch": 6.71, "learning_rate": 5.175915044590222e-06, "loss": 0.0684, "step": 21799 }, { "epoch": 6.71, "learning_rate": 5.17504215286538e-06, "loss": 0.0675, "step": 21800 }, { "epoch": 6.71, "learning_rate": 5.17416930905681e-06, "loss": 0.0937, "step": 21801 }, { "epoch": 6.71, "learning_rate": 5.173296513173192e-06, "loss": 0.0803, "step": 21802 }, { "epoch": 6.71, "learning_rate": 5.172423765223191e-06, "loss": 0.1382, "step": 21803 }, { "epoch": 6.71, "learning_rate": 5.1715510652154644e-06, "loss": 0.1222, "step": 21804 }, { "epoch": 6.71, "learning_rate": 5.17067841315869e-06, "loss": 0.0675, "step": 21805 }, { "epoch": 6.71, "learning_rate": 5.169805809061528e-06, "loss": 0.0882, "step": 21806 }, { "epoch": 6.71, "learning_rate": 5.168933252932647e-06, "loss": 0.0765, "step": 21807 }, { "epoch": 6.71, "learning_rate": 5.1680607447807116e-06, "loss": 0.0817, "step": 21808 }, { "epoch": 6.71, "learning_rate": 5.1671882846143815e-06, "loss": 0.0574, "step": 21809 }, { "epoch": 6.71, "learning_rate": 5.166315872442336e-06, "loss": 0.0835, "step": 21810 }, { "epoch": 6.71, "learning_rate": 5.165443508273218e-06, "loss": 0.0575, "step": 21811 }, { "epoch": 6.71, "learning_rate": 5.164571192115706e-06, "loss": 0.1245, "step": 21812 }, { "epoch": 6.71, "learning_rate": 5.163698923978459e-06, "loss": 0.1227, "step": 21813 }, { "epoch": 6.71, "learning_rate": 5.16282670387014e-06, "loss": 0.0533, "step": 21814 }, { "epoch": 6.71, "learning_rate": 5.161954531799409e-06, "loss": 0.0994, "step": 21815 }, { "epoch": 6.71, "learning_rate": 5.161082407774927e-06, "loss": 0.1357, "step": 21816 }, { "epoch": 6.71, "learning_rate": 5.1602103318053586e-06, "loss": 0.0796, "step": 21817 }, { "epoch": 6.71, "learning_rate": 5.159338303899361e-06, "loss": 0.0681, "step": 21818 }, { "epoch": 6.71, "learning_rate": 5.158466324065592e-06, "loss": 0.0826, "step": 21819 }, { "epoch": 6.71, "learning_rate": 5.1575943923127216e-06, "loss": 0.0843, "step": 21820 }, { "epoch": 6.71, "learning_rate": 5.156722508649395e-06, "loss": 0.1271, "step": 21821 }, { "epoch": 6.71, "learning_rate": 5.155850673084279e-06, "loss": 0.0689, "step": 21822 }, { "epoch": 6.71, "learning_rate": 5.154978885626033e-06, "loss": 0.0917, "step": 21823 }, { "epoch": 6.71, "learning_rate": 5.154107146283311e-06, "loss": 0.0984, "step": 21824 }, { "epoch": 6.72, "learning_rate": 5.153235455064771e-06, "loss": 0.0738, "step": 21825 }, { "epoch": 6.72, "learning_rate": 5.152363811979066e-06, "loss": 0.0722, "step": 21826 }, { "epoch": 6.72, "learning_rate": 5.1514922170348645e-06, "loss": 0.0757, "step": 21827 }, { "epoch": 6.72, "learning_rate": 5.150620670240809e-06, "loss": 0.0942, "step": 21828 }, { "epoch": 6.72, "learning_rate": 5.1497491716055555e-06, "loss": 0.0886, "step": 21829 }, { "epoch": 6.72, "learning_rate": 5.148877721137768e-06, "loss": 0.044, "step": 21830 }, { "epoch": 6.72, "learning_rate": 5.148006318846095e-06, "loss": 0.1347, "step": 21831 }, { "epoch": 6.72, "learning_rate": 5.147134964739191e-06, "loss": 0.0751, "step": 21832 }, { "epoch": 6.72, "learning_rate": 5.14626365882571e-06, "loss": 0.1101, "step": 21833 }, { "epoch": 6.72, "learning_rate": 5.1453924011143035e-06, "loss": 0.071, "step": 21834 }, { "epoch": 6.72, "learning_rate": 5.144521191613624e-06, "loss": 0.1587, "step": 21835 }, { "epoch": 6.72, "learning_rate": 5.143650030332321e-06, "loss": 0.0858, "step": 21836 }, { "epoch": 6.72, "learning_rate": 5.1427789172790565e-06, "loss": 0.0488, "step": 21837 }, { "epoch": 6.72, "learning_rate": 5.141907852462469e-06, "loss": 0.0617, "step": 21838 }, { "epoch": 6.72, "learning_rate": 5.141036835891211e-06, "loss": 0.0944, "step": 21839 }, { "epoch": 6.72, "learning_rate": 5.14016586757394e-06, "loss": 0.0732, "step": 21840 }, { "epoch": 6.72, "learning_rate": 5.1392949475193e-06, "loss": 0.0971, "step": 21841 }, { "epoch": 6.72, "learning_rate": 5.13842407573594e-06, "loss": 0.1603, "step": 21842 }, { "epoch": 6.72, "learning_rate": 5.137553252232511e-06, "loss": 0.0633, "step": 21843 }, { "epoch": 6.72, "learning_rate": 5.136682477017658e-06, "loss": 0.1208, "step": 21844 }, { "epoch": 6.72, "learning_rate": 5.1358117501000326e-06, "loss": 0.0465, "step": 21845 }, { "epoch": 6.72, "learning_rate": 5.134941071488273e-06, "loss": 0.1219, "step": 21846 }, { "epoch": 6.72, "learning_rate": 5.134070441191038e-06, "loss": 0.1215, "step": 21847 }, { "epoch": 6.72, "learning_rate": 5.1331998592169716e-06, "loss": 0.1216, "step": 21848 }, { "epoch": 6.72, "learning_rate": 5.132329325574707e-06, "loss": 0.1489, "step": 21849 }, { "epoch": 6.72, "learning_rate": 5.131458840272905e-06, "loss": 0.1229, "step": 21850 }, { "epoch": 6.72, "learning_rate": 5.130588403320201e-06, "loss": 0.0955, "step": 21851 }, { "epoch": 6.72, "learning_rate": 5.129718014725242e-06, "loss": 0.087, "step": 21852 }, { "epoch": 6.72, "learning_rate": 5.128847674496673e-06, "loss": 0.0691, "step": 21853 }, { "epoch": 6.72, "learning_rate": 5.127977382643137e-06, "loss": 0.0682, "step": 21854 }, { "epoch": 6.72, "learning_rate": 5.127107139173273e-06, "loss": 0.1074, "step": 21855 }, { "epoch": 6.72, "learning_rate": 5.1262369440957235e-06, "loss": 0.0976, "step": 21856 }, { "epoch": 6.72, "learning_rate": 5.125366797419138e-06, "loss": 0.0668, "step": 21857 }, { "epoch": 6.73, "learning_rate": 5.124496699152155e-06, "loss": 0.1322, "step": 21858 }, { "epoch": 6.73, "learning_rate": 5.123626649303406e-06, "loss": 0.1204, "step": 21859 }, { "epoch": 6.73, "learning_rate": 5.122756647881543e-06, "loss": 0.0625, "step": 21860 }, { "epoch": 6.73, "learning_rate": 5.121886694895201e-06, "loss": 0.0942, "step": 21861 }, { "epoch": 6.73, "learning_rate": 5.1210167903530195e-06, "loss": 0.089, "step": 21862 }, { "epoch": 6.73, "learning_rate": 5.120146934263638e-06, "loss": 0.1614, "step": 21863 }, { "epoch": 6.73, "learning_rate": 5.119277126635691e-06, "loss": 0.13, "step": 21864 }, { "epoch": 6.73, "learning_rate": 5.118407367477829e-06, "loss": 0.1192, "step": 21865 }, { "epoch": 6.73, "learning_rate": 5.117537656798672e-06, "loss": 0.0708, "step": 21866 }, { "epoch": 6.73, "learning_rate": 5.11666799460687e-06, "loss": 0.0933, "step": 21867 }, { "epoch": 6.73, "learning_rate": 5.115798380911056e-06, "loss": 0.0723, "step": 21868 }, { "epoch": 6.73, "learning_rate": 5.114928815719864e-06, "loss": 0.1244, "step": 21869 }, { "epoch": 6.73, "learning_rate": 5.114059299041932e-06, "loss": 0.0908, "step": 21870 }, { "epoch": 6.73, "learning_rate": 5.113189830885892e-06, "loss": 0.0811, "step": 21871 }, { "epoch": 6.73, "learning_rate": 5.112320411260385e-06, "loss": 0.0471, "step": 21872 }, { "epoch": 6.73, "learning_rate": 5.11145104017404e-06, "loss": 0.0657, "step": 21873 }, { "epoch": 6.73, "learning_rate": 5.110581717635487e-06, "loss": 0.084, "step": 21874 }, { "epoch": 6.73, "learning_rate": 5.109712443653371e-06, "loss": 0.0821, "step": 21875 }, { "epoch": 6.73, "learning_rate": 5.10884321823631e-06, "loss": 0.0995, "step": 21876 }, { "epoch": 6.73, "learning_rate": 5.1079740413929464e-06, "loss": 0.0751, "step": 21877 }, { "epoch": 6.73, "learning_rate": 5.10710491313191e-06, "loss": 0.1411, "step": 21878 }, { "epoch": 6.73, "learning_rate": 5.106235833461832e-06, "loss": 0.0409, "step": 21879 }, { "epoch": 6.73, "learning_rate": 5.105366802391343e-06, "loss": 0.0784, "step": 21880 }, { "epoch": 6.73, "learning_rate": 5.104497819929067e-06, "loss": 0.0659, "step": 21881 }, { "epoch": 6.73, "learning_rate": 5.103628886083649e-06, "loss": 0.0787, "step": 21882 }, { "epoch": 6.73, "learning_rate": 5.102760000863704e-06, "loss": 0.078, "step": 21883 }, { "epoch": 6.73, "learning_rate": 5.1018911642778615e-06, "loss": 0.1471, "step": 21884 }, { "epoch": 6.73, "learning_rate": 5.1010223763347565e-06, "loss": 0.0593, "step": 21885 }, { "epoch": 6.73, "learning_rate": 5.100153637043017e-06, "loss": 0.063, "step": 21886 }, { "epoch": 6.73, "learning_rate": 5.099284946411266e-06, "loss": 0.0685, "step": 21887 }, { "epoch": 6.73, "learning_rate": 5.098416304448133e-06, "loss": 0.1195, "step": 21888 }, { "epoch": 6.73, "learning_rate": 5.097547711162243e-06, "loss": 0.0989, "step": 21889 }, { "epoch": 6.74, "learning_rate": 5.096679166562221e-06, "loss": 0.0502, "step": 21890 }, { "epoch": 6.74, "learning_rate": 5.095810670656693e-06, "loss": 0.1074, "step": 21891 }, { "epoch": 6.74, "learning_rate": 5.094942223454288e-06, "loss": 0.1033, "step": 21892 }, { "epoch": 6.74, "learning_rate": 5.09407382496363e-06, "loss": 0.0909, "step": 21893 }, { "epoch": 6.74, "learning_rate": 5.093205475193334e-06, "loss": 0.1178, "step": 21894 }, { "epoch": 6.74, "learning_rate": 5.092337174152032e-06, "loss": 0.1187, "step": 21895 }, { "epoch": 6.74, "learning_rate": 5.091468921848346e-06, "loss": 0.0725, "step": 21896 }, { "epoch": 6.74, "learning_rate": 5.090600718290898e-06, "loss": 0.1123, "step": 21897 }, { "epoch": 6.74, "learning_rate": 5.089732563488308e-06, "loss": 0.044, "step": 21898 }, { "epoch": 6.74, "learning_rate": 5.088864457449201e-06, "loss": 0.043, "step": 21899 }, { "epoch": 6.74, "learning_rate": 5.087996400182193e-06, "loss": 0.0284, "step": 21900 }, { "epoch": 6.74, "learning_rate": 5.087128391695908e-06, "loss": 0.0867, "step": 21901 }, { "epoch": 6.74, "learning_rate": 5.086260431998967e-06, "loss": 0.0883, "step": 21902 }, { "epoch": 6.74, "learning_rate": 5.085392521099993e-06, "loss": 0.0649, "step": 21903 }, { "epoch": 6.74, "learning_rate": 5.084524659007594e-06, "loss": 0.0801, "step": 21904 }, { "epoch": 6.74, "learning_rate": 5.083656845730398e-06, "loss": 0.0727, "step": 21905 }, { "epoch": 6.74, "learning_rate": 5.082789081277021e-06, "loss": 0.0607, "step": 21906 }, { "epoch": 6.74, "learning_rate": 5.081921365656079e-06, "loss": 0.0829, "step": 21907 }, { "epoch": 6.74, "learning_rate": 5.081053698876193e-06, "loss": 0.0885, "step": 21908 }, { "epoch": 6.74, "learning_rate": 5.080186080945971e-06, "loss": 0.0988, "step": 21909 }, { "epoch": 6.74, "learning_rate": 5.079318511874045e-06, "loss": 0.1123, "step": 21910 }, { "epoch": 6.74, "learning_rate": 5.078450991669012e-06, "loss": 0.083, "step": 21911 }, { "epoch": 6.74, "learning_rate": 5.077583520339502e-06, "loss": 0.0918, "step": 21912 }, { "epoch": 6.74, "learning_rate": 5.076716097894123e-06, "loss": 0.1297, "step": 21913 }, { "epoch": 6.74, "learning_rate": 5.07584872434149e-06, "loss": 0.1598, "step": 21914 }, { "epoch": 6.74, "learning_rate": 5.074981399690219e-06, "loss": 0.036, "step": 21915 }, { "epoch": 6.74, "learning_rate": 5.074114123948921e-06, "loss": 0.1457, "step": 21916 }, { "epoch": 6.74, "learning_rate": 5.07324689712621e-06, "loss": 0.1213, "step": 21917 }, { "epoch": 6.74, "learning_rate": 5.072379719230697e-06, "loss": 0.1206, "step": 21918 }, { "epoch": 6.74, "learning_rate": 5.071512590270992e-06, "loss": 0.0718, "step": 21919 }, { "epoch": 6.74, "learning_rate": 5.070645510255716e-06, "loss": 0.0851, "step": 21920 }, { "epoch": 6.74, "learning_rate": 5.069778479193465e-06, "loss": 0.0876, "step": 21921 }, { "epoch": 6.74, "learning_rate": 5.068911497092863e-06, "loss": 0.0905, "step": 21922 }, { "epoch": 6.75, "learning_rate": 5.068044563962513e-06, "loss": 0.0736, "step": 21923 }, { "epoch": 6.75, "learning_rate": 5.067177679811026e-06, "loss": 0.1337, "step": 21924 }, { "epoch": 6.75, "learning_rate": 5.066310844647012e-06, "loss": 0.1152, "step": 21925 }, { "epoch": 6.75, "learning_rate": 5.065444058479074e-06, "loss": 0.1351, "step": 21926 }, { "epoch": 6.75, "learning_rate": 5.064577321315832e-06, "loss": 0.0723, "step": 21927 }, { "epoch": 6.75, "learning_rate": 5.0637106331658815e-06, "loss": 0.0667, "step": 21928 }, { "epoch": 6.75, "learning_rate": 5.062843994037831e-06, "loss": 0.056, "step": 21929 }, { "epoch": 6.75, "learning_rate": 5.061977403940292e-06, "loss": 0.0549, "step": 21930 }, { "epoch": 6.75, "learning_rate": 5.061110862881871e-06, "loss": 0.0519, "step": 21931 }, { "epoch": 6.75, "learning_rate": 5.0602443708711675e-06, "loss": 0.1947, "step": 21932 }, { "epoch": 6.75, "learning_rate": 5.059377927916793e-06, "loss": 0.0886, "step": 21933 }, { "epoch": 6.75, "learning_rate": 5.058511534027349e-06, "loss": 0.0809, "step": 21934 }, { "epoch": 6.75, "learning_rate": 5.057645189211437e-06, "loss": 0.0715, "step": 21935 }, { "epoch": 6.75, "learning_rate": 5.056778893477662e-06, "loss": 0.04, "step": 21936 }, { "epoch": 6.75, "learning_rate": 5.0559126468346354e-06, "loss": 0.1137, "step": 21937 }, { "epoch": 6.75, "learning_rate": 5.0550464492909475e-06, "loss": 0.1728, "step": 21938 }, { "epoch": 6.75, "learning_rate": 5.054180300855205e-06, "loss": 0.0743, "step": 21939 }, { "epoch": 6.75, "learning_rate": 5.0533142015360115e-06, "loss": 0.0696, "step": 21940 }, { "epoch": 6.75, "learning_rate": 5.0524481513419675e-06, "loss": 0.1203, "step": 21941 }, { "epoch": 6.75, "learning_rate": 5.051582150281673e-06, "loss": 0.0937, "step": 21942 }, { "epoch": 6.75, "learning_rate": 5.05071619836373e-06, "loss": 0.0695, "step": 21943 }, { "epoch": 6.75, "learning_rate": 5.049850295596734e-06, "loss": 0.1032, "step": 21944 }, { "epoch": 6.75, "learning_rate": 5.048984441989289e-06, "loss": 0.1425, "step": 21945 }, { "epoch": 6.75, "learning_rate": 5.048118637549987e-06, "loss": 0.1499, "step": 21946 }, { "epoch": 6.75, "learning_rate": 5.047252882287435e-06, "loss": 0.112, "step": 21947 }, { "epoch": 6.75, "learning_rate": 5.046387176210229e-06, "loss": 0.0668, "step": 21948 }, { "epoch": 6.75, "learning_rate": 5.045521519326957e-06, "loss": 0.075, "step": 21949 }, { "epoch": 6.75, "learning_rate": 5.044655911646226e-06, "loss": 0.1392, "step": 21950 }, { "epoch": 6.75, "learning_rate": 5.043790353176628e-06, "loss": 0.0718, "step": 21951 }, { "epoch": 6.75, "learning_rate": 5.042924843926761e-06, "loss": 0.1146, "step": 21952 }, { "epoch": 6.75, "learning_rate": 5.0420593839052175e-06, "loss": 0.0701, "step": 21953 }, { "epoch": 6.75, "learning_rate": 5.041193973120595e-06, "loss": 0.0958, "step": 21954 }, { "epoch": 6.76, "learning_rate": 5.040328611581487e-06, "loss": 0.066, "step": 21955 }, { "epoch": 6.76, "learning_rate": 5.039463299296481e-06, "loss": 0.0917, "step": 21956 }, { "epoch": 6.76, "learning_rate": 5.038598036274181e-06, "loss": 0.0847, "step": 21957 }, { "epoch": 6.76, "learning_rate": 5.037732822523178e-06, "loss": 0.048, "step": 21958 }, { "epoch": 6.76, "learning_rate": 5.036867658052056e-06, "loss": 0.0496, "step": 21959 }, { "epoch": 6.76, "learning_rate": 5.036002542869414e-06, "loss": 0.0874, "step": 21960 }, { "epoch": 6.76, "learning_rate": 5.035137476983843e-06, "loss": 0.097, "step": 21961 }, { "epoch": 6.76, "learning_rate": 5.034272460403932e-06, "loss": 0.1064, "step": 21962 }, { "epoch": 6.76, "learning_rate": 5.033407493138271e-06, "loss": 0.0559, "step": 21963 }, { "epoch": 6.76, "learning_rate": 5.032542575195449e-06, "loss": 0.1036, "step": 21964 }, { "epoch": 6.76, "learning_rate": 5.031677706584064e-06, "loss": 0.1447, "step": 21965 }, { "epoch": 6.76, "learning_rate": 5.0308128873126916e-06, "loss": 0.0822, "step": 21966 }, { "epoch": 6.76, "learning_rate": 5.02994811738993e-06, "loss": 0.0967, "step": 21967 }, { "epoch": 6.76, "learning_rate": 5.029083396824363e-06, "loss": 0.0551, "step": 21968 }, { "epoch": 6.76, "learning_rate": 5.028218725624581e-06, "loss": 0.0779, "step": 21969 }, { "epoch": 6.76, "learning_rate": 5.027354103799169e-06, "loss": 0.0817, "step": 21970 }, { "epoch": 6.76, "learning_rate": 5.02648953135671e-06, "loss": 0.1613, "step": 21971 }, { "epoch": 6.76, "learning_rate": 5.025625008305801e-06, "loss": 0.1059, "step": 21972 }, { "epoch": 6.76, "learning_rate": 5.024760534655012e-06, "loss": 0.0811, "step": 21973 }, { "epoch": 6.76, "learning_rate": 5.02389611041294e-06, "loss": 0.0653, "step": 21974 }, { "epoch": 6.76, "learning_rate": 5.023031735588171e-06, "loss": 0.0893, "step": 21975 }, { "epoch": 6.76, "learning_rate": 5.022167410189274e-06, "loss": 0.1172, "step": 21976 }, { "epoch": 6.76, "learning_rate": 5.021303134224848e-06, "loss": 0.0512, "step": 21977 }, { "epoch": 6.76, "learning_rate": 5.0204389077034685e-06, "loss": 0.0519, "step": 21978 }, { "epoch": 6.76, "learning_rate": 5.019574730633722e-06, "loss": 0.109, "step": 21979 }, { "epoch": 6.76, "learning_rate": 5.018710603024187e-06, "loss": 0.0984, "step": 21980 }, { "epoch": 6.76, "learning_rate": 5.017846524883443e-06, "loss": 0.0809, "step": 21981 }, { "epoch": 6.76, "learning_rate": 5.016982496220083e-06, "loss": 0.116, "step": 21982 }, { "epoch": 6.76, "learning_rate": 5.016118517042672e-06, "loss": 0.0925, "step": 21983 }, { "epoch": 6.76, "learning_rate": 5.0152545873598e-06, "loss": 0.1244, "step": 21984 }, { "epoch": 6.76, "learning_rate": 5.0143907071800445e-06, "loss": 0.0866, "step": 21985 }, { "epoch": 6.76, "learning_rate": 5.013526876511985e-06, "loss": 0.0598, "step": 21986 }, { "epoch": 6.76, "learning_rate": 5.012663095364199e-06, "loss": 0.0794, "step": 21987 }, { "epoch": 6.77, "learning_rate": 5.01179936374526e-06, "loss": 0.1061, "step": 21988 }, { "epoch": 6.77, "learning_rate": 5.010935681663758e-06, "loss": 0.0583, "step": 21989 }, { "epoch": 6.77, "learning_rate": 5.0100720491282605e-06, "loss": 0.0686, "step": 21990 }, { "epoch": 6.77, "learning_rate": 5.009208466147342e-06, "loss": 0.1228, "step": 21991 }, { "epoch": 6.77, "learning_rate": 5.00834493272959e-06, "loss": 0.0733, "step": 21992 }, { "epoch": 6.77, "learning_rate": 5.007481448883567e-06, "loss": 0.1132, "step": 21993 }, { "epoch": 6.77, "learning_rate": 5.006618014617856e-06, "loss": 0.1099, "step": 21994 }, { "epoch": 6.77, "learning_rate": 5.005754629941033e-06, "loss": 0.0534, "step": 21995 }, { "epoch": 6.77, "learning_rate": 5.004891294861666e-06, "loss": 0.0863, "step": 21996 }, { "epoch": 6.77, "learning_rate": 5.004028009388335e-06, "loss": 0.1092, "step": 21997 }, { "epoch": 6.77, "learning_rate": 5.0031647735296055e-06, "loss": 0.0469, "step": 21998 }, { "epoch": 6.77, "learning_rate": 5.002301587294063e-06, "loss": 0.0833, "step": 21999 }, { "epoch": 6.77, "learning_rate": 5.001438450690268e-06, "loss": 0.0954, "step": 22000 }, { "epoch": 6.77, "learning_rate": 5.000575363726792e-06, "loss": 0.0925, "step": 22001 }, { "epoch": 6.77, "learning_rate": 4.9997123264122134e-06, "loss": 0.1387, "step": 22002 }, { "epoch": 6.77, "learning_rate": 4.998849338755101e-06, "loss": 0.0778, "step": 22003 }, { "epoch": 6.77, "learning_rate": 4.997986400764022e-06, "loss": 0.0737, "step": 22004 }, { "epoch": 6.77, "learning_rate": 4.997123512447549e-06, "loss": 0.0726, "step": 22005 }, { "epoch": 6.77, "learning_rate": 4.99626067381425e-06, "loss": 0.1341, "step": 22006 }, { "epoch": 6.77, "learning_rate": 4.9953978848726955e-06, "loss": 0.127, "step": 22007 }, { "epoch": 6.77, "learning_rate": 4.994535145631447e-06, "loss": 0.0882, "step": 22008 }, { "epoch": 6.77, "learning_rate": 4.993672456099083e-06, "loss": 0.0717, "step": 22009 }, { "epoch": 6.77, "learning_rate": 4.992809816284168e-06, "loss": 0.0418, "step": 22010 }, { "epoch": 6.77, "learning_rate": 4.991947226195259e-06, "loss": 0.1318, "step": 22011 }, { "epoch": 6.77, "learning_rate": 4.991084685840933e-06, "loss": 0.0623, "step": 22012 }, { "epoch": 6.77, "learning_rate": 4.990222195229754e-06, "loss": 0.0994, "step": 22013 }, { "epoch": 6.77, "learning_rate": 4.989359754370284e-06, "loss": 0.0801, "step": 22014 }, { "epoch": 6.77, "learning_rate": 4.98849736327109e-06, "loss": 0.1011, "step": 22015 }, { "epoch": 6.77, "learning_rate": 4.987635021940736e-06, "loss": 0.0603, "step": 22016 }, { "epoch": 6.77, "learning_rate": 4.986772730387786e-06, "loss": 0.1109, "step": 22017 }, { "epoch": 6.77, "learning_rate": 4.9859104886208e-06, "loss": 0.1409, "step": 22018 }, { "epoch": 6.77, "learning_rate": 4.985048296648346e-06, "loss": 0.0908, "step": 22019 }, { "epoch": 6.78, "learning_rate": 4.9841861544789895e-06, "loss": 0.0543, "step": 22020 }, { "epoch": 6.78, "learning_rate": 4.9833240621212795e-06, "loss": 0.097, "step": 22021 }, { "epoch": 6.78, "learning_rate": 4.982462019583788e-06, "loss": 0.1243, "step": 22022 }, { "epoch": 6.78, "learning_rate": 4.9816000268750746e-06, "loss": 0.0541, "step": 22023 }, { "epoch": 6.78, "learning_rate": 4.980738084003696e-06, "loss": 0.0923, "step": 22024 }, { "epoch": 6.78, "learning_rate": 4.979876190978215e-06, "loss": 0.0778, "step": 22025 }, { "epoch": 6.78, "learning_rate": 4.9790143478071866e-06, "loss": 0.1143, "step": 22026 }, { "epoch": 6.78, "learning_rate": 4.978152554499181e-06, "loss": 0.1328, "step": 22027 }, { "epoch": 6.78, "learning_rate": 4.97729081106274e-06, "loss": 0.1225, "step": 22028 }, { "epoch": 6.78, "learning_rate": 4.976429117506435e-06, "loss": 0.0569, "step": 22029 }, { "epoch": 6.78, "learning_rate": 4.975567473838818e-06, "loss": 0.1177, "step": 22030 }, { "epoch": 6.78, "learning_rate": 4.974705880068448e-06, "loss": 0.0392, "step": 22031 }, { "epoch": 6.78, "learning_rate": 4.973844336203879e-06, "loss": 0.0957, "step": 22032 }, { "epoch": 6.78, "learning_rate": 4.9729828422536685e-06, "loss": 0.0839, "step": 22033 }, { "epoch": 6.78, "learning_rate": 4.972121398226371e-06, "loss": 0.0767, "step": 22034 }, { "epoch": 6.78, "learning_rate": 4.9712600041305425e-06, "loss": 0.1214, "step": 22035 }, { "epoch": 6.78, "learning_rate": 4.970398659974732e-06, "loss": 0.1245, "step": 22036 }, { "epoch": 6.78, "learning_rate": 4.969537365767507e-06, "loss": 0.0972, "step": 22037 }, { "epoch": 6.78, "learning_rate": 4.968676121517404e-06, "loss": 0.1027, "step": 22038 }, { "epoch": 6.78, "learning_rate": 4.967814927232988e-06, "loss": 0.1066, "step": 22039 }, { "epoch": 6.78, "learning_rate": 4.966953782922806e-06, "loss": 0.0638, "step": 22040 }, { "epoch": 6.78, "learning_rate": 4.9660926885954145e-06, "loss": 0.0602, "step": 22041 }, { "epoch": 6.78, "learning_rate": 4.965231644259359e-06, "loss": 0.0667, "step": 22042 }, { "epoch": 6.78, "learning_rate": 4.964370649923191e-06, "loss": 0.0519, "step": 22043 }, { "epoch": 6.78, "learning_rate": 4.963509705595472e-06, "loss": 0.0544, "step": 22044 }, { "epoch": 6.78, "learning_rate": 4.9626488112847384e-06, "loss": 0.0804, "step": 22045 }, { "epoch": 6.78, "learning_rate": 4.9617879669995415e-06, "loss": 0.0539, "step": 22046 }, { "epoch": 6.78, "learning_rate": 4.960927172748437e-06, "loss": 0.1049, "step": 22047 }, { "epoch": 6.78, "learning_rate": 4.960066428539969e-06, "loss": 0.0517, "step": 22048 }, { "epoch": 6.78, "learning_rate": 4.9592057343826875e-06, "loss": 0.1138, "step": 22049 }, { "epoch": 6.78, "learning_rate": 4.958345090285138e-06, "loss": 0.0751, "step": 22050 }, { "epoch": 6.78, "learning_rate": 4.957484496255869e-06, "loss": 0.0915, "step": 22051 }, { "epoch": 6.78, "learning_rate": 4.956623952303426e-06, "loss": 0.0773, "step": 22052 }, { "epoch": 6.79, "learning_rate": 4.955763458436351e-06, "loss": 0.0506, "step": 22053 }, { "epoch": 6.79, "learning_rate": 4.954903014663201e-06, "loss": 0.1146, "step": 22054 }, { "epoch": 6.79, "learning_rate": 4.95404262099251e-06, "loss": 0.0531, "step": 22055 }, { "epoch": 6.79, "learning_rate": 4.953182277432824e-06, "loss": 0.1107, "step": 22056 }, { "epoch": 6.79, "learning_rate": 4.952321983992693e-06, "loss": 0.1557, "step": 22057 }, { "epoch": 6.79, "learning_rate": 4.951461740680655e-06, "loss": 0.1021, "step": 22058 }, { "epoch": 6.79, "learning_rate": 4.950601547505256e-06, "loss": 0.077, "step": 22059 }, { "epoch": 6.79, "learning_rate": 4.949741404475037e-06, "loss": 0.0978, "step": 22060 }, { "epoch": 6.79, "learning_rate": 4.94888131159854e-06, "loss": 0.146, "step": 22061 }, { "epoch": 6.79, "learning_rate": 4.948021268884308e-06, "loss": 0.0708, "step": 22062 }, { "epoch": 6.79, "learning_rate": 4.947161276340876e-06, "loss": 0.0818, "step": 22063 }, { "epoch": 6.79, "learning_rate": 4.9463013339767925e-06, "loss": 0.0807, "step": 22064 }, { "epoch": 6.79, "learning_rate": 4.9454414418006e-06, "loss": 0.0592, "step": 22065 }, { "epoch": 6.79, "learning_rate": 4.944581599820823e-06, "loss": 0.1015, "step": 22066 }, { "epoch": 6.79, "learning_rate": 4.943721808046015e-06, "loss": 0.1291, "step": 22067 }, { "epoch": 6.79, "learning_rate": 4.9428620664847095e-06, "loss": 0.1527, "step": 22068 }, { "epoch": 6.79, "learning_rate": 4.942002375145444e-06, "loss": 0.1324, "step": 22069 }, { "epoch": 6.79, "learning_rate": 4.9411427340367555e-06, "loss": 0.1053, "step": 22070 }, { "epoch": 6.79, "learning_rate": 4.9402831431671834e-06, "loss": 0.0603, "step": 22071 }, { "epoch": 6.79, "learning_rate": 4.939423602545263e-06, "loss": 0.0631, "step": 22072 }, { "epoch": 6.79, "learning_rate": 4.938564112179526e-06, "loss": 0.1453, "step": 22073 }, { "epoch": 6.79, "learning_rate": 4.937704672078516e-06, "loss": 0.1483, "step": 22074 }, { "epoch": 6.79, "learning_rate": 4.936845282250767e-06, "loss": 0.1036, "step": 22075 }, { "epoch": 6.79, "learning_rate": 4.935985942704804e-06, "loss": 0.0931, "step": 22076 }, { "epoch": 6.79, "learning_rate": 4.935126653449172e-06, "loss": 0.0947, "step": 22077 }, { "epoch": 6.79, "learning_rate": 4.9342674144924e-06, "loss": 0.072, "step": 22078 }, { "epoch": 6.79, "learning_rate": 4.93340822584302e-06, "loss": 0.0892, "step": 22079 }, { "epoch": 6.79, "learning_rate": 4.932549087509567e-06, "loss": 0.0623, "step": 22080 }, { "epoch": 6.79, "learning_rate": 4.931689999500567e-06, "loss": 0.1004, "step": 22081 }, { "epoch": 6.79, "learning_rate": 4.930830961824566e-06, "loss": 0.0977, "step": 22082 }, { "epoch": 6.79, "learning_rate": 4.929971974490077e-06, "loss": 0.0484, "step": 22083 }, { "epoch": 6.79, "learning_rate": 4.929113037505642e-06, "loss": 0.081, "step": 22084 }, { "epoch": 6.8, "learning_rate": 4.9282541508797875e-06, "loss": 0.0745, "step": 22085 }, { "epoch": 6.8, "learning_rate": 4.927395314621045e-06, "loss": 0.1261, "step": 22086 }, { "epoch": 6.8, "learning_rate": 4.926536528737941e-06, "loss": 0.12, "step": 22087 }, { "epoch": 6.8, "learning_rate": 4.925677793239002e-06, "loss": 0.0669, "step": 22088 }, { "epoch": 6.8, "learning_rate": 4.924819108132767e-06, "loss": 0.0333, "step": 22089 }, { "epoch": 6.8, "learning_rate": 4.9239604734277515e-06, "loss": 0.0941, "step": 22090 }, { "epoch": 6.8, "learning_rate": 4.923101889132483e-06, "loss": 0.1089, "step": 22091 }, { "epoch": 6.8, "learning_rate": 4.922243355255498e-06, "loss": 0.0845, "step": 22092 }, { "epoch": 6.8, "learning_rate": 4.9213848718053105e-06, "loss": 0.068, "step": 22093 }, { "epoch": 6.8, "learning_rate": 4.920526438790455e-06, "loss": 0.1274, "step": 22094 }, { "epoch": 6.8, "learning_rate": 4.919668056219453e-06, "loss": 0.0739, "step": 22095 }, { "epoch": 6.8, "learning_rate": 4.91880972410083e-06, "loss": 0.1202, "step": 22096 }, { "epoch": 6.8, "learning_rate": 4.91795144244311e-06, "loss": 0.0945, "step": 22097 }, { "epoch": 6.8, "learning_rate": 4.917093211254811e-06, "loss": 0.1315, "step": 22098 }, { "epoch": 6.8, "learning_rate": 4.916235030544468e-06, "loss": 0.0707, "step": 22099 }, { "epoch": 6.8, "learning_rate": 4.915376900320594e-06, "loss": 0.0878, "step": 22100 }, { "epoch": 6.8, "learning_rate": 4.9145188205917095e-06, "loss": 0.0581, "step": 22101 }, { "epoch": 6.8, "learning_rate": 4.913660791366343e-06, "loss": 0.0832, "step": 22102 }, { "epoch": 6.8, "learning_rate": 4.9128028126530135e-06, "loss": 0.1208, "step": 22103 }, { "epoch": 6.8, "learning_rate": 4.9119448844602406e-06, "loss": 0.0988, "step": 22104 }, { "epoch": 6.8, "learning_rate": 4.911087006796543e-06, "loss": 0.0803, "step": 22105 }, { "epoch": 6.8, "learning_rate": 4.910229179670443e-06, "loss": 0.0974, "step": 22106 }, { "epoch": 6.8, "learning_rate": 4.909371403090457e-06, "loss": 0.0855, "step": 22107 }, { "epoch": 6.8, "learning_rate": 4.908513677065101e-06, "loss": 0.0541, "step": 22108 }, { "epoch": 6.8, "learning_rate": 4.907656001602901e-06, "loss": 0.0698, "step": 22109 }, { "epoch": 6.8, "learning_rate": 4.9067983767123736e-06, "loss": 0.0592, "step": 22110 }, { "epoch": 6.8, "learning_rate": 4.905940802402025e-06, "loss": 0.0913, "step": 22111 }, { "epoch": 6.8, "learning_rate": 4.905083278680383e-06, "loss": 0.1046, "step": 22112 }, { "epoch": 6.8, "learning_rate": 4.9042258055559584e-06, "loss": 0.1441, "step": 22113 }, { "epoch": 6.8, "learning_rate": 4.903368383037269e-06, "loss": 0.1481, "step": 22114 }, { "epoch": 6.8, "learning_rate": 4.902511011132829e-06, "loss": 0.0881, "step": 22115 }, { "epoch": 6.8, "learning_rate": 4.901653689851152e-06, "loss": 0.0859, "step": 22116 }, { "epoch": 6.8, "learning_rate": 4.900796419200753e-06, "loss": 0.0938, "step": 22117 }, { "epoch": 6.81, "learning_rate": 4.899939199190141e-06, "loss": 0.0911, "step": 22118 }, { "epoch": 6.81, "learning_rate": 4.899082029827838e-06, "loss": 0.0828, "step": 22119 }, { "epoch": 6.81, "learning_rate": 4.898224911122354e-06, "loss": 0.0907, "step": 22120 }, { "epoch": 6.81, "learning_rate": 4.897367843082191e-06, "loss": 0.088, "step": 22121 }, { "epoch": 6.81, "learning_rate": 4.8965108257158735e-06, "loss": 0.0849, "step": 22122 }, { "epoch": 6.81, "learning_rate": 4.895653859031906e-06, "loss": 0.0785, "step": 22123 }, { "epoch": 6.81, "learning_rate": 4.8947969430388e-06, "loss": 0.0636, "step": 22124 }, { "epoch": 6.81, "learning_rate": 4.893940077745065e-06, "loss": 0.0803, "step": 22125 }, { "epoch": 6.81, "learning_rate": 4.893083263159207e-06, "loss": 0.0542, "step": 22126 }, { "epoch": 6.81, "learning_rate": 4.892226499289747e-06, "loss": 0.1061, "step": 22127 }, { "epoch": 6.81, "learning_rate": 4.891369786145178e-06, "loss": 0.039, "step": 22128 }, { "epoch": 6.81, "learning_rate": 4.890513123734019e-06, "loss": 0.1083, "step": 22129 }, { "epoch": 6.81, "learning_rate": 4.889656512064773e-06, "loss": 0.0728, "step": 22130 }, { "epoch": 6.81, "learning_rate": 4.888799951145948e-06, "loss": 0.039, "step": 22131 }, { "epoch": 6.81, "learning_rate": 4.887943440986051e-06, "loss": 0.194, "step": 22132 }, { "epoch": 6.81, "learning_rate": 4.887086981593585e-06, "loss": 0.0773, "step": 22133 }, { "epoch": 6.81, "learning_rate": 4.886230572977059e-06, "loss": 0.0981, "step": 22134 }, { "epoch": 6.81, "learning_rate": 4.885374215144976e-06, "loss": 0.0822, "step": 22135 }, { "epoch": 6.81, "learning_rate": 4.884517908105837e-06, "loss": 0.097, "step": 22136 }, { "epoch": 6.81, "learning_rate": 4.883661651868158e-06, "loss": 0.0987, "step": 22137 }, { "epoch": 6.81, "learning_rate": 4.8828054464404255e-06, "loss": 0.04, "step": 22138 }, { "epoch": 6.81, "learning_rate": 4.881949291831155e-06, "loss": 0.0895, "step": 22139 }, { "epoch": 6.81, "learning_rate": 4.881093188048846e-06, "loss": 0.1078, "step": 22140 }, { "epoch": 6.81, "learning_rate": 4.880237135101998e-06, "loss": 0.0719, "step": 22141 }, { "epoch": 6.81, "learning_rate": 4.879381132999115e-06, "loss": 0.0426, "step": 22142 }, { "epoch": 6.81, "learning_rate": 4.878525181748692e-06, "loss": 0.0777, "step": 22143 }, { "epoch": 6.81, "learning_rate": 4.877669281359242e-06, "loss": 0.1191, "step": 22144 }, { "epoch": 6.81, "learning_rate": 4.8768134318392535e-06, "loss": 0.0408, "step": 22145 }, { "epoch": 6.81, "learning_rate": 4.875957633197224e-06, "loss": 0.1007, "step": 22146 }, { "epoch": 6.81, "learning_rate": 4.875101885441663e-06, "loss": 0.058, "step": 22147 }, { "epoch": 6.81, "learning_rate": 4.874246188581063e-06, "loss": 0.0504, "step": 22148 }, { "epoch": 6.81, "learning_rate": 4.873390542623922e-06, "loss": 0.0852, "step": 22149 }, { "epoch": 6.82, "learning_rate": 4.872534947578739e-06, "loss": 0.0477, "step": 22150 }, { "epoch": 6.82, "learning_rate": 4.871679403454008e-06, "loss": 0.1027, "step": 22151 }, { "epoch": 6.82, "learning_rate": 4.870823910258228e-06, "loss": 0.1048, "step": 22152 }, { "epoch": 6.82, "learning_rate": 4.869968467999889e-06, "loss": 0.112, "step": 22153 }, { "epoch": 6.82, "learning_rate": 4.8691130766875e-06, "loss": 0.0511, "step": 22154 }, { "epoch": 6.82, "learning_rate": 4.868257736329543e-06, "loss": 0.0817, "step": 22155 }, { "epoch": 6.82, "learning_rate": 4.867402446934512e-06, "loss": 0.139, "step": 22156 }, { "epoch": 6.82, "learning_rate": 4.8665472085109075e-06, "loss": 0.1004, "step": 22157 }, { "epoch": 6.82, "learning_rate": 4.865692021067222e-06, "loss": 0.1258, "step": 22158 }, { "epoch": 6.82, "learning_rate": 4.8648368846119465e-06, "loss": 0.075, "step": 22159 }, { "epoch": 6.82, "learning_rate": 4.863981799153573e-06, "loss": 0.1452, "step": 22160 }, { "epoch": 6.82, "learning_rate": 4.863126764700594e-06, "loss": 0.0697, "step": 22161 }, { "epoch": 6.82, "learning_rate": 4.8622717812615e-06, "loss": 0.1092, "step": 22162 }, { "epoch": 6.82, "learning_rate": 4.861416848844779e-06, "loss": 0.1216, "step": 22163 }, { "epoch": 6.82, "learning_rate": 4.860561967458928e-06, "loss": 0.0423, "step": 22164 }, { "epoch": 6.82, "learning_rate": 4.859707137112437e-06, "loss": 0.0916, "step": 22165 }, { "epoch": 6.82, "learning_rate": 4.858852357813784e-06, "loss": 0.0898, "step": 22166 }, { "epoch": 6.82, "learning_rate": 4.857997629571471e-06, "loss": 0.0734, "step": 22167 }, { "epoch": 6.82, "learning_rate": 4.857142952393978e-06, "loss": 0.1229, "step": 22168 }, { "epoch": 6.82, "learning_rate": 4.856288326289796e-06, "loss": 0.1671, "step": 22169 }, { "epoch": 6.82, "learning_rate": 4.855433751267408e-06, "loss": 0.0963, "step": 22170 }, { "epoch": 6.82, "learning_rate": 4.854579227335312e-06, "loss": 0.0258, "step": 22171 }, { "epoch": 6.82, "learning_rate": 4.853724754501983e-06, "loss": 0.099, "step": 22172 }, { "epoch": 6.82, "learning_rate": 4.852870332775905e-06, "loss": 0.0665, "step": 22173 }, { "epoch": 6.82, "learning_rate": 4.8520159621655725e-06, "loss": 0.1527, "step": 22174 }, { "epoch": 6.82, "learning_rate": 4.851161642679466e-06, "loss": 0.072, "step": 22175 }, { "epoch": 6.82, "learning_rate": 4.850307374326071e-06, "loss": 0.09, "step": 22176 }, { "epoch": 6.82, "learning_rate": 4.8494531571138685e-06, "loss": 0.1668, "step": 22177 }, { "epoch": 6.82, "learning_rate": 4.848598991051343e-06, "loss": 0.1098, "step": 22178 }, { "epoch": 6.82, "learning_rate": 4.847744876146978e-06, "loss": 0.0352, "step": 22179 }, { "epoch": 6.82, "learning_rate": 4.846890812409251e-06, "loss": 0.1015, "step": 22180 }, { "epoch": 6.82, "learning_rate": 4.846036799846651e-06, "loss": 0.0582, "step": 22181 }, { "epoch": 6.82, "learning_rate": 4.845182838467659e-06, "loss": 0.068, "step": 22182 }, { "epoch": 6.83, "learning_rate": 4.844328928280745e-06, "loss": 0.0537, "step": 22183 }, { "epoch": 6.83, "learning_rate": 4.8434750692944e-06, "loss": 0.0939, "step": 22184 }, { "epoch": 6.83, "learning_rate": 4.8426212615171e-06, "loss": 0.0906, "step": 22185 }, { "epoch": 6.83, "learning_rate": 4.841767504957324e-06, "loss": 0.1496, "step": 22186 }, { "epoch": 6.83, "learning_rate": 4.840913799623551e-06, "loss": 0.0819, "step": 22187 }, { "epoch": 6.83, "learning_rate": 4.840060145524254e-06, "loss": 0.0787, "step": 22188 }, { "epoch": 6.83, "learning_rate": 4.839206542667925e-06, "loss": 0.1224, "step": 22189 }, { "epoch": 6.83, "learning_rate": 4.838352991063022e-06, "loss": 0.105, "step": 22190 }, { "epoch": 6.83, "learning_rate": 4.837499490718035e-06, "loss": 0.0596, "step": 22191 }, { "epoch": 6.83, "learning_rate": 4.8366460416414394e-06, "loss": 0.1272, "step": 22192 }, { "epoch": 6.83, "learning_rate": 4.8357926438417e-06, "loss": 0.138, "step": 22193 }, { "epoch": 6.83, "learning_rate": 4.8349392973273035e-06, "loss": 0.1413, "step": 22194 }, { "epoch": 6.83, "learning_rate": 4.834086002106719e-06, "loss": 0.1082, "step": 22195 }, { "epoch": 6.83, "learning_rate": 4.833232758188423e-06, "loss": 0.0791, "step": 22196 }, { "epoch": 6.83, "learning_rate": 4.832379565580887e-06, "loss": 0.111, "step": 22197 }, { "epoch": 6.83, "learning_rate": 4.83152642429258e-06, "loss": 0.0714, "step": 22198 }, { "epoch": 6.83, "learning_rate": 4.830673334331986e-06, "loss": 0.0651, "step": 22199 }, { "epoch": 6.83, "learning_rate": 4.829820295707564e-06, "loss": 0.0714, "step": 22200 }, { "epoch": 6.83, "learning_rate": 4.828967308427795e-06, "loss": 0.088, "step": 22201 }, { "epoch": 6.83, "learning_rate": 4.828114372501146e-06, "loss": 0.1126, "step": 22202 }, { "epoch": 6.83, "learning_rate": 4.8272614879360875e-06, "loss": 0.1081, "step": 22203 }, { "epoch": 6.83, "learning_rate": 4.826408654741091e-06, "loss": 0.1437, "step": 22204 }, { "epoch": 6.83, "learning_rate": 4.8255558729246195e-06, "loss": 0.2047, "step": 22205 }, { "epoch": 6.83, "learning_rate": 4.8247031424951555e-06, "loss": 0.0807, "step": 22206 }, { "epoch": 6.83, "learning_rate": 4.823850463461155e-06, "loss": 0.0807, "step": 22207 }, { "epoch": 6.83, "learning_rate": 4.822997835831086e-06, "loss": 0.0663, "step": 22208 }, { "epoch": 6.83, "learning_rate": 4.822145259613423e-06, "loss": 0.0859, "step": 22209 }, { "epoch": 6.83, "learning_rate": 4.82129273481663e-06, "loss": 0.0803, "step": 22210 }, { "epoch": 6.83, "learning_rate": 4.820440261449173e-06, "loss": 0.0588, "step": 22211 }, { "epoch": 6.83, "learning_rate": 4.819587839519518e-06, "loss": 0.047, "step": 22212 }, { "epoch": 6.83, "learning_rate": 4.818735469036131e-06, "loss": 0.1372, "step": 22213 }, { "epoch": 6.83, "learning_rate": 4.817883150007474e-06, "loss": 0.1265, "step": 22214 }, { "epoch": 6.84, "learning_rate": 4.817030882442011e-06, "loss": 0.1047, "step": 22215 }, { "epoch": 6.84, "learning_rate": 4.816178666348216e-06, "loss": 0.1083, "step": 22216 }, { "epoch": 6.84, "learning_rate": 4.81532650173454e-06, "loss": 0.0588, "step": 22217 }, { "epoch": 6.84, "learning_rate": 4.814474388609446e-06, "loss": 0.0904, "step": 22218 }, { "epoch": 6.84, "learning_rate": 4.813622326981404e-06, "loss": 0.0719, "step": 22219 }, { "epoch": 6.84, "learning_rate": 4.812770316858872e-06, "loss": 0.1347, "step": 22220 }, { "epoch": 6.84, "learning_rate": 4.811918358250313e-06, "loss": 0.0985, "step": 22221 }, { "epoch": 6.84, "learning_rate": 4.811066451164185e-06, "loss": 0.0621, "step": 22222 }, { "epoch": 6.84, "learning_rate": 4.81021459560895e-06, "loss": 0.0986, "step": 22223 }, { "epoch": 6.84, "learning_rate": 4.809362791593067e-06, "loss": 0.111, "step": 22224 }, { "epoch": 6.84, "learning_rate": 4.808511039124991e-06, "loss": 0.122, "step": 22225 }, { "epoch": 6.84, "learning_rate": 4.80765933821319e-06, "loss": 0.1405, "step": 22226 }, { "epoch": 6.84, "learning_rate": 4.806807688866119e-06, "loss": 0.13, "step": 22227 }, { "epoch": 6.84, "learning_rate": 4.805956091092228e-06, "loss": 0.0625, "step": 22228 }, { "epoch": 6.84, "learning_rate": 4.805104544899983e-06, "loss": 0.0718, "step": 22229 }, { "epoch": 6.84, "learning_rate": 4.804253050297838e-06, "loss": 0.0522, "step": 22230 }, { "epoch": 6.84, "learning_rate": 4.803401607294249e-06, "loss": 0.154, "step": 22231 }, { "epoch": 6.84, "learning_rate": 4.80255021589767e-06, "loss": 0.0939, "step": 22232 }, { "epoch": 6.84, "learning_rate": 4.801698876116558e-06, "loss": 0.0844, "step": 22233 }, { "epoch": 6.84, "learning_rate": 4.800847587959369e-06, "loss": 0.0964, "step": 22234 }, { "epoch": 6.84, "learning_rate": 4.79999635143455e-06, "loss": 0.0762, "step": 22235 }, { "epoch": 6.84, "learning_rate": 4.799145166550563e-06, "loss": 0.1271, "step": 22236 }, { "epoch": 6.84, "learning_rate": 4.798294033315861e-06, "loss": 0.1253, "step": 22237 }, { "epoch": 6.84, "learning_rate": 4.797442951738887e-06, "loss": 0.1725, "step": 22238 }, { "epoch": 6.84, "learning_rate": 4.796591921828102e-06, "loss": 0.1085, "step": 22239 }, { "epoch": 6.84, "learning_rate": 4.795740943591955e-06, "loss": 0.0731, "step": 22240 }, { "epoch": 6.84, "learning_rate": 4.794890017038898e-06, "loss": 0.0491, "step": 22241 }, { "epoch": 6.84, "learning_rate": 4.794039142177379e-06, "loss": 0.0551, "step": 22242 }, { "epoch": 6.84, "learning_rate": 4.793188319015847e-06, "loss": 0.0728, "step": 22243 }, { "epoch": 6.84, "learning_rate": 4.79233754756276e-06, "loss": 0.0999, "step": 22244 }, { "epoch": 6.84, "learning_rate": 4.791486827826554e-06, "loss": 0.1189, "step": 22245 }, { "epoch": 6.84, "learning_rate": 4.790636159815686e-06, "loss": 0.1114, "step": 22246 }, { "epoch": 6.84, "learning_rate": 4.789785543538605e-06, "loss": 0.0727, "step": 22247 }, { "epoch": 6.85, "learning_rate": 4.788934979003752e-06, "loss": 0.1131, "step": 22248 }, { "epoch": 6.85, "learning_rate": 4.788084466219579e-06, "loss": 0.0529, "step": 22249 }, { "epoch": 6.85, "learning_rate": 4.787234005194531e-06, "loss": 0.1184, "step": 22250 }, { "epoch": 6.85, "learning_rate": 4.786383595937054e-06, "loss": 0.0721, "step": 22251 }, { "epoch": 6.85, "learning_rate": 4.785533238455592e-06, "loss": 0.0726, "step": 22252 }, { "epoch": 6.85, "learning_rate": 4.784682932758588e-06, "loss": 0.0838, "step": 22253 }, { "epoch": 6.85, "learning_rate": 4.783832678854495e-06, "loss": 0.0533, "step": 22254 }, { "epoch": 6.85, "learning_rate": 4.7829824767517444e-06, "loss": 0.1576, "step": 22255 }, { "epoch": 6.85, "learning_rate": 4.7821323264587895e-06, "loss": 0.0721, "step": 22256 }, { "epoch": 6.85, "learning_rate": 4.781282227984069e-06, "loss": 0.0896, "step": 22257 }, { "epoch": 6.85, "learning_rate": 4.780432181336026e-06, "loss": 0.0795, "step": 22258 }, { "epoch": 6.85, "learning_rate": 4.779582186523102e-06, "loss": 0.0505, "step": 22259 }, { "epoch": 6.85, "learning_rate": 4.778732243553734e-06, "loss": 0.1618, "step": 22260 }, { "epoch": 6.85, "learning_rate": 4.777882352436376e-06, "loss": 0.1559, "step": 22261 }, { "epoch": 6.85, "learning_rate": 4.777032513179454e-06, "loss": 0.0908, "step": 22262 }, { "epoch": 6.85, "learning_rate": 4.776182725791409e-06, "loss": 0.0849, "step": 22263 }, { "epoch": 6.85, "learning_rate": 4.7753329902806875e-06, "loss": 0.1732, "step": 22264 }, { "epoch": 6.85, "learning_rate": 4.774483306655725e-06, "loss": 0.1285, "step": 22265 }, { "epoch": 6.85, "learning_rate": 4.77363367492496e-06, "loss": 0.1073, "step": 22266 }, { "epoch": 6.85, "learning_rate": 4.7727840950968284e-06, "loss": 0.038, "step": 22267 }, { "epoch": 6.85, "learning_rate": 4.771934567179769e-06, "loss": 0.0655, "step": 22268 }, { "epoch": 6.85, "learning_rate": 4.771085091182217e-06, "loss": 0.1228, "step": 22269 }, { "epoch": 6.85, "learning_rate": 4.770235667112605e-06, "loss": 0.1001, "step": 22270 }, { "epoch": 6.85, "learning_rate": 4.769386294979381e-06, "loss": 0.0418, "step": 22271 }, { "epoch": 6.85, "learning_rate": 4.7685369747909685e-06, "loss": 0.0816, "step": 22272 }, { "epoch": 6.85, "learning_rate": 4.7676877065558e-06, "loss": 0.0782, "step": 22273 }, { "epoch": 6.85, "learning_rate": 4.76683849028232e-06, "loss": 0.12, "step": 22274 }, { "epoch": 6.85, "learning_rate": 4.765989325978956e-06, "loss": 0.0708, "step": 22275 }, { "epoch": 6.85, "learning_rate": 4.765140213654143e-06, "loss": 0.1099, "step": 22276 }, { "epoch": 6.85, "learning_rate": 4.764291153316311e-06, "loss": 0.0785, "step": 22277 }, { "epoch": 6.85, "learning_rate": 4.7634421449738935e-06, "loss": 0.0853, "step": 22278 }, { "epoch": 6.85, "learning_rate": 4.7625931886353215e-06, "loss": 0.0961, "step": 22279 }, { "epoch": 6.86, "learning_rate": 4.761744284309022e-06, "loss": 0.1246, "step": 22280 }, { "epoch": 6.86, "learning_rate": 4.760895432003434e-06, "loss": 0.0688, "step": 22281 }, { "epoch": 6.86, "learning_rate": 4.760046631726987e-06, "loss": 0.0825, "step": 22282 }, { "epoch": 6.86, "learning_rate": 4.759197883488098e-06, "loss": 0.1525, "step": 22283 }, { "epoch": 6.86, "learning_rate": 4.758349187295207e-06, "loss": 0.1057, "step": 22284 }, { "epoch": 6.86, "learning_rate": 4.7575005431567415e-06, "loss": 0.0859, "step": 22285 }, { "epoch": 6.86, "learning_rate": 4.756651951081126e-06, "loss": 0.1529, "step": 22286 }, { "epoch": 6.86, "learning_rate": 4.755803411076788e-06, "loss": 0.1477, "step": 22287 }, { "epoch": 6.86, "learning_rate": 4.754954923152153e-06, "loss": 0.0996, "step": 22288 }, { "epoch": 6.86, "learning_rate": 4.754106487315657e-06, "loss": 0.108, "step": 22289 }, { "epoch": 6.86, "learning_rate": 4.7532581035757105e-06, "loss": 0.12, "step": 22290 }, { "epoch": 6.86, "learning_rate": 4.75240977194075e-06, "loss": 0.1004, "step": 22291 }, { "epoch": 6.86, "learning_rate": 4.7515614924192026e-06, "loss": 0.0826, "step": 22292 }, { "epoch": 6.86, "learning_rate": 4.750713265019476e-06, "loss": 0.0785, "step": 22293 }, { "epoch": 6.86, "learning_rate": 4.749865089750011e-06, "loss": 0.1236, "step": 22294 }, { "epoch": 6.86, "learning_rate": 4.749016966619223e-06, "loss": 0.107, "step": 22295 }, { "epoch": 6.86, "learning_rate": 4.748168895635536e-06, "loss": 0.1391, "step": 22296 }, { "epoch": 6.86, "learning_rate": 4.747320876807372e-06, "loss": 0.084, "step": 22297 }, { "epoch": 6.86, "learning_rate": 4.74647291014315e-06, "loss": 0.085, "step": 22298 }, { "epoch": 6.86, "learning_rate": 4.7456249956513e-06, "loss": 0.0752, "step": 22299 }, { "epoch": 6.86, "learning_rate": 4.744777133340229e-06, "loss": 0.0938, "step": 22300 }, { "epoch": 6.86, "learning_rate": 4.7439293232183695e-06, "loss": 0.1027, "step": 22301 }, { "epoch": 6.86, "learning_rate": 4.743081565294134e-06, "loss": 0.077, "step": 22302 }, { "epoch": 6.86, "learning_rate": 4.7422338595759445e-06, "loss": 0.1149, "step": 22303 }, { "epoch": 6.86, "learning_rate": 4.741386206072219e-06, "loss": 0.1613, "step": 22304 }, { "epoch": 6.86, "learning_rate": 4.740538604791371e-06, "loss": 0.0425, "step": 22305 }, { "epoch": 6.86, "learning_rate": 4.739691055741829e-06, "loss": 0.1051, "step": 22306 }, { "epoch": 6.86, "learning_rate": 4.738843558931999e-06, "loss": 0.0999, "step": 22307 }, { "epoch": 6.86, "learning_rate": 4.737996114370298e-06, "loss": 0.0405, "step": 22308 }, { "epoch": 6.86, "learning_rate": 4.737148722065152e-06, "loss": 0.0899, "step": 22309 }, { "epoch": 6.86, "learning_rate": 4.736301382024963e-06, "loss": 0.0286, "step": 22310 }, { "epoch": 6.86, "learning_rate": 4.735454094258155e-06, "loss": 0.0861, "step": 22311 }, { "epoch": 6.86, "learning_rate": 4.734606858773141e-06, "loss": 0.1203, "step": 22312 }, { "epoch": 6.87, "learning_rate": 4.733759675578332e-06, "loss": 0.1165, "step": 22313 }, { "epoch": 6.87, "learning_rate": 4.732912544682144e-06, "loss": 0.1196, "step": 22314 }, { "epoch": 6.87, "learning_rate": 4.7320654660929854e-06, "loss": 0.1079, "step": 22315 }, { "epoch": 6.87, "learning_rate": 4.731218439819279e-06, "loss": 0.1247, "step": 22316 }, { "epoch": 6.87, "learning_rate": 4.730371465869426e-06, "loss": 0.0219, "step": 22317 }, { "epoch": 6.87, "learning_rate": 4.729524544251837e-06, "loss": 0.0917, "step": 22318 }, { "epoch": 6.87, "learning_rate": 4.72867767497493e-06, "loss": 0.0988, "step": 22319 }, { "epoch": 6.87, "learning_rate": 4.727830858047112e-06, "loss": 0.1386, "step": 22320 }, { "epoch": 6.87, "learning_rate": 4.726984093476792e-06, "loss": 0.1462, "step": 22321 }, { "epoch": 6.87, "learning_rate": 4.726137381272381e-06, "loss": 0.0513, "step": 22322 }, { "epoch": 6.87, "learning_rate": 4.725290721442285e-06, "loss": 0.0563, "step": 22323 }, { "epoch": 6.87, "learning_rate": 4.724444113994915e-06, "loss": 0.1494, "step": 22324 }, { "epoch": 6.87, "learning_rate": 4.7235975589386715e-06, "loss": 0.0827, "step": 22325 }, { "epoch": 6.87, "learning_rate": 4.7227510562819724e-06, "loss": 0.1063, "step": 22326 }, { "epoch": 6.87, "learning_rate": 4.721904606033222e-06, "loss": 0.1389, "step": 22327 }, { "epoch": 6.87, "learning_rate": 4.721058208200816e-06, "loss": 0.0963, "step": 22328 }, { "epoch": 6.87, "learning_rate": 4.72021186279317e-06, "loss": 0.1224, "step": 22329 }, { "epoch": 6.87, "learning_rate": 4.719365569818687e-06, "loss": 0.0944, "step": 22330 }, { "epoch": 6.87, "learning_rate": 4.718519329285771e-06, "loss": 0.0605, "step": 22331 }, { "epoch": 6.87, "learning_rate": 4.717673141202825e-06, "loss": 0.0898, "step": 22332 }, { "epoch": 6.87, "learning_rate": 4.716827005578254e-06, "loss": 0.067, "step": 22333 }, { "epoch": 6.87, "learning_rate": 4.715980922420459e-06, "loss": 0.0864, "step": 22334 }, { "epoch": 6.87, "learning_rate": 4.715134891737841e-06, "loss": 0.0638, "step": 22335 }, { "epoch": 6.87, "learning_rate": 4.714288913538806e-06, "loss": 0.1058, "step": 22336 }, { "epoch": 6.87, "learning_rate": 4.713442987831758e-06, "loss": 0.0636, "step": 22337 }, { "epoch": 6.87, "learning_rate": 4.712597114625086e-06, "loss": 0.1945, "step": 22338 }, { "epoch": 6.87, "learning_rate": 4.711751293927202e-06, "loss": 0.0781, "step": 22339 }, { "epoch": 6.87, "learning_rate": 4.7109055257465e-06, "loss": 0.0839, "step": 22340 }, { "epoch": 6.87, "learning_rate": 4.710059810091382e-06, "loss": 0.0943, "step": 22341 }, { "epoch": 6.87, "learning_rate": 4.709214146970245e-06, "loss": 0.0898, "step": 22342 }, { "epoch": 6.87, "learning_rate": 4.708368536391483e-06, "loss": 0.1264, "step": 22343 }, { "epoch": 6.87, "learning_rate": 4.707522978363508e-06, "loss": 0.142, "step": 22344 }, { "epoch": 6.88, "learning_rate": 4.706677472894699e-06, "loss": 0.0854, "step": 22345 }, { "epoch": 6.88, "learning_rate": 4.705832019993464e-06, "loss": 0.0894, "step": 22346 }, { "epoch": 6.88, "learning_rate": 4.704986619668197e-06, "loss": 0.0616, "step": 22347 }, { "epoch": 6.88, "learning_rate": 4.704141271927293e-06, "loss": 0.0971, "step": 22348 }, { "epoch": 6.88, "learning_rate": 4.703295976779147e-06, "loss": 0.0756, "step": 22349 }, { "epoch": 6.88, "learning_rate": 4.702450734232154e-06, "loss": 0.1064, "step": 22350 }, { "epoch": 6.88, "learning_rate": 4.701605544294706e-06, "loss": 0.1016, "step": 22351 }, { "epoch": 6.88, "learning_rate": 4.700760406975199e-06, "loss": 0.1141, "step": 22352 }, { "epoch": 6.88, "learning_rate": 4.699915322282021e-06, "loss": 0.0877, "step": 22353 }, { "epoch": 6.88, "learning_rate": 4.699070290223576e-06, "loss": 0.132, "step": 22354 }, { "epoch": 6.88, "learning_rate": 4.698225310808242e-06, "loss": 0.129, "step": 22355 }, { "epoch": 6.88, "learning_rate": 4.69738038404442e-06, "loss": 0.0497, "step": 22356 }, { "epoch": 6.88, "learning_rate": 4.696535509940499e-06, "loss": 0.1183, "step": 22357 }, { "epoch": 6.88, "learning_rate": 4.695690688504866e-06, "loss": 0.1063, "step": 22358 }, { "epoch": 6.88, "learning_rate": 4.694845919745914e-06, "loss": 0.1439, "step": 22359 }, { "epoch": 6.88, "learning_rate": 4.694001203672028e-06, "loss": 0.0653, "step": 22360 }, { "epoch": 6.88, "learning_rate": 4.693156540291608e-06, "loss": 0.0858, "step": 22361 }, { "epoch": 6.88, "learning_rate": 4.69231192961303e-06, "loss": 0.0758, "step": 22362 }, { "epoch": 6.88, "learning_rate": 4.691467371644682e-06, "loss": 0.1506, "step": 22363 }, { "epoch": 6.88, "learning_rate": 4.690622866394961e-06, "loss": 0.09, "step": 22364 }, { "epoch": 6.88, "learning_rate": 4.6897784138722465e-06, "loss": 0.1589, "step": 22365 }, { "epoch": 6.88, "learning_rate": 4.688934014084927e-06, "loss": 0.0978, "step": 22366 }, { "epoch": 6.88, "learning_rate": 4.688089667041384e-06, "loss": 0.1086, "step": 22367 }, { "epoch": 6.88, "learning_rate": 4.687245372750014e-06, "loss": 0.0844, "step": 22368 }, { "epoch": 6.88, "learning_rate": 4.686401131219189e-06, "loss": 0.0946, "step": 22369 }, { "epoch": 6.88, "learning_rate": 4.685556942457296e-06, "loss": 0.0911, "step": 22370 }, { "epoch": 6.88, "learning_rate": 4.684712806472727e-06, "loss": 0.0729, "step": 22371 }, { "epoch": 6.88, "learning_rate": 4.68386872327385e-06, "loss": 0.0892, "step": 22372 }, { "epoch": 6.88, "learning_rate": 4.683024692869062e-06, "loss": 0.0856, "step": 22373 }, { "epoch": 6.88, "learning_rate": 4.682180715266736e-06, "loss": 0.0523, "step": 22374 }, { "epoch": 6.88, "learning_rate": 4.681336790475259e-06, "loss": 0.1088, "step": 22375 }, { "epoch": 6.88, "learning_rate": 4.680492918503009e-06, "loss": 0.1314, "step": 22376 }, { "epoch": 6.88, "learning_rate": 4.679649099358363e-06, "loss": 0.0655, "step": 22377 }, { "epoch": 6.89, "learning_rate": 4.678805333049713e-06, "loss": 0.052, "step": 22378 }, { "epoch": 6.89, "learning_rate": 4.677961619585425e-06, "loss": 0.0225, "step": 22379 }, { "epoch": 6.89, "learning_rate": 4.67711795897388e-06, "loss": 0.0835, "step": 22380 }, { "epoch": 6.89, "learning_rate": 4.676274351223463e-06, "loss": 0.0637, "step": 22381 }, { "epoch": 6.89, "learning_rate": 4.675430796342549e-06, "loss": 0.1258, "step": 22382 }, { "epoch": 6.89, "learning_rate": 4.674587294339513e-06, "loss": 0.0861, "step": 22383 }, { "epoch": 6.89, "learning_rate": 4.6737438452227345e-06, "loss": 0.0676, "step": 22384 }, { "epoch": 6.89, "learning_rate": 4.672900449000588e-06, "loss": 0.0848, "step": 22385 }, { "epoch": 6.89, "learning_rate": 4.67205710568145e-06, "loss": 0.0569, "step": 22386 }, { "epoch": 6.89, "learning_rate": 4.671213815273691e-06, "loss": 0.1052, "step": 22387 }, { "epoch": 6.89, "learning_rate": 4.670370577785698e-06, "loss": 0.0483, "step": 22388 }, { "epoch": 6.89, "learning_rate": 4.669527393225833e-06, "loss": 0.0536, "step": 22389 }, { "epoch": 6.89, "learning_rate": 4.668684261602469e-06, "loss": 0.0822, "step": 22390 }, { "epoch": 6.89, "learning_rate": 4.6678411829239885e-06, "loss": 0.1123, "step": 22391 }, { "epoch": 6.89, "learning_rate": 4.66699815719876e-06, "loss": 0.0473, "step": 22392 }, { "epoch": 6.89, "learning_rate": 4.666155184435153e-06, "loss": 0.0945, "step": 22393 }, { "epoch": 6.89, "learning_rate": 4.665312264641542e-06, "loss": 0.1416, "step": 22394 }, { "epoch": 6.89, "learning_rate": 4.664469397826297e-06, "loss": 0.1568, "step": 22395 }, { "epoch": 6.89, "learning_rate": 4.663626583997789e-06, "loss": 0.088, "step": 22396 }, { "epoch": 6.89, "learning_rate": 4.662783823164383e-06, "loss": 0.2069, "step": 22397 }, { "epoch": 6.89, "learning_rate": 4.661941115334455e-06, "loss": 0.077, "step": 22398 }, { "epoch": 6.89, "learning_rate": 4.661098460516376e-06, "loss": 0.0525, "step": 22399 }, { "epoch": 6.89, "learning_rate": 4.660255858718502e-06, "loss": 0.0982, "step": 22400 }, { "epoch": 6.89, "learning_rate": 4.659413309949214e-06, "loss": 0.0684, "step": 22401 }, { "epoch": 6.89, "learning_rate": 4.658570814216872e-06, "loss": 0.1221, "step": 22402 }, { "epoch": 6.89, "learning_rate": 4.657728371529844e-06, "loss": 0.1094, "step": 22403 }, { "epoch": 6.89, "learning_rate": 4.656885981896498e-06, "loss": 0.0613, "step": 22404 }, { "epoch": 6.89, "learning_rate": 4.656043645325194e-06, "loss": 0.1049, "step": 22405 }, { "epoch": 6.89, "learning_rate": 4.6552013618243105e-06, "loss": 0.083, "step": 22406 }, { "epoch": 6.89, "learning_rate": 4.654359131402195e-06, "loss": 0.083, "step": 22407 }, { "epoch": 6.89, "learning_rate": 4.653516954067225e-06, "loss": 0.1225, "step": 22408 }, { "epoch": 6.89, "learning_rate": 4.652674829827762e-06, "loss": 0.1009, "step": 22409 }, { "epoch": 6.9, "learning_rate": 4.651832758692158e-06, "loss": 0.0685, "step": 22410 }, { "epoch": 6.9, "learning_rate": 4.650990740668788e-06, "loss": 0.0655, "step": 22411 }, { "epoch": 6.9, "learning_rate": 4.65014877576601e-06, "loss": 0.1499, "step": 22412 }, { "epoch": 6.9, "learning_rate": 4.6493068639921844e-06, "loss": 0.1004, "step": 22413 }, { "epoch": 6.9, "learning_rate": 4.648465005355675e-06, "loss": 0.1411, "step": 22414 }, { "epoch": 6.9, "learning_rate": 4.647623199864834e-06, "loss": 0.0809, "step": 22415 }, { "epoch": 6.9, "learning_rate": 4.6467814475280355e-06, "loss": 0.1428, "step": 22416 }, { "epoch": 6.9, "learning_rate": 4.645939748353625e-06, "loss": 0.1126, "step": 22417 }, { "epoch": 6.9, "learning_rate": 4.645098102349969e-06, "loss": 0.0942, "step": 22418 }, { "epoch": 6.9, "learning_rate": 4.644256509525426e-06, "loss": 0.0833, "step": 22419 }, { "epoch": 6.9, "learning_rate": 4.64341496988835e-06, "loss": 0.0481, "step": 22420 }, { "epoch": 6.9, "learning_rate": 4.642573483447101e-06, "loss": 0.0769, "step": 22421 }, { "epoch": 6.9, "learning_rate": 4.641732050210032e-06, "loss": 0.1091, "step": 22422 }, { "epoch": 6.9, "learning_rate": 4.640890670185508e-06, "loss": 0.0703, "step": 22423 }, { "epoch": 6.9, "learning_rate": 4.640049343381876e-06, "loss": 0.1162, "step": 22424 }, { "epoch": 6.9, "learning_rate": 4.63920806980749e-06, "loss": 0.0765, "step": 22425 }, { "epoch": 6.9, "learning_rate": 4.638366849470712e-06, "loss": 0.0567, "step": 22426 }, { "epoch": 6.9, "learning_rate": 4.637525682379893e-06, "loss": 0.1269, "step": 22427 }, { "epoch": 6.9, "learning_rate": 4.636684568543386e-06, "loss": 0.1041, "step": 22428 }, { "epoch": 6.9, "learning_rate": 4.635843507969544e-06, "loss": 0.0758, "step": 22429 }, { "epoch": 6.9, "learning_rate": 4.63500250066672e-06, "loss": 0.0793, "step": 22430 }, { "epoch": 6.9, "learning_rate": 4.634161546643265e-06, "loss": 0.1068, "step": 22431 }, { "epoch": 6.9, "learning_rate": 4.6333206459075285e-06, "loss": 0.0925, "step": 22432 }, { "epoch": 6.9, "learning_rate": 4.63247979846787e-06, "loss": 0.1439, "step": 22433 }, { "epoch": 6.9, "learning_rate": 4.631639004332632e-06, "loss": 0.0924, "step": 22434 }, { "epoch": 6.9, "learning_rate": 4.630798263510162e-06, "loss": 0.1178, "step": 22435 }, { "epoch": 6.9, "learning_rate": 4.629957576008817e-06, "loss": 0.0515, "step": 22436 }, { "epoch": 6.9, "learning_rate": 4.629116941836943e-06, "loss": 0.1094, "step": 22437 }, { "epoch": 6.9, "learning_rate": 4.628276361002887e-06, "loss": 0.0908, "step": 22438 }, { "epoch": 6.9, "learning_rate": 4.627435833514998e-06, "loss": 0.1526, "step": 22439 }, { "epoch": 6.9, "learning_rate": 4.626595359381623e-06, "loss": 0.1085, "step": 22440 }, { "epoch": 6.9, "learning_rate": 4.625754938611107e-06, "loss": 0.0931, "step": 22441 }, { "epoch": 6.9, "learning_rate": 4.624914571211795e-06, "loss": 0.1053, "step": 22442 }, { "epoch": 6.91, "learning_rate": 4.624074257192038e-06, "loss": 0.0989, "step": 22443 }, { "epoch": 6.91, "learning_rate": 4.623233996560183e-06, "loss": 0.1338, "step": 22444 }, { "epoch": 6.91, "learning_rate": 4.62239378932456e-06, "loss": 0.1441, "step": 22445 }, { "epoch": 6.91, "learning_rate": 4.62155363549353e-06, "loss": 0.1185, "step": 22446 }, { "epoch": 6.91, "learning_rate": 4.620713535075428e-06, "loss": 0.1193, "step": 22447 }, { "epoch": 6.91, "learning_rate": 4.619873488078597e-06, "loss": 0.0473, "step": 22448 }, { "epoch": 6.91, "learning_rate": 4.619033494511382e-06, "loss": 0.0913, "step": 22449 }, { "epoch": 6.91, "learning_rate": 4.618193554382123e-06, "loss": 0.0881, "step": 22450 }, { "epoch": 6.91, "learning_rate": 4.617353667699163e-06, "loss": 0.0915, "step": 22451 }, { "epoch": 6.91, "learning_rate": 4.616513834470837e-06, "loss": 0.0742, "step": 22452 }, { "epoch": 6.91, "learning_rate": 4.615674054705494e-06, "loss": 0.1365, "step": 22453 }, { "epoch": 6.91, "learning_rate": 4.614834328411474e-06, "loss": 0.0628, "step": 22454 }, { "epoch": 6.91, "learning_rate": 4.613994655597105e-06, "loss": 0.0819, "step": 22455 }, { "epoch": 6.91, "learning_rate": 4.613155036270736e-06, "loss": 0.0918, "step": 22456 }, { "epoch": 6.91, "learning_rate": 4.612315470440704e-06, "loss": 0.0421, "step": 22457 }, { "epoch": 6.91, "learning_rate": 4.611475958115342e-06, "loss": 0.083, "step": 22458 }, { "epoch": 6.91, "learning_rate": 4.610636499302992e-06, "loss": 0.1458, "step": 22459 }, { "epoch": 6.91, "learning_rate": 4.609797094011984e-06, "loss": 0.1037, "step": 22460 }, { "epoch": 6.91, "learning_rate": 4.608957742250667e-06, "loss": 0.0504, "step": 22461 }, { "epoch": 6.91, "learning_rate": 4.6081184440273596e-06, "loss": 0.1081, "step": 22462 }, { "epoch": 6.91, "learning_rate": 4.6072791993504094e-06, "loss": 0.1324, "step": 22463 }, { "epoch": 6.91, "learning_rate": 4.6064400082281466e-06, "loss": 0.0775, "step": 22464 }, { "epoch": 6.91, "learning_rate": 4.605600870668906e-06, "loss": 0.0468, "step": 22465 }, { "epoch": 6.91, "learning_rate": 4.60476178668102e-06, "loss": 0.093, "step": 22466 }, { "epoch": 6.91, "learning_rate": 4.603922756272823e-06, "loss": 0.0772, "step": 22467 }, { "epoch": 6.91, "learning_rate": 4.603083779452645e-06, "loss": 0.0653, "step": 22468 }, { "epoch": 6.91, "learning_rate": 4.602244856228819e-06, "loss": 0.0665, "step": 22469 }, { "epoch": 6.91, "learning_rate": 4.601405986609673e-06, "loss": 0.1193, "step": 22470 }, { "epoch": 6.91, "learning_rate": 4.600567170603548e-06, "loss": 0.0882, "step": 22471 }, { "epoch": 6.91, "learning_rate": 4.59972840821876e-06, "loss": 0.0845, "step": 22472 }, { "epoch": 6.91, "learning_rate": 4.59888969946365e-06, "loss": 0.0661, "step": 22473 }, { "epoch": 6.91, "learning_rate": 4.598051044346542e-06, "loss": 0.0233, "step": 22474 }, { "epoch": 6.92, "learning_rate": 4.597212442875767e-06, "loss": 0.1083, "step": 22475 }, { "epoch": 6.92, "learning_rate": 4.596373895059652e-06, "loss": 0.1244, "step": 22476 }, { "epoch": 6.92, "learning_rate": 4.595535400906519e-06, "loss": 0.0787, "step": 22477 }, { "epoch": 6.92, "learning_rate": 4.594696960424709e-06, "loss": 0.0572, "step": 22478 }, { "epoch": 6.92, "learning_rate": 4.593858573622536e-06, "loss": 0.0868, "step": 22479 }, { "epoch": 6.92, "learning_rate": 4.593020240508326e-06, "loss": 0.2819, "step": 22480 }, { "epoch": 6.92, "learning_rate": 4.592181961090411e-06, "loss": 0.093, "step": 22481 }, { "epoch": 6.92, "learning_rate": 4.5913437353771155e-06, "loss": 0.1095, "step": 22482 }, { "epoch": 6.92, "learning_rate": 4.590505563376759e-06, "loss": 0.1192, "step": 22483 }, { "epoch": 6.92, "learning_rate": 4.589667445097669e-06, "loss": 0.0705, "step": 22484 }, { "epoch": 6.92, "learning_rate": 4.588829380548168e-06, "loss": 0.079, "step": 22485 }, { "epoch": 6.92, "learning_rate": 4.587991369736578e-06, "loss": 0.0954, "step": 22486 }, { "epoch": 6.92, "learning_rate": 4.587153412671217e-06, "loss": 0.1427, "step": 22487 }, { "epoch": 6.92, "learning_rate": 4.586315509360419e-06, "loss": 0.1034, "step": 22488 }, { "epoch": 6.92, "learning_rate": 4.585477659812494e-06, "loss": 0.1101, "step": 22489 }, { "epoch": 6.92, "learning_rate": 4.5846398640357625e-06, "loss": 0.0442, "step": 22490 }, { "epoch": 6.92, "learning_rate": 4.583802122038551e-06, "loss": 0.1065, "step": 22491 }, { "epoch": 6.92, "learning_rate": 4.582964433829177e-06, "loss": 0.0906, "step": 22492 }, { "epoch": 6.92, "learning_rate": 4.582126799415958e-06, "loss": 0.0461, "step": 22493 }, { "epoch": 6.92, "learning_rate": 4.581289218807213e-06, "loss": 0.1094, "step": 22494 }, { "epoch": 6.92, "learning_rate": 4.5804516920112606e-06, "loss": 0.1273, "step": 22495 }, { "epoch": 6.92, "learning_rate": 4.5796142190364175e-06, "loss": 0.0873, "step": 22496 }, { "epoch": 6.92, "learning_rate": 4.5787767998909985e-06, "loss": 0.1375, "step": 22497 }, { "epoch": 6.92, "learning_rate": 4.577939434583324e-06, "loss": 0.074, "step": 22498 }, { "epoch": 6.92, "learning_rate": 4.577102123121713e-06, "loss": 0.0522, "step": 22499 }, { "epoch": 6.92, "learning_rate": 4.576264865514467e-06, "loss": 0.1151, "step": 22500 }, { "epoch": 6.92, "learning_rate": 4.5754276617699145e-06, "loss": 0.0934, "step": 22501 }, { "epoch": 6.92, "learning_rate": 4.574590511896364e-06, "loss": 0.0705, "step": 22502 }, { "epoch": 6.92, "learning_rate": 4.57375341590213e-06, "loss": 0.0516, "step": 22503 }, { "epoch": 6.92, "learning_rate": 4.5729163737955264e-06, "loss": 0.092, "step": 22504 }, { "epoch": 6.92, "learning_rate": 4.57207938558486e-06, "loss": 0.0925, "step": 22505 }, { "epoch": 6.92, "learning_rate": 4.571242451278456e-06, "loss": 0.1512, "step": 22506 }, { "epoch": 6.92, "learning_rate": 4.57040557088461e-06, "loss": 0.0806, "step": 22507 }, { "epoch": 6.93, "learning_rate": 4.569568744411644e-06, "loss": 0.0672, "step": 22508 }, { "epoch": 6.93, "learning_rate": 4.56873197186787e-06, "loss": 0.0538, "step": 22509 }, { "epoch": 6.93, "learning_rate": 4.567895253261585e-06, "loss": 0.0852, "step": 22510 }, { "epoch": 6.93, "learning_rate": 4.567058588601111e-06, "loss": 0.067, "step": 22511 }, { "epoch": 6.93, "learning_rate": 4.566221977894751e-06, "loss": 0.0831, "step": 22512 }, { "epoch": 6.93, "learning_rate": 4.565385421150817e-06, "loss": 0.0623, "step": 22513 }, { "epoch": 6.93, "learning_rate": 4.564548918377612e-06, "loss": 0.1166, "step": 22514 }, { "epoch": 6.93, "learning_rate": 4.563712469583443e-06, "loss": 0.0686, "step": 22515 }, { "epoch": 6.93, "learning_rate": 4.562876074776627e-06, "loss": 0.1128, "step": 22516 }, { "epoch": 6.93, "learning_rate": 4.5620397339654555e-06, "loss": 0.0848, "step": 22517 }, { "epoch": 6.93, "learning_rate": 4.5612034471582435e-06, "loss": 0.0587, "step": 22518 }, { "epoch": 6.93, "learning_rate": 4.560367214363295e-06, "loss": 0.1415, "step": 22519 }, { "epoch": 6.93, "learning_rate": 4.559531035588913e-06, "loss": 0.0658, "step": 22520 }, { "epoch": 6.93, "learning_rate": 4.558694910843402e-06, "loss": 0.1416, "step": 22521 }, { "epoch": 6.93, "learning_rate": 4.557858840135062e-06, "loss": 0.0924, "step": 22522 }, { "epoch": 6.93, "learning_rate": 4.557022823472206e-06, "loss": 0.0801, "step": 22523 }, { "epoch": 6.93, "learning_rate": 4.5561868608631264e-06, "loss": 0.1149, "step": 22524 }, { "epoch": 6.93, "learning_rate": 4.555350952316124e-06, "loss": 0.0872, "step": 22525 }, { "epoch": 6.93, "learning_rate": 4.554515097839511e-06, "loss": 0.0594, "step": 22526 }, { "epoch": 6.93, "learning_rate": 4.5536792974415785e-06, "loss": 0.0503, "step": 22527 }, { "epoch": 6.93, "learning_rate": 4.552843551130631e-06, "loss": 0.128, "step": 22528 }, { "epoch": 6.93, "learning_rate": 4.5520078589149685e-06, "loss": 0.0945, "step": 22529 }, { "epoch": 6.93, "learning_rate": 4.5511722208028875e-06, "loss": 0.1379, "step": 22530 }, { "epoch": 6.93, "learning_rate": 4.550336636802689e-06, "loss": 0.0826, "step": 22531 }, { "epoch": 6.93, "learning_rate": 4.549501106922667e-06, "loss": 0.0643, "step": 22532 }, { "epoch": 6.93, "learning_rate": 4.548665631171128e-06, "loss": 0.1061, "step": 22533 }, { "epoch": 6.93, "learning_rate": 4.5478302095563606e-06, "loss": 0.0972, "step": 22534 }, { "epoch": 6.93, "learning_rate": 4.546994842086662e-06, "loss": 0.109, "step": 22535 }, { "epoch": 6.93, "learning_rate": 4.546159528770332e-06, "loss": 0.1536, "step": 22536 }, { "epoch": 6.93, "learning_rate": 4.545324269615664e-06, "loss": 0.1251, "step": 22537 }, { "epoch": 6.93, "learning_rate": 4.544489064630954e-06, "loss": 0.0692, "step": 22538 }, { "epoch": 6.93, "learning_rate": 4.543653913824496e-06, "loss": 0.0676, "step": 22539 }, { "epoch": 6.94, "learning_rate": 4.542818817204583e-06, "loss": 0.074, "step": 22540 }, { "epoch": 6.94, "learning_rate": 4.541983774779508e-06, "loss": 0.0497, "step": 22541 }, { "epoch": 6.94, "learning_rate": 4.541148786557562e-06, "loss": 0.0958, "step": 22542 }, { "epoch": 6.94, "learning_rate": 4.540313852547042e-06, "loss": 0.0822, "step": 22543 }, { "epoch": 6.94, "learning_rate": 4.539478972756242e-06, "loss": 0.1149, "step": 22544 }, { "epoch": 6.94, "learning_rate": 4.53864414719344e-06, "loss": 0.1153, "step": 22545 }, { "epoch": 6.94, "learning_rate": 4.537809375866938e-06, "loss": 0.0839, "step": 22546 }, { "epoch": 6.94, "learning_rate": 4.536974658785025e-06, "loss": 0.0723, "step": 22547 }, { "epoch": 6.94, "learning_rate": 4.536139995955987e-06, "loss": 0.1271, "step": 22548 }, { "epoch": 6.94, "learning_rate": 4.5353053873881145e-06, "loss": 0.0716, "step": 22549 }, { "epoch": 6.94, "learning_rate": 4.534470833089697e-06, "loss": 0.0976, "step": 22550 }, { "epoch": 6.94, "learning_rate": 4.53363633306902e-06, "loss": 0.0457, "step": 22551 }, { "epoch": 6.94, "learning_rate": 4.53280188733437e-06, "loss": 0.0517, "step": 22552 }, { "epoch": 6.94, "learning_rate": 4.5319674958940385e-06, "loss": 0.1239, "step": 22553 }, { "epoch": 6.94, "learning_rate": 4.531133158756313e-06, "loss": 0.0972, "step": 22554 }, { "epoch": 6.94, "learning_rate": 4.530298875929469e-06, "loss": 0.1098, "step": 22555 }, { "epoch": 6.94, "learning_rate": 4.529464647421802e-06, "loss": 0.1, "step": 22556 }, { "epoch": 6.94, "learning_rate": 4.528630473241591e-06, "loss": 0.0852, "step": 22557 }, { "epoch": 6.94, "learning_rate": 4.527796353397124e-06, "loss": 0.1236, "step": 22558 }, { "epoch": 6.94, "learning_rate": 4.526962287896681e-06, "loss": 0.0622, "step": 22559 }, { "epoch": 6.94, "learning_rate": 4.526128276748544e-06, "loss": 0.0995, "step": 22560 }, { "epoch": 6.94, "learning_rate": 4.5252943199610046e-06, "loss": 0.1264, "step": 22561 }, { "epoch": 6.94, "learning_rate": 4.524460417542331e-06, "loss": 0.16, "step": 22562 }, { "epoch": 6.94, "learning_rate": 4.523626569500816e-06, "loss": 0.0717, "step": 22563 }, { "epoch": 6.94, "learning_rate": 4.5227927758447365e-06, "loss": 0.1417, "step": 22564 }, { "epoch": 6.94, "learning_rate": 4.521959036582372e-06, "loss": 0.1285, "step": 22565 }, { "epoch": 6.94, "learning_rate": 4.521125351722003e-06, "loss": 0.0606, "step": 22566 }, { "epoch": 6.94, "learning_rate": 4.520291721271909e-06, "loss": 0.0766, "step": 22567 }, { "epoch": 6.94, "learning_rate": 4.519458145240368e-06, "loss": 0.0947, "step": 22568 }, { "epoch": 6.94, "learning_rate": 4.518624623635655e-06, "loss": 0.1549, "step": 22569 }, { "epoch": 6.94, "learning_rate": 4.517791156466057e-06, "loss": 0.0505, "step": 22570 }, { "epoch": 6.94, "learning_rate": 4.516957743739847e-06, "loss": 0.0805, "step": 22571 }, { "epoch": 6.94, "learning_rate": 4.516124385465293e-06, "loss": 0.0817, "step": 22572 }, { "epoch": 6.95, "learning_rate": 4.515291081650681e-06, "loss": 0.0657, "step": 22573 }, { "epoch": 6.95, "learning_rate": 4.514457832304284e-06, "loss": 0.1297, "step": 22574 }, { "epoch": 6.95, "learning_rate": 4.513624637434377e-06, "loss": 0.1133, "step": 22575 }, { "epoch": 6.95, "learning_rate": 4.512791497049232e-06, "loss": 0.0814, "step": 22576 }, { "epoch": 6.95, "learning_rate": 4.511958411157123e-06, "loss": 0.0726, "step": 22577 }, { "epoch": 6.95, "learning_rate": 4.511125379766332e-06, "loss": 0.0943, "step": 22578 }, { "epoch": 6.95, "learning_rate": 4.5102924028851156e-06, "loss": 0.0786, "step": 22579 }, { "epoch": 6.95, "learning_rate": 4.509459480521761e-06, "loss": 0.1749, "step": 22580 }, { "epoch": 6.95, "learning_rate": 4.508626612684532e-06, "loss": 0.0873, "step": 22581 }, { "epoch": 6.95, "learning_rate": 4.507793799381702e-06, "loss": 0.1299, "step": 22582 }, { "epoch": 6.95, "learning_rate": 4.506961040621542e-06, "loss": 0.1656, "step": 22583 }, { "epoch": 6.95, "learning_rate": 4.506128336412317e-06, "loss": 0.0582, "step": 22584 }, { "epoch": 6.95, "learning_rate": 4.50529568676231e-06, "loss": 0.1431, "step": 22585 }, { "epoch": 6.95, "learning_rate": 4.504463091679776e-06, "loss": 0.071, "step": 22586 }, { "epoch": 6.95, "learning_rate": 4.503630551172985e-06, "loss": 0.0894, "step": 22587 }, { "epoch": 6.95, "learning_rate": 4.502798065250215e-06, "loss": 0.0912, "step": 22588 }, { "epoch": 6.95, "learning_rate": 4.501965633919719e-06, "loss": 0.0869, "step": 22589 }, { "epoch": 6.95, "learning_rate": 4.501133257189776e-06, "loss": 0.0529, "step": 22590 }, { "epoch": 6.95, "learning_rate": 4.500300935068647e-06, "loss": 0.121, "step": 22591 }, { "epoch": 6.95, "learning_rate": 4.499468667564599e-06, "loss": 0.1409, "step": 22592 }, { "epoch": 6.95, "learning_rate": 4.498636454685897e-06, "loss": 0.0561, "step": 22593 }, { "epoch": 6.95, "learning_rate": 4.4978042964408e-06, "loss": 0.1062, "step": 22594 }, { "epoch": 6.95, "learning_rate": 4.496972192837585e-06, "loss": 0.0739, "step": 22595 }, { "epoch": 6.95, "learning_rate": 4.496140143884505e-06, "loss": 0.074, "step": 22596 }, { "epoch": 6.95, "learning_rate": 4.495308149589822e-06, "loss": 0.0897, "step": 22597 }, { "epoch": 6.95, "learning_rate": 4.494476209961805e-06, "loss": 0.0773, "step": 22598 }, { "epoch": 6.95, "learning_rate": 4.4936443250087145e-06, "loss": 0.0964, "step": 22599 }, { "epoch": 6.95, "learning_rate": 4.492812494738811e-06, "loss": 0.1082, "step": 22600 }, { "epoch": 6.95, "learning_rate": 4.491980719160354e-06, "loss": 0.1227, "step": 22601 }, { "epoch": 6.95, "learning_rate": 4.491148998281605e-06, "loss": 0.1421, "step": 22602 }, { "epoch": 6.95, "learning_rate": 4.490317332110824e-06, "loss": 0.1263, "step": 22603 }, { "epoch": 6.95, "learning_rate": 4.489485720656266e-06, "loss": 0.0758, "step": 22604 }, { "epoch": 6.96, "learning_rate": 4.488654163926196e-06, "loss": 0.108, "step": 22605 }, { "epoch": 6.96, "learning_rate": 4.4878226619288745e-06, "loss": 0.0844, "step": 22606 }, { "epoch": 6.96, "learning_rate": 4.486991214672546e-06, "loss": 0.1085, "step": 22607 }, { "epoch": 6.96, "learning_rate": 4.48615982216548e-06, "loss": 0.1133, "step": 22608 }, { "epoch": 6.96, "learning_rate": 4.485328484415928e-06, "loss": 0.0892, "step": 22609 }, { "epoch": 6.96, "learning_rate": 4.484497201432147e-06, "loss": 0.0593, "step": 22610 }, { "epoch": 6.96, "learning_rate": 4.483665973222392e-06, "loss": 0.0418, "step": 22611 }, { "epoch": 6.96, "learning_rate": 4.482834799794919e-06, "loss": 0.0673, "step": 22612 }, { "epoch": 6.96, "learning_rate": 4.482003681157979e-06, "loss": 0.0772, "step": 22613 }, { "epoch": 6.96, "learning_rate": 4.481172617319826e-06, "loss": 0.0727, "step": 22614 }, { "epoch": 6.96, "learning_rate": 4.480341608288718e-06, "loss": 0.077, "step": 22615 }, { "epoch": 6.96, "learning_rate": 4.479510654072909e-06, "loss": 0.0812, "step": 22616 }, { "epoch": 6.96, "learning_rate": 4.478679754680639e-06, "loss": 0.0973, "step": 22617 }, { "epoch": 6.96, "learning_rate": 4.477848910120173e-06, "loss": 0.0907, "step": 22618 }, { "epoch": 6.96, "learning_rate": 4.477018120399756e-06, "loss": 0.1033, "step": 22619 }, { "epoch": 6.96, "learning_rate": 4.476187385527639e-06, "loss": 0.0993, "step": 22620 }, { "epoch": 6.96, "learning_rate": 4.475356705512071e-06, "loss": 0.0712, "step": 22621 }, { "epoch": 6.96, "learning_rate": 4.4745260803613e-06, "loss": 0.1057, "step": 22622 }, { "epoch": 6.96, "learning_rate": 4.4736955100835844e-06, "loss": 0.0882, "step": 22623 }, { "epoch": 6.96, "learning_rate": 4.472864994687159e-06, "loss": 0.0617, "step": 22624 }, { "epoch": 6.96, "learning_rate": 4.472034534180281e-06, "loss": 0.0876, "step": 22625 }, { "epoch": 6.96, "learning_rate": 4.471204128571199e-06, "loss": 0.1008, "step": 22626 }, { "epoch": 6.96, "learning_rate": 4.4703737778681474e-06, "loss": 0.1171, "step": 22627 }, { "epoch": 6.96, "learning_rate": 4.469543482079385e-06, "loss": 0.0852, "step": 22628 }, { "epoch": 6.96, "learning_rate": 4.468713241213152e-06, "loss": 0.0853, "step": 22629 }, { "epoch": 6.96, "learning_rate": 4.467883055277696e-06, "loss": 0.0356, "step": 22630 }, { "epoch": 6.96, "learning_rate": 4.467052924281258e-06, "loss": 0.122, "step": 22631 }, { "epoch": 6.96, "learning_rate": 4.46622284823208e-06, "loss": 0.0429, "step": 22632 }, { "epoch": 6.96, "learning_rate": 4.465392827138417e-06, "loss": 0.0698, "step": 22633 }, { "epoch": 6.96, "learning_rate": 4.464562861008497e-06, "loss": 0.0932, "step": 22634 }, { "epoch": 6.96, "learning_rate": 4.463732949850574e-06, "loss": 0.0916, "step": 22635 }, { "epoch": 6.96, "learning_rate": 4.462903093672883e-06, "loss": 0.0953, "step": 22636 }, { "epoch": 6.96, "learning_rate": 4.462073292483669e-06, "loss": 0.1499, "step": 22637 }, { "epoch": 6.97, "learning_rate": 4.461243546291171e-06, "loss": 0.1926, "step": 22638 }, { "epoch": 6.97, "learning_rate": 4.460413855103625e-06, "loss": 0.1124, "step": 22639 }, { "epoch": 6.97, "learning_rate": 4.459584218929283e-06, "loss": 0.106, "step": 22640 }, { "epoch": 6.97, "learning_rate": 4.458754637776371e-06, "loss": 0.1126, "step": 22641 }, { "epoch": 6.97, "learning_rate": 4.45792511165313e-06, "loss": 0.0909, "step": 22642 }, { "epoch": 6.97, "learning_rate": 4.457095640567804e-06, "loss": 0.0718, "step": 22643 }, { "epoch": 6.97, "learning_rate": 4.456266224528626e-06, "loss": 0.1337, "step": 22644 }, { "epoch": 6.97, "learning_rate": 4.455436863543834e-06, "loss": 0.103, "step": 22645 }, { "epoch": 6.97, "learning_rate": 4.454607557621663e-06, "loss": 0.0603, "step": 22646 }, { "epoch": 6.97, "learning_rate": 4.453778306770351e-06, "loss": 0.1113, "step": 22647 }, { "epoch": 6.97, "learning_rate": 4.4529491109981316e-06, "loss": 0.0819, "step": 22648 }, { "epoch": 6.97, "learning_rate": 4.452119970313236e-06, "loss": 0.0842, "step": 22649 }, { "epoch": 6.97, "learning_rate": 4.45129088472391e-06, "loss": 0.0964, "step": 22650 }, { "epoch": 6.97, "learning_rate": 4.4504618542383745e-06, "loss": 0.0523, "step": 22651 }, { "epoch": 6.97, "learning_rate": 4.449632878864866e-06, "loss": 0.0871, "step": 22652 }, { "epoch": 6.97, "learning_rate": 4.4488039586116194e-06, "loss": 0.1482, "step": 22653 }, { "epoch": 6.97, "learning_rate": 4.447975093486867e-06, "loss": 0.1209, "step": 22654 }, { "epoch": 6.97, "learning_rate": 4.4471462834988385e-06, "loss": 0.108, "step": 22655 }, { "epoch": 6.97, "learning_rate": 4.4463175286557654e-06, "loss": 0.0513, "step": 22656 }, { "epoch": 6.97, "learning_rate": 4.445488828965878e-06, "loss": 0.1031, "step": 22657 }, { "epoch": 6.97, "learning_rate": 4.444660184437405e-06, "loss": 0.0644, "step": 22658 }, { "epoch": 6.97, "learning_rate": 4.4438315950785725e-06, "loss": 0.0682, "step": 22659 }, { "epoch": 6.97, "learning_rate": 4.443003060897617e-06, "loss": 0.0708, "step": 22660 }, { "epoch": 6.97, "learning_rate": 4.442174581902767e-06, "loss": 0.0839, "step": 22661 }, { "epoch": 6.97, "learning_rate": 4.441346158102238e-06, "loss": 0.0892, "step": 22662 }, { "epoch": 6.97, "learning_rate": 4.440517789504268e-06, "loss": 0.0835, "step": 22663 }, { "epoch": 6.97, "learning_rate": 4.43968947611708e-06, "loss": 0.1313, "step": 22664 }, { "epoch": 6.97, "learning_rate": 4.4388612179489e-06, "loss": 0.0942, "step": 22665 }, { "epoch": 6.97, "learning_rate": 4.438033015007953e-06, "loss": 0.0759, "step": 22666 }, { "epoch": 6.97, "learning_rate": 4.437204867302465e-06, "loss": 0.0609, "step": 22667 }, { "epoch": 6.97, "learning_rate": 4.436376774840658e-06, "loss": 0.1399, "step": 22668 }, { "epoch": 6.97, "learning_rate": 4.435548737630756e-06, "loss": 0.0681, "step": 22669 }, { "epoch": 6.98, "learning_rate": 4.434720755680986e-06, "loss": 0.0755, "step": 22670 }, { "epoch": 6.98, "learning_rate": 4.433892828999571e-06, "loss": 0.0693, "step": 22671 }, { "epoch": 6.98, "learning_rate": 4.433064957594723e-06, "loss": 0.1091, "step": 22672 }, { "epoch": 6.98, "learning_rate": 4.432237141474674e-06, "loss": 0.0494, "step": 22673 }, { "epoch": 6.98, "learning_rate": 4.4314093806476424e-06, "loss": 0.1096, "step": 22674 }, { "epoch": 6.98, "learning_rate": 4.4305816751218475e-06, "loss": 0.0678, "step": 22675 }, { "epoch": 6.98, "learning_rate": 4.429754024905509e-06, "loss": 0.075, "step": 22676 }, { "epoch": 6.98, "learning_rate": 4.428926430006844e-06, "loss": 0.1375, "step": 22677 }, { "epoch": 6.98, "learning_rate": 4.428098890434081e-06, "loss": 0.0672, "step": 22678 }, { "epoch": 6.98, "learning_rate": 4.427271406195424e-06, "loss": 0.1122, "step": 22679 }, { "epoch": 6.98, "learning_rate": 4.426443977299101e-06, "loss": 0.1027, "step": 22680 }, { "epoch": 6.98, "learning_rate": 4.425616603753326e-06, "loss": 0.0592, "step": 22681 }, { "epoch": 6.98, "learning_rate": 4.4247892855663164e-06, "loss": 0.0937, "step": 22682 }, { "epoch": 6.98, "learning_rate": 4.423962022746288e-06, "loss": 0.0832, "step": 22683 }, { "epoch": 6.98, "learning_rate": 4.423134815301454e-06, "loss": 0.058, "step": 22684 }, { "epoch": 6.98, "learning_rate": 4.422307663240032e-06, "loss": 0.0877, "step": 22685 }, { "epoch": 6.98, "learning_rate": 4.421480566570235e-06, "loss": 0.0608, "step": 22686 }, { "epoch": 6.98, "learning_rate": 4.420653525300274e-06, "loss": 0.0752, "step": 22687 }, { "epoch": 6.98, "learning_rate": 4.419826539438373e-06, "loss": 0.0659, "step": 22688 }, { "epoch": 6.98, "learning_rate": 4.41899960899273e-06, "loss": 0.0741, "step": 22689 }, { "epoch": 6.98, "learning_rate": 4.4181727339715675e-06, "loss": 0.1009, "step": 22690 }, { "epoch": 6.98, "learning_rate": 4.417345914383096e-06, "loss": 0.0977, "step": 22691 }, { "epoch": 6.98, "learning_rate": 4.416519150235522e-06, "loss": 0.0655, "step": 22692 }, { "epoch": 6.98, "learning_rate": 4.41569244153706e-06, "loss": 0.0664, "step": 22693 }, { "epoch": 6.98, "learning_rate": 4.414865788295916e-06, "loss": 0.112, "step": 22694 }, { "epoch": 6.98, "learning_rate": 4.414039190520308e-06, "loss": 0.0547, "step": 22695 }, { "epoch": 6.98, "learning_rate": 4.413212648218436e-06, "loss": 0.0925, "step": 22696 }, { "epoch": 6.98, "learning_rate": 4.412386161398507e-06, "loss": 0.0855, "step": 22697 }, { "epoch": 6.98, "learning_rate": 4.411559730068737e-06, "loss": 0.058, "step": 22698 }, { "epoch": 6.98, "learning_rate": 4.41073335423733e-06, "loss": 0.0846, "step": 22699 }, { "epoch": 6.98, "learning_rate": 4.409907033912492e-06, "loss": 0.0583, "step": 22700 }, { "epoch": 6.98, "learning_rate": 4.409080769102428e-06, "loss": 0.0604, "step": 22701 }, { "epoch": 6.98, "learning_rate": 4.408254559815346e-06, "loss": 0.1151, "step": 22702 }, { "epoch": 6.99, "learning_rate": 4.407428406059449e-06, "loss": 0.1171, "step": 22703 }, { "epoch": 6.99, "learning_rate": 4.4066023078429395e-06, "loss": 0.1138, "step": 22704 }, { "epoch": 6.99, "learning_rate": 4.405776265174031e-06, "loss": 0.0802, "step": 22705 }, { "epoch": 6.99, "learning_rate": 4.404950278060916e-06, "loss": 0.0799, "step": 22706 }, { "epoch": 6.99, "learning_rate": 4.4041243465117975e-06, "loss": 0.0685, "step": 22707 }, { "epoch": 6.99, "learning_rate": 4.403298470534885e-06, "loss": 0.0373, "step": 22708 }, { "epoch": 6.99, "learning_rate": 4.4024726501383775e-06, "loss": 0.0756, "step": 22709 }, { "epoch": 6.99, "learning_rate": 4.401646885330475e-06, "loss": 0.0995, "step": 22710 }, { "epoch": 6.99, "learning_rate": 4.400821176119378e-06, "loss": 0.1638, "step": 22711 }, { "epoch": 6.99, "learning_rate": 4.399995522513288e-06, "loss": 0.0994, "step": 22712 }, { "epoch": 6.99, "learning_rate": 4.399169924520403e-06, "loss": 0.0993, "step": 22713 }, { "epoch": 6.99, "learning_rate": 4.398344382148919e-06, "loss": 0.0989, "step": 22714 }, { "epoch": 6.99, "learning_rate": 4.397518895407043e-06, "loss": 0.0959, "step": 22715 }, { "epoch": 6.99, "learning_rate": 4.396693464302969e-06, "loss": 0.0867, "step": 22716 }, { "epoch": 6.99, "learning_rate": 4.3958680888448866e-06, "loss": 0.1025, "step": 22717 }, { "epoch": 6.99, "learning_rate": 4.395042769041004e-06, "loss": 0.1212, "step": 22718 }, { "epoch": 6.99, "learning_rate": 4.39421750489951e-06, "loss": 0.0713, "step": 22719 }, { "epoch": 6.99, "learning_rate": 4.393392296428605e-06, "loss": 0.1174, "step": 22720 }, { "epoch": 6.99, "learning_rate": 4.39256714363648e-06, "loss": 0.1328, "step": 22721 }, { "epoch": 6.99, "learning_rate": 4.391742046531329e-06, "loss": 0.0815, "step": 22722 }, { "epoch": 6.99, "learning_rate": 4.390917005121353e-06, "loss": 0.1321, "step": 22723 }, { "epoch": 6.99, "learning_rate": 4.390092019414736e-06, "loss": 0.0813, "step": 22724 }, { "epoch": 6.99, "learning_rate": 4.389267089419678e-06, "loss": 0.0682, "step": 22725 }, { "epoch": 6.99, "learning_rate": 4.388442215144373e-06, "loss": 0.1118, "step": 22726 }, { "epoch": 6.99, "learning_rate": 4.387617396597e-06, "loss": 0.0554, "step": 22727 }, { "epoch": 6.99, "learning_rate": 4.386792633785764e-06, "loss": 0.1087, "step": 22728 }, { "epoch": 6.99, "learning_rate": 4.38596792671885e-06, "loss": 0.1032, "step": 22729 }, { "epoch": 6.99, "learning_rate": 4.385143275404447e-06, "loss": 0.1296, "step": 22730 }, { "epoch": 6.99, "learning_rate": 4.384318679850748e-06, "loss": 0.0886, "step": 22731 }, { "epoch": 6.99, "learning_rate": 4.3834941400659355e-06, "loss": 0.1154, "step": 22732 }, { "epoch": 6.99, "learning_rate": 4.38266965605821e-06, "loss": 0.0653, "step": 22733 }, { "epoch": 6.99, "learning_rate": 4.3818452278357445e-06, "loss": 0.1147, "step": 22734 }, { "epoch": 7.0, "learning_rate": 4.381020855406738e-06, "loss": 0.0758, "step": 22735 }, { "epoch": 7.0, "learning_rate": 4.380196538779374e-06, "loss": 0.0698, "step": 22736 }, { "epoch": 7.0, "learning_rate": 4.379372277961836e-06, "loss": 0.0666, "step": 22737 }, { "epoch": 7.0, "learning_rate": 4.378548072962311e-06, "loss": 0.0734, "step": 22738 }, { "epoch": 7.0, "learning_rate": 4.377723923788982e-06, "loss": 0.1063, "step": 22739 }, { "epoch": 7.0, "learning_rate": 4.376899830450042e-06, "loss": 0.0755, "step": 22740 }, { "epoch": 7.0, "learning_rate": 4.3760757929536665e-06, "loss": 0.0707, "step": 22741 }, { "epoch": 7.0, "learning_rate": 4.375251811308038e-06, "loss": 0.1058, "step": 22742 }, { "epoch": 7.0, "learning_rate": 4.3744278855213465e-06, "loss": 0.0887, "step": 22743 }, { "epoch": 7.0, "learning_rate": 4.37360401560177e-06, "loss": 0.055, "step": 22744 }, { "epoch": 7.0, "learning_rate": 4.372780201557491e-06, "loss": 0.1383, "step": 22745 }, { "epoch": 7.0, "learning_rate": 4.371956443396692e-06, "loss": 0.1133, "step": 22746 }, { "epoch": 7.0, "learning_rate": 4.371132741127553e-06, "loss": 0.0767, "step": 22747 }, { "epoch": 7.0, "learning_rate": 4.370309094758252e-06, "loss": 0.0826, "step": 22748 }, { "epoch": 7.0, "learning_rate": 4.369485504296967e-06, "loss": 0.1058, "step": 22749 }, { "epoch": 7.0, "learning_rate": 4.368661969751887e-06, "loss": 0.0653, "step": 22750 }, { "epoch": 7.0, "learning_rate": 4.367838491131181e-06, "loss": 0.0992, "step": 22751 }, { "epoch": 7.0, "learning_rate": 4.367015068443026e-06, "loss": 0.0922, "step": 22752 }, { "epoch": 7.0, "learning_rate": 4.3661917016956056e-06, "loss": 0.0583, "step": 22753 }, { "epoch": 7.0, "learning_rate": 4.365368390897095e-06, "loss": 0.0507, "step": 22754 }, { "epoch": 7.0, "learning_rate": 4.3645451360556676e-06, "loss": 0.0499, "step": 22755 }, { "epoch": 7.0, "learning_rate": 4.363721937179501e-06, "loss": 0.1142, "step": 22756 }, { "epoch": 7.0, "learning_rate": 4.362898794276772e-06, "loss": 0.0583, "step": 22757 }, { "epoch": 7.0, "learning_rate": 4.3620757073556516e-06, "loss": 0.0247, "step": 22758 }, { "epoch": 7.0, "learning_rate": 4.361252676424313e-06, "loss": 0.0483, "step": 22759 }, { "epoch": 7.0, "learning_rate": 4.360429701490935e-06, "loss": 0.0408, "step": 22760 }, { "epoch": 7.0, "learning_rate": 4.359606782563688e-06, "loss": 0.0307, "step": 22761 }, { "epoch": 7.0, "learning_rate": 4.358783919650743e-06, "loss": 0.0636, "step": 22762 }, { "epoch": 7.0, "learning_rate": 4.357961112760274e-06, "loss": 0.0604, "step": 22763 }, { "epoch": 7.0, "learning_rate": 4.35713836190045e-06, "loss": 0.0351, "step": 22764 }, { "epoch": 7.0, "learning_rate": 4.3563156670794435e-06, "loss": 0.0305, "step": 22765 }, { "epoch": 7.0, "learning_rate": 4.3554930283054195e-06, "loss": 0.0488, "step": 22766 }, { "epoch": 7.0, "learning_rate": 4.3546704455865596e-06, "loss": 0.0331, "step": 22767 }, { "epoch": 7.01, "learning_rate": 4.3538479189310215e-06, "loss": 0.0668, "step": 22768 }, { "epoch": 7.01, "learning_rate": 4.353025448346974e-06, "loss": 0.0784, "step": 22769 }, { "epoch": 7.01, "learning_rate": 4.352203033842591e-06, "loss": 0.0415, "step": 22770 }, { "epoch": 7.01, "learning_rate": 4.351380675426037e-06, "loss": 0.0332, "step": 22771 }, { "epoch": 7.01, "learning_rate": 4.350558373105479e-06, "loss": 0.0608, "step": 22772 }, { "epoch": 7.01, "learning_rate": 4.349736126889084e-06, "loss": 0.0435, "step": 22773 }, { "epoch": 7.01, "learning_rate": 4.3489139367850155e-06, "loss": 0.0469, "step": 22774 }, { "epoch": 7.01, "learning_rate": 4.348091802801439e-06, "loss": 0.053, "step": 22775 }, { "epoch": 7.01, "learning_rate": 4.3472697249465175e-06, "loss": 0.0628, "step": 22776 }, { "epoch": 7.01, "learning_rate": 4.3464477032284214e-06, "loss": 0.048, "step": 22777 }, { "epoch": 7.01, "learning_rate": 4.345625737655314e-06, "loss": 0.0526, "step": 22778 }, { "epoch": 7.01, "learning_rate": 4.344803828235346e-06, "loss": 0.0402, "step": 22779 }, { "epoch": 7.01, "learning_rate": 4.343981974976691e-06, "loss": 0.0286, "step": 22780 }, { "epoch": 7.01, "learning_rate": 4.3431601778875085e-06, "loss": 0.0277, "step": 22781 }, { "epoch": 7.01, "learning_rate": 4.342338436975959e-06, "loss": 0.0644, "step": 22782 }, { "epoch": 7.01, "learning_rate": 4.341516752250203e-06, "loss": 0.0276, "step": 22783 }, { "epoch": 7.01, "learning_rate": 4.340695123718402e-06, "loss": 0.0538, "step": 22784 }, { "epoch": 7.01, "learning_rate": 4.339873551388711e-06, "loss": 0.0519, "step": 22785 }, { "epoch": 7.01, "learning_rate": 4.339052035269291e-06, "loss": 0.0604, "step": 22786 }, { "epoch": 7.01, "learning_rate": 4.338230575368305e-06, "loss": 0.0648, "step": 22787 }, { "epoch": 7.01, "learning_rate": 4.33740917169391e-06, "loss": 0.0621, "step": 22788 }, { "epoch": 7.01, "learning_rate": 4.336587824254255e-06, "loss": 0.0599, "step": 22789 }, { "epoch": 7.01, "learning_rate": 4.335766533057505e-06, "loss": 0.0445, "step": 22790 }, { "epoch": 7.01, "learning_rate": 4.334945298111813e-06, "loss": 0.0401, "step": 22791 }, { "epoch": 7.01, "learning_rate": 4.3341241194253356e-06, "loss": 0.0854, "step": 22792 }, { "epoch": 7.01, "learning_rate": 4.333302997006229e-06, "loss": 0.0529, "step": 22793 }, { "epoch": 7.01, "learning_rate": 4.332481930862641e-06, "loss": 0.0302, "step": 22794 }, { "epoch": 7.01, "learning_rate": 4.33166092100274e-06, "loss": 0.0249, "step": 22795 }, { "epoch": 7.01, "learning_rate": 4.33083996743466e-06, "loss": 0.0497, "step": 22796 }, { "epoch": 7.01, "learning_rate": 4.330019070166569e-06, "loss": 0.0605, "step": 22797 }, { "epoch": 7.01, "learning_rate": 4.3291982292066144e-06, "loss": 0.0281, "step": 22798 }, { "epoch": 7.01, "learning_rate": 4.328377444562948e-06, "loss": 0.0516, "step": 22799 }, { "epoch": 7.02, "learning_rate": 4.32755671624372e-06, "loss": 0.0744, "step": 22800 }, { "epoch": 7.02, "learning_rate": 4.326736044257079e-06, "loss": 0.0656, "step": 22801 }, { "epoch": 7.02, "learning_rate": 4.325915428611185e-06, "loss": 0.0402, "step": 22802 }, { "epoch": 7.02, "learning_rate": 4.325094869314176e-06, "loss": 0.0689, "step": 22803 }, { "epoch": 7.02, "learning_rate": 4.3242743663742e-06, "loss": 0.0918, "step": 22804 }, { "epoch": 7.02, "learning_rate": 4.32345391979942e-06, "loss": 0.0288, "step": 22805 }, { "epoch": 7.02, "learning_rate": 4.322633529597966e-06, "loss": 0.0534, "step": 22806 }, { "epoch": 7.02, "learning_rate": 4.321813195777998e-06, "loss": 0.0492, "step": 22807 }, { "epoch": 7.02, "learning_rate": 4.320992918347656e-06, "loss": 0.0584, "step": 22808 }, { "epoch": 7.02, "learning_rate": 4.3201726973150904e-06, "loss": 0.0711, "step": 22809 }, { "epoch": 7.02, "learning_rate": 4.319352532688444e-06, "loss": 0.0414, "step": 22810 }, { "epoch": 7.02, "learning_rate": 4.318532424475859e-06, "loss": 0.0531, "step": 22811 }, { "epoch": 7.02, "learning_rate": 4.31771237268549e-06, "loss": 0.0505, "step": 22812 }, { "epoch": 7.02, "learning_rate": 4.31689237732547e-06, "loss": 0.025, "step": 22813 }, { "epoch": 7.02, "learning_rate": 4.3160724384039435e-06, "loss": 0.0427, "step": 22814 }, { "epoch": 7.02, "learning_rate": 4.3152525559290596e-06, "loss": 0.0513, "step": 22815 }, { "epoch": 7.02, "learning_rate": 4.314432729908956e-06, "loss": 0.0572, "step": 22816 }, { "epoch": 7.02, "learning_rate": 4.313612960351776e-06, "loss": 0.049, "step": 22817 }, { "epoch": 7.02, "learning_rate": 4.312793247265661e-06, "loss": 0.0626, "step": 22818 }, { "epoch": 7.02, "learning_rate": 4.31197359065875e-06, "loss": 0.0415, "step": 22819 }, { "epoch": 7.02, "learning_rate": 4.311153990539183e-06, "loss": 0.0906, "step": 22820 }, { "epoch": 7.02, "learning_rate": 4.310334446915096e-06, "loss": 0.0466, "step": 22821 }, { "epoch": 7.02, "learning_rate": 4.309514959794637e-06, "loss": 0.0497, "step": 22822 }, { "epoch": 7.02, "learning_rate": 4.308695529185941e-06, "loss": 0.0464, "step": 22823 }, { "epoch": 7.02, "learning_rate": 4.307876155097137e-06, "loss": 0.0643, "step": 22824 }, { "epoch": 7.02, "learning_rate": 4.307056837536373e-06, "loss": 0.0688, "step": 22825 }, { "epoch": 7.02, "learning_rate": 4.30623757651178e-06, "loss": 0.0394, "step": 22826 }, { "epoch": 7.02, "learning_rate": 4.3054183720314965e-06, "loss": 0.0405, "step": 22827 }, { "epoch": 7.02, "learning_rate": 4.304599224103657e-06, "loss": 0.0442, "step": 22828 }, { "epoch": 7.02, "learning_rate": 4.303780132736395e-06, "loss": 0.0692, "step": 22829 }, { "epoch": 7.02, "learning_rate": 4.302961097937847e-06, "loss": 0.0411, "step": 22830 }, { "epoch": 7.02, "learning_rate": 4.302142119716142e-06, "loss": 0.0383, "step": 22831 }, { "epoch": 7.02, "learning_rate": 4.3013231980794215e-06, "loss": 0.0456, "step": 22832 }, { "epoch": 7.03, "learning_rate": 4.300504333035815e-06, "loss": 0.0387, "step": 22833 }, { "epoch": 7.03, "learning_rate": 4.299685524593449e-06, "loss": 0.0588, "step": 22834 }, { "epoch": 7.03, "learning_rate": 4.298866772760461e-06, "loss": 0.0416, "step": 22835 }, { "epoch": 7.03, "learning_rate": 4.298048077544981e-06, "loss": 0.0573, "step": 22836 }, { "epoch": 7.03, "learning_rate": 4.2972294389551385e-06, "loss": 0.0682, "step": 22837 }, { "epoch": 7.03, "learning_rate": 4.296410856999062e-06, "loss": 0.05, "step": 22838 }, { "epoch": 7.03, "learning_rate": 4.2955923316848814e-06, "loss": 0.0452, "step": 22839 }, { "epoch": 7.03, "learning_rate": 4.294773863020732e-06, "loss": 0.0471, "step": 22840 }, { "epoch": 7.03, "learning_rate": 4.293955451014728e-06, "loss": 0.0426, "step": 22841 }, { "epoch": 7.03, "learning_rate": 4.293137095675011e-06, "loss": 0.0356, "step": 22842 }, { "epoch": 7.03, "learning_rate": 4.292318797009701e-06, "loss": 0.0309, "step": 22843 }, { "epoch": 7.03, "learning_rate": 4.291500555026926e-06, "loss": 0.0377, "step": 22844 }, { "epoch": 7.03, "learning_rate": 4.290682369734812e-06, "loss": 0.0758, "step": 22845 }, { "epoch": 7.03, "learning_rate": 4.289864241141483e-06, "loss": 0.0473, "step": 22846 }, { "epoch": 7.03, "learning_rate": 4.289046169255065e-06, "loss": 0.0434, "step": 22847 }, { "epoch": 7.03, "learning_rate": 4.288228154083682e-06, "loss": 0.0362, "step": 22848 }, { "epoch": 7.03, "learning_rate": 4.287410195635454e-06, "loss": 0.0469, "step": 22849 }, { "epoch": 7.03, "learning_rate": 4.286592293918515e-06, "loss": 0.0494, "step": 22850 }, { "epoch": 7.03, "learning_rate": 4.2857744489409725e-06, "loss": 0.0473, "step": 22851 }, { "epoch": 7.03, "learning_rate": 4.284956660710961e-06, "loss": 0.0737, "step": 22852 }, { "epoch": 7.03, "learning_rate": 4.284138929236595e-06, "loss": 0.0373, "step": 22853 }, { "epoch": 7.03, "learning_rate": 4.283321254525999e-06, "loss": 0.0368, "step": 22854 }, { "epoch": 7.03, "learning_rate": 4.2825036365872904e-06, "loss": 0.0405, "step": 22855 }, { "epoch": 7.03, "learning_rate": 4.281686075428587e-06, "loss": 0.0591, "step": 22856 }, { "epoch": 7.03, "learning_rate": 4.280868571058019e-06, "loss": 0.0521, "step": 22857 }, { "epoch": 7.03, "learning_rate": 4.280051123483693e-06, "loss": 0.0577, "step": 22858 }, { "epoch": 7.03, "learning_rate": 4.2792337327137265e-06, "loss": 0.0606, "step": 22859 }, { "epoch": 7.03, "learning_rate": 4.278416398756245e-06, "loss": 0.0416, "step": 22860 }, { "epoch": 7.03, "learning_rate": 4.277599121619363e-06, "loss": 0.0417, "step": 22861 }, { "epoch": 7.03, "learning_rate": 4.276781901311195e-06, "loss": 0.0286, "step": 22862 }, { "epoch": 7.03, "learning_rate": 4.275964737839857e-06, "loss": 0.0472, "step": 22863 }, { "epoch": 7.03, "learning_rate": 4.2751476312134655e-06, "loss": 0.0453, "step": 22864 }, { "epoch": 7.04, "learning_rate": 4.274330581440133e-06, "loss": 0.0661, "step": 22865 }, { "epoch": 7.04, "learning_rate": 4.273513588527973e-06, "loss": 0.0465, "step": 22866 }, { "epoch": 7.04, "learning_rate": 4.272696652485107e-06, "loss": 0.0215, "step": 22867 }, { "epoch": 7.04, "learning_rate": 4.271879773319637e-06, "loss": 0.0311, "step": 22868 }, { "epoch": 7.04, "learning_rate": 4.271062951039678e-06, "loss": 0.0663, "step": 22869 }, { "epoch": 7.04, "learning_rate": 4.270246185653347e-06, "loss": 0.0648, "step": 22870 }, { "epoch": 7.04, "learning_rate": 4.269429477168752e-06, "loss": 0.0411, "step": 22871 }, { "epoch": 7.04, "learning_rate": 4.2686128255940025e-06, "loss": 0.0662, "step": 22872 }, { "epoch": 7.04, "learning_rate": 4.267796230937212e-06, "loss": 0.0452, "step": 22873 }, { "epoch": 7.04, "learning_rate": 4.266979693206487e-06, "loss": 0.0593, "step": 22874 }, { "epoch": 7.04, "learning_rate": 4.2661632124099376e-06, "loss": 0.0763, "step": 22875 }, { "epoch": 7.04, "learning_rate": 4.2653467885556675e-06, "loss": 0.0492, "step": 22876 }, { "epoch": 7.04, "learning_rate": 4.264530421651792e-06, "loss": 0.0295, "step": 22877 }, { "epoch": 7.04, "learning_rate": 4.263714111706421e-06, "loss": 0.0252, "step": 22878 }, { "epoch": 7.04, "learning_rate": 4.262897858727647e-06, "loss": 0.0712, "step": 22879 }, { "epoch": 7.04, "learning_rate": 4.26208166272359e-06, "loss": 0.0273, "step": 22880 }, { "epoch": 7.04, "learning_rate": 4.2612655237023484e-06, "loss": 0.0555, "step": 22881 }, { "epoch": 7.04, "learning_rate": 4.26044944167203e-06, "loss": 0.0265, "step": 22882 }, { "epoch": 7.04, "learning_rate": 4.259633416640737e-06, "loss": 0.0448, "step": 22883 }, { "epoch": 7.04, "learning_rate": 4.258817448616577e-06, "loss": 0.0673, "step": 22884 }, { "epoch": 7.04, "learning_rate": 4.258001537607649e-06, "loss": 0.0379, "step": 22885 }, { "epoch": 7.04, "learning_rate": 4.2571856836220546e-06, "loss": 0.0373, "step": 22886 }, { "epoch": 7.04, "learning_rate": 4.256369886667904e-06, "loss": 0.0487, "step": 22887 }, { "epoch": 7.04, "learning_rate": 4.255554146753296e-06, "loss": 0.05, "step": 22888 }, { "epoch": 7.04, "learning_rate": 4.254738463886323e-06, "loss": 0.0363, "step": 22889 }, { "epoch": 7.04, "learning_rate": 4.2539228380750955e-06, "loss": 0.0371, "step": 22890 }, { "epoch": 7.04, "learning_rate": 4.253107269327709e-06, "loss": 0.0444, "step": 22891 }, { "epoch": 7.04, "learning_rate": 4.252291757652266e-06, "loss": 0.054, "step": 22892 }, { "epoch": 7.04, "learning_rate": 4.251476303056862e-06, "loss": 0.05, "step": 22893 }, { "epoch": 7.04, "learning_rate": 4.250660905549592e-06, "loss": 0.0604, "step": 22894 }, { "epoch": 7.04, "learning_rate": 4.2498455651385665e-06, "loss": 0.0536, "step": 22895 }, { "epoch": 7.04, "learning_rate": 4.2490302818318655e-06, "loss": 0.0691, "step": 22896 }, { "epoch": 7.04, "learning_rate": 4.248215055637598e-06, "loss": 0.079, "step": 22897 }, { "epoch": 7.05, "learning_rate": 4.247399886563856e-06, "loss": 0.0594, "step": 22898 }, { "epoch": 7.05, "learning_rate": 4.246584774618735e-06, "loss": 0.0641, "step": 22899 }, { "epoch": 7.05, "learning_rate": 4.24576971981033e-06, "loss": 0.0626, "step": 22900 }, { "epoch": 7.05, "learning_rate": 4.244954722146731e-06, "loss": 0.035, "step": 22901 }, { "epoch": 7.05, "learning_rate": 4.244139781636043e-06, "loss": 0.0693, "step": 22902 }, { "epoch": 7.05, "learning_rate": 4.243324898286349e-06, "loss": 0.0803, "step": 22903 }, { "epoch": 7.05, "learning_rate": 4.242510072105739e-06, "loss": 0.0598, "step": 22904 }, { "epoch": 7.05, "learning_rate": 4.241695303102318e-06, "loss": 0.069, "step": 22905 }, { "epoch": 7.05, "learning_rate": 4.240880591284163e-06, "loss": 0.0451, "step": 22906 }, { "epoch": 7.05, "learning_rate": 4.240065936659374e-06, "loss": 0.0464, "step": 22907 }, { "epoch": 7.05, "learning_rate": 4.23925133923604e-06, "loss": 0.0421, "step": 22908 }, { "epoch": 7.05, "learning_rate": 4.238436799022248e-06, "loss": 0.0323, "step": 22909 }, { "epoch": 7.05, "learning_rate": 4.237622316026091e-06, "loss": 0.0437, "step": 22910 }, { "epoch": 7.05, "learning_rate": 4.23680789025565e-06, "loss": 0.0701, "step": 22911 }, { "epoch": 7.05, "learning_rate": 4.235993521719025e-06, "loss": 0.0296, "step": 22912 }, { "epoch": 7.05, "learning_rate": 4.2351792104242935e-06, "loss": 0.0781, "step": 22913 }, { "epoch": 7.05, "learning_rate": 4.234364956379542e-06, "loss": 0.0625, "step": 22914 }, { "epoch": 7.05, "learning_rate": 4.233550759592864e-06, "loss": 0.0122, "step": 22915 }, { "epoch": 7.05, "learning_rate": 4.2327366200723404e-06, "loss": 0.0674, "step": 22916 }, { "epoch": 7.05, "learning_rate": 4.231922537826059e-06, "loss": 0.0956, "step": 22917 }, { "epoch": 7.05, "learning_rate": 4.231108512862101e-06, "loss": 0.0492, "step": 22918 }, { "epoch": 7.05, "learning_rate": 4.230294545188554e-06, "loss": 0.0501, "step": 22919 }, { "epoch": 7.05, "learning_rate": 4.229480634813499e-06, "loss": 0.05, "step": 22920 }, { "epoch": 7.05, "learning_rate": 4.228666781745016e-06, "loss": 0.0603, "step": 22921 }, { "epoch": 7.05, "learning_rate": 4.227852985991198e-06, "loss": 0.0698, "step": 22922 }, { "epoch": 7.05, "learning_rate": 4.227039247560115e-06, "loss": 0.0406, "step": 22923 }, { "epoch": 7.05, "learning_rate": 4.226225566459851e-06, "loss": 0.0484, "step": 22924 }, { "epoch": 7.05, "learning_rate": 4.225411942698491e-06, "loss": 0.0552, "step": 22925 }, { "epoch": 7.05, "learning_rate": 4.224598376284113e-06, "loss": 0.033, "step": 22926 }, { "epoch": 7.05, "learning_rate": 4.223784867224795e-06, "loss": 0.0868, "step": 22927 }, { "epoch": 7.05, "learning_rate": 4.222971415528617e-06, "loss": 0.0389, "step": 22928 }, { "epoch": 7.05, "learning_rate": 4.222158021203657e-06, "loss": 0.0548, "step": 22929 }, { "epoch": 7.06, "learning_rate": 4.221344684257993e-06, "loss": 0.0593, "step": 22930 }, { "epoch": 7.06, "learning_rate": 4.220531404699699e-06, "loss": 0.0593, "step": 22931 }, { "epoch": 7.06, "learning_rate": 4.2197181825368585e-06, "loss": 0.057, "step": 22932 }, { "epoch": 7.06, "learning_rate": 4.218905017777546e-06, "loss": 0.0489, "step": 22933 }, { "epoch": 7.06, "learning_rate": 4.218091910429828e-06, "loss": 0.0623, "step": 22934 }, { "epoch": 7.06, "learning_rate": 4.21727886050179e-06, "loss": 0.0543, "step": 22935 }, { "epoch": 7.06, "learning_rate": 4.216465868001502e-06, "loss": 0.0367, "step": 22936 }, { "epoch": 7.06, "learning_rate": 4.215652932937039e-06, "loss": 0.0696, "step": 22937 }, { "epoch": 7.06, "learning_rate": 4.214840055316473e-06, "loss": 0.0825, "step": 22938 }, { "epoch": 7.06, "learning_rate": 4.214027235147874e-06, "loss": 0.0518, "step": 22939 }, { "epoch": 7.06, "learning_rate": 4.2132144724393235e-06, "loss": 0.0626, "step": 22940 }, { "epoch": 7.06, "learning_rate": 4.212401767198879e-06, "loss": 0.0498, "step": 22941 }, { "epoch": 7.06, "learning_rate": 4.211589119434622e-06, "loss": 0.0281, "step": 22942 }, { "epoch": 7.06, "learning_rate": 4.210776529154625e-06, "loss": 0.0355, "step": 22943 }, { "epoch": 7.06, "learning_rate": 4.209963996366945e-06, "loss": 0.0726, "step": 22944 }, { "epoch": 7.06, "learning_rate": 4.209151521079662e-06, "loss": 0.0579, "step": 22945 }, { "epoch": 7.06, "learning_rate": 4.208339103300842e-06, "loss": 0.0604, "step": 22946 }, { "epoch": 7.06, "learning_rate": 4.207526743038551e-06, "loss": 0.0679, "step": 22947 }, { "epoch": 7.06, "learning_rate": 4.20671444030086e-06, "loss": 0.0509, "step": 22948 }, { "epoch": 7.06, "learning_rate": 4.205902195095829e-06, "loss": 0.0833, "step": 22949 }, { "epoch": 7.06, "learning_rate": 4.205090007431537e-06, "loss": 0.0468, "step": 22950 }, { "epoch": 7.06, "learning_rate": 4.204277877316035e-06, "loss": 0.0624, "step": 22951 }, { "epoch": 7.06, "learning_rate": 4.203465804757397e-06, "loss": 0.0454, "step": 22952 }, { "epoch": 7.06, "learning_rate": 4.202653789763688e-06, "loss": 0.0603, "step": 22953 }, { "epoch": 7.06, "learning_rate": 4.201841832342967e-06, "loss": 0.0489, "step": 22954 }, { "epoch": 7.06, "learning_rate": 4.201029932503303e-06, "loss": 0.0413, "step": 22955 }, { "epoch": 7.06, "learning_rate": 4.200218090252752e-06, "loss": 0.0536, "step": 22956 }, { "epoch": 7.06, "learning_rate": 4.199406305599387e-06, "loss": 0.0393, "step": 22957 }, { "epoch": 7.06, "learning_rate": 4.198594578551257e-06, "loss": 0.029, "step": 22958 }, { "epoch": 7.06, "learning_rate": 4.197782909116434e-06, "loss": 0.0562, "step": 22959 }, { "epoch": 7.06, "learning_rate": 4.196971297302972e-06, "loss": 0.0421, "step": 22960 }, { "epoch": 7.06, "learning_rate": 4.196159743118935e-06, "loss": 0.0512, "step": 22961 }, { "epoch": 7.06, "learning_rate": 4.19534824657238e-06, "loss": 0.0402, "step": 22962 }, { "epoch": 7.07, "learning_rate": 4.194536807671366e-06, "loss": 0.0326, "step": 22963 }, { "epoch": 7.07, "learning_rate": 4.193725426423953e-06, "loss": 0.0385, "step": 22964 }, { "epoch": 7.07, "learning_rate": 4.192914102838198e-06, "loss": 0.0416, "step": 22965 }, { "epoch": 7.07, "learning_rate": 4.1921028369221535e-06, "loss": 0.0569, "step": 22966 }, { "epoch": 7.07, "learning_rate": 4.191291628683888e-06, "loss": 0.0408, "step": 22967 }, { "epoch": 7.07, "learning_rate": 4.190480478131443e-06, "loss": 0.0633, "step": 22968 }, { "epoch": 7.07, "learning_rate": 4.1896693852728845e-06, "loss": 0.0271, "step": 22969 }, { "epoch": 7.07, "learning_rate": 4.1888583501162635e-06, "loss": 0.0479, "step": 22970 }, { "epoch": 7.07, "learning_rate": 4.188047372669636e-06, "loss": 0.0771, "step": 22971 }, { "epoch": 7.07, "learning_rate": 4.187236452941054e-06, "loss": 0.0311, "step": 22972 }, { "epoch": 7.07, "learning_rate": 4.186425590938568e-06, "loss": 0.0337, "step": 22973 }, { "epoch": 7.07, "learning_rate": 4.1856147866702405e-06, "loss": 0.0369, "step": 22974 }, { "epoch": 7.07, "learning_rate": 4.184804040144113e-06, "loss": 0.0409, "step": 22975 }, { "epoch": 7.07, "learning_rate": 4.183993351368238e-06, "loss": 0.0484, "step": 22976 }, { "epoch": 7.07, "learning_rate": 4.183182720350672e-06, "loss": 0.0344, "step": 22977 }, { "epoch": 7.07, "learning_rate": 4.182372147099463e-06, "loss": 0.0377, "step": 22978 }, { "epoch": 7.07, "learning_rate": 4.181561631622659e-06, "loss": 0.0514, "step": 22979 }, { "epoch": 7.07, "learning_rate": 4.180751173928312e-06, "loss": 0.0448, "step": 22980 }, { "epoch": 7.07, "learning_rate": 4.179940774024469e-06, "loss": 0.052, "step": 22981 }, { "epoch": 7.07, "learning_rate": 4.179130431919178e-06, "loss": 0.0629, "step": 22982 }, { "epoch": 7.07, "learning_rate": 4.178320147620482e-06, "loss": 0.0556, "step": 22983 }, { "epoch": 7.07, "learning_rate": 4.17750992113644e-06, "loss": 0.0836, "step": 22984 }, { "epoch": 7.07, "learning_rate": 4.176699752475087e-06, "loss": 0.0803, "step": 22985 }, { "epoch": 7.07, "learning_rate": 4.175889641644469e-06, "loss": 0.0394, "step": 22986 }, { "epoch": 7.07, "learning_rate": 4.175079588652638e-06, "loss": 0.068, "step": 22987 }, { "epoch": 7.07, "learning_rate": 4.174269593507635e-06, "loss": 0.0737, "step": 22988 }, { "epoch": 7.07, "learning_rate": 4.173459656217505e-06, "loss": 0.0213, "step": 22989 }, { "epoch": 7.07, "learning_rate": 4.172649776790288e-06, "loss": 0.0201, "step": 22990 }, { "epoch": 7.07, "learning_rate": 4.171839955234032e-06, "loss": 0.0603, "step": 22991 }, { "epoch": 7.07, "learning_rate": 4.171030191556775e-06, "loss": 0.0452, "step": 22992 }, { "epoch": 7.07, "learning_rate": 4.170220485766559e-06, "loss": 0.0561, "step": 22993 }, { "epoch": 7.07, "learning_rate": 4.169410837871427e-06, "loss": 0.0424, "step": 22994 }, { "epoch": 7.08, "learning_rate": 4.1686012478794245e-06, "loss": 0.0368, "step": 22995 }, { "epoch": 7.08, "learning_rate": 4.167791715798578e-06, "loss": 0.0568, "step": 22996 }, { "epoch": 7.08, "learning_rate": 4.166982241636939e-06, "loss": 0.0575, "step": 22997 }, { "epoch": 7.08, "learning_rate": 4.166172825402542e-06, "loss": 0.0624, "step": 22998 }, { "epoch": 7.08, "learning_rate": 4.165363467103426e-06, "loss": 0.049, "step": 22999 }, { "epoch": 7.08, "learning_rate": 4.1645541667476276e-06, "loss": 0.0583, "step": 23000 }, { "epoch": 7.08, "learning_rate": 4.163744924343183e-06, "loss": 0.0557, "step": 23001 }, { "epoch": 7.08, "learning_rate": 4.162935739898133e-06, "loss": 0.0403, "step": 23002 }, { "epoch": 7.08, "learning_rate": 4.1621266134205055e-06, "loss": 0.0498, "step": 23003 }, { "epoch": 7.08, "learning_rate": 4.161317544918345e-06, "loss": 0.0407, "step": 23004 }, { "epoch": 7.08, "learning_rate": 4.160508534399685e-06, "loss": 0.0287, "step": 23005 }, { "epoch": 7.08, "learning_rate": 4.159699581872552e-06, "loss": 0.0573, "step": 23006 }, { "epoch": 7.08, "learning_rate": 4.158890687344986e-06, "loss": 0.0464, "step": 23007 }, { "epoch": 7.08, "learning_rate": 4.158081850825021e-06, "loss": 0.0293, "step": 23008 }, { "epoch": 7.08, "learning_rate": 4.1572730723206855e-06, "loss": 0.0371, "step": 23009 }, { "epoch": 7.08, "learning_rate": 4.1564643518400135e-06, "loss": 0.0331, "step": 23010 }, { "epoch": 7.08, "learning_rate": 4.155655689391034e-06, "loss": 0.0893, "step": 23011 }, { "epoch": 7.08, "learning_rate": 4.1548470849817854e-06, "loss": 0.0727, "step": 23012 }, { "epoch": 7.08, "learning_rate": 4.154038538620285e-06, "loss": 0.0522, "step": 23013 }, { "epoch": 7.08, "learning_rate": 4.153230050314574e-06, "loss": 0.0399, "step": 23014 }, { "epoch": 7.08, "learning_rate": 4.1524216200726775e-06, "loss": 0.0529, "step": 23015 }, { "epoch": 7.08, "learning_rate": 4.1516132479026225e-06, "loss": 0.0469, "step": 23016 }, { "epoch": 7.08, "learning_rate": 4.1508049338124386e-06, "loss": 0.0701, "step": 23017 }, { "epoch": 7.08, "learning_rate": 4.149996677810149e-06, "loss": 0.0479, "step": 23018 }, { "epoch": 7.08, "learning_rate": 4.14918847990379e-06, "loss": 0.0342, "step": 23019 }, { "epoch": 7.08, "learning_rate": 4.14838034010138e-06, "loss": 0.0406, "step": 23020 }, { "epoch": 7.08, "learning_rate": 4.147572258410941e-06, "loss": 0.0334, "step": 23021 }, { "epoch": 7.08, "learning_rate": 4.146764234840511e-06, "loss": 0.0522, "step": 23022 }, { "epoch": 7.08, "learning_rate": 4.145956269398101e-06, "loss": 0.0342, "step": 23023 }, { "epoch": 7.08, "learning_rate": 4.145148362091742e-06, "loss": 0.0469, "step": 23024 }, { "epoch": 7.08, "learning_rate": 4.144340512929457e-06, "loss": 0.0522, "step": 23025 }, { "epoch": 7.08, "learning_rate": 4.143532721919265e-06, "loss": 0.0522, "step": 23026 }, { "epoch": 7.08, "learning_rate": 4.142724989069193e-06, "loss": 0.0327, "step": 23027 }, { "epoch": 7.09, "learning_rate": 4.141917314387256e-06, "loss": 0.0516, "step": 23028 }, { "epoch": 7.09, "learning_rate": 4.141109697881487e-06, "loss": 0.0235, "step": 23029 }, { "epoch": 7.09, "learning_rate": 4.140302139559893e-06, "loss": 0.0469, "step": 23030 }, { "epoch": 7.09, "learning_rate": 4.1394946394304965e-06, "loss": 0.0421, "step": 23031 }, { "epoch": 7.09, "learning_rate": 4.138687197501323e-06, "loss": 0.0467, "step": 23032 }, { "epoch": 7.09, "learning_rate": 4.137879813780388e-06, "loss": 0.0407, "step": 23033 }, { "epoch": 7.09, "learning_rate": 4.137072488275708e-06, "loss": 0.0496, "step": 23034 }, { "epoch": 7.09, "learning_rate": 4.136265220995303e-06, "loss": 0.0502, "step": 23035 }, { "epoch": 7.09, "learning_rate": 4.135458011947188e-06, "loss": 0.0498, "step": 23036 }, { "epoch": 7.09, "learning_rate": 4.13465086113938e-06, "loss": 0.1092, "step": 23037 }, { "epoch": 7.09, "learning_rate": 4.13384376857989e-06, "loss": 0.0273, "step": 23038 }, { "epoch": 7.09, "learning_rate": 4.133036734276743e-06, "loss": 0.028, "step": 23039 }, { "epoch": 7.09, "learning_rate": 4.132229758237951e-06, "loss": 0.05, "step": 23040 }, { "epoch": 7.09, "learning_rate": 4.1314228404715185e-06, "loss": 0.032, "step": 23041 }, { "epoch": 7.09, "learning_rate": 4.1306159809854685e-06, "loss": 0.0312, "step": 23042 }, { "epoch": 7.09, "learning_rate": 4.129809179787813e-06, "loss": 0.0391, "step": 23043 }, { "epoch": 7.09, "learning_rate": 4.129002436886561e-06, "loss": 0.0483, "step": 23044 }, { "epoch": 7.09, "learning_rate": 4.128195752289726e-06, "loss": 0.0516, "step": 23045 }, { "epoch": 7.09, "learning_rate": 4.127389126005319e-06, "loss": 0.0514, "step": 23046 }, { "epoch": 7.09, "learning_rate": 4.12658255804135e-06, "loss": 0.0441, "step": 23047 }, { "epoch": 7.09, "learning_rate": 4.125776048405826e-06, "loss": 0.0693, "step": 23048 }, { "epoch": 7.09, "learning_rate": 4.1249695971067635e-06, "loss": 0.0283, "step": 23049 }, { "epoch": 7.09, "learning_rate": 4.1241632041521705e-06, "loss": 0.0362, "step": 23050 }, { "epoch": 7.09, "learning_rate": 4.123356869550046e-06, "loss": 0.0612, "step": 23051 }, { "epoch": 7.09, "learning_rate": 4.122550593308406e-06, "loss": 0.0253, "step": 23052 }, { "epoch": 7.09, "learning_rate": 4.121744375435256e-06, "loss": 0.0318, "step": 23053 }, { "epoch": 7.09, "learning_rate": 4.120938215938602e-06, "loss": 0.0428, "step": 23054 }, { "epoch": 7.09, "learning_rate": 4.120132114826451e-06, "loss": 0.053, "step": 23055 }, { "epoch": 7.09, "learning_rate": 4.119326072106802e-06, "loss": 0.0366, "step": 23056 }, { "epoch": 7.09, "learning_rate": 4.118520087787673e-06, "loss": 0.0637, "step": 23057 }, { "epoch": 7.09, "learning_rate": 4.117714161877052e-06, "loss": 0.0389, "step": 23058 }, { "epoch": 7.09, "learning_rate": 4.116908294382956e-06, "loss": 0.0509, "step": 23059 }, { "epoch": 7.1, "learning_rate": 4.11610248531338e-06, "loss": 0.0395, "step": 23060 }, { "epoch": 7.1, "learning_rate": 4.115296734676331e-06, "loss": 0.0749, "step": 23061 }, { "epoch": 7.1, "learning_rate": 4.114491042479808e-06, "loss": 0.0405, "step": 23062 }, { "epoch": 7.1, "learning_rate": 4.113685408731813e-06, "loss": 0.0673, "step": 23063 }, { "epoch": 7.1, "learning_rate": 4.112879833440347e-06, "loss": 0.0445, "step": 23064 }, { "epoch": 7.1, "learning_rate": 4.1120743166134115e-06, "loss": 0.0375, "step": 23065 }, { "epoch": 7.1, "learning_rate": 4.111268858259e-06, "loss": 0.0735, "step": 23066 }, { "epoch": 7.1, "learning_rate": 4.110463458385122e-06, "loss": 0.0478, "step": 23067 }, { "epoch": 7.1, "learning_rate": 4.109658116999764e-06, "loss": 0.0543, "step": 23068 }, { "epoch": 7.1, "learning_rate": 4.108852834110932e-06, "loss": 0.0853, "step": 23069 }, { "epoch": 7.1, "learning_rate": 4.108047609726622e-06, "loss": 0.054, "step": 23070 }, { "epoch": 7.1, "learning_rate": 4.10724244385483e-06, "loss": 0.0709, "step": 23071 }, { "epoch": 7.1, "learning_rate": 4.10643733650355e-06, "loss": 0.0611, "step": 23072 }, { "epoch": 7.1, "learning_rate": 4.105632287680776e-06, "loss": 0.0494, "step": 23073 }, { "epoch": 7.1, "learning_rate": 4.104827297394513e-06, "loss": 0.0439, "step": 23074 }, { "epoch": 7.1, "learning_rate": 4.104022365652744e-06, "loss": 0.0623, "step": 23075 }, { "epoch": 7.1, "learning_rate": 4.103217492463463e-06, "loss": 0.0514, "step": 23076 }, { "epoch": 7.1, "learning_rate": 4.1024126778346705e-06, "loss": 0.051, "step": 23077 }, { "epoch": 7.1, "learning_rate": 4.101607921774357e-06, "loss": 0.0255, "step": 23078 }, { "epoch": 7.1, "learning_rate": 4.100803224290512e-06, "loss": 0.0472, "step": 23079 }, { "epoch": 7.1, "learning_rate": 4.099998585391128e-06, "loss": 0.0553, "step": 23080 }, { "epoch": 7.1, "learning_rate": 4.099194005084195e-06, "loss": 0.0355, "step": 23081 }, { "epoch": 7.1, "learning_rate": 4.0983894833777035e-06, "loss": 0.0377, "step": 23082 }, { "epoch": 7.1, "learning_rate": 4.097585020279641e-06, "loss": 0.0684, "step": 23083 }, { "epoch": 7.1, "learning_rate": 4.096780615798006e-06, "loss": 0.0499, "step": 23084 }, { "epoch": 7.1, "learning_rate": 4.095976269940777e-06, "loss": 0.0444, "step": 23085 }, { "epoch": 7.1, "learning_rate": 4.095171982715941e-06, "loss": 0.0457, "step": 23086 }, { "epoch": 7.1, "learning_rate": 4.0943677541314915e-06, "loss": 0.064, "step": 23087 }, { "epoch": 7.1, "learning_rate": 4.093563584195415e-06, "loss": 0.067, "step": 23088 }, { "epoch": 7.1, "learning_rate": 4.092759472915695e-06, "loss": 0.0711, "step": 23089 }, { "epoch": 7.1, "learning_rate": 4.091955420300318e-06, "loss": 0.0687, "step": 23090 }, { "epoch": 7.1, "learning_rate": 4.091151426357267e-06, "loss": 0.0534, "step": 23091 }, { "epoch": 7.1, "learning_rate": 4.09034749109453e-06, "loss": 0.0612, "step": 23092 }, { "epoch": 7.11, "learning_rate": 4.089543614520084e-06, "loss": 0.0414, "step": 23093 }, { "epoch": 7.11, "learning_rate": 4.088739796641922e-06, "loss": 0.0488, "step": 23094 }, { "epoch": 7.11, "learning_rate": 4.087936037468024e-06, "loss": 0.0675, "step": 23095 }, { "epoch": 7.11, "learning_rate": 4.087132337006363e-06, "loss": 0.0254, "step": 23096 }, { "epoch": 7.11, "learning_rate": 4.086328695264931e-06, "loss": 0.0449, "step": 23097 }, { "epoch": 7.11, "learning_rate": 4.085525112251706e-06, "loss": 0.0514, "step": 23098 }, { "epoch": 7.11, "learning_rate": 4.084721587974667e-06, "loss": 0.0315, "step": 23099 }, { "epoch": 7.11, "learning_rate": 4.083918122441795e-06, "loss": 0.0428, "step": 23100 }, { "epoch": 7.11, "learning_rate": 4.083114715661069e-06, "loss": 0.0824, "step": 23101 }, { "epoch": 7.11, "learning_rate": 4.082311367640465e-06, "loss": 0.0528, "step": 23102 }, { "epoch": 7.11, "learning_rate": 4.08150807838796e-06, "loss": 0.0559, "step": 23103 }, { "epoch": 7.11, "learning_rate": 4.080704847911539e-06, "loss": 0.0485, "step": 23104 }, { "epoch": 7.11, "learning_rate": 4.079901676219177e-06, "loss": 0.0403, "step": 23105 }, { "epoch": 7.11, "learning_rate": 4.079098563318841e-06, "loss": 0.0611, "step": 23106 }, { "epoch": 7.11, "learning_rate": 4.078295509218516e-06, "loss": 0.0808, "step": 23107 }, { "epoch": 7.11, "learning_rate": 4.077492513926174e-06, "loss": 0.015, "step": 23108 }, { "epoch": 7.11, "learning_rate": 4.076689577449791e-06, "loss": 0.0533, "step": 23109 }, { "epoch": 7.11, "learning_rate": 4.075886699797338e-06, "loss": 0.0707, "step": 23110 }, { "epoch": 7.11, "learning_rate": 4.0750838809767875e-06, "loss": 0.0654, "step": 23111 }, { "epoch": 7.11, "learning_rate": 4.074281120996121e-06, "loss": 0.076, "step": 23112 }, { "epoch": 7.11, "learning_rate": 4.073478419863298e-06, "loss": 0.0336, "step": 23113 }, { "epoch": 7.11, "learning_rate": 4.072675777586299e-06, "loss": 0.0772, "step": 23114 }, { "epoch": 7.11, "learning_rate": 4.071873194173093e-06, "loss": 0.0475, "step": 23115 }, { "epoch": 7.11, "learning_rate": 4.0710706696316495e-06, "loss": 0.0584, "step": 23116 }, { "epoch": 7.11, "learning_rate": 4.070268203969938e-06, "loss": 0.0216, "step": 23117 }, { "epoch": 7.11, "learning_rate": 4.069465797195925e-06, "loss": 0.0841, "step": 23118 }, { "epoch": 7.11, "learning_rate": 4.0686634493175895e-06, "loss": 0.0597, "step": 23119 }, { "epoch": 7.11, "learning_rate": 4.0678611603428884e-06, "loss": 0.0394, "step": 23120 }, { "epoch": 7.11, "learning_rate": 4.0670589302797905e-06, "loss": 0.0633, "step": 23121 }, { "epoch": 7.11, "learning_rate": 4.066256759136272e-06, "loss": 0.0376, "step": 23122 }, { "epoch": 7.11, "learning_rate": 4.065454646920286e-06, "loss": 0.0627, "step": 23123 }, { "epoch": 7.11, "learning_rate": 4.0646525936398086e-06, "loss": 0.059, "step": 23124 }, { "epoch": 7.12, "learning_rate": 4.0638505993028e-06, "loss": 0.0571, "step": 23125 }, { "epoch": 7.12, "learning_rate": 4.0630486639172265e-06, "loss": 0.044, "step": 23126 }, { "epoch": 7.12, "learning_rate": 4.062246787491051e-06, "loss": 0.0563, "step": 23127 }, { "epoch": 7.12, "learning_rate": 4.061444970032235e-06, "loss": 0.0446, "step": 23128 }, { "epoch": 7.12, "learning_rate": 4.060643211548749e-06, "loss": 0.0722, "step": 23129 }, { "epoch": 7.12, "learning_rate": 4.059841512048547e-06, "loss": 0.0639, "step": 23130 }, { "epoch": 7.12, "learning_rate": 4.0590398715395895e-06, "loss": 0.0281, "step": 23131 }, { "epoch": 7.12, "learning_rate": 4.058238290029845e-06, "loss": 0.0683, "step": 23132 }, { "epoch": 7.12, "learning_rate": 4.057436767527271e-06, "loss": 0.0677, "step": 23133 }, { "epoch": 7.12, "learning_rate": 4.056635304039826e-06, "loss": 0.0808, "step": 23134 }, { "epoch": 7.12, "learning_rate": 4.055833899575471e-06, "loss": 0.0791, "step": 23135 }, { "epoch": 7.12, "learning_rate": 4.055032554142163e-06, "loss": 0.1055, "step": 23136 }, { "epoch": 7.12, "learning_rate": 4.054231267747862e-06, "loss": 0.0394, "step": 23137 }, { "epoch": 7.12, "learning_rate": 4.0534300404005176e-06, "loss": 0.0706, "step": 23138 }, { "epoch": 7.12, "learning_rate": 4.0526288721081e-06, "loss": 0.0374, "step": 23139 }, { "epoch": 7.12, "learning_rate": 4.051827762878561e-06, "loss": 0.0642, "step": 23140 }, { "epoch": 7.12, "learning_rate": 4.051026712719848e-06, "loss": 0.0625, "step": 23141 }, { "epoch": 7.12, "learning_rate": 4.050225721639925e-06, "loss": 0.0444, "step": 23142 }, { "epoch": 7.12, "learning_rate": 4.049424789646745e-06, "loss": 0.085, "step": 23143 }, { "epoch": 7.12, "learning_rate": 4.048623916748261e-06, "loss": 0.0342, "step": 23144 }, { "epoch": 7.12, "learning_rate": 4.0478231029524265e-06, "loss": 0.0501, "step": 23145 }, { "epoch": 7.12, "learning_rate": 4.047022348267194e-06, "loss": 0.0538, "step": 23146 }, { "epoch": 7.12, "learning_rate": 4.046221652700516e-06, "loss": 0.0614, "step": 23147 }, { "epoch": 7.12, "learning_rate": 4.04542101626034e-06, "loss": 0.0703, "step": 23148 }, { "epoch": 7.12, "learning_rate": 4.0446204389546265e-06, "loss": 0.0635, "step": 23149 }, { "epoch": 7.12, "learning_rate": 4.043819920791322e-06, "loss": 0.11, "step": 23150 }, { "epoch": 7.12, "learning_rate": 4.043019461778369e-06, "loss": 0.0581, "step": 23151 }, { "epoch": 7.12, "learning_rate": 4.042219061923726e-06, "loss": 0.0752, "step": 23152 }, { "epoch": 7.12, "learning_rate": 4.041418721235341e-06, "loss": 0.0746, "step": 23153 }, { "epoch": 7.12, "learning_rate": 4.040618439721157e-06, "loss": 0.0378, "step": 23154 }, { "epoch": 7.12, "learning_rate": 4.039818217389121e-06, "loss": 0.0599, "step": 23155 }, { "epoch": 7.12, "learning_rate": 4.039018054247188e-06, "loss": 0.0415, "step": 23156 }, { "epoch": 7.12, "learning_rate": 4.038217950303301e-06, "loss": 0.0366, "step": 23157 }, { "epoch": 7.13, "learning_rate": 4.037417905565399e-06, "loss": 0.0776, "step": 23158 }, { "epoch": 7.13, "learning_rate": 4.036617920041435e-06, "loss": 0.051, "step": 23159 }, { "epoch": 7.13, "learning_rate": 4.035817993739351e-06, "loss": 0.0396, "step": 23160 }, { "epoch": 7.13, "learning_rate": 4.035018126667093e-06, "loss": 0.0574, "step": 23161 }, { "epoch": 7.13, "learning_rate": 4.034218318832601e-06, "loss": 0.0569, "step": 23162 }, { "epoch": 7.13, "learning_rate": 4.033418570243819e-06, "loss": 0.0464, "step": 23163 }, { "epoch": 7.13, "learning_rate": 4.032618880908689e-06, "loss": 0.0194, "step": 23164 }, { "epoch": 7.13, "learning_rate": 4.031819250835152e-06, "loss": 0.0391, "step": 23165 }, { "epoch": 7.13, "learning_rate": 4.031019680031152e-06, "loss": 0.0492, "step": 23166 }, { "epoch": 7.13, "learning_rate": 4.030220168504632e-06, "loss": 0.0705, "step": 23167 }, { "epoch": 7.13, "learning_rate": 4.0294207162635205e-06, "loss": 0.0355, "step": 23168 }, { "epoch": 7.13, "learning_rate": 4.0286213233157665e-06, "loss": 0.0595, "step": 23169 }, { "epoch": 7.13, "learning_rate": 4.0278219896693075e-06, "loss": 0.0881, "step": 23170 }, { "epoch": 7.13, "learning_rate": 4.027022715332079e-06, "loss": 0.0791, "step": 23171 }, { "epoch": 7.13, "learning_rate": 4.026223500312021e-06, "loss": 0.0128, "step": 23172 }, { "epoch": 7.13, "learning_rate": 4.025424344617065e-06, "loss": 0.0541, "step": 23173 }, { "epoch": 7.13, "learning_rate": 4.02462524825516e-06, "loss": 0.0401, "step": 23174 }, { "epoch": 7.13, "learning_rate": 4.023826211234224e-06, "loss": 0.0675, "step": 23175 }, { "epoch": 7.13, "learning_rate": 4.0230272335622065e-06, "loss": 0.0276, "step": 23176 }, { "epoch": 7.13, "learning_rate": 4.022228315247037e-06, "loss": 0.0376, "step": 23177 }, { "epoch": 7.13, "learning_rate": 4.021429456296649e-06, "loss": 0.0645, "step": 23178 }, { "epoch": 7.13, "learning_rate": 4.020630656718977e-06, "loss": 0.0345, "step": 23179 }, { "epoch": 7.13, "learning_rate": 4.019831916521954e-06, "loss": 0.089, "step": 23180 }, { "epoch": 7.13, "learning_rate": 4.01903323571351e-06, "loss": 0.046, "step": 23181 }, { "epoch": 7.13, "learning_rate": 4.01823461430158e-06, "loss": 0.0732, "step": 23182 }, { "epoch": 7.13, "learning_rate": 4.017436052294088e-06, "loss": 0.0602, "step": 23183 }, { "epoch": 7.13, "learning_rate": 4.016637549698977e-06, "loss": 0.0386, "step": 23184 }, { "epoch": 7.13, "learning_rate": 4.015839106524163e-06, "loss": 0.0294, "step": 23185 }, { "epoch": 7.13, "learning_rate": 4.015040722777585e-06, "loss": 0.0346, "step": 23186 }, { "epoch": 7.13, "learning_rate": 4.014242398467169e-06, "loss": 0.0826, "step": 23187 }, { "epoch": 7.13, "learning_rate": 4.013444133600842e-06, "loss": 0.0827, "step": 23188 }, { "epoch": 7.13, "learning_rate": 4.012645928186533e-06, "loss": 0.0626, "step": 23189 }, { "epoch": 7.14, "learning_rate": 4.011847782232164e-06, "loss": 0.0702, "step": 23190 }, { "epoch": 7.14, "learning_rate": 4.0110496957456725e-06, "loss": 0.0606, "step": 23191 }, { "epoch": 7.14, "learning_rate": 4.010251668734974e-06, "loss": 0.0467, "step": 23192 }, { "epoch": 7.14, "learning_rate": 4.009453701207992e-06, "loss": 0.0369, "step": 23193 }, { "epoch": 7.14, "learning_rate": 4.008655793172661e-06, "loss": 0.0466, "step": 23194 }, { "epoch": 7.14, "learning_rate": 4.007857944636898e-06, "loss": 0.0565, "step": 23195 }, { "epoch": 7.14, "learning_rate": 4.007060155608629e-06, "loss": 0.0282, "step": 23196 }, { "epoch": 7.14, "learning_rate": 4.0062624260957765e-06, "loss": 0.05, "step": 23197 }, { "epoch": 7.14, "learning_rate": 4.0054647561062625e-06, "loss": 0.039, "step": 23198 }, { "epoch": 7.14, "learning_rate": 4.004667145648006e-06, "loss": 0.0567, "step": 23199 }, { "epoch": 7.14, "learning_rate": 4.003869594728929e-06, "loss": 0.0359, "step": 23200 }, { "epoch": 7.14, "learning_rate": 4.003072103356958e-06, "loss": 0.0542, "step": 23201 }, { "epoch": 7.14, "learning_rate": 4.002274671540006e-06, "loss": 0.0628, "step": 23202 }, { "epoch": 7.14, "learning_rate": 4.001477299285989e-06, "loss": 0.0679, "step": 23203 }, { "epoch": 7.14, "learning_rate": 4.0006799866028364e-06, "loss": 0.0428, "step": 23204 }, { "epoch": 7.14, "learning_rate": 3.999882733498461e-06, "loss": 0.0539, "step": 23205 }, { "epoch": 7.14, "learning_rate": 3.9990855399807775e-06, "loss": 0.0279, "step": 23206 }, { "epoch": 7.14, "learning_rate": 3.998288406057707e-06, "loss": 0.0308, "step": 23207 }, { "epoch": 7.14, "learning_rate": 3.997491331737162e-06, "loss": 0.0582, "step": 23208 }, { "epoch": 7.14, "learning_rate": 3.9966943170270625e-06, "loss": 0.0595, "step": 23209 }, { "epoch": 7.14, "learning_rate": 3.995897361935316e-06, "loss": 0.0822, "step": 23210 }, { "epoch": 7.14, "learning_rate": 3.9951004664698465e-06, "loss": 0.0477, "step": 23211 }, { "epoch": 7.14, "learning_rate": 3.994303630638568e-06, "loss": 0.0392, "step": 23212 }, { "epoch": 7.14, "learning_rate": 3.993506854449382e-06, "loss": 0.0586, "step": 23213 }, { "epoch": 7.14, "learning_rate": 3.99271013791021e-06, "loss": 0.0737, "step": 23214 }, { "epoch": 7.14, "learning_rate": 3.991913481028965e-06, "loss": 0.0377, "step": 23215 }, { "epoch": 7.14, "learning_rate": 3.991116883813556e-06, "loss": 0.067, "step": 23216 }, { "epoch": 7.14, "learning_rate": 3.990320346271895e-06, "loss": 0.0415, "step": 23217 }, { "epoch": 7.14, "learning_rate": 3.989523868411886e-06, "loss": 0.0363, "step": 23218 }, { "epoch": 7.14, "learning_rate": 3.988727450241454e-06, "loss": 0.0586, "step": 23219 }, { "epoch": 7.14, "learning_rate": 3.98793109176849e-06, "loss": 0.0777, "step": 23220 }, { "epoch": 7.14, "learning_rate": 3.9871347930009154e-06, "loss": 0.048, "step": 23221 }, { "epoch": 7.14, "learning_rate": 3.986338553946637e-06, "loss": 0.0382, "step": 23222 }, { "epoch": 7.15, "learning_rate": 3.985542374613552e-06, "loss": 0.0504, "step": 23223 }, { "epoch": 7.15, "learning_rate": 3.984746255009577e-06, "loss": 0.0498, "step": 23224 }, { "epoch": 7.15, "learning_rate": 3.983950195142618e-06, "loss": 0.0332, "step": 23225 }, { "epoch": 7.15, "learning_rate": 3.9831541950205755e-06, "loss": 0.0553, "step": 23226 }, { "epoch": 7.15, "learning_rate": 3.982358254651359e-06, "loss": 0.038, "step": 23227 }, { "epoch": 7.15, "learning_rate": 3.981562374042867e-06, "loss": 0.0668, "step": 23228 }, { "epoch": 7.15, "learning_rate": 3.980766553203014e-06, "loss": 0.0733, "step": 23229 }, { "epoch": 7.15, "learning_rate": 3.979970792139689e-06, "loss": 0.0304, "step": 23230 }, { "epoch": 7.15, "learning_rate": 3.9791750908608066e-06, "loss": 0.0529, "step": 23231 }, { "epoch": 7.15, "learning_rate": 3.978379449374264e-06, "loss": 0.0478, "step": 23232 }, { "epoch": 7.15, "learning_rate": 3.977583867687964e-06, "loss": 0.0343, "step": 23233 }, { "epoch": 7.15, "learning_rate": 3.976788345809805e-06, "loss": 0.0235, "step": 23234 }, { "epoch": 7.15, "learning_rate": 3.975992883747688e-06, "loss": 0.0398, "step": 23235 }, { "epoch": 7.15, "learning_rate": 3.975197481509518e-06, "loss": 0.0585, "step": 23236 }, { "epoch": 7.15, "learning_rate": 3.974402139103189e-06, "loss": 0.0427, "step": 23237 }, { "epoch": 7.15, "learning_rate": 3.9736068565365935e-06, "loss": 0.04, "step": 23238 }, { "epoch": 7.15, "learning_rate": 3.972811633817645e-06, "loss": 0.0321, "step": 23239 }, { "epoch": 7.15, "learning_rate": 3.972016470954224e-06, "loss": 0.0611, "step": 23240 }, { "epoch": 7.15, "learning_rate": 3.971221367954239e-06, "loss": 0.0556, "step": 23241 }, { "epoch": 7.15, "learning_rate": 3.970426324825581e-06, "loss": 0.0422, "step": 23242 }, { "epoch": 7.15, "learning_rate": 3.969631341576148e-06, "loss": 0.0486, "step": 23243 }, { "epoch": 7.15, "learning_rate": 3.968836418213832e-06, "loss": 0.0494, "step": 23244 }, { "epoch": 7.15, "learning_rate": 3.968041554746527e-06, "loss": 0.0683, "step": 23245 }, { "epoch": 7.15, "learning_rate": 3.967246751182135e-06, "loss": 0.0553, "step": 23246 }, { "epoch": 7.15, "learning_rate": 3.9664520075285385e-06, "loss": 0.0558, "step": 23247 }, { "epoch": 7.15, "learning_rate": 3.965657323793631e-06, "loss": 0.0415, "step": 23248 }, { "epoch": 7.15, "learning_rate": 3.964862699985312e-06, "loss": 0.0476, "step": 23249 }, { "epoch": 7.15, "learning_rate": 3.964068136111467e-06, "loss": 0.0415, "step": 23250 }, { "epoch": 7.15, "learning_rate": 3.96327363217999e-06, "loss": 0.0941, "step": 23251 }, { "epoch": 7.15, "learning_rate": 3.962479188198769e-06, "loss": 0.0343, "step": 23252 }, { "epoch": 7.15, "learning_rate": 3.961684804175694e-06, "loss": 0.0341, "step": 23253 }, { "epoch": 7.15, "learning_rate": 3.960890480118653e-06, "loss": 0.0443, "step": 23254 }, { "epoch": 7.16, "learning_rate": 3.9600962160355324e-06, "loss": 0.0681, "step": 23255 }, { "epoch": 7.16, "learning_rate": 3.9593020119342275e-06, "loss": 0.0439, "step": 23256 }, { "epoch": 7.16, "learning_rate": 3.958507867822623e-06, "loss": 0.026, "step": 23257 }, { "epoch": 7.16, "learning_rate": 3.957713783708596e-06, "loss": 0.0704, "step": 23258 }, { "epoch": 7.16, "learning_rate": 3.956919759600045e-06, "loss": 0.0548, "step": 23259 }, { "epoch": 7.16, "learning_rate": 3.956125795504849e-06, "loss": 0.0429, "step": 23260 }, { "epoch": 7.16, "learning_rate": 3.955331891430895e-06, "loss": 0.0543, "step": 23261 }, { "epoch": 7.16, "learning_rate": 3.954538047386066e-06, "loss": 0.0342, "step": 23262 }, { "epoch": 7.16, "learning_rate": 3.953744263378245e-06, "loss": 0.0909, "step": 23263 }, { "epoch": 7.16, "learning_rate": 3.952950539415317e-06, "loss": 0.0496, "step": 23264 }, { "epoch": 7.16, "learning_rate": 3.952156875505159e-06, "loss": 0.0486, "step": 23265 }, { "epoch": 7.16, "learning_rate": 3.9513632716556605e-06, "loss": 0.0504, "step": 23266 }, { "epoch": 7.16, "learning_rate": 3.950569727874704e-06, "loss": 0.0591, "step": 23267 }, { "epoch": 7.16, "learning_rate": 3.949776244170157e-06, "loss": 0.0619, "step": 23268 }, { "epoch": 7.16, "learning_rate": 3.948982820549911e-06, "loss": 0.0842, "step": 23269 }, { "epoch": 7.16, "learning_rate": 3.948189457021844e-06, "loss": 0.0507, "step": 23270 }, { "epoch": 7.16, "learning_rate": 3.947396153593833e-06, "loss": 0.047, "step": 23271 }, { "epoch": 7.16, "learning_rate": 3.946602910273755e-06, "loss": 0.0664, "step": 23272 }, { "epoch": 7.16, "learning_rate": 3.945809727069487e-06, "loss": 0.0525, "step": 23273 }, { "epoch": 7.16, "learning_rate": 3.945016603988915e-06, "loss": 0.0745, "step": 23274 }, { "epoch": 7.16, "learning_rate": 3.944223541039901e-06, "loss": 0.0242, "step": 23275 }, { "epoch": 7.16, "learning_rate": 3.943430538230333e-06, "loss": 0.0835, "step": 23276 }, { "epoch": 7.16, "learning_rate": 3.942637595568081e-06, "loss": 0.0566, "step": 23277 }, { "epoch": 7.16, "learning_rate": 3.941844713061019e-06, "loss": 0.0479, "step": 23278 }, { "epoch": 7.16, "learning_rate": 3.941051890717024e-06, "loss": 0.068, "step": 23279 }, { "epoch": 7.16, "learning_rate": 3.940259128543967e-06, "loss": 0.0188, "step": 23280 }, { "epoch": 7.16, "learning_rate": 3.939466426549722e-06, "loss": 0.0345, "step": 23281 }, { "epoch": 7.16, "learning_rate": 3.938673784742161e-06, "loss": 0.0546, "step": 23282 }, { "epoch": 7.16, "learning_rate": 3.937881203129151e-06, "loss": 0.038, "step": 23283 }, { "epoch": 7.16, "learning_rate": 3.937088681718576e-06, "loss": 0.0489, "step": 23284 }, { "epoch": 7.16, "learning_rate": 3.936296220518291e-06, "loss": 0.0526, "step": 23285 }, { "epoch": 7.16, "learning_rate": 3.935503819536176e-06, "loss": 0.0406, "step": 23286 }, { "epoch": 7.16, "learning_rate": 3.934711478780098e-06, "loss": 0.0526, "step": 23287 }, { "epoch": 7.17, "learning_rate": 3.933919198257925e-06, "loss": 0.0461, "step": 23288 }, { "epoch": 7.17, "learning_rate": 3.933126977977524e-06, "loss": 0.0446, "step": 23289 }, { "epoch": 7.17, "learning_rate": 3.932334817946761e-06, "loss": 0.0638, "step": 23290 }, { "epoch": 7.17, "learning_rate": 3.931542718173512e-06, "loss": 0.0486, "step": 23291 }, { "epoch": 7.17, "learning_rate": 3.930750678665632e-06, "loss": 0.0401, "step": 23292 }, { "epoch": 7.17, "learning_rate": 3.9299586994309905e-06, "loss": 0.0511, "step": 23293 }, { "epoch": 7.17, "learning_rate": 3.9291667804774545e-06, "loss": 0.0572, "step": 23294 }, { "epoch": 7.17, "learning_rate": 3.9283749218128885e-06, "loss": 0.0204, "step": 23295 }, { "epoch": 7.17, "learning_rate": 3.927583123445156e-06, "loss": 0.0553, "step": 23296 }, { "epoch": 7.17, "learning_rate": 3.926791385382118e-06, "loss": 0.057, "step": 23297 }, { "epoch": 7.17, "learning_rate": 3.925999707631639e-06, "loss": 0.0631, "step": 23298 }, { "epoch": 7.17, "learning_rate": 3.925208090201581e-06, "loss": 0.0678, "step": 23299 }, { "epoch": 7.17, "learning_rate": 3.924416533099803e-06, "loss": 0.0564, "step": 23300 }, { "epoch": 7.17, "learning_rate": 3.923625036334175e-06, "loss": 0.0379, "step": 23301 }, { "epoch": 7.17, "learning_rate": 3.922833599912545e-06, "loss": 0.0497, "step": 23302 }, { "epoch": 7.17, "learning_rate": 3.922042223842777e-06, "loss": 0.0424, "step": 23303 }, { "epoch": 7.17, "learning_rate": 3.9212509081327346e-06, "loss": 0.0561, "step": 23304 }, { "epoch": 7.17, "learning_rate": 3.920459652790271e-06, "loss": 0.0582, "step": 23305 }, { "epoch": 7.17, "learning_rate": 3.919668457823248e-06, "loss": 0.0463, "step": 23306 }, { "epoch": 7.17, "learning_rate": 3.9188773232395205e-06, "loss": 0.0425, "step": 23307 }, { "epoch": 7.17, "learning_rate": 3.918086249046946e-06, "loss": 0.074, "step": 23308 }, { "epoch": 7.17, "learning_rate": 3.91729523525338e-06, "loss": 0.0406, "step": 23309 }, { "epoch": 7.17, "learning_rate": 3.916504281866675e-06, "loss": 0.0444, "step": 23310 }, { "epoch": 7.17, "learning_rate": 3.915713388894693e-06, "loss": 0.0799, "step": 23311 }, { "epoch": 7.17, "learning_rate": 3.9149225563452875e-06, "loss": 0.0272, "step": 23312 }, { "epoch": 7.17, "learning_rate": 3.914131784226301e-06, "loss": 0.0532, "step": 23313 }, { "epoch": 7.17, "learning_rate": 3.913341072545601e-06, "loss": 0.0605, "step": 23314 }, { "epoch": 7.17, "learning_rate": 3.912550421311032e-06, "loss": 0.0474, "step": 23315 }, { "epoch": 7.17, "learning_rate": 3.911759830530448e-06, "loss": 0.028, "step": 23316 }, { "epoch": 7.17, "learning_rate": 3.9109693002117e-06, "loss": 0.0401, "step": 23317 }, { "epoch": 7.17, "learning_rate": 3.9101788303626384e-06, "loss": 0.0243, "step": 23318 }, { "epoch": 7.17, "learning_rate": 3.909388420991113e-06, "loss": 0.0197, "step": 23319 }, { "epoch": 7.18, "learning_rate": 3.908598072104972e-06, "loss": 0.0662, "step": 23320 }, { "epoch": 7.18, "learning_rate": 3.9078077837120685e-06, "loss": 0.0357, "step": 23321 }, { "epoch": 7.18, "learning_rate": 3.907017555820253e-06, "loss": 0.0397, "step": 23322 }, { "epoch": 7.18, "learning_rate": 3.90622738843736e-06, "loss": 0.0517, "step": 23323 }, { "epoch": 7.18, "learning_rate": 3.905437281571251e-06, "loss": 0.0589, "step": 23324 }, { "epoch": 7.18, "learning_rate": 3.904647235229765e-06, "loss": 0.0642, "step": 23325 }, { "epoch": 7.18, "learning_rate": 3.9038572494207495e-06, "loss": 0.0339, "step": 23326 }, { "epoch": 7.18, "learning_rate": 3.90306732415205e-06, "loss": 0.0643, "step": 23327 }, { "epoch": 7.18, "learning_rate": 3.902277459431509e-06, "loss": 0.0194, "step": 23328 }, { "epoch": 7.18, "learning_rate": 3.901487655266978e-06, "loss": 0.0596, "step": 23329 }, { "epoch": 7.18, "learning_rate": 3.900697911666288e-06, "loss": 0.0415, "step": 23330 }, { "epoch": 7.18, "learning_rate": 3.8999082286372935e-06, "loss": 0.0705, "step": 23331 }, { "epoch": 7.18, "learning_rate": 3.899118606187832e-06, "loss": 0.0683, "step": 23332 }, { "epoch": 7.18, "learning_rate": 3.898329044325745e-06, "loss": 0.046, "step": 23333 }, { "epoch": 7.18, "learning_rate": 3.897539543058875e-06, "loss": 0.036, "step": 23334 }, { "epoch": 7.18, "learning_rate": 3.896750102395057e-06, "loss": 0.046, "step": 23335 }, { "epoch": 7.18, "learning_rate": 3.895960722342144e-06, "loss": 0.0775, "step": 23336 }, { "epoch": 7.18, "learning_rate": 3.8951714029079615e-06, "loss": 0.0516, "step": 23337 }, { "epoch": 7.18, "learning_rate": 3.8943821441003505e-06, "loss": 0.0295, "step": 23338 }, { "epoch": 7.18, "learning_rate": 3.8935929459271585e-06, "loss": 0.0324, "step": 23339 }, { "epoch": 7.18, "learning_rate": 3.8928038083962095e-06, "loss": 0.1186, "step": 23340 }, { "epoch": 7.18, "learning_rate": 3.892014731515351e-06, "loss": 0.0513, "step": 23341 }, { "epoch": 7.18, "learning_rate": 3.891225715292416e-06, "loss": 0.0278, "step": 23342 }, { "epoch": 7.18, "learning_rate": 3.890436759735239e-06, "loss": 0.0326, "step": 23343 }, { "epoch": 7.18, "learning_rate": 3.889647864851655e-06, "loss": 0.0341, "step": 23344 }, { "epoch": 7.18, "learning_rate": 3.888859030649498e-06, "loss": 0.0671, "step": 23345 }, { "epoch": 7.18, "learning_rate": 3.888070257136608e-06, "loss": 0.0738, "step": 23346 }, { "epoch": 7.18, "learning_rate": 3.88728154432081e-06, "loss": 0.0654, "step": 23347 }, { "epoch": 7.18, "learning_rate": 3.886492892209936e-06, "loss": 0.0535, "step": 23348 }, { "epoch": 7.18, "learning_rate": 3.885704300811827e-06, "loss": 0.0387, "step": 23349 }, { "epoch": 7.18, "learning_rate": 3.884915770134309e-06, "loss": 0.0753, "step": 23350 }, { "epoch": 7.18, "learning_rate": 3.884127300185212e-06, "loss": 0.0241, "step": 23351 }, { "epoch": 7.18, "learning_rate": 3.883338890972365e-06, "loss": 0.0999, "step": 23352 }, { "epoch": 7.19, "learning_rate": 3.882550542503607e-06, "loss": 0.0473, "step": 23353 }, { "epoch": 7.19, "learning_rate": 3.881762254786756e-06, "loss": 0.0421, "step": 23354 }, { "epoch": 7.19, "learning_rate": 3.880974027829641e-06, "loss": 0.0442, "step": 23355 }, { "epoch": 7.19, "learning_rate": 3.8801858616400985e-06, "loss": 0.064, "step": 23356 }, { "epoch": 7.19, "learning_rate": 3.879397756225949e-06, "loss": 0.0775, "step": 23357 }, { "epoch": 7.19, "learning_rate": 3.878609711595022e-06, "loss": 0.069, "step": 23358 }, { "epoch": 7.19, "learning_rate": 3.877821727755141e-06, "loss": 0.0641, "step": 23359 }, { "epoch": 7.19, "learning_rate": 3.877033804714133e-06, "loss": 0.1086, "step": 23360 }, { "epoch": 7.19, "learning_rate": 3.876245942479822e-06, "loss": 0.0384, "step": 23361 }, { "epoch": 7.19, "learning_rate": 3.87545814106003e-06, "loss": 0.0353, "step": 23362 }, { "epoch": 7.19, "learning_rate": 3.874670400462589e-06, "loss": 0.042, "step": 23363 }, { "epoch": 7.19, "learning_rate": 3.8738827206953135e-06, "loss": 0.0402, "step": 23364 }, { "epoch": 7.19, "learning_rate": 3.873095101766025e-06, "loss": 0.0456, "step": 23365 }, { "epoch": 7.19, "learning_rate": 3.872307543682553e-06, "loss": 0.0636, "step": 23366 }, { "epoch": 7.19, "learning_rate": 3.871520046452714e-06, "loss": 0.0552, "step": 23367 }, { "epoch": 7.19, "learning_rate": 3.870732610084329e-06, "loss": 0.0797, "step": 23368 }, { "epoch": 7.19, "learning_rate": 3.869945234585217e-06, "loss": 0.059, "step": 23369 }, { "epoch": 7.19, "learning_rate": 3.8691579199632e-06, "loss": 0.0292, "step": 23370 }, { "epoch": 7.19, "learning_rate": 3.8683706662260945e-06, "loss": 0.0308, "step": 23371 }, { "epoch": 7.19, "learning_rate": 3.867583473381716e-06, "loss": 0.0406, "step": 23372 }, { "epoch": 7.19, "learning_rate": 3.86679634143789e-06, "loss": 0.0537, "step": 23373 }, { "epoch": 7.19, "learning_rate": 3.866009270402431e-06, "loss": 0.0666, "step": 23374 }, { "epoch": 7.19, "learning_rate": 3.865222260283146e-06, "loss": 0.0711, "step": 23375 }, { "epoch": 7.19, "learning_rate": 3.8644353110878615e-06, "loss": 0.0598, "step": 23376 }, { "epoch": 7.19, "learning_rate": 3.8636484228243895e-06, "loss": 0.0738, "step": 23377 }, { "epoch": 7.19, "learning_rate": 3.862861595500544e-06, "loss": 0.0421, "step": 23378 }, { "epoch": 7.19, "learning_rate": 3.862074829124139e-06, "loss": 0.0211, "step": 23379 }, { "epoch": 7.19, "learning_rate": 3.861288123702989e-06, "loss": 0.0286, "step": 23380 }, { "epoch": 7.19, "learning_rate": 3.860501479244905e-06, "loss": 0.046, "step": 23381 }, { "epoch": 7.19, "learning_rate": 3.859714895757695e-06, "loss": 0.0736, "step": 23382 }, { "epoch": 7.19, "learning_rate": 3.858928373249179e-06, "loss": 0.0192, "step": 23383 }, { "epoch": 7.19, "learning_rate": 3.858141911727168e-06, "loss": 0.0605, "step": 23384 }, { "epoch": 7.2, "learning_rate": 3.857355511199461e-06, "loss": 0.0463, "step": 23385 }, { "epoch": 7.2, "learning_rate": 3.856569171673879e-06, "loss": 0.0534, "step": 23386 }, { "epoch": 7.2, "learning_rate": 3.855782893158227e-06, "loss": 0.0358, "step": 23387 }, { "epoch": 7.2, "learning_rate": 3.854996675660313e-06, "loss": 0.0335, "step": 23388 }, { "epoch": 7.2, "learning_rate": 3.854210519187946e-06, "loss": 0.0475, "step": 23389 }, { "epoch": 7.2, "learning_rate": 3.853424423748928e-06, "loss": 0.0304, "step": 23390 }, { "epoch": 7.2, "learning_rate": 3.852638389351079e-06, "loss": 0.021, "step": 23391 }, { "epoch": 7.2, "learning_rate": 3.8518524160021876e-06, "loss": 0.0414, "step": 23392 }, { "epoch": 7.2, "learning_rate": 3.851066503710072e-06, "loss": 0.0275, "step": 23393 }, { "epoch": 7.2, "learning_rate": 3.850280652482534e-06, "loss": 0.0818, "step": 23394 }, { "epoch": 7.2, "learning_rate": 3.849494862327376e-06, "loss": 0.0518, "step": 23395 }, { "epoch": 7.2, "learning_rate": 3.848709133252403e-06, "loss": 0.0678, "step": 23396 }, { "epoch": 7.2, "learning_rate": 3.847923465265418e-06, "loss": 0.0624, "step": 23397 }, { "epoch": 7.2, "learning_rate": 3.847137858374222e-06, "loss": 0.0576, "step": 23398 }, { "epoch": 7.2, "learning_rate": 3.846352312586618e-06, "loss": 0.0324, "step": 23399 }, { "epoch": 7.2, "learning_rate": 3.845566827910403e-06, "loss": 0.0661, "step": 23400 }, { "epoch": 7.2, "learning_rate": 3.84478140435339e-06, "loss": 0.0589, "step": 23401 }, { "epoch": 7.2, "learning_rate": 3.843996041923361e-06, "loss": 0.0473, "step": 23402 }, { "epoch": 7.2, "learning_rate": 3.84321074062813e-06, "loss": 0.0552, "step": 23403 }, { "epoch": 7.2, "learning_rate": 3.842425500475491e-06, "loss": 0.0504, "step": 23404 }, { "epoch": 7.2, "learning_rate": 3.84164032147324e-06, "loss": 0.0682, "step": 23405 }, { "epoch": 7.2, "learning_rate": 3.840855203629178e-06, "loss": 0.036, "step": 23406 }, { "epoch": 7.2, "learning_rate": 3.840070146951096e-06, "loss": 0.0336, "step": 23407 }, { "epoch": 7.2, "learning_rate": 3.839285151446802e-06, "loss": 0.0614, "step": 23408 }, { "epoch": 7.2, "learning_rate": 3.83850021712408e-06, "loss": 0.0372, "step": 23409 }, { "epoch": 7.2, "learning_rate": 3.837715343990727e-06, "loss": 0.0502, "step": 23410 }, { "epoch": 7.2, "learning_rate": 3.836930532054543e-06, "loss": 0.0434, "step": 23411 }, { "epoch": 7.2, "learning_rate": 3.836145781323318e-06, "loss": 0.0317, "step": 23412 }, { "epoch": 7.2, "learning_rate": 3.8353610918048475e-06, "loss": 0.0622, "step": 23413 }, { "epoch": 7.2, "learning_rate": 3.8345764635069225e-06, "loss": 0.0394, "step": 23414 }, { "epoch": 7.2, "learning_rate": 3.833791896437336e-06, "loss": 0.0521, "step": 23415 }, { "epoch": 7.2, "learning_rate": 3.833007390603878e-06, "loss": 0.0396, "step": 23416 }, { "epoch": 7.2, "learning_rate": 3.832222946014336e-06, "loss": 0.0134, "step": 23417 }, { "epoch": 7.21, "learning_rate": 3.831438562676513e-06, "loss": 0.0225, "step": 23418 }, { "epoch": 7.21, "learning_rate": 3.830654240598186e-06, "loss": 0.0427, "step": 23419 }, { "epoch": 7.21, "learning_rate": 3.829869979787145e-06, "loss": 0.0403, "step": 23420 }, { "epoch": 7.21, "learning_rate": 3.829085780251185e-06, "loss": 0.0637, "step": 23421 }, { "epoch": 7.21, "learning_rate": 3.828301641998091e-06, "loss": 0.0556, "step": 23422 }, { "epoch": 7.21, "learning_rate": 3.8275175650356485e-06, "loss": 0.0323, "step": 23423 }, { "epoch": 7.21, "learning_rate": 3.826733549371647e-06, "loss": 0.0536, "step": 23424 }, { "epoch": 7.21, "learning_rate": 3.825949595013871e-06, "loss": 0.0688, "step": 23425 }, { "epoch": 7.21, "learning_rate": 3.825165701970104e-06, "loss": 0.0553, "step": 23426 }, { "epoch": 7.21, "learning_rate": 3.82438187024813e-06, "loss": 0.0225, "step": 23427 }, { "epoch": 7.21, "learning_rate": 3.823598099855738e-06, "loss": 0.0451, "step": 23428 }, { "epoch": 7.21, "learning_rate": 3.822814390800713e-06, "loss": 0.067, "step": 23429 }, { "epoch": 7.21, "learning_rate": 3.822030743090828e-06, "loss": 0.0336, "step": 23430 }, { "epoch": 7.21, "learning_rate": 3.821247156733874e-06, "loss": 0.0501, "step": 23431 }, { "epoch": 7.21, "learning_rate": 3.820463631737631e-06, "loss": 0.0501, "step": 23432 }, { "epoch": 7.21, "learning_rate": 3.81968016810988e-06, "loss": 0.0378, "step": 23433 }, { "epoch": 7.21, "learning_rate": 3.8188967658584e-06, "loss": 0.0963, "step": 23434 }, { "epoch": 7.21, "learning_rate": 3.818113424990968e-06, "loss": 0.0482, "step": 23435 }, { "epoch": 7.21, "learning_rate": 3.817330145515374e-06, "loss": 0.0559, "step": 23436 }, { "epoch": 7.21, "learning_rate": 3.8165469274393825e-06, "loss": 0.0767, "step": 23437 }, { "epoch": 7.21, "learning_rate": 3.815763770770784e-06, "loss": 0.0411, "step": 23438 }, { "epoch": 7.21, "learning_rate": 3.8149806755173535e-06, "loss": 0.0464, "step": 23439 }, { "epoch": 7.21, "learning_rate": 3.8141976416868574e-06, "loss": 0.0634, "step": 23440 }, { "epoch": 7.21, "learning_rate": 3.8134146692870844e-06, "loss": 0.0505, "step": 23441 }, { "epoch": 7.21, "learning_rate": 3.8126317583258057e-06, "loss": 0.0755, "step": 23442 }, { "epoch": 7.21, "learning_rate": 3.811848908810795e-06, "loss": 0.0523, "step": 23443 }, { "epoch": 7.21, "learning_rate": 3.811066120749828e-06, "loss": 0.0498, "step": 23444 }, { "epoch": 7.21, "learning_rate": 3.8102833941506743e-06, "loss": 0.055, "step": 23445 }, { "epoch": 7.21, "learning_rate": 3.8095007290211196e-06, "loss": 0.0441, "step": 23446 }, { "epoch": 7.21, "learning_rate": 3.8087181253689198e-06, "loss": 0.0815, "step": 23447 }, { "epoch": 7.21, "learning_rate": 3.807935583201858e-06, "loss": 0.0472, "step": 23448 }, { "epoch": 7.21, "learning_rate": 3.807153102527704e-06, "loss": 0.0272, "step": 23449 }, { "epoch": 7.22, "learning_rate": 3.806370683354227e-06, "loss": 0.0266, "step": 23450 }, { "epoch": 7.22, "learning_rate": 3.8055883256891968e-06, "loss": 0.0527, "step": 23451 }, { "epoch": 7.22, "learning_rate": 3.80480602954038e-06, "loss": 0.0617, "step": 23452 }, { "epoch": 7.22, "learning_rate": 3.8040237949155567e-06, "loss": 0.0739, "step": 23453 }, { "epoch": 7.22, "learning_rate": 3.8032416218224834e-06, "loss": 0.0451, "step": 23454 }, { "epoch": 7.22, "learning_rate": 3.8024595102689286e-06, "loss": 0.0587, "step": 23455 }, { "epoch": 7.22, "learning_rate": 3.801677460262666e-06, "loss": 0.0501, "step": 23456 }, { "epoch": 7.22, "learning_rate": 3.800895471811459e-06, "loss": 0.0761, "step": 23457 }, { "epoch": 7.22, "learning_rate": 3.800113544923073e-06, "loss": 0.0281, "step": 23458 }, { "epoch": 7.22, "learning_rate": 3.7993316796052737e-06, "loss": 0.0452, "step": 23459 }, { "epoch": 7.22, "learning_rate": 3.7985498758658255e-06, "loss": 0.1148, "step": 23460 }, { "epoch": 7.22, "learning_rate": 3.7977681337124917e-06, "loss": 0.0187, "step": 23461 }, { "epoch": 7.22, "learning_rate": 3.7969864531530344e-06, "loss": 0.0874, "step": 23462 }, { "epoch": 7.22, "learning_rate": 3.796204834195224e-06, "loss": 0.059, "step": 23463 }, { "epoch": 7.22, "learning_rate": 3.7954232768468147e-06, "loss": 0.0649, "step": 23464 }, { "epoch": 7.22, "learning_rate": 3.7946417811155665e-06, "loss": 0.0566, "step": 23465 }, { "epoch": 7.22, "learning_rate": 3.7938603470092484e-06, "loss": 0.0526, "step": 23466 }, { "epoch": 7.22, "learning_rate": 3.7930789745356165e-06, "loss": 0.0208, "step": 23467 }, { "epoch": 7.22, "learning_rate": 3.792297663702431e-06, "loss": 0.0646, "step": 23468 }, { "epoch": 7.22, "learning_rate": 3.791516414517451e-06, "loss": 0.0549, "step": 23469 }, { "epoch": 7.22, "learning_rate": 3.7907352269884346e-06, "loss": 0.0888, "step": 23470 }, { "epoch": 7.22, "learning_rate": 3.78995410112314e-06, "loss": 0.0463, "step": 23471 }, { "epoch": 7.22, "learning_rate": 3.789173036929321e-06, "loss": 0.0698, "step": 23472 }, { "epoch": 7.22, "learning_rate": 3.7883920344147418e-06, "loss": 0.0674, "step": 23473 }, { "epoch": 7.22, "learning_rate": 3.787611093587158e-06, "loss": 0.026, "step": 23474 }, { "epoch": 7.22, "learning_rate": 3.7868302144543146e-06, "loss": 0.0475, "step": 23475 }, { "epoch": 7.22, "learning_rate": 3.7860493970239763e-06, "loss": 0.0628, "step": 23476 }, { "epoch": 7.22, "learning_rate": 3.7852686413038943e-06, "loss": 0.0396, "step": 23477 }, { "epoch": 7.22, "learning_rate": 3.7844879473018228e-06, "loss": 0.0556, "step": 23478 }, { "epoch": 7.22, "learning_rate": 3.783707315025514e-06, "loss": 0.033, "step": 23479 }, { "epoch": 7.22, "learning_rate": 3.782926744482721e-06, "loss": 0.0593, "step": 23480 }, { "epoch": 7.22, "learning_rate": 3.7821462356811956e-06, "loss": 0.0912, "step": 23481 }, { "epoch": 7.22, "learning_rate": 3.7813657886286847e-06, "loss": 0.065, "step": 23482 }, { "epoch": 7.23, "learning_rate": 3.7805854033329457e-06, "loss": 0.0638, "step": 23483 }, { "epoch": 7.23, "learning_rate": 3.7798050798017303e-06, "loss": 0.0401, "step": 23484 }, { "epoch": 7.23, "learning_rate": 3.7790248180427758e-06, "loss": 0.0527, "step": 23485 }, { "epoch": 7.23, "learning_rate": 3.778244618063841e-06, "loss": 0.0594, "step": 23486 }, { "epoch": 7.23, "learning_rate": 3.7774644798726725e-06, "loss": 0.0342, "step": 23487 }, { "epoch": 7.23, "learning_rate": 3.7766844034770155e-06, "loss": 0.0239, "step": 23488 }, { "epoch": 7.23, "learning_rate": 3.775904388884618e-06, "loss": 0.029, "step": 23489 }, { "epoch": 7.23, "learning_rate": 3.7751244361032234e-06, "loss": 0.0699, "step": 23490 }, { "epoch": 7.23, "learning_rate": 3.774344545140587e-06, "loss": 0.0325, "step": 23491 }, { "epoch": 7.23, "learning_rate": 3.77356471600444e-06, "loss": 0.0548, "step": 23492 }, { "epoch": 7.23, "learning_rate": 3.772784948702537e-06, "loss": 0.0376, "step": 23493 }, { "epoch": 7.23, "learning_rate": 3.772005243242619e-06, "loss": 0.0515, "step": 23494 }, { "epoch": 7.23, "learning_rate": 3.771225599632429e-06, "loss": 0.0523, "step": 23495 }, { "epoch": 7.23, "learning_rate": 3.770446017879709e-06, "loss": 0.0243, "step": 23496 }, { "epoch": 7.23, "learning_rate": 3.7696664979922004e-06, "loss": 0.0305, "step": 23497 }, { "epoch": 7.23, "learning_rate": 3.768887039977647e-06, "loss": 0.0473, "step": 23498 }, { "epoch": 7.23, "learning_rate": 3.768107643843787e-06, "loss": 0.0593, "step": 23499 }, { "epoch": 7.23, "learning_rate": 3.7673283095983593e-06, "loss": 0.034, "step": 23500 }, { "epoch": 7.23, "learning_rate": 3.766549037249112e-06, "loss": 0.0415, "step": 23501 }, { "epoch": 7.23, "learning_rate": 3.76576982680377e-06, "loss": 0.0719, "step": 23502 }, { "epoch": 7.23, "learning_rate": 3.764990678270084e-06, "loss": 0.0402, "step": 23503 }, { "epoch": 7.23, "learning_rate": 3.7642115916557866e-06, "loss": 0.017, "step": 23504 }, { "epoch": 7.23, "learning_rate": 3.7634325669686144e-06, "loss": 0.0574, "step": 23505 }, { "epoch": 7.23, "learning_rate": 3.7626536042163054e-06, "loss": 0.0426, "step": 23506 }, { "epoch": 7.23, "learning_rate": 3.7618747034065904e-06, "loss": 0.0542, "step": 23507 }, { "epoch": 7.23, "learning_rate": 3.761095864547216e-06, "loss": 0.0507, "step": 23508 }, { "epoch": 7.23, "learning_rate": 3.760317087645907e-06, "loss": 0.0212, "step": 23509 }, { "epoch": 7.23, "learning_rate": 3.7595383727103964e-06, "loss": 0.0414, "step": 23510 }, { "epoch": 7.23, "learning_rate": 3.758759719748424e-06, "loss": 0.061, "step": 23511 }, { "epoch": 7.23, "learning_rate": 3.7579811287677193e-06, "loss": 0.0541, "step": 23512 }, { "epoch": 7.23, "learning_rate": 3.757202599776015e-06, "loss": 0.0558, "step": 23513 }, { "epoch": 7.23, "learning_rate": 3.7564241327810436e-06, "loss": 0.0763, "step": 23514 }, { "epoch": 7.24, "learning_rate": 3.7556457277905334e-06, "loss": 0.0405, "step": 23515 }, { "epoch": 7.24, "learning_rate": 3.754867384812216e-06, "loss": 0.0721, "step": 23516 }, { "epoch": 7.24, "learning_rate": 3.7540891038538195e-06, "loss": 0.0197, "step": 23517 }, { "epoch": 7.24, "learning_rate": 3.7533108849230802e-06, "loss": 0.0605, "step": 23518 }, { "epoch": 7.24, "learning_rate": 3.752532728027717e-06, "loss": 0.0425, "step": 23519 }, { "epoch": 7.24, "learning_rate": 3.7517546331754594e-06, "loss": 0.0601, "step": 23520 }, { "epoch": 7.24, "learning_rate": 3.7509766003740387e-06, "loss": 0.0417, "step": 23521 }, { "epoch": 7.24, "learning_rate": 3.750198629631181e-06, "loss": 0.0474, "step": 23522 }, { "epoch": 7.24, "learning_rate": 3.7494207209546106e-06, "loss": 0.0264, "step": 23523 }, { "epoch": 7.24, "learning_rate": 3.748642874352053e-06, "loss": 0.0397, "step": 23524 }, { "epoch": 7.24, "learning_rate": 3.747865089831233e-06, "loss": 0.0526, "step": 23525 }, { "epoch": 7.24, "learning_rate": 3.7470873673998755e-06, "loss": 0.0364, "step": 23526 }, { "epoch": 7.24, "learning_rate": 3.7463097070656995e-06, "loss": 0.0603, "step": 23527 }, { "epoch": 7.24, "learning_rate": 3.745532108836435e-06, "loss": 0.0478, "step": 23528 }, { "epoch": 7.24, "learning_rate": 3.7447545727198044e-06, "loss": 0.0744, "step": 23529 }, { "epoch": 7.24, "learning_rate": 3.743977098723518e-06, "loss": 0.0646, "step": 23530 }, { "epoch": 7.24, "learning_rate": 3.743199686855309e-06, "loss": 0.0561, "step": 23531 }, { "epoch": 7.24, "learning_rate": 3.742422337122894e-06, "loss": 0.0576, "step": 23532 }, { "epoch": 7.24, "learning_rate": 3.7416450495339917e-06, "loss": 0.0518, "step": 23533 }, { "epoch": 7.24, "learning_rate": 3.7408678240963227e-06, "loss": 0.077, "step": 23534 }, { "epoch": 7.24, "learning_rate": 3.7400906608176002e-06, "loss": 0.0626, "step": 23535 }, { "epoch": 7.24, "learning_rate": 3.7393135597055552e-06, "loss": 0.0426, "step": 23536 }, { "epoch": 7.24, "learning_rate": 3.738536520767888e-06, "loss": 0.0336, "step": 23537 }, { "epoch": 7.24, "learning_rate": 3.7377595440123284e-06, "loss": 0.025, "step": 23538 }, { "epoch": 7.24, "learning_rate": 3.7369826294465904e-06, "loss": 0.062, "step": 23539 }, { "epoch": 7.24, "learning_rate": 3.736205777078381e-06, "loss": 0.0421, "step": 23540 }, { "epoch": 7.24, "learning_rate": 3.7354289869154247e-06, "loss": 0.0702, "step": 23541 }, { "epoch": 7.24, "learning_rate": 3.7346522589654323e-06, "loss": 0.0689, "step": 23542 }, { "epoch": 7.24, "learning_rate": 3.733875593236117e-06, "loss": 0.0649, "step": 23543 }, { "epoch": 7.24, "learning_rate": 3.7330989897351923e-06, "loss": 0.0674, "step": 23544 }, { "epoch": 7.24, "learning_rate": 3.732322448470366e-06, "loss": 0.059, "step": 23545 }, { "epoch": 7.24, "learning_rate": 3.7315459694493615e-06, "loss": 0.0762, "step": 23546 }, { "epoch": 7.24, "learning_rate": 3.730769552679876e-06, "loss": 0.06, "step": 23547 }, { "epoch": 7.25, "learning_rate": 3.72999319816963e-06, "loss": 0.0315, "step": 23548 }, { "epoch": 7.25, "learning_rate": 3.7292169059263305e-06, "loss": 0.0455, "step": 23549 }, { "epoch": 7.25, "learning_rate": 3.7284406759576864e-06, "loss": 0.0631, "step": 23550 }, { "epoch": 7.25, "learning_rate": 3.727664508271407e-06, "loss": 0.0323, "step": 23551 }, { "epoch": 7.25, "learning_rate": 3.726888402875195e-06, "loss": 0.0563, "step": 23552 }, { "epoch": 7.25, "learning_rate": 3.72611235977677e-06, "loss": 0.0513, "step": 23553 }, { "epoch": 7.25, "learning_rate": 3.7253363789838247e-06, "loss": 0.0755, "step": 23554 }, { "epoch": 7.25, "learning_rate": 3.7245604605040765e-06, "loss": 0.0198, "step": 23555 }, { "epoch": 7.25, "learning_rate": 3.7237846043452285e-06, "loss": 0.0315, "step": 23556 }, { "epoch": 7.25, "learning_rate": 3.7230088105149776e-06, "loss": 0.0584, "step": 23557 }, { "epoch": 7.25, "learning_rate": 3.722233079021037e-06, "loss": 0.0449, "step": 23558 }, { "epoch": 7.25, "learning_rate": 3.721457409871109e-06, "loss": 0.048, "step": 23559 }, { "epoch": 7.25, "learning_rate": 3.7206818030728943e-06, "loss": 0.0547, "step": 23560 }, { "epoch": 7.25, "learning_rate": 3.7199062586340963e-06, "loss": 0.0444, "step": 23561 }, { "epoch": 7.25, "learning_rate": 3.719130776562414e-06, "loss": 0.0579, "step": 23562 }, { "epoch": 7.25, "learning_rate": 3.7183553568655594e-06, "loss": 0.074, "step": 23563 }, { "epoch": 7.25, "learning_rate": 3.7175799995512175e-06, "loss": 0.0478, "step": 23564 }, { "epoch": 7.25, "learning_rate": 3.7168047046271007e-06, "loss": 0.0662, "step": 23565 }, { "epoch": 7.25, "learning_rate": 3.7160294721009026e-06, "loss": 0.0458, "step": 23566 }, { "epoch": 7.25, "learning_rate": 3.7152543019803246e-06, "loss": 0.0552, "step": 23567 }, { "epoch": 7.25, "learning_rate": 3.7144791942730617e-06, "loss": 0.0161, "step": 23568 }, { "epoch": 7.25, "learning_rate": 3.713704148986812e-06, "loss": 0.0535, "step": 23569 }, { "epoch": 7.25, "learning_rate": 3.71292916612928e-06, "loss": 0.0475, "step": 23570 }, { "epoch": 7.25, "learning_rate": 3.7121542457081515e-06, "loss": 0.0341, "step": 23571 }, { "epoch": 7.25, "learning_rate": 3.7113793877311233e-06, "loss": 0.0614, "step": 23572 }, { "epoch": 7.25, "learning_rate": 3.7106045922058974e-06, "loss": 0.0331, "step": 23573 }, { "epoch": 7.25, "learning_rate": 3.7098298591401637e-06, "loss": 0.064, "step": 23574 }, { "epoch": 7.25, "learning_rate": 3.709055188541617e-06, "loss": 0.0619, "step": 23575 }, { "epoch": 7.25, "learning_rate": 3.708280580417951e-06, "loss": 0.0757, "step": 23576 }, { "epoch": 7.25, "learning_rate": 3.7075060347768564e-06, "loss": 0.0533, "step": 23577 }, { "epoch": 7.25, "learning_rate": 3.706731551626026e-06, "loss": 0.0393, "step": 23578 }, { "epoch": 7.25, "learning_rate": 3.705957130973149e-06, "loss": 0.0336, "step": 23579 }, { "epoch": 7.26, "learning_rate": 3.7051827728259248e-06, "loss": 0.0607, "step": 23580 }, { "epoch": 7.26, "learning_rate": 3.704408477192034e-06, "loss": 0.0724, "step": 23581 }, { "epoch": 7.26, "learning_rate": 3.703634244079165e-06, "loss": 0.0637, "step": 23582 }, { "epoch": 7.26, "learning_rate": 3.702860073495015e-06, "loss": 0.0761, "step": 23583 }, { "epoch": 7.26, "learning_rate": 3.7020859654472676e-06, "loss": 0.0713, "step": 23584 }, { "epoch": 7.26, "learning_rate": 3.701311919943611e-06, "loss": 0.0357, "step": 23585 }, { "epoch": 7.26, "learning_rate": 3.700537936991733e-06, "loss": 0.0222, "step": 23586 }, { "epoch": 7.26, "learning_rate": 3.6997640165993178e-06, "loss": 0.0646, "step": 23587 }, { "epoch": 7.26, "learning_rate": 3.6989901587740527e-06, "loss": 0.0573, "step": 23588 }, { "epoch": 7.26, "learning_rate": 3.698216363523619e-06, "loss": 0.0492, "step": 23589 }, { "epoch": 7.26, "learning_rate": 3.697442630855709e-06, "loss": 0.0694, "step": 23590 }, { "epoch": 7.26, "learning_rate": 3.696668960778005e-06, "loss": 0.0807, "step": 23591 }, { "epoch": 7.26, "learning_rate": 3.69589535329818e-06, "loss": 0.0541, "step": 23592 }, { "epoch": 7.26, "learning_rate": 3.695121808423928e-06, "loss": 0.0607, "step": 23593 }, { "epoch": 7.26, "learning_rate": 3.6943483261629277e-06, "loss": 0.061, "step": 23594 }, { "epoch": 7.26, "learning_rate": 3.69357490652286e-06, "loss": 0.0625, "step": 23595 }, { "epoch": 7.26, "learning_rate": 3.6928015495114055e-06, "loss": 0.0451, "step": 23596 }, { "epoch": 7.26, "learning_rate": 3.692028255136244e-06, "loss": 0.0588, "step": 23597 }, { "epoch": 7.26, "learning_rate": 3.691255023405056e-06, "loss": 0.0389, "step": 23598 }, { "epoch": 7.26, "learning_rate": 3.6904818543255174e-06, "loss": 0.0163, "step": 23599 }, { "epoch": 7.26, "learning_rate": 3.6897087479053106e-06, "loss": 0.0341, "step": 23600 }, { "epoch": 7.26, "learning_rate": 3.688935704152117e-06, "loss": 0.0524, "step": 23601 }, { "epoch": 7.26, "learning_rate": 3.6881627230735996e-06, "loss": 0.0763, "step": 23602 }, { "epoch": 7.26, "learning_rate": 3.687389804677448e-06, "loss": 0.0333, "step": 23603 }, { "epoch": 7.26, "learning_rate": 3.6866169489713334e-06, "loss": 0.0882, "step": 23604 }, { "epoch": 7.26, "learning_rate": 3.685844155962931e-06, "loss": 0.0519, "step": 23605 }, { "epoch": 7.26, "learning_rate": 3.685071425659915e-06, "loss": 0.0708, "step": 23606 }, { "epoch": 7.26, "learning_rate": 3.6842987580699573e-06, "loss": 0.0459, "step": 23607 }, { "epoch": 7.26, "learning_rate": 3.6835261532007395e-06, "loss": 0.0534, "step": 23608 }, { "epoch": 7.26, "learning_rate": 3.6827536110599226e-06, "loss": 0.0462, "step": 23609 }, { "epoch": 7.26, "learning_rate": 3.6819811316551867e-06, "loss": 0.0422, "step": 23610 }, { "epoch": 7.26, "learning_rate": 3.681208714994202e-06, "loss": 0.0454, "step": 23611 }, { "epoch": 7.26, "learning_rate": 3.680436361084638e-06, "loss": 0.0543, "step": 23612 }, { "epoch": 7.27, "learning_rate": 3.6796640699341656e-06, "loss": 0.0396, "step": 23613 }, { "epoch": 7.27, "learning_rate": 3.6788918415504537e-06, "loss": 0.035, "step": 23614 }, { "epoch": 7.27, "learning_rate": 3.678119675941171e-06, "loss": 0.036, "step": 23615 }, { "epoch": 7.27, "learning_rate": 3.677347573113986e-06, "loss": 0.0519, "step": 23616 }, { "epoch": 7.27, "learning_rate": 3.676575533076565e-06, "loss": 0.0488, "step": 23617 }, { "epoch": 7.27, "learning_rate": 3.675803555836582e-06, "loss": 0.015, "step": 23618 }, { "epoch": 7.27, "learning_rate": 3.675031641401693e-06, "loss": 0.0655, "step": 23619 }, { "epoch": 7.27, "learning_rate": 3.6742597897795717e-06, "loss": 0.0744, "step": 23620 }, { "epoch": 7.27, "learning_rate": 3.6734880009778806e-06, "loss": 0.0426, "step": 23621 }, { "epoch": 7.27, "learning_rate": 3.6727162750042833e-06, "loss": 0.0492, "step": 23622 }, { "epoch": 7.27, "learning_rate": 3.6719446118664458e-06, "loss": 0.0746, "step": 23623 }, { "epoch": 7.27, "learning_rate": 3.6711730115720277e-06, "loss": 0.1021, "step": 23624 }, { "epoch": 7.27, "learning_rate": 3.6704014741286996e-06, "loss": 0.0336, "step": 23625 }, { "epoch": 7.27, "learning_rate": 3.6696299995441167e-06, "loss": 0.0631, "step": 23626 }, { "epoch": 7.27, "learning_rate": 3.6688585878259375e-06, "loss": 0.0604, "step": 23627 }, { "epoch": 7.27, "learning_rate": 3.668087238981831e-06, "loss": 0.0578, "step": 23628 }, { "epoch": 7.27, "learning_rate": 3.667315953019455e-06, "loss": 0.0623, "step": 23629 }, { "epoch": 7.27, "learning_rate": 3.6665447299464665e-06, "loss": 0.0367, "step": 23630 }, { "epoch": 7.27, "learning_rate": 3.6657735697705267e-06, "loss": 0.0189, "step": 23631 }, { "epoch": 7.27, "learning_rate": 3.6650024724992926e-06, "loss": 0.0223, "step": 23632 }, { "epoch": 7.27, "learning_rate": 3.664231438140423e-06, "loss": 0.0362, "step": 23633 }, { "epoch": 7.27, "learning_rate": 3.6634604667015717e-06, "loss": 0.0426, "step": 23634 }, { "epoch": 7.27, "learning_rate": 3.662689558190403e-06, "loss": 0.0783, "step": 23635 }, { "epoch": 7.27, "learning_rate": 3.661918712614566e-06, "loss": 0.051, "step": 23636 }, { "epoch": 7.27, "learning_rate": 3.6611479299817134e-06, "loss": 0.0394, "step": 23637 }, { "epoch": 7.27, "learning_rate": 3.660377210299506e-06, "loss": 0.0515, "step": 23638 }, { "epoch": 7.27, "learning_rate": 3.6596065535755975e-06, "loss": 0.0456, "step": 23639 }, { "epoch": 7.27, "learning_rate": 3.658835959817638e-06, "loss": 0.0576, "step": 23640 }, { "epoch": 7.27, "learning_rate": 3.658065429033283e-06, "loss": 0.0351, "step": 23641 }, { "epoch": 7.27, "learning_rate": 3.6572949612301824e-06, "loss": 0.0311, "step": 23642 }, { "epoch": 7.27, "learning_rate": 3.656524556415989e-06, "loss": 0.0582, "step": 23643 }, { "epoch": 7.27, "learning_rate": 3.6557542145983495e-06, "loss": 0.0559, "step": 23644 }, { "epoch": 7.28, "learning_rate": 3.6549839357849205e-06, "loss": 0.0619, "step": 23645 }, { "epoch": 7.28, "learning_rate": 3.6542137199833515e-06, "loss": 0.0592, "step": 23646 }, { "epoch": 7.28, "learning_rate": 3.6534435672012836e-06, "loss": 0.0772, "step": 23647 }, { "epoch": 7.28, "learning_rate": 3.6526734774463724e-06, "loss": 0.0519, "step": 23648 }, { "epoch": 7.28, "learning_rate": 3.651903450726264e-06, "loss": 0.0162, "step": 23649 }, { "epoch": 7.28, "learning_rate": 3.651133487048605e-06, "loss": 0.0494, "step": 23650 }, { "epoch": 7.28, "learning_rate": 3.650363586421042e-06, "loss": 0.0461, "step": 23651 }, { "epoch": 7.28, "learning_rate": 3.649593748851217e-06, "loss": 0.066, "step": 23652 }, { "epoch": 7.28, "learning_rate": 3.6488239743467847e-06, "loss": 0.0416, "step": 23653 }, { "epoch": 7.28, "learning_rate": 3.648054262915378e-06, "loss": 0.0841, "step": 23654 }, { "epoch": 7.28, "learning_rate": 3.64728461456465e-06, "loss": 0.0384, "step": 23655 }, { "epoch": 7.28, "learning_rate": 3.6465150293022433e-06, "loss": 0.0381, "step": 23656 }, { "epoch": 7.28, "learning_rate": 3.6457455071357918e-06, "loss": 0.0719, "step": 23657 }, { "epoch": 7.28, "learning_rate": 3.6449760480729467e-06, "loss": 0.0826, "step": 23658 }, { "epoch": 7.28, "learning_rate": 3.644206652121346e-06, "loss": 0.0452, "step": 23659 }, { "epoch": 7.28, "learning_rate": 3.643437319288632e-06, "loss": 0.055, "step": 23660 }, { "epoch": 7.28, "learning_rate": 3.642668049582443e-06, "loss": 0.0318, "step": 23661 }, { "epoch": 7.28, "learning_rate": 3.641898843010415e-06, "loss": 0.0538, "step": 23662 }, { "epoch": 7.28, "learning_rate": 3.6411296995802002e-06, "loss": 0.0575, "step": 23663 }, { "epoch": 7.28, "learning_rate": 3.64036061929942e-06, "loss": 0.0497, "step": 23664 }, { "epoch": 7.28, "learning_rate": 3.639591602175724e-06, "loss": 0.033, "step": 23665 }, { "epoch": 7.28, "learning_rate": 3.6388226482167445e-06, "loss": 0.0507, "step": 23666 }, { "epoch": 7.28, "learning_rate": 3.638053757430119e-06, "loss": 0.0432, "step": 23667 }, { "epoch": 7.28, "learning_rate": 3.6372849298234835e-06, "loss": 0.0099, "step": 23668 }, { "epoch": 7.28, "learning_rate": 3.6365161654044687e-06, "loss": 0.0478, "step": 23669 }, { "epoch": 7.28, "learning_rate": 3.63574746418072e-06, "loss": 0.0667, "step": 23670 }, { "epoch": 7.28, "learning_rate": 3.6349788261598618e-06, "loss": 0.0189, "step": 23671 }, { "epoch": 7.28, "learning_rate": 3.634210251349526e-06, "loss": 0.0276, "step": 23672 }, { "epoch": 7.28, "learning_rate": 3.633441739757353e-06, "loss": 0.0787, "step": 23673 }, { "epoch": 7.28, "learning_rate": 3.632673291390971e-06, "loss": 0.0253, "step": 23674 }, { "epoch": 7.28, "learning_rate": 3.6319049062580114e-06, "loss": 0.0607, "step": 23675 }, { "epoch": 7.28, "learning_rate": 3.631136584366105e-06, "loss": 0.0313, "step": 23676 }, { "epoch": 7.28, "learning_rate": 3.6303683257228827e-06, "loss": 0.0668, "step": 23677 }, { "epoch": 7.29, "learning_rate": 3.629600130335973e-06, "loss": 0.0328, "step": 23678 }, { "epoch": 7.29, "learning_rate": 3.6288319982130014e-06, "loss": 0.0552, "step": 23679 }, { "epoch": 7.29, "learning_rate": 3.6280639293616072e-06, "loss": 0.0272, "step": 23680 }, { "epoch": 7.29, "learning_rate": 3.627295923789407e-06, "loss": 0.0456, "step": 23681 }, { "epoch": 7.29, "learning_rate": 3.626527981504028e-06, "loss": 0.0536, "step": 23682 }, { "epoch": 7.29, "learning_rate": 3.625760102513103e-06, "loss": 0.0614, "step": 23683 }, { "epoch": 7.29, "learning_rate": 3.624992286824256e-06, "loss": 0.0698, "step": 23684 }, { "epoch": 7.29, "learning_rate": 3.62422453444511e-06, "loss": 0.0401, "step": 23685 }, { "epoch": 7.29, "learning_rate": 3.6234568453832906e-06, "loss": 0.0616, "step": 23686 }, { "epoch": 7.29, "learning_rate": 3.6226892196464225e-06, "loss": 0.0561, "step": 23687 }, { "epoch": 7.29, "learning_rate": 3.6219216572421266e-06, "loss": 0.0584, "step": 23688 }, { "epoch": 7.29, "learning_rate": 3.6211541581780252e-06, "loss": 0.0662, "step": 23689 }, { "epoch": 7.29, "learning_rate": 3.620386722461744e-06, "loss": 0.0722, "step": 23690 }, { "epoch": 7.29, "learning_rate": 3.619619350100906e-06, "loss": 0.0232, "step": 23691 }, { "epoch": 7.29, "learning_rate": 3.6188520411031214e-06, "loss": 0.0434, "step": 23692 }, { "epoch": 7.29, "learning_rate": 3.61808479547602e-06, "loss": 0.0513, "step": 23693 }, { "epoch": 7.29, "learning_rate": 3.617317613227219e-06, "loss": 0.0601, "step": 23694 }, { "epoch": 7.29, "learning_rate": 3.616550494364336e-06, "loss": 0.0423, "step": 23695 }, { "epoch": 7.29, "learning_rate": 3.6157834388949907e-06, "loss": 0.0641, "step": 23696 }, { "epoch": 7.29, "learning_rate": 3.615016446826799e-06, "loss": 0.0739, "step": 23697 }, { "epoch": 7.29, "learning_rate": 3.6142495181673787e-06, "loss": 0.0264, "step": 23698 }, { "epoch": 7.29, "learning_rate": 3.6134826529243416e-06, "loss": 0.0407, "step": 23699 }, { "epoch": 7.29, "learning_rate": 3.612715851105313e-06, "loss": 0.0411, "step": 23700 }, { "epoch": 7.29, "learning_rate": 3.6119491127179053e-06, "loss": 0.069, "step": 23701 }, { "epoch": 7.29, "learning_rate": 3.6111824377697225e-06, "loss": 0.0303, "step": 23702 }, { "epoch": 7.29, "learning_rate": 3.6104158262683907e-06, "loss": 0.0703, "step": 23703 }, { "epoch": 7.29, "learning_rate": 3.6096492782215186e-06, "loss": 0.0549, "step": 23704 }, { "epoch": 7.29, "learning_rate": 3.6088827936367177e-06, "loss": 0.069, "step": 23705 }, { "epoch": 7.29, "learning_rate": 3.608116372521602e-06, "loss": 0.0934, "step": 23706 }, { "epoch": 7.29, "learning_rate": 3.607350014883778e-06, "loss": 0.0667, "step": 23707 }, { "epoch": 7.29, "learning_rate": 3.6065837207308663e-06, "loss": 0.0555, "step": 23708 }, { "epoch": 7.29, "learning_rate": 3.6058174900704646e-06, "loss": 0.0228, "step": 23709 }, { "epoch": 7.3, "learning_rate": 3.6050513229101914e-06, "loss": 0.0583, "step": 23710 }, { "epoch": 7.3, "learning_rate": 3.6042852192576516e-06, "loss": 0.0324, "step": 23711 }, { "epoch": 7.3, "learning_rate": 3.603519179120455e-06, "loss": 0.0421, "step": 23712 }, { "epoch": 7.3, "learning_rate": 3.602753202506207e-06, "loss": 0.0528, "step": 23713 }, { "epoch": 7.3, "learning_rate": 3.601987289422516e-06, "loss": 0.0569, "step": 23714 }, { "epoch": 7.3, "learning_rate": 3.6012214398769886e-06, "loss": 0.0371, "step": 23715 }, { "epoch": 7.3, "learning_rate": 3.6004556538772293e-06, "loss": 0.0392, "step": 23716 }, { "epoch": 7.3, "learning_rate": 3.59968993143084e-06, "loss": 0.0577, "step": 23717 }, { "epoch": 7.3, "learning_rate": 3.5989242725454345e-06, "loss": 0.037, "step": 23718 }, { "epoch": 7.3, "learning_rate": 3.5981586772286036e-06, "loss": 0.0514, "step": 23719 }, { "epoch": 7.3, "learning_rate": 3.5973931454879605e-06, "loss": 0.059, "step": 23720 }, { "epoch": 7.3, "learning_rate": 3.5966276773311047e-06, "loss": 0.0532, "step": 23721 }, { "epoch": 7.3, "learning_rate": 3.595862272765638e-06, "loss": 0.0497, "step": 23722 }, { "epoch": 7.3, "learning_rate": 3.5950969317991613e-06, "loss": 0.037, "step": 23723 }, { "epoch": 7.3, "learning_rate": 3.5943316544392703e-06, "loss": 0.0474, "step": 23724 }, { "epoch": 7.3, "learning_rate": 3.5935664406935777e-06, "loss": 0.0984, "step": 23725 }, { "epoch": 7.3, "learning_rate": 3.5928012905696707e-06, "loss": 0.0458, "step": 23726 }, { "epoch": 7.3, "learning_rate": 3.5920362040751476e-06, "loss": 0.0441, "step": 23727 }, { "epoch": 7.3, "learning_rate": 3.591271181217616e-06, "loss": 0.0728, "step": 23728 }, { "epoch": 7.3, "learning_rate": 3.5905062220046672e-06, "loss": 0.0362, "step": 23729 }, { "epoch": 7.3, "learning_rate": 3.5897413264438985e-06, "loss": 0.0451, "step": 23730 }, { "epoch": 7.3, "learning_rate": 3.588976494542906e-06, "loss": 0.0509, "step": 23731 }, { "epoch": 7.3, "learning_rate": 3.5882117263092862e-06, "loss": 0.0584, "step": 23732 }, { "epoch": 7.3, "learning_rate": 3.587447021750633e-06, "loss": 0.0591, "step": 23733 }, { "epoch": 7.3, "learning_rate": 3.5866823808745377e-06, "loss": 0.0344, "step": 23734 }, { "epoch": 7.3, "learning_rate": 3.585917803688603e-06, "loss": 0.0812, "step": 23735 }, { "epoch": 7.3, "learning_rate": 3.5851532902004137e-06, "loss": 0.0393, "step": 23736 }, { "epoch": 7.3, "learning_rate": 3.5843888404175607e-06, "loss": 0.0504, "step": 23737 }, { "epoch": 7.3, "learning_rate": 3.5836244543476417e-06, "loss": 0.0765, "step": 23738 }, { "epoch": 7.3, "learning_rate": 3.582860131998246e-06, "loss": 0.0477, "step": 23739 }, { "epoch": 7.3, "learning_rate": 3.582095873376964e-06, "loss": 0.0565, "step": 23740 }, { "epoch": 7.3, "learning_rate": 3.5813316784913844e-06, "loss": 0.0438, "step": 23741 }, { "epoch": 7.3, "learning_rate": 3.5805675473490964e-06, "loss": 0.0474, "step": 23742 }, { "epoch": 7.31, "learning_rate": 3.57980347995769e-06, "loss": 0.0508, "step": 23743 }, { "epoch": 7.31, "learning_rate": 3.579039476324747e-06, "loss": 0.0716, "step": 23744 }, { "epoch": 7.31, "learning_rate": 3.578275536457865e-06, "loss": 0.0561, "step": 23745 }, { "epoch": 7.31, "learning_rate": 3.5775116603646244e-06, "loss": 0.0715, "step": 23746 }, { "epoch": 7.31, "learning_rate": 3.5767478480526117e-06, "loss": 0.0305, "step": 23747 }, { "epoch": 7.31, "learning_rate": 3.5759840995294136e-06, "loss": 0.0518, "step": 23748 }, { "epoch": 7.31, "learning_rate": 3.575220414802614e-06, "loss": 0.0502, "step": 23749 }, { "epoch": 7.31, "learning_rate": 3.574456793879796e-06, "loss": 0.0097, "step": 23750 }, { "epoch": 7.31, "learning_rate": 3.5736932367685416e-06, "loss": 0.0379, "step": 23751 }, { "epoch": 7.31, "learning_rate": 3.572929743476439e-06, "loss": 0.0567, "step": 23752 }, { "epoch": 7.31, "learning_rate": 3.5721663140110706e-06, "loss": 0.0426, "step": 23753 }, { "epoch": 7.31, "learning_rate": 3.5714029483800085e-06, "loss": 0.0137, "step": 23754 }, { "epoch": 7.31, "learning_rate": 3.570639646590842e-06, "loss": 0.0442, "step": 23755 }, { "epoch": 7.31, "learning_rate": 3.5698764086511505e-06, "loss": 0.0417, "step": 23756 }, { "epoch": 7.31, "learning_rate": 3.569113234568513e-06, "loss": 0.0377, "step": 23757 }, { "epoch": 7.31, "learning_rate": 3.568350124350507e-06, "loss": 0.0507, "step": 23758 }, { "epoch": 7.31, "learning_rate": 3.567587078004712e-06, "loss": 0.0717, "step": 23759 }, { "epoch": 7.31, "learning_rate": 3.5668240955387055e-06, "loss": 0.0215, "step": 23760 }, { "epoch": 7.31, "learning_rate": 3.5660611769600604e-06, "loss": 0.0535, "step": 23761 }, { "epoch": 7.31, "learning_rate": 3.5652983222763627e-06, "loss": 0.0476, "step": 23762 }, { "epoch": 7.31, "learning_rate": 3.564535531495186e-06, "loss": 0.0452, "step": 23763 }, { "epoch": 7.31, "learning_rate": 3.563772804624095e-06, "loss": 0.0197, "step": 23764 }, { "epoch": 7.31, "learning_rate": 3.5630101416706755e-06, "loss": 0.0401, "step": 23765 }, { "epoch": 7.31, "learning_rate": 3.5622475426424985e-06, "loss": 0.0437, "step": 23766 }, { "epoch": 7.31, "learning_rate": 3.561485007547135e-06, "loss": 0.0251, "step": 23767 }, { "epoch": 7.31, "learning_rate": 3.5607225363921615e-06, "loss": 0.0608, "step": 23768 }, { "epoch": 7.31, "learning_rate": 3.5599601291851427e-06, "loss": 0.0536, "step": 23769 }, { "epoch": 7.31, "learning_rate": 3.5591977859336625e-06, "loss": 0.0711, "step": 23770 }, { "epoch": 7.31, "learning_rate": 3.558435506645278e-06, "loss": 0.0465, "step": 23771 }, { "epoch": 7.31, "learning_rate": 3.557673291327568e-06, "loss": 0.0627, "step": 23772 }, { "epoch": 7.31, "learning_rate": 3.556911139988101e-06, "loss": 0.0469, "step": 23773 }, { "epoch": 7.31, "learning_rate": 3.556149052634443e-06, "loss": 0.0388, "step": 23774 }, { "epoch": 7.32, "learning_rate": 3.5553870292741644e-06, "loss": 0.0727, "step": 23775 }, { "epoch": 7.32, "learning_rate": 3.554625069914832e-06, "loss": 0.0749, "step": 23776 }, { "epoch": 7.32, "learning_rate": 3.553863174564014e-06, "loss": 0.0645, "step": 23777 }, { "epoch": 7.32, "learning_rate": 3.553101343229274e-06, "loss": 0.0427, "step": 23778 }, { "epoch": 7.32, "learning_rate": 3.552339575918177e-06, "loss": 0.0629, "step": 23779 }, { "epoch": 7.32, "learning_rate": 3.5515778726382967e-06, "loss": 0.0407, "step": 23780 }, { "epoch": 7.32, "learning_rate": 3.550816233397184e-06, "loss": 0.0535, "step": 23781 }, { "epoch": 7.32, "learning_rate": 3.5500546582024144e-06, "loss": 0.0523, "step": 23782 }, { "epoch": 7.32, "learning_rate": 3.5492931470615453e-06, "loss": 0.1084, "step": 23783 }, { "epoch": 7.32, "learning_rate": 3.5485316999821405e-06, "loss": 0.0565, "step": 23784 }, { "epoch": 7.32, "learning_rate": 3.5477703169717624e-06, "loss": 0.0552, "step": 23785 }, { "epoch": 7.32, "learning_rate": 3.5470089980379673e-06, "loss": 0.0396, "step": 23786 }, { "epoch": 7.32, "learning_rate": 3.546247743188328e-06, "loss": 0.0664, "step": 23787 }, { "epoch": 7.32, "learning_rate": 3.545486552430393e-06, "loss": 0.0688, "step": 23788 }, { "epoch": 7.32, "learning_rate": 3.544725425771721e-06, "loss": 0.0564, "step": 23789 }, { "epoch": 7.32, "learning_rate": 3.5439643632198795e-06, "loss": 0.0478, "step": 23790 }, { "epoch": 7.32, "learning_rate": 3.5432033647824206e-06, "loss": 0.0357, "step": 23791 }, { "epoch": 7.32, "learning_rate": 3.5424424304669038e-06, "loss": 0.0299, "step": 23792 }, { "epoch": 7.32, "learning_rate": 3.5416815602808854e-06, "loss": 0.0432, "step": 23793 }, { "epoch": 7.32, "learning_rate": 3.54092075423192e-06, "loss": 0.0378, "step": 23794 }, { "epoch": 7.32, "learning_rate": 3.540160012327565e-06, "loss": 0.0449, "step": 23795 }, { "epoch": 7.32, "learning_rate": 3.539399334575372e-06, "loss": 0.034, "step": 23796 }, { "epoch": 7.32, "learning_rate": 3.5386387209829044e-06, "loss": 0.0814, "step": 23797 }, { "epoch": 7.32, "learning_rate": 3.537878171557705e-06, "loss": 0.0166, "step": 23798 }, { "epoch": 7.32, "learning_rate": 3.5371176863073285e-06, "loss": 0.0564, "step": 23799 }, { "epoch": 7.32, "learning_rate": 3.536357265239333e-06, "loss": 0.0561, "step": 23800 }, { "epoch": 7.32, "learning_rate": 3.535596908361266e-06, "loss": 0.0847, "step": 23801 }, { "epoch": 7.32, "learning_rate": 3.5348366156806803e-06, "loss": 0.0293, "step": 23802 }, { "epoch": 7.32, "learning_rate": 3.5340763872051253e-06, "loss": 0.023, "step": 23803 }, { "epoch": 7.32, "learning_rate": 3.5333162229421512e-06, "loss": 0.0324, "step": 23804 }, { "epoch": 7.32, "learning_rate": 3.5325561228993065e-06, "loss": 0.0537, "step": 23805 }, { "epoch": 7.32, "learning_rate": 3.531796087084137e-06, "loss": 0.0313, "step": 23806 }, { "epoch": 7.32, "learning_rate": 3.531036115504197e-06, "loss": 0.0553, "step": 23807 }, { "epoch": 7.33, "learning_rate": 3.530276208167034e-06, "loss": 0.0787, "step": 23808 }, { "epoch": 7.33, "learning_rate": 3.5295163650801833e-06, "loss": 0.1316, "step": 23809 }, { "epoch": 7.33, "learning_rate": 3.528756586251203e-06, "loss": 0.077, "step": 23810 }, { "epoch": 7.33, "learning_rate": 3.5279968716876333e-06, "loss": 0.044, "step": 23811 }, { "epoch": 7.33, "learning_rate": 3.5272372213970184e-06, "loss": 0.0465, "step": 23812 }, { "epoch": 7.33, "learning_rate": 3.5264776353869046e-06, "loss": 0.0316, "step": 23813 }, { "epoch": 7.33, "learning_rate": 3.5257181136648344e-06, "loss": 0.0816, "step": 23814 }, { "epoch": 7.33, "learning_rate": 3.524958656238349e-06, "loss": 0.0671, "step": 23815 }, { "epoch": 7.33, "learning_rate": 3.524199263114989e-06, "loss": 0.0509, "step": 23816 }, { "epoch": 7.33, "learning_rate": 3.5234399343023016e-06, "loss": 0.0722, "step": 23817 }, { "epoch": 7.33, "learning_rate": 3.5226806698078287e-06, "loss": 0.0509, "step": 23818 }, { "epoch": 7.33, "learning_rate": 3.521921469639099e-06, "loss": 0.044, "step": 23819 }, { "epoch": 7.33, "learning_rate": 3.521162333803664e-06, "loss": 0.0481, "step": 23820 }, { "epoch": 7.33, "learning_rate": 3.520403262309057e-06, "loss": 0.0377, "step": 23821 }, { "epoch": 7.33, "learning_rate": 3.5196442551628174e-06, "loss": 0.0302, "step": 23822 }, { "epoch": 7.33, "learning_rate": 3.518885312372483e-06, "loss": 0.032, "step": 23823 }, { "epoch": 7.33, "learning_rate": 3.518126433945587e-06, "loss": 0.0494, "step": 23824 }, { "epoch": 7.33, "learning_rate": 3.5173676198896768e-06, "loss": 0.0801, "step": 23825 }, { "epoch": 7.33, "learning_rate": 3.5166088702122738e-06, "loss": 0.0512, "step": 23826 }, { "epoch": 7.33, "learning_rate": 3.5158501849209224e-06, "loss": 0.0536, "step": 23827 }, { "epoch": 7.33, "learning_rate": 3.5150915640231553e-06, "loss": 0.0368, "step": 23828 }, { "epoch": 7.33, "learning_rate": 3.514333007526506e-06, "loss": 0.0532, "step": 23829 }, { "epoch": 7.33, "learning_rate": 3.513574515438507e-06, "loss": 0.0501, "step": 23830 }, { "epoch": 7.33, "learning_rate": 3.5128160877666873e-06, "loss": 0.0476, "step": 23831 }, { "epoch": 7.33, "learning_rate": 3.5120577245185906e-06, "loss": 0.074, "step": 23832 }, { "epoch": 7.33, "learning_rate": 3.511299425701735e-06, "loss": 0.0342, "step": 23833 }, { "epoch": 7.33, "learning_rate": 3.5105411913236544e-06, "loss": 0.0613, "step": 23834 }, { "epoch": 7.33, "learning_rate": 3.509783021391887e-06, "loss": 0.0604, "step": 23835 }, { "epoch": 7.33, "learning_rate": 3.509024915913949e-06, "loss": 0.0494, "step": 23836 }, { "epoch": 7.33, "learning_rate": 3.508266874897379e-06, "loss": 0.0413, "step": 23837 }, { "epoch": 7.33, "learning_rate": 3.507508898349702e-06, "loss": 0.0625, "step": 23838 }, { "epoch": 7.33, "learning_rate": 3.5067509862784455e-06, "loss": 0.0547, "step": 23839 }, { "epoch": 7.34, "learning_rate": 3.505993138691135e-06, "loss": 0.0649, "step": 23840 }, { "epoch": 7.34, "learning_rate": 3.5052353555952955e-06, "loss": 0.0435, "step": 23841 }, { "epoch": 7.34, "learning_rate": 3.504477636998461e-06, "loss": 0.0342, "step": 23842 }, { "epoch": 7.34, "learning_rate": 3.503719982908147e-06, "loss": 0.0822, "step": 23843 }, { "epoch": 7.34, "learning_rate": 3.502962393331877e-06, "loss": 0.0784, "step": 23844 }, { "epoch": 7.34, "learning_rate": 3.5022048682771814e-06, "loss": 0.0434, "step": 23845 }, { "epoch": 7.34, "learning_rate": 3.50144740775158e-06, "loss": 0.0351, "step": 23846 }, { "epoch": 7.34, "learning_rate": 3.5006900117625953e-06, "loss": 0.0148, "step": 23847 }, { "epoch": 7.34, "learning_rate": 3.4999326803177493e-06, "loss": 0.0414, "step": 23848 }, { "epoch": 7.34, "learning_rate": 3.4991754134245617e-06, "loss": 0.0437, "step": 23849 }, { "epoch": 7.34, "learning_rate": 3.498418211090554e-06, "loss": 0.0529, "step": 23850 }, { "epoch": 7.34, "learning_rate": 3.497661073323242e-06, "loss": 0.0365, "step": 23851 }, { "epoch": 7.34, "learning_rate": 3.4969040001301513e-06, "loss": 0.0271, "step": 23852 }, { "epoch": 7.34, "learning_rate": 3.4961469915188007e-06, "loss": 0.0288, "step": 23853 }, { "epoch": 7.34, "learning_rate": 3.4953900474966985e-06, "loss": 0.0442, "step": 23854 }, { "epoch": 7.34, "learning_rate": 3.4946331680713706e-06, "loss": 0.0607, "step": 23855 }, { "epoch": 7.34, "learning_rate": 3.493876353250332e-06, "loss": 0.0741, "step": 23856 }, { "epoch": 7.34, "learning_rate": 3.4931196030410965e-06, "loss": 0.0677, "step": 23857 }, { "epoch": 7.34, "learning_rate": 3.4923629174511796e-06, "loss": 0.0753, "step": 23858 }, { "epoch": 7.34, "learning_rate": 3.491606296488098e-06, "loss": 0.0731, "step": 23859 }, { "epoch": 7.34, "learning_rate": 3.4908497401593634e-06, "loss": 0.0381, "step": 23860 }, { "epoch": 7.34, "learning_rate": 3.490093248472487e-06, "loss": 0.025, "step": 23861 }, { "epoch": 7.34, "learning_rate": 3.4893368214349864e-06, "loss": 0.0243, "step": 23862 }, { "epoch": 7.34, "learning_rate": 3.488580459054376e-06, "loss": 0.0555, "step": 23863 }, { "epoch": 7.34, "learning_rate": 3.487824161338157e-06, "loss": 0.0409, "step": 23864 }, { "epoch": 7.34, "learning_rate": 3.487067928293848e-06, "loss": 0.047, "step": 23865 }, { "epoch": 7.34, "learning_rate": 3.4863117599289562e-06, "loss": 0.0727, "step": 23866 }, { "epoch": 7.34, "learning_rate": 3.485555656250993e-06, "loss": 0.0496, "step": 23867 }, { "epoch": 7.34, "learning_rate": 3.4847996172674658e-06, "loss": 0.0316, "step": 23868 }, { "epoch": 7.34, "learning_rate": 3.484043642985879e-06, "loss": 0.058, "step": 23869 }, { "epoch": 7.34, "learning_rate": 3.4832877334137528e-06, "loss": 0.0502, "step": 23870 }, { "epoch": 7.34, "learning_rate": 3.482531888558577e-06, "loss": 0.0573, "step": 23871 }, { "epoch": 7.34, "learning_rate": 3.481776108427869e-06, "loss": 0.0464, "step": 23872 }, { "epoch": 7.35, "learning_rate": 3.481020393029136e-06, "loss": 0.034, "step": 23873 }, { "epoch": 7.35, "learning_rate": 3.480264742369872e-06, "loss": 0.0606, "step": 23874 }, { "epoch": 7.35, "learning_rate": 3.479509156457591e-06, "loss": 0.0288, "step": 23875 }, { "epoch": 7.35, "learning_rate": 3.478753635299793e-06, "loss": 0.0474, "step": 23876 }, { "epoch": 7.35, "learning_rate": 3.477998178903982e-06, "loss": 0.034, "step": 23877 }, { "epoch": 7.35, "learning_rate": 3.4772427872776606e-06, "loss": 0.0517, "step": 23878 }, { "epoch": 7.35, "learning_rate": 3.4764874604283263e-06, "loss": 0.0357, "step": 23879 }, { "epoch": 7.35, "learning_rate": 3.4757321983634895e-06, "loss": 0.052, "step": 23880 }, { "epoch": 7.35, "learning_rate": 3.474977001090638e-06, "loss": 0.0662, "step": 23881 }, { "epoch": 7.35, "learning_rate": 3.4742218686172836e-06, "loss": 0.0238, "step": 23882 }, { "epoch": 7.35, "learning_rate": 3.4734668009509187e-06, "loss": 0.0674, "step": 23883 }, { "epoch": 7.35, "learning_rate": 3.4727117980990444e-06, "loss": 0.0525, "step": 23884 }, { "epoch": 7.35, "learning_rate": 3.471956860069158e-06, "loss": 0.076, "step": 23885 }, { "epoch": 7.35, "learning_rate": 3.471201986868752e-06, "loss": 0.0483, "step": 23886 }, { "epoch": 7.35, "learning_rate": 3.4704471785053353e-06, "loss": 0.0304, "step": 23887 }, { "epoch": 7.35, "learning_rate": 3.4696924349863913e-06, "loss": 0.0454, "step": 23888 }, { "epoch": 7.35, "learning_rate": 3.468937756319417e-06, "loss": 0.0418, "step": 23889 }, { "epoch": 7.35, "learning_rate": 3.468183142511914e-06, "loss": 0.0279, "step": 23890 }, { "epoch": 7.35, "learning_rate": 3.4674285935713715e-06, "loss": 0.0404, "step": 23891 }, { "epoch": 7.35, "learning_rate": 3.4666741095052837e-06, "loss": 0.0497, "step": 23892 }, { "epoch": 7.35, "learning_rate": 3.4659196903211434e-06, "loss": 0.0603, "step": 23893 }, { "epoch": 7.35, "learning_rate": 3.4651653360264426e-06, "loss": 0.0549, "step": 23894 }, { "epoch": 7.35, "learning_rate": 3.464411046628673e-06, "loss": 0.0629, "step": 23895 }, { "epoch": 7.35, "learning_rate": 3.463656822135322e-06, "loss": 0.0346, "step": 23896 }, { "epoch": 7.35, "learning_rate": 3.4629026625538884e-06, "loss": 0.0642, "step": 23897 }, { "epoch": 7.35, "learning_rate": 3.462148567891854e-06, "loss": 0.0537, "step": 23898 }, { "epoch": 7.35, "learning_rate": 3.461394538156706e-06, "loss": 0.038, "step": 23899 }, { "epoch": 7.35, "learning_rate": 3.460640573355939e-06, "loss": 0.066, "step": 23900 }, { "epoch": 7.35, "learning_rate": 3.459886673497038e-06, "loss": 0.065, "step": 23901 }, { "epoch": 7.35, "learning_rate": 3.459132838587491e-06, "loss": 0.0407, "step": 23902 }, { "epoch": 7.35, "learning_rate": 3.458379068634783e-06, "loss": 0.0786, "step": 23903 }, { "epoch": 7.35, "learning_rate": 3.4576253636463996e-06, "loss": 0.0502, "step": 23904 }, { "epoch": 7.36, "learning_rate": 3.4568717236298257e-06, "loss": 0.0292, "step": 23905 }, { "epoch": 7.36, "learning_rate": 3.4561181485925423e-06, "loss": 0.054, "step": 23906 }, { "epoch": 7.36, "learning_rate": 3.4553646385420403e-06, "loss": 0.0442, "step": 23907 }, { "epoch": 7.36, "learning_rate": 3.454611193485803e-06, "loss": 0.0443, "step": 23908 }, { "epoch": 7.36, "learning_rate": 3.453857813431302e-06, "loss": 0.0496, "step": 23909 }, { "epoch": 7.36, "learning_rate": 3.4531044983860285e-06, "loss": 0.0334, "step": 23910 }, { "epoch": 7.36, "learning_rate": 3.452351248357462e-06, "loss": 0.0654, "step": 23911 }, { "epoch": 7.36, "learning_rate": 3.4515980633530823e-06, "loss": 0.0625, "step": 23912 }, { "epoch": 7.36, "learning_rate": 3.4508449433803683e-06, "loss": 0.0942, "step": 23913 }, { "epoch": 7.36, "learning_rate": 3.4500918884468006e-06, "loss": 0.0487, "step": 23914 }, { "epoch": 7.36, "learning_rate": 3.449338898559856e-06, "loss": 0.029, "step": 23915 }, { "epoch": 7.36, "learning_rate": 3.448585973727011e-06, "loss": 0.0471, "step": 23916 }, { "epoch": 7.36, "learning_rate": 3.4478331139557475e-06, "loss": 0.0383, "step": 23917 }, { "epoch": 7.36, "learning_rate": 3.447080319253544e-06, "loss": 0.051, "step": 23918 }, { "epoch": 7.36, "learning_rate": 3.4463275896278657e-06, "loss": 0.0487, "step": 23919 }, { "epoch": 7.36, "learning_rate": 3.445574925086198e-06, "loss": 0.0437, "step": 23920 }, { "epoch": 7.36, "learning_rate": 3.4448223256360112e-06, "loss": 0.041, "step": 23921 }, { "epoch": 7.36, "learning_rate": 3.444069791284781e-06, "loss": 0.0789, "step": 23922 }, { "epoch": 7.36, "learning_rate": 3.4433173220399796e-06, "loss": 0.0364, "step": 23923 }, { "epoch": 7.36, "learning_rate": 3.442564917909077e-06, "loss": 0.049, "step": 23924 }, { "epoch": 7.36, "learning_rate": 3.441812578899554e-06, "loss": 0.0594, "step": 23925 }, { "epoch": 7.36, "learning_rate": 3.4410603050188705e-06, "loss": 0.0272, "step": 23926 }, { "epoch": 7.36, "learning_rate": 3.440308096274506e-06, "loss": 0.0294, "step": 23927 }, { "epoch": 7.36, "learning_rate": 3.4395559526739273e-06, "loss": 0.0575, "step": 23928 }, { "epoch": 7.36, "learning_rate": 3.4388038742246055e-06, "loss": 0.0351, "step": 23929 }, { "epoch": 7.36, "learning_rate": 3.4380518609340076e-06, "loss": 0.0805, "step": 23930 }, { "epoch": 7.36, "learning_rate": 3.437299912809602e-06, "loss": 0.0375, "step": 23931 }, { "epoch": 7.36, "learning_rate": 3.436548029858856e-06, "loss": 0.0499, "step": 23932 }, { "epoch": 7.36, "learning_rate": 3.4357962120892383e-06, "loss": 0.0793, "step": 23933 }, { "epoch": 7.36, "learning_rate": 3.4350444595082098e-06, "loss": 0.0558, "step": 23934 }, { "epoch": 7.36, "learning_rate": 3.434292772123248e-06, "loss": 0.0487, "step": 23935 }, { "epoch": 7.36, "learning_rate": 3.4335411499418027e-06, "loss": 0.0387, "step": 23936 }, { "epoch": 7.36, "learning_rate": 3.4327895929713494e-06, "loss": 0.0353, "step": 23937 }, { "epoch": 7.37, "learning_rate": 3.432038101219348e-06, "loss": 0.0515, "step": 23938 }, { "epoch": 7.37, "learning_rate": 3.4312866746932603e-06, "loss": 0.0362, "step": 23939 }, { "epoch": 7.37, "learning_rate": 3.4305353134005514e-06, "loss": 0.0248, "step": 23940 }, { "epoch": 7.37, "learning_rate": 3.4297840173486774e-06, "loss": 0.0256, "step": 23941 }, { "epoch": 7.37, "learning_rate": 3.42903278654511e-06, "loss": 0.0303, "step": 23942 }, { "epoch": 7.37, "learning_rate": 3.428281620997296e-06, "loss": 0.0503, "step": 23943 }, { "epoch": 7.37, "learning_rate": 3.427530520712706e-06, "loss": 0.0828, "step": 23944 }, { "epoch": 7.37, "learning_rate": 3.426779485698796e-06, "loss": 0.0225, "step": 23945 }, { "epoch": 7.37, "learning_rate": 3.4260285159630224e-06, "loss": 0.0571, "step": 23946 }, { "epoch": 7.37, "learning_rate": 3.425277611512845e-06, "loss": 0.0394, "step": 23947 }, { "epoch": 7.37, "learning_rate": 3.4245267723557163e-06, "loss": 0.0667, "step": 23948 }, { "epoch": 7.37, "learning_rate": 3.423775998499105e-06, "loss": 0.0101, "step": 23949 }, { "epoch": 7.37, "learning_rate": 3.4230252899504546e-06, "loss": 0.0275, "step": 23950 }, { "epoch": 7.37, "learning_rate": 3.422274646717221e-06, "loss": 0.0758, "step": 23951 }, { "epoch": 7.37, "learning_rate": 3.4215240688068694e-06, "loss": 0.0604, "step": 23952 }, { "epoch": 7.37, "learning_rate": 3.4207735562268406e-06, "loss": 0.0358, "step": 23953 }, { "epoch": 7.37, "learning_rate": 3.420023108984597e-06, "loss": 0.0606, "step": 23954 }, { "epoch": 7.37, "learning_rate": 3.419272727087587e-06, "loss": 0.0771, "step": 23955 }, { "epoch": 7.37, "learning_rate": 3.418522410543266e-06, "loss": 0.0354, "step": 23956 }, { "epoch": 7.37, "learning_rate": 3.4177721593590806e-06, "loss": 0.0371, "step": 23957 }, { "epoch": 7.37, "learning_rate": 3.4170219735424827e-06, "loss": 0.0511, "step": 23958 }, { "epoch": 7.37, "learning_rate": 3.41627185310093e-06, "loss": 0.0563, "step": 23959 }, { "epoch": 7.37, "learning_rate": 3.415521798041861e-06, "loss": 0.0574, "step": 23960 }, { "epoch": 7.37, "learning_rate": 3.4147718083727264e-06, "loss": 0.0546, "step": 23961 }, { "epoch": 7.37, "learning_rate": 3.4140218841009808e-06, "loss": 0.0498, "step": 23962 }, { "epoch": 7.37, "learning_rate": 3.4132720252340666e-06, "loss": 0.0848, "step": 23963 }, { "epoch": 7.37, "learning_rate": 3.4125222317794327e-06, "loss": 0.0714, "step": 23964 }, { "epoch": 7.37, "learning_rate": 3.4117725037445237e-06, "loss": 0.0507, "step": 23965 }, { "epoch": 7.37, "learning_rate": 3.4110228411367863e-06, "loss": 0.0575, "step": 23966 }, { "epoch": 7.37, "learning_rate": 3.4102732439636645e-06, "loss": 0.0523, "step": 23967 }, { "epoch": 7.37, "learning_rate": 3.4095237122325984e-06, "loss": 0.0514, "step": 23968 }, { "epoch": 7.37, "learning_rate": 3.4087742459510396e-06, "loss": 0.0557, "step": 23969 }, { "epoch": 7.38, "learning_rate": 3.4080248451264307e-06, "loss": 0.0481, "step": 23970 }, { "epoch": 7.38, "learning_rate": 3.4072755097662035e-06, "loss": 0.0503, "step": 23971 }, { "epoch": 7.38, "learning_rate": 3.40652623987781e-06, "loss": 0.0389, "step": 23972 }, { "epoch": 7.38, "learning_rate": 3.405777035468688e-06, "loss": 0.0433, "step": 23973 }, { "epoch": 7.38, "learning_rate": 3.405027896546277e-06, "loss": 0.0827, "step": 23974 }, { "epoch": 7.38, "learning_rate": 3.404278823118017e-06, "loss": 0.0503, "step": 23975 }, { "epoch": 7.38, "learning_rate": 3.4035298151913477e-06, "loss": 0.0339, "step": 23976 }, { "epoch": 7.38, "learning_rate": 3.4027808727737065e-06, "loss": 0.0184, "step": 23977 }, { "epoch": 7.38, "learning_rate": 3.4020319958725277e-06, "loss": 0.0488, "step": 23978 }, { "epoch": 7.38, "learning_rate": 3.4012831844952554e-06, "loss": 0.0512, "step": 23979 }, { "epoch": 7.38, "learning_rate": 3.4005344386493266e-06, "loss": 0.0386, "step": 23980 }, { "epoch": 7.38, "learning_rate": 3.3997857583421656e-06, "loss": 0.0437, "step": 23981 }, { "epoch": 7.38, "learning_rate": 3.3990371435812185e-06, "loss": 0.028, "step": 23982 }, { "epoch": 7.38, "learning_rate": 3.3982885943739172e-06, "loss": 0.0441, "step": 23983 }, { "epoch": 7.38, "learning_rate": 3.397540110727694e-06, "loss": 0.0773, "step": 23984 }, { "epoch": 7.38, "learning_rate": 3.3967916926499823e-06, "loss": 0.0745, "step": 23985 }, { "epoch": 7.38, "learning_rate": 3.3960433401482107e-06, "loss": 0.0436, "step": 23986 }, { "epoch": 7.38, "learning_rate": 3.395295053229822e-06, "loss": 0.071, "step": 23987 }, { "epoch": 7.38, "learning_rate": 3.3945468319022345e-06, "loss": 0.0376, "step": 23988 }, { "epoch": 7.38, "learning_rate": 3.393798676172888e-06, "loss": 0.0317, "step": 23989 }, { "epoch": 7.38, "learning_rate": 3.393050586049209e-06, "loss": 0.0623, "step": 23990 }, { "epoch": 7.38, "learning_rate": 3.3923025615386274e-06, "loss": 0.0678, "step": 23991 }, { "epoch": 7.38, "learning_rate": 3.3915546026485703e-06, "loss": 0.0841, "step": 23992 }, { "epoch": 7.38, "learning_rate": 3.390806709386467e-06, "loss": 0.0562, "step": 23993 }, { "epoch": 7.38, "learning_rate": 3.3900588817597434e-06, "loss": 0.0086, "step": 23994 }, { "epoch": 7.38, "learning_rate": 3.3893111197758276e-06, "loss": 0.0876, "step": 23995 }, { "epoch": 7.38, "learning_rate": 3.3885634234421415e-06, "loss": 0.057, "step": 23996 }, { "epoch": 7.38, "learning_rate": 3.3878157927661215e-06, "loss": 0.0596, "step": 23997 }, { "epoch": 7.38, "learning_rate": 3.387068227755177e-06, "loss": 0.0638, "step": 23998 }, { "epoch": 7.38, "learning_rate": 3.386320728416743e-06, "loss": 0.0576, "step": 23999 }, { "epoch": 7.38, "learning_rate": 3.38557329475824e-06, "loss": 0.0605, "step": 24000 }, { "epoch": 7.38, "learning_rate": 3.3848259267870897e-06, "loss": 0.0583, "step": 24001 }, { "epoch": 7.38, "learning_rate": 3.384078624510715e-06, "loss": 0.0612, "step": 24002 }, { "epoch": 7.39, "learning_rate": 3.3833313879365337e-06, "loss": 0.0344, "step": 24003 }, { "epoch": 7.39, "learning_rate": 3.382584217071977e-06, "loss": 0.0266, "step": 24004 }, { "epoch": 7.39, "learning_rate": 3.381837111924453e-06, "loss": 0.0388, "step": 24005 }, { "epoch": 7.39, "learning_rate": 3.381090072501384e-06, "loss": 0.0676, "step": 24006 }, { "epoch": 7.39, "learning_rate": 3.3803430988101937e-06, "loss": 0.0355, "step": 24007 }, { "epoch": 7.39, "learning_rate": 3.3795961908582965e-06, "loss": 0.0527, "step": 24008 }, { "epoch": 7.39, "learning_rate": 3.3788493486531116e-06, "loss": 0.0501, "step": 24009 }, { "epoch": 7.39, "learning_rate": 3.3781025722020545e-06, "loss": 0.0729, "step": 24010 }, { "epoch": 7.39, "learning_rate": 3.3773558615125425e-06, "loss": 0.064, "step": 24011 }, { "epoch": 7.39, "learning_rate": 3.3766092165919896e-06, "loss": 0.0426, "step": 24012 }, { "epoch": 7.39, "learning_rate": 3.3758626374478077e-06, "loss": 0.0456, "step": 24013 }, { "epoch": 7.39, "learning_rate": 3.3751161240874233e-06, "loss": 0.1261, "step": 24014 }, { "epoch": 7.39, "learning_rate": 3.3743696765182367e-06, "loss": 0.0521, "step": 24015 }, { "epoch": 7.39, "learning_rate": 3.373623294747662e-06, "loss": 0.0286, "step": 24016 }, { "epoch": 7.39, "learning_rate": 3.3728769787831184e-06, "loss": 0.0527, "step": 24017 }, { "epoch": 7.39, "learning_rate": 3.372130728632015e-06, "loss": 0.0472, "step": 24018 }, { "epoch": 7.39, "learning_rate": 3.3713845443017613e-06, "loss": 0.0506, "step": 24019 }, { "epoch": 7.39, "learning_rate": 3.370638425799768e-06, "loss": 0.0516, "step": 24020 }, { "epoch": 7.39, "learning_rate": 3.3698923731334453e-06, "loss": 0.048, "step": 24021 }, { "epoch": 7.39, "learning_rate": 3.3691463863102025e-06, "loss": 0.0529, "step": 24022 }, { "epoch": 7.39, "learning_rate": 3.3684004653374424e-06, "loss": 0.0753, "step": 24023 }, { "epoch": 7.39, "learning_rate": 3.3676546102225827e-06, "loss": 0.0368, "step": 24024 }, { "epoch": 7.39, "learning_rate": 3.366908820973027e-06, "loss": 0.0699, "step": 24025 }, { "epoch": 7.39, "learning_rate": 3.3661630975961735e-06, "loss": 0.0547, "step": 24026 }, { "epoch": 7.39, "learning_rate": 3.365417440099439e-06, "loss": 0.0646, "step": 24027 }, { "epoch": 7.39, "learning_rate": 3.364671848490223e-06, "loss": 0.0373, "step": 24028 }, { "epoch": 7.39, "learning_rate": 3.36392632277593e-06, "loss": 0.0565, "step": 24029 }, { "epoch": 7.39, "learning_rate": 3.3631808629639663e-06, "loss": 0.0365, "step": 24030 }, { "epoch": 7.39, "learning_rate": 3.362435469061732e-06, "loss": 0.0386, "step": 24031 }, { "epoch": 7.39, "learning_rate": 3.361690141076631e-06, "loss": 0.051, "step": 24032 }, { "epoch": 7.39, "learning_rate": 3.3609448790160624e-06, "loss": 0.0543, "step": 24033 }, { "epoch": 7.39, "learning_rate": 3.360199682887433e-06, "loss": 0.0403, "step": 24034 }, { "epoch": 7.4, "learning_rate": 3.3594545526981424e-06, "loss": 0.0842, "step": 24035 }, { "epoch": 7.4, "learning_rate": 3.358709488455583e-06, "loss": 0.0434, "step": 24036 }, { "epoch": 7.4, "learning_rate": 3.3579644901671615e-06, "loss": 0.0414, "step": 24037 }, { "epoch": 7.4, "learning_rate": 3.3572195578402744e-06, "loss": 0.0638, "step": 24038 }, { "epoch": 7.4, "learning_rate": 3.3564746914823198e-06, "loss": 0.06, "step": 24039 }, { "epoch": 7.4, "learning_rate": 3.355729891100693e-06, "loss": 0.0661, "step": 24040 }, { "epoch": 7.4, "learning_rate": 3.3549851567027904e-06, "loss": 0.0391, "step": 24041 }, { "epoch": 7.4, "learning_rate": 3.3542404882960146e-06, "loss": 0.0392, "step": 24042 }, { "epoch": 7.4, "learning_rate": 3.3534958858877498e-06, "loss": 0.0249, "step": 24043 }, { "epoch": 7.4, "learning_rate": 3.3527513494853993e-06, "loss": 0.0486, "step": 24044 }, { "epoch": 7.4, "learning_rate": 3.3520068790963546e-06, "loss": 0.0641, "step": 24045 }, { "epoch": 7.4, "learning_rate": 3.3512624747280088e-06, "loss": 0.0556, "step": 24046 }, { "epoch": 7.4, "learning_rate": 3.3505181363877536e-06, "loss": 0.0618, "step": 24047 }, { "epoch": 7.4, "learning_rate": 3.349773864082978e-06, "loss": 0.0438, "step": 24048 }, { "epoch": 7.4, "learning_rate": 3.3490296578210845e-06, "loss": 0.053, "step": 24049 }, { "epoch": 7.4, "learning_rate": 3.348285517609453e-06, "loss": 0.0409, "step": 24050 }, { "epoch": 7.4, "learning_rate": 3.347541443455473e-06, "loss": 0.0647, "step": 24051 }, { "epoch": 7.4, "learning_rate": 3.3467974353665433e-06, "loss": 0.0264, "step": 24052 }, { "epoch": 7.4, "learning_rate": 3.346053493350042e-06, "loss": 0.0596, "step": 24053 }, { "epoch": 7.4, "learning_rate": 3.3453096174133635e-06, "loss": 0.041, "step": 24054 }, { "epoch": 7.4, "learning_rate": 3.3445658075638955e-06, "loss": 0.0365, "step": 24055 }, { "epoch": 7.4, "learning_rate": 3.3438220638090212e-06, "loss": 0.0291, "step": 24056 }, { "epoch": 7.4, "learning_rate": 3.343078386156129e-06, "loss": 0.0253, "step": 24057 }, { "epoch": 7.4, "learning_rate": 3.3423347746126e-06, "loss": 0.0317, "step": 24058 }, { "epoch": 7.4, "learning_rate": 3.34159122918583e-06, "loss": 0.0375, "step": 24059 }, { "epoch": 7.4, "learning_rate": 3.3408477498831917e-06, "loss": 0.0535, "step": 24060 }, { "epoch": 7.4, "learning_rate": 3.340104336712069e-06, "loss": 0.0611, "step": 24061 }, { "epoch": 7.4, "learning_rate": 3.339360989679852e-06, "loss": 0.0417, "step": 24062 }, { "epoch": 7.4, "learning_rate": 3.338617708793919e-06, "loss": 0.0566, "step": 24063 }, { "epoch": 7.4, "learning_rate": 3.3378744940616516e-06, "loss": 0.0378, "step": 24064 }, { "epoch": 7.4, "learning_rate": 3.33713134549043e-06, "loss": 0.0902, "step": 24065 }, { "epoch": 7.4, "learning_rate": 3.3363882630876355e-06, "loss": 0.0652, "step": 24066 }, { "epoch": 7.4, "learning_rate": 3.335645246860646e-06, "loss": 0.0561, "step": 24067 }, { "epoch": 7.41, "learning_rate": 3.334902296816838e-06, "loss": 0.0364, "step": 24068 }, { "epoch": 7.41, "learning_rate": 3.334159412963597e-06, "loss": 0.059, "step": 24069 }, { "epoch": 7.41, "learning_rate": 3.3334165953082987e-06, "loss": 0.029, "step": 24070 }, { "epoch": 7.41, "learning_rate": 3.3326738438583116e-06, "loss": 0.0493, "step": 24071 }, { "epoch": 7.41, "learning_rate": 3.3319311586210225e-06, "loss": 0.0478, "step": 24072 }, { "epoch": 7.41, "learning_rate": 3.3311885396038002e-06, "loss": 0.045, "step": 24073 }, { "epoch": 7.41, "learning_rate": 3.3304459868140237e-06, "loss": 0.079, "step": 24074 }, { "epoch": 7.41, "learning_rate": 3.329703500259065e-06, "loss": 0.0762, "step": 24075 }, { "epoch": 7.41, "learning_rate": 3.3289610799462978e-06, "loss": 0.0546, "step": 24076 }, { "epoch": 7.41, "learning_rate": 3.3282187258830943e-06, "loss": 0.0437, "step": 24077 }, { "epoch": 7.41, "learning_rate": 3.3274764380768265e-06, "loss": 0.0572, "step": 24078 }, { "epoch": 7.41, "learning_rate": 3.3267342165348694e-06, "loss": 0.056, "step": 24079 }, { "epoch": 7.41, "learning_rate": 3.3259920612645948e-06, "loss": 0.058, "step": 24080 }, { "epoch": 7.41, "learning_rate": 3.3252499722733646e-06, "loss": 0.029, "step": 24081 }, { "epoch": 7.41, "learning_rate": 3.324507949568556e-06, "loss": 0.0734, "step": 24082 }, { "epoch": 7.41, "learning_rate": 3.3237659931575363e-06, "loss": 0.0584, "step": 24083 }, { "epoch": 7.41, "learning_rate": 3.3230241030476728e-06, "loss": 0.0381, "step": 24084 }, { "epoch": 7.41, "learning_rate": 3.322282279246334e-06, "loss": 0.0311, "step": 24085 }, { "epoch": 7.41, "learning_rate": 3.321540521760883e-06, "loss": 0.0429, "step": 24086 }, { "epoch": 7.41, "learning_rate": 3.320798830598696e-06, "loss": 0.0508, "step": 24087 }, { "epoch": 7.41, "learning_rate": 3.3200572057671267e-06, "loss": 0.0473, "step": 24088 }, { "epoch": 7.41, "learning_rate": 3.319315647273548e-06, "loss": 0.0364, "step": 24089 }, { "epoch": 7.41, "learning_rate": 3.3185741551253226e-06, "loss": 0.0581, "step": 24090 }, { "epoch": 7.41, "learning_rate": 3.317832729329814e-06, "loss": 0.0232, "step": 24091 }, { "epoch": 7.41, "learning_rate": 3.317091369894384e-06, "loss": 0.034, "step": 24092 }, { "epoch": 7.41, "learning_rate": 3.3163500768263957e-06, "loss": 0.0679, "step": 24093 }, { "epoch": 7.41, "learning_rate": 3.315608850133212e-06, "loss": 0.0304, "step": 24094 }, { "epoch": 7.41, "learning_rate": 3.3148676898221922e-06, "loss": 0.0436, "step": 24095 }, { "epoch": 7.41, "learning_rate": 3.314126595900695e-06, "loss": 0.0512, "step": 24096 }, { "epoch": 7.41, "learning_rate": 3.3133855683760886e-06, "loss": 0.0333, "step": 24097 }, { "epoch": 7.41, "learning_rate": 3.3126446072557196e-06, "loss": 0.0493, "step": 24098 }, { "epoch": 7.41, "learning_rate": 3.3119037125469553e-06, "loss": 0.0434, "step": 24099 }, { "epoch": 7.42, "learning_rate": 3.3111628842571518e-06, "loss": 0.0344, "step": 24100 }, { "epoch": 7.42, "learning_rate": 3.3104221223936663e-06, "loss": 0.0496, "step": 24101 }, { "epoch": 7.42, "learning_rate": 3.309681426963853e-06, "loss": 0.0849, "step": 24102 }, { "epoch": 7.42, "learning_rate": 3.3089407979750665e-06, "loss": 0.0774, "step": 24103 }, { "epoch": 7.42, "learning_rate": 3.308200235434671e-06, "loss": 0.043, "step": 24104 }, { "epoch": 7.42, "learning_rate": 3.3074597393500116e-06, "loss": 0.0599, "step": 24105 }, { "epoch": 7.42, "learning_rate": 3.3067193097284412e-06, "loss": 0.0842, "step": 24106 }, { "epoch": 7.42, "learning_rate": 3.3059789465773206e-06, "loss": 0.0322, "step": 24107 }, { "epoch": 7.42, "learning_rate": 3.305238649903998e-06, "loss": 0.0501, "step": 24108 }, { "epoch": 7.42, "learning_rate": 3.304498419715826e-06, "loss": 0.051, "step": 24109 }, { "epoch": 7.42, "learning_rate": 3.303758256020154e-06, "loss": 0.0436, "step": 24110 }, { "epoch": 7.42, "learning_rate": 3.303018158824336e-06, "loss": 0.0408, "step": 24111 }, { "epoch": 7.42, "learning_rate": 3.3022781281357184e-06, "loss": 0.0581, "step": 24112 }, { "epoch": 7.42, "learning_rate": 3.3015381639616493e-06, "loss": 0.0287, "step": 24113 }, { "epoch": 7.42, "learning_rate": 3.300798266309486e-06, "loss": 0.0582, "step": 24114 }, { "epoch": 7.42, "learning_rate": 3.300058435186567e-06, "loss": 0.0589, "step": 24115 }, { "epoch": 7.42, "learning_rate": 3.299318670600239e-06, "loss": 0.0519, "step": 24116 }, { "epoch": 7.42, "learning_rate": 3.2985789725578563e-06, "loss": 0.0425, "step": 24117 }, { "epoch": 7.42, "learning_rate": 3.2978393410667596e-06, "loss": 0.0526, "step": 24118 }, { "epoch": 7.42, "learning_rate": 3.297099776134296e-06, "loss": 0.0445, "step": 24119 }, { "epoch": 7.42, "learning_rate": 3.2963602777678084e-06, "loss": 0.101, "step": 24120 }, { "epoch": 7.42, "learning_rate": 3.295620845974642e-06, "loss": 0.0508, "step": 24121 }, { "epoch": 7.42, "learning_rate": 3.294881480762139e-06, "loss": 0.0477, "step": 24122 }, { "epoch": 7.42, "learning_rate": 3.2941421821376397e-06, "loss": 0.0597, "step": 24123 }, { "epoch": 7.42, "learning_rate": 3.293402950108492e-06, "loss": 0.0433, "step": 24124 }, { "epoch": 7.42, "learning_rate": 3.2926637846820366e-06, "loss": 0.0652, "step": 24125 }, { "epoch": 7.42, "learning_rate": 3.2919246858656053e-06, "loss": 0.0361, "step": 24126 }, { "epoch": 7.42, "learning_rate": 3.2911856536665475e-06, "loss": 0.0587, "step": 24127 }, { "epoch": 7.42, "learning_rate": 3.2904466880922e-06, "loss": 0.051, "step": 24128 }, { "epoch": 7.42, "learning_rate": 3.2897077891498996e-06, "loss": 0.0535, "step": 24129 }, { "epoch": 7.42, "learning_rate": 3.288968956846986e-06, "loss": 0.0295, "step": 24130 }, { "epoch": 7.42, "learning_rate": 3.2882301911907955e-06, "loss": 0.0478, "step": 24131 }, { "epoch": 7.42, "learning_rate": 3.2874914921886646e-06, "loss": 0.0758, "step": 24132 }, { "epoch": 7.43, "learning_rate": 3.286752859847926e-06, "loss": 0.0352, "step": 24133 }, { "epoch": 7.43, "learning_rate": 3.286014294175923e-06, "loss": 0.0645, "step": 24134 }, { "epoch": 7.43, "learning_rate": 3.2852757951799887e-06, "loss": 0.0496, "step": 24135 }, { "epoch": 7.43, "learning_rate": 3.284537362867447e-06, "loss": 0.0688, "step": 24136 }, { "epoch": 7.43, "learning_rate": 3.283798997245643e-06, "loss": 0.034, "step": 24137 }, { "epoch": 7.43, "learning_rate": 3.2830606983219038e-06, "loss": 0.0376, "step": 24138 }, { "epoch": 7.43, "learning_rate": 3.2823224661035623e-06, "loss": 0.059, "step": 24139 }, { "epoch": 7.43, "learning_rate": 3.2815843005979475e-06, "loss": 0.055, "step": 24140 }, { "epoch": 7.43, "learning_rate": 3.280846201812394e-06, "loss": 0.0606, "step": 24141 }, { "epoch": 7.43, "learning_rate": 3.280108169754236e-06, "loss": 0.0382, "step": 24142 }, { "epoch": 7.43, "learning_rate": 3.2793702044307885e-06, "loss": 0.0649, "step": 24143 }, { "epoch": 7.43, "learning_rate": 3.278632305849393e-06, "loss": 0.0562, "step": 24144 }, { "epoch": 7.43, "learning_rate": 3.2778944740173736e-06, "loss": 0.0812, "step": 24145 }, { "epoch": 7.43, "learning_rate": 3.2771567089420563e-06, "loss": 0.0531, "step": 24146 }, { "epoch": 7.43, "learning_rate": 3.276419010630769e-06, "loss": 0.0911, "step": 24147 }, { "epoch": 7.43, "learning_rate": 3.2756813790908347e-06, "loss": 0.0625, "step": 24148 }, { "epoch": 7.43, "learning_rate": 3.2749438143295887e-06, "loss": 0.0342, "step": 24149 }, { "epoch": 7.43, "learning_rate": 3.2742063163543414e-06, "loss": 0.0377, "step": 24150 }, { "epoch": 7.43, "learning_rate": 3.2734688851724273e-06, "loss": 0.072, "step": 24151 }, { "epoch": 7.43, "learning_rate": 3.2727315207911703e-06, "loss": 0.0416, "step": 24152 }, { "epoch": 7.43, "learning_rate": 3.2719942232178824e-06, "loss": 0.0411, "step": 24153 }, { "epoch": 7.43, "learning_rate": 3.271256992459896e-06, "loss": 0.0481, "step": 24154 }, { "epoch": 7.43, "learning_rate": 3.2705198285245288e-06, "loss": 0.0421, "step": 24155 }, { "epoch": 7.43, "learning_rate": 3.2697827314191023e-06, "loss": 0.0309, "step": 24156 }, { "epoch": 7.43, "learning_rate": 3.269045701150936e-06, "loss": 0.0485, "step": 24157 }, { "epoch": 7.43, "learning_rate": 3.268308737727345e-06, "loss": 0.0666, "step": 24158 }, { "epoch": 7.43, "learning_rate": 3.26757184115566e-06, "loss": 0.0661, "step": 24159 }, { "epoch": 7.43, "learning_rate": 3.2668350114431844e-06, "loss": 0.0443, "step": 24160 }, { "epoch": 7.43, "learning_rate": 3.266098248597246e-06, "loss": 0.0502, "step": 24161 }, { "epoch": 7.43, "learning_rate": 3.265361552625158e-06, "loss": 0.0738, "step": 24162 }, { "epoch": 7.43, "learning_rate": 3.2646249235342363e-06, "loss": 0.051, "step": 24163 }, { "epoch": 7.43, "learning_rate": 3.2638883613317974e-06, "loss": 0.0465, "step": 24164 }, { "epoch": 7.44, "learning_rate": 3.2631518660251514e-06, "loss": 0.0353, "step": 24165 }, { "epoch": 7.44, "learning_rate": 3.2624154376216234e-06, "loss": 0.0527, "step": 24166 }, { "epoch": 7.44, "learning_rate": 3.2616790761285156e-06, "loss": 0.0217, "step": 24167 }, { "epoch": 7.44, "learning_rate": 3.2609427815531426e-06, "loss": 0.0439, "step": 24168 }, { "epoch": 7.44, "learning_rate": 3.2602065539028237e-06, "loss": 0.0776, "step": 24169 }, { "epoch": 7.44, "learning_rate": 3.2594703931848594e-06, "loss": 0.0475, "step": 24170 }, { "epoch": 7.44, "learning_rate": 3.2587342994065697e-06, "loss": 0.0426, "step": 24171 }, { "epoch": 7.44, "learning_rate": 3.2579982725752612e-06, "loss": 0.0354, "step": 24172 }, { "epoch": 7.44, "learning_rate": 3.2572623126982438e-06, "loss": 0.0536, "step": 24173 }, { "epoch": 7.44, "learning_rate": 3.256526419782826e-06, "loss": 0.0543, "step": 24174 }, { "epoch": 7.44, "learning_rate": 3.255790593836312e-06, "loss": 0.0704, "step": 24175 }, { "epoch": 7.44, "learning_rate": 3.2550548348660203e-06, "loss": 0.0669, "step": 24176 }, { "epoch": 7.44, "learning_rate": 3.2543191428792466e-06, "loss": 0.0417, "step": 24177 }, { "epoch": 7.44, "learning_rate": 3.2535835178832977e-06, "loss": 0.0271, "step": 24178 }, { "epoch": 7.44, "learning_rate": 3.2528479598854857e-06, "loss": 0.0503, "step": 24179 }, { "epoch": 7.44, "learning_rate": 3.2521124688931116e-06, "loss": 0.0507, "step": 24180 }, { "epoch": 7.44, "learning_rate": 3.2513770449134798e-06, "loss": 0.0501, "step": 24181 }, { "epoch": 7.44, "learning_rate": 3.2506416879538938e-06, "loss": 0.025, "step": 24182 }, { "epoch": 7.44, "learning_rate": 3.2499063980216562e-06, "loss": 0.0535, "step": 24183 }, { "epoch": 7.44, "learning_rate": 3.2491711751240686e-06, "loss": 0.0315, "step": 24184 }, { "epoch": 7.44, "learning_rate": 3.2484360192684294e-06, "loss": 0.0317, "step": 24185 }, { "epoch": 7.44, "learning_rate": 3.2477009304620468e-06, "loss": 0.0311, "step": 24186 }, { "epoch": 7.44, "learning_rate": 3.246965908712221e-06, "loss": 0.0691, "step": 24187 }, { "epoch": 7.44, "learning_rate": 3.2462309540262404e-06, "loss": 0.0548, "step": 24188 }, { "epoch": 7.44, "learning_rate": 3.2454960664114132e-06, "loss": 0.0463, "step": 24189 }, { "epoch": 7.44, "learning_rate": 3.2447612458750365e-06, "loss": 0.0282, "step": 24190 }, { "epoch": 7.44, "learning_rate": 3.244026492424406e-06, "loss": 0.0336, "step": 24191 }, { "epoch": 7.44, "learning_rate": 3.2432918060668195e-06, "loss": 0.0282, "step": 24192 }, { "epoch": 7.44, "learning_rate": 3.2425571868095716e-06, "loss": 0.0261, "step": 24193 }, { "epoch": 7.44, "learning_rate": 3.241822634659959e-06, "loss": 0.0378, "step": 24194 }, { "epoch": 7.44, "learning_rate": 3.2410881496252732e-06, "loss": 0.0428, "step": 24195 }, { "epoch": 7.44, "learning_rate": 3.240353731712814e-06, "loss": 0.0575, "step": 24196 }, { "epoch": 7.44, "learning_rate": 3.239619380929876e-06, "loss": 0.0112, "step": 24197 }, { "epoch": 7.45, "learning_rate": 3.2388850972837404e-06, "loss": 0.0742, "step": 24198 }, { "epoch": 7.45, "learning_rate": 3.2381508807817118e-06, "loss": 0.081, "step": 24199 }, { "epoch": 7.45, "learning_rate": 3.2374167314310758e-06, "loss": 0.0229, "step": 24200 }, { "epoch": 7.45, "learning_rate": 3.2366826492391245e-06, "loss": 0.0401, "step": 24201 }, { "epoch": 7.45, "learning_rate": 3.2359486342131476e-06, "loss": 0.0509, "step": 24202 }, { "epoch": 7.45, "learning_rate": 3.2352146863604317e-06, "loss": 0.0402, "step": 24203 }, { "epoch": 7.45, "learning_rate": 3.2344808056882747e-06, "loss": 0.0358, "step": 24204 }, { "epoch": 7.45, "learning_rate": 3.233746992203953e-06, "loss": 0.0471, "step": 24205 }, { "epoch": 7.45, "learning_rate": 3.2330132459147624e-06, "loss": 0.0588, "step": 24206 }, { "epoch": 7.45, "learning_rate": 3.232279566827987e-06, "loss": 0.0264, "step": 24207 }, { "epoch": 7.45, "learning_rate": 3.231545954950912e-06, "loss": 0.039, "step": 24208 }, { "epoch": 7.45, "learning_rate": 3.2308124102908245e-06, "loss": 0.0324, "step": 24209 }, { "epoch": 7.45, "learning_rate": 3.2300789328550086e-06, "loss": 0.0522, "step": 24210 }, { "epoch": 7.45, "learning_rate": 3.2293455226507474e-06, "loss": 0.0827, "step": 24211 }, { "epoch": 7.45, "learning_rate": 3.2286121796853264e-06, "loss": 0.052, "step": 24212 }, { "epoch": 7.45, "learning_rate": 3.227878903966023e-06, "loss": 0.0711, "step": 24213 }, { "epoch": 7.45, "learning_rate": 3.22714569550013e-06, "loss": 0.0488, "step": 24214 }, { "epoch": 7.45, "learning_rate": 3.2264125542949154e-06, "loss": 0.0395, "step": 24215 }, { "epoch": 7.45, "learning_rate": 3.2256794803576707e-06, "loss": 0.0595, "step": 24216 }, { "epoch": 7.45, "learning_rate": 3.2249464736956726e-06, "loss": 0.0491, "step": 24217 }, { "epoch": 7.45, "learning_rate": 3.2242135343161997e-06, "loss": 0.0627, "step": 24218 }, { "epoch": 7.45, "learning_rate": 3.22348066222653e-06, "loss": 0.0196, "step": 24219 }, { "epoch": 7.45, "learning_rate": 3.222747857433941e-06, "loss": 0.0472, "step": 24220 }, { "epoch": 7.45, "learning_rate": 3.2220151199457183e-06, "loss": 0.048, "step": 24221 }, { "epoch": 7.45, "learning_rate": 3.2212824497691285e-06, "loss": 0.0358, "step": 24222 }, { "epoch": 7.45, "learning_rate": 3.220549846911448e-06, "loss": 0.0304, "step": 24223 }, { "epoch": 7.45, "learning_rate": 3.219817311379959e-06, "loss": 0.0427, "step": 24224 }, { "epoch": 7.45, "learning_rate": 3.219817311379959e-06, "loss": 0.0526, "step": 24225 }, { "epoch": 7.45, "learning_rate": 3.2190848431819333e-06, "loss": 0.036, "step": 24226 }, { "epoch": 7.45, "learning_rate": 3.2183524423246436e-06, "loss": 0.0392, "step": 24227 }, { "epoch": 7.45, "learning_rate": 3.2176201088153658e-06, "loss": 0.0523, "step": 24228 }, { "epoch": 7.45, "learning_rate": 3.2168878426613703e-06, "loss": 0.0303, "step": 24229 }, { "epoch": 7.46, "learning_rate": 3.2161556438699303e-06, "loss": 0.045, "step": 24230 }, { "epoch": 7.46, "learning_rate": 3.2154235124483137e-06, "loss": 0.0543, "step": 24231 }, { "epoch": 7.46, "learning_rate": 3.2146914484038007e-06, "loss": 0.0505, "step": 24232 }, { "epoch": 7.46, "learning_rate": 3.2139594517436513e-06, "loss": 0.0314, "step": 24233 }, { "epoch": 7.46, "learning_rate": 3.2132275224751365e-06, "loss": 0.0314, "step": 24234 }, { "epoch": 7.46, "learning_rate": 3.2124956606055304e-06, "loss": 0.0595, "step": 24235 }, { "epoch": 7.46, "learning_rate": 3.211763866142097e-06, "loss": 0.0426, "step": 24236 }, { "epoch": 7.46, "learning_rate": 3.211032139092105e-06, "loss": 0.0441, "step": 24237 }, { "epoch": 7.46, "learning_rate": 3.210300479462821e-06, "loss": 0.0657, "step": 24238 }, { "epoch": 7.46, "learning_rate": 3.2095688872615117e-06, "loss": 0.0479, "step": 24239 }, { "epoch": 7.46, "learning_rate": 3.2088373624954405e-06, "loss": 0.0958, "step": 24240 }, { "epoch": 7.46, "learning_rate": 3.2081059051718697e-06, "loss": 0.0427, "step": 24241 }, { "epoch": 7.46, "learning_rate": 3.2073745152980704e-06, "loss": 0.0567, "step": 24242 }, { "epoch": 7.46, "learning_rate": 3.2066431928813068e-06, "loss": 0.0316, "step": 24243 }, { "epoch": 7.46, "learning_rate": 3.2059119379288293e-06, "loss": 0.0435, "step": 24244 }, { "epoch": 7.46, "learning_rate": 3.2051807504479115e-06, "loss": 0.0545, "step": 24245 }, { "epoch": 7.46, "learning_rate": 3.2044496304458117e-06, "loss": 0.0506, "step": 24246 }, { "epoch": 7.46, "learning_rate": 3.2037185779297896e-06, "loss": 0.0418, "step": 24247 }, { "epoch": 7.46, "learning_rate": 3.202987592907106e-06, "loss": 0.0564, "step": 24248 }, { "epoch": 7.46, "learning_rate": 3.202256675385019e-06, "loss": 0.0504, "step": 24249 }, { "epoch": 7.46, "learning_rate": 3.2015258253707893e-06, "loss": 0.0479, "step": 24250 }, { "epoch": 7.46, "learning_rate": 3.2007950428716705e-06, "loss": 0.0538, "step": 24251 }, { "epoch": 7.46, "learning_rate": 3.2000643278949263e-06, "loss": 0.0487, "step": 24252 }, { "epoch": 7.46, "learning_rate": 3.1993336804478126e-06, "loss": 0.0299, "step": 24253 }, { "epoch": 7.46, "learning_rate": 3.198603100537577e-06, "loss": 0.0569, "step": 24254 }, { "epoch": 7.46, "learning_rate": 3.1978725881714856e-06, "loss": 0.0528, "step": 24255 }, { "epoch": 7.46, "learning_rate": 3.197142143356787e-06, "loss": 0.0306, "step": 24256 }, { "epoch": 7.46, "learning_rate": 3.1964117661007378e-06, "loss": 0.0679, "step": 24257 }, { "epoch": 7.46, "learning_rate": 3.195681456410589e-06, "loss": 0.0681, "step": 24258 }, { "epoch": 7.46, "learning_rate": 3.194951214293591e-06, "loss": 0.0478, "step": 24259 }, { "epoch": 7.46, "learning_rate": 3.194221039757006e-06, "loss": 0.0522, "step": 24260 }, { "epoch": 7.46, "learning_rate": 3.193490932808072e-06, "loss": 0.038, "step": 24261 }, { "epoch": 7.46, "learning_rate": 3.19276089345405e-06, "loss": 0.038, "step": 24262 }, { "epoch": 7.47, "learning_rate": 3.192030921702186e-06, "loss": 0.039, "step": 24263 }, { "epoch": 7.47, "learning_rate": 3.1913010175597283e-06, "loss": 0.0482, "step": 24264 }, { "epoch": 7.47, "learning_rate": 3.190571181033928e-06, "loss": 0.0443, "step": 24265 }, { "epoch": 7.47, "learning_rate": 3.1898414121320277e-06, "loss": 0.0452, "step": 24266 }, { "epoch": 7.47, "learning_rate": 3.1891117108612847e-06, "loss": 0.0661, "step": 24267 }, { "epoch": 7.47, "learning_rate": 3.1883820772289364e-06, "loss": 0.0671, "step": 24268 }, { "epoch": 7.47, "learning_rate": 3.1876525112422283e-06, "loss": 0.046, "step": 24269 }, { "epoch": 7.47, "learning_rate": 3.186923012908416e-06, "loss": 0.0416, "step": 24270 }, { "epoch": 7.47, "learning_rate": 3.1861935822347314e-06, "loss": 0.054, "step": 24271 }, { "epoch": 7.47, "learning_rate": 3.1854642192284267e-06, "loss": 0.0904, "step": 24272 }, { "epoch": 7.47, "learning_rate": 3.184734923896744e-06, "loss": 0.063, "step": 24273 }, { "epoch": 7.47, "learning_rate": 3.1840056962469235e-06, "loss": 0.0703, "step": 24274 }, { "epoch": 7.47, "learning_rate": 3.1832765362862083e-06, "loss": 0.048, "step": 24275 }, { "epoch": 7.47, "learning_rate": 3.1825474440218363e-06, "loss": 0.0363, "step": 24276 }, { "epoch": 7.47, "learning_rate": 3.181818419461058e-06, "loss": 0.0451, "step": 24277 }, { "epoch": 7.47, "learning_rate": 3.1810894626111042e-06, "loss": 0.0695, "step": 24278 }, { "epoch": 7.47, "learning_rate": 3.180360573479212e-06, "loss": 0.0773, "step": 24279 }, { "epoch": 7.47, "learning_rate": 3.1796317520726284e-06, "loss": 0.0496, "step": 24280 }, { "epoch": 7.47, "learning_rate": 3.1789029983985877e-06, "loss": 0.0651, "step": 24281 }, { "epoch": 7.47, "learning_rate": 3.178174312464326e-06, "loss": 0.057, "step": 24282 }, { "epoch": 7.47, "learning_rate": 3.1774456942770804e-06, "loss": 0.0428, "step": 24283 }, { "epoch": 7.47, "learning_rate": 3.176717143844088e-06, "loss": 0.0478, "step": 24284 }, { "epoch": 7.47, "learning_rate": 3.175988661172582e-06, "loss": 0.0453, "step": 24285 }, { "epoch": 7.47, "learning_rate": 3.1752602462697947e-06, "loss": 0.0424, "step": 24286 }, { "epoch": 7.47, "learning_rate": 3.1745318991429665e-06, "loss": 0.0454, "step": 24287 }, { "epoch": 7.47, "learning_rate": 3.1738036197993294e-06, "loss": 0.0376, "step": 24288 }, { "epoch": 7.47, "learning_rate": 3.173075408246108e-06, "loss": 0.0302, "step": 24289 }, { "epoch": 7.47, "learning_rate": 3.172347264490543e-06, "loss": 0.0424, "step": 24290 }, { "epoch": 7.47, "learning_rate": 3.1716191885398615e-06, "loss": 0.0764, "step": 24291 }, { "epoch": 7.47, "learning_rate": 3.1708911804012956e-06, "loss": 0.0409, "step": 24292 }, { "epoch": 7.47, "learning_rate": 3.1701632400820735e-06, "loss": 0.0684, "step": 24293 }, { "epoch": 7.47, "learning_rate": 3.169435367589425e-06, "loss": 0.0375, "step": 24294 }, { "epoch": 7.48, "learning_rate": 3.1687075629305787e-06, "loss": 0.0332, "step": 24295 }, { "epoch": 7.48, "learning_rate": 3.1679798261127594e-06, "loss": 0.0439, "step": 24296 }, { "epoch": 7.48, "learning_rate": 3.167252157143199e-06, "loss": 0.0309, "step": 24297 }, { "epoch": 7.48, "learning_rate": 3.1665245560291267e-06, "loss": 0.0666, "step": 24298 }, { "epoch": 7.48, "learning_rate": 3.1657970227777556e-06, "loss": 0.0516, "step": 24299 }, { "epoch": 7.48, "learning_rate": 3.165069557396323e-06, "loss": 0.0266, "step": 24300 }, { "epoch": 7.48, "learning_rate": 3.1643421598920486e-06, "loss": 0.0457, "step": 24301 }, { "epoch": 7.48, "learning_rate": 3.163614830272156e-06, "loss": 0.0443, "step": 24302 }, { "epoch": 7.48, "learning_rate": 3.1628875685438697e-06, "loss": 0.0871, "step": 24303 }, { "epoch": 7.48, "learning_rate": 3.162160374714407e-06, "loss": 0.0375, "step": 24304 }, { "epoch": 7.48, "learning_rate": 3.1614332487910003e-06, "loss": 0.0409, "step": 24305 }, { "epoch": 7.48, "learning_rate": 3.160706190780858e-06, "loss": 0.0332, "step": 24306 }, { "epoch": 7.48, "learning_rate": 3.15997920069121e-06, "loss": 0.0703, "step": 24307 }, { "epoch": 7.48, "learning_rate": 3.1592522785292714e-06, "loss": 0.0401, "step": 24308 }, { "epoch": 7.48, "learning_rate": 3.1585254243022635e-06, "loss": 0.0914, "step": 24309 }, { "epoch": 7.48, "learning_rate": 3.157798638017402e-06, "loss": 0.0463, "step": 24310 }, { "epoch": 7.48, "learning_rate": 3.1570719196819065e-06, "loss": 0.0437, "step": 24311 }, { "epoch": 7.48, "learning_rate": 3.156345269302994e-06, "loss": 0.042, "step": 24312 }, { "epoch": 7.48, "learning_rate": 3.1556186868878804e-06, "loss": 0.0429, "step": 24313 }, { "epoch": 7.48, "learning_rate": 3.1548921724437775e-06, "loss": 0.0596, "step": 24314 }, { "epoch": 7.48, "learning_rate": 3.15416572597791e-06, "loss": 0.0617, "step": 24315 }, { "epoch": 7.48, "learning_rate": 3.1534393474974814e-06, "loss": 0.0233, "step": 24316 }, { "epoch": 7.48, "learning_rate": 3.1527130370097113e-06, "loss": 0.0518, "step": 24317 }, { "epoch": 7.48, "learning_rate": 3.151986794521812e-06, "loss": 0.0856, "step": 24318 }, { "epoch": 7.48, "learning_rate": 3.1512606200409956e-06, "loss": 0.0395, "step": 24319 }, { "epoch": 7.48, "learning_rate": 3.1505345135744737e-06, "loss": 0.0349, "step": 24320 }, { "epoch": 7.48, "learning_rate": 3.1498084751294523e-06, "loss": 0.0456, "step": 24321 }, { "epoch": 7.48, "learning_rate": 3.1490825047131523e-06, "loss": 0.0501, "step": 24322 }, { "epoch": 7.48, "learning_rate": 3.148356602332775e-06, "loss": 0.0745, "step": 24323 }, { "epoch": 7.48, "learning_rate": 3.1476307679955276e-06, "loss": 0.0435, "step": 24324 }, { "epoch": 7.48, "learning_rate": 3.1469050017086254e-06, "loss": 0.0691, "step": 24325 }, { "epoch": 7.48, "learning_rate": 3.146179303479272e-06, "loss": 0.0672, "step": 24326 }, { "epoch": 7.48, "learning_rate": 3.1454536733146755e-06, "loss": 0.0718, "step": 24327 }, { "epoch": 7.49, "learning_rate": 3.1447281112220405e-06, "loss": 0.0238, "step": 24328 }, { "epoch": 7.49, "learning_rate": 3.1440026172085738e-06, "loss": 0.068, "step": 24329 }, { "epoch": 7.49, "learning_rate": 3.1432771912814796e-06, "loss": 0.0355, "step": 24330 }, { "epoch": 7.49, "learning_rate": 3.14255183344796e-06, "loss": 0.0418, "step": 24331 }, { "epoch": 7.49, "learning_rate": 3.141826543715226e-06, "loss": 0.0344, "step": 24332 }, { "epoch": 7.49, "learning_rate": 3.141101322090472e-06, "loss": 0.0561, "step": 24333 }, { "epoch": 7.49, "learning_rate": 3.1403761685809007e-06, "loss": 0.0541, "step": 24334 }, { "epoch": 7.49, "learning_rate": 3.139651083193719e-06, "loss": 0.0375, "step": 24335 }, { "epoch": 7.49, "learning_rate": 3.1389260659361244e-06, "loss": 0.0915, "step": 24336 }, { "epoch": 7.49, "learning_rate": 3.138201116815317e-06, "loss": 0.0482, "step": 24337 }, { "epoch": 7.49, "learning_rate": 3.1374762358384935e-06, "loss": 0.0378, "step": 24338 }, { "epoch": 7.49, "learning_rate": 3.1367514230128625e-06, "loss": 0.0473, "step": 24339 }, { "epoch": 7.49, "learning_rate": 3.136026678345613e-06, "loss": 0.0609, "step": 24340 }, { "epoch": 7.49, "learning_rate": 3.1353020018439394e-06, "loss": 0.0182, "step": 24341 }, { "epoch": 7.49, "learning_rate": 3.134577393515048e-06, "loss": 0.0537, "step": 24342 }, { "epoch": 7.49, "learning_rate": 3.1338528533661306e-06, "loss": 0.0469, "step": 24343 }, { "epoch": 7.49, "learning_rate": 3.133128381404381e-06, "loss": 0.0334, "step": 24344 }, { "epoch": 7.49, "learning_rate": 3.1324039776369965e-06, "loss": 0.0598, "step": 24345 }, { "epoch": 7.49, "learning_rate": 3.131679642071169e-06, "loss": 0.0379, "step": 24346 }, { "epoch": 7.49, "learning_rate": 3.130955374714094e-06, "loss": 0.0473, "step": 24347 }, { "epoch": 7.49, "learning_rate": 3.1302311755729587e-06, "loss": 0.057, "step": 24348 }, { "epoch": 7.49, "learning_rate": 3.1295070446549657e-06, "loss": 0.0551, "step": 24349 }, { "epoch": 7.49, "learning_rate": 3.1287829819672964e-06, "loss": 0.0339, "step": 24350 }, { "epoch": 7.49, "learning_rate": 3.1280589875171417e-06, "loss": 0.0553, "step": 24351 }, { "epoch": 7.49, "learning_rate": 3.127335061311698e-06, "loss": 0.0361, "step": 24352 }, { "epoch": 7.49, "learning_rate": 3.126611203358152e-06, "loss": 0.0623, "step": 24353 }, { "epoch": 7.49, "learning_rate": 3.125887413663691e-06, "loss": 0.0645, "step": 24354 }, { "epoch": 7.49, "learning_rate": 3.125163692235502e-06, "loss": 0.0427, "step": 24355 }, { "epoch": 7.49, "learning_rate": 3.124440039080775e-06, "loss": 0.0193, "step": 24356 }, { "epoch": 7.49, "learning_rate": 3.1237164542066943e-06, "loss": 0.0575, "step": 24357 }, { "epoch": 7.49, "learning_rate": 3.1229929376204437e-06, "loss": 0.0364, "step": 24358 }, { "epoch": 7.49, "learning_rate": 3.122269489329215e-06, "loss": 0.0431, "step": 24359 }, { "epoch": 7.5, "learning_rate": 3.1215461093401913e-06, "loss": 0.0481, "step": 24360 }, { "epoch": 7.5, "learning_rate": 3.1208227976605475e-06, "loss": 0.0215, "step": 24361 }, { "epoch": 7.5, "learning_rate": 3.120099554297478e-06, "loss": 0.0382, "step": 24362 }, { "epoch": 7.5, "learning_rate": 3.11937637925816e-06, "loss": 0.0648, "step": 24363 }, { "epoch": 7.5, "learning_rate": 3.1186532725497754e-06, "loss": 0.0578, "step": 24364 }, { "epoch": 7.5, "learning_rate": 3.1179302341795058e-06, "loss": 0.0651, "step": 24365 }, { "epoch": 7.5, "learning_rate": 3.1172072641545294e-06, "loss": 0.0422, "step": 24366 }, { "epoch": 7.5, "learning_rate": 3.116484362482034e-06, "loss": 0.043, "step": 24367 }, { "epoch": 7.5, "learning_rate": 3.1157615291691855e-06, "loss": 0.0467, "step": 24368 }, { "epoch": 7.5, "learning_rate": 3.115038764223175e-06, "loss": 0.0575, "step": 24369 }, { "epoch": 7.5, "learning_rate": 3.1143160676511764e-06, "loss": 0.0396, "step": 24370 }, { "epoch": 7.5, "learning_rate": 3.1135934394603596e-06, "loss": 0.0536, "step": 24371 }, { "epoch": 7.5, "learning_rate": 3.1128708796579098e-06, "loss": 0.0396, "step": 24372 }, { "epoch": 7.5, "learning_rate": 3.1121483882509996e-06, "loss": 0.047, "step": 24373 }, { "epoch": 7.5, "learning_rate": 3.111425965246804e-06, "loss": 0.0403, "step": 24374 }, { "epoch": 7.5, "learning_rate": 3.110703610652497e-06, "loss": 0.0404, "step": 24375 }, { "epoch": 7.5, "learning_rate": 3.1099813244752497e-06, "loss": 0.0589, "step": 24376 }, { "epoch": 7.5, "learning_rate": 3.1092591067222444e-06, "loss": 0.0706, "step": 24377 }, { "epoch": 7.5, "learning_rate": 3.1085369574006395e-06, "loss": 0.0572, "step": 24378 }, { "epoch": 7.5, "learning_rate": 3.1078148765176175e-06, "loss": 0.0303, "step": 24379 }, { "epoch": 7.5, "learning_rate": 3.1070928640803464e-06, "loss": 0.0763, "step": 24380 }, { "epoch": 7.5, "learning_rate": 3.106370920095996e-06, "loss": 0.0398, "step": 24381 }, { "epoch": 7.5, "learning_rate": 3.1056490445717345e-06, "loss": 0.0229, "step": 24382 }, { "epoch": 7.5, "learning_rate": 3.1049272375147298e-06, "loss": 0.042, "step": 24383 }, { "epoch": 7.5, "learning_rate": 3.104205498932159e-06, "loss": 0.0467, "step": 24384 }, { "epoch": 7.5, "learning_rate": 3.1034838288311784e-06, "loss": 0.0712, "step": 24385 }, { "epoch": 7.5, "learning_rate": 3.1027622272189572e-06, "loss": 0.0397, "step": 24386 }, { "epoch": 7.5, "learning_rate": 3.102040694102667e-06, "loss": 0.0598, "step": 24387 }, { "epoch": 7.5, "learning_rate": 3.10131922948947e-06, "loss": 0.0515, "step": 24388 }, { "epoch": 7.5, "learning_rate": 3.1005978333865304e-06, "loss": 0.0331, "step": 24389 }, { "epoch": 7.5, "learning_rate": 3.099876505801014e-06, "loss": 0.0373, "step": 24390 }, { "epoch": 7.5, "learning_rate": 3.099155246740082e-06, "loss": 0.0795, "step": 24391 }, { "epoch": 7.5, "learning_rate": 3.0984340562108995e-06, "loss": 0.0684, "step": 24392 }, { "epoch": 7.51, "learning_rate": 3.0977129342206234e-06, "loss": 0.037, "step": 24393 }, { "epoch": 7.51, "learning_rate": 3.0969918807764263e-06, "loss": 0.0435, "step": 24394 }, { "epoch": 7.51, "learning_rate": 3.0962708958854593e-06, "loss": 0.066, "step": 24395 }, { "epoch": 7.51, "learning_rate": 3.0955499795548803e-06, "loss": 0.043, "step": 24396 }, { "epoch": 7.51, "learning_rate": 3.0948291317918575e-06, "loss": 0.0415, "step": 24397 }, { "epoch": 7.51, "learning_rate": 3.0941083526035452e-06, "loss": 0.0587, "step": 24398 }, { "epoch": 7.51, "learning_rate": 3.0933876419971008e-06, "loss": 0.0664, "step": 24399 }, { "epoch": 7.51, "learning_rate": 3.0926669999796833e-06, "loss": 0.0774, "step": 24400 }, { "epoch": 7.51, "learning_rate": 3.091946426558449e-06, "loss": 0.054, "step": 24401 }, { "epoch": 7.51, "learning_rate": 3.0912259217405516e-06, "loss": 0.0501, "step": 24402 }, { "epoch": 7.51, "learning_rate": 3.090505485533146e-06, "loss": 0.0632, "step": 24403 }, { "epoch": 7.51, "learning_rate": 3.0897851179433924e-06, "loss": 0.0733, "step": 24404 }, { "epoch": 7.51, "learning_rate": 3.0890648189784445e-06, "loss": 0.0973, "step": 24405 }, { "epoch": 7.51, "learning_rate": 3.0883445886454443e-06, "loss": 0.0372, "step": 24406 }, { "epoch": 7.51, "learning_rate": 3.0876244269515574e-06, "loss": 0.0503, "step": 24407 }, { "epoch": 7.51, "learning_rate": 3.0869043339039294e-06, "loss": 0.0776, "step": 24408 }, { "epoch": 7.51, "learning_rate": 3.0861843095097134e-06, "loss": 0.0725, "step": 24409 }, { "epoch": 7.51, "learning_rate": 3.0854643537760585e-06, "loss": 0.05, "step": 24410 }, { "epoch": 7.51, "learning_rate": 3.0847444667101156e-06, "loss": 0.04, "step": 24411 }, { "epoch": 7.51, "learning_rate": 3.0840246483190338e-06, "loss": 0.0516, "step": 24412 }, { "epoch": 7.51, "learning_rate": 3.083304898609957e-06, "loss": 0.0655, "step": 24413 }, { "epoch": 7.51, "learning_rate": 3.0825852175900406e-06, "loss": 0.0552, "step": 24414 }, { "epoch": 7.51, "learning_rate": 3.081865605266432e-06, "loss": 0.0518, "step": 24415 }, { "epoch": 7.51, "learning_rate": 3.0811460616462672e-06, "loss": 0.0212, "step": 24416 }, { "epoch": 7.51, "learning_rate": 3.080426586736702e-06, "loss": 0.0293, "step": 24417 }, { "epoch": 7.51, "learning_rate": 3.0797071805448774e-06, "loss": 0.0344, "step": 24418 }, { "epoch": 7.51, "learning_rate": 3.0789878430779394e-06, "loss": 0.042, "step": 24419 }, { "epoch": 7.51, "learning_rate": 3.0782685743430305e-06, "loss": 0.0633, "step": 24420 }, { "epoch": 7.51, "learning_rate": 3.0775493743472906e-06, "loss": 0.0609, "step": 24421 }, { "epoch": 7.51, "learning_rate": 3.0768302430978713e-06, "loss": 0.0427, "step": 24422 }, { "epoch": 7.51, "learning_rate": 3.0761111806019016e-06, "loss": 0.0194, "step": 24423 }, { "epoch": 7.51, "learning_rate": 3.075392186866534e-06, "loss": 0.0339, "step": 24424 }, { "epoch": 7.52, "learning_rate": 3.074673261898903e-06, "loss": 0.0502, "step": 24425 }, { "epoch": 7.52, "learning_rate": 3.073954405706148e-06, "loss": 0.0337, "step": 24426 }, { "epoch": 7.52, "learning_rate": 3.0732356182954103e-06, "loss": 0.0533, "step": 24427 }, { "epoch": 7.52, "learning_rate": 3.0725168996738264e-06, "loss": 0.0426, "step": 24428 }, { "epoch": 7.52, "learning_rate": 3.071798249848534e-06, "loss": 0.0603, "step": 24429 }, { "epoch": 7.52, "learning_rate": 3.07107966882667e-06, "loss": 0.0495, "step": 24430 }, { "epoch": 7.52, "learning_rate": 3.0703611566153677e-06, "loss": 0.0527, "step": 24431 }, { "epoch": 7.52, "learning_rate": 3.069642713221773e-06, "loss": 0.0281, "step": 24432 }, { "epoch": 7.52, "learning_rate": 3.0689243386530064e-06, "loss": 0.0208, "step": 24433 }, { "epoch": 7.52, "learning_rate": 3.0682060329162122e-06, "loss": 0.0526, "step": 24434 }, { "epoch": 7.52, "learning_rate": 3.06748779601852e-06, "loss": 0.1181, "step": 24435 }, { "epoch": 7.52, "learning_rate": 3.0667696279670635e-06, "loss": 0.0676, "step": 24436 }, { "epoch": 7.52, "learning_rate": 3.066051528768975e-06, "loss": 0.0403, "step": 24437 }, { "epoch": 7.52, "learning_rate": 3.065333498431381e-06, "loss": 0.053, "step": 24438 }, { "epoch": 7.52, "learning_rate": 3.0646155369614238e-06, "loss": 0.0592, "step": 24439 }, { "epoch": 7.52, "learning_rate": 3.0638976443662227e-06, "loss": 0.04, "step": 24440 }, { "epoch": 7.52, "learning_rate": 3.0631798206529085e-06, "loss": 0.06, "step": 24441 }, { "epoch": 7.52, "learning_rate": 3.062462065828613e-06, "loss": 0.0213, "step": 24442 }, { "epoch": 7.52, "learning_rate": 3.061744379900464e-06, "loss": 0.0534, "step": 24443 }, { "epoch": 7.52, "learning_rate": 3.0610267628755884e-06, "loss": 0.0526, "step": 24444 }, { "epoch": 7.52, "learning_rate": 3.060309214761111e-06, "loss": 0.0521, "step": 24445 }, { "epoch": 7.52, "learning_rate": 3.05959173556416e-06, "loss": 0.0447, "step": 24446 }, { "epoch": 7.52, "learning_rate": 3.0588743252918583e-06, "loss": 0.0509, "step": 24447 }, { "epoch": 7.52, "learning_rate": 3.058156983951329e-06, "loss": 0.0456, "step": 24448 }, { "epoch": 7.52, "learning_rate": 3.0574397115497056e-06, "loss": 0.0318, "step": 24449 }, { "epoch": 7.52, "learning_rate": 3.056722508094101e-06, "loss": 0.0675, "step": 24450 }, { "epoch": 7.52, "learning_rate": 3.0560053735916372e-06, "loss": 0.048, "step": 24451 }, { "epoch": 7.52, "learning_rate": 3.055288308049443e-06, "loss": 0.0307, "step": 24452 }, { "epoch": 7.52, "learning_rate": 3.0545713114746376e-06, "loss": 0.0614, "step": 24453 }, { "epoch": 7.52, "learning_rate": 3.0538543838743395e-06, "loss": 0.0582, "step": 24454 }, { "epoch": 7.52, "learning_rate": 3.053137525255668e-06, "loss": 0.0493, "step": 24455 }, { "epoch": 7.52, "learning_rate": 3.0524207356257453e-06, "loss": 0.0184, "step": 24456 }, { "epoch": 7.52, "learning_rate": 3.0517040149916853e-06, "loss": 0.0791, "step": 24457 }, { "epoch": 7.53, "learning_rate": 3.050987363360607e-06, "loss": 0.0498, "step": 24458 }, { "epoch": 7.53, "learning_rate": 3.0502707807396314e-06, "loss": 0.0276, "step": 24459 }, { "epoch": 7.53, "learning_rate": 3.0495542671358745e-06, "loss": 0.0271, "step": 24460 }, { "epoch": 7.53, "learning_rate": 3.048837822556443e-06, "loss": 0.0416, "step": 24461 }, { "epoch": 7.53, "learning_rate": 3.048121447008463e-06, "loss": 0.0761, "step": 24462 }, { "epoch": 7.53, "learning_rate": 3.0474051404990424e-06, "loss": 0.0461, "step": 24463 }, { "epoch": 7.53, "learning_rate": 3.0466889030352976e-06, "loss": 0.0407, "step": 24464 }, { "epoch": 7.53, "learning_rate": 3.0459727346243396e-06, "loss": 0.0646, "step": 24465 }, { "epoch": 7.53, "learning_rate": 3.045256635273278e-06, "loss": 0.0537, "step": 24466 }, { "epoch": 7.53, "learning_rate": 3.0445406049892347e-06, "loss": 0.026, "step": 24467 }, { "epoch": 7.53, "learning_rate": 3.0438246437793073e-06, "loss": 0.0535, "step": 24468 }, { "epoch": 7.53, "learning_rate": 3.043108751650615e-06, "loss": 0.0559, "step": 24469 }, { "epoch": 7.53, "learning_rate": 3.042392928610267e-06, "loss": 0.0749, "step": 24470 }, { "epoch": 7.53, "learning_rate": 3.041677174665364e-06, "loss": 0.0657, "step": 24471 }, { "epoch": 7.53, "learning_rate": 3.040961489823023e-06, "loss": 0.0627, "step": 24472 }, { "epoch": 7.53, "learning_rate": 3.0402458740903474e-06, "loss": 0.0556, "step": 24473 }, { "epoch": 7.53, "learning_rate": 3.0395303274744446e-06, "loss": 0.0633, "step": 24474 }, { "epoch": 7.53, "learning_rate": 3.038814849982421e-06, "loss": 0.0565, "step": 24475 }, { "epoch": 7.53, "learning_rate": 3.0380994416213785e-06, "loss": 0.0392, "step": 24476 }, { "epoch": 7.53, "learning_rate": 3.037384102398431e-06, "loss": 0.0459, "step": 24477 }, { "epoch": 7.53, "learning_rate": 3.0366688323206694e-06, "loss": 0.0672, "step": 24478 }, { "epoch": 7.53, "learning_rate": 3.0359536313952076e-06, "loss": 0.0294, "step": 24479 }, { "epoch": 7.53, "learning_rate": 3.035238499629145e-06, "loss": 0.0638, "step": 24480 }, { "epoch": 7.53, "learning_rate": 3.034523437029582e-06, "loss": 0.0294, "step": 24481 }, { "epoch": 7.53, "learning_rate": 3.033808443603622e-06, "loss": 0.0379, "step": 24482 }, { "epoch": 7.53, "learning_rate": 3.03309351935836e-06, "loss": 0.0568, "step": 24483 }, { "epoch": 7.53, "learning_rate": 3.032378664300907e-06, "loss": 0.026, "step": 24484 }, { "epoch": 7.53, "learning_rate": 3.0316638784383524e-06, "loss": 0.0396, "step": 24485 }, { "epoch": 7.53, "learning_rate": 3.0309491617777943e-06, "loss": 0.0398, "step": 24486 }, { "epoch": 7.53, "learning_rate": 3.03023451432634e-06, "loss": 0.0379, "step": 24487 }, { "epoch": 7.53, "learning_rate": 3.0295199360910754e-06, "loss": 0.0433, "step": 24488 }, { "epoch": 7.53, "learning_rate": 3.0288054270791036e-06, "loss": 0.0677, "step": 24489 }, { "epoch": 7.54, "learning_rate": 3.0280909872975194e-06, "loss": 0.0338, "step": 24490 }, { "epoch": 7.54, "learning_rate": 3.027376616753418e-06, "loss": 0.0466, "step": 24491 }, { "epoch": 7.54, "learning_rate": 3.0266623154538923e-06, "loss": 0.0453, "step": 24492 }, { "epoch": 7.54, "learning_rate": 3.025948083406033e-06, "loss": 0.0219, "step": 24493 }, { "epoch": 7.54, "learning_rate": 3.0252339206169433e-06, "loss": 0.0507, "step": 24494 }, { "epoch": 7.54, "learning_rate": 3.024519827093706e-06, "loss": 0.0678, "step": 24495 }, { "epoch": 7.54, "learning_rate": 3.0238058028434126e-06, "loss": 0.0489, "step": 24496 }, { "epoch": 7.54, "learning_rate": 3.0230918478731598e-06, "loss": 0.0551, "step": 24497 }, { "epoch": 7.54, "learning_rate": 3.022377962190035e-06, "loss": 0.0817, "step": 24498 }, { "epoch": 7.54, "learning_rate": 3.0216641458011276e-06, "loss": 0.0626, "step": 24499 }, { "epoch": 7.54, "learning_rate": 3.020950398713527e-06, "loss": 0.0522, "step": 24500 }, { "epoch": 7.54, "learning_rate": 3.0202367209343207e-06, "loss": 0.0575, "step": 24501 }, { "epoch": 7.54, "learning_rate": 3.019523112470597e-06, "loss": 0.0367, "step": 24502 }, { "epoch": 7.54, "learning_rate": 3.0188095733294388e-06, "loss": 0.0495, "step": 24503 }, { "epoch": 7.54, "learning_rate": 3.018096103517939e-06, "loss": 0.0571, "step": 24504 }, { "epoch": 7.54, "learning_rate": 3.0173827030431823e-06, "loss": 0.0557, "step": 24505 }, { "epoch": 7.54, "learning_rate": 3.0166693719122454e-06, "loss": 0.0914, "step": 24506 }, { "epoch": 7.54, "learning_rate": 3.01595611013222e-06, "loss": 0.0396, "step": 24507 }, { "epoch": 7.54, "learning_rate": 3.0152429177101872e-06, "loss": 0.0587, "step": 24508 }, { "epoch": 7.54, "learning_rate": 3.0145297946532292e-06, "loss": 0.0575, "step": 24509 }, { "epoch": 7.54, "learning_rate": 3.0138167409684293e-06, "loss": 0.0289, "step": 24510 }, { "epoch": 7.54, "learning_rate": 3.0131037566628673e-06, "loss": 0.0144, "step": 24511 }, { "epoch": 7.54, "learning_rate": 3.0123908417436245e-06, "loss": 0.0392, "step": 24512 }, { "epoch": 7.54, "learning_rate": 3.0116779962177765e-06, "loss": 0.0533, "step": 24513 }, { "epoch": 7.54, "learning_rate": 3.010965220092411e-06, "loss": 0.067, "step": 24514 }, { "epoch": 7.54, "learning_rate": 3.010252513374604e-06, "loss": 0.0333, "step": 24515 }, { "epoch": 7.54, "learning_rate": 3.009539876071427e-06, "loss": 0.0763, "step": 24516 }, { "epoch": 7.54, "learning_rate": 3.008827308189963e-06, "loss": 0.0632, "step": 24517 }, { "epoch": 7.54, "learning_rate": 3.0081148097372883e-06, "loss": 0.0315, "step": 24518 }, { "epoch": 7.54, "learning_rate": 3.0074023807204764e-06, "loss": 0.042, "step": 24519 }, { "epoch": 7.54, "learning_rate": 3.0066900211466045e-06, "loss": 0.039, "step": 24520 }, { "epoch": 7.54, "learning_rate": 3.0059777310227423e-06, "loss": 0.0716, "step": 24521 }, { "epoch": 7.54, "learning_rate": 3.0052655103559735e-06, "loss": 0.033, "step": 24522 }, { "epoch": 7.55, "learning_rate": 3.004553359153358e-06, "loss": 0.0709, "step": 24523 }, { "epoch": 7.55, "learning_rate": 3.0038412774219783e-06, "loss": 0.0384, "step": 24524 }, { "epoch": 7.55, "learning_rate": 3.0031292651689025e-06, "loss": 0.0515, "step": 24525 }, { "epoch": 7.55, "learning_rate": 3.002417322401202e-06, "loss": 0.0752, "step": 24526 }, { "epoch": 7.55, "learning_rate": 3.0017054491259455e-06, "loss": 0.0716, "step": 24527 }, { "epoch": 7.55, "learning_rate": 3.000993645350205e-06, "loss": 0.0814, "step": 24528 }, { "epoch": 7.55, "learning_rate": 3.0002819110810475e-06, "loss": 0.0376, "step": 24529 }, { "epoch": 7.55, "learning_rate": 2.9995702463255418e-06, "loss": 0.0408, "step": 24530 }, { "epoch": 7.55, "learning_rate": 2.9988586510907526e-06, "loss": 0.019, "step": 24531 }, { "epoch": 7.55, "learning_rate": 2.9981471253837545e-06, "loss": 0.0693, "step": 24532 }, { "epoch": 7.55, "learning_rate": 2.9974356692116036e-06, "loss": 0.033, "step": 24533 }, { "epoch": 7.55, "learning_rate": 2.9967242825813736e-06, "loss": 0.0352, "step": 24534 }, { "epoch": 7.55, "learning_rate": 2.9960129655001248e-06, "loss": 0.0316, "step": 24535 }, { "epoch": 7.55, "learning_rate": 2.9953017179749223e-06, "loss": 0.045, "step": 24536 }, { "epoch": 7.55, "learning_rate": 2.9945905400128305e-06, "loss": 0.0411, "step": 24537 }, { "epoch": 7.55, "learning_rate": 2.9938794316209076e-06, "loss": 0.0654, "step": 24538 }, { "epoch": 7.55, "learning_rate": 2.993168392806224e-06, "loss": 0.0334, "step": 24539 }, { "epoch": 7.55, "learning_rate": 2.9924574235758308e-06, "loss": 0.0276, "step": 24540 }, { "epoch": 7.55, "learning_rate": 2.9917465239367972e-06, "loss": 0.0638, "step": 24541 }, { "epoch": 7.55, "learning_rate": 2.9910356938961782e-06, "loss": 0.0401, "step": 24542 }, { "epoch": 7.55, "learning_rate": 2.990324933461035e-06, "loss": 0.061, "step": 24543 }, { "epoch": 7.55, "learning_rate": 2.989614242638427e-06, "loss": 0.0335, "step": 24544 }, { "epoch": 7.55, "learning_rate": 2.988903621435408e-06, "loss": 0.0544, "step": 24545 }, { "epoch": 7.55, "learning_rate": 2.98819306985904e-06, "loss": 0.0671, "step": 24546 }, { "epoch": 7.55, "learning_rate": 2.987482587916376e-06, "loss": 0.0871, "step": 24547 }, { "epoch": 7.55, "learning_rate": 2.9867721756144698e-06, "loss": 0.0567, "step": 24548 }, { "epoch": 7.55, "learning_rate": 2.986061832960385e-06, "loss": 0.0355, "step": 24549 }, { "epoch": 7.55, "learning_rate": 2.9853515599611647e-06, "loss": 0.047, "step": 24550 }, { "epoch": 7.55, "learning_rate": 2.984641356623872e-06, "loss": 0.0378, "step": 24551 }, { "epoch": 7.55, "learning_rate": 2.983931222955556e-06, "loss": 0.0395, "step": 24552 }, { "epoch": 7.55, "learning_rate": 2.9832211589632687e-06, "loss": 0.0788, "step": 24553 }, { "epoch": 7.55, "learning_rate": 2.982511164654063e-06, "loss": 0.0497, "step": 24554 }, { "epoch": 7.56, "learning_rate": 2.981801240034985e-06, "loss": 0.0513, "step": 24555 }, { "epoch": 7.56, "learning_rate": 2.981091385113095e-06, "loss": 0.0391, "step": 24556 }, { "epoch": 7.56, "learning_rate": 2.9803815998954334e-06, "loss": 0.0336, "step": 24557 }, { "epoch": 7.56, "learning_rate": 2.979671884389049e-06, "loss": 0.0286, "step": 24558 }, { "epoch": 7.56, "learning_rate": 2.9789622386009964e-06, "loss": 0.0182, "step": 24559 }, { "epoch": 7.56, "learning_rate": 2.978252662538319e-06, "loss": 0.0587, "step": 24560 }, { "epoch": 7.56, "learning_rate": 2.977543156208065e-06, "loss": 0.0311, "step": 24561 }, { "epoch": 7.56, "learning_rate": 2.9768337196172783e-06, "loss": 0.0361, "step": 24562 }, { "epoch": 7.56, "learning_rate": 2.9761243527730057e-06, "loss": 0.0308, "step": 24563 }, { "epoch": 7.56, "learning_rate": 2.9754150556822927e-06, "loss": 0.0298, "step": 24564 }, { "epoch": 7.56, "learning_rate": 2.9747058283521777e-06, "loss": 0.0585, "step": 24565 }, { "epoch": 7.56, "learning_rate": 2.973996670789716e-06, "loss": 0.0505, "step": 24566 }, { "epoch": 7.56, "learning_rate": 2.9732875830019392e-06, "loss": 0.06, "step": 24567 }, { "epoch": 7.56, "learning_rate": 2.9725785649958895e-06, "loss": 0.0392, "step": 24568 }, { "epoch": 7.56, "learning_rate": 2.9718696167786145e-06, "loss": 0.0428, "step": 24569 }, { "epoch": 7.56, "learning_rate": 2.9711607383571507e-06, "loss": 0.0436, "step": 24570 }, { "epoch": 7.56, "learning_rate": 2.9704519297385392e-06, "loss": 0.0272, "step": 24571 }, { "epoch": 7.56, "learning_rate": 2.9697431909298193e-06, "loss": 0.0682, "step": 24572 }, { "epoch": 7.56, "learning_rate": 2.969034521938028e-06, "loss": 0.05, "step": 24573 }, { "epoch": 7.56, "learning_rate": 2.9683259227702043e-06, "loss": 0.0506, "step": 24574 }, { "epoch": 7.56, "learning_rate": 2.967617393433382e-06, "loss": 0.0488, "step": 24575 }, { "epoch": 7.56, "learning_rate": 2.9669089339346026e-06, "loss": 0.052, "step": 24576 }, { "epoch": 7.56, "learning_rate": 2.9662005442809027e-06, "loss": 0.0528, "step": 24577 }, { "epoch": 7.56, "learning_rate": 2.9654922244793084e-06, "loss": 0.0298, "step": 24578 }, { "epoch": 7.56, "learning_rate": 2.9647839745368623e-06, "loss": 0.026, "step": 24579 }, { "epoch": 7.56, "learning_rate": 2.9640757944605947e-06, "loss": 0.0373, "step": 24580 }, { "epoch": 7.56, "learning_rate": 2.9633676842575386e-06, "loss": 0.0583, "step": 24581 }, { "epoch": 7.56, "learning_rate": 2.962659643934728e-06, "loss": 0.0519, "step": 24582 }, { "epoch": 7.56, "learning_rate": 2.9619516734991884e-06, "loss": 0.0462, "step": 24583 }, { "epoch": 7.56, "learning_rate": 2.9612437729579626e-06, "loss": 0.0217, "step": 24584 }, { "epoch": 7.56, "learning_rate": 2.9605359423180655e-06, "loss": 0.028, "step": 24585 }, { "epoch": 7.56, "learning_rate": 2.959828181586538e-06, "loss": 0.0967, "step": 24586 }, { "epoch": 7.56, "learning_rate": 2.9591204907704084e-06, "loss": 0.0429, "step": 24587 }, { "epoch": 7.57, "learning_rate": 2.9584128698766955e-06, "loss": 0.0523, "step": 24588 }, { "epoch": 7.57, "learning_rate": 2.957705318912435e-06, "loss": 0.0473, "step": 24589 }, { "epoch": 7.57, "learning_rate": 2.956997837884652e-06, "loss": 0.0515, "step": 24590 }, { "epoch": 7.57, "learning_rate": 2.95629042680037e-06, "loss": 0.0491, "step": 24591 }, { "epoch": 7.57, "learning_rate": 2.9555830856666167e-06, "loss": 0.0564, "step": 24592 }, { "epoch": 7.57, "learning_rate": 2.954875814490412e-06, "loss": 0.0564, "step": 24593 }, { "epoch": 7.57, "learning_rate": 2.9541686132787907e-06, "loss": 0.0553, "step": 24594 }, { "epoch": 7.57, "learning_rate": 2.953461482038762e-06, "loss": 0.0359, "step": 24595 }, { "epoch": 7.57, "learning_rate": 2.9527544207773583e-06, "loss": 0.0374, "step": 24596 }, { "epoch": 7.57, "learning_rate": 2.9520474295015975e-06, "loss": 0.0448, "step": 24597 }, { "epoch": 7.57, "learning_rate": 2.951340508218502e-06, "loss": 0.0597, "step": 24598 }, { "epoch": 7.57, "learning_rate": 2.950633656935091e-06, "loss": 0.0617, "step": 24599 }, { "epoch": 7.57, "learning_rate": 2.9499268756583823e-06, "loss": 0.0415, "step": 24600 }, { "epoch": 7.57, "learning_rate": 2.949220164395402e-06, "loss": 0.0367, "step": 24601 }, { "epoch": 7.57, "learning_rate": 2.9485135231531616e-06, "loss": 0.0222, "step": 24602 }, { "epoch": 7.57, "learning_rate": 2.9478069519386776e-06, "loss": 0.0497, "step": 24603 }, { "epoch": 7.57, "learning_rate": 2.9471004507589727e-06, "loss": 0.0473, "step": 24604 }, { "epoch": 7.57, "learning_rate": 2.946394019621062e-06, "loss": 0.038, "step": 24605 }, { "epoch": 7.57, "learning_rate": 2.945687658531957e-06, "loss": 0.1041, "step": 24606 }, { "epoch": 7.57, "learning_rate": 2.944981367498677e-06, "loss": 0.0305, "step": 24607 }, { "epoch": 7.57, "learning_rate": 2.9442751465282336e-06, "loss": 0.0564, "step": 24608 }, { "epoch": 7.57, "learning_rate": 2.9435689956276403e-06, "loss": 0.0338, "step": 24609 }, { "epoch": 7.57, "learning_rate": 2.9428629148039077e-06, "loss": 0.0485, "step": 24610 }, { "epoch": 7.57, "learning_rate": 2.9421569040640564e-06, "loss": 0.0531, "step": 24611 }, { "epoch": 7.57, "learning_rate": 2.941450963415089e-06, "loss": 0.0601, "step": 24612 }, { "epoch": 7.57, "learning_rate": 2.940745092864015e-06, "loss": 0.0713, "step": 24613 }, { "epoch": 7.57, "learning_rate": 2.940039292417851e-06, "loss": 0.0605, "step": 24614 }, { "epoch": 7.57, "learning_rate": 2.939333562083605e-06, "loss": 0.0597, "step": 24615 }, { "epoch": 7.57, "learning_rate": 2.938627901868283e-06, "loss": 0.075, "step": 24616 }, { "epoch": 7.57, "learning_rate": 2.9379223117788947e-06, "loss": 0.0371, "step": 24617 }, { "epoch": 7.57, "learning_rate": 2.937216791822445e-06, "loss": 0.044, "step": 24618 }, { "epoch": 7.57, "learning_rate": 2.9365113420059432e-06, "loss": 0.0457, "step": 24619 }, { "epoch": 7.58, "learning_rate": 2.93580596233639e-06, "loss": 0.0373, "step": 24620 }, { "epoch": 7.58, "learning_rate": 2.935100652820797e-06, "loss": 0.0368, "step": 24621 }, { "epoch": 7.58, "learning_rate": 2.934395413466169e-06, "loss": 0.0692, "step": 24622 }, { "epoch": 7.58, "learning_rate": 2.9336902442795003e-06, "loss": 0.0522, "step": 24623 }, { "epoch": 7.58, "learning_rate": 2.9329851452678038e-06, "loss": 0.0335, "step": 24624 }, { "epoch": 7.58, "learning_rate": 2.9322801164380767e-06, "loss": 0.0432, "step": 24625 }, { "epoch": 7.58, "learning_rate": 2.9315751577973227e-06, "loss": 0.0547, "step": 24626 }, { "epoch": 7.58, "learning_rate": 2.9308702693525417e-06, "loss": 0.0624, "step": 24627 }, { "epoch": 7.58, "learning_rate": 2.9301654511107335e-06, "loss": 0.06, "step": 24628 }, { "epoch": 7.58, "learning_rate": 2.929460703078899e-06, "loss": 0.0469, "step": 24629 }, { "epoch": 7.58, "learning_rate": 2.9287560252640323e-06, "loss": 0.0302, "step": 24630 }, { "epoch": 7.58, "learning_rate": 2.9280514176731377e-06, "loss": 0.0311, "step": 24631 }, { "epoch": 7.58, "learning_rate": 2.9273468803132156e-06, "loss": 0.0139, "step": 24632 }, { "epoch": 7.58, "learning_rate": 2.9266424131912495e-06, "loss": 0.0559, "step": 24633 }, { "epoch": 7.58, "learning_rate": 2.925938016314246e-06, "loss": 0.0962, "step": 24634 }, { "epoch": 7.58, "learning_rate": 2.9252336896891986e-06, "loss": 0.0563, "step": 24635 }, { "epoch": 7.58, "learning_rate": 2.9245294333231e-06, "loss": 0.0454, "step": 24636 }, { "epoch": 7.58, "learning_rate": 2.9238252472229468e-06, "loss": 0.0599, "step": 24637 }, { "epoch": 7.58, "learning_rate": 2.9231211313957254e-06, "loss": 0.075, "step": 24638 }, { "epoch": 7.58, "learning_rate": 2.9224170858484414e-06, "loss": 0.0321, "step": 24639 }, { "epoch": 7.58, "learning_rate": 2.9217131105880715e-06, "loss": 0.0507, "step": 24640 }, { "epoch": 7.58, "learning_rate": 2.9210092056216166e-06, "loss": 0.0618, "step": 24641 }, { "epoch": 7.58, "learning_rate": 2.9203053709560647e-06, "loss": 0.0413, "step": 24642 }, { "epoch": 7.58, "learning_rate": 2.9196016065984056e-06, "loss": 0.0495, "step": 24643 }, { "epoch": 7.58, "learning_rate": 2.9188979125556284e-06, "loss": 0.0208, "step": 24644 }, { "epoch": 7.58, "learning_rate": 2.9181942888347204e-06, "loss": 0.0616, "step": 24645 }, { "epoch": 7.58, "learning_rate": 2.9174907354426696e-06, "loss": 0.0408, "step": 24646 }, { "epoch": 7.58, "learning_rate": 2.916787252386464e-06, "loss": 0.061, "step": 24647 }, { "epoch": 7.58, "learning_rate": 2.916083839673086e-06, "loss": 0.0523, "step": 24648 }, { "epoch": 7.58, "learning_rate": 2.91538049730953e-06, "loss": 0.072, "step": 24649 }, { "epoch": 7.58, "learning_rate": 2.914677225302769e-06, "loss": 0.0268, "step": 24650 }, { "epoch": 7.58, "learning_rate": 2.9139740236597958e-06, "loss": 0.0853, "step": 24651 }, { "epoch": 7.58, "learning_rate": 2.9132708923875917e-06, "loss": 0.0634, "step": 24652 }, { "epoch": 7.59, "learning_rate": 2.9125678314931394e-06, "loss": 0.043, "step": 24653 }, { "epoch": 7.59, "learning_rate": 2.9118648409834205e-06, "loss": 0.0476, "step": 24654 }, { "epoch": 7.59, "learning_rate": 2.911161920865412e-06, "loss": 0.0415, "step": 24655 }, { "epoch": 7.59, "learning_rate": 2.9104590711461066e-06, "loss": 0.0519, "step": 24656 }, { "epoch": 7.59, "learning_rate": 2.909756291832473e-06, "loss": 0.0554, "step": 24657 }, { "epoch": 7.59, "learning_rate": 2.9090535829314904e-06, "loss": 0.0487, "step": 24658 }, { "epoch": 7.59, "learning_rate": 2.9083509444501433e-06, "loss": 0.0248, "step": 24659 }, { "epoch": 7.59, "learning_rate": 2.907648376395409e-06, "loss": 0.0425, "step": 24660 }, { "epoch": 7.59, "learning_rate": 2.9069458787742612e-06, "loss": 0.0344, "step": 24661 }, { "epoch": 7.59, "learning_rate": 2.906243451593679e-06, "loss": 0.0574, "step": 24662 }, { "epoch": 7.59, "learning_rate": 2.9055410948606365e-06, "loss": 0.0481, "step": 24663 }, { "epoch": 7.59, "learning_rate": 2.90483880858211e-06, "loss": 0.0475, "step": 24664 }, { "epoch": 7.59, "learning_rate": 2.90413659276507e-06, "loss": 0.0471, "step": 24665 }, { "epoch": 7.59, "learning_rate": 2.903434447416499e-06, "loss": 0.0479, "step": 24666 }, { "epoch": 7.59, "learning_rate": 2.902732372543362e-06, "loss": 0.0356, "step": 24667 }, { "epoch": 7.59, "learning_rate": 2.90203036815263e-06, "loss": 0.0594, "step": 24668 }, { "epoch": 7.59, "learning_rate": 2.9013284342512817e-06, "loss": 0.0316, "step": 24669 }, { "epoch": 7.59, "learning_rate": 2.900626570846283e-06, "loss": 0.0586, "step": 24670 }, { "epoch": 7.59, "learning_rate": 2.8999247779446073e-06, "loss": 0.0439, "step": 24671 }, { "epoch": 7.59, "learning_rate": 2.899223055553221e-06, "loss": 0.0491, "step": 24672 }, { "epoch": 7.59, "learning_rate": 2.8985214036790944e-06, "loss": 0.0588, "step": 24673 }, { "epoch": 7.59, "learning_rate": 2.8978198223291953e-06, "loss": 0.043, "step": 24674 }, { "epoch": 7.59, "learning_rate": 2.897118311510487e-06, "loss": 0.0487, "step": 24675 }, { "epoch": 7.59, "learning_rate": 2.8964168712299447e-06, "loss": 0.1187, "step": 24676 }, { "epoch": 7.59, "learning_rate": 2.8957155014945316e-06, "loss": 0.0506, "step": 24677 }, { "epoch": 7.59, "learning_rate": 2.895014202311205e-06, "loss": 0.0412, "step": 24678 }, { "epoch": 7.59, "learning_rate": 2.894312973686939e-06, "loss": 0.03, "step": 24679 }, { "epoch": 7.59, "learning_rate": 2.893611815628694e-06, "loss": 0.0697, "step": 24680 }, { "epoch": 7.59, "learning_rate": 2.8929107281434334e-06, "loss": 0.027, "step": 24681 }, { "epoch": 7.59, "learning_rate": 2.8922097112381185e-06, "loss": 0.052, "step": 24682 }, { "epoch": 7.59, "learning_rate": 2.8915087649197094e-06, "loss": 0.0469, "step": 24683 }, { "epoch": 7.59, "learning_rate": 2.890807889195175e-06, "loss": 0.0361, "step": 24684 }, { "epoch": 7.6, "learning_rate": 2.890107084071465e-06, "loss": 0.0391, "step": 24685 }, { "epoch": 7.6, "learning_rate": 2.8894063495555468e-06, "loss": 0.0535, "step": 24686 }, { "epoch": 7.6, "learning_rate": 2.8887056856543805e-06, "loss": 0.0694, "step": 24687 }, { "epoch": 7.6, "learning_rate": 2.8880050923749137e-06, "loss": 0.045, "step": 24688 }, { "epoch": 7.6, "learning_rate": 2.887304569724115e-06, "loss": 0.0409, "step": 24689 }, { "epoch": 7.6, "learning_rate": 2.8866041177089365e-06, "loss": 0.0336, "step": 24690 }, { "epoch": 7.6, "learning_rate": 2.8859037363363362e-06, "loss": 0.0439, "step": 24691 }, { "epoch": 7.6, "learning_rate": 2.8852034256132666e-06, "loss": 0.0398, "step": 24692 }, { "epoch": 7.6, "learning_rate": 2.884503185546682e-06, "loss": 0.0529, "step": 24693 }, { "epoch": 7.6, "learning_rate": 2.8838030161435447e-06, "loss": 0.0549, "step": 24694 }, { "epoch": 7.6, "learning_rate": 2.883102917410796e-06, "loss": 0.0639, "step": 24695 }, { "epoch": 7.6, "learning_rate": 2.882402889355398e-06, "loss": 0.0986, "step": 24696 }, { "epoch": 7.6, "learning_rate": 2.881702931984298e-06, "loss": 0.0714, "step": 24697 }, { "epoch": 7.6, "learning_rate": 2.881003045304448e-06, "loss": 0.0629, "step": 24698 }, { "epoch": 7.6, "learning_rate": 2.8803032293227995e-06, "loss": 0.042, "step": 24699 }, { "epoch": 7.6, "learning_rate": 2.8796034840462995e-06, "loss": 0.0497, "step": 24700 }, { "epoch": 7.6, "learning_rate": 2.8789038094819044e-06, "loss": 0.0494, "step": 24701 }, { "epoch": 7.6, "learning_rate": 2.8782042056365543e-06, "loss": 0.0775, "step": 24702 }, { "epoch": 7.6, "learning_rate": 2.8775046725171975e-06, "loss": 0.0312, "step": 24703 }, { "epoch": 7.6, "learning_rate": 2.8768052101307863e-06, "loss": 0.0492, "step": 24704 }, { "epoch": 7.6, "learning_rate": 2.876105818484264e-06, "loss": 0.0703, "step": 24705 }, { "epoch": 7.6, "learning_rate": 2.875406497584576e-06, "loss": 0.0553, "step": 24706 }, { "epoch": 7.6, "learning_rate": 2.8747072474386694e-06, "loss": 0.0225, "step": 24707 }, { "epoch": 7.6, "learning_rate": 2.8740080680534854e-06, "loss": 0.0196, "step": 24708 }, { "epoch": 7.6, "learning_rate": 2.8733089594359696e-06, "loss": 0.045, "step": 24709 }, { "epoch": 7.6, "learning_rate": 2.8726099215930604e-06, "loss": 0.0553, "step": 24710 }, { "epoch": 7.6, "learning_rate": 2.8719109545317102e-06, "loss": 0.0372, "step": 24711 }, { "epoch": 7.6, "learning_rate": 2.87121205825885e-06, "loss": 0.0391, "step": 24712 }, { "epoch": 7.6, "learning_rate": 2.870513232781421e-06, "loss": 0.0591, "step": 24713 }, { "epoch": 7.6, "learning_rate": 2.86981447810637e-06, "loss": 0.0499, "step": 24714 }, { "epoch": 7.6, "learning_rate": 2.8691157942406324e-06, "loss": 0.0409, "step": 24715 }, { "epoch": 7.6, "learning_rate": 2.8684171811911464e-06, "loss": 0.0399, "step": 24716 }, { "epoch": 7.6, "learning_rate": 2.8677186389648503e-06, "loss": 0.0776, "step": 24717 }, { "epoch": 7.61, "learning_rate": 2.8670201675686825e-06, "loss": 0.0611, "step": 24718 }, { "epoch": 7.61, "learning_rate": 2.8663217670095777e-06, "loss": 0.0585, "step": 24719 }, { "epoch": 7.61, "learning_rate": 2.8656234372944692e-06, "loss": 0.0626, "step": 24720 }, { "epoch": 7.61, "learning_rate": 2.864925178430299e-06, "loss": 0.0645, "step": 24721 }, { "epoch": 7.61, "learning_rate": 2.8642269904239996e-06, "loss": 0.0251, "step": 24722 }, { "epoch": 7.61, "learning_rate": 2.863528873282497e-06, "loss": 0.0554, "step": 24723 }, { "epoch": 7.61, "learning_rate": 2.8628308270127335e-06, "loss": 0.0422, "step": 24724 }, { "epoch": 7.61, "learning_rate": 2.8621328516216374e-06, "loss": 0.0363, "step": 24725 }, { "epoch": 7.61, "learning_rate": 2.8614349471161406e-06, "loss": 0.0508, "step": 24726 }, { "epoch": 7.61, "learning_rate": 2.860737113503174e-06, "loss": 0.0653, "step": 24727 }, { "epoch": 7.61, "learning_rate": 2.860039350789667e-06, "loss": 0.0441, "step": 24728 }, { "epoch": 7.61, "learning_rate": 2.8593416589825498e-06, "loss": 0.0779, "step": 24729 }, { "epoch": 7.61, "learning_rate": 2.858644038088748e-06, "loss": 0.0246, "step": 24730 }, { "epoch": 7.61, "learning_rate": 2.857946488115195e-06, "loss": 0.0192, "step": 24731 }, { "epoch": 7.61, "learning_rate": 2.8572490090688165e-06, "loss": 0.0657, "step": 24732 }, { "epoch": 7.61, "learning_rate": 2.8565516009565387e-06, "loss": 0.0229, "step": 24733 }, { "epoch": 7.61, "learning_rate": 2.8558542637852858e-06, "loss": 0.0603, "step": 24734 }, { "epoch": 7.61, "learning_rate": 2.855156997561985e-06, "loss": 0.0434, "step": 24735 }, { "epoch": 7.61, "learning_rate": 2.854459802293561e-06, "loss": 0.0589, "step": 24736 }, { "epoch": 7.61, "learning_rate": 2.853762677986932e-06, "loss": 0.0413, "step": 24737 }, { "epoch": 7.61, "learning_rate": 2.8530656246490306e-06, "loss": 0.0631, "step": 24738 }, { "epoch": 7.61, "learning_rate": 2.8523686422867757e-06, "loss": 0.0331, "step": 24739 }, { "epoch": 7.61, "learning_rate": 2.851671730907083e-06, "loss": 0.0545, "step": 24740 }, { "epoch": 7.61, "learning_rate": 2.85097489051688e-06, "loss": 0.0561, "step": 24741 }, { "epoch": 7.61, "learning_rate": 2.8502781211230855e-06, "loss": 0.0637, "step": 24742 }, { "epoch": 7.61, "learning_rate": 2.849581422732619e-06, "loss": 0.0635, "step": 24743 }, { "epoch": 7.61, "learning_rate": 2.848884795352398e-06, "loss": 0.0655, "step": 24744 }, { "epoch": 7.61, "learning_rate": 2.848188238989342e-06, "loss": 0.0443, "step": 24745 }, { "epoch": 7.61, "learning_rate": 2.8474917536503687e-06, "loss": 0.0515, "step": 24746 }, { "epoch": 7.61, "learning_rate": 2.8467953393423908e-06, "loss": 0.0521, "step": 24747 }, { "epoch": 7.61, "learning_rate": 2.8460989960723306e-06, "loss": 0.0262, "step": 24748 }, { "epoch": 7.61, "learning_rate": 2.845402723847104e-06, "loss": 0.0336, "step": 24749 }, { "epoch": 7.62, "learning_rate": 2.844706522673616e-06, "loss": 0.0586, "step": 24750 }, { "epoch": 7.62, "learning_rate": 2.8440103925587904e-06, "loss": 0.033, "step": 24751 }, { "epoch": 7.62, "learning_rate": 2.843314333509536e-06, "loss": 0.0769, "step": 24752 }, { "epoch": 7.62, "learning_rate": 2.842618345532766e-06, "loss": 0.0506, "step": 24753 }, { "epoch": 7.62, "learning_rate": 2.841922428635392e-06, "loss": 0.0609, "step": 24754 }, { "epoch": 7.62, "learning_rate": 2.8412265828243237e-06, "loss": 0.0656, "step": 24755 }, { "epoch": 7.62, "learning_rate": 2.8405308081064777e-06, "loss": 0.0621, "step": 24756 }, { "epoch": 7.62, "learning_rate": 2.839835104488754e-06, "loss": 0.059, "step": 24757 }, { "epoch": 7.62, "learning_rate": 2.8391394719780696e-06, "loss": 0.071, "step": 24758 }, { "epoch": 7.62, "learning_rate": 2.838443910581329e-06, "loss": 0.0636, "step": 24759 }, { "epoch": 7.62, "learning_rate": 2.8377484203054416e-06, "loss": 0.0561, "step": 24760 }, { "epoch": 7.62, "learning_rate": 2.8370530011573126e-06, "loss": 0.0662, "step": 24761 }, { "epoch": 7.62, "learning_rate": 2.836357653143846e-06, "loss": 0.0414, "step": 24762 }, { "epoch": 7.62, "learning_rate": 2.835662376271957e-06, "loss": 0.0684, "step": 24763 }, { "epoch": 7.62, "learning_rate": 2.8349671705485403e-06, "loss": 0.0405, "step": 24764 }, { "epoch": 7.62, "learning_rate": 2.8342720359805e-06, "loss": 0.0687, "step": 24765 }, { "epoch": 7.62, "learning_rate": 2.8335769725747476e-06, "loss": 0.0412, "step": 24766 }, { "epoch": 7.62, "learning_rate": 2.8328819803381757e-06, "loss": 0.0471, "step": 24767 }, { "epoch": 7.62, "learning_rate": 2.8321870592776933e-06, "loss": 0.042, "step": 24768 }, { "epoch": 7.62, "learning_rate": 2.8314922094001997e-06, "loss": 0.0275, "step": 24769 }, { "epoch": 7.62, "learning_rate": 2.830797430712595e-06, "loss": 0.0403, "step": 24770 }, { "epoch": 7.62, "learning_rate": 2.8301027232217804e-06, "loss": 0.0378, "step": 24771 }, { "epoch": 7.62, "learning_rate": 2.8294080869346495e-06, "loss": 0.0354, "step": 24772 }, { "epoch": 7.62, "learning_rate": 2.828713521858111e-06, "loss": 0.0716, "step": 24773 }, { "epoch": 7.62, "learning_rate": 2.8280190279990537e-06, "loss": 0.0513, "step": 24774 }, { "epoch": 7.62, "learning_rate": 2.8273246053643743e-06, "loss": 0.0453, "step": 24775 }, { "epoch": 7.62, "learning_rate": 2.8266302539609747e-06, "loss": 0.0498, "step": 24776 }, { "epoch": 7.62, "learning_rate": 2.825935973795748e-06, "loss": 0.0466, "step": 24777 }, { "epoch": 7.62, "learning_rate": 2.82524176487559e-06, "loss": 0.0395, "step": 24778 }, { "epoch": 7.62, "learning_rate": 2.8245476272073925e-06, "loss": 0.0253, "step": 24779 }, { "epoch": 7.62, "learning_rate": 2.823853560798051e-06, "loss": 0.0414, "step": 24780 }, { "epoch": 7.62, "learning_rate": 2.8231595656544563e-06, "loss": 0.0367, "step": 24781 }, { "epoch": 7.62, "learning_rate": 2.8224656417834994e-06, "loss": 0.0656, "step": 24782 }, { "epoch": 7.63, "learning_rate": 2.821771789192076e-06, "loss": 0.035, "step": 24783 }, { "epoch": 7.63, "learning_rate": 2.821078007887078e-06, "loss": 0.0503, "step": 24784 }, { "epoch": 7.63, "learning_rate": 2.820384297875386e-06, "loss": 0.0661, "step": 24785 }, { "epoch": 7.63, "learning_rate": 2.8196906591638975e-06, "loss": 0.0868, "step": 24786 }, { "epoch": 7.63, "learning_rate": 2.8189970917594978e-06, "loss": 0.1445, "step": 24787 }, { "epoch": 7.63, "learning_rate": 2.8183035956690753e-06, "loss": 0.0293, "step": 24788 }, { "epoch": 7.63, "learning_rate": 2.8176101708995174e-06, "loss": 0.037, "step": 24789 }, { "epoch": 7.63, "learning_rate": 2.8169168174577086e-06, "loss": 0.0336, "step": 24790 }, { "epoch": 7.63, "learning_rate": 2.8162235353505376e-06, "loss": 0.0302, "step": 24791 }, { "epoch": 7.63, "learning_rate": 2.815530324584884e-06, "loss": 0.0259, "step": 24792 }, { "epoch": 7.63, "learning_rate": 2.814837185167639e-06, "loss": 0.039, "step": 24793 }, { "epoch": 7.63, "learning_rate": 2.814144117105686e-06, "loss": 0.0365, "step": 24794 }, { "epoch": 7.63, "learning_rate": 2.8134511204058977e-06, "loss": 0.05, "step": 24795 }, { "epoch": 7.63, "learning_rate": 2.812758195075167e-06, "loss": 0.0363, "step": 24796 }, { "epoch": 7.63, "learning_rate": 2.8120653411203703e-06, "loss": 0.0513, "step": 24797 }, { "epoch": 7.63, "learning_rate": 2.81137255854839e-06, "loss": 0.0831, "step": 24798 }, { "epoch": 7.63, "learning_rate": 2.8106798473661057e-06, "loss": 0.0126, "step": 24799 }, { "epoch": 7.63, "learning_rate": 2.8099872075803925e-06, "loss": 0.0725, "step": 24800 }, { "epoch": 7.63, "learning_rate": 2.80929463919814e-06, "loss": 0.0512, "step": 24801 }, { "epoch": 7.63, "learning_rate": 2.808602142226212e-06, "loss": 0.0227, "step": 24802 }, { "epoch": 7.63, "learning_rate": 2.8079097166714963e-06, "loss": 0.0458, "step": 24803 }, { "epoch": 7.63, "learning_rate": 2.8072173625408683e-06, "loss": 0.039, "step": 24804 }, { "epoch": 7.63, "learning_rate": 2.8065250798411945e-06, "loss": 0.0263, "step": 24805 }, { "epoch": 7.63, "learning_rate": 2.8058328685793588e-06, "loss": 0.0348, "step": 24806 }, { "epoch": 7.63, "learning_rate": 2.8051407287622343e-06, "loss": 0.0266, "step": 24807 }, { "epoch": 7.63, "learning_rate": 2.8044486603966936e-06, "loss": 0.0382, "step": 24808 }, { "epoch": 7.63, "learning_rate": 2.8037566634896085e-06, "loss": 0.0669, "step": 24809 }, { "epoch": 7.63, "learning_rate": 2.803064738047848e-06, "loss": 0.0591, "step": 24810 }, { "epoch": 7.63, "learning_rate": 2.802372884078295e-06, "loss": 0.034, "step": 24811 }, { "epoch": 7.63, "learning_rate": 2.8016811015878063e-06, "loss": 0.0668, "step": 24812 }, { "epoch": 7.63, "learning_rate": 2.800989390583262e-06, "loss": 0.0696, "step": 24813 }, { "epoch": 7.63, "learning_rate": 2.800297751071528e-06, "loss": 0.0291, "step": 24814 }, { "epoch": 7.64, "learning_rate": 2.7996061830594714e-06, "loss": 0.0812, "step": 24815 }, { "epoch": 7.64, "learning_rate": 2.798914686553963e-06, "loss": 0.0184, "step": 24816 }, { "epoch": 7.64, "learning_rate": 2.7982232615618642e-06, "loss": 0.0598, "step": 24817 }, { "epoch": 7.64, "learning_rate": 2.797531908090052e-06, "loss": 0.0815, "step": 24818 }, { "epoch": 7.64, "learning_rate": 2.796840626145384e-06, "loss": 0.0327, "step": 24819 }, { "epoch": 7.64, "learning_rate": 2.7961494157347236e-06, "loss": 0.0576, "step": 24820 }, { "epoch": 7.64, "learning_rate": 2.795458276864941e-06, "loss": 0.0608, "step": 24821 }, { "epoch": 7.64, "learning_rate": 2.794767209542898e-06, "loss": 0.0519, "step": 24822 }, { "epoch": 7.64, "learning_rate": 2.7940762137754575e-06, "loss": 0.0419, "step": 24823 }, { "epoch": 7.64, "learning_rate": 2.793385289569481e-06, "loss": 0.0442, "step": 24824 }, { "epoch": 7.64, "learning_rate": 2.792694436931831e-06, "loss": 0.0162, "step": 24825 }, { "epoch": 7.64, "learning_rate": 2.7920036558693674e-06, "loss": 0.0498, "step": 24826 }, { "epoch": 7.64, "learning_rate": 2.791312946388948e-06, "loss": 0.0393, "step": 24827 }, { "epoch": 7.64, "learning_rate": 2.7906223084974405e-06, "loss": 0.0494, "step": 24828 }, { "epoch": 7.64, "learning_rate": 2.7899317422016937e-06, "loss": 0.0357, "step": 24829 }, { "epoch": 7.64, "learning_rate": 2.7892412475085685e-06, "loss": 0.051, "step": 24830 }, { "epoch": 7.64, "learning_rate": 2.788550824424925e-06, "loss": 0.0281, "step": 24831 }, { "epoch": 7.64, "learning_rate": 2.787860472957619e-06, "loss": 0.0474, "step": 24832 }, { "epoch": 7.64, "learning_rate": 2.787170193113504e-06, "loss": 0.069, "step": 24833 }, { "epoch": 7.64, "learning_rate": 2.786479984899438e-06, "loss": 0.0999, "step": 24834 }, { "epoch": 7.64, "learning_rate": 2.7857898483222733e-06, "loss": 0.057, "step": 24835 }, { "epoch": 7.64, "learning_rate": 2.785099783388864e-06, "loss": 0.0289, "step": 24836 }, { "epoch": 7.64, "learning_rate": 2.784409790106061e-06, "loss": 0.0711, "step": 24837 }, { "epoch": 7.64, "learning_rate": 2.783719868480721e-06, "loss": 0.0468, "step": 24838 }, { "epoch": 7.64, "learning_rate": 2.7830300185196968e-06, "loss": 0.0292, "step": 24839 }, { "epoch": 7.64, "learning_rate": 2.7823402402298304e-06, "loss": 0.1046, "step": 24840 }, { "epoch": 7.64, "learning_rate": 2.78165053361798e-06, "loss": 0.0585, "step": 24841 }, { "epoch": 7.64, "learning_rate": 2.780960898690992e-06, "loss": 0.0723, "step": 24842 }, { "epoch": 7.64, "learning_rate": 2.7802713354557166e-06, "loss": 0.0514, "step": 24843 }, { "epoch": 7.64, "learning_rate": 2.779581843919e-06, "loss": 0.0468, "step": 24844 }, { "epoch": 7.64, "learning_rate": 2.77889242408769e-06, "loss": 0.0537, "step": 24845 }, { "epoch": 7.64, "learning_rate": 2.778203075968634e-06, "loss": 0.0424, "step": 24846 }, { "epoch": 7.64, "learning_rate": 2.777513799568674e-06, "loss": 0.05, "step": 24847 }, { "epoch": 7.65, "learning_rate": 2.7768245948946615e-06, "loss": 0.0701, "step": 24848 }, { "epoch": 7.65, "learning_rate": 2.776135461953441e-06, "loss": 0.0336, "step": 24849 }, { "epoch": 7.65, "learning_rate": 2.7754464007518467e-06, "loss": 0.0203, "step": 24850 }, { "epoch": 7.65, "learning_rate": 2.7747574112967303e-06, "loss": 0.0238, "step": 24851 }, { "epoch": 7.65, "learning_rate": 2.7740684935949324e-06, "loss": 0.0161, "step": 24852 }, { "epoch": 7.65, "learning_rate": 2.773379647653295e-06, "loss": 0.0299, "step": 24853 }, { "epoch": 7.65, "learning_rate": 2.772690873478656e-06, "loss": 0.0477, "step": 24854 }, { "epoch": 7.65, "learning_rate": 2.7720021710778565e-06, "loss": 0.0567, "step": 24855 }, { "epoch": 7.65, "learning_rate": 2.7713135404577417e-06, "loss": 0.0676, "step": 24856 }, { "epoch": 7.65, "learning_rate": 2.7706249816251396e-06, "loss": 0.0427, "step": 24857 }, { "epoch": 7.65, "learning_rate": 2.7699364945868977e-06, "loss": 0.0262, "step": 24858 }, { "epoch": 7.65, "learning_rate": 2.76924807934985e-06, "loss": 0.0637, "step": 24859 }, { "epoch": 7.65, "learning_rate": 2.7685597359208326e-06, "loss": 0.0462, "step": 24860 }, { "epoch": 7.65, "learning_rate": 2.767871464306682e-06, "loss": 0.0404, "step": 24861 }, { "epoch": 7.65, "learning_rate": 2.7671832645142327e-06, "loss": 0.0835, "step": 24862 }, { "epoch": 7.65, "learning_rate": 2.76649513655032e-06, "loss": 0.0637, "step": 24863 }, { "epoch": 7.65, "learning_rate": 2.7658070804217764e-06, "loss": 0.0316, "step": 24864 }, { "epoch": 7.65, "learning_rate": 2.7651190961354336e-06, "loss": 0.0558, "step": 24865 }, { "epoch": 7.65, "learning_rate": 2.764431183698131e-06, "loss": 0.0619, "step": 24866 }, { "epoch": 7.65, "learning_rate": 2.7637433431166903e-06, "loss": 0.0772, "step": 24867 }, { "epoch": 7.65, "learning_rate": 2.763055574397949e-06, "loss": 0.0225, "step": 24868 }, { "epoch": 7.65, "learning_rate": 2.7623678775487374e-06, "loss": 0.0585, "step": 24869 }, { "epoch": 7.65, "learning_rate": 2.761680252575882e-06, "loss": 0.0396, "step": 24870 }, { "epoch": 7.65, "learning_rate": 2.7609926994862136e-06, "loss": 0.0649, "step": 24871 }, { "epoch": 7.65, "learning_rate": 2.760305218286555e-06, "loss": 0.0767, "step": 24872 }, { "epoch": 7.65, "learning_rate": 2.7596178089837445e-06, "loss": 0.0393, "step": 24873 }, { "epoch": 7.65, "learning_rate": 2.7589304715845998e-06, "loss": 0.0443, "step": 24874 }, { "epoch": 7.65, "learning_rate": 2.758243206095945e-06, "loss": 0.0317, "step": 24875 }, { "epoch": 7.65, "learning_rate": 2.757556012524614e-06, "loss": 0.0421, "step": 24876 }, { "epoch": 7.65, "learning_rate": 2.7568688908774243e-06, "loss": 0.0348, "step": 24877 }, { "epoch": 7.65, "learning_rate": 2.756181841161204e-06, "loss": 0.0434, "step": 24878 }, { "epoch": 7.65, "learning_rate": 2.7554948633827727e-06, "loss": 0.087, "step": 24879 }, { "epoch": 7.66, "learning_rate": 2.754807957548955e-06, "loss": 0.0463, "step": 24880 }, { "epoch": 7.66, "learning_rate": 2.754121123666571e-06, "loss": 0.0307, "step": 24881 }, { "epoch": 7.66, "learning_rate": 2.75343436174244e-06, "loss": 0.0514, "step": 24882 }, { "epoch": 7.66, "learning_rate": 2.75274767178339e-06, "loss": 0.0527, "step": 24883 }, { "epoch": 7.66, "learning_rate": 2.752061053796232e-06, "loss": 0.0302, "step": 24884 }, { "epoch": 7.66, "learning_rate": 2.751374507787785e-06, "loss": 0.0672, "step": 24885 }, { "epoch": 7.66, "learning_rate": 2.7506880337648723e-06, "loss": 0.026, "step": 24886 }, { "epoch": 7.66, "learning_rate": 2.750001631734308e-06, "loss": 0.0418, "step": 24887 }, { "epoch": 7.66, "learning_rate": 2.7493153017029094e-06, "loss": 0.0603, "step": 24888 }, { "epoch": 7.66, "learning_rate": 2.748629043677493e-06, "loss": 0.0297, "step": 24889 }, { "epoch": 7.66, "learning_rate": 2.7479428576648738e-06, "loss": 0.0782, "step": 24890 }, { "epoch": 7.66, "learning_rate": 2.7472567436718642e-06, "loss": 0.0483, "step": 24891 }, { "epoch": 7.66, "learning_rate": 2.746570701705277e-06, "loss": 0.0361, "step": 24892 }, { "epoch": 7.66, "learning_rate": 2.745884731771931e-06, "loss": 0.1772, "step": 24893 }, { "epoch": 7.66, "learning_rate": 2.7451988338786386e-06, "loss": 0.046, "step": 24894 }, { "epoch": 7.66, "learning_rate": 2.744513008032201e-06, "loss": 0.0319, "step": 24895 }, { "epoch": 7.66, "learning_rate": 2.7438272542394396e-06, "loss": 0.0591, "step": 24896 }, { "epoch": 7.66, "learning_rate": 2.743141572507161e-06, "loss": 0.0443, "step": 24897 }, { "epoch": 7.66, "learning_rate": 2.7424559628421745e-06, "loss": 0.0526, "step": 24898 }, { "epoch": 7.66, "learning_rate": 2.7417704252512888e-06, "loss": 0.0431, "step": 24899 }, { "epoch": 7.66, "learning_rate": 2.74108495974131e-06, "loss": 0.0251, "step": 24900 }, { "epoch": 7.66, "learning_rate": 2.740399566319053e-06, "loss": 0.0477, "step": 24901 }, { "epoch": 7.66, "learning_rate": 2.739714244991313e-06, "loss": 0.0355, "step": 24902 }, { "epoch": 7.66, "learning_rate": 2.7390289957649053e-06, "loss": 0.0391, "step": 24903 }, { "epoch": 7.66, "learning_rate": 2.738343818646634e-06, "loss": 0.0272, "step": 24904 }, { "epoch": 7.66, "learning_rate": 2.7376587136432954e-06, "loss": 0.0976, "step": 24905 }, { "epoch": 7.66, "learning_rate": 2.736973680761702e-06, "loss": 0.0396, "step": 24906 }, { "epoch": 7.66, "learning_rate": 2.736288720008653e-06, "loss": 0.0193, "step": 24907 }, { "epoch": 7.66, "learning_rate": 2.735603831390953e-06, "loss": 0.0875, "step": 24908 }, { "epoch": 7.66, "learning_rate": 2.7349190149154004e-06, "loss": 0.0684, "step": 24909 }, { "epoch": 7.66, "learning_rate": 2.7342342705887946e-06, "loss": 0.0285, "step": 24910 }, { "epoch": 7.66, "learning_rate": 2.7335495984179463e-06, "loss": 0.0387, "step": 24911 }, { "epoch": 7.66, "learning_rate": 2.7328649984096412e-06, "loss": 0.0318, "step": 24912 }, { "epoch": 7.67, "learning_rate": 2.7321804705706867e-06, "loss": 0.0394, "step": 24913 }, { "epoch": 7.67, "learning_rate": 2.7314960149078785e-06, "loss": 0.0497, "step": 24914 }, { "epoch": 7.67, "learning_rate": 2.730811631428013e-06, "loss": 0.0462, "step": 24915 }, { "epoch": 7.67, "learning_rate": 2.7301273201378884e-06, "loss": 0.0493, "step": 24916 }, { "epoch": 7.67, "learning_rate": 2.7294430810442964e-06, "loss": 0.027, "step": 24917 }, { "epoch": 7.67, "learning_rate": 2.7287589141540404e-06, "loss": 0.0495, "step": 24918 }, { "epoch": 7.67, "learning_rate": 2.728074819473908e-06, "loss": 0.0668, "step": 24919 }, { "epoch": 7.67, "learning_rate": 2.7273907970106916e-06, "loss": 0.0236, "step": 24920 }, { "epoch": 7.67, "learning_rate": 2.7267068467711898e-06, "loss": 0.0699, "step": 24921 }, { "epoch": 7.67, "learning_rate": 2.726022968762192e-06, "loss": 0.0621, "step": 24922 }, { "epoch": 7.67, "learning_rate": 2.7253391629904915e-06, "loss": 0.0298, "step": 24923 }, { "epoch": 7.67, "learning_rate": 2.724655429462876e-06, "loss": 0.0663, "step": 24924 }, { "epoch": 7.67, "learning_rate": 2.7239717681861378e-06, "loss": 0.0299, "step": 24925 }, { "epoch": 7.67, "learning_rate": 2.7232881791670652e-06, "loss": 0.0404, "step": 24926 }, { "epoch": 7.67, "learning_rate": 2.7226046624124447e-06, "loss": 0.0522, "step": 24927 }, { "epoch": 7.67, "learning_rate": 2.721921217929073e-06, "loss": 0.0362, "step": 24928 }, { "epoch": 7.67, "learning_rate": 2.7212378457237264e-06, "loss": 0.0601, "step": 24929 }, { "epoch": 7.67, "learning_rate": 2.7205545458031977e-06, "loss": 0.0302, "step": 24930 }, { "epoch": 7.67, "learning_rate": 2.719871318174272e-06, "loss": 0.0631, "step": 24931 }, { "epoch": 7.67, "learning_rate": 2.7191881628437335e-06, "loss": 0.0785, "step": 24932 }, { "epoch": 7.67, "learning_rate": 2.7185050798183666e-06, "loss": 0.0497, "step": 24933 }, { "epoch": 7.67, "learning_rate": 2.717822069104953e-06, "loss": 0.0409, "step": 24934 }, { "epoch": 7.67, "learning_rate": 2.7171391307102825e-06, "loss": 0.0863, "step": 24935 }, { "epoch": 7.67, "learning_rate": 2.7164562646411306e-06, "loss": 0.043, "step": 24936 }, { "epoch": 7.67, "learning_rate": 2.715773470904277e-06, "loss": 0.0695, "step": 24937 }, { "epoch": 7.67, "learning_rate": 2.7150907495065095e-06, "loss": 0.0708, "step": 24938 }, { "epoch": 7.67, "learning_rate": 2.7144081004546042e-06, "loss": 0.041, "step": 24939 }, { "epoch": 7.67, "learning_rate": 2.7137255237553415e-06, "loss": 0.0537, "step": 24940 }, { "epoch": 7.67, "learning_rate": 2.7130430194155e-06, "loss": 0.0863, "step": 24941 }, { "epoch": 7.67, "learning_rate": 2.7123605874418568e-06, "loss": 0.0326, "step": 24942 }, { "epoch": 7.67, "learning_rate": 2.711678227841189e-06, "loss": 0.0656, "step": 24943 }, { "epoch": 7.67, "learning_rate": 2.7109959406202713e-06, "loss": 0.0679, "step": 24944 }, { "epoch": 7.68, "learning_rate": 2.7103137257858867e-06, "loss": 0.0519, "step": 24945 }, { "epoch": 7.68, "learning_rate": 2.7096315833448027e-06, "loss": 0.0548, "step": 24946 }, { "epoch": 7.68, "learning_rate": 2.7089495133037934e-06, "loss": 0.0644, "step": 24947 }, { "epoch": 7.68, "learning_rate": 2.708267515669638e-06, "loss": 0.0313, "step": 24948 }, { "epoch": 7.68, "learning_rate": 2.7075855904491055e-06, "loss": 0.042, "step": 24949 }, { "epoch": 7.68, "learning_rate": 2.7069037376489706e-06, "loss": 0.03, "step": 24950 }, { "epoch": 7.68, "learning_rate": 2.706221957276002e-06, "loss": 0.0733, "step": 24951 }, { "epoch": 7.68, "learning_rate": 2.705540249336971e-06, "loss": 0.0728, "step": 24952 }, { "epoch": 7.68, "learning_rate": 2.704858613838649e-06, "loss": 0.037, "step": 24953 }, { "epoch": 7.68, "learning_rate": 2.704177050787802e-06, "loss": 0.0445, "step": 24954 }, { "epoch": 7.68, "learning_rate": 2.703495560191203e-06, "loss": 0.0495, "step": 24955 }, { "epoch": 7.68, "learning_rate": 2.7028141420556196e-06, "loss": 0.0498, "step": 24956 }, { "epoch": 7.68, "learning_rate": 2.702132796387813e-06, "loss": 0.0883, "step": 24957 }, { "epoch": 7.68, "learning_rate": 2.7014515231945557e-06, "loss": 0.054, "step": 24958 }, { "epoch": 7.68, "learning_rate": 2.700770322482612e-06, "loss": 0.018, "step": 24959 }, { "epoch": 7.68, "learning_rate": 2.700089194258745e-06, "loss": 0.0376, "step": 24960 }, { "epoch": 7.68, "learning_rate": 2.699408138529721e-06, "loss": 0.0651, "step": 24961 }, { "epoch": 7.68, "learning_rate": 2.698727155302302e-06, "loss": 0.0561, "step": 24962 }, { "epoch": 7.68, "learning_rate": 2.6980462445832524e-06, "loss": 0.0859, "step": 24963 }, { "epoch": 7.68, "learning_rate": 2.6973654063793296e-06, "loss": 0.064, "step": 24964 }, { "epoch": 7.68, "learning_rate": 2.6966846406973014e-06, "loss": 0.0293, "step": 24965 }, { "epoch": 7.68, "learning_rate": 2.696003947543929e-06, "loss": 0.0526, "step": 24966 }, { "epoch": 7.68, "learning_rate": 2.695323326925963e-06, "loss": 0.0456, "step": 24967 }, { "epoch": 7.68, "learning_rate": 2.6946427788501716e-06, "loss": 0.0627, "step": 24968 }, { "epoch": 7.68, "learning_rate": 2.6939623033233096e-06, "loss": 0.0504, "step": 24969 }, { "epoch": 7.68, "learning_rate": 2.693281900352136e-06, "loss": 0.0459, "step": 24970 }, { "epoch": 7.68, "learning_rate": 2.692601569943407e-06, "loss": 0.0343, "step": 24971 }, { "epoch": 7.68, "learning_rate": 2.6919213121038765e-06, "loss": 0.0591, "step": 24972 }, { "epoch": 7.68, "learning_rate": 2.6912411268403084e-06, "loss": 0.0498, "step": 24973 }, { "epoch": 7.68, "learning_rate": 2.690561014159445e-06, "loss": 0.0611, "step": 24974 }, { "epoch": 7.68, "learning_rate": 2.689880974068052e-06, "loss": 0.0638, "step": 24975 }, { "epoch": 7.68, "learning_rate": 2.689201006572877e-06, "loss": 0.0571, "step": 24976 }, { "epoch": 7.68, "learning_rate": 2.688521111680674e-06, "loss": 0.0426, "step": 24977 }, { "epoch": 7.69, "learning_rate": 2.6878412893981954e-06, "loss": 0.0322, "step": 24978 }, { "epoch": 7.69, "learning_rate": 2.687161539732188e-06, "loss": 0.0497, "step": 24979 }, { "epoch": 7.69, "learning_rate": 2.686481862689413e-06, "loss": 0.0471, "step": 24980 }, { "epoch": 7.69, "learning_rate": 2.6858022582766097e-06, "loss": 0.0486, "step": 24981 }, { "epoch": 7.69, "learning_rate": 2.6851227265005287e-06, "loss": 0.0316, "step": 24982 }, { "epoch": 7.69, "learning_rate": 2.6844432673679253e-06, "loss": 0.0296, "step": 24983 }, { "epoch": 7.69, "learning_rate": 2.683763880885538e-06, "loss": 0.0785, "step": 24984 }, { "epoch": 7.69, "learning_rate": 2.68308456706012e-06, "loss": 0.0385, "step": 24985 }, { "epoch": 7.69, "learning_rate": 2.682405325898416e-06, "loss": 0.0599, "step": 24986 }, { "epoch": 7.69, "learning_rate": 2.6817261574071717e-06, "loss": 0.0417, "step": 24987 }, { "epoch": 7.69, "learning_rate": 2.6810470615931316e-06, "loss": 0.0592, "step": 24988 }, { "epoch": 7.69, "learning_rate": 2.6803680384630357e-06, "loss": 0.0365, "step": 24989 }, { "epoch": 7.69, "learning_rate": 2.6796890880236383e-06, "loss": 0.0435, "step": 24990 }, { "epoch": 7.69, "learning_rate": 2.6790102102816707e-06, "loss": 0.0797, "step": 24991 }, { "epoch": 7.69, "learning_rate": 2.6783314052438758e-06, "loss": 0.043, "step": 24992 }, { "epoch": 7.69, "learning_rate": 2.6776526729170017e-06, "loss": 0.0868, "step": 24993 }, { "epoch": 7.69, "learning_rate": 2.6769740133077847e-06, "loss": 0.0562, "step": 24994 }, { "epoch": 7.69, "learning_rate": 2.6762954264229657e-06, "loss": 0.0625, "step": 24995 }, { "epoch": 7.69, "learning_rate": 2.675616912269281e-06, "loss": 0.036, "step": 24996 }, { "epoch": 7.69, "learning_rate": 2.674938470853472e-06, "loss": 0.0232, "step": 24997 }, { "epoch": 7.69, "learning_rate": 2.6742601021822746e-06, "loss": 0.0303, "step": 24998 }, { "epoch": 7.69, "learning_rate": 2.673581806262423e-06, "loss": 0.0344, "step": 24999 }, { "epoch": 7.69, "learning_rate": 2.67290358310066e-06, "loss": 0.0575, "step": 25000 }, { "epoch": 7.69, "learning_rate": 2.67222543270372e-06, "loss": 0.0732, "step": 25001 }, { "epoch": 7.69, "learning_rate": 2.6715473550783278e-06, "loss": 0.0583, "step": 25002 }, { "epoch": 7.69, "learning_rate": 2.6708693502312287e-06, "loss": 0.0519, "step": 25003 }, { "epoch": 7.69, "learning_rate": 2.6701914181691515e-06, "loss": 0.0458, "step": 25004 }, { "epoch": 7.69, "learning_rate": 2.6695135588988285e-06, "loss": 0.0886, "step": 25005 }, { "epoch": 7.69, "learning_rate": 2.668835772426992e-06, "loss": 0.0464, "step": 25006 }, { "epoch": 7.69, "learning_rate": 2.668158058760373e-06, "loss": 0.0486, "step": 25007 }, { "epoch": 7.69, "learning_rate": 2.6674804179057013e-06, "loss": 0.0517, "step": 25008 }, { "epoch": 7.69, "learning_rate": 2.666802849869704e-06, "loss": 0.0611, "step": 25009 }, { "epoch": 7.7, "learning_rate": 2.6661253546591158e-06, "loss": 0.0345, "step": 25010 }, { "epoch": 7.7, "learning_rate": 2.665447932280665e-06, "loss": 0.052, "step": 25011 }, { "epoch": 7.7, "learning_rate": 2.6647705827410697e-06, "loss": 0.0489, "step": 25012 }, { "epoch": 7.7, "learning_rate": 2.6640933060470663e-06, "loss": 0.0303, "step": 25013 }, { "epoch": 7.7, "learning_rate": 2.663416102205377e-06, "loss": 0.0454, "step": 25014 }, { "epoch": 7.7, "learning_rate": 2.662738971222728e-06, "loss": 0.0523, "step": 25015 }, { "epoch": 7.7, "learning_rate": 2.6620619131058432e-06, "loss": 0.0388, "step": 25016 }, { "epoch": 7.7, "learning_rate": 2.6613849278614436e-06, "loss": 0.0557, "step": 25017 }, { "epoch": 7.7, "learning_rate": 2.6607080154962607e-06, "loss": 0.0513, "step": 25018 }, { "epoch": 7.7, "learning_rate": 2.660031176017005e-06, "loss": 0.0537, "step": 25019 }, { "epoch": 7.7, "learning_rate": 2.659354409430408e-06, "loss": 0.0351, "step": 25020 }, { "epoch": 7.7, "learning_rate": 2.6586777157431865e-06, "loss": 0.0789, "step": 25021 }, { "epoch": 7.7, "learning_rate": 2.6580010949620615e-06, "loss": 0.044, "step": 25022 }, { "epoch": 7.7, "learning_rate": 2.6573245470937527e-06, "loss": 0.048, "step": 25023 }, { "epoch": 7.7, "learning_rate": 2.6566480721449773e-06, "loss": 0.0276, "step": 25024 }, { "epoch": 7.7, "learning_rate": 2.6559716701224547e-06, "loss": 0.0468, "step": 25025 }, { "epoch": 7.7, "learning_rate": 2.655295341032903e-06, "loss": 0.046, "step": 25026 }, { "epoch": 7.7, "learning_rate": 2.6546190848830333e-06, "loss": 0.0234, "step": 25027 }, { "epoch": 7.7, "learning_rate": 2.6539429016795726e-06, "loss": 0.0201, "step": 25028 }, { "epoch": 7.7, "learning_rate": 2.653266791429222e-06, "loss": 0.071, "step": 25029 }, { "epoch": 7.7, "learning_rate": 2.6525907541387075e-06, "loss": 0.0791, "step": 25030 }, { "epoch": 7.7, "learning_rate": 2.651914789814738e-06, "loss": 0.0378, "step": 25031 }, { "epoch": 7.7, "learning_rate": 2.651238898464027e-06, "loss": 0.0538, "step": 25032 }, { "epoch": 7.7, "learning_rate": 2.650563080093286e-06, "loss": 0.0374, "step": 25033 }, { "epoch": 7.7, "learning_rate": 2.6498873347092247e-06, "loss": 0.0206, "step": 25034 }, { "epoch": 7.7, "learning_rate": 2.6492116623185616e-06, "loss": 0.0405, "step": 25035 }, { "epoch": 7.7, "learning_rate": 2.648536062927999e-06, "loss": 0.0531, "step": 25036 }, { "epoch": 7.7, "learning_rate": 2.6478605365442446e-06, "loss": 0.041, "step": 25037 }, { "epoch": 7.7, "learning_rate": 2.647185083174014e-06, "loss": 0.0796, "step": 25038 }, { "epoch": 7.7, "learning_rate": 2.646509702824014e-06, "loss": 0.0373, "step": 25039 }, { "epoch": 7.7, "learning_rate": 2.645834395500948e-06, "loss": 0.044, "step": 25040 }, { "epoch": 7.7, "learning_rate": 2.6451591612115245e-06, "loss": 0.0626, "step": 25041 }, { "epoch": 7.7, "learning_rate": 2.6444839999624496e-06, "loss": 0.0357, "step": 25042 }, { "epoch": 7.71, "learning_rate": 2.643808911760427e-06, "loss": 0.06, "step": 25043 }, { "epoch": 7.71, "learning_rate": 2.643133896612159e-06, "loss": 0.0379, "step": 25044 }, { "epoch": 7.71, "learning_rate": 2.642458954524357e-06, "loss": 0.0427, "step": 25045 }, { "epoch": 7.71, "learning_rate": 2.641784085503717e-06, "loss": 0.063, "step": 25046 }, { "epoch": 7.71, "learning_rate": 2.641109289556938e-06, "loss": 0.0582, "step": 25047 }, { "epoch": 7.71, "learning_rate": 2.6404345666907295e-06, "loss": 0.0523, "step": 25048 }, { "epoch": 7.71, "learning_rate": 2.639759916911788e-06, "loss": 0.0459, "step": 25049 }, { "epoch": 7.71, "learning_rate": 2.6390853402268147e-06, "loss": 0.0788, "step": 25050 }, { "epoch": 7.71, "learning_rate": 2.6384108366425075e-06, "loss": 0.0526, "step": 25051 }, { "epoch": 7.71, "learning_rate": 2.6377364061655665e-06, "loss": 0.0597, "step": 25052 }, { "epoch": 7.71, "learning_rate": 2.6370620488026866e-06, "loss": 0.0525, "step": 25053 }, { "epoch": 7.71, "learning_rate": 2.636387764560564e-06, "loss": 0.0374, "step": 25054 }, { "epoch": 7.71, "learning_rate": 2.635713553445901e-06, "loss": 0.0597, "step": 25055 }, { "epoch": 7.71, "learning_rate": 2.6350394154653926e-06, "loss": 0.0652, "step": 25056 }, { "epoch": 7.71, "learning_rate": 2.634365350625724e-06, "loss": 0.0457, "step": 25057 }, { "epoch": 7.71, "learning_rate": 2.633691358933599e-06, "loss": 0.0541, "step": 25058 }, { "epoch": 7.71, "learning_rate": 2.6330174403957077e-06, "loss": 0.0326, "step": 25059 }, { "epoch": 7.71, "learning_rate": 2.6323435950187428e-06, "loss": 0.0323, "step": 25060 }, { "epoch": 7.71, "learning_rate": 2.631669822809396e-06, "loss": 0.0393, "step": 25061 }, { "epoch": 7.71, "learning_rate": 2.6309961237743587e-06, "loss": 0.0476, "step": 25062 }, { "epoch": 7.71, "learning_rate": 2.6303224979203214e-06, "loss": 0.0543, "step": 25063 }, { "epoch": 7.71, "learning_rate": 2.6296489452539707e-06, "loss": 0.0548, "step": 25064 }, { "epoch": 7.71, "learning_rate": 2.628975465782002e-06, "loss": 0.0265, "step": 25065 }, { "epoch": 7.71, "learning_rate": 2.628302059511104e-06, "loss": 0.0927, "step": 25066 }, { "epoch": 7.71, "learning_rate": 2.6276287264479537e-06, "loss": 0.0439, "step": 25067 }, { "epoch": 7.71, "learning_rate": 2.626955466599248e-06, "loss": 0.0712, "step": 25068 }, { "epoch": 7.71, "learning_rate": 2.626282279971669e-06, "loss": 0.0534, "step": 25069 }, { "epoch": 7.71, "learning_rate": 2.625609166571903e-06, "loss": 0.0239, "step": 25070 }, { "epoch": 7.71, "learning_rate": 2.624936126406635e-06, "loss": 0.0373, "step": 25071 }, { "epoch": 7.71, "learning_rate": 2.6242631594825452e-06, "loss": 0.0511, "step": 25072 }, { "epoch": 7.71, "learning_rate": 2.623590265806326e-06, "loss": 0.0455, "step": 25073 }, { "epoch": 7.71, "learning_rate": 2.6229174453846474e-06, "loss": 0.0649, "step": 25074 }, { "epoch": 7.72, "learning_rate": 2.6222446982242e-06, "loss": 0.0356, "step": 25075 }, { "epoch": 7.72, "learning_rate": 2.6215720243316636e-06, "loss": 0.0766, "step": 25076 }, { "epoch": 7.72, "learning_rate": 2.6208994237137166e-06, "loss": 0.0732, "step": 25077 }, { "epoch": 7.72, "learning_rate": 2.620226896377038e-06, "loss": 0.0401, "step": 25078 }, { "epoch": 7.72, "learning_rate": 2.6195544423283057e-06, "loss": 0.0371, "step": 25079 }, { "epoch": 7.72, "learning_rate": 2.6188820615742062e-06, "loss": 0.0682, "step": 25080 }, { "epoch": 7.72, "learning_rate": 2.6182097541214067e-06, "loss": 0.0504, "step": 25081 }, { "epoch": 7.72, "learning_rate": 2.617537519976585e-06, "loss": 0.0517, "step": 25082 }, { "epoch": 7.72, "learning_rate": 2.6168653591464254e-06, "loss": 0.1253, "step": 25083 }, { "epoch": 7.72, "learning_rate": 2.6161932716375916e-06, "loss": 0.072, "step": 25084 }, { "epoch": 7.72, "learning_rate": 2.615521257456767e-06, "loss": 0.027, "step": 25085 }, { "epoch": 7.72, "learning_rate": 2.6148493166106215e-06, "loss": 0.026, "step": 25086 }, { "epoch": 7.72, "learning_rate": 2.614177449105828e-06, "loss": 0.0335, "step": 25087 }, { "epoch": 7.72, "learning_rate": 2.61350565494906e-06, "loss": 0.0487, "step": 25088 }, { "epoch": 7.72, "learning_rate": 2.6128339341469855e-06, "loss": 0.0436, "step": 25089 }, { "epoch": 7.72, "learning_rate": 2.6121622867062844e-06, "loss": 0.0478, "step": 25090 }, { "epoch": 7.72, "learning_rate": 2.6114907126336176e-06, "loss": 0.056, "step": 25091 }, { "epoch": 7.72, "learning_rate": 2.6108192119356546e-06, "loss": 0.0285, "step": 25092 }, { "epoch": 7.72, "learning_rate": 2.610147784619069e-06, "loss": 0.0387, "step": 25093 }, { "epoch": 7.72, "learning_rate": 2.6094764306905284e-06, "loss": 0.0306, "step": 25094 }, { "epoch": 7.72, "learning_rate": 2.608805150156697e-06, "loss": 0.0425, "step": 25095 }, { "epoch": 7.72, "learning_rate": 2.6081339430242426e-06, "loss": 0.0661, "step": 25096 }, { "epoch": 7.72, "learning_rate": 2.60746280929983e-06, "loss": 0.0389, "step": 25097 }, { "epoch": 7.72, "learning_rate": 2.606791748990125e-06, "loss": 0.0414, "step": 25098 }, { "epoch": 7.72, "learning_rate": 2.6061207621017893e-06, "loss": 0.0428, "step": 25099 }, { "epoch": 7.72, "learning_rate": 2.605449848641495e-06, "loss": 0.0576, "step": 25100 }, { "epoch": 7.72, "learning_rate": 2.604779008615895e-06, "loss": 0.0452, "step": 25101 }, { "epoch": 7.72, "learning_rate": 2.604108242031652e-06, "loss": 0.0236, "step": 25102 }, { "epoch": 7.72, "learning_rate": 2.6034375488954343e-06, "loss": 0.0235, "step": 25103 }, { "epoch": 7.72, "learning_rate": 2.602766929213898e-06, "loss": 0.0396, "step": 25104 }, { "epoch": 7.72, "learning_rate": 2.602096382993704e-06, "loss": 0.0347, "step": 25105 }, { "epoch": 7.72, "learning_rate": 2.6014259102415106e-06, "loss": 0.0455, "step": 25106 }, { "epoch": 7.72, "learning_rate": 2.600755510963977e-06, "loss": 0.0775, "step": 25107 }, { "epoch": 7.73, "learning_rate": 2.6000851851677598e-06, "loss": 0.0467, "step": 25108 }, { "epoch": 7.73, "learning_rate": 2.599414932859514e-06, "loss": 0.0729, "step": 25109 }, { "epoch": 7.73, "learning_rate": 2.598744754045902e-06, "loss": 0.0582, "step": 25110 }, { "epoch": 7.73, "learning_rate": 2.5980746487335785e-06, "loss": 0.0597, "step": 25111 }, { "epoch": 7.73, "learning_rate": 2.59740461692919e-06, "loss": 0.0533, "step": 25112 }, { "epoch": 7.73, "learning_rate": 2.5967346586393983e-06, "loss": 0.0424, "step": 25113 }, { "epoch": 7.73, "learning_rate": 2.5960647738708553e-06, "loss": 0.0542, "step": 25114 }, { "epoch": 7.73, "learning_rate": 2.5953949626302123e-06, "loss": 0.0245, "step": 25115 }, { "epoch": 7.73, "learning_rate": 2.5947252249241218e-06, "loss": 0.0332, "step": 25116 }, { "epoch": 7.73, "learning_rate": 2.5940555607592312e-06, "loss": 0.0468, "step": 25117 }, { "epoch": 7.73, "learning_rate": 2.5933859701422005e-06, "loss": 0.0317, "step": 25118 }, { "epoch": 7.73, "learning_rate": 2.5927164530796677e-06, "loss": 0.0423, "step": 25119 }, { "epoch": 7.73, "learning_rate": 2.5920470095782902e-06, "loss": 0.0698, "step": 25120 }, { "epoch": 7.73, "learning_rate": 2.5913776396447155e-06, "loss": 0.0706, "step": 25121 }, { "epoch": 7.73, "learning_rate": 2.590708343285584e-06, "loss": 0.063, "step": 25122 }, { "epoch": 7.73, "learning_rate": 2.590039120507549e-06, "loss": 0.0503, "step": 25123 }, { "epoch": 7.73, "learning_rate": 2.589369971317256e-06, "loss": 0.057, "step": 25124 }, { "epoch": 7.73, "learning_rate": 2.588700895721348e-06, "loss": 0.0242, "step": 25125 }, { "epoch": 7.73, "learning_rate": 2.588031893726468e-06, "loss": 0.0343, "step": 25126 }, { "epoch": 7.73, "learning_rate": 2.5873629653392653e-06, "loss": 0.083, "step": 25127 }, { "epoch": 7.73, "learning_rate": 2.5866941105663836e-06, "loss": 0.0469, "step": 25128 }, { "epoch": 7.73, "learning_rate": 2.5860253294144555e-06, "loss": 0.0494, "step": 25129 }, { "epoch": 7.73, "learning_rate": 2.5853566218901303e-06, "loss": 0.0656, "step": 25130 }, { "epoch": 7.73, "learning_rate": 2.5846879880000496e-06, "loss": 0.0303, "step": 25131 }, { "epoch": 7.73, "learning_rate": 2.5840194277508503e-06, "loss": 0.0519, "step": 25132 }, { "epoch": 7.73, "learning_rate": 2.583350941149173e-06, "loss": 0.0632, "step": 25133 }, { "epoch": 7.73, "learning_rate": 2.5826825282016523e-06, "loss": 0.0423, "step": 25134 }, { "epoch": 7.73, "learning_rate": 2.5820141889149377e-06, "loss": 0.0493, "step": 25135 }, { "epoch": 7.73, "learning_rate": 2.5813459232956518e-06, "loss": 0.0435, "step": 25136 }, { "epoch": 7.73, "learning_rate": 2.5806777313504416e-06, "loss": 0.0394, "step": 25137 }, { "epoch": 7.73, "learning_rate": 2.5800096130859386e-06, "loss": 0.0375, "step": 25138 }, { "epoch": 7.73, "learning_rate": 2.5793415685087797e-06, "loss": 0.0644, "step": 25139 }, { "epoch": 7.74, "learning_rate": 2.578673597625597e-06, "loss": 0.0185, "step": 25140 }, { "epoch": 7.74, "learning_rate": 2.578005700443026e-06, "loss": 0.051, "step": 25141 }, { "epoch": 7.74, "learning_rate": 2.5773378769676983e-06, "loss": 0.065, "step": 25142 }, { "epoch": 7.74, "learning_rate": 2.5766701272062466e-06, "loss": 0.0662, "step": 25143 }, { "epoch": 7.74, "learning_rate": 2.5760024511652994e-06, "loss": 0.0598, "step": 25144 }, { "epoch": 7.74, "learning_rate": 2.575334848851495e-06, "loss": 0.0679, "step": 25145 }, { "epoch": 7.74, "learning_rate": 2.5746673202714534e-06, "loss": 0.0298, "step": 25146 }, { "epoch": 7.74, "learning_rate": 2.573999865431811e-06, "loss": 0.0801, "step": 25147 }, { "epoch": 7.74, "learning_rate": 2.573332484339194e-06, "loss": 0.0457, "step": 25148 }, { "epoch": 7.74, "learning_rate": 2.5726651770002297e-06, "loss": 0.0514, "step": 25149 }, { "epoch": 7.74, "learning_rate": 2.571997943421546e-06, "loss": 0.0792, "step": 25150 }, { "epoch": 7.74, "learning_rate": 2.5713307836097657e-06, "loss": 0.0608, "step": 25151 }, { "epoch": 7.74, "learning_rate": 2.5706636975715225e-06, "loss": 0.0362, "step": 25152 }, { "epoch": 7.74, "learning_rate": 2.569996685313434e-06, "loss": 0.0243, "step": 25153 }, { "epoch": 7.74, "learning_rate": 2.569329746842123e-06, "loss": 0.0473, "step": 25154 }, { "epoch": 7.74, "learning_rate": 2.568662882164218e-06, "loss": 0.0636, "step": 25155 }, { "epoch": 7.74, "learning_rate": 2.56799609128634e-06, "loss": 0.0296, "step": 25156 }, { "epoch": 7.74, "learning_rate": 2.5673293742151095e-06, "loss": 0.0347, "step": 25157 }, { "epoch": 7.74, "learning_rate": 2.5666627309571502e-06, "loss": 0.059, "step": 25158 }, { "epoch": 7.74, "learning_rate": 2.5659961615190788e-06, "loss": 0.0382, "step": 25159 }, { "epoch": 7.74, "learning_rate": 2.5653296659075176e-06, "loss": 0.0446, "step": 25160 }, { "epoch": 7.74, "learning_rate": 2.564663244129083e-06, "loss": 0.028, "step": 25161 }, { "epoch": 7.74, "learning_rate": 2.5639968961903996e-06, "loss": 0.0564, "step": 25162 }, { "epoch": 7.74, "learning_rate": 2.563330622098078e-06, "loss": 0.0408, "step": 25163 }, { "epoch": 7.74, "learning_rate": 2.562664421858734e-06, "loss": 0.0413, "step": 25164 }, { "epoch": 7.74, "learning_rate": 2.561998295478989e-06, "loss": 0.0091, "step": 25165 }, { "epoch": 7.74, "learning_rate": 2.5613322429654573e-06, "loss": 0.0511, "step": 25166 }, { "epoch": 7.74, "learning_rate": 2.5606662643247515e-06, "loss": 0.0415, "step": 25167 }, { "epoch": 7.74, "learning_rate": 2.560000359563486e-06, "loss": 0.0423, "step": 25168 }, { "epoch": 7.74, "learning_rate": 2.5593345286882732e-06, "loss": 0.0483, "step": 25169 }, { "epoch": 7.74, "learning_rate": 2.5586687717057267e-06, "loss": 0.0467, "step": 25170 }, { "epoch": 7.74, "learning_rate": 2.5580030886224537e-06, "loss": 0.0441, "step": 25171 }, { "epoch": 7.74, "learning_rate": 2.5573374794450724e-06, "loss": 0.0408, "step": 25172 }, { "epoch": 7.75, "learning_rate": 2.556671944180192e-06, "loss": 0.0668, "step": 25173 }, { "epoch": 7.75, "learning_rate": 2.5560064828344145e-06, "loss": 0.0455, "step": 25174 }, { "epoch": 7.75, "learning_rate": 2.555341095414354e-06, "loss": 0.0483, "step": 25175 }, { "epoch": 7.75, "learning_rate": 2.5546757819266197e-06, "loss": 0.0506, "step": 25176 }, { "epoch": 7.75, "learning_rate": 2.554010542377815e-06, "loss": 0.0468, "step": 25177 }, { "epoch": 7.75, "learning_rate": 2.55334537677455e-06, "loss": 0.0432, "step": 25178 }, { "epoch": 7.75, "learning_rate": 2.5526802851234268e-06, "loss": 0.0619, "step": 25179 }, { "epoch": 7.75, "learning_rate": 2.552015267431053e-06, "loss": 0.0615, "step": 25180 }, { "epoch": 7.75, "learning_rate": 2.5513503237040273e-06, "loss": 0.0446, "step": 25181 }, { "epoch": 7.75, "learning_rate": 2.5506854539489624e-06, "loss": 0.04, "step": 25182 }, { "epoch": 7.75, "learning_rate": 2.5500206581724584e-06, "loss": 0.0598, "step": 25183 }, { "epoch": 7.75, "learning_rate": 2.5493559363811093e-06, "loss": 0.0904, "step": 25184 }, { "epoch": 7.75, "learning_rate": 2.5486912885815262e-06, "loss": 0.0477, "step": 25185 }, { "epoch": 7.75, "learning_rate": 2.5480267147803064e-06, "loss": 0.0304, "step": 25186 }, { "epoch": 7.75, "learning_rate": 2.547362214984048e-06, "loss": 0.0585, "step": 25187 }, { "epoch": 7.75, "learning_rate": 2.5466977891993518e-06, "loss": 0.0761, "step": 25188 }, { "epoch": 7.75, "learning_rate": 2.5460334374328134e-06, "loss": 0.0587, "step": 25189 }, { "epoch": 7.75, "learning_rate": 2.545369159691038e-06, "loss": 0.066, "step": 25190 }, { "epoch": 7.75, "learning_rate": 2.5447049559806116e-06, "loss": 0.0393, "step": 25191 }, { "epoch": 7.75, "learning_rate": 2.5440408263081385e-06, "loss": 0.0391, "step": 25192 }, { "epoch": 7.75, "learning_rate": 2.5433767706802114e-06, "loss": 0.0481, "step": 25193 }, { "epoch": 7.75, "learning_rate": 2.5427127891034265e-06, "loss": 0.0517, "step": 25194 }, { "epoch": 7.75, "learning_rate": 2.542048881584376e-06, "loss": 0.041, "step": 25195 }, { "epoch": 7.75, "learning_rate": 2.5413850481296497e-06, "loss": 0.0403, "step": 25196 }, { "epoch": 7.75, "learning_rate": 2.54072128874585e-06, "loss": 0.034, "step": 25197 }, { "epoch": 7.75, "learning_rate": 2.5400576034395606e-06, "loss": 0.0401, "step": 25198 }, { "epoch": 7.75, "learning_rate": 2.5393939922173707e-06, "loss": 0.0517, "step": 25199 }, { "epoch": 7.75, "learning_rate": 2.5387304550858806e-06, "loss": 0.0383, "step": 25200 }, { "epoch": 7.75, "learning_rate": 2.5380669920516677e-06, "loss": 0.0445, "step": 25201 }, { "epoch": 7.75, "learning_rate": 2.537403603121329e-06, "loss": 0.087, "step": 25202 }, { "epoch": 7.75, "learning_rate": 2.5367402883014512e-06, "loss": 0.0696, "step": 25203 }, { "epoch": 7.75, "learning_rate": 2.53607704759862e-06, "loss": 0.143, "step": 25204 }, { "epoch": 7.76, "learning_rate": 2.535413881019423e-06, "loss": 0.0321, "step": 25205 }, { "epoch": 7.76, "learning_rate": 2.5347507885704424e-06, "loss": 0.0647, "step": 25206 }, { "epoch": 7.76, "learning_rate": 2.5340877702582734e-06, "loss": 0.0413, "step": 25207 }, { "epoch": 7.76, "learning_rate": 2.53342482608949e-06, "loss": 0.0582, "step": 25208 }, { "epoch": 7.76, "learning_rate": 2.532761956070676e-06, "loss": 0.0244, "step": 25209 }, { "epoch": 7.76, "learning_rate": 2.532099160208422e-06, "loss": 0.0412, "step": 25210 }, { "epoch": 7.76, "learning_rate": 2.5314364385093047e-06, "loss": 0.0615, "step": 25211 }, { "epoch": 7.76, "learning_rate": 2.530773790979908e-06, "loss": 0.0493, "step": 25212 }, { "epoch": 7.76, "learning_rate": 2.5301112176268104e-06, "loss": 0.0486, "step": 25213 }, { "epoch": 7.76, "learning_rate": 2.529448718456594e-06, "loss": 0.0493, "step": 25214 }, { "epoch": 7.76, "learning_rate": 2.5287862934758356e-06, "loss": 0.0341, "step": 25215 }, { "epoch": 7.76, "learning_rate": 2.528123942691113e-06, "loss": 0.0276, "step": 25216 }, { "epoch": 7.76, "learning_rate": 2.527461666109009e-06, "loss": 0.0271, "step": 25217 }, { "epoch": 7.76, "learning_rate": 2.526799463736099e-06, "loss": 0.0364, "step": 25218 }, { "epoch": 7.76, "learning_rate": 2.5261373355789532e-06, "loss": 0.0469, "step": 25219 }, { "epoch": 7.76, "learning_rate": 2.5254752816441552e-06, "loss": 0.0825, "step": 25220 }, { "epoch": 7.76, "learning_rate": 2.5248133019382747e-06, "loss": 0.0637, "step": 25221 }, { "epoch": 7.76, "learning_rate": 2.524151396467889e-06, "loss": 0.0612, "step": 25222 }, { "epoch": 7.76, "learning_rate": 2.5234895652395687e-06, "loss": 0.0618, "step": 25223 }, { "epoch": 7.76, "learning_rate": 2.522827808259889e-06, "loss": 0.0514, "step": 25224 }, { "epoch": 7.76, "learning_rate": 2.522166125535419e-06, "loss": 0.0516, "step": 25225 }, { "epoch": 7.76, "learning_rate": 2.521504517072728e-06, "loss": 0.057, "step": 25226 }, { "epoch": 7.76, "learning_rate": 2.5208429828783943e-06, "loss": 0.042, "step": 25227 }, { "epoch": 7.76, "learning_rate": 2.520181522958984e-06, "loss": 0.0497, "step": 25228 }, { "epoch": 7.76, "learning_rate": 2.5195201373210597e-06, "loss": 0.0582, "step": 25229 }, { "epoch": 7.76, "learning_rate": 2.5188588259711967e-06, "loss": 0.0906, "step": 25230 }, { "epoch": 7.76, "learning_rate": 2.5181975889159615e-06, "loss": 0.0807, "step": 25231 }, { "epoch": 7.76, "learning_rate": 2.5175364261619195e-06, "loss": 0.0909, "step": 25232 }, { "epoch": 7.76, "learning_rate": 2.5168753377156363e-06, "loss": 0.0711, "step": 25233 }, { "epoch": 7.76, "learning_rate": 2.5162143235836744e-06, "loss": 0.0633, "step": 25234 }, { "epoch": 7.76, "learning_rate": 2.515553383772609e-06, "loss": 0.0376, "step": 25235 }, { "epoch": 7.76, "learning_rate": 2.514892518288988e-06, "loss": 0.0508, "step": 25236 }, { "epoch": 7.76, "learning_rate": 2.5142317271393877e-06, "loss": 0.0615, "step": 25237 }, { "epoch": 7.77, "learning_rate": 2.5135710103303645e-06, "loss": 0.0499, "step": 25238 }, { "epoch": 7.77, "learning_rate": 2.5129103678684816e-06, "loss": 0.0437, "step": 25239 }, { "epoch": 7.77, "learning_rate": 2.512249799760299e-06, "loss": 0.0531, "step": 25240 }, { "epoch": 7.77, "learning_rate": 2.511589306012376e-06, "loss": 0.0597, "step": 25241 }, { "epoch": 7.77, "learning_rate": 2.5109288866312733e-06, "loss": 0.0519, "step": 25242 }, { "epoch": 7.77, "learning_rate": 2.5102685416235497e-06, "loss": 0.0338, "step": 25243 }, { "epoch": 7.77, "learning_rate": 2.509608270995758e-06, "loss": 0.0545, "step": 25244 }, { "epoch": 7.77, "learning_rate": 2.5089480747544657e-06, "loss": 0.0613, "step": 25245 }, { "epoch": 7.77, "learning_rate": 2.5082879529062165e-06, "loss": 0.09, "step": 25246 }, { "epoch": 7.77, "learning_rate": 2.5076279054575768e-06, "loss": 0.0513, "step": 25247 }, { "epoch": 7.77, "learning_rate": 2.5069679324150966e-06, "loss": 0.0595, "step": 25248 }, { "epoch": 7.77, "learning_rate": 2.5063080337853294e-06, "loss": 0.0286, "step": 25249 }, { "epoch": 7.77, "learning_rate": 2.5056482095748313e-06, "loss": 0.0545, "step": 25250 }, { "epoch": 7.77, "learning_rate": 2.5049884597901508e-06, "loss": 0.1474, "step": 25251 }, { "epoch": 7.77, "learning_rate": 2.504328784437847e-06, "loss": 0.076, "step": 25252 }, { "epoch": 7.77, "learning_rate": 2.503669183524464e-06, "loss": 0.0561, "step": 25253 }, { "epoch": 7.77, "learning_rate": 2.503009657056553e-06, "loss": 0.0316, "step": 25254 }, { "epoch": 7.77, "learning_rate": 2.502350205040668e-06, "loss": 0.0612, "step": 25255 }, { "epoch": 7.77, "learning_rate": 2.501690827483355e-06, "loss": 0.0558, "step": 25256 }, { "epoch": 7.77, "learning_rate": 2.501031524391163e-06, "loss": 0.0632, "step": 25257 }, { "epoch": 7.77, "learning_rate": 2.500372295770639e-06, "loss": 0.0197, "step": 25258 }, { "epoch": 7.77, "learning_rate": 2.4997131416283304e-06, "loss": 0.0553, "step": 25259 }, { "epoch": 7.77, "learning_rate": 2.499054061970783e-06, "loss": 0.0781, "step": 25260 }, { "epoch": 7.77, "learning_rate": 2.4983950568045377e-06, "loss": 0.0513, "step": 25261 }, { "epoch": 7.77, "learning_rate": 2.4977361261361497e-06, "loss": 0.06, "step": 25262 }, { "epoch": 7.77, "learning_rate": 2.4970772699721535e-06, "loss": 0.0424, "step": 25263 }, { "epoch": 7.77, "learning_rate": 2.4964184883190922e-06, "loss": 0.0419, "step": 25264 }, { "epoch": 7.77, "learning_rate": 2.495759781183513e-06, "loss": 0.0556, "step": 25265 }, { "epoch": 7.77, "learning_rate": 2.4951011485719556e-06, "loss": 0.0592, "step": 25266 }, { "epoch": 7.77, "learning_rate": 2.494442590490961e-06, "loss": 0.0352, "step": 25267 }, { "epoch": 7.77, "learning_rate": 2.4937841069470683e-06, "loss": 0.0433, "step": 25268 }, { "epoch": 7.77, "learning_rate": 2.4931256979468175e-06, "loss": 0.0517, "step": 25269 }, { "epoch": 7.78, "learning_rate": 2.492467363496747e-06, "loss": 0.0465, "step": 25270 }, { "epoch": 7.78, "learning_rate": 2.491809103603392e-06, "loss": 0.0384, "step": 25271 }, { "epoch": 7.78, "learning_rate": 2.491150918273295e-06, "loss": 0.0537, "step": 25272 }, { "epoch": 7.78, "learning_rate": 2.490492807512992e-06, "loss": 0.0491, "step": 25273 }, { "epoch": 7.78, "learning_rate": 2.4898347713290116e-06, "loss": 0.0489, "step": 25274 }, { "epoch": 7.78, "learning_rate": 2.4891768097278955e-06, "loss": 0.0373, "step": 25275 }, { "epoch": 7.78, "learning_rate": 2.4885189227161767e-06, "loss": 0.0686, "step": 25276 }, { "epoch": 7.78, "learning_rate": 2.487861110300388e-06, "loss": 0.0414, "step": 25277 }, { "epoch": 7.78, "learning_rate": 2.4872033724870613e-06, "loss": 0.0437, "step": 25278 }, { "epoch": 7.78, "learning_rate": 2.4865457092827294e-06, "loss": 0.079, "step": 25279 }, { "epoch": 7.78, "learning_rate": 2.4858881206939236e-06, "loss": 0.0463, "step": 25280 }, { "epoch": 7.78, "learning_rate": 2.485230606727169e-06, "loss": 0.0544, "step": 25281 }, { "epoch": 7.78, "learning_rate": 2.484573167389005e-06, "loss": 0.0254, "step": 25282 }, { "epoch": 7.78, "learning_rate": 2.483915802685959e-06, "loss": 0.0278, "step": 25283 }, { "epoch": 7.78, "learning_rate": 2.4832585126245503e-06, "loss": 0.0936, "step": 25284 }, { "epoch": 7.78, "learning_rate": 2.4826012972113145e-06, "loss": 0.0383, "step": 25285 }, { "epoch": 7.78, "learning_rate": 2.481944156452776e-06, "loss": 0.0377, "step": 25286 }, { "epoch": 7.78, "learning_rate": 2.481287090355461e-06, "loss": 0.0526, "step": 25287 }, { "epoch": 7.78, "learning_rate": 2.480630098925895e-06, "loss": 0.0606, "step": 25288 }, { "epoch": 7.78, "learning_rate": 2.4799731821705987e-06, "loss": 0.0531, "step": 25289 }, { "epoch": 7.78, "learning_rate": 2.4793163400961053e-06, "loss": 0.0601, "step": 25290 }, { "epoch": 7.78, "learning_rate": 2.478659572708926e-06, "loss": 0.0282, "step": 25291 }, { "epoch": 7.78, "learning_rate": 2.478002880015592e-06, "loss": 0.058, "step": 25292 }, { "epoch": 7.78, "learning_rate": 2.477346262022622e-06, "loss": 0.0427, "step": 25293 }, { "epoch": 7.78, "learning_rate": 2.4766897187365368e-06, "loss": 0.0319, "step": 25294 }, { "epoch": 7.78, "learning_rate": 2.476033250163856e-06, "loss": 0.0624, "step": 25295 }, { "epoch": 7.78, "learning_rate": 2.475376856311097e-06, "loss": 0.0375, "step": 25296 }, { "epoch": 7.78, "learning_rate": 2.474720537184786e-06, "loss": 0.0574, "step": 25297 }, { "epoch": 7.78, "learning_rate": 2.4740642927914327e-06, "loss": 0.0576, "step": 25298 }, { "epoch": 7.78, "learning_rate": 2.473408123137554e-06, "loss": 0.0623, "step": 25299 }, { "epoch": 7.78, "learning_rate": 2.472752028229676e-06, "loss": 0.0457, "step": 25300 }, { "epoch": 7.78, "learning_rate": 2.472096008074301e-06, "loss": 0.0301, "step": 25301 }, { "epoch": 7.78, "learning_rate": 2.4714400626779533e-06, "loss": 0.052, "step": 25302 }, { "epoch": 7.79, "learning_rate": 2.4707841920471454e-06, "loss": 0.0549, "step": 25303 }, { "epoch": 7.79, "learning_rate": 2.4701283961883894e-06, "loss": 0.0464, "step": 25304 }, { "epoch": 7.79, "learning_rate": 2.4694726751081978e-06, "loss": 0.0478, "step": 25305 }, { "epoch": 7.79, "learning_rate": 2.4688170288130795e-06, "loss": 0.0478, "step": 25306 }, { "epoch": 7.79, "learning_rate": 2.468161457309556e-06, "loss": 0.0459, "step": 25307 }, { "epoch": 7.79, "learning_rate": 2.4675059606041265e-06, "loss": 0.1019, "step": 25308 }, { "epoch": 7.79, "learning_rate": 2.4668505387033025e-06, "loss": 0.0558, "step": 25309 }, { "epoch": 7.79, "learning_rate": 2.4661951916135985e-06, "loss": 0.037, "step": 25310 }, { "epoch": 7.79, "learning_rate": 2.46553991934152e-06, "loss": 0.0391, "step": 25311 }, { "epoch": 7.79, "learning_rate": 2.4648847218935734e-06, "loss": 0.0512, "step": 25312 }, { "epoch": 7.79, "learning_rate": 2.4642295992762665e-06, "loss": 0.0375, "step": 25313 }, { "epoch": 7.79, "learning_rate": 2.463574551496105e-06, "loss": 0.0606, "step": 25314 }, { "epoch": 7.79, "learning_rate": 2.462919578559594e-06, "loss": 0.0367, "step": 25315 }, { "epoch": 7.79, "learning_rate": 2.4622646804732342e-06, "loss": 0.0383, "step": 25316 }, { "epoch": 7.79, "learning_rate": 2.4616098572435363e-06, "loss": 0.025, "step": 25317 }, { "epoch": 7.79, "learning_rate": 2.4609551088770035e-06, "loss": 0.0529, "step": 25318 }, { "epoch": 7.79, "learning_rate": 2.4603004353801284e-06, "loss": 0.053, "step": 25319 }, { "epoch": 7.79, "learning_rate": 2.459645836759422e-06, "loss": 0.0342, "step": 25320 }, { "epoch": 7.79, "learning_rate": 2.4589913130213815e-06, "loss": 0.0567, "step": 25321 }, { "epoch": 7.79, "learning_rate": 2.458336864172508e-06, "loss": 0.0508, "step": 25322 }, { "epoch": 7.79, "learning_rate": 2.4576824902192964e-06, "loss": 0.0558, "step": 25323 }, { "epoch": 7.79, "learning_rate": 2.4570281911682557e-06, "loss": 0.049, "step": 25324 }, { "epoch": 7.79, "learning_rate": 2.456373967025872e-06, "loss": 0.0571, "step": 25325 }, { "epoch": 7.79, "learning_rate": 2.4557198177986464e-06, "loss": 0.0471, "step": 25326 }, { "epoch": 7.79, "learning_rate": 2.455065743493078e-06, "loss": 0.0338, "step": 25327 }, { "epoch": 7.79, "learning_rate": 2.4544117441156602e-06, "loss": 0.0527, "step": 25328 }, { "epoch": 7.79, "learning_rate": 2.4537578196728886e-06, "loss": 0.0624, "step": 25329 }, { "epoch": 7.79, "learning_rate": 2.4531039701712566e-06, "loss": 0.0398, "step": 25330 }, { "epoch": 7.79, "learning_rate": 2.4524501956172563e-06, "loss": 0.0829, "step": 25331 }, { "epoch": 7.79, "learning_rate": 2.4517964960173836e-06, "loss": 0.0823, "step": 25332 }, { "epoch": 7.79, "learning_rate": 2.451142871378124e-06, "loss": 0.0508, "step": 25333 }, { "epoch": 7.79, "learning_rate": 2.4504893217059768e-06, "loss": 0.0743, "step": 25334 }, { "epoch": 7.8, "learning_rate": 2.44983584700743e-06, "loss": 0.0437, "step": 25335 }, { "epoch": 7.8, "learning_rate": 2.4491824472889667e-06, "loss": 0.0705, "step": 25336 }, { "epoch": 7.8, "learning_rate": 2.448529122557084e-06, "loss": 0.0631, "step": 25337 }, { "epoch": 7.8, "learning_rate": 2.447875872818266e-06, "loss": 0.0529, "step": 25338 }, { "epoch": 7.8, "learning_rate": 2.4472226980790015e-06, "loss": 0.048, "step": 25339 }, { "epoch": 7.8, "learning_rate": 2.446569598345776e-06, "loss": 0.052, "step": 25340 }, { "epoch": 7.8, "learning_rate": 2.4459165736250756e-06, "loss": 0.0412, "step": 25341 }, { "epoch": 7.8, "learning_rate": 2.445263623923386e-06, "loss": 0.0602, "step": 25342 }, { "epoch": 7.8, "learning_rate": 2.444610749247188e-06, "loss": 0.0572, "step": 25343 }, { "epoch": 7.8, "learning_rate": 2.443957949602973e-06, "loss": 0.0739, "step": 25344 }, { "epoch": 7.8, "learning_rate": 2.443305224997221e-06, "loss": 0.0444, "step": 25345 }, { "epoch": 7.8, "learning_rate": 2.442652575436406e-06, "loss": 0.0505, "step": 25346 }, { "epoch": 7.8, "learning_rate": 2.4420000009270205e-06, "loss": 0.055, "step": 25347 }, { "epoch": 7.8, "learning_rate": 2.4413475014755396e-06, "loss": 0.0658, "step": 25348 }, { "epoch": 7.8, "learning_rate": 2.440695077088445e-06, "loss": 0.0661, "step": 25349 }, { "epoch": 7.8, "learning_rate": 2.440042727772215e-06, "loss": 0.0534, "step": 25350 }, { "epoch": 7.8, "learning_rate": 2.4393904535333244e-06, "loss": 0.0918, "step": 25351 }, { "epoch": 7.8, "learning_rate": 2.4387382543782613e-06, "loss": 0.0423, "step": 25352 }, { "epoch": 7.8, "learning_rate": 2.4380861303134893e-06, "loss": 0.0317, "step": 25353 }, { "epoch": 7.8, "learning_rate": 2.4374340813454954e-06, "loss": 0.0523, "step": 25354 }, { "epoch": 7.8, "learning_rate": 2.4367821074807506e-06, "loss": 0.024, "step": 25355 }, { "epoch": 7.8, "learning_rate": 2.43613020872573e-06, "loss": 0.0434, "step": 25356 }, { "epoch": 7.8, "learning_rate": 2.435478385086907e-06, "loss": 0.0678, "step": 25357 }, { "epoch": 7.8, "learning_rate": 2.434826636570755e-06, "loss": 0.0241, "step": 25358 }, { "epoch": 7.8, "learning_rate": 2.4341749631837475e-06, "loss": 0.0494, "step": 25359 }, { "epoch": 7.8, "learning_rate": 2.433523364932355e-06, "loss": 0.0306, "step": 25360 }, { "epoch": 7.8, "learning_rate": 2.432871841823047e-06, "loss": 0.0476, "step": 25361 }, { "epoch": 7.8, "learning_rate": 2.432220393862301e-06, "loss": 0.0377, "step": 25362 }, { "epoch": 7.8, "learning_rate": 2.4315690210565745e-06, "loss": 0.0662, "step": 25363 }, { "epoch": 7.8, "learning_rate": 2.430917723412347e-06, "loss": 0.0531, "step": 25364 }, { "epoch": 7.8, "learning_rate": 2.4302665009360814e-06, "loss": 0.0681, "step": 25365 }, { "epoch": 7.8, "learning_rate": 2.429615353634247e-06, "loss": 0.0594, "step": 25366 }, { "epoch": 7.8, "learning_rate": 2.4289642815133086e-06, "loss": 0.0514, "step": 25367 }, { "epoch": 7.81, "learning_rate": 2.4283132845797297e-06, "loss": 0.0675, "step": 25368 }, { "epoch": 7.81, "learning_rate": 2.427662362839984e-06, "loss": 0.0615, "step": 25369 }, { "epoch": 7.81, "learning_rate": 2.427011516300527e-06, "loss": 0.0726, "step": 25370 }, { "epoch": 7.81, "learning_rate": 2.4263607449678216e-06, "loss": 0.0399, "step": 25371 }, { "epoch": 7.81, "learning_rate": 2.4257100488483377e-06, "loss": 0.0728, "step": 25372 }, { "epoch": 7.81, "learning_rate": 2.425059427948534e-06, "loss": 0.0155, "step": 25373 }, { "epoch": 7.81, "learning_rate": 2.42440888227487e-06, "loss": 0.0226, "step": 25374 }, { "epoch": 7.81, "learning_rate": 2.4237584118338086e-06, "loss": 0.0557, "step": 25375 }, { "epoch": 7.81, "learning_rate": 2.4231080166318087e-06, "loss": 0.0806, "step": 25376 }, { "epoch": 7.81, "learning_rate": 2.42245769667533e-06, "loss": 0.0381, "step": 25377 }, { "epoch": 7.81, "learning_rate": 2.421807451970826e-06, "loss": 0.0598, "step": 25378 }, { "epoch": 7.81, "learning_rate": 2.421157282524764e-06, "loss": 0.0739, "step": 25379 }, { "epoch": 7.81, "learning_rate": 2.4205071883435928e-06, "loss": 0.0711, "step": 25380 }, { "epoch": 7.81, "learning_rate": 2.419857169433768e-06, "loss": 0.0479, "step": 25381 }, { "epoch": 7.81, "learning_rate": 2.4192072258017496e-06, "loss": 0.0577, "step": 25382 }, { "epoch": 7.81, "learning_rate": 2.4185573574539913e-06, "loss": 0.0299, "step": 25383 }, { "epoch": 7.81, "learning_rate": 2.417907564396944e-06, "loss": 0.0333, "step": 25384 }, { "epoch": 7.81, "learning_rate": 2.417257846637063e-06, "loss": 0.0514, "step": 25385 }, { "epoch": 7.81, "learning_rate": 2.4166082041808013e-06, "loss": 0.0488, "step": 25386 }, { "epoch": 7.81, "learning_rate": 2.415958637034609e-06, "loss": 0.063, "step": 25387 }, { "epoch": 7.81, "learning_rate": 2.4153091452049324e-06, "loss": 0.0306, "step": 25388 }, { "epoch": 7.81, "learning_rate": 2.414659728698231e-06, "loss": 0.0374, "step": 25389 }, { "epoch": 7.81, "learning_rate": 2.414010387520951e-06, "loss": 0.0628, "step": 25390 }, { "epoch": 7.81, "learning_rate": 2.413361121679534e-06, "loss": 0.059, "step": 25391 }, { "epoch": 7.81, "learning_rate": 2.4127119311804358e-06, "loss": 0.0252, "step": 25392 }, { "epoch": 7.81, "learning_rate": 2.412062816030102e-06, "loss": 0.0593, "step": 25393 }, { "epoch": 7.81, "learning_rate": 2.4114137762349766e-06, "loss": 0.0574, "step": 25394 }, { "epoch": 7.81, "learning_rate": 2.410764811801506e-06, "loss": 0.0662, "step": 25395 }, { "epoch": 7.81, "learning_rate": 2.4101159227361326e-06, "loss": 0.084, "step": 25396 }, { "epoch": 7.81, "learning_rate": 2.409467109045309e-06, "loss": 0.0412, "step": 25397 }, { "epoch": 7.81, "learning_rate": 2.4088183707354673e-06, "loss": 0.0532, "step": 25398 }, { "epoch": 7.81, "learning_rate": 2.408169707813057e-06, "loss": 0.0365, "step": 25399 }, { "epoch": 7.82, "learning_rate": 2.407521120284523e-06, "loss": 0.0513, "step": 25400 }, { "epoch": 7.82, "learning_rate": 2.4068726081562953e-06, "loss": 0.0555, "step": 25401 }, { "epoch": 7.82, "learning_rate": 2.406224171434823e-06, "loss": 0.0231, "step": 25402 }, { "epoch": 7.82, "learning_rate": 2.4055758101265438e-06, "loss": 0.0606, "step": 25403 }, { "epoch": 7.82, "learning_rate": 2.4049275242378955e-06, "loss": 0.0369, "step": 25404 }, { "epoch": 7.82, "learning_rate": 2.4042793137753174e-06, "loss": 0.0304, "step": 25405 }, { "epoch": 7.82, "learning_rate": 2.403631178745243e-06, "loss": 0.0544, "step": 25406 }, { "epoch": 7.82, "learning_rate": 2.4029831191541176e-06, "loss": 0.0458, "step": 25407 }, { "epoch": 7.82, "learning_rate": 2.4023351350083657e-06, "loss": 0.0236, "step": 25408 }, { "epoch": 7.82, "learning_rate": 2.4016872263144308e-06, "loss": 0.0558, "step": 25409 }, { "epoch": 7.82, "learning_rate": 2.4010393930787445e-06, "loss": 0.0618, "step": 25410 }, { "epoch": 7.82, "learning_rate": 2.4003916353077416e-06, "loss": 0.0433, "step": 25411 }, { "epoch": 7.82, "learning_rate": 2.3997439530078526e-06, "loss": 0.0432, "step": 25412 }, { "epoch": 7.82, "learning_rate": 2.3990963461855075e-06, "loss": 0.0711, "step": 25413 }, { "epoch": 7.82, "learning_rate": 2.3984488148471473e-06, "loss": 0.037, "step": 25414 }, { "epoch": 7.82, "learning_rate": 2.3978013589991946e-06, "loss": 0.0379, "step": 25415 }, { "epoch": 7.82, "learning_rate": 2.3971539786480757e-06, "loss": 0.0606, "step": 25416 }, { "epoch": 7.82, "learning_rate": 2.3965066738002316e-06, "loss": 0.0652, "step": 25417 }, { "epoch": 7.82, "learning_rate": 2.3958594444620786e-06, "loss": 0.0334, "step": 25418 }, { "epoch": 7.82, "learning_rate": 2.395212290640051e-06, "loss": 0.0692, "step": 25419 }, { "epoch": 7.82, "learning_rate": 2.394565212340575e-06, "loss": 0.0374, "step": 25420 }, { "epoch": 7.82, "learning_rate": 2.393918209570075e-06, "loss": 0.0874, "step": 25421 }, { "epoch": 7.82, "learning_rate": 2.393271282334978e-06, "loss": 0.0271, "step": 25422 }, { "epoch": 7.82, "learning_rate": 2.392624430641706e-06, "loss": 0.0493, "step": 25423 }, { "epoch": 7.82, "learning_rate": 2.3919776544966876e-06, "loss": 0.0353, "step": 25424 }, { "epoch": 7.82, "learning_rate": 2.3913309539063424e-06, "loss": 0.0957, "step": 25425 }, { "epoch": 7.82, "learning_rate": 2.390684328877089e-06, "loss": 0.0213, "step": 25426 }, { "epoch": 7.82, "learning_rate": 2.390037779415356e-06, "loss": 0.0665, "step": 25427 }, { "epoch": 7.82, "learning_rate": 2.389391305527562e-06, "loss": 0.0304, "step": 25428 }, { "epoch": 7.82, "learning_rate": 2.388744907220126e-06, "loss": 0.0471, "step": 25429 }, { "epoch": 7.82, "learning_rate": 2.3880985844994674e-06, "loss": 0.0433, "step": 25430 }, { "epoch": 7.82, "learning_rate": 2.387452337372006e-06, "loss": 0.0329, "step": 25431 }, { "epoch": 7.82, "learning_rate": 2.386806165844159e-06, "loss": 0.0293, "step": 25432 }, { "epoch": 7.83, "learning_rate": 2.3861600699223397e-06, "loss": 0.0532, "step": 25433 }, { "epoch": 7.83, "learning_rate": 2.3855140496129714e-06, "loss": 0.0567, "step": 25434 }, { "epoch": 7.83, "learning_rate": 2.3848681049224686e-06, "loss": 0.0505, "step": 25435 }, { "epoch": 7.83, "learning_rate": 2.3842222358572387e-06, "loss": 0.065, "step": 25436 }, { "epoch": 7.83, "learning_rate": 2.383576442423704e-06, "loss": 0.0395, "step": 25437 }, { "epoch": 7.83, "learning_rate": 2.382930724628274e-06, "loss": 0.0514, "step": 25438 }, { "epoch": 7.83, "learning_rate": 2.3822850824773623e-06, "loss": 0.0759, "step": 25439 }, { "epoch": 7.83, "learning_rate": 2.381639515977381e-06, "loss": 0.0466, "step": 25440 }, { "epoch": 7.83, "learning_rate": 2.3809940251347396e-06, "loss": 0.0562, "step": 25441 }, { "epoch": 7.83, "learning_rate": 2.3803486099558504e-06, "loss": 0.0313, "step": 25442 }, { "epoch": 7.83, "learning_rate": 2.379703270447118e-06, "loss": 0.0488, "step": 25443 }, { "epoch": 7.83, "learning_rate": 2.3790580066149583e-06, "loss": 0.0561, "step": 25444 }, { "epoch": 7.83, "learning_rate": 2.3784128184657797e-06, "loss": 0.0553, "step": 25445 }, { "epoch": 7.83, "learning_rate": 2.37776770600598e-06, "loss": 0.0469, "step": 25446 }, { "epoch": 7.83, "learning_rate": 2.3771226692419745e-06, "loss": 0.0291, "step": 25447 }, { "epoch": 7.83, "learning_rate": 2.376477708180167e-06, "loss": 0.0592, "step": 25448 }, { "epoch": 7.83, "learning_rate": 2.375832822826962e-06, "loss": 0.0408, "step": 25449 }, { "epoch": 7.83, "learning_rate": 2.3751880131887627e-06, "loss": 0.0304, "step": 25450 }, { "epoch": 7.83, "learning_rate": 2.3745432792719715e-06, "loss": 0.0371, "step": 25451 }, { "epoch": 7.83, "learning_rate": 2.3738986210829997e-06, "loss": 0.0536, "step": 25452 }, { "epoch": 7.83, "learning_rate": 2.3732540386282356e-06, "loss": 0.0562, "step": 25453 }, { "epoch": 7.83, "learning_rate": 2.372609531914092e-06, "loss": 0.0228, "step": 25454 }, { "epoch": 7.83, "learning_rate": 2.3719651009469646e-06, "loss": 0.0383, "step": 25455 }, { "epoch": 7.83, "learning_rate": 2.3713207457332542e-06, "loss": 0.0507, "step": 25456 }, { "epoch": 7.83, "learning_rate": 2.3706764662793603e-06, "loss": 0.0486, "step": 25457 }, { "epoch": 7.83, "learning_rate": 2.3700322625916806e-06, "loss": 0.0251, "step": 25458 }, { "epoch": 7.83, "learning_rate": 2.3693881346766113e-06, "loss": 0.0399, "step": 25459 }, { "epoch": 7.83, "learning_rate": 2.3687440825405518e-06, "loss": 0.0335, "step": 25460 }, { "epoch": 7.83, "learning_rate": 2.368100106189892e-06, "loss": 0.0772, "step": 25461 }, { "epoch": 7.83, "learning_rate": 2.367456205631039e-06, "loss": 0.0466, "step": 25462 }, { "epoch": 7.83, "learning_rate": 2.3668123808703746e-06, "loss": 0.0303, "step": 25463 }, { "epoch": 7.83, "learning_rate": 2.3661686319143006e-06, "loss": 0.056, "step": 25464 }, { "epoch": 7.84, "learning_rate": 2.3655249587692073e-06, "loss": 0.0272, "step": 25465 }, { "epoch": 7.84, "learning_rate": 2.364881361441488e-06, "loss": 0.0633, "step": 25466 }, { "epoch": 7.84, "learning_rate": 2.3642378399375332e-06, "loss": 0.0848, "step": 25467 }, { "epoch": 7.84, "learning_rate": 2.363594394263731e-06, "loss": 0.0652, "step": 25468 }, { "epoch": 7.84, "learning_rate": 2.362951024426481e-06, "loss": 0.0725, "step": 25469 }, { "epoch": 7.84, "learning_rate": 2.362307730432162e-06, "loss": 0.0567, "step": 25470 }, { "epoch": 7.84, "learning_rate": 2.361664512287163e-06, "loss": 0.0519, "step": 25471 }, { "epoch": 7.84, "learning_rate": 2.361021369997879e-06, "loss": 0.0472, "step": 25472 }, { "epoch": 7.84, "learning_rate": 2.3603783035706927e-06, "loss": 0.0297, "step": 25473 }, { "epoch": 7.84, "learning_rate": 2.3597353130119903e-06, "loss": 0.0322, "step": 25474 }, { "epoch": 7.84, "learning_rate": 2.359092398328159e-06, "loss": 0.0307, "step": 25475 }, { "epoch": 7.84, "learning_rate": 2.3584495595255806e-06, "loss": 0.0269, "step": 25476 }, { "epoch": 7.84, "learning_rate": 2.3578067966106423e-06, "loss": 0.0464, "step": 25477 }, { "epoch": 7.84, "learning_rate": 2.3571641095897223e-06, "loss": 0.0516, "step": 25478 }, { "epoch": 7.84, "learning_rate": 2.356521498469212e-06, "loss": 0.0507, "step": 25479 }, { "epoch": 7.84, "learning_rate": 2.3558789632554844e-06, "loss": 0.0515, "step": 25480 }, { "epoch": 7.84, "learning_rate": 2.355236503954922e-06, "loss": 0.0369, "step": 25481 }, { "epoch": 7.84, "learning_rate": 2.3545941205739086e-06, "loss": 0.0633, "step": 25482 }, { "epoch": 7.84, "learning_rate": 2.353951813118822e-06, "loss": 0.0302, "step": 25483 }, { "epoch": 7.84, "learning_rate": 2.3533095815960417e-06, "loss": 0.0516, "step": 25484 }, { "epoch": 7.84, "learning_rate": 2.352667426011943e-06, "loss": 0.0291, "step": 25485 }, { "epoch": 7.84, "learning_rate": 2.3520253463729058e-06, "loss": 0.0496, "step": 25486 }, { "epoch": 7.84, "learning_rate": 2.351383342685306e-06, "loss": 0.0456, "step": 25487 }, { "epoch": 7.84, "learning_rate": 2.3507414149555153e-06, "loss": 0.0438, "step": 25488 }, { "epoch": 7.84, "learning_rate": 2.350099563189915e-06, "loss": 0.0665, "step": 25489 }, { "epoch": 7.84, "learning_rate": 2.3494577873948787e-06, "loss": 0.0566, "step": 25490 }, { "epoch": 7.84, "learning_rate": 2.3488160875767717e-06, "loss": 0.0196, "step": 25491 }, { "epoch": 7.84, "learning_rate": 2.3481744637419768e-06, "loss": 0.0572, "step": 25492 }, { "epoch": 7.84, "learning_rate": 2.3475329158968606e-06, "loss": 0.0568, "step": 25493 }, { "epoch": 7.84, "learning_rate": 2.3468914440477953e-06, "loss": 0.0649, "step": 25494 }, { "epoch": 7.84, "learning_rate": 2.3462500482011508e-06, "loss": 0.04, "step": 25495 }, { "epoch": 7.84, "learning_rate": 2.345608728363298e-06, "loss": 0.0496, "step": 25496 }, { "epoch": 7.84, "learning_rate": 2.3449674845406046e-06, "loss": 0.0467, "step": 25497 }, { "epoch": 7.85, "learning_rate": 2.344326316739437e-06, "loss": 0.0549, "step": 25498 }, { "epoch": 7.85, "learning_rate": 2.3436852249661667e-06, "loss": 0.0249, "step": 25499 }, { "epoch": 7.85, "learning_rate": 2.343044209227162e-06, "loss": 0.0498, "step": 25500 }, { "epoch": 7.85, "learning_rate": 2.342403269528779e-06, "loss": 0.0412, "step": 25501 }, { "epoch": 7.85, "learning_rate": 2.3417624058773915e-06, "loss": 0.0274, "step": 25502 }, { "epoch": 7.85, "learning_rate": 2.341121618279362e-06, "loss": 0.0909, "step": 25503 }, { "epoch": 7.85, "learning_rate": 2.340480906741053e-06, "loss": 0.0735, "step": 25504 }, { "epoch": 7.85, "learning_rate": 2.3398402712688285e-06, "loss": 0.0661, "step": 25505 }, { "epoch": 7.85, "learning_rate": 2.339199711869047e-06, "loss": 0.0336, "step": 25506 }, { "epoch": 7.85, "learning_rate": 2.338559228548077e-06, "loss": 0.1086, "step": 25507 }, { "epoch": 7.85, "learning_rate": 2.3379188213122706e-06, "loss": 0.0501, "step": 25508 }, { "epoch": 7.85, "learning_rate": 2.3372784901679947e-06, "loss": 0.0663, "step": 25509 }, { "epoch": 7.85, "learning_rate": 2.3366382351216057e-06, "loss": 0.0412, "step": 25510 }, { "epoch": 7.85, "learning_rate": 2.3359980561794603e-06, "loss": 0.0304, "step": 25511 }, { "epoch": 7.85, "learning_rate": 2.3353579533479187e-06, "loss": 0.0732, "step": 25512 }, { "epoch": 7.85, "learning_rate": 2.334717926633334e-06, "loss": 0.057, "step": 25513 }, { "epoch": 7.85, "learning_rate": 2.3340779760420695e-06, "loss": 0.0856, "step": 25514 }, { "epoch": 7.85, "learning_rate": 2.3334381015804742e-06, "loss": 0.0444, "step": 25515 }, { "epoch": 7.85, "learning_rate": 2.332798303254901e-06, "loss": 0.0486, "step": 25516 }, { "epoch": 7.85, "learning_rate": 2.332158581071712e-06, "loss": 0.0519, "step": 25517 }, { "epoch": 7.85, "learning_rate": 2.33151893503725e-06, "loss": 0.0533, "step": 25518 }, { "epoch": 7.85, "learning_rate": 2.3308793651578755e-06, "loss": 0.0455, "step": 25519 }, { "epoch": 7.85, "learning_rate": 2.3302398714399367e-06, "loss": 0.055, "step": 25520 }, { "epoch": 7.85, "learning_rate": 2.3296004538897843e-06, "loss": 0.0348, "step": 25521 }, { "epoch": 7.85, "learning_rate": 2.328961112513769e-06, "loss": 0.0403, "step": 25522 }, { "epoch": 7.85, "learning_rate": 2.3283218473182366e-06, "loss": 0.0337, "step": 25523 }, { "epoch": 7.85, "learning_rate": 2.3276826583095445e-06, "loss": 0.0431, "step": 25524 }, { "epoch": 7.85, "learning_rate": 2.32704354549403e-06, "loss": 0.0319, "step": 25525 }, { "epoch": 7.85, "learning_rate": 2.3264045088780463e-06, "loss": 0.0604, "step": 25526 }, { "epoch": 7.85, "learning_rate": 2.3257655484679376e-06, "loss": 0.048, "step": 25527 }, { "epoch": 7.85, "learning_rate": 2.3251266642700508e-06, "loss": 0.0473, "step": 25528 }, { "epoch": 7.85, "learning_rate": 2.324487856290729e-06, "loss": 0.0301, "step": 25529 }, { "epoch": 7.86, "learning_rate": 2.323849124536315e-06, "loss": 0.0436, "step": 25530 }, { "epoch": 7.86, "learning_rate": 2.3232104690131585e-06, "loss": 0.0688, "step": 25531 }, { "epoch": 7.86, "learning_rate": 2.322571889727594e-06, "loss": 0.0888, "step": 25532 }, { "epoch": 7.86, "learning_rate": 2.321933386685964e-06, "loss": 0.0253, "step": 25533 }, { "epoch": 7.86, "learning_rate": 2.3212949598946143e-06, "loss": 0.0671, "step": 25534 }, { "epoch": 7.86, "learning_rate": 2.3206566093598824e-06, "loss": 0.0492, "step": 25535 }, { "epoch": 7.86, "learning_rate": 2.3200183350881087e-06, "loss": 0.0599, "step": 25536 }, { "epoch": 7.86, "learning_rate": 2.3193801370856304e-06, "loss": 0.066, "step": 25537 }, { "epoch": 7.86, "learning_rate": 2.318742015358787e-06, "loss": 0.0331, "step": 25538 }, { "epoch": 7.86, "learning_rate": 2.318103969913914e-06, "loss": 0.0254, "step": 25539 }, { "epoch": 7.86, "learning_rate": 2.3174660007573446e-06, "loss": 0.0422, "step": 25540 }, { "epoch": 7.86, "learning_rate": 2.3168281078954256e-06, "loss": 0.0418, "step": 25541 }, { "epoch": 7.86, "learning_rate": 2.31619029133448e-06, "loss": 0.0377, "step": 25542 }, { "epoch": 7.86, "learning_rate": 2.3155525510808453e-06, "loss": 0.0821, "step": 25543 }, { "epoch": 7.86, "learning_rate": 2.3149148871408577e-06, "loss": 0.0494, "step": 25544 }, { "epoch": 7.86, "learning_rate": 2.3142772995208485e-06, "loss": 0.0479, "step": 25545 }, { "epoch": 7.86, "learning_rate": 2.313639788227149e-06, "loss": 0.0503, "step": 25546 }, { "epoch": 7.86, "learning_rate": 2.3130023532660896e-06, "loss": 0.077, "step": 25547 }, { "epoch": 7.86, "learning_rate": 2.312364994644002e-06, "loss": 0.0506, "step": 25548 }, { "epoch": 7.86, "learning_rate": 2.3117277123672154e-06, "loss": 0.0618, "step": 25549 }, { "epoch": 7.86, "learning_rate": 2.3110905064420553e-06, "loss": 0.0318, "step": 25550 }, { "epoch": 7.86, "learning_rate": 2.310453376874855e-06, "loss": 0.0899, "step": 25551 }, { "epoch": 7.86, "learning_rate": 2.3098163236719427e-06, "loss": 0.0639, "step": 25552 }, { "epoch": 7.86, "learning_rate": 2.309179346839637e-06, "loss": 0.0428, "step": 25553 }, { "epoch": 7.86, "learning_rate": 2.308542446384271e-06, "loss": 0.0675, "step": 25554 }, { "epoch": 7.86, "learning_rate": 2.3079056223121666e-06, "loss": 0.0694, "step": 25555 }, { "epoch": 7.86, "learning_rate": 2.307268874629649e-06, "loss": 0.0351, "step": 25556 }, { "epoch": 7.86, "learning_rate": 2.3066322033430412e-06, "loss": 0.0528, "step": 25557 }, { "epoch": 7.86, "learning_rate": 2.3059956084586656e-06, "loss": 0.0195, "step": 25558 }, { "epoch": 7.86, "learning_rate": 2.3053590899828446e-06, "loss": 0.024, "step": 25559 }, { "epoch": 7.86, "learning_rate": 2.3047226479218978e-06, "loss": 0.037, "step": 25560 }, { "epoch": 7.86, "learning_rate": 2.3040862822821485e-06, "loss": 0.0576, "step": 25561 }, { "epoch": 7.86, "learning_rate": 2.3034499930699194e-06, "loss": 0.0683, "step": 25562 }, { "epoch": 7.87, "learning_rate": 2.3028137802915185e-06, "loss": 0.033, "step": 25563 }, { "epoch": 7.87, "learning_rate": 2.302177643953274e-06, "loss": 0.0387, "step": 25564 }, { "epoch": 7.87, "learning_rate": 2.3015415840614997e-06, "loss": 0.0612, "step": 25565 }, { "epoch": 7.87, "learning_rate": 2.3009056006225127e-06, "loss": 0.0357, "step": 25566 }, { "epoch": 7.87, "learning_rate": 2.3002696936426294e-06, "loss": 0.0613, "step": 25567 }, { "epoch": 7.87, "learning_rate": 2.2996338631281613e-06, "loss": 0.0542, "step": 25568 }, { "epoch": 7.87, "learning_rate": 2.2989981090854306e-06, "loss": 0.0376, "step": 25569 }, { "epoch": 7.87, "learning_rate": 2.298362431520742e-06, "loss": 0.0417, "step": 25570 }, { "epoch": 7.87, "learning_rate": 2.2977268304404145e-06, "loss": 0.0637, "step": 25571 }, { "epoch": 7.87, "learning_rate": 2.297091305850757e-06, "loss": 0.066, "step": 25572 }, { "epoch": 7.87, "learning_rate": 2.296455857758084e-06, "loss": 0.0746, "step": 25573 }, { "epoch": 7.87, "learning_rate": 2.295820486168703e-06, "loss": 0.0691, "step": 25574 }, { "epoch": 7.87, "learning_rate": 2.2951851910889256e-06, "loss": 0.0588, "step": 25575 }, { "epoch": 7.87, "learning_rate": 2.29454997252506e-06, "loss": 0.042, "step": 25576 }, { "epoch": 7.87, "learning_rate": 2.2939148304834135e-06, "loss": 0.0499, "step": 25577 }, { "epoch": 7.87, "learning_rate": 2.293279764970293e-06, "loss": 0.03, "step": 25578 }, { "epoch": 7.87, "learning_rate": 2.2926447759920134e-06, "loss": 0.0721, "step": 25579 }, { "epoch": 7.87, "learning_rate": 2.292009863554868e-06, "loss": 0.0412, "step": 25580 }, { "epoch": 7.87, "learning_rate": 2.2913750276651724e-06, "loss": 0.0661, "step": 25581 }, { "epoch": 7.87, "learning_rate": 2.2907402683292268e-06, "loss": 0.0558, "step": 25582 }, { "epoch": 7.87, "learning_rate": 2.290105585553335e-06, "loss": 0.069, "step": 25583 }, { "epoch": 7.87, "learning_rate": 2.2894709793438006e-06, "loss": 0.026, "step": 25584 }, { "epoch": 7.87, "learning_rate": 2.2888364497069227e-06, "loss": 0.0734, "step": 25585 }, { "epoch": 7.87, "learning_rate": 2.288201996649011e-06, "loss": 0.0481, "step": 25586 }, { "epoch": 7.87, "learning_rate": 2.2875676201763584e-06, "loss": 0.0487, "step": 25587 }, { "epoch": 7.87, "learning_rate": 2.286933320295265e-06, "loss": 0.0534, "step": 25588 }, { "epoch": 7.87, "learning_rate": 2.2862990970120347e-06, "loss": 0.0336, "step": 25589 }, { "epoch": 7.87, "learning_rate": 2.2856649503329633e-06, "loss": 0.0582, "step": 25590 }, { "epoch": 7.87, "learning_rate": 2.2850308802643483e-06, "loss": 0.04, "step": 25591 }, { "epoch": 7.87, "learning_rate": 2.2843968868124876e-06, "loss": 0.0706, "step": 25592 }, { "epoch": 7.87, "learning_rate": 2.283762969983677e-06, "loss": 0.0609, "step": 25593 }, { "epoch": 7.87, "learning_rate": 2.283129129784212e-06, "loss": 0.0227, "step": 25594 }, { "epoch": 7.88, "learning_rate": 2.2824953662203832e-06, "loss": 0.0325, "step": 25595 }, { "epoch": 7.88, "learning_rate": 2.281861679298493e-06, "loss": 0.0501, "step": 25596 }, { "epoch": 7.88, "learning_rate": 2.2812280690248277e-06, "loss": 0.0458, "step": 25597 }, { "epoch": 7.88, "learning_rate": 2.280594535405678e-06, "loss": 0.0331, "step": 25598 }, { "epoch": 7.88, "learning_rate": 2.2799610784473425e-06, "loss": 0.0489, "step": 25599 }, { "epoch": 7.88, "learning_rate": 2.2793276981561084e-06, "loss": 0.0465, "step": 25600 }, { "epoch": 7.88, "learning_rate": 2.2786943945382654e-06, "loss": 0.071, "step": 25601 }, { "epoch": 7.88, "learning_rate": 2.278061167600104e-06, "loss": 0.0275, "step": 25602 }, { "epoch": 7.88, "learning_rate": 2.2774280173479115e-06, "loss": 0.0262, "step": 25603 }, { "epoch": 7.88, "learning_rate": 2.2767949437879765e-06, "loss": 0.0698, "step": 25604 }, { "epoch": 7.88, "learning_rate": 2.2761619469265818e-06, "loss": 0.0644, "step": 25605 }, { "epoch": 7.88, "learning_rate": 2.2755290267700212e-06, "loss": 0.0195, "step": 25606 }, { "epoch": 7.88, "learning_rate": 2.2748961833245786e-06, "loss": 0.0479, "step": 25607 }, { "epoch": 7.88, "learning_rate": 2.2742634165965317e-06, "loss": 0.045, "step": 25608 }, { "epoch": 7.88, "learning_rate": 2.273630726592172e-06, "loss": 0.0618, "step": 25609 }, { "epoch": 7.88, "learning_rate": 2.272998113317779e-06, "loss": 0.0233, "step": 25610 }, { "epoch": 7.88, "learning_rate": 2.272365576779637e-06, "loss": 0.0616, "step": 25611 }, { "epoch": 7.88, "learning_rate": 2.2717331169840275e-06, "loss": 0.0445, "step": 25612 }, { "epoch": 7.88, "learning_rate": 2.271100733937227e-06, "loss": 0.0505, "step": 25613 }, { "epoch": 7.88, "learning_rate": 2.2704684276455247e-06, "loss": 0.0467, "step": 25614 }, { "epoch": 7.88, "learning_rate": 2.2698361981151894e-06, "loss": 0.0215, "step": 25615 }, { "epoch": 7.88, "learning_rate": 2.269204045352509e-06, "loss": 0.0432, "step": 25616 }, { "epoch": 7.88, "learning_rate": 2.2685719693637588e-06, "loss": 0.0471, "step": 25617 }, { "epoch": 7.88, "learning_rate": 2.2679399701552085e-06, "loss": 0.0196, "step": 25618 }, { "epoch": 7.88, "learning_rate": 2.2673080477331442e-06, "loss": 0.038, "step": 25619 }, { "epoch": 7.88, "learning_rate": 2.266676202103837e-06, "loss": 0.0481, "step": 25620 }, { "epoch": 7.88, "learning_rate": 2.266044433273562e-06, "loss": 0.051, "step": 25621 }, { "epoch": 7.88, "learning_rate": 2.2654127412485947e-06, "loss": 0.0551, "step": 25622 }, { "epoch": 7.88, "learning_rate": 2.264781126035204e-06, "loss": 0.0706, "step": 25623 }, { "epoch": 7.88, "learning_rate": 2.2641495876396713e-06, "loss": 0.0631, "step": 25624 }, { "epoch": 7.88, "learning_rate": 2.263518126068257e-06, "loss": 0.0475, "step": 25625 }, { "epoch": 7.88, "learning_rate": 2.2628867413272402e-06, "loss": 0.0509, "step": 25626 }, { "epoch": 7.88, "learning_rate": 2.2622554334228896e-06, "loss": 0.0412, "step": 25627 }, { "epoch": 7.89, "learning_rate": 2.261624202361473e-06, "loss": 0.0674, "step": 25628 }, { "epoch": 7.89, "learning_rate": 2.2609930481492614e-06, "loss": 0.046, "step": 25629 }, { "epoch": 7.89, "learning_rate": 2.2603619707925174e-06, "loss": 0.0802, "step": 25630 }, { "epoch": 7.89, "learning_rate": 2.259730970297518e-06, "loss": 0.0369, "step": 25631 }, { "epoch": 7.89, "learning_rate": 2.2591000466705215e-06, "loss": 0.0275, "step": 25632 }, { "epoch": 7.89, "learning_rate": 2.258469199917792e-06, "loss": 0.0337, "step": 25633 }, { "epoch": 7.89, "learning_rate": 2.2578384300456014e-06, "loss": 0.067, "step": 25634 }, { "epoch": 7.89, "learning_rate": 2.25720773706021e-06, "loss": 0.0725, "step": 25635 }, { "epoch": 7.89, "learning_rate": 2.2565771209678822e-06, "loss": 0.0567, "step": 25636 }, { "epoch": 7.89, "learning_rate": 2.2559465817748803e-06, "loss": 0.0447, "step": 25637 }, { "epoch": 7.89, "learning_rate": 2.255316119487466e-06, "loss": 0.0273, "step": 25638 }, { "epoch": 7.89, "learning_rate": 2.2546857341119e-06, "loss": 0.0209, "step": 25639 }, { "epoch": 7.89, "learning_rate": 2.25405542565444e-06, "loss": 0.0495, "step": 25640 }, { "epoch": 7.89, "learning_rate": 2.253425194121355e-06, "loss": 0.0622, "step": 25641 }, { "epoch": 7.89, "learning_rate": 2.2527950395188934e-06, "loss": 0.082, "step": 25642 }, { "epoch": 7.89, "learning_rate": 2.2521649618533146e-06, "loss": 0.0712, "step": 25643 }, { "epoch": 7.89, "learning_rate": 2.251534961130881e-06, "loss": 0.028, "step": 25644 }, { "epoch": 7.89, "learning_rate": 2.250905037357848e-06, "loss": 0.0832, "step": 25645 }, { "epoch": 7.89, "learning_rate": 2.2502751905404675e-06, "loss": 0.0422, "step": 25646 }, { "epoch": 7.89, "learning_rate": 2.249645420684998e-06, "loss": 0.0543, "step": 25647 }, { "epoch": 7.89, "learning_rate": 2.249015727797692e-06, "loss": 0.0675, "step": 25648 }, { "epoch": 7.89, "learning_rate": 2.248386111884804e-06, "loss": 0.0604, "step": 25649 }, { "epoch": 7.89, "learning_rate": 2.247756572952583e-06, "loss": 0.036, "step": 25650 }, { "epoch": 7.89, "learning_rate": 2.2471271110072856e-06, "loss": 0.0236, "step": 25651 }, { "epoch": 7.89, "learning_rate": 2.2464977260551647e-06, "loss": 0.0537, "step": 25652 }, { "epoch": 7.89, "learning_rate": 2.2458684181024626e-06, "loss": 0.0525, "step": 25653 }, { "epoch": 7.89, "learning_rate": 2.245239187155435e-06, "loss": 0.0519, "step": 25654 }, { "epoch": 7.89, "learning_rate": 2.24461003322033e-06, "loss": 0.0707, "step": 25655 }, { "epoch": 7.89, "learning_rate": 2.2439809563033945e-06, "loss": 0.0393, "step": 25656 }, { "epoch": 7.89, "learning_rate": 2.2433519564108764e-06, "loss": 0.0477, "step": 25657 }, { "epoch": 7.89, "learning_rate": 2.2427230335490223e-06, "loss": 0.0345, "step": 25658 }, { "epoch": 7.89, "learning_rate": 2.2420941877240775e-06, "loss": 0.0304, "step": 25659 }, { "epoch": 7.9, "learning_rate": 2.2414654189422845e-06, "loss": 0.0372, "step": 25660 }, { "epoch": 7.9, "learning_rate": 2.2408367272098928e-06, "loss": 0.0479, "step": 25661 }, { "epoch": 7.9, "learning_rate": 2.2402081125331475e-06, "loss": 0.0661, "step": 25662 }, { "epoch": 7.9, "learning_rate": 2.2395795749182815e-06, "loss": 0.0802, "step": 25663 }, { "epoch": 7.9, "learning_rate": 2.2389511143715446e-06, "loss": 0.0285, "step": 25664 }, { "epoch": 7.9, "learning_rate": 2.238322730899176e-06, "loss": 0.0351, "step": 25665 }, { "epoch": 7.9, "learning_rate": 2.2376944245074162e-06, "loss": 0.0416, "step": 25666 }, { "epoch": 7.9, "learning_rate": 2.237066195202504e-06, "loss": 0.0488, "step": 25667 }, { "epoch": 7.9, "learning_rate": 2.2364380429906775e-06, "loss": 0.0747, "step": 25668 }, { "epoch": 7.9, "learning_rate": 2.235809967878181e-06, "loss": 0.0585, "step": 25669 }, { "epoch": 7.9, "learning_rate": 2.235181969871242e-06, "loss": 0.0289, "step": 25670 }, { "epoch": 7.9, "learning_rate": 2.2345540489761054e-06, "loss": 0.0603, "step": 25671 }, { "epoch": 7.9, "learning_rate": 2.233926205199004e-06, "loss": 0.0338, "step": 25672 }, { "epoch": 7.9, "learning_rate": 2.233298438546172e-06, "loss": 0.0376, "step": 25673 }, { "epoch": 7.9, "learning_rate": 2.232670749023845e-06, "loss": 0.0532, "step": 25674 }, { "epoch": 7.9, "learning_rate": 2.2320431366382556e-06, "loss": 0.0408, "step": 25675 }, { "epoch": 7.9, "learning_rate": 2.2314156013956377e-06, "loss": 0.049, "step": 25676 }, { "epoch": 7.9, "learning_rate": 2.230788143302223e-06, "loss": 0.0342, "step": 25677 }, { "epoch": 7.9, "learning_rate": 2.2301607623642386e-06, "loss": 0.0268, "step": 25678 }, { "epoch": 7.9, "learning_rate": 2.2295334585879235e-06, "loss": 0.0961, "step": 25679 }, { "epoch": 7.9, "learning_rate": 2.2289062319794985e-06, "loss": 0.059, "step": 25680 }, { "epoch": 7.9, "learning_rate": 2.2282790825451996e-06, "loss": 0.0479, "step": 25681 }, { "epoch": 7.9, "learning_rate": 2.2276520102912515e-06, "loss": 0.062, "step": 25682 }, { "epoch": 7.9, "learning_rate": 2.2270250152238816e-06, "loss": 0.0534, "step": 25683 }, { "epoch": 7.9, "learning_rate": 2.2263980973493183e-06, "loss": 0.0272, "step": 25684 }, { "epoch": 7.9, "learning_rate": 2.225771256673782e-06, "loss": 0.0505, "step": 25685 }, { "epoch": 7.9, "learning_rate": 2.2251444932035094e-06, "loss": 0.081, "step": 25686 }, { "epoch": 7.9, "learning_rate": 2.2245178069447137e-06, "loss": 0.0292, "step": 25687 }, { "epoch": 7.9, "learning_rate": 2.223891197903619e-06, "loss": 0.0366, "step": 25688 }, { "epoch": 7.9, "learning_rate": 2.2232646660864544e-06, "loss": 0.0655, "step": 25689 }, { "epoch": 7.9, "learning_rate": 2.2226382114994393e-06, "loss": 0.0982, "step": 25690 }, { "epoch": 7.9, "learning_rate": 2.222011834148795e-06, "loss": 0.0798, "step": 25691 }, { "epoch": 7.9, "learning_rate": 2.221385534040741e-06, "loss": 0.0336, "step": 25692 }, { "epoch": 7.91, "learning_rate": 2.2207593111814974e-06, "loss": 0.0464, "step": 25693 }, { "epoch": 7.91, "learning_rate": 2.220133165577284e-06, "loss": 0.0431, "step": 25694 }, { "epoch": 7.91, "learning_rate": 2.219507097234316e-06, "loss": 0.048, "step": 25695 }, { "epoch": 7.91, "learning_rate": 2.2188811061588177e-06, "loss": 0.034, "step": 25696 }, { "epoch": 7.91, "learning_rate": 2.2182551923569982e-06, "loss": 0.0948, "step": 25697 }, { "epoch": 7.91, "learning_rate": 2.2176293558350746e-06, "loss": 0.0528, "step": 25698 }, { "epoch": 7.91, "learning_rate": 2.2170035965992674e-06, "loss": 0.0506, "step": 25699 }, { "epoch": 7.91, "learning_rate": 2.2163779146557863e-06, "loss": 0.0759, "step": 25700 }, { "epoch": 7.91, "learning_rate": 2.2157523100108467e-06, "loss": 0.0766, "step": 25701 }, { "epoch": 7.91, "learning_rate": 2.21512678267066e-06, "loss": 0.0549, "step": 25702 }, { "epoch": 7.91, "learning_rate": 2.21450133264144e-06, "loss": 0.0566, "step": 25703 }, { "epoch": 7.91, "learning_rate": 2.213875959929397e-06, "loss": 0.0683, "step": 25704 }, { "epoch": 7.91, "learning_rate": 2.2132506645407393e-06, "loss": 0.0555, "step": 25705 }, { "epoch": 7.91, "learning_rate": 2.212625446481681e-06, "loss": 0.0516, "step": 25706 }, { "epoch": 7.91, "learning_rate": 2.2120003057584315e-06, "loss": 0.0573, "step": 25707 }, { "epoch": 7.91, "learning_rate": 2.2113752423771917e-06, "loss": 0.0367, "step": 25708 }, { "epoch": 7.91, "learning_rate": 2.2107502563441762e-06, "loss": 0.0327, "step": 25709 }, { "epoch": 7.91, "learning_rate": 2.2101253476655904e-06, "loss": 0.0794, "step": 25710 }, { "epoch": 7.91, "learning_rate": 2.209500516347639e-06, "loss": 0.0147, "step": 25711 }, { "epoch": 7.91, "learning_rate": 2.2088757623965263e-06, "loss": 0.0512, "step": 25712 }, { "epoch": 7.91, "learning_rate": 2.2082510858184558e-06, "loss": 0.0242, "step": 25713 }, { "epoch": 7.91, "learning_rate": 2.2076264866196393e-06, "loss": 0.0628, "step": 25714 }, { "epoch": 7.91, "learning_rate": 2.2070019648062667e-06, "loss": 0.0362, "step": 25715 }, { "epoch": 7.91, "learning_rate": 2.2063775203845504e-06, "loss": 0.0377, "step": 25716 }, { "epoch": 7.91, "learning_rate": 2.205753153360688e-06, "loss": 0.06, "step": 25717 }, { "epoch": 7.91, "learning_rate": 2.2051288637408794e-06, "loss": 0.0575, "step": 25718 }, { "epoch": 7.91, "learning_rate": 2.2045046515313263e-06, "loss": 0.0317, "step": 25719 }, { "epoch": 7.91, "learning_rate": 2.203880516738226e-06, "loss": 0.0485, "step": 25720 }, { "epoch": 7.91, "learning_rate": 2.2032564593677773e-06, "loss": 0.0836, "step": 25721 }, { "epoch": 7.91, "learning_rate": 2.2026324794261754e-06, "loss": 0.0363, "step": 25722 }, { "epoch": 7.91, "learning_rate": 2.202008576919622e-06, "loss": 0.0455, "step": 25723 }, { "epoch": 7.91, "learning_rate": 2.201384751854312e-06, "loss": 0.0395, "step": 25724 }, { "epoch": 7.92, "learning_rate": 2.2007610042364337e-06, "loss": 0.0544, "step": 25725 }, { "epoch": 7.92, "learning_rate": 2.2001373340721897e-06, "loss": 0.0341, "step": 25726 }, { "epoch": 7.92, "learning_rate": 2.1995137413677703e-06, "loss": 0.0401, "step": 25727 }, { "epoch": 7.92, "learning_rate": 2.1988902261293688e-06, "loss": 0.0666, "step": 25728 }, { "epoch": 7.92, "learning_rate": 2.198266788363177e-06, "loss": 0.0609, "step": 25729 }, { "epoch": 7.92, "learning_rate": 2.1976434280753833e-06, "loss": 0.0424, "step": 25730 }, { "epoch": 7.92, "learning_rate": 2.1970201452721873e-06, "loss": 0.0316, "step": 25731 }, { "epoch": 7.92, "learning_rate": 2.196396939959766e-06, "loss": 0.0257, "step": 25732 }, { "epoch": 7.92, "learning_rate": 2.1957738121443197e-06, "loss": 0.0279, "step": 25733 }, { "epoch": 7.92, "learning_rate": 2.1951507618320334e-06, "loss": 0.0691, "step": 25734 }, { "epoch": 7.92, "learning_rate": 2.194527789029088e-06, "loss": 0.0304, "step": 25735 }, { "epoch": 7.92, "learning_rate": 2.1939048937416783e-06, "loss": 0.0333, "step": 25736 }, { "epoch": 7.92, "learning_rate": 2.1932820759759875e-06, "loss": 0.0444, "step": 25737 }, { "epoch": 7.92, "learning_rate": 2.1926593357382e-06, "loss": 0.0628, "step": 25738 }, { "epoch": 7.92, "learning_rate": 2.1920366730345012e-06, "loss": 0.0467, "step": 25739 }, { "epoch": 7.92, "learning_rate": 2.191414087871071e-06, "loss": 0.1094, "step": 25740 }, { "epoch": 7.92, "learning_rate": 2.190791580254101e-06, "loss": 0.0337, "step": 25741 }, { "epoch": 7.92, "learning_rate": 2.190169150189763e-06, "loss": 0.0938, "step": 25742 }, { "epoch": 7.92, "learning_rate": 2.1895467976842457e-06, "loss": 0.0316, "step": 25743 }, { "epoch": 7.92, "learning_rate": 2.188924522743726e-06, "loss": 0.0397, "step": 25744 }, { "epoch": 7.92, "learning_rate": 2.188302325374386e-06, "loss": 0.0182, "step": 25745 }, { "epoch": 7.92, "learning_rate": 2.1876802055824032e-06, "loss": 0.0427, "step": 25746 }, { "epoch": 7.92, "learning_rate": 2.187058163373953e-06, "loss": 0.0314, "step": 25747 }, { "epoch": 7.92, "learning_rate": 2.1864361987552207e-06, "loss": 0.0632, "step": 25748 }, { "epoch": 7.92, "learning_rate": 2.185814311732376e-06, "loss": 0.0399, "step": 25749 }, { "epoch": 7.92, "learning_rate": 2.185192502311595e-06, "loss": 0.0302, "step": 25750 }, { "epoch": 7.92, "learning_rate": 2.184570770499056e-06, "loss": 0.0579, "step": 25751 }, { "epoch": 7.92, "learning_rate": 2.183949116300933e-06, "loss": 0.0758, "step": 25752 }, { "epoch": 7.92, "learning_rate": 2.1833275397233987e-06, "loss": 0.0448, "step": 25753 }, { "epoch": 7.92, "learning_rate": 2.1827060407726264e-06, "loss": 0.0587, "step": 25754 }, { "epoch": 7.92, "learning_rate": 2.1820846194547864e-06, "loss": 0.0739, "step": 25755 }, { "epoch": 7.92, "learning_rate": 2.1814632757760525e-06, "loss": 0.0706, "step": 25756 }, { "epoch": 7.92, "learning_rate": 2.180842009742591e-06, "loss": 0.0446, "step": 25757 }, { "epoch": 7.93, "learning_rate": 2.18022082136058e-06, "loss": 0.0219, "step": 25758 }, { "epoch": 7.93, "learning_rate": 2.179599710636181e-06, "loss": 0.0806, "step": 25759 }, { "epoch": 7.93, "learning_rate": 2.1789786775755604e-06, "loss": 0.0614, "step": 25760 }, { "epoch": 7.93, "learning_rate": 2.178357722184894e-06, "loss": 0.0478, "step": 25761 }, { "epoch": 7.93, "learning_rate": 2.1777368444703427e-06, "loss": 0.033, "step": 25762 }, { "epoch": 7.93, "learning_rate": 2.1771160444380745e-06, "loss": 0.05, "step": 25763 }, { "epoch": 7.93, "learning_rate": 2.176495322094254e-06, "loss": 0.0197, "step": 25764 }, { "epoch": 7.93, "learning_rate": 2.1758746774450444e-06, "loss": 0.0607, "step": 25765 }, { "epoch": 7.93, "learning_rate": 2.1752541104966107e-06, "loss": 0.0474, "step": 25766 }, { "epoch": 7.93, "learning_rate": 2.174633621255112e-06, "loss": 0.0231, "step": 25767 }, { "epoch": 7.93, "learning_rate": 2.1740132097267164e-06, "loss": 0.0244, "step": 25768 }, { "epoch": 7.93, "learning_rate": 2.173392875917586e-06, "loss": 0.0432, "step": 25769 }, { "epoch": 7.93, "learning_rate": 2.1727726198338705e-06, "loss": 0.0553, "step": 25770 }, { "epoch": 7.93, "learning_rate": 2.1721524414817406e-06, "loss": 0.0406, "step": 25771 }, { "epoch": 7.93, "learning_rate": 2.1715323408673506e-06, "loss": 0.0598, "step": 25772 }, { "epoch": 7.93, "learning_rate": 2.1709123179968607e-06, "loss": 0.0412, "step": 25773 }, { "epoch": 7.93, "learning_rate": 2.170292372876426e-06, "loss": 0.0599, "step": 25774 }, { "epoch": 7.93, "learning_rate": 2.1696725055122036e-06, "loss": 0.0899, "step": 25775 }, { "epoch": 7.93, "learning_rate": 2.16905271591035e-06, "loss": 0.0432, "step": 25776 }, { "epoch": 7.93, "learning_rate": 2.1684330040770183e-06, "loss": 0.05, "step": 25777 }, { "epoch": 7.93, "learning_rate": 2.167813370018367e-06, "loss": 0.0659, "step": 25778 }, { "epoch": 7.93, "learning_rate": 2.167193813740549e-06, "loss": 0.0466, "step": 25779 }, { "epoch": 7.93, "learning_rate": 2.1665743352497116e-06, "loss": 0.0437, "step": 25780 }, { "epoch": 7.93, "learning_rate": 2.1659549345520125e-06, "loss": 0.0383, "step": 25781 }, { "epoch": 7.93, "learning_rate": 2.1653356116536006e-06, "loss": 0.0573, "step": 25782 }, { "epoch": 7.93, "learning_rate": 2.164716366560626e-06, "loss": 0.0407, "step": 25783 }, { "epoch": 7.93, "learning_rate": 2.1640971992792416e-06, "loss": 0.0353, "step": 25784 }, { "epoch": 7.93, "learning_rate": 2.1634781098155888e-06, "loss": 0.0305, "step": 25785 }, { "epoch": 7.93, "learning_rate": 2.162859098175827e-06, "loss": 0.0258, "step": 25786 }, { "epoch": 7.93, "learning_rate": 2.1622401643660917e-06, "loss": 0.0494, "step": 25787 }, { "epoch": 7.93, "learning_rate": 2.1616213083925375e-06, "loss": 0.0439, "step": 25788 }, { "epoch": 7.93, "learning_rate": 2.161002530261309e-06, "loss": 0.0571, "step": 25789 }, { "epoch": 7.94, "learning_rate": 2.1603838299785486e-06, "loss": 0.0538, "step": 25790 }, { "epoch": 7.94, "learning_rate": 2.1597652075504026e-06, "loss": 0.0791, "step": 25791 }, { "epoch": 7.94, "learning_rate": 2.1591466629830138e-06, "loss": 0.0511, "step": 25792 }, { "epoch": 7.94, "learning_rate": 2.1585281962825257e-06, "loss": 0.038, "step": 25793 }, { "epoch": 7.94, "learning_rate": 2.1579098074550784e-06, "loss": 0.0442, "step": 25794 }, { "epoch": 7.94, "learning_rate": 2.1572914965068113e-06, "loss": 0.0598, "step": 25795 }, { "epoch": 7.94, "learning_rate": 2.1566732634438738e-06, "loss": 0.0802, "step": 25796 }, { "epoch": 7.94, "learning_rate": 2.1560551082723937e-06, "loss": 0.0445, "step": 25797 }, { "epoch": 7.94, "learning_rate": 2.1554370309985184e-06, "loss": 0.0499, "step": 25798 }, { "epoch": 7.94, "learning_rate": 2.1548190316283835e-06, "loss": 0.0254, "step": 25799 }, { "epoch": 7.94, "learning_rate": 2.154201110168126e-06, "loss": 0.0598, "step": 25800 }, { "epoch": 7.94, "learning_rate": 2.153583266623882e-06, "loss": 0.0375, "step": 25801 }, { "epoch": 7.94, "learning_rate": 2.1529655010017848e-06, "loss": 0.0473, "step": 25802 }, { "epoch": 7.94, "learning_rate": 2.1523478133079776e-06, "loss": 0.0366, "step": 25803 }, { "epoch": 7.94, "learning_rate": 2.1517302035485865e-06, "loss": 0.0313, "step": 25804 }, { "epoch": 7.94, "learning_rate": 2.1511126717297447e-06, "loss": 0.0532, "step": 25805 }, { "epoch": 7.94, "learning_rate": 2.150495217857591e-06, "loss": 0.071, "step": 25806 }, { "epoch": 7.94, "learning_rate": 2.149877841938255e-06, "loss": 0.0845, "step": 25807 }, { "epoch": 7.94, "learning_rate": 2.1492605439778657e-06, "loss": 0.0778, "step": 25808 }, { "epoch": 7.94, "learning_rate": 2.148643323982554e-06, "loss": 0.044, "step": 25809 }, { "epoch": 7.94, "learning_rate": 2.1480261819584525e-06, "loss": 0.0454, "step": 25810 }, { "epoch": 7.94, "learning_rate": 2.147409117911685e-06, "loss": 0.047, "step": 25811 }, { "epoch": 7.94, "learning_rate": 2.146792131848381e-06, "loss": 0.0559, "step": 25812 }, { "epoch": 7.94, "learning_rate": 2.1461752237746738e-06, "loss": 0.0313, "step": 25813 }, { "epoch": 7.94, "learning_rate": 2.145558393696683e-06, "loss": 0.0603, "step": 25814 }, { "epoch": 7.94, "learning_rate": 2.1449416416205327e-06, "loss": 0.0754, "step": 25815 }, { "epoch": 7.94, "learning_rate": 2.1443249675523536e-06, "loss": 0.089, "step": 25816 }, { "epoch": 7.94, "learning_rate": 2.1437083714982687e-06, "loss": 0.0446, "step": 25817 }, { "epoch": 7.94, "learning_rate": 2.1430918534643996e-06, "loss": 0.0676, "step": 25818 }, { "epoch": 7.94, "learning_rate": 2.142475413456869e-06, "loss": 0.0669, "step": 25819 }, { "epoch": 7.94, "learning_rate": 2.1418590514817995e-06, "loss": 0.0598, "step": 25820 }, { "epoch": 7.94, "learning_rate": 2.1412427675453128e-06, "loss": 0.029, "step": 25821 }, { "epoch": 7.94, "learning_rate": 2.140626561653525e-06, "loss": 0.074, "step": 25822 }, { "epoch": 7.95, "learning_rate": 2.1400104338125616e-06, "loss": 0.071, "step": 25823 }, { "epoch": 7.95, "learning_rate": 2.1393943840285414e-06, "loss": 0.0678, "step": 25824 }, { "epoch": 7.95, "learning_rate": 2.138778412307574e-06, "loss": 0.046, "step": 25825 }, { "epoch": 7.95, "learning_rate": 2.1381625186557842e-06, "loss": 0.0304, "step": 25826 }, { "epoch": 7.95, "learning_rate": 2.137546703079286e-06, "loss": 0.0407, "step": 25827 }, { "epoch": 7.95, "learning_rate": 2.1369309655841963e-06, "loss": 0.051, "step": 25828 }, { "epoch": 7.95, "learning_rate": 2.1363153061766297e-06, "loss": 0.0562, "step": 25829 }, { "epoch": 7.95, "learning_rate": 2.1356997248626954e-06, "loss": 0.0675, "step": 25830 }, { "epoch": 7.95, "learning_rate": 2.1350842216485167e-06, "loss": 0.0773, "step": 25831 }, { "epoch": 7.95, "learning_rate": 2.1344687965401944e-06, "loss": 0.0838, "step": 25832 }, { "epoch": 7.95, "learning_rate": 2.133853449543849e-06, "loss": 0.1276, "step": 25833 }, { "epoch": 7.95, "learning_rate": 2.133238180665591e-06, "loss": 0.0462, "step": 25834 }, { "epoch": 7.95, "learning_rate": 2.1326229899115235e-06, "loss": 0.0432, "step": 25835 }, { "epoch": 7.95, "learning_rate": 2.132007877287762e-06, "loss": 0.0319, "step": 25836 }, { "epoch": 7.95, "learning_rate": 2.131392842800415e-06, "loss": 0.0525, "step": 25837 }, { "epoch": 7.95, "learning_rate": 2.1307778864555873e-06, "loss": 0.0328, "step": 25838 }, { "epoch": 7.95, "learning_rate": 2.1301630082593895e-06, "loss": 0.0514, "step": 25839 }, { "epoch": 7.95, "learning_rate": 2.129548208217922e-06, "loss": 0.0312, "step": 25840 }, { "epoch": 7.95, "learning_rate": 2.1289334863373e-06, "loss": 0.037, "step": 25841 }, { "epoch": 7.95, "learning_rate": 2.128318842623618e-06, "loss": 0.0326, "step": 25842 }, { "epoch": 7.95, "learning_rate": 2.127704277082987e-06, "loss": 0.0534, "step": 25843 }, { "epoch": 7.95, "learning_rate": 2.127089789721507e-06, "loss": 0.0768, "step": 25844 }, { "epoch": 7.95, "learning_rate": 2.126475380545282e-06, "loss": 0.0389, "step": 25845 }, { "epoch": 7.95, "learning_rate": 2.125861049560414e-06, "loss": 0.0634, "step": 25846 }, { "epoch": 7.95, "learning_rate": 2.1252467967729984e-06, "loss": 0.0787, "step": 25847 }, { "epoch": 7.95, "learning_rate": 2.1246326221891458e-06, "loss": 0.0346, "step": 25848 }, { "epoch": 7.95, "learning_rate": 2.124018525814947e-06, "loss": 0.043, "step": 25849 }, { "epoch": 7.95, "learning_rate": 2.123404507656499e-06, "loss": 0.0603, "step": 25850 }, { "epoch": 7.95, "learning_rate": 2.1227905677199077e-06, "loss": 0.0257, "step": 25851 }, { "epoch": 7.95, "learning_rate": 2.1221767060112653e-06, "loss": 0.0465, "step": 25852 }, { "epoch": 7.95, "learning_rate": 2.1215629225366683e-06, "loss": 0.0868, "step": 25853 }, { "epoch": 7.95, "learning_rate": 2.120949217302213e-06, "loss": 0.0556, "step": 25854 }, { "epoch": 7.96, "learning_rate": 2.1203355903139934e-06, "loss": 0.0444, "step": 25855 }, { "epoch": 7.96, "learning_rate": 2.119722041578104e-06, "loss": 0.0477, "step": 25856 }, { "epoch": 7.96, "learning_rate": 2.1191085711006343e-06, "loss": 0.0414, "step": 25857 }, { "epoch": 7.96, "learning_rate": 2.1184951788876852e-06, "loss": 0.0642, "step": 25858 }, { "epoch": 7.96, "learning_rate": 2.117881864945339e-06, "loss": 0.0685, "step": 25859 }, { "epoch": 7.96, "learning_rate": 2.1172686292796886e-06, "loss": 0.0425, "step": 25860 }, { "epoch": 7.96, "learning_rate": 2.116655471896828e-06, "loss": 0.0555, "step": 25861 }, { "epoch": 7.96, "learning_rate": 2.116042392802845e-06, "loss": 0.0257, "step": 25862 }, { "epoch": 7.96, "learning_rate": 2.115429392003826e-06, "loss": 0.042, "step": 25863 }, { "epoch": 7.96, "learning_rate": 2.1148164695058603e-06, "loss": 0.0345, "step": 25864 }, { "epoch": 7.96, "learning_rate": 2.1142036253150344e-06, "loss": 0.0851, "step": 25865 }, { "epoch": 7.96, "learning_rate": 2.113590859437433e-06, "loss": 0.0739, "step": 25866 }, { "epoch": 7.96, "learning_rate": 2.112978171879141e-06, "loss": 0.0408, "step": 25867 }, { "epoch": 7.96, "learning_rate": 2.112365562646248e-06, "loss": 0.0388, "step": 25868 }, { "epoch": 7.96, "learning_rate": 2.1117530317448365e-06, "loss": 0.0511, "step": 25869 }, { "epoch": 7.96, "learning_rate": 2.1111405791809812e-06, "loss": 0.0507, "step": 25870 }, { "epoch": 7.96, "learning_rate": 2.1105282049607744e-06, "loss": 0.0253, "step": 25871 }, { "epoch": 7.96, "learning_rate": 2.109915909090293e-06, "loss": 0.0413, "step": 25872 }, { "epoch": 7.96, "learning_rate": 2.1093036915756193e-06, "loss": 0.0491, "step": 25873 }, { "epoch": 7.96, "learning_rate": 2.1086915524228313e-06, "loss": 0.074, "step": 25874 }, { "epoch": 7.96, "learning_rate": 2.1080794916380085e-06, "loss": 0.0378, "step": 25875 }, { "epoch": 7.96, "learning_rate": 2.1074675092272313e-06, "loss": 0.0496, "step": 25876 }, { "epoch": 7.96, "learning_rate": 2.1068556051965717e-06, "loss": 0.0358, "step": 25877 }, { "epoch": 7.96, "learning_rate": 2.106243779552114e-06, "loss": 0.0271, "step": 25878 }, { "epoch": 7.96, "learning_rate": 2.105632032299932e-06, "loss": 0.0823, "step": 25879 }, { "epoch": 7.96, "learning_rate": 2.1050203634460963e-06, "loss": 0.0381, "step": 25880 }, { "epoch": 7.96, "learning_rate": 2.1044087729966856e-06, "loss": 0.0224, "step": 25881 }, { "epoch": 7.96, "learning_rate": 2.103797260957774e-06, "loss": 0.0373, "step": 25882 }, { "epoch": 7.96, "learning_rate": 2.1031858273354322e-06, "loss": 0.0258, "step": 25883 }, { "epoch": 7.96, "learning_rate": 2.102574472135732e-06, "loss": 0.0546, "step": 25884 }, { "epoch": 7.96, "learning_rate": 2.1019631953647446e-06, "loss": 0.0354, "step": 25885 }, { "epoch": 7.96, "learning_rate": 2.101351997028547e-06, "loss": 0.0279, "step": 25886 }, { "epoch": 7.96, "learning_rate": 2.100740877133197e-06, "loss": 0.0244, "step": 25887 }, { "epoch": 7.97, "learning_rate": 2.1001298356847744e-06, "loss": 0.0524, "step": 25888 }, { "epoch": 7.97, "learning_rate": 2.0995188726893436e-06, "loss": 0.0356, "step": 25889 }, { "epoch": 7.97, "learning_rate": 2.0989079881529704e-06, "loss": 0.0327, "step": 25890 }, { "epoch": 7.97, "learning_rate": 2.0982971820817235e-06, "loss": 0.0736, "step": 25891 }, { "epoch": 7.97, "learning_rate": 2.0976864544816666e-06, "loss": 0.039, "step": 25892 }, { "epoch": 7.97, "learning_rate": 2.097075805358868e-06, "loss": 0.0668, "step": 25893 }, { "epoch": 7.97, "learning_rate": 2.0964652347193894e-06, "loss": 0.0584, "step": 25894 }, { "epoch": 7.97, "learning_rate": 2.095854742569292e-06, "loss": 0.0516, "step": 25895 }, { "epoch": 7.97, "learning_rate": 2.0952443289146475e-06, "loss": 0.0447, "step": 25896 }, { "epoch": 7.97, "learning_rate": 2.094633993761507e-06, "loss": 0.0339, "step": 25897 }, { "epoch": 7.97, "learning_rate": 2.0940237371159387e-06, "loss": 0.0339, "step": 25898 }, { "epoch": 7.97, "learning_rate": 2.093413558984002e-06, "loss": 0.046, "step": 25899 }, { "epoch": 7.97, "learning_rate": 2.092803459371755e-06, "loss": 0.0211, "step": 25900 }, { "epoch": 7.97, "learning_rate": 2.092193438285257e-06, "loss": 0.0694, "step": 25901 }, { "epoch": 7.97, "learning_rate": 2.0915834957305638e-06, "loss": 0.0599, "step": 25902 }, { "epoch": 7.97, "learning_rate": 2.0909736317137406e-06, "loss": 0.0503, "step": 25903 }, { "epoch": 7.97, "learning_rate": 2.090363846240836e-06, "loss": 0.0551, "step": 25904 }, { "epoch": 7.97, "learning_rate": 2.0897541393179042e-06, "loss": 0.0601, "step": 25905 }, { "epoch": 7.97, "learning_rate": 2.089144510951008e-06, "loss": 0.0508, "step": 25906 }, { "epoch": 7.97, "learning_rate": 2.088534961146197e-06, "loss": 0.0453, "step": 25907 }, { "epoch": 7.97, "learning_rate": 2.0879254899095257e-06, "loss": 0.0364, "step": 25908 }, { "epoch": 7.97, "learning_rate": 2.0873160972470462e-06, "loss": 0.0577, "step": 25909 }, { "epoch": 7.97, "learning_rate": 2.08670678316481e-06, "loss": 0.0448, "step": 25910 }, { "epoch": 7.97, "learning_rate": 2.0860975476688692e-06, "loss": 0.0707, "step": 25911 }, { "epoch": 7.97, "learning_rate": 2.08548839076527e-06, "loss": 0.0343, "step": 25912 }, { "epoch": 7.97, "learning_rate": 2.0848793124600707e-06, "loss": 0.0518, "step": 25913 }, { "epoch": 7.97, "learning_rate": 2.08427031275931e-06, "loss": 0.0826, "step": 25914 }, { "epoch": 7.97, "learning_rate": 2.083661391669043e-06, "loss": 0.043, "step": 25915 }, { "epoch": 7.97, "learning_rate": 2.0830525491953137e-06, "loss": 0.0223, "step": 25916 }, { "epoch": 7.97, "learning_rate": 2.082443785344169e-06, "loss": 0.037, "step": 25917 }, { "epoch": 7.97, "learning_rate": 2.081835100121655e-06, "loss": 0.0418, "step": 25918 }, { "epoch": 7.97, "learning_rate": 2.0812264935338132e-06, "loss": 0.0679, "step": 25919 }, { "epoch": 7.98, "learning_rate": 2.0806179655866964e-06, "loss": 0.0633, "step": 25920 }, { "epoch": 7.98, "learning_rate": 2.080009516286339e-06, "loss": 0.0252, "step": 25921 }, { "epoch": 7.98, "learning_rate": 2.079401145638783e-06, "loss": 0.0318, "step": 25922 }, { "epoch": 7.98, "learning_rate": 2.0787928536500767e-06, "loss": 0.0439, "step": 25923 }, { "epoch": 7.98, "learning_rate": 2.0781846403262572e-06, "loss": 0.0333, "step": 25924 }, { "epoch": 7.98, "learning_rate": 2.0775765056733667e-06, "loss": 0.0336, "step": 25925 }, { "epoch": 7.98, "learning_rate": 2.076968449697442e-06, "loss": 0.0412, "step": 25926 }, { "epoch": 7.98, "learning_rate": 2.0763604724045217e-06, "loss": 0.0604, "step": 25927 }, { "epoch": 7.98, "learning_rate": 2.0757525738006455e-06, "loss": 0.1003, "step": 25928 }, { "epoch": 7.98, "learning_rate": 2.0751447538918468e-06, "loss": 0.0761, "step": 25929 }, { "epoch": 7.98, "learning_rate": 2.074537012684168e-06, "loss": 0.0372, "step": 25930 }, { "epoch": 7.98, "learning_rate": 2.0739293501836424e-06, "loss": 0.0656, "step": 25931 }, { "epoch": 7.98, "learning_rate": 2.073321766396298e-06, "loss": 0.046, "step": 25932 }, { "epoch": 7.98, "learning_rate": 2.072714261328177e-06, "loss": 0.0394, "step": 25933 }, { "epoch": 7.98, "learning_rate": 2.072106834985309e-06, "loss": 0.0433, "step": 25934 }, { "epoch": 7.98, "learning_rate": 2.071499487373727e-06, "loss": 0.0466, "step": 25935 }, { "epoch": 7.98, "learning_rate": 2.0708922184994617e-06, "loss": 0.0284, "step": 25936 }, { "epoch": 7.98, "learning_rate": 2.070285028368545e-06, "loss": 0.0756, "step": 25937 }, { "epoch": 7.98, "learning_rate": 2.069677916987006e-06, "loss": 0.0745, "step": 25938 }, { "epoch": 7.98, "learning_rate": 2.069070884360872e-06, "loss": 0.0476, "step": 25939 }, { "epoch": 7.98, "learning_rate": 2.068463930496175e-06, "loss": 0.0581, "step": 25940 }, { "epoch": 7.98, "learning_rate": 2.067857055398944e-06, "loss": 0.0414, "step": 25941 }, { "epoch": 7.98, "learning_rate": 2.0672502590751985e-06, "loss": 0.0671, "step": 25942 }, { "epoch": 7.98, "learning_rate": 2.0666435415309703e-06, "loss": 0.0272, "step": 25943 }, { "epoch": 7.98, "learning_rate": 2.066036902772285e-06, "loss": 0.0433, "step": 25944 }, { "epoch": 7.98, "learning_rate": 2.0654303428051636e-06, "loss": 0.0373, "step": 25945 }, { "epoch": 7.98, "learning_rate": 2.064823861635633e-06, "loss": 0.0267, "step": 25946 }, { "epoch": 7.98, "learning_rate": 2.0642174592697105e-06, "loss": 0.0211, "step": 25947 }, { "epoch": 7.98, "learning_rate": 2.063611135713429e-06, "loss": 0.0271, "step": 25948 }, { "epoch": 7.98, "learning_rate": 2.0630048909727984e-06, "loss": 0.0429, "step": 25949 }, { "epoch": 7.98, "learning_rate": 2.0623987250538458e-06, "loss": 0.0544, "step": 25950 }, { "epoch": 7.98, "learning_rate": 2.0617926379625896e-06, "loss": 0.0312, "step": 25951 }, { "epoch": 7.98, "learning_rate": 2.06118662970505e-06, "loss": 0.102, "step": 25952 }, { "epoch": 7.99, "learning_rate": 2.0605807002872413e-06, "loss": 0.0246, "step": 25953 }, { "epoch": 7.99, "learning_rate": 2.059974849715185e-06, "loss": 0.0517, "step": 25954 }, { "epoch": 7.99, "learning_rate": 2.059369077994896e-06, "loss": 0.0493, "step": 25955 }, { "epoch": 7.99, "learning_rate": 2.0587633851323906e-06, "loss": 0.0378, "step": 25956 }, { "epoch": 7.99, "learning_rate": 2.0581577711336807e-06, "loss": 0.0332, "step": 25957 }, { "epoch": 7.99, "learning_rate": 2.0575522360047882e-06, "loss": 0.0557, "step": 25958 }, { "epoch": 7.99, "learning_rate": 2.0569467797517173e-06, "loss": 0.0537, "step": 25959 }, { "epoch": 7.99, "learning_rate": 2.056341402380488e-06, "loss": 0.0529, "step": 25960 }, { "epoch": 7.99, "learning_rate": 2.0557361038971102e-06, "loss": 0.0952, "step": 25961 }, { "epoch": 7.99, "learning_rate": 2.0551308843075935e-06, "loss": 0.0502, "step": 25962 }, { "epoch": 7.99, "learning_rate": 2.05452574361795e-06, "loss": 0.0757, "step": 25963 }, { "epoch": 7.99, "learning_rate": 2.0539206818341863e-06, "loss": 0.0667, "step": 25964 }, { "epoch": 7.99, "learning_rate": 2.0533156989623182e-06, "loss": 0.0663, "step": 25965 }, { "epoch": 7.99, "learning_rate": 2.052710795008347e-06, "loss": 0.0372, "step": 25966 }, { "epoch": 7.99, "learning_rate": 2.0521059699782788e-06, "loss": 0.0375, "step": 25967 }, { "epoch": 7.99, "learning_rate": 2.0515012238781263e-06, "loss": 0.035, "step": 25968 }, { "epoch": 7.99, "learning_rate": 2.0508965567138916e-06, "loss": 0.0317, "step": 25969 }, { "epoch": 7.99, "learning_rate": 2.05029196849158e-06, "loss": 0.0616, "step": 25970 }, { "epoch": 7.99, "learning_rate": 2.0496874592171966e-06, "loss": 0.0451, "step": 25971 }, { "epoch": 7.99, "learning_rate": 2.0490830288967443e-06, "loss": 0.0632, "step": 25972 }, { "epoch": 7.99, "learning_rate": 2.0484786775362252e-06, "loss": 0.0366, "step": 25973 }, { "epoch": 7.99, "learning_rate": 2.0478744051416376e-06, "loss": 0.0544, "step": 25974 }, { "epoch": 7.99, "learning_rate": 2.0472702117189924e-06, "loss": 0.0463, "step": 25975 }, { "epoch": 7.99, "learning_rate": 2.0466660972742803e-06, "loss": 0.0647, "step": 25976 }, { "epoch": 7.99, "learning_rate": 2.0460620618135007e-06, "loss": 0.0685, "step": 25977 }, { "epoch": 7.99, "learning_rate": 2.0454581053426592e-06, "loss": 0.0779, "step": 25978 }, { "epoch": 7.99, "learning_rate": 2.0448542278677486e-06, "loss": 0.0737, "step": 25979 }, { "epoch": 7.99, "learning_rate": 2.0442504293947683e-06, "loss": 0.0585, "step": 25980 }, { "epoch": 7.99, "learning_rate": 2.043646709929713e-06, "loss": 0.0354, "step": 25981 }, { "epoch": 7.99, "learning_rate": 2.0430430694785787e-06, "loss": 0.0421, "step": 25982 }, { "epoch": 7.99, "learning_rate": 2.0424395080473604e-06, "loss": 0.0484, "step": 25983 }, { "epoch": 7.99, "learning_rate": 2.041836025642048e-06, "loss": 0.0533, "step": 25984 }, { "epoch": 8.0, "learning_rate": 2.041232622268642e-06, "loss": 0.0297, "step": 25985 }, { "epoch": 8.0, "learning_rate": 2.0406292979331333e-06, "loss": 0.0385, "step": 25986 }, { "epoch": 8.0, "learning_rate": 2.040026052641506e-06, "loss": 0.0614, "step": 25987 }, { "epoch": 8.0, "learning_rate": 2.039422886399759e-06, "loss": 0.0341, "step": 25988 }, { "epoch": 8.0, "learning_rate": 2.038819799213878e-06, "loss": 0.0281, "step": 25989 }, { "epoch": 8.0, "learning_rate": 2.0382167910898552e-06, "loss": 0.0617, "step": 25990 }, { "epoch": 8.0, "learning_rate": 2.037613862033677e-06, "loss": 0.037, "step": 25991 }, { "epoch": 8.0, "learning_rate": 2.037011012051332e-06, "loss": 0.054, "step": 25992 }, { "epoch": 8.0, "learning_rate": 2.036408241148806e-06, "loss": 0.0717, "step": 25993 }, { "epoch": 8.0, "learning_rate": 2.035805549332083e-06, "loss": 0.0438, "step": 25994 }, { "epoch": 8.0, "learning_rate": 2.0352029366071537e-06, "loss": 0.0248, "step": 25995 }, { "epoch": 8.0, "learning_rate": 2.0346004029800027e-06, "loss": 0.0418, "step": 25996 }, { "epoch": 8.0, "learning_rate": 2.033997948456605e-06, "loss": 0.0603, "step": 25997 }, { "epoch": 8.0, "learning_rate": 2.033395573042952e-06, "loss": 0.0574, "step": 25998 }, { "epoch": 8.0, "learning_rate": 2.0327932767450244e-06, "loss": 0.0555, "step": 25999 }, { "epoch": 8.0, "learning_rate": 2.032191059568801e-06, "loss": 0.0517, "step": 26000 }, { "epoch": 8.0, "learning_rate": 2.0315889215202643e-06, "loss": 0.0547, "step": 26001 }, { "epoch": 8.0, "learning_rate": 2.0309868626053906e-06, "loss": 0.0225, "step": 26002 }, { "epoch": 8.0, "learning_rate": 2.0303848828301678e-06, "loss": 0.0221, "step": 26003 }, { "epoch": 8.0, "learning_rate": 2.0297829822005612e-06, "loss": 0.0267, "step": 26004 }, { "epoch": 8.0, "learning_rate": 2.0291811607225596e-06, "loss": 0.046, "step": 26005 }, { "epoch": 8.0, "learning_rate": 2.0285794184021344e-06, "loss": 0.0485, "step": 26006 }, { "epoch": 8.0, "learning_rate": 2.0279777552452616e-06, "loss": 0.036, "step": 26007 }, { "epoch": 8.0, "learning_rate": 2.0273761712579176e-06, "loss": 0.0225, "step": 26008 }, { "epoch": 8.0, "learning_rate": 2.0267746664460743e-06, "loss": 0.0522, "step": 26009 }, { "epoch": 8.0, "learning_rate": 2.026173240815711e-06, "loss": 0.0292, "step": 26010 }, { "epoch": 8.0, "learning_rate": 2.025571894372794e-06, "loss": 0.0137, "step": 26011 }, { "epoch": 8.0, "learning_rate": 2.024970627123295e-06, "loss": 0.0318, "step": 26012 }, { "epoch": 8.0, "learning_rate": 2.024369439073193e-06, "loss": 0.026, "step": 26013 }, { "epoch": 8.0, "learning_rate": 2.023768330228447e-06, "loss": 0.0469, "step": 26014 }, { "epoch": 8.0, "learning_rate": 2.023167300595036e-06, "loss": 0.0155, "step": 26015 }, { "epoch": 8.0, "learning_rate": 2.022566350178925e-06, "loss": 0.031, "step": 26016 }, { "epoch": 8.0, "learning_rate": 2.021965478986082e-06, "loss": 0.0224, "step": 26017 }, { "epoch": 8.01, "learning_rate": 2.0213646870224758e-06, "loss": 0.0233, "step": 26018 }, { "epoch": 8.01, "learning_rate": 2.0207639742940676e-06, "loss": 0.0601, "step": 26019 }, { "epoch": 8.01, "learning_rate": 2.0201633408068333e-06, "loss": 0.0228, "step": 26020 }, { "epoch": 8.01, "learning_rate": 2.019562786566729e-06, "loss": 0.0382, "step": 26021 }, { "epoch": 8.01, "learning_rate": 2.0189623115797176e-06, "loss": 0.0394, "step": 26022 }, { "epoch": 8.01, "learning_rate": 2.0183619158517687e-06, "loss": 0.044, "step": 26023 }, { "epoch": 8.01, "learning_rate": 2.017761599388842e-06, "loss": 0.0281, "step": 26024 }, { "epoch": 8.01, "learning_rate": 2.0171613621968998e-06, "loss": 0.0143, "step": 26025 }, { "epoch": 8.01, "learning_rate": 2.016561204281903e-06, "loss": 0.0234, "step": 26026 }, { "epoch": 8.01, "learning_rate": 2.01596112564981e-06, "loss": 0.0475, "step": 26027 }, { "epoch": 8.01, "learning_rate": 2.015361126306582e-06, "loss": 0.0288, "step": 26028 }, { "epoch": 8.01, "learning_rate": 2.0147612062581746e-06, "loss": 0.0342, "step": 26029 }, { "epoch": 8.01, "learning_rate": 2.014161365510553e-06, "loss": 0.0261, "step": 26030 }, { "epoch": 8.01, "learning_rate": 2.013561604069666e-06, "loss": 0.0199, "step": 26031 }, { "epoch": 8.01, "learning_rate": 2.01296192194147e-06, "loss": 0.0238, "step": 26032 }, { "epoch": 8.01, "learning_rate": 2.012362319131925e-06, "loss": 0.0288, "step": 26033 }, { "epoch": 8.01, "learning_rate": 2.0117627956469856e-06, "loss": 0.0332, "step": 26034 }, { "epoch": 8.01, "learning_rate": 2.011163351492603e-06, "loss": 0.027, "step": 26035 }, { "epoch": 8.01, "learning_rate": 2.0105639866747306e-06, "loss": 0.0307, "step": 26036 }, { "epoch": 8.01, "learning_rate": 2.0099647011993217e-06, "loss": 0.0344, "step": 26037 }, { "epoch": 8.01, "learning_rate": 2.0093654950723274e-06, "loss": 0.0345, "step": 26038 }, { "epoch": 8.01, "learning_rate": 2.0087663682996948e-06, "loss": 0.0186, "step": 26039 }, { "epoch": 8.01, "learning_rate": 2.00816732088738e-06, "loss": 0.0233, "step": 26040 }, { "epoch": 8.01, "learning_rate": 2.0075683528413316e-06, "loss": 0.0376, "step": 26041 }, { "epoch": 8.01, "learning_rate": 2.006969464167491e-06, "loss": 0.0448, "step": 26042 }, { "epoch": 8.01, "learning_rate": 2.006370654871812e-06, "loss": 0.0482, "step": 26043 }, { "epoch": 8.01, "learning_rate": 2.0057719249602403e-06, "loss": 0.0286, "step": 26044 }, { "epoch": 8.01, "learning_rate": 2.00517327443872e-06, "loss": 0.0322, "step": 26045 }, { "epoch": 8.01, "learning_rate": 2.004574703313198e-06, "loss": 0.0429, "step": 26046 }, { "epoch": 8.01, "learning_rate": 2.0039762115896154e-06, "loss": 0.0423, "step": 26047 }, { "epoch": 8.01, "learning_rate": 2.0033777992739235e-06, "loss": 0.0362, "step": 26048 }, { "epoch": 8.01, "learning_rate": 2.0027794663720545e-06, "loss": 0.0283, "step": 26049 }, { "epoch": 8.02, "learning_rate": 2.00218121288996e-06, "loss": 0.0301, "step": 26050 }, { "epoch": 8.02, "learning_rate": 2.001583038833578e-06, "loss": 0.0344, "step": 26051 }, { "epoch": 8.02, "learning_rate": 2.0009849442088423e-06, "loss": 0.0333, "step": 26052 }, { "epoch": 8.02, "learning_rate": 2.0003869290217025e-06, "loss": 0.0158, "step": 26053 }, { "epoch": 8.02, "learning_rate": 1.9997889932780924e-06, "loss": 0.0353, "step": 26054 }, { "epoch": 8.02, "learning_rate": 1.99919113698395e-06, "loss": 0.0145, "step": 26055 }, { "epoch": 8.02, "learning_rate": 1.9985933601452143e-06, "loss": 0.0408, "step": 26056 }, { "epoch": 8.02, "learning_rate": 1.9979956627678188e-06, "loss": 0.0293, "step": 26057 }, { "epoch": 8.02, "learning_rate": 1.997398044857706e-06, "loss": 0.0211, "step": 26058 }, { "epoch": 8.02, "learning_rate": 1.9968005064207996e-06, "loss": 0.036, "step": 26059 }, { "epoch": 8.02, "learning_rate": 1.9962030474630443e-06, "loss": 0.0423, "step": 26060 }, { "epoch": 8.02, "learning_rate": 1.9956056679903678e-06, "loss": 0.0311, "step": 26061 }, { "epoch": 8.02, "learning_rate": 1.995008368008705e-06, "loss": 0.0426, "step": 26062 }, { "epoch": 8.02, "learning_rate": 1.994411147523987e-06, "loss": 0.0369, "step": 26063 }, { "epoch": 8.02, "learning_rate": 1.993814006542142e-06, "loss": 0.0391, "step": 26064 }, { "epoch": 8.02, "learning_rate": 1.9932169450691075e-06, "loss": 0.0203, "step": 26065 }, { "epoch": 8.02, "learning_rate": 1.9926199631108045e-06, "loss": 0.0228, "step": 26066 }, { "epoch": 8.02, "learning_rate": 1.9920230606731637e-06, "loss": 0.0266, "step": 26067 }, { "epoch": 8.02, "learning_rate": 1.991426237762116e-06, "loss": 0.0111, "step": 26068 }, { "epoch": 8.02, "learning_rate": 1.990829494383587e-06, "loss": 0.0327, "step": 26069 }, { "epoch": 8.02, "learning_rate": 1.9902328305435024e-06, "loss": 0.0264, "step": 26070 }, { "epoch": 8.02, "learning_rate": 1.9896362462477893e-06, "loss": 0.0329, "step": 26071 }, { "epoch": 8.02, "learning_rate": 1.989039741502369e-06, "loss": 0.0346, "step": 26072 }, { "epoch": 8.02, "learning_rate": 1.988443316313168e-06, "loss": 0.0178, "step": 26073 }, { "epoch": 8.02, "learning_rate": 1.9878469706861058e-06, "loss": 0.0364, "step": 26074 }, { "epoch": 8.02, "learning_rate": 1.9872507046271126e-06, "loss": 0.0174, "step": 26075 }, { "epoch": 8.02, "learning_rate": 1.9866545181421016e-06, "loss": 0.0272, "step": 26076 }, { "epoch": 8.02, "learning_rate": 1.986058411236995e-06, "loss": 0.0395, "step": 26077 }, { "epoch": 8.02, "learning_rate": 1.9854623839177156e-06, "loss": 0.0211, "step": 26078 }, { "epoch": 8.02, "learning_rate": 1.9848664361901827e-06, "loss": 0.0247, "step": 26079 }, { "epoch": 8.02, "learning_rate": 1.9842705680603115e-06, "loss": 0.0319, "step": 26080 }, { "epoch": 8.02, "learning_rate": 1.9836747795340215e-06, "loss": 0.048, "step": 26081 }, { "epoch": 8.02, "learning_rate": 1.9830790706172297e-06, "loss": 0.0293, "step": 26082 }, { "epoch": 8.03, "learning_rate": 1.982483441315851e-06, "loss": 0.0204, "step": 26083 }, { "epoch": 8.03, "learning_rate": 1.9818878916357976e-06, "loss": 0.0599, "step": 26084 }, { "epoch": 8.03, "learning_rate": 1.981292421582991e-06, "loss": 0.0336, "step": 26085 }, { "epoch": 8.03, "learning_rate": 1.980697031163342e-06, "loss": 0.0339, "step": 26086 }, { "epoch": 8.03, "learning_rate": 1.980101720382758e-06, "loss": 0.0316, "step": 26087 }, { "epoch": 8.03, "learning_rate": 1.9795064892471583e-06, "loss": 0.0369, "step": 26088 }, { "epoch": 8.03, "learning_rate": 1.97891133776245e-06, "loss": 0.0182, "step": 26089 }, { "epoch": 8.03, "learning_rate": 1.9783162659345456e-06, "loss": 0.0313, "step": 26090 }, { "epoch": 8.03, "learning_rate": 1.977721273769353e-06, "loss": 0.0277, "step": 26091 }, { "epoch": 8.03, "learning_rate": 1.977126361272782e-06, "loss": 0.0245, "step": 26092 }, { "epoch": 8.03, "learning_rate": 1.97653152845074e-06, "loss": 0.0126, "step": 26093 }, { "epoch": 8.03, "learning_rate": 1.975936775309132e-06, "loss": 0.0506, "step": 26094 }, { "epoch": 8.03, "learning_rate": 1.97534210185387e-06, "loss": 0.0337, "step": 26095 }, { "epoch": 8.03, "learning_rate": 1.9747475080908586e-06, "loss": 0.0459, "step": 26096 }, { "epoch": 8.03, "learning_rate": 1.9741529940259963e-06, "loss": 0.0244, "step": 26097 }, { "epoch": 8.03, "learning_rate": 1.973558559665194e-06, "loss": 0.04, "step": 26098 }, { "epoch": 8.03, "learning_rate": 1.9729642050143526e-06, "loss": 0.0332, "step": 26099 }, { "epoch": 8.03, "learning_rate": 1.972369930079374e-06, "loss": 0.022, "step": 26100 }, { "epoch": 8.03, "learning_rate": 1.9717757348661617e-06, "loss": 0.0307, "step": 26101 }, { "epoch": 8.03, "learning_rate": 1.971181619380611e-06, "loss": 0.0423, "step": 26102 }, { "epoch": 8.03, "learning_rate": 1.9705875836286337e-06, "loss": 0.0201, "step": 26103 }, { "epoch": 8.03, "learning_rate": 1.969993627616115e-06, "loss": 0.0379, "step": 26104 }, { "epoch": 8.03, "learning_rate": 1.969399751348964e-06, "loss": 0.0375, "step": 26105 }, { "epoch": 8.03, "learning_rate": 1.968805954833073e-06, "loss": 0.0203, "step": 26106 }, { "epoch": 8.03, "learning_rate": 1.9682122380743417e-06, "loss": 0.0319, "step": 26107 }, { "epoch": 8.03, "learning_rate": 1.9676186010786657e-06, "loss": 0.0398, "step": 26108 }, { "epoch": 8.03, "learning_rate": 1.967025043851939e-06, "loss": 0.0374, "step": 26109 }, { "epoch": 8.03, "learning_rate": 1.9664315664000578e-06, "loss": 0.0313, "step": 26110 }, { "epoch": 8.03, "learning_rate": 1.965838168728912e-06, "loss": 0.0448, "step": 26111 }, { "epoch": 8.03, "learning_rate": 1.9652448508444e-06, "loss": 0.0201, "step": 26112 }, { "epoch": 8.03, "learning_rate": 1.964651612752415e-06, "loss": 0.0317, "step": 26113 }, { "epoch": 8.03, "learning_rate": 1.9640584544588393e-06, "loss": 0.0188, "step": 26114 }, { "epoch": 8.04, "learning_rate": 1.963465375969572e-06, "loss": 0.0113, "step": 26115 }, { "epoch": 8.04, "learning_rate": 1.9628723772905e-06, "loss": 0.0302, "step": 26116 }, { "epoch": 8.04, "learning_rate": 1.962279458427514e-06, "loss": 0.0454, "step": 26117 }, { "epoch": 8.04, "learning_rate": 1.961686619386499e-06, "loss": 0.0255, "step": 26118 }, { "epoch": 8.04, "learning_rate": 1.961093860173342e-06, "loss": 0.0151, "step": 26119 }, { "epoch": 8.04, "learning_rate": 1.9605011807939377e-06, "loss": 0.0229, "step": 26120 }, { "epoch": 8.04, "learning_rate": 1.959908581254161e-06, "loss": 0.0121, "step": 26121 }, { "epoch": 8.04, "learning_rate": 1.9593160615599038e-06, "loss": 0.0242, "step": 26122 }, { "epoch": 8.04, "learning_rate": 1.958723621717048e-06, "loss": 0.0418, "step": 26123 }, { "epoch": 8.04, "learning_rate": 1.9581312617314796e-06, "loss": 0.0207, "step": 26124 }, { "epoch": 8.04, "learning_rate": 1.9575389816090774e-06, "loss": 0.0338, "step": 26125 }, { "epoch": 8.04, "learning_rate": 1.9569467813557242e-06, "loss": 0.0281, "step": 26126 }, { "epoch": 8.04, "learning_rate": 1.9563546609773065e-06, "loss": 0.0185, "step": 26127 }, { "epoch": 8.04, "learning_rate": 1.955762620479699e-06, "loss": 0.0379, "step": 26128 }, { "epoch": 8.04, "learning_rate": 1.955170659868778e-06, "loss": 0.0297, "step": 26129 }, { "epoch": 8.04, "learning_rate": 1.9545787791504323e-06, "loss": 0.0199, "step": 26130 }, { "epoch": 8.04, "learning_rate": 1.95398697833053e-06, "loss": 0.0227, "step": 26131 }, { "epoch": 8.04, "learning_rate": 1.953395257414954e-06, "loss": 0.0406, "step": 26132 }, { "epoch": 8.04, "learning_rate": 1.95280361640958e-06, "loss": 0.0284, "step": 26133 }, { "epoch": 8.04, "learning_rate": 1.9522120553202817e-06, "loss": 0.0306, "step": 26134 }, { "epoch": 8.04, "learning_rate": 1.9516205741529347e-06, "loss": 0.0429, "step": 26135 }, { "epoch": 8.04, "learning_rate": 1.951029172913411e-06, "loss": 0.0309, "step": 26136 }, { "epoch": 8.04, "learning_rate": 1.9504378516075917e-06, "loss": 0.0147, "step": 26137 }, { "epoch": 8.04, "learning_rate": 1.9498466102413394e-06, "loss": 0.0294, "step": 26138 }, { "epoch": 8.04, "learning_rate": 1.9492554488205275e-06, "loss": 0.0461, "step": 26139 }, { "epoch": 8.04, "learning_rate": 1.948664367351031e-06, "loss": 0.0368, "step": 26140 }, { "epoch": 8.04, "learning_rate": 1.9480733658387175e-06, "loss": 0.0222, "step": 26141 }, { "epoch": 8.04, "learning_rate": 1.947482444289457e-06, "loss": 0.0248, "step": 26142 }, { "epoch": 8.04, "learning_rate": 1.9468916027091167e-06, "loss": 0.0248, "step": 26143 }, { "epoch": 8.04, "learning_rate": 1.9463008411035645e-06, "loss": 0.0183, "step": 26144 }, { "epoch": 8.04, "learning_rate": 1.945710159478669e-06, "loss": 0.0294, "step": 26145 }, { "epoch": 8.04, "learning_rate": 1.9451195578402905e-06, "loss": 0.0315, "step": 26146 }, { "epoch": 8.04, "learning_rate": 1.9445290361943014e-06, "loss": 0.0574, "step": 26147 }, { "epoch": 8.05, "learning_rate": 1.943938594546566e-06, "loss": 0.0262, "step": 26148 }, { "epoch": 8.05, "learning_rate": 1.94334823290294e-06, "loss": 0.0292, "step": 26149 }, { "epoch": 8.05, "learning_rate": 1.942757951269293e-06, "loss": 0.0162, "step": 26150 }, { "epoch": 8.05, "learning_rate": 1.9421677496514857e-06, "loss": 0.0244, "step": 26151 }, { "epoch": 8.05, "learning_rate": 1.941577628055379e-06, "loss": 0.0127, "step": 26152 }, { "epoch": 8.05, "learning_rate": 1.9409875864868332e-06, "loss": 0.0389, "step": 26153 }, { "epoch": 8.05, "learning_rate": 1.940397624951709e-06, "loss": 0.0263, "step": 26154 }, { "epoch": 8.05, "learning_rate": 1.939807743455863e-06, "loss": 0.0308, "step": 26155 }, { "epoch": 8.05, "learning_rate": 1.9392179420051526e-06, "loss": 0.0382, "step": 26156 }, { "epoch": 8.05, "learning_rate": 1.9386282206054395e-06, "loss": 0.024, "step": 26157 }, { "epoch": 8.05, "learning_rate": 1.93803857926258e-06, "loss": 0.0279, "step": 26158 }, { "epoch": 8.05, "learning_rate": 1.9374490179824237e-06, "loss": 0.0327, "step": 26159 }, { "epoch": 8.05, "learning_rate": 1.936859536770831e-06, "loss": 0.0229, "step": 26160 }, { "epoch": 8.05, "learning_rate": 1.9362701356336545e-06, "loss": 0.0279, "step": 26161 }, { "epoch": 8.05, "learning_rate": 1.935680814576747e-06, "loss": 0.0496, "step": 26162 }, { "epoch": 8.05, "learning_rate": 1.9350915736059616e-06, "loss": 0.0346, "step": 26163 }, { "epoch": 8.05, "learning_rate": 1.934502412727146e-06, "loss": 0.024, "step": 26164 }, { "epoch": 8.05, "learning_rate": 1.933913331946161e-06, "loss": 0.0479, "step": 26165 }, { "epoch": 8.05, "learning_rate": 1.9333243312688453e-06, "loss": 0.0285, "step": 26166 }, { "epoch": 8.05, "learning_rate": 1.9327354107010566e-06, "loss": 0.0599, "step": 26167 }, { "epoch": 8.05, "learning_rate": 1.9321465702486396e-06, "loss": 0.0317, "step": 26168 }, { "epoch": 8.05, "learning_rate": 1.931557809917444e-06, "loss": 0.0348, "step": 26169 }, { "epoch": 8.05, "learning_rate": 1.930969129713314e-06, "loss": 0.0354, "step": 26170 }, { "epoch": 8.05, "learning_rate": 1.9303805296420973e-06, "loss": 0.0537, "step": 26171 }, { "epoch": 8.05, "learning_rate": 1.929792009709639e-06, "loss": 0.0268, "step": 26172 }, { "epoch": 8.05, "learning_rate": 1.929203569921784e-06, "loss": 0.0386, "step": 26173 }, { "epoch": 8.05, "learning_rate": 1.9286152102843735e-06, "loss": 0.0181, "step": 26174 }, { "epoch": 8.05, "learning_rate": 1.9280269308032575e-06, "loss": 0.0236, "step": 26175 }, { "epoch": 8.05, "learning_rate": 1.927438731484269e-06, "loss": 0.019, "step": 26176 }, { "epoch": 8.05, "learning_rate": 1.9268506123332555e-06, "loss": 0.0252, "step": 26177 }, { "epoch": 8.05, "learning_rate": 1.926262573356057e-06, "loss": 0.0339, "step": 26178 }, { "epoch": 8.05, "learning_rate": 1.9256746145585103e-06, "loss": 0.0315, "step": 26179 }, { "epoch": 8.06, "learning_rate": 1.9250867359464575e-06, "loss": 0.0376, "step": 26180 }, { "epoch": 8.06, "learning_rate": 1.924498937525733e-06, "loss": 0.0745, "step": 26181 }, { "epoch": 8.06, "learning_rate": 1.923911219302181e-06, "loss": 0.0493, "step": 26182 }, { "epoch": 8.06, "learning_rate": 1.923323581281632e-06, "loss": 0.0242, "step": 26183 }, { "epoch": 8.06, "learning_rate": 1.922736023469921e-06, "loss": 0.0293, "step": 26184 }, { "epoch": 8.06, "learning_rate": 1.9221485458728874e-06, "loss": 0.0267, "step": 26185 }, { "epoch": 8.06, "learning_rate": 1.9215611484963635e-06, "loss": 0.026, "step": 26186 }, { "epoch": 8.06, "learning_rate": 1.9209738313461836e-06, "loss": 0.0179, "step": 26187 }, { "epoch": 8.06, "learning_rate": 1.9203865944281787e-06, "loss": 0.0557, "step": 26188 }, { "epoch": 8.06, "learning_rate": 1.919799437748181e-06, "loss": 0.0395, "step": 26189 }, { "epoch": 8.06, "learning_rate": 1.9192123613120216e-06, "loss": 0.0354, "step": 26190 }, { "epoch": 8.06, "learning_rate": 1.918625365125529e-06, "loss": 0.0299, "step": 26191 }, { "epoch": 8.06, "learning_rate": 1.9180384491945403e-06, "loss": 0.0166, "step": 26192 }, { "epoch": 8.06, "learning_rate": 1.9174516135248745e-06, "loss": 0.0416, "step": 26193 }, { "epoch": 8.06, "learning_rate": 1.9168648581223616e-06, "loss": 0.0318, "step": 26194 }, { "epoch": 8.06, "learning_rate": 1.9162781829928323e-06, "loss": 0.0128, "step": 26195 }, { "epoch": 8.06, "learning_rate": 1.915691588142111e-06, "loss": 0.0474, "step": 26196 }, { "epoch": 8.06, "learning_rate": 1.9151050735760237e-06, "loss": 0.0246, "step": 26197 }, { "epoch": 8.06, "learning_rate": 1.9145186393003935e-06, "loss": 0.0309, "step": 26198 }, { "epoch": 8.06, "learning_rate": 1.913932285321045e-06, "loss": 0.0167, "step": 26199 }, { "epoch": 8.06, "learning_rate": 1.913346011643802e-06, "loss": 0.0374, "step": 26200 }, { "epoch": 8.06, "learning_rate": 1.912759818274482e-06, "loss": 0.0398, "step": 26201 }, { "epoch": 8.06, "learning_rate": 1.912173705218914e-06, "loss": 0.0376, "step": 26202 }, { "epoch": 8.06, "learning_rate": 1.9115876724829174e-06, "loss": 0.0305, "step": 26203 }, { "epoch": 8.06, "learning_rate": 1.911001720072305e-06, "loss": 0.0316, "step": 26204 }, { "epoch": 8.06, "learning_rate": 1.9104158479929026e-06, "loss": 0.0372, "step": 26205 }, { "epoch": 8.06, "learning_rate": 1.9098300562505266e-06, "loss": 0.0075, "step": 26206 }, { "epoch": 8.06, "learning_rate": 1.9092443448509947e-06, "loss": 0.0372, "step": 26207 }, { "epoch": 8.06, "learning_rate": 1.9086587138001222e-06, "loss": 0.0202, "step": 26208 }, { "epoch": 8.06, "learning_rate": 1.9080731631037265e-06, "loss": 0.0935, "step": 26209 }, { "epoch": 8.06, "learning_rate": 1.9074876927676213e-06, "loss": 0.0207, "step": 26210 }, { "epoch": 8.06, "learning_rate": 1.9069023027976196e-06, "loss": 0.0277, "step": 26211 }, { "epoch": 8.06, "learning_rate": 1.9063169931995396e-06, "loss": 0.0496, "step": 26212 }, { "epoch": 8.07, "learning_rate": 1.905731763979194e-06, "loss": 0.0283, "step": 26213 }, { "epoch": 8.07, "learning_rate": 1.9051466151423858e-06, "loss": 0.0317, "step": 26214 }, { "epoch": 8.07, "learning_rate": 1.9045615466949353e-06, "loss": 0.0289, "step": 26215 }, { "epoch": 8.07, "learning_rate": 1.9039765586426496e-06, "loss": 0.0382, "step": 26216 }, { "epoch": 8.07, "learning_rate": 1.9033916509913376e-06, "loss": 0.0144, "step": 26217 }, { "epoch": 8.07, "learning_rate": 1.902806823746809e-06, "loss": 0.0135, "step": 26218 }, { "epoch": 8.07, "learning_rate": 1.902222076914869e-06, "loss": 0.0359, "step": 26219 }, { "epoch": 8.07, "learning_rate": 1.9016374105013324e-06, "loss": 0.041, "step": 26220 }, { "epoch": 8.07, "learning_rate": 1.9010528245119942e-06, "loss": 0.0324, "step": 26221 }, { "epoch": 8.07, "learning_rate": 1.9004683189526674e-06, "loss": 0.0658, "step": 26222 }, { "epoch": 8.07, "learning_rate": 1.8998838938291565e-06, "loss": 0.0212, "step": 26223 }, { "epoch": 8.07, "learning_rate": 1.8992995491472632e-06, "loss": 0.0165, "step": 26224 }, { "epoch": 8.07, "learning_rate": 1.898715284912791e-06, "loss": 0.0218, "step": 26225 }, { "epoch": 8.07, "learning_rate": 1.8981311011315395e-06, "loss": 0.0168, "step": 26226 }, { "epoch": 8.07, "learning_rate": 1.8975469978093185e-06, "loss": 0.04, "step": 26227 }, { "epoch": 8.07, "learning_rate": 1.8969629749519214e-06, "loss": 0.0591, "step": 26228 }, { "epoch": 8.07, "learning_rate": 1.8963790325651464e-06, "loss": 0.0325, "step": 26229 }, { "epoch": 8.07, "learning_rate": 1.8957951706548018e-06, "loss": 0.043, "step": 26230 }, { "epoch": 8.07, "learning_rate": 1.8952113892266755e-06, "loss": 0.0402, "step": 26231 }, { "epoch": 8.07, "learning_rate": 1.894627688286571e-06, "loss": 0.0481, "step": 26232 }, { "epoch": 8.07, "learning_rate": 1.894044067840285e-06, "loss": 0.0435, "step": 26233 }, { "epoch": 8.07, "learning_rate": 1.8934605278936113e-06, "loss": 0.0322, "step": 26234 }, { "epoch": 8.07, "learning_rate": 1.8928770684523456e-06, "loss": 0.0393, "step": 26235 }, { "epoch": 8.07, "learning_rate": 1.8922936895222798e-06, "loss": 0.0456, "step": 26236 }, { "epoch": 8.07, "learning_rate": 1.8917103911092161e-06, "loss": 0.0522, "step": 26237 }, { "epoch": 8.07, "learning_rate": 1.8911271732189372e-06, "loss": 0.029, "step": 26238 }, { "epoch": 8.07, "learning_rate": 1.8905440358572368e-06, "loss": 0.0341, "step": 26239 }, { "epoch": 8.07, "learning_rate": 1.8899609790299112e-06, "loss": 0.0358, "step": 26240 }, { "epoch": 8.07, "learning_rate": 1.8893780027427466e-06, "loss": 0.0376, "step": 26241 }, { "epoch": 8.07, "learning_rate": 1.8887951070015341e-06, "loss": 0.0235, "step": 26242 }, { "epoch": 8.07, "learning_rate": 1.8882122918120604e-06, "loss": 0.0169, "step": 26243 }, { "epoch": 8.07, "learning_rate": 1.8876295571801162e-06, "loss": 0.038, "step": 26244 }, { "epoch": 8.08, "learning_rate": 1.8870469031114868e-06, "loss": 0.0204, "step": 26245 }, { "epoch": 8.08, "learning_rate": 1.8864643296119556e-06, "loss": 0.0211, "step": 26246 }, { "epoch": 8.08, "learning_rate": 1.8858818366873145e-06, "loss": 0.0481, "step": 26247 }, { "epoch": 8.08, "learning_rate": 1.8852994243433465e-06, "loss": 0.0337, "step": 26248 }, { "epoch": 8.08, "learning_rate": 1.8847170925858293e-06, "loss": 0.0255, "step": 26249 }, { "epoch": 8.08, "learning_rate": 1.884134841420554e-06, "loss": 0.0231, "step": 26250 }, { "epoch": 8.08, "learning_rate": 1.8835526708532981e-06, "loss": 0.05, "step": 26251 }, { "epoch": 8.08, "learning_rate": 1.8829705808898458e-06, "loss": 0.0339, "step": 26252 }, { "epoch": 8.08, "learning_rate": 1.8823885715359757e-06, "loss": 0.024, "step": 26253 }, { "epoch": 8.08, "learning_rate": 1.8818066427974691e-06, "loss": 0.0193, "step": 26254 }, { "epoch": 8.08, "learning_rate": 1.8812247946801044e-06, "loss": 0.0332, "step": 26255 }, { "epoch": 8.08, "learning_rate": 1.880643027189657e-06, "loss": 0.022, "step": 26256 }, { "epoch": 8.08, "learning_rate": 1.88006134033191e-06, "loss": 0.036, "step": 26257 }, { "epoch": 8.08, "learning_rate": 1.8794797341126403e-06, "loss": 0.0355, "step": 26258 }, { "epoch": 8.08, "learning_rate": 1.878898208537615e-06, "loss": 0.0207, "step": 26259 }, { "epoch": 8.08, "learning_rate": 1.8783167636126187e-06, "loss": 0.0325, "step": 26260 }, { "epoch": 8.08, "learning_rate": 1.8777353993434221e-06, "loss": 0.0257, "step": 26261 }, { "epoch": 8.08, "learning_rate": 1.8771541157357975e-06, "loss": 0.0474, "step": 26262 }, { "epoch": 8.08, "learning_rate": 1.8765729127955202e-06, "loss": 0.0441, "step": 26263 }, { "epoch": 8.08, "learning_rate": 1.8759917905283575e-06, "loss": 0.0374, "step": 26264 }, { "epoch": 8.08, "learning_rate": 1.8754107489400885e-06, "loss": 0.0447, "step": 26265 }, { "epoch": 8.08, "learning_rate": 1.8748297880364742e-06, "loss": 0.0363, "step": 26266 }, { "epoch": 8.08, "learning_rate": 1.8742489078232894e-06, "loss": 0.0419, "step": 26267 }, { "epoch": 8.08, "learning_rate": 1.8736681083063035e-06, "loss": 0.0325, "step": 26268 }, { "epoch": 8.08, "learning_rate": 1.8730873894912827e-06, "loss": 0.0268, "step": 26269 }, { "epoch": 8.08, "learning_rate": 1.8725067513839934e-06, "loss": 0.0133, "step": 26270 }, { "epoch": 8.08, "learning_rate": 1.8719261939902023e-06, "loss": 0.0208, "step": 26271 }, { "epoch": 8.08, "learning_rate": 1.8713457173156757e-06, "loss": 0.0265, "step": 26272 }, { "epoch": 8.08, "learning_rate": 1.870765321366178e-06, "loss": 0.0207, "step": 26273 }, { "epoch": 8.08, "learning_rate": 1.8701850061474692e-06, "loss": 0.0188, "step": 26274 }, { "epoch": 8.08, "learning_rate": 1.869604771665322e-06, "loss": 0.0319, "step": 26275 }, { "epoch": 8.08, "learning_rate": 1.8690246179254877e-06, "loss": 0.0486, "step": 26276 }, { "epoch": 8.08, "learning_rate": 1.868444544933734e-06, "loss": 0.0415, "step": 26277 }, { "epoch": 8.09, "learning_rate": 1.8678645526958216e-06, "loss": 0.0327, "step": 26278 }, { "epoch": 8.09, "learning_rate": 1.8672846412175083e-06, "loss": 0.0612, "step": 26279 }, { "epoch": 8.09, "learning_rate": 1.8667048105045537e-06, "loss": 0.0243, "step": 26280 }, { "epoch": 8.09, "learning_rate": 1.8661250605627146e-06, "loss": 0.0243, "step": 26281 }, { "epoch": 8.09, "learning_rate": 1.865545391397754e-06, "loss": 0.0234, "step": 26282 }, { "epoch": 8.09, "learning_rate": 1.8649658030154226e-06, "loss": 0.0395, "step": 26283 }, { "epoch": 8.09, "learning_rate": 1.8643862954214754e-06, "loss": 0.0214, "step": 26284 }, { "epoch": 8.09, "learning_rate": 1.8638068686216726e-06, "loss": 0.0278, "step": 26285 }, { "epoch": 8.09, "learning_rate": 1.8632275226217665e-06, "loss": 0.0478, "step": 26286 }, { "epoch": 8.09, "learning_rate": 1.8626482574275107e-06, "loss": 0.0461, "step": 26287 }, { "epoch": 8.09, "learning_rate": 1.8620690730446567e-06, "loss": 0.0318, "step": 26288 }, { "epoch": 8.09, "learning_rate": 1.8614899694789567e-06, "loss": 0.01, "step": 26289 }, { "epoch": 8.09, "learning_rate": 1.860910946736163e-06, "loss": 0.0397, "step": 26290 }, { "epoch": 8.09, "learning_rate": 1.8603320048220208e-06, "loss": 0.0263, "step": 26291 }, { "epoch": 8.09, "learning_rate": 1.85975314374229e-06, "loss": 0.0367, "step": 26292 }, { "epoch": 8.09, "learning_rate": 1.8591743635027094e-06, "loss": 0.0315, "step": 26293 }, { "epoch": 8.09, "learning_rate": 1.8585956641090276e-06, "loss": 0.0392, "step": 26294 }, { "epoch": 8.09, "learning_rate": 1.8580170455669976e-06, "loss": 0.0394, "step": 26295 }, { "epoch": 8.09, "learning_rate": 1.8574385078823608e-06, "loss": 0.0257, "step": 26296 }, { "epoch": 8.09, "learning_rate": 1.8568600510608659e-06, "loss": 0.0392, "step": 26297 }, { "epoch": 8.09, "learning_rate": 1.8562816751082536e-06, "loss": 0.0227, "step": 26298 }, { "epoch": 8.09, "learning_rate": 1.8557033800302715e-06, "loss": 0.0312, "step": 26299 }, { "epoch": 8.09, "learning_rate": 1.8551251658326596e-06, "loss": 0.0314, "step": 26300 }, { "epoch": 8.09, "learning_rate": 1.854547032521159e-06, "loss": 0.0266, "step": 26301 }, { "epoch": 8.09, "learning_rate": 1.853968980101517e-06, "loss": 0.0224, "step": 26302 }, { "epoch": 8.09, "learning_rate": 1.8533910085794714e-06, "loss": 0.0286, "step": 26303 }, { "epoch": 8.09, "learning_rate": 1.8528131179607578e-06, "loss": 0.0437, "step": 26304 }, { "epoch": 8.09, "learning_rate": 1.8522353082511203e-06, "loss": 0.0401, "step": 26305 }, { "epoch": 8.09, "learning_rate": 1.8516575794562951e-06, "loss": 0.033, "step": 26306 }, { "epoch": 8.09, "learning_rate": 1.8510799315820206e-06, "loss": 0.0211, "step": 26307 }, { "epoch": 8.09, "learning_rate": 1.8505023646340304e-06, "loss": 0.0277, "step": 26308 }, { "epoch": 8.09, "learning_rate": 1.8499248786180667e-06, "loss": 0.0331, "step": 26309 }, { "epoch": 8.1, "learning_rate": 1.8493474735398575e-06, "loss": 0.0294, "step": 26310 }, { "epoch": 8.1, "learning_rate": 1.848770149405138e-06, "loss": 0.0397, "step": 26311 }, { "epoch": 8.1, "learning_rate": 1.8481929062196458e-06, "loss": 0.016, "step": 26312 }, { "epoch": 8.1, "learning_rate": 1.8476157439891108e-06, "loss": 0.0216, "step": 26313 }, { "epoch": 8.1, "learning_rate": 1.847038662719265e-06, "loss": 0.0262, "step": 26314 }, { "epoch": 8.1, "learning_rate": 1.8464616624158393e-06, "loss": 0.032, "step": 26315 }, { "epoch": 8.1, "learning_rate": 1.845884743084564e-06, "loss": 0.0358, "step": 26316 }, { "epoch": 8.1, "learning_rate": 1.8453079047311673e-06, "loss": 0.0369, "step": 26317 }, { "epoch": 8.1, "learning_rate": 1.8447311473613772e-06, "loss": 0.0472, "step": 26318 }, { "epoch": 8.1, "learning_rate": 1.8441544709809256e-06, "loss": 0.0523, "step": 26319 }, { "epoch": 8.1, "learning_rate": 1.843577875595538e-06, "loss": 0.0475, "step": 26320 }, { "epoch": 8.1, "learning_rate": 1.8430013612109355e-06, "loss": 0.0275, "step": 26321 }, { "epoch": 8.1, "learning_rate": 1.8424249278328488e-06, "loss": 0.0342, "step": 26322 }, { "epoch": 8.1, "learning_rate": 1.8418485754670013e-06, "loss": 0.0279, "step": 26323 }, { "epoch": 8.1, "learning_rate": 1.841272304119116e-06, "loss": 0.0159, "step": 26324 }, { "epoch": 8.1, "learning_rate": 1.8406961137949165e-06, "loss": 0.033, "step": 26325 }, { "epoch": 8.1, "learning_rate": 1.840120004500121e-06, "loss": 0.0391, "step": 26326 }, { "epoch": 8.1, "learning_rate": 1.83954397624046e-06, "loss": 0.0254, "step": 26327 }, { "epoch": 8.1, "learning_rate": 1.8389680290216428e-06, "loss": 0.0209, "step": 26328 }, { "epoch": 8.1, "learning_rate": 1.8383921628493972e-06, "loss": 0.0329, "step": 26329 }, { "epoch": 8.1, "learning_rate": 1.8378163777294421e-06, "loss": 0.0246, "step": 26330 }, { "epoch": 8.1, "learning_rate": 1.8372406736674874e-06, "loss": 0.0522, "step": 26331 }, { "epoch": 8.1, "learning_rate": 1.8366650506692596e-06, "loss": 0.0312, "step": 26332 }, { "epoch": 8.1, "learning_rate": 1.8360895087404696e-06, "loss": 0.0286, "step": 26333 }, { "epoch": 8.1, "learning_rate": 1.8355140478868362e-06, "loss": 0.0148, "step": 26334 }, { "epoch": 8.1, "learning_rate": 1.8349386681140723e-06, "loss": 0.0204, "step": 26335 }, { "epoch": 8.1, "learning_rate": 1.8343633694278895e-06, "loss": 0.0217, "step": 26336 }, { "epoch": 8.1, "learning_rate": 1.8337881518340107e-06, "loss": 0.0495, "step": 26337 }, { "epoch": 8.1, "learning_rate": 1.8332130153381355e-06, "loss": 0.0125, "step": 26338 }, { "epoch": 8.1, "learning_rate": 1.8326379599459842e-06, "loss": 0.0384, "step": 26339 }, { "epoch": 8.1, "learning_rate": 1.8320629856632654e-06, "loss": 0.036, "step": 26340 }, { "epoch": 8.1, "learning_rate": 1.8314880924956891e-06, "loss": 0.0282, "step": 26341 }, { "epoch": 8.1, "learning_rate": 1.830913280448965e-06, "loss": 0.0378, "step": 26342 }, { "epoch": 8.11, "learning_rate": 1.8303385495287972e-06, "loss": 0.0548, "step": 26343 }, { "epoch": 8.11, "learning_rate": 1.829763899740903e-06, "loss": 0.0221, "step": 26344 }, { "epoch": 8.11, "learning_rate": 1.8291893310909804e-06, "loss": 0.0368, "step": 26345 }, { "epoch": 8.11, "learning_rate": 1.8286148435847351e-06, "loss": 0.0241, "step": 26346 }, { "epoch": 8.11, "learning_rate": 1.8280404372278815e-06, "loss": 0.0257, "step": 26347 }, { "epoch": 8.11, "learning_rate": 1.8274661120261117e-06, "loss": 0.0221, "step": 26348 }, { "epoch": 8.11, "learning_rate": 1.8268918679851388e-06, "loss": 0.03, "step": 26349 }, { "epoch": 8.11, "learning_rate": 1.8263177051106618e-06, "loss": 0.0201, "step": 26350 }, { "epoch": 8.11, "learning_rate": 1.8257436234083824e-06, "loss": 0.0313, "step": 26351 }, { "epoch": 8.11, "learning_rate": 1.8251696228840032e-06, "loss": 0.023, "step": 26352 }, { "epoch": 8.11, "learning_rate": 1.8245957035432215e-06, "loss": 0.0573, "step": 26353 }, { "epoch": 8.11, "learning_rate": 1.824021865391744e-06, "loss": 0.0369, "step": 26354 }, { "epoch": 8.11, "learning_rate": 1.8234481084352619e-06, "loss": 0.0412, "step": 26355 }, { "epoch": 8.11, "learning_rate": 1.8228744326794735e-06, "loss": 0.0314, "step": 26356 }, { "epoch": 8.11, "learning_rate": 1.8223008381300812e-06, "loss": 0.0262, "step": 26357 }, { "epoch": 8.11, "learning_rate": 1.821727324792777e-06, "loss": 0.0471, "step": 26358 }, { "epoch": 8.11, "learning_rate": 1.8211538926732587e-06, "loss": 0.0261, "step": 26359 }, { "epoch": 8.11, "learning_rate": 1.8205805417772205e-06, "loss": 0.026, "step": 26360 }, { "epoch": 8.11, "learning_rate": 1.8200072721103545e-06, "loss": 0.0323, "step": 26361 }, { "epoch": 8.11, "learning_rate": 1.8194340836783565e-06, "loss": 0.0313, "step": 26362 }, { "epoch": 8.11, "learning_rate": 1.8188609764869147e-06, "loss": 0.0244, "step": 26363 }, { "epoch": 8.11, "learning_rate": 1.818287950541725e-06, "loss": 0.0324, "step": 26364 }, { "epoch": 8.11, "learning_rate": 1.8177150058484804e-06, "loss": 0.0425, "step": 26365 }, { "epoch": 8.11, "learning_rate": 1.817142142412861e-06, "loss": 0.0302, "step": 26366 }, { "epoch": 8.11, "learning_rate": 1.8165693602405643e-06, "loss": 0.0349, "step": 26367 }, { "epoch": 8.11, "learning_rate": 1.8159966593372757e-06, "loss": 0.0211, "step": 26368 }, { "epoch": 8.11, "learning_rate": 1.8154240397086831e-06, "loss": 0.028, "step": 26369 }, { "epoch": 8.11, "learning_rate": 1.8148515013604728e-06, "loss": 0.0434, "step": 26370 }, { "epoch": 8.11, "learning_rate": 1.8142790442983304e-06, "loss": 0.0386, "step": 26371 }, { "epoch": 8.11, "learning_rate": 1.8137066685279414e-06, "loss": 0.0337, "step": 26372 }, { "epoch": 8.11, "learning_rate": 1.813134374054988e-06, "loss": 0.0514, "step": 26373 }, { "epoch": 8.11, "learning_rate": 1.8125621608851574e-06, "loss": 0.0329, "step": 26374 }, { "epoch": 8.12, "learning_rate": 1.8119900290241331e-06, "loss": 0.0284, "step": 26375 }, { "epoch": 8.12, "learning_rate": 1.8114179784775886e-06, "loss": 0.0503, "step": 26376 }, { "epoch": 8.12, "learning_rate": 1.8108460092512137e-06, "loss": 0.0416, "step": 26377 }, { "epoch": 8.12, "learning_rate": 1.8102741213506847e-06, "loss": 0.0533, "step": 26378 }, { "epoch": 8.12, "learning_rate": 1.8097023147816817e-06, "loss": 0.0236, "step": 26379 }, { "epoch": 8.12, "learning_rate": 1.8091305895498835e-06, "loss": 0.0363, "step": 26380 }, { "epoch": 8.12, "learning_rate": 1.8085589456609641e-06, "loss": 0.0439, "step": 26381 }, { "epoch": 8.12, "learning_rate": 1.8079873831206096e-06, "loss": 0.0235, "step": 26382 }, { "epoch": 8.12, "learning_rate": 1.807415901934485e-06, "loss": 0.0411, "step": 26383 }, { "epoch": 8.12, "learning_rate": 1.8068445021082738e-06, "loss": 0.039, "step": 26384 }, { "epoch": 8.12, "learning_rate": 1.8062731836476477e-06, "loss": 0.0352, "step": 26385 }, { "epoch": 8.12, "learning_rate": 1.8057019465582803e-06, "loss": 0.0283, "step": 26386 }, { "epoch": 8.12, "learning_rate": 1.805130790845845e-06, "loss": 0.0241, "step": 26387 }, { "epoch": 8.12, "learning_rate": 1.8045597165160134e-06, "loss": 0.0365, "step": 26388 }, { "epoch": 8.12, "learning_rate": 1.8039887235744557e-06, "loss": 0.0262, "step": 26389 }, { "epoch": 8.12, "learning_rate": 1.8034178120268452e-06, "loss": 0.0259, "step": 26390 }, { "epoch": 8.12, "learning_rate": 1.8028469818788475e-06, "loss": 0.0492, "step": 26391 }, { "epoch": 8.12, "learning_rate": 1.8022762331361376e-06, "loss": 0.0241, "step": 26392 }, { "epoch": 8.12, "learning_rate": 1.801705565804377e-06, "loss": 0.0283, "step": 26393 }, { "epoch": 8.12, "learning_rate": 1.8011349798892363e-06, "loss": 0.028, "step": 26394 }, { "epoch": 8.12, "learning_rate": 1.8005644753963824e-06, "loss": 0.0269, "step": 26395 }, { "epoch": 8.12, "learning_rate": 1.7999940523314808e-06, "loss": 0.0237, "step": 26396 }, { "epoch": 8.12, "learning_rate": 1.7994237107001955e-06, "loss": 0.026, "step": 26397 }, { "epoch": 8.12, "learning_rate": 1.7988534505081867e-06, "loss": 0.0173, "step": 26398 }, { "epoch": 8.12, "learning_rate": 1.7982832717611277e-06, "loss": 0.02, "step": 26399 }, { "epoch": 8.12, "learning_rate": 1.7977131744646724e-06, "loss": 0.0299, "step": 26400 }, { "epoch": 8.12, "learning_rate": 1.7971431586244814e-06, "loss": 0.0443, "step": 26401 }, { "epoch": 8.12, "learning_rate": 1.7965732242462218e-06, "loss": 0.0203, "step": 26402 }, { "epoch": 8.12, "learning_rate": 1.7960033713355507e-06, "loss": 0.037, "step": 26403 }, { "epoch": 8.12, "learning_rate": 1.7954335998981277e-06, "loss": 0.0298, "step": 26404 }, { "epoch": 8.12, "learning_rate": 1.7948639099396092e-06, "loss": 0.0248, "step": 26405 }, { "epoch": 8.12, "learning_rate": 1.7942943014656543e-06, "loss": 0.0218, "step": 26406 }, { "epoch": 8.12, "learning_rate": 1.7937247744819208e-06, "loss": 0.0311, "step": 26407 }, { "epoch": 8.13, "learning_rate": 1.7931553289940607e-06, "loss": 0.0191, "step": 26408 }, { "epoch": 8.13, "learning_rate": 1.792585965007736e-06, "loss": 0.0303, "step": 26409 }, { "epoch": 8.13, "learning_rate": 1.7920166825285944e-06, "loss": 0.0337, "step": 26410 }, { "epoch": 8.13, "learning_rate": 1.7914474815622896e-06, "loss": 0.0308, "step": 26411 }, { "epoch": 8.13, "learning_rate": 1.7908783621144798e-06, "loss": 0.0445, "step": 26412 }, { "epoch": 8.13, "learning_rate": 1.7903093241908121e-06, "loss": 0.0547, "step": 26413 }, { "epoch": 8.13, "learning_rate": 1.7897403677969405e-06, "loss": 0.0239, "step": 26414 }, { "epoch": 8.13, "learning_rate": 1.7891714929385128e-06, "loss": 0.0165, "step": 26415 }, { "epoch": 8.13, "learning_rate": 1.7886026996211804e-06, "loss": 0.0239, "step": 26416 }, { "epoch": 8.13, "learning_rate": 1.7880339878505903e-06, "loss": 0.0395, "step": 26417 }, { "epoch": 8.13, "learning_rate": 1.7874653576323885e-06, "loss": 0.0584, "step": 26418 }, { "epoch": 8.13, "learning_rate": 1.7868968089722283e-06, "loss": 0.0192, "step": 26419 }, { "epoch": 8.13, "learning_rate": 1.7863283418757537e-06, "loss": 0.0275, "step": 26420 }, { "epoch": 8.13, "learning_rate": 1.785759956348604e-06, "loss": 0.0308, "step": 26421 }, { "epoch": 8.13, "learning_rate": 1.7851916523964297e-06, "loss": 0.0314, "step": 26422 }, { "epoch": 8.13, "learning_rate": 1.7846234300248743e-06, "loss": 0.025, "step": 26423 }, { "epoch": 8.13, "learning_rate": 1.7840552892395802e-06, "loss": 0.0245, "step": 26424 }, { "epoch": 8.13, "learning_rate": 1.783487230046188e-06, "loss": 0.0199, "step": 26425 }, { "epoch": 8.13, "learning_rate": 1.7829192524503414e-06, "loss": 0.0255, "step": 26426 }, { "epoch": 8.13, "learning_rate": 1.7823513564576788e-06, "loss": 0.0183, "step": 26427 }, { "epoch": 8.13, "learning_rate": 1.7817835420738382e-06, "loss": 0.0456, "step": 26428 }, { "epoch": 8.13, "learning_rate": 1.781215809304464e-06, "loss": 0.0217, "step": 26429 }, { "epoch": 8.13, "learning_rate": 1.7806481581551948e-06, "loss": 0.036, "step": 26430 }, { "epoch": 8.13, "learning_rate": 1.7800805886316585e-06, "loss": 0.0319, "step": 26431 }, { "epoch": 8.13, "learning_rate": 1.7795131007395006e-06, "loss": 0.0305, "step": 26432 }, { "epoch": 8.13, "learning_rate": 1.7789456944843542e-06, "loss": 0.0434, "step": 26433 }, { "epoch": 8.13, "learning_rate": 1.7783783698718527e-06, "loss": 0.0453, "step": 26434 }, { "epoch": 8.13, "learning_rate": 1.7778111269076314e-06, "loss": 0.0293, "step": 26435 }, { "epoch": 8.13, "learning_rate": 1.7772439655973217e-06, "loss": 0.0211, "step": 26436 }, { "epoch": 8.13, "learning_rate": 1.7766768859465623e-06, "loss": 0.041, "step": 26437 }, { "epoch": 8.13, "learning_rate": 1.7761098879609762e-06, "loss": 0.0232, "step": 26438 }, { "epoch": 8.13, "learning_rate": 1.7755429716461992e-06, "loss": 0.0376, "step": 26439 }, { "epoch": 8.14, "learning_rate": 1.774976137007861e-06, "loss": 0.0307, "step": 26440 }, { "epoch": 8.14, "learning_rate": 1.7744093840515907e-06, "loss": 0.0283, "step": 26441 }, { "epoch": 8.14, "learning_rate": 1.773842712783016e-06, "loss": 0.0392, "step": 26442 }, { "epoch": 8.14, "learning_rate": 1.773276123207761e-06, "loss": 0.0615, "step": 26443 }, { "epoch": 8.14, "learning_rate": 1.7727096153314627e-06, "loss": 0.0361, "step": 26444 }, { "epoch": 8.14, "learning_rate": 1.7721431891597362e-06, "loss": 0.0277, "step": 26445 }, { "epoch": 8.14, "learning_rate": 1.7715768446982097e-06, "loss": 0.0311, "step": 26446 }, { "epoch": 8.14, "learning_rate": 1.7710105819525136e-06, "loss": 0.0362, "step": 26447 }, { "epoch": 8.14, "learning_rate": 1.7704444009282606e-06, "loss": 0.0273, "step": 26448 }, { "epoch": 8.14, "learning_rate": 1.7698783016310827e-06, "loss": 0.0616, "step": 26449 }, { "epoch": 8.14, "learning_rate": 1.7693122840665978e-06, "loss": 0.0425, "step": 26450 }, { "epoch": 8.14, "learning_rate": 1.7687463482404276e-06, "loss": 0.0377, "step": 26451 }, { "epoch": 8.14, "learning_rate": 1.7681804941581925e-06, "loss": 0.0131, "step": 26452 }, { "epoch": 8.14, "learning_rate": 1.7676147218255092e-06, "loss": 0.0376, "step": 26453 }, { "epoch": 8.14, "learning_rate": 1.7670490312480037e-06, "loss": 0.0419, "step": 26454 }, { "epoch": 8.14, "learning_rate": 1.7664834224312865e-06, "loss": 0.0444, "step": 26455 }, { "epoch": 8.14, "learning_rate": 1.7659178953809752e-06, "loss": 0.0227, "step": 26456 }, { "epoch": 8.14, "learning_rate": 1.7653524501026898e-06, "loss": 0.0072, "step": 26457 }, { "epoch": 8.14, "learning_rate": 1.7647870866020445e-06, "loss": 0.0407, "step": 26458 }, { "epoch": 8.14, "learning_rate": 1.7642218048846527e-06, "loss": 0.0229, "step": 26459 }, { "epoch": 8.14, "learning_rate": 1.7636566049561287e-06, "loss": 0.0517, "step": 26460 }, { "epoch": 8.14, "learning_rate": 1.763091486822086e-06, "loss": 0.0387, "step": 26461 }, { "epoch": 8.14, "learning_rate": 1.7625264504881356e-06, "loss": 0.0362, "step": 26462 }, { "epoch": 8.14, "learning_rate": 1.7619614959598874e-06, "loss": 0.051, "step": 26463 }, { "epoch": 8.14, "learning_rate": 1.7613966232429559e-06, "loss": 0.0344, "step": 26464 }, { "epoch": 8.14, "learning_rate": 1.7608318323429518e-06, "loss": 0.0515, "step": 26465 }, { "epoch": 8.14, "learning_rate": 1.7602671232654755e-06, "loss": 0.0254, "step": 26466 }, { "epoch": 8.14, "learning_rate": 1.7597024960161436e-06, "loss": 0.059, "step": 26467 }, { "epoch": 8.14, "learning_rate": 1.7591379506005612e-06, "loss": 0.0419, "step": 26468 }, { "epoch": 8.14, "learning_rate": 1.7585734870243321e-06, "loss": 0.0372, "step": 26469 }, { "epoch": 8.14, "learning_rate": 1.7580091052930648e-06, "loss": 0.0156, "step": 26470 }, { "epoch": 8.14, "learning_rate": 1.757444805412364e-06, "loss": 0.0479, "step": 26471 }, { "epoch": 8.14, "learning_rate": 1.7568805873878315e-06, "loss": 0.0272, "step": 26472 }, { "epoch": 8.15, "learning_rate": 1.7563164512250697e-06, "loss": 0.0291, "step": 26473 }, { "epoch": 8.15, "learning_rate": 1.7557523969296852e-06, "loss": 0.0168, "step": 26474 }, { "epoch": 8.15, "learning_rate": 1.7551884245072813e-06, "loss": 0.0175, "step": 26475 }, { "epoch": 8.15, "learning_rate": 1.7546245339634494e-06, "loss": 0.0245, "step": 26476 }, { "epoch": 8.15, "learning_rate": 1.7540607253037966e-06, "loss": 0.0417, "step": 26477 }, { "epoch": 8.15, "learning_rate": 1.7534969985339212e-06, "loss": 0.0265, "step": 26478 }, { "epoch": 8.15, "learning_rate": 1.7529333536594217e-06, "loss": 0.042, "step": 26479 }, { "epoch": 8.15, "learning_rate": 1.7523697906858939e-06, "loss": 0.0397, "step": 26480 }, { "epoch": 8.15, "learning_rate": 1.7518063096189319e-06, "loss": 0.0384, "step": 26481 }, { "epoch": 8.15, "learning_rate": 1.7512429104641415e-06, "loss": 0.0348, "step": 26482 }, { "epoch": 8.15, "learning_rate": 1.7506795932271058e-06, "loss": 0.0152, "step": 26483 }, { "epoch": 8.15, "learning_rate": 1.750116357913426e-06, "loss": 0.0329, "step": 26484 }, { "epoch": 8.15, "learning_rate": 1.7495532045286955e-06, "loss": 0.0341, "step": 26485 }, { "epoch": 8.15, "learning_rate": 1.7489901330785053e-06, "loss": 0.0511, "step": 26486 }, { "epoch": 8.15, "learning_rate": 1.7484271435684474e-06, "loss": 0.0219, "step": 26487 }, { "epoch": 8.15, "learning_rate": 1.747864236004112e-06, "loss": 0.0404, "step": 26488 }, { "epoch": 8.15, "learning_rate": 1.7473014103910912e-06, "loss": 0.029, "step": 26489 }, { "epoch": 8.15, "learning_rate": 1.7467386667349728e-06, "loss": 0.0453, "step": 26490 }, { "epoch": 8.15, "learning_rate": 1.7461760050413435e-06, "loss": 0.0131, "step": 26491 }, { "epoch": 8.15, "learning_rate": 1.7456134253157976e-06, "loss": 0.0381, "step": 26492 }, { "epoch": 8.15, "learning_rate": 1.7450509275639139e-06, "loss": 0.0369, "step": 26493 }, { "epoch": 8.15, "learning_rate": 1.7444885117912845e-06, "loss": 0.0217, "step": 26494 }, { "epoch": 8.15, "learning_rate": 1.7439261780034932e-06, "loss": 0.0317, "step": 26495 }, { "epoch": 8.15, "learning_rate": 1.743363926206123e-06, "loss": 0.0247, "step": 26496 }, { "epoch": 8.15, "learning_rate": 1.7428017564047594e-06, "loss": 0.038, "step": 26497 }, { "epoch": 8.15, "learning_rate": 1.7422396686049813e-06, "loss": 0.022, "step": 26498 }, { "epoch": 8.15, "learning_rate": 1.7416776628123798e-06, "loss": 0.0147, "step": 26499 }, { "epoch": 8.15, "learning_rate": 1.7411157390325273e-06, "loss": 0.0262, "step": 26500 }, { "epoch": 8.15, "learning_rate": 1.7405538972710044e-06, "loss": 0.0343, "step": 26501 }, { "epoch": 8.15, "learning_rate": 1.7399921375333962e-06, "loss": 0.0279, "step": 26502 }, { "epoch": 8.15, "learning_rate": 1.7394304598252776e-06, "loss": 0.0231, "step": 26503 }, { "epoch": 8.15, "learning_rate": 1.7388688641522289e-06, "loss": 0.0154, "step": 26504 }, { "epoch": 8.16, "learning_rate": 1.7383073505198255e-06, "loss": 0.014, "step": 26505 }, { "epoch": 8.16, "learning_rate": 1.7377459189336442e-06, "loss": 0.0319, "step": 26506 }, { "epoch": 8.16, "learning_rate": 1.7371845693992618e-06, "loss": 0.0303, "step": 26507 }, { "epoch": 8.16, "learning_rate": 1.736623301922249e-06, "loss": 0.0236, "step": 26508 }, { "epoch": 8.16, "learning_rate": 1.736062116508187e-06, "loss": 0.0424, "step": 26509 }, { "epoch": 8.16, "learning_rate": 1.7355010131626393e-06, "loss": 0.0249, "step": 26510 }, { "epoch": 8.16, "learning_rate": 1.7349399918911858e-06, "loss": 0.0153, "step": 26511 }, { "epoch": 8.16, "learning_rate": 1.7343790526993965e-06, "loss": 0.0264, "step": 26512 }, { "epoch": 8.16, "learning_rate": 1.7338181955928401e-06, "loss": 0.0292, "step": 26513 }, { "epoch": 8.16, "learning_rate": 1.7332574205770868e-06, "loss": 0.0246, "step": 26514 }, { "epoch": 8.16, "learning_rate": 1.7326967276577044e-06, "loss": 0.0242, "step": 26515 }, { "epoch": 8.16, "learning_rate": 1.7321361168402684e-06, "loss": 0.0209, "step": 26516 }, { "epoch": 8.16, "learning_rate": 1.7315755881303377e-06, "loss": 0.0212, "step": 26517 }, { "epoch": 8.16, "learning_rate": 1.7310151415334798e-06, "loss": 0.0236, "step": 26518 }, { "epoch": 8.16, "learning_rate": 1.7304547770552638e-06, "loss": 0.0348, "step": 26519 }, { "epoch": 8.16, "learning_rate": 1.7298944947012542e-06, "loss": 0.0392, "step": 26520 }, { "epoch": 8.16, "learning_rate": 1.7293342944770142e-06, "loss": 0.0222, "step": 26521 }, { "epoch": 8.16, "learning_rate": 1.728774176388106e-06, "loss": 0.0225, "step": 26522 }, { "epoch": 8.16, "learning_rate": 1.7282141404400943e-06, "loss": 0.0321, "step": 26523 }, { "epoch": 8.16, "learning_rate": 1.7276541866385398e-06, "loss": 0.0274, "step": 26524 }, { "epoch": 8.16, "learning_rate": 1.7270943149890006e-06, "loss": 0.0279, "step": 26525 }, { "epoch": 8.16, "learning_rate": 1.7265345254970445e-06, "loss": 0.0301, "step": 26526 }, { "epoch": 8.16, "learning_rate": 1.7259748181682222e-06, "loss": 0.0302, "step": 26527 }, { "epoch": 8.16, "learning_rate": 1.7254151930080942e-06, "loss": 0.0348, "step": 26528 }, { "epoch": 8.16, "learning_rate": 1.7248556500222213e-06, "loss": 0.0269, "step": 26529 }, { "epoch": 8.16, "learning_rate": 1.7242961892161581e-06, "loss": 0.0098, "step": 26530 }, { "epoch": 8.16, "learning_rate": 1.723736810595461e-06, "loss": 0.0234, "step": 26531 }, { "epoch": 8.16, "learning_rate": 1.7231775141656858e-06, "loss": 0.0541, "step": 26532 }, { "epoch": 8.16, "learning_rate": 1.7226182999323849e-06, "loss": 0.022, "step": 26533 }, { "epoch": 8.16, "learning_rate": 1.7220591679011135e-06, "loss": 0.0265, "step": 26534 }, { "epoch": 8.16, "learning_rate": 1.7215001180774215e-06, "loss": 0.0233, "step": 26535 }, { "epoch": 8.16, "learning_rate": 1.720941150466865e-06, "loss": 0.0372, "step": 26536 }, { "epoch": 8.16, "learning_rate": 1.7203822650749957e-06, "loss": 0.0207, "step": 26537 }, { "epoch": 8.17, "learning_rate": 1.7198234619073562e-06, "loss": 0.0227, "step": 26538 }, { "epoch": 8.17, "learning_rate": 1.719264740969504e-06, "loss": 0.0292, "step": 26539 }, { "epoch": 8.17, "learning_rate": 1.7187061022669848e-06, "loss": 0.0213, "step": 26540 }, { "epoch": 8.17, "learning_rate": 1.7181475458053453e-06, "loss": 0.0158, "step": 26541 }, { "epoch": 8.17, "learning_rate": 1.7175890715901343e-06, "loss": 0.0533, "step": 26542 }, { "epoch": 8.17, "learning_rate": 1.7170306796268942e-06, "loss": 0.0304, "step": 26543 }, { "epoch": 8.17, "learning_rate": 1.7164723699211782e-06, "loss": 0.0631, "step": 26544 }, { "epoch": 8.17, "learning_rate": 1.715914142478521e-06, "loss": 0.0263, "step": 26545 }, { "epoch": 8.17, "learning_rate": 1.715355997304473e-06, "loss": 0.0131, "step": 26546 }, { "epoch": 8.17, "learning_rate": 1.7147979344045785e-06, "loss": 0.0237, "step": 26547 }, { "epoch": 8.17, "learning_rate": 1.7142399537843712e-06, "loss": 0.0267, "step": 26548 }, { "epoch": 8.17, "learning_rate": 1.7136820554494005e-06, "loss": 0.026, "step": 26549 }, { "epoch": 8.17, "learning_rate": 1.713124239405204e-06, "loss": 0.0196, "step": 26550 }, { "epoch": 8.17, "learning_rate": 1.7125665056573205e-06, "loss": 0.0303, "step": 26551 }, { "epoch": 8.17, "learning_rate": 1.71200885421129e-06, "loss": 0.0311, "step": 26552 }, { "epoch": 8.17, "learning_rate": 1.7114512850726473e-06, "loss": 0.0386, "step": 26553 }, { "epoch": 8.17, "learning_rate": 1.7108937982469376e-06, "loss": 0.0245, "step": 26554 }, { "epoch": 8.17, "learning_rate": 1.7103363937396867e-06, "loss": 0.0259, "step": 26555 }, { "epoch": 8.17, "learning_rate": 1.7097790715564377e-06, "loss": 0.0457, "step": 26556 }, { "epoch": 8.17, "learning_rate": 1.709221831702723e-06, "loss": 0.0369, "step": 26557 }, { "epoch": 8.17, "learning_rate": 1.7086646741840762e-06, "loss": 0.0214, "step": 26558 }, { "epoch": 8.17, "learning_rate": 1.7081075990060314e-06, "loss": 0.0235, "step": 26559 }, { "epoch": 8.17, "learning_rate": 1.7075506061741175e-06, "loss": 0.0295, "step": 26560 }, { "epoch": 8.17, "learning_rate": 1.7069936956938737e-06, "loss": 0.0423, "step": 26561 }, { "epoch": 8.17, "learning_rate": 1.7064368675708232e-06, "loss": 0.0375, "step": 26562 }, { "epoch": 8.17, "learning_rate": 1.7058801218104947e-06, "loss": 0.0297, "step": 26563 }, { "epoch": 8.17, "learning_rate": 1.705323458418424e-06, "loss": 0.0299, "step": 26564 }, { "epoch": 8.17, "learning_rate": 1.7047668774001369e-06, "loss": 0.0287, "step": 26565 }, { "epoch": 8.17, "learning_rate": 1.7042103787611585e-06, "loss": 0.0251, "step": 26566 }, { "epoch": 8.17, "learning_rate": 1.703653962507017e-06, "loss": 0.0489, "step": 26567 }, { "epoch": 8.17, "learning_rate": 1.7030976286432376e-06, "loss": 0.0195, "step": 26568 }, { "epoch": 8.17, "learning_rate": 1.7025413771753452e-06, "loss": 0.043, "step": 26569 }, { "epoch": 8.18, "learning_rate": 1.7019852081088616e-06, "loss": 0.0228, "step": 26570 }, { "epoch": 8.18, "learning_rate": 1.7014291214493184e-06, "loss": 0.0389, "step": 26571 }, { "epoch": 8.18, "learning_rate": 1.7008731172022296e-06, "loss": 0.0323, "step": 26572 }, { "epoch": 8.18, "learning_rate": 1.7003171953731157e-06, "loss": 0.0228, "step": 26573 }, { "epoch": 8.18, "learning_rate": 1.6997613559675053e-06, "loss": 0.0168, "step": 26574 }, { "epoch": 8.18, "learning_rate": 1.6992055989909129e-06, "loss": 0.0184, "step": 26575 }, { "epoch": 8.18, "learning_rate": 1.6986499244488607e-06, "loss": 0.0277, "step": 26576 }, { "epoch": 8.18, "learning_rate": 1.6980943323468645e-06, "loss": 0.0582, "step": 26577 }, { "epoch": 8.18, "learning_rate": 1.6975388226904433e-06, "loss": 0.0225, "step": 26578 }, { "epoch": 8.18, "learning_rate": 1.6969833954851123e-06, "loss": 0.0328, "step": 26579 }, { "epoch": 8.18, "learning_rate": 1.6964280507363872e-06, "loss": 0.0439, "step": 26580 }, { "epoch": 8.18, "learning_rate": 1.695872788449786e-06, "loss": 0.0403, "step": 26581 }, { "epoch": 8.18, "learning_rate": 1.6953176086308242e-06, "loss": 0.0399, "step": 26582 }, { "epoch": 8.18, "learning_rate": 1.6947625112850074e-06, "loss": 0.0288, "step": 26583 }, { "epoch": 8.18, "learning_rate": 1.6942074964178567e-06, "loss": 0.0219, "step": 26584 }, { "epoch": 8.18, "learning_rate": 1.693652564034879e-06, "loss": 0.0384, "step": 26585 }, { "epoch": 8.18, "learning_rate": 1.6930977141415872e-06, "loss": 0.0249, "step": 26586 }, { "epoch": 8.18, "learning_rate": 1.6925429467434906e-06, "loss": 0.0408, "step": 26587 }, { "epoch": 8.18, "learning_rate": 1.6919882618460992e-06, "loss": 0.0259, "step": 26588 }, { "epoch": 8.18, "learning_rate": 1.691433659454922e-06, "loss": 0.036, "step": 26589 }, { "epoch": 8.18, "learning_rate": 1.690879139575462e-06, "loss": 0.0174, "step": 26590 }, { "epoch": 8.18, "learning_rate": 1.6903247022132329e-06, "loss": 0.0267, "step": 26591 }, { "epoch": 8.18, "learning_rate": 1.6897703473737414e-06, "loss": 0.0258, "step": 26592 }, { "epoch": 8.18, "learning_rate": 1.6892160750624831e-06, "loss": 0.042, "step": 26593 }, { "epoch": 8.18, "learning_rate": 1.6886618852849723e-06, "loss": 0.0325, "step": 26594 }, { "epoch": 8.18, "learning_rate": 1.6881077780467082e-06, "loss": 0.0356, "step": 26595 }, { "epoch": 8.18, "learning_rate": 1.687553753353195e-06, "loss": 0.0359, "step": 26596 }, { "epoch": 8.18, "learning_rate": 1.6869998112099328e-06, "loss": 0.0287, "step": 26597 }, { "epoch": 8.18, "learning_rate": 1.6864459516224218e-06, "loss": 0.0398, "step": 26598 }, { "epoch": 8.18, "learning_rate": 1.6858921745961698e-06, "loss": 0.02, "step": 26599 }, { "epoch": 8.18, "learning_rate": 1.6853384801366669e-06, "loss": 0.0167, "step": 26600 }, { "epoch": 8.18, "learning_rate": 1.6847848682494172e-06, "loss": 0.0358, "step": 26601 }, { "epoch": 8.18, "learning_rate": 1.6842313389399178e-06, "loss": 0.0383, "step": 26602 }, { "epoch": 8.19, "learning_rate": 1.6836778922136642e-06, "loss": 0.038, "step": 26603 }, { "epoch": 8.19, "learning_rate": 1.6831245280761555e-06, "loss": 0.0354, "step": 26604 }, { "epoch": 8.19, "learning_rate": 1.6825712465328835e-06, "loss": 0.0196, "step": 26605 }, { "epoch": 8.19, "learning_rate": 1.6820180475893455e-06, "loss": 0.0166, "step": 26606 }, { "epoch": 8.19, "learning_rate": 1.6814649312510346e-06, "loss": 0.0404, "step": 26607 }, { "epoch": 8.19, "learning_rate": 1.6809118975234407e-06, "loss": 0.0493, "step": 26608 }, { "epoch": 8.19, "learning_rate": 1.680358946412064e-06, "loss": 0.0263, "step": 26609 }, { "epoch": 8.19, "learning_rate": 1.6798060779223858e-06, "loss": 0.0208, "step": 26610 }, { "epoch": 8.19, "learning_rate": 1.6792532920599025e-06, "loss": 0.0314, "step": 26611 }, { "epoch": 8.19, "learning_rate": 1.6787005888301045e-06, "loss": 0.0217, "step": 26612 }, { "epoch": 8.19, "learning_rate": 1.678147968238477e-06, "loss": 0.0186, "step": 26613 }, { "epoch": 8.19, "learning_rate": 1.6775954302905107e-06, "loss": 0.0198, "step": 26614 }, { "epoch": 8.19, "learning_rate": 1.6770429749916895e-06, "loss": 0.031, "step": 26615 }, { "epoch": 8.19, "learning_rate": 1.6764906023475069e-06, "loss": 0.0258, "step": 26616 }, { "epoch": 8.19, "learning_rate": 1.6759383123634421e-06, "loss": 0.0193, "step": 26617 }, { "epoch": 8.19, "learning_rate": 1.6753861050449783e-06, "loss": 0.0189, "step": 26618 }, { "epoch": 8.19, "learning_rate": 1.6748339803976054e-06, "loss": 0.0427, "step": 26619 }, { "epoch": 8.19, "learning_rate": 1.674281938426804e-06, "loss": 0.0311, "step": 26620 }, { "epoch": 8.19, "learning_rate": 1.6737299791380558e-06, "loss": 0.03, "step": 26621 }, { "epoch": 8.19, "learning_rate": 1.6731781025368422e-06, "loss": 0.035, "step": 26622 }, { "epoch": 8.19, "learning_rate": 1.6726263086286444e-06, "loss": 0.0501, "step": 26623 }, { "epoch": 8.19, "learning_rate": 1.6720745974189422e-06, "loss": 0.0375, "step": 26624 }, { "epoch": 8.19, "learning_rate": 1.6715229689132128e-06, "loss": 0.0249, "step": 26625 }, { "epoch": 8.19, "learning_rate": 1.6709714231169405e-06, "loss": 0.0287, "step": 26626 }, { "epoch": 8.19, "learning_rate": 1.6704199600355952e-06, "loss": 0.0283, "step": 26627 }, { "epoch": 8.19, "learning_rate": 1.669868579674655e-06, "loss": 0.0279, "step": 26628 }, { "epoch": 8.19, "learning_rate": 1.6693172820395986e-06, "loss": 0.037, "step": 26629 }, { "epoch": 8.19, "learning_rate": 1.6687660671358997e-06, "loss": 0.0178, "step": 26630 }, { "epoch": 8.19, "learning_rate": 1.6682149349690325e-06, "loss": 0.0605, "step": 26631 }, { "epoch": 8.19, "learning_rate": 1.6676638855444693e-06, "loss": 0.0271, "step": 26632 }, { "epoch": 8.19, "learning_rate": 1.6671129188676838e-06, "loss": 0.0547, "step": 26633 }, { "epoch": 8.19, "learning_rate": 1.6665620349441459e-06, "loss": 0.029, "step": 26634 }, { "epoch": 8.2, "learning_rate": 1.6660112337793256e-06, "loss": 0.0228, "step": 26635 }, { "epoch": 8.2, "learning_rate": 1.6654605153786961e-06, "loss": 0.0335, "step": 26636 }, { "epoch": 8.2, "learning_rate": 1.6649098797477282e-06, "loss": 0.0337, "step": 26637 }, { "epoch": 8.2, "learning_rate": 1.6643593268918812e-06, "loss": 0.0659, "step": 26638 }, { "epoch": 8.2, "learning_rate": 1.663808856816632e-06, "loss": 0.0321, "step": 26639 }, { "epoch": 8.2, "learning_rate": 1.6632584695274434e-06, "loss": 0.0384, "step": 26640 }, { "epoch": 8.2, "learning_rate": 1.6627081650297815e-06, "loss": 0.0556, "step": 26641 }, { "epoch": 8.2, "learning_rate": 1.6621579433291112e-06, "loss": 0.0292, "step": 26642 }, { "epoch": 8.2, "learning_rate": 1.6616078044308948e-06, "loss": 0.0276, "step": 26643 }, { "epoch": 8.2, "learning_rate": 1.661057748340602e-06, "loss": 0.0247, "step": 26644 }, { "epoch": 8.2, "learning_rate": 1.6605077750636867e-06, "loss": 0.0543, "step": 26645 }, { "epoch": 8.2, "learning_rate": 1.6599578846056185e-06, "loss": 0.0234, "step": 26646 }, { "epoch": 8.2, "learning_rate": 1.6594080769718558e-06, "loss": 0.0523, "step": 26647 }, { "epoch": 8.2, "learning_rate": 1.6588583521678536e-06, "loss": 0.0234, "step": 26648 }, { "epoch": 8.2, "learning_rate": 1.658308710199078e-06, "loss": 0.0208, "step": 26649 }, { "epoch": 8.2, "learning_rate": 1.6577591510709845e-06, "loss": 0.0286, "step": 26650 }, { "epoch": 8.2, "learning_rate": 1.657209674789031e-06, "loss": 0.023, "step": 26651 }, { "epoch": 8.2, "learning_rate": 1.6566602813586752e-06, "loss": 0.0214, "step": 26652 }, { "epoch": 8.2, "learning_rate": 1.6561109707853695e-06, "loss": 0.0221, "step": 26653 }, { "epoch": 8.2, "learning_rate": 1.6555617430745763e-06, "loss": 0.0403, "step": 26654 }, { "epoch": 8.2, "learning_rate": 1.65501259823174e-06, "loss": 0.0175, "step": 26655 }, { "epoch": 8.2, "learning_rate": 1.654463536262323e-06, "loss": 0.0273, "step": 26656 }, { "epoch": 8.2, "learning_rate": 1.653914557171774e-06, "loss": 0.0367, "step": 26657 }, { "epoch": 8.2, "learning_rate": 1.6533656609655456e-06, "loss": 0.0601, "step": 26658 }, { "epoch": 8.2, "learning_rate": 1.652816847649089e-06, "loss": 0.02, "step": 26659 }, { "epoch": 8.2, "learning_rate": 1.652268117227851e-06, "loss": 0.0331, "step": 26660 }, { "epoch": 8.2, "learning_rate": 1.6517194697072903e-06, "loss": 0.0276, "step": 26661 }, { "epoch": 8.2, "learning_rate": 1.651170905092846e-06, "loss": 0.0299, "step": 26662 }, { "epoch": 8.2, "learning_rate": 1.650622423389966e-06, "loss": 0.0325, "step": 26663 }, { "epoch": 8.2, "learning_rate": 1.650074024604106e-06, "loss": 0.0388, "step": 26664 }, { "epoch": 8.2, "learning_rate": 1.6495257087407012e-06, "loss": 0.0263, "step": 26665 }, { "epoch": 8.2, "learning_rate": 1.6489774758052047e-06, "loss": 0.0424, "step": 26666 }, { "epoch": 8.2, "learning_rate": 1.648429325803058e-06, "loss": 0.0152, "step": 26667 }, { "epoch": 8.21, "learning_rate": 1.6478812587397064e-06, "loss": 0.0243, "step": 26668 }, { "epoch": 8.21, "learning_rate": 1.6473332746205906e-06, "loss": 0.0241, "step": 26669 }, { "epoch": 8.21, "learning_rate": 1.6467853734511508e-06, "loss": 0.0517, "step": 26670 }, { "epoch": 8.21, "learning_rate": 1.6462375552368359e-06, "loss": 0.0291, "step": 26671 }, { "epoch": 8.21, "learning_rate": 1.6456898199830785e-06, "loss": 0.0333, "step": 26672 }, { "epoch": 8.21, "learning_rate": 1.6451421676953182e-06, "loss": 0.0635, "step": 26673 }, { "epoch": 8.21, "learning_rate": 1.644594598378999e-06, "loss": 0.0215, "step": 26674 }, { "epoch": 8.21, "learning_rate": 1.6440471120395563e-06, "loss": 0.0289, "step": 26675 }, { "epoch": 8.21, "learning_rate": 1.6434997086824267e-06, "loss": 0.0646, "step": 26676 }, { "epoch": 8.21, "learning_rate": 1.642952388313046e-06, "loss": 0.0269, "step": 26677 }, { "epoch": 8.21, "learning_rate": 1.642405150936851e-06, "loss": 0.0524, "step": 26678 }, { "epoch": 8.21, "learning_rate": 1.641857996559275e-06, "loss": 0.0301, "step": 26679 }, { "epoch": 8.21, "learning_rate": 1.6413109251857494e-06, "loss": 0.0214, "step": 26680 }, { "epoch": 8.21, "learning_rate": 1.6407639368217133e-06, "loss": 0.0226, "step": 26681 }, { "epoch": 8.21, "learning_rate": 1.6402170314725964e-06, "loss": 0.0305, "step": 26682 }, { "epoch": 8.21, "learning_rate": 1.6396702091438255e-06, "loss": 0.0236, "step": 26683 }, { "epoch": 8.21, "learning_rate": 1.6391234698408365e-06, "loss": 0.0377, "step": 26684 }, { "epoch": 8.21, "learning_rate": 1.6385768135690561e-06, "loss": 0.0146, "step": 26685 }, { "epoch": 8.21, "learning_rate": 1.6380302403339155e-06, "loss": 0.0195, "step": 26686 }, { "epoch": 8.21, "learning_rate": 1.6374837501408403e-06, "loss": 0.0282, "step": 26687 }, { "epoch": 8.21, "learning_rate": 1.6369373429952596e-06, "loss": 0.0149, "step": 26688 }, { "epoch": 8.21, "learning_rate": 1.6363910189025978e-06, "loss": 0.0454, "step": 26689 }, { "epoch": 8.21, "learning_rate": 1.6358447778682784e-06, "loss": 0.0207, "step": 26690 }, { "epoch": 8.21, "learning_rate": 1.6352986198977327e-06, "loss": 0.027, "step": 26691 }, { "epoch": 8.21, "learning_rate": 1.634752544996382e-06, "loss": 0.027, "step": 26692 }, { "epoch": 8.21, "learning_rate": 1.634206553169645e-06, "loss": 0.012, "step": 26693 }, { "epoch": 8.21, "learning_rate": 1.6336606444229476e-06, "loss": 0.0267, "step": 26694 }, { "epoch": 8.21, "learning_rate": 1.6331148187617118e-06, "loss": 0.0322, "step": 26695 }, { "epoch": 8.21, "learning_rate": 1.6325690761913571e-06, "loss": 0.037, "step": 26696 }, { "epoch": 8.21, "learning_rate": 1.632023416717302e-06, "loss": 0.023, "step": 26697 }, { "epoch": 8.21, "learning_rate": 1.631477840344965e-06, "loss": 0.0385, "step": 26698 }, { "epoch": 8.21, "learning_rate": 1.630932347079771e-06, "loss": 0.0251, "step": 26699 }, { "epoch": 8.22, "learning_rate": 1.6303869369271264e-06, "loss": 0.0308, "step": 26700 }, { "epoch": 8.22, "learning_rate": 1.6298416098924552e-06, "loss": 0.0344, "step": 26701 }, { "epoch": 8.22, "learning_rate": 1.6292963659811723e-06, "loss": 0.0333, "step": 26702 }, { "epoch": 8.22, "learning_rate": 1.6287512051986897e-06, "loss": 0.0235, "step": 26703 }, { "epoch": 8.22, "learning_rate": 1.6282061275504234e-06, "loss": 0.0222, "step": 26704 }, { "epoch": 8.22, "learning_rate": 1.627661133041787e-06, "loss": 0.0323, "step": 26705 }, { "epoch": 8.22, "learning_rate": 1.62711622167819e-06, "loss": 0.0227, "step": 26706 }, { "epoch": 8.22, "learning_rate": 1.626571393465044e-06, "loss": 0.0201, "step": 26707 }, { "epoch": 8.22, "learning_rate": 1.6260266484077636e-06, "loss": 0.0453, "step": 26708 }, { "epoch": 8.22, "learning_rate": 1.625481986511759e-06, "loss": 0.0631, "step": 26709 }, { "epoch": 8.22, "learning_rate": 1.624937407782431e-06, "loss": 0.0079, "step": 26710 }, { "epoch": 8.22, "learning_rate": 1.6243929122251955e-06, "loss": 0.0285, "step": 26711 }, { "epoch": 8.22, "learning_rate": 1.6238484998454584e-06, "loss": 0.0247, "step": 26712 }, { "epoch": 8.22, "learning_rate": 1.6233041706486253e-06, "loss": 0.032, "step": 26713 }, { "epoch": 8.22, "learning_rate": 1.6227599246401016e-06, "loss": 0.0385, "step": 26714 }, { "epoch": 8.22, "learning_rate": 1.6222157618252898e-06, "loss": 0.0328, "step": 26715 }, { "epoch": 8.22, "learning_rate": 1.6216716822096013e-06, "loss": 0.0343, "step": 26716 }, { "epoch": 8.22, "learning_rate": 1.6211276857984304e-06, "loss": 0.0238, "step": 26717 }, { "epoch": 8.22, "learning_rate": 1.6205837725971862e-06, "loss": 0.0321, "step": 26718 }, { "epoch": 8.22, "learning_rate": 1.6200399426112679e-06, "loss": 0.0387, "step": 26719 }, { "epoch": 8.22, "learning_rate": 1.6194961958460764e-06, "loss": 0.0206, "step": 26720 }, { "epoch": 8.22, "learning_rate": 1.618952532307011e-06, "loss": 0.0154, "step": 26721 }, { "epoch": 8.22, "learning_rate": 1.6184089519994718e-06, "loss": 0.0246, "step": 26722 }, { "epoch": 8.22, "learning_rate": 1.6178654549288553e-06, "loss": 0.0376, "step": 26723 }, { "epoch": 8.22, "learning_rate": 1.6173220411005608e-06, "loss": 0.0238, "step": 26724 }, { "epoch": 8.22, "learning_rate": 1.6167787105199816e-06, "loss": 0.027, "step": 26725 }, { "epoch": 8.22, "learning_rate": 1.6162354631925203e-06, "loss": 0.024, "step": 26726 }, { "epoch": 8.22, "learning_rate": 1.6156922991235634e-06, "loss": 0.0335, "step": 26727 }, { "epoch": 8.22, "learning_rate": 1.6151492183185113e-06, "loss": 0.025, "step": 26728 }, { "epoch": 8.22, "learning_rate": 1.614606220782755e-06, "loss": 0.015, "step": 26729 }, { "epoch": 8.22, "learning_rate": 1.6140633065216882e-06, "loss": 0.042, "step": 26730 }, { "epoch": 8.22, "learning_rate": 1.6135204755407009e-06, "loss": 0.0237, "step": 26731 }, { "epoch": 8.22, "learning_rate": 1.6129777278451807e-06, "loss": 0.0361, "step": 26732 }, { "epoch": 8.23, "learning_rate": 1.6124350634405273e-06, "loss": 0.0243, "step": 26733 }, { "epoch": 8.23, "learning_rate": 1.6118924823321225e-06, "loss": 0.0413, "step": 26734 }, { "epoch": 8.23, "learning_rate": 1.6113499845253523e-06, "loss": 0.0243, "step": 26735 }, { "epoch": 8.23, "learning_rate": 1.610807570025611e-06, "loss": 0.0312, "step": 26736 }, { "epoch": 8.23, "learning_rate": 1.6102652388382812e-06, "loss": 0.0413, "step": 26737 }, { "epoch": 8.23, "learning_rate": 1.6097229909687507e-06, "loss": 0.0251, "step": 26738 }, { "epoch": 8.23, "learning_rate": 1.609180826422404e-06, "loss": 0.0372, "step": 26739 }, { "epoch": 8.23, "learning_rate": 1.6086387452046237e-06, "loss": 0.0378, "step": 26740 }, { "epoch": 8.23, "learning_rate": 1.6080967473207954e-06, "loss": 0.0447, "step": 26741 }, { "epoch": 8.23, "learning_rate": 1.607554832776298e-06, "loss": 0.0282, "step": 26742 }, { "epoch": 8.23, "learning_rate": 1.6070130015765206e-06, "loss": 0.0293, "step": 26743 }, { "epoch": 8.23, "learning_rate": 1.6064712537268369e-06, "loss": 0.041, "step": 26744 }, { "epoch": 8.23, "learning_rate": 1.6059295892326266e-06, "loss": 0.0281, "step": 26745 }, { "epoch": 8.23, "learning_rate": 1.6053880080992746e-06, "loss": 0.0313, "step": 26746 }, { "epoch": 8.23, "learning_rate": 1.6048465103321553e-06, "loss": 0.0194, "step": 26747 }, { "epoch": 8.23, "learning_rate": 1.604305095936648e-06, "loss": 0.0371, "step": 26748 }, { "epoch": 8.23, "learning_rate": 1.6037637649181291e-06, "loss": 0.0272, "step": 26749 }, { "epoch": 8.23, "learning_rate": 1.6032225172819737e-06, "loss": 0.0239, "step": 26750 }, { "epoch": 8.23, "learning_rate": 1.6026813530335583e-06, "loss": 0.0272, "step": 26751 }, { "epoch": 8.23, "learning_rate": 1.602140272178253e-06, "loss": 0.0244, "step": 26752 }, { "epoch": 8.23, "learning_rate": 1.601599274721437e-06, "loss": 0.0268, "step": 26753 }, { "epoch": 8.23, "learning_rate": 1.6010583606684836e-06, "loss": 0.0265, "step": 26754 }, { "epoch": 8.23, "learning_rate": 1.6005175300247554e-06, "loss": 0.0471, "step": 26755 }, { "epoch": 8.23, "learning_rate": 1.5999767827956324e-06, "loss": 0.0283, "step": 26756 }, { "epoch": 8.23, "learning_rate": 1.5994361189864815e-06, "loss": 0.0369, "step": 26757 }, { "epoch": 8.23, "learning_rate": 1.5988955386026728e-06, "loss": 0.0228, "step": 26758 }, { "epoch": 8.23, "learning_rate": 1.598355041649573e-06, "loss": 0.0518, "step": 26759 }, { "epoch": 8.23, "learning_rate": 1.5978146281325491e-06, "loss": 0.0273, "step": 26760 }, { "epoch": 8.23, "learning_rate": 1.5972742980569745e-06, "loss": 0.0228, "step": 26761 }, { "epoch": 8.23, "learning_rate": 1.5967340514282049e-06, "loss": 0.0149, "step": 26762 }, { "epoch": 8.23, "learning_rate": 1.5961938882516136e-06, "loss": 0.0231, "step": 26763 }, { "epoch": 8.23, "learning_rate": 1.5956538085325645e-06, "loss": 0.0324, "step": 26764 }, { "epoch": 8.24, "learning_rate": 1.5951138122764132e-06, "loss": 0.0317, "step": 26765 }, { "epoch": 8.24, "learning_rate": 1.5945738994885318e-06, "loss": 0.0205, "step": 26766 }, { "epoch": 8.24, "learning_rate": 1.5940340701742773e-06, "loss": 0.0218, "step": 26767 }, { "epoch": 8.24, "learning_rate": 1.5934943243390123e-06, "loss": 0.0322, "step": 26768 }, { "epoch": 8.24, "learning_rate": 1.5929546619880953e-06, "loss": 0.0457, "step": 26769 }, { "epoch": 8.24, "learning_rate": 1.5924150831268847e-06, "loss": 0.0196, "step": 26770 }, { "epoch": 8.24, "learning_rate": 1.5918755877607461e-06, "loss": 0.016, "step": 26771 }, { "epoch": 8.24, "learning_rate": 1.5913361758950275e-06, "loss": 0.0287, "step": 26772 }, { "epoch": 8.24, "learning_rate": 1.590796847535092e-06, "loss": 0.0253, "step": 26773 }, { "epoch": 8.24, "learning_rate": 1.5902576026862949e-06, "loss": 0.0388, "step": 26774 }, { "epoch": 8.24, "learning_rate": 1.589718441353989e-06, "loss": 0.0389, "step": 26775 }, { "epoch": 8.24, "learning_rate": 1.5891793635435315e-06, "loss": 0.0341, "step": 26776 }, { "epoch": 8.24, "learning_rate": 1.5886403692602713e-06, "loss": 0.0226, "step": 26777 }, { "epoch": 8.24, "learning_rate": 1.58810145850957e-06, "loss": 0.0332, "step": 26778 }, { "epoch": 8.24, "learning_rate": 1.5875626312967706e-06, "loss": 0.0316, "step": 26779 }, { "epoch": 8.24, "learning_rate": 1.5870238876272249e-06, "loss": 0.0587, "step": 26780 }, { "epoch": 8.24, "learning_rate": 1.5864852275062892e-06, "loss": 0.0271, "step": 26781 }, { "epoch": 8.24, "learning_rate": 1.5859466509393085e-06, "loss": 0.0523, "step": 26782 }, { "epoch": 8.24, "learning_rate": 1.5854081579316316e-06, "loss": 0.0398, "step": 26783 }, { "epoch": 8.24, "learning_rate": 1.5848697484886078e-06, "loss": 0.0344, "step": 26784 }, { "epoch": 8.24, "learning_rate": 1.5843314226155825e-06, "loss": 0.0465, "step": 26785 }, { "epoch": 8.24, "learning_rate": 1.583793180317903e-06, "loss": 0.021, "step": 26786 }, { "epoch": 8.24, "learning_rate": 1.58325502160091e-06, "loss": 0.0186, "step": 26787 }, { "epoch": 8.24, "learning_rate": 1.5827169464699576e-06, "loss": 0.0212, "step": 26788 }, { "epoch": 8.24, "learning_rate": 1.5821789549303812e-06, "loss": 0.0261, "step": 26789 }, { "epoch": 8.24, "learning_rate": 1.5816410469875232e-06, "loss": 0.0229, "step": 26790 }, { "epoch": 8.24, "learning_rate": 1.5811032226467304e-06, "loss": 0.0189, "step": 26791 }, { "epoch": 8.24, "learning_rate": 1.580565481913342e-06, "loss": 0.0157, "step": 26792 }, { "epoch": 8.24, "learning_rate": 1.5800278247926982e-06, "loss": 0.0339, "step": 26793 }, { "epoch": 8.24, "learning_rate": 1.5794902512901378e-06, "loss": 0.029, "step": 26794 }, { "epoch": 8.24, "learning_rate": 1.578952761410999e-06, "loss": 0.0499, "step": 26795 }, { "epoch": 8.24, "learning_rate": 1.578415355160622e-06, "loss": 0.038, "step": 26796 }, { "epoch": 8.24, "learning_rate": 1.5778780325443378e-06, "loss": 0.0487, "step": 26797 }, { "epoch": 8.25, "learning_rate": 1.577340793567491e-06, "loss": 0.0277, "step": 26798 }, { "epoch": 8.25, "learning_rate": 1.5768036382354135e-06, "loss": 0.0281, "step": 26799 }, { "epoch": 8.25, "learning_rate": 1.576266566553435e-06, "loss": 0.0247, "step": 26800 }, { "epoch": 8.25, "learning_rate": 1.5757295785268955e-06, "loss": 0.04, "step": 26801 }, { "epoch": 8.25, "learning_rate": 1.5751926741611256e-06, "loss": 0.0202, "step": 26802 }, { "epoch": 8.25, "learning_rate": 1.5746558534614565e-06, "loss": 0.0239, "step": 26803 }, { "epoch": 8.25, "learning_rate": 1.5741191164332192e-06, "loss": 0.0279, "step": 26804 }, { "epoch": 8.25, "learning_rate": 1.5735824630817453e-06, "loss": 0.0308, "step": 26805 }, { "epoch": 8.25, "learning_rate": 1.5730458934123637e-06, "loss": 0.0251, "step": 26806 }, { "epoch": 8.25, "learning_rate": 1.5725094074304004e-06, "loss": 0.0281, "step": 26807 }, { "epoch": 8.25, "learning_rate": 1.5719730051411874e-06, "loss": 0.0242, "step": 26808 }, { "epoch": 8.25, "learning_rate": 1.571436686550053e-06, "loss": 0.05, "step": 26809 }, { "epoch": 8.25, "learning_rate": 1.5709004516623161e-06, "loss": 0.0421, "step": 26810 }, { "epoch": 8.25, "learning_rate": 1.570364300483307e-06, "loss": 0.0515, "step": 26811 }, { "epoch": 8.25, "learning_rate": 1.5698282330183512e-06, "loss": 0.0351, "step": 26812 }, { "epoch": 8.25, "learning_rate": 1.569292249272769e-06, "loss": 0.0378, "step": 26813 }, { "epoch": 8.25, "learning_rate": 1.5687563492518854e-06, "loss": 0.03, "step": 26814 }, { "epoch": 8.25, "learning_rate": 1.568220532961019e-06, "loss": 0.0242, "step": 26815 }, { "epoch": 8.25, "learning_rate": 1.567684800405499e-06, "loss": 0.0377, "step": 26816 }, { "epoch": 8.25, "learning_rate": 1.5671491515906355e-06, "loss": 0.016, "step": 26817 }, { "epoch": 8.25, "learning_rate": 1.5666135865217546e-06, "loss": 0.0254, "step": 26818 }, { "epoch": 8.25, "learning_rate": 1.5660781052041741e-06, "loss": 0.0234, "step": 26819 }, { "epoch": 8.25, "learning_rate": 1.5655427076432107e-06, "loss": 0.0225, "step": 26820 }, { "epoch": 8.25, "learning_rate": 1.5650073938441813e-06, "loss": 0.0231, "step": 26821 }, { "epoch": 8.25, "learning_rate": 1.5644721638124028e-06, "loss": 0.0224, "step": 26822 }, { "epoch": 8.25, "learning_rate": 1.56393701755319e-06, "loss": 0.0333, "step": 26823 }, { "epoch": 8.25, "learning_rate": 1.5634019550718571e-06, "loss": 0.02, "step": 26824 }, { "epoch": 8.25, "learning_rate": 1.562866976373716e-06, "loss": 0.0327, "step": 26825 }, { "epoch": 8.25, "learning_rate": 1.5623320814640862e-06, "loss": 0.0348, "step": 26826 }, { "epoch": 8.25, "learning_rate": 1.5617972703482708e-06, "loss": 0.0161, "step": 26827 }, { "epoch": 8.25, "learning_rate": 1.5612625430315875e-06, "loss": 0.0315, "step": 26828 }, { "epoch": 8.25, "learning_rate": 1.5607278995193454e-06, "loss": 0.0204, "step": 26829 }, { "epoch": 8.26, "learning_rate": 1.5601933398168523e-06, "loss": 0.0102, "step": 26830 }, { "epoch": 8.26, "learning_rate": 1.5596588639294186e-06, "loss": 0.0275, "step": 26831 }, { "epoch": 8.26, "learning_rate": 1.5591244718623477e-06, "loss": 0.041, "step": 26832 }, { "epoch": 8.26, "learning_rate": 1.5585901636209554e-06, "loss": 0.019, "step": 26833 }, { "epoch": 8.26, "learning_rate": 1.5580559392105399e-06, "loss": 0.0398, "step": 26834 }, { "epoch": 8.26, "learning_rate": 1.5575217986364078e-06, "loss": 0.0287, "step": 26835 }, { "epoch": 8.26, "learning_rate": 1.5569877419038659e-06, "loss": 0.04, "step": 26836 }, { "epoch": 8.26, "learning_rate": 1.556453769018218e-06, "loss": 0.0262, "step": 26837 }, { "epoch": 8.26, "learning_rate": 1.5559198799847664e-06, "loss": 0.0586, "step": 26838 }, { "epoch": 8.26, "learning_rate": 1.5553860748088123e-06, "loss": 0.032, "step": 26839 }, { "epoch": 8.26, "learning_rate": 1.5548523534956572e-06, "loss": 0.0193, "step": 26840 }, { "epoch": 8.26, "learning_rate": 1.5543187160506013e-06, "loss": 0.0355, "step": 26841 }, { "epoch": 8.26, "learning_rate": 1.5537851624789423e-06, "loss": 0.0231, "step": 26842 }, { "epoch": 8.26, "learning_rate": 1.5532516927859853e-06, "loss": 0.0328, "step": 26843 }, { "epoch": 8.26, "learning_rate": 1.5527183069770213e-06, "loss": 0.0397, "step": 26844 }, { "epoch": 8.26, "learning_rate": 1.5521850050573484e-06, "loss": 0.0391, "step": 26845 }, { "epoch": 8.26, "learning_rate": 1.5516517870322656e-06, "loss": 0.0324, "step": 26846 }, { "epoch": 8.26, "learning_rate": 1.5511186529070666e-06, "loss": 0.0226, "step": 26847 }, { "epoch": 8.26, "learning_rate": 1.5505856026870459e-06, "loss": 0.0317, "step": 26848 }, { "epoch": 8.26, "learning_rate": 1.5500526363774981e-06, "loss": 0.0491, "step": 26849 }, { "epoch": 8.26, "learning_rate": 1.5495197539837148e-06, "loss": 0.0204, "step": 26850 }, { "epoch": 8.26, "learning_rate": 1.5489869555109893e-06, "loss": 0.0283, "step": 26851 }, { "epoch": 8.26, "learning_rate": 1.5484542409646086e-06, "loss": 0.0277, "step": 26852 }, { "epoch": 8.26, "learning_rate": 1.5479216103498696e-06, "loss": 0.0293, "step": 26853 }, { "epoch": 8.26, "learning_rate": 1.5473890636720612e-06, "loss": 0.031, "step": 26854 }, { "epoch": 8.26, "learning_rate": 1.5468566009364649e-06, "loss": 0.0286, "step": 26855 }, { "epoch": 8.26, "learning_rate": 1.5463242221483742e-06, "loss": 0.0214, "step": 26856 }, { "epoch": 8.26, "learning_rate": 1.5457919273130762e-06, "loss": 0.0285, "step": 26857 }, { "epoch": 8.26, "learning_rate": 1.5452597164358563e-06, "loss": 0.0431, "step": 26858 }, { "epoch": 8.26, "learning_rate": 1.5447275895219993e-06, "loss": 0.0272, "step": 26859 }, { "epoch": 8.26, "learning_rate": 1.5441955465767866e-06, "loss": 0.0189, "step": 26860 }, { "epoch": 8.26, "learning_rate": 1.5436635876055107e-06, "loss": 0.0506, "step": 26861 }, { "epoch": 8.26, "learning_rate": 1.543131712613445e-06, "loss": 0.031, "step": 26862 }, { "epoch": 8.27, "learning_rate": 1.5425999216058762e-06, "loss": 0.0542, "step": 26863 }, { "epoch": 8.27, "learning_rate": 1.5420682145880883e-06, "loss": 0.0452, "step": 26864 }, { "epoch": 8.27, "learning_rate": 1.5415365915653535e-06, "loss": 0.0476, "step": 26865 }, { "epoch": 8.27, "learning_rate": 1.5410050525429577e-06, "loss": 0.0269, "step": 26866 }, { "epoch": 8.27, "learning_rate": 1.5404735975261775e-06, "loss": 0.0271, "step": 26867 }, { "epoch": 8.27, "learning_rate": 1.5399422265202912e-06, "loss": 0.0147, "step": 26868 }, { "epoch": 8.27, "learning_rate": 1.5394109395305757e-06, "loss": 0.0151, "step": 26869 }, { "epoch": 8.27, "learning_rate": 1.5388797365623053e-06, "loss": 0.0342, "step": 26870 }, { "epoch": 8.27, "learning_rate": 1.5383486176207606e-06, "loss": 0.0233, "step": 26871 }, { "epoch": 8.27, "learning_rate": 1.5378175827112097e-06, "loss": 0.0221, "step": 26872 }, { "epoch": 8.27, "learning_rate": 1.5372866318389302e-06, "loss": 0.0216, "step": 26873 }, { "epoch": 8.27, "learning_rate": 1.5367557650091935e-06, "loss": 0.0204, "step": 26874 }, { "epoch": 8.27, "learning_rate": 1.5362249822272735e-06, "loss": 0.0204, "step": 26875 }, { "epoch": 8.27, "learning_rate": 1.535694283498439e-06, "loss": 0.0264, "step": 26876 }, { "epoch": 8.27, "learning_rate": 1.535163668827958e-06, "loss": 0.0225, "step": 26877 }, { "epoch": 8.27, "learning_rate": 1.53463313822111e-06, "loss": 0.0415, "step": 26878 }, { "epoch": 8.27, "learning_rate": 1.5341026916831526e-06, "loss": 0.0157, "step": 26879 }, { "epoch": 8.27, "learning_rate": 1.533572329219356e-06, "loss": 0.0388, "step": 26880 }, { "epoch": 8.27, "learning_rate": 1.533042050834992e-06, "loss": 0.0222, "step": 26881 }, { "epoch": 8.27, "learning_rate": 1.5325118565353237e-06, "loss": 0.0216, "step": 26882 }, { "epoch": 8.27, "learning_rate": 1.531981746325616e-06, "loss": 0.0322, "step": 26883 }, { "epoch": 8.27, "learning_rate": 1.5314517202111345e-06, "loss": 0.022, "step": 26884 }, { "epoch": 8.27, "learning_rate": 1.5309217781971419e-06, "loss": 0.0317, "step": 26885 }, { "epoch": 8.27, "learning_rate": 1.5303919202889006e-06, "loss": 0.037, "step": 26886 }, { "epoch": 8.27, "learning_rate": 1.5298621464916718e-06, "loss": 0.0203, "step": 26887 }, { "epoch": 8.27, "learning_rate": 1.5293324568107237e-06, "loss": 0.0213, "step": 26888 }, { "epoch": 8.27, "learning_rate": 1.5288028512513076e-06, "loss": 0.0342, "step": 26889 }, { "epoch": 8.27, "learning_rate": 1.5282733298186846e-06, "loss": 0.0397, "step": 26890 }, { "epoch": 8.27, "learning_rate": 1.5277438925181175e-06, "loss": 0.0401, "step": 26891 }, { "epoch": 8.27, "learning_rate": 1.527214539354862e-06, "loss": 0.0216, "step": 26892 }, { "epoch": 8.27, "learning_rate": 1.5266852703341738e-06, "loss": 0.043, "step": 26893 }, { "epoch": 8.27, "learning_rate": 1.526156085461311e-06, "loss": 0.0241, "step": 26894 }, { "epoch": 8.28, "learning_rate": 1.5256269847415283e-06, "loss": 0.0311, "step": 26895 }, { "epoch": 8.28, "learning_rate": 1.5250979681800793e-06, "loss": 0.0315, "step": 26896 }, { "epoch": 8.28, "learning_rate": 1.5245690357822163e-06, "loss": 0.0393, "step": 26897 }, { "epoch": 8.28, "learning_rate": 1.5240401875531963e-06, "loss": 0.0244, "step": 26898 }, { "epoch": 8.28, "learning_rate": 1.5235114234982683e-06, "loss": 0.019, "step": 26899 }, { "epoch": 8.28, "learning_rate": 1.522982743622684e-06, "loss": 0.0193, "step": 26900 }, { "epoch": 8.28, "learning_rate": 1.5224541479316946e-06, "loss": 0.0288, "step": 26901 }, { "epoch": 8.28, "learning_rate": 1.521925636430548e-06, "loss": 0.0379, "step": 26902 }, { "epoch": 8.28, "learning_rate": 1.5213972091244932e-06, "loss": 0.037, "step": 26903 }, { "epoch": 8.28, "learning_rate": 1.5208688660187764e-06, "loss": 0.0307, "step": 26904 }, { "epoch": 8.28, "learning_rate": 1.520340607118651e-06, "loss": 0.0328, "step": 26905 }, { "epoch": 8.28, "learning_rate": 1.5198124324293562e-06, "loss": 0.0447, "step": 26906 }, { "epoch": 8.28, "learning_rate": 1.5192843419561366e-06, "loss": 0.0118, "step": 26907 }, { "epoch": 8.28, "learning_rate": 1.5187563357042423e-06, "loss": 0.017, "step": 26908 }, { "epoch": 8.28, "learning_rate": 1.5182284136789149e-06, "loss": 0.0356, "step": 26909 }, { "epoch": 8.28, "learning_rate": 1.5177005758853946e-06, "loss": 0.0511, "step": 26910 }, { "epoch": 8.28, "learning_rate": 1.5171728223289261e-06, "loss": 0.0201, "step": 26911 }, { "epoch": 8.28, "learning_rate": 1.5166451530147496e-06, "loss": 0.0248, "step": 26912 }, { "epoch": 8.28, "learning_rate": 1.5161175679481054e-06, "loss": 0.0261, "step": 26913 }, { "epoch": 8.28, "learning_rate": 1.5155900671342306e-06, "loss": 0.0471, "step": 26914 }, { "epoch": 8.28, "learning_rate": 1.5150626505783672e-06, "loss": 0.0181, "step": 26915 }, { "epoch": 8.28, "learning_rate": 1.5145353182857547e-06, "loss": 0.0357, "step": 26916 }, { "epoch": 8.28, "learning_rate": 1.5140080702616223e-06, "loss": 0.0334, "step": 26917 }, { "epoch": 8.28, "learning_rate": 1.513480906511212e-06, "loss": 0.0265, "step": 26918 }, { "epoch": 8.28, "learning_rate": 1.5129538270397592e-06, "loss": 0.0268, "step": 26919 }, { "epoch": 8.28, "learning_rate": 1.512426831852496e-06, "loss": 0.0387, "step": 26920 }, { "epoch": 8.28, "learning_rate": 1.511899920954656e-06, "loss": 0.1435, "step": 26921 }, { "epoch": 8.28, "learning_rate": 1.5113730943514736e-06, "loss": 0.0381, "step": 26922 }, { "epoch": 8.28, "learning_rate": 1.5108463520481797e-06, "loss": 0.0208, "step": 26923 }, { "epoch": 8.28, "learning_rate": 1.5103196940500031e-06, "loss": 0.0386, "step": 26924 }, { "epoch": 8.28, "learning_rate": 1.5097931203621785e-06, "loss": 0.0434, "step": 26925 }, { "epoch": 8.28, "learning_rate": 1.509266630989935e-06, "loss": 0.033, "step": 26926 }, { "epoch": 8.28, "learning_rate": 1.5087402259384943e-06, "loss": 0.0375, "step": 26927 }, { "epoch": 8.29, "learning_rate": 1.5082139052130918e-06, "loss": 0.0255, "step": 26928 }, { "epoch": 8.29, "learning_rate": 1.5076876688189512e-06, "loss": 0.0169, "step": 26929 }, { "epoch": 8.29, "learning_rate": 1.5071615167612996e-06, "loss": 0.0271, "step": 26930 }, { "epoch": 8.29, "learning_rate": 1.5066354490453606e-06, "loss": 0.0304, "step": 26931 }, { "epoch": 8.29, "learning_rate": 1.5061094656763564e-06, "loss": 0.0458, "step": 26932 }, { "epoch": 8.29, "learning_rate": 1.5055835666595186e-06, "loss": 0.0344, "step": 26933 }, { "epoch": 8.29, "learning_rate": 1.5050577520000608e-06, "loss": 0.0279, "step": 26934 }, { "epoch": 8.29, "learning_rate": 1.50453202170321e-06, "loss": 0.035, "step": 26935 }, { "epoch": 8.29, "learning_rate": 1.5040063757741863e-06, "loss": 0.0291, "step": 26936 }, { "epoch": 8.29, "learning_rate": 1.50348081421821e-06, "loss": 0.0129, "step": 26937 }, { "epoch": 8.29, "learning_rate": 1.5029553370404993e-06, "loss": 0.0206, "step": 26938 }, { "epoch": 8.29, "learning_rate": 1.5024299442462697e-06, "loss": 0.0431, "step": 26939 }, { "epoch": 8.29, "learning_rate": 1.5019046358407485e-06, "loss": 0.0341, "step": 26940 }, { "epoch": 8.29, "learning_rate": 1.5013794118291435e-06, "loss": 0.0158, "step": 26941 }, { "epoch": 8.29, "learning_rate": 1.500854272216672e-06, "loss": 0.0286, "step": 26942 }, { "epoch": 8.29, "learning_rate": 1.5003292170085538e-06, "loss": 0.0436, "step": 26943 }, { "epoch": 8.29, "learning_rate": 1.499804246209996e-06, "loss": 0.028, "step": 26944 }, { "epoch": 8.29, "learning_rate": 1.499279359826219e-06, "loss": 0.0208, "step": 26945 }, { "epoch": 8.29, "learning_rate": 1.4987545578624318e-06, "loss": 0.0263, "step": 26946 }, { "epoch": 8.29, "learning_rate": 1.498229840323847e-06, "loss": 0.0235, "step": 26947 }, { "epoch": 8.29, "learning_rate": 1.497705207215675e-06, "loss": 0.0359, "step": 26948 }, { "epoch": 8.29, "learning_rate": 1.4971806585431247e-06, "loss": 0.0355, "step": 26949 }, { "epoch": 8.29, "learning_rate": 1.4966561943114112e-06, "loss": 0.0302, "step": 26950 }, { "epoch": 8.29, "learning_rate": 1.4961318145257353e-06, "loss": 0.0402, "step": 26951 }, { "epoch": 8.29, "learning_rate": 1.4956075191913066e-06, "loss": 0.0122, "step": 26952 }, { "epoch": 8.29, "learning_rate": 1.4950833083133354e-06, "loss": 0.0398, "step": 26953 }, { "epoch": 8.29, "learning_rate": 1.4945591818970252e-06, "loss": 0.0311, "step": 26954 }, { "epoch": 8.29, "learning_rate": 1.4940351399475805e-06, "loss": 0.0248, "step": 26955 }, { "epoch": 8.29, "learning_rate": 1.4935111824702076e-06, "loss": 0.0336, "step": 26956 }, { "epoch": 8.29, "learning_rate": 1.4929873094701075e-06, "loss": 0.0347, "step": 26957 }, { "epoch": 8.29, "learning_rate": 1.492463520952484e-06, "loss": 0.0168, "step": 26958 }, { "epoch": 8.29, "learning_rate": 1.4919398169225352e-06, "loss": 0.0204, "step": 26959 }, { "epoch": 8.3, "learning_rate": 1.4914161973854714e-06, "loss": 0.0384, "step": 26960 }, { "epoch": 8.3, "learning_rate": 1.4908926623464838e-06, "loss": 0.0431, "step": 26961 }, { "epoch": 8.3, "learning_rate": 1.4903692118107716e-06, "loss": 0.0426, "step": 26962 }, { "epoch": 8.3, "learning_rate": 1.4898458457835386e-06, "loss": 0.0298, "step": 26963 }, { "epoch": 8.3, "learning_rate": 1.4893225642699782e-06, "loss": 0.0296, "step": 26964 }, { "epoch": 8.3, "learning_rate": 1.48879936727529e-06, "loss": 0.0182, "step": 26965 }, { "epoch": 8.3, "learning_rate": 1.488276254804668e-06, "loss": 0.0316, "step": 26966 }, { "epoch": 8.3, "learning_rate": 1.4877532268633065e-06, "loss": 0.0206, "step": 26967 }, { "epoch": 8.3, "learning_rate": 1.4872302834564012e-06, "loss": 0.0221, "step": 26968 }, { "epoch": 8.3, "learning_rate": 1.4867074245891423e-06, "loss": 0.0264, "step": 26969 }, { "epoch": 8.3, "learning_rate": 1.4861846502667276e-06, "loss": 0.0542, "step": 26970 }, { "epoch": 8.3, "learning_rate": 1.4856619604943478e-06, "loss": 0.0352, "step": 26971 }, { "epoch": 8.3, "learning_rate": 1.4851393552771864e-06, "loss": 0.0404, "step": 26972 }, { "epoch": 8.3, "learning_rate": 1.4846168346204425e-06, "loss": 0.0338, "step": 26973 }, { "epoch": 8.3, "learning_rate": 1.484094398529301e-06, "loss": 0.0228, "step": 26974 }, { "epoch": 8.3, "learning_rate": 1.4835720470089498e-06, "loss": 0.0161, "step": 26975 }, { "epoch": 8.3, "learning_rate": 1.4830497800645782e-06, "loss": 0.0314, "step": 26976 }, { "epoch": 8.3, "learning_rate": 1.4825275977013687e-06, "loss": 0.0598, "step": 26977 }, { "epoch": 8.3, "learning_rate": 1.4820054999245148e-06, "loss": 0.0294, "step": 26978 }, { "epoch": 8.3, "learning_rate": 1.4814834867391925e-06, "loss": 0.0459, "step": 26979 }, { "epoch": 8.3, "learning_rate": 1.480961558150592e-06, "loss": 0.0274, "step": 26980 }, { "epoch": 8.3, "learning_rate": 1.4804397141638972e-06, "loss": 0.0331, "step": 26981 }, { "epoch": 8.3, "learning_rate": 1.4799179547842823e-06, "loss": 0.0256, "step": 26982 }, { "epoch": 8.3, "learning_rate": 1.4793962800169381e-06, "loss": 0.021, "step": 26983 }, { "epoch": 8.3, "learning_rate": 1.4788746898670415e-06, "loss": 0.0157, "step": 26984 }, { "epoch": 8.3, "learning_rate": 1.4783531843397713e-06, "loss": 0.018, "step": 26985 }, { "epoch": 8.3, "learning_rate": 1.4778317634403082e-06, "loss": 0.0342, "step": 26986 }, { "epoch": 8.3, "learning_rate": 1.4773104271738282e-06, "loss": 0.0331, "step": 26987 }, { "epoch": 8.3, "learning_rate": 1.4767891755455155e-06, "loss": 0.0294, "step": 26988 }, { "epoch": 8.3, "learning_rate": 1.4762680085605363e-06, "loss": 0.0285, "step": 26989 }, { "epoch": 8.3, "learning_rate": 1.4757469262240731e-06, "loss": 0.0414, "step": 26990 }, { "epoch": 8.3, "learning_rate": 1.4752259285412995e-06, "loss": 0.0359, "step": 26991 }, { "epoch": 8.3, "learning_rate": 1.4747050155173903e-06, "loss": 0.0496, "step": 26992 }, { "epoch": 8.31, "learning_rate": 1.4741841871575159e-06, "loss": 0.0298, "step": 26993 }, { "epoch": 8.31, "learning_rate": 1.4736634434668484e-06, "loss": 0.0294, "step": 26994 }, { "epoch": 8.31, "learning_rate": 1.4731427844505653e-06, "loss": 0.0218, "step": 26995 }, { "epoch": 8.31, "learning_rate": 1.472622210113831e-06, "loss": 0.0401, "step": 26996 }, { "epoch": 8.31, "learning_rate": 1.4721017204618148e-06, "loss": 0.0234, "step": 26997 }, { "epoch": 8.31, "learning_rate": 1.4715813154996905e-06, "loss": 0.0177, "step": 26998 }, { "epoch": 8.31, "learning_rate": 1.4710609952326239e-06, "loss": 0.031, "step": 26999 }, { "epoch": 8.31, "learning_rate": 1.4705407596657818e-06, "loss": 0.0269, "step": 27000 }, { "epoch": 8.31, "learning_rate": 1.4700206088043312e-06, "loss": 0.0388, "step": 27001 }, { "epoch": 8.31, "learning_rate": 1.469500542653437e-06, "loss": 0.0339, "step": 27002 }, { "epoch": 8.31, "learning_rate": 1.468980561218265e-06, "loss": 0.0319, "step": 27003 }, { "epoch": 8.31, "learning_rate": 1.4684606645039755e-06, "loss": 0.0321, "step": 27004 }, { "epoch": 8.31, "learning_rate": 1.46794085251574e-06, "loss": 0.0265, "step": 27005 }, { "epoch": 8.31, "learning_rate": 1.4674211252587122e-06, "loss": 0.031, "step": 27006 }, { "epoch": 8.31, "learning_rate": 1.4669014827380534e-06, "loss": 0.0242, "step": 27007 }, { "epoch": 8.31, "learning_rate": 1.4663819249589306e-06, "loss": 0.0471, "step": 27008 }, { "epoch": 8.31, "learning_rate": 1.4658624519264997e-06, "loss": 0.027, "step": 27009 }, { "epoch": 8.31, "learning_rate": 1.4653430636459209e-06, "loss": 0.0297, "step": 27010 }, { "epoch": 8.31, "learning_rate": 1.4648237601223502e-06, "loss": 0.0209, "step": 27011 }, { "epoch": 8.31, "learning_rate": 1.464304541360946e-06, "loss": 0.0282, "step": 27012 }, { "epoch": 8.31, "learning_rate": 1.4637854073668645e-06, "loss": 0.0318, "step": 27013 }, { "epoch": 8.31, "learning_rate": 1.463266358145259e-06, "loss": 0.0224, "step": 27014 }, { "epoch": 8.31, "learning_rate": 1.462747393701288e-06, "loss": 0.0321, "step": 27015 }, { "epoch": 8.31, "learning_rate": 1.462228514040106e-06, "loss": 0.042, "step": 27016 }, { "epoch": 8.31, "learning_rate": 1.4617097191668595e-06, "loss": 0.0141, "step": 27017 }, { "epoch": 8.31, "learning_rate": 1.4611910090867065e-06, "loss": 0.0244, "step": 27018 }, { "epoch": 8.31, "learning_rate": 1.4606723838047964e-06, "loss": 0.0175, "step": 27019 }, { "epoch": 8.31, "learning_rate": 1.4601538433262796e-06, "loss": 0.0269, "step": 27020 }, { "epoch": 8.31, "learning_rate": 1.4596353876563052e-06, "loss": 0.016, "step": 27021 }, { "epoch": 8.31, "learning_rate": 1.4591170168000235e-06, "loss": 0.0522, "step": 27022 }, { "epoch": 8.31, "learning_rate": 1.4585987307625803e-06, "loss": 0.0268, "step": 27023 }, { "epoch": 8.31, "learning_rate": 1.4580805295491218e-06, "loss": 0.0096, "step": 27024 }, { "epoch": 8.32, "learning_rate": 1.457562413164799e-06, "loss": 0.0247, "step": 27025 }, { "epoch": 8.32, "learning_rate": 1.4570443816147561e-06, "loss": 0.0258, "step": 27026 }, { "epoch": 8.32, "learning_rate": 1.4565264349041307e-06, "loss": 0.0224, "step": 27027 }, { "epoch": 8.32, "learning_rate": 1.4560085730380745e-06, "loss": 0.0194, "step": 27028 }, { "epoch": 8.32, "learning_rate": 1.455490796021728e-06, "loss": 0.0307, "step": 27029 }, { "epoch": 8.32, "learning_rate": 1.4549731038602323e-06, "loss": 0.0262, "step": 27030 }, { "epoch": 8.32, "learning_rate": 1.45445549655873e-06, "loss": 0.0195, "step": 27031 }, { "epoch": 8.32, "learning_rate": 1.453937974122357e-06, "loss": 0.0096, "step": 27032 }, { "epoch": 8.32, "learning_rate": 1.4534205365562604e-06, "loss": 0.0274, "step": 27033 }, { "epoch": 8.32, "learning_rate": 1.4529031838655706e-06, "loss": 0.0352, "step": 27034 }, { "epoch": 8.32, "learning_rate": 1.4523859160554322e-06, "loss": 0.0208, "step": 27035 }, { "epoch": 8.32, "learning_rate": 1.4518687331309788e-06, "loss": 0.04, "step": 27036 }, { "epoch": 8.32, "learning_rate": 1.4513516350973478e-06, "loss": 0.0354, "step": 27037 }, { "epoch": 8.32, "learning_rate": 1.4508346219596725e-06, "loss": 0.0332, "step": 27038 }, { "epoch": 8.32, "learning_rate": 1.45031769372309e-06, "loss": 0.0352, "step": 27039 }, { "epoch": 8.32, "learning_rate": 1.4498008503927308e-06, "loss": 0.0313, "step": 27040 }, { "epoch": 8.32, "learning_rate": 1.4492840919737305e-06, "loss": 0.0348, "step": 27041 }, { "epoch": 8.32, "learning_rate": 1.4487674184712175e-06, "loss": 0.035, "step": 27042 }, { "epoch": 8.32, "learning_rate": 1.4482508298903286e-06, "loss": 0.0421, "step": 27043 }, { "epoch": 8.32, "learning_rate": 1.4477343262361864e-06, "loss": 0.0176, "step": 27044 }, { "epoch": 8.32, "learning_rate": 1.447217907513927e-06, "loss": 0.0213, "step": 27045 }, { "epoch": 8.32, "learning_rate": 1.4467015737286761e-06, "loss": 0.0288, "step": 27046 }, { "epoch": 8.32, "learning_rate": 1.4461853248855617e-06, "loss": 0.0233, "step": 27047 }, { "epoch": 8.32, "learning_rate": 1.44566916098971e-06, "loss": 0.0347, "step": 27048 }, { "epoch": 8.32, "learning_rate": 1.4451530820462444e-06, "loss": 0.0488, "step": 27049 }, { "epoch": 8.32, "learning_rate": 1.444637088060299e-06, "loss": 0.0378, "step": 27050 }, { "epoch": 8.32, "learning_rate": 1.4441211790369892e-06, "loss": 0.0434, "step": 27051 }, { "epoch": 8.32, "learning_rate": 1.44360535498144e-06, "loss": 0.0255, "step": 27052 }, { "epoch": 8.32, "learning_rate": 1.4430896158987773e-06, "loss": 0.0382, "step": 27053 }, { "epoch": 8.32, "learning_rate": 1.4425739617941203e-06, "loss": 0.0503, "step": 27054 }, { "epoch": 8.32, "learning_rate": 1.4420583926725918e-06, "loss": 0.0435, "step": 27055 }, { "epoch": 8.32, "learning_rate": 1.4415429085393096e-06, "loss": 0.0288, "step": 27056 }, { "epoch": 8.32, "learning_rate": 1.4410275093993942e-06, "loss": 0.0314, "step": 27057 }, { "epoch": 8.33, "learning_rate": 1.4405121952579638e-06, "loss": 0.0138, "step": 27058 }, { "epoch": 8.33, "learning_rate": 1.439996966120134e-06, "loss": 0.0269, "step": 27059 }, { "epoch": 8.33, "learning_rate": 1.4394818219910279e-06, "loss": 0.0283, "step": 27060 }, { "epoch": 8.33, "learning_rate": 1.4389667628757552e-06, "loss": 0.0219, "step": 27061 }, { "epoch": 8.33, "learning_rate": 1.4384517887794292e-06, "loss": 0.0161, "step": 27062 }, { "epoch": 8.33, "learning_rate": 1.4379368997071708e-06, "loss": 0.0535, "step": 27063 }, { "epoch": 8.33, "learning_rate": 1.4374220956640895e-06, "loss": 0.0298, "step": 27064 }, { "epoch": 8.33, "learning_rate": 1.4369073766552988e-06, "loss": 0.0613, "step": 27065 }, { "epoch": 8.33, "learning_rate": 1.4363927426859091e-06, "loss": 0.0284, "step": 27066 }, { "epoch": 8.33, "learning_rate": 1.4358781937610323e-06, "loss": 0.0302, "step": 27067 }, { "epoch": 8.33, "learning_rate": 1.435363729885778e-06, "loss": 0.03, "step": 27068 }, { "epoch": 8.33, "learning_rate": 1.4348493510652529e-06, "loss": 0.0172, "step": 27069 }, { "epoch": 8.33, "learning_rate": 1.43433505730457e-06, "loss": 0.0249, "step": 27070 }, { "epoch": 8.33, "learning_rate": 1.4338208486088356e-06, "loss": 0.0324, "step": 27071 }, { "epoch": 8.33, "learning_rate": 1.4333067249831523e-06, "loss": 0.0247, "step": 27072 }, { "epoch": 8.33, "learning_rate": 1.432792686432629e-06, "loss": 0.0355, "step": 27073 }, { "epoch": 8.33, "learning_rate": 1.432278732962371e-06, "loss": 0.0189, "step": 27074 }, { "epoch": 8.33, "learning_rate": 1.4317648645774817e-06, "loss": 0.0322, "step": 27075 }, { "epoch": 8.33, "learning_rate": 1.4312510812830639e-06, "loss": 0.0385, "step": 27076 }, { "epoch": 8.33, "learning_rate": 1.4307373830842174e-06, "loss": 0.0291, "step": 27077 }, { "epoch": 8.33, "learning_rate": 1.430223769986051e-06, "loss": 0.0295, "step": 27078 }, { "epoch": 8.33, "learning_rate": 1.4297102419936559e-06, "loss": 0.0244, "step": 27079 }, { "epoch": 8.33, "learning_rate": 1.429196799112139e-06, "loss": 0.0281, "step": 27080 }, { "epoch": 8.33, "learning_rate": 1.4286834413465999e-06, "loss": 0.034, "step": 27081 }, { "epoch": 8.33, "learning_rate": 1.4281701687021288e-06, "loss": 0.0249, "step": 27082 }, { "epoch": 8.33, "learning_rate": 1.4276569811838303e-06, "loss": 0.034, "step": 27083 }, { "epoch": 8.33, "learning_rate": 1.4271438787967984e-06, "loss": 0.0387, "step": 27084 }, { "epoch": 8.33, "learning_rate": 1.4266308615461288e-06, "loss": 0.0192, "step": 27085 }, { "epoch": 8.33, "learning_rate": 1.4261179294369164e-06, "loss": 0.032, "step": 27086 }, { "epoch": 8.33, "learning_rate": 1.4256050824742529e-06, "loss": 0.0074, "step": 27087 }, { "epoch": 8.33, "learning_rate": 1.4250923206632372e-06, "loss": 0.0233, "step": 27088 }, { "epoch": 8.33, "learning_rate": 1.4245796440089533e-06, "loss": 0.0214, "step": 27089 }, { "epoch": 8.34, "learning_rate": 1.424067052516499e-06, "loss": 0.0203, "step": 27090 }, { "epoch": 8.34, "learning_rate": 1.4235545461909628e-06, "loss": 0.0399, "step": 27091 }, { "epoch": 8.34, "learning_rate": 1.4230421250374338e-06, "loss": 0.0255, "step": 27092 }, { "epoch": 8.34, "learning_rate": 1.4225297890610024e-06, "loss": 0.042, "step": 27093 }, { "epoch": 8.34, "learning_rate": 1.4220175382667523e-06, "loss": 0.0229, "step": 27094 }, { "epoch": 8.34, "learning_rate": 1.4215053726597782e-06, "loss": 0.033, "step": 27095 }, { "epoch": 8.34, "learning_rate": 1.4209932922451574e-06, "loss": 0.0379, "step": 27096 }, { "epoch": 8.34, "learning_rate": 1.4204812970279824e-06, "loss": 0.0324, "step": 27097 }, { "epoch": 8.34, "learning_rate": 1.4199693870133357e-06, "loss": 0.0311, "step": 27098 }, { "epoch": 8.34, "learning_rate": 1.4194575622063e-06, "loss": 0.0296, "step": 27099 }, { "epoch": 8.34, "learning_rate": 1.418945822611959e-06, "loss": 0.0292, "step": 27100 }, { "epoch": 8.34, "learning_rate": 1.4184341682353942e-06, "loss": 0.0152, "step": 27101 }, { "epoch": 8.34, "learning_rate": 1.417922599081687e-06, "loss": 0.0235, "step": 27102 }, { "epoch": 8.34, "learning_rate": 1.4174111151559188e-06, "loss": 0.027, "step": 27103 }, { "epoch": 8.34, "learning_rate": 1.4168997164631648e-06, "loss": 0.0313, "step": 27104 }, { "epoch": 8.34, "learning_rate": 1.4163884030085117e-06, "loss": 0.0248, "step": 27105 }, { "epoch": 8.34, "learning_rate": 1.415877174797029e-06, "loss": 0.0222, "step": 27106 }, { "epoch": 8.34, "learning_rate": 1.415366031833799e-06, "loss": 0.0259, "step": 27107 }, { "epoch": 8.34, "learning_rate": 1.4148549741238959e-06, "loss": 0.0436, "step": 27108 }, { "epoch": 8.34, "learning_rate": 1.4143440016723953e-06, "loss": 0.024, "step": 27109 }, { "epoch": 8.34, "learning_rate": 1.413833114484372e-06, "loss": 0.0161, "step": 27110 }, { "epoch": 8.34, "learning_rate": 1.4133223125648966e-06, "loss": 0.0253, "step": 27111 }, { "epoch": 8.34, "learning_rate": 1.4128115959190492e-06, "loss": 0.0227, "step": 27112 }, { "epoch": 8.34, "learning_rate": 1.4123009645518938e-06, "loss": 0.0376, "step": 27113 }, { "epoch": 8.34, "learning_rate": 1.4117904184685028e-06, "loss": 0.0222, "step": 27114 }, { "epoch": 8.34, "learning_rate": 1.41127995767395e-06, "loss": 0.0392, "step": 27115 }, { "epoch": 8.34, "learning_rate": 1.4107695821733026e-06, "loss": 0.0188, "step": 27116 }, { "epoch": 8.34, "learning_rate": 1.4102592919716295e-06, "loss": 0.0263, "step": 27117 }, { "epoch": 8.34, "learning_rate": 1.4097490870739982e-06, "loss": 0.0392, "step": 27118 }, { "epoch": 8.34, "learning_rate": 1.4092389674854757e-06, "loss": 0.0195, "step": 27119 }, { "epoch": 8.34, "learning_rate": 1.4087289332111276e-06, "loss": 0.0219, "step": 27120 }, { "epoch": 8.34, "learning_rate": 1.4082189842560178e-06, "loss": 0.0548, "step": 27121 }, { "epoch": 8.34, "learning_rate": 1.4077091206252146e-06, "loss": 0.0092, "step": 27122 }, { "epoch": 8.35, "learning_rate": 1.4071993423237783e-06, "loss": 0.0223, "step": 27123 }, { "epoch": 8.35, "learning_rate": 1.4066896493567684e-06, "loss": 0.0336, "step": 27124 }, { "epoch": 8.35, "learning_rate": 1.4061800417292527e-06, "loss": 0.0339, "step": 27125 }, { "epoch": 8.35, "learning_rate": 1.4056705194462894e-06, "loss": 0.0251, "step": 27126 }, { "epoch": 8.35, "learning_rate": 1.4051610825129392e-06, "loss": 0.042, "step": 27127 }, { "epoch": 8.35, "learning_rate": 1.40465173093426e-06, "loss": 0.0293, "step": 27128 }, { "epoch": 8.35, "learning_rate": 1.4041424647153112e-06, "loss": 0.0516, "step": 27129 }, { "epoch": 8.35, "learning_rate": 1.4036332838611489e-06, "loss": 0.0583, "step": 27130 }, { "epoch": 8.35, "learning_rate": 1.4031241883768297e-06, "loss": 0.0232, "step": 27131 }, { "epoch": 8.35, "learning_rate": 1.4026151782674125e-06, "loss": 0.0363, "step": 27132 }, { "epoch": 8.35, "learning_rate": 1.4021062535379525e-06, "loss": 0.0273, "step": 27133 }, { "epoch": 8.35, "learning_rate": 1.4015974141934962e-06, "loss": 0.034, "step": 27134 }, { "epoch": 8.35, "learning_rate": 1.4010886602391049e-06, "loss": 0.035, "step": 27135 }, { "epoch": 8.35, "learning_rate": 1.400579991679828e-06, "loss": 0.0373, "step": 27136 }, { "epoch": 8.35, "learning_rate": 1.4000714085207167e-06, "loss": 0.0379, "step": 27137 }, { "epoch": 8.35, "learning_rate": 1.399562910766823e-06, "loss": 0.0287, "step": 27138 }, { "epoch": 8.35, "learning_rate": 1.399054498423197e-06, "loss": 0.0128, "step": 27139 }, { "epoch": 8.35, "learning_rate": 1.3985461714948844e-06, "loss": 0.0339, "step": 27140 }, { "epoch": 8.35, "learning_rate": 1.398037929986935e-06, "loss": 0.0236, "step": 27141 }, { "epoch": 8.35, "learning_rate": 1.3975297739043992e-06, "loss": 0.023, "step": 27142 }, { "epoch": 8.35, "learning_rate": 1.3970217032523226e-06, "loss": 0.0251, "step": 27143 }, { "epoch": 8.35, "learning_rate": 1.3965137180357447e-06, "loss": 0.0214, "step": 27144 }, { "epoch": 8.35, "learning_rate": 1.396005818259717e-06, "loss": 0.031, "step": 27145 }, { "epoch": 8.35, "learning_rate": 1.395498003929281e-06, "loss": 0.0408, "step": 27146 }, { "epoch": 8.35, "learning_rate": 1.3949902750494805e-06, "loss": 0.0482, "step": 27147 }, { "epoch": 8.35, "learning_rate": 1.3944826316253567e-06, "loss": 0.0291, "step": 27148 }, { "epoch": 8.35, "learning_rate": 1.3939750736619495e-06, "loss": 0.0239, "step": 27149 }, { "epoch": 8.35, "learning_rate": 1.3934676011643055e-06, "loss": 0.0299, "step": 27150 }, { "epoch": 8.35, "learning_rate": 1.3929602141374564e-06, "loss": 0.0472, "step": 27151 }, { "epoch": 8.35, "learning_rate": 1.392452912586446e-06, "loss": 0.0373, "step": 27152 }, { "epoch": 8.35, "learning_rate": 1.3919456965163125e-06, "loss": 0.0265, "step": 27153 }, { "epoch": 8.35, "learning_rate": 1.3914385659320907e-06, "loss": 0.0376, "step": 27154 }, { "epoch": 8.36, "learning_rate": 1.3909315208388185e-06, "loss": 0.0368, "step": 27155 }, { "epoch": 8.36, "learning_rate": 1.390424561241528e-06, "loss": 0.0398, "step": 27156 }, { "epoch": 8.36, "learning_rate": 1.3899176871452614e-06, "loss": 0.0237, "step": 27157 }, { "epoch": 8.36, "learning_rate": 1.3894108985550437e-06, "loss": 0.053, "step": 27158 }, { "epoch": 8.36, "learning_rate": 1.3889041954759108e-06, "loss": 0.0502, "step": 27159 }, { "epoch": 8.36, "learning_rate": 1.3883975779129e-06, "loss": 0.0209, "step": 27160 }, { "epoch": 8.36, "learning_rate": 1.3878910458710327e-06, "loss": 0.0428, "step": 27161 }, { "epoch": 8.36, "learning_rate": 1.387384599355347e-06, "loss": 0.0403, "step": 27162 }, { "epoch": 8.36, "learning_rate": 1.386878238370869e-06, "loss": 0.0328, "step": 27163 }, { "epoch": 8.36, "learning_rate": 1.386371962922629e-06, "loss": 0.0314, "step": 27164 }, { "epoch": 8.36, "learning_rate": 1.3858657730156544e-06, "loss": 0.0161, "step": 27165 }, { "epoch": 8.36, "learning_rate": 1.3853596686549686e-06, "loss": 0.0327, "step": 27166 }, { "epoch": 8.36, "learning_rate": 1.3848536498456054e-06, "loss": 0.0245, "step": 27167 }, { "epoch": 8.36, "learning_rate": 1.3843477165925846e-06, "loss": 0.0209, "step": 27168 }, { "epoch": 8.36, "learning_rate": 1.383841868900927e-06, "loss": 0.0441, "step": 27169 }, { "epoch": 8.36, "learning_rate": 1.3833361067756646e-06, "loss": 0.0301, "step": 27170 }, { "epoch": 8.36, "learning_rate": 1.3828304302218142e-06, "loss": 0.0346, "step": 27171 }, { "epoch": 8.36, "learning_rate": 1.382324839244401e-06, "loss": 0.0248, "step": 27172 }, { "epoch": 8.36, "learning_rate": 1.3818193338484443e-06, "loss": 0.0168, "step": 27173 }, { "epoch": 8.36, "learning_rate": 1.3813139140389632e-06, "loss": 0.027, "step": 27174 }, { "epoch": 8.36, "learning_rate": 1.3808085798209791e-06, "loss": 0.0264, "step": 27175 }, { "epoch": 8.36, "learning_rate": 1.3803033311995072e-06, "loss": 0.0495, "step": 27176 }, { "epoch": 8.36, "learning_rate": 1.3797981681795691e-06, "loss": 0.0303, "step": 27177 }, { "epoch": 8.36, "learning_rate": 1.3792930907661828e-06, "loss": 0.0316, "step": 27178 }, { "epoch": 8.36, "learning_rate": 1.3787880989643553e-06, "loss": 0.0325, "step": 27179 }, { "epoch": 8.36, "learning_rate": 1.3782831927791107e-06, "loss": 0.0195, "step": 27180 }, { "epoch": 8.36, "learning_rate": 1.3777783722154603e-06, "loss": 0.0402, "step": 27181 }, { "epoch": 8.36, "learning_rate": 1.3772736372784169e-06, "loss": 0.0264, "step": 27182 }, { "epoch": 8.36, "learning_rate": 1.3767689879729928e-06, "loss": 0.018, "step": 27183 }, { "epoch": 8.36, "learning_rate": 1.376264424304201e-06, "loss": 0.0143, "step": 27184 }, { "epoch": 8.36, "learning_rate": 1.375759946277051e-06, "loss": 0.0294, "step": 27185 }, { "epoch": 8.36, "learning_rate": 1.3752555538965507e-06, "loss": 0.0334, "step": 27186 }, { "epoch": 8.36, "learning_rate": 1.3747512471677138e-06, "loss": 0.0214, "step": 27187 }, { "epoch": 8.37, "learning_rate": 1.3742470260955477e-06, "loss": 0.0453, "step": 27188 }, { "epoch": 8.37, "learning_rate": 1.3737428906850546e-06, "loss": 0.0242, "step": 27189 }, { "epoch": 8.37, "learning_rate": 1.3732388409412466e-06, "loss": 0.0348, "step": 27190 }, { "epoch": 8.37, "learning_rate": 1.3727348768691285e-06, "loss": 0.0254, "step": 27191 }, { "epoch": 8.37, "learning_rate": 1.3722309984737036e-06, "loss": 0.0299, "step": 27192 }, { "epoch": 8.37, "learning_rate": 1.3717272057599762e-06, "loss": 0.045, "step": 27193 }, { "epoch": 8.37, "learning_rate": 1.3712234987329486e-06, "loss": 0.0356, "step": 27194 }, { "epoch": 8.37, "learning_rate": 1.3707198773976272e-06, "loss": 0.0358, "step": 27195 }, { "epoch": 8.37, "learning_rate": 1.3702163417590064e-06, "loss": 0.0226, "step": 27196 }, { "epoch": 8.37, "learning_rate": 1.3697128918220936e-06, "loss": 0.0345, "step": 27197 }, { "epoch": 8.37, "learning_rate": 1.369209527591885e-06, "loss": 0.0308, "step": 27198 }, { "epoch": 8.37, "learning_rate": 1.3687062490733804e-06, "loss": 0.0227, "step": 27199 }, { "epoch": 8.37, "learning_rate": 1.3682030562715776e-06, "loss": 0.0212, "step": 27200 }, { "epoch": 8.37, "learning_rate": 1.3676999491914745e-06, "loss": 0.0293, "step": 27201 }, { "epoch": 8.37, "learning_rate": 1.3671969278380648e-06, "loss": 0.0251, "step": 27202 }, { "epoch": 8.37, "learning_rate": 1.366693992216348e-06, "loss": 0.0424, "step": 27203 }, { "epoch": 8.37, "learning_rate": 1.3661911423313123e-06, "loss": 0.0273, "step": 27204 }, { "epoch": 8.37, "learning_rate": 1.3656883781879616e-06, "loss": 0.0307, "step": 27205 }, { "epoch": 8.37, "learning_rate": 1.3651856997912782e-06, "loss": 0.0434, "step": 27206 }, { "epoch": 8.37, "learning_rate": 1.3646831071462606e-06, "loss": 0.0325, "step": 27207 }, { "epoch": 8.37, "learning_rate": 1.3641806002578984e-06, "loss": 0.0262, "step": 27208 }, { "epoch": 8.37, "learning_rate": 1.3636781791311826e-06, "loss": 0.0243, "step": 27209 }, { "epoch": 8.37, "learning_rate": 1.3631758437711007e-06, "loss": 0.0174, "step": 27210 }, { "epoch": 8.37, "learning_rate": 1.362673594182642e-06, "loss": 0.02, "step": 27211 }, { "epoch": 8.37, "learning_rate": 1.3621714303707978e-06, "loss": 0.0282, "step": 27212 }, { "epoch": 8.37, "learning_rate": 1.361669352340551e-06, "loss": 0.0368, "step": 27213 }, { "epoch": 8.37, "learning_rate": 1.3611673600968855e-06, "loss": 0.0292, "step": 27214 }, { "epoch": 8.37, "learning_rate": 1.3606654536447927e-06, "loss": 0.0182, "step": 27215 }, { "epoch": 8.37, "learning_rate": 1.3601636329892543e-06, "loss": 0.0311, "step": 27216 }, { "epoch": 8.37, "learning_rate": 1.359661898135254e-06, "loss": 0.0185, "step": 27217 }, { "epoch": 8.37, "learning_rate": 1.3591602490877742e-06, "loss": 0.0284, "step": 27218 }, { "epoch": 8.37, "learning_rate": 1.3586586858517958e-06, "loss": 0.0457, "step": 27219 }, { "epoch": 8.38, "learning_rate": 1.3581572084323014e-06, "loss": 0.0376, "step": 27220 }, { "epoch": 8.38, "learning_rate": 1.357655816834269e-06, "loss": 0.016, "step": 27221 }, { "epoch": 8.38, "learning_rate": 1.3571545110626827e-06, "loss": 0.0352, "step": 27222 }, { "epoch": 8.38, "learning_rate": 1.356653291122516e-06, "loss": 0.0224, "step": 27223 }, { "epoch": 8.38, "learning_rate": 1.3561521570187453e-06, "loss": 0.0251, "step": 27224 }, { "epoch": 8.38, "learning_rate": 1.3556511087563529e-06, "loss": 0.0372, "step": 27225 }, { "epoch": 8.38, "learning_rate": 1.3551501463403115e-06, "loss": 0.0178, "step": 27226 }, { "epoch": 8.38, "learning_rate": 1.3546492697755963e-06, "loss": 0.0303, "step": 27227 }, { "epoch": 8.38, "learning_rate": 1.354148479067182e-06, "loss": 0.0261, "step": 27228 }, { "epoch": 8.38, "learning_rate": 1.3536477742200415e-06, "loss": 0.0297, "step": 27229 }, { "epoch": 8.38, "learning_rate": 1.3531471552391473e-06, "loss": 0.0311, "step": 27230 }, { "epoch": 8.38, "learning_rate": 1.3526466221294687e-06, "loss": 0.0305, "step": 27231 }, { "epoch": 8.38, "learning_rate": 1.3521461748959808e-06, "loss": 0.0327, "step": 27232 }, { "epoch": 8.38, "learning_rate": 1.3516458135436539e-06, "loss": 0.0231, "step": 27233 }, { "epoch": 8.38, "learning_rate": 1.3511455380774507e-06, "loss": 0.0158, "step": 27234 }, { "epoch": 8.38, "learning_rate": 1.3506453485023452e-06, "loss": 0.0258, "step": 27235 }, { "epoch": 8.38, "learning_rate": 1.350145244823302e-06, "loss": 0.0232, "step": 27236 }, { "epoch": 8.38, "learning_rate": 1.3496452270452898e-06, "loss": 0.0347, "step": 27237 }, { "epoch": 8.38, "learning_rate": 1.349145295173272e-06, "loss": 0.0291, "step": 27238 }, { "epoch": 8.38, "learning_rate": 1.3486454492122147e-06, "loss": 0.0196, "step": 27239 }, { "epoch": 8.38, "learning_rate": 1.3481456891670818e-06, "loss": 0.0287, "step": 27240 }, { "epoch": 8.38, "learning_rate": 1.3476460150428327e-06, "loss": 0.0415, "step": 27241 }, { "epoch": 8.38, "learning_rate": 1.3471464268444356e-06, "loss": 0.0332, "step": 27242 }, { "epoch": 8.38, "learning_rate": 1.346646924576852e-06, "loss": 0.0491, "step": 27243 }, { "epoch": 8.38, "learning_rate": 1.3461475082450349e-06, "loss": 0.0202, "step": 27244 }, { "epoch": 8.38, "learning_rate": 1.3456481778539499e-06, "loss": 0.0281, "step": 27245 }, { "epoch": 8.38, "learning_rate": 1.3451489334085555e-06, "loss": 0.0339, "step": 27246 }, { "epoch": 8.38, "learning_rate": 1.3446497749138087e-06, "loss": 0.0495, "step": 27247 }, { "epoch": 8.38, "learning_rate": 1.344150702374667e-06, "loss": 0.0249, "step": 27248 }, { "epoch": 8.38, "learning_rate": 1.3436517157960837e-06, "loss": 0.037, "step": 27249 }, { "epoch": 8.38, "learning_rate": 1.343152815183021e-06, "loss": 0.0431, "step": 27250 }, { "epoch": 8.38, "learning_rate": 1.3426540005404243e-06, "loss": 0.0289, "step": 27251 }, { "epoch": 8.38, "learning_rate": 1.3421552718732555e-06, "loss": 0.0286, "step": 27252 }, { "epoch": 8.39, "learning_rate": 1.341656629186464e-06, "loss": 0.0612, "step": 27253 }, { "epoch": 8.39, "learning_rate": 1.3411580724850015e-06, "loss": 0.0366, "step": 27254 }, { "epoch": 8.39, "learning_rate": 1.3406596017738195e-06, "loss": 0.0436, "step": 27255 }, { "epoch": 8.39, "learning_rate": 1.340161217057866e-06, "loss": 0.0462, "step": 27256 }, { "epoch": 8.39, "learning_rate": 1.3396629183420984e-06, "loss": 0.0242, "step": 27257 }, { "epoch": 8.39, "learning_rate": 1.339164705631456e-06, "loss": 0.0227, "step": 27258 }, { "epoch": 8.39, "learning_rate": 1.3386665789308885e-06, "loss": 0.0183, "step": 27259 }, { "epoch": 8.39, "learning_rate": 1.3381685382453492e-06, "loss": 0.0392, "step": 27260 }, { "epoch": 8.39, "learning_rate": 1.3376705835797744e-06, "loss": 0.0536, "step": 27261 }, { "epoch": 8.39, "learning_rate": 1.3371727149391156e-06, "loss": 0.0202, "step": 27262 }, { "epoch": 8.39, "learning_rate": 1.3366749323283167e-06, "loss": 0.0163, "step": 27263 }, { "epoch": 8.39, "learning_rate": 1.3361772357523195e-06, "loss": 0.0439, "step": 27264 }, { "epoch": 8.39, "learning_rate": 1.3356796252160676e-06, "loss": 0.0188, "step": 27265 }, { "epoch": 8.39, "learning_rate": 1.3351821007245003e-06, "loss": 0.0367, "step": 27266 }, { "epoch": 8.39, "learning_rate": 1.3346846622825649e-06, "loss": 0.0183, "step": 27267 }, { "epoch": 8.39, "learning_rate": 1.334187309895194e-06, "loss": 0.0447, "step": 27268 }, { "epoch": 8.39, "learning_rate": 1.3336900435673283e-06, "loss": 0.0359, "step": 27269 }, { "epoch": 8.39, "learning_rate": 1.3331928633039093e-06, "loss": 0.0321, "step": 27270 }, { "epoch": 8.39, "learning_rate": 1.332695769109873e-06, "loss": 0.0259, "step": 27271 }, { "epoch": 8.39, "learning_rate": 1.3321987609901553e-06, "loss": 0.0134, "step": 27272 }, { "epoch": 8.39, "learning_rate": 1.3317018389496927e-06, "loss": 0.019, "step": 27273 }, { "epoch": 8.39, "learning_rate": 1.3312050029934197e-06, "loss": 0.0267, "step": 27274 }, { "epoch": 8.39, "learning_rate": 1.3307082531262706e-06, "loss": 0.0516, "step": 27275 }, { "epoch": 8.39, "learning_rate": 1.3302115893531763e-06, "loss": 0.0355, "step": 27276 }, { "epoch": 8.39, "learning_rate": 1.3297150116790757e-06, "loss": 0.0349, "step": 27277 }, { "epoch": 8.39, "learning_rate": 1.3292185201088926e-06, "loss": 0.0364, "step": 27278 }, { "epoch": 8.39, "learning_rate": 1.3287221146475593e-06, "loss": 0.0256, "step": 27279 }, { "epoch": 8.39, "learning_rate": 1.3282257953000099e-06, "loss": 0.0318, "step": 27280 }, { "epoch": 8.39, "learning_rate": 1.3277295620711695e-06, "loss": 0.0362, "step": 27281 }, { "epoch": 8.39, "learning_rate": 1.3272334149659683e-06, "loss": 0.047, "step": 27282 }, { "epoch": 8.39, "learning_rate": 1.3267373539893313e-06, "loss": 0.035, "step": 27283 }, { "epoch": 8.39, "learning_rate": 1.326241379146186e-06, "loss": 0.0345, "step": 27284 }, { "epoch": 8.4, "learning_rate": 1.325745490441458e-06, "loss": 0.0207, "step": 27285 }, { "epoch": 8.4, "learning_rate": 1.3252496878800702e-06, "loss": 0.0283, "step": 27286 }, { "epoch": 8.4, "learning_rate": 1.324753971466949e-06, "loss": 0.0298, "step": 27287 }, { "epoch": 8.4, "learning_rate": 1.3242583412070186e-06, "loss": 0.0333, "step": 27288 }, { "epoch": 8.4, "learning_rate": 1.3237627971051947e-06, "loss": 0.0337, "step": 27289 }, { "epoch": 8.4, "learning_rate": 1.3232673391664042e-06, "loss": 0.0339, "step": 27290 }, { "epoch": 8.4, "learning_rate": 1.3227719673955664e-06, "loss": 0.0891, "step": 27291 }, { "epoch": 8.4, "learning_rate": 1.3222766817975995e-06, "loss": 0.015, "step": 27292 }, { "epoch": 8.4, "learning_rate": 1.321781482377421e-06, "loss": 0.0484, "step": 27293 }, { "epoch": 8.4, "learning_rate": 1.3212863691399525e-06, "loss": 0.0139, "step": 27294 }, { "epoch": 8.4, "learning_rate": 1.3207913420901108e-06, "loss": 0.0287, "step": 27295 }, { "epoch": 8.4, "learning_rate": 1.3202964012328056e-06, "loss": 0.0335, "step": 27296 }, { "epoch": 8.4, "learning_rate": 1.3198015465729597e-06, "loss": 0.039, "step": 27297 }, { "epoch": 8.4, "learning_rate": 1.3193067781154835e-06, "loss": 0.0229, "step": 27298 }, { "epoch": 8.4, "learning_rate": 1.3188120958652917e-06, "loss": 0.033, "step": 27299 }, { "epoch": 8.4, "learning_rate": 1.3183174998272962e-06, "loss": 0.039, "step": 27300 }, { "epoch": 8.4, "learning_rate": 1.31782299000641e-06, "loss": 0.0296, "step": 27301 }, { "epoch": 8.4, "learning_rate": 1.317328566407543e-06, "loss": 0.0355, "step": 27302 }, { "epoch": 8.4, "learning_rate": 1.316834229035603e-06, "loss": 0.0314, "step": 27303 }, { "epoch": 8.4, "learning_rate": 1.3163399778955032e-06, "loss": 0.0348, "step": 27304 }, { "epoch": 8.4, "learning_rate": 1.3158458129921535e-06, "loss": 0.0397, "step": 27305 }, { "epoch": 8.4, "learning_rate": 1.3153517343304544e-06, "loss": 0.0349, "step": 27306 }, { "epoch": 8.4, "learning_rate": 1.3148577419153186e-06, "loss": 0.0367, "step": 27307 }, { "epoch": 8.4, "learning_rate": 1.3143638357516497e-06, "loss": 0.0319, "step": 27308 }, { "epoch": 8.4, "learning_rate": 1.313870015844353e-06, "loss": 0.0201, "step": 27309 }, { "epoch": 8.4, "learning_rate": 1.313376282198332e-06, "loss": 0.03, "step": 27310 }, { "epoch": 8.4, "learning_rate": 1.3128826348184886e-06, "loss": 0.0183, "step": 27311 }, { "epoch": 8.4, "learning_rate": 1.3123890737097312e-06, "loss": 0.0245, "step": 27312 }, { "epoch": 8.4, "learning_rate": 1.311895598876952e-06, "loss": 0.0298, "step": 27313 }, { "epoch": 8.4, "learning_rate": 1.3114022103250601e-06, "loss": 0.0191, "step": 27314 }, { "epoch": 8.4, "learning_rate": 1.3109089080589509e-06, "loss": 0.0277, "step": 27315 }, { "epoch": 8.4, "learning_rate": 1.3104156920835254e-06, "loss": 0.0848, "step": 27316 }, { "epoch": 8.4, "learning_rate": 1.3099225624036792e-06, "loss": 0.0369, "step": 27317 }, { "epoch": 8.41, "learning_rate": 1.3094295190243123e-06, "loss": 0.0317, "step": 27318 }, { "epoch": 8.41, "learning_rate": 1.3089365619503191e-06, "loss": 0.0416, "step": 27319 }, { "epoch": 8.41, "learning_rate": 1.308443691186596e-06, "loss": 0.0236, "step": 27320 }, { "epoch": 8.41, "learning_rate": 1.3079509067380346e-06, "loss": 0.0235, "step": 27321 }, { "epoch": 8.41, "learning_rate": 1.3074582086095365e-06, "loss": 0.0109, "step": 27322 }, { "epoch": 8.41, "learning_rate": 1.3069655968059858e-06, "loss": 0.0495, "step": 27323 }, { "epoch": 8.41, "learning_rate": 1.3064730713322793e-06, "loss": 0.0483, "step": 27324 }, { "epoch": 8.41, "learning_rate": 1.305980632193309e-06, "loss": 0.0291, "step": 27325 }, { "epoch": 8.41, "learning_rate": 1.3054882793939616e-06, "loss": 0.0235, "step": 27326 }, { "epoch": 8.41, "learning_rate": 1.3049960129391303e-06, "loss": 0.0228, "step": 27327 }, { "epoch": 8.41, "learning_rate": 1.304503832833699e-06, "loss": 0.035, "step": 27328 }, { "epoch": 8.41, "learning_rate": 1.3040117390825635e-06, "loss": 0.0168, "step": 27329 }, { "epoch": 8.41, "learning_rate": 1.3035197316906046e-06, "loss": 0.0303, "step": 27330 }, { "epoch": 8.41, "learning_rate": 1.3030278106627058e-06, "loss": 0.0354, "step": 27331 }, { "epoch": 8.41, "learning_rate": 1.3025359760037592e-06, "loss": 0.0493, "step": 27332 }, { "epoch": 8.41, "learning_rate": 1.3020442277186463e-06, "loss": 0.0168, "step": 27333 }, { "epoch": 8.41, "learning_rate": 1.3015525658122507e-06, "loss": 0.0366, "step": 27334 }, { "epoch": 8.41, "learning_rate": 1.3010609902894545e-06, "loss": 0.0333, "step": 27335 }, { "epoch": 8.41, "learning_rate": 1.3005695011551401e-06, "loss": 0.0348, "step": 27336 }, { "epoch": 8.41, "learning_rate": 1.3000780984141881e-06, "loss": 0.0377, "step": 27337 }, { "epoch": 8.41, "learning_rate": 1.2995867820714757e-06, "loss": 0.0394, "step": 27338 }, { "epoch": 8.41, "learning_rate": 1.2990955521318904e-06, "loss": 0.0417, "step": 27339 }, { "epoch": 8.41, "learning_rate": 1.2986044086003025e-06, "loss": 0.0341, "step": 27340 }, { "epoch": 8.41, "learning_rate": 1.2981133514815903e-06, "loss": 0.0193, "step": 27341 }, { "epoch": 8.41, "learning_rate": 1.2976223807806355e-06, "loss": 0.0331, "step": 27342 }, { "epoch": 8.41, "learning_rate": 1.2971314965023095e-06, "loss": 0.0281, "step": 27343 }, { "epoch": 8.41, "learning_rate": 1.296640698651489e-06, "loss": 0.0125, "step": 27344 }, { "epoch": 8.41, "learning_rate": 1.2961499872330475e-06, "loss": 0.0377, "step": 27345 }, { "epoch": 8.41, "learning_rate": 1.2956593622518588e-06, "loss": 0.014, "step": 27346 }, { "epoch": 8.41, "learning_rate": 1.2951688237127945e-06, "loss": 0.0222, "step": 27347 }, { "epoch": 8.41, "learning_rate": 1.2946783716207234e-06, "loss": 0.022, "step": 27348 }, { "epoch": 8.41, "learning_rate": 1.2941880059805222e-06, "loss": 0.0195, "step": 27349 }, { "epoch": 8.42, "learning_rate": 1.2936977267970597e-06, "loss": 0.0406, "step": 27350 }, { "epoch": 8.42, "learning_rate": 1.293207534075198e-06, "loss": 0.0302, "step": 27351 }, { "epoch": 8.42, "learning_rate": 1.2927174278198128e-06, "loss": 0.0383, "step": 27352 }, { "epoch": 8.42, "learning_rate": 1.2922274080357677e-06, "loss": 0.0146, "step": 27353 }, { "epoch": 8.42, "learning_rate": 1.291737474727931e-06, "loss": 0.0306, "step": 27354 }, { "epoch": 8.42, "learning_rate": 1.2912476279011654e-06, "loss": 0.0307, "step": 27355 }, { "epoch": 8.42, "learning_rate": 1.2907578675603383e-06, "loss": 0.0233, "step": 27356 }, { "epoch": 8.42, "learning_rate": 1.2902681937103111e-06, "loss": 0.0311, "step": 27357 }, { "epoch": 8.42, "learning_rate": 1.289778606355946e-06, "loss": 0.0445, "step": 27358 }, { "epoch": 8.42, "learning_rate": 1.2892891055021095e-06, "loss": 0.0243, "step": 27359 }, { "epoch": 8.42, "learning_rate": 1.2887996911536627e-06, "loss": 0.0357, "step": 27360 }, { "epoch": 8.42, "learning_rate": 1.2883103633154582e-06, "loss": 0.0167, "step": 27361 }, { "epoch": 8.42, "learning_rate": 1.287821121992363e-06, "loss": 0.0135, "step": 27362 }, { "epoch": 8.42, "learning_rate": 1.2873319671892337e-06, "loss": 0.0294, "step": 27363 }, { "epoch": 8.42, "learning_rate": 1.2868428989109283e-06, "loss": 0.0257, "step": 27364 }, { "epoch": 8.42, "learning_rate": 1.286353917162303e-06, "loss": 0.0256, "step": 27365 }, { "epoch": 8.42, "learning_rate": 1.2858650219482116e-06, "loss": 0.0361, "step": 27366 }, { "epoch": 8.42, "learning_rate": 1.2853762132735158e-06, "loss": 0.0269, "step": 27367 }, { "epoch": 8.42, "learning_rate": 1.284887491143062e-06, "loss": 0.0465, "step": 27368 }, { "epoch": 8.42, "learning_rate": 1.2843988555617094e-06, "loss": 0.0309, "step": 27369 }, { "epoch": 8.42, "learning_rate": 1.2839103065343084e-06, "loss": 0.0339, "step": 27370 }, { "epoch": 8.42, "learning_rate": 1.2834218440657108e-06, "loss": 0.0496, "step": 27371 }, { "epoch": 8.42, "learning_rate": 1.2829334681607686e-06, "loss": 0.0444, "step": 27372 }, { "epoch": 8.42, "learning_rate": 1.2824451788243286e-06, "loss": 0.0343, "step": 27373 }, { "epoch": 8.42, "learning_rate": 1.281956976061246e-06, "loss": 0.0312, "step": 27374 }, { "epoch": 8.42, "learning_rate": 1.2814688598763648e-06, "loss": 0.0285, "step": 27375 }, { "epoch": 8.42, "learning_rate": 1.2809808302745298e-06, "loss": 0.0223, "step": 27376 }, { "epoch": 8.42, "learning_rate": 1.2804928872605948e-06, "loss": 0.0386, "step": 27377 }, { "epoch": 8.42, "learning_rate": 1.2800050308393974e-06, "loss": 0.0252, "step": 27378 }, { "epoch": 8.42, "learning_rate": 1.2795172610157891e-06, "loss": 0.0327, "step": 27379 }, { "epoch": 8.42, "learning_rate": 1.2790295777946115e-06, "loss": 0.029, "step": 27380 }, { "epoch": 8.42, "learning_rate": 1.2785419811807075e-06, "loss": 0.0406, "step": 27381 }, { "epoch": 8.42, "learning_rate": 1.27805447117892e-06, "loss": 0.0416, "step": 27382 }, { "epoch": 8.43, "learning_rate": 1.277567047794087e-06, "loss": 0.0321, "step": 27383 }, { "epoch": 8.43, "learning_rate": 1.2770797110310573e-06, "loss": 0.0351, "step": 27384 }, { "epoch": 8.43, "learning_rate": 1.2765924608946622e-06, "loss": 0.0254, "step": 27385 }, { "epoch": 8.43, "learning_rate": 1.2761052973897425e-06, "loss": 0.0171, "step": 27386 }, { "epoch": 8.43, "learning_rate": 1.2756182205211386e-06, "loss": 0.0394, "step": 27387 }, { "epoch": 8.43, "learning_rate": 1.275131230293687e-06, "loss": 0.038, "step": 27388 }, { "epoch": 8.43, "learning_rate": 1.2746443267122233e-06, "loss": 0.0314, "step": 27389 }, { "epoch": 8.43, "learning_rate": 1.2741575097815829e-06, "loss": 0.0479, "step": 27390 }, { "epoch": 8.43, "learning_rate": 1.2736707795066006e-06, "loss": 0.0378, "step": 27391 }, { "epoch": 8.43, "learning_rate": 1.2731841358921094e-06, "loss": 0.017, "step": 27392 }, { "epoch": 8.43, "learning_rate": 1.2726975789429408e-06, "loss": 0.0407, "step": 27393 }, { "epoch": 8.43, "learning_rate": 1.27221110866393e-06, "loss": 0.0263, "step": 27394 }, { "epoch": 8.43, "learning_rate": 1.2717247250599096e-06, "loss": 0.0584, "step": 27395 }, { "epoch": 8.43, "learning_rate": 1.2712384281357026e-06, "loss": 0.0254, "step": 27396 }, { "epoch": 8.43, "learning_rate": 1.270752217896144e-06, "loss": 0.049, "step": 27397 }, { "epoch": 8.43, "learning_rate": 1.270266094346062e-06, "loss": 0.0339, "step": 27398 }, { "epoch": 8.43, "learning_rate": 1.269780057490283e-06, "loss": 0.0502, "step": 27399 }, { "epoch": 8.43, "learning_rate": 1.269294107333634e-06, "loss": 0.0187, "step": 27400 }, { "epoch": 8.43, "learning_rate": 1.2688082438809413e-06, "loss": 0.0159, "step": 27401 }, { "epoch": 8.43, "learning_rate": 1.2683224671370286e-06, "loss": 0.0281, "step": 27402 }, { "epoch": 8.43, "learning_rate": 1.26783677710672e-06, "loss": 0.0455, "step": 27403 }, { "epoch": 8.43, "learning_rate": 1.267351173794842e-06, "loss": 0.042, "step": 27404 }, { "epoch": 8.43, "learning_rate": 1.2668656572062167e-06, "loss": 0.0327, "step": 27405 }, { "epoch": 8.43, "learning_rate": 1.2663802273456593e-06, "loss": 0.0221, "step": 27406 }, { "epoch": 8.43, "learning_rate": 1.2658948842179986e-06, "loss": 0.0489, "step": 27407 }, { "epoch": 8.43, "learning_rate": 1.2654096278280514e-06, "loss": 0.0399, "step": 27408 }, { "epoch": 8.43, "learning_rate": 1.2649244581806363e-06, "loss": 0.025, "step": 27409 }, { "epoch": 8.43, "learning_rate": 1.2644393752805717e-06, "loss": 0.0531, "step": 27410 }, { "epoch": 8.43, "learning_rate": 1.2639543791326726e-06, "loss": 0.0156, "step": 27411 }, { "epoch": 8.43, "learning_rate": 1.2634694697417626e-06, "loss": 0.0199, "step": 27412 }, { "epoch": 8.43, "learning_rate": 1.2629846471126483e-06, "loss": 0.0252, "step": 27413 }, { "epoch": 8.43, "learning_rate": 1.2624999112501513e-06, "loss": 0.0506, "step": 27414 }, { "epoch": 8.44, "learning_rate": 1.262015262159082e-06, "loss": 0.0304, "step": 27415 }, { "epoch": 8.44, "learning_rate": 1.2615306998442545e-06, "loss": 0.0195, "step": 27416 }, { "epoch": 8.44, "learning_rate": 1.2610462243104815e-06, "loss": 0.0328, "step": 27417 }, { "epoch": 8.44, "learning_rate": 1.2605618355625726e-06, "loss": 0.0143, "step": 27418 }, { "epoch": 8.44, "learning_rate": 1.2600775336053405e-06, "loss": 0.0351, "step": 27419 }, { "epoch": 8.44, "learning_rate": 1.2595933184435915e-06, "loss": 0.024, "step": 27420 }, { "epoch": 8.44, "learning_rate": 1.2591091900821362e-06, "loss": 0.025, "step": 27421 }, { "epoch": 8.44, "learning_rate": 1.258625148525785e-06, "loss": 0.0693, "step": 27422 }, { "epoch": 8.44, "learning_rate": 1.2581411937793386e-06, "loss": 0.0218, "step": 27423 }, { "epoch": 8.44, "learning_rate": 1.25765732584761e-06, "loss": 0.0285, "step": 27424 }, { "epoch": 8.44, "learning_rate": 1.2571735447354005e-06, "loss": 0.0279, "step": 27425 }, { "epoch": 8.44, "learning_rate": 1.2566898504475157e-06, "loss": 0.03, "step": 27426 }, { "epoch": 8.44, "learning_rate": 1.2562062429887578e-06, "loss": 0.0331, "step": 27427 }, { "epoch": 8.44, "learning_rate": 1.255722722363929e-06, "loss": 0.0197, "step": 27428 }, { "epoch": 8.44, "learning_rate": 1.2552392885778364e-06, "loss": 0.0239, "step": 27429 }, { "epoch": 8.44, "learning_rate": 1.2547559416352762e-06, "loss": 0.0097, "step": 27430 }, { "epoch": 8.44, "learning_rate": 1.2542726815410456e-06, "loss": 0.019, "step": 27431 }, { "epoch": 8.44, "learning_rate": 1.2537895082999507e-06, "loss": 0.0223, "step": 27432 }, { "epoch": 8.44, "learning_rate": 1.2533064219167868e-06, "loss": 0.0227, "step": 27433 }, { "epoch": 8.44, "learning_rate": 1.252823422396351e-06, "loss": 0.04, "step": 27434 }, { "epoch": 8.44, "learning_rate": 1.2523405097434393e-06, "loss": 0.0363, "step": 27435 }, { "epoch": 8.44, "learning_rate": 1.2518576839628494e-06, "loss": 0.0272, "step": 27436 }, { "epoch": 8.44, "learning_rate": 1.251374945059376e-06, "loss": 0.016, "step": 27437 }, { "epoch": 8.44, "learning_rate": 1.2508922930378086e-06, "loss": 0.0279, "step": 27438 }, { "epoch": 8.44, "learning_rate": 1.2504097279029491e-06, "loss": 0.0215, "step": 27439 }, { "epoch": 8.44, "learning_rate": 1.2499272496595826e-06, "loss": 0.0251, "step": 27440 }, { "epoch": 8.44, "learning_rate": 1.249444858312502e-06, "loss": 0.0373, "step": 27441 }, { "epoch": 8.44, "learning_rate": 1.2489625538664995e-06, "loss": 0.028, "step": 27442 }, { "epoch": 8.44, "learning_rate": 1.2484803363263641e-06, "loss": 0.0414, "step": 27443 }, { "epoch": 8.44, "learning_rate": 1.2479982056968842e-06, "loss": 0.0339, "step": 27444 }, { "epoch": 8.44, "learning_rate": 1.2475161619828492e-06, "loss": 0.0388, "step": 27445 }, { "epoch": 8.44, "learning_rate": 1.247034205189045e-06, "loss": 0.0277, "step": 27446 }, { "epoch": 8.44, "learning_rate": 1.246552335320258e-06, "loss": 0.0422, "step": 27447 }, { "epoch": 8.45, "learning_rate": 1.246070552381271e-06, "loss": 0.0377, "step": 27448 }, { "epoch": 8.45, "learning_rate": 1.2455888563768748e-06, "loss": 0.0229, "step": 27449 }, { "epoch": 8.45, "learning_rate": 1.245107247311851e-06, "loss": 0.022, "step": 27450 }, { "epoch": 8.45, "learning_rate": 1.2446257251909754e-06, "loss": 0.0192, "step": 27451 }, { "epoch": 8.45, "learning_rate": 1.2441442900190392e-06, "loss": 0.0457, "step": 27452 }, { "epoch": 8.45, "learning_rate": 1.2436629418008194e-06, "loss": 0.0347, "step": 27453 }, { "epoch": 8.45, "learning_rate": 1.2431816805410968e-06, "loss": 0.0188, "step": 27454 }, { "epoch": 8.45, "learning_rate": 1.2427005062446508e-06, "loss": 0.061, "step": 27455 }, { "epoch": 8.45, "learning_rate": 1.2422194189162595e-06, "loss": 0.0328, "step": 27456 }, { "epoch": 8.45, "learning_rate": 1.2417384185607006e-06, "loss": 0.0276, "step": 27457 }, { "epoch": 8.45, "learning_rate": 1.241257505182749e-06, "loss": 0.0412, "step": 27458 }, { "epoch": 8.45, "learning_rate": 1.2407766787871845e-06, "loss": 0.0389, "step": 27459 }, { "epoch": 8.45, "learning_rate": 1.2402959393787828e-06, "loss": 0.0184, "step": 27460 }, { "epoch": 8.45, "learning_rate": 1.2398152869623115e-06, "loss": 0.0283, "step": 27461 }, { "epoch": 8.45, "learning_rate": 1.23933472154255e-06, "loss": 0.0114, "step": 27462 }, { "epoch": 8.45, "learning_rate": 1.2388542431242689e-06, "loss": 0.0378, "step": 27463 }, { "epoch": 8.45, "learning_rate": 1.2383738517122401e-06, "loss": 0.0337, "step": 27464 }, { "epoch": 8.45, "learning_rate": 1.237893547311233e-06, "loss": 0.0234, "step": 27465 }, { "epoch": 8.45, "learning_rate": 1.2374133299260173e-06, "loss": 0.032, "step": 27466 }, { "epoch": 8.45, "learning_rate": 1.2369331995613664e-06, "loss": 0.0282, "step": 27467 }, { "epoch": 8.45, "learning_rate": 1.2364531562220406e-06, "loss": 0.0205, "step": 27468 }, { "epoch": 8.45, "learning_rate": 1.2359731999128144e-06, "loss": 0.0451, "step": 27469 }, { "epoch": 8.45, "learning_rate": 1.2354933306384509e-06, "loss": 0.0203, "step": 27470 }, { "epoch": 8.45, "learning_rate": 1.2350135484037173e-06, "loss": 0.0327, "step": 27471 }, { "epoch": 8.45, "learning_rate": 1.2345338532133766e-06, "loss": 0.0229, "step": 27472 }, { "epoch": 8.45, "learning_rate": 1.2340542450721915e-06, "loss": 0.0344, "step": 27473 }, { "epoch": 8.45, "learning_rate": 1.2335747239849306e-06, "loss": 0.011, "step": 27474 }, { "epoch": 8.45, "learning_rate": 1.2330952899563498e-06, "loss": 0.0109, "step": 27475 }, { "epoch": 8.45, "learning_rate": 1.23261594299121e-06, "loss": 0.0349, "step": 27476 }, { "epoch": 8.45, "learning_rate": 1.2321366830942794e-06, "loss": 0.0422, "step": 27477 }, { "epoch": 8.45, "learning_rate": 1.2316575102703077e-06, "loss": 0.0156, "step": 27478 }, { "epoch": 8.45, "learning_rate": 1.23117842452406e-06, "loss": 0.0202, "step": 27479 }, { "epoch": 8.46, "learning_rate": 1.2306994258602922e-06, "loss": 0.0239, "step": 27480 }, { "epoch": 8.46, "learning_rate": 1.2302205142837609e-06, "loss": 0.0256, "step": 27481 }, { "epoch": 8.46, "learning_rate": 1.2297416897992232e-06, "loss": 0.0196, "step": 27482 }, { "epoch": 8.46, "learning_rate": 1.2292629524114296e-06, "loss": 0.018, "step": 27483 }, { "epoch": 8.46, "learning_rate": 1.2287843021251445e-06, "loss": 0.0107, "step": 27484 }, { "epoch": 8.46, "learning_rate": 1.2283057389451113e-06, "loss": 0.0189, "step": 27485 }, { "epoch": 8.46, "learning_rate": 1.2278272628760857e-06, "loss": 0.0334, "step": 27486 }, { "epoch": 8.46, "learning_rate": 1.2273488739228213e-06, "loss": 0.04, "step": 27487 }, { "epoch": 8.46, "learning_rate": 1.2268705720900676e-06, "loss": 0.047, "step": 27488 }, { "epoch": 8.46, "learning_rate": 1.2263923573825753e-06, "loss": 0.0433, "step": 27489 }, { "epoch": 8.46, "learning_rate": 1.2259142298050896e-06, "loss": 0.0375, "step": 27490 }, { "epoch": 8.46, "learning_rate": 1.225436189362368e-06, "loss": 0.0218, "step": 27491 }, { "epoch": 8.46, "learning_rate": 1.2249582360591483e-06, "loss": 0.0455, "step": 27492 }, { "epoch": 8.46, "learning_rate": 1.2244803699001785e-06, "loss": 0.0298, "step": 27493 }, { "epoch": 8.46, "learning_rate": 1.2240025908902097e-06, "loss": 0.025, "step": 27494 }, { "epoch": 8.46, "learning_rate": 1.2235248990339821e-06, "loss": 0.0446, "step": 27495 }, { "epoch": 8.46, "learning_rate": 1.2230472943362414e-06, "loss": 0.0297, "step": 27496 }, { "epoch": 8.46, "learning_rate": 1.2225697768017309e-06, "loss": 0.0283, "step": 27497 }, { "epoch": 8.46, "learning_rate": 1.2220923464351908e-06, "loss": 0.0392, "step": 27498 }, { "epoch": 8.46, "learning_rate": 1.2216150032413643e-06, "loss": 0.0316, "step": 27499 }, { "epoch": 8.46, "learning_rate": 1.2211377472249896e-06, "loss": 0.0269, "step": 27500 }, { "epoch": 8.46, "learning_rate": 1.220660578390811e-06, "loss": 0.0218, "step": 27501 }, { "epoch": 8.46, "learning_rate": 1.2201834967435633e-06, "loss": 0.0237, "step": 27502 }, { "epoch": 8.46, "learning_rate": 1.2197065022879817e-06, "loss": 0.0354, "step": 27503 }, { "epoch": 8.46, "learning_rate": 1.219229595028809e-06, "loss": 0.0154, "step": 27504 }, { "epoch": 8.46, "learning_rate": 1.2187527749707805e-06, "loss": 0.0222, "step": 27505 }, { "epoch": 8.46, "learning_rate": 1.218276042118629e-06, "loss": 0.023, "step": 27506 }, { "epoch": 8.46, "learning_rate": 1.2177993964770896e-06, "loss": 0.0424, "step": 27507 }, { "epoch": 8.46, "learning_rate": 1.2173228380508961e-06, "loss": 0.0581, "step": 27508 }, { "epoch": 8.46, "learning_rate": 1.2168463668447817e-06, "loss": 0.0194, "step": 27509 }, { "epoch": 8.46, "learning_rate": 1.2163699828634746e-06, "loss": 0.0204, "step": 27510 }, { "epoch": 8.46, "learning_rate": 1.215893686111712e-06, "loss": 0.023, "step": 27511 }, { "epoch": 8.46, "learning_rate": 1.2154174765942227e-06, "loss": 0.0228, "step": 27512 }, { "epoch": 8.47, "learning_rate": 1.2149413543157295e-06, "loss": 0.0157, "step": 27513 }, { "epoch": 8.47, "learning_rate": 1.2144653192809675e-06, "loss": 0.0219, "step": 27514 }, { "epoch": 8.47, "learning_rate": 1.2139893714946616e-06, "loss": 0.0535, "step": 27515 }, { "epoch": 8.47, "learning_rate": 1.213513510961538e-06, "loss": 0.0341, "step": 27516 }, { "epoch": 8.47, "learning_rate": 1.2130377376863244e-06, "loss": 0.0192, "step": 27517 }, { "epoch": 8.47, "learning_rate": 1.2125620516737435e-06, "loss": 0.017, "step": 27518 }, { "epoch": 8.47, "learning_rate": 1.2120864529285203e-06, "loss": 0.049, "step": 27519 }, { "epoch": 8.47, "learning_rate": 1.2116109414553767e-06, "loss": 0.0317, "step": 27520 }, { "epoch": 8.47, "learning_rate": 1.2111355172590366e-06, "loss": 0.0224, "step": 27521 }, { "epoch": 8.47, "learning_rate": 1.2106601803442241e-06, "loss": 0.0283, "step": 27522 }, { "epoch": 8.47, "learning_rate": 1.210184930715652e-06, "loss": 0.0508, "step": 27523 }, { "epoch": 8.47, "learning_rate": 1.2097097683780468e-06, "loss": 0.0452, "step": 27524 }, { "epoch": 8.47, "learning_rate": 1.2092346933361242e-06, "loss": 0.0246, "step": 27525 }, { "epoch": 8.47, "learning_rate": 1.2087597055946032e-06, "loss": 0.021, "step": 27526 }, { "epoch": 8.47, "learning_rate": 1.2082848051582007e-06, "loss": 0.035, "step": 27527 }, { "epoch": 8.47, "learning_rate": 1.2078099920316311e-06, "loss": 0.0408, "step": 27528 }, { "epoch": 8.47, "learning_rate": 1.2073352662196158e-06, "loss": 0.0194, "step": 27529 }, { "epoch": 8.47, "learning_rate": 1.2068606277268591e-06, "loss": 0.0154, "step": 27530 }, { "epoch": 8.47, "learning_rate": 1.2063860765580836e-06, "loss": 0.0181, "step": 27531 }, { "epoch": 8.47, "learning_rate": 1.2059116127179993e-06, "loss": 0.0375, "step": 27532 }, { "epoch": 8.47, "learning_rate": 1.2054372362113164e-06, "loss": 0.0507, "step": 27533 }, { "epoch": 8.47, "learning_rate": 1.2049629470427481e-06, "loss": 0.0187, "step": 27534 }, { "epoch": 8.47, "learning_rate": 1.2044887452170029e-06, "loss": 0.0281, "step": 27535 }, { "epoch": 8.47, "learning_rate": 1.2040146307387901e-06, "loss": 0.0388, "step": 27536 }, { "epoch": 8.47, "learning_rate": 1.2035406036128194e-06, "loss": 0.0154, "step": 27537 }, { "epoch": 8.47, "learning_rate": 1.2030666638437948e-06, "loss": 0.0403, "step": 27538 }, { "epoch": 8.47, "learning_rate": 1.2025928114364306e-06, "loss": 0.0328, "step": 27539 }, { "epoch": 8.47, "learning_rate": 1.2021190463954235e-06, "loss": 0.0136, "step": 27540 }, { "epoch": 8.47, "learning_rate": 1.2016453687254847e-06, "loss": 0.0324, "step": 27541 }, { "epoch": 8.47, "learning_rate": 1.2011717784313159e-06, "loss": 0.0517, "step": 27542 }, { "epoch": 8.47, "learning_rate": 1.2006982755176212e-06, "loss": 0.0296, "step": 27543 }, { "epoch": 8.47, "learning_rate": 1.200224859989101e-06, "loss": 0.0153, "step": 27544 }, { "epoch": 8.48, "learning_rate": 1.199751531850457e-06, "loss": 0.0437, "step": 27545 }, { "epoch": 8.48, "learning_rate": 1.199278291106395e-06, "loss": 0.0863, "step": 27546 }, { "epoch": 8.48, "learning_rate": 1.1988051377616084e-06, "loss": 0.037, "step": 27547 }, { "epoch": 8.48, "learning_rate": 1.198332071820796e-06, "loss": 0.014, "step": 27548 }, { "epoch": 8.48, "learning_rate": 1.1978590932886603e-06, "loss": 0.0191, "step": 27549 }, { "epoch": 8.48, "learning_rate": 1.1973862021698956e-06, "loss": 0.0434, "step": 27550 }, { "epoch": 8.48, "learning_rate": 1.1969133984691983e-06, "loss": 0.0277, "step": 27551 }, { "epoch": 8.48, "learning_rate": 1.1964406821912656e-06, "loss": 0.0761, "step": 27552 }, { "epoch": 8.48, "learning_rate": 1.1959680533407892e-06, "loss": 0.0359, "step": 27553 }, { "epoch": 8.48, "learning_rate": 1.1954955119224644e-06, "loss": 0.0206, "step": 27554 }, { "epoch": 8.48, "learning_rate": 1.1950230579409817e-06, "loss": 0.0233, "step": 27555 }, { "epoch": 8.48, "learning_rate": 1.1945506914010385e-06, "loss": 0.0173, "step": 27556 }, { "epoch": 8.48, "learning_rate": 1.1940784123073213e-06, "loss": 0.0259, "step": 27557 }, { "epoch": 8.48, "learning_rate": 1.1936062206645183e-06, "loss": 0.0204, "step": 27558 }, { "epoch": 8.48, "learning_rate": 1.1931341164773246e-06, "loss": 0.0301, "step": 27559 }, { "epoch": 8.48, "learning_rate": 1.1926620997504245e-06, "loss": 0.0376, "step": 27560 }, { "epoch": 8.48, "learning_rate": 1.1921901704885075e-06, "loss": 0.0353, "step": 27561 }, { "epoch": 8.48, "learning_rate": 1.1917183286962598e-06, "loss": 0.0469, "step": 27562 }, { "epoch": 8.48, "learning_rate": 1.1912465743783674e-06, "loss": 0.0151, "step": 27563 }, { "epoch": 8.48, "learning_rate": 1.1907749075395147e-06, "loss": 0.0366, "step": 27564 }, { "epoch": 8.48, "learning_rate": 1.1903033281843835e-06, "loss": 0.0203, "step": 27565 }, { "epoch": 8.48, "learning_rate": 1.189831836317662e-06, "loss": 0.0211, "step": 27566 }, { "epoch": 8.48, "learning_rate": 1.1893604319440321e-06, "loss": 0.0474, "step": 27567 }, { "epoch": 8.48, "learning_rate": 1.1888891150681681e-06, "loss": 0.0367, "step": 27568 }, { "epoch": 8.48, "learning_rate": 1.1884178856947592e-06, "loss": 0.0435, "step": 27569 }, { "epoch": 8.48, "learning_rate": 1.187946743828481e-06, "loss": 0.0414, "step": 27570 }, { "epoch": 8.48, "learning_rate": 1.1874756894740137e-06, "loss": 0.0366, "step": 27571 }, { "epoch": 8.48, "learning_rate": 1.1870047226360337e-06, "loss": 0.0152, "step": 27572 }, { "epoch": 8.48, "learning_rate": 1.1865338433192175e-06, "loss": 0.0096, "step": 27573 }, { "epoch": 8.48, "learning_rate": 1.1860630515282467e-06, "loss": 0.0383, "step": 27574 }, { "epoch": 8.48, "learning_rate": 1.1855923472677888e-06, "loss": 0.0402, "step": 27575 }, { "epoch": 8.48, "learning_rate": 1.1851217305425256e-06, "loss": 0.0352, "step": 27576 }, { "epoch": 8.48, "learning_rate": 1.1846512013571288e-06, "loss": 0.0315, "step": 27577 }, { "epoch": 8.49, "learning_rate": 1.1841807597162657e-06, "loss": 0.0186, "step": 27578 }, { "epoch": 8.49, "learning_rate": 1.1837104056246152e-06, "loss": 0.0293, "step": 27579 }, { "epoch": 8.49, "learning_rate": 1.1832401390868465e-06, "loss": 0.0415, "step": 27580 }, { "epoch": 8.49, "learning_rate": 1.182769960107628e-06, "loss": 0.0304, "step": 27581 }, { "epoch": 8.49, "learning_rate": 1.1822998686916298e-06, "loss": 0.0346, "step": 27582 }, { "epoch": 8.49, "learning_rate": 1.181829864843519e-06, "loss": 0.0157, "step": 27583 }, { "epoch": 8.49, "learning_rate": 1.1813599485679684e-06, "loss": 0.0314, "step": 27584 }, { "epoch": 8.49, "learning_rate": 1.1808901198696365e-06, "loss": 0.0161, "step": 27585 }, { "epoch": 8.49, "learning_rate": 1.1804203787531965e-06, "loss": 0.0321, "step": 27586 }, { "epoch": 8.49, "learning_rate": 1.17995072522331e-06, "loss": 0.0264, "step": 27587 }, { "epoch": 8.49, "learning_rate": 1.1794811592846423e-06, "loss": 0.0252, "step": 27588 }, { "epoch": 8.49, "learning_rate": 1.179011680941855e-06, "loss": 0.0254, "step": 27589 }, { "epoch": 8.49, "learning_rate": 1.1785422901996102e-06, "loss": 0.0184, "step": 27590 }, { "epoch": 8.49, "learning_rate": 1.178072987062574e-06, "loss": 0.0211, "step": 27591 }, { "epoch": 8.49, "learning_rate": 1.1776037715354017e-06, "loss": 0.0345, "step": 27592 }, { "epoch": 8.49, "learning_rate": 1.1771346436227538e-06, "loss": 0.01, "step": 27593 }, { "epoch": 8.49, "learning_rate": 1.1766656033292933e-06, "loss": 0.0305, "step": 27594 }, { "epoch": 8.49, "learning_rate": 1.1761966506596711e-06, "loss": 0.0203, "step": 27595 }, { "epoch": 8.49, "learning_rate": 1.17572778561855e-06, "loss": 0.0409, "step": 27596 }, { "epoch": 8.49, "learning_rate": 1.1752590082105863e-06, "loss": 0.0414, "step": 27597 }, { "epoch": 8.49, "learning_rate": 1.174790318440432e-06, "loss": 0.031, "step": 27598 }, { "epoch": 8.49, "learning_rate": 1.174321716312744e-06, "loss": 0.0397, "step": 27599 }, { "epoch": 8.49, "learning_rate": 1.1738532018321735e-06, "loss": 0.0182, "step": 27600 }, { "epoch": 8.49, "learning_rate": 1.17338477500338e-06, "loss": 0.0224, "step": 27601 }, { "epoch": 8.49, "learning_rate": 1.1729164358310075e-06, "loss": 0.033, "step": 27602 }, { "epoch": 8.49, "learning_rate": 1.1724481843197078e-06, "loss": 0.022, "step": 27603 }, { "epoch": 8.49, "learning_rate": 1.1719800204741361e-06, "loss": 0.0501, "step": 27604 }, { "epoch": 8.49, "learning_rate": 1.1715119442989387e-06, "loss": 0.0276, "step": 27605 }, { "epoch": 8.49, "learning_rate": 1.1710439557987641e-06, "loss": 0.0267, "step": 27606 }, { "epoch": 8.49, "learning_rate": 1.1705760549782598e-06, "loss": 0.0339, "step": 27607 }, { "epoch": 8.49, "learning_rate": 1.170108241842074e-06, "loss": 0.0278, "step": 27608 }, { "epoch": 8.49, "learning_rate": 1.16964051639485e-06, "loss": 0.0288, "step": 27609 }, { "epoch": 8.5, "learning_rate": 1.1691728786412315e-06, "loss": 0.0311, "step": 27610 }, { "epoch": 8.5, "learning_rate": 1.1687053285858685e-06, "loss": 0.021, "step": 27611 }, { "epoch": 8.5, "learning_rate": 1.1682378662334014e-06, "loss": 0.0237, "step": 27612 }, { "epoch": 8.5, "learning_rate": 1.167770491588469e-06, "loss": 0.044, "step": 27613 }, { "epoch": 8.5, "learning_rate": 1.1673032046557164e-06, "loss": 0.0267, "step": 27614 }, { "epoch": 8.5, "learning_rate": 1.1668360054397842e-06, "loss": 0.0255, "step": 27615 }, { "epoch": 8.5, "learning_rate": 1.1663688939453122e-06, "loss": 0.0177, "step": 27616 }, { "epoch": 8.5, "learning_rate": 1.1659018701769375e-06, "loss": 0.017, "step": 27617 }, { "epoch": 8.5, "learning_rate": 1.165434934139299e-06, "loss": 0.0211, "step": 27618 }, { "epoch": 8.5, "learning_rate": 1.1649680858370337e-06, "loss": 0.0315, "step": 27619 }, { "epoch": 8.5, "learning_rate": 1.164501325274776e-06, "loss": 0.0329, "step": 27620 }, { "epoch": 8.5, "learning_rate": 1.1640346524571643e-06, "loss": 0.0213, "step": 27621 }, { "epoch": 8.5, "learning_rate": 1.163568067388835e-06, "loss": 0.025, "step": 27622 }, { "epoch": 8.5, "learning_rate": 1.1631015700744153e-06, "loss": 0.0308, "step": 27623 }, { "epoch": 8.5, "learning_rate": 1.1626351605185427e-06, "loss": 0.0293, "step": 27624 }, { "epoch": 8.5, "learning_rate": 1.1621688387258468e-06, "loss": 0.0219, "step": 27625 }, { "epoch": 8.5, "learning_rate": 1.1617026047009606e-06, "loss": 0.0284, "step": 27626 }, { "epoch": 8.5, "learning_rate": 1.1612364584485125e-06, "loss": 0.0161, "step": 27627 }, { "epoch": 8.5, "learning_rate": 1.1607703999731312e-06, "loss": 0.0497, "step": 27628 }, { "epoch": 8.5, "learning_rate": 1.1603044292794497e-06, "loss": 0.0246, "step": 27629 }, { "epoch": 8.5, "learning_rate": 1.1598385463720884e-06, "loss": 0.0292, "step": 27630 }, { "epoch": 8.5, "learning_rate": 1.1593727512556797e-06, "loss": 0.0281, "step": 27631 }, { "epoch": 8.5, "learning_rate": 1.158907043934847e-06, "loss": 0.0369, "step": 27632 }, { "epoch": 8.5, "learning_rate": 1.1584414244142162e-06, "loss": 0.0198, "step": 27633 }, { "epoch": 8.5, "learning_rate": 1.1579758926984098e-06, "loss": 0.031, "step": 27634 }, { "epoch": 8.5, "learning_rate": 1.157510448792052e-06, "loss": 0.0229, "step": 27635 }, { "epoch": 8.5, "learning_rate": 1.1570450926997657e-06, "loss": 0.0267, "step": 27636 }, { "epoch": 8.5, "learning_rate": 1.1565798244261706e-06, "loss": 0.0251, "step": 27637 }, { "epoch": 8.5, "learning_rate": 1.1561146439758862e-06, "loss": 0.0434, "step": 27638 }, { "epoch": 8.5, "learning_rate": 1.1556495513535393e-06, "loss": 0.0171, "step": 27639 }, { "epoch": 8.5, "learning_rate": 1.155184546563738e-06, "loss": 0.0409, "step": 27640 }, { "epoch": 8.5, "learning_rate": 1.1547196296111085e-06, "loss": 0.0324, "step": 27641 }, { "epoch": 8.5, "learning_rate": 1.1542548005002651e-06, "loss": 0.036, "step": 27642 }, { "epoch": 8.51, "learning_rate": 1.1537900592358243e-06, "loss": 0.0324, "step": 27643 }, { "epoch": 8.51, "learning_rate": 1.1533254058224008e-06, "loss": 0.0252, "step": 27644 }, { "epoch": 8.51, "learning_rate": 1.152860840264608e-06, "loss": 0.0303, "step": 27645 }, { "epoch": 8.51, "learning_rate": 1.1523963625670643e-06, "loss": 0.0309, "step": 27646 }, { "epoch": 8.51, "learning_rate": 1.151931972734377e-06, "loss": 0.0218, "step": 27647 }, { "epoch": 8.51, "learning_rate": 1.151467670771157e-06, "loss": 0.0435, "step": 27648 }, { "epoch": 8.51, "learning_rate": 1.1510034566820205e-06, "loss": 0.0329, "step": 27649 }, { "epoch": 8.51, "learning_rate": 1.1505393304715763e-06, "loss": 0.0385, "step": 27650 }, { "epoch": 8.51, "learning_rate": 1.1500752921444314e-06, "loss": 0.0271, "step": 27651 }, { "epoch": 8.51, "learning_rate": 1.1496113417051957e-06, "loss": 0.0361, "step": 27652 }, { "epoch": 8.51, "learning_rate": 1.1491474791584755e-06, "loss": 0.0311, "step": 27653 }, { "epoch": 8.51, "learning_rate": 1.1486837045088784e-06, "loss": 0.0239, "step": 27654 }, { "epoch": 8.51, "learning_rate": 1.148220017761007e-06, "loss": 0.0271, "step": 27655 }, { "epoch": 8.51, "learning_rate": 1.1477564189194745e-06, "loss": 0.0316, "step": 27656 }, { "epoch": 8.51, "learning_rate": 1.1472929079888762e-06, "loss": 0.0372, "step": 27657 }, { "epoch": 8.51, "learning_rate": 1.146829484973816e-06, "loss": 0.0457, "step": 27658 }, { "epoch": 8.51, "learning_rate": 1.1463661498789003e-06, "loss": 0.0262, "step": 27659 }, { "epoch": 8.51, "learning_rate": 1.1459029027087288e-06, "loss": 0.0392, "step": 27660 }, { "epoch": 8.51, "learning_rate": 1.1454397434679022e-06, "loss": 0.0339, "step": 27661 }, { "epoch": 8.51, "learning_rate": 1.144976672161019e-06, "loss": 0.0243, "step": 27662 }, { "epoch": 8.51, "learning_rate": 1.1445136887926777e-06, "loss": 0.0282, "step": 27663 }, { "epoch": 8.51, "learning_rate": 1.1440507933674784e-06, "loss": 0.0325, "step": 27664 }, { "epoch": 8.51, "learning_rate": 1.1435879858900134e-06, "loss": 0.0257, "step": 27665 }, { "epoch": 8.51, "learning_rate": 1.1431252663648851e-06, "loss": 0.0284, "step": 27666 }, { "epoch": 8.51, "learning_rate": 1.1426626347966862e-06, "loss": 0.0201, "step": 27667 }, { "epoch": 8.51, "learning_rate": 1.1422000911900078e-06, "loss": 0.0255, "step": 27668 }, { "epoch": 8.51, "learning_rate": 1.141737635549448e-06, "loss": 0.013, "step": 27669 }, { "epoch": 8.51, "learning_rate": 1.1412752678795969e-06, "loss": 0.0245, "step": 27670 }, { "epoch": 8.51, "learning_rate": 1.140812988185047e-06, "loss": 0.0276, "step": 27671 }, { "epoch": 8.51, "learning_rate": 1.1403507964703887e-06, "loss": 0.0358, "step": 27672 }, { "epoch": 8.51, "learning_rate": 1.1398886927402131e-06, "loss": 0.041, "step": 27673 }, { "epoch": 8.51, "learning_rate": 1.139426676999108e-06, "loss": 0.0286, "step": 27674 }, { "epoch": 8.52, "learning_rate": 1.1389647492516598e-06, "loss": 0.0381, "step": 27675 }, { "epoch": 8.52, "learning_rate": 1.1385029095024602e-06, "loss": 0.044, "step": 27676 }, { "epoch": 8.52, "learning_rate": 1.1380411577560968e-06, "loss": 0.0244, "step": 27677 }, { "epoch": 8.52, "learning_rate": 1.137579494017147e-06, "loss": 0.0368, "step": 27678 }, { "epoch": 8.52, "learning_rate": 1.1371179182902037e-06, "loss": 0.0427, "step": 27679 }, { "epoch": 8.52, "learning_rate": 1.1366564305798466e-06, "loss": 0.0649, "step": 27680 }, { "epoch": 8.52, "learning_rate": 1.13619503089066e-06, "loss": 0.0337, "step": 27681 }, { "epoch": 8.52, "learning_rate": 1.1357337192272266e-06, "loss": 0.0179, "step": 27682 }, { "epoch": 8.52, "learning_rate": 1.1352724955941242e-06, "loss": 0.0334, "step": 27683 }, { "epoch": 8.52, "learning_rate": 1.1348113599959409e-06, "loss": 0.0365, "step": 27684 }, { "epoch": 8.52, "learning_rate": 1.1343503124372467e-06, "loss": 0.0289, "step": 27685 }, { "epoch": 8.52, "learning_rate": 1.1338893529226257e-06, "loss": 0.0243, "step": 27686 }, { "epoch": 8.52, "learning_rate": 1.1334284814566565e-06, "loss": 0.0332, "step": 27687 }, { "epoch": 8.52, "learning_rate": 1.132967698043913e-06, "loss": 0.035, "step": 27688 }, { "epoch": 8.52, "learning_rate": 1.132507002688973e-06, "loss": 0.0267, "step": 27689 }, { "epoch": 8.52, "learning_rate": 1.1320463953964078e-06, "loss": 0.0413, "step": 27690 }, { "epoch": 8.52, "learning_rate": 1.1315858761707999e-06, "loss": 0.0247, "step": 27691 }, { "epoch": 8.52, "learning_rate": 1.1311254450167132e-06, "loss": 0.0337, "step": 27692 }, { "epoch": 8.52, "learning_rate": 1.1306651019387261e-06, "loss": 0.0194, "step": 27693 }, { "epoch": 8.52, "learning_rate": 1.1302048469414117e-06, "loss": 0.0217, "step": 27694 }, { "epoch": 8.52, "learning_rate": 1.1297446800293332e-06, "loss": 0.0458, "step": 27695 }, { "epoch": 8.52, "learning_rate": 1.1292846012070679e-06, "loss": 0.0214, "step": 27696 }, { "epoch": 8.52, "learning_rate": 1.1288246104791812e-06, "loss": 0.0289, "step": 27697 }, { "epoch": 8.52, "learning_rate": 1.1283647078502424e-06, "loss": 0.0248, "step": 27698 }, { "epoch": 8.52, "learning_rate": 1.1279048933248182e-06, "loss": 0.0569, "step": 27699 }, { "epoch": 8.52, "learning_rate": 1.1274451669074738e-06, "loss": 0.0472, "step": 27700 }, { "epoch": 8.52, "learning_rate": 1.1269855286027798e-06, "loss": 0.021, "step": 27701 }, { "epoch": 8.52, "learning_rate": 1.1265259784152926e-06, "loss": 0.0288, "step": 27702 }, { "epoch": 8.52, "learning_rate": 1.1260665163495821e-06, "loss": 0.0436, "step": 27703 }, { "epoch": 8.52, "learning_rate": 1.125607142410211e-06, "loss": 0.0168, "step": 27704 }, { "epoch": 8.52, "learning_rate": 1.1251478566017381e-06, "loss": 0.0239, "step": 27705 }, { "epoch": 8.52, "learning_rate": 1.1251478566017381e-06, "loss": 0.0268, "step": 27706 }, { "epoch": 8.52, "learning_rate": 1.1246886589287275e-06, "loss": 0.0267, "step": 27707 }, { "epoch": 8.53, "learning_rate": 1.1242295493957367e-06, "loss": 0.0212, "step": 27708 }, { "epoch": 8.53, "learning_rate": 1.1237705280073298e-06, "loss": 0.034, "step": 27709 }, { "epoch": 8.53, "learning_rate": 1.1233115947680594e-06, "loss": 0.0167, "step": 27710 }, { "epoch": 8.53, "learning_rate": 1.1228527496824848e-06, "loss": 0.0275, "step": 27711 }, { "epoch": 8.53, "learning_rate": 1.1223939927551664e-06, "loss": 0.0179, "step": 27712 }, { "epoch": 8.53, "learning_rate": 1.121935323990656e-06, "loss": 0.0184, "step": 27713 }, { "epoch": 8.53, "learning_rate": 1.1214767433935114e-06, "loss": 0.0195, "step": 27714 }, { "epoch": 8.53, "learning_rate": 1.1210182509682854e-06, "loss": 0.0223, "step": 27715 }, { "epoch": 8.53, "learning_rate": 1.12055984671953e-06, "loss": 0.0267, "step": 27716 }, { "epoch": 8.53, "learning_rate": 1.1201015306518005e-06, "loss": 0.0299, "step": 27717 }, { "epoch": 8.53, "learning_rate": 1.119643302769644e-06, "loss": 0.0254, "step": 27718 }, { "epoch": 8.53, "learning_rate": 1.1191851630776173e-06, "loss": 0.0377, "step": 27719 }, { "epoch": 8.53, "learning_rate": 1.1187271115802656e-06, "loss": 0.0398, "step": 27720 }, { "epoch": 8.53, "learning_rate": 1.1182691482821361e-06, "loss": 0.0211, "step": 27721 }, { "epoch": 8.53, "learning_rate": 1.1178112731877822e-06, "loss": 0.0415, "step": 27722 }, { "epoch": 8.53, "learning_rate": 1.1173534863017488e-06, "loss": 0.0409, "step": 27723 }, { "epoch": 8.53, "learning_rate": 1.11689578762858e-06, "loss": 0.0256, "step": 27724 }, { "epoch": 8.53, "learning_rate": 1.116438177172825e-06, "loss": 0.0361, "step": 27725 }, { "epoch": 8.53, "learning_rate": 1.115980654939025e-06, "loss": 0.0236, "step": 27726 }, { "epoch": 8.53, "learning_rate": 1.1155232209317258e-06, "loss": 0.0204, "step": 27727 }, { "epoch": 8.53, "learning_rate": 1.1150658751554667e-06, "loss": 0.0285, "step": 27728 }, { "epoch": 8.53, "learning_rate": 1.1146086176147942e-06, "loss": 0.0202, "step": 27729 }, { "epoch": 8.53, "learning_rate": 1.1141514483142491e-06, "loss": 0.0377, "step": 27730 }, { "epoch": 8.53, "learning_rate": 1.1136943672583643e-06, "loss": 0.0211, "step": 27731 }, { "epoch": 8.53, "learning_rate": 1.1132373744516878e-06, "loss": 0.038, "step": 27732 }, { "epoch": 8.53, "learning_rate": 1.1127804698987544e-06, "loss": 0.0313, "step": 27733 }, { "epoch": 8.53, "learning_rate": 1.1123236536041016e-06, "loss": 0.0284, "step": 27734 }, { "epoch": 8.53, "learning_rate": 1.1118669255722646e-06, "loss": 0.0189, "step": 27735 }, { "epoch": 8.53, "learning_rate": 1.1114102858077825e-06, "loss": 0.0406, "step": 27736 }, { "epoch": 8.53, "learning_rate": 1.1109537343151877e-06, "loss": 0.0346, "step": 27737 }, { "epoch": 8.53, "learning_rate": 1.1104972710990126e-06, "loss": 0.0285, "step": 27738 }, { "epoch": 8.53, "learning_rate": 1.1100408961637932e-06, "loss": 0.0322, "step": 27739 }, { "epoch": 8.54, "learning_rate": 1.109584609514064e-06, "loss": 0.0345, "step": 27740 }, { "epoch": 8.54, "learning_rate": 1.1091284111543499e-06, "loss": 0.0239, "step": 27741 }, { "epoch": 8.54, "learning_rate": 1.1086723010891852e-06, "loss": 0.0363, "step": 27742 }, { "epoch": 8.54, "learning_rate": 1.108216279323101e-06, "loss": 0.0315, "step": 27743 }, { "epoch": 8.54, "learning_rate": 1.1077603458606224e-06, "loss": 0.0482, "step": 27744 }, { "epoch": 8.54, "learning_rate": 1.10730450070628e-06, "loss": 0.0264, "step": 27745 }, { "epoch": 8.54, "learning_rate": 1.106848743864597e-06, "loss": 0.0218, "step": 27746 }, { "epoch": 8.54, "learning_rate": 1.1063930753401065e-06, "loss": 0.0286, "step": 27747 }, { "epoch": 8.54, "learning_rate": 1.1059374951373248e-06, "loss": 0.0229, "step": 27748 }, { "epoch": 8.54, "learning_rate": 1.1054820032607827e-06, "loss": 0.0266, "step": 27749 }, { "epoch": 8.54, "learning_rate": 1.1050265997150024e-06, "loss": 0.0184, "step": 27750 }, { "epoch": 8.54, "learning_rate": 1.1045712845045053e-06, "loss": 0.0355, "step": 27751 }, { "epoch": 8.54, "learning_rate": 1.1041160576338139e-06, "loss": 0.03, "step": 27752 }, { "epoch": 8.54, "learning_rate": 1.1036609191074488e-06, "loss": 0.0328, "step": 27753 }, { "epoch": 8.54, "learning_rate": 1.1032058689299297e-06, "loss": 0.0385, "step": 27754 }, { "epoch": 8.54, "learning_rate": 1.1027509071057762e-06, "loss": 0.0325, "step": 27755 }, { "epoch": 8.54, "learning_rate": 1.1022960336395039e-06, "loss": 0.0328, "step": 27756 }, { "epoch": 8.54, "learning_rate": 1.1018412485356366e-06, "loss": 0.0241, "step": 27757 }, { "epoch": 8.54, "learning_rate": 1.1013865517986822e-06, "loss": 0.0247, "step": 27758 }, { "epoch": 8.54, "learning_rate": 1.1009319434331623e-06, "loss": 0.0308, "step": 27759 }, { "epoch": 8.54, "learning_rate": 1.10047742344359e-06, "loss": 0.0331, "step": 27760 }, { "epoch": 8.54, "learning_rate": 1.1000229918344796e-06, "loss": 0.0336, "step": 27761 }, { "epoch": 8.54, "learning_rate": 1.099568648610343e-06, "loss": 0.0393, "step": 27762 }, { "epoch": 8.54, "learning_rate": 1.09911439377569e-06, "loss": 0.0281, "step": 27763 }, { "epoch": 8.54, "learning_rate": 1.0986602273350388e-06, "loss": 0.0288, "step": 27764 }, { "epoch": 8.54, "learning_rate": 1.098206149292893e-06, "loss": 0.0312, "step": 27765 }, { "epoch": 8.54, "learning_rate": 1.0977521596537621e-06, "loss": 0.0173, "step": 27766 }, { "epoch": 8.54, "learning_rate": 1.0972982584221592e-06, "loss": 0.0149, "step": 27767 }, { "epoch": 8.54, "learning_rate": 1.0968444456025896e-06, "loss": 0.0157, "step": 27768 }, { "epoch": 8.54, "learning_rate": 1.0963907211995607e-06, "loss": 0.0216, "step": 27769 }, { "epoch": 8.54, "learning_rate": 1.0959370852175766e-06, "loss": 0.0343, "step": 27770 }, { "epoch": 8.54, "learning_rate": 1.095483537661144e-06, "loss": 0.0326, "step": 27771 }, { "epoch": 8.54, "learning_rate": 1.0950300785347668e-06, "loss": 0.0447, "step": 27772 }, { "epoch": 8.55, "learning_rate": 1.0945767078429447e-06, "loss": 0.029, "step": 27773 }, { "epoch": 8.55, "learning_rate": 1.0941234255901888e-06, "loss": 0.0246, "step": 27774 }, { "epoch": 8.55, "learning_rate": 1.0936702317809922e-06, "loss": 0.0228, "step": 27775 }, { "epoch": 8.55, "learning_rate": 1.0932171264198578e-06, "loss": 0.0361, "step": 27776 }, { "epoch": 8.55, "learning_rate": 1.0927641095112874e-06, "loss": 0.0254, "step": 27777 }, { "epoch": 8.55, "learning_rate": 1.0923111810597776e-06, "loss": 0.0294, "step": 27778 }, { "epoch": 8.55, "learning_rate": 1.091858341069828e-06, "loss": 0.0344, "step": 27779 }, { "epoch": 8.55, "learning_rate": 1.0914055895459353e-06, "loss": 0.0282, "step": 27780 }, { "epoch": 8.55, "learning_rate": 1.0909529264925955e-06, "loss": 0.0234, "step": 27781 }, { "epoch": 8.55, "learning_rate": 1.0905003519143042e-06, "loss": 0.036, "step": 27782 }, { "epoch": 8.55, "learning_rate": 1.0900478658155534e-06, "loss": 0.0378, "step": 27783 }, { "epoch": 8.55, "learning_rate": 1.0895954682008402e-06, "loss": 0.0251, "step": 27784 }, { "epoch": 8.55, "learning_rate": 1.0891431590746592e-06, "loss": 0.0356, "step": 27785 }, { "epoch": 8.55, "learning_rate": 1.0886909384414945e-06, "loss": 0.0232, "step": 27786 }, { "epoch": 8.55, "learning_rate": 1.0882388063058437e-06, "loss": 0.0274, "step": 27787 }, { "epoch": 8.55, "learning_rate": 1.087786762672195e-06, "loss": 0.0184, "step": 27788 }, { "epoch": 8.55, "learning_rate": 1.0873348075450374e-06, "loss": 0.0247, "step": 27789 }, { "epoch": 8.55, "learning_rate": 1.0868829409288596e-06, "loss": 0.024, "step": 27790 }, { "epoch": 8.55, "learning_rate": 1.0864311628281465e-06, "loss": 0.0247, "step": 27791 }, { "epoch": 8.55, "learning_rate": 1.0859794732473904e-06, "loss": 0.0362, "step": 27792 }, { "epoch": 8.55, "learning_rate": 1.08552787219107e-06, "loss": 0.0285, "step": 27793 }, { "epoch": 8.55, "learning_rate": 1.085076359663676e-06, "loss": 0.0114, "step": 27794 }, { "epoch": 8.55, "learning_rate": 1.0846249356696925e-06, "loss": 0.0284, "step": 27795 }, { "epoch": 8.55, "learning_rate": 1.084173600213595e-06, "loss": 0.018, "step": 27796 }, { "epoch": 8.55, "learning_rate": 1.0837223532998731e-06, "loss": 0.0245, "step": 27797 }, { "epoch": 8.55, "learning_rate": 1.0832711949330055e-06, "loss": 0.0277, "step": 27798 }, { "epoch": 8.55, "learning_rate": 1.082820125117473e-06, "loss": 0.0345, "step": 27799 }, { "epoch": 8.55, "learning_rate": 1.0823691438577545e-06, "loss": 0.0153, "step": 27800 }, { "epoch": 8.55, "learning_rate": 1.0819182511583282e-06, "loss": 0.0266, "step": 27801 }, { "epoch": 8.55, "learning_rate": 1.0814674470236753e-06, "loss": 0.0254, "step": 27802 }, { "epoch": 8.55, "learning_rate": 1.0810167314582676e-06, "loss": 0.0567, "step": 27803 }, { "epoch": 8.55, "learning_rate": 1.0805661044665849e-06, "loss": 0.0437, "step": 27804 }, { "epoch": 8.56, "learning_rate": 1.0801155660531015e-06, "loss": 0.03, "step": 27805 }, { "epoch": 8.56, "learning_rate": 1.0796651162222916e-06, "loss": 0.0142, "step": 27806 }, { "epoch": 8.56, "learning_rate": 1.0792147549786281e-06, "loss": 0.0182, "step": 27807 }, { "epoch": 8.56, "learning_rate": 1.0787644823265819e-06, "loss": 0.0148, "step": 27808 }, { "epoch": 8.56, "learning_rate": 1.0783142982706297e-06, "loss": 0.0174, "step": 27809 }, { "epoch": 8.56, "learning_rate": 1.0778642028152375e-06, "loss": 0.0172, "step": 27810 }, { "epoch": 8.56, "learning_rate": 1.0774141959648743e-06, "loss": 0.0277, "step": 27811 }, { "epoch": 8.56, "learning_rate": 1.076964277724014e-06, "loss": 0.0226, "step": 27812 }, { "epoch": 8.56, "learning_rate": 1.0765144480971224e-06, "loss": 0.0187, "step": 27813 }, { "epoch": 8.56, "learning_rate": 1.0760647070886654e-06, "loss": 0.0353, "step": 27814 }, { "epoch": 8.56, "learning_rate": 1.075615054703112e-06, "loss": 0.0127, "step": 27815 }, { "epoch": 8.56, "learning_rate": 1.075165490944925e-06, "loss": 0.0317, "step": 27816 }, { "epoch": 8.56, "learning_rate": 1.07471601581857e-06, "loss": 0.0225, "step": 27817 }, { "epoch": 8.56, "learning_rate": 1.0742666293285086e-06, "loss": 0.0209, "step": 27818 }, { "epoch": 8.56, "learning_rate": 1.07381733147921e-06, "loss": 0.0144, "step": 27819 }, { "epoch": 8.56, "learning_rate": 1.0733681222751292e-06, "loss": 0.0442, "step": 27820 }, { "epoch": 8.56, "learning_rate": 1.0729190017207292e-06, "loss": 0.0419, "step": 27821 }, { "epoch": 8.56, "learning_rate": 1.0724699698204722e-06, "loss": 0.0334, "step": 27822 }, { "epoch": 8.56, "learning_rate": 1.0720210265788167e-06, "loss": 0.0311, "step": 27823 }, { "epoch": 8.56, "learning_rate": 1.0715721720002203e-06, "loss": 0.0389, "step": 27824 }, { "epoch": 8.56, "learning_rate": 1.0711234060891407e-06, "loss": 0.0212, "step": 27825 }, { "epoch": 8.56, "learning_rate": 1.0706747288500352e-06, "loss": 0.0282, "step": 27826 }, { "epoch": 8.56, "learning_rate": 1.0702261402873592e-06, "loss": 0.0296, "step": 27827 }, { "epoch": 8.56, "learning_rate": 1.069777640405566e-06, "loss": 0.0411, "step": 27828 }, { "epoch": 8.56, "learning_rate": 1.0693292292091118e-06, "loss": 0.0277, "step": 27829 }, { "epoch": 8.56, "learning_rate": 1.0688809067024519e-06, "loss": 0.0157, "step": 27830 }, { "epoch": 8.56, "learning_rate": 1.0684326728900319e-06, "loss": 0.0249, "step": 27831 }, { "epoch": 8.56, "learning_rate": 1.067984527776309e-06, "loss": 0.0407, "step": 27832 }, { "epoch": 8.56, "learning_rate": 1.067536471365732e-06, "loss": 0.0351, "step": 27833 }, { "epoch": 8.56, "learning_rate": 1.06708850366275e-06, "loss": 0.0492, "step": 27834 }, { "epoch": 8.56, "learning_rate": 1.066640624671812e-06, "loss": 0.0397, "step": 27835 }, { "epoch": 8.56, "learning_rate": 1.0661928343973649e-06, "loss": 0.0451, "step": 27836 }, { "epoch": 8.56, "learning_rate": 1.0657451328438573e-06, "loss": 0.0397, "step": 27837 }, { "epoch": 8.57, "learning_rate": 1.0652975200157333e-06, "loss": 0.0659, "step": 27838 }, { "epoch": 8.57, "learning_rate": 1.0648499959174408e-06, "loss": 0.0295, "step": 27839 }, { "epoch": 8.57, "learning_rate": 1.0644025605534247e-06, "loss": 0.0293, "step": 27840 }, { "epoch": 8.57, "learning_rate": 1.0639552139281217e-06, "loss": 0.0234, "step": 27841 }, { "epoch": 8.57, "learning_rate": 1.0635079560459826e-06, "loss": 0.0335, "step": 27842 }, { "epoch": 8.57, "learning_rate": 1.063060786911444e-06, "loss": 0.02, "step": 27843 }, { "epoch": 8.57, "learning_rate": 1.06261370652895e-06, "loss": 0.0139, "step": 27844 }, { "epoch": 8.57, "learning_rate": 1.062166714902938e-06, "loss": 0.0241, "step": 27845 }, { "epoch": 8.57, "learning_rate": 1.0617198120378459e-06, "loss": 0.052, "step": 27846 }, { "epoch": 8.57, "learning_rate": 1.0612729979381175e-06, "loss": 0.0315, "step": 27847 }, { "epoch": 8.57, "learning_rate": 1.060826272608183e-06, "loss": 0.0407, "step": 27848 }, { "epoch": 8.57, "learning_rate": 1.060379636052483e-06, "loss": 0.0234, "step": 27849 }, { "epoch": 8.57, "learning_rate": 1.0599330882754532e-06, "loss": 0.0352, "step": 27850 }, { "epoch": 8.57, "learning_rate": 1.0594866292815276e-06, "loss": 0.0345, "step": 27851 }, { "epoch": 8.57, "learning_rate": 1.0590402590751392e-06, "loss": 0.0182, "step": 27852 }, { "epoch": 8.57, "learning_rate": 1.0585939776607213e-06, "loss": 0.0154, "step": 27853 }, { "epoch": 8.57, "learning_rate": 1.0581477850427058e-06, "loss": 0.0296, "step": 27854 }, { "epoch": 8.57, "learning_rate": 1.0577016812255237e-06, "loss": 0.0282, "step": 27855 }, { "epoch": 8.57, "learning_rate": 1.0572556662136036e-06, "loss": 0.0196, "step": 27856 }, { "epoch": 8.57, "learning_rate": 1.0568097400113796e-06, "loss": 0.0252, "step": 27857 }, { "epoch": 8.57, "learning_rate": 1.0563639026232742e-06, "loss": 0.0301, "step": 27858 }, { "epoch": 8.57, "learning_rate": 1.05591815405372e-06, "loss": 0.028, "step": 27859 }, { "epoch": 8.57, "learning_rate": 1.0554724943071414e-06, "loss": 0.0302, "step": 27860 }, { "epoch": 8.57, "learning_rate": 1.055026923387965e-06, "loss": 0.0365, "step": 27861 }, { "epoch": 8.57, "learning_rate": 1.0545814413006138e-06, "loss": 0.0357, "step": 27862 }, { "epoch": 8.57, "learning_rate": 1.0541360480495121e-06, "loss": 0.0203, "step": 27863 }, { "epoch": 8.57, "learning_rate": 1.0536907436390886e-06, "loss": 0.0217, "step": 27864 }, { "epoch": 8.57, "learning_rate": 1.0532455280737585e-06, "loss": 0.0256, "step": 27865 }, { "epoch": 8.57, "learning_rate": 1.052800401357944e-06, "loss": 0.0312, "step": 27866 }, { "epoch": 8.57, "learning_rate": 1.0523553634960704e-06, "loss": 0.0274, "step": 27867 }, { "epoch": 8.57, "learning_rate": 1.051910414492553e-06, "loss": 0.0285, "step": 27868 }, { "epoch": 8.57, "learning_rate": 1.0514655543518126e-06, "loss": 0.0181, "step": 27869 }, { "epoch": 8.58, "learning_rate": 1.051020783078267e-06, "loss": 0.0272, "step": 27870 }, { "epoch": 8.58, "learning_rate": 1.0505761006763315e-06, "loss": 0.0355, "step": 27871 }, { "epoch": 8.58, "learning_rate": 1.0501315071504247e-06, "loss": 0.0295, "step": 27872 }, { "epoch": 8.58, "learning_rate": 1.0496870025049578e-06, "loss": 0.0237, "step": 27873 }, { "epoch": 8.58, "learning_rate": 1.0492425867443511e-06, "loss": 0.0205, "step": 27874 }, { "epoch": 8.58, "learning_rate": 1.048798259873014e-06, "loss": 0.0366, "step": 27875 }, { "epoch": 8.58, "learning_rate": 1.048354021895357e-06, "loss": 0.025, "step": 27876 }, { "epoch": 8.58, "learning_rate": 1.0479098728157977e-06, "loss": 0.0264, "step": 27877 }, { "epoch": 8.58, "learning_rate": 1.0474658126387427e-06, "loss": 0.0324, "step": 27878 }, { "epoch": 8.58, "learning_rate": 1.047021841368604e-06, "loss": 0.0263, "step": 27879 }, { "epoch": 8.58, "learning_rate": 1.046577959009789e-06, "loss": 0.0311, "step": 27880 }, { "epoch": 8.58, "learning_rate": 1.0461341655667067e-06, "loss": 0.0214, "step": 27881 }, { "epoch": 8.58, "learning_rate": 1.0456904610437645e-06, "loss": 0.0226, "step": 27882 }, { "epoch": 8.58, "learning_rate": 1.0452468454453656e-06, "loss": 0.0325, "step": 27883 }, { "epoch": 8.58, "learning_rate": 1.044803318775922e-06, "loss": 0.008, "step": 27884 }, { "epoch": 8.58, "learning_rate": 1.0443598810398336e-06, "loss": 0.0127, "step": 27885 }, { "epoch": 8.58, "learning_rate": 1.0439165322415057e-06, "loss": 0.0234, "step": 27886 }, { "epoch": 8.58, "learning_rate": 1.0434732723853414e-06, "loss": 0.0337, "step": 27887 }, { "epoch": 8.58, "learning_rate": 1.0430301014757416e-06, "loss": 0.0189, "step": 27888 }, { "epoch": 8.58, "learning_rate": 1.0425870195171073e-06, "loss": 0.051, "step": 27889 }, { "epoch": 8.58, "learning_rate": 1.0421440265138383e-06, "loss": 0.0299, "step": 27890 }, { "epoch": 8.58, "learning_rate": 1.041701122470339e-06, "loss": 0.0127, "step": 27891 }, { "epoch": 8.58, "learning_rate": 1.041258307391001e-06, "loss": 0.077, "step": 27892 }, { "epoch": 8.58, "learning_rate": 1.0408155812802224e-06, "loss": 0.0175, "step": 27893 }, { "epoch": 8.58, "learning_rate": 1.0403729441424048e-06, "loss": 0.0176, "step": 27894 }, { "epoch": 8.58, "learning_rate": 1.0399303959819408e-06, "loss": 0.0321, "step": 27895 }, { "epoch": 8.58, "learning_rate": 1.0394879368032263e-06, "loss": 0.0204, "step": 27896 }, { "epoch": 8.58, "learning_rate": 1.0390455666106547e-06, "loss": 0.023, "step": 27897 }, { "epoch": 8.58, "learning_rate": 1.0386032854086191e-06, "loss": 0.0195, "step": 27898 }, { "epoch": 8.58, "learning_rate": 1.0381610932015118e-06, "loss": 0.0142, "step": 27899 }, { "epoch": 8.58, "learning_rate": 1.0377189899937234e-06, "loss": 0.0228, "step": 27900 }, { "epoch": 8.58, "learning_rate": 1.037276975789646e-06, "loss": 0.0203, "step": 27901 }, { "epoch": 8.58, "learning_rate": 1.0368350505936708e-06, "loss": 0.0281, "step": 27902 }, { "epoch": 8.59, "learning_rate": 1.0363932144101807e-06, "loss": 0.0346, "step": 27903 }, { "epoch": 8.59, "learning_rate": 1.035951467243569e-06, "loss": 0.0363, "step": 27904 }, { "epoch": 8.59, "learning_rate": 1.0355098090982207e-06, "loss": 0.0377, "step": 27905 }, { "epoch": 8.59, "learning_rate": 1.0350682399785216e-06, "loss": 0.0192, "step": 27906 }, { "epoch": 8.59, "learning_rate": 1.034626759888857e-06, "loss": 0.0444, "step": 27907 }, { "epoch": 8.59, "learning_rate": 1.03418536883361e-06, "loss": 0.0267, "step": 27908 }, { "epoch": 8.59, "learning_rate": 1.0337440668171694e-06, "loss": 0.0177, "step": 27909 }, { "epoch": 8.59, "learning_rate": 1.0333028538439093e-06, "loss": 0.0367, "step": 27910 }, { "epoch": 8.59, "learning_rate": 1.0328617299182187e-06, "loss": 0.028, "step": 27911 }, { "epoch": 8.59, "learning_rate": 1.0324206950444771e-06, "loss": 0.0307, "step": 27912 }, { "epoch": 8.59, "learning_rate": 1.0319797492270577e-06, "loss": 0.0313, "step": 27913 }, { "epoch": 8.59, "learning_rate": 1.0315388924703485e-06, "loss": 0.0374, "step": 27914 }, { "epoch": 8.59, "learning_rate": 1.0310981247787222e-06, "loss": 0.0192, "step": 27915 }, { "epoch": 8.59, "learning_rate": 1.0306574461565577e-06, "loss": 0.0403, "step": 27916 }, { "epoch": 8.59, "learning_rate": 1.0302168566082315e-06, "loss": 0.0221, "step": 27917 }, { "epoch": 8.59, "learning_rate": 1.0297763561381168e-06, "loss": 0.0389, "step": 27918 }, { "epoch": 8.59, "learning_rate": 1.0293359447505936e-06, "loss": 0.0679, "step": 27919 }, { "epoch": 8.59, "learning_rate": 1.028895622450028e-06, "loss": 0.0179, "step": 27920 }, { "epoch": 8.59, "learning_rate": 1.0284553892407988e-06, "loss": 0.027, "step": 27921 }, { "epoch": 8.59, "learning_rate": 1.0280152451272774e-06, "loss": 0.0278, "step": 27922 }, { "epoch": 8.59, "learning_rate": 1.027575190113832e-06, "loss": 0.0265, "step": 27923 }, { "epoch": 8.59, "learning_rate": 1.027135224204835e-06, "loss": 0.0355, "step": 27924 }, { "epoch": 8.59, "learning_rate": 1.0266953474046526e-06, "loss": 0.0365, "step": 27925 }, { "epoch": 8.59, "learning_rate": 1.0262555597176593e-06, "loss": 0.0307, "step": 27926 }, { "epoch": 8.59, "learning_rate": 1.025815861148216e-06, "loss": 0.0309, "step": 27927 }, { "epoch": 8.59, "learning_rate": 1.0253762517006916e-06, "loss": 0.0384, "step": 27928 }, { "epoch": 8.59, "learning_rate": 1.0249367313794535e-06, "loss": 0.0359, "step": 27929 }, { "epoch": 8.59, "learning_rate": 1.0244973001888647e-06, "loss": 0.0485, "step": 27930 }, { "epoch": 8.59, "learning_rate": 1.024057958133291e-06, "loss": 0.0508, "step": 27931 }, { "epoch": 8.59, "learning_rate": 1.023618705217093e-06, "loss": 0.0232, "step": 27932 }, { "epoch": 8.59, "learning_rate": 1.0231795414446356e-06, "loss": 0.0367, "step": 27933 }, { "epoch": 8.59, "learning_rate": 1.022740466820278e-06, "loss": 0.0325, "step": 27934 }, { "epoch": 8.6, "learning_rate": 1.0223014813483788e-06, "loss": 0.0329, "step": 27935 }, { "epoch": 8.6, "learning_rate": 1.021862585033304e-06, "loss": 0.0241, "step": 27936 }, { "epoch": 8.6, "learning_rate": 1.0214237778794068e-06, "loss": 0.0272, "step": 27937 }, { "epoch": 8.6, "learning_rate": 1.020985059891043e-06, "loss": 0.0317, "step": 27938 }, { "epoch": 8.6, "learning_rate": 1.0205464310725765e-06, "loss": 0.0462, "step": 27939 }, { "epoch": 8.6, "learning_rate": 1.020107891428359e-06, "loss": 0.0309, "step": 27940 }, { "epoch": 8.6, "learning_rate": 1.0196694409627462e-06, "loss": 0.0487, "step": 27941 }, { "epoch": 8.6, "learning_rate": 1.019231079680093e-06, "loss": 0.024, "step": 27942 }, { "epoch": 8.6, "learning_rate": 1.0187928075847508e-06, "loss": 0.0531, "step": 27943 }, { "epoch": 8.6, "learning_rate": 1.0183546246810748e-06, "loss": 0.0196, "step": 27944 }, { "epoch": 8.6, "learning_rate": 1.0179165309734118e-06, "loss": 0.0331, "step": 27945 }, { "epoch": 8.6, "learning_rate": 1.017478526466119e-06, "loss": 0.0248, "step": 27946 }, { "epoch": 8.6, "learning_rate": 1.0170406111635445e-06, "loss": 0.0319, "step": 27947 }, { "epoch": 8.6, "learning_rate": 1.0166027850700312e-06, "loss": 0.015, "step": 27948 }, { "epoch": 8.6, "learning_rate": 1.0161650481899344e-06, "loss": 0.0402, "step": 27949 }, { "epoch": 8.6, "learning_rate": 1.0157274005275975e-06, "loss": 0.0284, "step": 27950 }, { "epoch": 8.6, "learning_rate": 1.015289842087368e-06, "loss": 0.025, "step": 27951 }, { "epoch": 8.6, "learning_rate": 1.0148523728735915e-06, "loss": 0.023, "step": 27952 }, { "epoch": 8.6, "learning_rate": 1.014414992890611e-06, "loss": 0.0302, "step": 27953 }, { "epoch": 8.6, "learning_rate": 1.013977702142771e-06, "loss": 0.024, "step": 27954 }, { "epoch": 8.6, "learning_rate": 1.0135405006344134e-06, "loss": 0.053, "step": 27955 }, { "epoch": 8.6, "learning_rate": 1.0131033883698816e-06, "loss": 0.0224, "step": 27956 }, { "epoch": 8.6, "learning_rate": 1.0126663653535184e-06, "loss": 0.0342, "step": 27957 }, { "epoch": 8.6, "learning_rate": 1.0122294315896564e-06, "loss": 0.026, "step": 27958 }, { "epoch": 8.6, "learning_rate": 1.0117925870826417e-06, "loss": 0.0192, "step": 27959 }, { "epoch": 8.6, "learning_rate": 1.011355831836811e-06, "loss": 0.0452, "step": 27960 }, { "epoch": 8.6, "learning_rate": 1.010919165856501e-06, "loss": 0.0367, "step": 27961 }, { "epoch": 8.6, "learning_rate": 1.010482589146048e-06, "loss": 0.0279, "step": 27962 }, { "epoch": 8.6, "learning_rate": 1.0100461017097862e-06, "loss": 0.018, "step": 27963 }, { "epoch": 8.6, "learning_rate": 1.0096097035520548e-06, "loss": 0.0222, "step": 27964 }, { "epoch": 8.6, "learning_rate": 1.009173394677182e-06, "loss": 0.0282, "step": 27965 }, { "epoch": 8.6, "learning_rate": 1.0087371750895058e-06, "loss": 0.0222, "step": 27966 }, { "epoch": 8.6, "learning_rate": 1.008301044793355e-06, "loss": 0.0361, "step": 27967 }, { "epoch": 8.61, "learning_rate": 1.0078650037930627e-06, "loss": 0.0251, "step": 27968 }, { "epoch": 8.61, "learning_rate": 1.0074290520929574e-06, "loss": 0.0141, "step": 27969 }, { "epoch": 8.61, "learning_rate": 1.0069931896973705e-06, "loss": 0.0247, "step": 27970 }, { "epoch": 8.61, "learning_rate": 1.0065574166106285e-06, "loss": 0.0156, "step": 27971 }, { "epoch": 8.61, "learning_rate": 1.0061217328370609e-06, "loss": 0.0386, "step": 27972 }, { "epoch": 8.61, "learning_rate": 1.0056861383809902e-06, "loss": 0.0416, "step": 27973 }, { "epoch": 8.61, "learning_rate": 1.0052506332467503e-06, "loss": 0.0298, "step": 27974 }, { "epoch": 8.61, "learning_rate": 1.0048152174386584e-06, "loss": 0.017, "step": 27975 }, { "epoch": 8.61, "learning_rate": 1.0043798909610425e-06, "loss": 0.0337, "step": 27976 }, { "epoch": 8.61, "learning_rate": 1.003944653818224e-06, "loss": 0.0402, "step": 27977 }, { "epoch": 8.61, "learning_rate": 1.0035095060145272e-06, "loss": 0.0308, "step": 27978 }, { "epoch": 8.61, "learning_rate": 1.003074447554273e-06, "loss": 0.0417, "step": 27979 }, { "epoch": 8.61, "learning_rate": 1.002639478441778e-06, "loss": 0.0658, "step": 27980 }, { "epoch": 8.61, "learning_rate": 1.0022045986813699e-06, "loss": 0.0257, "step": 27981 }, { "epoch": 8.61, "learning_rate": 1.0017698082773608e-06, "loss": 0.0416, "step": 27982 }, { "epoch": 8.61, "learning_rate": 1.0013351072340683e-06, "loss": 0.0288, "step": 27983 }, { "epoch": 8.61, "learning_rate": 1.0009004955558144e-06, "loss": 0.0217, "step": 27984 }, { "epoch": 8.61, "learning_rate": 1.0004659732469113e-06, "loss": 0.0137, "step": 27985 }, { "epoch": 8.61, "learning_rate": 1.0000315403116755e-06, "loss": 0.0436, "step": 27986 }, { "epoch": 8.61, "learning_rate": 9.995971967544217e-07, "loss": 0.0218, "step": 27987 }, { "epoch": 8.61, "learning_rate": 9.991629425794624e-07, "loss": 0.0356, "step": 27988 }, { "epoch": 8.61, "learning_rate": 9.987287777911103e-07, "loss": 0.0339, "step": 27989 }, { "epoch": 8.61, "learning_rate": 9.982947023936752e-07, "loss": 0.026, "step": 27990 }, { "epoch": 8.61, "learning_rate": 9.978607163914745e-07, "loss": 0.0717, "step": 27991 }, { "epoch": 8.61, "learning_rate": 9.974268197888104e-07, "loss": 0.0314, "step": 27992 }, { "epoch": 8.61, "learning_rate": 9.96993012589994e-07, "loss": 0.0341, "step": 27993 }, { "epoch": 8.61, "learning_rate": 9.965592947993364e-07, "loss": 0.0345, "step": 27994 }, { "epoch": 8.61, "learning_rate": 9.961256664211427e-07, "loss": 0.0323, "step": 27995 }, { "epoch": 8.61, "learning_rate": 9.956921274597186e-07, "loss": 0.0258, "step": 27996 }, { "epoch": 8.61, "learning_rate": 9.952586779193718e-07, "loss": 0.022, "step": 27997 }, { "epoch": 8.61, "learning_rate": 9.948253178044043e-07, "loss": 0.0481, "step": 27998 }, { "epoch": 8.61, "learning_rate": 9.943920471191204e-07, "loss": 0.0323, "step": 27999 }, { "epoch": 8.62, "learning_rate": 9.939588658678222e-07, "loss": 0.0259, "step": 28000 }, { "epoch": 8.62, "learning_rate": 9.935257740548143e-07, "loss": 0.0204, "step": 28001 }, { "epoch": 8.62, "learning_rate": 9.930927716843975e-07, "loss": 0.0459, "step": 28002 }, { "epoch": 8.62, "learning_rate": 9.926598587608672e-07, "loss": 0.0433, "step": 28003 }, { "epoch": 8.62, "learning_rate": 9.922270352885277e-07, "loss": 0.0229, "step": 28004 }, { "epoch": 8.62, "learning_rate": 9.917943012716758e-07, "loss": 0.0293, "step": 28005 }, { "epoch": 8.62, "learning_rate": 9.913616567146089e-07, "loss": 0.0101, "step": 28006 }, { "epoch": 8.62, "learning_rate": 9.909291016216228e-07, "loss": 0.0346, "step": 28007 }, { "epoch": 8.62, "learning_rate": 9.904966359970114e-07, "loss": 0.0173, "step": 28008 }, { "epoch": 8.62, "learning_rate": 9.90064259845077e-07, "loss": 0.0166, "step": 28009 }, { "epoch": 8.62, "learning_rate": 9.896319731701031e-07, "loss": 0.0192, "step": 28010 }, { "epoch": 8.62, "learning_rate": 9.891997759763905e-07, "loss": 0.0361, "step": 28011 }, { "epoch": 8.62, "learning_rate": 9.887676682682312e-07, "loss": 0.0256, "step": 28012 }, { "epoch": 8.62, "learning_rate": 9.883356500499085e-07, "loss": 0.026, "step": 28013 }, { "epoch": 8.62, "learning_rate": 9.879037213257214e-07, "loss": 0.0374, "step": 28014 }, { "epoch": 8.62, "learning_rate": 9.874718820999562e-07, "loss": 0.0464, "step": 28015 }, { "epoch": 8.62, "learning_rate": 9.870401323769008e-07, "loss": 0.0285, "step": 28016 }, { "epoch": 8.62, "learning_rate": 9.866084721608438e-07, "loss": 0.0265, "step": 28017 }, { "epoch": 8.62, "learning_rate": 9.861769014560696e-07, "loss": 0.0477, "step": 28018 }, { "epoch": 8.62, "learning_rate": 9.857454202668693e-07, "loss": 0.012, "step": 28019 }, { "epoch": 8.62, "learning_rate": 9.853140285975226e-07, "loss": 0.0312, "step": 28020 }, { "epoch": 8.62, "learning_rate": 9.848827264523163e-07, "loss": 0.0513, "step": 28021 }, { "epoch": 8.62, "learning_rate": 9.844515138355337e-07, "loss": 0.0184, "step": 28022 }, { "epoch": 8.62, "learning_rate": 9.840203907514555e-07, "loss": 0.0379, "step": 28023 }, { "epoch": 8.62, "learning_rate": 9.835893572043654e-07, "loss": 0.0291, "step": 28024 }, { "epoch": 8.62, "learning_rate": 9.831584131985405e-07, "loss": 0.0503, "step": 28025 }, { "epoch": 8.62, "learning_rate": 9.827275587382667e-07, "loss": 0.0338, "step": 28026 }, { "epoch": 8.62, "learning_rate": 9.822967938278172e-07, "loss": 0.0388, "step": 28027 }, { "epoch": 8.62, "learning_rate": 9.818661184714685e-07, "loss": 0.035, "step": 28028 }, { "epoch": 8.62, "learning_rate": 9.814355326735036e-07, "loss": 0.0385, "step": 28029 }, { "epoch": 8.62, "learning_rate": 9.810050364381952e-07, "loss": 0.0311, "step": 28030 }, { "epoch": 8.62, "learning_rate": 9.805746297698182e-07, "loss": 0.0316, "step": 28031 }, { "epoch": 8.62, "learning_rate": 9.801443126726485e-07, "loss": 0.0353, "step": 28032 }, { "epoch": 8.63, "learning_rate": 9.79714085150959e-07, "loss": 0.0211, "step": 28033 }, { "epoch": 8.63, "learning_rate": 9.79283947209021e-07, "loss": 0.0319, "step": 28034 }, { "epoch": 8.63, "learning_rate": 9.788538988511064e-07, "loss": 0.0236, "step": 28035 }, { "epoch": 8.63, "learning_rate": 9.7842394008149e-07, "loss": 0.0222, "step": 28036 }, { "epoch": 8.63, "learning_rate": 9.779940709044367e-07, "loss": 0.0353, "step": 28037 }, { "epoch": 8.63, "learning_rate": 9.775642913242145e-07, "loss": 0.0425, "step": 28038 }, { "epoch": 8.63, "learning_rate": 9.771346013450967e-07, "loss": 0.0403, "step": 28039 }, { "epoch": 8.63, "learning_rate": 9.767050009713476e-07, "loss": 0.0334, "step": 28040 }, { "epoch": 8.63, "learning_rate": 9.762754902072348e-07, "loss": 0.0236, "step": 28041 }, { "epoch": 8.63, "learning_rate": 9.758460690570227e-07, "loss": 0.0468, "step": 28042 }, { "epoch": 8.63, "learning_rate": 9.754167375249756e-07, "loss": 0.0233, "step": 28043 }, { "epoch": 8.63, "learning_rate": 9.749874956153572e-07, "loss": 0.0263, "step": 28044 }, { "epoch": 8.63, "learning_rate": 9.74558343332429e-07, "loss": 0.0317, "step": 28045 }, { "epoch": 8.63, "learning_rate": 9.74129280680457e-07, "loss": 0.0562, "step": 28046 }, { "epoch": 8.63, "learning_rate": 9.73700307663702e-07, "loss": 0.0098, "step": 28047 }, { "epoch": 8.63, "learning_rate": 9.732714242864172e-07, "loss": 0.0149, "step": 28048 }, { "epoch": 8.63, "learning_rate": 9.728426305528682e-07, "loss": 0.0421, "step": 28049 }, { "epoch": 8.63, "learning_rate": 9.724139264673116e-07, "loss": 0.0548, "step": 28050 }, { "epoch": 8.63, "learning_rate": 9.71985312034005e-07, "loss": 0.0279, "step": 28051 }, { "epoch": 8.63, "learning_rate": 9.715567872572051e-07, "loss": 0.0224, "step": 28052 }, { "epoch": 8.63, "learning_rate": 9.711283521411674e-07, "loss": 0.0381, "step": 28053 }, { "epoch": 8.63, "learning_rate": 9.707000066901452e-07, "loss": 0.0493, "step": 28054 }, { "epoch": 8.63, "learning_rate": 9.702717509083925e-07, "loss": 0.044, "step": 28055 }, { "epoch": 8.63, "learning_rate": 9.698435848001653e-07, "loss": 0.0353, "step": 28056 }, { "epoch": 8.63, "learning_rate": 9.694155083697154e-07, "loss": 0.036, "step": 28057 }, { "epoch": 8.63, "learning_rate": 9.689875216212886e-07, "loss": 0.032, "step": 28058 }, { "epoch": 8.63, "learning_rate": 9.68559624559141e-07, "loss": 0.0322, "step": 28059 }, { "epoch": 8.63, "learning_rate": 9.681318171875198e-07, "loss": 0.0347, "step": 28060 }, { "epoch": 8.63, "learning_rate": 9.677040995106734e-07, "loss": 0.0444, "step": 28061 }, { "epoch": 8.63, "learning_rate": 9.672764715328497e-07, "loss": 0.0388, "step": 28062 }, { "epoch": 8.63, "learning_rate": 9.668489332582942e-07, "loss": 0.0278, "step": 28063 }, { "epoch": 8.63, "learning_rate": 9.664214846912568e-07, "loss": 0.022, "step": 28064 }, { "epoch": 8.64, "learning_rate": 9.659941258359761e-07, "loss": 0.0219, "step": 28065 }, { "epoch": 8.64, "learning_rate": 9.655668566967026e-07, "loss": 0.0443, "step": 28066 }, { "epoch": 8.64, "learning_rate": 9.651396772776755e-07, "loss": 0.0343, "step": 28067 }, { "epoch": 8.64, "learning_rate": 9.647125875831386e-07, "loss": 0.0488, "step": 28068 }, { "epoch": 8.64, "learning_rate": 9.64285587617333e-07, "loss": 0.0473, "step": 28069 }, { "epoch": 8.64, "learning_rate": 9.638586773844993e-07, "loss": 0.0253, "step": 28070 }, { "epoch": 8.64, "learning_rate": 9.634318568888767e-07, "loss": 0.0159, "step": 28071 }, { "epoch": 8.64, "learning_rate": 9.63005126134704e-07, "loss": 0.029, "step": 28072 }, { "epoch": 8.64, "learning_rate": 9.625784851262165e-07, "loss": 0.0306, "step": 28073 }, { "epoch": 8.64, "learning_rate": 9.621519338676587e-07, "loss": 0.0318, "step": 28074 }, { "epoch": 8.64, "learning_rate": 9.617254723632574e-07, "loss": 0.0416, "step": 28075 }, { "epoch": 8.64, "learning_rate": 9.612991006172534e-07, "loss": 0.0252, "step": 28076 }, { "epoch": 8.64, "learning_rate": 9.608728186338812e-07, "loss": 0.0364, "step": 28077 }, { "epoch": 8.64, "learning_rate": 9.604466264173718e-07, "loss": 0.011, "step": 28078 }, { "epoch": 8.64, "learning_rate": 9.600205239719584e-07, "loss": 0.0325, "step": 28079 }, { "epoch": 8.64, "learning_rate": 9.5959451130187e-07, "loss": 0.0425, "step": 28080 }, { "epoch": 8.64, "learning_rate": 9.591685884113444e-07, "loss": 0.0327, "step": 28081 }, { "epoch": 8.64, "learning_rate": 9.587427553046036e-07, "loss": 0.0199, "step": 28082 }, { "epoch": 8.64, "learning_rate": 9.58317011985882e-07, "loss": 0.0374, "step": 28083 }, { "epoch": 8.64, "learning_rate": 9.578913584594052e-07, "loss": 0.0317, "step": 28084 }, { "epoch": 8.64, "learning_rate": 9.574657947294008e-07, "loss": 0.0469, "step": 28085 }, { "epoch": 8.64, "learning_rate": 9.570403208000945e-07, "loss": 0.0317, "step": 28086 }, { "epoch": 8.64, "learning_rate": 9.566149366757104e-07, "loss": 0.0229, "step": 28087 }, { "epoch": 8.64, "learning_rate": 9.561896423604777e-07, "loss": 0.0354, "step": 28088 }, { "epoch": 8.64, "learning_rate": 9.557644378586162e-07, "loss": 0.0304, "step": 28089 }, { "epoch": 8.64, "learning_rate": 9.55339323174347e-07, "loss": 0.0251, "step": 28090 }, { "epoch": 8.64, "learning_rate": 9.549142983118976e-07, "loss": 0.0403, "step": 28091 }, { "epoch": 8.64, "learning_rate": 9.544893632754816e-07, "loss": 0.014, "step": 28092 }, { "epoch": 8.64, "learning_rate": 9.540645180693243e-07, "loss": 0.036, "step": 28093 }, { "epoch": 8.64, "learning_rate": 9.536397626976434e-07, "loss": 0.0277, "step": 28094 }, { "epoch": 8.64, "learning_rate": 9.532150971646581e-07, "loss": 0.03, "step": 28095 }, { "epoch": 8.64, "learning_rate": 9.527905214745836e-07, "loss": 0.0213, "step": 28096 }, { "epoch": 8.64, "learning_rate": 9.523660356316366e-07, "loss": 0.0285, "step": 28097 }, { "epoch": 8.65, "learning_rate": 9.519416396400361e-07, "loss": 0.017, "step": 28098 }, { "epoch": 8.65, "learning_rate": 9.515173335039929e-07, "loss": 0.016, "step": 28099 }, { "epoch": 8.65, "learning_rate": 9.510931172277193e-07, "loss": 0.0329, "step": 28100 }, { "epoch": 8.65, "learning_rate": 9.506689908154331e-07, "loss": 0.0532, "step": 28101 }, { "epoch": 8.65, "learning_rate": 9.502449542713432e-07, "loss": 0.0304, "step": 28102 }, { "epoch": 8.65, "learning_rate": 9.498210075996628e-07, "loss": 0.0303, "step": 28103 }, { "epoch": 8.65, "learning_rate": 9.493971508045996e-07, "loss": 0.0222, "step": 28104 }, { "epoch": 8.65, "learning_rate": 9.489733838903648e-07, "loss": 0.0189, "step": 28105 }, { "epoch": 8.65, "learning_rate": 9.485497068611648e-07, "loss": 0.0246, "step": 28106 }, { "epoch": 8.65, "learning_rate": 9.481261197212077e-07, "loss": 0.0164, "step": 28107 }, { "epoch": 8.65, "learning_rate": 9.47702622474701e-07, "loss": 0.0355, "step": 28108 }, { "epoch": 8.65, "learning_rate": 9.472792151258536e-07, "loss": 0.0287, "step": 28109 }, { "epoch": 8.65, "learning_rate": 9.468558976788622e-07, "loss": 0.0342, "step": 28110 }, { "epoch": 8.65, "learning_rate": 9.464326701379368e-07, "loss": 0.0207, "step": 28111 }, { "epoch": 8.65, "learning_rate": 9.460095325072805e-07, "loss": 0.0341, "step": 28112 }, { "epoch": 8.65, "learning_rate": 9.455864847910922e-07, "loss": 0.0399, "step": 28113 }, { "epoch": 8.65, "learning_rate": 9.451635269935766e-07, "loss": 0.0437, "step": 28114 }, { "epoch": 8.65, "learning_rate": 9.447406591189312e-07, "loss": 0.0247, "step": 28115 }, { "epoch": 8.65, "learning_rate": 9.443178811713572e-07, "loss": 0.0225, "step": 28116 }, { "epoch": 8.65, "learning_rate": 9.438951931550499e-07, "loss": 0.0487, "step": 28117 }, { "epoch": 8.65, "learning_rate": 9.434725950742119e-07, "loss": 0.0179, "step": 28118 }, { "epoch": 8.65, "learning_rate": 9.430500869330406e-07, "loss": 0.0352, "step": 28119 }, { "epoch": 8.65, "learning_rate": 9.426276687357239e-07, "loss": 0.0105, "step": 28120 }, { "epoch": 8.65, "learning_rate": 9.42205340486465e-07, "loss": 0.0384, "step": 28121 }, { "epoch": 8.65, "learning_rate": 9.417831021894541e-07, "loss": 0.0392, "step": 28122 }, { "epoch": 8.65, "learning_rate": 9.413609538488866e-07, "loss": 0.0288, "step": 28123 }, { "epoch": 8.65, "learning_rate": 9.409388954689524e-07, "loss": 0.0203, "step": 28124 }, { "epoch": 8.65, "learning_rate": 9.405169270538417e-07, "loss": 0.0258, "step": 28125 }, { "epoch": 8.65, "learning_rate": 9.40095048607752e-07, "loss": 0.0235, "step": 28126 }, { "epoch": 8.65, "learning_rate": 9.396732601348645e-07, "loss": 0.023, "step": 28127 }, { "epoch": 8.65, "learning_rate": 9.392515616393738e-07, "loss": 0.0233, "step": 28128 }, { "epoch": 8.65, "learning_rate": 9.388299531254652e-07, "loss": 0.0282, "step": 28129 }, { "epoch": 8.66, "learning_rate": 9.384084345973266e-07, "loss": 0.0334, "step": 28130 }, { "epoch": 8.66, "learning_rate": 9.379870060591434e-07, "loss": 0.0435, "step": 28131 }, { "epoch": 8.66, "learning_rate": 9.375656675151013e-07, "loss": 0.0286, "step": 28132 }, { "epoch": 8.66, "learning_rate": 9.371444189693846e-07, "loss": 0.0396, "step": 28133 }, { "epoch": 8.66, "learning_rate": 9.367232604261756e-07, "loss": 0.0356, "step": 28134 }, { "epoch": 8.66, "learning_rate": 9.363021918896553e-07, "loss": 0.0268, "step": 28135 }, { "epoch": 8.66, "learning_rate": 9.358812133640127e-07, "loss": 0.03, "step": 28136 }, { "epoch": 8.66, "learning_rate": 9.354603248534189e-07, "loss": 0.0096, "step": 28137 }, { "epoch": 8.66, "learning_rate": 9.350395263620604e-07, "loss": 0.0268, "step": 28138 }, { "epoch": 8.66, "learning_rate": 9.34618817894114e-07, "loss": 0.0116, "step": 28139 }, { "epoch": 8.66, "learning_rate": 9.341981994537563e-07, "loss": 0.0336, "step": 28140 }, { "epoch": 8.66, "learning_rate": 9.337776710451674e-07, "loss": 0.044, "step": 28141 }, { "epoch": 8.66, "learning_rate": 9.333572326725193e-07, "loss": 0.0319, "step": 28142 }, { "epoch": 8.66, "learning_rate": 9.329368843399944e-07, "loss": 0.0261, "step": 28143 }, { "epoch": 8.66, "learning_rate": 9.325166260517593e-07, "loss": 0.0425, "step": 28144 }, { "epoch": 8.66, "learning_rate": 9.320964578119907e-07, "loss": 0.035, "step": 28145 }, { "epoch": 8.66, "learning_rate": 9.316763796248629e-07, "loss": 0.0178, "step": 28146 }, { "epoch": 8.66, "learning_rate": 9.312563914945461e-07, "loss": 0.0438, "step": 28147 }, { "epoch": 8.66, "learning_rate": 9.308364934252113e-07, "loss": 0.0316, "step": 28148 }, { "epoch": 8.66, "learning_rate": 9.304166854210284e-07, "loss": 0.03, "step": 28149 }, { "epoch": 8.66, "learning_rate": 9.299969674861664e-07, "loss": 0.0414, "step": 28150 }, { "epoch": 8.66, "learning_rate": 9.295773396247943e-07, "loss": 0.0294, "step": 28151 }, { "epoch": 8.66, "learning_rate": 9.291578018410762e-07, "loss": 0.0381, "step": 28152 }, { "epoch": 8.66, "learning_rate": 9.287383541391848e-07, "loss": 0.0331, "step": 28153 }, { "epoch": 8.66, "learning_rate": 9.283189965232808e-07, "loss": 0.0268, "step": 28154 }, { "epoch": 8.66, "learning_rate": 9.278997289975277e-07, "loss": 0.0292, "step": 28155 }, { "epoch": 8.66, "learning_rate": 9.274805515660923e-07, "loss": 0.0443, "step": 28156 }, { "epoch": 8.66, "learning_rate": 9.270614642331377e-07, "loss": 0.0262, "step": 28157 }, { "epoch": 8.66, "learning_rate": 9.266424670028241e-07, "loss": 0.0205, "step": 28158 }, { "epoch": 8.66, "learning_rate": 9.262235598793134e-07, "loss": 0.0243, "step": 28159 }, { "epoch": 8.66, "learning_rate": 9.258047428667649e-07, "loss": 0.0215, "step": 28160 }, { "epoch": 8.66, "learning_rate": 9.253860159693384e-07, "loss": 0.0354, "step": 28161 }, { "epoch": 8.66, "learning_rate": 9.249673791911907e-07, "loss": 0.0191, "step": 28162 }, { "epoch": 8.67, "learning_rate": 9.245488325364826e-07, "loss": 0.0287, "step": 28163 }, { "epoch": 8.67, "learning_rate": 9.241303760093712e-07, "loss": 0.0177, "step": 28164 }, { "epoch": 8.67, "learning_rate": 9.237120096140051e-07, "loss": 0.0285, "step": 28165 }, { "epoch": 8.67, "learning_rate": 9.232937333545455e-07, "loss": 0.0285, "step": 28166 }, { "epoch": 8.67, "learning_rate": 9.228755472351447e-07, "loss": 0.0147, "step": 28167 }, { "epoch": 8.67, "learning_rate": 9.224574512599549e-07, "loss": 0.0304, "step": 28168 }, { "epoch": 8.67, "learning_rate": 9.220394454331294e-07, "loss": 0.0264, "step": 28169 }, { "epoch": 8.67, "learning_rate": 9.216215297588182e-07, "loss": 0.0358, "step": 28170 }, { "epoch": 8.67, "learning_rate": 9.212037042411726e-07, "loss": 0.0446, "step": 28171 }, { "epoch": 8.67, "learning_rate": 9.207859688843402e-07, "loss": 0.032, "step": 28172 }, { "epoch": 8.67, "learning_rate": 9.20368323692471e-07, "loss": 0.0192, "step": 28173 }, { "epoch": 8.67, "learning_rate": 9.199507686697151e-07, "loss": 0.0424, "step": 28174 }, { "epoch": 8.67, "learning_rate": 9.195333038202137e-07, "loss": 0.03, "step": 28175 }, { "epoch": 8.67, "learning_rate": 9.191159291481156e-07, "loss": 0.0271, "step": 28176 }, { "epoch": 8.67, "learning_rate": 9.186986446575663e-07, "loss": 0.022, "step": 28177 }, { "epoch": 8.67, "learning_rate": 9.182814503527093e-07, "loss": 0.0438, "step": 28178 }, { "epoch": 8.67, "learning_rate": 9.178643462376868e-07, "loss": 0.0427, "step": 28179 }, { "epoch": 8.67, "learning_rate": 9.174473323166399e-07, "loss": 0.0448, "step": 28180 }, { "epoch": 8.67, "learning_rate": 9.170304085937165e-07, "loss": 0.0192, "step": 28181 }, { "epoch": 8.67, "learning_rate": 9.166135750730476e-07, "loss": 0.0144, "step": 28182 }, { "epoch": 8.67, "learning_rate": 9.161968317587788e-07, "loss": 0.0435, "step": 28183 }, { "epoch": 8.67, "learning_rate": 9.15780178655048e-07, "loss": 0.0126, "step": 28184 }, { "epoch": 8.67, "learning_rate": 9.153636157659929e-07, "loss": 0.0397, "step": 28185 }, { "epoch": 8.67, "learning_rate": 9.149471430957491e-07, "loss": 0.0393, "step": 28186 }, { "epoch": 8.67, "learning_rate": 9.14530760648451e-07, "loss": 0.038, "step": 28187 }, { "epoch": 8.67, "learning_rate": 9.141144684282399e-07, "loss": 0.0288, "step": 28188 }, { "epoch": 8.67, "learning_rate": 9.136982664392446e-07, "loss": 0.0488, "step": 28189 }, { "epoch": 8.67, "learning_rate": 9.132821546855964e-07, "loss": 0.0292, "step": 28190 }, { "epoch": 8.67, "learning_rate": 9.128661331714361e-07, "loss": 0.0232, "step": 28191 }, { "epoch": 8.67, "learning_rate": 9.124502019008863e-07, "loss": 0.0206, "step": 28192 }, { "epoch": 8.67, "learning_rate": 9.120343608780824e-07, "loss": 0.0209, "step": 28193 }, { "epoch": 8.67, "learning_rate": 9.116186101071533e-07, "loss": 0.0257, "step": 28194 }, { "epoch": 8.68, "learning_rate": 9.11202949592227e-07, "loss": 0.042, "step": 28195 }, { "epoch": 8.68, "learning_rate": 9.107873793374322e-07, "loss": 0.0122, "step": 28196 }, { "epoch": 8.68, "learning_rate": 9.103718993468946e-07, "loss": 0.0499, "step": 28197 }, { "epoch": 8.68, "learning_rate": 9.099565096247443e-07, "loss": 0.0314, "step": 28198 }, { "epoch": 8.68, "learning_rate": 9.095412101751011e-07, "loss": 0.0395, "step": 28199 }, { "epoch": 8.68, "learning_rate": 9.091260010020908e-07, "loss": 0.0229, "step": 28200 }, { "epoch": 8.68, "learning_rate": 9.08710882109839e-07, "loss": 0.0264, "step": 28201 }, { "epoch": 8.68, "learning_rate": 9.082958535024677e-07, "loss": 0.0287, "step": 28202 }, { "epoch": 8.68, "learning_rate": 9.078809151840972e-07, "loss": 0.0214, "step": 28203 }, { "epoch": 8.68, "learning_rate": 9.074660671588487e-07, "loss": 0.0411, "step": 28204 }, { "epoch": 8.68, "learning_rate": 9.07051309430842e-07, "loss": 0.0332, "step": 28205 }, { "epoch": 8.68, "learning_rate": 9.066366420041961e-07, "loss": 0.0205, "step": 28206 }, { "epoch": 8.68, "learning_rate": 9.062220648830267e-07, "loss": 0.0123, "step": 28207 }, { "epoch": 8.68, "learning_rate": 9.058075780714582e-07, "loss": 0.0352, "step": 28208 }, { "epoch": 8.68, "learning_rate": 9.053931815735995e-07, "loss": 0.0367, "step": 28209 }, { "epoch": 8.68, "learning_rate": 9.049788753935662e-07, "loss": 0.0243, "step": 28210 }, { "epoch": 8.68, "learning_rate": 9.045646595354762e-07, "loss": 0.0469, "step": 28211 }, { "epoch": 8.68, "learning_rate": 9.041505340034429e-07, "loss": 0.0384, "step": 28212 }, { "epoch": 8.68, "learning_rate": 9.037364988015773e-07, "loss": 0.0266, "step": 28213 }, { "epoch": 8.68, "learning_rate": 9.033225539339919e-07, "loss": 0.0453, "step": 28214 }, { "epoch": 8.68, "learning_rate": 9.029086994047964e-07, "loss": 0.0375, "step": 28215 }, { "epoch": 8.68, "learning_rate": 9.024949352181023e-07, "loss": 0.0328, "step": 28216 }, { "epoch": 8.68, "learning_rate": 9.020812613780161e-07, "loss": 0.0164, "step": 28217 }, { "epoch": 8.68, "learning_rate": 9.016676778886502e-07, "loss": 0.0161, "step": 28218 }, { "epoch": 8.68, "learning_rate": 9.012541847541112e-07, "loss": 0.0318, "step": 28219 }, { "epoch": 8.68, "learning_rate": 9.008407819784993e-07, "loss": 0.0107, "step": 28220 }, { "epoch": 8.68, "learning_rate": 9.004274695659265e-07, "loss": 0.0354, "step": 28221 }, { "epoch": 8.68, "learning_rate": 9.000142475204965e-07, "loss": 0.0306, "step": 28222 }, { "epoch": 8.68, "learning_rate": 8.996011158463103e-07, "loss": 0.0306, "step": 28223 }, { "epoch": 8.68, "learning_rate": 8.991880745474734e-07, "loss": 0.0208, "step": 28224 }, { "epoch": 8.68, "learning_rate": 8.98775123628084e-07, "loss": 0.049, "step": 28225 }, { "epoch": 8.68, "learning_rate": 8.983622630922495e-07, "loss": 0.0365, "step": 28226 }, { "epoch": 8.68, "learning_rate": 8.979494929440625e-07, "loss": 0.025, "step": 28227 }, { "epoch": 8.69, "learning_rate": 8.975368131876272e-07, "loss": 0.0214, "step": 28228 }, { "epoch": 8.69, "learning_rate": 8.971242238270416e-07, "loss": 0.0329, "step": 28229 }, { "epoch": 8.69, "learning_rate": 8.967117248663992e-07, "loss": 0.0356, "step": 28230 }, { "epoch": 8.69, "learning_rate": 8.962993163098011e-07, "loss": 0.0438, "step": 28231 }, { "epoch": 8.69, "learning_rate": 8.958869981613405e-07, "loss": 0.0213, "step": 28232 }, { "epoch": 8.69, "learning_rate": 8.954747704251132e-07, "loss": 0.0397, "step": 28233 }, { "epoch": 8.69, "learning_rate": 8.950626331052114e-07, "loss": 0.0268, "step": 28234 }, { "epoch": 8.69, "learning_rate": 8.946505862057286e-07, "loss": 0.0278, "step": 28235 }, { "epoch": 8.69, "learning_rate": 8.942386297307604e-07, "loss": 0.025, "step": 28236 }, { "epoch": 8.69, "learning_rate": 8.9382676368439e-07, "loss": 0.0324, "step": 28237 }, { "epoch": 8.69, "learning_rate": 8.934149880707154e-07, "loss": 0.0349, "step": 28238 }, { "epoch": 8.69, "learning_rate": 8.930033028938223e-07, "loss": 0.0147, "step": 28239 }, { "epoch": 8.69, "learning_rate": 8.925917081577995e-07, "loss": 0.0622, "step": 28240 }, { "epoch": 8.69, "learning_rate": 8.921802038667349e-07, "loss": 0.0421, "step": 28241 }, { "epoch": 8.69, "learning_rate": 8.91768790024713e-07, "loss": 0.0314, "step": 28242 }, { "epoch": 8.69, "learning_rate": 8.913574666358249e-07, "loss": 0.0471, "step": 28243 }, { "epoch": 8.69, "learning_rate": 8.909462337041508e-07, "loss": 0.0253, "step": 28244 }, { "epoch": 8.69, "learning_rate": 8.90535091233774e-07, "loss": 0.0225, "step": 28245 }, { "epoch": 8.69, "learning_rate": 8.901240392287802e-07, "loss": 0.0289, "step": 28246 }, { "epoch": 8.69, "learning_rate": 8.897130776932506e-07, "loss": 0.017, "step": 28247 }, { "epoch": 8.69, "learning_rate": 8.893022066312674e-07, "loss": 0.0229, "step": 28248 }, { "epoch": 8.69, "learning_rate": 8.888914260469095e-07, "loss": 0.0227, "step": 28249 }, { "epoch": 8.69, "learning_rate": 8.88480735944256e-07, "loss": 0.0386, "step": 28250 }, { "epoch": 8.69, "learning_rate": 8.880701363273869e-07, "loss": 0.045, "step": 28251 }, { "epoch": 8.69, "learning_rate": 8.876596272003768e-07, "loss": 0.0183, "step": 28252 }, { "epoch": 8.69, "learning_rate": 8.87249208567309e-07, "loss": 0.0186, "step": 28253 }, { "epoch": 8.69, "learning_rate": 8.868388804322537e-07, "loss": 0.0378, "step": 28254 }, { "epoch": 8.69, "learning_rate": 8.864286427992852e-07, "loss": 0.0426, "step": 28255 }, { "epoch": 8.69, "learning_rate": 8.860184956724805e-07, "loss": 0.0337, "step": 28256 }, { "epoch": 8.69, "learning_rate": 8.856084390559128e-07, "loss": 0.027, "step": 28257 }, { "epoch": 8.69, "learning_rate": 8.856084390559128e-07, "loss": 0.0316, "step": 28258 }, { "epoch": 8.69, "learning_rate": 8.851984729536534e-07, "loss": 0.0378, "step": 28259 }, { "epoch": 8.7, "learning_rate": 8.847885973697734e-07, "loss": 0.0327, "step": 28260 }, { "epoch": 8.7, "learning_rate": 8.84378812308343e-07, "loss": 0.0195, "step": 28261 }, { "epoch": 8.7, "learning_rate": 8.839691177734322e-07, "loss": 0.0452, "step": 28262 }, { "epoch": 8.7, "learning_rate": 8.835595137691078e-07, "loss": 0.0302, "step": 28263 }, { "epoch": 8.7, "learning_rate": 8.83150000299442e-07, "loss": 0.0338, "step": 28264 }, { "epoch": 8.7, "learning_rate": 8.827405773685005e-07, "loss": 0.014, "step": 28265 }, { "epoch": 8.7, "learning_rate": 8.823312449803434e-07, "loss": 0.0266, "step": 28266 }, { "epoch": 8.7, "learning_rate": 8.81922003139043e-07, "loss": 0.0201, "step": 28267 }, { "epoch": 8.7, "learning_rate": 8.815128518486604e-07, "loss": 0.0439, "step": 28268 }, { "epoch": 8.7, "learning_rate": 8.811037911132581e-07, "loss": 0.0192, "step": 28269 }, { "epoch": 8.7, "learning_rate": 8.806948209369004e-07, "loss": 0.0171, "step": 28270 }, { "epoch": 8.7, "learning_rate": 8.802859413236476e-07, "loss": 0.0214, "step": 28271 }, { "epoch": 8.7, "learning_rate": 8.798771522775607e-07, "loss": 0.0139, "step": 28272 }, { "epoch": 8.7, "learning_rate": 8.794684538026965e-07, "loss": 0.0333, "step": 28273 }, { "epoch": 8.7, "learning_rate": 8.790598459031174e-07, "loss": 0.0298, "step": 28274 }, { "epoch": 8.7, "learning_rate": 8.786513285828835e-07, "loss": 0.0122, "step": 28275 }, { "epoch": 8.7, "learning_rate": 8.782429018460436e-07, "loss": 0.0239, "step": 28276 }, { "epoch": 8.7, "learning_rate": 8.778345656966613e-07, "loss": 0.0437, "step": 28277 }, { "epoch": 8.7, "learning_rate": 8.774263201387889e-07, "loss": 0.0487, "step": 28278 }, { "epoch": 8.7, "learning_rate": 8.770181651764797e-07, "loss": 0.0137, "step": 28279 }, { "epoch": 8.7, "learning_rate": 8.766101008137872e-07, "loss": 0.0525, "step": 28280 }, { "epoch": 8.7, "learning_rate": 8.76202127054766e-07, "loss": 0.0328, "step": 28281 }, { "epoch": 8.7, "learning_rate": 8.757942439034695e-07, "loss": 0.0182, "step": 28282 }, { "epoch": 8.7, "learning_rate": 8.753864513639399e-07, "loss": 0.0371, "step": 28283 }, { "epoch": 8.7, "learning_rate": 8.749787494402351e-07, "loss": 0.0204, "step": 28284 }, { "epoch": 8.7, "learning_rate": 8.745711381364009e-07, "loss": 0.0255, "step": 28285 }, { "epoch": 8.7, "learning_rate": 8.74163617456486e-07, "loss": 0.0334, "step": 28286 }, { "epoch": 8.7, "learning_rate": 8.737561874045364e-07, "loss": 0.0174, "step": 28287 }, { "epoch": 8.7, "learning_rate": 8.733488479845997e-07, "loss": 0.0383, "step": 28288 }, { "epoch": 8.7, "learning_rate": 8.729415992007196e-07, "loss": 0.0516, "step": 28289 }, { "epoch": 8.7, "learning_rate": 8.725344410569392e-07, "loss": 0.0439, "step": 28290 }, { "epoch": 8.7, "learning_rate": 8.721273735573066e-07, "loss": 0.0273, "step": 28291 }, { "epoch": 8.7, "learning_rate": 8.71720396705863e-07, "loss": 0.0479, "step": 28292 }, { "epoch": 8.71, "learning_rate": 8.71313510506645e-07, "loss": 0.0302, "step": 28293 }, { "epoch": 8.71, "learning_rate": 8.709067149636985e-07, "loss": 0.0081, "step": 28294 }, { "epoch": 8.71, "learning_rate": 8.705000100810623e-07, "loss": 0.0403, "step": 28295 }, { "epoch": 8.71, "learning_rate": 8.700933958627755e-07, "loss": 0.0182, "step": 28296 }, { "epoch": 8.71, "learning_rate": 8.696868723128748e-07, "loss": 0.0433, "step": 28297 }, { "epoch": 8.71, "learning_rate": 8.692804394353971e-07, "loss": 0.0326, "step": 28298 }, { "epoch": 8.71, "learning_rate": 8.688740972343834e-07, "loss": 0.0284, "step": 28299 }, { "epoch": 8.71, "learning_rate": 8.684678457138618e-07, "loss": 0.0316, "step": 28300 }, { "epoch": 8.71, "learning_rate": 8.680616848778711e-07, "loss": 0.0221, "step": 28301 }, { "epoch": 8.71, "learning_rate": 8.676556147304449e-07, "loss": 0.0191, "step": 28302 }, { "epoch": 8.71, "learning_rate": 8.672496352756154e-07, "loss": 0.0274, "step": 28303 }, { "epoch": 8.71, "learning_rate": 8.66843746517414e-07, "loss": 0.0478, "step": 28304 }, { "epoch": 8.71, "learning_rate": 8.664379484598684e-07, "loss": 0.0206, "step": 28305 }, { "epoch": 8.71, "learning_rate": 8.660322411070166e-07, "loss": 0.037, "step": 28306 }, { "epoch": 8.71, "learning_rate": 8.656266244628808e-07, "loss": 0.03, "step": 28307 }, { "epoch": 8.71, "learning_rate": 8.65221098531489e-07, "loss": 0.0212, "step": 28308 }, { "epoch": 8.71, "learning_rate": 8.648156633168747e-07, "loss": 0.0286, "step": 28309 }, { "epoch": 8.71, "learning_rate": 8.644103188230557e-07, "loss": 0.0423, "step": 28310 }, { "epoch": 8.71, "learning_rate": 8.640050650540644e-07, "loss": 0.0324, "step": 28311 }, { "epoch": 8.71, "learning_rate": 8.635999020139229e-07, "loss": 0.0278, "step": 28312 }, { "epoch": 8.71, "learning_rate": 8.631948297066539e-07, "loss": 0.0247, "step": 28313 }, { "epoch": 8.71, "learning_rate": 8.627898481362817e-07, "loss": 0.0269, "step": 28314 }, { "epoch": 8.71, "learning_rate": 8.623849573068244e-07, "loss": 0.0348, "step": 28315 }, { "epoch": 8.71, "learning_rate": 8.619801572223108e-07, "loss": 0.0347, "step": 28316 }, { "epoch": 8.71, "learning_rate": 8.615754478867544e-07, "loss": 0.0297, "step": 28317 }, { "epoch": 8.71, "learning_rate": 8.611708293041732e-07, "loss": 0.0288, "step": 28318 }, { "epoch": 8.71, "learning_rate": 8.607663014785905e-07, "loss": 0.0281, "step": 28319 }, { "epoch": 8.71, "learning_rate": 8.603618644140222e-07, "loss": 0.0292, "step": 28320 }, { "epoch": 8.71, "learning_rate": 8.599575181144826e-07, "loss": 0.0237, "step": 28321 }, { "epoch": 8.71, "learning_rate": 8.595532625839898e-07, "loss": 0.0255, "step": 28322 }, { "epoch": 8.71, "learning_rate": 8.59149097826556e-07, "loss": 0.0505, "step": 28323 }, { "epoch": 8.71, "learning_rate": 8.587450238461969e-07, "loss": 0.0525, "step": 28324 }, { "epoch": 8.72, "learning_rate": 8.583410406469217e-07, "loss": 0.0582, "step": 28325 }, { "epoch": 8.72, "learning_rate": 8.579371482327481e-07, "loss": 0.0222, "step": 28326 }, { "epoch": 8.72, "learning_rate": 8.575333466076863e-07, "loss": 0.0276, "step": 28327 }, { "epoch": 8.72, "learning_rate": 8.571296357757397e-07, "loss": 0.0293, "step": 28328 }, { "epoch": 8.72, "learning_rate": 8.567260157409241e-07, "loss": 0.0201, "step": 28329 }, { "epoch": 8.72, "learning_rate": 8.563224865072461e-07, "loss": 0.0214, "step": 28330 }, { "epoch": 8.72, "learning_rate": 8.559190480787128e-07, "loss": 0.0357, "step": 28331 }, { "epoch": 8.72, "learning_rate": 8.555157004593306e-07, "loss": 0.0277, "step": 28332 }, { "epoch": 8.72, "learning_rate": 8.551124436531055e-07, "loss": 0.0296, "step": 28333 }, { "epoch": 8.72, "learning_rate": 8.54709277664042e-07, "loss": 0.0414, "step": 28334 }, { "epoch": 8.72, "learning_rate": 8.543062024961424e-07, "loss": 0.0417, "step": 28335 }, { "epoch": 8.72, "learning_rate": 8.539032181534124e-07, "loss": 0.0178, "step": 28336 }, { "epoch": 8.72, "learning_rate": 8.535003246398554e-07, "loss": 0.0448, "step": 28337 }, { "epoch": 8.72, "learning_rate": 8.530975219594651e-07, "loss": 0.0288, "step": 28338 }, { "epoch": 8.72, "learning_rate": 8.526948101162492e-07, "loss": 0.0144, "step": 28339 }, { "epoch": 8.72, "learning_rate": 8.522921891142034e-07, "loss": 0.0363, "step": 28340 }, { "epoch": 8.72, "learning_rate": 8.518896589573278e-07, "loss": 0.0455, "step": 28341 }, { "epoch": 8.72, "learning_rate": 8.514872196496182e-07, "loss": 0.0268, "step": 28342 }, { "epoch": 8.72, "learning_rate": 8.510848711950704e-07, "loss": 0.0171, "step": 28343 }, { "epoch": 8.72, "learning_rate": 8.506826135976865e-07, "loss": 0.027, "step": 28344 }, { "epoch": 8.72, "learning_rate": 8.502804468614512e-07, "loss": 0.0186, "step": 28345 }, { "epoch": 8.72, "learning_rate": 8.498783709903668e-07, "loss": 0.0259, "step": 28346 }, { "epoch": 8.72, "learning_rate": 8.494763859884225e-07, "loss": 0.0333, "step": 28347 }, { "epoch": 8.72, "learning_rate": 8.490744918596106e-07, "loss": 0.1028, "step": 28348 }, { "epoch": 8.72, "learning_rate": 8.486726886079233e-07, "loss": 0.0337, "step": 28349 }, { "epoch": 8.72, "learning_rate": 8.482709762373509e-07, "loss": 0.026, "step": 28350 }, { "epoch": 8.72, "learning_rate": 8.478693547518813e-07, "loss": 0.0211, "step": 28351 }, { "epoch": 8.72, "learning_rate": 8.474678241555046e-07, "loss": 0.0418, "step": 28352 }, { "epoch": 8.72, "learning_rate": 8.470663844522053e-07, "loss": 0.0319, "step": 28353 }, { "epoch": 8.72, "learning_rate": 8.46665035645976e-07, "loss": 0.026, "step": 28354 }, { "epoch": 8.72, "learning_rate": 8.462637777407956e-07, "loss": 0.0255, "step": 28355 }, { "epoch": 8.72, "learning_rate": 8.458626107406531e-07, "loss": 0.0574, "step": 28356 }, { "epoch": 8.72, "learning_rate": 8.454615346495321e-07, "loss": 0.0189, "step": 28357 }, { "epoch": 8.73, "learning_rate": 8.450605494714159e-07, "loss": 0.0179, "step": 28358 }, { "epoch": 8.73, "learning_rate": 8.446596552102848e-07, "loss": 0.0225, "step": 28359 }, { "epoch": 8.73, "learning_rate": 8.4425885187012e-07, "loss": 0.0296, "step": 28360 }, { "epoch": 8.73, "learning_rate": 8.438581394549061e-07, "loss": 0.0248, "step": 28361 }, { "epoch": 8.73, "learning_rate": 8.434575179686178e-07, "loss": 0.0272, "step": 28362 }, { "epoch": 8.73, "learning_rate": 8.430569874152339e-07, "loss": 0.0404, "step": 28363 }, { "epoch": 8.73, "learning_rate": 8.426565477987347e-07, "loss": 0.034, "step": 28364 }, { "epoch": 8.73, "learning_rate": 8.42256199123096e-07, "loss": 0.0361, "step": 28365 }, { "epoch": 8.73, "learning_rate": 8.418559413922933e-07, "loss": 0.0311, "step": 28366 }, { "epoch": 8.73, "learning_rate": 8.414557746103014e-07, "loss": 0.0232, "step": 28367 }, { "epoch": 8.73, "learning_rate": 8.410556987810937e-07, "loss": 0.0376, "step": 28368 }, { "epoch": 8.73, "learning_rate": 8.406557139086447e-07, "loss": 0.0343, "step": 28369 }, { "epoch": 8.73, "learning_rate": 8.402558199969246e-07, "loss": 0.0172, "step": 28370 }, { "epoch": 8.73, "learning_rate": 8.398560170499093e-07, "loss": 0.0285, "step": 28371 }, { "epoch": 8.73, "learning_rate": 8.394563050715631e-07, "loss": 0.0216, "step": 28372 }, { "epoch": 8.73, "learning_rate": 8.390566840658576e-07, "loss": 0.0239, "step": 28373 }, { "epoch": 8.73, "learning_rate": 8.38657154036765e-07, "loss": 0.0236, "step": 28374 }, { "epoch": 8.73, "learning_rate": 8.382577149882487e-07, "loss": 0.0324, "step": 28375 }, { "epoch": 8.73, "learning_rate": 8.378583669242779e-07, "loss": 0.0248, "step": 28376 }, { "epoch": 8.73, "learning_rate": 8.374591098488171e-07, "loss": 0.0424, "step": 28377 }, { "epoch": 8.73, "learning_rate": 8.370599437658311e-07, "loss": 0.0302, "step": 28378 }, { "epoch": 8.73, "learning_rate": 8.366608686792854e-07, "loss": 0.0305, "step": 28379 }, { "epoch": 8.73, "learning_rate": 8.362618845931414e-07, "loss": 0.0189, "step": 28380 }, { "epoch": 8.73, "learning_rate": 8.358629915113637e-07, "loss": 0.0205, "step": 28381 }, { "epoch": 8.73, "learning_rate": 8.354641894379134e-07, "loss": 0.027, "step": 28382 }, { "epoch": 8.73, "learning_rate": 8.350654783767475e-07, "loss": 0.014, "step": 28383 }, { "epoch": 8.73, "learning_rate": 8.346668583318285e-07, "loss": 0.0294, "step": 28384 }, { "epoch": 8.73, "learning_rate": 8.342683293071164e-07, "loss": 0.0396, "step": 28385 }, { "epoch": 8.73, "learning_rate": 8.338698913065657e-07, "loss": 0.0281, "step": 28386 }, { "epoch": 8.73, "learning_rate": 8.334715443341357e-07, "loss": 0.0362, "step": 28387 }, { "epoch": 8.73, "learning_rate": 8.330732883937808e-07, "loss": 0.0314, "step": 28388 }, { "epoch": 8.73, "learning_rate": 8.32675123489457e-07, "loss": 0.031, "step": 28389 }, { "epoch": 8.74, "learning_rate": 8.322770496251153e-07, "loss": 0.0161, "step": 28390 }, { "epoch": 8.74, "learning_rate": 8.31879066804715e-07, "loss": 0.041, "step": 28391 }, { "epoch": 8.74, "learning_rate": 8.31481175032206e-07, "loss": 0.025, "step": 28392 }, { "epoch": 8.74, "learning_rate": 8.310833743115354e-07, "loss": 0.0439, "step": 28393 }, { "epoch": 8.74, "learning_rate": 8.306856646466588e-07, "loss": 0.0274, "step": 28394 }, { "epoch": 8.74, "learning_rate": 8.302880460415252e-07, "loss": 0.0325, "step": 28395 }, { "epoch": 8.74, "learning_rate": 8.298905185000816e-07, "loss": 0.0213, "step": 28396 }, { "epoch": 8.74, "learning_rate": 8.29493082026277e-07, "loss": 0.0337, "step": 28397 }, { "epoch": 8.74, "learning_rate": 8.29095736624056e-07, "loss": 0.0223, "step": 28398 }, { "epoch": 8.74, "learning_rate": 8.286984822973699e-07, "loss": 0.0389, "step": 28399 }, { "epoch": 8.74, "learning_rate": 8.283013190501565e-07, "loss": 0.0312, "step": 28400 }, { "epoch": 8.74, "learning_rate": 8.279042468863663e-07, "loss": 0.0358, "step": 28401 }, { "epoch": 8.74, "learning_rate": 8.275072658099393e-07, "loss": 0.0291, "step": 28402 }, { "epoch": 8.74, "learning_rate": 8.271103758248189e-07, "loss": 0.0471, "step": 28403 }, { "epoch": 8.74, "learning_rate": 8.267135769349477e-07, "loss": 0.0227, "step": 28404 }, { "epoch": 8.74, "learning_rate": 8.263168691442624e-07, "loss": 0.0478, "step": 28405 }, { "epoch": 8.74, "learning_rate": 8.259202524567089e-07, "loss": 0.0344, "step": 28406 }, { "epoch": 8.74, "learning_rate": 8.255237268762206e-07, "loss": 0.0465, "step": 28407 }, { "epoch": 8.74, "learning_rate": 8.251272924067355e-07, "loss": 0.0167, "step": 28408 }, { "epoch": 8.74, "learning_rate": 8.247309490521949e-07, "loss": 0.0285, "step": 28409 }, { "epoch": 8.74, "learning_rate": 8.2433469681653e-07, "loss": 0.0499, "step": 28410 }, { "epoch": 8.74, "learning_rate": 8.239385357036789e-07, "loss": 0.043, "step": 28411 }, { "epoch": 8.74, "learning_rate": 8.23542465717575e-07, "loss": 0.025, "step": 28412 }, { "epoch": 8.74, "learning_rate": 8.231464868621519e-07, "loss": 0.0378, "step": 28413 }, { "epoch": 8.74, "learning_rate": 8.22750599141342e-07, "loss": 0.0418, "step": 28414 }, { "epoch": 8.74, "learning_rate": 8.223548025590755e-07, "loss": 0.0241, "step": 28415 }, { "epoch": 8.74, "learning_rate": 8.21959097119287e-07, "loss": 0.0357, "step": 28416 }, { "epoch": 8.74, "learning_rate": 8.215634828259012e-07, "loss": 0.0287, "step": 28417 }, { "epoch": 8.74, "learning_rate": 8.211679596828481e-07, "loss": 0.0301, "step": 28418 }, { "epoch": 8.74, "learning_rate": 8.207725276940582e-07, "loss": 0.019, "step": 28419 }, { "epoch": 8.74, "learning_rate": 8.203771868634569e-07, "loss": 0.0216, "step": 28420 }, { "epoch": 8.74, "learning_rate": 8.199819371949703e-07, "loss": 0.0181, "step": 28421 }, { "epoch": 8.74, "learning_rate": 8.195867786925227e-07, "loss": 0.0178, "step": 28422 }, { "epoch": 8.75, "learning_rate": 8.191917113600401e-07, "loss": 0.0305, "step": 28423 }, { "epoch": 8.75, "learning_rate": 8.187967352014458e-07, "loss": 0.0191, "step": 28424 }, { "epoch": 8.75, "learning_rate": 8.184018502206592e-07, "loss": 0.0341, "step": 28425 }, { "epoch": 8.75, "learning_rate": 8.180070564216058e-07, "loss": 0.0544, "step": 28426 }, { "epoch": 8.75, "learning_rate": 8.17612353808207e-07, "loss": 0.0144, "step": 28427 }, { "epoch": 8.75, "learning_rate": 8.172177423843763e-07, "loss": 0.0341, "step": 28428 }, { "epoch": 8.75, "learning_rate": 8.168232221540385e-07, "loss": 0.0232, "step": 28429 }, { "epoch": 8.75, "learning_rate": 8.16428793121109e-07, "loss": 0.0427, "step": 28430 }, { "epoch": 8.75, "learning_rate": 8.160344552895061e-07, "loss": 0.0674, "step": 28431 }, { "epoch": 8.75, "learning_rate": 8.156402086631443e-07, "loss": 0.0277, "step": 28432 }, { "epoch": 8.75, "learning_rate": 8.152460532459405e-07, "loss": 0.0129, "step": 28433 }, { "epoch": 8.75, "learning_rate": 8.148519890418083e-07, "loss": 0.0274, "step": 28434 }, { "epoch": 8.75, "learning_rate": 8.1445801605466e-07, "loss": 0.0173, "step": 28435 }, { "epoch": 8.75, "learning_rate": 8.140641342884104e-07, "loss": 0.0211, "step": 28436 }, { "epoch": 8.75, "learning_rate": 8.136703437469717e-07, "loss": 0.0316, "step": 28437 }, { "epoch": 8.75, "learning_rate": 8.132766444342488e-07, "loss": 0.0335, "step": 28438 }, { "epoch": 8.75, "learning_rate": 8.128830363541574e-07, "loss": 0.0525, "step": 28439 }, { "epoch": 8.75, "learning_rate": 8.124895195106053e-07, "loss": 0.0233, "step": 28440 }, { "epoch": 8.75, "learning_rate": 8.120960939074996e-07, "loss": 0.0628, "step": 28441 }, { "epoch": 8.75, "learning_rate": 8.11702759548747e-07, "loss": 0.0281, "step": 28442 }, { "epoch": 8.75, "learning_rate": 8.113095164382534e-07, "loss": 0.0199, "step": 28443 }, { "epoch": 8.75, "learning_rate": 8.109163645799267e-07, "loss": 0.0507, "step": 28444 }, { "epoch": 8.75, "learning_rate": 8.105233039776672e-07, "loss": 0.0289, "step": 28445 }, { "epoch": 8.75, "learning_rate": 8.101303346353817e-07, "loss": 0.0297, "step": 28446 }, { "epoch": 8.75, "learning_rate": 8.097374565569738e-07, "loss": 0.033, "step": 28447 }, { "epoch": 8.75, "learning_rate": 8.093446697463392e-07, "loss": 0.0335, "step": 28448 }, { "epoch": 8.75, "learning_rate": 8.089519742073848e-07, "loss": 0.0354, "step": 28449 }, { "epoch": 8.75, "learning_rate": 8.085593699440075e-07, "loss": 0.0158, "step": 28450 }, { "epoch": 8.75, "learning_rate": 8.081668569601064e-07, "loss": 0.0191, "step": 28451 }, { "epoch": 8.75, "learning_rate": 8.077744352595807e-07, "loss": 0.023, "step": 28452 }, { "epoch": 8.75, "learning_rate": 8.073821048463237e-07, "loss": 0.0311, "step": 28453 }, { "epoch": 8.75, "learning_rate": 8.069898657242393e-07, "loss": 0.0133, "step": 28454 }, { "epoch": 8.76, "learning_rate": 8.065977178972139e-07, "loss": 0.0538, "step": 28455 }, { "epoch": 8.76, "learning_rate": 8.062056613691483e-07, "loss": 0.0312, "step": 28456 }, { "epoch": 8.76, "learning_rate": 8.058136961439333e-07, "loss": 0.0432, "step": 28457 }, { "epoch": 8.76, "learning_rate": 8.054218222254618e-07, "loss": 0.0306, "step": 28458 }, { "epoch": 8.76, "learning_rate": 8.050300396176259e-07, "loss": 0.0277, "step": 28459 }, { "epoch": 8.76, "learning_rate": 8.046383483243136e-07, "loss": 0.0202, "step": 28460 }, { "epoch": 8.76, "learning_rate": 8.042467483494209e-07, "loss": 0.0257, "step": 28461 }, { "epoch": 8.76, "learning_rate": 8.038552396968324e-07, "loss": 0.0383, "step": 28462 }, { "epoch": 8.76, "learning_rate": 8.034638223704338e-07, "loss": 0.0251, "step": 28463 }, { "epoch": 8.76, "learning_rate": 8.030724963741177e-07, "loss": 0.0229, "step": 28464 }, { "epoch": 8.76, "learning_rate": 8.026812617117675e-07, "loss": 0.0522, "step": 28465 }, { "epoch": 8.76, "learning_rate": 8.022901183872689e-07, "loss": 0.0396, "step": 28466 }, { "epoch": 8.76, "learning_rate": 8.018990664045057e-07, "loss": 0.0238, "step": 28467 }, { "epoch": 8.76, "learning_rate": 8.015081057673623e-07, "loss": 0.0315, "step": 28468 }, { "epoch": 8.76, "learning_rate": 8.011172364797216e-07, "loss": 0.022, "step": 28469 }, { "epoch": 8.76, "learning_rate": 8.007264585454632e-07, "loss": 0.0331, "step": 28470 }, { "epoch": 8.76, "learning_rate": 8.003357719684723e-07, "loss": 0.0286, "step": 28471 }, { "epoch": 8.76, "learning_rate": 7.999451767526234e-07, "loss": 0.0248, "step": 28472 }, { "epoch": 8.76, "learning_rate": 7.995546729017978e-07, "loss": 0.0246, "step": 28473 }, { "epoch": 8.76, "learning_rate": 7.991642604198734e-07, "loss": 0.0344, "step": 28474 }, { "epoch": 8.76, "learning_rate": 7.987739393107297e-07, "loss": 0.0405, "step": 28475 }, { "epoch": 8.76, "learning_rate": 7.983837095782399e-07, "loss": 0.0321, "step": 28476 }, { "epoch": 8.76, "learning_rate": 7.979935712262776e-07, "loss": 0.0307, "step": 28477 }, { "epoch": 8.76, "learning_rate": 7.976035242587254e-07, "loss": 0.06, "step": 28478 }, { "epoch": 8.76, "learning_rate": 7.97213568679448e-07, "loss": 0.0374, "step": 28479 }, { "epoch": 8.76, "learning_rate": 7.9682370449232e-07, "loss": 0.0191, "step": 28480 }, { "epoch": 8.76, "learning_rate": 7.96433931701217e-07, "loss": 0.0215, "step": 28481 }, { "epoch": 8.76, "learning_rate": 7.960442503100075e-07, "loss": 0.0213, "step": 28482 }, { "epoch": 8.76, "learning_rate": 7.956546603225601e-07, "loss": 0.0377, "step": 28483 }, { "epoch": 8.76, "learning_rate": 7.952651617427454e-07, "loss": 0.0248, "step": 28484 }, { "epoch": 8.76, "learning_rate": 7.948757545744312e-07, "loss": 0.0204, "step": 28485 }, { "epoch": 8.76, "learning_rate": 7.944864388214835e-07, "loss": 0.0213, "step": 28486 }, { "epoch": 8.76, "learning_rate": 7.94097214487769e-07, "loss": 0.0428, "step": 28487 }, { "epoch": 8.77, "learning_rate": 7.937080815771569e-07, "loss": 0.0332, "step": 28488 }, { "epoch": 8.77, "learning_rate": 7.933190400935054e-07, "loss": 0.0363, "step": 28489 }, { "epoch": 8.77, "learning_rate": 7.9293009004068e-07, "loss": 0.0322, "step": 28490 }, { "epoch": 8.77, "learning_rate": 7.925412314225467e-07, "loss": 0.0247, "step": 28491 }, { "epoch": 8.77, "learning_rate": 7.921524642429634e-07, "loss": 0.0242, "step": 28492 }, { "epoch": 8.77, "learning_rate": 7.917637885057938e-07, "loss": 0.0334, "step": 28493 }, { "epoch": 8.77, "learning_rate": 7.913752042148959e-07, "loss": 0.0214, "step": 28494 }, { "epoch": 8.77, "learning_rate": 7.909867113741287e-07, "loss": 0.0426, "step": 28495 }, { "epoch": 8.77, "learning_rate": 7.905983099873504e-07, "loss": 0.0157, "step": 28496 }, { "epoch": 8.77, "learning_rate": 7.902100000584179e-07, "loss": 0.0508, "step": 28497 }, { "epoch": 8.77, "learning_rate": 7.89821781591189e-07, "loss": 0.0409, "step": 28498 }, { "epoch": 8.77, "learning_rate": 7.894336545895198e-07, "loss": 0.0181, "step": 28499 }, { "epoch": 8.77, "learning_rate": 7.890456190572604e-07, "loss": 0.0309, "step": 28500 }, { "epoch": 8.77, "learning_rate": 7.886576749982678e-07, "loss": 0.0426, "step": 28501 }, { "epoch": 8.77, "learning_rate": 7.882698224163954e-07, "loss": 0.0089, "step": 28502 }, { "epoch": 8.77, "learning_rate": 7.878820613154925e-07, "loss": 0.0854, "step": 28503 }, { "epoch": 8.77, "learning_rate": 7.874943916994104e-07, "loss": 0.0851, "step": 28504 }, { "epoch": 8.77, "learning_rate": 7.871068135719984e-07, "loss": 0.034, "step": 28505 }, { "epoch": 8.77, "learning_rate": 7.867193269371099e-07, "loss": 0.0286, "step": 28506 }, { "epoch": 8.77, "learning_rate": 7.863319317985863e-07, "loss": 0.0438, "step": 28507 }, { "epoch": 8.77, "learning_rate": 7.859446281602801e-07, "loss": 0.0173, "step": 28508 }, { "epoch": 8.77, "learning_rate": 7.855574160260371e-07, "loss": 0.0259, "step": 28509 }, { "epoch": 8.77, "learning_rate": 7.851702953996975e-07, "loss": 0.0286, "step": 28510 }, { "epoch": 8.77, "learning_rate": 7.847832662851118e-07, "loss": 0.0185, "step": 28511 }, { "epoch": 8.77, "learning_rate": 7.843963286861211e-07, "loss": 0.0232, "step": 28512 }, { "epoch": 8.77, "learning_rate": 7.84009482606568e-07, "loss": 0.0339, "step": 28513 }, { "epoch": 8.77, "learning_rate": 7.836227280502939e-07, "loss": 0.0305, "step": 28514 }, { "epoch": 8.77, "learning_rate": 7.832360650211402e-07, "loss": 0.0179, "step": 28515 }, { "epoch": 8.77, "learning_rate": 7.828494935229492e-07, "loss": 0.0578, "step": 28516 }, { "epoch": 8.77, "learning_rate": 7.824630135595535e-07, "loss": 0.0231, "step": 28517 }, { "epoch": 8.77, "learning_rate": 7.820766251347978e-07, "loss": 0.0278, "step": 28518 }, { "epoch": 8.77, "learning_rate": 7.816903282525157e-07, "loss": 0.0317, "step": 28519 }, { "epoch": 8.78, "learning_rate": 7.813041229165452e-07, "loss": 0.0283, "step": 28520 }, { "epoch": 8.78, "learning_rate": 7.809180091307211e-07, "loss": 0.0281, "step": 28521 }, { "epoch": 8.78, "learning_rate": 7.805319868988759e-07, "loss": 0.0232, "step": 28522 }, { "epoch": 8.78, "learning_rate": 7.801460562248486e-07, "loss": 0.0387, "step": 28523 }, { "epoch": 8.78, "learning_rate": 7.797602171124663e-07, "loss": 0.0245, "step": 28524 }, { "epoch": 8.78, "learning_rate": 7.793744695655603e-07, "loss": 0.0396, "step": 28525 }, { "epoch": 8.78, "learning_rate": 7.789888135879675e-07, "loss": 0.0268, "step": 28526 }, { "epoch": 8.78, "learning_rate": 7.786032491835116e-07, "loss": 0.0246, "step": 28527 }, { "epoch": 8.78, "learning_rate": 7.78217776356025e-07, "loss": 0.0472, "step": 28528 }, { "epoch": 8.78, "learning_rate": 7.778323951093358e-07, "loss": 0.052, "step": 28529 }, { "epoch": 8.78, "learning_rate": 7.774471054472688e-07, "loss": 0.0612, "step": 28530 }, { "epoch": 8.78, "learning_rate": 7.77061907373653e-07, "loss": 0.0174, "step": 28531 }, { "epoch": 8.78, "learning_rate": 7.766768008923109e-07, "loss": 0.0323, "step": 28532 }, { "epoch": 8.78, "learning_rate": 7.762917860070718e-07, "loss": 0.025, "step": 28533 }, { "epoch": 8.78, "learning_rate": 7.759068627217537e-07, "loss": 0.0348, "step": 28534 }, { "epoch": 8.78, "learning_rate": 7.755220310401812e-07, "loss": 0.0303, "step": 28535 }, { "epoch": 8.78, "learning_rate": 7.75137290966177e-07, "loss": 0.0454, "step": 28536 }, { "epoch": 8.78, "learning_rate": 7.747526425035623e-07, "loss": 0.0556, "step": 28537 }, { "epoch": 8.78, "learning_rate": 7.743680856561552e-07, "loss": 0.0381, "step": 28538 }, { "epoch": 8.78, "learning_rate": 7.739836204277762e-07, "loss": 0.0352, "step": 28539 }, { "epoch": 8.78, "learning_rate": 7.73599246822242e-07, "loss": 0.0357, "step": 28540 }, { "epoch": 8.78, "learning_rate": 7.732149648433718e-07, "loss": 0.0261, "step": 28541 }, { "epoch": 8.78, "learning_rate": 7.728307744949781e-07, "loss": 0.014, "step": 28542 }, { "epoch": 8.78, "learning_rate": 7.724466757808813e-07, "loss": 0.0261, "step": 28543 }, { "epoch": 8.78, "learning_rate": 7.720626687048938e-07, "loss": 0.0314, "step": 28544 }, { "epoch": 8.78, "learning_rate": 7.71678753270827e-07, "loss": 0.0446, "step": 28545 }, { "epoch": 8.78, "learning_rate": 7.712949294824956e-07, "loss": 0.0374, "step": 28546 }, { "epoch": 8.78, "learning_rate": 7.709111973437111e-07, "loss": 0.0312, "step": 28547 }, { "epoch": 8.78, "learning_rate": 7.705275568582848e-07, "loss": 0.0303, "step": 28548 }, { "epoch": 8.78, "learning_rate": 7.70144008030026e-07, "loss": 0.0227, "step": 28549 }, { "epoch": 8.78, "learning_rate": 7.697605508627426e-07, "loss": 0.03, "step": 28550 }, { "epoch": 8.78, "learning_rate": 7.693771853602449e-07, "loss": 0.031, "step": 28551 }, { "epoch": 8.78, "learning_rate": 7.689939115263367e-07, "loss": 0.0351, "step": 28552 }, { "epoch": 8.79, "learning_rate": 7.686107293648282e-07, "loss": 0.0145, "step": 28553 }, { "epoch": 8.79, "learning_rate": 7.682276388795251e-07, "loss": 0.0319, "step": 28554 }, { "epoch": 8.79, "learning_rate": 7.678446400742268e-07, "loss": 0.021, "step": 28555 }, { "epoch": 8.79, "learning_rate": 7.674617329527412e-07, "loss": 0.0315, "step": 28556 }, { "epoch": 8.79, "learning_rate": 7.670789175188687e-07, "loss": 0.038, "step": 28557 }, { "epoch": 8.79, "learning_rate": 7.666961937764128e-07, "loss": 0.0255, "step": 28558 }, { "epoch": 8.79, "learning_rate": 7.663135617291728e-07, "loss": 0.0319, "step": 28559 }, { "epoch": 8.79, "learning_rate": 7.659310213809479e-07, "loss": 0.0304, "step": 28560 }, { "epoch": 8.79, "learning_rate": 7.655485727355416e-07, "loss": 0.0193, "step": 28561 }, { "epoch": 8.79, "learning_rate": 7.651662157967454e-07, "loss": 0.028, "step": 28562 }, { "epoch": 8.79, "learning_rate": 7.647839505683608e-07, "loss": 0.0303, "step": 28563 }, { "epoch": 8.79, "learning_rate": 7.644017770541834e-07, "loss": 0.0325, "step": 28564 }, { "epoch": 8.79, "learning_rate": 7.640196952580081e-07, "loss": 0.0097, "step": 28565 }, { "epoch": 8.79, "learning_rate": 7.636377051836297e-07, "loss": 0.0258, "step": 28566 }, { "epoch": 8.79, "learning_rate": 7.632558068348417e-07, "loss": 0.0281, "step": 28567 }, { "epoch": 8.79, "learning_rate": 7.628740002154356e-07, "loss": 0.0351, "step": 28568 }, { "epoch": 8.79, "learning_rate": 7.624922853292039e-07, "loss": 0.0309, "step": 28569 }, { "epoch": 8.79, "learning_rate": 7.621106621799356e-07, "loss": 0.0236, "step": 28570 }, { "epoch": 8.79, "learning_rate": 7.617291307714259e-07, "loss": 0.0238, "step": 28571 }, { "epoch": 8.79, "learning_rate": 7.613476911074568e-07, "loss": 0.0357, "step": 28572 }, { "epoch": 8.79, "learning_rate": 7.609663431918213e-07, "loss": 0.0249, "step": 28573 }, { "epoch": 8.79, "learning_rate": 7.60585087028305e-07, "loss": 0.0389, "step": 28574 }, { "epoch": 8.79, "learning_rate": 7.602039226206947e-07, "loss": 0.0308, "step": 28575 }, { "epoch": 8.79, "learning_rate": 7.598228499727745e-07, "loss": 0.0329, "step": 28576 }, { "epoch": 8.79, "learning_rate": 7.594418690883276e-07, "loss": 0.03, "step": 28577 }, { "epoch": 8.79, "learning_rate": 7.590609799711424e-07, "loss": 0.0341, "step": 28578 }, { "epoch": 8.79, "learning_rate": 7.586801826249979e-07, "loss": 0.0154, "step": 28579 }, { "epoch": 8.79, "learning_rate": 7.582994770536733e-07, "loss": 0.0327, "step": 28580 }, { "epoch": 8.79, "learning_rate": 7.579188632609547e-07, "loss": 0.0258, "step": 28581 }, { "epoch": 8.79, "learning_rate": 7.575383412506187e-07, "loss": 0.0365, "step": 28582 }, { "epoch": 8.79, "learning_rate": 7.57157911026446e-07, "loss": 0.0216, "step": 28583 }, { "epoch": 8.79, "learning_rate": 7.567775725922133e-07, "loss": 0.027, "step": 28584 }, { "epoch": 8.8, "learning_rate": 7.563973259516976e-07, "loss": 0.013, "step": 28585 }, { "epoch": 8.8, "learning_rate": 7.560171711086761e-07, "loss": 0.0189, "step": 28586 }, { "epoch": 8.8, "learning_rate": 7.556371080669222e-07, "loss": 0.0262, "step": 28587 }, { "epoch": 8.8, "learning_rate": 7.552571368302153e-07, "loss": 0.0338, "step": 28588 }, { "epoch": 8.8, "learning_rate": 7.548772574023233e-07, "loss": 0.0356, "step": 28589 }, { "epoch": 8.8, "learning_rate": 7.544974697870188e-07, "loss": 0.0247, "step": 28590 }, { "epoch": 8.8, "learning_rate": 7.541177739880767e-07, "loss": 0.0262, "step": 28591 }, { "epoch": 8.8, "learning_rate": 7.537381700092661e-07, "loss": 0.0345, "step": 28592 }, { "epoch": 8.8, "learning_rate": 7.533586578543583e-07, "loss": 0.0201, "step": 28593 }, { "epoch": 8.8, "learning_rate": 7.529792375271206e-07, "loss": 0.0387, "step": 28594 }, { "epoch": 8.8, "learning_rate": 7.525999090313207e-07, "loss": 0.0351, "step": 28595 }, { "epoch": 8.8, "learning_rate": 7.52220672370727e-07, "loss": 0.0238, "step": 28596 }, { "epoch": 8.8, "learning_rate": 7.518415275491031e-07, "loss": 0.0288, "step": 28597 }, { "epoch": 8.8, "learning_rate": 7.51462474570217e-07, "loss": 0.0451, "step": 28598 }, { "epoch": 8.8, "learning_rate": 7.510835134378347e-07, "loss": 0.029, "step": 28599 }, { "epoch": 8.8, "learning_rate": 7.507046441557142e-07, "loss": 0.0242, "step": 28600 }, { "epoch": 8.8, "learning_rate": 7.503258667276225e-07, "loss": 0.0321, "step": 28601 }, { "epoch": 8.8, "learning_rate": 7.4994718115732e-07, "loss": 0.016, "step": 28602 }, { "epoch": 8.8, "learning_rate": 7.495685874485658e-07, "loss": 0.0239, "step": 28603 }, { "epoch": 8.8, "learning_rate": 7.491900856051227e-07, "loss": 0.0979, "step": 28604 }, { "epoch": 8.8, "learning_rate": 7.488116756307462e-07, "loss": 0.0319, "step": 28605 }, { "epoch": 8.8, "learning_rate": 7.484333575291968e-07, "loss": 0.0484, "step": 28606 }, { "epoch": 8.8, "learning_rate": 7.480551313042284e-07, "loss": 0.0604, "step": 28607 }, { "epoch": 8.8, "learning_rate": 7.476769969596021e-07, "loss": 0.0265, "step": 28608 }, { "epoch": 8.8, "learning_rate": 7.472989544990716e-07, "loss": 0.03, "step": 28609 }, { "epoch": 8.8, "learning_rate": 7.469210039263863e-07, "loss": 0.0428, "step": 28610 }, { "epoch": 8.8, "learning_rate": 7.465431452453054e-07, "loss": 0.0216, "step": 28611 }, { "epoch": 8.8, "learning_rate": 7.46165378459579e-07, "loss": 0.0354, "step": 28612 }, { "epoch": 8.8, "learning_rate": 7.457877035729588e-07, "loss": 0.0362, "step": 28613 }, { "epoch": 8.8, "learning_rate": 7.454101205891962e-07, "loss": 0.0647, "step": 28614 }, { "epoch": 8.8, "learning_rate": 7.45032629512038e-07, "loss": 0.0288, "step": 28615 }, { "epoch": 8.8, "learning_rate": 7.446552303452393e-07, "loss": 0.0284, "step": 28616 }, { "epoch": 8.8, "learning_rate": 7.442779230925412e-07, "loss": 0.0365, "step": 28617 }, { "epoch": 8.81, "learning_rate": 7.439007077576943e-07, "loss": 0.0394, "step": 28618 }, { "epoch": 8.81, "learning_rate": 7.435235843444455e-07, "loss": 0.0237, "step": 28619 }, { "epoch": 8.81, "learning_rate": 7.431465528565373e-07, "loss": 0.0242, "step": 28620 }, { "epoch": 8.81, "learning_rate": 7.427696132977158e-07, "loss": 0.0162, "step": 28621 }, { "epoch": 8.81, "learning_rate": 7.423927656717212e-07, "loss": 0.0409, "step": 28622 }, { "epoch": 8.81, "learning_rate": 7.420160099823026e-07, "loss": 0.0432, "step": 28623 }, { "epoch": 8.81, "learning_rate": 7.416393462331961e-07, "loss": 0.0357, "step": 28624 }, { "epoch": 8.81, "learning_rate": 7.412627744281409e-07, "loss": 0.0277, "step": 28625 }, { "epoch": 8.81, "learning_rate": 7.408862945708839e-07, "loss": 0.0246, "step": 28626 }, { "epoch": 8.81, "learning_rate": 7.405099066651556e-07, "loss": 0.0281, "step": 28627 }, { "epoch": 8.81, "learning_rate": 7.401336107146995e-07, "loss": 0.0373, "step": 28628 }, { "epoch": 8.81, "learning_rate": 7.397574067232505e-07, "loss": 0.042, "step": 28629 }, { "epoch": 8.81, "learning_rate": 7.393812946945444e-07, "loss": 0.0229, "step": 28630 }, { "epoch": 8.81, "learning_rate": 7.390052746323185e-07, "loss": 0.0513, "step": 28631 }, { "epoch": 8.81, "learning_rate": 7.386293465403027e-07, "loss": 0.0332, "step": 28632 }, { "epoch": 8.81, "learning_rate": 7.382535104222366e-07, "loss": 0.0369, "step": 28633 }, { "epoch": 8.81, "learning_rate": 7.37877766281847e-07, "loss": 0.0243, "step": 28634 }, { "epoch": 8.81, "learning_rate": 7.375021141228656e-07, "loss": 0.0289, "step": 28635 }, { "epoch": 8.81, "learning_rate": 7.371265539490269e-07, "loss": 0.0503, "step": 28636 }, { "epoch": 8.81, "learning_rate": 7.367510857640581e-07, "loss": 0.0226, "step": 28637 }, { "epoch": 8.81, "learning_rate": 7.363757095716884e-07, "loss": 0.0381, "step": 28638 }, { "epoch": 8.81, "learning_rate": 7.360004253756459e-07, "loss": 0.037, "step": 28639 }, { "epoch": 8.81, "learning_rate": 7.356252331796576e-07, "loss": 0.0198, "step": 28640 }, { "epoch": 8.81, "learning_rate": 7.352501329874473e-07, "loss": 0.0152, "step": 28641 }, { "epoch": 8.81, "learning_rate": 7.348751248027419e-07, "loss": 0.0519, "step": 28642 }, { "epoch": 8.81, "learning_rate": 7.345002086292663e-07, "loss": 0.035, "step": 28643 }, { "epoch": 8.81, "learning_rate": 7.341253844707452e-07, "loss": 0.0323, "step": 28644 }, { "epoch": 8.81, "learning_rate": 7.337506523308946e-07, "loss": 0.0674, "step": 28645 }, { "epoch": 8.81, "learning_rate": 7.333760122134415e-07, "loss": 0.0456, "step": 28646 }, { "epoch": 8.81, "learning_rate": 7.330014641221062e-07, "loss": 0.0275, "step": 28647 }, { "epoch": 8.81, "learning_rate": 7.326270080606058e-07, "loss": 0.0306, "step": 28648 }, { "epoch": 8.81, "learning_rate": 7.322526440326616e-07, "loss": 0.032, "step": 28649 }, { "epoch": 8.82, "learning_rate": 7.318783720419886e-07, "loss": 0.0385, "step": 28650 }, { "epoch": 8.82, "learning_rate": 7.31504192092306e-07, "loss": 0.0182, "step": 28651 }, { "epoch": 8.82, "learning_rate": 7.311301041873276e-07, "loss": 0.0216, "step": 28652 }, { "epoch": 8.82, "learning_rate": 7.307561083307702e-07, "loss": 0.035, "step": 28653 }, { "epoch": 8.82, "learning_rate": 7.303822045263498e-07, "loss": 0.0341, "step": 28654 }, { "epoch": 8.82, "learning_rate": 7.300083927777746e-07, "loss": 0.02, "step": 28655 }, { "epoch": 8.82, "learning_rate": 7.296346730887594e-07, "loss": 0.0332, "step": 28656 }, { "epoch": 8.82, "learning_rate": 7.292610454630178e-07, "loss": 0.0258, "step": 28657 }, { "epoch": 8.82, "learning_rate": 7.288875099042569e-07, "loss": 0.0359, "step": 28658 }, { "epoch": 8.82, "learning_rate": 7.285140664161872e-07, "loss": 0.0275, "step": 28659 }, { "epoch": 8.82, "learning_rate": 7.281407150025166e-07, "loss": 0.0492, "step": 28660 }, { "epoch": 8.82, "learning_rate": 7.277674556669578e-07, "loss": 0.0257, "step": 28661 }, { "epoch": 8.82, "learning_rate": 7.27394288413209e-07, "loss": 0.0259, "step": 28662 }, { "epoch": 8.82, "learning_rate": 7.270212132449839e-07, "loss": 0.0196, "step": 28663 }, { "epoch": 8.82, "learning_rate": 7.266482301659838e-07, "loss": 0.0172, "step": 28664 }, { "epoch": 8.82, "learning_rate": 7.262753391799127e-07, "loss": 0.0349, "step": 28665 }, { "epoch": 8.82, "learning_rate": 7.259025402904751e-07, "loss": 0.0291, "step": 28666 }, { "epoch": 8.82, "learning_rate": 7.255298335013728e-07, "loss": 0.0504, "step": 28667 }, { "epoch": 8.82, "learning_rate": 7.251572188163059e-07, "loss": 0.0174, "step": 28668 }, { "epoch": 8.82, "learning_rate": 7.247846962389771e-07, "loss": 0.0236, "step": 28669 }, { "epoch": 8.82, "learning_rate": 7.244122657730812e-07, "loss": 0.0318, "step": 28670 }, { "epoch": 8.82, "learning_rate": 7.240399274223253e-07, "loss": 0.0403, "step": 28671 }, { "epoch": 8.82, "learning_rate": 7.236676811903976e-07, "loss": 0.0101, "step": 28672 }, { "epoch": 8.82, "learning_rate": 7.232955270810005e-07, "loss": 0.0196, "step": 28673 }, { "epoch": 8.82, "learning_rate": 7.22923465097829e-07, "loss": 0.026, "step": 28674 }, { "epoch": 8.82, "learning_rate": 7.225514952445767e-07, "loss": 0.0193, "step": 28675 }, { "epoch": 8.82, "learning_rate": 7.221796175249396e-07, "loss": 0.0288, "step": 28676 }, { "epoch": 8.82, "learning_rate": 7.218078319426069e-07, "loss": 0.0259, "step": 28677 }, { "epoch": 8.82, "learning_rate": 7.21436138501278e-07, "loss": 0.0296, "step": 28678 }, { "epoch": 8.82, "learning_rate": 7.210645372046354e-07, "loss": 0.0216, "step": 28679 }, { "epoch": 8.82, "learning_rate": 7.206930280563751e-07, "loss": 0.0187, "step": 28680 }, { "epoch": 8.82, "learning_rate": 7.203216110601851e-07, "loss": 0.0306, "step": 28681 }, { "epoch": 8.82, "learning_rate": 7.199502862197549e-07, "loss": 0.0397, "step": 28682 }, { "epoch": 8.83, "learning_rate": 7.195790535387703e-07, "loss": 0.0327, "step": 28683 }, { "epoch": 8.83, "learning_rate": 7.192079130209195e-07, "loss": 0.0483, "step": 28684 }, { "epoch": 8.83, "learning_rate": 7.188368646698862e-07, "loss": 0.0251, "step": 28685 }, { "epoch": 8.83, "learning_rate": 7.184659084893575e-07, "loss": 0.0326, "step": 28686 }, { "epoch": 8.83, "learning_rate": 7.180950444830148e-07, "loss": 0.0354, "step": 28687 }, { "epoch": 8.83, "learning_rate": 7.177242726545464e-07, "loss": 0.0206, "step": 28688 }, { "epoch": 8.83, "learning_rate": 7.17353593007627e-07, "loss": 0.0236, "step": 28689 }, { "epoch": 8.83, "learning_rate": 7.169830055459437e-07, "loss": 0.0185, "step": 28690 }, { "epoch": 8.83, "learning_rate": 7.166125102731735e-07, "loss": 0.0172, "step": 28691 }, { "epoch": 8.83, "learning_rate": 7.162421071929981e-07, "loss": 0.0483, "step": 28692 }, { "epoch": 8.83, "learning_rate": 7.158717963090945e-07, "loss": 0.0252, "step": 28693 }, { "epoch": 8.83, "learning_rate": 7.155015776251383e-07, "loss": 0.0256, "step": 28694 }, { "epoch": 8.83, "learning_rate": 7.151314511448115e-07, "loss": 0.0407, "step": 28695 }, { "epoch": 8.83, "learning_rate": 7.147614168717854e-07, "loss": 0.0334, "step": 28696 }, { "epoch": 8.83, "learning_rate": 7.143914748097336e-07, "loss": 0.0345, "step": 28697 }, { "epoch": 8.83, "learning_rate": 7.140216249623355e-07, "loss": 0.0366, "step": 28698 }, { "epoch": 8.83, "learning_rate": 7.136518673332593e-07, "loss": 0.033, "step": 28699 }, { "epoch": 8.83, "learning_rate": 7.132822019261798e-07, "loss": 0.0256, "step": 28700 }, { "epoch": 8.83, "learning_rate": 7.129126287447663e-07, "loss": 0.0152, "step": 28701 }, { "epoch": 8.83, "learning_rate": 7.125431477926892e-07, "loss": 0.0342, "step": 28702 }, { "epoch": 8.83, "learning_rate": 7.121737590736188e-07, "loss": 0.0228, "step": 28703 }, { "epoch": 8.83, "learning_rate": 7.118044625912213e-07, "loss": 0.0246, "step": 28704 }, { "epoch": 8.83, "learning_rate": 7.114352583491691e-07, "loss": 0.0571, "step": 28705 }, { "epoch": 8.83, "learning_rate": 7.110661463511225e-07, "loss": 0.0259, "step": 28706 }, { "epoch": 8.83, "learning_rate": 7.1069712660075e-07, "loss": 0.0252, "step": 28707 }, { "epoch": 8.83, "learning_rate": 7.103281991017174e-07, "loss": 0.0344, "step": 28708 }, { "epoch": 8.83, "learning_rate": 7.099593638576874e-07, "loss": 0.0192, "step": 28709 }, { "epoch": 8.83, "learning_rate": 7.095906208723224e-07, "loss": 0.0417, "step": 28710 }, { "epoch": 8.83, "learning_rate": 7.092219701492853e-07, "loss": 0.0292, "step": 28711 }, { "epoch": 8.83, "learning_rate": 7.088534116922363e-07, "loss": 0.0402, "step": 28712 }, { "epoch": 8.83, "learning_rate": 7.08484945504836e-07, "loss": 0.0149, "step": 28713 }, { "epoch": 8.83, "learning_rate": 7.081165715907423e-07, "loss": 0.0689, "step": 28714 }, { "epoch": 8.84, "learning_rate": 7.077482899536159e-07, "loss": 0.0328, "step": 28715 }, { "epoch": 8.84, "learning_rate": 7.073801005971148e-07, "loss": 0.0307, "step": 28716 }, { "epoch": 8.84, "learning_rate": 7.070120035248906e-07, "loss": 0.0424, "step": 28717 }, { "epoch": 8.84, "learning_rate": 7.066439987406026e-07, "loss": 0.031, "step": 28718 }, { "epoch": 8.84, "learning_rate": 7.062760862479057e-07, "loss": 0.0321, "step": 28719 }, { "epoch": 8.84, "learning_rate": 7.059082660504524e-07, "loss": 0.0103, "step": 28720 }, { "epoch": 8.84, "learning_rate": 7.055405381518954e-07, "loss": 0.0288, "step": 28721 }, { "epoch": 8.84, "learning_rate": 7.051729025558851e-07, "loss": 0.0241, "step": 28722 }, { "epoch": 8.84, "learning_rate": 7.048053592660786e-07, "loss": 0.0249, "step": 28723 }, { "epoch": 8.84, "learning_rate": 7.044379082861174e-07, "loss": 0.0138, "step": 28724 }, { "epoch": 8.84, "learning_rate": 7.040705496196564e-07, "loss": 0.0178, "step": 28725 }, { "epoch": 8.84, "learning_rate": 7.037032832703439e-07, "loss": 0.037, "step": 28726 }, { "epoch": 8.84, "learning_rate": 7.033361092418223e-07, "loss": 0.0402, "step": 28727 }, { "epoch": 8.84, "learning_rate": 7.029690275377432e-07, "loss": 0.0301, "step": 28728 }, { "epoch": 8.84, "learning_rate": 7.026020381617493e-07, "loss": 0.0221, "step": 28729 }, { "epoch": 8.84, "learning_rate": 7.022351411174866e-07, "loss": 0.0276, "step": 28730 }, { "epoch": 8.84, "learning_rate": 7.018683364085976e-07, "loss": 0.0234, "step": 28731 }, { "epoch": 8.84, "learning_rate": 7.015016240387241e-07, "loss": 0.0277, "step": 28732 }, { "epoch": 8.84, "learning_rate": 7.011350040115128e-07, "loss": 0.015, "step": 28733 }, { "epoch": 8.84, "learning_rate": 7.007684763305967e-07, "loss": 0.0382, "step": 28734 }, { "epoch": 8.84, "learning_rate": 7.004020409996226e-07, "loss": 0.0307, "step": 28735 }, { "epoch": 8.84, "learning_rate": 7.000356980222268e-07, "loss": 0.0281, "step": 28736 }, { "epoch": 8.84, "learning_rate": 6.996694474020471e-07, "loss": 0.0495, "step": 28737 }, { "epoch": 8.84, "learning_rate": 6.993032891427221e-07, "loss": 0.0339, "step": 28738 }, { "epoch": 8.84, "learning_rate": 6.989372232478853e-07, "loss": 0.0217, "step": 28739 }, { "epoch": 8.84, "learning_rate": 6.985712497211772e-07, "loss": 0.0344, "step": 28740 }, { "epoch": 8.84, "learning_rate": 6.982053685662272e-07, "loss": 0.0629, "step": 28741 }, { "epoch": 8.84, "learning_rate": 6.978395797866688e-07, "loss": 0.0159, "step": 28742 }, { "epoch": 8.84, "learning_rate": 6.974738833861383e-07, "loss": 0.0347, "step": 28743 }, { "epoch": 8.84, "learning_rate": 6.971082793682659e-07, "loss": 0.0382, "step": 28744 }, { "epoch": 8.84, "learning_rate": 6.967427677366823e-07, "loss": 0.0241, "step": 28745 }, { "epoch": 8.84, "learning_rate": 6.963773484950164e-07, "loss": 0.0402, "step": 28746 }, { "epoch": 8.84, "learning_rate": 6.96012021646898e-07, "loss": 0.0371, "step": 28747 }, { "epoch": 8.85, "learning_rate": 6.956467871959549e-07, "loss": 0.0347, "step": 28748 }, { "epoch": 8.85, "learning_rate": 6.952816451458133e-07, "loss": 0.0395, "step": 28749 }, { "epoch": 8.85, "learning_rate": 6.949165955001036e-07, "loss": 0.0288, "step": 28750 }, { "epoch": 8.85, "learning_rate": 6.945516382624462e-07, "loss": 0.0264, "step": 28751 }, { "epoch": 8.85, "learning_rate": 6.94186773436466e-07, "loss": 0.0356, "step": 28752 }, { "epoch": 8.85, "learning_rate": 6.93822001025789e-07, "loss": 0.035, "step": 28753 }, { "epoch": 8.85, "learning_rate": 6.934573210340357e-07, "loss": 0.033, "step": 28754 }, { "epoch": 8.85, "learning_rate": 6.930927334648285e-07, "loss": 0.0174, "step": 28755 }, { "epoch": 8.85, "learning_rate": 6.927282383217893e-07, "loss": 0.0272, "step": 28756 }, { "epoch": 8.85, "learning_rate": 6.92363835608535e-07, "loss": 0.0269, "step": 28757 }, { "epoch": 8.85, "learning_rate": 6.919995253286859e-07, "loss": 0.0178, "step": 28758 }, { "epoch": 8.85, "learning_rate": 6.916353074858595e-07, "loss": 0.0353, "step": 28759 }, { "epoch": 8.85, "learning_rate": 6.912711820836748e-07, "loss": 0.0226, "step": 28760 }, { "epoch": 8.85, "learning_rate": 6.90907149125748e-07, "loss": 0.0339, "step": 28761 }, { "epoch": 8.85, "learning_rate": 6.905432086156883e-07, "loss": 0.0283, "step": 28762 }, { "epoch": 8.85, "learning_rate": 6.901793605571172e-07, "loss": 0.0381, "step": 28763 }, { "epoch": 8.85, "learning_rate": 6.898156049536453e-07, "loss": 0.0215, "step": 28764 }, { "epoch": 8.85, "learning_rate": 6.89451941808884e-07, "loss": 0.0355, "step": 28765 }, { "epoch": 8.85, "learning_rate": 6.890883711264462e-07, "loss": 0.0494, "step": 28766 }, { "epoch": 8.85, "learning_rate": 6.887248929099433e-07, "loss": 0.0335, "step": 28767 }, { "epoch": 8.85, "learning_rate": 6.883615071629823e-07, "loss": 0.0232, "step": 28768 }, { "epoch": 8.85, "learning_rate": 6.879982138891717e-07, "loss": 0.0136, "step": 28769 }, { "epoch": 8.85, "learning_rate": 6.876350130921239e-07, "loss": 0.0354, "step": 28770 }, { "epoch": 8.85, "learning_rate": 6.872719047754428e-07, "loss": 0.0247, "step": 28771 }, { "epoch": 8.85, "learning_rate": 6.869088889427334e-07, "loss": 0.0301, "step": 28772 }, { "epoch": 8.85, "learning_rate": 6.865459655976014e-07, "loss": 0.0407, "step": 28773 }, { "epoch": 8.85, "learning_rate": 6.861831347436532e-07, "loss": 0.0218, "step": 28774 }, { "epoch": 8.85, "learning_rate": 6.858203963844889e-07, "loss": 0.0196, "step": 28775 }, { "epoch": 8.85, "learning_rate": 6.854577505237137e-07, "loss": 0.0319, "step": 28776 }, { "epoch": 8.85, "learning_rate": 6.850951971649255e-07, "loss": 0.0259, "step": 28777 }, { "epoch": 8.85, "learning_rate": 6.847327363117296e-07, "loss": 0.0178, "step": 28778 }, { "epoch": 8.85, "learning_rate": 6.843703679677194e-07, "loss": 0.0428, "step": 28779 }, { "epoch": 8.86, "learning_rate": 6.840080921365e-07, "loss": 0.0101, "step": 28780 }, { "epoch": 8.86, "learning_rate": 6.836459088216651e-07, "loss": 0.0183, "step": 28781 }, { "epoch": 8.86, "learning_rate": 6.83283818026812e-07, "loss": 0.0379, "step": 28782 }, { "epoch": 8.86, "learning_rate": 6.829218197555376e-07, "loss": 0.0248, "step": 28783 }, { "epoch": 8.86, "learning_rate": 6.82559914011437e-07, "loss": 0.0598, "step": 28784 }, { "epoch": 8.86, "learning_rate": 6.821981007981027e-07, "loss": 0.0513, "step": 28785 }, { "epoch": 8.86, "learning_rate": 6.818363801191297e-07, "loss": 0.025, "step": 28786 }, { "epoch": 8.86, "learning_rate": 6.814747519781073e-07, "loss": 0.0335, "step": 28787 }, { "epoch": 8.86, "learning_rate": 6.811132163786316e-07, "loss": 0.0378, "step": 28788 }, { "epoch": 8.86, "learning_rate": 6.807517733242874e-07, "loss": 0.0258, "step": 28789 }, { "epoch": 8.86, "learning_rate": 6.803904228186697e-07, "loss": 0.025, "step": 28790 }, { "epoch": 8.86, "learning_rate": 6.800291648653623e-07, "loss": 0.0336, "step": 28791 }, { "epoch": 8.86, "learning_rate": 6.796679994679567e-07, "loss": 0.0232, "step": 28792 }, { "epoch": 8.86, "learning_rate": 6.793069266300367e-07, "loss": 0.039, "step": 28793 }, { "epoch": 8.86, "learning_rate": 6.789459463551873e-07, "loss": 0.0273, "step": 28794 }, { "epoch": 8.86, "learning_rate": 6.785850586469989e-07, "loss": 0.0201, "step": 28795 }, { "epoch": 8.86, "learning_rate": 6.782242635090497e-07, "loss": 0.0092, "step": 28796 }, { "epoch": 8.86, "learning_rate": 6.778635609449235e-07, "loss": 0.0254, "step": 28797 }, { "epoch": 8.86, "learning_rate": 6.775029509582054e-07, "loss": 0.033, "step": 28798 }, { "epoch": 8.86, "learning_rate": 6.771424335524745e-07, "loss": 0.0561, "step": 28799 }, { "epoch": 8.86, "learning_rate": 6.767820087313115e-07, "loss": 0.019, "step": 28800 }, { "epoch": 8.86, "learning_rate": 6.764216764982956e-07, "loss": 0.0219, "step": 28801 }, { "epoch": 8.86, "learning_rate": 6.760614368570051e-07, "loss": 0.0296, "step": 28802 }, { "epoch": 8.86, "learning_rate": 6.757012898110182e-07, "loss": 0.0139, "step": 28803 }, { "epoch": 8.86, "learning_rate": 6.753412353639099e-07, "loss": 0.0418, "step": 28804 }, { "epoch": 8.86, "learning_rate": 6.749812735192596e-07, "loss": 0.0361, "step": 28805 }, { "epoch": 8.86, "learning_rate": 6.746214042806376e-07, "loss": 0.0185, "step": 28806 }, { "epoch": 8.86, "learning_rate": 6.742616276516179e-07, "loss": 0.0218, "step": 28807 }, { "epoch": 8.86, "learning_rate": 6.739019436357774e-07, "loss": 0.0237, "step": 28808 }, { "epoch": 8.86, "learning_rate": 6.735423522366846e-07, "loss": 0.0267, "step": 28809 }, { "epoch": 8.86, "learning_rate": 6.731828534579132e-07, "loss": 0.0451, "step": 28810 }, { "epoch": 8.86, "learning_rate": 6.728234473030315e-07, "loss": 0.0271, "step": 28811 }, { "epoch": 8.86, "learning_rate": 6.724641337756088e-07, "loss": 0.0531, "step": 28812 }, { "epoch": 8.87, "learning_rate": 6.721049128792134e-07, "loss": 0.0436, "step": 28813 }, { "epoch": 8.87, "learning_rate": 6.717457846174125e-07, "loss": 0.0261, "step": 28814 }, { "epoch": 8.87, "learning_rate": 6.713867489937742e-07, "loss": 0.0341, "step": 28815 }, { "epoch": 8.87, "learning_rate": 6.710278060118647e-07, "loss": 0.013, "step": 28816 }, { "epoch": 8.87, "learning_rate": 6.706689556752433e-07, "loss": 0.0252, "step": 28817 }, { "epoch": 8.87, "learning_rate": 6.703101979874793e-07, "loss": 0.0505, "step": 28818 }, { "epoch": 8.87, "learning_rate": 6.699515329521333e-07, "loss": 0.0329, "step": 28819 }, { "epoch": 8.87, "learning_rate": 6.69592960572767e-07, "loss": 0.018, "step": 28820 }, { "epoch": 8.87, "learning_rate": 6.692344808529427e-07, "loss": 0.0283, "step": 28821 }, { "epoch": 8.87, "learning_rate": 6.68876093796218e-07, "loss": 0.0232, "step": 28822 }, { "epoch": 8.87, "learning_rate": 6.685177994061554e-07, "loss": 0.0381, "step": 28823 }, { "epoch": 8.87, "learning_rate": 6.681595976863086e-07, "loss": 0.0303, "step": 28824 }, { "epoch": 8.87, "learning_rate": 6.678014886402396e-07, "loss": 0.0429, "step": 28825 }, { "epoch": 8.87, "learning_rate": 6.67443472271504e-07, "loss": 0.0317, "step": 28826 }, { "epoch": 8.87, "learning_rate": 6.670855485836525e-07, "loss": 0.03, "step": 28827 }, { "epoch": 8.87, "learning_rate": 6.667277175802444e-07, "loss": 0.027, "step": 28828 }, { "epoch": 8.87, "learning_rate": 6.663699792648326e-07, "loss": 0.0291, "step": 28829 }, { "epoch": 8.87, "learning_rate": 6.660123336409685e-07, "loss": 0.0688, "step": 28830 }, { "epoch": 8.87, "learning_rate": 6.656547807122038e-07, "loss": 0.0197, "step": 28831 }, { "epoch": 8.87, "learning_rate": 6.652973204820889e-07, "loss": 0.0732, "step": 28832 }, { "epoch": 8.87, "learning_rate": 6.649399529541789e-07, "loss": 0.0136, "step": 28833 }, { "epoch": 8.87, "learning_rate": 6.645826781320141e-07, "loss": 0.0441, "step": 28834 }, { "epoch": 8.87, "learning_rate": 6.642254960191485e-07, "loss": 0.0392, "step": 28835 }, { "epoch": 8.87, "learning_rate": 6.638684066191292e-07, "loss": 0.0183, "step": 28836 }, { "epoch": 8.87, "learning_rate": 6.635114099355e-07, "loss": 0.0453, "step": 28837 }, { "epoch": 8.87, "learning_rate": 6.631545059718069e-07, "loss": 0.0479, "step": 28838 }, { "epoch": 8.87, "learning_rate": 6.627976947315939e-07, "loss": 0.0343, "step": 28839 }, { "epoch": 8.87, "learning_rate": 6.624409762184081e-07, "loss": 0.0336, "step": 28840 }, { "epoch": 8.87, "learning_rate": 6.620843504357877e-07, "loss": 0.0326, "step": 28841 }, { "epoch": 8.87, "learning_rate": 6.617278173872743e-07, "loss": 0.0353, "step": 28842 }, { "epoch": 8.87, "learning_rate": 6.613713770764129e-07, "loss": 0.0361, "step": 28843 }, { "epoch": 8.87, "learning_rate": 6.610150295067375e-07, "loss": 0.0336, "step": 28844 }, { "epoch": 8.88, "learning_rate": 6.606587746817916e-07, "loss": 0.0393, "step": 28845 }, { "epoch": 8.88, "learning_rate": 6.603026126051104e-07, "loss": 0.0226, "step": 28846 }, { "epoch": 8.88, "learning_rate": 6.599465432802332e-07, "loss": 0.0224, "step": 28847 }, { "epoch": 8.88, "learning_rate": 6.59590566710695e-07, "loss": 0.031, "step": 28848 }, { "epoch": 8.88, "learning_rate": 6.592346829000284e-07, "loss": 0.0215, "step": 28849 }, { "epoch": 8.88, "learning_rate": 6.588788918517741e-07, "loss": 0.0267, "step": 28850 }, { "epoch": 8.88, "learning_rate": 6.585231935694592e-07, "loss": 0.0212, "step": 28851 }, { "epoch": 8.88, "learning_rate": 6.581675880566163e-07, "loss": 0.0158, "step": 28852 }, { "epoch": 8.88, "learning_rate": 6.578120753167816e-07, "loss": 0.0377, "step": 28853 }, { "epoch": 8.88, "learning_rate": 6.574566553534823e-07, "loss": 0.0488, "step": 28854 }, { "epoch": 8.88, "learning_rate": 6.571013281702499e-07, "loss": 0.0245, "step": 28855 }, { "epoch": 8.88, "learning_rate": 6.567460937706105e-07, "loss": 0.0291, "step": 28856 }, { "epoch": 8.88, "learning_rate": 6.563909521580936e-07, "loss": 0.0291, "step": 28857 }, { "epoch": 8.88, "learning_rate": 6.560359033362263e-07, "loss": 0.0217, "step": 28858 }, { "epoch": 8.88, "learning_rate": 6.556809473085335e-07, "loss": 0.037, "step": 28859 }, { "epoch": 8.88, "learning_rate": 6.553260840785414e-07, "loss": 0.0265, "step": 28860 }, { "epoch": 8.88, "learning_rate": 6.549713136497748e-07, "loss": 0.0281, "step": 28861 }, { "epoch": 8.88, "learning_rate": 6.546166360257533e-07, "loss": 0.0337, "step": 28862 }, { "epoch": 8.88, "learning_rate": 6.542620512100029e-07, "loss": 0.0219, "step": 28863 }, { "epoch": 8.88, "learning_rate": 6.539075592060429e-07, "loss": 0.0384, "step": 28864 }, { "epoch": 8.88, "learning_rate": 6.53553160017395e-07, "loss": 0.0264, "step": 28865 }, { "epoch": 8.88, "learning_rate": 6.531988536475786e-07, "loss": 0.0271, "step": 28866 }, { "epoch": 8.88, "learning_rate": 6.52844640100111e-07, "loss": 0.0243, "step": 28867 }, { "epoch": 8.88, "learning_rate": 6.524905193785114e-07, "loss": 0.026, "step": 28868 }, { "epoch": 8.88, "learning_rate": 6.521364914862948e-07, "loss": 0.0248, "step": 28869 }, { "epoch": 8.88, "learning_rate": 6.517825564269786e-07, "loss": 0.0321, "step": 28870 }, { "epoch": 8.88, "learning_rate": 6.514287142040777e-07, "loss": 0.0195, "step": 28871 }, { "epoch": 8.88, "learning_rate": 6.510749648211056e-07, "loss": 0.0293, "step": 28872 }, { "epoch": 8.88, "learning_rate": 6.507213082815745e-07, "loss": 0.0286, "step": 28873 }, { "epoch": 8.88, "learning_rate": 6.503677445889977e-07, "loss": 0.0429, "step": 28874 }, { "epoch": 8.88, "learning_rate": 6.500142737468873e-07, "loss": 0.0264, "step": 28875 }, { "epoch": 8.88, "learning_rate": 6.496608957587492e-07, "loss": 0.0246, "step": 28876 }, { "epoch": 8.88, "learning_rate": 6.493076106280982e-07, "loss": 0.0273, "step": 28877 }, { "epoch": 8.89, "learning_rate": 6.489544183584418e-07, "loss": 0.0362, "step": 28878 }, { "epoch": 8.89, "learning_rate": 6.486013189532836e-07, "loss": 0.0516, "step": 28879 }, { "epoch": 8.89, "learning_rate": 6.482483124161343e-07, "loss": 0.0412, "step": 28880 }, { "epoch": 8.89, "learning_rate": 6.478953987504977e-07, "loss": 0.0392, "step": 28881 }, { "epoch": 8.89, "learning_rate": 6.475425779598787e-07, "loss": 0.0218, "step": 28882 }, { "epoch": 8.89, "learning_rate": 6.471898500477825e-07, "loss": 0.0277, "step": 28883 }, { "epoch": 8.89, "learning_rate": 6.468372150177093e-07, "loss": 0.0112, "step": 28884 }, { "epoch": 8.89, "learning_rate": 6.464846728731644e-07, "loss": 0.0131, "step": 28885 }, { "epoch": 8.89, "learning_rate": 6.461322236176438e-07, "loss": 0.0271, "step": 28886 }, { "epoch": 8.89, "learning_rate": 6.457798672546545e-07, "loss": 0.0067, "step": 28887 }, { "epoch": 8.89, "learning_rate": 6.454276037876927e-07, "loss": 0.0303, "step": 28888 }, { "epoch": 8.89, "learning_rate": 6.450754332202525e-07, "loss": 0.044, "step": 28889 }, { "epoch": 8.89, "learning_rate": 6.447233555558374e-07, "loss": 0.0358, "step": 28890 }, { "epoch": 8.89, "learning_rate": 6.443713707979415e-07, "loss": 0.0353, "step": 28891 }, { "epoch": 8.89, "learning_rate": 6.440194789500598e-07, "loss": 0.0248, "step": 28892 }, { "epoch": 8.89, "learning_rate": 6.436676800156883e-07, "loss": 0.0186, "step": 28893 }, { "epoch": 8.89, "learning_rate": 6.433159739983174e-07, "loss": 0.0328, "step": 28894 }, { "epoch": 8.89, "learning_rate": 6.429643609014447e-07, "loss": 0.0403, "step": 28895 }, { "epoch": 8.89, "learning_rate": 6.426128407285581e-07, "loss": 0.0095, "step": 28896 }, { "epoch": 8.89, "learning_rate": 6.422614134831506e-07, "loss": 0.0201, "step": 28897 }, { "epoch": 8.89, "learning_rate": 6.419100791687116e-07, "loss": 0.0327, "step": 28898 }, { "epoch": 8.89, "learning_rate": 6.415588377887305e-07, "loss": 0.0192, "step": 28899 }, { "epoch": 8.89, "learning_rate": 6.412076893466945e-07, "loss": 0.0289, "step": 28900 }, { "epoch": 8.89, "learning_rate": 6.408566338460931e-07, "loss": 0.022, "step": 28901 }, { "epoch": 8.89, "learning_rate": 6.405056712904089e-07, "loss": 0.0216, "step": 28902 }, { "epoch": 8.89, "learning_rate": 6.401548016831305e-07, "loss": 0.0483, "step": 28903 }, { "epoch": 8.89, "learning_rate": 6.398040250277393e-07, "loss": 0.0479, "step": 28904 }, { "epoch": 8.89, "learning_rate": 6.394533413277249e-07, "loss": 0.0501, "step": 28905 }, { "epoch": 8.89, "learning_rate": 6.39102750586561e-07, "loss": 0.0374, "step": 28906 }, { "epoch": 8.89, "learning_rate": 6.387522528077372e-07, "loss": 0.0205, "step": 28907 }, { "epoch": 8.89, "learning_rate": 6.384018479947296e-07, "loss": 0.0198, "step": 28908 }, { "epoch": 8.89, "learning_rate": 6.380515361510209e-07, "loss": 0.0089, "step": 28909 }, { "epoch": 8.9, "learning_rate": 6.377013172800872e-07, "loss": 0.0315, "step": 28910 }, { "epoch": 8.9, "learning_rate": 6.37351191385408e-07, "loss": 0.0409, "step": 28911 }, { "epoch": 8.9, "learning_rate": 6.370011584704617e-07, "loss": 0.0342, "step": 28912 }, { "epoch": 8.9, "learning_rate": 6.366512185387231e-07, "loss": 0.034, "step": 28913 }, { "epoch": 8.9, "learning_rate": 6.36301371593665e-07, "loss": 0.0229, "step": 28914 }, { "epoch": 8.9, "learning_rate": 6.359516176387659e-07, "loss": 0.0358, "step": 28915 }, { "epoch": 8.9, "learning_rate": 6.356019566774974e-07, "loss": 0.071, "step": 28916 }, { "epoch": 8.9, "learning_rate": 6.352523887133321e-07, "loss": 0.0368, "step": 28917 }, { "epoch": 8.9, "learning_rate": 6.349029137497409e-07, "loss": 0.0286, "step": 28918 }, { "epoch": 8.9, "learning_rate": 6.345535317901952e-07, "loss": 0.0423, "step": 28919 }, { "epoch": 8.9, "learning_rate": 6.342042428381645e-07, "loss": 0.0378, "step": 28920 }, { "epoch": 8.9, "learning_rate": 6.338550468971161e-07, "loss": 0.0511, "step": 28921 }, { "epoch": 8.9, "learning_rate": 6.335059439705226e-07, "loss": 0.0398, "step": 28922 }, { "epoch": 8.9, "learning_rate": 6.331569340618448e-07, "loss": 0.0231, "step": 28923 }, { "epoch": 8.9, "learning_rate": 6.32808017174551e-07, "loss": 0.0309, "step": 28924 }, { "epoch": 8.9, "learning_rate": 6.324591933121072e-07, "loss": 0.034, "step": 28925 }, { "epoch": 8.9, "learning_rate": 6.321104624779773e-07, "loss": 0.0293, "step": 28926 }, { "epoch": 8.9, "learning_rate": 6.317618246756241e-07, "loss": 0.0204, "step": 28927 }, { "epoch": 8.9, "learning_rate": 6.314132799085104e-07, "loss": 0.0414, "step": 28928 }, { "epoch": 8.9, "learning_rate": 6.310648281800969e-07, "loss": 0.0161, "step": 28929 }, { "epoch": 8.9, "learning_rate": 6.307164694938439e-07, "loss": 0.0386, "step": 28930 }, { "epoch": 8.9, "learning_rate": 6.303682038532111e-07, "loss": 0.0112, "step": 28931 }, { "epoch": 8.9, "learning_rate": 6.300200312616578e-07, "loss": 0.0258, "step": 28932 }, { "epoch": 8.9, "learning_rate": 6.296719517226424e-07, "loss": 0.0069, "step": 28933 }, { "epoch": 8.9, "learning_rate": 6.293239652396188e-07, "loss": 0.0275, "step": 28934 }, { "epoch": 8.9, "learning_rate": 6.289760718160454e-07, "loss": 0.0222, "step": 28935 }, { "epoch": 8.9, "learning_rate": 6.28628271455376e-07, "loss": 0.0563, "step": 28936 }, { "epoch": 8.9, "learning_rate": 6.282805641610646e-07, "loss": 0.0284, "step": 28937 }, { "epoch": 8.9, "learning_rate": 6.279329499365649e-07, "loss": 0.0271, "step": 28938 }, { "epoch": 8.9, "learning_rate": 6.275854287853267e-07, "loss": 0.0273, "step": 28939 }, { "epoch": 8.9, "learning_rate": 6.272380007108059e-07, "loss": 0.0132, "step": 28940 }, { "epoch": 8.9, "learning_rate": 6.268906657164476e-07, "loss": 0.0218, "step": 28941 }, { "epoch": 8.9, "learning_rate": 6.265434238057044e-07, "loss": 0.0348, "step": 28942 }, { "epoch": 8.91, "learning_rate": 6.26196274982025e-07, "loss": 0.0321, "step": 28943 }, { "epoch": 8.91, "learning_rate": 6.25849219248853e-07, "loss": 0.0298, "step": 28944 }, { "epoch": 8.91, "learning_rate": 6.255022566096402e-07, "loss": 0.0283, "step": 28945 }, { "epoch": 8.91, "learning_rate": 6.251553870678284e-07, "loss": 0.0391, "step": 28946 }, { "epoch": 8.91, "learning_rate": 6.248086106268647e-07, "loss": 0.0336, "step": 28947 }, { "epoch": 8.91, "learning_rate": 6.24461927290192e-07, "loss": 0.0266, "step": 28948 }, { "epoch": 8.91, "learning_rate": 6.241153370612507e-07, "loss": 0.0446, "step": 28949 }, { "epoch": 8.91, "learning_rate": 6.237688399434894e-07, "loss": 0.0565, "step": 28950 }, { "epoch": 8.91, "learning_rate": 6.234224359403407e-07, "loss": 0.0277, "step": 28951 }, { "epoch": 8.91, "learning_rate": 6.230761250552508e-07, "loss": 0.013, "step": 28952 }, { "epoch": 8.91, "learning_rate": 6.227299072916571e-07, "loss": 0.0381, "step": 28953 }, { "epoch": 8.91, "learning_rate": 6.223837826529988e-07, "loss": 0.0335, "step": 28954 }, { "epoch": 8.91, "learning_rate": 6.220377511427112e-07, "loss": 0.0222, "step": 28955 }, { "epoch": 8.91, "learning_rate": 6.216918127642301e-07, "loss": 0.0269, "step": 28956 }, { "epoch": 8.91, "learning_rate": 6.213459675209954e-07, "loss": 0.0351, "step": 28957 }, { "epoch": 8.91, "learning_rate": 6.210002154164374e-07, "loss": 0.0262, "step": 28958 }, { "epoch": 8.91, "learning_rate": 6.206545564539901e-07, "loss": 0.0289, "step": 28959 }, { "epoch": 8.91, "learning_rate": 6.203089906370885e-07, "loss": 0.0286, "step": 28960 }, { "epoch": 8.91, "learning_rate": 6.199635179691643e-07, "loss": 0.044, "step": 28961 }, { "epoch": 8.91, "learning_rate": 6.19618138453646e-07, "loss": 0.0355, "step": 28962 }, { "epoch": 8.91, "learning_rate": 6.19272852093965e-07, "loss": 0.0387, "step": 28963 }, { "epoch": 8.91, "learning_rate": 6.1892765889355e-07, "loss": 0.029, "step": 28964 }, { "epoch": 8.91, "learning_rate": 6.185825588558303e-07, "loss": 0.0295, "step": 28965 }, { "epoch": 8.91, "learning_rate": 6.182375519842299e-07, "loss": 0.0233, "step": 28966 }, { "epoch": 8.91, "learning_rate": 6.178926382821793e-07, "loss": 0.0464, "step": 28967 }, { "epoch": 8.91, "learning_rate": 6.175478177531002e-07, "loss": 0.0171, "step": 28968 }, { "epoch": 8.91, "learning_rate": 6.172030904004179e-07, "loss": 0.0385, "step": 28969 }, { "epoch": 8.91, "learning_rate": 6.16858456227557e-07, "loss": 0.0263, "step": 28970 }, { "epoch": 8.91, "learning_rate": 6.165139152379396e-07, "loss": 0.0274, "step": 28971 }, { "epoch": 8.91, "learning_rate": 6.161694674349872e-07, "loss": 0.0392, "step": 28972 }, { "epoch": 8.91, "learning_rate": 6.158251128221205e-07, "loss": 0.0347, "step": 28973 }, { "epoch": 8.91, "learning_rate": 6.154808514027588e-07, "loss": 0.0332, "step": 28974 }, { "epoch": 8.92, "learning_rate": 6.151366831803218e-07, "loss": 0.0352, "step": 28975 }, { "epoch": 8.92, "learning_rate": 6.147926081582256e-07, "loss": 0.0187, "step": 28976 }, { "epoch": 8.92, "learning_rate": 6.144486263398886e-07, "loss": 0.0506, "step": 28977 }, { "epoch": 8.92, "learning_rate": 6.14104737728729e-07, "loss": 0.0246, "step": 28978 }, { "epoch": 8.92, "learning_rate": 6.137609423281577e-07, "loss": 0.0352, "step": 28979 }, { "epoch": 8.92, "learning_rate": 6.134172401415905e-07, "loss": 0.0269, "step": 28980 }, { "epoch": 8.92, "learning_rate": 6.130736311724417e-07, "loss": 0.0283, "step": 28981 }, { "epoch": 8.92, "learning_rate": 6.127301154241239e-07, "loss": 0.0261, "step": 28982 }, { "epoch": 8.92, "learning_rate": 6.123866929000466e-07, "loss": 0.0285, "step": 28983 }, { "epoch": 8.92, "learning_rate": 6.120433636036216e-07, "loss": 0.0295, "step": 28984 }, { "epoch": 8.92, "learning_rate": 6.117001275382584e-07, "loss": 0.0466, "step": 28985 }, { "epoch": 8.92, "learning_rate": 6.11356984707363e-07, "loss": 0.0237, "step": 28986 }, { "epoch": 8.92, "learning_rate": 6.110139351143485e-07, "loss": 0.0416, "step": 28987 }, { "epoch": 8.92, "learning_rate": 6.106709787626198e-07, "loss": 0.0278, "step": 28988 }, { "epoch": 8.92, "learning_rate": 6.103281156555784e-07, "loss": 0.0248, "step": 28989 }, { "epoch": 8.92, "learning_rate": 6.099853457966342e-07, "loss": 0.0245, "step": 28990 }, { "epoch": 8.92, "learning_rate": 6.096426691891899e-07, "loss": 0.0236, "step": 28991 }, { "epoch": 8.92, "learning_rate": 6.093000858366471e-07, "loss": 0.0217, "step": 28992 }, { "epoch": 8.92, "learning_rate": 6.08957595742411e-07, "loss": 0.0302, "step": 28993 }, { "epoch": 8.92, "learning_rate": 6.086151989098787e-07, "loss": 0.0212, "step": 28994 }, { "epoch": 8.92, "learning_rate": 6.082728953424555e-07, "loss": 0.0238, "step": 28995 }, { "epoch": 8.92, "learning_rate": 6.079306850435351e-07, "loss": 0.0381, "step": 28996 }, { "epoch": 8.92, "learning_rate": 6.075885680165206e-07, "loss": 0.0338, "step": 28997 }, { "epoch": 8.92, "learning_rate": 6.072465442648079e-07, "loss": 0.0249, "step": 28998 }, { "epoch": 8.92, "learning_rate": 6.069046137917944e-07, "loss": 0.023, "step": 28999 }, { "epoch": 8.92, "learning_rate": 6.065627766008742e-07, "loss": 0.0267, "step": 29000 }, { "epoch": 8.92, "learning_rate": 6.062210326954432e-07, "loss": 0.021, "step": 29001 }, { "epoch": 8.92, "learning_rate": 6.058793820788944e-07, "loss": 0.0296, "step": 29002 }, { "epoch": 8.92, "learning_rate": 6.055378247546217e-07, "loss": 0.0425, "step": 29003 }, { "epoch": 8.92, "learning_rate": 6.051963607260148e-07, "loss": 0.0362, "step": 29004 }, { "epoch": 8.92, "learning_rate": 6.048549899964706e-07, "loss": 0.0298, "step": 29005 }, { "epoch": 8.92, "learning_rate": 6.045137125693712e-07, "loss": 0.0513, "step": 29006 }, { "epoch": 8.92, "learning_rate": 6.041725284481115e-07, "loss": 0.023, "step": 29007 }, { "epoch": 8.93, "learning_rate": 6.038314376360776e-07, "loss": 0.0438, "step": 29008 }, { "epoch": 8.93, "learning_rate": 6.03490440136657e-07, "loss": 0.0247, "step": 29009 }, { "epoch": 8.93, "learning_rate": 6.03149535953238e-07, "loss": 0.0405, "step": 29010 }, { "epoch": 8.93, "learning_rate": 6.028087250892012e-07, "loss": 0.0267, "step": 29011 }, { "epoch": 8.93, "learning_rate": 6.024680075479384e-07, "loss": 0.0172, "step": 29012 }, { "epoch": 8.93, "learning_rate": 6.021273833328279e-07, "loss": 0.0434, "step": 29013 }, { "epoch": 8.93, "learning_rate": 6.017868524472525e-07, "loss": 0.0296, "step": 29014 }, { "epoch": 8.93, "learning_rate": 6.014464148945964e-07, "loss": 0.0339, "step": 29015 }, { "epoch": 8.93, "learning_rate": 6.01106070678239e-07, "loss": 0.0265, "step": 29016 }, { "epoch": 8.93, "learning_rate": 6.007658198015609e-07, "loss": 0.0202, "step": 29017 }, { "epoch": 8.93, "learning_rate": 6.004256622679405e-07, "loss": 0.0173, "step": 29018 }, { "epoch": 8.93, "learning_rate": 6.000855980807574e-07, "loss": 0.0215, "step": 29019 }, { "epoch": 8.93, "learning_rate": 5.997456272433866e-07, "loss": 0.0257, "step": 29020 }, { "epoch": 8.93, "learning_rate": 5.994057497592032e-07, "loss": 0.0262, "step": 29021 }, { "epoch": 8.93, "learning_rate": 5.99065965631589e-07, "loss": 0.0455, "step": 29022 }, { "epoch": 8.93, "learning_rate": 5.987262748639111e-07, "loss": 0.0225, "step": 29023 }, { "epoch": 8.93, "learning_rate": 5.983866774595448e-07, "loss": 0.0467, "step": 29024 }, { "epoch": 8.93, "learning_rate": 5.980471734218651e-07, "loss": 0.0454, "step": 29025 }, { "epoch": 8.93, "learning_rate": 5.977077627542416e-07, "loss": 0.028, "step": 29026 }, { "epoch": 8.93, "learning_rate": 5.973684454600459e-07, "loss": 0.0226, "step": 29027 }, { "epoch": 8.93, "learning_rate": 5.970292215426476e-07, "loss": 0.0151, "step": 29028 }, { "epoch": 8.93, "learning_rate": 5.966900910054141e-07, "loss": 0.0313, "step": 29029 }, { "epoch": 8.93, "learning_rate": 5.963510538517159e-07, "loss": 0.0158, "step": 29030 }, { "epoch": 8.93, "learning_rate": 5.960121100849159e-07, "loss": 0.038, "step": 29031 }, { "epoch": 8.93, "learning_rate": 5.956732597083837e-07, "loss": 0.0207, "step": 29032 }, { "epoch": 8.93, "learning_rate": 5.953345027254854e-07, "loss": 0.0381, "step": 29033 }, { "epoch": 8.93, "learning_rate": 5.949958391395805e-07, "loss": 0.0364, "step": 29034 }, { "epoch": 8.93, "learning_rate": 5.946572689540353e-07, "loss": 0.0269, "step": 29035 }, { "epoch": 8.93, "learning_rate": 5.943187921722127e-07, "loss": 0.0217, "step": 29036 }, { "epoch": 8.93, "learning_rate": 5.939804087974721e-07, "loss": 0.0309, "step": 29037 }, { "epoch": 8.93, "learning_rate": 5.936421188331754e-07, "loss": 0.031, "step": 29038 }, { "epoch": 8.93, "learning_rate": 5.933039222826797e-07, "loss": 0.0414, "step": 29039 }, { "epoch": 8.94, "learning_rate": 5.929658191493492e-07, "loss": 0.0383, "step": 29040 }, { "epoch": 8.94, "learning_rate": 5.926278094365345e-07, "loss": 0.0224, "step": 29041 }, { "epoch": 8.94, "learning_rate": 5.922898931475973e-07, "loss": 0.0217, "step": 29042 }, { "epoch": 8.94, "learning_rate": 5.919520702858928e-07, "loss": 0.033, "step": 29043 }, { "epoch": 8.94, "learning_rate": 5.916143408547725e-07, "loss": 0.0166, "step": 29044 }, { "epoch": 8.94, "learning_rate": 5.91276704857594e-07, "loss": 0.0317, "step": 29045 }, { "epoch": 8.94, "learning_rate": 5.909391622977101e-07, "loss": 0.0101, "step": 29046 }, { "epoch": 8.94, "learning_rate": 5.906017131784714e-07, "loss": 0.0312, "step": 29047 }, { "epoch": 8.94, "learning_rate": 5.902643575032285e-07, "loss": 0.0379, "step": 29048 }, { "epoch": 8.94, "learning_rate": 5.899270952753333e-07, "loss": 0.0278, "step": 29049 }, { "epoch": 8.94, "learning_rate": 5.895899264981364e-07, "loss": 0.0441, "step": 29050 }, { "epoch": 8.94, "learning_rate": 5.892528511749817e-07, "loss": 0.0435, "step": 29051 }, { "epoch": 8.94, "learning_rate": 5.8891586930922e-07, "loss": 0.0366, "step": 29052 }, { "epoch": 8.94, "learning_rate": 5.885789809041986e-07, "loss": 0.0256, "step": 29053 }, { "epoch": 8.94, "learning_rate": 5.882421859632614e-07, "loss": 0.0369, "step": 29054 }, { "epoch": 8.94, "learning_rate": 5.879054844897536e-07, "loss": 0.0354, "step": 29055 }, { "epoch": 8.94, "learning_rate": 5.875688764870169e-07, "loss": 0.045, "step": 29056 }, { "epoch": 8.94, "learning_rate": 5.872323619583987e-07, "loss": 0.0213, "step": 29057 }, { "epoch": 8.94, "learning_rate": 5.868959409072372e-07, "loss": 0.03, "step": 29058 }, { "epoch": 8.94, "learning_rate": 5.865596133368723e-07, "loss": 0.0222, "step": 29059 }, { "epoch": 8.94, "learning_rate": 5.862233792506477e-07, "loss": 0.0342, "step": 29060 }, { "epoch": 8.94, "learning_rate": 5.858872386519021e-07, "loss": 0.0185, "step": 29061 }, { "epoch": 8.94, "learning_rate": 5.855511915439716e-07, "loss": 0.0162, "step": 29062 }, { "epoch": 8.94, "learning_rate": 5.852152379301946e-07, "loss": 0.0491, "step": 29063 }, { "epoch": 8.94, "learning_rate": 5.848793778139072e-07, "loss": 0.0451, "step": 29064 }, { "epoch": 8.94, "learning_rate": 5.845436111984437e-07, "loss": 0.0217, "step": 29065 }, { "epoch": 8.94, "learning_rate": 5.842079380871402e-07, "loss": 0.0354, "step": 29066 }, { "epoch": 8.94, "learning_rate": 5.838723584833316e-07, "loss": 0.0356, "step": 29067 }, { "epoch": 8.94, "learning_rate": 5.835368723903456e-07, "loss": 0.0343, "step": 29068 }, { "epoch": 8.94, "learning_rate": 5.832014798115182e-07, "loss": 0.0247, "step": 29069 }, { "epoch": 8.94, "learning_rate": 5.82866180750179e-07, "loss": 0.0279, "step": 29070 }, { "epoch": 8.94, "learning_rate": 5.825309752096586e-07, "loss": 0.0178, "step": 29071 }, { "epoch": 8.94, "learning_rate": 5.821958631932845e-07, "loss": 0.0354, "step": 29072 }, { "epoch": 8.95, "learning_rate": 5.818608447043828e-07, "loss": 0.0218, "step": 29073 }, { "epoch": 8.95, "learning_rate": 5.815259197462874e-07, "loss": 0.011, "step": 29074 }, { "epoch": 8.95, "learning_rate": 5.811910883223171e-07, "loss": 0.0296, "step": 29075 }, { "epoch": 8.95, "learning_rate": 5.80856350435799e-07, "loss": 0.0378, "step": 29076 }, { "epoch": 8.95, "learning_rate": 5.805217060900603e-07, "loss": 0.0232, "step": 29077 }, { "epoch": 8.95, "learning_rate": 5.80187155288422e-07, "loss": 0.0305, "step": 29078 }, { "epoch": 8.95, "learning_rate": 5.798526980342068e-07, "loss": 0.0404, "step": 29079 }, { "epoch": 8.95, "learning_rate": 5.795183343307354e-07, "loss": 0.0323, "step": 29080 }, { "epoch": 8.95, "learning_rate": 5.791840641813295e-07, "loss": 0.0426, "step": 29081 }, { "epoch": 8.95, "learning_rate": 5.78849887589309e-07, "loss": 0.0438, "step": 29082 }, { "epoch": 8.95, "learning_rate": 5.785158045579897e-07, "loss": 0.0265, "step": 29083 }, { "epoch": 8.95, "learning_rate": 5.781818150906948e-07, "loss": 0.0484, "step": 29084 }, { "epoch": 8.95, "learning_rate": 5.77847919190736e-07, "loss": 0.026, "step": 29085 }, { "epoch": 8.95, "learning_rate": 5.775141168614306e-07, "loss": 0.0451, "step": 29086 }, { "epoch": 8.95, "learning_rate": 5.77180408106095e-07, "loss": 0.0435, "step": 29087 }, { "epoch": 8.95, "learning_rate": 5.76846792928042e-07, "loss": 0.0339, "step": 29088 }, { "epoch": 8.95, "learning_rate": 5.765132713305854e-07, "loss": 0.0258, "step": 29089 }, { "epoch": 8.95, "learning_rate": 5.761798433170362e-07, "loss": 0.0295, "step": 29090 }, { "epoch": 8.95, "learning_rate": 5.758465088907072e-07, "loss": 0.0326, "step": 29091 }, { "epoch": 8.95, "learning_rate": 5.755132680549081e-07, "loss": 0.0139, "step": 29092 }, { "epoch": 8.95, "learning_rate": 5.751801208129459e-07, "loss": 0.0385, "step": 29093 }, { "epoch": 8.95, "learning_rate": 5.748470671681328e-07, "loss": 0.0263, "step": 29094 }, { "epoch": 8.95, "learning_rate": 5.745141071237759e-07, "loss": 0.0348, "step": 29095 }, { "epoch": 8.95, "learning_rate": 5.741812406831793e-07, "loss": 0.0424, "step": 29096 }, { "epoch": 8.95, "learning_rate": 5.738484678496503e-07, "loss": 0.0227, "step": 29097 }, { "epoch": 8.95, "learning_rate": 5.735157886264931e-07, "loss": 0.0364, "step": 29098 }, { "epoch": 8.95, "learning_rate": 5.731832030170126e-07, "loss": 0.0335, "step": 29099 }, { "epoch": 8.95, "learning_rate": 5.728507110245107e-07, "loss": 0.0287, "step": 29100 }, { "epoch": 8.95, "learning_rate": 5.725183126522893e-07, "loss": 0.0324, "step": 29101 }, { "epoch": 8.95, "learning_rate": 5.721860079036501e-07, "loss": 0.0203, "step": 29102 }, { "epoch": 8.95, "learning_rate": 5.718537967818915e-07, "loss": 0.0344, "step": 29103 }, { "epoch": 8.95, "learning_rate": 5.715216792903144e-07, "loss": 0.0224, "step": 29104 }, { "epoch": 8.96, "learning_rate": 5.711896554322194e-07, "loss": 0.03, "step": 29105 }, { "epoch": 8.96, "learning_rate": 5.708577252108971e-07, "loss": 0.0372, "step": 29106 }, { "epoch": 8.96, "learning_rate": 5.705258886296494e-07, "loss": 0.0201, "step": 29107 }, { "epoch": 8.96, "learning_rate": 5.701941456917703e-07, "loss": 0.0138, "step": 29108 }, { "epoch": 8.96, "learning_rate": 5.698624964005539e-07, "loss": 0.0384, "step": 29109 }, { "epoch": 8.96, "learning_rate": 5.695309407592941e-07, "loss": 0.0289, "step": 29110 }, { "epoch": 8.96, "learning_rate": 5.691994787712818e-07, "loss": 0.0222, "step": 29111 }, { "epoch": 8.96, "learning_rate": 5.688681104398131e-07, "loss": 0.067, "step": 29112 }, { "epoch": 8.96, "learning_rate": 5.685368357681731e-07, "loss": 0.0291, "step": 29113 }, { "epoch": 8.96, "learning_rate": 5.68205654759657e-07, "loss": 0.0374, "step": 29114 }, { "epoch": 8.96, "learning_rate": 5.678745674175501e-07, "loss": 0.0261, "step": 29115 }, { "epoch": 8.96, "learning_rate": 5.675435737451429e-07, "loss": 0.0271, "step": 29116 }, { "epoch": 8.96, "learning_rate": 5.672126737457207e-07, "loss": 0.0321, "step": 29117 }, { "epoch": 8.96, "learning_rate": 5.668818674225684e-07, "loss": 0.0336, "step": 29118 }, { "epoch": 8.96, "learning_rate": 5.665511547789759e-07, "loss": 0.0277, "step": 29119 }, { "epoch": 8.96, "learning_rate": 5.662205358182226e-07, "loss": 0.0345, "step": 29120 }, { "epoch": 8.96, "learning_rate": 5.658900105435938e-07, "loss": 0.0266, "step": 29121 }, { "epoch": 8.96, "learning_rate": 5.655595789583735e-07, "loss": 0.0346, "step": 29122 }, { "epoch": 8.96, "learning_rate": 5.6522924106584e-07, "loss": 0.0366, "step": 29123 }, { "epoch": 8.96, "learning_rate": 5.648989968692753e-07, "loss": 0.0388, "step": 29124 }, { "epoch": 8.96, "learning_rate": 5.6456884637196e-07, "loss": 0.0244, "step": 29125 }, { "epoch": 8.96, "learning_rate": 5.642387895771728e-07, "loss": 0.0343, "step": 29126 }, { "epoch": 8.96, "learning_rate": 5.639088264881898e-07, "loss": 0.0282, "step": 29127 }, { "epoch": 8.96, "learning_rate": 5.635789571082872e-07, "loss": 0.0303, "step": 29128 }, { "epoch": 8.96, "learning_rate": 5.632491814407459e-07, "loss": 0.0387, "step": 29129 }, { "epoch": 8.96, "learning_rate": 5.629194994888354e-07, "loss": 0.0347, "step": 29130 }, { "epoch": 8.96, "learning_rate": 5.625899112558308e-07, "loss": 0.0165, "step": 29131 }, { "epoch": 8.96, "learning_rate": 5.622604167450074e-07, "loss": 0.0197, "step": 29132 }, { "epoch": 8.96, "learning_rate": 5.619310159596358e-07, "loss": 0.0219, "step": 29133 }, { "epoch": 8.96, "learning_rate": 5.616017089029879e-07, "loss": 0.037, "step": 29134 }, { "epoch": 8.96, "learning_rate": 5.612724955783333e-07, "loss": 0.0267, "step": 29135 }, { "epoch": 8.96, "learning_rate": 5.609433759889426e-07, "loss": 0.0315, "step": 29136 }, { "epoch": 8.96, "learning_rate": 5.606143501380823e-07, "loss": 0.0325, "step": 29137 }, { "epoch": 8.97, "learning_rate": 5.602854180290207e-07, "loss": 0.0295, "step": 29138 }, { "epoch": 8.97, "learning_rate": 5.599565796650275e-07, "loss": 0.0252, "step": 29139 }, { "epoch": 8.97, "learning_rate": 5.596278350493634e-07, "loss": 0.027, "step": 29140 }, { "epoch": 8.97, "learning_rate": 5.592991841852946e-07, "loss": 0.01, "step": 29141 }, { "epoch": 8.97, "learning_rate": 5.589706270760875e-07, "loss": 0.0355, "step": 29142 }, { "epoch": 8.97, "learning_rate": 5.586421637250028e-07, "loss": 0.0269, "step": 29143 }, { "epoch": 8.97, "learning_rate": 5.583137941353034e-07, "loss": 0.031, "step": 29144 }, { "epoch": 8.97, "learning_rate": 5.57985518310249e-07, "loss": 0.0409, "step": 29145 }, { "epoch": 8.97, "learning_rate": 5.576573362531001e-07, "loss": 0.0141, "step": 29146 }, { "epoch": 8.97, "learning_rate": 5.573292479671166e-07, "loss": 0.0321, "step": 29147 }, { "epoch": 8.97, "learning_rate": 5.570012534555558e-07, "loss": 0.036, "step": 29148 }, { "epoch": 8.97, "learning_rate": 5.566733527216761e-07, "loss": 0.0315, "step": 29149 }, { "epoch": 8.97, "learning_rate": 5.563455457687339e-07, "loss": 0.0265, "step": 29150 }, { "epoch": 8.97, "learning_rate": 5.560178325999821e-07, "loss": 0.0314, "step": 29151 }, { "epoch": 8.97, "learning_rate": 5.556902132186781e-07, "loss": 0.0325, "step": 29152 }, { "epoch": 8.97, "learning_rate": 5.553626876280749e-07, "loss": 0.0244, "step": 29153 }, { "epoch": 8.97, "learning_rate": 5.550352558314242e-07, "loss": 0.0399, "step": 29154 }, { "epoch": 8.97, "learning_rate": 5.547079178319781e-07, "loss": 0.0207, "step": 29155 }, { "epoch": 8.97, "learning_rate": 5.543806736329859e-07, "loss": 0.0243, "step": 29156 }, { "epoch": 8.97, "learning_rate": 5.54053523237702e-07, "loss": 0.0206, "step": 29157 }, { "epoch": 8.97, "learning_rate": 5.537264666493703e-07, "loss": 0.0113, "step": 29158 }, { "epoch": 8.97, "learning_rate": 5.533995038712403e-07, "loss": 0.0318, "step": 29159 }, { "epoch": 8.97, "learning_rate": 5.530726349065618e-07, "loss": 0.0278, "step": 29160 }, { "epoch": 8.97, "learning_rate": 5.527458597585766e-07, "loss": 0.0503, "step": 29161 }, { "epoch": 8.97, "learning_rate": 5.524191784305321e-07, "loss": 0.0287, "step": 29162 }, { "epoch": 8.97, "learning_rate": 5.520925909256724e-07, "loss": 0.0343, "step": 29163 }, { "epoch": 8.97, "learning_rate": 5.517660972472416e-07, "loss": 0.0118, "step": 29164 }, { "epoch": 8.97, "learning_rate": 5.514396973984803e-07, "loss": 0.02, "step": 29165 }, { "epoch": 8.97, "learning_rate": 5.511133913826294e-07, "loss": 0.0319, "step": 29166 }, { "epoch": 8.97, "learning_rate": 5.50787179202934e-07, "loss": 0.056, "step": 29167 }, { "epoch": 8.97, "learning_rate": 5.504610608626281e-07, "loss": 0.0363, "step": 29168 }, { "epoch": 8.97, "learning_rate": 5.501350363649537e-07, "loss": 0.0188, "step": 29169 }, { "epoch": 8.98, "learning_rate": 5.498091057131483e-07, "loss": 0.0336, "step": 29170 }, { "epoch": 8.98, "learning_rate": 5.494832689104479e-07, "loss": 0.0236, "step": 29171 }, { "epoch": 8.98, "learning_rate": 5.491575259600879e-07, "loss": 0.0131, "step": 29172 }, { "epoch": 8.98, "learning_rate": 5.488318768653022e-07, "loss": 0.0116, "step": 29173 }, { "epoch": 8.98, "learning_rate": 5.485063216293296e-07, "loss": 0.0375, "step": 29174 }, { "epoch": 8.98, "learning_rate": 5.481808602553995e-07, "loss": 0.0388, "step": 29175 }, { "epoch": 8.98, "learning_rate": 5.478554927467417e-07, "loss": 0.0267, "step": 29176 }, { "epoch": 8.98, "learning_rate": 5.475302191065923e-07, "loss": 0.0301, "step": 29177 }, { "epoch": 8.98, "learning_rate": 5.47205039338179e-07, "loss": 0.0214, "step": 29178 }, { "epoch": 8.98, "learning_rate": 5.468799534447322e-07, "loss": 0.0146, "step": 29179 }, { "epoch": 8.98, "learning_rate": 5.465549614294796e-07, "loss": 0.0245, "step": 29180 }, { "epoch": 8.98, "learning_rate": 5.462300632956485e-07, "loss": 0.0329, "step": 29181 }, { "epoch": 8.98, "learning_rate": 5.459052590464664e-07, "loss": 0.0081, "step": 29182 }, { "epoch": 8.98, "learning_rate": 5.455805486851562e-07, "loss": 0.0227, "step": 29183 }, { "epoch": 8.98, "learning_rate": 5.452559322149475e-07, "loss": 0.0227, "step": 29184 }, { "epoch": 8.98, "learning_rate": 5.449314096390601e-07, "loss": 0.0403, "step": 29185 }, { "epoch": 8.98, "learning_rate": 5.446069809607157e-07, "loss": 0.0295, "step": 29186 }, { "epoch": 8.98, "learning_rate": 5.442826461831408e-07, "loss": 0.04, "step": 29187 }, { "epoch": 8.98, "learning_rate": 5.439584053095525e-07, "loss": 0.0245, "step": 29188 }, { "epoch": 8.98, "learning_rate": 5.436342583431731e-07, "loss": 0.0316, "step": 29189 }, { "epoch": 8.98, "learning_rate": 5.433102052872207e-07, "loss": 0.0221, "step": 29190 }, { "epoch": 8.98, "learning_rate": 5.429862461449132e-07, "loss": 0.0346, "step": 29191 }, { "epoch": 8.98, "learning_rate": 5.426623809194687e-07, "loss": 0.0314, "step": 29192 }, { "epoch": 8.98, "learning_rate": 5.423386096141003e-07, "loss": 0.0507, "step": 29193 }, { "epoch": 8.98, "learning_rate": 5.420149322320289e-07, "loss": 0.033, "step": 29194 }, { "epoch": 8.98, "learning_rate": 5.416913487764663e-07, "loss": 0.0233, "step": 29195 }, { "epoch": 8.98, "learning_rate": 5.413678592506233e-07, "loss": 0.0293, "step": 29196 }, { "epoch": 8.98, "learning_rate": 5.410444636577161e-07, "loss": 0.0362, "step": 29197 }, { "epoch": 8.98, "learning_rate": 5.407211620009545e-07, "loss": 0.015, "step": 29198 }, { "epoch": 8.98, "learning_rate": 5.403979542835502e-07, "loss": 0.0541, "step": 29199 }, { "epoch": 8.98, "learning_rate": 5.400748405087131e-07, "loss": 0.0393, "step": 29200 }, { "epoch": 8.98, "learning_rate": 5.397518206796504e-07, "loss": 0.0324, "step": 29201 }, { "epoch": 8.98, "learning_rate": 5.394288947995707e-07, "loss": 0.0366, "step": 29202 }, { "epoch": 8.99, "learning_rate": 5.391060628716805e-07, "loss": 0.0347, "step": 29203 }, { "epoch": 8.99, "learning_rate": 5.38783324899188e-07, "loss": 0.0524, "step": 29204 }, { "epoch": 8.99, "learning_rate": 5.384606808852977e-07, "loss": 0.0341, "step": 29205 }, { "epoch": 8.99, "learning_rate": 5.381381308332101e-07, "loss": 0.0318, "step": 29206 }, { "epoch": 8.99, "learning_rate": 5.378156747461328e-07, "loss": 0.0157, "step": 29207 }, { "epoch": 8.99, "learning_rate": 5.374933126272664e-07, "loss": 0.0211, "step": 29208 }, { "epoch": 8.99, "learning_rate": 5.371710444798117e-07, "loss": 0.043, "step": 29209 }, { "epoch": 8.99, "learning_rate": 5.368488703069708e-07, "loss": 0.0351, "step": 29210 }, { "epoch": 8.99, "learning_rate": 5.365267901119398e-07, "loss": 0.0204, "step": 29211 }, { "epoch": 8.99, "learning_rate": 5.362048038979228e-07, "loss": 0.0345, "step": 29212 }, { "epoch": 8.99, "learning_rate": 5.358829116681119e-07, "loss": 0.0306, "step": 29213 }, { "epoch": 8.99, "learning_rate": 5.355611134257066e-07, "loss": 0.0357, "step": 29214 }, { "epoch": 8.99, "learning_rate": 5.352394091739022e-07, "loss": 0.0342, "step": 29215 }, { "epoch": 8.99, "learning_rate": 5.349177989158938e-07, "loss": 0.016, "step": 29216 }, { "epoch": 8.99, "learning_rate": 5.345962826548745e-07, "loss": 0.0253, "step": 29217 }, { "epoch": 8.99, "learning_rate": 5.342748603940384e-07, "loss": 0.0178, "step": 29218 }, { "epoch": 8.99, "learning_rate": 5.339535321365763e-07, "loss": 0.0316, "step": 29219 }, { "epoch": 8.99, "learning_rate": 5.3363229788568e-07, "loss": 0.0172, "step": 29220 }, { "epoch": 8.99, "learning_rate": 5.333111576445383e-07, "loss": 0.0327, "step": 29221 }, { "epoch": 8.99, "learning_rate": 5.329901114163438e-07, "loss": 0.0251, "step": 29222 }, { "epoch": 8.99, "learning_rate": 5.326691592042798e-07, "loss": 0.0291, "step": 29223 }, { "epoch": 8.99, "learning_rate": 5.323483010115382e-07, "loss": 0.0244, "step": 29224 }, { "epoch": 8.99, "learning_rate": 5.320275368413031e-07, "loss": 0.0243, "step": 29225 }, { "epoch": 8.99, "learning_rate": 5.317068666967607e-07, "loss": 0.0128, "step": 29226 }, { "epoch": 8.99, "learning_rate": 5.313862905810951e-07, "loss": 0.0355, "step": 29227 }, { "epoch": 8.99, "learning_rate": 5.310658084974896e-07, "loss": 0.0288, "step": 29228 }, { "epoch": 8.99, "learning_rate": 5.307454204491303e-07, "loss": 0.0332, "step": 29229 }, { "epoch": 8.99, "learning_rate": 5.304251264391935e-07, "loss": 0.0284, "step": 29230 }, { "epoch": 8.99, "learning_rate": 5.301049264708613e-07, "loss": 0.0357, "step": 29231 }, { "epoch": 8.99, "learning_rate": 5.297848205473177e-07, "loss": 0.0347, "step": 29232 }, { "epoch": 8.99, "learning_rate": 5.29464808671738e-07, "loss": 0.0326, "step": 29233 }, { "epoch": 8.99, "learning_rate": 5.291448908473007e-07, "loss": 0.026, "step": 29234 }, { "epoch": 9.0, "learning_rate": 5.288250670771832e-07, "loss": 0.0203, "step": 29235 }, { "epoch": 9.0, "learning_rate": 5.285053373645621e-07, "loss": 0.0164, "step": 29236 }, { "epoch": 9.0, "learning_rate": 5.281857017126124e-07, "loss": 0.0228, "step": 29237 }, { "epoch": 9.0, "learning_rate": 5.278661601245061e-07, "loss": 0.0344, "step": 29238 }, { "epoch": 9.0, "learning_rate": 5.275467126034217e-07, "loss": 0.0306, "step": 29239 }, { "epoch": 9.0, "learning_rate": 5.272273591525279e-07, "loss": 0.0331, "step": 29240 }, { "epoch": 9.0, "learning_rate": 5.269080997749942e-07, "loss": 0.0421, "step": 29241 }, { "epoch": 9.0, "learning_rate": 5.265889344739961e-07, "loss": 0.0377, "step": 29242 }, { "epoch": 9.0, "learning_rate": 5.262698632527008e-07, "loss": 0.0314, "step": 29243 }, { "epoch": 9.0, "learning_rate": 5.259508861142759e-07, "loss": 0.0253, "step": 29244 }, { "epoch": 9.0, "learning_rate": 5.256320030618912e-07, "loss": 0.0589, "step": 29245 }, { "epoch": 9.0, "learning_rate": 5.253132140987116e-07, "loss": 0.0255, "step": 29246 }, { "epoch": 9.0, "learning_rate": 5.249945192279049e-07, "loss": 0.0357, "step": 29247 }, { "epoch": 9.0, "learning_rate": 5.246759184526328e-07, "loss": 0.0275, "step": 29248 }, { "epoch": 9.0, "learning_rate": 5.243574117760642e-07, "loss": 0.0252, "step": 29249 }, { "epoch": 9.0, "learning_rate": 5.240389992013606e-07, "loss": 0.0247, "step": 29250 }, { "epoch": 9.0, "learning_rate": 5.237206807316797e-07, "loss": 0.0271, "step": 29251 }, { "epoch": 9.0, "learning_rate": 5.234024563701867e-07, "loss": 0.0355, "step": 29252 }, { "epoch": 9.0, "learning_rate": 5.230843261200425e-07, "loss": 0.0212, "step": 29253 }, { "epoch": 9.0, "learning_rate": 5.227662899844043e-07, "loss": 0.0253, "step": 29254 }, { "epoch": 9.0, "learning_rate": 5.224483479664322e-07, "loss": 0.0175, "step": 29255 }, { "epoch": 9.0, "learning_rate": 5.221305000692811e-07, "loss": 0.0178, "step": 29256 }, { "epoch": 9.0, "learning_rate": 5.218127462961109e-07, "loss": 0.0289, "step": 29257 }, { "epoch": 9.0, "learning_rate": 5.214950866500734e-07, "loss": 0.0248, "step": 29258 }, { "epoch": 9.0, "learning_rate": 5.211775211343273e-07, "loss": 0.0441, "step": 29259 }, { "epoch": 9.0, "learning_rate": 5.208600497520255e-07, "loss": 0.0226, "step": 29260 }, { "epoch": 9.0, "learning_rate": 5.205426725063167e-07, "loss": 0.0178, "step": 29261 }, { "epoch": 9.0, "learning_rate": 5.202253894003573e-07, "loss": 0.0186, "step": 29262 }, { "epoch": 9.0, "learning_rate": 5.199082004372958e-07, "loss": 0.028, "step": 29263 }, { "epoch": 9.0, "learning_rate": 5.19591105620284e-07, "loss": 0.0299, "step": 29264 }, { "epoch": 9.0, "learning_rate": 5.192741049524674e-07, "loss": 0.0307, "step": 29265 }, { "epoch": 9.0, "learning_rate": 5.18957198436999e-07, "loss": 0.02, "step": 29266 }, { "epoch": 9.0, "learning_rate": 5.18640386077025e-07, "loss": 0.0093, "step": 29267 }, { "epoch": 9.01, "learning_rate": 5.183236678756864e-07, "loss": 0.0304, "step": 29268 }, { "epoch": 9.01, "learning_rate": 5.180070438361351e-07, "loss": 0.0261, "step": 29269 }, { "epoch": 9.01, "learning_rate": 5.176905139615118e-07, "loss": 0.0242, "step": 29270 }, { "epoch": 9.01, "learning_rate": 5.173740782549619e-07, "loss": 0.0375, "step": 29271 }, { "epoch": 9.01, "learning_rate": 5.170577367196262e-07, "loss": 0.026, "step": 29272 }, { "epoch": 9.01, "learning_rate": 5.167414893586453e-07, "loss": 0.0164, "step": 29273 }, { "epoch": 9.01, "learning_rate": 5.164253361751637e-07, "loss": 0.0193, "step": 29274 }, { "epoch": 9.01, "learning_rate": 5.161092771723164e-07, "loss": 0.0202, "step": 29275 }, { "epoch": 9.01, "learning_rate": 5.157933123532466e-07, "loss": 0.0131, "step": 29276 }, { "epoch": 9.01, "learning_rate": 5.154774417210895e-07, "loss": 0.0217, "step": 29277 }, { "epoch": 9.01, "learning_rate": 5.151616652789837e-07, "loss": 0.0187, "step": 29278 }, { "epoch": 9.01, "learning_rate": 5.148459830300623e-07, "loss": 0.0355, "step": 29279 }, { "epoch": 9.01, "learning_rate": 5.145303949774638e-07, "loss": 0.0384, "step": 29280 }, { "epoch": 9.01, "learning_rate": 5.142149011243192e-07, "loss": 0.0173, "step": 29281 }, { "epoch": 9.01, "learning_rate": 5.138995014737636e-07, "loss": 0.0241, "step": 29282 }, { "epoch": 9.01, "learning_rate": 5.135841960289267e-07, "loss": 0.0182, "step": 29283 }, { "epoch": 9.01, "learning_rate": 5.132689847929451e-07, "loss": 0.0234, "step": 29284 }, { "epoch": 9.01, "learning_rate": 5.129538677689427e-07, "loss": 0.0063, "step": 29285 }, { "epoch": 9.01, "learning_rate": 5.126388449600539e-07, "loss": 0.0242, "step": 29286 }, { "epoch": 9.01, "learning_rate": 5.123239163694049e-07, "loss": 0.0279, "step": 29287 }, { "epoch": 9.01, "learning_rate": 5.120090820001222e-07, "loss": 0.0135, "step": 29288 }, { "epoch": 9.01, "learning_rate": 5.116943418553355e-07, "loss": 0.024, "step": 29289 }, { "epoch": 9.01, "learning_rate": 5.113796959381667e-07, "loss": 0.0248, "step": 29290 }, { "epoch": 9.01, "learning_rate": 5.110651442517445e-07, "loss": 0.0269, "step": 29291 }, { "epoch": 9.01, "learning_rate": 5.107506867991896e-07, "loss": 0.0184, "step": 29292 }, { "epoch": 9.01, "learning_rate": 5.104363235836251e-07, "loss": 0.0095, "step": 29293 }, { "epoch": 9.01, "learning_rate": 5.101220546081753e-07, "loss": 0.0288, "step": 29294 }, { "epoch": 9.01, "learning_rate": 5.098078798759587e-07, "loss": 0.013, "step": 29295 }, { "epoch": 9.01, "learning_rate": 5.094937993900962e-07, "loss": 0.0338, "step": 29296 }, { "epoch": 9.01, "learning_rate": 5.091798131537085e-07, "loss": 0.0189, "step": 29297 }, { "epoch": 9.01, "learning_rate": 5.08865921169911e-07, "loss": 0.029, "step": 29298 }, { "epoch": 9.01, "learning_rate": 5.085521234418223e-07, "loss": 0.0108, "step": 29299 }, { "epoch": 9.02, "learning_rate": 5.082384199725576e-07, "loss": 0.0446, "step": 29300 }, { "epoch": 9.02, "learning_rate": 5.079248107652357e-07, "loss": 0.0224, "step": 29301 }, { "epoch": 9.02, "learning_rate": 5.076112958229673e-07, "loss": 0.0291, "step": 29302 }, { "epoch": 9.02, "learning_rate": 5.072978751488655e-07, "loss": 0.0312, "step": 29303 }, { "epoch": 9.02, "learning_rate": 5.069845487460456e-07, "loss": 0.0223, "step": 29304 }, { "epoch": 9.02, "learning_rate": 5.066713166176196e-07, "loss": 0.0182, "step": 29305 }, { "epoch": 9.02, "learning_rate": 5.06358178766696e-07, "loss": 0.0266, "step": 29306 }, { "epoch": 9.02, "learning_rate": 5.060451351963846e-07, "loss": 0.0186, "step": 29307 }, { "epoch": 9.02, "learning_rate": 5.057321859097952e-07, "loss": 0.0242, "step": 29308 }, { "epoch": 9.02, "learning_rate": 5.054193309100364e-07, "loss": 0.0351, "step": 29309 }, { "epoch": 9.02, "learning_rate": 5.051065702002111e-07, "loss": 0.02, "step": 29310 }, { "epoch": 9.02, "learning_rate": 5.047939037834304e-07, "loss": 0.0207, "step": 29311 }, { "epoch": 9.02, "learning_rate": 5.044813316627994e-07, "loss": 0.0272, "step": 29312 }, { "epoch": 9.02, "learning_rate": 5.041688538414169e-07, "loss": 0.0282, "step": 29313 }, { "epoch": 9.02, "learning_rate": 5.038564703223903e-07, "loss": 0.0297, "step": 29314 }, { "epoch": 9.02, "learning_rate": 5.035441811088204e-07, "loss": 0.0212, "step": 29315 }, { "epoch": 9.02, "learning_rate": 5.032319862038104e-07, "loss": 0.03, "step": 29316 }, { "epoch": 9.02, "learning_rate": 5.029198856104578e-07, "loss": 0.0112, "step": 29317 }, { "epoch": 9.02, "learning_rate": 5.026078793318645e-07, "loss": 0.0137, "step": 29318 }, { "epoch": 9.02, "learning_rate": 5.02295967371128e-07, "loss": 0.0332, "step": 29319 }, { "epoch": 9.02, "learning_rate": 5.019841497313438e-07, "loss": 0.0337, "step": 29320 }, { "epoch": 9.02, "learning_rate": 5.016724264156125e-07, "loss": 0.0176, "step": 29321 }, { "epoch": 9.02, "learning_rate": 5.013607974270296e-07, "loss": 0.0255, "step": 29322 }, { "epoch": 9.02, "learning_rate": 5.010492627686848e-07, "loss": 0.0123, "step": 29323 }, { "epoch": 9.02, "learning_rate": 5.007378224436765e-07, "loss": 0.0144, "step": 29324 }, { "epoch": 9.02, "learning_rate": 5.004264764550971e-07, "loss": 0.0234, "step": 29325 }, { "epoch": 9.02, "learning_rate": 5.001152248060382e-07, "loss": 0.0306, "step": 29326 }, { "epoch": 9.02, "learning_rate": 4.998040674995897e-07, "loss": 0.0285, "step": 29327 }, { "epoch": 9.02, "learning_rate": 4.994930045388414e-07, "loss": 0.0067, "step": 29328 }, { "epoch": 9.02, "learning_rate": 4.99182035926885e-07, "loss": 0.0207, "step": 29329 }, { "epoch": 9.02, "learning_rate": 4.988711616668062e-07, "loss": 0.0358, "step": 29330 }, { "epoch": 9.02, "learning_rate": 4.985603817616936e-07, "loss": 0.0282, "step": 29331 }, { "epoch": 9.02, "learning_rate": 4.982496962146333e-07, "loss": 0.0273, "step": 29332 }, { "epoch": 9.03, "learning_rate": 4.979391050287108e-07, "loss": 0.024, "step": 29333 }, { "epoch": 9.03, "learning_rate": 4.976286082070103e-07, "loss": 0.0149, "step": 29334 }, { "epoch": 9.03, "learning_rate": 4.973182057526149e-07, "loss": 0.0303, "step": 29335 }, { "epoch": 9.03, "learning_rate": 4.970078976686088e-07, "loss": 0.0476, "step": 29336 }, { "epoch": 9.03, "learning_rate": 4.966976839580728e-07, "loss": 0.0196, "step": 29337 }, { "epoch": 9.03, "learning_rate": 4.963875646240846e-07, "loss": 0.0303, "step": 29338 }, { "epoch": 9.03, "learning_rate": 4.960775396697293e-07, "loss": 0.023, "step": 29339 }, { "epoch": 9.03, "learning_rate": 4.957676090980812e-07, "loss": 0.0384, "step": 29340 }, { "epoch": 9.03, "learning_rate": 4.954577729122212e-07, "loss": 0.027, "step": 29341 }, { "epoch": 9.03, "learning_rate": 4.951480311152257e-07, "loss": 0.0263, "step": 29342 }, { "epoch": 9.03, "learning_rate": 4.94838383710169e-07, "loss": 0.0361, "step": 29343 }, { "epoch": 9.03, "learning_rate": 4.945288307001284e-07, "loss": 0.022, "step": 29344 }, { "epoch": 9.03, "learning_rate": 4.942193720881749e-07, "loss": 0.0173, "step": 29345 }, { "epoch": 9.03, "learning_rate": 4.939100078773873e-07, "loss": 0.0191, "step": 29346 }, { "epoch": 9.03, "learning_rate": 4.936007380708319e-07, "loss": 0.0367, "step": 29347 }, { "epoch": 9.03, "learning_rate": 4.932915626715817e-07, "loss": 0.0283, "step": 29348 }, { "epoch": 9.03, "learning_rate": 4.9298248168271e-07, "loss": 0.024, "step": 29349 }, { "epoch": 9.03, "learning_rate": 4.926734951072831e-07, "loss": 0.0197, "step": 29350 }, { "epoch": 9.03, "learning_rate": 4.923646029483709e-07, "loss": 0.0186, "step": 29351 }, { "epoch": 9.03, "learning_rate": 4.920558052090418e-07, "loss": 0.0168, "step": 29352 }, { "epoch": 9.03, "learning_rate": 4.917471018923603e-07, "loss": 0.0253, "step": 29353 }, { "epoch": 9.03, "learning_rate": 4.914384930013927e-07, "loss": 0.024, "step": 29354 }, { "epoch": 9.03, "learning_rate": 4.911299785392032e-07, "loss": 0.0174, "step": 29355 }, { "epoch": 9.03, "learning_rate": 4.908215585088583e-07, "loss": 0.0315, "step": 29356 }, { "epoch": 9.03, "learning_rate": 4.90513232913421e-07, "loss": 0.0254, "step": 29357 }, { "epoch": 9.03, "learning_rate": 4.902050017559479e-07, "loss": 0.0205, "step": 29358 }, { "epoch": 9.03, "learning_rate": 4.898968650395053e-07, "loss": 0.0251, "step": 29359 }, { "epoch": 9.03, "learning_rate": 4.89588822767152e-07, "loss": 0.027, "step": 29360 }, { "epoch": 9.03, "learning_rate": 4.892808749419465e-07, "loss": 0.0194, "step": 29361 }, { "epoch": 9.03, "learning_rate": 4.889730215669475e-07, "loss": 0.0168, "step": 29362 }, { "epoch": 9.03, "learning_rate": 4.886652626452115e-07, "loss": 0.0186, "step": 29363 }, { "epoch": 9.03, "learning_rate": 4.88357598179795e-07, "loss": 0.0312, "step": 29364 }, { "epoch": 9.04, "learning_rate": 4.880500281737533e-07, "loss": 0.0323, "step": 29365 }, { "epoch": 9.04, "learning_rate": 4.877425526301427e-07, "loss": 0.0245, "step": 29366 }, { "epoch": 9.04, "learning_rate": 4.874351715520154e-07, "loss": 0.0162, "step": 29367 }, { "epoch": 9.04, "learning_rate": 4.871278849424221e-07, "loss": 0.0212, "step": 29368 }, { "epoch": 9.04, "learning_rate": 4.868206928044183e-07, "loss": 0.0217, "step": 29369 }, { "epoch": 9.04, "learning_rate": 4.865135951410515e-07, "loss": 0.0185, "step": 29370 }, { "epoch": 9.04, "learning_rate": 4.862065919553727e-07, "loss": 0.026, "step": 29371 }, { "epoch": 9.04, "learning_rate": 4.858996832504314e-07, "loss": 0.012, "step": 29372 }, { "epoch": 9.04, "learning_rate": 4.855928690292722e-07, "loss": 0.0253, "step": 29373 }, { "epoch": 9.04, "learning_rate": 4.852861492949479e-07, "loss": 0.0169, "step": 29374 }, { "epoch": 9.04, "learning_rate": 4.849795240504984e-07, "loss": 0.0201, "step": 29375 }, { "epoch": 9.04, "learning_rate": 4.846729932989724e-07, "loss": 0.0226, "step": 29376 }, { "epoch": 9.04, "learning_rate": 4.843665570434153e-07, "loss": 0.0314, "step": 29377 }, { "epoch": 9.04, "learning_rate": 4.840602152868656e-07, "loss": 0.015, "step": 29378 }, { "epoch": 9.04, "learning_rate": 4.837539680323689e-07, "loss": 0.0253, "step": 29379 }, { "epoch": 9.04, "learning_rate": 4.834478152829658e-07, "loss": 0.0219, "step": 29380 }, { "epoch": 9.04, "learning_rate": 4.831417570416963e-07, "loss": 0.0131, "step": 29381 }, { "epoch": 9.04, "learning_rate": 4.828357933116001e-07, "loss": 0.0293, "step": 29382 }, { "epoch": 9.04, "learning_rate": 4.825299240957149e-07, "loss": 0.0263, "step": 29383 }, { "epoch": 9.04, "learning_rate": 4.822241493970814e-07, "loss": 0.0269, "step": 29384 }, { "epoch": 9.04, "learning_rate": 4.819184692187317e-07, "loss": 0.0223, "step": 29385 }, { "epoch": 9.04, "learning_rate": 4.816128835637046e-07, "loss": 0.0239, "step": 29386 }, { "epoch": 9.04, "learning_rate": 4.813073924350342e-07, "loss": 0.0179, "step": 29387 }, { "epoch": 9.04, "learning_rate": 4.810019958357548e-07, "loss": 0.0232, "step": 29388 }, { "epoch": 9.04, "learning_rate": 4.806966937688973e-07, "loss": 0.0309, "step": 29389 }, { "epoch": 9.04, "learning_rate": 4.803914862374947e-07, "loss": 0.0239, "step": 29390 }, { "epoch": 9.04, "learning_rate": 4.800863732445804e-07, "loss": 0.0187, "step": 29391 }, { "epoch": 9.04, "learning_rate": 4.797813547931807e-07, "loss": 0.0261, "step": 29392 }, { "epoch": 9.04, "learning_rate": 4.794764308863242e-07, "loss": 0.0169, "step": 29393 }, { "epoch": 9.04, "learning_rate": 4.791716015270431e-07, "loss": 0.01, "step": 29394 }, { "epoch": 9.04, "learning_rate": 4.788668667183627e-07, "loss": 0.0366, "step": 29395 }, { "epoch": 9.04, "learning_rate": 4.785622264633094e-07, "loss": 0.0259, "step": 29396 }, { "epoch": 9.04, "learning_rate": 4.782576807649086e-07, "loss": 0.0201, "step": 29397 }, { "epoch": 9.05, "learning_rate": 4.779532296261846e-07, "loss": 0.0287, "step": 29398 }, { "epoch": 9.05, "learning_rate": 4.776488730501605e-07, "loss": 0.0273, "step": 29399 }, { "epoch": 9.05, "learning_rate": 4.773446110398571e-07, "loss": 0.0178, "step": 29400 }, { "epoch": 9.05, "learning_rate": 4.770404435983012e-07, "loss": 0.0396, "step": 29401 }, { "epoch": 9.05, "learning_rate": 4.7673637072851e-07, "loss": 0.0249, "step": 29402 }, { "epoch": 9.05, "learning_rate": 4.764323924335013e-07, "loss": 0.0227, "step": 29403 }, { "epoch": 9.05, "learning_rate": 4.76128508716297e-07, "loss": 0.0125, "step": 29404 }, { "epoch": 9.05, "learning_rate": 4.758247195799148e-07, "loss": 0.0204, "step": 29405 }, { "epoch": 9.05, "learning_rate": 4.755210250273701e-07, "loss": 0.0165, "step": 29406 }, { "epoch": 9.05, "learning_rate": 4.752174250616803e-07, "loss": 0.0205, "step": 29407 }, { "epoch": 9.05, "learning_rate": 4.7491391968585987e-07, "loss": 0.0173, "step": 29408 }, { "epoch": 9.05, "learning_rate": 4.746105089029229e-07, "loss": 0.0315, "step": 29409 }, { "epoch": 9.05, "learning_rate": 4.743071927158804e-07, "loss": 0.0241, "step": 29410 }, { "epoch": 9.05, "learning_rate": 4.7400397112774886e-07, "loss": 0.0188, "step": 29411 }, { "epoch": 9.05, "learning_rate": 4.7370084414153805e-07, "loss": 0.0176, "step": 29412 }, { "epoch": 9.05, "learning_rate": 4.7339781176025447e-07, "loss": 0.0156, "step": 29413 }, { "epoch": 9.05, "learning_rate": 4.7309487398691234e-07, "loss": 0.0355, "step": 29414 }, { "epoch": 9.05, "learning_rate": 4.7279203082451817e-07, "loss": 0.0266, "step": 29415 }, { "epoch": 9.05, "learning_rate": 4.724892822760807e-07, "loss": 0.0139, "step": 29416 }, { "epoch": 9.05, "learning_rate": 4.721866283446042e-07, "loss": 0.0332, "step": 29417 }, { "epoch": 9.05, "learning_rate": 4.718840690330961e-07, "loss": 0.023, "step": 29418 }, { "epoch": 9.05, "learning_rate": 4.715816043445609e-07, "loss": 0.0237, "step": 29419 }, { "epoch": 9.05, "learning_rate": 4.712792342820005e-07, "loss": 0.0237, "step": 29420 }, { "epoch": 9.05, "learning_rate": 4.709769588484203e-07, "loss": 0.0219, "step": 29421 }, { "epoch": 9.05, "learning_rate": 4.706747780468224e-07, "loss": 0.0179, "step": 29422 }, { "epoch": 9.05, "learning_rate": 4.703726918802032e-07, "loss": 0.0278, "step": 29423 }, { "epoch": 9.05, "learning_rate": 4.700707003515681e-07, "loss": 0.0195, "step": 29424 }, { "epoch": 9.05, "learning_rate": 4.6976880346391365e-07, "loss": 0.0329, "step": 29425 }, { "epoch": 9.05, "learning_rate": 4.694670012202385e-07, "loss": 0.022, "step": 29426 }, { "epoch": 9.05, "learning_rate": 4.691652936235402e-07, "loss": 0.012, "step": 29427 }, { "epoch": 9.05, "learning_rate": 4.688636806768121e-07, "loss": 0.0256, "step": 29428 }, { "epoch": 9.05, "learning_rate": 4.6856216238305494e-07, "loss": 0.0268, "step": 29429 }, { "epoch": 9.06, "learning_rate": 4.6826073874525756e-07, "loss": 0.0197, "step": 29430 }, { "epoch": 9.06, "learning_rate": 4.679594097664164e-07, "loss": 0.0142, "step": 29431 }, { "epoch": 9.06, "learning_rate": 4.676581754495235e-07, "loss": 0.0244, "step": 29432 }, { "epoch": 9.06, "learning_rate": 4.6735703579757095e-07, "loss": 0.0373, "step": 29433 }, { "epoch": 9.06, "learning_rate": 4.6705599081354747e-07, "loss": 0.0283, "step": 29434 }, { "epoch": 9.06, "learning_rate": 4.6675504050044283e-07, "loss": 0.0297, "step": 29435 }, { "epoch": 9.06, "learning_rate": 4.6645418486124916e-07, "loss": 0.0355, "step": 29436 }, { "epoch": 9.06, "learning_rate": 4.661534238989518e-07, "loss": 0.033, "step": 29437 }, { "epoch": 9.06, "learning_rate": 4.65852757616535e-07, "loss": 0.033, "step": 29438 }, { "epoch": 9.06, "learning_rate": 4.655521860169898e-07, "loss": 0.0299, "step": 29439 }, { "epoch": 9.06, "learning_rate": 4.65251709103296e-07, "loss": 0.025, "step": 29440 }, { "epoch": 9.06, "learning_rate": 4.6495132687844224e-07, "loss": 0.0088, "step": 29441 }, { "epoch": 9.06, "learning_rate": 4.646510393454096e-07, "loss": 0.0271, "step": 29442 }, { "epoch": 9.06, "learning_rate": 4.6435084650718e-07, "loss": 0.0275, "step": 29443 }, { "epoch": 9.06, "learning_rate": 4.6405074836673445e-07, "loss": 0.0187, "step": 29444 }, { "epoch": 9.06, "learning_rate": 4.6375074492705173e-07, "loss": 0.0161, "step": 29445 }, { "epoch": 9.06, "learning_rate": 4.634508361911161e-07, "loss": 0.0075, "step": 29446 }, { "epoch": 9.06, "learning_rate": 4.631510221619018e-07, "loss": 0.0192, "step": 29447 }, { "epoch": 9.06, "learning_rate": 4.6285130284238645e-07, "loss": 0.0225, "step": 29448 }, { "epoch": 9.06, "learning_rate": 4.625516782355477e-07, "loss": 0.0147, "step": 29449 }, { "epoch": 9.06, "learning_rate": 4.6225214834436094e-07, "loss": 0.0143, "step": 29450 }, { "epoch": 9.06, "learning_rate": 4.6195271317180044e-07, "loss": 0.0303, "step": 29451 }, { "epoch": 9.06, "learning_rate": 4.6165337272084056e-07, "loss": 0.0132, "step": 29452 }, { "epoch": 9.06, "learning_rate": 4.6135412699445323e-07, "loss": 0.0179, "step": 29453 }, { "epoch": 9.06, "learning_rate": 4.610549759956107e-07, "loss": 0.004, "step": 29454 }, { "epoch": 9.06, "learning_rate": 4.607559197272815e-07, "loss": 0.0362, "step": 29455 }, { "epoch": 9.06, "learning_rate": 4.604569581924412e-07, "loss": 0.0519, "step": 29456 }, { "epoch": 9.06, "learning_rate": 4.60158091394054e-07, "loss": 0.0264, "step": 29457 }, { "epoch": 9.06, "learning_rate": 4.5985931933508757e-07, "loss": 0.023, "step": 29458 }, { "epoch": 9.06, "learning_rate": 4.595606420185117e-07, "loss": 0.0163, "step": 29459 }, { "epoch": 9.06, "learning_rate": 4.5926205944729073e-07, "loss": 0.0199, "step": 29460 }, { "epoch": 9.06, "learning_rate": 4.589635716243912e-07, "loss": 0.0305, "step": 29461 }, { "epoch": 9.06, "learning_rate": 4.586651785527751e-07, "loss": 0.0377, "step": 29462 }, { "epoch": 9.07, "learning_rate": 4.583668802354102e-07, "loss": 0.0269, "step": 29463 }, { "epoch": 9.07, "learning_rate": 4.5806867667525513e-07, "loss": 0.0247, "step": 29464 }, { "epoch": 9.07, "learning_rate": 4.5777056787526974e-07, "loss": 0.0349, "step": 29465 }, { "epoch": 9.07, "learning_rate": 4.574725538384195e-07, "loss": 0.0257, "step": 29466 }, { "epoch": 9.07, "learning_rate": 4.5717463456766087e-07, "loss": 0.0232, "step": 29467 }, { "epoch": 9.07, "learning_rate": 4.5687681006595265e-07, "loss": 0.023, "step": 29468 }, { "epoch": 9.07, "learning_rate": 4.565790803362535e-07, "loss": 0.0187, "step": 29469 }, { "epoch": 9.07, "learning_rate": 4.5628144538152006e-07, "loss": 0.0271, "step": 29470 }, { "epoch": 9.07, "learning_rate": 4.559839052047066e-07, "loss": 0.0102, "step": 29471 }, { "epoch": 9.07, "learning_rate": 4.556864598087685e-07, "loss": 0.0195, "step": 29472 }, { "epoch": 9.07, "learning_rate": 4.5538910919666115e-07, "loss": 0.0162, "step": 29473 }, { "epoch": 9.07, "learning_rate": 4.550918533713367e-07, "loss": 0.0233, "step": 29474 }, { "epoch": 9.07, "learning_rate": 4.54794692335746e-07, "loss": 0.0149, "step": 29475 }, { "epoch": 9.07, "learning_rate": 4.5449762609284134e-07, "loss": 0.022, "step": 29476 }, { "epoch": 9.07, "learning_rate": 4.542006546455735e-07, "loss": 0.029, "step": 29477 }, { "epoch": 9.07, "learning_rate": 4.539037779968902e-07, "loss": 0.0191, "step": 29478 }, { "epoch": 9.07, "learning_rate": 4.5360699614974024e-07, "loss": 0.021, "step": 29479 }, { "epoch": 9.07, "learning_rate": 4.533103091070723e-07, "loss": 0.0161, "step": 29480 }, { "epoch": 9.07, "learning_rate": 4.530137168718296e-07, "loss": 0.0248, "step": 29481 }, { "epoch": 9.07, "learning_rate": 4.5271721944695977e-07, "loss": 0.0306, "step": 29482 }, { "epoch": 9.07, "learning_rate": 4.524208168354072e-07, "loss": 0.02, "step": 29483 }, { "epoch": 9.07, "learning_rate": 4.521245090401172e-07, "loss": 0.0334, "step": 29484 }, { "epoch": 9.07, "learning_rate": 4.5182829606402743e-07, "loss": 0.0204, "step": 29485 }, { "epoch": 9.07, "learning_rate": 4.515321779100834e-07, "loss": 0.0155, "step": 29486 }, { "epoch": 9.07, "learning_rate": 4.5123615458122605e-07, "loss": 0.017, "step": 29487 }, { "epoch": 9.07, "learning_rate": 4.50940226080393e-07, "loss": 0.0244, "step": 29488 }, { "epoch": 9.07, "learning_rate": 4.506443924105253e-07, "loss": 0.0379, "step": 29489 }, { "epoch": 9.07, "learning_rate": 4.5034865357455827e-07, "loss": 0.0372, "step": 29490 }, { "epoch": 9.07, "learning_rate": 4.500530095754319e-07, "loss": 0.0337, "step": 29491 }, { "epoch": 9.07, "learning_rate": 4.497574604160793e-07, "loss": 0.014, "step": 29492 }, { "epoch": 9.07, "learning_rate": 4.4946200609943704e-07, "loss": 0.0234, "step": 29493 }, { "epoch": 9.07, "learning_rate": 4.4916664662843947e-07, "loss": 0.0382, "step": 29494 }, { "epoch": 9.08, "learning_rate": 4.4887138200601976e-07, "loss": 0.0182, "step": 29495 }, { "epoch": 9.08, "learning_rate": 4.4857621223510896e-07, "loss": 0.0113, "step": 29496 }, { "epoch": 9.08, "learning_rate": 4.482811373186402e-07, "loss": 0.0135, "step": 29497 }, { "epoch": 9.08, "learning_rate": 4.479861572595423e-07, "loss": 0.0263, "step": 29498 }, { "epoch": 9.08, "learning_rate": 4.4769127206074514e-07, "loss": 0.0194, "step": 29499 }, { "epoch": 9.08, "learning_rate": 4.4739648172517634e-07, "loss": 0.0236, "step": 29500 }, { "epoch": 9.08, "learning_rate": 4.4710178625576694e-07, "loss": 0.0275, "step": 29501 }, { "epoch": 9.08, "learning_rate": 4.4680718565543904e-07, "loss": 0.0192, "step": 29502 }, { "epoch": 9.08, "learning_rate": 4.465126799271213e-07, "loss": 0.0298, "step": 29503 }, { "epoch": 9.08, "learning_rate": 4.4621826907373713e-07, "loss": 0.0276, "step": 29504 }, { "epoch": 9.08, "learning_rate": 4.4592395309820956e-07, "loss": 0.0217, "step": 29505 }, { "epoch": 9.08, "learning_rate": 4.4562973200346413e-07, "loss": 0.0277, "step": 29506 }, { "epoch": 9.08, "learning_rate": 4.453356057924185e-07, "loss": 0.0202, "step": 29507 }, { "epoch": 9.08, "learning_rate": 4.450415744679992e-07, "loss": 0.0225, "step": 29508 }, { "epoch": 9.08, "learning_rate": 4.447476380331228e-07, "loss": 0.0187, "step": 29509 }, { "epoch": 9.08, "learning_rate": 4.4445379649070587e-07, "loss": 0.0252, "step": 29510 }, { "epoch": 9.08, "learning_rate": 4.4416004984367154e-07, "loss": 0.0165, "step": 29511 }, { "epoch": 9.08, "learning_rate": 4.4386639809493426e-07, "loss": 0.0371, "step": 29512 }, { "epoch": 9.08, "learning_rate": 4.4357284124741163e-07, "loss": 0.0227, "step": 29513 }, { "epoch": 9.08, "learning_rate": 4.4327937930401797e-07, "loss": 0.0219, "step": 29514 }, { "epoch": 9.08, "learning_rate": 4.4298601226766766e-07, "loss": 0.0213, "step": 29515 }, { "epoch": 9.08, "learning_rate": 4.4269274014127397e-07, "loss": 0.0309, "step": 29516 }, { "epoch": 9.08, "learning_rate": 4.423995629277489e-07, "loss": 0.0178, "step": 29517 }, { "epoch": 9.08, "learning_rate": 4.421064806300068e-07, "loss": 0.0325, "step": 29518 }, { "epoch": 9.08, "learning_rate": 4.4181349325095546e-07, "loss": 0.0256, "step": 29519 }, { "epoch": 9.08, "learning_rate": 4.4152060079350354e-07, "loss": 0.0241, "step": 29520 }, { "epoch": 9.08, "learning_rate": 4.412278032605621e-07, "loss": 0.0268, "step": 29521 }, { "epoch": 9.08, "learning_rate": 4.4093510065503884e-07, "loss": 0.0234, "step": 29522 }, { "epoch": 9.08, "learning_rate": 4.406424929798403e-07, "loss": 0.0289, "step": 29523 }, { "epoch": 9.08, "learning_rate": 4.403499802378708e-07, "loss": 0.0171, "step": 29524 }, { "epoch": 9.08, "learning_rate": 4.4005756243203803e-07, "loss": 0.0229, "step": 29525 }, { "epoch": 9.08, "learning_rate": 4.39765239565243e-07, "loss": 0.0353, "step": 29526 }, { "epoch": 9.08, "learning_rate": 4.394730116403889e-07, "loss": 0.0174, "step": 29527 }, { "epoch": 9.09, "learning_rate": 4.391808786603813e-07, "loss": 0.0336, "step": 29528 }, { "epoch": 9.09, "learning_rate": 4.3888884062812e-07, "loss": 0.0108, "step": 29529 }, { "epoch": 9.09, "learning_rate": 4.3859689754650157e-07, "loss": 0.0155, "step": 29530 }, { "epoch": 9.09, "learning_rate": 4.3830504941842935e-07, "loss": 0.0303, "step": 29531 }, { "epoch": 9.09, "learning_rate": 4.3801329624679976e-07, "loss": 0.0149, "step": 29532 }, { "epoch": 9.09, "learning_rate": 4.377216380345106e-07, "loss": 0.021, "step": 29533 }, { "epoch": 9.09, "learning_rate": 4.374300747844595e-07, "loss": 0.0353, "step": 29534 }, { "epoch": 9.09, "learning_rate": 4.371386064995409e-07, "loss": 0.0196, "step": 29535 }, { "epoch": 9.09, "learning_rate": 4.368472331826479e-07, "loss": 0.0259, "step": 29536 }, { "epoch": 9.09, "learning_rate": 4.36555954836676e-07, "loss": 0.0256, "step": 29537 }, { "epoch": 9.09, "learning_rate": 4.362647714645174e-07, "loss": 0.0114, "step": 29538 }, { "epoch": 9.09, "learning_rate": 4.359736830690653e-07, "loss": 0.0259, "step": 29539 }, { "epoch": 9.09, "learning_rate": 4.3568268965320736e-07, "loss": 0.0224, "step": 29540 }, { "epoch": 9.09, "learning_rate": 4.3539179121983466e-07, "loss": 0.0435, "step": 29541 }, { "epoch": 9.09, "learning_rate": 4.351009877718382e-07, "loss": 0.0128, "step": 29542 }, { "epoch": 9.09, "learning_rate": 4.348102793121034e-07, "loss": 0.033, "step": 29543 }, { "epoch": 9.09, "learning_rate": 4.3451966584351803e-07, "loss": 0.0226, "step": 29544 }, { "epoch": 9.09, "learning_rate": 4.3422914736896635e-07, "loss": 0.0283, "step": 29545 }, { "epoch": 9.09, "learning_rate": 4.339387238913395e-07, "loss": 0.0339, "step": 29546 }, { "epoch": 9.09, "learning_rate": 4.33648395413514e-07, "loss": 0.0213, "step": 29547 }, { "epoch": 9.09, "learning_rate": 4.333581619383775e-07, "loss": 0.0312, "step": 29548 }, { "epoch": 9.09, "learning_rate": 4.3306802346881116e-07, "loss": 0.0193, "step": 29549 }, { "epoch": 9.09, "learning_rate": 4.32777980007697e-07, "loss": 0.0244, "step": 29550 }, { "epoch": 9.09, "learning_rate": 4.3248803155791496e-07, "loss": 0.0275, "step": 29551 }, { "epoch": 9.09, "learning_rate": 4.3219817812234277e-07, "loss": 0.0129, "step": 29552 }, { "epoch": 9.09, "learning_rate": 4.3190841970386367e-07, "loss": 0.0282, "step": 29553 }, { "epoch": 9.09, "learning_rate": 4.3161875630535086e-07, "loss": 0.0244, "step": 29554 }, { "epoch": 9.09, "learning_rate": 4.313291879296799e-07, "loss": 0.0136, "step": 29555 }, { "epoch": 9.09, "learning_rate": 4.310397145797329e-07, "loss": 0.0373, "step": 29556 }, { "epoch": 9.09, "learning_rate": 4.3075033625837757e-07, "loss": 0.0159, "step": 29557 }, { "epoch": 9.09, "learning_rate": 4.304610529684916e-07, "loss": 0.017, "step": 29558 }, { "epoch": 9.09, "learning_rate": 4.3017186471294713e-07, "loss": 0.0268, "step": 29559 }, { "epoch": 9.1, "learning_rate": 4.2988277149461634e-07, "loss": 0.0076, "step": 29560 }, { "epoch": 9.1, "learning_rate": 4.295937733163691e-07, "loss": 0.0271, "step": 29561 }, { "epoch": 9.1, "learning_rate": 4.2930487018107425e-07, "loss": 0.0315, "step": 29562 }, { "epoch": 9.1, "learning_rate": 4.2901606209160616e-07, "loss": 0.0251, "step": 29563 }, { "epoch": 9.1, "learning_rate": 4.2872734905082814e-07, "loss": 0.019, "step": 29564 }, { "epoch": 9.1, "learning_rate": 4.284387310616067e-07, "loss": 0.0204, "step": 29565 }, { "epoch": 9.1, "learning_rate": 4.2815020812681184e-07, "loss": 0.0147, "step": 29566 }, { "epoch": 9.1, "learning_rate": 4.278617802493068e-07, "loss": 0.0228, "step": 29567 }, { "epoch": 9.1, "learning_rate": 4.2757344743195596e-07, "loss": 0.0184, "step": 29568 }, { "epoch": 9.1, "learning_rate": 4.2728520967762365e-07, "loss": 0.0279, "step": 29569 }, { "epoch": 9.1, "learning_rate": 4.269970669891721e-07, "loss": 0.0232, "step": 29570 }, { "epoch": 9.1, "learning_rate": 4.267090193694612e-07, "loss": 0.0231, "step": 29571 }, { "epoch": 9.1, "learning_rate": 4.26421066821352e-07, "loss": 0.0252, "step": 29572 }, { "epoch": 9.1, "learning_rate": 4.261332093477066e-07, "loss": 0.0183, "step": 29573 }, { "epoch": 9.1, "learning_rate": 4.2584544695138287e-07, "loss": 0.032, "step": 29574 }, { "epoch": 9.1, "learning_rate": 4.2555777963523506e-07, "loss": 0.0236, "step": 29575 }, { "epoch": 9.1, "learning_rate": 4.2527020740212424e-07, "loss": 0.0292, "step": 29576 }, { "epoch": 9.1, "learning_rate": 4.2498273025490365e-07, "loss": 0.0364, "step": 29577 }, { "epoch": 9.1, "learning_rate": 4.2469534819643e-07, "loss": 0.0258, "step": 29578 }, { "epoch": 9.1, "learning_rate": 4.2440806122955644e-07, "loss": 0.0214, "step": 29579 }, { "epoch": 9.1, "learning_rate": 4.2412086935713523e-07, "loss": 0.0208, "step": 29580 }, { "epoch": 9.1, "learning_rate": 4.238337725820196e-07, "loss": 0.0338, "step": 29581 }, { "epoch": 9.1, "learning_rate": 4.235467709070584e-07, "loss": 0.0157, "step": 29582 }, { "epoch": 9.1, "learning_rate": 4.23259864335106e-07, "loss": 0.0195, "step": 29583 }, { "epoch": 9.1, "learning_rate": 4.2297305286901016e-07, "loss": 0.0234, "step": 29584 }, { "epoch": 9.1, "learning_rate": 4.226863365116163e-07, "loss": 0.0202, "step": 29585 }, { "epoch": 9.1, "learning_rate": 4.2239971526577437e-07, "loss": 0.0483, "step": 29586 }, { "epoch": 9.1, "learning_rate": 4.2211318913433e-07, "loss": 0.0201, "step": 29587 }, { "epoch": 9.1, "learning_rate": 4.218267581201296e-07, "loss": 0.0277, "step": 29588 }, { "epoch": 9.1, "learning_rate": 4.2154042222601666e-07, "loss": 0.0249, "step": 29589 }, { "epoch": 9.1, "learning_rate": 4.212541814548343e-07, "loss": 0.0248, "step": 29590 }, { "epoch": 9.1, "learning_rate": 4.2096803580942813e-07, "loss": 0.0291, "step": 29591 }, { "epoch": 9.1, "learning_rate": 4.206819852926347e-07, "loss": 0.0127, "step": 29592 }, { "epoch": 9.11, "learning_rate": 4.2039602990729953e-07, "loss": 0.0288, "step": 29593 }, { "epoch": 9.11, "learning_rate": 4.201101696562604e-07, "loss": 0.0119, "step": 29594 }, { "epoch": 9.11, "learning_rate": 4.198244045423561e-07, "loss": 0.0172, "step": 29595 }, { "epoch": 9.11, "learning_rate": 4.195387345684254e-07, "loss": 0.0366, "step": 29596 }, { "epoch": 9.11, "learning_rate": 4.192531597373051e-07, "loss": 0.0173, "step": 29597 }, { "epoch": 9.11, "learning_rate": 4.189676800518294e-07, "loss": 0.0169, "step": 29598 }, { "epoch": 9.11, "learning_rate": 4.186822955148362e-07, "loss": 0.0196, "step": 29599 }, { "epoch": 9.11, "learning_rate": 4.1839700612915535e-07, "loss": 0.0157, "step": 29600 }, { "epoch": 9.11, "learning_rate": 4.1811181189762684e-07, "loss": 0.0174, "step": 29601 }, { "epoch": 9.11, "learning_rate": 4.1782671282307617e-07, "loss": 0.0279, "step": 29602 }, { "epoch": 9.11, "learning_rate": 4.1754170890833777e-07, "loss": 0.0181, "step": 29603 }, { "epoch": 9.11, "learning_rate": 4.172568001562427e-07, "loss": 0.0292, "step": 29604 }, { "epoch": 9.11, "learning_rate": 4.1697198656961866e-07, "loss": 0.0436, "step": 29605 }, { "epoch": 9.11, "learning_rate": 4.166872681512957e-07, "loss": 0.0271, "step": 29606 }, { "epoch": 9.11, "learning_rate": 4.164026449040981e-07, "loss": 0.0157, "step": 29607 }, { "epoch": 9.11, "learning_rate": 4.161181168308581e-07, "loss": 0.0122, "step": 29608 }, { "epoch": 9.11, "learning_rate": 4.158336839343957e-07, "loss": 0.0356, "step": 29609 }, { "epoch": 9.11, "learning_rate": 4.155493462175375e-07, "loss": 0.0245, "step": 29610 }, { "epoch": 9.11, "learning_rate": 4.152651036831079e-07, "loss": 0.033, "step": 29611 }, { "epoch": 9.11, "learning_rate": 4.149809563339302e-07, "loss": 0.0249, "step": 29612 }, { "epoch": 9.11, "learning_rate": 4.146969041728255e-07, "loss": 0.0345, "step": 29613 }, { "epoch": 9.11, "learning_rate": 4.1441294720261373e-07, "loss": 0.0133, "step": 29614 }, { "epoch": 9.11, "learning_rate": 4.1412908542611706e-07, "loss": 0.0214, "step": 29615 }, { "epoch": 9.11, "learning_rate": 4.138453188461522e-07, "loss": 0.0216, "step": 29616 }, { "epoch": 9.11, "learning_rate": 4.135616474655368e-07, "loss": 0.0222, "step": 29617 }, { "epoch": 9.11, "learning_rate": 4.1327807128709207e-07, "loss": 0.0156, "step": 29618 }, { "epoch": 9.11, "learning_rate": 4.1299459031363123e-07, "loss": 0.0178, "step": 29619 }, { "epoch": 9.11, "learning_rate": 4.1271120454796755e-07, "loss": 0.0125, "step": 29620 }, { "epoch": 9.11, "learning_rate": 4.124279139929188e-07, "loss": 0.0241, "step": 29621 }, { "epoch": 9.11, "learning_rate": 4.1214471865129724e-07, "loss": 0.031, "step": 29622 }, { "epoch": 9.11, "learning_rate": 4.11861618525915e-07, "loss": 0.0184, "step": 29623 }, { "epoch": 9.11, "learning_rate": 4.115786136195832e-07, "loss": 0.0212, "step": 29624 }, { "epoch": 9.12, "learning_rate": 4.1129570393511285e-07, "loss": 0.016, "step": 29625 }, { "epoch": 9.12, "learning_rate": 4.1101288947531404e-07, "loss": 0.0107, "step": 29626 }, { "epoch": 9.12, "learning_rate": 4.107301702429922e-07, "loss": 0.018, "step": 29627 }, { "epoch": 9.12, "learning_rate": 4.104475462409596e-07, "loss": 0.0207, "step": 29628 }, { "epoch": 9.12, "learning_rate": 4.1016501747202176e-07, "loss": 0.0126, "step": 29629 }, { "epoch": 9.12, "learning_rate": 4.0988258393898193e-07, "loss": 0.0254, "step": 29630 }, { "epoch": 9.12, "learning_rate": 4.0960024564464687e-07, "loss": 0.0255, "step": 29631 }, { "epoch": 9.12, "learning_rate": 4.093180025918209e-07, "loss": 0.0376, "step": 29632 }, { "epoch": 9.12, "learning_rate": 4.090358547833051e-07, "loss": 0.0288, "step": 29633 }, { "epoch": 9.12, "learning_rate": 4.0875380222190286e-07, "loss": 0.0204, "step": 29634 }, { "epoch": 9.12, "learning_rate": 4.0847184491041523e-07, "loss": 0.0184, "step": 29635 }, { "epoch": 9.12, "learning_rate": 4.0818998285164223e-07, "loss": 0.0187, "step": 29636 }, { "epoch": 9.12, "learning_rate": 4.079082160483816e-07, "loss": 0.0202, "step": 29637 }, { "epoch": 9.12, "learning_rate": 4.076265445034333e-07, "loss": 0.025, "step": 29638 }, { "epoch": 9.12, "learning_rate": 4.073449682195962e-07, "loss": 0.0182, "step": 29639 }, { "epoch": 9.12, "learning_rate": 4.070634871996615e-07, "loss": 0.0373, "step": 29640 }, { "epoch": 9.12, "learning_rate": 4.067821014464279e-07, "loss": 0.0228, "step": 29641 }, { "epoch": 9.12, "learning_rate": 4.0650081096269e-07, "loss": 0.0285, "step": 29642 }, { "epoch": 9.12, "learning_rate": 4.06219615751241e-07, "loss": 0.0185, "step": 29643 }, { "epoch": 9.12, "learning_rate": 4.059385158148732e-07, "loss": 0.0219, "step": 29644 }, { "epoch": 9.12, "learning_rate": 4.0565751115637654e-07, "loss": 0.0272, "step": 29645 }, { "epoch": 9.12, "learning_rate": 4.053766017785454e-07, "loss": 0.0229, "step": 29646 }, { "epoch": 9.12, "learning_rate": 4.0509578768416435e-07, "loss": 0.0198, "step": 29647 }, { "epoch": 9.12, "learning_rate": 4.0481506887602773e-07, "loss": 0.0222, "step": 29648 }, { "epoch": 9.12, "learning_rate": 4.0453444535692e-07, "loss": 0.0475, "step": 29649 }, { "epoch": 9.12, "learning_rate": 4.042539171296278e-07, "loss": 0.0212, "step": 29650 }, { "epoch": 9.12, "learning_rate": 4.0397348419693894e-07, "loss": 0.0173, "step": 29651 }, { "epoch": 9.12, "learning_rate": 4.036931465616356e-07, "loss": 0.0308, "step": 29652 }, { "epoch": 9.12, "learning_rate": 4.034129042265067e-07, "loss": 0.0188, "step": 29653 }, { "epoch": 9.12, "learning_rate": 4.0313275719433e-07, "loss": 0.0109, "step": 29654 }, { "epoch": 9.12, "learning_rate": 4.028527054678888e-07, "loss": 0.0188, "step": 29655 }, { "epoch": 9.12, "learning_rate": 4.025727490499687e-07, "loss": 0.0192, "step": 29656 }, { "epoch": 9.12, "learning_rate": 4.02292887943343e-07, "loss": 0.0261, "step": 29657 }, { "epoch": 9.13, "learning_rate": 4.0201312215079614e-07, "loss": 0.018, "step": 29658 }, { "epoch": 9.13, "learning_rate": 4.017334516751037e-07, "loss": 0.0275, "step": 29659 }, { "epoch": 9.13, "learning_rate": 4.0145387651904455e-07, "loss": 0.0242, "step": 29660 }, { "epoch": 9.13, "learning_rate": 4.011743966853954e-07, "loss": 0.0237, "step": 29661 }, { "epoch": 9.13, "learning_rate": 4.008950121769295e-07, "loss": 0.0187, "step": 29662 }, { "epoch": 9.13, "learning_rate": 4.006157229964247e-07, "loss": 0.0246, "step": 29663 }, { "epoch": 9.13, "learning_rate": 4.00336529146651e-07, "loss": 0.0246, "step": 29664 }, { "epoch": 9.13, "learning_rate": 4.0005743063038393e-07, "loss": 0.0389, "step": 29665 }, { "epoch": 9.13, "learning_rate": 3.9977842745039464e-07, "loss": 0.0489, "step": 29666 }, { "epoch": 9.13, "learning_rate": 3.994995196094531e-07, "loss": 0.0272, "step": 29667 }, { "epoch": 9.13, "learning_rate": 3.9922070711032934e-07, "loss": 0.0136, "step": 29668 }, { "epoch": 9.13, "learning_rate": 3.9894198995579114e-07, "loss": 0.0233, "step": 29669 }, { "epoch": 9.13, "learning_rate": 3.9866336814860964e-07, "loss": 0.0159, "step": 29670 }, { "epoch": 9.13, "learning_rate": 3.983848416915481e-07, "loss": 0.0333, "step": 29671 }, { "epoch": 9.13, "learning_rate": 3.981064105873744e-07, "loss": 0.0216, "step": 29672 }, { "epoch": 9.13, "learning_rate": 3.97828074838853e-07, "loss": 0.0327, "step": 29673 }, { "epoch": 9.13, "learning_rate": 3.975498344487494e-07, "loss": 0.0287, "step": 29674 }, { "epoch": 9.13, "learning_rate": 3.9727168941982584e-07, "loss": 0.0287, "step": 29675 }, { "epoch": 9.13, "learning_rate": 3.969936397548446e-07, "loss": 0.0197, "step": 29676 }, { "epoch": 9.13, "learning_rate": 3.967156854565668e-07, "loss": 0.0151, "step": 29677 }, { "epoch": 9.13, "learning_rate": 3.964378265277524e-07, "loss": 0.0308, "step": 29678 }, { "epoch": 9.13, "learning_rate": 3.961600629711615e-07, "loss": 0.0226, "step": 29679 }, { "epoch": 9.13, "learning_rate": 3.9588239478955294e-07, "loss": 0.0134, "step": 29680 }, { "epoch": 9.13, "learning_rate": 3.956048219856834e-07, "loss": 0.0212, "step": 29681 }, { "epoch": 9.13, "learning_rate": 3.9532734456230847e-07, "loss": 0.0258, "step": 29682 }, { "epoch": 9.13, "learning_rate": 3.9504996252218597e-07, "loss": 0.0275, "step": 29683 }, { "epoch": 9.13, "learning_rate": 3.947726758680692e-07, "loss": 0.0248, "step": 29684 }, { "epoch": 9.13, "learning_rate": 3.944954846027127e-07, "loss": 0.023, "step": 29685 }, { "epoch": 9.13, "learning_rate": 3.9421838872886864e-07, "loss": 0.028, "step": 29686 }, { "epoch": 9.13, "learning_rate": 3.939413882492882e-07, "loss": 0.0329, "step": 29687 }, { "epoch": 9.13, "learning_rate": 3.9366448316672356e-07, "loss": 0.0278, "step": 29688 }, { "epoch": 9.13, "learning_rate": 3.933876734839237e-07, "loss": 0.0282, "step": 29689 }, { "epoch": 9.14, "learning_rate": 3.931109592036386e-07, "loss": 0.0371, "step": 29690 }, { "epoch": 9.14, "learning_rate": 3.928343403286161e-07, "loss": 0.0208, "step": 29691 }, { "epoch": 9.14, "learning_rate": 3.925578168616007e-07, "loss": 0.0268, "step": 29692 }, { "epoch": 9.14, "learning_rate": 3.9228138880534227e-07, "loss": 0.0287, "step": 29693 }, { "epoch": 9.14, "learning_rate": 3.920050561625843e-07, "loss": 0.0145, "step": 29694 }, { "epoch": 9.14, "learning_rate": 3.917288189360713e-07, "loss": 0.024, "step": 29695 }, { "epoch": 9.14, "learning_rate": 3.914526771285454e-07, "loss": 0.0183, "step": 29696 }, { "epoch": 9.14, "learning_rate": 3.9117663074275116e-07, "loss": 0.0132, "step": 29697 }, { "epoch": 9.14, "learning_rate": 3.9090067978142745e-07, "loss": 0.0231, "step": 29698 }, { "epoch": 9.14, "learning_rate": 3.906248242473154e-07, "loss": 0.0198, "step": 29699 }, { "epoch": 9.14, "learning_rate": 3.903490641431573e-07, "loss": 0.0279, "step": 29700 }, { "epoch": 9.14, "learning_rate": 3.900733994716899e-07, "loss": 0.0412, "step": 29701 }, { "epoch": 9.14, "learning_rate": 3.8979783023564756e-07, "loss": 0.0161, "step": 29702 }, { "epoch": 9.14, "learning_rate": 3.895223564377715e-07, "loss": 0.0168, "step": 29703 }, { "epoch": 9.14, "learning_rate": 3.8924697808079505e-07, "loss": 0.0322, "step": 29704 }, { "epoch": 9.14, "learning_rate": 3.889716951674549e-07, "loss": 0.0075, "step": 29705 }, { "epoch": 9.14, "learning_rate": 3.886965077004823e-07, "loss": 0.0222, "step": 29706 }, { "epoch": 9.14, "learning_rate": 3.884214156826105e-07, "loss": 0.0363, "step": 29707 }, { "epoch": 9.14, "learning_rate": 3.8814641911657516e-07, "loss": 0.0231, "step": 29708 }, { "epoch": 9.14, "learning_rate": 3.8787151800510183e-07, "loss": 0.0173, "step": 29709 }, { "epoch": 9.14, "learning_rate": 3.87596712350925e-07, "loss": 0.0204, "step": 29710 }, { "epoch": 9.14, "learning_rate": 3.873220021567714e-07, "loss": 0.0321, "step": 29711 }, { "epoch": 9.14, "learning_rate": 3.8704738742536884e-07, "loss": 0.0342, "step": 29712 }, { "epoch": 9.14, "learning_rate": 3.8677286815944627e-07, "loss": 0.0314, "step": 29713 }, { "epoch": 9.14, "learning_rate": 3.864984443617281e-07, "loss": 0.0371, "step": 29714 }, { "epoch": 9.14, "learning_rate": 3.8622411603494114e-07, "loss": 0.0296, "step": 29715 }, { "epoch": 9.14, "learning_rate": 3.859498831818087e-07, "loss": 0.018, "step": 29716 }, { "epoch": 9.14, "learning_rate": 3.8567574580505306e-07, "loss": 0.0157, "step": 29717 }, { "epoch": 9.14, "learning_rate": 3.8540170390740097e-07, "loss": 0.0254, "step": 29718 }, { "epoch": 9.14, "learning_rate": 3.8512775749156903e-07, "loss": 0.0228, "step": 29719 }, { "epoch": 9.14, "learning_rate": 3.8485390656027967e-07, "loss": 0.0354, "step": 29720 }, { "epoch": 9.14, "learning_rate": 3.8458015111625393e-07, "loss": 0.0228, "step": 29721 }, { "epoch": 9.14, "learning_rate": 3.843064911622085e-07, "loss": 0.0322, "step": 29722 }, { "epoch": 9.15, "learning_rate": 3.840329267008613e-07, "loss": 0.0135, "step": 29723 }, { "epoch": 9.15, "learning_rate": 3.8375945773492907e-07, "loss": 0.0283, "step": 29724 }, { "epoch": 9.15, "learning_rate": 3.834860842671306e-07, "loss": 0.0168, "step": 29725 }, { "epoch": 9.15, "learning_rate": 3.832128063001761e-07, "loss": 0.0195, "step": 29726 }, { "epoch": 9.15, "learning_rate": 3.8293962383677995e-07, "loss": 0.0162, "step": 29727 }, { "epoch": 9.15, "learning_rate": 3.826665368796589e-07, "loss": 0.0386, "step": 29728 }, { "epoch": 9.15, "learning_rate": 3.8239354543152307e-07, "loss": 0.018, "step": 29729 }, { "epoch": 9.15, "learning_rate": 3.8212064949508245e-07, "loss": 0.0178, "step": 29730 }, { "epoch": 9.15, "learning_rate": 3.8184784907304704e-07, "loss": 0.031, "step": 29731 }, { "epoch": 9.15, "learning_rate": 3.815751441681281e-07, "loss": 0.029, "step": 29732 }, { "epoch": 9.15, "learning_rate": 3.8130253478303234e-07, "loss": 0.0214, "step": 29733 }, { "epoch": 9.15, "learning_rate": 3.810300209204665e-07, "loss": 0.02, "step": 29734 }, { "epoch": 9.15, "learning_rate": 3.807576025831394e-07, "loss": 0.0176, "step": 29735 }, { "epoch": 9.15, "learning_rate": 3.804852797737535e-07, "loss": 0.0142, "step": 29736 }, { "epoch": 9.15, "learning_rate": 3.8021305249501314e-07, "loss": 0.0213, "step": 29737 }, { "epoch": 9.15, "learning_rate": 3.7994092074962405e-07, "loss": 0.03, "step": 29738 }, { "epoch": 9.15, "learning_rate": 3.796688845402885e-07, "loss": 0.0139, "step": 29739 }, { "epoch": 9.15, "learning_rate": 3.793969438697065e-07, "loss": 0.0115, "step": 29740 }, { "epoch": 9.15, "learning_rate": 3.791250987405792e-07, "loss": 0.0222, "step": 29741 }, { "epoch": 9.15, "learning_rate": 3.788533491556079e-07, "loss": 0.02, "step": 29742 }, { "epoch": 9.15, "learning_rate": 3.785816951174881e-07, "loss": 0.0282, "step": 29743 }, { "epoch": 9.15, "learning_rate": 3.783101366289199e-07, "loss": 0.0348, "step": 29744 }, { "epoch": 9.15, "learning_rate": 3.7803867369260007e-07, "loss": 0.026, "step": 29745 }, { "epoch": 9.15, "learning_rate": 3.7776730631122526e-07, "loss": 0.0155, "step": 29746 }, { "epoch": 9.15, "learning_rate": 3.7749603448748673e-07, "loss": 0.0113, "step": 29747 }, { "epoch": 9.15, "learning_rate": 3.772248582240823e-07, "loss": 0.0273, "step": 29748 }, { "epoch": 9.15, "learning_rate": 3.769537775237042e-07, "loss": 0.0176, "step": 29749 }, { "epoch": 9.15, "learning_rate": 3.766827923890437e-07, "loss": 0.0211, "step": 29750 }, { "epoch": 9.15, "learning_rate": 3.7641190282279196e-07, "loss": 0.0271, "step": 29751 }, { "epoch": 9.15, "learning_rate": 3.761411088276379e-07, "loss": 0.0127, "step": 29752 }, { "epoch": 9.15, "learning_rate": 3.758704104062749e-07, "loss": 0.0247, "step": 29753 }, { "epoch": 9.15, "learning_rate": 3.755998075613865e-07, "loss": 0.0536, "step": 29754 }, { "epoch": 9.16, "learning_rate": 3.7532930029566264e-07, "loss": 0.029, "step": 29755 }, { "epoch": 9.16, "learning_rate": 3.7505888861179116e-07, "loss": 0.0364, "step": 29756 }, { "epoch": 9.16, "learning_rate": 3.747885725124523e-07, "loss": 0.0202, "step": 29757 }, { "epoch": 9.16, "learning_rate": 3.7451835200033484e-07, "loss": 0.0351, "step": 29758 }, { "epoch": 9.16, "learning_rate": 3.7424822707812225e-07, "loss": 0.0135, "step": 29759 }, { "epoch": 9.16, "learning_rate": 3.7397819774849574e-07, "loss": 0.0136, "step": 29760 }, { "epoch": 9.16, "learning_rate": 3.7370826401413765e-07, "loss": 0.0372, "step": 29761 }, { "epoch": 9.16, "learning_rate": 3.734384258777268e-07, "loss": 0.0236, "step": 29762 }, { "epoch": 9.16, "learning_rate": 3.7316868334194676e-07, "loss": 0.0151, "step": 29763 }, { "epoch": 9.16, "learning_rate": 3.728990364094709e-07, "loss": 0.02, "step": 29764 }, { "epoch": 9.16, "learning_rate": 3.7262948508298255e-07, "loss": 0.0135, "step": 29765 }, { "epoch": 9.16, "learning_rate": 3.7236002936515636e-07, "loss": 0.0207, "step": 29766 }, { "epoch": 9.16, "learning_rate": 3.7209066925866786e-07, "loss": 0.0279, "step": 29767 }, { "epoch": 9.16, "learning_rate": 3.7182140476619166e-07, "loss": 0.026, "step": 29768 }, { "epoch": 9.16, "learning_rate": 3.715522358904022e-07, "loss": 0.011, "step": 29769 }, { "epoch": 9.16, "learning_rate": 3.712831626339752e-07, "loss": 0.02, "step": 29770 }, { "epoch": 9.16, "learning_rate": 3.710141849995785e-07, "loss": 0.0217, "step": 29771 }, { "epoch": 9.16, "learning_rate": 3.7074530298988443e-07, "loss": 0.0292, "step": 29772 }, { "epoch": 9.16, "learning_rate": 3.7047651660756745e-07, "loss": 0.0295, "step": 29773 }, { "epoch": 9.16, "learning_rate": 3.702078258552899e-07, "loss": 0.0232, "step": 29774 }, { "epoch": 9.16, "learning_rate": 3.6993923073572525e-07, "loss": 0.0133, "step": 29775 }, { "epoch": 9.16, "learning_rate": 3.696707312515391e-07, "loss": 0.026, "step": 29776 }, { "epoch": 9.16, "learning_rate": 3.6940232740539927e-07, "loss": 0.0231, "step": 29777 }, { "epoch": 9.16, "learning_rate": 3.691340191999693e-07, "loss": 0.0256, "step": 29778 }, { "epoch": 9.16, "learning_rate": 3.688658066379136e-07, "loss": 0.0185, "step": 29779 }, { "epoch": 9.16, "learning_rate": 3.68597689721899e-07, "loss": 0.0206, "step": 29780 }, { "epoch": 9.16, "learning_rate": 3.683296684545845e-07, "loss": 0.0212, "step": 29781 }, { "epoch": 9.16, "learning_rate": 3.680617428386324e-07, "loss": 0.0265, "step": 29782 }, { "epoch": 9.16, "learning_rate": 3.67793912876705e-07, "loss": 0.0225, "step": 29783 }, { "epoch": 9.16, "learning_rate": 3.675261785714612e-07, "loss": 0.0245, "step": 29784 }, { "epoch": 9.16, "learning_rate": 3.6725853992556017e-07, "loss": 0.0227, "step": 29785 }, { "epoch": 9.16, "learning_rate": 3.6699099694166076e-07, "loss": 0.0183, "step": 29786 }, { "epoch": 9.16, "learning_rate": 3.667235496224175e-07, "loss": 0.0307, "step": 29787 }, { "epoch": 9.17, "learning_rate": 3.6645619797048835e-07, "loss": 0.0195, "step": 29788 }, { "epoch": 9.17, "learning_rate": 3.661889419885256e-07, "loss": 0.0145, "step": 29789 }, { "epoch": 9.17, "learning_rate": 3.659217816791882e-07, "loss": 0.0191, "step": 29790 }, { "epoch": 9.17, "learning_rate": 3.656547170451263e-07, "loss": 0.0144, "step": 29791 }, { "epoch": 9.17, "learning_rate": 3.6538774808899e-07, "loss": 0.0323, "step": 29792 }, { "epoch": 9.17, "learning_rate": 3.6512087481343493e-07, "loss": 0.0191, "step": 29793 }, { "epoch": 9.17, "learning_rate": 3.64854097221109e-07, "loss": 0.0191, "step": 29794 }, { "epoch": 9.17, "learning_rate": 3.6458741531466226e-07, "loss": 0.0172, "step": 29795 }, { "epoch": 9.17, "learning_rate": 3.643208290967415e-07, "loss": 0.0179, "step": 29796 }, { "epoch": 9.17, "learning_rate": 3.6405433856999684e-07, "loss": 0.0352, "step": 29797 }, { "epoch": 9.17, "learning_rate": 3.6378794373707283e-07, "loss": 0.0202, "step": 29798 }, { "epoch": 9.17, "learning_rate": 3.63521644600614e-07, "loss": 0.0156, "step": 29799 }, { "epoch": 9.17, "learning_rate": 3.6325544116326827e-07, "loss": 0.0136, "step": 29800 }, { "epoch": 9.17, "learning_rate": 3.6298933342767796e-07, "loss": 0.0194, "step": 29801 }, { "epoch": 9.17, "learning_rate": 3.6272332139648314e-07, "loss": 0.0188, "step": 29802 }, { "epoch": 9.17, "learning_rate": 3.624574050723295e-07, "loss": 0.0163, "step": 29803 }, { "epoch": 9.17, "learning_rate": 3.6219158445785494e-07, "loss": 0.0148, "step": 29804 }, { "epoch": 9.17, "learning_rate": 3.619258595557018e-07, "loss": 0.0209, "step": 29805 }, { "epoch": 9.17, "learning_rate": 3.6166023036850683e-07, "loss": 0.0253, "step": 29806 }, { "epoch": 9.17, "learning_rate": 3.613946968989068e-07, "loss": 0.0124, "step": 29807 }, { "epoch": 9.17, "learning_rate": 3.611292591495441e-07, "loss": 0.0283, "step": 29808 }, { "epoch": 9.17, "learning_rate": 3.608639171230488e-07, "loss": 0.0254, "step": 29809 }, { "epoch": 9.17, "learning_rate": 3.605986708220599e-07, "loss": 0.0275, "step": 29810 }, { "epoch": 9.17, "learning_rate": 3.6033352024920975e-07, "loss": 0.0143, "step": 29811 }, { "epoch": 9.17, "learning_rate": 3.600684654071318e-07, "loss": 0.019, "step": 29812 }, { "epoch": 9.17, "learning_rate": 3.598035062984584e-07, "loss": 0.0241, "step": 29813 }, { "epoch": 9.17, "learning_rate": 3.595386429258207e-07, "loss": 0.0299, "step": 29814 }, { "epoch": 9.17, "learning_rate": 3.59273875291849e-07, "loss": 0.0468, "step": 29815 }, { "epoch": 9.17, "learning_rate": 3.5900920339917323e-07, "loss": 0.0236, "step": 29816 }, { "epoch": 9.17, "learning_rate": 3.5874462725042026e-07, "loss": 0.0181, "step": 29817 }, { "epoch": 9.17, "learning_rate": 3.584801468482213e-07, "loss": 0.0187, "step": 29818 }, { "epoch": 9.17, "learning_rate": 3.582157621951987e-07, "loss": 0.0245, "step": 29819 }, { "epoch": 9.18, "learning_rate": 3.579514732939804e-07, "loss": 0.0198, "step": 29820 }, { "epoch": 9.18, "learning_rate": 3.5768728014718976e-07, "loss": 0.0259, "step": 29821 }, { "epoch": 9.18, "learning_rate": 3.5742318275745147e-07, "loss": 0.0376, "step": 29822 }, { "epoch": 9.18, "learning_rate": 3.5715918112738777e-07, "loss": 0.0201, "step": 29823 }, { "epoch": 9.18, "learning_rate": 3.5689527525961887e-07, "loss": 0.0233, "step": 29824 }, { "epoch": 9.18, "learning_rate": 3.566314651567704e-07, "loss": 0.0206, "step": 29825 }, { "epoch": 9.18, "learning_rate": 3.563677508214569e-07, "loss": 0.022, "step": 29826 }, { "epoch": 9.18, "learning_rate": 3.561041322562997e-07, "loss": 0.0263, "step": 29827 }, { "epoch": 9.18, "learning_rate": 3.5584060946391664e-07, "loss": 0.0235, "step": 29828 }, { "epoch": 9.18, "learning_rate": 3.555771824469245e-07, "loss": 0.0155, "step": 29829 }, { "epoch": 9.18, "learning_rate": 3.5531385120794015e-07, "loss": 0.028, "step": 29830 }, { "epoch": 9.18, "learning_rate": 3.550506157495781e-07, "loss": 0.011, "step": 29831 }, { "epoch": 9.18, "learning_rate": 3.5478747607445183e-07, "loss": 0.0164, "step": 29832 }, { "epoch": 9.18, "learning_rate": 3.5452443218517597e-07, "loss": 0.0377, "step": 29833 }, { "epoch": 9.18, "learning_rate": 3.5426148408436054e-07, "loss": 0.0139, "step": 29834 }, { "epoch": 9.18, "learning_rate": 3.5399863177462024e-07, "loss": 0.0239, "step": 29835 }, { "epoch": 9.18, "learning_rate": 3.537358752585629e-07, "loss": 0.0193, "step": 29836 }, { "epoch": 9.18, "learning_rate": 3.5347321453879757e-07, "loss": 0.0169, "step": 29837 }, { "epoch": 9.18, "learning_rate": 3.532106496179344e-07, "loss": 0.0503, "step": 29838 }, { "epoch": 9.18, "learning_rate": 3.5294818049858017e-07, "loss": 0.027, "step": 29839 }, { "epoch": 9.18, "learning_rate": 3.5268580718334166e-07, "loss": 0.0149, "step": 29840 }, { "epoch": 9.18, "learning_rate": 3.524235296748246e-07, "loss": 0.0307, "step": 29841 }, { "epoch": 9.18, "learning_rate": 3.5216134797563363e-07, "loss": 0.0131, "step": 29842 }, { "epoch": 9.18, "learning_rate": 3.51899262088371e-07, "loss": 0.018, "step": 29843 }, { "epoch": 9.18, "learning_rate": 3.5163727201564025e-07, "loss": 0.0216, "step": 29844 }, { "epoch": 9.18, "learning_rate": 3.513753777600448e-07, "loss": 0.0263, "step": 29845 }, { "epoch": 9.18, "learning_rate": 3.51113579324186e-07, "loss": 0.0257, "step": 29846 }, { "epoch": 9.18, "learning_rate": 3.5085187671065834e-07, "loss": 0.0142, "step": 29847 }, { "epoch": 9.18, "learning_rate": 3.5059026992206645e-07, "loss": 0.0254, "step": 29848 }, { "epoch": 9.18, "learning_rate": 3.5032875896100603e-07, "loss": 0.023, "step": 29849 }, { "epoch": 9.18, "learning_rate": 3.5006734383007504e-07, "loss": 0.0237, "step": 29850 }, { "epoch": 9.18, "learning_rate": 3.4980602453186904e-07, "loss": 0.027, "step": 29851 }, { "epoch": 9.18, "learning_rate": 3.495448010689817e-07, "loss": 0.0175, "step": 29852 }, { "epoch": 9.19, "learning_rate": 3.492836734440097e-07, "loss": 0.013, "step": 29853 }, { "epoch": 9.19, "learning_rate": 3.4902264165954436e-07, "loss": 0.041, "step": 29854 }, { "epoch": 9.19, "learning_rate": 3.487617057181791e-07, "loss": 0.0152, "step": 29855 }, { "epoch": 9.19, "learning_rate": 3.485008656225053e-07, "loss": 0.0336, "step": 29856 }, { "epoch": 9.19, "learning_rate": 3.48240121375113e-07, "loss": 0.0243, "step": 29857 }, { "epoch": 9.19, "learning_rate": 3.4797947297859126e-07, "loss": 0.0311, "step": 29858 }, { "epoch": 9.19, "learning_rate": 3.4771892043552914e-07, "loss": 0.0452, "step": 29859 }, { "epoch": 9.19, "learning_rate": 3.4745846374851454e-07, "loss": 0.0123, "step": 29860 }, { "epoch": 9.19, "learning_rate": 3.4719810292013214e-07, "loss": 0.0204, "step": 29861 }, { "epoch": 9.19, "learning_rate": 3.4693783795296866e-07, "loss": 0.0213, "step": 29862 }, { "epoch": 9.19, "learning_rate": 3.4667766884961205e-07, "loss": 0.0393, "step": 29863 }, { "epoch": 9.19, "learning_rate": 3.464175956126392e-07, "loss": 0.0145, "step": 29864 }, { "epoch": 9.19, "learning_rate": 3.46157618244638e-07, "loss": 0.0147, "step": 29865 }, { "epoch": 9.19, "learning_rate": 3.458977367481897e-07, "loss": 0.0272, "step": 29866 }, { "epoch": 9.19, "learning_rate": 3.456379511258734e-07, "loss": 0.0239, "step": 29867 }, { "epoch": 9.19, "learning_rate": 3.453782613802703e-07, "loss": 0.0324, "step": 29868 }, { "epoch": 9.19, "learning_rate": 3.451186675139584e-07, "loss": 0.022, "step": 29869 }, { "epoch": 9.19, "learning_rate": 3.448591695295189e-07, "loss": 0.0184, "step": 29870 }, { "epoch": 9.19, "learning_rate": 3.4459976742952316e-07, "loss": 0.0408, "step": 29871 }, { "epoch": 9.19, "learning_rate": 3.443404612165513e-07, "loss": 0.0204, "step": 29872 }, { "epoch": 9.19, "learning_rate": 3.44081250893179e-07, "loss": 0.0177, "step": 29873 }, { "epoch": 9.19, "learning_rate": 3.438221364619776e-07, "loss": 0.0183, "step": 29874 }, { "epoch": 9.19, "learning_rate": 3.435631179255217e-07, "loss": 0.0218, "step": 29875 }, { "epoch": 9.19, "learning_rate": 3.433041952863847e-07, "loss": 0.0187, "step": 29876 }, { "epoch": 9.19, "learning_rate": 3.4304536854713577e-07, "loss": 0.0292, "step": 29877 }, { "epoch": 9.19, "learning_rate": 3.427866377103473e-07, "loss": 0.0246, "step": 29878 }, { "epoch": 9.19, "learning_rate": 3.42528002778586e-07, "loss": 0.0186, "step": 29879 }, { "epoch": 9.19, "learning_rate": 3.422694637544255e-07, "loss": 0.035, "step": 29880 }, { "epoch": 9.19, "learning_rate": 3.42011020640427e-07, "loss": 0.0289, "step": 29881 }, { "epoch": 9.19, "learning_rate": 3.4175267343916187e-07, "loss": 0.0311, "step": 29882 }, { "epoch": 9.19, "learning_rate": 3.414944221531935e-07, "loss": 0.0163, "step": 29883 }, { "epoch": 9.19, "learning_rate": 3.412362667850877e-07, "loss": 0.0491, "step": 29884 }, { "epoch": 9.2, "learning_rate": 3.409782073374068e-07, "loss": 0.0196, "step": 29885 }, { "epoch": 9.2, "learning_rate": 3.4072024381271327e-07, "loss": 0.025, "step": 29886 }, { "epoch": 9.2, "learning_rate": 3.404623762135728e-07, "loss": 0.0426, "step": 29887 }, { "epoch": 9.2, "learning_rate": 3.4020460454254113e-07, "loss": 0.0273, "step": 29888 }, { "epoch": 9.2, "learning_rate": 3.399469288021806e-07, "loss": 0.0219, "step": 29889 }, { "epoch": 9.2, "learning_rate": 3.396893489950503e-07, "loss": 0.0302, "step": 29890 }, { "epoch": 9.2, "learning_rate": 3.3943186512370827e-07, "loss": 0.0177, "step": 29891 }, { "epoch": 9.2, "learning_rate": 3.391744771907113e-07, "loss": 0.0239, "step": 29892 }, { "epoch": 9.2, "learning_rate": 3.389171851986162e-07, "loss": 0.0296, "step": 29893 }, { "epoch": 9.2, "learning_rate": 3.3865998914997645e-07, "loss": 0.0252, "step": 29894 }, { "epoch": 9.2, "learning_rate": 3.384028890473479e-07, "loss": 0.021, "step": 29895 }, { "epoch": 9.2, "learning_rate": 3.381458848932806e-07, "loss": 0.017, "step": 29896 }, { "epoch": 9.2, "learning_rate": 3.378889766903326e-07, "loss": 0.0242, "step": 29897 }, { "epoch": 9.2, "learning_rate": 3.3763216444105075e-07, "loss": 0.0284, "step": 29898 }, { "epoch": 9.2, "learning_rate": 3.3737544814798515e-07, "loss": 0.0324, "step": 29899 }, { "epoch": 9.2, "learning_rate": 3.371188278136883e-07, "loss": 0.0167, "step": 29900 }, { "epoch": 9.2, "learning_rate": 3.368623034407059e-07, "loss": 0.0221, "step": 29901 }, { "epoch": 9.2, "learning_rate": 3.3660587503158815e-07, "loss": 0.0147, "step": 29902 }, { "epoch": 9.2, "learning_rate": 3.3634954258887964e-07, "loss": 0.014, "step": 29903 }, { "epoch": 9.2, "learning_rate": 3.3609330611512614e-07, "loss": 0.0439, "step": 29904 }, { "epoch": 9.2, "learning_rate": 3.358371656128723e-07, "loss": 0.0319, "step": 29905 }, { "epoch": 9.2, "learning_rate": 3.3558112108466156e-07, "loss": 0.0244, "step": 29906 }, { "epoch": 9.2, "learning_rate": 3.3532517253303755e-07, "loss": 0.0255, "step": 29907 }, { "epoch": 9.2, "learning_rate": 3.350693199605437e-07, "loss": 0.0226, "step": 29908 }, { "epoch": 9.2, "learning_rate": 3.3481356336971694e-07, "loss": 0.0142, "step": 29909 }, { "epoch": 9.2, "learning_rate": 3.3455790276310073e-07, "loss": 0.0149, "step": 29910 }, { "epoch": 9.2, "learning_rate": 3.3430233814323087e-07, "loss": 0.0245, "step": 29911 }, { "epoch": 9.2, "learning_rate": 3.3404686951264867e-07, "loss": 0.023, "step": 29912 }, { "epoch": 9.2, "learning_rate": 3.3379149687388866e-07, "loss": 0.0229, "step": 29913 }, { "epoch": 9.2, "learning_rate": 3.3353622022948786e-07, "loss": 0.0314, "step": 29914 }, { "epoch": 9.2, "learning_rate": 3.3328103958198076e-07, "loss": 0.0257, "step": 29915 }, { "epoch": 9.2, "learning_rate": 3.33025954933901e-07, "loss": 0.0335, "step": 29916 }, { "epoch": 9.2, "learning_rate": 3.3277096628778425e-07, "loss": 0.0145, "step": 29917 }, { "epoch": 9.21, "learning_rate": 3.32516073646163e-07, "loss": 0.033, "step": 29918 }, { "epoch": 9.21, "learning_rate": 3.322612770115641e-07, "loss": 0.0285, "step": 29919 }, { "epoch": 9.21, "learning_rate": 3.3200657638652214e-07, "loss": 0.0236, "step": 29920 }, { "epoch": 9.21, "learning_rate": 3.317519717735651e-07, "loss": 0.0118, "step": 29921 }, { "epoch": 9.21, "learning_rate": 3.31497463175221e-07, "loss": 0.0137, "step": 29922 }, { "epoch": 9.21, "learning_rate": 3.3124305059401896e-07, "loss": 0.0178, "step": 29923 }, { "epoch": 9.21, "learning_rate": 3.3098873403248243e-07, "loss": 0.0356, "step": 29924 }, { "epoch": 9.21, "learning_rate": 3.307345134931417e-07, "loss": 0.0241, "step": 29925 }, { "epoch": 9.21, "learning_rate": 3.3048038897851576e-07, "loss": 0.0169, "step": 29926 }, { "epoch": 9.21, "learning_rate": 3.3022636049113267e-07, "loss": 0.0284, "step": 29927 }, { "epoch": 9.21, "learning_rate": 3.299724280335148e-07, "loss": 0.0175, "step": 29928 }, { "epoch": 9.21, "learning_rate": 3.297185916081813e-07, "loss": 0.0343, "step": 29929 }, { "epoch": 9.21, "learning_rate": 3.294648512176557e-07, "loss": 0.0245, "step": 29930 }, { "epoch": 9.21, "learning_rate": 3.2921120686445594e-07, "loss": 0.0224, "step": 29931 }, { "epoch": 9.21, "learning_rate": 3.2895765855110227e-07, "loss": 0.0341, "step": 29932 }, { "epoch": 9.21, "learning_rate": 3.2870420628011267e-07, "loss": 0.0154, "step": 29933 }, { "epoch": 9.21, "learning_rate": 3.284508500540018e-07, "loss": 0.0249, "step": 29934 }, { "epoch": 9.21, "learning_rate": 3.2819758987528984e-07, "loss": 0.0228, "step": 29935 }, { "epoch": 9.21, "learning_rate": 3.27944425746487e-07, "loss": 0.0243, "step": 29936 }, { "epoch": 9.21, "learning_rate": 3.276913576701113e-07, "loss": 0.0238, "step": 29937 }, { "epoch": 9.21, "learning_rate": 3.2743838564867513e-07, "loss": 0.0156, "step": 29938 }, { "epoch": 9.21, "learning_rate": 3.271855096846899e-07, "loss": 0.0255, "step": 29939 }, { "epoch": 9.21, "learning_rate": 3.2693272978066795e-07, "loss": 0.0331, "step": 29940 }, { "epoch": 9.21, "learning_rate": 3.266800459391173e-07, "loss": 0.018, "step": 29941 }, { "epoch": 9.21, "learning_rate": 3.264274581625515e-07, "loss": 0.0359, "step": 29942 }, { "epoch": 9.21, "learning_rate": 3.2617496645347526e-07, "loss": 0.014, "step": 29943 }, { "epoch": 9.21, "learning_rate": 3.259225708143954e-07, "loss": 0.0357, "step": 29944 }, { "epoch": 9.21, "learning_rate": 3.2567027124782215e-07, "loss": 0.0188, "step": 29945 }, { "epoch": 9.21, "learning_rate": 3.2541806775625906e-07, "loss": 0.0142, "step": 29946 }, { "epoch": 9.21, "learning_rate": 3.2516596034221084e-07, "loss": 0.0171, "step": 29947 }, { "epoch": 9.21, "learning_rate": 3.2491394900818207e-07, "loss": 0.022, "step": 29948 }, { "epoch": 9.21, "learning_rate": 3.24662033756673e-07, "loss": 0.0239, "step": 29949 }, { "epoch": 9.22, "learning_rate": 3.244102145901884e-07, "loss": 0.0209, "step": 29950 }, { "epoch": 9.22, "learning_rate": 3.2415849151122614e-07, "loss": 0.0239, "step": 29951 }, { "epoch": 9.22, "learning_rate": 3.2390686452228983e-07, "loss": 0.025, "step": 29952 }, { "epoch": 9.22, "learning_rate": 3.236553336258741e-07, "loss": 0.0214, "step": 29953 }, { "epoch": 9.22, "learning_rate": 3.234038988244792e-07, "loss": 0.0338, "step": 29954 }, { "epoch": 9.22, "learning_rate": 3.2315256012060203e-07, "loss": 0.0251, "step": 29955 }, { "epoch": 9.22, "learning_rate": 3.2290131751673835e-07, "loss": 0.0261, "step": 29956 }, { "epoch": 9.22, "learning_rate": 3.22650171015384e-07, "loss": 0.0198, "step": 29957 }, { "epoch": 9.22, "learning_rate": 3.2239912061903133e-07, "loss": 0.0305, "step": 29958 }, { "epoch": 9.22, "learning_rate": 3.221481663301751e-07, "loss": 0.0255, "step": 29959 }, { "epoch": 9.22, "learning_rate": 3.2189730815130657e-07, "loss": 0.07, "step": 29960 }, { "epoch": 9.22, "learning_rate": 3.216465460849172e-07, "loss": 0.0133, "step": 29961 }, { "epoch": 9.22, "learning_rate": 3.213958801334982e-07, "loss": 0.0323, "step": 29962 }, { "epoch": 9.22, "learning_rate": 3.2114531029953877e-07, "loss": 0.0218, "step": 29963 }, { "epoch": 9.22, "learning_rate": 3.208948365855247e-07, "loss": 0.0173, "step": 29964 }, { "epoch": 9.22, "learning_rate": 3.2064445899394723e-07, "loss": 0.0205, "step": 29965 }, { "epoch": 9.22, "learning_rate": 3.2039417752729006e-07, "loss": 0.0144, "step": 29966 }, { "epoch": 9.22, "learning_rate": 3.2014399218804e-07, "loss": 0.0249, "step": 29967 }, { "epoch": 9.22, "learning_rate": 3.1989390297868183e-07, "loss": 0.0154, "step": 29968 }, { "epoch": 9.22, "learning_rate": 3.196439099016968e-07, "loss": 0.0401, "step": 29969 }, { "epoch": 9.22, "learning_rate": 3.1939401295957186e-07, "loss": 0.0228, "step": 29970 }, { "epoch": 9.22, "learning_rate": 3.1914421215478385e-07, "loss": 0.0209, "step": 29971 }, { "epoch": 9.22, "learning_rate": 3.1889450748981756e-07, "loss": 0.0327, "step": 29972 }, { "epoch": 9.22, "learning_rate": 3.18644898967152e-07, "loss": 0.0107, "step": 29973 }, { "epoch": 9.22, "learning_rate": 3.1839538658926307e-07, "loss": 0.0278, "step": 29974 }, { "epoch": 9.22, "learning_rate": 3.18145970358632e-07, "loss": 0.0178, "step": 29975 }, { "epoch": 9.22, "learning_rate": 3.178966502777348e-07, "loss": 0.0515, "step": 29976 }, { "epoch": 9.22, "learning_rate": 3.176474263490459e-07, "loss": 0.0277, "step": 29977 }, { "epoch": 9.22, "learning_rate": 3.1739829857504235e-07, "loss": 0.0371, "step": 29978 }, { "epoch": 9.22, "learning_rate": 3.171492669581966e-07, "loss": 0.0171, "step": 29979 }, { "epoch": 9.22, "learning_rate": 3.1690033150098443e-07, "loss": 0.0425, "step": 29980 }, { "epoch": 9.22, "learning_rate": 3.166514922058739e-07, "loss": 0.0217, "step": 29981 }, { "epoch": 9.22, "learning_rate": 3.1640274907534074e-07, "loss": 0.0206, "step": 29982 }, { "epoch": 9.23, "learning_rate": 3.1615410211185196e-07, "loss": 0.0321, "step": 29983 }, { "epoch": 9.23, "learning_rate": 3.159055513178777e-07, "loss": 0.0167, "step": 29984 }, { "epoch": 9.23, "learning_rate": 3.1565709669588607e-07, "loss": 0.0185, "step": 29985 }, { "epoch": 9.23, "learning_rate": 3.1540873824834505e-07, "loss": 0.0179, "step": 29986 }, { "epoch": 9.23, "learning_rate": 3.151604759777216e-07, "loss": 0.0232, "step": 29987 }, { "epoch": 9.23, "learning_rate": 3.1491230988648035e-07, "loss": 0.0288, "step": 29988 }, { "epoch": 9.23, "learning_rate": 3.1466423997708497e-07, "loss": 0.0176, "step": 29989 }, { "epoch": 9.23, "learning_rate": 3.1441626625200007e-07, "loss": 0.0115, "step": 29990 }, { "epoch": 9.23, "learning_rate": 3.1416838871368925e-07, "loss": 0.022, "step": 29991 }, { "epoch": 9.23, "learning_rate": 3.139206073646117e-07, "loss": 0.0129, "step": 29992 }, { "epoch": 9.23, "learning_rate": 3.13672922207231e-07, "loss": 0.0223, "step": 29993 }, { "epoch": 9.23, "learning_rate": 3.13425333244004e-07, "loss": 0.0277, "step": 29994 }, { "epoch": 9.23, "learning_rate": 3.131778404773922e-07, "loss": 0.0186, "step": 29995 }, { "epoch": 9.23, "learning_rate": 3.129304439098502e-07, "loss": 0.0271, "step": 29996 }, { "epoch": 9.23, "learning_rate": 3.1268314354383823e-07, "loss": 0.0338, "step": 29997 }, { "epoch": 9.23, "learning_rate": 3.1243593938181106e-07, "loss": 0.0208, "step": 29998 }, { "epoch": 9.23, "learning_rate": 3.1218883142622114e-07, "loss": 0.0316, "step": 29999 }, { "epoch": 9.23, "learning_rate": 3.119418196795265e-07, "loss": 0.0157, "step": 30000 }, { "epoch": 9.23, "learning_rate": 3.1169490414417746e-07, "loss": 0.0085, "step": 30001 }, { "epoch": 9.23, "learning_rate": 3.114480848226276e-07, "loss": 0.041, "step": 30002 }, { "epoch": 9.23, "learning_rate": 3.1120136171732816e-07, "loss": 0.0189, "step": 30003 }, { "epoch": 9.23, "learning_rate": 3.1095473483072733e-07, "loss": 0.0175, "step": 30004 }, { "epoch": 9.23, "learning_rate": 3.1070820416527646e-07, "loss": 0.0167, "step": 30005 }, { "epoch": 9.23, "learning_rate": 3.1046176972342246e-07, "loss": 0.0186, "step": 30006 }, { "epoch": 9.23, "learning_rate": 3.102154315076134e-07, "loss": 0.0171, "step": 30007 }, { "epoch": 9.23, "learning_rate": 3.0996918952029717e-07, "loss": 0.02, "step": 30008 }, { "epoch": 9.23, "learning_rate": 3.0972304376391426e-07, "loss": 0.0385, "step": 30009 }, { "epoch": 9.23, "learning_rate": 3.094769942409148e-07, "loss": 0.0234, "step": 30010 }, { "epoch": 9.23, "learning_rate": 3.092310409537391e-07, "loss": 0.0091, "step": 30011 }, { "epoch": 9.23, "learning_rate": 3.089851839048308e-07, "loss": 0.016, "step": 30012 }, { "epoch": 9.23, "learning_rate": 3.087394230966312e-07, "loss": 0.0152, "step": 30013 }, { "epoch": 9.23, "learning_rate": 3.0849375853158057e-07, "loss": 0.016, "step": 30014 }, { "epoch": 9.24, "learning_rate": 3.0824819021211814e-07, "loss": 0.0171, "step": 30015 }, { "epoch": 9.24, "learning_rate": 3.0800271814068303e-07, "loss": 0.0212, "step": 30016 }, { "epoch": 9.24, "learning_rate": 3.0775734231971443e-07, "loss": 0.0212, "step": 30017 }, { "epoch": 9.24, "learning_rate": 3.0751206275164925e-07, "loss": 0.0367, "step": 30018 }, { "epoch": 9.24, "learning_rate": 3.0726687943891996e-07, "loss": 0.0192, "step": 30019 }, { "epoch": 9.24, "learning_rate": 3.070217923839647e-07, "loss": 0.0159, "step": 30020 }, { "epoch": 9.24, "learning_rate": 3.0677680158921585e-07, "loss": 0.0088, "step": 30021 }, { "epoch": 9.24, "learning_rate": 3.0653190705710713e-07, "loss": 0.0098, "step": 30022 }, { "epoch": 9.24, "learning_rate": 3.0628710879007095e-07, "loss": 0.0154, "step": 30023 }, { "epoch": 9.24, "learning_rate": 3.060424067905354e-07, "loss": 0.0392, "step": 30024 }, { "epoch": 9.24, "learning_rate": 3.0579780106093636e-07, "loss": 0.0218, "step": 30025 }, { "epoch": 9.24, "learning_rate": 3.055532916036974e-07, "loss": 0.0272, "step": 30026 }, { "epoch": 9.24, "learning_rate": 3.053088784212499e-07, "loss": 0.0232, "step": 30027 }, { "epoch": 9.24, "learning_rate": 3.050645615160197e-07, "loss": 0.0109, "step": 30028 }, { "epoch": 9.24, "learning_rate": 3.0482034089043377e-07, "loss": 0.0206, "step": 30029 }, { "epoch": 9.24, "learning_rate": 3.045762165469168e-07, "loss": 0.0275, "step": 30030 }, { "epoch": 9.24, "learning_rate": 3.043321884878936e-07, "loss": 0.0287, "step": 30031 }, { "epoch": 9.24, "learning_rate": 3.0408825671578767e-07, "loss": 0.0145, "step": 30032 }, { "epoch": 9.24, "learning_rate": 3.038444212330216e-07, "loss": 0.0165, "step": 30033 }, { "epoch": 9.24, "learning_rate": 3.0360068204201567e-07, "loss": 0.0272, "step": 30034 }, { "epoch": 9.24, "learning_rate": 3.0335703914519345e-07, "loss": 0.0226, "step": 30035 }, { "epoch": 9.24, "learning_rate": 3.031134925449708e-07, "loss": 0.0309, "step": 30036 }, { "epoch": 9.24, "learning_rate": 3.0287004224376804e-07, "loss": 0.0314, "step": 30037 }, { "epoch": 9.24, "learning_rate": 3.026266882440043e-07, "loss": 0.0312, "step": 30038 }, { "epoch": 9.24, "learning_rate": 3.0238343054809325e-07, "loss": 0.0186, "step": 30039 }, { "epoch": 9.24, "learning_rate": 3.021402691584541e-07, "loss": 0.02, "step": 30040 }, { "epoch": 9.24, "learning_rate": 3.01897204077497e-07, "loss": 0.0284, "step": 30041 }, { "epoch": 9.24, "learning_rate": 3.016542353076424e-07, "loss": 0.0253, "step": 30042 }, { "epoch": 9.24, "learning_rate": 3.0141136285129825e-07, "loss": 0.0104, "step": 30043 }, { "epoch": 9.24, "learning_rate": 3.01168586710876e-07, "loss": 0.0279, "step": 30044 }, { "epoch": 9.24, "learning_rate": 3.009259068887893e-07, "loss": 0.0126, "step": 30045 }, { "epoch": 9.24, "learning_rate": 3.0068332338744843e-07, "loss": 0.0296, "step": 30046 }, { "epoch": 9.24, "learning_rate": 3.0044083620926036e-07, "loss": 0.0342, "step": 30047 }, { "epoch": 9.25, "learning_rate": 3.0019844535663425e-07, "loss": 0.0245, "step": 30048 }, { "epoch": 9.25, "learning_rate": 2.9995615083197705e-07, "loss": 0.0246, "step": 30049 }, { "epoch": 9.25, "learning_rate": 2.9971395263769576e-07, "loss": 0.0157, "step": 30050 }, { "epoch": 9.25, "learning_rate": 2.994718507761929e-07, "loss": 0.0152, "step": 30051 }, { "epoch": 9.25, "learning_rate": 2.992298452498776e-07, "loss": 0.023, "step": 30052 }, { "epoch": 9.25, "learning_rate": 2.989879360611481e-07, "loss": 0.0306, "step": 30053 }, { "epoch": 9.25, "learning_rate": 2.9874612321241116e-07, "loss": 0.0184, "step": 30054 }, { "epoch": 9.25, "learning_rate": 2.9850440670606493e-07, "loss": 0.0307, "step": 30055 }, { "epoch": 9.25, "learning_rate": 2.982627865445109e-07, "loss": 0.0332, "step": 30056 }, { "epoch": 9.25, "learning_rate": 2.9802126273015044e-07, "loss": 0.042, "step": 30057 }, { "epoch": 9.25, "learning_rate": 2.977798352653782e-07, "loss": 0.0296, "step": 30058 }, { "epoch": 9.25, "learning_rate": 2.9753850415259577e-07, "loss": 0.0236, "step": 30059 }, { "epoch": 9.25, "learning_rate": 2.972972693941978e-07, "loss": 0.0293, "step": 30060 }, { "epoch": 9.25, "learning_rate": 2.97056130992579e-07, "loss": 0.0233, "step": 30061 }, { "epoch": 9.25, "learning_rate": 2.9681508895013645e-07, "loss": 0.0335, "step": 30062 }, { "epoch": 9.25, "learning_rate": 2.9657414326926146e-07, "loss": 0.0198, "step": 30063 }, { "epoch": 9.25, "learning_rate": 2.9633329395235e-07, "loss": 0.0174, "step": 30064 }, { "epoch": 9.25, "learning_rate": 2.960925410017912e-07, "loss": 0.0338, "step": 30065 }, { "epoch": 9.25, "learning_rate": 2.9585188441997646e-07, "loss": 0.0261, "step": 30066 }, { "epoch": 9.25, "learning_rate": 2.9561132420929726e-07, "loss": 0.0406, "step": 30067 }, { "epoch": 9.25, "learning_rate": 2.9537086037213945e-07, "loss": 0.0242, "step": 30068 }, { "epoch": 9.25, "learning_rate": 2.951304929108956e-07, "loss": 0.0229, "step": 30069 }, { "epoch": 9.25, "learning_rate": 2.9489022182794927e-07, "loss": 0.0218, "step": 30070 }, { "epoch": 9.25, "learning_rate": 2.9465004712568637e-07, "loss": 0.027, "step": 30071 }, { "epoch": 9.25, "learning_rate": 2.944099688064939e-07, "loss": 0.0581, "step": 30072 }, { "epoch": 9.25, "learning_rate": 2.9416998687275543e-07, "loss": 0.0306, "step": 30073 }, { "epoch": 9.25, "learning_rate": 2.9393010132685475e-07, "loss": 0.0168, "step": 30074 }, { "epoch": 9.25, "learning_rate": 2.936903121711732e-07, "loss": 0.0162, "step": 30075 }, { "epoch": 9.25, "learning_rate": 2.9345061940809216e-07, "loss": 0.0376, "step": 30076 }, { "epoch": 9.25, "learning_rate": 2.9321102303999316e-07, "loss": 0.0168, "step": 30077 }, { "epoch": 9.25, "learning_rate": 2.929715230692531e-07, "loss": 0.013, "step": 30078 }, { "epoch": 9.25, "learning_rate": 2.927321194982524e-07, "loss": 0.0161, "step": 30079 }, { "epoch": 9.26, "learning_rate": 2.924928123293702e-07, "loss": 0.0383, "step": 30080 }, { "epoch": 9.26, "learning_rate": 2.9225360156497905e-07, "loss": 0.032, "step": 30081 }, { "epoch": 9.26, "learning_rate": 2.9201448720745706e-07, "loss": 0.0367, "step": 30082 }, { "epoch": 9.26, "learning_rate": 2.9177546925917896e-07, "loss": 0.0302, "step": 30083 }, { "epoch": 9.26, "learning_rate": 2.915365477225174e-07, "loss": 0.013, "step": 30084 }, { "epoch": 9.26, "learning_rate": 2.912977225998459e-07, "loss": 0.0193, "step": 30085 }, { "epoch": 9.26, "learning_rate": 2.9105899389353375e-07, "loss": 0.0302, "step": 30086 }, { "epoch": 9.26, "learning_rate": 2.9082036160595683e-07, "loss": 0.0143, "step": 30087 }, { "epoch": 9.26, "learning_rate": 2.905818257394799e-07, "loss": 0.0197, "step": 30088 }, { "epoch": 9.26, "learning_rate": 2.903433862964744e-07, "loss": 0.0152, "step": 30089 }, { "epoch": 9.26, "learning_rate": 2.9010504327930955e-07, "loss": 0.0331, "step": 30090 }, { "epoch": 9.26, "learning_rate": 2.898667966903479e-07, "loss": 0.0219, "step": 30091 }, { "epoch": 9.26, "learning_rate": 2.8962864653195865e-07, "loss": 0.0187, "step": 30092 }, { "epoch": 9.26, "learning_rate": 2.893905928065066e-07, "loss": 0.0275, "step": 30093 }, { "epoch": 9.26, "learning_rate": 2.891526355163565e-07, "loss": 0.0285, "step": 30094 }, { "epoch": 9.26, "learning_rate": 2.8891477466386987e-07, "loss": 0.0139, "step": 30095 }, { "epoch": 9.26, "learning_rate": 2.8867701025140803e-07, "loss": 0.0177, "step": 30096 }, { "epoch": 9.26, "learning_rate": 2.8843934228133586e-07, "loss": 0.0218, "step": 30097 }, { "epoch": 9.26, "learning_rate": 2.882017707560092e-07, "loss": 0.0372, "step": 30098 }, { "epoch": 9.26, "learning_rate": 2.879642956777917e-07, "loss": 0.0092, "step": 30099 }, { "epoch": 9.26, "learning_rate": 2.877269170490393e-07, "loss": 0.0381, "step": 30100 }, { "epoch": 9.26, "learning_rate": 2.87489634872109e-07, "loss": 0.0257, "step": 30101 }, { "epoch": 9.26, "learning_rate": 2.872524491493589e-07, "loss": 0.0185, "step": 30102 }, { "epoch": 9.26, "learning_rate": 2.870153598831416e-07, "loss": 0.0246, "step": 30103 }, { "epoch": 9.26, "learning_rate": 2.867783670758162e-07, "loss": 0.0273, "step": 30104 }, { "epoch": 9.26, "learning_rate": 2.86541470729732e-07, "loss": 0.0286, "step": 30105 }, { "epoch": 9.26, "learning_rate": 2.8630467084724276e-07, "loss": 0.0246, "step": 30106 }, { "epoch": 9.26, "learning_rate": 2.8606796743070096e-07, "loss": 0.0244, "step": 30107 }, { "epoch": 9.26, "learning_rate": 2.8583136048245697e-07, "loss": 0.033, "step": 30108 }, { "epoch": 9.26, "learning_rate": 2.855948500048611e-07, "loss": 0.0264, "step": 30109 }, { "epoch": 9.26, "learning_rate": 2.8535843600026036e-07, "loss": 0.0218, "step": 30110 }, { "epoch": 9.26, "learning_rate": 2.85122118471004e-07, "loss": 0.031, "step": 30111 }, { "epoch": 9.26, "learning_rate": 2.8488589741943907e-07, "loss": 0.0262, "step": 30112 }, { "epoch": 9.27, "learning_rate": 2.8464977284790917e-07, "loss": 0.0256, "step": 30113 }, { "epoch": 9.27, "learning_rate": 2.844137447587636e-07, "loss": 0.0169, "step": 30114 }, { "epoch": 9.27, "learning_rate": 2.841778131543427e-07, "loss": 0.0122, "step": 30115 }, { "epoch": 9.27, "learning_rate": 2.83941978036989e-07, "loss": 0.0334, "step": 30116 }, { "epoch": 9.27, "learning_rate": 2.837062394090473e-07, "loss": 0.0448, "step": 30117 }, { "epoch": 9.27, "learning_rate": 2.8347059727285687e-07, "loss": 0.0334, "step": 30118 }, { "epoch": 9.27, "learning_rate": 2.8323505163075916e-07, "loss": 0.0192, "step": 30119 }, { "epoch": 9.27, "learning_rate": 2.829996024850923e-07, "loss": 0.0301, "step": 30120 }, { "epoch": 9.27, "learning_rate": 2.827642498381955e-07, "loss": 0.0252, "step": 30121 }, { "epoch": 9.27, "learning_rate": 2.8252899369240473e-07, "loss": 0.0198, "step": 30122 }, { "epoch": 9.27, "learning_rate": 2.8229383405005584e-07, "loss": 0.0191, "step": 30123 }, { "epoch": 9.27, "learning_rate": 2.8205877091348587e-07, "loss": 0.0248, "step": 30124 }, { "epoch": 9.27, "learning_rate": 2.8182380428503077e-07, "loss": 0.0321, "step": 30125 }, { "epoch": 9.27, "learning_rate": 2.815889341670186e-07, "loss": 0.0307, "step": 30126 }, { "epoch": 9.27, "learning_rate": 2.813541605617875e-07, "loss": 0.0139, "step": 30127 }, { "epoch": 9.27, "learning_rate": 2.8111948347166574e-07, "loss": 0.0193, "step": 30128 }, { "epoch": 9.27, "learning_rate": 2.808849028989846e-07, "loss": 0.0132, "step": 30129 }, { "epoch": 9.27, "learning_rate": 2.8065041884607347e-07, "loss": 0.0186, "step": 30130 }, { "epoch": 9.27, "learning_rate": 2.804160313152615e-07, "loss": 0.0302, "step": 30131 }, { "epoch": 9.27, "learning_rate": 2.801817403088769e-07, "loss": 0.0197, "step": 30132 }, { "epoch": 9.27, "learning_rate": 2.799475458292444e-07, "loss": 0.0108, "step": 30133 }, { "epoch": 9.27, "learning_rate": 2.7971344787869114e-07, "loss": 0.0139, "step": 30134 }, { "epoch": 9.27, "learning_rate": 2.794794464595441e-07, "loss": 0.039, "step": 30135 }, { "epoch": 9.27, "learning_rate": 2.792455415741224e-07, "loss": 0.0283, "step": 30136 }, { "epoch": 9.27, "learning_rate": 2.790117332247522e-07, "loss": 0.0092, "step": 30137 }, { "epoch": 9.27, "learning_rate": 2.7877802141375366e-07, "loss": 0.0172, "step": 30138 }, { "epoch": 9.27, "learning_rate": 2.785444061434506e-07, "loss": 0.0346, "step": 30139 }, { "epoch": 9.27, "learning_rate": 2.7831088741616e-07, "loss": 0.0186, "step": 30140 }, { "epoch": 9.27, "learning_rate": 2.780774652342011e-07, "loss": 0.0204, "step": 30141 }, { "epoch": 9.27, "learning_rate": 2.778441395998954e-07, "loss": 0.0245, "step": 30142 }, { "epoch": 9.27, "learning_rate": 2.776109105155556e-07, "loss": 0.0272, "step": 30143 }, { "epoch": 9.27, "learning_rate": 2.7737777798349964e-07, "loss": 0.0315, "step": 30144 }, { "epoch": 9.28, "learning_rate": 2.771447420060447e-07, "loss": 0.0317, "step": 30145 }, { "epoch": 9.28, "learning_rate": 2.769118025855022e-07, "loss": 0.0139, "step": 30146 }, { "epoch": 9.28, "learning_rate": 2.76678959724187e-07, "loss": 0.0216, "step": 30147 }, { "epoch": 9.28, "learning_rate": 2.7644621342441056e-07, "loss": 0.0264, "step": 30148 }, { "epoch": 9.28, "learning_rate": 2.762135636884855e-07, "loss": 0.0193, "step": 30149 }, { "epoch": 9.28, "learning_rate": 2.7598101051872104e-07, "loss": 0.0199, "step": 30150 }, { "epoch": 9.28, "learning_rate": 2.7574855391742537e-07, "loss": 0.025, "step": 30151 }, { "epoch": 9.28, "learning_rate": 2.755161938869122e-07, "loss": 0.034, "step": 30152 }, { "epoch": 9.28, "learning_rate": 2.75283930429483e-07, "loss": 0.0236, "step": 30153 }, { "epoch": 9.28, "learning_rate": 2.75051763547447e-07, "loss": 0.0133, "step": 30154 }, { "epoch": 9.28, "learning_rate": 2.748196932431102e-07, "loss": 0.0211, "step": 30155 }, { "epoch": 9.28, "learning_rate": 2.745877195187774e-07, "loss": 0.0101, "step": 30156 }, { "epoch": 9.28, "learning_rate": 2.743558423767512e-07, "loss": 0.0224, "step": 30157 }, { "epoch": 9.28, "learning_rate": 2.7412406181933416e-07, "loss": 0.0732, "step": 30158 }, { "epoch": 9.28, "learning_rate": 2.7389237784883116e-07, "loss": 0.0141, "step": 30159 }, { "epoch": 9.28, "learning_rate": 2.7366079046753925e-07, "loss": 0.0231, "step": 30160 }, { "epoch": 9.28, "learning_rate": 2.734292996777588e-07, "loss": 0.0269, "step": 30161 }, { "epoch": 9.28, "learning_rate": 2.7319790548179124e-07, "loss": 0.0224, "step": 30162 }, { "epoch": 9.28, "learning_rate": 2.729666078819326e-07, "loss": 0.0207, "step": 30163 }, { "epoch": 9.28, "learning_rate": 2.7273540688047995e-07, "loss": 0.0135, "step": 30164 }, { "epoch": 9.28, "learning_rate": 2.7250430247972915e-07, "loss": 0.0267, "step": 30165 }, { "epoch": 9.28, "learning_rate": 2.7227329468197616e-07, "loss": 0.0259, "step": 30166 }, { "epoch": 9.28, "learning_rate": 2.7204238348951586e-07, "loss": 0.0205, "step": 30167 }, { "epoch": 9.28, "learning_rate": 2.7181156890463745e-07, "loss": 0.0185, "step": 30168 }, { "epoch": 9.28, "learning_rate": 2.7158085092963916e-07, "loss": 0.037, "step": 30169 }, { "epoch": 9.28, "learning_rate": 2.71350229566808e-07, "loss": 0.014, "step": 30170 }, { "epoch": 9.28, "learning_rate": 2.711197048184333e-07, "loss": 0.021, "step": 30171 }, { "epoch": 9.28, "learning_rate": 2.7088927668680765e-07, "loss": 0.0247, "step": 30172 }, { "epoch": 9.28, "learning_rate": 2.706589451742181e-07, "loss": 0.0166, "step": 30173 }, { "epoch": 9.28, "learning_rate": 2.704287102829528e-07, "loss": 0.0201, "step": 30174 }, { "epoch": 9.28, "learning_rate": 2.7019857201529665e-07, "loss": 0.025, "step": 30175 }, { "epoch": 9.28, "learning_rate": 2.6996853037353665e-07, "loss": 0.0267, "step": 30176 }, { "epoch": 9.28, "learning_rate": 2.6973858535995547e-07, "loss": 0.0295, "step": 30177 }, { "epoch": 9.29, "learning_rate": 2.695087369768379e-07, "loss": 0.0189, "step": 30178 }, { "epoch": 9.29, "learning_rate": 2.6927898522646655e-07, "loss": 0.0294, "step": 30179 }, { "epoch": 9.29, "learning_rate": 2.690493301111252e-07, "loss": 0.0319, "step": 30180 }, { "epoch": 9.29, "learning_rate": 2.688197716330887e-07, "loss": 0.015, "step": 30181 }, { "epoch": 9.29, "learning_rate": 2.6859030979464296e-07, "loss": 0.0066, "step": 30182 }, { "epoch": 9.29, "learning_rate": 2.6836094459806284e-07, "loss": 0.0435, "step": 30183 }, { "epoch": 9.29, "learning_rate": 2.681316760456287e-07, "loss": 0.0123, "step": 30184 }, { "epoch": 9.29, "learning_rate": 2.679025041396155e-07, "loss": 0.0058, "step": 30185 }, { "epoch": 9.29, "learning_rate": 2.6767342888229907e-07, "loss": 0.033, "step": 30186 }, { "epoch": 9.29, "learning_rate": 2.6744445027595657e-07, "loss": 0.0183, "step": 30187 }, { "epoch": 9.29, "learning_rate": 2.6721556832285836e-07, "loss": 0.04, "step": 30188 }, { "epoch": 9.29, "learning_rate": 2.6698678302528146e-07, "loss": 0.0219, "step": 30189 }, { "epoch": 9.29, "learning_rate": 2.667580943854964e-07, "loss": 0.0175, "step": 30190 }, { "epoch": 9.29, "learning_rate": 2.6652950240577234e-07, "loss": 0.0282, "step": 30191 }, { "epoch": 9.29, "learning_rate": 2.663010070883809e-07, "loss": 0.0174, "step": 30192 }, { "epoch": 9.29, "learning_rate": 2.6607260843559245e-07, "loss": 0.0107, "step": 30193 }, { "epoch": 9.29, "learning_rate": 2.6584430644967296e-07, "loss": 0.0142, "step": 30194 }, { "epoch": 9.29, "learning_rate": 2.656161011328917e-07, "loss": 0.0324, "step": 30195 }, { "epoch": 9.29, "learning_rate": 2.6538799248751245e-07, "loss": 0.0213, "step": 30196 }, { "epoch": 9.29, "learning_rate": 2.651599805158045e-07, "loss": 0.0244, "step": 30197 }, { "epoch": 9.29, "learning_rate": 2.6493206522002825e-07, "loss": 0.0264, "step": 30198 }, { "epoch": 9.29, "learning_rate": 2.647042466024485e-07, "loss": 0.0241, "step": 30199 }, { "epoch": 9.29, "learning_rate": 2.644765246653291e-07, "loss": 0.023, "step": 30200 }, { "epoch": 9.29, "learning_rate": 2.6424889941092933e-07, "loss": 0.0342, "step": 30201 }, { "epoch": 9.29, "learning_rate": 2.640213708415118e-07, "loss": 0.0202, "step": 30202 }, { "epoch": 9.29, "learning_rate": 2.637939389593336e-07, "loss": 0.0264, "step": 30203 }, { "epoch": 9.29, "learning_rate": 2.6356660376665623e-07, "loss": 0.0293, "step": 30204 }, { "epoch": 9.29, "learning_rate": 2.633393652657357e-07, "loss": 0.0254, "step": 30205 }, { "epoch": 9.29, "learning_rate": 2.631122234588268e-07, "loss": 0.0225, "step": 30206 }, { "epoch": 9.29, "learning_rate": 2.6288517834818896e-07, "loss": 0.013, "step": 30207 }, { "epoch": 9.29, "learning_rate": 2.626582299360747e-07, "loss": 0.0252, "step": 30208 }, { "epoch": 9.29, "learning_rate": 2.62431378224739e-07, "loss": 0.0275, "step": 30209 }, { "epoch": 9.3, "learning_rate": 2.622046232164344e-07, "loss": 0.022, "step": 30210 }, { "epoch": 9.3, "learning_rate": 2.619779649134124e-07, "loss": 0.0185, "step": 30211 }, { "epoch": 9.3, "learning_rate": 2.617514033179236e-07, "loss": 0.0181, "step": 30212 }, { "epoch": 9.3, "learning_rate": 2.6152493843221714e-07, "loss": 0.0283, "step": 30213 }, { "epoch": 9.3, "learning_rate": 2.612985702585458e-07, "loss": 0.0187, "step": 30214 }, { "epoch": 9.3, "learning_rate": 2.610722987991543e-07, "loss": 0.0203, "step": 30215 }, { "epoch": 9.3, "learning_rate": 2.608461240562887e-07, "loss": 0.026, "step": 30216 }, { "epoch": 9.3, "learning_rate": 2.606200460321984e-07, "loss": 0.0198, "step": 30217 }, { "epoch": 9.3, "learning_rate": 2.60394064729127e-07, "loss": 0.0166, "step": 30218 }, { "epoch": 9.3, "learning_rate": 2.6016818014931946e-07, "loss": 0.0339, "step": 30219 }, { "epoch": 9.3, "learning_rate": 2.5994239229501727e-07, "loss": 0.0167, "step": 30220 }, { "epoch": 9.3, "learning_rate": 2.5971670116846537e-07, "loss": 0.0143, "step": 30221 }, { "epoch": 9.3, "learning_rate": 2.594911067719019e-07, "loss": 0.0218, "step": 30222 }, { "epoch": 9.3, "learning_rate": 2.592656091075685e-07, "loss": 0.0114, "step": 30223 }, { "epoch": 9.3, "learning_rate": 2.590402081777066e-07, "loss": 0.0234, "step": 30224 }, { "epoch": 9.3, "learning_rate": 2.588149039845533e-07, "loss": 0.0104, "step": 30225 }, { "epoch": 9.3, "learning_rate": 2.585896965303436e-07, "loss": 0.0205, "step": 30226 }, { "epoch": 9.3, "learning_rate": 2.583645858173178e-07, "loss": 0.0154, "step": 30227 }, { "epoch": 9.3, "learning_rate": 2.5813957184770977e-07, "loss": 0.0171, "step": 30228 }, { "epoch": 9.3, "learning_rate": 2.579146546237532e-07, "loss": 0.0255, "step": 30229 }, { "epoch": 9.3, "learning_rate": 2.57689834147683e-07, "loss": 0.0474, "step": 30230 }, { "epoch": 9.3, "learning_rate": 2.574651104217318e-07, "loss": 0.0389, "step": 30231 }, { "epoch": 9.3, "learning_rate": 2.5724048344813125e-07, "loss": 0.0176, "step": 30232 }, { "epoch": 9.3, "learning_rate": 2.5701595322911055e-07, "loss": 0.0178, "step": 30233 }, { "epoch": 9.3, "learning_rate": 2.5679151976690243e-07, "loss": 0.036, "step": 30234 }, { "epoch": 9.3, "learning_rate": 2.5656718306373396e-07, "loss": 0.0187, "step": 30235 }, { "epoch": 9.3, "learning_rate": 2.5634294312183226e-07, "loss": 0.0192, "step": 30236 }, { "epoch": 9.3, "learning_rate": 2.561187999434256e-07, "loss": 0.0248, "step": 30237 }, { "epoch": 9.3, "learning_rate": 2.5589475353073987e-07, "loss": 0.029, "step": 30238 }, { "epoch": 9.3, "learning_rate": 2.556708038860001e-07, "loss": 0.0263, "step": 30239 }, { "epoch": 9.3, "learning_rate": 2.554469510114288e-07, "loss": 0.0183, "step": 30240 }, { "epoch": 9.3, "learning_rate": 2.5522319490924987e-07, "loss": 0.0281, "step": 30241 }, { "epoch": 9.3, "learning_rate": 2.54999535581687e-07, "loss": 0.0239, "step": 30242 }, { "epoch": 9.31, "learning_rate": 2.547759730309585e-07, "loss": 0.0302, "step": 30243 }, { "epoch": 9.31, "learning_rate": 2.5455250725928695e-07, "loss": 0.017, "step": 30244 }, { "epoch": 9.31, "learning_rate": 2.5432913826889063e-07, "loss": 0.021, "step": 30245 }, { "epoch": 9.31, "learning_rate": 2.5410586606198775e-07, "loss": 0.0183, "step": 30246 }, { "epoch": 9.31, "learning_rate": 2.538826906407954e-07, "loss": 0.019, "step": 30247 }, { "epoch": 9.31, "learning_rate": 2.5365961200753074e-07, "loss": 0.0138, "step": 30248 }, { "epoch": 9.31, "learning_rate": 2.5343663016440865e-07, "loss": 0.0272, "step": 30249 }, { "epoch": 9.31, "learning_rate": 2.532137451136418e-07, "loss": 0.0177, "step": 30250 }, { "epoch": 9.31, "learning_rate": 2.5299095685744734e-07, "loss": 0.0217, "step": 30251 }, { "epoch": 9.31, "learning_rate": 2.527682653980368e-07, "loss": 0.0211, "step": 30252 }, { "epoch": 9.31, "learning_rate": 2.5254567073761835e-07, "loss": 0.0137, "step": 30253 }, { "epoch": 9.31, "learning_rate": 2.5232317287840593e-07, "loss": 0.0209, "step": 30254 }, { "epoch": 9.31, "learning_rate": 2.521007718226076e-07, "loss": 0.019, "step": 30255 }, { "epoch": 9.31, "learning_rate": 2.5187846757243286e-07, "loss": 0.0129, "step": 30256 }, { "epoch": 9.31, "learning_rate": 2.516562601300898e-07, "loss": 0.0097, "step": 30257 }, { "epoch": 9.31, "learning_rate": 2.5143414949778235e-07, "loss": 0.0327, "step": 30258 }, { "epoch": 9.31, "learning_rate": 2.5121213567772085e-07, "loss": 0.0139, "step": 30259 }, { "epoch": 9.31, "learning_rate": 2.509902186721058e-07, "loss": 0.0281, "step": 30260 }, { "epoch": 9.31, "learning_rate": 2.507683984831433e-07, "loss": 0.0251, "step": 30261 }, { "epoch": 9.31, "learning_rate": 2.505466751130359e-07, "loss": 0.0195, "step": 30262 }, { "epoch": 9.31, "learning_rate": 2.503250485639852e-07, "loss": 0.0192, "step": 30263 }, { "epoch": 9.31, "learning_rate": 2.5010351883819283e-07, "loss": 0.0222, "step": 30264 }, { "epoch": 9.31, "learning_rate": 2.49882085937857e-07, "loss": 0.0177, "step": 30265 }, { "epoch": 9.31, "learning_rate": 2.496607498651804e-07, "loss": 0.0123, "step": 30266 }, { "epoch": 9.31, "learning_rate": 2.4943951062235683e-07, "loss": 0.0269, "step": 30267 }, { "epoch": 9.31, "learning_rate": 2.492183682115856e-07, "loss": 0.0317, "step": 30268 }, { "epoch": 9.31, "learning_rate": 2.4899732263506395e-07, "loss": 0.0331, "step": 30269 }, { "epoch": 9.31, "learning_rate": 2.4877637389498333e-07, "loss": 0.0233, "step": 30270 }, { "epoch": 9.31, "learning_rate": 2.4855552199354096e-07, "loss": 0.0196, "step": 30271 }, { "epoch": 9.31, "learning_rate": 2.483347669329306e-07, "loss": 0.0123, "step": 30272 }, { "epoch": 9.31, "learning_rate": 2.481141087153427e-07, "loss": 0.0223, "step": 30273 }, { "epoch": 9.31, "learning_rate": 2.4789354734297e-07, "loss": 0.0238, "step": 30274 }, { "epoch": 9.32, "learning_rate": 2.4767308281800074e-07, "loss": 0.0295, "step": 30275 }, { "epoch": 9.32, "learning_rate": 2.474527151426276e-07, "loss": 0.0138, "step": 30276 }, { "epoch": 9.32, "learning_rate": 2.472324443190355e-07, "loss": 0.0258, "step": 30277 }, { "epoch": 9.32, "learning_rate": 2.4701227034941376e-07, "loss": 0.0268, "step": 30278 }, { "epoch": 9.32, "learning_rate": 2.467921932359496e-07, "loss": 0.0343, "step": 30279 }, { "epoch": 9.32, "learning_rate": 2.465722129808279e-07, "loss": 0.023, "step": 30280 }, { "epoch": 9.32, "learning_rate": 2.4635232958623247e-07, "loss": 0.0225, "step": 30281 }, { "epoch": 9.32, "learning_rate": 2.461325430543482e-07, "loss": 0.0368, "step": 30282 }, { "epoch": 9.32, "learning_rate": 2.459128533873567e-07, "loss": 0.0282, "step": 30283 }, { "epoch": 9.32, "learning_rate": 2.456932605874407e-07, "loss": 0.0182, "step": 30284 }, { "epoch": 9.32, "learning_rate": 2.4547376465677954e-07, "loss": 0.0252, "step": 30285 }, { "epoch": 9.32, "learning_rate": 2.4525436559755475e-07, "loss": 0.0275, "step": 30286 }, { "epoch": 9.32, "learning_rate": 2.450350634119458e-07, "loss": 0.0249, "step": 30287 }, { "epoch": 9.32, "learning_rate": 2.448158581021265e-07, "loss": 0.0199, "step": 30288 }, { "epoch": 9.32, "learning_rate": 2.4459674967027725e-07, "loss": 0.0286, "step": 30289 }, { "epoch": 9.32, "learning_rate": 2.4437773811857304e-07, "loss": 0.0177, "step": 30290 }, { "epoch": 9.32, "learning_rate": 2.4415882344918983e-07, "loss": 0.0175, "step": 30291 }, { "epoch": 9.32, "learning_rate": 2.439400056642993e-07, "loss": 0.0297, "step": 30292 }, { "epoch": 9.32, "learning_rate": 2.4372128476607747e-07, "loss": 0.0151, "step": 30293 }, { "epoch": 9.32, "learning_rate": 2.435026607566937e-07, "loss": 0.0115, "step": 30294 }, { "epoch": 9.32, "learning_rate": 2.432841336383207e-07, "loss": 0.0261, "step": 30295 }, { "epoch": 9.32, "learning_rate": 2.43065703413129e-07, "loss": 0.0216, "step": 30296 }, { "epoch": 9.32, "learning_rate": 2.4284737008328676e-07, "loss": 0.0195, "step": 30297 }, { "epoch": 9.32, "learning_rate": 2.426291336509623e-07, "loss": 0.0215, "step": 30298 }, { "epoch": 9.32, "learning_rate": 2.424109941183228e-07, "loss": 0.0313, "step": 30299 }, { "epoch": 9.32, "learning_rate": 2.421929514875354e-07, "loss": 0.0266, "step": 30300 }, { "epoch": 9.32, "learning_rate": 2.419750057607662e-07, "loss": 0.0269, "step": 30301 }, { "epoch": 9.32, "learning_rate": 2.4175715694017663e-07, "loss": 0.0174, "step": 30302 }, { "epoch": 9.32, "learning_rate": 2.4153940502793185e-07, "loss": 0.0175, "step": 30303 }, { "epoch": 9.32, "learning_rate": 2.413217500261966e-07, "loss": 0.0144, "step": 30304 }, { "epoch": 9.32, "learning_rate": 2.411041919371282e-07, "loss": 0.0214, "step": 30305 }, { "epoch": 9.32, "learning_rate": 2.4088673076289037e-07, "loss": 0.0269, "step": 30306 }, { "epoch": 9.32, "learning_rate": 2.4066936650564254e-07, "loss": 0.0105, "step": 30307 }, { "epoch": 9.33, "learning_rate": 2.404520991675396e-07, "loss": 0.0245, "step": 30308 }, { "epoch": 9.33, "learning_rate": 2.4023492875074326e-07, "loss": 0.0288, "step": 30309 }, { "epoch": 9.33, "learning_rate": 2.400178552574095e-07, "loss": 0.0118, "step": 30310 }, { "epoch": 9.33, "learning_rate": 2.3980087868969327e-07, "loss": 0.026, "step": 30311 }, { "epoch": 9.33, "learning_rate": 2.395839990497495e-07, "loss": 0.0207, "step": 30312 }, { "epoch": 9.33, "learning_rate": 2.393672163397309e-07, "loss": 0.0373, "step": 30313 }, { "epoch": 9.33, "learning_rate": 2.391505305617936e-07, "loss": 0.0231, "step": 30314 }, { "epoch": 9.33, "learning_rate": 2.389339417180858e-07, "loss": 0.0144, "step": 30315 }, { "epoch": 9.33, "learning_rate": 2.387174498107614e-07, "loss": 0.0174, "step": 30316 }, { "epoch": 9.33, "learning_rate": 2.385010548419675e-07, "loss": 0.0221, "step": 30317 }, { "epoch": 9.33, "learning_rate": 2.3828475681385576e-07, "loss": 0.0241, "step": 30318 }, { "epoch": 9.33, "learning_rate": 2.3806855572857223e-07, "loss": 0.0379, "step": 30319 }, { "epoch": 9.33, "learning_rate": 2.378524515882641e-07, "loss": 0.0212, "step": 30320 }, { "epoch": 9.33, "learning_rate": 2.3763644439508072e-07, "loss": 0.0176, "step": 30321 }, { "epoch": 9.33, "learning_rate": 2.3742053415116372e-07, "loss": 0.0187, "step": 30322 }, { "epoch": 9.33, "learning_rate": 2.3720472085865697e-07, "loss": 0.0122, "step": 30323 }, { "epoch": 9.33, "learning_rate": 2.3698900451970542e-07, "loss": 0.0136, "step": 30324 }, { "epoch": 9.33, "learning_rate": 2.3677338513645176e-07, "loss": 0.0171, "step": 30325 }, { "epoch": 9.33, "learning_rate": 2.3655786271103542e-07, "loss": 0.0256, "step": 30326 }, { "epoch": 9.33, "learning_rate": 2.3634243724559915e-07, "loss": 0.0285, "step": 30327 }, { "epoch": 9.33, "learning_rate": 2.3612710874228007e-07, "loss": 0.0285, "step": 30328 }, { "epoch": 9.33, "learning_rate": 2.359118772032176e-07, "loss": 0.0274, "step": 30329 }, { "epoch": 9.33, "learning_rate": 2.3569674263054788e-07, "loss": 0.027, "step": 30330 }, { "epoch": 9.33, "learning_rate": 2.354817050264102e-07, "loss": 0.0318, "step": 30331 }, { "epoch": 9.33, "learning_rate": 2.352667643929385e-07, "loss": 0.029, "step": 30332 }, { "epoch": 9.33, "learning_rate": 2.3505192073226547e-07, "loss": 0.0187, "step": 30333 }, { "epoch": 9.33, "learning_rate": 2.3483717404652828e-07, "loss": 0.0194, "step": 30334 }, { "epoch": 9.33, "learning_rate": 2.3462252433785637e-07, "loss": 0.021, "step": 30335 }, { "epoch": 9.33, "learning_rate": 2.3440797160838358e-07, "loss": 0.0125, "step": 30336 }, { "epoch": 9.33, "learning_rate": 2.3419351586024041e-07, "loss": 0.0274, "step": 30337 }, { "epoch": 9.33, "learning_rate": 2.3397915709555518e-07, "loss": 0.0219, "step": 30338 }, { "epoch": 9.33, "learning_rate": 2.337648953164584e-07, "loss": 0.0379, "step": 30339 }, { "epoch": 9.34, "learning_rate": 2.3355073052507615e-07, "loss": 0.0172, "step": 30340 }, { "epoch": 9.34, "learning_rate": 2.3333666272353672e-07, "loss": 0.0269, "step": 30341 }, { "epoch": 9.34, "learning_rate": 2.3312269191396619e-07, "loss": 0.0147, "step": 30342 }, { "epoch": 9.34, "learning_rate": 2.329088180984862e-07, "loss": 0.03, "step": 30343 }, { "epoch": 9.34, "learning_rate": 2.3269504127922504e-07, "loss": 0.0332, "step": 30344 }, { "epoch": 9.34, "learning_rate": 2.3248136145830434e-07, "loss": 0.0163, "step": 30345 }, { "epoch": 9.34, "learning_rate": 2.322677786378458e-07, "loss": 0.0197, "step": 30346 }, { "epoch": 9.34, "learning_rate": 2.3205429281996982e-07, "loss": 0.0282, "step": 30347 }, { "epoch": 9.34, "learning_rate": 2.3184090400679703e-07, "loss": 0.0203, "step": 30348 }, { "epoch": 9.34, "learning_rate": 2.3162761220044684e-07, "loss": 0.0161, "step": 30349 }, { "epoch": 9.34, "learning_rate": 2.314144174030364e-07, "loss": 0.0209, "step": 30350 }, { "epoch": 9.34, "learning_rate": 2.31201319616684e-07, "loss": 0.019, "step": 30351 }, { "epoch": 9.34, "learning_rate": 2.3098831884350804e-07, "loss": 0.0112, "step": 30352 }, { "epoch": 9.34, "learning_rate": 2.3077541508561896e-07, "loss": 0.0197, "step": 30353 }, { "epoch": 9.34, "learning_rate": 2.30562608345134e-07, "loss": 0.0285, "step": 30354 }, { "epoch": 9.34, "learning_rate": 2.30349898624167e-07, "loss": 0.0191, "step": 30355 }, { "epoch": 9.34, "learning_rate": 2.3013728592482852e-07, "loss": 0.0213, "step": 30356 }, { "epoch": 9.34, "learning_rate": 2.2992477024923132e-07, "loss": 0.0264, "step": 30357 }, { "epoch": 9.34, "learning_rate": 2.297123515994848e-07, "loss": 0.0133, "step": 30358 }, { "epoch": 9.34, "learning_rate": 2.295000299777006e-07, "loss": 0.0125, "step": 30359 }, { "epoch": 9.34, "learning_rate": 2.2928780538598483e-07, "loss": 0.0181, "step": 30360 }, { "epoch": 9.34, "learning_rate": 2.290756778264469e-07, "loss": 0.0139, "step": 30361 }, { "epoch": 9.34, "learning_rate": 2.288636473011918e-07, "loss": 0.0225, "step": 30362 }, { "epoch": 9.34, "learning_rate": 2.2865171381232677e-07, "loss": 0.0209, "step": 30363 }, { "epoch": 9.34, "learning_rate": 2.284398773619556e-07, "loss": 0.0116, "step": 30364 }, { "epoch": 9.34, "learning_rate": 2.2822813795218112e-07, "loss": 0.0359, "step": 30365 }, { "epoch": 9.34, "learning_rate": 2.280164955851083e-07, "loss": 0.0215, "step": 30366 }, { "epoch": 9.34, "learning_rate": 2.2780495026283766e-07, "loss": 0.0189, "step": 30367 }, { "epoch": 9.34, "learning_rate": 2.2759350198746978e-07, "loss": 0.0291, "step": 30368 }, { "epoch": 9.34, "learning_rate": 2.2738215076110625e-07, "loss": 0.0197, "step": 30369 }, { "epoch": 9.34, "learning_rate": 2.2717089658584323e-07, "loss": 0.0373, "step": 30370 }, { "epoch": 9.34, "learning_rate": 2.2695973946378013e-07, "loss": 0.0376, "step": 30371 }, { "epoch": 9.34, "learning_rate": 2.2674867939701418e-07, "loss": 0.0223, "step": 30372 }, { "epoch": 9.35, "learning_rate": 2.2653771638764144e-07, "loss": 0.0153, "step": 30373 }, { "epoch": 9.35, "learning_rate": 2.2632685043775692e-07, "loss": 0.0262, "step": 30374 }, { "epoch": 9.35, "learning_rate": 2.2611608154945343e-07, "loss": 0.0206, "step": 30375 }, { "epoch": 9.35, "learning_rate": 2.25905409724827e-07, "loss": 0.0356, "step": 30376 }, { "epoch": 9.35, "learning_rate": 2.25694834965966e-07, "loss": 0.0338, "step": 30377 }, { "epoch": 9.35, "learning_rate": 2.254843572749632e-07, "loss": 0.0309, "step": 30378 }, { "epoch": 9.35, "learning_rate": 2.2527397665391026e-07, "loss": 0.0148, "step": 30379 }, { "epoch": 9.35, "learning_rate": 2.250636931048955e-07, "loss": 0.0222, "step": 30380 }, { "epoch": 9.35, "learning_rate": 2.2485350663000727e-07, "loss": 0.0417, "step": 30381 }, { "epoch": 9.35, "learning_rate": 2.2464341723133276e-07, "loss": 0.0224, "step": 30382 }, { "epoch": 9.35, "learning_rate": 2.2443342491095809e-07, "loss": 0.0346, "step": 30383 }, { "epoch": 9.35, "learning_rate": 2.242235296709694e-07, "loss": 0.0331, "step": 30384 }, { "epoch": 9.35, "learning_rate": 2.240137315134483e-07, "loss": 0.0193, "step": 30385 }, { "epoch": 9.35, "learning_rate": 2.2380403044048428e-07, "loss": 0.0211, "step": 30386 }, { "epoch": 9.35, "learning_rate": 2.2359442645415452e-07, "loss": 0.032, "step": 30387 }, { "epoch": 9.35, "learning_rate": 2.2338491955654184e-07, "loss": 0.0308, "step": 30388 }, { "epoch": 9.35, "learning_rate": 2.2317550974972679e-07, "loss": 0.0218, "step": 30389 }, { "epoch": 9.35, "learning_rate": 2.2296619703579102e-07, "loss": 0.018, "step": 30390 }, { "epoch": 9.35, "learning_rate": 2.2275698141681178e-07, "loss": 0.0316, "step": 30391 }, { "epoch": 9.35, "learning_rate": 2.2254786289486518e-07, "loss": 0.0274, "step": 30392 }, { "epoch": 9.35, "learning_rate": 2.2233884147203065e-07, "loss": 0.0139, "step": 30393 }, { "epoch": 9.35, "learning_rate": 2.2212991715038324e-07, "loss": 0.0201, "step": 30394 }, { "epoch": 9.35, "learning_rate": 2.2192108993199567e-07, "loss": 0.0162, "step": 30395 }, { "epoch": 9.35, "learning_rate": 2.2171235981894412e-07, "loss": 0.0099, "step": 30396 }, { "epoch": 9.35, "learning_rate": 2.2150372681330135e-07, "loss": 0.0218, "step": 30397 }, { "epoch": 9.35, "learning_rate": 2.2129519091713792e-07, "loss": 0.0286, "step": 30398 }, { "epoch": 9.35, "learning_rate": 2.210867521325255e-07, "loss": 0.0282, "step": 30399 }, { "epoch": 9.35, "learning_rate": 2.2087841046153468e-07, "loss": 0.0184, "step": 30400 }, { "epoch": 9.35, "learning_rate": 2.206701659062338e-07, "loss": 0.0216, "step": 30401 }, { "epoch": 9.35, "learning_rate": 2.204620184686912e-07, "loss": 0.0145, "step": 30402 }, { "epoch": 9.35, "learning_rate": 2.20253968150973e-07, "loss": 0.0225, "step": 30403 }, { "epoch": 9.35, "learning_rate": 2.2004601495514866e-07, "loss": 0.0166, "step": 30404 }, { "epoch": 9.36, "learning_rate": 2.1983815888327765e-07, "loss": 0.014, "step": 30405 }, { "epoch": 9.36, "learning_rate": 2.196303999374294e-07, "loss": 0.0309, "step": 30406 }, { "epoch": 9.36, "learning_rate": 2.1942273811966563e-07, "loss": 0.0296, "step": 30407 }, { "epoch": 9.36, "learning_rate": 2.192151734320458e-07, "loss": 0.0265, "step": 30408 }, { "epoch": 9.36, "learning_rate": 2.1900770587663488e-07, "loss": 0.0245, "step": 30409 }, { "epoch": 9.36, "learning_rate": 2.188003354554924e-07, "loss": 0.0333, "step": 30410 }, { "epoch": 9.36, "learning_rate": 2.1859306217067665e-07, "loss": 0.015, "step": 30411 }, { "epoch": 9.36, "learning_rate": 2.1838588602424604e-07, "loss": 0.0283, "step": 30412 }, { "epoch": 9.36, "learning_rate": 2.1817880701825888e-07, "loss": 0.0265, "step": 30413 }, { "epoch": 9.36, "learning_rate": 2.1797182515477244e-07, "loss": 0.0229, "step": 30414 }, { "epoch": 9.36, "learning_rate": 2.1776494043583952e-07, "loss": 0.017, "step": 30415 }, { "epoch": 9.36, "learning_rate": 2.1755815286351734e-07, "loss": 0.0268, "step": 30416 }, { "epoch": 9.36, "learning_rate": 2.173514624398576e-07, "loss": 0.0248, "step": 30417 }, { "epoch": 9.36, "learning_rate": 2.1714486916691534e-07, "loss": 0.0354, "step": 30418 }, { "epoch": 9.36, "learning_rate": 2.1693837304673892e-07, "loss": 0.0216, "step": 30419 }, { "epoch": 9.36, "learning_rate": 2.1673197408138115e-07, "loss": 0.0321, "step": 30420 }, { "epoch": 9.36, "learning_rate": 2.1652567227289257e-07, "loss": 0.018, "step": 30421 }, { "epoch": 9.36, "learning_rate": 2.1631946762332045e-07, "loss": 0.0321, "step": 30422 }, { "epoch": 9.36, "learning_rate": 2.1611336013471207e-07, "loss": 0.0147, "step": 30423 }, { "epoch": 9.36, "learning_rate": 2.1590734980911575e-07, "loss": 0.0263, "step": 30424 }, { "epoch": 9.36, "learning_rate": 2.1570143664857657e-07, "loss": 0.0167, "step": 30425 }, { "epoch": 9.36, "learning_rate": 2.1549562065513952e-07, "loss": 0.0212, "step": 30426 }, { "epoch": 9.36, "learning_rate": 2.1528990183084853e-07, "loss": 0.0361, "step": 30427 }, { "epoch": 9.36, "learning_rate": 2.1508428017774751e-07, "loss": 0.0242, "step": 30428 }, { "epoch": 9.36, "learning_rate": 2.1487875569787708e-07, "loss": 0.0163, "step": 30429 }, { "epoch": 9.36, "learning_rate": 2.1467332839327783e-07, "loss": 0.0366, "step": 30430 }, { "epoch": 9.36, "learning_rate": 2.1446799826599252e-07, "loss": 0.0124, "step": 30431 }, { "epoch": 9.36, "learning_rate": 2.1426276531805845e-07, "loss": 0.0266, "step": 30432 }, { "epoch": 9.36, "learning_rate": 2.1405762955151178e-07, "loss": 0.019, "step": 30433 }, { "epoch": 9.36, "learning_rate": 2.1385259096839416e-07, "loss": 0.0327, "step": 30434 }, { "epoch": 9.36, "learning_rate": 2.1364764957073848e-07, "loss": 0.0212, "step": 30435 }, { "epoch": 9.36, "learning_rate": 2.134428053605808e-07, "loss": 0.0212, "step": 30436 }, { "epoch": 9.36, "learning_rate": 2.1323805833995627e-07, "loss": 0.026, "step": 30437 }, { "epoch": 9.37, "learning_rate": 2.130334085108976e-07, "loss": 0.0114, "step": 30438 }, { "epoch": 9.37, "learning_rate": 2.1282885587543657e-07, "loss": 0.0147, "step": 30439 }, { "epoch": 9.37, "learning_rate": 2.1262440043560483e-07, "loss": 0.0074, "step": 30440 }, { "epoch": 9.37, "learning_rate": 2.1242004219343415e-07, "loss": 0.0261, "step": 30441 }, { "epoch": 9.37, "learning_rate": 2.1221578115095286e-07, "loss": 0.0207, "step": 30442 }, { "epoch": 9.37, "learning_rate": 2.1201161731018828e-07, "loss": 0.0202, "step": 30443 }, { "epoch": 9.37, "learning_rate": 2.1180755067317093e-07, "loss": 0.0441, "step": 30444 }, { "epoch": 9.37, "learning_rate": 2.1160358124192482e-07, "loss": 0.0252, "step": 30445 }, { "epoch": 9.37, "learning_rate": 2.1139970901847607e-07, "loss": 0.0332, "step": 30446 }, { "epoch": 9.37, "learning_rate": 2.1119593400484972e-07, "loss": 0.0218, "step": 30447 }, { "epoch": 9.37, "learning_rate": 2.1099225620307085e-07, "loss": 0.0378, "step": 30448 }, { "epoch": 9.37, "learning_rate": 2.1078867561515893e-07, "loss": 0.024, "step": 30449 }, { "epoch": 9.37, "learning_rate": 2.1058519224313788e-07, "loss": 0.0349, "step": 30450 }, { "epoch": 9.37, "learning_rate": 2.1038180608902837e-07, "loss": 0.0229, "step": 30451 }, { "epoch": 9.37, "learning_rate": 2.1017851715484984e-07, "loss": 0.01, "step": 30452 }, { "epoch": 9.37, "learning_rate": 2.0997532544262068e-07, "loss": 0.0184, "step": 30453 }, { "epoch": 9.37, "learning_rate": 2.0977223095435927e-07, "loss": 0.016, "step": 30454 }, { "epoch": 9.37, "learning_rate": 2.0956923369208403e-07, "loss": 0.0212, "step": 30455 }, { "epoch": 9.37, "learning_rate": 2.0936633365780779e-07, "loss": 0.0218, "step": 30456 }, { "epoch": 9.37, "learning_rate": 2.091635308535478e-07, "loss": 0.0198, "step": 30457 }, { "epoch": 9.37, "learning_rate": 2.089608252813169e-07, "loss": 0.0341, "step": 30458 }, { "epoch": 9.37, "learning_rate": 2.0875821694313014e-07, "loss": 0.0119, "step": 30459 }, { "epoch": 9.37, "learning_rate": 2.0855570584099704e-07, "loss": 0.0298, "step": 30460 }, { "epoch": 9.37, "learning_rate": 2.0835329197692932e-07, "loss": 0.0261, "step": 30461 }, { "epoch": 9.37, "learning_rate": 2.081509753529387e-07, "loss": 0.025, "step": 30462 }, { "epoch": 9.37, "learning_rate": 2.0794875597103358e-07, "loss": 0.0239, "step": 30463 }, { "epoch": 9.37, "learning_rate": 2.0774663383322123e-07, "loss": 0.0277, "step": 30464 }, { "epoch": 9.37, "learning_rate": 2.0754460894151008e-07, "loss": 0.0145, "step": 30465 }, { "epoch": 9.37, "learning_rate": 2.0734268129790513e-07, "loss": 0.015, "step": 30466 }, { "epoch": 9.37, "learning_rate": 2.071408509044126e-07, "loss": 0.062, "step": 30467 }, { "epoch": 9.37, "learning_rate": 2.0693911776303753e-07, "loss": 0.034, "step": 30468 }, { "epoch": 9.37, "learning_rate": 2.0673748187578279e-07, "loss": 0.0172, "step": 30469 }, { "epoch": 9.38, "learning_rate": 2.0653594324464897e-07, "loss": 0.0139, "step": 30470 }, { "epoch": 9.38, "learning_rate": 2.0633450187164007e-07, "loss": 0.0209, "step": 30471 }, { "epoch": 9.38, "learning_rate": 2.0613315775875665e-07, "loss": 0.0192, "step": 30472 }, { "epoch": 9.38, "learning_rate": 2.05931910907996e-07, "loss": 0.0098, "step": 30473 }, { "epoch": 9.38, "learning_rate": 2.0573076132135883e-07, "loss": 0.0266, "step": 30474 }, { "epoch": 9.38, "learning_rate": 2.0552970900084012e-07, "loss": 0.0155, "step": 30475 }, { "epoch": 9.38, "learning_rate": 2.0532875394844053e-07, "loss": 0.0261, "step": 30476 }, { "epoch": 9.38, "learning_rate": 2.0512789616615181e-07, "loss": 0.0216, "step": 30477 }, { "epoch": 9.38, "learning_rate": 2.0492713565597123e-07, "loss": 0.0265, "step": 30478 }, { "epoch": 9.38, "learning_rate": 2.0472647241989164e-07, "loss": 0.019, "step": 30479 }, { "epoch": 9.38, "learning_rate": 2.045259064599059e-07, "loss": 0.0109, "step": 30480 }, { "epoch": 9.38, "learning_rate": 2.0432543777800463e-07, "loss": 0.0247, "step": 30481 }, { "epoch": 9.38, "learning_rate": 2.0412506637617957e-07, "loss": 0.0174, "step": 30482 }, { "epoch": 9.38, "learning_rate": 2.0392479225642248e-07, "loss": 0.028, "step": 30483 }, { "epoch": 9.38, "learning_rate": 2.0372461542071953e-07, "loss": 0.0159, "step": 30484 }, { "epoch": 9.38, "learning_rate": 2.0352453587105914e-07, "loss": 0.0123, "step": 30485 }, { "epoch": 9.38, "learning_rate": 2.0332455360943083e-07, "loss": 0.0106, "step": 30486 }, { "epoch": 9.38, "learning_rate": 2.031246686378163e-07, "loss": 0.0289, "step": 30487 }, { "epoch": 9.38, "learning_rate": 2.0292488095820405e-07, "loss": 0.035, "step": 30488 }, { "epoch": 9.38, "learning_rate": 2.0272519057257578e-07, "loss": 0.0208, "step": 30489 }, { "epoch": 9.38, "learning_rate": 2.0252559748291767e-07, "loss": 0.0451, "step": 30490 }, { "epoch": 9.38, "learning_rate": 2.0232610169120814e-07, "loss": 0.0278, "step": 30491 }, { "epoch": 9.38, "learning_rate": 2.0212670319943118e-07, "loss": 0.0197, "step": 30492 }, { "epoch": 9.38, "learning_rate": 2.019274020095663e-07, "loss": 0.0189, "step": 30493 }, { "epoch": 9.38, "learning_rate": 2.017281981235919e-07, "loss": 0.027, "step": 30494 }, { "epoch": 9.38, "learning_rate": 2.0152909154348643e-07, "loss": 0.021, "step": 30495 }, { "epoch": 9.38, "learning_rate": 2.013300822712283e-07, "loss": 0.0304, "step": 30496 }, { "epoch": 9.38, "learning_rate": 2.011311703087937e-07, "loss": 0.0274, "step": 30497 }, { "epoch": 9.38, "learning_rate": 2.009323556581566e-07, "loss": 0.0186, "step": 30498 }, { "epoch": 9.38, "learning_rate": 2.0073363832129322e-07, "loss": 0.0298, "step": 30499 }, { "epoch": 9.38, "learning_rate": 2.005350183001753e-07, "loss": 0.0356, "step": 30500 }, { "epoch": 9.38, "learning_rate": 2.0033649559677682e-07, "loss": 0.0247, "step": 30501 }, { "epoch": 9.38, "learning_rate": 2.001380702130673e-07, "loss": 0.0334, "step": 30502 }, { "epoch": 9.39, "learning_rate": 1.9993974215101853e-07, "loss": 0.0186, "step": 30503 }, { "epoch": 9.39, "learning_rate": 1.9974151141260223e-07, "loss": 0.0301, "step": 30504 }, { "epoch": 9.39, "learning_rate": 1.9954337799978242e-07, "loss": 0.0236, "step": 30505 }, { "epoch": 9.39, "learning_rate": 1.9934534191453081e-07, "loss": 0.0439, "step": 30506 }, { "epoch": 9.39, "learning_rate": 1.9914740315881255e-07, "loss": 0.0376, "step": 30507 }, { "epoch": 9.39, "learning_rate": 1.9894956173459267e-07, "loss": 0.0257, "step": 30508 }, { "epoch": 9.39, "learning_rate": 1.9875181764383634e-07, "loss": 0.0298, "step": 30509 }, { "epoch": 9.39, "learning_rate": 1.9855417088850858e-07, "loss": 0.0184, "step": 30510 }, { "epoch": 9.39, "learning_rate": 1.9835662147057012e-07, "loss": 0.0273, "step": 30511 }, { "epoch": 9.39, "learning_rate": 1.9815916939198377e-07, "loss": 0.0222, "step": 30512 }, { "epoch": 9.39, "learning_rate": 1.979618146547113e-07, "loss": 0.0244, "step": 30513 }, { "epoch": 9.39, "learning_rate": 1.977645572607123e-07, "loss": 0.0266, "step": 30514 }, { "epoch": 9.39, "learning_rate": 1.9756739721194408e-07, "loss": 0.0267, "step": 30515 }, { "epoch": 9.39, "learning_rate": 1.9737033451036614e-07, "loss": 0.0205, "step": 30516 }, { "epoch": 9.39, "learning_rate": 1.9717336915793472e-07, "loss": 0.0263, "step": 30517 }, { "epoch": 9.39, "learning_rate": 1.9697650115660715e-07, "loss": 0.0227, "step": 30518 }, { "epoch": 9.39, "learning_rate": 1.9677973050833742e-07, "loss": 0.0272, "step": 30519 }, { "epoch": 9.39, "learning_rate": 1.9658305721507842e-07, "loss": 0.0253, "step": 30520 }, { "epoch": 9.39, "learning_rate": 1.9638648127878747e-07, "loss": 0.013, "step": 30521 }, { "epoch": 9.39, "learning_rate": 1.9619000270141186e-07, "loss": 0.0226, "step": 30522 }, { "epoch": 9.39, "learning_rate": 1.9599362148490564e-07, "loss": 0.0181, "step": 30523 }, { "epoch": 9.39, "learning_rate": 1.9579733763121943e-07, "loss": 0.0344, "step": 30524 }, { "epoch": 9.39, "learning_rate": 1.956011511423006e-07, "loss": 0.0287, "step": 30525 }, { "epoch": 9.39, "learning_rate": 1.9540506202009756e-07, "loss": 0.0191, "step": 30526 }, { "epoch": 9.39, "learning_rate": 1.952090702665599e-07, "loss": 0.0252, "step": 30527 }, { "epoch": 9.39, "learning_rate": 1.9501317588363155e-07, "loss": 0.0422, "step": 30528 }, { "epoch": 9.39, "learning_rate": 1.948173788732599e-07, "loss": 0.0148, "step": 30529 }, { "epoch": 9.39, "learning_rate": 1.9462167923738673e-07, "loss": 0.0277, "step": 30530 }, { "epoch": 9.39, "learning_rate": 1.9442607697795934e-07, "loss": 0.0386, "step": 30531 }, { "epoch": 9.39, "learning_rate": 1.942305720969162e-07, "loss": 0.0213, "step": 30532 }, { "epoch": 9.39, "learning_rate": 1.940351645962013e-07, "loss": 0.0333, "step": 30533 }, { "epoch": 9.39, "learning_rate": 1.938398544777542e-07, "loss": 0.0188, "step": 30534 }, { "epoch": 9.4, "learning_rate": 1.936446417435156e-07, "loss": 0.0248, "step": 30535 }, { "epoch": 9.4, "learning_rate": 1.9344952639542391e-07, "loss": 0.0378, "step": 30536 }, { "epoch": 9.4, "learning_rate": 1.932545084354154e-07, "loss": 0.0075, "step": 30537 }, { "epoch": 9.4, "learning_rate": 1.9305958786542954e-07, "loss": 0.0193, "step": 30538 }, { "epoch": 9.4, "learning_rate": 1.9286476468739934e-07, "loss": 0.0409, "step": 30539 }, { "epoch": 9.4, "learning_rate": 1.9267003890325875e-07, "loss": 0.024, "step": 30540 }, { "epoch": 9.4, "learning_rate": 1.9247541051494512e-07, "loss": 0.022, "step": 30541 }, { "epoch": 9.4, "learning_rate": 1.9228087952439022e-07, "loss": 0.0256, "step": 30542 }, { "epoch": 9.4, "learning_rate": 1.9208644593352366e-07, "loss": 0.0155, "step": 30543 }, { "epoch": 9.4, "learning_rate": 1.9189210974427941e-07, "loss": 0.034, "step": 30544 }, { "epoch": 9.4, "learning_rate": 1.9169787095858484e-07, "loss": 0.0206, "step": 30545 }, { "epoch": 9.4, "learning_rate": 1.915037295783706e-07, "loss": 0.0206, "step": 30546 }, { "epoch": 9.4, "learning_rate": 1.9130968560556294e-07, "loss": 0.023, "step": 30547 }, { "epoch": 9.4, "learning_rate": 1.9111573904209257e-07, "loss": 0.0215, "step": 30548 }, { "epoch": 9.4, "learning_rate": 1.9092188988988126e-07, "loss": 0.0174, "step": 30549 }, { "epoch": 9.4, "learning_rate": 1.9072813815085523e-07, "loss": 0.0172, "step": 30550 }, { "epoch": 9.4, "learning_rate": 1.9053448382694073e-07, "loss": 0.019, "step": 30551 }, { "epoch": 9.4, "learning_rate": 1.9034092692005846e-07, "loss": 0.0252, "step": 30552 }, { "epoch": 9.4, "learning_rate": 1.9014746743213243e-07, "loss": 0.0182, "step": 30553 }, { "epoch": 9.4, "learning_rate": 1.8995410536508218e-07, "loss": 0.0251, "step": 30554 }, { "epoch": 9.4, "learning_rate": 1.8976084072082957e-07, "loss": 0.0213, "step": 30555 }, { "epoch": 9.4, "learning_rate": 1.8956767350129302e-07, "loss": 0.0108, "step": 30556 }, { "epoch": 9.4, "learning_rate": 1.89374603708391e-07, "loss": 0.0193, "step": 30557 }, { "epoch": 9.4, "learning_rate": 1.8918163134404087e-07, "loss": 0.0313, "step": 30558 }, { "epoch": 9.4, "learning_rate": 1.8898875641016002e-07, "loss": 0.0267, "step": 30559 }, { "epoch": 9.4, "learning_rate": 1.887959789086613e-07, "loss": 0.0286, "step": 30560 }, { "epoch": 9.4, "learning_rate": 1.8860329884146212e-07, "loss": 0.0233, "step": 30561 }, { "epoch": 9.4, "learning_rate": 1.8841071621047314e-07, "loss": 0.0222, "step": 30562 }, { "epoch": 9.4, "learning_rate": 1.8821823101760949e-07, "loss": 0.0423, "step": 30563 }, { "epoch": 9.4, "learning_rate": 1.880258432647819e-07, "loss": 0.0274, "step": 30564 }, { "epoch": 9.4, "learning_rate": 1.8783355295389995e-07, "loss": 0.0235, "step": 30565 }, { "epoch": 9.4, "learning_rate": 1.8764136008687428e-07, "loss": 0.023, "step": 30566 }, { "epoch": 9.4, "learning_rate": 1.874492646656123e-07, "loss": 0.0242, "step": 30567 }, { "epoch": 9.41, "learning_rate": 1.872572666920236e-07, "loss": 0.0173, "step": 30568 }, { "epoch": 9.41, "learning_rate": 1.870653661680144e-07, "loss": 0.0198, "step": 30569 }, { "epoch": 9.41, "learning_rate": 1.8687356309548876e-07, "loss": 0.0346, "step": 30570 }, { "epoch": 9.41, "learning_rate": 1.8668185747635403e-07, "loss": 0.0296, "step": 30571 }, { "epoch": 9.41, "learning_rate": 1.8649024931251204e-07, "loss": 0.0243, "step": 30572 }, { "epoch": 9.41, "learning_rate": 1.8629873860586567e-07, "loss": 0.0192, "step": 30573 }, { "epoch": 9.41, "learning_rate": 1.861073253583179e-07, "loss": 0.0272, "step": 30574 }, { "epoch": 9.41, "learning_rate": 1.8591600957176936e-07, "loss": 0.0199, "step": 30575 }, { "epoch": 9.41, "learning_rate": 1.857247912481197e-07, "loss": 0.03, "step": 30576 }, { "epoch": 9.41, "learning_rate": 1.855336703892674e-07, "loss": 0.0271, "step": 30577 }, { "epoch": 9.41, "learning_rate": 1.8534264699711091e-07, "loss": 0.0216, "step": 30578 }, { "epoch": 9.41, "learning_rate": 1.8515172107354873e-07, "loss": 0.0245, "step": 30579 }, { "epoch": 9.41, "learning_rate": 1.849608926204749e-07, "loss": 0.0263, "step": 30580 }, { "epoch": 9.41, "learning_rate": 1.8477016163978456e-07, "loss": 0.0241, "step": 30581 }, { "epoch": 9.41, "learning_rate": 1.8457952813337287e-07, "loss": 0.058, "step": 30582 }, { "epoch": 9.41, "learning_rate": 1.8438899210313388e-07, "loss": 0.0245, "step": 30583 }, { "epoch": 9.41, "learning_rate": 1.8419855355095717e-07, "loss": 0.0241, "step": 30584 }, { "epoch": 9.41, "learning_rate": 1.8400821247873457e-07, "loss": 0.02, "step": 30585 }, { "epoch": 9.41, "learning_rate": 1.83817968888359e-07, "loss": 0.0347, "step": 30586 }, { "epoch": 9.41, "learning_rate": 1.8362782278171675e-07, "loss": 0.0195, "step": 30587 }, { "epoch": 9.41, "learning_rate": 1.834377741606974e-07, "loss": 0.0116, "step": 30588 }, { "epoch": 9.41, "learning_rate": 1.8324782302718835e-07, "loss": 0.0186, "step": 30589 }, { "epoch": 9.41, "learning_rate": 1.8305796938307586e-07, "loss": 0.0253, "step": 30590 }, { "epoch": 9.41, "learning_rate": 1.8286821323024507e-07, "loss": 0.0099, "step": 30591 }, { "epoch": 9.41, "learning_rate": 1.8267855457058004e-07, "loss": 0.0302, "step": 30592 }, { "epoch": 9.41, "learning_rate": 1.8248899340596705e-07, "loss": 0.0185, "step": 30593 }, { "epoch": 9.41, "learning_rate": 1.8229952973828458e-07, "loss": 0.0211, "step": 30594 }, { "epoch": 9.41, "learning_rate": 1.8211016356941559e-07, "loss": 0.0272, "step": 30595 }, { "epoch": 9.41, "learning_rate": 1.8192089490124188e-07, "loss": 0.0166, "step": 30596 }, { "epoch": 9.41, "learning_rate": 1.8173172373564306e-07, "loss": 0.0228, "step": 30597 }, { "epoch": 9.41, "learning_rate": 1.8154265007449657e-07, "loss": 0.0248, "step": 30598 }, { "epoch": 9.41, "learning_rate": 1.8135367391967974e-07, "loss": 0.0231, "step": 30599 }, { "epoch": 9.42, "learning_rate": 1.8116479527307108e-07, "loss": 0.0264, "step": 30600 }, { "epoch": 9.42, "learning_rate": 1.8097601413654464e-07, "loss": 0.0208, "step": 30601 }, { "epoch": 9.42, "learning_rate": 1.8078733051197561e-07, "loss": 0.0208, "step": 30602 }, { "epoch": 9.42, "learning_rate": 1.8059874440123804e-07, "loss": 0.0291, "step": 30603 }, { "epoch": 9.42, "learning_rate": 1.8041025580620597e-07, "loss": 0.0148, "step": 30604 }, { "epoch": 9.42, "learning_rate": 1.8022186472874903e-07, "loss": 0.0358, "step": 30605 }, { "epoch": 9.42, "learning_rate": 1.8003357117073904e-07, "loss": 0.0111, "step": 30606 }, { "epoch": 9.42, "learning_rate": 1.7984537513404677e-07, "loss": 0.0343, "step": 30607 }, { "epoch": 9.42, "learning_rate": 1.7965727662054068e-07, "loss": 0.022, "step": 30608 }, { "epoch": 9.42, "learning_rate": 1.7946927563208706e-07, "loss": 0.0313, "step": 30609 }, { "epoch": 9.42, "learning_rate": 1.792813721705555e-07, "loss": 0.0202, "step": 30610 }, { "epoch": 9.42, "learning_rate": 1.7909356623781127e-07, "loss": 0.0209, "step": 30611 }, { "epoch": 9.42, "learning_rate": 1.7890585783571723e-07, "loss": 0.0286, "step": 30612 }, { "epoch": 9.42, "learning_rate": 1.787182469661408e-07, "loss": 0.0237, "step": 30613 }, { "epoch": 9.42, "learning_rate": 1.7853073363094498e-07, "loss": 0.0301, "step": 30614 }, { "epoch": 9.42, "learning_rate": 1.7834331783198933e-07, "loss": 0.0228, "step": 30615 }, { "epoch": 9.42, "learning_rate": 1.7815599957113573e-07, "loss": 0.0193, "step": 30616 }, { "epoch": 9.42, "learning_rate": 1.77968778850246e-07, "loss": 0.0115, "step": 30617 }, { "epoch": 9.42, "learning_rate": 1.7778165567117865e-07, "loss": 0.0194, "step": 30618 }, { "epoch": 9.42, "learning_rate": 1.7759463003579225e-07, "loss": 0.0246, "step": 30619 }, { "epoch": 9.42, "learning_rate": 1.7740770194594193e-07, "loss": 0.0137, "step": 30620 }, { "epoch": 9.42, "learning_rate": 1.7722087140348844e-07, "loss": 0.0222, "step": 30621 }, { "epoch": 9.42, "learning_rate": 1.770341384102825e-07, "loss": 0.028, "step": 30622 }, { "epoch": 9.42, "learning_rate": 1.7684750296818154e-07, "loss": 0.028, "step": 30623 }, { "epoch": 9.42, "learning_rate": 1.7666096507903964e-07, "loss": 0.0136, "step": 30624 }, { "epoch": 9.42, "learning_rate": 1.7647452474470527e-07, "loss": 0.0288, "step": 30625 }, { "epoch": 9.42, "learning_rate": 1.7628818196703368e-07, "loss": 0.0274, "step": 30626 }, { "epoch": 9.42, "learning_rate": 1.7610193674787446e-07, "loss": 0.0189, "step": 30627 }, { "epoch": 9.42, "learning_rate": 1.7591578908907724e-07, "loss": 0.0191, "step": 30628 }, { "epoch": 9.42, "learning_rate": 1.7572973899248947e-07, "loss": 0.0213, "step": 30629 }, { "epoch": 9.42, "learning_rate": 1.755437864599596e-07, "loss": 0.0276, "step": 30630 }, { "epoch": 9.42, "learning_rate": 1.7535793149333514e-07, "loss": 0.0369, "step": 30631 }, { "epoch": 9.42, "learning_rate": 1.7517217409446008e-07, "loss": 0.0204, "step": 30632 }, { "epoch": 9.43, "learning_rate": 1.7498651426518075e-07, "loss": 0.0307, "step": 30633 }, { "epoch": 9.43, "learning_rate": 1.7480095200734125e-07, "loss": 0.0357, "step": 30634 }, { "epoch": 9.43, "learning_rate": 1.7461548732278234e-07, "loss": 0.0228, "step": 30635 }, { "epoch": 9.43, "learning_rate": 1.7443012021334803e-07, "loss": 0.017, "step": 30636 }, { "epoch": 9.43, "learning_rate": 1.742448506808758e-07, "loss": 0.0224, "step": 30637 }, { "epoch": 9.43, "learning_rate": 1.7405967872721085e-07, "loss": 0.0146, "step": 30638 }, { "epoch": 9.43, "learning_rate": 1.7387460435418834e-07, "loss": 0.028, "step": 30639 }, { "epoch": 9.43, "learning_rate": 1.7368962756364573e-07, "loss": 0.0131, "step": 30640 }, { "epoch": 9.43, "learning_rate": 1.735047483574215e-07, "loss": 0.0233, "step": 30641 }, { "epoch": 9.43, "learning_rate": 1.7331996673735198e-07, "loss": 0.0321, "step": 30642 }, { "epoch": 9.43, "learning_rate": 1.7313528270527236e-07, "loss": 0.0146, "step": 30643 }, { "epoch": 9.43, "learning_rate": 1.7295069626301565e-07, "loss": 0.0247, "step": 30644 }, { "epoch": 9.43, "learning_rate": 1.727662074124148e-07, "loss": 0.022, "step": 30645 }, { "epoch": 9.43, "learning_rate": 1.7258181615530389e-07, "loss": 0.0222, "step": 30646 }, { "epoch": 9.43, "learning_rate": 1.7239752249351037e-07, "loss": 0.0231, "step": 30647 }, { "epoch": 9.43, "learning_rate": 1.7221332642886946e-07, "loss": 0.0242, "step": 30648 }, { "epoch": 9.43, "learning_rate": 1.720292279632063e-07, "loss": 0.0158, "step": 30649 }, { "epoch": 9.43, "learning_rate": 1.718452270983506e-07, "loss": 0.0134, "step": 30650 }, { "epoch": 9.43, "learning_rate": 1.716613238361309e-07, "loss": 0.019, "step": 30651 }, { "epoch": 9.43, "learning_rate": 1.714775181783712e-07, "loss": 0.0246, "step": 30652 }, { "epoch": 9.43, "learning_rate": 1.7129381012689906e-07, "loss": 0.0372, "step": 30653 }, { "epoch": 9.43, "learning_rate": 1.7111019968353625e-07, "loss": 0.0137, "step": 30654 }, { "epoch": 9.43, "learning_rate": 1.7092668685010916e-07, "loss": 0.0234, "step": 30655 }, { "epoch": 9.43, "learning_rate": 1.707432716284385e-07, "loss": 0.0271, "step": 30656 }, { "epoch": 9.43, "learning_rate": 1.7055995402034507e-07, "loss": 0.0217, "step": 30657 }, { "epoch": 9.43, "learning_rate": 1.7037673402765076e-07, "loss": 0.016, "step": 30658 }, { "epoch": 9.43, "learning_rate": 1.7019361165217518e-07, "loss": 0.0177, "step": 30659 }, { "epoch": 9.43, "learning_rate": 1.7001058689573579e-07, "loss": 0.0202, "step": 30660 }, { "epoch": 9.43, "learning_rate": 1.6982765976015226e-07, "loss": 0.0144, "step": 30661 }, { "epoch": 9.43, "learning_rate": 1.6964483024723865e-07, "loss": 0.03, "step": 30662 }, { "epoch": 9.43, "learning_rate": 1.6946209835881244e-07, "loss": 0.0306, "step": 30663 }, { "epoch": 9.43, "learning_rate": 1.6927946409668662e-07, "loss": 0.0235, "step": 30664 }, { "epoch": 9.44, "learning_rate": 1.6909692746267748e-07, "loss": 0.0303, "step": 30665 }, { "epoch": 9.44, "learning_rate": 1.689144884585947e-07, "loss": 0.0238, "step": 30666 }, { "epoch": 9.44, "learning_rate": 1.687321470862524e-07, "loss": 0.0211, "step": 30667 }, { "epoch": 9.44, "learning_rate": 1.6854990334746025e-07, "loss": 0.0142, "step": 30668 }, { "epoch": 9.44, "learning_rate": 1.6836775724402898e-07, "loss": 0.0105, "step": 30669 }, { "epoch": 9.44, "learning_rate": 1.681857087777672e-07, "loss": 0.0201, "step": 30670 }, { "epoch": 9.44, "learning_rate": 1.680037579504823e-07, "loss": 0.0246, "step": 30671 }, { "epoch": 9.44, "learning_rate": 1.6782190476398174e-07, "loss": 0.0123, "step": 30672 }, { "epoch": 9.44, "learning_rate": 1.6764014922007078e-07, "loss": 0.0291, "step": 30673 }, { "epoch": 9.44, "learning_rate": 1.674584913205546e-07, "loss": 0.0324, "step": 30674 }, { "epoch": 9.44, "learning_rate": 1.672769310672384e-07, "loss": 0.01, "step": 30675 }, { "epoch": 9.44, "learning_rate": 1.6709546846192527e-07, "loss": 0.0273, "step": 30676 }, { "epoch": 9.44, "learning_rate": 1.669141035064137e-07, "loss": 0.0285, "step": 30677 }, { "epoch": 9.44, "learning_rate": 1.6673283620251002e-07, "loss": 0.0279, "step": 30678 }, { "epoch": 9.44, "learning_rate": 1.665516665520106e-07, "loss": 0.0221, "step": 30679 }, { "epoch": 9.44, "learning_rate": 1.6637059455671623e-07, "loss": 0.0317, "step": 30680 }, { "epoch": 9.44, "learning_rate": 1.6618962021842434e-07, "loss": 0.0273, "step": 30681 }, { "epoch": 9.44, "learning_rate": 1.6600874353893237e-07, "loss": 0.0325, "step": 30682 }, { "epoch": 9.44, "learning_rate": 1.6582796452003892e-07, "loss": 0.0291, "step": 30683 }, { "epoch": 9.44, "learning_rate": 1.6564728316353474e-07, "loss": 0.0244, "step": 30684 }, { "epoch": 9.44, "learning_rate": 1.654666994712184e-07, "loss": 0.0297, "step": 30685 }, { "epoch": 9.44, "learning_rate": 1.6528621344488181e-07, "loss": 0.0211, "step": 30686 }, { "epoch": 9.44, "learning_rate": 1.6510582508631467e-07, "loss": 0.0128, "step": 30687 }, { "epoch": 9.44, "learning_rate": 1.649255343973122e-07, "loss": 0.0214, "step": 30688 }, { "epoch": 9.44, "learning_rate": 1.6474534137966402e-07, "loss": 0.016, "step": 30689 }, { "epoch": 9.44, "learning_rate": 1.6456524603515879e-07, "loss": 0.0177, "step": 30690 }, { "epoch": 9.44, "learning_rate": 1.64385248365585e-07, "loss": 0.0132, "step": 30691 }, { "epoch": 9.44, "learning_rate": 1.6420534837272907e-07, "loss": 0.0041, "step": 30692 }, { "epoch": 9.44, "learning_rate": 1.6402554605838173e-07, "loss": 0.0185, "step": 30693 }, { "epoch": 9.44, "learning_rate": 1.638458414243238e-07, "loss": 0.0152, "step": 30694 }, { "epoch": 9.44, "learning_rate": 1.6366623447234277e-07, "loss": 0.0155, "step": 30695 }, { "epoch": 9.44, "learning_rate": 1.634867252042216e-07, "loss": 0.0204, "step": 30696 }, { "epoch": 9.44, "learning_rate": 1.6330731362174224e-07, "loss": 0.0097, "step": 30697 }, { "epoch": 9.45, "learning_rate": 1.6312799972668659e-07, "loss": 0.0239, "step": 30698 }, { "epoch": 9.45, "learning_rate": 1.6294878352083653e-07, "loss": 0.0404, "step": 30699 }, { "epoch": 9.45, "learning_rate": 1.627696650059718e-07, "loss": 0.041, "step": 30700 }, { "epoch": 9.45, "learning_rate": 1.6259064418386982e-07, "loss": 0.0277, "step": 30701 }, { "epoch": 9.45, "learning_rate": 1.6241172105630808e-07, "loss": 0.0221, "step": 30702 }, { "epoch": 9.45, "learning_rate": 1.622328956250674e-07, "loss": 0.0288, "step": 30703 }, { "epoch": 9.45, "learning_rate": 1.6205416789191852e-07, "loss": 0.0288, "step": 30704 }, { "epoch": 9.45, "learning_rate": 1.6187553785863897e-07, "loss": 0.0214, "step": 30705 }, { "epoch": 9.45, "learning_rate": 1.6169700552700284e-07, "loss": 0.0409, "step": 30706 }, { "epoch": 9.45, "learning_rate": 1.615185708987821e-07, "loss": 0.0155, "step": 30707 }, { "epoch": 9.45, "learning_rate": 1.6134023397575084e-07, "loss": 0.0177, "step": 30708 }, { "epoch": 9.45, "learning_rate": 1.6116199475967654e-07, "loss": 0.0292, "step": 30709 }, { "epoch": 9.45, "learning_rate": 1.6098385325233334e-07, "loss": 0.0075, "step": 30710 }, { "epoch": 9.45, "learning_rate": 1.6080580945548764e-07, "loss": 0.0227, "step": 30711 }, { "epoch": 9.45, "learning_rate": 1.6062786337090687e-07, "loss": 0.0288, "step": 30712 }, { "epoch": 9.45, "learning_rate": 1.6045001500036072e-07, "loss": 0.0182, "step": 30713 }, { "epoch": 9.45, "learning_rate": 1.602722643456156e-07, "loss": 0.0197, "step": 30714 }, { "epoch": 9.45, "learning_rate": 1.6009461140843452e-07, "loss": 0.0149, "step": 30715 }, { "epoch": 9.45, "learning_rate": 1.5991705619058163e-07, "loss": 0.009, "step": 30716 }, { "epoch": 9.45, "learning_rate": 1.5973959869382327e-07, "loss": 0.0266, "step": 30717 }, { "epoch": 9.45, "learning_rate": 1.5956223891991807e-07, "loss": 0.0296, "step": 30718 }, { "epoch": 9.45, "learning_rate": 1.5938497687062905e-07, "loss": 0.0129, "step": 30719 }, { "epoch": 9.45, "learning_rate": 1.59207812547717e-07, "loss": 0.0162, "step": 30720 }, { "epoch": 9.45, "learning_rate": 1.5903074595294165e-07, "loss": 0.0268, "step": 30721 }, { "epoch": 9.45, "learning_rate": 1.588537770880605e-07, "loss": 0.0364, "step": 30722 }, { "epoch": 9.45, "learning_rate": 1.5867690595482987e-07, "loss": 0.0174, "step": 30723 }, { "epoch": 9.45, "learning_rate": 1.5850013255500952e-07, "loss": 0.0258, "step": 30724 }, { "epoch": 9.45, "learning_rate": 1.5832345689035134e-07, "loss": 0.0257, "step": 30725 }, { "epoch": 9.45, "learning_rate": 1.5814687896261282e-07, "loss": 0.0239, "step": 30726 }, { "epoch": 9.45, "learning_rate": 1.5797039877354592e-07, "loss": 0.0334, "step": 30727 }, { "epoch": 9.45, "learning_rate": 1.5779401632490365e-07, "loss": 0.0278, "step": 30728 }, { "epoch": 9.45, "learning_rate": 1.5761773161843687e-07, "loss": 0.0215, "step": 30729 }, { "epoch": 9.46, "learning_rate": 1.5744154465589745e-07, "loss": 0.0253, "step": 30730 }, { "epoch": 9.46, "learning_rate": 1.5726545543903627e-07, "loss": 0.0285, "step": 30731 }, { "epoch": 9.46, "learning_rate": 1.5708946396959856e-07, "loss": 0.0134, "step": 30732 }, { "epoch": 9.46, "learning_rate": 1.5691357024933407e-07, "loss": 0.0209, "step": 30733 }, { "epoch": 9.46, "learning_rate": 1.5673777427999027e-07, "loss": 0.018, "step": 30734 }, { "epoch": 9.46, "learning_rate": 1.565620760633113e-07, "loss": 0.0281, "step": 30735 }, { "epoch": 9.46, "learning_rate": 1.5638647560104358e-07, "loss": 0.013, "step": 30736 }, { "epoch": 9.46, "learning_rate": 1.5621097289493015e-07, "loss": 0.0373, "step": 30737 }, { "epoch": 9.46, "learning_rate": 1.5603556794671403e-07, "loss": 0.0376, "step": 30738 }, { "epoch": 9.46, "learning_rate": 1.558602607581372e-07, "loss": 0.0277, "step": 30739 }, { "epoch": 9.46, "learning_rate": 1.5568505133094046e-07, "loss": 0.0139, "step": 30740 }, { "epoch": 9.46, "learning_rate": 1.5550993966686468e-07, "loss": 0.0129, "step": 30741 }, { "epoch": 9.46, "learning_rate": 1.553349257676473e-07, "loss": 0.0181, "step": 30742 }, { "epoch": 9.46, "learning_rate": 1.5516000963502698e-07, "loss": 0.0167, "step": 30743 }, { "epoch": 9.46, "learning_rate": 1.549851912707423e-07, "loss": 0.0162, "step": 30744 }, { "epoch": 9.46, "learning_rate": 1.5481047067652744e-07, "loss": 0.0263, "step": 30745 }, { "epoch": 9.46, "learning_rate": 1.5463584785411767e-07, "loss": 0.0215, "step": 30746 }, { "epoch": 9.46, "learning_rate": 1.5446132280524717e-07, "loss": 0.0116, "step": 30747 }, { "epoch": 9.46, "learning_rate": 1.5428689553165233e-07, "loss": 0.0282, "step": 30748 }, { "epoch": 9.46, "learning_rate": 1.5411256603505953e-07, "loss": 0.032, "step": 30749 }, { "epoch": 9.46, "learning_rate": 1.5393833431720516e-07, "loss": 0.0385, "step": 30750 }, { "epoch": 9.46, "learning_rate": 1.5376420037981676e-07, "loss": 0.0199, "step": 30751 }, { "epoch": 9.46, "learning_rate": 1.5359016422462514e-07, "loss": 0.0218, "step": 30752 }, { "epoch": 9.46, "learning_rate": 1.5341622585335668e-07, "loss": 0.0097, "step": 30753 }, { "epoch": 9.46, "learning_rate": 1.5324238526774006e-07, "loss": 0.0291, "step": 30754 }, { "epoch": 9.46, "learning_rate": 1.530686424695038e-07, "loss": 0.0292, "step": 30755 }, { "epoch": 9.46, "learning_rate": 1.5289499746036885e-07, "loss": 0.0205, "step": 30756 }, { "epoch": 9.46, "learning_rate": 1.5272145024206264e-07, "loss": 0.0328, "step": 30757 }, { "epoch": 9.46, "learning_rate": 1.5254800081630828e-07, "loss": 0.0192, "step": 30758 }, { "epoch": 9.46, "learning_rate": 1.5237464918482768e-07, "loss": 0.0265, "step": 30759 }, { "epoch": 9.46, "learning_rate": 1.5220139534934175e-07, "loss": 0.0178, "step": 30760 }, { "epoch": 9.46, "learning_rate": 1.520282393115735e-07, "loss": 0.0249, "step": 30761 }, { "epoch": 9.46, "learning_rate": 1.5185518107324048e-07, "loss": 0.0136, "step": 30762 }, { "epoch": 9.47, "learning_rate": 1.516822206360613e-07, "loss": 0.0154, "step": 30763 }, { "epoch": 9.47, "learning_rate": 1.5150935800175348e-07, "loss": 0.0447, "step": 30764 }, { "epoch": 9.47, "learning_rate": 1.5133659317203674e-07, "loss": 0.0431, "step": 30765 }, { "epoch": 9.47, "learning_rate": 1.5116392614862309e-07, "loss": 0.0127, "step": 30766 }, { "epoch": 9.47, "learning_rate": 1.5099135693322776e-07, "loss": 0.0204, "step": 30767 }, { "epoch": 9.47, "learning_rate": 1.5081888552756608e-07, "loss": 0.016, "step": 30768 }, { "epoch": 9.47, "learning_rate": 1.5064651193335112e-07, "loss": 0.0273, "step": 30769 }, { "epoch": 9.47, "learning_rate": 1.5047423615229262e-07, "loss": 0.0135, "step": 30770 }, { "epoch": 9.47, "learning_rate": 1.5030205818610255e-07, "loss": 0.0135, "step": 30771 }, { "epoch": 9.47, "learning_rate": 1.5012997803649178e-07, "loss": 0.0175, "step": 30772 }, { "epoch": 9.47, "learning_rate": 1.4995799570516666e-07, "loss": 0.0184, "step": 30773 }, { "epoch": 9.47, "learning_rate": 1.4978611119383703e-07, "loss": 0.0217, "step": 30774 }, { "epoch": 9.47, "learning_rate": 1.4961432450421033e-07, "loss": 0.0213, "step": 30775 }, { "epoch": 9.47, "learning_rate": 1.4944263563799188e-07, "loss": 0.0248, "step": 30776 }, { "epoch": 9.47, "learning_rate": 1.492710445968859e-07, "loss": 0.0208, "step": 30777 }, { "epoch": 9.47, "learning_rate": 1.4909955138259768e-07, "loss": 0.0163, "step": 30778 }, { "epoch": 9.47, "learning_rate": 1.489281559968292e-07, "loss": 0.0227, "step": 30779 }, { "epoch": 9.47, "learning_rate": 1.4875685844128352e-07, "loss": 0.0297, "step": 30780 }, { "epoch": 9.47, "learning_rate": 1.485856587176615e-07, "loss": 0.0188, "step": 30781 }, { "epoch": 9.47, "learning_rate": 1.4841455682766292e-07, "loss": 0.0246, "step": 30782 }, { "epoch": 9.47, "learning_rate": 1.4824355277298753e-07, "loss": 0.0243, "step": 30783 }, { "epoch": 9.47, "learning_rate": 1.4807264655533282e-07, "loss": 0.014, "step": 30784 }, { "epoch": 9.47, "learning_rate": 1.4790183817639746e-07, "loss": 0.0152, "step": 30785 }, { "epoch": 9.47, "learning_rate": 1.477311276378768e-07, "loss": 0.0286, "step": 30786 }, { "epoch": 9.47, "learning_rate": 1.4756051494146496e-07, "loss": 0.0222, "step": 30787 }, { "epoch": 9.47, "learning_rate": 1.4739000008885839e-07, "loss": 0.0326, "step": 30788 }, { "epoch": 9.47, "learning_rate": 1.472195830817491e-07, "loss": 0.0371, "step": 30789 }, { "epoch": 9.47, "learning_rate": 1.4704926392183017e-07, "loss": 0.0394, "step": 30790 }, { "epoch": 9.47, "learning_rate": 1.4687904261079355e-07, "loss": 0.0265, "step": 30791 }, { "epoch": 9.47, "learning_rate": 1.4670891915032682e-07, "loss": 0.0241, "step": 30792 }, { "epoch": 9.47, "learning_rate": 1.4653889354212414e-07, "loss": 0.0205, "step": 30793 }, { "epoch": 9.47, "learning_rate": 1.4636896578786974e-07, "loss": 0.0126, "step": 30794 }, { "epoch": 9.48, "learning_rate": 1.461991358892534e-07, "loss": 0.0184, "step": 30795 }, { "epoch": 9.48, "learning_rate": 1.460294038479615e-07, "loss": 0.033, "step": 30796 }, { "epoch": 9.48, "learning_rate": 1.4585976966567826e-07, "loss": 0.0163, "step": 30797 }, { "epoch": 9.48, "learning_rate": 1.4569023334409128e-07, "loss": 0.0482, "step": 30798 }, { "epoch": 9.48, "learning_rate": 1.4552079488488025e-07, "loss": 0.0259, "step": 30799 }, { "epoch": 9.48, "learning_rate": 1.4535145428973164e-07, "loss": 0.0176, "step": 30800 }, { "epoch": 9.48, "learning_rate": 1.4518221156032408e-07, "loss": 0.0151, "step": 30801 }, { "epoch": 9.48, "learning_rate": 1.450130666983396e-07, "loss": 0.0196, "step": 30802 }, { "epoch": 9.48, "learning_rate": 1.4484401970545903e-07, "loss": 0.0168, "step": 30803 }, { "epoch": 9.48, "learning_rate": 1.4467507058335882e-07, "loss": 0.0206, "step": 30804 }, { "epoch": 9.48, "learning_rate": 1.445062193337188e-07, "loss": 0.0297, "step": 30805 }, { "epoch": 9.48, "learning_rate": 1.4433746595821528e-07, "loss": 0.0171, "step": 30806 }, { "epoch": 9.48, "learning_rate": 1.441688104585237e-07, "loss": 0.0168, "step": 30807 }, { "epoch": 9.48, "learning_rate": 1.4400025283631936e-07, "loss": 0.0203, "step": 30808 }, { "epoch": 9.48, "learning_rate": 1.4383179309327533e-07, "loss": 0.0231, "step": 30809 }, { "epoch": 9.48, "learning_rate": 1.4366343123106697e-07, "loss": 0.0306, "step": 30810 }, { "epoch": 9.48, "learning_rate": 1.4349516725136404e-07, "loss": 0.0294, "step": 30811 }, { "epoch": 9.48, "learning_rate": 1.433270011558363e-07, "loss": 0.0426, "step": 30812 }, { "epoch": 9.48, "learning_rate": 1.4315893294615802e-07, "loss": 0.0398, "step": 30813 }, { "epoch": 9.48, "learning_rate": 1.429909626239945e-07, "loss": 0.0329, "step": 30814 }, { "epoch": 9.48, "learning_rate": 1.4282309019101548e-07, "loss": 0.0144, "step": 30815 }, { "epoch": 9.48, "learning_rate": 1.4265531564888858e-07, "loss": 0.0312, "step": 30816 }, { "epoch": 9.48, "learning_rate": 1.4248763899927797e-07, "loss": 0.0295, "step": 30817 }, { "epoch": 9.48, "learning_rate": 1.4232006024385126e-07, "loss": 0.0337, "step": 30818 }, { "epoch": 9.48, "learning_rate": 1.4215257938427042e-07, "loss": 0.0215, "step": 30819 }, { "epoch": 9.48, "learning_rate": 1.4198519642219966e-07, "loss": 0.028, "step": 30820 }, { "epoch": 9.48, "learning_rate": 1.418179113593021e-07, "loss": 0.0311, "step": 30821 }, { "epoch": 9.48, "learning_rate": 1.4165072419723758e-07, "loss": 0.0204, "step": 30822 }, { "epoch": 9.48, "learning_rate": 1.4148363493766803e-07, "loss": 0.0178, "step": 30823 }, { "epoch": 9.48, "learning_rate": 1.4131664358225105e-07, "loss": 0.0236, "step": 30824 }, { "epoch": 9.48, "learning_rate": 1.4114975013264643e-07, "loss": 0.0192, "step": 30825 }, { "epoch": 9.48, "learning_rate": 1.4098295459051059e-07, "loss": 0.0258, "step": 30826 }, { "epoch": 9.48, "learning_rate": 1.408162569575e-07, "loss": 0.0342, "step": 30827 }, { "epoch": 9.49, "learning_rate": 1.4064965723527114e-07, "loss": 0.0359, "step": 30828 }, { "epoch": 9.49, "learning_rate": 1.4048315542547707e-07, "loss": 0.0233, "step": 30829 }, { "epoch": 9.49, "learning_rate": 1.403167515297721e-07, "loss": 0.0216, "step": 30830 }, { "epoch": 9.49, "learning_rate": 1.4015044554981038e-07, "loss": 0.0203, "step": 30831 }, { "epoch": 9.49, "learning_rate": 1.3998423748723956e-07, "loss": 0.0239, "step": 30832 }, { "epoch": 9.49, "learning_rate": 1.398181273437138e-07, "loss": 0.0209, "step": 30833 }, { "epoch": 9.49, "learning_rate": 1.3965211512088182e-07, "loss": 0.0323, "step": 30834 }, { "epoch": 9.49, "learning_rate": 1.3948620082039121e-07, "loss": 0.0164, "step": 30835 }, { "epoch": 9.49, "learning_rate": 1.3932038444389063e-07, "loss": 0.0174, "step": 30836 }, { "epoch": 9.49, "learning_rate": 1.391546659930254e-07, "loss": 0.0144, "step": 30837 }, { "epoch": 9.49, "learning_rate": 1.3898904546944425e-07, "loss": 0.0225, "step": 30838 }, { "epoch": 9.49, "learning_rate": 1.3882352287478806e-07, "loss": 0.0289, "step": 30839 }, { "epoch": 9.49, "learning_rate": 1.3865809821070332e-07, "loss": 0.0284, "step": 30840 }, { "epoch": 9.49, "learning_rate": 1.384927714788331e-07, "loss": 0.0183, "step": 30841 }, { "epoch": 9.49, "learning_rate": 1.3832754268081726e-07, "loss": 0.0373, "step": 30842 }, { "epoch": 9.49, "learning_rate": 1.3816241181829891e-07, "loss": 0.0147, "step": 30843 }, { "epoch": 9.49, "learning_rate": 1.3799737889291565e-07, "loss": 0.022, "step": 30844 }, { "epoch": 9.49, "learning_rate": 1.3783244390630834e-07, "loss": 0.0238, "step": 30845 }, { "epoch": 9.49, "learning_rate": 1.376676068601124e-07, "loss": 0.025, "step": 30846 }, { "epoch": 9.49, "learning_rate": 1.3750286775596754e-07, "loss": 0.0266, "step": 30847 }, { "epoch": 9.49, "learning_rate": 1.3733822659551033e-07, "loss": 0.0354, "step": 30848 }, { "epoch": 9.49, "learning_rate": 1.3717368338037163e-07, "loss": 0.03, "step": 30849 }, { "epoch": 9.49, "learning_rate": 1.3700923811218903e-07, "loss": 0.0192, "step": 30850 }, { "epoch": 9.49, "learning_rate": 1.3684489079259565e-07, "loss": 0.0169, "step": 30851 }, { "epoch": 9.49, "learning_rate": 1.3668064142322136e-07, "loss": 0.0161, "step": 30852 }, { "epoch": 9.49, "learning_rate": 1.3651649000569922e-07, "loss": 0.017, "step": 30853 }, { "epoch": 9.49, "learning_rate": 1.3635243654165798e-07, "loss": 0.0192, "step": 30854 }, { "epoch": 9.49, "learning_rate": 1.3618848103272964e-07, "loss": 0.0147, "step": 30855 }, { "epoch": 9.49, "learning_rate": 1.3602462348053848e-07, "loss": 0.0156, "step": 30856 }, { "epoch": 9.49, "learning_rate": 1.3586086388671428e-07, "loss": 0.0066, "step": 30857 }, { "epoch": 9.49, "learning_rate": 1.3569720225288353e-07, "loss": 0.0324, "step": 30858 }, { "epoch": 9.49, "learning_rate": 1.3553363858066937e-07, "loss": 0.0194, "step": 30859 }, { "epoch": 9.5, "learning_rate": 1.353701728716994e-07, "loss": 0.0253, "step": 30860 }, { "epoch": 9.5, "learning_rate": 1.3520680512759343e-07, "loss": 0.0164, "step": 30861 }, { "epoch": 9.5, "learning_rate": 1.3504353534997682e-07, "loss": 0.0231, "step": 30862 }, { "epoch": 9.5, "learning_rate": 1.348803635404683e-07, "loss": 0.0477, "step": 30863 }, { "epoch": 9.5, "learning_rate": 1.3471728970068986e-07, "loss": 0.0225, "step": 30864 }, { "epoch": 9.5, "learning_rate": 1.3455431383226247e-07, "loss": 0.0134, "step": 30865 }, { "epoch": 9.5, "learning_rate": 1.343914359368015e-07, "loss": 0.0226, "step": 30866 }, { "epoch": 9.5, "learning_rate": 1.342286560159267e-07, "loss": 0.015, "step": 30867 }, { "epoch": 9.5, "learning_rate": 1.3406597407125354e-07, "loss": 0.0217, "step": 30868 }, { "epoch": 9.5, "learning_rate": 1.3390339010439847e-07, "loss": 0.0145, "step": 30869 }, { "epoch": 9.5, "learning_rate": 1.337409041169757e-07, "loss": 0.0327, "step": 30870 }, { "epoch": 9.5, "learning_rate": 1.3357851611059736e-07, "loss": 0.0268, "step": 30871 }, { "epoch": 9.5, "learning_rate": 1.3341622608687877e-07, "loss": 0.0198, "step": 30872 }, { "epoch": 9.5, "learning_rate": 1.3325403404743088e-07, "loss": 0.0157, "step": 30873 }, { "epoch": 9.5, "learning_rate": 1.3309193999386238e-07, "loss": 0.0224, "step": 30874 }, { "epoch": 9.5, "learning_rate": 1.3292994392778535e-07, "loss": 0.0172, "step": 30875 }, { "epoch": 9.5, "learning_rate": 1.3276804585080737e-07, "loss": 0.0262, "step": 30876 }, { "epoch": 9.5, "learning_rate": 1.3260624576453718e-07, "loss": 0.0327, "step": 30877 }, { "epoch": 9.5, "learning_rate": 1.3244454367058013e-07, "loss": 0.0156, "step": 30878 }, { "epoch": 9.5, "learning_rate": 1.3228293957054383e-07, "loss": 0.0235, "step": 30879 }, { "epoch": 9.5, "learning_rate": 1.3212143346603256e-07, "loss": 0.0078, "step": 30880 }, { "epoch": 9.5, "learning_rate": 1.3196002535864837e-07, "loss": 0.0464, "step": 30881 }, { "epoch": 9.5, "learning_rate": 1.3179871524999778e-07, "loss": 0.0181, "step": 30882 }, { "epoch": 9.5, "learning_rate": 1.3163750314167944e-07, "loss": 0.0113, "step": 30883 }, { "epoch": 9.5, "learning_rate": 1.3147638903529548e-07, "loss": 0.015, "step": 30884 }, { "epoch": 9.5, "learning_rate": 1.3131537293244678e-07, "loss": 0.0221, "step": 30885 }, { "epoch": 9.5, "learning_rate": 1.3115445483473211e-07, "loss": 0.0129, "step": 30886 }, { "epoch": 9.5, "learning_rate": 1.3099363474374793e-07, "loss": 0.0275, "step": 30887 }, { "epoch": 9.5, "learning_rate": 1.30832912661093e-07, "loss": 0.0248, "step": 30888 }, { "epoch": 9.5, "learning_rate": 1.3067228858836378e-07, "loss": 0.028, "step": 30889 }, { "epoch": 9.5, "learning_rate": 1.3051176252715348e-07, "loss": 0.0291, "step": 30890 }, { "epoch": 9.5, "learning_rate": 1.3035133447905745e-07, "loss": 0.0318, "step": 30891 }, { "epoch": 9.5, "learning_rate": 1.301910044456689e-07, "loss": 0.029, "step": 30892 }, { "epoch": 9.51, "learning_rate": 1.3003077242858096e-07, "loss": 0.0331, "step": 30893 }, { "epoch": 9.51, "learning_rate": 1.298706384293824e-07, "loss": 0.0261, "step": 30894 }, { "epoch": 9.51, "learning_rate": 1.2971060244966526e-07, "loss": 0.022, "step": 30895 }, { "epoch": 9.51, "learning_rate": 1.2955066449101938e-07, "loss": 0.0274, "step": 30896 }, { "epoch": 9.51, "learning_rate": 1.2939082455503128e-07, "loss": 0.0084, "step": 30897 }, { "epoch": 9.51, "learning_rate": 1.2923108264328966e-07, "loss": 0.0373, "step": 30898 }, { "epoch": 9.51, "learning_rate": 1.2907143875738105e-07, "loss": 0.0453, "step": 30899 }, { "epoch": 9.51, "learning_rate": 1.2891189289889083e-07, "loss": 0.0122, "step": 30900 }, { "epoch": 9.51, "learning_rate": 1.287524450694011e-07, "loss": 0.0195, "step": 30901 }, { "epoch": 9.51, "learning_rate": 1.2859309527049835e-07, "loss": 0.02, "step": 30902 }, { "epoch": 9.51, "learning_rate": 1.2843384350376465e-07, "loss": 0.0223, "step": 30903 }, { "epoch": 9.51, "learning_rate": 1.2827468977077874e-07, "loss": 0.0176, "step": 30904 }, { "epoch": 9.51, "learning_rate": 1.2811563407312489e-07, "loss": 0.0131, "step": 30905 }, { "epoch": 9.51, "learning_rate": 1.2795667641238074e-07, "loss": 0.0332, "step": 30906 }, { "epoch": 9.51, "learning_rate": 1.2779781679012505e-07, "loss": 0.0305, "step": 30907 }, { "epoch": 9.51, "learning_rate": 1.276390552079354e-07, "loss": 0.0153, "step": 30908 }, { "epoch": 9.51, "learning_rate": 1.2748039166738834e-07, "loss": 0.0147, "step": 30909 }, { "epoch": 9.51, "learning_rate": 1.273218261700604e-07, "loss": 0.0362, "step": 30910 }, { "epoch": 9.51, "learning_rate": 1.2716335871752472e-07, "loss": 0.0228, "step": 30911 }, { "epoch": 9.51, "learning_rate": 1.2700498931135674e-07, "loss": 0.0137, "step": 30912 }, { "epoch": 9.51, "learning_rate": 1.2684671795312854e-07, "loss": 0.0121, "step": 30913 }, { "epoch": 9.51, "learning_rate": 1.2668854464441104e-07, "loss": 0.0308, "step": 30914 }, { "epoch": 9.51, "learning_rate": 1.2653046938677638e-07, "loss": 0.0231, "step": 30915 }, { "epoch": 9.51, "learning_rate": 1.2637249218179325e-07, "loss": 0.0365, "step": 30916 }, { "epoch": 9.51, "learning_rate": 1.2621461303103266e-07, "loss": 0.0125, "step": 30917 }, { "epoch": 9.51, "learning_rate": 1.260568319360589e-07, "loss": 0.028, "step": 30918 }, { "epoch": 9.51, "learning_rate": 1.258991488984418e-07, "loss": 0.0329, "step": 30919 }, { "epoch": 9.51, "learning_rate": 1.257415639197468e-07, "loss": 0.019, "step": 30920 }, { "epoch": 9.51, "learning_rate": 1.2558407700153708e-07, "loss": 0.0233, "step": 30921 }, { "epoch": 9.51, "learning_rate": 1.2542668814537918e-07, "loss": 0.0365, "step": 30922 }, { "epoch": 9.51, "learning_rate": 1.2526939735283517e-07, "loss": 0.023, "step": 30923 }, { "epoch": 9.51, "learning_rate": 1.2511220462546603e-07, "loss": 0.0287, "step": 30924 }, { "epoch": 9.52, "learning_rate": 1.2495510996483384e-07, "loss": 0.0241, "step": 30925 }, { "epoch": 9.52, "learning_rate": 1.2479811337249848e-07, "loss": 0.0204, "step": 30926 }, { "epoch": 9.52, "learning_rate": 1.246412148500198e-07, "loss": 0.0287, "step": 30927 }, { "epoch": 9.52, "learning_rate": 1.2448441439895541e-07, "loss": 0.0221, "step": 30928 }, { "epoch": 9.52, "learning_rate": 1.243277120208608e-07, "loss": 0.0199, "step": 30929 }, { "epoch": 9.52, "learning_rate": 1.2417110771729468e-07, "loss": 0.0174, "step": 30930 }, { "epoch": 9.52, "learning_rate": 1.2401460148981136e-07, "loss": 0.0366, "step": 30931 }, { "epoch": 9.52, "learning_rate": 1.2385819333996518e-07, "loss": 0.0142, "step": 30932 }, { "epoch": 9.52, "learning_rate": 1.237018832693093e-07, "loss": 0.0343, "step": 30933 }, { "epoch": 9.52, "learning_rate": 1.2354567127939476e-07, "loss": 0.013, "step": 30934 }, { "epoch": 9.52, "learning_rate": 1.233895573717758e-07, "loss": 0.0325, "step": 30935 }, { "epoch": 9.52, "learning_rate": 1.2323354154799904e-07, "loss": 0.0219, "step": 30936 }, { "epoch": 9.52, "learning_rate": 1.2307762380961762e-07, "loss": 0.024, "step": 30937 }, { "epoch": 9.52, "learning_rate": 1.229218041581781e-07, "loss": 0.0259, "step": 30938 }, { "epoch": 9.52, "learning_rate": 1.2276608259522705e-07, "loss": 0.0207, "step": 30939 }, { "epoch": 9.52, "learning_rate": 1.2261045912231318e-07, "loss": 0.0179, "step": 30940 }, { "epoch": 9.52, "learning_rate": 1.2245493374097973e-07, "loss": 0.0182, "step": 30941 }, { "epoch": 9.52, "learning_rate": 1.2229950645277321e-07, "loss": 0.0278, "step": 30942 }, { "epoch": 9.52, "learning_rate": 1.2214417725923578e-07, "loss": 0.0347, "step": 30943 }, { "epoch": 9.52, "learning_rate": 1.2198894616190948e-07, "loss": 0.0144, "step": 30944 }, { "epoch": 9.52, "learning_rate": 1.2183381316233756e-07, "loss": 0.0113, "step": 30945 }, { "epoch": 9.52, "learning_rate": 1.2167877826205876e-07, "loss": 0.0329, "step": 30946 }, { "epoch": 9.52, "learning_rate": 1.215238414626152e-07, "loss": 0.0212, "step": 30947 }, { "epoch": 9.52, "learning_rate": 1.2136900276554453e-07, "loss": 0.0255, "step": 30948 }, { "epoch": 9.52, "learning_rate": 1.212142621723833e-07, "loss": 0.017, "step": 30949 }, { "epoch": 9.52, "learning_rate": 1.2105961968466805e-07, "loss": 0.0294, "step": 30950 }, { "epoch": 9.52, "learning_rate": 1.2090507530393757e-07, "loss": 0.0206, "step": 30951 }, { "epoch": 9.52, "learning_rate": 1.2075062903172286e-07, "loss": 0.022, "step": 30952 }, { "epoch": 9.52, "learning_rate": 1.2059628086956044e-07, "loss": 0.0334, "step": 30953 }, { "epoch": 9.52, "learning_rate": 1.204420308189802e-07, "loss": 0.0244, "step": 30954 }, { "epoch": 9.52, "learning_rate": 1.202878788815187e-07, "loss": 0.0167, "step": 30955 }, { "epoch": 9.52, "learning_rate": 1.201338250587014e-07, "loss": 0.0262, "step": 30956 }, { "epoch": 9.52, "learning_rate": 1.199798693520615e-07, "loss": 0.0167, "step": 30957 }, { "epoch": 9.53, "learning_rate": 1.198260117631278e-07, "loss": 0.0324, "step": 30958 }, { "epoch": 9.53, "learning_rate": 1.1967225229342794e-07, "loss": 0.0171, "step": 30959 }, { "epoch": 9.53, "learning_rate": 1.1951859094448848e-07, "loss": 0.0329, "step": 30960 }, { "epoch": 9.53, "learning_rate": 1.1936502771783488e-07, "loss": 0.0172, "step": 30961 }, { "epoch": 9.53, "learning_rate": 1.1921156261499255e-07, "loss": 0.0267, "step": 30962 }, { "epoch": 9.53, "learning_rate": 1.19058195637487e-07, "loss": 0.0344, "step": 30963 }, { "epoch": 9.53, "learning_rate": 1.1890492678683808e-07, "loss": 0.0258, "step": 30964 }, { "epoch": 9.53, "learning_rate": 1.1875175606457234e-07, "loss": 0.0207, "step": 30965 }, { "epoch": 9.53, "learning_rate": 1.1859868347220749e-07, "loss": 0.0303, "step": 30966 }, { "epoch": 9.53, "learning_rate": 1.1844570901126451e-07, "loss": 0.0246, "step": 30967 }, { "epoch": 9.53, "learning_rate": 1.182928326832622e-07, "loss": 0.0266, "step": 30968 }, { "epoch": 9.53, "learning_rate": 1.1814005448972043e-07, "loss": 0.0224, "step": 30969 }, { "epoch": 9.53, "learning_rate": 1.1798737443215469e-07, "loss": 0.0144, "step": 30970 }, { "epoch": 9.53, "learning_rate": 1.1783479251208152e-07, "loss": 0.0144, "step": 30971 }, { "epoch": 9.53, "learning_rate": 1.1768230873101749e-07, "loss": 0.0338, "step": 30972 }, { "epoch": 9.53, "learning_rate": 1.1752992309047584e-07, "loss": 0.0251, "step": 30973 }, { "epoch": 9.53, "learning_rate": 1.1737763559196869e-07, "loss": 0.0231, "step": 30974 }, { "epoch": 9.53, "learning_rate": 1.1722544623701038e-07, "loss": 0.0245, "step": 30975 }, { "epoch": 9.53, "learning_rate": 1.1707335502711192e-07, "loss": 0.0239, "step": 30976 }, { "epoch": 9.53, "learning_rate": 1.1692136196378435e-07, "loss": 0.0201, "step": 30977 }, { "epoch": 9.53, "learning_rate": 1.1676946704853531e-07, "loss": 0.0125, "step": 30978 }, { "epoch": 9.53, "learning_rate": 1.1661767028287363e-07, "loss": 0.0307, "step": 30979 }, { "epoch": 9.53, "learning_rate": 1.1646597166830809e-07, "loss": 0.0157, "step": 30980 }, { "epoch": 9.53, "learning_rate": 1.1631437120634304e-07, "loss": 0.0479, "step": 30981 }, { "epoch": 9.53, "learning_rate": 1.1616286889848726e-07, "loss": 0.0125, "step": 30982 }, { "epoch": 9.53, "learning_rate": 1.160114647462418e-07, "loss": 0.0115, "step": 30983 }, { "epoch": 9.53, "learning_rate": 1.1586015875111211e-07, "loss": 0.0173, "step": 30984 }, { "epoch": 9.53, "learning_rate": 1.1570895091460144e-07, "loss": 0.0252, "step": 30985 }, { "epoch": 9.53, "learning_rate": 1.1555784123820967e-07, "loss": 0.0196, "step": 30986 }, { "epoch": 9.53, "learning_rate": 1.1540682972343898e-07, "loss": 0.0139, "step": 30987 }, { "epoch": 9.53, "learning_rate": 1.1525591637178813e-07, "loss": 0.0324, "step": 30988 }, { "epoch": 9.53, "learning_rate": 1.1510510118475593e-07, "loss": 0.0298, "step": 30989 }, { "epoch": 9.54, "learning_rate": 1.1495438416384009e-07, "loss": 0.0224, "step": 30990 }, { "epoch": 9.54, "learning_rate": 1.1480376531053716e-07, "loss": 0.0144, "step": 30991 }, { "epoch": 9.54, "learning_rate": 1.1465324462634375e-07, "loss": 0.0179, "step": 30992 }, { "epoch": 9.54, "learning_rate": 1.1450282211275532e-07, "loss": 0.0172, "step": 30993 }, { "epoch": 9.54, "learning_rate": 1.1435249777126289e-07, "loss": 0.0332, "step": 30994 }, { "epoch": 9.54, "learning_rate": 1.1420227160336195e-07, "loss": 0.0147, "step": 30995 }, { "epoch": 9.54, "learning_rate": 1.140521436105435e-07, "loss": 0.0202, "step": 30996 }, { "epoch": 9.54, "learning_rate": 1.139021137942986e-07, "loss": 0.0157, "step": 30997 }, { "epoch": 9.54, "learning_rate": 1.1375218215611606e-07, "loss": 0.0086, "step": 30998 }, { "epoch": 9.54, "learning_rate": 1.1360234869748688e-07, "loss": 0.0074, "step": 30999 }, { "epoch": 9.54, "learning_rate": 1.1345261341989766e-07, "loss": 0.0098, "step": 31000 }, { "epoch": 9.54, "learning_rate": 1.1330297632483389e-07, "loss": 0.0273, "step": 31001 }, { "epoch": 9.54, "learning_rate": 1.1315343741378549e-07, "loss": 0.0139, "step": 31002 }, { "epoch": 9.54, "learning_rate": 1.1300399668823569e-07, "loss": 0.0132, "step": 31003 }, { "epoch": 9.54, "learning_rate": 1.1285465414966667e-07, "loss": 0.0308, "step": 31004 }, { "epoch": 9.54, "learning_rate": 1.1270540979956501e-07, "loss": 0.0274, "step": 31005 }, { "epoch": 9.54, "learning_rate": 1.1255626363940952e-07, "loss": 0.0253, "step": 31006 }, { "epoch": 9.54, "learning_rate": 1.1240721567068346e-07, "loss": 0.0232, "step": 31007 }, { "epoch": 9.54, "learning_rate": 1.1225826589486677e-07, "loss": 0.0223, "step": 31008 }, { "epoch": 9.54, "learning_rate": 1.1210941431343714e-07, "loss": 0.0174, "step": 31009 }, { "epoch": 9.54, "learning_rate": 1.1196066092787561e-07, "loss": 0.0245, "step": 31010 }, { "epoch": 9.54, "learning_rate": 1.1181200573965655e-07, "loss": 0.0111, "step": 31011 }, { "epoch": 9.54, "learning_rate": 1.116634487502577e-07, "loss": 0.0178, "step": 31012 }, { "epoch": 9.54, "learning_rate": 1.115149899611545e-07, "loss": 0.0194, "step": 31013 }, { "epoch": 9.54, "learning_rate": 1.1136662937382137e-07, "loss": 0.0203, "step": 31014 }, { "epoch": 9.54, "learning_rate": 1.1121836698973043e-07, "loss": 0.0215, "step": 31015 }, { "epoch": 9.54, "learning_rate": 1.1107020281035497e-07, "loss": 0.0249, "step": 31016 }, { "epoch": 9.54, "learning_rate": 1.1092213683716713e-07, "loss": 0.0314, "step": 31017 }, { "epoch": 9.54, "learning_rate": 1.1077416907163573e-07, "loss": 0.026, "step": 31018 }, { "epoch": 9.54, "learning_rate": 1.1062629951523074e-07, "loss": 0.0254, "step": 31019 }, { "epoch": 9.54, "learning_rate": 1.1047852816942206e-07, "loss": 0.0284, "step": 31020 }, { "epoch": 9.54, "learning_rate": 1.103308550356752e-07, "loss": 0.0195, "step": 31021 }, { "epoch": 9.54, "learning_rate": 1.1018328011545787e-07, "loss": 0.0143, "step": 31022 }, { "epoch": 9.55, "learning_rate": 1.1003580341023445e-07, "loss": 0.0125, "step": 31023 }, { "epoch": 9.55, "learning_rate": 1.0988842492147155e-07, "loss": 0.0226, "step": 31024 }, { "epoch": 9.55, "learning_rate": 1.0974114465063135e-07, "loss": 0.0275, "step": 31025 }, { "epoch": 9.55, "learning_rate": 1.0959396259917487e-07, "loss": 0.0188, "step": 31026 }, { "epoch": 9.55, "learning_rate": 1.0944687876856875e-07, "loss": 0.0284, "step": 31027 }, { "epoch": 9.55, "learning_rate": 1.0929989316026846e-07, "loss": 0.017, "step": 31028 }, { "epoch": 9.55, "learning_rate": 1.0915300577573618e-07, "loss": 0.0124, "step": 31029 }, { "epoch": 9.55, "learning_rate": 1.0900621661642962e-07, "loss": 0.0268, "step": 31030 }, { "epoch": 9.55, "learning_rate": 1.0885952568380764e-07, "loss": 0.0062, "step": 31031 }, { "epoch": 9.55, "learning_rate": 1.0871293297932683e-07, "loss": 0.031, "step": 31032 }, { "epoch": 9.55, "learning_rate": 1.0856643850444271e-07, "loss": 0.023, "step": 31033 }, { "epoch": 9.55, "learning_rate": 1.0842004226060965e-07, "loss": 0.014, "step": 31034 }, { "epoch": 9.55, "learning_rate": 1.0827374424928316e-07, "loss": 0.0204, "step": 31035 }, { "epoch": 9.55, "learning_rate": 1.081275444719132e-07, "loss": 0.0199, "step": 31036 }, { "epoch": 9.55, "learning_rate": 1.0798144292995416e-07, "loss": 0.01, "step": 31037 }, { "epoch": 9.55, "learning_rate": 1.0783543962485598e-07, "loss": 0.0091, "step": 31038 }, { "epoch": 9.55, "learning_rate": 1.0768953455806863e-07, "loss": 0.0349, "step": 31039 }, { "epoch": 9.55, "learning_rate": 1.0754372773104093e-07, "loss": 0.0154, "step": 31040 }, { "epoch": 9.55, "learning_rate": 1.0739801914522174e-07, "loss": 0.013, "step": 31041 }, { "epoch": 9.55, "learning_rate": 1.0725240880205768e-07, "loss": 0.0332, "step": 31042 }, { "epoch": 9.55, "learning_rate": 1.0710689670299423e-07, "loss": 0.0369, "step": 31043 }, { "epoch": 9.55, "learning_rate": 1.0696148284947694e-07, "loss": 0.0177, "step": 31044 }, { "epoch": 9.55, "learning_rate": 1.0681616724295019e-07, "loss": 0.0126, "step": 31045 }, { "epoch": 9.55, "learning_rate": 1.0667094988485616e-07, "loss": 0.0225, "step": 31046 }, { "epoch": 9.55, "learning_rate": 1.0652583077663703e-07, "loss": 0.0381, "step": 31047 }, { "epoch": 9.55, "learning_rate": 1.0638080991973609e-07, "loss": 0.0372, "step": 31048 }, { "epoch": 9.55, "learning_rate": 1.062358873155911e-07, "loss": 0.0222, "step": 31049 }, { "epoch": 9.55, "learning_rate": 1.0609106296564197e-07, "loss": 0.0283, "step": 31050 }, { "epoch": 9.55, "learning_rate": 1.0594633687132649e-07, "loss": 0.017, "step": 31051 }, { "epoch": 9.55, "learning_rate": 1.058017090340835e-07, "loss": 0.0253, "step": 31052 }, { "epoch": 9.55, "learning_rate": 1.0565717945534737e-07, "loss": 0.016, "step": 31053 }, { "epoch": 9.55, "learning_rate": 1.0551274813655476e-07, "loss": 0.0163, "step": 31054 }, { "epoch": 9.56, "learning_rate": 1.0536841507914119e-07, "loss": 0.033, "step": 31055 }, { "epoch": 9.56, "learning_rate": 1.0522418028453663e-07, "loss": 0.005, "step": 31056 }, { "epoch": 9.56, "learning_rate": 1.0508004375417546e-07, "loss": 0.0219, "step": 31057 }, { "epoch": 9.56, "learning_rate": 1.0493600548948879e-07, "loss": 0.0207, "step": 31058 }, { "epoch": 9.56, "learning_rate": 1.0479206549190767e-07, "loss": 0.0199, "step": 31059 }, { "epoch": 9.56, "learning_rate": 1.04648223762861e-07, "loss": 0.0109, "step": 31060 }, { "epoch": 9.56, "learning_rate": 1.0450448030377758e-07, "loss": 0.0229, "step": 31061 }, { "epoch": 9.56, "learning_rate": 1.0436083511608408e-07, "loss": 0.0185, "step": 31062 }, { "epoch": 9.56, "learning_rate": 1.0421728820120713e-07, "loss": 0.03, "step": 31063 }, { "epoch": 9.56, "learning_rate": 1.0407383956057337e-07, "loss": 0.0338, "step": 31064 }, { "epoch": 9.56, "learning_rate": 1.0393048919560722e-07, "loss": 0.0232, "step": 31065 }, { "epoch": 9.56, "learning_rate": 1.0378723710772975e-07, "loss": 0.0316, "step": 31066 }, { "epoch": 9.56, "learning_rate": 1.0364408329836762e-07, "loss": 0.0266, "step": 31067 }, { "epoch": 9.56, "learning_rate": 1.0350102776893966e-07, "loss": 0.0201, "step": 31068 }, { "epoch": 9.56, "learning_rate": 1.0335807052086811e-07, "loss": 0.0242, "step": 31069 }, { "epoch": 9.56, "learning_rate": 1.032152115555718e-07, "loss": 0.0227, "step": 31070 }, { "epoch": 9.56, "learning_rate": 1.0307245087446849e-07, "loss": 0.0212, "step": 31071 }, { "epoch": 9.56, "learning_rate": 1.0292978847897816e-07, "loss": 0.0316, "step": 31072 }, { "epoch": 9.56, "learning_rate": 1.0278722437051636e-07, "loss": 0.0219, "step": 31073 }, { "epoch": 9.56, "learning_rate": 1.026447585504986e-07, "loss": 0.0202, "step": 31074 }, { "epoch": 9.56, "learning_rate": 1.0250239102033931e-07, "loss": 0.0215, "step": 31075 }, { "epoch": 9.56, "learning_rate": 1.0236012178145404e-07, "loss": 0.0168, "step": 31076 }, { "epoch": 9.56, "learning_rate": 1.0221795083525498e-07, "loss": 0.0254, "step": 31077 }, { "epoch": 9.56, "learning_rate": 1.0207587818315323e-07, "loss": 0.0228, "step": 31078 }, { "epoch": 9.56, "learning_rate": 1.0193390382656099e-07, "loss": 0.0245, "step": 31079 }, { "epoch": 9.56, "learning_rate": 1.0179202776688601e-07, "loss": 0.0241, "step": 31080 }, { "epoch": 9.56, "learning_rate": 1.016502500055394e-07, "loss": 0.0454, "step": 31081 }, { "epoch": 9.56, "learning_rate": 1.0150857054392893e-07, "loss": 0.011, "step": 31082 }, { "epoch": 9.56, "learning_rate": 1.0136698938346012e-07, "loss": 0.0251, "step": 31083 }, { "epoch": 9.56, "learning_rate": 1.0122550652553964e-07, "loss": 0.0236, "step": 31084 }, { "epoch": 9.56, "learning_rate": 1.0108412197157413e-07, "loss": 0.041, "step": 31085 }, { "epoch": 9.56, "learning_rate": 1.0094283572296581e-07, "loss": 0.0246, "step": 31086 }, { "epoch": 9.56, "learning_rate": 1.00801647781118e-07, "loss": 0.0192, "step": 31087 }, { "epoch": 9.57, "learning_rate": 1.0066055814743292e-07, "loss": 0.0248, "step": 31088 }, { "epoch": 9.57, "learning_rate": 1.0051956682331387e-07, "loss": 0.0254, "step": 31089 }, { "epoch": 9.57, "learning_rate": 1.0037867381015753e-07, "loss": 0.0228, "step": 31090 }, { "epoch": 9.57, "learning_rate": 1.00237879109365e-07, "loss": 0.0267, "step": 31091 }, { "epoch": 9.57, "learning_rate": 1.0009718272233404e-07, "loss": 0.0239, "step": 31092 }, { "epoch": 9.57, "learning_rate": 9.995658465046243e-08, "loss": 0.0234, "step": 31093 }, { "epoch": 9.57, "learning_rate": 9.981608489514683e-08, "loss": 0.0373, "step": 31094 }, { "epoch": 9.57, "learning_rate": 9.967568345778056e-08, "loss": 0.0298, "step": 31095 }, { "epoch": 9.57, "learning_rate": 9.953538033975918e-08, "loss": 0.0049, "step": 31096 }, { "epoch": 9.57, "learning_rate": 9.939517554247713e-08, "loss": 0.0158, "step": 31097 }, { "epoch": 9.57, "learning_rate": 9.925506906732441e-08, "loss": 0.0113, "step": 31098 }, { "epoch": 9.57, "learning_rate": 9.911506091569434e-08, "loss": 0.0138, "step": 31099 }, { "epoch": 9.57, "learning_rate": 9.897515108897692e-08, "loss": 0.0113, "step": 31100 }, { "epoch": 9.57, "learning_rate": 9.883533958855995e-08, "loss": 0.0219, "step": 31101 }, { "epoch": 9.57, "learning_rate": 9.869562641583453e-08, "loss": 0.0216, "step": 31102 }, { "epoch": 9.57, "learning_rate": 9.85560115721862e-08, "loss": 0.0276, "step": 31103 }, { "epoch": 9.57, "learning_rate": 9.841649505900275e-08, "loss": 0.0198, "step": 31104 }, { "epoch": 9.57, "learning_rate": 9.827707687766863e-08, "loss": 0.022, "step": 31105 }, { "epoch": 9.57, "learning_rate": 9.81377570295694e-08, "loss": 0.0173, "step": 31106 }, { "epoch": 9.57, "learning_rate": 9.799853551608729e-08, "loss": 0.0164, "step": 31107 }, { "epoch": 9.57, "learning_rate": 9.785941233860563e-08, "loss": 0.0246, "step": 31108 }, { "epoch": 9.57, "learning_rate": 9.772038749850665e-08, "loss": 0.0525, "step": 31109 }, { "epoch": 9.57, "learning_rate": 9.758146099717036e-08, "loss": 0.0176, "step": 31110 }, { "epoch": 9.57, "learning_rate": 9.744263283597566e-08, "loss": 0.034, "step": 31111 }, { "epoch": 9.57, "learning_rate": 9.730390301630365e-08, "loss": 0.0223, "step": 31112 }, { "epoch": 9.57, "learning_rate": 9.716527153952882e-08, "loss": 0.0181, "step": 31113 }, { "epoch": 9.57, "learning_rate": 9.702673840703114e-08, "loss": 0.0274, "step": 31114 }, { "epoch": 9.57, "learning_rate": 9.688830362018397e-08, "loss": 0.0238, "step": 31115 }, { "epoch": 9.57, "learning_rate": 9.674996718036178e-08, "loss": 0.0128, "step": 31116 }, { "epoch": 9.57, "learning_rate": 9.661172908894123e-08, "loss": 0.0162, "step": 31117 }, { "epoch": 9.57, "learning_rate": 9.647358934729233e-08, "loss": 0.0177, "step": 31118 }, { "epoch": 9.57, "learning_rate": 9.633554795678957e-08, "loss": 0.0232, "step": 31119 }, { "epoch": 9.58, "learning_rate": 9.619760491880182e-08, "loss": 0.0204, "step": 31120 }, { "epoch": 9.58, "learning_rate": 9.605976023469798e-08, "loss": 0.0157, "step": 31121 }, { "epoch": 9.58, "learning_rate": 9.59220139058492e-08, "loss": 0.0335, "step": 31122 }, { "epoch": 9.58, "learning_rate": 9.578436593362328e-08, "loss": 0.0235, "step": 31123 }, { "epoch": 9.58, "learning_rate": 9.564681631938688e-08, "loss": 0.0163, "step": 31124 }, { "epoch": 9.58, "learning_rate": 9.550936506450448e-08, "loss": 0.0255, "step": 31125 }, { "epoch": 9.58, "learning_rate": 9.537201217034275e-08, "loss": 0.0278, "step": 31126 }, { "epoch": 9.58, "learning_rate": 9.523475763826618e-08, "loss": 0.0203, "step": 31127 }, { "epoch": 9.58, "learning_rate": 9.509760146963587e-08, "loss": 0.013, "step": 31128 }, { "epoch": 9.58, "learning_rate": 9.49605436658152e-08, "loss": 0.0285, "step": 31129 }, { "epoch": 9.58, "learning_rate": 9.48235842281653e-08, "loss": 0.0196, "step": 31130 }, { "epoch": 9.58, "learning_rate": 9.468672315804616e-08, "loss": 0.0426, "step": 31131 }, { "epoch": 9.58, "learning_rate": 9.454996045681675e-08, "loss": 0.0208, "step": 31132 }, { "epoch": 9.58, "learning_rate": 9.441329612583594e-08, "loss": 0.03, "step": 31133 }, { "epoch": 9.58, "learning_rate": 9.427673016646044e-08, "loss": 0.0196, "step": 31134 }, { "epoch": 9.58, "learning_rate": 9.414026258004583e-08, "loss": 0.0287, "step": 31135 }, { "epoch": 9.58, "learning_rate": 9.400389336794658e-08, "loss": 0.0315, "step": 31136 }, { "epoch": 9.58, "learning_rate": 9.386762253152049e-08, "loss": 0.0202, "step": 31137 }, { "epoch": 9.58, "learning_rate": 9.373145007211759e-08, "loss": 0.0288, "step": 31138 }, { "epoch": 9.58, "learning_rate": 9.359537599109125e-08, "loss": 0.0196, "step": 31139 }, { "epoch": 9.58, "learning_rate": 9.345940028979373e-08, "loss": 0.0182, "step": 31140 }, { "epoch": 9.58, "learning_rate": 9.332352296957393e-08, "loss": 0.0307, "step": 31141 }, { "epoch": 9.58, "learning_rate": 9.318774403178188e-08, "loss": 0.0192, "step": 31142 }, { "epoch": 9.58, "learning_rate": 9.30520634777643e-08, "loss": 0.0214, "step": 31143 }, { "epoch": 9.58, "learning_rate": 9.291648130887232e-08, "loss": 0.0342, "step": 31144 }, { "epoch": 9.58, "learning_rate": 9.278099752645042e-08, "loss": 0.0231, "step": 31145 }, { "epoch": 9.58, "learning_rate": 9.264561213184309e-08, "loss": 0.0305, "step": 31146 }, { "epoch": 9.58, "learning_rate": 9.251032512639591e-08, "loss": 0.022, "step": 31147 }, { "epoch": 9.58, "learning_rate": 9.237513651145224e-08, "loss": 0.0176, "step": 31148 }, { "epoch": 9.58, "learning_rate": 9.224004628835437e-08, "loss": 0.0184, "step": 31149 }, { "epoch": 9.58, "learning_rate": 9.210505445844453e-08, "loss": 0.0259, "step": 31150 }, { "epoch": 9.58, "learning_rate": 9.197016102306277e-08, "loss": 0.0293, "step": 31151 }, { "epoch": 9.58, "learning_rate": 9.183536598354803e-08, "loss": 0.0306, "step": 31152 }, { "epoch": 9.59, "learning_rate": 9.170066934124034e-08, "loss": 0.0138, "step": 31153 }, { "epoch": 9.59, "learning_rate": 9.15660710974764e-08, "loss": 0.0325, "step": 31154 }, { "epoch": 9.59, "learning_rate": 9.143157125359514e-08, "loss": 0.017, "step": 31155 }, { "epoch": 9.59, "learning_rate": 9.129716981092774e-08, "loss": 0.0233, "step": 31156 }, { "epoch": 9.59, "learning_rate": 9.11628667708131e-08, "loss": 0.0189, "step": 31157 }, { "epoch": 9.59, "learning_rate": 9.10286621345835e-08, "loss": 0.0207, "step": 31158 }, { "epoch": 9.59, "learning_rate": 9.089455590357121e-08, "loss": 0.0103, "step": 31159 }, { "epoch": 9.59, "learning_rate": 9.07605480791085e-08, "loss": 0.0244, "step": 31160 }, { "epoch": 9.59, "learning_rate": 9.062663866252541e-08, "loss": 0.0284, "step": 31161 }, { "epoch": 9.59, "learning_rate": 9.049282765515421e-08, "loss": 0.0227, "step": 31162 }, { "epoch": 9.59, "learning_rate": 9.035911505832051e-08, "loss": 0.018, "step": 31163 }, { "epoch": 9.59, "learning_rate": 9.022550087335435e-08, "loss": 0.0391, "step": 31164 }, { "epoch": 9.59, "learning_rate": 9.009198510158357e-08, "loss": 0.04, "step": 31165 }, { "epoch": 9.59, "learning_rate": 8.995856774433042e-08, "loss": 0.0334, "step": 31166 }, { "epoch": 9.59, "learning_rate": 8.982524880292277e-08, "loss": 0.0187, "step": 31167 }, { "epoch": 9.59, "learning_rate": 8.969202827868395e-08, "loss": 0.0379, "step": 31168 }, { "epoch": 9.59, "learning_rate": 8.955890617293738e-08, "loss": 0.0225, "step": 31169 }, { "epoch": 9.59, "learning_rate": 8.942588248700423e-08, "loss": 0.0261, "step": 31170 }, { "epoch": 9.59, "learning_rate": 8.929295722220565e-08, "loss": 0.0165, "step": 31171 }, { "epoch": 9.59, "learning_rate": 8.916013037986282e-08, "loss": 0.0174, "step": 31172 }, { "epoch": 9.59, "learning_rate": 8.902740196129244e-08, "loss": 0.0303, "step": 31173 }, { "epoch": 9.59, "learning_rate": 8.889477196781571e-08, "loss": 0.039, "step": 31174 }, { "epoch": 9.59, "learning_rate": 8.87622404007471e-08, "loss": 0.0255, "step": 31175 }, { "epoch": 9.59, "learning_rate": 8.862980726140446e-08, "loss": 0.0292, "step": 31176 }, { "epoch": 9.59, "learning_rate": 8.849747255110341e-08, "loss": 0.0184, "step": 31177 }, { "epoch": 9.59, "learning_rate": 8.836523627115623e-08, "loss": 0.0218, "step": 31178 }, { "epoch": 9.59, "learning_rate": 8.823309842287631e-08, "loss": 0.0198, "step": 31179 }, { "epoch": 9.59, "learning_rate": 8.810105900757814e-08, "loss": 0.0303, "step": 31180 }, { "epoch": 9.59, "learning_rate": 8.79691180265696e-08, "loss": 0.0254, "step": 31181 }, { "epoch": 9.59, "learning_rate": 8.783727548116516e-08, "loss": 0.0355, "step": 31182 }, { "epoch": 9.59, "learning_rate": 8.770553137267046e-08, "loss": 0.0276, "step": 31183 }, { "epoch": 9.59, "learning_rate": 8.757388570239555e-08, "loss": 0.0262, "step": 31184 }, { "epoch": 9.6, "learning_rate": 8.744233847164829e-08, "loss": 0.0169, "step": 31185 }, { "epoch": 9.6, "learning_rate": 8.731088968173318e-08, "loss": 0.02, "step": 31186 }, { "epoch": 9.6, "learning_rate": 8.717953933395695e-08, "loss": 0.0171, "step": 31187 }, { "epoch": 9.6, "learning_rate": 8.704828742962412e-08, "loss": 0.0135, "step": 31188 }, { "epoch": 9.6, "learning_rate": 8.69171339700381e-08, "loss": 0.0089, "step": 31189 }, { "epoch": 9.6, "learning_rate": 8.678607895650115e-08, "loss": 0.0194, "step": 31190 }, { "epoch": 9.6, "learning_rate": 8.665512239031337e-08, "loss": 0.0189, "step": 31191 }, { "epoch": 9.6, "learning_rate": 8.652426427277815e-08, "loss": 0.0526, "step": 31192 }, { "epoch": 9.6, "learning_rate": 8.639350460519335e-08, "loss": 0.0128, "step": 31193 }, { "epoch": 9.6, "learning_rate": 8.626284338885682e-08, "loss": 0.0201, "step": 31194 }, { "epoch": 9.6, "learning_rate": 8.613228062506752e-08, "loss": 0.0186, "step": 31195 }, { "epoch": 9.6, "learning_rate": 8.600181631512217e-08, "loss": 0.0136, "step": 31196 }, { "epoch": 9.6, "learning_rate": 8.587145046031531e-08, "loss": 0.0435, "step": 31197 }, { "epoch": 9.6, "learning_rate": 8.574118306194146e-08, "loss": 0.0316, "step": 31198 }, { "epoch": 9.6, "learning_rate": 8.561101412129624e-08, "loss": 0.0144, "step": 31199 }, { "epoch": 9.6, "learning_rate": 8.548094363966974e-08, "loss": 0.0294, "step": 31200 }, { "epoch": 9.6, "learning_rate": 8.535097161835537e-08, "loss": 0.0281, "step": 31201 }, { "epoch": 9.6, "learning_rate": 8.52210980586432e-08, "loss": 0.0355, "step": 31202 }, { "epoch": 9.6, "learning_rate": 8.509132296182332e-08, "loss": 0.0145, "step": 31203 }, { "epoch": 9.6, "learning_rate": 8.49616463291847e-08, "loss": 0.0218, "step": 31204 }, { "epoch": 9.6, "learning_rate": 8.483206816201516e-08, "loss": 0.0222, "step": 31205 }, { "epoch": 9.6, "learning_rate": 8.470258846160151e-08, "loss": 0.0226, "step": 31206 }, { "epoch": 9.6, "learning_rate": 8.457320722922935e-08, "loss": 0.0195, "step": 31207 }, { "epoch": 9.6, "learning_rate": 8.44439244661821e-08, "loss": 0.0273, "step": 31208 }, { "epoch": 9.6, "learning_rate": 8.431474017374653e-08, "loss": 0.0196, "step": 31209 }, { "epoch": 9.6, "learning_rate": 8.41856543532038e-08, "loss": 0.0421, "step": 31210 }, { "epoch": 9.6, "learning_rate": 8.405666700583626e-08, "loss": 0.0119, "step": 31211 }, { "epoch": 9.6, "learning_rate": 8.392777813292396e-08, "loss": 0.0292, "step": 31212 }, { "epoch": 9.6, "learning_rate": 8.379898773574924e-08, "loss": 0.028, "step": 31213 }, { "epoch": 9.6, "learning_rate": 8.367029581558883e-08, "loss": 0.0253, "step": 31214 }, { "epoch": 9.6, "learning_rate": 8.354170237372172e-08, "loss": 0.0306, "step": 31215 }, { "epoch": 9.6, "learning_rate": 8.341320741142356e-08, "loss": 0.0362, "step": 31216 }, { "epoch": 9.6, "learning_rate": 8.328481092997331e-08, "loss": 0.0265, "step": 31217 }, { "epoch": 9.61, "learning_rate": 8.31565129306433e-08, "loss": 0.0146, "step": 31218 }, { "epoch": 9.61, "learning_rate": 8.302831341471029e-08, "loss": 0.0129, "step": 31219 }, { "epoch": 9.61, "learning_rate": 8.290021238344547e-08, "loss": 0.0227, "step": 31220 }, { "epoch": 9.61, "learning_rate": 8.277220983812007e-08, "loss": 0.0116, "step": 31221 }, { "epoch": 9.61, "learning_rate": 8.264430578000748e-08, "loss": 0.0375, "step": 31222 }, { "epoch": 9.61, "learning_rate": 8.251650021037671e-08, "loss": 0.026, "step": 31223 }, { "epoch": 9.61, "learning_rate": 8.238879313049785e-08, "loss": 0.0205, "step": 31224 }, { "epoch": 9.61, "learning_rate": 8.226118454163767e-08, "loss": 0.0429, "step": 31225 }, { "epoch": 9.61, "learning_rate": 8.213367444506515e-08, "loss": 0.0201, "step": 31226 }, { "epoch": 9.61, "learning_rate": 8.200626284204593e-08, "loss": 0.0195, "step": 31227 }, { "epoch": 9.61, "learning_rate": 8.187894973384458e-08, "loss": 0.0114, "step": 31228 }, { "epoch": 9.61, "learning_rate": 8.175173512172674e-08, "loss": 0.0172, "step": 31229 }, { "epoch": 9.61, "learning_rate": 8.162461900695474e-08, "loss": 0.0184, "step": 31230 }, { "epoch": 9.61, "learning_rate": 8.149760139079088e-08, "loss": 0.0215, "step": 31231 }, { "epoch": 9.61, "learning_rate": 8.137068227449751e-08, "loss": 0.0251, "step": 31232 }, { "epoch": 9.61, "learning_rate": 8.124386165933473e-08, "loss": 0.0128, "step": 31233 }, { "epoch": 9.61, "learning_rate": 8.111713954656153e-08, "loss": 0.017, "step": 31234 }, { "epoch": 9.61, "learning_rate": 8.099051593743578e-08, "loss": 0.0154, "step": 31235 }, { "epoch": 9.61, "learning_rate": 8.086399083321761e-08, "loss": 0.0312, "step": 31236 }, { "epoch": 9.61, "learning_rate": 8.073756423516043e-08, "loss": 0.0283, "step": 31237 }, { "epoch": 9.61, "learning_rate": 8.061123614452215e-08, "loss": 0.0198, "step": 31238 }, { "epoch": 9.61, "learning_rate": 8.04850065625551e-08, "loss": 0.022, "step": 31239 }, { "epoch": 9.61, "learning_rate": 8.035887549051491e-08, "loss": 0.0298, "step": 31240 }, { "epoch": 9.61, "learning_rate": 8.023284292965283e-08, "loss": 0.0119, "step": 31241 }, { "epoch": 9.61, "learning_rate": 8.01069088812212e-08, "loss": 0.015, "step": 31242 }, { "epoch": 9.61, "learning_rate": 7.99810733464701e-08, "loss": 0.0307, "step": 31243 }, { "epoch": 9.61, "learning_rate": 7.985533632665076e-08, "loss": 0.0373, "step": 31244 }, { "epoch": 9.61, "learning_rate": 7.972969782300888e-08, "loss": 0.0162, "step": 31245 }, { "epoch": 9.61, "learning_rate": 7.960415783679453e-08, "loss": 0.0196, "step": 31246 }, { "epoch": 9.61, "learning_rate": 7.947871636925342e-08, "loss": 0.0232, "step": 31247 }, { "epoch": 9.61, "learning_rate": 7.93533734216323e-08, "loss": 0.0312, "step": 31248 }, { "epoch": 9.61, "learning_rate": 7.922812899517573e-08, "loss": 0.0188, "step": 31249 }, { "epoch": 9.62, "learning_rate": 7.910298309112608e-08, "loss": 0.0182, "step": 31250 }, { "epoch": 9.62, "learning_rate": 7.897793571072787e-08, "loss": 0.0284, "step": 31251 }, { "epoch": 9.62, "learning_rate": 7.885298685522235e-08, "loss": 0.0182, "step": 31252 }, { "epoch": 9.62, "learning_rate": 7.872813652584854e-08, "loss": 0.0097, "step": 31253 }, { "epoch": 9.62, "learning_rate": 7.860338472384987e-08, "loss": 0.0153, "step": 31254 }, { "epoch": 9.62, "learning_rate": 7.847873145046315e-08, "loss": 0.0304, "step": 31255 }, { "epoch": 9.62, "learning_rate": 7.835417670692624e-08, "loss": 0.0224, "step": 31256 }, { "epoch": 9.62, "learning_rate": 7.822972049447708e-08, "loss": 0.0139, "step": 31257 }, { "epoch": 9.62, "learning_rate": 7.810536281435022e-08, "loss": 0.0237, "step": 31258 }, { "epoch": 9.62, "learning_rate": 7.798110366778134e-08, "loss": 0.0169, "step": 31259 }, { "epoch": 9.62, "learning_rate": 7.7856943056005e-08, "loss": 0.0195, "step": 31260 }, { "epoch": 9.62, "learning_rate": 7.773288098025356e-08, "loss": 0.0157, "step": 31261 }, { "epoch": 9.62, "learning_rate": 7.760891744175936e-08, "loss": 0.0145, "step": 31262 }, { "epoch": 9.62, "learning_rate": 7.748505244175252e-08, "loss": 0.0335, "step": 31263 }, { "epoch": 9.62, "learning_rate": 7.73612859814643e-08, "loss": 0.008, "step": 31264 }, { "epoch": 9.62, "learning_rate": 7.723761806212371e-08, "loss": 0.0665, "step": 31265 }, { "epoch": 9.62, "learning_rate": 7.711404868495976e-08, "loss": 0.0309, "step": 31266 }, { "epoch": 9.62, "learning_rate": 7.6990577851197e-08, "loss": 0.0225, "step": 31267 }, { "epoch": 9.62, "learning_rate": 7.68672055620634e-08, "loss": 0.0054, "step": 31268 }, { "epoch": 9.62, "learning_rate": 7.674393181878459e-08, "loss": 0.031, "step": 31269 }, { "epoch": 9.62, "learning_rate": 7.662075662258295e-08, "loss": 0.0168, "step": 31270 }, { "epoch": 9.62, "learning_rate": 7.649767997468305e-08, "loss": 0.0195, "step": 31271 }, { "epoch": 9.62, "learning_rate": 7.637470187630836e-08, "loss": 0.0211, "step": 31272 }, { "epoch": 9.62, "learning_rate": 7.625182232867789e-08, "loss": 0.0209, "step": 31273 }, { "epoch": 9.62, "learning_rate": 7.61290413330118e-08, "loss": 0.0166, "step": 31274 }, { "epoch": 9.62, "learning_rate": 7.600635889053243e-08, "loss": 0.0138, "step": 31275 }, { "epoch": 9.62, "learning_rate": 7.588377500245437e-08, "loss": 0.0089, "step": 31276 }, { "epoch": 9.62, "learning_rate": 7.576128966999774e-08, "loss": 0.0203, "step": 31277 }, { "epoch": 9.62, "learning_rate": 7.563890289437825e-08, "loss": 0.0228, "step": 31278 }, { "epoch": 9.62, "learning_rate": 7.551661467681048e-08, "loss": 0.0202, "step": 31279 }, { "epoch": 9.62, "learning_rate": 7.539442501850791e-08, "loss": 0.0321, "step": 31280 }, { "epoch": 9.62, "learning_rate": 7.527233392068734e-08, "loss": 0.0085, "step": 31281 }, { "epoch": 9.62, "learning_rate": 7.515034138456001e-08, "loss": 0.0202, "step": 31282 }, { "epoch": 9.63, "learning_rate": 7.502844741133497e-08, "loss": 0.0375, "step": 31283 }, { "epoch": 9.63, "learning_rate": 7.490665200222569e-08, "loss": 0.0256, "step": 31284 }, { "epoch": 9.63, "learning_rate": 7.478495515844008e-08, "loss": 0.0154, "step": 31285 }, { "epoch": 9.63, "learning_rate": 7.46633568811872e-08, "loss": 0.0224, "step": 31286 }, { "epoch": 9.63, "learning_rate": 7.454185717167495e-08, "loss": 0.0112, "step": 31287 }, { "epoch": 9.63, "learning_rate": 7.442045603110903e-08, "loss": 0.0396, "step": 31288 }, { "epoch": 9.63, "learning_rate": 7.429915346069738e-08, "loss": 0.0127, "step": 31289 }, { "epoch": 9.63, "learning_rate": 7.417794946164125e-08, "loss": 0.0294, "step": 31290 }, { "epoch": 9.63, "learning_rate": 7.405684403514635e-08, "loss": 0.0298, "step": 31291 }, { "epoch": 9.63, "learning_rate": 7.393583718241504e-08, "loss": 0.0312, "step": 31292 }, { "epoch": 9.63, "learning_rate": 7.381492890464858e-08, "loss": 0.0197, "step": 31293 }, { "epoch": 9.63, "learning_rate": 7.369411920304937e-08, "loss": 0.0279, "step": 31294 }, { "epoch": 9.63, "learning_rate": 7.35734080788142e-08, "loss": 0.0178, "step": 31295 }, { "epoch": 9.63, "learning_rate": 7.345279553314543e-08, "loss": 0.0173, "step": 31296 }, { "epoch": 9.63, "learning_rate": 7.33322815672377e-08, "loss": 0.0152, "step": 31297 }, { "epoch": 9.63, "learning_rate": 7.321186618228893e-08, "loss": 0.0253, "step": 31298 }, { "epoch": 9.63, "learning_rate": 7.309154937949592e-08, "loss": 0.0347, "step": 31299 }, { "epoch": 9.63, "learning_rate": 7.297133116005218e-08, "loss": 0.0483, "step": 31300 }, { "epoch": 9.63, "learning_rate": 7.28512115251523e-08, "loss": 0.0272, "step": 31301 }, { "epoch": 9.63, "learning_rate": 7.273119047598976e-08, "loss": 0.0172, "step": 31302 }, { "epoch": 9.63, "learning_rate": 7.261126801375474e-08, "loss": 0.0211, "step": 31303 }, { "epoch": 9.63, "learning_rate": 7.24914441396396e-08, "loss": 0.0178, "step": 31304 }, { "epoch": 9.63, "learning_rate": 7.237171885483341e-08, "loss": 0.0322, "step": 31305 }, { "epoch": 9.63, "learning_rate": 7.225209216052741e-08, "loss": 0.0244, "step": 31306 }, { "epoch": 9.63, "learning_rate": 7.213256405790625e-08, "loss": 0.022, "step": 31307 }, { "epoch": 9.63, "learning_rate": 7.201313454815784e-08, "loss": 0.0258, "step": 31308 }, { "epoch": 9.63, "learning_rate": 7.189380363247011e-08, "loss": 0.0245, "step": 31309 }, { "epoch": 9.63, "learning_rate": 7.177457131202658e-08, "loss": 0.0275, "step": 31310 }, { "epoch": 9.63, "learning_rate": 7.165543758801186e-08, "loss": 0.0155, "step": 31311 }, { "epoch": 9.63, "learning_rate": 7.153640246160831e-08, "loss": 0.0186, "step": 31312 }, { "epoch": 9.63, "learning_rate": 7.141746593399945e-08, "loss": 0.0153, "step": 31313 }, { "epoch": 9.63, "learning_rate": 7.129862800636545e-08, "loss": 0.0232, "step": 31314 }, { "epoch": 9.64, "learning_rate": 7.117988867988534e-08, "loss": 0.0216, "step": 31315 }, { "epoch": 9.64, "learning_rate": 7.106124795574154e-08, "loss": 0.0265, "step": 31316 }, { "epoch": 9.64, "learning_rate": 7.094270583510976e-08, "loss": 0.0239, "step": 31317 }, { "epoch": 9.64, "learning_rate": 7.082426231916684e-08, "loss": 0.0276, "step": 31318 }, { "epoch": 9.64, "learning_rate": 7.070591740909072e-08, "loss": 0.0096, "step": 31319 }, { "epoch": 9.64, "learning_rate": 7.058767110605602e-08, "loss": 0.0186, "step": 31320 }, { "epoch": 9.64, "learning_rate": 7.046952341123625e-08, "loss": 0.0185, "step": 31321 }, { "epoch": 9.64, "learning_rate": 7.035147432580603e-08, "loss": 0.0268, "step": 31322 }, { "epoch": 9.64, "learning_rate": 7.023352385093774e-08, "loss": 0.0099, "step": 31323 }, { "epoch": 9.64, "learning_rate": 7.011567198780155e-08, "loss": 0.0169, "step": 31324 }, { "epoch": 9.64, "learning_rate": 6.999791873756767e-08, "loss": 0.019, "step": 31325 }, { "epoch": 9.64, "learning_rate": 6.988026410140625e-08, "loss": 0.0205, "step": 31326 }, { "epoch": 9.64, "learning_rate": 6.976270808048747e-08, "loss": 0.0351, "step": 31327 }, { "epoch": 9.64, "learning_rate": 6.964525067597484e-08, "loss": 0.0347, "step": 31328 }, { "epoch": 9.64, "learning_rate": 6.952789188903742e-08, "loss": 0.0324, "step": 31329 }, { "epoch": 9.64, "learning_rate": 6.941063172084094e-08, "loss": 0.0254, "step": 31330 }, { "epoch": 9.64, "learning_rate": 6.929347017254894e-08, "loss": 0.0224, "step": 31331 }, { "epoch": 9.64, "learning_rate": 6.917640724532493e-08, "loss": 0.0272, "step": 31332 }, { "epoch": 9.64, "learning_rate": 6.90594429403324e-08, "loss": 0.0177, "step": 31333 }, { "epoch": 9.64, "learning_rate": 6.894257725873155e-08, "loss": 0.0219, "step": 31334 }, { "epoch": 9.64, "learning_rate": 6.882581020168366e-08, "loss": 0.016, "step": 31335 }, { "epoch": 9.64, "learning_rate": 6.870914177034893e-08, "loss": 0.0169, "step": 31336 }, { "epoch": 9.64, "learning_rate": 6.859257196588532e-08, "loss": 0.0237, "step": 31337 }, { "epoch": 9.64, "learning_rate": 6.847610078944966e-08, "loss": 0.018, "step": 31338 }, { "epoch": 9.64, "learning_rate": 6.835972824219994e-08, "loss": 0.0347, "step": 31339 }, { "epoch": 9.64, "learning_rate": 6.824345432529078e-08, "loss": 0.0152, "step": 31340 }, { "epoch": 9.64, "learning_rate": 6.812727903987903e-08, "loss": 0.0269, "step": 31341 }, { "epoch": 9.64, "learning_rate": 6.8011202387116e-08, "loss": 0.0197, "step": 31342 }, { "epoch": 9.64, "learning_rate": 6.78952243681541e-08, "loss": 0.0173, "step": 31343 }, { "epoch": 9.64, "learning_rate": 6.777934498414907e-08, "loss": 0.0289, "step": 31344 }, { "epoch": 9.64, "learning_rate": 6.766356423624664e-08, "loss": 0.0414, "step": 31345 }, { "epoch": 9.64, "learning_rate": 6.754788212559926e-08, "loss": 0.0335, "step": 31346 }, { "epoch": 9.64, "learning_rate": 6.743229865335599e-08, "loss": 0.0299, "step": 31347 }, { "epoch": 9.65, "learning_rate": 6.731681382066369e-08, "loss": 0.023, "step": 31348 }, { "epoch": 9.65, "learning_rate": 6.720142762867032e-08, "loss": 0.0292, "step": 31349 }, { "epoch": 9.65, "learning_rate": 6.708614007852055e-08, "loss": 0.0151, "step": 31350 }, { "epoch": 9.65, "learning_rate": 6.69709511713601e-08, "loss": 0.0127, "step": 31351 }, { "epoch": 9.65, "learning_rate": 6.685586090833252e-08, "loss": 0.0191, "step": 31352 }, { "epoch": 9.65, "learning_rate": 6.674086929058133e-08, "loss": 0.0265, "step": 31353 }, { "epoch": 9.65, "learning_rate": 6.662597631924783e-08, "loss": 0.0092, "step": 31354 }, { "epoch": 9.65, "learning_rate": 6.651118199547224e-08, "loss": 0.0139, "step": 31355 }, { "epoch": 9.65, "learning_rate": 6.639648632039697e-08, "loss": 0.0267, "step": 31356 }, { "epoch": 9.65, "learning_rate": 6.628188929515888e-08, "loss": 0.032, "step": 31357 }, { "epoch": 9.65, "learning_rate": 6.616739092089708e-08, "loss": 0.027, "step": 31358 }, { "epoch": 9.65, "learning_rate": 6.605299119874953e-08, "loss": 0.0326, "step": 31359 }, { "epoch": 9.65, "learning_rate": 6.593869012984866e-08, "loss": 0.0275, "step": 31360 }, { "epoch": 9.65, "learning_rate": 6.582448771533468e-08, "loss": 0.0254, "step": 31361 }, { "epoch": 9.65, "learning_rate": 6.571038395633778e-08, "loss": 0.0236, "step": 31362 }, { "epoch": 9.65, "learning_rate": 6.559637885399262e-08, "loss": 0.0299, "step": 31363 }, { "epoch": 9.65, "learning_rate": 6.548247240943051e-08, "loss": 0.0151, "step": 31364 }, { "epoch": 9.65, "learning_rate": 6.536866462378499e-08, "loss": 0.0277, "step": 31365 }, { "epoch": 9.65, "learning_rate": 6.525495549818295e-08, "loss": 0.0084, "step": 31366 }, { "epoch": 9.65, "learning_rate": 6.51413450337568e-08, "loss": 0.0347, "step": 31367 }, { "epoch": 9.65, "learning_rate": 6.502783323163231e-08, "loss": 0.0188, "step": 31368 }, { "epoch": 9.65, "learning_rate": 6.491442009293858e-08, "loss": 0.0258, "step": 31369 }, { "epoch": 9.65, "learning_rate": 6.480110561880026e-08, "loss": 0.0202, "step": 31370 }, { "epoch": 9.65, "learning_rate": 6.468788981034424e-08, "loss": 0.0096, "step": 31371 }, { "epoch": 9.65, "learning_rate": 6.457477266869405e-08, "loss": 0.0188, "step": 31372 }, { "epoch": 9.65, "learning_rate": 6.446175419497214e-08, "loss": 0.0135, "step": 31373 }, { "epoch": 9.65, "learning_rate": 6.434883439030316e-08, "loss": 0.0256, "step": 31374 }, { "epoch": 9.65, "learning_rate": 6.42360132558062e-08, "loss": 0.0196, "step": 31375 }, { "epoch": 9.65, "learning_rate": 6.412329079260371e-08, "loss": 0.0205, "step": 31376 }, { "epoch": 9.65, "learning_rate": 6.401066700181368e-08, "loss": 0.0276, "step": 31377 }, { "epoch": 9.65, "learning_rate": 6.389814188455523e-08, "loss": 0.0176, "step": 31378 }, { "epoch": 9.65, "learning_rate": 6.378571544194522e-08, "loss": 0.0261, "step": 31379 }, { "epoch": 9.66, "learning_rate": 6.367338767509945e-08, "loss": 0.0141, "step": 31380 }, { "epoch": 9.66, "learning_rate": 6.356115858513589e-08, "loss": 0.043, "step": 31381 }, { "epoch": 9.66, "learning_rate": 6.344902817316811e-08, "loss": 0.0148, "step": 31382 }, { "epoch": 9.66, "learning_rate": 6.333699644030855e-08, "loss": 0.0167, "step": 31383 }, { "epoch": 9.66, "learning_rate": 6.322506338767076e-08, "loss": 0.0322, "step": 31384 }, { "epoch": 9.66, "learning_rate": 6.311322901636607e-08, "loss": 0.0213, "step": 31385 }, { "epoch": 9.66, "learning_rate": 6.300149332750472e-08, "loss": 0.0205, "step": 31386 }, { "epoch": 9.66, "learning_rate": 6.28898563221969e-08, "loss": 0.0302, "step": 31387 }, { "epoch": 9.66, "learning_rate": 6.277831800155176e-08, "loss": 0.0183, "step": 31388 }, { "epoch": 9.66, "learning_rate": 6.266687836667618e-08, "loss": 0.0169, "step": 31389 }, { "epoch": 9.66, "learning_rate": 6.255553741867593e-08, "loss": 0.0087, "step": 31390 }, { "epoch": 9.66, "learning_rate": 6.244429515865902e-08, "loss": 0.0156, "step": 31391 }, { "epoch": 9.66, "learning_rate": 6.23331515877279e-08, "loss": 0.0257, "step": 31392 }, { "epoch": 9.66, "learning_rate": 6.222210670698836e-08, "loss": 0.0142, "step": 31393 }, { "epoch": 9.66, "learning_rate": 6.211116051754174e-08, "loss": 0.0156, "step": 31394 }, { "epoch": 9.66, "learning_rate": 6.200031302049048e-08, "loss": 0.0168, "step": 31395 }, { "epoch": 9.66, "learning_rate": 6.188956421693592e-08, "loss": 0.0268, "step": 31396 }, { "epoch": 9.66, "learning_rate": 6.17789141079761e-08, "loss": 0.0132, "step": 31397 }, { "epoch": 9.66, "learning_rate": 6.166836269471122e-08, "loss": 0.0161, "step": 31398 }, { "epoch": 9.66, "learning_rate": 6.15579099782404e-08, "loss": 0.0213, "step": 31399 }, { "epoch": 9.66, "learning_rate": 6.144755595965724e-08, "loss": 0.0274, "step": 31400 }, { "epoch": 9.66, "learning_rate": 6.133730064005971e-08, "loss": 0.022, "step": 31401 }, { "epoch": 9.66, "learning_rate": 6.122714402054252e-08, "loss": 0.0175, "step": 31402 }, { "epoch": 9.66, "learning_rate": 6.111708610220035e-08, "loss": 0.0165, "step": 31403 }, { "epoch": 9.66, "learning_rate": 6.100712688612564e-08, "loss": 0.0147, "step": 31404 }, { "epoch": 9.66, "learning_rate": 6.089726637340864e-08, "loss": 0.0287, "step": 31405 }, { "epoch": 9.66, "learning_rate": 6.078750456514404e-08, "loss": 0.0261, "step": 31406 }, { "epoch": 9.66, "learning_rate": 6.067784146241873e-08, "loss": 0.0328, "step": 31407 }, { "epoch": 9.66, "learning_rate": 6.056827706632185e-08, "loss": 0.0325, "step": 31408 }, { "epoch": 9.66, "learning_rate": 6.045881137794363e-08, "loss": 0.017, "step": 31409 }, { "epoch": 9.66, "learning_rate": 6.034944439836987e-08, "loss": 0.0138, "step": 31410 }, { "epoch": 9.66, "learning_rate": 6.024017612868637e-08, "loss": 0.0243, "step": 31411 }, { "epoch": 9.66, "learning_rate": 6.013100656997895e-08, "loss": 0.02, "step": 31412 }, { "epoch": 9.67, "learning_rate": 6.002193572333226e-08, "loss": 0.0183, "step": 31413 }, { "epoch": 9.67, "learning_rate": 5.991296358982767e-08, "loss": 0.0325, "step": 31414 }, { "epoch": 9.67, "learning_rate": 5.980409017054767e-08, "loss": 0.0481, "step": 31415 }, { "epoch": 9.67, "learning_rate": 5.969531546657581e-08, "loss": 0.0169, "step": 31416 }, { "epoch": 9.67, "learning_rate": 5.958663947899013e-08, "loss": 0.0139, "step": 31417 }, { "epoch": 9.67, "learning_rate": 5.9478062208870865e-08, "loss": 0.0254, "step": 31418 }, { "epoch": 9.67, "learning_rate": 5.936958365729495e-08, "loss": 0.0404, "step": 31419 }, { "epoch": 9.67, "learning_rate": 5.9261203825341504e-08, "loss": 0.036, "step": 31420 }, { "epoch": 9.67, "learning_rate": 5.915292271408524e-08, "loss": 0.0184, "step": 31421 }, { "epoch": 9.67, "learning_rate": 5.9044740324603057e-08, "loss": 0.0245, "step": 31422 }, { "epoch": 9.67, "learning_rate": 5.893665665796744e-08, "loss": 0.0219, "step": 31423 }, { "epoch": 9.67, "learning_rate": 5.8828671715253085e-08, "loss": 0.0408, "step": 31424 }, { "epoch": 9.67, "learning_rate": 5.8720785497532464e-08, "loss": 0.0236, "step": 31425 }, { "epoch": 9.67, "learning_rate": 5.861299800587583e-08, "loss": 0.0289, "step": 31426 }, { "epoch": 9.67, "learning_rate": 5.850530924135567e-08, "loss": 0.0231, "step": 31427 }, { "epoch": 9.67, "learning_rate": 5.8397719205038894e-08, "loss": 0.034, "step": 31428 }, { "epoch": 9.67, "learning_rate": 5.829022789799577e-08, "loss": 0.0087, "step": 31429 }, { "epoch": 9.67, "learning_rate": 5.8182835321293205e-08, "loss": 0.023, "step": 31430 }, { "epoch": 9.67, "learning_rate": 5.8075541475998145e-08, "loss": 0.0141, "step": 31431 }, { "epoch": 9.67, "learning_rate": 5.7968346363175274e-08, "loss": 0.0202, "step": 31432 }, { "epoch": 9.67, "learning_rate": 5.786124998388931e-08, "loss": 0.0189, "step": 31433 }, { "epoch": 9.67, "learning_rate": 5.7754252339204955e-08, "loss": 0.036, "step": 31434 }, { "epoch": 9.67, "learning_rate": 5.764735343018246e-08, "loss": 0.018, "step": 31435 }, { "epoch": 9.67, "learning_rate": 5.7540553257886545e-08, "loss": 0.0196, "step": 31436 }, { "epoch": 9.67, "learning_rate": 5.743385182337524e-08, "loss": 0.0238, "step": 31437 }, { "epoch": 9.67, "learning_rate": 5.7327249127708816e-08, "loss": 0.0125, "step": 31438 }, { "epoch": 9.67, "learning_rate": 5.722074517194753e-08, "loss": 0.0255, "step": 31439 }, { "epoch": 9.67, "learning_rate": 5.71143399571461e-08, "loss": 0.0188, "step": 31440 }, { "epoch": 9.67, "learning_rate": 5.700803348436479e-08, "loss": 0.0159, "step": 31441 }, { "epoch": 9.67, "learning_rate": 5.690182575465608e-08, "loss": 0.0171, "step": 31442 }, { "epoch": 9.67, "learning_rate": 5.6795716769075803e-08, "loss": 0.0277, "step": 31443 }, { "epoch": 9.67, "learning_rate": 5.668970652867867e-08, "loss": 0.0147, "step": 31444 }, { "epoch": 9.68, "learning_rate": 5.6583795034517165e-08, "loss": 0.0406, "step": 31445 }, { "epoch": 9.68, "learning_rate": 5.647798228764156e-08, "loss": 0.019, "step": 31446 }, { "epoch": 9.68, "learning_rate": 5.637226828910436e-08, "loss": 0.0193, "step": 31447 }, { "epoch": 9.68, "learning_rate": 5.6266653039955823e-08, "loss": 0.0199, "step": 31448 }, { "epoch": 9.68, "learning_rate": 5.616113654124289e-08, "loss": 0.0203, "step": 31449 }, { "epoch": 9.68, "learning_rate": 5.6055718794013613e-08, "loss": 0.0217, "step": 31450 }, { "epoch": 9.68, "learning_rate": 5.595039979931716e-08, "loss": 0.0261, "step": 31451 }, { "epoch": 9.68, "learning_rate": 5.584517955819713e-08, "loss": 0.0321, "step": 31452 }, { "epoch": 9.68, "learning_rate": 5.5740058071699354e-08, "loss": 0.042, "step": 31453 }, { "epoch": 9.68, "learning_rate": 5.563503534086745e-08, "loss": 0.0118, "step": 31454 }, { "epoch": 9.68, "learning_rate": 5.5530111366745024e-08, "loss": 0.023, "step": 31455 }, { "epoch": 9.68, "learning_rate": 5.5425286150374567e-08, "loss": 0.0186, "step": 31456 }, { "epoch": 9.68, "learning_rate": 5.532055969279526e-08, "loss": 0.0228, "step": 31457 }, { "epoch": 9.68, "learning_rate": 5.521593199504738e-08, "loss": 0.0255, "step": 31458 }, { "epoch": 9.68, "learning_rate": 5.5111403058172306e-08, "loss": 0.0291, "step": 31459 }, { "epoch": 9.68, "learning_rate": 5.5006972883204776e-08, "loss": 0.038, "step": 31460 }, { "epoch": 9.68, "learning_rate": 5.490264147118507e-08, "loss": 0.0313, "step": 31461 }, { "epoch": 9.68, "learning_rate": 5.47984088231468e-08, "loss": 0.0156, "step": 31462 }, { "epoch": 9.68, "learning_rate": 5.46942749401258e-08, "loss": 0.0216, "step": 31463 }, { "epoch": 9.68, "learning_rate": 5.459023982315792e-08, "loss": 0.0254, "step": 31464 }, { "epoch": 9.68, "learning_rate": 5.448630347327344e-08, "loss": 0.052, "step": 31465 }, { "epoch": 9.68, "learning_rate": 5.4382465891507083e-08, "loss": 0.0307, "step": 31466 }, { "epoch": 9.68, "learning_rate": 5.427872707888804e-08, "loss": 0.0124, "step": 31467 }, { "epoch": 9.68, "learning_rate": 5.4175087036448805e-08, "loss": 0.021, "step": 31468 }, { "epoch": 9.68, "learning_rate": 5.407154576521634e-08, "loss": 0.027, "step": 31469 }, { "epoch": 9.68, "learning_rate": 5.3968103266219815e-08, "loss": 0.0197, "step": 31470 }, { "epoch": 9.68, "learning_rate": 5.386475954048731e-08, "loss": 0.0259, "step": 31471 }, { "epoch": 9.68, "learning_rate": 5.376151458904355e-08, "loss": 0.0094, "step": 31472 }, { "epoch": 9.68, "learning_rate": 5.365836841291439e-08, "loss": 0.0149, "step": 31473 }, { "epoch": 9.68, "learning_rate": 5.355532101312566e-08, "loss": 0.0189, "step": 31474 }, { "epoch": 9.68, "learning_rate": 5.345237239069878e-08, "loss": 0.0356, "step": 31475 }, { "epoch": 9.68, "learning_rate": 5.3349522546657374e-08, "loss": 0.026, "step": 31476 }, { "epoch": 9.68, "learning_rate": 5.324677148202173e-08, "loss": 0.0122, "step": 31477 }, { "epoch": 9.69, "learning_rate": 5.314411919781326e-08, "loss": 0.0278, "step": 31478 }, { "epoch": 9.69, "learning_rate": 5.3041565695050034e-08, "loss": 0.019, "step": 31479 }, { "epoch": 9.69, "learning_rate": 5.293911097475235e-08, "loss": 0.0224, "step": 31480 }, { "epoch": 9.69, "learning_rate": 5.283675503793606e-08, "loss": 0.033, "step": 31481 }, { "epoch": 9.69, "learning_rate": 5.273449788561924e-08, "loss": 0.0263, "step": 31482 }, { "epoch": 9.69, "learning_rate": 5.263233951881663e-08, "loss": 0.0265, "step": 31483 }, { "epoch": 9.69, "learning_rate": 5.2530279938541875e-08, "loss": 0.0315, "step": 31484 }, { "epoch": 9.69, "learning_rate": 5.2428319145809705e-08, "loss": 0.0148, "step": 31485 }, { "epoch": 9.69, "learning_rate": 5.232645714163265e-08, "loss": 0.0379, "step": 31486 }, { "epoch": 9.69, "learning_rate": 5.2224693927021006e-08, "loss": 0.0203, "step": 31487 }, { "epoch": 9.69, "learning_rate": 5.21230295029862e-08, "loss": 0.0251, "step": 31488 }, { "epoch": 9.69, "learning_rate": 5.202146387053963e-08, "loss": 0.0205, "step": 31489 }, { "epoch": 9.69, "learning_rate": 5.1919997030686065e-08, "loss": 0.0327, "step": 31490 }, { "epoch": 9.69, "learning_rate": 5.18186289844369e-08, "loss": 0.0251, "step": 31491 }, { "epoch": 9.69, "learning_rate": 5.17173597327969e-08, "loss": 0.0224, "step": 31492 }, { "epoch": 9.69, "learning_rate": 5.161618927677192e-08, "loss": 0.0276, "step": 31493 }, { "epoch": 9.69, "learning_rate": 5.1515117617366716e-08, "loss": 0.0233, "step": 31494 }, { "epoch": 9.69, "learning_rate": 5.141414475558604e-08, "loss": 0.0225, "step": 31495 }, { "epoch": 9.69, "learning_rate": 5.131327069243131e-08, "loss": 0.0208, "step": 31496 }, { "epoch": 9.69, "learning_rate": 5.121249542890394e-08, "loss": 0.0181, "step": 31497 }, { "epoch": 9.69, "learning_rate": 5.1111818966006477e-08, "loss": 0.0179, "step": 31498 }, { "epoch": 9.69, "learning_rate": 5.1011241304738115e-08, "loss": 0.0205, "step": 31499 }, { "epoch": 9.69, "learning_rate": 5.0910762446095826e-08, "loss": 0.0193, "step": 31500 }, { "epoch": 9.69, "learning_rate": 5.0810382391081047e-08, "loss": 0.0122, "step": 31501 }, { "epoch": 9.69, "learning_rate": 5.071010114068742e-08, "loss": 0.0163, "step": 31502 }, { "epoch": 9.69, "learning_rate": 5.060991869591192e-08, "loss": 0.0292, "step": 31503 }, { "epoch": 9.69, "learning_rate": 5.0509835057750426e-08, "loss": 0.0328, "step": 31504 }, { "epoch": 9.69, "learning_rate": 5.040985022719436e-08, "loss": 0.0266, "step": 31505 }, { "epoch": 9.69, "learning_rate": 5.03099642052407e-08, "loss": 0.0117, "step": 31506 }, { "epoch": 9.69, "learning_rate": 5.0210176992876444e-08, "loss": 0.0235, "step": 31507 }, { "epoch": 9.69, "learning_rate": 5.011048859109635e-08, "loss": 0.0232, "step": 31508 }, { "epoch": 9.69, "learning_rate": 5.0010899000889624e-08, "loss": 0.0261, "step": 31509 }, { "epoch": 9.7, "learning_rate": 4.991140822324436e-08, "loss": 0.0262, "step": 31510 }, { "epoch": 9.7, "learning_rate": 4.981201625914978e-08, "loss": 0.0205, "step": 31511 }, { "epoch": 9.7, "learning_rate": 4.9712723109590636e-08, "loss": 0.0268, "step": 31512 }, { "epoch": 9.7, "learning_rate": 4.9613528775556143e-08, "loss": 0.0224, "step": 31513 }, { "epoch": 9.7, "learning_rate": 4.9514433258028846e-08, "loss": 0.0238, "step": 31514 }, { "epoch": 9.7, "learning_rate": 4.9415436557994637e-08, "loss": 0.0245, "step": 31515 }, { "epoch": 9.7, "learning_rate": 4.931653867643604e-08, "loss": 0.0135, "step": 31516 }, { "epoch": 9.7, "learning_rate": 4.9217739614333406e-08, "loss": 0.0187, "step": 31517 }, { "epoch": 9.7, "learning_rate": 4.911903937267148e-08, "loss": 0.0331, "step": 31518 }, { "epoch": 9.7, "learning_rate": 4.902043795242728e-08, "loss": 0.0262, "step": 31519 }, { "epoch": 9.7, "learning_rate": 4.892193535458223e-08, "loss": 0.0327, "step": 31520 }, { "epoch": 9.7, "learning_rate": 4.882353158011221e-08, "loss": 0.0211, "step": 31521 }, { "epoch": 9.7, "learning_rate": 4.872522662999646e-08, "loss": 0.0218, "step": 31522 }, { "epoch": 9.7, "learning_rate": 4.8627020505210845e-08, "loss": 0.0324, "step": 31523 }, { "epoch": 9.7, "learning_rate": 4.852891320673014e-08, "loss": 0.0252, "step": 31524 }, { "epoch": 9.7, "learning_rate": 4.843090473552914e-08, "loss": 0.0138, "step": 31525 }, { "epoch": 9.7, "learning_rate": 4.833299509258038e-08, "loss": 0.0502, "step": 31526 }, { "epoch": 9.7, "learning_rate": 4.823518427885754e-08, "loss": 0.029, "step": 31527 }, { "epoch": 9.7, "learning_rate": 4.813747229532984e-08, "loss": 0.0305, "step": 31528 }, { "epoch": 9.7, "learning_rate": 4.803985914296983e-08, "loss": 0.0331, "step": 31529 }, { "epoch": 9.7, "learning_rate": 4.7942344822746736e-08, "loss": 0.0214, "step": 31530 }, { "epoch": 9.7, "learning_rate": 4.784492933562645e-08, "loss": 0.0209, "step": 31531 }, { "epoch": 9.7, "learning_rate": 4.7747612682579324e-08, "loss": 0.0257, "step": 31532 }, { "epoch": 9.7, "learning_rate": 4.765039486457013e-08, "loss": 0.0228, "step": 31533 }, { "epoch": 9.7, "learning_rate": 4.755327588256586e-08, "loss": 0.033, "step": 31534 }, { "epoch": 9.7, "learning_rate": 4.74562557375291e-08, "loss": 0.0376, "step": 31535 }, { "epoch": 9.7, "learning_rate": 4.735933443042462e-08, "loss": 0.014, "step": 31536 }, { "epoch": 9.7, "learning_rate": 4.7262511962213874e-08, "loss": 0.0237, "step": 31537 }, { "epoch": 9.7, "learning_rate": 4.716578833386054e-08, "loss": 0.0226, "step": 31538 }, { "epoch": 9.7, "learning_rate": 4.7069163546322737e-08, "loss": 0.0278, "step": 31539 }, { "epoch": 9.7, "learning_rate": 4.697263760056081e-08, "loss": 0.0153, "step": 31540 }, { "epoch": 9.7, "learning_rate": 4.6876210497533994e-08, "loss": 0.0188, "step": 31541 }, { "epoch": 9.7, "learning_rate": 4.677988223819929e-08, "loss": 0.0284, "step": 31542 }, { "epoch": 9.71, "learning_rate": 4.6683652823513725e-08, "loss": 0.0155, "step": 31543 }, { "epoch": 9.71, "learning_rate": 4.658752225443319e-08, "loss": 0.0168, "step": 31544 }, { "epoch": 9.71, "learning_rate": 4.649149053191138e-08, "loss": 0.0557, "step": 31545 }, { "epoch": 9.71, "learning_rate": 4.639555765690196e-08, "loss": 0.0263, "step": 31546 }, { "epoch": 9.71, "learning_rate": 4.6299723630358616e-08, "loss": 0.0264, "step": 31547 }, { "epoch": 9.71, "learning_rate": 4.6203988453232816e-08, "loss": 0.0249, "step": 31548 }, { "epoch": 9.71, "learning_rate": 4.610835212647491e-08, "loss": 0.0153, "step": 31549 }, { "epoch": 9.71, "learning_rate": 4.601281465103413e-08, "loss": 0.0476, "step": 31550 }, { "epoch": 9.71, "learning_rate": 4.5917376027861945e-08, "loss": 0.0268, "step": 31551 }, { "epoch": 9.71, "learning_rate": 4.582203625790205e-08, "loss": 0.0132, "step": 31552 }, { "epoch": 9.71, "learning_rate": 4.572679534210478e-08, "loss": 0.0309, "step": 31553 }, { "epoch": 9.71, "learning_rate": 4.563165328141383e-08, "loss": 0.024, "step": 31554 }, { "epoch": 9.71, "learning_rate": 4.553661007677512e-08, "loss": 0.019, "step": 31555 }, { "epoch": 9.71, "learning_rate": 4.5441665729132334e-08, "loss": 0.0197, "step": 31556 }, { "epoch": 9.71, "learning_rate": 4.534682023942805e-08, "loss": 0.0266, "step": 31557 }, { "epoch": 9.71, "learning_rate": 4.525207360860373e-08, "loss": 0.0238, "step": 31558 }, { "epoch": 9.71, "learning_rate": 4.515742583760196e-08, "loss": 0.0188, "step": 31559 }, { "epoch": 9.71, "learning_rate": 4.506287692735978e-08, "loss": 0.0196, "step": 31560 }, { "epoch": 9.71, "learning_rate": 4.496842687881975e-08, "loss": 0.0239, "step": 31561 }, { "epoch": 9.71, "learning_rate": 4.4874075692917793e-08, "loss": 0.0137, "step": 31562 }, { "epoch": 9.71, "learning_rate": 4.4779823370590945e-08, "loss": 0.0113, "step": 31563 }, { "epoch": 9.71, "learning_rate": 4.468566991277512e-08, "loss": 0.0187, "step": 31564 }, { "epoch": 9.71, "learning_rate": 4.459161532040623e-08, "loss": 0.0273, "step": 31565 }, { "epoch": 9.71, "learning_rate": 4.449765959441688e-08, "loss": 0.0151, "step": 31566 }, { "epoch": 9.71, "learning_rate": 4.440380273574185e-08, "loss": 0.0328, "step": 31567 }, { "epoch": 9.71, "learning_rate": 4.431004474531264e-08, "loss": 0.0212, "step": 31568 }, { "epoch": 9.71, "learning_rate": 4.421638562405961e-08, "loss": 0.0278, "step": 31569 }, { "epoch": 9.71, "learning_rate": 4.412282537291312e-08, "loss": 0.02, "step": 31570 }, { "epoch": 9.71, "learning_rate": 4.4029363992802444e-08, "loss": 0.0167, "step": 31571 }, { "epoch": 9.71, "learning_rate": 4.393600148465682e-08, "loss": 0.0234, "step": 31572 }, { "epoch": 9.71, "learning_rate": 4.384273784940218e-08, "loss": 0.0081, "step": 31573 }, { "epoch": 9.71, "learning_rate": 4.374957308796446e-08, "loss": 0.0225, "step": 31574 }, { "epoch": 9.72, "learning_rate": 4.365650720126957e-08, "loss": 0.0224, "step": 31575 }, { "epoch": 9.72, "learning_rate": 4.356354019024234e-08, "loss": 0.0171, "step": 31576 }, { "epoch": 9.72, "learning_rate": 4.347067205580424e-08, "loss": 0.0139, "step": 31577 }, { "epoch": 9.72, "learning_rate": 4.33779027988801e-08, "loss": 0.015, "step": 31578 }, { "epoch": 9.72, "learning_rate": 4.328523242038807e-08, "loss": 0.0233, "step": 31579 }, { "epoch": 9.72, "learning_rate": 4.319266092125074e-08, "loss": 0.0194, "step": 31580 }, { "epoch": 9.72, "learning_rate": 4.310018830238627e-08, "loss": 0.0194, "step": 31581 }, { "epoch": 9.72, "learning_rate": 4.3007814564713925e-08, "loss": 0.0277, "step": 31582 }, { "epoch": 9.72, "learning_rate": 4.291553970915074e-08, "loss": 0.0236, "step": 31583 }, { "epoch": 9.72, "learning_rate": 4.282336373661378e-08, "loss": 0.0172, "step": 31584 }, { "epoch": 9.72, "learning_rate": 4.273128664801563e-08, "loss": 0.027, "step": 31585 }, { "epoch": 9.72, "learning_rate": 4.263930844427444e-08, "loss": 0.0228, "step": 31586 }, { "epoch": 9.72, "learning_rate": 4.254742912630061e-08, "loss": 0.0248, "step": 31587 }, { "epoch": 9.72, "learning_rate": 4.2455648695008956e-08, "loss": 0.0346, "step": 31588 }, { "epoch": 9.72, "learning_rate": 4.236396715130986e-08, "loss": 0.0141, "step": 31589 }, { "epoch": 9.72, "learning_rate": 4.2272384496112597e-08, "loss": 0.0165, "step": 31590 }, { "epoch": 9.72, "learning_rate": 4.2180900730328655e-08, "loss": 0.0147, "step": 31591 }, { "epoch": 9.72, "learning_rate": 4.208951585486509e-08, "loss": 0.0265, "step": 31592 }, { "epoch": 9.72, "learning_rate": 4.199822987063118e-08, "loss": 0.0322, "step": 31593 }, { "epoch": 9.72, "learning_rate": 4.190704277853175e-08, "loss": 0.0165, "step": 31594 }, { "epoch": 9.72, "learning_rate": 4.181595457947385e-08, "loss": 0.0178, "step": 31595 }, { "epoch": 9.72, "learning_rate": 4.1724965274360096e-08, "loss": 0.0189, "step": 31596 }, { "epoch": 9.72, "learning_rate": 4.163407486409532e-08, "loss": 0.0267, "step": 31597 }, { "epoch": 9.72, "learning_rate": 4.154328334958324e-08, "loss": 0.0165, "step": 31598 }, { "epoch": 9.72, "learning_rate": 4.145259073172314e-08, "loss": 0.0337, "step": 31599 }, { "epoch": 9.72, "learning_rate": 4.136199701141763e-08, "loss": 0.0165, "step": 31600 }, { "epoch": 9.72, "learning_rate": 4.127150218956488e-08, "loss": 0.0103, "step": 31601 }, { "epoch": 9.72, "learning_rate": 4.1181106267065284e-08, "loss": 0.0262, "step": 31602 }, { "epoch": 9.72, "learning_rate": 4.109080924481479e-08, "loss": 0.013, "step": 31603 }, { "epoch": 9.72, "learning_rate": 4.100061112371157e-08, "loss": 0.0179, "step": 31604 }, { "epoch": 9.72, "learning_rate": 4.091051190465045e-08, "loss": 0.015, "step": 31605 }, { "epoch": 9.72, "learning_rate": 4.0820511588527404e-08, "loss": 0.0246, "step": 31606 }, { "epoch": 9.72, "learning_rate": 4.073061017623392e-08, "loss": 0.0126, "step": 31607 }, { "epoch": 9.73, "learning_rate": 4.064080766866485e-08, "loss": 0.0461, "step": 31608 }, { "epoch": 9.73, "learning_rate": 4.0551104066711697e-08, "loss": 0.1145, "step": 31609 }, { "epoch": 9.73, "learning_rate": 4.0461499371263757e-08, "loss": 0.0323, "step": 31610 }, { "epoch": 9.73, "learning_rate": 4.0371993583213644e-08, "loss": 0.034, "step": 31611 }, { "epoch": 9.73, "learning_rate": 4.0282586703447315e-08, "loss": 0.0197, "step": 31612 }, { "epoch": 9.73, "learning_rate": 4.0193278732855166e-08, "loss": 0.0219, "step": 31613 }, { "epoch": 9.73, "learning_rate": 4.010406967232316e-08, "loss": 0.0255, "step": 31614 }, { "epoch": 9.73, "learning_rate": 4.0014959522736154e-08, "loss": 0.0224, "step": 31615 }, { "epoch": 9.73, "learning_rate": 3.9925948284980086e-08, "loss": 0.0299, "step": 31616 }, { "epoch": 9.73, "learning_rate": 3.983703595993871e-08, "loss": 0.0228, "step": 31617 }, { "epoch": 9.73, "learning_rate": 3.974822254849575e-08, "loss": 0.0108, "step": 31618 }, { "epoch": 9.73, "learning_rate": 3.965950805153274e-08, "loss": 0.0275, "step": 31619 }, { "epoch": 9.73, "learning_rate": 3.9570892469930064e-08, "loss": 0.0171, "step": 31620 }, { "epoch": 9.73, "learning_rate": 3.9482375804568154e-08, "loss": 0.0297, "step": 31621 }, { "epoch": 9.73, "learning_rate": 3.939395805632629e-08, "loss": 0.0184, "step": 31622 }, { "epoch": 9.73, "learning_rate": 3.930563922608266e-08, "loss": 0.0153, "step": 31623 }, { "epoch": 9.73, "learning_rate": 3.9217419314713236e-08, "loss": 0.02, "step": 31624 }, { "epoch": 9.73, "learning_rate": 3.912929832309509e-08, "loss": 0.0391, "step": 31625 }, { "epoch": 9.73, "learning_rate": 3.904127625210419e-08, "loss": 0.0361, "step": 31626 }, { "epoch": 9.73, "learning_rate": 3.8953353102613166e-08, "loss": 0.0274, "step": 31627 }, { "epoch": 9.73, "learning_rate": 3.886552887549577e-08, "loss": 0.0309, "step": 31628 }, { "epoch": 9.73, "learning_rate": 3.877780357162353e-08, "loss": 0.0167, "step": 31629 }, { "epoch": 9.73, "learning_rate": 3.8690177191869074e-08, "loss": 0.0358, "step": 31630 }, { "epoch": 9.73, "learning_rate": 3.8602649737101705e-08, "loss": 0.0274, "step": 31631 }, { "epoch": 9.73, "learning_rate": 3.8515221208189626e-08, "loss": 0.0058, "step": 31632 }, { "epoch": 9.73, "learning_rate": 3.8427891606003244e-08, "loss": 0.0317, "step": 31633 }, { "epoch": 9.73, "learning_rate": 3.8340660931408536e-08, "loss": 0.018, "step": 31634 }, { "epoch": 9.73, "learning_rate": 3.825352918527259e-08, "loss": 0.0205, "step": 31635 }, { "epoch": 9.73, "learning_rate": 3.816649636845915e-08, "loss": 0.0288, "step": 31636 }, { "epoch": 9.73, "learning_rate": 3.8079562481834196e-08, "loss": 0.0197, "step": 31637 }, { "epoch": 9.73, "learning_rate": 3.7992727526259264e-08, "loss": 0.0294, "step": 31638 }, { "epoch": 9.73, "learning_rate": 3.7905991502598106e-08, "loss": 0.0227, "step": 31639 }, { "epoch": 9.74, "learning_rate": 3.7819354411713364e-08, "loss": 0.0313, "step": 31640 }, { "epoch": 9.74, "learning_rate": 3.773281625446212e-08, "loss": 0.0192, "step": 31641 }, { "epoch": 9.74, "learning_rate": 3.764637703170593e-08, "loss": 0.0283, "step": 31642 }, { "epoch": 9.74, "learning_rate": 3.7560036744302977e-08, "loss": 0.0174, "step": 31643 }, { "epoch": 9.74, "learning_rate": 3.747379539311036e-08, "loss": 0.0208, "step": 31644 }, { "epoch": 9.74, "learning_rate": 3.738765297898517e-08, "loss": 0.0251, "step": 31645 }, { "epoch": 9.74, "learning_rate": 3.7301609502782277e-08, "loss": 0.0177, "step": 31646 }, { "epoch": 9.74, "learning_rate": 3.7215664965356554e-08, "loss": 0.0171, "step": 31647 }, { "epoch": 9.74, "learning_rate": 3.712981936756066e-08, "loss": 0.0321, "step": 31648 }, { "epoch": 9.74, "learning_rate": 3.704407271024724e-08, "loss": 0.017, "step": 31649 }, { "epoch": 9.74, "learning_rate": 3.695842499426894e-08, "loss": 0.0296, "step": 31650 }, { "epoch": 9.74, "learning_rate": 3.687287622047619e-08, "loss": 0.0342, "step": 31651 }, { "epoch": 9.74, "learning_rate": 3.678742638971722e-08, "loss": 0.0305, "step": 31652 }, { "epoch": 9.74, "learning_rate": 3.670207550284133e-08, "loss": 0.0342, "step": 31653 }, { "epoch": 9.74, "learning_rate": 3.661682356069674e-08, "loss": 0.0225, "step": 31654 }, { "epoch": 9.74, "learning_rate": 3.653167056413054e-08, "loss": 0.0193, "step": 31655 }, { "epoch": 9.74, "learning_rate": 3.644661651398651e-08, "loss": 0.0248, "step": 31656 }, { "epoch": 9.74, "learning_rate": 3.6361661411109525e-08, "loss": 0.0268, "step": 31657 }, { "epoch": 9.74, "learning_rate": 3.627680525634558e-08, "loss": 0.0274, "step": 31658 }, { "epoch": 9.74, "learning_rate": 3.6192048050534e-08, "loss": 0.0214, "step": 31659 }, { "epoch": 9.74, "learning_rate": 3.6107389794519666e-08, "loss": 0.0194, "step": 31660 }, { "epoch": 9.74, "learning_rate": 3.60228304891419e-08, "loss": 0.0293, "step": 31661 }, { "epoch": 9.74, "learning_rate": 3.593837013524004e-08, "loss": 0.0186, "step": 31662 }, { "epoch": 9.74, "learning_rate": 3.585400873365341e-08, "loss": 0.018, "step": 31663 }, { "epoch": 9.74, "learning_rate": 3.576974628521912e-08, "loss": 0.0249, "step": 31664 }, { "epoch": 9.74, "learning_rate": 3.5685582790775386e-08, "loss": 0.0242, "step": 31665 }, { "epoch": 9.74, "learning_rate": 3.5601518251157096e-08, "loss": 0.024, "step": 31666 }, { "epoch": 9.74, "learning_rate": 3.551755266719914e-08, "loss": 0.0181, "step": 31667 }, { "epoch": 9.74, "learning_rate": 3.543368603973529e-08, "loss": 0.0226, "step": 31668 }, { "epoch": 9.74, "learning_rate": 3.5349918369598223e-08, "loss": 0.0261, "step": 31669 }, { "epoch": 9.74, "learning_rate": 3.52662496576206e-08, "loss": 0.0273, "step": 31670 }, { "epoch": 9.74, "learning_rate": 3.518267990463287e-08, "loss": 0.0325, "step": 31671 }, { "epoch": 9.74, "learning_rate": 3.509920911146547e-08, "loss": 0.008, "step": 31672 }, { "epoch": 9.75, "learning_rate": 3.501583727894553e-08, "loss": 0.0252, "step": 31673 }, { "epoch": 9.75, "learning_rate": 3.493256440790349e-08, "loss": 0.0276, "step": 31674 }, { "epoch": 9.75, "learning_rate": 3.484939049916425e-08, "loss": 0.0182, "step": 31675 }, { "epoch": 9.75, "learning_rate": 3.476631555355603e-08, "loss": 0.0276, "step": 31676 }, { "epoch": 9.75, "learning_rate": 3.468333957190151e-08, "loss": 0.0198, "step": 31677 }, { "epoch": 9.75, "learning_rate": 3.4600462555025585e-08, "loss": 0.0275, "step": 31678 }, { "epoch": 9.75, "learning_rate": 3.451768450375204e-08, "loss": 0.0349, "step": 31679 }, { "epoch": 9.75, "learning_rate": 3.4435005418902436e-08, "loss": 0.0274, "step": 31680 }, { "epoch": 9.75, "learning_rate": 3.435242530129723e-08, "loss": 0.0203, "step": 31681 }, { "epoch": 9.75, "learning_rate": 3.426994415175688e-08, "loss": 0.0424, "step": 31682 }, { "epoch": 9.75, "learning_rate": 3.4187561971101844e-08, "loss": 0.0161, "step": 31683 }, { "epoch": 9.75, "learning_rate": 3.410527876014702e-08, "loss": 0.0304, "step": 31684 }, { "epoch": 9.75, "learning_rate": 3.4023094519713976e-08, "loss": 0.0301, "step": 31685 }, { "epoch": 9.75, "learning_rate": 3.394100925061539e-08, "loss": 0.021, "step": 31686 }, { "epoch": 9.75, "learning_rate": 3.3859022953667274e-08, "loss": 0.0274, "step": 31687 }, { "epoch": 9.75, "learning_rate": 3.377713562968454e-08, "loss": 0.0134, "step": 31688 }, { "epoch": 9.75, "learning_rate": 3.369534727947987e-08, "loss": 0.0243, "step": 31689 }, { "epoch": 9.75, "learning_rate": 3.3613657903865945e-08, "loss": 0.0242, "step": 31690 }, { "epoch": 9.75, "learning_rate": 3.353206750365323e-08, "loss": 0.0248, "step": 31691 }, { "epoch": 9.75, "learning_rate": 3.345057607965219e-08, "loss": 0.0233, "step": 31692 }, { "epoch": 9.75, "learning_rate": 3.3369183632673275e-08, "loss": 0.0212, "step": 31693 }, { "epoch": 9.75, "learning_rate": 3.3287890163523626e-08, "loss": 0.0263, "step": 31694 }, { "epoch": 9.75, "learning_rate": 3.320669567301038e-08, "loss": 0.0251, "step": 31695 }, { "epoch": 9.75, "learning_rate": 3.312560016194066e-08, "loss": 0.0245, "step": 31696 }, { "epoch": 9.75, "learning_rate": 3.304460363111939e-08, "loss": 0.0315, "step": 31697 }, { "epoch": 9.75, "learning_rate": 3.2963706081350357e-08, "loss": 0.0208, "step": 31698 }, { "epoch": 9.75, "learning_rate": 3.2882907513438475e-08, "loss": 0.0184, "step": 31699 }, { "epoch": 9.75, "learning_rate": 3.2802207928185336e-08, "loss": 0.0209, "step": 31700 }, { "epoch": 9.75, "learning_rate": 3.272160732639251e-08, "loss": 0.0262, "step": 31701 }, { "epoch": 9.75, "learning_rate": 3.2641105708859366e-08, "loss": 0.0227, "step": 31702 }, { "epoch": 9.75, "learning_rate": 3.256070307638637e-08, "loss": 0.0131, "step": 31703 }, { "epoch": 9.75, "learning_rate": 3.248039942977177e-08, "loss": 0.0317, "step": 31704 }, { "epoch": 9.76, "learning_rate": 3.240019476981382e-08, "loss": 0.0113, "step": 31705 }, { "epoch": 9.76, "learning_rate": 3.232008909730855e-08, "loss": 0.034, "step": 31706 }, { "epoch": 9.76, "learning_rate": 3.224008241304977e-08, "loss": 0.0151, "step": 31707 }, { "epoch": 9.76, "learning_rate": 3.2160174717834615e-08, "loss": 0.0133, "step": 31708 }, { "epoch": 9.76, "learning_rate": 3.208036601245579e-08, "loss": 0.0119, "step": 31709 }, { "epoch": 9.76, "learning_rate": 3.200065629770599e-08, "loss": 0.0249, "step": 31710 }, { "epoch": 9.76, "learning_rate": 3.19210455743757e-08, "loss": 0.0168, "step": 31711 }, { "epoch": 9.76, "learning_rate": 3.18415338432565e-08, "loss": 0.029, "step": 31712 }, { "epoch": 9.76, "learning_rate": 3.1762121105136653e-08, "loss": 0.0354, "step": 31713 }, { "epoch": 9.76, "learning_rate": 3.168280736080664e-08, "loss": 0.0278, "step": 31714 }, { "epoch": 9.76, "learning_rate": 3.160359261105472e-08, "loss": 0.0115, "step": 31715 }, { "epoch": 9.76, "learning_rate": 3.152447685666471e-08, "loss": 0.0214, "step": 31716 }, { "epoch": 9.76, "learning_rate": 3.144546009842486e-08, "loss": 0.0233, "step": 31717 }, { "epoch": 9.76, "learning_rate": 3.136654233711789e-08, "loss": 0.0282, "step": 31718 }, { "epoch": 9.76, "learning_rate": 3.128772357352983e-08, "loss": 0.0342, "step": 31719 }, { "epoch": 9.76, "learning_rate": 3.120900380844116e-08, "loss": 0.0289, "step": 31720 }, { "epoch": 9.76, "learning_rate": 3.1130383042634604e-08, "loss": 0.0256, "step": 31721 }, { "epoch": 9.76, "learning_rate": 3.105186127689064e-08, "loss": 0.0181, "step": 31722 }, { "epoch": 9.76, "learning_rate": 3.097343851198975e-08, "loss": 0.0257, "step": 31723 }, { "epoch": 9.76, "learning_rate": 3.0895114748710206e-08, "loss": 0.0316, "step": 31724 }, { "epoch": 9.76, "learning_rate": 3.0816889987830276e-08, "loss": 0.0188, "step": 31725 }, { "epoch": 9.76, "learning_rate": 3.073876423012712e-08, "loss": 0.0298, "step": 31726 }, { "epoch": 9.76, "learning_rate": 3.066073747637455e-08, "loss": 0.0151, "step": 31727 }, { "epoch": 9.76, "learning_rate": 3.0582809727349725e-08, "loss": 0.0225, "step": 31728 }, { "epoch": 9.76, "learning_rate": 3.0504980983825375e-08, "loss": 0.0199, "step": 31729 }, { "epoch": 9.76, "learning_rate": 3.0427251246575303e-08, "loss": 0.0245, "step": 31730 }, { "epoch": 9.76, "learning_rate": 3.034962051637003e-08, "loss": 0.0281, "step": 31731 }, { "epoch": 9.76, "learning_rate": 3.027208879398113e-08, "loss": 0.0214, "step": 31732 }, { "epoch": 9.76, "learning_rate": 3.019465608018024e-08, "loss": 0.0283, "step": 31733 }, { "epoch": 9.76, "learning_rate": 3.0117322375733394e-08, "loss": 0.0158, "step": 31734 }, { "epoch": 9.76, "learning_rate": 3.004008768141109e-08, "loss": 0.0276, "step": 31735 }, { "epoch": 9.76, "learning_rate": 2.996295199797827e-08, "loss": 0.0353, "step": 31736 }, { "epoch": 9.76, "learning_rate": 2.988591532620322e-08, "loss": 0.0273, "step": 31737 }, { "epoch": 9.77, "learning_rate": 2.9808977666848646e-08, "loss": 0.0188, "step": 31738 }, { "epoch": 9.77, "learning_rate": 2.9732139020680616e-08, "loss": 0.031, "step": 31739 }, { "epoch": 9.77, "learning_rate": 2.9655399388460736e-08, "loss": 0.0181, "step": 31740 }, { "epoch": 9.77, "learning_rate": 2.9578758770951732e-08, "loss": 0.0242, "step": 31741 }, { "epoch": 9.77, "learning_rate": 2.9502217168914104e-08, "loss": 0.0119, "step": 31742 }, { "epoch": 9.77, "learning_rate": 2.9425774583109467e-08, "loss": 0.0267, "step": 31743 }, { "epoch": 9.77, "learning_rate": 2.9349431014296105e-08, "loss": 0.0254, "step": 31744 }, { "epoch": 9.77, "learning_rate": 2.9273186463231185e-08, "loss": 0.0208, "step": 31745 }, { "epoch": 9.77, "learning_rate": 2.9197040930674102e-08, "loss": 0.0233, "step": 31746 }, { "epoch": 9.77, "learning_rate": 2.9120994417378702e-08, "loss": 0.0222, "step": 31747 }, { "epoch": 9.77, "learning_rate": 2.904504692410215e-08, "loss": 0.0234, "step": 31748 }, { "epoch": 9.77, "learning_rate": 2.896919845159607e-08, "loss": 0.0274, "step": 31749 }, { "epoch": 9.77, "learning_rate": 2.8893449000617636e-08, "loss": 0.041, "step": 31750 }, { "epoch": 9.77, "learning_rate": 2.8817798571916244e-08, "loss": 0.0245, "step": 31751 }, { "epoch": 9.77, "learning_rate": 2.8742247166242408e-08, "loss": 0.0171, "step": 31752 }, { "epoch": 9.77, "learning_rate": 2.866679478434886e-08, "loss": 0.0273, "step": 31753 }, { "epoch": 9.77, "learning_rate": 2.8591441426983892e-08, "loss": 0.0211, "step": 31754 }, { "epoch": 9.77, "learning_rate": 2.8516187094895787e-08, "loss": 0.0377, "step": 31755 }, { "epoch": 9.77, "learning_rate": 2.844103178883284e-08, "loss": 0.0259, "step": 31756 }, { "epoch": 9.77, "learning_rate": 2.8365975509538902e-08, "loss": 0.0171, "step": 31757 }, { "epoch": 9.77, "learning_rate": 2.8291018257762258e-08, "loss": 0.0211, "step": 31758 }, { "epoch": 9.77, "learning_rate": 2.8216160034244544e-08, "loss": 0.0368, "step": 31759 }, { "epoch": 9.77, "learning_rate": 2.8141400839731826e-08, "loss": 0.0246, "step": 31760 }, { "epoch": 9.77, "learning_rate": 2.806674067496573e-08, "loss": 0.0289, "step": 31761 }, { "epoch": 9.77, "learning_rate": 2.7992179540685672e-08, "loss": 0.0271, "step": 31762 }, { "epoch": 9.77, "learning_rate": 2.7917717437635495e-08, "loss": 0.0276, "step": 31763 }, { "epoch": 9.77, "learning_rate": 2.7843354366551278e-08, "loss": 0.0265, "step": 31764 }, { "epoch": 9.77, "learning_rate": 2.776909032817465e-08, "loss": 0.0147, "step": 31765 }, { "epoch": 9.77, "learning_rate": 2.7694925323240584e-08, "loss": 0.0193, "step": 31766 }, { "epoch": 9.77, "learning_rate": 2.7620859352487372e-08, "loss": 0.0182, "step": 31767 }, { "epoch": 9.77, "learning_rate": 2.7546892416649985e-08, "loss": 0.0257, "step": 31768 }, { "epoch": 9.77, "learning_rate": 2.7473024516461166e-08, "loss": 0.0212, "step": 31769 }, { "epoch": 9.78, "learning_rate": 2.7399255652658107e-08, "loss": 0.0352, "step": 31770 }, { "epoch": 9.78, "learning_rate": 2.7325585825970225e-08, "loss": 0.042, "step": 31771 }, { "epoch": 9.78, "learning_rate": 2.7252015037131373e-08, "loss": 0.027, "step": 31772 }, { "epoch": 9.78, "learning_rate": 2.7178543286870972e-08, "loss": 0.0151, "step": 31773 }, { "epoch": 9.78, "learning_rate": 2.7105170575918438e-08, "loss": 0.0177, "step": 31774 }, { "epoch": 9.78, "learning_rate": 2.7031896905003184e-08, "loss": 0.0271, "step": 31775 }, { "epoch": 9.78, "learning_rate": 2.695872227485241e-08, "loss": 0.025, "step": 31776 }, { "epoch": 9.78, "learning_rate": 2.6885646686192203e-08, "loss": 0.0344, "step": 31777 }, { "epoch": 9.78, "learning_rate": 2.6812670139749753e-08, "loss": 0.0245, "step": 31778 }, { "epoch": 9.78, "learning_rate": 2.673979263624893e-08, "loss": 0.0236, "step": 31779 }, { "epoch": 9.78, "learning_rate": 2.6667014176413598e-08, "loss": 0.0214, "step": 31780 }, { "epoch": 9.78, "learning_rate": 2.6594334760966512e-08, "loss": 0.0155, "step": 31781 }, { "epoch": 9.78, "learning_rate": 2.6521754390629316e-08, "loss": 0.0285, "step": 31782 }, { "epoch": 9.78, "learning_rate": 2.644927306612366e-08, "loss": 0.023, "step": 31783 }, { "epoch": 9.78, "learning_rate": 2.6376890788167853e-08, "loss": 0.0348, "step": 31784 }, { "epoch": 9.78, "learning_rate": 2.6304607557481322e-08, "loss": 0.0282, "step": 31785 }, { "epoch": 9.78, "learning_rate": 2.623242337478238e-08, "loss": 0.0156, "step": 31786 }, { "epoch": 9.78, "learning_rate": 2.616033824078823e-08, "loss": 0.0261, "step": 31787 }, { "epoch": 9.78, "learning_rate": 2.608835215621275e-08, "loss": 0.0139, "step": 31788 }, { "epoch": 9.78, "learning_rate": 2.6016465121773138e-08, "loss": 0.028, "step": 31789 }, { "epoch": 9.78, "learning_rate": 2.594467713818216e-08, "loss": 0.0244, "step": 31790 }, { "epoch": 9.78, "learning_rate": 2.587298820615258e-08, "loss": 0.0192, "step": 31791 }, { "epoch": 9.78, "learning_rate": 2.5801398326397163e-08, "loss": 0.0245, "step": 31792 }, { "epoch": 9.78, "learning_rate": 2.572990749962645e-08, "loss": 0.0332, "step": 31793 }, { "epoch": 9.78, "learning_rate": 2.565851572654876e-08, "loss": 0.0139, "step": 31794 }, { "epoch": 9.78, "learning_rate": 2.558722300787686e-08, "loss": 0.0364, "step": 31795 }, { "epoch": 9.78, "learning_rate": 2.551602934431574e-08, "loss": 0.0426, "step": 31796 }, { "epoch": 9.78, "learning_rate": 2.5444934736572614e-08, "loss": 0.0436, "step": 31797 }, { "epoch": 9.78, "learning_rate": 2.537393918535358e-08, "loss": 0.026, "step": 31798 }, { "epoch": 9.78, "learning_rate": 2.5303042691365853e-08, "loss": 0.0256, "step": 31799 }, { "epoch": 9.78, "learning_rate": 2.5232245255311095e-08, "loss": 0.022, "step": 31800 }, { "epoch": 9.78, "learning_rate": 2.5161546877893185e-08, "loss": 0.0181, "step": 31801 }, { "epoch": 9.78, "learning_rate": 2.5090947559813782e-08, "loss": 0.0256, "step": 31802 }, { "epoch": 9.79, "learning_rate": 2.5020447301774554e-08, "loss": 0.0144, "step": 31803 }, { "epoch": 9.79, "learning_rate": 2.4950046104476045e-08, "loss": 0.035, "step": 31804 }, { "epoch": 9.79, "learning_rate": 2.487974396861659e-08, "loss": 0.0221, "step": 31805 }, { "epoch": 9.79, "learning_rate": 2.480954089489451e-08, "loss": 0.0256, "step": 31806 }, { "epoch": 9.79, "learning_rate": 2.473943688400704e-08, "loss": 0.0361, "step": 31807 }, { "epoch": 9.79, "learning_rate": 2.466943193665028e-08, "loss": 0.0153, "step": 31808 }, { "epoch": 9.79, "learning_rate": 2.459952605352034e-08, "loss": 0.007, "step": 31809 }, { "epoch": 9.79, "learning_rate": 2.452971923531e-08, "loss": 0.0412, "step": 31810 }, { "epoch": 9.79, "learning_rate": 2.4460011482713153e-08, "loss": 0.0324, "step": 31811 }, { "epoch": 9.79, "learning_rate": 2.4390402796422574e-08, "loss": 0.0327, "step": 31812 }, { "epoch": 9.79, "learning_rate": 2.4320893177128822e-08, "loss": 0.0115, "step": 31813 }, { "epoch": 9.79, "learning_rate": 2.425148262552246e-08, "loss": 0.0298, "step": 31814 }, { "epoch": 9.79, "learning_rate": 2.4182171142292932e-08, "loss": 0.0217, "step": 31815 }, { "epoch": 9.79, "learning_rate": 2.411295872812858e-08, "loss": 0.0193, "step": 31816 }, { "epoch": 9.79, "learning_rate": 2.4043845383716623e-08, "loss": 0.0398, "step": 31817 }, { "epoch": 9.79, "learning_rate": 2.397483110974319e-08, "loss": 0.0112, "step": 31818 }, { "epoch": 9.79, "learning_rate": 2.3905915906893284e-08, "loss": 0.0438, "step": 31819 }, { "epoch": 9.79, "learning_rate": 2.3837099775853022e-08, "loss": 0.0252, "step": 31820 }, { "epoch": 9.79, "learning_rate": 2.3768382717304085e-08, "loss": 0.0248, "step": 31821 }, { "epoch": 9.79, "learning_rate": 2.3699764731929254e-08, "loss": 0.0092, "step": 31822 }, { "epoch": 9.79, "learning_rate": 2.363124582041021e-08, "loss": 0.0166, "step": 31823 }, { "epoch": 9.79, "learning_rate": 2.3562825983427517e-08, "loss": 0.0237, "step": 31824 }, { "epoch": 9.79, "learning_rate": 2.3494505221660635e-08, "loss": 0.0195, "step": 31825 }, { "epoch": 9.79, "learning_rate": 2.3426283535787907e-08, "loss": 0.0214, "step": 31826 }, { "epoch": 9.79, "learning_rate": 2.335816092648657e-08, "loss": 0.024, "step": 31827 }, { "epoch": 9.79, "learning_rate": 2.3290137394433865e-08, "loss": 0.0266, "step": 31828 }, { "epoch": 9.79, "learning_rate": 2.32222129403048e-08, "loss": 0.0146, "step": 31829 }, { "epoch": 9.79, "learning_rate": 2.3154387564774396e-08, "loss": 0.0136, "step": 31830 }, { "epoch": 9.79, "learning_rate": 2.3086661268514333e-08, "loss": 0.0236, "step": 31831 }, { "epoch": 9.79, "learning_rate": 2.3019034052200738e-08, "loss": 0.0272, "step": 31832 }, { "epoch": 9.79, "learning_rate": 2.295150591650197e-08, "loss": 0.0226, "step": 31833 }, { "epoch": 9.79, "learning_rate": 2.2884076862089712e-08, "loss": 0.0203, "step": 31834 }, { "epoch": 9.8, "learning_rate": 2.2816746889634532e-08, "loss": 0.0319, "step": 31835 }, { "epoch": 9.8, "learning_rate": 2.2749515999803683e-08, "loss": 0.0358, "step": 31836 }, { "epoch": 9.8, "learning_rate": 2.2682384193266625e-08, "loss": 0.0147, "step": 31837 }, { "epoch": 9.8, "learning_rate": 2.2615351470687276e-08, "loss": 0.0244, "step": 31838 }, { "epoch": 9.8, "learning_rate": 2.254841783273398e-08, "loss": 0.0301, "step": 31839 }, { "epoch": 9.8, "learning_rate": 2.2481583280069553e-08, "loss": 0.0261, "step": 31840 }, { "epoch": 9.8, "learning_rate": 2.241484781335901e-08, "loss": 0.0198, "step": 31841 }, { "epoch": 9.8, "learning_rate": 2.2348211433265155e-08, "loss": 0.015, "step": 31842 }, { "epoch": 9.8, "learning_rate": 2.2281674140448574e-08, "loss": 0.0311, "step": 31843 }, { "epoch": 9.8, "learning_rate": 2.2215235935570955e-08, "loss": 0.025, "step": 31844 }, { "epoch": 9.8, "learning_rate": 2.2148896819291776e-08, "loss": 0.0265, "step": 31845 }, { "epoch": 9.8, "learning_rate": 2.2082656792269398e-08, "loss": 0.014, "step": 31846 }, { "epoch": 9.8, "learning_rate": 2.201651585516218e-08, "loss": 0.0275, "step": 31847 }, { "epoch": 9.8, "learning_rate": 2.1950474008628487e-08, "loss": 0.0163, "step": 31848 }, { "epoch": 9.8, "learning_rate": 2.1884531253321128e-08, "loss": 0.0238, "step": 31849 }, { "epoch": 9.8, "learning_rate": 2.1818687589896248e-08, "loss": 0.0173, "step": 31850 }, { "epoch": 9.8, "learning_rate": 2.1752943019008876e-08, "loss": 0.0195, "step": 31851 }, { "epoch": 9.8, "learning_rate": 2.1687297541310715e-08, "loss": 0.0216, "step": 31852 }, { "epoch": 9.8, "learning_rate": 2.1621751157454574e-08, "loss": 0.0191, "step": 31853 }, { "epoch": 9.8, "learning_rate": 2.1556303868089933e-08, "loss": 0.014, "step": 31854 }, { "epoch": 9.8, "learning_rate": 2.1490955673867387e-08, "loss": 0.0241, "step": 31855 }, { "epoch": 9.8, "learning_rate": 2.142570657543641e-08, "loss": 0.0386, "step": 31856 }, { "epoch": 9.8, "learning_rate": 2.1360556573445378e-08, "loss": 0.0212, "step": 31857 }, { "epoch": 9.8, "learning_rate": 2.129550566854044e-08, "loss": 0.0398, "step": 31858 }, { "epoch": 9.8, "learning_rate": 2.123055386136663e-08, "loss": 0.0258, "step": 31859 }, { "epoch": 9.8, "learning_rate": 2.1165701152571217e-08, "loss": 0.0194, "step": 31860 }, { "epoch": 9.8, "learning_rate": 2.110094754279701e-08, "loss": 0.0165, "step": 31861 }, { "epoch": 9.8, "learning_rate": 2.1036293032686837e-08, "loss": 0.0249, "step": 31862 }, { "epoch": 9.8, "learning_rate": 2.0971737622883515e-08, "loss": 0.0295, "step": 31863 }, { "epoch": 9.8, "learning_rate": 2.090728131402764e-08, "loss": 0.0257, "step": 31864 }, { "epoch": 9.8, "learning_rate": 2.084292410675981e-08, "loss": 0.0292, "step": 31865 }, { "epoch": 9.8, "learning_rate": 2.0778666001718405e-08, "loss": 0.0179, "step": 31866 }, { "epoch": 9.8, "learning_rate": 2.0714506999541807e-08, "loss": 0.0133, "step": 31867 }, { "epoch": 9.81, "learning_rate": 2.0650447100867277e-08, "loss": 0.0241, "step": 31868 }, { "epoch": 9.81, "learning_rate": 2.0586486306330977e-08, "loss": 0.0305, "step": 31869 }, { "epoch": 9.81, "learning_rate": 2.0522624616569064e-08, "loss": 0.0191, "step": 31870 }, { "epoch": 9.81, "learning_rate": 2.0458862032214367e-08, "loss": 0.0114, "step": 31871 }, { "epoch": 9.81, "learning_rate": 2.039519855390082e-08, "loss": 0.0101, "step": 31872 }, { "epoch": 9.81, "learning_rate": 2.033163418226014e-08, "loss": 0.0165, "step": 31873 }, { "epoch": 9.81, "learning_rate": 2.0268168917924047e-08, "loss": 0.0131, "step": 31874 }, { "epoch": 9.81, "learning_rate": 2.0204802761522034e-08, "loss": 0.0146, "step": 31875 }, { "epoch": 9.81, "learning_rate": 2.01415357136836e-08, "loss": 0.0147, "step": 31876 }, { "epoch": 9.81, "learning_rate": 2.0078367775038244e-08, "loss": 0.0168, "step": 31877 }, { "epoch": 9.81, "learning_rate": 2.0015298946213236e-08, "loss": 0.0187, "step": 31878 }, { "epoch": 9.81, "learning_rate": 1.9952329227832525e-08, "loss": 0.0204, "step": 31879 }, { "epoch": 9.81, "learning_rate": 1.9889458620524493e-08, "loss": 0.0382, "step": 31880 }, { "epoch": 9.81, "learning_rate": 1.982668712491087e-08, "loss": 0.0251, "step": 31881 }, { "epoch": 9.81, "learning_rate": 1.976401474161782e-08, "loss": 0.0179, "step": 31882 }, { "epoch": 9.81, "learning_rate": 1.970144147126485e-08, "loss": 0.0098, "step": 31883 }, { "epoch": 9.81, "learning_rate": 1.963896731447479e-08, "loss": 0.0254, "step": 31884 }, { "epoch": 9.81, "learning_rate": 1.9576592271868257e-08, "loss": 0.0288, "step": 31885 }, { "epoch": 9.81, "learning_rate": 1.9514316344063645e-08, "loss": 0.0181, "step": 31886 }, { "epoch": 9.81, "learning_rate": 1.9452139531681568e-08, "loss": 0.0304, "step": 31887 }, { "epoch": 9.81, "learning_rate": 1.9390061835337093e-08, "loss": 0.0271, "step": 31888 }, { "epoch": 9.81, "learning_rate": 1.93280832556475e-08, "loss": 0.0405, "step": 31889 }, { "epoch": 9.81, "learning_rate": 1.9266203793228965e-08, "loss": 0.0177, "step": 31890 }, { "epoch": 9.81, "learning_rate": 1.9204423448696553e-08, "loss": 0.0271, "step": 31891 }, { "epoch": 9.81, "learning_rate": 1.9142742222660883e-08, "loss": 0.0329, "step": 31892 }, { "epoch": 9.81, "learning_rate": 1.9081160115738127e-08, "loss": 0.0283, "step": 31893 }, { "epoch": 9.81, "learning_rate": 1.9019677128536694e-08, "loss": 0.0176, "step": 31894 }, { "epoch": 9.81, "learning_rate": 1.8958293261669426e-08, "loss": 0.0217, "step": 31895 }, { "epoch": 9.81, "learning_rate": 1.889700851574472e-08, "loss": 0.0279, "step": 31896 }, { "epoch": 9.81, "learning_rate": 1.883582289137209e-08, "loss": 0.0123, "step": 31897 }, { "epoch": 9.81, "learning_rate": 1.8774736389157723e-08, "loss": 0.019, "step": 31898 }, { "epoch": 9.81, "learning_rate": 1.8713749009710013e-08, "loss": 0.0431, "step": 31899 }, { "epoch": 9.82, "learning_rate": 1.865286075363404e-08, "loss": 0.0214, "step": 31900 }, { "epoch": 9.82, "learning_rate": 1.8592071621532648e-08, "loss": 0.03, "step": 31901 }, { "epoch": 9.82, "learning_rate": 1.8531381614013133e-08, "loss": 0.02, "step": 31902 }, { "epoch": 9.82, "learning_rate": 1.8470790731675014e-08, "loss": 0.0164, "step": 31903 }, { "epoch": 9.82, "learning_rate": 1.8410298975121144e-08, "loss": 0.0174, "step": 31904 }, { "epoch": 9.82, "learning_rate": 1.834990634495326e-08, "loss": 0.027, "step": 31905 }, { "epoch": 9.82, "learning_rate": 1.8289612841768666e-08, "loss": 0.0322, "step": 31906 }, { "epoch": 9.82, "learning_rate": 1.82294184661691e-08, "loss": 0.0116, "step": 31907 }, { "epoch": 9.82, "learning_rate": 1.8169323218749646e-08, "loss": 0.028, "step": 31908 }, { "epoch": 9.82, "learning_rate": 1.810932710010982e-08, "loss": 0.0228, "step": 31909 }, { "epoch": 9.82, "learning_rate": 1.804943011084248e-08, "loss": 0.0277, "step": 31910 }, { "epoch": 9.82, "learning_rate": 1.7989632251544932e-08, "loss": 0.032, "step": 31911 }, { "epoch": 9.82, "learning_rate": 1.7929933522810028e-08, "loss": 0.0191, "step": 31912 }, { "epoch": 9.82, "learning_rate": 1.787033392522952e-08, "loss": 0.0183, "step": 31913 }, { "epoch": 9.82, "learning_rate": 1.781083345939738e-08, "loss": 0.0128, "step": 31914 }, { "epoch": 9.82, "learning_rate": 1.7751432125903133e-08, "loss": 0.0272, "step": 31915 }, { "epoch": 9.82, "learning_rate": 1.769212992533742e-08, "loss": 0.026, "step": 31916 }, { "epoch": 9.82, "learning_rate": 1.7632926858288658e-08, "loss": 0.0397, "step": 31917 }, { "epoch": 9.82, "learning_rate": 1.7573822925345264e-08, "loss": 0.0445, "step": 31918 }, { "epoch": 9.82, "learning_rate": 1.7514818127093436e-08, "loss": 0.0247, "step": 31919 }, { "epoch": 9.82, "learning_rate": 1.7455912464119374e-08, "loss": 0.0209, "step": 31920 }, { "epoch": 9.82, "learning_rate": 1.7397105937008162e-08, "loss": 0.023, "step": 31921 }, { "epoch": 9.82, "learning_rate": 1.733839854634489e-08, "loss": 0.0375, "step": 31922 }, { "epoch": 9.82, "learning_rate": 1.727979029271132e-08, "loss": 0.0314, "step": 31923 }, { "epoch": 9.82, "learning_rate": 1.7221281176689196e-08, "loss": 0.0303, "step": 31924 }, { "epoch": 9.82, "learning_rate": 1.7162871198860286e-08, "loss": 0.034, "step": 31925 }, { "epoch": 9.82, "learning_rate": 1.7104560359805235e-08, "loss": 0.0157, "step": 31926 }, { "epoch": 9.82, "learning_rate": 1.7046348660101353e-08, "loss": 0.0259, "step": 31927 }, { "epoch": 9.82, "learning_rate": 1.698823610032929e-08, "loss": 0.0199, "step": 31928 }, { "epoch": 9.82, "learning_rate": 1.6930222681063037e-08, "loss": 0.0204, "step": 31929 }, { "epoch": 9.82, "learning_rate": 1.687230840288101e-08, "loss": 0.0284, "step": 31930 }, { "epoch": 9.82, "learning_rate": 1.6814493266357202e-08, "loss": 0.0278, "step": 31931 }, { "epoch": 9.82, "learning_rate": 1.6756777272067814e-08, "loss": 0.0261, "step": 31932 }, { "epoch": 9.83, "learning_rate": 1.6699160420583504e-08, "loss": 0.021, "step": 31933 }, { "epoch": 9.83, "learning_rate": 1.6641642712477147e-08, "loss": 0.0096, "step": 31934 }, { "epoch": 9.83, "learning_rate": 1.6584224148320505e-08, "loss": 0.0321, "step": 31935 }, { "epoch": 9.83, "learning_rate": 1.6526904728683124e-08, "loss": 0.0293, "step": 31936 }, { "epoch": 9.83, "learning_rate": 1.646968445413455e-08, "loss": 0.0135, "step": 31937 }, { "epoch": 9.83, "learning_rate": 1.6412563325243214e-08, "loss": 0.0264, "step": 31938 }, { "epoch": 9.83, "learning_rate": 1.635554134257644e-08, "loss": 0.0192, "step": 31939 }, { "epoch": 9.83, "learning_rate": 1.629861850670045e-08, "loss": 0.0103, "step": 31940 }, { "epoch": 9.83, "learning_rate": 1.6241794818180333e-08, "loss": 0.0241, "step": 31941 }, { "epoch": 9.83, "learning_rate": 1.6185070277580097e-08, "loss": 0.027, "step": 31942 }, { "epoch": 9.83, "learning_rate": 1.6128444885464834e-08, "loss": 0.0335, "step": 31943 }, { "epoch": 9.83, "learning_rate": 1.6071918642394102e-08, "loss": 0.0219, "step": 31944 }, { "epoch": 9.83, "learning_rate": 1.6015491548930785e-08, "loss": 0.024, "step": 31945 }, { "epoch": 9.83, "learning_rate": 1.5959163605635542e-08, "loss": 0.0275, "step": 31946 }, { "epoch": 9.83, "learning_rate": 1.5902934813067928e-08, "loss": 0.0214, "step": 31947 }, { "epoch": 9.83, "learning_rate": 1.5846805171785273e-08, "loss": 0.035, "step": 31948 }, { "epoch": 9.83, "learning_rate": 1.579077468234491e-08, "loss": 0.0147, "step": 31949 }, { "epoch": 9.83, "learning_rate": 1.5734843345304172e-08, "loss": 0.0154, "step": 31950 }, { "epoch": 9.83, "learning_rate": 1.567901116121817e-08, "loss": 0.0228, "step": 31951 }, { "epoch": 9.83, "learning_rate": 1.5623278130642017e-08, "loss": 0.0283, "step": 31952 }, { "epoch": 9.83, "learning_rate": 1.55676442541286e-08, "loss": 0.0273, "step": 31953 }, { "epoch": 9.83, "learning_rate": 1.5512109532229703e-08, "loss": 0.0107, "step": 31954 }, { "epoch": 9.83, "learning_rate": 1.5456673965498213e-08, "loss": 0.0138, "step": 31955 }, { "epoch": 9.83, "learning_rate": 1.540133755448259e-08, "loss": 0.0414, "step": 31956 }, { "epoch": 9.83, "learning_rate": 1.534610029973571e-08, "loss": 0.0401, "step": 31957 }, { "epoch": 9.83, "learning_rate": 1.5290962201802707e-08, "loss": 0.0137, "step": 31958 }, { "epoch": 9.83, "learning_rate": 1.5235923261233134e-08, "loss": 0.0192, "step": 31959 }, { "epoch": 9.83, "learning_rate": 1.5180983478573218e-08, "loss": 0.0235, "step": 31960 }, { "epoch": 9.83, "learning_rate": 1.5126142854368085e-08, "loss": 0.0229, "step": 31961 }, { "epoch": 9.83, "learning_rate": 1.5071401389162854e-08, "loss": 0.0188, "step": 31962 }, { "epoch": 9.83, "learning_rate": 1.501675908350153e-08, "loss": 0.0262, "step": 31963 }, { "epoch": 9.83, "learning_rate": 1.4962215937925906e-08, "loss": 0.0222, "step": 31964 }, { "epoch": 9.84, "learning_rate": 1.4907771952977767e-08, "loss": 0.0272, "step": 31965 }, { "epoch": 9.84, "learning_rate": 1.4853427129197794e-08, "loss": 0.0279, "step": 31966 }, { "epoch": 9.84, "learning_rate": 1.4799181467125557e-08, "loss": 0.0312, "step": 31967 }, { "epoch": 9.84, "learning_rate": 1.474503496730173e-08, "loss": 0.0406, "step": 31968 }, { "epoch": 9.84, "learning_rate": 1.4690987630260333e-08, "loss": 0.0225, "step": 31969 }, { "epoch": 9.84, "learning_rate": 1.4637039456542047e-08, "loss": 0.0315, "step": 31970 }, { "epoch": 9.84, "learning_rate": 1.4583190446679774e-08, "loss": 0.0207, "step": 31971 }, { "epoch": 9.84, "learning_rate": 1.4529440601208644e-08, "loss": 0.0184, "step": 31972 }, { "epoch": 9.84, "learning_rate": 1.4475789920663785e-08, "loss": 0.0121, "step": 31973 }, { "epoch": 9.84, "learning_rate": 1.4422238405576994e-08, "loss": 0.0325, "step": 31974 }, { "epoch": 9.84, "learning_rate": 1.4368786056481177e-08, "loss": 0.0196, "step": 31975 }, { "epoch": 9.84, "learning_rate": 1.43154328739048e-08, "loss": 0.0297, "step": 31976 }, { "epoch": 9.84, "learning_rate": 1.4262178858380771e-08, "loss": 0.0165, "step": 31977 }, { "epoch": 9.84, "learning_rate": 1.4209024010435334e-08, "loss": 0.0342, "step": 31978 }, { "epoch": 9.84, "learning_rate": 1.4155968330598069e-08, "loss": 0.0213, "step": 31979 }, { "epoch": 9.84, "learning_rate": 1.4103011819395218e-08, "loss": 0.0255, "step": 31980 }, { "epoch": 9.84, "learning_rate": 1.4050154477353029e-08, "loss": 0.0288, "step": 31981 }, { "epoch": 9.84, "learning_rate": 1.3997396304995526e-08, "loss": 0.0217, "step": 31982 }, { "epoch": 9.84, "learning_rate": 1.3944737302846733e-08, "loss": 0.0228, "step": 31983 }, { "epoch": 9.84, "learning_rate": 1.3892177471430679e-08, "loss": 0.0163, "step": 31984 }, { "epoch": 9.84, "learning_rate": 1.3839716811269165e-08, "loss": 0.0156, "step": 31985 }, { "epoch": 9.84, "learning_rate": 1.378735532288289e-08, "loss": 0.0248, "step": 31986 }, { "epoch": 9.84, "learning_rate": 1.3735093006791434e-08, "loss": 0.0119, "step": 31987 }, { "epoch": 9.84, "learning_rate": 1.3682929863514383e-08, "loss": 0.0281, "step": 31988 }, { "epoch": 9.84, "learning_rate": 1.3630865893569101e-08, "loss": 0.0243, "step": 31989 }, { "epoch": 9.84, "learning_rate": 1.3578901097472952e-08, "loss": 0.018, "step": 31990 }, { "epoch": 9.84, "learning_rate": 1.352703547574219e-08, "loss": 0.0147, "step": 31991 }, { "epoch": 9.84, "learning_rate": 1.3475269028891958e-08, "loss": 0.0291, "step": 31992 }, { "epoch": 9.84, "learning_rate": 1.3423601757436289e-08, "loss": 0.0296, "step": 31993 }, { "epoch": 9.84, "learning_rate": 1.3372033661886996e-08, "loss": 0.0237, "step": 31994 }, { "epoch": 9.84, "learning_rate": 1.3320564742759223e-08, "loss": 0.0152, "step": 31995 }, { "epoch": 9.84, "learning_rate": 1.3269195000560343e-08, "loss": 0.02, "step": 31996 }, { "epoch": 9.84, "learning_rate": 1.3217924435803276e-08, "loss": 0.0205, "step": 31997 }, { "epoch": 9.85, "learning_rate": 1.3166753048996505e-08, "loss": 0.0243, "step": 31998 }, { "epoch": 9.85, "learning_rate": 1.3115680840648515e-08, "loss": 0.0205, "step": 31999 }, { "epoch": 9.85, "learning_rate": 1.3064707811265564e-08, "loss": 0.0292, "step": 32000 }, { "epoch": 9.85, "learning_rate": 1.3013833961352807e-08, "loss": 0.027, "step": 32001 }, { "epoch": 9.85, "learning_rate": 1.2963059291418723e-08, "loss": 0.0132, "step": 32002 }, { "epoch": 9.85, "learning_rate": 1.2912383801965133e-08, "loss": 0.0278, "step": 32003 }, { "epoch": 9.85, "learning_rate": 1.2861807493496082e-08, "loss": 0.0152, "step": 32004 }, { "epoch": 9.85, "learning_rate": 1.2811330366514495e-08, "loss": 0.0204, "step": 32005 }, { "epoch": 9.85, "learning_rate": 1.276095242151998e-08, "loss": 0.0185, "step": 32006 }, { "epoch": 9.85, "learning_rate": 1.271067365901324e-08, "loss": 0.0297, "step": 32007 }, { "epoch": 9.85, "learning_rate": 1.266049407949499e-08, "loss": 0.0252, "step": 32008 }, { "epoch": 9.85, "learning_rate": 1.2610413683462608e-08, "loss": 0.0322, "step": 32009 }, { "epoch": 9.85, "learning_rate": 1.2560432471412365e-08, "loss": 0.0393, "step": 32010 }, { "epoch": 9.85, "learning_rate": 1.251055044384275e-08, "loss": 0.023, "step": 32011 }, { "epoch": 9.85, "learning_rate": 1.2460767601247814e-08, "loss": 0.0292, "step": 32012 }, { "epoch": 9.85, "learning_rate": 1.2411083944122715e-08, "loss": 0.0249, "step": 32013 }, { "epoch": 9.85, "learning_rate": 1.2361499472959282e-08, "loss": 0.048, "step": 32014 }, { "epoch": 9.85, "learning_rate": 1.2312014188251565e-08, "loss": 0.0362, "step": 32015 }, { "epoch": 9.85, "learning_rate": 1.2262628090491391e-08, "loss": 0.0132, "step": 32016 }, { "epoch": 9.85, "learning_rate": 1.2213341180167259e-08, "loss": 0.0276, "step": 32017 }, { "epoch": 9.85, "learning_rate": 1.216415345776989e-08, "loss": 0.0295, "step": 32018 }, { "epoch": 9.85, "learning_rate": 1.2115064923787778e-08, "loss": 0.0394, "step": 32019 }, { "epoch": 9.85, "learning_rate": 1.2066075578708314e-08, "loss": 0.01, "step": 32020 }, { "epoch": 9.85, "learning_rate": 1.2017185423016663e-08, "loss": 0.0286, "step": 32021 }, { "epoch": 9.85, "learning_rate": 1.1968394457201326e-08, "loss": 0.0106, "step": 32022 }, { "epoch": 9.85, "learning_rate": 1.1919702681744138e-08, "loss": 0.0278, "step": 32023 }, { "epoch": 9.85, "learning_rate": 1.1871110097129157e-08, "loss": 0.0256, "step": 32024 }, { "epoch": 9.85, "learning_rate": 1.1822616703840439e-08, "loss": 0.0319, "step": 32025 }, { "epoch": 9.85, "learning_rate": 1.1774222502357602e-08, "loss": 0.022, "step": 32026 }, { "epoch": 9.85, "learning_rate": 1.172592749316248e-08, "loss": 0.0414, "step": 32027 }, { "epoch": 9.85, "learning_rate": 1.1677731676733584e-08, "loss": 0.0237, "step": 32028 }, { "epoch": 9.85, "learning_rate": 1.1629635053551635e-08, "loss": 0.009, "step": 32029 }, { "epoch": 9.86, "learning_rate": 1.1581637624092923e-08, "loss": 0.0142, "step": 32030 }, { "epoch": 9.86, "learning_rate": 1.153373938883262e-08, "loss": 0.0228, "step": 32031 }, { "epoch": 9.86, "learning_rate": 1.1485940348249235e-08, "loss": 0.0091, "step": 32032 }, { "epoch": 9.86, "learning_rate": 1.1438240502815723e-08, "loss": 0.0279, "step": 32033 }, { "epoch": 9.86, "learning_rate": 1.1390639853007258e-08, "loss": 0.0187, "step": 32034 }, { "epoch": 9.86, "learning_rate": 1.1343138399294573e-08, "loss": 0.0327, "step": 32035 }, { "epoch": 9.86, "learning_rate": 1.1295736142149515e-08, "loss": 0.0105, "step": 32036 }, { "epoch": 9.86, "learning_rate": 1.124843308204504e-08, "loss": 0.0164, "step": 32037 }, { "epoch": 9.86, "learning_rate": 1.120122921944855e-08, "loss": 0.0201, "step": 32038 }, { "epoch": 9.86, "learning_rate": 1.115412455483078e-08, "loss": 0.0313, "step": 32039 }, { "epoch": 9.86, "learning_rate": 1.1107119088658025e-08, "loss": 0.0239, "step": 32040 }, { "epoch": 9.86, "learning_rate": 1.1060212821397686e-08, "loss": 0.0147, "step": 32041 }, { "epoch": 9.86, "learning_rate": 1.101340575351495e-08, "loss": 0.0231, "step": 32042 }, { "epoch": 9.86, "learning_rate": 1.0966697885476108e-08, "loss": 0.0106, "step": 32043 }, { "epoch": 9.86, "learning_rate": 1.0920089217744122e-08, "loss": 0.0204, "step": 32044 }, { "epoch": 9.86, "learning_rate": 1.087357975078085e-08, "loss": 0.0118, "step": 32045 }, { "epoch": 9.86, "learning_rate": 1.082716948505036e-08, "loss": 0.0169, "step": 32046 }, { "epoch": 9.86, "learning_rate": 1.0780858421011175e-08, "loss": 0.0095, "step": 32047 }, { "epoch": 9.86, "learning_rate": 1.0734646559125151e-08, "loss": 0.0149, "step": 32048 }, { "epoch": 9.86, "learning_rate": 1.0688533899850806e-08, "loss": 0.0331, "step": 32049 }, { "epoch": 9.86, "learning_rate": 1.0642520443645554e-08, "loss": 0.0416, "step": 32050 }, { "epoch": 9.86, "learning_rate": 1.0596606190965696e-08, "loss": 0.0295, "step": 32051 }, { "epoch": 9.86, "learning_rate": 1.0550791142268646e-08, "loss": 0.0384, "step": 32052 }, { "epoch": 9.86, "learning_rate": 1.0505075298008483e-08, "loss": 0.0117, "step": 32053 }, { "epoch": 9.86, "learning_rate": 1.0459458658640397e-08, "loss": 0.0203, "step": 32054 }, { "epoch": 9.86, "learning_rate": 1.0413941224615142e-08, "loss": 0.0173, "step": 32055 }, { "epoch": 9.86, "learning_rate": 1.03685229963868e-08, "loss": 0.0234, "step": 32056 }, { "epoch": 9.86, "learning_rate": 1.0323203974406115e-08, "loss": 0.0289, "step": 32057 }, { "epoch": 9.86, "learning_rate": 1.0277984159122734e-08, "loss": 0.031, "step": 32058 }, { "epoch": 9.86, "learning_rate": 1.0232863550985184e-08, "loss": 0.0362, "step": 32059 }, { "epoch": 9.86, "learning_rate": 1.0187842150443106e-08, "loss": 0.0268, "step": 32060 }, { "epoch": 9.86, "learning_rate": 1.0142919957941699e-08, "loss": 0.0219, "step": 32061 }, { "epoch": 9.86, "learning_rate": 1.0098096973928384e-08, "loss": 0.0259, "step": 32062 }, { "epoch": 9.87, "learning_rate": 1.005337319884725e-08, "loss": 0.0414, "step": 32063 }, { "epoch": 9.87, "learning_rate": 1.0008748633144604e-08, "loss": 0.025, "step": 32064 }, { "epoch": 9.87, "learning_rate": 9.96422327726121e-09, "loss": 0.0389, "step": 32065 }, { "epoch": 9.87, "learning_rate": 9.919797131640041e-09, "loss": 0.0235, "step": 32066 }, { "epoch": 9.87, "learning_rate": 9.875470196721859e-09, "loss": 0.0165, "step": 32067 }, { "epoch": 9.87, "learning_rate": 9.831242472948532e-09, "loss": 0.0244, "step": 32068 }, { "epoch": 9.87, "learning_rate": 9.787113960757488e-09, "loss": 0.0329, "step": 32069 }, { "epoch": 9.87, "learning_rate": 9.743084660588376e-09, "loss": 0.0223, "step": 32070 }, { "epoch": 9.87, "learning_rate": 9.699154572877511e-09, "loss": 0.0288, "step": 32071 }, { "epoch": 9.87, "learning_rate": 9.655323698061214e-09, "loss": 0.0302, "step": 32072 }, { "epoch": 9.87, "learning_rate": 9.61159203657469e-09, "loss": 0.0263, "step": 32073 }, { "epoch": 9.87, "learning_rate": 9.567959588853148e-09, "loss": 0.0072, "step": 32074 }, { "epoch": 9.87, "learning_rate": 9.524426355328465e-09, "loss": 0.024, "step": 32075 }, { "epoch": 9.87, "learning_rate": 9.480992336433625e-09, "loss": 0.02, "step": 32076 }, { "epoch": 9.87, "learning_rate": 9.437657532600508e-09, "loss": 0.0318, "step": 32077 }, { "epoch": 9.87, "learning_rate": 9.394421944258769e-09, "loss": 0.017, "step": 32078 }, { "epoch": 9.87, "learning_rate": 9.351285571839174e-09, "loss": 0.0243, "step": 32079 }, { "epoch": 9.87, "learning_rate": 9.30824841576805e-09, "loss": 0.0279, "step": 32080 }, { "epoch": 9.87, "learning_rate": 9.265310476472833e-09, "loss": 0.019, "step": 32081 }, { "epoch": 9.87, "learning_rate": 9.222471754382068e-09, "loss": 0.0247, "step": 32082 }, { "epoch": 9.87, "learning_rate": 9.17973224991986e-09, "loss": 0.0134, "step": 32083 }, { "epoch": 9.87, "learning_rate": 9.137091963510314e-09, "loss": 0.0285, "step": 32084 }, { "epoch": 9.87, "learning_rate": 9.094550895577536e-09, "loss": 0.0211, "step": 32085 }, { "epoch": 9.87, "learning_rate": 9.052109046543412e-09, "loss": 0.0272, "step": 32086 }, { "epoch": 9.87, "learning_rate": 9.009766416829824e-09, "loss": 0.0169, "step": 32087 }, { "epoch": 9.87, "learning_rate": 8.967523006857549e-09, "loss": 0.0224, "step": 32088 }, { "epoch": 9.87, "learning_rate": 8.925378817045139e-09, "loss": 0.0334, "step": 32089 }, { "epoch": 9.87, "learning_rate": 8.883333847812258e-09, "loss": 0.0196, "step": 32090 }, { "epoch": 9.87, "learning_rate": 8.841388099576354e-09, "loss": 0.0195, "step": 32091 }, { "epoch": 9.87, "learning_rate": 8.799541572752646e-09, "loss": 0.0319, "step": 32092 }, { "epoch": 9.87, "learning_rate": 8.75779426775858e-09, "loss": 0.0231, "step": 32093 }, { "epoch": 9.87, "learning_rate": 8.716146185007157e-09, "loss": 0.0285, "step": 32094 }, { "epoch": 9.88, "learning_rate": 8.674597324912492e-09, "loss": 0.0085, "step": 32095 }, { "epoch": 9.88, "learning_rate": 8.633147687887589e-09, "loss": 0.0281, "step": 32096 }, { "epoch": 9.88, "learning_rate": 8.591797274344338e-09, "loss": 0.0229, "step": 32097 }, { "epoch": 9.88, "learning_rate": 8.550546084692413e-09, "loss": 0.0214, "step": 32098 }, { "epoch": 9.88, "learning_rate": 8.509394119342595e-09, "loss": 0.0268, "step": 32099 }, { "epoch": 9.88, "learning_rate": 8.468341378702339e-09, "loss": 0.0454, "step": 32100 }, { "epoch": 9.88, "learning_rate": 8.427387863180204e-09, "loss": 0.0196, "step": 32101 }, { "epoch": 9.88, "learning_rate": 8.386533573182531e-09, "loss": 0.0269, "step": 32102 }, { "epoch": 9.88, "learning_rate": 8.345778509114556e-09, "loss": 0.0408, "step": 32103 }, { "epoch": 9.88, "learning_rate": 8.305122671381505e-09, "loss": 0.032, "step": 32104 }, { "epoch": 9.88, "learning_rate": 8.264566060388612e-09, "loss": 0.0255, "step": 32105 }, { "epoch": 9.88, "learning_rate": 8.224108676535557e-09, "loss": 0.0146, "step": 32106 }, { "epoch": 9.88, "learning_rate": 8.183750520226464e-09, "loss": 0.0302, "step": 32107 }, { "epoch": 9.88, "learning_rate": 8.143491591862118e-09, "loss": 0.0164, "step": 32108 }, { "epoch": 9.88, "learning_rate": 8.103331891841094e-09, "loss": 0.0129, "step": 32109 }, { "epoch": 9.88, "learning_rate": 8.063271420563068e-09, "loss": 0.0223, "step": 32110 }, { "epoch": 9.88, "learning_rate": 8.023310178424393e-09, "loss": 0.0312, "step": 32111 }, { "epoch": 9.88, "learning_rate": 7.983448165824747e-09, "loss": 0.0217, "step": 32112 }, { "epoch": 9.88, "learning_rate": 7.94368538315715e-09, "loss": 0.0328, "step": 32113 }, { "epoch": 9.88, "learning_rate": 7.904021830819063e-09, "loss": 0.0121, "step": 32114 }, { "epoch": 9.88, "learning_rate": 7.864457509202395e-09, "loss": 0.0254, "step": 32115 }, { "epoch": 9.88, "learning_rate": 7.824992418700161e-09, "loss": 0.0182, "step": 32116 }, { "epoch": 9.88, "learning_rate": 7.785626559705383e-09, "loss": 0.0228, "step": 32117 }, { "epoch": 9.88, "learning_rate": 7.74635993260775e-09, "loss": 0.0268, "step": 32118 }, { "epoch": 9.88, "learning_rate": 7.707192537798059e-09, "loss": 0.0215, "step": 32119 }, { "epoch": 9.88, "learning_rate": 7.668124375665997e-09, "loss": 0.0224, "step": 32120 }, { "epoch": 9.88, "learning_rate": 7.629155446597924e-09, "loss": 0.0225, "step": 32121 }, { "epoch": 9.88, "learning_rate": 7.590285750981307e-09, "loss": 0.021, "step": 32122 }, { "epoch": 9.88, "learning_rate": 7.551515289203615e-09, "loss": 0.0157, "step": 32123 }, { "epoch": 9.88, "learning_rate": 7.512844061647873e-09, "loss": 0.0199, "step": 32124 }, { "epoch": 9.88, "learning_rate": 7.474272068698219e-09, "loss": 0.0376, "step": 32125 }, { "epoch": 9.88, "learning_rate": 7.4357993107399e-09, "loss": 0.0309, "step": 32126 }, { "epoch": 9.88, "learning_rate": 7.397425788152612e-09, "loss": 0.0276, "step": 32127 }, { "epoch": 9.89, "learning_rate": 7.359151501318274e-09, "loss": 0.023, "step": 32128 }, { "epoch": 9.89, "learning_rate": 7.3209764506176895e-09, "loss": 0.0167, "step": 32129 }, { "epoch": 9.89, "learning_rate": 7.282900636428336e-09, "loss": 0.0085, "step": 32130 }, { "epoch": 9.89, "learning_rate": 7.2449240591299095e-09, "loss": 0.0209, "step": 32131 }, { "epoch": 9.89, "learning_rate": 7.207046719099886e-09, "loss": 0.0293, "step": 32132 }, { "epoch": 9.89, "learning_rate": 7.16926861671241e-09, "loss": 0.027, "step": 32133 }, { "epoch": 9.89, "learning_rate": 7.131589752343848e-09, "loss": 0.0184, "step": 32134 }, { "epoch": 9.89, "learning_rate": 7.0940101263683445e-09, "loss": 0.024, "step": 32135 }, { "epoch": 9.89, "learning_rate": 7.056529739158935e-09, "loss": 0.0242, "step": 32136 }, { "epoch": 9.89, "learning_rate": 7.0191485910886535e-09, "loss": 0.0219, "step": 32137 }, { "epoch": 9.89, "learning_rate": 6.981866682527206e-09, "loss": 0.0157, "step": 32138 }, { "epoch": 9.89, "learning_rate": 6.944684013846514e-09, "loss": 0.0362, "step": 32139 }, { "epoch": 9.89, "learning_rate": 6.907600585414065e-09, "loss": 0.0322, "step": 32140 }, { "epoch": 9.89, "learning_rate": 6.870616397599561e-09, "loss": 0.0337, "step": 32141 }, { "epoch": 9.89, "learning_rate": 6.833731450769376e-09, "loss": 0.0189, "step": 32142 }, { "epoch": 9.89, "learning_rate": 6.7969457452909944e-09, "loss": 0.0317, "step": 32143 }, { "epoch": 9.89, "learning_rate": 6.7602592815285695e-09, "loss": 0.0311, "step": 32144 }, { "epoch": 9.89, "learning_rate": 6.7236720598462535e-09, "loss": 0.0347, "step": 32145 }, { "epoch": 9.89, "learning_rate": 6.6871840806082e-09, "loss": 0.0127, "step": 32146 }, { "epoch": 9.89, "learning_rate": 6.650795344177452e-09, "loss": 0.0171, "step": 32147 }, { "epoch": 9.89, "learning_rate": 6.6145058509137216e-09, "loss": 0.0291, "step": 32148 }, { "epoch": 9.89, "learning_rate": 6.5783156011778315e-09, "loss": 0.0173, "step": 32149 }, { "epoch": 9.89, "learning_rate": 6.542224595329494e-09, "loss": 0.016, "step": 32150 }, { "epoch": 9.89, "learning_rate": 6.50623283372731e-09, "loss": 0.0146, "step": 32151 }, { "epoch": 9.89, "learning_rate": 6.470340316728774e-09, "loss": 0.0282, "step": 32152 }, { "epoch": 9.89, "learning_rate": 6.4345470446902645e-09, "loss": 0.0324, "step": 32153 }, { "epoch": 9.89, "learning_rate": 6.3988530179659446e-09, "loss": 0.0224, "step": 32154 }, { "epoch": 9.89, "learning_rate": 6.363258236912195e-09, "loss": 0.0323, "step": 32155 }, { "epoch": 9.89, "learning_rate": 6.327762701882068e-09, "loss": 0.0178, "step": 32156 }, { "epoch": 9.89, "learning_rate": 6.292366413227502e-09, "loss": 0.0225, "step": 32157 }, { "epoch": 9.89, "learning_rate": 6.257069371299329e-09, "loss": 0.0078, "step": 32158 }, { "epoch": 9.89, "learning_rate": 6.2218715764494895e-09, "loss": 0.0211, "step": 32159 }, { "epoch": 9.9, "learning_rate": 6.186773029027704e-09, "loss": 0.0169, "step": 32160 }, { "epoch": 9.9, "learning_rate": 6.151773729381472e-09, "loss": 0.0254, "step": 32161 }, { "epoch": 9.9, "learning_rate": 6.116873677858293e-09, "loss": 0.0201, "step": 32162 }, { "epoch": 9.9, "learning_rate": 6.082072874804556e-09, "loss": 0.0259, "step": 32163 }, { "epoch": 9.9, "learning_rate": 6.047371320567763e-09, "loss": 0.0272, "step": 32164 }, { "epoch": 9.9, "learning_rate": 6.012769015490971e-09, "loss": 0.0239, "step": 32165 }, { "epoch": 9.9, "learning_rate": 5.97826595991724e-09, "loss": 0.0197, "step": 32166 }, { "epoch": 9.9, "learning_rate": 5.943862154191849e-09, "loss": 0.0273, "step": 32167 }, { "epoch": 9.9, "learning_rate": 5.9095575986534146e-09, "loss": 0.0411, "step": 32168 }, { "epoch": 9.9, "learning_rate": 5.875352293643888e-09, "loss": 0.0136, "step": 32169 }, { "epoch": 9.9, "learning_rate": 5.841246239502996e-09, "loss": 0.0296, "step": 32170 }, { "epoch": 9.9, "learning_rate": 5.807239436570467e-09, "loss": 0.015, "step": 32171 }, { "epoch": 9.9, "learning_rate": 5.773331885181588e-09, "loss": 0.0293, "step": 32172 }, { "epoch": 9.9, "learning_rate": 5.739523585674978e-09, "loss": 0.022, "step": 32173 }, { "epoch": 9.9, "learning_rate": 5.705814538385923e-09, "loss": 0.0357, "step": 32174 }, { "epoch": 9.9, "learning_rate": 5.6722047436497115e-09, "loss": 0.029, "step": 32175 }, { "epoch": 9.9, "learning_rate": 5.638694201798301e-09, "loss": 0.0248, "step": 32176 }, { "epoch": 9.9, "learning_rate": 5.605282913166976e-09, "loss": 0.0251, "step": 32177 }, { "epoch": 9.9, "learning_rate": 5.571970878085475e-09, "loss": 0.0285, "step": 32178 }, { "epoch": 9.9, "learning_rate": 5.538758096885755e-09, "loss": 0.0246, "step": 32179 }, { "epoch": 9.9, "learning_rate": 5.50564456989755e-09, "loss": 0.0146, "step": 32180 }, { "epoch": 9.9, "learning_rate": 5.472630297448378e-09, "loss": 0.0216, "step": 32181 }, { "epoch": 9.9, "learning_rate": 5.4397152798690846e-09, "loss": 0.0156, "step": 32182 }, { "epoch": 9.9, "learning_rate": 5.4068995174827446e-09, "loss": 0.0157, "step": 32183 }, { "epoch": 9.9, "learning_rate": 5.3741830106179834e-09, "loss": 0.0318, "step": 32184 }, { "epoch": 9.9, "learning_rate": 5.341565759600098e-09, "loss": 0.021, "step": 32185 }, { "epoch": 9.9, "learning_rate": 5.3090477647499415e-09, "loss": 0.0269, "step": 32186 }, { "epoch": 9.9, "learning_rate": 5.27662902639281e-09, "loss": 0.0171, "step": 32187 }, { "epoch": 9.9, "learning_rate": 5.2443095448506674e-09, "loss": 0.0257, "step": 32188 }, { "epoch": 9.9, "learning_rate": 5.212089320444369e-09, "loss": 0.0098, "step": 32189 }, { "epoch": 9.9, "learning_rate": 5.179968353492548e-09, "loss": 0.0249, "step": 32190 }, { "epoch": 9.9, "learning_rate": 5.147946644316059e-09, "loss": 0.0137, "step": 32191 }, { "epoch": 9.9, "learning_rate": 5.116024193231317e-09, "loss": 0.0304, "step": 32192 }, { "epoch": 9.91, "learning_rate": 5.084201000555844e-09, "loss": 0.0203, "step": 32193 }, { "epoch": 9.91, "learning_rate": 5.052477066606054e-09, "loss": 0.0204, "step": 32194 }, { "epoch": 9.91, "learning_rate": 5.020852391697251e-09, "loss": 0.0195, "step": 32195 }, { "epoch": 9.91, "learning_rate": 4.989326976142516e-09, "loss": 0.0179, "step": 32196 }, { "epoch": 9.91, "learning_rate": 4.957900820254935e-09, "loss": 0.0211, "step": 32197 }, { "epoch": 9.91, "learning_rate": 4.926573924347589e-09, "loss": 0.0247, "step": 32198 }, { "epoch": 9.91, "learning_rate": 4.89534628873134e-09, "loss": 0.0155, "step": 32199 }, { "epoch": 9.91, "learning_rate": 4.8642179137159405e-09, "loss": 0.0067, "step": 32200 }, { "epoch": 9.91, "learning_rate": 4.833188799610033e-09, "loss": 0.0139, "step": 32201 }, { "epoch": 9.91, "learning_rate": 4.80225894672337e-09, "loss": 0.0306, "step": 32202 }, { "epoch": 9.91, "learning_rate": 4.771428355362373e-09, "loss": 0.0129, "step": 32203 }, { "epoch": 9.91, "learning_rate": 4.740697025832353e-09, "loss": 0.021, "step": 32204 }, { "epoch": 9.91, "learning_rate": 4.710064958438621e-09, "loss": 0.0103, "step": 32205 }, { "epoch": 9.91, "learning_rate": 4.679532153486488e-09, "loss": 0.0114, "step": 32206 }, { "epoch": 9.91, "learning_rate": 4.649098611277936e-09, "loss": 0.0313, "step": 32207 }, { "epoch": 9.91, "learning_rate": 4.618764332116055e-09, "loss": 0.0098, "step": 32208 }, { "epoch": 9.91, "learning_rate": 4.588529316302825e-09, "loss": 0.0145, "step": 32209 }, { "epoch": 9.91, "learning_rate": 4.558393564136898e-09, "loss": 0.0107, "step": 32210 }, { "epoch": 9.91, "learning_rate": 4.5283570759180325e-09, "loss": 0.0277, "step": 32211 }, { "epoch": 9.91, "learning_rate": 4.498419851944879e-09, "loss": 0.0284, "step": 32212 }, { "epoch": 9.91, "learning_rate": 4.468581892513868e-09, "loss": 0.0276, "step": 32213 }, { "epoch": 9.91, "learning_rate": 4.438843197922538e-09, "loss": 0.0412, "step": 32214 }, { "epoch": 9.91, "learning_rate": 4.409203768466208e-09, "loss": 0.0272, "step": 32215 }, { "epoch": 9.91, "learning_rate": 4.379663604437978e-09, "loss": 0.0228, "step": 32216 }, { "epoch": 9.91, "learning_rate": 4.350222706132057e-09, "loss": 0.0225, "step": 32217 }, { "epoch": 9.91, "learning_rate": 4.3208810738404326e-09, "loss": 0.0138, "step": 32218 }, { "epoch": 9.91, "learning_rate": 4.291638707855094e-09, "loss": 0.0231, "step": 32219 }, { "epoch": 9.91, "learning_rate": 4.262495608465811e-09, "loss": 0.0403, "step": 32220 }, { "epoch": 9.91, "learning_rate": 4.23345177596235e-09, "loss": 0.0238, "step": 32221 }, { "epoch": 9.91, "learning_rate": 4.204507210633368e-09, "loss": 0.0205, "step": 32222 }, { "epoch": 9.91, "learning_rate": 4.1756619127664155e-09, "loss": 0.0351, "step": 32223 }, { "epoch": 9.91, "learning_rate": 4.146915882646818e-09, "loss": 0.0222, "step": 32224 }, { "epoch": 9.92, "learning_rate": 4.118269120561014e-09, "loss": 0.016, "step": 32225 }, { "epoch": 9.92, "learning_rate": 4.089721626793219e-09, "loss": 0.0099, "step": 32226 }, { "epoch": 9.92, "learning_rate": 4.061273401627653e-09, "loss": 0.0205, "step": 32227 }, { "epoch": 9.92, "learning_rate": 4.032924445345199e-09, "loss": 0.0163, "step": 32228 }, { "epoch": 9.92, "learning_rate": 4.0046747582289655e-09, "loss": 0.0132, "step": 32229 }, { "epoch": 9.92, "learning_rate": 3.97652434055873e-09, "loss": 0.0236, "step": 32230 }, { "epoch": 9.92, "learning_rate": 3.948473192614266e-09, "loss": 0.0152, "step": 32231 }, { "epoch": 9.92, "learning_rate": 3.9205213146731315e-09, "loss": 0.0253, "step": 32232 }, { "epoch": 9.92, "learning_rate": 3.892668707015101e-09, "loss": 0.0147, "step": 32233 }, { "epoch": 9.92, "learning_rate": 3.864915369915512e-09, "loss": 0.0156, "step": 32234 }, { "epoch": 9.92, "learning_rate": 3.8372613036496975e-09, "loss": 0.0301, "step": 32235 }, { "epoch": 9.92, "learning_rate": 3.809706508492994e-09, "loss": 0.0379, "step": 32236 }, { "epoch": 9.92, "learning_rate": 3.782250984717406e-09, "loss": 0.0242, "step": 32237 }, { "epoch": 9.92, "learning_rate": 3.754894732598269e-09, "loss": 0.0205, "step": 32238 }, { "epoch": 9.92, "learning_rate": 3.727637752405367e-09, "loss": 0.014, "step": 32239 }, { "epoch": 9.92, "learning_rate": 3.7004800444095933e-09, "loss": 0.0252, "step": 32240 }, { "epoch": 9.92, "learning_rate": 3.673421608880734e-09, "loss": 0.021, "step": 32241 }, { "epoch": 9.92, "learning_rate": 3.6464624460885725e-09, "loss": 0.0161, "step": 32242 }, { "epoch": 9.92, "learning_rate": 3.6196025562995616e-09, "loss": 0.0185, "step": 32243 }, { "epoch": 9.92, "learning_rate": 3.5928419397801562e-09, "loss": 0.0249, "step": 32244 }, { "epoch": 9.92, "learning_rate": 3.566180596797919e-09, "loss": 0.0274, "step": 32245 }, { "epoch": 9.92, "learning_rate": 3.5396185276159733e-09, "loss": 0.026, "step": 32246 }, { "epoch": 9.92, "learning_rate": 3.513155732498552e-09, "loss": 0.0209, "step": 32247 }, { "epoch": 9.92, "learning_rate": 3.486792211707668e-09, "loss": 0.0244, "step": 32248 }, { "epoch": 9.92, "learning_rate": 3.460527965507554e-09, "loss": 0.019, "step": 32249 }, { "epoch": 9.92, "learning_rate": 3.4343629941568924e-09, "loss": 0.0184, "step": 32250 }, { "epoch": 9.92, "learning_rate": 3.408297297915475e-09, "loss": 0.0169, "step": 32251 }, { "epoch": 9.92, "learning_rate": 3.382330877044204e-09, "loss": 0.023, "step": 32252 }, { "epoch": 9.92, "learning_rate": 3.3564637317984318e-09, "loss": 0.0072, "step": 32253 }, { "epoch": 9.92, "learning_rate": 3.3306958624357287e-09, "loss": 0.0187, "step": 32254 }, { "epoch": 9.92, "learning_rate": 3.3050272692136674e-09, "loss": 0.0121, "step": 32255 }, { "epoch": 9.92, "learning_rate": 3.2794579523853787e-09, "loss": 0.0483, "step": 32256 }, { "epoch": 9.92, "learning_rate": 3.2539879122051033e-09, "loss": 0.0147, "step": 32257 }, { "epoch": 9.93, "learning_rate": 3.228617148925972e-09, "loss": 0.0399, "step": 32258 }, { "epoch": 9.93, "learning_rate": 3.2033456628000058e-09, "loss": 0.0223, "step": 32259 }, { "epoch": 9.93, "learning_rate": 3.1781734540781153e-09, "loss": 0.019, "step": 32260 }, { "epoch": 9.93, "learning_rate": 3.15310052301121e-09, "loss": 0.0259, "step": 32261 }, { "epoch": 9.93, "learning_rate": 3.12812686984687e-09, "loss": 0.0231, "step": 32262 }, { "epoch": 9.93, "learning_rate": 3.103252494832676e-09, "loss": 0.0146, "step": 32263 }, { "epoch": 9.93, "learning_rate": 3.0784773982173165e-09, "loss": 0.0349, "step": 32264 }, { "epoch": 9.93, "learning_rate": 3.053801580246152e-09, "loss": 0.0262, "step": 32265 }, { "epoch": 9.93, "learning_rate": 3.0292250411645406e-09, "loss": 0.032, "step": 32266 }, { "epoch": 9.93, "learning_rate": 3.004747781215622e-09, "loss": 0.0134, "step": 32267 }, { "epoch": 9.93, "learning_rate": 2.9803698006436454e-09, "loss": 0.0353, "step": 32268 }, { "epoch": 9.93, "learning_rate": 2.9560910996895287e-09, "loss": 0.0188, "step": 32269 }, { "epoch": 9.93, "learning_rate": 2.9319116785953006e-09, "loss": 0.0181, "step": 32270 }, { "epoch": 9.93, "learning_rate": 2.9078315376007694e-09, "loss": 0.0243, "step": 32271 }, { "epoch": 9.93, "learning_rate": 2.883850676944633e-09, "loss": 0.0218, "step": 32272 }, { "epoch": 9.93, "learning_rate": 2.859969096865589e-09, "loss": 0.0302, "step": 32273 }, { "epoch": 9.93, "learning_rate": 2.836186797601226e-09, "loss": 0.0181, "step": 32274 }, { "epoch": 9.93, "learning_rate": 2.8125037793869105e-09, "loss": 0.0474, "step": 32275 }, { "epoch": 9.93, "learning_rate": 2.78892004245801e-09, "loss": 0.0166, "step": 32276 }, { "epoch": 9.93, "learning_rate": 2.7654355870487816e-09, "loss": 0.0237, "step": 32277 }, { "epoch": 9.93, "learning_rate": 2.742050413392372e-09, "loss": 0.0172, "step": 32278 }, { "epoch": 9.93, "learning_rate": 2.7187645217219283e-09, "loss": 0.0408, "step": 32279 }, { "epoch": 9.93, "learning_rate": 2.6955779122672664e-09, "loss": 0.0281, "step": 32280 }, { "epoch": 9.93, "learning_rate": 2.672490585259313e-09, "loss": 0.033, "step": 32281 }, { "epoch": 9.93, "learning_rate": 2.6495025409267737e-09, "loss": 0.0263, "step": 32282 }, { "epoch": 9.93, "learning_rate": 2.626613779499465e-09, "loss": 0.0161, "step": 32283 }, { "epoch": 9.93, "learning_rate": 2.6038243012038724e-09, "loss": 0.0274, "step": 32284 }, { "epoch": 9.93, "learning_rate": 2.5811341062653706e-09, "loss": 0.0144, "step": 32285 }, { "epoch": 9.93, "learning_rate": 2.5585431949104457e-09, "loss": 0.0121, "step": 32286 }, { "epoch": 9.93, "learning_rate": 2.5360515673633625e-09, "loss": 0.0184, "step": 32287 }, { "epoch": 9.93, "learning_rate": 2.5136592238461653e-09, "loss": 0.0267, "step": 32288 }, { "epoch": 9.93, "learning_rate": 2.49136616458312e-09, "loss": 0.0202, "step": 32289 }, { "epoch": 9.94, "learning_rate": 2.4691723897951605e-09, "loss": 0.0215, "step": 32290 }, { "epoch": 9.94, "learning_rate": 2.4470778997021107e-09, "loss": 0.0178, "step": 32291 }, { "epoch": 9.94, "learning_rate": 2.4250826945226847e-09, "loss": 0.0179, "step": 32292 }, { "epoch": 9.94, "learning_rate": 2.403186774477817e-09, "loss": 0.0276, "step": 32293 }, { "epoch": 9.94, "learning_rate": 2.3813901397817806e-09, "loss": 0.0304, "step": 32294 }, { "epoch": 9.94, "learning_rate": 2.3596927906532894e-09, "loss": 0.0414, "step": 32295 }, { "epoch": 9.94, "learning_rate": 2.338094727307727e-09, "loss": 0.0255, "step": 32296 }, { "epoch": 9.94, "learning_rate": 2.316595949958256e-09, "loss": 0.0135, "step": 32297 }, { "epoch": 9.94, "learning_rate": 2.295196458819149e-09, "loss": 0.0231, "step": 32298 }, { "epoch": 9.94, "learning_rate": 2.2738962541035692e-09, "loss": 0.0138, "step": 32299 }, { "epoch": 9.94, "learning_rate": 2.2526953360213487e-09, "loss": 0.0344, "step": 32300 }, { "epoch": 9.94, "learning_rate": 2.2315937047845406e-09, "loss": 0.0189, "step": 32301 }, { "epoch": 9.94, "learning_rate": 2.210591360601866e-09, "loss": 0.0157, "step": 32302 }, { "epoch": 9.94, "learning_rate": 2.1896883036831573e-09, "loss": 0.0223, "step": 32303 }, { "epoch": 9.94, "learning_rate": 2.1688845342338063e-09, "loss": 0.0303, "step": 32304 }, { "epoch": 9.94, "learning_rate": 2.148180052462534e-09, "loss": 0.0164, "step": 32305 }, { "epoch": 9.94, "learning_rate": 2.1275748585747323e-09, "loss": 0.0246, "step": 32306 }, { "epoch": 9.94, "learning_rate": 2.107068952773572e-09, "loss": 0.0342, "step": 32307 }, { "epoch": 9.94, "learning_rate": 2.0866623352633343e-09, "loss": 0.0228, "step": 32308 }, { "epoch": 9.94, "learning_rate": 2.0663550062471893e-09, "loss": 0.0149, "step": 32309 }, { "epoch": 9.94, "learning_rate": 2.0461469659271984e-09, "loss": 0.0397, "step": 32310 }, { "epoch": 9.94, "learning_rate": 2.026038214502091e-09, "loss": 0.0231, "step": 32311 }, { "epoch": 9.94, "learning_rate": 2.0060287521739274e-09, "loss": 0.0263, "step": 32312 }, { "epoch": 9.94, "learning_rate": 1.986118579140328e-09, "loss": 0.0234, "step": 32313 }, { "epoch": 9.94, "learning_rate": 1.9663076955978024e-09, "loss": 0.0245, "step": 32314 }, { "epoch": 9.94, "learning_rate": 1.9465961017450796e-09, "loss": 0.0234, "step": 32315 }, { "epoch": 9.94, "learning_rate": 1.9269837977775595e-09, "loss": 0.0219, "step": 32316 }, { "epoch": 9.94, "learning_rate": 1.907470783889531e-09, "loss": 0.0303, "step": 32317 }, { "epoch": 9.94, "learning_rate": 1.888057060274173e-09, "loss": 0.023, "step": 32318 }, { "epoch": 9.94, "learning_rate": 1.8687426271246646e-09, "loss": 0.0133, "step": 32319 }, { "epoch": 9.94, "learning_rate": 1.8495274846341837e-09, "loss": 0.025, "step": 32320 }, { "epoch": 9.94, "learning_rate": 1.830411632991469e-09, "loss": 0.0292, "step": 32321 }, { "epoch": 9.94, "learning_rate": 1.8113950723874785e-09, "loss": 0.0134, "step": 32322 }, { "epoch": 9.95, "learning_rate": 1.7924778030120605e-09, "loss": 0.0304, "step": 32323 }, { "epoch": 9.95, "learning_rate": 1.773659825050622e-09, "loss": 0.0231, "step": 32324 }, { "epoch": 9.95, "learning_rate": 1.7549411386919014e-09, "loss": 0.029, "step": 32325 }, { "epoch": 9.95, "learning_rate": 1.7363217441213053e-09, "loss": 0.0242, "step": 32326 }, { "epoch": 9.95, "learning_rate": 1.7178016415242416e-09, "loss": 0.02, "step": 32327 }, { "epoch": 9.95, "learning_rate": 1.6993808310827864e-09, "loss": 0.0312, "step": 32328 }, { "epoch": 9.95, "learning_rate": 1.6810593129823472e-09, "loss": 0.0171, "step": 32329 }, { "epoch": 9.95, "learning_rate": 1.66283708740278e-09, "loss": 0.028, "step": 32330 }, { "epoch": 9.95, "learning_rate": 1.6447141545272717e-09, "loss": 0.0178, "step": 32331 }, { "epoch": 9.95, "learning_rate": 1.6266905145334577e-09, "loss": 0.0259, "step": 32332 }, { "epoch": 9.95, "learning_rate": 1.6087661676011945e-09, "loss": 0.0194, "step": 32333 }, { "epoch": 9.95, "learning_rate": 1.590941113909228e-09, "loss": 0.0257, "step": 32334 }, { "epoch": 9.95, "learning_rate": 1.5732153536340832e-09, "loss": 0.0328, "step": 32335 }, { "epoch": 9.95, "learning_rate": 1.5555888869522861e-09, "loss": 0.0218, "step": 32336 }, { "epoch": 9.95, "learning_rate": 1.5380617140370313e-09, "loss": 0.0254, "step": 32337 }, { "epoch": 9.95, "learning_rate": 1.520633835063734e-09, "loss": 0.0262, "step": 32338 }, { "epoch": 9.95, "learning_rate": 1.5033052502066992e-09, "loss": 0.0301, "step": 32339 }, { "epoch": 9.95, "learning_rate": 1.486075959635791e-09, "loss": 0.0235, "step": 32340 }, { "epoch": 9.95, "learning_rate": 1.4689459635230941e-09, "loss": 0.0187, "step": 32341 }, { "epoch": 9.95, "learning_rate": 1.4519152620384725e-09, "loss": 0.0178, "step": 32342 }, { "epoch": 9.95, "learning_rate": 1.4349838553517902e-09, "loss": 0.0508, "step": 32343 }, { "epoch": 9.95, "learning_rate": 1.4181517436306913e-09, "loss": 0.0232, "step": 32344 }, { "epoch": 9.95, "learning_rate": 1.401418927041709e-09, "loss": 0.0203, "step": 32345 }, { "epoch": 9.95, "learning_rate": 1.3847854057524869e-09, "loss": 0.0204, "step": 32346 }, { "epoch": 9.95, "learning_rate": 1.368251179927338e-09, "loss": 0.0229, "step": 32347 }, { "epoch": 9.95, "learning_rate": 1.351816249729465e-09, "loss": 0.0349, "step": 32348 }, { "epoch": 9.95, "learning_rate": 1.3354806153231815e-09, "loss": 0.0221, "step": 32349 }, { "epoch": 9.95, "learning_rate": 1.3192442768716895e-09, "loss": 0.0217, "step": 32350 }, { "epoch": 9.95, "learning_rate": 1.3031072345337514e-09, "loss": 0.0256, "step": 32351 }, { "epoch": 9.95, "learning_rate": 1.2870694884725698e-09, "loss": 0.0205, "step": 32352 }, { "epoch": 9.95, "learning_rate": 1.2711310388446862e-09, "loss": 0.0215, "step": 32353 }, { "epoch": 9.95, "learning_rate": 1.2552918858099727e-09, "loss": 0.0309, "step": 32354 }, { "epoch": 9.96, "learning_rate": 1.2395520295249707e-09, "loss": 0.0201, "step": 32355 }, { "epoch": 9.96, "learning_rate": 1.223911470147332e-09, "loss": 0.0139, "step": 32356 }, { "epoch": 9.96, "learning_rate": 1.2083702078302673e-09, "loss": 0.0203, "step": 32357 }, { "epoch": 9.96, "learning_rate": 1.192928242730318e-09, "loss": 0.0257, "step": 32358 }, { "epoch": 9.96, "learning_rate": 1.1775855749984744e-09, "loss": 0.0229, "step": 32359 }, { "epoch": 9.96, "learning_rate": 1.1623422047901678e-09, "loss": 0.0428, "step": 32360 }, { "epoch": 9.96, "learning_rate": 1.1471981322530579e-09, "loss": 0.0215, "step": 32361 }, { "epoch": 9.96, "learning_rate": 1.1321533575403554e-09, "loss": 0.0223, "step": 32362 }, { "epoch": 9.96, "learning_rate": 1.1172078808008301e-09, "loss": 0.0126, "step": 32363 }, { "epoch": 9.96, "learning_rate": 1.102361702182142e-09, "loss": 0.0227, "step": 32364 }, { "epoch": 9.96, "learning_rate": 1.0876148218330607e-09, "loss": 0.0242, "step": 32365 }, { "epoch": 9.96, "learning_rate": 1.0729672398990254e-09, "loss": 0.0251, "step": 32366 }, { "epoch": 9.96, "learning_rate": 1.0584189565254755e-09, "loss": 0.0193, "step": 32367 }, { "epoch": 9.96, "learning_rate": 1.0439699718567398e-09, "loss": 0.0311, "step": 32368 }, { "epoch": 9.96, "learning_rate": 1.0296202860371473e-09, "loss": 0.0191, "step": 32369 }, { "epoch": 9.96, "learning_rate": 1.0153698992088069e-09, "loss": 0.0298, "step": 32370 }, { "epoch": 9.96, "learning_rate": 1.0012188115127163e-09, "loss": 0.036, "step": 32371 }, { "epoch": 9.96, "learning_rate": 9.871670230898744e-10, "loss": 0.033, "step": 32372 }, { "epoch": 9.96, "learning_rate": 9.73214534080169e-10, "loss": 0.0242, "step": 32373 }, { "epoch": 9.96, "learning_rate": 9.593613446223782e-10, "loss": 0.0147, "step": 32374 }, { "epoch": 9.96, "learning_rate": 9.456074548519489e-10, "loss": 0.0341, "step": 32375 }, { "epoch": 9.96, "learning_rate": 9.319528649087695e-10, "loss": 0.0369, "step": 32376 }, { "epoch": 9.96, "learning_rate": 9.183975749260665e-10, "loss": 0.0221, "step": 32377 }, { "epoch": 9.96, "learning_rate": 9.049415850381771e-10, "loss": 0.0239, "step": 32378 }, { "epoch": 9.96, "learning_rate": 8.915848953805484e-10, "loss": 0.0336, "step": 32379 }, { "epoch": 9.96, "learning_rate": 8.783275060852969e-10, "loss": 0.0343, "step": 32380 }, { "epoch": 9.96, "learning_rate": 8.651694172834291e-10, "loss": 0.0326, "step": 32381 }, { "epoch": 9.96, "learning_rate": 8.521106291048408e-10, "loss": 0.0145, "step": 32382 }, { "epoch": 9.96, "learning_rate": 8.391511416816489e-10, "loss": 0.027, "step": 32383 }, { "epoch": 9.96, "learning_rate": 8.262909551404185e-10, "loss": 0.0142, "step": 32384 }, { "epoch": 9.96, "learning_rate": 8.135300696099357e-10, "loss": 0.0437, "step": 32385 }, { "epoch": 9.96, "learning_rate": 8.008684852167659e-10, "loss": 0.0216, "step": 32386 }, { "epoch": 9.96, "learning_rate": 7.883062020863641e-10, "loss": 0.0339, "step": 32387 }, { "epoch": 9.97, "learning_rate": 7.758432203430755e-10, "loss": 0.0188, "step": 32388 }, { "epoch": 9.97, "learning_rate": 7.634795401123552e-10, "loss": 0.0189, "step": 32389 }, { "epoch": 9.97, "learning_rate": 7.512151615152174e-10, "loss": 0.0266, "step": 32390 }, { "epoch": 9.97, "learning_rate": 7.39050084673787e-10, "loss": 0.0165, "step": 32391 }, { "epoch": 9.97, "learning_rate": 7.269843097090779e-10, "loss": 0.014, "step": 32392 }, { "epoch": 9.97, "learning_rate": 7.150178367421046e-10, "loss": 0.0161, "step": 32393 }, { "epoch": 9.97, "learning_rate": 7.031506658894405e-10, "loss": 0.011, "step": 32394 }, { "epoch": 9.97, "learning_rate": 6.913827972709897e-10, "loss": 0.0281, "step": 32395 }, { "epoch": 9.97, "learning_rate": 6.797142310022154e-10, "loss": 0.0227, "step": 32396 }, { "epoch": 9.97, "learning_rate": 6.681449671996909e-10, "loss": 0.0214, "step": 32397 }, { "epoch": 9.97, "learning_rate": 6.566750059777694e-10, "loss": 0.0382, "step": 32398 }, { "epoch": 9.97, "learning_rate": 6.453043474519138e-10, "loss": 0.0197, "step": 32399 }, { "epoch": 9.97, "learning_rate": 6.340329917320365e-10, "loss": 0.0497, "step": 32400 }, { "epoch": 9.97, "learning_rate": 6.228609389336004e-10, "loss": 0.0179, "step": 32401 }, { "epoch": 9.97, "learning_rate": 6.117881891654077e-10, "loss": 0.0308, "step": 32402 }, { "epoch": 9.97, "learning_rate": 6.008147425373701e-10, "loss": 0.0272, "step": 32403 }, { "epoch": 9.97, "learning_rate": 5.899405991593998e-10, "loss": 0.0259, "step": 32404 }, { "epoch": 9.97, "learning_rate": 5.791657591380783e-10, "loss": 0.025, "step": 32405 }, { "epoch": 9.97, "learning_rate": 5.684902225822076e-10, "loss": 0.0155, "step": 32406 }, { "epoch": 9.97, "learning_rate": 5.579139895972586e-10, "loss": 0.0201, "step": 32407 }, { "epoch": 9.97, "learning_rate": 5.474370602875923e-10, "loss": 0.0296, "step": 32408 }, { "epoch": 9.97, "learning_rate": 5.370594347575697e-10, "loss": 0.0238, "step": 32409 }, { "epoch": 9.97, "learning_rate": 5.267811131104417e-10, "loss": 0.0316, "step": 32410 }, { "epoch": 9.97, "learning_rate": 5.166020954483486e-10, "loss": 0.0108, "step": 32411 }, { "epoch": 9.97, "learning_rate": 5.065223818723208e-10, "loss": 0.0271, "step": 32412 }, { "epoch": 9.97, "learning_rate": 4.965419724811682e-10, "loss": 0.025, "step": 32413 }, { "epoch": 9.97, "learning_rate": 4.86660867375921e-10, "loss": 0.0192, "step": 32414 }, { "epoch": 9.97, "learning_rate": 4.768790666542789e-10, "loss": 0.0257, "step": 32415 }, { "epoch": 9.97, "learning_rate": 4.671965704128312e-10, "loss": 0.0185, "step": 32416 }, { "epoch": 9.97, "learning_rate": 4.5761337874816735e-10, "loss": 0.0212, "step": 32417 }, { "epoch": 9.97, "learning_rate": 4.4812949175465634e-10, "loss": 0.0196, "step": 32418 }, { "epoch": 9.97, "learning_rate": 4.3874490952777735e-10, "loss": 0.0275, "step": 32419 }, { "epoch": 9.98, "learning_rate": 4.2945963215967887e-10, "loss": 0.0278, "step": 32420 }, { "epoch": 9.98, "learning_rate": 4.2027365974250943e-10, "loss": 0.0326, "step": 32421 }, { "epoch": 9.98, "learning_rate": 4.1118699236841753e-10, "loss": 0.0211, "step": 32422 }, { "epoch": 9.98, "learning_rate": 4.021996301273312e-10, "loss": 0.0245, "step": 32423 }, { "epoch": 9.98, "learning_rate": 3.933115731080683e-10, "loss": 0.0204, "step": 32424 }, { "epoch": 9.98, "learning_rate": 3.8452282139944673e-10, "loss": 0.0258, "step": 32425 }, { "epoch": 9.98, "learning_rate": 3.7583337508806384e-10, "loss": 0.0323, "step": 32426 }, { "epoch": 9.98, "learning_rate": 3.67243234260517e-10, "loss": 0.0176, "step": 32427 }, { "epoch": 9.98, "learning_rate": 3.5875239900229343e-10, "loss": 0.0272, "step": 32428 }, { "epoch": 9.98, "learning_rate": 3.5036086939777003e-10, "loss": 0.0171, "step": 32429 }, { "epoch": 9.98, "learning_rate": 3.4206864552910334e-10, "loss": 0.013, "step": 32430 }, { "epoch": 9.98, "learning_rate": 3.338757274806703e-10, "loss": 0.0214, "step": 32431 }, { "epoch": 9.98, "learning_rate": 3.25782115332407e-10, "loss": 0.0275, "step": 32432 }, { "epoch": 9.98, "learning_rate": 3.177878091653597e-10, "loss": 0.0185, "step": 32433 }, { "epoch": 9.98, "learning_rate": 3.098928090594644e-10, "loss": 0.0134, "step": 32434 }, { "epoch": 9.98, "learning_rate": 3.0209711509132657e-10, "loss": 0.0211, "step": 32435 }, { "epoch": 9.98, "learning_rate": 2.9440072733866175e-10, "loss": 0.0345, "step": 32436 }, { "epoch": 9.98, "learning_rate": 2.868036458802959e-10, "loss": 0.0257, "step": 32437 }, { "epoch": 9.98, "learning_rate": 2.7930587078839333e-10, "loss": 0.0218, "step": 32438 }, { "epoch": 9.98, "learning_rate": 2.7190740213955957e-10, "loss": 0.0237, "step": 32439 }, { "epoch": 9.98, "learning_rate": 2.646082400070693e-10, "loss": 0.0357, "step": 32440 }, { "epoch": 9.98, "learning_rate": 2.5740838446308703e-10, "loss": 0.0362, "step": 32441 }, { "epoch": 9.98, "learning_rate": 2.50307835578667e-10, "loss": 0.0112, "step": 32442 }, { "epoch": 9.98, "learning_rate": 2.433065934248635e-10, "loss": 0.0299, "step": 32443 }, { "epoch": 9.98, "learning_rate": 2.364046580716206e-10, "loss": 0.0181, "step": 32444 }, { "epoch": 9.98, "learning_rate": 2.2960202958555166e-10, "loss": 0.0146, "step": 32445 }, { "epoch": 9.98, "learning_rate": 2.2289870803660073e-10, "loss": 0.0241, "step": 32446 }, { "epoch": 9.98, "learning_rate": 2.1629469349027098e-10, "loss": 0.0318, "step": 32447 }, { "epoch": 9.98, "learning_rate": 2.0978998601206558e-10, "loss": 0.0176, "step": 32448 }, { "epoch": 9.98, "learning_rate": 2.0338458566748763e-10, "loss": 0.0329, "step": 32449 }, { "epoch": 9.98, "learning_rate": 1.970784925187097e-10, "loss": 0.0318, "step": 32450 }, { "epoch": 9.98, "learning_rate": 1.9087170662790422e-10, "loss": 0.0296, "step": 32451 }, { "epoch": 9.98, "learning_rate": 1.8476422805946415e-10, "loss": 0.0283, "step": 32452 }, { "epoch": 9.99, "learning_rate": 1.787560568722313e-10, "loss": 0.0152, "step": 32453 }, { "epoch": 9.99, "learning_rate": 1.728471931250475e-10, "loss": 0.0272, "step": 32454 }, { "epoch": 9.99, "learning_rate": 1.67037636878975e-10, "loss": 0.0493, "step": 32455 }, { "epoch": 9.99, "learning_rate": 1.6132738818952498e-10, "loss": 0.0123, "step": 32456 }, { "epoch": 9.99, "learning_rate": 1.5571644711442902e-10, "loss": 0.0201, "step": 32457 }, { "epoch": 9.99, "learning_rate": 1.5020481370808805e-10, "loss": 0.0318, "step": 32458 }, { "epoch": 9.99, "learning_rate": 1.4479248802823365e-10, "loss": 0.0213, "step": 32459 }, { "epoch": 9.99, "learning_rate": 1.394794701259361e-10, "loss": 0.019, "step": 32460 }, { "epoch": 9.99, "learning_rate": 1.342657600544861e-10, "loss": 0.0211, "step": 32461 }, { "epoch": 9.99, "learning_rate": 1.2915135786606414e-10, "loss": 0.0236, "step": 32462 }, { "epoch": 9.99, "learning_rate": 1.2413626361063025e-10, "loss": 0.0138, "step": 32463 }, { "epoch": 9.99, "learning_rate": 1.1922047733925467e-10, "loss": 0.0198, "step": 32464 }, { "epoch": 9.99, "learning_rate": 1.1440399910078725e-10, "loss": 0.0178, "step": 32465 }, { "epoch": 9.99, "learning_rate": 1.0968682894185734e-10, "loss": 0.0126, "step": 32466 }, { "epoch": 9.99, "learning_rate": 1.0506896691020451e-10, "loss": 0.0249, "step": 32467 }, { "epoch": 9.99, "learning_rate": 1.0055041305134794e-10, "loss": 0.0156, "step": 32468 }, { "epoch": 9.99, "learning_rate": 9.613116740969652e-11, "loss": 0.0198, "step": 32469 }, { "epoch": 9.99, "learning_rate": 9.181123003076942e-11, "loss": 0.0209, "step": 32470 }, { "epoch": 9.99, "learning_rate": 8.759060095453465e-11, "loss": 0.0409, "step": 32471 }, { "epoch": 9.99, "learning_rate": 8.346928022651135e-11, "loss": 0.021, "step": 32472 }, { "epoch": 9.99, "learning_rate": 7.944726788444711e-11, "loss": 0.013, "step": 32473 }, { "epoch": 9.99, "learning_rate": 7.552456397053042e-11, "loss": 0.0205, "step": 32474 }, { "epoch": 9.99, "learning_rate": 7.170116852250886e-11, "loss": 0.0168, "step": 32475 }, { "epoch": 9.99, "learning_rate": 6.797708157924021e-11, "loss": 0.0179, "step": 32476 }, { "epoch": 9.99, "learning_rate": 6.435230317625163e-11, "loss": 0.0257, "step": 32477 }, { "epoch": 9.99, "learning_rate": 6.082683335129069e-11, "loss": 0.0207, "step": 32478 }, { "epoch": 9.99, "learning_rate": 5.740067213877432e-11, "loss": 0.0423, "step": 32479 }, { "epoch": 9.99, "learning_rate": 5.407381957200919e-11, "loss": 0.0215, "step": 32480 }, { "epoch": 9.99, "learning_rate": 5.0846275684302e-11, "loss": 0.026, "step": 32481 }, { "epoch": 9.99, "learning_rate": 4.771804050784923e-11, "loss": 0.0461, "step": 32482 }, { "epoch": 9.99, "learning_rate": 4.4689114073737105e-11, "loss": 0.0367, "step": 32483 }, { "epoch": 9.99, "learning_rate": 4.175949641194166e-11, "loss": 0.03, "step": 32484 }, { "epoch": 10.0, "learning_rate": 3.892918755243891e-11, "loss": 0.0306, "step": 32485 }, { "epoch": 10.0, "learning_rate": 3.619818752187421e-11, "loss": 0.0212, "step": 32486 }, { "epoch": 10.0, "learning_rate": 3.3566496349113355e-11, "loss": 0.0165, "step": 32487 }, { "epoch": 10.0, "learning_rate": 3.103411405858125e-11, "loss": 0.0201, "step": 32488 }, { "epoch": 10.0, "learning_rate": 2.8601040675813042e-11, "loss": 0.022, "step": 32489 }, { "epoch": 10.0, "learning_rate": 2.626727622634384e-11, "loss": 0.0326, "step": 32490 }, { "epoch": 10.0, "learning_rate": 2.4032820731267892e-11, "loss": 0.0185, "step": 32491 }, { "epoch": 10.0, "learning_rate": 2.1897674213899877e-11, "loss": 0.0311, "step": 32492 }, { "epoch": 10.0, "learning_rate": 1.9861836695334036e-11, "loss": 0.0114, "step": 32493 }, { "epoch": 10.0, "learning_rate": 1.7925308196664603e-11, "loss": 0.0204, "step": 32494 }, { "epoch": 10.0, "learning_rate": 1.608808873565515e-11, "loss": 0.0199, "step": 32495 }, { "epoch": 10.0, "learning_rate": 1.4350178331179465e-11, "loss": 0.0417, "step": 32496 }, { "epoch": 10.0, "learning_rate": 1.2711576999890896e-11, "loss": 0.0321, "step": 32497 }, { "epoch": 10.0, "learning_rate": 1.117228475955301e-11, "loss": 0.0161, "step": 32498 }, { "epoch": 10.0, "learning_rate": 9.732301623488482e-12, "loss": 0.0322, "step": 32499 }, { "epoch": 10.0, "learning_rate": 8.39162760835066e-12, "loss": 0.0144, "step": 32500 }, { "epoch": 10.0, "learning_rate": 1.0491594320006755e-05, "loss": 0.0247, "step": 32501 }, { "epoch": 10.0, "learning_rate": 1.0491096651784847e-05, "loss": 0.023, "step": 32502 }, { "epoch": 10.0, "learning_rate": 1.0490598982343678e-05, "loss": 0.0174, "step": 32503 }, { "epoch": 10.0, "learning_rate": 1.0490101311684485e-05, "loss": 0.0151, "step": 32504 }, { "epoch": 10.0, "learning_rate": 1.04896036398085e-05, "loss": 0.0209, "step": 32505 }, { "epoch": 10.0, "learning_rate": 1.0489105966716964e-05, "loss": 0.0123, "step": 32506 }, { "epoch": 10.0, "learning_rate": 1.0488608292411106e-05, "loss": 0.0224, "step": 32507 }, { "epoch": 10.0, "learning_rate": 1.0488110616892169e-05, "loss": 0.0173, "step": 32508 }, { "epoch": 10.0, "learning_rate": 1.0487612940161384e-05, "loss": 0.0061, "step": 32509 }, { "epoch": 10.0, "learning_rate": 1.0487115262219987e-05, "loss": 0.023, "step": 32510 }, { "epoch": 10.0, "learning_rate": 1.0486617583069214e-05, "loss": 0.0152, "step": 32511 }, { "epoch": 10.0, "learning_rate": 1.0486119902710301e-05, "loss": 0.0186, "step": 32512 }, { "epoch": 10.0, "learning_rate": 1.0485622221144485e-05, "loss": 0.0251, "step": 32513 }, { "epoch": 10.0, "learning_rate": 1.0485124538372997e-05, "loss": 0.0157, "step": 32514 }, { "epoch": 10.0, "learning_rate": 1.0484626854397078e-05, "loss": 0.0285, "step": 32515 }, { "epoch": 10.0, "learning_rate": 1.048412916921796e-05, "loss": 0.03, "step": 32516 }, { "epoch": 10.01, "learning_rate": 1.0483631482836878e-05, "loss": 0.0362, "step": 32517 }, { "epoch": 10.01, "learning_rate": 1.0483133795255072e-05, "loss": 0.0324, "step": 32518 }, { "epoch": 10.01, "learning_rate": 1.0482636106473777e-05, "loss": 0.0234, "step": 32519 }, { "epoch": 10.01, "learning_rate": 1.0482138416494221e-05, "loss": 0.025, "step": 32520 }, { "epoch": 10.01, "learning_rate": 1.0481640725317647e-05, "loss": 0.0268, "step": 32521 }, { "epoch": 10.01, "learning_rate": 1.0481143032945292e-05, "loss": 0.029, "step": 32522 }, { "epoch": 10.01, "learning_rate": 1.0480645339378386e-05, "loss": 0.0207, "step": 32523 }, { "epoch": 10.01, "learning_rate": 1.0480147644618168e-05, "loss": 0.034, "step": 32524 }, { "epoch": 10.01, "learning_rate": 1.047964994866587e-05, "loss": 0.0249, "step": 32525 }, { "epoch": 10.01, "learning_rate": 1.0479152251522737e-05, "loss": 0.0176, "step": 32526 }, { "epoch": 10.01, "learning_rate": 1.0478654553189994e-05, "loss": 0.0394, "step": 32527 }, { "epoch": 10.01, "learning_rate": 1.0478156853668882e-05, "loss": 0.0159, "step": 32528 }, { "epoch": 10.01, "learning_rate": 1.0477659152960636e-05, "loss": 0.0289, "step": 32529 }, { "epoch": 10.01, "learning_rate": 1.0477161451066486e-05, "loss": 0.0295, "step": 32530 }, { "epoch": 10.01, "learning_rate": 1.0476663747987678e-05, "loss": 0.0771, "step": 32531 }, { "epoch": 10.01, "learning_rate": 1.0476166043725442e-05, "loss": 0.0367, "step": 32532 }, { "epoch": 10.01, "learning_rate": 1.0475668338281013e-05, "loss": 0.0278, "step": 32533 }, { "epoch": 10.01, "learning_rate": 1.0475170631655627e-05, "loss": 0.0368, "step": 32534 }, { "epoch": 10.01, "learning_rate": 1.047467292385052e-05, "loss": 0.0138, "step": 32535 }, { "epoch": 10.01, "learning_rate": 1.0474175214866931e-05, "loss": 0.0345, "step": 32536 }, { "epoch": 10.01, "learning_rate": 1.047367750470609e-05, "loss": 0.0167, "step": 32537 }, { "epoch": 10.01, "learning_rate": 1.0473179793369237e-05, "loss": 0.0457, "step": 32538 }, { "epoch": 10.01, "learning_rate": 1.0472682080857606e-05, "loss": 0.0578, "step": 32539 }, { "epoch": 10.01, "learning_rate": 1.047218436717243e-05, "loss": 0.0495, "step": 32540 }, { "epoch": 10.01, "learning_rate": 1.0471686652314952e-05, "loss": 0.0239, "step": 32541 }, { "epoch": 10.01, "learning_rate": 1.0471188936286399e-05, "loss": 0.0142, "step": 32542 }, { "epoch": 10.01, "learning_rate": 1.0470691219088016e-05, "loss": 0.0465, "step": 32543 }, { "epoch": 10.01, "learning_rate": 1.047019350072103e-05, "loss": 0.0252, "step": 32544 }, { "epoch": 10.01, "learning_rate": 1.046969578118668e-05, "loss": 0.034, "step": 32545 }, { "epoch": 10.01, "learning_rate": 1.0469198060486203e-05, "loss": 0.0286, "step": 32546 }, { "epoch": 10.01, "learning_rate": 1.0468700338620835e-05, "loss": 0.0289, "step": 32547 }, { "epoch": 10.01, "learning_rate": 1.0468202615591807e-05, "loss": 0.039, "step": 32548 }, { "epoch": 10.02, "learning_rate": 1.0467704891400361e-05, "loss": 0.0295, "step": 32549 }, { "epoch": 10.02, "learning_rate": 1.046720716604773e-05, "loss": 0.0321, "step": 32550 }, { "epoch": 10.02, "learning_rate": 1.0466709439535148e-05, "loss": 0.0292, "step": 32551 }, { "epoch": 10.02, "learning_rate": 1.0466211711863851e-05, "loss": 0.0238, "step": 32552 }, { "epoch": 10.02, "learning_rate": 1.0465713983035082e-05, "loss": 0.0222, "step": 32553 }, { "epoch": 10.02, "learning_rate": 1.0465216253050066e-05, "loss": 0.0407, "step": 32554 }, { "epoch": 10.02, "learning_rate": 1.0464718521910043e-05, "loss": 0.0285, "step": 32555 }, { "epoch": 10.02, "learning_rate": 1.046422078961625e-05, "loss": 0.0147, "step": 32556 }, { "epoch": 10.02, "learning_rate": 1.0463723056169922e-05, "loss": 0.0328, "step": 32557 }, { "epoch": 10.02, "learning_rate": 1.0463225321572294e-05, "loss": 0.0478, "step": 32558 }, { "epoch": 10.02, "learning_rate": 1.0462727585824606e-05, "loss": 0.0215, "step": 32559 }, { "epoch": 10.02, "learning_rate": 1.0462229848928087e-05, "loss": 0.0273, "step": 32560 }, { "epoch": 10.02, "learning_rate": 1.0461732110883975e-05, "loss": 0.0377, "step": 32561 }, { "epoch": 10.02, "learning_rate": 1.0461234371693508e-05, "loss": 0.0324, "step": 32562 }, { "epoch": 10.02, "learning_rate": 1.046073663135792e-05, "loss": 0.0101, "step": 32563 }, { "epoch": 10.02, "learning_rate": 1.046023888987845e-05, "loss": 0.0489, "step": 32564 }, { "epoch": 10.02, "learning_rate": 1.0459741147256325e-05, "loss": 0.0171, "step": 32565 }, { "epoch": 10.02, "learning_rate": 1.0459243403492792e-05, "loss": 0.0431, "step": 32566 }, { "epoch": 10.02, "learning_rate": 1.045874565858908e-05, "loss": 0.0305, "step": 32567 }, { "epoch": 10.02, "learning_rate": 1.0458247912546427e-05, "loss": 0.0283, "step": 32568 }, { "epoch": 10.02, "learning_rate": 1.0457750165366064e-05, "loss": 0.0324, "step": 32569 }, { "epoch": 10.02, "learning_rate": 1.0457252417049236e-05, "loss": 0.0263, "step": 32570 }, { "epoch": 10.02, "learning_rate": 1.0456754667597171e-05, "loss": 0.0351, "step": 32571 }, { "epoch": 10.02, "learning_rate": 1.0456256917011104e-05, "loss": 0.0279, "step": 32572 }, { "epoch": 10.02, "learning_rate": 1.0455759165292278e-05, "loss": 0.0438, "step": 32573 }, { "epoch": 10.02, "learning_rate": 1.0455261412441927e-05, "loss": 0.0288, "step": 32574 }, { "epoch": 10.02, "learning_rate": 1.045476365846128e-05, "loss": 0.0207, "step": 32575 }, { "epoch": 10.02, "learning_rate": 1.0454265903351578e-05, "loss": 0.0284, "step": 32576 }, { "epoch": 10.02, "learning_rate": 1.045376814711406e-05, "loss": 0.0483, "step": 32577 }, { "epoch": 10.02, "learning_rate": 1.0453270389749956e-05, "loss": 0.0371, "step": 32578 }, { "epoch": 10.02, "learning_rate": 1.0452772631260504e-05, "loss": 0.0621, "step": 32579 }, { "epoch": 10.02, "learning_rate": 1.0452274871646937e-05, "loss": 0.0261, "step": 32580 }, { "epoch": 10.02, "learning_rate": 1.0451777110910497e-05, "loss": 0.0323, "step": 32581 }, { "epoch": 10.03, "learning_rate": 1.0451279349052412e-05, "loss": 0.0272, "step": 32582 }, { "epoch": 10.03, "learning_rate": 1.0450781586073926e-05, "loss": 0.0248, "step": 32583 }, { "epoch": 10.03, "learning_rate": 1.045028382197627e-05, "loss": 0.0329, "step": 32584 }, { "epoch": 10.03, "learning_rate": 1.044978605676068e-05, "loss": 0.0373, "step": 32585 }, { "epoch": 10.03, "learning_rate": 1.0449288290428394e-05, "loss": 0.0305, "step": 32586 }, { "epoch": 10.03, "learning_rate": 1.0448790522980642e-05, "loss": 0.0457, "step": 32587 }, { "epoch": 10.03, "learning_rate": 1.0448292754418669e-05, "loss": 0.0395, "step": 32588 }, { "epoch": 10.03, "learning_rate": 1.0447794984743705e-05, "loss": 0.0788, "step": 32589 }, { "epoch": 10.03, "learning_rate": 1.0447297213956984e-05, "loss": 0.0295, "step": 32590 }, { "epoch": 10.03, "learning_rate": 1.044679944205975e-05, "loss": 0.0352, "step": 32591 }, { "epoch": 10.03, "learning_rate": 1.0446301669053227e-05, "loss": 0.0217, "step": 32592 }, { "epoch": 10.03, "learning_rate": 1.044580389493866e-05, "loss": 0.0256, "step": 32593 }, { "epoch": 10.03, "learning_rate": 1.0445306119717282e-05, "loss": 0.0372, "step": 32594 }, { "epoch": 10.03, "learning_rate": 1.0444808343390332e-05, "loss": 0.0331, "step": 32595 }, { "epoch": 10.03, "learning_rate": 1.044431056595904e-05, "loss": 0.0156, "step": 32596 }, { "epoch": 10.03, "learning_rate": 1.0443812787424641e-05, "loss": 0.0389, "step": 32597 }, { "epoch": 10.03, "learning_rate": 1.0443315007788381e-05, "loss": 0.0205, "step": 32598 }, { "epoch": 10.03, "learning_rate": 1.0442817227051487e-05, "loss": 0.0212, "step": 32599 }, { "epoch": 10.03, "learning_rate": 1.0442319445215195e-05, "loss": 0.0371, "step": 32600 }, { "epoch": 10.03, "learning_rate": 1.0441821662280745e-05, "loss": 0.0233, "step": 32601 }, { "epoch": 10.03, "learning_rate": 1.044132387824937e-05, "loss": 0.0314, "step": 32602 }, { "epoch": 10.03, "learning_rate": 1.0440826093122309e-05, "loss": 0.0397, "step": 32603 }, { "epoch": 10.03, "learning_rate": 1.0440328306900791e-05, "loss": 0.0674, "step": 32604 }, { "epoch": 10.03, "learning_rate": 1.0439830519586063e-05, "loss": 0.0319, "step": 32605 }, { "epoch": 10.03, "learning_rate": 1.043933273117935e-05, "loss": 0.0632, "step": 32606 }, { "epoch": 10.03, "learning_rate": 1.0438834941681893e-05, "loss": 0.0313, "step": 32607 }, { "epoch": 10.03, "learning_rate": 1.043833715109493e-05, "loss": 0.0227, "step": 32608 }, { "epoch": 10.03, "learning_rate": 1.0437839359419688e-05, "loss": 0.0503, "step": 32609 }, { "epoch": 10.03, "learning_rate": 1.0437341566657413e-05, "loss": 0.0566, "step": 32610 }, { "epoch": 10.03, "learning_rate": 1.0436843772809338e-05, "loss": 0.0334, "step": 32611 }, { "epoch": 10.03, "learning_rate": 1.0436345977876697e-05, "loss": 0.0659, "step": 32612 }, { "epoch": 10.03, "learning_rate": 1.0435848181860724e-05, "loss": 0.0306, "step": 32613 }, { "epoch": 10.04, "learning_rate": 1.0435350384762656e-05, "loss": 0.0444, "step": 32614 }, { "epoch": 10.04, "learning_rate": 1.0434852586583737e-05, "loss": 0.0322, "step": 32615 }, { "epoch": 10.04, "learning_rate": 1.0434354787325194e-05, "loss": 0.0368, "step": 32616 }, { "epoch": 10.04, "learning_rate": 1.043385698698826e-05, "loss": 0.0392, "step": 32617 }, { "epoch": 10.04, "learning_rate": 1.043335918557418e-05, "loss": 0.0154, "step": 32618 }, { "epoch": 10.04, "learning_rate": 1.0432861383084187e-05, "loss": 0.0376, "step": 32619 }, { "epoch": 10.04, "learning_rate": 1.0432363579519516e-05, "loss": 0.0435, "step": 32620 }, { "epoch": 10.04, "learning_rate": 1.04318657748814e-05, "loss": 0.0423, "step": 32621 }, { "epoch": 10.04, "learning_rate": 1.043136796917108e-05, "loss": 0.04, "step": 32622 }, { "epoch": 10.04, "learning_rate": 1.0430870162389787e-05, "loss": 0.0228, "step": 32623 }, { "epoch": 10.04, "learning_rate": 1.043037235453876e-05, "loss": 0.0453, "step": 32624 }, { "epoch": 10.04, "learning_rate": 1.0429874545619236e-05, "loss": 0.0461, "step": 32625 }, { "epoch": 10.04, "learning_rate": 1.0429376735632453e-05, "loss": 0.0479, "step": 32626 }, { "epoch": 10.04, "learning_rate": 1.0428878924579638e-05, "loss": 0.0059, "step": 32627 }, { "epoch": 10.04, "learning_rate": 1.0428381112462035e-05, "loss": 0.0458, "step": 32628 }, { "epoch": 10.04, "learning_rate": 1.0427883299280875e-05, "loss": 0.0369, "step": 32629 }, { "epoch": 10.04, "learning_rate": 1.0427385485037398e-05, "loss": 0.0447, "step": 32630 }, { "epoch": 10.04, "learning_rate": 1.0426887669732837e-05, "loss": 0.0368, "step": 32631 }, { "epoch": 10.04, "learning_rate": 1.0426389853368429e-05, "loss": 0.0584, "step": 32632 }, { "epoch": 10.04, "learning_rate": 1.0425892035945411e-05, "loss": 0.0275, "step": 32633 }, { "epoch": 10.04, "learning_rate": 1.0425394217465015e-05, "loss": 0.0401, "step": 32634 }, { "epoch": 10.04, "learning_rate": 1.042489639792848e-05, "loss": 0.0807, "step": 32635 }, { "epoch": 10.04, "learning_rate": 1.0424398577337048e-05, "loss": 0.0305, "step": 32636 }, { "epoch": 10.04, "learning_rate": 1.0423900755691942e-05, "loss": 0.0539, "step": 32637 }, { "epoch": 10.04, "learning_rate": 1.0423402932994406e-05, "loss": 0.0594, "step": 32638 }, { "epoch": 10.04, "learning_rate": 1.0422905109245678e-05, "loss": 0.0571, "step": 32639 }, { "epoch": 10.04, "learning_rate": 1.0422407284446987e-05, "loss": 0.0386, "step": 32640 }, { "epoch": 10.04, "learning_rate": 1.0421909458599574e-05, "loss": 0.0426, "step": 32641 }, { "epoch": 10.04, "learning_rate": 1.0421411631704672e-05, "loss": 0.0636, "step": 32642 }, { "epoch": 10.04, "learning_rate": 1.0420913803763522e-05, "loss": 0.0429, "step": 32643 }, { "epoch": 10.04, "learning_rate": 1.0420415974777354e-05, "loss": 0.0413, "step": 32644 }, { "epoch": 10.04, "learning_rate": 1.0419918144747408e-05, "loss": 0.0328, "step": 32645 }, { "epoch": 10.04, "learning_rate": 1.0419420313674915e-05, "loss": 0.0389, "step": 32646 }, { "epoch": 10.05, "learning_rate": 1.0418922481561119e-05, "loss": 0.0169, "step": 32647 }, { "epoch": 10.05, "learning_rate": 1.0418424648407249e-05, "loss": 0.0473, "step": 32648 }, { "epoch": 10.05, "learning_rate": 1.0417926814214542e-05, "loss": 0.0249, "step": 32649 }, { "epoch": 10.05, "learning_rate": 1.0417428978984239e-05, "loss": 0.0336, "step": 32650 }, { "epoch": 10.05, "learning_rate": 1.041693114271757e-05, "loss": 0.0372, "step": 32651 }, { "epoch": 10.05, "learning_rate": 1.0416433305415772e-05, "loss": 0.0459, "step": 32652 }, { "epoch": 10.05, "learning_rate": 1.0415935467080087e-05, "loss": 0.0269, "step": 32653 }, { "epoch": 10.05, "learning_rate": 1.0415437627711741e-05, "loss": 0.0403, "step": 32654 }, { "epoch": 10.05, "learning_rate": 1.0414939787311979e-05, "loss": 0.0642, "step": 32655 }, { "epoch": 10.05, "learning_rate": 1.0414441945882033e-05, "loss": 0.0651, "step": 32656 }, { "epoch": 10.05, "learning_rate": 1.0413944103423139e-05, "loss": 0.071, "step": 32657 }, { "epoch": 10.05, "learning_rate": 1.0413446259936532e-05, "loss": 0.0405, "step": 32658 }, { "epoch": 10.05, "learning_rate": 1.041294841542345e-05, "loss": 0.0576, "step": 32659 }, { "epoch": 10.05, "learning_rate": 1.0412450569885131e-05, "loss": 0.0319, "step": 32660 }, { "epoch": 10.05, "learning_rate": 1.0411952723322806e-05, "loss": 0.0385, "step": 32661 }, { "epoch": 10.05, "learning_rate": 1.0411454875737712e-05, "loss": 0.0435, "step": 32662 }, { "epoch": 10.05, "learning_rate": 1.0410957027131089e-05, "loss": 0.0322, "step": 32663 }, { "epoch": 10.05, "learning_rate": 1.0410459177504171e-05, "loss": 0.0339, "step": 32664 }, { "epoch": 10.05, "learning_rate": 1.0409961326858192e-05, "loss": 0.0257, "step": 32665 }, { "epoch": 10.05, "learning_rate": 1.040946347519439e-05, "loss": 0.0386, "step": 32666 }, { "epoch": 10.05, "learning_rate": 1.0408965622514e-05, "loss": 0.0409, "step": 32667 }, { "epoch": 10.05, "learning_rate": 1.040846776881826e-05, "loss": 0.0482, "step": 32668 }, { "epoch": 10.05, "learning_rate": 1.04079699141084e-05, "loss": 0.0246, "step": 32669 }, { "epoch": 10.05, "learning_rate": 1.0407472058385668e-05, "loss": 0.0335, "step": 32670 }, { "epoch": 10.05, "learning_rate": 1.040697420165129e-05, "loss": 0.0389, "step": 32671 }, { "epoch": 10.05, "learning_rate": 1.0406476343906502e-05, "loss": 0.0402, "step": 32672 }, { "epoch": 10.05, "learning_rate": 1.0405978485152544e-05, "loss": 0.0301, "step": 32673 }, { "epoch": 10.05, "learning_rate": 1.0405480625390652e-05, "loss": 0.0626, "step": 32674 }, { "epoch": 10.05, "learning_rate": 1.0404982764622061e-05, "loss": 0.0654, "step": 32675 }, { "epoch": 10.05, "learning_rate": 1.0404484902848007e-05, "loss": 0.0281, "step": 32676 }, { "epoch": 10.05, "learning_rate": 1.0403987040069725e-05, "loss": 0.0288, "step": 32677 }, { "epoch": 10.05, "learning_rate": 1.0403489176288452e-05, "loss": 0.0557, "step": 32678 }, { "epoch": 10.06, "learning_rate": 1.0402991311505423e-05, "loss": 0.0453, "step": 32679 }, { "epoch": 10.06, "learning_rate": 1.0402493445721878e-05, "loss": 0.0546, "step": 32680 }, { "epoch": 10.06, "learning_rate": 1.040199557893905e-05, "loss": 0.0562, "step": 32681 }, { "epoch": 10.06, "learning_rate": 1.0401497711158173e-05, "loss": 0.0258, "step": 32682 }, { "epoch": 10.06, "learning_rate": 1.0400999842380488e-05, "loss": 0.0366, "step": 32683 }, { "epoch": 10.06, "learning_rate": 1.040050197260723e-05, "loss": 0.0555, "step": 32684 }, { "epoch": 10.06, "learning_rate": 1.040000410183963e-05, "loss": 0.0377, "step": 32685 }, { "epoch": 10.06, "learning_rate": 1.0399506230078928e-05, "loss": 0.0648, "step": 32686 }, { "epoch": 10.06, "learning_rate": 1.0399008357326361e-05, "loss": 0.0475, "step": 32687 }, { "epoch": 10.06, "learning_rate": 1.0398510483583164e-05, "loss": 0.0603, "step": 32688 }, { "epoch": 10.06, "learning_rate": 1.039801260885057e-05, "loss": 0.0396, "step": 32689 }, { "epoch": 10.06, "learning_rate": 1.0397514733129818e-05, "loss": 0.0428, "step": 32690 }, { "epoch": 10.06, "learning_rate": 1.0397016856422151e-05, "loss": 0.0367, "step": 32691 }, { "epoch": 10.06, "learning_rate": 1.0396518978728792e-05, "loss": 0.0382, "step": 32692 }, { "epoch": 10.06, "learning_rate": 1.0396021100050984e-05, "loss": 0.0766, "step": 32693 }, { "epoch": 10.06, "learning_rate": 1.0395523220389962e-05, "loss": 0.0529, "step": 32694 }, { "epoch": 10.06, "learning_rate": 1.0395025339746965e-05, "loss": 0.0436, "step": 32695 }, { "epoch": 10.06, "learning_rate": 1.0394527458123225e-05, "loss": 0.0417, "step": 32696 }, { "epoch": 10.06, "learning_rate": 1.0394029575519977e-05, "loss": 0.0888, "step": 32697 }, { "epoch": 10.06, "learning_rate": 1.0393531691938465e-05, "loss": 0.0691, "step": 32698 }, { "epoch": 10.06, "learning_rate": 1.0393033807379917e-05, "loss": 0.0587, "step": 32699 }, { "epoch": 10.06, "learning_rate": 1.0392535921845572e-05, "loss": 0.0363, "step": 32700 }, { "epoch": 10.06, "learning_rate": 1.0392038035336667e-05, "loss": 0.0187, "step": 32701 }, { "epoch": 10.06, "learning_rate": 1.0391540147854437e-05, "loss": 0.0391, "step": 32702 }, { "epoch": 10.06, "learning_rate": 1.0391042259400119e-05, "loss": 0.025, "step": 32703 }, { "epoch": 10.06, "learning_rate": 1.0390544369974947e-05, "loss": 0.0511, "step": 32704 }, { "epoch": 10.06, "learning_rate": 1.0390046479580161e-05, "loss": 0.0194, "step": 32705 }, { "epoch": 10.06, "learning_rate": 1.0389548588216993e-05, "loss": 0.035, "step": 32706 }, { "epoch": 10.06, "learning_rate": 1.0389050695886677e-05, "loss": 0.0833, "step": 32707 }, { "epoch": 10.06, "learning_rate": 1.0388552802590461e-05, "loss": 0.0189, "step": 32708 }, { "epoch": 10.06, "learning_rate": 1.0388054908329568e-05, "loss": 0.0281, "step": 32709 }, { "epoch": 10.06, "learning_rate": 1.038755701310524e-05, "loss": 0.052, "step": 32710 }, { "epoch": 10.06, "learning_rate": 1.0387059116918714e-05, "loss": 0.0392, "step": 32711 }, { "epoch": 10.07, "learning_rate": 1.0386561219771222e-05, "loss": 0.0302, "step": 32712 }, { "epoch": 10.07, "learning_rate": 1.0386063321664007e-05, "loss": 0.0505, "step": 32713 }, { "epoch": 10.07, "learning_rate": 1.0385565422598294e-05, "loss": 0.0156, "step": 32714 }, { "epoch": 10.07, "learning_rate": 1.0385067522575331e-05, "loss": 0.0438, "step": 32715 }, { "epoch": 10.07, "learning_rate": 1.038456962159635e-05, "loss": 0.0733, "step": 32716 }, { "epoch": 10.07, "learning_rate": 1.0384071719662582e-05, "loss": 0.042, "step": 32717 }, { "epoch": 10.07, "learning_rate": 1.038357381677527e-05, "loss": 0.0767, "step": 32718 }, { "epoch": 10.07, "learning_rate": 1.0383075912935647e-05, "loss": 0.0321, "step": 32719 }, { "epoch": 10.07, "learning_rate": 1.0383075912935647e-05, "loss": 0.0541, "step": 32720 }, { "epoch": 10.07, "learning_rate": 1.0382578008144953e-05, "loss": 0.0823, "step": 32721 }, { "epoch": 10.07, "learning_rate": 1.0382080102404417e-05, "loss": 0.0866, "step": 32722 }, { "epoch": 10.07, "learning_rate": 1.038158219571528e-05, "loss": 0.0452, "step": 32723 }, { "epoch": 10.07, "learning_rate": 1.0381084288078778e-05, "loss": 0.0352, "step": 32724 }, { "epoch": 10.07, "learning_rate": 1.0380586379496144e-05, "loss": 0.0642, "step": 32725 }, { "epoch": 10.07, "learning_rate": 1.0380088469968619e-05, "loss": 0.0465, "step": 32726 }, { "epoch": 10.07, "learning_rate": 1.037959055949744e-05, "loss": 0.067, "step": 32727 }, { "epoch": 10.07, "learning_rate": 1.0379092648083832e-05, "loss": 0.0354, "step": 32728 }, { "epoch": 10.07, "learning_rate": 1.0378594735729046e-05, "loss": 0.0408, "step": 32729 }, { "epoch": 10.07, "learning_rate": 1.037809682243431e-05, "loss": 0.0338, "step": 32730 }, { "epoch": 10.07, "learning_rate": 1.0377598908200859e-05, "loss": 0.0516, "step": 32731 }, { "epoch": 10.07, "learning_rate": 1.0377100993029934e-05, "loss": 0.0734, "step": 32732 }, { "epoch": 10.07, "learning_rate": 1.0376603076922768e-05, "loss": 0.0472, "step": 32733 }, { "epoch": 10.07, "learning_rate": 1.0376105159880599e-05, "loss": 0.0423, "step": 32734 }, { "epoch": 10.07, "learning_rate": 1.037560724190466e-05, "loss": 0.0363, "step": 32735 }, { "epoch": 10.07, "learning_rate": 1.037510932299619e-05, "loss": 0.0422, "step": 32736 }, { "epoch": 10.07, "learning_rate": 1.037461140315643e-05, "loss": 0.0255, "step": 32737 }, { "epoch": 10.07, "learning_rate": 1.0374113482386604e-05, "loss": 0.0512, "step": 32738 }, { "epoch": 10.07, "learning_rate": 1.0373615560687957e-05, "loss": 0.0523, "step": 32739 }, { "epoch": 10.07, "learning_rate": 1.0373117638061725e-05, "loss": 0.0594, "step": 32740 }, { "epoch": 10.07, "learning_rate": 1.037261971450914e-05, "loss": 0.0398, "step": 32741 }, { "epoch": 10.07, "learning_rate": 1.0372121790031443e-05, "loss": 0.0449, "step": 32742 }, { "epoch": 10.07, "learning_rate": 1.0371623864629866e-05, "loss": 0.0531, "step": 32743 }, { "epoch": 10.08, "learning_rate": 1.0371125938305648e-05, "loss": 0.067, "step": 32744 }, { "epoch": 10.08, "learning_rate": 1.0370628011060024e-05, "loss": 0.0384, "step": 32745 }, { "epoch": 10.08, "learning_rate": 1.0370130082894232e-05, "loss": 0.0465, "step": 32746 }, { "epoch": 10.08, "learning_rate": 1.0369632153809505e-05, "loss": 0.0245, "step": 32747 }, { "epoch": 10.08, "learning_rate": 1.0369134223807082e-05, "loss": 0.0571, "step": 32748 }, { "epoch": 10.08, "learning_rate": 1.0368636292888199e-05, "loss": 0.0391, "step": 32749 }, { "epoch": 10.08, "learning_rate": 1.036813836105409e-05, "loss": 0.0283, "step": 32750 }, { "epoch": 10.08, "learning_rate": 1.0367640428305994e-05, "loss": 0.0528, "step": 32751 }, { "epoch": 10.08, "learning_rate": 1.0367142494645146e-05, "loss": 0.0478, "step": 32752 }, { "epoch": 10.08, "learning_rate": 1.0366644560072778e-05, "loss": 0.0374, "step": 32753 }, { "epoch": 10.08, "learning_rate": 1.0366146624590137e-05, "loss": 0.0716, "step": 32754 }, { "epoch": 10.08, "learning_rate": 1.0365648688198446e-05, "loss": 0.0665, "step": 32755 }, { "epoch": 10.08, "learning_rate": 1.0365150750898952e-05, "loss": 0.0394, "step": 32756 }, { "epoch": 10.08, "learning_rate": 1.0364652812692886e-05, "loss": 0.043, "step": 32757 }, { "epoch": 10.08, "learning_rate": 1.0364154873581486e-05, "loss": 0.0433, "step": 32758 }, { "epoch": 10.08, "learning_rate": 1.0363656933565989e-05, "loss": 0.0708, "step": 32759 }, { "epoch": 10.08, "learning_rate": 1.0363158992647626e-05, "loss": 0.0586, "step": 32760 }, { "epoch": 10.08, "learning_rate": 1.0362661050827643e-05, "loss": 0.0474, "step": 32761 }, { "epoch": 10.08, "learning_rate": 1.0362163108107267e-05, "loss": 0.0455, "step": 32762 }, { "epoch": 10.08, "learning_rate": 1.0361665164487732e-05, "loss": 0.0299, "step": 32763 }, { "epoch": 10.08, "learning_rate": 1.0361167219970287e-05, "loss": 0.0479, "step": 32764 }, { "epoch": 10.08, "learning_rate": 1.0360669274556161e-05, "loss": 0.0386, "step": 32765 }, { "epoch": 10.08, "learning_rate": 1.0360171328246589e-05, "loss": 0.0448, "step": 32766 }, { "epoch": 10.08, "learning_rate": 1.0359673381042807e-05, "loss": 0.0761, "step": 32767 }, { "epoch": 10.08, "learning_rate": 1.0359175432946057e-05, "loss": 0.0716, "step": 32768 }, { "epoch": 10.08, "learning_rate": 1.0358677483957567e-05, "loss": 0.0471, "step": 32769 }, { "epoch": 10.08, "learning_rate": 1.0358179534078579e-05, "loss": 0.0363, "step": 32770 }, { "epoch": 10.08, "learning_rate": 1.0357681583310329e-05, "loss": 0.0808, "step": 32771 }, { "epoch": 10.08, "learning_rate": 1.0357183631654051e-05, "loss": 0.0367, "step": 32772 }, { "epoch": 10.08, "learning_rate": 1.035668567911098e-05, "loss": 0.0371, "step": 32773 }, { "epoch": 10.08, "learning_rate": 1.0356187725682359e-05, "loss": 0.0238, "step": 32774 }, { "epoch": 10.08, "learning_rate": 1.0355689771369417e-05, "loss": 0.0361, "step": 32775 }, { "epoch": 10.08, "learning_rate": 1.0355191816173395e-05, "loss": 0.03, "step": 32776 }, { "epoch": 10.09, "learning_rate": 1.0354693860095525e-05, "loss": 0.0554, "step": 32777 }, { "epoch": 10.09, "learning_rate": 1.035419590313705e-05, "loss": 0.0602, "step": 32778 }, { "epoch": 10.09, "learning_rate": 1.0353697945299198e-05, "loss": 0.0857, "step": 32779 }, { "epoch": 10.09, "learning_rate": 1.035319998658321e-05, "loss": 0.0421, "step": 32780 }, { "epoch": 10.09, "learning_rate": 1.0352702026990322e-05, "loss": 0.0594, "step": 32781 }, { "epoch": 10.09, "learning_rate": 1.0352204066521773e-05, "loss": 0.0231, "step": 32782 }, { "epoch": 10.09, "learning_rate": 1.035170610517879e-05, "loss": 0.0431, "step": 32783 }, { "epoch": 10.09, "learning_rate": 1.0351208142962621e-05, "loss": 0.045, "step": 32784 }, { "epoch": 10.09, "learning_rate": 1.0350710179874497e-05, "loss": 0.0399, "step": 32785 }, { "epoch": 10.09, "learning_rate": 1.0350212215915651e-05, "loss": 0.0492, "step": 32786 }, { "epoch": 10.09, "learning_rate": 1.0349714251087323e-05, "loss": 0.0465, "step": 32787 }, { "epoch": 10.09, "learning_rate": 1.0349216285390751e-05, "loss": 0.0384, "step": 32788 }, { "epoch": 10.09, "learning_rate": 1.034871831882717e-05, "loss": 0.0215, "step": 32789 }, { "epoch": 10.09, "learning_rate": 1.0348220351397808e-05, "loss": 0.0467, "step": 32790 }, { "epoch": 10.09, "learning_rate": 1.0347722383103914e-05, "loss": 0.0628, "step": 32791 }, { "epoch": 10.09, "learning_rate": 1.0347224413946721e-05, "loss": 0.0391, "step": 32792 }, { "epoch": 10.09, "learning_rate": 1.034672644392746e-05, "loss": 0.0293, "step": 32793 }, { "epoch": 10.09, "learning_rate": 1.0346228473047373e-05, "loss": 0.1166, "step": 32794 }, { "epoch": 10.09, "learning_rate": 1.0345730501307694e-05, "loss": 0.0506, "step": 32795 }, { "epoch": 10.09, "learning_rate": 1.0345232528709656e-05, "loss": 0.0379, "step": 32796 }, { "epoch": 10.09, "learning_rate": 1.0344734555254502e-05, "loss": 0.0369, "step": 32797 }, { "epoch": 10.09, "learning_rate": 1.0344236580943463e-05, "loss": 0.0598, "step": 32798 }, { "epoch": 10.09, "learning_rate": 1.034373860577778e-05, "loss": 0.0476, "step": 32799 }, { "epoch": 10.09, "learning_rate": 1.0343240629758683e-05, "loss": 0.0448, "step": 32800 }, { "epoch": 10.09, "learning_rate": 1.0342742652887417e-05, "loss": 0.0747, "step": 32801 }, { "epoch": 10.09, "learning_rate": 1.034224467516521e-05, "loss": 0.0627, "step": 32802 }, { "epoch": 10.09, "learning_rate": 1.0341746696593303e-05, "loss": 0.0226, "step": 32803 }, { "epoch": 10.09, "learning_rate": 1.034124871717293e-05, "loss": 0.048, "step": 32804 }, { "epoch": 10.09, "learning_rate": 1.0340750736905328e-05, "loss": 0.0312, "step": 32805 }, { "epoch": 10.09, "learning_rate": 1.0340252755791736e-05, "loss": 0.0357, "step": 32806 }, { "epoch": 10.09, "learning_rate": 1.0339754773833386e-05, "loss": 0.037, "step": 32807 }, { "epoch": 10.09, "learning_rate": 1.0339256791031515e-05, "loss": 0.0268, "step": 32808 }, { "epoch": 10.1, "learning_rate": 1.0338758807387369e-05, "loss": 0.0434, "step": 32809 }, { "epoch": 10.1, "learning_rate": 1.0338260822902166e-05, "loss": 0.0451, "step": 32810 }, { "epoch": 10.1, "learning_rate": 1.0337762837577157e-05, "loss": 0.0482, "step": 32811 }, { "epoch": 10.1, "learning_rate": 1.0337264851413575e-05, "loss": 0.0201, "step": 32812 }, { "epoch": 10.1, "learning_rate": 1.0336766864412655e-05, "loss": 0.0293, "step": 32813 }, { "epoch": 10.1, "learning_rate": 1.0336268876575632e-05, "loss": 0.0609, "step": 32814 }, { "epoch": 10.1, "learning_rate": 1.0335770887903742e-05, "loss": 0.0355, "step": 32815 }, { "epoch": 10.1, "learning_rate": 1.033527289839823e-05, "loss": 0.0492, "step": 32816 }, { "epoch": 10.1, "learning_rate": 1.033477490806032e-05, "loss": 0.0451, "step": 32817 }, { "epoch": 10.1, "learning_rate": 1.0334276916891257e-05, "loss": 0.038, "step": 32818 }, { "epoch": 10.1, "learning_rate": 1.0333778924892272e-05, "loss": 0.0658, "step": 32819 }, { "epoch": 10.1, "learning_rate": 1.0333280932064606e-05, "loss": 0.0531, "step": 32820 }, { "epoch": 10.1, "learning_rate": 1.0332782938409494e-05, "loss": 0.0639, "step": 32821 }, { "epoch": 10.1, "learning_rate": 1.0332284943928169e-05, "loss": 0.0462, "step": 32822 }, { "epoch": 10.1, "learning_rate": 1.0331786948621874e-05, "loss": 0.0463, "step": 32823 }, { "epoch": 10.1, "learning_rate": 1.033128895249184e-05, "loss": 0.0636, "step": 32824 }, { "epoch": 10.1, "learning_rate": 1.0330790955539303e-05, "loss": 0.0585, "step": 32825 }, { "epoch": 10.1, "learning_rate": 1.0330292957765502e-05, "loss": 0.0392, "step": 32826 }, { "epoch": 10.1, "learning_rate": 1.0329794959171672e-05, "loss": 0.0574, "step": 32827 }, { "epoch": 10.1, "learning_rate": 1.0329296959759052e-05, "loss": 0.0357, "step": 32828 }, { "epoch": 10.1, "learning_rate": 1.0328798959528873e-05, "loss": 0.0794, "step": 32829 }, { "epoch": 10.1, "learning_rate": 1.032830095848238e-05, "loss": 0.0642, "step": 32830 }, { "epoch": 10.1, "learning_rate": 1.0327802956620801e-05, "loss": 0.0376, "step": 32831 }, { "epoch": 10.1, "learning_rate": 1.0327304953945374e-05, "loss": 0.1005, "step": 32832 }, { "epoch": 10.1, "learning_rate": 1.0326806950457343e-05, "loss": 0.0462, "step": 32833 }, { "epoch": 10.1, "learning_rate": 1.0326308946157933e-05, "loss": 0.0649, "step": 32834 }, { "epoch": 10.1, "learning_rate": 1.0325810941048386e-05, "loss": 0.0209, "step": 32835 }, { "epoch": 10.1, "learning_rate": 1.032531293512994e-05, "loss": 0.0566, "step": 32836 }, { "epoch": 10.1, "learning_rate": 1.0324814928403828e-05, "loss": 0.0351, "step": 32837 }, { "epoch": 10.1, "learning_rate": 1.0324316920871292e-05, "loss": 0.0612, "step": 32838 }, { "epoch": 10.1, "learning_rate": 1.0323818912533561e-05, "loss": 0.07, "step": 32839 }, { "epoch": 10.1, "learning_rate": 1.0323320903391876e-05, "loss": 0.0445, "step": 32840 }, { "epoch": 10.1, "learning_rate": 1.0322822893447474e-05, "loss": 0.0739, "step": 32841 }, { "epoch": 10.11, "learning_rate": 1.0322324882701584e-05, "loss": 0.0539, "step": 32842 }, { "epoch": 10.11, "learning_rate": 1.032182687115545e-05, "loss": 0.066, "step": 32843 }, { "epoch": 10.11, "learning_rate": 1.0321328858810314e-05, "loss": 0.0558, "step": 32844 }, { "epoch": 10.11, "learning_rate": 1.0320830845667398e-05, "loss": 0.0641, "step": 32845 }, { "epoch": 10.11, "learning_rate": 1.0320332831727949e-05, "loss": 0.0309, "step": 32846 }, { "epoch": 10.11, "learning_rate": 1.0319834816993197e-05, "loss": 0.0614, "step": 32847 }, { "epoch": 10.11, "learning_rate": 1.0319336801464384e-05, "loss": 0.031, "step": 32848 }, { "epoch": 10.11, "learning_rate": 1.0318838785142743e-05, "loss": 0.041, "step": 32849 }, { "epoch": 10.11, "learning_rate": 1.0318340768029509e-05, "loss": 0.0692, "step": 32850 }, { "epoch": 10.11, "learning_rate": 1.0317842750125923e-05, "loss": 0.0303, "step": 32851 }, { "epoch": 10.11, "learning_rate": 1.0317344731433217e-05, "loss": 0.0433, "step": 32852 }, { "epoch": 10.11, "learning_rate": 1.0316846711952632e-05, "loss": 0.0465, "step": 32853 }, { "epoch": 10.11, "learning_rate": 1.0316348691685403e-05, "loss": 0.037, "step": 32854 }, { "epoch": 10.11, "learning_rate": 1.0315850670632763e-05, "loss": 0.0455, "step": 32855 }, { "epoch": 10.11, "learning_rate": 1.0315352648795951e-05, "loss": 0.0976, "step": 32856 }, { "epoch": 10.11, "learning_rate": 1.0314854626176205e-05, "loss": 0.0512, "step": 32857 }, { "epoch": 10.11, "learning_rate": 1.0314356602774758e-05, "loss": 0.068, "step": 32858 }, { "epoch": 10.11, "learning_rate": 1.0313858578592851e-05, "loss": 0.0349, "step": 32859 }, { "epoch": 10.11, "learning_rate": 1.0313360553631713e-05, "loss": 0.0439, "step": 32860 }, { "epoch": 10.11, "learning_rate": 1.0312862527892592e-05, "loss": 0.0302, "step": 32861 }, { "epoch": 10.11, "learning_rate": 1.031236450137671e-05, "loss": 0.0348, "step": 32862 }, { "epoch": 10.11, "learning_rate": 1.0311866474085317e-05, "loss": 0.0354, "step": 32863 }, { "epoch": 10.11, "learning_rate": 1.031136844601964e-05, "loss": 0.0373, "step": 32864 }, { "epoch": 10.11, "learning_rate": 1.0310870417180922e-05, "loss": 0.0255, "step": 32865 }, { "epoch": 10.11, "learning_rate": 1.0310372387570395e-05, "loss": 0.0578, "step": 32866 }, { "epoch": 10.11, "learning_rate": 1.0309874357189298e-05, "loss": 0.0993, "step": 32867 }, { "epoch": 10.11, "learning_rate": 1.0309376326038868e-05, "loss": 0.0377, "step": 32868 }, { "epoch": 10.11, "learning_rate": 1.0308878294120336e-05, "loss": 0.042, "step": 32869 }, { "epoch": 10.11, "learning_rate": 1.0308380261434942e-05, "loss": 0.1105, "step": 32870 }, { "epoch": 10.11, "learning_rate": 1.0307882227983928e-05, "loss": 0.0469, "step": 32871 }, { "epoch": 10.11, "learning_rate": 1.0307384193768521e-05, "loss": 0.0387, "step": 32872 }, { "epoch": 10.11, "learning_rate": 1.0306886158789964e-05, "loss": 0.0426, "step": 32873 }, { "epoch": 10.12, "learning_rate": 1.0306388123049492e-05, "loss": 0.0389, "step": 32874 }, { "epoch": 10.12, "learning_rate": 1.0305890086548337e-05, "loss": 0.0428, "step": 32875 }, { "epoch": 10.12, "learning_rate": 1.0305392049287744e-05, "loss": 0.0175, "step": 32876 }, { "epoch": 10.12, "learning_rate": 1.0304894011268942e-05, "loss": 0.0401, "step": 32877 }, { "epoch": 10.12, "learning_rate": 1.0304395972493172e-05, "loss": 0.0472, "step": 32878 }, { "epoch": 10.12, "learning_rate": 1.030389793296167e-05, "loss": 0.0348, "step": 32879 }, { "epoch": 10.12, "learning_rate": 1.0303399892675668e-05, "loss": 0.0433, "step": 32880 }, { "epoch": 10.12, "learning_rate": 1.0302901851636409e-05, "loss": 0.1085, "step": 32881 }, { "epoch": 10.12, "learning_rate": 1.0302403809845126e-05, "loss": 0.0487, "step": 32882 }, { "epoch": 10.12, "learning_rate": 1.0301905767303053e-05, "loss": 0.032, "step": 32883 }, { "epoch": 10.12, "learning_rate": 1.0301407724011432e-05, "loss": 0.0761, "step": 32884 }, { "epoch": 10.12, "learning_rate": 1.0300909679971495e-05, "loss": 0.0459, "step": 32885 }, { "epoch": 10.12, "learning_rate": 1.0300411635184482e-05, "loss": 0.0711, "step": 32886 }, { "epoch": 10.12, "learning_rate": 1.0299913589651627e-05, "loss": 0.0593, "step": 32887 }, { "epoch": 10.12, "learning_rate": 1.029941554337417e-05, "loss": 0.0689, "step": 32888 }, { "epoch": 10.12, "learning_rate": 1.0298917496353342e-05, "loss": 0.0493, "step": 32889 }, { "epoch": 10.12, "learning_rate": 1.0298419448590382e-05, "loss": 0.0439, "step": 32890 }, { "epoch": 10.12, "learning_rate": 1.0297921400086528e-05, "loss": 0.0416, "step": 32891 }, { "epoch": 10.12, "learning_rate": 1.0297423350843017e-05, "loss": 0.0369, "step": 32892 }, { "epoch": 10.12, "learning_rate": 1.0296925300861083e-05, "loss": 0.0493, "step": 32893 }, { "epoch": 10.12, "learning_rate": 1.0296427250141962e-05, "loss": 0.03, "step": 32894 }, { "epoch": 10.12, "learning_rate": 1.0295929198686894e-05, "loss": 0.0509, "step": 32895 }, { "epoch": 10.12, "learning_rate": 1.0295431146497112e-05, "loss": 0.0558, "step": 32896 }, { "epoch": 10.12, "learning_rate": 1.0294933093573853e-05, "loss": 0.0698, "step": 32897 }, { "epoch": 10.12, "learning_rate": 1.0294435039918357e-05, "loss": 0.0781, "step": 32898 }, { "epoch": 10.12, "learning_rate": 1.029393698553186e-05, "loss": 0.055, "step": 32899 }, { "epoch": 10.12, "learning_rate": 1.0293438930415592e-05, "loss": 0.0565, "step": 32900 }, { "epoch": 10.12, "learning_rate": 1.0292940874570799e-05, "loss": 0.0469, "step": 32901 }, { "epoch": 10.12, "learning_rate": 1.0292442817998709e-05, "loss": 0.0318, "step": 32902 }, { "epoch": 10.12, "learning_rate": 1.0291944760700563e-05, "loss": 0.0521, "step": 32903 }, { "epoch": 10.12, "learning_rate": 1.0291446702677598e-05, "loss": 0.0611, "step": 32904 }, { "epoch": 10.12, "learning_rate": 1.029094864393105e-05, "loss": 0.0409, "step": 32905 }, { "epoch": 10.12, "learning_rate": 1.0290450584462155e-05, "loss": 0.0267, "step": 32906 }, { "epoch": 10.13, "learning_rate": 1.0289952524272147e-05, "loss": 0.0255, "step": 32907 }, { "epoch": 10.13, "learning_rate": 1.0289454463362265e-05, "loss": 0.0729, "step": 32908 }, { "epoch": 10.13, "learning_rate": 1.028895640173375e-05, "loss": 0.0387, "step": 32909 }, { "epoch": 10.13, "learning_rate": 1.028845833938783e-05, "loss": 0.0606, "step": 32910 }, { "epoch": 10.13, "learning_rate": 1.0287960276325745e-05, "loss": 0.0196, "step": 32911 }, { "epoch": 10.13, "learning_rate": 1.0287462212548736e-05, "loss": 0.0425, "step": 32912 }, { "epoch": 10.13, "learning_rate": 1.0286964148058034e-05, "loss": 0.0214, "step": 32913 }, { "epoch": 10.13, "learning_rate": 1.0286466082854878e-05, "loss": 0.0605, "step": 32914 }, { "epoch": 10.13, "learning_rate": 1.0285968016940502e-05, "loss": 0.0337, "step": 32915 }, { "epoch": 10.13, "learning_rate": 1.0285469950316147e-05, "loss": 0.0196, "step": 32916 }, { "epoch": 10.13, "learning_rate": 1.0284971882983043e-05, "loss": 0.0693, "step": 32917 }, { "epoch": 10.13, "learning_rate": 1.0284473814942436e-05, "loss": 0.0326, "step": 32918 }, { "epoch": 10.13, "learning_rate": 1.0283975746195554e-05, "loss": 0.0564, "step": 32919 }, { "epoch": 10.13, "learning_rate": 1.0283477676743639e-05, "loss": 0.0239, "step": 32920 }, { "epoch": 10.13, "learning_rate": 1.0282979606587924e-05, "loss": 0.0412, "step": 32921 }, { "epoch": 10.13, "learning_rate": 1.0282481535729645e-05, "loss": 0.0765, "step": 32922 }, { "epoch": 10.13, "learning_rate": 1.0281983464170045e-05, "loss": 0.0519, "step": 32923 }, { "epoch": 10.13, "learning_rate": 1.0281485391910353e-05, "loss": 0.0457, "step": 32924 }, { "epoch": 10.13, "learning_rate": 1.0280987318951808e-05, "loss": 0.0334, "step": 32925 }, { "epoch": 10.13, "learning_rate": 1.028048924529565e-05, "loss": 0.0538, "step": 32926 }, { "epoch": 10.13, "learning_rate": 1.0279991170943109e-05, "loss": 0.0949, "step": 32927 }, { "epoch": 10.13, "learning_rate": 1.0279493095895427e-05, "loss": 0.0584, "step": 32928 }, { "epoch": 10.13, "learning_rate": 1.0278995020153842e-05, "loss": 0.0532, "step": 32929 }, { "epoch": 10.13, "learning_rate": 1.0278496943719585e-05, "loss": 0.0443, "step": 32930 }, { "epoch": 10.13, "learning_rate": 1.0277998866593894e-05, "loss": 0.0291, "step": 32931 }, { "epoch": 10.13, "learning_rate": 1.0277500788778008e-05, "loss": 0.0836, "step": 32932 }, { "epoch": 10.13, "learning_rate": 1.0277002710273163e-05, "loss": 0.0498, "step": 32933 }, { "epoch": 10.13, "learning_rate": 1.0276504631080595e-05, "loss": 0.0561, "step": 32934 }, { "epoch": 10.13, "learning_rate": 1.0276006551201538e-05, "loss": 0.0686, "step": 32935 }, { "epoch": 10.13, "learning_rate": 1.0275508470637234e-05, "loss": 0.0788, "step": 32936 }, { "epoch": 10.13, "learning_rate": 1.0275010389388915e-05, "loss": 0.0335, "step": 32937 }, { "epoch": 10.13, "learning_rate": 1.0274512307457822e-05, "loss": 0.0489, "step": 32938 }, { "epoch": 10.14, "learning_rate": 1.0274014224845186e-05, "loss": 0.0509, "step": 32939 }, { "epoch": 10.14, "learning_rate": 1.027351614155225e-05, "loss": 0.0177, "step": 32940 }, { "epoch": 10.14, "learning_rate": 1.0273018057580243e-05, "loss": 0.0516, "step": 32941 }, { "epoch": 10.14, "learning_rate": 1.0272519972930406e-05, "loss": 0.0419, "step": 32942 }, { "epoch": 10.14, "learning_rate": 1.0272021887603977e-05, "loss": 0.0502, "step": 32943 }, { "epoch": 10.14, "learning_rate": 1.0271523801602194e-05, "loss": 0.037, "step": 32944 }, { "epoch": 10.14, "learning_rate": 1.0271025714926286e-05, "loss": 0.0724, "step": 32945 }, { "epoch": 10.14, "learning_rate": 1.0270527627577495e-05, "loss": 0.0367, "step": 32946 }, { "epoch": 10.14, "learning_rate": 1.0270029539557058e-05, "loss": 0.0486, "step": 32947 }, { "epoch": 10.14, "learning_rate": 1.026953145086621e-05, "loss": 0.0604, "step": 32948 }, { "epoch": 10.14, "learning_rate": 1.0269033361506189e-05, "loss": 0.0766, "step": 32949 }, { "epoch": 10.14, "learning_rate": 1.0268535271478231e-05, "loss": 0.0643, "step": 32950 }, { "epoch": 10.14, "learning_rate": 1.026803718078357e-05, "loss": 0.0444, "step": 32951 }, { "epoch": 10.14, "learning_rate": 1.0267539089423445e-05, "loss": 0.0085, "step": 32952 }, { "epoch": 10.14, "learning_rate": 1.0267040997399094e-05, "loss": 0.0614, "step": 32953 }, { "epoch": 10.14, "learning_rate": 1.0266542904711753e-05, "loss": 0.0253, "step": 32954 }, { "epoch": 10.14, "learning_rate": 1.0266044811362655e-05, "loss": 0.036, "step": 32955 }, { "epoch": 10.14, "learning_rate": 1.0265546717353041e-05, "loss": 0.0395, "step": 32956 }, { "epoch": 10.14, "learning_rate": 1.0265048622684146e-05, "loss": 0.0657, "step": 32957 }, { "epoch": 10.14, "learning_rate": 1.0264550527357209e-05, "loss": 0.0735, "step": 32958 }, { "epoch": 10.14, "learning_rate": 1.026405243137346e-05, "loss": 0.0155, "step": 32959 }, { "epoch": 10.14, "learning_rate": 1.0263554334734143e-05, "loss": 0.0624, "step": 32960 }, { "epoch": 10.14, "learning_rate": 1.0263056237440492e-05, "loss": 0.0479, "step": 32961 }, { "epoch": 10.14, "learning_rate": 1.0262558139493742e-05, "loss": 0.0701, "step": 32962 }, { "epoch": 10.14, "learning_rate": 1.026206004089513e-05, "loss": 0.0547, "step": 32963 }, { "epoch": 10.14, "learning_rate": 1.0261561941645894e-05, "loss": 0.0269, "step": 32964 }, { "epoch": 10.14, "learning_rate": 1.0261063841747271e-05, "loss": 0.0782, "step": 32965 }, { "epoch": 10.14, "learning_rate": 1.02605657412005e-05, "loss": 0.019, "step": 32966 }, { "epoch": 10.14, "learning_rate": 1.026006764000681e-05, "loss": 0.0405, "step": 32967 }, { "epoch": 10.14, "learning_rate": 1.0259569538167444e-05, "loss": 0.03, "step": 32968 }, { "epoch": 10.14, "learning_rate": 1.0259071435683636e-05, "loss": 0.0528, "step": 32969 }, { "epoch": 10.14, "learning_rate": 1.025857333255662e-05, "loss": 0.0733, "step": 32970 }, { "epoch": 10.14, "learning_rate": 1.0258075228787645e-05, "loss": 0.0304, "step": 32971 }, { "epoch": 10.15, "learning_rate": 1.025757712437793e-05, "loss": 0.0578, "step": 32972 }, { "epoch": 10.15, "learning_rate": 1.0257079019328724e-05, "loss": 0.0341, "step": 32973 }, { "epoch": 10.15, "learning_rate": 1.0256580913641261e-05, "loss": 0.0299, "step": 32974 }, { "epoch": 10.15, "learning_rate": 1.0256082807316776e-05, "loss": 0.0588, "step": 32975 }, { "epoch": 10.15, "learning_rate": 1.0255584700356507e-05, "loss": 0.0641, "step": 32976 }, { "epoch": 10.15, "learning_rate": 1.0255086592761688e-05, "loss": 0.1113, "step": 32977 }, { "epoch": 10.15, "learning_rate": 1.0254588484533561e-05, "loss": 0.0243, "step": 32978 }, { "epoch": 10.15, "learning_rate": 1.0254090375673357e-05, "loss": 0.0389, "step": 32979 }, { "epoch": 10.15, "learning_rate": 1.0253592266182315e-05, "loss": 0.0633, "step": 32980 }, { "epoch": 10.15, "learning_rate": 1.0253094156061674e-05, "loss": 0.0462, "step": 32981 }, { "epoch": 10.15, "learning_rate": 1.0252596045312666e-05, "loss": 0.067, "step": 32982 }, { "epoch": 10.15, "learning_rate": 1.0252097933936532e-05, "loss": 0.0672, "step": 32983 }, { "epoch": 10.15, "learning_rate": 1.0251599821934506e-05, "loss": 0.0504, "step": 32984 }, { "epoch": 10.15, "learning_rate": 1.0251101709307825e-05, "loss": 0.0699, "step": 32985 }, { "epoch": 10.15, "learning_rate": 1.0250603596057727e-05, "loss": 0.0498, "step": 32986 }, { "epoch": 10.15, "learning_rate": 1.0250105482185444e-05, "loss": 0.052, "step": 32987 }, { "epoch": 10.15, "learning_rate": 1.0249607367692222e-05, "loss": 0.0821, "step": 32988 }, { "epoch": 10.15, "learning_rate": 1.0249109252579292e-05, "loss": 0.0617, "step": 32989 }, { "epoch": 10.15, "learning_rate": 1.0248611136847885e-05, "loss": 0.0423, "step": 32990 }, { "epoch": 10.15, "learning_rate": 1.024811302049925e-05, "loss": 0.1002, "step": 32991 }, { "epoch": 10.15, "learning_rate": 1.0247614903534615e-05, "loss": 0.0307, "step": 32992 }, { "epoch": 10.15, "learning_rate": 1.0247116785955218e-05, "loss": 0.0573, "step": 32993 }, { "epoch": 10.15, "learning_rate": 1.0246618667762299e-05, "loss": 0.028, "step": 32994 }, { "epoch": 10.15, "learning_rate": 1.0246120548957091e-05, "loss": 0.0192, "step": 32995 }, { "epoch": 10.15, "learning_rate": 1.0245622429540832e-05, "loss": 0.0779, "step": 32996 }, { "epoch": 10.15, "learning_rate": 1.0245124309514755e-05, "loss": 0.0481, "step": 32997 }, { "epoch": 10.15, "learning_rate": 1.0244626188880105e-05, "loss": 0.0412, "step": 32998 }, { "epoch": 10.15, "learning_rate": 1.0244128067638112e-05, "loss": 0.0527, "step": 32999 }, { "epoch": 10.15, "learning_rate": 1.0243629945790017e-05, "loss": 0.0352, "step": 33000 }, { "epoch": 10.15, "learning_rate": 1.0243131823337053e-05, "loss": 0.0416, "step": 33001 }, { "epoch": 10.15, "learning_rate": 1.024263370028046e-05, "loss": 0.0529, "step": 33002 }, { "epoch": 10.15, "learning_rate": 1.024213557662147e-05, "loss": 0.0421, "step": 33003 }, { "epoch": 10.16, "learning_rate": 1.0241637452361323e-05, "loss": 0.0267, "step": 33004 }, { "epoch": 10.16, "learning_rate": 1.024113932750126e-05, "loss": 0.07, "step": 33005 }, { "epoch": 10.16, "learning_rate": 1.024064120204251e-05, "loss": 0.0734, "step": 33006 }, { "epoch": 10.16, "learning_rate": 1.0240143075986309e-05, "loss": 0.0479, "step": 33007 }, { "epoch": 10.16, "learning_rate": 1.02396449493339e-05, "loss": 0.0552, "step": 33008 }, { "epoch": 10.16, "learning_rate": 1.0239146822086518e-05, "loss": 0.0462, "step": 33009 }, { "epoch": 10.16, "learning_rate": 1.0238648694245402e-05, "loss": 0.0352, "step": 33010 }, { "epoch": 10.16, "learning_rate": 1.0238150565811783e-05, "loss": 0.0487, "step": 33011 }, { "epoch": 10.16, "learning_rate": 1.0237652436786897e-05, "loss": 0.0521, "step": 33012 }, { "epoch": 10.16, "learning_rate": 1.0237154307171989e-05, "loss": 0.0658, "step": 33013 }, { "epoch": 10.16, "learning_rate": 1.0236656176968287e-05, "loss": 0.0416, "step": 33014 }, { "epoch": 10.16, "learning_rate": 1.0236158046177032e-05, "loss": 0.0431, "step": 33015 }, { "epoch": 10.16, "learning_rate": 1.0235659914799465e-05, "loss": 0.0519, "step": 33016 }, { "epoch": 10.16, "learning_rate": 1.0235161782836811e-05, "loss": 0.0334, "step": 33017 }, { "epoch": 10.16, "learning_rate": 1.0234663650290317e-05, "loss": 0.0517, "step": 33018 }, { "epoch": 10.16, "learning_rate": 1.0234165517161218e-05, "loss": 0.0673, "step": 33019 }, { "epoch": 10.16, "learning_rate": 1.0233667383450748e-05, "loss": 0.0661, "step": 33020 }, { "epoch": 10.16, "learning_rate": 1.0233169249160145e-05, "loss": 0.0449, "step": 33021 }, { "epoch": 10.16, "learning_rate": 1.0232671114290644e-05, "loss": 0.0637, "step": 33022 }, { "epoch": 10.16, "learning_rate": 1.0232172978843486e-05, "loss": 0.0414, "step": 33023 }, { "epoch": 10.16, "learning_rate": 1.0231674842819901e-05, "loss": 0.029, "step": 33024 }, { "epoch": 10.16, "learning_rate": 1.0231176706221134e-05, "loss": 0.0654, "step": 33025 }, { "epoch": 10.16, "learning_rate": 1.0230678569048418e-05, "loss": 0.0495, "step": 33026 }, { "epoch": 10.16, "learning_rate": 1.0230180431302984e-05, "loss": 0.0614, "step": 33027 }, { "epoch": 10.16, "learning_rate": 1.0229682292986078e-05, "loss": 0.1051, "step": 33028 }, { "epoch": 10.16, "learning_rate": 1.0229184154098935e-05, "loss": 0.057, "step": 33029 }, { "epoch": 10.16, "learning_rate": 1.0228686014642786e-05, "loss": 0.0684, "step": 33030 }, { "epoch": 10.16, "learning_rate": 1.0228187874618872e-05, "loss": 0.0505, "step": 33031 }, { "epoch": 10.16, "learning_rate": 1.0227689734028427e-05, "loss": 0.0504, "step": 33032 }, { "epoch": 10.16, "learning_rate": 1.0227191592872695e-05, "loss": 0.022, "step": 33033 }, { "epoch": 10.16, "learning_rate": 1.02266934511529e-05, "loss": 0.0358, "step": 33034 }, { "epoch": 10.16, "learning_rate": 1.0226195308870293e-05, "loss": 0.0674, "step": 33035 }, { "epoch": 10.16, "learning_rate": 1.02256971660261e-05, "loss": 0.0723, "step": 33036 }, { "epoch": 10.17, "learning_rate": 1.0225199022621564e-05, "loss": 0.0506, "step": 33037 }, { "epoch": 10.17, "learning_rate": 1.022470087865792e-05, "loss": 0.0487, "step": 33038 }, { "epoch": 10.17, "learning_rate": 1.0224202734136401e-05, "loss": 0.0243, "step": 33039 }, { "epoch": 10.17, "learning_rate": 1.0223704589058253e-05, "loss": 0.0641, "step": 33040 }, { "epoch": 10.17, "learning_rate": 1.0223206443424703e-05, "loss": 0.0828, "step": 33041 }, { "epoch": 10.17, "learning_rate": 1.022270829723699e-05, "loss": 0.0333, "step": 33042 }, { "epoch": 10.17, "learning_rate": 1.0222210150496353e-05, "loss": 0.0368, "step": 33043 }, { "epoch": 10.17, "learning_rate": 1.022171200320403e-05, "loss": 0.0409, "step": 33044 }, { "epoch": 10.17, "learning_rate": 1.0221213855361253e-05, "loss": 0.0432, "step": 33045 }, { "epoch": 10.17, "learning_rate": 1.0220715706969264e-05, "loss": 0.0245, "step": 33046 }, { "epoch": 10.17, "learning_rate": 1.0220217558029298e-05, "loss": 0.0498, "step": 33047 }, { "epoch": 10.17, "learning_rate": 1.021971940854259e-05, "loss": 0.0574, "step": 33048 }, { "epoch": 10.17, "learning_rate": 1.0219221258510377e-05, "loss": 0.0687, "step": 33049 }, { "epoch": 10.17, "learning_rate": 1.02187231079339e-05, "loss": 0.0765, "step": 33050 }, { "epoch": 10.17, "learning_rate": 1.0218224956814391e-05, "loss": 0.0355, "step": 33051 }, { "epoch": 10.17, "learning_rate": 1.0217726805153085e-05, "loss": 0.0573, "step": 33052 }, { "epoch": 10.17, "learning_rate": 1.0217228652951225e-05, "loss": 0.0501, "step": 33053 }, { "epoch": 10.17, "learning_rate": 1.0216730500210046e-05, "loss": 0.0511, "step": 33054 }, { "epoch": 10.17, "learning_rate": 1.0216232346930782e-05, "loss": 0.0362, "step": 33055 }, { "epoch": 10.17, "learning_rate": 1.021573419311467e-05, "loss": 0.0343, "step": 33056 }, { "epoch": 10.17, "learning_rate": 1.021523603876295e-05, "loss": 0.06, "step": 33057 }, { "epoch": 10.17, "learning_rate": 1.0214737883876858e-05, "loss": 0.0894, "step": 33058 }, { "epoch": 10.17, "learning_rate": 1.0214239728457626e-05, "loss": 0.0377, "step": 33059 }, { "epoch": 10.17, "learning_rate": 1.0213741572506497e-05, "loss": 0.0727, "step": 33060 }, { "epoch": 10.17, "learning_rate": 1.0213243416024708e-05, "loss": 0.0554, "step": 33061 }, { "epoch": 10.17, "learning_rate": 1.0212745259013488e-05, "loss": 0.016, "step": 33062 }, { "epoch": 10.17, "learning_rate": 1.021224710147408e-05, "loss": 0.091, "step": 33063 }, { "epoch": 10.17, "learning_rate": 1.0211748943407723e-05, "loss": 0.0533, "step": 33064 }, { "epoch": 10.17, "learning_rate": 1.0211250784815648e-05, "loss": 0.0568, "step": 33065 }, { "epoch": 10.17, "learning_rate": 1.0210752625699094e-05, "loss": 0.0322, "step": 33066 }, { "epoch": 10.17, "learning_rate": 1.0210254466059297e-05, "loss": 0.0315, "step": 33067 }, { "epoch": 10.17, "learning_rate": 1.0209756305897499e-05, "loss": 0.0727, "step": 33068 }, { "epoch": 10.18, "learning_rate": 1.0209258145214926e-05, "loss": 0.0364, "step": 33069 }, { "epoch": 10.18, "learning_rate": 1.0208759984012826e-05, "loss": 0.0466, "step": 33070 }, { "epoch": 10.18, "learning_rate": 1.0208261822292433e-05, "loss": 0.0563, "step": 33071 }, { "epoch": 10.18, "learning_rate": 1.0207763660054979e-05, "loss": 0.0216, "step": 33072 }, { "epoch": 10.18, "learning_rate": 1.0207265497301704e-05, "loss": 0.0609, "step": 33073 }, { "epoch": 10.18, "learning_rate": 1.0206767334033845e-05, "loss": 0.0505, "step": 33074 }, { "epoch": 10.18, "learning_rate": 1.020626917025264e-05, "loss": 0.0538, "step": 33075 }, { "epoch": 10.18, "learning_rate": 1.0205771005959321e-05, "loss": 0.0232, "step": 33076 }, { "epoch": 10.18, "learning_rate": 1.0205272841155129e-05, "loss": 0.0721, "step": 33077 }, { "epoch": 10.18, "learning_rate": 1.0204774675841302e-05, "loss": 0.0533, "step": 33078 }, { "epoch": 10.18, "learning_rate": 1.0204276510019074e-05, "loss": 0.0718, "step": 33079 }, { "epoch": 10.18, "learning_rate": 1.0203778343689682e-05, "loss": 0.0419, "step": 33080 }, { "epoch": 10.18, "learning_rate": 1.0203280176854362e-05, "loss": 0.0261, "step": 33081 }, { "epoch": 10.18, "learning_rate": 1.0202782009514354e-05, "loss": 0.058, "step": 33082 }, { "epoch": 10.18, "learning_rate": 1.0202283841670892e-05, "loss": 0.0643, "step": 33083 }, { "epoch": 10.18, "learning_rate": 1.0201785673325214e-05, "loss": 0.0593, "step": 33084 }, { "epoch": 10.18, "learning_rate": 1.0201287504478559e-05, "loss": 0.0333, "step": 33085 }, { "epoch": 10.18, "learning_rate": 1.0200789335132157e-05, "loss": 0.0481, "step": 33086 }, { "epoch": 10.18, "learning_rate": 1.0200291165287251e-05, "loss": 0.0579, "step": 33087 }, { "epoch": 10.18, "learning_rate": 1.019979299494508e-05, "loss": 0.0524, "step": 33088 }, { "epoch": 10.18, "learning_rate": 1.0199294824106868e-05, "loss": 0.0214, "step": 33089 }, { "epoch": 10.18, "learning_rate": 1.0198796652773868e-05, "loss": 0.039, "step": 33090 }, { "epoch": 10.18, "learning_rate": 1.0198298480947306e-05, "loss": 0.06, "step": 33091 }, { "epoch": 10.18, "learning_rate": 1.0197800308628425e-05, "loss": 0.0231, "step": 33092 }, { "epoch": 10.18, "learning_rate": 1.0197302135818459e-05, "loss": 0.0593, "step": 33093 }, { "epoch": 10.18, "learning_rate": 1.0196803962518643e-05, "loss": 0.0385, "step": 33094 }, { "epoch": 10.18, "learning_rate": 1.0196305788730216e-05, "loss": 0.043, "step": 33095 }, { "epoch": 10.18, "learning_rate": 1.0195807614454416e-05, "loss": 0.0472, "step": 33096 }, { "epoch": 10.18, "learning_rate": 1.0195309439692475e-05, "loss": 0.0381, "step": 33097 }, { "epoch": 10.18, "learning_rate": 1.0194811264445639e-05, "loss": 0.0555, "step": 33098 }, { "epoch": 10.18, "learning_rate": 1.0194313088715135e-05, "loss": 0.0617, "step": 33099 }, { "epoch": 10.18, "learning_rate": 1.0193814912502206e-05, "loss": 0.0362, "step": 33100 }, { "epoch": 10.18, "learning_rate": 1.0193316735808085e-05, "loss": 0.0284, "step": 33101 }, { "epoch": 10.19, "learning_rate": 1.0192818558634012e-05, "loss": 0.0261, "step": 33102 }, { "epoch": 10.19, "learning_rate": 1.0192320380981224e-05, "loss": 0.0354, "step": 33103 }, { "epoch": 10.19, "learning_rate": 1.0191822202850952e-05, "loss": 0.0551, "step": 33104 }, { "epoch": 10.19, "learning_rate": 1.0191324024244442e-05, "loss": 0.0397, "step": 33105 }, { "epoch": 10.19, "learning_rate": 1.0190825845162925e-05, "loss": 0.0364, "step": 33106 }, { "epoch": 10.19, "learning_rate": 1.0190327665607635e-05, "loss": 0.0523, "step": 33107 }, { "epoch": 10.19, "learning_rate": 1.0189829485579816e-05, "loss": 0.027, "step": 33108 }, { "epoch": 10.19, "learning_rate": 1.01893313050807e-05, "loss": 0.0529, "step": 33109 }, { "epoch": 10.19, "learning_rate": 1.0188833124111528e-05, "loss": 0.0614, "step": 33110 }, { "epoch": 10.19, "learning_rate": 1.0188334942673533e-05, "loss": 0.0367, "step": 33111 }, { "epoch": 10.19, "learning_rate": 1.0187836760767954e-05, "loss": 0.0415, "step": 33112 }, { "epoch": 10.19, "learning_rate": 1.0187338578396025e-05, "loss": 0.043, "step": 33113 }, { "epoch": 10.19, "learning_rate": 1.0186840395558984e-05, "loss": 0.0695, "step": 33114 }, { "epoch": 10.19, "learning_rate": 1.0186342212258069e-05, "loss": 0.0411, "step": 33115 }, { "epoch": 10.19, "learning_rate": 1.018584402849452e-05, "loss": 0.0526, "step": 33116 }, { "epoch": 10.19, "learning_rate": 1.0185345844269565e-05, "loss": 0.0348, "step": 33117 }, { "epoch": 10.19, "learning_rate": 1.018484765958445e-05, "loss": 0.0701, "step": 33118 }, { "epoch": 10.19, "learning_rate": 1.0184349474440409e-05, "loss": 0.0452, "step": 33119 }, { "epoch": 10.19, "learning_rate": 1.0183851288838676e-05, "loss": 0.0582, "step": 33120 }, { "epoch": 10.19, "learning_rate": 1.0183353102780492e-05, "loss": 0.0413, "step": 33121 }, { "epoch": 10.19, "learning_rate": 1.0182854916267087e-05, "loss": 0.033, "step": 33122 }, { "epoch": 10.19, "learning_rate": 1.0182356729299707e-05, "loss": 0.0601, "step": 33123 }, { "epoch": 10.19, "learning_rate": 1.0181858541879582e-05, "loss": 0.0418, "step": 33124 }, { "epoch": 10.19, "learning_rate": 1.0181360354007952e-05, "loss": 0.0656, "step": 33125 }, { "epoch": 10.19, "learning_rate": 1.0180862165686054e-05, "loss": 0.0759, "step": 33126 }, { "epoch": 10.19, "learning_rate": 1.0180363976915121e-05, "loss": 0.0309, "step": 33127 }, { "epoch": 10.19, "learning_rate": 1.0179865787696397e-05, "loss": 0.0772, "step": 33128 }, { "epoch": 10.19, "learning_rate": 1.0179367598031111e-05, "loss": 0.0378, "step": 33129 }, { "epoch": 10.19, "learning_rate": 1.0178869407920507e-05, "loss": 0.0602, "step": 33130 }, { "epoch": 10.19, "learning_rate": 1.0178371217365817e-05, "loss": 0.0714, "step": 33131 }, { "epoch": 10.19, "learning_rate": 1.0177873026368277e-05, "loss": 0.0544, "step": 33132 }, { "epoch": 10.19, "learning_rate": 1.017737483492913e-05, "loss": 0.0393, "step": 33133 }, { "epoch": 10.2, "learning_rate": 1.0176876643049607e-05, "loss": 0.0648, "step": 33134 }, { "epoch": 10.2, "learning_rate": 1.0176378450730949e-05, "loss": 0.0567, "step": 33135 }, { "epoch": 10.2, "learning_rate": 1.017588025797439e-05, "loss": 0.0306, "step": 33136 }, { "epoch": 10.2, "learning_rate": 1.0175382064781164e-05, "loss": 0.0526, "step": 33137 }, { "epoch": 10.2, "learning_rate": 1.0174883871152517e-05, "loss": 0.0582, "step": 33138 }, { "epoch": 10.2, "learning_rate": 1.0174385677089675e-05, "loss": 0.0522, "step": 33139 }, { "epoch": 10.2, "learning_rate": 1.0173887482593886e-05, "loss": 0.045, "step": 33140 }, { "epoch": 10.2, "learning_rate": 1.017338928766638e-05, "loss": 0.067, "step": 33141 }, { "epoch": 10.2, "learning_rate": 1.0172891092308393e-05, "loss": 0.0544, "step": 33142 }, { "epoch": 10.2, "learning_rate": 1.0172392896521163e-05, "loss": 0.076, "step": 33143 }, { "epoch": 10.2, "learning_rate": 1.0171894700305931e-05, "loss": 0.0493, "step": 33144 }, { "epoch": 10.2, "learning_rate": 1.017139650366393e-05, "loss": 0.061, "step": 33145 }, { "epoch": 10.2, "learning_rate": 1.0170898306596397e-05, "loss": 0.0578, "step": 33146 }, { "epoch": 10.2, "learning_rate": 1.0170400109104573e-05, "loss": 0.0441, "step": 33147 }, { "epoch": 10.2, "learning_rate": 1.0169901911189688e-05, "loss": 0.0604, "step": 33148 }, { "epoch": 10.2, "learning_rate": 1.016940371285298e-05, "loss": 0.0566, "step": 33149 }, { "epoch": 10.2, "learning_rate": 1.0168905514095692e-05, "loss": 0.0589, "step": 33150 }, { "epoch": 10.2, "learning_rate": 1.0168407314919057e-05, "loss": 0.0525, "step": 33151 }, { "epoch": 10.2, "learning_rate": 1.0167909115324311e-05, "loss": 0.049, "step": 33152 }, { "epoch": 10.2, "learning_rate": 1.0167410915312693e-05, "loss": 0.0264, "step": 33153 }, { "epoch": 10.2, "learning_rate": 1.0166912714885437e-05, "loss": 0.0541, "step": 33154 }, { "epoch": 10.2, "learning_rate": 1.0166414514043784e-05, "loss": 0.0763, "step": 33155 }, { "epoch": 10.2, "learning_rate": 1.0165916312788967e-05, "loss": 0.0597, "step": 33156 }, { "epoch": 10.2, "learning_rate": 1.0165418111122225e-05, "loss": 0.0656, "step": 33157 }, { "epoch": 10.2, "learning_rate": 1.0164919909044795e-05, "loss": 0.0308, "step": 33158 }, { "epoch": 10.2, "learning_rate": 1.016442170655791e-05, "loss": 0.0582, "step": 33159 }, { "epoch": 10.2, "learning_rate": 1.0163923503662813e-05, "loss": 0.0612, "step": 33160 }, { "epoch": 10.2, "learning_rate": 1.0163425300360741e-05, "loss": 0.0859, "step": 33161 }, { "epoch": 10.2, "learning_rate": 1.0162927096652923e-05, "loss": 0.0409, "step": 33162 }, { "epoch": 10.2, "learning_rate": 1.0162428892540603e-05, "loss": 0.0495, "step": 33163 }, { "epoch": 10.2, "learning_rate": 1.0161930688025018e-05, "loss": 0.0645, "step": 33164 }, { "epoch": 10.2, "learning_rate": 1.0161432483107398e-05, "loss": 0.06, "step": 33165 }, { "epoch": 10.2, "learning_rate": 1.016093427778899e-05, "loss": 0.0794, "step": 33166 }, { "epoch": 10.21, "learning_rate": 1.0160436072071023e-05, "loss": 0.0589, "step": 33167 }, { "epoch": 10.21, "learning_rate": 1.0159937865954737e-05, "loss": 0.0477, "step": 33168 }, { "epoch": 10.21, "learning_rate": 1.0159439659441363e-05, "loss": 0.0459, "step": 33169 }, { "epoch": 10.21, "learning_rate": 1.0158941452532149e-05, "loss": 0.0457, "step": 33170 }, { "epoch": 10.21, "learning_rate": 1.0158443245228329e-05, "loss": 0.0632, "step": 33171 }, { "epoch": 10.21, "learning_rate": 1.0157945037531132e-05, "loss": 0.0381, "step": 33172 }, { "epoch": 10.21, "learning_rate": 1.0157446829441802e-05, "loss": 0.0434, "step": 33173 }, { "epoch": 10.21, "learning_rate": 1.0156948620961574e-05, "loss": 0.0594, "step": 33174 }, { "epoch": 10.21, "learning_rate": 1.0156450412091683e-05, "loss": 0.0717, "step": 33175 }, { "epoch": 10.21, "learning_rate": 1.015595220283337e-05, "loss": 0.0738, "step": 33176 }, { "epoch": 10.21, "learning_rate": 1.0155453993187867e-05, "loss": 0.089, "step": 33177 }, { "epoch": 10.21, "learning_rate": 1.0154955783156418e-05, "loss": 0.0491, "step": 33178 }, { "epoch": 10.21, "learning_rate": 1.015445757274025e-05, "loss": 0.0375, "step": 33179 }, { "epoch": 10.21, "learning_rate": 1.0153959361940608e-05, "loss": 0.0438, "step": 33180 }, { "epoch": 10.21, "learning_rate": 1.0153461150758727e-05, "loss": 0.0377, "step": 33181 }, { "epoch": 10.21, "learning_rate": 1.0152962939195845e-05, "loss": 0.0466, "step": 33182 }, { "epoch": 10.21, "learning_rate": 1.0152464727253195e-05, "loss": 0.0318, "step": 33183 }, { "epoch": 10.21, "learning_rate": 1.0151966514932017e-05, "loss": 0.0284, "step": 33184 }, { "epoch": 10.21, "learning_rate": 1.0151468302233546e-05, "loss": 0.0333, "step": 33185 }, { "epoch": 10.21, "learning_rate": 1.0150970089159023e-05, "loss": 0.0811, "step": 33186 }, { "epoch": 10.21, "learning_rate": 1.0150471875709677e-05, "loss": 0.0236, "step": 33187 }, { "epoch": 10.21, "learning_rate": 1.0149973661886757e-05, "loss": 0.0572, "step": 33188 }, { "epoch": 10.21, "learning_rate": 1.0149475447691485e-05, "loss": 0.0426, "step": 33189 }, { "epoch": 10.21, "learning_rate": 1.014897723312511e-05, "loss": 0.0386, "step": 33190 }, { "epoch": 10.21, "learning_rate": 1.0148479018188864e-05, "loss": 0.0432, "step": 33191 }, { "epoch": 10.21, "learning_rate": 1.0147980802883986e-05, "loss": 0.1037, "step": 33192 }, { "epoch": 10.21, "learning_rate": 1.014748258721171e-05, "loss": 0.0249, "step": 33193 }, { "epoch": 10.21, "learning_rate": 1.0146984371173272e-05, "loss": 0.0278, "step": 33194 }, { "epoch": 10.21, "learning_rate": 1.0146486154769916e-05, "loss": 0.044, "step": 33195 }, { "epoch": 10.21, "learning_rate": 1.0145987938002873e-05, "loss": 0.1085, "step": 33196 }, { "epoch": 10.21, "learning_rate": 1.0145489720873381e-05, "loss": 0.0327, "step": 33197 }, { "epoch": 10.21, "learning_rate": 1.0144991503382676e-05, "loss": 0.0691, "step": 33198 }, { "epoch": 10.22, "learning_rate": 1.0144493285531998e-05, "loss": 0.0397, "step": 33199 }, { "epoch": 10.22, "learning_rate": 1.0143995067322581e-05, "loss": 0.0595, "step": 33200 }, { "epoch": 10.22, "learning_rate": 1.0143496848755661e-05, "loss": 0.0499, "step": 33201 }, { "epoch": 10.22, "learning_rate": 1.0142998629832484e-05, "loss": 0.0267, "step": 33202 }, { "epoch": 10.22, "learning_rate": 1.0142500410554274e-05, "loss": 0.0634, "step": 33203 }, { "epoch": 10.22, "learning_rate": 1.0142002190922273e-05, "loss": 0.0475, "step": 33204 }, { "epoch": 10.22, "learning_rate": 1.0141503970937726e-05, "loss": 0.0556, "step": 33205 }, { "epoch": 10.22, "learning_rate": 1.0141005750601853e-05, "loss": 0.0455, "step": 33206 }, { "epoch": 10.22, "learning_rate": 1.0140507529915905e-05, "loss": 0.0871, "step": 33207 }, { "epoch": 10.22, "learning_rate": 1.0140009308881117e-05, "loss": 0.078, "step": 33208 }, { "epoch": 10.22, "learning_rate": 1.0139511087498724e-05, "loss": 0.0813, "step": 33209 }, { "epoch": 10.22, "learning_rate": 1.013901286576996e-05, "loss": 0.0653, "step": 33210 }, { "epoch": 10.22, "learning_rate": 1.0138514643696062e-05, "loss": 0.048, "step": 33211 }, { "epoch": 10.22, "learning_rate": 1.0138016421278276e-05, "loss": 0.0596, "step": 33212 }, { "epoch": 10.22, "learning_rate": 1.0137518198517828e-05, "loss": 0.0643, "step": 33213 }, { "epoch": 10.22, "learning_rate": 1.013701997541596e-05, "loss": 0.0526, "step": 33214 }, { "epoch": 10.22, "learning_rate": 1.0136521751973908e-05, "loss": 0.0615, "step": 33215 }, { "epoch": 10.22, "learning_rate": 1.013602352819291e-05, "loss": 0.0398, "step": 33216 }, { "epoch": 10.22, "learning_rate": 1.0135525304074203e-05, "loss": 0.0617, "step": 33217 }, { "epoch": 10.22, "learning_rate": 1.0135027079619024e-05, "loss": 0.0724, "step": 33218 }, { "epoch": 10.22, "learning_rate": 1.0134528854828607e-05, "loss": 0.0303, "step": 33219 }, { "epoch": 10.22, "learning_rate": 1.0134030629704192e-05, "loss": 0.0412, "step": 33220 }, { "epoch": 10.22, "learning_rate": 1.0133532404247012e-05, "loss": 0.063, "step": 33221 }, { "epoch": 10.22, "learning_rate": 1.0133034178458315e-05, "loss": 0.0374, "step": 33222 }, { "epoch": 10.22, "learning_rate": 1.0132535952339325e-05, "loss": 0.099, "step": 33223 }, { "epoch": 10.22, "learning_rate": 1.0132037725891281e-05, "loss": 0.0624, "step": 33224 }, { "epoch": 10.22, "learning_rate": 1.0131539499115426e-05, "loss": 0.0208, "step": 33225 }, { "epoch": 10.22, "learning_rate": 1.0131041272012993e-05, "loss": 0.0263, "step": 33226 }, { "epoch": 10.22, "learning_rate": 1.0130543044585222e-05, "loss": 0.0559, "step": 33227 }, { "epoch": 10.22, "learning_rate": 1.0130044816833346e-05, "loss": 0.0414, "step": 33228 }, { "epoch": 10.22, "learning_rate": 1.0129546588758605e-05, "loss": 0.0607, "step": 33229 }, { "epoch": 10.22, "learning_rate": 1.0129048360362233e-05, "loss": 0.0367, "step": 33230 }, { "epoch": 10.22, "learning_rate": 1.0128550131645467e-05, "loss": 0.0492, "step": 33231 }, { "epoch": 10.23, "learning_rate": 1.012805190260955e-05, "loss": 0.0859, "step": 33232 }, { "epoch": 10.23, "learning_rate": 1.0127553673255714e-05, "loss": 0.0464, "step": 33233 }, { "epoch": 10.23, "learning_rate": 1.0127055443585195e-05, "loss": 0.0407, "step": 33234 }, { "epoch": 10.23, "learning_rate": 1.0126557213599234e-05, "loss": 0.0744, "step": 33235 }, { "epoch": 10.23, "learning_rate": 1.012605898329906e-05, "loss": 0.0473, "step": 33236 }, { "epoch": 10.23, "learning_rate": 1.012556075268592e-05, "loss": 0.0465, "step": 33237 }, { "epoch": 10.23, "learning_rate": 1.0125062521761047e-05, "loss": 0.0339, "step": 33238 }, { "epoch": 10.23, "learning_rate": 1.0124564290525673e-05, "loss": 0.0331, "step": 33239 }, { "epoch": 10.23, "learning_rate": 1.0124066058981046e-05, "loss": 0.0372, "step": 33240 }, { "epoch": 10.23, "learning_rate": 1.0123567827128392e-05, "loss": 0.049, "step": 33241 }, { "epoch": 10.23, "learning_rate": 1.0123069594968952e-05, "loss": 0.0964, "step": 33242 }, { "epoch": 10.23, "learning_rate": 1.0122571362503968e-05, "loss": 0.0264, "step": 33243 }, { "epoch": 10.23, "learning_rate": 1.0122073129734668e-05, "loss": 0.0537, "step": 33244 }, { "epoch": 10.23, "learning_rate": 1.0121574896662294e-05, "loss": 0.0794, "step": 33245 }, { "epoch": 10.23, "learning_rate": 1.0121076663288084e-05, "loss": 0.0068, "step": 33246 }, { "epoch": 10.23, "learning_rate": 1.0120578429613272e-05, "loss": 0.0463, "step": 33247 }, { "epoch": 10.23, "learning_rate": 1.0120080195639097e-05, "loss": 0.0604, "step": 33248 }, { "epoch": 10.23, "learning_rate": 1.0119581961366791e-05, "loss": 0.0685, "step": 33249 }, { "epoch": 10.23, "learning_rate": 1.0119083726797603e-05, "loss": 0.0534, "step": 33250 }, { "epoch": 10.23, "learning_rate": 1.0118585491932755e-05, "loss": 0.0633, "step": 33251 }, { "epoch": 10.23, "learning_rate": 1.0118087256773494e-05, "loss": 0.032, "step": 33252 }, { "epoch": 10.23, "learning_rate": 1.0117589021321055e-05, "loss": 0.0161, "step": 33253 }, { "epoch": 10.23, "learning_rate": 1.0117090785576676e-05, "loss": 0.0514, "step": 33254 }, { "epoch": 10.23, "learning_rate": 1.0116592549541588e-05, "loss": 0.0672, "step": 33255 }, { "epoch": 10.23, "learning_rate": 1.0116094313217032e-05, "loss": 0.0642, "step": 33256 }, { "epoch": 10.23, "learning_rate": 1.0115596076604251e-05, "loss": 0.0821, "step": 33257 }, { "epoch": 10.23, "learning_rate": 1.0115097839704472e-05, "loss": 0.0424, "step": 33258 }, { "epoch": 10.23, "learning_rate": 1.0114599602518934e-05, "loss": 0.0614, "step": 33259 }, { "epoch": 10.23, "learning_rate": 1.0114101365048879e-05, "loss": 0.0585, "step": 33260 }, { "epoch": 10.23, "learning_rate": 1.011360312729554e-05, "loss": 0.0473, "step": 33261 }, { "epoch": 10.23, "learning_rate": 1.0113104889260154e-05, "loss": 0.0575, "step": 33262 }, { "epoch": 10.23, "learning_rate": 1.0112606650943961e-05, "loss": 0.0927, "step": 33263 }, { "epoch": 10.24, "learning_rate": 1.0112108412348196e-05, "loss": 0.059, "step": 33264 }, { "epoch": 10.24, "learning_rate": 1.0111610173474096e-05, "loss": 0.0792, "step": 33265 }, { "epoch": 10.24, "learning_rate": 1.0111111934322895e-05, "loss": 0.0555, "step": 33266 }, { "epoch": 10.24, "learning_rate": 1.0110613694895838e-05, "loss": 0.0289, "step": 33267 }, { "epoch": 10.24, "learning_rate": 1.0110115455194157e-05, "loss": 0.0554, "step": 33268 }, { "epoch": 10.24, "learning_rate": 1.0109617215219083e-05, "loss": 0.0702, "step": 33269 }, { "epoch": 10.24, "learning_rate": 1.0109118974971863e-05, "loss": 0.041, "step": 33270 }, { "epoch": 10.24, "learning_rate": 1.0108620734453729e-05, "loss": 0.0454, "step": 33271 }, { "epoch": 10.24, "learning_rate": 1.0108122493665922e-05, "loss": 0.0298, "step": 33272 }, { "epoch": 10.24, "learning_rate": 1.0107624252609674e-05, "loss": 0.0551, "step": 33273 }, { "epoch": 10.24, "learning_rate": 1.010712601128622e-05, "loss": 0.0308, "step": 33274 }, { "epoch": 10.24, "learning_rate": 1.0106627769696806e-05, "loss": 0.058, "step": 33275 }, { "epoch": 10.24, "learning_rate": 1.0106129527842661e-05, "loss": 0.0424, "step": 33276 }, { "epoch": 10.24, "learning_rate": 1.0105631285725028e-05, "loss": 0.0683, "step": 33277 }, { "epoch": 10.24, "learning_rate": 1.0105133043345142e-05, "loss": 0.0758, "step": 33278 }, { "epoch": 10.24, "learning_rate": 1.0104634800704234e-05, "loss": 0.0424, "step": 33279 }, { "epoch": 10.24, "learning_rate": 1.010413655780355e-05, "loss": 0.0637, "step": 33280 }, { "epoch": 10.24, "learning_rate": 1.0103638314644322e-05, "loss": 0.0382, "step": 33281 }, { "epoch": 10.24, "learning_rate": 1.0103140071227786e-05, "loss": 0.0651, "step": 33282 }, { "epoch": 10.24, "learning_rate": 1.0102641827555183e-05, "loss": 0.0309, "step": 33283 }, { "epoch": 10.24, "learning_rate": 1.0102143583627748e-05, "loss": 0.0683, "step": 33284 }, { "epoch": 10.24, "learning_rate": 1.0101645339446717e-05, "loss": 0.0546, "step": 33285 }, { "epoch": 10.24, "learning_rate": 1.0101147095013328e-05, "loss": 0.0576, "step": 33286 }, { "epoch": 10.24, "learning_rate": 1.0100648850328818e-05, "loss": 0.0579, "step": 33287 }, { "epoch": 10.24, "learning_rate": 1.0100150605394427e-05, "loss": 0.0542, "step": 33288 }, { "epoch": 10.24, "learning_rate": 1.0099652360211386e-05, "loss": 0.0854, "step": 33289 }, { "epoch": 10.24, "learning_rate": 1.0099154114780936e-05, "loss": 0.0553, "step": 33290 }, { "epoch": 10.24, "learning_rate": 1.0098655869104312e-05, "loss": 0.036, "step": 33291 }, { "epoch": 10.24, "learning_rate": 1.0098157623182754e-05, "loss": 0.0489, "step": 33292 }, { "epoch": 10.24, "learning_rate": 1.0097659377017497e-05, "loss": 0.0563, "step": 33293 }, { "epoch": 10.24, "learning_rate": 1.0097161130609774e-05, "loss": 0.0504, "step": 33294 }, { "epoch": 10.24, "learning_rate": 1.0096662883960833e-05, "loss": 0.0447, "step": 33295 }, { "epoch": 10.24, "learning_rate": 1.0096164637071897e-05, "loss": 0.0424, "step": 33296 }, { "epoch": 10.25, "learning_rate": 1.0095666389944214e-05, "loss": 0.0253, "step": 33297 }, { "epoch": 10.25, "learning_rate": 1.0095168142579017e-05, "loss": 0.0872, "step": 33298 }, { "epoch": 10.25, "learning_rate": 1.0094669894977542e-05, "loss": 0.0568, "step": 33299 }, { "epoch": 10.25, "learning_rate": 1.0094171647141028e-05, "loss": 0.0534, "step": 33300 }, { "epoch": 10.25, "learning_rate": 1.0093673399070714e-05, "loss": 0.0257, "step": 33301 }, { "epoch": 10.25, "learning_rate": 1.0093175150767831e-05, "loss": 0.0415, "step": 33302 }, { "epoch": 10.25, "learning_rate": 1.009267690223362e-05, "loss": 0.0809, "step": 33303 }, { "epoch": 10.25, "learning_rate": 1.0092178653469313e-05, "loss": 0.0929, "step": 33304 }, { "epoch": 10.25, "learning_rate": 1.0091680404476158e-05, "loss": 0.0201, "step": 33305 }, { "epoch": 10.25, "learning_rate": 1.0091182155255381e-05, "loss": 0.0252, "step": 33306 }, { "epoch": 10.25, "learning_rate": 1.0090683905808225e-05, "loss": 0.0355, "step": 33307 }, { "epoch": 10.25, "learning_rate": 1.0090185656135925e-05, "loss": 0.0375, "step": 33308 }, { "epoch": 10.25, "learning_rate": 1.008968740623972e-05, "loss": 0.0254, "step": 33309 }, { "epoch": 10.25, "learning_rate": 1.0089189156120842e-05, "loss": 0.0381, "step": 33310 }, { "epoch": 10.25, "learning_rate": 1.0088690905780532e-05, "loss": 0.0709, "step": 33311 }, { "epoch": 10.25, "learning_rate": 1.0088192655220028e-05, "loss": 0.0602, "step": 33312 }, { "epoch": 10.25, "learning_rate": 1.0087694404440567e-05, "loss": 0.0695, "step": 33313 }, { "epoch": 10.25, "learning_rate": 1.0087196153443379e-05, "loss": 0.0265, "step": 33314 }, { "epoch": 10.25, "learning_rate": 1.008669790222971e-05, "loss": 0.0373, "step": 33315 }, { "epoch": 10.25, "learning_rate": 1.0086199650800795e-05, "loss": 0.0249, "step": 33316 }, { "epoch": 10.25, "learning_rate": 1.0085701399157866e-05, "loss": 0.0668, "step": 33317 }, { "epoch": 10.25, "learning_rate": 1.0085203147302166e-05, "loss": 0.0171, "step": 33318 }, { "epoch": 10.25, "learning_rate": 1.0084704895234929e-05, "loss": 0.0371, "step": 33319 }, { "epoch": 10.25, "learning_rate": 1.0084206642957393e-05, "loss": 0.0581, "step": 33320 }, { "epoch": 10.25, "learning_rate": 1.0083708390470793e-05, "loss": 0.0513, "step": 33321 }, { "epoch": 10.25, "learning_rate": 1.008321013777637e-05, "loss": 0.0185, "step": 33322 }, { "epoch": 10.25, "learning_rate": 1.0082711884875358e-05, "loss": 0.0531, "step": 33323 }, { "epoch": 10.25, "learning_rate": 1.008221363176899e-05, "loss": 0.0487, "step": 33324 }, { "epoch": 10.25, "learning_rate": 1.0081715378458512e-05, "loss": 0.0607, "step": 33325 }, { "epoch": 10.25, "learning_rate": 1.0081217124945157e-05, "loss": 0.0572, "step": 33326 }, { "epoch": 10.25, "learning_rate": 1.0080718871230162e-05, "loss": 0.0524, "step": 33327 }, { "epoch": 10.25, "learning_rate": 1.0080220617314764e-05, "loss": 0.055, "step": 33328 }, { "epoch": 10.26, "learning_rate": 1.0079722363200197e-05, "loss": 0.0136, "step": 33329 }, { "epoch": 10.26, "learning_rate": 1.0079224108887704e-05, "loss": 0.0791, "step": 33330 }, { "epoch": 10.26, "learning_rate": 1.0078725854378514e-05, "loss": 0.065, "step": 33331 }, { "epoch": 10.26, "learning_rate": 1.0078227599673875e-05, "loss": 0.0677, "step": 33332 }, { "epoch": 10.26, "learning_rate": 1.0077729344775018e-05, "loss": 0.0597, "step": 33333 }, { "epoch": 10.26, "learning_rate": 1.0077231089683175e-05, "loss": 0.0496, "step": 33334 }, { "epoch": 10.26, "learning_rate": 1.007673283439959e-05, "loss": 0.0249, "step": 33335 }, { "epoch": 10.26, "learning_rate": 1.00762345789255e-05, "loss": 0.0257, "step": 33336 }, { "epoch": 10.26, "learning_rate": 1.0075736323262138e-05, "loss": 0.0573, "step": 33337 }, { "epoch": 10.26, "learning_rate": 1.0075238067410745e-05, "loss": 0.0387, "step": 33338 }, { "epoch": 10.26, "learning_rate": 1.0074739811372552e-05, "loss": 0.08, "step": 33339 }, { "epoch": 10.26, "learning_rate": 1.0074241555148805e-05, "loss": 0.0734, "step": 33340 }, { "epoch": 10.26, "learning_rate": 1.0073743298740732e-05, "loss": 0.0631, "step": 33341 }, { "epoch": 10.26, "learning_rate": 1.0073245042149578e-05, "loss": 0.0997, "step": 33342 }, { "epoch": 10.26, "learning_rate": 1.0072746785376577e-05, "loss": 0.0438, "step": 33343 }, { "epoch": 10.26, "learning_rate": 1.007224852842296e-05, "loss": 0.0532, "step": 33344 }, { "epoch": 10.26, "learning_rate": 1.0071750271289975e-05, "loss": 0.0801, "step": 33345 }, { "epoch": 10.26, "learning_rate": 1.0071252013978852e-05, "loss": 0.0559, "step": 33346 }, { "epoch": 10.26, "learning_rate": 1.0070753756490827e-05, "loss": 0.0385, "step": 33347 }, { "epoch": 10.26, "learning_rate": 1.0070255498827141e-05, "loss": 0.05, "step": 33348 }, { "epoch": 10.26, "learning_rate": 1.006975724098903e-05, "loss": 0.0781, "step": 33349 }, { "epoch": 10.26, "learning_rate": 1.006925898297773e-05, "loss": 0.0516, "step": 33350 }, { "epoch": 10.26, "learning_rate": 1.0068760724794478e-05, "loss": 0.0438, "step": 33351 }, { "epoch": 10.26, "learning_rate": 1.0068262466440513e-05, "loss": 0.0365, "step": 33352 }, { "epoch": 10.26, "learning_rate": 1.006776420791707e-05, "loss": 0.0397, "step": 33353 }, { "epoch": 10.26, "learning_rate": 1.0067265949225387e-05, "loss": 0.0319, "step": 33354 }, { "epoch": 10.26, "learning_rate": 1.00667676903667e-05, "loss": 0.0734, "step": 33355 }, { "epoch": 10.26, "learning_rate": 1.0066269431342246e-05, "loss": 0.0689, "step": 33356 }, { "epoch": 10.26, "learning_rate": 1.0065771172153267e-05, "loss": 0.0368, "step": 33357 }, { "epoch": 10.26, "learning_rate": 1.0065272912800993e-05, "loss": 0.0843, "step": 33358 }, { "epoch": 10.26, "learning_rate": 1.0064774653286662e-05, "loss": 0.0224, "step": 33359 }, { "epoch": 10.26, "learning_rate": 1.0064276393611517e-05, "loss": 0.0577, "step": 33360 }, { "epoch": 10.26, "learning_rate": 1.0063778133776789e-05, "loss": 0.0778, "step": 33361 }, { "epoch": 10.27, "learning_rate": 1.0063279873783718e-05, "loss": 0.0624, "step": 33362 }, { "epoch": 10.27, "learning_rate": 1.0062781613633538e-05, "loss": 0.0366, "step": 33363 }, { "epoch": 10.27, "learning_rate": 1.006228335332749e-05, "loss": 0.0731, "step": 33364 }, { "epoch": 10.27, "learning_rate": 1.006178509286681e-05, "loss": 0.0355, "step": 33365 }, { "epoch": 10.27, "learning_rate": 1.006128683225273e-05, "loss": 0.0523, "step": 33366 }, { "epoch": 10.27, "learning_rate": 1.0060788571486498e-05, "loss": 0.0416, "step": 33367 }, { "epoch": 10.27, "learning_rate": 1.006029031056934e-05, "loss": 0.0317, "step": 33368 }, { "epoch": 10.27, "learning_rate": 1.0059792049502497e-05, "loss": 0.0563, "step": 33369 }, { "epoch": 10.27, "learning_rate": 1.0059293788287207e-05, "loss": 0.0536, "step": 33370 }, { "epoch": 10.27, "learning_rate": 1.0058795526924708e-05, "loss": 0.0535, "step": 33371 }, { "epoch": 10.27, "learning_rate": 1.0058297265416234e-05, "loss": 0.0615, "step": 33372 }, { "epoch": 10.27, "learning_rate": 1.0057799003763026e-05, "loss": 0.0563, "step": 33373 }, { "epoch": 10.27, "learning_rate": 1.0057300741966316e-05, "loss": 0.0182, "step": 33374 }, { "epoch": 10.27, "learning_rate": 1.0056802480027346e-05, "loss": 0.0639, "step": 33375 }, { "epoch": 10.27, "learning_rate": 1.0056304217947347e-05, "loss": 0.045, "step": 33376 }, { "epoch": 10.27, "learning_rate": 1.0055805955727562e-05, "loss": 0.0483, "step": 33377 }, { "epoch": 10.27, "learning_rate": 1.005530769336923e-05, "loss": 0.0466, "step": 33378 }, { "epoch": 10.27, "learning_rate": 1.0054809430873579e-05, "loss": 0.0671, "step": 33379 }, { "epoch": 10.27, "learning_rate": 1.0054311168241851e-05, "loss": 0.0342, "step": 33380 }, { "epoch": 10.27, "learning_rate": 1.0053812905475284e-05, "loss": 0.051, "step": 33381 }, { "epoch": 10.27, "learning_rate": 1.0053314642575114e-05, "loss": 0.0853, "step": 33382 }, { "epoch": 10.27, "learning_rate": 1.0052816379542579e-05, "loss": 0.0474, "step": 33383 }, { "epoch": 10.27, "learning_rate": 1.0052318116378915e-05, "loss": 0.0939, "step": 33384 }, { "epoch": 10.27, "learning_rate": 1.0051819853085358e-05, "loss": 0.0621, "step": 33385 }, { "epoch": 10.27, "learning_rate": 1.0051321589663146e-05, "loss": 0.0335, "step": 33386 }, { "epoch": 10.27, "learning_rate": 1.0050823326113517e-05, "loss": 0.0349, "step": 33387 }, { "epoch": 10.27, "learning_rate": 1.0050325062437708e-05, "loss": 0.0696, "step": 33388 }, { "epoch": 10.27, "learning_rate": 1.0049826798636955e-05, "loss": 0.0652, "step": 33389 }, { "epoch": 10.27, "learning_rate": 1.0049328534712495e-05, "loss": 0.0576, "step": 33390 }, { "epoch": 10.27, "learning_rate": 1.0048830270665566e-05, "loss": 0.1005, "step": 33391 }, { "epoch": 10.27, "learning_rate": 1.0048332006497406e-05, "loss": 0.0592, "step": 33392 }, { "epoch": 10.27, "learning_rate": 1.0047833742209248e-05, "loss": 0.0533, "step": 33393 }, { "epoch": 10.28, "learning_rate": 1.0047335477802331e-05, "loss": 0.0574, "step": 33394 }, { "epoch": 10.28, "learning_rate": 1.0046837213277898e-05, "loss": 0.0535, "step": 33395 }, { "epoch": 10.28, "learning_rate": 1.0046338948637175e-05, "loss": 0.0457, "step": 33396 }, { "epoch": 10.28, "learning_rate": 1.0045840683881409e-05, "loss": 0.0387, "step": 33397 }, { "epoch": 10.28, "learning_rate": 1.0045342419011832e-05, "loss": 0.0752, "step": 33398 }, { "epoch": 10.28, "learning_rate": 1.004484415402968e-05, "loss": 0.0354, "step": 33399 }, { "epoch": 10.28, "learning_rate": 1.0044345888936195e-05, "loss": 0.0416, "step": 33400 }, { "epoch": 10.28, "learning_rate": 1.004384762373261e-05, "loss": 0.0683, "step": 33401 }, { "epoch": 10.28, "learning_rate": 1.0043349358420162e-05, "loss": 0.0458, "step": 33402 }, { "epoch": 10.28, "learning_rate": 1.0042851093000088e-05, "loss": 0.0429, "step": 33403 }, { "epoch": 10.28, "learning_rate": 1.004235282747363e-05, "loss": 0.0301, "step": 33404 }, { "epoch": 10.28, "learning_rate": 1.0041854561842021e-05, "loss": 0.0399, "step": 33405 }, { "epoch": 10.28, "learning_rate": 1.0041356296106494e-05, "loss": 0.0591, "step": 33406 }, { "epoch": 10.28, "learning_rate": 1.0040858030268295e-05, "loss": 0.0441, "step": 33407 }, { "epoch": 10.28, "learning_rate": 1.0040359764328656e-05, "loss": 0.0534, "step": 33408 }, { "epoch": 10.28, "learning_rate": 1.0039861498288812e-05, "loss": 0.0766, "step": 33409 }, { "epoch": 10.28, "learning_rate": 1.0039363232150006e-05, "loss": 0.0422, "step": 33410 }, { "epoch": 10.28, "learning_rate": 1.0038864965913469e-05, "loss": 0.0317, "step": 33411 }, { "epoch": 10.28, "learning_rate": 1.0038366699580441e-05, "loss": 0.0455, "step": 33412 }, { "epoch": 10.28, "learning_rate": 1.003786843315216e-05, "loss": 0.0614, "step": 33413 }, { "epoch": 10.28, "learning_rate": 1.003737016662986e-05, "loss": 0.0548, "step": 33414 }, { "epoch": 10.28, "learning_rate": 1.0036871900014781e-05, "loss": 0.0161, "step": 33415 }, { "epoch": 10.28, "learning_rate": 1.0036373633308161e-05, "loss": 0.0602, "step": 33416 }, { "epoch": 10.28, "learning_rate": 1.0035875366511234e-05, "loss": 0.0569, "step": 33417 }, { "epoch": 10.28, "learning_rate": 1.0035377099625236e-05, "loss": 0.0496, "step": 33418 }, { "epoch": 10.28, "learning_rate": 1.003487883265141e-05, "loss": 0.0658, "step": 33419 }, { "epoch": 10.28, "learning_rate": 1.0034380565590987e-05, "loss": 0.0602, "step": 33420 }, { "epoch": 10.28, "learning_rate": 1.0033882298445206e-05, "loss": 0.0513, "step": 33421 }, { "epoch": 10.28, "learning_rate": 1.0033384031215304e-05, "loss": 0.0438, "step": 33422 }, { "epoch": 10.28, "learning_rate": 1.003288576390252e-05, "loss": 0.0448, "step": 33423 }, { "epoch": 10.28, "learning_rate": 1.003238749650809e-05, "loss": 0.074, "step": 33424 }, { "epoch": 10.28, "learning_rate": 1.0031889229033246e-05, "loss": 0.04, "step": 33425 }, { "epoch": 10.28, "learning_rate": 1.0031390961479234e-05, "loss": 0.055, "step": 33426 }, { "epoch": 10.29, "learning_rate": 1.0030892693847287e-05, "loss": 0.0602, "step": 33427 }, { "epoch": 10.29, "learning_rate": 1.0030394426138639e-05, "loss": 0.0268, "step": 33428 }, { "epoch": 10.29, "learning_rate": 1.0029896158354533e-05, "loss": 0.0699, "step": 33429 }, { "epoch": 10.29, "learning_rate": 1.0029397890496203e-05, "loss": 0.0503, "step": 33430 }, { "epoch": 10.29, "learning_rate": 1.002889962256488e-05, "loss": 0.076, "step": 33431 }, { "epoch": 10.29, "learning_rate": 1.0028401354561812e-05, "loss": 0.0509, "step": 33432 }, { "epoch": 10.29, "learning_rate": 1.0027903086488231e-05, "loss": 0.0586, "step": 33433 }, { "epoch": 10.29, "learning_rate": 1.0027404818345374e-05, "loss": 0.054, "step": 33434 }, { "epoch": 10.29, "learning_rate": 1.0026906550134476e-05, "loss": 0.0808, "step": 33435 }, { "epoch": 10.29, "learning_rate": 1.0026408281856779e-05, "loss": 0.0382, "step": 33436 }, { "epoch": 10.29, "learning_rate": 1.0025910013513519e-05, "loss": 0.0372, "step": 33437 }, { "epoch": 10.29, "learning_rate": 1.0025411745105924e-05, "loss": 0.0444, "step": 33438 }, { "epoch": 10.29, "learning_rate": 1.0024913476635245e-05, "loss": 0.0828, "step": 33439 }, { "epoch": 10.29, "learning_rate": 1.0024415208102713e-05, "loss": 0.0479, "step": 33440 }, { "epoch": 10.29, "learning_rate": 1.002391693950956e-05, "loss": 0.091, "step": 33441 }, { "epoch": 10.29, "learning_rate": 1.002341867085703e-05, "loss": 0.0899, "step": 33442 }, { "epoch": 10.29, "learning_rate": 1.0022920402146359e-05, "loss": 0.0566, "step": 33443 }, { "epoch": 10.29, "learning_rate": 1.0022422133378783e-05, "loss": 0.0838, "step": 33444 }, { "epoch": 10.29, "learning_rate": 1.0021923864555536e-05, "loss": 0.0364, "step": 33445 }, { "epoch": 10.29, "learning_rate": 1.002142559567786e-05, "loss": 0.0411, "step": 33446 }, { "epoch": 10.29, "learning_rate": 1.002092732674699e-05, "loss": 0.0372, "step": 33447 }, { "epoch": 10.29, "learning_rate": 1.002042905776416e-05, "loss": 0.0531, "step": 33448 }, { "epoch": 10.29, "learning_rate": 1.0019930788730615e-05, "loss": 0.0749, "step": 33449 }, { "epoch": 10.29, "learning_rate": 1.0019432519647585e-05, "loss": 0.0271, "step": 33450 }, { "epoch": 10.29, "learning_rate": 1.0018934250516308e-05, "loss": 0.0641, "step": 33451 }, { "epoch": 10.29, "learning_rate": 1.0018435981338023e-05, "loss": 0.0296, "step": 33452 }, { "epoch": 10.29, "learning_rate": 1.0017937712113967e-05, "loss": 0.032, "step": 33453 }, { "epoch": 10.29, "learning_rate": 1.0017439442845378e-05, "loss": 0.0599, "step": 33454 }, { "epoch": 10.29, "learning_rate": 1.001694117353349e-05, "loss": 0.0766, "step": 33455 }, { "epoch": 10.29, "learning_rate": 1.0016442904179539e-05, "loss": 0.0602, "step": 33456 }, { "epoch": 10.29, "learning_rate": 1.0015944634784772e-05, "loss": 0.0452, "step": 33457 }, { "epoch": 10.29, "learning_rate": 1.0015446365350413e-05, "loss": 0.0416, "step": 33458 }, { "epoch": 10.3, "learning_rate": 1.0014948095877705e-05, "loss": 0.0606, "step": 33459 }, { "epoch": 10.3, "learning_rate": 1.0014449826367889e-05, "loss": 0.048, "step": 33460 }, { "epoch": 10.3, "learning_rate": 1.0013951556822192e-05, "loss": 0.0952, "step": 33461 }, { "epoch": 10.3, "learning_rate": 1.001345328724186e-05, "loss": 0.073, "step": 33462 }, { "epoch": 10.3, "learning_rate": 1.0012955017628127e-05, "loss": 0.0515, "step": 33463 }, { "epoch": 10.3, "learning_rate": 1.0012456747982231e-05, "loss": 0.038, "step": 33464 }, { "epoch": 10.3, "learning_rate": 1.0011958478305407e-05, "loss": 0.0351, "step": 33465 }, { "epoch": 10.3, "learning_rate": 1.0011460208598893e-05, "loss": 0.0863, "step": 33466 }, { "epoch": 10.3, "learning_rate": 1.001096193886393e-05, "loss": 0.0283, "step": 33467 }, { "epoch": 10.3, "learning_rate": 1.0010463669101746e-05, "loss": 0.036, "step": 33468 }, { "epoch": 10.3, "learning_rate": 1.0009965399313586e-05, "loss": 0.0551, "step": 33469 }, { "epoch": 10.3, "learning_rate": 1.0009467129500685e-05, "loss": 0.0361, "step": 33470 }, { "epoch": 10.3, "learning_rate": 1.0008968859664281e-05, "loss": 0.0386, "step": 33471 }, { "epoch": 10.3, "learning_rate": 1.0008470589805608e-05, "loss": 0.0859, "step": 33472 }, { "epoch": 10.3, "learning_rate": 1.0007972319925903e-05, "loss": 0.0536, "step": 33473 }, { "epoch": 10.3, "learning_rate": 1.0007474050026409e-05, "loss": 0.061, "step": 33474 }, { "epoch": 10.3, "learning_rate": 1.0006975780108358e-05, "loss": 0.0873, "step": 33475 }, { "epoch": 10.3, "learning_rate": 1.0006477510172984e-05, "loss": 0.0442, "step": 33476 }, { "epoch": 10.3, "learning_rate": 1.0005979240221533e-05, "loss": 0.0972, "step": 33477 }, { "epoch": 10.3, "learning_rate": 1.0005480970255234e-05, "loss": 0.0524, "step": 33478 }, { "epoch": 10.3, "learning_rate": 1.000498270027533e-05, "loss": 0.0629, "step": 33479 }, { "epoch": 10.3, "learning_rate": 1.0004484430283052e-05, "loss": 0.0506, "step": 33480 }, { "epoch": 10.3, "learning_rate": 1.0003986160279645e-05, "loss": 0.0715, "step": 33481 }, { "epoch": 10.3, "learning_rate": 1.0003487890266337e-05, "loss": 0.0746, "step": 33482 }, { "epoch": 10.3, "learning_rate": 1.000298962024437e-05, "loss": 0.0594, "step": 33483 }, { "epoch": 10.3, "learning_rate": 1.0002491350214983e-05, "loss": 0.039, "step": 33484 }, { "epoch": 10.3, "learning_rate": 1.0001993080179411e-05, "loss": 0.0583, "step": 33485 }, { "epoch": 10.3, "learning_rate": 1.0001494810138885e-05, "loss": 0.0173, "step": 33486 }, { "epoch": 10.3, "learning_rate": 1.0000996540094652e-05, "loss": 0.0513, "step": 33487 }, { "epoch": 10.3, "learning_rate": 1.0000498270047946e-05, "loss": 0.0418, "step": 33488 }, { "epoch": 10.3, "learning_rate": 1e-05, "loss": 0.0407, "step": 33489 }, { "epoch": 10.3, "learning_rate": 9.999501729952056e-06, "loss": 0.015, "step": 33490 }, { "epoch": 10.3, "learning_rate": 9.99900345990535e-06, "loss": 0.0586, "step": 33491 }, { "epoch": 10.31, "learning_rate": 9.998505189861115e-06, "loss": 0.0779, "step": 33492 }, { "epoch": 10.31, "learning_rate": 9.998006919820594e-06, "loss": 0.0418, "step": 33493 }, { "epoch": 10.31, "learning_rate": 9.997508649785019e-06, "loss": 0.0395, "step": 33494 }, { "epoch": 10.31, "learning_rate": 9.997010379755635e-06, "loss": 0.0339, "step": 33495 }, { "epoch": 10.31, "learning_rate": 9.996512109733668e-06, "loss": 0.0599, "step": 33496 }, { "epoch": 10.31, "learning_rate": 9.996013839720357e-06, "loss": 0.0538, "step": 33497 }, { "epoch": 10.31, "learning_rate": 9.99551556971695e-06, "loss": 0.0841, "step": 33498 }, { "epoch": 10.31, "learning_rate": 9.995017299724675e-06, "loss": 0.0361, "step": 33499 }, { "epoch": 10.31, "learning_rate": 9.994519029744767e-06, "loss": 0.0449, "step": 33500 }, { "epoch": 10.31, "learning_rate": 9.994020759778472e-06, "loss": 0.0388, "step": 33501 }, { "epoch": 10.31, "learning_rate": 9.993522489827016e-06, "loss": 0.0563, "step": 33502 }, { "epoch": 10.31, "learning_rate": 9.993024219891645e-06, "loss": 0.0938, "step": 33503 }, { "epoch": 10.31, "learning_rate": 9.992525949973593e-06, "loss": 0.01, "step": 33504 }, { "epoch": 10.31, "learning_rate": 9.992027680074099e-06, "loss": 0.0469, "step": 33505 }, { "epoch": 10.31, "learning_rate": 9.991529410194397e-06, "loss": 0.0679, "step": 33506 }, { "epoch": 10.31, "learning_rate": 9.99103114033572e-06, "loss": 0.043, "step": 33507 }, { "epoch": 10.31, "learning_rate": 9.990532870499317e-06, "loss": 0.1019, "step": 33508 }, { "epoch": 10.31, "learning_rate": 9.990034600686415e-06, "loss": 0.03, "step": 33509 }, { "epoch": 10.31, "learning_rate": 9.989536330898256e-06, "loss": 0.0415, "step": 33510 }, { "epoch": 10.31, "learning_rate": 9.989038061136074e-06, "loss": 0.0482, "step": 33511 }, { "epoch": 10.31, "learning_rate": 9.988539791401107e-06, "loss": 0.0832, "step": 33512 }, { "epoch": 10.31, "learning_rate": 9.988041521694594e-06, "loss": 0.1358, "step": 33513 }, { "epoch": 10.31, "learning_rate": 9.987543252017772e-06, "loss": 0.0836, "step": 33514 }, { "epoch": 10.31, "learning_rate": 9.987044982371875e-06, "loss": 0.0568, "step": 33515 }, { "epoch": 10.31, "learning_rate": 9.986546712758145e-06, "loss": 0.0435, "step": 33516 }, { "epoch": 10.31, "learning_rate": 9.986048443177808e-06, "loss": 0.0221, "step": 33517 }, { "epoch": 10.31, "learning_rate": 9.985550173632115e-06, "loss": 0.0368, "step": 33518 }, { "epoch": 10.31, "learning_rate": 9.985051904122298e-06, "loss": 0.0746, "step": 33519 }, { "epoch": 10.31, "learning_rate": 9.984553634649592e-06, "loss": 0.081, "step": 33520 }, { "epoch": 10.31, "learning_rate": 9.984055365215233e-06, "loss": 0.0354, "step": 33521 }, { "epoch": 10.31, "learning_rate": 9.983557095820461e-06, "loss": 0.0591, "step": 33522 }, { "epoch": 10.31, "learning_rate": 9.983058826466512e-06, "loss": 0.0589, "step": 33523 }, { "epoch": 10.32, "learning_rate": 9.982560557154624e-06, "loss": 0.0506, "step": 33524 }, { "epoch": 10.32, "learning_rate": 9.982062287886034e-06, "loss": 0.0447, "step": 33525 }, { "epoch": 10.32, "learning_rate": 9.981564018661982e-06, "loss": 0.0779, "step": 33526 }, { "epoch": 10.32, "learning_rate": 9.981065749483692e-06, "loss": 0.0617, "step": 33527 }, { "epoch": 10.32, "learning_rate": 9.980567480352417e-06, "loss": 0.0746, "step": 33528 }, { "epoch": 10.32, "learning_rate": 9.98006921126939e-06, "loss": 0.087, "step": 33529 }, { "epoch": 10.32, "learning_rate": 9.979570942235842e-06, "loss": 0.0569, "step": 33530 }, { "epoch": 10.32, "learning_rate": 9.979072673253016e-06, "loss": 0.053, "step": 33531 }, { "epoch": 10.32, "learning_rate": 9.978574404322143e-06, "loss": 0.0446, "step": 33532 }, { "epoch": 10.32, "learning_rate": 9.978076135444466e-06, "loss": 0.057, "step": 33533 }, { "epoch": 10.32, "learning_rate": 9.977577866621222e-06, "loss": 0.0551, "step": 33534 }, { "epoch": 10.32, "learning_rate": 9.977079597853644e-06, "loss": 0.0631, "step": 33535 }, { "epoch": 10.32, "learning_rate": 9.976581329142973e-06, "loss": 0.0705, "step": 33536 }, { "epoch": 10.32, "learning_rate": 9.97608306049044e-06, "loss": 0.0468, "step": 33537 }, { "epoch": 10.32, "learning_rate": 9.97558479189729e-06, "loss": 0.0511, "step": 33538 }, { "epoch": 10.32, "learning_rate": 9.975086523364759e-06, "loss": 0.0431, "step": 33539 }, { "epoch": 10.32, "learning_rate": 9.974588254894077e-06, "loss": 0.0738, "step": 33540 }, { "epoch": 10.32, "learning_rate": 9.974089986486488e-06, "loss": 0.0724, "step": 33541 }, { "epoch": 10.32, "learning_rate": 9.973591718143223e-06, "loss": 0.0372, "step": 33542 }, { "epoch": 10.32, "learning_rate": 9.973093449865526e-06, "loss": 0.0964, "step": 33543 }, { "epoch": 10.32, "learning_rate": 9.97259518165463e-06, "loss": 0.0424, "step": 33544 }, { "epoch": 10.32, "learning_rate": 9.972096913511772e-06, "loss": 0.0533, "step": 33545 }, { "epoch": 10.32, "learning_rate": 9.97159864543819e-06, "loss": 0.0774, "step": 33546 }, { "epoch": 10.32, "learning_rate": 9.97110037743512e-06, "loss": 0.082, "step": 33547 }, { "epoch": 10.32, "learning_rate": 9.9706021095038e-06, "loss": 0.0459, "step": 33548 }, { "epoch": 10.32, "learning_rate": 9.97010384164547e-06, "loss": 0.0678, "step": 33549 }, { "epoch": 10.32, "learning_rate": 9.969605573861365e-06, "loss": 0.0713, "step": 33550 }, { "epoch": 10.32, "learning_rate": 9.969107306152718e-06, "loss": 0.0502, "step": 33551 }, { "epoch": 10.32, "learning_rate": 9.968609038520766e-06, "loss": 0.0747, "step": 33552 }, { "epoch": 10.32, "learning_rate": 9.968110770966755e-06, "loss": 0.0526, "step": 33553 }, { "epoch": 10.32, "learning_rate": 9.967612503491915e-06, "loss": 0.0783, "step": 33554 }, { "epoch": 10.32, "learning_rate": 9.967114236097484e-06, "loss": 0.0746, "step": 33555 }, { "epoch": 10.32, "learning_rate": 9.966615968784699e-06, "loss": 0.0698, "step": 33556 }, { "epoch": 10.33, "learning_rate": 9.966117701554795e-06, "loss": 0.0886, "step": 33557 }, { "epoch": 10.33, "learning_rate": 9.965619434409014e-06, "loss": 0.0467, "step": 33558 }, { "epoch": 10.33, "learning_rate": 9.965121167348592e-06, "loss": 0.0247, "step": 33559 }, { "epoch": 10.33, "learning_rate": 9.964622900374767e-06, "loss": 0.0736, "step": 33560 }, { "epoch": 10.33, "learning_rate": 9.964124633488771e-06, "loss": 0.0533, "step": 33561 }, { "epoch": 10.33, "learning_rate": 9.963626366691839e-06, "loss": 0.0567, "step": 33562 }, { "epoch": 10.33, "learning_rate": 9.96312809998522e-06, "loss": 0.1019, "step": 33563 }, { "epoch": 10.33, "learning_rate": 9.962629833370143e-06, "loss": 0.0729, "step": 33564 }, { "epoch": 10.33, "learning_rate": 9.962131566847844e-06, "loss": 0.0805, "step": 33565 }, { "epoch": 10.33, "learning_rate": 9.961633300419562e-06, "loss": 0.0474, "step": 33566 }, { "epoch": 10.33, "learning_rate": 9.961135034086533e-06, "loss": 0.065, "step": 33567 }, { "epoch": 10.33, "learning_rate": 9.960636767849998e-06, "loss": 0.0725, "step": 33568 }, { "epoch": 10.33, "learning_rate": 9.96013850171119e-06, "loss": 0.0577, "step": 33569 }, { "epoch": 10.33, "learning_rate": 9.959640235671347e-06, "loss": 0.089, "step": 33570 }, { "epoch": 10.33, "learning_rate": 9.95914196973171e-06, "loss": 0.0666, "step": 33571 }, { "epoch": 10.33, "learning_rate": 9.958643703893506e-06, "loss": 0.0331, "step": 33572 }, { "epoch": 10.33, "learning_rate": 9.95814543815798e-06, "loss": 0.0683, "step": 33573 }, { "epoch": 10.33, "learning_rate": 9.957647172526374e-06, "loss": 0.0283, "step": 33574 }, { "epoch": 10.33, "learning_rate": 9.957148906999913e-06, "loss": 0.0822, "step": 33575 }, { "epoch": 10.33, "learning_rate": 9.956650641579843e-06, "loss": 0.0759, "step": 33576 }, { "epoch": 10.33, "learning_rate": 9.956152376267394e-06, "loss": 0.0932, "step": 33577 }, { "epoch": 10.33, "learning_rate": 9.955654111063808e-06, "loss": 0.0548, "step": 33578 }, { "epoch": 10.33, "learning_rate": 9.955155845970322e-06, "loss": 0.0725, "step": 33579 }, { "epoch": 10.33, "learning_rate": 9.954657580988171e-06, "loss": 0.0816, "step": 33580 }, { "epoch": 10.33, "learning_rate": 9.954159316118596e-06, "loss": 0.0203, "step": 33581 }, { "epoch": 10.33, "learning_rate": 9.953661051362825e-06, "loss": 0.034, "step": 33582 }, { "epoch": 10.33, "learning_rate": 9.953162786722105e-06, "loss": 0.0382, "step": 33583 }, { "epoch": 10.33, "learning_rate": 9.95266452219767e-06, "loss": 0.0513, "step": 33584 }, { "epoch": 10.33, "learning_rate": 9.952166257790757e-06, "loss": 0.0842, "step": 33585 }, { "epoch": 10.33, "learning_rate": 9.951667993502599e-06, "loss": 0.0618, "step": 33586 }, { "epoch": 10.33, "learning_rate": 9.951169729334437e-06, "loss": 0.0831, "step": 33587 }, { "epoch": 10.33, "learning_rate": 9.950671465287509e-06, "loss": 0.0709, "step": 33588 }, { "epoch": 10.34, "learning_rate": 9.950173201363048e-06, "loss": 0.0481, "step": 33589 }, { "epoch": 10.34, "learning_rate": 9.949674937562295e-06, "loss": 0.0441, "step": 33590 }, { "epoch": 10.34, "learning_rate": 9.949176673886486e-06, "loss": 0.0665, "step": 33591 }, { "epoch": 10.34, "learning_rate": 9.948678410336854e-06, "loss": 0.0584, "step": 33592 }, { "epoch": 10.34, "learning_rate": 9.948180146914644e-06, "loss": 0.0419, "step": 33593 }, { "epoch": 10.34, "learning_rate": 9.947681883621089e-06, "loss": 0.0432, "step": 33594 }, { "epoch": 10.34, "learning_rate": 9.947183620457426e-06, "loss": 0.0683, "step": 33595 }, { "epoch": 10.34, "learning_rate": 9.946685357424891e-06, "loss": 0.0755, "step": 33596 }, { "epoch": 10.34, "learning_rate": 9.946187094524719e-06, "loss": 0.0789, "step": 33597 }, { "epoch": 10.34, "learning_rate": 9.945688831758152e-06, "loss": 0.0299, "step": 33598 }, { "epoch": 10.34, "learning_rate": 9.945190569126425e-06, "loss": 0.0484, "step": 33599 }, { "epoch": 10.34, "learning_rate": 9.944692306630775e-06, "loss": 0.0647, "step": 33600 }, { "epoch": 10.34, "learning_rate": 9.94419404427244e-06, "loss": 0.0749, "step": 33601 }, { "epoch": 10.34, "learning_rate": 9.943695782052654e-06, "loss": 0.0161, "step": 33602 }, { "epoch": 10.34, "learning_rate": 9.943197519972657e-06, "loss": 0.0623, "step": 33603 }, { "epoch": 10.34, "learning_rate": 9.942699258033686e-06, "loss": 0.0478, "step": 33604 }, { "epoch": 10.34, "learning_rate": 9.942200996236979e-06, "loss": 0.0347, "step": 33605 }, { "epoch": 10.34, "learning_rate": 9.941702734583771e-06, "loss": 0.0542, "step": 33606 }, { "epoch": 10.34, "learning_rate": 9.941204473075292e-06, "loss": 0.0621, "step": 33607 }, { "epoch": 10.34, "learning_rate": 9.940706211712794e-06, "loss": 0.0593, "step": 33608 }, { "epoch": 10.34, "learning_rate": 9.940207950497506e-06, "loss": 0.0879, "step": 33609 }, { "epoch": 10.34, "learning_rate": 9.939709689430665e-06, "loss": 0.0521, "step": 33610 }, { "epoch": 10.34, "learning_rate": 9.939211428513507e-06, "loss": 0.0643, "step": 33611 }, { "epoch": 10.34, "learning_rate": 9.93871316774727e-06, "loss": 0.0722, "step": 33612 }, { "epoch": 10.34, "learning_rate": 9.938214907133193e-06, "loss": 0.0625, "step": 33613 }, { "epoch": 10.34, "learning_rate": 9.937716646672513e-06, "loss": 0.0602, "step": 33614 }, { "epoch": 10.34, "learning_rate": 9.937218386366464e-06, "loss": 0.0568, "step": 33615 }, { "epoch": 10.34, "learning_rate": 9.936720126216289e-06, "loss": 0.0499, "step": 33616 }, { "epoch": 10.34, "learning_rate": 9.936221866223213e-06, "loss": 0.0506, "step": 33617 }, { "epoch": 10.34, "learning_rate": 9.935723606388486e-06, "loss": 0.102, "step": 33618 }, { "epoch": 10.34, "learning_rate": 9.935225346713341e-06, "loss": 0.0315, "step": 33619 }, { "epoch": 10.34, "learning_rate": 9.934727087199012e-06, "loss": 0.0778, "step": 33620 }, { "epoch": 10.34, "learning_rate": 9.934228827846738e-06, "loss": 0.0627, "step": 33621 }, { "epoch": 10.35, "learning_rate": 9.933730568657754e-06, "loss": 0.0267, "step": 33622 }, { "epoch": 10.35, "learning_rate": 9.933232309633303e-06, "loss": 0.0429, "step": 33623 }, { "epoch": 10.35, "learning_rate": 9.932734050774616e-06, "loss": 0.0688, "step": 33624 }, { "epoch": 10.35, "learning_rate": 9.932235792082932e-06, "loss": 0.0387, "step": 33625 }, { "epoch": 10.35, "learning_rate": 9.931737533559492e-06, "loss": 0.0583, "step": 33626 }, { "epoch": 10.35, "learning_rate": 9.931239275205523e-06, "loss": 0.0604, "step": 33627 }, { "epoch": 10.35, "learning_rate": 9.930741017022272e-06, "loss": 0.047, "step": 33628 }, { "epoch": 10.35, "learning_rate": 9.930242759010974e-06, "loss": 0.0458, "step": 33629 }, { "epoch": 10.35, "learning_rate": 9.929744501172864e-06, "loss": 0.0431, "step": 33630 }, { "epoch": 10.35, "learning_rate": 9.929246243509178e-06, "loss": 0.0672, "step": 33631 }, { "epoch": 10.35, "learning_rate": 9.928747986021153e-06, "loss": 0.0668, "step": 33632 }, { "epoch": 10.35, "learning_rate": 9.928249728710029e-06, "loss": 0.0474, "step": 33633 }, { "epoch": 10.35, "learning_rate": 9.927751471577041e-06, "loss": 0.0557, "step": 33634 }, { "epoch": 10.35, "learning_rate": 9.927253214623428e-06, "loss": 0.0537, "step": 33635 }, { "epoch": 10.35, "learning_rate": 9.926754957850426e-06, "loss": 0.0655, "step": 33636 }, { "epoch": 10.35, "learning_rate": 9.926256701259268e-06, "loss": 0.0541, "step": 33637 }, { "epoch": 10.35, "learning_rate": 9.925758444851196e-06, "loss": 0.0707, "step": 33638 }, { "epoch": 10.35, "learning_rate": 9.92526018862745e-06, "loss": 0.0789, "step": 33639 }, { "epoch": 10.35, "learning_rate": 9.92476193258926e-06, "loss": 0.0621, "step": 33640 }, { "epoch": 10.35, "learning_rate": 9.924263676737866e-06, "loss": 0.0844, "step": 33641 }, { "epoch": 10.35, "learning_rate": 9.923765421074504e-06, "loss": 0.0526, "step": 33642 }, { "epoch": 10.35, "learning_rate": 9.923267165600411e-06, "loss": 0.039, "step": 33643 }, { "epoch": 10.35, "learning_rate": 9.922768910316829e-06, "loss": 0.0514, "step": 33644 }, { "epoch": 10.35, "learning_rate": 9.922270655224989e-06, "loss": 0.0534, "step": 33645 }, { "epoch": 10.35, "learning_rate": 9.921772400326128e-06, "loss": 0.0334, "step": 33646 }, { "epoch": 10.35, "learning_rate": 9.921274145621484e-06, "loss": 0.0597, "step": 33647 }, { "epoch": 10.35, "learning_rate": 9.9207758911123e-06, "loss": 0.05, "step": 33648 }, { "epoch": 10.35, "learning_rate": 9.920277636799804e-06, "loss": 0.0653, "step": 33649 }, { "epoch": 10.35, "learning_rate": 9.919779382685242e-06, "loss": 0.049, "step": 33650 }, { "epoch": 10.35, "learning_rate": 9.919281128769843e-06, "loss": 0.0632, "step": 33651 }, { "epoch": 10.35, "learning_rate": 9.918782875054843e-06, "loss": 0.0323, "step": 33652 }, { "epoch": 10.35, "learning_rate": 9.91828462154149e-06, "loss": 0.0606, "step": 33653 }, { "epoch": 10.36, "learning_rate": 9.917786368231011e-06, "loss": 0.0772, "step": 33654 }, { "epoch": 10.36, "learning_rate": 9.917288115124649e-06, "loss": 0.0739, "step": 33655 }, { "epoch": 10.36, "learning_rate": 9.916789862223636e-06, "loss": 0.0734, "step": 33656 }, { "epoch": 10.36, "learning_rate": 9.916291609529209e-06, "loss": 0.0846, "step": 33657 }, { "epoch": 10.36, "learning_rate": 9.91579335704261e-06, "loss": 0.0554, "step": 33658 }, { "epoch": 10.36, "learning_rate": 9.915295104765073e-06, "loss": 0.0746, "step": 33659 }, { "epoch": 10.36, "learning_rate": 9.914796852697839e-06, "loss": 0.0732, "step": 33660 }, { "epoch": 10.36, "learning_rate": 9.914298600842139e-06, "loss": 0.062, "step": 33661 }, { "epoch": 10.36, "learning_rate": 9.913800349199207e-06, "loss": 0.0498, "step": 33662 }, { "epoch": 10.36, "learning_rate": 9.913302097770292e-06, "loss": 0.0631, "step": 33663 }, { "epoch": 10.36, "learning_rate": 9.912803846556623e-06, "loss": 0.0728, "step": 33664 }, { "epoch": 10.36, "learning_rate": 9.91230559555944e-06, "loss": 0.0407, "step": 33665 }, { "epoch": 10.36, "learning_rate": 9.911807344779975e-06, "loss": 0.0665, "step": 33666 }, { "epoch": 10.36, "learning_rate": 9.911309094219468e-06, "loss": 0.0405, "step": 33667 }, { "epoch": 10.36, "learning_rate": 9.910810843879161e-06, "loss": 0.0652, "step": 33668 }, { "epoch": 10.36, "learning_rate": 9.910312593760284e-06, "loss": 0.0823, "step": 33669 }, { "epoch": 10.36, "learning_rate": 9.909814343864078e-06, "loss": 0.0463, "step": 33670 }, { "epoch": 10.36, "learning_rate": 9.90931609419178e-06, "loss": 0.0987, "step": 33671 }, { "epoch": 10.36, "learning_rate": 9.908817844744619e-06, "loss": 0.0699, "step": 33672 }, { "epoch": 10.36, "learning_rate": 9.908319595523844e-06, "loss": 0.0482, "step": 33673 }, { "epoch": 10.36, "learning_rate": 9.907821346530689e-06, "loss": 0.0695, "step": 33674 }, { "epoch": 10.36, "learning_rate": 9.907323097766386e-06, "loss": 0.0551, "step": 33675 }, { "epoch": 10.36, "learning_rate": 9.906824849232176e-06, "loss": 0.0505, "step": 33676 }, { "epoch": 10.36, "learning_rate": 9.90632660092929e-06, "loss": 0.0635, "step": 33677 }, { "epoch": 10.36, "learning_rate": 9.905828352858973e-06, "loss": 0.0427, "step": 33678 }, { "epoch": 10.36, "learning_rate": 9.90533010502246e-06, "loss": 0.0341, "step": 33679 }, { "epoch": 10.36, "learning_rate": 9.904831857420985e-06, "loss": 0.0557, "step": 33680 }, { "epoch": 10.36, "learning_rate": 9.904333610055791e-06, "loss": 0.0841, "step": 33681 }, { "epoch": 10.36, "learning_rate": 9.903835362928103e-06, "loss": 0.0435, "step": 33682 }, { "epoch": 10.36, "learning_rate": 9.903337116039172e-06, "loss": 0.0716, "step": 33683 }, { "epoch": 10.36, "learning_rate": 9.90283886939023e-06, "loss": 0.0532, "step": 33684 }, { "epoch": 10.36, "learning_rate": 9.902340622982508e-06, "loss": 0.0381, "step": 33685 }, { "epoch": 10.36, "learning_rate": 9.90184237681725e-06, "loss": 0.0512, "step": 33686 }, { "epoch": 10.37, "learning_rate": 9.901344130895691e-06, "loss": 0.0588, "step": 33687 }, { "epoch": 10.37, "learning_rate": 9.900845885219068e-06, "loss": 0.0719, "step": 33688 }, { "epoch": 10.37, "learning_rate": 9.900347639788616e-06, "loss": 0.0514, "step": 33689 }, { "epoch": 10.37, "learning_rate": 9.899849394605578e-06, "loss": 0.065, "step": 33690 }, { "epoch": 10.37, "learning_rate": 9.89935114967118e-06, "loss": 0.0547, "step": 33691 }, { "epoch": 10.37, "learning_rate": 9.898852904986674e-06, "loss": 0.0528, "step": 33692 }, { "epoch": 10.37, "learning_rate": 9.898354660553285e-06, "loss": 0.1273, "step": 33693 }, { "epoch": 10.37, "learning_rate": 9.897856416372254e-06, "loss": 0.0786, "step": 33694 }, { "epoch": 10.37, "learning_rate": 9.897358172444822e-06, "loss": 0.087, "step": 33695 }, { "epoch": 10.37, "learning_rate": 9.896859928772214e-06, "loss": 0.0514, "step": 33696 }, { "epoch": 10.37, "learning_rate": 9.896361685355681e-06, "loss": 0.0684, "step": 33697 }, { "epoch": 10.37, "learning_rate": 9.895863442196454e-06, "loss": 0.0566, "step": 33698 }, { "epoch": 10.37, "learning_rate": 9.89536519929577e-06, "loss": 0.0571, "step": 33699 }, { "epoch": 10.37, "learning_rate": 9.894866956654863e-06, "loss": 0.069, "step": 33700 }, { "epoch": 10.37, "learning_rate": 9.894368714274973e-06, "loss": 0.0626, "step": 33701 }, { "epoch": 10.37, "learning_rate": 9.893870472157339e-06, "loss": 0.0393, "step": 33702 }, { "epoch": 10.37, "learning_rate": 9.893372230303196e-06, "loss": 0.097, "step": 33703 }, { "epoch": 10.37, "learning_rate": 9.892873988713781e-06, "loss": 0.099, "step": 33704 }, { "epoch": 10.37, "learning_rate": 9.892375747390333e-06, "loss": 0.0624, "step": 33705 }, { "epoch": 10.37, "learning_rate": 9.89187750633408e-06, "loss": 0.0323, "step": 33706 }, { "epoch": 10.37, "learning_rate": 9.89137926554627e-06, "loss": 0.0348, "step": 33707 }, { "epoch": 10.37, "learning_rate": 9.89088102502814e-06, "loss": 0.0574, "step": 33708 }, { "epoch": 10.37, "learning_rate": 9.89038278478092e-06, "loss": 0.0622, "step": 33709 }, { "epoch": 10.37, "learning_rate": 9.88988454480585e-06, "loss": 0.0553, "step": 33710 }, { "epoch": 10.37, "learning_rate": 9.889386305104163e-06, "loss": 0.087, "step": 33711 }, { "epoch": 10.37, "learning_rate": 9.888888065677105e-06, "loss": 0.0615, "step": 33712 }, { "epoch": 10.37, "learning_rate": 9.888389826525907e-06, "loss": 0.056, "step": 33713 }, { "epoch": 10.37, "learning_rate": 9.887891587651807e-06, "loss": 0.1396, "step": 33714 }, { "epoch": 10.37, "learning_rate": 9.887393349056042e-06, "loss": 0.0688, "step": 33715 }, { "epoch": 10.37, "learning_rate": 9.886895110739846e-06, "loss": 0.0457, "step": 33716 }, { "epoch": 10.37, "learning_rate": 9.88639687270446e-06, "loss": 0.075, "step": 33717 }, { "epoch": 10.37, "learning_rate": 9.885898634951126e-06, "loss": 0.0743, "step": 33718 }, { "epoch": 10.38, "learning_rate": 9.88540039748107e-06, "loss": 0.0462, "step": 33719 }, { "epoch": 10.38, "learning_rate": 9.884902160295533e-06, "loss": 0.0653, "step": 33720 }, { "epoch": 10.38, "learning_rate": 9.884403923395754e-06, "loss": 0.0422, "step": 33721 }, { "epoch": 10.38, "learning_rate": 9.88390568678297e-06, "loss": 0.0511, "step": 33722 }, { "epoch": 10.38, "learning_rate": 9.883407450458415e-06, "loss": 0.0595, "step": 33723 }, { "epoch": 10.38, "learning_rate": 9.882909214423329e-06, "loss": 0.0593, "step": 33724 }, { "epoch": 10.38, "learning_rate": 9.882410978678946e-06, "loss": 0.0357, "step": 33725 }, { "epoch": 10.38, "learning_rate": 9.881912743226505e-06, "loss": 0.0807, "step": 33726 }, { "epoch": 10.38, "learning_rate": 9.881414508067245e-06, "loss": 0.0822, "step": 33727 }, { "epoch": 10.38, "learning_rate": 9.8809162732024e-06, "loss": 0.061, "step": 33728 }, { "epoch": 10.38, "learning_rate": 9.88041803863321e-06, "loss": 0.0288, "step": 33729 }, { "epoch": 10.38, "learning_rate": 9.879919804360908e-06, "loss": 0.0547, "step": 33730 }, { "epoch": 10.38, "learning_rate": 9.87942157038673e-06, "loss": 0.0516, "step": 33731 }, { "epoch": 10.38, "learning_rate": 9.87892333671192e-06, "loss": 0.0544, "step": 33732 }, { "epoch": 10.38, "learning_rate": 9.878425103337708e-06, "loss": 0.0729, "step": 33733 }, { "epoch": 10.38, "learning_rate": 9.877926870265335e-06, "loss": 0.0563, "step": 33734 }, { "epoch": 10.38, "learning_rate": 9.877428637496035e-06, "loss": 0.0561, "step": 33735 }, { "epoch": 10.38, "learning_rate": 9.876930405031047e-06, "loss": 0.0614, "step": 33736 }, { "epoch": 10.38, "learning_rate": 9.87643217287161e-06, "loss": 0.0549, "step": 33737 }, { "epoch": 10.38, "learning_rate": 9.875933941018956e-06, "loss": 0.0704, "step": 33738 }, { "epoch": 10.38, "learning_rate": 9.875435709474328e-06, "loss": 0.0638, "step": 33739 }, { "epoch": 10.38, "learning_rate": 9.874937478238958e-06, "loss": 0.025, "step": 33740 }, { "epoch": 10.38, "learning_rate": 9.87443924731408e-06, "loss": 0.0632, "step": 33741 }, { "epoch": 10.38, "learning_rate": 9.873941016700941e-06, "loss": 0.0598, "step": 33742 }, { "epoch": 10.38, "learning_rate": 9.873442786400771e-06, "loss": 0.0513, "step": 33743 }, { "epoch": 10.38, "learning_rate": 9.872944556414808e-06, "loss": 0.039, "step": 33744 }, { "epoch": 10.38, "learning_rate": 9.87244632674429e-06, "loss": 0.0804, "step": 33745 }, { "epoch": 10.38, "learning_rate": 9.871948097390451e-06, "loss": 0.0537, "step": 33746 }, { "epoch": 10.38, "learning_rate": 9.871449868354533e-06, "loss": 0.0586, "step": 33747 }, { "epoch": 10.38, "learning_rate": 9.870951639637769e-06, "loss": 0.0634, "step": 33748 }, { "epoch": 10.38, "learning_rate": 9.870453411241399e-06, "loss": 0.0442, "step": 33749 }, { "epoch": 10.38, "learning_rate": 9.869955183166659e-06, "loss": 0.0363, "step": 33750 }, { "epoch": 10.38, "learning_rate": 9.86945695541478e-06, "loss": 0.0485, "step": 33751 }, { "epoch": 10.39, "learning_rate": 9.868958727987007e-06, "loss": 0.0715, "step": 33752 }, { "epoch": 10.39, "learning_rate": 9.868460500884578e-06, "loss": 0.0657, "step": 33753 }, { "epoch": 10.39, "learning_rate": 9.867962274108722e-06, "loss": 0.0773, "step": 33754 }, { "epoch": 10.39, "learning_rate": 9.86746404766068e-06, "loss": 0.0355, "step": 33755 }, { "epoch": 10.39, "learning_rate": 9.86696582154169e-06, "loss": 0.0555, "step": 33756 }, { "epoch": 10.39, "learning_rate": 9.866467595752987e-06, "loss": 0.0309, "step": 33757 }, { "epoch": 10.39, "learning_rate": 9.86596937029581e-06, "loss": 0.0435, "step": 33758 }, { "epoch": 10.39, "learning_rate": 9.865471145171396e-06, "loss": 0.0468, "step": 33759 }, { "epoch": 10.39, "learning_rate": 9.864972920380982e-06, "loss": 0.064, "step": 33760 }, { "epoch": 10.39, "learning_rate": 9.864474695925797e-06, "loss": 0.0877, "step": 33761 }, { "epoch": 10.39, "learning_rate": 9.86397647180709e-06, "loss": 0.071, "step": 33762 }, { "epoch": 10.39, "learning_rate": 9.863478248026093e-06, "loss": 0.052, "step": 33763 }, { "epoch": 10.39, "learning_rate": 9.862980024584044e-06, "loss": 0.0214, "step": 33764 }, { "epoch": 10.39, "learning_rate": 9.862481801482176e-06, "loss": 0.0705, "step": 33765 }, { "epoch": 10.39, "learning_rate": 9.861983578721727e-06, "loss": 0.0355, "step": 33766 }, { "epoch": 10.39, "learning_rate": 9.861485356303938e-06, "loss": 0.042, "step": 33767 }, { "epoch": 10.39, "learning_rate": 9.860987134230043e-06, "loss": 0.0558, "step": 33768 }, { "epoch": 10.39, "learning_rate": 9.86048891250128e-06, "loss": 0.0597, "step": 33769 }, { "epoch": 10.39, "learning_rate": 9.859990691118884e-06, "loss": 0.0381, "step": 33770 }, { "epoch": 10.39, "learning_rate": 9.859492470084093e-06, "loss": 0.0435, "step": 33771 }, { "epoch": 10.39, "learning_rate": 9.858994249398147e-06, "loss": 0.0817, "step": 33772 }, { "epoch": 10.39, "learning_rate": 9.85849602906228e-06, "loss": 0.0556, "step": 33773 }, { "epoch": 10.39, "learning_rate": 9.857997809077729e-06, "loss": 0.0375, "step": 33774 }, { "epoch": 10.39, "learning_rate": 9.857499589445731e-06, "loss": 0.0421, "step": 33775 }, { "epoch": 10.39, "learning_rate": 9.85700137016752e-06, "loss": 0.053, "step": 33776 }, { "epoch": 10.39, "learning_rate": 9.856503151244339e-06, "loss": 0.0647, "step": 33777 }, { "epoch": 10.39, "learning_rate": 9.856004932677422e-06, "loss": 0.0725, "step": 33778 }, { "epoch": 10.39, "learning_rate": 9.855506714468006e-06, "loss": 0.059, "step": 33779 }, { "epoch": 10.39, "learning_rate": 9.855008496617326e-06, "loss": 0.0586, "step": 33780 }, { "epoch": 10.39, "learning_rate": 9.85451027912662e-06, "loss": 0.0531, "step": 33781 }, { "epoch": 10.39, "learning_rate": 9.854012061997128e-06, "loss": 0.0446, "step": 33782 }, { "epoch": 10.39, "learning_rate": 9.853513845230085e-06, "loss": 0.0634, "step": 33783 }, { "epoch": 10.4, "learning_rate": 9.85301562882673e-06, "loss": 0.037, "step": 33784 }, { "epoch": 10.4, "learning_rate": 9.852517412788296e-06, "loss": 0.0201, "step": 33785 }, { "epoch": 10.4, "learning_rate": 9.852019197116015e-06, "loss": 0.0534, "step": 33786 }, { "epoch": 10.4, "learning_rate": 9.85152098181114e-06, "loss": 0.061, "step": 33787 }, { "epoch": 10.4, "learning_rate": 9.851022766874892e-06, "loss": 0.0785, "step": 33788 }, { "epoch": 10.4, "learning_rate": 9.850524552308517e-06, "loss": 0.055, "step": 33789 }, { "epoch": 10.4, "learning_rate": 9.850026338113248e-06, "loss": 0.0401, "step": 33790 }, { "epoch": 10.4, "learning_rate": 9.849528124290322e-06, "loss": 0.0211, "step": 33791 }, { "epoch": 10.4, "learning_rate": 9.84902991084098e-06, "loss": 0.082, "step": 33792 }, { "epoch": 10.4, "learning_rate": 9.848531697766455e-06, "loss": 0.0805, "step": 33793 }, { "epoch": 10.4, "learning_rate": 9.848033485067984e-06, "loss": 0.0682, "step": 33794 }, { "epoch": 10.4, "learning_rate": 9.84753527274681e-06, "loss": 0.0931, "step": 33795 }, { "epoch": 10.4, "learning_rate": 9.847037060804155e-06, "loss": 0.0573, "step": 33796 }, { "epoch": 10.4, "learning_rate": 9.846538849241274e-06, "loss": 0.0275, "step": 33797 }, { "epoch": 10.4, "learning_rate": 9.846040638059394e-06, "loss": 0.0238, "step": 33798 }, { "epoch": 10.4, "learning_rate": 9.845542427259752e-06, "loss": 0.086, "step": 33799 }, { "epoch": 10.4, "learning_rate": 9.845044216843587e-06, "loss": 0.0753, "step": 33800 }, { "epoch": 10.4, "learning_rate": 9.844546006812135e-06, "loss": 0.0548, "step": 33801 }, { "epoch": 10.4, "learning_rate": 9.844047797166633e-06, "loss": 0.0426, "step": 33802 }, { "epoch": 10.4, "learning_rate": 9.84354958790832e-06, "loss": 0.0805, "step": 33803 }, { "epoch": 10.4, "learning_rate": 9.843051379038431e-06, "loss": 0.0378, "step": 33804 }, { "epoch": 10.4, "learning_rate": 9.842553170558205e-06, "loss": 0.0488, "step": 33805 }, { "epoch": 10.4, "learning_rate": 9.84205496246887e-06, "loss": 0.0422, "step": 33806 }, { "epoch": 10.4, "learning_rate": 9.841556754771675e-06, "loss": 0.0474, "step": 33807 }, { "epoch": 10.4, "learning_rate": 9.841058547467853e-06, "loss": 0.0599, "step": 33808 }, { "epoch": 10.4, "learning_rate": 9.840560340558639e-06, "loss": 0.0935, "step": 33809 }, { "epoch": 10.4, "learning_rate": 9.84006213404527e-06, "loss": 0.0399, "step": 33810 }, { "epoch": 10.4, "learning_rate": 9.839563927928982e-06, "loss": 0.0586, "step": 33811 }, { "epoch": 10.4, "learning_rate": 9.839065722211014e-06, "loss": 0.1198, "step": 33812 }, { "epoch": 10.4, "learning_rate": 9.838567516892604e-06, "loss": 0.0655, "step": 33813 }, { "epoch": 10.4, "learning_rate": 9.838069311974986e-06, "loss": 0.066, "step": 33814 }, { "epoch": 10.4, "learning_rate": 9.8375711074594e-06, "loss": 0.0432, "step": 33815 }, { "epoch": 10.4, "learning_rate": 9.837072903347078e-06, "loss": 0.0576, "step": 33816 }, { "epoch": 10.41, "learning_rate": 9.836574699639262e-06, "loss": 0.0621, "step": 33817 }, { "epoch": 10.41, "learning_rate": 9.836076496337189e-06, "loss": 0.0826, "step": 33818 }, { "epoch": 10.41, "learning_rate": 9.835578293442092e-06, "loss": 0.061, "step": 33819 }, { "epoch": 10.41, "learning_rate": 9.83508009095521e-06, "loss": 0.063, "step": 33820 }, { "epoch": 10.41, "learning_rate": 9.834581888877778e-06, "loss": 0.0756, "step": 33821 }, { "epoch": 10.41, "learning_rate": 9.834083687211035e-06, "loss": 0.0214, "step": 33822 }, { "epoch": 10.41, "learning_rate": 9.83358548595622e-06, "loss": 0.06, "step": 33823 }, { "epoch": 10.41, "learning_rate": 9.833087285114565e-06, "loss": 0.0739, "step": 33824 }, { "epoch": 10.41, "learning_rate": 9.83258908468731e-06, "loss": 0.0678, "step": 33825 }, { "epoch": 10.41, "learning_rate": 9.83209088467569e-06, "loss": 0.0819, "step": 33826 }, { "epoch": 10.41, "learning_rate": 9.831592685080945e-06, "loss": 0.0542, "step": 33827 }, { "epoch": 10.41, "learning_rate": 9.83109448590431e-06, "loss": 0.0672, "step": 33828 }, { "epoch": 10.41, "learning_rate": 9.830596287147023e-06, "loss": 0.0476, "step": 33829 }, { "epoch": 10.41, "learning_rate": 9.830098088810317e-06, "loss": 0.0461, "step": 33830 }, { "epoch": 10.41, "learning_rate": 9.829599890895429e-06, "loss": 0.036, "step": 33831 }, { "epoch": 10.41, "learning_rate": 9.829101693403605e-06, "loss": 0.0745, "step": 33832 }, { "epoch": 10.41, "learning_rate": 9.828603496336072e-06, "loss": 0.0702, "step": 33833 }, { "epoch": 10.41, "learning_rate": 9.828105299694072e-06, "loss": 0.0437, "step": 33834 }, { "epoch": 10.41, "learning_rate": 9.827607103478838e-06, "loss": 0.0485, "step": 33835 }, { "epoch": 10.41, "learning_rate": 9.827108907691609e-06, "loss": 0.0568, "step": 33836 }, { "epoch": 10.41, "learning_rate": 9.826610712333623e-06, "loss": 0.0555, "step": 33837 }, { "epoch": 10.41, "learning_rate": 9.826112517406115e-06, "loss": 0.0704, "step": 33838 }, { "epoch": 10.41, "learning_rate": 9.825614322910328e-06, "loss": 0.0999, "step": 33839 }, { "epoch": 10.41, "learning_rate": 9.825116128847488e-06, "loss": 0.1119, "step": 33840 }, { "epoch": 10.41, "learning_rate": 9.824617935218835e-06, "loss": 0.0382, "step": 33841 }, { "epoch": 10.41, "learning_rate": 9.824119742025615e-06, "loss": 0.0585, "step": 33842 }, { "epoch": 10.41, "learning_rate": 9.823621549269056e-06, "loss": 0.037, "step": 33843 }, { "epoch": 10.41, "learning_rate": 9.823123356950396e-06, "loss": 0.0786, "step": 33844 }, { "epoch": 10.41, "learning_rate": 9.822625165070873e-06, "loss": 0.056, "step": 33845 }, { "epoch": 10.41, "learning_rate": 9.822126973631723e-06, "loss": 0.0703, "step": 33846 }, { "epoch": 10.41, "learning_rate": 9.821628782634187e-06, "loss": 0.0919, "step": 33847 }, { "epoch": 10.41, "learning_rate": 9.821130592079496e-06, "loss": 0.0362, "step": 33848 }, { "epoch": 10.42, "learning_rate": 9.82063240196889e-06, "loss": 0.0416, "step": 33849 }, { "epoch": 10.42, "learning_rate": 9.82013421230361e-06, "loss": 0.0294, "step": 33850 }, { "epoch": 10.42, "learning_rate": 9.819636023084879e-06, "loss": 0.0346, "step": 33851 }, { "epoch": 10.42, "learning_rate": 9.819137834313948e-06, "loss": 0.0471, "step": 33852 }, { "epoch": 10.42, "learning_rate": 9.818639645992052e-06, "loss": 0.0473, "step": 33853 }, { "epoch": 10.42, "learning_rate": 9.818141458120423e-06, "loss": 0.0622, "step": 33854 }, { "epoch": 10.42, "learning_rate": 9.817643270700298e-06, "loss": 0.0734, "step": 33855 }, { "epoch": 10.42, "learning_rate": 9.817145083732914e-06, "loss": 0.0416, "step": 33856 }, { "epoch": 10.42, "learning_rate": 9.816646897219513e-06, "loss": 0.0928, "step": 33857 }, { "epoch": 10.42, "learning_rate": 9.816148711161328e-06, "loss": 0.0637, "step": 33858 }, { "epoch": 10.42, "learning_rate": 9.815650525559595e-06, "loss": 0.0455, "step": 33859 }, { "epoch": 10.42, "learning_rate": 9.815152340415554e-06, "loss": 0.0706, "step": 33860 }, { "epoch": 10.42, "learning_rate": 9.814654155730435e-06, "loss": 0.0633, "step": 33861 }, { "epoch": 10.42, "learning_rate": 9.814155971505483e-06, "loss": 0.0354, "step": 33862 }, { "epoch": 10.42, "learning_rate": 9.813657787741933e-06, "loss": 0.0545, "step": 33863 }, { "epoch": 10.42, "learning_rate": 9.813159604441021e-06, "loss": 0.0697, "step": 33864 }, { "epoch": 10.42, "learning_rate": 9.81266142160398e-06, "loss": 0.0378, "step": 33865 }, { "epoch": 10.42, "learning_rate": 9.812163239232051e-06, "loss": 0.0401, "step": 33866 }, { "epoch": 10.42, "learning_rate": 9.81166505732647e-06, "loss": 0.0762, "step": 33867 }, { "epoch": 10.42, "learning_rate": 9.811166875888475e-06, "loss": 0.0624, "step": 33868 }, { "epoch": 10.42, "learning_rate": 9.810668694919303e-06, "loss": 0.047, "step": 33869 }, { "epoch": 10.42, "learning_rate": 9.810170514420188e-06, "loss": 0.0309, "step": 33870 }, { "epoch": 10.42, "learning_rate": 9.809672334392367e-06, "loss": 0.0584, "step": 33871 }, { "epoch": 10.42, "learning_rate": 9.809174154837079e-06, "loss": 0.0415, "step": 33872 }, { "epoch": 10.42, "learning_rate": 9.80867597575556e-06, "loss": 0.0611, "step": 33873 }, { "epoch": 10.42, "learning_rate": 9.808177797149052e-06, "loss": 0.0488, "step": 33874 }, { "epoch": 10.42, "learning_rate": 9.807679619018781e-06, "loss": 0.0515, "step": 33875 }, { "epoch": 10.42, "learning_rate": 9.80718144136599e-06, "loss": 0.0912, "step": 33876 }, { "epoch": 10.42, "learning_rate": 9.806683264191916e-06, "loss": 0.0738, "step": 33877 }, { "epoch": 10.42, "learning_rate": 9.806185087497797e-06, "loss": 0.0599, "step": 33878 }, { "epoch": 10.42, "learning_rate": 9.805686911284867e-06, "loss": 0.0847, "step": 33879 }, { "epoch": 10.42, "learning_rate": 9.805188735554364e-06, "loss": 0.0668, "step": 33880 }, { "epoch": 10.42, "learning_rate": 9.804690560307525e-06, "loss": 0.0663, "step": 33881 }, { "epoch": 10.43, "learning_rate": 9.804192385545586e-06, "loss": 0.0456, "step": 33882 }, { "epoch": 10.43, "learning_rate": 9.803694211269786e-06, "loss": 0.0304, "step": 33883 }, { "epoch": 10.43, "learning_rate": 9.803196037481362e-06, "loss": 0.089, "step": 33884 }, { "epoch": 10.43, "learning_rate": 9.802697864181546e-06, "loss": 0.0315, "step": 33885 }, { "epoch": 10.43, "learning_rate": 9.802199691371576e-06, "loss": 0.0473, "step": 33886 }, { "epoch": 10.43, "learning_rate": 9.801701519052695e-06, "loss": 0.0426, "step": 33887 }, { "epoch": 10.43, "learning_rate": 9.801203347226136e-06, "loss": 0.1029, "step": 33888 }, { "epoch": 10.43, "learning_rate": 9.800705175893133e-06, "loss": 0.0734, "step": 33889 }, { "epoch": 10.43, "learning_rate": 9.800207005054927e-06, "loss": 0.055, "step": 33890 }, { "epoch": 10.43, "learning_rate": 9.79970883471275e-06, "loss": 0.0725, "step": 33891 }, { "epoch": 10.43, "learning_rate": 9.799210664867844e-06, "loss": 0.0666, "step": 33892 }, { "epoch": 10.43, "learning_rate": 9.798712495521444e-06, "loss": 0.061, "step": 33893 }, { "epoch": 10.43, "learning_rate": 9.798214326674787e-06, "loss": 0.0391, "step": 33894 }, { "epoch": 10.43, "learning_rate": 9.797716158329113e-06, "loss": 0.0664, "step": 33895 }, { "epoch": 10.43, "learning_rate": 9.797217990485647e-06, "loss": 0.1115, "step": 33896 }, { "epoch": 10.43, "learning_rate": 9.79671982314564e-06, "loss": 0.0381, "step": 33897 }, { "epoch": 10.43, "learning_rate": 9.796221656310322e-06, "loss": 0.0565, "step": 33898 }, { "epoch": 10.43, "learning_rate": 9.795723489980931e-06, "loss": 0.047, "step": 33899 }, { "epoch": 10.43, "learning_rate": 9.795225324158702e-06, "loss": 0.0528, "step": 33900 }, { "epoch": 10.43, "learning_rate": 9.794727158844871e-06, "loss": 0.0583, "step": 33901 }, { "epoch": 10.43, "learning_rate": 9.79422899404068e-06, "loss": 0.0621, "step": 33902 }, { "epoch": 10.43, "learning_rate": 9.793730829747364e-06, "loss": 0.063, "step": 33903 }, { "epoch": 10.43, "learning_rate": 9.793232665966159e-06, "loss": 0.0377, "step": 33904 }, { "epoch": 10.43, "learning_rate": 9.792734502698301e-06, "loss": 0.0723, "step": 33905 }, { "epoch": 10.43, "learning_rate": 9.792236339945023e-06, "loss": 0.0354, "step": 33906 }, { "epoch": 10.43, "learning_rate": 9.791738177707569e-06, "loss": 0.0421, "step": 33907 }, { "epoch": 10.43, "learning_rate": 9.791240015987177e-06, "loss": 0.0386, "step": 33908 }, { "epoch": 10.43, "learning_rate": 9.790741854785076e-06, "loss": 0.0509, "step": 33909 }, { "epoch": 10.43, "learning_rate": 9.790243694102506e-06, "loss": 0.042, "step": 33910 }, { "epoch": 10.43, "learning_rate": 9.789745533940705e-06, "loss": 0.0384, "step": 33911 }, { "epoch": 10.43, "learning_rate": 9.78924737430091e-06, "loss": 0.033, "step": 33912 }, { "epoch": 10.43, "learning_rate": 9.788749215184355e-06, "loss": 0.0327, "step": 33913 }, { "epoch": 10.44, "learning_rate": 9.788251056592282e-06, "loss": 0.0495, "step": 33914 }, { "epoch": 10.44, "learning_rate": 9.787752898525922e-06, "loss": 0.0514, "step": 33915 }, { "epoch": 10.44, "learning_rate": 9.787254740986514e-06, "loss": 0.0906, "step": 33916 }, { "epoch": 10.44, "learning_rate": 9.786756583975296e-06, "loss": 0.0499, "step": 33917 }, { "epoch": 10.44, "learning_rate": 9.786258427493505e-06, "loss": 0.0631, "step": 33918 }, { "epoch": 10.44, "learning_rate": 9.785760271542377e-06, "loss": 0.0691, "step": 33919 }, { "epoch": 10.44, "learning_rate": 9.785262116123147e-06, "loss": 0.0375, "step": 33920 }, { "epoch": 10.44, "learning_rate": 9.784763961237053e-06, "loss": 0.0617, "step": 33921 }, { "epoch": 10.44, "learning_rate": 9.784265806885332e-06, "loss": 0.0663, "step": 33922 }, { "epoch": 10.44, "learning_rate": 9.783767653069222e-06, "loss": 0.0696, "step": 33923 }, { "epoch": 10.44, "learning_rate": 9.783269499789958e-06, "loss": 0.0557, "step": 33924 }, { "epoch": 10.44, "learning_rate": 9.782771347048776e-06, "loss": 0.0353, "step": 33925 }, { "epoch": 10.44, "learning_rate": 9.782273194846915e-06, "loss": 0.0743, "step": 33926 }, { "epoch": 10.44, "learning_rate": 9.781775043185612e-06, "loss": 0.0639, "step": 33927 }, { "epoch": 10.44, "learning_rate": 9.781276892066103e-06, "loss": 0.0602, "step": 33928 }, { "epoch": 10.44, "learning_rate": 9.780778741489626e-06, "loss": 0.0487, "step": 33929 }, { "epoch": 10.44, "learning_rate": 9.780280591457414e-06, "loss": 0.0587, "step": 33930 }, { "epoch": 10.44, "learning_rate": 9.779782441970702e-06, "loss": 0.0464, "step": 33931 }, { "epoch": 10.44, "learning_rate": 9.779284293030737e-06, "loss": 0.0359, "step": 33932 }, { "epoch": 10.44, "learning_rate": 9.778786144638748e-06, "loss": 0.0433, "step": 33933 }, { "epoch": 10.44, "learning_rate": 9.778287996795974e-06, "loss": 0.0182, "step": 33934 }, { "epoch": 10.44, "learning_rate": 9.77778984950365e-06, "loss": 0.0388, "step": 33935 }, { "epoch": 10.44, "learning_rate": 9.777291702763012e-06, "loss": 0.0573, "step": 33936 }, { "epoch": 10.44, "learning_rate": 9.7767935565753e-06, "loss": 0.0787, "step": 33937 }, { "epoch": 10.44, "learning_rate": 9.776295410941752e-06, "loss": 0.0578, "step": 33938 }, { "epoch": 10.44, "learning_rate": 9.775797265863602e-06, "loss": 0.075, "step": 33939 }, { "epoch": 10.44, "learning_rate": 9.775299121342086e-06, "loss": 0.0974, "step": 33940 }, { "epoch": 10.44, "learning_rate": 9.774800977378436e-06, "loss": 0.0476, "step": 33941 }, { "epoch": 10.44, "learning_rate": 9.774302833973903e-06, "loss": 0.0298, "step": 33942 }, { "epoch": 10.44, "learning_rate": 9.773804691129712e-06, "loss": 0.0334, "step": 33943 }, { "epoch": 10.44, "learning_rate": 9.773306548847102e-06, "loss": 0.0748, "step": 33944 }, { "epoch": 10.44, "learning_rate": 9.77280840712731e-06, "loss": 0.0828, "step": 33945 }, { "epoch": 10.44, "learning_rate": 9.772310265971575e-06, "loss": 0.0179, "step": 33946 }, { "epoch": 10.45, "learning_rate": 9.771812125381132e-06, "loss": 0.0692, "step": 33947 }, { "epoch": 10.45, "learning_rate": 9.771313985357217e-06, "loss": 0.0548, "step": 33948 }, { "epoch": 10.45, "learning_rate": 9.77081584590107e-06, "loss": 0.0721, "step": 33949 }, { "epoch": 10.45, "learning_rate": 9.770317707013927e-06, "loss": 0.0741, "step": 33950 }, { "epoch": 10.45, "learning_rate": 9.769819568697016e-06, "loss": 0.0512, "step": 33951 }, { "epoch": 10.45, "learning_rate": 9.769321430951585e-06, "loss": 0.0309, "step": 33952 }, { "epoch": 10.45, "learning_rate": 9.76882329377887e-06, "loss": 0.0714, "step": 33953 }, { "epoch": 10.45, "learning_rate": 9.7683251571801e-06, "loss": 0.0483, "step": 33954 }, { "epoch": 10.45, "learning_rate": 9.767827021156519e-06, "loss": 0.0608, "step": 33955 }, { "epoch": 10.45, "learning_rate": 9.767328885709358e-06, "loss": 0.0428, "step": 33956 }, { "epoch": 10.45, "learning_rate": 9.766830750839857e-06, "loss": 0.0484, "step": 33957 }, { "epoch": 10.45, "learning_rate": 9.766332616549254e-06, "loss": 0.0588, "step": 33958 }, { "epoch": 10.45, "learning_rate": 9.765834482838785e-06, "loss": 0.0558, "step": 33959 }, { "epoch": 10.45, "learning_rate": 9.765336349709687e-06, "loss": 0.0606, "step": 33960 }, { "epoch": 10.45, "learning_rate": 9.764838217163189e-06, "loss": 0.0471, "step": 33961 }, { "epoch": 10.45, "learning_rate": 9.764340085200539e-06, "loss": 0.0367, "step": 33962 }, { "epoch": 10.45, "learning_rate": 9.76384195382297e-06, "loss": 0.0486, "step": 33963 }, { "epoch": 10.45, "learning_rate": 9.763343823031718e-06, "loss": 0.1032, "step": 33964 }, { "epoch": 10.45, "learning_rate": 9.762845692828016e-06, "loss": 0.0484, "step": 33965 }, { "epoch": 10.45, "learning_rate": 9.762347563213105e-06, "loss": 0.0344, "step": 33966 }, { "epoch": 10.45, "learning_rate": 9.761849434188222e-06, "loss": 0.0363, "step": 33967 }, { "epoch": 10.45, "learning_rate": 9.761351305754603e-06, "loss": 0.045, "step": 33968 }, { "epoch": 10.45, "learning_rate": 9.760853177913483e-06, "loss": 0.0528, "step": 33969 }, { "epoch": 10.45, "learning_rate": 9.760355050666102e-06, "loss": 0.0734, "step": 33970 }, { "epoch": 10.45, "learning_rate": 9.759856924013693e-06, "loss": 0.0513, "step": 33971 }, { "epoch": 10.45, "learning_rate": 9.759358797957495e-06, "loss": 0.0773, "step": 33972 }, { "epoch": 10.45, "learning_rate": 9.758860672498745e-06, "loss": 0.0769, "step": 33973 }, { "epoch": 10.45, "learning_rate": 9.75836254763868e-06, "loss": 0.0586, "step": 33974 }, { "epoch": 10.45, "learning_rate": 9.757864423378534e-06, "loss": 0.0743, "step": 33975 }, { "epoch": 10.45, "learning_rate": 9.757366299719543e-06, "loss": 0.074, "step": 33976 }, { "epoch": 10.45, "learning_rate": 9.75686817666295e-06, "loss": 0.0592, "step": 33977 }, { "epoch": 10.45, "learning_rate": 9.756370054209987e-06, "loss": 0.095, "step": 33978 }, { "epoch": 10.46, "learning_rate": 9.75587193236189e-06, "loss": 0.0335, "step": 33979 }, { "epoch": 10.46, "learning_rate": 9.755373811119898e-06, "loss": 0.0558, "step": 33980 }, { "epoch": 10.46, "learning_rate": 9.754875690485247e-06, "loss": 0.041, "step": 33981 }, { "epoch": 10.46, "learning_rate": 9.754377570459173e-06, "loss": 0.0586, "step": 33982 }, { "epoch": 10.46, "learning_rate": 9.753879451042914e-06, "loss": 0.0915, "step": 33983 }, { "epoch": 10.46, "learning_rate": 9.753381332237706e-06, "loss": 0.0622, "step": 33984 }, { "epoch": 10.46, "learning_rate": 9.752883214044787e-06, "loss": 0.0239, "step": 33985 }, { "epoch": 10.46, "learning_rate": 9.752385096465386e-06, "loss": 0.0769, "step": 33986 }, { "epoch": 10.46, "learning_rate": 9.751886979500752e-06, "loss": 0.0326, "step": 33987 }, { "epoch": 10.46, "learning_rate": 9.751388863152116e-06, "loss": 0.1019, "step": 33988 }, { "epoch": 10.46, "learning_rate": 9.750890747420713e-06, "loss": 0.0583, "step": 33989 }, { "epoch": 10.46, "learning_rate": 9.750392632307782e-06, "loss": 0.0513, "step": 33990 }, { "epoch": 10.46, "learning_rate": 9.749894517814556e-06, "loss": 0.0416, "step": 33991 }, { "epoch": 10.46, "learning_rate": 9.749396403942277e-06, "loss": 0.0662, "step": 33992 }, { "epoch": 10.46, "learning_rate": 9.748898290692179e-06, "loss": 0.0759, "step": 33993 }, { "epoch": 10.46, "learning_rate": 9.748400178065499e-06, "loss": 0.0467, "step": 33994 }, { "epoch": 10.46, "learning_rate": 9.747902066063473e-06, "loss": 0.0546, "step": 33995 }, { "epoch": 10.46, "learning_rate": 9.747403954687334e-06, "loss": 0.0544, "step": 33996 }, { "epoch": 10.46, "learning_rate": 9.74690584393833e-06, "loss": 0.058, "step": 33997 }, { "epoch": 10.46, "learning_rate": 9.746407733817688e-06, "loss": 0.0613, "step": 33998 }, { "epoch": 10.46, "learning_rate": 9.745909624326648e-06, "loss": 0.1073, "step": 33999 }, { "epoch": 10.46, "learning_rate": 9.745411515466444e-06, "loss": 0.0384, "step": 34000 }, { "epoch": 10.46, "learning_rate": 9.744913407238314e-06, "loss": 0.0257, "step": 34001 }, { "epoch": 10.46, "learning_rate": 9.744415299643496e-06, "loss": 0.0583, "step": 34002 }, { "epoch": 10.46, "learning_rate": 9.743917192683226e-06, "loss": 0.0589, "step": 34003 }, { "epoch": 10.46, "learning_rate": 9.743419086358742e-06, "loss": 0.051, "step": 34004 }, { "epoch": 10.46, "learning_rate": 9.74292098067128e-06, "loss": 0.0422, "step": 34005 }, { "epoch": 10.46, "learning_rate": 9.74242287562207e-06, "loss": 0.0729, "step": 34006 }, { "epoch": 10.46, "learning_rate": 9.741924771212358e-06, "loss": 0.0508, "step": 34007 }, { "epoch": 10.46, "learning_rate": 9.741426667443381e-06, "loss": 0.065, "step": 34008 }, { "epoch": 10.46, "learning_rate": 9.740928564316369e-06, "loss": 0.1091, "step": 34009 }, { "epoch": 10.46, "learning_rate": 9.740430461832562e-06, "loss": 0.0649, "step": 34010 }, { "epoch": 10.46, "learning_rate": 9.739932359993193e-06, "loss": 0.0686, "step": 34011 }, { "epoch": 10.47, "learning_rate": 9.739434258799506e-06, "loss": 0.0506, "step": 34012 }, { "epoch": 10.47, "learning_rate": 9.73893615825273e-06, "loss": 0.046, "step": 34013 }, { "epoch": 10.47, "learning_rate": 9.738438058354108e-06, "loss": 0.04, "step": 34014 }, { "epoch": 10.47, "learning_rate": 9.737939959104875e-06, "loss": 0.0603, "step": 34015 }, { "epoch": 10.47, "learning_rate": 9.73744186050626e-06, "loss": 0.0438, "step": 34016 }, { "epoch": 10.47, "learning_rate": 9.736943762559511e-06, "loss": 0.0422, "step": 34017 }, { "epoch": 10.47, "learning_rate": 9.73644566526586e-06, "loss": 0.0623, "step": 34018 }, { "epoch": 10.47, "learning_rate": 9.735947568626544e-06, "loss": 0.0476, "step": 34019 }, { "epoch": 10.47, "learning_rate": 9.735449472642798e-06, "loss": 0.0516, "step": 34020 }, { "epoch": 10.47, "learning_rate": 9.734951377315856e-06, "loss": 0.0634, "step": 34021 }, { "epoch": 10.47, "learning_rate": 9.734453282646962e-06, "loss": 0.0663, "step": 34022 }, { "epoch": 10.47, "learning_rate": 9.733955188637347e-06, "loss": 0.0643, "step": 34023 }, { "epoch": 10.47, "learning_rate": 9.733457095288252e-06, "loss": 0.0664, "step": 34024 }, { "epoch": 10.47, "learning_rate": 9.732959002600909e-06, "loss": 0.0531, "step": 34025 }, { "epoch": 10.47, "learning_rate": 9.732460910576557e-06, "loss": 0.0526, "step": 34026 }, { "epoch": 10.47, "learning_rate": 9.731962819216433e-06, "loss": 0.0489, "step": 34027 }, { "epoch": 10.47, "learning_rate": 9.731464728521774e-06, "loss": 0.071, "step": 34028 }, { "epoch": 10.47, "learning_rate": 9.730966638493816e-06, "loss": 0.0274, "step": 34029 }, { "epoch": 10.47, "learning_rate": 9.730468549133794e-06, "loss": 0.0616, "step": 34030 }, { "epoch": 10.47, "learning_rate": 9.729970460442942e-06, "loss": 0.0474, "step": 34031 }, { "epoch": 10.47, "learning_rate": 9.729472372422507e-06, "loss": 0.051, "step": 34032 }, { "epoch": 10.47, "learning_rate": 9.728974285073717e-06, "loss": 0.0475, "step": 34033 }, { "epoch": 10.47, "learning_rate": 9.728476198397811e-06, "loss": 0.0621, "step": 34034 }, { "epoch": 10.47, "learning_rate": 9.727978112396026e-06, "loss": 0.0784, "step": 34035 }, { "epoch": 10.47, "learning_rate": 9.727480027069594e-06, "loss": 0.0275, "step": 34036 }, { "epoch": 10.47, "learning_rate": 9.72698194241976e-06, "loss": 0.0462, "step": 34037 }, { "epoch": 10.47, "learning_rate": 9.726483858447754e-06, "loss": 0.0748, "step": 34038 }, { "epoch": 10.47, "learning_rate": 9.725985775154818e-06, "loss": 0.061, "step": 34039 }, { "epoch": 10.47, "learning_rate": 9.725487692542184e-06, "loss": 0.1171, "step": 34040 }, { "epoch": 10.47, "learning_rate": 9.724989610611085e-06, "loss": 0.0503, "step": 34041 }, { "epoch": 10.47, "learning_rate": 9.72449152936277e-06, "loss": 0.0681, "step": 34042 }, { "epoch": 10.47, "learning_rate": 9.723993448798464e-06, "loss": 0.074, "step": 34043 }, { "epoch": 10.48, "learning_rate": 9.72349536891941e-06, "loss": 0.0726, "step": 34044 }, { "epoch": 10.48, "learning_rate": 9.72299728972684e-06, "loss": 0.0709, "step": 34045 }, { "epoch": 10.48, "learning_rate": 9.722499211221993e-06, "loss": 0.0557, "step": 34046 }, { "epoch": 10.48, "learning_rate": 9.722001133406107e-06, "loss": 0.0773, "step": 34047 }, { "epoch": 10.48, "learning_rate": 9.721503056280418e-06, "loss": 0.0309, "step": 34048 }, { "epoch": 10.48, "learning_rate": 9.721004979846161e-06, "loss": 0.0786, "step": 34049 }, { "epoch": 10.48, "learning_rate": 9.720506904104578e-06, "loss": 0.0594, "step": 34050 }, { "epoch": 10.48, "learning_rate": 9.720008829056893e-06, "loss": 0.0455, "step": 34051 }, { "epoch": 10.48, "learning_rate": 9.719510754704353e-06, "loss": 0.0391, "step": 34052 }, { "epoch": 10.48, "learning_rate": 9.719012681048196e-06, "loss": 0.0449, "step": 34053 }, { "epoch": 10.48, "learning_rate": 9.718514608089652e-06, "loss": 0.0359, "step": 34054 }, { "epoch": 10.48, "learning_rate": 9.71801653582996e-06, "loss": 0.0422, "step": 34055 }, { "epoch": 10.48, "learning_rate": 9.717518464270355e-06, "loss": 0.0475, "step": 34056 }, { "epoch": 10.48, "learning_rate": 9.717020393412079e-06, "loss": 0.0796, "step": 34057 }, { "epoch": 10.48, "learning_rate": 9.716522323256365e-06, "loss": 0.0232, "step": 34058 }, { "epoch": 10.48, "learning_rate": 9.716024253804448e-06, "loss": 0.0667, "step": 34059 }, { "epoch": 10.48, "learning_rate": 9.71552618505757e-06, "loss": 0.0361, "step": 34060 }, { "epoch": 10.48, "learning_rate": 9.715028117016955e-06, "loss": 0.0476, "step": 34061 }, { "epoch": 10.48, "learning_rate": 9.714530049683855e-06, "loss": 0.0573, "step": 34062 }, { "epoch": 10.48, "learning_rate": 9.714031983059502e-06, "loss": 0.0992, "step": 34063 }, { "epoch": 10.48, "learning_rate": 9.713533917145127e-06, "loss": 0.0731, "step": 34064 }, { "epoch": 10.48, "learning_rate": 9.713035851941972e-06, "loss": 0.0554, "step": 34065 }, { "epoch": 10.48, "learning_rate": 9.712537787451268e-06, "loss": 0.0682, "step": 34066 }, { "epoch": 10.48, "learning_rate": 9.712039723674257e-06, "loss": 0.0457, "step": 34067 }, { "epoch": 10.48, "learning_rate": 9.711541660612174e-06, "loss": 0.0792, "step": 34068 }, { "epoch": 10.48, "learning_rate": 9.711043598266255e-06, "loss": 0.0353, "step": 34069 }, { "epoch": 10.48, "learning_rate": 9.710545536637738e-06, "loss": 0.043, "step": 34070 }, { "epoch": 10.48, "learning_rate": 9.710047475727854e-06, "loss": 0.0833, "step": 34071 }, { "epoch": 10.48, "learning_rate": 9.709549415537848e-06, "loss": 0.0881, "step": 34072 }, { "epoch": 10.48, "learning_rate": 9.709051356068952e-06, "loss": 0.0651, "step": 34073 }, { "epoch": 10.48, "learning_rate": 9.708553297322407e-06, "loss": 0.1267, "step": 34074 }, { "epoch": 10.48, "learning_rate": 9.708055239299442e-06, "loss": 0.0519, "step": 34075 }, { "epoch": 10.48, "learning_rate": 9.707557182001294e-06, "loss": 0.0495, "step": 34076 }, { "epoch": 10.49, "learning_rate": 9.707059125429206e-06, "loss": 0.0653, "step": 34077 }, { "epoch": 10.49, "learning_rate": 9.70656106958441e-06, "loss": 0.0733, "step": 34078 }, { "epoch": 10.49, "learning_rate": 9.706063014468145e-06, "loss": 0.0603, "step": 34079 }, { "epoch": 10.49, "learning_rate": 9.705564960081646e-06, "loss": 0.046, "step": 34080 }, { "epoch": 10.49, "learning_rate": 9.705066906426147e-06, "loss": 0.0507, "step": 34081 }, { "epoch": 10.49, "learning_rate": 9.704568853502891e-06, "loss": 0.0704, "step": 34082 }, { "epoch": 10.49, "learning_rate": 9.70407080131311e-06, "loss": 0.0468, "step": 34083 }, { "epoch": 10.49, "learning_rate": 9.703572749858044e-06, "loss": 0.0473, "step": 34084 }, { "epoch": 10.49, "learning_rate": 9.703074699138919e-06, "loss": 0.0694, "step": 34085 }, { "epoch": 10.49, "learning_rate": 9.702576649156984e-06, "loss": 0.0562, "step": 34086 }, { "epoch": 10.49, "learning_rate": 9.702078599913475e-06, "loss": 0.066, "step": 34087 }, { "epoch": 10.49, "learning_rate": 9.701580551409622e-06, "loss": 0.0709, "step": 34088 }, { "epoch": 10.49, "learning_rate": 9.701082503646663e-06, "loss": 0.0495, "step": 34089 }, { "epoch": 10.49, "learning_rate": 9.700584456625834e-06, "loss": 0.1051, "step": 34090 }, { "epoch": 10.49, "learning_rate": 9.700086410348375e-06, "loss": 0.0854, "step": 34091 }, { "epoch": 10.49, "learning_rate": 9.69958836481552e-06, "loss": 0.0334, "step": 34092 }, { "epoch": 10.49, "learning_rate": 9.699090320028507e-06, "loss": 0.073, "step": 34093 }, { "epoch": 10.49, "learning_rate": 9.698592275988573e-06, "loss": 0.0633, "step": 34094 }, { "epoch": 10.49, "learning_rate": 9.698094232696947e-06, "loss": 0.0586, "step": 34095 }, { "epoch": 10.49, "learning_rate": 9.697596190154876e-06, "loss": 0.0633, "step": 34096 }, { "epoch": 10.49, "learning_rate": 9.697098148363594e-06, "loss": 0.0516, "step": 34097 }, { "epoch": 10.49, "learning_rate": 9.696600107324334e-06, "loss": 0.0387, "step": 34098 }, { "epoch": 10.49, "learning_rate": 9.696102067038334e-06, "loss": 0.0927, "step": 34099 }, { "epoch": 10.49, "learning_rate": 9.69560402750683e-06, "loss": 0.064, "step": 34100 }, { "epoch": 10.49, "learning_rate": 9.695105988731058e-06, "loss": 0.0538, "step": 34101 }, { "epoch": 10.49, "learning_rate": 9.694607950712258e-06, "loss": 0.0451, "step": 34102 }, { "epoch": 10.49, "learning_rate": 9.694109913451664e-06, "loss": 0.0463, "step": 34103 }, { "epoch": 10.49, "learning_rate": 9.693611876950512e-06, "loss": 0.0479, "step": 34104 }, { "epoch": 10.49, "learning_rate": 9.693113841210037e-06, "loss": 0.0396, "step": 34105 }, { "epoch": 10.49, "learning_rate": 9.69261580623148e-06, "loss": 0.0836, "step": 34106 }, { "epoch": 10.49, "learning_rate": 9.692117772016074e-06, "loss": 0.0405, "step": 34107 }, { "epoch": 10.49, "learning_rate": 9.69161973856506e-06, "loss": 0.0711, "step": 34108 }, { "epoch": 10.5, "learning_rate": 9.691121705879669e-06, "loss": 0.0739, "step": 34109 }, { "epoch": 10.5, "learning_rate": 9.690623673961134e-06, "loss": 0.072, "step": 34110 }, { "epoch": 10.5, "learning_rate": 9.690125642810706e-06, "loss": 0.1056, "step": 34111 }, { "epoch": 10.5, "learning_rate": 9.689627612429607e-06, "loss": 0.0885, "step": 34112 }, { "epoch": 10.5, "learning_rate": 9.689129582819081e-06, "loss": 0.0648, "step": 34113 }, { "epoch": 10.5, "learning_rate": 9.688631553980361e-06, "loss": 0.0244, "step": 34114 }, { "epoch": 10.5, "learning_rate": 9.688133525914685e-06, "loss": 0.0644, "step": 34115 }, { "epoch": 10.5, "learning_rate": 9.687635498623291e-06, "loss": 0.0757, "step": 34116 }, { "epoch": 10.5, "learning_rate": 9.687137472107412e-06, "loss": 0.0765, "step": 34117 }, { "epoch": 10.5, "learning_rate": 9.68663944636829e-06, "loss": 0.0422, "step": 34118 }, { "epoch": 10.5, "learning_rate": 9.686141421407154e-06, "loss": 0.0692, "step": 34119 }, { "epoch": 10.5, "learning_rate": 9.685643397225242e-06, "loss": 0.0671, "step": 34120 }, { "epoch": 10.5, "learning_rate": 9.6851453738238e-06, "loss": 0.0688, "step": 34121 }, { "epoch": 10.5, "learning_rate": 9.684647351204052e-06, "loss": 0.0939, "step": 34122 }, { "epoch": 10.5, "learning_rate": 9.68414932936724e-06, "loss": 0.0484, "step": 34123 }, { "epoch": 10.5, "learning_rate": 9.683651308314602e-06, "loss": 0.0623, "step": 34124 }, { "epoch": 10.5, "learning_rate": 9.68315328804737e-06, "loss": 0.0524, "step": 34125 }, { "epoch": 10.5, "learning_rate": 9.682655268566783e-06, "loss": 0.0558, "step": 34126 }, { "epoch": 10.5, "learning_rate": 9.682157249874078e-06, "loss": 0.0665, "step": 34127 }, { "epoch": 10.5, "learning_rate": 9.681659231970493e-06, "loss": 0.1043, "step": 34128 }, { "epoch": 10.5, "learning_rate": 9.681161214857262e-06, "loss": 0.0681, "step": 34129 }, { "epoch": 10.5, "learning_rate": 9.680663198535618e-06, "loss": 0.0541, "step": 34130 }, { "epoch": 10.5, "learning_rate": 9.680165183006803e-06, "loss": 0.0565, "step": 34131 }, { "epoch": 10.5, "learning_rate": 9.679667168272056e-06, "loss": 0.0627, "step": 34132 }, { "epoch": 10.5, "learning_rate": 9.679169154332605e-06, "loss": 0.0481, "step": 34133 }, { "epoch": 10.5, "learning_rate": 9.678671141189691e-06, "loss": 0.056, "step": 34134 }, { "epoch": 10.5, "learning_rate": 9.678173128844548e-06, "loss": 0.0699, "step": 34135 }, { "epoch": 10.5, "learning_rate": 9.677675117298416e-06, "loss": 0.0661, "step": 34136 }, { "epoch": 10.5, "learning_rate": 9.677177106552531e-06, "loss": 0.0609, "step": 34137 }, { "epoch": 10.5, "learning_rate": 9.676679096608127e-06, "loss": 0.036, "step": 34138 }, { "epoch": 10.5, "learning_rate": 9.676181087466444e-06, "loss": 0.0456, "step": 34139 }, { "epoch": 10.5, "learning_rate": 9.67568307912871e-06, "loss": 0.062, "step": 34140 }, { "epoch": 10.5, "learning_rate": 9.675185071596172e-06, "loss": 0.0565, "step": 34141 }, { "epoch": 10.51, "learning_rate": 9.674687064870062e-06, "loss": 0.0697, "step": 34142 }, { "epoch": 10.51, "learning_rate": 9.674189058951617e-06, "loss": 0.0748, "step": 34143 }, { "epoch": 10.51, "learning_rate": 9.673691053842072e-06, "loss": 0.0693, "step": 34144 }, { "epoch": 10.51, "learning_rate": 9.673193049542662e-06, "loss": 0.0617, "step": 34145 }, { "epoch": 10.51, "learning_rate": 9.672695046054626e-06, "loss": 0.0848, "step": 34146 }, { "epoch": 10.51, "learning_rate": 9.672197043379202e-06, "loss": 0.0174, "step": 34147 }, { "epoch": 10.51, "learning_rate": 9.671699041517623e-06, "loss": 0.0466, "step": 34148 }, { "epoch": 10.51, "learning_rate": 9.671201040471128e-06, "loss": 0.0499, "step": 34149 }, { "epoch": 10.51, "learning_rate": 9.67070304024095e-06, "loss": 0.0491, "step": 34150 }, { "epoch": 10.51, "learning_rate": 9.67020504082833e-06, "loss": 0.0465, "step": 34151 }, { "epoch": 10.51, "learning_rate": 9.669707042234502e-06, "loss": 0.0635, "step": 34152 }, { "epoch": 10.51, "learning_rate": 9.669209044460702e-06, "loss": 0.0605, "step": 34153 }, { "epoch": 10.51, "learning_rate": 9.668711047508167e-06, "loss": 0.1011, "step": 34154 }, { "epoch": 10.51, "learning_rate": 9.668213051378129e-06, "loss": 0.0674, "step": 34155 }, { "epoch": 10.51, "learning_rate": 9.667715056071831e-06, "loss": 0.0759, "step": 34156 }, { "epoch": 10.51, "learning_rate": 9.667217061590508e-06, "loss": 0.0639, "step": 34157 }, { "epoch": 10.51, "learning_rate": 9.666719067935395e-06, "loss": 0.0506, "step": 34158 }, { "epoch": 10.51, "learning_rate": 9.66622107510773e-06, "loss": 0.0732, "step": 34159 }, { "epoch": 10.51, "learning_rate": 9.665723083108744e-06, "loss": 0.0613, "step": 34160 }, { "epoch": 10.51, "learning_rate": 9.66522509193968e-06, "loss": 0.1285, "step": 34161 }, { "epoch": 10.51, "learning_rate": 9.664727101601773e-06, "loss": 0.0385, "step": 34162 }, { "epoch": 10.51, "learning_rate": 9.66422911209626e-06, "loss": 0.0434, "step": 34163 }, { "epoch": 10.51, "learning_rate": 9.663731123424373e-06, "loss": 0.0792, "step": 34164 }, { "epoch": 10.51, "learning_rate": 9.663233135587347e-06, "loss": 0.0522, "step": 34165 }, { "epoch": 10.51, "learning_rate": 9.662735148586429e-06, "loss": 0.0463, "step": 34166 }, { "epoch": 10.51, "learning_rate": 9.662237162422845e-06, "loss": 0.0952, "step": 34167 }, { "epoch": 10.51, "learning_rate": 9.661739177097836e-06, "loss": 0.1032, "step": 34168 }, { "epoch": 10.51, "learning_rate": 9.661241192612638e-06, "loss": 0.0892, "step": 34169 }, { "epoch": 10.51, "learning_rate": 9.660743208968484e-06, "loss": 0.0805, "step": 34170 }, { "epoch": 10.51, "learning_rate": 9.660245226166615e-06, "loss": 0.0363, "step": 34171 }, { "epoch": 10.51, "learning_rate": 9.659747244208267e-06, "loss": 0.0483, "step": 34172 }, { "epoch": 10.51, "learning_rate": 9.659249263094677e-06, "loss": 0.0611, "step": 34173 }, { "epoch": 10.52, "learning_rate": 9.658751282827075e-06, "loss": 0.0633, "step": 34174 }, { "epoch": 10.52, "learning_rate": 9.658253303406699e-06, "loss": 0.08, "step": 34175 }, { "epoch": 10.52, "learning_rate": 9.657755324834794e-06, "loss": 0.0413, "step": 34176 }, { "epoch": 10.52, "learning_rate": 9.657257347112588e-06, "loss": 0.0773, "step": 34177 }, { "epoch": 10.52, "learning_rate": 9.656759370241318e-06, "loss": 0.0273, "step": 34178 }, { "epoch": 10.52, "learning_rate": 9.656261394222223e-06, "loss": 0.0546, "step": 34179 }, { "epoch": 10.52, "learning_rate": 9.655763419056537e-06, "loss": 0.0397, "step": 34180 }, { "epoch": 10.52, "learning_rate": 9.6552654447455e-06, "loss": 0.0821, "step": 34181 }, { "epoch": 10.52, "learning_rate": 9.654767471290346e-06, "loss": 0.0562, "step": 34182 }, { "epoch": 10.52, "learning_rate": 9.65426949869231e-06, "loss": 0.056, "step": 34183 }, { "epoch": 10.52, "learning_rate": 9.653771526952632e-06, "loss": 0.068, "step": 34184 }, { "epoch": 10.52, "learning_rate": 9.65327355607254e-06, "loss": 0.0516, "step": 34185 }, { "epoch": 10.52, "learning_rate": 9.65277558605328e-06, "loss": 0.0709, "step": 34186 }, { "epoch": 10.52, "learning_rate": 9.652277616896087e-06, "loss": 0.0422, "step": 34187 }, { "epoch": 10.52, "learning_rate": 9.651779648602193e-06, "loss": 0.0686, "step": 34188 }, { "epoch": 10.52, "learning_rate": 9.651281681172837e-06, "loss": 0.0634, "step": 34189 }, { "epoch": 10.52, "learning_rate": 9.650783714609252e-06, "loss": 0.043, "step": 34190 }, { "epoch": 10.52, "learning_rate": 9.650285748912678e-06, "loss": 0.0311, "step": 34191 }, { "epoch": 10.52, "learning_rate": 9.649787784084352e-06, "loss": 0.0364, "step": 34192 }, { "epoch": 10.52, "learning_rate": 9.649289820125507e-06, "loss": 0.049, "step": 34193 }, { "epoch": 10.52, "learning_rate": 9.648791857037384e-06, "loss": 0.0604, "step": 34194 }, { "epoch": 10.52, "learning_rate": 9.648293894821209e-06, "loss": 0.0343, "step": 34195 }, { "epoch": 10.52, "learning_rate": 9.647795933478229e-06, "loss": 0.0869, "step": 34196 }, { "epoch": 10.52, "learning_rate": 9.64729797300968e-06, "loss": 0.0729, "step": 34197 }, { "epoch": 10.52, "learning_rate": 9.646800013416792e-06, "loss": 0.039, "step": 34198 }, { "epoch": 10.52, "learning_rate": 9.646302054700805e-06, "loss": 0.077, "step": 34199 }, { "epoch": 10.52, "learning_rate": 9.645804096862953e-06, "loss": 0.0595, "step": 34200 }, { "epoch": 10.52, "learning_rate": 9.645306139904476e-06, "loss": 0.0738, "step": 34201 }, { "epoch": 10.52, "learning_rate": 9.644808183826609e-06, "loss": 0.1186, "step": 34202 }, { "epoch": 10.52, "learning_rate": 9.644310228630585e-06, "loss": 0.0949, "step": 34203 }, { "epoch": 10.52, "learning_rate": 9.643812274317644e-06, "loss": 0.055, "step": 34204 }, { "epoch": 10.52, "learning_rate": 9.64331432088902e-06, "loss": 0.0914, "step": 34205 }, { "epoch": 10.52, "learning_rate": 9.64281636834595e-06, "loss": 0.0445, "step": 34206 }, { "epoch": 10.53, "learning_rate": 9.642318416689673e-06, "loss": 0.063, "step": 34207 }, { "epoch": 10.53, "learning_rate": 9.641820465921426e-06, "loss": 0.0961, "step": 34208 }, { "epoch": 10.53, "learning_rate": 9.641322516042437e-06, "loss": 0.1053, "step": 34209 }, { "epoch": 10.53, "learning_rate": 9.640824567053945e-06, "loss": 0.065, "step": 34210 }, { "epoch": 10.53, "learning_rate": 9.640326618957196e-06, "loss": 0.0341, "step": 34211 }, { "epoch": 10.53, "learning_rate": 9.639828671753415e-06, "loss": 0.0909, "step": 34212 }, { "epoch": 10.53, "learning_rate": 9.639330725443842e-06, "loss": 0.0709, "step": 34213 }, { "epoch": 10.53, "learning_rate": 9.638832780029716e-06, "loss": 0.0467, "step": 34214 }, { "epoch": 10.53, "learning_rate": 9.638334835512266e-06, "loss": 0.0914, "step": 34215 }, { "epoch": 10.53, "learning_rate": 9.637836891892738e-06, "loss": 0.0579, "step": 34216 }, { "epoch": 10.53, "learning_rate": 9.637338949172362e-06, "loss": 0.0527, "step": 34217 }, { "epoch": 10.53, "learning_rate": 9.636841007352378e-06, "loss": 0.0256, "step": 34218 }, { "epoch": 10.53, "learning_rate": 9.636343066434016e-06, "loss": 0.044, "step": 34219 }, { "epoch": 10.53, "learning_rate": 9.635845126418514e-06, "loss": 0.0895, "step": 34220 }, { "epoch": 10.53, "learning_rate": 9.635347187307116e-06, "loss": 0.0362, "step": 34221 }, { "epoch": 10.53, "learning_rate": 9.63484924910105e-06, "loss": 0.0681, "step": 34222 }, { "epoch": 10.53, "learning_rate": 9.634351311801557e-06, "loss": 0.073, "step": 34223 }, { "epoch": 10.53, "learning_rate": 9.633853375409868e-06, "loss": 0.043, "step": 34224 }, { "epoch": 10.53, "learning_rate": 9.633355439927222e-06, "loss": 0.0476, "step": 34225 }, { "epoch": 10.53, "learning_rate": 9.632857505354859e-06, "loss": 0.1061, "step": 34226 }, { "epoch": 10.53, "learning_rate": 9.632359571694009e-06, "loss": 0.0753, "step": 34227 }, { "epoch": 10.53, "learning_rate": 9.631861638945912e-06, "loss": 0.0612, "step": 34228 }, { "epoch": 10.53, "learning_rate": 9.631363707111808e-06, "loss": 0.0594, "step": 34229 }, { "epoch": 10.53, "learning_rate": 9.630865776192918e-06, "loss": 0.0313, "step": 34230 }, { "epoch": 10.53, "learning_rate": 9.630367846190498e-06, "loss": 0.1159, "step": 34231 }, { "epoch": 10.53, "learning_rate": 9.629869917105771e-06, "loss": 0.066, "step": 34232 }, { "epoch": 10.53, "learning_rate": 9.629371988939979e-06, "loss": 0.0438, "step": 34233 }, { "epoch": 10.53, "learning_rate": 9.628874061694355e-06, "loss": 0.0652, "step": 34234 }, { "epoch": 10.53, "learning_rate": 9.628376135370136e-06, "loss": 0.0814, "step": 34235 }, { "epoch": 10.53, "learning_rate": 9.627878209968559e-06, "loss": 0.048, "step": 34236 }, { "epoch": 10.53, "learning_rate": 9.627380285490863e-06, "loss": 0.0394, "step": 34237 }, { "epoch": 10.53, "learning_rate": 9.62688236193828e-06, "loss": 0.0517, "step": 34238 }, { "epoch": 10.54, "learning_rate": 9.626384439312048e-06, "loss": 0.0453, "step": 34239 }, { "epoch": 10.54, "learning_rate": 9.625886517613398e-06, "loss": 0.1146, "step": 34240 }, { "epoch": 10.54, "learning_rate": 9.625388596843575e-06, "loss": 0.0756, "step": 34241 }, { "epoch": 10.54, "learning_rate": 9.624890677003812e-06, "loss": 0.0538, "step": 34242 }, { "epoch": 10.54, "learning_rate": 9.624392758095344e-06, "loss": 0.0864, "step": 34243 }, { "epoch": 10.54, "learning_rate": 9.623894840119406e-06, "loss": 0.0493, "step": 34244 }, { "epoch": 10.54, "learning_rate": 9.623396923077235e-06, "loss": 0.0391, "step": 34245 }, { "epoch": 10.54, "learning_rate": 9.62289900697007e-06, "loss": 0.0534, "step": 34246 }, { "epoch": 10.54, "learning_rate": 9.622401091799143e-06, "loss": 0.0587, "step": 34247 }, { "epoch": 10.54, "learning_rate": 9.621903177565694e-06, "loss": 0.0623, "step": 34248 }, { "epoch": 10.54, "learning_rate": 9.621405264270957e-06, "loss": 0.0962, "step": 34249 }, { "epoch": 10.54, "learning_rate": 9.620907351916166e-06, "loss": 0.0708, "step": 34250 }, { "epoch": 10.54, "learning_rate": 9.620409440502564e-06, "loss": 0.0991, "step": 34251 }, { "epoch": 10.54, "learning_rate": 9.619911530031384e-06, "loss": 0.0564, "step": 34252 }, { "epoch": 10.54, "learning_rate": 9.619413620503857e-06, "loss": 0.0812, "step": 34253 }, { "epoch": 10.54, "learning_rate": 9.618915711921226e-06, "loss": 0.0424, "step": 34254 }, { "epoch": 10.54, "learning_rate": 9.618417804284722e-06, "loss": 0.0578, "step": 34255 }, { "epoch": 10.54, "learning_rate": 9.617919897595586e-06, "loss": 0.0381, "step": 34256 }, { "epoch": 10.54, "learning_rate": 9.617421991855052e-06, "loss": 0.0669, "step": 34257 }, { "epoch": 10.54, "learning_rate": 9.616924087064355e-06, "loss": 0.0919, "step": 34258 }, { "epoch": 10.54, "learning_rate": 9.616426183224734e-06, "loss": 0.062, "step": 34259 }, { "epoch": 10.54, "learning_rate": 9.61592828033742e-06, "loss": 0.0774, "step": 34260 }, { "epoch": 10.54, "learning_rate": 9.615430378403653e-06, "loss": 0.0446, "step": 34261 }, { "epoch": 10.54, "learning_rate": 9.61493247742467e-06, "loss": 0.0721, "step": 34262 }, { "epoch": 10.54, "learning_rate": 9.614434577401708e-06, "loss": 0.0584, "step": 34263 }, { "epoch": 10.54, "learning_rate": 9.613936678336e-06, "loss": 0.1012, "step": 34264 }, { "epoch": 10.54, "learning_rate": 9.613438780228777e-06, "loss": 0.0753, "step": 34265 }, { "epoch": 10.54, "learning_rate": 9.612940883081289e-06, "loss": 0.0759, "step": 34266 }, { "epoch": 10.54, "learning_rate": 9.612442986894763e-06, "loss": 0.0865, "step": 34267 }, { "epoch": 10.54, "learning_rate": 9.611945091670435e-06, "loss": 0.0583, "step": 34268 }, { "epoch": 10.54, "learning_rate": 9.611447197409544e-06, "loss": 0.0613, "step": 34269 }, { "epoch": 10.54, "learning_rate": 9.610949304113321e-06, "loss": 0.0496, "step": 34270 }, { "epoch": 10.54, "learning_rate": 9.61045141178301e-06, "loss": 0.0789, "step": 34271 }, { "epoch": 10.55, "learning_rate": 9.609953520419842e-06, "loss": 0.0936, "step": 34272 }, { "epoch": 10.55, "learning_rate": 9.609455630025058e-06, "loss": 0.05, "step": 34273 }, { "epoch": 10.55, "learning_rate": 9.608957740599888e-06, "loss": 0.07, "step": 34274 }, { "epoch": 10.55, "learning_rate": 9.608459852145563e-06, "loss": 0.0619, "step": 34275 }, { "epoch": 10.55, "learning_rate": 9.607961964663337e-06, "loss": 0.0978, "step": 34276 }, { "epoch": 10.55, "learning_rate": 9.60746407815443e-06, "loss": 0.0595, "step": 34277 }, { "epoch": 10.55, "learning_rate": 9.606966192620086e-06, "loss": 0.054, "step": 34278 }, { "epoch": 10.55, "learning_rate": 9.606468308061539e-06, "loss": 0.0759, "step": 34279 }, { "epoch": 10.55, "learning_rate": 9.605970424480021e-06, "loss": 0.0556, "step": 34280 }, { "epoch": 10.55, "learning_rate": 9.605472541876778e-06, "loss": 0.1037, "step": 34281 }, { "epoch": 10.55, "learning_rate": 9.604974660253039e-06, "loss": 0.0578, "step": 34282 }, { "epoch": 10.55, "learning_rate": 9.60447677961004e-06, "loss": 0.0361, "step": 34283 }, { "epoch": 10.55, "learning_rate": 9.603978899949021e-06, "loss": 0.0567, "step": 34284 }, { "epoch": 10.55, "learning_rate": 9.60348102127121e-06, "loss": 0.0272, "step": 34285 }, { "epoch": 10.55, "learning_rate": 9.602983143577852e-06, "loss": 0.0671, "step": 34286 }, { "epoch": 10.55, "learning_rate": 9.602485266870184e-06, "loss": 0.0419, "step": 34287 }, { "epoch": 10.55, "learning_rate": 9.601987391149434e-06, "loss": 0.0567, "step": 34288 }, { "epoch": 10.55, "learning_rate": 9.601489516416843e-06, "loss": 0.0445, "step": 34289 }, { "epoch": 10.55, "learning_rate": 9.600991642673642e-06, "loss": 0.0423, "step": 34290 }, { "epoch": 10.55, "learning_rate": 9.600493769921075e-06, "loss": 0.0355, "step": 34291 }, { "epoch": 10.55, "learning_rate": 9.599995898160374e-06, "loss": 0.0596, "step": 34292 }, { "epoch": 10.55, "learning_rate": 9.599498027392776e-06, "loss": 0.0649, "step": 34293 }, { "epoch": 10.55, "learning_rate": 9.599000157619517e-06, "loss": 0.0409, "step": 34294 }, { "epoch": 10.55, "learning_rate": 9.598502288841827e-06, "loss": 0.0668, "step": 34295 }, { "epoch": 10.55, "learning_rate": 9.598004421060952e-06, "loss": 0.0905, "step": 34296 }, { "epoch": 10.55, "learning_rate": 9.597506554278126e-06, "loss": 0.034, "step": 34297 }, { "epoch": 10.55, "learning_rate": 9.59700868849458e-06, "loss": 0.0505, "step": 34298 }, { "epoch": 10.55, "learning_rate": 9.596510823711553e-06, "loss": 0.065, "step": 34299 }, { "epoch": 10.55, "learning_rate": 9.596012959930278e-06, "loss": 0.0401, "step": 34300 }, { "epoch": 10.55, "learning_rate": 9.595515097151997e-06, "loss": 0.059, "step": 34301 }, { "epoch": 10.55, "learning_rate": 9.595017235377942e-06, "loss": 0.0807, "step": 34302 }, { "epoch": 10.55, "learning_rate": 9.594519374609352e-06, "loss": 0.0317, "step": 34303 }, { "epoch": 10.56, "learning_rate": 9.59402151484746e-06, "loss": 0.0606, "step": 34304 }, { "epoch": 10.56, "learning_rate": 9.5935236560935e-06, "loss": 0.0702, "step": 34305 }, { "epoch": 10.56, "learning_rate": 9.593025798348713e-06, "loss": 0.0727, "step": 34306 }, { "epoch": 10.56, "learning_rate": 9.592527941614334e-06, "loss": 0.021, "step": 34307 }, { "epoch": 10.56, "learning_rate": 9.592030085891602e-06, "loss": 0.0426, "step": 34308 }, { "epoch": 10.56, "learning_rate": 9.591532231181746e-06, "loss": 0.045, "step": 34309 }, { "epoch": 10.56, "learning_rate": 9.591034377486003e-06, "loss": 0.0319, "step": 34310 }, { "epoch": 10.56, "learning_rate": 9.590536524805613e-06, "loss": 0.0628, "step": 34311 }, { "epoch": 10.56, "learning_rate": 9.590038673141812e-06, "loss": 0.0919, "step": 34312 }, { "epoch": 10.56, "learning_rate": 9.589540822495832e-06, "loss": 0.0679, "step": 34313 }, { "epoch": 10.56, "learning_rate": 9.589042972868913e-06, "loss": 0.0668, "step": 34314 }, { "epoch": 10.56, "learning_rate": 9.58854512426229e-06, "loss": 0.0776, "step": 34315 }, { "epoch": 10.56, "learning_rate": 9.588047276677196e-06, "loss": 0.0631, "step": 34316 }, { "epoch": 10.56, "learning_rate": 9.587549430114872e-06, "loss": 0.0654, "step": 34317 }, { "epoch": 10.56, "learning_rate": 9.587051584576554e-06, "loss": 0.0473, "step": 34318 }, { "epoch": 10.56, "learning_rate": 9.586553740063473e-06, "loss": 0.0647, "step": 34319 }, { "epoch": 10.56, "learning_rate": 9.586055896576863e-06, "loss": 0.0919, "step": 34320 }, { "epoch": 10.56, "learning_rate": 9.58555805411797e-06, "loss": 0.0247, "step": 34321 }, { "epoch": 10.56, "learning_rate": 9.585060212688023e-06, "loss": 0.0642, "step": 34322 }, { "epoch": 10.56, "learning_rate": 9.58456237228826e-06, "loss": 0.0614, "step": 34323 }, { "epoch": 10.56, "learning_rate": 9.584064532919918e-06, "loss": 0.0431, "step": 34324 }, { "epoch": 10.56, "learning_rate": 9.583566694584228e-06, "loss": 0.0325, "step": 34325 }, { "epoch": 10.56, "learning_rate": 9.583068857282431e-06, "loss": 0.0744, "step": 34326 }, { "epoch": 10.56, "learning_rate": 9.582571021015764e-06, "loss": 0.0459, "step": 34327 }, { "epoch": 10.56, "learning_rate": 9.58207318578546e-06, "loss": 0.0681, "step": 34328 }, { "epoch": 10.56, "learning_rate": 9.581575351592756e-06, "loss": 0.0662, "step": 34329 }, { "epoch": 10.56, "learning_rate": 9.581077518438883e-06, "loss": 0.061, "step": 34330 }, { "epoch": 10.56, "learning_rate": 9.580579686325087e-06, "loss": 0.0974, "step": 34331 }, { "epoch": 10.56, "learning_rate": 9.580081855252597e-06, "loss": 0.0676, "step": 34332 }, { "epoch": 10.56, "learning_rate": 9.57958402522265e-06, "loss": 0.0238, "step": 34333 }, { "epoch": 10.56, "learning_rate": 9.579086196236483e-06, "loss": 0.0682, "step": 34334 }, { "epoch": 10.56, "learning_rate": 9.578588368295328e-06, "loss": 0.106, "step": 34335 }, { "epoch": 10.56, "learning_rate": 9.578090541400429e-06, "loss": 0.0708, "step": 34336 }, { "epoch": 10.57, "learning_rate": 9.577592715553016e-06, "loss": 0.0745, "step": 34337 }, { "epoch": 10.57, "learning_rate": 9.577094890754325e-06, "loss": 0.0502, "step": 34338 }, { "epoch": 10.57, "learning_rate": 9.5765970670056e-06, "loss": 0.0452, "step": 34339 }, { "epoch": 10.57, "learning_rate": 9.57609924430806e-06, "loss": 0.0281, "step": 34340 }, { "epoch": 10.57, "learning_rate": 9.575601422662955e-06, "loss": 0.0486, "step": 34341 }, { "epoch": 10.57, "learning_rate": 9.575103602071521e-06, "loss": 0.0276, "step": 34342 }, { "epoch": 10.57, "learning_rate": 9.574605782534988e-06, "loss": 0.0173, "step": 34343 }, { "epoch": 10.57, "learning_rate": 9.574107964054596e-06, "loss": 0.0694, "step": 34344 }, { "epoch": 10.57, "learning_rate": 9.573610146631574e-06, "loss": 0.0628, "step": 34345 }, { "epoch": 10.57, "learning_rate": 9.573112330267166e-06, "loss": 0.0947, "step": 34346 }, { "epoch": 10.57, "learning_rate": 9.572614514962607e-06, "loss": 0.0713, "step": 34347 }, { "epoch": 10.57, "learning_rate": 9.572116700719129e-06, "loss": 0.0515, "step": 34348 }, { "epoch": 10.57, "learning_rate": 9.571618887537969e-06, "loss": 0.1252, "step": 34349 }, { "epoch": 10.57, "learning_rate": 9.571121075420364e-06, "loss": 0.1008, "step": 34350 }, { "epoch": 10.57, "learning_rate": 9.57062326436755e-06, "loss": 0.0455, "step": 34351 }, { "epoch": 10.57, "learning_rate": 9.570125454380765e-06, "loss": 0.045, "step": 34352 }, { "epoch": 10.57, "learning_rate": 9.569627645461242e-06, "loss": 0.0232, "step": 34353 }, { "epoch": 10.57, "learning_rate": 9.569129837610216e-06, "loss": 0.008, "step": 34354 }, { "epoch": 10.57, "learning_rate": 9.568632030828922e-06, "loss": 0.0531, "step": 34355 }, { "epoch": 10.57, "learning_rate": 9.568134225118603e-06, "loss": 0.0585, "step": 34356 }, { "epoch": 10.57, "learning_rate": 9.567636420480489e-06, "loss": 0.0674, "step": 34357 }, { "epoch": 10.57, "learning_rate": 9.567138616915816e-06, "loss": 0.0939, "step": 34358 }, { "epoch": 10.57, "learning_rate": 9.566640814425823e-06, "loss": 0.0822, "step": 34359 }, { "epoch": 10.57, "learning_rate": 9.56614301301174e-06, "loss": 0.1014, "step": 34360 }, { "epoch": 10.57, "learning_rate": 9.56564521267481e-06, "loss": 0.064, "step": 34361 }, { "epoch": 10.57, "learning_rate": 9.565147413416266e-06, "loss": 0.037, "step": 34362 }, { "epoch": 10.57, "learning_rate": 9.564649615237345e-06, "loss": 0.0871, "step": 34363 }, { "epoch": 10.57, "learning_rate": 9.564151818139281e-06, "loss": 0.052, "step": 34364 }, { "epoch": 10.57, "learning_rate": 9.563654022123305e-06, "loss": 0.071, "step": 34365 }, { "epoch": 10.57, "learning_rate": 9.563156227190665e-06, "loss": 0.09, "step": 34366 }, { "epoch": 10.57, "learning_rate": 9.562658433342589e-06, "loss": 0.0569, "step": 34367 }, { "epoch": 10.57, "learning_rate": 9.562160640580313e-06, "loss": 0.0468, "step": 34368 }, { "epoch": 10.58, "learning_rate": 9.561662848905076e-06, "loss": 0.0607, "step": 34369 }, { "epoch": 10.58, "learning_rate": 9.561165058318108e-06, "loss": 0.0903, "step": 34370 }, { "epoch": 10.58, "learning_rate": 9.560667268820652e-06, "loss": 0.0346, "step": 34371 }, { "epoch": 10.58, "learning_rate": 9.56016948041394e-06, "loss": 0.0765, "step": 34372 }, { "epoch": 10.58, "learning_rate": 9.55967169309921e-06, "loss": 0.0722, "step": 34373 }, { "epoch": 10.58, "learning_rate": 9.559173906877696e-06, "loss": 0.0375, "step": 34374 }, { "epoch": 10.58, "learning_rate": 9.55867612175063e-06, "loss": 0.0331, "step": 34375 }, { "epoch": 10.58, "learning_rate": 9.558178337719259e-06, "loss": 0.0623, "step": 34376 }, { "epoch": 10.58, "learning_rate": 9.557680554784809e-06, "loss": 0.0845, "step": 34377 }, { "epoch": 10.58, "learning_rate": 9.557182772948518e-06, "loss": 0.0426, "step": 34378 }, { "epoch": 10.58, "learning_rate": 9.556684992211624e-06, "loss": 0.0879, "step": 34379 }, { "epoch": 10.58, "learning_rate": 9.556187212575359e-06, "loss": 0.0396, "step": 34380 }, { "epoch": 10.58, "learning_rate": 9.555689434040965e-06, "loss": 0.0593, "step": 34381 }, { "epoch": 10.58, "learning_rate": 9.555191656609673e-06, "loss": 0.0318, "step": 34382 }, { "epoch": 10.58, "learning_rate": 9.55469388028272e-06, "loss": 0.0359, "step": 34383 }, { "epoch": 10.58, "learning_rate": 9.554196105061345e-06, "loss": 0.0582, "step": 34384 }, { "epoch": 10.58, "learning_rate": 9.553698330946773e-06, "loss": 0.0787, "step": 34385 }, { "epoch": 10.58, "learning_rate": 9.553200557940254e-06, "loss": 0.0696, "step": 34386 }, { "epoch": 10.58, "learning_rate": 9.552702786043019e-06, "loss": 0.0527, "step": 34387 }, { "epoch": 10.58, "learning_rate": 9.5522050152563e-06, "loss": 0.0796, "step": 34388 }, { "epoch": 10.58, "learning_rate": 9.551707245581336e-06, "loss": 0.0537, "step": 34389 }, { "epoch": 10.58, "learning_rate": 9.55120947701936e-06, "loss": 0.1281, "step": 34390 }, { "epoch": 10.58, "learning_rate": 9.550711709571609e-06, "loss": 0.1021, "step": 34391 }, { "epoch": 10.58, "learning_rate": 9.550213943239323e-06, "loss": 0.1029, "step": 34392 }, { "epoch": 10.58, "learning_rate": 9.549716178023733e-06, "loss": 0.0439, "step": 34393 }, { "epoch": 10.58, "learning_rate": 9.54921841392608e-06, "loss": 0.0803, "step": 34394 }, { "epoch": 10.58, "learning_rate": 9.548720650947587e-06, "loss": 0.0479, "step": 34395 }, { "epoch": 10.58, "learning_rate": 9.548222889089505e-06, "loss": 0.0539, "step": 34396 }, { "epoch": 10.58, "learning_rate": 9.547725128353068e-06, "loss": 0.0805, "step": 34397 }, { "epoch": 10.58, "learning_rate": 9.547227368739503e-06, "loss": 0.037, "step": 34398 }, { "epoch": 10.58, "learning_rate": 9.54672961025005e-06, "loss": 0.0506, "step": 34399 }, { "epoch": 10.58, "learning_rate": 9.546231852885943e-06, "loss": 0.0779, "step": 34400 }, { "epoch": 10.58, "learning_rate": 9.545734096648423e-06, "loss": 0.0316, "step": 34401 }, { "epoch": 10.59, "learning_rate": 9.545236341538722e-06, "loss": 0.0653, "step": 34402 }, { "epoch": 10.59, "learning_rate": 9.544738587558078e-06, "loss": 0.0533, "step": 34403 }, { "epoch": 10.59, "learning_rate": 9.544240834707724e-06, "loss": 0.0463, "step": 34404 }, { "epoch": 10.59, "learning_rate": 9.543743082988896e-06, "loss": 0.0547, "step": 34405 }, { "epoch": 10.59, "learning_rate": 9.543245332402832e-06, "loss": 0.0406, "step": 34406 }, { "epoch": 10.59, "learning_rate": 9.542747582950768e-06, "loss": 0.0893, "step": 34407 }, { "epoch": 10.59, "learning_rate": 9.54224983463394e-06, "loss": 0.079, "step": 34408 }, { "epoch": 10.59, "learning_rate": 9.54175208745358e-06, "loss": 0.0617, "step": 34409 }, { "epoch": 10.59, "learning_rate": 9.541254341410924e-06, "loss": 0.0647, "step": 34410 }, { "epoch": 10.59, "learning_rate": 9.540756596507212e-06, "loss": 0.058, "step": 34411 }, { "epoch": 10.59, "learning_rate": 9.540258852743676e-06, "loss": 0.032, "step": 34412 }, { "epoch": 10.59, "learning_rate": 9.539761110121555e-06, "loss": 0.0735, "step": 34413 }, { "epoch": 10.59, "learning_rate": 9.539263368642084e-06, "loss": 0.0609, "step": 34414 }, { "epoch": 10.59, "learning_rate": 9.538765628306494e-06, "loss": 0.0634, "step": 34415 }, { "epoch": 10.59, "learning_rate": 9.538267889116027e-06, "loss": 0.0337, "step": 34416 }, { "epoch": 10.59, "learning_rate": 9.537770151071918e-06, "loss": 0.0616, "step": 34417 }, { "epoch": 10.59, "learning_rate": 9.5372724141754e-06, "loss": 0.0765, "step": 34418 }, { "epoch": 10.59, "learning_rate": 9.536774678427709e-06, "loss": 0.082, "step": 34419 }, { "epoch": 10.59, "learning_rate": 9.536276943830078e-06, "loss": 0.0669, "step": 34420 }, { "epoch": 10.59, "learning_rate": 9.535779210383753e-06, "loss": 0.0685, "step": 34421 }, { "epoch": 10.59, "learning_rate": 9.53528147808996e-06, "loss": 0.0749, "step": 34422 }, { "epoch": 10.59, "learning_rate": 9.534783746949939e-06, "loss": 0.0306, "step": 34423 }, { "epoch": 10.59, "learning_rate": 9.534286016964924e-06, "loss": 0.0451, "step": 34424 }, { "epoch": 10.59, "learning_rate": 9.533788288136149e-06, "loss": 0.0516, "step": 34425 }, { "epoch": 10.59, "learning_rate": 9.533290560464854e-06, "loss": 0.1216, "step": 34426 }, { "epoch": 10.59, "learning_rate": 9.532792833952273e-06, "loss": 0.0299, "step": 34427 }, { "epoch": 10.59, "learning_rate": 9.532295108599642e-06, "loss": 0.0602, "step": 34428 }, { "epoch": 10.59, "learning_rate": 9.531797384408196e-06, "loss": 0.0437, "step": 34429 }, { "epoch": 10.59, "learning_rate": 9.531299661379167e-06, "loss": 0.0638, "step": 34430 }, { "epoch": 10.59, "learning_rate": 9.5308019395138e-06, "loss": 0.0665, "step": 34431 }, { "epoch": 10.59, "learning_rate": 9.530304218813323e-06, "loss": 0.055, "step": 34432 }, { "epoch": 10.59, "learning_rate": 9.529806499278974e-06, "loss": 0.0768, "step": 34433 }, { "epoch": 10.6, "learning_rate": 9.52930878091199e-06, "loss": 0.0671, "step": 34434 }, { "epoch": 10.6, "learning_rate": 9.5288110637136e-06, "loss": 0.0566, "step": 34435 }, { "epoch": 10.6, "learning_rate": 9.528313347685052e-06, "loss": 0.0317, "step": 34436 }, { "epoch": 10.6, "learning_rate": 9.527815632827571e-06, "loss": 0.0458, "step": 34437 }, { "epoch": 10.6, "learning_rate": 9.527317919142398e-06, "loss": 0.1639, "step": 34438 }, { "epoch": 10.6, "learning_rate": 9.526820206630768e-06, "loss": 0.0635, "step": 34439 }, { "epoch": 10.6, "learning_rate": 9.52632249529391e-06, "loss": 0.0816, "step": 34440 }, { "epoch": 10.6, "learning_rate": 9.52582478513307e-06, "loss": 0.0431, "step": 34441 }, { "epoch": 10.6, "learning_rate": 9.525327076149482e-06, "loss": 0.0394, "step": 34442 }, { "epoch": 10.6, "learning_rate": 9.524829368344378e-06, "loss": 0.071, "step": 34443 }, { "epoch": 10.6, "learning_rate": 9.524331661718994e-06, "loss": 0.125, "step": 34444 }, { "epoch": 10.6, "learning_rate": 9.523833956274562e-06, "loss": 0.0662, "step": 34445 }, { "epoch": 10.6, "learning_rate": 9.523336252012325e-06, "loss": 0.061, "step": 34446 }, { "epoch": 10.6, "learning_rate": 9.522838548933515e-06, "loss": 0.057, "step": 34447 }, { "epoch": 10.6, "learning_rate": 9.52234084703937e-06, "loss": 0.0779, "step": 34448 }, { "epoch": 10.6, "learning_rate": 9.521843146331122e-06, "loss": 0.0707, "step": 34449 }, { "epoch": 10.6, "learning_rate": 9.521345446810008e-06, "loss": 0.0794, "step": 34450 }, { "epoch": 10.6, "learning_rate": 9.520847748477266e-06, "loss": 0.029, "step": 34451 }, { "epoch": 10.6, "learning_rate": 9.520350051334131e-06, "loss": 0.0851, "step": 34452 }, { "epoch": 10.6, "learning_rate": 9.519852355381835e-06, "loss": 0.0463, "step": 34453 }, { "epoch": 10.6, "learning_rate": 9.519354660621617e-06, "loss": 0.069, "step": 34454 }, { "epoch": 10.6, "learning_rate": 9.51885696705471e-06, "loss": 0.0297, "step": 34455 }, { "epoch": 10.6, "learning_rate": 9.518359274682354e-06, "loss": 0.0618, "step": 34456 }, { "epoch": 10.6, "learning_rate": 9.517861583505782e-06, "loss": 0.0652, "step": 34457 }, { "epoch": 10.6, "learning_rate": 9.517363893526228e-06, "loss": 0.0464, "step": 34458 }, { "epoch": 10.6, "learning_rate": 9.516866204744932e-06, "loss": 0.0647, "step": 34459 }, { "epoch": 10.6, "learning_rate": 9.516368517163123e-06, "loss": 0.0733, "step": 34460 }, { "epoch": 10.6, "learning_rate": 9.515870830782044e-06, "loss": 0.0464, "step": 34461 }, { "epoch": 10.6, "learning_rate": 9.515373145602926e-06, "loss": 0.1026, "step": 34462 }, { "epoch": 10.6, "learning_rate": 9.514875461627008e-06, "loss": 0.0725, "step": 34463 }, { "epoch": 10.6, "learning_rate": 9.514377778855521e-06, "loss": 0.0612, "step": 34464 }, { "epoch": 10.6, "learning_rate": 9.513880097289699e-06, "loss": 0.073, "step": 34465 }, { "epoch": 10.6, "learning_rate": 9.51338241693079e-06, "loss": 0.0764, "step": 34466 }, { "epoch": 10.61, "learning_rate": 9.512884737780017e-06, "loss": 0.0474, "step": 34467 }, { "epoch": 10.61, "learning_rate": 9.51238705983862e-06, "loss": 0.0429, "step": 34468 }, { "epoch": 10.61, "learning_rate": 9.511889383107835e-06, "loss": 0.0571, "step": 34469 }, { "epoch": 10.61, "learning_rate": 9.511391707588895e-06, "loss": 0.066, "step": 34470 }, { "epoch": 10.61, "learning_rate": 9.51089403328304e-06, "loss": 0.0952, "step": 34471 }, { "epoch": 10.61, "learning_rate": 9.510396360191501e-06, "loss": 0.0325, "step": 34472 }, { "epoch": 10.61, "learning_rate": 9.50989868831552e-06, "loss": 0.0655, "step": 34473 }, { "epoch": 10.61, "learning_rate": 9.509401017656327e-06, "loss": 0.0272, "step": 34474 }, { "epoch": 10.61, "learning_rate": 9.508903348215154e-06, "loss": 0.0418, "step": 34475 }, { "epoch": 10.61, "learning_rate": 9.508405679993248e-06, "loss": 0.0395, "step": 34476 }, { "epoch": 10.61, "learning_rate": 9.507908012991835e-06, "loss": 0.0791, "step": 34477 }, { "epoch": 10.61, "learning_rate": 9.507410347212157e-06, "loss": 0.0378, "step": 34478 }, { "epoch": 10.61, "learning_rate": 9.50691268265544e-06, "loss": 0.0749, "step": 34479 }, { "epoch": 10.61, "learning_rate": 9.50641501932293e-06, "loss": 0.0229, "step": 34480 }, { "epoch": 10.61, "learning_rate": 9.505917357215859e-06, "loss": 0.1014, "step": 34481 }, { "epoch": 10.61, "learning_rate": 9.505419696335461e-06, "loss": 0.0981, "step": 34482 }, { "epoch": 10.61, "learning_rate": 9.504922036682974e-06, "loss": 0.059, "step": 34483 }, { "epoch": 10.61, "learning_rate": 9.504424378259628e-06, "loss": 0.0781, "step": 34484 }, { "epoch": 10.61, "learning_rate": 9.503926721066664e-06, "loss": 0.0755, "step": 34485 }, { "epoch": 10.61, "learning_rate": 9.50342906510532e-06, "loss": 0.0851, "step": 34486 }, { "epoch": 10.61, "learning_rate": 9.502931410376826e-06, "loss": 0.0601, "step": 34487 }, { "epoch": 10.61, "learning_rate": 9.50243375688242e-06, "loss": 0.065, "step": 34488 }, { "epoch": 10.61, "learning_rate": 9.501936104623334e-06, "loss": 0.0514, "step": 34489 }, { "epoch": 10.61, "learning_rate": 9.501438453600808e-06, "loss": 0.0764, "step": 34490 }, { "epoch": 10.61, "learning_rate": 9.500940803816076e-06, "loss": 0.0461, "step": 34491 }, { "epoch": 10.61, "learning_rate": 9.500443155270374e-06, "loss": 0.0741, "step": 34492 }, { "epoch": 10.61, "learning_rate": 9.499945507964937e-06, "loss": 0.0442, "step": 34493 }, { "epoch": 10.61, "learning_rate": 9.499447861900997e-06, "loss": 0.0803, "step": 34494 }, { "epoch": 10.61, "learning_rate": 9.498950217079796e-06, "loss": 0.0641, "step": 34495 }, { "epoch": 10.61, "learning_rate": 9.498452573502565e-06, "loss": 0.039, "step": 34496 }, { "epoch": 10.61, "learning_rate": 9.497954931170545e-06, "loss": 0.0983, "step": 34497 }, { "epoch": 10.61, "learning_rate": 9.497457290084967e-06, "loss": 0.0577, "step": 34498 }, { "epoch": 10.62, "learning_rate": 9.49695965024706e-06, "loss": 0.0762, "step": 34499 }, { "epoch": 10.62, "learning_rate": 9.496462011658074e-06, "loss": 0.0607, "step": 34500 }, { "epoch": 10.62, "learning_rate": 9.495964374319235e-06, "loss": 0.0352, "step": 34501 }, { "epoch": 10.62, "learning_rate": 9.49546673823178e-06, "loss": 0.0286, "step": 34502 }, { "epoch": 10.62, "learning_rate": 9.494969103396946e-06, "loss": 0.0404, "step": 34503 }, { "epoch": 10.62, "learning_rate": 9.494471469815965e-06, "loss": 0.0588, "step": 34504 }, { "epoch": 10.62, "learning_rate": 9.493973837490076e-06, "loss": 0.0709, "step": 34505 }, { "epoch": 10.62, "learning_rate": 9.493476206420516e-06, "loss": 0.0737, "step": 34506 }, { "epoch": 10.62, "learning_rate": 9.492978576608516e-06, "loss": 0.0536, "step": 34507 }, { "epoch": 10.62, "learning_rate": 9.492480948055316e-06, "loss": 0.075, "step": 34508 }, { "epoch": 10.62, "learning_rate": 9.491983320762143e-06, "loss": 0.0888, "step": 34509 }, { "epoch": 10.62, "learning_rate": 9.491485694730245e-06, "loss": 0.048, "step": 34510 }, { "epoch": 10.62, "learning_rate": 9.490988069960849e-06, "loss": 0.0552, "step": 34511 }, { "epoch": 10.62, "learning_rate": 9.490490446455192e-06, "loss": 0.0833, "step": 34512 }, { "epoch": 10.62, "learning_rate": 9.489992824214509e-06, "loss": 0.0355, "step": 34513 }, { "epoch": 10.62, "learning_rate": 9.489495203240036e-06, "loss": 0.0807, "step": 34514 }, { "epoch": 10.62, "learning_rate": 9.488997583533009e-06, "loss": 0.0547, "step": 34515 }, { "epoch": 10.62, "learning_rate": 9.488499965094664e-06, "loss": 0.0455, "step": 34516 }, { "epoch": 10.62, "learning_rate": 9.488002347926235e-06, "loss": 0.1604, "step": 34517 }, { "epoch": 10.62, "learning_rate": 9.487504732028963e-06, "loss": 0.0845, "step": 34518 }, { "epoch": 10.62, "learning_rate": 9.487007117404069e-06, "loss": 0.0521, "step": 34519 }, { "epoch": 10.62, "learning_rate": 9.486509504052804e-06, "loss": 0.0607, "step": 34520 }, { "epoch": 10.62, "learning_rate": 9.486011891976399e-06, "loss": 0.0456, "step": 34521 }, { "epoch": 10.62, "learning_rate": 9.485514281176086e-06, "loss": 0.0283, "step": 34522 }, { "epoch": 10.62, "learning_rate": 9.485016671653102e-06, "loss": 0.0304, "step": 34523 }, { "epoch": 10.62, "learning_rate": 9.484519063408682e-06, "loss": 0.0602, "step": 34524 }, { "epoch": 10.62, "learning_rate": 9.484021456444063e-06, "loss": 0.0418, "step": 34525 }, { "epoch": 10.62, "learning_rate": 9.48352385076048e-06, "loss": 0.0805, "step": 34526 }, { "epoch": 10.62, "learning_rate": 9.483026246359167e-06, "loss": 0.049, "step": 34527 }, { "epoch": 10.62, "learning_rate": 9.482528643241363e-06, "loss": 0.0921, "step": 34528 }, { "epoch": 10.62, "learning_rate": 9.482031041408296e-06, "loss": 0.1705, "step": 34529 }, { "epoch": 10.62, "learning_rate": 9.481533440861209e-06, "loss": 0.0725, "step": 34530 }, { "epoch": 10.62, "learning_rate": 9.481035841601339e-06, "loss": 0.0572, "step": 34531 }, { "epoch": 10.63, "learning_rate": 9.480538243629912e-06, "loss": 0.0561, "step": 34532 }, { "epoch": 10.63, "learning_rate": 9.48004064694817e-06, "loss": 0.0537, "step": 34533 }, { "epoch": 10.63, "learning_rate": 9.479543051557344e-06, "loss": 0.1478, "step": 34534 }, { "epoch": 10.63, "learning_rate": 9.479045457458676e-06, "loss": 0.0479, "step": 34535 }, { "epoch": 10.63, "learning_rate": 9.478547864653395e-06, "loss": 0.0814, "step": 34536 }, { "epoch": 10.63, "learning_rate": 9.478050273142743e-06, "loss": 0.0189, "step": 34537 }, { "epoch": 10.63, "learning_rate": 9.477552682927948e-06, "loss": 0.0705, "step": 34538 }, { "epoch": 10.63, "learning_rate": 9.477055094010248e-06, "loss": 0.0552, "step": 34539 }, { "epoch": 10.63, "learning_rate": 9.476557506390881e-06, "loss": 0.0699, "step": 34540 }, { "epoch": 10.63, "learning_rate": 9.47605992007108e-06, "loss": 0.0574, "step": 34541 }, { "epoch": 10.63, "learning_rate": 9.475562335052086e-06, "loss": 0.0653, "step": 34542 }, { "epoch": 10.63, "learning_rate": 9.475064751335124e-06, "loss": 0.0804, "step": 34543 }, { "epoch": 10.63, "learning_rate": 9.474567168921433e-06, "loss": 0.0774, "step": 34544 }, { "epoch": 10.63, "learning_rate": 9.474069587812254e-06, "loss": 0.0589, "step": 34545 }, { "epoch": 10.63, "learning_rate": 9.473572008008817e-06, "loss": 0.0608, "step": 34546 }, { "epoch": 10.63, "learning_rate": 9.47307442951236e-06, "loss": 0.0765, "step": 34547 }, { "epoch": 10.63, "learning_rate": 9.472576852324116e-06, "loss": 0.0664, "step": 34548 }, { "epoch": 10.63, "learning_rate": 9.47207927644532e-06, "loss": 0.0511, "step": 34549 }, { "epoch": 10.63, "learning_rate": 9.47158170187721e-06, "loss": 0.0677, "step": 34550 }, { "epoch": 10.63, "learning_rate": 9.47108412862102e-06, "loss": 0.0538, "step": 34551 }, { "epoch": 10.63, "learning_rate": 9.47058655667799e-06, "loss": 0.0755, "step": 34552 }, { "epoch": 10.63, "learning_rate": 9.470088986049348e-06, "loss": 0.0707, "step": 34553 }, { "epoch": 10.63, "learning_rate": 9.469591416736326e-06, "loss": 0.0317, "step": 34554 }, { "epoch": 10.63, "learning_rate": 9.469093848740172e-06, "loss": 0.0648, "step": 34555 }, { "epoch": 10.63, "learning_rate": 9.468596282062114e-06, "loss": 0.0363, "step": 34556 }, { "epoch": 10.63, "learning_rate": 9.468098716703389e-06, "loss": 0.0665, "step": 34557 }, { "epoch": 10.63, "learning_rate": 9.46760115266523e-06, "loss": 0.0935, "step": 34558 }, { "epoch": 10.63, "learning_rate": 9.467103589948872e-06, "loss": 0.0481, "step": 34559 }, { "epoch": 10.63, "learning_rate": 9.466606028555556e-06, "loss": 0.0594, "step": 34560 }, { "epoch": 10.63, "learning_rate": 9.46610846848651e-06, "loss": 0.0586, "step": 34561 }, { "epoch": 10.63, "learning_rate": 9.465610909742974e-06, "loss": 0.0793, "step": 34562 }, { "epoch": 10.63, "learning_rate": 9.465113352326184e-06, "loss": 0.0479, "step": 34563 }, { "epoch": 10.64, "learning_rate": 9.46461579623737e-06, "loss": 0.0953, "step": 34564 }, { "epoch": 10.64, "learning_rate": 9.46411824147777e-06, "loss": 0.0402, "step": 34565 }, { "epoch": 10.64, "learning_rate": 9.463620688048626e-06, "loss": 0.1068, "step": 34566 }, { "epoch": 10.64, "learning_rate": 9.463123135951164e-06, "loss": 0.0745, "step": 34567 }, { "epoch": 10.64, "learning_rate": 9.462625585186621e-06, "loss": 0.052, "step": 34568 }, { "epoch": 10.64, "learning_rate": 9.462128035756234e-06, "loss": 0.0589, "step": 34569 }, { "epoch": 10.64, "learning_rate": 9.46163048766124e-06, "loss": 0.0872, "step": 34570 }, { "epoch": 10.64, "learning_rate": 9.46113294090287e-06, "loss": 0.0306, "step": 34571 }, { "epoch": 10.64, "learning_rate": 9.460635395482364e-06, "loss": 0.0645, "step": 34572 }, { "epoch": 10.64, "learning_rate": 9.460137851400956e-06, "loss": 0.0649, "step": 34573 }, { "epoch": 10.64, "learning_rate": 9.459640308659873e-06, "loss": 0.0589, "step": 34574 }, { "epoch": 10.64, "learning_rate": 9.459142767260362e-06, "loss": 0.0523, "step": 34575 }, { "epoch": 10.64, "learning_rate": 9.458645227203658e-06, "loss": 0.0441, "step": 34576 }, { "epoch": 10.64, "learning_rate": 9.458147688490989e-06, "loss": 0.0218, "step": 34577 }, { "epoch": 10.64, "learning_rate": 9.457650151123592e-06, "loss": 0.0876, "step": 34578 }, { "epoch": 10.64, "learning_rate": 9.457152615102702e-06, "loss": 0.043, "step": 34579 }, { "epoch": 10.64, "learning_rate": 9.456655080429557e-06, "loss": 0.0486, "step": 34580 }, { "epoch": 10.64, "learning_rate": 9.45615754710539e-06, "loss": 0.0429, "step": 34581 }, { "epoch": 10.64, "learning_rate": 9.455660015131441e-06, "loss": 0.0746, "step": 34582 }, { "epoch": 10.64, "learning_rate": 9.45516248450894e-06, "loss": 0.0515, "step": 34583 }, { "epoch": 10.64, "learning_rate": 9.454664955239119e-06, "loss": 0.0739, "step": 34584 }, { "epoch": 10.64, "learning_rate": 9.454167427323222e-06, "loss": 0.0269, "step": 34585 }, { "epoch": 10.64, "learning_rate": 9.45366990076248e-06, "loss": 0.0485, "step": 34586 }, { "epoch": 10.64, "learning_rate": 9.45317237555813e-06, "loss": 0.0981, "step": 34587 }, { "epoch": 10.64, "learning_rate": 9.452674851711404e-06, "loss": 0.0648, "step": 34588 }, { "epoch": 10.64, "learning_rate": 9.452177329223537e-06, "loss": 0.063, "step": 34589 }, { "epoch": 10.64, "learning_rate": 9.451679808095766e-06, "loss": 0.0662, "step": 34590 }, { "epoch": 10.64, "learning_rate": 9.451182288329328e-06, "loss": 0.0643, "step": 34591 }, { "epoch": 10.64, "learning_rate": 9.450684769925458e-06, "loss": 0.0474, "step": 34592 }, { "epoch": 10.64, "learning_rate": 9.450187252885387e-06, "loss": 0.0757, "step": 34593 }, { "epoch": 10.64, "learning_rate": 9.449689737210352e-06, "loss": 0.1022, "step": 34594 }, { "epoch": 10.64, "learning_rate": 9.449192222901591e-06, "loss": 0.0835, "step": 34595 }, { "epoch": 10.64, "learning_rate": 9.448694709960336e-06, "loss": 0.0586, "step": 34596 }, { "epoch": 10.65, "learning_rate": 9.448197198387827e-06, "loss": 0.0473, "step": 34597 }, { "epoch": 10.65, "learning_rate": 9.447699688185293e-06, "loss": 0.0525, "step": 34598 }, { "epoch": 10.65, "learning_rate": 9.447202179353967e-06, "loss": 0.0908, "step": 34599 }, { "epoch": 10.65, "learning_rate": 9.446704671895096e-06, "loss": 0.0526, "step": 34600 }, { "epoch": 10.65, "learning_rate": 9.446207165809907e-06, "loss": 0.0583, "step": 34601 }, { "epoch": 10.65, "learning_rate": 9.445709661099635e-06, "loss": 0.0552, "step": 34602 }, { "epoch": 10.65, "learning_rate": 9.445212157765516e-06, "loss": 0.0298, "step": 34603 }, { "epoch": 10.65, "learning_rate": 9.444714655808784e-06, "loss": 0.0829, "step": 34604 }, { "epoch": 10.65, "learning_rate": 9.444217155230678e-06, "loss": 0.072, "step": 34605 }, { "epoch": 10.65, "learning_rate": 9.443719656032431e-06, "loss": 0.1483, "step": 34606 }, { "epoch": 10.65, "learning_rate": 9.443222158215278e-06, "loss": 0.0791, "step": 34607 }, { "epoch": 10.65, "learning_rate": 9.442724661780458e-06, "loss": 0.0657, "step": 34608 }, { "epoch": 10.65, "learning_rate": 9.442227166729193e-06, "loss": 0.1037, "step": 34609 }, { "epoch": 10.65, "learning_rate": 9.441729673062736e-06, "loss": 0.0645, "step": 34610 }, { "epoch": 10.65, "learning_rate": 9.441232180782311e-06, "loss": 0.0645, "step": 34611 }, { "epoch": 10.65, "learning_rate": 9.440734689889156e-06, "loss": 0.0585, "step": 34612 }, { "epoch": 10.65, "learning_rate": 9.440237200384506e-06, "loss": 0.0656, "step": 34613 }, { "epoch": 10.65, "learning_rate": 9.439739712269594e-06, "loss": 0.0433, "step": 34614 }, { "epoch": 10.65, "learning_rate": 9.439242225545659e-06, "loss": 0.0678, "step": 34615 }, { "epoch": 10.65, "learning_rate": 9.438744740213936e-06, "loss": 0.0829, "step": 34616 }, { "epoch": 10.65, "learning_rate": 9.438247256275657e-06, "loss": 0.0594, "step": 34617 }, { "epoch": 10.65, "learning_rate": 9.437749773732061e-06, "loss": 0.0955, "step": 34618 }, { "epoch": 10.65, "learning_rate": 9.437252292584373e-06, "loss": 0.0405, "step": 34619 }, { "epoch": 10.65, "learning_rate": 9.436754812833843e-06, "loss": 0.0474, "step": 34620 }, { "epoch": 10.65, "learning_rate": 9.436257334481699e-06, "loss": 0.0639, "step": 34621 }, { "epoch": 10.65, "learning_rate": 9.435759857529174e-06, "loss": 0.0908, "step": 34622 }, { "epoch": 10.65, "learning_rate": 9.435262381977506e-06, "loss": 0.0587, "step": 34623 }, { "epoch": 10.65, "learning_rate": 9.434764907827927e-06, "loss": 0.0672, "step": 34624 }, { "epoch": 10.65, "learning_rate": 9.434267435081677e-06, "loss": 0.0597, "step": 34625 }, { "epoch": 10.65, "learning_rate": 9.433769963739987e-06, "loss": 0.0429, "step": 34626 }, { "epoch": 10.65, "learning_rate": 9.433272493804094e-06, "loss": 0.0325, "step": 34627 }, { "epoch": 10.65, "learning_rate": 9.432775025275233e-06, "loss": 0.0809, "step": 34628 }, { "epoch": 10.66, "learning_rate": 9.432277558154637e-06, "loss": 0.0387, "step": 34629 }, { "epoch": 10.66, "learning_rate": 9.431780092443543e-06, "loss": 0.0916, "step": 34630 }, { "epoch": 10.66, "learning_rate": 9.43128262814319e-06, "loss": 0.0593, "step": 34631 }, { "epoch": 10.66, "learning_rate": 9.430785165254805e-06, "loss": 0.0587, "step": 34632 }, { "epoch": 10.66, "learning_rate": 9.43028770377963e-06, "loss": 0.0549, "step": 34633 }, { "epoch": 10.66, "learning_rate": 9.429790243718892e-06, "loss": 0.1052, "step": 34634 }, { "epoch": 10.66, "learning_rate": 9.429292785073833e-06, "loss": 0.0734, "step": 34635 }, { "epoch": 10.66, "learning_rate": 9.428795327845688e-06, "loss": 0.0931, "step": 34636 }, { "epoch": 10.66, "learning_rate": 9.42829787203569e-06, "loss": 0.0435, "step": 34637 }, { "epoch": 10.66, "learning_rate": 9.427800417645073e-06, "loss": 0.071, "step": 34638 }, { "epoch": 10.66, "learning_rate": 9.427302964675072e-06, "loss": 0.0913, "step": 34639 }, { "epoch": 10.66, "learning_rate": 9.426805513126926e-06, "loss": 0.0907, "step": 34640 }, { "epoch": 10.66, "learning_rate": 9.426308063001867e-06, "loss": 0.0626, "step": 34641 }, { "epoch": 10.66, "learning_rate": 9.425810614301132e-06, "loss": 0.0806, "step": 34642 }, { "epoch": 10.66, "learning_rate": 9.425313167025954e-06, "loss": 0.0316, "step": 34643 }, { "epoch": 10.66, "learning_rate": 9.42481572117756e-06, "loss": 0.1079, "step": 34644 }, { "epoch": 10.66, "learning_rate": 9.424318276757205e-06, "loss": 0.0694, "step": 34645 }, { "epoch": 10.66, "learning_rate": 9.423820833766108e-06, "loss": 0.0706, "step": 34646 }, { "epoch": 10.66, "learning_rate": 9.423323392205509e-06, "loss": 0.0432, "step": 34647 }, { "epoch": 10.66, "learning_rate": 9.422825952076641e-06, "loss": 0.037, "step": 34648 }, { "epoch": 10.66, "learning_rate": 9.422328513380741e-06, "loss": 0.049, "step": 34649 }, { "epoch": 10.66, "learning_rate": 9.421831076119043e-06, "loss": 0.0581, "step": 34650 }, { "epoch": 10.66, "learning_rate": 9.421333640292785e-06, "loss": 0.0919, "step": 34651 }, { "epoch": 10.66, "learning_rate": 9.420836205903201e-06, "loss": 0.0536, "step": 34652 }, { "epoch": 10.66, "learning_rate": 9.420338772951521e-06, "loss": 0.0478, "step": 34653 }, { "epoch": 10.66, "learning_rate": 9.41984134143898e-06, "loss": 0.0503, "step": 34654 }, { "epoch": 10.66, "learning_rate": 9.419343911366825e-06, "loss": 0.08, "step": 34655 }, { "epoch": 10.66, "learning_rate": 9.418846482736279e-06, "loss": 0.0448, "step": 34656 }, { "epoch": 10.66, "learning_rate": 9.418349055548578e-06, "loss": 0.0439, "step": 34657 }, { "epoch": 10.66, "learning_rate": 9.417851629804962e-06, "loss": 0.0465, "step": 34658 }, { "epoch": 10.66, "learning_rate": 9.417354205506663e-06, "loss": 0.0755, "step": 34659 }, { "epoch": 10.66, "learning_rate": 9.416856782654915e-06, "loss": 0.0707, "step": 34660 }, { "epoch": 10.66, "learning_rate": 9.416359361250956e-06, "loss": 0.0406, "step": 34661 }, { "epoch": 10.67, "learning_rate": 9.415861941296021e-06, "loss": 0.0609, "step": 34662 }, { "epoch": 10.67, "learning_rate": 9.415364522791343e-06, "loss": 0.0739, "step": 34663 }, { "epoch": 10.67, "learning_rate": 9.414867105738152e-06, "loss": 0.025, "step": 34664 }, { "epoch": 10.67, "learning_rate": 9.414369690137691e-06, "loss": 0.0579, "step": 34665 }, { "epoch": 10.67, "learning_rate": 9.413872275991196e-06, "loss": 0.0772, "step": 34666 }, { "epoch": 10.67, "learning_rate": 9.413374863299896e-06, "loss": 0.0507, "step": 34667 }, { "epoch": 10.67, "learning_rate": 9.412877452065028e-06, "loss": 0.1079, "step": 34668 }, { "epoch": 10.67, "learning_rate": 9.412380042287824e-06, "loss": 0.0972, "step": 34669 }, { "epoch": 10.67, "learning_rate": 9.411882633969525e-06, "loss": 0.0693, "step": 34670 }, { "epoch": 10.67, "learning_rate": 9.411385227111361e-06, "loss": 0.0542, "step": 34671 }, { "epoch": 10.67, "learning_rate": 9.410887821714571e-06, "loss": 0.0575, "step": 34672 }, { "epoch": 10.67, "learning_rate": 9.41039041778039e-06, "loss": 0.0621, "step": 34673 }, { "epoch": 10.67, "learning_rate": 9.409893015310041e-06, "loss": 0.0475, "step": 34674 }, { "epoch": 10.67, "learning_rate": 9.409395614304776e-06, "loss": 0.0474, "step": 34675 }, { "epoch": 10.67, "learning_rate": 9.408898214765824e-06, "loss": 0.0975, "step": 34676 }, { "epoch": 10.67, "learning_rate": 9.408400816694417e-06, "loss": 0.0441, "step": 34677 }, { "epoch": 10.67, "learning_rate": 9.407903420091789e-06, "loss": 0.0261, "step": 34678 }, { "epoch": 10.67, "learning_rate": 9.407406024959176e-06, "loss": 0.0541, "step": 34679 }, { "epoch": 10.67, "learning_rate": 9.406908631297816e-06, "loss": 0.0485, "step": 34680 }, { "epoch": 10.67, "learning_rate": 9.406411239108943e-06, "loss": 0.113, "step": 34681 }, { "epoch": 10.67, "learning_rate": 9.40591384839379e-06, "loss": 0.0888, "step": 34682 }, { "epoch": 10.67, "learning_rate": 9.405416459153593e-06, "loss": 0.0311, "step": 34683 }, { "epoch": 10.67, "learning_rate": 9.404919071389586e-06, "loss": 0.0962, "step": 34684 }, { "epoch": 10.67, "learning_rate": 9.404421685103004e-06, "loss": 0.0562, "step": 34685 }, { "epoch": 10.67, "learning_rate": 9.403924300295085e-06, "loss": 0.105, "step": 34686 }, { "epoch": 10.67, "learning_rate": 9.40342691696706e-06, "loss": 0.0676, "step": 34687 }, { "epoch": 10.67, "learning_rate": 9.402929535120166e-06, "loss": 0.0686, "step": 34688 }, { "epoch": 10.67, "learning_rate": 9.402432154755634e-06, "loss": 0.0838, "step": 34689 }, { "epoch": 10.67, "learning_rate": 9.401934775874704e-06, "loss": 0.0392, "step": 34690 }, { "epoch": 10.67, "learning_rate": 9.401437398478609e-06, "loss": 0.0625, "step": 34691 }, { "epoch": 10.67, "learning_rate": 9.400940022568582e-06, "loss": 0.0256, "step": 34692 }, { "epoch": 10.67, "learning_rate": 9.400442648145861e-06, "loss": 0.0527, "step": 34693 }, { "epoch": 10.68, "learning_rate": 9.399945275211677e-06, "loss": 0.0341, "step": 34694 }, { "epoch": 10.68, "learning_rate": 9.399447903767269e-06, "loss": 0.0262, "step": 34695 }, { "epoch": 10.68, "learning_rate": 9.39895053381387e-06, "loss": 0.0588, "step": 34696 }, { "epoch": 10.68, "learning_rate": 9.398453165352717e-06, "loss": 0.0557, "step": 34697 }, { "epoch": 10.68, "learning_rate": 9.39795579838504e-06, "loss": 0.0436, "step": 34698 }, { "epoch": 10.68, "learning_rate": 9.397458432912073e-06, "loss": 0.0588, "step": 34699 }, { "epoch": 10.68, "learning_rate": 9.39696106893506e-06, "loss": 0.0634, "step": 34700 }, { "epoch": 10.68, "learning_rate": 9.396463706455229e-06, "loss": 0.042, "step": 34701 }, { "epoch": 10.68, "learning_rate": 9.395966345473814e-06, "loss": 0.0955, "step": 34702 }, { "epoch": 10.68, "learning_rate": 9.395468985992052e-06, "loss": 0.0822, "step": 34703 }, { "epoch": 10.68, "learning_rate": 9.394971628011177e-06, "loss": 0.0619, "step": 34704 }, { "epoch": 10.68, "learning_rate": 9.394474271532425e-06, "loss": 0.0574, "step": 34705 }, { "epoch": 10.68, "learning_rate": 9.393976916557032e-06, "loss": 0.0651, "step": 34706 }, { "epoch": 10.68, "learning_rate": 9.393479563086231e-06, "loss": 0.1073, "step": 34707 }, { "epoch": 10.68, "learning_rate": 9.392982211121255e-06, "loss": 0.0809, "step": 34708 }, { "epoch": 10.68, "learning_rate": 9.392484860663337e-06, "loss": 0.0911, "step": 34709 }, { "epoch": 10.68, "learning_rate": 9.391987511713722e-06, "loss": 0.0625, "step": 34710 }, { "epoch": 10.68, "learning_rate": 9.391490164273635e-06, "loss": 0.032, "step": 34711 }, { "epoch": 10.68, "learning_rate": 9.390992818344314e-06, "loss": 0.0727, "step": 34712 }, { "epoch": 10.68, "learning_rate": 9.390495473926993e-06, "loss": 0.0549, "step": 34713 }, { "epoch": 10.68, "learning_rate": 9.389998131022906e-06, "loss": 0.0766, "step": 34714 }, { "epoch": 10.68, "learning_rate": 9.389500789633293e-06, "loss": 0.073, "step": 34715 }, { "epoch": 10.68, "learning_rate": 9.389003449759383e-06, "loss": 0.0691, "step": 34716 }, { "epoch": 10.68, "learning_rate": 9.388506111402413e-06, "loss": 0.0109, "step": 34717 }, { "epoch": 10.68, "learning_rate": 9.388008774563621e-06, "loss": 0.0441, "step": 34718 }, { "epoch": 10.68, "learning_rate": 9.38751143924423e-06, "loss": 0.0583, "step": 34719 }, { "epoch": 10.68, "learning_rate": 9.387014105445488e-06, "loss": 0.0443, "step": 34720 }, { "epoch": 10.68, "learning_rate": 9.386516773168626e-06, "loss": 0.0852, "step": 34721 }, { "epoch": 10.68, "learning_rate": 9.386019442414876e-06, "loss": 0.0558, "step": 34722 }, { "epoch": 10.68, "learning_rate": 9.385522113185474e-06, "loss": 0.0968, "step": 34723 }, { "epoch": 10.68, "learning_rate": 9.385024785481653e-06, "loss": 0.0607, "step": 34724 }, { "epoch": 10.68, "learning_rate": 9.384527459304653e-06, "loss": 0.1058, "step": 34725 }, { "epoch": 10.68, "learning_rate": 9.384030134655703e-06, "loss": 0.0677, "step": 34726 }, { "epoch": 10.69, "learning_rate": 9.383532811536041e-06, "loss": 0.0434, "step": 34727 }, { "epoch": 10.69, "learning_rate": 9.383035489946905e-06, "loss": 0.0426, "step": 34728 }, { "epoch": 10.69, "learning_rate": 9.382538169889516e-06, "loss": 0.058, "step": 34729 }, { "epoch": 10.69, "learning_rate": 9.382040851365125e-06, "loss": 0.1336, "step": 34730 }, { "epoch": 10.69, "learning_rate": 9.38154353437496e-06, "loss": 0.0371, "step": 34731 }, { "epoch": 10.69, "learning_rate": 9.381046218920256e-06, "loss": 0.0894, "step": 34732 }, { "epoch": 10.69, "learning_rate": 9.380548905002245e-06, "loss": 0.0592, "step": 34733 }, { "epoch": 10.69, "learning_rate": 9.380051592622164e-06, "loss": 0.0891, "step": 34734 }, { "epoch": 10.69, "learning_rate": 9.379554281781248e-06, "loss": 0.0638, "step": 34735 }, { "epoch": 10.69, "learning_rate": 9.379056972480732e-06, "loss": 0.0271, "step": 34736 }, { "epoch": 10.69, "learning_rate": 9.37855966472185e-06, "loss": 0.0549, "step": 34737 }, { "epoch": 10.69, "learning_rate": 9.378062358505838e-06, "loss": 0.0834, "step": 34738 }, { "epoch": 10.69, "learning_rate": 9.377565053833927e-06, "loss": 0.0641, "step": 34739 }, { "epoch": 10.69, "learning_rate": 9.377067750707354e-06, "loss": 0.0718, "step": 34740 }, { "epoch": 10.69, "learning_rate": 9.376570449127357e-06, "loss": 0.0848, "step": 34741 }, { "epoch": 10.69, "learning_rate": 9.376073149095169e-06, "loss": 0.0332, "step": 34742 }, { "epoch": 10.69, "learning_rate": 9.37557585061202e-06, "loss": 0.0783, "step": 34743 }, { "epoch": 10.69, "learning_rate": 9.375078553679143e-06, "loss": 0.0666, "step": 34744 }, { "epoch": 10.69, "learning_rate": 9.374581258297786e-06, "loss": 0.0509, "step": 34745 }, { "epoch": 10.69, "learning_rate": 9.374083964469172e-06, "loss": 0.0627, "step": 34746 }, { "epoch": 10.69, "learning_rate": 9.373586672194538e-06, "loss": 0.0508, "step": 34747 }, { "epoch": 10.69, "learning_rate": 9.37308938147512e-06, "loss": 0.0606, "step": 34748 }, { "epoch": 10.69, "learning_rate": 9.37259209231215e-06, "loss": 0.0433, "step": 34749 }, { "epoch": 10.69, "learning_rate": 9.372094804706867e-06, "loss": 0.0887, "step": 34750 }, { "epoch": 10.69, "learning_rate": 9.371597518660502e-06, "loss": 0.1348, "step": 34751 }, { "epoch": 10.69, "learning_rate": 9.371100234174295e-06, "loss": 0.0587, "step": 34752 }, { "epoch": 10.69, "learning_rate": 9.370602951249474e-06, "loss": 0.049, "step": 34753 }, { "epoch": 10.69, "learning_rate": 9.37010566988727e-06, "loss": 0.0559, "step": 34754 }, { "epoch": 10.69, "learning_rate": 9.369608390088933e-06, "loss": 0.0946, "step": 34755 }, { "epoch": 10.69, "learning_rate": 9.369111111855686e-06, "loss": 0.0593, "step": 34756 }, { "epoch": 10.69, "learning_rate": 9.368613835188765e-06, "loss": 0.0391, "step": 34757 }, { "epoch": 10.69, "learning_rate": 9.368116560089404e-06, "loss": 0.0533, "step": 34758 }, { "epoch": 10.7, "learning_rate": 9.36761928655884e-06, "loss": 0.0481, "step": 34759 }, { "epoch": 10.7, "learning_rate": 9.367122014598307e-06, "loss": 0.093, "step": 34760 }, { "epoch": 10.7, "learning_rate": 9.366624744209041e-06, "loss": 0.0563, "step": 34761 }, { "epoch": 10.7, "learning_rate": 9.366127475392273e-06, "loss": 0.051, "step": 34762 }, { "epoch": 10.7, "learning_rate": 9.365630208149244e-06, "loss": 0.085, "step": 34763 }, { "epoch": 10.7, "learning_rate": 9.365132942481177e-06, "loss": 0.1187, "step": 34764 }, { "epoch": 10.7, "learning_rate": 9.364635678389318e-06, "loss": 0.0186, "step": 34765 }, { "epoch": 10.7, "learning_rate": 9.364138415874898e-06, "loss": 0.0995, "step": 34766 }, { "epoch": 10.7, "learning_rate": 9.36364115493915e-06, "loss": 0.0452, "step": 34767 }, { "epoch": 10.7, "learning_rate": 9.363143895583308e-06, "loss": 0.0304, "step": 34768 }, { "epoch": 10.7, "learning_rate": 9.362646637808608e-06, "loss": 0.0486, "step": 34769 }, { "epoch": 10.7, "learning_rate": 9.362149381616287e-06, "loss": 0.055, "step": 34770 }, { "epoch": 10.7, "learning_rate": 9.361652127007575e-06, "loss": 0.059, "step": 34771 }, { "epoch": 10.7, "learning_rate": 9.36115487398371e-06, "loss": 0.0609, "step": 34772 }, { "epoch": 10.7, "learning_rate": 9.360657622545929e-06, "loss": 0.0291, "step": 34773 }, { "epoch": 10.7, "learning_rate": 9.360160372695453e-06, "loss": 0.0712, "step": 34774 }, { "epoch": 10.7, "learning_rate": 9.359663124433531e-06, "loss": 0.0387, "step": 34775 }, { "epoch": 10.7, "learning_rate": 9.359165877761396e-06, "loss": 0.0343, "step": 34776 }, { "epoch": 10.7, "learning_rate": 9.358668632680277e-06, "loss": 0.0533, "step": 34777 }, { "epoch": 10.7, "learning_rate": 9.358171389191412e-06, "loss": 0.0575, "step": 34778 }, { "epoch": 10.7, "learning_rate": 9.357674147296032e-06, "loss": 0.061, "step": 34779 }, { "epoch": 10.7, "learning_rate": 9.357176906995374e-06, "loss": 0.0953, "step": 34780 }, { "epoch": 10.7, "learning_rate": 9.356679668290675e-06, "loss": 0.0977, "step": 34781 }, { "epoch": 10.7, "learning_rate": 9.356182431183165e-06, "loss": 0.0789, "step": 34782 }, { "epoch": 10.7, "learning_rate": 9.355685195674082e-06, "loss": 0.0593, "step": 34783 }, { "epoch": 10.7, "learning_rate": 9.355187961764655e-06, "loss": 0.079, "step": 34784 }, { "epoch": 10.7, "learning_rate": 9.354690729456126e-06, "loss": 0.0911, "step": 34785 }, { "epoch": 10.7, "learning_rate": 9.354193498749728e-06, "loss": 0.1215, "step": 34786 }, { "epoch": 10.7, "learning_rate": 9.353696269646692e-06, "loss": 0.0543, "step": 34787 }, { "epoch": 10.7, "learning_rate": 9.353199042148252e-06, "loss": 0.1077, "step": 34788 }, { "epoch": 10.7, "learning_rate": 9.352701816255643e-06, "loss": 0.0618, "step": 34789 }, { "epoch": 10.7, "learning_rate": 9.352204591970105e-06, "loss": 0.0365, "step": 34790 }, { "epoch": 10.7, "learning_rate": 9.351707369292866e-06, "loss": 0.0311, "step": 34791 }, { "epoch": 10.71, "learning_rate": 9.351210148225162e-06, "loss": 0.0663, "step": 34792 }, { "epoch": 10.71, "learning_rate": 9.350712928768231e-06, "loss": 0.1217, "step": 34793 }, { "epoch": 10.71, "learning_rate": 9.350215710923301e-06, "loss": 0.0644, "step": 34794 }, { "epoch": 10.71, "learning_rate": 9.349718494691613e-06, "loss": 0.0735, "step": 34795 }, { "epoch": 10.71, "learning_rate": 9.3492212800744e-06, "loss": 0.0389, "step": 34796 }, { "epoch": 10.71, "learning_rate": 9.348724067072893e-06, "loss": 0.0673, "step": 34797 }, { "epoch": 10.71, "learning_rate": 9.34822685568833e-06, "loss": 0.0404, "step": 34798 }, { "epoch": 10.71, "learning_rate": 9.34772964592194e-06, "loss": 0.0821, "step": 34799 }, { "epoch": 10.71, "learning_rate": 9.347232437774966e-06, "loss": 0.0926, "step": 34800 }, { "epoch": 10.71, "learning_rate": 9.346735231248637e-06, "loss": 0.084, "step": 34801 }, { "epoch": 10.71, "learning_rate": 9.346238026344186e-06, "loss": 0.0729, "step": 34802 }, { "epoch": 10.71, "learning_rate": 9.345740823062852e-06, "loss": 0.0788, "step": 34803 }, { "epoch": 10.71, "learning_rate": 9.345243621405865e-06, "loss": 0.0185, "step": 34804 }, { "epoch": 10.71, "learning_rate": 9.344746421374462e-06, "loss": 0.0793, "step": 34805 }, { "epoch": 10.71, "learning_rate": 9.344249222969878e-06, "loss": 0.1063, "step": 34806 }, { "epoch": 10.71, "learning_rate": 9.34375202619335e-06, "loss": 0.0606, "step": 34807 }, { "epoch": 10.71, "learning_rate": 9.343254831046105e-06, "loss": 0.0652, "step": 34808 }, { "epoch": 10.71, "learning_rate": 9.342757637529377e-06, "loss": 0.0294, "step": 34809 }, { "epoch": 10.71, "learning_rate": 9.342260445644412e-06, "loss": 0.047, "step": 34810 }, { "epoch": 10.71, "learning_rate": 9.341763255392434e-06, "loss": 0.058, "step": 34811 }, { "epoch": 10.71, "learning_rate": 9.34126606677468e-06, "loss": 0.0434, "step": 34812 }, { "epoch": 10.71, "learning_rate": 9.340768879792386e-06, "loss": 0.1082, "step": 34813 }, { "epoch": 10.71, "learning_rate": 9.340271694446782e-06, "loss": 0.0463, "step": 34814 }, { "epoch": 10.71, "learning_rate": 9.339774510739107e-06, "loss": 0.0691, "step": 34815 }, { "epoch": 10.71, "learning_rate": 9.339277328670596e-06, "loss": 0.0477, "step": 34816 }, { "epoch": 10.71, "learning_rate": 9.33878014824248e-06, "loss": 0.0493, "step": 34817 }, { "epoch": 10.71, "learning_rate": 9.338282969455997e-06, "loss": 0.0451, "step": 34818 }, { "epoch": 10.71, "learning_rate": 9.337785792312372e-06, "loss": 0.1105, "step": 34819 }, { "epoch": 10.71, "learning_rate": 9.337288616812852e-06, "loss": 0.0735, "step": 34820 }, { "epoch": 10.71, "learning_rate": 9.336791442958667e-06, "loss": 0.0569, "step": 34821 }, { "epoch": 10.71, "learning_rate": 9.336294270751047e-06, "loss": 0.052, "step": 34822 }, { "epoch": 10.71, "learning_rate": 9.335797100191231e-06, "loss": 0.06, "step": 34823 }, { "epoch": 10.72, "learning_rate": 9.335299931280448e-06, "loss": 0.0814, "step": 34824 }, { "epoch": 10.72, "learning_rate": 9.33480276401994e-06, "loss": 0.035, "step": 34825 }, { "epoch": 10.72, "learning_rate": 9.334305598410937e-06, "loss": 0.0636, "step": 34826 }, { "epoch": 10.72, "learning_rate": 9.333808434454671e-06, "loss": 0.0352, "step": 34827 }, { "epoch": 10.72, "learning_rate": 9.333311272152385e-06, "loss": 0.0562, "step": 34828 }, { "epoch": 10.72, "learning_rate": 9.332814111505299e-06, "loss": 0.0925, "step": 34829 }, { "epoch": 10.72, "learning_rate": 9.33231695251466e-06, "loss": 0.034, "step": 34830 }, { "epoch": 10.72, "learning_rate": 9.331819795181701e-06, "loss": 0.0314, "step": 34831 }, { "epoch": 10.72, "learning_rate": 9.331322639507651e-06, "loss": 0.0212, "step": 34832 }, { "epoch": 10.72, "learning_rate": 9.330825485493746e-06, "loss": 0.0508, "step": 34833 }, { "epoch": 10.72, "learning_rate": 9.330328333141219e-06, "loss": 0.0636, "step": 34834 }, { "epoch": 10.72, "learning_rate": 9.329831182451308e-06, "loss": 0.0675, "step": 34835 }, { "epoch": 10.72, "learning_rate": 9.329334033425245e-06, "loss": 0.0859, "step": 34836 }, { "epoch": 10.72, "learning_rate": 9.328836886064267e-06, "loss": 0.0578, "step": 34837 }, { "epoch": 10.72, "learning_rate": 9.328339740369605e-06, "loss": 0.1199, "step": 34838 }, { "epoch": 10.72, "learning_rate": 9.327842596342492e-06, "loss": 0.1186, "step": 34839 }, { "epoch": 10.72, "learning_rate": 9.327345453984167e-06, "loss": 0.089, "step": 34840 }, { "epoch": 10.72, "learning_rate": 9.32684831329586e-06, "loss": 0.0576, "step": 34841 }, { "epoch": 10.72, "learning_rate": 9.326351174278813e-06, "loss": 0.145, "step": 34842 }, { "epoch": 10.72, "learning_rate": 9.325854036934249e-06, "loss": 0.0891, "step": 34843 }, { "epoch": 10.72, "learning_rate": 9.325356901263407e-06, "loss": 0.069, "step": 34844 }, { "epoch": 10.72, "learning_rate": 9.324859767267525e-06, "loss": 0.0746, "step": 34845 }, { "epoch": 10.72, "learning_rate": 9.324362634947833e-06, "loss": 0.0987, "step": 34846 }, { "epoch": 10.72, "learning_rate": 9.323865504305566e-06, "loss": 0.0327, "step": 34847 }, { "epoch": 10.72, "learning_rate": 9.32336837534196e-06, "loss": 0.0631, "step": 34848 }, { "epoch": 10.72, "learning_rate": 9.322871248058245e-06, "loss": 0.1086, "step": 34849 }, { "epoch": 10.72, "learning_rate": 9.322374122455658e-06, "loss": 0.077, "step": 34850 }, { "epoch": 10.72, "learning_rate": 9.321876998535436e-06, "loss": 0.0872, "step": 34851 }, { "epoch": 10.72, "learning_rate": 9.321379876298813e-06, "loss": 0.0688, "step": 34852 }, { "epoch": 10.72, "learning_rate": 9.320882755747017e-06, "loss": 0.0936, "step": 34853 }, { "epoch": 10.72, "learning_rate": 9.320385636881283e-06, "loss": 0.0544, "step": 34854 }, { "epoch": 10.72, "learning_rate": 9.319888519702854e-06, "loss": 0.1034, "step": 34855 }, { "epoch": 10.72, "learning_rate": 9.319391404212957e-06, "loss": 0.036, "step": 34856 }, { "epoch": 10.73, "learning_rate": 9.318894290412827e-06, "loss": 0.0629, "step": 34857 }, { "epoch": 10.73, "learning_rate": 9.3183971783037e-06, "loss": 0.0692, "step": 34858 }, { "epoch": 10.73, "learning_rate": 9.317900067886805e-06, "loss": 0.067, "step": 34859 }, { "epoch": 10.73, "learning_rate": 9.317402959163384e-06, "loss": 0.0723, "step": 34860 }, { "epoch": 10.73, "learning_rate": 9.316905852134668e-06, "loss": 0.0255, "step": 34861 }, { "epoch": 10.73, "learning_rate": 9.31640874680189e-06, "loss": 0.055, "step": 34862 }, { "epoch": 10.73, "learning_rate": 9.315911643166286e-06, "loss": 0.0316, "step": 34863 }, { "epoch": 10.73, "learning_rate": 9.315414541229082e-06, "loss": 0.0862, "step": 34864 }, { "epoch": 10.73, "learning_rate": 9.314917440991527e-06, "loss": 0.0846, "step": 34865 }, { "epoch": 10.73, "learning_rate": 9.314420342454844e-06, "loss": 0.076, "step": 34866 }, { "epoch": 10.73, "learning_rate": 9.313923245620271e-06, "loss": 0.0435, "step": 34867 }, { "epoch": 10.73, "learning_rate": 9.313426150489043e-06, "loss": 0.0363, "step": 34868 }, { "epoch": 10.73, "learning_rate": 9.31292905706239e-06, "loss": 0.0435, "step": 34869 }, { "epoch": 10.73, "learning_rate": 9.31243196534155e-06, "loss": 0.0611, "step": 34870 }, { "epoch": 10.73, "learning_rate": 9.311934875327755e-06, "loss": 0.0532, "step": 34871 }, { "epoch": 10.73, "learning_rate": 9.311437787022243e-06, "loss": 0.0769, "step": 34872 }, { "epoch": 10.73, "learning_rate": 9.310940700426242e-06, "loss": 0.0508, "step": 34873 }, { "epoch": 10.73, "learning_rate": 9.310443615540992e-06, "loss": 0.0849, "step": 34874 }, { "epoch": 10.73, "learning_rate": 9.309946532367724e-06, "loss": 0.0381, "step": 34875 }, { "epoch": 10.73, "learning_rate": 9.309449450907674e-06, "loss": 0.0608, "step": 34876 }, { "epoch": 10.73, "learning_rate": 9.308952371162076e-06, "loss": 0.0411, "step": 34877 }, { "epoch": 10.73, "learning_rate": 9.308455293132156e-06, "loss": 0.0315, "step": 34878 }, { "epoch": 10.73, "learning_rate": 9.307958216819162e-06, "loss": 0.0799, "step": 34879 }, { "epoch": 10.73, "learning_rate": 9.307461142224318e-06, "loss": 0.0885, "step": 34880 }, { "epoch": 10.73, "learning_rate": 9.306964069348863e-06, "loss": 0.0787, "step": 34881 }, { "epoch": 10.73, "learning_rate": 9.306466998194029e-06, "loss": 0.0855, "step": 34882 }, { "epoch": 10.73, "learning_rate": 9.305969928761048e-06, "loss": 0.0561, "step": 34883 }, { "epoch": 10.73, "learning_rate": 9.305472861051157e-06, "loss": 0.0588, "step": 34884 }, { "epoch": 10.73, "learning_rate": 9.304975795065594e-06, "loss": 0.056, "step": 34885 }, { "epoch": 10.73, "learning_rate": 9.304478730805587e-06, "loss": 0.0846, "step": 34886 }, { "epoch": 10.73, "learning_rate": 9.30398166827237e-06, "loss": 0.0539, "step": 34887 }, { "epoch": 10.73, "learning_rate": 9.303484607467176e-06, "loss": 0.0665, "step": 34888 }, { "epoch": 10.74, "learning_rate": 9.302987548391248e-06, "loss": 0.0465, "step": 34889 }, { "epoch": 10.74, "learning_rate": 9.302490491045812e-06, "loss": 0.0544, "step": 34890 }, { "epoch": 10.74, "learning_rate": 9.301993435432104e-06, "loss": 0.0931, "step": 34891 }, { "epoch": 10.74, "learning_rate": 9.301496381551358e-06, "loss": 0.0312, "step": 34892 }, { "epoch": 10.74, "learning_rate": 9.300999329404805e-06, "loss": 0.0846, "step": 34893 }, { "epoch": 10.74, "learning_rate": 9.300502278993687e-06, "loss": 0.0396, "step": 34894 }, { "epoch": 10.74, "learning_rate": 9.30000523031923e-06, "loss": 0.0543, "step": 34895 }, { "epoch": 10.74, "learning_rate": 9.299508183382674e-06, "loss": 0.0455, "step": 34896 }, { "epoch": 10.74, "learning_rate": 9.299011138185252e-06, "loss": 0.0418, "step": 34897 }, { "epoch": 10.74, "learning_rate": 9.298514094728188e-06, "loss": 0.0282, "step": 34898 }, { "epoch": 10.74, "learning_rate": 9.29801705301273e-06, "loss": 0.0854, "step": 34899 }, { "epoch": 10.74, "learning_rate": 9.29752001304011e-06, "loss": 0.0607, "step": 34900 }, { "epoch": 10.74, "learning_rate": 9.297022974811553e-06, "loss": 0.0448, "step": 34901 }, { "epoch": 10.74, "learning_rate": 9.2965259383283e-06, "loss": 0.0568, "step": 34902 }, { "epoch": 10.74, "learning_rate": 9.296028903591582e-06, "loss": 0.0873, "step": 34903 }, { "epoch": 10.74, "learning_rate": 9.295531870602636e-06, "loss": 0.0597, "step": 34904 }, { "epoch": 10.74, "learning_rate": 9.295034839362694e-06, "loss": 0.0445, "step": 34905 }, { "epoch": 10.74, "learning_rate": 9.29453780987299e-06, "loss": 0.0901, "step": 34906 }, { "epoch": 10.74, "learning_rate": 9.294040782134764e-06, "loss": 0.0603, "step": 34907 }, { "epoch": 10.74, "learning_rate": 9.293543756149235e-06, "loss": 0.0692, "step": 34908 }, { "epoch": 10.74, "learning_rate": 9.293046731917651e-06, "loss": 0.0927, "step": 34909 }, { "epoch": 10.74, "learning_rate": 9.292549709441243e-06, "loss": 0.0488, "step": 34910 }, { "epoch": 10.74, "learning_rate": 9.292052688721243e-06, "loss": 0.035, "step": 34911 }, { "epoch": 10.74, "learning_rate": 9.291555669758883e-06, "loss": 0.0612, "step": 34912 }, { "epoch": 10.74, "learning_rate": 9.291058652555399e-06, "loss": 0.0751, "step": 34913 }, { "epoch": 10.74, "learning_rate": 9.290561637112027e-06, "loss": 0.086, "step": 34914 }, { "epoch": 10.74, "learning_rate": 9.290064623429999e-06, "loss": 0.0641, "step": 34915 }, { "epoch": 10.74, "learning_rate": 9.289567611510549e-06, "loss": 0.0511, "step": 34916 }, { "epoch": 10.74, "learning_rate": 9.289070601354912e-06, "loss": 0.0419, "step": 34917 }, { "epoch": 10.74, "learning_rate": 9.288573592964318e-06, "loss": 0.0728, "step": 34918 }, { "epoch": 10.74, "learning_rate": 9.288076586340005e-06, "loss": 0.0762, "step": 34919 }, { "epoch": 10.74, "learning_rate": 9.287579581483209e-06, "loss": 0.0755, "step": 34920 }, { "epoch": 10.74, "learning_rate": 9.287082578395161e-06, "loss": 0.044, "step": 34921 }, { "epoch": 10.75, "learning_rate": 9.286585577077092e-06, "loss": 0.0977, "step": 34922 }, { "epoch": 10.75, "learning_rate": 9.286088577530236e-06, "loss": 0.0538, "step": 34923 }, { "epoch": 10.75, "learning_rate": 9.285591579755834e-06, "loss": 0.034, "step": 34924 }, { "epoch": 10.75, "learning_rate": 9.285094583755116e-06, "loss": 0.0914, "step": 34925 }, { "epoch": 10.75, "learning_rate": 9.284597589529313e-06, "loss": 0.0793, "step": 34926 }, { "epoch": 10.75, "learning_rate": 9.284100597079662e-06, "loss": 0.0746, "step": 34927 }, { "epoch": 10.75, "learning_rate": 9.283603606407394e-06, "loss": 0.045, "step": 34928 }, { "epoch": 10.75, "learning_rate": 9.283106617513748e-06, "loss": 0.1072, "step": 34929 }, { "epoch": 10.75, "learning_rate": 9.282609630399954e-06, "loss": 0.0391, "step": 34930 }, { "epoch": 10.75, "learning_rate": 9.28211264506725e-06, "loss": 0.0845, "step": 34931 }, { "epoch": 10.75, "learning_rate": 9.281615661516866e-06, "loss": 0.0656, "step": 34932 }, { "epoch": 10.75, "learning_rate": 9.28111867975003e-06, "loss": 0.1076, "step": 34933 }, { "epoch": 10.75, "learning_rate": 9.280621699767988e-06, "loss": 0.069, "step": 34934 }, { "epoch": 10.75, "learning_rate": 9.280124721571969e-06, "loss": 0.0725, "step": 34935 }, { "epoch": 10.75, "learning_rate": 9.279627745163204e-06, "loss": 0.0463, "step": 34936 }, { "epoch": 10.75, "learning_rate": 9.27913077054293e-06, "loss": 0.1133, "step": 34937 }, { "epoch": 10.75, "learning_rate": 9.278633797712379e-06, "loss": 0.0781, "step": 34938 }, { "epoch": 10.75, "learning_rate": 9.278136826672786e-06, "loss": 0.0669, "step": 34939 }, { "epoch": 10.75, "learning_rate": 9.277639857425387e-06, "loss": 0.0259, "step": 34940 }, { "epoch": 10.75, "learning_rate": 9.277142889971412e-06, "loss": 0.0507, "step": 34941 }, { "epoch": 10.75, "learning_rate": 9.276645924312098e-06, "loss": 0.0511, "step": 34942 }, { "epoch": 10.75, "learning_rate": 9.27614896044867e-06, "loss": 0.0785, "step": 34943 }, { "epoch": 10.75, "learning_rate": 9.275651998382377e-06, "loss": 0.0662, "step": 34944 }, { "epoch": 10.75, "learning_rate": 9.275155038114442e-06, "loss": 0.0378, "step": 34945 }, { "epoch": 10.75, "learning_rate": 9.274658079646104e-06, "loss": 0.134, "step": 34946 }, { "epoch": 10.75, "learning_rate": 9.274161122978593e-06, "loss": 0.0765, "step": 34947 }, { "epoch": 10.75, "learning_rate": 9.27366416811314e-06, "loss": 0.0768, "step": 34948 }, { "epoch": 10.75, "learning_rate": 9.273167215050987e-06, "loss": 0.0711, "step": 34949 }, { "epoch": 10.75, "learning_rate": 9.272670263793363e-06, "loss": 0.0602, "step": 34950 }, { "epoch": 10.75, "learning_rate": 9.272173314341505e-06, "loss": 0.0341, "step": 34951 }, { "epoch": 10.75, "learning_rate": 9.271676366696644e-06, "loss": 0.0875, "step": 34952 }, { "epoch": 10.75, "learning_rate": 9.271179420860008e-06, "loss": 0.0804, "step": 34953 }, { "epoch": 10.76, "learning_rate": 9.270682476832842e-06, "loss": 0.0926, "step": 34954 }, { "epoch": 10.76, "learning_rate": 9.270185534616377e-06, "loss": 0.0849, "step": 34955 }, { "epoch": 10.76, "learning_rate": 9.269688594211842e-06, "loss": 0.0795, "step": 34956 }, { "epoch": 10.76, "learning_rate": 9.269191655620474e-06, "loss": 0.069, "step": 34957 }, { "epoch": 10.76, "learning_rate": 9.268694718843503e-06, "loss": 0.0673, "step": 34958 }, { "epoch": 10.76, "learning_rate": 9.26819778388217e-06, "loss": 0.0502, "step": 34959 }, { "epoch": 10.76, "learning_rate": 9.267700850737703e-06, "loss": 0.0743, "step": 34960 }, { "epoch": 10.76, "learning_rate": 9.267203919411336e-06, "loss": 0.0655, "step": 34961 }, { "epoch": 10.76, "learning_rate": 9.266706989904305e-06, "loss": 0.0542, "step": 34962 }, { "epoch": 10.76, "learning_rate": 9.266210062217843e-06, "loss": 0.0751, "step": 34963 }, { "epoch": 10.76, "learning_rate": 9.265713136353184e-06, "loss": 0.0449, "step": 34964 }, { "epoch": 10.76, "learning_rate": 9.265216212311562e-06, "loss": 0.0579, "step": 34965 }, { "epoch": 10.76, "learning_rate": 9.26471929009421e-06, "loss": 0.0509, "step": 34966 }, { "epoch": 10.76, "learning_rate": 9.264222369702362e-06, "loss": 0.057, "step": 34967 }, { "epoch": 10.76, "learning_rate": 9.263725451137248e-06, "loss": 0.1215, "step": 34968 }, { "epoch": 10.76, "learning_rate": 9.263228534400108e-06, "loss": 0.0617, "step": 34969 }, { "epoch": 10.76, "learning_rate": 9.262731619492174e-06, "loss": 0.065, "step": 34970 }, { "epoch": 10.76, "learning_rate": 9.262234706414677e-06, "loss": 0.0687, "step": 34971 }, { "epoch": 10.76, "learning_rate": 9.261737795168853e-06, "loss": 0.0827, "step": 34972 }, { "epoch": 10.76, "learning_rate": 9.261240885755934e-06, "loss": 0.0915, "step": 34973 }, { "epoch": 10.76, "learning_rate": 9.260743978177155e-06, "loss": 0.0971, "step": 34974 }, { "epoch": 10.76, "learning_rate": 9.260247072433748e-06, "loss": 0.038, "step": 34975 }, { "epoch": 10.76, "learning_rate": 9.259750168526953e-06, "loss": 0.0452, "step": 34976 }, { "epoch": 10.76, "learning_rate": 9.259253266457998e-06, "loss": 0.0799, "step": 34977 }, { "epoch": 10.76, "learning_rate": 9.25875636622811e-06, "loss": 0.0963, "step": 34978 }, { "epoch": 10.76, "learning_rate": 9.258259467838537e-06, "loss": 0.1017, "step": 34979 }, { "epoch": 10.76, "learning_rate": 9.257762571290505e-06, "loss": 0.1252, "step": 34980 }, { "epoch": 10.76, "learning_rate": 9.257265676585247e-06, "loss": 0.0668, "step": 34981 }, { "epoch": 10.76, "learning_rate": 9.256768783723999e-06, "loss": 0.0826, "step": 34982 }, { "epoch": 10.76, "learning_rate": 9.256271892707992e-06, "loss": 0.0597, "step": 34983 }, { "epoch": 10.76, "learning_rate": 9.255775003538462e-06, "loss": 0.0622, "step": 34984 }, { "epoch": 10.76, "learning_rate": 9.255278116216643e-06, "loss": 0.0649, "step": 34985 }, { "epoch": 10.76, "learning_rate": 9.254781230743768e-06, "loss": 0.0621, "step": 34986 }, { "epoch": 10.77, "learning_rate": 9.254284347121071e-06, "loss": 0.0638, "step": 34987 }, { "epoch": 10.77, "learning_rate": 9.25378746534978e-06, "loss": 0.042, "step": 34988 }, { "epoch": 10.77, "learning_rate": 9.253290585431139e-06, "loss": 0.0536, "step": 34989 }, { "epoch": 10.77, "learning_rate": 9.252793707366374e-06, "loss": 0.0672, "step": 34990 }, { "epoch": 10.77, "learning_rate": 9.252296831156723e-06, "loss": 0.0482, "step": 34991 }, { "epoch": 10.77, "learning_rate": 9.251799956803414e-06, "loss": 0.0511, "step": 34992 }, { "epoch": 10.77, "learning_rate": 9.251303084307683e-06, "loss": 0.0876, "step": 34993 }, { "epoch": 10.77, "learning_rate": 9.250806213670767e-06, "loss": 0.0579, "step": 34994 }, { "epoch": 10.77, "learning_rate": 9.250309344893898e-06, "loss": 0.0642, "step": 34995 }, { "epoch": 10.77, "learning_rate": 9.249812477978309e-06, "loss": 0.0757, "step": 34996 }, { "epoch": 10.77, "learning_rate": 9.249315612925235e-06, "loss": 0.0897, "step": 34997 }, { "epoch": 10.77, "learning_rate": 9.2488187497359e-06, "loss": 0.0778, "step": 34998 }, { "epoch": 10.77, "learning_rate": 9.248321888411552e-06, "loss": 0.0366, "step": 34999 }, { "epoch": 10.77, "learning_rate": 9.247825028953419e-06, "loss": 0.0611, "step": 35000 }, { "epoch": 10.77, "learning_rate": 9.247328171362731e-06, "loss": 0.0372, "step": 35001 }, { "epoch": 10.77, "learning_rate": 9.246831315640727e-06, "loss": 0.0862, "step": 35002 }, { "epoch": 10.77, "learning_rate": 9.246334461788634e-06, "loss": 0.0748, "step": 35003 }, { "epoch": 10.77, "learning_rate": 9.24583760980769e-06, "loss": 0.0634, "step": 35004 }, { "epoch": 10.77, "learning_rate": 9.24534075969913e-06, "loss": 0.0844, "step": 35005 }, { "epoch": 10.77, "learning_rate": 9.244843911464185e-06, "loss": 0.0788, "step": 35006 }, { "epoch": 10.77, "learning_rate": 9.24434706510409e-06, "loss": 0.064, "step": 35007 }, { "epoch": 10.77, "learning_rate": 9.243850220620071e-06, "loss": 0.0564, "step": 35008 }, { "epoch": 10.77, "learning_rate": 9.243353378013372e-06, "loss": 0.0588, "step": 35009 }, { "epoch": 10.77, "learning_rate": 9.242856537285227e-06, "loss": 0.0344, "step": 35010 }, { "epoch": 10.77, "learning_rate": 9.242359698436862e-06, "loss": 0.0506, "step": 35011 }, { "epoch": 10.77, "learning_rate": 9.241862861469514e-06, "loss": 0.039, "step": 35012 }, { "epoch": 10.77, "learning_rate": 9.241366026384414e-06, "loss": 0.1038, "step": 35013 }, { "epoch": 10.77, "learning_rate": 9.2408691931828e-06, "loss": 0.0893, "step": 35014 }, { "epoch": 10.77, "learning_rate": 9.2403723618659e-06, "loss": 0.0734, "step": 35015 }, { "epoch": 10.77, "learning_rate": 9.239875532434954e-06, "loss": 0.0703, "step": 35016 }, { "epoch": 10.77, "learning_rate": 9.239378704891191e-06, "loss": 0.0523, "step": 35017 }, { "epoch": 10.77, "learning_rate": 9.238881879235844e-06, "loss": 0.0702, "step": 35018 }, { "epoch": 10.78, "learning_rate": 9.23838505547015e-06, "loss": 0.064, "step": 35019 }, { "epoch": 10.78, "learning_rate": 9.23788823359534e-06, "loss": 0.0839, "step": 35020 }, { "epoch": 10.78, "learning_rate": 9.23739141361265e-06, "loss": 0.0901, "step": 35021 }, { "epoch": 10.78, "learning_rate": 9.23689459552331e-06, "loss": 0.0931, "step": 35022 }, { "epoch": 10.78, "learning_rate": 9.236397779328554e-06, "loss": 0.0176, "step": 35023 }, { "epoch": 10.78, "learning_rate": 9.235900965029616e-06, "loss": 0.0693, "step": 35024 }, { "epoch": 10.78, "learning_rate": 9.235404152627731e-06, "loss": 0.0682, "step": 35025 }, { "epoch": 10.78, "learning_rate": 9.234907342124134e-06, "loss": 0.034, "step": 35026 }, { "epoch": 10.78, "learning_rate": 9.234410533520051e-06, "loss": 0.0279, "step": 35027 }, { "epoch": 10.78, "learning_rate": 9.233913726816721e-06, "loss": 0.0695, "step": 35028 }, { "epoch": 10.78, "learning_rate": 9.23341692201538e-06, "loss": 0.0715, "step": 35029 }, { "epoch": 10.78, "learning_rate": 9.232920119117256e-06, "loss": 0.0314, "step": 35030 }, { "epoch": 10.78, "learning_rate": 9.232423318123587e-06, "loss": 0.0508, "step": 35031 }, { "epoch": 10.78, "learning_rate": 9.231926519035601e-06, "loss": 0.039, "step": 35032 }, { "epoch": 10.78, "learning_rate": 9.231429721854533e-06, "loss": 0.0439, "step": 35033 }, { "epoch": 10.78, "learning_rate": 9.230932926581622e-06, "loss": 0.0588, "step": 35034 }, { "epoch": 10.78, "learning_rate": 9.230436133218095e-06, "loss": 0.0763, "step": 35035 }, { "epoch": 10.78, "learning_rate": 9.229939341765188e-06, "loss": 0.0443, "step": 35036 }, { "epoch": 10.78, "learning_rate": 9.229442552224135e-06, "loss": 0.0875, "step": 35037 }, { "epoch": 10.78, "learning_rate": 9.228945764596164e-06, "loss": 0.0723, "step": 35038 }, { "epoch": 10.78, "learning_rate": 9.228448978882517e-06, "loss": 0.0673, "step": 35039 }, { "epoch": 10.78, "learning_rate": 9.227952195084421e-06, "loss": 0.046, "step": 35040 }, { "epoch": 10.78, "learning_rate": 9.227455413203115e-06, "loss": 0.0904, "step": 35041 }, { "epoch": 10.78, "learning_rate": 9.226958633239829e-06, "loss": 0.051, "step": 35042 }, { "epoch": 10.78, "learning_rate": 9.226461855195789e-06, "loss": 0.0555, "step": 35043 }, { "epoch": 10.78, "learning_rate": 9.225965079072244e-06, "loss": 0.0427, "step": 35044 }, { "epoch": 10.78, "learning_rate": 9.225468304870415e-06, "loss": 0.0493, "step": 35045 }, { "epoch": 10.78, "learning_rate": 9.224971532591542e-06, "loss": 0.0352, "step": 35046 }, { "epoch": 10.78, "learning_rate": 9.224474762236853e-06, "loss": 0.0659, "step": 35047 }, { "epoch": 10.78, "learning_rate": 9.223977993807584e-06, "loss": 0.0742, "step": 35048 }, { "epoch": 10.78, "learning_rate": 9.22348122730497e-06, "loss": 0.0764, "step": 35049 }, { "epoch": 10.78, "learning_rate": 9.222984462730241e-06, "loss": 0.0523, "step": 35050 }, { "epoch": 10.78, "learning_rate": 9.222487700084633e-06, "loss": 0.1024, "step": 35051 }, { "epoch": 10.79, "learning_rate": 9.221990939369381e-06, "loss": 0.0643, "step": 35052 }, { "epoch": 10.79, "learning_rate": 9.22149418058571e-06, "loss": 0.0588, "step": 35053 }, { "epoch": 10.79, "learning_rate": 9.220997423734863e-06, "loss": 0.0798, "step": 35054 }, { "epoch": 10.79, "learning_rate": 9.220500668818072e-06, "loss": 0.0283, "step": 35055 }, { "epoch": 10.79, "learning_rate": 9.220003915836564e-06, "loss": 0.0546, "step": 35056 }, { "epoch": 10.79, "learning_rate": 9.219507164791577e-06, "loss": 0.0374, "step": 35057 }, { "epoch": 10.79, "learning_rate": 9.219010415684341e-06, "loss": 0.0843, "step": 35058 }, { "epoch": 10.79, "learning_rate": 9.218513668516094e-06, "loss": 0.101, "step": 35059 }, { "epoch": 10.79, "learning_rate": 9.218016923288067e-06, "loss": 0.0661, "step": 35060 }, { "epoch": 10.79, "learning_rate": 9.217520180001492e-06, "loss": 0.0601, "step": 35061 }, { "epoch": 10.79, "learning_rate": 9.217023438657606e-06, "loss": 0.0959, "step": 35062 }, { "epoch": 10.79, "learning_rate": 9.216526699257635e-06, "loss": 0.0565, "step": 35063 }, { "epoch": 10.79, "learning_rate": 9.216029961802819e-06, "loss": 0.0518, "step": 35064 }, { "epoch": 10.79, "learning_rate": 9.215533226294393e-06, "loss": 0.0834, "step": 35065 }, { "epoch": 10.79, "learning_rate": 9.215036492733583e-06, "loss": 0.0278, "step": 35066 }, { "epoch": 10.79, "learning_rate": 9.214539761121627e-06, "loss": 0.0779, "step": 35067 }, { "epoch": 10.79, "learning_rate": 9.214043031459754e-06, "loss": 0.0672, "step": 35068 }, { "epoch": 10.79, "learning_rate": 9.213546303749203e-06, "loss": 0.048, "step": 35069 }, { "epoch": 10.79, "learning_rate": 9.213049577991204e-06, "loss": 0.0472, "step": 35070 }, { "epoch": 10.79, "learning_rate": 9.212552854186992e-06, "loss": 0.0621, "step": 35071 }, { "epoch": 10.79, "learning_rate": 9.212056132337798e-06, "loss": 0.0516, "step": 35072 }, { "epoch": 10.79, "learning_rate": 9.211559412444853e-06, "loss": 0.0495, "step": 35073 }, { "epoch": 10.79, "learning_rate": 9.211062694509396e-06, "loss": 0.0919, "step": 35074 }, { "epoch": 10.79, "learning_rate": 9.21056597853266e-06, "loss": 0.0858, "step": 35075 }, { "epoch": 10.79, "learning_rate": 9.210069264515876e-06, "loss": 0.0689, "step": 35076 }, { "epoch": 10.79, "learning_rate": 9.209572552460275e-06, "loss": 0.0705, "step": 35077 }, { "epoch": 10.79, "learning_rate": 9.209075842367088e-06, "loss": 0.085, "step": 35078 }, { "epoch": 10.79, "learning_rate": 9.208579134237559e-06, "loss": 0.1051, "step": 35079 }, { "epoch": 10.79, "learning_rate": 9.208082428072913e-06, "loss": 0.048, "step": 35080 }, { "epoch": 10.79, "learning_rate": 9.207585723874384e-06, "loss": 0.036, "step": 35081 }, { "epoch": 10.79, "learning_rate": 9.207089021643207e-06, "loss": 0.061, "step": 35082 }, { "epoch": 10.79, "learning_rate": 9.206592321380611e-06, "loss": 0.0764, "step": 35083 }, { "epoch": 10.8, "learning_rate": 9.206095623087834e-06, "loss": 0.0827, "step": 35084 }, { "epoch": 10.8, "learning_rate": 9.205598926766108e-06, "loss": 0.0709, "step": 35085 }, { "epoch": 10.8, "learning_rate": 9.205102232416669e-06, "loss": 0.0669, "step": 35086 }, { "epoch": 10.8, "learning_rate": 9.204605540040745e-06, "loss": 0.0918, "step": 35087 }, { "epoch": 10.8, "learning_rate": 9.204108849639565e-06, "loss": 0.0783, "step": 35088 }, { "epoch": 10.8, "learning_rate": 9.203612161214375e-06, "loss": 0.067, "step": 35089 }, { "epoch": 10.8, "learning_rate": 9.2031154747664e-06, "loss": 0.0666, "step": 35090 }, { "epoch": 10.8, "learning_rate": 9.202618790296873e-06, "loss": 0.0645, "step": 35091 }, { "epoch": 10.8, "learning_rate": 9.20212210780703e-06, "loss": 0.0616, "step": 35092 }, { "epoch": 10.8, "learning_rate": 9.2016254272981e-06, "loss": 0.0981, "step": 35093 }, { "epoch": 10.8, "learning_rate": 9.201128748771322e-06, "loss": 0.0862, "step": 35094 }, { "epoch": 10.8, "learning_rate": 9.200632072227925e-06, "loss": 0.0097, "step": 35095 }, { "epoch": 10.8, "learning_rate": 9.200135397669144e-06, "loss": 0.0311, "step": 35096 }, { "epoch": 10.8, "learning_rate": 9.199638725096212e-06, "loss": 0.0666, "step": 35097 }, { "epoch": 10.8, "learning_rate": 9.199142054510354e-06, "loss": 0.055, "step": 35098 }, { "epoch": 10.8, "learning_rate": 9.198645385912817e-06, "loss": 0.0735, "step": 35099 }, { "epoch": 10.8, "learning_rate": 9.198148719304828e-06, "loss": 0.0371, "step": 35100 }, { "epoch": 10.8, "learning_rate": 9.197652054687619e-06, "loss": 0.0494, "step": 35101 }, { "epoch": 10.8, "learning_rate": 9.197155392062421e-06, "loss": 0.0958, "step": 35102 }, { "epoch": 10.8, "learning_rate": 9.19665873143047e-06, "loss": 0.0417, "step": 35103 }, { "epoch": 10.8, "learning_rate": 9.196162072793e-06, "loss": 0.0456, "step": 35104 }, { "epoch": 10.8, "learning_rate": 9.195665416151243e-06, "loss": 0.0358, "step": 35105 }, { "epoch": 10.8, "learning_rate": 9.195168761506432e-06, "loss": 0.0256, "step": 35106 }, { "epoch": 10.8, "learning_rate": 9.1946721088598e-06, "loss": 0.0452, "step": 35107 }, { "epoch": 10.8, "learning_rate": 9.194175458212577e-06, "loss": 0.062, "step": 35108 }, { "epoch": 10.8, "learning_rate": 9.193678809566003e-06, "loss": 0.0978, "step": 35109 }, { "epoch": 10.8, "learning_rate": 9.193182162921307e-06, "loss": 0.1203, "step": 35110 }, { "epoch": 10.8, "learning_rate": 9.192685518279721e-06, "loss": 0.0438, "step": 35111 }, { "epoch": 10.8, "learning_rate": 9.19218887564248e-06, "loss": 0.0497, "step": 35112 }, { "epoch": 10.8, "learning_rate": 9.191692235010814e-06, "loss": 0.0747, "step": 35113 }, { "epoch": 10.8, "learning_rate": 9.19119559638596e-06, "loss": 0.0597, "step": 35114 }, { "epoch": 10.8, "learning_rate": 9.19069895976915e-06, "loss": 0.0897, "step": 35115 }, { "epoch": 10.8, "learning_rate": 9.190202325161614e-06, "loss": 0.0447, "step": 35116 }, { "epoch": 10.81, "learning_rate": 9.18970569256459e-06, "loss": 0.0407, "step": 35117 }, { "epoch": 10.81, "learning_rate": 9.189209061979305e-06, "loss": 0.105, "step": 35118 }, { "epoch": 10.81, "learning_rate": 9.188712433406996e-06, "loss": 0.0919, "step": 35119 }, { "epoch": 10.81, "learning_rate": 9.188215806848898e-06, "loss": 0.0479, "step": 35120 }, { "epoch": 10.81, "learning_rate": 9.187719182306243e-06, "loss": 0.0414, "step": 35121 }, { "epoch": 10.81, "learning_rate": 9.187222559780259e-06, "loss": 0.0378, "step": 35122 }, { "epoch": 10.81, "learning_rate": 9.18672593927218e-06, "loss": 0.0667, "step": 35123 }, { "epoch": 10.81, "learning_rate": 9.186229320783244e-06, "loss": 0.0593, "step": 35124 }, { "epoch": 10.81, "learning_rate": 9.185732704314682e-06, "loss": 0.0754, "step": 35125 }, { "epoch": 10.81, "learning_rate": 9.185236089867726e-06, "loss": 0.0376, "step": 35126 }, { "epoch": 10.81, "learning_rate": 9.184739477443608e-06, "loss": 0.109, "step": 35127 }, { "epoch": 10.81, "learning_rate": 9.18424286704356e-06, "loss": 0.0294, "step": 35128 }, { "epoch": 10.81, "learning_rate": 9.18374625866882e-06, "loss": 0.0734, "step": 35129 }, { "epoch": 10.81, "learning_rate": 9.183249652320618e-06, "loss": 0.0809, "step": 35130 }, { "epoch": 10.81, "learning_rate": 9.182753048000188e-06, "loss": 0.0436, "step": 35131 }, { "epoch": 10.81, "learning_rate": 9.18225644570876e-06, "loss": 0.0818, "step": 35132 }, { "epoch": 10.81, "learning_rate": 9.181759845447566e-06, "loss": 0.0801, "step": 35133 }, { "epoch": 10.81, "learning_rate": 9.181263247217847e-06, "loss": 0.037, "step": 35134 }, { "epoch": 10.81, "learning_rate": 9.180766651020828e-06, "loss": 0.0562, "step": 35135 }, { "epoch": 10.81, "learning_rate": 9.180270056857746e-06, "loss": 0.066, "step": 35136 }, { "epoch": 10.81, "learning_rate": 9.179773464729832e-06, "loss": 0.0409, "step": 35137 }, { "epoch": 10.81, "learning_rate": 9.179276874638315e-06, "loss": 0.0555, "step": 35138 }, { "epoch": 10.81, "learning_rate": 9.178780286584437e-06, "loss": 0.0192, "step": 35139 }, { "epoch": 10.81, "learning_rate": 9.178283700569424e-06, "loss": 0.0925, "step": 35140 }, { "epoch": 10.81, "learning_rate": 9.177787116594514e-06, "loss": 0.0548, "step": 35141 }, { "epoch": 10.81, "learning_rate": 9.177290534660937e-06, "loss": 0.1418, "step": 35142 }, { "epoch": 10.81, "learning_rate": 9.17679395476992e-06, "loss": 0.0513, "step": 35143 }, { "epoch": 10.81, "learning_rate": 9.176297376922706e-06, "loss": 0.0574, "step": 35144 }, { "epoch": 10.81, "learning_rate": 9.175800801120525e-06, "loss": 0.0574, "step": 35145 }, { "epoch": 10.81, "learning_rate": 9.175304227364606e-06, "loss": 0.0446, "step": 35146 }, { "epoch": 10.81, "learning_rate": 9.174807655656185e-06, "loss": 0.0793, "step": 35147 }, { "epoch": 10.81, "learning_rate": 9.174311085996492e-06, "loss": 0.0903, "step": 35148 }, { "epoch": 10.82, "learning_rate": 9.173814518386764e-06, "loss": 0.049, "step": 35149 }, { "epoch": 10.82, "learning_rate": 9.173317952828232e-06, "loss": 0.0498, "step": 35150 }, { "epoch": 10.82, "learning_rate": 9.172821389322128e-06, "loss": 0.0753, "step": 35151 }, { "epoch": 10.82, "learning_rate": 9.17232482786969e-06, "loss": 0.0592, "step": 35152 }, { "epoch": 10.82, "learning_rate": 9.171828268472137e-06, "loss": 0.0404, "step": 35153 }, { "epoch": 10.82, "learning_rate": 9.171331711130717e-06, "loss": 0.0755, "step": 35154 }, { "epoch": 10.82, "learning_rate": 9.170835155846659e-06, "loss": 0.076, "step": 35155 }, { "epoch": 10.82, "learning_rate": 9.17033860262119e-06, "loss": 0.0837, "step": 35156 }, { "epoch": 10.82, "learning_rate": 9.169842051455549e-06, "loss": 0.0442, "step": 35157 }, { "epoch": 10.82, "learning_rate": 9.169345502350962e-06, "loss": 0.0347, "step": 35158 }, { "epoch": 10.82, "learning_rate": 9.16884895530867e-06, "loss": 0.0818, "step": 35159 }, { "epoch": 10.82, "learning_rate": 9.1683524103299e-06, "loss": 0.0588, "step": 35160 }, { "epoch": 10.82, "learning_rate": 9.16785586741589e-06, "loss": 0.043, "step": 35161 }, { "epoch": 10.82, "learning_rate": 9.167359326567867e-06, "loss": 0.0603, "step": 35162 }, { "epoch": 10.82, "learning_rate": 9.166862787787065e-06, "loss": 0.0605, "step": 35163 }, { "epoch": 10.82, "learning_rate": 9.166366251074719e-06, "loss": 0.0549, "step": 35164 }, { "epoch": 10.82, "learning_rate": 9.165869716432066e-06, "loss": 0.0647, "step": 35165 }, { "epoch": 10.82, "learning_rate": 9.165373183860329e-06, "loss": 0.0486, "step": 35166 }, { "epoch": 10.82, "learning_rate": 9.164876653360746e-06, "loss": 0.063, "step": 35167 }, { "epoch": 10.82, "learning_rate": 9.164380124934546e-06, "loss": 0.0803, "step": 35168 }, { "epoch": 10.82, "learning_rate": 9.163883598582968e-06, "loss": 0.0572, "step": 35169 }, { "epoch": 10.82, "learning_rate": 9.163387074307241e-06, "loss": 0.0421, "step": 35170 }, { "epoch": 10.82, "learning_rate": 9.1628905521086e-06, "loss": 0.0758, "step": 35171 }, { "epoch": 10.82, "learning_rate": 9.162394031988275e-06, "loss": 0.0426, "step": 35172 }, { "epoch": 10.82, "learning_rate": 9.161897513947498e-06, "loss": 0.0862, "step": 35173 }, { "epoch": 10.82, "learning_rate": 9.161400997987505e-06, "loss": 0.0799, "step": 35174 }, { "epoch": 10.82, "learning_rate": 9.160904484109528e-06, "loss": 0.0751, "step": 35175 }, { "epoch": 10.82, "learning_rate": 9.160407972314799e-06, "loss": 0.0542, "step": 35176 }, { "epoch": 10.82, "learning_rate": 9.15991146260455e-06, "loss": 0.1151, "step": 35177 }, { "epoch": 10.82, "learning_rate": 9.159414954980011e-06, "loss": 0.0753, "step": 35178 }, { "epoch": 10.82, "learning_rate": 9.158918449442425e-06, "loss": 0.0321, "step": 35179 }, { "epoch": 10.82, "learning_rate": 9.158421945993014e-06, "loss": 0.0421, "step": 35180 }, { "epoch": 10.82, "learning_rate": 9.157925444633014e-06, "loss": 0.0705, "step": 35181 }, { "epoch": 10.83, "learning_rate": 9.15742894536366e-06, "loss": 0.0654, "step": 35182 }, { "epoch": 10.83, "learning_rate": 9.156932448186179e-06, "loss": 0.0404, "step": 35183 }, { "epoch": 10.83, "learning_rate": 9.15643595310181e-06, "loss": 0.0754, "step": 35184 }, { "epoch": 10.83, "learning_rate": 9.155939460111783e-06, "loss": 0.0592, "step": 35185 }, { "epoch": 10.83, "learning_rate": 9.155442969217333e-06, "loss": 0.0972, "step": 35186 }, { "epoch": 10.83, "learning_rate": 9.154946480419688e-06, "loss": 0.0841, "step": 35187 }, { "epoch": 10.83, "learning_rate": 9.154449993720081e-06, "loss": 0.038, "step": 35188 }, { "epoch": 10.83, "learning_rate": 9.15395350911975e-06, "loss": 0.0821, "step": 35189 }, { "epoch": 10.83, "learning_rate": 9.153457026619923e-06, "loss": 0.0869, "step": 35190 }, { "epoch": 10.83, "learning_rate": 9.152960546221835e-06, "loss": 0.083, "step": 35191 }, { "epoch": 10.83, "learning_rate": 9.152464067926717e-06, "loss": 0.066, "step": 35192 }, { "epoch": 10.83, "learning_rate": 9.1519675917358e-06, "loss": 0.1146, "step": 35193 }, { "epoch": 10.83, "learning_rate": 9.151471117650321e-06, "loss": 0.0777, "step": 35194 }, { "epoch": 10.83, "learning_rate": 9.150974645671511e-06, "loss": 0.1038, "step": 35195 }, { "epoch": 10.83, "learning_rate": 9.150478175800602e-06, "loss": 0.0613, "step": 35196 }, { "epoch": 10.83, "learning_rate": 9.149981708038829e-06, "loss": 0.0787, "step": 35197 }, { "epoch": 10.83, "learning_rate": 9.149485242387414e-06, "loss": 0.1115, "step": 35198 }, { "epoch": 10.83, "learning_rate": 9.148988778847604e-06, "loss": 0.0521, "step": 35199 }, { "epoch": 10.83, "learning_rate": 9.148492317420627e-06, "loss": 0.0407, "step": 35200 }, { "epoch": 10.83, "learning_rate": 9.147995858107712e-06, "loss": 0.1239, "step": 35201 }, { "epoch": 10.83, "learning_rate": 9.147499400910093e-06, "loss": 0.0607, "step": 35202 }, { "epoch": 10.83, "learning_rate": 9.147002945829003e-06, "loss": 0.0392, "step": 35203 }, { "epoch": 10.83, "learning_rate": 9.146506492865674e-06, "loss": 0.0262, "step": 35204 }, { "epoch": 10.83, "learning_rate": 9.146010042021341e-06, "loss": 0.0648, "step": 35205 }, { "epoch": 10.83, "learning_rate": 9.145513593297234e-06, "loss": 0.0692, "step": 35206 }, { "epoch": 10.83, "learning_rate": 9.14501714669459e-06, "loss": 0.0589, "step": 35207 }, { "epoch": 10.83, "learning_rate": 9.144520702214632e-06, "loss": 0.0434, "step": 35208 }, { "epoch": 10.83, "learning_rate": 9.1440242598586e-06, "loss": 0.0303, "step": 35209 }, { "epoch": 10.83, "learning_rate": 9.143527819627729e-06, "loss": 0.0471, "step": 35210 }, { "epoch": 10.83, "learning_rate": 9.143031381523243e-06, "loss": 0.0371, "step": 35211 }, { "epoch": 10.83, "learning_rate": 9.142534945546381e-06, "loss": 0.056, "step": 35212 }, { "epoch": 10.83, "learning_rate": 9.142534945546381e-06, "loss": 0.0856, "step": 35213 }, { "epoch": 10.84, "learning_rate": 9.142038511698372e-06, "loss": 0.0579, "step": 35214 }, { "epoch": 10.84, "learning_rate": 9.141542079980453e-06, "loss": 0.1083, "step": 35215 }, { "epoch": 10.84, "learning_rate": 9.141045650393851e-06, "loss": 0.0507, "step": 35216 }, { "epoch": 10.84, "learning_rate": 9.140549222939803e-06, "loss": 0.0658, "step": 35217 }, { "epoch": 10.84, "learning_rate": 9.140052797619538e-06, "loss": 0.0839, "step": 35218 }, { "epoch": 10.84, "learning_rate": 9.139556374434288e-06, "loss": 0.0619, "step": 35219 }, { "epoch": 10.84, "learning_rate": 9.139059953385293e-06, "loss": 0.0631, "step": 35220 }, { "epoch": 10.84, "learning_rate": 9.138563534473776e-06, "loss": 0.0356, "step": 35221 }, { "epoch": 10.84, "learning_rate": 9.138067117700977e-06, "loss": 0.0756, "step": 35222 }, { "epoch": 10.84, "learning_rate": 9.137570703068122e-06, "loss": 0.0536, "step": 35223 }, { "epoch": 10.84, "learning_rate": 9.137074290576446e-06, "loss": 0.0429, "step": 35224 }, { "epoch": 10.84, "learning_rate": 9.136577880227183e-06, "loss": 0.1038, "step": 35225 }, { "epoch": 10.84, "learning_rate": 9.136081472021565e-06, "loss": 0.0952, "step": 35226 }, { "epoch": 10.84, "learning_rate": 9.135585065960824e-06, "loss": 0.0551, "step": 35227 }, { "epoch": 10.84, "learning_rate": 9.13508866204619e-06, "loss": 0.0916, "step": 35228 }, { "epoch": 10.84, "learning_rate": 9.134592260278898e-06, "loss": 0.0429, "step": 35229 }, { "epoch": 10.84, "learning_rate": 9.134095860660182e-06, "loss": 0.0795, "step": 35230 }, { "epoch": 10.84, "learning_rate": 9.133599463191272e-06, "loss": 0.0691, "step": 35231 }, { "epoch": 10.84, "learning_rate": 9.133103067873403e-06, "loss": 0.0624, "step": 35232 }, { "epoch": 10.84, "learning_rate": 9.132606674707803e-06, "loss": 0.0539, "step": 35233 }, { "epoch": 10.84, "learning_rate": 9.132110283695702e-06, "loss": 0.0594, "step": 35234 }, { "epoch": 10.84, "learning_rate": 9.131613894838345e-06, "loss": 0.0545, "step": 35235 }, { "epoch": 10.84, "learning_rate": 9.131117508136953e-06, "loss": 0.0483, "step": 35236 }, { "epoch": 10.84, "learning_rate": 9.130621123592764e-06, "loss": 0.0478, "step": 35237 }, { "epoch": 10.84, "learning_rate": 9.130124741207007e-06, "loss": 0.0918, "step": 35238 }, { "epoch": 10.84, "learning_rate": 9.129628360980915e-06, "loss": 0.0965, "step": 35239 }, { "epoch": 10.84, "learning_rate": 9.129131982915721e-06, "loss": 0.0783, "step": 35240 }, { "epoch": 10.84, "learning_rate": 9.12863560701266e-06, "loss": 0.1062, "step": 35241 }, { "epoch": 10.84, "learning_rate": 9.128139233272958e-06, "loss": 0.0854, "step": 35242 }, { "epoch": 10.84, "learning_rate": 9.127642861697858e-06, "loss": 0.0519, "step": 35243 }, { "epoch": 10.84, "learning_rate": 9.127146492288576e-06, "loss": 0.0706, "step": 35244 }, { "epoch": 10.84, "learning_rate": 9.126650125046361e-06, "loss": 0.0465, "step": 35245 }, { "epoch": 10.84, "learning_rate": 9.126153759972437e-06, "loss": 0.0376, "step": 35246 }, { "epoch": 10.85, "learning_rate": 9.125657397068038e-06, "loss": 0.0695, "step": 35247 }, { "epoch": 10.85, "learning_rate": 9.125161036334395e-06, "loss": 0.0636, "step": 35248 }, { "epoch": 10.85, "learning_rate": 9.12466467777274e-06, "loss": 0.0456, "step": 35249 }, { "epoch": 10.85, "learning_rate": 9.124168321384306e-06, "loss": 0.0472, "step": 35250 }, { "epoch": 10.85, "learning_rate": 9.123671967170328e-06, "loss": 0.0351, "step": 35251 }, { "epoch": 10.85, "learning_rate": 9.123175615132038e-06, "loss": 0.0961, "step": 35252 }, { "epoch": 10.85, "learning_rate": 9.122679265270666e-06, "loss": 0.0826, "step": 35253 }, { "epoch": 10.85, "learning_rate": 9.12218291758744e-06, "loss": 0.1067, "step": 35254 }, { "epoch": 10.85, "learning_rate": 9.1216865720836e-06, "loss": 0.1137, "step": 35255 }, { "epoch": 10.85, "learning_rate": 9.12119022876038e-06, "loss": 0.0464, "step": 35256 }, { "epoch": 10.85, "learning_rate": 9.120693887619003e-06, "loss": 0.0729, "step": 35257 }, { "epoch": 10.85, "learning_rate": 9.120197548660707e-06, "loss": 0.1149, "step": 35258 }, { "epoch": 10.85, "learning_rate": 9.119701211886723e-06, "loss": 0.0662, "step": 35259 }, { "epoch": 10.85, "learning_rate": 9.119204877298281e-06, "loss": 0.0965, "step": 35260 }, { "epoch": 10.85, "learning_rate": 9.11870854489662e-06, "loss": 0.0559, "step": 35261 }, { "epoch": 10.85, "learning_rate": 9.118212214682968e-06, "loss": 0.0651, "step": 35262 }, { "epoch": 10.85, "learning_rate": 9.117715886658558e-06, "loss": 0.0583, "step": 35263 }, { "epoch": 10.85, "learning_rate": 9.117219560824616e-06, "loss": 0.0621, "step": 35264 }, { "epoch": 10.85, "learning_rate": 9.116723237182383e-06, "loss": 0.0538, "step": 35265 }, { "epoch": 10.85, "learning_rate": 9.116226915733091e-06, "loss": 0.0653, "step": 35266 }, { "epoch": 10.85, "learning_rate": 9.115730596477967e-06, "loss": 0.0876, "step": 35267 }, { "epoch": 10.85, "learning_rate": 9.115234279418241e-06, "loss": 0.0401, "step": 35268 }, { "epoch": 10.85, "learning_rate": 9.114737964555157e-06, "loss": 0.0265, "step": 35269 }, { "epoch": 10.85, "learning_rate": 9.114241651889936e-06, "loss": 0.0794, "step": 35270 }, { "epoch": 10.85, "learning_rate": 9.113745341423816e-06, "loss": 0.0718, "step": 35271 }, { "epoch": 10.85, "learning_rate": 9.113249033158028e-06, "loss": 0.0536, "step": 35272 }, { "epoch": 10.85, "learning_rate": 9.112752727093799e-06, "loss": 0.0999, "step": 35273 }, { "epoch": 10.85, "learning_rate": 9.11225642323237e-06, "loss": 0.0659, "step": 35274 }, { "epoch": 10.85, "learning_rate": 9.111760121574968e-06, "loss": 0.0419, "step": 35275 }, { "epoch": 10.85, "learning_rate": 9.111263822122827e-06, "loss": 0.0726, "step": 35276 }, { "epoch": 10.85, "learning_rate": 9.11076752487718e-06, "loss": 0.0583, "step": 35277 }, { "epoch": 10.85, "learning_rate": 9.11027122983925e-06, "loss": 0.051, "step": 35278 }, { "epoch": 10.86, "learning_rate": 9.109774937010283e-06, "loss": 0.0367, "step": 35279 }, { "epoch": 10.86, "learning_rate": 9.109278646391505e-06, "loss": 0.0707, "step": 35280 }, { "epoch": 10.86, "learning_rate": 9.108782357984147e-06, "loss": 0.0837, "step": 35281 }, { "epoch": 10.86, "learning_rate": 9.108286071789443e-06, "loss": 0.116, "step": 35282 }, { "epoch": 10.86, "learning_rate": 9.107789787808621e-06, "loss": 0.0578, "step": 35283 }, { "epoch": 10.86, "learning_rate": 9.107293506042918e-06, "loss": 0.0702, "step": 35284 }, { "epoch": 10.86, "learning_rate": 9.106797226493565e-06, "loss": 0.1037, "step": 35285 }, { "epoch": 10.86, "learning_rate": 9.106300949161793e-06, "loss": 0.052, "step": 35286 }, { "epoch": 10.86, "learning_rate": 9.105804674048839e-06, "loss": 0.0375, "step": 35287 }, { "epoch": 10.86, "learning_rate": 9.105308401155923e-06, "loss": 0.0548, "step": 35288 }, { "epoch": 10.86, "learning_rate": 9.10481213048429e-06, "loss": 0.0966, "step": 35289 }, { "epoch": 10.86, "learning_rate": 9.10431586203517e-06, "loss": 0.0811, "step": 35290 }, { "epoch": 10.86, "learning_rate": 9.103819595809788e-06, "loss": 0.0574, "step": 35291 }, { "epoch": 10.86, "learning_rate": 9.103323331809381e-06, "loss": 0.0475, "step": 35292 }, { "epoch": 10.86, "learning_rate": 9.102827070035179e-06, "loss": 0.0733, "step": 35293 }, { "epoch": 10.86, "learning_rate": 9.102330810488417e-06, "loss": 0.0514, "step": 35294 }, { "epoch": 10.86, "learning_rate": 9.101834553170326e-06, "loss": 0.0355, "step": 35295 }, { "epoch": 10.86, "learning_rate": 9.101338298082137e-06, "loss": 0.0548, "step": 35296 }, { "epoch": 10.86, "learning_rate": 9.100842045225084e-06, "loss": 0.0459, "step": 35297 }, { "epoch": 10.86, "learning_rate": 9.100345794600395e-06, "loss": 0.0618, "step": 35298 }, { "epoch": 10.86, "learning_rate": 9.099849546209307e-06, "loss": 0.0575, "step": 35299 }, { "epoch": 10.86, "learning_rate": 9.09935330005305e-06, "loss": 0.0606, "step": 35300 }, { "epoch": 10.86, "learning_rate": 9.098857056132857e-06, "loss": 0.065, "step": 35301 }, { "epoch": 10.86, "learning_rate": 9.098360814449958e-06, "loss": 0.0404, "step": 35302 }, { "epoch": 10.86, "learning_rate": 9.097864575005583e-06, "loss": 0.0811, "step": 35303 }, { "epoch": 10.86, "learning_rate": 9.09736833780097e-06, "loss": 0.0779, "step": 35304 }, { "epoch": 10.86, "learning_rate": 9.096872102837348e-06, "loss": 0.0623, "step": 35305 }, { "epoch": 10.86, "learning_rate": 9.09637587011595e-06, "loss": 0.0486, "step": 35306 }, { "epoch": 10.86, "learning_rate": 9.095879639638006e-06, "loss": 0.0749, "step": 35307 }, { "epoch": 10.86, "learning_rate": 9.095383411404747e-06, "loss": 0.0831, "step": 35308 }, { "epoch": 10.86, "learning_rate": 9.094887185417409e-06, "loss": 0.092, "step": 35309 }, { "epoch": 10.86, "learning_rate": 9.094390961677223e-06, "loss": 0.0769, "step": 35310 }, { "epoch": 10.86, "learning_rate": 9.093894740185421e-06, "loss": 0.044, "step": 35311 }, { "epoch": 10.87, "learning_rate": 9.093398520943233e-06, "loss": 0.0684, "step": 35312 }, { "epoch": 10.87, "learning_rate": 9.092902303951889e-06, "loss": 0.0828, "step": 35313 }, { "epoch": 10.87, "learning_rate": 9.092406089212628e-06, "loss": 0.0635, "step": 35314 }, { "epoch": 10.87, "learning_rate": 9.091909876726676e-06, "loss": 0.0841, "step": 35315 }, { "epoch": 10.87, "learning_rate": 9.091413666495267e-06, "loss": 0.0635, "step": 35316 }, { "epoch": 10.87, "learning_rate": 9.090917458519635e-06, "loss": 0.1137, "step": 35317 }, { "epoch": 10.87, "learning_rate": 9.090421252801006e-06, "loss": 0.0769, "step": 35318 }, { "epoch": 10.87, "learning_rate": 9.089925049340619e-06, "loss": 0.0724, "step": 35319 }, { "epoch": 10.87, "learning_rate": 9.089428848139702e-06, "loss": 0.0465, "step": 35320 }, { "epoch": 10.87, "learning_rate": 9.08893264919949e-06, "loss": 0.0581, "step": 35321 }, { "epoch": 10.87, "learning_rate": 9.08843645252121e-06, "loss": 0.081, "step": 35322 }, { "epoch": 10.87, "learning_rate": 9.087940258106093e-06, "loss": 0.064, "step": 35323 }, { "epoch": 10.87, "learning_rate": 9.08744406595538e-06, "loss": 0.0772, "step": 35324 }, { "epoch": 10.87, "learning_rate": 9.086947876070295e-06, "loss": 0.0751, "step": 35325 }, { "epoch": 10.87, "learning_rate": 9.086451688452074e-06, "loss": 0.0591, "step": 35326 }, { "epoch": 10.87, "learning_rate": 9.085955503101947e-06, "loss": 0.0216, "step": 35327 }, { "epoch": 10.87, "learning_rate": 9.085459320021142e-06, "loss": 0.0804, "step": 35328 }, { "epoch": 10.87, "learning_rate": 9.084963139210899e-06, "loss": 0.0812, "step": 35329 }, { "epoch": 10.87, "learning_rate": 9.084466960672444e-06, "loss": 0.0471, "step": 35330 }, { "epoch": 10.87, "learning_rate": 9.083970784407012e-06, "loss": 0.0513, "step": 35331 }, { "epoch": 10.87, "learning_rate": 9.083474610415836e-06, "loss": 0.0881, "step": 35332 }, { "epoch": 10.87, "learning_rate": 9.082978438700138e-06, "loss": 0.074, "step": 35333 }, { "epoch": 10.87, "learning_rate": 9.082482269261163e-06, "loss": 0.0334, "step": 35334 }, { "epoch": 10.87, "learning_rate": 9.081986102100137e-06, "loss": 0.0483, "step": 35335 }, { "epoch": 10.87, "learning_rate": 9.081489937218293e-06, "loss": 0.101, "step": 35336 }, { "epoch": 10.87, "learning_rate": 9.080993774616859e-06, "loss": 0.1156, "step": 35337 }, { "epoch": 10.87, "learning_rate": 9.08049761429707e-06, "loss": 0.0245, "step": 35338 }, { "epoch": 10.87, "learning_rate": 9.080001456260157e-06, "loss": 0.0507, "step": 35339 }, { "epoch": 10.87, "learning_rate": 9.079505300507354e-06, "loss": 0.0705, "step": 35340 }, { "epoch": 10.87, "learning_rate": 9.079009147039892e-06, "loss": 0.0553, "step": 35341 }, { "epoch": 10.87, "learning_rate": 9.078512995859002e-06, "loss": 0.0609, "step": 35342 }, { "epoch": 10.87, "learning_rate": 9.07801684696591e-06, "loss": 0.0806, "step": 35343 }, { "epoch": 10.88, "learning_rate": 9.07752070036186e-06, "loss": 0.0435, "step": 35344 }, { "epoch": 10.88, "learning_rate": 9.077024556048077e-06, "loss": 0.027, "step": 35345 }, { "epoch": 10.88, "learning_rate": 9.076528414025792e-06, "loss": 0.0615, "step": 35346 }, { "epoch": 10.88, "learning_rate": 9.076032274296238e-06, "loss": 0.061, "step": 35347 }, { "epoch": 10.88, "learning_rate": 9.075536136860644e-06, "loss": 0.0781, "step": 35348 }, { "epoch": 10.88, "learning_rate": 9.075040001720247e-06, "loss": 0.0576, "step": 35349 }, { "epoch": 10.88, "learning_rate": 9.074543868876277e-06, "loss": 0.0719, "step": 35350 }, { "epoch": 10.88, "learning_rate": 9.074047738329965e-06, "loss": 0.0577, "step": 35351 }, { "epoch": 10.88, "learning_rate": 9.073551610082543e-06, "loss": 0.0676, "step": 35352 }, { "epoch": 10.88, "learning_rate": 9.073055484135238e-06, "loss": 0.0858, "step": 35353 }, { "epoch": 10.88, "learning_rate": 9.072559360489292e-06, "loss": 0.0667, "step": 35354 }, { "epoch": 10.88, "learning_rate": 9.072063239145929e-06, "loss": 0.0975, "step": 35355 }, { "epoch": 10.88, "learning_rate": 9.071567120106386e-06, "loss": 0.0771, "step": 35356 }, { "epoch": 10.88, "learning_rate": 9.071071003371888e-06, "loss": 0.044, "step": 35357 }, { "epoch": 10.88, "learning_rate": 9.070574888943668e-06, "loss": 0.1122, "step": 35358 }, { "epoch": 10.88, "learning_rate": 9.070078776822965e-06, "loss": 0.0452, "step": 35359 }, { "epoch": 10.88, "learning_rate": 9.069582667011003e-06, "loss": 0.0858, "step": 35360 }, { "epoch": 10.88, "learning_rate": 9.069086559509017e-06, "loss": 0.0844, "step": 35361 }, { "epoch": 10.88, "learning_rate": 9.06859045431824e-06, "loss": 0.0611, "step": 35362 }, { "epoch": 10.88, "learning_rate": 9.068094351439896e-06, "loss": 0.0404, "step": 35363 }, { "epoch": 10.88, "learning_rate": 9.067598250875227e-06, "loss": 0.0392, "step": 35364 }, { "epoch": 10.88, "learning_rate": 9.06710215262546e-06, "loss": 0.0863, "step": 35365 }, { "epoch": 10.88, "learning_rate": 9.066606056691829e-06, "loss": 0.0244, "step": 35366 }, { "epoch": 10.88, "learning_rate": 9.06610996307556e-06, "loss": 0.0313, "step": 35367 }, { "epoch": 10.88, "learning_rate": 9.065613871777886e-06, "loss": 0.0672, "step": 35368 }, { "epoch": 10.88, "learning_rate": 9.065117782800046e-06, "loss": 0.0746, "step": 35369 }, { "epoch": 10.88, "learning_rate": 9.064621696143266e-06, "loss": 0.0457, "step": 35370 }, { "epoch": 10.88, "learning_rate": 9.064125611808776e-06, "loss": 0.0534, "step": 35371 }, { "epoch": 10.88, "learning_rate": 9.06362952979781e-06, "loss": 0.0858, "step": 35372 }, { "epoch": 10.88, "learning_rate": 9.063133450111599e-06, "loss": 0.0545, "step": 35373 }, { "epoch": 10.88, "learning_rate": 9.062637372751377e-06, "loss": 0.0603, "step": 35374 }, { "epoch": 10.88, "learning_rate": 9.062141297718372e-06, "loss": 0.0627, "step": 35375 }, { "epoch": 10.88, "learning_rate": 9.061645225013817e-06, "loss": 0.0624, "step": 35376 }, { "epoch": 10.89, "learning_rate": 9.061149154638949e-06, "loss": 0.065, "step": 35377 }, { "epoch": 10.89, "learning_rate": 9.060653086594986e-06, "loss": 0.0522, "step": 35378 }, { "epoch": 10.89, "learning_rate": 9.060157020883173e-06, "loss": 0.0558, "step": 35379 }, { "epoch": 10.89, "learning_rate": 9.059660957504737e-06, "loss": 0.1193, "step": 35380 }, { "epoch": 10.89, "learning_rate": 9.05916489646091e-06, "loss": 0.0706, "step": 35381 }, { "epoch": 10.89, "learning_rate": 9.058668837752921e-06, "loss": 0.0454, "step": 35382 }, { "epoch": 10.89, "learning_rate": 9.058172781382003e-06, "loss": 0.0758, "step": 35383 }, { "epoch": 10.89, "learning_rate": 9.05767672734939e-06, "loss": 0.0729, "step": 35384 }, { "epoch": 10.89, "learning_rate": 9.05718067565631e-06, "loss": 0.0444, "step": 35385 }, { "epoch": 10.89, "learning_rate": 9.056684626303998e-06, "loss": 0.0763, "step": 35386 }, { "epoch": 10.89, "learning_rate": 9.056188579293685e-06, "loss": 0.0403, "step": 35387 }, { "epoch": 10.89, "learning_rate": 9.055692534626595e-06, "loss": 0.0386, "step": 35388 }, { "epoch": 10.89, "learning_rate": 9.05519649230397e-06, "loss": 0.1046, "step": 35389 }, { "epoch": 10.89, "learning_rate": 9.05470045232704e-06, "loss": 0.0986, "step": 35390 }, { "epoch": 10.89, "learning_rate": 9.054204414697033e-06, "loss": 0.1101, "step": 35391 }, { "epoch": 10.89, "learning_rate": 9.053708379415178e-06, "loss": 0.0475, "step": 35392 }, { "epoch": 10.89, "learning_rate": 9.053212346482711e-06, "loss": 0.0812, "step": 35393 }, { "epoch": 10.89, "learning_rate": 9.052716315900862e-06, "loss": 0.0759, "step": 35394 }, { "epoch": 10.89, "learning_rate": 9.052220287670865e-06, "loss": 0.0669, "step": 35395 }, { "epoch": 10.89, "learning_rate": 9.05172426179395e-06, "loss": 0.0292, "step": 35396 }, { "epoch": 10.89, "learning_rate": 9.051228238271346e-06, "loss": 0.0415, "step": 35397 }, { "epoch": 10.89, "learning_rate": 9.050732217104286e-06, "loss": 0.1026, "step": 35398 }, { "epoch": 10.89, "learning_rate": 9.050236198294003e-06, "loss": 0.0846, "step": 35399 }, { "epoch": 10.89, "learning_rate": 9.049740181841731e-06, "loss": 0.0631, "step": 35400 }, { "epoch": 10.89, "learning_rate": 9.049244167748694e-06, "loss": 0.0651, "step": 35401 }, { "epoch": 10.89, "learning_rate": 9.048748156016128e-06, "loss": 0.0869, "step": 35402 }, { "epoch": 10.89, "learning_rate": 9.048252146645262e-06, "loss": 0.0888, "step": 35403 }, { "epoch": 10.89, "learning_rate": 9.047756139637332e-06, "loss": 0.0672, "step": 35404 }, { "epoch": 10.89, "learning_rate": 9.047260134993566e-06, "loss": 0.046, "step": 35405 }, { "epoch": 10.89, "learning_rate": 9.046764132715197e-06, "loss": 0.0645, "step": 35406 }, { "epoch": 10.89, "learning_rate": 9.046268132803456e-06, "loss": 0.0698, "step": 35407 }, { "epoch": 10.89, "learning_rate": 9.04577213525957e-06, "loss": 0.0552, "step": 35408 }, { "epoch": 10.9, "learning_rate": 9.045276140084778e-06, "loss": 0.0651, "step": 35409 }, { "epoch": 10.9, "learning_rate": 9.044780147280307e-06, "loss": 0.0938, "step": 35410 }, { "epoch": 10.9, "learning_rate": 9.044284156847393e-06, "loss": 0.0889, "step": 35411 }, { "epoch": 10.9, "learning_rate": 9.043788168787261e-06, "loss": 0.083, "step": 35412 }, { "epoch": 10.9, "learning_rate": 9.04329218310114e-06, "loss": 0.0747, "step": 35413 }, { "epoch": 10.9, "learning_rate": 9.042796199790274e-06, "loss": 0.0696, "step": 35414 }, { "epoch": 10.9, "learning_rate": 9.042300218855887e-06, "loss": 0.076, "step": 35415 }, { "epoch": 10.9, "learning_rate": 9.041804240299206e-06, "loss": 0.0549, "step": 35416 }, { "epoch": 10.9, "learning_rate": 9.04130826412147e-06, "loss": 0.051, "step": 35417 }, { "epoch": 10.9, "learning_rate": 9.040812290323903e-06, "loss": 0.0722, "step": 35418 }, { "epoch": 10.9, "learning_rate": 9.040316318907745e-06, "loss": 0.0567, "step": 35419 }, { "epoch": 10.9, "learning_rate": 9.03982034987422e-06, "loss": 0.1037, "step": 35420 }, { "epoch": 10.9, "learning_rate": 9.039324383224565e-06, "loss": 0.103, "step": 35421 }, { "epoch": 10.9, "learning_rate": 9.038828418960008e-06, "loss": 0.0782, "step": 35422 }, { "epoch": 10.9, "learning_rate": 9.038332457081776e-06, "loss": 0.0768, "step": 35423 }, { "epoch": 10.9, "learning_rate": 9.037836497591111e-06, "loss": 0.0595, "step": 35424 }, { "epoch": 10.9, "learning_rate": 9.037340540489239e-06, "loss": 0.0504, "step": 35425 }, { "epoch": 10.9, "learning_rate": 9.036844585777386e-06, "loss": 0.0647, "step": 35426 }, { "epoch": 10.9, "learning_rate": 9.036348633456791e-06, "loss": 0.0664, "step": 35427 }, { "epoch": 10.9, "learning_rate": 9.035852683528681e-06, "loss": 0.0493, "step": 35428 }, { "epoch": 10.9, "learning_rate": 9.03535673599429e-06, "loss": 0.0836, "step": 35429 }, { "epoch": 10.9, "learning_rate": 9.034860790854848e-06, "loss": 0.0612, "step": 35430 }, { "epoch": 10.9, "learning_rate": 9.034364848111589e-06, "loss": 0.0995, "step": 35431 }, { "epoch": 10.9, "learning_rate": 9.033868907765742e-06, "loss": 0.0937, "step": 35432 }, { "epoch": 10.9, "learning_rate": 9.03337296981853e-06, "loss": 0.0783, "step": 35433 }, { "epoch": 10.9, "learning_rate": 9.032877034271199e-06, "loss": 0.09, "step": 35434 }, { "epoch": 10.9, "learning_rate": 9.032381101124975e-06, "loss": 0.0554, "step": 35435 }, { "epoch": 10.9, "learning_rate": 9.031885170381087e-06, "loss": 0.0508, "step": 35436 }, { "epoch": 10.9, "learning_rate": 9.031389242040764e-06, "loss": 0.0653, "step": 35437 }, { "epoch": 10.9, "learning_rate": 9.03089331610524e-06, "loss": 0.0411, "step": 35438 }, { "epoch": 10.9, "learning_rate": 9.03039739257575e-06, "loss": 0.0391, "step": 35439 }, { "epoch": 10.9, "learning_rate": 9.02990147145352e-06, "loss": 0.0509, "step": 35440 }, { "epoch": 10.9, "learning_rate": 9.029405552739783e-06, "loss": 0.0372, "step": 35441 }, { "epoch": 10.91, "learning_rate": 9.028909636435775e-06, "loss": 0.0525, "step": 35442 }, { "epoch": 10.91, "learning_rate": 9.028413722542715e-06, "loss": 0.0602, "step": 35443 }, { "epoch": 10.91, "learning_rate": 9.027917811061843e-06, "loss": 0.0708, "step": 35444 }, { "epoch": 10.91, "learning_rate": 9.027421901994395e-06, "loss": 0.0588, "step": 35445 }, { "epoch": 10.91, "learning_rate": 9.026925995341593e-06, "loss": 0.0343, "step": 35446 }, { "epoch": 10.91, "learning_rate": 9.026430091104671e-06, "loss": 0.0631, "step": 35447 }, { "epoch": 10.91, "learning_rate": 9.025934189284858e-06, "loss": 0.0608, "step": 35448 }, { "epoch": 10.91, "learning_rate": 9.025438289883391e-06, "loss": 0.1173, "step": 35449 }, { "epoch": 10.91, "learning_rate": 9.024942392901498e-06, "loss": 0.0754, "step": 35450 }, { "epoch": 10.91, "learning_rate": 9.024446498340408e-06, "loss": 0.0991, "step": 35451 }, { "epoch": 10.91, "learning_rate": 9.023950606201357e-06, "loss": 0.0973, "step": 35452 }, { "epoch": 10.91, "learning_rate": 9.023454716485572e-06, "loss": 0.0723, "step": 35453 }, { "epoch": 10.91, "learning_rate": 9.022958829194284e-06, "loss": 0.0902, "step": 35454 }, { "epoch": 10.91, "learning_rate": 9.02246294432873e-06, "loss": 0.0632, "step": 35455 }, { "epoch": 10.91, "learning_rate": 9.021967061890137e-06, "loss": 0.0598, "step": 35456 }, { "epoch": 10.91, "learning_rate": 9.021471181879735e-06, "loss": 0.0679, "step": 35457 }, { "epoch": 10.91, "learning_rate": 9.020975304298752e-06, "loss": 0.0894, "step": 35458 }, { "epoch": 10.91, "learning_rate": 9.020479429148427e-06, "loss": 0.0748, "step": 35459 }, { "epoch": 10.91, "learning_rate": 9.019983556429987e-06, "loss": 0.0888, "step": 35460 }, { "epoch": 10.91, "learning_rate": 9.019487686144663e-06, "loss": 0.081, "step": 35461 }, { "epoch": 10.91, "learning_rate": 9.018991818293688e-06, "loss": 0.0366, "step": 35462 }, { "epoch": 10.91, "learning_rate": 9.018495952878291e-06, "loss": 0.0604, "step": 35463 }, { "epoch": 10.91, "learning_rate": 9.018000089899704e-06, "loss": 0.1075, "step": 35464 }, { "epoch": 10.91, "learning_rate": 9.017504229359157e-06, "loss": 0.0374, "step": 35465 }, { "epoch": 10.91, "learning_rate": 9.017008371257887e-06, "loss": 0.07, "step": 35466 }, { "epoch": 10.91, "learning_rate": 9.016512515597117e-06, "loss": 0.0491, "step": 35467 }, { "epoch": 10.91, "learning_rate": 9.016016662378076e-06, "loss": 0.0524, "step": 35468 }, { "epoch": 10.91, "learning_rate": 9.015520811602008e-06, "loss": 0.0892, "step": 35469 }, { "epoch": 10.91, "learning_rate": 9.015024963270133e-06, "loss": 0.0696, "step": 35470 }, { "epoch": 10.91, "learning_rate": 9.014529117383687e-06, "loss": 0.0345, "step": 35471 }, { "epoch": 10.91, "learning_rate": 9.014033273943898e-06, "loss": 0.0935, "step": 35472 }, { "epoch": 10.91, "learning_rate": 9.013537432952e-06, "loss": 0.0876, "step": 35473 }, { "epoch": 10.92, "learning_rate": 9.01304159440922e-06, "loss": 0.0325, "step": 35474 }, { "epoch": 10.92, "learning_rate": 9.012545758316794e-06, "loss": 0.1527, "step": 35475 }, { "epoch": 10.92, "learning_rate": 9.012049924675951e-06, "loss": 0.0899, "step": 35476 }, { "epoch": 10.92, "learning_rate": 9.011554093487924e-06, "loss": 0.0749, "step": 35477 }, { "epoch": 10.92, "learning_rate": 9.011058264753933e-06, "loss": 0.0347, "step": 35478 }, { "epoch": 10.92, "learning_rate": 9.010562438475225e-06, "loss": 0.067, "step": 35479 }, { "epoch": 10.92, "learning_rate": 9.010066614653023e-06, "loss": 0.0684, "step": 35480 }, { "epoch": 10.92, "learning_rate": 9.009570793288558e-06, "loss": 0.044, "step": 35481 }, { "epoch": 10.92, "learning_rate": 9.009074974383062e-06, "loss": 0.0713, "step": 35482 }, { "epoch": 10.92, "learning_rate": 9.008579157937763e-06, "loss": 0.0973, "step": 35483 }, { "epoch": 10.92, "learning_rate": 9.008083343953898e-06, "loss": 0.077, "step": 35484 }, { "epoch": 10.92, "learning_rate": 9.007587532432692e-06, "loss": 0.0732, "step": 35485 }, { "epoch": 10.92, "learning_rate": 9.00709172337538e-06, "loss": 0.063, "step": 35486 }, { "epoch": 10.92, "learning_rate": 9.006595916783196e-06, "loss": 0.036, "step": 35487 }, { "epoch": 10.92, "learning_rate": 9.006100112657357e-06, "loss": 0.0585, "step": 35488 }, { "epoch": 10.92, "learning_rate": 9.005604310999109e-06, "loss": 0.0604, "step": 35489 }, { "epoch": 10.92, "learning_rate": 9.00510851180968e-06, "loss": 0.107, "step": 35490 }, { "epoch": 10.92, "learning_rate": 9.004612715090294e-06, "loss": 0.063, "step": 35491 }, { "epoch": 10.92, "learning_rate": 9.004116920842188e-06, "loss": 0.0629, "step": 35492 }, { "epoch": 10.92, "learning_rate": 9.00362112906659e-06, "loss": 0.0657, "step": 35493 }, { "epoch": 10.92, "learning_rate": 9.003125339764733e-06, "loss": 0.0631, "step": 35494 }, { "epoch": 10.92, "learning_rate": 9.002629552937846e-06, "loss": 0.1534, "step": 35495 }, { "epoch": 10.92, "learning_rate": 9.002133768587163e-06, "loss": 0.0611, "step": 35496 }, { "epoch": 10.92, "learning_rate": 9.001637986713913e-06, "loss": 0.0377, "step": 35497 }, { "epoch": 10.92, "learning_rate": 9.001142207319323e-06, "loss": 0.0662, "step": 35498 }, { "epoch": 10.92, "learning_rate": 9.000646430404631e-06, "loss": 0.0582, "step": 35499 }, { "epoch": 10.92, "learning_rate": 9.000150655971065e-06, "loss": 0.0515, "step": 35500 }, { "epoch": 10.92, "learning_rate": 8.999654884019854e-06, "loss": 0.1107, "step": 35501 }, { "epoch": 10.92, "learning_rate": 8.999159114552232e-06, "loss": 0.0466, "step": 35502 }, { "epoch": 10.92, "learning_rate": 8.998663347569424e-06, "loss": 0.0391, "step": 35503 }, { "epoch": 10.92, "learning_rate": 8.998167583072667e-06, "loss": 0.0582, "step": 35504 }, { "epoch": 10.92, "learning_rate": 8.99767182106319e-06, "loss": 0.1375, "step": 35505 }, { "epoch": 10.92, "learning_rate": 8.997176061542224e-06, "loss": 0.0659, "step": 35506 }, { "epoch": 10.93, "learning_rate": 8.996680304511e-06, "loss": 0.0502, "step": 35507 }, { "epoch": 10.93, "learning_rate": 8.996184549970746e-06, "loss": 0.0422, "step": 35508 }, { "epoch": 10.93, "learning_rate": 8.995688797922697e-06, "loss": 0.088, "step": 35509 }, { "epoch": 10.93, "learning_rate": 8.99519304836808e-06, "loss": 0.0861, "step": 35510 }, { "epoch": 10.93, "learning_rate": 8.994697301308133e-06, "loss": 0.0569, "step": 35511 }, { "epoch": 10.93, "learning_rate": 8.994201556744079e-06, "loss": 0.0926, "step": 35512 }, { "epoch": 10.93, "learning_rate": 8.993705814677146e-06, "loss": 0.0556, "step": 35513 }, { "epoch": 10.93, "learning_rate": 8.993210075108578e-06, "loss": 0.0185, "step": 35514 }, { "epoch": 10.93, "learning_rate": 8.992714338039595e-06, "loss": 0.0407, "step": 35515 }, { "epoch": 10.93, "learning_rate": 8.992218603471428e-06, "loss": 0.0674, "step": 35516 }, { "epoch": 10.93, "learning_rate": 8.991722871405313e-06, "loss": 0.0602, "step": 35517 }, { "epoch": 10.93, "learning_rate": 8.991227141842477e-06, "loss": 0.074, "step": 35518 }, { "epoch": 10.93, "learning_rate": 8.990731414784153e-06, "loss": 0.0626, "step": 35519 }, { "epoch": 10.93, "learning_rate": 8.990235690231572e-06, "loss": 0.0529, "step": 35520 }, { "epoch": 10.93, "learning_rate": 8.989739968185965e-06, "loss": 0.0765, "step": 35521 }, { "epoch": 10.93, "learning_rate": 8.989244248648559e-06, "loss": 0.064, "step": 35522 }, { "epoch": 10.93, "learning_rate": 8.988748531620582e-06, "loss": 0.0468, "step": 35523 }, { "epoch": 10.93, "learning_rate": 8.988252817103277e-06, "loss": 0.1112, "step": 35524 }, { "epoch": 10.93, "learning_rate": 8.987757105097866e-06, "loss": 0.0751, "step": 35525 }, { "epoch": 10.93, "learning_rate": 8.98726139560558e-06, "loss": 0.0552, "step": 35526 }, { "epoch": 10.93, "learning_rate": 8.986765688627652e-06, "loss": 0.0988, "step": 35527 }, { "epoch": 10.93, "learning_rate": 8.986269984165309e-06, "loss": 0.0147, "step": 35528 }, { "epoch": 10.93, "learning_rate": 8.985774282219786e-06, "loss": 0.0572, "step": 35529 }, { "epoch": 10.93, "learning_rate": 8.985278582792314e-06, "loss": 0.0808, "step": 35530 }, { "epoch": 10.93, "learning_rate": 8.984782885884119e-06, "loss": 0.0695, "step": 35531 }, { "epoch": 10.93, "learning_rate": 8.984287191496439e-06, "loss": 0.0747, "step": 35532 }, { "epoch": 10.93, "learning_rate": 8.98379149963049e-06, "loss": 0.0789, "step": 35533 }, { "epoch": 10.93, "learning_rate": 8.98329581028752e-06, "loss": 0.1035, "step": 35534 }, { "epoch": 10.93, "learning_rate": 8.982800123468755e-06, "loss": 0.02, "step": 35535 }, { "epoch": 10.93, "learning_rate": 8.98230443917542e-06, "loss": 0.0707, "step": 35536 }, { "epoch": 10.93, "learning_rate": 8.981808757408748e-06, "loss": 0.0647, "step": 35537 }, { "epoch": 10.93, "learning_rate": 8.981313078169969e-06, "loss": 0.0307, "step": 35538 }, { "epoch": 10.94, "learning_rate": 8.980817401460317e-06, "loss": 0.0686, "step": 35539 }, { "epoch": 10.94, "learning_rate": 8.98032172728102e-06, "loss": 0.0579, "step": 35540 }, { "epoch": 10.94, "learning_rate": 8.97982605563331e-06, "loss": 0.0262, "step": 35541 }, { "epoch": 10.94, "learning_rate": 8.97933038651842e-06, "loss": 0.0722, "step": 35542 }, { "epoch": 10.94, "learning_rate": 8.978834719937569e-06, "loss": 0.1113, "step": 35543 }, { "epoch": 10.94, "learning_rate": 8.978339055892001e-06, "loss": 0.0743, "step": 35544 }, { "epoch": 10.94, "learning_rate": 8.977843394382945e-06, "loss": 0.1042, "step": 35545 }, { "epoch": 10.94, "learning_rate": 8.977347735411625e-06, "loss": 0.0758, "step": 35546 }, { "epoch": 10.94, "learning_rate": 8.976852078979274e-06, "loss": 0.0332, "step": 35547 }, { "epoch": 10.94, "learning_rate": 8.976356425087124e-06, "loss": 0.0509, "step": 35548 }, { "epoch": 10.94, "learning_rate": 8.975860773736405e-06, "loss": 0.0592, "step": 35549 }, { "epoch": 10.94, "learning_rate": 8.975365124928349e-06, "loss": 0.0504, "step": 35550 }, { "epoch": 10.94, "learning_rate": 8.974869478664183e-06, "loss": 0.0649, "step": 35551 }, { "epoch": 10.94, "learning_rate": 8.974373834945141e-06, "loss": 0.0272, "step": 35552 }, { "epoch": 10.94, "learning_rate": 8.973878193772451e-06, "loss": 0.0761, "step": 35553 }, { "epoch": 10.94, "learning_rate": 8.973382555147347e-06, "loss": 0.0648, "step": 35554 }, { "epoch": 10.94, "learning_rate": 8.972886919071055e-06, "loss": 0.052, "step": 35555 }, { "epoch": 10.94, "learning_rate": 8.972391285544813e-06, "loss": 0.1041, "step": 35556 }, { "epoch": 10.94, "learning_rate": 8.971895654569842e-06, "loss": 0.0594, "step": 35557 }, { "epoch": 10.94, "learning_rate": 8.971400026147375e-06, "loss": 0.0622, "step": 35558 }, { "epoch": 10.94, "learning_rate": 8.970904400278648e-06, "loss": 0.0556, "step": 35559 }, { "epoch": 10.94, "learning_rate": 8.970408776964888e-06, "loss": 0.067, "step": 35560 }, { "epoch": 10.94, "learning_rate": 8.969913156207323e-06, "loss": 0.0318, "step": 35561 }, { "epoch": 10.94, "learning_rate": 8.969417538007189e-06, "loss": 0.0629, "step": 35562 }, { "epoch": 10.94, "learning_rate": 8.968921922365709e-06, "loss": 0.0483, "step": 35563 }, { "epoch": 10.94, "learning_rate": 8.968426309284121e-06, "loss": 0.0819, "step": 35564 }, { "epoch": 10.94, "learning_rate": 8.967930698763651e-06, "loss": 0.1012, "step": 35565 }, { "epoch": 10.94, "learning_rate": 8.967435090805536e-06, "loss": 0.0848, "step": 35566 }, { "epoch": 10.94, "learning_rate": 8.966939485410996e-06, "loss": 0.0823, "step": 35567 }, { "epoch": 10.94, "learning_rate": 8.966443882581264e-06, "loss": 0.0459, "step": 35568 }, { "epoch": 10.94, "learning_rate": 8.965948282317579e-06, "loss": 0.0585, "step": 35569 }, { "epoch": 10.94, "learning_rate": 8.965452684621164e-06, "loss": 0.0478, "step": 35570 }, { "epoch": 10.94, "learning_rate": 8.964957089493251e-06, "loss": 0.0816, "step": 35571 }, { "epoch": 10.95, "learning_rate": 8.96446149693507e-06, "loss": 0.0563, "step": 35572 }, { "epoch": 10.95, "learning_rate": 8.963965906947852e-06, "loss": 0.0535, "step": 35573 }, { "epoch": 10.95, "learning_rate": 8.963470319532826e-06, "loss": 0.0893, "step": 35574 }, { "epoch": 10.95, "learning_rate": 8.962974734691226e-06, "loss": 0.0588, "step": 35575 }, { "epoch": 10.95, "learning_rate": 8.96247915242428e-06, "loss": 0.0789, "step": 35576 }, { "epoch": 10.95, "learning_rate": 8.96198357273322e-06, "loss": 0.0357, "step": 35577 }, { "epoch": 10.95, "learning_rate": 8.961487995619269e-06, "loss": 0.0525, "step": 35578 }, { "epoch": 10.95, "learning_rate": 8.960992421083669e-06, "loss": 0.0735, "step": 35579 }, { "epoch": 10.95, "learning_rate": 8.960496849127645e-06, "loss": 0.047, "step": 35580 }, { "epoch": 10.95, "learning_rate": 8.960001279752424e-06, "loss": 0.1003, "step": 35581 }, { "epoch": 10.95, "learning_rate": 8.959505712959241e-06, "loss": 0.0635, "step": 35582 }, { "epoch": 10.95, "learning_rate": 8.959010148749324e-06, "loss": 0.0583, "step": 35583 }, { "epoch": 10.95, "learning_rate": 8.958514587123903e-06, "loss": 0.0856, "step": 35584 }, { "epoch": 10.95, "learning_rate": 8.958019028084212e-06, "loss": 0.0821, "step": 35585 }, { "epoch": 10.95, "learning_rate": 8.957523471631477e-06, "loss": 0.0696, "step": 35586 }, { "epoch": 10.95, "learning_rate": 8.957027917766934e-06, "loss": 0.0359, "step": 35587 }, { "epoch": 10.95, "learning_rate": 8.956532366491802e-06, "loss": 0.0492, "step": 35588 }, { "epoch": 10.95, "learning_rate": 8.956036817807324e-06, "loss": 0.0256, "step": 35589 }, { "epoch": 10.95, "learning_rate": 8.955541271714727e-06, "loss": 0.1013, "step": 35590 }, { "epoch": 10.95, "learning_rate": 8.955045728215238e-06, "loss": 0.0784, "step": 35591 }, { "epoch": 10.95, "learning_rate": 8.954550187310087e-06, "loss": 0.0605, "step": 35592 }, { "epoch": 10.95, "learning_rate": 8.954054649000505e-06, "loss": 0.0614, "step": 35593 }, { "epoch": 10.95, "learning_rate": 8.953559113287726e-06, "loss": 0.0919, "step": 35594 }, { "epoch": 10.95, "learning_rate": 8.953063580172975e-06, "loss": 0.0336, "step": 35595 }, { "epoch": 10.95, "learning_rate": 8.952568049657488e-06, "loss": 0.0769, "step": 35596 }, { "epoch": 10.95, "learning_rate": 8.95207252174249e-06, "loss": 0.0828, "step": 35597 }, { "epoch": 10.95, "learning_rate": 8.951576996429214e-06, "loss": 0.1372, "step": 35598 }, { "epoch": 10.95, "learning_rate": 8.95108147371889e-06, "loss": 0.058, "step": 35599 }, { "epoch": 10.95, "learning_rate": 8.95058595361275e-06, "loss": 0.035, "step": 35600 }, { "epoch": 10.95, "learning_rate": 8.95009043611202e-06, "loss": 0.0531, "step": 35601 }, { "epoch": 10.95, "learning_rate": 8.949594921217933e-06, "loss": 0.0539, "step": 35602 }, { "epoch": 10.95, "learning_rate": 8.949099408931716e-06, "loss": 0.0511, "step": 35603 }, { "epoch": 10.96, "learning_rate": 8.948603899254603e-06, "loss": 0.0633, "step": 35604 }, { "epoch": 10.96, "learning_rate": 8.948108392187824e-06, "loss": 0.1, "step": 35605 }, { "epoch": 10.96, "learning_rate": 8.94761288773261e-06, "loss": 0.0595, "step": 35606 }, { "epoch": 10.96, "learning_rate": 8.947117385890187e-06, "loss": 0.0554, "step": 35607 }, { "epoch": 10.96, "learning_rate": 8.946621886661786e-06, "loss": 0.0371, "step": 35608 }, { "epoch": 10.96, "learning_rate": 8.94612639004864e-06, "loss": 0.0785, "step": 35609 }, { "epoch": 10.96, "learning_rate": 8.94563089605198e-06, "loss": 0.0569, "step": 35610 }, { "epoch": 10.96, "learning_rate": 8.945135404673037e-06, "loss": 0.0524, "step": 35611 }, { "epoch": 10.96, "learning_rate": 8.944639915913034e-06, "loss": 0.1049, "step": 35612 }, { "epoch": 10.96, "learning_rate": 8.944144429773203e-06, "loss": 0.0769, "step": 35613 }, { "epoch": 10.96, "learning_rate": 8.943648946254781e-06, "loss": 0.0601, "step": 35614 }, { "epoch": 10.96, "learning_rate": 8.943153465358994e-06, "loss": 0.0553, "step": 35615 }, { "epoch": 10.96, "learning_rate": 8.942657987087071e-06, "loss": 0.0691, "step": 35616 }, { "epoch": 10.96, "learning_rate": 8.942162511440245e-06, "loss": 0.0647, "step": 35617 }, { "epoch": 10.96, "learning_rate": 8.94166703841974e-06, "loss": 0.051, "step": 35618 }, { "epoch": 10.96, "learning_rate": 8.941171568026794e-06, "loss": 0.0649, "step": 35619 }, { "epoch": 10.96, "learning_rate": 8.940676100262633e-06, "loss": 0.0419, "step": 35620 }, { "epoch": 10.96, "learning_rate": 8.940180635128489e-06, "loss": 0.0742, "step": 35621 }, { "epoch": 10.96, "learning_rate": 8.939685172625588e-06, "loss": 0.0553, "step": 35622 }, { "epoch": 10.96, "learning_rate": 8.93918971275516e-06, "loss": 0.0545, "step": 35623 }, { "epoch": 10.96, "learning_rate": 8.938694255518444e-06, "loss": 0.0458, "step": 35624 }, { "epoch": 10.96, "learning_rate": 8.938198800916662e-06, "loss": 0.1036, "step": 35625 }, { "epoch": 10.96, "learning_rate": 8.937703348951046e-06, "loss": 0.0783, "step": 35626 }, { "epoch": 10.96, "learning_rate": 8.937207899622826e-06, "loss": 0.074, "step": 35627 }, { "epoch": 10.96, "learning_rate": 8.936712452933231e-06, "loss": 0.147, "step": 35628 }, { "epoch": 10.96, "learning_rate": 8.936217008883494e-06, "loss": 0.0375, "step": 35629 }, { "epoch": 10.96, "learning_rate": 8.935721567474843e-06, "loss": 0.0703, "step": 35630 }, { "epoch": 10.96, "learning_rate": 8.93522612870851e-06, "loss": 0.088, "step": 35631 }, { "epoch": 10.96, "learning_rate": 8.934730692585723e-06, "loss": 0.0709, "step": 35632 }, { "epoch": 10.96, "learning_rate": 8.934235259107708e-06, "loss": 0.0563, "step": 35633 }, { "epoch": 10.96, "learning_rate": 8.933739828275703e-06, "loss": 0.0942, "step": 35634 }, { "epoch": 10.96, "learning_rate": 8.933244400090937e-06, "loss": 0.0874, "step": 35635 }, { "epoch": 10.96, "learning_rate": 8.932748974554636e-06, "loss": 0.0934, "step": 35636 }, { "epoch": 10.97, "learning_rate": 8.932253551668031e-06, "loss": 0.0546, "step": 35637 }, { "epoch": 10.97, "learning_rate": 8.93175813143235e-06, "loss": 0.0515, "step": 35638 }, { "epoch": 10.97, "learning_rate": 8.931262713848829e-06, "loss": 0.054, "step": 35639 }, { "epoch": 10.97, "learning_rate": 8.930767298918692e-06, "loss": 0.0459, "step": 35640 }, { "epoch": 10.97, "learning_rate": 8.930271886643173e-06, "loss": 0.0557, "step": 35641 }, { "epoch": 10.97, "learning_rate": 8.929776477023504e-06, "loss": 0.0558, "step": 35642 }, { "epoch": 10.97, "learning_rate": 8.929281070060903e-06, "loss": 0.0598, "step": 35643 }, { "epoch": 10.97, "learning_rate": 8.928785665756615e-06, "loss": 0.0466, "step": 35644 }, { "epoch": 10.97, "learning_rate": 8.928290264111863e-06, "loss": 0.0341, "step": 35645 }, { "epoch": 10.97, "learning_rate": 8.927794865127876e-06, "loss": 0.0878, "step": 35646 }, { "epoch": 10.97, "learning_rate": 8.927299468805886e-06, "loss": 0.0549, "step": 35647 }, { "epoch": 10.97, "learning_rate": 8.92680407514712e-06, "loss": 0.0938, "step": 35648 }, { "epoch": 10.97, "learning_rate": 8.926308684152813e-06, "loss": 0.0689, "step": 35649 }, { "epoch": 10.97, "learning_rate": 8.92581329582419e-06, "loss": 0.0924, "step": 35650 }, { "epoch": 10.97, "learning_rate": 8.925317910162485e-06, "loss": 0.084, "step": 35651 }, { "epoch": 10.97, "learning_rate": 8.924822527168925e-06, "loss": 0.0668, "step": 35652 }, { "epoch": 10.97, "learning_rate": 8.924327146844738e-06, "loss": 0.0489, "step": 35653 }, { "epoch": 10.97, "learning_rate": 8.92383176919116e-06, "loss": 0.163, "step": 35654 }, { "epoch": 10.97, "learning_rate": 8.923336394209416e-06, "loss": 0.0799, "step": 35655 }, { "epoch": 10.97, "learning_rate": 8.92284102190074e-06, "loss": 0.0389, "step": 35656 }, { "epoch": 10.97, "learning_rate": 8.92234565226636e-06, "loss": 0.0236, "step": 35657 }, { "epoch": 10.97, "learning_rate": 8.9218502853075e-06, "loss": 0.0257, "step": 35658 }, { "epoch": 10.97, "learning_rate": 8.921354921025398e-06, "loss": 0.1257, "step": 35659 }, { "epoch": 10.97, "learning_rate": 8.920859559421283e-06, "loss": 0.0368, "step": 35660 }, { "epoch": 10.97, "learning_rate": 8.92036420049638e-06, "loss": 0.1036, "step": 35661 }, { "epoch": 10.97, "learning_rate": 8.91986884425192e-06, "loss": 0.0776, "step": 35662 }, { "epoch": 10.97, "learning_rate": 8.919373490689137e-06, "loss": 0.0758, "step": 35663 }, { "epoch": 10.97, "learning_rate": 8.91887813980926e-06, "loss": 0.1015, "step": 35664 }, { "epoch": 10.97, "learning_rate": 8.918382791613514e-06, "loss": 0.0826, "step": 35665 }, { "epoch": 10.97, "learning_rate": 8.917887446103137e-06, "loss": 0.0659, "step": 35666 }, { "epoch": 10.97, "learning_rate": 8.917392103279346e-06, "loss": 0.0513, "step": 35667 }, { "epoch": 10.97, "learning_rate": 8.916896763143382e-06, "loss": 0.1154, "step": 35668 }, { "epoch": 10.98, "learning_rate": 8.916401425696473e-06, "loss": 0.1017, "step": 35669 }, { "epoch": 10.98, "learning_rate": 8.915906090939848e-06, "loss": 0.0565, "step": 35670 }, { "epoch": 10.98, "learning_rate": 8.915410758874732e-06, "loss": 0.0982, "step": 35671 }, { "epoch": 10.98, "learning_rate": 8.91491542950236e-06, "loss": 0.044, "step": 35672 }, { "epoch": 10.98, "learning_rate": 8.91442010282396e-06, "loss": 0.0677, "step": 35673 }, { "epoch": 10.98, "learning_rate": 8.913924778840764e-06, "loss": 0.1323, "step": 35674 }, { "epoch": 10.98, "learning_rate": 8.913429457553999e-06, "loss": 0.0421, "step": 35675 }, { "epoch": 10.98, "learning_rate": 8.912934138964897e-06, "loss": 0.0603, "step": 35676 }, { "epoch": 10.98, "learning_rate": 8.912438823074683e-06, "loss": 0.0588, "step": 35677 }, { "epoch": 10.98, "learning_rate": 8.911943509884592e-06, "loss": 0.0705, "step": 35678 }, { "epoch": 10.98, "learning_rate": 8.911448199395856e-06, "loss": 0.0596, "step": 35679 }, { "epoch": 10.98, "learning_rate": 8.910952891609698e-06, "loss": 0.1228, "step": 35680 }, { "epoch": 10.98, "learning_rate": 8.91045758652735e-06, "loss": 0.0955, "step": 35681 }, { "epoch": 10.98, "learning_rate": 8.909962284150039e-06, "loss": 0.0631, "step": 35682 }, { "epoch": 10.98, "learning_rate": 8.909466984479e-06, "loss": 0.0745, "step": 35683 }, { "epoch": 10.98, "learning_rate": 8.908971687515462e-06, "loss": 0.0414, "step": 35684 }, { "epoch": 10.98, "learning_rate": 8.908476393260652e-06, "loss": 0.0838, "step": 35685 }, { "epoch": 10.98, "learning_rate": 8.907981101715802e-06, "loss": 0.0656, "step": 35686 }, { "epoch": 10.98, "learning_rate": 8.907485812882137e-06, "loss": 0.038, "step": 35687 }, { "epoch": 10.98, "learning_rate": 8.906990526760893e-06, "loss": 0.0402, "step": 35688 }, { "epoch": 10.98, "learning_rate": 8.906495243353297e-06, "loss": 0.051, "step": 35689 }, { "epoch": 10.98, "learning_rate": 8.90599996266058e-06, "loss": 0.0409, "step": 35690 }, { "epoch": 10.98, "learning_rate": 8.905504684683969e-06, "loss": 0.0633, "step": 35691 }, { "epoch": 10.98, "learning_rate": 8.90500940942469e-06, "loss": 0.0365, "step": 35692 }, { "epoch": 10.98, "learning_rate": 8.904514136883984e-06, "loss": 0.0297, "step": 35693 }, { "epoch": 10.98, "learning_rate": 8.904018867063071e-06, "loss": 0.0624, "step": 35694 }, { "epoch": 10.98, "learning_rate": 8.903523599963184e-06, "loss": 0.0318, "step": 35695 }, { "epoch": 10.98, "learning_rate": 8.903028335585552e-06, "loss": 0.0907, "step": 35696 }, { "epoch": 10.98, "learning_rate": 8.902533073931404e-06, "loss": 0.049, "step": 35697 }, { "epoch": 10.98, "learning_rate": 8.902037815001973e-06, "loss": 0.0591, "step": 35698 }, { "epoch": 10.98, "learning_rate": 8.901542558798485e-06, "loss": 0.0715, "step": 35699 }, { "epoch": 10.98, "learning_rate": 8.901047305322172e-06, "loss": 0.0706, "step": 35700 }, { "epoch": 10.98, "learning_rate": 8.900552054574261e-06, "loss": 0.0599, "step": 35701 }, { "epoch": 10.99, "learning_rate": 8.90005680655598e-06, "loss": 0.048, "step": 35702 }, { "epoch": 10.99, "learning_rate": 8.899561561268566e-06, "loss": 0.0589, "step": 35703 }, { "epoch": 10.99, "learning_rate": 8.899066318713242e-06, "loss": 0.0528, "step": 35704 }, { "epoch": 10.99, "learning_rate": 8.89857107889124e-06, "loss": 0.0523, "step": 35705 }, { "epoch": 10.99, "learning_rate": 8.89807584180379e-06, "loss": 0.0654, "step": 35706 }, { "epoch": 10.99, "learning_rate": 8.897580607452117e-06, "loss": 0.0565, "step": 35707 }, { "epoch": 10.99, "learning_rate": 8.897085375837457e-06, "loss": 0.0824, "step": 35708 }, { "epoch": 10.99, "learning_rate": 8.896590146961038e-06, "loss": 0.0682, "step": 35709 }, { "epoch": 10.99, "learning_rate": 8.896094920824086e-06, "loss": 0.1156, "step": 35710 }, { "epoch": 10.99, "learning_rate": 8.895599697427834e-06, "loss": 0.0696, "step": 35711 }, { "epoch": 10.99, "learning_rate": 8.895104476773506e-06, "loss": 0.0307, "step": 35712 }, { "epoch": 10.99, "learning_rate": 8.89460925886234e-06, "loss": 0.0841, "step": 35713 }, { "epoch": 10.99, "learning_rate": 8.894114043695562e-06, "loss": 0.0405, "step": 35714 }, { "epoch": 10.99, "learning_rate": 8.893618831274399e-06, "loss": 0.077, "step": 35715 }, { "epoch": 10.99, "learning_rate": 8.893123621600083e-06, "loss": 0.0586, "step": 35716 }, { "epoch": 10.99, "learning_rate": 8.89262841467384e-06, "loss": 0.0553, "step": 35717 }, { "epoch": 10.99, "learning_rate": 8.892133210496904e-06, "loss": 0.0568, "step": 35718 }, { "epoch": 10.99, "learning_rate": 8.891638009070502e-06, "loss": 0.0503, "step": 35719 }, { "epoch": 10.99, "learning_rate": 8.891142810395865e-06, "loss": 0.0851, "step": 35720 }, { "epoch": 10.99, "learning_rate": 8.890647614474223e-06, "loss": 0.0904, "step": 35721 }, { "epoch": 10.99, "learning_rate": 8.890152421306799e-06, "loss": 0.0654, "step": 35722 }, { "epoch": 10.99, "learning_rate": 8.88965723089483e-06, "loss": 0.0999, "step": 35723 }, { "epoch": 10.99, "learning_rate": 8.889162043239543e-06, "loss": 0.0798, "step": 35724 }, { "epoch": 10.99, "learning_rate": 8.888666858342169e-06, "loss": 0.0611, "step": 35725 }, { "epoch": 10.99, "learning_rate": 8.888171676203933e-06, "loss": 0.0836, "step": 35726 }, { "epoch": 10.99, "learning_rate": 8.887676496826066e-06, "loss": 0.0269, "step": 35727 }, { "epoch": 10.99, "learning_rate": 8.887181320209801e-06, "loss": 0.0377, "step": 35728 }, { "epoch": 10.99, "learning_rate": 8.886686146356364e-06, "loss": 0.0487, "step": 35729 }, { "epoch": 10.99, "learning_rate": 8.886190975266985e-06, "loss": 0.0741, "step": 35730 }, { "epoch": 10.99, "learning_rate": 8.885695806942893e-06, "loss": 0.0591, "step": 35731 }, { "epoch": 10.99, "learning_rate": 8.885200641385317e-06, "loss": 0.0461, "step": 35732 }, { "epoch": 10.99, "learning_rate": 8.884705478595489e-06, "loss": 0.077, "step": 35733 }, { "epoch": 11.0, "learning_rate": 8.884210318574637e-06, "loss": 0.0401, "step": 35734 }, { "epoch": 11.0, "learning_rate": 8.883715161323991e-06, "loss": 0.0581, "step": 35735 }, { "epoch": 11.0, "learning_rate": 8.883220006844778e-06, "loss": 0.0795, "step": 35736 }, { "epoch": 11.0, "learning_rate": 8.882724855138227e-06, "loss": 0.0422, "step": 35737 }, { "epoch": 11.0, "learning_rate": 8.88222970620557e-06, "loss": 0.0622, "step": 35738 }, { "epoch": 11.0, "learning_rate": 8.881734560048037e-06, "loss": 0.0635, "step": 35739 }, { "epoch": 11.0, "learning_rate": 8.881239416666854e-06, "loss": 0.0822, "step": 35740 }, { "epoch": 11.0, "learning_rate": 8.880744276063252e-06, "loss": 0.0694, "step": 35741 }, { "epoch": 11.0, "learning_rate": 8.880249138238458e-06, "loss": 0.0737, "step": 35742 }, { "epoch": 11.0, "learning_rate": 8.879754003193706e-06, "loss": 0.0873, "step": 35743 }, { "epoch": 11.0, "learning_rate": 8.879258870930223e-06, "loss": 0.0904, "step": 35744 }, { "epoch": 11.0, "learning_rate": 8.87876374144924e-06, "loss": 0.1076, "step": 35745 }, { "epoch": 11.0, "learning_rate": 8.878268614751983e-06, "loss": 0.0706, "step": 35746 }, { "epoch": 11.0, "learning_rate": 8.877773490839677e-06, "loss": 0.07, "step": 35747 }, { "epoch": 11.0, "learning_rate": 8.877278369713563e-06, "loss": 0.0532, "step": 35748 }, { "epoch": 11.0, "learning_rate": 8.876783251374864e-06, "loss": 0.0831, "step": 35749 }, { "epoch": 11.0, "learning_rate": 8.876288135824808e-06, "loss": 0.0627, "step": 35750 }, { "epoch": 11.0, "learning_rate": 8.875793023064626e-06, "loss": 0.0566, "step": 35751 }, { "epoch": 11.0, "learning_rate": 8.875297913095544e-06, "loss": 0.0315, "step": 35752 }, { "epoch": 11.0, "learning_rate": 8.874802805918797e-06, "loss": 0.0425, "step": 35753 }, { "epoch": 11.0, "learning_rate": 8.874307701535611e-06, "loss": 0.041, "step": 35754 }, { "epoch": 11.0, "learning_rate": 8.873812599947215e-06, "loss": 0.0253, "step": 35755 }, { "epoch": 11.0, "learning_rate": 8.873317501154843e-06, "loss": 0.0732, "step": 35756 }, { "epoch": 11.0, "learning_rate": 8.87282240515971e-06, "loss": 0.061, "step": 35757 }, { "epoch": 11.0, "learning_rate": 8.872327311963065e-06, "loss": 0.0567, "step": 35758 }, { "epoch": 11.0, "learning_rate": 8.871832221566124e-06, "loss": 0.043, "step": 35759 }, { "epoch": 11.0, "learning_rate": 8.871337133970118e-06, "loss": 0.0441, "step": 35760 }, { "epoch": 11.0, "learning_rate": 8.870842049176278e-06, "loss": 0.0352, "step": 35761 }, { "epoch": 11.0, "learning_rate": 8.870346967185832e-06, "loss": 0.027, "step": 35762 }, { "epoch": 11.0, "learning_rate": 8.869851888000011e-06, "loss": 0.0253, "step": 35763 }, { "epoch": 11.0, "learning_rate": 8.869356811620043e-06, "loss": 0.0348, "step": 35764 }, { "epoch": 11.0, "learning_rate": 8.868861738047158e-06, "loss": 0.0481, "step": 35765 }, { "epoch": 11.0, "learning_rate": 8.868366667282586e-06, "loss": 0.0427, "step": 35766 }, { "epoch": 11.01, "learning_rate": 8.86787159932755e-06, "loss": 0.0471, "step": 35767 }, { "epoch": 11.01, "learning_rate": 8.867376534183285e-06, "loss": 0.0476, "step": 35768 }, { "epoch": 11.01, "learning_rate": 8.866881471851022e-06, "loss": 0.034, "step": 35769 }, { "epoch": 11.01, "learning_rate": 8.866386412331985e-06, "loss": 0.0367, "step": 35770 }, { "epoch": 11.01, "learning_rate": 8.865891355627404e-06, "loss": 0.0464, "step": 35771 }, { "epoch": 11.01, "learning_rate": 8.865396301738507e-06, "loss": 0.0635, "step": 35772 }, { "epoch": 11.01, "learning_rate": 8.864901250666527e-06, "loss": 0.0588, "step": 35773 }, { "epoch": 11.01, "learning_rate": 8.864406202412693e-06, "loss": 0.0551, "step": 35774 }, { "epoch": 11.01, "learning_rate": 8.863911156978231e-06, "loss": 0.0369, "step": 35775 }, { "epoch": 11.01, "learning_rate": 8.863416114364372e-06, "loss": 0.0531, "step": 35776 }, { "epoch": 11.01, "learning_rate": 8.862921074572342e-06, "loss": 0.0392, "step": 35777 }, { "epoch": 11.01, "learning_rate": 8.862426037603373e-06, "loss": 0.0357, "step": 35778 }, { "epoch": 11.01, "learning_rate": 8.861931003458697e-06, "loss": 0.0749, "step": 35779 }, { "epoch": 11.01, "learning_rate": 8.861435972139538e-06, "loss": 0.0515, "step": 35780 }, { "epoch": 11.01, "learning_rate": 8.860940943647126e-06, "loss": 0.0536, "step": 35781 }, { "epoch": 11.01, "learning_rate": 8.860445917982687e-06, "loss": 0.0388, "step": 35782 }, { "epoch": 11.01, "learning_rate": 8.859950895147458e-06, "loss": 0.0534, "step": 35783 }, { "epoch": 11.01, "learning_rate": 8.859455875142661e-06, "loss": 0.0764, "step": 35784 }, { "epoch": 11.01, "learning_rate": 8.85896085796953e-06, "loss": 0.0522, "step": 35785 }, { "epoch": 11.01, "learning_rate": 8.85846584362929e-06, "loss": 0.0648, "step": 35786 }, { "epoch": 11.01, "learning_rate": 8.85797083212317e-06, "loss": 0.042, "step": 35787 }, { "epoch": 11.01, "learning_rate": 8.857475823452403e-06, "loss": 0.0411, "step": 35788 }, { "epoch": 11.01, "learning_rate": 8.856980817618214e-06, "loss": 0.0383, "step": 35789 }, { "epoch": 11.01, "learning_rate": 8.856485814621838e-06, "loss": 0.0525, "step": 35790 }, { "epoch": 11.01, "learning_rate": 8.855990814464497e-06, "loss": 0.044, "step": 35791 }, { "epoch": 11.01, "learning_rate": 8.855495817147417e-06, "loss": 0.0526, "step": 35792 }, { "epoch": 11.01, "learning_rate": 8.855000822671838e-06, "loss": 0.103, "step": 35793 }, { "epoch": 11.01, "learning_rate": 8.854505831038983e-06, "loss": 0.0565, "step": 35794 }, { "epoch": 11.01, "learning_rate": 8.85401084225008e-06, "loss": 0.0593, "step": 35795 }, { "epoch": 11.01, "learning_rate": 8.853515856306359e-06, "loss": 0.0477, "step": 35796 }, { "epoch": 11.01, "learning_rate": 8.853020873209048e-06, "loss": 0.0303, "step": 35797 }, { "epoch": 11.01, "learning_rate": 8.852525892959379e-06, "loss": 0.0545, "step": 35798 }, { "epoch": 11.02, "learning_rate": 8.852030915558578e-06, "loss": 0.0612, "step": 35799 }, { "epoch": 11.02, "learning_rate": 8.851535941007877e-06, "loss": 0.072, "step": 35800 }, { "epoch": 11.02, "learning_rate": 8.8510409693085e-06, "loss": 0.0506, "step": 35801 }, { "epoch": 11.02, "learning_rate": 8.850546000461676e-06, "loss": 0.0662, "step": 35802 }, { "epoch": 11.02, "learning_rate": 8.850051034468642e-06, "loss": 0.0449, "step": 35803 }, { "epoch": 11.02, "learning_rate": 8.849556071330618e-06, "loss": 0.0473, "step": 35804 }, { "epoch": 11.02, "learning_rate": 8.849061111048838e-06, "loss": 0.059, "step": 35805 }, { "epoch": 11.02, "learning_rate": 8.848566153624527e-06, "loss": 0.0671, "step": 35806 }, { "epoch": 11.02, "learning_rate": 8.848071199058916e-06, "loss": 0.042, "step": 35807 }, { "epoch": 11.02, "learning_rate": 8.847576247353234e-06, "loss": 0.0582, "step": 35808 }, { "epoch": 11.02, "learning_rate": 8.84708129850871e-06, "loss": 0.0333, "step": 35809 }, { "epoch": 11.02, "learning_rate": 8.846586352526572e-06, "loss": 0.0394, "step": 35810 }, { "epoch": 11.02, "learning_rate": 8.846091409408053e-06, "loss": 0.0527, "step": 35811 }, { "epoch": 11.02, "learning_rate": 8.84559646915437e-06, "loss": 0.0586, "step": 35812 }, { "epoch": 11.02, "learning_rate": 8.845101531766763e-06, "loss": 0.057, "step": 35813 }, { "epoch": 11.02, "learning_rate": 8.844606597246461e-06, "loss": 0.0632, "step": 35814 }, { "epoch": 11.02, "learning_rate": 8.844111665594687e-06, "loss": 0.0406, "step": 35815 }, { "epoch": 11.02, "learning_rate": 8.843616736812673e-06, "loss": 0.0436, "step": 35816 }, { "epoch": 11.02, "learning_rate": 8.843121810901643e-06, "loss": 0.0415, "step": 35817 }, { "epoch": 11.02, "learning_rate": 8.842626887862832e-06, "loss": 0.0434, "step": 35818 }, { "epoch": 11.02, "learning_rate": 8.842131967697469e-06, "loss": 0.054, "step": 35819 }, { "epoch": 11.02, "learning_rate": 8.841637050406778e-06, "loss": 0.0584, "step": 35820 }, { "epoch": 11.02, "learning_rate": 8.84114213599199e-06, "loss": 0.0594, "step": 35821 }, { "epoch": 11.02, "learning_rate": 8.84064722445433e-06, "loss": 0.0335, "step": 35822 }, { "epoch": 11.02, "learning_rate": 8.840152315795036e-06, "loss": 0.0538, "step": 35823 }, { "epoch": 11.02, "learning_rate": 8.839657410015331e-06, "loss": 0.0364, "step": 35824 }, { "epoch": 11.02, "learning_rate": 8.839162507116443e-06, "loss": 0.0778, "step": 35825 }, { "epoch": 11.02, "learning_rate": 8.8386676070996e-06, "loss": 0.0586, "step": 35826 }, { "epoch": 11.02, "learning_rate": 8.838172709966031e-06, "loss": 0.0802, "step": 35827 }, { "epoch": 11.02, "learning_rate": 8.837677815716967e-06, "loss": 0.0586, "step": 35828 }, { "epoch": 11.02, "learning_rate": 8.837182924353638e-06, "loss": 0.0353, "step": 35829 }, { "epoch": 11.02, "learning_rate": 8.836688035877268e-06, "loss": 0.0347, "step": 35830 }, { "epoch": 11.02, "learning_rate": 8.836193150289089e-06, "loss": 0.0605, "step": 35831 }, { "epoch": 11.03, "learning_rate": 8.835698267590327e-06, "loss": 0.0309, "step": 35832 }, { "epoch": 11.03, "learning_rate": 8.835203387782213e-06, "loss": 0.0639, "step": 35833 }, { "epoch": 11.03, "learning_rate": 8.834708510865976e-06, "loss": 0.0334, "step": 35834 }, { "epoch": 11.03, "learning_rate": 8.834213636842846e-06, "loss": 0.0606, "step": 35835 }, { "epoch": 11.03, "learning_rate": 8.833718765714046e-06, "loss": 0.033, "step": 35836 }, { "epoch": 11.03, "learning_rate": 8.833223897480807e-06, "loss": 0.0489, "step": 35837 }, { "epoch": 11.03, "learning_rate": 8.83272903214436e-06, "loss": 0.0711, "step": 35838 }, { "epoch": 11.03, "learning_rate": 8.832234169705932e-06, "loss": 0.0409, "step": 35839 }, { "epoch": 11.03, "learning_rate": 8.831739310166752e-06, "loss": 0.027, "step": 35840 }, { "epoch": 11.03, "learning_rate": 8.831244453528048e-06, "loss": 0.0479, "step": 35841 }, { "epoch": 11.03, "learning_rate": 8.830749599791047e-06, "loss": 0.0941, "step": 35842 }, { "epoch": 11.03, "learning_rate": 8.830254748956983e-06, "loss": 0.0593, "step": 35843 }, { "epoch": 11.03, "learning_rate": 8.82975990102708e-06, "loss": 0.0421, "step": 35844 }, { "epoch": 11.03, "learning_rate": 8.829265056002569e-06, "loss": 0.0555, "step": 35845 }, { "epoch": 11.03, "learning_rate": 8.828770213884675e-06, "loss": 0.0609, "step": 35846 }, { "epoch": 11.03, "learning_rate": 8.828275374674626e-06, "loss": 0.0775, "step": 35847 }, { "epoch": 11.03, "learning_rate": 8.82778053837366e-06, "loss": 0.0358, "step": 35848 }, { "epoch": 11.03, "learning_rate": 8.827285704982994e-06, "loss": 0.0342, "step": 35849 }, { "epoch": 11.03, "learning_rate": 8.826790874503865e-06, "loss": 0.0453, "step": 35850 }, { "epoch": 11.03, "learning_rate": 8.826296046937495e-06, "loss": 0.0413, "step": 35851 }, { "epoch": 11.03, "learning_rate": 8.825801222285115e-06, "loss": 0.0552, "step": 35852 }, { "epoch": 11.03, "learning_rate": 8.825306400547955e-06, "loss": 0.1056, "step": 35853 }, { "epoch": 11.03, "learning_rate": 8.824811581727243e-06, "loss": 0.074, "step": 35854 }, { "epoch": 11.03, "learning_rate": 8.824316765824207e-06, "loss": 0.0418, "step": 35855 }, { "epoch": 11.03, "learning_rate": 8.823821952840079e-06, "loss": 0.0494, "step": 35856 }, { "epoch": 11.03, "learning_rate": 8.823327142776076e-06, "loss": 0.0448, "step": 35857 }, { "epoch": 11.03, "learning_rate": 8.822832335633441e-06, "loss": 0.0582, "step": 35858 }, { "epoch": 11.03, "learning_rate": 8.822337531413395e-06, "loss": 0.0389, "step": 35859 }, { "epoch": 11.03, "learning_rate": 8.821842730117166e-06, "loss": 0.0356, "step": 35860 }, { "epoch": 11.03, "learning_rate": 8.821347931745986e-06, "loss": 0.0471, "step": 35861 }, { "epoch": 11.03, "learning_rate": 8.820853136301078e-06, "loss": 0.0221, "step": 35862 }, { "epoch": 11.03, "learning_rate": 8.820358343783675e-06, "loss": 0.0727, "step": 35863 }, { "epoch": 11.04, "learning_rate": 8.819863554195005e-06, "loss": 0.0609, "step": 35864 }, { "epoch": 11.04, "learning_rate": 8.819368767536296e-06, "loss": 0.0358, "step": 35865 }, { "epoch": 11.04, "learning_rate": 8.818873983808777e-06, "loss": 0.0198, "step": 35866 }, { "epoch": 11.04, "learning_rate": 8.81837920301367e-06, "loss": 0.0348, "step": 35867 }, { "epoch": 11.04, "learning_rate": 8.817884425152211e-06, "loss": 0.042, "step": 35868 }, { "epoch": 11.04, "learning_rate": 8.817389650225631e-06, "loss": 0.0134, "step": 35869 }, { "epoch": 11.04, "learning_rate": 8.816894878235151e-06, "loss": 0.0592, "step": 35870 }, { "epoch": 11.04, "learning_rate": 8.816400109182002e-06, "loss": 0.0465, "step": 35871 }, { "epoch": 11.04, "learning_rate": 8.815905343067412e-06, "loss": 0.0748, "step": 35872 }, { "epoch": 11.04, "learning_rate": 8.815410579892609e-06, "loss": 0.0554, "step": 35873 }, { "epoch": 11.04, "learning_rate": 8.814915819658823e-06, "loss": 0.033, "step": 35874 }, { "epoch": 11.04, "learning_rate": 8.814421062367284e-06, "loss": 0.047, "step": 35875 }, { "epoch": 11.04, "learning_rate": 8.813926308019218e-06, "loss": 0.0515, "step": 35876 }, { "epoch": 11.04, "learning_rate": 8.813431556615845e-06, "loss": 0.0617, "step": 35877 }, { "epoch": 11.04, "learning_rate": 8.812936808158407e-06, "loss": 0.0354, "step": 35878 }, { "epoch": 11.04, "learning_rate": 8.812442062648131e-06, "loss": 0.0625, "step": 35879 }, { "epoch": 11.04, "learning_rate": 8.811947320086237e-06, "loss": 0.0382, "step": 35880 }, { "epoch": 11.04, "learning_rate": 8.811452580473958e-06, "loss": 0.1183, "step": 35881 }, { "epoch": 11.04, "learning_rate": 8.81095784381252e-06, "loss": 0.0505, "step": 35882 }, { "epoch": 11.04, "learning_rate": 8.810463110103155e-06, "loss": 0.0528, "step": 35883 }, { "epoch": 11.04, "learning_rate": 8.80996837934709e-06, "loss": 0.0643, "step": 35884 }, { "epoch": 11.04, "learning_rate": 8.809473651545553e-06, "loss": 0.0327, "step": 35885 }, { "epoch": 11.04, "learning_rate": 8.80897892669977e-06, "loss": 0.0222, "step": 35886 }, { "epoch": 11.04, "learning_rate": 8.80848420481097e-06, "loss": 0.0412, "step": 35887 }, { "epoch": 11.04, "learning_rate": 8.807989485880387e-06, "loss": 0.043, "step": 35888 }, { "epoch": 11.04, "learning_rate": 8.807494769909242e-06, "loss": 0.0477, "step": 35889 }, { "epoch": 11.04, "learning_rate": 8.807000056898768e-06, "loss": 0.0468, "step": 35890 }, { "epoch": 11.04, "learning_rate": 8.80650534685019e-06, "loss": 0.0465, "step": 35891 }, { "epoch": 11.04, "learning_rate": 8.806010639764733e-06, "loss": 0.0739, "step": 35892 }, { "epoch": 11.04, "learning_rate": 8.805515935643635e-06, "loss": 0.0683, "step": 35893 }, { "epoch": 11.04, "learning_rate": 8.805021234488119e-06, "loss": 0.0571, "step": 35894 }, { "epoch": 11.04, "learning_rate": 8.804526536299413e-06, "loss": 0.0381, "step": 35895 }, { "epoch": 11.04, "learning_rate": 8.804031841078743e-06, "loss": 0.036, "step": 35896 }, { "epoch": 11.05, "learning_rate": 8.803537148827339e-06, "loss": 0.0394, "step": 35897 }, { "epoch": 11.05, "learning_rate": 8.803042459546431e-06, "loss": 0.0328, "step": 35898 }, { "epoch": 11.05, "learning_rate": 8.802547773237248e-06, "loss": 0.072, "step": 35899 }, { "epoch": 11.05, "learning_rate": 8.802053089901016e-06, "loss": 0.0566, "step": 35900 }, { "epoch": 11.05, "learning_rate": 8.801558409538962e-06, "loss": 0.0657, "step": 35901 }, { "epoch": 11.05, "learning_rate": 8.801063732152311e-06, "loss": 0.0203, "step": 35902 }, { "epoch": 11.05, "learning_rate": 8.800569057742302e-06, "loss": 0.0348, "step": 35903 }, { "epoch": 11.05, "learning_rate": 8.800074386310153e-06, "loss": 0.0272, "step": 35904 }, { "epoch": 11.05, "learning_rate": 8.799579717857098e-06, "loss": 0.0311, "step": 35905 }, { "epoch": 11.05, "learning_rate": 8.799085052384362e-06, "loss": 0.0509, "step": 35906 }, { "epoch": 11.05, "learning_rate": 8.798590389893172e-06, "loss": 0.0356, "step": 35907 }, { "epoch": 11.05, "learning_rate": 8.79809573038476e-06, "loss": 0.0313, "step": 35908 }, { "epoch": 11.05, "learning_rate": 8.797601073860353e-06, "loss": 0.0441, "step": 35909 }, { "epoch": 11.05, "learning_rate": 8.797106420321178e-06, "loss": 0.0247, "step": 35910 }, { "epoch": 11.05, "learning_rate": 8.796611769768467e-06, "loss": 0.0419, "step": 35911 }, { "epoch": 11.05, "learning_rate": 8.796117122203436e-06, "loss": 0.0316, "step": 35912 }, { "epoch": 11.05, "learning_rate": 8.795622477627326e-06, "loss": 0.0401, "step": 35913 }, { "epoch": 11.05, "learning_rate": 8.795127836041364e-06, "loss": 0.0326, "step": 35914 }, { "epoch": 11.05, "learning_rate": 8.79463319744677e-06, "loss": 0.0721, "step": 35915 }, { "epoch": 11.05, "learning_rate": 8.79413856184478e-06, "loss": 0.0263, "step": 35916 }, { "epoch": 11.05, "learning_rate": 8.793643929236617e-06, "loss": 0.0504, "step": 35917 }, { "epoch": 11.05, "learning_rate": 8.793149299623512e-06, "loss": 0.0463, "step": 35918 }, { "epoch": 11.05, "learning_rate": 8.792654673006691e-06, "loss": 0.0326, "step": 35919 }, { "epoch": 11.05, "learning_rate": 8.792160049387383e-06, "loss": 0.032, "step": 35920 }, { "epoch": 11.05, "learning_rate": 8.79166542876682e-06, "loss": 0.0568, "step": 35921 }, { "epoch": 11.05, "learning_rate": 8.791170811146217e-06, "loss": 0.0366, "step": 35922 }, { "epoch": 11.05, "learning_rate": 8.790676196526816e-06, "loss": 0.0943, "step": 35923 }, { "epoch": 11.05, "learning_rate": 8.790181584909844e-06, "loss": 0.0286, "step": 35924 }, { "epoch": 11.05, "learning_rate": 8.789686976296521e-06, "loss": 0.063, "step": 35925 }, { "epoch": 11.05, "learning_rate": 8.78919237068808e-06, "loss": 0.0876, "step": 35926 }, { "epoch": 11.05, "learning_rate": 8.788697768085744e-06, "loss": 0.0183, "step": 35927 }, { "epoch": 11.05, "learning_rate": 8.788203168490748e-06, "loss": 0.0359, "step": 35928 }, { "epoch": 11.06, "learning_rate": 8.787708571904317e-06, "loss": 0.102, "step": 35929 }, { "epoch": 11.06, "learning_rate": 8.78721397832768e-06, "loss": 0.0289, "step": 35930 }, { "epoch": 11.06, "learning_rate": 8.786719387762061e-06, "loss": 0.0413, "step": 35931 }, { "epoch": 11.06, "learning_rate": 8.786224800208691e-06, "loss": 0.0614, "step": 35932 }, { "epoch": 11.06, "learning_rate": 8.785730215668798e-06, "loss": 0.0367, "step": 35933 }, { "epoch": 11.06, "learning_rate": 8.785235634143612e-06, "loss": 0.0411, "step": 35934 }, { "epoch": 11.06, "learning_rate": 8.784741055634359e-06, "loss": 0.0323, "step": 35935 }, { "epoch": 11.06, "learning_rate": 8.784246480142263e-06, "loss": 0.0285, "step": 35936 }, { "epoch": 11.06, "learning_rate": 8.783751907668554e-06, "loss": 0.041, "step": 35937 }, { "epoch": 11.06, "learning_rate": 8.783257338214464e-06, "loss": 0.0238, "step": 35938 }, { "epoch": 11.06, "learning_rate": 8.782762771781217e-06, "loss": 0.0716, "step": 35939 }, { "epoch": 11.06, "learning_rate": 8.782268208370042e-06, "loss": 0.0437, "step": 35940 }, { "epoch": 11.06, "learning_rate": 8.781773647982169e-06, "loss": 0.045, "step": 35941 }, { "epoch": 11.06, "learning_rate": 8.781279090618818e-06, "loss": 0.0245, "step": 35942 }, { "epoch": 11.06, "learning_rate": 8.780784536281226e-06, "loss": 0.0403, "step": 35943 }, { "epoch": 11.06, "learning_rate": 8.780289984970617e-06, "loss": 0.0494, "step": 35944 }, { "epoch": 11.06, "learning_rate": 8.779795436688224e-06, "loss": 0.046, "step": 35945 }, { "epoch": 11.06, "learning_rate": 8.779300891435266e-06, "loss": 0.0622, "step": 35946 }, { "epoch": 11.06, "learning_rate": 8.778806349212968e-06, "loss": 0.0532, "step": 35947 }, { "epoch": 11.06, "learning_rate": 8.778311810022573e-06, "loss": 0.0348, "step": 35948 }, { "epoch": 11.06, "learning_rate": 8.777817273865299e-06, "loss": 0.036, "step": 35949 }, { "epoch": 11.06, "learning_rate": 8.777322740742376e-06, "loss": 0.0608, "step": 35950 }, { "epoch": 11.06, "learning_rate": 8.77682821065503e-06, "loss": 0.0482, "step": 35951 }, { "epoch": 11.06, "learning_rate": 8.776333683604486e-06, "loss": 0.0486, "step": 35952 }, { "epoch": 11.06, "learning_rate": 8.77583915959198e-06, "loss": 0.0391, "step": 35953 }, { "epoch": 11.06, "learning_rate": 8.775344638618733e-06, "loss": 0.0471, "step": 35954 }, { "epoch": 11.06, "learning_rate": 8.774850120685979e-06, "loss": 0.0316, "step": 35955 }, { "epoch": 11.06, "learning_rate": 8.77435560579494e-06, "loss": 0.0572, "step": 35956 }, { "epoch": 11.06, "learning_rate": 8.77386109394684e-06, "loss": 0.0593, "step": 35957 }, { "epoch": 11.06, "learning_rate": 8.77336658514292e-06, "loss": 0.0599, "step": 35958 }, { "epoch": 11.06, "learning_rate": 8.772872079384397e-06, "loss": 0.0211, "step": 35959 }, { "epoch": 11.06, "learning_rate": 8.772377576672502e-06, "loss": 0.0229, "step": 35960 }, { "epoch": 11.06, "learning_rate": 8.771883077008463e-06, "loss": 0.052, "step": 35961 }, { "epoch": 11.07, "learning_rate": 8.771388580393505e-06, "loss": 0.0411, "step": 35962 }, { "epoch": 11.07, "learning_rate": 8.770894086828859e-06, "loss": 0.0314, "step": 35963 }, { "epoch": 11.07, "learning_rate": 8.770399596315752e-06, "loss": 0.0172, "step": 35964 }, { "epoch": 11.07, "learning_rate": 8.769905108855412e-06, "loss": 0.0691, "step": 35965 }, { "epoch": 11.07, "learning_rate": 8.769410624449069e-06, "loss": 0.0793, "step": 35966 }, { "epoch": 11.07, "learning_rate": 8.76891614309794e-06, "loss": 0.016, "step": 35967 }, { "epoch": 11.07, "learning_rate": 8.768421664803265e-06, "loss": 0.0337, "step": 35968 }, { "epoch": 11.07, "learning_rate": 8.767927189566268e-06, "loss": 0.0311, "step": 35969 }, { "epoch": 11.07, "learning_rate": 8.767432717388174e-06, "loss": 0.0374, "step": 35970 }, { "epoch": 11.07, "learning_rate": 8.766938248270213e-06, "loss": 0.0349, "step": 35971 }, { "epoch": 11.07, "learning_rate": 8.76644378221361e-06, "loss": 0.0505, "step": 35972 }, { "epoch": 11.07, "learning_rate": 8.765949319219595e-06, "loss": 0.0232, "step": 35973 }, { "epoch": 11.07, "learning_rate": 8.765454859289397e-06, "loss": 0.0359, "step": 35974 }, { "epoch": 11.07, "learning_rate": 8.76496040242424e-06, "loss": 0.0626, "step": 35975 }, { "epoch": 11.07, "learning_rate": 8.764465948625359e-06, "loss": 0.0562, "step": 35976 }, { "epoch": 11.07, "learning_rate": 8.763971497893966e-06, "loss": 0.0594, "step": 35977 }, { "epoch": 11.07, "learning_rate": 8.763477050231304e-06, "loss": 0.0485, "step": 35978 }, { "epoch": 11.07, "learning_rate": 8.762982605638598e-06, "loss": 0.0766, "step": 35979 }, { "epoch": 11.07, "learning_rate": 8.762488164117069e-06, "loss": 0.0569, "step": 35980 }, { "epoch": 11.07, "learning_rate": 8.761993725667948e-06, "loss": 0.0881, "step": 35981 }, { "epoch": 11.07, "learning_rate": 8.761499290292461e-06, "loss": 0.0331, "step": 35982 }, { "epoch": 11.07, "learning_rate": 8.761004857991841e-06, "loss": 0.0549, "step": 35983 }, { "epoch": 11.07, "learning_rate": 8.760510428767311e-06, "loss": 0.052, "step": 35984 }, { "epoch": 11.07, "learning_rate": 8.7600160026201e-06, "loss": 0.0588, "step": 35985 }, { "epoch": 11.07, "learning_rate": 8.759521579551434e-06, "loss": 0.0198, "step": 35986 }, { "epoch": 11.07, "learning_rate": 8.759027159562541e-06, "loss": 0.0257, "step": 35987 }, { "epoch": 11.07, "learning_rate": 8.75853274265465e-06, "loss": 0.0386, "step": 35988 }, { "epoch": 11.07, "learning_rate": 8.758038328828987e-06, "loss": 0.0228, "step": 35989 }, { "epoch": 11.07, "learning_rate": 8.757543918086783e-06, "loss": 0.0289, "step": 35990 }, { "epoch": 11.07, "learning_rate": 8.757049510429259e-06, "loss": 0.0424, "step": 35991 }, { "epoch": 11.07, "learning_rate": 8.756555105857643e-06, "loss": 0.0434, "step": 35992 }, { "epoch": 11.07, "learning_rate": 8.756060704373173e-06, "loss": 0.0584, "step": 35993 }, { "epoch": 11.08, "learning_rate": 8.755566305977064e-06, "loss": 0.0541, "step": 35994 }, { "epoch": 11.08, "learning_rate": 8.75507191067055e-06, "loss": 0.0531, "step": 35995 }, { "epoch": 11.08, "learning_rate": 8.754577518454857e-06, "loss": 0.0518, "step": 35996 }, { "epoch": 11.08, "learning_rate": 8.754083129331211e-06, "loss": 0.0414, "step": 35997 }, { "epoch": 11.08, "learning_rate": 8.75358874330084e-06, "loss": 0.0649, "step": 35998 }, { "epoch": 11.08, "learning_rate": 8.753094360364973e-06, "loss": 0.0424, "step": 35999 }, { "epoch": 11.08, "learning_rate": 8.75259998052484e-06, "loss": 0.0438, "step": 36000 }, { "epoch": 11.08, "learning_rate": 8.752105603781662e-06, "loss": 0.0611, "step": 36001 }, { "epoch": 11.08, "learning_rate": 8.751611230136666e-06, "loss": 0.0651, "step": 36002 }, { "epoch": 11.08, "learning_rate": 8.751116859591089e-06, "loss": 0.0476, "step": 36003 }, { "epoch": 11.08, "learning_rate": 8.750622492146148e-06, "loss": 0.0457, "step": 36004 }, { "epoch": 11.08, "learning_rate": 8.750128127803077e-06, "loss": 0.1275, "step": 36005 }, { "epoch": 11.08, "learning_rate": 8.7496337665631e-06, "loss": 0.0511, "step": 36006 }, { "epoch": 11.08, "learning_rate": 8.749139408427443e-06, "loss": 0.0239, "step": 36007 }, { "epoch": 11.08, "learning_rate": 8.748645053397337e-06, "loss": 0.0564, "step": 36008 }, { "epoch": 11.08, "learning_rate": 8.748150701474008e-06, "loss": 0.0556, "step": 36009 }, { "epoch": 11.08, "learning_rate": 8.747656352658685e-06, "loss": 0.0406, "step": 36010 }, { "epoch": 11.08, "learning_rate": 8.747162006952595e-06, "loss": 0.0534, "step": 36011 }, { "epoch": 11.08, "learning_rate": 8.746667664356957e-06, "loss": 0.0591, "step": 36012 }, { "epoch": 11.08, "learning_rate": 8.746173324873013e-06, "loss": 0.0663, "step": 36013 }, { "epoch": 11.08, "learning_rate": 8.74567898850198e-06, "loss": 0.0573, "step": 36014 }, { "epoch": 11.08, "learning_rate": 8.745184655245088e-06, "loss": 0.0612, "step": 36015 }, { "epoch": 11.08, "learning_rate": 8.744690325103561e-06, "loss": 0.0375, "step": 36016 }, { "epoch": 11.08, "learning_rate": 8.744195998078632e-06, "loss": 0.3091, "step": 36017 }, { "epoch": 11.08, "learning_rate": 8.743701674171526e-06, "loss": 0.0481, "step": 36018 }, { "epoch": 11.08, "learning_rate": 8.743207353383469e-06, "loss": 0.0937, "step": 36019 }, { "epoch": 11.08, "learning_rate": 8.74271303571569e-06, "loss": 0.0485, "step": 36020 }, { "epoch": 11.08, "learning_rate": 8.74221872116942e-06, "loss": 0.0378, "step": 36021 }, { "epoch": 11.08, "learning_rate": 8.741724409745873e-06, "loss": 0.028, "step": 36022 }, { "epoch": 11.08, "learning_rate": 8.741230101446289e-06, "loss": 0.0178, "step": 36023 }, { "epoch": 11.08, "learning_rate": 8.740735796271896e-06, "loss": 0.0535, "step": 36024 }, { "epoch": 11.08, "learning_rate": 8.740241494223911e-06, "loss": 0.0407, "step": 36025 }, { "epoch": 11.08, "learning_rate": 8.739747195303569e-06, "loss": 0.0836, "step": 36026 }, { "epoch": 11.09, "learning_rate": 8.739252899512091e-06, "loss": 0.0475, "step": 36027 }, { "epoch": 11.09, "learning_rate": 8.738758606850713e-06, "loss": 0.0319, "step": 36028 }, { "epoch": 11.09, "learning_rate": 8.738264317320655e-06, "loss": 0.037, "step": 36029 }, { "epoch": 11.09, "learning_rate": 8.737770030923148e-06, "loss": 0.0306, "step": 36030 }, { "epoch": 11.09, "learning_rate": 8.737275747659416e-06, "loss": 0.058, "step": 36031 }, { "epoch": 11.09, "learning_rate": 8.736781467530688e-06, "loss": 0.0291, "step": 36032 }, { "epoch": 11.09, "learning_rate": 8.736287190538194e-06, "loss": 0.0438, "step": 36033 }, { "epoch": 11.09, "learning_rate": 8.735792916683157e-06, "loss": 0.0462, "step": 36034 }, { "epoch": 11.09, "learning_rate": 8.735298645966806e-06, "loss": 0.0392, "step": 36035 }, { "epoch": 11.09, "learning_rate": 8.734804378390366e-06, "loss": 0.0692, "step": 36036 }, { "epoch": 11.09, "learning_rate": 8.734310113955064e-06, "loss": 0.0331, "step": 36037 }, { "epoch": 11.09, "learning_rate": 8.73381585266213e-06, "loss": 0.0425, "step": 36038 }, { "epoch": 11.09, "learning_rate": 8.73332159451279e-06, "loss": 0.0573, "step": 36039 }, { "epoch": 11.09, "learning_rate": 8.732827339508272e-06, "loss": 0.0549, "step": 36040 }, { "epoch": 11.09, "learning_rate": 8.732333087649802e-06, "loss": 0.0381, "step": 36041 }, { "epoch": 11.09, "learning_rate": 8.731838838938604e-06, "loss": 0.0597, "step": 36042 }, { "epoch": 11.09, "learning_rate": 8.731344593375912e-06, "loss": 0.1023, "step": 36043 }, { "epoch": 11.09, "learning_rate": 8.730850350962948e-06, "loss": 0.0787, "step": 36044 }, { "epoch": 11.09, "learning_rate": 8.730356111700944e-06, "loss": 0.0591, "step": 36045 }, { "epoch": 11.09, "learning_rate": 8.72986187559112e-06, "loss": 0.0537, "step": 36046 }, { "epoch": 11.09, "learning_rate": 8.729367642634703e-06, "loss": 0.0827, "step": 36047 }, { "epoch": 11.09, "learning_rate": 8.728873412832928e-06, "loss": 0.0451, "step": 36048 }, { "epoch": 11.09, "learning_rate": 8.728379186187019e-06, "loss": 0.0434, "step": 36049 }, { "epoch": 11.09, "learning_rate": 8.7278849626982e-06, "loss": 0.0457, "step": 36050 }, { "epoch": 11.09, "learning_rate": 8.727390742367698e-06, "loss": 0.037, "step": 36051 }, { "epoch": 11.09, "learning_rate": 8.726896525196741e-06, "loss": 0.0606, "step": 36052 }, { "epoch": 11.09, "learning_rate": 8.72640231118656e-06, "loss": 0.0492, "step": 36053 }, { "epoch": 11.09, "learning_rate": 8.725908100338377e-06, "loss": 0.0483, "step": 36054 }, { "epoch": 11.09, "learning_rate": 8.725413892653424e-06, "loss": 0.0909, "step": 36055 }, { "epoch": 11.09, "learning_rate": 8.724919688132918e-06, "loss": 0.0411, "step": 36056 }, { "epoch": 11.09, "learning_rate": 8.724425486778096e-06, "loss": 0.0519, "step": 36057 }, { "epoch": 11.09, "learning_rate": 8.723931288590184e-06, "loss": 0.0492, "step": 36058 }, { "epoch": 11.1, "learning_rate": 8.723437093570406e-06, "loss": 0.0662, "step": 36059 }, { "epoch": 11.1, "learning_rate": 8.722942901719989e-06, "loss": 0.0544, "step": 36060 }, { "epoch": 11.1, "learning_rate": 8.722448713040158e-06, "loss": 0.0671, "step": 36061 }, { "epoch": 11.1, "learning_rate": 8.721954527532143e-06, "loss": 0.0378, "step": 36062 }, { "epoch": 11.1, "learning_rate": 8.721460345197173e-06, "loss": 0.0437, "step": 36063 }, { "epoch": 11.1, "learning_rate": 8.720966166036471e-06, "loss": 0.0696, "step": 36064 }, { "epoch": 11.1, "learning_rate": 8.720471990051265e-06, "loss": 0.0483, "step": 36065 }, { "epoch": 11.1, "learning_rate": 8.71997781724278e-06, "loss": 0.0319, "step": 36066 }, { "epoch": 11.1, "learning_rate": 8.719483647612247e-06, "loss": 0.0436, "step": 36067 }, { "epoch": 11.1, "learning_rate": 8.718989481160892e-06, "loss": 0.0364, "step": 36068 }, { "epoch": 11.1, "learning_rate": 8.718495317889941e-06, "loss": 0.0356, "step": 36069 }, { "epoch": 11.1, "learning_rate": 8.71800115780062e-06, "loss": 0.0226, "step": 36070 }, { "epoch": 11.1, "learning_rate": 8.717507000894152e-06, "loss": 0.0585, "step": 36071 }, { "epoch": 11.1, "learning_rate": 8.717012847171775e-06, "loss": 0.0522, "step": 36072 }, { "epoch": 11.1, "learning_rate": 8.716518696634707e-06, "loss": 0.0352, "step": 36073 }, { "epoch": 11.1, "learning_rate": 8.716024549284176e-06, "loss": 0.0471, "step": 36074 }, { "epoch": 11.1, "learning_rate": 8.71553040512141e-06, "loss": 0.0511, "step": 36075 }, { "epoch": 11.1, "learning_rate": 8.715036264147635e-06, "loss": 0.0388, "step": 36076 }, { "epoch": 11.1, "learning_rate": 8.71454212636408e-06, "loss": 0.091, "step": 36077 }, { "epoch": 11.1, "learning_rate": 8.71404799177197e-06, "loss": 0.0458, "step": 36078 }, { "epoch": 11.1, "learning_rate": 8.713553860372534e-06, "loss": 0.05, "step": 36079 }, { "epoch": 11.1, "learning_rate": 8.713059732166996e-06, "loss": 0.0262, "step": 36080 }, { "epoch": 11.1, "learning_rate": 8.712565607156579e-06, "loss": 0.0684, "step": 36081 }, { "epoch": 11.1, "learning_rate": 8.712071485342521e-06, "loss": 0.0529, "step": 36082 }, { "epoch": 11.1, "learning_rate": 8.71157736672604e-06, "loss": 0.0313, "step": 36083 }, { "epoch": 11.1, "learning_rate": 8.711083251308366e-06, "loss": 0.0563, "step": 36084 }, { "epoch": 11.1, "learning_rate": 8.710589139090722e-06, "loss": 0.0385, "step": 36085 }, { "epoch": 11.1, "learning_rate": 8.710095030074336e-06, "loss": 0.0756, "step": 36086 }, { "epoch": 11.1, "learning_rate": 8.709600924260441e-06, "loss": 0.0351, "step": 36087 }, { "epoch": 11.1, "learning_rate": 8.709106821650259e-06, "loss": 0.056, "step": 36088 }, { "epoch": 11.1, "learning_rate": 8.708612722245014e-06, "loss": 0.0671, "step": 36089 }, { "epoch": 11.1, "learning_rate": 8.708118626045939e-06, "loss": 0.0419, "step": 36090 }, { "epoch": 11.1, "learning_rate": 8.707624533054251e-06, "loss": 0.0323, "step": 36091 }, { "epoch": 11.11, "learning_rate": 8.707130443271188e-06, "loss": 0.0355, "step": 36092 }, { "epoch": 11.11, "learning_rate": 8.706636356697972e-06, "loss": 0.0391, "step": 36093 }, { "epoch": 11.11, "learning_rate": 8.706142273335827e-06, "loss": 0.0448, "step": 36094 }, { "epoch": 11.11, "learning_rate": 8.705648193185982e-06, "loss": 0.073, "step": 36095 }, { "epoch": 11.11, "learning_rate": 8.705154116249663e-06, "loss": 0.0217, "step": 36096 }, { "epoch": 11.11, "learning_rate": 8.704660042528098e-06, "loss": 0.0638, "step": 36097 }, { "epoch": 11.11, "learning_rate": 8.704165972022514e-06, "loss": 0.0385, "step": 36098 }, { "epoch": 11.11, "learning_rate": 8.703671904734136e-06, "loss": 0.0528, "step": 36099 }, { "epoch": 11.11, "learning_rate": 8.703177840664193e-06, "loss": 0.0801, "step": 36100 }, { "epoch": 11.11, "learning_rate": 8.702683779813904e-06, "loss": 0.0403, "step": 36101 }, { "epoch": 11.11, "learning_rate": 8.702189722184505e-06, "loss": 0.0623, "step": 36102 }, { "epoch": 11.11, "learning_rate": 8.701695667777221e-06, "loss": 0.0471, "step": 36103 }, { "epoch": 11.11, "learning_rate": 8.701201616593275e-06, "loss": 0.05, "step": 36104 }, { "epoch": 11.11, "learning_rate": 8.700707568633896e-06, "loss": 0.0465, "step": 36105 }, { "epoch": 11.11, "learning_rate": 8.700213523900307e-06, "loss": 0.0312, "step": 36106 }, { "epoch": 11.11, "learning_rate": 8.69971948239374e-06, "loss": 0.0653, "step": 36107 }, { "epoch": 11.11, "learning_rate": 8.699225444115418e-06, "loss": 0.07, "step": 36108 }, { "epoch": 11.11, "learning_rate": 8.69873140906657e-06, "loss": 0.0435, "step": 36109 }, { "epoch": 11.11, "learning_rate": 8.698237377248419e-06, "loss": 0.059, "step": 36110 }, { "epoch": 11.11, "learning_rate": 8.697743348662192e-06, "loss": 0.0463, "step": 36111 }, { "epoch": 11.11, "learning_rate": 8.697249323309121e-06, "loss": 0.0373, "step": 36112 }, { "epoch": 11.11, "learning_rate": 8.69675530119043e-06, "loss": 0.0352, "step": 36113 }, { "epoch": 11.11, "learning_rate": 8.696261282307342e-06, "loss": 0.0503, "step": 36114 }, { "epoch": 11.11, "learning_rate": 8.695767266661085e-06, "loss": 0.0363, "step": 36115 }, { "epoch": 11.11, "learning_rate": 8.695273254252884e-06, "loss": 0.049, "step": 36116 }, { "epoch": 11.11, "learning_rate": 8.694779245083972e-06, "loss": 0.024, "step": 36117 }, { "epoch": 11.11, "learning_rate": 8.69428523915557e-06, "loss": 0.0429, "step": 36118 }, { "epoch": 11.11, "learning_rate": 8.693791236468905e-06, "loss": 0.0406, "step": 36119 }, { "epoch": 11.11, "learning_rate": 8.693297237025206e-06, "loss": 0.0515, "step": 36120 }, { "epoch": 11.11, "learning_rate": 8.692803240825693e-06, "loss": 0.0632, "step": 36121 }, { "epoch": 11.11, "learning_rate": 8.692309247871603e-06, "loss": 0.0308, "step": 36122 }, { "epoch": 11.11, "learning_rate": 8.691815258164153e-06, "loss": 0.0551, "step": 36123 }, { "epoch": 11.12, "learning_rate": 8.691321271704576e-06, "loss": 0.0361, "step": 36124 }, { "epoch": 11.12, "learning_rate": 8.690827288494094e-06, "loss": 0.0418, "step": 36125 }, { "epoch": 11.12, "learning_rate": 8.690333308533932e-06, "loss": 0.0605, "step": 36126 }, { "epoch": 11.12, "learning_rate": 8.689839331825322e-06, "loss": 0.0306, "step": 36127 }, { "epoch": 11.12, "learning_rate": 8.689345358369489e-06, "loss": 0.0354, "step": 36128 }, { "epoch": 11.12, "learning_rate": 8.688851388167658e-06, "loss": 0.0332, "step": 36129 }, { "epoch": 11.12, "learning_rate": 8.688357421221054e-06, "loss": 0.0532, "step": 36130 }, { "epoch": 11.12, "learning_rate": 8.687863457530903e-06, "loss": 0.0571, "step": 36131 }, { "epoch": 11.12, "learning_rate": 8.687369497098435e-06, "loss": 0.0379, "step": 36132 }, { "epoch": 11.12, "learning_rate": 8.686875539924875e-06, "loss": 0.0382, "step": 36133 }, { "epoch": 11.12, "learning_rate": 8.686381586011452e-06, "loss": 0.0443, "step": 36134 }, { "epoch": 11.12, "learning_rate": 8.685887635359387e-06, "loss": 0.0763, "step": 36135 }, { "epoch": 11.12, "learning_rate": 8.685393687969905e-06, "loss": 0.0447, "step": 36136 }, { "epoch": 11.12, "learning_rate": 8.68489974384424e-06, "loss": 0.0581, "step": 36137 }, { "epoch": 11.12, "learning_rate": 8.684405802983615e-06, "loss": 0.0433, "step": 36138 }, { "epoch": 11.12, "learning_rate": 8.683911865389254e-06, "loss": 0.0545, "step": 36139 }, { "epoch": 11.12, "learning_rate": 8.683417931062385e-06, "loss": 0.025, "step": 36140 }, { "epoch": 11.12, "learning_rate": 8.682924000004234e-06, "loss": 0.0562, "step": 36141 }, { "epoch": 11.12, "learning_rate": 8.682430072216029e-06, "loss": 0.0454, "step": 36142 }, { "epoch": 11.12, "learning_rate": 8.681936147698994e-06, "loss": 0.0321, "step": 36143 }, { "epoch": 11.12, "learning_rate": 8.681442226454358e-06, "loss": 0.0317, "step": 36144 }, { "epoch": 11.12, "learning_rate": 8.680948308483346e-06, "loss": 0.0395, "step": 36145 }, { "epoch": 11.12, "learning_rate": 8.680454393787178e-06, "loss": 0.0194, "step": 36146 }, { "epoch": 11.12, "learning_rate": 8.679960482367088e-06, "loss": 0.018, "step": 36147 }, { "epoch": 11.12, "learning_rate": 8.679466574224305e-06, "loss": 0.0561, "step": 36148 }, { "epoch": 11.12, "learning_rate": 8.67897266936005e-06, "loss": 0.0426, "step": 36149 }, { "epoch": 11.12, "learning_rate": 8.678478767775547e-06, "loss": 0.037, "step": 36150 }, { "epoch": 11.12, "learning_rate": 8.677984869472022e-06, "loss": 0.0266, "step": 36151 }, { "epoch": 11.12, "learning_rate": 8.677490974450708e-06, "loss": 0.0609, "step": 36152 }, { "epoch": 11.12, "learning_rate": 8.676997082712827e-06, "loss": 0.0504, "step": 36153 }, { "epoch": 11.12, "learning_rate": 8.676503194259606e-06, "loss": 0.028, "step": 36154 }, { "epoch": 11.12, "learning_rate": 8.676009309092273e-06, "loss": 0.0864, "step": 36155 }, { "epoch": 11.12, "learning_rate": 8.675515427212044e-06, "loss": 0.0529, "step": 36156 }, { "epoch": 11.13, "learning_rate": 8.675021548620158e-06, "loss": 0.058, "step": 36157 }, { "epoch": 11.13, "learning_rate": 8.67452767331784e-06, "loss": 0.0541, "step": 36158 }, { "epoch": 11.13, "learning_rate": 8.674033801306308e-06, "loss": 0.0587, "step": 36159 }, { "epoch": 11.13, "learning_rate": 8.673539932586794e-06, "loss": 0.0501, "step": 36160 }, { "epoch": 11.13, "learning_rate": 8.673046067160521e-06, "loss": 0.0688, "step": 36161 }, { "epoch": 11.13, "learning_rate": 8.672552205028719e-06, "loss": 0.0363, "step": 36162 }, { "epoch": 11.13, "learning_rate": 8.672058346192611e-06, "loss": 0.0419, "step": 36163 }, { "epoch": 11.13, "learning_rate": 8.671564490653423e-06, "loss": 0.0368, "step": 36164 }, { "epoch": 11.13, "learning_rate": 8.671070638412385e-06, "loss": 0.0362, "step": 36165 }, { "epoch": 11.13, "learning_rate": 8.670576789470717e-06, "loss": 0.0527, "step": 36166 }, { "epoch": 11.13, "learning_rate": 8.670082943829649e-06, "loss": 0.0872, "step": 36167 }, { "epoch": 11.13, "learning_rate": 8.66958910149041e-06, "loss": 0.0149, "step": 36168 }, { "epoch": 11.13, "learning_rate": 8.669095262454222e-06, "loss": 0.0348, "step": 36169 }, { "epoch": 11.13, "learning_rate": 8.66860142672231e-06, "loss": 0.0518, "step": 36170 }, { "epoch": 11.13, "learning_rate": 8.6681075942959e-06, "loss": 0.045, "step": 36171 }, { "epoch": 11.13, "learning_rate": 8.667613765176222e-06, "loss": 0.0698, "step": 36172 }, { "epoch": 11.13, "learning_rate": 8.6671199393645e-06, "loss": 0.0473, "step": 36173 }, { "epoch": 11.13, "learning_rate": 8.66662611686196e-06, "loss": 0.038, "step": 36174 }, { "epoch": 11.13, "learning_rate": 8.666132297669827e-06, "loss": 0.0368, "step": 36175 }, { "epoch": 11.13, "learning_rate": 8.665638481789324e-06, "loss": 0.0577, "step": 36176 }, { "epoch": 11.13, "learning_rate": 8.665144669221687e-06, "loss": 0.0605, "step": 36177 }, { "epoch": 11.13, "learning_rate": 8.664650859968133e-06, "loss": 0.048, "step": 36178 }, { "epoch": 11.13, "learning_rate": 8.664157054029895e-06, "loss": 0.0685, "step": 36179 }, { "epoch": 11.13, "learning_rate": 8.663663251408191e-06, "loss": 0.0334, "step": 36180 }, { "epoch": 11.13, "learning_rate": 8.663169452104248e-06, "loss": 0.0542, "step": 36181 }, { "epoch": 11.13, "learning_rate": 8.662675656119301e-06, "loss": 0.0319, "step": 36182 }, { "epoch": 11.13, "learning_rate": 8.662181863454568e-06, "loss": 0.0586, "step": 36183 }, { "epoch": 11.13, "learning_rate": 8.661688074111276e-06, "loss": 0.0403, "step": 36184 }, { "epoch": 11.13, "learning_rate": 8.66119428809065e-06, "loss": 0.0372, "step": 36185 }, { "epoch": 11.13, "learning_rate": 8.660700505393919e-06, "loss": 0.0495, "step": 36186 }, { "epoch": 11.13, "learning_rate": 8.660206726022307e-06, "loss": 0.0731, "step": 36187 }, { "epoch": 11.13, "learning_rate": 8.659712949977043e-06, "loss": 0.0499, "step": 36188 }, { "epoch": 11.14, "learning_rate": 8.659219177259348e-06, "loss": 0.0581, "step": 36189 }, { "epoch": 11.14, "learning_rate": 8.658725407870454e-06, "loss": 0.0612, "step": 36190 }, { "epoch": 11.14, "learning_rate": 8.658231641811576e-06, "loss": 0.0332, "step": 36191 }, { "epoch": 11.14, "learning_rate": 8.657737879083954e-06, "loss": 0.0615, "step": 36192 }, { "epoch": 11.14, "learning_rate": 8.657244119688806e-06, "loss": 0.0479, "step": 36193 }, { "epoch": 11.14, "learning_rate": 8.656750363627356e-06, "loss": 0.0341, "step": 36194 }, { "epoch": 11.14, "learning_rate": 8.656256610900834e-06, "loss": 0.0579, "step": 36195 }, { "epoch": 11.14, "learning_rate": 8.655762861510464e-06, "loss": 0.0379, "step": 36196 }, { "epoch": 11.14, "learning_rate": 8.655269115457474e-06, "loss": 0.0719, "step": 36197 }, { "epoch": 11.14, "learning_rate": 8.654775372743086e-06, "loss": 0.0727, "step": 36198 }, { "epoch": 11.14, "learning_rate": 8.65428163336853e-06, "loss": 0.0587, "step": 36199 }, { "epoch": 11.14, "learning_rate": 8.653787897335032e-06, "loss": 0.0611, "step": 36200 }, { "epoch": 11.14, "learning_rate": 8.653294164643809e-06, "loss": 0.0543, "step": 36201 }, { "epoch": 11.14, "learning_rate": 8.652800435296098e-06, "loss": 0.0645, "step": 36202 }, { "epoch": 11.14, "learning_rate": 8.652306709293121e-06, "loss": 0.0473, "step": 36203 }, { "epoch": 11.14, "learning_rate": 8.651812986636103e-06, "loss": 0.0213, "step": 36204 }, { "epoch": 11.14, "learning_rate": 8.65131926732627e-06, "loss": 0.0491, "step": 36205 }, { "epoch": 11.14, "learning_rate": 8.650825551364844e-06, "loss": 0.0431, "step": 36206 }, { "epoch": 11.14, "learning_rate": 8.650331838753057e-06, "loss": 0.049, "step": 36207 }, { "epoch": 11.14, "learning_rate": 8.649838129492132e-06, "loss": 0.0446, "step": 36208 }, { "epoch": 11.14, "learning_rate": 8.649344423583295e-06, "loss": 0.0868, "step": 36209 }, { "epoch": 11.14, "learning_rate": 8.648850721027773e-06, "loss": 0.0701, "step": 36210 }, { "epoch": 11.14, "learning_rate": 8.648357021826786e-06, "loss": 0.0544, "step": 36211 }, { "epoch": 11.14, "learning_rate": 8.647863325981568e-06, "loss": 0.0275, "step": 36212 }, { "epoch": 11.14, "learning_rate": 8.647369633493344e-06, "loss": 0.0629, "step": 36213 }, { "epoch": 11.14, "learning_rate": 8.646875944363332e-06, "loss": 0.0567, "step": 36214 }, { "epoch": 11.14, "learning_rate": 8.646382258592762e-06, "loss": 0.0289, "step": 36215 }, { "epoch": 11.14, "learning_rate": 8.64588857618286e-06, "loss": 0.0719, "step": 36216 }, { "epoch": 11.14, "learning_rate": 8.645394897134852e-06, "loss": 0.0431, "step": 36217 }, { "epoch": 11.14, "learning_rate": 8.644901221449964e-06, "loss": 0.0556, "step": 36218 }, { "epoch": 11.14, "learning_rate": 8.64440754912942e-06, "loss": 0.0243, "step": 36219 }, { "epoch": 11.14, "learning_rate": 8.643913880174449e-06, "loss": 0.0351, "step": 36220 }, { "epoch": 11.14, "learning_rate": 8.64342021458627e-06, "loss": 0.0386, "step": 36221 }, { "epoch": 11.15, "learning_rate": 8.642926552366115e-06, "loss": 0.0572, "step": 36222 }, { "epoch": 11.15, "learning_rate": 8.64243289351521e-06, "loss": 0.0549, "step": 36223 }, { "epoch": 11.15, "learning_rate": 8.641939238034777e-06, "loss": 0.0665, "step": 36224 }, { "epoch": 11.15, "learning_rate": 8.641445585926042e-06, "loss": 0.0756, "step": 36225 }, { "epoch": 11.15, "learning_rate": 8.640951937190228e-06, "loss": 0.061, "step": 36226 }, { "epoch": 11.15, "learning_rate": 8.64045829182857e-06, "loss": 0.0574, "step": 36227 }, { "epoch": 11.15, "learning_rate": 8.639964649842285e-06, "loss": 0.0466, "step": 36228 }, { "epoch": 11.15, "learning_rate": 8.6394710112326e-06, "loss": 0.0779, "step": 36229 }, { "epoch": 11.15, "learning_rate": 8.638977376000743e-06, "loss": 0.0892, "step": 36230 }, { "epoch": 11.15, "learning_rate": 8.638483744147935e-06, "loss": 0.0293, "step": 36231 }, { "epoch": 11.15, "learning_rate": 8.637990115675409e-06, "loss": 0.086, "step": 36232 }, { "epoch": 11.15, "learning_rate": 8.637496490584385e-06, "loss": 0.0391, "step": 36233 }, { "epoch": 11.15, "learning_rate": 8.637002868876092e-06, "loss": 0.0318, "step": 36234 }, { "epoch": 11.15, "learning_rate": 8.636509250551752e-06, "loss": 0.0643, "step": 36235 }, { "epoch": 11.15, "learning_rate": 8.636015635612586e-06, "loss": 0.0488, "step": 36236 }, { "epoch": 11.15, "learning_rate": 8.635522024059833e-06, "loss": 0.0291, "step": 36237 }, { "epoch": 11.15, "learning_rate": 8.635028415894708e-06, "loss": 0.0574, "step": 36238 }, { "epoch": 11.15, "learning_rate": 8.634534811118442e-06, "loss": 0.0649, "step": 36239 }, { "epoch": 11.15, "learning_rate": 8.634041209732255e-06, "loss": 0.0543, "step": 36240 }, { "epoch": 11.15, "learning_rate": 8.633547611737376e-06, "loss": 0.0719, "step": 36241 }, { "epoch": 11.15, "learning_rate": 8.633054017135029e-06, "loss": 0.0488, "step": 36242 }, { "epoch": 11.15, "learning_rate": 8.632560425926442e-06, "loss": 0.0268, "step": 36243 }, { "epoch": 11.15, "learning_rate": 8.632066838112838e-06, "loss": 0.0321, "step": 36244 }, { "epoch": 11.15, "learning_rate": 8.631573253695446e-06, "loss": 0.0448, "step": 36245 }, { "epoch": 11.15, "learning_rate": 8.631079672675483e-06, "loss": 0.0385, "step": 36246 }, { "epoch": 11.15, "learning_rate": 8.630586095054185e-06, "loss": 0.0485, "step": 36247 }, { "epoch": 11.15, "learning_rate": 8.630092520832772e-06, "loss": 0.0786, "step": 36248 }, { "epoch": 11.15, "learning_rate": 8.62959895001247e-06, "loss": 0.0759, "step": 36249 }, { "epoch": 11.15, "learning_rate": 8.629105382594503e-06, "loss": 0.0284, "step": 36250 }, { "epoch": 11.15, "learning_rate": 8.628611818580097e-06, "loss": 0.0346, "step": 36251 }, { "epoch": 11.15, "learning_rate": 8.62811825797048e-06, "loss": 0.0518, "step": 36252 }, { "epoch": 11.15, "learning_rate": 8.627624700766875e-06, "loss": 0.1268, "step": 36253 }, { "epoch": 11.16, "learning_rate": 8.627131146970507e-06, "loss": 0.056, "step": 36254 }, { "epoch": 11.16, "learning_rate": 8.626637596582605e-06, "loss": 0.0337, "step": 36255 }, { "epoch": 11.16, "learning_rate": 8.626144049604386e-06, "loss": 0.0331, "step": 36256 }, { "epoch": 11.16, "learning_rate": 8.625650506037084e-06, "loss": 0.0403, "step": 36257 }, { "epoch": 11.16, "learning_rate": 8.625156965881925e-06, "loss": 0.0469, "step": 36258 }, { "epoch": 11.16, "learning_rate": 8.624663429140128e-06, "loss": 0.0527, "step": 36259 }, { "epoch": 11.16, "learning_rate": 8.624169895812919e-06, "loss": 0.0397, "step": 36260 }, { "epoch": 11.16, "learning_rate": 8.623676365901526e-06, "loss": 0.0451, "step": 36261 }, { "epoch": 11.16, "learning_rate": 8.623182839407174e-06, "loss": 0.0535, "step": 36262 }, { "epoch": 11.16, "learning_rate": 8.622689316331088e-06, "loss": 0.0528, "step": 36263 }, { "epoch": 11.16, "learning_rate": 8.622195796674492e-06, "loss": 0.0561, "step": 36264 }, { "epoch": 11.16, "learning_rate": 8.621702280438615e-06, "loss": 0.0459, "step": 36265 }, { "epoch": 11.16, "learning_rate": 8.621208767624676e-06, "loss": 0.0442, "step": 36266 }, { "epoch": 11.16, "learning_rate": 8.620715258233906e-06, "loss": 0.0752, "step": 36267 }, { "epoch": 11.16, "learning_rate": 8.620221752267528e-06, "loss": 0.0328, "step": 36268 }, { "epoch": 11.16, "learning_rate": 8.619728249726771e-06, "loss": 0.0445, "step": 36269 }, { "epoch": 11.16, "learning_rate": 8.619234750612853e-06, "loss": 0.0266, "step": 36270 }, { "epoch": 11.16, "learning_rate": 8.618741254927001e-06, "loss": 0.0314, "step": 36271 }, { "epoch": 11.16, "learning_rate": 8.618247762670445e-06, "loss": 0.041, "step": 36272 }, { "epoch": 11.16, "learning_rate": 8.617754273844407e-06, "loss": 0.0386, "step": 36273 }, { "epoch": 11.16, "learning_rate": 8.617260788450111e-06, "loss": 0.0915, "step": 36274 }, { "epoch": 11.16, "learning_rate": 8.616767306488785e-06, "loss": 0.0219, "step": 36275 }, { "epoch": 11.16, "learning_rate": 8.61627382796165e-06, "loss": 0.0423, "step": 36276 }, { "epoch": 11.16, "learning_rate": 8.615780352869938e-06, "loss": 0.056, "step": 36277 }, { "epoch": 11.16, "learning_rate": 8.615286881214869e-06, "loss": 0.0273, "step": 36278 }, { "epoch": 11.16, "learning_rate": 8.614793412997671e-06, "loss": 0.0746, "step": 36279 }, { "epoch": 11.16, "learning_rate": 8.614299948219566e-06, "loss": 0.062, "step": 36280 }, { "epoch": 11.16, "learning_rate": 8.613806486881775e-06, "loss": 0.0531, "step": 36281 }, { "epoch": 11.16, "learning_rate": 8.613313028985539e-06, "loss": 0.0373, "step": 36282 }, { "epoch": 11.16, "learning_rate": 8.612819574532067e-06, "loss": 0.0403, "step": 36283 }, { "epoch": 11.16, "learning_rate": 8.612326123522591e-06, "loss": 0.019, "step": 36284 }, { "epoch": 11.16, "learning_rate": 8.611832675958335e-06, "loss": 0.0477, "step": 36285 }, { "epoch": 11.16, "learning_rate": 8.611339231840522e-06, "loss": 0.0517, "step": 36286 }, { "epoch": 11.17, "learning_rate": 8.610845791170383e-06, "loss": 0.0372, "step": 36287 }, { "epoch": 11.17, "learning_rate": 8.610352353949138e-06, "loss": 0.062, "step": 36288 }, { "epoch": 11.17, "learning_rate": 8.609858920178015e-06, "loss": 0.0578, "step": 36289 }, { "epoch": 11.17, "learning_rate": 8.609365489858237e-06, "loss": 0.0907, "step": 36290 }, { "epoch": 11.17, "learning_rate": 8.608872062991025e-06, "loss": 0.027, "step": 36291 }, { "epoch": 11.17, "learning_rate": 8.608378639577614e-06, "loss": 0.0365, "step": 36292 }, { "epoch": 11.17, "learning_rate": 8.607885219619223e-06, "loss": 0.0637, "step": 36293 }, { "epoch": 11.17, "learning_rate": 8.607391803117078e-06, "loss": 0.0178, "step": 36294 }, { "epoch": 11.17, "learning_rate": 8.606898390072403e-06, "loss": 0.0464, "step": 36295 }, { "epoch": 11.17, "learning_rate": 8.60640498048642e-06, "loss": 0.0628, "step": 36296 }, { "epoch": 11.17, "learning_rate": 8.605911574360362e-06, "loss": 0.0461, "step": 36297 }, { "epoch": 11.17, "learning_rate": 8.605418171695447e-06, "loss": 0.0766, "step": 36298 }, { "epoch": 11.17, "learning_rate": 8.604924772492905e-06, "loss": 0.0434, "step": 36299 }, { "epoch": 11.17, "learning_rate": 8.604431376753961e-06, "loss": 0.0355, "step": 36300 }, { "epoch": 11.17, "learning_rate": 8.60393798447983e-06, "loss": 0.0366, "step": 36301 }, { "epoch": 11.17, "learning_rate": 8.603444595671751e-06, "loss": 0.0347, "step": 36302 }, { "epoch": 11.17, "learning_rate": 8.602951210330942e-06, "loss": 0.0374, "step": 36303 }, { "epoch": 11.17, "learning_rate": 8.602457828458628e-06, "loss": 0.0478, "step": 36304 }, { "epoch": 11.17, "learning_rate": 8.601964450056033e-06, "loss": 0.0451, "step": 36305 }, { "epoch": 11.17, "learning_rate": 8.601471075124383e-06, "loss": 0.03, "step": 36306 }, { "epoch": 11.17, "learning_rate": 8.600977703664904e-06, "loss": 0.0334, "step": 36307 }, { "epoch": 11.17, "learning_rate": 8.600484335678821e-06, "loss": 0.0358, "step": 36308 }, { "epoch": 11.17, "learning_rate": 8.599990971167359e-06, "loss": 0.0337, "step": 36309 }, { "epoch": 11.17, "learning_rate": 8.59949761013174e-06, "loss": 0.0341, "step": 36310 }, { "epoch": 11.17, "learning_rate": 8.599004252573191e-06, "loss": 0.0282, "step": 36311 }, { "epoch": 11.17, "learning_rate": 8.598510898492936e-06, "loss": 0.0336, "step": 36312 }, { "epoch": 11.17, "learning_rate": 8.598017547892206e-06, "loss": 0.0389, "step": 36313 }, { "epoch": 11.17, "learning_rate": 8.597524200772215e-06, "loss": 0.0621, "step": 36314 }, { "epoch": 11.17, "learning_rate": 8.597030857134196e-06, "loss": 0.0406, "step": 36315 }, { "epoch": 11.17, "learning_rate": 8.596537516979369e-06, "loss": 0.049, "step": 36316 }, { "epoch": 11.17, "learning_rate": 8.596044180308961e-06, "loss": 0.0468, "step": 36317 }, { "epoch": 11.17, "learning_rate": 8.595550847124197e-06, "loss": 0.0597, "step": 36318 }, { "epoch": 11.18, "learning_rate": 8.595057517426304e-06, "loss": 0.0387, "step": 36319 }, { "epoch": 11.18, "learning_rate": 8.594564191216503e-06, "loss": 0.0628, "step": 36320 }, { "epoch": 11.18, "learning_rate": 8.594070868496018e-06, "loss": 0.0565, "step": 36321 }, { "epoch": 11.18, "learning_rate": 8.593577549266078e-06, "loss": 0.0246, "step": 36322 }, { "epoch": 11.18, "learning_rate": 8.593084233527905e-06, "loss": 0.0344, "step": 36323 }, { "epoch": 11.18, "learning_rate": 8.592590921282728e-06, "loss": 0.0442, "step": 36324 }, { "epoch": 11.18, "learning_rate": 8.592097612531766e-06, "loss": 0.0377, "step": 36325 }, { "epoch": 11.18, "learning_rate": 8.591604307276242e-06, "loss": 0.0417, "step": 36326 }, { "epoch": 11.18, "learning_rate": 8.591111005517391e-06, "loss": 0.0323, "step": 36327 }, { "epoch": 11.18, "learning_rate": 8.59061770725643e-06, "loss": 0.0471, "step": 36328 }, { "epoch": 11.18, "learning_rate": 8.590124412494584e-06, "loss": 0.0457, "step": 36329 }, { "epoch": 11.18, "learning_rate": 8.58963112123308e-06, "loss": 0.0393, "step": 36330 }, { "epoch": 11.18, "learning_rate": 8.589137833473137e-06, "loss": 0.0478, "step": 36331 }, { "epoch": 11.18, "learning_rate": 8.58864454921599e-06, "loss": 0.0209, "step": 36332 }, { "epoch": 11.18, "learning_rate": 8.588151268462857e-06, "loss": 0.0352, "step": 36333 }, { "epoch": 11.18, "learning_rate": 8.587657991214965e-06, "loss": 0.0408, "step": 36334 }, { "epoch": 11.18, "learning_rate": 8.587164717473537e-06, "loss": 0.04, "step": 36335 }, { "epoch": 11.18, "learning_rate": 8.586671447239792e-06, "loss": 0.0483, "step": 36336 }, { "epoch": 11.18, "learning_rate": 8.586178180514968e-06, "loss": 0.0672, "step": 36337 }, { "epoch": 11.18, "learning_rate": 8.58568491730028e-06, "loss": 0.0285, "step": 36338 }, { "epoch": 11.18, "learning_rate": 8.585191657596955e-06, "loss": 0.0786, "step": 36339 }, { "epoch": 11.18, "learning_rate": 8.584698401406219e-06, "loss": 0.0567, "step": 36340 }, { "epoch": 11.18, "learning_rate": 8.584205148729292e-06, "loss": 0.1291, "step": 36341 }, { "epoch": 11.18, "learning_rate": 8.583711899567403e-06, "loss": 0.0496, "step": 36342 }, { "epoch": 11.18, "learning_rate": 8.583218653921777e-06, "loss": 0.0306, "step": 36343 }, { "epoch": 11.18, "learning_rate": 8.582725411793638e-06, "loss": 0.0584, "step": 36344 }, { "epoch": 11.18, "learning_rate": 8.58223217318421e-06, "loss": 0.052, "step": 36345 }, { "epoch": 11.18, "learning_rate": 8.581738938094712e-06, "loss": 0.053, "step": 36346 }, { "epoch": 11.18, "learning_rate": 8.581245706526378e-06, "loss": 0.0496, "step": 36347 }, { "epoch": 11.18, "learning_rate": 8.580752478480431e-06, "loss": 0.0691, "step": 36348 }, { "epoch": 11.18, "learning_rate": 8.58025925395809e-06, "loss": 0.0518, "step": 36349 }, { "epoch": 11.18, "learning_rate": 8.579766032960582e-06, "loss": 0.0602, "step": 36350 }, { "epoch": 11.18, "learning_rate": 8.579272815489131e-06, "loss": 0.0253, "step": 36351 }, { "epoch": 11.19, "learning_rate": 8.578779601544965e-06, "loss": 0.0661, "step": 36352 }, { "epoch": 11.19, "learning_rate": 8.578286391129304e-06, "loss": 0.0353, "step": 36353 }, { "epoch": 11.19, "learning_rate": 8.577793184243376e-06, "loss": 0.0309, "step": 36354 }, { "epoch": 11.19, "learning_rate": 8.577299980888407e-06, "loss": 0.0579, "step": 36355 }, { "epoch": 11.19, "learning_rate": 8.576806781065612e-06, "loss": 0.0282, "step": 36356 }, { "epoch": 11.19, "learning_rate": 8.576313584776226e-06, "loss": 0.0642, "step": 36357 }, { "epoch": 11.19, "learning_rate": 8.575820392021473e-06, "loss": 0.0299, "step": 36358 }, { "epoch": 11.19, "learning_rate": 8.575327202802569e-06, "loss": 0.0259, "step": 36359 }, { "epoch": 11.19, "learning_rate": 8.574834017120746e-06, "loss": 0.015, "step": 36360 }, { "epoch": 11.19, "learning_rate": 8.574340834977222e-06, "loss": 0.0541, "step": 36361 }, { "epoch": 11.19, "learning_rate": 8.573847656373228e-06, "loss": 0.0448, "step": 36362 }, { "epoch": 11.19, "learning_rate": 8.573354481309986e-06, "loss": 0.0405, "step": 36363 }, { "epoch": 11.19, "learning_rate": 8.57286130978872e-06, "loss": 0.0449, "step": 36364 }, { "epoch": 11.19, "learning_rate": 8.572368141810656e-06, "loss": 0.044, "step": 36365 }, { "epoch": 11.19, "learning_rate": 8.571874977377012e-06, "loss": 0.0511, "step": 36366 }, { "epoch": 11.19, "learning_rate": 8.571381816489023e-06, "loss": 0.0512, "step": 36367 }, { "epoch": 11.19, "learning_rate": 8.570888659147904e-06, "loss": 0.062, "step": 36368 }, { "epoch": 11.19, "learning_rate": 8.570395505354888e-06, "loss": 0.0236, "step": 36369 }, { "epoch": 11.19, "learning_rate": 8.569902355111192e-06, "loss": 0.0532, "step": 36370 }, { "epoch": 11.19, "learning_rate": 8.56940920841804e-06, "loss": 0.046, "step": 36371 }, { "epoch": 11.19, "learning_rate": 8.568916065276663e-06, "loss": 0.0933, "step": 36372 }, { "epoch": 11.19, "learning_rate": 8.56842292568828e-06, "loss": 0.0674, "step": 36373 }, { "epoch": 11.19, "learning_rate": 8.567929789654118e-06, "loss": 0.0558, "step": 36374 }, { "epoch": 11.19, "learning_rate": 8.5674366571754e-06, "loss": 0.0721, "step": 36375 }, { "epoch": 11.19, "learning_rate": 8.566943528253348e-06, "loss": 0.0773, "step": 36376 }, { "epoch": 11.19, "learning_rate": 8.566450402889192e-06, "loss": 0.0706, "step": 36377 }, { "epoch": 11.19, "learning_rate": 8.565957281084151e-06, "loss": 0.0865, "step": 36378 }, { "epoch": 11.19, "learning_rate": 8.565464162839456e-06, "loss": 0.0104, "step": 36379 }, { "epoch": 11.19, "learning_rate": 8.564971048156323e-06, "loss": 0.0471, "step": 36380 }, { "epoch": 11.19, "learning_rate": 8.564477937035976e-06, "loss": 0.0581, "step": 36381 }, { "epoch": 11.19, "learning_rate": 8.56398482947965e-06, "loss": 0.0327, "step": 36382 }, { "epoch": 11.19, "learning_rate": 8.563491725488562e-06, "loss": 0.0579, "step": 36383 }, { "epoch": 11.2, "learning_rate": 8.562998625063934e-06, "loss": 0.0315, "step": 36384 }, { "epoch": 11.2, "learning_rate": 8.562505528206995e-06, "loss": 0.0537, "step": 36385 }, { "epoch": 11.2, "learning_rate": 8.562012434918965e-06, "loss": 0.0564, "step": 36386 }, { "epoch": 11.2, "learning_rate": 8.56151934520107e-06, "loss": 0.0707, "step": 36387 }, { "epoch": 11.2, "learning_rate": 8.561026259054538e-06, "loss": 0.0411, "step": 36388 }, { "epoch": 11.2, "learning_rate": 8.560533176480588e-06, "loss": 0.0731, "step": 36389 }, { "epoch": 11.2, "learning_rate": 8.560040097480448e-06, "loss": 0.0504, "step": 36390 }, { "epoch": 11.2, "learning_rate": 8.559547022055335e-06, "loss": 0.0682, "step": 36391 }, { "epoch": 11.2, "learning_rate": 8.559053950206485e-06, "loss": 0.0642, "step": 36392 }, { "epoch": 11.2, "learning_rate": 8.558560881935111e-06, "loss": 0.049, "step": 36393 }, { "epoch": 11.2, "learning_rate": 8.558067817242444e-06, "loss": 0.0622, "step": 36394 }, { "epoch": 11.2, "learning_rate": 8.557574756129707e-06, "loss": 0.0569, "step": 36395 }, { "epoch": 11.2, "learning_rate": 8.557081698598119e-06, "loss": 0.0373, "step": 36396 }, { "epoch": 11.2, "learning_rate": 8.55658864464891e-06, "loss": 0.0378, "step": 36397 }, { "epoch": 11.2, "learning_rate": 8.556095594283302e-06, "loss": 0.0515, "step": 36398 }, { "epoch": 11.2, "learning_rate": 8.55560254750252e-06, "loss": 0.055, "step": 36399 }, { "epoch": 11.2, "learning_rate": 8.55510950430779e-06, "loss": 0.0452, "step": 36400 }, { "epoch": 11.2, "learning_rate": 8.554616464700327e-06, "loss": 0.0406, "step": 36401 }, { "epoch": 11.2, "learning_rate": 8.554123428681367e-06, "loss": 0.0521, "step": 36402 }, { "epoch": 11.2, "learning_rate": 8.553630396252129e-06, "loss": 0.0381, "step": 36403 }, { "epoch": 11.2, "learning_rate": 8.553137367413835e-06, "loss": 0.0499, "step": 36404 }, { "epoch": 11.2, "learning_rate": 8.552644342167713e-06, "loss": 0.025, "step": 36405 }, { "epoch": 11.2, "learning_rate": 8.55215132051498e-06, "loss": 0.0404, "step": 36406 }, { "epoch": 11.2, "learning_rate": 8.551658302456868e-06, "loss": 0.0304, "step": 36407 }, { "epoch": 11.2, "learning_rate": 8.551165287994598e-06, "loss": 0.033, "step": 36408 }, { "epoch": 11.2, "learning_rate": 8.550672277129393e-06, "loss": 0.0266, "step": 36409 }, { "epoch": 11.2, "learning_rate": 8.55017926986248e-06, "loss": 0.0387, "step": 36410 }, { "epoch": 11.2, "learning_rate": 8.549686266195078e-06, "loss": 0.04, "step": 36411 }, { "epoch": 11.2, "learning_rate": 8.549193266128416e-06, "loss": 0.0453, "step": 36412 }, { "epoch": 11.2, "learning_rate": 8.548700269663718e-06, "loss": 0.0191, "step": 36413 }, { "epoch": 11.2, "learning_rate": 8.548207276802203e-06, "loss": 0.0496, "step": 36414 }, { "epoch": 11.2, "learning_rate": 8.5477142875451e-06, "loss": 0.0666, "step": 36415 }, { "epoch": 11.2, "learning_rate": 8.547221301893628e-06, "loss": 0.0414, "step": 36416 }, { "epoch": 11.21, "learning_rate": 8.546728319849016e-06, "loss": 0.0391, "step": 36417 }, { "epoch": 11.21, "learning_rate": 8.546235341412486e-06, "loss": 0.0398, "step": 36418 }, { "epoch": 11.21, "learning_rate": 8.54574236658526e-06, "loss": 0.0692, "step": 36419 }, { "epoch": 11.21, "learning_rate": 8.545249395368566e-06, "loss": 0.0569, "step": 36420 }, { "epoch": 11.21, "learning_rate": 8.544756427763622e-06, "loss": 0.0597, "step": 36421 }, { "epoch": 11.21, "learning_rate": 8.544263463771658e-06, "loss": 0.07, "step": 36422 }, { "epoch": 11.21, "learning_rate": 8.543770503393895e-06, "loss": 0.0394, "step": 36423 }, { "epoch": 11.21, "learning_rate": 8.54327754663156e-06, "loss": 0.0475, "step": 36424 }, { "epoch": 11.21, "learning_rate": 8.542784593485871e-06, "loss": 0.0522, "step": 36425 }, { "epoch": 11.21, "learning_rate": 8.542291643958052e-06, "loss": 0.0848, "step": 36426 }, { "epoch": 11.21, "learning_rate": 8.541798698049335e-06, "loss": 0.0418, "step": 36427 }, { "epoch": 11.21, "learning_rate": 8.541305755760938e-06, "loss": 0.0633, "step": 36428 }, { "epoch": 11.21, "learning_rate": 8.540812817094084e-06, "loss": 0.0477, "step": 36429 }, { "epoch": 11.21, "learning_rate": 8.54031988205e-06, "loss": 0.0395, "step": 36430 }, { "epoch": 11.21, "learning_rate": 8.539826950629905e-06, "loss": 0.018, "step": 36431 }, { "epoch": 11.21, "learning_rate": 8.539334022835028e-06, "loss": 0.0518, "step": 36432 }, { "epoch": 11.21, "learning_rate": 8.538841098666591e-06, "loss": 0.0463, "step": 36433 }, { "epoch": 11.21, "learning_rate": 8.538348178125822e-06, "loss": 0.0555, "step": 36434 }, { "epoch": 11.21, "learning_rate": 8.537855261213936e-06, "loss": 0.0567, "step": 36435 }, { "epoch": 11.21, "learning_rate": 8.537362347932156e-06, "loss": 0.0461, "step": 36436 }, { "epoch": 11.21, "learning_rate": 8.536869438281718e-06, "loss": 0.07, "step": 36437 }, { "epoch": 11.21, "learning_rate": 8.536376532263838e-06, "loss": 0.0479, "step": 36438 }, { "epoch": 11.21, "learning_rate": 8.53588362987974e-06, "loss": 0.0374, "step": 36439 }, { "epoch": 11.21, "learning_rate": 8.535390731130648e-06, "loss": 0.0452, "step": 36440 }, { "epoch": 11.21, "learning_rate": 8.534897836017784e-06, "loss": 0.035, "step": 36441 }, { "epoch": 11.21, "learning_rate": 8.534404944542376e-06, "loss": 0.0303, "step": 36442 }, { "epoch": 11.21, "learning_rate": 8.533912056705645e-06, "loss": 0.0593, "step": 36443 }, { "epoch": 11.21, "learning_rate": 8.533419172508814e-06, "loss": 0.0508, "step": 36444 }, { "epoch": 11.21, "learning_rate": 8.532926291953113e-06, "loss": 0.0389, "step": 36445 }, { "epoch": 11.21, "learning_rate": 8.532433415039751e-06, "loss": 0.0518, "step": 36446 }, { "epoch": 11.21, "learning_rate": 8.531940541769967e-06, "loss": 0.0449, "step": 36447 }, { "epoch": 11.21, "learning_rate": 8.53144767214498e-06, "loss": 0.0445, "step": 36448 }, { "epoch": 11.22, "learning_rate": 8.530954806166012e-06, "loss": 0.1019, "step": 36449 }, { "epoch": 11.22, "learning_rate": 8.530461943834284e-06, "loss": 0.0543, "step": 36450 }, { "epoch": 11.22, "learning_rate": 8.529969085151024e-06, "loss": 0.0215, "step": 36451 }, { "epoch": 11.22, "learning_rate": 8.529476230117455e-06, "loss": 0.0471, "step": 36452 }, { "epoch": 11.22, "learning_rate": 8.528983378734802e-06, "loss": 0.0555, "step": 36453 }, { "epoch": 11.22, "learning_rate": 8.528490531004285e-06, "loss": 0.0349, "step": 36454 }, { "epoch": 11.22, "learning_rate": 8.527997686927127e-06, "loss": 0.0524, "step": 36455 }, { "epoch": 11.22, "learning_rate": 8.527504846504558e-06, "loss": 0.0563, "step": 36456 }, { "epoch": 11.22, "learning_rate": 8.527012009737795e-06, "loss": 0.034, "step": 36457 }, { "epoch": 11.22, "learning_rate": 8.526519176628068e-06, "loss": 0.0449, "step": 36458 }, { "epoch": 11.22, "learning_rate": 8.526026347176593e-06, "loss": 0.0466, "step": 36459 }, { "epoch": 11.22, "learning_rate": 8.525533521384594e-06, "loss": 0.0525, "step": 36460 }, { "epoch": 11.22, "learning_rate": 8.525040699253304e-06, "loss": 0.0304, "step": 36461 }, { "epoch": 11.22, "learning_rate": 8.524547880783937e-06, "loss": 0.0836, "step": 36462 }, { "epoch": 11.22, "learning_rate": 8.524055065977721e-06, "loss": 0.044, "step": 36463 }, { "epoch": 11.22, "learning_rate": 8.523562254835878e-06, "loss": 0.094, "step": 36464 }, { "epoch": 11.22, "learning_rate": 8.523069447359631e-06, "loss": 0.0498, "step": 36465 }, { "epoch": 11.22, "learning_rate": 8.522576643550205e-06, "loss": 0.0485, "step": 36466 }, { "epoch": 11.22, "learning_rate": 8.522083843408823e-06, "loss": 0.0463, "step": 36467 }, { "epoch": 11.22, "learning_rate": 8.52159104693671e-06, "loss": 0.0344, "step": 36468 }, { "epoch": 11.22, "learning_rate": 8.521098254135087e-06, "loss": 0.0397, "step": 36469 }, { "epoch": 11.22, "learning_rate": 8.520605465005174e-06, "loss": 0.0563, "step": 36470 }, { "epoch": 11.22, "learning_rate": 8.520112679548206e-06, "loss": 0.0363, "step": 36471 }, { "epoch": 11.22, "learning_rate": 8.519619897765395e-06, "loss": 0.0379, "step": 36472 }, { "epoch": 11.22, "learning_rate": 8.519127119657971e-06, "loss": 0.0479, "step": 36473 }, { "epoch": 11.22, "learning_rate": 8.518634345227153e-06, "loss": 0.0598, "step": 36474 }, { "epoch": 11.22, "learning_rate": 8.518141574474167e-06, "loss": 0.0424, "step": 36475 }, { "epoch": 11.22, "learning_rate": 8.517648807400234e-06, "loss": 0.0432, "step": 36476 }, { "epoch": 11.22, "learning_rate": 8.517156044006582e-06, "loss": 0.0704, "step": 36477 }, { "epoch": 11.22, "learning_rate": 8.516663284294432e-06, "loss": 0.0178, "step": 36478 }, { "epoch": 11.22, "learning_rate": 8.51617052826501e-06, "loss": 0.0617, "step": 36479 }, { "epoch": 11.22, "learning_rate": 8.515677775919528e-06, "loss": 0.0463, "step": 36480 }, { "epoch": 11.22, "learning_rate": 8.515185027259223e-06, "loss": 0.0488, "step": 36481 }, { "epoch": 11.23, "learning_rate": 8.514692282285315e-06, "loss": 0.049, "step": 36482 }, { "epoch": 11.23, "learning_rate": 8.514199540999025e-06, "loss": 0.0582, "step": 36483 }, { "epoch": 11.23, "learning_rate": 8.513706803401576e-06, "loss": 0.0599, "step": 36484 }, { "epoch": 11.23, "learning_rate": 8.51321406949419e-06, "loss": 0.052, "step": 36485 }, { "epoch": 11.23, "learning_rate": 8.512721339278095e-06, "loss": 0.0279, "step": 36486 }, { "epoch": 11.23, "learning_rate": 8.512228612754511e-06, "loss": 0.0264, "step": 36487 }, { "epoch": 11.23, "learning_rate": 8.511735889924664e-06, "loss": 0.0626, "step": 36488 }, { "epoch": 11.23, "learning_rate": 8.511243170789774e-06, "loss": 0.042, "step": 36489 }, { "epoch": 11.23, "learning_rate": 8.510750455351065e-06, "loss": 0.0686, "step": 36490 }, { "epoch": 11.23, "learning_rate": 8.510257743609763e-06, "loss": 0.05, "step": 36491 }, { "epoch": 11.23, "learning_rate": 8.509765035567092e-06, "loss": 0.0503, "step": 36492 }, { "epoch": 11.23, "learning_rate": 8.50927233122427e-06, "loss": 0.0672, "step": 36493 }, { "epoch": 11.23, "learning_rate": 8.508779630582522e-06, "loss": 0.0505, "step": 36494 }, { "epoch": 11.23, "learning_rate": 8.50828693364307e-06, "loss": 0.0566, "step": 36495 }, { "epoch": 11.23, "learning_rate": 8.507794240407143e-06, "loss": 0.0454, "step": 36496 }, { "epoch": 11.23, "learning_rate": 8.50730155087596e-06, "loss": 0.0307, "step": 36497 }, { "epoch": 11.23, "learning_rate": 8.506808865050746e-06, "loss": 0.0332, "step": 36498 }, { "epoch": 11.23, "learning_rate": 8.506316182932722e-06, "loss": 0.0412, "step": 36499 }, { "epoch": 11.23, "learning_rate": 8.505823504523109e-06, "loss": 0.0628, "step": 36500 }, { "epoch": 11.23, "learning_rate": 8.505330829823137e-06, "loss": 0.0499, "step": 36501 }, { "epoch": 11.23, "learning_rate": 8.504838158834023e-06, "loss": 0.0856, "step": 36502 }, { "epoch": 11.23, "learning_rate": 8.504345491556998e-06, "loss": 0.0626, "step": 36503 }, { "epoch": 11.23, "learning_rate": 8.503852827993277e-06, "loss": 0.0635, "step": 36504 }, { "epoch": 11.23, "learning_rate": 8.503360168144081e-06, "loss": 0.0529, "step": 36505 }, { "epoch": 11.23, "learning_rate": 8.502867512010645e-06, "loss": 0.0106, "step": 36506 }, { "epoch": 11.23, "learning_rate": 8.502374859594184e-06, "loss": 0.0795, "step": 36507 }, { "epoch": 11.23, "learning_rate": 8.501882210895921e-06, "loss": 0.0378, "step": 36508 }, { "epoch": 11.23, "learning_rate": 8.501389565917082e-06, "loss": 0.0375, "step": 36509 }, { "epoch": 11.23, "learning_rate": 8.500896924658886e-06, "loss": 0.0668, "step": 36510 }, { "epoch": 11.23, "learning_rate": 8.500404287122561e-06, "loss": 0.0466, "step": 36511 }, { "epoch": 11.23, "learning_rate": 8.499911653309327e-06, "loss": 0.1223, "step": 36512 }, { "epoch": 11.23, "learning_rate": 8.499419023220413e-06, "loss": 0.041, "step": 36513 }, { "epoch": 11.24, "learning_rate": 8.498926396857032e-06, "loss": 0.0561, "step": 36514 }, { "epoch": 11.24, "learning_rate": 8.498433774220408e-06, "loss": 0.0518, "step": 36515 }, { "epoch": 11.24, "learning_rate": 8.497941155311776e-06, "loss": 0.0519, "step": 36516 }, { "epoch": 11.24, "learning_rate": 8.497448540132347e-06, "loss": 0.0425, "step": 36517 }, { "epoch": 11.24, "learning_rate": 8.496955928683351e-06, "loss": 0.0383, "step": 36518 }, { "epoch": 11.24, "learning_rate": 8.496463320966004e-06, "loss": 0.0526, "step": 36519 }, { "epoch": 11.24, "learning_rate": 8.495970716981536e-06, "loss": 0.0485, "step": 36520 }, { "epoch": 11.24, "learning_rate": 8.495478116731166e-06, "loss": 0.0332, "step": 36521 }, { "epoch": 11.24, "learning_rate": 8.49498552021612e-06, "loss": 0.0806, "step": 36522 }, { "epoch": 11.24, "learning_rate": 8.494492927437619e-06, "loss": 0.0443, "step": 36523 }, { "epoch": 11.24, "learning_rate": 8.494000338396888e-06, "loss": 0.0416, "step": 36524 }, { "epoch": 11.24, "learning_rate": 8.49350775309514e-06, "loss": 0.0423, "step": 36525 }, { "epoch": 11.24, "learning_rate": 8.493015171533612e-06, "loss": 0.0283, "step": 36526 }, { "epoch": 11.24, "learning_rate": 8.492522593713523e-06, "loss": 0.0678, "step": 36527 }, { "epoch": 11.24, "learning_rate": 8.492030019636092e-06, "loss": 0.0309, "step": 36528 }, { "epoch": 11.24, "learning_rate": 8.491537449302545e-06, "loss": 0.0481, "step": 36529 }, { "epoch": 11.24, "learning_rate": 8.491044882714101e-06, "loss": 0.0409, "step": 36530 }, { "epoch": 11.24, "learning_rate": 8.490552319871986e-06, "loss": 0.052, "step": 36531 }, { "epoch": 11.24, "learning_rate": 8.490059760777425e-06, "loss": 0.058, "step": 36532 }, { "epoch": 11.24, "learning_rate": 8.489567205431638e-06, "loss": 0.0473, "step": 36533 }, { "epoch": 11.24, "learning_rate": 8.48907465383585e-06, "loss": 0.0422, "step": 36534 }, { "epoch": 11.24, "learning_rate": 8.488582105991277e-06, "loss": 0.027, "step": 36535 }, { "epoch": 11.24, "learning_rate": 8.48808956189915e-06, "loss": 0.0498, "step": 36536 }, { "epoch": 11.24, "learning_rate": 8.487597021560692e-06, "loss": 0.0715, "step": 36537 }, { "epoch": 11.24, "learning_rate": 8.48710448497712e-06, "loss": 0.0414, "step": 36538 }, { "epoch": 11.24, "learning_rate": 8.486611952149661e-06, "loss": 0.0712, "step": 36539 }, { "epoch": 11.24, "learning_rate": 8.486119423079534e-06, "loss": 0.0536, "step": 36540 }, { "epoch": 11.24, "learning_rate": 8.485626897767964e-06, "loss": 0.0317, "step": 36541 }, { "epoch": 11.24, "learning_rate": 8.485134376216178e-06, "loss": 0.0452, "step": 36542 }, { "epoch": 11.24, "learning_rate": 8.484641858425394e-06, "loss": 0.0339, "step": 36543 }, { "epoch": 11.24, "learning_rate": 8.484149344396833e-06, "loss": 0.0342, "step": 36544 }, { "epoch": 11.24, "learning_rate": 8.48365683413172e-06, "loss": 0.0244, "step": 36545 }, { "epoch": 11.24, "learning_rate": 8.483164327631282e-06, "loss": 0.0675, "step": 36546 }, { "epoch": 11.25, "learning_rate": 8.482671824896737e-06, "loss": 0.0895, "step": 36547 }, { "epoch": 11.25, "learning_rate": 8.482179325929309e-06, "loss": 0.0481, "step": 36548 }, { "epoch": 11.25, "learning_rate": 8.481686830730222e-06, "loss": 0.0371, "step": 36549 }, { "epoch": 11.25, "learning_rate": 8.481194339300692e-06, "loss": 0.0309, "step": 36550 }, { "epoch": 11.25, "learning_rate": 8.480701851641951e-06, "loss": 0.0504, "step": 36551 }, { "epoch": 11.25, "learning_rate": 8.480209367755217e-06, "loss": 0.1048, "step": 36552 }, { "epoch": 11.25, "learning_rate": 8.479716887641714e-06, "loss": 0.0517, "step": 36553 }, { "epoch": 11.25, "learning_rate": 8.479224411302664e-06, "loss": 0.0315, "step": 36554 }, { "epoch": 11.25, "learning_rate": 8.478731938739286e-06, "loss": 0.027, "step": 36555 }, { "epoch": 11.25, "learning_rate": 8.47823946995281e-06, "loss": 0.0318, "step": 36556 }, { "epoch": 11.25, "learning_rate": 8.477747004944455e-06, "loss": 0.0511, "step": 36557 }, { "epoch": 11.25, "learning_rate": 8.477254543715446e-06, "loss": 0.0597, "step": 36558 }, { "epoch": 11.25, "learning_rate": 8.476762086267002e-06, "loss": 0.0585, "step": 36559 }, { "epoch": 11.25, "learning_rate": 8.476269632600342e-06, "loss": 0.0431, "step": 36560 }, { "epoch": 11.25, "learning_rate": 8.4757771827167e-06, "loss": 0.0635, "step": 36561 }, { "epoch": 11.25, "learning_rate": 8.475284736617291e-06, "loss": 0.0411, "step": 36562 }, { "epoch": 11.25, "learning_rate": 8.474792294303338e-06, "loss": 0.0566, "step": 36563 }, { "epoch": 11.25, "learning_rate": 8.474299855776065e-06, "loss": 0.0535, "step": 36564 }, { "epoch": 11.25, "learning_rate": 8.473807421036692e-06, "loss": 0.0388, "step": 36565 }, { "epoch": 11.25, "learning_rate": 8.473314990086445e-06, "loss": 0.0579, "step": 36566 }, { "epoch": 11.25, "learning_rate": 8.472822562926546e-06, "loss": 0.0558, "step": 36567 }, { "epoch": 11.25, "learning_rate": 8.47233013955822e-06, "loss": 0.0425, "step": 36568 }, { "epoch": 11.25, "learning_rate": 8.471837719982683e-06, "loss": 0.0455, "step": 36569 }, { "epoch": 11.25, "learning_rate": 8.471345304201157e-06, "loss": 0.0494, "step": 36570 }, { "epoch": 11.25, "learning_rate": 8.470852892214875e-06, "loss": 0.0329, "step": 36571 }, { "epoch": 11.25, "learning_rate": 8.470360484025051e-06, "loss": 0.0322, "step": 36572 }, { "epoch": 11.25, "learning_rate": 8.469868079632909e-06, "loss": 0.1064, "step": 36573 }, { "epoch": 11.25, "learning_rate": 8.469375679039673e-06, "loss": 0.0504, "step": 36574 }, { "epoch": 11.25, "learning_rate": 8.468883282246562e-06, "loss": 0.0316, "step": 36575 }, { "epoch": 11.25, "learning_rate": 8.468390889254803e-06, "loss": 0.0565, "step": 36576 }, { "epoch": 11.25, "learning_rate": 8.467898500065616e-06, "loss": 0.0791, "step": 36577 }, { "epoch": 11.25, "learning_rate": 8.467406114680224e-06, "loss": 0.0661, "step": 36578 }, { "epoch": 11.26, "learning_rate": 8.466913733099852e-06, "loss": 0.0763, "step": 36579 }, { "epoch": 11.26, "learning_rate": 8.466421355325715e-06, "loss": 0.0748, "step": 36580 }, { "epoch": 11.26, "learning_rate": 8.465928981359043e-06, "loss": 0.0468, "step": 36581 }, { "epoch": 11.26, "learning_rate": 8.465436611201057e-06, "loss": 0.043, "step": 36582 }, { "epoch": 11.26, "learning_rate": 8.464944244852977e-06, "loss": 0.0378, "step": 36583 }, { "epoch": 11.26, "learning_rate": 8.464451882316028e-06, "loss": 0.0728, "step": 36584 }, { "epoch": 11.26, "learning_rate": 8.463959523591428e-06, "loss": 0.0435, "step": 36585 }, { "epoch": 11.26, "learning_rate": 8.463467168680404e-06, "loss": 0.0534, "step": 36586 }, { "epoch": 11.26, "learning_rate": 8.462974817584177e-06, "loss": 0.051, "step": 36587 }, { "epoch": 11.26, "learning_rate": 8.462482470303968e-06, "loss": 0.059, "step": 36588 }, { "epoch": 11.26, "learning_rate": 8.461990126841004e-06, "loss": 0.044, "step": 36589 }, { "epoch": 11.26, "learning_rate": 8.461497787196497e-06, "loss": 0.0528, "step": 36590 }, { "epoch": 11.26, "learning_rate": 8.461005451371682e-06, "loss": 0.0385, "step": 36591 }, { "epoch": 11.26, "learning_rate": 8.460513119367776e-06, "loss": 0.0544, "step": 36592 }, { "epoch": 11.26, "learning_rate": 8.460020791185998e-06, "loss": 0.0301, "step": 36593 }, { "epoch": 11.26, "learning_rate": 8.459528466827576e-06, "loss": 0.0374, "step": 36594 }, { "epoch": 11.26, "learning_rate": 8.459036146293725e-06, "loss": 0.0264, "step": 36595 }, { "epoch": 11.26, "learning_rate": 8.458543829585674e-06, "loss": 0.0279, "step": 36596 }, { "epoch": 11.26, "learning_rate": 8.458051516704644e-06, "loss": 0.0118, "step": 36597 }, { "epoch": 11.26, "learning_rate": 8.457559207651856e-06, "loss": 0.0237, "step": 36598 }, { "epoch": 11.26, "learning_rate": 8.457066902428534e-06, "loss": 0.0517, "step": 36599 }, { "epoch": 11.26, "learning_rate": 8.456574601035896e-06, "loss": 0.034, "step": 36600 }, { "epoch": 11.26, "learning_rate": 8.456082303475169e-06, "loss": 0.0442, "step": 36601 }, { "epoch": 11.26, "learning_rate": 8.455590009747572e-06, "loss": 0.0564, "step": 36602 }, { "epoch": 11.26, "learning_rate": 8.455097719854333e-06, "loss": 0.0308, "step": 36603 }, { "epoch": 11.26, "learning_rate": 8.454605433796666e-06, "loss": 0.0874, "step": 36604 }, { "epoch": 11.26, "learning_rate": 8.454113151575795e-06, "loss": 0.0505, "step": 36605 }, { "epoch": 11.26, "learning_rate": 8.45362087319295e-06, "loss": 0.0548, "step": 36606 }, { "epoch": 11.26, "learning_rate": 8.453128598649343e-06, "loss": 0.0592, "step": 36607 }, { "epoch": 11.26, "learning_rate": 8.452636327946203e-06, "loss": 0.0525, "step": 36608 }, { "epoch": 11.26, "learning_rate": 8.452144061084749e-06, "loss": 0.0473, "step": 36609 }, { "epoch": 11.26, "learning_rate": 8.451651798066203e-06, "loss": 0.0371, "step": 36610 }, { "epoch": 11.26, "learning_rate": 8.45115953889179e-06, "loss": 0.0352, "step": 36611 }, { "epoch": 11.27, "learning_rate": 8.450667283562729e-06, "loss": 0.0843, "step": 36612 }, { "epoch": 11.27, "learning_rate": 8.450175032080245e-06, "loss": 0.0412, "step": 36613 }, { "epoch": 11.27, "learning_rate": 8.449682784445558e-06, "loss": 0.0626, "step": 36614 }, { "epoch": 11.27, "learning_rate": 8.449190540659888e-06, "loss": 0.1026, "step": 36615 }, { "epoch": 11.27, "learning_rate": 8.448698300724464e-06, "loss": 0.0565, "step": 36616 }, { "epoch": 11.27, "learning_rate": 8.448206064640502e-06, "loss": 0.0554, "step": 36617 }, { "epoch": 11.27, "learning_rate": 8.447713832409226e-06, "loss": 0.0781, "step": 36618 }, { "epoch": 11.27, "learning_rate": 8.447221604031858e-06, "loss": 0.0828, "step": 36619 }, { "epoch": 11.27, "learning_rate": 8.446729379509618e-06, "loss": 0.0596, "step": 36620 }, { "epoch": 11.27, "learning_rate": 8.446237158843733e-06, "loss": 0.0511, "step": 36621 }, { "epoch": 11.27, "learning_rate": 8.44574494203542e-06, "loss": 0.0273, "step": 36622 }, { "epoch": 11.27, "learning_rate": 8.445252729085907e-06, "loss": 0.0428, "step": 36623 }, { "epoch": 11.27, "learning_rate": 8.444760519996413e-06, "loss": 0.0706, "step": 36624 }, { "epoch": 11.27, "learning_rate": 8.444268314768152e-06, "loss": 0.0764, "step": 36625 }, { "epoch": 11.27, "learning_rate": 8.443776113402358e-06, "loss": 0.0761, "step": 36626 }, { "epoch": 11.27, "learning_rate": 8.44328391590025e-06, "loss": 0.0434, "step": 36627 }, { "epoch": 11.27, "learning_rate": 8.442791722263048e-06, "loss": 0.0714, "step": 36628 }, { "epoch": 11.27, "learning_rate": 8.442299532491973e-06, "loss": 0.0299, "step": 36629 }, { "epoch": 11.27, "learning_rate": 8.441807346588247e-06, "loss": 0.0573, "step": 36630 }, { "epoch": 11.27, "learning_rate": 8.441315164553095e-06, "loss": 0.0489, "step": 36631 }, { "epoch": 11.27, "learning_rate": 8.440822986387735e-06, "loss": 0.0295, "step": 36632 }, { "epoch": 11.27, "learning_rate": 8.440330812093395e-06, "loss": 0.0498, "step": 36633 }, { "epoch": 11.27, "learning_rate": 8.439838641671292e-06, "loss": 0.0684, "step": 36634 }, { "epoch": 11.27, "learning_rate": 8.439346475122645e-06, "loss": 0.0549, "step": 36635 }, { "epoch": 11.27, "learning_rate": 8.438854312448683e-06, "loss": 0.1004, "step": 36636 }, { "epoch": 11.27, "learning_rate": 8.438362153650628e-06, "loss": 0.041, "step": 36637 }, { "epoch": 11.27, "learning_rate": 8.437869998729695e-06, "loss": 0.0332, "step": 36638 }, { "epoch": 11.27, "learning_rate": 8.437377847687111e-06, "loss": 0.0507, "step": 36639 }, { "epoch": 11.27, "learning_rate": 8.436885700524094e-06, "loss": 0.0673, "step": 36640 }, { "epoch": 11.27, "learning_rate": 8.436393557241872e-06, "loss": 0.0416, "step": 36641 }, { "epoch": 11.27, "learning_rate": 8.43590141784166e-06, "loss": 0.0427, "step": 36642 }, { "epoch": 11.27, "learning_rate": 8.435409282324686e-06, "loss": 0.0479, "step": 36643 }, { "epoch": 11.28, "learning_rate": 8.434917150692168e-06, "loss": 0.0783, "step": 36644 }, { "epoch": 11.28, "learning_rate": 8.434425022945326e-06, "loss": 0.0715, "step": 36645 }, { "epoch": 11.28, "learning_rate": 8.433932899085387e-06, "loss": 0.0682, "step": 36646 }, { "epoch": 11.28, "learning_rate": 8.433440779113574e-06, "loss": 0.0603, "step": 36647 }, { "epoch": 11.28, "learning_rate": 8.432948663031102e-06, "loss": 0.0637, "step": 36648 }, { "epoch": 11.28, "learning_rate": 8.432456550839196e-06, "loss": 0.0463, "step": 36649 }, { "epoch": 11.28, "learning_rate": 8.431964442539075e-06, "loss": 0.0441, "step": 36650 }, { "epoch": 11.28, "learning_rate": 8.431472338131967e-06, "loss": 0.0538, "step": 36651 }, { "epoch": 11.28, "learning_rate": 8.43098023761909e-06, "loss": 0.0419, "step": 36652 }, { "epoch": 11.28, "learning_rate": 8.430488141001665e-06, "loss": 0.0238, "step": 36653 }, { "epoch": 11.28, "learning_rate": 8.429996048280915e-06, "loss": 0.0469, "step": 36654 }, { "epoch": 11.28, "learning_rate": 8.429503959458061e-06, "loss": 0.0409, "step": 36655 }, { "epoch": 11.28, "learning_rate": 8.429011874534325e-06, "loss": 0.0368, "step": 36656 }, { "epoch": 11.28, "learning_rate": 8.429011874534325e-06, "loss": 0.0563, "step": 36657 }, { "epoch": 11.28, "learning_rate": 8.42851979351093e-06, "loss": 0.066, "step": 36658 }, { "epoch": 11.28, "learning_rate": 8.4280277163891e-06, "loss": 0.0531, "step": 36659 }, { "epoch": 11.28, "learning_rate": 8.427535643170051e-06, "loss": 0.047, "step": 36660 }, { "epoch": 11.28, "learning_rate": 8.427043573855001e-06, "loss": 0.0575, "step": 36661 }, { "epoch": 11.28, "learning_rate": 8.426551508445187e-06, "loss": 0.0298, "step": 36662 }, { "epoch": 11.28, "learning_rate": 8.426059446941817e-06, "loss": 0.0154, "step": 36663 }, { "epoch": 11.28, "learning_rate": 8.425567389346118e-06, "loss": 0.0353, "step": 36664 }, { "epoch": 11.28, "learning_rate": 8.42507533565931e-06, "loss": 0.0386, "step": 36665 }, { "epoch": 11.28, "learning_rate": 8.424583285882614e-06, "loss": 0.0535, "step": 36666 }, { "epoch": 11.28, "learning_rate": 8.424091240017253e-06, "loss": 0.0567, "step": 36667 }, { "epoch": 11.28, "learning_rate": 8.423599198064452e-06, "loss": 0.0582, "step": 36668 }, { "epoch": 11.28, "learning_rate": 8.42310716002543e-06, "loss": 0.0543, "step": 36669 }, { "epoch": 11.28, "learning_rate": 8.422615125901404e-06, "loss": 0.0606, "step": 36670 }, { "epoch": 11.28, "learning_rate": 8.422123095693598e-06, "loss": 0.0685, "step": 36671 }, { "epoch": 11.28, "learning_rate": 8.42163106940324e-06, "loss": 0.0609, "step": 36672 }, { "epoch": 11.28, "learning_rate": 8.421139047031543e-06, "loss": 0.0469, "step": 36673 }, { "epoch": 11.28, "learning_rate": 8.420647028579734e-06, "loss": 0.1194, "step": 36674 }, { "epoch": 11.28, "learning_rate": 8.420155014049032e-06, "loss": 0.1133, "step": 36675 }, { "epoch": 11.28, "learning_rate": 8.419663003440657e-06, "loss": 0.0334, "step": 36676 }, { "epoch": 11.29, "learning_rate": 8.419170996755835e-06, "loss": 0.0879, "step": 36677 }, { "epoch": 11.29, "learning_rate": 8.418678993995785e-06, "loss": 0.0719, "step": 36678 }, { "epoch": 11.29, "learning_rate": 8.418186995161729e-06, "loss": 0.0472, "step": 36679 }, { "epoch": 11.29, "learning_rate": 8.417695000254891e-06, "loss": 0.0284, "step": 36680 }, { "epoch": 11.29, "learning_rate": 8.417203009276482e-06, "loss": 0.0746, "step": 36681 }, { "epoch": 11.29, "learning_rate": 8.416711022227736e-06, "loss": 0.1309, "step": 36682 }, { "epoch": 11.29, "learning_rate": 8.416219039109872e-06, "loss": 0.0511, "step": 36683 }, { "epoch": 11.29, "learning_rate": 8.415727059924107e-06, "loss": 0.0456, "step": 36684 }, { "epoch": 11.29, "learning_rate": 8.415235084671664e-06, "loss": 0.0492, "step": 36685 }, { "epoch": 11.29, "learning_rate": 8.414743113353764e-06, "loss": 0.0383, "step": 36686 }, { "epoch": 11.29, "learning_rate": 8.414251145971632e-06, "loss": 0.041, "step": 36687 }, { "epoch": 11.29, "learning_rate": 8.413759182526485e-06, "loss": 0.0787, "step": 36688 }, { "epoch": 11.29, "learning_rate": 8.413267223019548e-06, "loss": 0.055, "step": 36689 }, { "epoch": 11.29, "learning_rate": 8.412775267452042e-06, "loss": 0.0737, "step": 36690 }, { "epoch": 11.29, "learning_rate": 8.41228331582518e-06, "loss": 0.0578, "step": 36691 }, { "epoch": 11.29, "learning_rate": 8.411791368140197e-06, "loss": 0.0561, "step": 36692 }, { "epoch": 11.29, "learning_rate": 8.411299424398309e-06, "loss": 0.0313, "step": 36693 }, { "epoch": 11.29, "learning_rate": 8.410807484600732e-06, "loss": 0.0373, "step": 36694 }, { "epoch": 11.29, "learning_rate": 8.410315548748695e-06, "loss": 0.0503, "step": 36695 }, { "epoch": 11.29, "learning_rate": 8.40982361684341e-06, "loss": 0.0456, "step": 36696 }, { "epoch": 11.29, "learning_rate": 8.40933168888611e-06, "loss": 0.0438, "step": 36697 }, { "epoch": 11.29, "learning_rate": 8.408839764878009e-06, "loss": 0.057, "step": 36698 }, { "epoch": 11.29, "learning_rate": 8.408347844820328e-06, "loss": 0.0395, "step": 36699 }, { "epoch": 11.29, "learning_rate": 8.407855928714294e-06, "loss": 0.0365, "step": 36700 }, { "epoch": 11.29, "learning_rate": 8.40736401656112e-06, "loss": 0.042, "step": 36701 }, { "epoch": 11.29, "learning_rate": 8.406872108362034e-06, "loss": 0.0484, "step": 36702 }, { "epoch": 11.29, "learning_rate": 8.406380204118257e-06, "loss": 0.0501, "step": 36703 }, { "epoch": 11.29, "learning_rate": 8.405888303831009e-06, "loss": 0.0351, "step": 36704 }, { "epoch": 11.29, "learning_rate": 8.405396407501506e-06, "loss": 0.0221, "step": 36705 }, { "epoch": 11.29, "learning_rate": 8.404904515130977e-06, "loss": 0.0495, "step": 36706 }, { "epoch": 11.29, "learning_rate": 8.404412626720638e-06, "loss": 0.0403, "step": 36707 }, { "epoch": 11.29, "learning_rate": 8.403920742271713e-06, "loss": 0.0712, "step": 36708 }, { "epoch": 11.3, "learning_rate": 8.403428861785423e-06, "loss": 0.032, "step": 36709 }, { "epoch": 11.3, "learning_rate": 8.402936985262989e-06, "loss": 0.0395, "step": 36710 }, { "epoch": 11.3, "learning_rate": 8.40244511270563e-06, "loss": 0.0243, "step": 36711 }, { "epoch": 11.3, "learning_rate": 8.401953244114572e-06, "loss": 0.0622, "step": 36712 }, { "epoch": 11.3, "learning_rate": 8.401461379491032e-06, "loss": 0.0563, "step": 36713 }, { "epoch": 11.3, "learning_rate": 8.400969518836234e-06, "loss": 0.0582, "step": 36714 }, { "epoch": 11.3, "learning_rate": 8.400477662151397e-06, "loss": 0.0848, "step": 36715 }, { "epoch": 11.3, "learning_rate": 8.399985809437737e-06, "loss": 0.0438, "step": 36716 }, { "epoch": 11.3, "learning_rate": 8.399493960696487e-06, "loss": 0.0249, "step": 36717 }, { "epoch": 11.3, "learning_rate": 8.399002115928862e-06, "loss": 0.0489, "step": 36718 }, { "epoch": 11.3, "learning_rate": 8.398510275136081e-06, "loss": 0.082, "step": 36719 }, { "epoch": 11.3, "learning_rate": 8.398018438319368e-06, "loss": 0.0363, "step": 36720 }, { "epoch": 11.3, "learning_rate": 8.397526605479942e-06, "loss": 0.0452, "step": 36721 }, { "epoch": 11.3, "learning_rate": 8.397034776619028e-06, "loss": 0.0505, "step": 36722 }, { "epoch": 11.3, "learning_rate": 8.396542951737842e-06, "loss": 0.0409, "step": 36723 }, { "epoch": 11.3, "learning_rate": 8.396051130837608e-06, "loss": 0.0366, "step": 36724 }, { "epoch": 11.3, "learning_rate": 8.39555931391955e-06, "loss": 0.0437, "step": 36725 }, { "epoch": 11.3, "learning_rate": 8.39506750098488e-06, "loss": 0.0583, "step": 36726 }, { "epoch": 11.3, "learning_rate": 8.394575692034831e-06, "loss": 0.0981, "step": 36727 }, { "epoch": 11.3, "learning_rate": 8.394083887070614e-06, "loss": 0.0575, "step": 36728 }, { "epoch": 11.3, "learning_rate": 8.393592086093456e-06, "loss": 0.0541, "step": 36729 }, { "epoch": 11.3, "learning_rate": 8.393100289104574e-06, "loss": 0.0766, "step": 36730 }, { "epoch": 11.3, "learning_rate": 8.392608496105188e-06, "loss": 0.0855, "step": 36731 }, { "epoch": 11.3, "learning_rate": 8.392116707096526e-06, "loss": 0.0295, "step": 36732 }, { "epoch": 11.3, "learning_rate": 8.391624922079802e-06, "loss": 0.0368, "step": 36733 }, { "epoch": 11.3, "learning_rate": 8.391133141056243e-06, "loss": 0.0328, "step": 36734 }, { "epoch": 11.3, "learning_rate": 8.390641364027066e-06, "loss": 0.072, "step": 36735 }, { "epoch": 11.3, "learning_rate": 8.390149590993488e-06, "loss": 0.0619, "step": 36736 }, { "epoch": 11.3, "learning_rate": 8.389657821956738e-06, "loss": 0.0413, "step": 36737 }, { "epoch": 11.3, "learning_rate": 8.389166056918035e-06, "loss": 0.0538, "step": 36738 }, { "epoch": 11.3, "learning_rate": 8.388674295878599e-06, "loss": 0.0697, "step": 36739 }, { "epoch": 11.3, "learning_rate": 8.388182538839648e-06, "loss": 0.059, "step": 36740 }, { "epoch": 11.3, "learning_rate": 8.387690785802403e-06, "loss": 0.0372, "step": 36741 }, { "epoch": 11.31, "learning_rate": 8.38719903676809e-06, "loss": 0.054, "step": 36742 }, { "epoch": 11.31, "learning_rate": 8.386707291737925e-06, "loss": 0.0544, "step": 36743 }, { "epoch": 11.31, "learning_rate": 8.386215550713133e-06, "loss": 0.0503, "step": 36744 }, { "epoch": 11.31, "learning_rate": 8.385723813694933e-06, "loss": 0.04, "step": 36745 }, { "epoch": 11.31, "learning_rate": 8.385232080684543e-06, "loss": 0.0436, "step": 36746 }, { "epoch": 11.31, "learning_rate": 8.384740351683188e-06, "loss": 0.0577, "step": 36747 }, { "epoch": 11.31, "learning_rate": 8.38424862669209e-06, "loss": 0.0517, "step": 36748 }, { "epoch": 11.31, "learning_rate": 8.383756905712465e-06, "loss": 0.0519, "step": 36749 }, { "epoch": 11.31, "learning_rate": 8.383265188745535e-06, "loss": 0.0403, "step": 36750 }, { "epoch": 11.31, "learning_rate": 8.382773475792521e-06, "loss": 0.0858, "step": 36751 }, { "epoch": 11.31, "learning_rate": 8.382281766854646e-06, "loss": 0.0317, "step": 36752 }, { "epoch": 11.31, "learning_rate": 8.381790061933129e-06, "loss": 0.0635, "step": 36753 }, { "epoch": 11.31, "learning_rate": 8.38129836102919e-06, "loss": 0.0805, "step": 36754 }, { "epoch": 11.31, "learning_rate": 8.380806664144052e-06, "loss": 0.0488, "step": 36755 }, { "epoch": 11.31, "learning_rate": 8.380314971278932e-06, "loss": 0.0515, "step": 36756 }, { "epoch": 11.31, "learning_rate": 8.379823282435056e-06, "loss": 0.0424, "step": 36757 }, { "epoch": 11.31, "learning_rate": 8.37933159761364e-06, "loss": 0.0435, "step": 36758 }, { "epoch": 11.31, "learning_rate": 8.378839916815911e-06, "loss": 0.0749, "step": 36759 }, { "epoch": 11.31, "learning_rate": 8.378348240043083e-06, "loss": 0.0671, "step": 36760 }, { "epoch": 11.31, "learning_rate": 8.377856567296374e-06, "loss": 0.0619, "step": 36761 }, { "epoch": 11.31, "learning_rate": 8.377364898577017e-06, "loss": 0.0518, "step": 36762 }, { "epoch": 11.31, "learning_rate": 8.376873233886224e-06, "loss": 0.0701, "step": 36763 }, { "epoch": 11.31, "learning_rate": 8.376381573225217e-06, "loss": 0.0256, "step": 36764 }, { "epoch": 11.31, "learning_rate": 8.375889916595215e-06, "loss": 0.0238, "step": 36765 }, { "epoch": 11.31, "learning_rate": 8.37539826399744e-06, "loss": 0.0455, "step": 36766 }, { "epoch": 11.31, "learning_rate": 8.374906615433114e-06, "loss": 0.0649, "step": 36767 }, { "epoch": 11.31, "learning_rate": 8.374414970903458e-06, "loss": 0.0475, "step": 36768 }, { "epoch": 11.31, "learning_rate": 8.373923330409693e-06, "loss": 0.0643, "step": 36769 }, { "epoch": 11.31, "learning_rate": 8.373431693953036e-06, "loss": 0.0478, "step": 36770 }, { "epoch": 11.31, "learning_rate": 8.372940061534704e-06, "loss": 0.0252, "step": 36771 }, { "epoch": 11.31, "learning_rate": 8.372448433155931e-06, "loss": 0.0654, "step": 36772 }, { "epoch": 11.31, "learning_rate": 8.371956808817928e-06, "loss": 0.0375, "step": 36773 }, { "epoch": 11.32, "learning_rate": 8.371465188521916e-06, "loss": 0.063, "step": 36774 }, { "epoch": 11.32, "learning_rate": 8.37097357226912e-06, "loss": 0.048, "step": 36775 }, { "epoch": 11.32, "learning_rate": 8.370481960060753e-06, "loss": 0.0443, "step": 36776 }, { "epoch": 11.32, "learning_rate": 8.369990351898041e-06, "loss": 0.0332, "step": 36777 }, { "epoch": 11.32, "learning_rate": 8.369498747782205e-06, "loss": 0.0734, "step": 36778 }, { "epoch": 11.32, "learning_rate": 8.369007147714463e-06, "loss": 0.077, "step": 36779 }, { "epoch": 11.32, "learning_rate": 8.36851555169604e-06, "loss": 0.0639, "step": 36780 }, { "epoch": 11.32, "learning_rate": 8.368023959728146e-06, "loss": 0.0638, "step": 36781 }, { "epoch": 11.32, "learning_rate": 8.367532371812012e-06, "loss": 0.0661, "step": 36782 }, { "epoch": 11.32, "learning_rate": 8.36704078794886e-06, "loss": 0.0438, "step": 36783 }, { "epoch": 11.32, "learning_rate": 8.3665492081399e-06, "loss": 0.0422, "step": 36784 }, { "epoch": 11.32, "learning_rate": 8.36605763238636e-06, "loss": 0.03, "step": 36785 }, { "epoch": 11.32, "learning_rate": 8.365566060689455e-06, "loss": 0.0894, "step": 36786 }, { "epoch": 11.32, "learning_rate": 8.36507449305041e-06, "loss": 0.0654, "step": 36787 }, { "epoch": 11.32, "learning_rate": 8.364582929470448e-06, "loss": 0.0342, "step": 36788 }, { "epoch": 11.32, "learning_rate": 8.364091369950783e-06, "loss": 0.0449, "step": 36789 }, { "epoch": 11.32, "learning_rate": 8.363599814492641e-06, "loss": 0.0547, "step": 36790 }, { "epoch": 11.32, "learning_rate": 8.363108263097232e-06, "loss": 0.0339, "step": 36791 }, { "epoch": 11.32, "learning_rate": 8.36261671576579e-06, "loss": 0.0576, "step": 36792 }, { "epoch": 11.32, "learning_rate": 8.36212517249953e-06, "loss": 0.0155, "step": 36793 }, { "epoch": 11.32, "learning_rate": 8.36163363329967e-06, "loss": 0.0827, "step": 36794 }, { "epoch": 11.32, "learning_rate": 8.361142098167432e-06, "loss": 0.0725, "step": 36795 }, { "epoch": 11.32, "learning_rate": 8.360650567104034e-06, "loss": 0.0252, "step": 36796 }, { "epoch": 11.32, "learning_rate": 8.360159040110701e-06, "loss": 0.0356, "step": 36797 }, { "epoch": 11.32, "learning_rate": 8.359667517188652e-06, "loss": 0.0521, "step": 36798 }, { "epoch": 11.32, "learning_rate": 8.359175998339104e-06, "loss": 0.0239, "step": 36799 }, { "epoch": 11.32, "learning_rate": 8.358684483563283e-06, "loss": 0.0828, "step": 36800 }, { "epoch": 11.32, "learning_rate": 8.358192972862401e-06, "loss": 0.0597, "step": 36801 }, { "epoch": 11.32, "learning_rate": 8.357701466237686e-06, "loss": 0.0245, "step": 36802 }, { "epoch": 11.32, "learning_rate": 8.357209963690355e-06, "loss": 0.0322, "step": 36803 }, { "epoch": 11.32, "learning_rate": 8.356718465221632e-06, "loss": 0.0787, "step": 36804 }, { "epoch": 11.32, "learning_rate": 8.35622697083273e-06, "loss": 0.064, "step": 36805 }, { "epoch": 11.32, "learning_rate": 8.355735480524874e-06, "loss": 0.0383, "step": 36806 }, { "epoch": 11.33, "learning_rate": 8.355243994299285e-06, "loss": 0.0303, "step": 36807 }, { "epoch": 11.33, "learning_rate": 8.354752512157181e-06, "loss": 0.0562, "step": 36808 }, { "epoch": 11.33, "learning_rate": 8.354261034099782e-06, "loss": 0.0448, "step": 36809 }, { "epoch": 11.33, "learning_rate": 8.35376956012831e-06, "loss": 0.0485, "step": 36810 }, { "epoch": 11.33, "learning_rate": 8.353278090243983e-06, "loss": 0.0463, "step": 36811 }, { "epoch": 11.33, "learning_rate": 8.352786624448024e-06, "loss": 0.0596, "step": 36812 }, { "epoch": 11.33, "learning_rate": 8.35229516274165e-06, "loss": 0.0331, "step": 36813 }, { "epoch": 11.33, "learning_rate": 8.351803705126087e-06, "loss": 0.0423, "step": 36814 }, { "epoch": 11.33, "learning_rate": 8.35131225160255e-06, "loss": 0.0463, "step": 36815 }, { "epoch": 11.33, "learning_rate": 8.350820802172256e-06, "loss": 0.0239, "step": 36816 }, { "epoch": 11.33, "learning_rate": 8.350329356836434e-06, "loss": 0.0365, "step": 36817 }, { "epoch": 11.33, "learning_rate": 8.349837915596297e-06, "loss": 0.0311, "step": 36818 }, { "epoch": 11.33, "learning_rate": 8.349346478453068e-06, "loss": 0.0327, "step": 36819 }, { "epoch": 11.33, "learning_rate": 8.348855045407967e-06, "loss": 0.0478, "step": 36820 }, { "epoch": 11.33, "learning_rate": 8.348363616462212e-06, "loss": 0.0793, "step": 36821 }, { "epoch": 11.33, "learning_rate": 8.347872191617027e-06, "loss": 0.0351, "step": 36822 }, { "epoch": 11.33, "learning_rate": 8.34738077087363e-06, "loss": 0.049, "step": 36823 }, { "epoch": 11.33, "learning_rate": 8.34688935423324e-06, "loss": 0.0298, "step": 36824 }, { "epoch": 11.33, "learning_rate": 8.346397941697082e-06, "loss": 0.035, "step": 36825 }, { "epoch": 11.33, "learning_rate": 8.345906533266364e-06, "loss": 0.0419, "step": 36826 }, { "epoch": 11.33, "learning_rate": 8.345415128942322e-06, "loss": 0.0537, "step": 36827 }, { "epoch": 11.33, "learning_rate": 8.344923728726165e-06, "loss": 0.0406, "step": 36828 }, { "epoch": 11.33, "learning_rate": 8.344432332619115e-06, "loss": 0.039, "step": 36829 }, { "epoch": 11.33, "learning_rate": 8.343940940622394e-06, "loss": 0.0407, "step": 36830 }, { "epoch": 11.33, "learning_rate": 8.34344955273722e-06, "loss": 0.0616, "step": 36831 }, { "epoch": 11.33, "learning_rate": 8.342958168964816e-06, "loss": 0.0202, "step": 36832 }, { "epoch": 11.33, "learning_rate": 8.3424667893064e-06, "loss": 0.0727, "step": 36833 }, { "epoch": 11.33, "learning_rate": 8.341975413763193e-06, "loss": 0.0742, "step": 36834 }, { "epoch": 11.33, "learning_rate": 8.341484042336417e-06, "loss": 0.0424, "step": 36835 }, { "epoch": 11.33, "learning_rate": 8.34099267502728e-06, "loss": 0.0658, "step": 36836 }, { "epoch": 11.33, "learning_rate": 8.340501311837016e-06, "loss": 0.0423, "step": 36837 }, { "epoch": 11.33, "learning_rate": 8.340009952766843e-06, "loss": 0.0298, "step": 36838 }, { "epoch": 11.34, "learning_rate": 8.339518597817975e-06, "loss": 0.0632, "step": 36839 }, { "epoch": 11.34, "learning_rate": 8.339027246991633e-06, "loss": 0.0422, "step": 36840 }, { "epoch": 11.34, "learning_rate": 8.338535900289038e-06, "loss": 0.0652, "step": 36841 }, { "epoch": 11.34, "learning_rate": 8.338044557711413e-06, "loss": 0.0282, "step": 36842 }, { "epoch": 11.34, "learning_rate": 8.337553219259975e-06, "loss": 0.0742, "step": 36843 }, { "epoch": 11.34, "learning_rate": 8.337061884935944e-06, "loss": 0.0593, "step": 36844 }, { "epoch": 11.34, "learning_rate": 8.336570554740538e-06, "loss": 0.0435, "step": 36845 }, { "epoch": 11.34, "learning_rate": 8.33607922867498e-06, "loss": 0.0523, "step": 36846 }, { "epoch": 11.34, "learning_rate": 8.33558790674049e-06, "loss": 0.0462, "step": 36847 }, { "epoch": 11.34, "learning_rate": 8.335096588938288e-06, "loss": 0.1079, "step": 36848 }, { "epoch": 11.34, "learning_rate": 8.33460527526959e-06, "loss": 0.0485, "step": 36849 }, { "epoch": 11.34, "learning_rate": 8.334113965735614e-06, "loss": 0.0366, "step": 36850 }, { "epoch": 11.34, "learning_rate": 8.333622660337592e-06, "loss": 0.0537, "step": 36851 }, { "epoch": 11.34, "learning_rate": 8.33313135907673e-06, "loss": 0.0485, "step": 36852 }, { "epoch": 11.34, "learning_rate": 8.332640061954256e-06, "loss": 0.044, "step": 36853 }, { "epoch": 11.34, "learning_rate": 8.332148768971388e-06, "loss": 0.0817, "step": 36854 }, { "epoch": 11.34, "learning_rate": 8.331657480129341e-06, "loss": 0.031, "step": 36855 }, { "epoch": 11.34, "learning_rate": 8.331166195429342e-06, "loss": 0.0473, "step": 36856 }, { "epoch": 11.34, "learning_rate": 8.330674914872606e-06, "loss": 0.0138, "step": 36857 }, { "epoch": 11.34, "learning_rate": 8.330183638460356e-06, "loss": 0.0475, "step": 36858 }, { "epoch": 11.34, "learning_rate": 8.329692366193811e-06, "loss": 0.0507, "step": 36859 }, { "epoch": 11.34, "learning_rate": 8.329201098074182e-06, "loss": 0.0721, "step": 36860 }, { "epoch": 11.34, "learning_rate": 8.328709834102703e-06, "loss": 0.0356, "step": 36861 }, { "epoch": 11.34, "learning_rate": 8.328218574280588e-06, "loss": 0.0253, "step": 36862 }, { "epoch": 11.34, "learning_rate": 8.327727318609052e-06, "loss": 0.0737, "step": 36863 }, { "epoch": 11.34, "learning_rate": 8.32723606708932e-06, "loss": 0.0609, "step": 36864 }, { "epoch": 11.34, "learning_rate": 8.326744819722608e-06, "loss": 0.025, "step": 36865 }, { "epoch": 11.34, "learning_rate": 8.326253576510138e-06, "loss": 0.0475, "step": 36866 }, { "epoch": 11.34, "learning_rate": 8.32576233745313e-06, "loss": 0.0583, "step": 36867 }, { "epoch": 11.34, "learning_rate": 8.325271102552803e-06, "loss": 0.1579, "step": 36868 }, { "epoch": 11.34, "learning_rate": 8.324779871810378e-06, "loss": 0.0695, "step": 36869 }, { "epoch": 11.34, "learning_rate": 8.324288645227066e-06, "loss": 0.0427, "step": 36870 }, { "epoch": 11.34, "learning_rate": 8.3237974228041e-06, "loss": 0.0578, "step": 36871 }, { "epoch": 11.35, "learning_rate": 8.323306204542694e-06, "loss": 0.0527, "step": 36872 }, { "epoch": 11.35, "learning_rate": 8.322814990444064e-06, "loss": 0.0534, "step": 36873 }, { "epoch": 11.35, "learning_rate": 8.322323780509432e-06, "loss": 0.0549, "step": 36874 }, { "epoch": 11.35, "learning_rate": 8.321832574740017e-06, "loss": 0.0441, "step": 36875 }, { "epoch": 11.35, "learning_rate": 8.32134137313704e-06, "loss": 0.0436, "step": 36876 }, { "epoch": 11.35, "learning_rate": 8.32085017570172e-06, "loss": 0.0533, "step": 36877 }, { "epoch": 11.35, "learning_rate": 8.320358982435276e-06, "loss": 0.0456, "step": 36878 }, { "epoch": 11.35, "learning_rate": 8.31986779333893e-06, "loss": 0.0704, "step": 36879 }, { "epoch": 11.35, "learning_rate": 8.319376608413896e-06, "loss": 0.0514, "step": 36880 }, { "epoch": 11.35, "learning_rate": 8.318885427661399e-06, "loss": 0.0217, "step": 36881 }, { "epoch": 11.35, "learning_rate": 8.318394251082655e-06, "loss": 0.0875, "step": 36882 }, { "epoch": 11.35, "learning_rate": 8.317903078678888e-06, "loss": 0.0389, "step": 36883 }, { "epoch": 11.35, "learning_rate": 8.317411910451313e-06, "loss": 0.0467, "step": 36884 }, { "epoch": 11.35, "learning_rate": 8.316920746401147e-06, "loss": 0.0831, "step": 36885 }, { "epoch": 11.35, "learning_rate": 8.316429586529616e-06, "loss": 0.0281, "step": 36886 }, { "epoch": 11.35, "learning_rate": 8.315938430837935e-06, "loss": 0.0427, "step": 36887 }, { "epoch": 11.35, "learning_rate": 8.315447279327326e-06, "loss": 0.0749, "step": 36888 }, { "epoch": 11.35, "learning_rate": 8.314956131999008e-06, "loss": 0.0476, "step": 36889 }, { "epoch": 11.35, "learning_rate": 8.314464988854196e-06, "loss": 0.0373, "step": 36890 }, { "epoch": 11.35, "learning_rate": 8.313973849894116e-06, "loss": 0.0668, "step": 36891 }, { "epoch": 11.35, "learning_rate": 8.313482715119985e-06, "loss": 0.042, "step": 36892 }, { "epoch": 11.35, "learning_rate": 8.312991584533022e-06, "loss": 0.0403, "step": 36893 }, { "epoch": 11.35, "learning_rate": 8.312500458134446e-06, "loss": 0.0728, "step": 36894 }, { "epoch": 11.35, "learning_rate": 8.31200933592547e-06, "loss": 0.1299, "step": 36895 }, { "epoch": 11.35, "learning_rate": 8.311518217907328e-06, "loss": 0.0665, "step": 36896 }, { "epoch": 11.35, "learning_rate": 8.311027104081228e-06, "loss": 0.0482, "step": 36897 }, { "epoch": 11.35, "learning_rate": 8.310535994448393e-06, "loss": 0.0536, "step": 36898 }, { "epoch": 11.35, "learning_rate": 8.310044889010042e-06, "loss": 0.0607, "step": 36899 }, { "epoch": 11.35, "learning_rate": 8.309553787767391e-06, "loss": 0.0905, "step": 36900 }, { "epoch": 11.35, "learning_rate": 8.309062690721664e-06, "loss": 0.0293, "step": 36901 }, { "epoch": 11.35, "learning_rate": 8.30857159787408e-06, "loss": 0.0376, "step": 36902 }, { "epoch": 11.35, "learning_rate": 8.308080509225855e-06, "loss": 0.0701, "step": 36903 }, { "epoch": 11.36, "learning_rate": 8.307589424778214e-06, "loss": 0.0671, "step": 36904 }, { "epoch": 11.36, "learning_rate": 8.307098344532362e-06, "loss": 0.081, "step": 36905 }, { "epoch": 11.36, "learning_rate": 8.30660726848954e-06, "loss": 0.0778, "step": 36906 }, { "epoch": 11.36, "learning_rate": 8.306116196650949e-06, "loss": 0.0589, "step": 36907 }, { "epoch": 11.36, "learning_rate": 8.305625129017818e-06, "loss": 0.0493, "step": 36908 }, { "epoch": 11.36, "learning_rate": 8.30513406559136e-06, "loss": 0.0499, "step": 36909 }, { "epoch": 11.36, "learning_rate": 8.304643006372797e-06, "loss": 0.0556, "step": 36910 }, { "epoch": 11.36, "learning_rate": 8.304151951363352e-06, "loss": 0.0473, "step": 36911 }, { "epoch": 11.36, "learning_rate": 8.303660900564239e-06, "loss": 0.0501, "step": 36912 }, { "epoch": 11.36, "learning_rate": 8.303169853976678e-06, "loss": 0.0426, "step": 36913 }, { "epoch": 11.36, "learning_rate": 8.302678811601892e-06, "loss": 0.0358, "step": 36914 }, { "epoch": 11.36, "learning_rate": 8.30218777344109e-06, "loss": 0.0376, "step": 36915 }, { "epoch": 11.36, "learning_rate": 8.301696739495502e-06, "loss": 0.0375, "step": 36916 }, { "epoch": 11.36, "learning_rate": 8.301205709766345e-06, "loss": 0.0364, "step": 36917 }, { "epoch": 11.36, "learning_rate": 8.300714684254834e-06, "loss": 0.0297, "step": 36918 }, { "epoch": 11.36, "learning_rate": 8.300223662962192e-06, "loss": 0.0219, "step": 36919 }, { "epoch": 11.36, "learning_rate": 8.299732645889633e-06, "loss": 0.0411, "step": 36920 }, { "epoch": 11.36, "learning_rate": 8.299241633038381e-06, "loss": 0.0555, "step": 36921 }, { "epoch": 11.36, "learning_rate": 8.298750624409655e-06, "loss": 0.0578, "step": 36922 }, { "epoch": 11.36, "learning_rate": 8.29825962000467e-06, "loss": 0.045, "step": 36923 }, { "epoch": 11.36, "learning_rate": 8.297768619824653e-06, "loss": 0.0462, "step": 36924 }, { "epoch": 11.36, "learning_rate": 8.29727762387081e-06, "loss": 0.06, "step": 36925 }, { "epoch": 11.36, "learning_rate": 8.296786632144373e-06, "loss": 0.0483, "step": 36926 }, { "epoch": 11.36, "learning_rate": 8.296295644646556e-06, "loss": 0.0541, "step": 36927 }, { "epoch": 11.36, "learning_rate": 8.295804661378576e-06, "loss": 0.0603, "step": 36928 }, { "epoch": 11.36, "learning_rate": 8.295313682341654e-06, "loss": 0.0518, "step": 36929 }, { "epoch": 11.36, "learning_rate": 8.294822707537006e-06, "loss": 0.0272, "step": 36930 }, { "epoch": 11.36, "learning_rate": 8.294331736965855e-06, "loss": 0.0571, "step": 36931 }, { "epoch": 11.36, "learning_rate": 8.29384077062942e-06, "loss": 0.0767, "step": 36932 }, { "epoch": 11.36, "learning_rate": 8.293349808528916e-06, "loss": 0.0425, "step": 36933 }, { "epoch": 11.36, "learning_rate": 8.292858850665567e-06, "loss": 0.041, "step": 36934 }, { "epoch": 11.36, "learning_rate": 8.292367897040585e-06, "loss": 0.0235, "step": 36935 }, { "epoch": 11.36, "learning_rate": 8.291876947655197e-06, "loss": 0.0162, "step": 36936 }, { "epoch": 11.37, "learning_rate": 8.291386002510617e-06, "loss": 0.0373, "step": 36937 }, { "epoch": 11.37, "learning_rate": 8.290895061608067e-06, "loss": 0.0399, "step": 36938 }, { "epoch": 11.37, "learning_rate": 8.290404124948763e-06, "loss": 0.0636, "step": 36939 }, { "epoch": 11.37, "learning_rate": 8.289913192533918e-06, "loss": 0.0495, "step": 36940 }, { "epoch": 11.37, "learning_rate": 8.289422264364765e-06, "loss": 0.0398, "step": 36941 }, { "epoch": 11.37, "learning_rate": 8.288931340442514e-06, "loss": 0.0865, "step": 36942 }, { "epoch": 11.37, "learning_rate": 8.288440420768385e-06, "loss": 0.0471, "step": 36943 }, { "epoch": 11.37, "learning_rate": 8.287949505343598e-06, "loss": 0.0525, "step": 36944 }, { "epoch": 11.37, "learning_rate": 8.287458594169367e-06, "loss": 0.0507, "step": 36945 }, { "epoch": 11.37, "learning_rate": 8.286967687246916e-06, "loss": 0.046, "step": 36946 }, { "epoch": 11.37, "learning_rate": 8.286476784577465e-06, "loss": 0.0461, "step": 36947 }, { "epoch": 11.37, "learning_rate": 8.28598588616223e-06, "loss": 0.0334, "step": 36948 }, { "epoch": 11.37, "learning_rate": 8.285494992002429e-06, "loss": 0.0523, "step": 36949 }, { "epoch": 11.37, "learning_rate": 8.285004102099277e-06, "loss": 0.0588, "step": 36950 }, { "epoch": 11.37, "learning_rate": 8.284513216454005e-06, "loss": 0.0644, "step": 36951 }, { "epoch": 11.37, "learning_rate": 8.28402233506782e-06, "loss": 0.0297, "step": 36952 }, { "epoch": 11.37, "learning_rate": 8.283531457941947e-06, "loss": 0.0495, "step": 36953 }, { "epoch": 11.37, "learning_rate": 8.283040585077601e-06, "loss": 0.0408, "step": 36954 }, { "epoch": 11.37, "learning_rate": 8.282549716476e-06, "loss": 0.069, "step": 36955 }, { "epoch": 11.37, "learning_rate": 8.282058852138369e-06, "loss": 0.0332, "step": 36956 }, { "epoch": 11.37, "learning_rate": 8.281567992065922e-06, "loss": 0.049, "step": 36957 }, { "epoch": 11.37, "learning_rate": 8.28107713625988e-06, "loss": 0.0698, "step": 36958 }, { "epoch": 11.37, "learning_rate": 8.280586284721458e-06, "loss": 0.0282, "step": 36959 }, { "epoch": 11.37, "learning_rate": 8.280095437451873e-06, "loss": 0.0384, "step": 36960 }, { "epoch": 11.37, "learning_rate": 8.279604594452351e-06, "loss": 0.0144, "step": 36961 }, { "epoch": 11.37, "learning_rate": 8.27911375572411e-06, "loss": 0.0403, "step": 36962 }, { "epoch": 11.37, "learning_rate": 8.278622921268364e-06, "loss": 0.0629, "step": 36963 }, { "epoch": 11.37, "learning_rate": 8.278132091086332e-06, "loss": 0.046, "step": 36964 }, { "epoch": 11.37, "learning_rate": 8.277641265179232e-06, "loss": 0.0674, "step": 36965 }, { "epoch": 11.37, "learning_rate": 8.277150443548285e-06, "loss": 0.0383, "step": 36966 }, { "epoch": 11.37, "learning_rate": 8.276659626194711e-06, "loss": 0.0466, "step": 36967 }, { "epoch": 11.37, "learning_rate": 8.276168813119726e-06, "loss": 0.0413, "step": 36968 }, { "epoch": 11.38, "learning_rate": 8.275678004324551e-06, "loss": 0.0407, "step": 36969 }, { "epoch": 11.38, "learning_rate": 8.275187199810396e-06, "loss": 0.0538, "step": 36970 }, { "epoch": 11.38, "learning_rate": 8.274696399578492e-06, "loss": 0.0637, "step": 36971 }, { "epoch": 11.38, "learning_rate": 8.274205603630052e-06, "loss": 0.0478, "step": 36972 }, { "epoch": 11.38, "learning_rate": 8.273714811966293e-06, "loss": 0.0489, "step": 36973 }, { "epoch": 11.38, "learning_rate": 8.273224024588434e-06, "loss": 0.0482, "step": 36974 }, { "epoch": 11.38, "learning_rate": 8.272733241497694e-06, "loss": 0.0769, "step": 36975 }, { "epoch": 11.38, "learning_rate": 8.272242462695291e-06, "loss": 0.0613, "step": 36976 }, { "epoch": 11.38, "learning_rate": 8.271751688182447e-06, "loss": 0.0349, "step": 36977 }, { "epoch": 11.38, "learning_rate": 8.271260917960375e-06, "loss": 0.0445, "step": 36978 }, { "epoch": 11.38, "learning_rate": 8.270770152030298e-06, "loss": 0.0516, "step": 36979 }, { "epoch": 11.38, "learning_rate": 8.27027939039343e-06, "loss": 0.0743, "step": 36980 }, { "epoch": 11.38, "learning_rate": 8.269788633050994e-06, "loss": 0.0571, "step": 36981 }, { "epoch": 11.38, "learning_rate": 8.269297880004205e-06, "loss": 0.0759, "step": 36982 }, { "epoch": 11.38, "learning_rate": 8.268807131254288e-06, "loss": 0.051, "step": 36983 }, { "epoch": 11.38, "learning_rate": 8.268316386802451e-06, "loss": 0.0364, "step": 36984 }, { "epoch": 11.38, "learning_rate": 8.267825646649916e-06, "loss": 0.0443, "step": 36985 }, { "epoch": 11.38, "learning_rate": 8.267334910797907e-06, "loss": 0.0469, "step": 36986 }, { "epoch": 11.38, "learning_rate": 8.266844179247636e-06, "loss": 0.0839, "step": 36987 }, { "epoch": 11.38, "learning_rate": 8.266353452000326e-06, "loss": 0.0403, "step": 36988 }, { "epoch": 11.38, "learning_rate": 8.26586272905719e-06, "loss": 0.0566, "step": 36989 }, { "epoch": 11.38, "learning_rate": 8.26537201041945e-06, "loss": 0.0453, "step": 36990 }, { "epoch": 11.38, "learning_rate": 8.264881296088325e-06, "loss": 0.0382, "step": 36991 }, { "epoch": 11.38, "learning_rate": 8.264390586065031e-06, "loss": 0.0518, "step": 36992 }, { "epoch": 11.38, "learning_rate": 8.26389988035079e-06, "loss": 0.0555, "step": 36993 }, { "epoch": 11.38, "learning_rate": 8.263409178946816e-06, "loss": 0.0507, "step": 36994 }, { "epoch": 11.38, "learning_rate": 8.262918481854322e-06, "loss": 0.0565, "step": 36995 }, { "epoch": 11.38, "learning_rate": 8.262427789074543e-06, "loss": 0.0686, "step": 36996 }, { "epoch": 11.38, "learning_rate": 8.261937100608684e-06, "loss": 0.0315, "step": 36997 }, { "epoch": 11.38, "learning_rate": 8.261446416457966e-06, "loss": 0.0688, "step": 36998 }, { "epoch": 11.38, "learning_rate": 8.260955736623608e-06, "loss": 0.0605, "step": 36999 }, { "epoch": 11.38, "learning_rate": 8.260465061106826e-06, "loss": 0.0585, "step": 37000 }, { "epoch": 11.38, "learning_rate": 8.259974389908842e-06, "loss": 0.0445, "step": 37001 }, { "epoch": 11.39, "learning_rate": 8.259483723030874e-06, "loss": 0.0374, "step": 37002 }, { "epoch": 11.39, "learning_rate": 8.258993060474138e-06, "loss": 0.0324, "step": 37003 }, { "epoch": 11.39, "learning_rate": 8.258502402239854e-06, "loss": 0.0843, "step": 37004 }, { "epoch": 11.39, "learning_rate": 8.258011748329234e-06, "loss": 0.0745, "step": 37005 }, { "epoch": 11.39, "learning_rate": 8.257521098743507e-06, "loss": 0.0427, "step": 37006 }, { "epoch": 11.39, "learning_rate": 8.257030453483884e-06, "loss": 0.0539, "step": 37007 }, { "epoch": 11.39, "learning_rate": 8.256539812551582e-06, "loss": 0.0758, "step": 37008 }, { "epoch": 11.39, "learning_rate": 8.256049175947824e-06, "loss": 0.0237, "step": 37009 }, { "epoch": 11.39, "learning_rate": 8.255558543673823e-06, "loss": 0.0299, "step": 37010 }, { "epoch": 11.39, "learning_rate": 8.255067915730802e-06, "loss": 0.0383, "step": 37011 }, { "epoch": 11.39, "learning_rate": 8.254577292119977e-06, "loss": 0.0523, "step": 37012 }, { "epoch": 11.39, "learning_rate": 8.254086672842566e-06, "loss": 0.0402, "step": 37013 }, { "epoch": 11.39, "learning_rate": 8.253596057899788e-06, "loss": 0.0545, "step": 37014 }, { "epoch": 11.39, "learning_rate": 8.253105447292855e-06, "loss": 0.0528, "step": 37015 }, { "epoch": 11.39, "learning_rate": 8.252614841022996e-06, "loss": 0.0509, "step": 37016 }, { "epoch": 11.39, "learning_rate": 8.252124239091423e-06, "loss": 0.065, "step": 37017 }, { "epoch": 11.39, "learning_rate": 8.251633641499353e-06, "loss": 0.0414, "step": 37018 }, { "epoch": 11.39, "learning_rate": 8.251143048248005e-06, "loss": 0.0362, "step": 37019 }, { "epoch": 11.39, "learning_rate": 8.250652459338596e-06, "loss": 0.0518, "step": 37020 }, { "epoch": 11.39, "learning_rate": 8.250161874772347e-06, "loss": 0.0254, "step": 37021 }, { "epoch": 11.39, "learning_rate": 8.249671294550473e-06, "loss": 0.0398, "step": 37022 }, { "epoch": 11.39, "learning_rate": 8.249180718674197e-06, "loss": 0.0544, "step": 37023 }, { "epoch": 11.39, "learning_rate": 8.24869014714473e-06, "loss": 0.0456, "step": 37024 }, { "epoch": 11.39, "learning_rate": 8.248199579963292e-06, "loss": 0.0403, "step": 37025 }, { "epoch": 11.39, "learning_rate": 8.247709017131104e-06, "loss": 0.0638, "step": 37026 }, { "epoch": 11.39, "learning_rate": 8.247218458649385e-06, "loss": 0.0349, "step": 37027 }, { "epoch": 11.39, "learning_rate": 8.246727904519348e-06, "loss": 0.0629, "step": 37028 }, { "epoch": 11.39, "learning_rate": 8.246237354742211e-06, "loss": 0.058, "step": 37029 }, { "epoch": 11.39, "learning_rate": 8.245746809319194e-06, "loss": 0.0424, "step": 37030 }, { "epoch": 11.39, "learning_rate": 8.245256268251516e-06, "loss": 0.0445, "step": 37031 }, { "epoch": 11.39, "learning_rate": 8.244765731540393e-06, "loss": 0.0391, "step": 37032 }, { "epoch": 11.39, "learning_rate": 8.244275199187045e-06, "loss": 0.0571, "step": 37033 }, { "epoch": 11.4, "learning_rate": 8.243784671192687e-06, "loss": 0.0752, "step": 37034 }, { "epoch": 11.4, "learning_rate": 8.243294147558536e-06, "loss": 0.0165, "step": 37035 }, { "epoch": 11.4, "learning_rate": 8.242803628285814e-06, "loss": 0.0598, "step": 37036 }, { "epoch": 11.4, "learning_rate": 8.242313113375739e-06, "loss": 0.0536, "step": 37037 }, { "epoch": 11.4, "learning_rate": 8.241822602829527e-06, "loss": 0.0553, "step": 37038 }, { "epoch": 11.4, "learning_rate": 8.241332096648393e-06, "loss": 0.0686, "step": 37039 }, { "epoch": 11.4, "learning_rate": 8.240841594833554e-06, "loss": 0.0628, "step": 37040 }, { "epoch": 11.4, "learning_rate": 8.240351097386238e-06, "loss": 0.0806, "step": 37041 }, { "epoch": 11.4, "learning_rate": 8.239860604307653e-06, "loss": 0.0785, "step": 37042 }, { "epoch": 11.4, "learning_rate": 8.239370115599019e-06, "loss": 0.0594, "step": 37043 }, { "epoch": 11.4, "learning_rate": 8.238879631261556e-06, "loss": 0.0954, "step": 37044 }, { "epoch": 11.4, "learning_rate": 8.238389151296475e-06, "loss": 0.0637, "step": 37045 }, { "epoch": 11.4, "learning_rate": 8.237898675705003e-06, "loss": 0.0522, "step": 37046 }, { "epoch": 11.4, "learning_rate": 8.237408204488353e-06, "loss": 0.0388, "step": 37047 }, { "epoch": 11.4, "learning_rate": 8.236917737647747e-06, "loss": 0.0381, "step": 37048 }, { "epoch": 11.4, "learning_rate": 8.236427275184394e-06, "loss": 0.0855, "step": 37049 }, { "epoch": 11.4, "learning_rate": 8.235936817099515e-06, "loss": 0.0503, "step": 37050 }, { "epoch": 11.4, "learning_rate": 8.235446363394334e-06, "loss": 0.0167, "step": 37051 }, { "epoch": 11.4, "learning_rate": 8.234955914070062e-06, "loss": 0.1168, "step": 37052 }, { "epoch": 11.4, "learning_rate": 8.234465469127919e-06, "loss": 0.083, "step": 37053 }, { "epoch": 11.4, "learning_rate": 8.233975028569123e-06, "loss": 0.0517, "step": 37054 }, { "epoch": 11.4, "learning_rate": 8.233484592394887e-06, "loss": 0.0379, "step": 37055 }, { "epoch": 11.4, "learning_rate": 8.232994160606436e-06, "loss": 0.0344, "step": 37056 }, { "epoch": 11.4, "learning_rate": 8.232503733204985e-06, "loss": 0.0547, "step": 37057 }, { "epoch": 11.4, "learning_rate": 8.232013310191748e-06, "loss": 0.037, "step": 37058 }, { "epoch": 11.4, "learning_rate": 8.23152289156795e-06, "loss": 0.0458, "step": 37059 }, { "epoch": 11.4, "learning_rate": 8.231032477334796e-06, "loss": 0.0464, "step": 37060 }, { "epoch": 11.4, "learning_rate": 8.230542067493517e-06, "loss": 0.0366, "step": 37061 }, { "epoch": 11.4, "learning_rate": 8.230051662045326e-06, "loss": 0.0386, "step": 37062 }, { "epoch": 11.4, "learning_rate": 8.229561260991437e-06, "loss": 0.0609, "step": 37063 }, { "epoch": 11.4, "learning_rate": 8.229070864333072e-06, "loss": 0.0599, "step": 37064 }, { "epoch": 11.4, "learning_rate": 8.228580472071443e-06, "loss": 0.0382, "step": 37065 }, { "epoch": 11.4, "learning_rate": 8.228090084207773e-06, "loss": 0.0356, "step": 37066 }, { "epoch": 11.41, "learning_rate": 8.227599700743279e-06, "loss": 0.0468, "step": 37067 }, { "epoch": 11.41, "learning_rate": 8.227109321679177e-06, "loss": 0.104, "step": 37068 }, { "epoch": 11.41, "learning_rate": 8.226618947016685e-06, "loss": 0.0381, "step": 37069 }, { "epoch": 11.41, "learning_rate": 8.226128576757015e-06, "loss": 0.0816, "step": 37070 }, { "epoch": 11.41, "learning_rate": 8.225638210901394e-06, "loss": 0.0669, "step": 37071 }, { "epoch": 11.41, "learning_rate": 8.225147849451035e-06, "loss": 0.0638, "step": 37072 }, { "epoch": 11.41, "learning_rate": 8.224657492407156e-06, "loss": 0.0602, "step": 37073 }, { "epoch": 11.41, "learning_rate": 8.224167139770973e-06, "loss": 0.0477, "step": 37074 }, { "epoch": 11.41, "learning_rate": 8.223676791543702e-06, "loss": 0.0742, "step": 37075 }, { "epoch": 11.41, "learning_rate": 8.223186447726564e-06, "loss": 0.0611, "step": 37076 }, { "epoch": 11.41, "learning_rate": 8.222696108320776e-06, "loss": 0.0323, "step": 37077 }, { "epoch": 11.41, "learning_rate": 8.222205773327555e-06, "loss": 0.0395, "step": 37078 }, { "epoch": 11.41, "learning_rate": 8.221715442748117e-06, "loss": 0.0501, "step": 37079 }, { "epoch": 11.41, "learning_rate": 8.221225116583677e-06, "loss": 0.0701, "step": 37080 }, { "epoch": 11.41, "learning_rate": 8.220734794835459e-06, "loss": 0.0716, "step": 37081 }, { "epoch": 11.41, "learning_rate": 8.220244477504677e-06, "loss": 0.0815, "step": 37082 }, { "epoch": 11.41, "learning_rate": 8.219754164592548e-06, "loss": 0.0574, "step": 37083 }, { "epoch": 11.41, "learning_rate": 8.21926385610029e-06, "loss": 0.044, "step": 37084 }, { "epoch": 11.41, "learning_rate": 8.218773552029115e-06, "loss": 0.0339, "step": 37085 }, { "epoch": 11.41, "learning_rate": 8.21828325238025e-06, "loss": 0.0529, "step": 37086 }, { "epoch": 11.41, "learning_rate": 8.217792957154907e-06, "loss": 0.0399, "step": 37087 }, { "epoch": 11.41, "learning_rate": 8.217302666354301e-06, "loss": 0.0492, "step": 37088 }, { "epoch": 11.41, "learning_rate": 8.216812379979655e-06, "loss": 0.0633, "step": 37089 }, { "epoch": 11.41, "learning_rate": 8.21632209803218e-06, "loss": 0.0358, "step": 37090 }, { "epoch": 11.41, "learning_rate": 8.215831820513098e-06, "loss": 0.0483, "step": 37091 }, { "epoch": 11.41, "learning_rate": 8.215341547423624e-06, "loss": 0.0466, "step": 37092 }, { "epoch": 11.41, "learning_rate": 8.21485127876498e-06, "loss": 0.0267, "step": 37093 }, { "epoch": 11.41, "learning_rate": 8.214361014538374e-06, "loss": 0.0425, "step": 37094 }, { "epoch": 11.41, "learning_rate": 8.213870754745025e-06, "loss": 0.0688, "step": 37095 }, { "epoch": 11.41, "learning_rate": 8.21338049938616e-06, "loss": 0.054, "step": 37096 }, { "epoch": 11.41, "learning_rate": 8.21289024846299e-06, "loss": 0.0337, "step": 37097 }, { "epoch": 11.41, "learning_rate": 8.212400001976729e-06, "loss": 0.0365, "step": 37098 }, { "epoch": 11.42, "learning_rate": 8.211909759928597e-06, "loss": 0.0607, "step": 37099 }, { "epoch": 11.42, "learning_rate": 8.211419522319808e-06, "loss": 0.0293, "step": 37100 }, { "epoch": 11.42, "learning_rate": 8.210929289151587e-06, "loss": 0.0471, "step": 37101 }, { "epoch": 11.42, "learning_rate": 8.210439060425144e-06, "loss": 0.0252, "step": 37102 }, { "epoch": 11.42, "learning_rate": 8.2099488361417e-06, "loss": 0.0326, "step": 37103 }, { "epoch": 11.42, "learning_rate": 8.209458616302472e-06, "loss": 0.0613, "step": 37104 }, { "epoch": 11.42, "learning_rate": 8.208968400908668e-06, "loss": 0.0539, "step": 37105 }, { "epoch": 11.42, "learning_rate": 8.20847818996152e-06, "loss": 0.0283, "step": 37106 }, { "epoch": 11.42, "learning_rate": 8.207987983462236e-06, "loss": 0.0612, "step": 37107 }, { "epoch": 11.42, "learning_rate": 8.207497781412033e-06, "loss": 0.0469, "step": 37108 }, { "epoch": 11.42, "learning_rate": 8.207007583812131e-06, "loss": 0.0393, "step": 37109 }, { "epoch": 11.42, "learning_rate": 8.206517390663743e-06, "loss": 0.0586, "step": 37110 }, { "epoch": 11.42, "learning_rate": 8.206027201968092e-06, "loss": 0.0554, "step": 37111 }, { "epoch": 11.42, "learning_rate": 8.205537017726391e-06, "loss": 0.0427, "step": 37112 }, { "epoch": 11.42, "learning_rate": 8.205046837939858e-06, "loss": 0.0495, "step": 37113 }, { "epoch": 11.42, "learning_rate": 8.204556662609713e-06, "loss": 0.062, "step": 37114 }, { "epoch": 11.42, "learning_rate": 8.204066491737163e-06, "loss": 0.0673, "step": 37115 }, { "epoch": 11.42, "learning_rate": 8.203576325323435e-06, "loss": 0.0785, "step": 37116 }, { "epoch": 11.42, "learning_rate": 8.203086163369744e-06, "loss": 0.0269, "step": 37117 }, { "epoch": 11.42, "learning_rate": 8.202596005877307e-06, "loss": 0.0385, "step": 37118 }, { "epoch": 11.42, "learning_rate": 8.202105852847336e-06, "loss": 0.0354, "step": 37119 }, { "epoch": 11.42, "learning_rate": 8.20161570428105e-06, "loss": 0.037, "step": 37120 }, { "epoch": 11.42, "learning_rate": 8.20112556017967e-06, "loss": 0.0421, "step": 37121 }, { "epoch": 11.42, "learning_rate": 8.200635420544411e-06, "loss": 0.0625, "step": 37122 }, { "epoch": 11.42, "learning_rate": 8.200145285376489e-06, "loss": 0.0608, "step": 37123 }, { "epoch": 11.42, "learning_rate": 8.199655154677123e-06, "loss": 0.0457, "step": 37124 }, { "epoch": 11.42, "learning_rate": 8.19916502844752e-06, "loss": 0.078, "step": 37125 }, { "epoch": 11.42, "learning_rate": 8.19867490668891e-06, "loss": 0.0382, "step": 37126 }, { "epoch": 11.42, "learning_rate": 8.198184789402508e-06, "loss": 0.0663, "step": 37127 }, { "epoch": 11.42, "learning_rate": 8.197694676589523e-06, "loss": 0.0285, "step": 37128 }, { "epoch": 11.42, "learning_rate": 8.197204568251178e-06, "loss": 0.0463, "step": 37129 }, { "epoch": 11.42, "learning_rate": 8.196714464388687e-06, "loss": 0.0683, "step": 37130 }, { "epoch": 11.42, "learning_rate": 8.196224365003267e-06, "loss": 0.0762, "step": 37131 }, { "epoch": 11.43, "learning_rate": 8.195734270096137e-06, "loss": 0.0424, "step": 37132 }, { "epoch": 11.43, "learning_rate": 8.195244179668512e-06, "loss": 0.0607, "step": 37133 }, { "epoch": 11.43, "learning_rate": 8.19475409372161e-06, "loss": 0.0548, "step": 37134 }, { "epoch": 11.43, "learning_rate": 8.194264012256643e-06, "loss": 0.0745, "step": 37135 }, { "epoch": 11.43, "learning_rate": 8.193773935274836e-06, "loss": 0.0549, "step": 37136 }, { "epoch": 11.43, "learning_rate": 8.193283862777401e-06, "loss": 0.052, "step": 37137 }, { "epoch": 11.43, "learning_rate": 8.192793794765556e-06, "loss": 0.0688, "step": 37138 }, { "epoch": 11.43, "learning_rate": 8.192303731240515e-06, "loss": 0.0584, "step": 37139 }, { "epoch": 11.43, "learning_rate": 8.191813672203493e-06, "loss": 0.0369, "step": 37140 }, { "epoch": 11.43, "learning_rate": 8.191323617655717e-06, "loss": 0.0283, "step": 37141 }, { "epoch": 11.43, "learning_rate": 8.190833567598392e-06, "loss": 0.0472, "step": 37142 }, { "epoch": 11.43, "learning_rate": 8.190343522032742e-06, "loss": 0.0519, "step": 37143 }, { "epoch": 11.43, "learning_rate": 8.189853480959982e-06, "loss": 0.0467, "step": 37144 }, { "epoch": 11.43, "learning_rate": 8.189363444381323e-06, "loss": 0.074, "step": 37145 }, { "epoch": 11.43, "learning_rate": 8.18887341229799e-06, "loss": 0.0433, "step": 37146 }, { "epoch": 11.43, "learning_rate": 8.188383384711195e-06, "loss": 0.0477, "step": 37147 }, { "epoch": 11.43, "learning_rate": 8.18789336162216e-06, "loss": 0.0595, "step": 37148 }, { "epoch": 11.43, "learning_rate": 8.187403343032093e-06, "loss": 0.0572, "step": 37149 }, { "epoch": 11.43, "learning_rate": 8.186913328942213e-06, "loss": 0.0318, "step": 37150 }, { "epoch": 11.43, "learning_rate": 8.186423319353742e-06, "loss": 0.0476, "step": 37151 }, { "epoch": 11.43, "learning_rate": 8.185933314267891e-06, "loss": 0.0492, "step": 37152 }, { "epoch": 11.43, "learning_rate": 8.185443313685881e-06, "loss": 0.0465, "step": 37153 }, { "epoch": 11.43, "learning_rate": 8.184953317608924e-06, "loss": 0.0782, "step": 37154 }, { "epoch": 11.43, "learning_rate": 8.184463326038236e-06, "loss": 0.0592, "step": 37155 }, { "epoch": 11.43, "learning_rate": 8.18397333897504e-06, "loss": 0.0839, "step": 37156 }, { "epoch": 11.43, "learning_rate": 8.183483356420547e-06, "loss": 0.0432, "step": 37157 }, { "epoch": 11.43, "learning_rate": 8.182993378375976e-06, "loss": 0.0654, "step": 37158 }, { "epoch": 11.43, "learning_rate": 8.182503404842546e-06, "loss": 0.0235, "step": 37159 }, { "epoch": 11.43, "learning_rate": 8.182013435821461e-06, "loss": 0.0989, "step": 37160 }, { "epoch": 11.43, "learning_rate": 8.181523471313952e-06, "loss": 0.0525, "step": 37161 }, { "epoch": 11.43, "learning_rate": 8.181033511321232e-06, "loss": 0.0587, "step": 37162 }, { "epoch": 11.43, "learning_rate": 8.180543555844514e-06, "loss": 0.0795, "step": 37163 }, { "epoch": 11.44, "learning_rate": 8.180053604885016e-06, "loss": 0.0469, "step": 37164 }, { "epoch": 11.44, "learning_rate": 8.179563658443953e-06, "loss": 0.0473, "step": 37165 }, { "epoch": 11.44, "learning_rate": 8.179073716522542e-06, "loss": 0.0518, "step": 37166 }, { "epoch": 11.44, "learning_rate": 8.178583779122001e-06, "loss": 0.0499, "step": 37167 }, { "epoch": 11.44, "learning_rate": 8.178093846243546e-06, "loss": 0.0405, "step": 37168 }, { "epoch": 11.44, "learning_rate": 8.177603917888394e-06, "loss": 0.0335, "step": 37169 }, { "epoch": 11.44, "learning_rate": 8.177113994057756e-06, "loss": 0.0641, "step": 37170 }, { "epoch": 11.44, "learning_rate": 8.176624074752854e-06, "loss": 0.0827, "step": 37171 }, { "epoch": 11.44, "learning_rate": 8.176134159974907e-06, "loss": 0.0467, "step": 37172 }, { "epoch": 11.44, "learning_rate": 8.175644249725124e-06, "loss": 0.0584, "step": 37173 }, { "epoch": 11.44, "learning_rate": 8.175154344004724e-06, "loss": 0.0361, "step": 37174 }, { "epoch": 11.44, "learning_rate": 8.174664442814923e-06, "loss": 0.046, "step": 37175 }, { "epoch": 11.44, "learning_rate": 8.17417454615694e-06, "loss": 0.0556, "step": 37176 }, { "epoch": 11.44, "learning_rate": 8.17368465403199e-06, "loss": 0.0548, "step": 37177 }, { "epoch": 11.44, "learning_rate": 8.173194766441286e-06, "loss": 0.052, "step": 37178 }, { "epoch": 11.44, "learning_rate": 8.17270488338605e-06, "loss": 0.041, "step": 37179 }, { "epoch": 11.44, "learning_rate": 8.17221500486749e-06, "loss": 0.0138, "step": 37180 }, { "epoch": 11.44, "learning_rate": 8.171725130886831e-06, "loss": 0.047, "step": 37181 }, { "epoch": 11.44, "learning_rate": 8.171235261445289e-06, "loss": 0.0782, "step": 37182 }, { "epoch": 11.44, "learning_rate": 8.170745396544072e-06, "loss": 0.0336, "step": 37183 }, { "epoch": 11.44, "learning_rate": 8.170255536184404e-06, "loss": 0.0595, "step": 37184 }, { "epoch": 11.44, "learning_rate": 8.169765680367496e-06, "loss": 0.0322, "step": 37185 }, { "epoch": 11.44, "learning_rate": 8.169275829094566e-06, "loss": 0.0392, "step": 37186 }, { "epoch": 11.44, "learning_rate": 8.168785982366831e-06, "loss": 0.0626, "step": 37187 }, { "epoch": 11.44, "learning_rate": 8.168296140185508e-06, "loss": 0.019, "step": 37188 }, { "epoch": 11.44, "learning_rate": 8.167806302551812e-06, "loss": 0.0269, "step": 37189 }, { "epoch": 11.44, "learning_rate": 8.167316469466957e-06, "loss": 0.0434, "step": 37190 }, { "epoch": 11.44, "learning_rate": 8.166826640932161e-06, "loss": 0.0564, "step": 37191 }, { "epoch": 11.44, "learning_rate": 8.166336816948643e-06, "loss": 0.0778, "step": 37192 }, { "epoch": 11.44, "learning_rate": 8.165846997517617e-06, "loss": 0.0513, "step": 37193 }, { "epoch": 11.44, "learning_rate": 8.165357182640297e-06, "loss": 0.0385, "step": 37194 }, { "epoch": 11.44, "learning_rate": 8.164867372317896e-06, "loss": 0.0424, "step": 37195 }, { "epoch": 11.44, "learning_rate": 8.16437756655164e-06, "loss": 0.0514, "step": 37196 }, { "epoch": 11.45, "learning_rate": 8.163887765342738e-06, "loss": 0.047, "step": 37197 }, { "epoch": 11.45, "learning_rate": 8.163397968692409e-06, "loss": 0.0438, "step": 37198 }, { "epoch": 11.45, "learning_rate": 8.162908176601866e-06, "loss": 0.0698, "step": 37199 }, { "epoch": 11.45, "learning_rate": 8.162418389072325e-06, "loss": 0.0387, "step": 37200 }, { "epoch": 11.45, "learning_rate": 8.161928606105007e-06, "loss": 0.047, "step": 37201 }, { "epoch": 11.45, "learning_rate": 8.161438827701123e-06, "loss": 0.0345, "step": 37202 }, { "epoch": 11.45, "learning_rate": 8.160949053861894e-06, "loss": 0.0521, "step": 37203 }, { "epoch": 11.45, "learning_rate": 8.16045928458853e-06, "loss": 0.0437, "step": 37204 }, { "epoch": 11.45, "learning_rate": 8.159969519882247e-06, "loss": 0.0808, "step": 37205 }, { "epoch": 11.45, "learning_rate": 8.159479759744267e-06, "loss": 0.0351, "step": 37206 }, { "epoch": 11.45, "learning_rate": 8.158990004175803e-06, "loss": 0.0538, "step": 37207 }, { "epoch": 11.45, "learning_rate": 8.15850025317807e-06, "loss": 0.0505, "step": 37208 }, { "epoch": 11.45, "learning_rate": 8.158010506752283e-06, "loss": 0.0433, "step": 37209 }, { "epoch": 11.45, "learning_rate": 8.15752076489966e-06, "loss": 0.0322, "step": 37210 }, { "epoch": 11.45, "learning_rate": 8.157031027621416e-06, "loss": 0.0402, "step": 37211 }, { "epoch": 11.45, "learning_rate": 8.156541294918767e-06, "loss": 0.0735, "step": 37212 }, { "epoch": 11.45, "learning_rate": 8.15605156679293e-06, "loss": 0.0588, "step": 37213 }, { "epoch": 11.45, "learning_rate": 8.155561843245123e-06, "loss": 0.0485, "step": 37214 }, { "epoch": 11.45, "learning_rate": 8.155072124276551e-06, "loss": 0.0519, "step": 37215 }, { "epoch": 11.45, "learning_rate": 8.154582409888441e-06, "loss": 0.0555, "step": 37216 }, { "epoch": 11.45, "learning_rate": 8.154092700082008e-06, "loss": 0.0836, "step": 37217 }, { "epoch": 11.45, "learning_rate": 8.153602994858463e-06, "loss": 0.0893, "step": 37218 }, { "epoch": 11.45, "learning_rate": 8.153113294219024e-06, "loss": 0.0378, "step": 37219 }, { "epoch": 11.45, "learning_rate": 8.152623598164906e-06, "loss": 0.0436, "step": 37220 }, { "epoch": 11.45, "learning_rate": 8.152133906697328e-06, "loss": 0.0407, "step": 37221 }, { "epoch": 11.45, "learning_rate": 8.1516442198175e-06, "loss": 0.0503, "step": 37222 }, { "epoch": 11.45, "learning_rate": 8.151154537526644e-06, "loss": 0.0626, "step": 37223 }, { "epoch": 11.45, "learning_rate": 8.150664859825973e-06, "loss": 0.0691, "step": 37224 }, { "epoch": 11.45, "learning_rate": 8.150175186716699e-06, "loss": 0.0484, "step": 37225 }, { "epoch": 11.45, "learning_rate": 8.149685518200044e-06, "loss": 0.0459, "step": 37226 }, { "epoch": 11.45, "learning_rate": 8.149195854277223e-06, "loss": 0.0822, "step": 37227 }, { "epoch": 11.45, "learning_rate": 8.148706194949448e-06, "loss": 0.0382, "step": 37228 }, { "epoch": 11.46, "learning_rate": 8.148216540217935e-06, "loss": 0.0501, "step": 37229 }, { "epoch": 11.46, "learning_rate": 8.1477268900839e-06, "loss": 0.0605, "step": 37230 }, { "epoch": 11.46, "learning_rate": 8.147237244548563e-06, "loss": 0.0291, "step": 37231 }, { "epoch": 11.46, "learning_rate": 8.146747603613135e-06, "loss": 0.0257, "step": 37232 }, { "epoch": 11.46, "learning_rate": 8.146257967278833e-06, "loss": 0.0471, "step": 37233 }, { "epoch": 11.46, "learning_rate": 8.145768335546874e-06, "loss": 0.096, "step": 37234 }, { "epoch": 11.46, "learning_rate": 8.14527870841847e-06, "loss": 0.048, "step": 37235 }, { "epoch": 11.46, "learning_rate": 8.14478908589484e-06, "loss": 0.0386, "step": 37236 }, { "epoch": 11.46, "learning_rate": 8.144299467977199e-06, "loss": 0.0668, "step": 37237 }, { "epoch": 11.46, "learning_rate": 8.143809854666765e-06, "loss": 0.0459, "step": 37238 }, { "epoch": 11.46, "learning_rate": 8.143320245964743e-06, "loss": 0.0475, "step": 37239 }, { "epoch": 11.46, "learning_rate": 8.142830641872362e-06, "loss": 0.0522, "step": 37240 }, { "epoch": 11.46, "learning_rate": 8.142341042390832e-06, "loss": 0.0369, "step": 37241 }, { "epoch": 11.46, "learning_rate": 8.141851447521367e-06, "loss": 0.0477, "step": 37242 }, { "epoch": 11.46, "learning_rate": 8.141361857265185e-06, "loss": 0.0709, "step": 37243 }, { "epoch": 11.46, "learning_rate": 8.140872271623499e-06, "loss": 0.0394, "step": 37244 }, { "epoch": 11.46, "learning_rate": 8.140382690597526e-06, "loss": 0.0505, "step": 37245 }, { "epoch": 11.46, "learning_rate": 8.139893114188482e-06, "loss": 0.0669, "step": 37246 }, { "epoch": 11.46, "learning_rate": 8.139403542397582e-06, "loss": 0.082, "step": 37247 }, { "epoch": 11.46, "learning_rate": 8.138913975226044e-06, "loss": 0.0488, "step": 37248 }, { "epoch": 11.46, "learning_rate": 8.138424412675074e-06, "loss": 0.0637, "step": 37249 }, { "epoch": 11.46, "learning_rate": 8.137934854745897e-06, "loss": 0.0656, "step": 37250 }, { "epoch": 11.46, "learning_rate": 8.13744530143973e-06, "loss": 0.0807, "step": 37251 }, { "epoch": 11.46, "learning_rate": 8.136955752757783e-06, "loss": 0.0618, "step": 37252 }, { "epoch": 11.46, "learning_rate": 8.13646620870127e-06, "loss": 0.037, "step": 37253 }, { "epoch": 11.46, "learning_rate": 8.135976669271408e-06, "loss": 0.0475, "step": 37254 }, { "epoch": 11.46, "learning_rate": 8.135487134469416e-06, "loss": 0.0274, "step": 37255 }, { "epoch": 11.46, "learning_rate": 8.134997604296506e-06, "loss": 0.0797, "step": 37256 }, { "epoch": 11.46, "learning_rate": 8.134508078753896e-06, "loss": 0.0803, "step": 37257 }, { "epoch": 11.46, "learning_rate": 8.134018557842798e-06, "loss": 0.0411, "step": 37258 }, { "epoch": 11.46, "learning_rate": 8.133529041564427e-06, "loss": 0.0808, "step": 37259 }, { "epoch": 11.46, "learning_rate": 8.133039529920002e-06, "loss": 0.0425, "step": 37260 }, { "epoch": 11.46, "learning_rate": 8.132550022910737e-06, "loss": 0.0449, "step": 37261 }, { "epoch": 11.47, "learning_rate": 8.132060520537849e-06, "loss": 0.0405, "step": 37262 }, { "epoch": 11.47, "learning_rate": 8.131571022802548e-06, "loss": 0.0642, "step": 37263 }, { "epoch": 11.47, "learning_rate": 8.13108152970605e-06, "loss": 0.0657, "step": 37264 }, { "epoch": 11.47, "learning_rate": 8.130592041249576e-06, "loss": 0.0448, "step": 37265 }, { "epoch": 11.47, "learning_rate": 8.130102557434338e-06, "loss": 0.0692, "step": 37266 }, { "epoch": 11.47, "learning_rate": 8.129613078261551e-06, "loss": 0.0432, "step": 37267 }, { "epoch": 11.47, "learning_rate": 8.129123603732431e-06, "loss": 0.0695, "step": 37268 }, { "epoch": 11.47, "learning_rate": 8.12863413384819e-06, "loss": 0.0425, "step": 37269 }, { "epoch": 11.47, "learning_rate": 8.12814466861005e-06, "loss": 0.0365, "step": 37270 }, { "epoch": 11.47, "learning_rate": 8.127655208019219e-06, "loss": 0.0483, "step": 37271 }, { "epoch": 11.47, "learning_rate": 8.127165752076918e-06, "loss": 0.0478, "step": 37272 }, { "epoch": 11.47, "learning_rate": 8.126676300784358e-06, "loss": 0.0667, "step": 37273 }, { "epoch": 11.47, "learning_rate": 8.126186854142752e-06, "loss": 0.0245, "step": 37274 }, { "epoch": 11.47, "learning_rate": 8.125697412153326e-06, "loss": 0.0368, "step": 37275 }, { "epoch": 11.47, "learning_rate": 8.125207974817284e-06, "loss": 0.0477, "step": 37276 }, { "epoch": 11.47, "learning_rate": 8.124718542135846e-06, "loss": 0.0826, "step": 37277 }, { "epoch": 11.47, "learning_rate": 8.124229114110227e-06, "loss": 0.0513, "step": 37278 }, { "epoch": 11.47, "learning_rate": 8.12373969074164e-06, "loss": 0.0466, "step": 37279 }, { "epoch": 11.47, "learning_rate": 8.1232502720313e-06, "loss": 0.0827, "step": 37280 }, { "epoch": 11.47, "learning_rate": 8.122760857980426e-06, "loss": 0.0505, "step": 37281 }, { "epoch": 11.47, "learning_rate": 8.122271448590234e-06, "loss": 0.0734, "step": 37282 }, { "epoch": 11.47, "learning_rate": 8.121782043861932e-06, "loss": 0.0654, "step": 37283 }, { "epoch": 11.47, "learning_rate": 8.121292643796734e-06, "loss": 0.0351, "step": 37284 }, { "epoch": 11.47, "learning_rate": 8.120803248395868e-06, "loss": 0.0714, "step": 37285 }, { "epoch": 11.47, "learning_rate": 8.120313857660537e-06, "loss": 0.0347, "step": 37286 }, { "epoch": 11.47, "learning_rate": 8.119824471591962e-06, "loss": 0.0542, "step": 37287 }, { "epoch": 11.47, "learning_rate": 8.119335090191354e-06, "loss": 0.0364, "step": 37288 }, { "epoch": 11.47, "learning_rate": 8.11884571345993e-06, "loss": 0.0674, "step": 37289 }, { "epoch": 11.47, "learning_rate": 8.118356341398906e-06, "loss": 0.03, "step": 37290 }, { "epoch": 11.47, "learning_rate": 8.117866974009496e-06, "loss": 0.0512, "step": 37291 }, { "epoch": 11.47, "learning_rate": 8.117377611292914e-06, "loss": 0.086, "step": 37292 }, { "epoch": 11.47, "learning_rate": 8.116888253250378e-06, "loss": 0.0409, "step": 37293 }, { "epoch": 11.48, "learning_rate": 8.116398899883096e-06, "loss": 0.0557, "step": 37294 }, { "epoch": 11.48, "learning_rate": 8.11590955119229e-06, "loss": 0.0678, "step": 37295 }, { "epoch": 11.48, "learning_rate": 8.115420207179177e-06, "loss": 0.0571, "step": 37296 }, { "epoch": 11.48, "learning_rate": 8.114930867844963e-06, "loss": 0.0348, "step": 37297 }, { "epoch": 11.48, "learning_rate": 8.114441533190869e-06, "loss": 0.0479, "step": 37298 }, { "epoch": 11.48, "learning_rate": 8.113952203218106e-06, "loss": 0.0193, "step": 37299 }, { "epoch": 11.48, "learning_rate": 8.113462877927893e-06, "loss": 0.0466, "step": 37300 }, { "epoch": 11.48, "learning_rate": 8.112973557321445e-06, "loss": 0.0477, "step": 37301 }, { "epoch": 11.48, "learning_rate": 8.112484241399972e-06, "loss": 0.0843, "step": 37302 }, { "epoch": 11.48, "learning_rate": 8.111994930164694e-06, "loss": 0.0811, "step": 37303 }, { "epoch": 11.48, "learning_rate": 8.11150562361682e-06, "loss": 0.0479, "step": 37304 }, { "epoch": 11.48, "learning_rate": 8.11101632175757e-06, "loss": 0.0551, "step": 37305 }, { "epoch": 11.48, "learning_rate": 8.110527024588162e-06, "loss": 0.0293, "step": 37306 }, { "epoch": 11.48, "learning_rate": 8.1100377321098e-06, "loss": 0.0498, "step": 37307 }, { "epoch": 11.48, "learning_rate": 8.109548444323708e-06, "loss": 0.0428, "step": 37308 }, { "epoch": 11.48, "learning_rate": 8.109059161231095e-06, "loss": 0.0568, "step": 37309 }, { "epoch": 11.48, "learning_rate": 8.10856988283318e-06, "loss": 0.0888, "step": 37310 }, { "epoch": 11.48, "learning_rate": 8.108080609131176e-06, "loss": 0.029, "step": 37311 }, { "epoch": 11.48, "learning_rate": 8.107591340126297e-06, "loss": 0.0617, "step": 37312 }, { "epoch": 11.48, "learning_rate": 8.10710207581976e-06, "loss": 0.0345, "step": 37313 }, { "epoch": 11.48, "learning_rate": 8.106612816212775e-06, "loss": 0.0933, "step": 37314 }, { "epoch": 11.48, "learning_rate": 8.106123561306562e-06, "loss": 0.043, "step": 37315 }, { "epoch": 11.48, "learning_rate": 8.105634311102336e-06, "loss": 0.0494, "step": 37316 }, { "epoch": 11.48, "learning_rate": 8.105145065601309e-06, "loss": 0.0426, "step": 37317 }, { "epoch": 11.48, "learning_rate": 8.104655824804694e-06, "loss": 0.0305, "step": 37318 }, { "epoch": 11.48, "learning_rate": 8.104166588713707e-06, "loss": 0.0368, "step": 37319 }, { "epoch": 11.48, "learning_rate": 8.103677357329564e-06, "loss": 0.0822, "step": 37320 }, { "epoch": 11.48, "learning_rate": 8.103188130653478e-06, "loss": 0.0268, "step": 37321 }, { "epoch": 11.48, "learning_rate": 8.102698908686665e-06, "loss": 0.0726, "step": 37322 }, { "epoch": 11.48, "learning_rate": 8.10220969143034e-06, "loss": 0.0617, "step": 37323 }, { "epoch": 11.48, "learning_rate": 8.101720478885715e-06, "loss": 0.0465, "step": 37324 }, { "epoch": 11.48, "learning_rate": 8.101231271054009e-06, "loss": 0.0383, "step": 37325 }, { "epoch": 11.48, "learning_rate": 8.100742067936432e-06, "loss": 0.0462, "step": 37326 }, { "epoch": 11.49, "learning_rate": 8.100252869534203e-06, "loss": 0.0427, "step": 37327 }, { "epoch": 11.49, "learning_rate": 8.099763675848533e-06, "loss": 0.0355, "step": 37328 }, { "epoch": 11.49, "learning_rate": 8.099274486880631e-06, "loss": 0.037, "step": 37329 }, { "epoch": 11.49, "learning_rate": 8.098785302631726e-06, "loss": 0.0796, "step": 37330 }, { "epoch": 11.49, "learning_rate": 8.098296123103023e-06, "loss": 0.06, "step": 37331 }, { "epoch": 11.49, "learning_rate": 8.097806948295736e-06, "loss": 0.0605, "step": 37332 }, { "epoch": 11.49, "learning_rate": 8.097317778211084e-06, "loss": 0.0571, "step": 37333 }, { "epoch": 11.49, "learning_rate": 8.096828612850276e-06, "loss": 0.0275, "step": 37334 }, { "epoch": 11.49, "learning_rate": 8.09633945221453e-06, "loss": 0.0331, "step": 37335 }, { "epoch": 11.49, "learning_rate": 8.09585029630506e-06, "loss": 0.0508, "step": 37336 }, { "epoch": 11.49, "learning_rate": 8.095361145123084e-06, "loss": 0.0587, "step": 37337 }, { "epoch": 11.49, "learning_rate": 8.09487199866981e-06, "loss": 0.079, "step": 37338 }, { "epoch": 11.49, "learning_rate": 8.094382856946452e-06, "loss": 0.0517, "step": 37339 }, { "epoch": 11.49, "learning_rate": 8.093893719954234e-06, "loss": 0.0562, "step": 37340 }, { "epoch": 11.49, "learning_rate": 8.09340458769436e-06, "loss": 0.0476, "step": 37341 }, { "epoch": 11.49, "learning_rate": 8.092915460168051e-06, "loss": 0.0402, "step": 37342 }, { "epoch": 11.49, "learning_rate": 8.092426337376516e-06, "loss": 0.0393, "step": 37343 }, { "epoch": 11.49, "learning_rate": 8.091937219320973e-06, "loss": 0.0318, "step": 37344 }, { "epoch": 11.49, "learning_rate": 8.091448106002636e-06, "loss": 0.0626, "step": 37345 }, { "epoch": 11.49, "learning_rate": 8.090958997422718e-06, "loss": 0.0399, "step": 37346 }, { "epoch": 11.49, "learning_rate": 8.090469893582434e-06, "loss": 0.0691, "step": 37347 }, { "epoch": 11.49, "learning_rate": 8.089980794483002e-06, "loss": 0.0405, "step": 37348 }, { "epoch": 11.49, "learning_rate": 8.089491700125626e-06, "loss": 0.039, "step": 37349 }, { "epoch": 11.49, "learning_rate": 8.089002610511532e-06, "loss": 0.0596, "step": 37350 }, { "epoch": 11.49, "learning_rate": 8.08851352564193e-06, "loss": 0.0759, "step": 37351 }, { "epoch": 11.49, "learning_rate": 8.088024445518033e-06, "loss": 0.0608, "step": 37352 }, { "epoch": 11.49, "learning_rate": 8.087535370141054e-06, "loss": 0.0645, "step": 37353 }, { "epoch": 11.49, "learning_rate": 8.087046299512209e-06, "loss": 0.0485, "step": 37354 }, { "epoch": 11.49, "learning_rate": 8.086557233632712e-06, "loss": 0.043, "step": 37355 }, { "epoch": 11.49, "learning_rate": 8.086068172503778e-06, "loss": 0.0797, "step": 37356 }, { "epoch": 11.49, "learning_rate": 8.085579116126623e-06, "loss": 0.043, "step": 37357 }, { "epoch": 11.49, "learning_rate": 8.085090064502458e-06, "loss": 0.0246, "step": 37358 }, { "epoch": 11.5, "learning_rate": 8.084601017632495e-06, "loss": 0.0581, "step": 37359 }, { "epoch": 11.5, "learning_rate": 8.084111975517955e-06, "loss": 0.0465, "step": 37360 }, { "epoch": 11.5, "learning_rate": 8.08362293816005e-06, "loss": 0.0357, "step": 37361 }, { "epoch": 11.5, "learning_rate": 8.083133905559989e-06, "loss": 0.0278, "step": 37362 }, { "epoch": 11.5, "learning_rate": 8.08264487771899e-06, "loss": 0.0522, "step": 37363 }, { "epoch": 11.5, "learning_rate": 8.082155854638267e-06, "loss": 0.0338, "step": 37364 }, { "epoch": 11.5, "learning_rate": 8.081666836319034e-06, "loss": 0.0433, "step": 37365 }, { "epoch": 11.5, "learning_rate": 8.081177822762506e-06, "loss": 0.0437, "step": 37366 }, { "epoch": 11.5, "learning_rate": 8.080688813969896e-06, "loss": 0.0432, "step": 37367 }, { "epoch": 11.5, "learning_rate": 8.080199809942418e-06, "loss": 0.0336, "step": 37368 }, { "epoch": 11.5, "learning_rate": 8.079710810681285e-06, "loss": 0.0689, "step": 37369 }, { "epoch": 11.5, "learning_rate": 8.079221816187714e-06, "loss": 0.0722, "step": 37370 }, { "epoch": 11.5, "learning_rate": 8.078732826462917e-06, "loss": 0.0425, "step": 37371 }, { "epoch": 11.5, "learning_rate": 8.078243841508111e-06, "loss": 0.0244, "step": 37372 }, { "epoch": 11.5, "learning_rate": 8.077754861324506e-06, "loss": 0.0168, "step": 37373 }, { "epoch": 11.5, "learning_rate": 8.077265885913312e-06, "loss": 0.0692, "step": 37374 }, { "epoch": 11.5, "learning_rate": 8.076776915275756e-06, "loss": 0.0382, "step": 37375 }, { "epoch": 11.5, "learning_rate": 8.076287949413039e-06, "loss": 0.07, "step": 37376 }, { "epoch": 11.5, "learning_rate": 8.075798988326384e-06, "loss": 0.0436, "step": 37377 }, { "epoch": 11.5, "learning_rate": 8.075310032017e-06, "loss": 0.0354, "step": 37378 }, { "epoch": 11.5, "learning_rate": 8.0748210804861e-06, "loss": 0.1238, "step": 37379 }, { "epoch": 11.5, "learning_rate": 8.074332133734904e-06, "loss": 0.0572, "step": 37380 }, { "epoch": 11.5, "learning_rate": 8.07384319176462e-06, "loss": 0.0565, "step": 37381 }, { "epoch": 11.5, "learning_rate": 8.073354254576468e-06, "loss": 0.0451, "step": 37382 }, { "epoch": 11.5, "learning_rate": 8.072865322171654e-06, "loss": 0.0409, "step": 37383 }, { "epoch": 11.5, "learning_rate": 8.072376394551393e-06, "loss": 0.1014, "step": 37384 }, { "epoch": 11.5, "learning_rate": 8.071887471716908e-06, "loss": 0.049, "step": 37385 }, { "epoch": 11.5, "learning_rate": 8.071398553669405e-06, "loss": 0.0662, "step": 37386 }, { "epoch": 11.5, "learning_rate": 8.070909640410099e-06, "loss": 0.0552, "step": 37387 }, { "epoch": 11.5, "learning_rate": 8.070420731940202e-06, "loss": 0.048, "step": 37388 }, { "epoch": 11.5, "learning_rate": 8.06993182826093e-06, "loss": 0.0553, "step": 37389 }, { "epoch": 11.5, "learning_rate": 8.0694429293735e-06, "loss": 0.044, "step": 37390 }, { "epoch": 11.5, "learning_rate": 8.068954035279121e-06, "loss": 0.0447, "step": 37391 }, { "epoch": 11.51, "learning_rate": 8.06846514597901e-06, "loss": 0.0635, "step": 37392 }, { "epoch": 11.51, "learning_rate": 8.06797626147438e-06, "loss": 0.0512, "step": 37393 }, { "epoch": 11.51, "learning_rate": 8.06748738176644e-06, "loss": 0.0442, "step": 37394 }, { "epoch": 11.51, "learning_rate": 8.06699850685641e-06, "loss": 0.0704, "step": 37395 }, { "epoch": 11.51, "learning_rate": 8.066509636745505e-06, "loss": 0.0223, "step": 37396 }, { "epoch": 11.51, "learning_rate": 8.066020771434932e-06, "loss": 0.0637, "step": 37397 }, { "epoch": 11.51, "learning_rate": 8.065531910925907e-06, "loss": 0.0453, "step": 37398 }, { "epoch": 11.51, "learning_rate": 8.065043055219646e-06, "loss": 0.0567, "step": 37399 }, { "epoch": 11.51, "learning_rate": 8.06455420431736e-06, "loss": 0.0635, "step": 37400 }, { "epoch": 11.51, "learning_rate": 8.064065358220268e-06, "loss": 0.0502, "step": 37401 }, { "epoch": 11.51, "learning_rate": 8.063576516929577e-06, "loss": 0.0677, "step": 37402 }, { "epoch": 11.51, "learning_rate": 8.063087680446507e-06, "loss": 0.0883, "step": 37403 }, { "epoch": 11.51, "learning_rate": 8.062598848772261e-06, "loss": 0.0544, "step": 37404 }, { "epoch": 11.51, "learning_rate": 8.062110021908063e-06, "loss": 0.0659, "step": 37405 }, { "epoch": 11.51, "learning_rate": 8.061621199855127e-06, "loss": 0.056, "step": 37406 }, { "epoch": 11.51, "learning_rate": 8.061132382614661e-06, "loss": 0.0675, "step": 37407 }, { "epoch": 11.51, "learning_rate": 8.060643570187879e-06, "loss": 0.0626, "step": 37408 }, { "epoch": 11.51, "learning_rate": 8.060154762575994e-06, "loss": 0.0573, "step": 37409 }, { "epoch": 11.51, "learning_rate": 8.059665959780225e-06, "loss": 0.0492, "step": 37410 }, { "epoch": 11.51, "learning_rate": 8.059177161801783e-06, "loss": 0.0343, "step": 37411 }, { "epoch": 11.51, "learning_rate": 8.05868836864188e-06, "loss": 0.0343, "step": 37412 }, { "epoch": 11.51, "learning_rate": 8.058199580301731e-06, "loss": 0.0447, "step": 37413 }, { "epoch": 11.51, "learning_rate": 8.057710796782546e-06, "loss": 0.0213, "step": 37414 }, { "epoch": 11.51, "learning_rate": 8.057222018085544e-06, "loss": 0.0556, "step": 37415 }, { "epoch": 11.51, "learning_rate": 8.056733244211935e-06, "loss": 0.0845, "step": 37416 }, { "epoch": 11.51, "learning_rate": 8.056244475162938e-06, "loss": 0.0683, "step": 37417 }, { "epoch": 11.51, "learning_rate": 8.055755710939757e-06, "loss": 0.037, "step": 37418 }, { "epoch": 11.51, "learning_rate": 8.05526695154361e-06, "loss": 0.0294, "step": 37419 }, { "epoch": 11.51, "learning_rate": 8.054778196975714e-06, "loss": 0.0577, "step": 37420 }, { "epoch": 11.51, "learning_rate": 8.054289447237277e-06, "loss": 0.0234, "step": 37421 }, { "epoch": 11.51, "learning_rate": 8.053800702329515e-06, "loss": 0.0435, "step": 37422 }, { "epoch": 11.51, "learning_rate": 8.053311962253643e-06, "loss": 0.0551, "step": 37423 }, { "epoch": 11.52, "learning_rate": 8.052823227010869e-06, "loss": 0.0317, "step": 37424 }, { "epoch": 11.52, "learning_rate": 8.052334496602412e-06, "loss": 0.0194, "step": 37425 }, { "epoch": 11.52, "learning_rate": 8.051845771029483e-06, "loss": 0.0613, "step": 37426 }, { "epoch": 11.52, "learning_rate": 8.0513570502933e-06, "loss": 0.0639, "step": 37427 }, { "epoch": 11.52, "learning_rate": 8.050868334395067e-06, "loss": 0.0214, "step": 37428 }, { "epoch": 11.52, "learning_rate": 8.050379623336e-06, "loss": 0.0439, "step": 37429 }, { "epoch": 11.52, "learning_rate": 8.049890917117322e-06, "loss": 0.0598, "step": 37430 }, { "epoch": 11.52, "learning_rate": 8.049402215740237e-06, "loss": 0.0768, "step": 37431 }, { "epoch": 11.52, "learning_rate": 8.04891351920596e-06, "loss": 0.0432, "step": 37432 }, { "epoch": 11.52, "learning_rate": 8.048424827515704e-06, "loss": 0.0334, "step": 37433 }, { "epoch": 11.52, "learning_rate": 8.04793614067068e-06, "loss": 0.0644, "step": 37434 }, { "epoch": 11.52, "learning_rate": 8.047447458672107e-06, "loss": 0.1196, "step": 37435 }, { "epoch": 11.52, "learning_rate": 8.046958781521198e-06, "loss": 0.0754, "step": 37436 }, { "epoch": 11.52, "learning_rate": 8.04647010921916e-06, "loss": 0.0788, "step": 37437 }, { "epoch": 11.52, "learning_rate": 8.045981441767215e-06, "loss": 0.0306, "step": 37438 }, { "epoch": 11.52, "learning_rate": 8.045492779166564e-06, "loss": 0.0325, "step": 37439 }, { "epoch": 11.52, "learning_rate": 8.045004121418434e-06, "loss": 0.0793, "step": 37440 }, { "epoch": 11.52, "learning_rate": 8.044515468524028e-06, "loss": 0.045, "step": 37441 }, { "epoch": 11.52, "learning_rate": 8.044026820484565e-06, "loss": 0.0557, "step": 37442 }, { "epoch": 11.52, "learning_rate": 8.043538177301256e-06, "loss": 0.1321, "step": 37443 }, { "epoch": 11.52, "learning_rate": 8.04304953897531e-06, "loss": 0.0435, "step": 37444 }, { "epoch": 11.52, "learning_rate": 8.04256090550795e-06, "loss": 0.0616, "step": 37445 }, { "epoch": 11.52, "learning_rate": 8.042072276900382e-06, "loss": 0.0308, "step": 37446 }, { "epoch": 11.52, "learning_rate": 8.041583653153819e-06, "loss": 0.0407, "step": 37447 }, { "epoch": 11.52, "learning_rate": 8.04109503426948e-06, "loss": 0.0621, "step": 37448 }, { "epoch": 11.52, "learning_rate": 8.040606420248567e-06, "loss": 0.0437, "step": 37449 }, { "epoch": 11.52, "learning_rate": 8.040117811092305e-06, "loss": 0.1013, "step": 37450 }, { "epoch": 11.52, "learning_rate": 8.039629206801903e-06, "loss": 0.043, "step": 37451 }, { "epoch": 11.52, "learning_rate": 8.039140607378571e-06, "loss": 0.052, "step": 37452 }, { "epoch": 11.52, "learning_rate": 8.038652012823525e-06, "loss": 0.0536, "step": 37453 }, { "epoch": 11.52, "learning_rate": 8.038163423137974e-06, "loss": 0.0599, "step": 37454 }, { "epoch": 11.52, "learning_rate": 8.037674838323138e-06, "loss": 0.081, "step": 37455 }, { "epoch": 11.52, "learning_rate": 8.037186258380226e-06, "loss": 0.0599, "step": 37456 }, { "epoch": 11.53, "learning_rate": 8.036697683310451e-06, "loss": 0.0275, "step": 37457 }, { "epoch": 11.53, "learning_rate": 8.036209113115027e-06, "loss": 0.0352, "step": 37458 }, { "epoch": 11.53, "learning_rate": 8.035720547795163e-06, "loss": 0.0529, "step": 37459 }, { "epoch": 11.53, "learning_rate": 8.035231987352078e-06, "loss": 0.0593, "step": 37460 }, { "epoch": 11.53, "learning_rate": 8.034743431786985e-06, "loss": 0.055, "step": 37461 }, { "epoch": 11.53, "learning_rate": 8.034254881101091e-06, "loss": 0.0838, "step": 37462 }, { "epoch": 11.53, "learning_rate": 8.033766335295613e-06, "loss": 0.0282, "step": 37463 }, { "epoch": 11.53, "learning_rate": 8.03327779437176e-06, "loss": 0.0332, "step": 37464 }, { "epoch": 11.53, "learning_rate": 8.032789258330751e-06, "loss": 0.0453, "step": 37465 }, { "epoch": 11.53, "learning_rate": 8.032300727173797e-06, "loss": 0.0472, "step": 37466 }, { "epoch": 11.53, "learning_rate": 8.031812200902108e-06, "loss": 0.03, "step": 37467 }, { "epoch": 11.53, "learning_rate": 8.0313236795169e-06, "loss": 0.0847, "step": 37468 }, { "epoch": 11.53, "learning_rate": 8.030835163019382e-06, "loss": 0.0345, "step": 37469 }, { "epoch": 11.53, "learning_rate": 8.03034665141077e-06, "loss": 0.0333, "step": 37470 }, { "epoch": 11.53, "learning_rate": 8.029858144692278e-06, "loss": 0.0465, "step": 37471 }, { "epoch": 11.53, "learning_rate": 8.029369642865118e-06, "loss": 0.0674, "step": 37472 }, { "epoch": 11.53, "learning_rate": 8.028881145930502e-06, "loss": 0.0475, "step": 37473 }, { "epoch": 11.53, "learning_rate": 8.028392653889637e-06, "loss": 0.0354, "step": 37474 }, { "epoch": 11.53, "learning_rate": 8.027904166743749e-06, "loss": 0.0476, "step": 37475 }, { "epoch": 11.53, "learning_rate": 8.02741568449404e-06, "loss": 0.0472, "step": 37476 }, { "epoch": 11.53, "learning_rate": 8.026927207141726e-06, "loss": 0.0296, "step": 37477 }, { "epoch": 11.53, "learning_rate": 8.02643873468802e-06, "loss": 0.0646, "step": 37478 }, { "epoch": 11.53, "learning_rate": 8.025950267134133e-06, "loss": 0.0474, "step": 37479 }, { "epoch": 11.53, "learning_rate": 8.025461804481281e-06, "loss": 0.0309, "step": 37480 }, { "epoch": 11.53, "learning_rate": 8.024973346730675e-06, "loss": 0.0249, "step": 37481 }, { "epoch": 11.53, "learning_rate": 8.02448489388353e-06, "loss": 0.0549, "step": 37482 }, { "epoch": 11.53, "learning_rate": 8.023996445941053e-06, "loss": 0.071, "step": 37483 }, { "epoch": 11.53, "learning_rate": 8.023508002904456e-06, "loss": 0.0423, "step": 37484 }, { "epoch": 11.53, "learning_rate": 8.023019564774963e-06, "loss": 0.06, "step": 37485 }, { "epoch": 11.53, "learning_rate": 8.02253113155378e-06, "loss": 0.0593, "step": 37486 }, { "epoch": 11.53, "learning_rate": 8.022042703242115e-06, "loss": 0.035, "step": 37487 }, { "epoch": 11.53, "learning_rate": 8.021554279841185e-06, "loss": 0.0441, "step": 37488 }, { "epoch": 11.54, "learning_rate": 8.021065861352203e-06, "loss": 0.0452, "step": 37489 }, { "epoch": 11.54, "learning_rate": 8.020577447776381e-06, "loss": 0.0619, "step": 37490 }, { "epoch": 11.54, "learning_rate": 8.020089039114932e-06, "loss": 0.0733, "step": 37491 }, { "epoch": 11.54, "learning_rate": 8.019600635369068e-06, "loss": 0.0322, "step": 37492 }, { "epoch": 11.54, "learning_rate": 8.019112236540005e-06, "loss": 0.0713, "step": 37493 }, { "epoch": 11.54, "learning_rate": 8.018623842628946e-06, "loss": 0.0568, "step": 37494 }, { "epoch": 11.54, "learning_rate": 8.018135453637112e-06, "loss": 0.0539, "step": 37495 }, { "epoch": 11.54, "learning_rate": 8.017647069565718e-06, "loss": 0.0206, "step": 37496 }, { "epoch": 11.54, "learning_rate": 8.017158690415968e-06, "loss": 0.0747, "step": 37497 }, { "epoch": 11.54, "learning_rate": 8.016670316189078e-06, "loss": 0.0451, "step": 37498 }, { "epoch": 11.54, "learning_rate": 8.016181946886259e-06, "loss": 0.0391, "step": 37499 }, { "epoch": 11.54, "learning_rate": 8.015693582508728e-06, "loss": 0.0653, "step": 37500 }, { "epoch": 11.54, "learning_rate": 8.015205223057696e-06, "loss": 0.0351, "step": 37501 }, { "epoch": 11.54, "learning_rate": 8.014716868534372e-06, "loss": 0.0878, "step": 37502 }, { "epoch": 11.54, "learning_rate": 8.014228518939975e-06, "loss": 0.0666, "step": 37503 }, { "epoch": 11.54, "learning_rate": 8.013740174275706e-06, "loss": 0.0331, "step": 37504 }, { "epoch": 11.54, "learning_rate": 8.013251834542788e-06, "loss": 0.0553, "step": 37505 }, { "epoch": 11.54, "learning_rate": 8.012763499742435e-06, "loss": 0.0613, "step": 37506 }, { "epoch": 11.54, "learning_rate": 8.01227516987585e-06, "loss": 0.0459, "step": 37507 }, { "epoch": 11.54, "learning_rate": 8.01178684494425e-06, "loss": 0.0602, "step": 37508 }, { "epoch": 11.54, "learning_rate": 8.011298524948845e-06, "loss": 0.036, "step": 37509 }, { "epoch": 11.54, "learning_rate": 8.010810209890853e-06, "loss": 0.0217, "step": 37510 }, { "epoch": 11.54, "learning_rate": 8.010321899771483e-06, "loss": 0.0472, "step": 37511 }, { "epoch": 11.54, "learning_rate": 8.009833594591948e-06, "loss": 0.0457, "step": 37512 }, { "epoch": 11.54, "learning_rate": 8.009345294353457e-06, "loss": 0.0535, "step": 37513 }, { "epoch": 11.54, "learning_rate": 8.008856999057225e-06, "loss": 0.0461, "step": 37514 }, { "epoch": 11.54, "learning_rate": 8.008368708704464e-06, "loss": 0.0493, "step": 37515 }, { "epoch": 11.54, "learning_rate": 8.00788042329639e-06, "loss": 0.0445, "step": 37516 }, { "epoch": 11.54, "learning_rate": 8.00739214283421e-06, "loss": 0.0744, "step": 37517 }, { "epoch": 11.54, "learning_rate": 8.006903867319138e-06, "loss": 0.0317, "step": 37518 }, { "epoch": 11.54, "learning_rate": 8.006415596752387e-06, "loss": 0.0681, "step": 37519 }, { "epoch": 11.54, "learning_rate": 8.005927331135167e-06, "loss": 0.072, "step": 37520 }, { "epoch": 11.54, "learning_rate": 8.005439070468692e-06, "loss": 0.0285, "step": 37521 }, { "epoch": 11.55, "learning_rate": 8.004950814754175e-06, "loss": 0.0363, "step": 37522 }, { "epoch": 11.55, "learning_rate": 8.004462563992828e-06, "loss": 0.0556, "step": 37523 }, { "epoch": 11.55, "learning_rate": 8.00397431818586e-06, "loss": 0.048, "step": 37524 }, { "epoch": 11.55, "learning_rate": 8.003486077334486e-06, "loss": 0.0537, "step": 37525 }, { "epoch": 11.55, "learning_rate": 8.00299784143992e-06, "loss": 0.033, "step": 37526 }, { "epoch": 11.55, "learning_rate": 8.002509610503374e-06, "loss": 0.0561, "step": 37527 }, { "epoch": 11.55, "learning_rate": 8.002021384526056e-06, "loss": 0.0435, "step": 37528 }, { "epoch": 11.55, "learning_rate": 8.001533163509175e-06, "loss": 0.0492, "step": 37529 }, { "epoch": 11.55, "learning_rate": 8.001044947453955e-06, "loss": 0.0602, "step": 37530 }, { "epoch": 11.55, "learning_rate": 8.0005567363616e-06, "loss": 0.0246, "step": 37531 }, { "epoch": 11.55, "learning_rate": 8.000068530233322e-06, "loss": 0.0459, "step": 37532 }, { "epoch": 11.55, "learning_rate": 7.999580329070337e-06, "loss": 0.0918, "step": 37533 }, { "epoch": 11.55, "learning_rate": 7.999092132873851e-06, "loss": 0.0619, "step": 37534 }, { "epoch": 11.55, "learning_rate": 7.998603941645082e-06, "loss": 0.0478, "step": 37535 }, { "epoch": 11.55, "learning_rate": 7.998115755385241e-06, "loss": 0.0742, "step": 37536 }, { "epoch": 11.55, "learning_rate": 7.997627574095538e-06, "loss": 0.0471, "step": 37537 }, { "epoch": 11.55, "learning_rate": 7.997139397777187e-06, "loss": 0.0553, "step": 37538 }, { "epoch": 11.55, "learning_rate": 7.996651226431394e-06, "loss": 0.0607, "step": 37539 }, { "epoch": 11.55, "learning_rate": 7.996163060059381e-06, "loss": 0.0537, "step": 37540 }, { "epoch": 11.55, "learning_rate": 7.995674898662356e-06, "loss": 0.043, "step": 37541 }, { "epoch": 11.55, "learning_rate": 7.995186742241527e-06, "loss": 0.0487, "step": 37542 }, { "epoch": 11.55, "learning_rate": 7.99469859079811e-06, "loss": 0.0227, "step": 37543 }, { "epoch": 11.55, "learning_rate": 7.994210444333312e-06, "loss": 0.0856, "step": 37544 }, { "epoch": 11.55, "learning_rate": 7.993722302848353e-06, "loss": 0.0233, "step": 37545 }, { "epoch": 11.55, "learning_rate": 7.99323416634444e-06, "loss": 0.0557, "step": 37546 }, { "epoch": 11.55, "learning_rate": 7.992746034822783e-06, "loss": 0.0349, "step": 37547 }, { "epoch": 11.55, "learning_rate": 7.992257908284602e-06, "loss": 0.0179, "step": 37548 }, { "epoch": 11.55, "learning_rate": 7.991769786731095e-06, "loss": 0.0495, "step": 37549 }, { "epoch": 11.55, "learning_rate": 7.991281670163488e-06, "loss": 0.0476, "step": 37550 }, { "epoch": 11.55, "learning_rate": 7.990793558582988e-06, "loss": 0.0707, "step": 37551 }, { "epoch": 11.55, "learning_rate": 7.990305451990805e-06, "loss": 0.0438, "step": 37552 }, { "epoch": 11.55, "learning_rate": 7.989817350388151e-06, "loss": 0.0315, "step": 37553 }, { "epoch": 11.56, "learning_rate": 7.989329253776235e-06, "loss": 0.0502, "step": 37554 }, { "epoch": 11.56, "learning_rate": 7.988841162156276e-06, "loss": 0.0625, "step": 37555 }, { "epoch": 11.56, "learning_rate": 7.988353075529482e-06, "loss": 0.0661, "step": 37556 }, { "epoch": 11.56, "learning_rate": 7.987864993897066e-06, "loss": 0.0599, "step": 37557 }, { "epoch": 11.56, "learning_rate": 7.987376917260237e-06, "loss": 0.0296, "step": 37558 }, { "epoch": 11.56, "learning_rate": 7.986888845620208e-06, "loss": 0.0743, "step": 37559 }, { "epoch": 11.56, "learning_rate": 7.986400778978192e-06, "loss": 0.0463, "step": 37560 }, { "epoch": 11.56, "learning_rate": 7.985912717335405e-06, "loss": 0.0386, "step": 37561 }, { "epoch": 11.56, "learning_rate": 7.985424660693048e-06, "loss": 0.0459, "step": 37562 }, { "epoch": 11.56, "learning_rate": 7.984936609052341e-06, "loss": 0.102, "step": 37563 }, { "epoch": 11.56, "learning_rate": 7.984448562414489e-06, "loss": 0.071, "step": 37564 }, { "epoch": 11.56, "learning_rate": 7.983960520780712e-06, "loss": 0.0454, "step": 37565 }, { "epoch": 11.56, "learning_rate": 7.983472484152216e-06, "loss": 0.0623, "step": 37566 }, { "epoch": 11.56, "learning_rate": 7.982984452530214e-06, "loss": 0.0382, "step": 37567 }, { "epoch": 11.56, "learning_rate": 7.982496425915919e-06, "loss": 0.0555, "step": 37568 }, { "epoch": 11.56, "learning_rate": 7.982008404310538e-06, "loss": 0.0641, "step": 37569 }, { "epoch": 11.56, "learning_rate": 7.981520387715289e-06, "loss": 0.0286, "step": 37570 }, { "epoch": 11.56, "learning_rate": 7.981032376131382e-06, "loss": 0.0321, "step": 37571 }, { "epoch": 11.56, "learning_rate": 7.980544369560029e-06, "loss": 0.0755, "step": 37572 }, { "epoch": 11.56, "learning_rate": 7.980056368002435e-06, "loss": 0.0193, "step": 37573 }, { "epoch": 11.56, "learning_rate": 7.979568371459816e-06, "loss": 0.0402, "step": 37574 }, { "epoch": 11.56, "learning_rate": 7.979080379933388e-06, "loss": 0.0448, "step": 37575 }, { "epoch": 11.56, "learning_rate": 7.978592393424356e-06, "loss": 0.0344, "step": 37576 }, { "epoch": 11.56, "learning_rate": 7.978104411933935e-06, "loss": 0.068, "step": 37577 }, { "epoch": 11.56, "learning_rate": 7.977616435463336e-06, "loss": 0.0371, "step": 37578 }, { "epoch": 11.56, "learning_rate": 7.977128464013768e-06, "loss": 0.0446, "step": 37579 }, { "epoch": 11.56, "learning_rate": 7.976640497586447e-06, "loss": 0.0528, "step": 37580 }, { "epoch": 11.56, "learning_rate": 7.976152536182583e-06, "loss": 0.075, "step": 37581 }, { "epoch": 11.56, "learning_rate": 7.975664579803387e-06, "loss": 0.0361, "step": 37582 }, { "epoch": 11.56, "learning_rate": 7.975176628450068e-06, "loss": 0.0478, "step": 37583 }, { "epoch": 11.56, "learning_rate": 7.974688682123837e-06, "loss": 0.0615, "step": 37584 }, { "epoch": 11.56, "learning_rate": 7.974200740825913e-06, "loss": 0.0416, "step": 37585 }, { "epoch": 11.56, "learning_rate": 7.9737128045575e-06, "loss": 0.064, "step": 37586 }, { "epoch": 11.57, "learning_rate": 7.973224873319813e-06, "loss": 0.0508, "step": 37587 }, { "epoch": 11.57, "learning_rate": 7.972736947114062e-06, "loss": 0.0537, "step": 37588 }, { "epoch": 11.57, "learning_rate": 7.972249025941455e-06, "loss": 0.0314, "step": 37589 }, { "epoch": 11.57, "learning_rate": 7.97176110980321e-06, "loss": 0.0391, "step": 37590 }, { "epoch": 11.57, "learning_rate": 7.971273198700537e-06, "loss": 0.0591, "step": 37591 }, { "epoch": 11.57, "learning_rate": 7.970785292634644e-06, "loss": 0.0665, "step": 37592 }, { "epoch": 11.57, "learning_rate": 7.970297391606748e-06, "loss": 0.0512, "step": 37593 }, { "epoch": 11.57, "learning_rate": 7.969809495618047e-06, "loss": 0.1163, "step": 37594 }, { "epoch": 11.57, "learning_rate": 7.969321604669768e-06, "loss": 0.0302, "step": 37595 }, { "epoch": 11.57, "learning_rate": 7.968833718763117e-06, "loss": 0.0662, "step": 37596 }, { "epoch": 11.57, "learning_rate": 7.968345837899302e-06, "loss": 0.0563, "step": 37597 }, { "epoch": 11.57, "learning_rate": 7.967857962079538e-06, "loss": 0.0651, "step": 37598 }, { "epoch": 11.57, "learning_rate": 7.967370091305033e-06, "loss": 0.0234, "step": 37599 }, { "epoch": 11.57, "learning_rate": 7.966882225577e-06, "loss": 0.0439, "step": 37600 }, { "epoch": 11.57, "learning_rate": 7.966394364896653e-06, "loss": 0.053, "step": 37601 }, { "epoch": 11.57, "learning_rate": 7.9659065092652e-06, "loss": 0.0409, "step": 37602 }, { "epoch": 11.57, "learning_rate": 7.965418658683853e-06, "loss": 0.0753, "step": 37603 }, { "epoch": 11.57, "learning_rate": 7.964930813153817e-06, "loss": 0.0405, "step": 37604 }, { "epoch": 11.57, "learning_rate": 7.964442972676313e-06, "loss": 0.0395, "step": 37605 }, { "epoch": 11.57, "learning_rate": 7.963955137252552e-06, "loss": 0.0299, "step": 37606 }, { "epoch": 11.57, "learning_rate": 7.963467306883739e-06, "loss": 0.044, "step": 37607 }, { "epoch": 11.57, "learning_rate": 7.96297948157109e-06, "loss": 0.0435, "step": 37608 }, { "epoch": 11.57, "learning_rate": 7.962491661315807e-06, "loss": 0.0333, "step": 37609 }, { "epoch": 11.57, "learning_rate": 7.962003846119113e-06, "loss": 0.0539, "step": 37610 }, { "epoch": 11.57, "learning_rate": 7.961516035982212e-06, "loss": 0.028, "step": 37611 }, { "epoch": 11.57, "learning_rate": 7.96102823090632e-06, "loss": 0.055, "step": 37612 }, { "epoch": 11.57, "learning_rate": 7.960540430892642e-06, "loss": 0.0959, "step": 37613 }, { "epoch": 11.57, "learning_rate": 7.960052635942393e-06, "loss": 0.0207, "step": 37614 }, { "epoch": 11.57, "learning_rate": 7.959564846056785e-06, "loss": 0.0472, "step": 37615 }, { "epoch": 11.57, "learning_rate": 7.959077061237026e-06, "loss": 0.0344, "step": 37616 }, { "epoch": 11.57, "learning_rate": 7.95858928148433e-06, "loss": 0.0655, "step": 37617 }, { "epoch": 11.57, "learning_rate": 7.958101506799906e-06, "loss": 0.1247, "step": 37618 }, { "epoch": 11.58, "learning_rate": 7.957613737184963e-06, "loss": 0.0552, "step": 37619 }, { "epoch": 11.58, "learning_rate": 7.957125972640717e-06, "loss": 0.0186, "step": 37620 }, { "epoch": 11.58, "learning_rate": 7.956638213168376e-06, "loss": 0.065, "step": 37621 }, { "epoch": 11.58, "learning_rate": 7.956150458769151e-06, "loss": 0.0538, "step": 37622 }, { "epoch": 11.58, "learning_rate": 7.955662709444254e-06, "loss": 0.03, "step": 37623 }, { "epoch": 11.58, "learning_rate": 7.955174965194892e-06, "loss": 0.0318, "step": 37624 }, { "epoch": 11.58, "learning_rate": 7.954687226022285e-06, "loss": 0.0696, "step": 37625 }, { "epoch": 11.58, "learning_rate": 7.954199491927635e-06, "loss": 0.0335, "step": 37626 }, { "epoch": 11.58, "learning_rate": 7.953711762912159e-06, "loss": 0.0345, "step": 37627 }, { "epoch": 11.58, "learning_rate": 7.953224038977062e-06, "loss": 0.057, "step": 37628 }, { "epoch": 11.58, "learning_rate": 7.952736320123556e-06, "loss": 0.0435, "step": 37629 }, { "epoch": 11.58, "learning_rate": 7.952248606352856e-06, "loss": 0.0746, "step": 37630 }, { "epoch": 11.58, "learning_rate": 7.951760897666173e-06, "loss": 0.0461, "step": 37631 }, { "epoch": 11.58, "learning_rate": 7.951273194064714e-06, "loss": 0.0394, "step": 37632 }, { "epoch": 11.58, "learning_rate": 7.95078549554969e-06, "loss": 0.0807, "step": 37633 }, { "epoch": 11.58, "learning_rate": 7.950297802122313e-06, "loss": 0.0369, "step": 37634 }, { "epoch": 11.58, "learning_rate": 7.949810113783795e-06, "loss": 0.0399, "step": 37635 }, { "epoch": 11.58, "learning_rate": 7.949322430535345e-06, "loss": 0.0666, "step": 37636 }, { "epoch": 11.58, "learning_rate": 7.948834752378176e-06, "loss": 0.0557, "step": 37637 }, { "epoch": 11.58, "learning_rate": 7.948347079313494e-06, "loss": 0.0571, "step": 37638 }, { "epoch": 11.58, "learning_rate": 7.947859411342516e-06, "loss": 0.0586, "step": 37639 }, { "epoch": 11.58, "learning_rate": 7.947371748466453e-06, "loss": 0.0189, "step": 37640 }, { "epoch": 11.58, "learning_rate": 7.946884090686508e-06, "loss": 0.0338, "step": 37641 }, { "epoch": 11.58, "learning_rate": 7.946396438003898e-06, "loss": 0.0383, "step": 37642 }, { "epoch": 11.58, "learning_rate": 7.945908790419831e-06, "loss": 0.0375, "step": 37643 }, { "epoch": 11.58, "learning_rate": 7.94542114793552e-06, "loss": 0.0618, "step": 37644 }, { "epoch": 11.58, "learning_rate": 7.944933510552173e-06, "loss": 0.0595, "step": 37645 }, { "epoch": 11.58, "learning_rate": 7.944445878271002e-06, "loss": 0.0622, "step": 37646 }, { "epoch": 11.58, "learning_rate": 7.94395825109322e-06, "loss": 0.0646, "step": 37647 }, { "epoch": 11.58, "learning_rate": 7.943470629020032e-06, "loss": 0.0373, "step": 37648 }, { "epoch": 11.58, "learning_rate": 7.942983012052652e-06, "loss": 0.0567, "step": 37649 }, { "epoch": 11.58, "learning_rate": 7.942495400192293e-06, "loss": 0.0531, "step": 37650 }, { "epoch": 11.58, "learning_rate": 7.942007793440165e-06, "loss": 0.0477, "step": 37651 }, { "epoch": 11.59, "learning_rate": 7.941520191797475e-06, "loss": 0.0527, "step": 37652 }, { "epoch": 11.59, "learning_rate": 7.941032595265431e-06, "loss": 0.0594, "step": 37653 }, { "epoch": 11.59, "learning_rate": 7.940545003845253e-06, "loss": 0.0736, "step": 37654 }, { "epoch": 11.59, "learning_rate": 7.940057417538147e-06, "loss": 0.059, "step": 37655 }, { "epoch": 11.59, "learning_rate": 7.93956983634532e-06, "loss": 0.0379, "step": 37656 }, { "epoch": 11.59, "learning_rate": 7.939082260267989e-06, "loss": 0.045, "step": 37657 }, { "epoch": 11.59, "learning_rate": 7.938594689307358e-06, "loss": 0.0682, "step": 37658 }, { "epoch": 11.59, "learning_rate": 7.93810712346464e-06, "loss": 0.028, "step": 37659 }, { "epoch": 11.59, "learning_rate": 7.937619562741048e-06, "loss": 0.0723, "step": 37660 }, { "epoch": 11.59, "learning_rate": 7.937132007137794e-06, "loss": 0.0441, "step": 37661 }, { "epoch": 11.59, "learning_rate": 7.936644456656082e-06, "loss": 0.0423, "step": 37662 }, { "epoch": 11.59, "learning_rate": 7.936156911297122e-06, "loss": 0.0687, "step": 37663 }, { "epoch": 11.59, "learning_rate": 7.935669371062132e-06, "loss": 0.0425, "step": 37664 }, { "epoch": 11.59, "learning_rate": 7.93518183595232e-06, "loss": 0.0817, "step": 37665 }, { "epoch": 11.59, "learning_rate": 7.934694305968893e-06, "loss": 0.0729, "step": 37666 }, { "epoch": 11.59, "learning_rate": 7.934206781113061e-06, "loss": 0.0828, "step": 37667 }, { "epoch": 11.59, "learning_rate": 7.933719261386036e-06, "loss": 0.0237, "step": 37668 }, { "epoch": 11.59, "learning_rate": 7.933231746789033e-06, "loss": 0.0445, "step": 37669 }, { "epoch": 11.59, "learning_rate": 7.932744237323256e-06, "loss": 0.0268, "step": 37670 }, { "epoch": 11.59, "learning_rate": 7.932256732989919e-06, "loss": 0.0662, "step": 37671 }, { "epoch": 11.59, "learning_rate": 7.931769233790234e-06, "loss": 0.0525, "step": 37672 }, { "epoch": 11.59, "learning_rate": 7.9312817397254e-06, "loss": 0.0409, "step": 37673 }, { "epoch": 11.59, "learning_rate": 7.930794250796643e-06, "loss": 0.0747, "step": 37674 }, { "epoch": 11.59, "learning_rate": 7.930306767005165e-06, "loss": 0.0351, "step": 37675 }, { "epoch": 11.59, "learning_rate": 7.929819288352176e-06, "loss": 0.0563, "step": 37676 }, { "epoch": 11.59, "learning_rate": 7.929331814838889e-06, "loss": 0.0294, "step": 37677 }, { "epoch": 11.59, "learning_rate": 7.92884434646651e-06, "loss": 0.0302, "step": 37678 }, { "epoch": 11.59, "learning_rate": 7.928356883236252e-06, "loss": 0.0691, "step": 37679 }, { "epoch": 11.59, "learning_rate": 7.927869425149329e-06, "loss": 0.0732, "step": 37680 }, { "epoch": 11.59, "learning_rate": 7.927381972206945e-06, "loss": 0.0507, "step": 37681 }, { "epoch": 11.59, "learning_rate": 7.926894524410316e-06, "loss": 0.0471, "step": 37682 }, { "epoch": 11.59, "learning_rate": 7.926407081760643e-06, "loss": 0.0853, "step": 37683 }, { "epoch": 11.6, "learning_rate": 7.925919644259145e-06, "loss": 0.0265, "step": 37684 }, { "epoch": 11.6, "learning_rate": 7.925432211907033e-06, "loss": 0.0576, "step": 37685 }, { "epoch": 11.6, "learning_rate": 7.92494478470551e-06, "loss": 0.074, "step": 37686 }, { "epoch": 11.6, "learning_rate": 7.92445736265579e-06, "loss": 0.0514, "step": 37687 }, { "epoch": 11.6, "learning_rate": 7.923969945759082e-06, "loss": 0.048, "step": 37688 }, { "epoch": 11.6, "learning_rate": 7.9234825340166e-06, "loss": 0.0322, "step": 37689 }, { "epoch": 11.6, "learning_rate": 7.922995127429547e-06, "loss": 0.0328, "step": 37690 }, { "epoch": 11.6, "learning_rate": 7.92250772599914e-06, "loss": 0.0326, "step": 37691 }, { "epoch": 11.6, "learning_rate": 7.922020329726587e-06, "loss": 0.054, "step": 37692 }, { "epoch": 11.6, "learning_rate": 7.921532938613094e-06, "loss": 0.0465, "step": 37693 }, { "epoch": 11.6, "learning_rate": 7.921045552659876e-06, "loss": 0.0352, "step": 37694 }, { "epoch": 11.6, "learning_rate": 7.920558171868142e-06, "loss": 0.0586, "step": 37695 }, { "epoch": 11.6, "learning_rate": 7.920070796239104e-06, "loss": 0.045, "step": 37696 }, { "epoch": 11.6, "learning_rate": 7.919583425773965e-06, "loss": 0.0714, "step": 37697 }, { "epoch": 11.6, "learning_rate": 7.91909606047394e-06, "loss": 0.0492, "step": 37698 }, { "epoch": 11.6, "learning_rate": 7.91860870034024e-06, "loss": 0.0344, "step": 37699 }, { "epoch": 11.6, "learning_rate": 7.918121345374073e-06, "loss": 0.048, "step": 37700 }, { "epoch": 11.6, "learning_rate": 7.91763399557665e-06, "loss": 0.0557, "step": 37701 }, { "epoch": 11.6, "learning_rate": 7.91714665094918e-06, "loss": 0.0635, "step": 37702 }, { "epoch": 11.6, "learning_rate": 7.916659311492871e-06, "loss": 0.0722, "step": 37703 }, { "epoch": 11.6, "learning_rate": 7.916171977208938e-06, "loss": 0.033, "step": 37704 }, { "epoch": 11.6, "learning_rate": 7.915684648098587e-06, "loss": 0.0491, "step": 37705 }, { "epoch": 11.6, "learning_rate": 7.915197324163033e-06, "loss": 0.0293, "step": 37706 }, { "epoch": 11.6, "learning_rate": 7.914710005403479e-06, "loss": 0.0552, "step": 37707 }, { "epoch": 11.6, "learning_rate": 7.914222691821134e-06, "loss": 0.1035, "step": 37708 }, { "epoch": 11.6, "learning_rate": 7.913735383417217e-06, "loss": 0.0507, "step": 37709 }, { "epoch": 11.6, "learning_rate": 7.913248080192931e-06, "loss": 0.0556, "step": 37710 }, { "epoch": 11.6, "learning_rate": 7.912760782149489e-06, "loss": 0.0598, "step": 37711 }, { "epoch": 11.6, "learning_rate": 7.912273489288097e-06, "loss": 0.0483, "step": 37712 }, { "epoch": 11.6, "learning_rate": 7.911786201609965e-06, "loss": 0.0597, "step": 37713 }, { "epoch": 11.6, "learning_rate": 7.91129891911631e-06, "loss": 0.0509, "step": 37714 }, { "epoch": 11.6, "learning_rate": 7.910811641808334e-06, "loss": 0.0478, "step": 37715 }, { "epoch": 11.6, "learning_rate": 7.91032436968725e-06, "loss": 0.0118, "step": 37716 }, { "epoch": 11.61, "learning_rate": 7.90983710275427e-06, "loss": 0.047, "step": 37717 }, { "epoch": 11.61, "learning_rate": 7.909349841010594e-06, "loss": 0.0527, "step": 37718 }, { "epoch": 11.61, "learning_rate": 7.908862584457444e-06, "loss": 0.0528, "step": 37719 }, { "epoch": 11.61, "learning_rate": 7.908375333096024e-06, "loss": 0.0721, "step": 37720 }, { "epoch": 11.61, "learning_rate": 7.907888086927544e-06, "loss": 0.0746, "step": 37721 }, { "epoch": 11.61, "learning_rate": 7.907400845953214e-06, "loss": 0.1122, "step": 37722 }, { "epoch": 11.61, "learning_rate": 7.90691361017424e-06, "loss": 0.031, "step": 37723 }, { "epoch": 11.61, "learning_rate": 7.90642637959184e-06, "loss": 0.0693, "step": 37724 }, { "epoch": 11.61, "learning_rate": 7.905939154207217e-06, "loss": 0.0642, "step": 37725 }, { "epoch": 11.61, "learning_rate": 7.905451934021581e-06, "loss": 0.0943, "step": 37726 }, { "epoch": 11.61, "learning_rate": 7.904964719036148e-06, "loss": 0.0445, "step": 37727 }, { "epoch": 11.61, "learning_rate": 7.904477509252116e-06, "loss": 0.0435, "step": 37728 }, { "epoch": 11.61, "learning_rate": 7.903990304670704e-06, "loss": 0.0797, "step": 37729 }, { "epoch": 11.61, "learning_rate": 7.903503105293122e-06, "loss": 0.0562, "step": 37730 }, { "epoch": 11.61, "learning_rate": 7.903015911120574e-06, "loss": 0.0477, "step": 37731 }, { "epoch": 11.61, "learning_rate": 7.902528722154274e-06, "loss": 0.0599, "step": 37732 }, { "epoch": 11.61, "learning_rate": 7.902041538395426e-06, "loss": 0.0345, "step": 37733 }, { "epoch": 11.61, "learning_rate": 7.901554359845246e-06, "loss": 0.0837, "step": 37734 }, { "epoch": 11.61, "learning_rate": 7.90106718650494e-06, "loss": 0.0548, "step": 37735 }, { "epoch": 11.61, "learning_rate": 7.90058001837572e-06, "loss": 0.0637, "step": 37736 }, { "epoch": 11.61, "learning_rate": 7.900092855458795e-06, "loss": 0.0618, "step": 37737 }, { "epoch": 11.61, "learning_rate": 7.899605697755367e-06, "loss": 0.0449, "step": 37738 }, { "epoch": 11.61, "learning_rate": 7.899118545266655e-06, "loss": 0.0514, "step": 37739 }, { "epoch": 11.61, "learning_rate": 7.898631397993869e-06, "loss": 0.0559, "step": 37740 }, { "epoch": 11.61, "learning_rate": 7.898144255938212e-06, "loss": 0.0482, "step": 37741 }, { "epoch": 11.61, "learning_rate": 7.897657119100896e-06, "loss": 0.0494, "step": 37742 }, { "epoch": 11.61, "learning_rate": 7.897169987483129e-06, "loss": 0.0252, "step": 37743 }, { "epoch": 11.61, "learning_rate": 7.896682861086124e-06, "loss": 0.0924, "step": 37744 }, { "epoch": 11.61, "learning_rate": 7.896195739911088e-06, "loss": 0.0583, "step": 37745 }, { "epoch": 11.61, "learning_rate": 7.895708623959231e-06, "loss": 0.0502, "step": 37746 }, { "epoch": 11.61, "learning_rate": 7.895221513231762e-06, "loss": 0.0741, "step": 37747 }, { "epoch": 11.61, "learning_rate": 7.894734407729887e-06, "loss": 0.0509, "step": 37748 }, { "epoch": 11.62, "learning_rate": 7.894247307454823e-06, "loss": 0.0998, "step": 37749 }, { "epoch": 11.62, "learning_rate": 7.893760212407775e-06, "loss": 0.0304, "step": 37750 }, { "epoch": 11.62, "learning_rate": 7.893273122589954e-06, "loss": 0.0252, "step": 37751 }, { "epoch": 11.62, "learning_rate": 7.892786038002567e-06, "loss": 0.0462, "step": 37752 }, { "epoch": 11.62, "learning_rate": 7.89229895864682e-06, "loss": 0.0362, "step": 37753 }, { "epoch": 11.62, "learning_rate": 7.891811884523933e-06, "loss": 0.0492, "step": 37754 }, { "epoch": 11.62, "learning_rate": 7.891324815635104e-06, "loss": 0.0821, "step": 37755 }, { "epoch": 11.62, "learning_rate": 7.89083775198155e-06, "loss": 0.0427, "step": 37756 }, { "epoch": 11.62, "learning_rate": 7.890350693564475e-06, "loss": 0.0477, "step": 37757 }, { "epoch": 11.62, "learning_rate": 7.88986364038509e-06, "loss": 0.0362, "step": 37758 }, { "epoch": 11.62, "learning_rate": 7.889376592444605e-06, "loss": 0.053, "step": 37759 }, { "epoch": 11.62, "learning_rate": 7.88888954974423e-06, "loss": 0.0337, "step": 37760 }, { "epoch": 11.62, "learning_rate": 7.888402512285175e-06, "loss": 0.0505, "step": 37761 }, { "epoch": 11.62, "learning_rate": 7.887915480068646e-06, "loss": 0.0783, "step": 37762 }, { "epoch": 11.62, "learning_rate": 7.887428453095848e-06, "loss": 0.0239, "step": 37763 }, { "epoch": 11.62, "learning_rate": 7.886941431368003e-06, "loss": 0.0736, "step": 37764 }, { "epoch": 11.62, "learning_rate": 7.886454414886311e-06, "loss": 0.0769, "step": 37765 }, { "epoch": 11.62, "learning_rate": 7.885967403651982e-06, "loss": 0.072, "step": 37766 }, { "epoch": 11.62, "learning_rate": 7.885480397666225e-06, "loss": 0.0559, "step": 37767 }, { "epoch": 11.62, "learning_rate": 7.88499339693025e-06, "loss": 0.0834, "step": 37768 }, { "epoch": 11.62, "learning_rate": 7.884506401445267e-06, "loss": 0.0372, "step": 37769 }, { "epoch": 11.62, "learning_rate": 7.884019411212486e-06, "loss": 0.0558, "step": 37770 }, { "epoch": 11.62, "learning_rate": 7.883532426233112e-06, "loss": 0.0545, "step": 37771 }, { "epoch": 11.62, "learning_rate": 7.88304544650836e-06, "loss": 0.0353, "step": 37772 }, { "epoch": 11.62, "learning_rate": 7.882558472039428e-06, "loss": 0.0334, "step": 37773 }, { "epoch": 11.62, "learning_rate": 7.882071502827539e-06, "loss": 0.0367, "step": 37774 }, { "epoch": 11.62, "learning_rate": 7.881584538873895e-06, "loss": 0.0495, "step": 37775 }, { "epoch": 11.62, "learning_rate": 7.881097580179705e-06, "loss": 0.0444, "step": 37776 }, { "epoch": 11.62, "learning_rate": 7.880610626746177e-06, "loss": 0.0372, "step": 37777 }, { "epoch": 11.62, "learning_rate": 7.88012367857452e-06, "loss": 0.0577, "step": 37778 }, { "epoch": 11.62, "learning_rate": 7.879636735665948e-06, "loss": 0.0564, "step": 37779 }, { "epoch": 11.62, "learning_rate": 7.879149798021666e-06, "loss": 0.0369, "step": 37780 }, { "epoch": 11.62, "learning_rate": 7.87866286564288e-06, "loss": 0.0366, "step": 37781 }, { "epoch": 11.63, "learning_rate": 7.878175938530808e-06, "loss": 0.0232, "step": 37782 }, { "epoch": 11.63, "learning_rate": 7.877689016686645e-06, "loss": 0.0795, "step": 37783 }, { "epoch": 11.63, "learning_rate": 7.877202100111612e-06, "loss": 0.0434, "step": 37784 }, { "epoch": 11.63, "learning_rate": 7.876715188806918e-06, "loss": 0.0362, "step": 37785 }, { "epoch": 11.63, "learning_rate": 7.876228282773765e-06, "loss": 0.0499, "step": 37786 }, { "epoch": 11.63, "learning_rate": 7.875741382013364e-06, "loss": 0.0492, "step": 37787 }, { "epoch": 11.63, "learning_rate": 7.875254486526922e-06, "loss": 0.0652, "step": 37788 }, { "epoch": 11.63, "learning_rate": 7.874767596315653e-06, "loss": 0.0251, "step": 37789 }, { "epoch": 11.63, "learning_rate": 7.874280711380762e-06, "loss": 0.0783, "step": 37790 }, { "epoch": 11.63, "learning_rate": 7.87379383172346e-06, "loss": 0.0511, "step": 37791 }, { "epoch": 11.63, "learning_rate": 7.873306957344954e-06, "loss": 0.0413, "step": 37792 }, { "epoch": 11.63, "learning_rate": 7.872820088246452e-06, "loss": 0.044, "step": 37793 }, { "epoch": 11.63, "learning_rate": 7.872333224429166e-06, "loss": 0.0923, "step": 37794 }, { "epoch": 11.63, "learning_rate": 7.871846365894305e-06, "loss": 0.023, "step": 37795 }, { "epoch": 11.63, "learning_rate": 7.871359512643076e-06, "loss": 0.1006, "step": 37796 }, { "epoch": 11.63, "learning_rate": 7.870872664676684e-06, "loss": 0.0406, "step": 37797 }, { "epoch": 11.63, "learning_rate": 7.87038582199634e-06, "loss": 0.0701, "step": 37798 }, { "epoch": 11.63, "learning_rate": 7.869898984603256e-06, "loss": 0.0194, "step": 37799 }, { "epoch": 11.63, "learning_rate": 7.86941215249864e-06, "loss": 0.0149, "step": 37800 }, { "epoch": 11.63, "learning_rate": 7.8689253256837e-06, "loss": 0.0306, "step": 37801 }, { "epoch": 11.63, "learning_rate": 7.868438504159641e-06, "loss": 0.0786, "step": 37802 }, { "epoch": 11.63, "learning_rate": 7.867951687927674e-06, "loss": 0.0347, "step": 37803 }, { "epoch": 11.63, "learning_rate": 7.86746487698901e-06, "loss": 0.036, "step": 37804 }, { "epoch": 11.63, "learning_rate": 7.866978071344856e-06, "loss": 0.0824, "step": 37805 }, { "epoch": 11.63, "learning_rate": 7.866491270996421e-06, "loss": 0.056, "step": 37806 }, { "epoch": 11.63, "learning_rate": 7.866004475944913e-06, "loss": 0.0679, "step": 37807 }, { "epoch": 11.63, "learning_rate": 7.865517686191536e-06, "loss": 0.0489, "step": 37808 }, { "epoch": 11.63, "learning_rate": 7.865030901737508e-06, "loss": 0.0437, "step": 37809 }, { "epoch": 11.63, "learning_rate": 7.864544122584032e-06, "loss": 0.0362, "step": 37810 }, { "epoch": 11.63, "learning_rate": 7.864057348732316e-06, "loss": 0.0338, "step": 37811 }, { "epoch": 11.63, "learning_rate": 7.863570580183572e-06, "loss": 0.0428, "step": 37812 }, { "epoch": 11.63, "learning_rate": 7.863083816939003e-06, "loss": 0.0361, "step": 37813 }, { "epoch": 11.64, "learning_rate": 7.86259705899982e-06, "loss": 0.0454, "step": 37814 }, { "epoch": 11.64, "learning_rate": 7.862110306367236e-06, "loss": 0.0265, "step": 37815 }, { "epoch": 11.64, "learning_rate": 7.861623559042457e-06, "loss": 0.077, "step": 37816 }, { "epoch": 11.64, "learning_rate": 7.861136817026688e-06, "loss": 0.0451, "step": 37817 }, { "epoch": 11.64, "learning_rate": 7.860650080321135e-06, "loss": 0.0629, "step": 37818 }, { "epoch": 11.64, "learning_rate": 7.860163348927018e-06, "loss": 0.0178, "step": 37819 }, { "epoch": 11.64, "learning_rate": 7.859676622845535e-06, "loss": 0.0479, "step": 37820 }, { "epoch": 11.64, "learning_rate": 7.859189902077901e-06, "loss": 0.0418, "step": 37821 }, { "epoch": 11.64, "learning_rate": 7.858703186625319e-06, "loss": 0.0818, "step": 37822 }, { "epoch": 11.64, "learning_rate": 7.858216476488999e-06, "loss": 0.0439, "step": 37823 }, { "epoch": 11.64, "learning_rate": 7.857729771670151e-06, "loss": 0.0473, "step": 37824 }, { "epoch": 11.64, "learning_rate": 7.857243072169983e-06, "loss": 0.0453, "step": 37825 }, { "epoch": 11.64, "learning_rate": 7.856756377989702e-06, "loss": 0.0643, "step": 37826 }, { "epoch": 11.64, "learning_rate": 7.856269689130519e-06, "loss": 0.0226, "step": 37827 }, { "epoch": 11.64, "learning_rate": 7.855783005593635e-06, "loss": 0.0659, "step": 37828 }, { "epoch": 11.64, "learning_rate": 7.855296327380267e-06, "loss": 0.0417, "step": 37829 }, { "epoch": 11.64, "learning_rate": 7.854809654491622e-06, "loss": 0.0431, "step": 37830 }, { "epoch": 11.64, "learning_rate": 7.854322986928906e-06, "loss": 0.0647, "step": 37831 }, { "epoch": 11.64, "learning_rate": 7.853836324693327e-06, "loss": 0.0616, "step": 37832 }, { "epoch": 11.64, "learning_rate": 7.85334966778609e-06, "loss": 0.0612, "step": 37833 }, { "epoch": 11.64, "learning_rate": 7.852863016208412e-06, "loss": 0.0511, "step": 37834 }, { "epoch": 11.64, "learning_rate": 7.852376369961495e-06, "loss": 0.0545, "step": 37835 }, { "epoch": 11.64, "learning_rate": 7.851889729046548e-06, "loss": 0.0829, "step": 37836 }, { "epoch": 11.64, "learning_rate": 7.851403093464782e-06, "loss": 0.0456, "step": 37837 }, { "epoch": 11.64, "learning_rate": 7.850916463217396e-06, "loss": 0.0489, "step": 37838 }, { "epoch": 11.64, "learning_rate": 7.85042983830561e-06, "loss": 0.0471, "step": 37839 }, { "epoch": 11.64, "learning_rate": 7.84994321873063e-06, "loss": 0.0491, "step": 37840 }, { "epoch": 11.64, "learning_rate": 7.849456604493658e-06, "loss": 0.0658, "step": 37841 }, { "epoch": 11.64, "learning_rate": 7.848969995595905e-06, "loss": 0.0438, "step": 37842 }, { "epoch": 11.64, "learning_rate": 7.84848339203858e-06, "loss": 0.0316, "step": 37843 }, { "epoch": 11.64, "learning_rate": 7.84799679382289e-06, "loss": 0.0609, "step": 37844 }, { "epoch": 11.64, "learning_rate": 7.847510200950045e-06, "loss": 0.041, "step": 37845 }, { "epoch": 11.64, "learning_rate": 7.847023613421251e-06, "loss": 0.0497, "step": 37846 }, { "epoch": 11.65, "learning_rate": 7.84653703123772e-06, "loss": 0.0344, "step": 37847 }, { "epoch": 11.65, "learning_rate": 7.846050454400653e-06, "loss": 0.0526, "step": 37848 }, { "epoch": 11.65, "learning_rate": 7.845563882911263e-06, "loss": 0.0428, "step": 37849 }, { "epoch": 11.65, "learning_rate": 7.845077316770757e-06, "loss": 0.0391, "step": 37850 }, { "epoch": 11.65, "learning_rate": 7.844590755980346e-06, "loss": 0.0832, "step": 37851 }, { "epoch": 11.65, "learning_rate": 7.844104200541235e-06, "loss": 0.0385, "step": 37852 }, { "epoch": 11.65, "learning_rate": 7.843617650454625e-06, "loss": 0.0529, "step": 37853 }, { "epoch": 11.65, "learning_rate": 7.843131105721739e-06, "loss": 0.065, "step": 37854 }, { "epoch": 11.65, "learning_rate": 7.842644566343773e-06, "loss": 0.0491, "step": 37855 }, { "epoch": 11.65, "learning_rate": 7.84215803232194e-06, "loss": 0.0461, "step": 37856 }, { "epoch": 11.65, "learning_rate": 7.841671503657448e-06, "loss": 0.0665, "step": 37857 }, { "epoch": 11.65, "learning_rate": 7.8411849803515e-06, "loss": 0.0383, "step": 37858 }, { "epoch": 11.65, "learning_rate": 7.84069846240531e-06, "loss": 0.0234, "step": 37859 }, { "epoch": 11.65, "learning_rate": 7.840211949820085e-06, "loss": 0.0279, "step": 37860 }, { "epoch": 11.65, "learning_rate": 7.839725442597033e-06, "loss": 0.0628, "step": 37861 }, { "epoch": 11.65, "learning_rate": 7.839238940737356e-06, "loss": 0.0441, "step": 37862 }, { "epoch": 11.65, "learning_rate": 7.838752444242264e-06, "loss": 0.0377, "step": 37863 }, { "epoch": 11.65, "learning_rate": 7.838265953112973e-06, "loss": 0.0925, "step": 37864 }, { "epoch": 11.65, "learning_rate": 7.837779467350683e-06, "loss": 0.0525, "step": 37865 }, { "epoch": 11.65, "learning_rate": 7.837292986956604e-06, "loss": 0.0261, "step": 37866 }, { "epoch": 11.65, "learning_rate": 7.836806511931943e-06, "loss": 0.0799, "step": 37867 }, { "epoch": 11.65, "learning_rate": 7.836320042277906e-06, "loss": 0.0453, "step": 37868 }, { "epoch": 11.65, "learning_rate": 7.835833577995706e-06, "loss": 0.0303, "step": 37869 }, { "epoch": 11.65, "learning_rate": 7.835347119086547e-06, "loss": 0.0572, "step": 37870 }, { "epoch": 11.65, "learning_rate": 7.834860665551638e-06, "loss": 0.0441, "step": 37871 }, { "epoch": 11.65, "learning_rate": 7.834374217392188e-06, "loss": 0.0778, "step": 37872 }, { "epoch": 11.65, "learning_rate": 7.833887774609397e-06, "loss": 0.0563, "step": 37873 }, { "epoch": 11.65, "learning_rate": 7.833401337204483e-06, "loss": 0.0324, "step": 37874 }, { "epoch": 11.65, "learning_rate": 7.83291490517865e-06, "loss": 0.038, "step": 37875 }, { "epoch": 11.65, "learning_rate": 7.832428478533105e-06, "loss": 0.0351, "step": 37876 }, { "epoch": 11.65, "learning_rate": 7.831942057269055e-06, "loss": 0.0523, "step": 37877 }, { "epoch": 11.65, "learning_rate": 7.831455641387705e-06, "loss": 0.0604, "step": 37878 }, { "epoch": 11.66, "learning_rate": 7.830969230890269e-06, "loss": 0.0772, "step": 37879 }, { "epoch": 11.66, "learning_rate": 7.830482825777951e-06, "loss": 0.0265, "step": 37880 }, { "epoch": 11.66, "learning_rate": 7.82999642605196e-06, "loss": 0.0537, "step": 37881 }, { "epoch": 11.66, "learning_rate": 7.829510031713506e-06, "loss": 0.0529, "step": 37882 }, { "epoch": 11.66, "learning_rate": 7.829023642763785e-06, "loss": 0.0448, "step": 37883 }, { "epoch": 11.66, "learning_rate": 7.828537259204019e-06, "loss": 0.0622, "step": 37884 }, { "epoch": 11.66, "learning_rate": 7.828050881035412e-06, "loss": 0.0612, "step": 37885 }, { "epoch": 11.66, "learning_rate": 7.827564508259165e-06, "loss": 0.0865, "step": 37886 }, { "epoch": 11.66, "learning_rate": 7.82707814087649e-06, "loss": 0.0501, "step": 37887 }, { "epoch": 11.66, "learning_rate": 7.826591778888594e-06, "loss": 0.0542, "step": 37888 }, { "epoch": 11.66, "learning_rate": 7.826105422296687e-06, "loss": 0.0781, "step": 37889 }, { "epoch": 11.66, "learning_rate": 7.82561907110197e-06, "loss": 0.0284, "step": 37890 }, { "epoch": 11.66, "learning_rate": 7.82513272530566e-06, "loss": 0.0386, "step": 37891 }, { "epoch": 11.66, "learning_rate": 7.824646384908955e-06, "loss": 0.0668, "step": 37892 }, { "epoch": 11.66, "learning_rate": 7.824160049913066e-06, "loss": 0.0568, "step": 37893 }, { "epoch": 11.66, "learning_rate": 7.823673720319202e-06, "loss": 0.0293, "step": 37894 }, { "epoch": 11.66, "learning_rate": 7.823187396128572e-06, "loss": 0.0589, "step": 37895 }, { "epoch": 11.66, "learning_rate": 7.82270107734238e-06, "loss": 0.0513, "step": 37896 }, { "epoch": 11.66, "learning_rate": 7.822214763961835e-06, "loss": 0.0554, "step": 37897 }, { "epoch": 11.66, "learning_rate": 7.82172845598814e-06, "loss": 0.0642, "step": 37898 }, { "epoch": 11.66, "learning_rate": 7.821242153422507e-06, "loss": 0.0161, "step": 37899 }, { "epoch": 11.66, "learning_rate": 7.820755856266145e-06, "loss": 0.0704, "step": 37900 }, { "epoch": 11.66, "learning_rate": 7.820269564520258e-06, "loss": 0.0485, "step": 37901 }, { "epoch": 11.66, "learning_rate": 7.819783278186053e-06, "loss": 0.0392, "step": 37902 }, { "epoch": 11.66, "learning_rate": 7.819296997264736e-06, "loss": 0.0291, "step": 37903 }, { "epoch": 11.66, "learning_rate": 7.81881072175752e-06, "loss": 0.0637, "step": 37904 }, { "epoch": 11.66, "learning_rate": 7.818324451665608e-06, "loss": 0.0547, "step": 37905 }, { "epoch": 11.66, "learning_rate": 7.81783818699021e-06, "loss": 0.0463, "step": 37906 }, { "epoch": 11.66, "learning_rate": 7.81735192773253e-06, "loss": 0.0412, "step": 37907 }, { "epoch": 11.66, "learning_rate": 7.816865673893774e-06, "loss": 0.0662, "step": 37908 }, { "epoch": 11.66, "learning_rate": 7.816379425475155e-06, "loss": 0.0593, "step": 37909 }, { "epoch": 11.66, "learning_rate": 7.815893182477878e-06, "loss": 0.0409, "step": 37910 }, { "epoch": 11.66, "learning_rate": 7.815406944903148e-06, "loss": 0.0777, "step": 37911 }, { "epoch": 11.67, "learning_rate": 7.814920712752174e-06, "loss": 0.0649, "step": 37912 }, { "epoch": 11.67, "learning_rate": 7.81443448602616e-06, "loss": 0.0419, "step": 37913 }, { "epoch": 11.67, "learning_rate": 7.813948264726318e-06, "loss": 0.0655, "step": 37914 }, { "epoch": 11.67, "learning_rate": 7.813462048853855e-06, "loss": 0.0591, "step": 37915 }, { "epoch": 11.67, "learning_rate": 7.812975838409977e-06, "loss": 0.0581, "step": 37916 }, { "epoch": 11.67, "learning_rate": 7.812489633395888e-06, "loss": 0.063, "step": 37917 }, { "epoch": 11.67, "learning_rate": 7.812003433812795e-06, "loss": 0.074, "step": 37918 }, { "epoch": 11.67, "learning_rate": 7.811517239661912e-06, "loss": 0.0491, "step": 37919 }, { "epoch": 11.67, "learning_rate": 7.81103105094444e-06, "loss": 0.0503, "step": 37920 }, { "epoch": 11.67, "learning_rate": 7.810544867661589e-06, "loss": 0.052, "step": 37921 }, { "epoch": 11.67, "learning_rate": 7.810058689814564e-06, "loss": 0.0463, "step": 37922 }, { "epoch": 11.67, "learning_rate": 7.80957251740457e-06, "loss": 0.0501, "step": 37923 }, { "epoch": 11.67, "learning_rate": 7.80908635043282e-06, "loss": 0.0934, "step": 37924 }, { "epoch": 11.67, "learning_rate": 7.808600188900517e-06, "loss": 0.0385, "step": 37925 }, { "epoch": 11.67, "learning_rate": 7.80811403280887e-06, "loss": 0.0627, "step": 37926 }, { "epoch": 11.67, "learning_rate": 7.807627882159087e-06, "loss": 0.0476, "step": 37927 }, { "epoch": 11.67, "learning_rate": 7.807141736952367e-06, "loss": 0.0383, "step": 37928 }, { "epoch": 11.67, "learning_rate": 7.806655597189925e-06, "loss": 0.0498, "step": 37929 }, { "epoch": 11.67, "learning_rate": 7.80616946287297e-06, "loss": 0.0387, "step": 37930 }, { "epoch": 11.67, "learning_rate": 7.805683334002702e-06, "loss": 0.0411, "step": 37931 }, { "epoch": 11.67, "learning_rate": 7.80519721058033e-06, "loss": 0.0687, "step": 37932 }, { "epoch": 11.67, "learning_rate": 7.804711092607061e-06, "loss": 0.0347, "step": 37933 }, { "epoch": 11.67, "learning_rate": 7.804224980084103e-06, "loss": 0.0408, "step": 37934 }, { "epoch": 11.67, "learning_rate": 7.803738873012663e-06, "loss": 0.0404, "step": 37935 }, { "epoch": 11.67, "learning_rate": 7.803252771393948e-06, "loss": 0.0448, "step": 37936 }, { "epoch": 11.67, "learning_rate": 7.802766675229166e-06, "loss": 0.0395, "step": 37937 }, { "epoch": 11.67, "learning_rate": 7.802280584519515e-06, "loss": 0.0274, "step": 37938 }, { "epoch": 11.67, "learning_rate": 7.801794499266212e-06, "loss": 0.1171, "step": 37939 }, { "epoch": 11.67, "learning_rate": 7.801308419470466e-06, "loss": 0.0617, "step": 37940 }, { "epoch": 11.67, "learning_rate": 7.800822345133475e-06, "loss": 0.0538, "step": 37941 }, { "epoch": 11.67, "learning_rate": 7.800336276256449e-06, "loss": 0.0467, "step": 37942 }, { "epoch": 11.67, "learning_rate": 7.799850212840592e-06, "loss": 0.0667, "step": 37943 }, { "epoch": 11.68, "learning_rate": 7.799364154887117e-06, "loss": 0.0437, "step": 37944 }, { "epoch": 11.68, "learning_rate": 7.798878102397227e-06, "loss": 0.0521, "step": 37945 }, { "epoch": 11.68, "learning_rate": 7.798392055372129e-06, "loss": 0.0291, "step": 37946 }, { "epoch": 11.68, "learning_rate": 7.79790601381303e-06, "loss": 0.1039, "step": 37947 }, { "epoch": 11.68, "learning_rate": 7.797419977721135e-06, "loss": 0.0558, "step": 37948 }, { "epoch": 11.68, "learning_rate": 7.796933947097654e-06, "loss": 0.0599, "step": 37949 }, { "epoch": 11.68, "learning_rate": 7.796447921943793e-06, "loss": 0.0291, "step": 37950 }, { "epoch": 11.68, "learning_rate": 7.79596190226076e-06, "loss": 0.0321, "step": 37951 }, { "epoch": 11.68, "learning_rate": 7.795475888049757e-06, "loss": 0.0617, "step": 37952 }, { "epoch": 11.68, "learning_rate": 7.794989879311991e-06, "loss": 0.0501, "step": 37953 }, { "epoch": 11.68, "learning_rate": 7.794503876048672e-06, "loss": 0.0351, "step": 37954 }, { "epoch": 11.68, "learning_rate": 7.794017878261006e-06, "loss": 0.0425, "step": 37955 }, { "epoch": 11.68, "learning_rate": 7.793531885950199e-06, "loss": 0.0269, "step": 37956 }, { "epoch": 11.68, "learning_rate": 7.793045899117457e-06, "loss": 0.0532, "step": 37957 }, { "epoch": 11.68, "learning_rate": 7.792559917763985e-06, "loss": 0.0701, "step": 37958 }, { "epoch": 11.68, "learning_rate": 7.792073941890994e-06, "loss": 0.0535, "step": 37959 }, { "epoch": 11.68, "learning_rate": 7.791587971499687e-06, "loss": 0.0375, "step": 37960 }, { "epoch": 11.68, "learning_rate": 7.791102006591274e-06, "loss": 0.0707, "step": 37961 }, { "epoch": 11.68, "learning_rate": 7.790616047166959e-06, "loss": 0.0488, "step": 37962 }, { "epoch": 11.68, "learning_rate": 7.790130093227943e-06, "loss": 0.0398, "step": 37963 }, { "epoch": 11.68, "learning_rate": 7.789644144775445e-06, "loss": 0.0811, "step": 37964 }, { "epoch": 11.68, "learning_rate": 7.789158201810663e-06, "loss": 0.0722, "step": 37965 }, { "epoch": 11.68, "learning_rate": 7.788672264334804e-06, "loss": 0.0417, "step": 37966 }, { "epoch": 11.68, "learning_rate": 7.788186332349075e-06, "loss": 0.0733, "step": 37967 }, { "epoch": 11.68, "learning_rate": 7.787700405854683e-06, "loss": 0.0336, "step": 37968 }, { "epoch": 11.68, "learning_rate": 7.787214484852837e-06, "loss": 0.0389, "step": 37969 }, { "epoch": 11.68, "learning_rate": 7.78672856934474e-06, "loss": 0.0858, "step": 37970 }, { "epoch": 11.68, "learning_rate": 7.786242659331598e-06, "loss": 0.0358, "step": 37971 }, { "epoch": 11.68, "learning_rate": 7.785756754814623e-06, "loss": 0.0352, "step": 37972 }, { "epoch": 11.68, "learning_rate": 7.785270855795008e-06, "loss": 0.0472, "step": 37973 }, { "epoch": 11.68, "learning_rate": 7.784784962273977e-06, "loss": 0.0267, "step": 37974 }, { "epoch": 11.68, "learning_rate": 7.784299074252724e-06, "loss": 0.0562, "step": 37975 }, { "epoch": 11.68, "learning_rate": 7.78381319173246e-06, "loss": 0.0295, "step": 37976 }, { "epoch": 11.69, "learning_rate": 7.783327314714391e-06, "loss": 0.0531, "step": 37977 }, { "epoch": 11.69, "learning_rate": 7.782841443199718e-06, "loss": 0.0715, "step": 37978 }, { "epoch": 11.69, "learning_rate": 7.782355577189657e-06, "loss": 0.0421, "step": 37979 }, { "epoch": 11.69, "learning_rate": 7.781869716685408e-06, "loss": 0.0298, "step": 37980 }, { "epoch": 11.69, "learning_rate": 7.781383861688179e-06, "loss": 0.0347, "step": 37981 }, { "epoch": 11.69, "learning_rate": 7.780898012199177e-06, "loss": 0.0552, "step": 37982 }, { "epoch": 11.69, "learning_rate": 7.7804121682196e-06, "loss": 0.0649, "step": 37983 }, { "epoch": 11.69, "learning_rate": 7.779926329750668e-06, "loss": 0.04, "step": 37984 }, { "epoch": 11.69, "learning_rate": 7.779440496793582e-06, "loss": 0.037, "step": 37985 }, { "epoch": 11.69, "learning_rate": 7.778954669349544e-06, "loss": 0.0719, "step": 37986 }, { "epoch": 11.69, "learning_rate": 7.778468847419761e-06, "loss": 0.0297, "step": 37987 }, { "epoch": 11.69, "learning_rate": 7.77798303100544e-06, "loss": 0.0383, "step": 37988 }, { "epoch": 11.69, "learning_rate": 7.777497220107791e-06, "loss": 0.045, "step": 37989 }, { "epoch": 11.69, "learning_rate": 7.777011414728016e-06, "loss": 0.0195, "step": 37990 }, { "epoch": 11.69, "learning_rate": 7.776525614867323e-06, "loss": 0.084, "step": 37991 }, { "epoch": 11.69, "learning_rate": 7.776039820526918e-06, "loss": 0.0609, "step": 37992 }, { "epoch": 11.69, "learning_rate": 7.775554031708004e-06, "loss": 0.0457, "step": 37993 }, { "epoch": 11.69, "learning_rate": 7.77506824841179e-06, "loss": 0.0522, "step": 37994 }, { "epoch": 11.69, "learning_rate": 7.774582470639487e-06, "loss": 0.0526, "step": 37995 }, { "epoch": 11.69, "learning_rate": 7.774096698392292e-06, "loss": 0.0456, "step": 37996 }, { "epoch": 11.69, "learning_rate": 7.773610931671415e-06, "loss": 0.0383, "step": 37997 }, { "epoch": 11.69, "learning_rate": 7.77312517047806e-06, "loss": 0.0612, "step": 37998 }, { "epoch": 11.69, "learning_rate": 7.772639414813436e-06, "loss": 0.0272, "step": 37999 }, { "epoch": 11.69, "learning_rate": 7.772153664678749e-06, "loss": 0.0763, "step": 38000 }, { "epoch": 11.69, "learning_rate": 7.771667920075204e-06, "loss": 0.0326, "step": 38001 }, { "epoch": 11.69, "learning_rate": 7.771182181004005e-06, "loss": 0.057, "step": 38002 }, { "epoch": 11.69, "learning_rate": 7.77069644746636e-06, "loss": 0.082, "step": 38003 }, { "epoch": 11.69, "learning_rate": 7.770210719463475e-06, "loss": 0.0414, "step": 38004 }, { "epoch": 11.69, "learning_rate": 7.769724996996556e-06, "loss": 0.0572, "step": 38005 }, { "epoch": 11.69, "learning_rate": 7.76923928006681e-06, "loss": 0.0534, "step": 38006 }, { "epoch": 11.69, "learning_rate": 7.768753568675443e-06, "loss": 0.0496, "step": 38007 }, { "epoch": 11.69, "learning_rate": 7.76826786282365e-06, "loss": 0.0678, "step": 38008 }, { "epoch": 11.7, "learning_rate": 7.767782162512656e-06, "loss": 0.0292, "step": 38009 }, { "epoch": 11.7, "learning_rate": 7.767296467743655e-06, "loss": 0.0477, "step": 38010 }, { "epoch": 11.7, "learning_rate": 7.766810778517854e-06, "loss": 0.0358, "step": 38011 }, { "epoch": 11.7, "learning_rate": 7.76632509483646e-06, "loss": 0.1237, "step": 38012 }, { "epoch": 11.7, "learning_rate": 7.765839416700675e-06, "loss": 0.0628, "step": 38013 }, { "epoch": 11.7, "learning_rate": 7.765353744111712e-06, "loss": 0.0418, "step": 38014 }, { "epoch": 11.7, "learning_rate": 7.764868077070772e-06, "loss": 0.0777, "step": 38015 }, { "epoch": 11.7, "learning_rate": 7.764382415579067e-06, "loss": 0.0582, "step": 38016 }, { "epoch": 11.7, "learning_rate": 7.763896759637793e-06, "loss": 0.0683, "step": 38017 }, { "epoch": 11.7, "learning_rate": 7.763411109248157e-06, "loss": 0.0428, "step": 38018 }, { "epoch": 11.7, "learning_rate": 7.762925464411375e-06, "loss": 0.0225, "step": 38019 }, { "epoch": 11.7, "learning_rate": 7.762439825128643e-06, "loss": 0.0369, "step": 38020 }, { "epoch": 11.7, "learning_rate": 7.76195419140117e-06, "loss": 0.0424, "step": 38021 }, { "epoch": 11.7, "learning_rate": 7.761468563230163e-06, "loss": 0.0594, "step": 38022 }, { "epoch": 11.7, "learning_rate": 7.760982940616823e-06, "loss": 0.0345, "step": 38023 }, { "epoch": 11.7, "learning_rate": 7.760497323562359e-06, "loss": 0.069, "step": 38024 }, { "epoch": 11.7, "learning_rate": 7.760011712067978e-06, "loss": 0.0303, "step": 38025 }, { "epoch": 11.7, "learning_rate": 7.759526106134885e-06, "loss": 0.0649, "step": 38026 }, { "epoch": 11.7, "learning_rate": 7.75904050576428e-06, "loss": 0.0339, "step": 38027 }, { "epoch": 11.7, "learning_rate": 7.758554910957378e-06, "loss": 0.079, "step": 38028 }, { "epoch": 11.7, "learning_rate": 7.758069321715378e-06, "loss": 0.0461, "step": 38029 }, { "epoch": 11.7, "learning_rate": 7.75758373803949e-06, "loss": 0.0448, "step": 38030 }, { "epoch": 11.7, "learning_rate": 7.757098159930915e-06, "loss": 0.0466, "step": 38031 }, { "epoch": 11.7, "learning_rate": 7.75661258739086e-06, "loss": 0.0576, "step": 38032 }, { "epoch": 11.7, "learning_rate": 7.756127020420532e-06, "loss": 0.0342, "step": 38033 }, { "epoch": 11.7, "learning_rate": 7.755641459021135e-06, "loss": 0.0402, "step": 38034 }, { "epoch": 11.7, "learning_rate": 7.755155903193876e-06, "loss": 0.0861, "step": 38035 }, { "epoch": 11.7, "learning_rate": 7.754670352939959e-06, "loss": 0.0708, "step": 38036 }, { "epoch": 11.7, "learning_rate": 7.754184808260589e-06, "loss": 0.061, "step": 38037 }, { "epoch": 11.7, "learning_rate": 7.753699269156974e-06, "loss": 0.0287, "step": 38038 }, { "epoch": 11.7, "learning_rate": 7.753213735630318e-06, "loss": 0.0379, "step": 38039 }, { "epoch": 11.7, "learning_rate": 7.75272820768183e-06, "loss": 0.0562, "step": 38040 }, { "epoch": 11.7, "learning_rate": 7.752242685312709e-06, "loss": 0.0252, "step": 38041 }, { "epoch": 11.71, "learning_rate": 7.751757168524159e-06, "loss": 0.0588, "step": 38042 }, { "epoch": 11.71, "learning_rate": 7.751271657317397e-06, "loss": 0.0904, "step": 38043 }, { "epoch": 11.71, "learning_rate": 7.750786151693618e-06, "loss": 0.0732, "step": 38044 }, { "epoch": 11.71, "learning_rate": 7.750300651654031e-06, "loss": 0.058, "step": 38045 }, { "epoch": 11.71, "learning_rate": 7.749815157199843e-06, "loss": 0.0815, "step": 38046 }, { "epoch": 11.71, "learning_rate": 7.749329668332253e-06, "loss": 0.056, "step": 38047 }, { "epoch": 11.71, "learning_rate": 7.748844185052474e-06, "loss": 0.0657, "step": 38048 }, { "epoch": 11.71, "learning_rate": 7.748358707361706e-06, "loss": 0.0386, "step": 38049 }, { "epoch": 11.71, "learning_rate": 7.747873235261157e-06, "loss": 0.0477, "step": 38050 }, { "epoch": 11.71, "learning_rate": 7.747387768752035e-06, "loss": 0.0227, "step": 38051 }, { "epoch": 11.71, "learning_rate": 7.746902307835535e-06, "loss": 0.0324, "step": 38052 }, { "epoch": 11.71, "learning_rate": 7.746416852512874e-06, "loss": 0.0351, "step": 38053 }, { "epoch": 11.71, "learning_rate": 7.745931402785252e-06, "loss": 0.0175, "step": 38054 }, { "epoch": 11.71, "learning_rate": 7.745445958653875e-06, "loss": 0.052, "step": 38055 }, { "epoch": 11.71, "learning_rate": 7.744960520119946e-06, "loss": 0.0469, "step": 38056 }, { "epoch": 11.71, "learning_rate": 7.744475087184671e-06, "loss": 0.0682, "step": 38057 }, { "epoch": 11.71, "learning_rate": 7.74398965984926e-06, "loss": 0.0194, "step": 38058 }, { "epoch": 11.71, "learning_rate": 7.743504238114911e-06, "loss": 0.0259, "step": 38059 }, { "epoch": 11.71, "learning_rate": 7.743018821982835e-06, "loss": 0.048, "step": 38060 }, { "epoch": 11.71, "learning_rate": 7.742533411454238e-06, "loss": 0.047, "step": 38061 }, { "epoch": 11.71, "learning_rate": 7.742048006530313e-06, "loss": 0.0985, "step": 38062 }, { "epoch": 11.71, "learning_rate": 7.74156260721228e-06, "loss": 0.0423, "step": 38063 }, { "epoch": 11.71, "learning_rate": 7.741077213501339e-06, "loss": 0.0428, "step": 38064 }, { "epoch": 11.71, "learning_rate": 7.740591825398693e-06, "loss": 0.0386, "step": 38065 }, { "epoch": 11.71, "learning_rate": 7.740106442905548e-06, "loss": 0.0634, "step": 38066 }, { "epoch": 11.71, "learning_rate": 7.739621066023108e-06, "loss": 0.0527, "step": 38067 }, { "epoch": 11.71, "learning_rate": 7.739135694752582e-06, "loss": 0.0461, "step": 38068 }, { "epoch": 11.71, "learning_rate": 7.73865032909517e-06, "loss": 0.0644, "step": 38069 }, { "epoch": 11.71, "learning_rate": 7.738164969052082e-06, "loss": 0.0674, "step": 38070 }, { "epoch": 11.71, "learning_rate": 7.73767961462452e-06, "loss": 0.0557, "step": 38071 }, { "epoch": 11.71, "learning_rate": 7.737194265813687e-06, "loss": 0.0522, "step": 38072 }, { "epoch": 11.71, "learning_rate": 7.736708922620795e-06, "loss": 0.0468, "step": 38073 }, { "epoch": 11.72, "learning_rate": 7.736223585047044e-06, "loss": 0.038, "step": 38074 }, { "epoch": 11.72, "learning_rate": 7.735738253093638e-06, "loss": 0.0362, "step": 38075 }, { "epoch": 11.72, "learning_rate": 7.735252926761784e-06, "loss": 0.0682, "step": 38076 }, { "epoch": 11.72, "learning_rate": 7.734767606052684e-06, "loss": 0.05, "step": 38077 }, { "epoch": 11.72, "learning_rate": 7.734282290967548e-06, "loss": 0.0899, "step": 38078 }, { "epoch": 11.72, "learning_rate": 7.733796981507577e-06, "loss": 0.0498, "step": 38079 }, { "epoch": 11.72, "learning_rate": 7.733311677673979e-06, "loss": 0.0211, "step": 38080 }, { "epoch": 11.72, "learning_rate": 7.732826379467956e-06, "loss": 0.0275, "step": 38081 }, { "epoch": 11.72, "learning_rate": 7.732341086890711e-06, "loss": 0.0277, "step": 38082 }, { "epoch": 11.72, "learning_rate": 7.731855799943455e-06, "loss": 0.0506, "step": 38083 }, { "epoch": 11.72, "learning_rate": 7.73137051862739e-06, "loss": 0.069, "step": 38084 }, { "epoch": 11.72, "learning_rate": 7.730885242943721e-06, "loss": 0.0697, "step": 38085 }, { "epoch": 11.72, "learning_rate": 7.730399972893651e-06, "loss": 0.0437, "step": 38086 }, { "epoch": 11.72, "learning_rate": 7.729914708478383e-06, "loss": 0.0757, "step": 38087 }, { "epoch": 11.72, "learning_rate": 7.729429449699128e-06, "loss": 0.0615, "step": 38088 }, { "epoch": 11.72, "learning_rate": 7.728944196557087e-06, "loss": 0.0332, "step": 38089 }, { "epoch": 11.72, "learning_rate": 7.728458949053465e-06, "loss": 0.0378, "step": 38090 }, { "epoch": 11.72, "learning_rate": 7.727973707189469e-06, "loss": 0.0616, "step": 38091 }, { "epoch": 11.72, "learning_rate": 7.727488470966298e-06, "loss": 0.045, "step": 38092 }, { "epoch": 11.72, "learning_rate": 7.727003240385163e-06, "loss": 0.0382, "step": 38093 }, { "epoch": 11.72, "learning_rate": 7.726518015447266e-06, "loss": 0.051, "step": 38094 }, { "epoch": 11.72, "learning_rate": 7.726032796153814e-06, "loss": 0.0788, "step": 38095 }, { "epoch": 11.72, "learning_rate": 7.725547582506007e-06, "loss": 0.0395, "step": 38096 }, { "epoch": 11.72, "learning_rate": 7.725062374505048e-06, "loss": 0.0381, "step": 38097 }, { "epoch": 11.72, "learning_rate": 7.724577172152153e-06, "loss": 0.0453, "step": 38098 }, { "epoch": 11.72, "learning_rate": 7.724091975448517e-06, "loss": 0.0526, "step": 38099 }, { "epoch": 11.72, "learning_rate": 7.723606784395345e-06, "loss": 0.0218, "step": 38100 }, { "epoch": 11.72, "learning_rate": 7.723121598993846e-06, "loss": 0.0288, "step": 38101 }, { "epoch": 11.72, "learning_rate": 7.72263641924522e-06, "loss": 0.0282, "step": 38102 }, { "epoch": 11.72, "learning_rate": 7.722151245150676e-06, "loss": 0.0644, "step": 38103 }, { "epoch": 11.72, "learning_rate": 7.721666076711416e-06, "loss": 0.0378, "step": 38104 }, { "epoch": 11.72, "learning_rate": 7.721180913928645e-06, "loss": 0.0674, "step": 38105 }, { "epoch": 11.72, "learning_rate": 7.720695756803569e-06, "loss": 0.0363, "step": 38106 }, { "epoch": 11.73, "learning_rate": 7.720210605337385e-06, "loss": 0.0334, "step": 38107 }, { "epoch": 11.73, "learning_rate": 7.719725459531308e-06, "loss": 0.043, "step": 38108 }, { "epoch": 11.73, "learning_rate": 7.71924031938654e-06, "loss": 0.0373, "step": 38109 }, { "epoch": 11.73, "learning_rate": 7.71875518490428e-06, "loss": 0.0657, "step": 38110 }, { "epoch": 11.73, "learning_rate": 7.718270056085738e-06, "loss": 0.0469, "step": 38111 }, { "epoch": 11.73, "learning_rate": 7.717784932932112e-06, "loss": 0.0447, "step": 38112 }, { "epoch": 11.73, "learning_rate": 7.717299815444614e-06, "loss": 0.0451, "step": 38113 }, { "epoch": 11.73, "learning_rate": 7.716814703624445e-06, "loss": 0.0645, "step": 38114 }, { "epoch": 11.73, "learning_rate": 7.716329597472811e-06, "loss": 0.0502, "step": 38115 }, { "epoch": 11.73, "learning_rate": 7.715844496990914e-06, "loss": 0.0311, "step": 38116 }, { "epoch": 11.73, "learning_rate": 7.715359402179954e-06, "loss": 0.0373, "step": 38117 }, { "epoch": 11.73, "learning_rate": 7.714874313041148e-06, "loss": 0.0644, "step": 38118 }, { "epoch": 11.73, "learning_rate": 7.714389229575692e-06, "loss": 0.0763, "step": 38119 }, { "epoch": 11.73, "learning_rate": 7.71390415178479e-06, "loss": 0.075, "step": 38120 }, { "epoch": 11.73, "learning_rate": 7.713419079669647e-06, "loss": 0.0582, "step": 38121 }, { "epoch": 11.73, "learning_rate": 7.712934013231466e-06, "loss": 0.0378, "step": 38122 }, { "epoch": 11.73, "learning_rate": 7.712448952471457e-06, "loss": 0.0383, "step": 38123 }, { "epoch": 11.73, "learning_rate": 7.71196389739082e-06, "loss": 0.0667, "step": 38124 }, { "epoch": 11.73, "learning_rate": 7.711478847990759e-06, "loss": 0.0571, "step": 38125 }, { "epoch": 11.73, "learning_rate": 7.71099380427248e-06, "loss": 0.0646, "step": 38126 }, { "epoch": 11.73, "learning_rate": 7.710508766237182e-06, "loss": 0.0382, "step": 38127 }, { "epoch": 11.73, "learning_rate": 7.710023733886078e-06, "loss": 0.0301, "step": 38128 }, { "epoch": 11.73, "learning_rate": 7.709538707220366e-06, "loss": 0.0396, "step": 38129 }, { "epoch": 11.73, "learning_rate": 7.709053686241256e-06, "loss": 0.0698, "step": 38130 }, { "epoch": 11.73, "learning_rate": 7.708568670949944e-06, "loss": 0.0373, "step": 38131 }, { "epoch": 11.73, "learning_rate": 7.708083661347637e-06, "loss": 0.1017, "step": 38132 }, { "epoch": 11.73, "learning_rate": 7.707598657435544e-06, "loss": 0.0532, "step": 38133 }, { "epoch": 11.73, "learning_rate": 7.707113659214863e-06, "loss": 0.0864, "step": 38134 }, { "epoch": 11.73, "learning_rate": 7.706628666686802e-06, "loss": 0.0682, "step": 38135 }, { "epoch": 11.73, "learning_rate": 7.706143679852562e-06, "loss": 0.0323, "step": 38136 }, { "epoch": 11.73, "learning_rate": 7.705658698713348e-06, "loss": 0.0418, "step": 38137 }, { "epoch": 11.73, "learning_rate": 7.705173723270368e-06, "loss": 0.0457, "step": 38138 }, { "epoch": 11.74, "learning_rate": 7.704688753524823e-06, "loss": 0.0444, "step": 38139 }, { "epoch": 11.74, "learning_rate": 7.704203789477917e-06, "loss": 0.0813, "step": 38140 }, { "epoch": 11.74, "learning_rate": 7.703718831130854e-06, "loss": 0.022, "step": 38141 }, { "epoch": 11.74, "learning_rate": 7.703233878484832e-06, "loss": 0.041, "step": 38142 }, { "epoch": 11.74, "learning_rate": 7.702748931541068e-06, "loss": 0.0583, "step": 38143 }, { "epoch": 11.74, "learning_rate": 7.702263990300758e-06, "loss": 0.0516, "step": 38144 }, { "epoch": 11.74, "learning_rate": 7.701779054765107e-06, "loss": 0.0495, "step": 38145 }, { "epoch": 11.74, "learning_rate": 7.701294124935319e-06, "loss": 0.0657, "step": 38146 }, { "epoch": 11.74, "learning_rate": 7.700809200812596e-06, "loss": 0.0506, "step": 38147 }, { "epoch": 11.74, "learning_rate": 7.700324282398146e-06, "loss": 0.1099, "step": 38148 }, { "epoch": 11.74, "learning_rate": 7.69983936969317e-06, "loss": 0.0626, "step": 38149 }, { "epoch": 11.74, "learning_rate": 7.699354462698871e-06, "loss": 0.0392, "step": 38150 }, { "epoch": 11.74, "learning_rate": 7.69886956141646e-06, "loss": 0.0612, "step": 38151 }, { "epoch": 11.74, "learning_rate": 7.698384665847127e-06, "loss": 0.0384, "step": 38152 }, { "epoch": 11.74, "learning_rate": 7.697899775992093e-06, "loss": 0.0645, "step": 38153 }, { "epoch": 11.74, "learning_rate": 7.697414891852549e-06, "loss": 0.058, "step": 38154 }, { "epoch": 11.74, "learning_rate": 7.696930013429703e-06, "loss": 0.0391, "step": 38155 }, { "epoch": 11.74, "learning_rate": 7.69644514072476e-06, "loss": 0.0469, "step": 38156 }, { "epoch": 11.74, "learning_rate": 7.69596027373892e-06, "loss": 0.0658, "step": 38157 }, { "epoch": 11.74, "learning_rate": 7.695475412473393e-06, "loss": 0.0835, "step": 38158 }, { "epoch": 11.74, "learning_rate": 7.694990556929378e-06, "loss": 0.0388, "step": 38159 }, { "epoch": 11.74, "learning_rate": 7.694505707108079e-06, "loss": 0.0727, "step": 38160 }, { "epoch": 11.74, "learning_rate": 7.694020863010705e-06, "loss": 0.0483, "step": 38161 }, { "epoch": 11.74, "learning_rate": 7.693536024638449e-06, "loss": 0.0081, "step": 38162 }, { "epoch": 11.74, "learning_rate": 7.693051191992525e-06, "loss": 0.0652, "step": 38163 }, { "epoch": 11.74, "learning_rate": 7.692566365074135e-06, "loss": 0.0657, "step": 38164 }, { "epoch": 11.74, "learning_rate": 7.692081543884477e-06, "loss": 0.0575, "step": 38165 }, { "epoch": 11.74, "learning_rate": 7.69159672842476e-06, "loss": 0.034, "step": 38166 }, { "epoch": 11.74, "learning_rate": 7.691111918696183e-06, "loss": 0.0468, "step": 38167 }, { "epoch": 11.74, "learning_rate": 7.690627114699956e-06, "loss": 0.0557, "step": 38168 }, { "epoch": 11.74, "learning_rate": 7.690142316437278e-06, "loss": 0.0506, "step": 38169 }, { "epoch": 11.74, "learning_rate": 7.689657523909356e-06, "loss": 0.0432, "step": 38170 }, { "epoch": 11.74, "learning_rate": 7.689172737117389e-06, "loss": 0.0302, "step": 38171 }, { "epoch": 11.75, "learning_rate": 7.688687956062582e-06, "loss": 0.041, "step": 38172 }, { "epoch": 11.75, "learning_rate": 7.688203180746142e-06, "loss": 0.0657, "step": 38173 }, { "epoch": 11.75, "learning_rate": 7.687718411169273e-06, "loss": 0.0474, "step": 38174 }, { "epoch": 11.75, "learning_rate": 7.687233647333172e-06, "loss": 0.0207, "step": 38175 }, { "epoch": 11.75, "learning_rate": 7.686748889239049e-06, "loss": 0.0486, "step": 38176 }, { "epoch": 11.75, "learning_rate": 7.6862641368881e-06, "loss": 0.0672, "step": 38177 }, { "epoch": 11.75, "learning_rate": 7.685779390281536e-06, "loss": 0.0204, "step": 38178 }, { "epoch": 11.75, "learning_rate": 7.68529464942056e-06, "loss": 0.0588, "step": 38179 }, { "epoch": 11.75, "learning_rate": 7.68480991430637e-06, "loss": 0.0467, "step": 38180 }, { "epoch": 11.75, "learning_rate": 7.684325184940176e-06, "loss": 0.0707, "step": 38181 }, { "epoch": 11.75, "learning_rate": 7.683840461323175e-06, "loss": 0.1115, "step": 38182 }, { "epoch": 11.75, "learning_rate": 7.683355743456575e-06, "loss": 0.0281, "step": 38183 }, { "epoch": 11.75, "learning_rate": 7.682871031341579e-06, "loss": 0.068, "step": 38184 }, { "epoch": 11.75, "learning_rate": 7.682386324979392e-06, "loss": 0.0402, "step": 38185 }, { "epoch": 11.75, "learning_rate": 7.68190162437121e-06, "loss": 0.0459, "step": 38186 }, { "epoch": 11.75, "learning_rate": 7.68141692951824e-06, "loss": 0.0456, "step": 38187 }, { "epoch": 11.75, "learning_rate": 7.680932240421692e-06, "loss": 0.0371, "step": 38188 }, { "epoch": 11.75, "learning_rate": 7.680447557082763e-06, "loss": 0.0555, "step": 38189 }, { "epoch": 11.75, "learning_rate": 7.679962879502656e-06, "loss": 0.0491, "step": 38190 }, { "epoch": 11.75, "learning_rate": 7.679478207682578e-06, "loss": 0.0411, "step": 38191 }, { "epoch": 11.75, "learning_rate": 7.678993541623725e-06, "loss": 0.055, "step": 38192 }, { "epoch": 11.75, "learning_rate": 7.678508881327309e-06, "loss": 0.0277, "step": 38193 }, { "epoch": 11.75, "learning_rate": 7.67802422679453e-06, "loss": 0.0319, "step": 38194 }, { "epoch": 11.75, "learning_rate": 7.677539578026593e-06, "loss": 0.0561, "step": 38195 }, { "epoch": 11.75, "learning_rate": 7.677054935024696e-06, "loss": 0.0474, "step": 38196 }, { "epoch": 11.75, "learning_rate": 7.676570297790042e-06, "loss": 0.0451, "step": 38197 }, { "epoch": 11.75, "learning_rate": 7.676085666323843e-06, "loss": 0.0367, "step": 38198 }, { "epoch": 11.75, "learning_rate": 7.675601040627294e-06, "loss": 0.0623, "step": 38199 }, { "epoch": 11.75, "learning_rate": 7.675116420701603e-06, "loss": 0.0514, "step": 38200 }, { "epoch": 11.75, "learning_rate": 7.67463180654797e-06, "loss": 0.0244, "step": 38201 }, { "epoch": 11.75, "learning_rate": 7.674147198167597e-06, "loss": 0.0603, "step": 38202 }, { "epoch": 11.75, "learning_rate": 7.673662595561691e-06, "loss": 0.0328, "step": 38203 }, { "epoch": 11.76, "learning_rate": 7.673177998731456e-06, "loss": 0.0622, "step": 38204 }, { "epoch": 11.76, "learning_rate": 7.672693407678092e-06, "loss": 0.0417, "step": 38205 }, { "epoch": 11.76, "learning_rate": 7.672208822402802e-06, "loss": 0.0574, "step": 38206 }, { "epoch": 11.76, "learning_rate": 7.671724242906786e-06, "loss": 0.0356, "step": 38207 }, { "epoch": 11.76, "learning_rate": 7.671239669191257e-06, "loss": 0.0571, "step": 38208 }, { "epoch": 11.76, "learning_rate": 7.670755101257412e-06, "loss": 0.0542, "step": 38209 }, { "epoch": 11.76, "learning_rate": 7.670270539106452e-06, "loss": 0.0652, "step": 38210 }, { "epoch": 11.76, "learning_rate": 7.669785982739582e-06, "loss": 0.0647, "step": 38211 }, { "epoch": 11.76, "learning_rate": 7.669301432158004e-06, "loss": 0.0239, "step": 38212 }, { "epoch": 11.76, "learning_rate": 7.668816887362924e-06, "loss": 0.0884, "step": 38213 }, { "epoch": 11.76, "learning_rate": 7.668332348355542e-06, "loss": 0.04, "step": 38214 }, { "epoch": 11.76, "learning_rate": 7.667847815137064e-06, "loss": 0.0417, "step": 38215 }, { "epoch": 11.76, "learning_rate": 7.667363287708692e-06, "loss": 0.0774, "step": 38216 }, { "epoch": 11.76, "learning_rate": 7.666878766071623e-06, "loss": 0.0546, "step": 38217 }, { "epoch": 11.76, "learning_rate": 7.666394250227068e-06, "loss": 0.0509, "step": 38218 }, { "epoch": 11.76, "learning_rate": 7.665909740176231e-06, "loss": 0.0624, "step": 38219 }, { "epoch": 11.76, "learning_rate": 7.665425235920307e-06, "loss": 0.0991, "step": 38220 }, { "epoch": 11.76, "learning_rate": 7.664940737460502e-06, "loss": 0.0597, "step": 38221 }, { "epoch": 11.76, "learning_rate": 7.66445624479802e-06, "loss": 0.049, "step": 38222 }, { "epoch": 11.76, "learning_rate": 7.663971757934064e-06, "loss": 0.0548, "step": 38223 }, { "epoch": 11.76, "learning_rate": 7.663487276869837e-06, "loss": 0.0497, "step": 38224 }, { "epoch": 11.76, "learning_rate": 7.663002801606541e-06, "loss": 0.0401, "step": 38225 }, { "epoch": 11.76, "learning_rate": 7.662518332145377e-06, "loss": 0.0529, "step": 38226 }, { "epoch": 11.76, "learning_rate": 7.66203386848755e-06, "loss": 0.0569, "step": 38227 }, { "epoch": 11.76, "learning_rate": 7.661549410634265e-06, "loss": 0.0352, "step": 38228 }, { "epoch": 11.76, "learning_rate": 7.66106495858672e-06, "loss": 0.04, "step": 38229 }, { "epoch": 11.76, "learning_rate": 7.660580512346126e-06, "loss": 0.0388, "step": 38230 }, { "epoch": 11.76, "learning_rate": 7.660096071913675e-06, "loss": 0.0407, "step": 38231 }, { "epoch": 11.76, "learning_rate": 7.659611637290573e-06, "loss": 0.0287, "step": 38232 }, { "epoch": 11.76, "learning_rate": 7.659127208478027e-06, "loss": 0.047, "step": 38233 }, { "epoch": 11.76, "learning_rate": 7.658642785477236e-06, "loss": 0.0543, "step": 38234 }, { "epoch": 11.76, "learning_rate": 7.658158368289406e-06, "loss": 0.0461, "step": 38235 }, { "epoch": 11.76, "learning_rate": 7.657673956915735e-06, "loss": 0.0505, "step": 38236 }, { "epoch": 11.77, "learning_rate": 7.657189551357427e-06, "loss": 0.0575, "step": 38237 }, { "epoch": 11.77, "learning_rate": 7.656705151615688e-06, "loss": 0.0398, "step": 38238 }, { "epoch": 11.77, "learning_rate": 7.656220757691718e-06, "loss": 0.0428, "step": 38239 }, { "epoch": 11.77, "learning_rate": 7.655736369586723e-06, "loss": 0.0219, "step": 38240 }, { "epoch": 11.77, "learning_rate": 7.6552519873019e-06, "loss": 0.0373, "step": 38241 }, { "epoch": 11.77, "learning_rate": 7.65476761083845e-06, "loss": 0.0459, "step": 38242 }, { "epoch": 11.77, "learning_rate": 7.654283240197584e-06, "loss": 0.0235, "step": 38243 }, { "epoch": 11.77, "learning_rate": 7.6537988753805e-06, "loss": 0.0344, "step": 38244 }, { "epoch": 11.77, "learning_rate": 7.653314516388402e-06, "loss": 0.0653, "step": 38245 }, { "epoch": 11.77, "learning_rate": 7.65283016322249e-06, "loss": 0.0494, "step": 38246 }, { "epoch": 11.77, "learning_rate": 7.652345815883966e-06, "loss": 0.0609, "step": 38247 }, { "epoch": 11.77, "learning_rate": 7.651861474374036e-06, "loss": 0.0762, "step": 38248 }, { "epoch": 11.77, "learning_rate": 7.651377138693903e-06, "loss": 0.0437, "step": 38249 }, { "epoch": 11.77, "learning_rate": 7.650892808844765e-06, "loss": 0.0545, "step": 38250 }, { "epoch": 11.77, "learning_rate": 7.650408484827831e-06, "loss": 0.0625, "step": 38251 }, { "epoch": 11.77, "learning_rate": 7.649924166644291e-06, "loss": 0.0395, "step": 38252 }, { "epoch": 11.77, "learning_rate": 7.649439854295364e-06, "loss": 0.0317, "step": 38253 }, { "epoch": 11.77, "learning_rate": 7.648955547782239e-06, "loss": 0.0445, "step": 38254 }, { "epoch": 11.77, "learning_rate": 7.648471247106127e-06, "loss": 0.0593, "step": 38255 }, { "epoch": 11.77, "learning_rate": 7.647986952268224e-06, "loss": 0.0847, "step": 38256 }, { "epoch": 11.77, "learning_rate": 7.647502663269735e-06, "loss": 0.0561, "step": 38257 }, { "epoch": 11.77, "learning_rate": 7.647018380111865e-06, "loss": 0.0185, "step": 38258 }, { "epoch": 11.77, "learning_rate": 7.646534102795814e-06, "loss": 0.0396, "step": 38259 }, { "epoch": 11.77, "learning_rate": 7.646049831322783e-06, "loss": 0.0519, "step": 38260 }, { "epoch": 11.77, "learning_rate": 7.64556556569398e-06, "loss": 0.0684, "step": 38261 }, { "epoch": 11.77, "learning_rate": 7.645081305910596e-06, "loss": 0.0531, "step": 38262 }, { "epoch": 11.77, "learning_rate": 7.644597051973843e-06, "loss": 0.0314, "step": 38263 }, { "epoch": 11.77, "learning_rate": 7.644112803884923e-06, "loss": 0.0424, "step": 38264 }, { "epoch": 11.77, "learning_rate": 7.643628561645034e-06, "loss": 0.054, "step": 38265 }, { "epoch": 11.77, "learning_rate": 7.64314432525538e-06, "loss": 0.0298, "step": 38266 }, { "epoch": 11.77, "learning_rate": 7.642660094717162e-06, "loss": 0.0363, "step": 38267 }, { "epoch": 11.77, "learning_rate": 7.642175870031584e-06, "loss": 0.0541, "step": 38268 }, { "epoch": 11.78, "learning_rate": 7.641691651199848e-06, "loss": 0.0296, "step": 38269 }, { "epoch": 11.78, "learning_rate": 7.641207438223157e-06, "loss": 0.0453, "step": 38270 }, { "epoch": 11.78, "learning_rate": 7.640723231102712e-06, "loss": 0.0683, "step": 38271 }, { "epoch": 11.78, "learning_rate": 7.64023902983971e-06, "loss": 0.0388, "step": 38272 }, { "epoch": 11.78, "learning_rate": 7.639754834435365e-06, "loss": 0.0494, "step": 38273 }, { "epoch": 11.78, "learning_rate": 7.639270644890873e-06, "loss": 0.0552, "step": 38274 }, { "epoch": 11.78, "learning_rate": 7.638786461207433e-06, "loss": 0.0496, "step": 38275 }, { "epoch": 11.78, "learning_rate": 7.638302283386248e-06, "loss": 0.0497, "step": 38276 }, { "epoch": 11.78, "learning_rate": 7.637818111428522e-06, "loss": 0.0531, "step": 38277 }, { "epoch": 11.78, "learning_rate": 7.637333945335458e-06, "loss": 0.0434, "step": 38278 }, { "epoch": 11.78, "learning_rate": 7.636849785108256e-06, "loss": 0.0424, "step": 38279 }, { "epoch": 11.78, "learning_rate": 7.63636563074812e-06, "loss": 0.039, "step": 38280 }, { "epoch": 11.78, "learning_rate": 7.63588148225625e-06, "loss": 0.0712, "step": 38281 }, { "epoch": 11.78, "learning_rate": 7.635397339633847e-06, "loss": 0.0744, "step": 38282 }, { "epoch": 11.78, "learning_rate": 7.634913202882119e-06, "loss": 0.0748, "step": 38283 }, { "epoch": 11.78, "learning_rate": 7.634429072002263e-06, "loss": 0.0448, "step": 38284 }, { "epoch": 11.78, "learning_rate": 7.633944946995483e-06, "loss": 0.1012, "step": 38285 }, { "epoch": 11.78, "learning_rate": 7.633460827862977e-06, "loss": 0.0447, "step": 38286 }, { "epoch": 11.78, "learning_rate": 7.632976714605947e-06, "loss": 0.0531, "step": 38287 }, { "epoch": 11.78, "learning_rate": 7.632492607225604e-06, "loss": 0.0245, "step": 38288 }, { "epoch": 11.78, "learning_rate": 7.632008505723141e-06, "loss": 0.0976, "step": 38289 }, { "epoch": 11.78, "learning_rate": 7.631524410099763e-06, "loss": 0.0899, "step": 38290 }, { "epoch": 11.78, "learning_rate": 7.631040320356671e-06, "loss": 0.0545, "step": 38291 }, { "epoch": 11.78, "learning_rate": 7.630556236495066e-06, "loss": 0.0407, "step": 38292 }, { "epoch": 11.78, "learning_rate": 7.630072158516151e-06, "loss": 0.0444, "step": 38293 }, { "epoch": 11.78, "learning_rate": 7.62958808642113e-06, "loss": 0.0927, "step": 38294 }, { "epoch": 11.78, "learning_rate": 7.629104020211204e-06, "loss": 0.0371, "step": 38295 }, { "epoch": 11.78, "learning_rate": 7.628619959887572e-06, "loss": 0.0598, "step": 38296 }, { "epoch": 11.78, "learning_rate": 7.628135905451433e-06, "loss": 0.0444, "step": 38297 }, { "epoch": 11.78, "learning_rate": 7.627651856903999e-06, "loss": 0.0864, "step": 38298 }, { "epoch": 11.78, "learning_rate": 7.627167814246464e-06, "loss": 0.0564, "step": 38299 }, { "epoch": 11.78, "learning_rate": 7.6266837774800315e-06, "loss": 0.0738, "step": 38300 }, { "epoch": 11.78, "learning_rate": 7.6261997466059035e-06, "loss": 0.0399, "step": 38301 }, { "epoch": 11.79, "learning_rate": 7.62571572162528e-06, "loss": 0.0729, "step": 38302 }, { "epoch": 11.79, "learning_rate": 7.6252317025393655e-06, "loss": 0.0424, "step": 38303 }, { "epoch": 11.79, "learning_rate": 7.62474768934936e-06, "loss": 0.0489, "step": 38304 }, { "epoch": 11.79, "learning_rate": 7.624263682056467e-06, "loss": 0.0376, "step": 38305 }, { "epoch": 11.79, "learning_rate": 7.623779680661888e-06, "loss": 0.0443, "step": 38306 }, { "epoch": 11.79, "learning_rate": 7.623295685166817e-06, "loss": 0.0503, "step": 38307 }, { "epoch": 11.79, "learning_rate": 7.622811695572467e-06, "loss": 0.0397, "step": 38308 }, { "epoch": 11.79, "learning_rate": 7.622327711880037e-06, "loss": 0.0682, "step": 38309 }, { "epoch": 11.79, "learning_rate": 7.621843734090724e-06, "loss": 0.0193, "step": 38310 }, { "epoch": 11.79, "learning_rate": 7.6213597622057315e-06, "loss": 0.0468, "step": 38311 }, { "epoch": 11.79, "learning_rate": 7.620875796226258e-06, "loss": 0.0699, "step": 38312 }, { "epoch": 11.79, "learning_rate": 7.620391836153512e-06, "loss": 0.0541, "step": 38313 }, { "epoch": 11.79, "learning_rate": 7.619907881988692e-06, "loss": 0.0191, "step": 38314 }, { "epoch": 11.79, "learning_rate": 7.619423933732999e-06, "loss": 0.044, "step": 38315 }, { "epoch": 11.79, "learning_rate": 7.618939991387636e-06, "loss": 0.0338, "step": 38316 }, { "epoch": 11.79, "learning_rate": 7.618456054953796e-06, "loss": 0.0634, "step": 38317 }, { "epoch": 11.79, "learning_rate": 7.617972124432693e-06, "loss": 0.0545, "step": 38318 }, { "epoch": 11.79, "learning_rate": 7.6174881998255255e-06, "loss": 0.0618, "step": 38319 }, { "epoch": 11.79, "learning_rate": 7.61700428113349e-06, "loss": 0.0336, "step": 38320 }, { "epoch": 11.79, "learning_rate": 7.61652036835779e-06, "loss": 0.0561, "step": 38321 }, { "epoch": 11.79, "learning_rate": 7.616036461499626e-06, "loss": 0.013, "step": 38322 }, { "epoch": 11.79, "learning_rate": 7.6155525605602024e-06, "loss": 0.0845, "step": 38323 }, { "epoch": 11.79, "learning_rate": 7.615068665540719e-06, "loss": 0.1032, "step": 38324 }, { "epoch": 11.79, "learning_rate": 7.614584776442377e-06, "loss": 0.0352, "step": 38325 }, { "epoch": 11.79, "learning_rate": 7.6141008932663786e-06, "loss": 0.0634, "step": 38326 }, { "epoch": 11.79, "learning_rate": 7.613617016013921e-06, "loss": 0.0305, "step": 38327 }, { "epoch": 11.79, "learning_rate": 7.613133144686212e-06, "loss": 0.0522, "step": 38328 }, { "epoch": 11.79, "learning_rate": 7.61264927928445e-06, "loss": 0.0479, "step": 38329 }, { "epoch": 11.79, "learning_rate": 7.612165419809839e-06, "loss": 0.0426, "step": 38330 }, { "epoch": 11.79, "learning_rate": 7.6116815662635735e-06, "loss": 0.0291, "step": 38331 }, { "epoch": 11.79, "learning_rate": 7.611197718646859e-06, "loss": 0.0618, "step": 38332 }, { "epoch": 11.79, "learning_rate": 7.610713876960897e-06, "loss": 0.0392, "step": 38333 }, { "epoch": 11.8, "learning_rate": 7.610230041206888e-06, "loss": 0.0778, "step": 38334 }, { "epoch": 11.8, "learning_rate": 7.6097462113860355e-06, "loss": 0.078, "step": 38335 }, { "epoch": 11.8, "learning_rate": 7.609262387499536e-06, "loss": 0.0358, "step": 38336 }, { "epoch": 11.8, "learning_rate": 7.608778569548594e-06, "loss": 0.0521, "step": 38337 }, { "epoch": 11.8, "learning_rate": 7.608294757534411e-06, "loss": 0.0908, "step": 38338 }, { "epoch": 11.8, "learning_rate": 7.607810951458188e-06, "loss": 0.0627, "step": 38339 }, { "epoch": 11.8, "learning_rate": 7.607327151321127e-06, "loss": 0.0626, "step": 38340 }, { "epoch": 11.8, "learning_rate": 7.606843357124426e-06, "loss": 0.0447, "step": 38341 }, { "epoch": 11.8, "learning_rate": 7.606359568869282e-06, "loss": 0.0283, "step": 38342 }, { "epoch": 11.8, "learning_rate": 7.60587578655691e-06, "loss": 0.0778, "step": 38343 }, { "epoch": 11.8, "learning_rate": 7.6053920101885e-06, "loss": 0.0751, "step": 38344 }, { "epoch": 11.8, "learning_rate": 7.604908239765258e-06, "loss": 0.0519, "step": 38345 }, { "epoch": 11.8, "learning_rate": 7.604424475288381e-06, "loss": 0.0347, "step": 38346 }, { "epoch": 11.8, "learning_rate": 7.60394071675907e-06, "loss": 0.0468, "step": 38347 }, { "epoch": 11.8, "learning_rate": 7.60345696417853e-06, "loss": 0.051, "step": 38348 }, { "epoch": 11.8, "learning_rate": 7.602973217547962e-06, "loss": 0.0392, "step": 38349 }, { "epoch": 11.8, "learning_rate": 7.602489476868565e-06, "loss": 0.0263, "step": 38350 }, { "epoch": 11.8, "learning_rate": 7.602005742141542e-06, "loss": 0.0595, "step": 38351 }, { "epoch": 11.8, "learning_rate": 7.601522013368086e-06, "loss": 0.0876, "step": 38352 }, { "epoch": 11.8, "learning_rate": 7.60103829054941e-06, "loss": 0.0307, "step": 38353 }, { "epoch": 11.8, "learning_rate": 7.600554573686707e-06, "loss": 0.0784, "step": 38354 }, { "epoch": 11.8, "learning_rate": 7.600070862781181e-06, "loss": 0.0635, "step": 38355 }, { "epoch": 11.8, "learning_rate": 7.599587157834032e-06, "loss": 0.0461, "step": 38356 }, { "epoch": 11.8, "learning_rate": 7.599103458846458e-06, "loss": 0.0509, "step": 38357 }, { "epoch": 11.8, "learning_rate": 7.598619765819666e-06, "loss": 0.0577, "step": 38358 }, { "epoch": 11.8, "learning_rate": 7.598136078754853e-06, "loss": 0.0401, "step": 38359 }, { "epoch": 11.8, "learning_rate": 7.597652397653221e-06, "loss": 0.0519, "step": 38360 }, { "epoch": 11.8, "learning_rate": 7.597168722515973e-06, "loss": 0.0123, "step": 38361 }, { "epoch": 11.8, "learning_rate": 7.596685053344299e-06, "loss": 0.0594, "step": 38362 }, { "epoch": 11.8, "learning_rate": 7.5962013901394135e-06, "loss": 0.0657, "step": 38363 }, { "epoch": 11.8, "learning_rate": 7.595717732902515e-06, "loss": 0.0355, "step": 38364 }, { "epoch": 11.8, "learning_rate": 7.595234081634798e-06, "loss": 0.0569, "step": 38365 }, { "epoch": 11.8, "learning_rate": 7.594750436337467e-06, "loss": 0.0764, "step": 38366 }, { "epoch": 11.81, "learning_rate": 7.59426679701172e-06, "loss": 0.079, "step": 38367 }, { "epoch": 11.81, "learning_rate": 7.593783163658761e-06, "loss": 0.0402, "step": 38368 }, { "epoch": 11.81, "learning_rate": 7.593299536279791e-06, "loss": 0.0903, "step": 38369 }, { "epoch": 11.81, "learning_rate": 7.592815914876009e-06, "loss": 0.071, "step": 38370 }, { "epoch": 11.81, "learning_rate": 7.592332299448618e-06, "loss": 0.0668, "step": 38371 }, { "epoch": 11.81, "learning_rate": 7.591848689998811e-06, "loss": 0.0553, "step": 38372 }, { "epoch": 11.81, "learning_rate": 7.5913650865277986e-06, "loss": 0.0918, "step": 38373 }, { "epoch": 11.81, "learning_rate": 7.590881489036779e-06, "loss": 0.0443, "step": 38374 }, { "epoch": 11.81, "learning_rate": 7.59039789752695e-06, "loss": 0.0367, "step": 38375 }, { "epoch": 11.81, "learning_rate": 7.589914311999512e-06, "loss": 0.0435, "step": 38376 }, { "epoch": 11.81, "learning_rate": 7.589430732455666e-06, "loss": 0.0317, "step": 38377 }, { "epoch": 11.81, "learning_rate": 7.588947158896615e-06, "loss": 0.0386, "step": 38378 }, { "epoch": 11.81, "learning_rate": 7.588463591323559e-06, "loss": 0.0436, "step": 38379 }, { "epoch": 11.81, "learning_rate": 7.587980029737698e-06, "loss": 0.0492, "step": 38380 }, { "epoch": 11.81, "learning_rate": 7.587496474140231e-06, "loss": 0.0324, "step": 38381 }, { "epoch": 11.81, "learning_rate": 7.587012924532359e-06, "loss": 0.0449, "step": 38382 }, { "epoch": 11.81, "learning_rate": 7.586529380915285e-06, "loss": 0.0448, "step": 38383 }, { "epoch": 11.81, "learning_rate": 7.586045843290207e-06, "loss": 0.0623, "step": 38384 }, { "epoch": 11.81, "learning_rate": 7.58556231165833e-06, "loss": 0.0319, "step": 38385 }, { "epoch": 11.81, "learning_rate": 7.585078786020848e-06, "loss": 0.078, "step": 38386 }, { "epoch": 11.81, "learning_rate": 7.58459526637896e-06, "loss": 0.041, "step": 38387 }, { "epoch": 11.81, "learning_rate": 7.584111752733878e-06, "loss": 0.0407, "step": 38388 }, { "epoch": 11.81, "learning_rate": 7.5836282450867925e-06, "loss": 0.0668, "step": 38389 }, { "epoch": 11.81, "learning_rate": 7.583144743438906e-06, "loss": 0.0409, "step": 38390 }, { "epoch": 11.81, "learning_rate": 7.582661247791421e-06, "loss": 0.0432, "step": 38391 }, { "epoch": 11.81, "learning_rate": 7.582177758145532e-06, "loss": 0.0974, "step": 38392 }, { "epoch": 11.81, "learning_rate": 7.581694274502449e-06, "loss": 0.0698, "step": 38393 }, { "epoch": 11.81, "learning_rate": 7.5812107968633656e-06, "loss": 0.0576, "step": 38394 }, { "epoch": 11.81, "learning_rate": 7.580727325229486e-06, "loss": 0.0327, "step": 38395 }, { "epoch": 11.81, "learning_rate": 7.580243859602006e-06, "loss": 0.0644, "step": 38396 }, { "epoch": 11.81, "learning_rate": 7.5797603999821235e-06, "loss": 0.0394, "step": 38397 }, { "epoch": 11.81, "learning_rate": 7.579276946371049e-06, "loss": 0.032, "step": 38398 }, { "epoch": 11.82, "learning_rate": 7.578793498769976e-06, "loss": 0.0693, "step": 38399 }, { "epoch": 11.82, "learning_rate": 7.5783100571801075e-06, "loss": 0.0527, "step": 38400 }, { "epoch": 11.82, "learning_rate": 7.57782662160264e-06, "loss": 0.0739, "step": 38401 }, { "epoch": 11.82, "learning_rate": 7.577343192038775e-06, "loss": 0.0519, "step": 38402 }, { "epoch": 11.82, "learning_rate": 7.576859768489716e-06, "loss": 0.0563, "step": 38403 }, { "epoch": 11.82, "learning_rate": 7.57637635095666e-06, "loss": 0.0585, "step": 38404 }, { "epoch": 11.82, "learning_rate": 7.57589293944081e-06, "loss": 0.0335, "step": 38405 }, { "epoch": 11.82, "learning_rate": 7.575409533943364e-06, "loss": 0.0342, "step": 38406 }, { "epoch": 11.82, "learning_rate": 7.574926134465518e-06, "loss": 0.0379, "step": 38407 }, { "epoch": 11.82, "learning_rate": 7.5744427410084794e-06, "loss": 0.0488, "step": 38408 }, { "epoch": 11.82, "learning_rate": 7.573959353573449e-06, "loss": 0.0959, "step": 38409 }, { "epoch": 11.82, "learning_rate": 7.573475972161622e-06, "loss": 0.0442, "step": 38410 }, { "epoch": 11.82, "learning_rate": 7.572992596774198e-06, "loss": 0.0663, "step": 38411 }, { "epoch": 11.82, "learning_rate": 7.572509227412378e-06, "loss": 0.0355, "step": 38412 }, { "epoch": 11.82, "learning_rate": 7.572025864077365e-06, "loss": 0.0332, "step": 38413 }, { "epoch": 11.82, "learning_rate": 7.571542506770357e-06, "loss": 0.0425, "step": 38414 }, { "epoch": 11.82, "learning_rate": 7.571059155492555e-06, "loss": 0.0784, "step": 38415 }, { "epoch": 11.82, "learning_rate": 7.57057581024516e-06, "loss": 0.0363, "step": 38416 }, { "epoch": 11.82, "learning_rate": 7.570092471029363e-06, "loss": 0.0313, "step": 38417 }, { "epoch": 11.82, "learning_rate": 7.569609137846376e-06, "loss": 0.0604, "step": 38418 }, { "epoch": 11.82, "learning_rate": 7.569125810697397e-06, "loss": 0.0477, "step": 38419 }, { "epoch": 11.82, "learning_rate": 7.568642489583622e-06, "loss": 0.0561, "step": 38420 }, { "epoch": 11.82, "learning_rate": 7.568159174506246e-06, "loss": 0.0478, "step": 38421 }, { "epoch": 11.82, "learning_rate": 7.567675865466484e-06, "loss": 0.0595, "step": 38422 }, { "epoch": 11.82, "learning_rate": 7.567192562465523e-06, "loss": 0.0281, "step": 38423 }, { "epoch": 11.82, "learning_rate": 7.566709265504567e-06, "loss": 0.0177, "step": 38424 }, { "epoch": 11.82, "learning_rate": 7.566225974584816e-06, "loss": 0.0436, "step": 38425 }, { "epoch": 11.82, "learning_rate": 7.5657426897074694e-06, "loss": 0.0745, "step": 38426 }, { "epoch": 11.82, "learning_rate": 7.565259410873727e-06, "loss": 0.0496, "step": 38427 }, { "epoch": 11.82, "learning_rate": 7.564776138084791e-06, "loss": 0.043, "step": 38428 }, { "epoch": 11.82, "learning_rate": 7.564292871341861e-06, "loss": 0.0673, "step": 38429 }, { "epoch": 11.82, "learning_rate": 7.563809610646132e-06, "loss": 0.0711, "step": 38430 }, { "epoch": 11.82, "learning_rate": 7.5633263559988035e-06, "loss": 0.0394, "step": 38431 }, { "epoch": 11.83, "learning_rate": 7.562843107401086e-06, "loss": 0.094, "step": 38432 }, { "epoch": 11.83, "learning_rate": 7.562359864854168e-06, "loss": 0.0542, "step": 38433 }, { "epoch": 11.83, "learning_rate": 7.5618766283592545e-06, "loss": 0.0206, "step": 38434 }, { "epoch": 11.83, "learning_rate": 7.561393397917545e-06, "loss": 0.0287, "step": 38435 }, { "epoch": 11.83, "learning_rate": 7.5609101735302344e-06, "loss": 0.0444, "step": 38436 }, { "epoch": 11.83, "learning_rate": 7.56042695519853e-06, "loss": 0.0228, "step": 38437 }, { "epoch": 11.83, "learning_rate": 7.559943742923626e-06, "loss": 0.0616, "step": 38438 }, { "epoch": 11.83, "learning_rate": 7.559460536706724e-06, "loss": 0.0354, "step": 38439 }, { "epoch": 11.83, "learning_rate": 7.558977336549027e-06, "loss": 0.0214, "step": 38440 }, { "epoch": 11.83, "learning_rate": 7.558494142451724e-06, "loss": 0.0233, "step": 38441 }, { "epoch": 11.83, "learning_rate": 7.558010954416026e-06, "loss": 0.0596, "step": 38442 }, { "epoch": 11.83, "learning_rate": 7.557527772443132e-06, "loss": 0.0455, "step": 38443 }, { "epoch": 11.83, "learning_rate": 7.557044596534234e-06, "loss": 0.0581, "step": 38444 }, { "epoch": 11.83, "learning_rate": 7.556561426690536e-06, "loss": 0.0311, "step": 38445 }, { "epoch": 11.83, "learning_rate": 7.556078262913236e-06, "loss": 0.0601, "step": 38446 }, { "epoch": 11.83, "learning_rate": 7.555595105203537e-06, "loss": 0.0607, "step": 38447 }, { "epoch": 11.83, "learning_rate": 7.555111953562635e-06, "loss": 0.0322, "step": 38448 }, { "epoch": 11.83, "learning_rate": 7.554628807991733e-06, "loss": 0.0501, "step": 38449 }, { "epoch": 11.83, "learning_rate": 7.554145668492029e-06, "loss": 0.0319, "step": 38450 }, { "epoch": 11.83, "learning_rate": 7.553662535064715e-06, "loss": 0.0544, "step": 38451 }, { "epoch": 11.83, "learning_rate": 7.553179407711003e-06, "loss": 0.0472, "step": 38452 }, { "epoch": 11.83, "learning_rate": 7.5526962864320884e-06, "loss": 0.0313, "step": 38453 }, { "epoch": 11.83, "learning_rate": 7.552213171229168e-06, "loss": 0.0191, "step": 38454 }, { "epoch": 11.83, "learning_rate": 7.5517300621034406e-06, "loss": 0.0456, "step": 38455 }, { "epoch": 11.83, "learning_rate": 7.551246959056106e-06, "loss": 0.0268, "step": 38456 }, { "epoch": 11.83, "learning_rate": 7.550763862088369e-06, "loss": 0.0552, "step": 38457 }, { "epoch": 11.83, "learning_rate": 7.550280771201423e-06, "loss": 0.0565, "step": 38458 }, { "epoch": 11.83, "learning_rate": 7.5497976863964694e-06, "loss": 0.0682, "step": 38459 }, { "epoch": 11.83, "learning_rate": 7.549314607674708e-06, "loss": 0.055, "step": 38460 }, { "epoch": 11.83, "learning_rate": 7.548831535037336e-06, "loss": 0.0364, "step": 38461 }, { "epoch": 11.83, "learning_rate": 7.548348468485556e-06, "loss": 0.0662, "step": 38462 }, { "epoch": 11.83, "learning_rate": 7.547865408020566e-06, "loss": 0.0341, "step": 38463 }, { "epoch": 11.84, "learning_rate": 7.547382353643568e-06, "loss": 0.0184, "step": 38464 }, { "epoch": 11.84, "learning_rate": 7.546899305355755e-06, "loss": 0.0386, "step": 38465 }, { "epoch": 11.84, "learning_rate": 7.546416263158326e-06, "loss": 0.0221, "step": 38466 }, { "epoch": 11.84, "learning_rate": 7.545933227052491e-06, "loss": 0.0457, "step": 38467 }, { "epoch": 11.84, "learning_rate": 7.545450197039439e-06, "loss": 0.029, "step": 38468 }, { "epoch": 11.84, "learning_rate": 7.544967173120372e-06, "loss": 0.0405, "step": 38469 }, { "epoch": 11.84, "learning_rate": 7.544484155296492e-06, "loss": 0.039, "step": 38470 }, { "epoch": 11.84, "learning_rate": 7.5440011435689915e-06, "loss": 0.0137, "step": 38471 }, { "epoch": 11.84, "learning_rate": 7.5435181379390765e-06, "loss": 0.0754, "step": 38472 }, { "epoch": 11.84, "learning_rate": 7.5430351384079444e-06, "loss": 0.0543, "step": 38473 }, { "epoch": 11.84, "learning_rate": 7.5425521449767955e-06, "loss": 0.0643, "step": 38474 }, { "epoch": 11.84, "learning_rate": 7.542069157646825e-06, "loss": 0.0556, "step": 38475 }, { "epoch": 11.84, "learning_rate": 7.541586176419231e-06, "loss": 0.0529, "step": 38476 }, { "epoch": 11.84, "learning_rate": 7.54110320129522e-06, "loss": 0.0964, "step": 38477 }, { "epoch": 11.84, "learning_rate": 7.540620232275986e-06, "loss": 0.0712, "step": 38478 }, { "epoch": 11.84, "learning_rate": 7.54013726936273e-06, "loss": 0.0508, "step": 38479 }, { "epoch": 11.84, "learning_rate": 7.539654312556648e-06, "loss": 0.0333, "step": 38480 }, { "epoch": 11.84, "learning_rate": 7.539171361858939e-06, "loss": 0.0652, "step": 38481 }, { "epoch": 11.84, "learning_rate": 7.538688417270808e-06, "loss": 0.0724, "step": 38482 }, { "epoch": 11.84, "learning_rate": 7.538205478793448e-06, "loss": 0.0674, "step": 38483 }, { "epoch": 11.84, "learning_rate": 7.537722546428061e-06, "loss": 0.0283, "step": 38484 }, { "epoch": 11.84, "learning_rate": 7.537239620175848e-06, "loss": 0.0499, "step": 38485 }, { "epoch": 11.84, "learning_rate": 7.536756700037998e-06, "loss": 0.0538, "step": 38486 }, { "epoch": 11.84, "learning_rate": 7.536273786015721e-06, "loss": 0.0306, "step": 38487 }, { "epoch": 11.84, "learning_rate": 7.535790878110215e-06, "loss": 0.0474, "step": 38488 }, { "epoch": 11.84, "learning_rate": 7.535307976322674e-06, "loss": 0.0363, "step": 38489 }, { "epoch": 11.84, "learning_rate": 7.534825080654298e-06, "loss": 0.0258, "step": 38490 }, { "epoch": 11.84, "learning_rate": 7.534342191106284e-06, "loss": 0.0366, "step": 38491 }, { "epoch": 11.84, "learning_rate": 7.533859307679836e-06, "loss": 0.0664, "step": 38492 }, { "epoch": 11.84, "learning_rate": 7.533376430376151e-06, "loss": 0.0615, "step": 38493 }, { "epoch": 11.84, "learning_rate": 7.532893559196428e-06, "loss": 0.045, "step": 38494 }, { "epoch": 11.84, "learning_rate": 7.532410694141866e-06, "loss": 0.0526, "step": 38495 }, { "epoch": 11.84, "learning_rate": 7.531927835213657e-06, "loss": 0.0381, "step": 38496 }, { "epoch": 11.85, "learning_rate": 7.531444982413011e-06, "loss": 0.0404, "step": 38497 }, { "epoch": 11.85, "learning_rate": 7.5309621357411224e-06, "loss": 0.0534, "step": 38498 }, { "epoch": 11.85, "learning_rate": 7.5304792951991875e-06, "loss": 0.0413, "step": 38499 }, { "epoch": 11.85, "learning_rate": 7.529996460788407e-06, "loss": 0.0591, "step": 38500 }, { "epoch": 11.85, "learning_rate": 7.529513632509977e-06, "loss": 0.0627, "step": 38501 }, { "epoch": 11.85, "learning_rate": 7.529030810365101e-06, "loss": 0.0526, "step": 38502 }, { "epoch": 11.85, "learning_rate": 7.528547994354976e-06, "loss": 0.0392, "step": 38503 }, { "epoch": 11.85, "learning_rate": 7.5280651844807995e-06, "loss": 0.0366, "step": 38504 }, { "epoch": 11.85, "learning_rate": 7.52758238074377e-06, "loss": 0.043, "step": 38505 }, { "epoch": 11.85, "learning_rate": 7.527099583145085e-06, "loss": 0.0616, "step": 38506 }, { "epoch": 11.85, "learning_rate": 7.526616791685948e-06, "loss": 0.0192, "step": 38507 }, { "epoch": 11.85, "learning_rate": 7.526134006367556e-06, "loss": 0.0584, "step": 38508 }, { "epoch": 11.85, "learning_rate": 7.525651227191105e-06, "loss": 0.028, "step": 38509 }, { "epoch": 11.85, "learning_rate": 7.525168454157794e-06, "loss": 0.0436, "step": 38510 }, { "epoch": 11.85, "learning_rate": 7.5246856872688216e-06, "loss": 0.0361, "step": 38511 }, { "epoch": 11.85, "learning_rate": 7.524202926525388e-06, "loss": 0.064, "step": 38512 }, { "epoch": 11.85, "learning_rate": 7.523720171928692e-06, "loss": 0.0658, "step": 38513 }, { "epoch": 11.85, "learning_rate": 7.523237423479932e-06, "loss": 0.0459, "step": 38514 }, { "epoch": 11.85, "learning_rate": 7.522754681180304e-06, "loss": 0.0627, "step": 38515 }, { "epoch": 11.85, "learning_rate": 7.5222719450310075e-06, "loss": 0.0646, "step": 38516 }, { "epoch": 11.85, "learning_rate": 7.521789215033243e-06, "loss": 0.0455, "step": 38517 }, { "epoch": 11.85, "learning_rate": 7.521306491188208e-06, "loss": 0.0407, "step": 38518 }, { "epoch": 11.85, "learning_rate": 7.520823773497103e-06, "loss": 0.0337, "step": 38519 }, { "epoch": 11.85, "learning_rate": 7.520341061961123e-06, "loss": 0.0869, "step": 38520 }, { "epoch": 11.85, "learning_rate": 7.519858356581462e-06, "loss": 0.059, "step": 38521 }, { "epoch": 11.85, "learning_rate": 7.519375657359331e-06, "loss": 0.0855, "step": 38522 }, { "epoch": 11.85, "learning_rate": 7.51889296429592e-06, "loss": 0.0416, "step": 38523 }, { "epoch": 11.85, "learning_rate": 7.518410277392428e-06, "loss": 0.0416, "step": 38524 }, { "epoch": 11.85, "learning_rate": 7.517927596650054e-06, "loss": 0.0625, "step": 38525 }, { "epoch": 11.85, "learning_rate": 7.517444922069995e-06, "loss": 0.0686, "step": 38526 }, { "epoch": 11.85, "learning_rate": 7.516962253653452e-06, "loss": 0.0537, "step": 38527 }, { "epoch": 11.85, "learning_rate": 7.516479591401624e-06, "loss": 0.0649, "step": 38528 }, { "epoch": 11.86, "learning_rate": 7.5159969353157105e-06, "loss": 0.0376, "step": 38529 }, { "epoch": 11.86, "learning_rate": 7.515514285396903e-06, "loss": 0.0927, "step": 38530 }, { "epoch": 11.86, "learning_rate": 7.5150316416464e-06, "loss": 0.0342, "step": 38531 }, { "epoch": 11.86, "learning_rate": 7.51454900406541e-06, "loss": 0.0458, "step": 38532 }, { "epoch": 11.86, "learning_rate": 7.5140663726551226e-06, "loss": 0.0757, "step": 38533 }, { "epoch": 11.86, "learning_rate": 7.513583747416738e-06, "loss": 0.0603, "step": 38534 }, { "epoch": 11.86, "learning_rate": 7.513101128351454e-06, "loss": 0.0384, "step": 38535 }, { "epoch": 11.86, "learning_rate": 7.512618515460468e-06, "loss": 0.0552, "step": 38536 }, { "epoch": 11.86, "learning_rate": 7.512135908744982e-06, "loss": 0.0498, "step": 38537 }, { "epoch": 11.86, "learning_rate": 7.511653308206192e-06, "loss": 0.0264, "step": 38538 }, { "epoch": 11.86, "learning_rate": 7.511170713845295e-06, "loss": 0.0483, "step": 38539 }, { "epoch": 11.86, "learning_rate": 7.5106881256634936e-06, "loss": 0.0619, "step": 38540 }, { "epoch": 11.86, "learning_rate": 7.510205543661975e-06, "loss": 0.0698, "step": 38541 }, { "epoch": 11.86, "learning_rate": 7.50972296784195e-06, "loss": 0.047, "step": 38542 }, { "epoch": 11.86, "learning_rate": 7.509240398204614e-06, "loss": 0.0457, "step": 38543 }, { "epoch": 11.86, "learning_rate": 7.508757834751159e-06, "loss": 0.0457, "step": 38544 }, { "epoch": 11.86, "learning_rate": 7.508275277482787e-06, "loss": 0.0399, "step": 38545 }, { "epoch": 11.86, "learning_rate": 7.507792726400694e-06, "loss": 0.0452, "step": 38546 }, { "epoch": 11.86, "learning_rate": 7.507310181506082e-06, "loss": 0.0554, "step": 38547 }, { "epoch": 11.86, "learning_rate": 7.506827642800146e-06, "loss": 0.0344, "step": 38548 }, { "epoch": 11.86, "learning_rate": 7.5063451102840855e-06, "loss": 0.0528, "step": 38549 }, { "epoch": 11.86, "learning_rate": 7.5058625839591e-06, "loss": 0.0607, "step": 38550 }, { "epoch": 11.86, "learning_rate": 7.505380063826379e-06, "loss": 0.0784, "step": 38551 }, { "epoch": 11.86, "learning_rate": 7.504897549887131e-06, "loss": 0.0364, "step": 38552 }, { "epoch": 11.86, "learning_rate": 7.504415042142552e-06, "loss": 0.0494, "step": 38553 }, { "epoch": 11.86, "learning_rate": 7.503932540593835e-06, "loss": 0.0637, "step": 38554 }, { "epoch": 11.86, "learning_rate": 7.503450045242181e-06, "loss": 0.0734, "step": 38555 }, { "epoch": 11.86, "learning_rate": 7.502967556088785e-06, "loss": 0.1279, "step": 38556 }, { "epoch": 11.86, "learning_rate": 7.5024850731348495e-06, "loss": 0.0613, "step": 38557 }, { "epoch": 11.86, "learning_rate": 7.50200259638157e-06, "loss": 0.0524, "step": 38558 }, { "epoch": 11.86, "learning_rate": 7.501520125830146e-06, "loss": 0.0564, "step": 38559 }, { "epoch": 11.86, "learning_rate": 7.501037661481773e-06, "loss": 0.076, "step": 38560 }, { "epoch": 11.86, "learning_rate": 7.500555203337647e-06, "loss": 0.0544, "step": 38561 }, { "epoch": 11.87, "learning_rate": 7.500072751398972e-06, "loss": 0.0681, "step": 38562 }, { "epoch": 11.87, "learning_rate": 7.499590305666942e-06, "loss": 0.0425, "step": 38563 }, { "epoch": 11.87, "learning_rate": 7.499107866142757e-06, "loss": 0.0183, "step": 38564 }, { "epoch": 11.87, "learning_rate": 7.498625432827611e-06, "loss": 0.0363, "step": 38565 }, { "epoch": 11.87, "learning_rate": 7.4981430057227015e-06, "loss": 0.0403, "step": 38566 }, { "epoch": 11.87, "learning_rate": 7.49766058482923e-06, "loss": 0.0427, "step": 38567 }, { "epoch": 11.87, "learning_rate": 7.497178170148394e-06, "loss": 0.0336, "step": 38568 }, { "epoch": 11.87, "learning_rate": 7.49669576168139e-06, "loss": 0.0327, "step": 38569 }, { "epoch": 11.87, "learning_rate": 7.496213359429414e-06, "loss": 0.0353, "step": 38570 }, { "epoch": 11.87, "learning_rate": 7.4957309633936635e-06, "loss": 0.0254, "step": 38571 }, { "epoch": 11.87, "learning_rate": 7.49524857357534e-06, "loss": 0.0598, "step": 38572 }, { "epoch": 11.87, "learning_rate": 7.4947661899756395e-06, "loss": 0.1021, "step": 38573 }, { "epoch": 11.87, "learning_rate": 7.49428381259576e-06, "loss": 0.04, "step": 38574 }, { "epoch": 11.87, "learning_rate": 7.4938014414368985e-06, "loss": 0.0613, "step": 38575 }, { "epoch": 11.87, "learning_rate": 7.4933190765002476e-06, "loss": 0.0558, "step": 38576 }, { "epoch": 11.87, "learning_rate": 7.4928367177870145e-06, "loss": 0.0352, "step": 38577 }, { "epoch": 11.87, "learning_rate": 7.492354365298391e-06, "loss": 0.045, "step": 38578 }, { "epoch": 11.87, "learning_rate": 7.491872019035574e-06, "loss": 0.0419, "step": 38579 }, { "epoch": 11.87, "learning_rate": 7.491389678999764e-06, "loss": 0.0628, "step": 38580 }, { "epoch": 11.87, "learning_rate": 7.490907345192156e-06, "loss": 0.0843, "step": 38581 }, { "epoch": 11.87, "learning_rate": 7.490425017613949e-06, "loss": 0.0491, "step": 38582 }, { "epoch": 11.87, "learning_rate": 7.48994269626634e-06, "loss": 0.0391, "step": 38583 }, { "epoch": 11.87, "learning_rate": 7.489460381150527e-06, "loss": 0.0183, "step": 38584 }, { "epoch": 11.87, "learning_rate": 7.48897807226771e-06, "loss": 0.0793, "step": 38585 }, { "epoch": 11.87, "learning_rate": 7.488495769619077e-06, "loss": 0.0298, "step": 38586 }, { "epoch": 11.87, "learning_rate": 7.488013473205838e-06, "loss": 0.0453, "step": 38587 }, { "epoch": 11.87, "learning_rate": 7.487531183029182e-06, "loss": 0.0586, "step": 38588 }, { "epoch": 11.87, "learning_rate": 7.487048899090308e-06, "loss": 0.0256, "step": 38589 }, { "epoch": 11.87, "learning_rate": 7.486566621390414e-06, "loss": 0.0282, "step": 38590 }, { "epoch": 11.87, "learning_rate": 7.486084349930698e-06, "loss": 0.0846, "step": 38591 }, { "epoch": 11.87, "learning_rate": 7.485602084712356e-06, "loss": 0.0726, "step": 38592 }, { "epoch": 11.87, "learning_rate": 7.485119825736587e-06, "loss": 0.042, "step": 38593 }, { "epoch": 11.88, "learning_rate": 7.4846375730045885e-06, "loss": 0.0467, "step": 38594 }, { "epoch": 11.88, "learning_rate": 7.4841553265175585e-06, "loss": 0.0736, "step": 38595 }, { "epoch": 11.88, "learning_rate": 7.483673086276686e-06, "loss": 0.0562, "step": 38596 }, { "epoch": 11.88, "learning_rate": 7.483190852283178e-06, "loss": 0.1085, "step": 38597 }, { "epoch": 11.88, "learning_rate": 7.482708624538233e-06, "loss": 0.0516, "step": 38598 }, { "epoch": 11.88, "learning_rate": 7.4822264030430406e-06, "loss": 0.0504, "step": 38599 }, { "epoch": 11.88, "learning_rate": 7.4817441877988005e-06, "loss": 0.037, "step": 38600 }, { "epoch": 11.88, "learning_rate": 7.481261978806709e-06, "loss": 0.0407, "step": 38601 }, { "epoch": 11.88, "learning_rate": 7.4807797760679675e-06, "loss": 0.0632, "step": 38602 }, { "epoch": 11.88, "learning_rate": 7.480297579583771e-06, "loss": 0.0412, "step": 38603 }, { "epoch": 11.88, "learning_rate": 7.479815389355316e-06, "loss": 0.0551, "step": 38604 }, { "epoch": 11.88, "learning_rate": 7.479333205383799e-06, "loss": 0.0426, "step": 38605 }, { "epoch": 11.88, "learning_rate": 7.478851027670417e-06, "loss": 0.0487, "step": 38606 }, { "epoch": 11.88, "learning_rate": 7.4783688562163694e-06, "loss": 0.0548, "step": 38607 }, { "epoch": 11.88, "learning_rate": 7.477886691022856e-06, "loss": 0.0292, "step": 38608 }, { "epoch": 11.88, "learning_rate": 7.477404532091066e-06, "loss": 0.0426, "step": 38609 }, { "epoch": 11.88, "learning_rate": 7.476922379422202e-06, "loss": 0.0433, "step": 38610 }, { "epoch": 11.88, "learning_rate": 7.476440233017457e-06, "loss": 0.0628, "step": 38611 }, { "epoch": 11.88, "learning_rate": 7.475958092878032e-06, "loss": 0.0505, "step": 38612 }, { "epoch": 11.88, "learning_rate": 7.475475959005123e-06, "loss": 0.0729, "step": 38613 }, { "epoch": 11.88, "learning_rate": 7.474993831399926e-06, "loss": 0.06, "step": 38614 }, { "epoch": 11.88, "learning_rate": 7.474511710063639e-06, "loss": 0.0398, "step": 38615 }, { "epoch": 11.88, "learning_rate": 7.474029594997456e-06, "loss": 0.0397, "step": 38616 }, { "epoch": 11.88, "learning_rate": 7.47354748620258e-06, "loss": 0.0501, "step": 38617 }, { "epoch": 11.88, "learning_rate": 7.473065383680205e-06, "loss": 0.0594, "step": 38618 }, { "epoch": 11.88, "learning_rate": 7.472583287431527e-06, "loss": 0.0582, "step": 38619 }, { "epoch": 11.88, "learning_rate": 7.4721011974577425e-06, "loss": 0.0525, "step": 38620 }, { "epoch": 11.88, "learning_rate": 7.471619113760044e-06, "loss": 0.079, "step": 38621 }, { "epoch": 11.88, "learning_rate": 7.471137036339642e-06, "loss": 0.0615, "step": 38622 }, { "epoch": 11.88, "learning_rate": 7.470654965197722e-06, "loss": 0.0414, "step": 38623 }, { "epoch": 11.88, "learning_rate": 7.470172900335484e-06, "loss": 0.0262, "step": 38624 }, { "epoch": 11.88, "learning_rate": 7.469690841754124e-06, "loss": 0.0607, "step": 38625 }, { "epoch": 11.88, "learning_rate": 7.469208789454838e-06, "loss": 0.0522, "step": 38626 }, { "epoch": 11.89, "learning_rate": 7.468726743438825e-06, "loss": 0.0598, "step": 38627 }, { "epoch": 11.89, "learning_rate": 7.4682447037072835e-06, "loss": 0.0414, "step": 38628 }, { "epoch": 11.89, "learning_rate": 7.467762670261408e-06, "loss": 0.0454, "step": 38629 }, { "epoch": 11.89, "learning_rate": 7.467280643102393e-06, "loss": 0.0461, "step": 38630 }, { "epoch": 11.89, "learning_rate": 7.4667986222314345e-06, "loss": 0.0509, "step": 38631 }, { "epoch": 11.89, "learning_rate": 7.4663166076497376e-06, "loss": 0.0487, "step": 38632 }, { "epoch": 11.89, "learning_rate": 7.465834599358493e-06, "loss": 0.0853, "step": 38633 }, { "epoch": 11.89, "learning_rate": 7.465352597358895e-06, "loss": 0.0426, "step": 38634 }, { "epoch": 11.89, "learning_rate": 7.464870601652146e-06, "loss": 0.054, "step": 38635 }, { "epoch": 11.89, "learning_rate": 7.464388612239435e-06, "loss": 0.0669, "step": 38636 }, { "epoch": 11.89, "learning_rate": 7.463906629121966e-06, "loss": 0.0516, "step": 38637 }, { "epoch": 11.89, "learning_rate": 7.463424652300934e-06, "loss": 0.0672, "step": 38638 }, { "epoch": 11.89, "learning_rate": 7.462942681777537e-06, "loss": 0.0296, "step": 38639 }, { "epoch": 11.89, "learning_rate": 7.462460717552968e-06, "loss": 0.0406, "step": 38640 }, { "epoch": 11.89, "learning_rate": 7.461978759628421e-06, "loss": 0.035, "step": 38641 }, { "epoch": 11.89, "learning_rate": 7.461496808005099e-06, "loss": 0.061, "step": 38642 }, { "epoch": 11.89, "learning_rate": 7.4610148626841995e-06, "loss": 0.0227, "step": 38643 }, { "epoch": 11.89, "learning_rate": 7.460532923666912e-06, "loss": 0.0488, "step": 38644 }, { "epoch": 11.89, "learning_rate": 7.460050990954437e-06, "loss": 0.0692, "step": 38645 }, { "epoch": 11.89, "learning_rate": 7.459569064547969e-06, "loss": 0.0828, "step": 38646 }, { "epoch": 11.89, "learning_rate": 7.459087144448709e-06, "loss": 0.0545, "step": 38647 }, { "epoch": 11.89, "learning_rate": 7.45860523065785e-06, "loss": 0.0092, "step": 38648 }, { "epoch": 11.89, "learning_rate": 7.458123323176588e-06, "loss": 0.0494, "step": 38649 }, { "epoch": 11.89, "learning_rate": 7.457641422006123e-06, "loss": 0.0389, "step": 38650 }, { "epoch": 11.89, "learning_rate": 7.4571595271476435e-06, "loss": 0.029, "step": 38651 }, { "epoch": 11.89, "learning_rate": 7.456677638602355e-06, "loss": 0.0522, "step": 38652 }, { "epoch": 11.89, "learning_rate": 7.456195756371454e-06, "loss": 0.0824, "step": 38653 }, { "epoch": 11.89, "learning_rate": 7.455713880456129e-06, "loss": 0.0298, "step": 38654 }, { "epoch": 11.89, "learning_rate": 7.45523201085758e-06, "loss": 0.0296, "step": 38655 }, { "epoch": 11.89, "learning_rate": 7.454750147577004e-06, "loss": 0.0402, "step": 38656 }, { "epoch": 11.89, "learning_rate": 7.454268290615598e-06, "loss": 0.0317, "step": 38657 }, { "epoch": 11.89, "learning_rate": 7.453786439974557e-06, "loss": 0.0433, "step": 38658 }, { "epoch": 11.9, "learning_rate": 7.45330459565508e-06, "loss": 0.06, "step": 38659 }, { "epoch": 11.9, "learning_rate": 7.45282275765836e-06, "loss": 0.0269, "step": 38660 }, { "epoch": 11.9, "learning_rate": 7.4523409259855905e-06, "loss": 0.0247, "step": 38661 }, { "epoch": 11.9, "learning_rate": 7.451859100637976e-06, "loss": 0.0467, "step": 38662 }, { "epoch": 11.9, "learning_rate": 7.451377281616707e-06, "loss": 0.0658, "step": 38663 }, { "epoch": 11.9, "learning_rate": 7.450895468922985e-06, "loss": 0.0802, "step": 38664 }, { "epoch": 11.9, "learning_rate": 7.450413662557998e-06, "loss": 0.0325, "step": 38665 }, { "epoch": 11.9, "learning_rate": 7.449931862522946e-06, "loss": 0.0413, "step": 38666 }, { "epoch": 11.9, "learning_rate": 7.449450068819028e-06, "loss": 0.0667, "step": 38667 }, { "epoch": 11.9, "learning_rate": 7.448968281447437e-06, "loss": 0.024, "step": 38668 }, { "epoch": 11.9, "learning_rate": 7.44848650040937e-06, "loss": 0.0482, "step": 38669 }, { "epoch": 11.9, "learning_rate": 7.448004725706024e-06, "loss": 0.0652, "step": 38670 }, { "epoch": 11.9, "learning_rate": 7.447522957338591e-06, "loss": 0.0714, "step": 38671 }, { "epoch": 11.9, "learning_rate": 7.447041195308273e-06, "loss": 0.0298, "step": 38672 }, { "epoch": 11.9, "learning_rate": 7.446559439616263e-06, "loss": 0.0704, "step": 38673 }, { "epoch": 11.9, "learning_rate": 7.44607769026376e-06, "loss": 0.0386, "step": 38674 }, { "epoch": 11.9, "learning_rate": 7.445595947251956e-06, "loss": 0.0726, "step": 38675 }, { "epoch": 11.9, "learning_rate": 7.445114210582044e-06, "loss": 0.0365, "step": 38676 }, { "epoch": 11.9, "learning_rate": 7.444632480255231e-06, "loss": 0.0513, "step": 38677 }, { "epoch": 11.9, "learning_rate": 7.444150756272704e-06, "loss": 0.0324, "step": 38678 }, { "epoch": 11.9, "learning_rate": 7.443669038635663e-06, "loss": 0.0328, "step": 38679 }, { "epoch": 11.9, "learning_rate": 7.443187327345301e-06, "loss": 0.071, "step": 38680 }, { "epoch": 11.9, "learning_rate": 7.442705622402815e-06, "loss": 0.0488, "step": 38681 }, { "epoch": 11.9, "learning_rate": 7.442223923809402e-06, "loss": 0.0482, "step": 38682 }, { "epoch": 11.9, "learning_rate": 7.441742231566258e-06, "loss": 0.0799, "step": 38683 }, { "epoch": 11.9, "learning_rate": 7.44126054567458e-06, "loss": 0.0423, "step": 38684 }, { "epoch": 11.9, "learning_rate": 7.440778866135563e-06, "loss": 0.0273, "step": 38685 }, { "epoch": 11.9, "learning_rate": 7.440297192950395e-06, "loss": 0.0521, "step": 38686 }, { "epoch": 11.9, "learning_rate": 7.439815526120286e-06, "loss": 0.0154, "step": 38687 }, { "epoch": 11.9, "learning_rate": 7.439333865646423e-06, "loss": 0.0465, "step": 38688 }, { "epoch": 11.9, "learning_rate": 7.438852211530003e-06, "loss": 0.038, "step": 38689 }, { "epoch": 11.9, "learning_rate": 7.438370563772224e-06, "loss": 0.0522, "step": 38690 }, { "epoch": 11.9, "learning_rate": 7.4378889223742766e-06, "loss": 0.0773, "step": 38691 }, { "epoch": 11.91, "learning_rate": 7.437407287337364e-06, "loss": 0.1045, "step": 38692 }, { "epoch": 11.91, "learning_rate": 7.436925658662677e-06, "loss": 0.0654, "step": 38693 }, { "epoch": 11.91, "learning_rate": 7.4364440363514125e-06, "loss": 0.0386, "step": 38694 }, { "epoch": 11.91, "learning_rate": 7.435962420404768e-06, "loss": 0.0438, "step": 38695 }, { "epoch": 11.91, "learning_rate": 7.4354808108239325e-06, "loss": 0.0123, "step": 38696 }, { "epoch": 11.91, "learning_rate": 7.43499920761011e-06, "loss": 0.0613, "step": 38697 }, { "epoch": 11.91, "learning_rate": 7.434517610764495e-06, "loss": 0.0556, "step": 38698 }, { "epoch": 11.91, "learning_rate": 7.434036020288281e-06, "loss": 0.0723, "step": 38699 }, { "epoch": 11.91, "learning_rate": 7.4335544361826616e-06, "loss": 0.0502, "step": 38700 }, { "epoch": 11.91, "learning_rate": 7.433072858448834e-06, "loss": 0.0662, "step": 38701 }, { "epoch": 11.91, "learning_rate": 7.432591287087996e-06, "loss": 0.03, "step": 38702 }, { "epoch": 11.91, "learning_rate": 7.4321097221013414e-06, "loss": 0.0567, "step": 38703 }, { "epoch": 11.91, "learning_rate": 7.431628163490067e-06, "loss": 0.0338, "step": 38704 }, { "epoch": 11.91, "learning_rate": 7.4311466112553665e-06, "loss": 0.0437, "step": 38705 }, { "epoch": 11.91, "learning_rate": 7.430665065398435e-06, "loss": 0.0447, "step": 38706 }, { "epoch": 11.91, "learning_rate": 7.4301835259204714e-06, "loss": 0.0738, "step": 38707 }, { "epoch": 11.91, "learning_rate": 7.429701992822671e-06, "loss": 0.0388, "step": 38708 }, { "epoch": 11.91, "learning_rate": 7.429220466106225e-06, "loss": 0.0703, "step": 38709 }, { "epoch": 11.91, "learning_rate": 7.428738945772332e-06, "loss": 0.0718, "step": 38710 }, { "epoch": 11.91, "learning_rate": 7.428257431822186e-06, "loss": 0.0539, "step": 38711 }, { "epoch": 11.91, "learning_rate": 7.427775924256984e-06, "loss": 0.0274, "step": 38712 }, { "epoch": 11.91, "learning_rate": 7.427294423077922e-06, "loss": 0.0588, "step": 38713 }, { "epoch": 11.91, "learning_rate": 7.426812928286193e-06, "loss": 0.0426, "step": 38714 }, { "epoch": 11.91, "learning_rate": 7.426331439882995e-06, "loss": 0.0747, "step": 38715 }, { "epoch": 11.91, "learning_rate": 7.42584995786952e-06, "loss": 0.0591, "step": 38716 }, { "epoch": 11.91, "learning_rate": 7.4253684822469684e-06, "loss": 0.0462, "step": 38717 }, { "epoch": 11.91, "learning_rate": 7.424887013016531e-06, "loss": 0.0402, "step": 38718 }, { "epoch": 11.91, "learning_rate": 7.424405550179408e-06, "loss": 0.105, "step": 38719 }, { "epoch": 11.91, "learning_rate": 7.423924093736789e-06, "loss": 0.0575, "step": 38720 }, { "epoch": 11.91, "learning_rate": 7.423442643689869e-06, "loss": 0.0385, "step": 38721 }, { "epoch": 11.91, "learning_rate": 7.422961200039852e-06, "loss": 0.04, "step": 38722 }, { "epoch": 11.91, "learning_rate": 7.422479762787925e-06, "loss": 0.0485, "step": 38723 }, { "epoch": 11.92, "learning_rate": 7.421998331935287e-06, "loss": 0.0822, "step": 38724 }, { "epoch": 11.92, "learning_rate": 7.42151690748313e-06, "loss": 0.0365, "step": 38725 }, { "epoch": 11.92, "learning_rate": 7.421035489432652e-06, "loss": 0.0427, "step": 38726 }, { "epoch": 11.92, "learning_rate": 7.420554077785048e-06, "loss": 0.0911, "step": 38727 }, { "epoch": 11.92, "learning_rate": 7.420072672541514e-06, "loss": 0.0507, "step": 38728 }, { "epoch": 11.92, "learning_rate": 7.419591273703245e-06, "loss": 0.0224, "step": 38729 }, { "epoch": 11.92, "learning_rate": 7.419109881271434e-06, "loss": 0.0531, "step": 38730 }, { "epoch": 11.92, "learning_rate": 7.418628495247271e-06, "loss": 0.0348, "step": 38731 }, { "epoch": 11.92, "learning_rate": 7.418147115631966e-06, "loss": 0.0393, "step": 38732 }, { "epoch": 11.92, "learning_rate": 7.417665742426703e-06, "loss": 0.0564, "step": 38733 }, { "epoch": 11.92, "learning_rate": 7.41718437563268e-06, "loss": 0.073, "step": 38734 }, { "epoch": 11.92, "learning_rate": 7.416703015251092e-06, "loss": 0.0654, "step": 38735 }, { "epoch": 11.92, "learning_rate": 7.416221661283132e-06, "loss": 0.0534, "step": 38736 }, { "epoch": 11.92, "learning_rate": 7.4157403137299976e-06, "loss": 0.033, "step": 38737 }, { "epoch": 11.92, "learning_rate": 7.415258972592885e-06, "loss": 0.0387, "step": 38738 }, { "epoch": 11.92, "learning_rate": 7.4147776378729875e-06, "loss": 0.0636, "step": 38739 }, { "epoch": 11.92, "learning_rate": 7.4142963095715e-06, "loss": 0.0766, "step": 38740 }, { "epoch": 11.92, "learning_rate": 7.413814987689614e-06, "loss": 0.0611, "step": 38741 }, { "epoch": 11.92, "learning_rate": 7.413333672228531e-06, "loss": 0.0628, "step": 38742 }, { "epoch": 11.92, "learning_rate": 7.4128523631894464e-06, "loss": 0.0417, "step": 38743 }, { "epoch": 11.92, "learning_rate": 7.412371060573549e-06, "loss": 0.1028, "step": 38744 }, { "epoch": 11.92, "learning_rate": 7.411889764382036e-06, "loss": 0.0647, "step": 38745 }, { "epoch": 11.92, "learning_rate": 7.411408474616102e-06, "loss": 0.0793, "step": 38746 }, { "epoch": 11.92, "learning_rate": 7.410927191276946e-06, "loss": 0.0345, "step": 38747 }, { "epoch": 11.92, "learning_rate": 7.410445914365758e-06, "loss": 0.0452, "step": 38748 }, { "epoch": 11.92, "learning_rate": 7.409964643883735e-06, "loss": 0.0552, "step": 38749 }, { "epoch": 11.92, "learning_rate": 7.409483379832074e-06, "loss": 0.0643, "step": 38750 }, { "epoch": 11.92, "learning_rate": 7.409002122211961e-06, "loss": 0.0455, "step": 38751 }, { "epoch": 11.92, "learning_rate": 7.408520871024601e-06, "loss": 0.1032, "step": 38752 }, { "epoch": 11.92, "learning_rate": 7.408039626271187e-06, "loss": 0.025, "step": 38753 }, { "epoch": 11.92, "learning_rate": 7.407558387952911e-06, "loss": 0.0468, "step": 38754 }, { "epoch": 11.92, "learning_rate": 7.407077156070967e-06, "loss": 0.0327, "step": 38755 }, { "epoch": 11.92, "learning_rate": 7.40659593062655e-06, "loss": 0.0714, "step": 38756 }, { "epoch": 11.93, "learning_rate": 7.406114711620859e-06, "loss": 0.0683, "step": 38757 }, { "epoch": 11.93, "learning_rate": 7.405633499055085e-06, "loss": 0.0595, "step": 38758 }, { "epoch": 11.93, "learning_rate": 7.405152292930424e-06, "loss": 0.0376, "step": 38759 }, { "epoch": 11.93, "learning_rate": 7.40467109324807e-06, "loss": 0.0566, "step": 38760 }, { "epoch": 11.93, "learning_rate": 7.404189900009215e-06, "loss": 0.0502, "step": 38761 }, { "epoch": 11.93, "learning_rate": 7.4037087132150605e-06, "loss": 0.0495, "step": 38762 }, { "epoch": 11.93, "learning_rate": 7.403227532866797e-06, "loss": 0.0444, "step": 38763 }, { "epoch": 11.93, "learning_rate": 7.40274635896562e-06, "loss": 0.0378, "step": 38764 }, { "epoch": 11.93, "learning_rate": 7.402265191512723e-06, "loss": 0.0693, "step": 38765 }, { "epoch": 11.93, "learning_rate": 7.401784030509299e-06, "loss": 0.0524, "step": 38766 }, { "epoch": 11.93, "learning_rate": 7.401302875956547e-06, "loss": 0.0502, "step": 38767 }, { "epoch": 11.93, "learning_rate": 7.400821727855658e-06, "loss": 0.0418, "step": 38768 }, { "epoch": 11.93, "learning_rate": 7.40034058620783e-06, "loss": 0.0378, "step": 38769 }, { "epoch": 11.93, "learning_rate": 7.399859451014254e-06, "loss": 0.0391, "step": 38770 }, { "epoch": 11.93, "learning_rate": 7.399378322276125e-06, "loss": 0.0748, "step": 38771 }, { "epoch": 11.93, "learning_rate": 7.398897199994639e-06, "loss": 0.0366, "step": 38772 }, { "epoch": 11.93, "learning_rate": 7.398416084170992e-06, "loss": 0.048, "step": 38773 }, { "epoch": 11.93, "learning_rate": 7.397934974806378e-06, "loss": 0.0504, "step": 38774 }, { "epoch": 11.93, "learning_rate": 7.3974538719019876e-06, "loss": 0.0217, "step": 38775 }, { "epoch": 11.93, "learning_rate": 7.396972775459013e-06, "loss": 0.0506, "step": 38776 }, { "epoch": 11.93, "learning_rate": 7.39649168547866e-06, "loss": 0.0681, "step": 38777 }, { "epoch": 11.93, "learning_rate": 7.396010601962114e-06, "loss": 0.0592, "step": 38778 }, { "epoch": 11.93, "learning_rate": 7.3955295249105694e-06, "loss": 0.0799, "step": 38779 }, { "epoch": 11.93, "learning_rate": 7.395048454325227e-06, "loss": 0.0565, "step": 38780 }, { "epoch": 11.93, "learning_rate": 7.394567390207272e-06, "loss": 0.0486, "step": 38781 }, { "epoch": 11.93, "learning_rate": 7.394086332557907e-06, "loss": 0.0499, "step": 38782 }, { "epoch": 11.93, "learning_rate": 7.393605281378322e-06, "loss": 0.0491, "step": 38783 }, { "epoch": 11.93, "learning_rate": 7.393124236669714e-06, "loss": 0.0506, "step": 38784 }, { "epoch": 11.93, "learning_rate": 7.392643198433278e-06, "loss": 0.0513, "step": 38785 }, { "epoch": 11.93, "learning_rate": 7.392162166670198e-06, "loss": 0.0668, "step": 38786 }, { "epoch": 11.93, "learning_rate": 7.391681141381683e-06, "loss": 0.0411, "step": 38787 }, { "epoch": 11.93, "learning_rate": 7.39120012256892e-06, "loss": 0.0589, "step": 38788 }, { "epoch": 11.94, "learning_rate": 7.390719110233102e-06, "loss": 0.0461, "step": 38789 }, { "epoch": 11.94, "learning_rate": 7.390238104375427e-06, "loss": 0.0588, "step": 38790 }, { "epoch": 11.94, "learning_rate": 7.389757104997084e-06, "loss": 0.047, "step": 38791 }, { "epoch": 11.94, "learning_rate": 7.389276112099272e-06, "loss": 0.0503, "step": 38792 }, { "epoch": 11.94, "learning_rate": 7.388795125683183e-06, "loss": 0.0422, "step": 38793 }, { "epoch": 11.94, "learning_rate": 7.388314145750013e-06, "loss": 0.0644, "step": 38794 }, { "epoch": 11.94, "learning_rate": 7.387833172300958e-06, "loss": 0.0409, "step": 38795 }, { "epoch": 11.94, "learning_rate": 7.3873522053372e-06, "loss": 0.0477, "step": 38796 }, { "epoch": 11.94, "learning_rate": 7.3868712448599475e-06, "loss": 0.0498, "step": 38797 }, { "epoch": 11.94, "learning_rate": 7.386390290870392e-06, "loss": 0.0666, "step": 38798 }, { "epoch": 11.94, "learning_rate": 7.385909343369722e-06, "loss": 0.0499, "step": 38799 }, { "epoch": 11.94, "learning_rate": 7.385428402359135e-06, "loss": 0.0397, "step": 38800 }, { "epoch": 11.94, "learning_rate": 7.384947467839823e-06, "loss": 0.0304, "step": 38801 }, { "epoch": 11.94, "learning_rate": 7.384466539812981e-06, "loss": 0.0273, "step": 38802 }, { "epoch": 11.94, "learning_rate": 7.3839856182798056e-06, "loss": 0.0703, "step": 38803 }, { "epoch": 11.94, "learning_rate": 7.383504703241488e-06, "loss": 0.0723, "step": 38804 }, { "epoch": 11.94, "learning_rate": 7.383023794699222e-06, "loss": 0.073, "step": 38805 }, { "epoch": 11.94, "learning_rate": 7.382542892654201e-06, "loss": 0.0253, "step": 38806 }, { "epoch": 11.94, "learning_rate": 7.382061997107623e-06, "loss": 0.0567, "step": 38807 }, { "epoch": 11.94, "learning_rate": 7.38158110806068e-06, "loss": 0.0488, "step": 38808 }, { "epoch": 11.94, "learning_rate": 7.3811002255145655e-06, "loss": 0.0416, "step": 38809 }, { "epoch": 11.94, "learning_rate": 7.380619349470472e-06, "loss": 0.0352, "step": 38810 }, { "epoch": 11.94, "learning_rate": 7.380138479929591e-06, "loss": 0.0567, "step": 38811 }, { "epoch": 11.94, "learning_rate": 7.379657616893124e-06, "loss": 0.0304, "step": 38812 }, { "epoch": 11.94, "learning_rate": 7.379176760362259e-06, "loss": 0.0534, "step": 38813 }, { "epoch": 11.94, "learning_rate": 7.378695910338191e-06, "loss": 0.0462, "step": 38814 }, { "epoch": 11.94, "learning_rate": 7.378215066822114e-06, "loss": 0.0416, "step": 38815 }, { "epoch": 11.94, "learning_rate": 7.377734229815225e-06, "loss": 0.0689, "step": 38816 }, { "epoch": 11.94, "learning_rate": 7.377253399318714e-06, "loss": 0.0724, "step": 38817 }, { "epoch": 11.94, "learning_rate": 7.376772575333775e-06, "loss": 0.0721, "step": 38818 }, { "epoch": 11.94, "learning_rate": 7.376291757861604e-06, "loss": 0.107, "step": 38819 }, { "epoch": 11.94, "learning_rate": 7.37581094690339e-06, "loss": 0.0327, "step": 38820 }, { "epoch": 11.94, "learning_rate": 7.375330142460331e-06, "loss": 0.0698, "step": 38821 }, { "epoch": 11.95, "learning_rate": 7.3748493445336235e-06, "loss": 0.0473, "step": 38822 }, { "epoch": 11.95, "learning_rate": 7.374368553124454e-06, "loss": 0.0513, "step": 38823 }, { "epoch": 11.95, "learning_rate": 7.37388776823402e-06, "loss": 0.0711, "step": 38824 }, { "epoch": 11.95, "learning_rate": 7.373406989863514e-06, "loss": 0.0512, "step": 38825 }, { "epoch": 11.95, "learning_rate": 7.372926218014131e-06, "loss": 0.0427, "step": 38826 }, { "epoch": 11.95, "learning_rate": 7.372445452687065e-06, "loss": 0.057, "step": 38827 }, { "epoch": 11.95, "learning_rate": 7.371964693883507e-06, "loss": 0.0388, "step": 38828 }, { "epoch": 11.95, "learning_rate": 7.371483941604656e-06, "loss": 0.0421, "step": 38829 }, { "epoch": 11.95, "learning_rate": 7.371003195851693e-06, "loss": 0.1475, "step": 38830 }, { "epoch": 11.95, "learning_rate": 7.370522456625827e-06, "loss": 0.0485, "step": 38831 }, { "epoch": 11.95, "learning_rate": 7.370041723928246e-06, "loss": 0.0541, "step": 38832 }, { "epoch": 11.95, "learning_rate": 7.369560997760141e-06, "loss": 0.0646, "step": 38833 }, { "epoch": 11.95, "learning_rate": 7.3690802781227056e-06, "loss": 0.0606, "step": 38834 }, { "epoch": 11.95, "learning_rate": 7.368599565017134e-06, "loss": 0.1005, "step": 38835 }, { "epoch": 11.95, "learning_rate": 7.3681188584446205e-06, "loss": 0.0589, "step": 38836 }, { "epoch": 11.95, "learning_rate": 7.367638158406359e-06, "loss": 0.0295, "step": 38837 }, { "epoch": 11.95, "learning_rate": 7.367157464903543e-06, "loss": 0.0957, "step": 38838 }, { "epoch": 11.95, "learning_rate": 7.366676777937365e-06, "loss": 0.0401, "step": 38839 }, { "epoch": 11.95, "learning_rate": 7.366196097509016e-06, "loss": 0.054, "step": 38840 }, { "epoch": 11.95, "learning_rate": 7.365715423619695e-06, "loss": 0.0242, "step": 38841 }, { "epoch": 11.95, "learning_rate": 7.365234756270592e-06, "loss": 0.0719, "step": 38842 }, { "epoch": 11.95, "learning_rate": 7.364754095462903e-06, "loss": 0.0445, "step": 38843 }, { "epoch": 11.95, "learning_rate": 7.3642734411978166e-06, "loss": 0.0582, "step": 38844 }, { "epoch": 11.95, "learning_rate": 7.363792793476527e-06, "loss": 0.074, "step": 38845 }, { "epoch": 11.95, "learning_rate": 7.36331215230023e-06, "loss": 0.0651, "step": 38846 }, { "epoch": 11.95, "learning_rate": 7.362831517670121e-06, "loss": 0.0545, "step": 38847 }, { "epoch": 11.95, "learning_rate": 7.362350889587389e-06, "loss": 0.0621, "step": 38848 }, { "epoch": 11.95, "learning_rate": 7.3618702680532275e-06, "loss": 0.0805, "step": 38849 }, { "epoch": 11.95, "learning_rate": 7.361389653068829e-06, "loss": 0.0326, "step": 38850 }, { "epoch": 11.95, "learning_rate": 7.360909044635391e-06, "loss": 0.0689, "step": 38851 }, { "epoch": 11.95, "learning_rate": 7.360428442754105e-06, "loss": 0.0304, "step": 38852 }, { "epoch": 11.95, "learning_rate": 7.359947847426165e-06, "loss": 0.0453, "step": 38853 }, { "epoch": 11.96, "learning_rate": 7.359467258652762e-06, "loss": 0.0766, "step": 38854 }, { "epoch": 11.96, "learning_rate": 7.358986676435085e-06, "loss": 0.0417, "step": 38855 }, { "epoch": 11.96, "learning_rate": 7.358506100774338e-06, "loss": 0.052, "step": 38856 }, { "epoch": 11.96, "learning_rate": 7.3580255316717065e-06, "loss": 0.0638, "step": 38857 }, { "epoch": 11.96, "learning_rate": 7.357544969128385e-06, "loss": 0.0672, "step": 38858 }, { "epoch": 11.96, "learning_rate": 7.357064413145567e-06, "loss": 0.0439, "step": 38859 }, { "epoch": 11.96, "learning_rate": 7.356583863724442e-06, "loss": 0.0237, "step": 38860 }, { "epoch": 11.96, "learning_rate": 7.356103320866211e-06, "loss": 0.0838, "step": 38861 }, { "epoch": 11.96, "learning_rate": 7.355622784572063e-06, "loss": 0.0755, "step": 38862 }, { "epoch": 11.96, "learning_rate": 7.3551422548431885e-06, "loss": 0.0441, "step": 38863 }, { "epoch": 11.96, "learning_rate": 7.354661731680785e-06, "loss": 0.0231, "step": 38864 }, { "epoch": 11.96, "learning_rate": 7.354181215086039e-06, "loss": 0.039, "step": 38865 }, { "epoch": 11.96, "learning_rate": 7.353700705060154e-06, "loss": 0.0672, "step": 38866 }, { "epoch": 11.96, "learning_rate": 7.353220201604313e-06, "loss": 0.0437, "step": 38867 }, { "epoch": 11.96, "learning_rate": 7.352739704719714e-06, "loss": 0.0601, "step": 38868 }, { "epoch": 11.96, "learning_rate": 7.352259214407547e-06, "loss": 0.0545, "step": 38869 }, { "epoch": 11.96, "learning_rate": 7.3517787306690055e-06, "loss": 0.0462, "step": 38870 }, { "epoch": 11.96, "learning_rate": 7.351298253505285e-06, "loss": 0.0455, "step": 38871 }, { "epoch": 11.96, "learning_rate": 7.350817782917578e-06, "loss": 0.0565, "step": 38872 }, { "epoch": 11.96, "learning_rate": 7.350337318907075e-06, "loss": 0.0419, "step": 38873 }, { "epoch": 11.96, "learning_rate": 7.3498568614749735e-06, "loss": 0.0735, "step": 38874 }, { "epoch": 11.96, "learning_rate": 7.349376410622455e-06, "loss": 0.0567, "step": 38875 }, { "epoch": 11.96, "learning_rate": 7.348895966350727e-06, "loss": 0.0369, "step": 38876 }, { "epoch": 11.96, "learning_rate": 7.348415528660975e-06, "loss": 0.0504, "step": 38877 }, { "epoch": 11.96, "learning_rate": 7.3479350975543925e-06, "loss": 0.0703, "step": 38878 }, { "epoch": 11.96, "learning_rate": 7.347454673032171e-06, "loss": 0.0539, "step": 38879 }, { "epoch": 11.96, "learning_rate": 7.3469742550955026e-06, "loss": 0.091, "step": 38880 }, { "epoch": 11.96, "learning_rate": 7.3464938437455835e-06, "loss": 0.0493, "step": 38881 }, { "epoch": 11.96, "learning_rate": 7.3460134389836055e-06, "loss": 0.0401, "step": 38882 }, { "epoch": 11.96, "learning_rate": 7.345533040810759e-06, "loss": 0.0533, "step": 38883 }, { "epoch": 11.96, "learning_rate": 7.345052649228241e-06, "loss": 0.0532, "step": 38884 }, { "epoch": 11.96, "learning_rate": 7.344572264237237e-06, "loss": 0.0259, "step": 38885 }, { "epoch": 11.96, "learning_rate": 7.344091885838949e-06, "loss": 0.0501, "step": 38886 }, { "epoch": 11.97, "learning_rate": 7.343611514034564e-06, "loss": 0.0611, "step": 38887 }, { "epoch": 11.97, "learning_rate": 7.343131148825275e-06, "loss": 0.0532, "step": 38888 }, { "epoch": 11.97, "learning_rate": 7.342650790212275e-06, "loss": 0.0449, "step": 38889 }, { "epoch": 11.97, "learning_rate": 7.342170438196754e-06, "loss": 0.0615, "step": 38890 }, { "epoch": 11.97, "learning_rate": 7.341690092779909e-06, "loss": 0.0516, "step": 38891 }, { "epoch": 11.97, "learning_rate": 7.341209753962933e-06, "loss": 0.0399, "step": 38892 }, { "epoch": 11.97, "learning_rate": 7.340729421747015e-06, "loss": 0.0709, "step": 38893 }, { "epoch": 11.97, "learning_rate": 7.340249096133348e-06, "loss": 0.0316, "step": 38894 }, { "epoch": 11.97, "learning_rate": 7.339768777123124e-06, "loss": 0.1015, "step": 38895 }, { "epoch": 11.97, "learning_rate": 7.33928846471754e-06, "loss": 0.0658, "step": 38896 }, { "epoch": 11.97, "learning_rate": 7.338808158917785e-06, "loss": 0.0999, "step": 38897 }, { "epoch": 11.97, "learning_rate": 7.338327859725054e-06, "loss": 0.0336, "step": 38898 }, { "epoch": 11.97, "learning_rate": 7.337847567140534e-06, "loss": 0.0725, "step": 38899 }, { "epoch": 11.97, "learning_rate": 7.3373672811654175e-06, "loss": 0.0548, "step": 38900 }, { "epoch": 11.97, "learning_rate": 7.3368870018009075e-06, "loss": 0.0434, "step": 38901 }, { "epoch": 11.97, "learning_rate": 7.336406729048187e-06, "loss": 0.043, "step": 38902 }, { "epoch": 11.97, "learning_rate": 7.33592646290845e-06, "loss": 0.0499, "step": 38903 }, { "epoch": 11.97, "learning_rate": 7.335446203382888e-06, "loss": 0.0492, "step": 38904 }, { "epoch": 11.97, "learning_rate": 7.334965950472693e-06, "loss": 0.0212, "step": 38905 }, { "epoch": 11.97, "learning_rate": 7.3344857041790616e-06, "loss": 0.0508, "step": 38906 }, { "epoch": 11.97, "learning_rate": 7.334005464503184e-06, "loss": 0.0655, "step": 38907 }, { "epoch": 11.97, "learning_rate": 7.333525231446255e-06, "loss": 0.0417, "step": 38908 }, { "epoch": 11.97, "learning_rate": 7.33304500500946e-06, "loss": 0.0548, "step": 38909 }, { "epoch": 11.97, "learning_rate": 7.332564785193992e-06, "loss": 0.0378, "step": 38910 }, { "epoch": 11.97, "learning_rate": 7.332084572001053e-06, "loss": 0.0508, "step": 38911 }, { "epoch": 11.97, "learning_rate": 7.331604365431826e-06, "loss": 0.0666, "step": 38912 }, { "epoch": 11.97, "learning_rate": 7.331124165487507e-06, "loss": 0.0511, "step": 38913 }, { "epoch": 11.97, "learning_rate": 7.330643972169287e-06, "loss": 0.0498, "step": 38914 }, { "epoch": 11.97, "learning_rate": 7.330163785478356e-06, "loss": 0.0997, "step": 38915 }, { "epoch": 11.97, "learning_rate": 7.32968360541591e-06, "loss": 0.09, "step": 38916 }, { "epoch": 11.97, "learning_rate": 7.3292034319831405e-06, "loss": 0.0396, "step": 38917 }, { "epoch": 11.97, "learning_rate": 7.328723265181239e-06, "loss": 0.069, "step": 38918 }, { "epoch": 11.98, "learning_rate": 7.328243105011399e-06, "loss": 0.0458, "step": 38919 }, { "epoch": 11.98, "learning_rate": 7.327762951474805e-06, "loss": 0.0662, "step": 38920 }, { "epoch": 11.98, "learning_rate": 7.3272828045726595e-06, "loss": 0.0402, "step": 38921 }, { "epoch": 11.98, "learning_rate": 7.326802664306152e-06, "loss": 0.0325, "step": 38922 }, { "epoch": 11.98, "learning_rate": 7.326322530676471e-06, "loss": 0.0501, "step": 38923 }, { "epoch": 11.98, "learning_rate": 7.325842403684811e-06, "loss": 0.0456, "step": 38924 }, { "epoch": 11.98, "learning_rate": 7.32536228333236e-06, "loss": 0.0687, "step": 38925 }, { "epoch": 11.98, "learning_rate": 7.324882169620315e-06, "loss": 0.0609, "step": 38926 }, { "epoch": 11.98, "learning_rate": 7.324402062549869e-06, "loss": 0.0376, "step": 38927 }, { "epoch": 11.98, "learning_rate": 7.323921962122209e-06, "loss": 0.0373, "step": 38928 }, { "epoch": 11.98, "learning_rate": 7.323441868338532e-06, "loss": 0.0295, "step": 38929 }, { "epoch": 11.98, "learning_rate": 7.322961781200022e-06, "loss": 0.0548, "step": 38930 }, { "epoch": 11.98, "learning_rate": 7.32248170070788e-06, "loss": 0.0566, "step": 38931 }, { "epoch": 11.98, "learning_rate": 7.322001626863295e-06, "loss": 0.0233, "step": 38932 }, { "epoch": 11.98, "learning_rate": 7.321521559667457e-06, "loss": 0.0442, "step": 38933 }, { "epoch": 11.98, "learning_rate": 7.321041499121558e-06, "loss": 0.0187, "step": 38934 }, { "epoch": 11.98, "learning_rate": 7.32056144522679e-06, "loss": 0.063, "step": 38935 }, { "epoch": 11.98, "learning_rate": 7.320081397984347e-06, "loss": 0.0389, "step": 38936 }, { "epoch": 11.98, "learning_rate": 7.319601357395419e-06, "loss": 0.0411, "step": 38937 }, { "epoch": 11.98, "learning_rate": 7.319121323461198e-06, "loss": 0.0224, "step": 38938 }, { "epoch": 11.98, "learning_rate": 7.3186412961828765e-06, "loss": 0.0287, "step": 38939 }, { "epoch": 11.98, "learning_rate": 7.318161275561644e-06, "loss": 0.0495, "step": 38940 }, { "epoch": 11.98, "learning_rate": 7.3176812615986955e-06, "loss": 0.0262, "step": 38941 }, { "epoch": 11.98, "learning_rate": 7.317201254295222e-06, "loss": 0.0369, "step": 38942 }, { "epoch": 11.98, "learning_rate": 7.316721253652417e-06, "loss": 0.031, "step": 38943 }, { "epoch": 11.98, "learning_rate": 7.316241259671466e-06, "loss": 0.0528, "step": 38944 }, { "epoch": 11.98, "learning_rate": 7.315761272353563e-06, "loss": 0.0496, "step": 38945 }, { "epoch": 11.98, "learning_rate": 7.315281291699904e-06, "loss": 0.0614, "step": 38946 }, { "epoch": 11.98, "learning_rate": 7.314801317711677e-06, "loss": 0.0264, "step": 38947 }, { "epoch": 11.98, "learning_rate": 7.314321350390075e-06, "loss": 0.0402, "step": 38948 }, { "epoch": 11.98, "learning_rate": 7.31384138973629e-06, "loss": 0.0694, "step": 38949 }, { "epoch": 11.98, "learning_rate": 7.313361435751507e-06, "loss": 0.0574, "step": 38950 }, { "epoch": 11.98, "learning_rate": 7.312881488436928e-06, "loss": 0.0861, "step": 38951 }, { "epoch": 11.99, "learning_rate": 7.31240154779374e-06, "loss": 0.0433, "step": 38952 }, { "epoch": 11.99, "learning_rate": 7.311921613823136e-06, "loss": 0.0634, "step": 38953 }, { "epoch": 11.99, "learning_rate": 7.311441686526304e-06, "loss": 0.0578, "step": 38954 }, { "epoch": 11.99, "learning_rate": 7.310961765904432e-06, "loss": 0.0589, "step": 38955 }, { "epoch": 11.99, "learning_rate": 7.310481851958725e-06, "loss": 0.0389, "step": 38956 }, { "epoch": 11.99, "learning_rate": 7.310001944690364e-06, "loss": 0.0404, "step": 38957 }, { "epoch": 11.99, "learning_rate": 7.309522044100544e-06, "loss": 0.0318, "step": 38958 }, { "epoch": 11.99, "learning_rate": 7.309042150190453e-06, "loss": 0.0454, "step": 38959 }, { "epoch": 11.99, "learning_rate": 7.308562262961283e-06, "loss": 0.0536, "step": 38960 }, { "epoch": 11.99, "learning_rate": 7.308082382414231e-06, "loss": 0.0541, "step": 38961 }, { "epoch": 11.99, "learning_rate": 7.3076025085504845e-06, "loss": 0.0491, "step": 38962 }, { "epoch": 11.99, "learning_rate": 7.3071226413712335e-06, "loss": 0.0511, "step": 38963 }, { "epoch": 11.99, "learning_rate": 7.3066427808776754e-06, "loss": 0.046, "step": 38964 }, { "epoch": 11.99, "learning_rate": 7.30616292707099e-06, "loss": 0.0407, "step": 38965 }, { "epoch": 11.99, "learning_rate": 7.305683079952381e-06, "loss": 0.0662, "step": 38966 }, { "epoch": 11.99, "learning_rate": 7.305203239523033e-06, "loss": 0.0454, "step": 38967 }, { "epoch": 11.99, "learning_rate": 7.30472340578414e-06, "loss": 0.0264, "step": 38968 }, { "epoch": 11.99, "learning_rate": 7.304243578736891e-06, "loss": 0.06, "step": 38969 }, { "epoch": 11.99, "learning_rate": 7.303763758382477e-06, "loss": 0.0476, "step": 38970 }, { "epoch": 11.99, "learning_rate": 7.3032839447220924e-06, "loss": 0.0515, "step": 38971 }, { "epoch": 11.99, "learning_rate": 7.302804137756927e-06, "loss": 0.0563, "step": 38972 }, { "epoch": 11.99, "learning_rate": 7.302324337488171e-06, "loss": 0.054, "step": 38973 }, { "epoch": 11.99, "learning_rate": 7.30184454391702e-06, "loss": 0.0435, "step": 38974 }, { "epoch": 11.99, "learning_rate": 7.301364757044655e-06, "loss": 0.0401, "step": 38975 }, { "epoch": 11.99, "learning_rate": 7.300884976872277e-06, "loss": 0.0402, "step": 38976 }, { "epoch": 11.99, "learning_rate": 7.300405203401076e-06, "loss": 0.0306, "step": 38977 }, { "epoch": 11.99, "learning_rate": 7.29992543663224e-06, "loss": 0.035, "step": 38978 }, { "epoch": 11.99, "learning_rate": 7.299445676566961e-06, "loss": 0.0406, "step": 38979 }, { "epoch": 11.99, "learning_rate": 7.298965923206427e-06, "loss": 0.0286, "step": 38980 }, { "epoch": 11.99, "learning_rate": 7.2984861765518355e-06, "loss": 0.0681, "step": 38981 }, { "epoch": 11.99, "learning_rate": 7.298006436604374e-06, "loss": 0.0732, "step": 38982 }, { "epoch": 11.99, "learning_rate": 7.297526703365236e-06, "loss": 0.0634, "step": 38983 }, { "epoch": 12.0, "learning_rate": 7.2970469768356115e-06, "loss": 0.0734, "step": 38984 }, { "epoch": 12.0, "learning_rate": 7.296567257016684e-06, "loss": 0.0412, "step": 38985 }, { "epoch": 12.0, "learning_rate": 7.2960875439096555e-06, "loss": 0.0314, "step": 38986 }, { "epoch": 12.0, "learning_rate": 7.295607837515715e-06, "loss": 0.076, "step": 38987 }, { "epoch": 12.0, "learning_rate": 7.295128137836048e-06, "loss": 0.0605, "step": 38988 }, { "epoch": 12.0, "learning_rate": 7.2946484448718506e-06, "loss": 0.0537, "step": 38989 }, { "epoch": 12.0, "learning_rate": 7.294168758624307e-06, "loss": 0.0436, "step": 38990 }, { "epoch": 12.0, "learning_rate": 7.293689079094617e-06, "loss": 0.0489, "step": 38991 }, { "epoch": 12.0, "learning_rate": 7.293209406283967e-06, "loss": 0.0759, "step": 38992 }, { "epoch": 12.0, "learning_rate": 7.292729740193549e-06, "loss": 0.0416, "step": 38993 }, { "epoch": 12.0, "learning_rate": 7.2922500808245525e-06, "loss": 0.0777, "step": 38994 }, { "epoch": 12.0, "learning_rate": 7.291770428178167e-06, "loss": 0.0586, "step": 38995 }, { "epoch": 12.0, "learning_rate": 7.291290782255587e-06, "loss": 0.0472, "step": 38996 }, { "epoch": 12.0, "learning_rate": 7.290811143058003e-06, "loss": 0.0375, "step": 38997 }, { "epoch": 12.0, "learning_rate": 7.2903315105866055e-06, "loss": 0.071, "step": 38998 }, { "epoch": 12.0, "learning_rate": 7.2898518848425824e-06, "loss": 0.0614, "step": 38999 }, { "epoch": 12.0, "learning_rate": 7.289372265827122e-06, "loss": 0.0553, "step": 39000 }, { "epoch": 12.0, "learning_rate": 7.288892653541426e-06, "loss": 0.0582, "step": 39001 }, { "epoch": 12.0, "learning_rate": 7.288413047986677e-06, "loss": 0.0242, "step": 39002 }, { "epoch": 12.0, "learning_rate": 7.287933449164068e-06, "loss": 0.0327, "step": 39003 }, { "epoch": 12.0, "learning_rate": 7.287453857074789e-06, "loss": 0.0441, "step": 39004 }, { "epoch": 12.0, "learning_rate": 7.2869742717200265e-06, "loss": 0.0299, "step": 39005 }, { "epoch": 12.0, "learning_rate": 7.28649469310098e-06, "loss": 0.0203, "step": 39006 }, { "epoch": 12.0, "learning_rate": 7.286015121218835e-06, "loss": 0.024, "step": 39007 }, { "epoch": 12.0, "learning_rate": 7.2855355560747835e-06, "loss": 0.032, "step": 39008 }, { "epoch": 12.0, "learning_rate": 7.285055997670014e-06, "loss": 0.0437, "step": 39009 }, { "epoch": 12.0, "learning_rate": 7.284576446005716e-06, "loss": 0.0358, "step": 39010 }, { "epoch": 12.0, "learning_rate": 7.284096901083086e-06, "loss": 0.0298, "step": 39011 }, { "epoch": 12.0, "learning_rate": 7.2836173629033125e-06, "loss": 0.0244, "step": 39012 }, { "epoch": 12.0, "learning_rate": 7.283137831467583e-06, "loss": 0.0195, "step": 39013 }, { "epoch": 12.0, "learning_rate": 7.282658306777089e-06, "loss": 0.032, "step": 39014 }, { "epoch": 12.0, "learning_rate": 7.2821787888330205e-06, "loss": 0.0128, "step": 39015 }, { "epoch": 12.0, "learning_rate": 7.2816992776365714e-06, "loss": 0.055, "step": 39016 }, { "epoch": 12.01, "learning_rate": 7.281219773188931e-06, "loss": 0.0435, "step": 39017 }, { "epoch": 12.01, "learning_rate": 7.2807402754912885e-06, "loss": 0.0216, "step": 39018 }, { "epoch": 12.01, "learning_rate": 7.280260784544837e-06, "loss": 0.0202, "step": 39019 }, { "epoch": 12.01, "learning_rate": 7.279781300350758e-06, "loss": 0.0554, "step": 39020 }, { "epoch": 12.01, "learning_rate": 7.279301822910253e-06, "loss": 0.0463, "step": 39021 }, { "epoch": 12.01, "learning_rate": 7.278822352224511e-06, "loss": 0.0258, "step": 39022 }, { "epoch": 12.01, "learning_rate": 7.278342888294716e-06, "loss": 0.0337, "step": 39023 }, { "epoch": 12.01, "learning_rate": 7.277863431122062e-06, "loss": 0.028, "step": 39024 }, { "epoch": 12.01, "learning_rate": 7.2773839807077374e-06, "loss": 0.0271, "step": 39025 }, { "epoch": 12.01, "learning_rate": 7.276904537052938e-06, "loss": 0.0275, "step": 39026 }, { "epoch": 12.01, "learning_rate": 7.276425100158848e-06, "loss": 0.0239, "step": 39027 }, { "epoch": 12.01, "learning_rate": 7.275945670026662e-06, "loss": 0.0457, "step": 39028 }, { "epoch": 12.01, "learning_rate": 7.27546624665757e-06, "loss": 0.0274, "step": 39029 }, { "epoch": 12.01, "learning_rate": 7.274986830052755e-06, "loss": 0.0476, "step": 39030 }, { "epoch": 12.01, "learning_rate": 7.2745074202134165e-06, "loss": 0.0269, "step": 39031 }, { "epoch": 12.01, "learning_rate": 7.274028017140744e-06, "loss": 0.0224, "step": 39032 }, { "epoch": 12.01, "learning_rate": 7.273548620835922e-06, "loss": 0.0174, "step": 39033 }, { "epoch": 12.01, "learning_rate": 7.273069231300145e-06, "loss": 0.0206, "step": 39034 }, { "epoch": 12.01, "learning_rate": 7.272589848534599e-06, "loss": 0.0347, "step": 39035 }, { "epoch": 12.01, "learning_rate": 7.272110472540479e-06, "loss": 0.0318, "step": 39036 }, { "epoch": 12.01, "learning_rate": 7.271631103318975e-06, "loss": 0.0412, "step": 39037 }, { "epoch": 12.01, "learning_rate": 7.271151740871274e-06, "loss": 0.0314, "step": 39038 }, { "epoch": 12.01, "learning_rate": 7.270672385198567e-06, "loss": 0.0522, "step": 39039 }, { "epoch": 12.01, "learning_rate": 7.2701930363020435e-06, "loss": 0.0231, "step": 39040 }, { "epoch": 12.01, "learning_rate": 7.269713694182896e-06, "loss": 0.036, "step": 39041 }, { "epoch": 12.01, "learning_rate": 7.269234358842314e-06, "loss": 0.0165, "step": 39042 }, { "epoch": 12.01, "learning_rate": 7.268755030281489e-06, "loss": 0.0215, "step": 39043 }, { "epoch": 12.01, "learning_rate": 7.268275708501606e-06, "loss": 0.0373, "step": 39044 }, { "epoch": 12.01, "learning_rate": 7.267796393503858e-06, "loss": 0.0415, "step": 39045 }, { "epoch": 12.01, "learning_rate": 7.267317085289436e-06, "loss": 0.0282, "step": 39046 }, { "epoch": 12.01, "learning_rate": 7.266837783859528e-06, "loss": 0.023, "step": 39047 }, { "epoch": 12.01, "learning_rate": 7.266358489215326e-06, "loss": 0.0435, "step": 39048 }, { "epoch": 12.01, "learning_rate": 7.265879201358019e-06, "loss": 0.0184, "step": 39049 }, { "epoch": 12.02, "learning_rate": 7.265399920288794e-06, "loss": 0.0351, "step": 39050 }, { "epoch": 12.02, "learning_rate": 7.264920646008846e-06, "loss": 0.0462, "step": 39051 }, { "epoch": 12.02, "learning_rate": 7.264441378519363e-06, "loss": 0.0367, "step": 39052 }, { "epoch": 12.02, "learning_rate": 7.263962117821537e-06, "loss": 0.0394, "step": 39053 }, { "epoch": 12.02, "learning_rate": 7.2634828639165525e-06, "loss": 0.0188, "step": 39054 }, { "epoch": 12.02, "learning_rate": 7.263003616805599e-06, "loss": 0.0284, "step": 39055 }, { "epoch": 12.02, "learning_rate": 7.2625243764898746e-06, "loss": 0.0198, "step": 39056 }, { "epoch": 12.02, "learning_rate": 7.262045142970563e-06, "loss": 0.0362, "step": 39057 }, { "epoch": 12.02, "learning_rate": 7.261565916248856e-06, "loss": 0.0274, "step": 39058 }, { "epoch": 12.02, "learning_rate": 7.261086696325941e-06, "loss": 0.0273, "step": 39059 }, { "epoch": 12.02, "learning_rate": 7.260607483203008e-06, "loss": 0.0336, "step": 39060 }, { "epoch": 12.02, "learning_rate": 7.2601282768812495e-06, "loss": 0.0395, "step": 39061 }, { "epoch": 12.02, "learning_rate": 7.259649077361854e-06, "loss": 0.0228, "step": 39062 }, { "epoch": 12.02, "learning_rate": 7.2591698846460146e-06, "loss": 0.0359, "step": 39063 }, { "epoch": 12.02, "learning_rate": 7.258690698734913e-06, "loss": 0.0385, "step": 39064 }, { "epoch": 12.02, "learning_rate": 7.258211519629742e-06, "loss": 0.0205, "step": 39065 }, { "epoch": 12.02, "learning_rate": 7.257732347331698e-06, "loss": 0.023, "step": 39066 }, { "epoch": 12.02, "learning_rate": 7.257253181841962e-06, "loss": 0.0345, "step": 39067 }, { "epoch": 12.02, "learning_rate": 7.256774023161728e-06, "loss": 0.0307, "step": 39068 }, { "epoch": 12.02, "learning_rate": 7.256294871292184e-06, "loss": 0.0351, "step": 39069 }, { "epoch": 12.02, "learning_rate": 7.2558157262345186e-06, "loss": 0.0371, "step": 39070 }, { "epoch": 12.02, "learning_rate": 7.255336587989924e-06, "loss": 0.0322, "step": 39071 }, { "epoch": 12.02, "learning_rate": 7.254857456559589e-06, "loss": 0.0481, "step": 39072 }, { "epoch": 12.02, "learning_rate": 7.254378331944703e-06, "loss": 0.0333, "step": 39073 }, { "epoch": 12.02, "learning_rate": 7.253899214146458e-06, "loss": 0.0274, "step": 39074 }, { "epoch": 12.02, "learning_rate": 7.253420103166036e-06, "loss": 0.0302, "step": 39075 }, { "epoch": 12.02, "learning_rate": 7.252940999004633e-06, "loss": 0.0268, "step": 39076 }, { "epoch": 12.02, "learning_rate": 7.2524619016634404e-06, "loss": 0.0309, "step": 39077 }, { "epoch": 12.02, "learning_rate": 7.2519828111436405e-06, "loss": 0.0215, "step": 39078 }, { "epoch": 12.02, "learning_rate": 7.251503727446428e-06, "loss": 0.0286, "step": 39079 }, { "epoch": 12.02, "learning_rate": 7.251024650572989e-06, "loss": 0.0358, "step": 39080 }, { "epoch": 12.02, "learning_rate": 7.250545580524515e-06, "loss": 0.0177, "step": 39081 }, { "epoch": 12.03, "learning_rate": 7.250066517302197e-06, "loss": 0.0297, "step": 39082 }, { "epoch": 12.03, "learning_rate": 7.249587460907221e-06, "loss": 0.0307, "step": 39083 }, { "epoch": 12.03, "learning_rate": 7.2491084113407816e-06, "loss": 0.0252, "step": 39084 }, { "epoch": 12.03, "learning_rate": 7.248629368604057e-06, "loss": 0.0602, "step": 39085 }, { "epoch": 12.03, "learning_rate": 7.248150332698248e-06, "loss": 0.0278, "step": 39086 }, { "epoch": 12.03, "learning_rate": 7.247671303624542e-06, "loss": 0.0157, "step": 39087 }, { "epoch": 12.03, "learning_rate": 7.2471922813841245e-06, "loss": 0.0403, "step": 39088 }, { "epoch": 12.03, "learning_rate": 7.246713265978187e-06, "loss": 0.0244, "step": 39089 }, { "epoch": 12.03, "learning_rate": 7.246234257407914e-06, "loss": 0.0152, "step": 39090 }, { "epoch": 12.03, "learning_rate": 7.245755255674503e-06, "loss": 0.0205, "step": 39091 }, { "epoch": 12.03, "learning_rate": 7.245276260779138e-06, "loss": 0.0411, "step": 39092 }, { "epoch": 12.03, "learning_rate": 7.2447972727230095e-06, "loss": 0.072, "step": 39093 }, { "epoch": 12.03, "learning_rate": 7.244318291507308e-06, "loss": 0.0316, "step": 39094 }, { "epoch": 12.03, "learning_rate": 7.243839317133218e-06, "loss": 0.0209, "step": 39095 }, { "epoch": 12.03, "learning_rate": 7.243360349601933e-06, "loss": 0.0594, "step": 39096 }, { "epoch": 12.03, "learning_rate": 7.2428813889146435e-06, "loss": 0.0271, "step": 39097 }, { "epoch": 12.03, "learning_rate": 7.242402435072537e-06, "loss": 0.0398, "step": 39098 }, { "epoch": 12.03, "learning_rate": 7.241923488076799e-06, "loss": 0.0214, "step": 39099 }, { "epoch": 12.03, "learning_rate": 7.241444547928619e-06, "loss": 0.0248, "step": 39100 }, { "epoch": 12.03, "learning_rate": 7.240965614629194e-06, "loss": 0.031, "step": 39101 }, { "epoch": 12.03, "learning_rate": 7.240486688179704e-06, "loss": 0.0365, "step": 39102 }, { "epoch": 12.03, "learning_rate": 7.2400077685813445e-06, "loss": 0.0189, "step": 39103 }, { "epoch": 12.03, "learning_rate": 7.239528855835299e-06, "loss": 0.052, "step": 39104 }, { "epoch": 12.03, "learning_rate": 7.239049949942758e-06, "loss": 0.0599, "step": 39105 }, { "epoch": 12.03, "learning_rate": 7.238571050904914e-06, "loss": 0.0566, "step": 39106 }, { "epoch": 12.03, "learning_rate": 7.2380921587229536e-06, "loss": 0.0249, "step": 39107 }, { "epoch": 12.03, "learning_rate": 7.2376132733980675e-06, "loss": 0.0392, "step": 39108 }, { "epoch": 12.03, "learning_rate": 7.237134394931442e-06, "loss": 0.0444, "step": 39109 }, { "epoch": 12.03, "learning_rate": 7.2366555233242605e-06, "loss": 0.0126, "step": 39110 }, { "epoch": 12.03, "learning_rate": 7.236176658577727e-06, "loss": 0.0347, "step": 39111 }, { "epoch": 12.03, "learning_rate": 7.235697800693019e-06, "loss": 0.0439, "step": 39112 }, { "epoch": 12.03, "learning_rate": 7.235218949671328e-06, "loss": 0.0204, "step": 39113 }, { "epoch": 12.03, "learning_rate": 7.234740105513842e-06, "loss": 0.0281, "step": 39114 }, { "epoch": 12.04, "learning_rate": 7.23426126822175e-06, "loss": 0.0243, "step": 39115 }, { "epoch": 12.04, "learning_rate": 7.233782437796243e-06, "loss": 0.0154, "step": 39116 }, { "epoch": 12.04, "learning_rate": 7.23330361423851e-06, "loss": 0.0295, "step": 39117 }, { "epoch": 12.04, "learning_rate": 7.232824797549736e-06, "loss": 0.019, "step": 39118 }, { "epoch": 12.04, "learning_rate": 7.232345987731115e-06, "loss": 0.0261, "step": 39119 }, { "epoch": 12.04, "learning_rate": 7.231867184783826e-06, "loss": 0.0174, "step": 39120 }, { "epoch": 12.04, "learning_rate": 7.231388388709072e-06, "loss": 0.017, "step": 39121 }, { "epoch": 12.04, "learning_rate": 7.230909599508031e-06, "loss": 0.0438, "step": 39122 }, { "epoch": 12.04, "learning_rate": 7.230430817181896e-06, "loss": 0.0504, "step": 39123 }, { "epoch": 12.04, "learning_rate": 7.229952041731854e-06, "loss": 0.0298, "step": 39124 }, { "epoch": 12.04, "learning_rate": 7.229473273159092e-06, "loss": 0.0248, "step": 39125 }, { "epoch": 12.04, "learning_rate": 7.228994511464804e-06, "loss": 0.0308, "step": 39126 }, { "epoch": 12.04, "learning_rate": 7.228515756650174e-06, "loss": 0.0335, "step": 39127 }, { "epoch": 12.04, "learning_rate": 7.228037008716393e-06, "loss": 0.0324, "step": 39128 }, { "epoch": 12.04, "learning_rate": 7.227558267664652e-06, "loss": 0.0289, "step": 39129 }, { "epoch": 12.04, "learning_rate": 7.227079533496128e-06, "loss": 0.0147, "step": 39130 }, { "epoch": 12.04, "learning_rate": 7.226600806212024e-06, "loss": 0.0115, "step": 39131 }, { "epoch": 12.04, "learning_rate": 7.226122085813524e-06, "loss": 0.0329, "step": 39132 }, { "epoch": 12.04, "learning_rate": 7.225643372301812e-06, "loss": 0.0299, "step": 39133 }, { "epoch": 12.04, "learning_rate": 7.225164665678082e-06, "loss": 0.032, "step": 39134 }, { "epoch": 12.04, "learning_rate": 7.224685965943515e-06, "loss": 0.0233, "step": 39135 }, { "epoch": 12.04, "learning_rate": 7.2242072730993085e-06, "loss": 0.0427, "step": 39136 }, { "epoch": 12.04, "learning_rate": 7.223728587146646e-06, "loss": 0.0312, "step": 39137 }, { "epoch": 12.04, "learning_rate": 7.223249908086718e-06, "loss": 0.0412, "step": 39138 }, { "epoch": 12.04, "learning_rate": 7.222771235920712e-06, "loss": 0.0309, "step": 39139 }, { "epoch": 12.04, "learning_rate": 7.222292570649813e-06, "loss": 0.0317, "step": 39140 }, { "epoch": 12.04, "learning_rate": 7.221813912275215e-06, "loss": 0.0186, "step": 39141 }, { "epoch": 12.04, "learning_rate": 7.221335260798107e-06, "loss": 0.0372, "step": 39142 }, { "epoch": 12.04, "learning_rate": 7.220856616219671e-06, "loss": 0.055, "step": 39143 }, { "epoch": 12.04, "learning_rate": 7.2203779785411e-06, "loss": 0.0369, "step": 39144 }, { "epoch": 12.04, "learning_rate": 7.219899347763579e-06, "loss": 0.0543, "step": 39145 }, { "epoch": 12.04, "learning_rate": 7.219420723888301e-06, "loss": 0.0391, "step": 39146 }, { "epoch": 12.05, "learning_rate": 7.218942106916451e-06, "loss": 0.0326, "step": 39147 }, { "epoch": 12.05, "learning_rate": 7.218463496849218e-06, "loss": 0.0303, "step": 39148 }, { "epoch": 12.05, "learning_rate": 7.217984893687789e-06, "loss": 0.044, "step": 39149 }, { "epoch": 12.05, "learning_rate": 7.217506297433353e-06, "loss": 0.0423, "step": 39150 }, { "epoch": 12.05, "learning_rate": 7.217027708087101e-06, "loss": 0.0237, "step": 39151 }, { "epoch": 12.05, "learning_rate": 7.2165491256502186e-06, "loss": 0.0352, "step": 39152 }, { "epoch": 12.05, "learning_rate": 7.216070550123896e-06, "loss": 0.0289, "step": 39153 }, { "epoch": 12.05, "learning_rate": 7.215591981509318e-06, "loss": 0.0152, "step": 39154 }, { "epoch": 12.05, "learning_rate": 7.215113419807669e-06, "loss": 0.0431, "step": 39155 }, { "epoch": 12.05, "learning_rate": 7.21463486502015e-06, "loss": 0.0295, "step": 39156 }, { "epoch": 12.05, "learning_rate": 7.21415631714794e-06, "loss": 0.0511, "step": 39157 }, { "epoch": 12.05, "learning_rate": 7.213677776192228e-06, "loss": 0.0166, "step": 39158 }, { "epoch": 12.05, "learning_rate": 7.213199242154203e-06, "loss": 0.035, "step": 39159 }, { "epoch": 12.05, "learning_rate": 7.212720715035052e-06, "loss": 0.03, "step": 39160 }, { "epoch": 12.05, "learning_rate": 7.212242194835965e-06, "loss": 0.0288, "step": 39161 }, { "epoch": 12.05, "learning_rate": 7.21176368155813e-06, "loss": 0.033, "step": 39162 }, { "epoch": 12.05, "learning_rate": 7.211285175202736e-06, "loss": 0.0234, "step": 39163 }, { "epoch": 12.05, "learning_rate": 7.210806675770966e-06, "loss": 0.0386, "step": 39164 }, { "epoch": 12.05, "learning_rate": 7.210328183264007e-06, "loss": 0.0333, "step": 39165 }, { "epoch": 12.05, "learning_rate": 7.209849697683058e-06, "loss": 0.0598, "step": 39166 }, { "epoch": 12.05, "learning_rate": 7.209371219029298e-06, "loss": 0.0471, "step": 39167 }, { "epoch": 12.05, "learning_rate": 7.208892747303917e-06, "loss": 0.0262, "step": 39168 }, { "epoch": 12.05, "learning_rate": 7.208414282508103e-06, "loss": 0.0415, "step": 39169 }, { "epoch": 12.05, "learning_rate": 7.20793582464304e-06, "loss": 0.0093, "step": 39170 }, { "epoch": 12.05, "learning_rate": 7.207457373709923e-06, "loss": 0.0611, "step": 39171 }, { "epoch": 12.05, "learning_rate": 7.2069789297099355e-06, "loss": 0.0447, "step": 39172 }, { "epoch": 12.05, "learning_rate": 7.2065004926442685e-06, "loss": 0.0452, "step": 39173 }, { "epoch": 12.05, "learning_rate": 7.206022062514108e-06, "loss": 0.0381, "step": 39174 }, { "epoch": 12.05, "learning_rate": 7.205543639320635e-06, "loss": 0.0234, "step": 39175 }, { "epoch": 12.05, "learning_rate": 7.20506522306505e-06, "loss": 0.0165, "step": 39176 }, { "epoch": 12.05, "learning_rate": 7.204586813748536e-06, "loss": 0.0167, "step": 39177 }, { "epoch": 12.05, "learning_rate": 7.204108411372275e-06, "loss": 0.0288, "step": 39178 }, { "epoch": 12.05, "learning_rate": 7.2036300159374614e-06, "loss": 0.0201, "step": 39179 }, { "epoch": 12.06, "learning_rate": 7.2031516274452775e-06, "loss": 0.0247, "step": 39180 }, { "epoch": 12.06, "learning_rate": 7.202673245896916e-06, "loss": 0.0286, "step": 39181 }, { "epoch": 12.06, "learning_rate": 7.202194871293564e-06, "loss": 0.0285, "step": 39182 }, { "epoch": 12.06, "learning_rate": 7.201716503636407e-06, "loss": 0.0362, "step": 39183 }, { "epoch": 12.06, "learning_rate": 7.201238142926634e-06, "loss": 0.0242, "step": 39184 }, { "epoch": 12.06, "learning_rate": 7.2007597891654275e-06, "loss": 0.0507, "step": 39185 }, { "epoch": 12.06, "learning_rate": 7.200281442353985e-06, "loss": 0.0213, "step": 39186 }, { "epoch": 12.06, "learning_rate": 7.19980310249349e-06, "loss": 0.0484, "step": 39187 }, { "epoch": 12.06, "learning_rate": 7.199324769585126e-06, "loss": 0.0344, "step": 39188 }, { "epoch": 12.06, "learning_rate": 7.198846443630085e-06, "loss": 0.0206, "step": 39189 }, { "epoch": 12.06, "learning_rate": 7.198368124629551e-06, "loss": 0.0331, "step": 39190 }, { "epoch": 12.06, "learning_rate": 7.197889812584715e-06, "loss": 0.0253, "step": 39191 }, { "epoch": 12.06, "learning_rate": 7.197411507496763e-06, "loss": 0.028, "step": 39192 }, { "epoch": 12.06, "learning_rate": 7.196933209366883e-06, "loss": 0.0627, "step": 39193 }, { "epoch": 12.06, "learning_rate": 7.1964549181962615e-06, "loss": 0.0372, "step": 39194 }, { "epoch": 12.06, "learning_rate": 7.195976633986086e-06, "loss": 0.0461, "step": 39195 }, { "epoch": 12.06, "learning_rate": 7.195498356737545e-06, "loss": 0.0305, "step": 39196 }, { "epoch": 12.06, "learning_rate": 7.195020086451827e-06, "loss": 0.0158, "step": 39197 }, { "epoch": 12.06, "learning_rate": 7.194541823130119e-06, "loss": 0.0257, "step": 39198 }, { "epoch": 12.06, "learning_rate": 7.194063566773606e-06, "loss": 0.0281, "step": 39199 }, { "epoch": 12.06, "learning_rate": 7.193585317383474e-06, "loss": 0.0202, "step": 39200 }, { "epoch": 12.06, "learning_rate": 7.1931070749609155e-06, "loss": 0.0257, "step": 39201 }, { "epoch": 12.06, "learning_rate": 7.192628839507116e-06, "loss": 0.0279, "step": 39202 }, { "epoch": 12.06, "learning_rate": 7.192150611023261e-06, "loss": 0.0122, "step": 39203 }, { "epoch": 12.06, "learning_rate": 7.1916723895105404e-06, "loss": 0.0277, "step": 39204 }, { "epoch": 12.06, "learning_rate": 7.191194174970137e-06, "loss": 0.0254, "step": 39205 }, { "epoch": 12.06, "learning_rate": 7.190715967403243e-06, "loss": 0.0584, "step": 39206 }, { "epoch": 12.06, "learning_rate": 7.190237766811044e-06, "loss": 0.0273, "step": 39207 }, { "epoch": 12.06, "learning_rate": 7.18975957319473e-06, "loss": 0.0284, "step": 39208 }, { "epoch": 12.06, "learning_rate": 7.189281386555479e-06, "loss": 0.0308, "step": 39209 }, { "epoch": 12.06, "learning_rate": 7.188803206894488e-06, "loss": 0.0953, "step": 39210 }, { "epoch": 12.06, "learning_rate": 7.188325034212944e-06, "loss": 0.0375, "step": 39211 }, { "epoch": 12.07, "learning_rate": 7.187846868512028e-06, "loss": 0.0298, "step": 39212 }, { "epoch": 12.07, "learning_rate": 7.18736870979293e-06, "loss": 0.0571, "step": 39213 }, { "epoch": 12.07, "learning_rate": 7.186890558056836e-06, "loss": 0.0383, "step": 39214 }, { "epoch": 12.07, "learning_rate": 7.1864124133049365e-06, "loss": 0.0271, "step": 39215 }, { "epoch": 12.07, "learning_rate": 7.185934275538416e-06, "loss": 0.0432, "step": 39216 }, { "epoch": 12.07, "learning_rate": 7.1854561447584615e-06, "loss": 0.0257, "step": 39217 }, { "epoch": 12.07, "learning_rate": 7.1849780209662614e-06, "loss": 0.0212, "step": 39218 }, { "epoch": 12.07, "learning_rate": 7.184499904162999e-06, "loss": 0.0402, "step": 39219 }, { "epoch": 12.07, "learning_rate": 7.184021794349867e-06, "loss": 0.0396, "step": 39220 }, { "epoch": 12.07, "learning_rate": 7.183543691528052e-06, "loss": 0.0447, "step": 39221 }, { "epoch": 12.07, "learning_rate": 7.1830655956987375e-06, "loss": 0.0312, "step": 39222 }, { "epoch": 12.07, "learning_rate": 7.18258750686311e-06, "loss": 0.0444, "step": 39223 }, { "epoch": 12.07, "learning_rate": 7.182109425022357e-06, "loss": 0.0366, "step": 39224 }, { "epoch": 12.07, "learning_rate": 7.181631350177667e-06, "loss": 0.0339, "step": 39225 }, { "epoch": 12.07, "learning_rate": 7.181153282330229e-06, "loss": 0.0297, "step": 39226 }, { "epoch": 12.07, "learning_rate": 7.180675221481226e-06, "loss": 0.0368, "step": 39227 }, { "epoch": 12.07, "learning_rate": 7.180197167631846e-06, "loss": 0.0301, "step": 39228 }, { "epoch": 12.07, "learning_rate": 7.179719120783275e-06, "loss": 0.0285, "step": 39229 }, { "epoch": 12.07, "learning_rate": 7.179241080936703e-06, "loss": 0.0135, "step": 39230 }, { "epoch": 12.07, "learning_rate": 7.1787630480933155e-06, "loss": 0.0417, "step": 39231 }, { "epoch": 12.07, "learning_rate": 7.178285022254301e-06, "loss": 0.0244, "step": 39232 }, { "epoch": 12.07, "learning_rate": 7.17780700342084e-06, "loss": 0.0528, "step": 39233 }, { "epoch": 12.07, "learning_rate": 7.177328991594121e-06, "loss": 0.0239, "step": 39234 }, { "epoch": 12.07, "learning_rate": 7.176850986775339e-06, "loss": 0.0232, "step": 39235 }, { "epoch": 12.07, "learning_rate": 7.176372988965673e-06, "loss": 0.0414, "step": 39236 }, { "epoch": 12.07, "learning_rate": 7.175894998166311e-06, "loss": 0.0139, "step": 39237 }, { "epoch": 12.07, "learning_rate": 7.175417014378441e-06, "loss": 0.0269, "step": 39238 }, { "epoch": 12.07, "learning_rate": 7.1749390376032464e-06, "loss": 0.053, "step": 39239 }, { "epoch": 12.07, "learning_rate": 7.17446106784192e-06, "loss": 0.0224, "step": 39240 }, { "epoch": 12.07, "learning_rate": 7.173983105095643e-06, "loss": 0.0417, "step": 39241 }, { "epoch": 12.07, "learning_rate": 7.173505149365607e-06, "loss": 0.0338, "step": 39242 }, { "epoch": 12.07, "learning_rate": 7.173027200652996e-06, "loss": 0.0357, "step": 39243 }, { "epoch": 12.07, "learning_rate": 7.172549258958989e-06, "loss": 0.0345, "step": 39244 }, { "epoch": 12.08, "learning_rate": 7.172071324284786e-06, "loss": 0.0238, "step": 39245 }, { "epoch": 12.08, "learning_rate": 7.1715933966315675e-06, "loss": 0.0177, "step": 39246 }, { "epoch": 12.08, "learning_rate": 7.171115476000519e-06, "loss": 0.0279, "step": 39247 }, { "epoch": 12.08, "learning_rate": 7.1706375623928286e-06, "loss": 0.0407, "step": 39248 }, { "epoch": 12.08, "learning_rate": 7.17015965580968e-06, "loss": 0.0282, "step": 39249 }, { "epoch": 12.08, "learning_rate": 7.169681756252265e-06, "loss": 0.0433, "step": 39250 }, { "epoch": 12.08, "learning_rate": 7.169203863721765e-06, "loss": 0.0421, "step": 39251 }, { "epoch": 12.08, "learning_rate": 7.168725978219371e-06, "loss": 0.0347, "step": 39252 }, { "epoch": 12.08, "learning_rate": 7.168248099746268e-06, "loss": 0.0348, "step": 39253 }, { "epoch": 12.08, "learning_rate": 7.1677702283036344e-06, "loss": 0.0681, "step": 39254 }, { "epoch": 12.08, "learning_rate": 7.16729236389267e-06, "loss": 0.0371, "step": 39255 }, { "epoch": 12.08, "learning_rate": 7.166814506514555e-06, "loss": 0.0491, "step": 39256 }, { "epoch": 12.08, "learning_rate": 7.166336656170476e-06, "loss": 0.05, "step": 39257 }, { "epoch": 12.08, "learning_rate": 7.165858812861617e-06, "loss": 0.0362, "step": 39258 }, { "epoch": 12.08, "learning_rate": 7.165380976589166e-06, "loss": 0.0566, "step": 39259 }, { "epoch": 12.08, "learning_rate": 7.164903147354312e-06, "loss": 0.0354, "step": 39260 }, { "epoch": 12.08, "learning_rate": 7.1644253251582374e-06, "loss": 0.0237, "step": 39261 }, { "epoch": 12.08, "learning_rate": 7.163947510002132e-06, "loss": 0.0238, "step": 39262 }, { "epoch": 12.08, "learning_rate": 7.163469701887182e-06, "loss": 0.0281, "step": 39263 }, { "epoch": 12.08, "learning_rate": 7.162991900814564e-06, "loss": 0.0352, "step": 39264 }, { "epoch": 12.08, "learning_rate": 7.162514106785479e-06, "loss": 0.031, "step": 39265 }, { "epoch": 12.08, "learning_rate": 7.162036319801106e-06, "loss": 0.025, "step": 39266 }, { "epoch": 12.08, "learning_rate": 7.161558539862632e-06, "loss": 0.0301, "step": 39267 }, { "epoch": 12.08, "learning_rate": 7.1610807669712425e-06, "loss": 0.0248, "step": 39268 }, { "epoch": 12.08, "learning_rate": 7.160603001128121e-06, "loss": 0.0405, "step": 39269 }, { "epoch": 12.08, "learning_rate": 7.1601252423344604e-06, "loss": 0.0244, "step": 39270 }, { "epoch": 12.08, "learning_rate": 7.159647490591443e-06, "loss": 0.0303, "step": 39271 }, { "epoch": 12.08, "learning_rate": 7.159169745900254e-06, "loss": 0.0319, "step": 39272 }, { "epoch": 12.08, "learning_rate": 7.15869200826208e-06, "loss": 0.0398, "step": 39273 }, { "epoch": 12.08, "learning_rate": 7.158214277678109e-06, "loss": 0.0287, "step": 39274 }, { "epoch": 12.08, "learning_rate": 7.157736554149524e-06, "loss": 0.0272, "step": 39275 }, { "epoch": 12.08, "learning_rate": 7.157258837677514e-06, "loss": 0.0191, "step": 39276 }, { "epoch": 12.09, "learning_rate": 7.156781128263268e-06, "loss": 0.0554, "step": 39277 }, { "epoch": 12.09, "learning_rate": 7.156303425907965e-06, "loss": 0.0386, "step": 39278 }, { "epoch": 12.09, "learning_rate": 7.155825730612791e-06, "loss": 0.0225, "step": 39279 }, { "epoch": 12.09, "learning_rate": 7.155348042378937e-06, "loss": 0.027, "step": 39280 }, { "epoch": 12.09, "learning_rate": 7.154870361207587e-06, "loss": 0.0343, "step": 39281 }, { "epoch": 12.09, "learning_rate": 7.154392687099927e-06, "loss": 0.0206, "step": 39282 }, { "epoch": 12.09, "learning_rate": 7.153915020057143e-06, "loss": 0.0303, "step": 39283 }, { "epoch": 12.09, "learning_rate": 7.15343736008042e-06, "loss": 0.0288, "step": 39284 }, { "epoch": 12.09, "learning_rate": 7.152959707170944e-06, "loss": 0.0665, "step": 39285 }, { "epoch": 12.09, "learning_rate": 7.152482061329903e-06, "loss": 0.0164, "step": 39286 }, { "epoch": 12.09, "learning_rate": 7.152004422558483e-06, "loss": 0.0289, "step": 39287 }, { "epoch": 12.09, "learning_rate": 7.1515267908578665e-06, "loss": 0.0339, "step": 39288 }, { "epoch": 12.09, "learning_rate": 7.151049166229237e-06, "loss": 0.0398, "step": 39289 }, { "epoch": 12.09, "learning_rate": 7.1505715486737905e-06, "loss": 0.018, "step": 39290 }, { "epoch": 12.09, "learning_rate": 7.1500939381927044e-06, "loss": 0.0128, "step": 39291 }, { "epoch": 12.09, "learning_rate": 7.1496163347871675e-06, "loss": 0.0478, "step": 39292 }, { "epoch": 12.09, "learning_rate": 7.149138738458363e-06, "loss": 0.0261, "step": 39293 }, { "epoch": 12.09, "learning_rate": 7.1486611492074784e-06, "loss": 0.0416, "step": 39294 }, { "epoch": 12.09, "learning_rate": 7.1481835670357005e-06, "loss": 0.0233, "step": 39295 }, { "epoch": 12.09, "learning_rate": 7.1477059919442135e-06, "loss": 0.056, "step": 39296 }, { "epoch": 12.09, "learning_rate": 7.147228423934205e-06, "loss": 0.0241, "step": 39297 }, { "epoch": 12.09, "learning_rate": 7.14675086300686e-06, "loss": 0.0369, "step": 39298 }, { "epoch": 12.09, "learning_rate": 7.146273309163358e-06, "loss": 0.0442, "step": 39299 }, { "epoch": 12.09, "learning_rate": 7.145795762404896e-06, "loss": 0.0354, "step": 39300 }, { "epoch": 12.09, "learning_rate": 7.1453182227326516e-06, "loss": 0.0487, "step": 39301 }, { "epoch": 12.09, "learning_rate": 7.144840690147812e-06, "loss": 0.0187, "step": 39302 }, { "epoch": 12.09, "learning_rate": 7.144363164651564e-06, "loss": 0.0546, "step": 39303 }, { "epoch": 12.09, "learning_rate": 7.14388564624509e-06, "loss": 0.0443, "step": 39304 }, { "epoch": 12.09, "learning_rate": 7.143408134929579e-06, "loss": 0.0142, "step": 39305 }, { "epoch": 12.09, "learning_rate": 7.142930630706217e-06, "loss": 0.032, "step": 39306 }, { "epoch": 12.09, "learning_rate": 7.142453133576188e-06, "loss": 0.0336, "step": 39307 }, { "epoch": 12.09, "learning_rate": 7.141975643540679e-06, "loss": 0.0232, "step": 39308 }, { "epoch": 12.09, "learning_rate": 7.1414981606008685e-06, "loss": 0.0332, "step": 39309 }, { "epoch": 12.1, "learning_rate": 7.14102068475795e-06, "loss": 0.0369, "step": 39310 }, { "epoch": 12.1, "learning_rate": 7.140543216013109e-06, "loss": 0.0248, "step": 39311 }, { "epoch": 12.1, "learning_rate": 7.1400657543675264e-06, "loss": 0.055, "step": 39312 }, { "epoch": 12.1, "learning_rate": 7.139588299822391e-06, "loss": 0.0319, "step": 39313 }, { "epoch": 12.1, "learning_rate": 7.139110852378883e-06, "loss": 0.0199, "step": 39314 }, { "epoch": 12.1, "learning_rate": 7.138633412038194e-06, "loss": 0.0247, "step": 39315 }, { "epoch": 12.1, "learning_rate": 7.1381559788015065e-06, "loss": 0.0333, "step": 39316 }, { "epoch": 12.1, "learning_rate": 7.137678552670007e-06, "loss": 0.0387, "step": 39317 }, { "epoch": 12.1, "learning_rate": 7.137201133644877e-06, "loss": 0.0413, "step": 39318 }, { "epoch": 12.1, "learning_rate": 7.136723721727306e-06, "loss": 0.0183, "step": 39319 }, { "epoch": 12.1, "learning_rate": 7.1362463169184804e-06, "loss": 0.0203, "step": 39320 }, { "epoch": 12.1, "learning_rate": 7.135768919219583e-06, "loss": 0.0333, "step": 39321 }, { "epoch": 12.1, "learning_rate": 7.135291528631798e-06, "loss": 0.0389, "step": 39322 }, { "epoch": 12.1, "learning_rate": 7.134814145156312e-06, "loss": 0.0476, "step": 39323 }, { "epoch": 12.1, "learning_rate": 7.1343367687943085e-06, "loss": 0.0631, "step": 39324 }, { "epoch": 12.1, "learning_rate": 7.133859399546976e-06, "loss": 0.029, "step": 39325 }, { "epoch": 12.1, "learning_rate": 7.133382037415497e-06, "loss": 0.013, "step": 39326 }, { "epoch": 12.1, "learning_rate": 7.13290468240106e-06, "loss": 0.0314, "step": 39327 }, { "epoch": 12.1, "learning_rate": 7.132427334504846e-06, "loss": 0.0317, "step": 39328 }, { "epoch": 12.1, "learning_rate": 7.131949993728039e-06, "loss": 0.0706, "step": 39329 }, { "epoch": 12.1, "learning_rate": 7.13147266007183e-06, "loss": 0.0452, "step": 39330 }, { "epoch": 12.1, "learning_rate": 7.130995333537401e-06, "loss": 0.0287, "step": 39331 }, { "epoch": 12.1, "learning_rate": 7.13051801412594e-06, "loss": 0.0369, "step": 39332 }, { "epoch": 12.1, "learning_rate": 7.130040701838625e-06, "loss": 0.0396, "step": 39333 }, { "epoch": 12.1, "learning_rate": 7.129563396676643e-06, "loss": 0.015, "step": 39334 }, { "epoch": 12.1, "learning_rate": 7.129086098641187e-06, "loss": 0.0218, "step": 39335 }, { "epoch": 12.1, "learning_rate": 7.128608807733434e-06, "loss": 0.0259, "step": 39336 }, { "epoch": 12.1, "learning_rate": 7.128131523954571e-06, "loss": 0.0252, "step": 39337 }, { "epoch": 12.1, "learning_rate": 7.127654247305783e-06, "loss": 0.0222, "step": 39338 }, { "epoch": 12.1, "learning_rate": 7.127176977788254e-06, "loss": 0.0177, "step": 39339 }, { "epoch": 12.1, "learning_rate": 7.126699715403171e-06, "loss": 0.042, "step": 39340 }, { "epoch": 12.1, "learning_rate": 7.126222460151719e-06, "loss": 0.024, "step": 39341 }, { "epoch": 12.11, "learning_rate": 7.125745212035084e-06, "loss": 0.0352, "step": 39342 }, { "epoch": 12.11, "learning_rate": 7.125267971054446e-06, "loss": 0.0299, "step": 39343 }, { "epoch": 12.11, "learning_rate": 7.124790737210989e-06, "loss": 0.0158, "step": 39344 }, { "epoch": 12.11, "learning_rate": 7.124313510505907e-06, "loss": 0.0339, "step": 39345 }, { "epoch": 12.11, "learning_rate": 7.123836290940378e-06, "loss": 0.0351, "step": 39346 }, { "epoch": 12.11, "learning_rate": 7.123359078515587e-06, "loss": 0.029, "step": 39347 }, { "epoch": 12.11, "learning_rate": 7.122881873232721e-06, "loss": 0.0319, "step": 39348 }, { "epoch": 12.11, "learning_rate": 7.122404675092961e-06, "loss": 0.0363, "step": 39349 }, { "epoch": 12.11, "learning_rate": 7.121927484097497e-06, "loss": 0.0437, "step": 39350 }, { "epoch": 12.11, "learning_rate": 7.12145030024751e-06, "loss": 0.0317, "step": 39351 }, { "epoch": 12.11, "learning_rate": 7.120973123544188e-06, "loss": 0.0441, "step": 39352 }, { "epoch": 12.11, "learning_rate": 7.1204959539887135e-06, "loss": 0.014, "step": 39353 }, { "epoch": 12.11, "learning_rate": 7.120018791582266e-06, "loss": 0.0279, "step": 39354 }, { "epoch": 12.11, "learning_rate": 7.119541636326039e-06, "loss": 0.0415, "step": 39355 }, { "epoch": 12.11, "learning_rate": 7.119064488221217e-06, "loss": 0.0569, "step": 39356 }, { "epoch": 12.11, "learning_rate": 7.1185873472689786e-06, "loss": 0.02, "step": 39357 }, { "epoch": 12.11, "learning_rate": 7.11811021347051e-06, "loss": 0.0375, "step": 39358 }, { "epoch": 12.11, "learning_rate": 7.117633086826996e-06, "loss": 0.0202, "step": 39359 }, { "epoch": 12.11, "learning_rate": 7.117155967339623e-06, "loss": 0.0243, "step": 39360 }, { "epoch": 12.11, "learning_rate": 7.116678855009576e-06, "loss": 0.0243, "step": 39361 }, { "epoch": 12.11, "learning_rate": 7.116201749838037e-06, "loss": 0.0243, "step": 39362 }, { "epoch": 12.11, "learning_rate": 7.115724651826193e-06, "loss": 0.0277, "step": 39363 }, { "epoch": 12.11, "learning_rate": 7.115247560975221e-06, "loss": 0.0313, "step": 39364 }, { "epoch": 12.11, "learning_rate": 7.114770477286317e-06, "loss": 0.0147, "step": 39365 }, { "epoch": 12.11, "learning_rate": 7.114293400760661e-06, "loss": 0.0262, "step": 39366 }, { "epoch": 12.11, "learning_rate": 7.113816331399435e-06, "loss": 0.0339, "step": 39367 }, { "epoch": 12.11, "learning_rate": 7.113339269203825e-06, "loss": 0.0253, "step": 39368 }, { "epoch": 12.11, "learning_rate": 7.112862214175014e-06, "loss": 0.0233, "step": 39369 }, { "epoch": 12.11, "learning_rate": 7.112385166314189e-06, "loss": 0.0131, "step": 39370 }, { "epoch": 12.11, "learning_rate": 7.111908125622533e-06, "loss": 0.0322, "step": 39371 }, { "epoch": 12.11, "learning_rate": 7.1114310921012305e-06, "loss": 0.0312, "step": 39372 }, { "epoch": 12.11, "learning_rate": 7.110954065751466e-06, "loss": 0.0254, "step": 39373 }, { "epoch": 12.11, "learning_rate": 7.110477046574422e-06, "loss": 0.0343, "step": 39374 }, { "epoch": 12.12, "learning_rate": 7.110000034571286e-06, "loss": 0.0211, "step": 39375 }, { "epoch": 12.12, "learning_rate": 7.109523029743242e-06, "loss": 0.0376, "step": 39376 }, { "epoch": 12.12, "learning_rate": 7.109046032091474e-06, "loss": 0.0164, "step": 39377 }, { "epoch": 12.12, "learning_rate": 7.108569041617164e-06, "loss": 0.0317, "step": 39378 }, { "epoch": 12.12, "learning_rate": 7.108092058321494e-06, "loss": 0.0248, "step": 39379 }, { "epoch": 12.12, "learning_rate": 7.107615082205654e-06, "loss": 0.0346, "step": 39380 }, { "epoch": 12.12, "learning_rate": 7.107138113270828e-06, "loss": 0.0332, "step": 39381 }, { "epoch": 12.12, "learning_rate": 7.106661151518197e-06, "loss": 0.0268, "step": 39382 }, { "epoch": 12.12, "learning_rate": 7.106184196948947e-06, "loss": 0.0292, "step": 39383 }, { "epoch": 12.12, "learning_rate": 7.105707249564258e-06, "loss": 0.063, "step": 39384 }, { "epoch": 12.12, "learning_rate": 7.105230309365321e-06, "loss": 0.0606, "step": 39385 }, { "epoch": 12.12, "learning_rate": 7.104753376353315e-06, "loss": 0.0252, "step": 39386 }, { "epoch": 12.12, "learning_rate": 7.10427645052943e-06, "loss": 0.0509, "step": 39387 }, { "epoch": 12.12, "learning_rate": 7.103799531894843e-06, "loss": 0.051, "step": 39388 }, { "epoch": 12.12, "learning_rate": 7.1033226204507364e-06, "loss": 0.026, "step": 39389 }, { "epoch": 12.12, "learning_rate": 7.102845716198306e-06, "loss": 0.0332, "step": 39390 }, { "epoch": 12.12, "learning_rate": 7.102368819138725e-06, "loss": 0.0391, "step": 39391 }, { "epoch": 12.12, "learning_rate": 7.101891929273182e-06, "loss": 0.0379, "step": 39392 }, { "epoch": 12.12, "learning_rate": 7.1014150466028605e-06, "loss": 0.019, "step": 39393 }, { "epoch": 12.12, "learning_rate": 7.100938171128942e-06, "loss": 0.0279, "step": 39394 }, { "epoch": 12.12, "learning_rate": 7.100461302852614e-06, "loss": 0.0374, "step": 39395 }, { "epoch": 12.12, "learning_rate": 7.0999844417750584e-06, "loss": 0.0357, "step": 39396 }, { "epoch": 12.12, "learning_rate": 7.09950758789746e-06, "loss": 0.0316, "step": 39397 }, { "epoch": 12.12, "learning_rate": 7.099030741221005e-06, "loss": 0.0209, "step": 39398 }, { "epoch": 12.12, "learning_rate": 7.0985539017468676e-06, "loss": 0.0358, "step": 39399 }, { "epoch": 12.12, "learning_rate": 7.0980770694762455e-06, "loss": 0.0169, "step": 39400 }, { "epoch": 12.12, "learning_rate": 7.097600244410313e-06, "loss": 0.0175, "step": 39401 }, { "epoch": 12.12, "learning_rate": 7.097123426550257e-06, "loss": 0.0197, "step": 39402 }, { "epoch": 12.12, "learning_rate": 7.0966466158972605e-06, "loss": 0.0338, "step": 39403 }, { "epoch": 12.12, "learning_rate": 7.096169812452507e-06, "loss": 0.047, "step": 39404 }, { "epoch": 12.12, "learning_rate": 7.095693016217182e-06, "loss": 0.0264, "step": 39405 }, { "epoch": 12.12, "learning_rate": 7.095216227192467e-06, "loss": 0.0417, "step": 39406 }, { "epoch": 12.13, "learning_rate": 7.094739445379549e-06, "loss": 0.0332, "step": 39407 }, { "epoch": 12.13, "learning_rate": 7.094262670779611e-06, "loss": 0.0359, "step": 39408 }, { "epoch": 12.13, "learning_rate": 7.09378590339383e-06, "loss": 0.0952, "step": 39409 }, { "epoch": 12.13, "learning_rate": 7.093309143223398e-06, "loss": 0.0336, "step": 39410 }, { "epoch": 12.13, "learning_rate": 7.0928323902695e-06, "loss": 0.036, "step": 39411 }, { "epoch": 12.13, "learning_rate": 7.092355644533311e-06, "loss": 0.0365, "step": 39412 }, { "epoch": 12.13, "learning_rate": 7.091878906016018e-06, "loss": 0.0248, "step": 39413 }, { "epoch": 12.13, "learning_rate": 7.091402174718807e-06, "loss": 0.0367, "step": 39414 }, { "epoch": 12.13, "learning_rate": 7.090925450642861e-06, "loss": 0.0223, "step": 39415 }, { "epoch": 12.13, "learning_rate": 7.090448733789362e-06, "loss": 0.0195, "step": 39416 }, { "epoch": 12.13, "learning_rate": 7.089972024159495e-06, "loss": 0.0241, "step": 39417 }, { "epoch": 12.13, "learning_rate": 7.089495321754444e-06, "loss": 0.0333, "step": 39418 }, { "epoch": 12.13, "learning_rate": 7.089018626575388e-06, "loss": 0.0374, "step": 39419 }, { "epoch": 12.13, "learning_rate": 7.0885419386235165e-06, "loss": 0.02, "step": 39420 }, { "epoch": 12.13, "learning_rate": 7.088065257900012e-06, "loss": 0.0298, "step": 39421 }, { "epoch": 12.13, "learning_rate": 7.0875885844060555e-06, "loss": 0.0364, "step": 39422 }, { "epoch": 12.13, "learning_rate": 7.087111918142831e-06, "loss": 0.0388, "step": 39423 }, { "epoch": 12.13, "learning_rate": 7.08663525911152e-06, "loss": 0.0244, "step": 39424 }, { "epoch": 12.13, "learning_rate": 7.086158607313309e-06, "loss": 0.0275, "step": 39425 }, { "epoch": 12.13, "learning_rate": 7.085681962749382e-06, "loss": 0.0232, "step": 39426 }, { "epoch": 12.13, "learning_rate": 7.08520532542092e-06, "loss": 0.013, "step": 39427 }, { "epoch": 12.13, "learning_rate": 7.084728695329111e-06, "loss": 0.0193, "step": 39428 }, { "epoch": 12.13, "learning_rate": 7.084252072475129e-06, "loss": 0.0495, "step": 39429 }, { "epoch": 12.13, "learning_rate": 7.083775456860166e-06, "loss": 0.0342, "step": 39430 }, { "epoch": 12.13, "learning_rate": 7.083298848485403e-06, "loss": 0.0297, "step": 39431 }, { "epoch": 12.13, "learning_rate": 7.082822247352024e-06, "loss": 0.0195, "step": 39432 }, { "epoch": 12.13, "learning_rate": 7.082345653461209e-06, "loss": 0.0464, "step": 39433 }, { "epoch": 12.13, "learning_rate": 7.081869066814138e-06, "loss": 0.0176, "step": 39434 }, { "epoch": 12.13, "learning_rate": 7.081392487412006e-06, "loss": 0.0474, "step": 39435 }, { "epoch": 12.13, "learning_rate": 7.080915915255989e-06, "loss": 0.0374, "step": 39436 }, { "epoch": 12.13, "learning_rate": 7.08043935034727e-06, "loss": 0.0664, "step": 39437 }, { "epoch": 12.13, "learning_rate": 7.079962792687033e-06, "loss": 0.0583, "step": 39438 }, { "epoch": 12.13, "learning_rate": 7.0794862422764584e-06, "loss": 0.0222, "step": 39439 }, { "epoch": 12.14, "learning_rate": 7.079009699116736e-06, "loss": 0.0202, "step": 39440 }, { "epoch": 12.14, "learning_rate": 7.078533163209043e-06, "loss": 0.0565, "step": 39441 }, { "epoch": 12.14, "learning_rate": 7.078056634554569e-06, "loss": 0.0343, "step": 39442 }, { "epoch": 12.14, "learning_rate": 7.077580113154488e-06, "loss": 0.016, "step": 39443 }, { "epoch": 12.14, "learning_rate": 7.077103599009984e-06, "loss": 0.0456, "step": 39444 }, { "epoch": 12.14, "learning_rate": 7.076627092122251e-06, "loss": 0.0129, "step": 39445 }, { "epoch": 12.14, "learning_rate": 7.076150592492464e-06, "loss": 0.0298, "step": 39446 }, { "epoch": 12.14, "learning_rate": 7.075674100121803e-06, "loss": 0.0256, "step": 39447 }, { "epoch": 12.14, "learning_rate": 7.075197615011458e-06, "loss": 0.0363, "step": 39448 }, { "epoch": 12.14, "learning_rate": 7.074721137162606e-06, "loss": 0.0276, "step": 39449 }, { "epoch": 12.14, "learning_rate": 7.0742446665764356e-06, "loss": 0.0233, "step": 39450 }, { "epoch": 12.14, "learning_rate": 7.073768203254126e-06, "loss": 0.0368, "step": 39451 }, { "epoch": 12.14, "learning_rate": 7.07329174719686e-06, "loss": 0.0491, "step": 39452 }, { "epoch": 12.14, "learning_rate": 7.072815298405824e-06, "loss": 0.0316, "step": 39453 }, { "epoch": 12.14, "learning_rate": 7.072338856882194e-06, "loss": 0.0265, "step": 39454 }, { "epoch": 12.14, "learning_rate": 7.07186242262716e-06, "loss": 0.0199, "step": 39455 }, { "epoch": 12.14, "learning_rate": 7.071385995641905e-06, "loss": 0.0315, "step": 39456 }, { "epoch": 12.14, "learning_rate": 7.070909575927607e-06, "loss": 0.0392, "step": 39457 }, { "epoch": 12.14, "learning_rate": 7.070433163485451e-06, "loss": 0.0227, "step": 39458 }, { "epoch": 12.14, "learning_rate": 7.069956758316616e-06, "loss": 0.0346, "step": 39459 }, { "epoch": 12.14, "learning_rate": 7.069480360422291e-06, "loss": 0.0272, "step": 39460 }, { "epoch": 12.14, "learning_rate": 7.069003969803658e-06, "loss": 0.0262, "step": 39461 }, { "epoch": 12.14, "learning_rate": 7.068527586461897e-06, "loss": 0.0352, "step": 39462 }, { "epoch": 12.14, "learning_rate": 7.0680512103981936e-06, "loss": 0.0305, "step": 39463 }, { "epoch": 12.14, "learning_rate": 7.067574841613722e-06, "loss": 0.033, "step": 39464 }, { "epoch": 12.14, "learning_rate": 7.0670984801096764e-06, "loss": 0.0297, "step": 39465 }, { "epoch": 12.14, "learning_rate": 7.066622125887237e-06, "loss": 0.0443, "step": 39466 }, { "epoch": 12.14, "learning_rate": 7.066145778947581e-06, "loss": 0.0385, "step": 39467 }, { "epoch": 12.14, "learning_rate": 7.065669439291895e-06, "loss": 0.0308, "step": 39468 }, { "epoch": 12.14, "learning_rate": 7.0651931069213574e-06, "loss": 0.0249, "step": 39469 }, { "epoch": 12.14, "learning_rate": 7.0647167818371575e-06, "loss": 0.0356, "step": 39470 }, { "epoch": 12.14, "learning_rate": 7.064240464040472e-06, "loss": 0.0348, "step": 39471 }, { "epoch": 12.15, "learning_rate": 7.063764153532489e-06, "loss": 0.0202, "step": 39472 }, { "epoch": 12.15, "learning_rate": 7.063287850314387e-06, "loss": 0.0379, "step": 39473 }, { "epoch": 12.15, "learning_rate": 7.062811554387348e-06, "loss": 0.0552, "step": 39474 }, { "epoch": 12.15, "learning_rate": 7.062335265752557e-06, "loss": 0.0206, "step": 39475 }, { "epoch": 12.15, "learning_rate": 7.061858984411196e-06, "loss": 0.0413, "step": 39476 }, { "epoch": 12.15, "learning_rate": 7.061382710364449e-06, "loss": 0.0348, "step": 39477 }, { "epoch": 12.15, "learning_rate": 7.060906443613495e-06, "loss": 0.0403, "step": 39478 }, { "epoch": 12.15, "learning_rate": 7.0604301841595145e-06, "loss": 0.0195, "step": 39479 }, { "epoch": 12.15, "learning_rate": 7.059953932003697e-06, "loss": 0.0135, "step": 39480 }, { "epoch": 12.15, "learning_rate": 7.05947768714722e-06, "loss": 0.0165, "step": 39481 }, { "epoch": 12.15, "learning_rate": 7.059001449591268e-06, "loss": 0.0432, "step": 39482 }, { "epoch": 12.15, "learning_rate": 7.058525219337022e-06, "loss": 0.0446, "step": 39483 }, { "epoch": 12.15, "learning_rate": 7.0580489963856646e-06, "loss": 0.0632, "step": 39484 }, { "epoch": 12.15, "learning_rate": 7.057572780738379e-06, "loss": 0.0482, "step": 39485 }, { "epoch": 12.15, "learning_rate": 7.057096572396347e-06, "loss": 0.0239, "step": 39486 }, { "epoch": 12.15, "learning_rate": 7.056620371360753e-06, "loss": 0.0326, "step": 39487 }, { "epoch": 12.15, "learning_rate": 7.056144177632774e-06, "loss": 0.0281, "step": 39488 }, { "epoch": 12.15, "learning_rate": 7.055667991213593e-06, "loss": 0.0288, "step": 39489 }, { "epoch": 12.15, "learning_rate": 7.0551918121044e-06, "loss": 0.0543, "step": 39490 }, { "epoch": 12.15, "learning_rate": 7.054715640306369e-06, "loss": 0.0369, "step": 39491 }, { "epoch": 12.15, "learning_rate": 7.0542394758206854e-06, "loss": 0.0257, "step": 39492 }, { "epoch": 12.15, "learning_rate": 7.053763318648532e-06, "loss": 0.0295, "step": 39493 }, { "epoch": 12.15, "learning_rate": 7.053287168791087e-06, "loss": 0.0298, "step": 39494 }, { "epoch": 12.15, "learning_rate": 7.052811026249539e-06, "loss": 0.0343, "step": 39495 }, { "epoch": 12.15, "learning_rate": 7.052334891025065e-06, "loss": 0.0616, "step": 39496 }, { "epoch": 12.15, "learning_rate": 7.05185876311885e-06, "loss": 0.0376, "step": 39497 }, { "epoch": 12.15, "learning_rate": 7.051382642532077e-06, "loss": 0.0309, "step": 39498 }, { "epoch": 12.15, "learning_rate": 7.050906529265918e-06, "loss": 0.0551, "step": 39499 }, { "epoch": 12.15, "learning_rate": 7.05043042332157e-06, "loss": 0.0189, "step": 39500 }, { "epoch": 12.15, "learning_rate": 7.049954324700208e-06, "loss": 0.0233, "step": 39501 }, { "epoch": 12.15, "learning_rate": 7.0494782334030125e-06, "loss": 0.0311, "step": 39502 }, { "epoch": 12.15, "learning_rate": 7.049002149431167e-06, "loss": 0.0289, "step": 39503 }, { "epoch": 12.15, "learning_rate": 7.048526072785851e-06, "loss": 0.0189, "step": 39504 }, { "epoch": 12.16, "learning_rate": 7.048050003468252e-06, "loss": 0.0304, "step": 39505 }, { "epoch": 12.16, "learning_rate": 7.047573941479549e-06, "loss": 0.0192, "step": 39506 }, { "epoch": 12.16, "learning_rate": 7.047097886820924e-06, "loss": 0.0306, "step": 39507 }, { "epoch": 12.16, "learning_rate": 7.046621839493562e-06, "loss": 0.024, "step": 39508 }, { "epoch": 12.16, "learning_rate": 7.046145799498633e-06, "loss": 0.0161, "step": 39509 }, { "epoch": 12.16, "learning_rate": 7.045669766837333e-06, "loss": 0.0437, "step": 39510 }, { "epoch": 12.16, "learning_rate": 7.0451937415108405e-06, "loss": 0.03, "step": 39511 }, { "epoch": 12.16, "learning_rate": 7.044717723520334e-06, "loss": 0.0211, "step": 39512 }, { "epoch": 12.16, "learning_rate": 7.044241712866996e-06, "loss": 0.0363, "step": 39513 }, { "epoch": 12.16, "learning_rate": 7.0437657095520065e-06, "loss": 0.0329, "step": 39514 }, { "epoch": 12.16, "learning_rate": 7.0432897135765535e-06, "loss": 0.0293, "step": 39515 }, { "epoch": 12.16, "learning_rate": 7.0428137249418125e-06, "loss": 0.0573, "step": 39516 }, { "epoch": 12.16, "learning_rate": 7.04233774364897e-06, "loss": 0.0356, "step": 39517 }, { "epoch": 12.16, "learning_rate": 7.041861769699205e-06, "loss": 0.0423, "step": 39518 }, { "epoch": 12.16, "learning_rate": 7.0413858030936966e-06, "loss": 0.0301, "step": 39519 }, { "epoch": 12.16, "learning_rate": 7.040909843833632e-06, "loss": 0.0421, "step": 39520 }, { "epoch": 12.16, "learning_rate": 7.0404338919201935e-06, "loss": 0.0268, "step": 39521 }, { "epoch": 12.16, "learning_rate": 7.039957947354557e-06, "loss": 0.0438, "step": 39522 }, { "epoch": 12.16, "learning_rate": 7.039482010137908e-06, "loss": 0.0371, "step": 39523 }, { "epoch": 12.16, "learning_rate": 7.039006080271425e-06, "loss": 0.0266, "step": 39524 }, { "epoch": 12.16, "learning_rate": 7.038530157756292e-06, "loss": 0.1289, "step": 39525 }, { "epoch": 12.16, "learning_rate": 7.038054242593691e-06, "loss": 0.0327, "step": 39526 }, { "epoch": 12.16, "learning_rate": 7.037578334784803e-06, "loss": 0.0261, "step": 39527 }, { "epoch": 12.16, "learning_rate": 7.037102434330809e-06, "loss": 0.0428, "step": 39528 }, { "epoch": 12.16, "learning_rate": 7.036626541232889e-06, "loss": 0.0334, "step": 39529 }, { "epoch": 12.16, "learning_rate": 7.0361506554922285e-06, "loss": 0.0266, "step": 39530 }, { "epoch": 12.16, "learning_rate": 7.035674777110006e-06, "loss": 0.0301, "step": 39531 }, { "epoch": 12.16, "learning_rate": 7.035198906087408e-06, "loss": 0.0285, "step": 39532 }, { "epoch": 12.16, "learning_rate": 7.034723042425609e-06, "loss": 0.0498, "step": 39533 }, { "epoch": 12.16, "learning_rate": 7.034247186125789e-06, "loss": 0.0681, "step": 39534 }, { "epoch": 12.16, "learning_rate": 7.033771337189139e-06, "loss": 0.0363, "step": 39535 }, { "epoch": 12.16, "learning_rate": 7.033295495616834e-06, "loss": 0.0287, "step": 39536 }, { "epoch": 12.17, "learning_rate": 7.032819661410055e-06, "loss": 0.0172, "step": 39537 }, { "epoch": 12.17, "learning_rate": 7.032343834569987e-06, "loss": 0.0266, "step": 39538 }, { "epoch": 12.17, "learning_rate": 7.031868015097806e-06, "loss": 0.0363, "step": 39539 }, { "epoch": 12.17, "learning_rate": 7.031392202994698e-06, "loss": 0.0278, "step": 39540 }, { "epoch": 12.17, "learning_rate": 7.0309163982618444e-06, "loss": 0.037, "step": 39541 }, { "epoch": 12.17, "learning_rate": 7.030440600900425e-06, "loss": 0.0332, "step": 39542 }, { "epoch": 12.17, "learning_rate": 7.029964810911621e-06, "loss": 0.0254, "step": 39543 }, { "epoch": 12.17, "learning_rate": 7.029489028296609e-06, "loss": 0.0234, "step": 39544 }, { "epoch": 12.17, "learning_rate": 7.029013253056581e-06, "loss": 0.0166, "step": 39545 }, { "epoch": 12.17, "learning_rate": 7.02853748519271e-06, "loss": 0.0323, "step": 39546 }, { "epoch": 12.17, "learning_rate": 7.028061724706179e-06, "loss": 0.0348, "step": 39547 }, { "epoch": 12.17, "learning_rate": 7.02758597159817e-06, "loss": 0.0563, "step": 39548 }, { "epoch": 12.17, "learning_rate": 7.027110225869862e-06, "loss": 0.0311, "step": 39549 }, { "epoch": 12.17, "learning_rate": 7.026634487522439e-06, "loss": 0.0208, "step": 39550 }, { "epoch": 12.17, "learning_rate": 7.026158756557081e-06, "loss": 0.0369, "step": 39551 }, { "epoch": 12.17, "learning_rate": 7.0256830329749694e-06, "loss": 0.0307, "step": 39552 }, { "epoch": 12.17, "learning_rate": 7.025207316777287e-06, "loss": 0.045, "step": 39553 }, { "epoch": 12.17, "learning_rate": 7.0247316079652074e-06, "loss": 0.0429, "step": 39554 }, { "epoch": 12.17, "learning_rate": 7.02425590653992e-06, "loss": 0.0313, "step": 39555 }, { "epoch": 12.17, "learning_rate": 7.023780212502604e-06, "loss": 0.0316, "step": 39556 }, { "epoch": 12.17, "learning_rate": 7.02330452585444e-06, "loss": 0.0332, "step": 39557 }, { "epoch": 12.17, "learning_rate": 7.022828846596607e-06, "loss": 0.0417, "step": 39558 }, { "epoch": 12.17, "learning_rate": 7.022353174730285e-06, "loss": 0.0351, "step": 39559 }, { "epoch": 12.17, "learning_rate": 7.021877510256661e-06, "loss": 0.0404, "step": 39560 }, { "epoch": 12.17, "learning_rate": 7.0214018531769105e-06, "loss": 0.0372, "step": 39561 }, { "epoch": 12.17, "learning_rate": 7.020926203492218e-06, "loss": 0.052, "step": 39562 }, { "epoch": 12.17, "learning_rate": 7.020450561203762e-06, "loss": 0.0182, "step": 39563 }, { "epoch": 12.17, "learning_rate": 7.019974926312721e-06, "loss": 0.0353, "step": 39564 }, { "epoch": 12.17, "learning_rate": 7.01949929882028e-06, "loss": 0.0255, "step": 39565 }, { "epoch": 12.17, "learning_rate": 7.019023678727622e-06, "loss": 0.0577, "step": 39566 }, { "epoch": 12.17, "learning_rate": 7.018548066035923e-06, "loss": 0.0312, "step": 39567 }, { "epoch": 12.17, "learning_rate": 7.018072460746365e-06, "loss": 0.0327, "step": 39568 }, { "epoch": 12.17, "learning_rate": 7.017596862860127e-06, "loss": 0.0303, "step": 39569 }, { "epoch": 12.18, "learning_rate": 7.0171212723783935e-06, "loss": 0.0372, "step": 39570 }, { "epoch": 12.18, "learning_rate": 7.016645689302345e-06, "loss": 0.0226, "step": 39571 }, { "epoch": 12.18, "learning_rate": 7.016170113633161e-06, "loss": 0.0265, "step": 39572 }, { "epoch": 12.18, "learning_rate": 7.015694545372021e-06, "loss": 0.0455, "step": 39573 }, { "epoch": 12.18, "learning_rate": 7.015218984520105e-06, "loss": 0.0363, "step": 39574 }, { "epoch": 12.18, "learning_rate": 7.0147434310785975e-06, "loss": 0.0306, "step": 39575 }, { "epoch": 12.18, "learning_rate": 7.0142678850486775e-06, "loss": 0.0133, "step": 39576 }, { "epoch": 12.18, "learning_rate": 7.013792346431528e-06, "loss": 0.025, "step": 39577 }, { "epoch": 12.18, "learning_rate": 7.0133168152283245e-06, "loss": 0.0303, "step": 39578 }, { "epoch": 12.18, "learning_rate": 7.012841291440249e-06, "loss": 0.0203, "step": 39579 }, { "epoch": 12.18, "learning_rate": 7.0123657750684835e-06, "loss": 0.0381, "step": 39580 }, { "epoch": 12.18, "learning_rate": 7.0118902661142095e-06, "loss": 0.0344, "step": 39581 }, { "epoch": 12.18, "learning_rate": 7.011414764578607e-06, "loss": 0.0433, "step": 39582 }, { "epoch": 12.18, "learning_rate": 7.010939270462856e-06, "loss": 0.0236, "step": 39583 }, { "epoch": 12.18, "learning_rate": 7.010463783768133e-06, "loss": 0.0465, "step": 39584 }, { "epoch": 12.18, "learning_rate": 7.009988304495627e-06, "loss": 0.0228, "step": 39585 }, { "epoch": 12.18, "learning_rate": 7.009512832646513e-06, "loss": 0.056, "step": 39586 }, { "epoch": 12.18, "learning_rate": 7.009037368221974e-06, "loss": 0.0437, "step": 39587 }, { "epoch": 12.18, "learning_rate": 7.008561911223186e-06, "loss": 0.0267, "step": 39588 }, { "epoch": 12.18, "learning_rate": 7.008086461651329e-06, "loss": 0.0182, "step": 39589 }, { "epoch": 12.18, "learning_rate": 7.007611019507593e-06, "loss": 0.0195, "step": 39590 }, { "epoch": 12.18, "learning_rate": 7.00713558479315e-06, "loss": 0.0323, "step": 39591 }, { "epoch": 12.18, "learning_rate": 7.006660157509183e-06, "loss": 0.0335, "step": 39592 }, { "epoch": 12.18, "learning_rate": 7.0061847376568715e-06, "loss": 0.0404, "step": 39593 }, { "epoch": 12.18, "learning_rate": 7.005709325237394e-06, "loss": 0.0356, "step": 39594 }, { "epoch": 12.18, "learning_rate": 7.005233920251935e-06, "loss": 0.0278, "step": 39595 }, { "epoch": 12.18, "learning_rate": 7.004758522701673e-06, "loss": 0.0395, "step": 39596 }, { "epoch": 12.18, "learning_rate": 7.00428313258779e-06, "loss": 0.0703, "step": 39597 }, { "epoch": 12.18, "learning_rate": 7.003807749911462e-06, "loss": 0.0269, "step": 39598 }, { "epoch": 12.18, "learning_rate": 7.003332374673868e-06, "loss": 0.0199, "step": 39599 }, { "epoch": 12.18, "learning_rate": 7.002857006876198e-06, "loss": 0.0209, "step": 39600 }, { "epoch": 12.18, "learning_rate": 7.002381646519625e-06, "loss": 0.032, "step": 39601 }, { "epoch": 12.19, "learning_rate": 7.001906293605329e-06, "loss": 0.0309, "step": 39602 }, { "epoch": 12.19, "learning_rate": 7.001430948134489e-06, "loss": 0.0341, "step": 39603 }, { "epoch": 12.19, "learning_rate": 7.000955610108291e-06, "loss": 0.0458, "step": 39604 }, { "epoch": 12.19, "learning_rate": 7.00048027952791e-06, "loss": 0.0247, "step": 39605 }, { "epoch": 12.19, "learning_rate": 7.000004956394528e-06, "loss": 0.0098, "step": 39606 }, { "epoch": 12.19, "learning_rate": 6.999529640709326e-06, "loss": 0.0305, "step": 39607 }, { "epoch": 12.19, "learning_rate": 6.999054332473481e-06, "loss": 0.0181, "step": 39608 }, { "epoch": 12.19, "learning_rate": 6.998579031688176e-06, "loss": 0.023, "step": 39609 }, { "epoch": 12.19, "learning_rate": 6.9981037383545915e-06, "loss": 0.018, "step": 39610 }, { "epoch": 12.19, "learning_rate": 6.997628452473908e-06, "loss": 0.0231, "step": 39611 }, { "epoch": 12.19, "learning_rate": 6.9971531740472995e-06, "loss": 0.0161, "step": 39612 }, { "epoch": 12.19, "learning_rate": 6.996677903075948e-06, "loss": 0.0126, "step": 39613 }, { "epoch": 12.19, "learning_rate": 6.9962026395610416e-06, "loss": 0.0436, "step": 39614 }, { "epoch": 12.19, "learning_rate": 6.995727383503752e-06, "loss": 0.0478, "step": 39615 }, { "epoch": 12.19, "learning_rate": 6.995252134905261e-06, "loss": 0.0314, "step": 39616 }, { "epoch": 12.19, "learning_rate": 6.994776893766749e-06, "loss": 0.0347, "step": 39617 }, { "epoch": 12.19, "learning_rate": 6.994301660089394e-06, "loss": 0.0144, "step": 39618 }, { "epoch": 12.19, "learning_rate": 6.99382643387438e-06, "loss": 0.0462, "step": 39619 }, { "epoch": 12.19, "learning_rate": 6.993351215122883e-06, "loss": 0.0102, "step": 39620 }, { "epoch": 12.19, "learning_rate": 6.992876003836089e-06, "loss": 0.047, "step": 39621 }, { "epoch": 12.19, "learning_rate": 6.992400800015169e-06, "loss": 0.0241, "step": 39622 }, { "epoch": 12.19, "learning_rate": 6.991925603661304e-06, "loss": 0.0159, "step": 39623 }, { "epoch": 12.19, "learning_rate": 6.991450414775682e-06, "loss": 0.0311, "step": 39624 }, { "epoch": 12.19, "learning_rate": 6.9909752333594765e-06, "loss": 0.0326, "step": 39625 }, { "epoch": 12.19, "learning_rate": 6.990500059413868e-06, "loss": 0.0247, "step": 39626 }, { "epoch": 12.19, "learning_rate": 6.990024892940036e-06, "loss": 0.0238, "step": 39627 }, { "epoch": 12.19, "learning_rate": 6.989549733939158e-06, "loss": 0.0422, "step": 39628 }, { "epoch": 12.19, "learning_rate": 6.98907458241242e-06, "loss": 0.0191, "step": 39629 }, { "epoch": 12.19, "learning_rate": 6.988599438360998e-06, "loss": 0.0351, "step": 39630 }, { "epoch": 12.19, "learning_rate": 6.988124301786071e-06, "loss": 0.0336, "step": 39631 }, { "epoch": 12.19, "learning_rate": 6.9876491726888215e-06, "loss": 0.0365, "step": 39632 }, { "epoch": 12.19, "learning_rate": 6.987174051070422e-06, "loss": 0.0293, "step": 39633 }, { "epoch": 12.19, "learning_rate": 6.986698936932059e-06, "loss": 0.0368, "step": 39634 }, { "epoch": 12.2, "learning_rate": 6.986223830274914e-06, "loss": 0.0482, "step": 39635 }, { "epoch": 12.2, "learning_rate": 6.9857487311001605e-06, "loss": 0.0405, "step": 39636 }, { "epoch": 12.2, "learning_rate": 6.98527363940898e-06, "loss": 0.0452, "step": 39637 }, { "epoch": 12.2, "learning_rate": 6.98479855520255e-06, "loss": 0.0426, "step": 39638 }, { "epoch": 12.2, "learning_rate": 6.984323478482056e-06, "loss": 0.0143, "step": 39639 }, { "epoch": 12.2, "learning_rate": 6.983848409248672e-06, "loss": 0.0286, "step": 39640 }, { "epoch": 12.2, "learning_rate": 6.98337334750358e-06, "loss": 0.023, "step": 39641 }, { "epoch": 12.2, "learning_rate": 6.98289829324796e-06, "loss": 0.0445, "step": 39642 }, { "epoch": 12.2, "learning_rate": 6.982423246482983e-06, "loss": 0.0255, "step": 39643 }, { "epoch": 12.2, "learning_rate": 6.981948207209842e-06, "loss": 0.0293, "step": 39644 }, { "epoch": 12.2, "learning_rate": 6.9814731754297114e-06, "loss": 0.0281, "step": 39645 }, { "epoch": 12.2, "learning_rate": 6.980998151143766e-06, "loss": 0.032, "step": 39646 }, { "epoch": 12.2, "learning_rate": 6.98052313435319e-06, "loss": 0.0592, "step": 39647 }, { "epoch": 12.2, "learning_rate": 6.980048125059158e-06, "loss": 0.0333, "step": 39648 }, { "epoch": 12.2, "learning_rate": 6.979573123262854e-06, "loss": 0.0317, "step": 39649 }, { "epoch": 12.2, "learning_rate": 6.9790981289654556e-06, "loss": 0.0567, "step": 39650 }, { "epoch": 12.2, "learning_rate": 6.9786231421681435e-06, "loss": 0.0322, "step": 39651 }, { "epoch": 12.2, "learning_rate": 6.978148162872095e-06, "loss": 0.0235, "step": 39652 }, { "epoch": 12.2, "learning_rate": 6.977673191078487e-06, "loss": 0.0306, "step": 39653 }, { "epoch": 12.2, "learning_rate": 6.977198226788504e-06, "loss": 0.0318, "step": 39654 }, { "epoch": 12.2, "learning_rate": 6.976723270003324e-06, "loss": 0.0319, "step": 39655 }, { "epoch": 12.2, "learning_rate": 6.976248320724126e-06, "loss": 0.0278, "step": 39656 }, { "epoch": 12.2, "learning_rate": 6.975773378952086e-06, "loss": 0.0181, "step": 39657 }, { "epoch": 12.2, "learning_rate": 6.975298444688385e-06, "loss": 0.0337, "step": 39658 }, { "epoch": 12.2, "learning_rate": 6.974823517934204e-06, "loss": 0.047, "step": 39659 }, { "epoch": 12.2, "learning_rate": 6.974348598690719e-06, "loss": 0.0372, "step": 39660 }, { "epoch": 12.2, "learning_rate": 6.973873686959111e-06, "loss": 0.0291, "step": 39661 }, { "epoch": 12.2, "learning_rate": 6.973398782740561e-06, "loss": 0.0568, "step": 39662 }, { "epoch": 12.2, "learning_rate": 6.972923886036242e-06, "loss": 0.0429, "step": 39663 }, { "epoch": 12.2, "learning_rate": 6.972448996847341e-06, "loss": 0.024, "step": 39664 }, { "epoch": 12.2, "learning_rate": 6.971974115175031e-06, "loss": 0.0304, "step": 39665 }, { "epoch": 12.2, "learning_rate": 6.971499241020495e-06, "loss": 0.0612, "step": 39666 }, { "epoch": 12.21, "learning_rate": 6.971024374384908e-06, "loss": 0.0201, "step": 39667 }, { "epoch": 12.21, "learning_rate": 6.970549515269447e-06, "loss": 0.0168, "step": 39668 }, { "epoch": 12.21, "learning_rate": 6.970074663675302e-06, "loss": 0.0277, "step": 39669 }, { "epoch": 12.21, "learning_rate": 6.969599819603641e-06, "loss": 0.0376, "step": 39670 }, { "epoch": 12.21, "learning_rate": 6.969124983055649e-06, "loss": 0.0316, "step": 39671 }, { "epoch": 12.21, "learning_rate": 6.968650154032499e-06, "loss": 0.0362, "step": 39672 }, { "epoch": 12.21, "learning_rate": 6.968175332535374e-06, "loss": 0.0301, "step": 39673 }, { "epoch": 12.21, "learning_rate": 6.967700518565454e-06, "loss": 0.0252, "step": 39674 }, { "epoch": 12.21, "learning_rate": 6.967225712123915e-06, "loss": 0.0521, "step": 39675 }, { "epoch": 12.21, "learning_rate": 6.966750913211941e-06, "loss": 0.0332, "step": 39676 }, { "epoch": 12.21, "learning_rate": 6.966276121830704e-06, "loss": 0.0311, "step": 39677 }, { "epoch": 12.21, "learning_rate": 6.96580133798138e-06, "loss": 0.0251, "step": 39678 }, { "epoch": 12.21, "learning_rate": 6.96532656166516e-06, "loss": 0.034, "step": 39679 }, { "epoch": 12.21, "learning_rate": 6.964851792883216e-06, "loss": 0.0514, "step": 39680 }, { "epoch": 12.21, "learning_rate": 6.964377031636724e-06, "loss": 0.0415, "step": 39681 }, { "epoch": 12.21, "learning_rate": 6.9639022779268675e-06, "loss": 0.0352, "step": 39682 }, { "epoch": 12.21, "learning_rate": 6.963427531754819e-06, "loss": 0.0241, "step": 39683 }, { "epoch": 12.21, "learning_rate": 6.962952793121765e-06, "loss": 0.033, "step": 39684 }, { "epoch": 12.21, "learning_rate": 6.9624780620288805e-06, "loss": 0.0355, "step": 39685 }, { "epoch": 12.21, "learning_rate": 6.962003338477343e-06, "loss": 0.0275, "step": 39686 }, { "epoch": 12.21, "learning_rate": 6.9615286224683356e-06, "loss": 0.0428, "step": 39687 }, { "epoch": 12.21, "learning_rate": 6.961053914003026e-06, "loss": 0.0169, "step": 39688 }, { "epoch": 12.21, "learning_rate": 6.960579213082606e-06, "loss": 0.0348, "step": 39689 }, { "epoch": 12.21, "learning_rate": 6.960104519708249e-06, "loss": 0.0154, "step": 39690 }, { "epoch": 12.21, "learning_rate": 6.959629833881131e-06, "loss": 0.0554, "step": 39691 }, { "epoch": 12.21, "learning_rate": 6.959155155602433e-06, "loss": 0.0142, "step": 39692 }, { "epoch": 12.21, "learning_rate": 6.95868048487333e-06, "loss": 0.0329, "step": 39693 }, { "epoch": 12.21, "learning_rate": 6.958205821695006e-06, "loss": 0.023, "step": 39694 }, { "epoch": 12.21, "learning_rate": 6.957731166068638e-06, "loss": 0.0394, "step": 39695 }, { "epoch": 12.21, "learning_rate": 6.957256517995401e-06, "loss": 0.0369, "step": 39696 }, { "epoch": 12.21, "learning_rate": 6.956781877476479e-06, "loss": 0.0404, "step": 39697 }, { "epoch": 12.21, "learning_rate": 6.956307244513043e-06, "loss": 0.0269, "step": 39698 }, { "epoch": 12.21, "learning_rate": 6.9558326191062775e-06, "loss": 0.018, "step": 39699 }, { "epoch": 12.22, "learning_rate": 6.955358001257362e-06, "loss": 0.066, "step": 39700 }, { "epoch": 12.22, "learning_rate": 6.9548833909674685e-06, "loss": 0.0397, "step": 39701 }, { "epoch": 12.22, "learning_rate": 6.954408788237779e-06, "loss": 0.0305, "step": 39702 }, { "epoch": 12.22, "learning_rate": 6.953934193069469e-06, "loss": 0.0416, "step": 39703 }, { "epoch": 12.22, "learning_rate": 6.953459605463721e-06, "loss": 0.0331, "step": 39704 }, { "epoch": 12.22, "learning_rate": 6.9529850254217135e-06, "loss": 0.0224, "step": 39705 }, { "epoch": 12.22, "learning_rate": 6.952510452944621e-06, "loss": 0.0336, "step": 39706 }, { "epoch": 12.22, "learning_rate": 6.952035888033625e-06, "loss": 0.0227, "step": 39707 }, { "epoch": 12.22, "learning_rate": 6.951561330689899e-06, "loss": 0.0415, "step": 39708 }, { "epoch": 12.22, "learning_rate": 6.951086780914626e-06, "loss": 0.0367, "step": 39709 }, { "epoch": 12.22, "learning_rate": 6.950612238708983e-06, "loss": 0.0328, "step": 39710 }, { "epoch": 12.22, "learning_rate": 6.95013770407415e-06, "loss": 0.0207, "step": 39711 }, { "epoch": 12.22, "learning_rate": 6.949663177011301e-06, "loss": 0.0389, "step": 39712 }, { "epoch": 12.22, "learning_rate": 6.9491886575216115e-06, "loss": 0.0257, "step": 39713 }, { "epoch": 12.22, "learning_rate": 6.94871414560627e-06, "loss": 0.0425, "step": 39714 }, { "epoch": 12.22, "learning_rate": 6.948239641266448e-06, "loss": 0.0368, "step": 39715 }, { "epoch": 12.22, "learning_rate": 6.947765144503323e-06, "loss": 0.0267, "step": 39716 }, { "epoch": 12.22, "learning_rate": 6.9472906553180745e-06, "loss": 0.036, "step": 39717 }, { "epoch": 12.22, "learning_rate": 6.946816173711878e-06, "loss": 0.033, "step": 39718 }, { "epoch": 12.22, "learning_rate": 6.946341699685917e-06, "loss": 0.0287, "step": 39719 }, { "epoch": 12.22, "learning_rate": 6.9458672332413645e-06, "loss": 0.0191, "step": 39720 }, { "epoch": 12.22, "learning_rate": 6.945392774379403e-06, "loss": 0.0156, "step": 39721 }, { "epoch": 12.22, "learning_rate": 6.944918323101206e-06, "loss": 0.037, "step": 39722 }, { "epoch": 12.22, "learning_rate": 6.944443879407949e-06, "loss": 0.0388, "step": 39723 }, { "epoch": 12.22, "learning_rate": 6.94396944330082e-06, "loss": 0.0154, "step": 39724 }, { "epoch": 12.22, "learning_rate": 6.94349501478099e-06, "loss": 0.0263, "step": 39725 }, { "epoch": 12.22, "learning_rate": 6.943020593849637e-06, "loss": 0.0468, "step": 39726 }, { "epoch": 12.22, "learning_rate": 6.94254618050794e-06, "loss": 0.0113, "step": 39727 }, { "epoch": 12.22, "learning_rate": 6.942071774757074e-06, "loss": 0.0128, "step": 39728 }, { "epoch": 12.22, "learning_rate": 6.941597376598222e-06, "loss": 0.0194, "step": 39729 }, { "epoch": 12.22, "learning_rate": 6.941122986032559e-06, "loss": 0.0432, "step": 39730 }, { "epoch": 12.22, "learning_rate": 6.940648603061263e-06, "loss": 0.0421, "step": 39731 }, { "epoch": 12.23, "learning_rate": 6.940174227685515e-06, "loss": 0.0212, "step": 39732 }, { "epoch": 12.23, "learning_rate": 6.939699859906481e-06, "loss": 0.0487, "step": 39733 }, { "epoch": 12.23, "learning_rate": 6.939225499725355e-06, "loss": 0.0179, "step": 39734 }, { "epoch": 12.23, "learning_rate": 6.938751147143303e-06, "loss": 0.0369, "step": 39735 }, { "epoch": 12.23, "learning_rate": 6.9382768021615085e-06, "loss": 0.0251, "step": 39736 }, { "epoch": 12.23, "learning_rate": 6.937802464781147e-06, "loss": 0.0284, "step": 39737 }, { "epoch": 12.23, "learning_rate": 6.937328135003394e-06, "loss": 0.0413, "step": 39738 }, { "epoch": 12.23, "learning_rate": 6.936853812829432e-06, "loss": 0.0311, "step": 39739 }, { "epoch": 12.23, "learning_rate": 6.936379498260436e-06, "loss": 0.0565, "step": 39740 }, { "epoch": 12.23, "learning_rate": 6.935905191297584e-06, "loss": 0.0457, "step": 39741 }, { "epoch": 12.23, "learning_rate": 6.935430891942054e-06, "loss": 0.0341, "step": 39742 }, { "epoch": 12.23, "learning_rate": 6.934956600195019e-06, "loss": 0.0155, "step": 39743 }, { "epoch": 12.23, "learning_rate": 6.934482316057663e-06, "loss": 0.0374, "step": 39744 }, { "epoch": 12.23, "learning_rate": 6.934008039531163e-06, "loss": 0.0107, "step": 39745 }, { "epoch": 12.23, "learning_rate": 6.933533770616694e-06, "loss": 0.0307, "step": 39746 }, { "epoch": 12.23, "learning_rate": 6.933059509315432e-06, "loss": 0.0274, "step": 39747 }, { "epoch": 12.23, "learning_rate": 6.932585255628556e-06, "loss": 0.036, "step": 39748 }, { "epoch": 12.23, "learning_rate": 6.932111009557245e-06, "loss": 0.0218, "step": 39749 }, { "epoch": 12.23, "learning_rate": 6.931636771102674e-06, "loss": 0.0409, "step": 39750 }, { "epoch": 12.23, "learning_rate": 6.931162540266024e-06, "loss": 0.0191, "step": 39751 }, { "epoch": 12.23, "learning_rate": 6.930688317048469e-06, "loss": 0.0195, "step": 39752 }, { "epoch": 12.23, "learning_rate": 6.930214101451186e-06, "loss": 0.0347, "step": 39753 }, { "epoch": 12.23, "learning_rate": 6.929739893475355e-06, "loss": 0.0256, "step": 39754 }, { "epoch": 12.23, "learning_rate": 6.9292656931221555e-06, "loss": 0.0283, "step": 39755 }, { "epoch": 12.23, "learning_rate": 6.928791500392758e-06, "loss": 0.0382, "step": 39756 }, { "epoch": 12.23, "learning_rate": 6.928317315288344e-06, "loss": 0.0281, "step": 39757 }, { "epoch": 12.23, "learning_rate": 6.927843137810087e-06, "loss": 0.0291, "step": 39758 }, { "epoch": 12.23, "learning_rate": 6.9273689679591705e-06, "loss": 0.0328, "step": 39759 }, { "epoch": 12.23, "learning_rate": 6.926894805736768e-06, "loss": 0.0482, "step": 39760 }, { "epoch": 12.23, "learning_rate": 6.9264206511440565e-06, "loss": 0.05, "step": 39761 }, { "epoch": 12.23, "learning_rate": 6.925946504182215e-06, "loss": 0.0124, "step": 39762 }, { "epoch": 12.23, "learning_rate": 6.925472364852417e-06, "loss": 0.0427, "step": 39763 }, { "epoch": 12.23, "learning_rate": 6.924998233155844e-06, "loss": 0.0152, "step": 39764 }, { "epoch": 12.24, "learning_rate": 6.924524109093673e-06, "loss": 0.0441, "step": 39765 }, { "epoch": 12.24, "learning_rate": 6.92404999266708e-06, "loss": 0.0143, "step": 39766 }, { "epoch": 12.24, "learning_rate": 6.923575883877241e-06, "loss": 0.066, "step": 39767 }, { "epoch": 12.24, "learning_rate": 6.923101782725328e-06, "loss": 0.0516, "step": 39768 }, { "epoch": 12.24, "learning_rate": 6.922627689212531e-06, "loss": 0.0362, "step": 39769 }, { "epoch": 12.24, "learning_rate": 6.922153603340016e-06, "loss": 0.0225, "step": 39770 }, { "epoch": 12.24, "learning_rate": 6.921679525108966e-06, "loss": 0.0381, "step": 39771 }, { "epoch": 12.24, "learning_rate": 6.921205454520554e-06, "loss": 0.0406, "step": 39772 }, { "epoch": 12.24, "learning_rate": 6.920731391575959e-06, "loss": 0.0346, "step": 39773 }, { "epoch": 12.24, "learning_rate": 6.920257336276358e-06, "loss": 0.0806, "step": 39774 }, { "epoch": 12.24, "learning_rate": 6.919783288622928e-06, "loss": 0.0286, "step": 39775 }, { "epoch": 12.24, "learning_rate": 6.919309248616848e-06, "loss": 0.0316, "step": 39776 }, { "epoch": 12.24, "learning_rate": 6.918835216259291e-06, "loss": 0.0519, "step": 39777 }, { "epoch": 12.24, "learning_rate": 6.91836119155143e-06, "loss": 0.0322, "step": 39778 }, { "epoch": 12.24, "learning_rate": 6.917887174494455e-06, "loss": 0.0263, "step": 39779 }, { "epoch": 12.24, "learning_rate": 6.917413165089533e-06, "loss": 0.0319, "step": 39780 }, { "epoch": 12.24, "learning_rate": 6.916939163337844e-06, "loss": 0.0328, "step": 39781 }, { "epoch": 12.24, "learning_rate": 6.916465169240562e-06, "loss": 0.0379, "step": 39782 }, { "epoch": 12.24, "learning_rate": 6.915991182798865e-06, "loss": 0.0339, "step": 39783 }, { "epoch": 12.24, "learning_rate": 6.9155172040139306e-06, "loss": 0.0178, "step": 39784 }, { "epoch": 12.24, "learning_rate": 6.9150432328869375e-06, "loss": 0.03, "step": 39785 }, { "epoch": 12.24, "learning_rate": 6.91456926941906e-06, "loss": 0.0339, "step": 39786 }, { "epoch": 12.24, "learning_rate": 6.914095313611476e-06, "loss": 0.0376, "step": 39787 }, { "epoch": 12.24, "learning_rate": 6.913621365465357e-06, "loss": 0.0383, "step": 39788 }, { "epoch": 12.24, "learning_rate": 6.913147424981887e-06, "loss": 0.0402, "step": 39789 }, { "epoch": 12.24, "learning_rate": 6.912673492162241e-06, "loss": 0.0538, "step": 39790 }, { "epoch": 12.24, "learning_rate": 6.912199567007595e-06, "loss": 0.0304, "step": 39791 }, { "epoch": 12.24, "learning_rate": 6.911725649519123e-06, "loss": 0.0242, "step": 39792 }, { "epoch": 12.24, "learning_rate": 6.9112517396980015e-06, "loss": 0.0209, "step": 39793 }, { "epoch": 12.24, "learning_rate": 6.910777837545413e-06, "loss": 0.0264, "step": 39794 }, { "epoch": 12.24, "learning_rate": 6.910303943062528e-06, "loss": 0.0221, "step": 39795 }, { "epoch": 12.24, "learning_rate": 6.909830056250527e-06, "loss": 0.0246, "step": 39796 }, { "epoch": 12.25, "learning_rate": 6.909356177110587e-06, "loss": 0.0292, "step": 39797 }, { "epoch": 12.25, "learning_rate": 6.908882305643876e-06, "loss": 0.0355, "step": 39798 }, { "epoch": 12.25, "learning_rate": 6.908408441851579e-06, "loss": 0.0156, "step": 39799 }, { "epoch": 12.25, "learning_rate": 6.907934585734875e-06, "loss": 0.015, "step": 39800 }, { "epoch": 12.25, "learning_rate": 6.907460737294933e-06, "loss": 0.0388, "step": 39801 }, { "epoch": 12.25, "learning_rate": 6.906986896532932e-06, "loss": 0.016, "step": 39802 }, { "epoch": 12.25, "learning_rate": 6.906513063450047e-06, "loss": 0.0487, "step": 39803 }, { "epoch": 12.25, "learning_rate": 6.906039238047457e-06, "loss": 0.0247, "step": 39804 }, { "epoch": 12.25, "learning_rate": 6.905565420326339e-06, "loss": 0.0415, "step": 39805 }, { "epoch": 12.25, "learning_rate": 6.9050916102878676e-06, "loss": 0.042, "step": 39806 }, { "epoch": 12.25, "learning_rate": 6.904617807933218e-06, "loss": 0.0119, "step": 39807 }, { "epoch": 12.25, "learning_rate": 6.904144013263567e-06, "loss": 0.0436, "step": 39808 }, { "epoch": 12.25, "learning_rate": 6.903670226280095e-06, "loss": 0.0257, "step": 39809 }, { "epoch": 12.25, "learning_rate": 6.903196446983973e-06, "loss": 0.009, "step": 39810 }, { "epoch": 12.25, "learning_rate": 6.902722675376382e-06, "loss": 0.0486, "step": 39811 }, { "epoch": 12.25, "learning_rate": 6.902248911458494e-06, "loss": 0.0534, "step": 39812 }, { "epoch": 12.25, "learning_rate": 6.901775155231485e-06, "loss": 0.0313, "step": 39813 }, { "epoch": 12.25, "learning_rate": 6.901301406696534e-06, "loss": 0.0294, "step": 39814 }, { "epoch": 12.25, "learning_rate": 6.900827665854817e-06, "loss": 0.0226, "step": 39815 }, { "epoch": 12.25, "learning_rate": 6.900353932707509e-06, "loss": 0.0249, "step": 39816 }, { "epoch": 12.25, "learning_rate": 6.899880207255787e-06, "loss": 0.0377, "step": 39817 }, { "epoch": 12.25, "learning_rate": 6.899406489500824e-06, "loss": 0.0151, "step": 39818 }, { "epoch": 12.25, "learning_rate": 6.898932779443801e-06, "loss": 0.0263, "step": 39819 }, { "epoch": 12.25, "learning_rate": 6.898459077085893e-06, "loss": 0.018, "step": 39820 }, { "epoch": 12.25, "learning_rate": 6.897985382428276e-06, "loss": 0.025, "step": 39821 }, { "epoch": 12.25, "learning_rate": 6.897511695472124e-06, "loss": 0.0312, "step": 39822 }, { "epoch": 12.25, "learning_rate": 6.8970380162186065e-06, "loss": 0.0199, "step": 39823 }, { "epoch": 12.25, "learning_rate": 6.896564344668917e-06, "loss": 0.0226, "step": 39824 }, { "epoch": 12.25, "learning_rate": 6.896090680824218e-06, "loss": 0.0156, "step": 39825 }, { "epoch": 12.25, "learning_rate": 6.895617024685689e-06, "loss": 0.0531, "step": 39826 }, { "epoch": 12.25, "learning_rate": 6.895143376254507e-06, "loss": 0.0362, "step": 39827 }, { "epoch": 12.25, "learning_rate": 6.894669735531844e-06, "loss": 0.0456, "step": 39828 }, { "epoch": 12.25, "learning_rate": 6.894196102518881e-06, "loss": 0.0266, "step": 39829 }, { "epoch": 12.26, "learning_rate": 6.893722477216792e-06, "loss": 0.024, "step": 39830 }, { "epoch": 12.26, "learning_rate": 6.893248859626753e-06, "loss": 0.0375, "step": 39831 }, { "epoch": 12.26, "learning_rate": 6.892775249749942e-06, "loss": 0.0529, "step": 39832 }, { "epoch": 12.26, "learning_rate": 6.892301647587525e-06, "loss": 0.0325, "step": 39833 }, { "epoch": 12.26, "learning_rate": 6.891828053140691e-06, "loss": 0.0374, "step": 39834 }, { "epoch": 12.26, "learning_rate": 6.891354466410609e-06, "loss": 0.0327, "step": 39835 }, { "epoch": 12.26, "learning_rate": 6.890880887398455e-06, "loss": 0.0263, "step": 39836 }, { "epoch": 12.26, "learning_rate": 6.8904073161054065e-06, "loss": 0.0385, "step": 39837 }, { "epoch": 12.26, "learning_rate": 6.889933752532636e-06, "loss": 0.0183, "step": 39838 }, { "epoch": 12.26, "learning_rate": 6.8894601966813235e-06, "loss": 0.0276, "step": 39839 }, { "epoch": 12.26, "learning_rate": 6.888986648552643e-06, "loss": 0.0367, "step": 39840 }, { "epoch": 12.26, "learning_rate": 6.888513108147768e-06, "loss": 0.0433, "step": 39841 }, { "epoch": 12.26, "learning_rate": 6.888039575467881e-06, "loss": 0.0401, "step": 39842 }, { "epoch": 12.26, "learning_rate": 6.887566050514145e-06, "loss": 0.0182, "step": 39843 }, { "epoch": 12.26, "learning_rate": 6.887092533287748e-06, "loss": 0.058, "step": 39844 }, { "epoch": 12.26, "learning_rate": 6.886619023789862e-06, "loss": 0.0439, "step": 39845 }, { "epoch": 12.26, "learning_rate": 6.886145522021661e-06, "loss": 0.0279, "step": 39846 }, { "epoch": 12.26, "learning_rate": 6.885672027984321e-06, "loss": 0.0303, "step": 39847 }, { "epoch": 12.26, "learning_rate": 6.885198541679016e-06, "loss": 0.0109, "step": 39848 }, { "epoch": 12.26, "learning_rate": 6.8847250631069255e-06, "loss": 0.0317, "step": 39849 }, { "epoch": 12.26, "learning_rate": 6.884251592269221e-06, "loss": 0.0307, "step": 39850 }, { "epoch": 12.26, "learning_rate": 6.883778129167081e-06, "loss": 0.0752, "step": 39851 }, { "epoch": 12.26, "learning_rate": 6.88330467380168e-06, "loss": 0.0436, "step": 39852 }, { "epoch": 12.26, "learning_rate": 6.8828312261741914e-06, "loss": 0.0241, "step": 39853 }, { "epoch": 12.26, "learning_rate": 6.8823577862857935e-06, "loss": 0.0453, "step": 39854 }, { "epoch": 12.26, "learning_rate": 6.881884354137663e-06, "loss": 0.0422, "step": 39855 }, { "epoch": 12.26, "learning_rate": 6.881410929730972e-06, "loss": 0.0287, "step": 39856 }, { "epoch": 12.26, "learning_rate": 6.880937513066897e-06, "loss": 0.0365, "step": 39857 }, { "epoch": 12.26, "learning_rate": 6.880464104146609e-06, "loss": 0.0159, "step": 39858 }, { "epoch": 12.26, "learning_rate": 6.879990702971292e-06, "loss": 0.0301, "step": 39859 }, { "epoch": 12.26, "learning_rate": 6.879517309542116e-06, "loss": 0.0489, "step": 39860 }, { "epoch": 12.26, "learning_rate": 6.8790439238602576e-06, "loss": 0.028, "step": 39861 }, { "epoch": 12.27, "learning_rate": 6.878570545926893e-06, "loss": 0.033, "step": 39862 }, { "epoch": 12.27, "learning_rate": 6.878097175743192e-06, "loss": 0.0355, "step": 39863 }, { "epoch": 12.27, "learning_rate": 6.877623813310337e-06, "loss": 0.031, "step": 39864 }, { "epoch": 12.27, "learning_rate": 6.8771504586295e-06, "loss": 0.0326, "step": 39865 }, { "epoch": 12.27, "learning_rate": 6.8766771117018585e-06, "loss": 0.0475, "step": 39866 }, { "epoch": 12.27, "learning_rate": 6.876203772528583e-06, "loss": 0.0443, "step": 39867 }, { "epoch": 12.27, "learning_rate": 6.8757304411108485e-06, "loss": 0.0161, "step": 39868 }, { "epoch": 12.27, "learning_rate": 6.875257117449839e-06, "loss": 0.0293, "step": 39869 }, { "epoch": 12.27, "learning_rate": 6.8747838015467215e-06, "loss": 0.0373, "step": 39870 }, { "epoch": 12.27, "learning_rate": 6.874310493402674e-06, "loss": 0.064, "step": 39871 }, { "epoch": 12.27, "learning_rate": 6.87383719301887e-06, "loss": 0.0294, "step": 39872 }, { "epoch": 12.27, "learning_rate": 6.8733639003964835e-06, "loss": 0.0315, "step": 39873 }, { "epoch": 12.27, "learning_rate": 6.872890615536694e-06, "loss": 0.0351, "step": 39874 }, { "epoch": 12.27, "learning_rate": 6.872417338440674e-06, "loss": 0.0598, "step": 39875 }, { "epoch": 12.27, "learning_rate": 6.871944069109599e-06, "loss": 0.0337, "step": 39876 }, { "epoch": 12.27, "learning_rate": 6.871470807544644e-06, "loss": 0.0296, "step": 39877 }, { "epoch": 12.27, "learning_rate": 6.870997553746979e-06, "loss": 0.0393, "step": 39878 }, { "epoch": 12.27, "learning_rate": 6.8705243077177875e-06, "loss": 0.0421, "step": 39879 }, { "epoch": 12.27, "learning_rate": 6.87005106945824e-06, "loss": 0.0083, "step": 39880 }, { "epoch": 12.27, "learning_rate": 6.869577838969512e-06, "loss": 0.0312, "step": 39881 }, { "epoch": 12.27, "learning_rate": 6.869104616252777e-06, "loss": 0.0341, "step": 39882 }, { "epoch": 12.27, "learning_rate": 6.8686314013092116e-06, "loss": 0.042, "step": 39883 }, { "epoch": 12.27, "learning_rate": 6.868158194139991e-06, "loss": 0.0134, "step": 39884 }, { "epoch": 12.27, "learning_rate": 6.8676849947462884e-06, "loss": 0.0221, "step": 39885 }, { "epoch": 12.27, "learning_rate": 6.867211803129281e-06, "loss": 0.0293, "step": 39886 }, { "epoch": 12.27, "learning_rate": 6.866738619290143e-06, "loss": 0.0289, "step": 39887 }, { "epoch": 12.27, "learning_rate": 6.866265443230042e-06, "loss": 0.0311, "step": 39888 }, { "epoch": 12.27, "learning_rate": 6.865792274950165e-06, "loss": 0.0235, "step": 39889 }, { "epoch": 12.27, "learning_rate": 6.865319114451682e-06, "loss": 0.0364, "step": 39890 }, { "epoch": 12.27, "learning_rate": 6.864845961735764e-06, "loss": 0.0151, "step": 39891 }, { "epoch": 12.27, "learning_rate": 6.864372816803588e-06, "loss": 0.0571, "step": 39892 }, { "epoch": 12.27, "learning_rate": 6.8638996796563275e-06, "loss": 0.025, "step": 39893 }, { "epoch": 12.27, "learning_rate": 6.863426550295161e-06, "loss": 0.0234, "step": 39894 }, { "epoch": 12.28, "learning_rate": 6.86295342872126e-06, "loss": 0.0429, "step": 39895 }, { "epoch": 12.28, "learning_rate": 6.8624803149358e-06, "loss": 0.0313, "step": 39896 }, { "epoch": 12.28, "learning_rate": 6.862007208939958e-06, "loss": 0.0275, "step": 39897 }, { "epoch": 12.28, "learning_rate": 6.8615341107349e-06, "loss": 0.0406, "step": 39898 }, { "epoch": 12.28, "learning_rate": 6.861061020321811e-06, "loss": 0.0471, "step": 39899 }, { "epoch": 12.28, "learning_rate": 6.860587937701862e-06, "loss": 0.0162, "step": 39900 }, { "epoch": 12.28, "learning_rate": 6.860114862876226e-06, "loss": 0.0247, "step": 39901 }, { "epoch": 12.28, "learning_rate": 6.859641795846078e-06, "loss": 0.0296, "step": 39902 }, { "epoch": 12.28, "learning_rate": 6.859168736612592e-06, "loss": 0.0373, "step": 39903 }, { "epoch": 12.28, "learning_rate": 6.858695685176942e-06, "loss": 0.0278, "step": 39904 }, { "epoch": 12.28, "learning_rate": 6.858222641540306e-06, "loss": 0.0462, "step": 39905 }, { "epoch": 12.28, "learning_rate": 6.857749605703856e-06, "loss": 0.0563, "step": 39906 }, { "epoch": 12.28, "learning_rate": 6.857276577668768e-06, "loss": 0.0402, "step": 39907 }, { "epoch": 12.28, "learning_rate": 6.856803557436211e-06, "loss": 0.0285, "step": 39908 }, { "epoch": 12.28, "learning_rate": 6.856330545007366e-06, "loss": 0.037, "step": 39909 }, { "epoch": 12.28, "learning_rate": 6.855857540383404e-06, "loss": 0.0277, "step": 39910 }, { "epoch": 12.28, "learning_rate": 6.855384543565503e-06, "loss": 0.0184, "step": 39911 }, { "epoch": 12.28, "learning_rate": 6.854911554554832e-06, "loss": 0.0207, "step": 39912 }, { "epoch": 12.28, "learning_rate": 6.8544385733525665e-06, "loss": 0.0326, "step": 39913 }, { "epoch": 12.28, "learning_rate": 6.853965599959883e-06, "loss": 0.0464, "step": 39914 }, { "epoch": 12.28, "learning_rate": 6.853492634377956e-06, "loss": 0.0328, "step": 39915 }, { "epoch": 12.28, "learning_rate": 6.853019676607957e-06, "loss": 0.0339, "step": 39916 }, { "epoch": 12.28, "learning_rate": 6.852546726651063e-06, "loss": 0.0128, "step": 39917 }, { "epoch": 12.28, "learning_rate": 6.852073784508444e-06, "loss": 0.0331, "step": 39918 }, { "epoch": 12.28, "learning_rate": 6.8516008501812795e-06, "loss": 0.0232, "step": 39919 }, { "epoch": 12.28, "learning_rate": 6.851127923670742e-06, "loss": 0.0295, "step": 39920 }, { "epoch": 12.28, "learning_rate": 6.850655004978006e-06, "loss": 0.0403, "step": 39921 }, { "epoch": 12.28, "learning_rate": 6.850182094104243e-06, "loss": 0.0393, "step": 39922 }, { "epoch": 12.28, "learning_rate": 6.8497091910506244e-06, "loss": 0.0295, "step": 39923 }, { "epoch": 12.28, "learning_rate": 6.849236295818334e-06, "loss": 0.0392, "step": 39924 }, { "epoch": 12.28, "learning_rate": 6.848763408408541e-06, "loss": 0.0338, "step": 39925 }, { "epoch": 12.28, "learning_rate": 6.848290528822417e-06, "loss": 0.0321, "step": 39926 }, { "epoch": 12.29, "learning_rate": 6.847817657061138e-06, "loss": 0.0271, "step": 39927 }, { "epoch": 12.29, "learning_rate": 6.847344793125875e-06, "loss": 0.0479, "step": 39928 }, { "epoch": 12.29, "learning_rate": 6.846871937017809e-06, "loss": 0.0543, "step": 39929 }, { "epoch": 12.29, "learning_rate": 6.846399088738108e-06, "loss": 0.0277, "step": 39930 }, { "epoch": 12.29, "learning_rate": 6.845926248287948e-06, "loss": 0.0364, "step": 39931 }, { "epoch": 12.29, "learning_rate": 6.845453415668507e-06, "loss": 0.0501, "step": 39932 }, { "epoch": 12.29, "learning_rate": 6.844980590880947e-06, "loss": 0.0343, "step": 39933 }, { "epoch": 12.29, "learning_rate": 6.844507773926456e-06, "loss": 0.0272, "step": 39934 }, { "epoch": 12.29, "learning_rate": 6.8440349648061986e-06, "loss": 0.0276, "step": 39935 }, { "epoch": 12.29, "learning_rate": 6.843562163521353e-06, "loss": 0.0235, "step": 39936 }, { "epoch": 12.29, "learning_rate": 6.843089370073091e-06, "loss": 0.0173, "step": 39937 }, { "epoch": 12.29, "learning_rate": 6.8426165844625845e-06, "loss": 0.0417, "step": 39938 }, { "epoch": 12.29, "learning_rate": 6.842143806691013e-06, "loss": 0.0326, "step": 39939 }, { "epoch": 12.29, "learning_rate": 6.841671036759548e-06, "loss": 0.0369, "step": 39940 }, { "epoch": 12.29, "learning_rate": 6.8411982746693594e-06, "loss": 0.0289, "step": 39941 }, { "epoch": 12.29, "learning_rate": 6.840725520421627e-06, "loss": 0.0222, "step": 39942 }, { "epoch": 12.29, "learning_rate": 6.840252774017516e-06, "loss": 0.0304, "step": 39943 }, { "epoch": 12.29, "learning_rate": 6.83978003545821e-06, "loss": 0.0304, "step": 39944 }, { "epoch": 12.29, "learning_rate": 6.83930730474488e-06, "loss": 0.0478, "step": 39945 }, { "epoch": 12.29, "learning_rate": 6.838834581878695e-06, "loss": 0.0307, "step": 39946 }, { "epoch": 12.29, "learning_rate": 6.838361866860832e-06, "loss": 0.0388, "step": 39947 }, { "epoch": 12.29, "learning_rate": 6.837889159692462e-06, "loss": 0.0184, "step": 39948 }, { "epoch": 12.29, "learning_rate": 6.837416460374762e-06, "loss": 0.0338, "step": 39949 }, { "epoch": 12.29, "learning_rate": 6.836943768908905e-06, "loss": 0.0217, "step": 39950 }, { "epoch": 12.29, "learning_rate": 6.836471085296064e-06, "loss": 0.0481, "step": 39951 }, { "epoch": 12.29, "learning_rate": 6.835998409537412e-06, "loss": 0.0723, "step": 39952 }, { "epoch": 12.29, "learning_rate": 6.835525741634121e-06, "loss": 0.0427, "step": 39953 }, { "epoch": 12.29, "learning_rate": 6.8350530815873694e-06, "loss": 0.0245, "step": 39954 }, { "epoch": 12.29, "learning_rate": 6.83458042939833e-06, "loss": 0.033, "step": 39955 }, { "epoch": 12.29, "learning_rate": 6.83410778506817e-06, "loss": 0.0324, "step": 39956 }, { "epoch": 12.29, "learning_rate": 6.833635148598068e-06, "loss": 0.0404, "step": 39957 }, { "epoch": 12.29, "learning_rate": 6.833162519989195e-06, "loss": 0.0449, "step": 39958 }, { "epoch": 12.29, "learning_rate": 6.832689899242726e-06, "loss": 0.0235, "step": 39959 }, { "epoch": 12.3, "learning_rate": 6.8322172863598355e-06, "loss": 0.0232, "step": 39960 }, { "epoch": 12.3, "learning_rate": 6.8317446813416956e-06, "loss": 0.0606, "step": 39961 }, { "epoch": 12.3, "learning_rate": 6.83127208418948e-06, "loss": 0.0458, "step": 39962 }, { "epoch": 12.3, "learning_rate": 6.8307994949043576e-06, "loss": 0.0344, "step": 39963 }, { "epoch": 12.3, "learning_rate": 6.830326913487508e-06, "loss": 0.0286, "step": 39964 }, { "epoch": 12.3, "learning_rate": 6.829854339940103e-06, "loss": 0.0303, "step": 39965 }, { "epoch": 12.3, "learning_rate": 6.8293817742633175e-06, "loss": 0.0209, "step": 39966 }, { "epoch": 12.3, "learning_rate": 6.828909216458319e-06, "loss": 0.0272, "step": 39967 }, { "epoch": 12.3, "learning_rate": 6.828436666526281e-06, "loss": 0.0377, "step": 39968 }, { "epoch": 12.3, "learning_rate": 6.827964124468385e-06, "loss": 0.0248, "step": 39969 }, { "epoch": 12.3, "learning_rate": 6.827491590285798e-06, "loss": 0.0344, "step": 39970 }, { "epoch": 12.3, "learning_rate": 6.827019063979693e-06, "loss": 0.0128, "step": 39971 }, { "epoch": 12.3, "learning_rate": 6.826546545551244e-06, "loss": 0.0402, "step": 39972 }, { "epoch": 12.3, "learning_rate": 6.826074035001624e-06, "loss": 0.0246, "step": 39973 }, { "epoch": 12.3, "learning_rate": 6.825601532332007e-06, "loss": 0.0516, "step": 39974 }, { "epoch": 12.3, "learning_rate": 6.825129037543567e-06, "loss": 0.0186, "step": 39975 }, { "epoch": 12.3, "learning_rate": 6.824656550637476e-06, "loss": 0.0365, "step": 39976 }, { "epoch": 12.3, "learning_rate": 6.8241840716149054e-06, "loss": 0.0183, "step": 39977 }, { "epoch": 12.3, "learning_rate": 6.823711600477025e-06, "loss": 0.037, "step": 39978 }, { "epoch": 12.3, "learning_rate": 6.8232391372250195e-06, "loss": 0.021, "step": 39979 }, { "epoch": 12.3, "learning_rate": 6.822766681860053e-06, "loss": 0.0522, "step": 39980 }, { "epoch": 12.3, "learning_rate": 6.8222942343833e-06, "loss": 0.0226, "step": 39981 }, { "epoch": 12.3, "learning_rate": 6.821821794795933e-06, "loss": 0.0363, "step": 39982 }, { "epoch": 12.3, "learning_rate": 6.821349363099125e-06, "loss": 0.0376, "step": 39983 }, { "epoch": 12.3, "learning_rate": 6.82087693929405e-06, "loss": 0.0505, "step": 39984 }, { "epoch": 12.3, "learning_rate": 6.820404523381882e-06, "loss": 0.0256, "step": 39985 }, { "epoch": 12.3, "learning_rate": 6.819932115363791e-06, "loss": 0.0342, "step": 39986 }, { "epoch": 12.3, "learning_rate": 6.819459715240955e-06, "loss": 0.0524, "step": 39987 }, { "epoch": 12.3, "learning_rate": 6.818987323014534e-06, "loss": 0.0248, "step": 39988 }, { "epoch": 12.3, "learning_rate": 6.818514938685716e-06, "loss": 0.0223, "step": 39989 }, { "epoch": 12.3, "learning_rate": 6.81804256225567e-06, "loss": 0.0397, "step": 39990 }, { "epoch": 12.3, "learning_rate": 6.8175701937255645e-06, "loss": 0.0251, "step": 39991 }, { "epoch": 12.31, "learning_rate": 6.817097833096573e-06, "loss": 0.0283, "step": 39992 }, { "epoch": 12.31, "learning_rate": 6.816625480369868e-06, "loss": 0.0202, "step": 39993 }, { "epoch": 12.31, "learning_rate": 6.816153135546626e-06, "loss": 0.0471, "step": 39994 }, { "epoch": 12.31, "learning_rate": 6.815680798628017e-06, "loss": 0.0264, "step": 39995 }, { "epoch": 12.31, "learning_rate": 6.815208469615216e-06, "loss": 0.0658, "step": 39996 }, { "epoch": 12.31, "learning_rate": 6.814736148509389e-06, "loss": 0.0236, "step": 39997 }, { "epoch": 12.31, "learning_rate": 6.814263835311718e-06, "loss": 0.0384, "step": 39998 }, { "epoch": 12.31, "learning_rate": 6.813791530023369e-06, "loss": 0.0224, "step": 39999 }, { "epoch": 12.31, "learning_rate": 6.813319232645519e-06, "loss": 0.048, "step": 40000 }, { "epoch": 12.31, "learning_rate": 6.812846943179336e-06, "loss": 0.0439, "step": 40001 }, { "epoch": 12.31, "learning_rate": 6.812374661625993e-06, "loss": 0.0395, "step": 40002 }, { "epoch": 12.31, "learning_rate": 6.811902387986668e-06, "loss": 0.031, "step": 40003 }, { "epoch": 12.31, "learning_rate": 6.811430122262529e-06, "loss": 0.0319, "step": 40004 }, { "epoch": 12.31, "learning_rate": 6.810957864454749e-06, "loss": 0.0741, "step": 40005 }, { "epoch": 12.31, "learning_rate": 6.810485614564502e-06, "loss": 0.0222, "step": 40006 }, { "epoch": 12.31, "learning_rate": 6.810013372592956e-06, "loss": 0.0357, "step": 40007 }, { "epoch": 12.31, "learning_rate": 6.809541138541289e-06, "loss": 0.0451, "step": 40008 }, { "epoch": 12.31, "learning_rate": 6.80906891241067e-06, "loss": 0.0405, "step": 40009 }, { "epoch": 12.31, "learning_rate": 6.808596694202275e-06, "loss": 0.039, "step": 40010 }, { "epoch": 12.31, "learning_rate": 6.808124483917276e-06, "loss": 0.0292, "step": 40011 }, { "epoch": 12.31, "learning_rate": 6.807652281556836e-06, "loss": 0.0537, "step": 40012 }, { "epoch": 12.31, "learning_rate": 6.8071800871221406e-06, "loss": 0.0267, "step": 40013 }, { "epoch": 12.31, "learning_rate": 6.806707900614355e-06, "loss": 0.0388, "step": 40014 }, { "epoch": 12.31, "learning_rate": 6.806235722034653e-06, "loss": 0.0134, "step": 40015 }, { "epoch": 12.31, "learning_rate": 6.805763551384205e-06, "loss": 0.038, "step": 40016 }, { "epoch": 12.31, "learning_rate": 6.805291388664186e-06, "loss": 0.0282, "step": 40017 }, { "epoch": 12.31, "learning_rate": 6.804819233875767e-06, "loss": 0.0596, "step": 40018 }, { "epoch": 12.31, "learning_rate": 6.80434708702012e-06, "loss": 0.0563, "step": 40019 }, { "epoch": 12.31, "learning_rate": 6.803874948098419e-06, "loss": 0.0574, "step": 40020 }, { "epoch": 12.31, "learning_rate": 6.8034028171118375e-06, "loss": 0.0111, "step": 40021 }, { "epoch": 12.31, "learning_rate": 6.802930694061538e-06, "loss": 0.0296, "step": 40022 }, { "epoch": 12.31, "learning_rate": 6.802458578948703e-06, "loss": 0.0213, "step": 40023 }, { "epoch": 12.31, "learning_rate": 6.801986471774504e-06, "loss": 0.025, "step": 40024 }, { "epoch": 12.32, "learning_rate": 6.801514372540109e-06, "loss": 0.0217, "step": 40025 }, { "epoch": 12.32, "learning_rate": 6.80104228124669e-06, "loss": 0.0507, "step": 40026 }, { "epoch": 12.32, "learning_rate": 6.80057019789542e-06, "loss": 0.038, "step": 40027 }, { "epoch": 12.32, "learning_rate": 6.800098122487473e-06, "loss": 0.0281, "step": 40028 }, { "epoch": 12.32, "learning_rate": 6.79962605502402e-06, "loss": 0.0251, "step": 40029 }, { "epoch": 12.32, "learning_rate": 6.799153995506234e-06, "loss": 0.0359, "step": 40030 }, { "epoch": 12.32, "learning_rate": 6.798681943935284e-06, "loss": 0.0483, "step": 40031 }, { "epoch": 12.32, "learning_rate": 6.798209900312343e-06, "loss": 0.0553, "step": 40032 }, { "epoch": 12.32, "learning_rate": 6.797737864638583e-06, "loss": 0.0238, "step": 40033 }, { "epoch": 12.32, "learning_rate": 6.797265836915181e-06, "loss": 0.0384, "step": 40034 }, { "epoch": 12.32, "learning_rate": 6.796793817143301e-06, "loss": 0.0602, "step": 40035 }, { "epoch": 12.32, "learning_rate": 6.7963218053241195e-06, "loss": 0.0237, "step": 40036 }, { "epoch": 12.32, "learning_rate": 6.795849801458805e-06, "loss": 0.0232, "step": 40037 }, { "epoch": 12.32, "learning_rate": 6.7953778055485335e-06, "loss": 0.0177, "step": 40038 }, { "epoch": 12.32, "learning_rate": 6.794905817594475e-06, "loss": 0.0381, "step": 40039 }, { "epoch": 12.32, "learning_rate": 6.794433837597802e-06, "loss": 0.0552, "step": 40040 }, { "epoch": 12.32, "learning_rate": 6.793961865559684e-06, "loss": 0.0245, "step": 40041 }, { "epoch": 12.32, "learning_rate": 6.793489901481294e-06, "loss": 0.0248, "step": 40042 }, { "epoch": 12.32, "learning_rate": 6.793017945363804e-06, "loss": 0.0308, "step": 40043 }, { "epoch": 12.32, "learning_rate": 6.792545997208388e-06, "loss": 0.0295, "step": 40044 }, { "epoch": 12.32, "learning_rate": 6.792074057016218e-06, "loss": 0.0224, "step": 40045 }, { "epoch": 12.32, "learning_rate": 6.791602124788459e-06, "loss": 0.0322, "step": 40046 }, { "epoch": 12.32, "learning_rate": 6.791130200526285e-06, "loss": 0.0149, "step": 40047 }, { "epoch": 12.32, "learning_rate": 6.790658284230874e-06, "loss": 0.0386, "step": 40048 }, { "epoch": 12.32, "learning_rate": 6.790186375903393e-06, "loss": 0.0184, "step": 40049 }, { "epoch": 12.32, "learning_rate": 6.789714475545013e-06, "loss": 0.0492, "step": 40050 }, { "epoch": 12.32, "learning_rate": 6.789242583156905e-06, "loss": 0.0273, "step": 40051 }, { "epoch": 12.32, "learning_rate": 6.788770698740243e-06, "loss": 0.0541, "step": 40052 }, { "epoch": 12.32, "learning_rate": 6.788298822296197e-06, "loss": 0.0383, "step": 40053 }, { "epoch": 12.32, "learning_rate": 6.7878269538259404e-06, "loss": 0.0251, "step": 40054 }, { "epoch": 12.32, "learning_rate": 6.787355093330645e-06, "loss": 0.0319, "step": 40055 }, { "epoch": 12.32, "learning_rate": 6.786883240811479e-06, "loss": 0.0288, "step": 40056 }, { "epoch": 12.33, "learning_rate": 6.7864113962696125e-06, "loss": 0.0311, "step": 40057 }, { "epoch": 12.33, "learning_rate": 6.785939559706225e-06, "loss": 0.056, "step": 40058 }, { "epoch": 12.33, "learning_rate": 6.785467731122482e-06, "loss": 0.0257, "step": 40059 }, { "epoch": 12.33, "learning_rate": 6.7849959105195565e-06, "loss": 0.0307, "step": 40060 }, { "epoch": 12.33, "learning_rate": 6.784524097898619e-06, "loss": 0.0303, "step": 40061 }, { "epoch": 12.33, "learning_rate": 6.7840522932608375e-06, "loss": 0.0297, "step": 40062 }, { "epoch": 12.33, "learning_rate": 6.783580496607391e-06, "loss": 0.0475, "step": 40063 }, { "epoch": 12.33, "learning_rate": 6.783108707939447e-06, "loss": 0.0282, "step": 40064 }, { "epoch": 12.33, "learning_rate": 6.782636927258175e-06, "loss": 0.0216, "step": 40065 }, { "epoch": 12.33, "learning_rate": 6.782165154564753e-06, "loss": 0.035, "step": 40066 }, { "epoch": 12.33, "learning_rate": 6.781693389860338e-06, "loss": 0.1351, "step": 40067 }, { "epoch": 12.33, "learning_rate": 6.781221633146118e-06, "loss": 0.0235, "step": 40068 }, { "epoch": 12.33, "learning_rate": 6.780749884423257e-06, "loss": 0.0452, "step": 40069 }, { "epoch": 12.33, "learning_rate": 6.780278143692925e-06, "loss": 0.0331, "step": 40070 }, { "epoch": 12.33, "learning_rate": 6.7798064109562935e-06, "loss": 0.0235, "step": 40071 }, { "epoch": 12.33, "learning_rate": 6.779334686214533e-06, "loss": 0.0285, "step": 40072 }, { "epoch": 12.33, "learning_rate": 6.778862969468818e-06, "loss": 0.0651, "step": 40073 }, { "epoch": 12.33, "learning_rate": 6.778391260720319e-06, "loss": 0.0249, "step": 40074 }, { "epoch": 12.33, "learning_rate": 6.777919559970205e-06, "loss": 0.0162, "step": 40075 }, { "epoch": 12.33, "learning_rate": 6.777447867219651e-06, "loss": 0.0213, "step": 40076 }, { "epoch": 12.33, "learning_rate": 6.776976182469817e-06, "loss": 0.0274, "step": 40077 }, { "epoch": 12.33, "learning_rate": 6.776504505721888e-06, "loss": 0.0533, "step": 40078 }, { "epoch": 12.33, "learning_rate": 6.7760328369770314e-06, "loss": 0.0231, "step": 40079 }, { "epoch": 12.33, "learning_rate": 6.775561176236413e-06, "loss": 0.0412, "step": 40080 }, { "epoch": 12.33, "learning_rate": 6.775089523501208e-06, "loss": 0.0233, "step": 40081 }, { "epoch": 12.33, "learning_rate": 6.774617878772582e-06, "loss": 0.0518, "step": 40082 }, { "epoch": 12.33, "learning_rate": 6.7741462420517155e-06, "loss": 0.0348, "step": 40083 }, { "epoch": 12.33, "learning_rate": 6.773674613339774e-06, "loss": 0.0277, "step": 40084 }, { "epoch": 12.33, "learning_rate": 6.773202992637926e-06, "loss": 0.0431, "step": 40085 }, { "epoch": 12.33, "learning_rate": 6.772731379947346e-06, "loss": 0.0342, "step": 40086 }, { "epoch": 12.33, "learning_rate": 6.772259775269203e-06, "loss": 0.0299, "step": 40087 }, { "epoch": 12.33, "learning_rate": 6.77178817860467e-06, "loss": 0.0321, "step": 40088 }, { "epoch": 12.33, "learning_rate": 6.771316589954917e-06, "loss": 0.0148, "step": 40089 }, { "epoch": 12.34, "learning_rate": 6.770845009321117e-06, "loss": 0.0258, "step": 40090 }, { "epoch": 12.34, "learning_rate": 6.770373436704436e-06, "loss": 0.02, "step": 40091 }, { "epoch": 12.34, "learning_rate": 6.769901872106044e-06, "loss": 0.0351, "step": 40092 }, { "epoch": 12.34, "learning_rate": 6.769430315527117e-06, "loss": 0.0231, "step": 40093 }, { "epoch": 12.34, "learning_rate": 6.768958766968825e-06, "loss": 0.038, "step": 40094 }, { "epoch": 12.34, "learning_rate": 6.768487226432336e-06, "loss": 0.0425, "step": 40095 }, { "epoch": 12.34, "learning_rate": 6.7680156939188235e-06, "loss": 0.0297, "step": 40096 }, { "epoch": 12.34, "learning_rate": 6.767544169429453e-06, "loss": 0.0277, "step": 40097 }, { "epoch": 12.34, "learning_rate": 6.767072652965402e-06, "loss": 0.0858, "step": 40098 }, { "epoch": 12.34, "learning_rate": 6.7666011445278375e-06, "loss": 0.0421, "step": 40099 }, { "epoch": 12.34, "learning_rate": 6.766129644117932e-06, "loss": 0.0274, "step": 40100 }, { "epoch": 12.34, "learning_rate": 6.765658151736853e-06, "loss": 0.0456, "step": 40101 }, { "epoch": 12.34, "learning_rate": 6.76518666738577e-06, "loss": 0.037, "step": 40102 }, { "epoch": 12.34, "learning_rate": 6.764715191065862e-06, "loss": 0.0481, "step": 40103 }, { "epoch": 12.34, "learning_rate": 6.764243722778292e-06, "loss": 0.0488, "step": 40104 }, { "epoch": 12.34, "learning_rate": 6.763772262524233e-06, "loss": 0.0387, "step": 40105 }, { "epoch": 12.34, "learning_rate": 6.763300810304853e-06, "loss": 0.0391, "step": 40106 }, { "epoch": 12.34, "learning_rate": 6.7628293661213255e-06, "loss": 0.0372, "step": 40107 }, { "epoch": 12.34, "learning_rate": 6.76235792997482e-06, "loss": 0.0525, "step": 40108 }, { "epoch": 12.34, "learning_rate": 6.761886501866507e-06, "loss": 0.0292, "step": 40109 }, { "epoch": 12.34, "learning_rate": 6.761415081797558e-06, "loss": 0.0349, "step": 40110 }, { "epoch": 12.34, "learning_rate": 6.760943669769142e-06, "loss": 0.0412, "step": 40111 }, { "epoch": 12.34, "learning_rate": 6.760472265782425e-06, "loss": 0.0213, "step": 40112 }, { "epoch": 12.34, "learning_rate": 6.760000869838589e-06, "loss": 0.0309, "step": 40113 }, { "epoch": 12.34, "learning_rate": 6.759529481938794e-06, "loss": 0.0384, "step": 40114 }, { "epoch": 12.34, "learning_rate": 6.759058102084214e-06, "loss": 0.0159, "step": 40115 }, { "epoch": 12.34, "learning_rate": 6.75858673027602e-06, "loss": 0.0342, "step": 40116 }, { "epoch": 12.34, "learning_rate": 6.758115366515378e-06, "loss": 0.0446, "step": 40117 }, { "epoch": 12.34, "learning_rate": 6.757644010803464e-06, "loss": 0.0181, "step": 40118 }, { "epoch": 12.34, "learning_rate": 6.757172663141445e-06, "loss": 0.0314, "step": 40119 }, { "epoch": 12.34, "learning_rate": 6.756701323530492e-06, "loss": 0.0252, "step": 40120 }, { "epoch": 12.34, "learning_rate": 6.756229991971779e-06, "loss": 0.025, "step": 40121 }, { "epoch": 12.35, "learning_rate": 6.7557586684664655e-06, "loss": 0.0262, "step": 40122 }, { "epoch": 12.35, "learning_rate": 6.7552873530157325e-06, "loss": 0.0483, "step": 40123 }, { "epoch": 12.35, "learning_rate": 6.754816045620749e-06, "loss": 0.0369, "step": 40124 }, { "epoch": 12.35, "learning_rate": 6.75434474628268e-06, "loss": 0.0253, "step": 40125 }, { "epoch": 12.35, "learning_rate": 6.753873455002698e-06, "loss": 0.0446, "step": 40126 }, { "epoch": 12.35, "learning_rate": 6.753402171781969e-06, "loss": 0.0164, "step": 40127 }, { "epoch": 12.35, "learning_rate": 6.752930896621671e-06, "loss": 0.0227, "step": 40128 }, { "epoch": 12.35, "learning_rate": 6.75245962952297e-06, "loss": 0.0445, "step": 40129 }, { "epoch": 12.35, "learning_rate": 6.751988370487037e-06, "loss": 0.02, "step": 40130 }, { "epoch": 12.35, "learning_rate": 6.7515171195150405e-06, "loss": 0.0475, "step": 40131 }, { "epoch": 12.35, "learning_rate": 6.751045876608149e-06, "loss": 0.0447, "step": 40132 }, { "epoch": 12.35, "learning_rate": 6.750574641767538e-06, "loss": 0.0195, "step": 40133 }, { "epoch": 12.35, "learning_rate": 6.750103414994374e-06, "loss": 0.0381, "step": 40134 }, { "epoch": 12.35, "learning_rate": 6.749632196289827e-06, "loss": 0.0625, "step": 40135 }, { "epoch": 12.35, "learning_rate": 6.7491609856550656e-06, "loss": 0.0345, "step": 40136 }, { "epoch": 12.35, "learning_rate": 6.74868978309126e-06, "loss": 0.0176, "step": 40137 }, { "epoch": 12.35, "learning_rate": 6.748218588599582e-06, "loss": 0.0273, "step": 40138 }, { "epoch": 12.35, "learning_rate": 6.747747402181202e-06, "loss": 0.0296, "step": 40139 }, { "epoch": 12.35, "learning_rate": 6.747276223837289e-06, "loss": 0.0378, "step": 40140 }, { "epoch": 12.35, "learning_rate": 6.7468050535690096e-06, "loss": 0.0301, "step": 40141 }, { "epoch": 12.35, "learning_rate": 6.7463338913775365e-06, "loss": 0.0303, "step": 40142 }, { "epoch": 12.35, "learning_rate": 6.7458627372640396e-06, "loss": 0.038, "step": 40143 }, { "epoch": 12.35, "learning_rate": 6.745391591229689e-06, "loss": 0.0378, "step": 40144 }, { "epoch": 12.35, "learning_rate": 6.744920453275656e-06, "loss": 0.0504, "step": 40145 }, { "epoch": 12.35, "learning_rate": 6.744449323403106e-06, "loss": 0.0281, "step": 40146 }, { "epoch": 12.35, "learning_rate": 6.743978201613206e-06, "loss": 0.047, "step": 40147 }, { "epoch": 12.35, "learning_rate": 6.7435070879071374e-06, "loss": 0.042, "step": 40148 }, { "epoch": 12.35, "learning_rate": 6.743035982286059e-06, "loss": 0.0288, "step": 40149 }, { "epoch": 12.35, "learning_rate": 6.742564884751146e-06, "loss": 0.0278, "step": 40150 }, { "epoch": 12.35, "learning_rate": 6.742093795303566e-06, "loss": 0.0113, "step": 40151 }, { "epoch": 12.35, "learning_rate": 6.7416227139444854e-06, "loss": 0.076, "step": 40152 }, { "epoch": 12.35, "learning_rate": 6.741151640675081e-06, "loss": 0.0301, "step": 40153 }, { "epoch": 12.35, "learning_rate": 6.740680575496516e-06, "loss": 0.0282, "step": 40154 }, { "epoch": 12.36, "learning_rate": 6.740209518409967e-06, "loss": 0.0266, "step": 40155 }, { "epoch": 12.36, "learning_rate": 6.739738469416596e-06, "loss": 0.074, "step": 40156 }, { "epoch": 12.36, "learning_rate": 6.739267428517571e-06, "loss": 0.029, "step": 40157 }, { "epoch": 12.36, "learning_rate": 6.738796395714072e-06, "loss": 0.0363, "step": 40158 }, { "epoch": 12.36, "learning_rate": 6.738325371007262e-06, "loss": 0.0354, "step": 40159 }, { "epoch": 12.36, "learning_rate": 6.737854354398308e-06, "loss": 0.0271, "step": 40160 }, { "epoch": 12.36, "learning_rate": 6.737383345888384e-06, "loss": 0.0318, "step": 40161 }, { "epoch": 12.36, "learning_rate": 6.736912345478654e-06, "loss": 0.0288, "step": 40162 }, { "epoch": 12.36, "learning_rate": 6.736441353170295e-06, "loss": 0.0173, "step": 40163 }, { "epoch": 12.36, "learning_rate": 6.735970368964472e-06, "loss": 0.0296, "step": 40164 }, { "epoch": 12.36, "learning_rate": 6.7354993928623546e-06, "loss": 0.0828, "step": 40165 }, { "epoch": 12.36, "learning_rate": 6.735028424865113e-06, "loss": 0.0347, "step": 40166 }, { "epoch": 12.36, "learning_rate": 6.73455746497391e-06, "loss": 0.0142, "step": 40167 }, { "epoch": 12.36, "learning_rate": 6.734086513189924e-06, "loss": 0.0354, "step": 40168 }, { "epoch": 12.36, "learning_rate": 6.733615569514325e-06, "loss": 0.0292, "step": 40169 }, { "epoch": 12.36, "learning_rate": 6.733144633948274e-06, "loss": 0.0279, "step": 40170 }, { "epoch": 12.36, "learning_rate": 6.7326737064929444e-06, "loss": 0.0282, "step": 40171 }, { "epoch": 12.36, "learning_rate": 6.732202787149502e-06, "loss": 0.0381, "step": 40172 }, { "epoch": 12.36, "learning_rate": 6.731731875919123e-06, "loss": 0.0206, "step": 40173 }, { "epoch": 12.36, "learning_rate": 6.731260972802972e-06, "loss": 0.0176, "step": 40174 }, { "epoch": 12.36, "learning_rate": 6.7307900778022184e-06, "loss": 0.0295, "step": 40175 }, { "epoch": 12.36, "learning_rate": 6.730319190918036e-06, "loss": 0.0352, "step": 40176 }, { "epoch": 12.36, "learning_rate": 6.72984831215158e-06, "loss": 0.0251, "step": 40177 }, { "epoch": 12.36, "learning_rate": 6.729377441504035e-06, "loss": 0.022, "step": 40178 }, { "epoch": 12.36, "learning_rate": 6.728906578976566e-06, "loss": 0.0704, "step": 40179 }, { "epoch": 12.36, "learning_rate": 6.728435724570337e-06, "loss": 0.0336, "step": 40180 }, { "epoch": 12.36, "learning_rate": 6.727964878286521e-06, "loss": 0.0521, "step": 40181 }, { "epoch": 12.36, "learning_rate": 6.727494040126284e-06, "loss": 0.0363, "step": 40182 }, { "epoch": 12.36, "learning_rate": 6.7270232100908e-06, "loss": 0.027, "step": 40183 }, { "epoch": 12.36, "learning_rate": 6.7265523881812335e-06, "loss": 0.0404, "step": 40184 }, { "epoch": 12.36, "learning_rate": 6.726081574398755e-06, "loss": 0.0289, "step": 40185 }, { "epoch": 12.36, "learning_rate": 6.725610768744535e-06, "loss": 0.0147, "step": 40186 }, { "epoch": 12.37, "learning_rate": 6.725139971219738e-06, "loss": 0.0198, "step": 40187 }, { "epoch": 12.37, "learning_rate": 6.724669181825536e-06, "loss": 0.0581, "step": 40188 }, { "epoch": 12.37, "learning_rate": 6.724198400563099e-06, "loss": 0.0221, "step": 40189 }, { "epoch": 12.37, "learning_rate": 6.723727627433596e-06, "loss": 0.0376, "step": 40190 }, { "epoch": 12.37, "learning_rate": 6.723256862438192e-06, "loss": 0.0191, "step": 40191 }, { "epoch": 12.37, "learning_rate": 6.722786105578056e-06, "loss": 0.0393, "step": 40192 }, { "epoch": 12.37, "learning_rate": 6.72231535685436e-06, "loss": 0.0409, "step": 40193 }, { "epoch": 12.37, "learning_rate": 6.721844616268271e-06, "loss": 0.0265, "step": 40194 }, { "epoch": 12.37, "learning_rate": 6.721373883820959e-06, "loss": 0.0358, "step": 40195 }, { "epoch": 12.37, "learning_rate": 6.720903159513592e-06, "loss": 0.028, "step": 40196 }, { "epoch": 12.37, "learning_rate": 6.720432443347335e-06, "loss": 0.0569, "step": 40197 }, { "epoch": 12.37, "learning_rate": 6.719961735323364e-06, "loss": 0.0629, "step": 40198 }, { "epoch": 12.37, "learning_rate": 6.719491035442842e-06, "loss": 0.0421, "step": 40199 }, { "epoch": 12.37, "learning_rate": 6.719020343706942e-06, "loss": 0.0308, "step": 40200 }, { "epoch": 12.37, "learning_rate": 6.718549660116829e-06, "loss": 0.0109, "step": 40201 }, { "epoch": 12.37, "learning_rate": 6.718078984673666e-06, "loss": 0.0446, "step": 40202 }, { "epoch": 12.37, "learning_rate": 6.717608317378637e-06, "loss": 0.0384, "step": 40203 }, { "epoch": 12.37, "learning_rate": 6.717137658232897e-06, "loss": 0.0381, "step": 40204 }, { "epoch": 12.37, "learning_rate": 6.7166670072376205e-06, "loss": 0.0241, "step": 40205 }, { "epoch": 12.37, "learning_rate": 6.7161963643939755e-06, "loss": 0.0517, "step": 40206 }, { "epoch": 12.37, "learning_rate": 6.715725729703127e-06, "loss": 0.0297, "step": 40207 }, { "epoch": 12.37, "learning_rate": 6.715255103166247e-06, "loss": 0.0248, "step": 40208 }, { "epoch": 12.37, "learning_rate": 6.714784484784505e-06, "loss": 0.0328, "step": 40209 }, { "epoch": 12.37, "learning_rate": 6.714313874559067e-06, "loss": 0.0384, "step": 40210 }, { "epoch": 12.37, "learning_rate": 6.713843272491103e-06, "loss": 0.0401, "step": 40211 }, { "epoch": 12.37, "learning_rate": 6.713372678581773e-06, "loss": 0.0418, "step": 40212 }, { "epoch": 12.37, "learning_rate": 6.712902092832261e-06, "loss": 0.0221, "step": 40213 }, { "epoch": 12.37, "learning_rate": 6.712431515243724e-06, "loss": 0.0253, "step": 40214 }, { "epoch": 12.37, "learning_rate": 6.711960945817333e-06, "loss": 0.0471, "step": 40215 }, { "epoch": 12.37, "learning_rate": 6.711490384554257e-06, "loss": 0.0276, "step": 40216 }, { "epoch": 12.37, "learning_rate": 6.711019831455661e-06, "loss": 0.0219, "step": 40217 }, { "epoch": 12.37, "learning_rate": 6.710549286522718e-06, "loss": 0.035, "step": 40218 }, { "epoch": 12.37, "learning_rate": 6.710078749756596e-06, "loss": 0.0216, "step": 40219 }, { "epoch": 12.38, "learning_rate": 6.709608221158461e-06, "loss": 0.0217, "step": 40220 }, { "epoch": 12.38, "learning_rate": 6.709137700729482e-06, "loss": 0.0271, "step": 40221 }, { "epoch": 12.38, "learning_rate": 6.708667188470823e-06, "loss": 0.0353, "step": 40222 }, { "epoch": 12.38, "learning_rate": 6.708196684383659e-06, "loss": 0.0458, "step": 40223 }, { "epoch": 12.38, "learning_rate": 6.707726188469157e-06, "loss": 0.014, "step": 40224 }, { "epoch": 12.38, "learning_rate": 6.707255700728482e-06, "loss": 0.0346, "step": 40225 }, { "epoch": 12.38, "learning_rate": 6.7067852211628035e-06, "loss": 0.0304, "step": 40226 }, { "epoch": 12.38, "learning_rate": 6.706314749773287e-06, "loss": 0.033, "step": 40227 }, { "epoch": 12.38, "learning_rate": 6.705844286561105e-06, "loss": 0.0359, "step": 40228 }, { "epoch": 12.38, "learning_rate": 6.705373831527424e-06, "loss": 0.0379, "step": 40229 }, { "epoch": 12.38, "learning_rate": 6.704903384673411e-06, "loss": 0.0301, "step": 40230 }, { "epoch": 12.38, "learning_rate": 6.704432946000237e-06, "loss": 0.0277, "step": 40231 }, { "epoch": 12.38, "learning_rate": 6.703962515509062e-06, "loss": 0.0134, "step": 40232 }, { "epoch": 12.38, "learning_rate": 6.703492093201061e-06, "loss": 0.0367, "step": 40233 }, { "epoch": 12.38, "learning_rate": 6.703021679077405e-06, "loss": 0.0214, "step": 40234 }, { "epoch": 12.38, "learning_rate": 6.702551273139255e-06, "loss": 0.0342, "step": 40235 }, { "epoch": 12.38, "learning_rate": 6.702080875387781e-06, "loss": 0.0514, "step": 40236 }, { "epoch": 12.38, "learning_rate": 6.701610485824148e-06, "loss": 0.0404, "step": 40237 }, { "epoch": 12.38, "learning_rate": 6.7011401044495304e-06, "loss": 0.0391, "step": 40238 }, { "epoch": 12.38, "learning_rate": 6.700669731265092e-06, "loss": 0.0587, "step": 40239 }, { "epoch": 12.38, "learning_rate": 6.700199366272001e-06, "loss": 0.0286, "step": 40240 }, { "epoch": 12.38, "learning_rate": 6.699729009471425e-06, "loss": 0.0286, "step": 40241 }, { "epoch": 12.38, "learning_rate": 6.699258660864532e-06, "loss": 0.0488, "step": 40242 }, { "epoch": 12.38, "learning_rate": 6.698788320452489e-06, "loss": 0.0318, "step": 40243 }, { "epoch": 12.38, "learning_rate": 6.6983179882364665e-06, "loss": 0.0269, "step": 40244 }, { "epoch": 12.38, "learning_rate": 6.697847664217633e-06, "loss": 0.0255, "step": 40245 }, { "epoch": 12.38, "learning_rate": 6.697377348397151e-06, "loss": 0.0305, "step": 40246 }, { "epoch": 12.38, "learning_rate": 6.696907040776186e-06, "loss": 0.0213, "step": 40247 }, { "epoch": 12.38, "learning_rate": 6.696436741355916e-06, "loss": 0.0236, "step": 40248 }, { "epoch": 12.38, "learning_rate": 6.695966450137501e-06, "loss": 0.0448, "step": 40249 }, { "epoch": 12.38, "learning_rate": 6.6954961671221115e-06, "loss": 0.0221, "step": 40250 }, { "epoch": 12.38, "learning_rate": 6.695025892310913e-06, "loss": 0.0316, "step": 40251 }, { "epoch": 12.39, "learning_rate": 6.694555625705075e-06, "loss": 0.0398, "step": 40252 }, { "epoch": 12.39, "learning_rate": 6.694085367305763e-06, "loss": 0.0314, "step": 40253 }, { "epoch": 12.39, "learning_rate": 6.693615117114147e-06, "loss": 0.0248, "step": 40254 }, { "epoch": 12.39, "learning_rate": 6.693144875131397e-06, "loss": 0.0458, "step": 40255 }, { "epoch": 12.39, "learning_rate": 6.692674641358672e-06, "loss": 0.0202, "step": 40256 }, { "epoch": 12.39, "learning_rate": 6.692204415797141e-06, "loss": 0.0215, "step": 40257 }, { "epoch": 12.39, "learning_rate": 6.691734198447981e-06, "loss": 0.0237, "step": 40258 }, { "epoch": 12.39, "learning_rate": 6.691263989312351e-06, "loss": 0.0426, "step": 40259 }, { "epoch": 12.39, "learning_rate": 6.690793788391421e-06, "loss": 0.022, "step": 40260 }, { "epoch": 12.39, "learning_rate": 6.690323595686358e-06, "loss": 0.0213, "step": 40261 }, { "epoch": 12.39, "learning_rate": 6.689853411198326e-06, "loss": 0.0501, "step": 40262 }, { "epoch": 12.39, "learning_rate": 6.6893832349284995e-06, "loss": 0.0359, "step": 40263 }, { "epoch": 12.39, "learning_rate": 6.68891306687804e-06, "loss": 0.0284, "step": 40264 }, { "epoch": 12.39, "learning_rate": 6.6884429070481175e-06, "loss": 0.0403, "step": 40265 }, { "epoch": 12.39, "learning_rate": 6.6879727554398995e-06, "loss": 0.0239, "step": 40266 }, { "epoch": 12.39, "learning_rate": 6.687502612054549e-06, "loss": 0.0211, "step": 40267 }, { "epoch": 12.39, "learning_rate": 6.687032476893238e-06, "loss": 0.0313, "step": 40268 }, { "epoch": 12.39, "learning_rate": 6.6865623499571355e-06, "loss": 0.0263, "step": 40269 }, { "epoch": 12.39, "learning_rate": 6.686092231247402e-06, "loss": 0.026, "step": 40270 }, { "epoch": 12.39, "learning_rate": 6.685622120765209e-06, "loss": 0.0291, "step": 40271 }, { "epoch": 12.39, "learning_rate": 6.68515201851172e-06, "loss": 0.0335, "step": 40272 }, { "epoch": 12.39, "learning_rate": 6.684681924488108e-06, "loss": 0.0201, "step": 40273 }, { "epoch": 12.39, "learning_rate": 6.684211838695537e-06, "loss": 0.0238, "step": 40274 }, { "epoch": 12.39, "learning_rate": 6.683741761135174e-06, "loss": 0.0443, "step": 40275 }, { "epoch": 12.39, "learning_rate": 6.683271691808187e-06, "loss": 0.0189, "step": 40276 }, { "epoch": 12.39, "learning_rate": 6.682801630715737e-06, "loss": 0.0296, "step": 40277 }, { "epoch": 12.39, "learning_rate": 6.682331577859001e-06, "loss": 0.0352, "step": 40278 }, { "epoch": 12.39, "learning_rate": 6.6818615332391424e-06, "loss": 0.0273, "step": 40279 }, { "epoch": 12.39, "learning_rate": 6.681391496857326e-06, "loss": 0.0369, "step": 40280 }, { "epoch": 12.39, "learning_rate": 6.680921468714718e-06, "loss": 0.0466, "step": 40281 }, { "epoch": 12.39, "learning_rate": 6.680451448812486e-06, "loss": 0.0511, "step": 40282 }, { "epoch": 12.39, "learning_rate": 6.679981437151801e-06, "loss": 0.0241, "step": 40283 }, { "epoch": 12.39, "learning_rate": 6.679511433733827e-06, "loss": 0.018, "step": 40284 }, { "epoch": 12.4, "learning_rate": 6.67904143855973e-06, "loss": 0.0373, "step": 40285 }, { "epoch": 12.4, "learning_rate": 6.678571451630678e-06, "loss": 0.0285, "step": 40286 }, { "epoch": 12.4, "learning_rate": 6.678101472947836e-06, "loss": 0.0296, "step": 40287 }, { "epoch": 12.4, "learning_rate": 6.677631502512374e-06, "loss": 0.0271, "step": 40288 }, { "epoch": 12.4, "learning_rate": 6.677161540325461e-06, "loss": 0.0176, "step": 40289 }, { "epoch": 12.4, "learning_rate": 6.676691586388255e-06, "loss": 0.0564, "step": 40290 }, { "epoch": 12.4, "learning_rate": 6.67622164070193e-06, "loss": 0.0341, "step": 40291 }, { "epoch": 12.4, "learning_rate": 6.6757517032676475e-06, "loss": 0.0331, "step": 40292 }, { "epoch": 12.4, "learning_rate": 6.675281774086582e-06, "loss": 0.0243, "step": 40293 }, { "epoch": 12.4, "learning_rate": 6.6748118531598924e-06, "loss": 0.0163, "step": 40294 }, { "epoch": 12.4, "learning_rate": 6.67434194048875e-06, "loss": 0.0247, "step": 40295 }, { "epoch": 12.4, "learning_rate": 6.673872036074321e-06, "loss": 0.045, "step": 40296 }, { "epoch": 12.4, "learning_rate": 6.673402139917766e-06, "loss": 0.0543, "step": 40297 }, { "epoch": 12.4, "learning_rate": 6.672932252020261e-06, "loss": 0.0258, "step": 40298 }, { "epoch": 12.4, "learning_rate": 6.672462372382968e-06, "loss": 0.0262, "step": 40299 }, { "epoch": 12.4, "learning_rate": 6.671992501007055e-06, "loss": 0.019, "step": 40300 }, { "epoch": 12.4, "learning_rate": 6.671522637893686e-06, "loss": 0.0343, "step": 40301 }, { "epoch": 12.4, "learning_rate": 6.671052783044025e-06, "loss": 0.057, "step": 40302 }, { "epoch": 12.4, "learning_rate": 6.670582936459249e-06, "loss": 0.0357, "step": 40303 }, { "epoch": 12.4, "learning_rate": 6.670113098140514e-06, "loss": 0.0183, "step": 40304 }, { "epoch": 12.4, "learning_rate": 6.669643268088992e-06, "loss": 0.0259, "step": 40305 }, { "epoch": 12.4, "learning_rate": 6.6691734463058476e-06, "loss": 0.0465, "step": 40306 }, { "epoch": 12.4, "learning_rate": 6.668703632792246e-06, "loss": 0.026, "step": 40307 }, { "epoch": 12.4, "learning_rate": 6.6682338275493576e-06, "loss": 0.0237, "step": 40308 }, { "epoch": 12.4, "learning_rate": 6.667764030578346e-06, "loss": 0.0403, "step": 40309 }, { "epoch": 12.4, "learning_rate": 6.667294241880378e-06, "loss": 0.043, "step": 40310 }, { "epoch": 12.4, "learning_rate": 6.66682446145662e-06, "loss": 0.0323, "step": 40311 }, { "epoch": 12.4, "learning_rate": 6.666354689308234e-06, "loss": 0.0143, "step": 40312 }, { "epoch": 12.4, "learning_rate": 6.665884925436397e-06, "loss": 0.0639, "step": 40313 }, { "epoch": 12.4, "learning_rate": 6.665415169842267e-06, "loss": 0.032, "step": 40314 }, { "epoch": 12.4, "learning_rate": 6.6649454225270105e-06, "loss": 0.017, "step": 40315 }, { "epoch": 12.4, "learning_rate": 6.664475683491797e-06, "loss": 0.0324, "step": 40316 }, { "epoch": 12.41, "learning_rate": 6.664005952737789e-06, "loss": 0.0332, "step": 40317 }, { "epoch": 12.41, "learning_rate": 6.663536230266156e-06, "loss": 0.0157, "step": 40318 }, { "epoch": 12.41, "learning_rate": 6.6630665160780624e-06, "loss": 0.0324, "step": 40319 }, { "epoch": 12.41, "learning_rate": 6.662596810174677e-06, "loss": 0.0271, "step": 40320 }, { "epoch": 12.41, "learning_rate": 6.662127112557164e-06, "loss": 0.0168, "step": 40321 }, { "epoch": 12.41, "learning_rate": 6.661657423226685e-06, "loss": 0.0327, "step": 40322 }, { "epoch": 12.41, "learning_rate": 6.661187742184414e-06, "loss": 0.0364, "step": 40323 }, { "epoch": 12.41, "learning_rate": 6.6607180694315165e-06, "loss": 0.0302, "step": 40324 }, { "epoch": 12.41, "learning_rate": 6.660248404969154e-06, "loss": 0.0201, "step": 40325 }, { "epoch": 12.41, "learning_rate": 6.6597787487984935e-06, "loss": 0.0462, "step": 40326 }, { "epoch": 12.41, "learning_rate": 6.659309100920699e-06, "loss": 0.0258, "step": 40327 }, { "epoch": 12.41, "learning_rate": 6.658839461336943e-06, "loss": 0.0464, "step": 40328 }, { "epoch": 12.41, "learning_rate": 6.6583698300483864e-06, "loss": 0.0333, "step": 40329 }, { "epoch": 12.41, "learning_rate": 6.657900207056198e-06, "loss": 0.0236, "step": 40330 }, { "epoch": 12.41, "learning_rate": 6.657430592361544e-06, "loss": 0.0183, "step": 40331 }, { "epoch": 12.41, "learning_rate": 6.656960985965581e-06, "loss": 0.0276, "step": 40332 }, { "epoch": 12.41, "learning_rate": 6.656491387869489e-06, "loss": 0.0396, "step": 40333 }, { "epoch": 12.41, "learning_rate": 6.656021798074429e-06, "loss": 0.0431, "step": 40334 }, { "epoch": 12.41, "learning_rate": 6.655552216581562e-06, "loss": 0.0209, "step": 40335 }, { "epoch": 12.41, "learning_rate": 6.655082643392059e-06, "loss": 0.0164, "step": 40336 }, { "epoch": 12.41, "learning_rate": 6.654613078507081e-06, "loss": 0.0212, "step": 40337 }, { "epoch": 12.41, "learning_rate": 6.654143521927799e-06, "loss": 0.0402, "step": 40338 }, { "epoch": 12.41, "learning_rate": 6.653673973655377e-06, "loss": 0.0506, "step": 40339 }, { "epoch": 12.41, "learning_rate": 6.653204433690981e-06, "loss": 0.0373, "step": 40340 }, { "epoch": 12.41, "learning_rate": 6.652734902035775e-06, "loss": 0.031, "step": 40341 }, { "epoch": 12.41, "learning_rate": 6.652265378690923e-06, "loss": 0.0359, "step": 40342 }, { "epoch": 12.41, "learning_rate": 6.651795863657598e-06, "loss": 0.0187, "step": 40343 }, { "epoch": 12.41, "learning_rate": 6.651326356936959e-06, "loss": 0.033, "step": 40344 }, { "epoch": 12.41, "learning_rate": 6.650856858530177e-06, "loss": 0.0479, "step": 40345 }, { "epoch": 12.41, "learning_rate": 6.650387368438412e-06, "loss": 0.0444, "step": 40346 }, { "epoch": 12.41, "learning_rate": 6.649917886662829e-06, "loss": 0.031, "step": 40347 }, { "epoch": 12.41, "learning_rate": 6.649448413204602e-06, "loss": 0.0407, "step": 40348 }, { "epoch": 12.41, "learning_rate": 6.648978948064889e-06, "loss": 0.0324, "step": 40349 }, { "epoch": 12.42, "learning_rate": 6.648509491244859e-06, "loss": 0.0211, "step": 40350 }, { "epoch": 12.42, "learning_rate": 6.648040042745675e-06, "loss": 0.0307, "step": 40351 }, { "epoch": 12.42, "learning_rate": 6.647570602568503e-06, "loss": 0.0376, "step": 40352 }, { "epoch": 12.42, "learning_rate": 6.6471011707145105e-06, "loss": 0.0233, "step": 40353 }, { "epoch": 12.42, "learning_rate": 6.646631747184862e-06, "loss": 0.0387, "step": 40354 }, { "epoch": 12.42, "learning_rate": 6.646162331980726e-06, "loss": 0.0408, "step": 40355 }, { "epoch": 12.42, "learning_rate": 6.645692925103262e-06, "loss": 0.023, "step": 40356 }, { "epoch": 12.42, "learning_rate": 6.645223526553634e-06, "loss": 0.0357, "step": 40357 }, { "epoch": 12.42, "learning_rate": 6.644754136333018e-06, "loss": 0.0405, "step": 40358 }, { "epoch": 12.42, "learning_rate": 6.644284754442569e-06, "loss": 0.0303, "step": 40359 }, { "epoch": 12.42, "learning_rate": 6.643815380883458e-06, "loss": 0.0271, "step": 40360 }, { "epoch": 12.42, "learning_rate": 6.64334601565685e-06, "loss": 0.0461, "step": 40361 }, { "epoch": 12.42, "learning_rate": 6.6428766587639045e-06, "loss": 0.0189, "step": 40362 }, { "epoch": 12.42, "learning_rate": 6.642407310205794e-06, "loss": 0.0309, "step": 40363 }, { "epoch": 12.42, "learning_rate": 6.641937969983682e-06, "loss": 0.0295, "step": 40364 }, { "epoch": 12.42, "learning_rate": 6.641468638098732e-06, "loss": 0.0577, "step": 40365 }, { "epoch": 12.42, "learning_rate": 6.640999314552112e-06, "loss": 0.0407, "step": 40366 }, { "epoch": 12.42, "learning_rate": 6.640529999344978e-06, "loss": 0.0204, "step": 40367 }, { "epoch": 12.42, "learning_rate": 6.6400606924785095e-06, "loss": 0.0344, "step": 40368 }, { "epoch": 12.42, "learning_rate": 6.639591393953861e-06, "loss": 0.0207, "step": 40369 }, { "epoch": 12.42, "learning_rate": 6.639122103772204e-06, "loss": 0.0441, "step": 40370 }, { "epoch": 12.42, "learning_rate": 6.638652821934699e-06, "loss": 0.0337, "step": 40371 }, { "epoch": 12.42, "learning_rate": 6.638183548442511e-06, "loss": 0.0449, "step": 40372 }, { "epoch": 12.42, "learning_rate": 6.637714283296807e-06, "loss": 0.023, "step": 40373 }, { "epoch": 12.42, "learning_rate": 6.637245026498755e-06, "loss": 0.0217, "step": 40374 }, { "epoch": 12.42, "learning_rate": 6.636775778049516e-06, "loss": 0.0485, "step": 40375 }, { "epoch": 12.42, "learning_rate": 6.636306537950259e-06, "loss": 0.0258, "step": 40376 }, { "epoch": 12.42, "learning_rate": 6.635837306202137e-06, "loss": 0.0358, "step": 40377 }, { "epoch": 12.42, "learning_rate": 6.6353680828063306e-06, "loss": 0.0264, "step": 40378 }, { "epoch": 12.42, "learning_rate": 6.634898867764e-06, "loss": 0.0287, "step": 40379 }, { "epoch": 12.42, "learning_rate": 6.634429661076307e-06, "loss": 0.032, "step": 40380 }, { "epoch": 12.42, "learning_rate": 6.633960462744415e-06, "loss": 0.02, "step": 40381 }, { "epoch": 12.43, "learning_rate": 6.633491272769492e-06, "loss": 0.0248, "step": 40382 }, { "epoch": 12.43, "learning_rate": 6.633022091152704e-06, "loss": 0.0462, "step": 40383 }, { "epoch": 12.43, "learning_rate": 6.632552917895214e-06, "loss": 0.0306, "step": 40384 }, { "epoch": 12.43, "learning_rate": 6.6320837529981884e-06, "loss": 0.0389, "step": 40385 }, { "epoch": 12.43, "learning_rate": 6.63161459646279e-06, "loss": 0.0406, "step": 40386 }, { "epoch": 12.43, "learning_rate": 6.631145448290184e-06, "loss": 0.0547, "step": 40387 }, { "epoch": 12.43, "learning_rate": 6.6306763084815365e-06, "loss": 0.0537, "step": 40388 }, { "epoch": 12.43, "learning_rate": 6.630207177038013e-06, "loss": 0.022, "step": 40389 }, { "epoch": 12.43, "learning_rate": 6.629738053960774e-06, "loss": 0.0263, "step": 40390 }, { "epoch": 12.43, "learning_rate": 6.629268939250984e-06, "loss": 0.034, "step": 40391 }, { "epoch": 12.43, "learning_rate": 6.628799832909816e-06, "loss": 0.0258, "step": 40392 }, { "epoch": 12.43, "learning_rate": 6.628330734938427e-06, "loss": 0.0235, "step": 40393 }, { "epoch": 12.43, "learning_rate": 6.627861645337984e-06, "loss": 0.0306, "step": 40394 }, { "epoch": 12.43, "learning_rate": 6.627392564109652e-06, "loss": 0.023, "step": 40395 }, { "epoch": 12.43, "learning_rate": 6.6269234912545925e-06, "loss": 0.0397, "step": 40396 }, { "epoch": 12.43, "learning_rate": 6.626454426773974e-06, "loss": 0.0336, "step": 40397 }, { "epoch": 12.43, "learning_rate": 6.625985370668961e-06, "loss": 0.0455, "step": 40398 }, { "epoch": 12.43, "learning_rate": 6.625516322940716e-06, "loss": 0.0461, "step": 40399 }, { "epoch": 12.43, "learning_rate": 6.625047283590406e-06, "loss": 0.0305, "step": 40400 }, { "epoch": 12.43, "learning_rate": 6.624578252619188e-06, "loss": 0.0407, "step": 40401 }, { "epoch": 12.43, "learning_rate": 6.624109230028237e-06, "loss": 0.0303, "step": 40402 }, { "epoch": 12.43, "learning_rate": 6.623640215818714e-06, "loss": 0.0308, "step": 40403 }, { "epoch": 12.43, "learning_rate": 6.62317120999178e-06, "loss": 0.0242, "step": 40404 }, { "epoch": 12.43, "learning_rate": 6.622702212548602e-06, "loss": 0.0272, "step": 40405 }, { "epoch": 12.43, "learning_rate": 6.62223322349034e-06, "loss": 0.0489, "step": 40406 }, { "epoch": 12.43, "learning_rate": 6.621764242818167e-06, "loss": 0.0358, "step": 40407 }, { "epoch": 12.43, "learning_rate": 6.6212952705332416e-06, "loss": 0.0294, "step": 40408 }, { "epoch": 12.43, "learning_rate": 6.620826306636729e-06, "loss": 0.021, "step": 40409 }, { "epoch": 12.43, "learning_rate": 6.620357351129796e-06, "loss": 0.0551, "step": 40410 }, { "epoch": 12.43, "learning_rate": 6.619888404013597e-06, "loss": 0.0273, "step": 40411 }, { "epoch": 12.43, "learning_rate": 6.6194194652893095e-06, "loss": 0.0236, "step": 40412 }, { "epoch": 12.43, "learning_rate": 6.618950534958094e-06, "loss": 0.0193, "step": 40413 }, { "epoch": 12.43, "learning_rate": 6.61848161302111e-06, "loss": 0.0457, "step": 40414 }, { "epoch": 12.44, "learning_rate": 6.618012699479526e-06, "loss": 0.0257, "step": 40415 }, { "epoch": 12.44, "learning_rate": 6.617543794334501e-06, "loss": 0.033, "step": 40416 }, { "epoch": 12.44, "learning_rate": 6.617074897587206e-06, "loss": 0.0389, "step": 40417 }, { "epoch": 12.44, "learning_rate": 6.616606009238802e-06, "loss": 0.0551, "step": 40418 }, { "epoch": 12.44, "learning_rate": 6.616137129290452e-06, "loss": 0.0378, "step": 40419 }, { "epoch": 12.44, "learning_rate": 6.615668257743322e-06, "loss": 0.0241, "step": 40420 }, { "epoch": 12.44, "learning_rate": 6.6151993945985735e-06, "loss": 0.0254, "step": 40421 }, { "epoch": 12.44, "learning_rate": 6.614730539857375e-06, "loss": 0.0339, "step": 40422 }, { "epoch": 12.44, "learning_rate": 6.614261693520887e-06, "loss": 0.044, "step": 40423 }, { "epoch": 12.44, "learning_rate": 6.613792855590276e-06, "loss": 0.0445, "step": 40424 }, { "epoch": 12.44, "learning_rate": 6.613324026066703e-06, "loss": 0.0298, "step": 40425 }, { "epoch": 12.44, "learning_rate": 6.612855204951332e-06, "loss": 0.0205, "step": 40426 }, { "epoch": 12.44, "learning_rate": 6.612386392245329e-06, "loss": 0.0296, "step": 40427 }, { "epoch": 12.44, "learning_rate": 6.611917587949859e-06, "loss": 0.0519, "step": 40428 }, { "epoch": 12.44, "learning_rate": 6.611448792066083e-06, "loss": 0.0216, "step": 40429 }, { "epoch": 12.44, "learning_rate": 6.610980004595167e-06, "loss": 0.021, "step": 40430 }, { "epoch": 12.44, "learning_rate": 6.610511225538272e-06, "loss": 0.027, "step": 40431 }, { "epoch": 12.44, "learning_rate": 6.610042454896564e-06, "loss": 0.029, "step": 40432 }, { "epoch": 12.44, "learning_rate": 6.609573692671209e-06, "loss": 0.0316, "step": 40433 }, { "epoch": 12.44, "learning_rate": 6.609104938863369e-06, "loss": 0.0759, "step": 40434 }, { "epoch": 12.44, "learning_rate": 6.608636193474207e-06, "loss": 0.0359, "step": 40435 }, { "epoch": 12.44, "learning_rate": 6.608167456504882e-06, "loss": 0.0164, "step": 40436 }, { "epoch": 12.44, "learning_rate": 6.607698727956568e-06, "loss": 0.0463, "step": 40437 }, { "epoch": 12.44, "learning_rate": 6.607230007830422e-06, "loss": 0.0213, "step": 40438 }, { "epoch": 12.44, "learning_rate": 6.606761296127609e-06, "loss": 0.0288, "step": 40439 }, { "epoch": 12.44, "learning_rate": 6.606292592849294e-06, "loss": 0.0213, "step": 40440 }, { "epoch": 12.44, "learning_rate": 6.605823897996637e-06, "loss": 0.0301, "step": 40441 }, { "epoch": 12.44, "learning_rate": 6.605355211570805e-06, "loss": 0.041, "step": 40442 }, { "epoch": 12.44, "learning_rate": 6.604886533572963e-06, "loss": 0.0296, "step": 40443 }, { "epoch": 12.44, "learning_rate": 6.604417864004271e-06, "loss": 0.0237, "step": 40444 }, { "epoch": 12.44, "learning_rate": 6.603949202865897e-06, "loss": 0.0464, "step": 40445 }, { "epoch": 12.44, "learning_rate": 6.603480550158995e-06, "loss": 0.028, "step": 40446 }, { "epoch": 12.45, "learning_rate": 6.603011905884741e-06, "loss": 0.0296, "step": 40447 }, { "epoch": 12.45, "learning_rate": 6.60254327004429e-06, "loss": 0.0668, "step": 40448 }, { "epoch": 12.45, "learning_rate": 6.602074642638809e-06, "loss": 0.039, "step": 40449 }, { "epoch": 12.45, "learning_rate": 6.601606023669461e-06, "loss": 0.0206, "step": 40450 }, { "epoch": 12.45, "learning_rate": 6.601137413137406e-06, "loss": 0.026, "step": 40451 }, { "epoch": 12.45, "learning_rate": 6.6006688110438134e-06, "loss": 0.0175, "step": 40452 }, { "epoch": 12.45, "learning_rate": 6.600200217389844e-06, "loss": 0.0275, "step": 40453 }, { "epoch": 12.45, "learning_rate": 6.599731632176661e-06, "loss": 0.0259, "step": 40454 }, { "epoch": 12.45, "learning_rate": 6.599263055405429e-06, "loss": 0.0336, "step": 40455 }, { "epoch": 12.45, "learning_rate": 6.598794487077304e-06, "loss": 0.0333, "step": 40456 }, { "epoch": 12.45, "learning_rate": 6.598325927193459e-06, "loss": 0.0376, "step": 40457 }, { "epoch": 12.45, "learning_rate": 6.597857375755055e-06, "loss": 0.0316, "step": 40458 }, { "epoch": 12.45, "learning_rate": 6.597388832763254e-06, "loss": 0.0474, "step": 40459 }, { "epoch": 12.45, "learning_rate": 6.596920298219218e-06, "loss": 0.0408, "step": 40460 }, { "epoch": 12.45, "learning_rate": 6.596451772124109e-06, "loss": 0.0241, "step": 40461 }, { "epoch": 12.45, "learning_rate": 6.595983254479096e-06, "loss": 0.0436, "step": 40462 }, { "epoch": 12.45, "learning_rate": 6.5955147452853385e-06, "loss": 0.0414, "step": 40463 }, { "epoch": 12.45, "learning_rate": 6.595046244543999e-06, "loss": 0.0363, "step": 40464 }, { "epoch": 12.45, "learning_rate": 6.594577752256244e-06, "loss": 0.015, "step": 40465 }, { "epoch": 12.45, "learning_rate": 6.59410926842323e-06, "loss": 0.0353, "step": 40466 }, { "epoch": 12.45, "learning_rate": 6.593640793046128e-06, "loss": 0.0223, "step": 40467 }, { "epoch": 12.45, "learning_rate": 6.593172326126098e-06, "loss": 0.0295, "step": 40468 }, { "epoch": 12.45, "learning_rate": 6.592703867664302e-06, "loss": 0.0264, "step": 40469 }, { "epoch": 12.45, "learning_rate": 6.592235417661903e-06, "loss": 0.0432, "step": 40470 }, { "epoch": 12.45, "learning_rate": 6.591766976120063e-06, "loss": 0.0342, "step": 40471 }, { "epoch": 12.45, "learning_rate": 6.591298543039949e-06, "loss": 0.0118, "step": 40472 }, { "epoch": 12.45, "learning_rate": 6.590830118422722e-06, "loss": 0.0366, "step": 40473 }, { "epoch": 12.45, "learning_rate": 6.5903617022695435e-06, "loss": 0.0231, "step": 40474 }, { "epoch": 12.45, "learning_rate": 6.589893294581579e-06, "loss": 0.0331, "step": 40475 }, { "epoch": 12.45, "learning_rate": 6.589424895359988e-06, "loss": 0.0364, "step": 40476 }, { "epoch": 12.45, "learning_rate": 6.588956504605937e-06, "loss": 0.0174, "step": 40477 }, { "epoch": 12.45, "learning_rate": 6.588488122320586e-06, "loss": 0.0305, "step": 40478 }, { "epoch": 12.45, "learning_rate": 6.588019748505105e-06, "loss": 0.019, "step": 40479 }, { "epoch": 12.46, "learning_rate": 6.587551383160646e-06, "loss": 0.0353, "step": 40480 }, { "epoch": 12.46, "learning_rate": 6.587083026288374e-06, "loss": 0.0271, "step": 40481 }, { "epoch": 12.46, "learning_rate": 6.58661467788946e-06, "loss": 0.0464, "step": 40482 }, { "epoch": 12.46, "learning_rate": 6.58614633796506e-06, "loss": 0.0165, "step": 40483 }, { "epoch": 12.46, "learning_rate": 6.58567800651634e-06, "loss": 0.0453, "step": 40484 }, { "epoch": 12.46, "learning_rate": 6.585209683544459e-06, "loss": 0.0309, "step": 40485 }, { "epoch": 12.46, "learning_rate": 6.58474136905058e-06, "loss": 0.0203, "step": 40486 }, { "epoch": 12.46, "learning_rate": 6.58427306303587e-06, "loss": 0.0379, "step": 40487 }, { "epoch": 12.46, "learning_rate": 6.583804765501489e-06, "loss": 0.0406, "step": 40488 }, { "epoch": 12.46, "learning_rate": 6.583336476448602e-06, "loss": 0.0218, "step": 40489 }, { "epoch": 12.46, "learning_rate": 6.582868195878367e-06, "loss": 0.026, "step": 40490 }, { "epoch": 12.46, "learning_rate": 6.582399923791944e-06, "loss": 0.0229, "step": 40491 }, { "epoch": 12.46, "learning_rate": 6.581931660190508e-06, "loss": 0.0458, "step": 40492 }, { "epoch": 12.46, "learning_rate": 6.581463405075212e-06, "loss": 0.0258, "step": 40493 }, { "epoch": 12.46, "learning_rate": 6.580995158447221e-06, "loss": 0.0321, "step": 40494 }, { "epoch": 12.46, "learning_rate": 6.580526920307697e-06, "loss": 0.0437, "step": 40495 }, { "epoch": 12.46, "learning_rate": 6.580058690657801e-06, "loss": 0.0415, "step": 40496 }, { "epoch": 12.46, "learning_rate": 6.579590469498699e-06, "loss": 0.0398, "step": 40497 }, { "epoch": 12.46, "learning_rate": 6.579122256831551e-06, "loss": 0.0199, "step": 40498 }, { "epoch": 12.46, "learning_rate": 6.578654052657523e-06, "loss": 0.0279, "step": 40499 }, { "epoch": 12.46, "learning_rate": 6.5781858569777745e-06, "loss": 0.0313, "step": 40500 }, { "epoch": 12.46, "learning_rate": 6.577717669793462e-06, "loss": 0.0276, "step": 40501 }, { "epoch": 12.46, "learning_rate": 6.5772494911057575e-06, "loss": 0.0222, "step": 40502 }, { "epoch": 12.46, "learning_rate": 6.5767813209158225e-06, "loss": 0.0341, "step": 40503 }, { "epoch": 12.46, "learning_rate": 6.5763131592248144e-06, "loss": 0.0332, "step": 40504 }, { "epoch": 12.46, "learning_rate": 6.575845006033899e-06, "loss": 0.0383, "step": 40505 }, { "epoch": 12.46, "learning_rate": 6.575376861344233e-06, "loss": 0.0384, "step": 40506 }, { "epoch": 12.46, "learning_rate": 6.574908725156985e-06, "loss": 0.0333, "step": 40507 }, { "epoch": 12.46, "learning_rate": 6.574440597473317e-06, "loss": 0.0297, "step": 40508 }, { "epoch": 12.46, "learning_rate": 6.573972478294389e-06, "loss": 0.0158, "step": 40509 }, { "epoch": 12.46, "learning_rate": 6.5735043676213665e-06, "loss": 0.0367, "step": 40510 }, { "epoch": 12.46, "learning_rate": 6.5730362654554015e-06, "loss": 0.0247, "step": 40511 }, { "epoch": 12.47, "learning_rate": 6.572568171797669e-06, "loss": 0.0182, "step": 40512 }, { "epoch": 12.47, "learning_rate": 6.572100086649326e-06, "loss": 0.0293, "step": 40513 }, { "epoch": 12.47, "learning_rate": 6.571632010011533e-06, "loss": 0.0332, "step": 40514 }, { "epoch": 12.47, "learning_rate": 6.571163941885454e-06, "loss": 0.0523, "step": 40515 }, { "epoch": 12.47, "learning_rate": 6.570695882272248e-06, "loss": 0.0359, "step": 40516 }, { "epoch": 12.47, "learning_rate": 6.570227831173082e-06, "loss": 0.0276, "step": 40517 }, { "epoch": 12.47, "learning_rate": 6.569759788589114e-06, "loss": 0.0279, "step": 40518 }, { "epoch": 12.47, "learning_rate": 6.569291754521509e-06, "loss": 0.0331, "step": 40519 }, { "epoch": 12.47, "learning_rate": 6.568823728971428e-06, "loss": 0.0378, "step": 40520 }, { "epoch": 12.47, "learning_rate": 6.568355711940031e-06, "loss": 0.0187, "step": 40521 }, { "epoch": 12.47, "learning_rate": 6.567887703428484e-06, "loss": 0.0357, "step": 40522 }, { "epoch": 12.47, "learning_rate": 6.567419703437945e-06, "loss": 0.0556, "step": 40523 }, { "epoch": 12.47, "learning_rate": 6.566951711969581e-06, "loss": 0.0331, "step": 40524 }, { "epoch": 12.47, "learning_rate": 6.5664837290245465e-06, "loss": 0.0374, "step": 40525 }, { "epoch": 12.47, "learning_rate": 6.566015754604006e-06, "loss": 0.0205, "step": 40526 }, { "epoch": 12.47, "learning_rate": 6.565547788709126e-06, "loss": 0.0412, "step": 40527 }, { "epoch": 12.47, "learning_rate": 6.565079831341063e-06, "loss": 0.0377, "step": 40528 }, { "epoch": 12.47, "learning_rate": 6.564611882500982e-06, "loss": 0.025, "step": 40529 }, { "epoch": 12.47, "learning_rate": 6.564143942190042e-06, "loss": 0.0412, "step": 40530 }, { "epoch": 12.47, "learning_rate": 6.563676010409405e-06, "loss": 0.0279, "step": 40531 }, { "epoch": 12.47, "learning_rate": 6.563208087160238e-06, "loss": 0.0338, "step": 40532 }, { "epoch": 12.47, "learning_rate": 6.562740172443696e-06, "loss": 0.0501, "step": 40533 }, { "epoch": 12.47, "learning_rate": 6.562272266260947e-06, "loss": 0.0478, "step": 40534 }, { "epoch": 12.47, "learning_rate": 6.561804368613147e-06, "loss": 0.0183, "step": 40535 }, { "epoch": 12.47, "learning_rate": 6.561336479501455e-06, "loss": 0.0173, "step": 40536 }, { "epoch": 12.47, "learning_rate": 6.560868598927046e-06, "loss": 0.0273, "step": 40537 }, { "epoch": 12.47, "learning_rate": 6.560400726891069e-06, "loss": 0.0245, "step": 40538 }, { "epoch": 12.47, "learning_rate": 6.559932863394689e-06, "loss": 0.035, "step": 40539 }, { "epoch": 12.47, "learning_rate": 6.5594650084390694e-06, "loss": 0.0486, "step": 40540 }, { "epoch": 12.47, "learning_rate": 6.558997162025369e-06, "loss": 0.0309, "step": 40541 }, { "epoch": 12.47, "learning_rate": 6.558529324154752e-06, "loss": 0.0787, "step": 40542 }, { "epoch": 12.47, "learning_rate": 6.558061494828381e-06, "loss": 0.0443, "step": 40543 }, { "epoch": 12.47, "learning_rate": 6.557593674047413e-06, "loss": 0.0198, "step": 40544 }, { "epoch": 12.48, "learning_rate": 6.557125861813015e-06, "loss": 0.0291, "step": 40545 }, { "epoch": 12.48, "learning_rate": 6.5566580581263396e-06, "loss": 0.0306, "step": 40546 }, { "epoch": 12.48, "learning_rate": 6.556190262988559e-06, "loss": 0.0329, "step": 40547 }, { "epoch": 12.48, "learning_rate": 6.555722476400827e-06, "loss": 0.0701, "step": 40548 }, { "epoch": 12.48, "learning_rate": 6.555254698364308e-06, "loss": 0.0542, "step": 40549 }, { "epoch": 12.48, "learning_rate": 6.554786928880165e-06, "loss": 0.0307, "step": 40550 }, { "epoch": 12.48, "learning_rate": 6.554319167949553e-06, "loss": 0.0546, "step": 40551 }, { "epoch": 12.48, "learning_rate": 6.55385141557364e-06, "loss": 0.0294, "step": 40552 }, { "epoch": 12.48, "learning_rate": 6.553383671753586e-06, "loss": 0.0193, "step": 40553 }, { "epoch": 12.48, "learning_rate": 6.5529159364905496e-06, "loss": 0.0422, "step": 40554 }, { "epoch": 12.48, "learning_rate": 6.552448209785697e-06, "loss": 0.0253, "step": 40555 }, { "epoch": 12.48, "learning_rate": 6.551980491640179e-06, "loss": 0.0479, "step": 40556 }, { "epoch": 12.48, "learning_rate": 6.551512782055168e-06, "loss": 0.0099, "step": 40557 }, { "epoch": 12.48, "learning_rate": 6.551045081031824e-06, "loss": 0.049, "step": 40558 }, { "epoch": 12.48, "learning_rate": 6.550577388571302e-06, "loss": 0.0287, "step": 40559 }, { "epoch": 12.48, "learning_rate": 6.5501097046747674e-06, "loss": 0.0437, "step": 40560 }, { "epoch": 12.48, "learning_rate": 6.549642029343377e-06, "loss": 0.0463, "step": 40561 }, { "epoch": 12.48, "learning_rate": 6.549174362578299e-06, "loss": 0.0141, "step": 40562 }, { "epoch": 12.48, "learning_rate": 6.54870670438069e-06, "loss": 0.0415, "step": 40563 }, { "epoch": 12.48, "learning_rate": 6.548239054751712e-06, "loss": 0.036, "step": 40564 }, { "epoch": 12.48, "learning_rate": 6.5477714136925254e-06, "loss": 0.0419, "step": 40565 }, { "epoch": 12.48, "learning_rate": 6.547303781204289e-06, "loss": 0.0184, "step": 40566 }, { "epoch": 12.48, "learning_rate": 6.54683615728817e-06, "loss": 0.029, "step": 40567 }, { "epoch": 12.48, "learning_rate": 6.546368541945327e-06, "loss": 0.0255, "step": 40568 }, { "epoch": 12.48, "learning_rate": 6.545900935176919e-06, "loss": 0.0161, "step": 40569 }, { "epoch": 12.48, "learning_rate": 6.5454333369841064e-06, "loss": 0.0329, "step": 40570 }, { "epoch": 12.48, "learning_rate": 6.54496574736805e-06, "loss": 0.0263, "step": 40571 }, { "epoch": 12.48, "learning_rate": 6.5444981663299135e-06, "loss": 0.0428, "step": 40572 }, { "epoch": 12.48, "learning_rate": 6.544030593870858e-06, "loss": 0.0372, "step": 40573 }, { "epoch": 12.48, "learning_rate": 6.5435630299920415e-06, "loss": 0.0355, "step": 40574 }, { "epoch": 12.48, "learning_rate": 6.5430954746946276e-06, "loss": 0.0497, "step": 40575 }, { "epoch": 12.48, "learning_rate": 6.542627927979772e-06, "loss": 0.0469, "step": 40576 }, { "epoch": 12.49, "learning_rate": 6.5421603898486415e-06, "loss": 0.0396, "step": 40577 }, { "epoch": 12.49, "learning_rate": 6.541692860302395e-06, "loss": 0.0263, "step": 40578 }, { "epoch": 12.49, "learning_rate": 6.541225339342195e-06, "loss": 0.0812, "step": 40579 }, { "epoch": 12.49, "learning_rate": 6.540757826969197e-06, "loss": 0.0337, "step": 40580 }, { "epoch": 12.49, "learning_rate": 6.540290323184561e-06, "loss": 0.0354, "step": 40581 }, { "epoch": 12.49, "learning_rate": 6.539822827989459e-06, "loss": 0.0364, "step": 40582 }, { "epoch": 12.49, "learning_rate": 6.539355341385041e-06, "loss": 0.0324, "step": 40583 }, { "epoch": 12.49, "learning_rate": 6.538887863372472e-06, "loss": 0.0203, "step": 40584 }, { "epoch": 12.49, "learning_rate": 6.538420393952909e-06, "loss": 0.019, "step": 40585 }, { "epoch": 12.49, "learning_rate": 6.537952933127513e-06, "loss": 0.0333, "step": 40586 }, { "epoch": 12.49, "learning_rate": 6.537485480897451e-06, "loss": 0.0275, "step": 40587 }, { "epoch": 12.49, "learning_rate": 6.537018037263878e-06, "loss": 0.0235, "step": 40588 }, { "epoch": 12.49, "learning_rate": 6.5365506022279564e-06, "loss": 0.0304, "step": 40589 }, { "epoch": 12.49, "learning_rate": 6.536083175790846e-06, "loss": 0.0337, "step": 40590 }, { "epoch": 12.49, "learning_rate": 6.5356157579537015e-06, "loss": 0.0348, "step": 40591 }, { "epoch": 12.49, "learning_rate": 6.535148348717694e-06, "loss": 0.0195, "step": 40592 }, { "epoch": 12.49, "learning_rate": 6.53468094808398e-06, "loss": 0.0291, "step": 40593 }, { "epoch": 12.49, "learning_rate": 6.534213556053718e-06, "loss": 0.0446, "step": 40594 }, { "epoch": 12.49, "learning_rate": 6.533746172628069e-06, "loss": 0.0259, "step": 40595 }, { "epoch": 12.49, "learning_rate": 6.533278797808191e-06, "loss": 0.0314, "step": 40596 }, { "epoch": 12.49, "learning_rate": 6.532811431595251e-06, "loss": 0.0403, "step": 40597 }, { "epoch": 12.49, "learning_rate": 6.5323440739904045e-06, "loss": 0.0538, "step": 40598 }, { "epoch": 12.49, "learning_rate": 6.531876724994812e-06, "loss": 0.021, "step": 40599 }, { "epoch": 12.49, "learning_rate": 6.531409384609638e-06, "loss": 0.023, "step": 40600 }, { "epoch": 12.49, "learning_rate": 6.5309420528360335e-06, "loss": 0.0423, "step": 40601 }, { "epoch": 12.49, "learning_rate": 6.530474729675167e-06, "loss": 0.0306, "step": 40602 }, { "epoch": 12.49, "learning_rate": 6.5300074151282e-06, "loss": 0.0252, "step": 40603 }, { "epoch": 12.49, "learning_rate": 6.529540109196286e-06, "loss": 0.0391, "step": 40604 }, { "epoch": 12.49, "learning_rate": 6.5290728118805895e-06, "loss": 0.0653, "step": 40605 }, { "epoch": 12.49, "learning_rate": 6.5286055231822665e-06, "loss": 0.0165, "step": 40606 }, { "epoch": 12.49, "learning_rate": 6.528138243102482e-06, "loss": 0.0353, "step": 40607 }, { "epoch": 12.49, "learning_rate": 6.527670971642395e-06, "loss": 0.0504, "step": 40608 }, { "epoch": 12.49, "learning_rate": 6.5272037088031645e-06, "loss": 0.0376, "step": 40609 }, { "epoch": 12.5, "learning_rate": 6.526736454585954e-06, "loss": 0.0398, "step": 40610 }, { "epoch": 12.5, "learning_rate": 6.5262692089919135e-06, "loss": 0.0504, "step": 40611 }, { "epoch": 12.5, "learning_rate": 6.525801972022213e-06, "loss": 0.0361, "step": 40612 }, { "epoch": 12.5, "learning_rate": 6.525334743678014e-06, "loss": 0.0086, "step": 40613 }, { "epoch": 12.5, "learning_rate": 6.524867523960468e-06, "loss": 0.0415, "step": 40614 }, { "epoch": 12.5, "learning_rate": 6.52440031287074e-06, "loss": 0.0272, "step": 40615 }, { "epoch": 12.5, "learning_rate": 6.5239331104099855e-06, "loss": 0.0505, "step": 40616 }, { "epoch": 12.5, "learning_rate": 6.523465916579372e-06, "loss": 0.0269, "step": 40617 }, { "epoch": 12.5, "learning_rate": 6.522998731380055e-06, "loss": 0.0412, "step": 40618 }, { "epoch": 12.5, "learning_rate": 6.522531554813194e-06, "loss": 0.0166, "step": 40619 }, { "epoch": 12.5, "learning_rate": 6.52206438687995e-06, "loss": 0.0188, "step": 40620 }, { "epoch": 12.5, "learning_rate": 6.521597227581481e-06, "loss": 0.0275, "step": 40621 }, { "epoch": 12.5, "learning_rate": 6.521130076918949e-06, "loss": 0.03, "step": 40622 }, { "epoch": 12.5, "learning_rate": 6.520662934893514e-06, "loss": 0.0404, "step": 40623 }, { "epoch": 12.5, "learning_rate": 6.520195801506338e-06, "loss": 0.0438, "step": 40624 }, { "epoch": 12.5, "learning_rate": 6.519728676758575e-06, "loss": 0.0294, "step": 40625 }, { "epoch": 12.5, "learning_rate": 6.519261560651384e-06, "loss": 0.025, "step": 40626 }, { "epoch": 12.5, "learning_rate": 6.5187944531859325e-06, "loss": 0.0255, "step": 40627 }, { "epoch": 12.5, "learning_rate": 6.518327354363374e-06, "loss": 0.0356, "step": 40628 }, { "epoch": 12.5, "learning_rate": 6.517860264184872e-06, "loss": 0.0275, "step": 40629 }, { "epoch": 12.5, "learning_rate": 6.517393182651582e-06, "loss": 0.0324, "step": 40630 }, { "epoch": 12.5, "learning_rate": 6.516926109764665e-06, "loss": 0.0336, "step": 40631 }, { "epoch": 12.5, "learning_rate": 6.516459045525283e-06, "loss": 0.0485, "step": 40632 }, { "epoch": 12.5, "learning_rate": 6.5159919899345946e-06, "loss": 0.0335, "step": 40633 }, { "epoch": 12.5, "learning_rate": 6.515524942993761e-06, "loss": 0.0364, "step": 40634 }, { "epoch": 12.5, "learning_rate": 6.515057904703936e-06, "loss": 0.0386, "step": 40635 }, { "epoch": 12.5, "learning_rate": 6.51459087506628e-06, "loss": 0.0302, "step": 40636 }, { "epoch": 12.5, "learning_rate": 6.514123854081962e-06, "loss": 0.0497, "step": 40637 }, { "epoch": 12.5, "learning_rate": 6.513656841752132e-06, "loss": 0.0265, "step": 40638 }, { "epoch": 12.5, "learning_rate": 6.513189838077951e-06, "loss": 0.0542, "step": 40639 }, { "epoch": 12.5, "learning_rate": 6.512722843060582e-06, "loss": 0.0415, "step": 40640 }, { "epoch": 12.5, "learning_rate": 6.5122558567011775e-06, "loss": 0.0258, "step": 40641 }, { "epoch": 12.51, "learning_rate": 6.511788879000905e-06, "loss": 0.0203, "step": 40642 }, { "epoch": 12.51, "learning_rate": 6.5113219099609195e-06, "loss": 0.0385, "step": 40643 }, { "epoch": 12.51, "learning_rate": 6.5108549495823815e-06, "loss": 0.0334, "step": 40644 }, { "epoch": 12.51, "learning_rate": 6.510387997866452e-06, "loss": 0.0394, "step": 40645 }, { "epoch": 12.51, "learning_rate": 6.509921054814283e-06, "loss": 0.069, "step": 40646 }, { "epoch": 12.51, "learning_rate": 6.509454120427043e-06, "loss": 0.0559, "step": 40647 }, { "epoch": 12.51, "learning_rate": 6.508987194705888e-06, "loss": 0.0416, "step": 40648 }, { "epoch": 12.51, "learning_rate": 6.508520277651975e-06, "loss": 0.0306, "step": 40649 }, { "epoch": 12.51, "learning_rate": 6.508053369266466e-06, "loss": 0.0341, "step": 40650 }, { "epoch": 12.51, "learning_rate": 6.5075864695505165e-06, "loss": 0.0262, "step": 40651 }, { "epoch": 12.51, "learning_rate": 6.5071195785052895e-06, "loss": 0.0181, "step": 40652 }, { "epoch": 12.51, "learning_rate": 6.506652696131945e-06, "loss": 0.0358, "step": 40653 }, { "epoch": 12.51, "learning_rate": 6.506185822431639e-06, "loss": 0.0277, "step": 40654 }, { "epoch": 12.51, "learning_rate": 6.505718957405533e-06, "loss": 0.0328, "step": 40655 }, { "epoch": 12.51, "learning_rate": 6.50525210105478e-06, "loss": 0.0347, "step": 40656 }, { "epoch": 12.51, "learning_rate": 6.504785253380547e-06, "loss": 0.029, "step": 40657 }, { "epoch": 12.51, "learning_rate": 6.504318414383992e-06, "loss": 0.0421, "step": 40658 }, { "epoch": 12.51, "learning_rate": 6.5038515840662696e-06, "loss": 0.053, "step": 40659 }, { "epoch": 12.51, "learning_rate": 6.503384762428542e-06, "loss": 0.0554, "step": 40660 }, { "epoch": 12.51, "learning_rate": 6.502917949471965e-06, "loss": 0.0386, "step": 40661 }, { "epoch": 12.51, "learning_rate": 6.502451145197701e-06, "loss": 0.0303, "step": 40662 }, { "epoch": 12.51, "learning_rate": 6.501984349606909e-06, "loss": 0.0392, "step": 40663 }, { "epoch": 12.51, "learning_rate": 6.501517562700747e-06, "loss": 0.0416, "step": 40664 }, { "epoch": 12.51, "learning_rate": 6.501050784480374e-06, "loss": 0.0428, "step": 40665 }, { "epoch": 12.51, "learning_rate": 6.500584014946947e-06, "loss": 0.022, "step": 40666 }, { "epoch": 12.51, "learning_rate": 6.500117254101627e-06, "loss": 0.0445, "step": 40667 }, { "epoch": 12.51, "learning_rate": 6.499650501945575e-06, "loss": 0.0487, "step": 40668 }, { "epoch": 12.51, "learning_rate": 6.499183758479944e-06, "loss": 0.0326, "step": 40669 }, { "epoch": 12.51, "learning_rate": 6.498717023705898e-06, "loss": 0.0179, "step": 40670 }, { "epoch": 12.51, "learning_rate": 6.498250297624591e-06, "loss": 0.0453, "step": 40671 }, { "epoch": 12.51, "learning_rate": 6.4977835802371845e-06, "loss": 0.0759, "step": 40672 }, { "epoch": 12.51, "learning_rate": 6.49731687154484e-06, "loss": 0.0154, "step": 40673 }, { "epoch": 12.51, "learning_rate": 6.496850171548712e-06, "loss": 0.0247, "step": 40674 }, { "epoch": 12.52, "learning_rate": 6.4963834802499615e-06, "loss": 0.0351, "step": 40675 }, { "epoch": 12.52, "learning_rate": 6.495916797649743e-06, "loss": 0.0477, "step": 40676 }, { "epoch": 12.52, "learning_rate": 6.4954501237492205e-06, "loss": 0.0288, "step": 40677 }, { "epoch": 12.52, "learning_rate": 6.494983458549551e-06, "loss": 0.0279, "step": 40678 }, { "epoch": 12.52, "learning_rate": 6.494516802051896e-06, "loss": 0.0455, "step": 40679 }, { "epoch": 12.52, "learning_rate": 6.494050154257408e-06, "loss": 0.0156, "step": 40680 }, { "epoch": 12.52, "learning_rate": 6.493583515167242e-06, "loss": 0.0229, "step": 40681 }, { "epoch": 12.52, "learning_rate": 6.493116884782571e-06, "loss": 0.0255, "step": 40682 }, { "epoch": 12.52, "learning_rate": 6.492650263104543e-06, "loss": 0.031, "step": 40683 }, { "epoch": 12.52, "learning_rate": 6.492183650134318e-06, "loss": 0.066, "step": 40684 }, { "epoch": 12.52, "learning_rate": 6.491717045873056e-06, "loss": 0.0493, "step": 40685 }, { "epoch": 12.52, "learning_rate": 6.491250450321913e-06, "loss": 0.0242, "step": 40686 }, { "epoch": 12.52, "learning_rate": 6.490783863482052e-06, "loss": 0.0242, "step": 40687 }, { "epoch": 12.52, "learning_rate": 6.490317285354628e-06, "loss": 0.0372, "step": 40688 }, { "epoch": 12.52, "learning_rate": 6.4898507159408016e-06, "loss": 0.0213, "step": 40689 }, { "epoch": 12.52, "learning_rate": 6.4893841552417266e-06, "loss": 0.0329, "step": 40690 }, { "epoch": 12.52, "learning_rate": 6.488917603258562e-06, "loss": 0.0291, "step": 40691 }, { "epoch": 12.52, "learning_rate": 6.488451059992473e-06, "loss": 0.0454, "step": 40692 }, { "epoch": 12.52, "learning_rate": 6.487984525444613e-06, "loss": 0.0188, "step": 40693 }, { "epoch": 12.52, "learning_rate": 6.487517999616139e-06, "loss": 0.0249, "step": 40694 }, { "epoch": 12.52, "learning_rate": 6.487051482508212e-06, "loss": 0.0624, "step": 40695 }, { "epoch": 12.52, "learning_rate": 6.486584974121988e-06, "loss": 0.0457, "step": 40696 }, { "epoch": 12.52, "learning_rate": 6.4861184744586255e-06, "loss": 0.0323, "step": 40697 }, { "epoch": 12.52, "learning_rate": 6.485651983519286e-06, "loss": 0.0147, "step": 40698 }, { "epoch": 12.52, "learning_rate": 6.485185501305125e-06, "loss": 0.0393, "step": 40699 }, { "epoch": 12.52, "learning_rate": 6.484719027817301e-06, "loss": 0.0465, "step": 40700 }, { "epoch": 12.52, "learning_rate": 6.484252563056967e-06, "loss": 0.0588, "step": 40701 }, { "epoch": 12.52, "learning_rate": 6.483786107025289e-06, "loss": 0.0452, "step": 40702 }, { "epoch": 12.52, "learning_rate": 6.483319659723426e-06, "loss": 0.0386, "step": 40703 }, { "epoch": 12.52, "learning_rate": 6.482853221152529e-06, "loss": 0.047, "step": 40704 }, { "epoch": 12.52, "learning_rate": 6.4823867913137605e-06, "loss": 0.0372, "step": 40705 }, { "epoch": 12.52, "learning_rate": 6.481920370208274e-06, "loss": 0.0245, "step": 40706 }, { "epoch": 12.53, "learning_rate": 6.481453957837234e-06, "loss": 0.0273, "step": 40707 }, { "epoch": 12.53, "learning_rate": 6.480987554201794e-06, "loss": 0.057, "step": 40708 }, { "epoch": 12.53, "learning_rate": 6.480521159303114e-06, "loss": 0.0267, "step": 40709 }, { "epoch": 12.53, "learning_rate": 6.4800547731423525e-06, "loss": 0.034, "step": 40710 }, { "epoch": 12.53, "learning_rate": 6.479588395720661e-06, "loss": 0.0198, "step": 40711 }, { "epoch": 12.53, "learning_rate": 6.4791220270392065e-06, "loss": 0.0428, "step": 40712 }, { "epoch": 12.53, "learning_rate": 6.478655667099143e-06, "loss": 0.0202, "step": 40713 }, { "epoch": 12.53, "learning_rate": 6.478189315901629e-06, "loss": 0.0486, "step": 40714 }, { "epoch": 12.53, "learning_rate": 6.477722973447819e-06, "loss": 0.0391, "step": 40715 }, { "epoch": 12.53, "learning_rate": 6.477256639738872e-06, "loss": 0.0336, "step": 40716 }, { "epoch": 12.53, "learning_rate": 6.476790314775949e-06, "loss": 0.0462, "step": 40717 }, { "epoch": 12.53, "learning_rate": 6.476323998560206e-06, "loss": 0.0187, "step": 40718 }, { "epoch": 12.53, "learning_rate": 6.475857691092801e-06, "loss": 0.0323, "step": 40719 }, { "epoch": 12.53, "learning_rate": 6.47539139237489e-06, "loss": 0.01, "step": 40720 }, { "epoch": 12.53, "learning_rate": 6.474925102407631e-06, "loss": 0.0377, "step": 40721 }, { "epoch": 12.53, "learning_rate": 6.474458821192184e-06, "loss": 0.0468, "step": 40722 }, { "epoch": 12.53, "learning_rate": 6.473992548729706e-06, "loss": 0.0403, "step": 40723 }, { "epoch": 12.53, "learning_rate": 6.4735262850213566e-06, "loss": 0.0227, "step": 40724 }, { "epoch": 12.53, "learning_rate": 6.473060030068288e-06, "loss": 0.0246, "step": 40725 }, { "epoch": 12.53, "learning_rate": 6.472593783871657e-06, "loss": 0.0221, "step": 40726 }, { "epoch": 12.53, "learning_rate": 6.472127546432628e-06, "loss": 0.0273, "step": 40727 }, { "epoch": 12.53, "learning_rate": 6.471661317752355e-06, "loss": 0.0404, "step": 40728 }, { "epoch": 12.53, "learning_rate": 6.471195097831997e-06, "loss": 0.0482, "step": 40729 }, { "epoch": 12.53, "learning_rate": 6.470728886672708e-06, "loss": 0.0323, "step": 40730 }, { "epoch": 12.53, "learning_rate": 6.470262684275648e-06, "loss": 0.0199, "step": 40731 }, { "epoch": 12.53, "learning_rate": 6.469796490641974e-06, "loss": 0.0266, "step": 40732 }, { "epoch": 12.53, "learning_rate": 6.469330305772845e-06, "loss": 0.0365, "step": 40733 }, { "epoch": 12.53, "learning_rate": 6.468864129669418e-06, "loss": 0.0389, "step": 40734 }, { "epoch": 12.53, "learning_rate": 6.468397962332848e-06, "loss": 0.035, "step": 40735 }, { "epoch": 12.53, "learning_rate": 6.46793180376429e-06, "loss": 0.037, "step": 40736 }, { "epoch": 12.53, "learning_rate": 6.4674656539649104e-06, "loss": 0.0275, "step": 40737 }, { "epoch": 12.53, "learning_rate": 6.466999512935861e-06, "loss": 0.0473, "step": 40738 }, { "epoch": 12.53, "learning_rate": 6.4665333806782974e-06, "loss": 0.0434, "step": 40739 }, { "epoch": 12.54, "learning_rate": 6.46606725719338e-06, "loss": 0.0683, "step": 40740 }, { "epoch": 12.54, "learning_rate": 6.465601142482263e-06, "loss": 0.0252, "step": 40741 }, { "epoch": 12.54, "learning_rate": 6.4651350365461065e-06, "loss": 0.0389, "step": 40742 }, { "epoch": 12.54, "learning_rate": 6.464668939386068e-06, "loss": 0.0239, "step": 40743 }, { "epoch": 12.54, "learning_rate": 6.4642028510033e-06, "loss": 0.0461, "step": 40744 }, { "epoch": 12.54, "learning_rate": 6.46373677139897e-06, "loss": 0.0276, "step": 40745 }, { "epoch": 12.54, "learning_rate": 6.463270700574219e-06, "loss": 0.0442, "step": 40746 }, { "epoch": 12.54, "learning_rate": 6.46280463853022e-06, "loss": 0.0245, "step": 40747 }, { "epoch": 12.54, "learning_rate": 6.462338585268122e-06, "loss": 0.0398, "step": 40748 }, { "epoch": 12.54, "learning_rate": 6.461872540789084e-06, "loss": 0.0472, "step": 40749 }, { "epoch": 12.54, "learning_rate": 6.461406505094261e-06, "loss": 0.0533, "step": 40750 }, { "epoch": 12.54, "learning_rate": 6.4609404781848115e-06, "loss": 0.0372, "step": 40751 }, { "epoch": 12.54, "learning_rate": 6.460474460061894e-06, "loss": 0.0391, "step": 40752 }, { "epoch": 12.54, "learning_rate": 6.4600084507266636e-06, "loss": 0.0396, "step": 40753 }, { "epoch": 12.54, "learning_rate": 6.4595424501802785e-06, "loss": 0.0365, "step": 40754 }, { "epoch": 12.54, "learning_rate": 6.4590764584238964e-06, "loss": 0.0284, "step": 40755 }, { "epoch": 12.54, "learning_rate": 6.458610475458666e-06, "loss": 0.0247, "step": 40756 }, { "epoch": 12.54, "learning_rate": 6.458144501285757e-06, "loss": 0.0368, "step": 40757 }, { "epoch": 12.54, "learning_rate": 6.4576785359063225e-06, "loss": 0.0304, "step": 40758 }, { "epoch": 12.54, "learning_rate": 6.457212579321515e-06, "loss": 0.0294, "step": 40759 }, { "epoch": 12.54, "learning_rate": 6.456746631532492e-06, "loss": 0.0189, "step": 40760 }, { "epoch": 12.54, "learning_rate": 6.456280692540411e-06, "loss": 0.0362, "step": 40761 }, { "epoch": 12.54, "learning_rate": 6.4558147623464315e-06, "loss": 0.0571, "step": 40762 }, { "epoch": 12.54, "learning_rate": 6.455348840951708e-06, "loss": 0.0248, "step": 40763 }, { "epoch": 12.54, "learning_rate": 6.4548829283573985e-06, "loss": 0.0433, "step": 40764 }, { "epoch": 12.54, "learning_rate": 6.454417024564659e-06, "loss": 0.0416, "step": 40765 }, { "epoch": 12.54, "learning_rate": 6.453951129574644e-06, "loss": 0.0139, "step": 40766 }, { "epoch": 12.54, "learning_rate": 6.453485243388513e-06, "loss": 0.0158, "step": 40767 }, { "epoch": 12.54, "learning_rate": 6.453019366007427e-06, "loss": 0.0766, "step": 40768 }, { "epoch": 12.54, "learning_rate": 6.452553497432533e-06, "loss": 0.0168, "step": 40769 }, { "epoch": 12.54, "learning_rate": 6.452087637664993e-06, "loss": 0.0476, "step": 40770 }, { "epoch": 12.54, "learning_rate": 6.4516217867059615e-06, "loss": 0.0098, "step": 40771 }, { "epoch": 12.55, "learning_rate": 6.451155944556598e-06, "loss": 0.0365, "step": 40772 }, { "epoch": 12.55, "learning_rate": 6.450690111218058e-06, "loss": 0.0485, "step": 40773 }, { "epoch": 12.55, "learning_rate": 6.450224286691496e-06, "loss": 0.0313, "step": 40774 }, { "epoch": 12.55, "learning_rate": 6.4497584709780715e-06, "loss": 0.0397, "step": 40775 }, { "epoch": 12.55, "learning_rate": 6.449292664078939e-06, "loss": 0.0373, "step": 40776 }, { "epoch": 12.55, "learning_rate": 6.448826865995255e-06, "loss": 0.0221, "step": 40777 }, { "epoch": 12.55, "learning_rate": 6.448361076728178e-06, "loss": 0.0433, "step": 40778 }, { "epoch": 12.55, "learning_rate": 6.447895296278864e-06, "loss": 0.0193, "step": 40779 }, { "epoch": 12.55, "learning_rate": 6.4474295246484675e-06, "loss": 0.0401, "step": 40780 }, { "epoch": 12.55, "learning_rate": 6.446963761838141e-06, "loss": 0.0309, "step": 40781 }, { "epoch": 12.55, "learning_rate": 6.4464980078490515e-06, "loss": 0.0267, "step": 40782 }, { "epoch": 12.55, "learning_rate": 6.446032262682347e-06, "loss": 0.1444, "step": 40783 }, { "epoch": 12.55, "learning_rate": 6.445566526339187e-06, "loss": 0.0345, "step": 40784 }, { "epoch": 12.55, "learning_rate": 6.445100798820725e-06, "loss": 0.0256, "step": 40785 }, { "epoch": 12.55, "learning_rate": 6.444635080128121e-06, "loss": 0.025, "step": 40786 }, { "epoch": 12.55, "learning_rate": 6.444169370262531e-06, "loss": 0.0296, "step": 40787 }, { "epoch": 12.55, "learning_rate": 6.443703669225109e-06, "loss": 0.0361, "step": 40788 }, { "epoch": 12.55, "learning_rate": 6.443237977017014e-06, "loss": 0.03, "step": 40789 }, { "epoch": 12.55, "learning_rate": 6.4427722936393936e-06, "loss": 0.0642, "step": 40790 }, { "epoch": 12.55, "learning_rate": 6.442306619093415e-06, "loss": 0.0342, "step": 40791 }, { "epoch": 12.55, "learning_rate": 6.441840953380232e-06, "loss": 0.0423, "step": 40792 }, { "epoch": 12.55, "learning_rate": 6.441375296500997e-06, "loss": 0.0518, "step": 40793 }, { "epoch": 12.55, "learning_rate": 6.4409096484568674e-06, "loss": 0.0362, "step": 40794 }, { "epoch": 12.55, "learning_rate": 6.440444009248997e-06, "loss": 0.0319, "step": 40795 }, { "epoch": 12.55, "learning_rate": 6.439978378878547e-06, "loss": 0.025, "step": 40796 }, { "epoch": 12.55, "learning_rate": 6.439512757346671e-06, "loss": 0.0437, "step": 40797 }, { "epoch": 12.55, "learning_rate": 6.439047144654526e-06, "loss": 0.0284, "step": 40798 }, { "epoch": 12.55, "learning_rate": 6.438581540803266e-06, "loss": 0.0223, "step": 40799 }, { "epoch": 12.55, "learning_rate": 6.438115945794046e-06, "loss": 0.018, "step": 40800 }, { "epoch": 12.55, "learning_rate": 6.437650359628025e-06, "loss": 0.0225, "step": 40801 }, { "epoch": 12.55, "learning_rate": 6.437184782306358e-06, "loss": 0.0237, "step": 40802 }, { "epoch": 12.55, "learning_rate": 6.436719213830204e-06, "loss": 0.0347, "step": 40803 }, { "epoch": 12.55, "learning_rate": 6.436253654200712e-06, "loss": 0.0524, "step": 40804 }, { "epoch": 12.56, "learning_rate": 6.435788103419041e-06, "loss": 0.0306, "step": 40805 }, { "epoch": 12.56, "learning_rate": 6.435322561486348e-06, "loss": 0.0256, "step": 40806 }, { "epoch": 12.56, "learning_rate": 6.4348570284037895e-06, "loss": 0.0191, "step": 40807 }, { "epoch": 12.56, "learning_rate": 6.434391504172518e-06, "loss": 0.0228, "step": 40808 }, { "epoch": 12.56, "learning_rate": 6.433925988793692e-06, "loss": 0.0489, "step": 40809 }, { "epoch": 12.56, "learning_rate": 6.4334604822684645e-06, "loss": 0.0288, "step": 40810 }, { "epoch": 12.56, "learning_rate": 6.432994984597996e-06, "loss": 0.0392, "step": 40811 }, { "epoch": 12.56, "learning_rate": 6.4325294957834375e-06, "loss": 0.0435, "step": 40812 }, { "epoch": 12.56, "learning_rate": 6.43206401582595e-06, "loss": 0.0421, "step": 40813 }, { "epoch": 12.56, "learning_rate": 6.4315985447266825e-06, "loss": 0.0185, "step": 40814 }, { "epoch": 12.56, "learning_rate": 6.431133082486789e-06, "loss": 0.0447, "step": 40815 }, { "epoch": 12.56, "learning_rate": 6.430667629107439e-06, "loss": 0.0435, "step": 40816 }, { "epoch": 12.56, "learning_rate": 6.430202184589775e-06, "loss": 0.0266, "step": 40817 }, { "epoch": 12.56, "learning_rate": 6.429736748934956e-06, "loss": 0.0225, "step": 40818 }, { "epoch": 12.56, "learning_rate": 6.429271322144139e-06, "loss": 0.0166, "step": 40819 }, { "epoch": 12.56, "learning_rate": 6.428805904218476e-06, "loss": 0.0345, "step": 40820 }, { "epoch": 12.56, "learning_rate": 6.428340495159129e-06, "loss": 0.011, "step": 40821 }, { "epoch": 12.56, "learning_rate": 6.4278750949672485e-06, "loss": 0.0282, "step": 40822 }, { "epoch": 12.56, "learning_rate": 6.42740970364399e-06, "loss": 0.021, "step": 40823 }, { "epoch": 12.56, "learning_rate": 6.426944321190513e-06, "loss": 0.0311, "step": 40824 }, { "epoch": 12.56, "learning_rate": 6.426478947607965e-06, "loss": 0.0348, "step": 40825 }, { "epoch": 12.56, "learning_rate": 6.426013582897511e-06, "loss": 0.0198, "step": 40826 }, { "epoch": 12.56, "learning_rate": 6.4255482270603006e-06, "loss": 0.0113, "step": 40827 }, { "epoch": 12.56, "learning_rate": 6.42508288009749e-06, "loss": 0.0135, "step": 40828 }, { "epoch": 12.56, "learning_rate": 6.4246175420102345e-06, "loss": 0.0246, "step": 40829 }, { "epoch": 12.56, "learning_rate": 6.424152212799688e-06, "loss": 0.0344, "step": 40830 }, { "epoch": 12.56, "learning_rate": 6.423686892467009e-06, "loss": 0.0251, "step": 40831 }, { "epoch": 12.56, "learning_rate": 6.423221581013353e-06, "loss": 0.0137, "step": 40832 }, { "epoch": 12.56, "learning_rate": 6.422756278439872e-06, "loss": 0.0657, "step": 40833 }, { "epoch": 12.56, "learning_rate": 6.422290984747726e-06, "loss": 0.0435, "step": 40834 }, { "epoch": 12.56, "learning_rate": 6.42182569993806e-06, "loss": 0.0413, "step": 40835 }, { "epoch": 12.56, "learning_rate": 6.421360424012039e-06, "loss": 0.0373, "step": 40836 }, { "epoch": 12.57, "learning_rate": 6.420895156970819e-06, "loss": 0.0313, "step": 40837 }, { "epoch": 12.57, "learning_rate": 6.42042989881555e-06, "loss": 0.0306, "step": 40838 }, { "epoch": 12.57, "learning_rate": 6.419964649547388e-06, "loss": 0.0328, "step": 40839 }, { "epoch": 12.57, "learning_rate": 6.4194994091674855e-06, "loss": 0.0237, "step": 40840 }, { "epoch": 12.57, "learning_rate": 6.4190341776770034e-06, "loss": 0.0488, "step": 40841 }, { "epoch": 12.57, "learning_rate": 6.4185689550770935e-06, "loss": 0.0298, "step": 40842 }, { "epoch": 12.57, "learning_rate": 6.418103741368913e-06, "loss": 0.0399, "step": 40843 }, { "epoch": 12.57, "learning_rate": 6.417638536553617e-06, "loss": 0.0367, "step": 40844 }, { "epoch": 12.57, "learning_rate": 6.41717334063235e-06, "loss": 0.0482, "step": 40845 }, { "epoch": 12.57, "learning_rate": 6.416708153606282e-06, "loss": 0.0325, "step": 40846 }, { "epoch": 12.57, "learning_rate": 6.416242975476563e-06, "loss": 0.0619, "step": 40847 }, { "epoch": 12.57, "learning_rate": 6.415777806244345e-06, "loss": 0.0392, "step": 40848 }, { "epoch": 12.57, "learning_rate": 6.415312645910785e-06, "loss": 0.0338, "step": 40849 }, { "epoch": 12.57, "learning_rate": 6.414847494477034e-06, "loss": 0.037, "step": 40850 }, { "epoch": 12.57, "learning_rate": 6.414382351944253e-06, "loss": 0.0308, "step": 40851 }, { "epoch": 12.57, "learning_rate": 6.413917218313592e-06, "loss": 0.0377, "step": 40852 }, { "epoch": 12.57, "learning_rate": 6.41345209358621e-06, "loss": 0.0419, "step": 40853 }, { "epoch": 12.57, "learning_rate": 6.412986977763259e-06, "loss": 0.0211, "step": 40854 }, { "epoch": 12.57, "learning_rate": 6.412521870845893e-06, "loss": 0.0312, "step": 40855 }, { "epoch": 12.57, "learning_rate": 6.412056772835269e-06, "loss": 0.0298, "step": 40856 }, { "epoch": 12.57, "learning_rate": 6.4115916837325405e-06, "loss": 0.0471, "step": 40857 }, { "epoch": 12.57, "learning_rate": 6.411126603538865e-06, "loss": 0.0356, "step": 40858 }, { "epoch": 12.57, "learning_rate": 6.410661532255393e-06, "loss": 0.0435, "step": 40859 }, { "epoch": 12.57, "learning_rate": 6.4101964698832765e-06, "loss": 0.0222, "step": 40860 }, { "epoch": 12.57, "learning_rate": 6.40973141642368e-06, "loss": 0.0308, "step": 40861 }, { "epoch": 12.57, "learning_rate": 6.409266371877751e-06, "loss": 0.0328, "step": 40862 }, { "epoch": 12.57, "learning_rate": 6.408801336246645e-06, "loss": 0.0328, "step": 40863 }, { "epoch": 12.57, "learning_rate": 6.408336309531518e-06, "loss": 0.0496, "step": 40864 }, { "epoch": 12.57, "learning_rate": 6.407871291733521e-06, "loss": 0.0252, "step": 40865 }, { "epoch": 12.57, "learning_rate": 6.407406282853812e-06, "loss": 0.0441, "step": 40866 }, { "epoch": 12.57, "learning_rate": 6.406941282893546e-06, "loss": 0.0593, "step": 40867 }, { "epoch": 12.57, "learning_rate": 6.406476291853878e-06, "loss": 0.127, "step": 40868 }, { "epoch": 12.57, "learning_rate": 6.40601130973596e-06, "loss": 0.046, "step": 40869 }, { "epoch": 12.58, "learning_rate": 6.405546336540941e-06, "loss": 0.0295, "step": 40870 }, { "epoch": 12.58, "learning_rate": 6.405081372269988e-06, "loss": 0.0266, "step": 40871 }, { "epoch": 12.58, "learning_rate": 6.4046164169242475e-06, "loss": 0.0141, "step": 40872 }, { "epoch": 12.58, "learning_rate": 6.404151470504873e-06, "loss": 0.0253, "step": 40873 }, { "epoch": 12.58, "learning_rate": 6.403686533013023e-06, "loss": 0.03, "step": 40874 }, { "epoch": 12.58, "learning_rate": 6.403221604449847e-06, "loss": 0.0432, "step": 40875 }, { "epoch": 12.58, "learning_rate": 6.402756684816505e-06, "loss": 0.0361, "step": 40876 }, { "epoch": 12.58, "learning_rate": 6.4022917741141475e-06, "loss": 0.0185, "step": 40877 }, { "epoch": 12.58, "learning_rate": 6.4018268723439304e-06, "loss": 0.026, "step": 40878 }, { "epoch": 12.58, "learning_rate": 6.401361979507008e-06, "loss": 0.0289, "step": 40879 }, { "epoch": 12.58, "learning_rate": 6.400897095604528e-06, "loss": 0.0319, "step": 40880 }, { "epoch": 12.58, "learning_rate": 6.400432220637654e-06, "loss": 0.0244, "step": 40881 }, { "epoch": 12.58, "learning_rate": 6.399967354607539e-06, "loss": 0.0508, "step": 40882 }, { "epoch": 12.58, "learning_rate": 6.399502497515333e-06, "loss": 0.0254, "step": 40883 }, { "epoch": 12.58, "learning_rate": 6.399037649362191e-06, "loss": 0.0339, "step": 40884 }, { "epoch": 12.58, "learning_rate": 6.398572810149266e-06, "loss": 0.0277, "step": 40885 }, { "epoch": 12.58, "learning_rate": 6.398107979877714e-06, "loss": 0.03, "step": 40886 }, { "epoch": 12.58, "learning_rate": 6.397643158548692e-06, "loss": 0.0349, "step": 40887 }, { "epoch": 12.58, "learning_rate": 6.397178346163348e-06, "loss": 0.0603, "step": 40888 }, { "epoch": 12.58, "learning_rate": 6.396713542722843e-06, "loss": 0.0284, "step": 40889 }, { "epoch": 12.58, "learning_rate": 6.3962487482283195e-06, "loss": 0.0428, "step": 40890 }, { "epoch": 12.58, "learning_rate": 6.395783962680944e-06, "loss": 0.0466, "step": 40891 }, { "epoch": 12.58, "learning_rate": 6.395319186081866e-06, "loss": 0.0294, "step": 40892 }, { "epoch": 12.58, "learning_rate": 6.394854418432237e-06, "loss": 0.0564, "step": 40893 }, { "epoch": 12.58, "learning_rate": 6.394389659733214e-06, "loss": 0.0328, "step": 40894 }, { "epoch": 12.58, "learning_rate": 6.393924909985945e-06, "loss": 0.0362, "step": 40895 }, { "epoch": 12.58, "learning_rate": 6.393460169191592e-06, "loss": 0.0375, "step": 40896 }, { "epoch": 12.58, "learning_rate": 6.392995437351305e-06, "loss": 0.0294, "step": 40897 }, { "epoch": 12.58, "learning_rate": 6.392530714466236e-06, "loss": 0.0409, "step": 40898 }, { "epoch": 12.58, "learning_rate": 6.392066000537544e-06, "loss": 0.0457, "step": 40899 }, { "epoch": 12.58, "learning_rate": 6.391601295566375e-06, "loss": 0.0226, "step": 40900 }, { "epoch": 12.58, "learning_rate": 6.39113659955389e-06, "loss": 0.0484, "step": 40901 }, { "epoch": 12.59, "learning_rate": 6.39067191250124e-06, "loss": 0.0172, "step": 40902 }, { "epoch": 12.59, "learning_rate": 6.39020723440958e-06, "loss": 0.0268, "step": 40903 }, { "epoch": 12.59, "learning_rate": 6.3897425652800606e-06, "loss": 0.0268, "step": 40904 }, { "epoch": 12.59, "learning_rate": 6.389277905113835e-06, "loss": 0.0255, "step": 40905 }, { "epoch": 12.59, "learning_rate": 6.388813253912061e-06, "loss": 0.0393, "step": 40906 }, { "epoch": 12.59, "learning_rate": 6.38834861167589e-06, "loss": 0.0385, "step": 40907 }, { "epoch": 12.59, "learning_rate": 6.387883978406476e-06, "loss": 0.0291, "step": 40908 }, { "epoch": 12.59, "learning_rate": 6.3874193541049725e-06, "loss": 0.056, "step": 40909 }, { "epoch": 12.59, "learning_rate": 6.38695473877253e-06, "loss": 0.031, "step": 40910 }, { "epoch": 12.59, "learning_rate": 6.386490132410308e-06, "loss": 0.0241, "step": 40911 }, { "epoch": 12.59, "learning_rate": 6.386025535019456e-06, "loss": 0.043, "step": 40912 }, { "epoch": 12.59, "learning_rate": 6.385560946601131e-06, "loss": 0.0199, "step": 40913 }, { "epoch": 12.59, "learning_rate": 6.38509636715648e-06, "loss": 0.0385, "step": 40914 }, { "epoch": 12.59, "learning_rate": 6.384631796686658e-06, "loss": 0.0275, "step": 40915 }, { "epoch": 12.59, "learning_rate": 6.384167235192827e-06, "loss": 0.0555, "step": 40916 }, { "epoch": 12.59, "learning_rate": 6.383702682676129e-06, "loss": 0.0455, "step": 40917 }, { "epoch": 12.59, "learning_rate": 6.3832381391377255e-06, "loss": 0.0518, "step": 40918 }, { "epoch": 12.59, "learning_rate": 6.3827736045787646e-06, "loss": 0.0232, "step": 40919 }, { "epoch": 12.59, "learning_rate": 6.382309079000402e-06, "loss": 0.0449, "step": 40920 }, { "epoch": 12.59, "learning_rate": 6.381844562403789e-06, "loss": 0.024, "step": 40921 }, { "epoch": 12.59, "learning_rate": 6.381380054790084e-06, "loss": 0.0593, "step": 40922 }, { "epoch": 12.59, "learning_rate": 6.380915556160438e-06, "loss": 0.0148, "step": 40923 }, { "epoch": 12.59, "learning_rate": 6.380451066516e-06, "loss": 0.0279, "step": 40924 }, { "epoch": 12.59, "learning_rate": 6.379986585857922e-06, "loss": 0.0251, "step": 40925 }, { "epoch": 12.59, "learning_rate": 6.379522114187369e-06, "loss": 0.0602, "step": 40926 }, { "epoch": 12.59, "learning_rate": 6.3790576515054824e-06, "loss": 0.0433, "step": 40927 }, { "epoch": 12.59, "learning_rate": 6.378593197813421e-06, "loss": 0.0605, "step": 40928 }, { "epoch": 12.59, "learning_rate": 6.378128753112336e-06, "loss": 0.0283, "step": 40929 }, { "epoch": 12.59, "learning_rate": 6.377664317403379e-06, "loss": 0.0318, "step": 40930 }, { "epoch": 12.59, "learning_rate": 6.377199890687706e-06, "loss": 0.0244, "step": 40931 }, { "epoch": 12.59, "learning_rate": 6.376735472966471e-06, "loss": 0.0207, "step": 40932 }, { "epoch": 12.59, "learning_rate": 6.3762710642408245e-06, "loss": 0.0268, "step": 40933 }, { "epoch": 12.59, "learning_rate": 6.375806664511921e-06, "loss": 0.0283, "step": 40934 }, { "epoch": 12.6, "learning_rate": 6.375342273780907e-06, "loss": 0.0455, "step": 40935 }, { "epoch": 12.6, "learning_rate": 6.374877892048944e-06, "loss": 0.0214, "step": 40936 }, { "epoch": 12.6, "learning_rate": 6.374413519317185e-06, "loss": 0.0238, "step": 40937 }, { "epoch": 12.6, "learning_rate": 6.3739491555867785e-06, "loss": 0.0518, "step": 40938 }, { "epoch": 12.6, "learning_rate": 6.373484800858877e-06, "loss": 0.0079, "step": 40939 }, { "epoch": 12.6, "learning_rate": 6.373020455134633e-06, "loss": 0.0272, "step": 40940 }, { "epoch": 12.6, "learning_rate": 6.3725561184152054e-06, "loss": 0.0235, "step": 40941 }, { "epoch": 12.6, "learning_rate": 6.372091790701742e-06, "loss": 0.048, "step": 40942 }, { "epoch": 12.6, "learning_rate": 6.371627471995396e-06, "loss": 0.0294, "step": 40943 }, { "epoch": 12.6, "learning_rate": 6.371163162297322e-06, "loss": 0.0114, "step": 40944 }, { "epoch": 12.6, "learning_rate": 6.370698861608668e-06, "loss": 0.069, "step": 40945 }, { "epoch": 12.6, "learning_rate": 6.3702345699305935e-06, "loss": 0.0247, "step": 40946 }, { "epoch": 12.6, "learning_rate": 6.36977028726425e-06, "loss": 0.0252, "step": 40947 }, { "epoch": 12.6, "learning_rate": 6.369306013610785e-06, "loss": 0.0165, "step": 40948 }, { "epoch": 12.6, "learning_rate": 6.368841748971355e-06, "loss": 0.039, "step": 40949 }, { "epoch": 12.6, "learning_rate": 6.368377493347111e-06, "loss": 0.0328, "step": 40950 }, { "epoch": 12.6, "learning_rate": 6.367913246739208e-06, "loss": 0.0519, "step": 40951 }, { "epoch": 12.6, "learning_rate": 6.367449009148796e-06, "loss": 0.0518, "step": 40952 }, { "epoch": 12.6, "learning_rate": 6.36698478057703e-06, "loss": 0.0322, "step": 40953 }, { "epoch": 12.6, "learning_rate": 6.3665205610250615e-06, "loss": 0.0354, "step": 40954 }, { "epoch": 12.6, "learning_rate": 6.366056350494041e-06, "loss": 0.0356, "step": 40955 }, { "epoch": 12.6, "learning_rate": 6.365592148985124e-06, "loss": 0.0285, "step": 40956 }, { "epoch": 12.6, "learning_rate": 6.365127956499462e-06, "loss": 0.0352, "step": 40957 }, { "epoch": 12.6, "learning_rate": 6.364663773038211e-06, "loss": 0.02, "step": 40958 }, { "epoch": 12.6, "learning_rate": 6.364199598602516e-06, "loss": 0.0466, "step": 40959 }, { "epoch": 12.6, "learning_rate": 6.36373543319353e-06, "loss": 0.0224, "step": 40960 }, { "epoch": 12.6, "learning_rate": 6.363271276812414e-06, "loss": 0.0456, "step": 40961 }, { "epoch": 12.6, "learning_rate": 6.3628071294603135e-06, "loss": 0.0252, "step": 40962 }, { "epoch": 12.6, "learning_rate": 6.362342991138381e-06, "loss": 0.0275, "step": 40963 }, { "epoch": 12.6, "learning_rate": 6.361878861847772e-06, "loss": 0.0203, "step": 40964 }, { "epoch": 12.6, "learning_rate": 6.3614147415896355e-06, "loss": 0.0229, "step": 40965 }, { "epoch": 12.6, "learning_rate": 6.360950630365126e-06, "loss": 0.0393, "step": 40966 }, { "epoch": 12.61, "learning_rate": 6.360486528175395e-06, "loss": 0.0225, "step": 40967 }, { "epoch": 12.61, "learning_rate": 6.360022435021597e-06, "loss": 0.0362, "step": 40968 }, { "epoch": 12.61, "learning_rate": 6.359558350904879e-06, "loss": 0.0283, "step": 40969 }, { "epoch": 12.61, "learning_rate": 6.3590942758263945e-06, "loss": 0.0427, "step": 40970 }, { "epoch": 12.61, "learning_rate": 6.358630209787301e-06, "loss": 0.0567, "step": 40971 }, { "epoch": 12.61, "learning_rate": 6.358166152788745e-06, "loss": 0.0305, "step": 40972 }, { "epoch": 12.61, "learning_rate": 6.357702104831882e-06, "loss": 0.0287, "step": 40973 }, { "epoch": 12.61, "learning_rate": 6.357238065917864e-06, "loss": 0.0206, "step": 40974 }, { "epoch": 12.61, "learning_rate": 6.356774036047837e-06, "loss": 0.0321, "step": 40975 }, { "epoch": 12.61, "learning_rate": 6.35631001522296e-06, "loss": 0.0213, "step": 40976 }, { "epoch": 12.61, "learning_rate": 6.355846003444385e-06, "loss": 0.0324, "step": 40977 }, { "epoch": 12.61, "learning_rate": 6.35538200071326e-06, "loss": 0.0202, "step": 40978 }, { "epoch": 12.61, "learning_rate": 6.354918007030743e-06, "loss": 0.0322, "step": 40979 }, { "epoch": 12.61, "learning_rate": 6.354454022397974e-06, "loss": 0.0316, "step": 40980 }, { "epoch": 12.61, "learning_rate": 6.35399004681612e-06, "loss": 0.0354, "step": 40981 }, { "epoch": 12.61, "learning_rate": 6.353526080286323e-06, "loss": 0.0414, "step": 40982 }, { "epoch": 12.61, "learning_rate": 6.353062122809739e-06, "loss": 0.0621, "step": 40983 }, { "epoch": 12.61, "learning_rate": 6.352598174387516e-06, "loss": 0.0359, "step": 40984 }, { "epoch": 12.61, "learning_rate": 6.3521342350208085e-06, "loss": 0.0451, "step": 40985 }, { "epoch": 12.61, "learning_rate": 6.351670304710771e-06, "loss": 0.0316, "step": 40986 }, { "epoch": 12.61, "learning_rate": 6.35120638345855e-06, "loss": 0.0222, "step": 40987 }, { "epoch": 12.61, "learning_rate": 6.350742471265302e-06, "loss": 0.0308, "step": 40988 }, { "epoch": 12.61, "learning_rate": 6.350278568132178e-06, "loss": 0.0323, "step": 40989 }, { "epoch": 12.61, "learning_rate": 6.349814674060322e-06, "loss": 0.031, "step": 40990 }, { "epoch": 12.61, "learning_rate": 6.349350789050896e-06, "loss": 0.0396, "step": 40991 }, { "epoch": 12.61, "learning_rate": 6.3488869131050505e-06, "loss": 0.0364, "step": 40992 }, { "epoch": 12.61, "learning_rate": 6.348423046223933e-06, "loss": 0.0426, "step": 40993 }, { "epoch": 12.61, "learning_rate": 6.347959188408694e-06, "loss": 0.0197, "step": 40994 }, { "epoch": 12.61, "learning_rate": 6.347495339660489e-06, "loss": 0.0481, "step": 40995 }, { "epoch": 12.61, "learning_rate": 6.347031499980468e-06, "loss": 0.0361, "step": 40996 }, { "epoch": 12.61, "learning_rate": 6.346567669369786e-06, "loss": 0.0377, "step": 40997 }, { "epoch": 12.61, "learning_rate": 6.346103847829588e-06, "loss": 0.0463, "step": 40998 }, { "epoch": 12.61, "learning_rate": 6.34564003536103e-06, "loss": 0.0261, "step": 40999 }, { "epoch": 12.62, "learning_rate": 6.345176231965262e-06, "loss": 0.0144, "step": 41000 }, { "epoch": 12.62, "learning_rate": 6.344712437643436e-06, "loss": 0.0253, "step": 41001 }, { "epoch": 12.62, "learning_rate": 6.344248652396707e-06, "loss": 0.0343, "step": 41002 }, { "epoch": 12.62, "learning_rate": 6.343784876226222e-06, "loss": 0.0391, "step": 41003 }, { "epoch": 12.62, "learning_rate": 6.343321109133132e-06, "loss": 0.0533, "step": 41004 }, { "epoch": 12.62, "learning_rate": 6.342857351118588e-06, "loss": 0.0228, "step": 41005 }, { "epoch": 12.62, "learning_rate": 6.342393602183745e-06, "loss": 0.0286, "step": 41006 }, { "epoch": 12.62, "learning_rate": 6.3419298623297545e-06, "loss": 0.012, "step": 41007 }, { "epoch": 12.62, "learning_rate": 6.341466131557764e-06, "loss": 0.0495, "step": 41008 }, { "epoch": 12.62, "learning_rate": 6.3410024098689284e-06, "loss": 0.0516, "step": 41009 }, { "epoch": 12.62, "learning_rate": 6.3405386972643935e-06, "loss": 0.022, "step": 41010 }, { "epoch": 12.62, "learning_rate": 6.340074993745318e-06, "loss": 0.0137, "step": 41011 }, { "epoch": 12.62, "learning_rate": 6.339611299312849e-06, "loss": 0.0247, "step": 41012 }, { "epoch": 12.62, "learning_rate": 6.33914761396814e-06, "loss": 0.0295, "step": 41013 }, { "epoch": 12.62, "learning_rate": 6.338683937712337e-06, "loss": 0.0377, "step": 41014 }, { "epoch": 12.62, "learning_rate": 6.3382202705465935e-06, "loss": 0.018, "step": 41015 }, { "epoch": 12.62, "learning_rate": 6.337756612472067e-06, "loss": 0.0845, "step": 41016 }, { "epoch": 12.62, "learning_rate": 6.337292963489901e-06, "loss": 0.0327, "step": 41017 }, { "epoch": 12.62, "learning_rate": 6.33682932360125e-06, "loss": 0.0316, "step": 41018 }, { "epoch": 12.62, "learning_rate": 6.3363656928072636e-06, "loss": 0.0195, "step": 41019 }, { "epoch": 12.62, "learning_rate": 6.335902071109092e-06, "loss": 0.0222, "step": 41020 }, { "epoch": 12.62, "learning_rate": 6.33543845850789e-06, "loss": 0.0205, "step": 41021 }, { "epoch": 12.62, "learning_rate": 6.3349748550048055e-06, "loss": 0.027, "step": 41022 }, { "epoch": 12.62, "learning_rate": 6.334511260600993e-06, "loss": 0.0378, "step": 41023 }, { "epoch": 12.62, "learning_rate": 6.3340476752975975e-06, "loss": 0.0577, "step": 41024 }, { "epoch": 12.62, "learning_rate": 6.3335840990957705e-06, "loss": 0.0234, "step": 41025 }, { "epoch": 12.62, "learning_rate": 6.333120531996672e-06, "loss": 0.0487, "step": 41026 }, { "epoch": 12.62, "learning_rate": 6.332656974001443e-06, "loss": 0.0283, "step": 41027 }, { "epoch": 12.62, "learning_rate": 6.33219342511124e-06, "loss": 0.0444, "step": 41028 }, { "epoch": 12.62, "learning_rate": 6.331729885327211e-06, "loss": 0.0309, "step": 41029 }, { "epoch": 12.62, "learning_rate": 6.331266354650507e-06, "loss": 0.0234, "step": 41030 }, { "epoch": 12.62, "learning_rate": 6.33080283308228e-06, "loss": 0.0443, "step": 41031 }, { "epoch": 12.63, "learning_rate": 6.330339320623681e-06, "loss": 0.0217, "step": 41032 }, { "epoch": 12.63, "learning_rate": 6.32987581727586e-06, "loss": 0.0231, "step": 41033 }, { "epoch": 12.63, "learning_rate": 6.32941232303997e-06, "loss": 0.0279, "step": 41034 }, { "epoch": 12.63, "learning_rate": 6.328948837917153e-06, "loss": 0.0261, "step": 41035 }, { "epoch": 12.63, "learning_rate": 6.328485361908572e-06, "loss": 0.0285, "step": 41036 }, { "epoch": 12.63, "learning_rate": 6.3280218950153725e-06, "loss": 0.0231, "step": 41037 }, { "epoch": 12.63, "learning_rate": 6.327558437238703e-06, "loss": 0.0411, "step": 41038 }, { "epoch": 12.63, "learning_rate": 6.327094988579716e-06, "loss": 0.0443, "step": 41039 }, { "epoch": 12.63, "learning_rate": 6.32663154903956e-06, "loss": 0.0348, "step": 41040 }, { "epoch": 12.63, "learning_rate": 6.326168118619391e-06, "loss": 0.0206, "step": 41041 }, { "epoch": 12.63, "learning_rate": 6.325704697320354e-06, "loss": 0.0446, "step": 41042 }, { "epoch": 12.63, "learning_rate": 6.3252412851436015e-06, "loss": 0.0229, "step": 41043 }, { "epoch": 12.63, "learning_rate": 6.324777882090287e-06, "loss": 0.0405, "step": 41044 }, { "epoch": 12.63, "learning_rate": 6.3243144881615535e-06, "loss": 0.0442, "step": 41045 }, { "epoch": 12.63, "learning_rate": 6.323851103358558e-06, "loss": 0.0296, "step": 41046 }, { "epoch": 12.63, "learning_rate": 6.323387727682453e-06, "loss": 0.0322, "step": 41047 }, { "epoch": 12.63, "learning_rate": 6.322924361134382e-06, "loss": 0.0179, "step": 41048 }, { "epoch": 12.63, "learning_rate": 6.322461003715499e-06, "loss": 0.0394, "step": 41049 }, { "epoch": 12.63, "learning_rate": 6.3219976554269515e-06, "loss": 0.0419, "step": 41050 }, { "epoch": 12.63, "learning_rate": 6.321534316269895e-06, "loss": 0.0521, "step": 41051 }, { "epoch": 12.63, "learning_rate": 6.321070986245475e-06, "loss": 0.0356, "step": 41052 }, { "epoch": 12.63, "learning_rate": 6.320607665354845e-06, "loss": 0.0224, "step": 41053 }, { "epoch": 12.63, "learning_rate": 6.320144353599156e-06, "loss": 0.0274, "step": 41054 }, { "epoch": 12.63, "learning_rate": 6.319681050979553e-06, "loss": 0.0355, "step": 41055 }, { "epoch": 12.63, "learning_rate": 6.319217757497192e-06, "loss": 0.0276, "step": 41056 }, { "epoch": 12.63, "learning_rate": 6.318754473153221e-06, "loss": 0.0392, "step": 41057 }, { "epoch": 12.63, "learning_rate": 6.3182911979487915e-06, "loss": 0.038, "step": 41058 }, { "epoch": 12.63, "learning_rate": 6.3178279318850525e-06, "loss": 0.0329, "step": 41059 }, { "epoch": 12.63, "learning_rate": 6.317364674963149e-06, "loss": 0.0282, "step": 41060 }, { "epoch": 12.63, "learning_rate": 6.316901427184241e-06, "loss": 0.0234, "step": 41061 }, { "epoch": 12.63, "learning_rate": 6.316438188549472e-06, "loss": 0.0401, "step": 41062 }, { "epoch": 12.63, "learning_rate": 6.315974959059994e-06, "loss": 0.042, "step": 41063 }, { "epoch": 12.63, "learning_rate": 6.315511738716958e-06, "loss": 0.0396, "step": 41064 }, { "epoch": 12.64, "learning_rate": 6.31504852752151e-06, "loss": 0.0459, "step": 41065 }, { "epoch": 12.64, "learning_rate": 6.3145853254748065e-06, "loss": 0.0224, "step": 41066 }, { "epoch": 12.64, "learning_rate": 6.314122132577994e-06, "loss": 0.0459, "step": 41067 }, { "epoch": 12.64, "learning_rate": 6.313658948832223e-06, "loss": 0.0499, "step": 41068 }, { "epoch": 12.64, "learning_rate": 6.3131957742386415e-06, "loss": 0.0247, "step": 41069 }, { "epoch": 12.64, "learning_rate": 6.3127326087983974e-06, "loss": 0.02, "step": 41070 }, { "epoch": 12.64, "learning_rate": 6.312269452512651e-06, "loss": 0.0341, "step": 41071 }, { "epoch": 12.64, "learning_rate": 6.311806305382542e-06, "loss": 0.0608, "step": 41072 }, { "epoch": 12.64, "learning_rate": 6.311343167409224e-06, "loss": 0.0454, "step": 41073 }, { "epoch": 12.64, "learning_rate": 6.310880038593846e-06, "loss": 0.0174, "step": 41074 }, { "epoch": 12.64, "learning_rate": 6.310416918937557e-06, "loss": 0.0221, "step": 41075 }, { "epoch": 12.64, "learning_rate": 6.30995380844151e-06, "loss": 0.0256, "step": 41076 }, { "epoch": 12.64, "learning_rate": 6.309490707106854e-06, "loss": 0.0294, "step": 41077 }, { "epoch": 12.64, "learning_rate": 6.309027614934735e-06, "loss": 0.0169, "step": 41078 }, { "epoch": 12.64, "learning_rate": 6.308564531926309e-06, "loss": 0.0276, "step": 41079 }, { "epoch": 12.64, "learning_rate": 6.308101458082716e-06, "loss": 0.0185, "step": 41080 }, { "epoch": 12.64, "learning_rate": 6.3076383934051174e-06, "loss": 0.0403, "step": 41081 }, { "epoch": 12.64, "learning_rate": 6.307175337894657e-06, "loss": 0.0235, "step": 41082 }, { "epoch": 12.64, "learning_rate": 6.306712291552484e-06, "loss": 0.0268, "step": 41083 }, { "epoch": 12.64, "learning_rate": 6.306249254379746e-06, "loss": 0.0254, "step": 41084 }, { "epoch": 12.64, "learning_rate": 6.305786226377596e-06, "loss": 0.039, "step": 41085 }, { "epoch": 12.64, "learning_rate": 6.3053232075471846e-06, "loss": 0.0285, "step": 41086 }, { "epoch": 12.64, "learning_rate": 6.30486019788966e-06, "loss": 0.0344, "step": 41087 }, { "epoch": 12.64, "learning_rate": 6.304397197406169e-06, "loss": 0.0364, "step": 41088 }, { "epoch": 12.64, "learning_rate": 6.303934206097868e-06, "loss": 0.0353, "step": 41089 }, { "epoch": 12.64, "learning_rate": 6.303471223965894e-06, "loss": 0.0427, "step": 41090 }, { "epoch": 12.64, "learning_rate": 6.303008251011409e-06, "loss": 0.0167, "step": 41091 }, { "epoch": 12.64, "learning_rate": 6.3025452872355606e-06, "loss": 0.0307, "step": 41092 }, { "epoch": 12.64, "learning_rate": 6.302082332639493e-06, "loss": 0.0199, "step": 41093 }, { "epoch": 12.64, "learning_rate": 6.301619387224358e-06, "loss": 0.0419, "step": 41094 }, { "epoch": 12.64, "learning_rate": 6.301156450991304e-06, "loss": 0.0204, "step": 41095 }, { "epoch": 12.64, "learning_rate": 6.300693523941481e-06, "loss": 0.0278, "step": 41096 }, { "epoch": 12.65, "learning_rate": 6.300230606076041e-06, "loss": 0.029, "step": 41097 }, { "epoch": 12.65, "learning_rate": 6.29976769739613e-06, "loss": 0.0288, "step": 41098 }, { "epoch": 12.65, "learning_rate": 6.299304797902898e-06, "loss": 0.0308, "step": 41099 }, { "epoch": 12.65, "learning_rate": 6.298841907597493e-06, "loss": 0.023, "step": 41100 }, { "epoch": 12.65, "learning_rate": 6.298379026481068e-06, "loss": 0.0283, "step": 41101 }, { "epoch": 12.65, "learning_rate": 6.29791615455477e-06, "loss": 0.0366, "step": 41102 }, { "epoch": 12.65, "learning_rate": 6.297453291819749e-06, "loss": 0.0251, "step": 41103 }, { "epoch": 12.65, "learning_rate": 6.296990438277151e-06, "loss": 0.0434, "step": 41104 }, { "epoch": 12.65, "learning_rate": 6.296527593928126e-06, "loss": 0.0384, "step": 41105 }, { "epoch": 12.65, "learning_rate": 6.296064758773828e-06, "loss": 0.041, "step": 41106 }, { "epoch": 12.65, "learning_rate": 6.295601932815401e-06, "loss": 0.0137, "step": 41107 }, { "epoch": 12.65, "learning_rate": 6.295139116053997e-06, "loss": 0.041, "step": 41108 }, { "epoch": 12.65, "learning_rate": 6.2946763084907625e-06, "loss": 0.0368, "step": 41109 }, { "epoch": 12.65, "learning_rate": 6.2942135101268465e-06, "loss": 0.0196, "step": 41110 }, { "epoch": 12.65, "learning_rate": 6.293750720963401e-06, "loss": 0.0469, "step": 41111 }, { "epoch": 12.65, "learning_rate": 6.2932879410015735e-06, "loss": 0.0293, "step": 41112 }, { "epoch": 12.65, "learning_rate": 6.2928251702425154e-06, "loss": 0.0372, "step": 41113 }, { "epoch": 12.65, "learning_rate": 6.292362408687369e-06, "loss": 0.0235, "step": 41114 }, { "epoch": 12.65, "learning_rate": 6.2918996563372854e-06, "loss": 0.0686, "step": 41115 }, { "epoch": 12.65, "learning_rate": 6.291436913193421e-06, "loss": 0.0495, "step": 41116 }, { "epoch": 12.65, "learning_rate": 6.290974179256915e-06, "loss": 0.0215, "step": 41117 }, { "epoch": 12.65, "learning_rate": 6.2905114545289225e-06, "loss": 0.0162, "step": 41118 }, { "epoch": 12.65, "learning_rate": 6.290048739010588e-06, "loss": 0.0426, "step": 41119 }, { "epoch": 12.65, "learning_rate": 6.289586032703062e-06, "loss": 0.0579, "step": 41120 }, { "epoch": 12.65, "learning_rate": 6.289123335607495e-06, "loss": 0.0625, "step": 41121 }, { "epoch": 12.65, "learning_rate": 6.2886606477250345e-06, "loss": 0.0305, "step": 41122 }, { "epoch": 12.65, "learning_rate": 6.288197969056831e-06, "loss": 0.0422, "step": 41123 }, { "epoch": 12.65, "learning_rate": 6.287735299604029e-06, "loss": 0.0355, "step": 41124 }, { "epoch": 12.65, "learning_rate": 6.287272639367775e-06, "loss": 0.0458, "step": 41125 }, { "epoch": 12.65, "learning_rate": 6.28680998834923e-06, "loss": 0.0305, "step": 41126 }, { "epoch": 12.65, "learning_rate": 6.28634734654953e-06, "loss": 0.0337, "step": 41127 }, { "epoch": 12.65, "learning_rate": 6.28588471396983e-06, "loss": 0.0297, "step": 41128 }, { "epoch": 12.65, "learning_rate": 6.285422090611277e-06, "loss": 0.0321, "step": 41129 }, { "epoch": 12.66, "learning_rate": 6.284959476475017e-06, "loss": 0.0413, "step": 41130 }, { "epoch": 12.66, "learning_rate": 6.284496871562204e-06, "loss": 0.0347, "step": 41131 }, { "epoch": 12.66, "learning_rate": 6.284034275873982e-06, "loss": 0.03, "step": 41132 }, { "epoch": 12.66, "learning_rate": 6.2835716894115026e-06, "loss": 0.0387, "step": 41133 }, { "epoch": 12.66, "learning_rate": 6.2831091121759135e-06, "loss": 0.0274, "step": 41134 }, { "epoch": 12.66, "learning_rate": 6.282646544168358e-06, "loss": 0.0155, "step": 41135 }, { "epoch": 12.66, "learning_rate": 6.282183985389991e-06, "loss": 0.0694, "step": 41136 }, { "epoch": 12.66, "learning_rate": 6.281721435841963e-06, "loss": 0.0317, "step": 41137 }, { "epoch": 12.66, "learning_rate": 6.281258895525415e-06, "loss": 0.0287, "step": 41138 }, { "epoch": 12.66, "learning_rate": 6.280796364441499e-06, "loss": 0.0171, "step": 41139 }, { "epoch": 12.66, "learning_rate": 6.2803338425913615e-06, "loss": 0.0751, "step": 41140 }, { "epoch": 12.66, "learning_rate": 6.2803338425913615e-06, "loss": 0.0398, "step": 41141 }, { "epoch": 12.66, "learning_rate": 6.279871329976152e-06, "loss": 0.0535, "step": 41142 }, { "epoch": 12.66, "learning_rate": 6.279408826597022e-06, "loss": 0.0401, "step": 41143 }, { "epoch": 12.66, "learning_rate": 6.278946332455116e-06, "loss": 0.0216, "step": 41144 }, { "epoch": 12.66, "learning_rate": 6.278483847551584e-06, "loss": 0.0396, "step": 41145 }, { "epoch": 12.66, "learning_rate": 6.278021371887568e-06, "loss": 0.0703, "step": 41146 }, { "epoch": 12.66, "learning_rate": 6.277558905464225e-06, "loss": 0.0363, "step": 41147 }, { "epoch": 12.66, "learning_rate": 6.277096448282703e-06, "loss": 0.0465, "step": 41148 }, { "epoch": 12.66, "learning_rate": 6.276634000344144e-06, "loss": 0.0343, "step": 41149 }, { "epoch": 12.66, "learning_rate": 6.276171561649699e-06, "loss": 0.014, "step": 41150 }, { "epoch": 12.66, "learning_rate": 6.2757091322005134e-06, "loss": 0.0316, "step": 41151 }, { "epoch": 12.66, "learning_rate": 6.2752467119977415e-06, "loss": 0.0292, "step": 41152 }, { "epoch": 12.66, "learning_rate": 6.274784301042527e-06, "loss": 0.0383, "step": 41153 }, { "epoch": 12.66, "learning_rate": 6.274321899336018e-06, "loss": 0.0337, "step": 41154 }, { "epoch": 12.66, "learning_rate": 6.273859506879365e-06, "loss": 0.0233, "step": 41155 }, { "epoch": 12.66, "learning_rate": 6.273397123673711e-06, "loss": 0.0429, "step": 41156 }, { "epoch": 12.66, "learning_rate": 6.2729347497202095e-06, "loss": 0.0273, "step": 41157 }, { "epoch": 12.66, "learning_rate": 6.272472385020007e-06, "loss": 0.0389, "step": 41158 }, { "epoch": 12.66, "learning_rate": 6.272010029574251e-06, "loss": 0.0583, "step": 41159 }, { "epoch": 12.66, "learning_rate": 6.271547683384089e-06, "loss": 0.0287, "step": 41160 }, { "epoch": 12.66, "learning_rate": 6.271085346450666e-06, "loss": 0.0401, "step": 41161 }, { "epoch": 12.67, "learning_rate": 6.270623018775135e-06, "loss": 0.0307, "step": 41162 }, { "epoch": 12.67, "learning_rate": 6.270160700358642e-06, "loss": 0.0205, "step": 41163 }, { "epoch": 12.67, "learning_rate": 6.269698391202333e-06, "loss": 0.0436, "step": 41164 }, { "epoch": 12.67, "learning_rate": 6.269236091307359e-06, "loss": 0.0337, "step": 41165 }, { "epoch": 12.67, "learning_rate": 6.268773800674863e-06, "loss": 0.037, "step": 41166 }, { "epoch": 12.67, "learning_rate": 6.268311519305999e-06, "loss": 0.0403, "step": 41167 }, { "epoch": 12.67, "learning_rate": 6.267849247201909e-06, "loss": 0.0181, "step": 41168 }, { "epoch": 12.67, "learning_rate": 6.267386984363748e-06, "loss": 0.0459, "step": 41169 }, { "epoch": 12.67, "learning_rate": 6.266924730792654e-06, "loss": 0.0397, "step": 41170 }, { "epoch": 12.67, "learning_rate": 6.266462486489777e-06, "loss": 0.0263, "step": 41171 }, { "epoch": 12.67, "learning_rate": 6.266000251456275e-06, "loss": 0.0447, "step": 41172 }, { "epoch": 12.67, "learning_rate": 6.265538025693283e-06, "loss": 0.0305, "step": 41173 }, { "epoch": 12.67, "learning_rate": 6.265075809201954e-06, "loss": 0.0229, "step": 41174 }, { "epoch": 12.67, "learning_rate": 6.264613601983435e-06, "loss": 0.0292, "step": 41175 }, { "epoch": 12.67, "learning_rate": 6.264151404038871e-06, "loss": 0.0197, "step": 41176 }, { "epoch": 12.67, "learning_rate": 6.263689215369414e-06, "loss": 0.0643, "step": 41177 }, { "epoch": 12.67, "learning_rate": 6.263227035976209e-06, "loss": 0.0441, "step": 41178 }, { "epoch": 12.67, "learning_rate": 6.2627648658604045e-06, "loss": 0.0369, "step": 41179 }, { "epoch": 12.67, "learning_rate": 6.262302705023147e-06, "loss": 0.0261, "step": 41180 }, { "epoch": 12.67, "learning_rate": 6.2618405534655835e-06, "loss": 0.0334, "step": 41181 }, { "epoch": 12.67, "learning_rate": 6.261378411188865e-06, "loss": 0.0323, "step": 41182 }, { "epoch": 12.67, "learning_rate": 6.2609162781941345e-06, "loss": 0.0252, "step": 41183 }, { "epoch": 12.67, "learning_rate": 6.26045415448254e-06, "loss": 0.0483, "step": 41184 }, { "epoch": 12.67, "learning_rate": 6.259992040055229e-06, "loss": 0.0381, "step": 41185 }, { "epoch": 12.67, "learning_rate": 6.259529934913351e-06, "loss": 0.0297, "step": 41186 }, { "epoch": 12.67, "learning_rate": 6.25906783905805e-06, "loss": 0.0344, "step": 41187 }, { "epoch": 12.67, "learning_rate": 6.258605752490477e-06, "loss": 0.0229, "step": 41188 }, { "epoch": 12.67, "learning_rate": 6.258143675211777e-06, "loss": 0.0309, "step": 41189 }, { "epoch": 12.67, "learning_rate": 6.257681607223096e-06, "loss": 0.03, "step": 41190 }, { "epoch": 12.67, "learning_rate": 6.257219548525584e-06, "loss": 0.0411, "step": 41191 }, { "epoch": 12.67, "learning_rate": 6.256757499120387e-06, "loss": 0.0371, "step": 41192 }, { "epoch": 12.67, "learning_rate": 6.256295459008654e-06, "loss": 0.0332, "step": 41193 }, { "epoch": 12.67, "learning_rate": 6.255833428191527e-06, "loss": 0.0059, "step": 41194 }, { "epoch": 12.68, "learning_rate": 6.255371406670153e-06, "loss": 0.0384, "step": 41195 }, { "epoch": 12.68, "learning_rate": 6.2549093944456884e-06, "loss": 0.0365, "step": 41196 }, { "epoch": 12.68, "learning_rate": 6.254447391519271e-06, "loss": 0.0356, "step": 41197 }, { "epoch": 12.68, "learning_rate": 6.253985397892053e-06, "loss": 0.0703, "step": 41198 }, { "epoch": 12.68, "learning_rate": 6.253523413565179e-06, "loss": 0.053, "step": 41199 }, { "epoch": 12.68, "learning_rate": 6.253061438539793e-06, "loss": 0.0377, "step": 41200 }, { "epoch": 12.68, "learning_rate": 6.2525994728170495e-06, "loss": 0.0453, "step": 41201 }, { "epoch": 12.68, "learning_rate": 6.252137516398089e-06, "loss": 0.035, "step": 41202 }, { "epoch": 12.68, "learning_rate": 6.251675569284064e-06, "loss": 0.0295, "step": 41203 }, { "epoch": 12.68, "learning_rate": 6.251213631476114e-06, "loss": 0.0172, "step": 41204 }, { "epoch": 12.68, "learning_rate": 6.250751702975388e-06, "loss": 0.0237, "step": 41205 }, { "epoch": 12.68, "learning_rate": 6.2502897837830405e-06, "loss": 0.0239, "step": 41206 }, { "epoch": 12.68, "learning_rate": 6.249827873900209e-06, "loss": 0.0266, "step": 41207 }, { "epoch": 12.68, "learning_rate": 6.249365973328046e-06, "loss": 0.0343, "step": 41208 }, { "epoch": 12.68, "learning_rate": 6.248904082067695e-06, "loss": 0.0827, "step": 41209 }, { "epoch": 12.68, "learning_rate": 6.248442200120302e-06, "loss": 0.0357, "step": 41210 }, { "epoch": 12.68, "learning_rate": 6.247980327487017e-06, "loss": 0.0402, "step": 41211 }, { "epoch": 12.68, "learning_rate": 6.247518464168986e-06, "loss": 0.0136, "step": 41212 }, { "epoch": 12.68, "learning_rate": 6.247056610167357e-06, "loss": 0.0388, "step": 41213 }, { "epoch": 12.68, "learning_rate": 6.246594765483274e-06, "loss": 0.0404, "step": 41214 }, { "epoch": 12.68, "learning_rate": 6.24613293011788e-06, "loss": 0.0311, "step": 41215 }, { "epoch": 12.68, "learning_rate": 6.2456711040723285e-06, "loss": 0.0252, "step": 41216 }, { "epoch": 12.68, "learning_rate": 6.245209287347767e-06, "loss": 0.0426, "step": 41217 }, { "epoch": 12.68, "learning_rate": 6.244747479945334e-06, "loss": 0.0348, "step": 41218 }, { "epoch": 12.68, "learning_rate": 6.244285681866182e-06, "loss": 0.0411, "step": 41219 }, { "epoch": 12.68, "learning_rate": 6.2438238931114556e-06, "loss": 0.0391, "step": 41220 }, { "epoch": 12.68, "learning_rate": 6.243362113682303e-06, "loss": 0.0175, "step": 41221 }, { "epoch": 12.68, "learning_rate": 6.242900343579869e-06, "loss": 0.0354, "step": 41222 }, { "epoch": 12.68, "learning_rate": 6.242438582805301e-06, "loss": 0.0113, "step": 41223 }, { "epoch": 12.68, "learning_rate": 6.241976831359746e-06, "loss": 0.0386, "step": 41224 }, { "epoch": 12.68, "learning_rate": 6.241515089244343e-06, "loss": 0.0286, "step": 41225 }, { "epoch": 12.68, "learning_rate": 6.2410533564602515e-06, "loss": 0.0228, "step": 41226 }, { "epoch": 12.69, "learning_rate": 6.2405916330086106e-06, "loss": 0.0214, "step": 41227 }, { "epoch": 12.69, "learning_rate": 6.240129918890567e-06, "loss": 0.0516, "step": 41228 }, { "epoch": 12.69, "learning_rate": 6.239668214107266e-06, "loss": 0.0481, "step": 41229 }, { "epoch": 12.69, "learning_rate": 6.239206518659853e-06, "loss": 0.0363, "step": 41230 }, { "epoch": 12.69, "learning_rate": 6.23874483254948e-06, "loss": 0.041, "step": 41231 }, { "epoch": 12.69, "learning_rate": 6.2382831557772875e-06, "loss": 0.028, "step": 41232 }, { "epoch": 12.69, "learning_rate": 6.237821488344425e-06, "loss": 0.0714, "step": 41233 }, { "epoch": 12.69, "learning_rate": 6.237359830252036e-06, "loss": 0.0471, "step": 41234 }, { "epoch": 12.69, "learning_rate": 6.236898181501268e-06, "loss": 0.0229, "step": 41235 }, { "epoch": 12.69, "learning_rate": 6.236436542093267e-06, "loss": 0.0301, "step": 41236 }, { "epoch": 12.69, "learning_rate": 6.235974912029181e-06, "loss": 0.0554, "step": 41237 }, { "epoch": 12.69, "learning_rate": 6.235513291310158e-06, "loss": 0.0396, "step": 41238 }, { "epoch": 12.69, "learning_rate": 6.235051679937336e-06, "loss": 0.038, "step": 41239 }, { "epoch": 12.69, "learning_rate": 6.234590077911865e-06, "loss": 0.0365, "step": 41240 }, { "epoch": 12.69, "learning_rate": 6.234128485234894e-06, "loss": 0.0272, "step": 41241 }, { "epoch": 12.69, "learning_rate": 6.233666901907567e-06, "loss": 0.047, "step": 41242 }, { "epoch": 12.69, "learning_rate": 6.2332053279310275e-06, "loss": 0.0327, "step": 41243 }, { "epoch": 12.69, "learning_rate": 6.232743763306426e-06, "loss": 0.0184, "step": 41244 }, { "epoch": 12.69, "learning_rate": 6.232282208034903e-06, "loss": 0.043, "step": 41245 }, { "epoch": 12.69, "learning_rate": 6.231820662117609e-06, "loss": 0.018, "step": 41246 }, { "epoch": 12.69, "learning_rate": 6.231359125555689e-06, "loss": 0.0231, "step": 41247 }, { "epoch": 12.69, "learning_rate": 6.230897598350292e-06, "loss": 0.0353, "step": 41248 }, { "epoch": 12.69, "learning_rate": 6.230436080502556e-06, "loss": 0.0186, "step": 41249 }, { "epoch": 12.69, "learning_rate": 6.229974572013627e-06, "loss": 0.082, "step": 41250 }, { "epoch": 12.69, "learning_rate": 6.229513072884662e-06, "loss": 0.0254, "step": 41251 }, { "epoch": 12.69, "learning_rate": 6.229051583116796e-06, "loss": 0.031, "step": 41252 }, { "epoch": 12.69, "learning_rate": 6.2285901027111806e-06, "loss": 0.0308, "step": 41253 }, { "epoch": 12.69, "learning_rate": 6.228128631668957e-06, "loss": 0.031, "step": 41254 }, { "epoch": 12.69, "learning_rate": 6.227667169991272e-06, "loss": 0.0432, "step": 41255 }, { "epoch": 12.69, "learning_rate": 6.227205717679274e-06, "loss": 0.0253, "step": 41256 }, { "epoch": 12.69, "learning_rate": 6.2267442747341085e-06, "loss": 0.0225, "step": 41257 }, { "epoch": 12.69, "learning_rate": 6.226282841156919e-06, "loss": 0.025, "step": 41258 }, { "epoch": 12.69, "learning_rate": 6.225821416948854e-06, "loss": 0.0416, "step": 41259 }, { "epoch": 12.7, "learning_rate": 6.2253600021110495e-06, "loss": 0.0469, "step": 41260 }, { "epoch": 12.7, "learning_rate": 6.2248985966446665e-06, "loss": 0.0263, "step": 41261 }, { "epoch": 12.7, "learning_rate": 6.224437200550839e-06, "loss": 0.0432, "step": 41262 }, { "epoch": 12.7, "learning_rate": 6.223975813830716e-06, "loss": 0.0414, "step": 41263 }, { "epoch": 12.7, "learning_rate": 6.223514436485445e-06, "loss": 0.0322, "step": 41264 }, { "epoch": 12.7, "learning_rate": 6.223053068516165e-06, "loss": 0.0222, "step": 41265 }, { "epoch": 12.7, "learning_rate": 6.22259170992403e-06, "loss": 0.0323, "step": 41266 }, { "epoch": 12.7, "learning_rate": 6.222130360710181e-06, "loss": 0.0301, "step": 41267 }, { "epoch": 12.7, "learning_rate": 6.2216690208757625e-06, "loss": 0.0249, "step": 41268 }, { "epoch": 12.7, "learning_rate": 6.2212076904219245e-06, "loss": 0.0283, "step": 41269 }, { "epoch": 12.7, "learning_rate": 6.220746369349804e-06, "loss": 0.0454, "step": 41270 }, { "epoch": 12.7, "learning_rate": 6.2202850576605534e-06, "loss": 0.0377, "step": 41271 }, { "epoch": 12.7, "learning_rate": 6.2198237553553185e-06, "loss": 0.0122, "step": 41272 }, { "epoch": 12.7, "learning_rate": 6.219362462435241e-06, "loss": 0.0401, "step": 41273 }, { "epoch": 12.7, "learning_rate": 6.218901178901466e-06, "loss": 0.0316, "step": 41274 }, { "epoch": 12.7, "learning_rate": 6.2184399047551396e-06, "loss": 0.0189, "step": 41275 }, { "epoch": 12.7, "learning_rate": 6.217978639997408e-06, "loss": 0.0337, "step": 41276 }, { "epoch": 12.7, "learning_rate": 6.217517384629417e-06, "loss": 0.0207, "step": 41277 }, { "epoch": 12.7, "learning_rate": 6.2170561386523095e-06, "loss": 0.0451, "step": 41278 }, { "epoch": 12.7, "learning_rate": 6.216594902067233e-06, "loss": 0.0464, "step": 41279 }, { "epoch": 12.7, "learning_rate": 6.2161336748753286e-06, "loss": 0.0362, "step": 41280 }, { "epoch": 12.7, "learning_rate": 6.215672457077746e-06, "loss": 0.0208, "step": 41281 }, { "epoch": 12.7, "learning_rate": 6.21521124867563e-06, "loss": 0.029, "step": 41282 }, { "epoch": 12.7, "learning_rate": 6.214750049670124e-06, "loss": 0.0367, "step": 41283 }, { "epoch": 12.7, "learning_rate": 6.214288860062372e-06, "loss": 0.0354, "step": 41284 }, { "epoch": 12.7, "learning_rate": 6.213827679853517e-06, "loss": 0.0292, "step": 41285 }, { "epoch": 12.7, "learning_rate": 6.2133665090447115e-06, "loss": 0.0329, "step": 41286 }, { "epoch": 12.7, "learning_rate": 6.212905347637095e-06, "loss": 0.0229, "step": 41287 }, { "epoch": 12.7, "learning_rate": 6.212444195631812e-06, "loss": 0.039, "step": 41288 }, { "epoch": 12.7, "learning_rate": 6.211983053030011e-06, "loss": 0.0473, "step": 41289 }, { "epoch": 12.7, "learning_rate": 6.211521919832833e-06, "loss": 0.033, "step": 41290 }, { "epoch": 12.7, "learning_rate": 6.211060796041426e-06, "loss": 0.052, "step": 41291 }, { "epoch": 12.71, "learning_rate": 6.210599681656933e-06, "loss": 0.0332, "step": 41292 }, { "epoch": 12.71, "learning_rate": 6.210138576680502e-06, "loss": 0.0506, "step": 41293 }, { "epoch": 12.71, "learning_rate": 6.2096774811132745e-06, "loss": 0.0254, "step": 41294 }, { "epoch": 12.71, "learning_rate": 6.20921639495639e-06, "loss": 0.031, "step": 41295 }, { "epoch": 12.71, "learning_rate": 6.208755318211006e-06, "loss": 0.0236, "step": 41296 }, { "epoch": 12.71, "learning_rate": 6.2082942508782576e-06, "loss": 0.0396, "step": 41297 }, { "epoch": 12.71, "learning_rate": 6.207833192959294e-06, "loss": 0.0376, "step": 41298 }, { "epoch": 12.71, "learning_rate": 6.207372144455257e-06, "loss": 0.0363, "step": 41299 }, { "epoch": 12.71, "learning_rate": 6.20691110536729e-06, "loss": 0.0369, "step": 41300 }, { "epoch": 12.71, "learning_rate": 6.206450075696544e-06, "loss": 0.0436, "step": 41301 }, { "epoch": 12.71, "learning_rate": 6.20598905544416e-06, "loss": 0.0313, "step": 41302 }, { "epoch": 12.71, "learning_rate": 6.205528044611282e-06, "loss": 0.051, "step": 41303 }, { "epoch": 12.71, "learning_rate": 6.205067043199054e-06, "loss": 0.0425, "step": 41304 }, { "epoch": 12.71, "learning_rate": 6.204606051208617e-06, "loss": 0.0419, "step": 41305 }, { "epoch": 12.71, "learning_rate": 6.204145068641128e-06, "loss": 0.0352, "step": 41306 }, { "epoch": 12.71, "learning_rate": 6.20368409549772e-06, "loss": 0.0194, "step": 41307 }, { "epoch": 12.71, "learning_rate": 6.203223131779539e-06, "loss": 0.0532, "step": 41308 }, { "epoch": 12.71, "learning_rate": 6.202762177487733e-06, "loss": 0.0466, "step": 41309 }, { "epoch": 12.71, "learning_rate": 6.202301232623444e-06, "loss": 0.0368, "step": 41310 }, { "epoch": 12.71, "learning_rate": 6.201840297187818e-06, "loss": 0.0193, "step": 41311 }, { "epoch": 12.71, "learning_rate": 6.201379371181999e-06, "loss": 0.0205, "step": 41312 }, { "epoch": 12.71, "learning_rate": 6.200918454607131e-06, "loss": 0.0339, "step": 41313 }, { "epoch": 12.71, "learning_rate": 6.20045754746436e-06, "loss": 0.0401, "step": 41314 }, { "epoch": 12.71, "learning_rate": 6.199996649754822e-06, "loss": 0.0267, "step": 41315 }, { "epoch": 12.71, "learning_rate": 6.199535761479674e-06, "loss": 0.0418, "step": 41316 }, { "epoch": 12.71, "learning_rate": 6.1990748826400535e-06, "loss": 0.0321, "step": 41317 }, { "epoch": 12.71, "learning_rate": 6.198614013237105e-06, "loss": 0.0427, "step": 41318 }, { "epoch": 12.71, "learning_rate": 6.198153153271973e-06, "loss": 0.0356, "step": 41319 }, { "epoch": 12.71, "learning_rate": 6.197692302745799e-06, "loss": 0.0244, "step": 41320 }, { "epoch": 12.71, "learning_rate": 6.197231461659732e-06, "loss": 0.0447, "step": 41321 }, { "epoch": 12.71, "learning_rate": 6.196770630014915e-06, "loss": 0.0354, "step": 41322 }, { "epoch": 12.71, "learning_rate": 6.196309807812491e-06, "loss": 0.0191, "step": 41323 }, { "epoch": 12.71, "learning_rate": 6.195848995053605e-06, "loss": 0.0283, "step": 41324 }, { "epoch": 12.72, "learning_rate": 6.1953881917393954e-06, "loss": 0.0276, "step": 41325 }, { "epoch": 12.72, "learning_rate": 6.194927397871016e-06, "loss": 0.0185, "step": 41326 }, { "epoch": 12.72, "learning_rate": 6.1944666134496056e-06, "loss": 0.0087, "step": 41327 }, { "epoch": 12.72, "learning_rate": 6.194005838476309e-06, "loss": 0.0226, "step": 41328 }, { "epoch": 12.72, "learning_rate": 6.193545072952269e-06, "loss": 0.0272, "step": 41329 }, { "epoch": 12.72, "learning_rate": 6.1930843168786285e-06, "loss": 0.0249, "step": 41330 }, { "epoch": 12.72, "learning_rate": 6.192623570256535e-06, "loss": 0.0113, "step": 41331 }, { "epoch": 12.72, "learning_rate": 6.192162833087132e-06, "loss": 0.0514, "step": 41332 }, { "epoch": 12.72, "learning_rate": 6.19170210537156e-06, "loss": 0.0274, "step": 41333 }, { "epoch": 12.72, "learning_rate": 6.191241387110967e-06, "loss": 0.0305, "step": 41334 }, { "epoch": 12.72, "learning_rate": 6.190780678306492e-06, "loss": 0.0247, "step": 41335 }, { "epoch": 12.72, "learning_rate": 6.190319978959283e-06, "loss": 0.0193, "step": 41336 }, { "epoch": 12.72, "learning_rate": 6.189859289070483e-06, "loss": 0.027, "step": 41337 }, { "epoch": 12.72, "learning_rate": 6.189398608641237e-06, "loss": 0.0189, "step": 41338 }, { "epoch": 12.72, "learning_rate": 6.188937937672686e-06, "loss": 0.0229, "step": 41339 }, { "epoch": 12.72, "learning_rate": 6.188477276165971e-06, "loss": 0.0291, "step": 41340 }, { "epoch": 12.72, "learning_rate": 6.188016624122243e-06, "loss": 0.025, "step": 41341 }, { "epoch": 12.72, "learning_rate": 6.187555981542641e-06, "loss": 0.0371, "step": 41342 }, { "epoch": 12.72, "learning_rate": 6.18709534842831e-06, "loss": 0.0344, "step": 41343 }, { "epoch": 12.72, "learning_rate": 6.186634724780394e-06, "loss": 0.0191, "step": 41344 }, { "epoch": 12.72, "learning_rate": 6.186174110600033e-06, "loss": 0.0112, "step": 41345 }, { "epoch": 12.72, "learning_rate": 6.185713505888376e-06, "loss": 0.036, "step": 41346 }, { "epoch": 12.72, "learning_rate": 6.1852529106465635e-06, "loss": 0.0354, "step": 41347 }, { "epoch": 12.72, "learning_rate": 6.184792324875742e-06, "loss": 0.0411, "step": 41348 }, { "epoch": 12.72, "learning_rate": 6.18433174857705e-06, "loss": 0.0256, "step": 41349 }, { "epoch": 12.72, "learning_rate": 6.1838711817516305e-06, "loss": 0.0152, "step": 41350 }, { "epoch": 12.72, "learning_rate": 6.183410624400635e-06, "loss": 0.0328, "step": 41351 }, { "epoch": 12.72, "learning_rate": 6.182950076525201e-06, "loss": 0.0365, "step": 41352 }, { "epoch": 12.72, "learning_rate": 6.182489538126473e-06, "loss": 0.0429, "step": 41353 }, { "epoch": 12.72, "learning_rate": 6.1820290092055935e-06, "loss": 0.036, "step": 41354 }, { "epoch": 12.72, "learning_rate": 6.181568489763705e-06, "loss": 0.0205, "step": 41355 }, { "epoch": 12.72, "learning_rate": 6.181107979801955e-06, "loss": 0.0398, "step": 41356 }, { "epoch": 12.73, "learning_rate": 6.180647479321484e-06, "loss": 0.0365, "step": 41357 }, { "epoch": 12.73, "learning_rate": 6.180186988323436e-06, "loss": 0.0367, "step": 41358 }, { "epoch": 12.73, "learning_rate": 6.179726506808954e-06, "loss": 0.0357, "step": 41359 }, { "epoch": 12.73, "learning_rate": 6.179266034779177e-06, "loss": 0.0279, "step": 41360 }, { "epoch": 12.73, "learning_rate": 6.178805572235258e-06, "loss": 0.0567, "step": 41361 }, { "epoch": 12.73, "learning_rate": 6.178345119178333e-06, "loss": 0.0332, "step": 41362 }, { "epoch": 12.73, "learning_rate": 6.177884675609547e-06, "loss": 0.0261, "step": 41363 }, { "epoch": 12.73, "learning_rate": 6.177424241530042e-06, "loss": 0.0385, "step": 41364 }, { "epoch": 12.73, "learning_rate": 6.176963816940961e-06, "loss": 0.0392, "step": 41365 }, { "epoch": 12.73, "learning_rate": 6.176503401843449e-06, "loss": 0.0473, "step": 41366 }, { "epoch": 12.73, "learning_rate": 6.176042996238649e-06, "loss": 0.0418, "step": 41367 }, { "epoch": 12.73, "learning_rate": 6.175582600127703e-06, "loss": 0.0356, "step": 41368 }, { "epoch": 12.73, "learning_rate": 6.175122213511757e-06, "loss": 0.0478, "step": 41369 }, { "epoch": 12.73, "learning_rate": 6.174661836391945e-06, "loss": 0.0335, "step": 41370 }, { "epoch": 12.73, "learning_rate": 6.174201468769421e-06, "loss": 0.0306, "step": 41371 }, { "epoch": 12.73, "learning_rate": 6.173741110645324e-06, "loss": 0.0213, "step": 41372 }, { "epoch": 12.73, "learning_rate": 6.173280762020795e-06, "loss": 0.0494, "step": 41373 }, { "epoch": 12.73, "learning_rate": 6.172820422896979e-06, "loss": 0.0444, "step": 41374 }, { "epoch": 12.73, "learning_rate": 6.172360093275015e-06, "loss": 0.0553, "step": 41375 }, { "epoch": 12.73, "learning_rate": 6.1718997731560515e-06, "loss": 0.0386, "step": 41376 }, { "epoch": 12.73, "learning_rate": 6.171439462541229e-06, "loss": 0.048, "step": 41377 }, { "epoch": 12.73, "learning_rate": 6.1709791614316895e-06, "loss": 0.0467, "step": 41378 }, { "epoch": 12.73, "learning_rate": 6.170518869828577e-06, "loss": 0.0574, "step": 41379 }, { "epoch": 12.73, "learning_rate": 6.170058587733031e-06, "loss": 0.0262, "step": 41380 }, { "epoch": 12.73, "learning_rate": 6.169598315146199e-06, "loss": 0.0224, "step": 41381 }, { "epoch": 12.73, "learning_rate": 6.169138052069226e-06, "loss": 0.0507, "step": 41382 }, { "epoch": 12.73, "learning_rate": 6.168677798503246e-06, "loss": 0.0343, "step": 41383 }, { "epoch": 12.73, "learning_rate": 6.168217554449407e-06, "loss": 0.0205, "step": 41384 }, { "epoch": 12.73, "learning_rate": 6.167757319908848e-06, "loss": 0.0329, "step": 41385 }, { "epoch": 12.73, "learning_rate": 6.1672970948827185e-06, "loss": 0.0196, "step": 41386 }, { "epoch": 12.73, "learning_rate": 6.1668368793721545e-06, "loss": 0.0373, "step": 41387 }, { "epoch": 12.73, "learning_rate": 6.166376673378302e-06, "loss": 0.0247, "step": 41388 }, { "epoch": 12.73, "learning_rate": 6.1659164769023024e-06, "loss": 0.031, "step": 41389 }, { "epoch": 12.74, "learning_rate": 6.165456289945299e-06, "loss": 0.0304, "step": 41390 }, { "epoch": 12.74, "learning_rate": 6.164996112508434e-06, "loss": 0.0301, "step": 41391 }, { "epoch": 12.74, "learning_rate": 6.1645359445928485e-06, "loss": 0.0415, "step": 41392 }, { "epoch": 12.74, "learning_rate": 6.164075786199691e-06, "loss": 0.0312, "step": 41393 }, { "epoch": 12.74, "learning_rate": 6.163615637330095e-06, "loss": 0.0484, "step": 41394 }, { "epoch": 12.74, "learning_rate": 6.163155497985204e-06, "loss": 0.0288, "step": 41395 }, { "epoch": 12.74, "learning_rate": 6.16269536816617e-06, "loss": 0.0446, "step": 41396 }, { "epoch": 12.74, "learning_rate": 6.162235247874126e-06, "loss": 0.0277, "step": 41397 }, { "epoch": 12.74, "learning_rate": 6.1617751371102176e-06, "loss": 0.0201, "step": 41398 }, { "epoch": 12.74, "learning_rate": 6.161315035875586e-06, "loss": 0.0227, "step": 41399 }, { "epoch": 12.74, "learning_rate": 6.160854944171372e-06, "loss": 0.0244, "step": 41400 }, { "epoch": 12.74, "learning_rate": 6.160394861998724e-06, "loss": 0.0549, "step": 41401 }, { "epoch": 12.74, "learning_rate": 6.15993478935878e-06, "loss": 0.0391, "step": 41402 }, { "epoch": 12.74, "learning_rate": 6.159474726252684e-06, "loss": 0.0437, "step": 41403 }, { "epoch": 12.74, "learning_rate": 6.159014672681576e-06, "loss": 0.0248, "step": 41404 }, { "epoch": 12.74, "learning_rate": 6.1585546286465936e-06, "loss": 0.0269, "step": 41405 }, { "epoch": 12.74, "learning_rate": 6.1580945941488915e-06, "loss": 0.0594, "step": 41406 }, { "epoch": 12.74, "learning_rate": 6.157634569189601e-06, "loss": 0.0168, "step": 41407 }, { "epoch": 12.74, "learning_rate": 6.15717455376987e-06, "loss": 0.051, "step": 41408 }, { "epoch": 12.74, "learning_rate": 6.156714547890838e-06, "loss": 0.0398, "step": 41409 }, { "epoch": 12.74, "learning_rate": 6.156254551553647e-06, "loss": 0.0442, "step": 41410 }, { "epoch": 12.74, "learning_rate": 6.1557945647594385e-06, "loss": 0.017, "step": 41411 }, { "epoch": 12.74, "learning_rate": 6.1553345875093586e-06, "loss": 0.0209, "step": 41412 }, { "epoch": 12.74, "learning_rate": 6.1548746198045454e-06, "loss": 0.0244, "step": 41413 }, { "epoch": 12.74, "learning_rate": 6.1544146616461434e-06, "loss": 0.047, "step": 41414 }, { "epoch": 12.74, "learning_rate": 6.153954713035288e-06, "loss": 0.0383, "step": 41415 }, { "epoch": 12.74, "learning_rate": 6.153494773973129e-06, "loss": 0.0315, "step": 41416 }, { "epoch": 12.74, "learning_rate": 6.153034844460808e-06, "loss": 0.0396, "step": 41417 }, { "epoch": 12.74, "learning_rate": 6.1525749244994626e-06, "loss": 0.0415, "step": 41418 }, { "epoch": 12.74, "learning_rate": 6.1521150140902365e-06, "loss": 0.0309, "step": 41419 }, { "epoch": 12.74, "learning_rate": 6.15165511323427e-06, "loss": 0.0255, "step": 41420 }, { "epoch": 12.74, "learning_rate": 6.151195221932706e-06, "loss": 0.0293, "step": 41421 }, { "epoch": 12.75, "learning_rate": 6.1507353401866896e-06, "loss": 0.015, "step": 41422 }, { "epoch": 12.75, "learning_rate": 6.150275467997358e-06, "loss": 0.0532, "step": 41423 }, { "epoch": 12.75, "learning_rate": 6.149815605365858e-06, "loss": 0.0499, "step": 41424 }, { "epoch": 12.75, "learning_rate": 6.14935575229332e-06, "loss": 0.0331, "step": 41425 }, { "epoch": 12.75, "learning_rate": 6.148895908780899e-06, "loss": 0.0304, "step": 41426 }, { "epoch": 12.75, "learning_rate": 6.148436074829732e-06, "loss": 0.0261, "step": 41427 }, { "epoch": 12.75, "learning_rate": 6.147976250440959e-06, "loss": 0.0453, "step": 41428 }, { "epoch": 12.75, "learning_rate": 6.147516435615721e-06, "loss": 0.0352, "step": 41429 }, { "epoch": 12.75, "learning_rate": 6.147056630355161e-06, "loss": 0.0445, "step": 41430 }, { "epoch": 12.75, "learning_rate": 6.146596834660421e-06, "loss": 0.024, "step": 41431 }, { "epoch": 12.75, "learning_rate": 6.146137048532643e-06, "loss": 0.0283, "step": 41432 }, { "epoch": 12.75, "learning_rate": 6.145677271972969e-06, "loss": 0.0254, "step": 41433 }, { "epoch": 12.75, "learning_rate": 6.145217504982537e-06, "loss": 0.0264, "step": 41434 }, { "epoch": 12.75, "learning_rate": 6.14475774756249e-06, "loss": 0.0463, "step": 41435 }, { "epoch": 12.75, "learning_rate": 6.1442979997139715e-06, "loss": 0.0255, "step": 41436 }, { "epoch": 12.75, "learning_rate": 6.143838261438122e-06, "loss": 0.0548, "step": 41437 }, { "epoch": 12.75, "learning_rate": 6.143378532736084e-06, "loss": 0.0188, "step": 41438 }, { "epoch": 12.75, "learning_rate": 6.142918813608997e-06, "loss": 0.0381, "step": 41439 }, { "epoch": 12.75, "learning_rate": 6.142459104057998e-06, "loss": 0.025, "step": 41440 }, { "epoch": 12.75, "learning_rate": 6.141999404084238e-06, "loss": 0.0324, "step": 41441 }, { "epoch": 12.75, "learning_rate": 6.141539713688851e-06, "loss": 0.0246, "step": 41442 }, { "epoch": 12.75, "learning_rate": 6.141080032872983e-06, "loss": 0.0377, "step": 41443 }, { "epoch": 12.75, "learning_rate": 6.1406203616377715e-06, "loss": 0.0333, "step": 41444 }, { "epoch": 12.75, "learning_rate": 6.140160699984356e-06, "loss": 0.0189, "step": 41445 }, { "epoch": 12.75, "learning_rate": 6.139701047913885e-06, "loss": 0.0266, "step": 41446 }, { "epoch": 12.75, "learning_rate": 6.139241405427495e-06, "loss": 0.0178, "step": 41447 }, { "epoch": 12.75, "learning_rate": 6.13878177252633e-06, "loss": 0.0349, "step": 41448 }, { "epoch": 12.75, "learning_rate": 6.138322149211526e-06, "loss": 0.0514, "step": 41449 }, { "epoch": 12.75, "learning_rate": 6.1378625354842245e-06, "loss": 0.0242, "step": 41450 }, { "epoch": 12.75, "learning_rate": 6.137402931345573e-06, "loss": 0.0195, "step": 41451 }, { "epoch": 12.75, "learning_rate": 6.136943336796708e-06, "loss": 0.0157, "step": 41452 }, { "epoch": 12.75, "learning_rate": 6.136483751838772e-06, "loss": 0.0322, "step": 41453 }, { "epoch": 12.75, "learning_rate": 6.136024176472904e-06, "loss": 0.0409, "step": 41454 }, { "epoch": 12.76, "learning_rate": 6.135564610700244e-06, "loss": 0.0328, "step": 41455 }, { "epoch": 12.76, "learning_rate": 6.135105054521939e-06, "loss": 0.0236, "step": 41456 }, { "epoch": 12.76, "learning_rate": 6.1346455079391254e-06, "loss": 0.0336, "step": 41457 }, { "epoch": 12.76, "learning_rate": 6.134185970952944e-06, "loss": 0.0137, "step": 41458 }, { "epoch": 12.76, "learning_rate": 6.13372644356454e-06, "loss": 0.0326, "step": 41459 }, { "epoch": 12.76, "learning_rate": 6.133266925775043e-06, "loss": 0.0605, "step": 41460 }, { "epoch": 12.76, "learning_rate": 6.13280741758561e-06, "loss": 0.032, "step": 41461 }, { "epoch": 12.76, "learning_rate": 6.13234791899737e-06, "loss": 0.0227, "step": 41462 }, { "epoch": 12.76, "learning_rate": 6.131888430011469e-06, "loss": 0.035, "step": 41463 }, { "epoch": 12.76, "learning_rate": 6.131428950629044e-06, "loss": 0.0469, "step": 41464 }, { "epoch": 12.76, "learning_rate": 6.130969480851239e-06, "loss": 0.036, "step": 41465 }, { "epoch": 12.76, "learning_rate": 6.130510020679194e-06, "loss": 0.0132, "step": 41466 }, { "epoch": 12.76, "learning_rate": 6.130050570114049e-06, "loss": 0.0295, "step": 41467 }, { "epoch": 12.76, "learning_rate": 6.129591129156945e-06, "loss": 0.0588, "step": 41468 }, { "epoch": 12.76, "learning_rate": 6.129131697809025e-06, "loss": 0.0307, "step": 41469 }, { "epoch": 12.76, "learning_rate": 6.128672276071423e-06, "loss": 0.0216, "step": 41470 }, { "epoch": 12.76, "learning_rate": 6.128212863945287e-06, "loss": 0.0359, "step": 41471 }, { "epoch": 12.76, "learning_rate": 6.127753461431756e-06, "loss": 0.0336, "step": 41472 }, { "epoch": 12.76, "learning_rate": 6.127294068531968e-06, "loss": 0.0142, "step": 41473 }, { "epoch": 12.76, "learning_rate": 6.126834685247065e-06, "loss": 0.0347, "step": 41474 }, { "epoch": 12.76, "learning_rate": 6.126375311578183e-06, "loss": 0.0273, "step": 41475 }, { "epoch": 12.76, "learning_rate": 6.125915947526472e-06, "loss": 0.0443, "step": 41476 }, { "epoch": 12.76, "learning_rate": 6.125456593093066e-06, "loss": 0.0276, "step": 41477 }, { "epoch": 12.76, "learning_rate": 6.124997248279107e-06, "loss": 0.0263, "step": 41478 }, { "epoch": 12.76, "learning_rate": 6.124537913085736e-06, "loss": 0.0454, "step": 41479 }, { "epoch": 12.76, "learning_rate": 6.124078587514087e-06, "loss": 0.0288, "step": 41480 }, { "epoch": 12.76, "learning_rate": 6.123619271565311e-06, "loss": 0.0254, "step": 41481 }, { "epoch": 12.76, "learning_rate": 6.123159965240543e-06, "loss": 0.0385, "step": 41482 }, { "epoch": 12.76, "learning_rate": 6.122700668540924e-06, "loss": 0.019, "step": 41483 }, { "epoch": 12.76, "learning_rate": 6.122241381467594e-06, "loss": 0.0303, "step": 41484 }, { "epoch": 12.76, "learning_rate": 6.121782104021691e-06, "loss": 0.0432, "step": 41485 }, { "epoch": 12.76, "learning_rate": 6.121322836204359e-06, "loss": 0.0294, "step": 41486 }, { "epoch": 12.77, "learning_rate": 6.120863578016736e-06, "loss": 0.0469, "step": 41487 }, { "epoch": 12.77, "learning_rate": 6.120404329459963e-06, "loss": 0.0235, "step": 41488 }, { "epoch": 12.77, "learning_rate": 6.119945090535182e-06, "loss": 0.0305, "step": 41489 }, { "epoch": 12.77, "learning_rate": 6.119485861243528e-06, "loss": 0.03, "step": 41490 }, { "epoch": 12.77, "learning_rate": 6.1190266415861465e-06, "loss": 0.0457, "step": 41491 }, { "epoch": 12.77, "learning_rate": 6.118567431564175e-06, "loss": 0.0296, "step": 41492 }, { "epoch": 12.77, "learning_rate": 6.118108231178758e-06, "loss": 0.0605, "step": 41493 }, { "epoch": 12.77, "learning_rate": 6.1176490404310285e-06, "loss": 0.0345, "step": 41494 }, { "epoch": 12.77, "learning_rate": 6.117189859322126e-06, "loss": 0.0347, "step": 41495 }, { "epoch": 12.77, "learning_rate": 6.1167306878532e-06, "loss": 0.036, "step": 41496 }, { "epoch": 12.77, "learning_rate": 6.116271526025383e-06, "loss": 0.0418, "step": 41497 }, { "epoch": 12.77, "learning_rate": 6.115812373839815e-06, "loss": 0.03, "step": 41498 }, { "epoch": 12.77, "learning_rate": 6.115353231297641e-06, "loss": 0.0504, "step": 41499 }, { "epoch": 12.77, "learning_rate": 6.114894098399994e-06, "loss": 0.0218, "step": 41500 }, { "epoch": 12.77, "learning_rate": 6.114434975148019e-06, "loss": 0.0588, "step": 41501 }, { "epoch": 12.77, "learning_rate": 6.113975861542856e-06, "loss": 0.0239, "step": 41502 }, { "epoch": 12.77, "learning_rate": 6.113516757585645e-06, "loss": 0.0313, "step": 41503 }, { "epoch": 12.77, "learning_rate": 6.11305766327752e-06, "loss": 0.0451, "step": 41504 }, { "epoch": 12.77, "learning_rate": 6.112598578619623e-06, "loss": 0.0402, "step": 41505 }, { "epoch": 12.77, "learning_rate": 6.112139503613101e-06, "loss": 0.0506, "step": 41506 }, { "epoch": 12.77, "learning_rate": 6.111680438259087e-06, "loss": 0.0364, "step": 41507 }, { "epoch": 12.77, "learning_rate": 6.111221382558722e-06, "loss": 0.0193, "step": 41508 }, { "epoch": 12.77, "learning_rate": 6.110762336513145e-06, "loss": 0.0293, "step": 41509 }, { "epoch": 12.77, "learning_rate": 6.110303300123496e-06, "loss": 0.0588, "step": 41510 }, { "epoch": 12.77, "learning_rate": 6.109844273390917e-06, "loss": 0.0293, "step": 41511 }, { "epoch": 12.77, "learning_rate": 6.109385256316545e-06, "loss": 0.059, "step": 41512 }, { "epoch": 12.77, "learning_rate": 6.108926248901521e-06, "loss": 0.0262, "step": 41513 }, { "epoch": 12.77, "learning_rate": 6.108467251146986e-06, "loss": 0.0353, "step": 41514 }, { "epoch": 12.77, "learning_rate": 6.108008263054073e-06, "loss": 0.0196, "step": 41515 }, { "epoch": 12.77, "learning_rate": 6.107549284623928e-06, "loss": 0.0307, "step": 41516 }, { "epoch": 12.77, "learning_rate": 6.107090315857693e-06, "loss": 0.0418, "step": 41517 }, { "epoch": 12.77, "learning_rate": 6.1066313567565e-06, "loss": 0.0302, "step": 41518 }, { "epoch": 12.77, "learning_rate": 6.106172407321491e-06, "loss": 0.0255, "step": 41519 }, { "epoch": 12.78, "learning_rate": 6.105713467553807e-06, "loss": 0.0315, "step": 41520 }, { "epoch": 12.78, "learning_rate": 6.105254537454586e-06, "loss": 0.0491, "step": 41521 }, { "epoch": 12.78, "learning_rate": 6.104795617024968e-06, "loss": 0.0446, "step": 41522 }, { "epoch": 12.78, "learning_rate": 6.104336706266094e-06, "loss": 0.0364, "step": 41523 }, { "epoch": 12.78, "learning_rate": 6.103877805179102e-06, "loss": 0.0377, "step": 41524 }, { "epoch": 12.78, "learning_rate": 6.103418913765127e-06, "loss": 0.0139, "step": 41525 }, { "epoch": 12.78, "learning_rate": 6.102960032025313e-06, "loss": 0.0426, "step": 41526 }, { "epoch": 12.78, "learning_rate": 6.102501159960804e-06, "loss": 0.0289, "step": 41527 }, { "epoch": 12.78, "learning_rate": 6.1020422975727304e-06, "loss": 0.0395, "step": 41528 }, { "epoch": 12.78, "learning_rate": 6.101583444862236e-06, "loss": 0.0251, "step": 41529 }, { "epoch": 12.78, "learning_rate": 6.1011246018304545e-06, "loss": 0.0282, "step": 41530 }, { "epoch": 12.78, "learning_rate": 6.100665768478534e-06, "loss": 0.039, "step": 41531 }, { "epoch": 12.78, "learning_rate": 6.100206944807608e-06, "loss": 0.022, "step": 41532 }, { "epoch": 12.78, "learning_rate": 6.0997481308188176e-06, "loss": 0.0268, "step": 41533 }, { "epoch": 12.78, "learning_rate": 6.0992893265133e-06, "loss": 0.0433, "step": 41534 }, { "epoch": 12.78, "learning_rate": 6.098830531892195e-06, "loss": 0.0517, "step": 41535 }, { "epoch": 12.78, "learning_rate": 6.098371746956643e-06, "loss": 0.0482, "step": 41536 }, { "epoch": 12.78, "learning_rate": 6.097912971707784e-06, "loss": 0.0306, "step": 41537 }, { "epoch": 12.78, "learning_rate": 6.097454206146753e-06, "loss": 0.0293, "step": 41538 }, { "epoch": 12.78, "learning_rate": 6.0969954502746916e-06, "loss": 0.0422, "step": 41539 }, { "epoch": 12.78, "learning_rate": 6.096536704092737e-06, "loss": 0.0245, "step": 41540 }, { "epoch": 12.78, "learning_rate": 6.09607796760203e-06, "loss": 0.0238, "step": 41541 }, { "epoch": 12.78, "learning_rate": 6.09561924080371e-06, "loss": 0.0682, "step": 41542 }, { "epoch": 12.78, "learning_rate": 6.095160523698913e-06, "loss": 0.0564, "step": 41543 }, { "epoch": 12.78, "learning_rate": 6.0947018162887816e-06, "loss": 0.0496, "step": 41544 }, { "epoch": 12.78, "learning_rate": 6.09424311857445e-06, "loss": 0.017, "step": 41545 }, { "epoch": 12.78, "learning_rate": 6.093784430557062e-06, "loss": 0.0685, "step": 41546 }, { "epoch": 12.78, "learning_rate": 6.0933257522377545e-06, "loss": 0.0579, "step": 41547 }, { "epoch": 12.78, "learning_rate": 6.092867083617667e-06, "loss": 0.0571, "step": 41548 }, { "epoch": 12.78, "learning_rate": 6.092408424697935e-06, "loss": 0.0221, "step": 41549 }, { "epoch": 12.78, "learning_rate": 6.0919497754796975e-06, "loss": 0.0397, "step": 41550 }, { "epoch": 12.78, "learning_rate": 6.0914911359640994e-06, "loss": 0.0191, "step": 41551 }, { "epoch": 12.79, "learning_rate": 6.091032506152274e-06, "loss": 0.0356, "step": 41552 }, { "epoch": 12.79, "learning_rate": 6.090573886045361e-06, "loss": 0.0469, "step": 41553 }, { "epoch": 12.79, "learning_rate": 6.090115275644498e-06, "loss": 0.0393, "step": 41554 }, { "epoch": 12.79, "learning_rate": 6.089656674950824e-06, "loss": 0.0586, "step": 41555 }, { "epoch": 12.79, "learning_rate": 6.0891980839654795e-06, "loss": 0.036, "step": 41556 }, { "epoch": 12.79, "learning_rate": 6.088739502689603e-06, "loss": 0.0397, "step": 41557 }, { "epoch": 12.79, "learning_rate": 6.088280931124332e-06, "loss": 0.0218, "step": 41558 }, { "epoch": 12.79, "learning_rate": 6.087822369270804e-06, "loss": 0.0434, "step": 41559 }, { "epoch": 12.79, "learning_rate": 6.087363817130154e-06, "loss": 0.0338, "step": 41560 }, { "epoch": 12.79, "learning_rate": 6.086905274703531e-06, "loss": 0.0469, "step": 41561 }, { "epoch": 12.79, "learning_rate": 6.086446741992064e-06, "loss": 0.0152, "step": 41562 }, { "epoch": 12.79, "learning_rate": 6.085988218996897e-06, "loss": 0.0226, "step": 41563 }, { "epoch": 12.79, "learning_rate": 6.085529705719164e-06, "loss": 0.0457, "step": 41564 }, { "epoch": 12.79, "learning_rate": 6.0850712021600044e-06, "loss": 0.0315, "step": 41565 }, { "epoch": 12.79, "learning_rate": 6.0846127083205606e-06, "loss": 0.0329, "step": 41566 }, { "epoch": 12.79, "learning_rate": 6.084154224201966e-06, "loss": 0.0321, "step": 41567 }, { "epoch": 12.79, "learning_rate": 6.083695749805362e-06, "loss": 0.0309, "step": 41568 }, { "epoch": 12.79, "learning_rate": 6.0832372851318865e-06, "loss": 0.0552, "step": 41569 }, { "epoch": 12.79, "learning_rate": 6.082778830182672e-06, "loss": 0.0443, "step": 41570 }, { "epoch": 12.79, "learning_rate": 6.082320384958866e-06, "loss": 0.0432, "step": 41571 }, { "epoch": 12.79, "learning_rate": 6.0818619494616026e-06, "loss": 0.0423, "step": 41572 }, { "epoch": 12.79, "learning_rate": 6.08140352369202e-06, "loss": 0.0119, "step": 41573 }, { "epoch": 12.79, "learning_rate": 6.0809451076512505e-06, "loss": 0.0404, "step": 41574 }, { "epoch": 12.79, "learning_rate": 6.080486701340444e-06, "loss": 0.0317, "step": 41575 }, { "epoch": 12.79, "learning_rate": 6.08002830476073e-06, "loss": 0.0487, "step": 41576 }, { "epoch": 12.79, "learning_rate": 6.079569917913249e-06, "loss": 0.0358, "step": 41577 }, { "epoch": 12.79, "learning_rate": 6.07911154079914e-06, "loss": 0.0251, "step": 41578 }, { "epoch": 12.79, "learning_rate": 6.078653173419537e-06, "loss": 0.0235, "step": 41579 }, { "epoch": 12.79, "learning_rate": 6.078194815775583e-06, "loss": 0.0314, "step": 41580 }, { "epoch": 12.79, "learning_rate": 6.0777364678684134e-06, "loss": 0.0297, "step": 41581 }, { "epoch": 12.79, "learning_rate": 6.0772781296991715e-06, "loss": 0.0278, "step": 41582 }, { "epoch": 12.79, "learning_rate": 6.076819801268985e-06, "loss": 0.0358, "step": 41583 }, { "epoch": 12.79, "learning_rate": 6.076361482578995e-06, "loss": 0.0351, "step": 41584 }, { "epoch": 12.8, "learning_rate": 6.0759031736303485e-06, "loss": 0.0481, "step": 41585 }, { "epoch": 12.8, "learning_rate": 6.075444874424172e-06, "loss": 0.0389, "step": 41586 }, { "epoch": 12.8, "learning_rate": 6.074986584961611e-06, "loss": 0.0251, "step": 41587 }, { "epoch": 12.8, "learning_rate": 6.074528305243798e-06, "loss": 0.0428, "step": 41588 }, { "epoch": 12.8, "learning_rate": 6.07407003527187e-06, "loss": 0.0103, "step": 41589 }, { "epoch": 12.8, "learning_rate": 6.073611775046973e-06, "loss": 0.0279, "step": 41590 }, { "epoch": 12.8, "learning_rate": 6.073153524570236e-06, "loss": 0.0526, "step": 41591 }, { "epoch": 12.8, "learning_rate": 6.072695283842802e-06, "loss": 0.037, "step": 41592 }, { "epoch": 12.8, "learning_rate": 6.072237052865809e-06, "loss": 0.038, "step": 41593 }, { "epoch": 12.8, "learning_rate": 6.0717788316403866e-06, "loss": 0.0295, "step": 41594 }, { "epoch": 12.8, "learning_rate": 6.0713206201676825e-06, "loss": 0.0261, "step": 41595 }, { "epoch": 12.8, "learning_rate": 6.0708624184488305e-06, "loss": 0.0472, "step": 41596 }, { "epoch": 12.8, "learning_rate": 6.070404226484967e-06, "loss": 0.0192, "step": 41597 }, { "epoch": 12.8, "learning_rate": 6.06994604427723e-06, "loss": 0.0303, "step": 41598 }, { "epoch": 12.8, "learning_rate": 6.069487871826756e-06, "loss": 0.0182, "step": 41599 }, { "epoch": 12.8, "learning_rate": 6.069029709134686e-06, "loss": 0.0418, "step": 41600 }, { "epoch": 12.8, "learning_rate": 6.068571556202154e-06, "loss": 0.031, "step": 41601 }, { "epoch": 12.8, "learning_rate": 6.068113413030301e-06, "loss": 0.0316, "step": 41602 }, { "epoch": 12.8, "learning_rate": 6.067655279620264e-06, "loss": 0.0369, "step": 41603 }, { "epoch": 12.8, "learning_rate": 6.067197155973172e-06, "loss": 0.0606, "step": 41604 }, { "epoch": 12.8, "learning_rate": 6.066739042090172e-06, "loss": 0.0373, "step": 41605 }, { "epoch": 12.8, "learning_rate": 6.0662809379724045e-06, "loss": 0.0358, "step": 41606 }, { "epoch": 12.8, "learning_rate": 6.065822843620997e-06, "loss": 0.0577, "step": 41607 }, { "epoch": 12.8, "learning_rate": 6.065364759037089e-06, "loss": 0.0244, "step": 41608 }, { "epoch": 12.8, "learning_rate": 6.064906684221819e-06, "loss": 0.0406, "step": 41609 }, { "epoch": 12.8, "learning_rate": 6.064448619176326e-06, "loss": 0.03, "step": 41610 }, { "epoch": 12.8, "learning_rate": 6.0639905639017475e-06, "loss": 0.0382, "step": 41611 }, { "epoch": 12.8, "learning_rate": 6.0635325183992185e-06, "loss": 0.0125, "step": 41612 }, { "epoch": 12.8, "learning_rate": 6.063074482669878e-06, "loss": 0.0384, "step": 41613 }, { "epoch": 12.8, "learning_rate": 6.062616456714858e-06, "loss": 0.0396, "step": 41614 }, { "epoch": 12.8, "learning_rate": 6.062158440535304e-06, "loss": 0.0214, "step": 41615 }, { "epoch": 12.8, "learning_rate": 6.061700434132349e-06, "loss": 0.0304, "step": 41616 }, { "epoch": 12.81, "learning_rate": 6.061242437507131e-06, "loss": 0.0466, "step": 41617 }, { "epoch": 12.81, "learning_rate": 6.060784450660784e-06, "loss": 0.0287, "step": 41618 }, { "epoch": 12.81, "learning_rate": 6.060326473594445e-06, "loss": 0.0351, "step": 41619 }, { "epoch": 12.81, "learning_rate": 6.059868506309257e-06, "loss": 0.0414, "step": 41620 }, { "epoch": 12.81, "learning_rate": 6.059410548806351e-06, "loss": 0.0364, "step": 41621 }, { "epoch": 12.81, "learning_rate": 6.058952601086867e-06, "loss": 0.053, "step": 41622 }, { "epoch": 12.81, "learning_rate": 6.058494663151943e-06, "loss": 0.0374, "step": 41623 }, { "epoch": 12.81, "learning_rate": 6.05803673500271e-06, "loss": 0.0322, "step": 41624 }, { "epoch": 12.81, "learning_rate": 6.0575788166403125e-06, "loss": 0.0378, "step": 41625 }, { "epoch": 12.81, "learning_rate": 6.057120908065883e-06, "loss": 0.034, "step": 41626 }, { "epoch": 12.81, "learning_rate": 6.056663009280561e-06, "loss": 0.0409, "step": 41627 }, { "epoch": 12.81, "learning_rate": 6.056205120285481e-06, "loss": 0.0435, "step": 41628 }, { "epoch": 12.81, "learning_rate": 6.055747241081775e-06, "loss": 0.0401, "step": 41629 }, { "epoch": 12.81, "learning_rate": 6.0552893716705925e-06, "loss": 0.0242, "step": 41630 }, { "epoch": 12.81, "learning_rate": 6.054831512053061e-06, "loss": 0.0308, "step": 41631 }, { "epoch": 12.81, "learning_rate": 6.05437366223032e-06, "loss": 0.038, "step": 41632 }, { "epoch": 12.81, "learning_rate": 6.053915822203504e-06, "loss": 0.0313, "step": 41633 }, { "epoch": 12.81, "learning_rate": 6.053457991973749e-06, "loss": 0.051, "step": 41634 }, { "epoch": 12.81, "learning_rate": 6.053000171542196e-06, "loss": 0.0457, "step": 41635 }, { "epoch": 12.81, "learning_rate": 6.052542360909981e-06, "loss": 0.0218, "step": 41636 }, { "epoch": 12.81, "learning_rate": 6.0520845600782395e-06, "loss": 0.0186, "step": 41637 }, { "epoch": 12.81, "learning_rate": 6.051626769048105e-06, "loss": 0.0665, "step": 41638 }, { "epoch": 12.81, "learning_rate": 6.051168987820715e-06, "loss": 0.0522, "step": 41639 }, { "epoch": 12.81, "learning_rate": 6.050711216397212e-06, "loss": 0.0146, "step": 41640 }, { "epoch": 12.81, "learning_rate": 6.050253454778728e-06, "loss": 0.0393, "step": 41641 }, { "epoch": 12.81, "learning_rate": 6.049795702966399e-06, "loss": 0.0344, "step": 41642 }, { "epoch": 12.81, "learning_rate": 6.049337960961362e-06, "loss": 0.0627, "step": 41643 }, { "epoch": 12.81, "learning_rate": 6.048880228764753e-06, "loss": 0.0373, "step": 41644 }, { "epoch": 12.81, "learning_rate": 6.0484225063777094e-06, "loss": 0.0301, "step": 41645 }, { "epoch": 12.81, "learning_rate": 6.047964793801369e-06, "loss": 0.0401, "step": 41646 }, { "epoch": 12.81, "learning_rate": 6.047507091036866e-06, "loss": 0.0147, "step": 41647 }, { "epoch": 12.81, "learning_rate": 6.047049398085338e-06, "loss": 0.0333, "step": 41648 }, { "epoch": 12.81, "learning_rate": 6.046591714947917e-06, "loss": 0.031, "step": 41649 }, { "epoch": 12.82, "learning_rate": 6.046134041625746e-06, "loss": 0.0384, "step": 41650 }, { "epoch": 12.82, "learning_rate": 6.04567637811996e-06, "loss": 0.0292, "step": 41651 }, { "epoch": 12.82, "learning_rate": 6.045218724431692e-06, "loss": 0.0283, "step": 41652 }, { "epoch": 12.82, "learning_rate": 6.044761080562078e-06, "loss": 0.0269, "step": 41653 }, { "epoch": 12.82, "learning_rate": 6.044303446512257e-06, "loss": 0.0277, "step": 41654 }, { "epoch": 12.82, "learning_rate": 6.043845822283363e-06, "loss": 0.0519, "step": 41655 }, { "epoch": 12.82, "learning_rate": 6.043388207876535e-06, "loss": 0.0262, "step": 41656 }, { "epoch": 12.82, "learning_rate": 6.042930603292907e-06, "loss": 0.0135, "step": 41657 }, { "epoch": 12.82, "learning_rate": 6.042473008533618e-06, "loss": 0.0397, "step": 41658 }, { "epoch": 12.82, "learning_rate": 6.0420154235997955e-06, "loss": 0.029, "step": 41659 }, { "epoch": 12.82, "learning_rate": 6.0415578484925855e-06, "loss": 0.0424, "step": 41660 }, { "epoch": 12.82, "learning_rate": 6.041100283213122e-06, "loss": 0.0421, "step": 41661 }, { "epoch": 12.82, "learning_rate": 6.040642727762537e-06, "loss": 0.0389, "step": 41662 }, { "epoch": 12.82, "learning_rate": 6.040185182141969e-06, "loss": 0.0421, "step": 41663 }, { "epoch": 12.82, "learning_rate": 6.039727646352551e-06, "loss": 0.0369, "step": 41664 }, { "epoch": 12.82, "learning_rate": 6.039270120395424e-06, "loss": 0.0309, "step": 41665 }, { "epoch": 12.82, "learning_rate": 6.038812604271722e-06, "loss": 0.0355, "step": 41666 }, { "epoch": 12.82, "learning_rate": 6.03835509798258e-06, "loss": 0.0343, "step": 41667 }, { "epoch": 12.82, "learning_rate": 6.037897601529135e-06, "loss": 0.0206, "step": 41668 }, { "epoch": 12.82, "learning_rate": 6.037440114912521e-06, "loss": 0.0226, "step": 41669 }, { "epoch": 12.82, "learning_rate": 6.036982638133875e-06, "loss": 0.0176, "step": 41670 }, { "epoch": 12.82, "learning_rate": 6.036525171194332e-06, "loss": 0.0346, "step": 41671 }, { "epoch": 12.82, "learning_rate": 6.036067714095032e-06, "loss": 0.0357, "step": 41672 }, { "epoch": 12.82, "learning_rate": 6.035610266837104e-06, "loss": 0.0379, "step": 41673 }, { "epoch": 12.82, "learning_rate": 6.035152829421686e-06, "loss": 0.0253, "step": 41674 }, { "epoch": 12.82, "learning_rate": 6.034695401849917e-06, "loss": 0.0531, "step": 41675 }, { "epoch": 12.82, "learning_rate": 6.03423798412293e-06, "loss": 0.0228, "step": 41676 }, { "epoch": 12.82, "learning_rate": 6.033780576241861e-06, "loss": 0.0293, "step": 41677 }, { "epoch": 12.82, "learning_rate": 6.033323178207845e-06, "loss": 0.0253, "step": 41678 }, { "epoch": 12.82, "learning_rate": 6.032865790022015e-06, "loss": 0.0417, "step": 41679 }, { "epoch": 12.82, "learning_rate": 6.032408411685515e-06, "loss": 0.0342, "step": 41680 }, { "epoch": 12.82, "learning_rate": 6.031951043199473e-06, "loss": 0.022, "step": 41681 }, { "epoch": 12.83, "learning_rate": 6.0314936845650296e-06, "loss": 0.0438, "step": 41682 }, { "epoch": 12.83, "learning_rate": 6.0310363357833155e-06, "loss": 0.0445, "step": 41683 }, { "epoch": 12.83, "learning_rate": 6.0305789968554626e-06, "loss": 0.0272, "step": 41684 }, { "epoch": 12.83, "learning_rate": 6.03012166778262e-06, "loss": 0.0318, "step": 41685 }, { "epoch": 12.83, "learning_rate": 6.029664348565912e-06, "loss": 0.0509, "step": 41686 }, { "epoch": 12.83, "learning_rate": 6.0292070392064774e-06, "loss": 0.0389, "step": 41687 }, { "epoch": 12.83, "learning_rate": 6.02874973970545e-06, "loss": 0.039, "step": 41688 }, { "epoch": 12.83, "learning_rate": 6.028292450063966e-06, "loss": 0.0364, "step": 41689 }, { "epoch": 12.83, "learning_rate": 6.027835170283163e-06, "loss": 0.0404, "step": 41690 }, { "epoch": 12.83, "learning_rate": 6.027377900364174e-06, "loss": 0.0239, "step": 41691 }, { "epoch": 12.83, "learning_rate": 6.026920640308135e-06, "loss": 0.0331, "step": 41692 }, { "epoch": 12.83, "learning_rate": 6.026463390116182e-06, "loss": 0.0176, "step": 41693 }, { "epoch": 12.83, "learning_rate": 6.0260061497894455e-06, "loss": 0.0241, "step": 41694 }, { "epoch": 12.83, "learning_rate": 6.0255489193290675e-06, "loss": 0.0571, "step": 41695 }, { "epoch": 12.83, "learning_rate": 6.025091698736179e-06, "loss": 0.0191, "step": 41696 }, { "epoch": 12.83, "learning_rate": 6.0246344880119165e-06, "loss": 0.0447, "step": 41697 }, { "epoch": 12.83, "learning_rate": 6.024177287157415e-06, "loss": 0.0322, "step": 41698 }, { "epoch": 12.83, "learning_rate": 6.0237200961738074e-06, "loss": 0.0321, "step": 41699 }, { "epoch": 12.83, "learning_rate": 6.023262915062232e-06, "loss": 0.0379, "step": 41700 }, { "epoch": 12.83, "learning_rate": 6.022805743823824e-06, "loss": 0.0201, "step": 41701 }, { "epoch": 12.83, "learning_rate": 6.022348582459716e-06, "loss": 0.0372, "step": 41702 }, { "epoch": 12.83, "learning_rate": 6.021891430971047e-06, "loss": 0.0338, "step": 41703 }, { "epoch": 12.83, "learning_rate": 6.021434289358943e-06, "loss": 0.0407, "step": 41704 }, { "epoch": 12.83, "learning_rate": 6.0209771576245475e-06, "loss": 0.0225, "step": 41705 }, { "epoch": 12.83, "learning_rate": 6.020520035768998e-06, "loss": 0.0283, "step": 41706 }, { "epoch": 12.83, "learning_rate": 6.02006292379342e-06, "loss": 0.0496, "step": 41707 }, { "epoch": 12.83, "learning_rate": 6.019605821698953e-06, "loss": 0.0396, "step": 41708 }, { "epoch": 12.83, "learning_rate": 6.01914872948673e-06, "loss": 0.0249, "step": 41709 }, { "epoch": 12.83, "learning_rate": 6.018691647157891e-06, "loss": 0.0369, "step": 41710 }, { "epoch": 12.83, "learning_rate": 6.018234574713565e-06, "loss": 0.0173, "step": 41711 }, { "epoch": 12.83, "learning_rate": 6.01777751215489e-06, "loss": 0.0205, "step": 41712 }, { "epoch": 12.83, "learning_rate": 6.017320459483e-06, "loss": 0.0307, "step": 41713 }, { "epoch": 12.83, "learning_rate": 6.016863416699028e-06, "loss": 0.0502, "step": 41714 }, { "epoch": 12.84, "learning_rate": 6.016406383804112e-06, "loss": 0.0302, "step": 41715 }, { "epoch": 12.84, "learning_rate": 6.015949360799387e-06, "loss": 0.0425, "step": 41716 }, { "epoch": 12.84, "learning_rate": 6.015492347685983e-06, "loss": 0.0416, "step": 41717 }, { "epoch": 12.84, "learning_rate": 6.0150353444650375e-06, "loss": 0.0498, "step": 41718 }, { "epoch": 12.84, "learning_rate": 6.014578351137683e-06, "loss": 0.0359, "step": 41719 }, { "epoch": 12.84, "learning_rate": 6.014121367705059e-06, "loss": 0.0431, "step": 41720 }, { "epoch": 12.84, "learning_rate": 6.013664394168297e-06, "loss": 0.0356, "step": 41721 }, { "epoch": 12.84, "learning_rate": 6.01320743052853e-06, "loss": 0.0335, "step": 41722 }, { "epoch": 12.84, "learning_rate": 6.012750476786895e-06, "loss": 0.0158, "step": 41723 }, { "epoch": 12.84, "learning_rate": 6.012293532944525e-06, "loss": 0.0199, "step": 41724 }, { "epoch": 12.84, "learning_rate": 6.011836599002555e-06, "loss": 0.0306, "step": 41725 }, { "epoch": 12.84, "learning_rate": 6.011379674962121e-06, "loss": 0.0367, "step": 41726 }, { "epoch": 12.84, "learning_rate": 6.010922760824357e-06, "loss": 0.0435, "step": 41727 }, { "epoch": 12.84, "learning_rate": 6.010465856590395e-06, "loss": 0.0283, "step": 41728 }, { "epoch": 12.84, "learning_rate": 6.010008962261367e-06, "loss": 0.036, "step": 41729 }, { "epoch": 12.84, "learning_rate": 6.009552077838417e-06, "loss": 0.0275, "step": 41730 }, { "epoch": 12.84, "learning_rate": 6.00909520332267e-06, "loss": 0.025, "step": 41731 }, { "epoch": 12.84, "learning_rate": 6.008638338715265e-06, "loss": 0.0339, "step": 41732 }, { "epoch": 12.84, "learning_rate": 6.008181484017335e-06, "loss": 0.029, "step": 41733 }, { "epoch": 12.84, "learning_rate": 6.00772463923001e-06, "loss": 0.0302, "step": 41734 }, { "epoch": 12.84, "learning_rate": 6.007267804354433e-06, "loss": 0.0477, "step": 41735 }, { "epoch": 12.84, "learning_rate": 6.006810979391733e-06, "loss": 0.023, "step": 41736 }, { "epoch": 12.84, "learning_rate": 6.006354164343047e-06, "loss": 0.0256, "step": 41737 }, { "epoch": 12.84, "learning_rate": 6.005897359209505e-06, "loss": 0.0379, "step": 41738 }, { "epoch": 12.84, "learning_rate": 6.005440563992239e-06, "loss": 0.0356, "step": 41739 }, { "epoch": 12.84, "learning_rate": 6.004983778692392e-06, "loss": 0.0516, "step": 41740 }, { "epoch": 12.84, "learning_rate": 6.004527003311094e-06, "loss": 0.0245, "step": 41741 }, { "epoch": 12.84, "learning_rate": 6.004070237849475e-06, "loss": 0.0175, "step": 41742 }, { "epoch": 12.84, "learning_rate": 6.003613482308675e-06, "loss": 0.0376, "step": 41743 }, { "epoch": 12.84, "learning_rate": 6.003156736689821e-06, "loss": 0.0453, "step": 41744 }, { "epoch": 12.84, "learning_rate": 6.002700000994055e-06, "loss": 0.0273, "step": 41745 }, { "epoch": 12.84, "learning_rate": 6.002243275222506e-06, "loss": 0.0387, "step": 41746 }, { "epoch": 12.85, "learning_rate": 6.00178655937631e-06, "loss": 0.0405, "step": 41747 }, { "epoch": 12.85, "learning_rate": 6.0013298534566035e-06, "loss": 0.0248, "step": 41748 }, { "epoch": 12.85, "learning_rate": 6.000873157464508e-06, "loss": 0.0408, "step": 41749 }, { "epoch": 12.85, "learning_rate": 6.000416471401171e-06, "loss": 0.0373, "step": 41750 }, { "epoch": 12.85, "learning_rate": 5.999959795267726e-06, "loss": 0.0298, "step": 41751 }, { "epoch": 12.85, "learning_rate": 5.999503129065298e-06, "loss": 0.0316, "step": 41752 }, { "epoch": 12.85, "learning_rate": 5.999046472795027e-06, "loss": 0.0374, "step": 41753 }, { "epoch": 12.85, "learning_rate": 5.998589826458041e-06, "loss": 0.042, "step": 41754 }, { "epoch": 12.85, "learning_rate": 5.998133190055482e-06, "loss": 0.102, "step": 41755 }, { "epoch": 12.85, "learning_rate": 5.997676563588478e-06, "loss": 0.0201, "step": 41756 }, { "epoch": 12.85, "learning_rate": 5.997219947058164e-06, "loss": 0.0312, "step": 41757 }, { "epoch": 12.85, "learning_rate": 5.996763340465677e-06, "loss": 0.0272, "step": 41758 }, { "epoch": 12.85, "learning_rate": 5.9963067438121395e-06, "loss": 0.0233, "step": 41759 }, { "epoch": 12.85, "learning_rate": 5.995850157098698e-06, "loss": 0.023, "step": 41760 }, { "epoch": 12.85, "learning_rate": 5.995393580326483e-06, "loss": 0.0223, "step": 41761 }, { "epoch": 12.85, "learning_rate": 5.994937013496624e-06, "loss": 0.0263, "step": 41762 }, { "epoch": 12.85, "learning_rate": 5.994480456610258e-06, "loss": 0.0288, "step": 41763 }, { "epoch": 12.85, "learning_rate": 5.994023909668514e-06, "loss": 0.0288, "step": 41764 }, { "epoch": 12.85, "learning_rate": 5.993567372672531e-06, "loss": 0.0342, "step": 41765 }, { "epoch": 12.85, "learning_rate": 5.993110845623439e-06, "loss": 0.0459, "step": 41766 }, { "epoch": 12.85, "learning_rate": 5.9926543285223734e-06, "loss": 0.0181, "step": 41767 }, { "epoch": 12.85, "learning_rate": 5.992197821370466e-06, "loss": 0.0475, "step": 41768 }, { "epoch": 12.85, "learning_rate": 5.991741324168849e-06, "loss": 0.017, "step": 41769 }, { "epoch": 12.85, "learning_rate": 5.991284836918662e-06, "loss": 0.0184, "step": 41770 }, { "epoch": 12.85, "learning_rate": 5.990828359621032e-06, "loss": 0.0454, "step": 41771 }, { "epoch": 12.85, "learning_rate": 5.990371892277096e-06, "loss": 0.0317, "step": 41772 }, { "epoch": 12.85, "learning_rate": 5.989915434887985e-06, "loss": 0.0324, "step": 41773 }, { "epoch": 12.85, "learning_rate": 5.98945898745483e-06, "loss": 0.0335, "step": 41774 }, { "epoch": 12.85, "learning_rate": 5.98900254997877e-06, "loss": 0.0295, "step": 41775 }, { "epoch": 12.85, "learning_rate": 5.988546122460934e-06, "loss": 0.0511, "step": 41776 }, { "epoch": 12.85, "learning_rate": 5.988089704902458e-06, "loss": 0.0631, "step": 41777 }, { "epoch": 12.85, "learning_rate": 5.987633297304471e-06, "loss": 0.0294, "step": 41778 }, { "epoch": 12.85, "learning_rate": 5.98717689966811e-06, "loss": 0.0481, "step": 41779 }, { "epoch": 12.86, "learning_rate": 5.986720511994508e-06, "loss": 0.0182, "step": 41780 }, { "epoch": 12.86, "learning_rate": 5.986264134284796e-06, "loss": 0.0524, "step": 41781 }, { "epoch": 12.86, "learning_rate": 5.9858077665401105e-06, "loss": 0.0395, "step": 41782 }, { "epoch": 12.86, "learning_rate": 5.985351408761582e-06, "loss": 0.0322, "step": 41783 }, { "epoch": 12.86, "learning_rate": 5.9848950609503375e-06, "loss": 0.0398, "step": 41784 }, { "epoch": 12.86, "learning_rate": 5.984438723107522e-06, "loss": 0.0249, "step": 41785 }, { "epoch": 12.86, "learning_rate": 5.983982395234261e-06, "loss": 0.0239, "step": 41786 }, { "epoch": 12.86, "learning_rate": 5.98352607733169e-06, "loss": 0.0359, "step": 41787 }, { "epoch": 12.86, "learning_rate": 5.98306976940094e-06, "loss": 0.0379, "step": 41788 }, { "epoch": 12.86, "learning_rate": 5.982613471443143e-06, "loss": 0.0224, "step": 41789 }, { "epoch": 12.86, "learning_rate": 5.982157183459436e-06, "loss": 0.032, "step": 41790 }, { "epoch": 12.86, "learning_rate": 5.9817009054509485e-06, "loss": 0.0263, "step": 41791 }, { "epoch": 12.86, "learning_rate": 5.981244637418815e-06, "loss": 0.0242, "step": 41792 }, { "epoch": 12.86, "learning_rate": 5.980788379364171e-06, "loss": 0.0447, "step": 41793 }, { "epoch": 12.86, "learning_rate": 5.980332131288138e-06, "loss": 0.0171, "step": 41794 }, { "epoch": 12.86, "learning_rate": 5.979875893191863e-06, "loss": 0.0217, "step": 41795 }, { "epoch": 12.86, "learning_rate": 5.979419665076471e-06, "loss": 0.0342, "step": 41796 }, { "epoch": 12.86, "learning_rate": 5.978963446943095e-06, "loss": 0.0509, "step": 41797 }, { "epoch": 12.86, "learning_rate": 5.97850723879287e-06, "loss": 0.0346, "step": 41798 }, { "epoch": 12.86, "learning_rate": 5.9780510406269245e-06, "loss": 0.0562, "step": 41799 }, { "epoch": 12.86, "learning_rate": 5.977594852446397e-06, "loss": 0.0388, "step": 41800 }, { "epoch": 12.86, "learning_rate": 5.977138674252416e-06, "loss": 0.0431, "step": 41801 }, { "epoch": 12.86, "learning_rate": 5.9766825060461164e-06, "loss": 0.0264, "step": 41802 }, { "epoch": 12.86, "learning_rate": 5.97622634782863e-06, "loss": 0.0432, "step": 41803 }, { "epoch": 12.86, "learning_rate": 5.975770199601083e-06, "loss": 0.0165, "step": 41804 }, { "epoch": 12.86, "learning_rate": 5.975314061364619e-06, "loss": 0.0376, "step": 41805 }, { "epoch": 12.86, "learning_rate": 5.974857933120366e-06, "loss": 0.0393, "step": 41806 }, { "epoch": 12.86, "learning_rate": 5.974401814869453e-06, "loss": 0.0357, "step": 41807 }, { "epoch": 12.86, "learning_rate": 5.973945706613016e-06, "loss": 0.0388, "step": 41808 }, { "epoch": 12.86, "learning_rate": 5.973489608352185e-06, "loss": 0.0346, "step": 41809 }, { "epoch": 12.86, "learning_rate": 5.9730335200880955e-06, "loss": 0.046, "step": 41810 }, { "epoch": 12.86, "learning_rate": 5.972577441821878e-06, "loss": 0.0325, "step": 41811 }, { "epoch": 12.87, "learning_rate": 5.972121373554665e-06, "loss": 0.0247, "step": 41812 }, { "epoch": 12.87, "learning_rate": 5.971665315287587e-06, "loss": 0.0287, "step": 41813 }, { "epoch": 12.87, "learning_rate": 5.9712092670217784e-06, "loss": 0.019, "step": 41814 }, { "epoch": 12.87, "learning_rate": 5.970753228758371e-06, "loss": 0.0258, "step": 41815 }, { "epoch": 12.87, "learning_rate": 5.9702972004985e-06, "loss": 0.0275, "step": 41816 }, { "epoch": 12.87, "learning_rate": 5.969841182243293e-06, "loss": 0.0492, "step": 41817 }, { "epoch": 12.87, "learning_rate": 5.969385173993885e-06, "loss": 0.0531, "step": 41818 }, { "epoch": 12.87, "learning_rate": 5.968929175751403e-06, "loss": 0.0505, "step": 41819 }, { "epoch": 12.87, "learning_rate": 5.968473187516986e-06, "loss": 0.0231, "step": 41820 }, { "epoch": 12.87, "learning_rate": 5.968017209291763e-06, "loss": 0.0251, "step": 41821 }, { "epoch": 12.87, "learning_rate": 5.967561241076867e-06, "loss": 0.0452, "step": 41822 }, { "epoch": 12.87, "learning_rate": 5.967105282873428e-06, "loss": 0.0303, "step": 41823 }, { "epoch": 12.87, "learning_rate": 5.9666493346825775e-06, "loss": 0.0626, "step": 41824 }, { "epoch": 12.87, "learning_rate": 5.966193396505452e-06, "loss": 0.0179, "step": 41825 }, { "epoch": 12.87, "learning_rate": 5.96573746834318e-06, "loss": 0.0455, "step": 41826 }, { "epoch": 12.87, "learning_rate": 5.9652815501968975e-06, "loss": 0.0463, "step": 41827 }, { "epoch": 12.87, "learning_rate": 5.964825642067731e-06, "loss": 0.0696, "step": 41828 }, { "epoch": 12.87, "learning_rate": 5.96436974395681e-06, "loss": 0.0404, "step": 41829 }, { "epoch": 12.87, "learning_rate": 5.963913855865278e-06, "loss": 0.0322, "step": 41830 }, { "epoch": 12.87, "learning_rate": 5.963457977794256e-06, "loss": 0.0247, "step": 41831 }, { "epoch": 12.87, "learning_rate": 5.963002109744881e-06, "loss": 0.0339, "step": 41832 }, { "epoch": 12.87, "learning_rate": 5.962546251718282e-06, "loss": 0.0214, "step": 41833 }, { "epoch": 12.87, "learning_rate": 5.962090403715592e-06, "loss": 0.0263, "step": 41834 }, { "epoch": 12.87, "learning_rate": 5.961634565737944e-06, "loss": 0.037, "step": 41835 }, { "epoch": 12.87, "learning_rate": 5.9611787377864684e-06, "loss": 0.0315, "step": 41836 }, { "epoch": 12.87, "learning_rate": 5.960722919862301e-06, "loss": 0.0182, "step": 41837 }, { "epoch": 12.87, "learning_rate": 5.960267111966565e-06, "loss": 0.0862, "step": 41838 }, { "epoch": 12.87, "learning_rate": 5.959811314100394e-06, "loss": 0.0279, "step": 41839 }, { "epoch": 12.87, "learning_rate": 5.959355526264928e-06, "loss": 0.0235, "step": 41840 }, { "epoch": 12.87, "learning_rate": 5.958899748461291e-06, "loss": 0.0375, "step": 41841 }, { "epoch": 12.87, "learning_rate": 5.958443980690617e-06, "loss": 0.033, "step": 41842 }, { "epoch": 12.87, "learning_rate": 5.957988222954036e-06, "loss": 0.0284, "step": 41843 }, { "epoch": 12.87, "learning_rate": 5.957532475252679e-06, "loss": 0.0264, "step": 41844 }, { "epoch": 12.88, "learning_rate": 5.9570767375876815e-06, "loss": 0.0588, "step": 41845 }, { "epoch": 12.88, "learning_rate": 5.956621009960173e-06, "loss": 0.0257, "step": 41846 }, { "epoch": 12.88, "learning_rate": 5.956165292371283e-06, "loss": 0.0198, "step": 41847 }, { "epoch": 12.88, "learning_rate": 5.955709584822147e-06, "loss": 0.0287, "step": 41848 }, { "epoch": 12.88, "learning_rate": 5.955253887313888e-06, "loss": 0.0182, "step": 41849 }, { "epoch": 12.88, "learning_rate": 5.954798199847647e-06, "loss": 0.0327, "step": 41850 }, { "epoch": 12.88, "learning_rate": 5.954342522424553e-06, "loss": 0.0381, "step": 41851 }, { "epoch": 12.88, "learning_rate": 5.953886855045734e-06, "loss": 0.0461, "step": 41852 }, { "epoch": 12.88, "learning_rate": 5.953431197712325e-06, "loss": 0.0226, "step": 41853 }, { "epoch": 12.88, "learning_rate": 5.95297555042545e-06, "loss": 0.0358, "step": 41854 }, { "epoch": 12.88, "learning_rate": 5.952519913186251e-06, "loss": 0.0408, "step": 41855 }, { "epoch": 12.88, "learning_rate": 5.9520642859958525e-06, "loss": 0.039, "step": 41856 }, { "epoch": 12.88, "learning_rate": 5.951608668855387e-06, "loss": 0.0644, "step": 41857 }, { "epoch": 12.88, "learning_rate": 5.951153061765989e-06, "loss": 0.0188, "step": 41858 }, { "epoch": 12.88, "learning_rate": 5.950697464728778e-06, "loss": 0.0472, "step": 41859 }, { "epoch": 12.88, "learning_rate": 5.950241877744899e-06, "loss": 0.0341, "step": 41860 }, { "epoch": 12.88, "learning_rate": 5.9497863008154785e-06, "loss": 0.0373, "step": 41861 }, { "epoch": 12.88, "learning_rate": 5.949330733941647e-06, "loss": 0.0362, "step": 41862 }, { "epoch": 12.88, "learning_rate": 5.948875177124533e-06, "loss": 0.0346, "step": 41863 }, { "epoch": 12.88, "learning_rate": 5.948419630365269e-06, "loss": 0.0295, "step": 41864 }, { "epoch": 12.88, "learning_rate": 5.947964093664988e-06, "loss": 0.0329, "step": 41865 }, { "epoch": 12.88, "learning_rate": 5.94750856702482e-06, "loss": 0.0356, "step": 41866 }, { "epoch": 12.88, "learning_rate": 5.947053050445897e-06, "loss": 0.0428, "step": 41867 }, { "epoch": 12.88, "learning_rate": 5.946597543929347e-06, "loss": 0.0198, "step": 41868 }, { "epoch": 12.88, "learning_rate": 5.946142047476302e-06, "loss": 0.0405, "step": 41869 }, { "epoch": 12.88, "learning_rate": 5.945686561087894e-06, "loss": 0.0443, "step": 41870 }, { "epoch": 12.88, "learning_rate": 5.9452310847652536e-06, "loss": 0.0198, "step": 41871 }, { "epoch": 12.88, "learning_rate": 5.944775618509513e-06, "loss": 0.0198, "step": 41872 }, { "epoch": 12.88, "learning_rate": 5.944320162321801e-06, "loss": 0.0478, "step": 41873 }, { "epoch": 12.88, "learning_rate": 5.943864716203245e-06, "loss": 0.0266, "step": 41874 }, { "epoch": 12.88, "learning_rate": 5.943409280154981e-06, "loss": 0.0393, "step": 41875 }, { "epoch": 12.88, "learning_rate": 5.94295385417814e-06, "loss": 0.0384, "step": 41876 }, { "epoch": 12.89, "learning_rate": 5.942498438273849e-06, "loss": 0.045, "step": 41877 }, { "epoch": 12.89, "learning_rate": 5.942043032443241e-06, "loss": 0.041, "step": 41878 }, { "epoch": 12.89, "learning_rate": 5.941587636687445e-06, "loss": 0.031, "step": 41879 }, { "epoch": 12.89, "learning_rate": 5.9411322510075945e-06, "loss": 0.0407, "step": 41880 }, { "epoch": 12.89, "learning_rate": 5.940676875404818e-06, "loss": 0.044, "step": 41881 }, { "epoch": 12.89, "learning_rate": 5.940221509880249e-06, "loss": 0.0267, "step": 41882 }, { "epoch": 12.89, "learning_rate": 5.9397661544350125e-06, "loss": 0.0249, "step": 41883 }, { "epoch": 12.89, "learning_rate": 5.939310809070237e-06, "loss": 0.017, "step": 41884 }, { "epoch": 12.89, "learning_rate": 5.938855473787067e-06, "loss": 0.0298, "step": 41885 }, { "epoch": 12.89, "learning_rate": 5.93840014858662e-06, "loss": 0.0302, "step": 41886 }, { "epoch": 12.89, "learning_rate": 5.937944833470032e-06, "loss": 0.0308, "step": 41887 }, { "epoch": 12.89, "learning_rate": 5.9374895284384315e-06, "loss": 0.0357, "step": 41888 }, { "epoch": 12.89, "learning_rate": 5.937034233492946e-06, "loss": 0.0279, "step": 41889 }, { "epoch": 12.89, "learning_rate": 5.936578948634713e-06, "loss": 0.0348, "step": 41890 }, { "epoch": 12.89, "learning_rate": 5.9361236738648575e-06, "loss": 0.0346, "step": 41891 }, { "epoch": 12.89, "learning_rate": 5.9356684091845116e-06, "loss": 0.0218, "step": 41892 }, { "epoch": 12.89, "learning_rate": 5.935213154594808e-06, "loss": 0.024, "step": 41893 }, { "epoch": 12.89, "learning_rate": 5.934757910096867e-06, "loss": 0.0285, "step": 41894 }, { "epoch": 12.89, "learning_rate": 5.934302675691833e-06, "loss": 0.0477, "step": 41895 }, { "epoch": 12.89, "learning_rate": 5.933847451380828e-06, "loss": 0.0502, "step": 41896 }, { "epoch": 12.89, "learning_rate": 5.933392237164983e-06, "loss": 0.0335, "step": 41897 }, { "epoch": 12.89, "learning_rate": 5.9329370330454285e-06, "loss": 0.0296, "step": 41898 }, { "epoch": 12.89, "learning_rate": 5.932481839023292e-06, "loss": 0.0186, "step": 41899 }, { "epoch": 12.89, "learning_rate": 5.932026655099709e-06, "loss": 0.0217, "step": 41900 }, { "epoch": 12.89, "learning_rate": 5.931571481275809e-06, "loss": 0.0234, "step": 41901 }, { "epoch": 12.89, "learning_rate": 5.931116317552718e-06, "loss": 0.0254, "step": 41902 }, { "epoch": 12.89, "learning_rate": 5.930661163931572e-06, "loss": 0.0149, "step": 41903 }, { "epoch": 12.89, "learning_rate": 5.93020602041349e-06, "loss": 0.0227, "step": 41904 }, { "epoch": 12.89, "learning_rate": 5.929750886999614e-06, "loss": 0.0429, "step": 41905 }, { "epoch": 12.89, "learning_rate": 5.929295763691072e-06, "loss": 0.0507, "step": 41906 }, { "epoch": 12.89, "learning_rate": 5.928840650488988e-06, "loss": 0.0226, "step": 41907 }, { "epoch": 12.89, "learning_rate": 5.9283855473944955e-06, "loss": 0.0201, "step": 41908 }, { "epoch": 12.89, "learning_rate": 5.9279304544087215e-06, "loss": 0.0412, "step": 41909 }, { "epoch": 12.9, "learning_rate": 5.927475371532801e-06, "loss": 0.0299, "step": 41910 }, { "epoch": 12.9, "learning_rate": 5.927020298767861e-06, "loss": 0.0236, "step": 41911 }, { "epoch": 12.9, "learning_rate": 5.92656523611503e-06, "loss": 0.0187, "step": 41912 }, { "epoch": 12.9, "learning_rate": 5.926110183575441e-06, "loss": 0.0223, "step": 41913 }, { "epoch": 12.9, "learning_rate": 5.925655141150219e-06, "loss": 0.0156, "step": 41914 }, { "epoch": 12.9, "learning_rate": 5.925200108840501e-06, "loss": 0.0354, "step": 41915 }, { "epoch": 12.9, "learning_rate": 5.924745086647413e-06, "loss": 0.0203, "step": 41916 }, { "epoch": 12.9, "learning_rate": 5.924290074572083e-06, "loss": 0.0242, "step": 41917 }, { "epoch": 12.9, "learning_rate": 5.923835072615642e-06, "loss": 0.0133, "step": 41918 }, { "epoch": 12.9, "learning_rate": 5.9233800807792174e-06, "loss": 0.0256, "step": 41919 }, { "epoch": 12.9, "learning_rate": 5.922925099063943e-06, "loss": 0.0243, "step": 41920 }, { "epoch": 12.9, "learning_rate": 5.9224701274709454e-06, "loss": 0.021, "step": 41921 }, { "epoch": 12.9, "learning_rate": 5.922015166001356e-06, "loss": 0.0323, "step": 41922 }, { "epoch": 12.9, "learning_rate": 5.921560214656303e-06, "loss": 0.0178, "step": 41923 }, { "epoch": 12.9, "learning_rate": 5.921105273436916e-06, "loss": 0.0519, "step": 41924 }, { "epoch": 12.9, "learning_rate": 5.920650342344326e-06, "loss": 0.0208, "step": 41925 }, { "epoch": 12.9, "learning_rate": 5.920195421379662e-06, "loss": 0.0387, "step": 41926 }, { "epoch": 12.9, "learning_rate": 5.919740510544054e-06, "loss": 0.086, "step": 41927 }, { "epoch": 12.9, "learning_rate": 5.9192856098386295e-06, "loss": 0.0394, "step": 41928 }, { "epoch": 12.9, "learning_rate": 5.918830719264514e-06, "loss": 0.0341, "step": 41929 }, { "epoch": 12.9, "learning_rate": 5.918375838822847e-06, "loss": 0.0396, "step": 41930 }, { "epoch": 12.9, "learning_rate": 5.9179209685147525e-06, "loss": 0.0423, "step": 41931 }, { "epoch": 12.9, "learning_rate": 5.917466108341357e-06, "loss": 0.0186, "step": 41932 }, { "epoch": 12.9, "learning_rate": 5.917011258303793e-06, "loss": 0.0336, "step": 41933 }, { "epoch": 12.9, "learning_rate": 5.916556418403189e-06, "loss": 0.0358, "step": 41934 }, { "epoch": 12.9, "learning_rate": 5.916101588640675e-06, "loss": 0.0599, "step": 41935 }, { "epoch": 12.9, "learning_rate": 5.9156467690173806e-06, "loss": 0.0333, "step": 41936 }, { "epoch": 12.9, "learning_rate": 5.915191959534436e-06, "loss": 0.0314, "step": 41937 }, { "epoch": 12.9, "learning_rate": 5.914737160192967e-06, "loss": 0.0257, "step": 41938 }, { "epoch": 12.9, "learning_rate": 5.9142823709941e-06, "loss": 0.0189, "step": 41939 }, { "epoch": 12.9, "learning_rate": 5.913827591938974e-06, "loss": 0.0313, "step": 41940 }, { "epoch": 12.9, "learning_rate": 5.91337282302871e-06, "loss": 0.0435, "step": 41941 }, { "epoch": 12.91, "learning_rate": 5.912918064264441e-06, "loss": 0.0496, "step": 41942 }, { "epoch": 12.91, "learning_rate": 5.912463315647293e-06, "loss": 0.0336, "step": 41943 }, { "epoch": 12.91, "learning_rate": 5.9120085771783966e-06, "loss": 0.0436, "step": 41944 }, { "epoch": 12.91, "learning_rate": 5.911553848858881e-06, "loss": 0.0292, "step": 41945 }, { "epoch": 12.91, "learning_rate": 5.9110991306898745e-06, "loss": 0.0297, "step": 41946 }, { "epoch": 12.91, "learning_rate": 5.910644422672509e-06, "loss": 0.0493, "step": 41947 }, { "epoch": 12.91, "learning_rate": 5.910189724807911e-06, "loss": 0.0574, "step": 41948 }, { "epoch": 12.91, "learning_rate": 5.909735037097204e-06, "loss": 0.0248, "step": 41949 }, { "epoch": 12.91, "learning_rate": 5.909280359541526e-06, "loss": 0.0414, "step": 41950 }, { "epoch": 12.91, "learning_rate": 5.908825692142004e-06, "loss": 0.0416, "step": 41951 }, { "epoch": 12.91, "learning_rate": 5.908371034899764e-06, "loss": 0.0452, "step": 41952 }, { "epoch": 12.91, "learning_rate": 5.907916387815934e-06, "loss": 0.0348, "step": 41953 }, { "epoch": 12.91, "learning_rate": 5.907461750891643e-06, "loss": 0.0399, "step": 41954 }, { "epoch": 12.91, "learning_rate": 5.9070071241280235e-06, "loss": 0.0322, "step": 41955 }, { "epoch": 12.91, "learning_rate": 5.906552507526202e-06, "loss": 0.0526, "step": 41956 }, { "epoch": 12.91, "learning_rate": 5.906097901087306e-06, "loss": 0.0353, "step": 41957 }, { "epoch": 12.91, "learning_rate": 5.905643304812468e-06, "loss": 0.0225, "step": 41958 }, { "epoch": 12.91, "learning_rate": 5.9051887187028076e-06, "loss": 0.0244, "step": 41959 }, { "epoch": 12.91, "learning_rate": 5.904734142759465e-06, "loss": 0.0441, "step": 41960 }, { "epoch": 12.91, "learning_rate": 5.904279576983564e-06, "loss": 0.0415, "step": 41961 }, { "epoch": 12.91, "learning_rate": 5.90382502137623e-06, "loss": 0.0394, "step": 41962 }, { "epoch": 12.91, "learning_rate": 5.903370475938595e-06, "loss": 0.0302, "step": 41963 }, { "epoch": 12.91, "learning_rate": 5.902915940671785e-06, "loss": 0.0403, "step": 41964 }, { "epoch": 12.91, "learning_rate": 5.902461415576932e-06, "loss": 0.0303, "step": 41965 }, { "epoch": 12.91, "learning_rate": 5.9020069006551635e-06, "loss": 0.0282, "step": 41966 }, { "epoch": 12.91, "learning_rate": 5.901552395907606e-06, "loss": 0.0153, "step": 41967 }, { "epoch": 12.91, "learning_rate": 5.901097901335388e-06, "loss": 0.0234, "step": 41968 }, { "epoch": 12.91, "learning_rate": 5.90064341693964e-06, "loss": 0.029, "step": 41969 }, { "epoch": 12.91, "learning_rate": 5.90018894272149e-06, "loss": 0.0373, "step": 41970 }, { "epoch": 12.91, "learning_rate": 5.899734478682064e-06, "loss": 0.032, "step": 41971 }, { "epoch": 12.91, "learning_rate": 5.899280024822496e-06, "loss": 0.0274, "step": 41972 }, { "epoch": 12.91, "learning_rate": 5.898825581143902e-06, "loss": 0.0452, "step": 41973 }, { "epoch": 12.91, "learning_rate": 5.898371147647426e-06, "loss": 0.022, "step": 41974 }, { "epoch": 12.92, "learning_rate": 5.897916724334185e-06, "loss": 0.022, "step": 41975 }, { "epoch": 12.92, "learning_rate": 5.897462311205313e-06, "loss": 0.0306, "step": 41976 }, { "epoch": 12.92, "learning_rate": 5.897007908261935e-06, "loss": 0.0262, "step": 41977 }, { "epoch": 12.92, "learning_rate": 5.896553515505178e-06, "loss": 0.0128, "step": 41978 }, { "epoch": 12.92, "learning_rate": 5.896099132936175e-06, "loss": 0.036, "step": 41979 }, { "epoch": 12.92, "learning_rate": 5.8956447605560505e-06, "loss": 0.0217, "step": 41980 }, { "epoch": 12.92, "learning_rate": 5.895190398365935e-06, "loss": 0.0278, "step": 41981 }, { "epoch": 12.92, "learning_rate": 5.894736046366956e-06, "loss": 0.0383, "step": 41982 }, { "epoch": 12.92, "learning_rate": 5.894281704560235e-06, "loss": 0.0464, "step": 41983 }, { "epoch": 12.92, "learning_rate": 5.89382737294691e-06, "loss": 0.0118, "step": 41984 }, { "epoch": 12.92, "learning_rate": 5.893373051528106e-06, "loss": 0.018, "step": 41985 }, { "epoch": 12.92, "learning_rate": 5.8929187403049485e-06, "loss": 0.0486, "step": 41986 }, { "epoch": 12.92, "learning_rate": 5.892464439278566e-06, "loss": 0.0302, "step": 41987 }, { "epoch": 12.92, "learning_rate": 5.892010148450085e-06, "loss": 0.026, "step": 41988 }, { "epoch": 12.92, "learning_rate": 5.891555867820637e-06, "loss": 0.0249, "step": 41989 }, { "epoch": 12.92, "learning_rate": 5.891101597391348e-06, "loss": 0.0376, "step": 41990 }, { "epoch": 12.92, "learning_rate": 5.890647337163347e-06, "loss": 0.0178, "step": 41991 }, { "epoch": 12.92, "learning_rate": 5.890193087137759e-06, "loss": 0.0382, "step": 41992 }, { "epoch": 12.92, "learning_rate": 5.8897388473157134e-06, "loss": 0.0343, "step": 41993 }, { "epoch": 12.92, "learning_rate": 5.889284617698339e-06, "loss": 0.0236, "step": 41994 }, { "epoch": 12.92, "learning_rate": 5.888830398286765e-06, "loss": 0.0158, "step": 41995 }, { "epoch": 12.92, "learning_rate": 5.888376189082116e-06, "loss": 0.0455, "step": 41996 }, { "epoch": 12.92, "learning_rate": 5.8879219900855185e-06, "loss": 0.032, "step": 41997 }, { "epoch": 12.92, "learning_rate": 5.887467801298101e-06, "loss": 0.064, "step": 41998 }, { "epoch": 12.92, "learning_rate": 5.8870136227209955e-06, "loss": 0.0238, "step": 41999 }, { "epoch": 12.92, "learning_rate": 5.8865594543553254e-06, "loss": 0.0249, "step": 42000 }, { "epoch": 12.92, "learning_rate": 5.886105296202218e-06, "loss": 0.0207, "step": 42001 }, { "epoch": 12.92, "learning_rate": 5.8856511482628035e-06, "loss": 0.0222, "step": 42002 }, { "epoch": 12.92, "learning_rate": 5.885197010538206e-06, "loss": 0.038, "step": 42003 }, { "epoch": 12.92, "learning_rate": 5.884742883029556e-06, "loss": 0.0295, "step": 42004 }, { "epoch": 12.92, "learning_rate": 5.884288765737981e-06, "loss": 0.0402, "step": 42005 }, { "epoch": 12.92, "learning_rate": 5.883834658664609e-06, "loss": 0.0582, "step": 42006 }, { "epoch": 12.93, "learning_rate": 5.8833805618105635e-06, "loss": 0.0378, "step": 42007 }, { "epoch": 12.93, "learning_rate": 5.882926475176971e-06, "loss": 0.0241, "step": 42008 }, { "epoch": 12.93, "learning_rate": 5.882472398764968e-06, "loss": 0.0263, "step": 42009 }, { "epoch": 12.93, "learning_rate": 5.882018332575674e-06, "loss": 0.0288, "step": 42010 }, { "epoch": 12.93, "learning_rate": 5.882018332575674e-06, "loss": 0.0229, "step": 42011 }, { "epoch": 12.93, "learning_rate": 5.881564276610218e-06, "loss": 0.0308, "step": 42012 }, { "epoch": 12.93, "learning_rate": 5.881110230869726e-06, "loss": 0.0374, "step": 42013 }, { "epoch": 12.93, "learning_rate": 5.880656195355328e-06, "loss": 0.0264, "step": 42014 }, { "epoch": 12.93, "learning_rate": 5.88020217006815e-06, "loss": 0.0263, "step": 42015 }, { "epoch": 12.93, "learning_rate": 5.879748155009321e-06, "loss": 0.0172, "step": 42016 }, { "epoch": 12.93, "learning_rate": 5.879294150179967e-06, "loss": 0.0244, "step": 42017 }, { "epoch": 12.93, "learning_rate": 5.878840155581213e-06, "loss": 0.0648, "step": 42018 }, { "epoch": 12.93, "learning_rate": 5.878386171214184e-06, "loss": 0.0301, "step": 42019 }, { "epoch": 12.93, "learning_rate": 5.877932197080017e-06, "loss": 0.0316, "step": 42020 }, { "epoch": 12.93, "learning_rate": 5.877478233179832e-06, "loss": 0.0312, "step": 42021 }, { "epoch": 12.93, "learning_rate": 5.877024279514756e-06, "loss": 0.0291, "step": 42022 }, { "epoch": 12.93, "learning_rate": 5.8765703360859185e-06, "loss": 0.0383, "step": 42023 }, { "epoch": 12.93, "learning_rate": 5.876116402894442e-06, "loss": 0.0699, "step": 42024 }, { "epoch": 12.93, "learning_rate": 5.875662479941459e-06, "loss": 0.0241, "step": 42025 }, { "epoch": 12.93, "learning_rate": 5.875208567228094e-06, "loss": 0.0496, "step": 42026 }, { "epoch": 12.93, "learning_rate": 5.874754664755473e-06, "loss": 0.0361, "step": 42027 }, { "epoch": 12.93, "learning_rate": 5.874300772524728e-06, "loss": 0.0401, "step": 42028 }, { "epoch": 12.93, "learning_rate": 5.873846890536976e-06, "loss": 0.0214, "step": 42029 }, { "epoch": 12.93, "learning_rate": 5.873393018793352e-06, "loss": 0.0418, "step": 42030 }, { "epoch": 12.93, "learning_rate": 5.872939157294983e-06, "loss": 0.0278, "step": 42031 }, { "epoch": 12.93, "learning_rate": 5.872485306042993e-06, "loss": 0.0126, "step": 42032 }, { "epoch": 12.93, "learning_rate": 5.872031465038508e-06, "loss": 0.0294, "step": 42033 }, { "epoch": 12.93, "learning_rate": 5.871577634282655e-06, "loss": 0.0321, "step": 42034 }, { "epoch": 12.93, "learning_rate": 5.871123813776563e-06, "loss": 0.0273, "step": 42035 }, { "epoch": 12.93, "learning_rate": 5.870670003521357e-06, "loss": 0.0524, "step": 42036 }, { "epoch": 12.93, "learning_rate": 5.870216203518165e-06, "loss": 0.0384, "step": 42037 }, { "epoch": 12.93, "learning_rate": 5.869762413768114e-06, "loss": 0.02, "step": 42038 }, { "epoch": 12.93, "learning_rate": 5.869308634272325e-06, "loss": 0.0486, "step": 42039 }, { "epoch": 12.94, "learning_rate": 5.868854865031932e-06, "loss": 0.0184, "step": 42040 }, { "epoch": 12.94, "learning_rate": 5.868401106048061e-06, "loss": 0.0471, "step": 42041 }, { "epoch": 12.94, "learning_rate": 5.8679473573218335e-06, "loss": 0.0285, "step": 42042 }, { "epoch": 12.94, "learning_rate": 5.867493618854379e-06, "loss": 0.0506, "step": 42043 }, { "epoch": 12.94, "learning_rate": 5.867039890646822e-06, "loss": 0.0342, "step": 42044 }, { "epoch": 12.94, "learning_rate": 5.866586172700292e-06, "loss": 0.0318, "step": 42045 }, { "epoch": 12.94, "learning_rate": 5.866132465015916e-06, "loss": 0.0319, "step": 42046 }, { "epoch": 12.94, "learning_rate": 5.865678767594817e-06, "loss": 0.0815, "step": 42047 }, { "epoch": 12.94, "learning_rate": 5.865225080438124e-06, "loss": 0.0256, "step": 42048 }, { "epoch": 12.94, "learning_rate": 5.8647714035469595e-06, "loss": 0.0161, "step": 42049 }, { "epoch": 12.94, "learning_rate": 5.864317736922456e-06, "loss": 0.0288, "step": 42050 }, { "epoch": 12.94, "learning_rate": 5.863864080565735e-06, "loss": 0.023, "step": 42051 }, { "epoch": 12.94, "learning_rate": 5.863410434477927e-06, "loss": 0.0292, "step": 42052 }, { "epoch": 12.94, "learning_rate": 5.862956798660154e-06, "loss": 0.0281, "step": 42053 }, { "epoch": 12.94, "learning_rate": 5.862503173113542e-06, "loss": 0.0375, "step": 42054 }, { "epoch": 12.94, "learning_rate": 5.862049557839221e-06, "loss": 0.0515, "step": 42055 }, { "epoch": 12.94, "learning_rate": 5.861595952838315e-06, "loss": 0.0249, "step": 42056 }, { "epoch": 12.94, "learning_rate": 5.861142358111951e-06, "loss": 0.0254, "step": 42057 }, { "epoch": 12.94, "learning_rate": 5.860688773661256e-06, "loss": 0.0409, "step": 42058 }, { "epoch": 12.94, "learning_rate": 5.860235199487351e-06, "loss": 0.046, "step": 42059 }, { "epoch": 12.94, "learning_rate": 5.8597816355913685e-06, "loss": 0.0479, "step": 42060 }, { "epoch": 12.94, "learning_rate": 5.8593280819744314e-06, "loss": 0.0187, "step": 42061 }, { "epoch": 12.94, "learning_rate": 5.85887453863767e-06, "loss": 0.02, "step": 42062 }, { "epoch": 12.94, "learning_rate": 5.858421005582204e-06, "loss": 0.0439, "step": 42063 }, { "epoch": 12.94, "learning_rate": 5.857967482809157e-06, "loss": 0.0187, "step": 42064 }, { "epoch": 12.94, "learning_rate": 5.8575139703196684e-06, "loss": 0.0281, "step": 42065 }, { "epoch": 12.94, "learning_rate": 5.857060468114852e-06, "loss": 0.0381, "step": 42066 }, { "epoch": 12.94, "learning_rate": 5.856606976195838e-06, "loss": 0.0856, "step": 42067 }, { "epoch": 12.94, "learning_rate": 5.856153494563752e-06, "loss": 0.0129, "step": 42068 }, { "epoch": 12.94, "learning_rate": 5.855700023219717e-06, "loss": 0.015, "step": 42069 }, { "epoch": 12.94, "learning_rate": 5.855246562164864e-06, "loss": 0.0362, "step": 42070 }, { "epoch": 12.94, "learning_rate": 5.854793111400317e-06, "loss": 0.0572, "step": 42071 }, { "epoch": 12.95, "learning_rate": 5.8543396709272005e-06, "loss": 0.0164, "step": 42072 }, { "epoch": 12.95, "learning_rate": 5.853886240746643e-06, "loss": 0.0268, "step": 42073 }, { "epoch": 12.95, "learning_rate": 5.853432820859763e-06, "loss": 0.0343, "step": 42074 }, { "epoch": 12.95, "learning_rate": 5.852979411267696e-06, "loss": 0.0318, "step": 42075 }, { "epoch": 12.95, "learning_rate": 5.852526011971562e-06, "loss": 0.0356, "step": 42076 }, { "epoch": 12.95, "learning_rate": 5.852072622972488e-06, "loss": 0.0599, "step": 42077 }, { "epoch": 12.95, "learning_rate": 5.851619244271599e-06, "loss": 0.0248, "step": 42078 }, { "epoch": 12.95, "learning_rate": 5.85116587587002e-06, "loss": 0.0371, "step": 42079 }, { "epoch": 12.95, "learning_rate": 5.850712517768877e-06, "loss": 0.0384, "step": 42080 }, { "epoch": 12.95, "learning_rate": 5.850259169969299e-06, "loss": 0.016, "step": 42081 }, { "epoch": 12.95, "learning_rate": 5.849805832472408e-06, "loss": 0.0133, "step": 42082 }, { "epoch": 12.95, "learning_rate": 5.849352505279332e-06, "loss": 0.022, "step": 42083 }, { "epoch": 12.95, "learning_rate": 5.848899188391189e-06, "loss": 0.0272, "step": 42084 }, { "epoch": 12.95, "learning_rate": 5.8484458818091135e-06, "loss": 0.0175, "step": 42085 }, { "epoch": 12.95, "learning_rate": 5.84799258553423e-06, "loss": 0.0136, "step": 42086 }, { "epoch": 12.95, "learning_rate": 5.847539299567659e-06, "loss": 0.041, "step": 42087 }, { "epoch": 12.95, "learning_rate": 5.847086023910531e-06, "loss": 0.0317, "step": 42088 }, { "epoch": 12.95, "learning_rate": 5.8466327585639635e-06, "loss": 0.0233, "step": 42089 }, { "epoch": 12.95, "learning_rate": 5.846179503529091e-06, "loss": 0.0241, "step": 42090 }, { "epoch": 12.95, "learning_rate": 5.845726258807035e-06, "loss": 0.0347, "step": 42091 }, { "epoch": 12.95, "learning_rate": 5.8452730243989205e-06, "loss": 0.0377, "step": 42092 }, { "epoch": 12.95, "learning_rate": 5.844819800305874e-06, "loss": 0.0254, "step": 42093 }, { "epoch": 12.95, "learning_rate": 5.844366586529015e-06, "loss": 0.0428, "step": 42094 }, { "epoch": 12.95, "learning_rate": 5.843913383069477e-06, "loss": 0.0331, "step": 42095 }, { "epoch": 12.95, "learning_rate": 5.843460189928384e-06, "loss": 0.0329, "step": 42096 }, { "epoch": 12.95, "learning_rate": 5.843007007106855e-06, "loss": 0.0223, "step": 42097 }, { "epoch": 12.95, "learning_rate": 5.842553834606021e-06, "loss": 0.0252, "step": 42098 }, { "epoch": 12.95, "learning_rate": 5.842100672427002e-06, "loss": 0.0186, "step": 42099 }, { "epoch": 12.95, "learning_rate": 5.84164752057093e-06, "loss": 0.0275, "step": 42100 }, { "epoch": 12.95, "learning_rate": 5.841194379038924e-06, "loss": 0.0409, "step": 42101 }, { "epoch": 12.95, "learning_rate": 5.840741247832112e-06, "loss": 0.0397, "step": 42102 }, { "epoch": 12.95, "learning_rate": 5.840288126951618e-06, "loss": 0.0634, "step": 42103 }, { "epoch": 12.95, "learning_rate": 5.839835016398565e-06, "loss": 0.0213, "step": 42104 }, { "epoch": 12.96, "learning_rate": 5.839381916174082e-06, "loss": 0.0688, "step": 42105 }, { "epoch": 12.96, "learning_rate": 5.838928826279292e-06, "loss": 0.0317, "step": 42106 }, { "epoch": 12.96, "learning_rate": 5.838475746715319e-06, "loss": 0.0218, "step": 42107 }, { "epoch": 12.96, "learning_rate": 5.838022677483291e-06, "loss": 0.0263, "step": 42108 }, { "epoch": 12.96, "learning_rate": 5.83756961858433e-06, "loss": 0.0291, "step": 42109 }, { "epoch": 12.96, "learning_rate": 5.837116570019561e-06, "loss": 0.025, "step": 42110 }, { "epoch": 12.96, "learning_rate": 5.83666353179011e-06, "loss": 0.0419, "step": 42111 }, { "epoch": 12.96, "learning_rate": 5.836210503897099e-06, "loss": 0.0279, "step": 42112 }, { "epoch": 12.96, "learning_rate": 5.835757486341658e-06, "loss": 0.0284, "step": 42113 }, { "epoch": 12.96, "learning_rate": 5.8353044791249015e-06, "loss": 0.0302, "step": 42114 }, { "epoch": 12.96, "learning_rate": 5.83485148224797e-06, "loss": 0.0228, "step": 42115 }, { "epoch": 12.96, "learning_rate": 5.834398495711977e-06, "loss": 0.0427, "step": 42116 }, { "epoch": 12.96, "learning_rate": 5.8339455195180485e-06, "loss": 0.0343, "step": 42117 }, { "epoch": 12.96, "learning_rate": 5.83349255366731e-06, "loss": 0.0187, "step": 42118 }, { "epoch": 12.96, "learning_rate": 5.833039598160881e-06, "loss": 0.0523, "step": 42119 }, { "epoch": 12.96, "learning_rate": 5.832586652999898e-06, "loss": 0.0514, "step": 42120 }, { "epoch": 12.96, "learning_rate": 5.8321337181854775e-06, "loss": 0.0224, "step": 42121 }, { "epoch": 12.96, "learning_rate": 5.831680793718748e-06, "loss": 0.017, "step": 42122 }, { "epoch": 12.96, "learning_rate": 5.8312278796008295e-06, "loss": 0.0425, "step": 42123 }, { "epoch": 12.96, "learning_rate": 5.830774975832842e-06, "loss": 0.0337, "step": 42124 }, { "epoch": 12.96, "learning_rate": 5.830322082415922e-06, "loss": 0.0288, "step": 42125 }, { "epoch": 12.96, "learning_rate": 5.829869199351188e-06, "loss": 0.0367, "step": 42126 }, { "epoch": 12.96, "learning_rate": 5.8294163266397654e-06, "loss": 0.0434, "step": 42127 }, { "epoch": 12.96, "learning_rate": 5.828963464282776e-06, "loss": 0.0378, "step": 42128 }, { "epoch": 12.96, "learning_rate": 5.828510612281346e-06, "loss": 0.0361, "step": 42129 }, { "epoch": 12.96, "learning_rate": 5.828057770636601e-06, "loss": 0.0389, "step": 42130 }, { "epoch": 12.96, "learning_rate": 5.827604939349662e-06, "loss": 0.0189, "step": 42131 }, { "epoch": 12.96, "learning_rate": 5.827152118421657e-06, "loss": 0.0361, "step": 42132 }, { "epoch": 12.96, "learning_rate": 5.826699307853707e-06, "loss": 0.0263, "step": 42133 }, { "epoch": 12.96, "learning_rate": 5.826246507646933e-06, "loss": 0.0236, "step": 42134 }, { "epoch": 12.96, "learning_rate": 5.825793717802469e-06, "loss": 0.0358, "step": 42135 }, { "epoch": 12.96, "learning_rate": 5.825340938321438e-06, "loss": 0.0307, "step": 42136 }, { "epoch": 12.97, "learning_rate": 5.824888169204955e-06, "loss": 0.0497, "step": 42137 }, { "epoch": 12.97, "learning_rate": 5.82443541045415e-06, "loss": 0.0308, "step": 42138 }, { "epoch": 12.97, "learning_rate": 5.823982662070142e-06, "loss": 0.0283, "step": 42139 }, { "epoch": 12.97, "learning_rate": 5.823529924054062e-06, "loss": 0.0364, "step": 42140 }, { "epoch": 12.97, "learning_rate": 5.8230771964070324e-06, "loss": 0.0338, "step": 42141 }, { "epoch": 12.97, "learning_rate": 5.822624479130176e-06, "loss": 0.0362, "step": 42142 }, { "epoch": 12.97, "learning_rate": 5.822171772224621e-06, "loss": 0.0457, "step": 42143 }, { "epoch": 12.97, "learning_rate": 5.821719075691479e-06, "loss": 0.0205, "step": 42144 }, { "epoch": 12.97, "learning_rate": 5.8212663895318855e-06, "loss": 0.0275, "step": 42145 }, { "epoch": 12.97, "learning_rate": 5.820813713746961e-06, "loss": 0.0325, "step": 42146 }, { "epoch": 12.97, "learning_rate": 5.82036104833783e-06, "loss": 0.0432, "step": 42147 }, { "epoch": 12.97, "learning_rate": 5.819908393305616e-06, "loss": 0.0386, "step": 42148 }, { "epoch": 12.97, "learning_rate": 5.819455748651441e-06, "loss": 0.026, "step": 42149 }, { "epoch": 12.97, "learning_rate": 5.819003114376431e-06, "loss": 0.0417, "step": 42150 }, { "epoch": 12.97, "learning_rate": 5.818550490481711e-06, "loss": 0.0322, "step": 42151 }, { "epoch": 12.97, "learning_rate": 5.818097876968401e-06, "loss": 0.0324, "step": 42152 }, { "epoch": 12.97, "learning_rate": 5.817645273837627e-06, "loss": 0.025, "step": 42153 }, { "epoch": 12.97, "learning_rate": 5.817192681090509e-06, "loss": 0.0475, "step": 42154 }, { "epoch": 12.97, "learning_rate": 5.8167400987281766e-06, "loss": 0.0312, "step": 42155 }, { "epoch": 12.97, "learning_rate": 5.8162875267517515e-06, "loss": 0.0423, "step": 42156 }, { "epoch": 12.97, "learning_rate": 5.815834965162359e-06, "loss": 0.0356, "step": 42157 }, { "epoch": 12.97, "learning_rate": 5.815382413961119e-06, "loss": 0.0273, "step": 42158 }, { "epoch": 12.97, "learning_rate": 5.814929873149151e-06, "loss": 0.0242, "step": 42159 }, { "epoch": 12.97, "learning_rate": 5.814477342727587e-06, "loss": 0.02, "step": 42160 }, { "epoch": 12.97, "learning_rate": 5.814024822697548e-06, "loss": 0.0218, "step": 42161 }, { "epoch": 12.97, "learning_rate": 5.8135723130601575e-06, "loss": 0.0147, "step": 42162 }, { "epoch": 12.97, "learning_rate": 5.813119813816538e-06, "loss": 0.0256, "step": 42163 }, { "epoch": 12.97, "learning_rate": 5.812667324967813e-06, "loss": 0.0173, "step": 42164 }, { "epoch": 12.97, "learning_rate": 5.812214846515106e-06, "loss": 0.0381, "step": 42165 }, { "epoch": 12.97, "learning_rate": 5.811762378459541e-06, "loss": 0.033, "step": 42166 }, { "epoch": 12.97, "learning_rate": 5.811309920802242e-06, "loss": 0.0365, "step": 42167 }, { "epoch": 12.97, "learning_rate": 5.810857473544329e-06, "loss": 0.0228, "step": 42168 }, { "epoch": 12.97, "learning_rate": 5.810405036686925e-06, "loss": 0.0344, "step": 42169 }, { "epoch": 12.98, "learning_rate": 5.809952610231162e-06, "loss": 0.0469, "step": 42170 }, { "epoch": 12.98, "learning_rate": 5.809500194178154e-06, "loss": 0.0259, "step": 42171 }, { "epoch": 12.98, "learning_rate": 5.80904778852903e-06, "loss": 0.0466, "step": 42172 }, { "epoch": 12.98, "learning_rate": 5.8085953932849074e-06, "loss": 0.0297, "step": 42173 }, { "epoch": 12.98, "learning_rate": 5.8081430084469095e-06, "loss": 0.016, "step": 42174 }, { "epoch": 12.98, "learning_rate": 5.807690634016166e-06, "loss": 0.0277, "step": 42175 }, { "epoch": 12.98, "learning_rate": 5.807238269993796e-06, "loss": 0.0278, "step": 42176 }, { "epoch": 12.98, "learning_rate": 5.806785916380924e-06, "loss": 0.0228, "step": 42177 }, { "epoch": 12.98, "learning_rate": 5.806333573178675e-06, "loss": 0.0121, "step": 42178 }, { "epoch": 12.98, "learning_rate": 5.8058812403881625e-06, "loss": 0.0261, "step": 42179 }, { "epoch": 12.98, "learning_rate": 5.80542891801052e-06, "loss": 0.0327, "step": 42180 }, { "epoch": 12.98, "learning_rate": 5.804976606046866e-06, "loss": 0.0195, "step": 42181 }, { "epoch": 12.98, "learning_rate": 5.804524304498325e-06, "loss": 0.0528, "step": 42182 }, { "epoch": 12.98, "learning_rate": 5.804072013366017e-06, "loss": 0.0131, "step": 42183 }, { "epoch": 12.98, "learning_rate": 5.803619732651069e-06, "loss": 0.0175, "step": 42184 }, { "epoch": 12.98, "learning_rate": 5.8031674623546016e-06, "loss": 0.0446, "step": 42185 }, { "epoch": 12.98, "learning_rate": 5.802715202477738e-06, "loss": 0.031, "step": 42186 }, { "epoch": 12.98, "learning_rate": 5.8022629530216e-06, "loss": 0.0276, "step": 42187 }, { "epoch": 12.98, "learning_rate": 5.801810713987313e-06, "loss": 0.0279, "step": 42188 }, { "epoch": 12.98, "learning_rate": 5.801358485375994e-06, "loss": 0.0452, "step": 42189 }, { "epoch": 12.98, "learning_rate": 5.800906267188773e-06, "loss": 0.0402, "step": 42190 }, { "epoch": 12.98, "learning_rate": 5.800454059426774e-06, "loss": 0.0313, "step": 42191 }, { "epoch": 12.98, "learning_rate": 5.800001862091112e-06, "loss": 0.028, "step": 42192 }, { "epoch": 12.98, "learning_rate": 5.799549675182912e-06, "loss": 0.0334, "step": 42193 }, { "epoch": 12.98, "learning_rate": 5.799097498703294e-06, "loss": 0.0723, "step": 42194 }, { "epoch": 12.98, "learning_rate": 5.798645332653389e-06, "loss": 0.023, "step": 42195 }, { "epoch": 12.98, "learning_rate": 5.798193177034315e-06, "loss": 0.0384, "step": 42196 }, { "epoch": 12.98, "learning_rate": 5.797741031847194e-06, "loss": 0.041, "step": 42197 }, { "epoch": 12.98, "learning_rate": 5.797288897093149e-06, "loss": 0.0513, "step": 42198 }, { "epoch": 12.98, "learning_rate": 5.796836772773303e-06, "loss": 0.0403, "step": 42199 }, { "epoch": 12.98, "learning_rate": 5.796384658888778e-06, "loss": 0.0344, "step": 42200 }, { "epoch": 12.98, "learning_rate": 5.795932555440697e-06, "loss": 0.0481, "step": 42201 }, { "epoch": 12.99, "learning_rate": 5.795480462430182e-06, "loss": 0.0645, "step": 42202 }, { "epoch": 12.99, "learning_rate": 5.795028379858355e-06, "loss": 0.0419, "step": 42203 }, { "epoch": 12.99, "learning_rate": 5.794576307726339e-06, "loss": 0.0454, "step": 42204 }, { "epoch": 12.99, "learning_rate": 5.794124246035258e-06, "loss": 0.0416, "step": 42205 }, { "epoch": 12.99, "learning_rate": 5.793672194786231e-06, "loss": 0.0244, "step": 42206 }, { "epoch": 12.99, "learning_rate": 5.793220153980381e-06, "loss": 0.0349, "step": 42207 }, { "epoch": 12.99, "learning_rate": 5.7927681236188325e-06, "loss": 0.0318, "step": 42208 }, { "epoch": 12.99, "learning_rate": 5.792316103702703e-06, "loss": 0.0333, "step": 42209 }, { "epoch": 12.99, "learning_rate": 5.791864094233122e-06, "loss": 0.0283, "step": 42210 }, { "epoch": 12.99, "learning_rate": 5.791412095211207e-06, "loss": 0.0392, "step": 42211 }, { "epoch": 12.99, "learning_rate": 5.790960106638086e-06, "loss": 0.0357, "step": 42212 }, { "epoch": 12.99, "learning_rate": 5.790508128514872e-06, "loss": 0.0225, "step": 42213 }, { "epoch": 12.99, "learning_rate": 5.790056160842688e-06, "loss": 0.0333, "step": 42214 }, { "epoch": 12.99, "learning_rate": 5.789604203622663e-06, "loss": 0.0397, "step": 42215 }, { "epoch": 12.99, "learning_rate": 5.789152256855917e-06, "loss": 0.0375, "step": 42216 }, { "epoch": 12.99, "learning_rate": 5.788700320543569e-06, "loss": 0.0376, "step": 42217 }, { "epoch": 12.99, "learning_rate": 5.788248394686743e-06, "loss": 0.0264, "step": 42218 }, { "epoch": 12.99, "learning_rate": 5.78779647928656e-06, "loss": 0.027, "step": 42219 }, { "epoch": 12.99, "learning_rate": 5.787344574344145e-06, "loss": 0.0312, "step": 42220 }, { "epoch": 12.99, "learning_rate": 5.786892679860616e-06, "loss": 0.0592, "step": 42221 }, { "epoch": 12.99, "learning_rate": 5.786440795837098e-06, "loss": 0.0358, "step": 42222 }, { "epoch": 12.99, "learning_rate": 5.785988922274711e-06, "loss": 0.0295, "step": 42223 }, { "epoch": 12.99, "learning_rate": 5.785537059174573e-06, "loss": 0.0244, "step": 42224 }, { "epoch": 12.99, "learning_rate": 5.785085206537819e-06, "loss": 0.0203, "step": 42225 }, { "epoch": 12.99, "learning_rate": 5.784633364365557e-06, "loss": 0.034, "step": 42226 }, { "epoch": 12.99, "learning_rate": 5.784181532658915e-06, "loss": 0.0219, "step": 42227 }, { "epoch": 12.99, "learning_rate": 5.783729711419014e-06, "loss": 0.0299, "step": 42228 }, { "epoch": 12.99, "learning_rate": 5.783277900646971e-06, "loss": 0.0303, "step": 42229 }, { "epoch": 12.99, "learning_rate": 5.782826100343916e-06, "loss": 0.0422, "step": 42230 }, { "epoch": 12.99, "learning_rate": 5.782374310510967e-06, "loss": 0.0275, "step": 42231 }, { "epoch": 12.99, "learning_rate": 5.781922531149244e-06, "loss": 0.0319, "step": 42232 }, { "epoch": 12.99, "learning_rate": 5.781470762259877e-06, "loss": 0.0245, "step": 42233 }, { "epoch": 12.99, "learning_rate": 5.781019003843971e-06, "loss": 0.0238, "step": 42234 }, { "epoch": 13.0, "learning_rate": 5.780567255902662e-06, "loss": 0.0466, "step": 42235 }, { "epoch": 13.0, "learning_rate": 5.780115518437067e-06, "loss": 0.0282, "step": 42236 }, { "epoch": 13.0, "learning_rate": 5.779663791448307e-06, "loss": 0.0185, "step": 42237 }, { "epoch": 13.0, "learning_rate": 5.779212074937505e-06, "loss": 0.0443, "step": 42238 }, { "epoch": 13.0, "learning_rate": 5.7787603689057805e-06, "loss": 0.0199, "step": 42239 }, { "epoch": 13.0, "learning_rate": 5.778308673354256e-06, "loss": 0.0212, "step": 42240 }, { "epoch": 13.0, "learning_rate": 5.777856988284054e-06, "loss": 0.0153, "step": 42241 }, { "epoch": 13.0, "learning_rate": 5.777405313696294e-06, "loss": 0.0264, "step": 42242 }, { "epoch": 13.0, "learning_rate": 5.776953649592099e-06, "loss": 0.0337, "step": 42243 }, { "epoch": 13.0, "learning_rate": 5.7765019959725855e-06, "loss": 0.0315, "step": 42244 }, { "epoch": 13.0, "learning_rate": 5.776050352838882e-06, "loss": 0.0448, "step": 42245 }, { "epoch": 13.0, "learning_rate": 5.77559872019211e-06, "loss": 0.03, "step": 42246 }, { "epoch": 13.0, "learning_rate": 5.775147098033385e-06, "loss": 0.0264, "step": 42247 }, { "epoch": 13.0, "learning_rate": 5.77469548636383e-06, "loss": 0.0509, "step": 42248 }, { "epoch": 13.0, "learning_rate": 5.774243885184565e-06, "loss": 0.0168, "step": 42249 }, { "epoch": 13.0, "learning_rate": 5.773792294496715e-06, "loss": 0.0211, "step": 42250 }, { "epoch": 13.0, "learning_rate": 5.773340714301401e-06, "loss": 0.0251, "step": 42251 }, { "epoch": 13.0, "learning_rate": 5.772889144599741e-06, "loss": 0.0406, "step": 42252 }, { "epoch": 13.0, "learning_rate": 5.772437585392858e-06, "loss": 0.0238, "step": 42253 }, { "epoch": 13.0, "learning_rate": 5.771986036681874e-06, "loss": 0.0287, "step": 42254 }, { "epoch": 13.0, "learning_rate": 5.771534498467908e-06, "loss": 0.0205, "step": 42255 }, { "epoch": 13.0, "learning_rate": 5.771082970752081e-06, "loss": 0.0246, "step": 42256 }, { "epoch": 13.0, "learning_rate": 5.770631453535515e-06, "loss": 0.023, "step": 42257 }, { "epoch": 13.0, "learning_rate": 5.770179946819332e-06, "loss": 0.0113, "step": 42258 }, { "epoch": 13.0, "learning_rate": 5.769728450604652e-06, "loss": 0.0329, "step": 42259 }, { "epoch": 13.0, "learning_rate": 5.769276964892594e-06, "loss": 0.0192, "step": 42260 }, { "epoch": 13.0, "learning_rate": 5.7688254896842825e-06, "loss": 0.0244, "step": 42261 }, { "epoch": 13.0, "learning_rate": 5.7683740249808365e-06, "loss": 0.0146, "step": 42262 }, { "epoch": 13.0, "learning_rate": 5.7679225707833765e-06, "loss": 0.0231, "step": 42263 }, { "epoch": 13.0, "learning_rate": 5.76747112709302e-06, "loss": 0.0207, "step": 42264 }, { "epoch": 13.0, "learning_rate": 5.767019693910896e-06, "loss": 0.0244, "step": 42265 }, { "epoch": 13.0, "learning_rate": 5.76656827123812e-06, "loss": 0.022, "step": 42266 }, { "epoch": 13.01, "learning_rate": 5.766116859075818e-06, "loss": 0.0129, "step": 42267 }, { "epoch": 13.01, "learning_rate": 5.765665457425102e-06, "loss": 0.0175, "step": 42268 }, { "epoch": 13.01, "learning_rate": 5.765214066287094e-06, "loss": 0.028, "step": 42269 }, { "epoch": 13.01, "learning_rate": 5.764762685662921e-06, "loss": 0.0248, "step": 42270 }, { "epoch": 13.01, "learning_rate": 5.764311315553701e-06, "loss": 0.0244, "step": 42271 }, { "epoch": 13.01, "learning_rate": 5.763859955960554e-06, "loss": 0.0082, "step": 42272 }, { "epoch": 13.01, "learning_rate": 5.763408606884602e-06, "loss": 0.0212, "step": 42273 }, { "epoch": 13.01, "learning_rate": 5.762957268326962e-06, "loss": 0.0155, "step": 42274 }, { "epoch": 13.01, "learning_rate": 5.762505940288758e-06, "loss": 0.0185, "step": 42275 }, { "epoch": 13.01, "learning_rate": 5.762054622771109e-06, "loss": 0.0176, "step": 42276 }, { "epoch": 13.01, "learning_rate": 5.761603315775136e-06, "loss": 0.0121, "step": 42277 }, { "epoch": 13.01, "learning_rate": 5.7611520193019585e-06, "loss": 0.0184, "step": 42278 }, { "epoch": 13.01, "learning_rate": 5.760700733352696e-06, "loss": 0.0245, "step": 42279 }, { "epoch": 13.01, "learning_rate": 5.760249457928477e-06, "loss": 0.0099, "step": 42280 }, { "epoch": 13.01, "learning_rate": 5.759798193030413e-06, "loss": 0.0202, "step": 42281 }, { "epoch": 13.01, "learning_rate": 5.759346938659624e-06, "loss": 0.0266, "step": 42282 }, { "epoch": 13.01, "learning_rate": 5.758895694817237e-06, "loss": 0.0169, "step": 42283 }, { "epoch": 13.01, "learning_rate": 5.758444461504362e-06, "loss": 0.0188, "step": 42284 }, { "epoch": 13.01, "learning_rate": 5.75799323872213e-06, "loss": 0.0094, "step": 42285 }, { "epoch": 13.01, "learning_rate": 5.7575420264716586e-06, "loss": 0.008, "step": 42286 }, { "epoch": 13.01, "learning_rate": 5.757090824754066e-06, "loss": 0.0211, "step": 42287 }, { "epoch": 13.01, "learning_rate": 5.756639633570478e-06, "loss": 0.0161, "step": 42288 }, { "epoch": 13.01, "learning_rate": 5.7561884529220005e-06, "loss": 0.0151, "step": 42289 }, { "epoch": 13.01, "learning_rate": 5.755737282809767e-06, "loss": 0.0227, "step": 42290 }, { "epoch": 13.01, "learning_rate": 5.755286123234894e-06, "loss": 0.012, "step": 42291 }, { "epoch": 13.01, "learning_rate": 5.754834974198501e-06, "loss": 0.0252, "step": 42292 }, { "epoch": 13.01, "learning_rate": 5.754383835701709e-06, "loss": 0.0174, "step": 42293 }, { "epoch": 13.01, "learning_rate": 5.753932707745635e-06, "loss": 0.0299, "step": 42294 }, { "epoch": 13.01, "learning_rate": 5.753481590331405e-06, "loss": 0.0177, "step": 42295 }, { "epoch": 13.01, "learning_rate": 5.753030483460132e-06, "loss": 0.021, "step": 42296 }, { "epoch": 13.01, "learning_rate": 5.7525793871329415e-06, "loss": 0.0218, "step": 42297 }, { "epoch": 13.01, "learning_rate": 5.752128301350951e-06, "loss": 0.0228, "step": 42298 }, { "epoch": 13.01, "learning_rate": 5.751677226115277e-06, "loss": 0.0202, "step": 42299 }, { "epoch": 13.02, "learning_rate": 5.751226161427046e-06, "loss": 0.0265, "step": 42300 }, { "epoch": 13.02, "learning_rate": 5.750775107287379e-06, "loss": 0.0183, "step": 42301 }, { "epoch": 13.02, "learning_rate": 5.750324063697389e-06, "loss": 0.027, "step": 42302 }, { "epoch": 13.02, "learning_rate": 5.7498730306581975e-06, "loss": 0.0215, "step": 42303 }, { "epoch": 13.02, "learning_rate": 5.749422008170922e-06, "loss": 0.0166, "step": 42304 }, { "epoch": 13.02, "learning_rate": 5.74897099623669e-06, "loss": 0.013, "step": 42305 }, { "epoch": 13.02, "learning_rate": 5.7485199948566175e-06, "loss": 0.0213, "step": 42306 }, { "epoch": 13.02, "learning_rate": 5.748069004031822e-06, "loss": 0.0349, "step": 42307 }, { "epoch": 13.02, "learning_rate": 5.747618023763426e-06, "loss": 0.0071, "step": 42308 }, { "epoch": 13.02, "learning_rate": 5.747167054052548e-06, "loss": 0.0152, "step": 42309 }, { "epoch": 13.02, "learning_rate": 5.7467160949003085e-06, "loss": 0.0217, "step": 42310 }, { "epoch": 13.02, "learning_rate": 5.746265146307825e-06, "loss": 0.0206, "step": 42311 }, { "epoch": 13.02, "learning_rate": 5.745814208276219e-06, "loss": 0.0198, "step": 42312 }, { "epoch": 13.02, "learning_rate": 5.745363280806608e-06, "loss": 0.0426, "step": 42313 }, { "epoch": 13.02, "learning_rate": 5.7449123639001105e-06, "loss": 0.0283, "step": 42314 }, { "epoch": 13.02, "learning_rate": 5.744461457557857e-06, "loss": 0.0209, "step": 42315 }, { "epoch": 13.02, "learning_rate": 5.744010561780953e-06, "loss": 0.0138, "step": 42316 }, { "epoch": 13.02, "learning_rate": 5.7435596765705235e-06, "loss": 0.0189, "step": 42317 }, { "epoch": 13.02, "learning_rate": 5.7431088019276884e-06, "loss": 0.0127, "step": 42318 }, { "epoch": 13.02, "learning_rate": 5.742657937853562e-06, "loss": 0.0261, "step": 42319 }, { "epoch": 13.02, "learning_rate": 5.742207084349274e-06, "loss": 0.0162, "step": 42320 }, { "epoch": 13.02, "learning_rate": 5.741756241415935e-06, "loss": 0.0249, "step": 42321 }, { "epoch": 13.02, "learning_rate": 5.7413054090546735e-06, "loss": 0.02, "step": 42322 }, { "epoch": 13.02, "learning_rate": 5.740854587266598e-06, "loss": 0.0344, "step": 42323 }, { "epoch": 13.02, "learning_rate": 5.740403776052827e-06, "loss": 0.0258, "step": 42324 }, { "epoch": 13.02, "learning_rate": 5.739952975414491e-06, "loss": 0.0157, "step": 42325 }, { "epoch": 13.02, "learning_rate": 5.739502185352703e-06, "loss": 0.0449, "step": 42326 }, { "epoch": 13.02, "learning_rate": 5.739051405868581e-06, "loss": 0.0161, "step": 42327 }, { "epoch": 13.02, "learning_rate": 5.738600636963247e-06, "loss": 0.0209, "step": 42328 }, { "epoch": 13.02, "learning_rate": 5.738149878637818e-06, "loss": 0.0184, "step": 42329 }, { "epoch": 13.02, "learning_rate": 5.737699130893415e-06, "loss": 0.0229, "step": 42330 }, { "epoch": 13.02, "learning_rate": 5.737248393731155e-06, "loss": 0.0173, "step": 42331 }, { "epoch": 13.03, "learning_rate": 5.7367976671521584e-06, "loss": 0.0169, "step": 42332 }, { "epoch": 13.03, "learning_rate": 5.736346951157544e-06, "loss": 0.0376, "step": 42333 }, { "epoch": 13.03, "learning_rate": 5.735896245748427e-06, "loss": 0.0145, "step": 42334 }, { "epoch": 13.03, "learning_rate": 5.735445550925934e-06, "loss": 0.0304, "step": 42335 }, { "epoch": 13.03, "learning_rate": 5.734994866691182e-06, "loss": 0.0245, "step": 42336 }, { "epoch": 13.03, "learning_rate": 5.734544193045286e-06, "loss": 0.0228, "step": 42337 }, { "epoch": 13.03, "learning_rate": 5.734093529989366e-06, "loss": 0.022, "step": 42338 }, { "epoch": 13.03, "learning_rate": 5.733642877524539e-06, "loss": 0.0164, "step": 42339 }, { "epoch": 13.03, "learning_rate": 5.7331922356519286e-06, "loss": 0.0238, "step": 42340 }, { "epoch": 13.03, "learning_rate": 5.732741604372654e-06, "loss": 0.0187, "step": 42341 }, { "epoch": 13.03, "learning_rate": 5.732290983687829e-06, "loss": 0.0199, "step": 42342 }, { "epoch": 13.03, "learning_rate": 5.731840373598581e-06, "loss": 0.0199, "step": 42343 }, { "epoch": 13.03, "learning_rate": 5.731389774106013e-06, "loss": 0.0209, "step": 42344 }, { "epoch": 13.03, "learning_rate": 5.730939185211258e-06, "loss": 0.0284, "step": 42345 }, { "epoch": 13.03, "learning_rate": 5.73048860691543e-06, "loss": 0.0207, "step": 42346 }, { "epoch": 13.03, "learning_rate": 5.730038039219649e-06, "loss": 0.0211, "step": 42347 }, { "epoch": 13.03, "learning_rate": 5.729587482125031e-06, "loss": 0.0156, "step": 42348 }, { "epoch": 13.03, "learning_rate": 5.729136935632696e-06, "loss": 0.0192, "step": 42349 }, { "epoch": 13.03, "learning_rate": 5.728686399743764e-06, "loss": 0.0198, "step": 42350 }, { "epoch": 13.03, "learning_rate": 5.728235874459351e-06, "loss": 0.0215, "step": 42351 }, { "epoch": 13.03, "learning_rate": 5.727785359780577e-06, "loss": 0.026, "step": 42352 }, { "epoch": 13.03, "learning_rate": 5.727334855708559e-06, "loss": 0.0232, "step": 42353 }, { "epoch": 13.03, "learning_rate": 5.726884362244414e-06, "loss": 0.0169, "step": 42354 }, { "epoch": 13.03, "learning_rate": 5.726433879389268e-06, "loss": 0.0173, "step": 42355 }, { "epoch": 13.03, "learning_rate": 5.725983407144233e-06, "loss": 0.0183, "step": 42356 }, { "epoch": 13.03, "learning_rate": 5.725532945510434e-06, "loss": 0.0132, "step": 42357 }, { "epoch": 13.03, "learning_rate": 5.725082494488979e-06, "loss": 0.0181, "step": 42358 }, { "epoch": 13.03, "learning_rate": 5.724632054080989e-06, "loss": 0.0139, "step": 42359 }, { "epoch": 13.03, "learning_rate": 5.7241816242875885e-06, "loss": 0.0123, "step": 42360 }, { "epoch": 13.03, "learning_rate": 5.723731205109893e-06, "loss": 0.0197, "step": 42361 }, { "epoch": 13.03, "learning_rate": 5.723280796549018e-06, "loss": 0.0255, "step": 42362 }, { "epoch": 13.03, "learning_rate": 5.722830398606085e-06, "loss": 0.0087, "step": 42363 }, { "epoch": 13.03, "learning_rate": 5.722380011282211e-06, "loss": 0.0359, "step": 42364 }, { "epoch": 13.04, "learning_rate": 5.721929634578516e-06, "loss": 0.0204, "step": 42365 }, { "epoch": 13.04, "learning_rate": 5.721479268496114e-06, "loss": 0.0309, "step": 42366 }, { "epoch": 13.04, "learning_rate": 5.721028913036126e-06, "loss": 0.0261, "step": 42367 }, { "epoch": 13.04, "learning_rate": 5.720578568199666e-06, "loss": 0.0297, "step": 42368 }, { "epoch": 13.04, "learning_rate": 5.7201282339878606e-06, "loss": 0.019, "step": 42369 }, { "epoch": 13.04, "learning_rate": 5.719677910401825e-06, "loss": 0.0127, "step": 42370 }, { "epoch": 13.04, "learning_rate": 5.719227597442672e-06, "loss": 0.027, "step": 42371 }, { "epoch": 13.04, "learning_rate": 5.718777295111524e-06, "loss": 0.0188, "step": 42372 }, { "epoch": 13.04, "learning_rate": 5.718327003409492e-06, "loss": 0.0144, "step": 42373 }, { "epoch": 13.04, "learning_rate": 5.717876722337705e-06, "loss": 0.029, "step": 42374 }, { "epoch": 13.04, "learning_rate": 5.7174264518972756e-06, "loss": 0.0385, "step": 42375 }, { "epoch": 13.04, "learning_rate": 5.716976192089321e-06, "loss": 0.0173, "step": 42376 }, { "epoch": 13.04, "learning_rate": 5.716525942914961e-06, "loss": 0.0262, "step": 42377 }, { "epoch": 13.04, "learning_rate": 5.716075704375311e-06, "loss": 0.0268, "step": 42378 }, { "epoch": 13.04, "learning_rate": 5.715625476471491e-06, "loss": 0.0154, "step": 42379 }, { "epoch": 13.04, "learning_rate": 5.715175259204617e-06, "loss": 0.0114, "step": 42380 }, { "epoch": 13.04, "learning_rate": 5.714725052575809e-06, "loss": 0.0236, "step": 42381 }, { "epoch": 13.04, "learning_rate": 5.714274856586183e-06, "loss": 0.0277, "step": 42382 }, { "epoch": 13.04, "learning_rate": 5.713824671236856e-06, "loss": 0.0139, "step": 42383 }, { "epoch": 13.04, "learning_rate": 5.713374496528948e-06, "loss": 0.0183, "step": 42384 }, { "epoch": 13.04, "learning_rate": 5.712924332463575e-06, "loss": 0.0217, "step": 42385 }, { "epoch": 13.04, "learning_rate": 5.712474179041856e-06, "loss": 0.0338, "step": 42386 }, { "epoch": 13.04, "learning_rate": 5.712024036264907e-06, "loss": 0.019, "step": 42387 }, { "epoch": 13.04, "learning_rate": 5.711573904133842e-06, "loss": 0.0243, "step": 42388 }, { "epoch": 13.04, "learning_rate": 5.711123782649788e-06, "loss": 0.0205, "step": 42389 }, { "epoch": 13.04, "learning_rate": 5.710673671813856e-06, "loss": 0.0127, "step": 42390 }, { "epoch": 13.04, "learning_rate": 5.7102235716271695e-06, "loss": 0.0195, "step": 42391 }, { "epoch": 13.04, "learning_rate": 5.709773482090837e-06, "loss": 0.0305, "step": 42392 }, { "epoch": 13.04, "learning_rate": 5.709323403205977e-06, "loss": 0.0142, "step": 42393 }, { "epoch": 13.04, "learning_rate": 5.708873334973715e-06, "loss": 0.0206, "step": 42394 }, { "epoch": 13.04, "learning_rate": 5.708423277395162e-06, "loss": 0.0195, "step": 42395 }, { "epoch": 13.04, "learning_rate": 5.707973230471437e-06, "loss": 0.0219, "step": 42396 }, { "epoch": 13.05, "learning_rate": 5.707523194203659e-06, "loss": 0.03, "step": 42397 }, { "epoch": 13.05, "learning_rate": 5.707073168592943e-06, "loss": 0.0184, "step": 42398 }, { "epoch": 13.05, "learning_rate": 5.706623153640406e-06, "loss": 0.0135, "step": 42399 }, { "epoch": 13.05, "learning_rate": 5.706173149347167e-06, "loss": 0.0216, "step": 42400 }, { "epoch": 13.05, "learning_rate": 5.705723155714342e-06, "loss": 0.0249, "step": 42401 }, { "epoch": 13.05, "learning_rate": 5.70527317274305e-06, "loss": 0.0102, "step": 42402 }, { "epoch": 13.05, "learning_rate": 5.704823200434402e-06, "loss": 0.0345, "step": 42403 }, { "epoch": 13.05, "learning_rate": 5.7043732387895275e-06, "loss": 0.0159, "step": 42404 }, { "epoch": 13.05, "learning_rate": 5.703923287809534e-06, "loss": 0.0151, "step": 42405 }, { "epoch": 13.05, "learning_rate": 5.7034733474955385e-06, "loss": 0.0255, "step": 42406 }, { "epoch": 13.05, "learning_rate": 5.703023417848662e-06, "loss": 0.0394, "step": 42407 }, { "epoch": 13.05, "learning_rate": 5.702573498870015e-06, "loss": 0.0215, "step": 42408 }, { "epoch": 13.05, "learning_rate": 5.702123590560724e-06, "loss": 0.0187, "step": 42409 }, { "epoch": 13.05, "learning_rate": 5.701673692921901e-06, "loss": 0.015, "step": 42410 }, { "epoch": 13.05, "learning_rate": 5.701223805954664e-06, "loss": 0.0239, "step": 42411 }, { "epoch": 13.05, "learning_rate": 5.700773929660133e-06, "loss": 0.0133, "step": 42412 }, { "epoch": 13.05, "learning_rate": 5.700324064039413e-06, "loss": 0.0116, "step": 42413 }, { "epoch": 13.05, "learning_rate": 5.699874209093635e-06, "loss": 0.017, "step": 42414 }, { "epoch": 13.05, "learning_rate": 5.699424364823908e-06, "loss": 0.0209, "step": 42415 }, { "epoch": 13.05, "learning_rate": 5.698974531231352e-06, "loss": 0.0164, "step": 42416 }, { "epoch": 13.05, "learning_rate": 5.698524708317082e-06, "loss": 0.0357, "step": 42417 }, { "epoch": 13.05, "learning_rate": 5.698074896082215e-06, "loss": 0.0139, "step": 42418 }, { "epoch": 13.05, "learning_rate": 5.69762509452787e-06, "loss": 0.0147, "step": 42419 }, { "epoch": 13.05, "learning_rate": 5.697175303655161e-06, "loss": 0.0107, "step": 42420 }, { "epoch": 13.05, "learning_rate": 5.696725523465205e-06, "loss": 0.0247, "step": 42421 }, { "epoch": 13.05, "learning_rate": 5.69627575395912e-06, "loss": 0.0215, "step": 42422 }, { "epoch": 13.05, "learning_rate": 5.695825995138019e-06, "loss": 0.0238, "step": 42423 }, { "epoch": 13.05, "learning_rate": 5.695376247003025e-06, "loss": 0.013, "step": 42424 }, { "epoch": 13.05, "learning_rate": 5.6949265095552545e-06, "loss": 0.0367, "step": 42425 }, { "epoch": 13.05, "learning_rate": 5.694476782795818e-06, "loss": 0.0163, "step": 42426 }, { "epoch": 13.05, "learning_rate": 5.694027066725835e-06, "loss": 0.0128, "step": 42427 }, { "epoch": 13.05, "learning_rate": 5.6935773613464165e-06, "loss": 0.0299, "step": 42428 }, { "epoch": 13.05, "learning_rate": 5.69312766665869e-06, "loss": 0.0298, "step": 42429 }, { "epoch": 13.06, "learning_rate": 5.692677982663766e-06, "loss": 0.0096, "step": 42430 }, { "epoch": 13.06, "learning_rate": 5.692228309362761e-06, "loss": 0.0259, "step": 42431 }, { "epoch": 13.06, "learning_rate": 5.691778646756792e-06, "loss": 0.0285, "step": 42432 }, { "epoch": 13.06, "learning_rate": 5.6913289948469745e-06, "loss": 0.0141, "step": 42433 }, { "epoch": 13.06, "learning_rate": 5.690879353634427e-06, "loss": 0.023, "step": 42434 }, { "epoch": 13.06, "learning_rate": 5.6904297231202635e-06, "loss": 0.0085, "step": 42435 }, { "epoch": 13.06, "learning_rate": 5.6899801033056e-06, "loss": 0.0142, "step": 42436 }, { "epoch": 13.06, "learning_rate": 5.689530494191555e-06, "loss": 0.0293, "step": 42437 }, { "epoch": 13.06, "learning_rate": 5.689080895779244e-06, "loss": 0.0276, "step": 42438 }, { "epoch": 13.06, "learning_rate": 5.688631308069783e-06, "loss": 0.0203, "step": 42439 }, { "epoch": 13.06, "learning_rate": 5.688181731064288e-06, "loss": 0.0145, "step": 42440 }, { "epoch": 13.06, "learning_rate": 5.687732164763875e-06, "loss": 0.0315, "step": 42441 }, { "epoch": 13.06, "learning_rate": 5.6872826091696595e-06, "loss": 0.0222, "step": 42442 }, { "epoch": 13.06, "learning_rate": 5.686833064282757e-06, "loss": 0.0422, "step": 42443 }, { "epoch": 13.06, "learning_rate": 5.686383530104288e-06, "loss": 0.0217, "step": 42444 }, { "epoch": 13.06, "learning_rate": 5.685934006635365e-06, "loss": 0.0085, "step": 42445 }, { "epoch": 13.06, "learning_rate": 5.685484493877111e-06, "loss": 0.02, "step": 42446 }, { "epoch": 13.06, "learning_rate": 5.6850349918306295e-06, "loss": 0.0304, "step": 42447 }, { "epoch": 13.06, "learning_rate": 5.68458550049704e-06, "loss": 0.0213, "step": 42448 }, { "epoch": 13.06, "learning_rate": 5.684136019877465e-06, "loss": 0.0185, "step": 42449 }, { "epoch": 13.06, "learning_rate": 5.683686549973018e-06, "loss": 0.0113, "step": 42450 }, { "epoch": 13.06, "learning_rate": 5.683237090784812e-06, "loss": 0.0147, "step": 42451 }, { "epoch": 13.06, "learning_rate": 5.682787642313965e-06, "loss": 0.0251, "step": 42452 }, { "epoch": 13.06, "learning_rate": 5.682338204561593e-06, "loss": 0.0294, "step": 42453 }, { "epoch": 13.06, "learning_rate": 5.68188877752881e-06, "loss": 0.0291, "step": 42454 }, { "epoch": 13.06, "learning_rate": 5.681439361216734e-06, "loss": 0.0108, "step": 42455 }, { "epoch": 13.06, "learning_rate": 5.6809899556264795e-06, "loss": 0.0128, "step": 42456 }, { "epoch": 13.06, "learning_rate": 5.680540560759162e-06, "loss": 0.0177, "step": 42457 }, { "epoch": 13.06, "learning_rate": 5.680091176615894e-06, "loss": 0.0245, "step": 42458 }, { "epoch": 13.06, "learning_rate": 5.679641803197804e-06, "loss": 0.0207, "step": 42459 }, { "epoch": 13.06, "learning_rate": 5.679192440505994e-06, "loss": 0.0212, "step": 42460 }, { "epoch": 13.06, "learning_rate": 5.678743088541584e-06, "loss": 0.019, "step": 42461 }, { "epoch": 13.07, "learning_rate": 5.67829374730569e-06, "loss": 0.0125, "step": 42462 }, { "epoch": 13.07, "learning_rate": 5.677844416799424e-06, "loss": 0.0176, "step": 42463 }, { "epoch": 13.07, "learning_rate": 5.677395097023909e-06, "loss": 0.0249, "step": 42464 }, { "epoch": 13.07, "learning_rate": 5.676945787980257e-06, "loss": 0.0212, "step": 42465 }, { "epoch": 13.07, "learning_rate": 5.676496489669583e-06, "loss": 0.0187, "step": 42466 }, { "epoch": 13.07, "learning_rate": 5.676047202093005e-06, "loss": 0.0139, "step": 42467 }, { "epoch": 13.07, "learning_rate": 5.67559792525163e-06, "loss": 0.0286, "step": 42468 }, { "epoch": 13.07, "learning_rate": 5.675148659146583e-06, "loss": 0.0233, "step": 42469 }, { "epoch": 13.07, "learning_rate": 5.674699403778976e-06, "loss": 0.0137, "step": 42470 }, { "epoch": 13.07, "learning_rate": 5.674250159149924e-06, "loss": 0.0172, "step": 42471 }, { "epoch": 13.07, "learning_rate": 5.673800925260543e-06, "loss": 0.0118, "step": 42472 }, { "epoch": 13.07, "learning_rate": 5.673351702111949e-06, "loss": 0.0136, "step": 42473 }, { "epoch": 13.07, "learning_rate": 5.672902489705254e-06, "loss": 0.01, "step": 42474 }, { "epoch": 13.07, "learning_rate": 5.672453288041577e-06, "loss": 0.0252, "step": 42475 }, { "epoch": 13.07, "learning_rate": 5.672004097122033e-06, "loss": 0.0227, "step": 42476 }, { "epoch": 13.07, "learning_rate": 5.671554916947734e-06, "loss": 0.0147, "step": 42477 }, { "epoch": 13.07, "learning_rate": 5.671105747519794e-06, "loss": 0.0241, "step": 42478 }, { "epoch": 13.07, "learning_rate": 5.670656588839334e-06, "loss": 0.0214, "step": 42479 }, { "epoch": 13.07, "learning_rate": 5.6702074409074715e-06, "loss": 0.0119, "step": 42480 }, { "epoch": 13.07, "learning_rate": 5.669758303725314e-06, "loss": 0.0319, "step": 42481 }, { "epoch": 13.07, "learning_rate": 5.669309177293978e-06, "loss": 0.0145, "step": 42482 }, { "epoch": 13.07, "learning_rate": 5.6688600616145765e-06, "loss": 0.0196, "step": 42483 }, { "epoch": 13.07, "learning_rate": 5.6684109566882305e-06, "loss": 0.0186, "step": 42484 }, { "epoch": 13.07, "learning_rate": 5.667961862516054e-06, "loss": 0.0265, "step": 42485 }, { "epoch": 13.07, "learning_rate": 5.667512779099158e-06, "loss": 0.0155, "step": 42486 }, { "epoch": 13.07, "learning_rate": 5.667063706438661e-06, "loss": 0.0249, "step": 42487 }, { "epoch": 13.07, "learning_rate": 5.6666146445356775e-06, "loss": 0.0227, "step": 42488 }, { "epoch": 13.07, "learning_rate": 5.666165593391322e-06, "loss": 0.0144, "step": 42489 }, { "epoch": 13.07, "learning_rate": 5.665716553006707e-06, "loss": 0.0205, "step": 42490 }, { "epoch": 13.07, "learning_rate": 5.66526752338295e-06, "loss": 0.0155, "step": 42491 }, { "epoch": 13.07, "learning_rate": 5.6648185045211656e-06, "loss": 0.0123, "step": 42492 }, { "epoch": 13.07, "learning_rate": 5.664369496422463e-06, "loss": 0.0233, "step": 42493 }, { "epoch": 13.07, "learning_rate": 5.663920499087971e-06, "loss": 0.0146, "step": 42494 }, { "epoch": 13.08, "learning_rate": 5.6634715125187925e-06, "loss": 0.0173, "step": 42495 }, { "epoch": 13.08, "learning_rate": 5.663022536716044e-06, "loss": 0.0115, "step": 42496 }, { "epoch": 13.08, "learning_rate": 5.662573571680841e-06, "loss": 0.0181, "step": 42497 }, { "epoch": 13.08, "learning_rate": 5.662124617414295e-06, "loss": 0.0198, "step": 42498 }, { "epoch": 13.08, "learning_rate": 5.661675673917529e-06, "loss": 0.0344, "step": 42499 }, { "epoch": 13.08, "learning_rate": 5.661226741191651e-06, "loss": 0.0238, "step": 42500 }, { "epoch": 13.08, "learning_rate": 5.660777819237782e-06, "loss": 0.035, "step": 42501 }, { "epoch": 13.08, "learning_rate": 5.6603289080570274e-06, "loss": 0.021, "step": 42502 }, { "epoch": 13.08, "learning_rate": 5.6598800076505025e-06, "loss": 0.0252, "step": 42503 }, { "epoch": 13.08, "learning_rate": 5.659431118019329e-06, "loss": 0.0173, "step": 42504 }, { "epoch": 13.08, "learning_rate": 5.65898223916462e-06, "loss": 0.0139, "step": 42505 }, { "epoch": 13.08, "learning_rate": 5.658533371087487e-06, "loss": 0.0355, "step": 42506 }, { "epoch": 13.08, "learning_rate": 5.6580845137890435e-06, "loss": 0.0161, "step": 42507 }, { "epoch": 13.08, "learning_rate": 5.6576356672704065e-06, "loss": 0.0313, "step": 42508 }, { "epoch": 13.08, "learning_rate": 5.6571868315326906e-06, "loss": 0.0184, "step": 42509 }, { "epoch": 13.08, "learning_rate": 5.656738006577007e-06, "loss": 0.0466, "step": 42510 }, { "epoch": 13.08, "learning_rate": 5.656289192404474e-06, "loss": 0.0227, "step": 42511 }, { "epoch": 13.08, "learning_rate": 5.655840389016203e-06, "loss": 0.0168, "step": 42512 }, { "epoch": 13.08, "learning_rate": 5.655391596413304e-06, "loss": 0.0125, "step": 42513 }, { "epoch": 13.08, "learning_rate": 5.654942814596902e-06, "loss": 0.0205, "step": 42514 }, { "epoch": 13.08, "learning_rate": 5.654494043568109e-06, "loss": 0.0226, "step": 42515 }, { "epoch": 13.08, "learning_rate": 5.654045283328031e-06, "loss": 0.018, "step": 42516 }, { "epoch": 13.08, "learning_rate": 5.653596533877789e-06, "loss": 0.0194, "step": 42517 }, { "epoch": 13.08, "learning_rate": 5.65314779521849e-06, "loss": 0.0118, "step": 42518 }, { "epoch": 13.08, "learning_rate": 5.652699067351257e-06, "loss": 0.0206, "step": 42519 }, { "epoch": 13.08, "learning_rate": 5.6522503502772e-06, "loss": 0.0193, "step": 42520 }, { "epoch": 13.08, "learning_rate": 5.651801643997433e-06, "loss": 0.0329, "step": 42521 }, { "epoch": 13.08, "learning_rate": 5.651352948513075e-06, "loss": 0.0134, "step": 42522 }, { "epoch": 13.08, "learning_rate": 5.650904263825227e-06, "loss": 0.0213, "step": 42523 }, { "epoch": 13.08, "learning_rate": 5.650455589935015e-06, "loss": 0.0084, "step": 42524 }, { "epoch": 13.08, "learning_rate": 5.65000692684355e-06, "loss": 0.0252, "step": 42525 }, { "epoch": 13.08, "learning_rate": 5.6495582745519454e-06, "loss": 0.0119, "step": 42526 }, { "epoch": 13.09, "learning_rate": 5.649109633061313e-06, "loss": 0.0169, "step": 42527 }, { "epoch": 13.09, "learning_rate": 5.648661002372769e-06, "loss": 0.0406, "step": 42528 }, { "epoch": 13.09, "learning_rate": 5.648212382487427e-06, "loss": 0.0333, "step": 42529 }, { "epoch": 13.09, "learning_rate": 5.647763773406401e-06, "loss": 0.0137, "step": 42530 }, { "epoch": 13.09, "learning_rate": 5.647315175130803e-06, "loss": 0.0078, "step": 42531 }, { "epoch": 13.09, "learning_rate": 5.646866587661748e-06, "loss": 0.0128, "step": 42532 }, { "epoch": 13.09, "learning_rate": 5.646418011000347e-06, "loss": 0.0279, "step": 42533 }, { "epoch": 13.09, "learning_rate": 5.64596944514772e-06, "loss": 0.0143, "step": 42534 }, { "epoch": 13.09, "learning_rate": 5.645520890104976e-06, "loss": 0.022, "step": 42535 }, { "epoch": 13.09, "learning_rate": 5.6450723458732345e-06, "loss": 0.0172, "step": 42536 }, { "epoch": 13.09, "learning_rate": 5.6446238124536e-06, "loss": 0.0222, "step": 42537 }, { "epoch": 13.09, "learning_rate": 5.644175289847187e-06, "loss": 0.0272, "step": 42538 }, { "epoch": 13.09, "learning_rate": 5.643726778055117e-06, "loss": 0.0406, "step": 42539 }, { "epoch": 13.09, "learning_rate": 5.643278277078497e-06, "loss": 0.021, "step": 42540 }, { "epoch": 13.09, "learning_rate": 5.6428297869184444e-06, "loss": 0.0281, "step": 42541 }, { "epoch": 13.09, "learning_rate": 5.64238130757607e-06, "loss": 0.02, "step": 42542 }, { "epoch": 13.09, "learning_rate": 5.641932839052488e-06, "loss": 0.0119, "step": 42543 }, { "epoch": 13.09, "learning_rate": 5.641484381348812e-06, "loss": 0.014, "step": 42544 }, { "epoch": 13.09, "learning_rate": 5.641035934466154e-06, "loss": 0.0304, "step": 42545 }, { "epoch": 13.09, "learning_rate": 5.64058749840563e-06, "loss": 0.0193, "step": 42546 }, { "epoch": 13.09, "learning_rate": 5.640139073168351e-06, "loss": 0.0256, "step": 42547 }, { "epoch": 13.09, "learning_rate": 5.639690658755429e-06, "loss": 0.0147, "step": 42548 }, { "epoch": 13.09, "learning_rate": 5.639242255167987e-06, "loss": 0.0251, "step": 42549 }, { "epoch": 13.09, "learning_rate": 5.638793862407126e-06, "loss": 0.0091, "step": 42550 }, { "epoch": 13.09, "learning_rate": 5.6383454804739654e-06, "loss": 0.0167, "step": 42551 }, { "epoch": 13.09, "learning_rate": 5.637897109369617e-06, "loss": 0.0195, "step": 42552 }, { "epoch": 13.09, "learning_rate": 5.637448749095189e-06, "loss": 0.0258, "step": 42553 }, { "epoch": 13.09, "learning_rate": 5.637000399651804e-06, "loss": 0.0344, "step": 42554 }, { "epoch": 13.09, "learning_rate": 5.63655206104057e-06, "loss": 0.0138, "step": 42555 }, { "epoch": 13.09, "learning_rate": 5.636103733262601e-06, "loss": 0.0222, "step": 42556 }, { "epoch": 13.09, "learning_rate": 5.635655416319016e-06, "loss": 0.0309, "step": 42557 }, { "epoch": 13.09, "learning_rate": 5.635207110210912e-06, "loss": 0.0549, "step": 42558 }, { "epoch": 13.09, "learning_rate": 5.634758814939417e-06, "loss": 0.0148, "step": 42559 }, { "epoch": 13.1, "learning_rate": 5.634310530505639e-06, "loss": 0.0278, "step": 42560 }, { "epoch": 13.1, "learning_rate": 5.633862256910691e-06, "loss": 0.0108, "step": 42561 }, { "epoch": 13.1, "learning_rate": 5.633413994155685e-06, "loss": 0.0199, "step": 42562 }, { "epoch": 13.1, "learning_rate": 5.632965742241736e-06, "loss": 0.0265, "step": 42563 }, { "epoch": 13.1, "learning_rate": 5.632517501169955e-06, "loss": 0.0283, "step": 42564 }, { "epoch": 13.1, "learning_rate": 5.632069270941456e-06, "loss": 0.0145, "step": 42565 }, { "epoch": 13.1, "learning_rate": 5.631621051557351e-06, "loss": 0.0147, "step": 42566 }, { "epoch": 13.1, "learning_rate": 5.631172843018754e-06, "loss": 0.0253, "step": 42567 }, { "epoch": 13.1, "learning_rate": 5.630724645326773e-06, "loss": 0.0194, "step": 42568 }, { "epoch": 13.1, "learning_rate": 5.6302764584825285e-06, "loss": 0.0308, "step": 42569 }, { "epoch": 13.1, "learning_rate": 5.6298282824871335e-06, "loss": 0.028, "step": 42570 }, { "epoch": 13.1, "learning_rate": 5.629380117341692e-06, "loss": 0.0134, "step": 42571 }, { "epoch": 13.1, "learning_rate": 5.628931963047322e-06, "loss": 0.0238, "step": 42572 }, { "epoch": 13.1, "learning_rate": 5.628483819605132e-06, "loss": 0.0108, "step": 42573 }, { "epoch": 13.1, "learning_rate": 5.6280356870162416e-06, "loss": 0.0341, "step": 42574 }, { "epoch": 13.1, "learning_rate": 5.62758756528176e-06, "loss": 0.0162, "step": 42575 }, { "epoch": 13.1, "learning_rate": 5.627139454402799e-06, "loss": 0.0159, "step": 42576 }, { "epoch": 13.1, "learning_rate": 5.626691354380473e-06, "loss": 0.0114, "step": 42577 }, { "epoch": 13.1, "learning_rate": 5.626243265215893e-06, "loss": 0.0244, "step": 42578 }, { "epoch": 13.1, "learning_rate": 5.625795186910172e-06, "loss": 0.022, "step": 42579 }, { "epoch": 13.1, "learning_rate": 5.625347119464422e-06, "loss": 0.0489, "step": 42580 }, { "epoch": 13.1, "learning_rate": 5.624899062879756e-06, "loss": 0.0191, "step": 42581 }, { "epoch": 13.1, "learning_rate": 5.624451017157286e-06, "loss": 0.0154, "step": 42582 }, { "epoch": 13.1, "learning_rate": 5.624002982298125e-06, "loss": 0.0134, "step": 42583 }, { "epoch": 13.1, "learning_rate": 5.623554958303384e-06, "loss": 0.023, "step": 42584 }, { "epoch": 13.1, "learning_rate": 5.623106945174176e-06, "loss": 0.0156, "step": 42585 }, { "epoch": 13.1, "learning_rate": 5.622658942911614e-06, "loss": 0.0226, "step": 42586 }, { "epoch": 13.1, "learning_rate": 5.62221095151681e-06, "loss": 0.0153, "step": 42587 }, { "epoch": 13.1, "learning_rate": 5.621762970990871e-06, "loss": 0.0201, "step": 42588 }, { "epoch": 13.1, "learning_rate": 5.62131500133492e-06, "loss": 0.0273, "step": 42589 }, { "epoch": 13.1, "learning_rate": 5.620867042550062e-06, "loss": 0.0207, "step": 42590 }, { "epoch": 13.1, "learning_rate": 5.620419094637415e-06, "loss": 0.0101, "step": 42591 }, { "epoch": 13.11, "learning_rate": 5.619971157598083e-06, "loss": 0.0221, "step": 42592 }, { "epoch": 13.11, "learning_rate": 5.619523231433177e-06, "loss": 0.0101, "step": 42593 }, { "epoch": 13.11, "learning_rate": 5.6190753161438185e-06, "loss": 0.0241, "step": 42594 }, { "epoch": 13.11, "learning_rate": 5.618627411731114e-06, "loss": 0.0105, "step": 42595 }, { "epoch": 13.11, "learning_rate": 5.618179518196177e-06, "loss": 0.0158, "step": 42596 }, { "epoch": 13.11, "learning_rate": 5.61773163554012e-06, "loss": 0.0155, "step": 42597 }, { "epoch": 13.11, "learning_rate": 5.617283763764052e-06, "loss": 0.0313, "step": 42598 }, { "epoch": 13.11, "learning_rate": 5.616835902869089e-06, "loss": 0.0653, "step": 42599 }, { "epoch": 13.11, "learning_rate": 5.616388052856339e-06, "loss": 0.0122, "step": 42600 }, { "epoch": 13.11, "learning_rate": 5.615940213726917e-06, "loss": 0.034, "step": 42601 }, { "epoch": 13.11, "learning_rate": 5.6154923854819334e-06, "loss": 0.0128, "step": 42602 }, { "epoch": 13.11, "learning_rate": 5.615044568122495e-06, "loss": 0.0198, "step": 42603 }, { "epoch": 13.11, "learning_rate": 5.614596761649728e-06, "loss": 0.028, "step": 42604 }, { "epoch": 13.11, "learning_rate": 5.6141489660647295e-06, "loss": 0.0204, "step": 42605 }, { "epoch": 13.11, "learning_rate": 5.613701181368618e-06, "loss": 0.0248, "step": 42606 }, { "epoch": 13.11, "learning_rate": 5.613253407562505e-06, "loss": 0.0326, "step": 42607 }, { "epoch": 13.11, "learning_rate": 5.6128056446474944e-06, "loss": 0.0239, "step": 42608 }, { "epoch": 13.11, "learning_rate": 5.6123578926247105e-06, "loss": 0.0206, "step": 42609 }, { "epoch": 13.11, "learning_rate": 5.611910151495259e-06, "loss": 0.012, "step": 42610 }, { "epoch": 13.11, "learning_rate": 5.611462421260251e-06, "loss": 0.0171, "step": 42611 }, { "epoch": 13.11, "learning_rate": 5.611014701920803e-06, "loss": 0.0147, "step": 42612 }, { "epoch": 13.11, "learning_rate": 5.610566993478014e-06, "loss": 0.0199, "step": 42613 }, { "epoch": 13.11, "learning_rate": 5.610119295933008e-06, "loss": 0.0217, "step": 42614 }, { "epoch": 13.11, "learning_rate": 5.609671609286892e-06, "loss": 0.0435, "step": 42615 }, { "epoch": 13.11, "learning_rate": 5.609223933540778e-06, "loss": 0.0214, "step": 42616 }, { "epoch": 13.11, "learning_rate": 5.608776268695777e-06, "loss": 0.0285, "step": 42617 }, { "epoch": 13.11, "learning_rate": 5.608328614753001e-06, "loss": 0.0183, "step": 42618 }, { "epoch": 13.11, "learning_rate": 5.6078809717135615e-06, "loss": 0.0291, "step": 42619 }, { "epoch": 13.11, "learning_rate": 5.60743333957857e-06, "loss": 0.0186, "step": 42620 }, { "epoch": 13.11, "learning_rate": 5.606985718349136e-06, "loss": 0.0101, "step": 42621 }, { "epoch": 13.11, "learning_rate": 5.606538108026372e-06, "loss": 0.0212, "step": 42622 }, { "epoch": 13.11, "learning_rate": 5.606090508611386e-06, "loss": 0.0199, "step": 42623 }, { "epoch": 13.11, "learning_rate": 5.605642920105296e-06, "loss": 0.0293, "step": 42624 }, { "epoch": 13.12, "learning_rate": 5.6051953425092154e-06, "loss": 0.0228, "step": 42625 }, { "epoch": 13.12, "learning_rate": 5.604747775824245e-06, "loss": 0.0186, "step": 42626 }, { "epoch": 13.12, "learning_rate": 5.6043002200515015e-06, "loss": 0.0233, "step": 42627 }, { "epoch": 13.12, "learning_rate": 5.603852675192092e-06, "loss": 0.04, "step": 42628 }, { "epoch": 13.12, "learning_rate": 5.603405141247133e-06, "loss": 0.0175, "step": 42629 }, { "epoch": 13.12, "learning_rate": 5.602957618217735e-06, "loss": 0.0272, "step": 42630 }, { "epoch": 13.12, "learning_rate": 5.602510106105008e-06, "loss": 0.0203, "step": 42631 }, { "epoch": 13.12, "learning_rate": 5.602062604910064e-06, "loss": 0.0228, "step": 42632 }, { "epoch": 13.12, "learning_rate": 5.601615114634011e-06, "loss": 0.0262, "step": 42633 }, { "epoch": 13.12, "learning_rate": 5.601167635277962e-06, "loss": 0.0322, "step": 42634 }, { "epoch": 13.12, "learning_rate": 5.600720166843029e-06, "loss": 0.0114, "step": 42635 }, { "epoch": 13.12, "learning_rate": 5.600272709330321e-06, "loss": 0.0104, "step": 42636 }, { "epoch": 13.12, "learning_rate": 5.59982526274095e-06, "loss": 0.016, "step": 42637 }, { "epoch": 13.12, "learning_rate": 5.5993778270760266e-06, "loss": 0.018, "step": 42638 }, { "epoch": 13.12, "learning_rate": 5.5989304023366615e-06, "loss": 0.0338, "step": 42639 }, { "epoch": 13.12, "learning_rate": 5.5984829885239665e-06, "loss": 0.0204, "step": 42640 }, { "epoch": 13.12, "learning_rate": 5.598035585639051e-06, "loss": 0.0243, "step": 42641 }, { "epoch": 13.12, "learning_rate": 5.597588193683026e-06, "loss": 0.0161, "step": 42642 }, { "epoch": 13.12, "learning_rate": 5.597140812656999e-06, "loss": 0.0196, "step": 42643 }, { "epoch": 13.12, "learning_rate": 5.596693442562089e-06, "loss": 0.023, "step": 42644 }, { "epoch": 13.12, "learning_rate": 5.596246083399402e-06, "loss": 0.0179, "step": 42645 }, { "epoch": 13.12, "learning_rate": 5.595798735170053e-06, "loss": 0.0186, "step": 42646 }, { "epoch": 13.12, "learning_rate": 5.595351397875143e-06, "loss": 0.0146, "step": 42647 }, { "epoch": 13.12, "learning_rate": 5.594904071515784e-06, "loss": 0.0308, "step": 42648 }, { "epoch": 13.12, "learning_rate": 5.594456756093096e-06, "loss": 0.0191, "step": 42649 }, { "epoch": 13.12, "learning_rate": 5.594009451608183e-06, "loss": 0.0158, "step": 42650 }, { "epoch": 13.12, "learning_rate": 5.593562158062158e-06, "loss": 0.0148, "step": 42651 }, { "epoch": 13.12, "learning_rate": 5.5931148754561295e-06, "loss": 0.0196, "step": 42652 }, { "epoch": 13.12, "learning_rate": 5.5926676037912086e-06, "loss": 0.0123, "step": 42653 }, { "epoch": 13.12, "learning_rate": 5.5922203430685065e-06, "loss": 0.0191, "step": 42654 }, { "epoch": 13.12, "learning_rate": 5.591773093289132e-06, "loss": 0.0147, "step": 42655 }, { "epoch": 13.12, "learning_rate": 5.591325854454197e-06, "loss": 0.0214, "step": 42656 }, { "epoch": 13.13, "learning_rate": 5.59087862656481e-06, "loss": 0.0213, "step": 42657 }, { "epoch": 13.13, "learning_rate": 5.590431409622081e-06, "loss": 0.0442, "step": 42658 }, { "epoch": 13.13, "learning_rate": 5.589984203627129e-06, "loss": 0.0172, "step": 42659 }, { "epoch": 13.13, "learning_rate": 5.5895370085810544e-06, "loss": 0.027, "step": 42660 }, { "epoch": 13.13, "learning_rate": 5.589089824484968e-06, "loss": 0.0214, "step": 42661 }, { "epoch": 13.13, "learning_rate": 5.588642651339984e-06, "loss": 0.0269, "step": 42662 }, { "epoch": 13.13, "learning_rate": 5.588195489147208e-06, "loss": 0.0126, "step": 42663 }, { "epoch": 13.13, "learning_rate": 5.587748337907754e-06, "loss": 0.0167, "step": 42664 }, { "epoch": 13.13, "learning_rate": 5.587301197622734e-06, "loss": 0.0227, "step": 42665 }, { "epoch": 13.13, "learning_rate": 5.586854068293255e-06, "loss": 0.0313, "step": 42666 }, { "epoch": 13.13, "learning_rate": 5.586406949920432e-06, "loss": 0.0117, "step": 42667 }, { "epoch": 13.13, "learning_rate": 5.585959842505362e-06, "loss": 0.021, "step": 42668 }, { "epoch": 13.13, "learning_rate": 5.585512746049169e-06, "loss": 0.0096, "step": 42669 }, { "epoch": 13.13, "learning_rate": 5.5850656605529554e-06, "loss": 0.0087, "step": 42670 }, { "epoch": 13.13, "learning_rate": 5.5846185860178344e-06, "loss": 0.0228, "step": 42671 }, { "epoch": 13.13, "learning_rate": 5.584171522444916e-06, "loss": 0.017, "step": 42672 }, { "epoch": 13.13, "learning_rate": 5.5837244698353085e-06, "loss": 0.0279, "step": 42673 }, { "epoch": 13.13, "learning_rate": 5.583277428190124e-06, "loss": 0.0112, "step": 42674 }, { "epoch": 13.13, "learning_rate": 5.582830397510468e-06, "loss": 0.0238, "step": 42675 }, { "epoch": 13.13, "learning_rate": 5.5823833777974555e-06, "loss": 0.0151, "step": 42676 }, { "epoch": 13.13, "learning_rate": 5.581936369052194e-06, "loss": 0.0367, "step": 42677 }, { "epoch": 13.13, "learning_rate": 5.581489371275789e-06, "loss": 0.0192, "step": 42678 }, { "epoch": 13.13, "learning_rate": 5.581042384469359e-06, "loss": 0.0102, "step": 42679 }, { "epoch": 13.13, "learning_rate": 5.580595408634013e-06, "loss": 0.0156, "step": 42680 }, { "epoch": 13.13, "learning_rate": 5.580148443770853e-06, "loss": 0.0329, "step": 42681 }, { "epoch": 13.13, "learning_rate": 5.579701489880993e-06, "loss": 0.0214, "step": 42682 }, { "epoch": 13.13, "learning_rate": 5.579254546965539e-06, "loss": 0.0264, "step": 42683 }, { "epoch": 13.13, "learning_rate": 5.5788076150256075e-06, "loss": 0.0304, "step": 42684 }, { "epoch": 13.13, "learning_rate": 5.578360694062306e-06, "loss": 0.0368, "step": 42685 }, { "epoch": 13.13, "learning_rate": 5.577913784076741e-06, "loss": 0.0142, "step": 42686 }, { "epoch": 13.13, "learning_rate": 5.5774668850700255e-06, "loss": 0.0455, "step": 42687 }, { "epoch": 13.13, "learning_rate": 5.577019997043267e-06, "loss": 0.0181, "step": 42688 }, { "epoch": 13.13, "learning_rate": 5.576573119997575e-06, "loss": 0.0165, "step": 42689 }, { "epoch": 13.14, "learning_rate": 5.57612625393406e-06, "loss": 0.0333, "step": 42690 }, { "epoch": 13.14, "learning_rate": 5.57567939885383e-06, "loss": 0.021, "step": 42691 }, { "epoch": 13.14, "learning_rate": 5.575232554757997e-06, "loss": 0.024, "step": 42692 }, { "epoch": 13.14, "learning_rate": 5.574785721647663e-06, "loss": 0.0167, "step": 42693 }, { "epoch": 13.14, "learning_rate": 5.5743388995239525e-06, "loss": 0.0117, "step": 42694 }, { "epoch": 13.14, "learning_rate": 5.57389208838796e-06, "loss": 0.019, "step": 42695 }, { "epoch": 13.14, "learning_rate": 5.5734452882407994e-06, "loss": 0.0363, "step": 42696 }, { "epoch": 13.14, "learning_rate": 5.572998499083583e-06, "loss": 0.0119, "step": 42697 }, { "epoch": 13.14, "learning_rate": 5.572551720917413e-06, "loss": 0.0194, "step": 42698 }, { "epoch": 13.14, "learning_rate": 5.572104953743407e-06, "loss": 0.0098, "step": 42699 }, { "epoch": 13.14, "learning_rate": 5.5716581975626715e-06, "loss": 0.028, "step": 42700 }, { "epoch": 13.14, "learning_rate": 5.571211452376319e-06, "loss": 0.0178, "step": 42701 }, { "epoch": 13.14, "learning_rate": 5.57076471818545e-06, "loss": 0.0218, "step": 42702 }, { "epoch": 13.14, "learning_rate": 5.570317994991173e-06, "loss": 0.0195, "step": 42703 }, { "epoch": 13.14, "learning_rate": 5.569871282794608e-06, "loss": 0.0095, "step": 42704 }, { "epoch": 13.14, "learning_rate": 5.569424581596857e-06, "loss": 0.0179, "step": 42705 }, { "epoch": 13.14, "learning_rate": 5.568977891399032e-06, "loss": 0.0144, "step": 42706 }, { "epoch": 13.14, "learning_rate": 5.5685312122022396e-06, "loss": 0.0145, "step": 42707 }, { "epoch": 13.14, "learning_rate": 5.5680845440075885e-06, "loss": 0.0321, "step": 42708 }, { "epoch": 13.14, "learning_rate": 5.5676378868161885e-06, "loss": 0.0165, "step": 42709 }, { "epoch": 13.14, "learning_rate": 5.567191240629151e-06, "loss": 0.0151, "step": 42710 }, { "epoch": 13.14, "learning_rate": 5.56674460544758e-06, "loss": 0.0147, "step": 42711 }, { "epoch": 13.14, "learning_rate": 5.5662979812725885e-06, "loss": 0.0193, "step": 42712 }, { "epoch": 13.14, "learning_rate": 5.5658513681052796e-06, "loss": 0.0372, "step": 42713 }, { "epoch": 13.14, "learning_rate": 5.565404765946769e-06, "loss": 0.0207, "step": 42714 }, { "epoch": 13.14, "learning_rate": 5.5649581747981695e-06, "loss": 0.0268, "step": 42715 }, { "epoch": 13.14, "learning_rate": 5.5645115946605775e-06, "loss": 0.0615, "step": 42716 }, { "epoch": 13.14, "learning_rate": 5.564065025535107e-06, "loss": 0.0129, "step": 42717 }, { "epoch": 13.14, "learning_rate": 5.563618467422864e-06, "loss": 0.0117, "step": 42718 }, { "epoch": 13.14, "learning_rate": 5.5631719203249635e-06, "loss": 0.0273, "step": 42719 }, { "epoch": 13.14, "learning_rate": 5.562725384242512e-06, "loss": 0.0209, "step": 42720 }, { "epoch": 13.14, "learning_rate": 5.562278859176617e-06, "loss": 0.0175, "step": 42721 }, { "epoch": 13.15, "learning_rate": 5.5618323451283905e-06, "loss": 0.0185, "step": 42722 }, { "epoch": 13.15, "learning_rate": 5.56138584209893e-06, "loss": 0.026, "step": 42723 }, { "epoch": 13.15, "learning_rate": 5.560939350089356e-06, "loss": 0.0322, "step": 42724 }, { "epoch": 13.15, "learning_rate": 5.560492869100773e-06, "loss": 0.0358, "step": 42725 }, { "epoch": 13.15, "learning_rate": 5.56004639913429e-06, "loss": 0.0158, "step": 42726 }, { "epoch": 13.15, "learning_rate": 5.559599940191014e-06, "loss": 0.0286, "step": 42727 }, { "epoch": 13.15, "learning_rate": 5.559153492272054e-06, "loss": 0.0247, "step": 42728 }, { "epoch": 13.15, "learning_rate": 5.558707055378519e-06, "loss": 0.0248, "step": 42729 }, { "epoch": 13.15, "learning_rate": 5.558260629511518e-06, "loss": 0.0154, "step": 42730 }, { "epoch": 13.15, "learning_rate": 5.557814214672157e-06, "loss": 0.0108, "step": 42731 }, { "epoch": 13.15, "learning_rate": 5.557367810861547e-06, "loss": 0.0164, "step": 42732 }, { "epoch": 13.15, "learning_rate": 5.55692141808079e-06, "loss": 0.0153, "step": 42733 }, { "epoch": 13.15, "learning_rate": 5.556475036331002e-06, "loss": 0.0157, "step": 42734 }, { "epoch": 13.15, "learning_rate": 5.556028665613291e-06, "loss": 0.0188, "step": 42735 }, { "epoch": 13.15, "learning_rate": 5.555582305928766e-06, "loss": 0.0176, "step": 42736 }, { "epoch": 13.15, "learning_rate": 5.555135957278528e-06, "loss": 0.0224, "step": 42737 }, { "epoch": 13.15, "learning_rate": 5.554689619663684e-06, "loss": 0.0136, "step": 42738 }, { "epoch": 13.15, "learning_rate": 5.554243293085353e-06, "loss": 0.0148, "step": 42739 }, { "epoch": 13.15, "learning_rate": 5.553796977544637e-06, "loss": 0.0149, "step": 42740 }, { "epoch": 13.15, "learning_rate": 5.553350673042643e-06, "loss": 0.0261, "step": 42741 }, { "epoch": 13.15, "learning_rate": 5.5529043795804815e-06, "loss": 0.0723, "step": 42742 }, { "epoch": 13.15, "learning_rate": 5.55245809715926e-06, "loss": 0.0133, "step": 42743 }, { "epoch": 13.15, "learning_rate": 5.5520118257800855e-06, "loss": 0.0254, "step": 42744 }, { "epoch": 13.15, "learning_rate": 5.551565565444066e-06, "loss": 0.0175, "step": 42745 }, { "epoch": 13.15, "learning_rate": 5.551119316152311e-06, "loss": 0.0195, "step": 42746 }, { "epoch": 13.15, "learning_rate": 5.550673077905926e-06, "loss": 0.0287, "step": 42747 }, { "epoch": 13.15, "learning_rate": 5.550226850706018e-06, "loss": 0.0159, "step": 42748 }, { "epoch": 13.15, "learning_rate": 5.549780634553704e-06, "loss": 0.019, "step": 42749 }, { "epoch": 13.15, "learning_rate": 5.549334429450082e-06, "loss": 0.0375, "step": 42750 }, { "epoch": 13.15, "learning_rate": 5.548888235396262e-06, "loss": 0.0285, "step": 42751 }, { "epoch": 13.15, "learning_rate": 5.548442052393352e-06, "loss": 0.0217, "step": 42752 }, { "epoch": 13.15, "learning_rate": 5.547995880442456e-06, "loss": 0.019, "step": 42753 }, { "epoch": 13.15, "learning_rate": 5.547549719544692e-06, "loss": 0.0202, "step": 42754 }, { "epoch": 13.16, "learning_rate": 5.54710356970116e-06, "loss": 0.0236, "step": 42755 }, { "epoch": 13.16, "learning_rate": 5.5466574309129725e-06, "loss": 0.0279, "step": 42756 }, { "epoch": 13.16, "learning_rate": 5.546211303181228e-06, "loss": 0.0307, "step": 42757 }, { "epoch": 13.16, "learning_rate": 5.545765186507044e-06, "loss": 0.0201, "step": 42758 }, { "epoch": 13.16, "learning_rate": 5.545319080891522e-06, "loss": 0.0167, "step": 42759 }, { "epoch": 13.16, "learning_rate": 5.544872986335774e-06, "loss": 0.041, "step": 42760 }, { "epoch": 13.16, "learning_rate": 5.544426902840905e-06, "loss": 0.035, "step": 42761 }, { "epoch": 13.16, "learning_rate": 5.5439808304080225e-06, "loss": 0.0405, "step": 42762 }, { "epoch": 13.16, "learning_rate": 5.543534769038235e-06, "loss": 0.0193, "step": 42763 }, { "epoch": 13.16, "learning_rate": 5.543088718732647e-06, "loss": 0.0238, "step": 42764 }, { "epoch": 13.16, "learning_rate": 5.5426426794923705e-06, "loss": 0.0261, "step": 42765 }, { "epoch": 13.16, "learning_rate": 5.54219665131851e-06, "loss": 0.0143, "step": 42766 }, { "epoch": 13.16, "learning_rate": 5.5417506342121685e-06, "loss": 0.0095, "step": 42767 }, { "epoch": 13.16, "learning_rate": 5.541304628174464e-06, "loss": 0.0309, "step": 42768 }, { "epoch": 13.16, "learning_rate": 5.540858633206496e-06, "loss": 0.0231, "step": 42769 }, { "epoch": 13.16, "learning_rate": 5.540412649309379e-06, "loss": 0.0259, "step": 42770 }, { "epoch": 13.16, "learning_rate": 5.539966676484212e-06, "loss": 0.0069, "step": 42771 }, { "epoch": 13.16, "learning_rate": 5.5395207147321e-06, "loss": 0.0793, "step": 42772 }, { "epoch": 13.16, "learning_rate": 5.53907476405416e-06, "loss": 0.0282, "step": 42773 }, { "epoch": 13.16, "learning_rate": 5.538628824451495e-06, "loss": 0.025, "step": 42774 }, { "epoch": 13.16, "learning_rate": 5.538182895925212e-06, "loss": 0.0276, "step": 42775 }, { "epoch": 13.16, "learning_rate": 5.537736978476417e-06, "loss": 0.0248, "step": 42776 }, { "epoch": 13.16, "learning_rate": 5.537291072106219e-06, "loss": 0.0461, "step": 42777 }, { "epoch": 13.16, "learning_rate": 5.536845176815724e-06, "loss": 0.0177, "step": 42778 }, { "epoch": 13.16, "learning_rate": 5.536399292606038e-06, "loss": 0.0083, "step": 42779 }, { "epoch": 13.16, "learning_rate": 5.53595341947827e-06, "loss": 0.0215, "step": 42780 }, { "epoch": 13.16, "learning_rate": 5.535507557433527e-06, "loss": 0.0302, "step": 42781 }, { "epoch": 13.16, "learning_rate": 5.535061706472911e-06, "loss": 0.0211, "step": 42782 }, { "epoch": 13.16, "learning_rate": 5.53461586659754e-06, "loss": 0.0233, "step": 42783 }, { "epoch": 13.16, "learning_rate": 5.534170037808511e-06, "loss": 0.0247, "step": 42784 }, { "epoch": 13.16, "learning_rate": 5.533724220106933e-06, "loss": 0.0079, "step": 42785 }, { "epoch": 13.16, "learning_rate": 5.533278413493915e-06, "loss": 0.0224, "step": 42786 }, { "epoch": 13.17, "learning_rate": 5.532832617970557e-06, "loss": 0.0153, "step": 42787 }, { "epoch": 13.17, "learning_rate": 5.5323868335379775e-06, "loss": 0.0208, "step": 42788 }, { "epoch": 13.17, "learning_rate": 5.531941060197275e-06, "loss": 0.0408, "step": 42789 }, { "epoch": 13.17, "learning_rate": 5.531495297949558e-06, "loss": 0.0344, "step": 42790 }, { "epoch": 13.17, "learning_rate": 5.531049546795939e-06, "loss": 0.0176, "step": 42791 }, { "epoch": 13.17, "learning_rate": 5.53060380673751e-06, "loss": 0.013, "step": 42792 }, { "epoch": 13.17, "learning_rate": 5.530158077775391e-06, "loss": 0.0165, "step": 42793 }, { "epoch": 13.17, "learning_rate": 5.529712359910685e-06, "loss": 0.0157, "step": 42794 }, { "epoch": 13.17, "learning_rate": 5.5292666531444985e-06, "loss": 0.0192, "step": 42795 }, { "epoch": 13.17, "learning_rate": 5.528820957477936e-06, "loss": 0.0118, "step": 42796 }, { "epoch": 13.17, "learning_rate": 5.5283752729121075e-06, "loss": 0.0214, "step": 42797 }, { "epoch": 13.17, "learning_rate": 5.527929599448116e-06, "loss": 0.0171, "step": 42798 }, { "epoch": 13.17, "learning_rate": 5.52748393708707e-06, "loss": 0.0227, "step": 42799 }, { "epoch": 13.17, "learning_rate": 5.5270382858300775e-06, "loss": 0.0259, "step": 42800 }, { "epoch": 13.17, "learning_rate": 5.526592645678241e-06, "loss": 0.0191, "step": 42801 }, { "epoch": 13.17, "learning_rate": 5.526147016632666e-06, "loss": 0.0145, "step": 42802 }, { "epoch": 13.17, "learning_rate": 5.525701398694465e-06, "loss": 0.0169, "step": 42803 }, { "epoch": 13.17, "learning_rate": 5.525255791864747e-06, "loss": 0.0193, "step": 42804 }, { "epoch": 13.17, "learning_rate": 5.5248101961446065e-06, "loss": 0.0124, "step": 42805 }, { "epoch": 13.17, "learning_rate": 5.524364611535157e-06, "loss": 0.0185, "step": 42806 }, { "epoch": 13.17, "learning_rate": 5.5239190380374986e-06, "loss": 0.0203, "step": 42807 }, { "epoch": 13.17, "learning_rate": 5.5234734756527475e-06, "loss": 0.0377, "step": 42808 }, { "epoch": 13.17, "learning_rate": 5.523027924382004e-06, "loss": 0.019, "step": 42809 }, { "epoch": 13.17, "learning_rate": 5.522582384226376e-06, "loss": 0.0283, "step": 42810 }, { "epoch": 13.17, "learning_rate": 5.522136855186968e-06, "loss": 0.0234, "step": 42811 }, { "epoch": 13.17, "learning_rate": 5.521691337264887e-06, "loss": 0.0384, "step": 42812 }, { "epoch": 13.17, "learning_rate": 5.52124583046124e-06, "loss": 0.03, "step": 42813 }, { "epoch": 13.17, "learning_rate": 5.520800334777132e-06, "loss": 0.0204, "step": 42814 }, { "epoch": 13.17, "learning_rate": 5.520354850213669e-06, "loss": 0.0209, "step": 42815 }, { "epoch": 13.17, "learning_rate": 5.519909376771957e-06, "loss": 0.0248, "step": 42816 }, { "epoch": 13.17, "learning_rate": 5.519463914453103e-06, "loss": 0.0118, "step": 42817 }, { "epoch": 13.17, "learning_rate": 5.519018463258212e-06, "loss": 0.0197, "step": 42818 }, { "epoch": 13.17, "learning_rate": 5.5185730231883895e-06, "loss": 0.0142, "step": 42819 }, { "epoch": 13.18, "learning_rate": 5.518127594244742e-06, "loss": 0.0237, "step": 42820 }, { "epoch": 13.18, "learning_rate": 5.517682176428376e-06, "loss": 0.0247, "step": 42821 }, { "epoch": 13.18, "learning_rate": 5.5172367697403925e-06, "loss": 0.0377, "step": 42822 }, { "epoch": 13.18, "learning_rate": 5.516791374181906e-06, "loss": 0.0279, "step": 42823 }, { "epoch": 13.18, "learning_rate": 5.516345989754018e-06, "loss": 0.0163, "step": 42824 }, { "epoch": 13.18, "learning_rate": 5.515900616457836e-06, "loss": 0.0157, "step": 42825 }, { "epoch": 13.18, "learning_rate": 5.515455254294461e-06, "loss": 0.0189, "step": 42826 }, { "epoch": 13.18, "learning_rate": 5.515009903264998e-06, "loss": 0.0082, "step": 42827 }, { "epoch": 13.18, "learning_rate": 5.514564563370562e-06, "loss": 0.0143, "step": 42828 }, { "epoch": 13.18, "learning_rate": 5.514119234612251e-06, "loss": 0.024, "step": 42829 }, { "epoch": 13.18, "learning_rate": 5.513673916991173e-06, "loss": 0.0342, "step": 42830 }, { "epoch": 13.18, "learning_rate": 5.513228610508432e-06, "loss": 0.0146, "step": 42831 }, { "epoch": 13.18, "learning_rate": 5.512783315165136e-06, "loss": 0.0237, "step": 42832 }, { "epoch": 13.18, "learning_rate": 5.512338030962389e-06, "loss": 0.02, "step": 42833 }, { "epoch": 13.18, "learning_rate": 5.5118927579012985e-06, "loss": 0.0266, "step": 42834 }, { "epoch": 13.18, "learning_rate": 5.511447495982967e-06, "loss": 0.0273, "step": 42835 }, { "epoch": 13.18, "learning_rate": 5.511002245208501e-06, "loss": 0.0103, "step": 42836 }, { "epoch": 13.18, "learning_rate": 5.510557005579001e-06, "loss": 0.0333, "step": 42837 }, { "epoch": 13.18, "learning_rate": 5.510111777095588e-06, "loss": 0.0209, "step": 42838 }, { "epoch": 13.18, "learning_rate": 5.509666559759353e-06, "loss": 0.0198, "step": 42839 }, { "epoch": 13.18, "learning_rate": 5.509221353571404e-06, "loss": 0.0197, "step": 42840 }, { "epoch": 13.18, "learning_rate": 5.508776158532847e-06, "loss": 0.0111, "step": 42841 }, { "epoch": 13.18, "learning_rate": 5.5083309746447845e-06, "loss": 0.0261, "step": 42842 }, { "epoch": 13.18, "learning_rate": 5.50788580190833e-06, "loss": 0.0126, "step": 42843 }, { "epoch": 13.18, "learning_rate": 5.507440640324584e-06, "loss": 0.0105, "step": 42844 }, { "epoch": 13.18, "learning_rate": 5.506995489894652e-06, "loss": 0.019, "step": 42845 }, { "epoch": 13.18, "learning_rate": 5.506550350619642e-06, "loss": 0.0276, "step": 42846 }, { "epoch": 13.18, "learning_rate": 5.506105222500649e-06, "loss": 0.0205, "step": 42847 }, { "epoch": 13.18, "learning_rate": 5.505660105538789e-06, "loss": 0.0113, "step": 42848 }, { "epoch": 13.18, "learning_rate": 5.505214999735164e-06, "loss": 0.0159, "step": 42849 }, { "epoch": 13.18, "learning_rate": 5.504769905090876e-06, "loss": 0.0161, "step": 42850 }, { "epoch": 13.18, "learning_rate": 5.504324821607035e-06, "loss": 0.0196, "step": 42851 }, { "epoch": 13.19, "learning_rate": 5.503879749284742e-06, "loss": 0.0452, "step": 42852 }, { "epoch": 13.19, "learning_rate": 5.503434688125104e-06, "loss": 0.0239, "step": 42853 }, { "epoch": 13.19, "learning_rate": 5.5029896381292254e-06, "loss": 0.0148, "step": 42854 }, { "epoch": 13.19, "learning_rate": 5.5025445992982114e-06, "loss": 0.0512, "step": 42855 }, { "epoch": 13.19, "learning_rate": 5.502099571633166e-06, "loss": 0.0343, "step": 42856 }, { "epoch": 13.19, "learning_rate": 5.501654555135192e-06, "loss": 0.0216, "step": 42857 }, { "epoch": 13.19, "learning_rate": 5.501209549805399e-06, "loss": 0.0164, "step": 42858 }, { "epoch": 13.19, "learning_rate": 5.5007645556448955e-06, "loss": 0.0353, "step": 42859 }, { "epoch": 13.19, "learning_rate": 5.500319572654777e-06, "loss": 0.0142, "step": 42860 }, { "epoch": 13.19, "learning_rate": 5.49987460083615e-06, "loss": 0.0131, "step": 42861 }, { "epoch": 13.19, "learning_rate": 5.499429640190119e-06, "loss": 0.0248, "step": 42862 }, { "epoch": 13.19, "learning_rate": 5.498984690717795e-06, "loss": 0.018, "step": 42863 }, { "epoch": 13.19, "learning_rate": 5.498539752420278e-06, "loss": 0.0305, "step": 42864 }, { "epoch": 13.19, "learning_rate": 5.498094825298672e-06, "loss": 0.0184, "step": 42865 }, { "epoch": 13.19, "learning_rate": 5.497649909354084e-06, "loss": 0.0216, "step": 42866 }, { "epoch": 13.19, "learning_rate": 5.497205004587618e-06, "loss": 0.0288, "step": 42867 }, { "epoch": 13.19, "learning_rate": 5.496760111000376e-06, "loss": 0.0238, "step": 42868 }, { "epoch": 13.19, "learning_rate": 5.496315228593468e-06, "loss": 0.0182, "step": 42869 }, { "epoch": 13.19, "learning_rate": 5.495870357367992e-06, "loss": 0.0184, "step": 42870 }, { "epoch": 13.19, "learning_rate": 5.495425497325058e-06, "loss": 0.0136, "step": 42871 }, { "epoch": 13.19, "learning_rate": 5.494980648465762e-06, "loss": 0.0329, "step": 42872 }, { "epoch": 13.19, "learning_rate": 5.494535810791224e-06, "loss": 0.0165, "step": 42873 }, { "epoch": 13.19, "learning_rate": 5.494090984302534e-06, "loss": 0.0355, "step": 42874 }, { "epoch": 13.19, "learning_rate": 5.493646169000802e-06, "loss": 0.0251, "step": 42875 }, { "epoch": 13.19, "learning_rate": 5.493201364887131e-06, "loss": 0.0145, "step": 42876 }, { "epoch": 13.19, "learning_rate": 5.492756571962622e-06, "loss": 0.0109, "step": 42877 }, { "epoch": 13.19, "learning_rate": 5.492311790228389e-06, "loss": 0.0225, "step": 42878 }, { "epoch": 13.19, "learning_rate": 5.491867019685528e-06, "loss": 0.0205, "step": 42879 }, { "epoch": 13.19, "learning_rate": 5.491422260335151e-06, "loss": 0.0232, "step": 42880 }, { "epoch": 13.19, "learning_rate": 5.490977512178352e-06, "loss": 0.0236, "step": 42881 }, { "epoch": 13.19, "learning_rate": 5.490532775216239e-06, "loss": 0.0064, "step": 42882 }, { "epoch": 13.19, "learning_rate": 5.49008804944992e-06, "loss": 0.0296, "step": 42883 }, { "epoch": 13.19, "learning_rate": 5.4896433348804965e-06, "loss": 0.0254, "step": 42884 }, { "epoch": 13.2, "learning_rate": 5.4891986315090716e-06, "loss": 0.0171, "step": 42885 }, { "epoch": 13.2, "learning_rate": 5.488753939336752e-06, "loss": 0.0187, "step": 42886 }, { "epoch": 13.2, "learning_rate": 5.488309258364639e-06, "loss": 0.0216, "step": 42887 }, { "epoch": 13.2, "learning_rate": 5.487864588593839e-06, "loss": 0.0127, "step": 42888 }, { "epoch": 13.2, "learning_rate": 5.487419930025454e-06, "loss": 0.035, "step": 42889 }, { "epoch": 13.2, "learning_rate": 5.486975282660589e-06, "loss": 0.0157, "step": 42890 }, { "epoch": 13.2, "learning_rate": 5.486530646500348e-06, "loss": 0.0308, "step": 42891 }, { "epoch": 13.2, "learning_rate": 5.486086021545829e-06, "loss": 0.0218, "step": 42892 }, { "epoch": 13.2, "learning_rate": 5.485641407798151e-06, "loss": 0.0228, "step": 42893 }, { "epoch": 13.2, "learning_rate": 5.485196805258404e-06, "loss": 0.0137, "step": 42894 }, { "epoch": 13.2, "learning_rate": 5.4847522139276965e-06, "loss": 0.0283, "step": 42895 }, { "epoch": 13.2, "learning_rate": 5.484307633807131e-06, "loss": 0.0333, "step": 42896 }, { "epoch": 13.2, "learning_rate": 5.48386306489781e-06, "loss": 0.0273, "step": 42897 }, { "epoch": 13.2, "learning_rate": 5.483418507200842e-06, "loss": 0.0419, "step": 42898 }, { "epoch": 13.2, "learning_rate": 5.482973960717328e-06, "loss": 0.0241, "step": 42899 }, { "epoch": 13.2, "learning_rate": 5.482529425448373e-06, "loss": 0.0256, "step": 42900 }, { "epoch": 13.2, "learning_rate": 5.482084901395084e-06, "loss": 0.0196, "step": 42901 }, { "epoch": 13.2, "learning_rate": 5.481640388558551e-06, "loss": 0.0178, "step": 42902 }, { "epoch": 13.2, "learning_rate": 5.4811958869398905e-06, "loss": 0.0248, "step": 42903 }, { "epoch": 13.2, "learning_rate": 5.4807513965402024e-06, "loss": 0.0344, "step": 42904 }, { "epoch": 13.2, "learning_rate": 5.4803069173605915e-06, "loss": 0.0168, "step": 42905 }, { "epoch": 13.2, "learning_rate": 5.4798624494021605e-06, "loss": 0.028, "step": 42906 }, { "epoch": 13.2, "learning_rate": 5.479417992666011e-06, "loss": 0.0183, "step": 42907 }, { "epoch": 13.2, "learning_rate": 5.478973547153248e-06, "loss": 0.0237, "step": 42908 }, { "epoch": 13.2, "learning_rate": 5.478529112864974e-06, "loss": 0.0106, "step": 42909 }, { "epoch": 13.2, "learning_rate": 5.478084689802295e-06, "loss": 0.0151, "step": 42910 }, { "epoch": 13.2, "learning_rate": 5.4776402779663115e-06, "loss": 0.0146, "step": 42911 }, { "epoch": 13.2, "learning_rate": 5.4771958773581256e-06, "loss": 0.0183, "step": 42912 }, { "epoch": 13.2, "learning_rate": 5.476751487978846e-06, "loss": 0.015, "step": 42913 }, { "epoch": 13.2, "learning_rate": 5.476307109829577e-06, "loss": 0.0418, "step": 42914 }, { "epoch": 13.2, "learning_rate": 5.475862742911413e-06, "loss": 0.0244, "step": 42915 }, { "epoch": 13.2, "learning_rate": 5.475418387225464e-06, "loss": 0.0199, "step": 42916 }, { "epoch": 13.21, "learning_rate": 5.474974042772826e-06, "loss": 0.0139, "step": 42917 }, { "epoch": 13.21, "learning_rate": 5.4745297095546125e-06, "loss": 0.0089, "step": 42918 }, { "epoch": 13.21, "learning_rate": 5.47408538757192e-06, "loss": 0.0245, "step": 42919 }, { "epoch": 13.21, "learning_rate": 5.473641076825855e-06, "loss": 0.0105, "step": 42920 }, { "epoch": 13.21, "learning_rate": 5.473196777317518e-06, "loss": 0.0427, "step": 42921 }, { "epoch": 13.21, "learning_rate": 5.4727524890480135e-06, "loss": 0.037, "step": 42922 }, { "epoch": 13.21, "learning_rate": 5.472308212018445e-06, "loss": 0.0245, "step": 42923 }, { "epoch": 13.21, "learning_rate": 5.471863946229911e-06, "loss": 0.0143, "step": 42924 }, { "epoch": 13.21, "learning_rate": 5.471419691683522e-06, "loss": 0.0274, "step": 42925 }, { "epoch": 13.21, "learning_rate": 5.470975448380376e-06, "loss": 0.0137, "step": 42926 }, { "epoch": 13.21, "learning_rate": 5.470531216321573e-06, "loss": 0.02, "step": 42927 }, { "epoch": 13.21, "learning_rate": 5.470086995508226e-06, "loss": 0.0126, "step": 42928 }, { "epoch": 13.21, "learning_rate": 5.4696427859414295e-06, "loss": 0.014, "step": 42929 }, { "epoch": 13.21, "learning_rate": 5.469198587622289e-06, "loss": 0.0286, "step": 42930 }, { "epoch": 13.21, "learning_rate": 5.468754400551905e-06, "loss": 0.0177, "step": 42931 }, { "epoch": 13.21, "learning_rate": 5.468310224731381e-06, "loss": 0.0207, "step": 42932 }, { "epoch": 13.21, "learning_rate": 5.467866060161824e-06, "loss": 0.0444, "step": 42933 }, { "epoch": 13.21, "learning_rate": 5.467421906844333e-06, "loss": 0.0221, "step": 42934 }, { "epoch": 13.21, "learning_rate": 5.466977764780015e-06, "loss": 0.0173, "step": 42935 }, { "epoch": 13.21, "learning_rate": 5.466533633969966e-06, "loss": 0.0206, "step": 42936 }, { "epoch": 13.21, "learning_rate": 5.466089514415288e-06, "loss": 0.026, "step": 42937 }, { "epoch": 13.21, "learning_rate": 5.465645406117091e-06, "loss": 0.0221, "step": 42938 }, { "epoch": 13.21, "learning_rate": 5.4652013090764746e-06, "loss": 0.0416, "step": 42939 }, { "epoch": 13.21, "learning_rate": 5.46475722329454e-06, "loss": 0.0181, "step": 42940 }, { "epoch": 13.21, "learning_rate": 5.464313148772391e-06, "loss": 0.0114, "step": 42941 }, { "epoch": 13.21, "learning_rate": 5.463869085511128e-06, "loss": 0.0115, "step": 42942 }, { "epoch": 13.21, "learning_rate": 5.4634250335118575e-06, "loss": 0.0266, "step": 42943 }, { "epoch": 13.21, "learning_rate": 5.4629809927756794e-06, "loss": 0.013, "step": 42944 }, { "epoch": 13.21, "learning_rate": 5.462536963303696e-06, "loss": 0.0329, "step": 42945 }, { "epoch": 13.21, "learning_rate": 5.462092945097011e-06, "loss": 0.0106, "step": 42946 }, { "epoch": 13.21, "learning_rate": 5.4616489381567195e-06, "loss": 0.0141, "step": 42947 }, { "epoch": 13.21, "learning_rate": 5.461204942483935e-06, "loss": 0.019, "step": 42948 }, { "epoch": 13.21, "learning_rate": 5.46076095807976e-06, "loss": 0.0186, "step": 42949 }, { "epoch": 13.22, "learning_rate": 5.460316984945286e-06, "loss": 0.0309, "step": 42950 }, { "epoch": 13.22, "learning_rate": 5.459873023081624e-06, "loss": 0.0131, "step": 42951 }, { "epoch": 13.22, "learning_rate": 5.4594290724898665e-06, "loss": 0.0117, "step": 42952 }, { "epoch": 13.22, "learning_rate": 5.4589851331711285e-06, "loss": 0.0366, "step": 42953 }, { "epoch": 13.22, "learning_rate": 5.458541205126504e-06, "loss": 0.0258, "step": 42954 }, { "epoch": 13.22, "learning_rate": 5.4580972883570996e-06, "loss": 0.0134, "step": 42955 }, { "epoch": 13.22, "learning_rate": 5.4576533828640184e-06, "loss": 0.0297, "step": 42956 }, { "epoch": 13.22, "learning_rate": 5.457209488648351e-06, "loss": 0.0189, "step": 42957 }, { "epoch": 13.22, "learning_rate": 5.456765605711212e-06, "loss": 0.0189, "step": 42958 }, { "epoch": 13.22, "learning_rate": 5.456321734053698e-06, "loss": 0.0354, "step": 42959 }, { "epoch": 13.22, "learning_rate": 5.455877873676913e-06, "loss": 0.0401, "step": 42960 }, { "epoch": 13.22, "learning_rate": 5.455434024581959e-06, "loss": 0.0159, "step": 42961 }, { "epoch": 13.22, "learning_rate": 5.454990186769937e-06, "loss": 0.0166, "step": 42962 }, { "epoch": 13.22, "learning_rate": 5.454546360241948e-06, "loss": 0.0198, "step": 42963 }, { "epoch": 13.22, "learning_rate": 5.454102544999096e-06, "loss": 0.0264, "step": 42964 }, { "epoch": 13.22, "learning_rate": 5.453658741042481e-06, "loss": 0.0131, "step": 42965 }, { "epoch": 13.22, "learning_rate": 5.453214948373206e-06, "loss": 0.0179, "step": 42966 }, { "epoch": 13.22, "learning_rate": 5.452771166992369e-06, "loss": 0.0153, "step": 42967 }, { "epoch": 13.22, "learning_rate": 5.452327396901078e-06, "loss": 0.0132, "step": 42968 }, { "epoch": 13.22, "learning_rate": 5.451883638100434e-06, "loss": 0.0207, "step": 42969 }, { "epoch": 13.22, "learning_rate": 5.451439890591539e-06, "loss": 0.0179, "step": 42970 }, { "epoch": 13.22, "learning_rate": 5.45099615437549e-06, "loss": 0.0217, "step": 42971 }, { "epoch": 13.22, "learning_rate": 5.450552429453386e-06, "loss": 0.0242, "step": 42972 }, { "epoch": 13.22, "learning_rate": 5.450108715826339e-06, "loss": 0.0393, "step": 42973 }, { "epoch": 13.22, "learning_rate": 5.4496650134954445e-06, "loss": 0.0101, "step": 42974 }, { "epoch": 13.22, "learning_rate": 5.449221322461805e-06, "loss": 0.0324, "step": 42975 }, { "epoch": 13.22, "learning_rate": 5.448777642726522e-06, "loss": 0.0245, "step": 42976 }, { "epoch": 13.22, "learning_rate": 5.448333974290698e-06, "loss": 0.0287, "step": 42977 }, { "epoch": 13.22, "learning_rate": 5.447890317155433e-06, "loss": 0.0138, "step": 42978 }, { "epoch": 13.22, "learning_rate": 5.44744667132183e-06, "loss": 0.0138, "step": 42979 }, { "epoch": 13.22, "learning_rate": 5.4470030367909874e-06, "loss": 0.022, "step": 42980 }, { "epoch": 13.22, "learning_rate": 5.446559413564011e-06, "loss": 0.025, "step": 42981 }, { "epoch": 13.23, "learning_rate": 5.446115801641995e-06, "loss": 0.0155, "step": 42982 }, { "epoch": 13.23, "learning_rate": 5.445672201026054e-06, "loss": 0.0277, "step": 42983 }, { "epoch": 13.23, "learning_rate": 5.445228611717277e-06, "loss": 0.0185, "step": 42984 }, { "epoch": 13.23, "learning_rate": 5.4447850337167705e-06, "loss": 0.0334, "step": 42985 }, { "epoch": 13.23, "learning_rate": 5.444341467025633e-06, "loss": 0.0235, "step": 42986 }, { "epoch": 13.23, "learning_rate": 5.443897911644964e-06, "loss": 0.031, "step": 42987 }, { "epoch": 13.23, "learning_rate": 5.443454367575873e-06, "loss": 0.0194, "step": 42988 }, { "epoch": 13.23, "learning_rate": 5.443010834819456e-06, "loss": 0.0144, "step": 42989 }, { "epoch": 13.23, "learning_rate": 5.442567313376814e-06, "loss": 0.0187, "step": 42990 }, { "epoch": 13.23, "learning_rate": 5.442123803249053e-06, "loss": 0.018, "step": 42991 }, { "epoch": 13.23, "learning_rate": 5.441680304437263e-06, "loss": 0.032, "step": 42992 }, { "epoch": 13.23, "learning_rate": 5.441236816942555e-06, "loss": 0.0135, "step": 42993 }, { "epoch": 13.23, "learning_rate": 5.440793340766027e-06, "loss": 0.0228, "step": 42994 }, { "epoch": 13.23, "learning_rate": 5.440349875908779e-06, "loss": 0.0225, "step": 42995 }, { "epoch": 13.23, "learning_rate": 5.439906422371914e-06, "loss": 0.0237, "step": 42996 }, { "epoch": 13.23, "learning_rate": 5.439462980156531e-06, "loss": 0.0251, "step": 42997 }, { "epoch": 13.23, "learning_rate": 5.4390195492637335e-06, "loss": 0.0249, "step": 42998 }, { "epoch": 13.23, "learning_rate": 5.43857612969462e-06, "loss": 0.0203, "step": 42999 }, { "epoch": 13.23, "learning_rate": 5.438132721450293e-06, "loss": 0.0119, "step": 43000 }, { "epoch": 13.23, "learning_rate": 5.4376893245318514e-06, "loss": 0.021, "step": 43001 }, { "epoch": 13.23, "learning_rate": 5.437245938940394e-06, "loss": 0.0398, "step": 43002 }, { "epoch": 13.23, "learning_rate": 5.436802564677028e-06, "loss": 0.0609, "step": 43003 }, { "epoch": 13.23, "learning_rate": 5.436359201742857e-06, "loss": 0.0115, "step": 43004 }, { "epoch": 13.23, "learning_rate": 5.435915850138971e-06, "loss": 0.0193, "step": 43005 }, { "epoch": 13.23, "learning_rate": 5.435472509866475e-06, "loss": 0.0223, "step": 43006 }, { "epoch": 13.23, "learning_rate": 5.435029180926465e-06, "loss": 0.0157, "step": 43007 }, { "epoch": 13.23, "learning_rate": 5.434585863320052e-06, "loss": 0.0146, "step": 43008 }, { "epoch": 13.23, "learning_rate": 5.434142557048332e-06, "loss": 0.028, "step": 43009 }, { "epoch": 13.23, "learning_rate": 5.433699262112405e-06, "loss": 0.025, "step": 43010 }, { "epoch": 13.23, "learning_rate": 5.433255978513372e-06, "loss": 0.0172, "step": 43011 }, { "epoch": 13.23, "learning_rate": 5.432812706252333e-06, "loss": 0.0241, "step": 43012 }, { "epoch": 13.23, "learning_rate": 5.432369445330388e-06, "loss": 0.022, "step": 43013 }, { "epoch": 13.23, "learning_rate": 5.431926195748638e-06, "loss": 0.022, "step": 43014 }, { "epoch": 13.24, "learning_rate": 5.431482957508185e-06, "loss": 0.0324, "step": 43015 }, { "epoch": 13.24, "learning_rate": 5.431039730610128e-06, "loss": 0.0124, "step": 43016 }, { "epoch": 13.24, "learning_rate": 5.4305965150555665e-06, "loss": 0.0208, "step": 43017 }, { "epoch": 13.24, "learning_rate": 5.430153310845603e-06, "loss": 0.0126, "step": 43018 }, { "epoch": 13.24, "learning_rate": 5.429710117981337e-06, "loss": 0.0158, "step": 43019 }, { "epoch": 13.24, "learning_rate": 5.4292669364638685e-06, "loss": 0.0201, "step": 43020 }, { "epoch": 13.24, "learning_rate": 5.428823766294298e-06, "loss": 0.0198, "step": 43021 }, { "epoch": 13.24, "learning_rate": 5.42838060747372e-06, "loss": 0.0212, "step": 43022 }, { "epoch": 13.24, "learning_rate": 5.427937460003247e-06, "loss": 0.018, "step": 43023 }, { "epoch": 13.24, "learning_rate": 5.42749432388397e-06, "loss": 0.0158, "step": 43024 }, { "epoch": 13.24, "learning_rate": 5.427051199117e-06, "loss": 0.0269, "step": 43025 }, { "epoch": 13.24, "learning_rate": 5.426608085703422e-06, "loss": 0.0141, "step": 43026 }, { "epoch": 13.24, "learning_rate": 5.42616498364434e-06, "loss": 0.0178, "step": 43027 }, { "epoch": 13.24, "learning_rate": 5.42572189294086e-06, "loss": 0.0262, "step": 43028 }, { "epoch": 13.24, "learning_rate": 5.425278813594081e-06, "loss": 0.0271, "step": 43029 }, { "epoch": 13.24, "learning_rate": 5.4248357456051e-06, "loss": 0.0261, "step": 43030 }, { "epoch": 13.24, "learning_rate": 5.4243926889750195e-06, "loss": 0.0302, "step": 43031 }, { "epoch": 13.24, "learning_rate": 5.423949643704938e-06, "loss": 0.064, "step": 43032 }, { "epoch": 13.24, "learning_rate": 5.423506609795956e-06, "loss": 0.0178, "step": 43033 }, { "epoch": 13.24, "learning_rate": 5.423063587249173e-06, "loss": 0.0285, "step": 43034 }, { "epoch": 13.24, "learning_rate": 5.422620576065689e-06, "loss": 0.0145, "step": 43035 }, { "epoch": 13.24, "learning_rate": 5.4221775762466055e-06, "loss": 0.0182, "step": 43036 }, { "epoch": 13.24, "learning_rate": 5.421734587793015e-06, "loss": 0.0231, "step": 43037 }, { "epoch": 13.24, "learning_rate": 5.421291610706032e-06, "loss": 0.0133, "step": 43038 }, { "epoch": 13.24, "learning_rate": 5.420848644986745e-06, "loss": 0.0253, "step": 43039 }, { "epoch": 13.24, "learning_rate": 5.420405690636255e-06, "loss": 0.025, "step": 43040 }, { "epoch": 13.24, "learning_rate": 5.419962747655663e-06, "loss": 0.0168, "step": 43041 }, { "epoch": 13.24, "learning_rate": 5.419519816046065e-06, "loss": 0.0093, "step": 43042 }, { "epoch": 13.24, "learning_rate": 5.419076895808568e-06, "loss": 0.0292, "step": 43043 }, { "epoch": 13.24, "learning_rate": 5.4186339869442684e-06, "loss": 0.0117, "step": 43044 }, { "epoch": 13.24, "learning_rate": 5.418191089454266e-06, "loss": 0.0195, "step": 43045 }, { "epoch": 13.24, "learning_rate": 5.417748203339663e-06, "loss": 0.0211, "step": 43046 }, { "epoch": 13.25, "learning_rate": 5.417305328601549e-06, "loss": 0.0121, "step": 43047 }, { "epoch": 13.25, "learning_rate": 5.416862465241033e-06, "loss": 0.0137, "step": 43048 }, { "epoch": 13.25, "learning_rate": 5.416419613259213e-06, "loss": 0.0174, "step": 43049 }, { "epoch": 13.25, "learning_rate": 5.415976772657188e-06, "loss": 0.0116, "step": 43050 }, { "epoch": 13.25, "learning_rate": 5.415533943436055e-06, "loss": 0.0186, "step": 43051 }, { "epoch": 13.25, "learning_rate": 5.415091125596916e-06, "loss": 0.0187, "step": 43052 }, { "epoch": 13.25, "learning_rate": 5.4146483191408695e-06, "loss": 0.0152, "step": 43053 }, { "epoch": 13.25, "learning_rate": 5.414205524069016e-06, "loss": 0.0276, "step": 43054 }, { "epoch": 13.25, "learning_rate": 5.413762740382454e-06, "loss": 0.0265, "step": 43055 }, { "epoch": 13.25, "learning_rate": 5.4133199680822815e-06, "loss": 0.0343, "step": 43056 }, { "epoch": 13.25, "learning_rate": 5.412877207169596e-06, "loss": 0.0203, "step": 43057 }, { "epoch": 13.25, "learning_rate": 5.412434457645503e-06, "loss": 0.0123, "step": 43058 }, { "epoch": 13.25, "learning_rate": 5.411991719511102e-06, "loss": 0.0224, "step": 43059 }, { "epoch": 13.25, "learning_rate": 5.411548992767486e-06, "loss": 0.0096, "step": 43060 }, { "epoch": 13.25, "learning_rate": 5.411106277415756e-06, "loss": 0.0247, "step": 43061 }, { "epoch": 13.25, "learning_rate": 5.410663573457009e-06, "loss": 0.0194, "step": 43062 }, { "epoch": 13.25, "learning_rate": 5.41022088089235e-06, "loss": 0.0158, "step": 43063 }, { "epoch": 13.25, "learning_rate": 5.4097781997228745e-06, "loss": 0.0169, "step": 43064 }, { "epoch": 13.25, "learning_rate": 5.4093355299496845e-06, "loss": 0.0275, "step": 43065 }, { "epoch": 13.25, "learning_rate": 5.408892871573874e-06, "loss": 0.0126, "step": 43066 }, { "epoch": 13.25, "learning_rate": 5.4084502245965465e-06, "loss": 0.0112, "step": 43067 }, { "epoch": 13.25, "learning_rate": 5.4080075890188e-06, "loss": 0.0222, "step": 43068 }, { "epoch": 13.25, "learning_rate": 5.4075649648417314e-06, "loss": 0.0204, "step": 43069 }, { "epoch": 13.25, "learning_rate": 5.407122352066442e-06, "loss": 0.0105, "step": 43070 }, { "epoch": 13.25, "learning_rate": 5.406679750694028e-06, "loss": 0.0133, "step": 43071 }, { "epoch": 13.25, "learning_rate": 5.406237160725591e-06, "loss": 0.0227, "step": 43072 }, { "epoch": 13.25, "learning_rate": 5.40579458216223e-06, "loss": 0.0166, "step": 43073 }, { "epoch": 13.25, "learning_rate": 5.405352015005039e-06, "loss": 0.0152, "step": 43074 }, { "epoch": 13.25, "learning_rate": 5.404909459255123e-06, "loss": 0.0301, "step": 43075 }, { "epoch": 13.25, "learning_rate": 5.404466914913577e-06, "loss": 0.029, "step": 43076 }, { "epoch": 13.25, "learning_rate": 5.4040243819814966e-06, "loss": 0.0107, "step": 43077 }, { "epoch": 13.25, "learning_rate": 5.40358186045999e-06, "loss": 0.034, "step": 43078 }, { "epoch": 13.25, "learning_rate": 5.403139350350148e-06, "loss": 0.0132, "step": 43079 }, { "epoch": 13.26, "learning_rate": 5.402696851653077e-06, "loss": 0.0242, "step": 43080 }, { "epoch": 13.26, "learning_rate": 5.402254364369868e-06, "loss": 0.0242, "step": 43081 }, { "epoch": 13.26, "learning_rate": 5.401811888501616e-06, "loss": 0.0196, "step": 43082 }, { "epoch": 13.26, "learning_rate": 5.40136942404943e-06, "loss": 0.0162, "step": 43083 }, { "epoch": 13.26, "learning_rate": 5.400926971014403e-06, "loss": 0.0279, "step": 43084 }, { "epoch": 13.26, "learning_rate": 5.400484529397636e-06, "loss": 0.0174, "step": 43085 }, { "epoch": 13.26, "learning_rate": 5.4000420992002245e-06, "loss": 0.0227, "step": 43086 }, { "epoch": 13.26, "learning_rate": 5.399599680423269e-06, "loss": 0.0312, "step": 43087 }, { "epoch": 13.26, "learning_rate": 5.399157273067868e-06, "loss": 0.012, "step": 43088 }, { "epoch": 13.26, "learning_rate": 5.398714877135118e-06, "loss": 0.0195, "step": 43089 }, { "epoch": 13.26, "learning_rate": 5.3982724926261185e-06, "loss": 0.0137, "step": 43090 }, { "epoch": 13.26, "learning_rate": 5.397830119541969e-06, "loss": 0.0359, "step": 43091 }, { "epoch": 13.26, "learning_rate": 5.397387757883761e-06, "loss": 0.0109, "step": 43092 }, { "epoch": 13.26, "learning_rate": 5.3969454076526076e-06, "loss": 0.0312, "step": 43093 }, { "epoch": 13.26, "learning_rate": 5.396503068849593e-06, "loss": 0.0273, "step": 43094 }, { "epoch": 13.26, "learning_rate": 5.396060741475821e-06, "loss": 0.0116, "step": 43095 }, { "epoch": 13.26, "learning_rate": 5.39561842553239e-06, "loss": 0.009, "step": 43096 }, { "epoch": 13.26, "learning_rate": 5.395176121020391e-06, "loss": 0.0175, "step": 43097 }, { "epoch": 13.26, "learning_rate": 5.394733827940933e-06, "loss": 0.0265, "step": 43098 }, { "epoch": 13.26, "learning_rate": 5.3942915462951095e-06, "loss": 0.0393, "step": 43099 }, { "epoch": 13.26, "learning_rate": 5.3938492760840176e-06, "loss": 0.0208, "step": 43100 }, { "epoch": 13.26, "learning_rate": 5.393407017308761e-06, "loss": 0.0214, "step": 43101 }, { "epoch": 13.26, "learning_rate": 5.392964769970426e-06, "loss": 0.024, "step": 43102 }, { "epoch": 13.26, "learning_rate": 5.392522534070119e-06, "loss": 0.018, "step": 43103 }, { "epoch": 13.26, "learning_rate": 5.392080309608938e-06, "loss": 0.0128, "step": 43104 }, { "epoch": 13.26, "learning_rate": 5.391638096587979e-06, "loss": 0.031, "step": 43105 }, { "epoch": 13.26, "learning_rate": 5.391195895008341e-06, "loss": 0.013, "step": 43106 }, { "epoch": 13.26, "learning_rate": 5.390753704871119e-06, "loss": 0.0288, "step": 43107 }, { "epoch": 13.26, "learning_rate": 5.390311526177414e-06, "loss": 0.0236, "step": 43108 }, { "epoch": 13.26, "learning_rate": 5.389869358928323e-06, "loss": 0.0178, "step": 43109 }, { "epoch": 13.26, "learning_rate": 5.389427203124944e-06, "loss": 0.0207, "step": 43110 }, { "epoch": 13.26, "learning_rate": 5.388985058768374e-06, "loss": 0.015, "step": 43111 }, { "epoch": 13.27, "learning_rate": 5.388542925859708e-06, "loss": 0.0112, "step": 43112 }, { "epoch": 13.27, "learning_rate": 5.3881008044000495e-06, "loss": 0.0168, "step": 43113 }, { "epoch": 13.27, "learning_rate": 5.387658694390496e-06, "loss": 0.0156, "step": 43114 }, { "epoch": 13.27, "learning_rate": 5.387216595832141e-06, "loss": 0.0162, "step": 43115 }, { "epoch": 13.27, "learning_rate": 5.386774508726083e-06, "loss": 0.0202, "step": 43116 }, { "epoch": 13.27, "learning_rate": 5.386332433073415e-06, "loss": 0.0297, "step": 43117 }, { "epoch": 13.27, "learning_rate": 5.3858903688752436e-06, "loss": 0.0072, "step": 43118 }, { "epoch": 13.27, "learning_rate": 5.3854483161326645e-06, "loss": 0.0143, "step": 43119 }, { "epoch": 13.27, "learning_rate": 5.385006274846771e-06, "loss": 0.0108, "step": 43120 }, { "epoch": 13.27, "learning_rate": 5.384564245018664e-06, "loss": 0.0229, "step": 43121 }, { "epoch": 13.27, "learning_rate": 5.384122226649439e-06, "loss": 0.0187, "step": 43122 }, { "epoch": 13.27, "learning_rate": 5.383680219740195e-06, "loss": 0.0107, "step": 43123 }, { "epoch": 13.27, "learning_rate": 5.383238224292029e-06, "loss": 0.0186, "step": 43124 }, { "epoch": 13.27, "learning_rate": 5.382796240306037e-06, "loss": 0.0299, "step": 43125 }, { "epoch": 13.27, "learning_rate": 5.382354267783316e-06, "loss": 0.0234, "step": 43126 }, { "epoch": 13.27, "learning_rate": 5.381912306724962e-06, "loss": 0.0143, "step": 43127 }, { "epoch": 13.27, "learning_rate": 5.381470357132084e-06, "loss": 0.0212, "step": 43128 }, { "epoch": 13.27, "learning_rate": 5.381028419005763e-06, "loss": 0.0244, "step": 43129 }, { "epoch": 13.27, "learning_rate": 5.380586492347104e-06, "loss": 0.0278, "step": 43130 }, { "epoch": 13.27, "learning_rate": 5.380144577157205e-06, "loss": 0.0166, "step": 43131 }, { "epoch": 13.27, "learning_rate": 5.379702673437157e-06, "loss": 0.0179, "step": 43132 }, { "epoch": 13.27, "learning_rate": 5.3792607811880644e-06, "loss": 0.0143, "step": 43133 }, { "epoch": 13.27, "learning_rate": 5.378818900411022e-06, "loss": 0.0115, "step": 43134 }, { "epoch": 13.27, "learning_rate": 5.378377031107131e-06, "loss": 0.0227, "step": 43135 }, { "epoch": 13.27, "learning_rate": 5.37793517327748e-06, "loss": 0.0079, "step": 43136 }, { "epoch": 13.27, "learning_rate": 5.3774933269231654e-06, "loss": 0.0212, "step": 43137 }, { "epoch": 13.27, "learning_rate": 5.377051492045293e-06, "loss": 0.0141, "step": 43138 }, { "epoch": 13.27, "learning_rate": 5.376609668644954e-06, "loss": 0.0253, "step": 43139 }, { "epoch": 13.27, "learning_rate": 5.376167856723249e-06, "loss": 0.0258, "step": 43140 }, { "epoch": 13.27, "learning_rate": 5.375726056281272e-06, "loss": 0.0147, "step": 43141 }, { "epoch": 13.27, "learning_rate": 5.37528426732012e-06, "loss": 0.0204, "step": 43142 }, { "epoch": 13.27, "learning_rate": 5.37484248984089e-06, "loss": 0.028, "step": 43143 }, { "epoch": 13.27, "learning_rate": 5.374400723844681e-06, "loss": 0.0103, "step": 43144 }, { "epoch": 13.28, "learning_rate": 5.373958969332587e-06, "loss": 0.0269, "step": 43145 }, { "epoch": 13.28, "learning_rate": 5.373517226305704e-06, "loss": 0.0122, "step": 43146 }, { "epoch": 13.28, "learning_rate": 5.373075494765129e-06, "loss": 0.0143, "step": 43147 }, { "epoch": 13.28, "learning_rate": 5.3726337747119625e-06, "loss": 0.0201, "step": 43148 }, { "epoch": 13.28, "learning_rate": 5.372192066147304e-06, "loss": 0.0166, "step": 43149 }, { "epoch": 13.28, "learning_rate": 5.371750369072241e-06, "loss": 0.0319, "step": 43150 }, { "epoch": 13.28, "learning_rate": 5.371308683487869e-06, "loss": 0.0267, "step": 43151 }, { "epoch": 13.28, "learning_rate": 5.370867009395294e-06, "loss": 0.0135, "step": 43152 }, { "epoch": 13.28, "learning_rate": 5.370425346795609e-06, "loss": 0.02, "step": 43153 }, { "epoch": 13.28, "learning_rate": 5.369983695689908e-06, "loss": 0.0246, "step": 43154 }, { "epoch": 13.28, "learning_rate": 5.369542056079291e-06, "loss": 0.0144, "step": 43155 }, { "epoch": 13.28, "learning_rate": 5.369100427964853e-06, "loss": 0.0299, "step": 43156 }, { "epoch": 13.28, "learning_rate": 5.368658811347689e-06, "loss": 0.0251, "step": 43157 }, { "epoch": 13.28, "learning_rate": 5.368217206228898e-06, "loss": 0.0174, "step": 43158 }, { "epoch": 13.28, "learning_rate": 5.367775612609574e-06, "loss": 0.0292, "step": 43159 }, { "epoch": 13.28, "learning_rate": 5.367334030490814e-06, "loss": 0.0216, "step": 43160 }, { "epoch": 13.28, "learning_rate": 5.366892459873711e-06, "loss": 0.0254, "step": 43161 }, { "epoch": 13.28, "learning_rate": 5.366450900759372e-06, "loss": 0.0216, "step": 43162 }, { "epoch": 13.28, "learning_rate": 5.366009353148884e-06, "loss": 0.0252, "step": 43163 }, { "epoch": 13.28, "learning_rate": 5.365567817043345e-06, "loss": 0.0199, "step": 43164 }, { "epoch": 13.28, "learning_rate": 5.365126292443852e-06, "loss": 0.016, "step": 43165 }, { "epoch": 13.28, "learning_rate": 5.364684779351495e-06, "loss": 0.0221, "step": 43166 }, { "epoch": 13.28, "learning_rate": 5.364243277767383e-06, "loss": 0.0129, "step": 43167 }, { "epoch": 13.28, "learning_rate": 5.363801787692603e-06, "loss": 0.0191, "step": 43168 }, { "epoch": 13.28, "learning_rate": 5.363360309128254e-06, "loss": 0.0247, "step": 43169 }, { "epoch": 13.28, "learning_rate": 5.362918842075435e-06, "loss": 0.0204, "step": 43170 }, { "epoch": 13.28, "learning_rate": 5.36247738653523e-06, "loss": 0.0139, "step": 43171 }, { "epoch": 13.28, "learning_rate": 5.362035942508749e-06, "loss": 0.0147, "step": 43172 }, { "epoch": 13.28, "learning_rate": 5.361594509997081e-06, "loss": 0.0217, "step": 43173 }, { "epoch": 13.28, "learning_rate": 5.361153089001323e-06, "loss": 0.0139, "step": 43174 }, { "epoch": 13.28, "learning_rate": 5.360711679522573e-06, "loss": 0.0274, "step": 43175 }, { "epoch": 13.28, "learning_rate": 5.360270281561924e-06, "loss": 0.0431, "step": 43176 }, { "epoch": 13.29, "learning_rate": 5.359828895120474e-06, "loss": 0.0127, "step": 43177 }, { "epoch": 13.29, "learning_rate": 5.359387520199317e-06, "loss": 0.023, "step": 43178 }, { "epoch": 13.29, "learning_rate": 5.35894615679955e-06, "loss": 0.0311, "step": 43179 }, { "epoch": 13.29, "learning_rate": 5.358504804922269e-06, "loss": 0.0137, "step": 43180 }, { "epoch": 13.29, "learning_rate": 5.358063464568563e-06, "loss": 0.0209, "step": 43181 }, { "epoch": 13.29, "learning_rate": 5.35762213573954e-06, "loss": 0.0306, "step": 43182 }, { "epoch": 13.29, "learning_rate": 5.357180818436294e-06, "loss": 0.0173, "step": 43183 }, { "epoch": 13.29, "learning_rate": 5.356739512659912e-06, "loss": 0.0353, "step": 43184 }, { "epoch": 13.29, "learning_rate": 5.356298218411493e-06, "loss": 0.0237, "step": 43185 }, { "epoch": 13.29, "learning_rate": 5.355856935692131e-06, "loss": 0.0147, "step": 43186 }, { "epoch": 13.29, "learning_rate": 5.355415664502928e-06, "loss": 0.0217, "step": 43187 }, { "epoch": 13.29, "learning_rate": 5.354974404844975e-06, "loss": 0.0155, "step": 43188 }, { "epoch": 13.29, "learning_rate": 5.354533156719368e-06, "loss": 0.018, "step": 43189 }, { "epoch": 13.29, "learning_rate": 5.3540919201272045e-06, "loss": 0.0177, "step": 43190 }, { "epoch": 13.29, "learning_rate": 5.353650695069577e-06, "loss": 0.0213, "step": 43191 }, { "epoch": 13.29, "learning_rate": 5.353209481547582e-06, "loss": 0.0134, "step": 43192 }, { "epoch": 13.29, "learning_rate": 5.352768279562315e-06, "loss": 0.0322, "step": 43193 }, { "epoch": 13.29, "learning_rate": 5.352327089114873e-06, "loss": 0.0207, "step": 43194 }, { "epoch": 13.29, "learning_rate": 5.351885910206348e-06, "loss": 0.0139, "step": 43195 }, { "epoch": 13.29, "learning_rate": 5.351444742837839e-06, "loss": 0.0069, "step": 43196 }, { "epoch": 13.29, "learning_rate": 5.3510035870104385e-06, "loss": 0.0276, "step": 43197 }, { "epoch": 13.29, "learning_rate": 5.350562442725242e-06, "loss": 0.0196, "step": 43198 }, { "epoch": 13.29, "learning_rate": 5.350121309983347e-06, "loss": 0.0312, "step": 43199 }, { "epoch": 13.29, "learning_rate": 5.349680188785846e-06, "loss": 0.0346, "step": 43200 }, { "epoch": 13.29, "learning_rate": 5.3492390791338324e-06, "loss": 0.0126, "step": 43201 }, { "epoch": 13.29, "learning_rate": 5.348797981028408e-06, "loss": 0.0196, "step": 43202 }, { "epoch": 13.29, "learning_rate": 5.348356894470664e-06, "loss": 0.0368, "step": 43203 }, { "epoch": 13.29, "learning_rate": 5.3479158194617e-06, "loss": 0.0234, "step": 43204 }, { "epoch": 13.29, "learning_rate": 5.347474756002603e-06, "loss": 0.0162, "step": 43205 }, { "epoch": 13.29, "learning_rate": 5.347033704094467e-06, "loss": 0.0255, "step": 43206 }, { "epoch": 13.29, "learning_rate": 5.346592663738397e-06, "loss": 0.0122, "step": 43207 }, { "epoch": 13.29, "learning_rate": 5.346151634935484e-06, "loss": 0.0137, "step": 43208 }, { "epoch": 13.29, "learning_rate": 5.3457106176868186e-06, "loss": 0.0144, "step": 43209 }, { "epoch": 13.3, "learning_rate": 5.345269611993502e-06, "loss": 0.0156, "step": 43210 }, { "epoch": 13.3, "learning_rate": 5.344828617856625e-06, "loss": 0.0171, "step": 43211 }, { "epoch": 13.3, "learning_rate": 5.344387635277283e-06, "loss": 0.0338, "step": 43212 }, { "epoch": 13.3, "learning_rate": 5.343946664256572e-06, "loss": 0.0257, "step": 43213 }, { "epoch": 13.3, "learning_rate": 5.343505704795586e-06, "loss": 0.0318, "step": 43214 }, { "epoch": 13.3, "learning_rate": 5.34306475689542e-06, "loss": 0.0255, "step": 43215 }, { "epoch": 13.3, "learning_rate": 5.342623820557165e-06, "loss": 0.0273, "step": 43216 }, { "epoch": 13.3, "learning_rate": 5.342182895781926e-06, "loss": 0.0206, "step": 43217 }, { "epoch": 13.3, "learning_rate": 5.341741982570789e-06, "loss": 0.0368, "step": 43218 }, { "epoch": 13.3, "learning_rate": 5.34130108092485e-06, "loss": 0.0125, "step": 43219 }, { "epoch": 13.3, "learning_rate": 5.340860190845205e-06, "loss": 0.0331, "step": 43220 }, { "epoch": 13.3, "learning_rate": 5.340419312332943e-06, "loss": 0.021, "step": 43221 }, { "epoch": 13.3, "learning_rate": 5.339978445389169e-06, "loss": 0.0185, "step": 43222 }, { "epoch": 13.3, "learning_rate": 5.339537590014972e-06, "loss": 0.0166, "step": 43223 }, { "epoch": 13.3, "learning_rate": 5.339096746211445e-06, "loss": 0.0247, "step": 43224 }, { "epoch": 13.3, "learning_rate": 5.338655913979691e-06, "loss": 0.0133, "step": 43225 }, { "epoch": 13.3, "learning_rate": 5.338215093320788e-06, "loss": 0.0126, "step": 43226 }, { "epoch": 13.3, "learning_rate": 5.337774284235844e-06, "loss": 0.018, "step": 43227 }, { "epoch": 13.3, "learning_rate": 5.337333486725951e-06, "loss": 0.0203, "step": 43228 }, { "epoch": 13.3, "learning_rate": 5.3368927007922e-06, "loss": 0.0075, "step": 43229 }, { "epoch": 13.3, "learning_rate": 5.336451926435688e-06, "loss": 0.0418, "step": 43230 }, { "epoch": 13.3, "learning_rate": 5.336011163657509e-06, "loss": 0.0086, "step": 43231 }, { "epoch": 13.3, "learning_rate": 5.335570412458757e-06, "loss": 0.02, "step": 43232 }, { "epoch": 13.3, "learning_rate": 5.335129672840527e-06, "loss": 0.0068, "step": 43233 }, { "epoch": 13.3, "learning_rate": 5.334688944803909e-06, "loss": 0.0212, "step": 43234 }, { "epoch": 13.3, "learning_rate": 5.334248228350004e-06, "loss": 0.0176, "step": 43235 }, { "epoch": 13.3, "learning_rate": 5.333807523479898e-06, "loss": 0.0384, "step": 43236 }, { "epoch": 13.3, "learning_rate": 5.333366830194694e-06, "loss": 0.0271, "step": 43237 }, { "epoch": 13.3, "learning_rate": 5.332926148495485e-06, "loss": 0.026, "step": 43238 }, { "epoch": 13.3, "learning_rate": 5.3324854783833606e-06, "loss": 0.0191, "step": 43239 }, { "epoch": 13.3, "learning_rate": 5.332044819859414e-06, "loss": 0.0421, "step": 43240 }, { "epoch": 13.3, "learning_rate": 5.331604172924738e-06, "loss": 0.0296, "step": 43241 }, { "epoch": 13.31, "learning_rate": 5.331163537580435e-06, "loss": 0.0265, "step": 43242 }, { "epoch": 13.31, "learning_rate": 5.330722913827594e-06, "loss": 0.0207, "step": 43243 }, { "epoch": 13.31, "learning_rate": 5.330282301667309e-06, "loss": 0.0253, "step": 43244 }, { "epoch": 13.31, "learning_rate": 5.329841701100674e-06, "loss": 0.0201, "step": 43245 }, { "epoch": 13.31, "learning_rate": 5.329401112128784e-06, "loss": 0.0262, "step": 43246 }, { "epoch": 13.31, "learning_rate": 5.328960534752731e-06, "loss": 0.0096, "step": 43247 }, { "epoch": 13.31, "learning_rate": 5.32851996897361e-06, "loss": 0.0183, "step": 43248 }, { "epoch": 13.31, "learning_rate": 5.328079414792515e-06, "loss": 0.0124, "step": 43249 }, { "epoch": 13.31, "learning_rate": 5.3276388722105385e-06, "loss": 0.0365, "step": 43250 }, { "epoch": 13.31, "learning_rate": 5.3271983412287745e-06, "loss": 0.0247, "step": 43251 }, { "epoch": 13.31, "learning_rate": 5.3267578218483185e-06, "loss": 0.0138, "step": 43252 }, { "epoch": 13.31, "learning_rate": 5.326317314070263e-06, "loss": 0.0172, "step": 43253 }, { "epoch": 13.31, "learning_rate": 5.325876817895701e-06, "loss": 0.0179, "step": 43254 }, { "epoch": 13.31, "learning_rate": 5.325436333325727e-06, "loss": 0.0212, "step": 43255 }, { "epoch": 13.31, "learning_rate": 5.32499586036143e-06, "loss": 0.0211, "step": 43256 }, { "epoch": 13.31, "learning_rate": 5.324555399003912e-06, "loss": 0.0233, "step": 43257 }, { "epoch": 13.31, "learning_rate": 5.3241149492542635e-06, "loss": 0.0233, "step": 43258 }, { "epoch": 13.31, "learning_rate": 5.323674511113579e-06, "loss": 0.023, "step": 43259 }, { "epoch": 13.31, "learning_rate": 5.323234084582948e-06, "loss": 0.024, "step": 43260 }, { "epoch": 13.31, "learning_rate": 5.322793669663462e-06, "loss": 0.0081, "step": 43261 }, { "epoch": 13.31, "learning_rate": 5.32235326635622e-06, "loss": 0.0075, "step": 43262 }, { "epoch": 13.31, "learning_rate": 5.321912874662316e-06, "loss": 0.0278, "step": 43263 }, { "epoch": 13.31, "learning_rate": 5.32147249458284e-06, "loss": 0.0243, "step": 43264 }, { "epoch": 13.31, "learning_rate": 5.321032126118888e-06, "loss": 0.0148, "step": 43265 }, { "epoch": 13.31, "learning_rate": 5.32059176927155e-06, "loss": 0.0494, "step": 43266 }, { "epoch": 13.31, "learning_rate": 5.320151424041922e-06, "loss": 0.0251, "step": 43267 }, { "epoch": 13.31, "learning_rate": 5.319711090431097e-06, "loss": 0.0264, "step": 43268 }, { "epoch": 13.31, "learning_rate": 5.3192707684401665e-06, "loss": 0.0242, "step": 43269 }, { "epoch": 13.31, "learning_rate": 5.318830458070226e-06, "loss": 0.0186, "step": 43270 }, { "epoch": 13.31, "learning_rate": 5.318390159322362e-06, "loss": 0.0157, "step": 43271 }, { "epoch": 13.31, "learning_rate": 5.3179498721976806e-06, "loss": 0.0154, "step": 43272 }, { "epoch": 13.31, "learning_rate": 5.317509596697266e-06, "loss": 0.0138, "step": 43273 }, { "epoch": 13.31, "learning_rate": 5.317069332822211e-06, "loss": 0.0206, "step": 43274 }, { "epoch": 13.32, "learning_rate": 5.3166290805736095e-06, "loss": 0.0191, "step": 43275 }, { "epoch": 13.32, "learning_rate": 5.3161888399525545e-06, "loss": 0.0217, "step": 43276 }, { "epoch": 13.32, "learning_rate": 5.315748610960141e-06, "loss": 0.0276, "step": 43277 }, { "epoch": 13.32, "learning_rate": 5.315308393597462e-06, "loss": 0.0186, "step": 43278 }, { "epoch": 13.32, "learning_rate": 5.31486818786561e-06, "loss": 0.0319, "step": 43279 }, { "epoch": 13.32, "learning_rate": 5.314427993765679e-06, "loss": 0.0298, "step": 43280 }, { "epoch": 13.32, "learning_rate": 5.313987811298753e-06, "loss": 0.0144, "step": 43281 }, { "epoch": 13.32, "learning_rate": 5.313547640465937e-06, "loss": 0.0103, "step": 43282 }, { "epoch": 13.32, "learning_rate": 5.313107481268318e-06, "loss": 0.0192, "step": 43283 }, { "epoch": 13.32, "learning_rate": 5.31266733370699e-06, "loss": 0.0237, "step": 43284 }, { "epoch": 13.32, "learning_rate": 5.3122271977830456e-06, "loss": 0.0233, "step": 43285 }, { "epoch": 13.32, "learning_rate": 5.3117870734975765e-06, "loss": 0.0213, "step": 43286 }, { "epoch": 13.32, "learning_rate": 5.311346960851677e-06, "loss": 0.0164, "step": 43287 }, { "epoch": 13.32, "learning_rate": 5.310906859846439e-06, "loss": 0.0098, "step": 43288 }, { "epoch": 13.32, "learning_rate": 5.310466770482953e-06, "loss": 0.0251, "step": 43289 }, { "epoch": 13.32, "learning_rate": 5.310026692762316e-06, "loss": 0.0134, "step": 43290 }, { "epoch": 13.32, "learning_rate": 5.309586626685613e-06, "loss": 0.0267, "step": 43291 }, { "epoch": 13.32, "learning_rate": 5.309146572253947e-06, "loss": 0.0103, "step": 43292 }, { "epoch": 13.32, "learning_rate": 5.308706529468408e-06, "loss": 0.0243, "step": 43293 }, { "epoch": 13.32, "learning_rate": 5.3082664983300835e-06, "loss": 0.0217, "step": 43294 }, { "epoch": 13.32, "learning_rate": 5.307826478840068e-06, "loss": 0.0206, "step": 43295 }, { "epoch": 13.32, "learning_rate": 5.30738647099945e-06, "loss": 0.0205, "step": 43296 }, { "epoch": 13.32, "learning_rate": 5.306946474809331e-06, "loss": 0.0201, "step": 43297 }, { "epoch": 13.32, "learning_rate": 5.3065064902707985e-06, "loss": 0.0198, "step": 43298 }, { "epoch": 13.32, "learning_rate": 5.306066517384945e-06, "loss": 0.0258, "step": 43299 }, { "epoch": 13.32, "learning_rate": 5.305626556152863e-06, "loss": 0.0184, "step": 43300 }, { "epoch": 13.32, "learning_rate": 5.305186606575644e-06, "loss": 0.0328, "step": 43301 }, { "epoch": 13.32, "learning_rate": 5.304746668654382e-06, "loss": 0.0122, "step": 43302 }, { "epoch": 13.32, "learning_rate": 5.304306742390167e-06, "loss": 0.0111, "step": 43303 }, { "epoch": 13.32, "learning_rate": 5.3038668277840935e-06, "loss": 0.0208, "step": 43304 }, { "epoch": 13.32, "learning_rate": 5.303426924837252e-06, "loss": 0.0146, "step": 43305 }, { "epoch": 13.32, "learning_rate": 5.302987033550731e-06, "loss": 0.0339, "step": 43306 }, { "epoch": 13.33, "learning_rate": 5.302547153925636e-06, "loss": 0.0156, "step": 43307 }, { "epoch": 13.33, "learning_rate": 5.302107285963045e-06, "loss": 0.0112, "step": 43308 }, { "epoch": 13.33, "learning_rate": 5.301667429664056e-06, "loss": 0.0293, "step": 43309 }, { "epoch": 13.33, "learning_rate": 5.30122758502976e-06, "loss": 0.015, "step": 43310 }, { "epoch": 13.33, "learning_rate": 5.300787752061245e-06, "loss": 0.0096, "step": 43311 }, { "epoch": 13.33, "learning_rate": 5.300347930759611e-06, "loss": 0.012, "step": 43312 }, { "epoch": 13.33, "learning_rate": 5.299908121125945e-06, "loss": 0.0207, "step": 43313 }, { "epoch": 13.33, "learning_rate": 5.299468323161346e-06, "loss": 0.0218, "step": 43314 }, { "epoch": 13.33, "learning_rate": 5.299028536866895e-06, "loss": 0.0089, "step": 43315 }, { "epoch": 13.33, "learning_rate": 5.298588762243684e-06, "loss": 0.0198, "step": 43316 }, { "epoch": 13.33, "learning_rate": 5.298148999292815e-06, "loss": 0.0156, "step": 43317 }, { "epoch": 13.33, "learning_rate": 5.2977092480153734e-06, "loss": 0.0197, "step": 43318 }, { "epoch": 13.33, "learning_rate": 5.2972695084124525e-06, "loss": 0.0087, "step": 43319 }, { "epoch": 13.33, "learning_rate": 5.296829780485144e-06, "loss": 0.0351, "step": 43320 }, { "epoch": 13.33, "learning_rate": 5.2963900642345385e-06, "loss": 0.017, "step": 43321 }, { "epoch": 13.33, "learning_rate": 5.295950359661729e-06, "loss": 0.0262, "step": 43322 }, { "epoch": 13.33, "learning_rate": 5.295510666767807e-06, "loss": 0.0105, "step": 43323 }, { "epoch": 13.33, "learning_rate": 5.295070985553863e-06, "loss": 0.0186, "step": 43324 }, { "epoch": 13.33, "learning_rate": 5.294631316020988e-06, "loss": 0.0268, "step": 43325 }, { "epoch": 13.33, "learning_rate": 5.294191658170273e-06, "loss": 0.029, "step": 43326 }, { "epoch": 13.33, "learning_rate": 5.2937520120028156e-06, "loss": 0.0151, "step": 43327 }, { "epoch": 13.33, "learning_rate": 5.293312377519706e-06, "loss": 0.0156, "step": 43328 }, { "epoch": 13.33, "learning_rate": 5.292872754722029e-06, "loss": 0.0142, "step": 43329 }, { "epoch": 13.33, "learning_rate": 5.29243314361088e-06, "loss": 0.0372, "step": 43330 }, { "epoch": 13.33, "learning_rate": 5.291993544187347e-06, "loss": 0.0155, "step": 43331 }, { "epoch": 13.33, "learning_rate": 5.291553956452529e-06, "loss": 0.0221, "step": 43332 }, { "epoch": 13.33, "learning_rate": 5.291114380407512e-06, "loss": 0.0202, "step": 43333 }, { "epoch": 13.33, "learning_rate": 5.2906748160533895e-06, "loss": 0.016, "step": 43334 }, { "epoch": 13.33, "learning_rate": 5.290235263391255e-06, "loss": 0.0214, "step": 43335 }, { "epoch": 13.33, "learning_rate": 5.2897957224221884e-06, "loss": 0.0172, "step": 43336 }, { "epoch": 13.33, "learning_rate": 5.289356193147295e-06, "loss": 0.0192, "step": 43337 }, { "epoch": 13.33, "learning_rate": 5.288916675567659e-06, "loss": 0.0172, "step": 43338 }, { "epoch": 13.33, "learning_rate": 5.288477169684372e-06, "loss": 0.0101, "step": 43339 }, { "epoch": 13.34, "learning_rate": 5.288037675498528e-06, "loss": 0.0271, "step": 43340 }, { "epoch": 13.34, "learning_rate": 5.287598193011214e-06, "loss": 0.0243, "step": 43341 }, { "epoch": 13.34, "learning_rate": 5.287158722223524e-06, "loss": 0.023, "step": 43342 }, { "epoch": 13.34, "learning_rate": 5.286719263136549e-06, "loss": 0.0347, "step": 43343 }, { "epoch": 13.34, "learning_rate": 5.28627981575138e-06, "loss": 0.0115, "step": 43344 }, { "epoch": 13.34, "learning_rate": 5.285840380069106e-06, "loss": 0.0077, "step": 43345 }, { "epoch": 13.34, "learning_rate": 5.285400956090816e-06, "loss": 0.0154, "step": 43346 }, { "epoch": 13.34, "learning_rate": 5.284961543817608e-06, "loss": 0.0272, "step": 43347 }, { "epoch": 13.34, "learning_rate": 5.28452214325057e-06, "loss": 0.0236, "step": 43348 }, { "epoch": 13.34, "learning_rate": 5.2840827543907954e-06, "loss": 0.0158, "step": 43349 }, { "epoch": 13.34, "learning_rate": 5.283643377239369e-06, "loss": 0.0323, "step": 43350 }, { "epoch": 13.34, "learning_rate": 5.283204011797381e-06, "loss": 0.0246, "step": 43351 }, { "epoch": 13.34, "learning_rate": 5.282764658065929e-06, "loss": 0.0257, "step": 43352 }, { "epoch": 13.34, "learning_rate": 5.2823253160461e-06, "loss": 0.0183, "step": 43353 }, { "epoch": 13.34, "learning_rate": 5.281885985738986e-06, "loss": 0.0082, "step": 43354 }, { "epoch": 13.34, "learning_rate": 5.281446667145677e-06, "loss": 0.0294, "step": 43355 }, { "epoch": 13.34, "learning_rate": 5.281007360267263e-06, "loss": 0.0225, "step": 43356 }, { "epoch": 13.34, "learning_rate": 5.280568065104836e-06, "loss": 0.0137, "step": 43357 }, { "epoch": 13.34, "learning_rate": 5.280128781659488e-06, "loss": 0.0163, "step": 43358 }, { "epoch": 13.34, "learning_rate": 5.279689509932307e-06, "loss": 0.0234, "step": 43359 }, { "epoch": 13.34, "learning_rate": 5.279250249924384e-06, "loss": 0.0183, "step": 43360 }, { "epoch": 13.34, "learning_rate": 5.278811001636806e-06, "loss": 0.0081, "step": 43361 }, { "epoch": 13.34, "learning_rate": 5.278371765070673e-06, "loss": 0.0336, "step": 43362 }, { "epoch": 13.34, "learning_rate": 5.277932540227069e-06, "loss": 0.0152, "step": 43363 }, { "epoch": 13.34, "learning_rate": 5.277493327107085e-06, "loss": 0.0253, "step": 43364 }, { "epoch": 13.34, "learning_rate": 5.277054125711811e-06, "loss": 0.025, "step": 43365 }, { "epoch": 13.34, "learning_rate": 5.276614936042336e-06, "loss": 0.0301, "step": 43366 }, { "epoch": 13.34, "learning_rate": 5.276175758099755e-06, "loss": 0.0301, "step": 43367 }, { "epoch": 13.34, "learning_rate": 5.275736591885157e-06, "loss": 0.0125, "step": 43368 }, { "epoch": 13.34, "learning_rate": 5.275297437399636e-06, "loss": 0.0272, "step": 43369 }, { "epoch": 13.34, "learning_rate": 5.274858294644272e-06, "loss": 0.0196, "step": 43370 }, { "epoch": 13.34, "learning_rate": 5.274419163620159e-06, "loss": 0.0254, "step": 43371 }, { "epoch": 13.35, "learning_rate": 5.273980044328392e-06, "loss": 0.039, "step": 43372 }, { "epoch": 13.35, "learning_rate": 5.273540936770059e-06, "loss": 0.0189, "step": 43373 }, { "epoch": 13.35, "learning_rate": 5.273101840946249e-06, "loss": 0.0225, "step": 43374 }, { "epoch": 13.35, "learning_rate": 5.272662756858053e-06, "loss": 0.0364, "step": 43375 }, { "epoch": 13.35, "learning_rate": 5.272223684506561e-06, "loss": 0.0141, "step": 43376 }, { "epoch": 13.35, "learning_rate": 5.271784623892862e-06, "loss": 0.029, "step": 43377 }, { "epoch": 13.35, "learning_rate": 5.271345575018049e-06, "loss": 0.0323, "step": 43378 }, { "epoch": 13.35, "learning_rate": 5.270906537883209e-06, "loss": 0.0241, "step": 43379 }, { "epoch": 13.35, "learning_rate": 5.270467512489433e-06, "loss": 0.0265, "step": 43380 }, { "epoch": 13.35, "learning_rate": 5.270028498837808e-06, "loss": 0.028, "step": 43381 }, { "epoch": 13.35, "learning_rate": 5.26958949692943e-06, "loss": 0.0227, "step": 43382 }, { "epoch": 13.35, "learning_rate": 5.269150506765392e-06, "loss": 0.0275, "step": 43383 }, { "epoch": 13.35, "learning_rate": 5.268711528346771e-06, "loss": 0.0173, "step": 43384 }, { "epoch": 13.35, "learning_rate": 5.268272561674666e-06, "loss": 0.015, "step": 43385 }, { "epoch": 13.35, "learning_rate": 5.26783360675016e-06, "loss": 0.0261, "step": 43386 }, { "epoch": 13.35, "learning_rate": 5.267394663574351e-06, "loss": 0.0101, "step": 43387 }, { "epoch": 13.35, "learning_rate": 5.266955732148325e-06, "loss": 0.0212, "step": 43388 }, { "epoch": 13.35, "learning_rate": 5.266516812473173e-06, "loss": 0.0161, "step": 43389 }, { "epoch": 13.35, "learning_rate": 5.266077904549987e-06, "loss": 0.0127, "step": 43390 }, { "epoch": 13.35, "learning_rate": 5.265639008379846e-06, "loss": 0.0124, "step": 43391 }, { "epoch": 13.35, "learning_rate": 5.2652001239638495e-06, "loss": 0.0297, "step": 43392 }, { "epoch": 13.35, "learning_rate": 5.2647612513030855e-06, "loss": 0.02, "step": 43393 }, { "epoch": 13.35, "learning_rate": 5.264322390398643e-06, "loss": 0.0327, "step": 43394 }, { "epoch": 13.35, "learning_rate": 5.263883541251611e-06, "loss": 0.0139, "step": 43395 }, { "epoch": 13.35, "learning_rate": 5.263444703863079e-06, "loss": 0.0781, "step": 43396 }, { "epoch": 13.35, "learning_rate": 5.2630058782341375e-06, "loss": 0.0239, "step": 43397 }, { "epoch": 13.35, "learning_rate": 5.262567064365875e-06, "loss": 0.0243, "step": 43398 }, { "epoch": 13.35, "learning_rate": 5.26212826225938e-06, "loss": 0.01, "step": 43399 }, { "epoch": 13.35, "learning_rate": 5.2616894719157455e-06, "loss": 0.0407, "step": 43400 }, { "epoch": 13.35, "learning_rate": 5.261250693336054e-06, "loss": 0.021, "step": 43401 }, { "epoch": 13.35, "learning_rate": 5.260811926521403e-06, "loss": 0.019, "step": 43402 }, { "epoch": 13.35, "learning_rate": 5.260373171472879e-06, "loss": 0.0274, "step": 43403 }, { "epoch": 13.35, "learning_rate": 5.2599344281915745e-06, "loss": 0.0319, "step": 43404 }, { "epoch": 13.36, "learning_rate": 5.259495696678571e-06, "loss": 0.0217, "step": 43405 }, { "epoch": 13.36, "learning_rate": 5.259056976934957e-06, "loss": 0.0214, "step": 43406 }, { "epoch": 13.36, "learning_rate": 5.25861826896183e-06, "loss": 0.0274, "step": 43407 }, { "epoch": 13.36, "learning_rate": 5.258179572760277e-06, "loss": 0.0114, "step": 43408 }, { "epoch": 13.36, "learning_rate": 5.2577408883313844e-06, "loss": 0.0075, "step": 43409 }, { "epoch": 13.36, "learning_rate": 5.257302215676243e-06, "loss": 0.0277, "step": 43410 }, { "epoch": 13.36, "learning_rate": 5.2568635547959435e-06, "loss": 0.0279, "step": 43411 }, { "epoch": 13.36, "learning_rate": 5.2564249056915704e-06, "loss": 0.0234, "step": 43412 }, { "epoch": 13.36, "learning_rate": 5.255986268364217e-06, "loss": 0.0547, "step": 43413 }, { "epoch": 13.36, "learning_rate": 5.255547642814971e-06, "loss": 0.0128, "step": 43414 }, { "epoch": 13.36, "learning_rate": 5.25510902904492e-06, "loss": 0.0182, "step": 43415 }, { "epoch": 13.36, "learning_rate": 5.2546704270551506e-06, "loss": 0.016, "step": 43416 }, { "epoch": 13.36, "learning_rate": 5.254231836846764e-06, "loss": 0.0078, "step": 43417 }, { "epoch": 13.36, "learning_rate": 5.253793258420835e-06, "loss": 0.0262, "step": 43418 }, { "epoch": 13.36, "learning_rate": 5.2533546917784585e-06, "loss": 0.0142, "step": 43419 }, { "epoch": 13.36, "learning_rate": 5.2529161369207225e-06, "loss": 0.0309, "step": 43420 }, { "epoch": 13.36, "learning_rate": 5.252477593848713e-06, "loss": 0.017, "step": 43421 }, { "epoch": 13.36, "learning_rate": 5.252039062563523e-06, "loss": 0.0174, "step": 43422 }, { "epoch": 13.36, "learning_rate": 5.251600543066242e-06, "loss": 0.0244, "step": 43423 }, { "epoch": 13.36, "learning_rate": 5.251162035357957e-06, "loss": 0.0437, "step": 43424 }, { "epoch": 13.36, "learning_rate": 5.2507235394397595e-06, "loss": 0.029, "step": 43425 }, { "epoch": 13.36, "learning_rate": 5.250285055312727e-06, "loss": 0.0157, "step": 43426 }, { "epoch": 13.36, "learning_rate": 5.249846582977962e-06, "loss": 0.0578, "step": 43427 }, { "epoch": 13.36, "learning_rate": 5.249408122436545e-06, "loss": 0.0239, "step": 43428 }, { "epoch": 13.36, "learning_rate": 5.248969673689568e-06, "loss": 0.0196, "step": 43429 }, { "epoch": 13.36, "learning_rate": 5.248531236738119e-06, "loss": 0.0288, "step": 43430 }, { "epoch": 13.36, "learning_rate": 5.248092811583286e-06, "loss": 0.0302, "step": 43431 }, { "epoch": 13.36, "learning_rate": 5.247654398226157e-06, "loss": 0.0242, "step": 43432 }, { "epoch": 13.36, "learning_rate": 5.247215996667821e-06, "loss": 0.0196, "step": 43433 }, { "epoch": 13.36, "learning_rate": 5.246777606909366e-06, "loss": 0.0224, "step": 43434 }, { "epoch": 13.36, "learning_rate": 5.24633922895188e-06, "loss": 0.0237, "step": 43435 }, { "epoch": 13.36, "learning_rate": 5.245900862796449e-06, "loss": 0.0142, "step": 43436 }, { "epoch": 13.37, "learning_rate": 5.245462508444168e-06, "loss": 0.0347, "step": 43437 }, { "epoch": 13.37, "learning_rate": 5.245024165896126e-06, "loss": 0.0276, "step": 43438 }, { "epoch": 13.37, "learning_rate": 5.244585835153404e-06, "loss": 0.0297, "step": 43439 }, { "epoch": 13.37, "learning_rate": 5.244147516217093e-06, "loss": 0.0209, "step": 43440 }, { "epoch": 13.37, "learning_rate": 5.243709209088277e-06, "loss": 0.0141, "step": 43441 }, { "epoch": 13.37, "learning_rate": 5.243270913768054e-06, "loss": 0.0376, "step": 43442 }, { "epoch": 13.37, "learning_rate": 5.2428326302575065e-06, "loss": 0.0168, "step": 43443 }, { "epoch": 13.37, "learning_rate": 5.242394358557722e-06, "loss": 0.0673, "step": 43444 }, { "epoch": 13.37, "learning_rate": 5.24195609866979e-06, "loss": 0.0295, "step": 43445 }, { "epoch": 13.37, "learning_rate": 5.241517850594799e-06, "loss": 0.0278, "step": 43446 }, { "epoch": 13.37, "learning_rate": 5.241079614333836e-06, "loss": 0.0208, "step": 43447 }, { "epoch": 13.37, "learning_rate": 5.240641389887989e-06, "loss": 0.0046, "step": 43448 }, { "epoch": 13.37, "learning_rate": 5.240203177258347e-06, "loss": 0.0211, "step": 43449 }, { "epoch": 13.37, "learning_rate": 5.239764976445998e-06, "loss": 0.0215, "step": 43450 }, { "epoch": 13.37, "learning_rate": 5.239326787452028e-06, "loss": 0.0092, "step": 43451 }, { "epoch": 13.37, "learning_rate": 5.238888610277526e-06, "loss": 0.0248, "step": 43452 }, { "epoch": 13.37, "learning_rate": 5.2384504449235794e-06, "loss": 0.0173, "step": 43453 }, { "epoch": 13.37, "learning_rate": 5.238012291391279e-06, "loss": 0.0313, "step": 43454 }, { "epoch": 13.37, "learning_rate": 5.237574149681708e-06, "loss": 0.0254, "step": 43455 }, { "epoch": 13.37, "learning_rate": 5.237136019795954e-06, "loss": 0.0215, "step": 43456 }, { "epoch": 13.37, "learning_rate": 5.23669790173511e-06, "loss": 0.0154, "step": 43457 }, { "epoch": 13.37, "learning_rate": 5.236259795500263e-06, "loss": 0.0239, "step": 43458 }, { "epoch": 13.37, "learning_rate": 5.235821701092501e-06, "loss": 0.0186, "step": 43459 }, { "epoch": 13.37, "learning_rate": 5.235383618512906e-06, "loss": 0.0132, "step": 43460 }, { "epoch": 13.37, "learning_rate": 5.234945547762564e-06, "loss": 0.0104, "step": 43461 }, { "epoch": 13.37, "learning_rate": 5.234507488842573e-06, "loss": 0.0121, "step": 43462 }, { "epoch": 13.37, "learning_rate": 5.2340694417540135e-06, "loss": 0.0277, "step": 43463 }, { "epoch": 13.37, "learning_rate": 5.2336314064979766e-06, "loss": 0.0216, "step": 43464 }, { "epoch": 13.37, "learning_rate": 5.2331933830755456e-06, "loss": 0.0161, "step": 43465 }, { "epoch": 13.37, "learning_rate": 5.232755371487812e-06, "loss": 0.0243, "step": 43466 }, { "epoch": 13.37, "learning_rate": 5.232317371735862e-06, "loss": 0.0187, "step": 43467 }, { "epoch": 13.37, "learning_rate": 5.231879383820782e-06, "loss": 0.023, "step": 43468 }, { "epoch": 13.37, "learning_rate": 5.231441407743659e-06, "loss": 0.025, "step": 43469 }, { "epoch": 13.38, "learning_rate": 5.231003443505582e-06, "loss": 0.0045, "step": 43470 }, { "epoch": 13.38, "learning_rate": 5.230565491107636e-06, "loss": 0.0098, "step": 43471 }, { "epoch": 13.38, "learning_rate": 5.230127550550916e-06, "loss": 0.0257, "step": 43472 }, { "epoch": 13.38, "learning_rate": 5.229689621836499e-06, "loss": 0.0165, "step": 43473 }, { "epoch": 13.38, "learning_rate": 5.229251704965477e-06, "loss": 0.0142, "step": 43474 }, { "epoch": 13.38, "learning_rate": 5.2288137999389364e-06, "loss": 0.0363, "step": 43475 }, { "epoch": 13.38, "learning_rate": 5.2283759067579635e-06, "loss": 0.0101, "step": 43476 }, { "epoch": 13.38, "learning_rate": 5.2279380254236475e-06, "loss": 0.0332, "step": 43477 }, { "epoch": 13.38, "learning_rate": 5.227500155937078e-06, "loss": 0.0075, "step": 43478 }, { "epoch": 13.38, "learning_rate": 5.227062298299337e-06, "loss": 0.0316, "step": 43479 }, { "epoch": 13.38, "learning_rate": 5.2266244525115175e-06, "loss": 0.0174, "step": 43480 }, { "epoch": 13.38, "learning_rate": 5.2261866185746945e-06, "loss": 0.0238, "step": 43481 }, { "epoch": 13.38, "learning_rate": 5.225748796489966e-06, "loss": 0.0305, "step": 43482 }, { "epoch": 13.38, "learning_rate": 5.225310986258419e-06, "loss": 0.0199, "step": 43483 }, { "epoch": 13.38, "learning_rate": 5.224873187881136e-06, "loss": 0.0276, "step": 43484 }, { "epoch": 13.38, "learning_rate": 5.224435401359207e-06, "loss": 0.0076, "step": 43485 }, { "epoch": 13.38, "learning_rate": 5.223997626693716e-06, "loss": 0.0171, "step": 43486 }, { "epoch": 13.38, "learning_rate": 5.223559863885752e-06, "loss": 0.0192, "step": 43487 }, { "epoch": 13.38, "learning_rate": 5.223122112936401e-06, "loss": 0.0232, "step": 43488 }, { "epoch": 13.38, "learning_rate": 5.222684373846749e-06, "loss": 0.0256, "step": 43489 }, { "epoch": 13.38, "learning_rate": 5.222246646617886e-06, "loss": 0.0232, "step": 43490 }, { "epoch": 13.38, "learning_rate": 5.22180893125089e-06, "loss": 0.0293, "step": 43491 }, { "epoch": 13.38, "learning_rate": 5.221371227746859e-06, "loss": 0.0235, "step": 43492 }, { "epoch": 13.38, "learning_rate": 5.22093353610688e-06, "loss": 0.0179, "step": 43493 }, { "epoch": 13.38, "learning_rate": 5.22049585633203e-06, "loss": 0.0186, "step": 43494 }, { "epoch": 13.38, "learning_rate": 5.2200581884234e-06, "loss": 0.0207, "step": 43495 }, { "epoch": 13.38, "learning_rate": 5.219620532382072e-06, "loss": 0.0173, "step": 43496 }, { "epoch": 13.38, "learning_rate": 5.219182888209145e-06, "loss": 0.0176, "step": 43497 }, { "epoch": 13.38, "learning_rate": 5.218745255905695e-06, "loss": 0.0154, "step": 43498 }, { "epoch": 13.38, "learning_rate": 5.2183076354728125e-06, "loss": 0.0328, "step": 43499 }, { "epoch": 13.38, "learning_rate": 5.217870026911582e-06, "loss": 0.0219, "step": 43500 }, { "epoch": 13.38, "learning_rate": 5.217432430223092e-06, "loss": 0.0243, "step": 43501 }, { "epoch": 13.39, "learning_rate": 5.2169948454084275e-06, "loss": 0.0251, "step": 43502 }, { "epoch": 13.39, "learning_rate": 5.216557272468675e-06, "loss": 0.012, "step": 43503 }, { "epoch": 13.39, "learning_rate": 5.216119711404922e-06, "loss": 0.0146, "step": 43504 }, { "epoch": 13.39, "learning_rate": 5.215682162218254e-06, "loss": 0.012, "step": 43505 }, { "epoch": 13.39, "learning_rate": 5.215244624909756e-06, "loss": 0.025, "step": 43506 }, { "epoch": 13.39, "learning_rate": 5.214807099480518e-06, "loss": 0.015, "step": 43507 }, { "epoch": 13.39, "learning_rate": 5.214369585931622e-06, "loss": 0.0189, "step": 43508 }, { "epoch": 13.39, "learning_rate": 5.2139320842641575e-06, "loss": 0.0168, "step": 43509 }, { "epoch": 13.39, "learning_rate": 5.2134945944792095e-06, "loss": 0.0173, "step": 43510 }, { "epoch": 13.39, "learning_rate": 5.2130571165778586e-06, "loss": 0.0226, "step": 43511 }, { "epoch": 13.39, "learning_rate": 5.2126196505612015e-06, "loss": 0.0113, "step": 43512 }, { "epoch": 13.39, "learning_rate": 5.212182196430317e-06, "loss": 0.0356, "step": 43513 }, { "epoch": 13.39, "learning_rate": 5.211744754186301e-06, "loss": 0.0194, "step": 43514 }, { "epoch": 13.39, "learning_rate": 5.211307323830226e-06, "loss": 0.0242, "step": 43515 }, { "epoch": 13.39, "learning_rate": 5.210869905363178e-06, "loss": 0.0252, "step": 43516 }, { "epoch": 13.39, "learning_rate": 5.210432498786255e-06, "loss": 0.0312, "step": 43517 }, { "epoch": 13.39, "learning_rate": 5.209995104100538e-06, "loss": 0.0132, "step": 43518 }, { "epoch": 13.39, "learning_rate": 5.20955772130711e-06, "loss": 0.0324, "step": 43519 }, { "epoch": 13.39, "learning_rate": 5.20912035040706e-06, "loss": 0.0297, "step": 43520 }, { "epoch": 13.39, "learning_rate": 5.20868299140147e-06, "loss": 0.0145, "step": 43521 }, { "epoch": 13.39, "learning_rate": 5.2082456442914305e-06, "loss": 0.0064, "step": 43522 }, { "epoch": 13.39, "learning_rate": 5.207808309078026e-06, "loss": 0.0194, "step": 43523 }, { "epoch": 13.39, "learning_rate": 5.20737098576234e-06, "loss": 0.0289, "step": 43524 }, { "epoch": 13.39, "learning_rate": 5.20693367434546e-06, "loss": 0.0265, "step": 43525 }, { "epoch": 13.39, "learning_rate": 5.206496374828468e-06, "loss": 0.0153, "step": 43526 }, { "epoch": 13.39, "learning_rate": 5.206059087212461e-06, "loss": 0.022, "step": 43527 }, { "epoch": 13.39, "learning_rate": 5.205621811498514e-06, "loss": 0.0155, "step": 43528 }, { "epoch": 13.39, "learning_rate": 5.205184547687714e-06, "loss": 0.0195, "step": 43529 }, { "epoch": 13.39, "learning_rate": 5.204747295781147e-06, "loss": 0.0143, "step": 43530 }, { "epoch": 13.39, "learning_rate": 5.204310055779897e-06, "loss": 0.0126, "step": 43531 }, { "epoch": 13.39, "learning_rate": 5.203872827685058e-06, "loss": 0.0064, "step": 43532 }, { "epoch": 13.39, "learning_rate": 5.203435611497708e-06, "loss": 0.0384, "step": 43533 }, { "epoch": 13.39, "learning_rate": 5.202998407218936e-06, "loss": 0.0346, "step": 43534 }, { "epoch": 13.4, "learning_rate": 5.2025612148498265e-06, "loss": 0.0125, "step": 43535 }, { "epoch": 13.4, "learning_rate": 5.202124034391458e-06, "loss": 0.0163, "step": 43536 }, { "epoch": 13.4, "learning_rate": 5.201686865844926e-06, "loss": 0.0173, "step": 43537 }, { "epoch": 13.4, "learning_rate": 5.201249709211312e-06, "loss": 0.0129, "step": 43538 }, { "epoch": 13.4, "learning_rate": 5.2008125644917014e-06, "loss": 0.0261, "step": 43539 }, { "epoch": 13.4, "learning_rate": 5.200375431687179e-06, "loss": 0.0239, "step": 43540 }, { "epoch": 13.4, "learning_rate": 5.199938310798831e-06, "loss": 0.0271, "step": 43541 }, { "epoch": 13.4, "learning_rate": 5.199501201827741e-06, "loss": 0.0182, "step": 43542 }, { "epoch": 13.4, "learning_rate": 5.1990641047749955e-06, "loss": 0.0199, "step": 43543 }, { "epoch": 13.4, "learning_rate": 5.198627019641681e-06, "loss": 0.0157, "step": 43544 }, { "epoch": 13.4, "learning_rate": 5.198189946428877e-06, "loss": 0.0247, "step": 43545 }, { "epoch": 13.4, "learning_rate": 5.1977528851376756e-06, "loss": 0.021, "step": 43546 }, { "epoch": 13.4, "learning_rate": 5.197315835769159e-06, "loss": 0.0177, "step": 43547 }, { "epoch": 13.4, "learning_rate": 5.196878798324418e-06, "loss": 0.0308, "step": 43548 }, { "epoch": 13.4, "learning_rate": 5.1964417728045265e-06, "loss": 0.041, "step": 43549 }, { "epoch": 13.4, "learning_rate": 5.196004759210572e-06, "loss": 0.0238, "step": 43550 }, { "epoch": 13.4, "learning_rate": 5.195567757543647e-06, "loss": 0.0205, "step": 43551 }, { "epoch": 13.4, "learning_rate": 5.19513076780483e-06, "loss": 0.0195, "step": 43552 }, { "epoch": 13.4, "learning_rate": 5.19469378999521e-06, "loss": 0.0382, "step": 43553 }, { "epoch": 13.4, "learning_rate": 5.194256824115869e-06, "loss": 0.0295, "step": 43554 }, { "epoch": 13.4, "learning_rate": 5.193819870167893e-06, "loss": 0.03, "step": 43555 }, { "epoch": 13.4, "learning_rate": 5.193382928152366e-06, "loss": 0.0183, "step": 43556 }, { "epoch": 13.4, "learning_rate": 5.192945998070373e-06, "loss": 0.0145, "step": 43557 }, { "epoch": 13.4, "learning_rate": 5.192509079923e-06, "loss": 0.0138, "step": 43558 }, { "epoch": 13.4, "learning_rate": 5.192072173711331e-06, "loss": 0.031, "step": 43559 }, { "epoch": 13.4, "learning_rate": 5.191635279436447e-06, "loss": 0.0303, "step": 43560 }, { "epoch": 13.4, "learning_rate": 5.1911983970994395e-06, "loss": 0.0195, "step": 43561 }, { "epoch": 13.4, "learning_rate": 5.190761526701395e-06, "loss": 0.0159, "step": 43562 }, { "epoch": 13.4, "learning_rate": 5.1903246682433875e-06, "loss": 0.0124, "step": 43563 }, { "epoch": 13.4, "learning_rate": 5.189887821726509e-06, "loss": 0.0284, "step": 43564 }, { "epoch": 13.4, "learning_rate": 5.189450987151836e-06, "loss": 0.0101, "step": 43565 }, { "epoch": 13.4, "learning_rate": 5.189014164520465e-06, "loss": 0.0177, "step": 43566 }, { "epoch": 13.41, "learning_rate": 5.1885773538334746e-06, "loss": 0.0093, "step": 43567 }, { "epoch": 13.41, "learning_rate": 5.18814055509195e-06, "loss": 0.0193, "step": 43568 }, { "epoch": 13.41, "learning_rate": 5.187703768296978e-06, "loss": 0.0173, "step": 43569 }, { "epoch": 13.41, "learning_rate": 5.187266993449633e-06, "loss": 0.0175, "step": 43570 }, { "epoch": 13.41, "learning_rate": 5.18683023055101e-06, "loss": 0.0102, "step": 43571 }, { "epoch": 13.41, "learning_rate": 5.186393479602189e-06, "loss": 0.0129, "step": 43572 }, { "epoch": 13.41, "learning_rate": 5.185956740604257e-06, "loss": 0.0131, "step": 43573 }, { "epoch": 13.41, "learning_rate": 5.185520013558296e-06, "loss": 0.0092, "step": 43574 }, { "epoch": 13.41, "learning_rate": 5.185083298465392e-06, "loss": 0.0081, "step": 43575 }, { "epoch": 13.41, "learning_rate": 5.184646595326627e-06, "loss": 0.0252, "step": 43576 }, { "epoch": 13.41, "learning_rate": 5.184209904143086e-06, "loss": 0.0271, "step": 43577 }, { "epoch": 13.41, "learning_rate": 5.183773224915853e-06, "loss": 0.0278, "step": 43578 }, { "epoch": 13.41, "learning_rate": 5.183336557646015e-06, "loss": 0.0258, "step": 43579 }, { "epoch": 13.41, "learning_rate": 5.182899902334648e-06, "loss": 0.0161, "step": 43580 }, { "epoch": 13.41, "learning_rate": 5.1824632589828465e-06, "loss": 0.0294, "step": 43581 }, { "epoch": 13.41, "learning_rate": 5.18202662759169e-06, "loss": 0.0237, "step": 43582 }, { "epoch": 13.41, "learning_rate": 5.181590008162265e-06, "loss": 0.0262, "step": 43583 }, { "epoch": 13.41, "learning_rate": 5.181153400695651e-06, "loss": 0.0194, "step": 43584 }, { "epoch": 13.41, "learning_rate": 5.180716805192929e-06, "loss": 0.0285, "step": 43585 }, { "epoch": 13.41, "learning_rate": 5.180280221655191e-06, "loss": 0.0287, "step": 43586 }, { "epoch": 13.41, "learning_rate": 5.1798436500835205e-06, "loss": 0.0278, "step": 43587 }, { "epoch": 13.41, "learning_rate": 5.179407090478996e-06, "loss": 0.0196, "step": 43588 }, { "epoch": 13.41, "learning_rate": 5.178970542842705e-06, "loss": 0.0166, "step": 43589 }, { "epoch": 13.41, "learning_rate": 5.178534007175732e-06, "loss": 0.0173, "step": 43590 }, { "epoch": 13.41, "learning_rate": 5.178097483479157e-06, "loss": 0.0112, "step": 43591 }, { "epoch": 13.41, "learning_rate": 5.177660971754068e-06, "loss": 0.0164, "step": 43592 }, { "epoch": 13.41, "learning_rate": 5.177224472001545e-06, "loss": 0.0223, "step": 43593 }, { "epoch": 13.41, "learning_rate": 5.1767879842226745e-06, "loss": 0.0186, "step": 43594 }, { "epoch": 13.41, "learning_rate": 5.176351508418536e-06, "loss": 0.0228, "step": 43595 }, { "epoch": 13.41, "learning_rate": 5.175915044590222e-06, "loss": 0.0161, "step": 43596 }, { "epoch": 13.41, "learning_rate": 5.175478592738808e-06, "loss": 0.0444, "step": 43597 }, { "epoch": 13.41, "learning_rate": 5.17504215286538e-06, "loss": 0.025, "step": 43598 }, { "epoch": 13.41, "learning_rate": 5.17460572497102e-06, "loss": 0.0171, "step": 43599 }, { "epoch": 13.42, "learning_rate": 5.17416930905681e-06, "loss": 0.0251, "step": 43600 }, { "epoch": 13.42, "learning_rate": 5.173732905123841e-06, "loss": 0.0434, "step": 43601 }, { "epoch": 13.42, "learning_rate": 5.173296513173192e-06, "loss": 0.0167, "step": 43602 }, { "epoch": 13.42, "learning_rate": 5.172860133205946e-06, "loss": 0.0291, "step": 43603 }, { "epoch": 13.42, "learning_rate": 5.172423765223191e-06, "loss": 0.0117, "step": 43604 }, { "epoch": 13.42, "learning_rate": 5.1719874092259996e-06, "loss": 0.0154, "step": 43605 }, { "epoch": 13.42, "learning_rate": 5.1715510652154644e-06, "loss": 0.0215, "step": 43606 }, { "epoch": 13.42, "learning_rate": 5.171114733192667e-06, "loss": 0.0317, "step": 43607 }, { "epoch": 13.42, "learning_rate": 5.17067841315869e-06, "loss": 0.0173, "step": 43608 }, { "epoch": 13.42, "learning_rate": 5.170242105114616e-06, "loss": 0.0201, "step": 43609 }, { "epoch": 13.42, "learning_rate": 5.169805809061528e-06, "loss": 0.0276, "step": 43610 }, { "epoch": 13.42, "learning_rate": 5.169369525000512e-06, "loss": 0.02, "step": 43611 }, { "epoch": 13.42, "learning_rate": 5.168933252932647e-06, "loss": 0.027, "step": 43612 }, { "epoch": 13.42, "learning_rate": 5.16849699285902e-06, "loss": 0.0247, "step": 43613 }, { "epoch": 13.42, "learning_rate": 5.1680607447807116e-06, "loss": 0.0299, "step": 43614 }, { "epoch": 13.42, "learning_rate": 5.167624508698803e-06, "loss": 0.0173, "step": 43615 }, { "epoch": 13.42, "learning_rate": 5.1671882846143815e-06, "loss": 0.0175, "step": 43616 }, { "epoch": 13.42, "learning_rate": 5.166752072528535e-06, "loss": 0.0336, "step": 43617 }, { "epoch": 13.42, "learning_rate": 5.166315872442336e-06, "loss": 0.0185, "step": 43618 }, { "epoch": 13.42, "learning_rate": 5.165879684356869e-06, "loss": 0.0135, "step": 43619 }, { "epoch": 13.42, "learning_rate": 5.165443508273218e-06, "loss": 0.0249, "step": 43620 }, { "epoch": 13.42, "learning_rate": 5.16500734419247e-06, "loss": 0.027, "step": 43621 }, { "epoch": 13.42, "learning_rate": 5.164571192115706e-06, "loss": 0.029, "step": 43622 }, { "epoch": 13.42, "learning_rate": 5.164135052044008e-06, "loss": 0.0299, "step": 43623 }, { "epoch": 13.42, "learning_rate": 5.163698923978459e-06, "loss": 0.0223, "step": 43624 }, { "epoch": 13.42, "learning_rate": 5.163262807920142e-06, "loss": 0.016, "step": 43625 }, { "epoch": 13.42, "learning_rate": 5.16282670387014e-06, "loss": 0.0169, "step": 43626 }, { "epoch": 13.42, "learning_rate": 5.162390611829534e-06, "loss": 0.0392, "step": 43627 }, { "epoch": 13.42, "learning_rate": 5.161954531799409e-06, "loss": 0.015, "step": 43628 }, { "epoch": 13.42, "learning_rate": 5.161518463780846e-06, "loss": 0.0409, "step": 43629 }, { "epoch": 13.42, "learning_rate": 5.161082407774927e-06, "loss": 0.0193, "step": 43630 }, { "epoch": 13.42, "learning_rate": 5.160646363782737e-06, "loss": 0.0223, "step": 43631 }, { "epoch": 13.43, "learning_rate": 5.1602103318053586e-06, "loss": 0.0411, "step": 43632 }, { "epoch": 13.43, "learning_rate": 5.1597743118438725e-06, "loss": 0.0284, "step": 43633 }, { "epoch": 13.43, "learning_rate": 5.159338303899361e-06, "loss": 0.0219, "step": 43634 }, { "epoch": 13.43, "learning_rate": 5.158902307972904e-06, "loss": 0.0141, "step": 43635 }, { "epoch": 13.43, "learning_rate": 5.158466324065592e-06, "loss": 0.0113, "step": 43636 }, { "epoch": 13.43, "learning_rate": 5.1580303521785024e-06, "loss": 0.0317, "step": 43637 }, { "epoch": 13.43, "learning_rate": 5.1575943923127216e-06, "loss": 0.0171, "step": 43638 }, { "epoch": 13.43, "learning_rate": 5.157158444469326e-06, "loss": 0.0148, "step": 43639 }, { "epoch": 13.43, "learning_rate": 5.156722508649395e-06, "loss": 0.0201, "step": 43640 }, { "epoch": 13.43, "learning_rate": 5.15628658485402e-06, "loss": 0.017, "step": 43641 }, { "epoch": 13.43, "learning_rate": 5.155850673084279e-06, "loss": 0.0373, "step": 43642 }, { "epoch": 13.43, "learning_rate": 5.155414773341256e-06, "loss": 0.0195, "step": 43643 }, { "epoch": 13.43, "learning_rate": 5.154978885626033e-06, "loss": 0.0298, "step": 43644 }, { "epoch": 13.43, "learning_rate": 5.15454300993969e-06, "loss": 0.026, "step": 43645 }, { "epoch": 13.43, "learning_rate": 5.154107146283311e-06, "loss": 0.0263, "step": 43646 }, { "epoch": 13.43, "learning_rate": 5.153671294657978e-06, "loss": 0.0153, "step": 43647 }, { "epoch": 13.43, "learning_rate": 5.153235455064771e-06, "loss": 0.0215, "step": 43648 }, { "epoch": 13.43, "learning_rate": 5.1527996275047745e-06, "loss": 0.0184, "step": 43649 }, { "epoch": 13.43, "learning_rate": 5.152363811979066e-06, "loss": 0.0833, "step": 43650 }, { "epoch": 13.43, "learning_rate": 5.15192800848874e-06, "loss": 0.0074, "step": 43651 }, { "epoch": 13.43, "learning_rate": 5.1514922170348645e-06, "loss": 0.0143, "step": 43652 }, { "epoch": 13.43, "learning_rate": 5.151056437618527e-06, "loss": 0.0345, "step": 43653 }, { "epoch": 13.43, "learning_rate": 5.150620670240809e-06, "loss": 0.0299, "step": 43654 }, { "epoch": 13.43, "learning_rate": 5.150184914902789e-06, "loss": 0.0127, "step": 43655 }, { "epoch": 13.43, "learning_rate": 5.1497491716055555e-06, "loss": 0.0205, "step": 43656 }, { "epoch": 13.43, "learning_rate": 5.1493134403501884e-06, "loss": 0.0247, "step": 43657 }, { "epoch": 13.43, "learning_rate": 5.148877721137768e-06, "loss": 0.0161, "step": 43658 }, { "epoch": 13.43, "learning_rate": 5.1484420139693805e-06, "loss": 0.0123, "step": 43659 }, { "epoch": 13.43, "learning_rate": 5.148006318846095e-06, "loss": 0.0296, "step": 43660 }, { "epoch": 13.43, "learning_rate": 5.147570635769006e-06, "loss": 0.0245, "step": 43661 }, { "epoch": 13.43, "learning_rate": 5.147134964739191e-06, "loss": 0.0154, "step": 43662 }, { "epoch": 13.43, "learning_rate": 5.146699305757732e-06, "loss": 0.0091, "step": 43663 }, { "epoch": 13.43, "learning_rate": 5.14626365882571e-06, "loss": 0.0514, "step": 43664 }, { "epoch": 13.44, "learning_rate": 5.145828023944206e-06, "loss": 0.0172, "step": 43665 }, { "epoch": 13.44, "learning_rate": 5.1453924011143035e-06, "loss": 0.0281, "step": 43666 }, { "epoch": 13.44, "learning_rate": 5.144956790337081e-06, "loss": 0.0118, "step": 43667 }, { "epoch": 13.44, "learning_rate": 5.144521191613624e-06, "loss": 0.0296, "step": 43668 }, { "epoch": 13.44, "learning_rate": 5.14408560494501e-06, "loss": 0.0189, "step": 43669 }, { "epoch": 13.44, "learning_rate": 5.143650030332321e-06, "loss": 0.019, "step": 43670 }, { "epoch": 13.44, "learning_rate": 5.143214467776641e-06, "loss": 0.0167, "step": 43671 }, { "epoch": 13.44, "learning_rate": 5.1427789172790565e-06, "loss": 0.0272, "step": 43672 }, { "epoch": 13.44, "learning_rate": 5.142343378840637e-06, "loss": 0.0162, "step": 43673 }, { "epoch": 13.44, "learning_rate": 5.141907852462469e-06, "loss": 0.0139, "step": 43674 }, { "epoch": 13.44, "learning_rate": 5.141472338145631e-06, "loss": 0.0236, "step": 43675 }, { "epoch": 13.44, "learning_rate": 5.141036835891211e-06, "loss": 0.0307, "step": 43676 }, { "epoch": 13.44, "learning_rate": 5.140601345700287e-06, "loss": 0.0198, "step": 43677 }, { "epoch": 13.44, "learning_rate": 5.14016586757394e-06, "loss": 0.0117, "step": 43678 }, { "epoch": 13.44, "learning_rate": 5.13973040151325e-06, "loss": 0.0274, "step": 43679 }, { "epoch": 13.44, "learning_rate": 5.1392949475193e-06, "loss": 0.0269, "step": 43680 }, { "epoch": 13.44, "learning_rate": 5.13885950559317e-06, "loss": 0.0251, "step": 43681 }, { "epoch": 13.44, "learning_rate": 5.13842407573594e-06, "loss": 0.0217, "step": 43682 }, { "epoch": 13.44, "learning_rate": 5.1379886579486934e-06, "loss": 0.0124, "step": 43683 }, { "epoch": 13.44, "learning_rate": 5.137553252232511e-06, "loss": 0.0173, "step": 43684 }, { "epoch": 13.44, "learning_rate": 5.137117858588472e-06, "loss": 0.0182, "step": 43685 }, { "epoch": 13.44, "learning_rate": 5.136682477017658e-06, "loss": 0.0282, "step": 43686 }, { "epoch": 13.44, "learning_rate": 5.136247107521151e-06, "loss": 0.0119, "step": 43687 }, { "epoch": 13.44, "learning_rate": 5.1358117501000326e-06, "loss": 0.0191, "step": 43688 }, { "epoch": 13.44, "learning_rate": 5.135376404755381e-06, "loss": 0.0253, "step": 43689 }, { "epoch": 13.44, "learning_rate": 5.134941071488273e-06, "loss": 0.0132, "step": 43690 }, { "epoch": 13.44, "learning_rate": 5.134505750299801e-06, "loss": 0.0118, "step": 43691 }, { "epoch": 13.44, "learning_rate": 5.134070441191038e-06, "loss": 0.0133, "step": 43692 }, { "epoch": 13.44, "learning_rate": 5.1336351441630715e-06, "loss": 0.0135, "step": 43693 }, { "epoch": 13.44, "learning_rate": 5.1331998592169716e-06, "loss": 0.0217, "step": 43694 }, { "epoch": 13.44, "learning_rate": 5.132764586353821e-06, "loss": 0.0338, "step": 43695 }, { "epoch": 13.44, "learning_rate": 5.132329325574707e-06, "loss": 0.0343, "step": 43696 }, { "epoch": 13.45, "learning_rate": 5.131894076880709e-06, "loss": 0.0299, "step": 43697 }, { "epoch": 13.45, "learning_rate": 5.131458840272905e-06, "loss": 0.0182, "step": 43698 }, { "epoch": 13.45, "learning_rate": 5.131023615752375e-06, "loss": 0.0144, "step": 43699 }, { "epoch": 13.45, "learning_rate": 5.130588403320201e-06, "loss": 0.0352, "step": 43700 }, { "epoch": 13.45, "learning_rate": 5.130153202977464e-06, "loss": 0.0322, "step": 43701 }, { "epoch": 13.45, "learning_rate": 5.129718014725242e-06, "loss": 0.033, "step": 43702 }, { "epoch": 13.45, "learning_rate": 5.129282838564619e-06, "loss": 0.0166, "step": 43703 }, { "epoch": 13.45, "learning_rate": 5.128847674496673e-06, "loss": 0.0043, "step": 43704 }, { "epoch": 13.45, "learning_rate": 5.1284125225224815e-06, "loss": 0.0264, "step": 43705 }, { "epoch": 13.45, "learning_rate": 5.127977382643137e-06, "loss": 0.0195, "step": 43706 }, { "epoch": 13.45, "learning_rate": 5.127542254859705e-06, "loss": 0.0216, "step": 43707 }, { "epoch": 13.45, "learning_rate": 5.127107139173273e-06, "loss": 0.011, "step": 43708 }, { "epoch": 13.45, "learning_rate": 5.126672035584921e-06, "loss": 0.0411, "step": 43709 }, { "epoch": 13.45, "learning_rate": 5.1262369440957235e-06, "loss": 0.0198, "step": 43710 }, { "epoch": 13.45, "learning_rate": 5.125801864706771e-06, "loss": 0.0214, "step": 43711 }, { "epoch": 13.45, "learning_rate": 5.125366797419138e-06, "loss": 0.0202, "step": 43712 }, { "epoch": 13.45, "learning_rate": 5.124931742233905e-06, "loss": 0.0321, "step": 43713 }, { "epoch": 13.45, "learning_rate": 5.124496699152155e-06, "loss": 0.0246, "step": 43714 }, { "epoch": 13.45, "learning_rate": 5.124061668174958e-06, "loss": 0.0403, "step": 43715 }, { "epoch": 13.45, "learning_rate": 5.123626649303406e-06, "loss": 0.0211, "step": 43716 }, { "epoch": 13.45, "learning_rate": 5.123191642538574e-06, "loss": 0.0229, "step": 43717 }, { "epoch": 13.45, "learning_rate": 5.122756647881543e-06, "loss": 0.0206, "step": 43718 }, { "epoch": 13.45, "learning_rate": 5.1223216653333915e-06, "loss": 0.0202, "step": 43719 }, { "epoch": 13.45, "learning_rate": 5.121886694895201e-06, "loss": 0.0338, "step": 43720 }, { "epoch": 13.45, "learning_rate": 5.12145173656805e-06, "loss": 0.018, "step": 43721 }, { "epoch": 13.45, "learning_rate": 5.1210167903530195e-06, "loss": 0.0114, "step": 43722 }, { "epoch": 13.45, "learning_rate": 5.1205818562511884e-06, "loss": 0.0221, "step": 43723 }, { "epoch": 13.45, "learning_rate": 5.120146934263638e-06, "loss": 0.0094, "step": 43724 }, { "epoch": 13.45, "learning_rate": 5.1197120243914434e-06, "loss": 0.0343, "step": 43725 }, { "epoch": 13.45, "learning_rate": 5.119277126635691e-06, "loss": 0.0231, "step": 43726 }, { "epoch": 13.45, "learning_rate": 5.1188422409974625e-06, "loss": 0.02, "step": 43727 }, { "epoch": 13.45, "learning_rate": 5.118407367477829e-06, "loss": 0.0202, "step": 43728 }, { "epoch": 13.45, "learning_rate": 5.117972506077873e-06, "loss": 0.0186, "step": 43729 }, { "epoch": 13.46, "learning_rate": 5.117537656798672e-06, "loss": 0.0199, "step": 43730 }, { "epoch": 13.46, "learning_rate": 5.117102819641312e-06, "loss": 0.0109, "step": 43731 }, { "epoch": 13.46, "learning_rate": 5.11666799460687e-06, "loss": 0.022, "step": 43732 }, { "epoch": 13.46, "learning_rate": 5.116233181696424e-06, "loss": 0.0213, "step": 43733 }, { "epoch": 13.46, "learning_rate": 5.115798380911056e-06, "loss": 0.0251, "step": 43734 }, { "epoch": 13.46, "learning_rate": 5.115363592251843e-06, "loss": 0.0184, "step": 43735 }, { "epoch": 13.46, "learning_rate": 5.114928815719864e-06, "loss": 0.0249, "step": 43736 }, { "epoch": 13.46, "learning_rate": 5.114494051316201e-06, "loss": 0.0237, "step": 43737 }, { "epoch": 13.46, "learning_rate": 5.114059299041932e-06, "loss": 0.0185, "step": 43738 }, { "epoch": 13.46, "learning_rate": 5.113624558898137e-06, "loss": 0.0202, "step": 43739 }, { "epoch": 13.46, "learning_rate": 5.113189830885892e-06, "loss": 0.0197, "step": 43740 }, { "epoch": 13.46, "learning_rate": 5.1127551150062845e-06, "loss": 0.0332, "step": 43741 }, { "epoch": 13.46, "learning_rate": 5.112320411260385e-06, "loss": 0.0252, "step": 43742 }, { "epoch": 13.46, "learning_rate": 5.1118857196492775e-06, "loss": 0.0257, "step": 43743 }, { "epoch": 13.46, "learning_rate": 5.11145104017404e-06, "loss": 0.0326, "step": 43744 }, { "epoch": 13.46, "learning_rate": 5.111016372835747e-06, "loss": 0.028, "step": 43745 }, { "epoch": 13.46, "learning_rate": 5.110581717635487e-06, "loss": 0.014, "step": 43746 }, { "epoch": 13.46, "learning_rate": 5.110147074574334e-06, "loss": 0.0126, "step": 43747 }, { "epoch": 13.46, "learning_rate": 5.109712443653371e-06, "loss": 0.0128, "step": 43748 }, { "epoch": 13.46, "learning_rate": 5.109277824873671e-06, "loss": 0.0377, "step": 43749 }, { "epoch": 13.46, "learning_rate": 5.10884321823631e-06, "loss": 0.0219, "step": 43750 }, { "epoch": 13.46, "learning_rate": 5.108408623742378e-06, "loss": 0.0197, "step": 43751 }, { "epoch": 13.46, "learning_rate": 5.1079740413929464e-06, "loss": 0.0303, "step": 43752 }, { "epoch": 13.46, "learning_rate": 5.107539471189099e-06, "loss": 0.0251, "step": 43753 }, { "epoch": 13.46, "learning_rate": 5.10710491313191e-06, "loss": 0.0229, "step": 43754 }, { "epoch": 13.46, "learning_rate": 5.1066703672224625e-06, "loss": 0.0215, "step": 43755 }, { "epoch": 13.46, "learning_rate": 5.106235833461832e-06, "loss": 0.029, "step": 43756 }, { "epoch": 13.46, "learning_rate": 5.105801311851099e-06, "loss": 0.023, "step": 43757 }, { "epoch": 13.46, "learning_rate": 5.105366802391343e-06, "loss": 0.0207, "step": 43758 }, { "epoch": 13.46, "learning_rate": 5.104932305083641e-06, "loss": 0.0211, "step": 43759 }, { "epoch": 13.46, "learning_rate": 5.104497819929067e-06, "loss": 0.0164, "step": 43760 }, { "epoch": 13.46, "learning_rate": 5.10406334692871e-06, "loss": 0.0233, "step": 43761 }, { "epoch": 13.47, "learning_rate": 5.103628886083649e-06, "loss": 0.0115, "step": 43762 }, { "epoch": 13.47, "learning_rate": 5.103194437394952e-06, "loss": 0.0209, "step": 43763 }, { "epoch": 13.47, "learning_rate": 5.102760000863704e-06, "loss": 0.0188, "step": 43764 }, { "epoch": 13.47, "learning_rate": 5.1023255764909774e-06, "loss": 0.0221, "step": 43765 }, { "epoch": 13.47, "learning_rate": 5.1018911642778615e-06, "loss": 0.0171, "step": 43766 }, { "epoch": 13.47, "learning_rate": 5.101456764225427e-06, "loss": 0.0139, "step": 43767 }, { "epoch": 13.47, "learning_rate": 5.1010223763347565e-06, "loss": 0.0197, "step": 43768 }, { "epoch": 13.47, "learning_rate": 5.100588000606932e-06, "loss": 0.0187, "step": 43769 }, { "epoch": 13.47, "learning_rate": 5.100153637043017e-06, "loss": 0.0341, "step": 43770 }, { "epoch": 13.47, "learning_rate": 5.099719285644103e-06, "loss": 0.0255, "step": 43771 }, { "epoch": 13.47, "learning_rate": 5.099284946411266e-06, "loss": 0.0173, "step": 43772 }, { "epoch": 13.47, "learning_rate": 5.098850619345584e-06, "loss": 0.0623, "step": 43773 }, { "epoch": 13.47, "learning_rate": 5.098416304448133e-06, "loss": 0.022, "step": 43774 }, { "epoch": 13.47, "learning_rate": 5.097982001719994e-06, "loss": 0.0342, "step": 43775 }, { "epoch": 13.47, "learning_rate": 5.097547711162243e-06, "loss": 0.0208, "step": 43776 }, { "epoch": 13.47, "learning_rate": 5.0971134327759596e-06, "loss": 0.0293, "step": 43777 }, { "epoch": 13.47, "learning_rate": 5.096679166562221e-06, "loss": 0.0176, "step": 43778 }, { "epoch": 13.47, "learning_rate": 5.0962449125221084e-06, "loss": 0.0086, "step": 43779 }, { "epoch": 13.47, "learning_rate": 5.095810670656693e-06, "loss": 0.0209, "step": 43780 }, { "epoch": 13.47, "learning_rate": 5.095376440967061e-06, "loss": 0.0222, "step": 43781 }, { "epoch": 13.47, "learning_rate": 5.094942223454288e-06, "loss": 0.0155, "step": 43782 }, { "epoch": 13.47, "learning_rate": 5.094508018119454e-06, "loss": 0.0234, "step": 43783 }, { "epoch": 13.47, "learning_rate": 5.09407382496363e-06, "loss": 0.0217, "step": 43784 }, { "epoch": 13.47, "learning_rate": 5.093639643987895e-06, "loss": 0.0175, "step": 43785 }, { "epoch": 13.47, "learning_rate": 5.093205475193334e-06, "loss": 0.0174, "step": 43786 }, { "epoch": 13.47, "learning_rate": 5.09277131858102e-06, "loss": 0.0211, "step": 43787 }, { "epoch": 13.47, "learning_rate": 5.092337174152032e-06, "loss": 0.0206, "step": 43788 }, { "epoch": 13.47, "learning_rate": 5.091903041907449e-06, "loss": 0.0216, "step": 43789 }, { "epoch": 13.47, "learning_rate": 5.091468921848346e-06, "loss": 0.0231, "step": 43790 }, { "epoch": 13.47, "learning_rate": 5.091034813975804e-06, "loss": 0.0177, "step": 43791 }, { "epoch": 13.47, "learning_rate": 5.090600718290898e-06, "loss": 0.0348, "step": 43792 }, { "epoch": 13.47, "learning_rate": 5.0901666347947065e-06, "loss": 0.0109, "step": 43793 }, { "epoch": 13.47, "learning_rate": 5.089732563488308e-06, "loss": 0.0261, "step": 43794 }, { "epoch": 13.48, "learning_rate": 5.089298504372777e-06, "loss": 0.0141, "step": 43795 }, { "epoch": 13.48, "learning_rate": 5.088864457449201e-06, "loss": 0.0136, "step": 43796 }, { "epoch": 13.48, "learning_rate": 5.088430422718646e-06, "loss": 0.0226, "step": 43797 }, { "epoch": 13.48, "learning_rate": 5.087996400182193e-06, "loss": 0.027, "step": 43798 }, { "epoch": 13.48, "learning_rate": 5.087562389840923e-06, "loss": 0.0088, "step": 43799 }, { "epoch": 13.48, "learning_rate": 5.087128391695908e-06, "loss": 0.0372, "step": 43800 }, { "epoch": 13.48, "learning_rate": 5.086694405748231e-06, "loss": 0.0114, "step": 43801 }, { "epoch": 13.48, "learning_rate": 5.086260431998967e-06, "loss": 0.0221, "step": 43802 }, { "epoch": 13.48, "learning_rate": 5.0858264704491944e-06, "loss": 0.0255, "step": 43803 }, { "epoch": 13.48, "learning_rate": 5.085392521099993e-06, "loss": 0.0173, "step": 43804 }, { "epoch": 13.48, "learning_rate": 5.084958583952431e-06, "loss": 0.0149, "step": 43805 }, { "epoch": 13.48, "learning_rate": 5.084524659007594e-06, "loss": 0.0162, "step": 43806 }, { "epoch": 13.48, "learning_rate": 5.0840907462665564e-06, "loss": 0.0254, "step": 43807 }, { "epoch": 13.48, "learning_rate": 5.083656845730398e-06, "loss": 0.021, "step": 43808 }, { "epoch": 13.48, "learning_rate": 5.0832229574001935e-06, "loss": 0.0301, "step": 43809 }, { "epoch": 13.48, "learning_rate": 5.082789081277021e-06, "loss": 0.0183, "step": 43810 }, { "epoch": 13.48, "learning_rate": 5.082355217361957e-06, "loss": 0.0118, "step": 43811 }, { "epoch": 13.48, "learning_rate": 5.081921365656079e-06, "loss": 0.0317, "step": 43812 }, { "epoch": 13.48, "learning_rate": 5.0814875261604656e-06, "loss": 0.0143, "step": 43813 }, { "epoch": 13.48, "learning_rate": 5.081053698876193e-06, "loss": 0.0353, "step": 43814 }, { "epoch": 13.48, "learning_rate": 5.080619883804333e-06, "loss": 0.0318, "step": 43815 }, { "epoch": 13.48, "learning_rate": 5.080186080945971e-06, "loss": 0.0145, "step": 43816 }, { "epoch": 13.48, "learning_rate": 5.079752290302185e-06, "loss": 0.0349, "step": 43817 }, { "epoch": 13.48, "learning_rate": 5.079318511874045e-06, "loss": 0.0181, "step": 43818 }, { "epoch": 13.48, "learning_rate": 5.078884745662629e-06, "loss": 0.0271, "step": 43819 }, { "epoch": 13.48, "learning_rate": 5.078450991669012e-06, "loss": 0.0165, "step": 43820 }, { "epoch": 13.48, "learning_rate": 5.0780172498942785e-06, "loss": 0.0277, "step": 43821 }, { "epoch": 13.48, "learning_rate": 5.077583520339502e-06, "loss": 0.0207, "step": 43822 }, { "epoch": 13.48, "learning_rate": 5.077149803005757e-06, "loss": 0.0218, "step": 43823 }, { "epoch": 13.48, "learning_rate": 5.076716097894123e-06, "loss": 0.0095, "step": 43824 }, { "epoch": 13.48, "learning_rate": 5.076282405005675e-06, "loss": 0.0169, "step": 43825 }, { "epoch": 13.48, "learning_rate": 5.07584872434149e-06, "loss": 0.0242, "step": 43826 }, { "epoch": 13.49, "learning_rate": 5.075415055902646e-06, "loss": 0.0147, "step": 43827 }, { "epoch": 13.49, "learning_rate": 5.074981399690219e-06, "loss": 0.0327, "step": 43828 }, { "epoch": 13.49, "learning_rate": 5.0745477557052845e-06, "loss": 0.0147, "step": 43829 }, { "epoch": 13.49, "learning_rate": 5.074114123948921e-06, "loss": 0.0294, "step": 43830 }, { "epoch": 13.49, "learning_rate": 5.073680504422203e-06, "loss": 0.0164, "step": 43831 }, { "epoch": 13.49, "learning_rate": 5.07324689712621e-06, "loss": 0.0175, "step": 43832 }, { "epoch": 13.49, "learning_rate": 5.072813302062015e-06, "loss": 0.0218, "step": 43833 }, { "epoch": 13.49, "learning_rate": 5.072379719230697e-06, "loss": 0.0149, "step": 43834 }, { "epoch": 13.49, "learning_rate": 5.071946148633328e-06, "loss": 0.0399, "step": 43835 }, { "epoch": 13.49, "learning_rate": 5.071512590270992e-06, "loss": 0.0161, "step": 43836 }, { "epoch": 13.49, "learning_rate": 5.07107904414476e-06, "loss": 0.0155, "step": 43837 }, { "epoch": 13.49, "learning_rate": 5.070645510255716e-06, "loss": 0.0205, "step": 43838 }, { "epoch": 13.49, "learning_rate": 5.070211988604925e-06, "loss": 0.0247, "step": 43839 }, { "epoch": 13.49, "learning_rate": 5.069778479193465e-06, "loss": 0.0192, "step": 43840 }, { "epoch": 13.49, "learning_rate": 5.069344982022421e-06, "loss": 0.0237, "step": 43841 }, { "epoch": 13.49, "learning_rate": 5.068911497092863e-06, "loss": 0.0099, "step": 43842 }, { "epoch": 13.49, "learning_rate": 5.068478024405868e-06, "loss": 0.0143, "step": 43843 }, { "epoch": 13.49, "learning_rate": 5.068044563962513e-06, "loss": 0.0167, "step": 43844 }, { "epoch": 13.49, "learning_rate": 5.067611115763874e-06, "loss": 0.0199, "step": 43845 }, { "epoch": 13.49, "learning_rate": 5.067177679811026e-06, "loss": 0.0139, "step": 43846 }, { "epoch": 13.49, "learning_rate": 5.066744256105047e-06, "loss": 0.0261, "step": 43847 }, { "epoch": 13.49, "learning_rate": 5.066310844647012e-06, "loss": 0.0093, "step": 43848 }, { "epoch": 13.49, "learning_rate": 5.065877445437997e-06, "loss": 0.0379, "step": 43849 }, { "epoch": 13.49, "learning_rate": 5.065444058479074e-06, "loss": 0.0251, "step": 43850 }, { "epoch": 13.49, "learning_rate": 5.065010683771331e-06, "loss": 0.0172, "step": 43851 }, { "epoch": 13.49, "learning_rate": 5.064577321315832e-06, "loss": 0.028, "step": 43852 }, { "epoch": 13.49, "learning_rate": 5.064143971113657e-06, "loss": 0.036, "step": 43853 }, { "epoch": 13.49, "learning_rate": 5.0637106331658815e-06, "loss": 0.0209, "step": 43854 }, { "epoch": 13.49, "learning_rate": 5.063277307473579e-06, "loss": 0.0304, "step": 43855 }, { "epoch": 13.49, "learning_rate": 5.062843994037831e-06, "loss": 0.0106, "step": 43856 }, { "epoch": 13.49, "learning_rate": 5.062410692859711e-06, "loss": 0.0129, "step": 43857 }, { "epoch": 13.49, "learning_rate": 5.061977403940292e-06, "loss": 0.015, "step": 43858 }, { "epoch": 13.49, "learning_rate": 5.061544127280658e-06, "loss": 0.0325, "step": 43859 }, { "epoch": 13.5, "learning_rate": 5.061110862881871e-06, "loss": 0.0225, "step": 43860 }, { "epoch": 13.5, "learning_rate": 5.0606776107450176e-06, "loss": 0.0164, "step": 43861 }, { "epoch": 13.5, "learning_rate": 5.0602443708711675e-06, "loss": 0.0508, "step": 43862 }, { "epoch": 13.5, "learning_rate": 5.059811143261402e-06, "loss": 0.0258, "step": 43863 }, { "epoch": 13.5, "learning_rate": 5.059377927916793e-06, "loss": 0.0393, "step": 43864 }, { "epoch": 13.5, "learning_rate": 5.058944724838417e-06, "loss": 0.019, "step": 43865 }, { "epoch": 13.5, "learning_rate": 5.058511534027349e-06, "loss": 0.0142, "step": 43866 }, { "epoch": 13.5, "learning_rate": 5.058078355484663e-06, "loss": 0.0224, "step": 43867 }, { "epoch": 13.5, "learning_rate": 5.057645189211437e-06, "loss": 0.0155, "step": 43868 }, { "epoch": 13.5, "learning_rate": 5.057212035208747e-06, "loss": 0.0111, "step": 43869 }, { "epoch": 13.5, "learning_rate": 5.056778893477662e-06, "loss": 0.0194, "step": 43870 }, { "epoch": 13.5, "learning_rate": 5.0563457640192655e-06, "loss": 0.013, "step": 43871 }, { "epoch": 13.5, "learning_rate": 5.0559126468346354e-06, "loss": 0.0278, "step": 43872 }, { "epoch": 13.5, "learning_rate": 5.055479541924837e-06, "loss": 0.011, "step": 43873 }, { "epoch": 13.5, "learning_rate": 5.0550464492909475e-06, "loss": 0.0147, "step": 43874 }, { "epoch": 13.5, "learning_rate": 5.054613368934044e-06, "loss": 0.0279, "step": 43875 }, { "epoch": 13.5, "learning_rate": 5.054180300855205e-06, "loss": 0.0127, "step": 43876 }, { "epoch": 13.5, "learning_rate": 5.0537472450555024e-06, "loss": 0.0288, "step": 43877 }, { "epoch": 13.5, "learning_rate": 5.0533142015360115e-06, "loss": 0.0151, "step": 43878 }, { "epoch": 13.5, "learning_rate": 5.052881170297809e-06, "loss": 0.0109, "step": 43879 }, { "epoch": 13.5, "learning_rate": 5.0524481513419675e-06, "loss": 0.0197, "step": 43880 }, { "epoch": 13.5, "learning_rate": 5.052015144669564e-06, "loss": 0.0209, "step": 43881 }, { "epoch": 13.5, "learning_rate": 5.051582150281673e-06, "loss": 0.0171, "step": 43882 }, { "epoch": 13.5, "learning_rate": 5.05114916817937e-06, "loss": 0.0221, "step": 43883 }, { "epoch": 13.5, "learning_rate": 5.05071619836373e-06, "loss": 0.0223, "step": 43884 }, { "epoch": 13.5, "learning_rate": 5.050283240835826e-06, "loss": 0.0235, "step": 43885 }, { "epoch": 13.5, "learning_rate": 5.049850295596734e-06, "loss": 0.0228, "step": 43886 }, { "epoch": 13.5, "learning_rate": 5.049417362647531e-06, "loss": 0.0187, "step": 43887 }, { "epoch": 13.5, "learning_rate": 5.048984441989289e-06, "loss": 0.0269, "step": 43888 }, { "epoch": 13.5, "learning_rate": 5.048551533623084e-06, "loss": 0.0247, "step": 43889 }, { "epoch": 13.5, "learning_rate": 5.048118637549987e-06, "loss": 0.0097, "step": 43890 }, { "epoch": 13.5, "learning_rate": 5.04768575377108e-06, "loss": 0.0671, "step": 43891 }, { "epoch": 13.51, "learning_rate": 5.047252882287435e-06, "loss": 0.0306, "step": 43892 }, { "epoch": 13.51, "learning_rate": 5.046820023100129e-06, "loss": 0.0795, "step": 43893 }, { "epoch": 13.51, "learning_rate": 5.046387176210229e-06, "loss": 0.0469, "step": 43894 }, { "epoch": 13.51, "learning_rate": 5.045954341618811e-06, "loss": 0.0282, "step": 43895 }, { "epoch": 13.51, "learning_rate": 5.045521519326957e-06, "loss": 0.0261, "step": 43896 }, { "epoch": 13.51, "learning_rate": 5.045088709335737e-06, "loss": 0.0285, "step": 43897 }, { "epoch": 13.51, "learning_rate": 5.044655911646226e-06, "loss": 0.0279, "step": 43898 }, { "epoch": 13.51, "learning_rate": 5.044223126259498e-06, "loss": 0.0321, "step": 43899 }, { "epoch": 13.51, "learning_rate": 5.043790353176628e-06, "loss": 0.0207, "step": 43900 }, { "epoch": 13.51, "learning_rate": 5.043357592398691e-06, "loss": 0.0241, "step": 43901 }, { "epoch": 13.51, "learning_rate": 5.042924843926761e-06, "loss": 0.0225, "step": 43902 }, { "epoch": 13.51, "learning_rate": 5.042492107761912e-06, "loss": 0.0204, "step": 43903 }, { "epoch": 13.51, "learning_rate": 5.0420593839052175e-06, "loss": 0.0253, "step": 43904 }, { "epoch": 13.51, "learning_rate": 5.0416266723577515e-06, "loss": 0.0128, "step": 43905 }, { "epoch": 13.51, "learning_rate": 5.041193973120595e-06, "loss": 0.0264, "step": 43906 }, { "epoch": 13.51, "learning_rate": 5.040761286194814e-06, "loss": 0.0261, "step": 43907 }, { "epoch": 13.51, "learning_rate": 5.040328611581487e-06, "loss": 0.0168, "step": 43908 }, { "epoch": 13.51, "learning_rate": 5.0398959492816844e-06, "loss": 0.0267, "step": 43909 }, { "epoch": 13.51, "learning_rate": 5.039463299296481e-06, "loss": 0.0232, "step": 43910 }, { "epoch": 13.51, "learning_rate": 5.039030661626956e-06, "loss": 0.0123, "step": 43911 }, { "epoch": 13.51, "learning_rate": 5.038598036274181e-06, "loss": 0.0299, "step": 43912 }, { "epoch": 13.51, "learning_rate": 5.03816542323923e-06, "loss": 0.0216, "step": 43913 }, { "epoch": 13.51, "learning_rate": 5.037732822523178e-06, "loss": 0.0145, "step": 43914 }, { "epoch": 13.51, "learning_rate": 5.037300234127091e-06, "loss": 0.0154, "step": 43915 }, { "epoch": 13.51, "learning_rate": 5.036867658052056e-06, "loss": 0.0237, "step": 43916 }, { "epoch": 13.51, "learning_rate": 5.036435094299138e-06, "loss": 0.0285, "step": 43917 }, { "epoch": 13.51, "learning_rate": 5.036002542869414e-06, "loss": 0.0176, "step": 43918 }, { "epoch": 13.51, "learning_rate": 5.035570003763958e-06, "loss": 0.0172, "step": 43919 }, { "epoch": 13.51, "learning_rate": 5.035137476983843e-06, "loss": 0.0163, "step": 43920 }, { "epoch": 13.51, "learning_rate": 5.034704962530142e-06, "loss": 0.04, "step": 43921 }, { "epoch": 13.51, "learning_rate": 5.034272460403932e-06, "loss": 0.0333, "step": 43922 }, { "epoch": 13.51, "learning_rate": 5.033839970606283e-06, "loss": 0.0132, "step": 43923 }, { "epoch": 13.51, "learning_rate": 5.033407493138271e-06, "loss": 0.0461, "step": 43924 }, { "epoch": 13.52, "learning_rate": 5.032975028000966e-06, "loss": 0.0255, "step": 43925 }, { "epoch": 13.52, "learning_rate": 5.032542575195449e-06, "loss": 0.0047, "step": 43926 }, { "epoch": 13.52, "learning_rate": 5.032110134722793e-06, "loss": 0.0212, "step": 43927 }, { "epoch": 13.52, "learning_rate": 5.031677706584064e-06, "loss": 0.0055, "step": 43928 }, { "epoch": 13.52, "learning_rate": 5.03124529078034e-06, "loss": 0.0163, "step": 43929 }, { "epoch": 13.52, "learning_rate": 5.0308128873126916e-06, "loss": 0.0238, "step": 43930 }, { "epoch": 13.52, "learning_rate": 5.030380496182198e-06, "loss": 0.0157, "step": 43931 }, { "epoch": 13.52, "learning_rate": 5.02994811738993e-06, "loss": 0.0146, "step": 43932 }, { "epoch": 13.52, "learning_rate": 5.02951575093696e-06, "loss": 0.0304, "step": 43933 }, { "epoch": 13.52, "learning_rate": 5.029083396824363e-06, "loss": 0.0203, "step": 43934 }, { "epoch": 13.52, "learning_rate": 5.028651055053213e-06, "loss": 0.0185, "step": 43935 }, { "epoch": 13.52, "learning_rate": 5.028218725624581e-06, "loss": 0.0268, "step": 43936 }, { "epoch": 13.52, "learning_rate": 5.027786408539542e-06, "loss": 0.0185, "step": 43937 }, { "epoch": 13.52, "learning_rate": 5.027354103799169e-06, "loss": 0.0359, "step": 43938 }, { "epoch": 13.52, "learning_rate": 5.026921811404532e-06, "loss": 0.0167, "step": 43939 }, { "epoch": 13.52, "learning_rate": 5.02648953135671e-06, "loss": 0.0247, "step": 43940 }, { "epoch": 13.52, "learning_rate": 5.026057263656778e-06, "loss": 0.04, "step": 43941 }, { "epoch": 13.52, "learning_rate": 5.025625008305801e-06, "loss": 0.021, "step": 43942 }, { "epoch": 13.52, "learning_rate": 5.025192765304856e-06, "loss": 0.0229, "step": 43943 }, { "epoch": 13.52, "learning_rate": 5.024760534655012e-06, "loss": 0.0377, "step": 43944 }, { "epoch": 13.52, "learning_rate": 5.02432831635735e-06, "loss": 0.0144, "step": 43945 }, { "epoch": 13.52, "learning_rate": 5.02389611041294e-06, "loss": 0.0179, "step": 43946 }, { "epoch": 13.52, "learning_rate": 5.023463916822855e-06, "loss": 0.0225, "step": 43947 }, { "epoch": 13.52, "learning_rate": 5.023031735588171e-06, "loss": 0.0223, "step": 43948 }, { "epoch": 13.52, "learning_rate": 5.0225995667099495e-06, "loss": 0.0067, "step": 43949 }, { "epoch": 13.52, "learning_rate": 5.022167410189274e-06, "loss": 0.0196, "step": 43950 }, { "epoch": 13.52, "learning_rate": 5.021735266027216e-06, "loss": 0.0214, "step": 43951 }, { "epoch": 13.52, "learning_rate": 5.021303134224848e-06, "loss": 0.0177, "step": 43952 }, { "epoch": 13.52, "learning_rate": 5.020871014783241e-06, "loss": 0.0132, "step": 43953 }, { "epoch": 13.52, "learning_rate": 5.0204389077034685e-06, "loss": 0.0343, "step": 43954 }, { "epoch": 13.52, "learning_rate": 5.020006812986605e-06, "loss": 0.0331, "step": 43955 }, { "epoch": 13.52, "learning_rate": 5.019574730633722e-06, "loss": 0.0248, "step": 43956 }, { "epoch": 13.53, "learning_rate": 5.0191426606458915e-06, "loss": 0.0206, "step": 43957 }, { "epoch": 13.53, "learning_rate": 5.018710603024187e-06, "loss": 0.0335, "step": 43958 }, { "epoch": 13.53, "learning_rate": 5.018278557769678e-06, "loss": 0.0245, "step": 43959 }, { "epoch": 13.53, "learning_rate": 5.017846524883443e-06, "loss": 0.0191, "step": 43960 }, { "epoch": 13.53, "learning_rate": 5.017414504366553e-06, "loss": 0.0135, "step": 43961 }, { "epoch": 13.53, "learning_rate": 5.016982496220083e-06, "loss": 0.0404, "step": 43962 }, { "epoch": 13.53, "learning_rate": 5.016550500445099e-06, "loss": 0.0131, "step": 43963 }, { "epoch": 13.53, "learning_rate": 5.016118517042672e-06, "loss": 0.0281, "step": 43964 }, { "epoch": 13.53, "learning_rate": 5.015686546013882e-06, "loss": 0.0153, "step": 43965 }, { "epoch": 13.53, "learning_rate": 5.0152545873598e-06, "loss": 0.009, "step": 43966 }, { "epoch": 13.53, "learning_rate": 5.014822641081496e-06, "loss": 0.0235, "step": 43967 }, { "epoch": 13.53, "learning_rate": 5.0143907071800445e-06, "loss": 0.0128, "step": 43968 }, { "epoch": 13.53, "learning_rate": 5.013958785656516e-06, "loss": 0.084, "step": 43969 }, { "epoch": 13.53, "learning_rate": 5.013526876511985e-06, "loss": 0.0262, "step": 43970 }, { "epoch": 13.53, "learning_rate": 5.013094979747522e-06, "loss": 0.019, "step": 43971 }, { "epoch": 13.53, "learning_rate": 5.012663095364199e-06, "loss": 0.0224, "step": 43972 }, { "epoch": 13.53, "learning_rate": 5.012231223363089e-06, "loss": 0.0186, "step": 43973 }, { "epoch": 13.53, "learning_rate": 5.01179936374526e-06, "loss": 0.0075, "step": 43974 }, { "epoch": 13.53, "learning_rate": 5.011367516511796e-06, "loss": 0.0237, "step": 43975 }, { "epoch": 13.53, "learning_rate": 5.010935681663758e-06, "loss": 0.0102, "step": 43976 }, { "epoch": 13.53, "learning_rate": 5.010503859202223e-06, "loss": 0.0237, "step": 43977 }, { "epoch": 13.53, "learning_rate": 5.0100720491282605e-06, "loss": 0.0292, "step": 43978 }, { "epoch": 13.53, "learning_rate": 5.00964025144294e-06, "loss": 0.0204, "step": 43979 }, { "epoch": 13.53, "learning_rate": 5.009208466147342e-06, "loss": 0.0157, "step": 43980 }, { "epoch": 13.53, "learning_rate": 5.008776693242533e-06, "loss": 0.0281, "step": 43981 }, { "epoch": 13.53, "learning_rate": 5.00834493272959e-06, "loss": 0.0193, "step": 43982 }, { "epoch": 13.53, "learning_rate": 5.007913184609578e-06, "loss": 0.0265, "step": 43983 }, { "epoch": 13.53, "learning_rate": 5.007481448883567e-06, "loss": 0.0425, "step": 43984 }, { "epoch": 13.53, "learning_rate": 5.007049725552638e-06, "loss": 0.0184, "step": 43985 }, { "epoch": 13.53, "learning_rate": 5.006618014617856e-06, "loss": 0.0271, "step": 43986 }, { "epoch": 13.53, "learning_rate": 5.006186316080298e-06, "loss": 0.0268, "step": 43987 }, { "epoch": 13.53, "learning_rate": 5.005754629941033e-06, "loss": 0.0297, "step": 43988 }, { "epoch": 13.53, "learning_rate": 5.005322956201132e-06, "loss": 0.0134, "step": 43989 }, { "epoch": 13.54, "learning_rate": 5.004891294861666e-06, "loss": 0.0337, "step": 43990 }, { "epoch": 13.54, "learning_rate": 5.004459645923712e-06, "loss": 0.0125, "step": 43991 }, { "epoch": 13.54, "learning_rate": 5.004028009388335e-06, "loss": 0.0191, "step": 43992 }, { "epoch": 13.54, "learning_rate": 5.003596385256611e-06, "loss": 0.0266, "step": 43993 }, { "epoch": 13.54, "learning_rate": 5.0031647735296055e-06, "loss": 0.0341, "step": 43994 }, { "epoch": 13.54, "learning_rate": 5.0027331742083996e-06, "loss": 0.0221, "step": 43995 }, { "epoch": 13.54, "learning_rate": 5.002301587294063e-06, "loss": 0.023, "step": 43996 }, { "epoch": 13.54, "learning_rate": 5.0018700127876606e-06, "loss": 0.0151, "step": 43997 }, { "epoch": 13.54, "learning_rate": 5.001438450690268e-06, "loss": 0.0309, "step": 43998 }, { "epoch": 13.54, "learning_rate": 5.0010069010029515e-06, "loss": 0.0156, "step": 43999 }, { "epoch": 13.54, "learning_rate": 5.000575363726792e-06, "loss": 0.0175, "step": 44000 }, { "epoch": 13.54, "learning_rate": 5.0001438388628544e-06, "loss": 0.029, "step": 44001 }, { "epoch": 13.54, "learning_rate": 4.9997123264122134e-06, "loss": 0.0465, "step": 44002 }, { "epoch": 13.54, "learning_rate": 4.9992808263759414e-06, "loss": 0.0271, "step": 44003 }, { "epoch": 13.54, "learning_rate": 4.998849338755101e-06, "loss": 0.0155, "step": 44004 }, { "epoch": 13.54, "learning_rate": 4.998417863550771e-06, "loss": 0.0136, "step": 44005 }, { "epoch": 13.54, "learning_rate": 4.997986400764022e-06, "loss": 0.0207, "step": 44006 }, { "epoch": 13.54, "learning_rate": 4.997554950395924e-06, "loss": 0.0308, "step": 44007 }, { "epoch": 13.54, "learning_rate": 4.997123512447549e-06, "loss": 0.0255, "step": 44008 }, { "epoch": 13.54, "learning_rate": 4.996692086919968e-06, "loss": 0.0178, "step": 44009 }, { "epoch": 13.54, "learning_rate": 4.99626067381425e-06, "loss": 0.0126, "step": 44010 }, { "epoch": 13.54, "learning_rate": 4.995829273131469e-06, "loss": 0.0187, "step": 44011 }, { "epoch": 13.54, "learning_rate": 4.9953978848726955e-06, "loss": 0.0148, "step": 44012 }, { "epoch": 13.54, "learning_rate": 4.994966509038999e-06, "loss": 0.0183, "step": 44013 }, { "epoch": 13.54, "learning_rate": 4.994535145631447e-06, "loss": 0.0233, "step": 44014 }, { "epoch": 13.54, "learning_rate": 4.9941037946511195e-06, "loss": 0.0253, "step": 44015 }, { "epoch": 13.54, "learning_rate": 4.993672456099083e-06, "loss": 0.0184, "step": 44016 }, { "epoch": 13.54, "learning_rate": 4.993241129976411e-06, "loss": 0.0115, "step": 44017 }, { "epoch": 13.54, "learning_rate": 4.992809816284168e-06, "loss": 0.0243, "step": 44018 }, { "epoch": 13.54, "learning_rate": 4.992378515023425e-06, "loss": 0.0172, "step": 44019 }, { "epoch": 13.54, "learning_rate": 4.991947226195259e-06, "loss": 0.0149, "step": 44020 }, { "epoch": 13.54, "learning_rate": 4.991515949800738e-06, "loss": 0.0085, "step": 44021 }, { "epoch": 13.55, "learning_rate": 4.991084685840933e-06, "loss": 0.0197, "step": 44022 }, { "epoch": 13.55, "learning_rate": 4.990653434316915e-06, "loss": 0.0259, "step": 44023 }, { "epoch": 13.55, "learning_rate": 4.990222195229754e-06, "loss": 0.0133, "step": 44024 }, { "epoch": 13.55, "learning_rate": 4.9897909685805194e-06, "loss": 0.0106, "step": 44025 }, { "epoch": 13.55, "learning_rate": 4.989359754370284e-06, "loss": 0.0269, "step": 44026 }, { "epoch": 13.55, "learning_rate": 4.988928552600118e-06, "loss": 0.0211, "step": 44027 }, { "epoch": 13.55, "learning_rate": 4.98849736327109e-06, "loss": 0.026, "step": 44028 }, { "epoch": 13.55, "learning_rate": 4.98806618638427e-06, "loss": 0.0241, "step": 44029 }, { "epoch": 13.55, "learning_rate": 4.987635021940736e-06, "loss": 0.022, "step": 44030 }, { "epoch": 13.55, "learning_rate": 4.987203869941551e-06, "loss": 0.021, "step": 44031 }, { "epoch": 13.55, "learning_rate": 4.986772730387786e-06, "loss": 0.0304, "step": 44032 }, { "epoch": 13.55, "learning_rate": 4.986341603280513e-06, "loss": 0.0277, "step": 44033 }, { "epoch": 13.55, "learning_rate": 4.9859104886208e-06, "loss": 0.0258, "step": 44034 }, { "epoch": 13.55, "learning_rate": 4.985479386409722e-06, "loss": 0.0167, "step": 44035 }, { "epoch": 13.55, "learning_rate": 4.985048296648346e-06, "loss": 0.0114, "step": 44036 }, { "epoch": 13.55, "learning_rate": 4.984617219337744e-06, "loss": 0.0104, "step": 44037 }, { "epoch": 13.55, "learning_rate": 4.9841861544789895e-06, "loss": 0.0252, "step": 44038 }, { "epoch": 13.55, "learning_rate": 4.98375510207314e-06, "loss": 0.0199, "step": 44039 }, { "epoch": 13.55, "learning_rate": 4.9833240621212795e-06, "loss": 0.0266, "step": 44040 }, { "epoch": 13.55, "learning_rate": 4.982893034624472e-06, "loss": 0.0181, "step": 44041 }, { "epoch": 13.55, "learning_rate": 4.982462019583788e-06, "loss": 0.0098, "step": 44042 }, { "epoch": 13.55, "learning_rate": 4.982031017000299e-06, "loss": 0.0313, "step": 44043 }, { "epoch": 13.55, "learning_rate": 4.9816000268750746e-06, "loss": 0.0233, "step": 44044 }, { "epoch": 13.55, "learning_rate": 4.981169049209182e-06, "loss": 0.0142, "step": 44045 }, { "epoch": 13.55, "learning_rate": 4.980738084003696e-06, "loss": 0.0354, "step": 44046 }, { "epoch": 13.55, "learning_rate": 4.980307131259684e-06, "loss": 0.0266, "step": 44047 }, { "epoch": 13.55, "learning_rate": 4.979876190978215e-06, "loss": 0.0217, "step": 44048 }, { "epoch": 13.55, "learning_rate": 4.979445263160357e-06, "loss": 0.0281, "step": 44049 }, { "epoch": 13.55, "learning_rate": 4.9790143478071866e-06, "loss": 0.0174, "step": 44050 }, { "epoch": 13.55, "learning_rate": 4.978583444919773e-06, "loss": 0.0175, "step": 44051 }, { "epoch": 13.55, "learning_rate": 4.978152554499181e-06, "loss": 0.0263, "step": 44052 }, { "epoch": 13.55, "learning_rate": 4.977721676546481e-06, "loss": 0.0198, "step": 44053 }, { "epoch": 13.55, "learning_rate": 4.97729081106274e-06, "loss": 0.0228, "step": 44054 }, { "epoch": 13.56, "learning_rate": 4.976859958049036e-06, "loss": 0.0443, "step": 44055 }, { "epoch": 13.56, "learning_rate": 4.976429117506435e-06, "loss": 0.0221, "step": 44056 }, { "epoch": 13.56, "learning_rate": 4.975998289436006e-06, "loss": 0.0324, "step": 44057 }, { "epoch": 13.56, "learning_rate": 4.975567473838818e-06, "loss": 0.0241, "step": 44058 }, { "epoch": 13.56, "learning_rate": 4.975136670715943e-06, "loss": 0.0188, "step": 44059 }, { "epoch": 13.56, "learning_rate": 4.974705880068448e-06, "loss": 0.0291, "step": 44060 }, { "epoch": 13.56, "learning_rate": 4.974275101897404e-06, "loss": 0.0352, "step": 44061 }, { "epoch": 13.56, "learning_rate": 4.973844336203879e-06, "loss": 0.0352, "step": 44062 }, { "epoch": 13.56, "learning_rate": 4.973413582988944e-06, "loss": 0.0313, "step": 44063 }, { "epoch": 13.56, "learning_rate": 4.9729828422536685e-06, "loss": 0.0419, "step": 44064 }, { "epoch": 13.56, "learning_rate": 4.972552113999121e-06, "loss": 0.0262, "step": 44065 }, { "epoch": 13.56, "learning_rate": 4.972121398226371e-06, "loss": 0.0137, "step": 44066 }, { "epoch": 13.56, "learning_rate": 4.971690694936489e-06, "loss": 0.0114, "step": 44067 }, { "epoch": 13.56, "learning_rate": 4.9712600041305425e-06, "loss": 0.0205, "step": 44068 }, { "epoch": 13.56, "learning_rate": 4.970829325809598e-06, "loss": 0.036, "step": 44069 }, { "epoch": 13.56, "learning_rate": 4.970398659974732e-06, "loss": 0.026, "step": 44070 }, { "epoch": 13.56, "learning_rate": 4.96996800662701e-06, "loss": 0.0236, "step": 44071 }, { "epoch": 13.56, "learning_rate": 4.969537365767507e-06, "loss": 0.0099, "step": 44072 }, { "epoch": 13.56, "learning_rate": 4.969106737397281e-06, "loss": 0.0156, "step": 44073 }, { "epoch": 13.56, "learning_rate": 4.968676121517404e-06, "loss": 0.0152, "step": 44074 }, { "epoch": 13.56, "learning_rate": 4.96824551812895e-06, "loss": 0.0125, "step": 44075 }, { "epoch": 13.56, "learning_rate": 4.967814927232988e-06, "loss": 0.0138, "step": 44076 }, { "epoch": 13.56, "learning_rate": 4.967384348830583e-06, "loss": 0.017, "step": 44077 }, { "epoch": 13.56, "learning_rate": 4.966953782922806e-06, "loss": 0.0191, "step": 44078 }, { "epoch": 13.56, "learning_rate": 4.966523229510727e-06, "loss": 0.0181, "step": 44079 }, { "epoch": 13.56, "learning_rate": 4.9660926885954145e-06, "loss": 0.0204, "step": 44080 }, { "epoch": 13.56, "learning_rate": 4.965662160177935e-06, "loss": 0.0118, "step": 44081 }, { "epoch": 13.56, "learning_rate": 4.965231644259359e-06, "loss": 0.0241, "step": 44082 }, { "epoch": 13.56, "learning_rate": 4.964801140840757e-06, "loss": 0.0164, "step": 44083 }, { "epoch": 13.56, "learning_rate": 4.964370649923191e-06, "loss": 0.0241, "step": 44084 }, { "epoch": 13.56, "learning_rate": 4.963940171507744e-06, "loss": 0.0148, "step": 44085 }, { "epoch": 13.56, "learning_rate": 4.963509705595472e-06, "loss": 0.0216, "step": 44086 }, { "epoch": 13.57, "learning_rate": 4.963079252187447e-06, "loss": 0.0193, "step": 44087 }, { "epoch": 13.57, "learning_rate": 4.9626488112847384e-06, "loss": 0.0191, "step": 44088 }, { "epoch": 13.57, "learning_rate": 4.962218382888411e-06, "loss": 0.0164, "step": 44089 }, { "epoch": 13.57, "learning_rate": 4.9617879669995415e-06, "loss": 0.0196, "step": 44090 }, { "epoch": 13.57, "learning_rate": 4.961357563619194e-06, "loss": 0.0116, "step": 44091 }, { "epoch": 13.57, "learning_rate": 4.960927172748437e-06, "loss": 0.0199, "step": 44092 }, { "epoch": 13.57, "learning_rate": 4.960496794388343e-06, "loss": 0.0184, "step": 44093 }, { "epoch": 13.57, "learning_rate": 4.960066428539969e-06, "loss": 0.0115, "step": 44094 }, { "epoch": 13.57, "learning_rate": 4.959636075204396e-06, "loss": 0.0303, "step": 44095 }, { "epoch": 13.57, "learning_rate": 4.9592057343826875e-06, "loss": 0.0217, "step": 44096 }, { "epoch": 13.57, "learning_rate": 4.958775406075912e-06, "loss": 0.0191, "step": 44097 }, { "epoch": 13.57, "learning_rate": 4.958345090285138e-06, "loss": 0.0213, "step": 44098 }, { "epoch": 13.57, "learning_rate": 4.957914787011434e-06, "loss": 0.0157, "step": 44099 }, { "epoch": 13.57, "learning_rate": 4.957484496255869e-06, "loss": 0.0198, "step": 44100 }, { "epoch": 13.57, "learning_rate": 4.957054218019511e-06, "loss": 0.0219, "step": 44101 }, { "epoch": 13.57, "learning_rate": 4.956623952303426e-06, "loss": 0.0175, "step": 44102 }, { "epoch": 13.57, "learning_rate": 4.956193699108685e-06, "loss": 0.0252, "step": 44103 }, { "epoch": 13.57, "learning_rate": 4.955763458436351e-06, "loss": 0.0205, "step": 44104 }, { "epoch": 13.57, "learning_rate": 4.9553332302875e-06, "loss": 0.0241, "step": 44105 }, { "epoch": 13.57, "learning_rate": 4.954903014663201e-06, "loss": 0.0239, "step": 44106 }, { "epoch": 13.57, "learning_rate": 4.954472811564515e-06, "loss": 0.0232, "step": 44107 }, { "epoch": 13.57, "learning_rate": 4.95404262099251e-06, "loss": 0.0177, "step": 44108 }, { "epoch": 13.57, "learning_rate": 4.953612442948255e-06, "loss": 0.0195, "step": 44109 }, { "epoch": 13.57, "learning_rate": 4.953182277432824e-06, "loss": 0.0133, "step": 44110 }, { "epoch": 13.57, "learning_rate": 4.952752124447281e-06, "loss": 0.024, "step": 44111 }, { "epoch": 13.57, "learning_rate": 4.952321983992693e-06, "loss": 0.013, "step": 44112 }, { "epoch": 13.57, "learning_rate": 4.951891856070128e-06, "loss": 0.0239, "step": 44113 }, { "epoch": 13.57, "learning_rate": 4.951461740680655e-06, "loss": 0.0159, "step": 44114 }, { "epoch": 13.57, "learning_rate": 4.951031637825342e-06, "loss": 0.0213, "step": 44115 }, { "epoch": 13.57, "learning_rate": 4.950601547505256e-06, "loss": 0.033, "step": 44116 }, { "epoch": 13.57, "learning_rate": 4.950171469721466e-06, "loss": 0.0199, "step": 44117 }, { "epoch": 13.57, "learning_rate": 4.949741404475037e-06, "loss": 0.0348, "step": 44118 }, { "epoch": 13.57, "learning_rate": 4.9493113517670396e-06, "loss": 0.0464, "step": 44119 }, { "epoch": 13.58, "learning_rate": 4.94888131159854e-06, "loss": 0.0253, "step": 44120 }, { "epoch": 13.58, "learning_rate": 4.948451283970608e-06, "loss": 0.0267, "step": 44121 }, { "epoch": 13.58, "learning_rate": 4.948021268884308e-06, "loss": 0.0128, "step": 44122 }, { "epoch": 13.58, "learning_rate": 4.947591266340709e-06, "loss": 0.0131, "step": 44123 }, { "epoch": 13.58, "learning_rate": 4.947161276340876e-06, "loss": 0.0388, "step": 44124 }, { "epoch": 13.58, "learning_rate": 4.946731298885883e-06, "loss": 0.0312, "step": 44125 }, { "epoch": 13.58, "learning_rate": 4.9463013339767925e-06, "loss": 0.0221, "step": 44126 }, { "epoch": 13.58, "learning_rate": 4.945871381614679e-06, "loss": 0.0103, "step": 44127 }, { "epoch": 13.58, "learning_rate": 4.9454414418006e-06, "loss": 0.0155, "step": 44128 }, { "epoch": 13.58, "learning_rate": 4.945011514535622e-06, "loss": 0.0326, "step": 44129 }, { "epoch": 13.58, "learning_rate": 4.944581599820823e-06, "loss": 0.0186, "step": 44130 }, { "epoch": 13.58, "learning_rate": 4.944151697657266e-06, "loss": 0.0209, "step": 44131 }, { "epoch": 13.58, "learning_rate": 4.943721808046015e-06, "loss": 0.0204, "step": 44132 }, { "epoch": 13.58, "learning_rate": 4.943291930988141e-06, "loss": 0.0195, "step": 44133 }, { "epoch": 13.58, "learning_rate": 4.9428620664847095e-06, "loss": 0.0248, "step": 44134 }, { "epoch": 13.58, "learning_rate": 4.942432214536787e-06, "loss": 0.0644, "step": 44135 }, { "epoch": 13.58, "learning_rate": 4.942002375145444e-06, "loss": 0.0183, "step": 44136 }, { "epoch": 13.58, "learning_rate": 4.941572548311744e-06, "loss": 0.0256, "step": 44137 }, { "epoch": 13.58, "learning_rate": 4.9411427340367555e-06, "loss": 0.0159, "step": 44138 }, { "epoch": 13.58, "learning_rate": 4.940712932321543e-06, "loss": 0.0261, "step": 44139 }, { "epoch": 13.58, "learning_rate": 4.9402831431671834e-06, "loss": 0.0214, "step": 44140 }, { "epoch": 13.58, "learning_rate": 4.9398533665747326e-06, "loss": 0.0206, "step": 44141 }, { "epoch": 13.58, "learning_rate": 4.939423602545263e-06, "loss": 0.0124, "step": 44142 }, { "epoch": 13.58, "learning_rate": 4.93899385107984e-06, "loss": 0.0246, "step": 44143 }, { "epoch": 13.58, "learning_rate": 4.938564112179526e-06, "loss": 0.0257, "step": 44144 }, { "epoch": 13.58, "learning_rate": 4.938134385845396e-06, "loss": 0.0313, "step": 44145 }, { "epoch": 13.58, "learning_rate": 4.937704672078516e-06, "loss": 0.0289, "step": 44146 }, { "epoch": 13.58, "learning_rate": 4.937274970879948e-06, "loss": 0.0503, "step": 44147 }, { "epoch": 13.58, "learning_rate": 4.936845282250767e-06, "loss": 0.0165, "step": 44148 }, { "epoch": 13.58, "learning_rate": 4.9364156061920265e-06, "loss": 0.0276, "step": 44149 }, { "epoch": 13.58, "learning_rate": 4.935985942704804e-06, "loss": 0.0251, "step": 44150 }, { "epoch": 13.58, "learning_rate": 4.935556291790164e-06, "loss": 0.0257, "step": 44151 }, { "epoch": 13.59, "learning_rate": 4.935126653449172e-06, "loss": 0.0115, "step": 44152 }, { "epoch": 13.59, "learning_rate": 4.934697027682894e-06, "loss": 0.0183, "step": 44153 }, { "epoch": 13.59, "learning_rate": 4.9342674144924e-06, "loss": 0.0191, "step": 44154 }, { "epoch": 13.59, "learning_rate": 4.9338378138787526e-06, "loss": 0.0333, "step": 44155 }, { "epoch": 13.59, "learning_rate": 4.93340822584302e-06, "loss": 0.0133, "step": 44156 }, { "epoch": 13.59, "learning_rate": 4.93297865038627e-06, "loss": 0.0226, "step": 44157 }, { "epoch": 13.59, "learning_rate": 4.932549087509567e-06, "loss": 0.0224, "step": 44158 }, { "epoch": 13.59, "learning_rate": 4.9321195372139756e-06, "loss": 0.0305, "step": 44159 }, { "epoch": 13.59, "learning_rate": 4.931689999500567e-06, "loss": 0.042, "step": 44160 }, { "epoch": 13.59, "learning_rate": 4.931260474370412e-06, "loss": 0.035, "step": 44161 }, { "epoch": 13.59, "learning_rate": 4.930830961824566e-06, "loss": 0.0205, "step": 44162 }, { "epoch": 13.59, "learning_rate": 4.930401461864099e-06, "loss": 0.0222, "step": 44163 }, { "epoch": 13.59, "learning_rate": 4.929971974490077e-06, "loss": 0.0264, "step": 44164 }, { "epoch": 13.59, "learning_rate": 4.929542499703569e-06, "loss": 0.0263, "step": 44165 }, { "epoch": 13.59, "learning_rate": 4.929113037505642e-06, "loss": 0.026, "step": 44166 }, { "epoch": 13.59, "learning_rate": 4.928683587897359e-06, "loss": 0.0171, "step": 44167 }, { "epoch": 13.59, "learning_rate": 4.9282541508797875e-06, "loss": 0.0159, "step": 44168 }, { "epoch": 13.59, "learning_rate": 4.927824726453995e-06, "loss": 0.0159, "step": 44169 }, { "epoch": 13.59, "learning_rate": 4.927395314621045e-06, "loss": 0.0197, "step": 44170 }, { "epoch": 13.59, "learning_rate": 4.926965915382005e-06, "loss": 0.0165, "step": 44171 }, { "epoch": 13.59, "learning_rate": 4.926536528737941e-06, "loss": 0.0195, "step": 44172 }, { "epoch": 13.59, "learning_rate": 4.92610715468992e-06, "loss": 0.0058, "step": 44173 }, { "epoch": 13.59, "learning_rate": 4.925677793239002e-06, "loss": 0.0183, "step": 44174 }, { "epoch": 13.59, "learning_rate": 4.9252484443862655e-06, "loss": 0.0159, "step": 44175 }, { "epoch": 13.59, "learning_rate": 4.924819108132767e-06, "loss": 0.0279, "step": 44176 }, { "epoch": 13.59, "learning_rate": 4.924389784479573e-06, "loss": 0.015, "step": 44177 }, { "epoch": 13.59, "learning_rate": 4.9239604734277515e-06, "loss": 0.0203, "step": 44178 }, { "epoch": 13.59, "learning_rate": 4.923531174978363e-06, "loss": 0.0184, "step": 44179 }, { "epoch": 13.59, "learning_rate": 4.923101889132483e-06, "loss": 0.0227, "step": 44180 }, { "epoch": 13.59, "learning_rate": 4.922672615891171e-06, "loss": 0.033, "step": 44181 }, { "epoch": 13.59, "learning_rate": 4.922243355255498e-06, "loss": 0.0209, "step": 44182 }, { "epoch": 13.59, "learning_rate": 4.921814107226523e-06, "loss": 0.0339, "step": 44183 }, { "epoch": 13.59, "learning_rate": 4.9213848718053105e-06, "loss": 0.019, "step": 44184 }, { "epoch": 13.6, "learning_rate": 4.920955648992934e-06, "loss": 0.0309, "step": 44185 }, { "epoch": 13.6, "learning_rate": 4.920526438790455e-06, "loss": 0.0122, "step": 44186 }, { "epoch": 13.6, "learning_rate": 4.92009724119894e-06, "loss": 0.0161, "step": 44187 }, { "epoch": 13.6, "learning_rate": 4.919668056219453e-06, "loss": 0.0351, "step": 44188 }, { "epoch": 13.6, "learning_rate": 4.919238883853061e-06, "loss": 0.0094, "step": 44189 }, { "epoch": 13.6, "learning_rate": 4.91880972410083e-06, "loss": 0.0358, "step": 44190 }, { "epoch": 13.6, "learning_rate": 4.918380576963824e-06, "loss": 0.0176, "step": 44191 }, { "epoch": 13.6, "learning_rate": 4.91795144244311e-06, "loss": 0.0248, "step": 44192 }, { "epoch": 13.6, "learning_rate": 4.91752232053975e-06, "loss": 0.0174, "step": 44193 }, { "epoch": 13.6, "learning_rate": 4.917093211254811e-06, "loss": 0.0223, "step": 44194 }, { "epoch": 13.6, "learning_rate": 4.916664114589361e-06, "loss": 0.0178, "step": 44195 }, { "epoch": 13.6, "learning_rate": 4.916235030544468e-06, "loss": 0.0207, "step": 44196 }, { "epoch": 13.6, "learning_rate": 4.915805959121189e-06, "loss": 0.0193, "step": 44197 }, { "epoch": 13.6, "learning_rate": 4.915376900320594e-06, "loss": 0.0242, "step": 44198 }, { "epoch": 13.6, "learning_rate": 4.914947854143743e-06, "loss": 0.012, "step": 44199 }, { "epoch": 13.6, "learning_rate": 4.9145188205917095e-06, "loss": 0.0143, "step": 44200 }, { "epoch": 13.6, "learning_rate": 4.9140897996655545e-06, "loss": 0.0265, "step": 44201 }, { "epoch": 13.6, "learning_rate": 4.913660791366343e-06, "loss": 0.0129, "step": 44202 }, { "epoch": 13.6, "learning_rate": 4.913231795695145e-06, "loss": 0.0194, "step": 44203 }, { "epoch": 13.6, "learning_rate": 4.9128028126530135e-06, "loss": 0.0172, "step": 44204 }, { "epoch": 13.6, "learning_rate": 4.912373842241025e-06, "loss": 0.0232, "step": 44205 }, { "epoch": 13.6, "learning_rate": 4.9119448844602406e-06, "loss": 0.0206, "step": 44206 }, { "epoch": 13.6, "learning_rate": 4.911515939311724e-06, "loss": 0.0149, "step": 44207 }, { "epoch": 13.6, "learning_rate": 4.911087006796543e-06, "loss": 0.0358, "step": 44208 }, { "epoch": 13.6, "learning_rate": 4.91065808691576e-06, "loss": 0.0201, "step": 44209 }, { "epoch": 13.6, "learning_rate": 4.910229179670443e-06, "loss": 0.0208, "step": 44210 }, { "epoch": 13.6, "learning_rate": 4.909800285061653e-06, "loss": 0.027, "step": 44211 }, { "epoch": 13.6, "learning_rate": 4.909371403090457e-06, "loss": 0.0161, "step": 44212 }, { "epoch": 13.6, "learning_rate": 4.90894253375792e-06, "loss": 0.0214, "step": 44213 }, { "epoch": 13.6, "learning_rate": 4.908513677065101e-06, "loss": 0.0196, "step": 44214 }, { "epoch": 13.6, "learning_rate": 4.908084833013075e-06, "loss": 0.0348, "step": 44215 }, { "epoch": 13.6, "learning_rate": 4.907656001602901e-06, "loss": 0.021, "step": 44216 }, { "epoch": 13.61, "learning_rate": 4.907227182835649e-06, "loss": 0.0202, "step": 44217 }, { "epoch": 13.61, "learning_rate": 4.9067983767123736e-06, "loss": 0.0379, "step": 44218 }, { "epoch": 13.61, "learning_rate": 4.9063695832341415e-06, "loss": 0.0143, "step": 44219 }, { "epoch": 13.61, "learning_rate": 4.905940802402025e-06, "loss": 0.0174, "step": 44220 }, { "epoch": 13.61, "learning_rate": 4.905512034217084e-06, "loss": 0.0075, "step": 44221 }, { "epoch": 13.61, "learning_rate": 4.905083278680383e-06, "loss": 0.014, "step": 44222 }, { "epoch": 13.61, "learning_rate": 4.904654535792986e-06, "loss": 0.0362, "step": 44223 }, { "epoch": 13.61, "learning_rate": 4.9042258055559584e-06, "loss": 0.0107, "step": 44224 }, { "epoch": 13.61, "learning_rate": 4.903797087970366e-06, "loss": 0.0215, "step": 44225 }, { "epoch": 13.61, "learning_rate": 4.903368383037269e-06, "loss": 0.0203, "step": 44226 }, { "epoch": 13.61, "learning_rate": 4.902939690757736e-06, "loss": 0.0324, "step": 44227 }, { "epoch": 13.61, "learning_rate": 4.902511011132829e-06, "loss": 0.0116, "step": 44228 }, { "epoch": 13.61, "learning_rate": 4.902082344163609e-06, "loss": 0.02, "step": 44229 }, { "epoch": 13.61, "learning_rate": 4.901653689851152e-06, "loss": 0.0069, "step": 44230 }, { "epoch": 13.61, "learning_rate": 4.901225048196511e-06, "loss": 0.0253, "step": 44231 }, { "epoch": 13.61, "learning_rate": 4.900796419200753e-06, "loss": 0.0181, "step": 44232 }, { "epoch": 13.61, "learning_rate": 4.900367802864944e-06, "loss": 0.0247, "step": 44233 }, { "epoch": 13.61, "learning_rate": 4.899939199190141e-06, "loss": 0.0121, "step": 44234 }, { "epoch": 13.61, "learning_rate": 4.899510608177419e-06, "loss": 0.0204, "step": 44235 }, { "epoch": 13.61, "learning_rate": 4.899082029827838e-06, "loss": 0.0167, "step": 44236 }, { "epoch": 13.61, "learning_rate": 4.89865346414246e-06, "loss": 0.0222, "step": 44237 }, { "epoch": 13.61, "learning_rate": 4.898224911122354e-06, "loss": 0.0185, "step": 44238 }, { "epoch": 13.61, "learning_rate": 4.897796370768572e-06, "loss": 0.03, "step": 44239 }, { "epoch": 13.61, "learning_rate": 4.897367843082191e-06, "loss": 0.0257, "step": 44240 }, { "epoch": 13.61, "learning_rate": 4.89693932806427e-06, "loss": 0.0316, "step": 44241 }, { "epoch": 13.61, "learning_rate": 4.8965108257158735e-06, "loss": 0.0443, "step": 44242 }, { "epoch": 13.61, "learning_rate": 4.896082336038064e-06, "loss": 0.0177, "step": 44243 }, { "epoch": 13.61, "learning_rate": 4.895653859031906e-06, "loss": 0.0209, "step": 44244 }, { "epoch": 13.61, "learning_rate": 4.895225394698463e-06, "loss": 0.0165, "step": 44245 }, { "epoch": 13.61, "learning_rate": 4.8947969430388e-06, "loss": 0.0293, "step": 44246 }, { "epoch": 13.61, "learning_rate": 4.894368504053979e-06, "loss": 0.0111, "step": 44247 }, { "epoch": 13.61, "learning_rate": 4.893940077745065e-06, "loss": 0.0127, "step": 44248 }, { "epoch": 13.61, "learning_rate": 4.893511664113118e-06, "loss": 0.0179, "step": 44249 }, { "epoch": 13.62, "learning_rate": 4.893083263159207e-06, "loss": 0.0143, "step": 44250 }, { "epoch": 13.62, "learning_rate": 4.892654874884398e-06, "loss": 0.0152, "step": 44251 }, { "epoch": 13.62, "learning_rate": 4.892226499289747e-06, "loss": 0.0197, "step": 44252 }, { "epoch": 13.62, "learning_rate": 4.89179813637632e-06, "loss": 0.0385, "step": 44253 }, { "epoch": 13.62, "learning_rate": 4.891369786145178e-06, "loss": 0.0389, "step": 44254 }, { "epoch": 13.62, "learning_rate": 4.890941448597391e-06, "loss": 0.0225, "step": 44255 }, { "epoch": 13.62, "learning_rate": 4.890513123734019e-06, "loss": 0.0057, "step": 44256 }, { "epoch": 13.62, "learning_rate": 4.890084811556125e-06, "loss": 0.0503, "step": 44257 }, { "epoch": 13.62, "learning_rate": 4.889656512064773e-06, "loss": 0.0198, "step": 44258 }, { "epoch": 13.62, "learning_rate": 4.889228225261026e-06, "loss": 0.024, "step": 44259 }, { "epoch": 13.62, "learning_rate": 4.888799951145948e-06, "loss": 0.0303, "step": 44260 }, { "epoch": 13.62, "learning_rate": 4.888371689720601e-06, "loss": 0.0245, "step": 44261 }, { "epoch": 13.62, "learning_rate": 4.887943440986051e-06, "loss": 0.023, "step": 44262 }, { "epoch": 13.62, "learning_rate": 4.887515204943357e-06, "loss": 0.0165, "step": 44263 }, { "epoch": 13.62, "learning_rate": 4.887086981593585e-06, "loss": 0.0283, "step": 44264 }, { "epoch": 13.62, "learning_rate": 4.886658770937798e-06, "loss": 0.0366, "step": 44265 }, { "epoch": 13.62, "learning_rate": 4.886230572977059e-06, "loss": 0.0121, "step": 44266 }, { "epoch": 13.62, "learning_rate": 4.88580238771243e-06, "loss": 0.0253, "step": 44267 }, { "epoch": 13.62, "learning_rate": 4.885374215144976e-06, "loss": 0.0117, "step": 44268 }, { "epoch": 13.62, "learning_rate": 4.8849460552757544e-06, "loss": 0.0345, "step": 44269 }, { "epoch": 13.62, "learning_rate": 4.884517908105837e-06, "loss": 0.0158, "step": 44270 }, { "epoch": 13.62, "learning_rate": 4.884089773636282e-06, "loss": 0.0174, "step": 44271 }, { "epoch": 13.62, "learning_rate": 4.883661651868158e-06, "loss": 0.0226, "step": 44272 }, { "epoch": 13.62, "learning_rate": 4.883233542802518e-06, "loss": 0.0123, "step": 44273 }, { "epoch": 13.62, "learning_rate": 4.8828054464404255e-06, "loss": 0.0213, "step": 44274 }, { "epoch": 13.62, "learning_rate": 4.882377362782952e-06, "loss": 0.0372, "step": 44275 }, { "epoch": 13.62, "learning_rate": 4.881949291831155e-06, "loss": 0.0219, "step": 44276 }, { "epoch": 13.62, "learning_rate": 4.881521233586099e-06, "loss": 0.0288, "step": 44277 }, { "epoch": 13.62, "learning_rate": 4.881093188048846e-06, "loss": 0.032, "step": 44278 }, { "epoch": 13.62, "learning_rate": 4.880665155220458e-06, "loss": 0.036, "step": 44279 }, { "epoch": 13.62, "learning_rate": 4.880237135101998e-06, "loss": 0.0175, "step": 44280 }, { "epoch": 13.62, "learning_rate": 4.879809127694531e-06, "loss": 0.0173, "step": 44281 }, { "epoch": 13.63, "learning_rate": 4.879381132999115e-06, "loss": 0.0487, "step": 44282 }, { "epoch": 13.63, "learning_rate": 4.878953151016816e-06, "loss": 0.0362, "step": 44283 }, { "epoch": 13.63, "learning_rate": 4.878525181748692e-06, "loss": 0.0264, "step": 44284 }, { "epoch": 13.63, "learning_rate": 4.878097225195817e-06, "loss": 0.0122, "step": 44285 }, { "epoch": 13.63, "learning_rate": 4.877669281359242e-06, "loss": 0.0134, "step": 44286 }, { "epoch": 13.63, "learning_rate": 4.877241350240033e-06, "loss": 0.017, "step": 44287 }, { "epoch": 13.63, "learning_rate": 4.8768134318392535e-06, "loss": 0.0349, "step": 44288 }, { "epoch": 13.63, "learning_rate": 4.8763855261579605e-06, "loss": 0.029, "step": 44289 }, { "epoch": 13.63, "learning_rate": 4.875957633197224e-06, "loss": 0.0189, "step": 44290 }, { "epoch": 13.63, "learning_rate": 4.875529752958105e-06, "loss": 0.0262, "step": 44291 }, { "epoch": 13.63, "learning_rate": 4.875101885441663e-06, "loss": 0.0188, "step": 44292 }, { "epoch": 13.63, "learning_rate": 4.874674030648966e-06, "loss": 0.0189, "step": 44293 }, { "epoch": 13.63, "learning_rate": 4.874246188581063e-06, "loss": 0.0126, "step": 44294 }, { "epoch": 13.63, "learning_rate": 4.87381835923903e-06, "loss": 0.028, "step": 44295 }, { "epoch": 13.63, "learning_rate": 4.873390542623922e-06, "loss": 0.0175, "step": 44296 }, { "epoch": 13.63, "learning_rate": 4.872962738736805e-06, "loss": 0.0307, "step": 44297 }, { "epoch": 13.63, "learning_rate": 4.872534947578739e-06, "loss": 0.0142, "step": 44298 }, { "epoch": 13.63, "learning_rate": 4.872107169150786e-06, "loss": 0.0053, "step": 44299 }, { "epoch": 13.63, "learning_rate": 4.871679403454008e-06, "loss": 0.0191, "step": 44300 }, { "epoch": 13.63, "learning_rate": 4.871251650489469e-06, "loss": 0.0178, "step": 44301 }, { "epoch": 13.63, "learning_rate": 4.870823910258228e-06, "loss": 0.0176, "step": 44302 }, { "epoch": 13.63, "learning_rate": 4.870396182761349e-06, "loss": 0.0133, "step": 44303 }, { "epoch": 13.63, "learning_rate": 4.869968467999889e-06, "loss": 0.0252, "step": 44304 }, { "epoch": 13.63, "learning_rate": 4.86954076597492e-06, "loss": 0.0175, "step": 44305 }, { "epoch": 13.63, "learning_rate": 4.8691130766875e-06, "loss": 0.0194, "step": 44306 }, { "epoch": 13.63, "learning_rate": 4.8686854001386865e-06, "loss": 0.0101, "step": 44307 }, { "epoch": 13.63, "learning_rate": 4.868257736329543e-06, "loss": 0.0185, "step": 44308 }, { "epoch": 13.63, "learning_rate": 4.867830085261128e-06, "loss": 0.024, "step": 44309 }, { "epoch": 13.63, "learning_rate": 4.867402446934512e-06, "loss": 0.0186, "step": 44310 }, { "epoch": 13.63, "learning_rate": 4.866974821350751e-06, "loss": 0.0176, "step": 44311 }, { "epoch": 13.63, "learning_rate": 4.8665472085109075e-06, "loss": 0.0346, "step": 44312 }, { "epoch": 13.63, "learning_rate": 4.866119608416045e-06, "loss": 0.0216, "step": 44313 }, { "epoch": 13.63, "learning_rate": 4.865692021067222e-06, "loss": 0.0242, "step": 44314 }, { "epoch": 13.64, "learning_rate": 4.865264446465502e-06, "loss": 0.0181, "step": 44315 }, { "epoch": 13.64, "learning_rate": 4.8648368846119465e-06, "loss": 0.0143, "step": 44316 }, { "epoch": 13.64, "learning_rate": 4.864409335507616e-06, "loss": 0.021, "step": 44317 }, { "epoch": 13.64, "learning_rate": 4.863981799153573e-06, "loss": 0.0119, "step": 44318 }, { "epoch": 13.64, "learning_rate": 4.863554275550878e-06, "loss": 0.0142, "step": 44319 }, { "epoch": 13.64, "learning_rate": 4.863126764700594e-06, "loss": 0.0186, "step": 44320 }, { "epoch": 13.64, "learning_rate": 4.862699266603781e-06, "loss": 0.0293, "step": 44321 }, { "epoch": 13.64, "learning_rate": 4.8622717812615e-06, "loss": 0.0379, "step": 44322 }, { "epoch": 13.64, "learning_rate": 4.861844308674813e-06, "loss": 0.0265, "step": 44323 }, { "epoch": 13.64, "learning_rate": 4.861416848844779e-06, "loss": 0.0039, "step": 44324 }, { "epoch": 13.64, "learning_rate": 4.860989401772465e-06, "loss": 0.0234, "step": 44325 }, { "epoch": 13.64, "learning_rate": 4.860561967458928e-06, "loss": 0.0059, "step": 44326 }, { "epoch": 13.64, "learning_rate": 4.860134545905235e-06, "loss": 0.0235, "step": 44327 }, { "epoch": 13.64, "learning_rate": 4.859707137112437e-06, "loss": 0.0235, "step": 44328 }, { "epoch": 13.64, "learning_rate": 4.859279741081598e-06, "loss": 0.0213, "step": 44329 }, { "epoch": 13.64, "learning_rate": 4.858852357813784e-06, "loss": 0.0297, "step": 44330 }, { "epoch": 13.64, "learning_rate": 4.858424987310056e-06, "loss": 0.0376, "step": 44331 }, { "epoch": 13.64, "learning_rate": 4.857997629571471e-06, "loss": 0.0214, "step": 44332 }, { "epoch": 13.64, "learning_rate": 4.857570284599091e-06, "loss": 0.0154, "step": 44333 }, { "epoch": 13.64, "learning_rate": 4.857142952393978e-06, "loss": 0.0278, "step": 44334 }, { "epoch": 13.64, "learning_rate": 4.856715632957193e-06, "loss": 0.0219, "step": 44335 }, { "epoch": 13.64, "learning_rate": 4.856288326289796e-06, "loss": 0.0173, "step": 44336 }, { "epoch": 13.64, "learning_rate": 4.855861032392849e-06, "loss": 0.0265, "step": 44337 }, { "epoch": 13.64, "learning_rate": 4.855433751267408e-06, "loss": 0.0273, "step": 44338 }, { "epoch": 13.64, "learning_rate": 4.855006482914543e-06, "loss": 0.0236, "step": 44339 }, { "epoch": 13.64, "learning_rate": 4.854579227335312e-06, "loss": 0.0349, "step": 44340 }, { "epoch": 13.64, "learning_rate": 4.854151984530771e-06, "loss": 0.0184, "step": 44341 }, { "epoch": 13.64, "learning_rate": 4.853724754501983e-06, "loss": 0.0149, "step": 44342 }, { "epoch": 13.64, "learning_rate": 4.853297537250005e-06, "loss": 0.0188, "step": 44343 }, { "epoch": 13.64, "learning_rate": 4.852870332775905e-06, "loss": 0.0205, "step": 44344 }, { "epoch": 13.64, "learning_rate": 4.85244314108074e-06, "loss": 0.017, "step": 44345 }, { "epoch": 13.64, "learning_rate": 4.8520159621655725e-06, "loss": 0.0135, "step": 44346 }, { "epoch": 13.65, "learning_rate": 4.851588796031461e-06, "loss": 0.0195, "step": 44347 }, { "epoch": 13.65, "learning_rate": 4.851161642679466e-06, "loss": 0.0349, "step": 44348 }, { "epoch": 13.65, "learning_rate": 4.85073450211065e-06, "loss": 0.0132, "step": 44349 }, { "epoch": 13.65, "learning_rate": 4.850307374326071e-06, "loss": 0.0159, "step": 44350 }, { "epoch": 13.65, "learning_rate": 4.84988025932679e-06, "loss": 0.0318, "step": 44351 }, { "epoch": 13.65, "learning_rate": 4.8494531571138685e-06, "loss": 0.0238, "step": 44352 }, { "epoch": 13.65, "learning_rate": 4.849026067688363e-06, "loss": 0.0154, "step": 44353 }, { "epoch": 13.65, "learning_rate": 4.848598991051343e-06, "loss": 0.0183, "step": 44354 }, { "epoch": 13.65, "learning_rate": 4.8481719272038604e-06, "loss": 0.0297, "step": 44355 }, { "epoch": 13.65, "learning_rate": 4.847744876146978e-06, "loss": 0.0104, "step": 44356 }, { "epoch": 13.65, "learning_rate": 4.847317837881757e-06, "loss": 0.0144, "step": 44357 }, { "epoch": 13.65, "learning_rate": 4.846890812409251e-06, "loss": 0.017, "step": 44358 }, { "epoch": 13.65, "learning_rate": 4.84646379973053e-06, "loss": 0.0168, "step": 44359 }, { "epoch": 13.65, "learning_rate": 4.846036799846651e-06, "loss": 0.0357, "step": 44360 }, { "epoch": 13.65, "learning_rate": 4.845609812758676e-06, "loss": 0.0271, "step": 44361 }, { "epoch": 13.65, "learning_rate": 4.845182838467659e-06, "loss": 0.0374, "step": 44362 }, { "epoch": 13.65, "learning_rate": 4.84475587697466e-06, "loss": 0.0231, "step": 44363 }, { "epoch": 13.65, "learning_rate": 4.844328928280745e-06, "loss": 0.0146, "step": 44364 }, { "epoch": 13.65, "learning_rate": 4.843901992386972e-06, "loss": 0.0066, "step": 44365 }, { "epoch": 13.65, "learning_rate": 4.8434750692944e-06, "loss": 0.0214, "step": 44366 }, { "epoch": 13.65, "learning_rate": 4.84304815900409e-06, "loss": 0.0215, "step": 44367 }, { "epoch": 13.65, "learning_rate": 4.8426212615171e-06, "loss": 0.0495, "step": 44368 }, { "epoch": 13.65, "learning_rate": 4.842194376834492e-06, "loss": 0.0131, "step": 44369 }, { "epoch": 13.65, "learning_rate": 4.841767504957324e-06, "loss": 0.0103, "step": 44370 }, { "epoch": 13.65, "learning_rate": 4.841340645886658e-06, "loss": 0.0269, "step": 44371 }, { "epoch": 13.65, "learning_rate": 4.840913799623551e-06, "loss": 0.0145, "step": 44372 }, { "epoch": 13.65, "learning_rate": 4.840486966169061e-06, "loss": 0.0349, "step": 44373 }, { "epoch": 13.65, "learning_rate": 4.840060145524254e-06, "loss": 0.0255, "step": 44374 }, { "epoch": 13.65, "learning_rate": 4.839633337690192e-06, "loss": 0.0274, "step": 44375 }, { "epoch": 13.65, "learning_rate": 4.839206542667925e-06, "loss": 0.018, "step": 44376 }, { "epoch": 13.65, "learning_rate": 4.838779760458516e-06, "loss": 0.0259, "step": 44377 }, { "epoch": 13.65, "learning_rate": 4.838352991063022e-06, "loss": 0.0204, "step": 44378 }, { "epoch": 13.65, "learning_rate": 4.83792623448251e-06, "loss": 0.0215, "step": 44379 }, { "epoch": 13.66, "learning_rate": 4.837499490718035e-06, "loss": 0.0295, "step": 44380 }, { "epoch": 13.66, "learning_rate": 4.837072759770657e-06, "loss": 0.0176, "step": 44381 }, { "epoch": 13.66, "learning_rate": 4.8366460416414394e-06, "loss": 0.0168, "step": 44382 }, { "epoch": 13.66, "learning_rate": 4.836219336331431e-06, "loss": 0.0326, "step": 44383 }, { "epoch": 13.66, "learning_rate": 4.8357926438417e-06, "loss": 0.0151, "step": 44384 }, { "epoch": 13.66, "learning_rate": 4.835365964173306e-06, "loss": 0.0184, "step": 44385 }, { "epoch": 13.66, "learning_rate": 4.8349392973273035e-06, "loss": 0.0244, "step": 44386 }, { "epoch": 13.66, "learning_rate": 4.834512643304755e-06, "loss": 0.0347, "step": 44387 }, { "epoch": 13.66, "learning_rate": 4.834086002106719e-06, "loss": 0.01, "step": 44388 }, { "epoch": 13.66, "learning_rate": 4.833659373734256e-06, "loss": 0.0087, "step": 44389 }, { "epoch": 13.66, "learning_rate": 4.833232758188423e-06, "loss": 0.0088, "step": 44390 }, { "epoch": 13.66, "learning_rate": 4.83280615547028e-06, "loss": 0.0151, "step": 44391 }, { "epoch": 13.66, "learning_rate": 4.832379565580887e-06, "loss": 0.013, "step": 44392 }, { "epoch": 13.66, "learning_rate": 4.831952988521298e-06, "loss": 0.0218, "step": 44393 }, { "epoch": 13.66, "learning_rate": 4.83152642429258e-06, "loss": 0.0202, "step": 44394 }, { "epoch": 13.66, "learning_rate": 4.83109987289579e-06, "loss": 0.0157, "step": 44395 }, { "epoch": 13.66, "learning_rate": 4.830673334331986e-06, "loss": 0.0218, "step": 44396 }, { "epoch": 13.66, "learning_rate": 4.8302468086022255e-06, "loss": 0.0134, "step": 44397 }, { "epoch": 13.66, "learning_rate": 4.829820295707564e-06, "loss": 0.0201, "step": 44398 }, { "epoch": 13.66, "learning_rate": 4.829393795649069e-06, "loss": 0.0452, "step": 44399 }, { "epoch": 13.66, "learning_rate": 4.828967308427795e-06, "loss": 0.0186, "step": 44400 }, { "epoch": 13.66, "learning_rate": 4.828540834044801e-06, "loss": 0.0206, "step": 44401 }, { "epoch": 13.66, "learning_rate": 4.828114372501146e-06, "loss": 0.0143, "step": 44402 }, { "epoch": 13.66, "learning_rate": 4.827687923797889e-06, "loss": 0.0237, "step": 44403 }, { "epoch": 13.66, "learning_rate": 4.8272614879360875e-06, "loss": 0.0212, "step": 44404 }, { "epoch": 13.66, "learning_rate": 4.8268350649168024e-06, "loss": 0.0184, "step": 44405 }, { "epoch": 13.66, "learning_rate": 4.826408654741091e-06, "loss": 0.0304, "step": 44406 }, { "epoch": 13.66, "learning_rate": 4.825982257410011e-06, "loss": 0.0198, "step": 44407 }, { "epoch": 13.66, "learning_rate": 4.8255558729246195e-06, "loss": 0.0327, "step": 44408 }, { "epoch": 13.66, "learning_rate": 4.825129501285985e-06, "loss": 0.0111, "step": 44409 }, { "epoch": 13.66, "learning_rate": 4.8247031424951555e-06, "loss": 0.0174, "step": 44410 }, { "epoch": 13.66, "learning_rate": 4.824276796553192e-06, "loss": 0.024, "step": 44411 }, { "epoch": 13.67, "learning_rate": 4.823850463461155e-06, "loss": 0.0312, "step": 44412 }, { "epoch": 13.67, "learning_rate": 4.823424143220097e-06, "loss": 0.0258, "step": 44413 }, { "epoch": 13.67, "learning_rate": 4.822997835831086e-06, "loss": 0.0166, "step": 44414 }, { "epoch": 13.67, "learning_rate": 4.822571541295174e-06, "loss": 0.0276, "step": 44415 }, { "epoch": 13.67, "learning_rate": 4.822145259613423e-06, "loss": 0.0223, "step": 44416 }, { "epoch": 13.67, "learning_rate": 4.821718990786892e-06, "loss": 0.0206, "step": 44417 }, { "epoch": 13.67, "learning_rate": 4.82129273481663e-06, "loss": 0.0249, "step": 44418 }, { "epoch": 13.67, "learning_rate": 4.820866491703706e-06, "loss": 0.0096, "step": 44419 }, { "epoch": 13.67, "learning_rate": 4.820440261449173e-06, "loss": 0.0252, "step": 44420 }, { "epoch": 13.67, "learning_rate": 4.8200140440540925e-06, "loss": 0.0174, "step": 44421 }, { "epoch": 13.67, "learning_rate": 4.819587839519518e-06, "loss": 0.0211, "step": 44422 }, { "epoch": 13.67, "learning_rate": 4.819161647846512e-06, "loss": 0.0307, "step": 44423 }, { "epoch": 13.67, "learning_rate": 4.818735469036131e-06, "loss": 0.0229, "step": 44424 }, { "epoch": 13.67, "learning_rate": 4.818309303089432e-06, "loss": 0.022, "step": 44425 }, { "epoch": 13.67, "learning_rate": 4.817883150007474e-06, "loss": 0.0157, "step": 44426 }, { "epoch": 13.67, "learning_rate": 4.817457009791316e-06, "loss": 0.0456, "step": 44427 }, { "epoch": 13.67, "learning_rate": 4.817030882442011e-06, "loss": 0.0211, "step": 44428 }, { "epoch": 13.67, "learning_rate": 4.816604767960626e-06, "loss": 0.0145, "step": 44429 }, { "epoch": 13.67, "learning_rate": 4.816178666348216e-06, "loss": 0.0236, "step": 44430 }, { "epoch": 13.67, "learning_rate": 4.8157525776058335e-06, "loss": 0.0202, "step": 44431 }, { "epoch": 13.67, "learning_rate": 4.81532650173454e-06, "loss": 0.0196, "step": 44432 }, { "epoch": 13.67, "learning_rate": 4.8149004387353884e-06, "loss": 0.0294, "step": 44433 }, { "epoch": 13.67, "learning_rate": 4.814474388609446e-06, "loss": 0.0272, "step": 44434 }, { "epoch": 13.67, "learning_rate": 4.814048351357765e-06, "loss": 0.0097, "step": 44435 }, { "epoch": 13.67, "learning_rate": 4.813622326981404e-06, "loss": 0.0192, "step": 44436 }, { "epoch": 13.67, "learning_rate": 4.813196315481421e-06, "loss": 0.016, "step": 44437 }, { "epoch": 13.67, "learning_rate": 4.812770316858872e-06, "loss": 0.0391, "step": 44438 }, { "epoch": 13.67, "learning_rate": 4.8123443311148175e-06, "loss": 0.0124, "step": 44439 }, { "epoch": 13.67, "learning_rate": 4.811918358250313e-06, "loss": 0.0194, "step": 44440 }, { "epoch": 13.67, "learning_rate": 4.811492398266417e-06, "loss": 0.0206, "step": 44441 }, { "epoch": 13.67, "learning_rate": 4.811066451164185e-06, "loss": 0.014, "step": 44442 }, { "epoch": 13.67, "learning_rate": 4.810640516944678e-06, "loss": 0.0186, "step": 44443 }, { "epoch": 13.67, "learning_rate": 4.81021459560895e-06, "loss": 0.0243, "step": 44444 }, { "epoch": 13.68, "learning_rate": 4.809788687158061e-06, "loss": 0.034, "step": 44445 }, { "epoch": 13.68, "learning_rate": 4.809362791593067e-06, "loss": 0.0381, "step": 44446 }, { "epoch": 13.68, "learning_rate": 4.808936908915026e-06, "loss": 0.0099, "step": 44447 }, { "epoch": 13.68, "learning_rate": 4.808511039124991e-06, "loss": 0.0177, "step": 44448 }, { "epoch": 13.68, "learning_rate": 4.808085182224028e-06, "loss": 0.0167, "step": 44449 }, { "epoch": 13.68, "learning_rate": 4.80765933821319e-06, "loss": 0.025, "step": 44450 }, { "epoch": 13.68, "learning_rate": 4.807233507093537e-06, "loss": 0.0227, "step": 44451 }, { "epoch": 13.68, "learning_rate": 4.806807688866119e-06, "loss": 0.0267, "step": 44452 }, { "epoch": 13.68, "learning_rate": 4.806381883531995e-06, "loss": 0.0231, "step": 44453 }, { "epoch": 13.68, "learning_rate": 4.805956091092228e-06, "loss": 0.0153, "step": 44454 }, { "epoch": 13.68, "learning_rate": 4.8055303115478714e-06, "loss": 0.0202, "step": 44455 }, { "epoch": 13.68, "learning_rate": 4.805104544899983e-06, "loss": 0.0351, "step": 44456 }, { "epoch": 13.68, "learning_rate": 4.80467879114962e-06, "loss": 0.0362, "step": 44457 }, { "epoch": 13.68, "learning_rate": 4.804253050297838e-06, "loss": 0.0158, "step": 44458 }, { "epoch": 13.68, "learning_rate": 4.803827322345696e-06, "loss": 0.0273, "step": 44459 }, { "epoch": 13.68, "learning_rate": 4.803401607294249e-06, "loss": 0.0258, "step": 44460 }, { "epoch": 13.68, "learning_rate": 4.802975905144554e-06, "loss": 0.0137, "step": 44461 }, { "epoch": 13.68, "learning_rate": 4.80255021589767e-06, "loss": 0.0078, "step": 44462 }, { "epoch": 13.68, "learning_rate": 4.802124539554649e-06, "loss": 0.0178, "step": 44463 }, { "epoch": 13.68, "learning_rate": 4.801698876116558e-06, "loss": 0.026, "step": 44464 }, { "epoch": 13.68, "learning_rate": 4.801273225584445e-06, "loss": 0.0389, "step": 44465 }, { "epoch": 13.68, "learning_rate": 4.800847587959369e-06, "loss": 0.021, "step": 44466 }, { "epoch": 13.68, "learning_rate": 4.800421963242385e-06, "loss": 0.0195, "step": 44467 }, { "epoch": 13.68, "learning_rate": 4.79999635143455e-06, "loss": 0.043, "step": 44468 }, { "epoch": 13.68, "learning_rate": 4.7995707525369236e-06, "loss": 0.0301, "step": 44469 }, { "epoch": 13.68, "learning_rate": 4.799145166550563e-06, "loss": 0.0223, "step": 44470 }, { "epoch": 13.68, "learning_rate": 4.798719593476522e-06, "loss": 0.034, "step": 44471 }, { "epoch": 13.68, "learning_rate": 4.798294033315861e-06, "loss": 0.0303, "step": 44472 }, { "epoch": 13.68, "learning_rate": 4.797868486069627e-06, "loss": 0.0197, "step": 44473 }, { "epoch": 13.68, "learning_rate": 4.797442951738887e-06, "loss": 0.0136, "step": 44474 }, { "epoch": 13.68, "learning_rate": 4.797017430324693e-06, "loss": 0.0125, "step": 44475 }, { "epoch": 13.68, "learning_rate": 4.796591921828102e-06, "loss": 0.0218, "step": 44476 }, { "epoch": 13.69, "learning_rate": 4.79616642625017e-06, "loss": 0.025, "step": 44477 }, { "epoch": 13.69, "learning_rate": 4.795740943591955e-06, "loss": 0.0147, "step": 44478 }, { "epoch": 13.69, "learning_rate": 4.7953154738545124e-06, "loss": 0.0117, "step": 44479 }, { "epoch": 13.69, "learning_rate": 4.794890017038898e-06, "loss": 0.0168, "step": 44480 }, { "epoch": 13.69, "learning_rate": 4.794464573146168e-06, "loss": 0.0236, "step": 44481 }, { "epoch": 13.69, "learning_rate": 4.794039142177379e-06, "loss": 0.0227, "step": 44482 }, { "epoch": 13.69, "learning_rate": 4.793613724133584e-06, "loss": 0.0167, "step": 44483 }, { "epoch": 13.69, "learning_rate": 4.793188319015847e-06, "loss": 0.0157, "step": 44484 }, { "epoch": 13.69, "learning_rate": 4.7927629268252215e-06, "loss": 0.0116, "step": 44485 }, { "epoch": 13.69, "learning_rate": 4.79233754756276e-06, "loss": 0.0107, "step": 44486 }, { "epoch": 13.69, "learning_rate": 4.791912181229519e-06, "loss": 0.0369, "step": 44487 }, { "epoch": 13.69, "learning_rate": 4.791486827826554e-06, "loss": 0.0244, "step": 44488 }, { "epoch": 13.69, "learning_rate": 4.791061487354924e-06, "loss": 0.0159, "step": 44489 }, { "epoch": 13.69, "learning_rate": 4.790636159815686e-06, "loss": 0.025, "step": 44490 }, { "epoch": 13.69, "learning_rate": 4.790210845209894e-06, "loss": 0.0138, "step": 44491 }, { "epoch": 13.69, "learning_rate": 4.789785543538605e-06, "loss": 0.0138, "step": 44492 }, { "epoch": 13.69, "learning_rate": 4.7893602548028715e-06, "loss": 0.0285, "step": 44493 }, { "epoch": 13.69, "learning_rate": 4.788934979003752e-06, "loss": 0.0175, "step": 44494 }, { "epoch": 13.69, "learning_rate": 4.788509716142303e-06, "loss": 0.0226, "step": 44495 }, { "epoch": 13.69, "learning_rate": 4.788084466219579e-06, "loss": 0.0202, "step": 44496 }, { "epoch": 13.69, "learning_rate": 4.7876592292366374e-06, "loss": 0.023, "step": 44497 }, { "epoch": 13.69, "learning_rate": 4.787234005194531e-06, "loss": 0.0156, "step": 44498 }, { "epoch": 13.69, "learning_rate": 4.786808794094319e-06, "loss": 0.0098, "step": 44499 }, { "epoch": 13.69, "learning_rate": 4.786383595937054e-06, "loss": 0.0201, "step": 44500 }, { "epoch": 13.69, "learning_rate": 4.785958410723793e-06, "loss": 0.0123, "step": 44501 }, { "epoch": 13.69, "learning_rate": 4.785533238455592e-06, "loss": 0.0116, "step": 44502 }, { "epoch": 13.69, "learning_rate": 4.785108079133503e-06, "loss": 0.0235, "step": 44503 }, { "epoch": 13.69, "learning_rate": 4.784682932758588e-06, "loss": 0.0193, "step": 44504 }, { "epoch": 13.69, "learning_rate": 4.784257799331898e-06, "loss": 0.0461, "step": 44505 }, { "epoch": 13.69, "learning_rate": 4.783832678854495e-06, "loss": 0.0168, "step": 44506 }, { "epoch": 13.69, "learning_rate": 4.783407571327425e-06, "loss": 0.0228, "step": 44507 }, { "epoch": 13.69, "learning_rate": 4.7829824767517444e-06, "loss": 0.0266, "step": 44508 }, { "epoch": 13.69, "learning_rate": 4.7825573951285155e-06, "loss": 0.0261, "step": 44509 }, { "epoch": 13.7, "learning_rate": 4.7821323264587895e-06, "loss": 0.0162, "step": 44510 }, { "epoch": 13.7, "learning_rate": 4.781707270743622e-06, "loss": 0.0165, "step": 44511 }, { "epoch": 13.7, "learning_rate": 4.781282227984069e-06, "loss": 0.0113, "step": 44512 }, { "epoch": 13.7, "learning_rate": 4.780857198181186e-06, "loss": 0.0128, "step": 44513 }, { "epoch": 13.7, "learning_rate": 4.780432181336026e-06, "loss": 0.0318, "step": 44514 }, { "epoch": 13.7, "learning_rate": 4.780007177449647e-06, "loss": 0.0187, "step": 44515 }, { "epoch": 13.7, "learning_rate": 4.779582186523102e-06, "loss": 0.0348, "step": 44516 }, { "epoch": 13.7, "learning_rate": 4.779157208557448e-06, "loss": 0.0109, "step": 44517 }, { "epoch": 13.7, "learning_rate": 4.778732243553734e-06, "loss": 0.0229, "step": 44518 }, { "epoch": 13.7, "learning_rate": 4.778307291513029e-06, "loss": 0.0238, "step": 44519 }, { "epoch": 13.7, "learning_rate": 4.777882352436376e-06, "loss": 0.023, "step": 44520 }, { "epoch": 13.7, "learning_rate": 4.777457426324832e-06, "loss": 0.0418, "step": 44521 }, { "epoch": 13.7, "learning_rate": 4.777032513179454e-06, "loss": 0.0173, "step": 44522 }, { "epoch": 13.7, "learning_rate": 4.7766076130012914e-06, "loss": 0.0048, "step": 44523 }, { "epoch": 13.7, "learning_rate": 4.776182725791409e-06, "loss": 0.0217, "step": 44524 }, { "epoch": 13.7, "learning_rate": 4.775757851550857e-06, "loss": 0.0211, "step": 44525 }, { "epoch": 13.7, "learning_rate": 4.7753329902806875e-06, "loss": 0.028, "step": 44526 }, { "epoch": 13.7, "learning_rate": 4.774908141981963e-06, "loss": 0.0198, "step": 44527 }, { "epoch": 13.7, "learning_rate": 4.774483306655725e-06, "loss": 0.0283, "step": 44528 }, { "epoch": 13.7, "learning_rate": 4.77405848430304e-06, "loss": 0.0223, "step": 44529 }, { "epoch": 13.7, "learning_rate": 4.77363367492496e-06, "loss": 0.0207, "step": 44530 }, { "epoch": 13.7, "learning_rate": 4.773208878522537e-06, "loss": 0.0264, "step": 44531 }, { "epoch": 13.7, "learning_rate": 4.7727840950968284e-06, "loss": 0.0127, "step": 44532 }, { "epoch": 13.7, "learning_rate": 4.772359324648888e-06, "loss": 0.0224, "step": 44533 }, { "epoch": 13.7, "learning_rate": 4.771934567179769e-06, "loss": 0.0142, "step": 44534 }, { "epoch": 13.7, "learning_rate": 4.771509822690527e-06, "loss": 0.0251, "step": 44535 }, { "epoch": 13.7, "learning_rate": 4.771085091182217e-06, "loss": 0.014, "step": 44536 }, { "epoch": 13.7, "learning_rate": 4.770660372655893e-06, "loss": 0.027, "step": 44537 }, { "epoch": 13.7, "learning_rate": 4.770235667112605e-06, "loss": 0.0319, "step": 44538 }, { "epoch": 13.7, "learning_rate": 4.769810974553417e-06, "loss": 0.0244, "step": 44539 }, { "epoch": 13.7, "learning_rate": 4.769386294979381e-06, "loss": 0.0207, "step": 44540 }, { "epoch": 13.7, "learning_rate": 4.768961628391545e-06, "loss": 0.0269, "step": 44541 }, { "epoch": 13.71, "learning_rate": 4.7685369747909685e-06, "loss": 0.0325, "step": 44542 }, { "epoch": 13.71, "learning_rate": 4.7681123341787e-06, "loss": 0.0163, "step": 44543 }, { "epoch": 13.71, "learning_rate": 4.7676877065558e-06, "loss": 0.016, "step": 44544 }, { "epoch": 13.71, "learning_rate": 4.767263091923323e-06, "loss": 0.0155, "step": 44545 }, { "epoch": 13.71, "learning_rate": 4.76683849028232e-06, "loss": 0.0345, "step": 44546 }, { "epoch": 13.71, "learning_rate": 4.766413901633846e-06, "loss": 0.0096, "step": 44547 }, { "epoch": 13.71, "learning_rate": 4.765989325978956e-06, "loss": 0.027, "step": 44548 }, { "epoch": 13.71, "learning_rate": 4.765564763318703e-06, "loss": 0.0048, "step": 44549 }, { "epoch": 13.71, "learning_rate": 4.765140213654143e-06, "loss": 0.0142, "step": 44550 }, { "epoch": 13.71, "learning_rate": 4.764715676986327e-06, "loss": 0.029, "step": 44551 }, { "epoch": 13.71, "learning_rate": 4.764291153316311e-06, "loss": 0.02, "step": 44552 }, { "epoch": 13.71, "learning_rate": 4.7638666426451445e-06, "loss": 0.0127, "step": 44553 }, { "epoch": 13.71, "learning_rate": 4.7634421449738935e-06, "loss": 0.0211, "step": 44554 }, { "epoch": 13.71, "learning_rate": 4.7630176603035996e-06, "loss": 0.0274, "step": 44555 }, { "epoch": 13.71, "learning_rate": 4.7625931886353215e-06, "loss": 0.0173, "step": 44556 }, { "epoch": 13.71, "learning_rate": 4.762168729970112e-06, "loss": 0.0176, "step": 44557 }, { "epoch": 13.71, "learning_rate": 4.761744284309022e-06, "loss": 0.0173, "step": 44558 }, { "epoch": 13.71, "learning_rate": 4.761319851653113e-06, "loss": 0.0278, "step": 44559 }, { "epoch": 13.71, "learning_rate": 4.760895432003434e-06, "loss": 0.008, "step": 44560 }, { "epoch": 13.71, "learning_rate": 4.760471025361043e-06, "loss": 0.019, "step": 44561 }, { "epoch": 13.71, "learning_rate": 4.760046631726987e-06, "loss": 0.0169, "step": 44562 }, { "epoch": 13.71, "learning_rate": 4.759622251102317e-06, "loss": 0.0145, "step": 44563 }, { "epoch": 13.71, "learning_rate": 4.759197883488098e-06, "loss": 0.0212, "step": 44564 }, { "epoch": 13.71, "learning_rate": 4.758773528885376e-06, "loss": 0.013, "step": 44565 }, { "epoch": 13.71, "learning_rate": 4.758349187295207e-06, "loss": 0.0114, "step": 44566 }, { "epoch": 13.71, "learning_rate": 4.757924858718645e-06, "loss": 0.0204, "step": 44567 }, { "epoch": 13.71, "learning_rate": 4.7575005431567415e-06, "loss": 0.016, "step": 44568 }, { "epoch": 13.71, "learning_rate": 4.75707624061055e-06, "loss": 0.0126, "step": 44569 }, { "epoch": 13.71, "learning_rate": 4.756651951081126e-06, "loss": 0.0259, "step": 44570 }, { "epoch": 13.71, "learning_rate": 4.75622767456952e-06, "loss": 0.0316, "step": 44571 }, { "epoch": 13.71, "learning_rate": 4.755803411076788e-06, "loss": 0.0181, "step": 44572 }, { "epoch": 13.71, "learning_rate": 4.7553791606039786e-06, "loss": 0.0214, "step": 44573 }, { "epoch": 13.71, "learning_rate": 4.754954923152153e-06, "loss": 0.0237, "step": 44574 }, { "epoch": 13.72, "learning_rate": 4.754530698722364e-06, "loss": 0.028, "step": 44575 }, { "epoch": 13.72, "learning_rate": 4.754106487315657e-06, "loss": 0.0197, "step": 44576 }, { "epoch": 13.72, "learning_rate": 4.753682288933089e-06, "loss": 0.0207, "step": 44577 }, { "epoch": 13.72, "learning_rate": 4.7532581035757105e-06, "loss": 0.0058, "step": 44578 }, { "epoch": 13.72, "learning_rate": 4.752833931244582e-06, "loss": 0.0125, "step": 44579 }, { "epoch": 13.72, "learning_rate": 4.75240977194075e-06, "loss": 0.0209, "step": 44580 }, { "epoch": 13.72, "learning_rate": 4.751985625665272e-06, "loss": 0.0196, "step": 44581 }, { "epoch": 13.72, "learning_rate": 4.7515614924192026e-06, "loss": 0.0177, "step": 44582 }, { "epoch": 13.72, "learning_rate": 4.751137372203582e-06, "loss": 0.0217, "step": 44583 }, { "epoch": 13.72, "learning_rate": 4.750713265019476e-06, "loss": 0.0296, "step": 44584 }, { "epoch": 13.72, "learning_rate": 4.7502891708679356e-06, "loss": 0.0104, "step": 44585 }, { "epoch": 13.72, "learning_rate": 4.749865089750011e-06, "loss": 0.0274, "step": 44586 }, { "epoch": 13.72, "learning_rate": 4.749441021666755e-06, "loss": 0.0227, "step": 44587 }, { "epoch": 13.72, "learning_rate": 4.749016966619223e-06, "loss": 0.0237, "step": 44588 }, { "epoch": 13.72, "learning_rate": 4.748592924608465e-06, "loss": 0.0282, "step": 44589 }, { "epoch": 13.72, "learning_rate": 4.748168895635536e-06, "loss": 0.0153, "step": 44590 }, { "epoch": 13.72, "learning_rate": 4.747744879701487e-06, "loss": 0.0186, "step": 44591 }, { "epoch": 13.72, "learning_rate": 4.747320876807372e-06, "loss": 0.0299, "step": 44592 }, { "epoch": 13.72, "learning_rate": 4.7468968869542395e-06, "loss": 0.0239, "step": 44593 }, { "epoch": 13.72, "learning_rate": 4.74647291014315e-06, "loss": 0.0372, "step": 44594 }, { "epoch": 13.72, "learning_rate": 4.74604894637515e-06, "loss": 0.0237, "step": 44595 }, { "epoch": 13.72, "learning_rate": 4.7456249956513e-06, "loss": 0.012, "step": 44596 }, { "epoch": 13.72, "learning_rate": 4.745201057972642e-06, "loss": 0.0268, "step": 44597 }, { "epoch": 13.72, "learning_rate": 4.744777133340229e-06, "loss": 0.0195, "step": 44598 }, { "epoch": 13.72, "learning_rate": 4.7443532217551225e-06, "loss": 0.017, "step": 44599 }, { "epoch": 13.72, "learning_rate": 4.7439293232183695e-06, "loss": 0.0143, "step": 44600 }, { "epoch": 13.72, "learning_rate": 4.743505437731022e-06, "loss": 0.0186, "step": 44601 }, { "epoch": 13.72, "learning_rate": 4.743081565294134e-06, "loss": 0.0276, "step": 44602 }, { "epoch": 13.72, "learning_rate": 4.742657705908758e-06, "loss": 0.036, "step": 44603 }, { "epoch": 13.72, "learning_rate": 4.7422338595759445e-06, "loss": 0.0202, "step": 44604 }, { "epoch": 13.72, "learning_rate": 4.7418100262967485e-06, "loss": 0.0198, "step": 44605 }, { "epoch": 13.72, "learning_rate": 4.741386206072219e-06, "loss": 0.0209, "step": 44606 }, { "epoch": 13.73, "learning_rate": 4.74096239890341e-06, "loss": 0.0215, "step": 44607 }, { "epoch": 13.73, "learning_rate": 4.740538604791371e-06, "loss": 0.0144, "step": 44608 }, { "epoch": 13.73, "learning_rate": 4.740114823737163e-06, "loss": 0.0219, "step": 44609 }, { "epoch": 13.73, "learning_rate": 4.739691055741829e-06, "loss": 0.0163, "step": 44610 }, { "epoch": 13.73, "learning_rate": 4.739267300806423e-06, "loss": 0.0112, "step": 44611 }, { "epoch": 13.73, "learning_rate": 4.738843558931999e-06, "loss": 0.0216, "step": 44612 }, { "epoch": 13.73, "learning_rate": 4.738419830119604e-06, "loss": 0.0144, "step": 44613 }, { "epoch": 13.73, "learning_rate": 4.737996114370298e-06, "loss": 0.0323, "step": 44614 }, { "epoch": 13.73, "learning_rate": 4.737572411685129e-06, "loss": 0.0142, "step": 44615 }, { "epoch": 13.73, "learning_rate": 4.737148722065152e-06, "loss": 0.0286, "step": 44616 }, { "epoch": 13.73, "learning_rate": 4.736725045511412e-06, "loss": 0.0146, "step": 44617 }, { "epoch": 13.73, "learning_rate": 4.736301382024963e-06, "loss": 0.0207, "step": 44618 }, { "epoch": 13.73, "learning_rate": 4.735877731606861e-06, "loss": 0.0375, "step": 44619 }, { "epoch": 13.73, "learning_rate": 4.735454094258155e-06, "loss": 0.0241, "step": 44620 }, { "epoch": 13.73, "learning_rate": 4.7350304699798976e-06, "loss": 0.0291, "step": 44621 }, { "epoch": 13.73, "learning_rate": 4.734606858773141e-06, "loss": 0.0255, "step": 44622 }, { "epoch": 13.73, "learning_rate": 4.734183260638935e-06, "loss": 0.018, "step": 44623 }, { "epoch": 13.73, "learning_rate": 4.733759675578332e-06, "loss": 0.0319, "step": 44624 }, { "epoch": 13.73, "learning_rate": 4.7333361035923855e-06, "loss": 0.0335, "step": 44625 }, { "epoch": 13.73, "learning_rate": 4.732912544682144e-06, "loss": 0.0165, "step": 44626 }, { "epoch": 13.73, "learning_rate": 4.732488998848662e-06, "loss": 0.0253, "step": 44627 }, { "epoch": 13.73, "learning_rate": 4.7320654660929854e-06, "loss": 0.0241, "step": 44628 }, { "epoch": 13.73, "learning_rate": 4.7316419464161744e-06, "loss": 0.0122, "step": 44629 }, { "epoch": 13.73, "learning_rate": 4.731218439819279e-06, "loss": 0.0204, "step": 44630 }, { "epoch": 13.73, "learning_rate": 4.730794946303344e-06, "loss": 0.0296, "step": 44631 }, { "epoch": 13.73, "learning_rate": 4.730371465869426e-06, "loss": 0.0167, "step": 44632 }, { "epoch": 13.73, "learning_rate": 4.729947998518571e-06, "loss": 0.0331, "step": 44633 }, { "epoch": 13.73, "learning_rate": 4.729524544251837e-06, "loss": 0.0356, "step": 44634 }, { "epoch": 13.73, "learning_rate": 4.729101103070273e-06, "loss": 0.0176, "step": 44635 }, { "epoch": 13.73, "learning_rate": 4.72867767497493e-06, "loss": 0.0177, "step": 44636 }, { "epoch": 13.73, "learning_rate": 4.728254259966862e-06, "loss": 0.018, "step": 44637 }, { "epoch": 13.73, "learning_rate": 4.727830858047112e-06, "loss": 0.0255, "step": 44638 }, { "epoch": 13.73, "learning_rate": 4.727407469216739e-06, "loss": 0.0171, "step": 44639 }, { "epoch": 13.74, "learning_rate": 4.726984093476792e-06, "loss": 0.0256, "step": 44640 }, { "epoch": 13.74, "learning_rate": 4.726560730828322e-06, "loss": 0.0125, "step": 44641 }, { "epoch": 13.74, "learning_rate": 4.726137381272381e-06, "loss": 0.0157, "step": 44642 }, { "epoch": 13.74, "learning_rate": 4.725714044810018e-06, "loss": 0.0308, "step": 44643 }, { "epoch": 13.74, "learning_rate": 4.725290721442285e-06, "loss": 0.0259, "step": 44644 }, { "epoch": 13.74, "learning_rate": 4.724867411170234e-06, "loss": 0.0099, "step": 44645 }, { "epoch": 13.74, "learning_rate": 4.724444113994915e-06, "loss": 0.0109, "step": 44646 }, { "epoch": 13.74, "learning_rate": 4.724020829917378e-06, "loss": 0.0178, "step": 44647 }, { "epoch": 13.74, "learning_rate": 4.7235975589386715e-06, "loss": 0.0201, "step": 44648 }, { "epoch": 13.74, "learning_rate": 4.723174301059854e-06, "loss": 0.0181, "step": 44649 }, { "epoch": 13.74, "learning_rate": 4.7227510562819724e-06, "loss": 0.0089, "step": 44650 }, { "epoch": 13.74, "learning_rate": 4.72232782460608e-06, "loss": 0.0245, "step": 44651 }, { "epoch": 13.74, "learning_rate": 4.721904606033222e-06, "loss": 0.0268, "step": 44652 }, { "epoch": 13.74, "learning_rate": 4.721481400564447e-06, "loss": 0.0135, "step": 44653 }, { "epoch": 13.74, "learning_rate": 4.721058208200816e-06, "loss": 0.0312, "step": 44654 }, { "epoch": 13.74, "learning_rate": 4.720635028943373e-06, "loss": 0.0284, "step": 44655 }, { "epoch": 13.74, "learning_rate": 4.72021186279317e-06, "loss": 0.0217, "step": 44656 }, { "epoch": 13.74, "learning_rate": 4.719788709751259e-06, "loss": 0.014, "step": 44657 }, { "epoch": 13.74, "learning_rate": 4.719365569818687e-06, "loss": 0.0229, "step": 44658 }, { "epoch": 13.74, "learning_rate": 4.718942442996508e-06, "loss": 0.0231, "step": 44659 }, { "epoch": 13.74, "learning_rate": 4.718519329285771e-06, "loss": 0.0294, "step": 44660 }, { "epoch": 13.74, "learning_rate": 4.7180962286875265e-06, "loss": 0.012, "step": 44661 }, { "epoch": 13.74, "learning_rate": 4.717673141202825e-06, "loss": 0.0157, "step": 44662 }, { "epoch": 13.74, "learning_rate": 4.7172500668327135e-06, "loss": 0.0096, "step": 44663 }, { "epoch": 13.74, "learning_rate": 4.716827005578254e-06, "loss": 0.0162, "step": 44664 }, { "epoch": 13.74, "learning_rate": 4.716403957440484e-06, "loss": 0.0098, "step": 44665 }, { "epoch": 13.74, "learning_rate": 4.715980922420459e-06, "loss": 0.0136, "step": 44666 }, { "epoch": 13.74, "learning_rate": 4.715557900519229e-06, "loss": 0.0166, "step": 44667 }, { "epoch": 13.74, "learning_rate": 4.715134891737841e-06, "loss": 0.0169, "step": 44668 }, { "epoch": 13.74, "learning_rate": 4.71471189607735e-06, "loss": 0.0216, "step": 44669 }, { "epoch": 13.74, "learning_rate": 4.714288913538806e-06, "loss": 0.047, "step": 44670 }, { "epoch": 13.74, "learning_rate": 4.713865944123257e-06, "loss": 0.029, "step": 44671 }, { "epoch": 13.75, "learning_rate": 4.713442987831758e-06, "loss": 0.0095, "step": 44672 }, { "epoch": 13.75, "learning_rate": 4.713020044665348e-06, "loss": 0.0335, "step": 44673 }, { "epoch": 13.75, "learning_rate": 4.712597114625086e-06, "loss": 0.0254, "step": 44674 }, { "epoch": 13.75, "learning_rate": 4.712174197712021e-06, "loss": 0.0161, "step": 44675 }, { "epoch": 13.75, "learning_rate": 4.711751293927202e-06, "loss": 0.0115, "step": 44676 }, { "epoch": 13.75, "learning_rate": 4.711328403271678e-06, "loss": 0.0168, "step": 44677 }, { "epoch": 13.75, "learning_rate": 4.7109055257465e-06, "loss": 0.023, "step": 44678 }, { "epoch": 13.75, "learning_rate": 4.710482661352718e-06, "loss": 0.0189, "step": 44679 }, { "epoch": 13.75, "learning_rate": 4.710059810091382e-06, "loss": 0.0189, "step": 44680 }, { "epoch": 13.75, "learning_rate": 4.709636971963541e-06, "loss": 0.0164, "step": 44681 }, { "epoch": 13.75, "learning_rate": 4.709214146970245e-06, "loss": 0.0203, "step": 44682 }, { "epoch": 13.75, "learning_rate": 4.70879133511254e-06, "loss": 0.0224, "step": 44683 }, { "epoch": 13.75, "learning_rate": 4.708368536391483e-06, "loss": 0.0257, "step": 44684 }, { "epoch": 13.75, "learning_rate": 4.707945750808125e-06, "loss": 0.013, "step": 44685 }, { "epoch": 13.75, "learning_rate": 4.707522978363508e-06, "loss": 0.0326, "step": 44686 }, { "epoch": 13.75, "learning_rate": 4.7071002190586836e-06, "loss": 0.027, "step": 44687 }, { "epoch": 13.75, "learning_rate": 4.706677472894699e-06, "loss": 0.0198, "step": 44688 }, { "epoch": 13.75, "learning_rate": 4.70625473987261e-06, "loss": 0.0187, "step": 44689 }, { "epoch": 13.75, "learning_rate": 4.705832019993464e-06, "loss": 0.0284, "step": 44690 }, { "epoch": 13.75, "learning_rate": 4.70540931325831e-06, "loss": 0.0241, "step": 44691 }, { "epoch": 13.75, "learning_rate": 4.704986619668197e-06, "loss": 0.0153, "step": 44692 }, { "epoch": 13.75, "learning_rate": 4.704563939224175e-06, "loss": 0.0172, "step": 44693 }, { "epoch": 13.75, "learning_rate": 4.704141271927293e-06, "loss": 0.0241, "step": 44694 }, { "epoch": 13.75, "learning_rate": 4.7037186177786e-06, "loss": 0.0171, "step": 44695 }, { "epoch": 13.75, "learning_rate": 4.703295976779147e-06, "loss": 0.0151, "step": 44696 }, { "epoch": 13.75, "learning_rate": 4.70287334892998e-06, "loss": 0.0226, "step": 44697 }, { "epoch": 13.75, "learning_rate": 4.702450734232154e-06, "loss": 0.0205, "step": 44698 }, { "epoch": 13.75, "learning_rate": 4.702028132686711e-06, "loss": 0.0218, "step": 44699 }, { "epoch": 13.75, "learning_rate": 4.701605544294706e-06, "loss": 0.0161, "step": 44700 }, { "epoch": 13.75, "learning_rate": 4.701182969057185e-06, "loss": 0.0203, "step": 44701 }, { "epoch": 13.75, "learning_rate": 4.700760406975199e-06, "loss": 0.0122, "step": 44702 }, { "epoch": 13.75, "learning_rate": 4.700337858049792e-06, "loss": 0.0232, "step": 44703 }, { "epoch": 13.75, "learning_rate": 4.699915322282021e-06, "loss": 0.0376, "step": 44704 }, { "epoch": 13.76, "learning_rate": 4.699492799672932e-06, "loss": 0.0183, "step": 44705 }, { "epoch": 13.76, "learning_rate": 4.699070290223576e-06, "loss": 0.0216, "step": 44706 }, { "epoch": 13.76, "learning_rate": 4.6986477939349976e-06, "loss": 0.02, "step": 44707 }, { "epoch": 13.76, "learning_rate": 4.698225310808242e-06, "loss": 0.0134, "step": 44708 }, { "epoch": 13.76, "learning_rate": 4.697802840844368e-06, "loss": 0.0257, "step": 44709 }, { "epoch": 13.76, "learning_rate": 4.69738038404442e-06, "loss": 0.0199, "step": 44710 }, { "epoch": 13.76, "learning_rate": 4.696957940409447e-06, "loss": 0.0198, "step": 44711 }, { "epoch": 13.76, "learning_rate": 4.696535509940499e-06, "loss": 0.0242, "step": 44712 }, { "epoch": 13.76, "learning_rate": 4.696113092638621e-06, "loss": 0.0154, "step": 44713 }, { "epoch": 13.76, "learning_rate": 4.695690688504866e-06, "loss": 0.0101, "step": 44714 }, { "epoch": 13.76, "learning_rate": 4.695268297540281e-06, "loss": 0.0191, "step": 44715 }, { "epoch": 13.76, "learning_rate": 4.694845919745914e-06, "loss": 0.0192, "step": 44716 }, { "epoch": 13.76, "learning_rate": 4.694423555122816e-06, "loss": 0.0261, "step": 44717 }, { "epoch": 13.76, "learning_rate": 4.694001203672028e-06, "loss": 0.0268, "step": 44718 }, { "epoch": 13.76, "learning_rate": 4.693578865394613e-06, "loss": 0.0219, "step": 44719 }, { "epoch": 13.76, "learning_rate": 4.693156540291608e-06, "loss": 0.0269, "step": 44720 }, { "epoch": 13.76, "learning_rate": 4.692734228364064e-06, "loss": 0.0114, "step": 44721 }, { "epoch": 13.76, "learning_rate": 4.69231192961303e-06, "loss": 0.0181, "step": 44722 }, { "epoch": 13.76, "learning_rate": 4.691889644039551e-06, "loss": 0.0238, "step": 44723 }, { "epoch": 13.76, "learning_rate": 4.691467371644682e-06, "loss": 0.0301, "step": 44724 }, { "epoch": 13.76, "learning_rate": 4.69104511242947e-06, "loss": 0.0137, "step": 44725 }, { "epoch": 13.76, "learning_rate": 4.690622866394961e-06, "loss": 0.0169, "step": 44726 }, { "epoch": 13.76, "learning_rate": 4.690200633542203e-06, "loss": 0.0274, "step": 44727 }, { "epoch": 13.76, "learning_rate": 4.6897784138722465e-06, "loss": 0.0206, "step": 44728 }, { "epoch": 13.76, "learning_rate": 4.689356207386139e-06, "loss": 0.0144, "step": 44729 }, { "epoch": 13.76, "learning_rate": 4.688934014084927e-06, "loss": 0.0124, "step": 44730 }, { "epoch": 13.76, "learning_rate": 4.6885118339696614e-06, "loss": 0.0241, "step": 44731 }, { "epoch": 13.76, "learning_rate": 4.688089667041384e-06, "loss": 0.0206, "step": 44732 }, { "epoch": 13.76, "learning_rate": 4.687667513301156e-06, "loss": 0.0224, "step": 44733 }, { "epoch": 13.76, "learning_rate": 4.687245372750014e-06, "loss": 0.0332, "step": 44734 }, { "epoch": 13.76, "learning_rate": 4.686823245389009e-06, "loss": 0.0174, "step": 44735 }, { "epoch": 13.76, "learning_rate": 4.686401131219189e-06, "loss": 0.0148, "step": 44736 }, { "epoch": 13.77, "learning_rate": 4.6859790302416e-06, "loss": 0.0226, "step": 44737 }, { "epoch": 13.77, "learning_rate": 4.685556942457296e-06, "loss": 0.0415, "step": 44738 }, { "epoch": 13.77, "learning_rate": 4.68513486786732e-06, "loss": 0.0157, "step": 44739 }, { "epoch": 13.77, "learning_rate": 4.684712806472727e-06, "loss": 0.0298, "step": 44740 }, { "epoch": 13.77, "learning_rate": 4.684290758274554e-06, "loss": 0.0163, "step": 44741 }, { "epoch": 13.77, "learning_rate": 4.68386872327385e-06, "loss": 0.0151, "step": 44742 }, { "epoch": 13.77, "learning_rate": 4.683446701471671e-06, "loss": 0.0218, "step": 44743 }, { "epoch": 13.77, "learning_rate": 4.683024692869062e-06, "loss": 0.0348, "step": 44744 }, { "epoch": 13.77, "learning_rate": 4.6826026974670665e-06, "loss": 0.0124, "step": 44745 }, { "epoch": 13.77, "learning_rate": 4.682180715266736e-06, "loss": 0.0178, "step": 44746 }, { "epoch": 13.77, "learning_rate": 4.681758746269118e-06, "loss": 0.0169, "step": 44747 }, { "epoch": 13.77, "learning_rate": 4.681336790475259e-06, "loss": 0.0128, "step": 44748 }, { "epoch": 13.77, "learning_rate": 4.680914847886206e-06, "loss": 0.0157, "step": 44749 }, { "epoch": 13.77, "learning_rate": 4.680492918503009e-06, "loss": 0.0365, "step": 44750 }, { "epoch": 13.77, "learning_rate": 4.680071002326714e-06, "loss": 0.0241, "step": 44751 }, { "epoch": 13.77, "learning_rate": 4.679649099358363e-06, "loss": 0.0235, "step": 44752 }, { "epoch": 13.77, "learning_rate": 4.679227209599017e-06, "loss": 0.0165, "step": 44753 }, { "epoch": 13.77, "learning_rate": 4.678805333049713e-06, "loss": 0.0197, "step": 44754 }, { "epoch": 13.77, "learning_rate": 4.678383469711499e-06, "loss": 0.0218, "step": 44755 }, { "epoch": 13.77, "learning_rate": 4.677961619585425e-06, "loss": 0.024, "step": 44756 }, { "epoch": 13.77, "learning_rate": 4.677539782672534e-06, "loss": 0.0189, "step": 44757 }, { "epoch": 13.77, "learning_rate": 4.67711795897388e-06, "loss": 0.0086, "step": 44758 }, { "epoch": 13.77, "learning_rate": 4.676696148490508e-06, "loss": 0.0188, "step": 44759 }, { "epoch": 13.77, "learning_rate": 4.676274351223463e-06, "loss": 0.0334, "step": 44760 }, { "epoch": 13.77, "learning_rate": 4.6758525671737984e-06, "loss": 0.0279, "step": 44761 }, { "epoch": 13.77, "learning_rate": 4.675430796342549e-06, "loss": 0.021, "step": 44762 }, { "epoch": 13.77, "learning_rate": 4.675009038730773e-06, "loss": 0.0205, "step": 44763 }, { "epoch": 13.77, "learning_rate": 4.674587294339513e-06, "loss": 0.0104, "step": 44764 }, { "epoch": 13.77, "learning_rate": 4.674165563169819e-06, "loss": 0.0261, "step": 44765 }, { "epoch": 13.77, "learning_rate": 4.6737438452227345e-06, "loss": 0.0848, "step": 44766 }, { "epoch": 13.77, "learning_rate": 4.673322140499309e-06, "loss": 0.0141, "step": 44767 }, { "epoch": 13.77, "learning_rate": 4.672900449000588e-06, "loss": 0.0166, "step": 44768 }, { "epoch": 13.77, "learning_rate": 4.6724787707276195e-06, "loss": 0.0419, "step": 44769 }, { "epoch": 13.78, "learning_rate": 4.67205710568145e-06, "loss": 0.0167, "step": 44770 }, { "epoch": 13.78, "learning_rate": 4.671635453863126e-06, "loss": 0.018, "step": 44771 }, { "epoch": 13.78, "learning_rate": 4.671213815273691e-06, "loss": 0.0294, "step": 44772 }, { "epoch": 13.78, "learning_rate": 4.6707921899142e-06, "loss": 0.0314, "step": 44773 }, { "epoch": 13.78, "learning_rate": 4.670370577785698e-06, "loss": 0.0464, "step": 44774 }, { "epoch": 13.78, "learning_rate": 4.669948978889225e-06, "loss": 0.0114, "step": 44775 }, { "epoch": 13.78, "learning_rate": 4.669527393225833e-06, "loss": 0.0185, "step": 44776 }, { "epoch": 13.78, "learning_rate": 4.669105820796564e-06, "loss": 0.0168, "step": 44777 }, { "epoch": 13.78, "learning_rate": 4.668684261602469e-06, "loss": 0.011, "step": 44778 }, { "epoch": 13.78, "learning_rate": 4.668262715644596e-06, "loss": 0.0276, "step": 44779 }, { "epoch": 13.78, "learning_rate": 4.6678411829239885e-06, "loss": 0.02, "step": 44780 }, { "epoch": 13.78, "learning_rate": 4.667419663441695e-06, "loss": 0.0336, "step": 44781 }, { "epoch": 13.78, "learning_rate": 4.66699815719876e-06, "loss": 0.0204, "step": 44782 }, { "epoch": 13.78, "learning_rate": 4.666576664196231e-06, "loss": 0.0216, "step": 44783 }, { "epoch": 13.78, "learning_rate": 4.666155184435153e-06, "loss": 0.0136, "step": 44784 }, { "epoch": 13.78, "learning_rate": 4.665733717916576e-06, "loss": 0.0279, "step": 44785 }, { "epoch": 13.78, "learning_rate": 4.665312264641542e-06, "loss": 0.0149, "step": 44786 }, { "epoch": 13.78, "learning_rate": 4.664890824611098e-06, "loss": 0.0138, "step": 44787 }, { "epoch": 13.78, "learning_rate": 4.664469397826297e-06, "loss": 0.0167, "step": 44788 }, { "epoch": 13.78, "learning_rate": 4.664047984288178e-06, "loss": 0.0168, "step": 44789 }, { "epoch": 13.78, "learning_rate": 4.663626583997789e-06, "loss": 0.0228, "step": 44790 }, { "epoch": 13.78, "learning_rate": 4.663205196956177e-06, "loss": 0.0161, "step": 44791 }, { "epoch": 13.78, "learning_rate": 4.662783823164383e-06, "loss": 0.0141, "step": 44792 }, { "epoch": 13.78, "learning_rate": 4.662362462623461e-06, "loss": 0.0163, "step": 44793 }, { "epoch": 13.78, "learning_rate": 4.661941115334455e-06, "loss": 0.0304, "step": 44794 }, { "epoch": 13.78, "learning_rate": 4.661519781298414e-06, "loss": 0.0306, "step": 44795 }, { "epoch": 13.78, "learning_rate": 4.661098460516376e-06, "loss": 0.0426, "step": 44796 }, { "epoch": 13.78, "learning_rate": 4.660677152989388e-06, "loss": 0.0146, "step": 44797 }, { "epoch": 13.78, "learning_rate": 4.660255858718502e-06, "loss": 0.0255, "step": 44798 }, { "epoch": 13.78, "learning_rate": 4.659834577704761e-06, "loss": 0.0272, "step": 44799 }, { "epoch": 13.78, "learning_rate": 4.659413309949214e-06, "loss": 0.0227, "step": 44800 }, { "epoch": 13.78, "learning_rate": 4.658992055452901e-06, "loss": 0.0308, "step": 44801 }, { "epoch": 13.79, "learning_rate": 4.658570814216872e-06, "loss": 0.0202, "step": 44802 }, { "epoch": 13.79, "learning_rate": 4.65814958624217e-06, "loss": 0.0206, "step": 44803 }, { "epoch": 13.79, "learning_rate": 4.657728371529844e-06, "loss": 0.0292, "step": 44804 }, { "epoch": 13.79, "learning_rate": 4.657307170080938e-06, "loss": 0.0119, "step": 44805 }, { "epoch": 13.79, "learning_rate": 4.656885981896498e-06, "loss": 0.0223, "step": 44806 }, { "epoch": 13.79, "learning_rate": 4.6564648069775665e-06, "loss": 0.0225, "step": 44807 }, { "epoch": 13.79, "learning_rate": 4.656043645325194e-06, "loss": 0.0108, "step": 44808 }, { "epoch": 13.79, "learning_rate": 4.65562249694043e-06, "loss": 0.0166, "step": 44809 }, { "epoch": 13.79, "learning_rate": 4.6552013618243105e-06, "loss": 0.0162, "step": 44810 }, { "epoch": 13.79, "learning_rate": 4.654780239977885e-06, "loss": 0.029, "step": 44811 }, { "epoch": 13.79, "learning_rate": 4.654359131402195e-06, "loss": 0.0183, "step": 44812 }, { "epoch": 13.79, "learning_rate": 4.653938036098295e-06, "loss": 0.0163, "step": 44813 }, { "epoch": 13.79, "learning_rate": 4.653516954067225e-06, "loss": 0.0169, "step": 44814 }, { "epoch": 13.79, "learning_rate": 4.653095885310031e-06, "loss": 0.0088, "step": 44815 }, { "epoch": 13.79, "learning_rate": 4.652674829827762e-06, "loss": 0.0102, "step": 44816 }, { "epoch": 13.79, "learning_rate": 4.652253787621452e-06, "loss": 0.0176, "step": 44817 }, { "epoch": 13.79, "learning_rate": 4.651832758692158e-06, "loss": 0.0307, "step": 44818 }, { "epoch": 13.79, "learning_rate": 4.651411743040921e-06, "loss": 0.0171, "step": 44819 }, { "epoch": 13.79, "learning_rate": 4.650990740668788e-06, "loss": 0.0137, "step": 44820 }, { "epoch": 13.79, "learning_rate": 4.650569751576802e-06, "loss": 0.0098, "step": 44821 }, { "epoch": 13.79, "learning_rate": 4.65014877576601e-06, "loss": 0.0341, "step": 44822 }, { "epoch": 13.79, "learning_rate": 4.649727813237456e-06, "loss": 0.0701, "step": 44823 }, { "epoch": 13.79, "learning_rate": 4.6493068639921844e-06, "loss": 0.0096, "step": 44824 }, { "epoch": 13.79, "learning_rate": 4.648885928031242e-06, "loss": 0.0413, "step": 44825 }, { "epoch": 13.79, "learning_rate": 4.648465005355675e-06, "loss": 0.0231, "step": 44826 }, { "epoch": 13.79, "learning_rate": 4.648044095966521e-06, "loss": 0.0363, "step": 44827 }, { "epoch": 13.79, "learning_rate": 4.647623199864834e-06, "loss": 0.0167, "step": 44828 }, { "epoch": 13.79, "learning_rate": 4.647202317051656e-06, "loss": 0.0162, "step": 44829 }, { "epoch": 13.79, "learning_rate": 4.6467814475280355e-06, "loss": 0.0171, "step": 44830 }, { "epoch": 13.79, "learning_rate": 4.6463605912950105e-06, "loss": 0.0348, "step": 44831 }, { "epoch": 13.79, "learning_rate": 4.645939748353625e-06, "loss": 0.0317, "step": 44832 }, { "epoch": 13.79, "learning_rate": 4.645518918704931e-06, "loss": 0.0136, "step": 44833 }, { "epoch": 13.79, "learning_rate": 4.645098102349969e-06, "loss": 0.0144, "step": 44834 }, { "epoch": 13.8, "learning_rate": 4.6446772992897874e-06, "loss": 0.0212, "step": 44835 }, { "epoch": 13.8, "learning_rate": 4.644256509525426e-06, "loss": 0.0237, "step": 44836 }, { "epoch": 13.8, "learning_rate": 4.643835733057932e-06, "loss": 0.0205, "step": 44837 }, { "epoch": 13.8, "learning_rate": 4.64341496988835e-06, "loss": 0.0233, "step": 44838 }, { "epoch": 13.8, "learning_rate": 4.642994220017726e-06, "loss": 0.0135, "step": 44839 }, { "epoch": 13.8, "learning_rate": 4.642573483447101e-06, "loss": 0.0192, "step": 44840 }, { "epoch": 13.8, "learning_rate": 4.642152760177524e-06, "loss": 0.0163, "step": 44841 }, { "epoch": 13.8, "learning_rate": 4.641732050210032e-06, "loss": 0.0227, "step": 44842 }, { "epoch": 13.8, "learning_rate": 4.641311353545682e-06, "loss": 0.0438, "step": 44843 }, { "epoch": 13.8, "learning_rate": 4.640890670185508e-06, "loss": 0.0214, "step": 44844 }, { "epoch": 13.8, "learning_rate": 4.640470000130558e-06, "loss": 0.0295, "step": 44845 }, { "epoch": 13.8, "learning_rate": 4.640049343381876e-06, "loss": 0.0166, "step": 44846 }, { "epoch": 13.8, "learning_rate": 4.639628699940503e-06, "loss": 0.0209, "step": 44847 }, { "epoch": 13.8, "learning_rate": 4.63920806980749e-06, "loss": 0.0263, "step": 44848 }, { "epoch": 13.8, "learning_rate": 4.638787452983878e-06, "loss": 0.0181, "step": 44849 }, { "epoch": 13.8, "learning_rate": 4.638366849470712e-06, "loss": 0.0373, "step": 44850 }, { "epoch": 13.8, "learning_rate": 4.637946259269039e-06, "loss": 0.0414, "step": 44851 }, { "epoch": 13.8, "learning_rate": 4.637525682379893e-06, "loss": 0.0163, "step": 44852 }, { "epoch": 13.8, "learning_rate": 4.637105118804328e-06, "loss": 0.0204, "step": 44853 }, { "epoch": 13.8, "learning_rate": 4.636684568543386e-06, "loss": 0.017, "step": 44854 }, { "epoch": 13.8, "learning_rate": 4.63626403159811e-06, "loss": 0.0177, "step": 44855 }, { "epoch": 13.8, "learning_rate": 4.635843507969544e-06, "loss": 0.0137, "step": 44856 }, { "epoch": 13.8, "learning_rate": 4.6354229976587324e-06, "loss": 0.0176, "step": 44857 }, { "epoch": 13.8, "learning_rate": 4.63500250066672e-06, "loss": 0.0138, "step": 44858 }, { "epoch": 13.8, "learning_rate": 4.63458201699455e-06, "loss": 0.0328, "step": 44859 }, { "epoch": 13.8, "learning_rate": 4.634161546643265e-06, "loss": 0.022, "step": 44860 }, { "epoch": 13.8, "learning_rate": 4.633741089613911e-06, "loss": 0.0282, "step": 44861 }, { "epoch": 13.8, "learning_rate": 4.6333206459075285e-06, "loss": 0.0091, "step": 44862 }, { "epoch": 13.8, "learning_rate": 4.632900215525167e-06, "loss": 0.0231, "step": 44863 }, { "epoch": 13.8, "learning_rate": 4.63247979846787e-06, "loss": 0.0332, "step": 44864 }, { "epoch": 13.8, "learning_rate": 4.632059394736676e-06, "loss": 0.0213, "step": 44865 }, { "epoch": 13.8, "learning_rate": 4.631639004332632e-06, "loss": 0.0199, "step": 44866 }, { "epoch": 13.81, "learning_rate": 4.631218627256777e-06, "loss": 0.0121, "step": 44867 }, { "epoch": 13.81, "learning_rate": 4.630798263510162e-06, "loss": 0.0413, "step": 44868 }, { "epoch": 13.81, "learning_rate": 4.630377913093828e-06, "loss": 0.0334, "step": 44869 }, { "epoch": 13.81, "learning_rate": 4.629957576008817e-06, "loss": 0.0138, "step": 44870 }, { "epoch": 13.81, "learning_rate": 4.6295372522561744e-06, "loss": 0.0106, "step": 44871 }, { "epoch": 13.81, "learning_rate": 4.629116941836943e-06, "loss": 0.0193, "step": 44872 }, { "epoch": 13.81, "learning_rate": 4.628696644752166e-06, "loss": 0.0115, "step": 44873 }, { "epoch": 13.81, "learning_rate": 4.628276361002887e-06, "loss": 0.0154, "step": 44874 }, { "epoch": 13.81, "learning_rate": 4.627856090590149e-06, "loss": 0.012, "step": 44875 }, { "epoch": 13.81, "learning_rate": 4.627435833514998e-06, "loss": 0.022, "step": 44876 }, { "epoch": 13.81, "learning_rate": 4.627015589778474e-06, "loss": 0.0208, "step": 44877 }, { "epoch": 13.81, "learning_rate": 4.626595359381623e-06, "loss": 0.0174, "step": 44878 }, { "epoch": 13.81, "learning_rate": 4.626175142325486e-06, "loss": 0.0278, "step": 44879 }, { "epoch": 13.81, "learning_rate": 4.625754938611107e-06, "loss": 0.0159, "step": 44880 }, { "epoch": 13.81, "learning_rate": 4.625334748239531e-06, "loss": 0.0197, "step": 44881 }, { "epoch": 13.81, "learning_rate": 4.624914571211795e-06, "loss": 0.0203, "step": 44882 }, { "epoch": 13.81, "learning_rate": 4.624494407528951e-06, "loss": 0.0239, "step": 44883 }, { "epoch": 13.81, "learning_rate": 4.624074257192038e-06, "loss": 0.0309, "step": 44884 }, { "epoch": 13.81, "learning_rate": 4.623654120202104e-06, "loss": 0.015, "step": 44885 }, { "epoch": 13.81, "learning_rate": 4.623233996560183e-06, "loss": 0.022, "step": 44886 }, { "epoch": 13.81, "learning_rate": 4.6228138862673175e-06, "loss": 0.0271, "step": 44887 }, { "epoch": 13.81, "learning_rate": 4.62239378932456e-06, "loss": 0.0171, "step": 44888 }, { "epoch": 13.81, "learning_rate": 4.621973705732951e-06, "loss": 0.0406, "step": 44889 }, { "epoch": 13.81, "learning_rate": 4.62155363549353e-06, "loss": 0.0145, "step": 44890 }, { "epoch": 13.81, "learning_rate": 4.6211335786073406e-06, "loss": 0.019, "step": 44891 }, { "epoch": 13.81, "learning_rate": 4.620713535075428e-06, "loss": 0.0206, "step": 44892 }, { "epoch": 13.81, "learning_rate": 4.620293504898832e-06, "loss": 0.0143, "step": 44893 }, { "epoch": 13.81, "learning_rate": 4.619873488078597e-06, "loss": 0.0207, "step": 44894 }, { "epoch": 13.81, "learning_rate": 4.619453484615766e-06, "loss": 0.0189, "step": 44895 }, { "epoch": 13.81, "learning_rate": 4.619033494511382e-06, "loss": 0.0411, "step": 44896 }, { "epoch": 13.81, "learning_rate": 4.618613517766483e-06, "loss": 0.0154, "step": 44897 }, { "epoch": 13.81, "learning_rate": 4.618193554382123e-06, "loss": 0.0167, "step": 44898 }, { "epoch": 13.81, "learning_rate": 4.617773604359334e-06, "loss": 0.0155, "step": 44899 }, { "epoch": 13.82, "learning_rate": 4.617353667699163e-06, "loss": 0.0212, "step": 44900 }, { "epoch": 13.82, "learning_rate": 4.616933744402651e-06, "loss": 0.0222, "step": 44901 }, { "epoch": 13.82, "learning_rate": 4.616513834470837e-06, "loss": 0.0154, "step": 44902 }, { "epoch": 13.82, "learning_rate": 4.616093937904772e-06, "loss": 0.0178, "step": 44903 }, { "epoch": 13.82, "learning_rate": 4.615674054705494e-06, "loss": 0.0188, "step": 44904 }, { "epoch": 13.82, "learning_rate": 4.615254184874046e-06, "loss": 0.0158, "step": 44905 }, { "epoch": 13.82, "learning_rate": 4.614834328411474e-06, "loss": 0.0383, "step": 44906 }, { "epoch": 13.82, "learning_rate": 4.614414485318809e-06, "loss": 0.0118, "step": 44907 }, { "epoch": 13.82, "learning_rate": 4.613994655597105e-06, "loss": 0.0146, "step": 44908 }, { "epoch": 13.82, "learning_rate": 4.6135748392474e-06, "loss": 0.0169, "step": 44909 }, { "epoch": 13.82, "learning_rate": 4.613155036270736e-06, "loss": 0.0193, "step": 44910 }, { "epoch": 13.82, "learning_rate": 4.6127352466681574e-06, "loss": 0.0143, "step": 44911 }, { "epoch": 13.82, "learning_rate": 4.612315470440704e-06, "loss": 0.0115, "step": 44912 }, { "epoch": 13.82, "learning_rate": 4.611895707589418e-06, "loss": 0.0368, "step": 44913 }, { "epoch": 13.82, "learning_rate": 4.611475958115342e-06, "loss": 0.0205, "step": 44914 }, { "epoch": 13.82, "learning_rate": 4.6110562220195195e-06, "loss": 0.0201, "step": 44915 }, { "epoch": 13.82, "learning_rate": 4.610636499302992e-06, "loss": 0.0163, "step": 44916 }, { "epoch": 13.82, "learning_rate": 4.610216789966797e-06, "loss": 0.0155, "step": 44917 }, { "epoch": 13.82, "learning_rate": 4.609797094011984e-06, "loss": 0.0054, "step": 44918 }, { "epoch": 13.82, "learning_rate": 4.609377411439596e-06, "loss": 0.0247, "step": 44919 }, { "epoch": 13.82, "learning_rate": 4.608957742250667e-06, "loss": 0.0086, "step": 44920 }, { "epoch": 13.82, "learning_rate": 4.608538086446242e-06, "loss": 0.0226, "step": 44921 }, { "epoch": 13.82, "learning_rate": 4.6081184440273596e-06, "loss": 0.0223, "step": 44922 }, { "epoch": 13.82, "learning_rate": 4.60769881499507e-06, "loss": 0.0144, "step": 44923 }, { "epoch": 13.82, "learning_rate": 4.6072791993504094e-06, "loss": 0.0189, "step": 44924 }, { "epoch": 13.82, "learning_rate": 4.606859597094422e-06, "loss": 0.0215, "step": 44925 }, { "epoch": 13.82, "learning_rate": 4.6064400082281466e-06, "loss": 0.0208, "step": 44926 }, { "epoch": 13.82, "learning_rate": 4.606020432752628e-06, "loss": 0.0127, "step": 44927 }, { "epoch": 13.82, "learning_rate": 4.605600870668906e-06, "loss": 0.0111, "step": 44928 }, { "epoch": 13.82, "learning_rate": 4.605181321978023e-06, "loss": 0.0267, "step": 44929 }, { "epoch": 13.82, "learning_rate": 4.60476178668102e-06, "loss": 0.011, "step": 44930 }, { "epoch": 13.82, "learning_rate": 4.604342264778939e-06, "loss": 0.0297, "step": 44931 }, { "epoch": 13.83, "learning_rate": 4.603922756272823e-06, "loss": 0.0095, "step": 44932 }, { "epoch": 13.83, "learning_rate": 4.60350326116371e-06, "loss": 0.0201, "step": 44933 }, { "epoch": 13.83, "learning_rate": 4.603083779452645e-06, "loss": 0.0184, "step": 44934 }, { "epoch": 13.83, "learning_rate": 4.6026643111406666e-06, "loss": 0.0196, "step": 44935 }, { "epoch": 13.83, "learning_rate": 4.602244856228819e-06, "loss": 0.0123, "step": 44936 }, { "epoch": 13.83, "learning_rate": 4.601825414718137e-06, "loss": 0.0348, "step": 44937 }, { "epoch": 13.83, "learning_rate": 4.601405986609673e-06, "loss": 0.0164, "step": 44938 }, { "epoch": 13.83, "learning_rate": 4.600986571904461e-06, "loss": 0.0207, "step": 44939 }, { "epoch": 13.83, "learning_rate": 4.600567170603548e-06, "loss": 0.0167, "step": 44940 }, { "epoch": 13.83, "learning_rate": 4.600147782707966e-06, "loss": 0.0274, "step": 44941 }, { "epoch": 13.83, "learning_rate": 4.59972840821876e-06, "loss": 0.0388, "step": 44942 }, { "epoch": 13.83, "learning_rate": 4.599309047136975e-06, "loss": 0.0256, "step": 44943 }, { "epoch": 13.83, "learning_rate": 4.59888969946365e-06, "loss": 0.017, "step": 44944 }, { "epoch": 13.83, "learning_rate": 4.598470365199825e-06, "loss": 0.026, "step": 44945 }, { "epoch": 13.83, "learning_rate": 4.598051044346542e-06, "loss": 0.0286, "step": 44946 }, { "epoch": 13.83, "learning_rate": 4.597631736904843e-06, "loss": 0.0163, "step": 44947 }, { "epoch": 13.83, "learning_rate": 4.597212442875767e-06, "loss": 0.0309, "step": 44948 }, { "epoch": 13.83, "learning_rate": 4.596793162260356e-06, "loss": 0.0182, "step": 44949 }, { "epoch": 13.83, "learning_rate": 4.596373895059652e-06, "loss": 0.044, "step": 44950 }, { "epoch": 13.83, "learning_rate": 4.5959546412746934e-06, "loss": 0.0139, "step": 44951 }, { "epoch": 13.83, "learning_rate": 4.595535400906519e-06, "loss": 0.0161, "step": 44952 }, { "epoch": 13.83, "learning_rate": 4.595116173956181e-06, "loss": 0.0228, "step": 44953 }, { "epoch": 13.83, "learning_rate": 4.594696960424709e-06, "loss": 0.0253, "step": 44954 }, { "epoch": 13.83, "learning_rate": 4.594277760313147e-06, "loss": 0.0298, "step": 44955 }, { "epoch": 13.83, "learning_rate": 4.593858573622536e-06, "loss": 0.0244, "step": 44956 }, { "epoch": 13.83, "learning_rate": 4.593439400353913e-06, "loss": 0.0374, "step": 44957 }, { "epoch": 13.83, "learning_rate": 4.593020240508326e-06, "loss": 0.0255, "step": 44958 }, { "epoch": 13.83, "learning_rate": 4.592601094086811e-06, "loss": 0.0175, "step": 44959 }, { "epoch": 13.83, "learning_rate": 4.592181961090411e-06, "loss": 0.0221, "step": 44960 }, { "epoch": 13.83, "learning_rate": 4.59176284152017e-06, "loss": 0.022, "step": 44961 }, { "epoch": 13.83, "learning_rate": 4.5913437353771155e-06, "loss": 0.018, "step": 44962 }, { "epoch": 13.83, "learning_rate": 4.590924642662299e-06, "loss": 0.0143, "step": 44963 }, { "epoch": 13.83, "learning_rate": 4.590505563376759e-06, "loss": 0.0184, "step": 44964 }, { "epoch": 13.84, "learning_rate": 4.590086497521535e-06, "loss": 0.0283, "step": 44965 }, { "epoch": 13.84, "learning_rate": 4.589667445097669e-06, "loss": 0.0134, "step": 44966 }, { "epoch": 13.84, "learning_rate": 4.589248406106199e-06, "loss": 0.0174, "step": 44967 }, { "epoch": 13.84, "learning_rate": 4.588829380548168e-06, "loss": 0.0116, "step": 44968 }, { "epoch": 13.84, "learning_rate": 4.588410368424613e-06, "loss": 0.0168, "step": 44969 }, { "epoch": 13.84, "learning_rate": 4.587991369736578e-06, "loss": 0.0285, "step": 44970 }, { "epoch": 13.84, "learning_rate": 4.5875723844851e-06, "loss": 0.0198, "step": 44971 }, { "epoch": 13.84, "learning_rate": 4.587153412671217e-06, "loss": 0.0249, "step": 44972 }, { "epoch": 13.84, "learning_rate": 4.5867344542959765e-06, "loss": 0.0211, "step": 44973 }, { "epoch": 13.84, "learning_rate": 4.586315509360419e-06, "loss": 0.025, "step": 44974 }, { "epoch": 13.84, "learning_rate": 4.585896577865577e-06, "loss": 0.0148, "step": 44975 }, { "epoch": 13.84, "learning_rate": 4.585477659812494e-06, "loss": 0.0208, "step": 44976 }, { "epoch": 13.84, "learning_rate": 4.585058755202206e-06, "loss": 0.0115, "step": 44977 }, { "epoch": 13.84, "learning_rate": 4.5846398640357625e-06, "loss": 0.0277, "step": 44978 }, { "epoch": 13.84, "learning_rate": 4.584220986314197e-06, "loss": 0.0555, "step": 44979 }, { "epoch": 13.84, "learning_rate": 4.583802122038551e-06, "loss": 0.0304, "step": 44980 }, { "epoch": 13.84, "learning_rate": 4.583383271209864e-06, "loss": 0.03, "step": 44981 }, { "epoch": 13.84, "learning_rate": 4.582964433829177e-06, "loss": 0.0213, "step": 44982 }, { "epoch": 13.84, "learning_rate": 4.582545609897528e-06, "loss": 0.0147, "step": 44983 }, { "epoch": 13.84, "learning_rate": 4.582126799415958e-06, "loss": 0.0204, "step": 44984 }, { "epoch": 13.84, "learning_rate": 4.581708002385506e-06, "loss": 0.0127, "step": 44985 }, { "epoch": 13.84, "learning_rate": 4.581289218807213e-06, "loss": 0.0232, "step": 44986 }, { "epoch": 13.84, "learning_rate": 4.580870448682114e-06, "loss": 0.0183, "step": 44987 }, { "epoch": 13.84, "learning_rate": 4.5804516920112606e-06, "loss": 0.0296, "step": 44988 }, { "epoch": 13.84, "learning_rate": 4.580032948795681e-06, "loss": 0.0131, "step": 44989 }, { "epoch": 13.84, "learning_rate": 4.5796142190364175e-06, "loss": 0.0198, "step": 44990 }, { "epoch": 13.84, "learning_rate": 4.5791955027345115e-06, "loss": 0.0242, "step": 44991 }, { "epoch": 13.84, "learning_rate": 4.5787767998909985e-06, "loss": 0.0176, "step": 44992 }, { "epoch": 13.84, "learning_rate": 4.578358110506923e-06, "loss": 0.0186, "step": 44993 }, { "epoch": 13.84, "learning_rate": 4.577939434583324e-06, "loss": 0.0159, "step": 44994 }, { "epoch": 13.84, "learning_rate": 4.577520772121244e-06, "loss": 0.0103, "step": 44995 }, { "epoch": 13.84, "learning_rate": 4.577102123121713e-06, "loss": 0.0215, "step": 44996 }, { "epoch": 13.85, "learning_rate": 4.576683487585771e-06, "loss": 0.0095, "step": 44997 }, { "epoch": 13.85, "learning_rate": 4.576264865514467e-06, "loss": 0.0149, "step": 44998 }, { "epoch": 13.85, "learning_rate": 4.5758462569088345e-06, "loss": 0.0247, "step": 44999 }, { "epoch": 13.85, "learning_rate": 4.5754276617699145e-06, "loss": 0.0147, "step": 45000 }, { "epoch": 13.85, "learning_rate": 4.575009080098745e-06, "loss": 0.0137, "step": 45001 }, { "epoch": 13.85, "learning_rate": 4.574590511896364e-06, "loss": 0.0161, "step": 45002 }, { "epoch": 13.85, "learning_rate": 4.574171957163814e-06, "loss": 0.0099, "step": 45003 }, { "epoch": 13.85, "learning_rate": 4.57375341590213e-06, "loss": 0.015, "step": 45004 }, { "epoch": 13.85, "learning_rate": 4.573334888112355e-06, "loss": 0.0244, "step": 45005 }, { "epoch": 13.85, "learning_rate": 4.5729163737955264e-06, "loss": 0.0286, "step": 45006 }, { "epoch": 13.85, "learning_rate": 4.572497872952679e-06, "loss": 0.028, "step": 45007 }, { "epoch": 13.85, "learning_rate": 4.57207938558486e-06, "loss": 0.0207, "step": 45008 }, { "epoch": 13.85, "learning_rate": 4.571660911693108e-06, "loss": 0.0142, "step": 45009 }, { "epoch": 13.85, "learning_rate": 4.571242451278456e-06, "loss": 0.0269, "step": 45010 }, { "epoch": 13.85, "learning_rate": 4.570824004341945e-06, "loss": 0.0178, "step": 45011 }, { "epoch": 13.85, "learning_rate": 4.57040557088461e-06, "loss": 0.0069, "step": 45012 }, { "epoch": 13.85, "learning_rate": 4.569987150907499e-06, "loss": 0.0167, "step": 45013 }, { "epoch": 13.85, "learning_rate": 4.569568744411644e-06, "loss": 0.0239, "step": 45014 }, { "epoch": 13.85, "learning_rate": 4.569150351398089e-06, "loss": 0.0319, "step": 45015 }, { "epoch": 13.85, "learning_rate": 4.56873197186787e-06, "loss": 0.0204, "step": 45016 }, { "epoch": 13.85, "learning_rate": 4.56831360582202e-06, "loss": 0.021, "step": 45017 }, { "epoch": 13.85, "learning_rate": 4.567895253261585e-06, "loss": 0.0153, "step": 45018 }, { "epoch": 13.85, "learning_rate": 4.567476914187603e-06, "loss": 0.0492, "step": 45019 }, { "epoch": 13.85, "learning_rate": 4.567058588601111e-06, "loss": 0.0114, "step": 45020 }, { "epoch": 13.85, "learning_rate": 4.566640276503148e-06, "loss": 0.0295, "step": 45021 }, { "epoch": 13.85, "learning_rate": 4.566221977894751e-06, "loss": 0.0117, "step": 45022 }, { "epoch": 13.85, "learning_rate": 4.565803692776962e-06, "loss": 0.0256, "step": 45023 }, { "epoch": 13.85, "learning_rate": 4.565385421150817e-06, "loss": 0.0119, "step": 45024 }, { "epoch": 13.85, "learning_rate": 4.564967163017354e-06, "loss": 0.0241, "step": 45025 }, { "epoch": 13.85, "learning_rate": 4.564548918377612e-06, "loss": 0.0115, "step": 45026 }, { "epoch": 13.85, "learning_rate": 4.5641306872326275e-06, "loss": 0.0149, "step": 45027 }, { "epoch": 13.85, "learning_rate": 4.563712469583443e-06, "loss": 0.0141, "step": 45028 }, { "epoch": 13.85, "learning_rate": 4.563294265431096e-06, "loss": 0.0182, "step": 45029 }, { "epoch": 13.86, "learning_rate": 4.562876074776627e-06, "loss": 0.0216, "step": 45030 }, { "epoch": 13.86, "learning_rate": 4.562457897621067e-06, "loss": 0.0296, "step": 45031 }, { "epoch": 13.86, "learning_rate": 4.5620397339654555e-06, "loss": 0.0196, "step": 45032 }, { "epoch": 13.86, "learning_rate": 4.5616215838108355e-06, "loss": 0.0118, "step": 45033 }, { "epoch": 13.86, "learning_rate": 4.5612034471582435e-06, "loss": 0.0232, "step": 45034 }, { "epoch": 13.86, "learning_rate": 4.560785324008718e-06, "loss": 0.0294, "step": 45035 }, { "epoch": 13.86, "learning_rate": 4.560367214363295e-06, "loss": 0.0215, "step": 45036 }, { "epoch": 13.86, "learning_rate": 4.5599491182230134e-06, "loss": 0.0162, "step": 45037 }, { "epoch": 13.86, "learning_rate": 4.559531035588913e-06, "loss": 0.0145, "step": 45038 }, { "epoch": 13.86, "learning_rate": 4.559112966462029e-06, "loss": 0.0173, "step": 45039 }, { "epoch": 13.86, "learning_rate": 4.558694910843402e-06, "loss": 0.0162, "step": 45040 }, { "epoch": 13.86, "learning_rate": 4.558276868734068e-06, "loss": 0.0162, "step": 45041 }, { "epoch": 13.86, "learning_rate": 4.557858840135062e-06, "loss": 0.0088, "step": 45042 }, { "epoch": 13.86, "learning_rate": 4.557440825047432e-06, "loss": 0.0313, "step": 45043 }, { "epoch": 13.86, "learning_rate": 4.557022823472206e-06, "loss": 0.0342, "step": 45044 }, { "epoch": 13.86, "learning_rate": 4.556604835410425e-06, "loss": 0.0298, "step": 45045 }, { "epoch": 13.86, "learning_rate": 4.5561868608631264e-06, "loss": 0.0189, "step": 45046 }, { "epoch": 13.86, "learning_rate": 4.555768899831345e-06, "loss": 0.016, "step": 45047 }, { "epoch": 13.86, "learning_rate": 4.555350952316124e-06, "loss": 0.0156, "step": 45048 }, { "epoch": 13.86, "learning_rate": 4.5549330183185e-06, "loss": 0.0267, "step": 45049 }, { "epoch": 13.86, "learning_rate": 4.554515097839511e-06, "loss": 0.0218, "step": 45050 }, { "epoch": 13.86, "learning_rate": 4.554097190880194e-06, "loss": 0.0133, "step": 45051 }, { "epoch": 13.86, "learning_rate": 4.5536792974415785e-06, "loss": 0.0201, "step": 45052 }, { "epoch": 13.86, "learning_rate": 4.553261417524713e-06, "loss": 0.0391, "step": 45053 }, { "epoch": 13.86, "learning_rate": 4.552843551130631e-06, "loss": 0.0126, "step": 45054 }, { "epoch": 13.86, "learning_rate": 4.552425698260371e-06, "loss": 0.0119, "step": 45055 }, { "epoch": 13.86, "learning_rate": 4.5520078589149685e-06, "loss": 0.0147, "step": 45056 }, { "epoch": 13.86, "learning_rate": 4.551590033095461e-06, "loss": 0.0206, "step": 45057 }, { "epoch": 13.86, "learning_rate": 4.5511722208028875e-06, "loss": 0.0233, "step": 45058 }, { "epoch": 13.86, "learning_rate": 4.550754422038285e-06, "loss": 0.0328, "step": 45059 }, { "epoch": 13.86, "learning_rate": 4.550336636802689e-06, "loss": 0.0232, "step": 45060 }, { "epoch": 13.86, "learning_rate": 4.549918865097138e-06, "loss": 0.0292, "step": 45061 }, { "epoch": 13.87, "learning_rate": 4.549501106922667e-06, "loss": 0.0434, "step": 45062 }, { "epoch": 13.87, "learning_rate": 4.549083362280318e-06, "loss": 0.0207, "step": 45063 }, { "epoch": 13.87, "learning_rate": 4.548665631171128e-06, "loss": 0.0233, "step": 45064 }, { "epoch": 13.87, "learning_rate": 4.54824791359613e-06, "loss": 0.0284, "step": 45065 }, { "epoch": 13.87, "learning_rate": 4.5478302095563606e-06, "loss": 0.0116, "step": 45066 }, { "epoch": 13.87, "learning_rate": 4.5474125190528575e-06, "loss": 0.0109, "step": 45067 }, { "epoch": 13.87, "learning_rate": 4.546994842086662e-06, "loss": 0.0179, "step": 45068 }, { "epoch": 13.87, "learning_rate": 4.546577178658808e-06, "loss": 0.0336, "step": 45069 }, { "epoch": 13.87, "learning_rate": 4.546159528770332e-06, "loss": 0.0092, "step": 45070 }, { "epoch": 13.87, "learning_rate": 4.545741892422272e-06, "loss": 0.017, "step": 45071 }, { "epoch": 13.87, "learning_rate": 4.545324269615664e-06, "loss": 0.0232, "step": 45072 }, { "epoch": 13.87, "learning_rate": 4.544906660351547e-06, "loss": 0.0184, "step": 45073 }, { "epoch": 13.87, "learning_rate": 4.544489064630954e-06, "loss": 0.0137, "step": 45074 }, { "epoch": 13.87, "learning_rate": 4.544071482454926e-06, "loss": 0.0259, "step": 45075 }, { "epoch": 13.87, "learning_rate": 4.543653913824496e-06, "loss": 0.0195, "step": 45076 }, { "epoch": 13.87, "learning_rate": 4.543236358740702e-06, "loss": 0.0284, "step": 45077 }, { "epoch": 13.87, "learning_rate": 4.542818817204583e-06, "loss": 0.0223, "step": 45078 }, { "epoch": 13.87, "learning_rate": 4.542401289217172e-06, "loss": 0.022, "step": 45079 }, { "epoch": 13.87, "learning_rate": 4.541983774779508e-06, "loss": 0.0485, "step": 45080 }, { "epoch": 13.87, "learning_rate": 4.541566273892627e-06, "loss": 0.0197, "step": 45081 }, { "epoch": 13.87, "learning_rate": 4.541148786557562e-06, "loss": 0.0198, "step": 45082 }, { "epoch": 13.87, "learning_rate": 4.540731312775356e-06, "loss": 0.0177, "step": 45083 }, { "epoch": 13.87, "learning_rate": 4.540313852547042e-06, "loss": 0.0287, "step": 45084 }, { "epoch": 13.87, "learning_rate": 4.53989640587366e-06, "loss": 0.0108, "step": 45085 }, { "epoch": 13.87, "learning_rate": 4.539478972756242e-06, "loss": 0.0362, "step": 45086 }, { "epoch": 13.87, "learning_rate": 4.53906155319582e-06, "loss": 0.0207, "step": 45087 }, { "epoch": 13.87, "learning_rate": 4.53864414719344e-06, "loss": 0.0115, "step": 45088 }, { "epoch": 13.87, "learning_rate": 4.5382267547501335e-06, "loss": 0.0144, "step": 45089 }, { "epoch": 13.87, "learning_rate": 4.537809375866938e-06, "loss": 0.0188, "step": 45090 }, { "epoch": 13.87, "learning_rate": 4.53739201054489e-06, "loss": 0.022, "step": 45091 }, { "epoch": 13.87, "learning_rate": 4.536974658785025e-06, "loss": 0.0161, "step": 45092 }, { "epoch": 13.87, "learning_rate": 4.536557320588378e-06, "loss": 0.02, "step": 45093 }, { "epoch": 13.87, "learning_rate": 4.536139995955987e-06, "loss": 0.024, "step": 45094 }, { "epoch": 13.88, "learning_rate": 4.535722684888888e-06, "loss": 0.0153, "step": 45095 }, { "epoch": 13.88, "learning_rate": 4.5353053873881145e-06, "loss": 0.0211, "step": 45096 }, { "epoch": 13.88, "learning_rate": 4.5348881034547035e-06, "loss": 0.0195, "step": 45097 }, { "epoch": 13.88, "learning_rate": 4.534470833089697e-06, "loss": 0.0167, "step": 45098 }, { "epoch": 13.88, "learning_rate": 4.5340535762941225e-06, "loss": 0.018, "step": 45099 }, { "epoch": 13.88, "learning_rate": 4.53363633306902e-06, "loss": 0.0168, "step": 45100 }, { "epoch": 13.88, "learning_rate": 4.533219103415426e-06, "loss": 0.0163, "step": 45101 }, { "epoch": 13.88, "learning_rate": 4.53280188733437e-06, "loss": 0.012, "step": 45102 }, { "epoch": 13.88, "learning_rate": 4.532384684826897e-06, "loss": 0.0099, "step": 45103 }, { "epoch": 13.88, "learning_rate": 4.5319674958940385e-06, "loss": 0.0284, "step": 45104 }, { "epoch": 13.88, "learning_rate": 4.531550320536832e-06, "loss": 0.0214, "step": 45105 }, { "epoch": 13.88, "learning_rate": 4.531133158756313e-06, "loss": 0.0169, "step": 45106 }, { "epoch": 13.88, "learning_rate": 4.5307160105535106e-06, "loss": 0.0294, "step": 45107 }, { "epoch": 13.88, "learning_rate": 4.530298875929469e-06, "loss": 0.0306, "step": 45108 }, { "epoch": 13.88, "learning_rate": 4.529881754885221e-06, "loss": 0.0158, "step": 45109 }, { "epoch": 13.88, "learning_rate": 4.529464647421802e-06, "loss": 0.0172, "step": 45110 }, { "epoch": 13.88, "learning_rate": 4.529047553540247e-06, "loss": 0.0228, "step": 45111 }, { "epoch": 13.88, "learning_rate": 4.528630473241591e-06, "loss": 0.0126, "step": 45112 }, { "epoch": 13.88, "learning_rate": 4.528213406526872e-06, "loss": 0.0244, "step": 45113 }, { "epoch": 13.88, "learning_rate": 4.527796353397124e-06, "loss": 0.033, "step": 45114 }, { "epoch": 13.88, "learning_rate": 4.527379313853381e-06, "loss": 0.0192, "step": 45115 }, { "epoch": 13.88, "learning_rate": 4.526962287896681e-06, "loss": 0.0124, "step": 45116 }, { "epoch": 13.88, "learning_rate": 4.5265452755280545e-06, "loss": 0.0213, "step": 45117 }, { "epoch": 13.88, "learning_rate": 4.526128276748544e-06, "loss": 0.0156, "step": 45118 }, { "epoch": 13.88, "learning_rate": 4.525711291559185e-06, "loss": 0.0221, "step": 45119 }, { "epoch": 13.88, "learning_rate": 4.5252943199610046e-06, "loss": 0.0243, "step": 45120 }, { "epoch": 13.88, "learning_rate": 4.524877361955039e-06, "loss": 0.0198, "step": 45121 }, { "epoch": 13.88, "learning_rate": 4.524460417542331e-06, "loss": 0.0173, "step": 45122 }, { "epoch": 13.88, "learning_rate": 4.524043486723911e-06, "loss": 0.0108, "step": 45123 }, { "epoch": 13.88, "learning_rate": 4.523626569500816e-06, "loss": 0.0247, "step": 45124 }, { "epoch": 13.88, "learning_rate": 4.523209665874079e-06, "loss": 0.021, "step": 45125 }, { "epoch": 13.88, "learning_rate": 4.5227927758447365e-06, "loss": 0.0293, "step": 45126 }, { "epoch": 13.89, "learning_rate": 4.522375899413822e-06, "loss": 0.0393, "step": 45127 }, { "epoch": 13.89, "learning_rate": 4.521959036582372e-06, "loss": 0.0275, "step": 45128 }, { "epoch": 13.89, "learning_rate": 4.521542187351421e-06, "loss": 0.0161, "step": 45129 }, { "epoch": 13.89, "learning_rate": 4.521125351722003e-06, "loss": 0.0185, "step": 45130 }, { "epoch": 13.89, "learning_rate": 4.52070852969515e-06, "loss": 0.021, "step": 45131 }, { "epoch": 13.89, "learning_rate": 4.520291721271909e-06, "loss": 0.0245, "step": 45132 }, { "epoch": 13.89, "learning_rate": 4.519874926453303e-06, "loss": 0.0109, "step": 45133 }, { "epoch": 13.89, "learning_rate": 4.519458145240368e-06, "loss": 0.0083, "step": 45134 }, { "epoch": 13.89, "learning_rate": 4.519041377634142e-06, "loss": 0.0256, "step": 45135 }, { "epoch": 13.89, "learning_rate": 4.518624623635655e-06, "loss": 0.0171, "step": 45136 }, { "epoch": 13.89, "learning_rate": 4.5182078832459505e-06, "loss": 0.0213, "step": 45137 }, { "epoch": 13.89, "learning_rate": 4.517791156466057e-06, "loss": 0.023, "step": 45138 }, { "epoch": 13.89, "learning_rate": 4.5173744432970094e-06, "loss": 0.0187, "step": 45139 }, { "epoch": 13.89, "learning_rate": 4.516957743739847e-06, "loss": 0.0164, "step": 45140 }, { "epoch": 13.89, "learning_rate": 4.516541057795593e-06, "loss": 0.0239, "step": 45141 }, { "epoch": 13.89, "learning_rate": 4.516124385465293e-06, "loss": 0.0125, "step": 45142 }, { "epoch": 13.89, "learning_rate": 4.5157077267499775e-06, "loss": 0.0242, "step": 45143 }, { "epoch": 13.89, "learning_rate": 4.515291081650681e-06, "loss": 0.0141, "step": 45144 }, { "epoch": 13.89, "learning_rate": 4.514874450168439e-06, "loss": 0.0224, "step": 45145 }, { "epoch": 13.89, "learning_rate": 4.514457832304284e-06, "loss": 0.03, "step": 45146 }, { "epoch": 13.89, "learning_rate": 4.514041228059253e-06, "loss": 0.0265, "step": 45147 }, { "epoch": 13.89, "learning_rate": 4.513624637434377e-06, "loss": 0.0095, "step": 45148 }, { "epoch": 13.89, "learning_rate": 4.513208060430693e-06, "loss": 0.0274, "step": 45149 }, { "epoch": 13.89, "learning_rate": 4.512791497049232e-06, "loss": 0.0175, "step": 45150 }, { "epoch": 13.89, "learning_rate": 4.5123749472910296e-06, "loss": 0.0156, "step": 45151 }, { "epoch": 13.89, "learning_rate": 4.511958411157123e-06, "loss": 0.0193, "step": 45152 }, { "epoch": 13.89, "learning_rate": 4.511541888648549e-06, "loss": 0.0224, "step": 45153 }, { "epoch": 13.89, "learning_rate": 4.511125379766332e-06, "loss": 0.0233, "step": 45154 }, { "epoch": 13.89, "learning_rate": 4.510708884511511e-06, "loss": 0.0175, "step": 45155 }, { "epoch": 13.89, "learning_rate": 4.5102924028851156e-06, "loss": 0.0185, "step": 45156 }, { "epoch": 13.89, "learning_rate": 4.509875934888189e-06, "loss": 0.0184, "step": 45157 }, { "epoch": 13.89, "learning_rate": 4.509459480521761e-06, "loss": 0.0317, "step": 45158 }, { "epoch": 13.89, "learning_rate": 4.5090430397868634e-06, "loss": 0.0379, "step": 45159 }, { "epoch": 13.9, "learning_rate": 4.508626612684532e-06, "loss": 0.0141, "step": 45160 }, { "epoch": 13.9, "learning_rate": 4.5082101992158e-06, "loss": 0.0283, "step": 45161 }, { "epoch": 13.9, "learning_rate": 4.507793799381702e-06, "loss": 0.0359, "step": 45162 }, { "epoch": 13.9, "learning_rate": 4.507377413183271e-06, "loss": 0.0766, "step": 45163 }, { "epoch": 13.9, "learning_rate": 4.506961040621542e-06, "loss": 0.0288, "step": 45164 }, { "epoch": 13.9, "learning_rate": 4.5065446816975475e-06, "loss": 0.012, "step": 45165 }, { "epoch": 13.9, "learning_rate": 4.506128336412317e-06, "loss": 0.0203, "step": 45166 }, { "epoch": 13.9, "learning_rate": 4.5057120047668975e-06, "loss": 0.0257, "step": 45167 }, { "epoch": 13.9, "learning_rate": 4.50529568676231e-06, "loss": 0.0238, "step": 45168 }, { "epoch": 13.9, "learning_rate": 4.504879382399592e-06, "loss": 0.0394, "step": 45169 }, { "epoch": 13.9, "learning_rate": 4.504463091679776e-06, "loss": 0.0307, "step": 45170 }, { "epoch": 13.9, "learning_rate": 4.504046814603893e-06, "loss": 0.0216, "step": 45171 }, { "epoch": 13.9, "learning_rate": 4.503630551172985e-06, "loss": 0.0287, "step": 45172 }, { "epoch": 13.9, "learning_rate": 4.503214301388079e-06, "loss": 0.0203, "step": 45173 }, { "epoch": 13.9, "learning_rate": 4.502798065250215e-06, "loss": 0.018, "step": 45174 }, { "epoch": 13.9, "learning_rate": 4.502381842760417e-06, "loss": 0.0276, "step": 45175 }, { "epoch": 13.9, "learning_rate": 4.501965633919719e-06, "loss": 0.0216, "step": 45176 }, { "epoch": 13.9, "learning_rate": 4.501549438729162e-06, "loss": 0.017, "step": 45177 }, { "epoch": 13.9, "learning_rate": 4.501133257189776e-06, "loss": 0.0245, "step": 45178 }, { "epoch": 13.9, "learning_rate": 4.500717089302593e-06, "loss": 0.0249, "step": 45179 }, { "epoch": 13.9, "learning_rate": 4.500300935068647e-06, "loss": 0.0216, "step": 45180 }, { "epoch": 13.9, "learning_rate": 4.499884794488971e-06, "loss": 0.0082, "step": 45181 }, { "epoch": 13.9, "learning_rate": 4.499468667564599e-06, "loss": 0.0321, "step": 45182 }, { "epoch": 13.9, "learning_rate": 4.4990525542965635e-06, "loss": 0.0106, "step": 45183 }, { "epoch": 13.9, "learning_rate": 4.498636454685897e-06, "loss": 0.0176, "step": 45184 }, { "epoch": 13.9, "learning_rate": 4.498220368733632e-06, "loss": 0.0151, "step": 45185 }, { "epoch": 13.9, "learning_rate": 4.4978042964408e-06, "loss": 0.0194, "step": 45186 }, { "epoch": 13.9, "learning_rate": 4.4973882378084395e-06, "loss": 0.0149, "step": 45187 }, { "epoch": 13.9, "learning_rate": 4.496972192837585e-06, "loss": 0.0165, "step": 45188 }, { "epoch": 13.9, "learning_rate": 4.496556161529262e-06, "loss": 0.0248, "step": 45189 }, { "epoch": 13.9, "learning_rate": 4.496140143884505e-06, "loss": 0.0298, "step": 45190 }, { "epoch": 13.9, "learning_rate": 4.495724139904345e-06, "loss": 0.0198, "step": 45191 }, { "epoch": 13.91, "learning_rate": 4.495308149589822e-06, "loss": 0.0201, "step": 45192 }, { "epoch": 13.91, "learning_rate": 4.494892172941965e-06, "loss": 0.0229, "step": 45193 }, { "epoch": 13.91, "learning_rate": 4.494476209961805e-06, "loss": 0.0219, "step": 45194 }, { "epoch": 13.91, "learning_rate": 4.4940602606503825e-06, "loss": 0.0214, "step": 45195 }, { "epoch": 13.91, "learning_rate": 4.4936443250087145e-06, "loss": 0.007, "step": 45196 }, { "epoch": 13.91, "learning_rate": 4.493228403037848e-06, "loss": 0.0167, "step": 45197 }, { "epoch": 13.91, "learning_rate": 4.492812494738811e-06, "loss": 0.0324, "step": 45198 }, { "epoch": 13.91, "learning_rate": 4.492396600112635e-06, "loss": 0.0243, "step": 45199 }, { "epoch": 13.91, "learning_rate": 4.491980719160354e-06, "loss": 0.0201, "step": 45200 }, { "epoch": 13.91, "learning_rate": 4.4915648518830005e-06, "loss": 0.018, "step": 45201 }, { "epoch": 13.91, "learning_rate": 4.491148998281605e-06, "loss": 0.016, "step": 45202 }, { "epoch": 13.91, "learning_rate": 4.4907331583572025e-06, "loss": 0.0265, "step": 45203 }, { "epoch": 13.91, "learning_rate": 4.490317332110824e-06, "loss": 0.0176, "step": 45204 }, { "epoch": 13.91, "learning_rate": 4.489901519543502e-06, "loss": 0.0148, "step": 45205 }, { "epoch": 13.91, "learning_rate": 4.489485720656266e-06, "loss": 0.012, "step": 45206 }, { "epoch": 13.91, "learning_rate": 4.489069935450155e-06, "loss": 0.0184, "step": 45207 }, { "epoch": 13.91, "learning_rate": 4.488654163926196e-06, "loss": 0.0196, "step": 45208 }, { "epoch": 13.91, "learning_rate": 4.488238406085428e-06, "loss": 0.0161, "step": 45209 }, { "epoch": 13.91, "learning_rate": 4.4878226619288745e-06, "loss": 0.0251, "step": 45210 }, { "epoch": 13.91, "learning_rate": 4.4874069314575675e-06, "loss": 0.0164, "step": 45211 }, { "epoch": 13.91, "learning_rate": 4.486991214672546e-06, "loss": 0.0105, "step": 45212 }, { "epoch": 13.91, "learning_rate": 4.48657551157484e-06, "loss": 0.0265, "step": 45213 }, { "epoch": 13.91, "learning_rate": 4.48615982216548e-06, "loss": 0.0255, "step": 45214 }, { "epoch": 13.91, "learning_rate": 4.485744146445499e-06, "loss": 0.0167, "step": 45215 }, { "epoch": 13.91, "learning_rate": 4.485328484415928e-06, "loss": 0.0232, "step": 45216 }, { "epoch": 13.91, "learning_rate": 4.4849128360778e-06, "loss": 0.038, "step": 45217 }, { "epoch": 13.91, "learning_rate": 4.484497201432147e-06, "loss": 0.016, "step": 45218 }, { "epoch": 13.91, "learning_rate": 4.4840815804800005e-06, "loss": 0.0156, "step": 45219 }, { "epoch": 13.91, "learning_rate": 4.483665973222392e-06, "loss": 0.0134, "step": 45220 }, { "epoch": 13.91, "learning_rate": 4.483250379660351e-06, "loss": 0.0271, "step": 45221 }, { "epoch": 13.91, "learning_rate": 4.482834799794919e-06, "loss": 0.0072, "step": 45222 }, { "epoch": 13.91, "learning_rate": 4.482419233627116e-06, "loss": 0.0281, "step": 45223 }, { "epoch": 13.91, "learning_rate": 4.482003681157979e-06, "loss": 0.0409, "step": 45224 }, { "epoch": 13.92, "learning_rate": 4.48158814238854e-06, "loss": 0.0243, "step": 45225 }, { "epoch": 13.92, "learning_rate": 4.481172617319826e-06, "loss": 0.0075, "step": 45226 }, { "epoch": 13.92, "learning_rate": 4.480757105952877e-06, "loss": 0.0163, "step": 45227 }, { "epoch": 13.92, "learning_rate": 4.480341608288718e-06, "loss": 0.0296, "step": 45228 }, { "epoch": 13.92, "learning_rate": 4.4799261243283875e-06, "loss": 0.0246, "step": 45229 }, { "epoch": 13.92, "learning_rate": 4.479510654072909e-06, "loss": 0.011, "step": 45230 }, { "epoch": 13.92, "learning_rate": 4.479095197523313e-06, "loss": 0.0218, "step": 45231 }, { "epoch": 13.92, "learning_rate": 4.478679754680639e-06, "loss": 0.0131, "step": 45232 }, { "epoch": 13.92, "learning_rate": 4.478264325545915e-06, "loss": 0.0282, "step": 45233 }, { "epoch": 13.92, "learning_rate": 4.477848910120173e-06, "loss": 0.035, "step": 45234 }, { "epoch": 13.92, "learning_rate": 4.477433508404442e-06, "loss": 0.018, "step": 45235 }, { "epoch": 13.92, "learning_rate": 4.477018120399756e-06, "loss": 0.0214, "step": 45236 }, { "epoch": 13.92, "learning_rate": 4.476602746107144e-06, "loss": 0.0257, "step": 45237 }, { "epoch": 13.92, "learning_rate": 4.476187385527639e-06, "loss": 0.0319, "step": 45238 }, { "epoch": 13.92, "learning_rate": 4.4757720386622705e-06, "loss": 0.0141, "step": 45239 }, { "epoch": 13.92, "learning_rate": 4.475356705512071e-06, "loss": 0.0131, "step": 45240 }, { "epoch": 13.92, "learning_rate": 4.474941386078068e-06, "loss": 0.0193, "step": 45241 }, { "epoch": 13.92, "learning_rate": 4.4745260803613e-06, "loss": 0.025, "step": 45242 }, { "epoch": 13.92, "learning_rate": 4.474110788362798e-06, "loss": 0.0174, "step": 45243 }, { "epoch": 13.92, "learning_rate": 4.4736955100835844e-06, "loss": 0.0175, "step": 45244 }, { "epoch": 13.92, "learning_rate": 4.473280245524696e-06, "loss": 0.032, "step": 45245 }, { "epoch": 13.92, "learning_rate": 4.472864994687159e-06, "loss": 0.0124, "step": 45246 }, { "epoch": 13.92, "learning_rate": 4.47244975757201e-06, "loss": 0.0294, "step": 45247 }, { "epoch": 13.92, "learning_rate": 4.472034534180281e-06, "loss": 0.0147, "step": 45248 }, { "epoch": 13.92, "learning_rate": 4.471619324512998e-06, "loss": 0.0208, "step": 45249 }, { "epoch": 13.92, "learning_rate": 4.471204128571199e-06, "loss": 0.0188, "step": 45250 }, { "epoch": 13.92, "learning_rate": 4.4707889463559e-06, "loss": 0.018, "step": 45251 }, { "epoch": 13.92, "learning_rate": 4.4703737778681474e-06, "loss": 0.0253, "step": 45252 }, { "epoch": 13.92, "learning_rate": 4.469958623108966e-06, "loss": 0.0463, "step": 45253 }, { "epoch": 13.92, "learning_rate": 4.469543482079385e-06, "loss": 0.0301, "step": 45254 }, { "epoch": 13.92, "learning_rate": 4.469128354780436e-06, "loss": 0.0176, "step": 45255 }, { "epoch": 13.92, "learning_rate": 4.468713241213152e-06, "loss": 0.0256, "step": 45256 }, { "epoch": 13.93, "learning_rate": 4.468298141378561e-06, "loss": 0.0208, "step": 45257 }, { "epoch": 13.93, "learning_rate": 4.467883055277696e-06, "loss": 0.0161, "step": 45258 }, { "epoch": 13.93, "learning_rate": 4.467467982911583e-06, "loss": 0.0205, "step": 45259 }, { "epoch": 13.93, "learning_rate": 4.467052924281258e-06, "loss": 0.0182, "step": 45260 }, { "epoch": 13.93, "learning_rate": 4.466637879387744e-06, "loss": 0.0254, "step": 45261 }, { "epoch": 13.93, "learning_rate": 4.46622284823208e-06, "loss": 0.0159, "step": 45262 }, { "epoch": 13.93, "learning_rate": 4.465807830815293e-06, "loss": 0.0122, "step": 45263 }, { "epoch": 13.93, "learning_rate": 4.465392827138417e-06, "loss": 0.0275, "step": 45264 }, { "epoch": 13.93, "learning_rate": 4.464977837202475e-06, "loss": 0.0312, "step": 45265 }, { "epoch": 13.93, "learning_rate": 4.464562861008497e-06, "loss": 0.0319, "step": 45266 }, { "epoch": 13.93, "learning_rate": 4.464147898557521e-06, "loss": 0.0219, "step": 45267 }, { "epoch": 13.93, "learning_rate": 4.463732949850574e-06, "loss": 0.0111, "step": 45268 }, { "epoch": 13.93, "learning_rate": 4.463318014888684e-06, "loss": 0.0161, "step": 45269 }, { "epoch": 13.93, "learning_rate": 4.462903093672883e-06, "loss": 0.024, "step": 45270 }, { "epoch": 13.93, "learning_rate": 4.462488186204201e-06, "loss": 0.0276, "step": 45271 }, { "epoch": 13.93, "learning_rate": 4.462073292483669e-06, "loss": 0.0219, "step": 45272 }, { "epoch": 13.93, "learning_rate": 4.4616584125123155e-06, "loss": 0.0279, "step": 45273 }, { "epoch": 13.93, "learning_rate": 4.461243546291171e-06, "loss": 0.0196, "step": 45274 }, { "epoch": 13.93, "learning_rate": 4.460828693821265e-06, "loss": 0.0172, "step": 45275 }, { "epoch": 13.93, "learning_rate": 4.460413855103625e-06, "loss": 0.0193, "step": 45276 }, { "epoch": 13.93, "learning_rate": 4.459999030139292e-06, "loss": 0.0296, "step": 45277 }, { "epoch": 13.93, "learning_rate": 4.459584218929283e-06, "loss": 0.0192, "step": 45278 }, { "epoch": 13.93, "learning_rate": 4.459169421474633e-06, "loss": 0.0204, "step": 45279 }, { "epoch": 13.93, "learning_rate": 4.458754637776371e-06, "loss": 0.0132, "step": 45280 }, { "epoch": 13.93, "learning_rate": 4.458339867835525e-06, "loss": 0.0169, "step": 45281 }, { "epoch": 13.93, "learning_rate": 4.45792511165313e-06, "loss": 0.0099, "step": 45282 }, { "epoch": 13.93, "learning_rate": 4.4575103692302125e-06, "loss": 0.0227, "step": 45283 }, { "epoch": 13.93, "learning_rate": 4.457095640567804e-06, "loss": 0.0213, "step": 45284 }, { "epoch": 13.93, "learning_rate": 4.4566809256669345e-06, "loss": 0.0223, "step": 45285 }, { "epoch": 13.93, "learning_rate": 4.456266224528626e-06, "loss": 0.0187, "step": 45286 }, { "epoch": 13.93, "learning_rate": 4.455851537153917e-06, "loss": 0.0171, "step": 45287 }, { "epoch": 13.93, "learning_rate": 4.455436863543834e-06, "loss": 0.012, "step": 45288 }, { "epoch": 13.93, "learning_rate": 4.455022203699406e-06, "loss": 0.0118, "step": 45289 }, { "epoch": 13.94, "learning_rate": 4.454607557621663e-06, "loss": 0.0255, "step": 45290 }, { "epoch": 13.94, "learning_rate": 4.454192925311635e-06, "loss": 0.0236, "step": 45291 }, { "epoch": 13.94, "learning_rate": 4.453778306770351e-06, "loss": 0.0133, "step": 45292 }, { "epoch": 13.94, "learning_rate": 4.45336370199884e-06, "loss": 0.0141, "step": 45293 }, { "epoch": 13.94, "learning_rate": 4.4529491109981316e-06, "loss": 0.0264, "step": 45294 }, { "epoch": 13.94, "learning_rate": 4.452534533769255e-06, "loss": 0.0164, "step": 45295 }, { "epoch": 13.94, "learning_rate": 4.452119970313236e-06, "loss": 0.0268, "step": 45296 }, { "epoch": 13.94, "learning_rate": 4.451705420631111e-06, "loss": 0.0049, "step": 45297 }, { "epoch": 13.94, "learning_rate": 4.45129088472391e-06, "loss": 0.0159, "step": 45298 }, { "epoch": 13.94, "learning_rate": 4.450876362592653e-06, "loss": 0.0296, "step": 45299 }, { "epoch": 13.94, "learning_rate": 4.4504618542383745e-06, "loss": 0.0133, "step": 45300 }, { "epoch": 13.94, "learning_rate": 4.4500473596621e-06, "loss": 0.0166, "step": 45301 }, { "epoch": 13.94, "learning_rate": 4.449632878864866e-06, "loss": 0.0177, "step": 45302 }, { "epoch": 13.94, "learning_rate": 4.449218411847696e-06, "loss": 0.0244, "step": 45303 }, { "epoch": 13.94, "learning_rate": 4.4488039586116194e-06, "loss": 0.0117, "step": 45304 }, { "epoch": 13.94, "learning_rate": 4.448389519157668e-06, "loss": 0.0347, "step": 45305 }, { "epoch": 13.94, "learning_rate": 4.447975093486867e-06, "loss": 0.0348, "step": 45306 }, { "epoch": 13.94, "learning_rate": 4.447560681600248e-06, "loss": 0.0173, "step": 45307 }, { "epoch": 13.94, "learning_rate": 4.4471462834988385e-06, "loss": 0.0183, "step": 45308 }, { "epoch": 13.94, "learning_rate": 4.446731899183668e-06, "loss": 0.0181, "step": 45309 }, { "epoch": 13.94, "learning_rate": 4.4463175286557654e-06, "loss": 0.021, "step": 45310 }, { "epoch": 13.94, "learning_rate": 4.445903171916159e-06, "loss": 0.0204, "step": 45311 }, { "epoch": 13.94, "learning_rate": 4.445488828965878e-06, "loss": 0.0185, "step": 45312 }, { "epoch": 13.94, "learning_rate": 4.44507449980595e-06, "loss": 0.0127, "step": 45313 }, { "epoch": 13.94, "learning_rate": 4.444660184437405e-06, "loss": 0.0128, "step": 45314 }, { "epoch": 13.94, "learning_rate": 4.4442458828612715e-06, "loss": 0.0215, "step": 45315 }, { "epoch": 13.94, "learning_rate": 4.4438315950785725e-06, "loss": 0.0185, "step": 45316 }, { "epoch": 13.94, "learning_rate": 4.443417321090346e-06, "loss": 0.0159, "step": 45317 }, { "epoch": 13.94, "learning_rate": 4.443003060897617e-06, "loss": 0.0124, "step": 45318 }, { "epoch": 13.94, "learning_rate": 4.442588814501417e-06, "loss": 0.0226, "step": 45319 }, { "epoch": 13.94, "learning_rate": 4.442174581902767e-06, "loss": 0.0347, "step": 45320 }, { "epoch": 13.94, "learning_rate": 4.441760363102695e-06, "loss": 0.0425, "step": 45321 }, { "epoch": 13.95, "learning_rate": 4.441346158102238e-06, "loss": 0.0232, "step": 45322 }, { "epoch": 13.95, "learning_rate": 4.440931966902419e-06, "loss": 0.0198, "step": 45323 }, { "epoch": 13.95, "learning_rate": 4.440517789504268e-06, "loss": 0.0126, "step": 45324 }, { "epoch": 13.95, "learning_rate": 4.440103625908811e-06, "loss": 0.0172, "step": 45325 }, { "epoch": 13.95, "learning_rate": 4.43968947611708e-06, "loss": 0.0272, "step": 45326 }, { "epoch": 13.95, "learning_rate": 4.439275340130099e-06, "loss": 0.0164, "step": 45327 }, { "epoch": 13.95, "learning_rate": 4.4388612179489e-06, "loss": 0.0273, "step": 45328 }, { "epoch": 13.95, "learning_rate": 4.438447109574507e-06, "loss": 0.017, "step": 45329 }, { "epoch": 13.95, "learning_rate": 4.438033015007953e-06, "loss": 0.0289, "step": 45330 }, { "epoch": 13.95, "learning_rate": 4.437618934250258e-06, "loss": 0.0102, "step": 45331 }, { "epoch": 13.95, "learning_rate": 4.437204867302465e-06, "loss": 0.0095, "step": 45332 }, { "epoch": 13.95, "learning_rate": 4.436790814165587e-06, "loss": 0.0312, "step": 45333 }, { "epoch": 13.95, "learning_rate": 4.436376774840658e-06, "loss": 0.0233, "step": 45334 }, { "epoch": 13.95, "learning_rate": 4.435962749328707e-06, "loss": 0.0205, "step": 45335 }, { "epoch": 13.95, "learning_rate": 4.435548737630756e-06, "loss": 0.0276, "step": 45336 }, { "epoch": 13.95, "learning_rate": 4.43513473974784e-06, "loss": 0.0347, "step": 45337 }, { "epoch": 13.95, "learning_rate": 4.434720755680986e-06, "loss": 0.0361, "step": 45338 }, { "epoch": 13.95, "learning_rate": 4.434306785431218e-06, "loss": 0.0175, "step": 45339 }, { "epoch": 13.95, "learning_rate": 4.433892828999571e-06, "loss": 0.014, "step": 45340 }, { "epoch": 13.95, "learning_rate": 4.433478886387059e-06, "loss": 0.0215, "step": 45341 }, { "epoch": 13.95, "learning_rate": 4.433064957594723e-06, "loss": 0.0064, "step": 45342 }, { "epoch": 13.95, "learning_rate": 4.432651042623586e-06, "loss": 0.0159, "step": 45343 }, { "epoch": 13.95, "learning_rate": 4.432237141474674e-06, "loss": 0.0411, "step": 45344 }, { "epoch": 13.95, "learning_rate": 4.431823254149018e-06, "loss": 0.0322, "step": 45345 }, { "epoch": 13.95, "learning_rate": 4.4314093806476424e-06, "loss": 0.0112, "step": 45346 }, { "epoch": 13.95, "learning_rate": 4.4309955209715776e-06, "loss": 0.0206, "step": 45347 }, { "epoch": 13.95, "learning_rate": 4.4305816751218475e-06, "loss": 0.0144, "step": 45348 }, { "epoch": 13.95, "learning_rate": 4.430167843099483e-06, "loss": 0.0132, "step": 45349 }, { "epoch": 13.95, "learning_rate": 4.429754024905509e-06, "loss": 0.0123, "step": 45350 }, { "epoch": 13.95, "learning_rate": 4.4293402205409515e-06, "loss": 0.0231, "step": 45351 }, { "epoch": 13.95, "learning_rate": 4.428926430006844e-06, "loss": 0.0289, "step": 45352 }, { "epoch": 13.95, "learning_rate": 4.428512653304214e-06, "loss": 0.0123, "step": 45353 }, { "epoch": 13.95, "learning_rate": 4.428098890434081e-06, "loss": 0.0222, "step": 45354 }, { "epoch": 13.96, "learning_rate": 4.427685141397476e-06, "loss": 0.022, "step": 45355 }, { "epoch": 13.96, "learning_rate": 4.427271406195424e-06, "loss": 0.0097, "step": 45356 }, { "epoch": 13.96, "learning_rate": 4.4268576848289566e-06, "loss": 0.0176, "step": 45357 }, { "epoch": 13.96, "learning_rate": 4.426443977299101e-06, "loss": 0.0144, "step": 45358 }, { "epoch": 13.96, "learning_rate": 4.426030283606882e-06, "loss": 0.0178, "step": 45359 }, { "epoch": 13.96, "learning_rate": 4.425616603753326e-06, "loss": 0.0176, "step": 45360 }, { "epoch": 13.96, "learning_rate": 4.425202937739463e-06, "loss": 0.0626, "step": 45361 }, { "epoch": 13.96, "learning_rate": 4.4247892855663164e-06, "loss": 0.0144, "step": 45362 }, { "epoch": 13.96, "learning_rate": 4.424375647234915e-06, "loss": 0.0201, "step": 45363 }, { "epoch": 13.96, "learning_rate": 4.423962022746288e-06, "loss": 0.0208, "step": 45364 }, { "epoch": 13.96, "learning_rate": 4.4235484121014575e-06, "loss": 0.0129, "step": 45365 }, { "epoch": 13.96, "learning_rate": 4.423134815301454e-06, "loss": 0.0241, "step": 45366 }, { "epoch": 13.96, "learning_rate": 4.422721232347304e-06, "loss": 0.0199, "step": 45367 }, { "epoch": 13.96, "learning_rate": 4.422307663240032e-06, "loss": 0.016, "step": 45368 }, { "epoch": 13.96, "learning_rate": 4.421894107980667e-06, "loss": 0.018, "step": 45369 }, { "epoch": 13.96, "learning_rate": 4.421480566570235e-06, "loss": 0.02, "step": 45370 }, { "epoch": 13.96, "learning_rate": 4.42106703900976e-06, "loss": 0.0205, "step": 45371 }, { "epoch": 13.96, "learning_rate": 4.420653525300274e-06, "loss": 0.0198, "step": 45372 }, { "epoch": 13.96, "learning_rate": 4.420240025442802e-06, "loss": 0.0181, "step": 45373 }, { "epoch": 13.96, "learning_rate": 4.419826539438373e-06, "loss": 0.021, "step": 45374 }, { "epoch": 13.96, "learning_rate": 4.419413067288006e-06, "loss": 0.0304, "step": 45375 }, { "epoch": 13.96, "learning_rate": 4.41899960899273e-06, "loss": 0.0263, "step": 45376 }, { "epoch": 13.96, "learning_rate": 4.4185861645535756e-06, "loss": 0.021, "step": 45377 }, { "epoch": 13.96, "learning_rate": 4.4181727339715675e-06, "loss": 0.0207, "step": 45378 }, { "epoch": 13.96, "learning_rate": 4.417759317247732e-06, "loss": 0.0273, "step": 45379 }, { "epoch": 13.96, "learning_rate": 4.417345914383096e-06, "loss": 0.0134, "step": 45380 }, { "epoch": 13.96, "learning_rate": 4.416932525378683e-06, "loss": 0.0167, "step": 45381 }, { "epoch": 13.96, "learning_rate": 4.416519150235522e-06, "loss": 0.0189, "step": 45382 }, { "epoch": 13.96, "learning_rate": 4.4161057889546385e-06, "loss": 0.0142, "step": 45383 }, { "epoch": 13.96, "learning_rate": 4.41569244153706e-06, "loss": 0.0302, "step": 45384 }, { "epoch": 13.96, "learning_rate": 4.415279107983812e-06, "loss": 0.0129, "step": 45385 }, { "epoch": 13.96, "learning_rate": 4.414865788295916e-06, "loss": 0.0301, "step": 45386 }, { "epoch": 13.97, "learning_rate": 4.41445248247441e-06, "loss": 0.0185, "step": 45387 }, { "epoch": 13.97, "learning_rate": 4.414039190520308e-06, "loss": 0.0224, "step": 45388 }, { "epoch": 13.97, "learning_rate": 4.413625912434642e-06, "loss": 0.0127, "step": 45389 }, { "epoch": 13.97, "learning_rate": 4.413212648218436e-06, "loss": 0.0085, "step": 45390 }, { "epoch": 13.97, "learning_rate": 4.412799397872714e-06, "loss": 0.0125, "step": 45391 }, { "epoch": 13.97, "learning_rate": 4.412386161398507e-06, "loss": 0.0259, "step": 45392 }, { "epoch": 13.97, "learning_rate": 4.4119729387968405e-06, "loss": 0.0122, "step": 45393 }, { "epoch": 13.97, "learning_rate": 4.411559730068737e-06, "loss": 0.0142, "step": 45394 }, { "epoch": 13.97, "learning_rate": 4.411146535215229e-06, "loss": 0.0363, "step": 45395 }, { "epoch": 13.97, "learning_rate": 4.41073335423733e-06, "loss": 0.0197, "step": 45396 }, { "epoch": 13.97, "learning_rate": 4.410320187136077e-06, "loss": 0.0234, "step": 45397 }, { "epoch": 13.97, "learning_rate": 4.409907033912492e-06, "loss": 0.0299, "step": 45398 }, { "epoch": 13.97, "learning_rate": 4.4094938945676e-06, "loss": 0.0102, "step": 45399 }, { "epoch": 13.97, "learning_rate": 4.409080769102428e-06, "loss": 0.0091, "step": 45400 }, { "epoch": 13.97, "learning_rate": 4.408667657518002e-06, "loss": 0.0219, "step": 45401 }, { "epoch": 13.97, "learning_rate": 4.408254559815346e-06, "loss": 0.0649, "step": 45402 }, { "epoch": 13.97, "learning_rate": 4.407841475995487e-06, "loss": 0.0085, "step": 45403 }, { "epoch": 13.97, "learning_rate": 4.407428406059449e-06, "loss": 0.0266, "step": 45404 }, { "epoch": 13.97, "learning_rate": 4.407015350008261e-06, "loss": 0.0247, "step": 45405 }, { "epoch": 13.97, "learning_rate": 4.4066023078429395e-06, "loss": 0.0165, "step": 45406 }, { "epoch": 13.97, "learning_rate": 4.406189279564521e-06, "loss": 0.0174, "step": 45407 }, { "epoch": 13.97, "learning_rate": 4.405776265174031e-06, "loss": 0.0252, "step": 45408 }, { "epoch": 13.97, "learning_rate": 4.405363264672486e-06, "loss": 0.0099, "step": 45409 }, { "epoch": 13.97, "learning_rate": 4.404950278060916e-06, "loss": 0.0264, "step": 45410 }, { "epoch": 13.97, "learning_rate": 4.404537305340342e-06, "loss": 0.0123, "step": 45411 }, { "epoch": 13.97, "learning_rate": 4.4041243465117975e-06, "loss": 0.0132, "step": 45412 }, { "epoch": 13.97, "learning_rate": 4.4037114015763035e-06, "loss": 0.0246, "step": 45413 }, { "epoch": 13.97, "learning_rate": 4.403298470534885e-06, "loss": 0.0126, "step": 45414 }, { "epoch": 13.97, "learning_rate": 4.402885553388568e-06, "loss": 0.0147, "step": 45415 }, { "epoch": 13.97, "learning_rate": 4.4024726501383775e-06, "loss": 0.0256, "step": 45416 }, { "epoch": 13.97, "learning_rate": 4.402059760785338e-06, "loss": 0.0324, "step": 45417 }, { "epoch": 13.97, "learning_rate": 4.401646885330475e-06, "loss": 0.01, "step": 45418 }, { "epoch": 13.97, "learning_rate": 4.401234023774813e-06, "loss": 0.0275, "step": 45419 }, { "epoch": 13.98, "learning_rate": 4.400821176119378e-06, "loss": 0.0346, "step": 45420 }, { "epoch": 13.98, "learning_rate": 4.400408342365191e-06, "loss": 0.0137, "step": 45421 }, { "epoch": 13.98, "learning_rate": 4.399995522513288e-06, "loss": 0.0137, "step": 45422 }, { "epoch": 13.98, "learning_rate": 4.399582716564682e-06, "loss": 0.0249, "step": 45423 }, { "epoch": 13.98, "learning_rate": 4.399169924520403e-06, "loss": 0.0248, "step": 45424 }, { "epoch": 13.98, "learning_rate": 4.398757146381475e-06, "loss": 0.0072, "step": 45425 }, { "epoch": 13.98, "learning_rate": 4.398344382148919e-06, "loss": 0.022, "step": 45426 }, { "epoch": 13.98, "learning_rate": 4.397931631823767e-06, "loss": 0.0155, "step": 45427 }, { "epoch": 13.98, "learning_rate": 4.397518895407043e-06, "loss": 0.0224, "step": 45428 }, { "epoch": 13.98, "learning_rate": 4.39710617289977e-06, "loss": 0.0206, "step": 45429 }, { "epoch": 13.98, "learning_rate": 4.396693464302969e-06, "loss": 0.0162, "step": 45430 }, { "epoch": 13.98, "learning_rate": 4.396280769617665e-06, "loss": 0.0252, "step": 45431 }, { "epoch": 13.98, "learning_rate": 4.3958680888448866e-06, "loss": 0.0152, "step": 45432 }, { "epoch": 13.98, "learning_rate": 4.395455421985658e-06, "loss": 0.0149, "step": 45433 }, { "epoch": 13.98, "learning_rate": 4.395042769041004e-06, "loss": 0.0321, "step": 45434 }, { "epoch": 13.98, "learning_rate": 4.394630130011946e-06, "loss": 0.028, "step": 45435 }, { "epoch": 13.98, "learning_rate": 4.39421750489951e-06, "loss": 0.0316, "step": 45436 }, { "epoch": 13.98, "learning_rate": 4.393804893704722e-06, "loss": 0.027, "step": 45437 }, { "epoch": 13.98, "learning_rate": 4.393392296428605e-06, "loss": 0.0172, "step": 45438 }, { "epoch": 13.98, "learning_rate": 4.392979713072183e-06, "loss": 0.0289, "step": 45439 }, { "epoch": 13.98, "learning_rate": 4.39256714363648e-06, "loss": 0.0181, "step": 45440 }, { "epoch": 13.98, "learning_rate": 4.392154588122518e-06, "loss": 0.0238, "step": 45441 }, { "epoch": 13.98, "learning_rate": 4.391742046531329e-06, "loss": 0.0231, "step": 45442 }, { "epoch": 13.98, "learning_rate": 4.391329518863935e-06, "loss": 0.0254, "step": 45443 }, { "epoch": 13.98, "learning_rate": 4.390917005121353e-06, "loss": 0.0239, "step": 45444 }, { "epoch": 13.98, "learning_rate": 4.3905045053046145e-06, "loss": 0.0206, "step": 45445 }, { "epoch": 13.98, "learning_rate": 4.390092019414736e-06, "loss": 0.0155, "step": 45446 }, { "epoch": 13.98, "learning_rate": 4.389679547452751e-06, "loss": 0.016, "step": 45447 }, { "epoch": 13.98, "learning_rate": 4.389267089419678e-06, "loss": 0.0141, "step": 45448 }, { "epoch": 13.98, "learning_rate": 4.388854645316542e-06, "loss": 0.0278, "step": 45449 }, { "epoch": 13.98, "learning_rate": 4.388442215144373e-06, "loss": 0.0193, "step": 45450 }, { "epoch": 13.98, "learning_rate": 4.38802979890418e-06, "loss": 0.0302, "step": 45451 }, { "epoch": 13.99, "learning_rate": 4.387617396597e-06, "loss": 0.0175, "step": 45452 }, { "epoch": 13.99, "learning_rate": 4.3872050082238535e-06, "loss": 0.0116, "step": 45453 }, { "epoch": 13.99, "learning_rate": 4.386792633785764e-06, "loss": 0.027, "step": 45454 }, { "epoch": 13.99, "learning_rate": 4.386380273283754e-06, "loss": 0.0149, "step": 45455 }, { "epoch": 13.99, "learning_rate": 4.38596792671885e-06, "loss": 0.018, "step": 45456 }, { "epoch": 13.99, "learning_rate": 4.385555594092072e-06, "loss": 0.0399, "step": 45457 }, { "epoch": 13.99, "learning_rate": 4.385143275404447e-06, "loss": 0.0208, "step": 45458 }, { "epoch": 13.99, "learning_rate": 4.384730970656999e-06, "loss": 0.0185, "step": 45459 }, { "epoch": 13.99, "learning_rate": 4.384318679850748e-06, "loss": 0.0113, "step": 45460 }, { "epoch": 13.99, "learning_rate": 4.383906402986717e-06, "loss": 0.0352, "step": 45461 }, { "epoch": 13.99, "learning_rate": 4.3834941400659355e-06, "loss": 0.0154, "step": 45462 }, { "epoch": 13.99, "learning_rate": 4.383081891089425e-06, "loss": 0.019, "step": 45463 }, { "epoch": 13.99, "learning_rate": 4.38266965605821e-06, "loss": 0.0175, "step": 45464 }, { "epoch": 13.99, "learning_rate": 4.382257434973309e-06, "loss": 0.0191, "step": 45465 }, { "epoch": 13.99, "learning_rate": 4.3818452278357445e-06, "loss": 0.0175, "step": 45466 }, { "epoch": 13.99, "learning_rate": 4.381433034646548e-06, "loss": 0.0186, "step": 45467 }, { "epoch": 13.99, "learning_rate": 4.381020855406738e-06, "loss": 0.0106, "step": 45468 }, { "epoch": 13.99, "learning_rate": 4.380608690117339e-06, "loss": 0.0235, "step": 45469 }, { "epoch": 13.99, "learning_rate": 4.380196538779374e-06, "loss": 0.0238, "step": 45470 }, { "epoch": 13.99, "learning_rate": 4.379784401393864e-06, "loss": 0.0171, "step": 45471 }, { "epoch": 13.99, "learning_rate": 4.379372277961836e-06, "loss": 0.0186, "step": 45472 }, { "epoch": 13.99, "learning_rate": 4.37896016848431e-06, "loss": 0.0114, "step": 45473 }, { "epoch": 13.99, "learning_rate": 4.378548072962311e-06, "loss": 0.0284, "step": 45474 }, { "epoch": 13.99, "learning_rate": 4.3781359913968615e-06, "loss": 0.0361, "step": 45475 }, { "epoch": 13.99, "learning_rate": 4.377723923788982e-06, "loss": 0.0241, "step": 45476 }, { "epoch": 13.99, "learning_rate": 4.377311870139704e-06, "loss": 0.0227, "step": 45477 }, { "epoch": 13.99, "learning_rate": 4.376899830450042e-06, "loss": 0.0095, "step": 45478 }, { "epoch": 13.99, "learning_rate": 4.376487804721022e-06, "loss": 0.0092, "step": 45479 }, { "epoch": 13.99, "learning_rate": 4.3760757929536665e-06, "loss": 0.0154, "step": 45480 }, { "epoch": 13.99, "learning_rate": 4.375663795148995e-06, "loss": 0.0217, "step": 45481 }, { "epoch": 13.99, "learning_rate": 4.375251811308038e-06, "loss": 0.0336, "step": 45482 }, { "epoch": 13.99, "learning_rate": 4.374839841431814e-06, "loss": 0.0379, "step": 45483 }, { "epoch": 13.99, "learning_rate": 4.3744278855213465e-06, "loss": 0.0119, "step": 45484 }, { "epoch": 14.0, "learning_rate": 4.374015943577661e-06, "loss": 0.0061, "step": 45485 }, { "epoch": 14.0, "learning_rate": 4.37360401560177e-06, "loss": 0.022, "step": 45486 }, { "epoch": 14.0, "learning_rate": 4.373192101594708e-06, "loss": 0.0095, "step": 45487 }, { "epoch": 14.0, "learning_rate": 4.372780201557491e-06, "loss": 0.0249, "step": 45488 }, { "epoch": 14.0, "learning_rate": 4.3723683154911455e-06, "loss": 0.0119, "step": 45489 }, { "epoch": 14.0, "learning_rate": 4.371956443396692e-06, "loss": 0.0164, "step": 45490 }, { "epoch": 14.0, "learning_rate": 4.371544585275153e-06, "loss": 0.0198, "step": 45491 }, { "epoch": 14.0, "learning_rate": 4.371132741127553e-06, "loss": 0.0197, "step": 45492 }, { "epoch": 14.0, "learning_rate": 4.370720910954911e-06, "loss": 0.0284, "step": 45493 }, { "epoch": 14.0, "learning_rate": 4.370309094758252e-06, "loss": 0.0225, "step": 45494 }, { "epoch": 14.0, "learning_rate": 4.369897292538597e-06, "loss": 0.0245, "step": 45495 }, { "epoch": 14.0, "learning_rate": 4.369485504296967e-06, "loss": 0.0165, "step": 45496 }, { "epoch": 14.0, "learning_rate": 4.36907373003439e-06, "loss": 0.0266, "step": 45497 }, { "epoch": 14.0, "learning_rate": 4.368661969751887e-06, "loss": 0.0159, "step": 45498 }, { "epoch": 14.0, "learning_rate": 4.368250223450476e-06, "loss": 0.013, "step": 45499 }, { "epoch": 14.0, "learning_rate": 4.367838491131181e-06, "loss": 0.0256, "step": 45500 }, { "epoch": 14.0, "learning_rate": 4.367426772795022e-06, "loss": 0.0161, "step": 45501 }, { "epoch": 14.0, "learning_rate": 4.367015068443026e-06, "loss": 0.0091, "step": 45502 }, { "epoch": 14.0, "learning_rate": 4.366603378076213e-06, "loss": 0.0076, "step": 45503 }, { "epoch": 14.0, "learning_rate": 4.3661917016956056e-06, "loss": 0.0154, "step": 45504 }, { "epoch": 14.0, "learning_rate": 4.3657800393022255e-06, "loss": 0.0088, "step": 45505 }, { "epoch": 14.0, "learning_rate": 4.365368390897095e-06, "loss": 0.0063, "step": 45506 }, { "epoch": 14.0, "learning_rate": 4.364956756481234e-06, "loss": 0.0144, "step": 45507 }, { "epoch": 14.0, "learning_rate": 4.3645451360556676e-06, "loss": 0.0072, "step": 45508 }, { "epoch": 14.0, "learning_rate": 4.364133529621416e-06, "loss": 0.0228, "step": 45509 }, { "epoch": 14.0, "learning_rate": 4.363721937179501e-06, "loss": 0.0128, "step": 45510 }, { "epoch": 14.0, "learning_rate": 4.363310358730947e-06, "loss": 0.0225, "step": 45511 }, { "epoch": 14.0, "learning_rate": 4.362898794276772e-06, "loss": 0.0148, "step": 45512 }, { "epoch": 14.0, "learning_rate": 4.362487243818e-06, "loss": 0.0156, "step": 45513 }, { "epoch": 14.0, "learning_rate": 4.3620757073556516e-06, "loss": 0.0192, "step": 45514 }, { "epoch": 14.0, "learning_rate": 4.3616641848907456e-06, "loss": 0.017, "step": 45515 }, { "epoch": 14.0, "learning_rate": 4.361252676424313e-06, "loss": 0.014, "step": 45516 }, { "epoch": 14.01, "learning_rate": 4.360841181957368e-06, "loss": 0.0184, "step": 45517 }, { "epoch": 14.01, "learning_rate": 4.360429701490935e-06, "loss": 0.0191, "step": 45518 }, { "epoch": 14.01, "learning_rate": 4.360018235026038e-06, "loss": 0.0136, "step": 45519 }, { "epoch": 14.01, "learning_rate": 4.359606782563688e-06, "loss": 0.0167, "step": 45520 }, { "epoch": 14.01, "learning_rate": 4.359195344104916e-06, "loss": 0.0041, "step": 45521 }, { "epoch": 14.01, "learning_rate": 4.358783919650743e-06, "loss": 0.0108, "step": 45522 }, { "epoch": 14.01, "learning_rate": 4.358372509202189e-06, "loss": 0.0107, "step": 45523 }, { "epoch": 14.01, "learning_rate": 4.357961112760274e-06, "loss": 0.0177, "step": 45524 }, { "epoch": 14.01, "learning_rate": 4.35754973032602e-06, "loss": 0.0153, "step": 45525 }, { "epoch": 14.01, "learning_rate": 4.35713836190045e-06, "loss": 0.0092, "step": 45526 }, { "epoch": 14.01, "learning_rate": 4.3567270074845845e-06, "loss": 0.0124, "step": 45527 }, { "epoch": 14.01, "learning_rate": 4.3563156670794435e-06, "loss": 0.0104, "step": 45528 }, { "epoch": 14.01, "learning_rate": 4.355904340686049e-06, "loss": 0.0177, "step": 45529 }, { "epoch": 14.01, "learning_rate": 4.3554930283054195e-06, "loss": 0.0147, "step": 45530 }, { "epoch": 14.01, "learning_rate": 4.355081729938583e-06, "loss": 0.0065, "step": 45531 }, { "epoch": 14.01, "learning_rate": 4.3546704455865596e-06, "loss": 0.008, "step": 45532 }, { "epoch": 14.01, "learning_rate": 4.354259175250365e-06, "loss": 0.0108, "step": 45533 }, { "epoch": 14.01, "learning_rate": 4.3538479189310215e-06, "loss": 0.0054, "step": 45534 }, { "epoch": 14.01, "learning_rate": 4.353436676629549e-06, "loss": 0.0255, "step": 45535 }, { "epoch": 14.01, "learning_rate": 4.353025448346974e-06, "loss": 0.0126, "step": 45536 }, { "epoch": 14.01, "learning_rate": 4.352614234084314e-06, "loss": 0.0134, "step": 45537 }, { "epoch": 14.01, "learning_rate": 4.352203033842591e-06, "loss": 0.0277, "step": 45538 }, { "epoch": 14.01, "learning_rate": 4.351791847622825e-06, "loss": 0.0113, "step": 45539 }, { "epoch": 14.01, "learning_rate": 4.351380675426037e-06, "loss": 0.0142, "step": 45540 }, { "epoch": 14.01, "learning_rate": 4.350969517253248e-06, "loss": 0.0058, "step": 45541 }, { "epoch": 14.01, "learning_rate": 4.350558373105479e-06, "loss": 0.0191, "step": 45542 }, { "epoch": 14.01, "learning_rate": 4.350147242983751e-06, "loss": 0.0111, "step": 45543 }, { "epoch": 14.01, "learning_rate": 4.349736126889084e-06, "loss": 0.0118, "step": 45544 }, { "epoch": 14.01, "learning_rate": 4.349325024822498e-06, "loss": 0.0101, "step": 45545 }, { "epoch": 14.01, "learning_rate": 4.3489139367850155e-06, "loss": 0.0081, "step": 45546 }, { "epoch": 14.01, "learning_rate": 4.348502862777656e-06, "loss": 0.0169, "step": 45547 }, { "epoch": 14.01, "learning_rate": 4.348091802801439e-06, "loss": 0.0195, "step": 45548 }, { "epoch": 14.01, "learning_rate": 4.347680756857389e-06, "loss": 0.0131, "step": 45549 }, { "epoch": 14.02, "learning_rate": 4.3472697249465175e-06, "loss": 0.0155, "step": 45550 }, { "epoch": 14.02, "learning_rate": 4.3468587070698574e-06, "loss": 0.0123, "step": 45551 }, { "epoch": 14.02, "learning_rate": 4.3464477032284214e-06, "loss": 0.0184, "step": 45552 }, { "epoch": 14.02, "learning_rate": 4.346036713423236e-06, "loss": 0.0117, "step": 45553 }, { "epoch": 14.02, "learning_rate": 4.345625737655314e-06, "loss": 0.0134, "step": 45554 }, { "epoch": 14.02, "learning_rate": 4.345214775925674e-06, "loss": 0.0095, "step": 45555 }, { "epoch": 14.02, "learning_rate": 4.344803828235346e-06, "loss": 0.0137, "step": 45556 }, { "epoch": 14.02, "learning_rate": 4.344392894585344e-06, "loss": 0.0125, "step": 45557 }, { "epoch": 14.02, "learning_rate": 4.343981974976691e-06, "loss": 0.0122, "step": 45558 }, { "epoch": 14.02, "learning_rate": 4.343571069410406e-06, "loss": 0.0069, "step": 45559 }, { "epoch": 14.02, "learning_rate": 4.3431601778875085e-06, "loss": 0.0106, "step": 45560 }, { "epoch": 14.02, "learning_rate": 4.342749300409021e-06, "loss": 0.0062, "step": 45561 }, { "epoch": 14.02, "learning_rate": 4.342338436975959e-06, "loss": 0.0105, "step": 45562 }, { "epoch": 14.02, "learning_rate": 4.341927587589347e-06, "loss": 0.0132, "step": 45563 }, { "epoch": 14.02, "learning_rate": 4.341516752250203e-06, "loss": 0.0098, "step": 45564 }, { "epoch": 14.02, "learning_rate": 4.3411059309595434e-06, "loss": 0.0122, "step": 45565 }, { "epoch": 14.02, "learning_rate": 4.340695123718402e-06, "loss": 0.0105, "step": 45566 }, { "epoch": 14.02, "learning_rate": 4.340284330527782e-06, "loss": 0.0123, "step": 45567 }, { "epoch": 14.02, "learning_rate": 4.339873551388711e-06, "loss": 0.014, "step": 45568 }, { "epoch": 14.02, "learning_rate": 4.339462786302209e-06, "loss": 0.0134, "step": 45569 }, { "epoch": 14.02, "learning_rate": 4.339052035269291e-06, "loss": 0.0114, "step": 45570 }, { "epoch": 14.02, "learning_rate": 4.3386412982909846e-06, "loss": 0.0087, "step": 45571 }, { "epoch": 14.02, "learning_rate": 4.338230575368305e-06, "loss": 0.0129, "step": 45572 }, { "epoch": 14.02, "learning_rate": 4.337819866502272e-06, "loss": 0.0101, "step": 45573 }, { "epoch": 14.02, "learning_rate": 4.33740917169391e-06, "loss": 0.0122, "step": 45574 }, { "epoch": 14.02, "learning_rate": 4.336998490944228e-06, "loss": 0.0051, "step": 45575 }, { "epoch": 14.02, "learning_rate": 4.336587824254255e-06, "loss": 0.0076, "step": 45576 }, { "epoch": 14.02, "learning_rate": 4.3361771716250065e-06, "loss": 0.0092, "step": 45577 }, { "epoch": 14.02, "learning_rate": 4.335766533057505e-06, "loss": 0.0223, "step": 45578 }, { "epoch": 14.02, "learning_rate": 4.335355908552767e-06, "loss": 0.0094, "step": 45579 }, { "epoch": 14.02, "learning_rate": 4.334945298111813e-06, "loss": 0.0076, "step": 45580 }, { "epoch": 14.02, "learning_rate": 4.3345347017356635e-06, "loss": 0.0089, "step": 45581 }, { "epoch": 14.03, "learning_rate": 4.3341241194253356e-06, "loss": 0.0214, "step": 45582 }, { "epoch": 14.03, "learning_rate": 4.3337135511818514e-06, "loss": 0.0182, "step": 45583 }, { "epoch": 14.03, "learning_rate": 4.333302997006229e-06, "loss": 0.0237, "step": 45584 }, { "epoch": 14.03, "learning_rate": 4.332892456899483e-06, "loss": 0.0186, "step": 45585 }, { "epoch": 14.03, "learning_rate": 4.332481930862641e-06, "loss": 0.019, "step": 45586 }, { "epoch": 14.03, "learning_rate": 4.332071418896723e-06, "loss": 0.0121, "step": 45587 }, { "epoch": 14.03, "learning_rate": 4.33166092100274e-06, "loss": 0.0153, "step": 45588 }, { "epoch": 14.03, "learning_rate": 4.331250437181713e-06, "loss": 0.016, "step": 45589 }, { "epoch": 14.03, "learning_rate": 4.33083996743466e-06, "loss": 0.0195, "step": 45590 }, { "epoch": 14.03, "learning_rate": 4.3304295117626076e-06, "loss": 0.0071, "step": 45591 }, { "epoch": 14.03, "learning_rate": 4.330019070166569e-06, "loss": 0.0194, "step": 45592 }, { "epoch": 14.03, "learning_rate": 4.329608642647566e-06, "loss": 0.0117, "step": 45593 }, { "epoch": 14.03, "learning_rate": 4.3291982292066144e-06, "loss": 0.0208, "step": 45594 }, { "epoch": 14.03, "learning_rate": 4.328787829844736e-06, "loss": 0.0101, "step": 45595 }, { "epoch": 14.03, "learning_rate": 4.328377444562948e-06, "loss": 0.0152, "step": 45596 }, { "epoch": 14.03, "learning_rate": 4.32796707336227e-06, "loss": 0.0172, "step": 45597 }, { "epoch": 14.03, "learning_rate": 4.32755671624372e-06, "loss": 0.0292, "step": 45598 }, { "epoch": 14.03, "learning_rate": 4.3271463732083185e-06, "loss": 0.0076, "step": 45599 }, { "epoch": 14.03, "learning_rate": 4.326736044257079e-06, "loss": 0.007, "step": 45600 }, { "epoch": 14.03, "learning_rate": 4.326325729391032e-06, "loss": 0.0159, "step": 45601 }, { "epoch": 14.03, "learning_rate": 4.325915428611185e-06, "loss": 0.0193, "step": 45602 }, { "epoch": 14.03, "learning_rate": 4.325505141918559e-06, "loss": 0.0078, "step": 45603 }, { "epoch": 14.03, "learning_rate": 4.325094869314176e-06, "loss": 0.0094, "step": 45604 }, { "epoch": 14.03, "learning_rate": 4.324684610799047e-06, "loss": 0.015, "step": 45605 }, { "epoch": 14.03, "learning_rate": 4.3242743663742e-06, "loss": 0.0094, "step": 45606 }, { "epoch": 14.03, "learning_rate": 4.3238641360406505e-06, "loss": 0.0109, "step": 45607 }, { "epoch": 14.03, "learning_rate": 4.32345391979942e-06, "loss": 0.0097, "step": 45608 }, { "epoch": 14.03, "learning_rate": 4.323043717651518e-06, "loss": 0.0113, "step": 45609 }, { "epoch": 14.03, "learning_rate": 4.322633529597966e-06, "loss": 0.0093, "step": 45610 }, { "epoch": 14.03, "learning_rate": 4.322223355639788e-06, "loss": 0.0084, "step": 45611 }, { "epoch": 14.03, "learning_rate": 4.321813195777998e-06, "loss": 0.0312, "step": 45612 }, { "epoch": 14.03, "learning_rate": 4.321403050013614e-06, "loss": 0.0192, "step": 45613 }, { "epoch": 14.03, "learning_rate": 4.320992918347656e-06, "loss": 0.0122, "step": 45614 }, { "epoch": 14.04, "learning_rate": 4.320582800781142e-06, "loss": 0.0068, "step": 45615 }, { "epoch": 14.04, "learning_rate": 4.3201726973150904e-06, "loss": 0.0221, "step": 45616 }, { "epoch": 14.04, "learning_rate": 4.3197626079505185e-06, "loss": 0.0119, "step": 45617 }, { "epoch": 14.04, "learning_rate": 4.319352532688444e-06, "loss": 0.0106, "step": 45618 }, { "epoch": 14.04, "learning_rate": 4.318942471529886e-06, "loss": 0.0245, "step": 45619 }, { "epoch": 14.04, "learning_rate": 4.318532424475859e-06, "loss": 0.006, "step": 45620 }, { "epoch": 14.04, "learning_rate": 4.318122391527386e-06, "loss": 0.0151, "step": 45621 }, { "epoch": 14.04, "learning_rate": 4.31771237268549e-06, "loss": 0.0086, "step": 45622 }, { "epoch": 14.04, "learning_rate": 4.3173023679511764e-06, "loss": 0.0103, "step": 45623 }, { "epoch": 14.04, "learning_rate": 4.31689237732547e-06, "loss": 0.0095, "step": 45624 }, { "epoch": 14.04, "learning_rate": 4.316482400809383e-06, "loss": 0.0134, "step": 45625 }, { "epoch": 14.04, "learning_rate": 4.3160724384039435e-06, "loss": 0.0127, "step": 45626 }, { "epoch": 14.04, "learning_rate": 4.315662490110162e-06, "loss": 0.0147, "step": 45627 }, { "epoch": 14.04, "learning_rate": 4.3152525559290596e-06, "loss": 0.0165, "step": 45628 }, { "epoch": 14.04, "learning_rate": 4.314842635861655e-06, "loss": 0.0081, "step": 45629 }, { "epoch": 14.04, "learning_rate": 4.314432729908956e-06, "loss": 0.0112, "step": 45630 }, { "epoch": 14.04, "learning_rate": 4.3140228380719925e-06, "loss": 0.0283, "step": 45631 }, { "epoch": 14.04, "learning_rate": 4.313612960351776e-06, "loss": 0.0064, "step": 45632 }, { "epoch": 14.04, "learning_rate": 4.313203096749327e-06, "loss": 0.0115, "step": 45633 }, { "epoch": 14.04, "learning_rate": 4.312793247265661e-06, "loss": 0.009, "step": 45634 }, { "epoch": 14.04, "learning_rate": 4.312383411901796e-06, "loss": 0.0096, "step": 45635 }, { "epoch": 14.04, "learning_rate": 4.31197359065875e-06, "loss": 0.0088, "step": 45636 }, { "epoch": 14.04, "learning_rate": 4.31156378353754e-06, "loss": 0.0239, "step": 45637 }, { "epoch": 14.04, "learning_rate": 4.311153990539183e-06, "loss": 0.0158, "step": 45638 }, { "epoch": 14.04, "learning_rate": 4.310744211664698e-06, "loss": 0.0068, "step": 45639 }, { "epoch": 14.04, "learning_rate": 4.310334446915096e-06, "loss": 0.0175, "step": 45640 }, { "epoch": 14.04, "learning_rate": 4.309924696291405e-06, "loss": 0.0256, "step": 45641 }, { "epoch": 14.04, "learning_rate": 4.309514959794637e-06, "loss": 0.0151, "step": 45642 }, { "epoch": 14.04, "learning_rate": 4.309105237425812e-06, "loss": 0.0207, "step": 45643 }, { "epoch": 14.04, "learning_rate": 4.308695529185941e-06, "loss": 0.0143, "step": 45644 }, { "epoch": 14.04, "learning_rate": 4.308285835076042e-06, "loss": 0.0209, "step": 45645 }, { "epoch": 14.04, "learning_rate": 4.307876155097137e-06, "loss": 0.0079, "step": 45646 }, { "epoch": 14.05, "learning_rate": 4.307466489250241e-06, "loss": 0.0171, "step": 45647 }, { "epoch": 14.05, "learning_rate": 4.307056837536373e-06, "loss": 0.0108, "step": 45648 }, { "epoch": 14.05, "learning_rate": 4.306647199956546e-06, "loss": 0.0116, "step": 45649 }, { "epoch": 14.05, "learning_rate": 4.30623757651178e-06, "loss": 0.0109, "step": 45650 }, { "epoch": 14.05, "learning_rate": 4.305827967203091e-06, "loss": 0.0166, "step": 45651 }, { "epoch": 14.05, "learning_rate": 4.3054183720314965e-06, "loss": 0.0153, "step": 45652 }, { "epoch": 14.05, "learning_rate": 4.305008790998013e-06, "loss": 0.0196, "step": 45653 }, { "epoch": 14.05, "learning_rate": 4.304599224103657e-06, "loss": 0.0109, "step": 45654 }, { "epoch": 14.05, "learning_rate": 4.3041896713494415e-06, "loss": 0.0111, "step": 45655 }, { "epoch": 14.05, "learning_rate": 4.303780132736395e-06, "loss": 0.0294, "step": 45656 }, { "epoch": 14.05, "learning_rate": 4.303370608265524e-06, "loss": 0.0078, "step": 45657 }, { "epoch": 14.05, "learning_rate": 4.302961097937847e-06, "loss": 0.0119, "step": 45658 }, { "epoch": 14.05, "learning_rate": 4.302551601754382e-06, "loss": 0.02, "step": 45659 }, { "epoch": 14.05, "learning_rate": 4.302142119716142e-06, "loss": 0.0099, "step": 45660 }, { "epoch": 14.05, "learning_rate": 4.301732651824151e-06, "loss": 0.0144, "step": 45661 }, { "epoch": 14.05, "learning_rate": 4.3013231980794215e-06, "loss": 0.0264, "step": 45662 }, { "epoch": 14.05, "learning_rate": 4.30091375848297e-06, "loss": 0.0044, "step": 45663 }, { "epoch": 14.05, "learning_rate": 4.300504333035815e-06, "loss": 0.0235, "step": 45664 }, { "epoch": 14.05, "learning_rate": 4.300094921738966e-06, "loss": 0.0182, "step": 45665 }, { "epoch": 14.05, "learning_rate": 4.299685524593449e-06, "loss": 0.0176, "step": 45666 }, { "epoch": 14.05, "learning_rate": 4.299276141600275e-06, "loss": 0.0243, "step": 45667 }, { "epoch": 14.05, "learning_rate": 4.298866772760461e-06, "loss": 0.0104, "step": 45668 }, { "epoch": 14.05, "learning_rate": 4.298457418075024e-06, "loss": 0.0148, "step": 45669 }, { "epoch": 14.05, "learning_rate": 4.298048077544981e-06, "loss": 0.0078, "step": 45670 }, { "epoch": 14.05, "learning_rate": 4.297638751171347e-06, "loss": 0.0133, "step": 45671 }, { "epoch": 14.05, "learning_rate": 4.2972294389551385e-06, "loss": 0.0178, "step": 45672 }, { "epoch": 14.05, "learning_rate": 4.296820140897371e-06, "loss": 0.0095, "step": 45673 }, { "epoch": 14.05, "learning_rate": 4.296410856999062e-06, "loss": 0.0091, "step": 45674 }, { "epoch": 14.05, "learning_rate": 4.296001587261225e-06, "loss": 0.013, "step": 45675 }, { "epoch": 14.05, "learning_rate": 4.2955923316848814e-06, "loss": 0.0135, "step": 45676 }, { "epoch": 14.05, "learning_rate": 4.2951830902710466e-06, "loss": 0.0144, "step": 45677 }, { "epoch": 14.05, "learning_rate": 4.294773863020732e-06, "loss": 0.0077, "step": 45678 }, { "epoch": 14.05, "learning_rate": 4.294364649934955e-06, "loss": 0.0149, "step": 45679 }, { "epoch": 14.06, "learning_rate": 4.293955451014728e-06, "loss": 0.0084, "step": 45680 }, { "epoch": 14.06, "learning_rate": 4.293546266261077e-06, "loss": 0.0105, "step": 45681 }, { "epoch": 14.06, "learning_rate": 4.293137095675011e-06, "loss": 0.0371, "step": 45682 }, { "epoch": 14.06, "learning_rate": 4.292727939257547e-06, "loss": 0.0071, "step": 45683 }, { "epoch": 14.06, "learning_rate": 4.292318797009701e-06, "loss": 0.0106, "step": 45684 }, { "epoch": 14.06, "learning_rate": 4.291909668932489e-06, "loss": 0.0134, "step": 45685 }, { "epoch": 14.06, "learning_rate": 4.291500555026926e-06, "loss": 0.0154, "step": 45686 }, { "epoch": 14.06, "learning_rate": 4.291091455294028e-06, "loss": 0.0188, "step": 45687 }, { "epoch": 14.06, "learning_rate": 4.290682369734812e-06, "loss": 0.018, "step": 45688 }, { "epoch": 14.06, "learning_rate": 4.290273298350292e-06, "loss": 0.011, "step": 45689 }, { "epoch": 14.06, "learning_rate": 4.289864241141483e-06, "loss": 0.0269, "step": 45690 }, { "epoch": 14.06, "learning_rate": 4.2894551981094024e-06, "loss": 0.0099, "step": 45691 }, { "epoch": 14.06, "learning_rate": 4.289046169255065e-06, "loss": 0.0109, "step": 45692 }, { "epoch": 14.06, "learning_rate": 4.288637154579487e-06, "loss": 0.0102, "step": 45693 }, { "epoch": 14.06, "learning_rate": 4.288228154083682e-06, "loss": 0.0088, "step": 45694 }, { "epoch": 14.06, "learning_rate": 4.287819167768664e-06, "loss": 0.0096, "step": 45695 }, { "epoch": 14.06, "learning_rate": 4.287410195635454e-06, "loss": 0.0111, "step": 45696 }, { "epoch": 14.06, "learning_rate": 4.287001237685066e-06, "loss": 0.0129, "step": 45697 }, { "epoch": 14.06, "learning_rate": 4.286592293918515e-06, "loss": 0.0081, "step": 45698 }, { "epoch": 14.06, "learning_rate": 4.286183364336813e-06, "loss": 0.0084, "step": 45699 }, { "epoch": 14.06, "learning_rate": 4.2857744489409725e-06, "loss": 0.0157, "step": 45700 }, { "epoch": 14.06, "learning_rate": 4.285365547732019e-06, "loss": 0.02, "step": 45701 }, { "epoch": 14.06, "learning_rate": 4.284956660710961e-06, "loss": 0.0065, "step": 45702 }, { "epoch": 14.06, "learning_rate": 4.2845477878788145e-06, "loss": 0.0182, "step": 45703 }, { "epoch": 14.06, "learning_rate": 4.284138929236595e-06, "loss": 0.0191, "step": 45704 }, { "epoch": 14.06, "learning_rate": 4.2837300847853194e-06, "loss": 0.0129, "step": 45705 }, { "epoch": 14.06, "learning_rate": 4.283321254525999e-06, "loss": 0.0124, "step": 45706 }, { "epoch": 14.06, "learning_rate": 4.282912438459651e-06, "loss": 0.023, "step": 45707 }, { "epoch": 14.06, "learning_rate": 4.2825036365872904e-06, "loss": 0.0165, "step": 45708 }, { "epoch": 14.06, "learning_rate": 4.282094848909932e-06, "loss": 0.0134, "step": 45709 }, { "epoch": 14.06, "learning_rate": 4.281686075428587e-06, "loss": 0.0123, "step": 45710 }, { "epoch": 14.06, "learning_rate": 4.281277316144281e-06, "loss": 0.0173, "step": 45711 }, { "epoch": 14.07, "learning_rate": 4.280868571058019e-06, "loss": 0.0078, "step": 45712 }, { "epoch": 14.07, "learning_rate": 4.280459840170818e-06, "loss": 0.0126, "step": 45713 }, { "epoch": 14.07, "learning_rate": 4.280051123483693e-06, "loss": 0.0218, "step": 45714 }, { "epoch": 14.07, "learning_rate": 4.279642420997655e-06, "loss": 0.0046, "step": 45715 }, { "epoch": 14.07, "learning_rate": 4.2792337327137265e-06, "loss": 0.0227, "step": 45716 }, { "epoch": 14.07, "learning_rate": 4.278825058632918e-06, "loss": 0.0127, "step": 45717 }, { "epoch": 14.07, "learning_rate": 4.278416398756245e-06, "loss": 0.0152, "step": 45718 }, { "epoch": 14.07, "learning_rate": 4.278007753084726e-06, "loss": 0.0075, "step": 45719 }, { "epoch": 14.07, "learning_rate": 4.277599121619363e-06, "loss": 0.0188, "step": 45720 }, { "epoch": 14.07, "learning_rate": 4.277190504361183e-06, "loss": 0.0228, "step": 45721 }, { "epoch": 14.07, "learning_rate": 4.276781901311195e-06, "loss": 0.0132, "step": 45722 }, { "epoch": 14.07, "learning_rate": 4.276373312470416e-06, "loss": 0.018, "step": 45723 }, { "epoch": 14.07, "learning_rate": 4.275964737839857e-06, "loss": 0.0099, "step": 45724 }, { "epoch": 14.07, "learning_rate": 4.2755561774205366e-06, "loss": 0.0227, "step": 45725 }, { "epoch": 14.07, "learning_rate": 4.2751476312134655e-06, "loss": 0.0106, "step": 45726 }, { "epoch": 14.07, "learning_rate": 4.274739099219661e-06, "loss": 0.018, "step": 45727 }, { "epoch": 14.07, "learning_rate": 4.274330581440133e-06, "loss": 0.0136, "step": 45728 }, { "epoch": 14.07, "learning_rate": 4.273922077875902e-06, "loss": 0.0091, "step": 45729 }, { "epoch": 14.07, "learning_rate": 4.273513588527973e-06, "loss": 0.0124, "step": 45730 }, { "epoch": 14.07, "learning_rate": 4.27310511339737e-06, "loss": 0.0133, "step": 45731 }, { "epoch": 14.07, "learning_rate": 4.272696652485107e-06, "loss": 0.0119, "step": 45732 }, { "epoch": 14.07, "learning_rate": 4.27228820579219e-06, "loss": 0.0137, "step": 45733 }, { "epoch": 14.07, "learning_rate": 4.271879773319637e-06, "loss": 0.0154, "step": 45734 }, { "epoch": 14.07, "learning_rate": 4.27147135506846e-06, "loss": 0.0122, "step": 45735 }, { "epoch": 14.07, "learning_rate": 4.271062951039678e-06, "loss": 0.0142, "step": 45736 }, { "epoch": 14.07, "learning_rate": 4.270654561234302e-06, "loss": 0.014, "step": 45737 }, { "epoch": 14.07, "learning_rate": 4.270246185653347e-06, "loss": 0.0249, "step": 45738 }, { "epoch": 14.07, "learning_rate": 4.2698378242978254e-06, "loss": 0.0096, "step": 45739 }, { "epoch": 14.07, "learning_rate": 4.269429477168752e-06, "loss": 0.0104, "step": 45740 }, { "epoch": 14.07, "learning_rate": 4.2690211442671395e-06, "loss": 0.0187, "step": 45741 }, { "epoch": 14.07, "learning_rate": 4.2686128255940025e-06, "loss": 0.0049, "step": 45742 }, { "epoch": 14.07, "learning_rate": 4.268204521150357e-06, "loss": 0.0139, "step": 45743 }, { "epoch": 14.07, "learning_rate": 4.267796230937212e-06, "loss": 0.0054, "step": 45744 }, { "epoch": 14.08, "learning_rate": 4.267387954955584e-06, "loss": 0.0099, "step": 45745 }, { "epoch": 14.08, "learning_rate": 4.266979693206487e-06, "loss": 0.0097, "step": 45746 }, { "epoch": 14.08, "learning_rate": 4.266571445690934e-06, "loss": 0.0125, "step": 45747 }, { "epoch": 14.08, "learning_rate": 4.2661632124099376e-06, "loss": 0.0142, "step": 45748 }, { "epoch": 14.08, "learning_rate": 4.265754993364512e-06, "loss": 0.0072, "step": 45749 }, { "epoch": 14.08, "learning_rate": 4.2653467885556675e-06, "loss": 0.0509, "step": 45750 }, { "epoch": 14.08, "learning_rate": 4.264938597984424e-06, "loss": 0.0154, "step": 45751 }, { "epoch": 14.08, "learning_rate": 4.264530421651792e-06, "loss": 0.0242, "step": 45752 }, { "epoch": 14.08, "learning_rate": 4.264122259558789e-06, "loss": 0.0113, "step": 45753 }, { "epoch": 14.08, "learning_rate": 4.263714111706421e-06, "loss": 0.0131, "step": 45754 }, { "epoch": 14.08, "learning_rate": 4.2633059780957e-06, "loss": 0.0109, "step": 45755 }, { "epoch": 14.08, "learning_rate": 4.262897858727647e-06, "loss": 0.0119, "step": 45756 }, { "epoch": 14.08, "learning_rate": 4.262489753603273e-06, "loss": 0.0123, "step": 45757 }, { "epoch": 14.08, "learning_rate": 4.26208166272359e-06, "loss": 0.0129, "step": 45758 }, { "epoch": 14.08, "learning_rate": 4.26167358608961e-06, "loss": 0.0119, "step": 45759 }, { "epoch": 14.08, "learning_rate": 4.2612655237023484e-06, "loss": 0.0082, "step": 45760 }, { "epoch": 14.08, "learning_rate": 4.260857475562817e-06, "loss": 0.01, "step": 45761 }, { "epoch": 14.08, "learning_rate": 4.26044944167203e-06, "loss": 0.0128, "step": 45762 }, { "epoch": 14.08, "learning_rate": 4.2600414220309984e-06, "loss": 0.0121, "step": 45763 }, { "epoch": 14.08, "learning_rate": 4.259633416640737e-06, "loss": 0.0092, "step": 45764 }, { "epoch": 14.08, "learning_rate": 4.259225425502256e-06, "loss": 0.0102, "step": 45765 }, { "epoch": 14.08, "learning_rate": 4.258817448616577e-06, "loss": 0.0123, "step": 45766 }, { "epoch": 14.08, "learning_rate": 4.258409485984702e-06, "loss": 0.0155, "step": 45767 }, { "epoch": 14.08, "learning_rate": 4.258001537607649e-06, "loss": 0.0254, "step": 45768 }, { "epoch": 14.08, "learning_rate": 4.257593603486431e-06, "loss": 0.0039, "step": 45769 }, { "epoch": 14.08, "learning_rate": 4.2571856836220546e-06, "loss": 0.0102, "step": 45770 }, { "epoch": 14.08, "learning_rate": 4.256777778015543e-06, "loss": 0.0181, "step": 45771 }, { "epoch": 14.08, "learning_rate": 4.256369886667904e-06, "loss": 0.0108, "step": 45772 }, { "epoch": 14.08, "learning_rate": 4.255962009580149e-06, "loss": 0.0087, "step": 45773 }, { "epoch": 14.08, "learning_rate": 4.255554146753296e-06, "loss": 0.0122, "step": 45774 }, { "epoch": 14.08, "learning_rate": 4.255146298188346e-06, "loss": 0.0119, "step": 45775 }, { "epoch": 14.08, "learning_rate": 4.254738463886323e-06, "loss": 0.0126, "step": 45776 }, { "epoch": 14.09, "learning_rate": 4.254330643848234e-06, "loss": 0.0121, "step": 45777 }, { "epoch": 14.09, "learning_rate": 4.2539228380750955e-06, "loss": 0.0124, "step": 45778 }, { "epoch": 14.09, "learning_rate": 4.253515046567917e-06, "loss": 0.0088, "step": 45779 }, { "epoch": 14.09, "learning_rate": 4.253107269327709e-06, "loss": 0.0123, "step": 45780 }, { "epoch": 14.09, "learning_rate": 4.252699506355488e-06, "loss": 0.0099, "step": 45781 }, { "epoch": 14.09, "learning_rate": 4.252291757652266e-06, "loss": 0.0116, "step": 45782 }, { "epoch": 14.09, "learning_rate": 4.251884023219053e-06, "loss": 0.0099, "step": 45783 }, { "epoch": 14.09, "learning_rate": 4.251476303056862e-06, "loss": 0.0134, "step": 45784 }, { "epoch": 14.09, "learning_rate": 4.251068597166702e-06, "loss": 0.0143, "step": 45785 }, { "epoch": 14.09, "learning_rate": 4.250660905549592e-06, "loss": 0.0124, "step": 45786 }, { "epoch": 14.09, "learning_rate": 4.2502532282065455e-06, "loss": 0.0132, "step": 45787 }, { "epoch": 14.09, "learning_rate": 4.2498455651385665e-06, "loss": 0.0087, "step": 45788 }, { "epoch": 14.09, "learning_rate": 4.24943791634667e-06, "loss": 0.0127, "step": 45789 }, { "epoch": 14.09, "learning_rate": 4.2490302818318655e-06, "loss": 0.0101, "step": 45790 }, { "epoch": 14.09, "learning_rate": 4.248622661595172e-06, "loss": 0.0199, "step": 45791 }, { "epoch": 14.09, "learning_rate": 4.248215055637598e-06, "loss": 0.0099, "step": 45792 }, { "epoch": 14.09, "learning_rate": 4.247807463960155e-06, "loss": 0.0059, "step": 45793 }, { "epoch": 14.09, "learning_rate": 4.247399886563856e-06, "loss": 0.0104, "step": 45794 }, { "epoch": 14.09, "learning_rate": 4.246992323449712e-06, "loss": 0.0193, "step": 45795 }, { "epoch": 14.09, "learning_rate": 4.246584774618735e-06, "loss": 0.0156, "step": 45796 }, { "epoch": 14.09, "learning_rate": 4.246177240071937e-06, "loss": 0.0034, "step": 45797 }, { "epoch": 14.09, "learning_rate": 4.24576971981033e-06, "loss": 0.0428, "step": 45798 }, { "epoch": 14.09, "learning_rate": 4.245362213834925e-06, "loss": 0.0226, "step": 45799 }, { "epoch": 14.09, "learning_rate": 4.244954722146731e-06, "loss": 0.0123, "step": 45800 }, { "epoch": 14.09, "learning_rate": 4.244547244746771e-06, "loss": 0.0096, "step": 45801 }, { "epoch": 14.09, "learning_rate": 4.244139781636043e-06, "loss": 0.0088, "step": 45802 }, { "epoch": 14.09, "learning_rate": 4.243732332815565e-06, "loss": 0.0288, "step": 45803 }, { "epoch": 14.09, "learning_rate": 4.243324898286349e-06, "loss": 0.0126, "step": 45804 }, { "epoch": 14.09, "learning_rate": 4.2429174780493995e-06, "loss": 0.0062, "step": 45805 }, { "epoch": 14.09, "learning_rate": 4.242510072105739e-06, "loss": 0.0148, "step": 45806 }, { "epoch": 14.09, "learning_rate": 4.242102680456374e-06, "loss": 0.0101, "step": 45807 }, { "epoch": 14.09, "learning_rate": 4.241695303102318e-06, "loss": 0.0094, "step": 45808 }, { "epoch": 14.09, "learning_rate": 4.241287940044577e-06, "loss": 0.0159, "step": 45809 }, { "epoch": 14.1, "learning_rate": 4.240880591284163e-06, "loss": 0.0243, "step": 45810 }, { "epoch": 14.1, "learning_rate": 4.240473256822093e-06, "loss": 0.0073, "step": 45811 }, { "epoch": 14.1, "learning_rate": 4.240065936659374e-06, "loss": 0.0179, "step": 45812 }, { "epoch": 14.1, "learning_rate": 4.23965863079702e-06, "loss": 0.0102, "step": 45813 }, { "epoch": 14.1, "learning_rate": 4.23925133923604e-06, "loss": 0.0122, "step": 45814 }, { "epoch": 14.1, "learning_rate": 4.238844061977446e-06, "loss": 0.0242, "step": 45815 }, { "epoch": 14.1, "learning_rate": 4.238436799022248e-06, "loss": 0.0186, "step": 45816 }, { "epoch": 14.1, "learning_rate": 4.238029550371459e-06, "loss": 0.0074, "step": 45817 }, { "epoch": 14.1, "learning_rate": 4.237622316026091e-06, "loss": 0.0107, "step": 45818 }, { "epoch": 14.1, "learning_rate": 4.237215095987152e-06, "loss": 0.0112, "step": 45819 }, { "epoch": 14.1, "learning_rate": 4.23680789025565e-06, "loss": 0.0167, "step": 45820 }, { "epoch": 14.1, "learning_rate": 4.236400698832603e-06, "loss": 0.0139, "step": 45821 }, { "epoch": 14.1, "learning_rate": 4.235993521719025e-06, "loss": 0.0088, "step": 45822 }, { "epoch": 14.1, "learning_rate": 4.235586358915917e-06, "loss": 0.0096, "step": 45823 }, { "epoch": 14.1, "learning_rate": 4.2351792104242935e-06, "loss": 0.0077, "step": 45824 }, { "epoch": 14.1, "learning_rate": 4.234772076245163e-06, "loss": 0.0156, "step": 45825 }, { "epoch": 14.1, "learning_rate": 4.234364956379542e-06, "loss": 0.0113, "step": 45826 }, { "epoch": 14.1, "learning_rate": 4.233957850828439e-06, "loss": 0.0085, "step": 45827 }, { "epoch": 14.1, "learning_rate": 4.233550759592864e-06, "loss": 0.0142, "step": 45828 }, { "epoch": 14.1, "learning_rate": 4.23314368267383e-06, "loss": 0.0143, "step": 45829 }, { "epoch": 14.1, "learning_rate": 4.2327366200723404e-06, "loss": 0.0073, "step": 45830 }, { "epoch": 14.1, "learning_rate": 4.232329571789414e-06, "loss": 0.0156, "step": 45831 }, { "epoch": 14.1, "learning_rate": 4.231922537826059e-06, "loss": 0.0112, "step": 45832 }, { "epoch": 14.1, "learning_rate": 4.231515518183283e-06, "loss": 0.0122, "step": 45833 }, { "epoch": 14.1, "learning_rate": 4.231108512862101e-06, "loss": 0.0154, "step": 45834 }, { "epoch": 14.1, "learning_rate": 4.23070152186352e-06, "loss": 0.0058, "step": 45835 }, { "epoch": 14.1, "learning_rate": 4.230294545188554e-06, "loss": 0.0061, "step": 45836 }, { "epoch": 14.1, "learning_rate": 4.229887582838209e-06, "loss": 0.011, "step": 45837 }, { "epoch": 14.1, "learning_rate": 4.229480634813499e-06, "loss": 0.0122, "step": 45838 }, { "epoch": 14.1, "learning_rate": 4.229073701115431e-06, "loss": 0.0146, "step": 45839 }, { "epoch": 14.1, "learning_rate": 4.228666781745016e-06, "loss": 0.0212, "step": 45840 }, { "epoch": 14.1, "learning_rate": 4.2282598767032664e-06, "loss": 0.0082, "step": 45841 }, { "epoch": 14.11, "learning_rate": 4.227852985991198e-06, "loss": 0.0073, "step": 45842 }, { "epoch": 14.11, "learning_rate": 4.2274461096098085e-06, "loss": 0.0178, "step": 45843 }, { "epoch": 14.11, "learning_rate": 4.227039247560115e-06, "loss": 0.0188, "step": 45844 }, { "epoch": 14.11, "learning_rate": 4.226632399843124e-06, "loss": 0.012, "step": 45845 }, { "epoch": 14.11, "learning_rate": 4.226225566459851e-06, "loss": 0.0097, "step": 45846 }, { "epoch": 14.11, "learning_rate": 4.225818747411302e-06, "loss": 0.0119, "step": 45847 }, { "epoch": 14.11, "learning_rate": 4.225411942698491e-06, "loss": 0.0113, "step": 45848 }, { "epoch": 14.11, "learning_rate": 4.225005152322423e-06, "loss": 0.0201, "step": 45849 }, { "epoch": 14.11, "learning_rate": 4.224598376284113e-06, "loss": 0.0108, "step": 45850 }, { "epoch": 14.11, "learning_rate": 4.224191614584565e-06, "loss": 0.0176, "step": 45851 }, { "epoch": 14.11, "learning_rate": 4.223784867224795e-06, "loss": 0.0161, "step": 45852 }, { "epoch": 14.11, "learning_rate": 4.223378134205808e-06, "loss": 0.0242, "step": 45853 }, { "epoch": 14.11, "learning_rate": 4.222971415528617e-06, "loss": 0.0155, "step": 45854 }, { "epoch": 14.11, "learning_rate": 4.222564711194226e-06, "loss": 0.0066, "step": 45855 }, { "epoch": 14.11, "learning_rate": 4.222158021203657e-06, "loss": 0.0165, "step": 45856 }, { "epoch": 14.11, "learning_rate": 4.221751345557908e-06, "loss": 0.0103, "step": 45857 }, { "epoch": 14.11, "learning_rate": 4.221344684257993e-06, "loss": 0.0061, "step": 45858 }, { "epoch": 14.11, "learning_rate": 4.220938037304921e-06, "loss": 0.0153, "step": 45859 }, { "epoch": 14.11, "learning_rate": 4.220531404699699e-06, "loss": 0.0119, "step": 45860 }, { "epoch": 14.11, "learning_rate": 4.2201247864433425e-06, "loss": 0.0116, "step": 45861 }, { "epoch": 14.11, "learning_rate": 4.2197181825368585e-06, "loss": 0.022, "step": 45862 }, { "epoch": 14.11, "learning_rate": 4.219311592981258e-06, "loss": 0.0203, "step": 45863 }, { "epoch": 14.11, "learning_rate": 4.218905017777546e-06, "loss": 0.0092, "step": 45864 }, { "epoch": 14.11, "learning_rate": 4.218498456926731e-06, "loss": 0.0151, "step": 45865 }, { "epoch": 14.11, "learning_rate": 4.218091910429828e-06, "loss": 0.0487, "step": 45866 }, { "epoch": 14.11, "learning_rate": 4.217685378287844e-06, "loss": 0.0125, "step": 45867 }, { "epoch": 14.11, "learning_rate": 4.21727886050179e-06, "loss": 0.0055, "step": 45868 }, { "epoch": 14.11, "learning_rate": 4.2168723570726735e-06, "loss": 0.0184, "step": 45869 }, { "epoch": 14.11, "learning_rate": 4.216465868001502e-06, "loss": 0.0064, "step": 45870 }, { "epoch": 14.11, "learning_rate": 4.216059393289288e-06, "loss": 0.0081, "step": 45871 }, { "epoch": 14.11, "learning_rate": 4.215652932937039e-06, "loss": 0.0172, "step": 45872 }, { "epoch": 14.11, "learning_rate": 4.215246486945764e-06, "loss": 0.0097, "step": 45873 }, { "epoch": 14.11, "learning_rate": 4.214840055316473e-06, "loss": 0.0191, "step": 45874 }, { "epoch": 14.12, "learning_rate": 4.214433638050171e-06, "loss": 0.0138, "step": 45875 }, { "epoch": 14.12, "learning_rate": 4.214027235147874e-06, "loss": 0.0091, "step": 45876 }, { "epoch": 14.12, "learning_rate": 4.21362084661059e-06, "loss": 0.0055, "step": 45877 }, { "epoch": 14.12, "learning_rate": 4.2132144724393235e-06, "loss": 0.0233, "step": 45878 }, { "epoch": 14.12, "learning_rate": 4.212808112635084e-06, "loss": 0.0143, "step": 45879 }, { "epoch": 14.12, "learning_rate": 4.212401767198879e-06, "loss": 0.0137, "step": 45880 }, { "epoch": 14.12, "learning_rate": 4.211995436131724e-06, "loss": 0.0169, "step": 45881 }, { "epoch": 14.12, "learning_rate": 4.211589119434622e-06, "loss": 0.0054, "step": 45882 }, { "epoch": 14.12, "learning_rate": 4.2111828171085865e-06, "loss": 0.0196, "step": 45883 }, { "epoch": 14.12, "learning_rate": 4.210776529154625e-06, "loss": 0.0107, "step": 45884 }, { "epoch": 14.12, "learning_rate": 4.210370255573738e-06, "loss": 0.023, "step": 45885 }, { "epoch": 14.12, "learning_rate": 4.209963996366945e-06, "loss": 0.0105, "step": 45886 }, { "epoch": 14.12, "learning_rate": 4.20955775153525e-06, "loss": 0.0186, "step": 45887 }, { "epoch": 14.12, "learning_rate": 4.209151521079662e-06, "loss": 0.008, "step": 45888 }, { "epoch": 14.12, "learning_rate": 4.20874530500119e-06, "loss": 0.017, "step": 45889 }, { "epoch": 14.12, "learning_rate": 4.208339103300842e-06, "loss": 0.0131, "step": 45890 }, { "epoch": 14.12, "learning_rate": 4.207932915979627e-06, "loss": 0.0236, "step": 45891 }, { "epoch": 14.12, "learning_rate": 4.207526743038551e-06, "loss": 0.0159, "step": 45892 }, { "epoch": 14.12, "learning_rate": 4.207120584478627e-06, "loss": 0.0082, "step": 45893 }, { "epoch": 14.12, "learning_rate": 4.20671444030086e-06, "loss": 0.0138, "step": 45894 }, { "epoch": 14.12, "learning_rate": 4.206308310506255e-06, "loss": 0.019, "step": 45895 }, { "epoch": 14.12, "learning_rate": 4.205902195095829e-06, "loss": 0.0111, "step": 45896 }, { "epoch": 14.12, "learning_rate": 4.205496094070586e-06, "loss": 0.0086, "step": 45897 }, { "epoch": 14.12, "learning_rate": 4.205090007431537e-06, "loss": 0.013, "step": 45898 }, { "epoch": 14.12, "learning_rate": 4.204683935179683e-06, "loss": 0.0081, "step": 45899 }, { "epoch": 14.12, "learning_rate": 4.204277877316035e-06, "loss": 0.0196, "step": 45900 }, { "epoch": 14.12, "learning_rate": 4.203871833841604e-06, "loss": 0.0107, "step": 45901 }, { "epoch": 14.12, "learning_rate": 4.203465804757397e-06, "loss": 0.013, "step": 45902 }, { "epoch": 14.12, "learning_rate": 4.2030597900644224e-06, "loss": 0.0087, "step": 45903 }, { "epoch": 14.12, "learning_rate": 4.202653789763688e-06, "loss": 0.0054, "step": 45904 }, { "epoch": 14.12, "learning_rate": 4.2022478038562e-06, "loss": 0.0199, "step": 45905 }, { "epoch": 14.12, "learning_rate": 4.201841832342967e-06, "loss": 0.0096, "step": 45906 }, { "epoch": 14.13, "learning_rate": 4.201435875225e-06, "loss": 0.0102, "step": 45907 }, { "epoch": 14.13, "learning_rate": 4.201029932503303e-06, "loss": 0.0202, "step": 45908 }, { "epoch": 14.13, "learning_rate": 4.200624004178883e-06, "loss": 0.0162, "step": 45909 }, { "epoch": 14.13, "learning_rate": 4.200218090252752e-06, "loss": 0.0111, "step": 45910 }, { "epoch": 14.13, "learning_rate": 4.19981219072592e-06, "loss": 0.0063, "step": 45911 }, { "epoch": 14.13, "learning_rate": 4.199406305599387e-06, "loss": 0.0188, "step": 45912 }, { "epoch": 14.13, "learning_rate": 4.1990004348741656e-06, "loss": 0.015, "step": 45913 }, { "epoch": 14.13, "learning_rate": 4.198594578551257e-06, "loss": 0.0157, "step": 45914 }, { "epoch": 14.13, "learning_rate": 4.198188736631679e-06, "loss": 0.0125, "step": 45915 }, { "epoch": 14.13, "learning_rate": 4.197782909116434e-06, "loss": 0.0121, "step": 45916 }, { "epoch": 14.13, "learning_rate": 4.197377096006529e-06, "loss": 0.0298, "step": 45917 }, { "epoch": 14.13, "learning_rate": 4.196971297302972e-06, "loss": 0.0192, "step": 45918 }, { "epoch": 14.13, "learning_rate": 4.1965655130067715e-06, "loss": 0.0065, "step": 45919 }, { "epoch": 14.13, "learning_rate": 4.196159743118935e-06, "loss": 0.0055, "step": 45920 }, { "epoch": 14.13, "learning_rate": 4.1957539876404685e-06, "loss": 0.0148, "step": 45921 }, { "epoch": 14.13, "learning_rate": 4.19534824657238e-06, "loss": 0.0128, "step": 45922 }, { "epoch": 14.13, "learning_rate": 4.1949425199156765e-06, "loss": 0.0218, "step": 45923 }, { "epoch": 14.13, "learning_rate": 4.194536807671366e-06, "loss": 0.0072, "step": 45924 }, { "epoch": 14.13, "learning_rate": 4.194131109840457e-06, "loss": 0.0092, "step": 45925 }, { "epoch": 14.13, "learning_rate": 4.193725426423953e-06, "loss": 0.011, "step": 45926 }, { "epoch": 14.13, "learning_rate": 4.193319757422865e-06, "loss": 0.0184, "step": 45927 }, { "epoch": 14.13, "learning_rate": 4.192914102838198e-06, "loss": 0.0181, "step": 45928 }, { "epoch": 14.13, "learning_rate": 4.192508462670956e-06, "loss": 0.0154, "step": 45929 }, { "epoch": 14.13, "learning_rate": 4.1921028369221535e-06, "loss": 0.0226, "step": 45930 }, { "epoch": 14.13, "learning_rate": 4.191697225592795e-06, "loss": 0.0123, "step": 45931 }, { "epoch": 14.13, "learning_rate": 4.191291628683888e-06, "loss": 0.0108, "step": 45932 }, { "epoch": 14.13, "learning_rate": 4.190886046196435e-06, "loss": 0.0132, "step": 45933 }, { "epoch": 14.13, "learning_rate": 4.190480478131443e-06, "loss": 0.0099, "step": 45934 }, { "epoch": 14.13, "learning_rate": 4.1900749244899256e-06, "loss": 0.013, "step": 45935 }, { "epoch": 14.13, "learning_rate": 4.1896693852728845e-06, "loss": 0.0072, "step": 45936 }, { "epoch": 14.13, "learning_rate": 4.189263860481329e-06, "loss": 0.0076, "step": 45937 }, { "epoch": 14.13, "learning_rate": 4.1888583501162635e-06, "loss": 0.012, "step": 45938 }, { "epoch": 14.13, "learning_rate": 4.188452854178699e-06, "loss": 0.0191, "step": 45939 }, { "epoch": 14.14, "learning_rate": 4.188047372669636e-06, "loss": 0.006, "step": 45940 }, { "epoch": 14.14, "learning_rate": 4.1876419055900866e-06, "loss": 0.0188, "step": 45941 }, { "epoch": 14.14, "learning_rate": 4.187236452941054e-06, "loss": 0.0093, "step": 45942 }, { "epoch": 14.14, "learning_rate": 4.186831014723547e-06, "loss": 0.0189, "step": 45943 }, { "epoch": 14.14, "learning_rate": 4.186425590938568e-06, "loss": 0.0103, "step": 45944 }, { "epoch": 14.14, "learning_rate": 4.186020181587134e-06, "loss": 0.0143, "step": 45945 }, { "epoch": 14.14, "learning_rate": 4.1856147866702405e-06, "loss": 0.0248, "step": 45946 }, { "epoch": 14.14, "learning_rate": 4.185209406188898e-06, "loss": 0.0148, "step": 45947 }, { "epoch": 14.14, "learning_rate": 4.184804040144113e-06, "loss": 0.009, "step": 45948 }, { "epoch": 14.14, "learning_rate": 4.184398688536888e-06, "loss": 0.0104, "step": 45949 }, { "epoch": 14.14, "learning_rate": 4.183993351368238e-06, "loss": 0.0158, "step": 45950 }, { "epoch": 14.14, "learning_rate": 4.183588028639163e-06, "loss": 0.0078, "step": 45951 }, { "epoch": 14.14, "learning_rate": 4.183182720350672e-06, "loss": 0.0071, "step": 45952 }, { "epoch": 14.14, "learning_rate": 4.182777426503774e-06, "loss": 0.0094, "step": 45953 }, { "epoch": 14.14, "learning_rate": 4.182372147099463e-06, "loss": 0.0125, "step": 45954 }, { "epoch": 14.14, "learning_rate": 4.181966882138757e-06, "loss": 0.0271, "step": 45955 }, { "epoch": 14.14, "learning_rate": 4.181561631622659e-06, "loss": 0.0262, "step": 45956 }, { "epoch": 14.14, "learning_rate": 4.181156395552176e-06, "loss": 0.017, "step": 45957 }, { "epoch": 14.14, "learning_rate": 4.180751173928312e-06, "loss": 0.0097, "step": 45958 }, { "epoch": 14.14, "learning_rate": 4.180345966752075e-06, "loss": 0.0169, "step": 45959 }, { "epoch": 14.14, "learning_rate": 4.179940774024469e-06, "loss": 0.0133, "step": 45960 }, { "epoch": 14.14, "learning_rate": 4.179535595746501e-06, "loss": 0.011, "step": 45961 }, { "epoch": 14.14, "learning_rate": 4.179130431919178e-06, "loss": 0.0114, "step": 45962 }, { "epoch": 14.14, "learning_rate": 4.178725282543503e-06, "loss": 0.0137, "step": 45963 }, { "epoch": 14.14, "learning_rate": 4.178320147620482e-06, "loss": 0.0085, "step": 45964 }, { "epoch": 14.14, "learning_rate": 4.177915027151126e-06, "loss": 0.0107, "step": 45965 }, { "epoch": 14.14, "learning_rate": 4.17750992113644e-06, "loss": 0.0149, "step": 45966 }, { "epoch": 14.14, "learning_rate": 4.177104829577424e-06, "loss": 0.0084, "step": 45967 }, { "epoch": 14.14, "learning_rate": 4.176699752475087e-06, "loss": 0.0076, "step": 45968 }, { "epoch": 14.14, "learning_rate": 4.176294689830431e-06, "loss": 0.0167, "step": 45969 }, { "epoch": 14.14, "learning_rate": 4.175889641644469e-06, "loss": 0.0221, "step": 45970 }, { "epoch": 14.14, "learning_rate": 4.175484607918203e-06, "loss": 0.021, "step": 45971 }, { "epoch": 14.15, "learning_rate": 4.175079588652638e-06, "loss": 0.0283, "step": 45972 }, { "epoch": 14.15, "learning_rate": 4.17467458384878e-06, "loss": 0.0163, "step": 45973 }, { "epoch": 14.15, "learning_rate": 4.174269593507635e-06, "loss": 0.0047, "step": 45974 }, { "epoch": 14.15, "learning_rate": 4.173864617630208e-06, "loss": 0.0076, "step": 45975 }, { "epoch": 14.15, "learning_rate": 4.173459656217505e-06, "loss": 0.0065, "step": 45976 }, { "epoch": 14.15, "learning_rate": 4.1730547092705295e-06, "loss": 0.01, "step": 45977 }, { "epoch": 14.15, "learning_rate": 4.172649776790288e-06, "loss": 0.0129, "step": 45978 }, { "epoch": 14.15, "learning_rate": 4.1722448587777875e-06, "loss": 0.0103, "step": 45979 }, { "epoch": 14.15, "learning_rate": 4.171839955234032e-06, "loss": 0.0076, "step": 45980 }, { "epoch": 14.15, "learning_rate": 4.171435066160025e-06, "loss": 0.0242, "step": 45981 }, { "epoch": 14.15, "learning_rate": 4.171030191556775e-06, "loss": 0.0413, "step": 45982 }, { "epoch": 14.15, "learning_rate": 4.170625331425285e-06, "loss": 0.0285, "step": 45983 }, { "epoch": 14.15, "learning_rate": 4.170220485766559e-06, "loss": 0.0094, "step": 45984 }, { "epoch": 14.15, "learning_rate": 4.169815654581605e-06, "loss": 0.0428, "step": 45985 }, { "epoch": 14.15, "learning_rate": 4.169410837871427e-06, "loss": 0.011, "step": 45986 }, { "epoch": 14.15, "learning_rate": 4.169006035637034e-06, "loss": 0.0127, "step": 45987 }, { "epoch": 14.15, "learning_rate": 4.1686012478794245e-06, "loss": 0.0109, "step": 45988 }, { "epoch": 14.15, "learning_rate": 4.168196474599602e-06, "loss": 0.0135, "step": 45989 }, { "epoch": 14.15, "learning_rate": 4.167791715798578e-06, "loss": 0.0098, "step": 45990 }, { "epoch": 14.15, "learning_rate": 4.167386971477356e-06, "loss": 0.0128, "step": 45991 }, { "epoch": 14.15, "learning_rate": 4.166982241636939e-06, "loss": 0.0122, "step": 45992 }, { "epoch": 14.15, "learning_rate": 4.1665775262783325e-06, "loss": 0.0196, "step": 45993 }, { "epoch": 14.15, "learning_rate": 4.166172825402542e-06, "loss": 0.0199, "step": 45994 }, { "epoch": 14.15, "learning_rate": 4.165768139010572e-06, "loss": 0.0068, "step": 45995 }, { "epoch": 14.15, "learning_rate": 4.165363467103426e-06, "loss": 0.0121, "step": 45996 }, { "epoch": 14.15, "learning_rate": 4.16495880968211e-06, "loss": 0.0141, "step": 45997 }, { "epoch": 14.15, "learning_rate": 4.1645541667476276e-06, "loss": 0.0096, "step": 45998 }, { "epoch": 14.15, "learning_rate": 4.16414953830098e-06, "loss": 0.0144, "step": 45999 }, { "epoch": 14.15, "learning_rate": 4.163744924343183e-06, "loss": 0.0145, "step": 46000 }, { "epoch": 14.15, "learning_rate": 4.163340324875232e-06, "loss": 0.0097, "step": 46001 }, { "epoch": 14.15, "learning_rate": 4.162935739898133e-06, "loss": 0.0082, "step": 46002 }, { "epoch": 14.15, "learning_rate": 4.16253116941289e-06, "loss": 0.0113, "step": 46003 }, { "epoch": 14.15, "learning_rate": 4.1621266134205055e-06, "loss": 0.0057, "step": 46004 }, { "epoch": 14.16, "learning_rate": 4.16172207192199e-06, "loss": 0.0133, "step": 46005 }, { "epoch": 14.16, "learning_rate": 4.161317544918345e-06, "loss": 0.0104, "step": 46006 }, { "epoch": 14.16, "learning_rate": 4.160913032410574e-06, "loss": 0.0128, "step": 46007 }, { "epoch": 14.16, "learning_rate": 4.160508534399685e-06, "loss": 0.007, "step": 46008 }, { "epoch": 14.16, "learning_rate": 4.160104050886673e-06, "loss": 0.0062, "step": 46009 }, { "epoch": 14.16, "learning_rate": 4.159699581872552e-06, "loss": 0.0137, "step": 46010 }, { "epoch": 14.16, "learning_rate": 4.159295127358321e-06, "loss": 0.0142, "step": 46011 }, { "epoch": 14.16, "learning_rate": 4.158890687344986e-06, "loss": 0.0135, "step": 46012 }, { "epoch": 14.16, "learning_rate": 4.158486261833552e-06, "loss": 0.0087, "step": 46013 }, { "epoch": 14.16, "learning_rate": 4.158081850825021e-06, "loss": 0.0083, "step": 46014 }, { "epoch": 14.16, "learning_rate": 4.157677454320397e-06, "loss": 0.0094, "step": 46015 }, { "epoch": 14.16, "learning_rate": 4.1572730723206855e-06, "loss": 0.0127, "step": 46016 }, { "epoch": 14.16, "learning_rate": 4.15686870482689e-06, "loss": 0.0314, "step": 46017 }, { "epoch": 14.16, "learning_rate": 4.1564643518400135e-06, "loss": 0.0133, "step": 46018 }, { "epoch": 14.16, "learning_rate": 4.156060013361058e-06, "loss": 0.0134, "step": 46019 }, { "epoch": 14.16, "learning_rate": 4.155655689391034e-06, "loss": 0.0076, "step": 46020 }, { "epoch": 14.16, "learning_rate": 4.155251379930943e-06, "loss": 0.0233, "step": 46021 }, { "epoch": 14.16, "learning_rate": 4.1548470849817854e-06, "loss": 0.0121, "step": 46022 }, { "epoch": 14.16, "learning_rate": 4.154442804544566e-06, "loss": 0.02, "step": 46023 }, { "epoch": 14.16, "learning_rate": 4.154038538620285e-06, "loss": 0.0104, "step": 46024 }, { "epoch": 14.16, "learning_rate": 4.153634287209955e-06, "loss": 0.01, "step": 46025 }, { "epoch": 14.16, "learning_rate": 4.153230050314574e-06, "loss": 0.0168, "step": 46026 }, { "epoch": 14.16, "learning_rate": 4.152825827935147e-06, "loss": 0.0112, "step": 46027 }, { "epoch": 14.16, "learning_rate": 4.1524216200726775e-06, "loss": 0.0114, "step": 46028 }, { "epoch": 14.16, "learning_rate": 4.152017426728168e-06, "loss": 0.008, "step": 46029 }, { "epoch": 14.16, "learning_rate": 4.1516132479026225e-06, "loss": 0.0156, "step": 46030 }, { "epoch": 14.16, "learning_rate": 4.151209083597045e-06, "loss": 0.0139, "step": 46031 }, { "epoch": 14.16, "learning_rate": 4.1508049338124386e-06, "loss": 0.0162, "step": 46032 }, { "epoch": 14.16, "learning_rate": 4.150400798549806e-06, "loss": 0.0216, "step": 46033 }, { "epoch": 14.16, "learning_rate": 4.149996677810149e-06, "loss": 0.0256, "step": 46034 }, { "epoch": 14.16, "learning_rate": 4.149592571594479e-06, "loss": 0.0119, "step": 46035 }, { "epoch": 14.16, "learning_rate": 4.14918847990379e-06, "loss": 0.0074, "step": 46036 }, { "epoch": 14.17, "learning_rate": 4.14878440273909e-06, "loss": 0.0103, "step": 46037 }, { "epoch": 14.17, "learning_rate": 4.14838034010138e-06, "loss": 0.0142, "step": 46038 }, { "epoch": 14.17, "learning_rate": 4.14797629199166e-06, "loss": 0.0083, "step": 46039 }, { "epoch": 14.17, "learning_rate": 4.147572258410941e-06, "loss": 0.0096, "step": 46040 }, { "epoch": 14.17, "learning_rate": 4.147168239360223e-06, "loss": 0.0104, "step": 46041 }, { "epoch": 14.17, "learning_rate": 4.146764234840511e-06, "loss": 0.0146, "step": 46042 }, { "epoch": 14.17, "learning_rate": 4.146360244852803e-06, "loss": 0.0265, "step": 46043 }, { "epoch": 14.17, "learning_rate": 4.145956269398101e-06, "loss": 0.0055, "step": 46044 }, { "epoch": 14.17, "learning_rate": 4.145552308477413e-06, "loss": 0.0105, "step": 46045 }, { "epoch": 14.17, "learning_rate": 4.145148362091742e-06, "loss": 0.0171, "step": 46046 }, { "epoch": 14.17, "learning_rate": 4.144744430242088e-06, "loss": 0.0103, "step": 46047 }, { "epoch": 14.17, "learning_rate": 4.144340512929457e-06, "loss": 0.0118, "step": 46048 }, { "epoch": 14.17, "learning_rate": 4.143936610154848e-06, "loss": 0.0212, "step": 46049 }, { "epoch": 14.17, "learning_rate": 4.143532721919265e-06, "loss": 0.0115, "step": 46050 }, { "epoch": 14.17, "learning_rate": 4.143128848223714e-06, "loss": 0.0176, "step": 46051 }, { "epoch": 14.17, "learning_rate": 4.142724989069193e-06, "loss": 0.0142, "step": 46052 }, { "epoch": 14.17, "learning_rate": 4.142321144456708e-06, "loss": 0.0106, "step": 46053 }, { "epoch": 14.17, "learning_rate": 4.141917314387256e-06, "loss": 0.0112, "step": 46054 }, { "epoch": 14.17, "learning_rate": 4.141513498861848e-06, "loss": 0.0124, "step": 46055 }, { "epoch": 14.17, "learning_rate": 4.141109697881487e-06, "loss": 0.0109, "step": 46056 }, { "epoch": 14.17, "learning_rate": 4.140705911447166e-06, "loss": 0.0122, "step": 46057 }, { "epoch": 14.17, "learning_rate": 4.140302139559893e-06, "loss": 0.0176, "step": 46058 }, { "epoch": 14.17, "learning_rate": 4.1398983822206676e-06, "loss": 0.0106, "step": 46059 }, { "epoch": 14.17, "learning_rate": 4.1394946394304965e-06, "loss": 0.0086, "step": 46060 }, { "epoch": 14.17, "learning_rate": 4.139090911190381e-06, "loss": 0.0124, "step": 46061 }, { "epoch": 14.17, "learning_rate": 4.138687197501323e-06, "loss": 0.0174, "step": 46062 }, { "epoch": 14.17, "learning_rate": 4.138283498364328e-06, "loss": 0.0295, "step": 46063 }, { "epoch": 14.17, "learning_rate": 4.137879813780388e-06, "loss": 0.0079, "step": 46064 }, { "epoch": 14.17, "learning_rate": 4.137476143750515e-06, "loss": 0.0077, "step": 46065 }, { "epoch": 14.17, "learning_rate": 4.137072488275708e-06, "loss": 0.0135, "step": 46066 }, { "epoch": 14.17, "learning_rate": 4.136668847356971e-06, "loss": 0.0222, "step": 46067 }, { "epoch": 14.17, "learning_rate": 4.136265220995303e-06, "loss": 0.0088, "step": 46068 }, { "epoch": 14.17, "learning_rate": 4.135861609191708e-06, "loss": 0.0146, "step": 46069 }, { "epoch": 14.18, "learning_rate": 4.135458011947188e-06, "loss": 0.0059, "step": 46070 }, { "epoch": 14.18, "learning_rate": 4.135054429262745e-06, "loss": 0.009, "step": 46071 }, { "epoch": 14.18, "learning_rate": 4.13465086113938e-06, "loss": 0.0129, "step": 46072 }, { "epoch": 14.18, "learning_rate": 4.134247307578096e-06, "loss": 0.0155, "step": 46073 }, { "epoch": 14.18, "learning_rate": 4.13384376857989e-06, "loss": 0.0117, "step": 46074 }, { "epoch": 14.18, "learning_rate": 4.1334402441457735e-06, "loss": 0.0129, "step": 46075 }, { "epoch": 14.18, "learning_rate": 4.133036734276743e-06, "loss": 0.0163, "step": 46076 }, { "epoch": 14.18, "learning_rate": 4.132633238973803e-06, "loss": 0.013, "step": 46077 }, { "epoch": 14.18, "learning_rate": 4.132229758237951e-06, "loss": 0.0183, "step": 46078 }, { "epoch": 14.18, "learning_rate": 4.131826292070186e-06, "loss": 0.0057, "step": 46079 }, { "epoch": 14.18, "learning_rate": 4.1314228404715185e-06, "loss": 0.0223, "step": 46080 }, { "epoch": 14.18, "learning_rate": 4.131019403442945e-06, "loss": 0.0157, "step": 46081 }, { "epoch": 14.18, "learning_rate": 4.1306159809854685e-06, "loss": 0.0068, "step": 46082 }, { "epoch": 14.18, "learning_rate": 4.130212573100091e-06, "loss": 0.0273, "step": 46083 }, { "epoch": 14.18, "learning_rate": 4.129809179787813e-06, "loss": 0.0206, "step": 46084 }, { "epoch": 14.18, "learning_rate": 4.129405801049635e-06, "loss": 0.0065, "step": 46085 }, { "epoch": 14.18, "learning_rate": 4.129002436886561e-06, "loss": 0.0223, "step": 46086 }, { "epoch": 14.18, "learning_rate": 4.128599087299591e-06, "loss": 0.0142, "step": 46087 }, { "epoch": 14.18, "learning_rate": 4.128195752289726e-06, "loss": 0.0107, "step": 46088 }, { "epoch": 14.18, "learning_rate": 4.127792431857965e-06, "loss": 0.0138, "step": 46089 }, { "epoch": 14.18, "learning_rate": 4.127389126005319e-06, "loss": 0.0151, "step": 46090 }, { "epoch": 14.18, "learning_rate": 4.12698583473278e-06, "loss": 0.0063, "step": 46091 }, { "epoch": 14.18, "learning_rate": 4.12658255804135e-06, "loss": 0.0143, "step": 46092 }, { "epoch": 14.18, "learning_rate": 4.126179295932033e-06, "loss": 0.0113, "step": 46093 }, { "epoch": 14.18, "learning_rate": 4.125776048405826e-06, "loss": 0.0137, "step": 46094 }, { "epoch": 14.18, "learning_rate": 4.1253728154637375e-06, "loss": 0.0152, "step": 46095 }, { "epoch": 14.18, "learning_rate": 4.1249695971067635e-06, "loss": 0.0099, "step": 46096 }, { "epoch": 14.18, "learning_rate": 4.124566393335906e-06, "loss": 0.0167, "step": 46097 }, { "epoch": 14.18, "learning_rate": 4.1241632041521705e-06, "loss": 0.0247, "step": 46098 }, { "epoch": 14.18, "learning_rate": 4.123760029556546e-06, "loss": 0.0229, "step": 46099 }, { "epoch": 14.18, "learning_rate": 4.123356869550046e-06, "loss": 0.0135, "step": 46100 }, { "epoch": 14.18, "learning_rate": 4.122953724133665e-06, "loss": 0.0053, "step": 46101 }, { "epoch": 14.19, "learning_rate": 4.122550593308406e-06, "loss": 0.016, "step": 46102 }, { "epoch": 14.19, "learning_rate": 4.12214747707527e-06, "loss": 0.0188, "step": 46103 }, { "epoch": 14.19, "learning_rate": 4.121744375435256e-06, "loss": 0.014, "step": 46104 }, { "epoch": 14.19, "learning_rate": 4.121341288389366e-06, "loss": 0.0124, "step": 46105 }, { "epoch": 14.19, "learning_rate": 4.120938215938602e-06, "loss": 0.0102, "step": 46106 }, { "epoch": 14.19, "learning_rate": 4.1205351580839634e-06, "loss": 0.015, "step": 46107 }, { "epoch": 14.19, "learning_rate": 4.120132114826451e-06, "loss": 0.0163, "step": 46108 }, { "epoch": 14.19, "learning_rate": 4.11972908616706e-06, "loss": 0.0072, "step": 46109 }, { "epoch": 14.19, "learning_rate": 4.119326072106802e-06, "loss": 0.0098, "step": 46110 }, { "epoch": 14.19, "learning_rate": 4.118923072646674e-06, "loss": 0.0139, "step": 46111 }, { "epoch": 14.19, "learning_rate": 4.118520087787673e-06, "loss": 0.0136, "step": 46112 }, { "epoch": 14.19, "learning_rate": 4.1181171175308e-06, "loss": 0.0146, "step": 46113 }, { "epoch": 14.19, "learning_rate": 4.117714161877052e-06, "loss": 0.0093, "step": 46114 }, { "epoch": 14.19, "learning_rate": 4.117311220827438e-06, "loss": 0.0263, "step": 46115 }, { "epoch": 14.19, "learning_rate": 4.116908294382956e-06, "loss": 0.007, "step": 46116 }, { "epoch": 14.19, "learning_rate": 4.116505382544602e-06, "loss": 0.0084, "step": 46117 }, { "epoch": 14.19, "learning_rate": 4.11610248531338e-06, "loss": 0.0093, "step": 46118 }, { "epoch": 14.19, "learning_rate": 4.11569960269029e-06, "loss": 0.0241, "step": 46119 }, { "epoch": 14.19, "learning_rate": 4.115296734676331e-06, "loss": 0.008, "step": 46120 }, { "epoch": 14.19, "learning_rate": 4.1148938812725035e-06, "loss": 0.0101, "step": 46121 }, { "epoch": 14.19, "learning_rate": 4.114491042479808e-06, "loss": 0.0047, "step": 46122 }, { "epoch": 14.19, "learning_rate": 4.114088218299244e-06, "loss": 0.0077, "step": 46123 }, { "epoch": 14.19, "learning_rate": 4.113685408731813e-06, "loss": 0.0264, "step": 46124 }, { "epoch": 14.19, "learning_rate": 4.113282613778514e-06, "loss": 0.0064, "step": 46125 }, { "epoch": 14.19, "learning_rate": 4.112879833440347e-06, "loss": 0.0113, "step": 46126 }, { "epoch": 14.19, "learning_rate": 4.1124770677183135e-06, "loss": 0.0182, "step": 46127 }, { "epoch": 14.19, "learning_rate": 4.1120743166134115e-06, "loss": 0.0101, "step": 46128 }, { "epoch": 14.19, "learning_rate": 4.111671580126637e-06, "loss": 0.0152, "step": 46129 }, { "epoch": 14.19, "learning_rate": 4.111268858259e-06, "loss": 0.0136, "step": 46130 }, { "epoch": 14.19, "learning_rate": 4.110866151011493e-06, "loss": 0.014, "step": 46131 }, { "epoch": 14.19, "learning_rate": 4.110463458385122e-06, "loss": 0.0071, "step": 46132 }, { "epoch": 14.19, "learning_rate": 4.11006078038088e-06, "loss": 0.0076, "step": 46133 }, { "epoch": 14.19, "learning_rate": 4.109658116999764e-06, "loss": 0.0196, "step": 46134 }, { "epoch": 14.2, "learning_rate": 4.109255468242783e-06, "loss": 0.0116, "step": 46135 }, { "epoch": 14.2, "learning_rate": 4.108852834110932e-06, "loss": 0.0075, "step": 46136 }, { "epoch": 14.2, "learning_rate": 4.108450214605213e-06, "loss": 0.0082, "step": 46137 }, { "epoch": 14.2, "learning_rate": 4.108047609726622e-06, "loss": 0.0106, "step": 46138 }, { "epoch": 14.2, "learning_rate": 4.1076450194761615e-06, "loss": 0.013, "step": 46139 }, { "epoch": 14.2, "learning_rate": 4.10724244385483e-06, "loss": 0.0091, "step": 46140 }, { "epoch": 14.2, "learning_rate": 4.106839882863626e-06, "loss": 0.0115, "step": 46141 }, { "epoch": 14.2, "learning_rate": 4.10643733650355e-06, "loss": 0.0209, "step": 46142 }, { "epoch": 14.2, "learning_rate": 4.106034804775602e-06, "loss": 0.0137, "step": 46143 }, { "epoch": 14.2, "learning_rate": 4.105632287680776e-06, "loss": 0.0119, "step": 46144 }, { "epoch": 14.2, "learning_rate": 4.105229785220083e-06, "loss": 0.006, "step": 46145 }, { "epoch": 14.2, "learning_rate": 4.104827297394513e-06, "loss": 0.0112, "step": 46146 }, { "epoch": 14.2, "learning_rate": 4.104424824205067e-06, "loss": 0.0119, "step": 46147 }, { "epoch": 14.2, "learning_rate": 4.104022365652744e-06, "loss": 0.0172, "step": 46148 }, { "epoch": 14.2, "learning_rate": 4.1036199217385404e-06, "loss": 0.0121, "step": 46149 }, { "epoch": 14.2, "learning_rate": 4.103217492463463e-06, "loss": 0.0181, "step": 46150 }, { "epoch": 14.2, "learning_rate": 4.102815077828507e-06, "loss": 0.0115, "step": 46151 }, { "epoch": 14.2, "learning_rate": 4.1024126778346705e-06, "loss": 0.017, "step": 46152 }, { "epoch": 14.2, "learning_rate": 4.102010292482958e-06, "loss": 0.0101, "step": 46153 }, { "epoch": 14.2, "learning_rate": 4.101607921774357e-06, "loss": 0.0138, "step": 46154 }, { "epoch": 14.2, "learning_rate": 4.101205565709876e-06, "loss": 0.0114, "step": 46155 }, { "epoch": 14.2, "learning_rate": 4.100803224290512e-06, "loss": 0.0107, "step": 46156 }, { "epoch": 14.2, "learning_rate": 4.100400897517263e-06, "loss": 0.007, "step": 46157 }, { "epoch": 14.2, "learning_rate": 4.099998585391128e-06, "loss": 0.0149, "step": 46158 }, { "epoch": 14.2, "learning_rate": 4.099596287913106e-06, "loss": 0.0113, "step": 46159 }, { "epoch": 14.2, "learning_rate": 4.099194005084195e-06, "loss": 0.0187, "step": 46160 }, { "epoch": 14.2, "learning_rate": 4.098791736905395e-06, "loss": 0.0148, "step": 46161 }, { "epoch": 14.2, "learning_rate": 4.0983894833777035e-06, "loss": 0.015, "step": 46162 }, { "epoch": 14.2, "learning_rate": 4.097987244502121e-06, "loss": 0.0198, "step": 46163 }, { "epoch": 14.2, "learning_rate": 4.097585020279641e-06, "loss": 0.0061, "step": 46164 }, { "epoch": 14.2, "learning_rate": 4.09718281071127e-06, "loss": 0.0217, "step": 46165 }, { "epoch": 14.2, "learning_rate": 4.096780615798006e-06, "loss": 0.0133, "step": 46166 }, { "epoch": 14.21, "learning_rate": 4.096378435540841e-06, "loss": 0.0133, "step": 46167 }, { "epoch": 14.21, "learning_rate": 4.095976269940777e-06, "loss": 0.0085, "step": 46168 }, { "epoch": 14.21, "learning_rate": 4.095574118998807e-06, "loss": 0.0218, "step": 46169 }, { "epoch": 14.21, "learning_rate": 4.095171982715941e-06, "loss": 0.0085, "step": 46170 }, { "epoch": 14.21, "learning_rate": 4.094769861093169e-06, "loss": 0.0056, "step": 46171 }, { "epoch": 14.21, "learning_rate": 4.0943677541314915e-06, "loss": 0.0098, "step": 46172 }, { "epoch": 14.21, "learning_rate": 4.093965661831908e-06, "loss": 0.0167, "step": 46173 }, { "epoch": 14.21, "learning_rate": 4.093563584195415e-06, "loss": 0.0196, "step": 46174 }, { "epoch": 14.21, "learning_rate": 4.093161521223012e-06, "loss": 0.01, "step": 46175 }, { "epoch": 14.21, "learning_rate": 4.092759472915695e-06, "loss": 0.0063, "step": 46176 }, { "epoch": 14.21, "learning_rate": 4.0923574392744635e-06, "loss": 0.0125, "step": 46177 }, { "epoch": 14.21, "learning_rate": 4.091955420300318e-06, "loss": 0.0107, "step": 46178 }, { "epoch": 14.21, "learning_rate": 4.0915534159942515e-06, "loss": 0.0106, "step": 46179 }, { "epoch": 14.21, "learning_rate": 4.091151426357267e-06, "loss": 0.0199, "step": 46180 }, { "epoch": 14.21, "learning_rate": 4.09074945139036e-06, "loss": 0.0116, "step": 46181 }, { "epoch": 14.21, "learning_rate": 4.09034749109453e-06, "loss": 0.0179, "step": 46182 }, { "epoch": 14.21, "learning_rate": 4.089945545470772e-06, "loss": 0.0085, "step": 46183 }, { "epoch": 14.21, "learning_rate": 4.089543614520084e-06, "loss": 0.0119, "step": 46184 }, { "epoch": 14.21, "learning_rate": 4.089141698243468e-06, "loss": 0.0075, "step": 46185 }, { "epoch": 14.21, "learning_rate": 4.088739796641922e-06, "loss": 0.0113, "step": 46186 }, { "epoch": 14.21, "learning_rate": 4.088337909716443e-06, "loss": 0.0119, "step": 46187 }, { "epoch": 14.21, "learning_rate": 4.087936037468024e-06, "loss": 0.0096, "step": 46188 }, { "epoch": 14.21, "learning_rate": 4.087534179897663e-06, "loss": 0.01, "step": 46189 }, { "epoch": 14.21, "learning_rate": 4.087132337006363e-06, "loss": 0.0189, "step": 46190 }, { "epoch": 14.21, "learning_rate": 4.08673050879512e-06, "loss": 0.0159, "step": 46191 }, { "epoch": 14.21, "learning_rate": 4.086328695264931e-06, "loss": 0.0127, "step": 46192 }, { "epoch": 14.21, "learning_rate": 4.085926896416794e-06, "loss": 0.0112, "step": 46193 }, { "epoch": 14.21, "learning_rate": 4.085525112251706e-06, "loss": 0.0083, "step": 46194 }, { "epoch": 14.21, "learning_rate": 4.085123342770665e-06, "loss": 0.0141, "step": 46195 }, { "epoch": 14.21, "learning_rate": 4.084721587974667e-06, "loss": 0.0195, "step": 46196 }, { "epoch": 14.21, "learning_rate": 4.084319847864712e-06, "loss": 0.0132, "step": 46197 }, { "epoch": 14.21, "learning_rate": 4.083918122441795e-06, "loss": 0.0246, "step": 46198 }, { "epoch": 14.21, "learning_rate": 4.083516411706911e-06, "loss": 0.0122, "step": 46199 }, { "epoch": 14.22, "learning_rate": 4.083114715661069e-06, "loss": 0.0081, "step": 46200 }, { "epoch": 14.22, "learning_rate": 4.082713034305252e-06, "loss": 0.0271, "step": 46201 }, { "epoch": 14.22, "learning_rate": 4.082311367640465e-06, "loss": 0.0084, "step": 46202 }, { "epoch": 14.22, "learning_rate": 4.081909715667704e-06, "loss": 0.0149, "step": 46203 }, { "epoch": 14.22, "learning_rate": 4.08150807838796e-06, "loss": 0.0105, "step": 46204 }, { "epoch": 14.22, "learning_rate": 4.081106455802241e-06, "loss": 0.0083, "step": 46205 }, { "epoch": 14.22, "learning_rate": 4.080704847911539e-06, "loss": 0.0117, "step": 46206 }, { "epoch": 14.22, "learning_rate": 4.080303254716851e-06, "loss": 0.015, "step": 46207 }, { "epoch": 14.22, "learning_rate": 4.079901676219177e-06, "loss": 0.0084, "step": 46208 }, { "epoch": 14.22, "learning_rate": 4.079500112419504e-06, "loss": 0.0111, "step": 46209 }, { "epoch": 14.22, "learning_rate": 4.079098563318841e-06, "loss": 0.0103, "step": 46210 }, { "epoch": 14.22, "learning_rate": 4.07869702891818e-06, "loss": 0.0114, "step": 46211 }, { "epoch": 14.22, "learning_rate": 4.078295509218516e-06, "loss": 0.014, "step": 46212 }, { "epoch": 14.22, "learning_rate": 4.07789400422085e-06, "loss": 0.0218, "step": 46213 }, { "epoch": 14.22, "learning_rate": 4.077492513926174e-06, "loss": 0.0148, "step": 46214 }, { "epoch": 14.22, "learning_rate": 4.077091038335489e-06, "loss": 0.0043, "step": 46215 }, { "epoch": 14.22, "learning_rate": 4.076689577449791e-06, "loss": 0.0299, "step": 46216 }, { "epoch": 14.22, "learning_rate": 4.076288131270074e-06, "loss": 0.0077, "step": 46217 }, { "epoch": 14.22, "learning_rate": 4.075886699797338e-06, "loss": 0.0046, "step": 46218 }, { "epoch": 14.22, "learning_rate": 4.075485283032574e-06, "loss": 0.0067, "step": 46219 }, { "epoch": 14.22, "learning_rate": 4.0750838809767875e-06, "loss": 0.0043, "step": 46220 }, { "epoch": 14.22, "learning_rate": 4.074682493630973e-06, "loss": 0.0139, "step": 46221 }, { "epoch": 14.22, "learning_rate": 4.074281120996121e-06, "loss": 0.0185, "step": 46222 }, { "epoch": 14.22, "learning_rate": 4.0738797630732315e-06, "loss": 0.0217, "step": 46223 }, { "epoch": 14.22, "learning_rate": 4.073478419863298e-06, "loss": 0.0114, "step": 46224 }, { "epoch": 14.22, "learning_rate": 4.073077091367323e-06, "loss": 0.0102, "step": 46225 }, { "epoch": 14.22, "learning_rate": 4.072675777586299e-06, "loss": 0.024, "step": 46226 }, { "epoch": 14.22, "learning_rate": 4.072274478521223e-06, "loss": 0.0151, "step": 46227 }, { "epoch": 14.22, "learning_rate": 4.071873194173093e-06, "loss": 0.0177, "step": 46228 }, { "epoch": 14.22, "learning_rate": 4.0714719245429025e-06, "loss": 0.0159, "step": 46229 }, { "epoch": 14.22, "learning_rate": 4.0710706696316495e-06, "loss": 0.0155, "step": 46230 }, { "epoch": 14.22, "learning_rate": 4.070669429440329e-06, "loss": 0.0102, "step": 46231 }, { "epoch": 14.23, "learning_rate": 4.070268203969938e-06, "loss": 0.01, "step": 46232 }, { "epoch": 14.23, "learning_rate": 4.069866993221473e-06, "loss": 0.0136, "step": 46233 }, { "epoch": 14.23, "learning_rate": 4.069465797195925e-06, "loss": 0.0117, "step": 46234 }, { "epoch": 14.23, "learning_rate": 4.069064615894302e-06, "loss": 0.0203, "step": 46235 }, { "epoch": 14.23, "learning_rate": 4.0686634493175895e-06, "loss": 0.0099, "step": 46236 }, { "epoch": 14.23, "learning_rate": 4.068262297466786e-06, "loss": 0.0182, "step": 46237 }, { "epoch": 14.23, "learning_rate": 4.0678611603428884e-06, "loss": 0.0071, "step": 46238 }, { "epoch": 14.23, "learning_rate": 4.06746003794689e-06, "loss": 0.0189, "step": 46239 }, { "epoch": 14.23, "learning_rate": 4.0670589302797905e-06, "loss": 0.0185, "step": 46240 }, { "epoch": 14.23, "learning_rate": 4.066657837342586e-06, "loss": 0.0112, "step": 46241 }, { "epoch": 14.23, "learning_rate": 4.066256759136272e-06, "loss": 0.0085, "step": 46242 }, { "epoch": 14.23, "learning_rate": 4.065855695661841e-06, "loss": 0.0194, "step": 46243 }, { "epoch": 14.23, "learning_rate": 4.065454646920286e-06, "loss": 0.0261, "step": 46244 }, { "epoch": 14.23, "learning_rate": 4.065053612912612e-06, "loss": 0.0098, "step": 46245 }, { "epoch": 14.23, "learning_rate": 4.0646525936398086e-06, "loss": 0.0148, "step": 46246 }, { "epoch": 14.23, "learning_rate": 4.064251589102872e-06, "loss": 0.011, "step": 46247 }, { "epoch": 14.23, "learning_rate": 4.0638505993028e-06, "loss": 0.0058, "step": 46248 }, { "epoch": 14.23, "learning_rate": 4.0634496242405865e-06, "loss": 0.0197, "step": 46249 }, { "epoch": 14.23, "learning_rate": 4.0630486639172265e-06, "loss": 0.0132, "step": 46250 }, { "epoch": 14.23, "learning_rate": 4.062647718333716e-06, "loss": 0.0189, "step": 46251 }, { "epoch": 14.23, "learning_rate": 4.062246787491051e-06, "loss": 0.0189, "step": 46252 }, { "epoch": 14.23, "learning_rate": 4.061845871390226e-06, "loss": 0.0159, "step": 46253 }, { "epoch": 14.23, "learning_rate": 4.061444970032235e-06, "loss": 0.0135, "step": 46254 }, { "epoch": 14.23, "learning_rate": 4.0610440834180765e-06, "loss": 0.0154, "step": 46255 }, { "epoch": 14.23, "learning_rate": 4.060643211548749e-06, "loss": 0.0081, "step": 46256 }, { "epoch": 14.23, "learning_rate": 4.060242354425239e-06, "loss": 0.0167, "step": 46257 }, { "epoch": 14.23, "learning_rate": 4.059841512048547e-06, "loss": 0.013, "step": 46258 }, { "epoch": 14.23, "learning_rate": 4.0594406844196636e-06, "loss": 0.0105, "step": 46259 }, { "epoch": 14.23, "learning_rate": 4.0590398715395895e-06, "loss": 0.0129, "step": 46260 }, { "epoch": 14.23, "learning_rate": 4.058639073409319e-06, "loss": 0.0154, "step": 46261 }, { "epoch": 14.23, "learning_rate": 4.058238290029845e-06, "loss": 0.0239, "step": 46262 }, { "epoch": 14.23, "learning_rate": 4.057837521402168e-06, "loss": 0.0306, "step": 46263 }, { "epoch": 14.23, "learning_rate": 4.057436767527271e-06, "loss": 0.012, "step": 46264 }, { "epoch": 14.24, "learning_rate": 4.057036028406161e-06, "loss": 0.0145, "step": 46265 }, { "epoch": 14.24, "learning_rate": 4.056635304039826e-06, "loss": 0.0232, "step": 46266 }, { "epoch": 14.24, "learning_rate": 4.056234594429265e-06, "loss": 0.0154, "step": 46267 }, { "epoch": 14.24, "learning_rate": 4.055833899575471e-06, "loss": 0.0139, "step": 46268 }, { "epoch": 14.24, "learning_rate": 4.055433219479439e-06, "loss": 0.024, "step": 46269 }, { "epoch": 14.24, "learning_rate": 4.055032554142163e-06, "loss": 0.0146, "step": 46270 }, { "epoch": 14.24, "learning_rate": 4.054631903564638e-06, "loss": 0.015, "step": 46271 }, { "epoch": 14.24, "learning_rate": 4.054231267747862e-06, "loss": 0.0113, "step": 46272 }, { "epoch": 14.24, "learning_rate": 4.0538306466928234e-06, "loss": 0.0192, "step": 46273 }, { "epoch": 14.24, "learning_rate": 4.0534300404005176e-06, "loss": 0.0083, "step": 46274 }, { "epoch": 14.24, "learning_rate": 4.053029448871947e-06, "loss": 0.0153, "step": 46275 }, { "epoch": 14.24, "learning_rate": 4.0526288721081e-06, "loss": 0.0205, "step": 46276 }, { "epoch": 14.24, "learning_rate": 4.052228310109976e-06, "loss": 0.0184, "step": 46277 }, { "epoch": 14.24, "learning_rate": 4.051827762878561e-06, "loss": 0.0074, "step": 46278 }, { "epoch": 14.24, "learning_rate": 4.051427230414852e-06, "loss": 0.0108, "step": 46279 }, { "epoch": 14.24, "learning_rate": 4.051026712719848e-06, "loss": 0.0119, "step": 46280 }, { "epoch": 14.24, "learning_rate": 4.050626209794542e-06, "loss": 0.0125, "step": 46281 }, { "epoch": 14.24, "learning_rate": 4.050225721639925e-06, "loss": 0.0056, "step": 46282 }, { "epoch": 14.24, "learning_rate": 4.049825248256996e-06, "loss": 0.014, "step": 46283 }, { "epoch": 14.24, "learning_rate": 4.049424789646745e-06, "loss": 0.0087, "step": 46284 }, { "epoch": 14.24, "learning_rate": 4.049024345810169e-06, "loss": 0.0452, "step": 46285 }, { "epoch": 14.24, "learning_rate": 4.048623916748261e-06, "loss": 0.0209, "step": 46286 }, { "epoch": 14.24, "learning_rate": 4.048223502462015e-06, "loss": 0.0196, "step": 46287 }, { "epoch": 14.24, "learning_rate": 4.0478231029524265e-06, "loss": 0.0046, "step": 46288 }, { "epoch": 14.24, "learning_rate": 4.0474227182204836e-06, "loss": 0.0217, "step": 46289 }, { "epoch": 14.24, "learning_rate": 4.047022348267194e-06, "loss": 0.0159, "step": 46290 }, { "epoch": 14.24, "learning_rate": 4.046621993093538e-06, "loss": 0.0148, "step": 46291 }, { "epoch": 14.24, "learning_rate": 4.046221652700516e-06, "loss": 0.0141, "step": 46292 }, { "epoch": 14.24, "learning_rate": 4.04582132708912e-06, "loss": 0.0102, "step": 46293 }, { "epoch": 14.24, "learning_rate": 4.04542101626034e-06, "loss": 0.0122, "step": 46294 }, { "epoch": 14.24, "learning_rate": 4.045020720215179e-06, "loss": 0.016, "step": 46295 }, { "epoch": 14.24, "learning_rate": 4.0446204389546265e-06, "loss": 0.0185, "step": 46296 }, { "epoch": 14.25, "learning_rate": 4.044220172479675e-06, "loss": 0.0153, "step": 46297 }, { "epoch": 14.25, "learning_rate": 4.043819920791322e-06, "loss": 0.0276, "step": 46298 }, { "epoch": 14.25, "learning_rate": 4.043419683890552e-06, "loss": 0.0086, "step": 46299 }, { "epoch": 14.25, "learning_rate": 4.043019461778369e-06, "loss": 0.0154, "step": 46300 }, { "epoch": 14.25, "learning_rate": 4.042619254455763e-06, "loss": 0.0092, "step": 46301 }, { "epoch": 14.25, "learning_rate": 4.042219061923726e-06, "loss": 0.0089, "step": 46302 }, { "epoch": 14.25, "learning_rate": 4.041818884183255e-06, "loss": 0.0071, "step": 46303 }, { "epoch": 14.25, "learning_rate": 4.041418721235341e-06, "loss": 0.0079, "step": 46304 }, { "epoch": 14.25, "learning_rate": 4.041018573080976e-06, "loss": 0.0157, "step": 46305 }, { "epoch": 14.25, "learning_rate": 4.040618439721157e-06, "loss": 0.0163, "step": 46306 }, { "epoch": 14.25, "learning_rate": 4.040218321156875e-06, "loss": 0.0099, "step": 46307 }, { "epoch": 14.25, "learning_rate": 4.039818217389121e-06, "loss": 0.0096, "step": 46308 }, { "epoch": 14.25, "learning_rate": 4.0394181284188946e-06, "loss": 0.0174, "step": 46309 }, { "epoch": 14.25, "learning_rate": 4.039018054247188e-06, "loss": 0.0199, "step": 46310 }, { "epoch": 14.25, "learning_rate": 4.038617994874994e-06, "loss": 0.005, "step": 46311 }, { "epoch": 14.25, "learning_rate": 4.038217950303301e-06, "loss": 0.0116, "step": 46312 }, { "epoch": 14.25, "learning_rate": 4.037817920533103e-06, "loss": 0.0221, "step": 46313 }, { "epoch": 14.25, "learning_rate": 4.037417905565399e-06, "loss": 0.0116, "step": 46314 }, { "epoch": 14.25, "learning_rate": 4.037017905401179e-06, "loss": 0.0158, "step": 46315 }, { "epoch": 14.25, "learning_rate": 4.036617920041435e-06, "loss": 0.0082, "step": 46316 }, { "epoch": 14.25, "learning_rate": 4.036217949487162e-06, "loss": 0.0146, "step": 46317 }, { "epoch": 14.25, "learning_rate": 4.035817993739351e-06, "loss": 0.0099, "step": 46318 }, { "epoch": 14.25, "learning_rate": 4.035418052798999e-06, "loss": 0.0136, "step": 46319 }, { "epoch": 14.25, "learning_rate": 4.035018126667093e-06, "loss": 0.0153, "step": 46320 }, { "epoch": 14.25, "learning_rate": 4.034618215344629e-06, "loss": 0.0158, "step": 46321 }, { "epoch": 14.25, "learning_rate": 4.034218318832601e-06, "loss": 0.0085, "step": 46322 }, { "epoch": 14.25, "learning_rate": 4.033818437131997e-06, "loss": 0.0109, "step": 46323 }, { "epoch": 14.25, "learning_rate": 4.033418570243819e-06, "loss": 0.0088, "step": 46324 }, { "epoch": 14.25, "learning_rate": 4.033018718169051e-06, "loss": 0.0054, "step": 46325 }, { "epoch": 14.25, "learning_rate": 4.032618880908689e-06, "loss": 0.0131, "step": 46326 }, { "epoch": 14.25, "learning_rate": 4.032219058463726e-06, "loss": 0.0098, "step": 46327 }, { "epoch": 14.25, "learning_rate": 4.031819250835152e-06, "loss": 0.015, "step": 46328 }, { "epoch": 14.25, "learning_rate": 4.031419458023964e-06, "loss": 0.0132, "step": 46329 }, { "epoch": 14.26, "learning_rate": 4.031019680031152e-06, "loss": 0.0111, "step": 46330 }, { "epoch": 14.26, "learning_rate": 4.03061991685771e-06, "loss": 0.0146, "step": 46331 }, { "epoch": 14.26, "learning_rate": 4.030220168504632e-06, "loss": 0.0179, "step": 46332 }, { "epoch": 14.26, "learning_rate": 4.0298204349729005e-06, "loss": 0.0217, "step": 46333 }, { "epoch": 14.26, "learning_rate": 4.0294207162635205e-06, "loss": 0.0144, "step": 46334 }, { "epoch": 14.26, "learning_rate": 4.029021012377477e-06, "loss": 0.0194, "step": 46335 }, { "epoch": 14.26, "learning_rate": 4.0286213233157665e-06, "loss": 0.0074, "step": 46336 }, { "epoch": 14.26, "learning_rate": 4.028221649079379e-06, "loss": 0.0204, "step": 46337 }, { "epoch": 14.26, "learning_rate": 4.0278219896693075e-06, "loss": 0.0167, "step": 46338 }, { "epoch": 14.26, "learning_rate": 4.027422345086544e-06, "loss": 0.0136, "step": 46339 }, { "epoch": 14.26, "learning_rate": 4.027022715332079e-06, "loss": 0.0171, "step": 46340 }, { "epoch": 14.26, "learning_rate": 4.026623100406908e-06, "loss": 0.015, "step": 46341 }, { "epoch": 14.26, "learning_rate": 4.026223500312021e-06, "loss": 0.0274, "step": 46342 }, { "epoch": 14.26, "learning_rate": 4.025823915048407e-06, "loss": 0.0094, "step": 46343 }, { "epoch": 14.26, "learning_rate": 4.025424344617065e-06, "loss": 0.0162, "step": 46344 }, { "epoch": 14.26, "learning_rate": 4.0250247890189875e-06, "loss": 0.0112, "step": 46345 }, { "epoch": 14.26, "learning_rate": 4.02462524825516e-06, "loss": 0.0168, "step": 46346 }, { "epoch": 14.26, "learning_rate": 4.024225722326575e-06, "loss": 0.0157, "step": 46347 }, { "epoch": 14.26, "learning_rate": 4.023826211234224e-06, "loss": 0.0182, "step": 46348 }, { "epoch": 14.26, "learning_rate": 4.023426714979105e-06, "loss": 0.0216, "step": 46349 }, { "epoch": 14.26, "learning_rate": 4.0230272335622065e-06, "loss": 0.0098, "step": 46350 }, { "epoch": 14.26, "learning_rate": 4.0226277669845195e-06, "loss": 0.0137, "step": 46351 }, { "epoch": 14.26, "learning_rate": 4.022228315247037e-06, "loss": 0.0197, "step": 46352 }, { "epoch": 14.26, "learning_rate": 4.021828878350749e-06, "loss": 0.0173, "step": 46353 }, { "epoch": 14.26, "learning_rate": 4.021429456296649e-06, "loss": 0.0089, "step": 46354 }, { "epoch": 14.26, "learning_rate": 4.0210300490857275e-06, "loss": 0.0115, "step": 46355 }, { "epoch": 14.26, "learning_rate": 4.020630656718977e-06, "loss": 0.0192, "step": 46356 }, { "epoch": 14.26, "learning_rate": 4.020231279197389e-06, "loss": 0.0125, "step": 46357 }, { "epoch": 14.26, "learning_rate": 4.019831916521954e-06, "loss": 0.0173, "step": 46358 }, { "epoch": 14.26, "learning_rate": 4.019432568693663e-06, "loss": 0.014, "step": 46359 }, { "epoch": 14.26, "learning_rate": 4.01903323571351e-06, "loss": 0.0124, "step": 46360 }, { "epoch": 14.26, "learning_rate": 4.018633917582485e-06, "loss": 0.0125, "step": 46361 }, { "epoch": 14.27, "learning_rate": 4.01823461430158e-06, "loss": 0.0114, "step": 46362 }, { "epoch": 14.27, "learning_rate": 4.017835325871781e-06, "loss": 0.0222, "step": 46363 }, { "epoch": 14.27, "learning_rate": 4.017436052294088e-06, "loss": 0.0145, "step": 46364 }, { "epoch": 14.27, "learning_rate": 4.017036793569488e-06, "loss": 0.022, "step": 46365 }, { "epoch": 14.27, "learning_rate": 4.016637549698977e-06, "loss": 0.0109, "step": 46366 }, { "epoch": 14.27, "learning_rate": 4.0162383206835385e-06, "loss": 0.0149, "step": 46367 }, { "epoch": 14.27, "learning_rate": 4.015839106524163e-06, "loss": 0.0225, "step": 46368 }, { "epoch": 14.27, "learning_rate": 4.0154399072218495e-06, "loss": 0.021, "step": 46369 }, { "epoch": 14.27, "learning_rate": 4.015040722777585e-06, "loss": 0.0072, "step": 46370 }, { "epoch": 14.27, "learning_rate": 4.014641553192361e-06, "loss": 0.0248, "step": 46371 }, { "epoch": 14.27, "learning_rate": 4.014242398467169e-06, "loss": 0.0205, "step": 46372 }, { "epoch": 14.27, "learning_rate": 4.013843258603e-06, "loss": 0.0086, "step": 46373 }, { "epoch": 14.27, "learning_rate": 4.013444133600842e-06, "loss": 0.025, "step": 46374 }, { "epoch": 14.27, "learning_rate": 4.0130450234616905e-06, "loss": 0.0132, "step": 46375 }, { "epoch": 14.27, "learning_rate": 4.012645928186533e-06, "loss": 0.0142, "step": 46376 }, { "epoch": 14.27, "learning_rate": 4.012246847776362e-06, "loss": 0.0051, "step": 46377 }, { "epoch": 14.27, "learning_rate": 4.011847782232164e-06, "loss": 0.0135, "step": 46378 }, { "epoch": 14.27, "learning_rate": 4.011448731554941e-06, "loss": 0.0084, "step": 46379 }, { "epoch": 14.27, "learning_rate": 4.0110496957456725e-06, "loss": 0.0141, "step": 46380 }, { "epoch": 14.27, "learning_rate": 4.010650674805353e-06, "loss": 0.0236, "step": 46381 }, { "epoch": 14.27, "learning_rate": 4.010251668734974e-06, "loss": 0.0067, "step": 46382 }, { "epoch": 14.27, "learning_rate": 4.009852677535521e-06, "loss": 0.0122, "step": 46383 }, { "epoch": 14.27, "learning_rate": 4.009453701207992e-06, "loss": 0.0242, "step": 46384 }, { "epoch": 14.27, "learning_rate": 4.009054739753375e-06, "loss": 0.0117, "step": 46385 }, { "epoch": 14.27, "learning_rate": 4.008655793172661e-06, "loss": 0.0174, "step": 46386 }, { "epoch": 14.27, "learning_rate": 4.008256861466842e-06, "loss": 0.0183, "step": 46387 }, { "epoch": 14.27, "learning_rate": 4.007857944636898e-06, "loss": 0.0179, "step": 46388 }, { "epoch": 14.27, "learning_rate": 4.007459042683832e-06, "loss": 0.0167, "step": 46389 }, { "epoch": 14.27, "learning_rate": 4.007060155608629e-06, "loss": 0.018, "step": 46390 }, { "epoch": 14.27, "learning_rate": 4.006661283412281e-06, "loss": 0.0107, "step": 46391 }, { "epoch": 14.27, "learning_rate": 4.0062624260957765e-06, "loss": 0.0182, "step": 46392 }, { "epoch": 14.27, "learning_rate": 4.005863583660107e-06, "loss": 0.0037, "step": 46393 }, { "epoch": 14.27, "learning_rate": 4.0054647561062625e-06, "loss": 0.0073, "step": 46394 }, { "epoch": 14.28, "learning_rate": 4.005065943435232e-06, "loss": 0.0135, "step": 46395 }, { "epoch": 14.28, "learning_rate": 4.004667145648006e-06, "loss": 0.0294, "step": 46396 }, { "epoch": 14.28, "learning_rate": 4.0042683627455755e-06, "loss": 0.0181, "step": 46397 }, { "epoch": 14.28, "learning_rate": 4.003869594728929e-06, "loss": 0.0081, "step": 46398 }, { "epoch": 14.28, "learning_rate": 4.003470841599058e-06, "loss": 0.0149, "step": 46399 }, { "epoch": 14.28, "learning_rate": 4.003072103356958e-06, "loss": 0.0216, "step": 46400 }, { "epoch": 14.28, "learning_rate": 4.00267338000361e-06, "loss": 0.0145, "step": 46401 }, { "epoch": 14.28, "learning_rate": 4.002274671540006e-06, "loss": 0.0124, "step": 46402 }, { "epoch": 14.28, "learning_rate": 4.001875977967133e-06, "loss": 0.0095, "step": 46403 }, { "epoch": 14.28, "learning_rate": 4.001477299285989e-06, "loss": 0.0208, "step": 46404 }, { "epoch": 14.28, "learning_rate": 4.001078635497561e-06, "loss": 0.0248, "step": 46405 }, { "epoch": 14.28, "learning_rate": 4.0006799866028364e-06, "loss": 0.0115, "step": 46406 }, { "epoch": 14.28, "learning_rate": 4.000281352602806e-06, "loss": 0.0124, "step": 46407 }, { "epoch": 14.28, "learning_rate": 3.999882733498461e-06, "loss": 0.0102, "step": 46408 }, { "epoch": 14.28, "learning_rate": 3.999484129290787e-06, "loss": 0.0151, "step": 46409 }, { "epoch": 14.28, "learning_rate": 3.9990855399807775e-06, "loss": 0.0105, "step": 46410 }, { "epoch": 14.28, "learning_rate": 3.998686965569421e-06, "loss": 0.0074, "step": 46411 }, { "epoch": 14.28, "learning_rate": 3.998288406057707e-06, "loss": 0.0121, "step": 46412 }, { "epoch": 14.28, "learning_rate": 3.99788986144662e-06, "loss": 0.01, "step": 46413 }, { "epoch": 14.28, "learning_rate": 3.997491331737162e-06, "loss": 0.0139, "step": 46414 }, { "epoch": 14.28, "learning_rate": 3.997092816930313e-06, "loss": 0.0239, "step": 46415 }, { "epoch": 14.28, "learning_rate": 3.9966943170270625e-06, "loss": 0.017, "step": 46416 }, { "epoch": 14.28, "learning_rate": 3.996295832028402e-06, "loss": 0.0116, "step": 46417 }, { "epoch": 14.28, "learning_rate": 3.995897361935316e-06, "loss": 0.0116, "step": 46418 }, { "epoch": 14.28, "learning_rate": 3.995498906748803e-06, "loss": 0.0126, "step": 46419 }, { "epoch": 14.28, "learning_rate": 3.9951004664698465e-06, "loss": 0.013, "step": 46420 }, { "epoch": 14.28, "learning_rate": 3.994702041099441e-06, "loss": 0.0189, "step": 46421 }, { "epoch": 14.28, "learning_rate": 3.994303630638568e-06, "loss": 0.007, "step": 46422 }, { "epoch": 14.28, "learning_rate": 3.9939052350882156e-06, "loss": 0.0082, "step": 46423 }, { "epoch": 14.28, "learning_rate": 3.993506854449382e-06, "loss": 0.011, "step": 46424 }, { "epoch": 14.28, "learning_rate": 3.99310848872305e-06, "loss": 0.0172, "step": 46425 }, { "epoch": 14.28, "learning_rate": 3.99271013791021e-06, "loss": 0.0175, "step": 46426 }, { "epoch": 14.29, "learning_rate": 3.9923118020118535e-06, "loss": 0.0165, "step": 46427 }, { "epoch": 14.29, "learning_rate": 3.991913481028965e-06, "loss": 0.0153, "step": 46428 }, { "epoch": 14.29, "learning_rate": 3.991515174962537e-06, "loss": 0.0187, "step": 46429 }, { "epoch": 14.29, "learning_rate": 3.991116883813556e-06, "loss": 0.0096, "step": 46430 }, { "epoch": 14.29, "learning_rate": 3.990718607583013e-06, "loss": 0.0151, "step": 46431 }, { "epoch": 14.29, "learning_rate": 3.990320346271895e-06, "loss": 0.016, "step": 46432 }, { "epoch": 14.29, "learning_rate": 3.989922099881187e-06, "loss": 0.0381, "step": 46433 }, { "epoch": 14.29, "learning_rate": 3.989523868411886e-06, "loss": 0.0181, "step": 46434 }, { "epoch": 14.29, "learning_rate": 3.989125651864982e-06, "loss": 0.0171, "step": 46435 }, { "epoch": 14.29, "learning_rate": 3.988727450241454e-06, "loss": 0.0103, "step": 46436 }, { "epoch": 14.29, "learning_rate": 3.988329263542295e-06, "loss": 0.0147, "step": 46437 }, { "epoch": 14.29, "learning_rate": 3.98793109176849e-06, "loss": 0.012, "step": 46438 }, { "epoch": 14.29, "learning_rate": 3.9875329349210354e-06, "loss": 0.0143, "step": 46439 }, { "epoch": 14.29, "learning_rate": 3.9871347930009154e-06, "loss": 0.0226, "step": 46440 }, { "epoch": 14.29, "learning_rate": 3.9867366660091176e-06, "loss": 0.013, "step": 46441 }, { "epoch": 14.29, "learning_rate": 3.986338553946637e-06, "loss": 0.0138, "step": 46442 }, { "epoch": 14.29, "learning_rate": 3.985940456814448e-06, "loss": 0.0094, "step": 46443 }, { "epoch": 14.29, "learning_rate": 3.985542374613552e-06, "loss": 0.0122, "step": 46444 }, { "epoch": 14.29, "learning_rate": 3.985144307344932e-06, "loss": 0.0069, "step": 46445 }, { "epoch": 14.29, "learning_rate": 3.984746255009577e-06, "loss": 0.0119, "step": 46446 }, { "epoch": 14.29, "learning_rate": 3.984348217608477e-06, "loss": 0.0109, "step": 46447 }, { "epoch": 14.29, "learning_rate": 3.983950195142618e-06, "loss": 0.0148, "step": 46448 }, { "epoch": 14.29, "learning_rate": 3.983552187612988e-06, "loss": 0.0098, "step": 46449 }, { "epoch": 14.29, "learning_rate": 3.9831541950205755e-06, "loss": 0.0096, "step": 46450 }, { "epoch": 14.29, "learning_rate": 3.982756217366371e-06, "loss": 0.012, "step": 46451 }, { "epoch": 14.29, "learning_rate": 3.982358254651359e-06, "loss": 0.0199, "step": 46452 }, { "epoch": 14.29, "learning_rate": 3.981960306876526e-06, "loss": 0.0117, "step": 46453 }, { "epoch": 14.29, "learning_rate": 3.981562374042867e-06, "loss": 0.018, "step": 46454 }, { "epoch": 14.29, "learning_rate": 3.9811644561513655e-06, "loss": 0.012, "step": 46455 }, { "epoch": 14.29, "learning_rate": 3.980766553203014e-06, "loss": 0.0126, "step": 46456 }, { "epoch": 14.29, "learning_rate": 3.980368665198791e-06, "loss": 0.0109, "step": 46457 }, { "epoch": 14.29, "learning_rate": 3.979970792139689e-06, "loss": 0.0221, "step": 46458 }, { "epoch": 14.29, "learning_rate": 3.979572934026699e-06, "loss": 0.0093, "step": 46459 }, { "epoch": 14.3, "learning_rate": 3.9791750908608066e-06, "loss": 0.0183, "step": 46460 }, { "epoch": 14.3, "learning_rate": 3.978777262642999e-06, "loss": 0.0104, "step": 46461 }, { "epoch": 14.3, "learning_rate": 3.978379449374264e-06, "loss": 0.016, "step": 46462 }, { "epoch": 14.3, "learning_rate": 3.97798165105559e-06, "loss": 0.0101, "step": 46463 }, { "epoch": 14.3, "learning_rate": 3.977583867687964e-06, "loss": 0.0135, "step": 46464 }, { "epoch": 14.3, "learning_rate": 3.977186099272373e-06, "loss": 0.0122, "step": 46465 }, { "epoch": 14.3, "learning_rate": 3.976788345809805e-06, "loss": 0.0222, "step": 46466 }, { "epoch": 14.3, "learning_rate": 3.976390607301249e-06, "loss": 0.0091, "step": 46467 }, { "epoch": 14.3, "learning_rate": 3.975992883747688e-06, "loss": 0.024, "step": 46468 }, { "epoch": 14.3, "learning_rate": 3.9755951751501184e-06, "loss": 0.0203, "step": 46469 }, { "epoch": 14.3, "learning_rate": 3.975197481509518e-06, "loss": 0.0267, "step": 46470 }, { "epoch": 14.3, "learning_rate": 3.97479980282688e-06, "loss": 0.0155, "step": 46471 }, { "epoch": 14.3, "learning_rate": 3.974402139103189e-06, "loss": 0.0135, "step": 46472 }, { "epoch": 14.3, "learning_rate": 3.9740044903394285e-06, "loss": 0.0204, "step": 46473 }, { "epoch": 14.3, "learning_rate": 3.9736068565365935e-06, "loss": 0.0098, "step": 46474 }, { "epoch": 14.3, "learning_rate": 3.973209237695669e-06, "loss": 0.0149, "step": 46475 }, { "epoch": 14.3, "learning_rate": 3.972811633817645e-06, "loss": 0.0168, "step": 46476 }, { "epoch": 14.3, "learning_rate": 3.972414044903501e-06, "loss": 0.026, "step": 46477 }, { "epoch": 14.3, "learning_rate": 3.972016470954224e-06, "loss": 0.0062, "step": 46478 }, { "epoch": 14.3, "learning_rate": 3.971618911970808e-06, "loss": 0.0144, "step": 46479 }, { "epoch": 14.3, "learning_rate": 3.971221367954239e-06, "loss": 0.0173, "step": 46480 }, { "epoch": 14.3, "learning_rate": 3.9708238389055e-06, "loss": 0.0087, "step": 46481 }, { "epoch": 14.3, "learning_rate": 3.970426324825581e-06, "loss": 0.0124, "step": 46482 }, { "epoch": 14.3, "learning_rate": 3.970028825715468e-06, "loss": 0.0177, "step": 46483 }, { "epoch": 14.3, "learning_rate": 3.969631341576148e-06, "loss": 0.0258, "step": 46484 }, { "epoch": 14.3, "learning_rate": 3.969233872408606e-06, "loss": 0.0051, "step": 46485 }, { "epoch": 14.3, "learning_rate": 3.968836418213832e-06, "loss": 0.0143, "step": 46486 }, { "epoch": 14.3, "learning_rate": 3.968438978992811e-06, "loss": 0.0079, "step": 46487 }, { "epoch": 14.3, "learning_rate": 3.968041554746527e-06, "loss": 0.0184, "step": 46488 }, { "epoch": 14.3, "learning_rate": 3.967644145475972e-06, "loss": 0.0121, "step": 46489 }, { "epoch": 14.3, "learning_rate": 3.967246751182135e-06, "loss": 0.0167, "step": 46490 }, { "epoch": 14.3, "learning_rate": 3.9668493718659924e-06, "loss": 0.0133, "step": 46491 }, { "epoch": 14.31, "learning_rate": 3.9664520075285385e-06, "loss": 0.0093, "step": 46492 }, { "epoch": 14.31, "learning_rate": 3.966054658170754e-06, "loss": 0.0134, "step": 46493 }, { "epoch": 14.31, "learning_rate": 3.965657323793631e-06, "loss": 0.0122, "step": 46494 }, { "epoch": 14.31, "learning_rate": 3.965260004398155e-06, "loss": 0.0058, "step": 46495 }, { "epoch": 14.31, "learning_rate": 3.964862699985312e-06, "loss": 0.0148, "step": 46496 }, { "epoch": 14.31, "learning_rate": 3.964465410556091e-06, "loss": 0.019, "step": 46497 }, { "epoch": 14.31, "learning_rate": 3.964068136111467e-06, "loss": 0.0136, "step": 46498 }, { "epoch": 14.31, "learning_rate": 3.96367087665244e-06, "loss": 0.0081, "step": 46499 }, { "epoch": 14.31, "learning_rate": 3.96327363217999e-06, "loss": 0.0126, "step": 46500 }, { "epoch": 14.31, "learning_rate": 3.9628764026951036e-06, "loss": 0.0092, "step": 46501 }, { "epoch": 14.31, "learning_rate": 3.962479188198769e-06, "loss": 0.0081, "step": 46502 }, { "epoch": 14.31, "learning_rate": 3.96208198869197e-06, "loss": 0.0294, "step": 46503 }, { "epoch": 14.31, "learning_rate": 3.961684804175694e-06, "loss": 0.0098, "step": 46504 }, { "epoch": 14.31, "learning_rate": 3.9612876346509265e-06, "loss": 0.0151, "step": 46505 }, { "epoch": 14.31, "learning_rate": 3.960890480118653e-06, "loss": 0.0056, "step": 46506 }, { "epoch": 14.31, "learning_rate": 3.960493340579861e-06, "loss": 0.0157, "step": 46507 }, { "epoch": 14.31, "learning_rate": 3.9600962160355324e-06, "loss": 0.0054, "step": 46508 }, { "epoch": 14.31, "learning_rate": 3.9596991064866606e-06, "loss": 0.0108, "step": 46509 }, { "epoch": 14.31, "learning_rate": 3.9593020119342275e-06, "loss": 0.0073, "step": 46510 }, { "epoch": 14.31, "learning_rate": 3.958904932379222e-06, "loss": 0.0168, "step": 46511 }, { "epoch": 14.31, "learning_rate": 3.958507867822623e-06, "loss": 0.008, "step": 46512 }, { "epoch": 14.31, "learning_rate": 3.958110818265417e-06, "loss": 0.0109, "step": 46513 }, { "epoch": 14.31, "learning_rate": 3.957713783708596e-06, "loss": 0.0204, "step": 46514 }, { "epoch": 14.31, "learning_rate": 3.957316764153144e-06, "loss": 0.0084, "step": 46515 }, { "epoch": 14.31, "learning_rate": 3.956919759600045e-06, "loss": 0.032, "step": 46516 }, { "epoch": 14.31, "learning_rate": 3.956522770050285e-06, "loss": 0.0059, "step": 46517 }, { "epoch": 14.31, "learning_rate": 3.956125795504849e-06, "loss": 0.0117, "step": 46518 }, { "epoch": 14.31, "learning_rate": 3.955728835964724e-06, "loss": 0.0129, "step": 46519 }, { "epoch": 14.31, "learning_rate": 3.955331891430895e-06, "loss": 0.0107, "step": 46520 }, { "epoch": 14.31, "learning_rate": 3.955331891430895e-06, "loss": 0.0196, "step": 46521 }, { "epoch": 14.31, "learning_rate": 3.954934961904347e-06, "loss": 0.02, "step": 46522 }, { "epoch": 14.31, "learning_rate": 3.954538047386066e-06, "loss": 0.0214, "step": 46523 }, { "epoch": 14.31, "learning_rate": 3.954141147877033e-06, "loss": 0.006, "step": 46524 }, { "epoch": 14.32, "learning_rate": 3.953744263378245e-06, "loss": 0.0162, "step": 46525 }, { "epoch": 14.32, "learning_rate": 3.9533473938906765e-06, "loss": 0.0105, "step": 46526 }, { "epoch": 14.32, "learning_rate": 3.952950539415317e-06, "loss": 0.0164, "step": 46527 }, { "epoch": 14.32, "learning_rate": 3.95255369995315e-06, "loss": 0.012, "step": 46528 }, { "epoch": 14.32, "learning_rate": 3.952156875505159e-06, "loss": 0.0129, "step": 46529 }, { "epoch": 14.32, "learning_rate": 3.9517600660723355e-06, "loss": 0.0062, "step": 46530 }, { "epoch": 14.32, "learning_rate": 3.9513632716556605e-06, "loss": 0.0135, "step": 46531 }, { "epoch": 14.32, "learning_rate": 3.950966492256121e-06, "loss": 0.0135, "step": 46532 }, { "epoch": 14.32, "learning_rate": 3.950569727874704e-06, "loss": 0.0066, "step": 46533 }, { "epoch": 14.32, "learning_rate": 3.950172978512383e-06, "loss": 0.0129, "step": 46534 }, { "epoch": 14.32, "learning_rate": 3.949776244170157e-06, "loss": 0.0115, "step": 46535 }, { "epoch": 14.32, "learning_rate": 3.9493795248490045e-06, "loss": 0.0094, "step": 46536 }, { "epoch": 14.32, "learning_rate": 3.948982820549911e-06, "loss": 0.0118, "step": 46537 }, { "epoch": 14.32, "learning_rate": 3.948586131273864e-06, "loss": 0.0157, "step": 46538 }, { "epoch": 14.32, "learning_rate": 3.948189457021844e-06, "loss": 0.0195, "step": 46539 }, { "epoch": 14.32, "learning_rate": 3.947792797794839e-06, "loss": 0.0073, "step": 46540 }, { "epoch": 14.32, "learning_rate": 3.947396153593833e-06, "loss": 0.0158, "step": 46541 }, { "epoch": 14.32, "learning_rate": 3.94699952441981e-06, "loss": 0.0219, "step": 46542 }, { "epoch": 14.32, "learning_rate": 3.946602910273755e-06, "loss": 0.0083, "step": 46543 }, { "epoch": 14.32, "learning_rate": 3.94620631115665e-06, "loss": 0.0121, "step": 46544 }, { "epoch": 14.32, "learning_rate": 3.945809727069487e-06, "loss": 0.0199, "step": 46545 }, { "epoch": 14.32, "learning_rate": 3.945413158013249e-06, "loss": 0.0126, "step": 46546 }, { "epoch": 14.32, "learning_rate": 3.945016603988915e-06, "loss": 0.0192, "step": 46547 }, { "epoch": 14.32, "learning_rate": 3.944620064997472e-06, "loss": 0.0165, "step": 46548 }, { "epoch": 14.32, "learning_rate": 3.944223541039901e-06, "loss": 0.0219, "step": 46549 }, { "epoch": 14.32, "learning_rate": 3.943827032117195e-06, "loss": 0.0079, "step": 46550 }, { "epoch": 14.32, "learning_rate": 3.943430538230333e-06, "loss": 0.0131, "step": 46551 }, { "epoch": 14.32, "learning_rate": 3.943034059380299e-06, "loss": 0.0131, "step": 46552 }, { "epoch": 14.32, "learning_rate": 3.942637595568081e-06, "loss": 0.0112, "step": 46553 }, { "epoch": 14.32, "learning_rate": 3.9422411467946585e-06, "loss": 0.0119, "step": 46554 }, { "epoch": 14.32, "learning_rate": 3.941844713061019e-06, "loss": 0.0066, "step": 46555 }, { "epoch": 14.32, "learning_rate": 3.941448294368146e-06, "loss": 0.0091, "step": 46556 }, { "epoch": 14.33, "learning_rate": 3.941051890717024e-06, "loss": 0.0122, "step": 46557 }, { "epoch": 14.33, "learning_rate": 3.940655502108635e-06, "loss": 0.0106, "step": 46558 }, { "epoch": 14.33, "learning_rate": 3.940259128543967e-06, "loss": 0.0224, "step": 46559 }, { "epoch": 14.33, "learning_rate": 3.939862770024001e-06, "loss": 0.0124, "step": 46560 }, { "epoch": 14.33, "learning_rate": 3.939466426549722e-06, "loss": 0.0149, "step": 46561 }, { "epoch": 14.33, "learning_rate": 3.939070098122114e-06, "loss": 0.0117, "step": 46562 }, { "epoch": 14.33, "learning_rate": 3.938673784742161e-06, "loss": 0.007, "step": 46563 }, { "epoch": 14.33, "learning_rate": 3.938277486410843e-06, "loss": 0.0155, "step": 46564 }, { "epoch": 14.33, "learning_rate": 3.937881203129151e-06, "loss": 0.0106, "step": 46565 }, { "epoch": 14.33, "learning_rate": 3.937484934898067e-06, "loss": 0.013, "step": 46566 }, { "epoch": 14.33, "learning_rate": 3.937088681718576e-06, "loss": 0.0198, "step": 46567 }, { "epoch": 14.33, "learning_rate": 3.9366924435916565e-06, "loss": 0.0224, "step": 46568 }, { "epoch": 14.33, "learning_rate": 3.936296220518291e-06, "loss": 0.0088, "step": 46569 }, { "epoch": 14.33, "learning_rate": 3.935900012499472e-06, "loss": 0.0107, "step": 46570 }, { "epoch": 14.33, "learning_rate": 3.935503819536176e-06, "loss": 0.0173, "step": 46571 }, { "epoch": 14.33, "learning_rate": 3.935107641629391e-06, "loss": 0.0098, "step": 46572 }, { "epoch": 14.33, "learning_rate": 3.934711478780098e-06, "loss": 0.0066, "step": 46573 }, { "epoch": 14.33, "learning_rate": 3.934315330989281e-06, "loss": 0.0114, "step": 46574 }, { "epoch": 14.33, "learning_rate": 3.933919198257925e-06, "loss": 0.0129, "step": 46575 }, { "epoch": 14.33, "learning_rate": 3.93352308058701e-06, "loss": 0.0102, "step": 46576 }, { "epoch": 14.33, "learning_rate": 3.933126977977524e-06, "loss": 0.0173, "step": 46577 }, { "epoch": 14.33, "learning_rate": 3.932730890430447e-06, "loss": 0.0067, "step": 46578 }, { "epoch": 14.33, "learning_rate": 3.932334817946761e-06, "loss": 0.0167, "step": 46579 }, { "epoch": 14.33, "learning_rate": 3.931938760527458e-06, "loss": 0.0136, "step": 46580 }, { "epoch": 14.33, "learning_rate": 3.931542718173512e-06, "loss": 0.0124, "step": 46581 }, { "epoch": 14.33, "learning_rate": 3.9311466908859085e-06, "loss": 0.019, "step": 46582 }, { "epoch": 14.33, "learning_rate": 3.930750678665632e-06, "loss": 0.0118, "step": 46583 }, { "epoch": 14.33, "learning_rate": 3.930354681513663e-06, "loss": 0.0201, "step": 46584 }, { "epoch": 14.33, "learning_rate": 3.9299586994309905e-06, "loss": 0.0076, "step": 46585 }, { "epoch": 14.33, "learning_rate": 3.929562732418592e-06, "loss": 0.0151, "step": 46586 }, { "epoch": 14.33, "learning_rate": 3.9291667804774545e-06, "loss": 0.0101, "step": 46587 }, { "epoch": 14.33, "learning_rate": 3.9287708436085624e-06, "loss": 0.0115, "step": 46588 }, { "epoch": 14.33, "learning_rate": 3.9283749218128885e-06, "loss": 0.0087, "step": 46589 }, { "epoch": 14.34, "learning_rate": 3.927979015091426e-06, "loss": 0.0165, "step": 46590 }, { "epoch": 14.34, "learning_rate": 3.927583123445156e-06, "loss": 0.0192, "step": 46591 }, { "epoch": 14.34, "learning_rate": 3.927187246875058e-06, "loss": 0.0128, "step": 46592 }, { "epoch": 14.34, "learning_rate": 3.926791385382118e-06, "loss": 0.0139, "step": 46593 }, { "epoch": 14.34, "learning_rate": 3.926395538967317e-06, "loss": 0.0124, "step": 46594 }, { "epoch": 14.34, "learning_rate": 3.925999707631639e-06, "loss": 0.0079, "step": 46595 }, { "epoch": 14.34, "learning_rate": 3.9256038913760665e-06, "loss": 0.0164, "step": 46596 }, { "epoch": 14.34, "learning_rate": 3.925208090201581e-06, "loss": 0.011, "step": 46597 }, { "epoch": 14.34, "learning_rate": 3.924812304109167e-06, "loss": 0.0189, "step": 46598 }, { "epoch": 14.34, "learning_rate": 3.924416533099803e-06, "loss": 0.0103, "step": 46599 }, { "epoch": 14.34, "learning_rate": 3.924020777174477e-06, "loss": 0.0141, "step": 46600 }, { "epoch": 14.34, "learning_rate": 3.923625036334175e-06, "loss": 0.02, "step": 46601 }, { "epoch": 14.34, "learning_rate": 3.923229310579868e-06, "loss": 0.0196, "step": 46602 }, { "epoch": 14.34, "learning_rate": 3.922833599912545e-06, "loss": 0.0088, "step": 46603 }, { "epoch": 14.34, "learning_rate": 3.922437904333185e-06, "loss": 0.0124, "step": 46604 }, { "epoch": 14.34, "learning_rate": 3.922042223842777e-06, "loss": 0.0059, "step": 46605 }, { "epoch": 14.34, "learning_rate": 3.921646558442299e-06, "loss": 0.0199, "step": 46606 }, { "epoch": 14.34, "learning_rate": 3.9212509081327346e-06, "loss": 0.01, "step": 46607 }, { "epoch": 14.34, "learning_rate": 3.920855272915064e-06, "loss": 0.0126, "step": 46608 }, { "epoch": 14.34, "learning_rate": 3.920459652790271e-06, "loss": 0.0136, "step": 46609 }, { "epoch": 14.34, "learning_rate": 3.9200640477593384e-06, "loss": 0.0115, "step": 46610 }, { "epoch": 14.34, "learning_rate": 3.919668457823248e-06, "loss": 0.0097, "step": 46611 }, { "epoch": 14.34, "learning_rate": 3.919272882982982e-06, "loss": 0.0164, "step": 46612 }, { "epoch": 14.34, "learning_rate": 3.9188773232395205e-06, "loss": 0.0088, "step": 46613 }, { "epoch": 14.34, "learning_rate": 3.918481778593849e-06, "loss": 0.0143, "step": 46614 }, { "epoch": 14.34, "learning_rate": 3.918086249046946e-06, "loss": 0.0131, "step": 46615 }, { "epoch": 14.34, "learning_rate": 3.917690734599796e-06, "loss": 0.0126, "step": 46616 }, { "epoch": 14.34, "learning_rate": 3.91729523525338e-06, "loss": 0.0095, "step": 46617 }, { "epoch": 14.34, "learning_rate": 3.916899751008681e-06, "loss": 0.0074, "step": 46618 }, { "epoch": 14.34, "learning_rate": 3.916504281866675e-06, "loss": 0.017, "step": 46619 }, { "epoch": 14.34, "learning_rate": 3.916108827828353e-06, "loss": 0.0194, "step": 46620 }, { "epoch": 14.34, "learning_rate": 3.915713388894693e-06, "loss": 0.0126, "step": 46621 }, { "epoch": 14.35, "learning_rate": 3.9153179650666796e-06, "loss": 0.0125, "step": 46622 }, { "epoch": 14.35, "learning_rate": 3.9149225563452875e-06, "loss": 0.0117, "step": 46623 }, { "epoch": 14.35, "learning_rate": 3.914527162731498e-06, "loss": 0.0129, "step": 46624 }, { "epoch": 14.35, "learning_rate": 3.914131784226301e-06, "loss": 0.0165, "step": 46625 }, { "epoch": 14.35, "learning_rate": 3.913736420830676e-06, "loss": 0.0192, "step": 46626 }, { "epoch": 14.35, "learning_rate": 3.913341072545601e-06, "loss": 0.0157, "step": 46627 }, { "epoch": 14.35, "learning_rate": 3.9129457393720595e-06, "loss": 0.0195, "step": 46628 }, { "epoch": 14.35, "learning_rate": 3.912550421311032e-06, "loss": 0.0143, "step": 46629 }, { "epoch": 14.35, "learning_rate": 3.912155118363501e-06, "loss": 0.0132, "step": 46630 }, { "epoch": 14.35, "learning_rate": 3.911759830530448e-06, "loss": 0.0328, "step": 46631 }, { "epoch": 14.35, "learning_rate": 3.911364557812853e-06, "loss": 0.0214, "step": 46632 }, { "epoch": 14.35, "learning_rate": 3.9109693002117e-06, "loss": 0.0135, "step": 46633 }, { "epoch": 14.35, "learning_rate": 3.9105740577279635e-06, "loss": 0.003, "step": 46634 }, { "epoch": 14.35, "learning_rate": 3.9101788303626384e-06, "loss": 0.0153, "step": 46635 }, { "epoch": 14.35, "learning_rate": 3.909783618116694e-06, "loss": 0.0233, "step": 46636 }, { "epoch": 14.35, "learning_rate": 3.909388420991113e-06, "loss": 0.0217, "step": 46637 }, { "epoch": 14.35, "learning_rate": 3.908993238986881e-06, "loss": 0.0141, "step": 46638 }, { "epoch": 14.35, "learning_rate": 3.908598072104972e-06, "loss": 0.013, "step": 46639 }, { "epoch": 14.35, "learning_rate": 3.908202920346375e-06, "loss": 0.0189, "step": 46640 }, { "epoch": 14.35, "learning_rate": 3.9078077837120685e-06, "loss": 0.0182, "step": 46641 }, { "epoch": 14.35, "learning_rate": 3.9074126622030326e-06, "loss": 0.0095, "step": 46642 }, { "epoch": 14.35, "learning_rate": 3.907017555820253e-06, "loss": 0.0158, "step": 46643 }, { "epoch": 14.35, "learning_rate": 3.9066224645646976e-06, "loss": 0.0131, "step": 46644 }, { "epoch": 14.35, "learning_rate": 3.90622738843736e-06, "loss": 0.0128, "step": 46645 }, { "epoch": 14.35, "learning_rate": 3.905832327439217e-06, "loss": 0.0158, "step": 46646 }, { "epoch": 14.35, "learning_rate": 3.905437281571251e-06, "loss": 0.0159, "step": 46647 }, { "epoch": 14.35, "learning_rate": 3.905042250834439e-06, "loss": 0.0077, "step": 46648 }, { "epoch": 14.35, "learning_rate": 3.904647235229765e-06, "loss": 0.0112, "step": 46649 }, { "epoch": 14.35, "learning_rate": 3.904252234758208e-06, "loss": 0.0245, "step": 46650 }, { "epoch": 14.35, "learning_rate": 3.9038572494207495e-06, "loss": 0.0226, "step": 46651 }, { "epoch": 14.35, "learning_rate": 3.903462279218369e-06, "loss": 0.0389, "step": 46652 }, { "epoch": 14.35, "learning_rate": 3.90306732415205e-06, "loss": 0.0078, "step": 46653 }, { "epoch": 14.35, "learning_rate": 3.902672384222767e-06, "loss": 0.0147, "step": 46654 }, { "epoch": 14.36, "learning_rate": 3.902277459431509e-06, "loss": 0.0068, "step": 46655 }, { "epoch": 14.36, "learning_rate": 3.901882549779253e-06, "loss": 0.017, "step": 46656 }, { "epoch": 14.36, "learning_rate": 3.901487655266978e-06, "loss": 0.0115, "step": 46657 }, { "epoch": 14.36, "learning_rate": 3.901092775895663e-06, "loss": 0.0169, "step": 46658 }, { "epoch": 14.36, "learning_rate": 3.900697911666288e-06, "loss": 0.0159, "step": 46659 }, { "epoch": 14.36, "learning_rate": 3.900303062579839e-06, "loss": 0.0037, "step": 46660 }, { "epoch": 14.36, "learning_rate": 3.8999082286372935e-06, "loss": 0.0144, "step": 46661 }, { "epoch": 14.36, "learning_rate": 3.89951340983963e-06, "loss": 0.0179, "step": 46662 }, { "epoch": 14.36, "learning_rate": 3.899118606187832e-06, "loss": 0.0234, "step": 46663 }, { "epoch": 14.36, "learning_rate": 3.898723817682877e-06, "loss": 0.0172, "step": 46664 }, { "epoch": 14.36, "learning_rate": 3.898329044325745e-06, "loss": 0.0074, "step": 46665 }, { "epoch": 14.36, "learning_rate": 3.897934286117417e-06, "loss": 0.0171, "step": 46666 }, { "epoch": 14.36, "learning_rate": 3.897539543058875e-06, "loss": 0.013, "step": 46667 }, { "epoch": 14.36, "learning_rate": 3.897144815151096e-06, "loss": 0.0215, "step": 46668 }, { "epoch": 14.36, "learning_rate": 3.896750102395057e-06, "loss": 0.0124, "step": 46669 }, { "epoch": 14.36, "learning_rate": 3.89635540479175e-06, "loss": 0.0196, "step": 46670 }, { "epoch": 14.36, "learning_rate": 3.895960722342144e-06, "loss": 0.0135, "step": 46671 }, { "epoch": 14.36, "learning_rate": 3.895566055047221e-06, "loss": 0.0106, "step": 46672 }, { "epoch": 14.36, "learning_rate": 3.8951714029079615e-06, "loss": 0.0193, "step": 46673 }, { "epoch": 14.36, "learning_rate": 3.894776765925342e-06, "loss": 0.0151, "step": 46674 }, { "epoch": 14.36, "learning_rate": 3.8943821441003505e-06, "loss": 0.0181, "step": 46675 }, { "epoch": 14.36, "learning_rate": 3.893987537433961e-06, "loss": 0.006, "step": 46676 }, { "epoch": 14.36, "learning_rate": 3.8935929459271585e-06, "loss": 0.007, "step": 46677 }, { "epoch": 14.36, "learning_rate": 3.893198369580915e-06, "loss": 0.0148, "step": 46678 }, { "epoch": 14.36, "learning_rate": 3.8928038083962095e-06, "loss": 0.0242, "step": 46679 }, { "epoch": 14.36, "learning_rate": 3.8924092623740296e-06, "loss": 0.0128, "step": 46680 }, { "epoch": 14.36, "learning_rate": 3.892014731515351e-06, "loss": 0.0125, "step": 46681 }, { "epoch": 14.36, "learning_rate": 3.891620215821154e-06, "loss": 0.0125, "step": 46682 }, { "epoch": 14.36, "learning_rate": 3.891225715292416e-06, "loss": 0.0172, "step": 46683 }, { "epoch": 14.36, "learning_rate": 3.890831229930117e-06, "loss": 0.0117, "step": 46684 }, { "epoch": 14.36, "learning_rate": 3.890436759735239e-06, "loss": 0.02, "step": 46685 }, { "epoch": 14.36, "learning_rate": 3.890042304708758e-06, "loss": 0.0103, "step": 46686 }, { "epoch": 14.37, "learning_rate": 3.889647864851655e-06, "loss": 0.0143, "step": 46687 }, { "epoch": 14.37, "learning_rate": 3.88925344016491e-06, "loss": 0.0102, "step": 46688 }, { "epoch": 14.37, "learning_rate": 3.888859030649498e-06, "loss": 0.0059, "step": 46689 }, { "epoch": 14.37, "learning_rate": 3.888464636306404e-06, "loss": 0.0333, "step": 46690 }, { "epoch": 14.37, "learning_rate": 3.888070257136608e-06, "loss": 0.0098, "step": 46691 }, { "epoch": 14.37, "learning_rate": 3.887675893141083e-06, "loss": 0.013, "step": 46692 }, { "epoch": 14.37, "learning_rate": 3.88728154432081e-06, "loss": 0.017, "step": 46693 }, { "epoch": 14.37, "learning_rate": 3.886887210676767e-06, "loss": 0.0156, "step": 46694 }, { "epoch": 14.37, "learning_rate": 3.886492892209936e-06, "loss": 0.0098, "step": 46695 }, { "epoch": 14.37, "learning_rate": 3.886098588921298e-06, "loss": 0.0106, "step": 46696 }, { "epoch": 14.37, "learning_rate": 3.885704300811827e-06, "loss": 0.0095, "step": 46697 }, { "epoch": 14.37, "learning_rate": 3.885310027882504e-06, "loss": 0.0156, "step": 46698 }, { "epoch": 14.37, "learning_rate": 3.884915770134309e-06, "loss": 0.0109, "step": 46699 }, { "epoch": 14.37, "learning_rate": 3.884521527568218e-06, "loss": 0.0114, "step": 46700 }, { "epoch": 14.37, "learning_rate": 3.884127300185212e-06, "loss": 0.0185, "step": 46701 }, { "epoch": 14.37, "learning_rate": 3.88373308798627e-06, "loss": 0.0063, "step": 46702 }, { "epoch": 14.37, "learning_rate": 3.883338890972365e-06, "loss": 0.0321, "step": 46703 }, { "epoch": 14.37, "learning_rate": 3.8829447091444886e-06, "loss": 0.0648, "step": 46704 }, { "epoch": 14.37, "learning_rate": 3.882550542503607e-06, "loss": 0.0118, "step": 46705 }, { "epoch": 14.37, "learning_rate": 3.882156391050703e-06, "loss": 0.0169, "step": 46706 }, { "epoch": 14.37, "learning_rate": 3.881762254786756e-06, "loss": 0.01, "step": 46707 }, { "epoch": 14.37, "learning_rate": 3.881368133712741e-06, "loss": 0.0074, "step": 46708 }, { "epoch": 14.37, "learning_rate": 3.880974027829641e-06, "loss": 0.0207, "step": 46709 }, { "epoch": 14.37, "learning_rate": 3.880579937138436e-06, "loss": 0.0103, "step": 46710 }, { "epoch": 14.37, "learning_rate": 3.8801858616400985e-06, "loss": 0.011, "step": 46711 }, { "epoch": 14.37, "learning_rate": 3.879791801335614e-06, "loss": 0.0131, "step": 46712 }, { "epoch": 14.37, "learning_rate": 3.879397756225949e-06, "loss": 0.0098, "step": 46713 }, { "epoch": 14.37, "learning_rate": 3.879003726312094e-06, "loss": 0.0151, "step": 46714 }, { "epoch": 14.37, "learning_rate": 3.878609711595022e-06, "loss": 0.0114, "step": 46715 }, { "epoch": 14.37, "learning_rate": 3.878215712075711e-06, "loss": 0.0151, "step": 46716 }, { "epoch": 14.37, "learning_rate": 3.877821727755141e-06, "loss": 0.0171, "step": 46717 }, { "epoch": 14.37, "learning_rate": 3.877427758634289e-06, "loss": 0.0166, "step": 46718 }, { "epoch": 14.37, "learning_rate": 3.877033804714133e-06, "loss": 0.0168, "step": 46719 }, { "epoch": 14.38, "learning_rate": 3.876639865995652e-06, "loss": 0.0162, "step": 46720 }, { "epoch": 14.38, "learning_rate": 3.876245942479822e-06, "loss": 0.0073, "step": 46721 }, { "epoch": 14.38, "learning_rate": 3.8758520341676235e-06, "loss": 0.0112, "step": 46722 }, { "epoch": 14.38, "learning_rate": 3.87545814106003e-06, "loss": 0.0222, "step": 46723 }, { "epoch": 14.38, "learning_rate": 3.875064263158026e-06, "loss": 0.0107, "step": 46724 }, { "epoch": 14.38, "learning_rate": 3.874670400462589e-06, "loss": 0.0138, "step": 46725 }, { "epoch": 14.38, "learning_rate": 3.874276552974692e-06, "loss": 0.0127, "step": 46726 }, { "epoch": 14.38, "learning_rate": 3.8738827206953135e-06, "loss": 0.0092, "step": 46727 }, { "epoch": 14.38, "learning_rate": 3.87348890362543e-06, "loss": 0.0381, "step": 46728 }, { "epoch": 14.38, "learning_rate": 3.873095101766025e-06, "loss": 0.0095, "step": 46729 }, { "epoch": 14.38, "learning_rate": 3.872701315118074e-06, "loss": 0.0098, "step": 46730 }, { "epoch": 14.38, "learning_rate": 3.872307543682553e-06, "loss": 0.0057, "step": 46731 }, { "epoch": 14.38, "learning_rate": 3.87191378746044e-06, "loss": 0.0266, "step": 46732 }, { "epoch": 14.38, "learning_rate": 3.871520046452714e-06, "loss": 0.0123, "step": 46733 }, { "epoch": 14.38, "learning_rate": 3.87112632066035e-06, "loss": 0.0194, "step": 46734 }, { "epoch": 14.38, "learning_rate": 3.870732610084329e-06, "loss": 0.0133, "step": 46735 }, { "epoch": 14.38, "learning_rate": 3.8703389147256255e-06, "loss": 0.0133, "step": 46736 }, { "epoch": 14.38, "learning_rate": 3.869945234585217e-06, "loss": 0.0225, "step": 46737 }, { "epoch": 14.38, "learning_rate": 3.8695515696640825e-06, "loss": 0.0074, "step": 46738 }, { "epoch": 14.38, "learning_rate": 3.8691579199632e-06, "loss": 0.0091, "step": 46739 }, { "epoch": 14.38, "learning_rate": 3.868764285483545e-06, "loss": 0.0124, "step": 46740 }, { "epoch": 14.38, "learning_rate": 3.8683706662260945e-06, "loss": 0.0092, "step": 46741 }, { "epoch": 14.38, "learning_rate": 3.867977062191828e-06, "loss": 0.0138, "step": 46742 }, { "epoch": 14.38, "learning_rate": 3.867583473381716e-06, "loss": 0.0154, "step": 46743 }, { "epoch": 14.38, "learning_rate": 3.867189899796746e-06, "loss": 0.0048, "step": 46744 }, { "epoch": 14.38, "learning_rate": 3.86679634143789e-06, "loss": 0.0186, "step": 46745 }, { "epoch": 14.38, "learning_rate": 3.866402798306128e-06, "loss": 0.009, "step": 46746 }, { "epoch": 14.38, "learning_rate": 3.866009270402431e-06, "loss": 0.0138, "step": 46747 }, { "epoch": 14.38, "learning_rate": 3.865615757727775e-06, "loss": 0.0102, "step": 46748 }, { "epoch": 14.38, "learning_rate": 3.865222260283146e-06, "loss": 0.0116, "step": 46749 }, { "epoch": 14.38, "learning_rate": 3.864828778069516e-06, "loss": 0.0215, "step": 46750 }, { "epoch": 14.38, "learning_rate": 3.8644353110878615e-06, "loss": 0.0042, "step": 46751 }, { "epoch": 14.39, "learning_rate": 3.864041859339161e-06, "loss": 0.009, "step": 46752 }, { "epoch": 14.39, "learning_rate": 3.8636484228243895e-06, "loss": 0.014, "step": 46753 }, { "epoch": 14.39, "learning_rate": 3.863255001544526e-06, "loss": 0.0102, "step": 46754 }, { "epoch": 14.39, "learning_rate": 3.862861595500544e-06, "loss": 0.0135, "step": 46755 }, { "epoch": 14.39, "learning_rate": 3.862468204693424e-06, "loss": 0.021, "step": 46756 }, { "epoch": 14.39, "learning_rate": 3.862074829124139e-06, "loss": 0.0121, "step": 46757 }, { "epoch": 14.39, "learning_rate": 3.8616814687936656e-06, "loss": 0.0295, "step": 46758 }, { "epoch": 14.39, "learning_rate": 3.861288123702989e-06, "loss": 0.0235, "step": 46759 }, { "epoch": 14.39, "learning_rate": 3.860894793853075e-06, "loss": 0.0124, "step": 46760 }, { "epoch": 14.39, "learning_rate": 3.860501479244905e-06, "loss": 0.0131, "step": 46761 }, { "epoch": 14.39, "learning_rate": 3.860108179879454e-06, "loss": 0.0228, "step": 46762 }, { "epoch": 14.39, "learning_rate": 3.859714895757695e-06, "loss": 0.0109, "step": 46763 }, { "epoch": 14.39, "learning_rate": 3.859321626880612e-06, "loss": 0.0129, "step": 46764 }, { "epoch": 14.39, "learning_rate": 3.858928373249179e-06, "loss": 0.0286, "step": 46765 }, { "epoch": 14.39, "learning_rate": 3.858535134864371e-06, "loss": 0.0106, "step": 46766 }, { "epoch": 14.39, "learning_rate": 3.858141911727168e-06, "loss": 0.0197, "step": 46767 }, { "epoch": 14.39, "learning_rate": 3.857748703838536e-06, "loss": 0.0059, "step": 46768 }, { "epoch": 14.39, "learning_rate": 3.857355511199461e-06, "loss": 0.0138, "step": 46769 }, { "epoch": 14.39, "learning_rate": 3.8569623338109164e-06, "loss": 0.0093, "step": 46770 }, { "epoch": 14.39, "learning_rate": 3.856569171673879e-06, "loss": 0.007, "step": 46771 }, { "epoch": 14.39, "learning_rate": 3.856176024789323e-06, "loss": 0.0109, "step": 46772 }, { "epoch": 14.39, "learning_rate": 3.855782893158227e-06, "loss": 0.0052, "step": 46773 }, { "epoch": 14.39, "learning_rate": 3.855389776781564e-06, "loss": 0.0124, "step": 46774 }, { "epoch": 14.39, "learning_rate": 3.854996675660313e-06, "loss": 0.0061, "step": 46775 }, { "epoch": 14.39, "learning_rate": 3.854603589795448e-06, "loss": 0.0096, "step": 46776 }, { "epoch": 14.39, "learning_rate": 3.854210519187946e-06, "loss": 0.0174, "step": 46777 }, { "epoch": 14.39, "learning_rate": 3.853817463838778e-06, "loss": 0.0205, "step": 46778 }, { "epoch": 14.39, "learning_rate": 3.853424423748928e-06, "loss": 0.0172, "step": 46779 }, { "epoch": 14.39, "learning_rate": 3.853031398919372e-06, "loss": 0.009, "step": 46780 }, { "epoch": 14.39, "learning_rate": 3.852638389351079e-06, "loss": 0.0091, "step": 46781 }, { "epoch": 14.39, "learning_rate": 3.852245395045026e-06, "loss": 0.0084, "step": 46782 }, { "epoch": 14.39, "learning_rate": 3.8518524160021876e-06, "loss": 0.0151, "step": 46783 }, { "epoch": 14.39, "learning_rate": 3.851459452223546e-06, "loss": 0.0117, "step": 46784 }, { "epoch": 14.4, "learning_rate": 3.851066503710072e-06, "loss": 0.012, "step": 46785 }, { "epoch": 14.4, "learning_rate": 3.850673570462744e-06, "loss": 0.0086, "step": 46786 }, { "epoch": 14.4, "learning_rate": 3.850280652482534e-06, "loss": 0.015, "step": 46787 }, { "epoch": 14.4, "learning_rate": 3.84988774977042e-06, "loss": 0.0098, "step": 46788 }, { "epoch": 14.4, "learning_rate": 3.849494862327376e-06, "loss": 0.012, "step": 46789 }, { "epoch": 14.4, "learning_rate": 3.849101990154379e-06, "loss": 0.01, "step": 46790 }, { "epoch": 14.4, "learning_rate": 3.848709133252403e-06, "loss": 0.0213, "step": 46791 }, { "epoch": 14.4, "learning_rate": 3.848316291622425e-06, "loss": 0.0194, "step": 46792 }, { "epoch": 14.4, "learning_rate": 3.847923465265418e-06, "loss": 0.018, "step": 46793 }, { "epoch": 14.4, "learning_rate": 3.847530654182359e-06, "loss": 0.0128, "step": 46794 }, { "epoch": 14.4, "learning_rate": 3.847137858374222e-06, "loss": 0.0106, "step": 46795 }, { "epoch": 14.4, "learning_rate": 3.846745077841984e-06, "loss": 0.0159, "step": 46796 }, { "epoch": 14.4, "learning_rate": 3.846352312586618e-06, "loss": 0.0106, "step": 46797 }, { "epoch": 14.4, "learning_rate": 3.845959562609097e-06, "loss": 0.0084, "step": 46798 }, { "epoch": 14.4, "learning_rate": 3.845566827910403e-06, "loss": 0.0195, "step": 46799 }, { "epoch": 14.4, "learning_rate": 3.845174108491508e-06, "loss": 0.0096, "step": 46800 }, { "epoch": 14.4, "learning_rate": 3.84478140435339e-06, "loss": 0.0313, "step": 46801 }, { "epoch": 14.4, "learning_rate": 3.844388715497015e-06, "loss": 0.011, "step": 46802 }, { "epoch": 14.4, "learning_rate": 3.843996041923361e-06, "loss": 0.0106, "step": 46803 }, { "epoch": 14.4, "learning_rate": 3.8436033836334096e-06, "loss": 0.0125, "step": 46804 }, { "epoch": 14.4, "learning_rate": 3.84321074062813e-06, "loss": 0.0156, "step": 46805 }, { "epoch": 14.4, "learning_rate": 3.842818112908498e-06, "loss": 0.0173, "step": 46806 }, { "epoch": 14.4, "learning_rate": 3.842425500475491e-06, "loss": 0.0123, "step": 46807 }, { "epoch": 14.4, "learning_rate": 3.84203290333008e-06, "loss": 0.0122, "step": 46808 }, { "epoch": 14.4, "learning_rate": 3.84164032147324e-06, "loss": 0.0101, "step": 46809 }, { "epoch": 14.4, "learning_rate": 3.841247754905949e-06, "loss": 0.0149, "step": 46810 }, { "epoch": 14.4, "learning_rate": 3.840855203629178e-06, "loss": 0.0068, "step": 46811 }, { "epoch": 14.4, "learning_rate": 3.8404626676439035e-06, "loss": 0.0178, "step": 46812 }, { "epoch": 14.4, "learning_rate": 3.840070146951096e-06, "loss": 0.0117, "step": 46813 }, { "epoch": 14.4, "learning_rate": 3.8396776415517415e-06, "loss": 0.0059, "step": 46814 }, { "epoch": 14.4, "learning_rate": 3.839285151446802e-06, "loss": 0.0123, "step": 46815 }, { "epoch": 14.4, "learning_rate": 3.838892676637257e-06, "loss": 0.0081, "step": 46816 }, { "epoch": 14.41, "learning_rate": 3.83850021712408e-06, "loss": 0.015, "step": 46817 }, { "epoch": 14.41, "learning_rate": 3.8381077729082425e-06, "loss": 0.0262, "step": 46818 }, { "epoch": 14.41, "learning_rate": 3.837715343990727e-06, "loss": 0.0117, "step": 46819 }, { "epoch": 14.41, "learning_rate": 3.837322930372503e-06, "loss": 0.0102, "step": 46820 }, { "epoch": 14.41, "learning_rate": 3.836930532054543e-06, "loss": 0.0119, "step": 46821 }, { "epoch": 14.41, "learning_rate": 3.8365381490378275e-06, "loss": 0.0102, "step": 46822 }, { "epoch": 14.41, "learning_rate": 3.836145781323318e-06, "loss": 0.0065, "step": 46823 }, { "epoch": 14.41, "learning_rate": 3.835753428912003e-06, "loss": 0.0131, "step": 46824 }, { "epoch": 14.41, "learning_rate": 3.8353610918048475e-06, "loss": 0.0177, "step": 46825 }, { "epoch": 14.41, "learning_rate": 3.8349687700028305e-06, "loss": 0.0142, "step": 46826 }, { "epoch": 14.41, "learning_rate": 3.8345764635069225e-06, "loss": 0.0183, "step": 46827 }, { "epoch": 14.41, "learning_rate": 3.834184172318099e-06, "loss": 0.0087, "step": 46828 }, { "epoch": 14.41, "learning_rate": 3.833791896437336e-06, "loss": 0.0201, "step": 46829 }, { "epoch": 14.41, "learning_rate": 3.833399635865603e-06, "loss": 0.0133, "step": 46830 }, { "epoch": 14.41, "learning_rate": 3.833007390603878e-06, "loss": 0.0111, "step": 46831 }, { "epoch": 14.41, "learning_rate": 3.832615160653131e-06, "loss": 0.01, "step": 46832 }, { "epoch": 14.41, "learning_rate": 3.832222946014336e-06, "loss": 0.0142, "step": 46833 }, { "epoch": 14.41, "learning_rate": 3.831830746688472e-06, "loss": 0.0077, "step": 46834 }, { "epoch": 14.41, "learning_rate": 3.831438562676513e-06, "loss": 0.0231, "step": 46835 }, { "epoch": 14.41, "learning_rate": 3.831046393979425e-06, "loss": 0.0185, "step": 46836 }, { "epoch": 14.41, "learning_rate": 3.830654240598186e-06, "loss": 0.0176, "step": 46837 }, { "epoch": 14.41, "learning_rate": 3.830262102533766e-06, "loss": 0.014, "step": 46838 }, { "epoch": 14.41, "learning_rate": 3.829869979787145e-06, "loss": 0.0304, "step": 46839 }, { "epoch": 14.41, "learning_rate": 3.829477872359294e-06, "loss": 0.0159, "step": 46840 }, { "epoch": 14.41, "learning_rate": 3.829085780251185e-06, "loss": 0.0304, "step": 46841 }, { "epoch": 14.41, "learning_rate": 3.828693703463793e-06, "loss": 0.0132, "step": 46842 }, { "epoch": 14.41, "learning_rate": 3.828301641998091e-06, "loss": 0.0105, "step": 46843 }, { "epoch": 14.41, "learning_rate": 3.827909595855051e-06, "loss": 0.0138, "step": 46844 }, { "epoch": 14.41, "learning_rate": 3.8275175650356485e-06, "loss": 0.0085, "step": 46845 }, { "epoch": 14.41, "learning_rate": 3.827125549540856e-06, "loss": 0.0157, "step": 46846 }, { "epoch": 14.41, "learning_rate": 3.826733549371647e-06, "loss": 0.0145, "step": 46847 }, { "epoch": 14.41, "learning_rate": 3.82634156452899e-06, "loss": 0.0096, "step": 46848 }, { "epoch": 14.41, "learning_rate": 3.825949595013871e-06, "loss": 0.0186, "step": 46849 }, { "epoch": 14.42, "learning_rate": 3.825557640827249e-06, "loss": 0.0172, "step": 46850 }, { "epoch": 14.42, "learning_rate": 3.825165701970104e-06, "loss": 0.0143, "step": 46851 }, { "epoch": 14.42, "learning_rate": 3.824773778443407e-06, "loss": 0.0127, "step": 46852 }, { "epoch": 14.42, "learning_rate": 3.82438187024813e-06, "loss": 0.017, "step": 46853 }, { "epoch": 14.42, "learning_rate": 3.82398997738525e-06, "loss": 0.0158, "step": 46854 }, { "epoch": 14.42, "learning_rate": 3.823598099855738e-06, "loss": 0.016, "step": 46855 }, { "epoch": 14.42, "learning_rate": 3.82320623766057e-06, "loss": 0.0144, "step": 46856 }, { "epoch": 14.42, "learning_rate": 3.822814390800713e-06, "loss": 0.0161, "step": 46857 }, { "epoch": 14.42, "learning_rate": 3.822422559277139e-06, "loss": 0.0127, "step": 46858 }, { "epoch": 14.42, "learning_rate": 3.822030743090828e-06, "loss": 0.0238, "step": 46859 }, { "epoch": 14.42, "learning_rate": 3.821638942242748e-06, "loss": 0.0128, "step": 46860 }, { "epoch": 14.42, "learning_rate": 3.821247156733874e-06, "loss": 0.0134, "step": 46861 }, { "epoch": 14.42, "learning_rate": 3.8208553865651775e-06, "loss": 0.0098, "step": 46862 }, { "epoch": 14.42, "learning_rate": 3.820463631737631e-06, "loss": 0.0119, "step": 46863 }, { "epoch": 14.42, "learning_rate": 3.820071892252208e-06, "loss": 0.0292, "step": 46864 }, { "epoch": 14.42, "learning_rate": 3.81968016810988e-06, "loss": 0.0092, "step": 46865 }, { "epoch": 14.42, "learning_rate": 3.819288459311619e-06, "loss": 0.0085, "step": 46866 }, { "epoch": 14.42, "learning_rate": 3.8188967658584e-06, "loss": 0.0485, "step": 46867 }, { "epoch": 14.42, "learning_rate": 3.818505087751189e-06, "loss": 0.0118, "step": 46868 }, { "epoch": 14.42, "learning_rate": 3.818113424990968e-06, "loss": 0.0161, "step": 46869 }, { "epoch": 14.42, "learning_rate": 3.817721777578708e-06, "loss": 0.0185, "step": 46870 }, { "epoch": 14.42, "learning_rate": 3.817330145515374e-06, "loss": 0.0078, "step": 46871 }, { "epoch": 14.42, "learning_rate": 3.8169385288019435e-06, "loss": 0.0265, "step": 46872 }, { "epoch": 14.42, "learning_rate": 3.8165469274393825e-06, "loss": 0.0168, "step": 46873 }, { "epoch": 14.42, "learning_rate": 3.8161553414286746e-06, "loss": 0.0091, "step": 46874 }, { "epoch": 14.42, "learning_rate": 3.815763770770784e-06, "loss": 0.0131, "step": 46875 }, { "epoch": 14.42, "learning_rate": 3.815372215466685e-06, "loss": 0.0134, "step": 46876 }, { "epoch": 14.42, "learning_rate": 3.8149806755173535e-06, "loss": 0.0162, "step": 46877 }, { "epoch": 14.42, "learning_rate": 3.8145891509237497e-06, "loss": 0.0229, "step": 46878 }, { "epoch": 14.42, "learning_rate": 3.8141976416868574e-06, "loss": 0.0202, "step": 46879 }, { "epoch": 14.42, "learning_rate": 3.813806147807645e-06, "loss": 0.0104, "step": 46880 }, { "epoch": 14.42, "learning_rate": 3.8134146692870844e-06, "loss": 0.014, "step": 46881 }, { "epoch": 14.43, "learning_rate": 3.813023206126147e-06, "loss": 0.0085, "step": 46882 }, { "epoch": 14.43, "learning_rate": 3.8126317583258057e-06, "loss": 0.0193, "step": 46883 }, { "epoch": 14.43, "learning_rate": 3.81224032588703e-06, "loss": 0.0027, "step": 46884 }, { "epoch": 14.43, "learning_rate": 3.811848908810795e-06, "loss": 0.0247, "step": 46885 }, { "epoch": 14.43, "learning_rate": 3.81145750709807e-06, "loss": 0.0074, "step": 46886 }, { "epoch": 14.43, "learning_rate": 3.811066120749828e-06, "loss": 0.0095, "step": 46887 }, { "epoch": 14.43, "learning_rate": 3.810674749767036e-06, "loss": 0.0188, "step": 46888 }, { "epoch": 14.43, "learning_rate": 3.8102833941506743e-06, "loss": 0.0047, "step": 46889 }, { "epoch": 14.43, "learning_rate": 3.80989205390171e-06, "loss": 0.0159, "step": 46890 }, { "epoch": 14.43, "learning_rate": 3.8095007290211196e-06, "loss": 0.014, "step": 46891 }, { "epoch": 14.43, "learning_rate": 3.8091094195098656e-06, "loss": 0.0224, "step": 46892 }, { "epoch": 14.43, "learning_rate": 3.8087181253689198e-06, "loss": 0.0121, "step": 46893 }, { "epoch": 14.43, "learning_rate": 3.8083268465992616e-06, "loss": 0.0125, "step": 46894 }, { "epoch": 14.43, "learning_rate": 3.807935583201858e-06, "loss": 0.0093, "step": 46895 }, { "epoch": 14.43, "learning_rate": 3.807544335177682e-06, "loss": 0.0125, "step": 46896 }, { "epoch": 14.43, "learning_rate": 3.807153102527704e-06, "loss": 0.028, "step": 46897 }, { "epoch": 14.43, "learning_rate": 3.8067618852528955e-06, "loss": 0.0148, "step": 46898 }, { "epoch": 14.43, "learning_rate": 3.806370683354227e-06, "loss": 0.0092, "step": 46899 }, { "epoch": 14.43, "learning_rate": 3.80597949683267e-06, "loss": 0.0222, "step": 46900 }, { "epoch": 14.43, "learning_rate": 3.8055883256891968e-06, "loss": 0.0135, "step": 46901 }, { "epoch": 14.43, "learning_rate": 3.805197169924778e-06, "loss": 0.012, "step": 46902 }, { "epoch": 14.43, "learning_rate": 3.80480602954038e-06, "loss": 0.0068, "step": 46903 }, { "epoch": 14.43, "learning_rate": 3.804414904536986e-06, "loss": 0.0286, "step": 46904 }, { "epoch": 14.43, "learning_rate": 3.8040237949155567e-06, "loss": 0.0125, "step": 46905 }, { "epoch": 14.43, "learning_rate": 3.8036327006770655e-06, "loss": 0.0125, "step": 46906 }, { "epoch": 14.43, "learning_rate": 3.8032416218224834e-06, "loss": 0.0034, "step": 46907 }, { "epoch": 14.43, "learning_rate": 3.8028505583527787e-06, "loss": 0.0125, "step": 46908 }, { "epoch": 14.43, "learning_rate": 3.8024595102689286e-06, "loss": 0.0086, "step": 46909 }, { "epoch": 14.43, "learning_rate": 3.802068477571901e-06, "loss": 0.0086, "step": 46910 }, { "epoch": 14.43, "learning_rate": 3.801677460262666e-06, "loss": 0.0146, "step": 46911 }, { "epoch": 14.43, "learning_rate": 3.801286458342198e-06, "loss": 0.0071, "step": 46912 }, { "epoch": 14.43, "learning_rate": 3.800895471811459e-06, "loss": 0.0176, "step": 46913 }, { "epoch": 14.43, "learning_rate": 3.800504500671428e-06, "loss": 0.0153, "step": 46914 }, { "epoch": 14.44, "learning_rate": 3.800113544923073e-06, "loss": 0.0138, "step": 46915 }, { "epoch": 14.44, "learning_rate": 3.799722604567365e-06, "loss": 0.0196, "step": 46916 }, { "epoch": 14.44, "learning_rate": 3.7993316796052737e-06, "loss": 0.0136, "step": 46917 }, { "epoch": 14.44, "learning_rate": 3.79894077003777e-06, "loss": 0.0083, "step": 46918 }, { "epoch": 14.44, "learning_rate": 3.7985498758658255e-06, "loss": 0.0155, "step": 46919 }, { "epoch": 14.44, "learning_rate": 3.7981589970904086e-06, "loss": 0.0214, "step": 46920 }, { "epoch": 14.44, "learning_rate": 3.7977681337124917e-06, "loss": 0.0285, "step": 46921 }, { "epoch": 14.44, "learning_rate": 3.7973772857330447e-06, "loss": 0.0144, "step": 46922 }, { "epoch": 14.44, "learning_rate": 3.7969864531530344e-06, "loss": 0.0065, "step": 46923 }, { "epoch": 14.44, "learning_rate": 3.7965956359734367e-06, "loss": 0.0093, "step": 46924 }, { "epoch": 14.44, "learning_rate": 3.796204834195224e-06, "loss": 0.0162, "step": 46925 }, { "epoch": 14.44, "learning_rate": 3.7958140478193595e-06, "loss": 0.0113, "step": 46926 }, { "epoch": 14.44, "learning_rate": 3.7954232768468147e-06, "loss": 0.01, "step": 46927 }, { "epoch": 14.44, "learning_rate": 3.7950325212785576e-06, "loss": 0.0141, "step": 46928 }, { "epoch": 14.44, "learning_rate": 3.7946417811155665e-06, "loss": 0.0137, "step": 46929 }, { "epoch": 14.44, "learning_rate": 3.7942510563588063e-06, "loss": 0.0188, "step": 46930 }, { "epoch": 14.44, "learning_rate": 3.7938603470092484e-06, "loss": 0.0113, "step": 46931 }, { "epoch": 14.44, "learning_rate": 3.7934696530678614e-06, "loss": 0.0106, "step": 46932 }, { "epoch": 14.44, "learning_rate": 3.7930789745356165e-06, "loss": 0.0112, "step": 46933 }, { "epoch": 14.44, "learning_rate": 3.7926883114134827e-06, "loss": 0.0096, "step": 46934 }, { "epoch": 14.44, "learning_rate": 3.792297663702431e-06, "loss": 0.0173, "step": 46935 }, { "epoch": 14.44, "learning_rate": 3.79190703140343e-06, "loss": 0.0146, "step": 46936 }, { "epoch": 14.44, "learning_rate": 3.791516414517451e-06, "loss": 0.0132, "step": 46937 }, { "epoch": 14.44, "learning_rate": 3.791125813045462e-06, "loss": 0.0156, "step": 46938 }, { "epoch": 14.44, "learning_rate": 3.7907352269884346e-06, "loss": 0.014, "step": 46939 }, { "epoch": 14.44, "learning_rate": 3.790344656347338e-06, "loss": 0.0159, "step": 46940 }, { "epoch": 14.44, "learning_rate": 3.78995410112314e-06, "loss": 0.0105, "step": 46941 }, { "epoch": 14.44, "learning_rate": 3.7895635613168123e-06, "loss": 0.0122, "step": 46942 }, { "epoch": 14.44, "learning_rate": 3.789173036929321e-06, "loss": 0.013, "step": 46943 }, { "epoch": 14.44, "learning_rate": 3.788782527961642e-06, "loss": 0.0106, "step": 46944 }, { "epoch": 14.44, "learning_rate": 3.7883920344147418e-06, "loss": 0.0089, "step": 46945 }, { "epoch": 14.44, "learning_rate": 3.7880015562895935e-06, "loss": 0.0091, "step": 46946 }, { "epoch": 14.45, "learning_rate": 3.787611093587158e-06, "loss": 0.0306, "step": 46947 }, { "epoch": 14.45, "learning_rate": 3.787220646308406e-06, "loss": 0.0241, "step": 46948 }, { "epoch": 14.45, "learning_rate": 3.7868302144543146e-06, "loss": 0.0134, "step": 46949 }, { "epoch": 14.45, "learning_rate": 3.7864397980258472e-06, "loss": 0.0068, "step": 46950 }, { "epoch": 14.45, "learning_rate": 3.7860493970239763e-06, "loss": 0.0121, "step": 46951 }, { "epoch": 14.45, "learning_rate": 3.7856590114496683e-06, "loss": 0.0103, "step": 46952 }, { "epoch": 14.45, "learning_rate": 3.7852686413038943e-06, "loss": 0.0107, "step": 46953 }, { "epoch": 14.45, "learning_rate": 3.784878286587622e-06, "loss": 0.0169, "step": 46954 }, { "epoch": 14.45, "learning_rate": 3.7844879473018228e-06, "loss": 0.0173, "step": 46955 }, { "epoch": 14.45, "learning_rate": 3.7840976234474636e-06, "loss": 0.0161, "step": 46956 }, { "epoch": 14.45, "learning_rate": 3.783707315025514e-06, "loss": 0.0107, "step": 46957 }, { "epoch": 14.45, "learning_rate": 3.78331702203694e-06, "loss": 0.0099, "step": 46958 }, { "epoch": 14.45, "learning_rate": 3.782926744482721e-06, "loss": 0.0157, "step": 46959 }, { "epoch": 14.45, "learning_rate": 3.7825364823638157e-06, "loss": 0.0081, "step": 46960 }, { "epoch": 14.45, "learning_rate": 3.7821462356811956e-06, "loss": 0.0135, "step": 46961 }, { "epoch": 14.45, "learning_rate": 3.7817560044358302e-06, "loss": 0.0122, "step": 46962 }, { "epoch": 14.45, "learning_rate": 3.7813657886286847e-06, "loss": 0.0068, "step": 46963 }, { "epoch": 14.45, "learning_rate": 3.780975588260735e-06, "loss": 0.0141, "step": 46964 }, { "epoch": 14.45, "learning_rate": 3.7805854033329457e-06, "loss": 0.0179, "step": 46965 }, { "epoch": 14.45, "learning_rate": 3.7801952338462865e-06, "loss": 0.0119, "step": 46966 }, { "epoch": 14.45, "learning_rate": 3.7798050798017303e-06, "loss": 0.0092, "step": 46967 }, { "epoch": 14.45, "learning_rate": 3.7794149412002336e-06, "loss": 0.0079, "step": 46968 }, { "epoch": 14.45, "learning_rate": 3.7790248180427758e-06, "loss": 0.0045, "step": 46969 }, { "epoch": 14.45, "learning_rate": 3.778634710330322e-06, "loss": 0.0173, "step": 46970 }, { "epoch": 14.45, "learning_rate": 3.778244618063841e-06, "loss": 0.0097, "step": 46971 }, { "epoch": 14.45, "learning_rate": 3.7778545412443013e-06, "loss": 0.0123, "step": 46972 }, { "epoch": 14.45, "learning_rate": 3.7774644798726725e-06, "loss": 0.0144, "step": 46973 }, { "epoch": 14.45, "learning_rate": 3.77707443394992e-06, "loss": 0.0051, "step": 46974 }, { "epoch": 14.45, "learning_rate": 3.7766844034770155e-06, "loss": 0.0108, "step": 46975 }, { "epoch": 14.45, "learning_rate": 3.7762943884549254e-06, "loss": 0.0089, "step": 46976 }, { "epoch": 14.45, "learning_rate": 3.775904388884618e-06, "loss": 0.0177, "step": 46977 }, { "epoch": 14.45, "learning_rate": 3.775514404767059e-06, "loss": 0.0147, "step": 46978 }, { "epoch": 14.45, "learning_rate": 3.7751244361032234e-06, "loss": 0.0173, "step": 46979 }, { "epoch": 14.46, "learning_rate": 3.7747344828940793e-06, "loss": 0.0131, "step": 46980 }, { "epoch": 14.46, "learning_rate": 3.774344545140587e-06, "loss": 0.0111, "step": 46981 }, { "epoch": 14.46, "learning_rate": 3.7739546228437186e-06, "loss": 0.0117, "step": 46982 }, { "epoch": 14.46, "learning_rate": 3.77356471600444e-06, "loss": 0.0203, "step": 46983 }, { "epoch": 14.46, "learning_rate": 3.7731748246237244e-06, "loss": 0.0467, "step": 46984 }, { "epoch": 14.46, "learning_rate": 3.772784948702537e-06, "loss": 0.0115, "step": 46985 }, { "epoch": 14.46, "learning_rate": 3.7723950882418457e-06, "loss": 0.0126, "step": 46986 }, { "epoch": 14.46, "learning_rate": 3.772005243242619e-06, "loss": 0.0105, "step": 46987 }, { "epoch": 14.46, "learning_rate": 3.7716154137058236e-06, "loss": 0.0277, "step": 46988 }, { "epoch": 14.46, "learning_rate": 3.771225599632429e-06, "loss": 0.0247, "step": 46989 }, { "epoch": 14.46, "learning_rate": 3.7708358010234003e-06, "loss": 0.0143, "step": 46990 }, { "epoch": 14.46, "learning_rate": 3.770446017879709e-06, "loss": 0.0091, "step": 46991 }, { "epoch": 14.46, "learning_rate": 3.7700562502023185e-06, "loss": 0.0046, "step": 46992 }, { "epoch": 14.46, "learning_rate": 3.7696664979922004e-06, "loss": 0.0175, "step": 46993 }, { "epoch": 14.46, "learning_rate": 3.769276761250321e-06, "loss": 0.0107, "step": 46994 }, { "epoch": 14.46, "learning_rate": 3.768887039977647e-06, "loss": 0.0257, "step": 46995 }, { "epoch": 14.46, "learning_rate": 3.7684973341751464e-06, "loss": 0.0097, "step": 46996 }, { "epoch": 14.46, "learning_rate": 3.768107643843787e-06, "loss": 0.011, "step": 46997 }, { "epoch": 14.46, "learning_rate": 3.7677179689845334e-06, "loss": 0.0162, "step": 46998 }, { "epoch": 14.46, "learning_rate": 3.7673283095983593e-06, "loss": 0.014, "step": 46999 }, { "epoch": 14.46, "learning_rate": 3.7669386656862283e-06, "loss": 0.0104, "step": 47000 }, { "epoch": 14.46, "learning_rate": 3.766549037249112e-06, "loss": 0.0083, "step": 47001 }, { "epoch": 14.46, "learning_rate": 3.7661594242879694e-06, "loss": 0.0275, "step": 47002 }, { "epoch": 14.46, "learning_rate": 3.76576982680377e-06, "loss": 0.0147, "step": 47003 }, { "epoch": 14.46, "learning_rate": 3.7653802447974873e-06, "loss": 0.0126, "step": 47004 }, { "epoch": 14.46, "learning_rate": 3.764990678270084e-06, "loss": 0.007, "step": 47005 }, { "epoch": 14.46, "learning_rate": 3.764601127222528e-06, "loss": 0.0163, "step": 47006 }, { "epoch": 14.46, "learning_rate": 3.7642115916557866e-06, "loss": 0.0097, "step": 47007 }, { "epoch": 14.46, "learning_rate": 3.763822071570826e-06, "loss": 0.0156, "step": 47008 }, { "epoch": 14.46, "learning_rate": 3.7634325669686144e-06, "loss": 0.0118, "step": 47009 }, { "epoch": 14.46, "learning_rate": 3.7630430778501192e-06, "loss": 0.012, "step": 47010 }, { "epoch": 14.46, "learning_rate": 3.7626536042163054e-06, "loss": 0.0245, "step": 47011 }, { "epoch": 14.47, "learning_rate": 3.762264146068142e-06, "loss": 0.0069, "step": 47012 }, { "epoch": 14.47, "learning_rate": 3.7618747034065904e-06, "loss": 0.0117, "step": 47013 }, { "epoch": 14.47, "learning_rate": 3.7614852762326303e-06, "loss": 0.014, "step": 47014 }, { "epoch": 14.47, "learning_rate": 3.761095864547216e-06, "loss": 0.0121, "step": 47015 }, { "epoch": 14.47, "learning_rate": 3.7607064683513205e-06, "loss": 0.0114, "step": 47016 }, { "epoch": 14.47, "learning_rate": 3.760317087645907e-06, "loss": 0.0135, "step": 47017 }, { "epoch": 14.47, "learning_rate": 3.759927722431941e-06, "loss": 0.0103, "step": 47018 }, { "epoch": 14.47, "learning_rate": 3.7595383727103964e-06, "loss": 0.0184, "step": 47019 }, { "epoch": 14.47, "learning_rate": 3.759149038482235e-06, "loss": 0.0073, "step": 47020 }, { "epoch": 14.47, "learning_rate": 3.758759719748424e-06, "loss": 0.0129, "step": 47021 }, { "epoch": 14.47, "learning_rate": 3.7583704165099335e-06, "loss": 0.0123, "step": 47022 }, { "epoch": 14.47, "learning_rate": 3.7579811287677193e-06, "loss": 0.0114, "step": 47023 }, { "epoch": 14.47, "learning_rate": 3.75759185652276e-06, "loss": 0.0134, "step": 47024 }, { "epoch": 14.47, "learning_rate": 3.757202599776015e-06, "loss": 0.0278, "step": 47025 }, { "epoch": 14.47, "learning_rate": 3.756813358528455e-06, "loss": 0.016, "step": 47026 }, { "epoch": 14.47, "learning_rate": 3.7564241327810436e-06, "loss": 0.0153, "step": 47027 }, { "epoch": 14.47, "learning_rate": 3.756034922534747e-06, "loss": 0.0071, "step": 47028 }, { "epoch": 14.47, "learning_rate": 3.7556457277905334e-06, "loss": 0.0169, "step": 47029 }, { "epoch": 14.47, "learning_rate": 3.7552565485493676e-06, "loss": 0.0133, "step": 47030 }, { "epoch": 14.47, "learning_rate": 3.754867384812216e-06, "loss": 0.0222, "step": 47031 }, { "epoch": 14.47, "learning_rate": 3.7544782365800457e-06, "loss": 0.0144, "step": 47032 }, { "epoch": 14.47, "learning_rate": 3.7540891038538195e-06, "loss": 0.0173, "step": 47033 }, { "epoch": 14.47, "learning_rate": 3.753699986634508e-06, "loss": 0.0229, "step": 47034 }, { "epoch": 14.47, "learning_rate": 3.7533108849230802e-06, "loss": 0.016, "step": 47035 }, { "epoch": 14.47, "learning_rate": 3.7529217987204936e-06, "loss": 0.0101, "step": 47036 }, { "epoch": 14.47, "learning_rate": 3.752532728027717e-06, "loss": 0.0237, "step": 47037 }, { "epoch": 14.47, "learning_rate": 3.752143672845715e-06, "loss": 0.0094, "step": 47038 }, { "epoch": 14.47, "learning_rate": 3.7517546331754594e-06, "loss": 0.0199, "step": 47039 }, { "epoch": 14.47, "learning_rate": 3.7513656090179116e-06, "loss": 0.0033, "step": 47040 }, { "epoch": 14.47, "learning_rate": 3.7509766003740387e-06, "loss": 0.0056, "step": 47041 }, { "epoch": 14.47, "learning_rate": 3.750587607244808e-06, "loss": 0.0136, "step": 47042 }, { "epoch": 14.47, "learning_rate": 3.750198629631181e-06, "loss": 0.0096, "step": 47043 }, { "epoch": 14.47, "learning_rate": 3.7498096675341276e-06, "loss": 0.0157, "step": 47044 }, { "epoch": 14.48, "learning_rate": 3.7494207209546106e-06, "loss": 0.0091, "step": 47045 }, { "epoch": 14.48, "learning_rate": 3.7490317898935978e-06, "loss": 0.0178, "step": 47046 }, { "epoch": 14.48, "learning_rate": 3.748642874352053e-06, "loss": 0.014, "step": 47047 }, { "epoch": 14.48, "learning_rate": 3.74825397433094e-06, "loss": 0.0129, "step": 47048 }, { "epoch": 14.48, "learning_rate": 3.747865089831233e-06, "loss": 0.0093, "step": 47049 }, { "epoch": 14.48, "learning_rate": 3.7474762208538882e-06, "loss": 0.0091, "step": 47050 }, { "epoch": 14.48, "learning_rate": 3.7470873673998755e-06, "loss": 0.013, "step": 47051 }, { "epoch": 14.48, "learning_rate": 3.7466985294701585e-06, "loss": 0.0205, "step": 47052 }, { "epoch": 14.48, "learning_rate": 3.7463097070656995e-06, "loss": 0.0081, "step": 47053 }, { "epoch": 14.48, "learning_rate": 3.745920900187471e-06, "loss": 0.0179, "step": 47054 }, { "epoch": 14.48, "learning_rate": 3.745532108836435e-06, "loss": 0.0117, "step": 47055 }, { "epoch": 14.48, "learning_rate": 3.7451433330135588e-06, "loss": 0.0129, "step": 47056 }, { "epoch": 14.48, "learning_rate": 3.7447545727198044e-06, "loss": 0.0145, "step": 47057 }, { "epoch": 14.48, "learning_rate": 3.7443658279561324e-06, "loss": 0.0176, "step": 47058 }, { "epoch": 14.48, "learning_rate": 3.743977098723518e-06, "loss": 0.0179, "step": 47059 }, { "epoch": 14.48, "learning_rate": 3.7435883850229226e-06, "loss": 0.0122, "step": 47060 }, { "epoch": 14.48, "learning_rate": 3.743199686855309e-06, "loss": 0.019, "step": 47061 }, { "epoch": 14.48, "learning_rate": 3.742811004221645e-06, "loss": 0.0178, "step": 47062 }, { "epoch": 14.48, "learning_rate": 3.742422337122894e-06, "loss": 0.0118, "step": 47063 }, { "epoch": 14.48, "learning_rate": 3.7420336855600214e-06, "loss": 0.0118, "step": 47064 }, { "epoch": 14.48, "learning_rate": 3.7416450495339917e-06, "loss": 0.014, "step": 47065 }, { "epoch": 14.48, "learning_rate": 3.741256429045771e-06, "loss": 0.0204, "step": 47066 }, { "epoch": 14.48, "learning_rate": 3.7408678240963227e-06, "loss": 0.0168, "step": 47067 }, { "epoch": 14.48, "learning_rate": 3.740479234686608e-06, "loss": 0.0169, "step": 47068 }, { "epoch": 14.48, "learning_rate": 3.7400906608176002e-06, "loss": 0.0099, "step": 47069 }, { "epoch": 14.48, "learning_rate": 3.7397021024902637e-06, "loss": 0.0119, "step": 47070 }, { "epoch": 14.48, "learning_rate": 3.7393135597055552e-06, "loss": 0.0132, "step": 47071 }, { "epoch": 14.48, "learning_rate": 3.738925032464442e-06, "loss": 0.013, "step": 47072 }, { "epoch": 14.48, "learning_rate": 3.738536520767888e-06, "loss": 0.0131, "step": 47073 }, { "epoch": 14.48, "learning_rate": 3.738148024616863e-06, "loss": 0.0072, "step": 47074 }, { "epoch": 14.48, "learning_rate": 3.7377595440123284e-06, "loss": 0.0106, "step": 47075 }, { "epoch": 14.48, "learning_rate": 3.7373710789552485e-06, "loss": 0.009, "step": 47076 }, { "epoch": 14.49, "learning_rate": 3.7369826294465904e-06, "loss": 0.0059, "step": 47077 }, { "epoch": 14.49, "learning_rate": 3.736594195487311e-06, "loss": 0.0088, "step": 47078 }, { "epoch": 14.49, "learning_rate": 3.736205777078381e-06, "loss": 0.016, "step": 47079 }, { "epoch": 14.49, "learning_rate": 3.7358173742207647e-06, "loss": 0.0055, "step": 47080 }, { "epoch": 14.49, "learning_rate": 3.7354289869154247e-06, "loss": 0.0122, "step": 47081 }, { "epoch": 14.49, "learning_rate": 3.7350406151633256e-06, "loss": 0.0108, "step": 47082 }, { "epoch": 14.49, "learning_rate": 3.7346522589654323e-06, "loss": 0.0177, "step": 47083 }, { "epoch": 14.49, "learning_rate": 3.734263918322708e-06, "loss": 0.0119, "step": 47084 }, { "epoch": 14.49, "learning_rate": 3.733875593236117e-06, "loss": 0.0135, "step": 47085 }, { "epoch": 14.49, "learning_rate": 3.7334872837066237e-06, "loss": 0.0339, "step": 47086 }, { "epoch": 14.49, "learning_rate": 3.7330989897351923e-06, "loss": 0.0151, "step": 47087 }, { "epoch": 14.49, "learning_rate": 3.732710711322782e-06, "loss": 0.012, "step": 47088 }, { "epoch": 14.49, "learning_rate": 3.732322448470366e-06, "loss": 0.0219, "step": 47089 }, { "epoch": 14.49, "learning_rate": 3.7319342011789074e-06, "loss": 0.0167, "step": 47090 }, { "epoch": 14.49, "learning_rate": 3.7315459694493615e-06, "loss": 0.0146, "step": 47091 }, { "epoch": 14.49, "learning_rate": 3.7311577532826946e-06, "loss": 0.0242, "step": 47092 }, { "epoch": 14.49, "learning_rate": 3.730769552679876e-06, "loss": 0.0138, "step": 47093 }, { "epoch": 14.49, "learning_rate": 3.7303813676418665e-06, "loss": 0.0156, "step": 47094 }, { "epoch": 14.49, "learning_rate": 3.72999319816963e-06, "loss": 0.0092, "step": 47095 }, { "epoch": 14.49, "learning_rate": 3.7296050442641306e-06, "loss": 0.0125, "step": 47096 }, { "epoch": 14.49, "learning_rate": 3.7292169059263305e-06, "loss": 0.0176, "step": 47097 }, { "epoch": 14.49, "learning_rate": 3.7288287831571946e-06, "loss": 0.0169, "step": 47098 }, { "epoch": 14.49, "learning_rate": 3.7284406759576864e-06, "loss": 0.0206, "step": 47099 }, { "epoch": 14.49, "learning_rate": 3.7280525843287698e-06, "loss": 0.0149, "step": 47100 }, { "epoch": 14.49, "learning_rate": 3.727664508271407e-06, "loss": 0.0103, "step": 47101 }, { "epoch": 14.49, "learning_rate": 3.7272764477865587e-06, "loss": 0.0238, "step": 47102 }, { "epoch": 14.49, "learning_rate": 3.726888402875195e-06, "loss": 0.0073, "step": 47103 }, { "epoch": 14.49, "learning_rate": 3.726500373538281e-06, "loss": 0.0167, "step": 47104 }, { "epoch": 14.49, "learning_rate": 3.72611235977677e-06, "loss": 0.0093, "step": 47105 }, { "epoch": 14.49, "learning_rate": 3.7257243615916317e-06, "loss": 0.0105, "step": 47106 }, { "epoch": 14.49, "learning_rate": 3.7253363789838247e-06, "loss": 0.0127, "step": 47107 }, { "epoch": 14.49, "learning_rate": 3.72494841195432e-06, "loss": 0.0076, "step": 47108 }, { "epoch": 14.49, "learning_rate": 3.7245604605040765e-06, "loss": 0.0135, "step": 47109 }, { "epoch": 14.5, "learning_rate": 3.7241725246340564e-06, "loss": 0.0231, "step": 47110 }, { "epoch": 14.5, "learning_rate": 3.7237846043452285e-06, "loss": 0.0036, "step": 47111 }, { "epoch": 14.5, "learning_rate": 3.723396699638544e-06, "loss": 0.024, "step": 47112 }, { "epoch": 14.5, "learning_rate": 3.7230088105149776e-06, "loss": 0.018, "step": 47113 }, { "epoch": 14.5, "learning_rate": 3.7226209369754875e-06, "loss": 0.0213, "step": 47114 }, { "epoch": 14.5, "learning_rate": 3.722233079021037e-06, "loss": 0.0147, "step": 47115 }, { "epoch": 14.5, "learning_rate": 3.7218452366525905e-06, "loss": 0.0102, "step": 47116 }, { "epoch": 14.5, "learning_rate": 3.721457409871109e-06, "loss": 0.0093, "step": 47117 }, { "epoch": 14.5, "learning_rate": 3.721069598677556e-06, "loss": 0.0192, "step": 47118 }, { "epoch": 14.5, "learning_rate": 3.7206818030728943e-06, "loss": 0.0114, "step": 47119 }, { "epoch": 14.5, "learning_rate": 3.7202940230580876e-06, "loss": 0.0427, "step": 47120 }, { "epoch": 14.5, "learning_rate": 3.7199062586340963e-06, "loss": 0.0082, "step": 47121 }, { "epoch": 14.5, "learning_rate": 3.7195185098018825e-06, "loss": 0.0164, "step": 47122 }, { "epoch": 14.5, "learning_rate": 3.719130776562414e-06, "loss": 0.0048, "step": 47123 }, { "epoch": 14.5, "learning_rate": 3.7187430589166516e-06, "loss": 0.0157, "step": 47124 }, { "epoch": 14.5, "learning_rate": 3.7183553568655594e-06, "loss": 0.0094, "step": 47125 }, { "epoch": 14.5, "learning_rate": 3.717967670410094e-06, "loss": 0.0224, "step": 47126 }, { "epoch": 14.5, "learning_rate": 3.7175799995512175e-06, "loss": 0.0109, "step": 47127 }, { "epoch": 14.5, "learning_rate": 3.7171923442899003e-06, "loss": 0.0269, "step": 47128 }, { "epoch": 14.5, "learning_rate": 3.7168047046271007e-06, "loss": 0.0172, "step": 47129 }, { "epoch": 14.5, "learning_rate": 3.71641708056378e-06, "loss": 0.0197, "step": 47130 }, { "epoch": 14.5, "learning_rate": 3.7160294721009026e-06, "loss": 0.0089, "step": 47131 }, { "epoch": 14.5, "learning_rate": 3.71564187923943e-06, "loss": 0.0164, "step": 47132 }, { "epoch": 14.5, "learning_rate": 3.7152543019803246e-06, "loss": 0.012, "step": 47133 }, { "epoch": 14.5, "learning_rate": 3.7148667403245477e-06, "loss": 0.0195, "step": 47134 }, { "epoch": 14.5, "learning_rate": 3.7144791942730617e-06, "loss": 0.0115, "step": 47135 }, { "epoch": 14.5, "learning_rate": 3.714091663826831e-06, "loss": 0.0238, "step": 47136 }, { "epoch": 14.5, "learning_rate": 3.713704148986812e-06, "loss": 0.0067, "step": 47137 }, { "epoch": 14.5, "learning_rate": 3.7133166497539773e-06, "loss": 0.0228, "step": 47138 }, { "epoch": 14.5, "learning_rate": 3.71292916612928e-06, "loss": 0.0223, "step": 47139 }, { "epoch": 14.5, "learning_rate": 3.712541698113683e-06, "loss": 0.0158, "step": 47140 }, { "epoch": 14.5, "learning_rate": 3.7121542457081515e-06, "loss": 0.0097, "step": 47141 }, { "epoch": 14.51, "learning_rate": 3.7117668089136416e-06, "loss": 0.0166, "step": 47142 }, { "epoch": 14.51, "learning_rate": 3.7113793877311233e-06, "loss": 0.0095, "step": 47143 }, { "epoch": 14.51, "learning_rate": 3.7109919821615546e-06, "loss": 0.0141, "step": 47144 }, { "epoch": 14.51, "learning_rate": 3.7106045922058974e-06, "loss": 0.0104, "step": 47145 }, { "epoch": 14.51, "learning_rate": 3.7102172178651164e-06, "loss": 0.0025, "step": 47146 }, { "epoch": 14.51, "learning_rate": 3.7098298591401637e-06, "loss": 0.0076, "step": 47147 }, { "epoch": 14.51, "learning_rate": 3.709442516032011e-06, "loss": 0.0109, "step": 47148 }, { "epoch": 14.51, "learning_rate": 3.709055188541617e-06, "loss": 0.0105, "step": 47149 }, { "epoch": 14.51, "learning_rate": 3.7086678766699435e-06, "loss": 0.0155, "step": 47150 }, { "epoch": 14.51, "learning_rate": 3.708280580417951e-06, "loss": 0.007, "step": 47151 }, { "epoch": 14.51, "learning_rate": 3.7078932997866014e-06, "loss": 0.0195, "step": 47152 }, { "epoch": 14.51, "learning_rate": 3.7075060347768564e-06, "loss": 0.0165, "step": 47153 }, { "epoch": 14.51, "learning_rate": 3.7071187853896774e-06, "loss": 0.0123, "step": 47154 }, { "epoch": 14.51, "learning_rate": 3.706731551626026e-06, "loss": 0.0162, "step": 47155 }, { "epoch": 14.51, "learning_rate": 3.706344333486864e-06, "loss": 0.0089, "step": 47156 }, { "epoch": 14.51, "learning_rate": 3.705957130973149e-06, "loss": 0.0108, "step": 47157 }, { "epoch": 14.51, "learning_rate": 3.7055699440858484e-06, "loss": 0.0135, "step": 47158 }, { "epoch": 14.51, "learning_rate": 3.7051827728259248e-06, "loss": 0.0152, "step": 47159 }, { "epoch": 14.51, "learning_rate": 3.704795617194331e-06, "loss": 0.0224, "step": 47160 }, { "epoch": 14.51, "learning_rate": 3.704408477192034e-06, "loss": 0.0309, "step": 47161 }, { "epoch": 14.51, "learning_rate": 3.704021352819989e-06, "loss": 0.0162, "step": 47162 }, { "epoch": 14.51, "learning_rate": 3.703634244079165e-06, "loss": 0.022, "step": 47163 }, { "epoch": 14.51, "learning_rate": 3.7032471509705203e-06, "loss": 0.008, "step": 47164 }, { "epoch": 14.51, "learning_rate": 3.702860073495015e-06, "loss": 0.0136, "step": 47165 }, { "epoch": 14.51, "learning_rate": 3.70247301165361e-06, "loss": 0.0141, "step": 47166 }, { "epoch": 14.51, "learning_rate": 3.7020859654472676e-06, "loss": 0.0133, "step": 47167 }, { "epoch": 14.51, "learning_rate": 3.7016989348769474e-06, "loss": 0.014, "step": 47168 }, { "epoch": 14.51, "learning_rate": 3.701311919943611e-06, "loss": 0.0155, "step": 47169 }, { "epoch": 14.51, "learning_rate": 3.700924920648219e-06, "loss": 0.0127, "step": 47170 }, { "epoch": 14.51, "learning_rate": 3.700537936991733e-06, "loss": 0.0222, "step": 47171 }, { "epoch": 14.51, "learning_rate": 3.7001509689751113e-06, "loss": 0.016, "step": 47172 }, { "epoch": 14.51, "learning_rate": 3.6997640165993178e-06, "loss": 0.0128, "step": 47173 }, { "epoch": 14.51, "learning_rate": 3.6993770798653107e-06, "loss": 0.0083, "step": 47174 }, { "epoch": 14.52, "learning_rate": 3.6989901587740527e-06, "loss": 0.0105, "step": 47175 }, { "epoch": 14.52, "learning_rate": 3.6986032533265027e-06, "loss": 0.0139, "step": 47176 }, { "epoch": 14.52, "learning_rate": 3.698216363523619e-06, "loss": 0.0118, "step": 47177 }, { "epoch": 14.52, "learning_rate": 3.697829489366368e-06, "loss": 0.0137, "step": 47178 }, { "epoch": 14.52, "learning_rate": 3.697442630855709e-06, "loss": 0.0147, "step": 47179 }, { "epoch": 14.52, "learning_rate": 3.6970557879926027e-06, "loss": 0.012, "step": 47180 }, { "epoch": 14.52, "learning_rate": 3.696668960778005e-06, "loss": 0.015, "step": 47181 }, { "epoch": 14.52, "learning_rate": 3.6962821492128742e-06, "loss": 0.0156, "step": 47182 }, { "epoch": 14.52, "learning_rate": 3.69589535329818e-06, "loss": 0.0162, "step": 47183 }, { "epoch": 14.52, "learning_rate": 3.6955085730348784e-06, "loss": 0.0164, "step": 47184 }, { "epoch": 14.52, "learning_rate": 3.695121808423928e-06, "loss": 0.0079, "step": 47185 }, { "epoch": 14.52, "learning_rate": 3.694735059466291e-06, "loss": 0.0196, "step": 47186 }, { "epoch": 14.52, "learning_rate": 3.6943483261629277e-06, "loss": 0.0111, "step": 47187 }, { "epoch": 14.52, "learning_rate": 3.6939616085147976e-06, "loss": 0.0257, "step": 47188 }, { "epoch": 14.52, "learning_rate": 3.69357490652286e-06, "loss": 0.0135, "step": 47189 }, { "epoch": 14.52, "learning_rate": 3.6931882201880766e-06, "loss": 0.0186, "step": 47190 }, { "epoch": 14.52, "learning_rate": 3.6928015495114055e-06, "loss": 0.013, "step": 47191 }, { "epoch": 14.52, "learning_rate": 3.6924148944938054e-06, "loss": 0.0134, "step": 47192 }, { "epoch": 14.52, "learning_rate": 3.692028255136244e-06, "loss": 0.0122, "step": 47193 }, { "epoch": 14.52, "learning_rate": 3.6916416314396743e-06, "loss": 0.0044, "step": 47194 }, { "epoch": 14.52, "learning_rate": 3.691255023405056e-06, "loss": 0.0231, "step": 47195 }, { "epoch": 14.52, "learning_rate": 3.690868431033352e-06, "loss": 0.0126, "step": 47196 }, { "epoch": 14.52, "learning_rate": 3.6904818543255174e-06, "loss": 0.0131, "step": 47197 }, { "epoch": 14.52, "learning_rate": 3.6900952932825175e-06, "loss": 0.0099, "step": 47198 }, { "epoch": 14.52, "learning_rate": 3.6897087479053106e-06, "loss": 0.0087, "step": 47199 }, { "epoch": 14.52, "learning_rate": 3.6893222181948564e-06, "loss": 0.0107, "step": 47200 }, { "epoch": 14.52, "learning_rate": 3.688935704152117e-06, "loss": 0.0236, "step": 47201 }, { "epoch": 14.52, "learning_rate": 3.6885492057780413e-06, "loss": 0.0159, "step": 47202 }, { "epoch": 14.52, "learning_rate": 3.6881627230735996e-06, "loss": 0.0233, "step": 47203 }, { "epoch": 14.52, "learning_rate": 3.687776256039749e-06, "loss": 0.0285, "step": 47204 }, { "epoch": 14.52, "learning_rate": 3.687389804677448e-06, "loss": 0.0055, "step": 47205 }, { "epoch": 14.52, "learning_rate": 3.687003368987656e-06, "loss": 0.0146, "step": 47206 }, { "epoch": 14.53, "learning_rate": 3.6866169489713334e-06, "loss": 0.0255, "step": 47207 }, { "epoch": 14.53, "learning_rate": 3.686230544629439e-06, "loss": 0.0259, "step": 47208 }, { "epoch": 14.53, "learning_rate": 3.685844155962931e-06, "loss": 0.0087, "step": 47209 }, { "epoch": 14.53, "learning_rate": 3.68545778297277e-06, "loss": 0.0108, "step": 47210 }, { "epoch": 14.53, "learning_rate": 3.685071425659915e-06, "loss": 0.0133, "step": 47211 }, { "epoch": 14.53, "learning_rate": 3.6846850840253222e-06, "loss": 0.0167, "step": 47212 }, { "epoch": 14.53, "learning_rate": 3.6842987580699573e-06, "loss": 0.0157, "step": 47213 }, { "epoch": 14.53, "learning_rate": 3.6839124477947786e-06, "loss": 0.0091, "step": 47214 }, { "epoch": 14.53, "learning_rate": 3.6835261532007395e-06, "loss": 0.0175, "step": 47215 }, { "epoch": 14.53, "learning_rate": 3.6831398742888026e-06, "loss": 0.011, "step": 47216 }, { "epoch": 14.53, "learning_rate": 3.6827536110599226e-06, "loss": 0.0158, "step": 47217 }, { "epoch": 14.53, "learning_rate": 3.682367363515066e-06, "loss": 0.0149, "step": 47218 }, { "epoch": 14.53, "learning_rate": 3.6819811316551867e-06, "loss": 0.0187, "step": 47219 }, { "epoch": 14.53, "learning_rate": 3.6815949154812468e-06, "loss": 0.0148, "step": 47220 }, { "epoch": 14.53, "learning_rate": 3.681208714994202e-06, "loss": 0.0102, "step": 47221 }, { "epoch": 14.53, "learning_rate": 3.680822530195013e-06, "loss": 0.0144, "step": 47222 }, { "epoch": 14.53, "learning_rate": 3.680436361084638e-06, "loss": 0.009, "step": 47223 }, { "epoch": 14.53, "learning_rate": 3.6800502076640353e-06, "loss": 0.009, "step": 47224 }, { "epoch": 14.53, "learning_rate": 3.6796640699341656e-06, "loss": 0.0102, "step": 47225 }, { "epoch": 14.53, "learning_rate": 3.679277947895985e-06, "loss": 0.0149, "step": 47226 }, { "epoch": 14.53, "learning_rate": 3.6788918415504537e-06, "loss": 0.0214, "step": 47227 }, { "epoch": 14.53, "learning_rate": 3.678505750898529e-06, "loss": 0.0157, "step": 47228 }, { "epoch": 14.53, "learning_rate": 3.678119675941171e-06, "loss": 0.0101, "step": 47229 }, { "epoch": 14.53, "learning_rate": 3.6777336166793377e-06, "loss": 0.0164, "step": 47230 }, { "epoch": 14.53, "learning_rate": 3.677347573113986e-06, "loss": 0.0135, "step": 47231 }, { "epoch": 14.53, "learning_rate": 3.676961545246074e-06, "loss": 0.0087, "step": 47232 }, { "epoch": 14.53, "learning_rate": 3.676575533076565e-06, "loss": 0.0084, "step": 47233 }, { "epoch": 14.53, "learning_rate": 3.6761895366064135e-06, "loss": 0.0142, "step": 47234 }, { "epoch": 14.53, "learning_rate": 3.675803555836582e-06, "loss": 0.0086, "step": 47235 }, { "epoch": 14.53, "learning_rate": 3.675417590768022e-06, "loss": 0.015, "step": 47236 }, { "epoch": 14.53, "learning_rate": 3.675031641401693e-06, "loss": 0.0123, "step": 47237 }, { "epoch": 14.53, "learning_rate": 3.6746457077385567e-06, "loss": 0.0066, "step": 47238 }, { "epoch": 14.53, "learning_rate": 3.6742597897795717e-06, "loss": 0.0166, "step": 47239 }, { "epoch": 14.54, "learning_rate": 3.673873887525693e-06, "loss": 0.012, "step": 47240 }, { "epoch": 14.54, "learning_rate": 3.6734880009778806e-06, "loss": 0.0146, "step": 47241 }, { "epoch": 14.54, "learning_rate": 3.6731021301370905e-06, "loss": 0.0114, "step": 47242 }, { "epoch": 14.54, "learning_rate": 3.6727162750042833e-06, "loss": 0.0163, "step": 47243 }, { "epoch": 14.54, "learning_rate": 3.6723304355804157e-06, "loss": 0.0209, "step": 47244 }, { "epoch": 14.54, "learning_rate": 3.6719446118664458e-06, "loss": 0.0132, "step": 47245 }, { "epoch": 14.54, "learning_rate": 3.6715588038633323e-06, "loss": 0.0071, "step": 47246 }, { "epoch": 14.54, "learning_rate": 3.6711730115720277e-06, "loss": 0.0126, "step": 47247 }, { "epoch": 14.54, "learning_rate": 3.6707872349935016e-06, "loss": 0.0128, "step": 47248 }, { "epoch": 14.54, "learning_rate": 3.6704014741286996e-06, "loss": 0.0186, "step": 47249 }, { "epoch": 14.54, "learning_rate": 3.6700157289785864e-06, "loss": 0.0187, "step": 47250 }, { "epoch": 14.54, "learning_rate": 3.6696299995441167e-06, "loss": 0.0142, "step": 47251 }, { "epoch": 14.54, "learning_rate": 3.6692442858262455e-06, "loss": 0.0244, "step": 47252 }, { "epoch": 14.54, "learning_rate": 3.6688585878259375e-06, "loss": 0.0065, "step": 47253 }, { "epoch": 14.54, "learning_rate": 3.668472905544147e-06, "loss": 0.0079, "step": 47254 }, { "epoch": 14.54, "learning_rate": 3.668087238981831e-06, "loss": 0.0118, "step": 47255 }, { "epoch": 14.54, "learning_rate": 3.6677015881399513e-06, "loss": 0.0094, "step": 47256 }, { "epoch": 14.54, "learning_rate": 3.667315953019455e-06, "loss": 0.0073, "step": 47257 }, { "epoch": 14.54, "learning_rate": 3.6669303336213082e-06, "loss": 0.0185, "step": 47258 }, { "epoch": 14.54, "learning_rate": 3.6665447299464665e-06, "loss": 0.004, "step": 47259 }, { "epoch": 14.54, "learning_rate": 3.6661591419958876e-06, "loss": 0.013, "step": 47260 }, { "epoch": 14.54, "learning_rate": 3.6657735697705267e-06, "loss": 0.0074, "step": 47261 }, { "epoch": 14.54, "learning_rate": 3.6653880132713425e-06, "loss": 0.009, "step": 47262 }, { "epoch": 14.54, "learning_rate": 3.6650024724992926e-06, "loss": 0.0065, "step": 47263 }, { "epoch": 14.54, "learning_rate": 3.6646169474553338e-06, "loss": 0.011, "step": 47264 }, { "epoch": 14.54, "learning_rate": 3.664231438140423e-06, "loss": 0.0092, "step": 47265 }, { "epoch": 14.54, "learning_rate": 3.6638459445555174e-06, "loss": 0.0221, "step": 47266 }, { "epoch": 14.54, "learning_rate": 3.6634604667015717e-06, "loss": 0.024, "step": 47267 }, { "epoch": 14.54, "learning_rate": 3.6630750045795472e-06, "loss": 0.009, "step": 47268 }, { "epoch": 14.54, "learning_rate": 3.662689558190403e-06, "loss": 0.0092, "step": 47269 }, { "epoch": 14.54, "learning_rate": 3.662304127535089e-06, "loss": 0.0113, "step": 47270 }, { "epoch": 14.54, "learning_rate": 3.661918712614566e-06, "loss": 0.0239, "step": 47271 }, { "epoch": 14.55, "learning_rate": 3.661533313429786e-06, "loss": 0.011, "step": 47272 }, { "epoch": 14.55, "learning_rate": 3.6611479299817134e-06, "loss": 0.0075, "step": 47273 }, { "epoch": 14.55, "learning_rate": 3.6607625622713005e-06, "loss": 0.0227, "step": 47274 }, { "epoch": 14.55, "learning_rate": 3.660377210299506e-06, "loss": 0.0113, "step": 47275 }, { "epoch": 14.55, "learning_rate": 3.6599918740672867e-06, "loss": 0.02, "step": 47276 }, { "epoch": 14.55, "learning_rate": 3.6596065535755975e-06, "loss": 0.0105, "step": 47277 }, { "epoch": 14.55, "learning_rate": 3.659221248825395e-06, "loss": 0.0173, "step": 47278 }, { "epoch": 14.55, "learning_rate": 3.658835959817638e-06, "loss": 0.0162, "step": 47279 }, { "epoch": 14.55, "learning_rate": 3.658450686553282e-06, "loss": 0.0123, "step": 47280 }, { "epoch": 14.55, "learning_rate": 3.658065429033283e-06, "loss": 0.0167, "step": 47281 }, { "epoch": 14.55, "learning_rate": 3.6576801872585945e-06, "loss": 0.0129, "step": 47282 }, { "epoch": 14.55, "learning_rate": 3.6572949612301824e-06, "loss": 0.0126, "step": 47283 }, { "epoch": 14.55, "learning_rate": 3.6569097509489938e-06, "loss": 0.0174, "step": 47284 }, { "epoch": 14.55, "learning_rate": 3.656524556415989e-06, "loss": 0.0158, "step": 47285 }, { "epoch": 14.55, "learning_rate": 3.656139377632123e-06, "loss": 0.0119, "step": 47286 }, { "epoch": 14.55, "learning_rate": 3.6557542145983495e-06, "loss": 0.0236, "step": 47287 }, { "epoch": 14.55, "learning_rate": 3.6553690673156306e-06, "loss": 0.0122, "step": 47288 }, { "epoch": 14.55, "learning_rate": 3.6549839357849205e-06, "loss": 0.014, "step": 47289 }, { "epoch": 14.55, "learning_rate": 3.654598820007178e-06, "loss": 0.0092, "step": 47290 }, { "epoch": 14.55, "learning_rate": 3.6542137199833515e-06, "loss": 0.0091, "step": 47291 }, { "epoch": 14.55, "learning_rate": 3.6538286357143993e-06, "loss": 0.0096, "step": 47292 }, { "epoch": 14.55, "learning_rate": 3.6534435672012836e-06, "loss": 0.018, "step": 47293 }, { "epoch": 14.55, "learning_rate": 3.653058514444956e-06, "loss": 0.0166, "step": 47294 }, { "epoch": 14.55, "learning_rate": 3.6526734774463724e-06, "loss": 0.0185, "step": 47295 }, { "epoch": 14.55, "learning_rate": 3.6522884562064908e-06, "loss": 0.0139, "step": 47296 }, { "epoch": 14.55, "learning_rate": 3.651903450726264e-06, "loss": 0.0136, "step": 47297 }, { "epoch": 14.55, "learning_rate": 3.6515184610066514e-06, "loss": 0.0169, "step": 47298 }, { "epoch": 14.55, "learning_rate": 3.651133487048605e-06, "loss": 0.0098, "step": 47299 }, { "epoch": 14.55, "learning_rate": 3.650748528853084e-06, "loss": 0.0113, "step": 47300 }, { "epoch": 14.55, "learning_rate": 3.650363586421042e-06, "loss": 0.0173, "step": 47301 }, { "epoch": 14.55, "learning_rate": 3.6499786597534315e-06, "loss": 0.0255, "step": 47302 }, { "epoch": 14.55, "learning_rate": 3.649593748851217e-06, "loss": 0.0091, "step": 47303 }, { "epoch": 14.55, "learning_rate": 3.6492088537153515e-06, "loss": 0.0137, "step": 47304 }, { "epoch": 14.56, "learning_rate": 3.6488239743467847e-06, "loss": 0.0108, "step": 47305 }, { "epoch": 14.56, "learning_rate": 3.648439110746477e-06, "loss": 0.0082, "step": 47306 }, { "epoch": 14.56, "learning_rate": 3.648054262915378e-06, "loss": 0.0123, "step": 47307 }, { "epoch": 14.56, "learning_rate": 3.6476694308544514e-06, "loss": 0.0081, "step": 47308 }, { "epoch": 14.56, "learning_rate": 3.64728461456465e-06, "loss": 0.0142, "step": 47309 }, { "epoch": 14.56, "learning_rate": 3.646899814046927e-06, "loss": 0.0139, "step": 47310 }, { "epoch": 14.56, "learning_rate": 3.6465150293022433e-06, "loss": 0.0144, "step": 47311 }, { "epoch": 14.56, "learning_rate": 3.646130260331543e-06, "loss": 0.0197, "step": 47312 }, { "epoch": 14.56, "learning_rate": 3.6457455071357918e-06, "loss": 0.0172, "step": 47313 }, { "epoch": 14.56, "learning_rate": 3.645360769715941e-06, "loss": 0.0167, "step": 47314 }, { "epoch": 14.56, "learning_rate": 3.6449760480729467e-06, "loss": 0.0099, "step": 47315 }, { "epoch": 14.56, "learning_rate": 3.6445913422077627e-06, "loss": 0.0201, "step": 47316 }, { "epoch": 14.56, "learning_rate": 3.644206652121346e-06, "loss": 0.0124, "step": 47317 }, { "epoch": 14.56, "learning_rate": 3.643821977814651e-06, "loss": 0.0101, "step": 47318 }, { "epoch": 14.56, "learning_rate": 3.643437319288632e-06, "loss": 0.0279, "step": 47319 }, { "epoch": 14.56, "learning_rate": 3.6430526765442443e-06, "loss": 0.0092, "step": 47320 }, { "epoch": 14.56, "learning_rate": 3.642668049582443e-06, "loss": 0.0205, "step": 47321 }, { "epoch": 14.56, "learning_rate": 3.6422834384041792e-06, "loss": 0.0231, "step": 47322 }, { "epoch": 14.56, "learning_rate": 3.641898843010415e-06, "loss": 0.0164, "step": 47323 }, { "epoch": 14.56, "learning_rate": 3.6415142634021026e-06, "loss": 0.019, "step": 47324 }, { "epoch": 14.56, "learning_rate": 3.6411296995802002e-06, "loss": 0.0079, "step": 47325 }, { "epoch": 14.56, "learning_rate": 3.6407451515456537e-06, "loss": 0.0121, "step": 47326 }, { "epoch": 14.56, "learning_rate": 3.64036061929942e-06, "loss": 0.0228, "step": 47327 }, { "epoch": 14.56, "learning_rate": 3.6399761028424596e-06, "loss": 0.0069, "step": 47328 }, { "epoch": 14.56, "learning_rate": 3.639591602175724e-06, "loss": 0.0111, "step": 47329 }, { "epoch": 14.56, "learning_rate": 3.6392071173001674e-06, "loss": 0.0267, "step": 47330 }, { "epoch": 14.56, "learning_rate": 3.6388226482167445e-06, "loss": 0.0282, "step": 47331 }, { "epoch": 14.56, "learning_rate": 3.63843819492641e-06, "loss": 0.0105, "step": 47332 }, { "epoch": 14.56, "learning_rate": 3.638053757430119e-06, "loss": 0.0104, "step": 47333 }, { "epoch": 14.56, "learning_rate": 3.637669335728825e-06, "loss": 0.0109, "step": 47334 }, { "epoch": 14.56, "learning_rate": 3.6372849298234835e-06, "loss": 0.0078, "step": 47335 }, { "epoch": 14.56, "learning_rate": 3.636900539715048e-06, "loss": 0.0108, "step": 47336 }, { "epoch": 14.57, "learning_rate": 3.6365161654044687e-06, "loss": 0.0089, "step": 47337 }, { "epoch": 14.57, "learning_rate": 3.636131806892712e-06, "loss": 0.012, "step": 47338 }, { "epoch": 14.57, "learning_rate": 3.63574746418072e-06, "loss": 0.0088, "step": 47339 }, { "epoch": 14.57, "learning_rate": 3.635363137269452e-06, "loss": 0.0204, "step": 47340 }, { "epoch": 14.57, "learning_rate": 3.6349788261598618e-06, "loss": 0.007, "step": 47341 }, { "epoch": 14.57, "learning_rate": 3.6345945308528997e-06, "loss": 0.0112, "step": 47342 }, { "epoch": 14.57, "learning_rate": 3.634210251349526e-06, "loss": 0.0098, "step": 47343 }, { "epoch": 14.57, "learning_rate": 3.633825987650693e-06, "loss": 0.01, "step": 47344 }, { "epoch": 14.57, "learning_rate": 3.633441739757353e-06, "loss": 0.0088, "step": 47345 }, { "epoch": 14.57, "learning_rate": 3.6330575076704644e-06, "loss": 0.0091, "step": 47346 }, { "epoch": 14.57, "learning_rate": 3.632673291390971e-06, "loss": 0.0177, "step": 47347 }, { "epoch": 14.57, "learning_rate": 3.6322890909198362e-06, "loss": 0.0249, "step": 47348 }, { "epoch": 14.57, "learning_rate": 3.6319049062580114e-06, "loss": 0.0067, "step": 47349 }, { "epoch": 14.57, "learning_rate": 3.6315207374064496e-06, "loss": 0.0262, "step": 47350 }, { "epoch": 14.57, "learning_rate": 3.631136584366105e-06, "loss": 0.0069, "step": 47351 }, { "epoch": 14.57, "learning_rate": 3.6307524471379317e-06, "loss": 0.0216, "step": 47352 }, { "epoch": 14.57, "learning_rate": 3.6303683257228827e-06, "loss": 0.0079, "step": 47353 }, { "epoch": 14.57, "learning_rate": 3.6299842201219117e-06, "loss": 0.0134, "step": 47354 }, { "epoch": 14.57, "learning_rate": 3.629600130335973e-06, "loss": 0.0123, "step": 47355 }, { "epoch": 14.57, "learning_rate": 3.6292160563660196e-06, "loss": 0.0121, "step": 47356 }, { "epoch": 14.57, "learning_rate": 3.6288319982130014e-06, "loss": 0.0047, "step": 47357 }, { "epoch": 14.57, "learning_rate": 3.6284479558778796e-06, "loss": 0.0116, "step": 47358 }, { "epoch": 14.57, "learning_rate": 3.6280639293616072e-06, "loss": 0.0084, "step": 47359 }, { "epoch": 14.57, "learning_rate": 3.6276799186651303e-06, "loss": 0.017, "step": 47360 }, { "epoch": 14.57, "learning_rate": 3.627295923789407e-06, "loss": 0.0164, "step": 47361 }, { "epoch": 14.57, "learning_rate": 3.626911944735385e-06, "loss": 0.0131, "step": 47362 }, { "epoch": 14.57, "learning_rate": 3.626527981504028e-06, "loss": 0.0136, "step": 47363 }, { "epoch": 14.57, "learning_rate": 3.626144034096283e-06, "loss": 0.014, "step": 47364 }, { "epoch": 14.57, "learning_rate": 3.625760102513103e-06, "loss": 0.0087, "step": 47365 }, { "epoch": 14.57, "learning_rate": 3.625376186755444e-06, "loss": 0.0126, "step": 47366 }, { "epoch": 14.57, "learning_rate": 3.624992286824256e-06, "loss": 0.0091, "step": 47367 }, { "epoch": 14.57, "learning_rate": 3.624608402720494e-06, "loss": 0.0171, "step": 47368 }, { "epoch": 14.57, "learning_rate": 3.62422453444511e-06, "loss": 0.0077, "step": 47369 }, { "epoch": 14.58, "learning_rate": 3.623840681999058e-06, "loss": 0.0157, "step": 47370 }, { "epoch": 14.58, "learning_rate": 3.6234568453832906e-06, "loss": 0.0157, "step": 47371 }, { "epoch": 14.58, "learning_rate": 3.623073024598761e-06, "loss": 0.0226, "step": 47372 }, { "epoch": 14.58, "learning_rate": 3.6226892196464225e-06, "loss": 0.0116, "step": 47373 }, { "epoch": 14.58, "learning_rate": 3.6223054305272264e-06, "loss": 0.0094, "step": 47374 }, { "epoch": 14.58, "learning_rate": 3.6219216572421266e-06, "loss": 0.0138, "step": 47375 }, { "epoch": 14.58, "learning_rate": 3.621537899792077e-06, "loss": 0.0127, "step": 47376 }, { "epoch": 14.58, "learning_rate": 3.6211541581780252e-06, "loss": 0.0124, "step": 47377 }, { "epoch": 14.58, "learning_rate": 3.6207704324009308e-06, "loss": 0.0242, "step": 47378 }, { "epoch": 14.58, "learning_rate": 3.620386722461744e-06, "loss": 0.0128, "step": 47379 }, { "epoch": 14.58, "learning_rate": 3.6200030283614205e-06, "loss": 0.0238, "step": 47380 }, { "epoch": 14.58, "learning_rate": 3.619619350100906e-06, "loss": 0.0137, "step": 47381 }, { "epoch": 14.58, "learning_rate": 3.619235687681153e-06, "loss": 0.0175, "step": 47382 }, { "epoch": 14.58, "learning_rate": 3.6188520411031214e-06, "loss": 0.0084, "step": 47383 }, { "epoch": 14.58, "learning_rate": 3.6184684103677594e-06, "loss": 0.01, "step": 47384 }, { "epoch": 14.58, "learning_rate": 3.61808479547602e-06, "loss": 0.0077, "step": 47385 }, { "epoch": 14.58, "learning_rate": 3.6177011964288555e-06, "loss": 0.0153, "step": 47386 }, { "epoch": 14.58, "learning_rate": 3.617317613227219e-06, "loss": 0.018, "step": 47387 }, { "epoch": 14.58, "learning_rate": 3.6169340458720613e-06, "loss": 0.0194, "step": 47388 }, { "epoch": 14.58, "learning_rate": 3.616550494364336e-06, "loss": 0.0086, "step": 47389 }, { "epoch": 14.58, "learning_rate": 3.6161669587049953e-06, "loss": 0.0133, "step": 47390 }, { "epoch": 14.58, "learning_rate": 3.6157834388949907e-06, "loss": 0.0227, "step": 47391 }, { "epoch": 14.58, "learning_rate": 3.615399934935271e-06, "loss": 0.0135, "step": 47392 }, { "epoch": 14.58, "learning_rate": 3.615016446826799e-06, "loss": 0.0071, "step": 47393 }, { "epoch": 14.58, "learning_rate": 3.6146329745705157e-06, "loss": 0.0128, "step": 47394 }, { "epoch": 14.58, "learning_rate": 3.6142495181673787e-06, "loss": 0.0101, "step": 47395 }, { "epoch": 14.58, "learning_rate": 3.6138660776183378e-06, "loss": 0.0069, "step": 47396 }, { "epoch": 14.58, "learning_rate": 3.6134826529243416e-06, "loss": 0.0183, "step": 47397 }, { "epoch": 14.58, "learning_rate": 3.613099244086351e-06, "loss": 0.0222, "step": 47398 }, { "epoch": 14.58, "learning_rate": 3.612715851105313e-06, "loss": 0.0157, "step": 47399 }, { "epoch": 14.58, "learning_rate": 3.6123324739821784e-06, "loss": 0.013, "step": 47400 }, { "epoch": 14.58, "learning_rate": 3.6119491127179053e-06, "loss": 0.0109, "step": 47401 }, { "epoch": 14.59, "learning_rate": 3.6115657673134317e-06, "loss": 0.0188, "step": 47402 }, { "epoch": 14.59, "learning_rate": 3.6111824377697225e-06, "loss": 0.0116, "step": 47403 }, { "epoch": 14.59, "learning_rate": 3.610799124087725e-06, "loss": 0.0062, "step": 47404 }, { "epoch": 14.59, "learning_rate": 3.6104158262683907e-06, "loss": 0.0177, "step": 47405 }, { "epoch": 14.59, "learning_rate": 3.6100325443126717e-06, "loss": 0.0231, "step": 47406 }, { "epoch": 14.59, "learning_rate": 3.6096492782215186e-06, "loss": 0.0083, "step": 47407 }, { "epoch": 14.59, "learning_rate": 3.6092660279958838e-06, "loss": 0.0142, "step": 47408 }, { "epoch": 14.59, "learning_rate": 3.6088827936367177e-06, "loss": 0.0118, "step": 47409 }, { "epoch": 14.59, "learning_rate": 3.6084995751449736e-06, "loss": 0.0095, "step": 47410 }, { "epoch": 14.59, "learning_rate": 3.608116372521602e-06, "loss": 0.0154, "step": 47411 }, { "epoch": 14.59, "learning_rate": 3.60773318576755e-06, "loss": 0.0139, "step": 47412 }, { "epoch": 14.59, "learning_rate": 3.607350014883778e-06, "loss": 0.0062, "step": 47413 }, { "epoch": 14.59, "learning_rate": 3.606966859871235e-06, "loss": 0.0097, "step": 47414 }, { "epoch": 14.59, "learning_rate": 3.6065837207308663e-06, "loss": 0.0163, "step": 47415 }, { "epoch": 14.59, "learning_rate": 3.606200597463627e-06, "loss": 0.0107, "step": 47416 }, { "epoch": 14.59, "learning_rate": 3.6058174900704646e-06, "loss": 0.0126, "step": 47417 }, { "epoch": 14.59, "learning_rate": 3.6054343985523365e-06, "loss": 0.0064, "step": 47418 }, { "epoch": 14.59, "learning_rate": 3.6050513229101914e-06, "loss": 0.007, "step": 47419 }, { "epoch": 14.59, "learning_rate": 3.604668263144979e-06, "loss": 0.0118, "step": 47420 }, { "epoch": 14.59, "learning_rate": 3.6042852192576516e-06, "loss": 0.0188, "step": 47421 }, { "epoch": 14.59, "learning_rate": 3.60390219124916e-06, "loss": 0.0111, "step": 47422 }, { "epoch": 14.59, "learning_rate": 3.603519179120455e-06, "loss": 0.02, "step": 47423 }, { "epoch": 14.59, "learning_rate": 3.6031361828724863e-06, "loss": 0.0164, "step": 47424 }, { "epoch": 14.59, "learning_rate": 3.602753202506207e-06, "loss": 0.0102, "step": 47425 }, { "epoch": 14.59, "learning_rate": 3.6023702380225665e-06, "loss": 0.0174, "step": 47426 }, { "epoch": 14.59, "learning_rate": 3.601987289422516e-06, "loss": 0.0099, "step": 47427 }, { "epoch": 14.59, "learning_rate": 3.601604356707007e-06, "loss": 0.0142, "step": 47428 }, { "epoch": 14.59, "learning_rate": 3.6012214398769886e-06, "loss": 0.0107, "step": 47429 }, { "epoch": 14.59, "learning_rate": 3.6008385389334122e-06, "loss": 0.0105, "step": 47430 }, { "epoch": 14.59, "learning_rate": 3.6004556538772293e-06, "loss": 0.0126, "step": 47431 }, { "epoch": 14.59, "learning_rate": 3.6000727847093854e-06, "loss": 0.0177, "step": 47432 }, { "epoch": 14.59, "learning_rate": 3.59968993143084e-06, "loss": 0.0144, "step": 47433 }, { "epoch": 14.59, "learning_rate": 3.5993070940425377e-06, "loss": 0.0121, "step": 47434 }, { "epoch": 14.6, "learning_rate": 3.5989242725454345e-06, "loss": 0.0064, "step": 47435 }, { "epoch": 14.6, "learning_rate": 3.5985414669404728e-06, "loss": 0.0127, "step": 47436 }, { "epoch": 14.6, "learning_rate": 3.5981586772286036e-06, "loss": 0.0071, "step": 47437 }, { "epoch": 14.6, "learning_rate": 3.597775903410784e-06, "loss": 0.013, "step": 47438 }, { "epoch": 14.6, "learning_rate": 3.5973931454879605e-06, "loss": 0.0144, "step": 47439 }, { "epoch": 14.6, "learning_rate": 3.5970104034610843e-06, "loss": 0.0182, "step": 47440 }, { "epoch": 14.6, "learning_rate": 3.5966276773311047e-06, "loss": 0.0109, "step": 47441 }, { "epoch": 14.6, "learning_rate": 3.596244967098973e-06, "loss": 0.0149, "step": 47442 }, { "epoch": 14.6, "learning_rate": 3.595862272765638e-06, "loss": 0.009, "step": 47443 }, { "epoch": 14.6, "learning_rate": 3.595479594332051e-06, "loss": 0.0104, "step": 47444 }, { "epoch": 14.6, "learning_rate": 3.5950969317991613e-06, "loss": 0.0107, "step": 47445 }, { "epoch": 14.6, "learning_rate": 3.5947142851679185e-06, "loss": 0.0178, "step": 47446 }, { "epoch": 14.6, "learning_rate": 3.5943316544392703e-06, "loss": 0.0078, "step": 47447 }, { "epoch": 14.6, "learning_rate": 3.5939490396141774e-06, "loss": 0.0118, "step": 47448 }, { "epoch": 14.6, "learning_rate": 3.5935664406935777e-06, "loss": 0.0085, "step": 47449 }, { "epoch": 14.6, "learning_rate": 3.593183857678425e-06, "loss": 0.0293, "step": 47450 }, { "epoch": 14.6, "learning_rate": 3.5928012905696707e-06, "loss": 0.0145, "step": 47451 }, { "epoch": 14.6, "learning_rate": 3.59241873936826e-06, "loss": 0.0115, "step": 47452 }, { "epoch": 14.6, "learning_rate": 3.5920362040751476e-06, "loss": 0.0113, "step": 47453 }, { "epoch": 14.6, "learning_rate": 3.5916536846912843e-06, "loss": 0.0185, "step": 47454 }, { "epoch": 14.6, "learning_rate": 3.591271181217616e-06, "loss": 0.014, "step": 47455 }, { "epoch": 14.6, "learning_rate": 3.5908886936550967e-06, "loss": 0.0172, "step": 47456 }, { "epoch": 14.6, "learning_rate": 3.5905062220046672e-06, "loss": 0.0092, "step": 47457 }, { "epoch": 14.6, "learning_rate": 3.590123766267285e-06, "loss": 0.0128, "step": 47458 }, { "epoch": 14.6, "learning_rate": 3.5897413264438985e-06, "loss": 0.0082, "step": 47459 }, { "epoch": 14.6, "learning_rate": 3.5893589025354558e-06, "loss": 0.015, "step": 47460 }, { "epoch": 14.6, "learning_rate": 3.588976494542906e-06, "loss": 0.0086, "step": 47461 }, { "epoch": 14.6, "learning_rate": 3.5885941024672e-06, "loss": 0.0148, "step": 47462 }, { "epoch": 14.6, "learning_rate": 3.5882117263092862e-06, "loss": 0.0118, "step": 47463 }, { "epoch": 14.6, "learning_rate": 3.5878293660701146e-06, "loss": 0.0123, "step": 47464 }, { "epoch": 14.6, "learning_rate": 3.587447021750633e-06, "loss": 0.0131, "step": 47465 }, { "epoch": 14.6, "learning_rate": 3.5870646933517928e-06, "loss": 0.0157, "step": 47466 }, { "epoch": 14.61, "learning_rate": 3.5866823808745377e-06, "loss": 0.0225, "step": 47467 }, { "epoch": 14.61, "learning_rate": 3.586300084319825e-06, "loss": 0.0111, "step": 47468 }, { "epoch": 14.61, "learning_rate": 3.585917803688603e-06, "loss": 0.0169, "step": 47469 }, { "epoch": 14.61, "learning_rate": 3.585535538981816e-06, "loss": 0.0357, "step": 47470 }, { "epoch": 14.61, "learning_rate": 3.5851532902004137e-06, "loss": 0.0096, "step": 47471 }, { "epoch": 14.61, "learning_rate": 3.5847710573453433e-06, "loss": 0.0116, "step": 47472 }, { "epoch": 14.61, "learning_rate": 3.5843888404175607e-06, "loss": 0.009, "step": 47473 }, { "epoch": 14.61, "learning_rate": 3.5840066394180107e-06, "loss": 0.0109, "step": 47474 }, { "epoch": 14.61, "learning_rate": 3.5836244543476417e-06, "loss": 0.0116, "step": 47475 }, { "epoch": 14.61, "learning_rate": 3.5832422852074043e-06, "loss": 0.0081, "step": 47476 }, { "epoch": 14.61, "learning_rate": 3.582860131998246e-06, "loss": 0.019, "step": 47477 }, { "epoch": 14.61, "learning_rate": 3.5824779947211164e-06, "loss": 0.0114, "step": 47478 }, { "epoch": 14.61, "learning_rate": 3.582095873376964e-06, "loss": 0.0155, "step": 47479 }, { "epoch": 14.61, "learning_rate": 3.581713767966737e-06, "loss": 0.009, "step": 47480 }, { "epoch": 14.61, "learning_rate": 3.5813316784913844e-06, "loss": 0.008, "step": 47481 }, { "epoch": 14.61, "learning_rate": 3.580949604951851e-06, "loss": 0.0127, "step": 47482 }, { "epoch": 14.61, "learning_rate": 3.5805675473490964e-06, "loss": 0.0102, "step": 47483 }, { "epoch": 14.61, "learning_rate": 3.5801855056840586e-06, "loss": 0.0062, "step": 47484 }, { "epoch": 14.61, "learning_rate": 3.57980347995769e-06, "loss": 0.0284, "step": 47485 }, { "epoch": 14.61, "learning_rate": 3.579421470170934e-06, "loss": 0.0084, "step": 47486 }, { "epoch": 14.61, "learning_rate": 3.579039476324747e-06, "loss": 0.0136, "step": 47487 }, { "epoch": 14.61, "learning_rate": 3.578657498420075e-06, "loss": 0.0087, "step": 47488 }, { "epoch": 14.61, "learning_rate": 3.578275536457865e-06, "loss": 0.0122, "step": 47489 }, { "epoch": 14.61, "learning_rate": 3.5778935904390688e-06, "loss": 0.0068, "step": 47490 }, { "epoch": 14.61, "learning_rate": 3.5775116603646244e-06, "loss": 0.0058, "step": 47491 }, { "epoch": 14.61, "learning_rate": 3.5771297462354905e-06, "loss": 0.0105, "step": 47492 }, { "epoch": 14.61, "learning_rate": 3.5767478480526117e-06, "loss": 0.0155, "step": 47493 }, { "epoch": 14.61, "learning_rate": 3.576365965816937e-06, "loss": 0.016, "step": 47494 }, { "epoch": 14.61, "learning_rate": 3.5759840995294136e-06, "loss": 0.0116, "step": 47495 }, { "epoch": 14.61, "learning_rate": 3.5756022491909904e-06, "loss": 0.0138, "step": 47496 }, { "epoch": 14.61, "learning_rate": 3.575220414802614e-06, "loss": 0.0143, "step": 47497 }, { "epoch": 14.61, "learning_rate": 3.5748385963652333e-06, "loss": 0.0192, "step": 47498 }, { "epoch": 14.61, "learning_rate": 3.574456793879796e-06, "loss": 0.0122, "step": 47499 }, { "epoch": 14.62, "learning_rate": 3.5740750073472508e-06, "loss": 0.0128, "step": 47500 }, { "epoch": 14.62, "learning_rate": 3.5736932367685416e-06, "loss": 0.0071, "step": 47501 }, { "epoch": 14.62, "learning_rate": 3.5733114821446225e-06, "loss": 0.0156, "step": 47502 }, { "epoch": 14.62, "learning_rate": 3.572929743476439e-06, "loss": 0.0288, "step": 47503 }, { "epoch": 14.62, "learning_rate": 3.572548020764941e-06, "loss": 0.0103, "step": 47504 }, { "epoch": 14.62, "learning_rate": 3.5721663140110706e-06, "loss": 0.0096, "step": 47505 }, { "epoch": 14.62, "learning_rate": 3.571784623215775e-06, "loss": 0.0107, "step": 47506 }, { "epoch": 14.62, "learning_rate": 3.5714029483800085e-06, "loss": 0.016, "step": 47507 }, { "epoch": 14.62, "learning_rate": 3.5710212895047146e-06, "loss": 0.021, "step": 47508 }, { "epoch": 14.62, "learning_rate": 3.570639646590842e-06, "loss": 0.0113, "step": 47509 }, { "epoch": 14.62, "learning_rate": 3.570258019639339e-06, "loss": 0.0094, "step": 47510 }, { "epoch": 14.62, "learning_rate": 3.5698764086511505e-06, "loss": 0.0111, "step": 47511 }, { "epoch": 14.62, "learning_rate": 3.5694948136272266e-06, "loss": 0.0102, "step": 47512 }, { "epoch": 14.62, "learning_rate": 3.569113234568513e-06, "loss": 0.0136, "step": 47513 }, { "epoch": 14.62, "learning_rate": 3.5687316714759567e-06, "loss": 0.0167, "step": 47514 }, { "epoch": 14.62, "learning_rate": 3.568350124350507e-06, "loss": 0.0235, "step": 47515 }, { "epoch": 14.62, "learning_rate": 3.5679685931931064e-06, "loss": 0.0168, "step": 47516 }, { "epoch": 14.62, "learning_rate": 3.567587078004712e-06, "loss": 0.0219, "step": 47517 }, { "epoch": 14.62, "learning_rate": 3.5672055787862613e-06, "loss": 0.0168, "step": 47518 }, { "epoch": 14.62, "learning_rate": 3.5668240955387055e-06, "loss": 0.0091, "step": 47519 }, { "epoch": 14.62, "learning_rate": 3.5664426282629905e-06, "loss": 0.0129, "step": 47520 }, { "epoch": 14.62, "learning_rate": 3.5660611769600604e-06, "loss": 0.0224, "step": 47521 }, { "epoch": 14.62, "learning_rate": 3.5656797416308707e-06, "loss": 0.0222, "step": 47522 }, { "epoch": 14.62, "learning_rate": 3.5652983222763627e-06, "loss": 0.0112, "step": 47523 }, { "epoch": 14.62, "learning_rate": 3.564916918897484e-06, "loss": 0.0145, "step": 47524 }, { "epoch": 14.62, "learning_rate": 3.564535531495186e-06, "loss": 0.0149, "step": 47525 }, { "epoch": 14.62, "learning_rate": 3.564154160070403e-06, "loss": 0.0104, "step": 47526 }, { "epoch": 14.62, "learning_rate": 3.563772804624095e-06, "loss": 0.0125, "step": 47527 }, { "epoch": 14.62, "learning_rate": 3.5633914651572033e-06, "loss": 0.0123, "step": 47528 }, { "epoch": 14.62, "learning_rate": 3.5630101416706755e-06, "loss": 0.0124, "step": 47529 }, { "epoch": 14.62, "learning_rate": 3.562628834165458e-06, "loss": 0.0053, "step": 47530 }, { "epoch": 14.62, "learning_rate": 3.5622475426424985e-06, "loss": 0.0101, "step": 47531 }, { "epoch": 14.63, "learning_rate": 3.5618662671027416e-06, "loss": 0.0149, "step": 47532 }, { "epoch": 14.63, "learning_rate": 3.561485007547135e-06, "loss": 0.0153, "step": 47533 }, { "epoch": 14.63, "learning_rate": 3.5611037639766267e-06, "loss": 0.0123, "step": 47534 }, { "epoch": 14.63, "learning_rate": 3.5607225363921615e-06, "loss": 0.0082, "step": 47535 }, { "epoch": 14.63, "learning_rate": 3.5603413247946817e-06, "loss": 0.0089, "step": 47536 }, { "epoch": 14.63, "learning_rate": 3.5599601291851427e-06, "loss": 0.0115, "step": 47537 }, { "epoch": 14.63, "learning_rate": 3.5595789495644894e-06, "loss": 0.0165, "step": 47538 }, { "epoch": 14.63, "learning_rate": 3.5591977859336625e-06, "loss": 0.0139, "step": 47539 }, { "epoch": 14.63, "learning_rate": 3.5588166382936106e-06, "loss": 0.0157, "step": 47540 }, { "epoch": 14.63, "learning_rate": 3.558435506645278e-06, "loss": 0.0187, "step": 47541 }, { "epoch": 14.63, "learning_rate": 3.558054390989616e-06, "loss": 0.0186, "step": 47542 }, { "epoch": 14.63, "learning_rate": 3.557673291327568e-06, "loss": 0.0206, "step": 47543 }, { "epoch": 14.63, "learning_rate": 3.5572922076600814e-06, "loss": 0.0203, "step": 47544 }, { "epoch": 14.63, "learning_rate": 3.556911139988101e-06, "loss": 0.0117, "step": 47545 }, { "epoch": 14.63, "learning_rate": 3.556530088312573e-06, "loss": 0.0074, "step": 47546 }, { "epoch": 14.63, "learning_rate": 3.556149052634443e-06, "loss": 0.0059, "step": 47547 }, { "epoch": 14.63, "learning_rate": 3.5557680329546586e-06, "loss": 0.0143, "step": 47548 }, { "epoch": 14.63, "learning_rate": 3.5553870292741644e-06, "loss": 0.0129, "step": 47549 }, { "epoch": 14.63, "learning_rate": 3.555006041593907e-06, "loss": 0.01, "step": 47550 }, { "epoch": 14.63, "learning_rate": 3.554625069914832e-06, "loss": 0.0273, "step": 47551 }, { "epoch": 14.63, "learning_rate": 3.5542441142378857e-06, "loss": 0.0123, "step": 47552 }, { "epoch": 14.63, "learning_rate": 3.553863174564014e-06, "loss": 0.0061, "step": 47553 }, { "epoch": 14.63, "learning_rate": 3.553482250894161e-06, "loss": 0.0102, "step": 47554 }, { "epoch": 14.63, "learning_rate": 3.553101343229274e-06, "loss": 0.0092, "step": 47555 }, { "epoch": 14.63, "learning_rate": 3.552720451570295e-06, "loss": 0.0152, "step": 47556 }, { "epoch": 14.63, "learning_rate": 3.552339575918177e-06, "loss": 0.0091, "step": 47557 }, { "epoch": 14.63, "learning_rate": 3.551958716273861e-06, "loss": 0.0111, "step": 47558 }, { "epoch": 14.63, "learning_rate": 3.5515778726382967e-06, "loss": 0.0158, "step": 47559 }, { "epoch": 14.63, "learning_rate": 3.5511970450124223e-06, "loss": 0.0115, "step": 47560 }, { "epoch": 14.63, "learning_rate": 3.550816233397184e-06, "loss": 0.0124, "step": 47561 }, { "epoch": 14.63, "learning_rate": 3.550435437793535e-06, "loss": 0.0165, "step": 47562 }, { "epoch": 14.63, "learning_rate": 3.5500546582024144e-06, "loss": 0.0071, "step": 47563 }, { "epoch": 14.63, "learning_rate": 3.549673894624769e-06, "loss": 0.0208, "step": 47564 }, { "epoch": 14.64, "learning_rate": 3.5492931470615453e-06, "loss": 0.0175, "step": 47565 }, { "epoch": 14.64, "learning_rate": 3.5489124155136876e-06, "loss": 0.014, "step": 47566 }, { "epoch": 14.64, "learning_rate": 3.5485316999821405e-06, "loss": 0.0124, "step": 47567 }, { "epoch": 14.64, "learning_rate": 3.548151000467851e-06, "loss": 0.0164, "step": 47568 }, { "epoch": 14.64, "learning_rate": 3.5477703169717624e-06, "loss": 0.0112, "step": 47569 }, { "epoch": 14.64, "learning_rate": 3.547389649494821e-06, "loss": 0.0164, "step": 47570 }, { "epoch": 14.64, "learning_rate": 3.5470089980379673e-06, "loss": 0.0129, "step": 47571 }, { "epoch": 14.64, "learning_rate": 3.5466283626021594e-06, "loss": 0.0183, "step": 47572 }, { "epoch": 14.64, "learning_rate": 3.546247743188328e-06, "loss": 0.0157, "step": 47573 }, { "epoch": 14.64, "learning_rate": 3.545867139797424e-06, "loss": 0.0217, "step": 47574 }, { "epoch": 14.64, "learning_rate": 3.545486552430393e-06, "loss": 0.0119, "step": 47575 }, { "epoch": 14.64, "learning_rate": 3.545105981088175e-06, "loss": 0.0204, "step": 47576 }, { "epoch": 14.64, "learning_rate": 3.544725425771721e-06, "loss": 0.0164, "step": 47577 }, { "epoch": 14.64, "learning_rate": 3.544344886481974e-06, "loss": 0.0155, "step": 47578 }, { "epoch": 14.64, "learning_rate": 3.5439643632198795e-06, "loss": 0.0134, "step": 47579 }, { "epoch": 14.64, "learning_rate": 3.5435838559863832e-06, "loss": 0.0157, "step": 47580 }, { "epoch": 14.64, "learning_rate": 3.5432033647824206e-06, "loss": 0.0133, "step": 47581 }, { "epoch": 14.64, "learning_rate": 3.542822889608947e-06, "loss": 0.006, "step": 47582 }, { "epoch": 14.64, "learning_rate": 3.5424424304669038e-06, "loss": 0.0085, "step": 47583 }, { "epoch": 14.64, "learning_rate": 3.5420619873572347e-06, "loss": 0.0257, "step": 47584 }, { "epoch": 14.64, "learning_rate": 3.5416815602808854e-06, "loss": 0.0247, "step": 47585 }, { "epoch": 14.64, "learning_rate": 3.541301149238798e-06, "loss": 0.0109, "step": 47586 }, { "epoch": 14.64, "learning_rate": 3.54092075423192e-06, "loss": 0.0111, "step": 47587 }, { "epoch": 14.64, "learning_rate": 3.5405403752611943e-06, "loss": 0.0101, "step": 47588 }, { "epoch": 14.64, "learning_rate": 3.540160012327565e-06, "loss": 0.0153, "step": 47589 }, { "epoch": 14.64, "learning_rate": 3.5397796654319773e-06, "loss": 0.0195, "step": 47590 }, { "epoch": 14.64, "learning_rate": 3.539399334575372e-06, "loss": 0.015, "step": 47591 }, { "epoch": 14.64, "learning_rate": 3.5390190197586992e-06, "loss": 0.0149, "step": 47592 }, { "epoch": 14.64, "learning_rate": 3.5386387209829044e-06, "loss": 0.0142, "step": 47593 }, { "epoch": 14.64, "learning_rate": 3.538258438248924e-06, "loss": 0.0098, "step": 47594 }, { "epoch": 14.64, "learning_rate": 3.537878171557705e-06, "loss": 0.0166, "step": 47595 }, { "epoch": 14.64, "learning_rate": 3.53749792091019e-06, "loss": 0.0199, "step": 47596 }, { "epoch": 14.65, "learning_rate": 3.5371176863073285e-06, "loss": 0.013, "step": 47597 }, { "epoch": 14.65, "learning_rate": 3.536737467750061e-06, "loss": 0.0095, "step": 47598 }, { "epoch": 14.65, "learning_rate": 3.536357265239333e-06, "loss": 0.0212, "step": 47599 }, { "epoch": 14.65, "learning_rate": 3.535977078776086e-06, "loss": 0.0155, "step": 47600 }, { "epoch": 14.65, "learning_rate": 3.535596908361266e-06, "loss": 0.0256, "step": 47601 }, { "epoch": 14.65, "learning_rate": 3.535216753995816e-06, "loss": 0.0138, "step": 47602 }, { "epoch": 14.65, "learning_rate": 3.5348366156806803e-06, "loss": 0.0081, "step": 47603 }, { "epoch": 14.65, "learning_rate": 3.5344564934168023e-06, "loss": 0.0096, "step": 47604 }, { "epoch": 14.65, "learning_rate": 3.5340763872051253e-06, "loss": 0.0113, "step": 47605 }, { "epoch": 14.65, "learning_rate": 3.5336962970465938e-06, "loss": 0.0152, "step": 47606 }, { "epoch": 14.65, "learning_rate": 3.5333162229421512e-06, "loss": 0.017, "step": 47607 }, { "epoch": 14.65, "learning_rate": 3.5329361648927405e-06, "loss": 0.0143, "step": 47608 }, { "epoch": 14.65, "learning_rate": 3.5325561228993065e-06, "loss": 0.0134, "step": 47609 }, { "epoch": 14.65, "learning_rate": 3.532176096962792e-06, "loss": 0.0092, "step": 47610 }, { "epoch": 14.65, "learning_rate": 3.531796087084137e-06, "loss": 0.0122, "step": 47611 }, { "epoch": 14.65, "learning_rate": 3.5314160932642925e-06, "loss": 0.0093, "step": 47612 }, { "epoch": 14.65, "learning_rate": 3.531036115504197e-06, "loss": 0.0657, "step": 47613 }, { "epoch": 14.65, "learning_rate": 3.530656153804799e-06, "loss": 0.0057, "step": 47614 }, { "epoch": 14.65, "learning_rate": 3.530276208167034e-06, "loss": 0.0112, "step": 47615 }, { "epoch": 14.65, "learning_rate": 3.5298962785918458e-06, "loss": 0.005, "step": 47616 }, { "epoch": 14.65, "learning_rate": 3.5295163650801833e-06, "loss": 0.0093, "step": 47617 }, { "epoch": 14.65, "learning_rate": 3.5291364676329886e-06, "loss": 0.0196, "step": 47618 }, { "epoch": 14.65, "learning_rate": 3.528756586251203e-06, "loss": 0.008, "step": 47619 }, { "epoch": 14.65, "learning_rate": 3.52837672093577e-06, "loss": 0.0202, "step": 47620 }, { "epoch": 14.65, "learning_rate": 3.5279968716876333e-06, "loss": 0.0094, "step": 47621 }, { "epoch": 14.65, "learning_rate": 3.5276170385077345e-06, "loss": 0.0086, "step": 47622 }, { "epoch": 14.65, "learning_rate": 3.5272372213970184e-06, "loss": 0.0167, "step": 47623 }, { "epoch": 14.65, "learning_rate": 3.526857420356428e-06, "loss": 0.0124, "step": 47624 }, { "epoch": 14.65, "learning_rate": 3.5264776353869046e-06, "loss": 0.0175, "step": 47625 }, { "epoch": 14.65, "learning_rate": 3.5260978664893895e-06, "loss": 0.0174, "step": 47626 }, { "epoch": 14.65, "learning_rate": 3.5257181136648344e-06, "loss": 0.0144, "step": 47627 }, { "epoch": 14.65, "learning_rate": 3.5253383769141724e-06, "loss": 0.0091, "step": 47628 }, { "epoch": 14.65, "learning_rate": 3.524958656238349e-06, "loss": 0.0177, "step": 47629 }, { "epoch": 14.66, "learning_rate": 3.524578951638309e-06, "loss": 0.0198, "step": 47630 }, { "epoch": 14.66, "learning_rate": 3.524199263114989e-06, "loss": 0.0123, "step": 47631 }, { "epoch": 14.66, "learning_rate": 3.523819590669341e-06, "loss": 0.0162, "step": 47632 }, { "epoch": 14.66, "learning_rate": 3.5234399343023016e-06, "loss": 0.0157, "step": 47633 }, { "epoch": 14.66, "learning_rate": 3.5230602940148163e-06, "loss": 0.0184, "step": 47634 }, { "epoch": 14.66, "learning_rate": 3.5226806698078287e-06, "loss": 0.0236, "step": 47635 }, { "epoch": 14.66, "learning_rate": 3.5223010616822716e-06, "loss": 0.0078, "step": 47636 }, { "epoch": 14.66, "learning_rate": 3.521921469639099e-06, "loss": 0.0143, "step": 47637 }, { "epoch": 14.66, "learning_rate": 3.5215418936792487e-06, "loss": 0.0073, "step": 47638 }, { "epoch": 14.66, "learning_rate": 3.521162333803664e-06, "loss": 0.0071, "step": 47639 }, { "epoch": 14.66, "learning_rate": 3.520782790013285e-06, "loss": 0.0125, "step": 47640 }, { "epoch": 14.66, "learning_rate": 3.520403262309057e-06, "loss": 0.0118, "step": 47641 }, { "epoch": 14.66, "learning_rate": 3.5200237506919198e-06, "loss": 0.018, "step": 47642 }, { "epoch": 14.66, "learning_rate": 3.5196442551628174e-06, "loss": 0.013, "step": 47643 }, { "epoch": 14.66, "learning_rate": 3.5192647757226903e-06, "loss": 0.0062, "step": 47644 }, { "epoch": 14.66, "learning_rate": 3.518885312372483e-06, "loss": 0.0179, "step": 47645 }, { "epoch": 14.66, "learning_rate": 3.518505865113132e-06, "loss": 0.0135, "step": 47646 }, { "epoch": 14.66, "learning_rate": 3.518126433945587e-06, "loss": 0.0131, "step": 47647 }, { "epoch": 14.66, "learning_rate": 3.517747018870791e-06, "loss": 0.0143, "step": 47648 }, { "epoch": 14.66, "learning_rate": 3.5173676198896768e-06, "loss": 0.013, "step": 47649 }, { "epoch": 14.66, "learning_rate": 3.5169882370031916e-06, "loss": 0.017, "step": 47650 }, { "epoch": 14.66, "learning_rate": 3.5166088702122738e-06, "loss": 0.0261, "step": 47651 }, { "epoch": 14.66, "learning_rate": 3.516229519517871e-06, "loss": 0.0064, "step": 47652 }, { "epoch": 14.66, "learning_rate": 3.5158501849209224e-06, "loss": 0.0112, "step": 47653 }, { "epoch": 14.66, "learning_rate": 3.51547086642237e-06, "loss": 0.0091, "step": 47654 }, { "epoch": 14.66, "learning_rate": 3.5150915640231553e-06, "loss": 0.0174, "step": 47655 }, { "epoch": 14.66, "learning_rate": 3.5147122777242203e-06, "loss": 0.0375, "step": 47656 }, { "epoch": 14.66, "learning_rate": 3.514333007526506e-06, "loss": 0.0133, "step": 47657 }, { "epoch": 14.66, "learning_rate": 3.5139537534309543e-06, "loss": 0.0069, "step": 47658 }, { "epoch": 14.66, "learning_rate": 3.513574515438507e-06, "loss": 0.0124, "step": 47659 }, { "epoch": 14.66, "learning_rate": 3.513195293550106e-06, "loss": 0.0088, "step": 47660 }, { "epoch": 14.66, "learning_rate": 3.5128160877666873e-06, "loss": 0.0094, "step": 47661 }, { "epoch": 14.67, "learning_rate": 3.5124368980892055e-06, "loss": 0.0288, "step": 47662 }, { "epoch": 14.67, "learning_rate": 3.5120577245185906e-06, "loss": 0.0066, "step": 47663 }, { "epoch": 14.67, "learning_rate": 3.511678567055786e-06, "loss": 0.0251, "step": 47664 }, { "epoch": 14.67, "learning_rate": 3.511299425701735e-06, "loss": 0.0208, "step": 47665 }, { "epoch": 14.67, "learning_rate": 3.5109203004573754e-06, "loss": 0.0087, "step": 47666 }, { "epoch": 14.67, "learning_rate": 3.5105411913236544e-06, "loss": 0.012, "step": 47667 }, { "epoch": 14.67, "learning_rate": 3.5101620983015097e-06, "loss": 0.011, "step": 47668 }, { "epoch": 14.67, "learning_rate": 3.509783021391887e-06, "loss": 0.015, "step": 47669 }, { "epoch": 14.67, "learning_rate": 3.5094039605957197e-06, "loss": 0.0131, "step": 47670 }, { "epoch": 14.67, "learning_rate": 3.509024915913949e-06, "loss": 0.0181, "step": 47671 }, { "epoch": 14.67, "learning_rate": 3.5086458873475227e-06, "loss": 0.0093, "step": 47672 }, { "epoch": 14.67, "learning_rate": 3.508266874897379e-06, "loss": 0.0163, "step": 47673 }, { "epoch": 14.67, "learning_rate": 3.507887878564458e-06, "loss": 0.0142, "step": 47674 }, { "epoch": 14.67, "learning_rate": 3.507508898349702e-06, "loss": 0.0224, "step": 47675 }, { "epoch": 14.67, "learning_rate": 3.50712993425405e-06, "loss": 0.012, "step": 47676 }, { "epoch": 14.67, "learning_rate": 3.5067509862784455e-06, "loss": 0.007, "step": 47677 }, { "epoch": 14.67, "learning_rate": 3.5063720544238266e-06, "loss": 0.0172, "step": 47678 }, { "epoch": 14.67, "learning_rate": 3.505993138691135e-06, "loss": 0.0084, "step": 47679 }, { "epoch": 14.67, "learning_rate": 3.505614239081313e-06, "loss": 0.0094, "step": 47680 }, { "epoch": 14.67, "learning_rate": 3.5052353555952955e-06, "loss": 0.0126, "step": 47681 }, { "epoch": 14.67, "learning_rate": 3.5048564882340308e-06, "loss": 0.0195, "step": 47682 }, { "epoch": 14.67, "learning_rate": 3.504477636998461e-06, "loss": 0.0123, "step": 47683 }, { "epoch": 14.67, "learning_rate": 3.5040988018895182e-06, "loss": 0.0125, "step": 47684 }, { "epoch": 14.67, "learning_rate": 3.503719982908147e-06, "loss": 0.0186, "step": 47685 }, { "epoch": 14.67, "learning_rate": 3.503341180055284e-06, "loss": 0.019, "step": 47686 }, { "epoch": 14.67, "learning_rate": 3.502962393331877e-06, "loss": 0.0136, "step": 47687 }, { "epoch": 14.67, "learning_rate": 3.502583622738862e-06, "loss": 0.0142, "step": 47688 }, { "epoch": 14.67, "learning_rate": 3.5022048682771814e-06, "loss": 0.0098, "step": 47689 }, { "epoch": 14.67, "learning_rate": 3.5018261299477772e-06, "loss": 0.0155, "step": 47690 }, { "epoch": 14.67, "learning_rate": 3.50144740775158e-06, "loss": 0.0113, "step": 47691 }, { "epoch": 14.67, "learning_rate": 3.5010687016895404e-06, "loss": 0.0214, "step": 47692 }, { "epoch": 14.67, "learning_rate": 3.5006900117625953e-06, "loss": 0.0097, "step": 47693 }, { "epoch": 14.67, "learning_rate": 3.5003113379716846e-06, "loss": 0.0164, "step": 47694 }, { "epoch": 14.68, "learning_rate": 3.4999326803177493e-06, "loss": 0.0113, "step": 47695 }, { "epoch": 14.68, "learning_rate": 3.499554038801728e-06, "loss": 0.0208, "step": 47696 }, { "epoch": 14.68, "learning_rate": 3.4991754134245617e-06, "loss": 0.0108, "step": 47697 }, { "epoch": 14.68, "learning_rate": 3.49879680418719e-06, "loss": 0.0099, "step": 47698 }, { "epoch": 14.68, "learning_rate": 3.498418211090554e-06, "loss": 0.0056, "step": 47699 }, { "epoch": 14.68, "learning_rate": 3.498039634135593e-06, "loss": 0.0131, "step": 47700 }, { "epoch": 14.68, "learning_rate": 3.497661073323242e-06, "loss": 0.0116, "step": 47701 }, { "epoch": 14.68, "learning_rate": 3.49728252865445e-06, "loss": 0.0128, "step": 47702 }, { "epoch": 14.68, "learning_rate": 3.4969040001301513e-06, "loss": 0.0194, "step": 47703 }, { "epoch": 14.68, "learning_rate": 3.4965254877512912e-06, "loss": 0.0133, "step": 47704 }, { "epoch": 14.68, "learning_rate": 3.4961469915188007e-06, "loss": 0.011, "step": 47705 }, { "epoch": 14.68, "learning_rate": 3.4957685114336214e-06, "loss": 0.0068, "step": 47706 }, { "epoch": 14.68, "learning_rate": 3.4953900474966985e-06, "loss": 0.0088, "step": 47707 }, { "epoch": 14.68, "learning_rate": 3.4950115997089685e-06, "loss": 0.0095, "step": 47708 }, { "epoch": 14.68, "learning_rate": 3.4946331680713706e-06, "loss": 0.0232, "step": 47709 }, { "epoch": 14.68, "learning_rate": 3.4942547525848457e-06, "loss": 0.0112, "step": 47710 }, { "epoch": 14.68, "learning_rate": 3.493876353250332e-06, "loss": 0.012, "step": 47711 }, { "epoch": 14.68, "learning_rate": 3.4934979700687686e-06, "loss": 0.0192, "step": 47712 }, { "epoch": 14.68, "learning_rate": 3.4931196030410965e-06, "loss": 0.0091, "step": 47713 }, { "epoch": 14.68, "learning_rate": 3.4927412521682535e-06, "loss": 0.0092, "step": 47714 }, { "epoch": 14.68, "learning_rate": 3.4923629174511796e-06, "loss": 0.0118, "step": 47715 }, { "epoch": 14.68, "learning_rate": 3.491984598890812e-06, "loss": 0.0077, "step": 47716 }, { "epoch": 14.68, "learning_rate": 3.491606296488098e-06, "loss": 0.0126, "step": 47717 }, { "epoch": 14.68, "learning_rate": 3.491228010243968e-06, "loss": 0.0181, "step": 47718 }, { "epoch": 14.68, "learning_rate": 3.4908497401593634e-06, "loss": 0.0068, "step": 47719 }, { "epoch": 14.68, "learning_rate": 3.490471486235224e-06, "loss": 0.0176, "step": 47720 }, { "epoch": 14.68, "learning_rate": 3.490093248472487e-06, "loss": 0.0177, "step": 47721 }, { "epoch": 14.68, "learning_rate": 3.4897150268720957e-06, "loss": 0.0161, "step": 47722 }, { "epoch": 14.68, "learning_rate": 3.4893368214349864e-06, "loss": 0.0114, "step": 47723 }, { "epoch": 14.68, "learning_rate": 3.488958632162103e-06, "loss": 0.0113, "step": 47724 }, { "epoch": 14.68, "learning_rate": 3.488580459054376e-06, "loss": 0.033, "step": 47725 }, { "epoch": 14.68, "learning_rate": 3.488202302112745e-06, "loss": 0.0077, "step": 47726 }, { "epoch": 14.69, "learning_rate": 3.487824161338157e-06, "loss": 0.0146, "step": 47727 }, { "epoch": 14.69, "learning_rate": 3.4874460367315442e-06, "loss": 0.0121, "step": 47728 }, { "epoch": 14.69, "learning_rate": 3.487067928293848e-06, "loss": 0.0237, "step": 47729 }, { "epoch": 14.69, "learning_rate": 3.486689836026006e-06, "loss": 0.0195, "step": 47730 }, { "epoch": 14.69, "learning_rate": 3.4863117599289562e-06, "loss": 0.0331, "step": 47731 }, { "epoch": 14.69, "learning_rate": 3.48593370000364e-06, "loss": 0.013, "step": 47732 }, { "epoch": 14.69, "learning_rate": 3.485555656250993e-06, "loss": 0.0133, "step": 47733 }, { "epoch": 14.69, "learning_rate": 3.485177628671956e-06, "loss": 0.0081, "step": 47734 }, { "epoch": 14.69, "learning_rate": 3.4847996172674658e-06, "loss": 0.0103, "step": 47735 }, { "epoch": 14.69, "learning_rate": 3.4844216220384586e-06, "loss": 0.0067, "step": 47736 }, { "epoch": 14.69, "learning_rate": 3.484043642985879e-06, "loss": 0.0116, "step": 47737 }, { "epoch": 14.69, "learning_rate": 3.483665680110667e-06, "loss": 0.017, "step": 47738 }, { "epoch": 14.69, "learning_rate": 3.4832877334137528e-06, "loss": 0.0194, "step": 47739 }, { "epoch": 14.69, "learning_rate": 3.4829098028960773e-06, "loss": 0.0111, "step": 47740 }, { "epoch": 14.69, "learning_rate": 3.482531888558577e-06, "loss": 0.0101, "step": 47741 }, { "epoch": 14.69, "learning_rate": 3.4821539904021963e-06, "loss": 0.006, "step": 47742 }, { "epoch": 14.69, "learning_rate": 3.481776108427869e-06, "loss": 0.0157, "step": 47743 }, { "epoch": 14.69, "learning_rate": 3.4813982426365355e-06, "loss": 0.0136, "step": 47744 }, { "epoch": 14.69, "learning_rate": 3.481020393029136e-06, "loss": 0.0139, "step": 47745 }, { "epoch": 14.69, "learning_rate": 3.4806425596065985e-06, "loss": 0.0053, "step": 47746 }, { "epoch": 14.69, "learning_rate": 3.480264742369872e-06, "loss": 0.0077, "step": 47747 }, { "epoch": 14.69, "learning_rate": 3.47988694131989e-06, "loss": 0.0138, "step": 47748 }, { "epoch": 14.69, "learning_rate": 3.479509156457591e-06, "loss": 0.0114, "step": 47749 }, { "epoch": 14.69, "learning_rate": 3.479131387783913e-06, "loss": 0.0206, "step": 47750 }, { "epoch": 14.69, "learning_rate": 3.478753635299793e-06, "loss": 0.0214, "step": 47751 }, { "epoch": 14.69, "learning_rate": 3.478375899006171e-06, "loss": 0.0151, "step": 47752 }, { "epoch": 14.69, "learning_rate": 3.477998178903982e-06, "loss": 0.0157, "step": 47753 }, { "epoch": 14.69, "learning_rate": 3.4776204749941667e-06, "loss": 0.0176, "step": 47754 }, { "epoch": 14.69, "learning_rate": 3.4772427872776606e-06, "loss": 0.0119, "step": 47755 }, { "epoch": 14.69, "learning_rate": 3.476865115755399e-06, "loss": 0.0211, "step": 47756 }, { "epoch": 14.69, "learning_rate": 3.4764874604283263e-06, "loss": 0.0149, "step": 47757 }, { "epoch": 14.69, "learning_rate": 3.4761098212973755e-06, "loss": 0.0134, "step": 47758 }, { "epoch": 14.69, "learning_rate": 3.4757321983634895e-06, "loss": 0.0111, "step": 47759 }, { "epoch": 14.7, "learning_rate": 3.4753545916275976e-06, "loss": 0.0102, "step": 47760 }, { "epoch": 14.7, "learning_rate": 3.474977001090638e-06, "loss": 0.0067, "step": 47761 }, { "epoch": 14.7, "learning_rate": 3.4745994267535556e-06, "loss": 0.0121, "step": 47762 }, { "epoch": 14.7, "learning_rate": 3.4742218686172836e-06, "loss": 0.0163, "step": 47763 }, { "epoch": 14.7, "learning_rate": 3.4738443266827583e-06, "loss": 0.0113, "step": 47764 }, { "epoch": 14.7, "learning_rate": 3.4734668009509187e-06, "loss": 0.0155, "step": 47765 }, { "epoch": 14.7, "learning_rate": 3.473089291422702e-06, "loss": 0.0137, "step": 47766 }, { "epoch": 14.7, "learning_rate": 3.4727117980990444e-06, "loss": 0.011, "step": 47767 }, { "epoch": 14.7, "learning_rate": 3.4723343209808834e-06, "loss": 0.0107, "step": 47768 }, { "epoch": 14.7, "learning_rate": 3.471956860069158e-06, "loss": 0.0147, "step": 47769 }, { "epoch": 14.7, "learning_rate": 3.4715794153648017e-06, "loss": 0.0126, "step": 47770 }, { "epoch": 14.7, "learning_rate": 3.471201986868752e-06, "loss": 0.0068, "step": 47771 }, { "epoch": 14.7, "learning_rate": 3.4708245745819534e-06, "loss": 0.0265, "step": 47772 }, { "epoch": 14.7, "learning_rate": 3.4704471785053353e-06, "loss": 0.0115, "step": 47773 }, { "epoch": 14.7, "learning_rate": 3.4700697986398347e-06, "loss": 0.0193, "step": 47774 }, { "epoch": 14.7, "learning_rate": 3.4696924349863913e-06, "loss": 0.0088, "step": 47775 }, { "epoch": 14.7, "learning_rate": 3.469315087545937e-06, "loss": 0.0122, "step": 47776 }, { "epoch": 14.7, "learning_rate": 3.468937756319417e-06, "loss": 0.0062, "step": 47777 }, { "epoch": 14.7, "learning_rate": 3.468560441307763e-06, "loss": 0.0097, "step": 47778 }, { "epoch": 14.7, "learning_rate": 3.468183142511914e-06, "loss": 0.0137, "step": 47779 }, { "epoch": 14.7, "learning_rate": 3.4678058599328067e-06, "loss": 0.0103, "step": 47780 }, { "epoch": 14.7, "learning_rate": 3.4674285935713715e-06, "loss": 0.0088, "step": 47781 }, { "epoch": 14.7, "learning_rate": 3.4670513434285525e-06, "loss": 0.0071, "step": 47782 }, { "epoch": 14.7, "learning_rate": 3.4666741095052837e-06, "loss": 0.0122, "step": 47783 }, { "epoch": 14.7, "learning_rate": 3.466296891802502e-06, "loss": 0.0297, "step": 47784 }, { "epoch": 14.7, "learning_rate": 3.4659196903211434e-06, "loss": 0.0134, "step": 47785 }, { "epoch": 14.7, "learning_rate": 3.465542505062145e-06, "loss": 0.0164, "step": 47786 }, { "epoch": 14.7, "learning_rate": 3.4651653360264426e-06, "loss": 0.0232, "step": 47787 }, { "epoch": 14.7, "learning_rate": 3.4647881832149734e-06, "loss": 0.0253, "step": 47788 }, { "epoch": 14.7, "learning_rate": 3.464411046628673e-06, "loss": 0.0116, "step": 47789 }, { "epoch": 14.7, "learning_rate": 3.4640339262684788e-06, "loss": 0.0155, "step": 47790 }, { "epoch": 14.7, "learning_rate": 3.463656822135322e-06, "loss": 0.01, "step": 47791 }, { "epoch": 14.71, "learning_rate": 3.463279734230146e-06, "loss": 0.0123, "step": 47792 }, { "epoch": 14.71, "learning_rate": 3.4629026625538884e-06, "loss": 0.0058, "step": 47793 }, { "epoch": 14.71, "learning_rate": 3.4625256071074776e-06, "loss": 0.0078, "step": 47794 }, { "epoch": 14.71, "learning_rate": 3.462148567891854e-06, "loss": 0.0139, "step": 47795 }, { "epoch": 14.71, "learning_rate": 3.461771544907949e-06, "loss": 0.0105, "step": 47796 }, { "epoch": 14.71, "learning_rate": 3.461394538156706e-06, "loss": 0.0211, "step": 47797 }, { "epoch": 14.71, "learning_rate": 3.4610175476390573e-06, "loss": 0.0085, "step": 47798 }, { "epoch": 14.71, "learning_rate": 3.460640573355939e-06, "loss": 0.0146, "step": 47799 }, { "epoch": 14.71, "learning_rate": 3.460263615308288e-06, "loss": 0.0355, "step": 47800 }, { "epoch": 14.71, "learning_rate": 3.459886673497038e-06, "loss": 0.0136, "step": 47801 }, { "epoch": 14.71, "learning_rate": 3.459509747923128e-06, "loss": 0.011, "step": 47802 }, { "epoch": 14.71, "learning_rate": 3.459132838587491e-06, "loss": 0.0143, "step": 47803 }, { "epoch": 14.71, "learning_rate": 3.458755945491065e-06, "loss": 0.01, "step": 47804 }, { "epoch": 14.71, "learning_rate": 3.458379068634783e-06, "loss": 0.0153, "step": 47805 }, { "epoch": 14.71, "learning_rate": 3.458002208019583e-06, "loss": 0.0107, "step": 47806 }, { "epoch": 14.71, "learning_rate": 3.4576253636463996e-06, "loss": 0.0146, "step": 47807 }, { "epoch": 14.71, "learning_rate": 3.4572485355161676e-06, "loss": 0.0124, "step": 47808 }, { "epoch": 14.71, "learning_rate": 3.4568717236298257e-06, "loss": 0.0181, "step": 47809 }, { "epoch": 14.71, "learning_rate": 3.4564949279883064e-06, "loss": 0.0134, "step": 47810 }, { "epoch": 14.71, "learning_rate": 3.4561181485925423e-06, "loss": 0.0092, "step": 47811 }, { "epoch": 14.71, "learning_rate": 3.4557413854434763e-06, "loss": 0.0195, "step": 47812 }, { "epoch": 14.71, "learning_rate": 3.4553646385420403e-06, "loss": 0.0108, "step": 47813 }, { "epoch": 14.71, "learning_rate": 3.4549879078891722e-06, "loss": 0.0214, "step": 47814 }, { "epoch": 14.71, "learning_rate": 3.454611193485803e-06, "loss": 0.0185, "step": 47815 }, { "epoch": 14.71, "learning_rate": 3.4542344953328642e-06, "loss": 0.0076, "step": 47816 }, { "epoch": 14.71, "learning_rate": 3.453857813431302e-06, "loss": 0.0111, "step": 47817 }, { "epoch": 14.71, "learning_rate": 3.453481147782044e-06, "loss": 0.028, "step": 47818 }, { "epoch": 14.71, "learning_rate": 3.4531044983860285e-06, "loss": 0.0123, "step": 47819 }, { "epoch": 14.71, "learning_rate": 3.4527278652441896e-06, "loss": 0.0103, "step": 47820 }, { "epoch": 14.71, "learning_rate": 3.452351248357462e-06, "loss": 0.0159, "step": 47821 }, { "epoch": 14.71, "learning_rate": 3.451974647726781e-06, "loss": 0.0158, "step": 47822 }, { "epoch": 14.71, "learning_rate": 3.4515980633530823e-06, "loss": 0.0105, "step": 47823 }, { "epoch": 14.71, "learning_rate": 3.451221495237299e-06, "loss": 0.0157, "step": 47824 }, { "epoch": 14.72, "learning_rate": 3.4508449433803683e-06, "loss": 0.0107, "step": 47825 }, { "epoch": 14.72, "learning_rate": 3.4504684077832197e-06, "loss": 0.0102, "step": 47826 }, { "epoch": 14.72, "learning_rate": 3.4500918884468006e-06, "loss": 0.0084, "step": 47827 }, { "epoch": 14.72, "learning_rate": 3.449715385372032e-06, "loss": 0.0133, "step": 47828 }, { "epoch": 14.72, "learning_rate": 3.449338898559856e-06, "loss": 0.0105, "step": 47829 }, { "epoch": 14.72, "learning_rate": 3.4489624280112046e-06, "loss": 0.0067, "step": 47830 }, { "epoch": 14.72, "learning_rate": 3.448585973727011e-06, "loss": 0.0167, "step": 47831 }, { "epoch": 14.72, "learning_rate": 3.4482095357082147e-06, "loss": 0.0124, "step": 47832 }, { "epoch": 14.72, "learning_rate": 3.4478331139557475e-06, "loss": 0.0087, "step": 47833 }, { "epoch": 14.72, "learning_rate": 3.4474567084705455e-06, "loss": 0.0306, "step": 47834 }, { "epoch": 14.72, "learning_rate": 3.447080319253544e-06, "loss": 0.074, "step": 47835 }, { "epoch": 14.72, "learning_rate": 3.4467039463056697e-06, "loss": 0.0132, "step": 47836 }, { "epoch": 14.72, "learning_rate": 3.4463275896278657e-06, "loss": 0.0175, "step": 47837 }, { "epoch": 14.72, "learning_rate": 3.4459512492210635e-06, "loss": 0.0077, "step": 47838 }, { "epoch": 14.72, "learning_rate": 3.445574925086198e-06, "loss": 0.0112, "step": 47839 }, { "epoch": 14.72, "learning_rate": 3.4451986172242024e-06, "loss": 0.0089, "step": 47840 }, { "epoch": 14.72, "learning_rate": 3.4448223256360112e-06, "loss": 0.0105, "step": 47841 }, { "epoch": 14.72, "learning_rate": 3.4444460503225594e-06, "loss": 0.0249, "step": 47842 }, { "epoch": 14.72, "learning_rate": 3.444069791284781e-06, "loss": 0.0071, "step": 47843 }, { "epoch": 14.72, "learning_rate": 3.44369354852361e-06, "loss": 0.0132, "step": 47844 }, { "epoch": 14.72, "learning_rate": 3.4433173220399796e-06, "loss": 0.0087, "step": 47845 }, { "epoch": 14.72, "learning_rate": 3.442941111834822e-06, "loss": 0.0165, "step": 47846 }, { "epoch": 14.72, "learning_rate": 3.442564917909077e-06, "loss": 0.0145, "step": 47847 }, { "epoch": 14.72, "learning_rate": 3.4421887402636776e-06, "loss": 0.0062, "step": 47848 }, { "epoch": 14.72, "learning_rate": 3.441812578899554e-06, "loss": 0.0092, "step": 47849 }, { "epoch": 14.72, "learning_rate": 3.441436433817641e-06, "loss": 0.0161, "step": 47850 }, { "epoch": 14.72, "learning_rate": 3.4410603050188705e-06, "loss": 0.0144, "step": 47851 }, { "epoch": 14.72, "learning_rate": 3.4406841925041813e-06, "loss": 0.0182, "step": 47852 }, { "epoch": 14.72, "learning_rate": 3.440308096274506e-06, "loss": 0.01, "step": 47853 }, { "epoch": 14.72, "learning_rate": 3.439932016330777e-06, "loss": 0.0079, "step": 47854 }, { "epoch": 14.72, "learning_rate": 3.4395559526739273e-06, "loss": 0.0211, "step": 47855 }, { "epoch": 14.72, "learning_rate": 3.4391799053048924e-06, "loss": 0.0113, "step": 47856 }, { "epoch": 14.73, "learning_rate": 3.4388038742246055e-06, "loss": 0.0084, "step": 47857 }, { "epoch": 14.73, "learning_rate": 3.438427859433998e-06, "loss": 0.0057, "step": 47858 }, { "epoch": 14.73, "learning_rate": 3.4380518609340076e-06, "loss": 0.0085, "step": 47859 }, { "epoch": 14.73, "learning_rate": 3.437675878725564e-06, "loss": 0.0124, "step": 47860 }, { "epoch": 14.73, "learning_rate": 3.437299912809602e-06, "loss": 0.01, "step": 47861 }, { "epoch": 14.73, "learning_rate": 3.4369239631870544e-06, "loss": 0.0116, "step": 47862 }, { "epoch": 14.73, "learning_rate": 3.436548029858856e-06, "loss": 0.0115, "step": 47863 }, { "epoch": 14.73, "learning_rate": 3.4361721128259396e-06, "loss": 0.0116, "step": 47864 }, { "epoch": 14.73, "learning_rate": 3.4357962120892383e-06, "loss": 0.014, "step": 47865 }, { "epoch": 14.73, "learning_rate": 3.4354203276496822e-06, "loss": 0.0092, "step": 47866 }, { "epoch": 14.73, "learning_rate": 3.4350444595082098e-06, "loss": 0.0158, "step": 47867 }, { "epoch": 14.73, "learning_rate": 3.434668607665753e-06, "loss": 0.0105, "step": 47868 }, { "epoch": 14.73, "learning_rate": 3.434292772123248e-06, "loss": 0.0201, "step": 47869 }, { "epoch": 14.73, "learning_rate": 3.4339169528816206e-06, "loss": 0.0146, "step": 47870 }, { "epoch": 14.73, "learning_rate": 3.4335411499418027e-06, "loss": 0.0126, "step": 47871 }, { "epoch": 14.73, "learning_rate": 3.4331653633047356e-06, "loss": 0.0098, "step": 47872 }, { "epoch": 14.73, "learning_rate": 3.4327895929713494e-06, "loss": 0.0182, "step": 47873 }, { "epoch": 14.73, "learning_rate": 3.4324138389425753e-06, "loss": 0.0187, "step": 47874 }, { "epoch": 14.73, "learning_rate": 3.432038101219348e-06, "loss": 0.0181, "step": 47875 }, { "epoch": 14.73, "learning_rate": 3.431662379802598e-06, "loss": 0.01, "step": 47876 }, { "epoch": 14.73, "learning_rate": 3.4312866746932603e-06, "loss": 0.0231, "step": 47877 }, { "epoch": 14.73, "learning_rate": 3.4309109858922675e-06, "loss": 0.0147, "step": 47878 }, { "epoch": 14.73, "learning_rate": 3.4305353134005514e-06, "loss": 0.0104, "step": 47879 }, { "epoch": 14.73, "learning_rate": 3.430159657219041e-06, "loss": 0.0245, "step": 47880 }, { "epoch": 14.73, "learning_rate": 3.4297840173486774e-06, "loss": 0.0164, "step": 47881 }, { "epoch": 14.73, "learning_rate": 3.429408393790391e-06, "loss": 0.0116, "step": 47882 }, { "epoch": 14.73, "learning_rate": 3.42903278654511e-06, "loss": 0.0212, "step": 47883 }, { "epoch": 14.73, "learning_rate": 3.4286571956137683e-06, "loss": 0.013, "step": 47884 }, { "epoch": 14.73, "learning_rate": 3.428281620997296e-06, "loss": 0.0153, "step": 47885 }, { "epoch": 14.73, "learning_rate": 3.427906062696632e-06, "loss": 0.0123, "step": 47886 }, { "epoch": 14.73, "learning_rate": 3.427530520712706e-06, "loss": 0.0051, "step": 47887 }, { "epoch": 14.73, "learning_rate": 3.42715499504645e-06, "loss": 0.0277, "step": 47888 }, { "epoch": 14.73, "learning_rate": 3.426779485698796e-06, "loss": 0.013, "step": 47889 }, { "epoch": 14.74, "learning_rate": 3.4264039926706748e-06, "loss": 0.0088, "step": 47890 }, { "epoch": 14.74, "learning_rate": 3.4260285159630224e-06, "loss": 0.027, "step": 47891 }, { "epoch": 14.74, "learning_rate": 3.4256530555767673e-06, "loss": 0.0179, "step": 47892 }, { "epoch": 14.74, "learning_rate": 3.425277611512845e-06, "loss": 0.0113, "step": 47893 }, { "epoch": 14.74, "learning_rate": 3.4249021837721853e-06, "loss": 0.022, "step": 47894 }, { "epoch": 14.74, "learning_rate": 3.4245267723557163e-06, "loss": 0.0171, "step": 47895 }, { "epoch": 14.74, "learning_rate": 3.424151377264383e-06, "loss": 0.0117, "step": 47896 }, { "epoch": 14.74, "learning_rate": 3.423775998499105e-06, "loss": 0.0093, "step": 47897 }, { "epoch": 14.74, "learning_rate": 3.4234006360608173e-06, "loss": 0.0112, "step": 47898 }, { "epoch": 14.74, "learning_rate": 3.4230252899504546e-06, "loss": 0.0052, "step": 47899 }, { "epoch": 14.74, "learning_rate": 3.4226499601689424e-06, "loss": 0.0098, "step": 47900 }, { "epoch": 14.74, "learning_rate": 3.422274646717221e-06, "loss": 0.0083, "step": 47901 }, { "epoch": 14.74, "learning_rate": 3.4218993495962183e-06, "loss": 0.0102, "step": 47902 }, { "epoch": 14.74, "learning_rate": 3.4215240688068694e-06, "loss": 0.0073, "step": 47903 }, { "epoch": 14.74, "learning_rate": 3.4211488043500996e-06, "loss": 0.0156, "step": 47904 }, { "epoch": 14.74, "learning_rate": 3.4207735562268406e-06, "loss": 0.0259, "step": 47905 }, { "epoch": 14.74, "learning_rate": 3.42039832443803e-06, "loss": 0.0136, "step": 47906 }, { "epoch": 14.74, "learning_rate": 3.420023108984597e-06, "loss": 0.0147, "step": 47907 }, { "epoch": 14.74, "learning_rate": 3.4196479098674717e-06, "loss": 0.0142, "step": 47908 }, { "epoch": 14.74, "learning_rate": 3.419272727087587e-06, "loss": 0.0134, "step": 47909 }, { "epoch": 14.74, "learning_rate": 3.4188975606458754e-06, "loss": 0.0128, "step": 47910 }, { "epoch": 14.74, "learning_rate": 3.418522410543266e-06, "loss": 0.0052, "step": 47911 }, { "epoch": 14.74, "learning_rate": 3.41814727678069e-06, "loss": 0.0201, "step": 47912 }, { "epoch": 14.74, "learning_rate": 3.4177721593590806e-06, "loss": 0.0172, "step": 47913 }, { "epoch": 14.74, "learning_rate": 3.417397058279369e-06, "loss": 0.0152, "step": 47914 }, { "epoch": 14.74, "learning_rate": 3.4170219735424827e-06, "loss": 0.0175, "step": 47915 }, { "epoch": 14.74, "learning_rate": 3.416646905149359e-06, "loss": 0.0149, "step": 47916 }, { "epoch": 14.74, "learning_rate": 3.41627185310093e-06, "loss": 0.0118, "step": 47917 }, { "epoch": 14.74, "learning_rate": 3.415896817398119e-06, "loss": 0.0136, "step": 47918 }, { "epoch": 14.74, "learning_rate": 3.415521798041861e-06, "loss": 0.0119, "step": 47919 }, { "epoch": 14.74, "learning_rate": 3.4151467950330853e-06, "loss": 0.0116, "step": 47920 }, { "epoch": 14.74, "learning_rate": 3.4147718083727264e-06, "loss": 0.0178, "step": 47921 }, { "epoch": 14.75, "learning_rate": 3.4143968380617153e-06, "loss": 0.0077, "step": 47922 }, { "epoch": 14.75, "learning_rate": 3.4140218841009808e-06, "loss": 0.0147, "step": 47923 }, { "epoch": 14.75, "learning_rate": 3.413646946491458e-06, "loss": 0.009, "step": 47924 }, { "epoch": 14.75, "learning_rate": 3.4132720252340666e-06, "loss": 0.0112, "step": 47925 }, { "epoch": 14.75, "learning_rate": 3.41289712032975e-06, "loss": 0.0088, "step": 47926 }, { "epoch": 14.75, "learning_rate": 3.4125222317794327e-06, "loss": 0.0116, "step": 47927 }, { "epoch": 14.75, "learning_rate": 3.4121473595840473e-06, "loss": 0.0143, "step": 47928 }, { "epoch": 14.75, "learning_rate": 3.4117725037445237e-06, "loss": 0.0241, "step": 47929 }, { "epoch": 14.75, "learning_rate": 3.411397664261793e-06, "loss": 0.0119, "step": 47930 }, { "epoch": 14.75, "learning_rate": 3.4110228411367863e-06, "loss": 0.0107, "step": 47931 }, { "epoch": 14.75, "learning_rate": 3.410648034370433e-06, "loss": 0.0256, "step": 47932 }, { "epoch": 14.75, "learning_rate": 3.4102732439636645e-06, "loss": 0.013, "step": 47933 }, { "epoch": 14.75, "learning_rate": 3.40989846991741e-06, "loss": 0.0138, "step": 47934 }, { "epoch": 14.75, "learning_rate": 3.4095237122325984e-06, "loss": 0.0136, "step": 47935 }, { "epoch": 14.75, "learning_rate": 3.4091489709101654e-06, "loss": 0.0173, "step": 47936 }, { "epoch": 14.75, "learning_rate": 3.4087742459510396e-06, "loss": 0.0077, "step": 47937 }, { "epoch": 14.75, "learning_rate": 3.408399537356153e-06, "loss": 0.0133, "step": 47938 }, { "epoch": 14.75, "learning_rate": 3.4080248451264307e-06, "loss": 0.0219, "step": 47939 }, { "epoch": 14.75, "learning_rate": 3.4076501692628017e-06, "loss": 0.0112, "step": 47940 }, { "epoch": 14.75, "learning_rate": 3.4072755097662035e-06, "loss": 0.0142, "step": 47941 }, { "epoch": 14.75, "learning_rate": 3.4069008666375625e-06, "loss": 0.017, "step": 47942 }, { "epoch": 14.75, "learning_rate": 3.40652623987781e-06, "loss": 0.0117, "step": 47943 }, { "epoch": 14.75, "learning_rate": 3.4061516294878747e-06, "loss": 0.0143, "step": 47944 }, { "epoch": 14.75, "learning_rate": 3.405777035468688e-06, "loss": 0.0031, "step": 47945 }, { "epoch": 14.75, "learning_rate": 3.4054024578211785e-06, "loss": 0.0206, "step": 47946 }, { "epoch": 14.75, "learning_rate": 3.405027896546277e-06, "loss": 0.0093, "step": 47947 }, { "epoch": 14.75, "learning_rate": 3.404653351644913e-06, "loss": 0.0094, "step": 47948 }, { "epoch": 14.75, "learning_rate": 3.404278823118017e-06, "loss": 0.0161, "step": 47949 }, { "epoch": 14.75, "learning_rate": 3.403904310966515e-06, "loss": 0.0135, "step": 47950 }, { "epoch": 14.75, "learning_rate": 3.4035298151913477e-06, "loss": 0.0077, "step": 47951 }, { "epoch": 14.75, "learning_rate": 3.4031553357934333e-06, "loss": 0.0167, "step": 47952 }, { "epoch": 14.75, "learning_rate": 3.4027808727737065e-06, "loss": 0.0099, "step": 47953 }, { "epoch": 14.75, "learning_rate": 3.4024064261330956e-06, "loss": 0.0081, "step": 47954 }, { "epoch": 14.76, "learning_rate": 3.4020319958725277e-06, "loss": 0.0141, "step": 47955 }, { "epoch": 14.76, "learning_rate": 3.4016575819929388e-06, "loss": 0.0155, "step": 47956 }, { "epoch": 14.76, "learning_rate": 3.4012831844952554e-06, "loss": 0.0253, "step": 47957 }, { "epoch": 14.76, "learning_rate": 3.4009088033804073e-06, "loss": 0.0109, "step": 47958 }, { "epoch": 14.76, "learning_rate": 3.4005344386493266e-06, "loss": 0.0099, "step": 47959 }, { "epoch": 14.76, "learning_rate": 3.4001600903029332e-06, "loss": 0.0127, "step": 47960 }, { "epoch": 14.76, "learning_rate": 3.3997857583421656e-06, "loss": 0.0352, "step": 47961 }, { "epoch": 14.76, "learning_rate": 3.3994114427679516e-06, "loss": 0.0086, "step": 47962 }, { "epoch": 14.76, "learning_rate": 3.3990371435812185e-06, "loss": 0.0144, "step": 47963 }, { "epoch": 14.76, "learning_rate": 3.3986628607828974e-06, "loss": 0.0162, "step": 47964 }, { "epoch": 14.76, "learning_rate": 3.3982885943739172e-06, "loss": 0.0103, "step": 47965 }, { "epoch": 14.76, "learning_rate": 3.3979143443552054e-06, "loss": 0.013, "step": 47966 }, { "epoch": 14.76, "learning_rate": 3.397540110727694e-06, "loss": 0.0143, "step": 47967 }, { "epoch": 14.76, "learning_rate": 3.397165893492309e-06, "loss": 0.0114, "step": 47968 }, { "epoch": 14.76, "learning_rate": 3.3967916926499823e-06, "loss": 0.0155, "step": 47969 }, { "epoch": 14.76, "learning_rate": 3.3964175082016372e-06, "loss": 0.0193, "step": 47970 }, { "epoch": 14.76, "learning_rate": 3.3960433401482107e-06, "loss": 0.0093, "step": 47971 }, { "epoch": 14.76, "learning_rate": 3.3956691884906324e-06, "loss": 0.0232, "step": 47972 }, { "epoch": 14.76, "learning_rate": 3.395295053229822e-06, "loss": 0.0194, "step": 47973 }, { "epoch": 14.76, "learning_rate": 3.3949209343667155e-06, "loss": 0.0068, "step": 47974 }, { "epoch": 14.76, "learning_rate": 3.3945468319022345e-06, "loss": 0.0108, "step": 47975 }, { "epoch": 14.76, "learning_rate": 3.3941727458373177e-06, "loss": 0.0211, "step": 47976 }, { "epoch": 14.76, "learning_rate": 3.393798676172888e-06, "loss": 0.0154, "step": 47977 }, { "epoch": 14.76, "learning_rate": 3.3934246229098753e-06, "loss": 0.0114, "step": 47978 }, { "epoch": 14.76, "learning_rate": 3.393050586049209e-06, "loss": 0.0034, "step": 47979 }, { "epoch": 14.76, "learning_rate": 3.392676565591817e-06, "loss": 0.0098, "step": 47980 }, { "epoch": 14.76, "learning_rate": 3.3923025615386274e-06, "loss": 0.0103, "step": 47981 }, { "epoch": 14.76, "learning_rate": 3.391928573890568e-06, "loss": 0.0168, "step": 47982 }, { "epoch": 14.76, "learning_rate": 3.3915546026485703e-06, "loss": 0.0086, "step": 47983 }, { "epoch": 14.76, "learning_rate": 3.39118064781356e-06, "loss": 0.0063, "step": 47984 }, { "epoch": 14.76, "learning_rate": 3.390806709386467e-06, "loss": 0.0139, "step": 47985 }, { "epoch": 14.76, "learning_rate": 3.3904327873682186e-06, "loss": 0.0145, "step": 47986 }, { "epoch": 14.77, "learning_rate": 3.3900588817597434e-06, "loss": 0.009, "step": 47987 }, { "epoch": 14.77, "learning_rate": 3.3896849925619713e-06, "loss": 0.0114, "step": 47988 }, { "epoch": 14.77, "learning_rate": 3.3893111197758276e-06, "loss": 0.0147, "step": 47989 }, { "epoch": 14.77, "learning_rate": 3.38893726340224e-06, "loss": 0.0142, "step": 47990 }, { "epoch": 14.77, "learning_rate": 3.3885634234421415e-06, "loss": 0.0171, "step": 47991 }, { "epoch": 14.77, "learning_rate": 3.3881895998964587e-06, "loss": 0.0245, "step": 47992 }, { "epoch": 14.77, "learning_rate": 3.3878157927661215e-06, "loss": 0.0134, "step": 47993 }, { "epoch": 14.77, "learning_rate": 3.3874420020520515e-06, "loss": 0.0201, "step": 47994 }, { "epoch": 14.77, "learning_rate": 3.387068227755177e-06, "loss": 0.0117, "step": 47995 }, { "epoch": 14.77, "learning_rate": 3.3866944698764325e-06, "loss": 0.0152, "step": 47996 }, { "epoch": 14.77, "learning_rate": 3.386320728416743e-06, "loss": 0.0143, "step": 47997 }, { "epoch": 14.77, "learning_rate": 3.385947003377037e-06, "loss": 0.0163, "step": 47998 }, { "epoch": 14.77, "learning_rate": 3.38557329475824e-06, "loss": 0.0115, "step": 47999 }, { "epoch": 14.77, "learning_rate": 3.385199602561282e-06, "loss": 0.0326, "step": 48000 }, { "epoch": 14.77, "learning_rate": 3.3848259267870897e-06, "loss": 0.0114, "step": 48001 }, { "epoch": 14.77, "learning_rate": 3.3844522674365922e-06, "loss": 0.0112, "step": 48002 }, { "epoch": 14.77, "learning_rate": 3.384078624510715e-06, "loss": 0.0165, "step": 48003 }, { "epoch": 14.77, "learning_rate": 3.3837049980103876e-06, "loss": 0.0074, "step": 48004 }, { "epoch": 14.77, "learning_rate": 3.3833313879365337e-06, "loss": 0.0108, "step": 48005 }, { "epoch": 14.77, "learning_rate": 3.3829577942900915e-06, "loss": 0.0211, "step": 48006 }, { "epoch": 14.77, "learning_rate": 3.382584217071977e-06, "loss": 0.0116, "step": 48007 }, { "epoch": 14.77, "learning_rate": 3.3822106562831213e-06, "loss": 0.017, "step": 48008 }, { "epoch": 14.77, "learning_rate": 3.381837111924453e-06, "loss": 0.0164, "step": 48009 }, { "epoch": 14.77, "learning_rate": 3.381463583996896e-06, "loss": 0.0129, "step": 48010 }, { "epoch": 14.77, "learning_rate": 3.381090072501384e-06, "loss": 0.0328, "step": 48011 }, { "epoch": 14.77, "learning_rate": 3.3807165774388405e-06, "loss": 0.011, "step": 48012 }, { "epoch": 14.77, "learning_rate": 3.3803430988101937e-06, "loss": 0.0053, "step": 48013 }, { "epoch": 14.77, "learning_rate": 3.379969636616374e-06, "loss": 0.008, "step": 48014 }, { "epoch": 14.77, "learning_rate": 3.3795961908582965e-06, "loss": 0.0149, "step": 48015 }, { "epoch": 14.77, "learning_rate": 3.379222761536902e-06, "loss": 0.0118, "step": 48016 }, { "epoch": 14.77, "learning_rate": 3.3788493486531116e-06, "loss": 0.009, "step": 48017 }, { "epoch": 14.77, "learning_rate": 3.3784759522078538e-06, "loss": 0.015, "step": 48018 }, { "epoch": 14.77, "learning_rate": 3.3781025722020545e-06, "loss": 0.013, "step": 48019 }, { "epoch": 14.78, "learning_rate": 3.377729208636642e-06, "loss": 0.0111, "step": 48020 }, { "epoch": 14.78, "learning_rate": 3.3773558615125425e-06, "loss": 0.0168, "step": 48021 }, { "epoch": 14.78, "learning_rate": 3.3769825308306826e-06, "loss": 0.0182, "step": 48022 }, { "epoch": 14.78, "learning_rate": 3.3766092165919896e-06, "loss": 0.0059, "step": 48023 }, { "epoch": 14.78, "learning_rate": 3.3762359187973905e-06, "loss": 0.0059, "step": 48024 }, { "epoch": 14.78, "learning_rate": 3.3758626374478077e-06, "loss": 0.0183, "step": 48025 }, { "epoch": 14.78, "learning_rate": 3.375489372544176e-06, "loss": 0.0163, "step": 48026 }, { "epoch": 14.78, "learning_rate": 3.3751161240874233e-06, "loss": 0.0072, "step": 48027 }, { "epoch": 14.78, "learning_rate": 3.374742892078465e-06, "loss": 0.0143, "step": 48028 }, { "epoch": 14.78, "learning_rate": 3.3743696765182367e-06, "loss": 0.0071, "step": 48029 }, { "epoch": 14.78, "learning_rate": 3.373996477407657e-06, "loss": 0.0198, "step": 48030 }, { "epoch": 14.78, "learning_rate": 3.373623294747662e-06, "loss": 0.0156, "step": 48031 }, { "epoch": 14.78, "learning_rate": 3.373250128539174e-06, "loss": 0.0161, "step": 48032 }, { "epoch": 14.78, "learning_rate": 3.3728769787831184e-06, "loss": 0.008, "step": 48033 }, { "epoch": 14.78, "learning_rate": 3.3725038454804237e-06, "loss": 0.008, "step": 48034 }, { "epoch": 14.78, "learning_rate": 3.372130728632015e-06, "loss": 0.0074, "step": 48035 }, { "epoch": 14.78, "learning_rate": 3.3717576282388185e-06, "loss": 0.0132, "step": 48036 }, { "epoch": 14.78, "learning_rate": 3.3713845443017613e-06, "loss": 0.0113, "step": 48037 }, { "epoch": 14.78, "learning_rate": 3.3710114768217695e-06, "loss": 0.0097, "step": 48038 }, { "epoch": 14.78, "learning_rate": 3.370638425799768e-06, "loss": 0.0329, "step": 48039 }, { "epoch": 14.78, "learning_rate": 3.3702653912366856e-06, "loss": 0.0129, "step": 48040 }, { "epoch": 14.78, "learning_rate": 3.3698923731334453e-06, "loss": 0.0065, "step": 48041 }, { "epoch": 14.78, "learning_rate": 3.3695193714909756e-06, "loss": 0.0128, "step": 48042 }, { "epoch": 14.78, "learning_rate": 3.3691463863102025e-06, "loss": 0.0086, "step": 48043 }, { "epoch": 14.78, "learning_rate": 3.3687734175920505e-06, "loss": 0.009, "step": 48044 }, { "epoch": 14.78, "learning_rate": 3.3684004653374424e-06, "loss": 0.0081, "step": 48045 }, { "epoch": 14.78, "learning_rate": 3.368027529547313e-06, "loss": 0.013, "step": 48046 }, { "epoch": 14.78, "learning_rate": 3.3676546102225827e-06, "loss": 0.0218, "step": 48047 }, { "epoch": 14.78, "learning_rate": 3.3672817073641807e-06, "loss": 0.0087, "step": 48048 }, { "epoch": 14.78, "learning_rate": 3.366908820973027e-06, "loss": 0.0115, "step": 48049 }, { "epoch": 14.78, "learning_rate": 3.3665359510500473e-06, "loss": 0.0159, "step": 48050 }, { "epoch": 14.78, "learning_rate": 3.3661630975961735e-06, "loss": 0.0142, "step": 48051 }, { "epoch": 14.79, "learning_rate": 3.3657902606123293e-06, "loss": 0.0087, "step": 48052 }, { "epoch": 14.79, "learning_rate": 3.365417440099439e-06, "loss": 0.016, "step": 48053 }, { "epoch": 14.79, "learning_rate": 3.3650446360584276e-06, "loss": 0.0193, "step": 48054 }, { "epoch": 14.79, "learning_rate": 3.364671848490223e-06, "loss": 0.0222, "step": 48055 }, { "epoch": 14.79, "learning_rate": 3.3642990773957483e-06, "loss": 0.0167, "step": 48056 }, { "epoch": 14.79, "learning_rate": 3.36392632277593e-06, "loss": 0.0101, "step": 48057 }, { "epoch": 14.79, "learning_rate": 3.3635535846316948e-06, "loss": 0.016, "step": 48058 }, { "epoch": 14.79, "learning_rate": 3.3631808629639663e-06, "loss": 0.0198, "step": 48059 }, { "epoch": 14.79, "learning_rate": 3.362808157773666e-06, "loss": 0.0217, "step": 48060 }, { "epoch": 14.79, "learning_rate": 3.362435469061732e-06, "loss": 0.0095, "step": 48061 }, { "epoch": 14.79, "learning_rate": 3.3620627968290777e-06, "loss": 0.0147, "step": 48062 }, { "epoch": 14.79, "learning_rate": 3.361690141076631e-06, "loss": 0.0125, "step": 48063 }, { "epoch": 14.79, "learning_rate": 3.361317501805319e-06, "loss": 0.0139, "step": 48064 }, { "epoch": 14.79, "learning_rate": 3.3609448790160624e-06, "loss": 0.0167, "step": 48065 }, { "epoch": 14.79, "learning_rate": 3.3605722727097923e-06, "loss": 0.0094, "step": 48066 }, { "epoch": 14.79, "learning_rate": 3.360199682887433e-06, "loss": 0.0176, "step": 48067 }, { "epoch": 14.79, "learning_rate": 3.3598271095499056e-06, "loss": 0.0173, "step": 48068 }, { "epoch": 14.79, "learning_rate": 3.3594545526981424e-06, "loss": 0.0219, "step": 48069 }, { "epoch": 14.79, "learning_rate": 3.359082012333057e-06, "loss": 0.0094, "step": 48070 }, { "epoch": 14.79, "learning_rate": 3.358709488455583e-06, "loss": 0.0076, "step": 48071 }, { "epoch": 14.79, "learning_rate": 3.3583369810666423e-06, "loss": 0.0249, "step": 48072 }, { "epoch": 14.79, "learning_rate": 3.3579644901671615e-06, "loss": 0.0149, "step": 48073 }, { "epoch": 14.79, "learning_rate": 3.357592015758064e-06, "loss": 0.0153, "step": 48074 }, { "epoch": 14.79, "learning_rate": 3.3572195578402744e-06, "loss": 0.0137, "step": 48075 }, { "epoch": 14.79, "learning_rate": 3.3568471164147175e-06, "loss": 0.0208, "step": 48076 }, { "epoch": 14.79, "learning_rate": 3.3564746914823198e-06, "loss": 0.0124, "step": 48077 }, { "epoch": 14.79, "learning_rate": 3.356102283044003e-06, "loss": 0.0135, "step": 48078 }, { "epoch": 14.79, "learning_rate": 3.355729891100693e-06, "loss": 0.0136, "step": 48079 }, { "epoch": 14.79, "learning_rate": 3.3553575156533125e-06, "loss": 0.0139, "step": 48080 }, { "epoch": 14.79, "learning_rate": 3.3549851567027904e-06, "loss": 0.025, "step": 48081 }, { "epoch": 14.79, "learning_rate": 3.354612814250051e-06, "loss": 0.0109, "step": 48082 }, { "epoch": 14.79, "learning_rate": 3.3542404882960146e-06, "loss": 0.0168, "step": 48083 }, { "epoch": 14.79, "learning_rate": 3.353868178841607e-06, "loss": 0.0135, "step": 48084 }, { "epoch": 14.8, "learning_rate": 3.3534958858877498e-06, "loss": 0.0165, "step": 48085 }, { "epoch": 14.8, "learning_rate": 3.3531236094353738e-06, "loss": 0.0117, "step": 48086 }, { "epoch": 14.8, "learning_rate": 3.3527513494853993e-06, "loss": 0.0198, "step": 48087 }, { "epoch": 14.8, "learning_rate": 3.352379106038751e-06, "loss": 0.0147, "step": 48088 }, { "epoch": 14.8, "learning_rate": 3.3520068790963546e-06, "loss": 0.0231, "step": 48089 }, { "epoch": 14.8, "learning_rate": 3.3516346686591316e-06, "loss": 0.0065, "step": 48090 }, { "epoch": 14.8, "learning_rate": 3.3512624747280088e-06, "loss": 0.0095, "step": 48091 }, { "epoch": 14.8, "learning_rate": 3.350890297303907e-06, "loss": 0.0175, "step": 48092 }, { "epoch": 14.8, "learning_rate": 3.3505181363877536e-06, "loss": 0.0099, "step": 48093 }, { "epoch": 14.8, "learning_rate": 3.3501459919804703e-06, "loss": 0.0197, "step": 48094 }, { "epoch": 14.8, "learning_rate": 3.349773864082978e-06, "loss": 0.0099, "step": 48095 }, { "epoch": 14.8, "learning_rate": 3.349401752696212e-06, "loss": 0.0234, "step": 48096 }, { "epoch": 14.8, "learning_rate": 3.3490296578210845e-06, "loss": 0.0104, "step": 48097 }, { "epoch": 14.8, "learning_rate": 3.348657579458524e-06, "loss": 0.0138, "step": 48098 }, { "epoch": 14.8, "learning_rate": 3.348285517609453e-06, "loss": 0.0099, "step": 48099 }, { "epoch": 14.8, "learning_rate": 3.347913472274792e-06, "loss": 0.0102, "step": 48100 }, { "epoch": 14.8, "learning_rate": 3.347541443455473e-06, "loss": 0.0049, "step": 48101 }, { "epoch": 14.8, "learning_rate": 3.347169431152414e-06, "loss": 0.0121, "step": 48102 }, { "epoch": 14.8, "learning_rate": 3.3467974353665433e-06, "loss": 0.0166, "step": 48103 }, { "epoch": 14.8, "learning_rate": 3.3464254560987787e-06, "loss": 0.0117, "step": 48104 }, { "epoch": 14.8, "learning_rate": 3.346053493350042e-06, "loss": 0.0097, "step": 48105 }, { "epoch": 14.8, "learning_rate": 3.3456815471212634e-06, "loss": 0.0099, "step": 48106 }, { "epoch": 14.8, "learning_rate": 3.3453096174133635e-06, "loss": 0.0085, "step": 48107 }, { "epoch": 14.8, "learning_rate": 3.344937704227267e-06, "loss": 0.0142, "step": 48108 }, { "epoch": 14.8, "learning_rate": 3.3445658075638955e-06, "loss": 0.0072, "step": 48109 }, { "epoch": 14.8, "learning_rate": 3.344193927424172e-06, "loss": 0.0104, "step": 48110 }, { "epoch": 14.8, "learning_rate": 3.3438220638090212e-06, "loss": 0.0185, "step": 48111 }, { "epoch": 14.8, "learning_rate": 3.343450216719366e-06, "loss": 0.0125, "step": 48112 }, { "epoch": 14.8, "learning_rate": 3.343078386156129e-06, "loss": 0.0165, "step": 48113 }, { "epoch": 14.8, "learning_rate": 3.3427065721202345e-06, "loss": 0.0188, "step": 48114 }, { "epoch": 14.8, "learning_rate": 3.3423347746126e-06, "loss": 0.0102, "step": 48115 }, { "epoch": 14.8, "learning_rate": 3.341962993634158e-06, "loss": 0.0075, "step": 48116 }, { "epoch": 14.81, "learning_rate": 3.34159122918583e-06, "loss": 0.0057, "step": 48117 }, { "epoch": 14.81, "learning_rate": 3.341219481268533e-06, "loss": 0.0175, "step": 48118 }, { "epoch": 14.81, "learning_rate": 3.3408477498831917e-06, "loss": 0.0078, "step": 48119 }, { "epoch": 14.81, "learning_rate": 3.3404760350307274e-06, "loss": 0.0152, "step": 48120 }, { "epoch": 14.81, "learning_rate": 3.340104336712069e-06, "loss": 0.0168, "step": 48121 }, { "epoch": 14.81, "learning_rate": 3.339732654928136e-06, "loss": 0.0116, "step": 48122 }, { "epoch": 14.81, "learning_rate": 3.339360989679852e-06, "loss": 0.0167, "step": 48123 }, { "epoch": 14.81, "learning_rate": 3.3389893409681417e-06, "loss": 0.012, "step": 48124 }, { "epoch": 14.81, "learning_rate": 3.338617708793919e-06, "loss": 0.0166, "step": 48125 }, { "epoch": 14.81, "learning_rate": 3.3382460931581164e-06, "loss": 0.0126, "step": 48126 }, { "epoch": 14.81, "learning_rate": 3.3378744940616516e-06, "loss": 0.0115, "step": 48127 }, { "epoch": 14.81, "learning_rate": 3.337502911505448e-06, "loss": 0.0163, "step": 48128 }, { "epoch": 14.81, "learning_rate": 3.33713134549043e-06, "loss": 0.009, "step": 48129 }, { "epoch": 14.81, "learning_rate": 3.336759796017518e-06, "loss": 0.0252, "step": 48130 }, { "epoch": 14.81, "learning_rate": 3.3363882630876355e-06, "loss": 0.0103, "step": 48131 }, { "epoch": 14.81, "learning_rate": 3.3360167467017036e-06, "loss": 0.0167, "step": 48132 }, { "epoch": 14.81, "learning_rate": 3.335645246860646e-06, "loss": 0.0137, "step": 48133 }, { "epoch": 14.81, "learning_rate": 3.335273763565384e-06, "loss": 0.0199, "step": 48134 }, { "epoch": 14.81, "learning_rate": 3.334902296816838e-06, "loss": 0.0076, "step": 48135 }, { "epoch": 14.81, "learning_rate": 3.3345308466159356e-06, "loss": 0.0107, "step": 48136 }, { "epoch": 14.81, "learning_rate": 3.334159412963597e-06, "loss": 0.0071, "step": 48137 }, { "epoch": 14.81, "learning_rate": 3.3337879958607455e-06, "loss": 0.0142, "step": 48138 }, { "epoch": 14.81, "learning_rate": 3.3334165953082987e-06, "loss": 0.0194, "step": 48139 }, { "epoch": 14.81, "learning_rate": 3.3330452113071777e-06, "loss": 0.0188, "step": 48140 }, { "epoch": 14.81, "learning_rate": 3.3326738438583116e-06, "loss": 0.0184, "step": 48141 }, { "epoch": 14.81, "learning_rate": 3.332302492962619e-06, "loss": 0.0096, "step": 48142 }, { "epoch": 14.81, "learning_rate": 3.3319311586210225e-06, "loss": 0.0093, "step": 48143 }, { "epoch": 14.81, "learning_rate": 3.331559840834442e-06, "loss": 0.0164, "step": 48144 }, { "epoch": 14.81, "learning_rate": 3.3311885396038002e-06, "loss": 0.0105, "step": 48145 }, { "epoch": 14.81, "learning_rate": 3.330817254930021e-06, "loss": 0.0168, "step": 48146 }, { "epoch": 14.81, "learning_rate": 3.3304459868140237e-06, "loss": 0.0105, "step": 48147 }, { "epoch": 14.81, "learning_rate": 3.330074735256731e-06, "loss": 0.0069, "step": 48148 }, { "epoch": 14.81, "learning_rate": 3.329703500259065e-06, "loss": 0.0105, "step": 48149 }, { "epoch": 14.82, "learning_rate": 3.329332281821943e-06, "loss": 0.0081, "step": 48150 }, { "epoch": 14.82, "learning_rate": 3.3289610799462978e-06, "loss": 0.0142, "step": 48151 }, { "epoch": 14.82, "learning_rate": 3.3285898946330396e-06, "loss": 0.0115, "step": 48152 }, { "epoch": 14.82, "learning_rate": 3.3282187258830943e-06, "loss": 0.0056, "step": 48153 }, { "epoch": 14.82, "learning_rate": 3.327847573697385e-06, "loss": 0.0084, "step": 48154 }, { "epoch": 14.82, "learning_rate": 3.3274764380768265e-06, "loss": 0.0237, "step": 48155 }, { "epoch": 14.82, "learning_rate": 3.327105319022348e-06, "loss": 0.0243, "step": 48156 }, { "epoch": 14.82, "learning_rate": 3.3267342165348694e-06, "loss": 0.0212, "step": 48157 }, { "epoch": 14.82, "learning_rate": 3.32636313061531e-06, "loss": 0.0154, "step": 48158 }, { "epoch": 14.82, "learning_rate": 3.3259920612645948e-06, "loss": 0.0139, "step": 48159 }, { "epoch": 14.82, "learning_rate": 3.325621008483636e-06, "loss": 0.0129, "step": 48160 }, { "epoch": 14.82, "learning_rate": 3.3252499722733646e-06, "loss": 0.0212, "step": 48161 }, { "epoch": 14.82, "learning_rate": 3.324878952634697e-06, "loss": 0.0076, "step": 48162 }, { "epoch": 14.82, "learning_rate": 3.324507949568556e-06, "loss": 0.0232, "step": 48163 }, { "epoch": 14.82, "learning_rate": 3.324136963075861e-06, "loss": 0.0074, "step": 48164 }, { "epoch": 14.82, "learning_rate": 3.3237659931575363e-06, "loss": 0.0188, "step": 48165 }, { "epoch": 14.82, "learning_rate": 3.3233950398144985e-06, "loss": 0.0145, "step": 48166 }, { "epoch": 14.82, "learning_rate": 3.3230241030476728e-06, "loss": 0.0092, "step": 48167 }, { "epoch": 14.82, "learning_rate": 3.322653182857977e-06, "loss": 0.0177, "step": 48168 }, { "epoch": 14.82, "learning_rate": 3.322282279246334e-06, "loss": 0.0064, "step": 48169 }, { "epoch": 14.82, "learning_rate": 3.321911392213659e-06, "loss": 0.0246, "step": 48170 }, { "epoch": 14.82, "learning_rate": 3.321540521760883e-06, "loss": 0.0102, "step": 48171 }, { "epoch": 14.82, "learning_rate": 3.3211696678889236e-06, "loss": 0.0174, "step": 48172 }, { "epoch": 14.82, "learning_rate": 3.320798830598696e-06, "loss": 0.0152, "step": 48173 }, { "epoch": 14.82, "learning_rate": 3.3204280098911247e-06, "loss": 0.0087, "step": 48174 }, { "epoch": 14.82, "learning_rate": 3.3200572057671267e-06, "loss": 0.0149, "step": 48175 }, { "epoch": 14.82, "learning_rate": 3.3196864182276276e-06, "loss": 0.0105, "step": 48176 }, { "epoch": 14.82, "learning_rate": 3.319315647273548e-06, "loss": 0.0098, "step": 48177 }, { "epoch": 14.82, "learning_rate": 3.3189448929058055e-06, "loss": 0.0127, "step": 48178 }, { "epoch": 14.82, "learning_rate": 3.3185741551253226e-06, "loss": 0.0156, "step": 48179 }, { "epoch": 14.82, "learning_rate": 3.318203433933018e-06, "loss": 0.0146, "step": 48180 }, { "epoch": 14.82, "learning_rate": 3.317832729329814e-06, "loss": 0.0123, "step": 48181 }, { "epoch": 14.83, "learning_rate": 3.317462041316628e-06, "loss": 0.0144, "step": 48182 }, { "epoch": 14.83, "learning_rate": 3.317091369894384e-06, "loss": 0.0118, "step": 48183 }, { "epoch": 14.83, "learning_rate": 3.3167207150640003e-06, "loss": 0.0077, "step": 48184 }, { "epoch": 14.83, "learning_rate": 3.3163500768263957e-06, "loss": 0.0066, "step": 48185 }, { "epoch": 14.83, "learning_rate": 3.315979455182493e-06, "loss": 0.0111, "step": 48186 }, { "epoch": 14.83, "learning_rate": 3.315608850133212e-06, "loss": 0.014, "step": 48187 }, { "epoch": 14.83, "learning_rate": 3.315238261679472e-06, "loss": 0.0175, "step": 48188 }, { "epoch": 14.83, "learning_rate": 3.3148676898221922e-06, "loss": 0.008, "step": 48189 }, { "epoch": 14.83, "learning_rate": 3.314497134562291e-06, "loss": 0.0126, "step": 48190 }, { "epoch": 14.83, "learning_rate": 3.314126595900695e-06, "loss": 0.0088, "step": 48191 }, { "epoch": 14.83, "learning_rate": 3.313756073838319e-06, "loss": 0.0146, "step": 48192 }, { "epoch": 14.83, "learning_rate": 3.3133855683760886e-06, "loss": 0.0143, "step": 48193 }, { "epoch": 14.83, "learning_rate": 3.313015079514915e-06, "loss": 0.0203, "step": 48194 }, { "epoch": 14.83, "learning_rate": 3.3126446072557196e-06, "loss": 0.0137, "step": 48195 }, { "epoch": 14.83, "learning_rate": 3.3122741515994274e-06, "loss": 0.0116, "step": 48196 }, { "epoch": 14.83, "learning_rate": 3.3119037125469553e-06, "loss": 0.0192, "step": 48197 }, { "epoch": 14.83, "learning_rate": 3.311533290099224e-06, "loss": 0.021, "step": 48198 }, { "epoch": 14.83, "learning_rate": 3.3111628842571518e-06, "loss": 0.0059, "step": 48199 }, { "epoch": 14.83, "learning_rate": 3.3107924950216596e-06, "loss": 0.0245, "step": 48200 }, { "epoch": 14.83, "learning_rate": 3.3104221223936663e-06, "loss": 0.0231, "step": 48201 }, { "epoch": 14.83, "learning_rate": 3.3100517663740904e-06, "loss": 0.0108, "step": 48202 }, { "epoch": 14.83, "learning_rate": 3.309681426963853e-06, "loss": 0.0279, "step": 48203 }, { "epoch": 14.83, "learning_rate": 3.3093111041638728e-06, "loss": 0.0199, "step": 48204 }, { "epoch": 14.83, "learning_rate": 3.3089407979750665e-06, "loss": 0.0096, "step": 48205 }, { "epoch": 14.83, "learning_rate": 3.308570508398363e-06, "loss": 0.0226, "step": 48206 }, { "epoch": 14.83, "learning_rate": 3.308200235434671e-06, "loss": 0.0163, "step": 48207 }, { "epoch": 14.83, "learning_rate": 3.3078299790849145e-06, "loss": 0.0191, "step": 48208 }, { "epoch": 14.83, "learning_rate": 3.3074597393500116e-06, "loss": 0.0141, "step": 48209 }, { "epoch": 14.83, "learning_rate": 3.307089516230878e-06, "loss": 0.0091, "step": 48210 }, { "epoch": 14.83, "learning_rate": 3.3067193097284412e-06, "loss": 0.0135, "step": 48211 }, { "epoch": 14.83, "learning_rate": 3.306349119843616e-06, "loss": 0.0138, "step": 48212 }, { "epoch": 14.83, "learning_rate": 3.3059789465773206e-06, "loss": 0.0083, "step": 48213 }, { "epoch": 14.83, "learning_rate": 3.3056087899304777e-06, "loss": 0.0071, "step": 48214 }, { "epoch": 14.84, "learning_rate": 3.305238649903998e-06, "loss": 0.0088, "step": 48215 }, { "epoch": 14.84, "learning_rate": 3.304868526498808e-06, "loss": 0.0075, "step": 48216 }, { "epoch": 14.84, "learning_rate": 3.304498419715826e-06, "loss": 0.009, "step": 48217 }, { "epoch": 14.84, "learning_rate": 3.304128329555968e-06, "loss": 0.0121, "step": 48218 }, { "epoch": 14.84, "learning_rate": 3.303758256020154e-06, "loss": 0.0181, "step": 48219 }, { "epoch": 14.84, "learning_rate": 3.303388199109304e-06, "loss": 0.0197, "step": 48220 }, { "epoch": 14.84, "learning_rate": 3.303018158824336e-06, "loss": 0.018, "step": 48221 }, { "epoch": 14.84, "learning_rate": 3.3026481351661677e-06, "loss": 0.01, "step": 48222 }, { "epoch": 14.84, "learning_rate": 3.3022781281357184e-06, "loss": 0.0094, "step": 48223 }, { "epoch": 14.84, "learning_rate": 3.3019081377339077e-06, "loss": 0.0162, "step": 48224 }, { "epoch": 14.84, "learning_rate": 3.3015381639616493e-06, "loss": 0.0129, "step": 48225 }, { "epoch": 14.84, "learning_rate": 3.3011682068198694e-06, "loss": 0.0224, "step": 48226 }, { "epoch": 14.84, "learning_rate": 3.300798266309486e-06, "loss": 0.0101, "step": 48227 }, { "epoch": 14.84, "learning_rate": 3.300428342431412e-06, "loss": 0.0081, "step": 48228 }, { "epoch": 14.84, "learning_rate": 3.300058435186567e-06, "loss": 0.007, "step": 48229 }, { "epoch": 14.84, "learning_rate": 3.2996885445758688e-06, "loss": 0.0258, "step": 48230 }, { "epoch": 14.84, "learning_rate": 3.299318670600239e-06, "loss": 0.0127, "step": 48231 }, { "epoch": 14.84, "learning_rate": 3.2989488132605963e-06, "loss": 0.0226, "step": 48232 }, { "epoch": 14.84, "learning_rate": 3.2985789725578563e-06, "loss": 0.0214, "step": 48233 }, { "epoch": 14.84, "learning_rate": 3.2982091484929378e-06, "loss": 0.0136, "step": 48234 }, { "epoch": 14.84, "learning_rate": 3.2978393410667596e-06, "loss": 0.0195, "step": 48235 }, { "epoch": 14.84, "learning_rate": 3.297469550280239e-06, "loss": 0.0154, "step": 48236 }, { "epoch": 14.84, "learning_rate": 3.297099776134296e-06, "loss": 0.0101, "step": 48237 }, { "epoch": 14.84, "learning_rate": 3.2967300186298456e-06, "loss": 0.0201, "step": 48238 }, { "epoch": 14.84, "learning_rate": 3.2963602777678084e-06, "loss": 0.0056, "step": 48239 }, { "epoch": 14.84, "learning_rate": 3.295990553549101e-06, "loss": 0.0092, "step": 48240 }, { "epoch": 14.84, "learning_rate": 3.295620845974642e-06, "loss": 0.0127, "step": 48241 }, { "epoch": 14.84, "learning_rate": 3.295251155045348e-06, "loss": 0.0123, "step": 48242 }, { "epoch": 14.84, "learning_rate": 3.294881480762139e-06, "loss": 0.0227, "step": 48243 }, { "epoch": 14.84, "learning_rate": 3.294511823125931e-06, "loss": 0.0104, "step": 48244 }, { "epoch": 14.84, "learning_rate": 3.2941421821376397e-06, "loss": 0.0186, "step": 48245 }, { "epoch": 14.84, "learning_rate": 3.2937725577981884e-06, "loss": 0.0045, "step": 48246 }, { "epoch": 14.85, "learning_rate": 3.293402950108492e-06, "loss": 0.0079, "step": 48247 }, { "epoch": 14.85, "learning_rate": 3.2930333590694707e-06, "loss": 0.0075, "step": 48248 }, { "epoch": 14.85, "learning_rate": 3.2926637846820366e-06, "loss": 0.0115, "step": 48249 }, { "epoch": 14.85, "learning_rate": 3.292294226947107e-06, "loss": 0.0231, "step": 48250 }, { "epoch": 14.85, "learning_rate": 3.2919246858656053e-06, "loss": 0.0297, "step": 48251 }, { "epoch": 14.85, "learning_rate": 3.2915551614384465e-06, "loss": 0.0117, "step": 48252 }, { "epoch": 14.85, "learning_rate": 3.2911856536665475e-06, "loss": 0.0119, "step": 48253 }, { "epoch": 14.85, "learning_rate": 3.2908161625508262e-06, "loss": 0.0124, "step": 48254 }, { "epoch": 14.85, "learning_rate": 3.2904466880922e-06, "loss": 0.0138, "step": 48255 }, { "epoch": 14.85, "learning_rate": 3.290077230291585e-06, "loss": 0.013, "step": 48256 }, { "epoch": 14.85, "learning_rate": 3.2897077891498996e-06, "loss": 0.0083, "step": 48257 }, { "epoch": 14.85, "learning_rate": 3.289338364668061e-06, "loss": 0.0143, "step": 48258 }, { "epoch": 14.85, "learning_rate": 3.288968956846986e-06, "loss": 0.0073, "step": 48259 }, { "epoch": 14.85, "learning_rate": 3.288599565687588e-06, "loss": 0.0146, "step": 48260 }, { "epoch": 14.85, "learning_rate": 3.2882301911907955e-06, "loss": 0.0208, "step": 48261 }, { "epoch": 14.85, "learning_rate": 3.2878608333575136e-06, "loss": 0.007, "step": 48262 }, { "epoch": 14.85, "learning_rate": 3.2874914921886646e-06, "loss": 0.0174, "step": 48263 }, { "epoch": 14.85, "learning_rate": 3.287122167685164e-06, "loss": 0.0203, "step": 48264 }, { "epoch": 14.85, "learning_rate": 3.286752859847926e-06, "loss": 0.0086, "step": 48265 }, { "epoch": 14.85, "learning_rate": 3.2863835686778746e-06, "loss": 0.0104, "step": 48266 }, { "epoch": 14.85, "learning_rate": 3.286014294175923e-06, "loss": 0.0092, "step": 48267 }, { "epoch": 14.85, "learning_rate": 3.285645036342987e-06, "loss": 0.014, "step": 48268 }, { "epoch": 14.85, "learning_rate": 3.2852757951799887e-06, "loss": 0.0145, "step": 48269 }, { "epoch": 14.85, "learning_rate": 3.2849065706878327e-06, "loss": 0.0246, "step": 48270 }, { "epoch": 14.85, "learning_rate": 3.284537362867447e-06, "loss": 0.0169, "step": 48271 }, { "epoch": 14.85, "learning_rate": 3.2841681717197447e-06, "loss": 0.0048, "step": 48272 }, { "epoch": 14.85, "learning_rate": 3.283798997245643e-06, "loss": 0.0122, "step": 48273 }, { "epoch": 14.85, "learning_rate": 3.2834298394460528e-06, "loss": 0.0099, "step": 48274 }, { "epoch": 14.85, "learning_rate": 3.2830606983219038e-06, "loss": 0.0035, "step": 48275 }, { "epoch": 14.85, "learning_rate": 3.2826915738741005e-06, "loss": 0.0159, "step": 48276 }, { "epoch": 14.85, "learning_rate": 3.2823224661035623e-06, "loss": 0.0139, "step": 48277 }, { "epoch": 14.85, "learning_rate": 3.2819533750112063e-06, "loss": 0.0103, "step": 48278 }, { "epoch": 14.85, "learning_rate": 3.2815843005979475e-06, "loss": 0.0098, "step": 48279 }, { "epoch": 14.86, "learning_rate": 3.2812152428647046e-06, "loss": 0.0074, "step": 48280 }, { "epoch": 14.86, "learning_rate": 3.280846201812394e-06, "loss": 0.0052, "step": 48281 }, { "epoch": 14.86, "learning_rate": 3.280477177441935e-06, "loss": 0.0127, "step": 48282 }, { "epoch": 14.86, "learning_rate": 3.280108169754236e-06, "loss": 0.0153, "step": 48283 }, { "epoch": 14.86, "learning_rate": 3.279739178750212e-06, "loss": 0.0099, "step": 48284 }, { "epoch": 14.86, "learning_rate": 3.2793702044307885e-06, "loss": 0.0085, "step": 48285 }, { "epoch": 14.86, "learning_rate": 3.2790012467968767e-06, "loss": 0.0186, "step": 48286 }, { "epoch": 14.86, "learning_rate": 3.278632305849393e-06, "loss": 0.0198, "step": 48287 }, { "epoch": 14.86, "learning_rate": 3.278263381589253e-06, "loss": 0.0147, "step": 48288 }, { "epoch": 14.86, "learning_rate": 3.2778944740173736e-06, "loss": 0.017, "step": 48289 }, { "epoch": 14.86, "learning_rate": 3.2775255831346686e-06, "loss": 0.0198, "step": 48290 }, { "epoch": 14.86, "learning_rate": 3.2771567089420563e-06, "loss": 0.0125, "step": 48291 }, { "epoch": 14.86, "learning_rate": 3.276787851440452e-06, "loss": 0.0159, "step": 48292 }, { "epoch": 14.86, "learning_rate": 3.276419010630769e-06, "loss": 0.0032, "step": 48293 }, { "epoch": 14.86, "learning_rate": 3.276050186513924e-06, "loss": 0.0157, "step": 48294 }, { "epoch": 14.86, "learning_rate": 3.2756813790908347e-06, "loss": 0.0104, "step": 48295 }, { "epoch": 14.86, "learning_rate": 3.2753125883624206e-06, "loss": 0.013, "step": 48296 }, { "epoch": 14.86, "learning_rate": 3.2749438143295887e-06, "loss": 0.0081, "step": 48297 }, { "epoch": 14.86, "learning_rate": 3.274575056993259e-06, "loss": 0.0064, "step": 48298 }, { "epoch": 14.86, "learning_rate": 3.2742063163543414e-06, "loss": 0.0122, "step": 48299 }, { "epoch": 14.86, "learning_rate": 3.273837592413761e-06, "loss": 0.0221, "step": 48300 }, { "epoch": 14.86, "learning_rate": 3.2734688851724273e-06, "loss": 0.0186, "step": 48301 }, { "epoch": 14.86, "learning_rate": 3.2731001946312578e-06, "loss": 0.0091, "step": 48302 }, { "epoch": 14.86, "learning_rate": 3.2727315207911703e-06, "loss": 0.008, "step": 48303 }, { "epoch": 14.86, "learning_rate": 3.2723628636530703e-06, "loss": 0.0103, "step": 48304 }, { "epoch": 14.86, "learning_rate": 3.2719942232178824e-06, "loss": 0.0225, "step": 48305 }, { "epoch": 14.86, "learning_rate": 3.27162559948652e-06, "loss": 0.0256, "step": 48306 }, { "epoch": 14.86, "learning_rate": 3.271256992459896e-06, "loss": 0.014, "step": 48307 }, { "epoch": 14.86, "learning_rate": 3.2708884021389278e-06, "loss": 0.0176, "step": 48308 }, { "epoch": 14.86, "learning_rate": 3.2705198285245288e-06, "loss": 0.0117, "step": 48309 }, { "epoch": 14.86, "learning_rate": 3.2701512716176153e-06, "loss": 0.0096, "step": 48310 }, { "epoch": 14.86, "learning_rate": 3.2697827314191023e-06, "loss": 0.0099, "step": 48311 }, { "epoch": 14.87, "learning_rate": 3.269414207929904e-06, "loss": 0.0092, "step": 48312 }, { "epoch": 14.87, "learning_rate": 3.269045701150936e-06, "loss": 0.0049, "step": 48313 }, { "epoch": 14.87, "learning_rate": 3.268677211083109e-06, "loss": 0.0074, "step": 48314 }, { "epoch": 14.87, "learning_rate": 3.268308737727345e-06, "loss": 0.0201, "step": 48315 }, { "epoch": 14.87, "learning_rate": 3.2679402810845563e-06, "loss": 0.0166, "step": 48316 }, { "epoch": 14.87, "learning_rate": 3.26757184115566e-06, "loss": 0.0093, "step": 48317 }, { "epoch": 14.87, "learning_rate": 3.2672034179415646e-06, "loss": 0.0141, "step": 48318 }, { "epoch": 14.87, "learning_rate": 3.2668350114431844e-06, "loss": 0.0153, "step": 48319 }, { "epoch": 14.87, "learning_rate": 3.2664666216614416e-06, "loss": 0.0118, "step": 48320 }, { "epoch": 14.87, "learning_rate": 3.266098248597246e-06, "loss": 0.0207, "step": 48321 }, { "epoch": 14.87, "learning_rate": 3.2657298922515137e-06, "loss": 0.0167, "step": 48322 }, { "epoch": 14.87, "learning_rate": 3.265361552625158e-06, "loss": 0.0137, "step": 48323 }, { "epoch": 14.87, "learning_rate": 3.2649932297190945e-06, "loss": 0.0044, "step": 48324 }, { "epoch": 14.87, "learning_rate": 3.2646249235342363e-06, "loss": 0.0058, "step": 48325 }, { "epoch": 14.87, "learning_rate": 3.2642566340715e-06, "loss": 0.0114, "step": 48326 }, { "epoch": 14.87, "learning_rate": 3.2638883613317974e-06, "loss": 0.0086, "step": 48327 }, { "epoch": 14.87, "learning_rate": 3.2635201053160447e-06, "loss": 0.0116, "step": 48328 }, { "epoch": 14.87, "learning_rate": 3.2631518660251514e-06, "loss": 0.0069, "step": 48329 }, { "epoch": 14.87, "learning_rate": 3.2627836434600435e-06, "loss": 0.0218, "step": 48330 }, { "epoch": 14.87, "learning_rate": 3.2624154376216234e-06, "loss": 0.0112, "step": 48331 }, { "epoch": 14.87, "learning_rate": 3.262047248510809e-06, "loss": 0.0088, "step": 48332 }, { "epoch": 14.87, "learning_rate": 3.2616790761285156e-06, "loss": 0.0098, "step": 48333 }, { "epoch": 14.87, "learning_rate": 3.2613109204756533e-06, "loss": 0.009, "step": 48334 }, { "epoch": 14.87, "learning_rate": 3.2609427815531426e-06, "loss": 0.017, "step": 48335 }, { "epoch": 14.87, "learning_rate": 3.260574659361894e-06, "loss": 0.0124, "step": 48336 }, { "epoch": 14.87, "learning_rate": 3.2602065539028237e-06, "loss": 0.0137, "step": 48337 }, { "epoch": 14.87, "learning_rate": 3.259838465176841e-06, "loss": 0.015, "step": 48338 }, { "epoch": 14.87, "learning_rate": 3.2594703931848594e-06, "loss": 0.0089, "step": 48339 }, { "epoch": 14.87, "learning_rate": 3.2591023379277986e-06, "loss": 0.0052, "step": 48340 }, { "epoch": 14.87, "learning_rate": 3.2587342994065697e-06, "loss": 0.0269, "step": 48341 }, { "epoch": 14.87, "learning_rate": 3.258366277622086e-06, "loss": 0.0095, "step": 48342 }, { "epoch": 14.87, "learning_rate": 3.2579982725752612e-06, "loss": 0.0273, "step": 48343 }, { "epoch": 14.87, "learning_rate": 3.2576302842670095e-06, "loss": 0.0182, "step": 48344 }, { "epoch": 14.88, "learning_rate": 3.2572623126982438e-06, "loss": 0.0127, "step": 48345 }, { "epoch": 14.88, "learning_rate": 3.2568943578698786e-06, "loss": 0.0119, "step": 48346 }, { "epoch": 14.88, "learning_rate": 3.256526419782826e-06, "loss": 0.0224, "step": 48347 }, { "epoch": 14.88, "learning_rate": 3.2561584984380014e-06, "loss": 0.0244, "step": 48348 }, { "epoch": 14.88, "learning_rate": 3.255790593836312e-06, "loss": 0.0122, "step": 48349 }, { "epoch": 14.88, "learning_rate": 3.2554227059786812e-06, "loss": 0.0059, "step": 48350 }, { "epoch": 14.88, "learning_rate": 3.2550548348660203e-06, "loss": 0.0299, "step": 48351 }, { "epoch": 14.88, "learning_rate": 3.2546869804992363e-06, "loss": 0.0035, "step": 48352 }, { "epoch": 14.88, "learning_rate": 3.2543191428792466e-06, "loss": 0.0089, "step": 48353 }, { "epoch": 14.88, "learning_rate": 3.253951322006961e-06, "loss": 0.0131, "step": 48354 }, { "epoch": 14.88, "learning_rate": 3.2535835178832977e-06, "loss": 0.0182, "step": 48355 }, { "epoch": 14.88, "learning_rate": 3.253215730509168e-06, "loss": 0.0153, "step": 48356 }, { "epoch": 14.88, "learning_rate": 3.2528479598854857e-06, "loss": 0.0066, "step": 48357 }, { "epoch": 14.88, "learning_rate": 3.2524802060131656e-06, "loss": 0.0183, "step": 48358 }, { "epoch": 14.88, "learning_rate": 3.2521124688931116e-06, "loss": 0.0221, "step": 48359 }, { "epoch": 14.88, "learning_rate": 3.2517447485262466e-06, "loss": 0.0135, "step": 48360 }, { "epoch": 14.88, "learning_rate": 3.2513770449134798e-06, "loss": 0.0125, "step": 48361 }, { "epoch": 14.88, "learning_rate": 3.2510093580557243e-06, "loss": 0.0145, "step": 48362 }, { "epoch": 14.88, "learning_rate": 3.2506416879538938e-06, "loss": 0.0136, "step": 48363 }, { "epoch": 14.88, "learning_rate": 3.250274034608899e-06, "loss": 0.0099, "step": 48364 }, { "epoch": 14.88, "learning_rate": 3.2499063980216562e-06, "loss": 0.0104, "step": 48365 }, { "epoch": 14.88, "learning_rate": 3.249538778193074e-06, "loss": 0.008, "step": 48366 }, { "epoch": 14.88, "learning_rate": 3.2491711751240686e-06, "loss": 0.0077, "step": 48367 }, { "epoch": 14.88, "learning_rate": 3.24880358881555e-06, "loss": 0.0174, "step": 48368 }, { "epoch": 14.88, "learning_rate": 3.2484360192684294e-06, "loss": 0.0171, "step": 48369 }, { "epoch": 14.88, "learning_rate": 3.2480684664836258e-06, "loss": 0.0215, "step": 48370 }, { "epoch": 14.88, "learning_rate": 3.2477009304620468e-06, "loss": 0.0083, "step": 48371 }, { "epoch": 14.88, "learning_rate": 3.2473334112046105e-06, "loss": 0.0178, "step": 48372 }, { "epoch": 14.88, "learning_rate": 3.246965908712221e-06, "loss": 0.0295, "step": 48373 }, { "epoch": 14.88, "learning_rate": 3.2465984229857905e-06, "loss": 0.0109, "step": 48374 }, { "epoch": 14.88, "learning_rate": 3.2462309540262404e-06, "loss": 0.015, "step": 48375 }, { "epoch": 14.88, "learning_rate": 3.2458635018344763e-06, "loss": 0.0213, "step": 48376 }, { "epoch": 14.89, "learning_rate": 3.2454960664114132e-06, "loss": 0.0119, "step": 48377 }, { "epoch": 14.89, "learning_rate": 3.2451286477579623e-06, "loss": 0.0139, "step": 48378 }, { "epoch": 14.89, "learning_rate": 3.2447612458750365e-06, "loss": 0.0085, "step": 48379 }, { "epoch": 14.89, "learning_rate": 3.2443938607635472e-06, "loss": 0.0158, "step": 48380 }, { "epoch": 14.89, "learning_rate": 3.244026492424406e-06, "loss": 0.0119, "step": 48381 }, { "epoch": 14.89, "learning_rate": 3.243659140858526e-06, "loss": 0.0164, "step": 48382 }, { "epoch": 14.89, "learning_rate": 3.2432918060668195e-06, "loss": 0.0039, "step": 48383 }, { "epoch": 14.89, "learning_rate": 3.2429244880501933e-06, "loss": 0.0094, "step": 48384 }, { "epoch": 14.89, "learning_rate": 3.2425571868095716e-06, "loss": 0.0097, "step": 48385 }, { "epoch": 14.89, "learning_rate": 3.2421899023458546e-06, "loss": 0.0106, "step": 48386 }, { "epoch": 14.89, "learning_rate": 3.241822634659959e-06, "loss": 0.0169, "step": 48387 }, { "epoch": 14.89, "learning_rate": 3.2414553837527953e-06, "loss": 0.0175, "step": 48388 }, { "epoch": 14.89, "learning_rate": 3.2410881496252732e-06, "loss": 0.0112, "step": 48389 }, { "epoch": 14.89, "learning_rate": 3.2407209322783095e-06, "loss": 0.015, "step": 48390 }, { "epoch": 14.89, "learning_rate": 3.240353731712814e-06, "loss": 0.0128, "step": 48391 }, { "epoch": 14.89, "learning_rate": 3.2399865479296978e-06, "loss": 0.0229, "step": 48392 }, { "epoch": 14.89, "learning_rate": 3.239619380929876e-06, "loss": 0.0064, "step": 48393 }, { "epoch": 14.89, "learning_rate": 3.2392522307142493e-06, "loss": 0.0166, "step": 48394 }, { "epoch": 14.89, "learning_rate": 3.2388850972837404e-06, "loss": 0.0149, "step": 48395 }, { "epoch": 14.89, "learning_rate": 3.2385179806392575e-06, "loss": 0.0177, "step": 48396 }, { "epoch": 14.89, "learning_rate": 3.2381508807817118e-06, "loss": 0.0122, "step": 48397 }, { "epoch": 14.89, "learning_rate": 3.237783797712014e-06, "loss": 0.008, "step": 48398 }, { "epoch": 14.89, "learning_rate": 3.2374167314310758e-06, "loss": 0.0074, "step": 48399 }, { "epoch": 14.89, "learning_rate": 3.237049681939809e-06, "loss": 0.0148, "step": 48400 }, { "epoch": 14.89, "learning_rate": 3.2366826492391245e-06, "loss": 0.0142, "step": 48401 }, { "epoch": 14.89, "learning_rate": 3.236315633329934e-06, "loss": 0.01, "step": 48402 }, { "epoch": 14.89, "learning_rate": 3.2359486342131476e-06, "loss": 0.0029, "step": 48403 }, { "epoch": 14.89, "learning_rate": 3.2355816518896743e-06, "loss": 0.0129, "step": 48404 }, { "epoch": 14.89, "learning_rate": 3.2352146863604317e-06, "loss": 0.0066, "step": 48405 }, { "epoch": 14.89, "learning_rate": 3.234847737626331e-06, "loss": 0.0103, "step": 48406 }, { "epoch": 14.89, "learning_rate": 3.2344808056882747e-06, "loss": 0.016, "step": 48407 }, { "epoch": 14.89, "learning_rate": 3.23411389054718e-06, "loss": 0.0147, "step": 48408 }, { "epoch": 14.89, "learning_rate": 3.233746992203953e-06, "loss": 0.0171, "step": 48409 }, { "epoch": 14.9, "learning_rate": 3.2333801106595107e-06, "loss": 0.0075, "step": 48410 }, { "epoch": 14.9, "learning_rate": 3.2330132459147624e-06, "loss": 0.0042, "step": 48411 }, { "epoch": 14.9, "learning_rate": 3.232646397970617e-06, "loss": 0.0093, "step": 48412 }, { "epoch": 14.9, "learning_rate": 3.232279566827987e-06, "loss": 0.0179, "step": 48413 }, { "epoch": 14.9, "learning_rate": 3.2319127524877813e-06, "loss": 0.0078, "step": 48414 }, { "epoch": 14.9, "learning_rate": 3.231545954950912e-06, "loss": 0.0092, "step": 48415 }, { "epoch": 14.9, "learning_rate": 3.23117917421829e-06, "loss": 0.0119, "step": 48416 }, { "epoch": 14.9, "learning_rate": 3.2308124102908245e-06, "loss": 0.0161, "step": 48417 }, { "epoch": 14.9, "learning_rate": 3.230445663169427e-06, "loss": 0.0152, "step": 48418 }, { "epoch": 14.9, "learning_rate": 3.2300789328550086e-06, "loss": 0.0214, "step": 48419 }, { "epoch": 14.9, "learning_rate": 3.229712219348479e-06, "loss": 0.0093, "step": 48420 }, { "epoch": 14.9, "learning_rate": 3.2293455226507474e-06, "loss": 0.0217, "step": 48421 }, { "epoch": 14.9, "learning_rate": 3.2289788427627265e-06, "loss": 0.0173, "step": 48422 }, { "epoch": 14.9, "learning_rate": 3.2286121796853264e-06, "loss": 0.01, "step": 48423 }, { "epoch": 14.9, "learning_rate": 3.228245533419453e-06, "loss": 0.0205, "step": 48424 }, { "epoch": 14.9, "learning_rate": 3.227878903966023e-06, "loss": 0.0185, "step": 48425 }, { "epoch": 14.9, "learning_rate": 3.2275122913259436e-06, "loss": 0.0175, "step": 48426 }, { "epoch": 14.9, "learning_rate": 3.22714569550013e-06, "loss": 0.0071, "step": 48427 }, { "epoch": 14.9, "learning_rate": 3.2267791164894836e-06, "loss": 0.0227, "step": 48428 }, { "epoch": 14.9, "learning_rate": 3.2264125542949154e-06, "loss": 0.0127, "step": 48429 }, { "epoch": 14.9, "learning_rate": 3.2260460089173428e-06, "loss": 0.005, "step": 48430 }, { "epoch": 14.9, "learning_rate": 3.2256794803576707e-06, "loss": 0.0081, "step": 48431 }, { "epoch": 14.9, "learning_rate": 3.2253129686168105e-06, "loss": 0.0268, "step": 48432 }, { "epoch": 14.9, "learning_rate": 3.2249464736956726e-06, "loss": 0.012, "step": 48433 }, { "epoch": 14.9, "learning_rate": 3.224579995595165e-06, "loss": 0.0137, "step": 48434 }, { "epoch": 14.9, "learning_rate": 3.2242135343161997e-06, "loss": 0.0112, "step": 48435 }, { "epoch": 14.9, "learning_rate": 3.2238470898596842e-06, "loss": 0.0137, "step": 48436 }, { "epoch": 14.9, "learning_rate": 3.22348066222653e-06, "loss": 0.0185, "step": 48437 }, { "epoch": 14.9, "learning_rate": 3.223114251417647e-06, "loss": 0.0123, "step": 48438 }, { "epoch": 14.9, "learning_rate": 3.222747857433941e-06, "loss": 0.0177, "step": 48439 }, { "epoch": 14.9, "learning_rate": 3.222381480276331e-06, "loss": 0.0068, "step": 48440 }, { "epoch": 14.9, "learning_rate": 3.2220151199457183e-06, "loss": 0.012, "step": 48441 }, { "epoch": 14.91, "learning_rate": 3.2216487764430136e-06, "loss": 0.0138, "step": 48442 }, { "epoch": 14.91, "learning_rate": 3.2212824497691285e-06, "loss": 0.0169, "step": 48443 }, { "epoch": 14.91, "learning_rate": 3.2209161399249677e-06, "loss": 0.0074, "step": 48444 }, { "epoch": 14.91, "learning_rate": 3.220549846911448e-06, "loss": 0.0156, "step": 48445 }, { "epoch": 14.91, "learning_rate": 3.220183570729475e-06, "loss": 0.0199, "step": 48446 }, { "epoch": 14.91, "learning_rate": 3.219817311379959e-06, "loss": 0.0099, "step": 48447 }, { "epoch": 14.91, "learning_rate": 3.2194510688638116e-06, "loss": 0.0076, "step": 48448 }, { "epoch": 14.91, "learning_rate": 3.2190848431819333e-06, "loss": 0.0083, "step": 48449 }, { "epoch": 14.91, "learning_rate": 3.2187186343352415e-06, "loss": 0.0101, "step": 48450 }, { "epoch": 14.91, "learning_rate": 3.2183524423246436e-06, "loss": 0.0164, "step": 48451 }, { "epoch": 14.91, "learning_rate": 3.2179862671510486e-06, "loss": 0.0098, "step": 48452 }, { "epoch": 14.91, "learning_rate": 3.2176201088153658e-06, "loss": 0.0225, "step": 48453 }, { "epoch": 14.91, "learning_rate": 3.217253967318503e-06, "loss": 0.0129, "step": 48454 }, { "epoch": 14.91, "learning_rate": 3.2168878426613703e-06, "loss": 0.0345, "step": 48455 }, { "epoch": 14.91, "learning_rate": 3.2165217348448763e-06, "loss": 0.0107, "step": 48456 }, { "epoch": 14.91, "learning_rate": 3.2161556438699303e-06, "loss": 0.0092, "step": 48457 }, { "epoch": 14.91, "learning_rate": 3.215789569737441e-06, "loss": 0.0071, "step": 48458 }, { "epoch": 14.91, "learning_rate": 3.2154235124483137e-06, "loss": 0.0118, "step": 48459 }, { "epoch": 14.91, "learning_rate": 3.2150574720034633e-06, "loss": 0.008, "step": 48460 }, { "epoch": 14.91, "learning_rate": 3.2146914484038007e-06, "loss": 0.0126, "step": 48461 }, { "epoch": 14.91, "learning_rate": 3.214325441650226e-06, "loss": 0.0177, "step": 48462 }, { "epoch": 14.91, "learning_rate": 3.2139594517436513e-06, "loss": 0.0206, "step": 48463 }, { "epoch": 14.91, "learning_rate": 3.213593478684983e-06, "loss": 0.0175, "step": 48464 }, { "epoch": 14.91, "learning_rate": 3.2132275224751365e-06, "loss": 0.0121, "step": 48465 }, { "epoch": 14.91, "learning_rate": 3.2128615831150155e-06, "loss": 0.01, "step": 48466 }, { "epoch": 14.91, "learning_rate": 3.2124956606055304e-06, "loss": 0.0059, "step": 48467 }, { "epoch": 14.91, "learning_rate": 3.212129754947587e-06, "loss": 0.0078, "step": 48468 }, { "epoch": 14.91, "learning_rate": 3.211763866142097e-06, "loss": 0.0133, "step": 48469 }, { "epoch": 14.91, "learning_rate": 3.2113979941899666e-06, "loss": 0.0101, "step": 48470 }, { "epoch": 14.91, "learning_rate": 3.211032139092105e-06, "loss": 0.0078, "step": 48471 }, { "epoch": 14.91, "learning_rate": 3.210666300849421e-06, "loss": 0.0175, "step": 48472 }, { "epoch": 14.91, "learning_rate": 3.210300479462821e-06, "loss": 0.01, "step": 48473 }, { "epoch": 14.91, "learning_rate": 3.209934674933216e-06, "loss": 0.0078, "step": 48474 }, { "epoch": 14.92, "learning_rate": 3.2095688872615117e-06, "loss": 0.0126, "step": 48475 }, { "epoch": 14.92, "learning_rate": 3.209203116448617e-06, "loss": 0.0111, "step": 48476 }, { "epoch": 14.92, "learning_rate": 3.2088373624954405e-06, "loss": 0.0153, "step": 48477 }, { "epoch": 14.92, "learning_rate": 3.20847162540289e-06, "loss": 0.0174, "step": 48478 }, { "epoch": 14.92, "learning_rate": 3.2081059051718697e-06, "loss": 0.0138, "step": 48479 }, { "epoch": 14.92, "learning_rate": 3.2077402018032954e-06, "loss": 0.0259, "step": 48480 }, { "epoch": 14.92, "learning_rate": 3.2073745152980704e-06, "loss": 0.0152, "step": 48481 }, { "epoch": 14.92, "learning_rate": 3.207008845657108e-06, "loss": 0.0098, "step": 48482 }, { "epoch": 14.92, "learning_rate": 3.2066431928813068e-06, "loss": 0.0149, "step": 48483 }, { "epoch": 14.92, "learning_rate": 3.206277556971575e-06, "loss": 0.0083, "step": 48484 }, { "epoch": 14.92, "learning_rate": 3.2059119379288293e-06, "loss": 0.017, "step": 48485 }, { "epoch": 14.92, "learning_rate": 3.205546335753973e-06, "loss": 0.0123, "step": 48486 }, { "epoch": 14.92, "learning_rate": 3.2051807504479115e-06, "loss": 0.0123, "step": 48487 }, { "epoch": 14.92, "learning_rate": 3.2048151820115557e-06, "loss": 0.0069, "step": 48488 }, { "epoch": 14.92, "learning_rate": 3.2044496304458117e-06, "loss": 0.0094, "step": 48489 }, { "epoch": 14.92, "learning_rate": 3.204084095751587e-06, "loss": 0.0135, "step": 48490 }, { "epoch": 14.92, "learning_rate": 3.2037185779297896e-06, "loss": 0.0152, "step": 48491 }, { "epoch": 14.92, "learning_rate": 3.203353076981327e-06, "loss": 0.0109, "step": 48492 }, { "epoch": 14.92, "learning_rate": 3.202987592907106e-06, "loss": 0.0067, "step": 48493 }, { "epoch": 14.92, "learning_rate": 3.2026221257080316e-06, "loss": 0.0101, "step": 48494 }, { "epoch": 14.92, "learning_rate": 3.202256675385019e-06, "loss": 0.0198, "step": 48495 }, { "epoch": 14.92, "learning_rate": 3.201891241938969e-06, "loss": 0.0098, "step": 48496 }, { "epoch": 14.92, "learning_rate": 3.2015258253707893e-06, "loss": 0.0224, "step": 48497 }, { "epoch": 14.92, "learning_rate": 3.2011604256813887e-06, "loss": 0.0152, "step": 48498 }, { "epoch": 14.92, "learning_rate": 3.2007950428716705e-06, "loss": 0.0171, "step": 48499 }, { "epoch": 14.92, "learning_rate": 3.200429676942548e-06, "loss": 0.0093, "step": 48500 }, { "epoch": 14.92, "learning_rate": 3.2000643278949263e-06, "loss": 0.0239, "step": 48501 }, { "epoch": 14.92, "learning_rate": 3.1996989957297107e-06, "loss": 0.0114, "step": 48502 }, { "epoch": 14.92, "learning_rate": 3.1993336804478126e-06, "loss": 0.0208, "step": 48503 }, { "epoch": 14.92, "learning_rate": 3.198968382050129e-06, "loss": 0.0126, "step": 48504 }, { "epoch": 14.92, "learning_rate": 3.198603100537577e-06, "loss": 0.013, "step": 48505 }, { "epoch": 14.92, "learning_rate": 3.198237835911061e-06, "loss": 0.0094, "step": 48506 }, { "epoch": 14.93, "learning_rate": 3.1978725881714856e-06, "loss": 0.0194, "step": 48507 }, { "epoch": 14.93, "learning_rate": 3.197507357319759e-06, "loss": 0.0085, "step": 48508 }, { "epoch": 14.93, "learning_rate": 3.197142143356787e-06, "loss": 0.0139, "step": 48509 }, { "epoch": 14.93, "learning_rate": 3.1967769462834774e-06, "loss": 0.0083, "step": 48510 }, { "epoch": 14.93, "learning_rate": 3.1964117661007378e-06, "loss": 0.0117, "step": 48511 }, { "epoch": 14.93, "learning_rate": 3.1960466028094715e-06, "loss": 0.0096, "step": 48512 }, { "epoch": 14.93, "learning_rate": 3.195681456410589e-06, "loss": 0.0072, "step": 48513 }, { "epoch": 14.93, "learning_rate": 3.1953163269049915e-06, "loss": 0.0091, "step": 48514 }, { "epoch": 14.93, "learning_rate": 3.194951214293591e-06, "loss": 0.0128, "step": 48515 }, { "epoch": 14.93, "learning_rate": 3.1945861185772963e-06, "loss": 0.0121, "step": 48516 }, { "epoch": 14.93, "learning_rate": 3.194221039757006e-06, "loss": 0.0067, "step": 48517 }, { "epoch": 14.93, "learning_rate": 3.193855977833631e-06, "loss": 0.0091, "step": 48518 }, { "epoch": 14.93, "learning_rate": 3.193490932808072e-06, "loss": 0.0128, "step": 48519 }, { "epoch": 14.93, "learning_rate": 3.193125904681245e-06, "loss": 0.0192, "step": 48520 }, { "epoch": 14.93, "learning_rate": 3.19276089345405e-06, "loss": 0.0102, "step": 48521 }, { "epoch": 14.93, "learning_rate": 3.192395899127395e-06, "loss": 0.0178, "step": 48522 }, { "epoch": 14.93, "learning_rate": 3.192030921702186e-06, "loss": 0.0091, "step": 48523 }, { "epoch": 14.93, "learning_rate": 3.191665961179328e-06, "loss": 0.0111, "step": 48524 }, { "epoch": 14.93, "learning_rate": 3.1913010175597283e-06, "loss": 0.0077, "step": 48525 }, { "epoch": 14.93, "learning_rate": 3.1909360908442933e-06, "loss": 0.013, "step": 48526 }, { "epoch": 14.93, "learning_rate": 3.190571181033928e-06, "loss": 0.0151, "step": 48527 }, { "epoch": 14.93, "learning_rate": 3.1902062881295383e-06, "loss": 0.0101, "step": 48528 }, { "epoch": 14.93, "learning_rate": 3.1898414121320277e-06, "loss": 0.0162, "step": 48529 }, { "epoch": 14.93, "learning_rate": 3.18947655304231e-06, "loss": 0.0092, "step": 48530 }, { "epoch": 14.93, "learning_rate": 3.1891117108612847e-06, "loss": 0.0112, "step": 48531 }, { "epoch": 14.93, "learning_rate": 3.1887468855898574e-06, "loss": 0.0145, "step": 48532 }, { "epoch": 14.93, "learning_rate": 3.1883820772289364e-06, "loss": 0.0076, "step": 48533 }, { "epoch": 14.93, "learning_rate": 3.188017285779422e-06, "loss": 0.0166, "step": 48534 }, { "epoch": 14.93, "learning_rate": 3.1876525112422283e-06, "loss": 0.0142, "step": 48535 }, { "epoch": 14.93, "learning_rate": 3.187287753618257e-06, "loss": 0.0157, "step": 48536 }, { "epoch": 14.93, "learning_rate": 3.186923012908416e-06, "loss": 0.0071, "step": 48537 }, { "epoch": 14.93, "learning_rate": 3.186558289113606e-06, "loss": 0.0065, "step": 48538 }, { "epoch": 14.93, "learning_rate": 3.1861935822347314e-06, "loss": 0.0163, "step": 48539 }, { "epoch": 14.94, "learning_rate": 3.185828892272704e-06, "loss": 0.0111, "step": 48540 }, { "epoch": 14.94, "learning_rate": 3.1854642192284267e-06, "loss": 0.0136, "step": 48541 }, { "epoch": 14.94, "learning_rate": 3.185099563102805e-06, "loss": 0.0135, "step": 48542 }, { "epoch": 14.94, "learning_rate": 3.184734923896744e-06, "loss": 0.0084, "step": 48543 }, { "epoch": 14.94, "learning_rate": 3.1843703016111484e-06, "loss": 0.0063, "step": 48544 }, { "epoch": 14.94, "learning_rate": 3.1840056962469235e-06, "loss": 0.0099, "step": 48545 }, { "epoch": 14.94, "learning_rate": 3.1836411078049757e-06, "loss": 0.0184, "step": 48546 }, { "epoch": 14.94, "learning_rate": 3.1832765362862083e-06, "loss": 0.0064, "step": 48547 }, { "epoch": 14.94, "learning_rate": 3.1829119816915276e-06, "loss": 0.018, "step": 48548 }, { "epoch": 14.94, "learning_rate": 3.1825474440218363e-06, "loss": 0.0082, "step": 48549 }, { "epoch": 14.94, "learning_rate": 3.1821829232780455e-06, "loss": 0.0092, "step": 48550 }, { "epoch": 14.94, "learning_rate": 3.181818419461058e-06, "loss": 0.0106, "step": 48551 }, { "epoch": 14.94, "learning_rate": 3.1814539325717753e-06, "loss": 0.0091, "step": 48552 }, { "epoch": 14.94, "learning_rate": 3.1810894626111042e-06, "loss": 0.0178, "step": 48553 }, { "epoch": 14.94, "learning_rate": 3.180725009579946e-06, "loss": 0.0298, "step": 48554 }, { "epoch": 14.94, "learning_rate": 3.180360573479212e-06, "loss": 0.0079, "step": 48555 }, { "epoch": 14.94, "learning_rate": 3.1799961543098047e-06, "loss": 0.0082, "step": 48556 }, { "epoch": 14.94, "learning_rate": 3.1796317520726284e-06, "loss": 0.0093, "step": 48557 }, { "epoch": 14.94, "learning_rate": 3.1792673667685915e-06, "loss": 0.0114, "step": 48558 }, { "epoch": 14.94, "learning_rate": 3.1789029983985877e-06, "loss": 0.014, "step": 48559 }, { "epoch": 14.94, "learning_rate": 3.1785386469635317e-06, "loss": 0.0156, "step": 48560 }, { "epoch": 14.94, "learning_rate": 3.178174312464326e-06, "loss": 0.0145, "step": 48561 }, { "epoch": 14.94, "learning_rate": 3.1778099949018736e-06, "loss": 0.0177, "step": 48562 }, { "epoch": 14.94, "learning_rate": 3.1774456942770804e-06, "loss": 0.011, "step": 48563 }, { "epoch": 14.94, "learning_rate": 3.1770814105908497e-06, "loss": 0.0148, "step": 48564 }, { "epoch": 14.94, "learning_rate": 3.176717143844088e-06, "loss": 0.0181, "step": 48565 }, { "epoch": 14.94, "learning_rate": 3.1763528940376966e-06, "loss": 0.0253, "step": 48566 }, { "epoch": 14.94, "learning_rate": 3.175988661172582e-06, "loss": 0.0071, "step": 48567 }, { "epoch": 14.94, "learning_rate": 3.1756244452496475e-06, "loss": 0.0211, "step": 48568 }, { "epoch": 14.94, "learning_rate": 3.1752602462697947e-06, "loss": 0.0096, "step": 48569 }, { "epoch": 14.94, "learning_rate": 3.174896064233933e-06, "loss": 0.0125, "step": 48570 }, { "epoch": 14.94, "learning_rate": 3.1745318991429665e-06, "loss": 0.0125, "step": 48571 }, { "epoch": 14.95, "learning_rate": 3.1741677509977986e-06, "loss": 0.0175, "step": 48572 }, { "epoch": 14.95, "learning_rate": 3.1738036197993294e-06, "loss": 0.0091, "step": 48573 }, { "epoch": 14.95, "learning_rate": 3.1734395055484623e-06, "loss": 0.0134, "step": 48574 }, { "epoch": 14.95, "learning_rate": 3.173075408246108e-06, "loss": 0.0202, "step": 48575 }, { "epoch": 14.95, "learning_rate": 3.1727113278931664e-06, "loss": 0.0172, "step": 48576 }, { "epoch": 14.95, "learning_rate": 3.172347264490543e-06, "loss": 0.0235, "step": 48577 }, { "epoch": 14.95, "learning_rate": 3.1719832180391395e-06, "loss": 0.009, "step": 48578 }, { "epoch": 14.95, "learning_rate": 3.1716191885398615e-06, "loss": 0.0153, "step": 48579 }, { "epoch": 14.95, "learning_rate": 3.1712551759936126e-06, "loss": 0.0132, "step": 48580 }, { "epoch": 14.95, "learning_rate": 3.1708911804012956e-06, "loss": 0.0166, "step": 48581 }, { "epoch": 14.95, "learning_rate": 3.1705272017638144e-06, "loss": 0.0146, "step": 48582 }, { "epoch": 14.95, "learning_rate": 3.1701632400820735e-06, "loss": 0.0104, "step": 48583 }, { "epoch": 14.95, "learning_rate": 3.169799295356972e-06, "loss": 0.0141, "step": 48584 }, { "epoch": 14.95, "learning_rate": 3.169435367589425e-06, "loss": 0.0074, "step": 48585 }, { "epoch": 14.95, "learning_rate": 3.169071456780325e-06, "loss": 0.0116, "step": 48586 }, { "epoch": 14.95, "learning_rate": 3.1687075629305787e-06, "loss": 0.0088, "step": 48587 }, { "epoch": 14.95, "learning_rate": 3.16834368604109e-06, "loss": 0.006, "step": 48588 }, { "epoch": 14.95, "learning_rate": 3.1679798261127594e-06, "loss": 0.0169, "step": 48589 }, { "epoch": 14.95, "learning_rate": 3.167615983146496e-06, "loss": 0.0146, "step": 48590 }, { "epoch": 14.95, "learning_rate": 3.167252157143199e-06, "loss": 0.0145, "step": 48591 }, { "epoch": 14.95, "learning_rate": 3.1668883481037736e-06, "loss": 0.0092, "step": 48592 }, { "epoch": 14.95, "learning_rate": 3.1665245560291267e-06, "loss": 0.0121, "step": 48593 }, { "epoch": 14.95, "learning_rate": 3.1661607809201487e-06, "loss": 0.0103, "step": 48594 }, { "epoch": 14.95, "learning_rate": 3.1657970227777556e-06, "loss": 0.0074, "step": 48595 }, { "epoch": 14.95, "learning_rate": 3.1654332816028466e-06, "loss": 0.012, "step": 48596 }, { "epoch": 14.95, "learning_rate": 3.165069557396323e-06, "loss": 0.0133, "step": 48597 }, { "epoch": 14.95, "learning_rate": 3.16470585015909e-06, "loss": 0.0148, "step": 48598 }, { "epoch": 14.95, "learning_rate": 3.1643421598920486e-06, "loss": 0.0084, "step": 48599 }, { "epoch": 14.95, "learning_rate": 3.163978486596103e-06, "loss": 0.0176, "step": 48600 }, { "epoch": 14.95, "learning_rate": 3.163614830272156e-06, "loss": 0.0055, "step": 48601 }, { "epoch": 14.95, "learning_rate": 3.1632511909211107e-06, "loss": 0.0174, "step": 48602 }, { "epoch": 14.95, "learning_rate": 3.1628875685438697e-06, "loss": 0.013, "step": 48603 }, { "epoch": 14.95, "learning_rate": 3.1625239631413317e-06, "loss": 0.0184, "step": 48604 }, { "epoch": 14.96, "learning_rate": 3.162160374714407e-06, "loss": 0.0124, "step": 48605 }, { "epoch": 14.96, "learning_rate": 3.161796803263998e-06, "loss": 0.019, "step": 48606 }, { "epoch": 14.96, "learning_rate": 3.1614332487910003e-06, "loss": 0.0101, "step": 48607 }, { "epoch": 14.96, "learning_rate": 3.1610697112963205e-06, "loss": 0.009, "step": 48608 }, { "epoch": 14.96, "learning_rate": 3.160706190780858e-06, "loss": 0.0137, "step": 48609 }, { "epoch": 14.96, "learning_rate": 3.1603426872455214e-06, "loss": 0.0138, "step": 48610 }, { "epoch": 14.96, "learning_rate": 3.15997920069121e-06, "loss": 0.0094, "step": 48611 }, { "epoch": 14.96, "learning_rate": 3.159615731118826e-06, "loss": 0.0141, "step": 48612 }, { "epoch": 14.96, "learning_rate": 3.1592522785292714e-06, "loss": 0.0122, "step": 48613 }, { "epoch": 14.96, "learning_rate": 3.15888884292345e-06, "loss": 0.0112, "step": 48614 }, { "epoch": 14.96, "learning_rate": 3.1585254243022635e-06, "loss": 0.0149, "step": 48615 }, { "epoch": 14.96, "learning_rate": 3.158162022666613e-06, "loss": 0.0148, "step": 48616 }, { "epoch": 14.96, "learning_rate": 3.157798638017402e-06, "loss": 0.0133, "step": 48617 }, { "epoch": 14.96, "learning_rate": 3.1574352703555335e-06, "loss": 0.0183, "step": 48618 }, { "epoch": 14.96, "learning_rate": 3.1570719196819065e-06, "loss": 0.0131, "step": 48619 }, { "epoch": 14.96, "learning_rate": 3.1567085859974265e-06, "loss": 0.0117, "step": 48620 }, { "epoch": 14.96, "learning_rate": 3.156345269302994e-06, "loss": 0.0075, "step": 48621 }, { "epoch": 14.96, "learning_rate": 3.1559819695995118e-06, "loss": 0.0115, "step": 48622 }, { "epoch": 14.96, "learning_rate": 3.1556186868878804e-06, "loss": 0.0108, "step": 48623 }, { "epoch": 14.96, "learning_rate": 3.155255421169e-06, "loss": 0.008, "step": 48624 }, { "epoch": 14.96, "learning_rate": 3.1548921724437775e-06, "loss": 0.0077, "step": 48625 }, { "epoch": 14.96, "learning_rate": 3.1545289407131128e-06, "loss": 0.008, "step": 48626 }, { "epoch": 14.96, "learning_rate": 3.15416572597791e-06, "loss": 0.0154, "step": 48627 }, { "epoch": 14.96, "learning_rate": 3.153802528239066e-06, "loss": 0.0148, "step": 48628 }, { "epoch": 14.96, "learning_rate": 3.1534393474974814e-06, "loss": 0.0067, "step": 48629 }, { "epoch": 14.96, "learning_rate": 3.1530761837540635e-06, "loss": 0.0131, "step": 48630 }, { "epoch": 14.96, "learning_rate": 3.1527130370097113e-06, "loss": 0.0068, "step": 48631 }, { "epoch": 14.96, "learning_rate": 3.1523499072653273e-06, "loss": 0.0213, "step": 48632 }, { "epoch": 14.96, "learning_rate": 3.151986794521812e-06, "loss": 0.0077, "step": 48633 }, { "epoch": 14.96, "learning_rate": 3.1516236987800684e-06, "loss": 0.0118, "step": 48634 }, { "epoch": 14.96, "learning_rate": 3.1512606200409956e-06, "loss": 0.0124, "step": 48635 }, { "epoch": 14.96, "learning_rate": 3.150897558305497e-06, "loss": 0.0146, "step": 48636 }, { "epoch": 14.97, "learning_rate": 3.1505345135744737e-06, "loss": 0.0191, "step": 48637 }, { "epoch": 14.97, "learning_rate": 3.1501714858488254e-06, "loss": 0.0118, "step": 48638 }, { "epoch": 14.97, "learning_rate": 3.1498084751294523e-06, "loss": 0.0077, "step": 48639 }, { "epoch": 14.97, "learning_rate": 3.1494454814172637e-06, "loss": 0.0163, "step": 48640 }, { "epoch": 14.97, "learning_rate": 3.1490825047131523e-06, "loss": 0.0139, "step": 48641 }, { "epoch": 14.97, "learning_rate": 3.148719545018023e-06, "loss": 0.0063, "step": 48642 }, { "epoch": 14.97, "learning_rate": 3.148356602332775e-06, "loss": 0.0126, "step": 48643 }, { "epoch": 14.97, "learning_rate": 3.1479936766583073e-06, "loss": 0.0086, "step": 48644 }, { "epoch": 14.97, "learning_rate": 3.1476307679955276e-06, "loss": 0.0094, "step": 48645 }, { "epoch": 14.97, "learning_rate": 3.1472678763453335e-06, "loss": 0.008, "step": 48646 }, { "epoch": 14.97, "learning_rate": 3.1469050017086254e-06, "loss": 0.0076, "step": 48647 }, { "epoch": 14.97, "learning_rate": 3.1465421440863074e-06, "loss": 0.0115, "step": 48648 }, { "epoch": 14.97, "learning_rate": 3.146179303479272e-06, "loss": 0.0231, "step": 48649 }, { "epoch": 14.97, "learning_rate": 3.1458164798884285e-06, "loss": 0.0132, "step": 48650 }, { "epoch": 14.97, "learning_rate": 3.1454536733146755e-06, "loss": 0.0098, "step": 48651 }, { "epoch": 14.97, "learning_rate": 3.145090883758912e-06, "loss": 0.0135, "step": 48652 }, { "epoch": 14.97, "learning_rate": 3.1447281112220405e-06, "loss": 0.0138, "step": 48653 }, { "epoch": 14.97, "learning_rate": 3.1443653557049615e-06, "loss": 0.0212, "step": 48654 }, { "epoch": 14.97, "learning_rate": 3.1440026172085738e-06, "loss": 0.007, "step": 48655 }, { "epoch": 14.97, "learning_rate": 3.14363989573378e-06, "loss": 0.0163, "step": 48656 }, { "epoch": 14.97, "learning_rate": 3.1432771912814796e-06, "loss": 0.0239, "step": 48657 }, { "epoch": 14.97, "learning_rate": 3.1429145038525744e-06, "loss": 0.0077, "step": 48658 }, { "epoch": 14.97, "learning_rate": 3.14255183344796e-06, "loss": 0.0079, "step": 48659 }, { "epoch": 14.97, "learning_rate": 3.1421891800685435e-06, "loss": 0.0293, "step": 48660 }, { "epoch": 14.97, "learning_rate": 3.141826543715226e-06, "loss": 0.0153, "step": 48661 }, { "epoch": 14.97, "learning_rate": 3.141463924388901e-06, "loss": 0.0091, "step": 48662 }, { "epoch": 14.97, "learning_rate": 3.141101322090472e-06, "loss": 0.0117, "step": 48663 }, { "epoch": 14.97, "learning_rate": 3.1407387368208365e-06, "loss": 0.0176, "step": 48664 }, { "epoch": 14.97, "learning_rate": 3.1403761685809007e-06, "loss": 0.0083, "step": 48665 }, { "epoch": 14.97, "learning_rate": 3.1400136173715613e-06, "loss": 0.0119, "step": 48666 }, { "epoch": 14.97, "learning_rate": 3.139651083193719e-06, "loss": 0.0172, "step": 48667 }, { "epoch": 14.97, "learning_rate": 3.1392885660482732e-06, "loss": 0.0101, "step": 48668 }, { "epoch": 14.97, "learning_rate": 3.1389260659361244e-06, "loss": 0.0137, "step": 48669 }, { "epoch": 14.98, "learning_rate": 3.138563582858173e-06, "loss": 0.0077, "step": 48670 }, { "epoch": 14.98, "learning_rate": 3.138201116815317e-06, "loss": 0.0147, "step": 48671 }, { "epoch": 14.98, "learning_rate": 3.1378386678084595e-06, "loss": 0.0158, "step": 48672 }, { "epoch": 14.98, "learning_rate": 3.1374762358384935e-06, "loss": 0.0286, "step": 48673 }, { "epoch": 14.98, "learning_rate": 3.137113820906332e-06, "loss": 0.0197, "step": 48674 }, { "epoch": 14.98, "learning_rate": 3.1367514230128625e-06, "loss": 0.0059, "step": 48675 }, { "epoch": 14.98, "learning_rate": 3.1363890421589894e-06, "loss": 0.0123, "step": 48676 }, { "epoch": 14.98, "learning_rate": 3.136026678345613e-06, "loss": 0.01, "step": 48677 }, { "epoch": 14.98, "learning_rate": 3.135664331573627e-06, "loss": 0.0099, "step": 48678 }, { "epoch": 14.98, "learning_rate": 3.1353020018439394e-06, "loss": 0.0164, "step": 48679 }, { "epoch": 14.98, "learning_rate": 3.134939689157447e-06, "loss": 0.0194, "step": 48680 }, { "epoch": 14.98, "learning_rate": 3.134577393515048e-06, "loss": 0.0071, "step": 48681 }, { "epoch": 14.98, "learning_rate": 3.134215114917646e-06, "loss": 0.0087, "step": 48682 }, { "epoch": 14.98, "learning_rate": 3.1338528533661306e-06, "loss": 0.0074, "step": 48683 }, { "epoch": 14.98, "learning_rate": 3.1334906088614104e-06, "loss": 0.0182, "step": 48684 }, { "epoch": 14.98, "learning_rate": 3.133128381404381e-06, "loss": 0.0097, "step": 48685 }, { "epoch": 14.98, "learning_rate": 3.1327661709959432e-06, "loss": 0.0153, "step": 48686 }, { "epoch": 14.98, "learning_rate": 3.1324039776369965e-06, "loss": 0.0097, "step": 48687 }, { "epoch": 14.98, "learning_rate": 3.1320418013284393e-06, "loss": 0.0087, "step": 48688 }, { "epoch": 14.98, "learning_rate": 3.131679642071169e-06, "loss": 0.0124, "step": 48689 }, { "epoch": 14.98, "learning_rate": 3.131317499866089e-06, "loss": 0.0095, "step": 48690 }, { "epoch": 14.98, "learning_rate": 3.130955374714094e-06, "loss": 0.0185, "step": 48691 }, { "epoch": 14.98, "learning_rate": 3.1305932666160853e-06, "loss": 0.0058, "step": 48692 }, { "epoch": 14.98, "learning_rate": 3.1302311755729587e-06, "loss": 0.0102, "step": 48693 }, { "epoch": 14.98, "learning_rate": 3.12986910158562e-06, "loss": 0.0103, "step": 48694 }, { "epoch": 14.98, "learning_rate": 3.1295070446549657e-06, "loss": 0.0131, "step": 48695 }, { "epoch": 14.98, "learning_rate": 3.1291450047818915e-06, "loss": 0.0148, "step": 48696 }, { "epoch": 14.98, "learning_rate": 3.1287829819672964e-06, "loss": 0.0189, "step": 48697 }, { "epoch": 14.98, "learning_rate": 3.1284209762120787e-06, "loss": 0.0139, "step": 48698 }, { "epoch": 14.98, "learning_rate": 3.1280589875171417e-06, "loss": 0.0084, "step": 48699 }, { "epoch": 14.98, "learning_rate": 3.127697015883383e-06, "loss": 0.0138, "step": 48700 }, { "epoch": 14.98, "learning_rate": 3.127335061311698e-06, "loss": 0.0223, "step": 48701 }, { "epoch": 14.99, "learning_rate": 3.126973123802989e-06, "loss": 0.0121, "step": 48702 }, { "epoch": 14.99, "learning_rate": 3.126611203358152e-06, "loss": 0.0081, "step": 48703 }, { "epoch": 14.99, "learning_rate": 3.126249299978086e-06, "loss": 0.0142, "step": 48704 }, { "epoch": 14.99, "learning_rate": 3.125887413663691e-06, "loss": 0.0304, "step": 48705 }, { "epoch": 14.99, "learning_rate": 3.1255255444158626e-06, "loss": 0.0103, "step": 48706 }, { "epoch": 14.99, "learning_rate": 3.125163692235502e-06, "loss": 0.0094, "step": 48707 }, { "epoch": 14.99, "learning_rate": 3.1248018571235037e-06, "loss": 0.022, "step": 48708 }, { "epoch": 14.99, "learning_rate": 3.124440039080775e-06, "loss": 0.0083, "step": 48709 }, { "epoch": 14.99, "learning_rate": 3.1240782381082046e-06, "loss": 0.0173, "step": 48710 }, { "epoch": 14.99, "learning_rate": 3.1237164542066943e-06, "loss": 0.0072, "step": 48711 }, { "epoch": 14.99, "learning_rate": 3.1233546873771424e-06, "loss": 0.0093, "step": 48712 }, { "epoch": 14.99, "learning_rate": 3.1229929376204437e-06, "loss": 0.0133, "step": 48713 }, { "epoch": 14.99, "learning_rate": 3.122631204937503e-06, "loss": 0.0069, "step": 48714 }, { "epoch": 14.99, "learning_rate": 3.122269489329215e-06, "loss": 0.0143, "step": 48715 }, { "epoch": 14.99, "learning_rate": 3.12190779079648e-06, "loss": 0.0075, "step": 48716 }, { "epoch": 14.99, "learning_rate": 3.1215461093401913e-06, "loss": 0.0104, "step": 48717 }, { "epoch": 14.99, "learning_rate": 3.121184444961246e-06, "loss": 0.0085, "step": 48718 }, { "epoch": 14.99, "learning_rate": 3.1208227976605475e-06, "loss": 0.0121, "step": 48719 }, { "epoch": 14.99, "learning_rate": 3.120461167438993e-06, "loss": 0.0144, "step": 48720 }, { "epoch": 14.99, "learning_rate": 3.120099554297478e-06, "loss": 0.0188, "step": 48721 }, { "epoch": 14.99, "learning_rate": 3.1197379582369e-06, "loss": 0.0134, "step": 48722 }, { "epoch": 14.99, "learning_rate": 3.11937637925816e-06, "loss": 0.0116, "step": 48723 }, { "epoch": 14.99, "learning_rate": 3.1190148173621515e-06, "loss": 0.014, "step": 48724 }, { "epoch": 14.99, "learning_rate": 3.1186532725497754e-06, "loss": 0.0171, "step": 48725 }, { "epoch": 14.99, "learning_rate": 3.1182917448219273e-06, "loss": 0.0159, "step": 48726 }, { "epoch": 14.99, "learning_rate": 3.1179302341795058e-06, "loss": 0.0127, "step": 48727 }, { "epoch": 14.99, "learning_rate": 3.1175687406234045e-06, "loss": 0.0237, "step": 48728 }, { "epoch": 14.99, "learning_rate": 3.1172072641545294e-06, "loss": 0.0099, "step": 48729 }, { "epoch": 14.99, "learning_rate": 3.116845804773775e-06, "loss": 0.0166, "step": 48730 }, { "epoch": 14.99, "learning_rate": 3.116484362482034e-06, "loss": 0.0116, "step": 48731 }, { "epoch": 14.99, "learning_rate": 3.116122937280206e-06, "loss": 0.0131, "step": 48732 }, { "epoch": 14.99, "learning_rate": 3.1157615291691855e-06, "loss": 0.0204, "step": 48733 }, { "epoch": 14.99, "learning_rate": 3.1154001381498775e-06, "loss": 0.0109, "step": 48734 }, { "epoch": 15.0, "learning_rate": 3.115038764223175e-06, "loss": 0.008, "step": 48735 }, { "epoch": 15.0, "learning_rate": 3.1146774073899743e-06, "loss": 0.0081, "step": 48736 }, { "epoch": 15.0, "learning_rate": 3.1143160676511764e-06, "loss": 0.0082, "step": 48737 }, { "epoch": 15.0, "learning_rate": 3.11395474500767e-06, "loss": 0.0114, "step": 48738 }, { "epoch": 15.0, "learning_rate": 3.1135934394603596e-06, "loss": 0.0072, "step": 48739 }, { "epoch": 15.0, "learning_rate": 3.1132321510101406e-06, "loss": 0.0052, "step": 48740 }, { "epoch": 15.0, "learning_rate": 3.1128708796579098e-06, "loss": 0.0107, "step": 48741 }, { "epoch": 15.0, "learning_rate": 3.112509625404564e-06, "loss": 0.008, "step": 48742 }, { "epoch": 15.0, "learning_rate": 3.1121483882509996e-06, "loss": 0.0112, "step": 48743 }, { "epoch": 15.0, "learning_rate": 3.111787168198114e-06, "loss": 0.0246, "step": 48744 }, { "epoch": 15.0, "learning_rate": 3.111425965246804e-06, "loss": 0.006, "step": 48745 }, { "epoch": 15.0, "learning_rate": 3.1110647793979666e-06, "loss": 0.0135, "step": 48746 }, { "epoch": 15.0, "learning_rate": 3.110703610652497e-06, "loss": 0.0096, "step": 48747 }, { "epoch": 15.0, "learning_rate": 3.1103424590112906e-06, "loss": 0.025, "step": 48748 }, { "epoch": 15.0, "learning_rate": 3.1099813244752497e-06, "loss": 0.0163, "step": 48749 }, { "epoch": 15.0, "learning_rate": 3.1096202070452675e-06, "loss": 0.0073, "step": 48750 }, { "epoch": 15.0, "learning_rate": 3.1092591067222444e-06, "loss": 0.0134, "step": 48751 }, { "epoch": 15.0, "learning_rate": 3.1088980235070688e-06, "loss": 0.0067, "step": 48752 }, { "epoch": 15.0, "learning_rate": 3.1085369574006395e-06, "loss": 0.0101, "step": 48753 }, { "epoch": 15.0, "learning_rate": 3.1081759084038577e-06, "loss": 0.0074, "step": 48754 }, { "epoch": 15.0, "learning_rate": 3.1078148765176175e-06, "loss": 0.0078, "step": 48755 }, { "epoch": 15.0, "learning_rate": 3.107453861742815e-06, "loss": 0.0107, "step": 48756 }, { "epoch": 15.0, "learning_rate": 3.1070928640803464e-06, "loss": 0.0083, "step": 48757 }, { "epoch": 15.0, "learning_rate": 3.106731883531108e-06, "loss": 0.009, "step": 48758 }, { "epoch": 15.0, "learning_rate": 3.106370920095996e-06, "loss": 0.0056, "step": 48759 }, { "epoch": 15.0, "learning_rate": 3.1060099737759064e-06, "loss": 0.0123, "step": 48760 }, { "epoch": 15.0, "learning_rate": 3.1056490445717345e-06, "loss": 0.0121, "step": 48761 }, { "epoch": 15.0, "learning_rate": 3.1052881324843786e-06, "loss": 0.0072, "step": 48762 }, { "epoch": 15.0, "learning_rate": 3.1049272375147298e-06, "loss": 0.006, "step": 48763 }, { "epoch": 15.0, "learning_rate": 3.1045663596636943e-06, "loss": 0.0059, "step": 48764 }, { "epoch": 15.0, "learning_rate": 3.104205498932159e-06, "loss": 0.0202, "step": 48765 }, { "epoch": 15.0, "learning_rate": 3.103844655321021e-06, "loss": 0.0067, "step": 48766 }, { "epoch": 15.01, "learning_rate": 3.1034838288311784e-06, "loss": 0.0124, "step": 48767 }, { "epoch": 15.01, "learning_rate": 3.1031230194635233e-06, "loss": 0.0122, "step": 48768 }, { "epoch": 15.01, "learning_rate": 3.1027622272189572e-06, "loss": 0.0114, "step": 48769 }, { "epoch": 15.01, "learning_rate": 3.1024014520983735e-06, "loss": 0.0052, "step": 48770 }, { "epoch": 15.01, "learning_rate": 3.102040694102667e-06, "loss": 0.0077, "step": 48771 }, { "epoch": 15.01, "learning_rate": 3.101679953232737e-06, "loss": 0.0086, "step": 48772 }, { "epoch": 15.01, "learning_rate": 3.10131922948947e-06, "loss": 0.0031, "step": 48773 }, { "epoch": 15.01, "learning_rate": 3.1009585228737705e-06, "loss": 0.0077, "step": 48774 }, { "epoch": 15.01, "learning_rate": 3.1005978333865304e-06, "loss": 0.0087, "step": 48775 }, { "epoch": 15.01, "learning_rate": 3.1002371610286462e-06, "loss": 0.0071, "step": 48776 }, { "epoch": 15.01, "learning_rate": 3.099876505801014e-06, "loss": 0.0074, "step": 48777 }, { "epoch": 15.01, "learning_rate": 3.0995158677045266e-06, "loss": 0.0078, "step": 48778 }, { "epoch": 15.01, "learning_rate": 3.099155246740082e-06, "loss": 0.0085, "step": 48779 }, { "epoch": 15.01, "learning_rate": 3.0987946429085747e-06, "loss": 0.01, "step": 48780 }, { "epoch": 15.01, "learning_rate": 3.0984340562108995e-06, "loss": 0.0142, "step": 48781 }, { "epoch": 15.01, "learning_rate": 3.0980734866479525e-06, "loss": 0.0094, "step": 48782 }, { "epoch": 15.01, "learning_rate": 3.0977129342206234e-06, "loss": 0.006, "step": 48783 }, { "epoch": 15.01, "learning_rate": 3.097352398929817e-06, "loss": 0.0099, "step": 48784 }, { "epoch": 15.01, "learning_rate": 3.0969918807764263e-06, "loss": 0.0065, "step": 48785 }, { "epoch": 15.01, "learning_rate": 3.0966313797613413e-06, "loss": 0.0114, "step": 48786 }, { "epoch": 15.01, "learning_rate": 3.0962708958854593e-06, "loss": 0.0091, "step": 48787 }, { "epoch": 15.01, "learning_rate": 3.095910429149671e-06, "loss": 0.0066, "step": 48788 }, { "epoch": 15.01, "learning_rate": 3.0955499795548803e-06, "loss": 0.0102, "step": 48789 }, { "epoch": 15.01, "learning_rate": 3.0951895471019776e-06, "loss": 0.0095, "step": 48790 }, { "epoch": 15.01, "learning_rate": 3.0948291317918575e-06, "loss": 0.0101, "step": 48791 }, { "epoch": 15.01, "learning_rate": 3.0944687336254154e-06, "loss": 0.0046, "step": 48792 }, { "epoch": 15.01, "learning_rate": 3.0941083526035452e-06, "loss": 0.0088, "step": 48793 }, { "epoch": 15.01, "learning_rate": 3.0937479887271415e-06, "loss": 0.0161, "step": 48794 }, { "epoch": 15.01, "learning_rate": 3.0933876419971008e-06, "loss": 0.0069, "step": 48795 }, { "epoch": 15.01, "learning_rate": 3.0930273124143163e-06, "loss": 0.0064, "step": 48796 }, { "epoch": 15.01, "learning_rate": 3.0926669999796833e-06, "loss": 0.0091, "step": 48797 }, { "epoch": 15.01, "learning_rate": 3.092306704694096e-06, "loss": 0.0095, "step": 48798 }, { "epoch": 15.01, "learning_rate": 3.091946426558449e-06, "loss": 0.009, "step": 48799 }, { "epoch": 15.02, "learning_rate": 3.0915861655736356e-06, "loss": 0.0062, "step": 48800 }, { "epoch": 15.02, "learning_rate": 3.0912259217405516e-06, "loss": 0.0088, "step": 48801 }, { "epoch": 15.02, "learning_rate": 3.0908656950600924e-06, "loss": 0.0043, "step": 48802 }, { "epoch": 15.02, "learning_rate": 3.090505485533146e-06, "loss": 0.0031, "step": 48803 }, { "epoch": 15.02, "learning_rate": 3.090145293160616e-06, "loss": 0.0094, "step": 48804 }, { "epoch": 15.02, "learning_rate": 3.0897851179433924e-06, "loss": 0.0039, "step": 48805 }, { "epoch": 15.02, "learning_rate": 3.089424959882372e-06, "loss": 0.0075, "step": 48806 }, { "epoch": 15.02, "learning_rate": 3.0890648189784445e-06, "loss": 0.0054, "step": 48807 }, { "epoch": 15.02, "learning_rate": 3.0887046952325016e-06, "loss": 0.0059, "step": 48808 }, { "epoch": 15.02, "learning_rate": 3.0883445886454443e-06, "loss": 0.012, "step": 48809 }, { "epoch": 15.02, "learning_rate": 3.0879844992181653e-06, "loss": 0.0084, "step": 48810 }, { "epoch": 15.02, "learning_rate": 3.0876244269515574e-06, "loss": 0.0049, "step": 48811 }, { "epoch": 15.02, "learning_rate": 3.0872643718465147e-06, "loss": 0.0251, "step": 48812 }, { "epoch": 15.02, "learning_rate": 3.0869043339039294e-06, "loss": 0.0072, "step": 48813 }, { "epoch": 15.02, "learning_rate": 3.0865443131246986e-06, "loss": 0.0112, "step": 48814 }, { "epoch": 15.02, "learning_rate": 3.0861843095097134e-06, "loss": 0.0046, "step": 48815 }, { "epoch": 15.02, "learning_rate": 3.0858243230598683e-06, "loss": 0.0057, "step": 48816 }, { "epoch": 15.02, "learning_rate": 3.0854643537760585e-06, "loss": 0.0082, "step": 48817 }, { "epoch": 15.02, "learning_rate": 3.0851044016591734e-06, "loss": 0.0127, "step": 48818 }, { "epoch": 15.02, "learning_rate": 3.0847444667101156e-06, "loss": 0.0101, "step": 48819 }, { "epoch": 15.02, "learning_rate": 3.0843845489297698e-06, "loss": 0.0096, "step": 48820 }, { "epoch": 15.02, "learning_rate": 3.0840246483190338e-06, "loss": 0.0068, "step": 48821 }, { "epoch": 15.02, "learning_rate": 3.0836647648787987e-06, "loss": 0.0104, "step": 48822 }, { "epoch": 15.02, "learning_rate": 3.083304898609957e-06, "loss": 0.0073, "step": 48823 }, { "epoch": 15.02, "learning_rate": 3.0829450495134074e-06, "loss": 0.0067, "step": 48824 }, { "epoch": 15.02, "learning_rate": 3.0825852175900406e-06, "loss": 0.0113, "step": 48825 }, { "epoch": 15.02, "learning_rate": 3.0822254028407504e-06, "loss": 0.011, "step": 48826 }, { "epoch": 15.02, "learning_rate": 3.081865605266432e-06, "loss": 0.0336, "step": 48827 }, { "epoch": 15.02, "learning_rate": 3.0815058248679708e-06, "loss": 0.0049, "step": 48828 }, { "epoch": 15.02, "learning_rate": 3.0811460616462672e-06, "loss": 0.0091, "step": 48829 }, { "epoch": 15.02, "learning_rate": 3.080786315602213e-06, "loss": 0.0083, "step": 48830 }, { "epoch": 15.02, "learning_rate": 3.080426586736702e-06, "loss": 0.0092, "step": 48831 }, { "epoch": 15.03, "learning_rate": 3.0800668750506256e-06, "loss": 0.0086, "step": 48832 }, { "epoch": 15.03, "learning_rate": 3.0797071805448774e-06, "loss": 0.003, "step": 48833 }, { "epoch": 15.03, "learning_rate": 3.0793475032203513e-06, "loss": 0.0107, "step": 48834 }, { "epoch": 15.03, "learning_rate": 3.0789878430779394e-06, "loss": 0.0078, "step": 48835 }, { "epoch": 15.03, "learning_rate": 3.0786282001185353e-06, "loss": 0.0103, "step": 48836 }, { "epoch": 15.03, "learning_rate": 3.0782685743430305e-06, "loss": 0.0097, "step": 48837 }, { "epoch": 15.03, "learning_rate": 3.077908965752315e-06, "loss": 0.0043, "step": 48838 }, { "epoch": 15.03, "learning_rate": 3.0775493743472906e-06, "loss": 0.0122, "step": 48839 }, { "epoch": 15.03, "learning_rate": 3.077189800128847e-06, "loss": 0.0103, "step": 48840 }, { "epoch": 15.03, "learning_rate": 3.0768302430978713e-06, "loss": 0.0086, "step": 48841 }, { "epoch": 15.03, "learning_rate": 3.07647070325526e-06, "loss": 0.0114, "step": 48842 }, { "epoch": 15.03, "learning_rate": 3.0761111806019016e-06, "loss": 0.0084, "step": 48843 }, { "epoch": 15.03, "learning_rate": 3.0757516751386974e-06, "loss": 0.0095, "step": 48844 }, { "epoch": 15.03, "learning_rate": 3.075392186866534e-06, "loss": 0.0095, "step": 48845 }, { "epoch": 15.03, "learning_rate": 3.075032715786305e-06, "loss": 0.0105, "step": 48846 }, { "epoch": 15.03, "learning_rate": 3.074673261898903e-06, "loss": 0.0101, "step": 48847 }, { "epoch": 15.03, "learning_rate": 3.0743138252052196e-06, "loss": 0.0122, "step": 48848 }, { "epoch": 15.03, "learning_rate": 3.073954405706148e-06, "loss": 0.0076, "step": 48849 }, { "epoch": 15.03, "learning_rate": 3.0735950034025807e-06, "loss": 0.0069, "step": 48850 }, { "epoch": 15.03, "learning_rate": 3.0732356182954103e-06, "loss": 0.0047, "step": 48851 }, { "epoch": 15.03, "learning_rate": 3.072876250385528e-06, "loss": 0.005, "step": 48852 }, { "epoch": 15.03, "learning_rate": 3.0725168996738264e-06, "loss": 0.004, "step": 48853 }, { "epoch": 15.03, "learning_rate": 3.0721575661611967e-06, "loss": 0.0084, "step": 48854 }, { "epoch": 15.03, "learning_rate": 3.071798249848534e-06, "loss": 0.0079, "step": 48855 }, { "epoch": 15.03, "learning_rate": 3.0714389507367272e-06, "loss": 0.007, "step": 48856 }, { "epoch": 15.03, "learning_rate": 3.07107966882667e-06, "loss": 0.0083, "step": 48857 }, { "epoch": 15.03, "learning_rate": 3.070720404119251e-06, "loss": 0.0067, "step": 48858 }, { "epoch": 15.03, "learning_rate": 3.0703611566153677e-06, "loss": 0.0104, "step": 48859 }, { "epoch": 15.03, "learning_rate": 3.070001926315911e-06, "loss": 0.0085, "step": 48860 }, { "epoch": 15.03, "learning_rate": 3.069642713221773e-06, "loss": 0.0079, "step": 48861 }, { "epoch": 15.03, "learning_rate": 3.069283517333841e-06, "loss": 0.0109, "step": 48862 }, { "epoch": 15.03, "learning_rate": 3.0689243386530064e-06, "loss": 0.0131, "step": 48863 }, { "epoch": 15.03, "learning_rate": 3.0685651771801672e-06, "loss": 0.0074, "step": 48864 }, { "epoch": 15.04, "learning_rate": 3.0682060329162122e-06, "loss": 0.0086, "step": 48865 }, { "epoch": 15.04, "learning_rate": 3.067846905862032e-06, "loss": 0.0052, "step": 48866 }, { "epoch": 15.04, "learning_rate": 3.06748779601852e-06, "loss": 0.0053, "step": 48867 }, { "epoch": 15.04, "learning_rate": 3.067128703386567e-06, "loss": 0.0072, "step": 48868 }, { "epoch": 15.04, "learning_rate": 3.0667696279670635e-06, "loss": 0.0053, "step": 48869 }, { "epoch": 15.04, "learning_rate": 3.0664105697609025e-06, "loss": 0.0162, "step": 48870 }, { "epoch": 15.04, "learning_rate": 3.066051528768975e-06, "loss": 0.0098, "step": 48871 }, { "epoch": 15.04, "learning_rate": 3.0656925049921717e-06, "loss": 0.0133, "step": 48872 }, { "epoch": 15.04, "learning_rate": 3.065333498431381e-06, "loss": 0.0094, "step": 48873 }, { "epoch": 15.04, "learning_rate": 3.064974509087505e-06, "loss": 0.0215, "step": 48874 }, { "epoch": 15.04, "learning_rate": 3.0646155369614238e-06, "loss": 0.064, "step": 48875 }, { "epoch": 15.04, "learning_rate": 3.0642565820540325e-06, "loss": 0.0131, "step": 48876 }, { "epoch": 15.04, "learning_rate": 3.0638976443662227e-06, "loss": 0.0089, "step": 48877 }, { "epoch": 15.04, "learning_rate": 3.0635387238988823e-06, "loss": 0.0097, "step": 48878 }, { "epoch": 15.04, "learning_rate": 3.0631798206529085e-06, "loss": 0.0063, "step": 48879 }, { "epoch": 15.04, "learning_rate": 3.062820934629188e-06, "loss": 0.009, "step": 48880 }, { "epoch": 15.04, "learning_rate": 3.062462065828613e-06, "loss": 0.0107, "step": 48881 }, { "epoch": 15.04, "learning_rate": 3.0621032142520777e-06, "loss": 0.0104, "step": 48882 }, { "epoch": 15.04, "learning_rate": 3.061744379900464e-06, "loss": 0.006, "step": 48883 }, { "epoch": 15.04, "learning_rate": 3.0613855627746713e-06, "loss": 0.0095, "step": 48884 }, { "epoch": 15.04, "learning_rate": 3.0610267628755884e-06, "loss": 0.0128, "step": 48885 }, { "epoch": 15.04, "learning_rate": 3.060667980204104e-06, "loss": 0.0064, "step": 48886 }, { "epoch": 15.04, "learning_rate": 3.060309214761111e-06, "loss": 0.0077, "step": 48887 }, { "epoch": 15.04, "learning_rate": 3.0599504665475e-06, "loss": 0.0081, "step": 48888 }, { "epoch": 15.04, "learning_rate": 3.05959173556416e-06, "loss": 0.0089, "step": 48889 }, { "epoch": 15.04, "learning_rate": 3.0592330218119824e-06, "loss": 0.0148, "step": 48890 }, { "epoch": 15.04, "learning_rate": 3.0588743252918583e-06, "loss": 0.0099, "step": 48891 }, { "epoch": 15.04, "learning_rate": 3.0585156460046784e-06, "loss": 0.0076, "step": 48892 }, { "epoch": 15.04, "learning_rate": 3.058156983951329e-06, "loss": 0.0168, "step": 48893 }, { "epoch": 15.04, "learning_rate": 3.0577983391327082e-06, "loss": 0.006, "step": 48894 }, { "epoch": 15.04, "learning_rate": 3.0574397115497056e-06, "loss": 0.0047, "step": 48895 }, { "epoch": 15.04, "learning_rate": 3.057081101203205e-06, "loss": 0.015, "step": 48896 }, { "epoch": 15.05, "learning_rate": 3.056722508094101e-06, "loss": 0.0091, "step": 48897 }, { "epoch": 15.05, "learning_rate": 3.0563639322232797e-06, "loss": 0.0101, "step": 48898 }, { "epoch": 15.05, "learning_rate": 3.0560053735916372e-06, "loss": 0.0122, "step": 48899 }, { "epoch": 15.05, "learning_rate": 3.055646832200062e-06, "loss": 0.0103, "step": 48900 }, { "epoch": 15.05, "learning_rate": 3.055288308049443e-06, "loss": 0.0069, "step": 48901 }, { "epoch": 15.05, "learning_rate": 3.0549298011406724e-06, "loss": 0.0042, "step": 48902 }, { "epoch": 15.05, "learning_rate": 3.0545713114746376e-06, "loss": 0.0055, "step": 48903 }, { "epoch": 15.05, "learning_rate": 3.0542128390522298e-06, "loss": 0.0058, "step": 48904 }, { "epoch": 15.05, "learning_rate": 3.0538543838743395e-06, "loss": 0.0138, "step": 48905 }, { "epoch": 15.05, "learning_rate": 3.0534959459418554e-06, "loss": 0.0174, "step": 48906 }, { "epoch": 15.05, "learning_rate": 3.053137525255668e-06, "loss": 0.0118, "step": 48907 }, { "epoch": 15.05, "learning_rate": 3.052779121816665e-06, "loss": 0.0036, "step": 48908 }, { "epoch": 15.05, "learning_rate": 3.0524207356257453e-06, "loss": 0.0051, "step": 48909 }, { "epoch": 15.05, "learning_rate": 3.052062366683787e-06, "loss": 0.0043, "step": 48910 }, { "epoch": 15.05, "learning_rate": 3.0517040149916853e-06, "loss": 0.0091, "step": 48911 }, { "epoch": 15.05, "learning_rate": 3.0513456805503305e-06, "loss": 0.0064, "step": 48912 }, { "epoch": 15.05, "learning_rate": 3.050987363360607e-06, "loss": 0.0056, "step": 48913 }, { "epoch": 15.05, "learning_rate": 3.0506290634234116e-06, "loss": 0.0045, "step": 48914 }, { "epoch": 15.05, "learning_rate": 3.0502707807396314e-06, "loss": 0.0062, "step": 48915 }, { "epoch": 15.05, "learning_rate": 3.049912515310157e-06, "loss": 0.0102, "step": 48916 }, { "epoch": 15.05, "learning_rate": 3.0495542671358745e-06, "loss": 0.0117, "step": 48917 }, { "epoch": 15.05, "learning_rate": 3.049196036217671e-06, "loss": 0.0052, "step": 48918 }, { "epoch": 15.05, "learning_rate": 3.048837822556443e-06, "loss": 0.0048, "step": 48919 }, { "epoch": 15.05, "learning_rate": 3.048479626153078e-06, "loss": 0.0057, "step": 48920 }, { "epoch": 15.05, "learning_rate": 3.048121447008463e-06, "loss": 0.0072, "step": 48921 }, { "epoch": 15.05, "learning_rate": 3.0477632851234872e-06, "loss": 0.0063, "step": 48922 }, { "epoch": 15.05, "learning_rate": 3.0474051404990424e-06, "loss": 0.008, "step": 48923 }, { "epoch": 15.05, "learning_rate": 3.0470470131360165e-06, "loss": 0.0059, "step": 48924 }, { "epoch": 15.05, "learning_rate": 3.0466889030352976e-06, "loss": 0.0085, "step": 48925 }, { "epoch": 15.05, "learning_rate": 3.046330810197775e-06, "loss": 0.0122, "step": 48926 }, { "epoch": 15.05, "learning_rate": 3.0459727346243396e-06, "loss": 0.0093, "step": 48927 }, { "epoch": 15.05, "learning_rate": 3.0456146763158755e-06, "loss": 0.0077, "step": 48928 }, { "epoch": 15.05, "learning_rate": 3.045256635273278e-06, "loss": 0.0076, "step": 48929 }, { "epoch": 15.06, "learning_rate": 3.044898611497438e-06, "loss": 0.0103, "step": 48930 }, { "epoch": 15.06, "learning_rate": 3.0445406049892347e-06, "loss": 0.0071, "step": 48931 }, { "epoch": 15.06, "learning_rate": 3.0441826157495625e-06, "loss": 0.004, "step": 48932 }, { "epoch": 15.06, "learning_rate": 3.0438246437793073e-06, "loss": 0.0094, "step": 48933 }, { "epoch": 15.06, "learning_rate": 3.043466689079363e-06, "loss": 0.0077, "step": 48934 }, { "epoch": 15.06, "learning_rate": 3.043108751650615e-06, "loss": 0.0055, "step": 48935 }, { "epoch": 15.06, "learning_rate": 3.0427508314939523e-06, "loss": 0.0088, "step": 48936 }, { "epoch": 15.06, "learning_rate": 3.042392928610267e-06, "loss": 0.0088, "step": 48937 }, { "epoch": 15.06, "learning_rate": 3.0420350430004388e-06, "loss": 0.0074, "step": 48938 }, { "epoch": 15.06, "learning_rate": 3.041677174665364e-06, "loss": 0.012, "step": 48939 }, { "epoch": 15.06, "learning_rate": 3.041319323605929e-06, "loss": 0.0067, "step": 48940 }, { "epoch": 15.06, "learning_rate": 3.040961489823023e-06, "loss": 0.009, "step": 48941 }, { "epoch": 15.06, "learning_rate": 3.0406036733175326e-06, "loss": 0.0084, "step": 48942 }, { "epoch": 15.06, "learning_rate": 3.0402458740903474e-06, "loss": 0.0091, "step": 48943 }, { "epoch": 15.06, "learning_rate": 3.0398880921423556e-06, "loss": 0.0085, "step": 48944 }, { "epoch": 15.06, "learning_rate": 3.0395303274744446e-06, "loss": 0.0078, "step": 48945 }, { "epoch": 15.06, "learning_rate": 3.0391725800875037e-06, "loss": 0.0067, "step": 48946 }, { "epoch": 15.06, "learning_rate": 3.038814849982421e-06, "loss": 0.0053, "step": 48947 }, { "epoch": 15.06, "learning_rate": 3.0384571371600812e-06, "loss": 0.0068, "step": 48948 }, { "epoch": 15.06, "learning_rate": 3.0380994416213785e-06, "loss": 0.005, "step": 48949 }, { "epoch": 15.06, "learning_rate": 3.037741763367201e-06, "loss": 0.0081, "step": 48950 }, { "epoch": 15.06, "learning_rate": 3.037384102398431e-06, "loss": 0.009, "step": 48951 }, { "epoch": 15.06, "learning_rate": 3.037026458715959e-06, "loss": 0.0097, "step": 48952 }, { "epoch": 15.06, "learning_rate": 3.0366688323206694e-06, "loss": 0.0065, "step": 48953 }, { "epoch": 15.06, "learning_rate": 3.0363112232134584e-06, "loss": 0.0114, "step": 48954 }, { "epoch": 15.06, "learning_rate": 3.0359536313952076e-06, "loss": 0.0101, "step": 48955 }, { "epoch": 15.06, "learning_rate": 3.0355960568668077e-06, "loss": 0.0104, "step": 48956 }, { "epoch": 15.06, "learning_rate": 3.035238499629145e-06, "loss": 0.0092, "step": 48957 }, { "epoch": 15.06, "learning_rate": 3.0348809596831075e-06, "loss": 0.0133, "step": 48958 }, { "epoch": 15.06, "learning_rate": 3.034523437029582e-06, "loss": 0.0075, "step": 48959 }, { "epoch": 15.06, "learning_rate": 3.0341659316694583e-06, "loss": 0.0106, "step": 48960 }, { "epoch": 15.06, "learning_rate": 3.033808443603622e-06, "loss": 0.008, "step": 48961 }, { "epoch": 15.07, "learning_rate": 3.033450972832961e-06, "loss": 0.0043, "step": 48962 }, { "epoch": 15.07, "learning_rate": 3.03309351935836e-06, "loss": 0.0068, "step": 48963 }, { "epoch": 15.07, "learning_rate": 3.032736083180716e-06, "loss": 0.0085, "step": 48964 }, { "epoch": 15.07, "learning_rate": 3.032378664300907e-06, "loss": 0.0082, "step": 48965 }, { "epoch": 15.07, "learning_rate": 3.032021262719823e-06, "loss": 0.0068, "step": 48966 }, { "epoch": 15.07, "learning_rate": 3.0316638784383524e-06, "loss": 0.0128, "step": 48967 }, { "epoch": 15.07, "learning_rate": 3.031306511457378e-06, "loss": 0.0112, "step": 48968 }, { "epoch": 15.07, "learning_rate": 3.0309491617777943e-06, "loss": 0.0031, "step": 48969 }, { "epoch": 15.07, "learning_rate": 3.0305918294004854e-06, "loss": 0.0145, "step": 48970 }, { "epoch": 15.07, "learning_rate": 3.03023451432634e-06, "loss": 0.0065, "step": 48971 }, { "epoch": 15.07, "learning_rate": 3.029877216556242e-06, "loss": 0.009, "step": 48972 }, { "epoch": 15.07, "learning_rate": 3.0295199360910754e-06, "loss": 0.0092, "step": 48973 }, { "epoch": 15.07, "learning_rate": 3.029162672931735e-06, "loss": 0.0057, "step": 48974 }, { "epoch": 15.07, "learning_rate": 3.0288054270791036e-06, "loss": 0.0162, "step": 48975 }, { "epoch": 15.07, "learning_rate": 3.0284481985340697e-06, "loss": 0.0036, "step": 48976 }, { "epoch": 15.07, "learning_rate": 3.0280909872975194e-06, "loss": 0.0072, "step": 48977 }, { "epoch": 15.07, "learning_rate": 3.02773379337034e-06, "loss": 0.0133, "step": 48978 }, { "epoch": 15.07, "learning_rate": 3.027376616753418e-06, "loss": 0.0077, "step": 48979 }, { "epoch": 15.07, "learning_rate": 3.0270194574476397e-06, "loss": 0.0105, "step": 48980 }, { "epoch": 15.07, "learning_rate": 3.0266623154538923e-06, "loss": 0.0088, "step": 48981 }, { "epoch": 15.07, "learning_rate": 3.026305190773062e-06, "loss": 0.0119, "step": 48982 }, { "epoch": 15.07, "learning_rate": 3.025948083406033e-06, "loss": 0.0112, "step": 48983 }, { "epoch": 15.07, "learning_rate": 3.0255909933536974e-06, "loss": 0.0133, "step": 48984 }, { "epoch": 15.07, "learning_rate": 3.0252339206169433e-06, "loss": 0.0081, "step": 48985 }, { "epoch": 15.07, "learning_rate": 3.02487686519665e-06, "loss": 0.0058, "step": 48986 }, { "epoch": 15.07, "learning_rate": 3.024519827093706e-06, "loss": 0.0106, "step": 48987 }, { "epoch": 15.07, "learning_rate": 3.0241628063089957e-06, "loss": 0.0139, "step": 48988 }, { "epoch": 15.07, "learning_rate": 3.0238058028434126e-06, "loss": 0.0056, "step": 48989 }, { "epoch": 15.07, "learning_rate": 3.023448816697838e-06, "loss": 0.0062, "step": 48990 }, { "epoch": 15.07, "learning_rate": 3.0230918478731598e-06, "loss": 0.0064, "step": 48991 }, { "epoch": 15.07, "learning_rate": 3.0227348963702663e-06, "loss": 0.0148, "step": 48992 }, { "epoch": 15.07, "learning_rate": 3.022377962190035e-06, "loss": 0.0042, "step": 48993 }, { "epoch": 15.07, "learning_rate": 3.022021045333361e-06, "loss": 0.0052, "step": 48994 }, { "epoch": 15.08, "learning_rate": 3.0216641458011276e-06, "loss": 0.0057, "step": 48995 }, { "epoch": 15.08, "learning_rate": 3.0213072635942207e-06, "loss": 0.006, "step": 48996 }, { "epoch": 15.08, "learning_rate": 3.020950398713527e-06, "loss": 0.0091, "step": 48997 }, { "epoch": 15.08, "learning_rate": 3.0205935511599316e-06, "loss": 0.0116, "step": 48998 }, { "epoch": 15.08, "learning_rate": 3.0202367209343207e-06, "loss": 0.0102, "step": 48999 }, { "epoch": 15.08, "learning_rate": 3.0198799080375808e-06, "loss": 0.0057, "step": 49000 }, { "epoch": 15.08, "learning_rate": 3.019523112470597e-06, "loss": 0.0082, "step": 49001 }, { "epoch": 15.08, "learning_rate": 3.0191663342342557e-06, "loss": 0.0095, "step": 49002 }, { "epoch": 15.08, "learning_rate": 3.0188095733294388e-06, "loss": 0.0062, "step": 49003 }, { "epoch": 15.08, "learning_rate": 3.0184528297570394e-06, "loss": 0.0088, "step": 49004 }, { "epoch": 15.08, "learning_rate": 3.018096103517939e-06, "loss": 0.0128, "step": 49005 }, { "epoch": 15.08, "learning_rate": 3.0177393946130273e-06, "loss": 0.0066, "step": 49006 }, { "epoch": 15.08, "learning_rate": 3.0173827030431823e-06, "loss": 0.0105, "step": 49007 }, { "epoch": 15.08, "learning_rate": 3.0170260288092913e-06, "loss": 0.0108, "step": 49008 }, { "epoch": 15.08, "learning_rate": 3.0166693719122454e-06, "loss": 0.0086, "step": 49009 }, { "epoch": 15.08, "learning_rate": 3.0163127323529263e-06, "loss": 0.0051, "step": 49010 }, { "epoch": 15.08, "learning_rate": 3.01595611013222e-06, "loss": 0.0103, "step": 49011 }, { "epoch": 15.08, "learning_rate": 3.015599505251011e-06, "loss": 0.0047, "step": 49012 }, { "epoch": 15.08, "learning_rate": 3.0152429177101872e-06, "loss": 0.0078, "step": 49013 }, { "epoch": 15.08, "learning_rate": 3.0148863475106315e-06, "loss": 0.0047, "step": 49014 }, { "epoch": 15.08, "learning_rate": 3.0145297946532292e-06, "loss": 0.0065, "step": 49015 }, { "epoch": 15.08, "learning_rate": 3.014173259138867e-06, "loss": 0.0149, "step": 49016 }, { "epoch": 15.08, "learning_rate": 3.0138167409684293e-06, "loss": 0.0079, "step": 49017 }, { "epoch": 15.08, "learning_rate": 3.013460240142797e-06, "loss": 0.0108, "step": 49018 }, { "epoch": 15.08, "learning_rate": 3.0131037566628673e-06, "loss": 0.0087, "step": 49019 }, { "epoch": 15.08, "learning_rate": 3.012747290529513e-06, "loss": 0.0056, "step": 49020 }, { "epoch": 15.08, "learning_rate": 3.0123908417436245e-06, "loss": 0.0062, "step": 49021 }, { "epoch": 15.08, "learning_rate": 3.012034410306085e-06, "loss": 0.0082, "step": 49022 }, { "epoch": 15.08, "learning_rate": 3.0116779962177765e-06, "loss": 0.0113, "step": 49023 }, { "epoch": 15.08, "learning_rate": 3.0113215994795917e-06, "loss": 0.0121, "step": 49024 }, { "epoch": 15.08, "learning_rate": 3.010965220092411e-06, "loss": 0.0119, "step": 49025 }, { "epoch": 15.08, "learning_rate": 3.010608858057119e-06, "loss": 0.009, "step": 49026 }, { "epoch": 15.09, "learning_rate": 3.010252513374604e-06, "loss": 0.0083, "step": 49027 }, { "epoch": 15.09, "learning_rate": 3.009896186045742e-06, "loss": 0.0175, "step": 49028 }, { "epoch": 15.09, "learning_rate": 3.009539876071427e-06, "loss": 0.0094, "step": 49029 }, { "epoch": 15.09, "learning_rate": 3.0091835834525386e-06, "loss": 0.0049, "step": 49030 }, { "epoch": 15.09, "learning_rate": 3.008827308189963e-06, "loss": 0.0089, "step": 49031 }, { "epoch": 15.09, "learning_rate": 3.0084710502845847e-06, "loss": 0.0213, "step": 49032 }, { "epoch": 15.09, "learning_rate": 3.0081148097372883e-06, "loss": 0.0104, "step": 49033 }, { "epoch": 15.09, "learning_rate": 3.0077585865489567e-06, "loss": 0.0119, "step": 49034 }, { "epoch": 15.09, "learning_rate": 3.0074023807204764e-06, "loss": 0.0147, "step": 49035 }, { "epoch": 15.09, "learning_rate": 3.0070461922527305e-06, "loss": 0.0108, "step": 49036 }, { "epoch": 15.09, "learning_rate": 3.0066900211466045e-06, "loss": 0.0094, "step": 49037 }, { "epoch": 15.09, "learning_rate": 3.0063338674029772e-06, "loss": 0.0069, "step": 49038 }, { "epoch": 15.09, "learning_rate": 3.0059777310227423e-06, "loss": 0.0081, "step": 49039 }, { "epoch": 15.09, "learning_rate": 3.0056216120067815e-06, "loss": 0.0063, "step": 49040 }, { "epoch": 15.09, "learning_rate": 3.0052655103559735e-06, "loss": 0.0053, "step": 49041 }, { "epoch": 15.09, "learning_rate": 3.0049094260712053e-06, "loss": 0.0097, "step": 49042 }, { "epoch": 15.09, "learning_rate": 3.004553359153358e-06, "loss": 0.0064, "step": 49043 }, { "epoch": 15.09, "learning_rate": 3.0041973096033215e-06, "loss": 0.0055, "step": 49044 }, { "epoch": 15.09, "learning_rate": 3.0038412774219783e-06, "loss": 0.0072, "step": 49045 }, { "epoch": 15.09, "learning_rate": 3.003485262610211e-06, "loss": 0.0108, "step": 49046 }, { "epoch": 15.09, "learning_rate": 3.0031292651689025e-06, "loss": 0.0212, "step": 49047 }, { "epoch": 15.09, "learning_rate": 3.002773285098938e-06, "loss": 0.007, "step": 49048 }, { "epoch": 15.09, "learning_rate": 3.002417322401202e-06, "loss": 0.0056, "step": 49049 }, { "epoch": 15.09, "learning_rate": 3.002061377076576e-06, "loss": 0.0083, "step": 49050 }, { "epoch": 15.09, "learning_rate": 3.0017054491259455e-06, "loss": 0.0087, "step": 49051 }, { "epoch": 15.09, "learning_rate": 3.0013495385501944e-06, "loss": 0.02, "step": 49052 }, { "epoch": 15.09, "learning_rate": 3.000993645350205e-06, "loss": 0.0085, "step": 49053 }, { "epoch": 15.09, "learning_rate": 3.000637769526862e-06, "loss": 0.0123, "step": 49054 }, { "epoch": 15.09, "learning_rate": 3.0002819110810475e-06, "loss": 0.0113, "step": 49055 }, { "epoch": 15.09, "learning_rate": 2.9999260700136467e-06, "loss": 0.0066, "step": 49056 }, { "epoch": 15.09, "learning_rate": 2.9995702463255418e-06, "loss": 0.0075, "step": 49057 }, { "epoch": 15.09, "learning_rate": 2.999214440017614e-06, "loss": 0.0127, "step": 49058 }, { "epoch": 15.09, "learning_rate": 2.9988586510907526e-06, "loss": 0.0116, "step": 49059 }, { "epoch": 15.1, "learning_rate": 2.998502879545837e-06, "loss": 0.0082, "step": 49060 }, { "epoch": 15.1, "learning_rate": 2.9981471253837545e-06, "loss": 0.0091, "step": 49061 }, { "epoch": 15.1, "learning_rate": 2.99779138860538e-06, "loss": 0.0057, "step": 49062 }, { "epoch": 15.1, "learning_rate": 2.9974356692116036e-06, "loss": 0.008, "step": 49063 }, { "epoch": 15.1, "learning_rate": 2.997079967203307e-06, "loss": 0.0072, "step": 49064 }, { "epoch": 15.1, "learning_rate": 2.9967242825813736e-06, "loss": 0.0031, "step": 49065 }, { "epoch": 15.1, "learning_rate": 2.9963686153466842e-06, "loss": 0.0058, "step": 49066 }, { "epoch": 15.1, "learning_rate": 2.9960129655001248e-06, "loss": 0.0193, "step": 49067 }, { "epoch": 15.1, "learning_rate": 2.9956573330425766e-06, "loss": 0.0057, "step": 49068 }, { "epoch": 15.1, "learning_rate": 2.9953017179749223e-06, "loss": 0.0032, "step": 49069 }, { "epoch": 15.1, "learning_rate": 2.9949461202980467e-06, "loss": 0.0137, "step": 49070 }, { "epoch": 15.1, "learning_rate": 2.9945905400128305e-06, "loss": 0.0074, "step": 49071 }, { "epoch": 15.1, "learning_rate": 2.994234977120154e-06, "loss": 0.0084, "step": 49072 }, { "epoch": 15.1, "learning_rate": 2.9938794316209076e-06, "loss": 0.0074, "step": 49073 }, { "epoch": 15.1, "learning_rate": 2.993523903515971e-06, "loss": 0.0048, "step": 49074 }, { "epoch": 15.1, "learning_rate": 2.993168392806224e-06, "loss": 0.0113, "step": 49075 }, { "epoch": 15.1, "learning_rate": 2.9928128994925508e-06, "loss": 0.0066, "step": 49076 }, { "epoch": 15.1, "learning_rate": 2.9924574235758308e-06, "loss": 0.0084, "step": 49077 }, { "epoch": 15.1, "learning_rate": 2.9921019650569527e-06, "loss": 0.007, "step": 49078 }, { "epoch": 15.1, "learning_rate": 2.9917465239367972e-06, "loss": 0.005, "step": 49079 }, { "epoch": 15.1, "learning_rate": 2.9913911002162443e-06, "loss": 0.0071, "step": 49080 }, { "epoch": 15.1, "learning_rate": 2.9910356938961782e-06, "loss": 0.0061, "step": 49081 }, { "epoch": 15.1, "learning_rate": 2.9906803049774812e-06, "loss": 0.0093, "step": 49082 }, { "epoch": 15.1, "learning_rate": 2.990324933461035e-06, "loss": 0.0052, "step": 49083 }, { "epoch": 15.1, "learning_rate": 2.9899695793477236e-06, "loss": 0.0067, "step": 49084 }, { "epoch": 15.1, "learning_rate": 2.989614242638427e-06, "loss": 0.0145, "step": 49085 }, { "epoch": 15.1, "learning_rate": 2.9892589233340275e-06, "loss": 0.0062, "step": 49086 }, { "epoch": 15.1, "learning_rate": 2.988903621435408e-06, "loss": 0.0089, "step": 49087 }, { "epoch": 15.1, "learning_rate": 2.988548336943452e-06, "loss": 0.0062, "step": 49088 }, { "epoch": 15.1, "learning_rate": 2.98819306985904e-06, "loss": 0.0086, "step": 49089 }, { "epoch": 15.1, "learning_rate": 2.987837820183054e-06, "loss": 0.0068, "step": 49090 }, { "epoch": 15.1, "learning_rate": 2.987482587916376e-06, "loss": 0.0064, "step": 49091 }, { "epoch": 15.11, "learning_rate": 2.987127373059885e-06, "loss": 0.0103, "step": 49092 }, { "epoch": 15.11, "learning_rate": 2.9867721756144698e-06, "loss": 0.0103, "step": 49093 }, { "epoch": 15.11, "learning_rate": 2.9864169955810085e-06, "loss": 0.0064, "step": 49094 }, { "epoch": 15.11, "learning_rate": 2.986061832960385e-06, "loss": 0.0079, "step": 49095 }, { "epoch": 15.11, "learning_rate": 2.9857066877534767e-06, "loss": 0.0081, "step": 49096 }, { "epoch": 15.11, "learning_rate": 2.9853515599611647e-06, "loss": 0.0046, "step": 49097 }, { "epoch": 15.11, "learning_rate": 2.9849964495843375e-06, "loss": 0.0031, "step": 49098 }, { "epoch": 15.11, "learning_rate": 2.984641356623872e-06, "loss": 0.0084, "step": 49099 }, { "epoch": 15.11, "learning_rate": 2.9842862810806516e-06, "loss": 0.0072, "step": 49100 }, { "epoch": 15.11, "learning_rate": 2.983931222955556e-06, "loss": 0.0061, "step": 49101 }, { "epoch": 15.11, "learning_rate": 2.983576182249468e-06, "loss": 0.0073, "step": 49102 }, { "epoch": 15.11, "learning_rate": 2.9832211589632687e-06, "loss": 0.008, "step": 49103 }, { "epoch": 15.11, "learning_rate": 2.9828661530978397e-06, "loss": 0.009, "step": 49104 }, { "epoch": 15.11, "learning_rate": 2.982511164654063e-06, "loss": 0.0076, "step": 49105 }, { "epoch": 15.11, "learning_rate": 2.982156193632818e-06, "loss": 0.004, "step": 49106 }, { "epoch": 15.11, "learning_rate": 2.981801240034985e-06, "loss": 0.0106, "step": 49107 }, { "epoch": 15.11, "learning_rate": 2.9814463038614527e-06, "loss": 0.0174, "step": 49108 }, { "epoch": 15.11, "learning_rate": 2.981091385113095e-06, "loss": 0.0108, "step": 49109 }, { "epoch": 15.11, "learning_rate": 2.980736483790795e-06, "loss": 0.0077, "step": 49110 }, { "epoch": 15.11, "learning_rate": 2.9803815998954334e-06, "loss": 0.0065, "step": 49111 }, { "epoch": 15.11, "learning_rate": 2.98002673342789e-06, "loss": 0.0081, "step": 49112 }, { "epoch": 15.11, "learning_rate": 2.98002673342789e-06, "loss": 0.0279, "step": 49113 }, { "epoch": 15.11, "learning_rate": 2.979671884389049e-06, "loss": 0.0086, "step": 49114 }, { "epoch": 15.11, "learning_rate": 2.979317052779791e-06, "loss": 0.0085, "step": 49115 }, { "epoch": 15.11, "learning_rate": 2.9789622386009964e-06, "loss": 0.0137, "step": 49116 }, { "epoch": 15.11, "learning_rate": 2.9786074418535483e-06, "loss": 0.0072, "step": 49117 }, { "epoch": 15.11, "learning_rate": 2.978252662538319e-06, "loss": 0.008, "step": 49118 }, { "epoch": 15.11, "learning_rate": 2.977897900656198e-06, "loss": 0.0111, "step": 49119 }, { "epoch": 15.11, "learning_rate": 2.977543156208065e-06, "loss": 0.0041, "step": 49120 }, { "epoch": 15.11, "learning_rate": 2.977188429194797e-06, "loss": 0.0222, "step": 49121 }, { "epoch": 15.11, "learning_rate": 2.9768337196172783e-06, "loss": 0.0059, "step": 49122 }, { "epoch": 15.11, "learning_rate": 2.976479027476388e-06, "loss": 0.0157, "step": 49123 }, { "epoch": 15.11, "learning_rate": 2.9761243527730057e-06, "loss": 0.009, "step": 49124 }, { "epoch": 15.12, "learning_rate": 2.975769695508014e-06, "loss": 0.0151, "step": 49125 }, { "epoch": 15.12, "learning_rate": 2.9754150556822927e-06, "loss": 0.0071, "step": 49126 }, { "epoch": 15.12, "learning_rate": 2.9750604332967216e-06, "loss": 0.0065, "step": 49127 }, { "epoch": 15.12, "learning_rate": 2.9747058283521777e-06, "loss": 0.006, "step": 49128 }, { "epoch": 15.12, "learning_rate": 2.974351240849549e-06, "loss": 0.005, "step": 49129 }, { "epoch": 15.12, "learning_rate": 2.973996670789716e-06, "loss": 0.0106, "step": 49130 }, { "epoch": 15.12, "learning_rate": 2.973642118173551e-06, "loss": 0.0111, "step": 49131 }, { "epoch": 15.12, "learning_rate": 2.9732875830019392e-06, "loss": 0.0072, "step": 49132 }, { "epoch": 15.12, "learning_rate": 2.972933065275756e-06, "loss": 0.0069, "step": 49133 }, { "epoch": 15.12, "learning_rate": 2.9725785649958895e-06, "loss": 0.012, "step": 49134 }, { "epoch": 15.12, "learning_rate": 2.972224082163215e-06, "loss": 0.0145, "step": 49135 }, { "epoch": 15.12, "learning_rate": 2.9718696167786145e-06, "loss": 0.0079, "step": 49136 }, { "epoch": 15.12, "learning_rate": 2.9715151688429657e-06, "loss": 0.0095, "step": 49137 }, { "epoch": 15.12, "learning_rate": 2.9711607383571507e-06, "loss": 0.0093, "step": 49138 }, { "epoch": 15.12, "learning_rate": 2.9708063253220487e-06, "loss": 0.0091, "step": 49139 }, { "epoch": 15.12, "learning_rate": 2.9704519297385392e-06, "loss": 0.0072, "step": 49140 }, { "epoch": 15.12, "learning_rate": 2.9700975516075037e-06, "loss": 0.0066, "step": 49141 }, { "epoch": 15.12, "learning_rate": 2.9697431909298193e-06, "loss": 0.0088, "step": 49142 }, { "epoch": 15.12, "learning_rate": 2.9693888477063647e-06, "loss": 0.0074, "step": 49143 }, { "epoch": 15.12, "learning_rate": 2.969034521938028e-06, "loss": 0.0055, "step": 49144 }, { "epoch": 15.12, "learning_rate": 2.968680213625681e-06, "loss": 0.0074, "step": 49145 }, { "epoch": 15.12, "learning_rate": 2.9683259227702043e-06, "loss": 0.0094, "step": 49146 }, { "epoch": 15.12, "learning_rate": 2.9679716493724795e-06, "loss": 0.0076, "step": 49147 }, { "epoch": 15.12, "learning_rate": 2.967617393433382e-06, "loss": 0.0148, "step": 49148 }, { "epoch": 15.12, "learning_rate": 2.967263154953798e-06, "loss": 0.009, "step": 49149 }, { "epoch": 15.12, "learning_rate": 2.9669089339346026e-06, "loss": 0.0063, "step": 49150 }, { "epoch": 15.12, "learning_rate": 2.9665547303766806e-06, "loss": 0.0084, "step": 49151 }, { "epoch": 15.12, "learning_rate": 2.9662005442809027e-06, "loss": 0.008, "step": 49152 }, { "epoch": 15.12, "learning_rate": 2.9658463756481503e-06, "loss": 0.0073, "step": 49153 }, { "epoch": 15.12, "learning_rate": 2.9654922244793084e-06, "loss": 0.0249, "step": 49154 }, { "epoch": 15.12, "learning_rate": 2.9651380907752525e-06, "loss": 0.008, "step": 49155 }, { "epoch": 15.12, "learning_rate": 2.9647839745368623e-06, "loss": 0.0196, "step": 49156 }, { "epoch": 15.13, "learning_rate": 2.964429875765016e-06, "loss": 0.0096, "step": 49157 }, { "epoch": 15.13, "learning_rate": 2.9640757944605947e-06, "loss": 0.0083, "step": 49158 }, { "epoch": 15.13, "learning_rate": 2.963721730624476e-06, "loss": 0.0082, "step": 49159 }, { "epoch": 15.13, "learning_rate": 2.9633676842575386e-06, "loss": 0.0066, "step": 49160 }, { "epoch": 15.13, "learning_rate": 2.9630136553606636e-06, "loss": 0.0107, "step": 49161 }, { "epoch": 15.13, "learning_rate": 2.962659643934728e-06, "loss": 0.0074, "step": 49162 }, { "epoch": 15.13, "learning_rate": 2.9623056499806067e-06, "loss": 0.0072, "step": 49163 }, { "epoch": 15.13, "learning_rate": 2.9619516734991884e-06, "loss": 0.0058, "step": 49164 }, { "epoch": 15.13, "learning_rate": 2.9615977144913487e-06, "loss": 0.0049, "step": 49165 }, { "epoch": 15.13, "learning_rate": 2.9612437729579626e-06, "loss": 0.0073, "step": 49166 }, { "epoch": 15.13, "learning_rate": 2.9608898488999093e-06, "loss": 0.0057, "step": 49167 }, { "epoch": 15.13, "learning_rate": 2.9605359423180655e-06, "loss": 0.0093, "step": 49168 }, { "epoch": 15.13, "learning_rate": 2.9601820532133173e-06, "loss": 0.0039, "step": 49169 }, { "epoch": 15.13, "learning_rate": 2.959828181586538e-06, "loss": 0.0068, "step": 49170 }, { "epoch": 15.13, "learning_rate": 2.9594743274386084e-06, "loss": 0.0096, "step": 49171 }, { "epoch": 15.13, "learning_rate": 2.9591204907704084e-06, "loss": 0.0058, "step": 49172 }, { "epoch": 15.13, "learning_rate": 2.958766671582809e-06, "loss": 0.027, "step": 49173 }, { "epoch": 15.13, "learning_rate": 2.9584128698766955e-06, "loss": 0.0057, "step": 49174 }, { "epoch": 15.13, "learning_rate": 2.9580590856529445e-06, "loss": 0.0066, "step": 49175 }, { "epoch": 15.13, "learning_rate": 2.957705318912435e-06, "loss": 0.0092, "step": 49176 }, { "epoch": 15.13, "learning_rate": 2.957351569656045e-06, "loss": 0.0085, "step": 49177 }, { "epoch": 15.13, "learning_rate": 2.956997837884652e-06, "loss": 0.0089, "step": 49178 }, { "epoch": 15.13, "learning_rate": 2.9566441235991338e-06, "loss": 0.0054, "step": 49179 }, { "epoch": 15.13, "learning_rate": 2.95629042680037e-06, "loss": 0.0091, "step": 49180 }, { "epoch": 15.13, "learning_rate": 2.9559367474892385e-06, "loss": 0.0044, "step": 49181 }, { "epoch": 15.13, "learning_rate": 2.9555830856666167e-06, "loss": 0.0204, "step": 49182 }, { "epoch": 15.13, "learning_rate": 2.9552294413333804e-06, "loss": 0.0101, "step": 49183 }, { "epoch": 15.13, "learning_rate": 2.954875814490412e-06, "loss": 0.0085, "step": 49184 }, { "epoch": 15.13, "learning_rate": 2.9545222051385892e-06, "loss": 0.0092, "step": 49185 }, { "epoch": 15.13, "learning_rate": 2.9541686132787907e-06, "loss": 0.0095, "step": 49186 }, { "epoch": 15.13, "learning_rate": 2.953815038911889e-06, "loss": 0.0082, "step": 49187 }, { "epoch": 15.13, "learning_rate": 2.953461482038762e-06, "loss": 0.0083, "step": 49188 }, { "epoch": 15.13, "learning_rate": 2.953107942660294e-06, "loss": 0.0089, "step": 49189 }, { "epoch": 15.14, "learning_rate": 2.9527544207773583e-06, "loss": 0.0081, "step": 49190 }, { "epoch": 15.14, "learning_rate": 2.952400916390833e-06, "loss": 0.0136, "step": 49191 }, { "epoch": 15.14, "learning_rate": 2.9520474295015975e-06, "loss": 0.0134, "step": 49192 }, { "epoch": 15.14, "learning_rate": 2.951693960110528e-06, "loss": 0.0174, "step": 49193 }, { "epoch": 15.14, "learning_rate": 2.951340508218502e-06, "loss": 0.0049, "step": 49194 }, { "epoch": 15.14, "learning_rate": 2.950987073826397e-06, "loss": 0.0103, "step": 49195 }, { "epoch": 15.14, "learning_rate": 2.950633656935091e-06, "loss": 0.0086, "step": 49196 }, { "epoch": 15.14, "learning_rate": 2.9502802575454615e-06, "loss": 0.0072, "step": 49197 }, { "epoch": 15.14, "learning_rate": 2.9499268756583823e-06, "loss": 0.0084, "step": 49198 }, { "epoch": 15.14, "learning_rate": 2.949573511274739e-06, "loss": 0.0061, "step": 49199 }, { "epoch": 15.14, "learning_rate": 2.949220164395402e-06, "loss": 0.0138, "step": 49200 }, { "epoch": 15.14, "learning_rate": 2.9488668350212503e-06, "loss": 0.0234, "step": 49201 }, { "epoch": 15.14, "learning_rate": 2.9485135231531616e-06, "loss": 0.0094, "step": 49202 }, { "epoch": 15.14, "learning_rate": 2.94816022879201e-06, "loss": 0.0088, "step": 49203 }, { "epoch": 15.14, "learning_rate": 2.9478069519386776e-06, "loss": 0.0085, "step": 49204 }, { "epoch": 15.14, "learning_rate": 2.94745369259404e-06, "loss": 0.0076, "step": 49205 }, { "epoch": 15.14, "learning_rate": 2.9471004507589727e-06, "loss": 0.0104, "step": 49206 }, { "epoch": 15.14, "learning_rate": 2.9467472264343577e-06, "loss": 0.0068, "step": 49207 }, { "epoch": 15.14, "learning_rate": 2.946394019621062e-06, "loss": 0.0254, "step": 49208 }, { "epoch": 15.14, "learning_rate": 2.9460408303199696e-06, "loss": 0.0107, "step": 49209 }, { "epoch": 15.14, "learning_rate": 2.945687658531957e-06, "loss": 0.0088, "step": 49210 }, { "epoch": 15.14, "learning_rate": 2.9453345042579006e-06, "loss": 0.0069, "step": 49211 }, { "epoch": 15.14, "learning_rate": 2.944981367498677e-06, "loss": 0.0031, "step": 49212 }, { "epoch": 15.14, "learning_rate": 2.9446282482551625e-06, "loss": 0.0106, "step": 49213 }, { "epoch": 15.14, "learning_rate": 2.9442751465282336e-06, "loss": 0.0077, "step": 49214 }, { "epoch": 15.14, "learning_rate": 2.943922062318767e-06, "loss": 0.0099, "step": 49215 }, { "epoch": 15.14, "learning_rate": 2.9435689956276403e-06, "loss": 0.0064, "step": 49216 }, { "epoch": 15.14, "learning_rate": 2.9432159464557296e-06, "loss": 0.0139, "step": 49217 }, { "epoch": 15.14, "learning_rate": 2.9428629148039077e-06, "loss": 0.0123, "step": 49218 }, { "epoch": 15.14, "learning_rate": 2.942509900673057e-06, "loss": 0.0112, "step": 49219 }, { "epoch": 15.14, "learning_rate": 2.9421569040640564e-06, "loss": 0.0068, "step": 49220 }, { "epoch": 15.14, "learning_rate": 2.941803924977773e-06, "loss": 0.0072, "step": 49221 }, { "epoch": 15.15, "learning_rate": 2.941450963415089e-06, "loss": 0.0145, "step": 49222 }, { "epoch": 15.15, "learning_rate": 2.9410980193768744e-06, "loss": 0.0064, "step": 49223 }, { "epoch": 15.15, "learning_rate": 2.940745092864015e-06, "loss": 0.0151, "step": 49224 }, { "epoch": 15.15, "learning_rate": 2.940392183877382e-06, "loss": 0.0093, "step": 49225 }, { "epoch": 15.15, "learning_rate": 2.940039292417851e-06, "loss": 0.0059, "step": 49226 }, { "epoch": 15.15, "learning_rate": 2.9396864184863007e-06, "loss": 0.0063, "step": 49227 }, { "epoch": 15.15, "learning_rate": 2.939333562083605e-06, "loss": 0.0083, "step": 49228 }, { "epoch": 15.15, "learning_rate": 2.9389807232106404e-06, "loss": 0.0049, "step": 49229 }, { "epoch": 15.15, "learning_rate": 2.938627901868283e-06, "loss": 0.009, "step": 49230 }, { "epoch": 15.15, "learning_rate": 2.9382750980574093e-06, "loss": 0.0157, "step": 49231 }, { "epoch": 15.15, "learning_rate": 2.9379223117788947e-06, "loss": 0.0096, "step": 49232 }, { "epoch": 15.15, "learning_rate": 2.9375695430336147e-06, "loss": 0.0053, "step": 49233 }, { "epoch": 15.15, "learning_rate": 2.937216791822445e-06, "loss": 0.006, "step": 49234 }, { "epoch": 15.15, "learning_rate": 2.9368640581462627e-06, "loss": 0.0225, "step": 49235 }, { "epoch": 15.15, "learning_rate": 2.9365113420059432e-06, "loss": 0.0074, "step": 49236 }, { "epoch": 15.15, "learning_rate": 2.9361586434023614e-06, "loss": 0.0078, "step": 49237 }, { "epoch": 15.15, "learning_rate": 2.93580596233639e-06, "loss": 0.0049, "step": 49238 }, { "epoch": 15.15, "learning_rate": 2.9354532988089113e-06, "loss": 0.0109, "step": 49239 }, { "epoch": 15.15, "learning_rate": 2.935100652820797e-06, "loss": 0.0112, "step": 49240 }, { "epoch": 15.15, "learning_rate": 2.9347480243729263e-06, "loss": 0.0053, "step": 49241 }, { "epoch": 15.15, "learning_rate": 2.934395413466169e-06, "loss": 0.0459, "step": 49242 }, { "epoch": 15.15, "learning_rate": 2.9340428201014005e-06, "loss": 0.0072, "step": 49243 }, { "epoch": 15.15, "learning_rate": 2.9336902442795003e-06, "loss": 0.015, "step": 49244 }, { "epoch": 15.15, "learning_rate": 2.933337686001343e-06, "loss": 0.0085, "step": 49245 }, { "epoch": 15.15, "learning_rate": 2.9329851452678038e-06, "loss": 0.0072, "step": 49246 }, { "epoch": 15.15, "learning_rate": 2.9326326220797563e-06, "loss": 0.0099, "step": 49247 }, { "epoch": 15.15, "learning_rate": 2.9322801164380767e-06, "loss": 0.0067, "step": 49248 }, { "epoch": 15.15, "learning_rate": 2.9319276283436405e-06, "loss": 0.0109, "step": 49249 }, { "epoch": 15.15, "learning_rate": 2.9315751577973227e-06, "loss": 0.0071, "step": 49250 }, { "epoch": 15.15, "learning_rate": 2.9312227047999974e-06, "loss": 0.0114, "step": 49251 }, { "epoch": 15.15, "learning_rate": 2.9308702693525417e-06, "loss": 0.0097, "step": 49252 }, { "epoch": 15.15, "learning_rate": 2.930517851455825e-06, "loss": 0.007, "step": 49253 }, { "epoch": 15.15, "learning_rate": 2.9301654511107335e-06, "loss": 0.0075, "step": 49254 }, { "epoch": 15.16, "learning_rate": 2.9298130683181323e-06, "loss": 0.0121, "step": 49255 }, { "epoch": 15.16, "learning_rate": 2.929460703078899e-06, "loss": 0.0088, "step": 49256 }, { "epoch": 15.16, "learning_rate": 2.929108355393908e-06, "loss": 0.0316, "step": 49257 }, { "epoch": 15.16, "learning_rate": 2.9287560252640323e-06, "loss": 0.0069, "step": 49258 }, { "epoch": 15.16, "learning_rate": 2.928403712690152e-06, "loss": 0.0052, "step": 49259 }, { "epoch": 15.16, "learning_rate": 2.9280514176731377e-06, "loss": 0.0156, "step": 49260 }, { "epoch": 15.16, "learning_rate": 2.9276991402138667e-06, "loss": 0.0073, "step": 49261 }, { "epoch": 15.16, "learning_rate": 2.9273468803132156e-06, "loss": 0.0074, "step": 49262 }, { "epoch": 15.16, "learning_rate": 2.9269946379720483e-06, "loss": 0.0111, "step": 49263 }, { "epoch": 15.16, "learning_rate": 2.9266424131912495e-06, "loss": 0.0071, "step": 49264 }, { "epoch": 15.16, "learning_rate": 2.92629020597169e-06, "loss": 0.0065, "step": 49265 }, { "epoch": 15.16, "learning_rate": 2.925938016314246e-06, "loss": 0.0097, "step": 49266 }, { "epoch": 15.16, "learning_rate": 2.9255858442197905e-06, "loss": 0.0127, "step": 49267 }, { "epoch": 15.16, "learning_rate": 2.9252336896891986e-06, "loss": 0.0068, "step": 49268 }, { "epoch": 15.16, "learning_rate": 2.924881552723343e-06, "loss": 0.0073, "step": 49269 }, { "epoch": 15.16, "learning_rate": 2.9245294333231e-06, "loss": 0.0129, "step": 49270 }, { "epoch": 15.16, "learning_rate": 2.9241773314893427e-06, "loss": 0.0072, "step": 49271 }, { "epoch": 15.16, "learning_rate": 2.9238252472229468e-06, "loss": 0.0056, "step": 49272 }, { "epoch": 15.16, "learning_rate": 2.9234731805247807e-06, "loss": 0.0095, "step": 49273 }, { "epoch": 15.16, "learning_rate": 2.9231211313957254e-06, "loss": 0.0046, "step": 49274 }, { "epoch": 15.16, "learning_rate": 2.9227690998366564e-06, "loss": 0.0104, "step": 49275 }, { "epoch": 15.16, "learning_rate": 2.9224170858484414e-06, "loss": 0.0093, "step": 49276 }, { "epoch": 15.16, "learning_rate": 2.9220650894319557e-06, "loss": 0.0132, "step": 49277 }, { "epoch": 15.16, "learning_rate": 2.9217131105880715e-06, "loss": 0.0272, "step": 49278 }, { "epoch": 15.16, "learning_rate": 2.9213611493176675e-06, "loss": 0.0069, "step": 49279 }, { "epoch": 15.16, "learning_rate": 2.9210092056216166e-06, "loss": 0.0124, "step": 49280 }, { "epoch": 15.16, "learning_rate": 2.9206572795007904e-06, "loss": 0.0056, "step": 49281 }, { "epoch": 15.16, "learning_rate": 2.9203053709560647e-06, "loss": 0.0077, "step": 49282 }, { "epoch": 15.16, "learning_rate": 2.9199534799883124e-06, "loss": 0.0033, "step": 49283 }, { "epoch": 15.16, "learning_rate": 2.9196016065984056e-06, "loss": 0.0024, "step": 49284 }, { "epoch": 15.16, "learning_rate": 2.91924975078722e-06, "loss": 0.0128, "step": 49285 }, { "epoch": 15.16, "learning_rate": 2.9188979125556284e-06, "loss": 0.0074, "step": 49286 }, { "epoch": 15.17, "learning_rate": 2.918546091904504e-06, "loss": 0.0074, "step": 49287 }, { "epoch": 15.17, "learning_rate": 2.9181942888347204e-06, "loss": 0.0091, "step": 49288 }, { "epoch": 15.17, "learning_rate": 2.9178425033471513e-06, "loss": 0.0084, "step": 49289 }, { "epoch": 15.17, "learning_rate": 2.9174907354426696e-06, "loss": 0.0169, "step": 49290 }, { "epoch": 15.17, "learning_rate": 2.91713898512215e-06, "loss": 0.0068, "step": 49291 }, { "epoch": 15.17, "learning_rate": 2.916787252386464e-06, "loss": 0.008, "step": 49292 }, { "epoch": 15.17, "learning_rate": 2.916435537236483e-06, "loss": 0.0067, "step": 49293 }, { "epoch": 15.17, "learning_rate": 2.916083839673086e-06, "loss": 0.0098, "step": 49294 }, { "epoch": 15.17, "learning_rate": 2.9157321596971434e-06, "loss": 0.0103, "step": 49295 }, { "epoch": 15.17, "learning_rate": 2.91538049730953e-06, "loss": 0.0063, "step": 49296 }, { "epoch": 15.17, "learning_rate": 2.9150288525111146e-06, "loss": 0.0049, "step": 49297 }, { "epoch": 15.17, "learning_rate": 2.914677225302769e-06, "loss": 0.012, "step": 49298 }, { "epoch": 15.17, "learning_rate": 2.914325615685373e-06, "loss": 0.0081, "step": 49299 }, { "epoch": 15.17, "learning_rate": 2.9139740236597958e-06, "loss": 0.008, "step": 49300 }, { "epoch": 15.17, "learning_rate": 2.9136224492269116e-06, "loss": 0.007, "step": 49301 }, { "epoch": 15.17, "learning_rate": 2.9132708923875917e-06, "loss": 0.0067, "step": 49302 }, { "epoch": 15.17, "learning_rate": 2.91291935314271e-06, "loss": 0.0062, "step": 49303 }, { "epoch": 15.17, "learning_rate": 2.9125678314931394e-06, "loss": 0.005, "step": 49304 }, { "epoch": 15.17, "learning_rate": 2.912216327439752e-06, "loss": 0.0067, "step": 49305 }, { "epoch": 15.17, "learning_rate": 2.9118648409834205e-06, "loss": 0.0117, "step": 49306 }, { "epoch": 15.17, "learning_rate": 2.9115133721250168e-06, "loss": 0.0109, "step": 49307 }, { "epoch": 15.17, "learning_rate": 2.911161920865412e-06, "loss": 0.0045, "step": 49308 }, { "epoch": 15.17, "learning_rate": 2.910810487205488e-06, "loss": 0.0144, "step": 49309 }, { "epoch": 15.17, "learning_rate": 2.9104590711461066e-06, "loss": 0.0075, "step": 49310 }, { "epoch": 15.17, "learning_rate": 2.9101076726881437e-06, "loss": 0.0112, "step": 49311 }, { "epoch": 15.17, "learning_rate": 2.909756291832473e-06, "loss": 0.011, "step": 49312 }, { "epoch": 15.17, "learning_rate": 2.909404928579962e-06, "loss": 0.0043, "step": 49313 }, { "epoch": 15.17, "learning_rate": 2.9090535829314904e-06, "loss": 0.0062, "step": 49314 }, { "epoch": 15.17, "learning_rate": 2.908702254887926e-06, "loss": 0.0073, "step": 49315 }, { "epoch": 15.17, "learning_rate": 2.9083509444501433e-06, "loss": 0.0073, "step": 49316 }, { "epoch": 15.17, "learning_rate": 2.907999651619017e-06, "loss": 0.0053, "step": 49317 }, { "epoch": 15.17, "learning_rate": 2.907648376395409e-06, "loss": 0.0116, "step": 49318 }, { "epoch": 15.17, "learning_rate": 2.9072971187802e-06, "loss": 0.0063, "step": 49319 }, { "epoch": 15.18, "learning_rate": 2.9069458787742612e-06, "loss": 0.0081, "step": 49320 }, { "epoch": 15.18, "learning_rate": 2.906594656378464e-06, "loss": 0.0082, "step": 49321 }, { "epoch": 15.18, "learning_rate": 2.906243451593679e-06, "loss": 0.0211, "step": 49322 }, { "epoch": 15.18, "learning_rate": 2.90589226442078e-06, "loss": 0.0075, "step": 49323 }, { "epoch": 15.18, "learning_rate": 2.9055410948606365e-06, "loss": 0.0138, "step": 49324 }, { "epoch": 15.18, "learning_rate": 2.905189942914123e-06, "loss": 0.0093, "step": 49325 }, { "epoch": 15.18, "learning_rate": 2.90483880858211e-06, "loss": 0.0092, "step": 49326 }, { "epoch": 15.18, "learning_rate": 2.9044876918654697e-06, "loss": 0.0146, "step": 49327 }, { "epoch": 15.18, "learning_rate": 2.90413659276507e-06, "loss": 0.0054, "step": 49328 }, { "epoch": 15.18, "learning_rate": 2.9037855112817893e-06, "loss": 0.0076, "step": 49329 }, { "epoch": 15.18, "learning_rate": 2.903434447416499e-06, "loss": 0.0076, "step": 49330 }, { "epoch": 15.18, "learning_rate": 2.9030834011700648e-06, "loss": 0.0084, "step": 49331 }, { "epoch": 15.18, "learning_rate": 2.902732372543362e-06, "loss": 0.0233, "step": 49332 }, { "epoch": 15.18, "learning_rate": 2.9023813615372577e-06, "loss": 0.0086, "step": 49333 }, { "epoch": 15.18, "learning_rate": 2.90203036815263e-06, "loss": 0.0075, "step": 49334 }, { "epoch": 15.18, "learning_rate": 2.9016793923903474e-06, "loss": 0.0087, "step": 49335 }, { "epoch": 15.18, "learning_rate": 2.9013284342512817e-06, "loss": 0.0151, "step": 49336 }, { "epoch": 15.18, "learning_rate": 2.900977493736303e-06, "loss": 0.0084, "step": 49337 }, { "epoch": 15.18, "learning_rate": 2.900626570846283e-06, "loss": 0.0125, "step": 49338 }, { "epoch": 15.18, "learning_rate": 2.900275665582095e-06, "loss": 0.0125, "step": 49339 }, { "epoch": 15.18, "learning_rate": 2.8999247779446073e-06, "loss": 0.0047, "step": 49340 }, { "epoch": 15.18, "learning_rate": 2.899573907934692e-06, "loss": 0.0084, "step": 49341 }, { "epoch": 15.18, "learning_rate": 2.899223055553221e-06, "loss": 0.01, "step": 49342 }, { "epoch": 15.18, "learning_rate": 2.8988722208010623e-06, "loss": 0.0054, "step": 49343 }, { "epoch": 15.18, "learning_rate": 2.8985214036790944e-06, "loss": 0.0153, "step": 49344 }, { "epoch": 15.18, "learning_rate": 2.898170604188181e-06, "loss": 0.0087, "step": 49345 }, { "epoch": 15.18, "learning_rate": 2.8978198223291953e-06, "loss": 0.0073, "step": 49346 }, { "epoch": 15.18, "learning_rate": 2.8974690581030085e-06, "loss": 0.0134, "step": 49347 }, { "epoch": 15.18, "learning_rate": 2.897118311510487e-06, "loss": 0.0068, "step": 49348 }, { "epoch": 15.18, "learning_rate": 2.8967675825525097e-06, "loss": 0.0043, "step": 49349 }, { "epoch": 15.18, "learning_rate": 2.8964168712299447e-06, "loss": 0.0123, "step": 49350 }, { "epoch": 15.18, "learning_rate": 2.896066177543664e-06, "loss": 0.0123, "step": 49351 }, { "epoch": 15.19, "learning_rate": 2.8957155014945316e-06, "loss": 0.0131, "step": 49352 }, { "epoch": 15.19, "learning_rate": 2.8953648430834204e-06, "loss": 0.0102, "step": 49353 }, { "epoch": 15.19, "learning_rate": 2.895014202311205e-06, "loss": 0.0059, "step": 49354 }, { "epoch": 15.19, "learning_rate": 2.8946635791787546e-06, "loss": 0.0053, "step": 49355 }, { "epoch": 15.19, "learning_rate": 2.894312973686939e-06, "loss": 0.0088, "step": 49356 }, { "epoch": 15.19, "learning_rate": 2.8939623858366284e-06, "loss": 0.0076, "step": 49357 }, { "epoch": 15.19, "learning_rate": 2.893611815628694e-06, "loss": 0.0076, "step": 49358 }, { "epoch": 15.19, "learning_rate": 2.8932612630640057e-06, "loss": 0.0093, "step": 49359 }, { "epoch": 15.19, "learning_rate": 2.8929107281434334e-06, "loss": 0.0086, "step": 49360 }, { "epoch": 15.19, "learning_rate": 2.8925602108678474e-06, "loss": 0.0171, "step": 49361 }, { "epoch": 15.19, "learning_rate": 2.8922097112381185e-06, "loss": 0.0111, "step": 49362 }, { "epoch": 15.19, "learning_rate": 2.891859229255113e-06, "loss": 0.0089, "step": 49363 }, { "epoch": 15.19, "learning_rate": 2.8915087649197094e-06, "loss": 0.0141, "step": 49364 }, { "epoch": 15.19, "learning_rate": 2.891158318232775e-06, "loss": 0.0082, "step": 49365 }, { "epoch": 15.19, "learning_rate": 2.890807889195175e-06, "loss": 0.0058, "step": 49366 }, { "epoch": 15.19, "learning_rate": 2.890457477807783e-06, "loss": 0.0128, "step": 49367 }, { "epoch": 15.19, "learning_rate": 2.890107084071465e-06, "loss": 0.0076, "step": 49368 }, { "epoch": 15.19, "learning_rate": 2.889756707987097e-06, "loss": 0.0085, "step": 49369 }, { "epoch": 15.19, "learning_rate": 2.8894063495555468e-06, "loss": 0.0094, "step": 49370 }, { "epoch": 15.19, "learning_rate": 2.8890560087776843e-06, "loss": 0.007, "step": 49371 }, { "epoch": 15.19, "learning_rate": 2.8887056856543805e-06, "loss": 0.0138, "step": 49372 }, { "epoch": 15.19, "learning_rate": 2.8883553801864973e-06, "loss": 0.0171, "step": 49373 }, { "epoch": 15.19, "learning_rate": 2.8880050923749137e-06, "loss": 0.0136, "step": 49374 }, { "epoch": 15.19, "learning_rate": 2.8876548222204968e-06, "loss": 0.0086, "step": 49375 }, { "epoch": 15.19, "learning_rate": 2.887304569724115e-06, "loss": 0.0107, "step": 49376 }, { "epoch": 15.19, "learning_rate": 2.886954334886639e-06, "loss": 0.0135, "step": 49377 }, { "epoch": 15.19, "learning_rate": 2.8866041177089365e-06, "loss": 0.0072, "step": 49378 }, { "epoch": 15.19, "learning_rate": 2.8862539181918794e-06, "loss": 0.0068, "step": 49379 }, { "epoch": 15.19, "learning_rate": 2.8859037363363362e-06, "loss": 0.0067, "step": 49380 }, { "epoch": 15.19, "learning_rate": 2.885553572143175e-06, "loss": 0.0102, "step": 49381 }, { "epoch": 15.19, "learning_rate": 2.8852034256132666e-06, "loss": 0.0087, "step": 49382 }, { "epoch": 15.19, "learning_rate": 2.8848532967474775e-06, "loss": 0.008, "step": 49383 }, { "epoch": 15.19, "learning_rate": 2.884503185546682e-06, "loss": 0.0109, "step": 49384 }, { "epoch": 15.2, "learning_rate": 2.884153092011748e-06, "loss": 0.006, "step": 49385 }, { "epoch": 15.2, "learning_rate": 2.8838030161435447e-06, "loss": 0.0042, "step": 49386 }, { "epoch": 15.2, "learning_rate": 2.883452957942938e-06, "loss": 0.0048, "step": 49387 }, { "epoch": 15.2, "learning_rate": 2.883102917410796e-06, "loss": 0.008, "step": 49388 }, { "epoch": 15.2, "learning_rate": 2.8827528945479945e-06, "loss": 0.0159, "step": 49389 }, { "epoch": 15.2, "learning_rate": 2.882402889355398e-06, "loss": 0.0127, "step": 49390 }, { "epoch": 15.2, "learning_rate": 2.882052901833876e-06, "loss": 0.009, "step": 49391 }, { "epoch": 15.2, "learning_rate": 2.881702931984298e-06, "loss": 0.0089, "step": 49392 }, { "epoch": 15.2, "learning_rate": 2.8813529798075323e-06, "loss": 0.0062, "step": 49393 }, { "epoch": 15.2, "learning_rate": 2.881003045304448e-06, "loss": 0.0066, "step": 49394 }, { "epoch": 15.2, "learning_rate": 2.8806531284759144e-06, "loss": 0.0101, "step": 49395 }, { "epoch": 15.2, "learning_rate": 2.8803032293227995e-06, "loss": 0.0072, "step": 49396 }, { "epoch": 15.2, "learning_rate": 2.879953347845973e-06, "loss": 0.0037, "step": 49397 }, { "epoch": 15.2, "learning_rate": 2.8796034840462995e-06, "loss": 0.0077, "step": 49398 }, { "epoch": 15.2, "learning_rate": 2.879253637924656e-06, "loss": 0.0115, "step": 49399 }, { "epoch": 15.2, "learning_rate": 2.8789038094819044e-06, "loss": 0.0119, "step": 49400 }, { "epoch": 15.2, "learning_rate": 2.8785539987189138e-06, "loss": 0.0079, "step": 49401 }, { "epoch": 15.2, "learning_rate": 2.8782042056365543e-06, "loss": 0.01, "step": 49402 }, { "epoch": 15.2, "learning_rate": 2.8778544302356904e-06, "loss": 0.006, "step": 49403 }, { "epoch": 15.2, "learning_rate": 2.8775046725171975e-06, "loss": 0.0116, "step": 49404 }, { "epoch": 15.2, "learning_rate": 2.877154932481939e-06, "loss": 0.0047, "step": 49405 }, { "epoch": 15.2, "learning_rate": 2.8768052101307863e-06, "loss": 0.0102, "step": 49406 }, { "epoch": 15.2, "learning_rate": 2.8764555054646083e-06, "loss": 0.0129, "step": 49407 }, { "epoch": 15.2, "learning_rate": 2.876105818484264e-06, "loss": 0.0121, "step": 49408 }, { "epoch": 15.2, "learning_rate": 2.8757561491906317e-06, "loss": 0.0059, "step": 49409 }, { "epoch": 15.2, "learning_rate": 2.875406497584576e-06, "loss": 0.0067, "step": 49410 }, { "epoch": 15.2, "learning_rate": 2.875056863666966e-06, "loss": 0.008, "step": 49411 }, { "epoch": 15.2, "learning_rate": 2.8747072474386694e-06, "loss": 0.0101, "step": 49412 }, { "epoch": 15.2, "learning_rate": 2.874357648900553e-06, "loss": 0.0036, "step": 49413 }, { "epoch": 15.2, "learning_rate": 2.8740080680534854e-06, "loss": 0.0069, "step": 49414 }, { "epoch": 15.2, "learning_rate": 2.8736585048983356e-06, "loss": 0.0068, "step": 49415 }, { "epoch": 15.2, "learning_rate": 2.8733089594359696e-06, "loss": 0.0092, "step": 49416 }, { "epoch": 15.21, "learning_rate": 2.872959431667256e-06, "loss": 0.0075, "step": 49417 }, { "epoch": 15.21, "learning_rate": 2.8726099215930604e-06, "loss": 0.0072, "step": 49418 }, { "epoch": 15.21, "learning_rate": 2.8722604292142552e-06, "loss": 0.0062, "step": 49419 }, { "epoch": 15.21, "learning_rate": 2.8719109545317102e-06, "loss": 0.0104, "step": 49420 }, { "epoch": 15.21, "learning_rate": 2.8715614975462847e-06, "loss": 0.0109, "step": 49421 }, { "epoch": 15.21, "learning_rate": 2.87121205825885e-06, "loss": 0.0065, "step": 49422 }, { "epoch": 15.21, "learning_rate": 2.8708626366702706e-06, "loss": 0.0083, "step": 49423 }, { "epoch": 15.21, "learning_rate": 2.870513232781421e-06, "loss": 0.0108, "step": 49424 }, { "epoch": 15.21, "learning_rate": 2.870163846593165e-06, "loss": 0.0061, "step": 49425 }, { "epoch": 15.21, "learning_rate": 2.86981447810637e-06, "loss": 0.0099, "step": 49426 }, { "epoch": 15.21, "learning_rate": 2.869465127321903e-06, "loss": 0.0066, "step": 49427 }, { "epoch": 15.21, "learning_rate": 2.8691157942406324e-06, "loss": 0.0108, "step": 49428 }, { "epoch": 15.21, "learning_rate": 2.868766478863424e-06, "loss": 0.0106, "step": 49429 }, { "epoch": 15.21, "learning_rate": 2.8684171811911464e-06, "loss": 0.0112, "step": 49430 }, { "epoch": 15.21, "learning_rate": 2.868067901224666e-06, "loss": 0.0043, "step": 49431 }, { "epoch": 15.21, "learning_rate": 2.8677186389648503e-06, "loss": 0.0066, "step": 49432 }, { "epoch": 15.21, "learning_rate": 2.867369394412567e-06, "loss": 0.0099, "step": 49433 }, { "epoch": 15.21, "learning_rate": 2.8670201675686825e-06, "loss": 0.0052, "step": 49434 }, { "epoch": 15.21, "learning_rate": 2.8666709584340636e-06, "loss": 0.0107, "step": 49435 }, { "epoch": 15.21, "learning_rate": 2.8663217670095777e-06, "loss": 0.018, "step": 49436 }, { "epoch": 15.21, "learning_rate": 2.8659725932960916e-06, "loss": 0.0063, "step": 49437 }, { "epoch": 15.21, "learning_rate": 2.8656234372944692e-06, "loss": 0.0148, "step": 49438 }, { "epoch": 15.21, "learning_rate": 2.8652742990055836e-06, "loss": 0.0098, "step": 49439 }, { "epoch": 15.21, "learning_rate": 2.864925178430299e-06, "loss": 0.0078, "step": 49440 }, { "epoch": 15.21, "learning_rate": 2.864576075569484e-06, "loss": 0.0093, "step": 49441 }, { "epoch": 15.21, "learning_rate": 2.8642269904239996e-06, "loss": 0.0171, "step": 49442 }, { "epoch": 15.21, "learning_rate": 2.8638779229947135e-06, "loss": 0.0132, "step": 49443 }, { "epoch": 15.21, "learning_rate": 2.863528873282497e-06, "loss": 0.0053, "step": 49444 }, { "epoch": 15.21, "learning_rate": 2.863179841288215e-06, "loss": 0.0067, "step": 49445 }, { "epoch": 15.21, "learning_rate": 2.8628308270127335e-06, "loss": 0.0108, "step": 49446 }, { "epoch": 15.21, "learning_rate": 2.8624818304569193e-06, "loss": 0.0066, "step": 49447 }, { "epoch": 15.21, "learning_rate": 2.8621328516216374e-06, "loss": 0.0089, "step": 49448 }, { "epoch": 15.21, "learning_rate": 2.8617838905077557e-06, "loss": 0.0047, "step": 49449 }, { "epoch": 15.22, "learning_rate": 2.8614349471161406e-06, "loss": 0.0122, "step": 49450 }, { "epoch": 15.22, "learning_rate": 2.8610860214476578e-06, "loss": 0.0074, "step": 49451 }, { "epoch": 15.22, "learning_rate": 2.860737113503174e-06, "loss": 0.0051, "step": 49452 }, { "epoch": 15.22, "learning_rate": 2.860388223283551e-06, "loss": 0.0061, "step": 49453 }, { "epoch": 15.22, "learning_rate": 2.860039350789667e-06, "loss": 0.0178, "step": 49454 }, { "epoch": 15.22, "learning_rate": 2.859690496022377e-06, "loss": 0.0117, "step": 49455 }, { "epoch": 15.22, "learning_rate": 2.8593416589825498e-06, "loss": 0.0072, "step": 49456 }, { "epoch": 15.22, "learning_rate": 2.8589928396710496e-06, "loss": 0.01, "step": 49457 }, { "epoch": 15.22, "learning_rate": 2.858644038088748e-06, "loss": 0.0095, "step": 49458 }, { "epoch": 15.22, "learning_rate": 2.8582952542365083e-06, "loss": 0.0099, "step": 49459 }, { "epoch": 15.22, "learning_rate": 2.857946488115195e-06, "loss": 0.0165, "step": 49460 }, { "epoch": 15.22, "learning_rate": 2.857597739725677e-06, "loss": 0.0072, "step": 49461 }, { "epoch": 15.22, "learning_rate": 2.8572490090688165e-06, "loss": 0.0048, "step": 49462 }, { "epoch": 15.22, "learning_rate": 2.856900296145483e-06, "loss": 0.0058, "step": 49463 }, { "epoch": 15.22, "learning_rate": 2.8565516009565387e-06, "loss": 0.0293, "step": 49464 }, { "epoch": 15.22, "learning_rate": 2.856202923502851e-06, "loss": 0.0069, "step": 49465 }, { "epoch": 15.22, "learning_rate": 2.8558542637852858e-06, "loss": 0.0052, "step": 49466 }, { "epoch": 15.22, "learning_rate": 2.8555056218047097e-06, "loss": 0.0036, "step": 49467 }, { "epoch": 15.22, "learning_rate": 2.855156997561985e-06, "loss": 0.0061, "step": 49468 }, { "epoch": 15.22, "learning_rate": 2.8548083910579818e-06, "loss": 0.0079, "step": 49469 }, { "epoch": 15.22, "learning_rate": 2.854459802293561e-06, "loss": 0.0141, "step": 49470 }, { "epoch": 15.22, "learning_rate": 2.8541112312695908e-06, "loss": 0.0075, "step": 49471 }, { "epoch": 15.22, "learning_rate": 2.853762677986932e-06, "loss": 0.0068, "step": 49472 }, { "epoch": 15.22, "learning_rate": 2.853414142446458e-06, "loss": 0.0056, "step": 49473 }, { "epoch": 15.22, "learning_rate": 2.8530656246490306e-06, "loss": 0.0111, "step": 49474 }, { "epoch": 15.22, "learning_rate": 2.8527171245955163e-06, "loss": 0.0091, "step": 49475 }, { "epoch": 15.22, "learning_rate": 2.8523686422867757e-06, "loss": 0.0079, "step": 49476 }, { "epoch": 15.22, "learning_rate": 2.8520201777236742e-06, "loss": 0.0065, "step": 49477 }, { "epoch": 15.22, "learning_rate": 2.851671730907083e-06, "loss": 0.007, "step": 49478 }, { "epoch": 15.22, "learning_rate": 2.851323301837863e-06, "loss": 0.0135, "step": 49479 }, { "epoch": 15.22, "learning_rate": 2.85097489051688e-06, "loss": 0.0046, "step": 49480 }, { "epoch": 15.22, "learning_rate": 2.8506264969449993e-06, "loss": 0.0236, "step": 49481 }, { "epoch": 15.23, "learning_rate": 2.8502781211230855e-06, "loss": 0.0094, "step": 49482 }, { "epoch": 15.23, "learning_rate": 2.849929763052004e-06, "loss": 0.0084, "step": 49483 }, { "epoch": 15.23, "learning_rate": 2.849581422732619e-06, "loss": 0.0097, "step": 49484 }, { "epoch": 15.23, "learning_rate": 2.849233100165795e-06, "loss": 0.0115, "step": 49485 }, { "epoch": 15.23, "learning_rate": 2.848884795352398e-06, "loss": 0.0062, "step": 49486 }, { "epoch": 15.23, "learning_rate": 2.848536508293289e-06, "loss": 0.0121, "step": 49487 }, { "epoch": 15.23, "learning_rate": 2.848188238989342e-06, "loss": 0.0197, "step": 49488 }, { "epoch": 15.23, "learning_rate": 2.847839987441413e-06, "loss": 0.0156, "step": 49489 }, { "epoch": 15.23, "learning_rate": 2.8474917536503687e-06, "loss": 0.0049, "step": 49490 }, { "epoch": 15.23, "learning_rate": 2.847143537617074e-06, "loss": 0.0054, "step": 49491 }, { "epoch": 15.23, "learning_rate": 2.8467953393423908e-06, "loss": 0.0063, "step": 49492 }, { "epoch": 15.23, "learning_rate": 2.8464471588271892e-06, "loss": 0.0037, "step": 49493 }, { "epoch": 15.23, "learning_rate": 2.8460989960723306e-06, "loss": 0.0083, "step": 49494 }, { "epoch": 15.23, "learning_rate": 2.84575085107868e-06, "loss": 0.006, "step": 49495 }, { "epoch": 15.23, "learning_rate": 2.845402723847104e-06, "loss": 0.0046, "step": 49496 }, { "epoch": 15.23, "learning_rate": 2.8450546143784585e-06, "loss": 0.004, "step": 49497 }, { "epoch": 15.23, "learning_rate": 2.844706522673616e-06, "loss": 0.0111, "step": 49498 }, { "epoch": 15.23, "learning_rate": 2.8443584487334385e-06, "loss": 0.0137, "step": 49499 }, { "epoch": 15.23, "learning_rate": 2.8440103925587904e-06, "loss": 0.0128, "step": 49500 }, { "epoch": 15.23, "learning_rate": 2.843662354150534e-06, "loss": 0.0068, "step": 49501 }, { "epoch": 15.23, "learning_rate": 2.843314333509536e-06, "loss": 0.0065, "step": 49502 }, { "epoch": 15.23, "learning_rate": 2.8429663306366585e-06, "loss": 0.0064, "step": 49503 }, { "epoch": 15.23, "learning_rate": 2.842618345532766e-06, "loss": 0.0065, "step": 49504 }, { "epoch": 15.23, "learning_rate": 2.8422703781987226e-06, "loss": 0.0072, "step": 49505 }, { "epoch": 15.23, "learning_rate": 2.841922428635392e-06, "loss": 0.0122, "step": 49506 }, { "epoch": 15.23, "learning_rate": 2.8415744968436353e-06, "loss": 0.0079, "step": 49507 }, { "epoch": 15.23, "learning_rate": 2.8412265828243237e-06, "loss": 0.0076, "step": 49508 }, { "epoch": 15.23, "learning_rate": 2.840878686578318e-06, "loss": 0.0089, "step": 49509 }, { "epoch": 15.23, "learning_rate": 2.8405308081064777e-06, "loss": 0.007, "step": 49510 }, { "epoch": 15.23, "learning_rate": 2.8401829474096696e-06, "loss": 0.0137, "step": 49511 }, { "epoch": 15.23, "learning_rate": 2.839835104488754e-06, "loss": 0.0046, "step": 49512 }, { "epoch": 15.23, "learning_rate": 2.8394872793446003e-06, "loss": 0.0064, "step": 49513 }, { "epoch": 15.23, "learning_rate": 2.8391394719780696e-06, "loss": 0.0114, "step": 49514 }, { "epoch": 15.24, "learning_rate": 2.8387916823900243e-06, "loss": 0.009, "step": 49515 }, { "epoch": 15.24, "learning_rate": 2.838443910581329e-06, "loss": 0.0084, "step": 49516 }, { "epoch": 15.24, "learning_rate": 2.8380961565528474e-06, "loss": 0.0073, "step": 49517 }, { "epoch": 15.24, "learning_rate": 2.8377484203054416e-06, "loss": 0.0075, "step": 49518 }, { "epoch": 15.24, "learning_rate": 2.8374007018399764e-06, "loss": 0.0063, "step": 49519 }, { "epoch": 15.24, "learning_rate": 2.8370530011573126e-06, "loss": 0.0097, "step": 49520 }, { "epoch": 15.24, "learning_rate": 2.836705318258317e-06, "loss": 0.0044, "step": 49521 }, { "epoch": 15.24, "learning_rate": 2.836357653143846e-06, "loss": 0.0076, "step": 49522 }, { "epoch": 15.24, "learning_rate": 2.8360100058147756e-06, "loss": 0.0075, "step": 49523 }, { "epoch": 15.24, "learning_rate": 2.835662376271957e-06, "loss": 0.0056, "step": 49524 }, { "epoch": 15.24, "learning_rate": 2.835314764516258e-06, "loss": 0.0151, "step": 49525 }, { "epoch": 15.24, "learning_rate": 2.8349671705485403e-06, "loss": 0.0236, "step": 49526 }, { "epoch": 15.24, "learning_rate": 2.8346195943696643e-06, "loss": 0.0115, "step": 49527 }, { "epoch": 15.24, "learning_rate": 2.8342720359805e-06, "loss": 0.0069, "step": 49528 }, { "epoch": 15.24, "learning_rate": 2.8339244953819057e-06, "loss": 0.0246, "step": 49529 }, { "epoch": 15.24, "learning_rate": 2.8335769725747476e-06, "loss": 0.0117, "step": 49530 }, { "epoch": 15.24, "learning_rate": 2.833229467559884e-06, "loss": 0.0088, "step": 49531 }, { "epoch": 15.24, "learning_rate": 2.8328819803381757e-06, "loss": 0.0107, "step": 49532 }, { "epoch": 15.24, "learning_rate": 2.832534510910492e-06, "loss": 0.0072, "step": 49533 }, { "epoch": 15.24, "learning_rate": 2.8321870592776933e-06, "loss": 0.0045, "step": 49534 }, { "epoch": 15.24, "learning_rate": 2.831839625440642e-06, "loss": 0.0078, "step": 49535 }, { "epoch": 15.24, "learning_rate": 2.8314922094001997e-06, "loss": 0.0145, "step": 49536 }, { "epoch": 15.24, "learning_rate": 2.8311448111572304e-06, "loss": 0.0143, "step": 49537 }, { "epoch": 15.24, "learning_rate": 2.830797430712595e-06, "loss": 0.0043, "step": 49538 }, { "epoch": 15.24, "learning_rate": 2.8304500680671574e-06, "loss": 0.0088, "step": 49539 }, { "epoch": 15.24, "learning_rate": 2.8301027232217804e-06, "loss": 0.0067, "step": 49540 }, { "epoch": 15.24, "learning_rate": 2.8297553961773238e-06, "loss": 0.0058, "step": 49541 }, { "epoch": 15.24, "learning_rate": 2.8294080869346495e-06, "loss": 0.0089, "step": 49542 }, { "epoch": 15.24, "learning_rate": 2.829060795494625e-06, "loss": 0.014, "step": 49543 }, { "epoch": 15.24, "learning_rate": 2.828713521858111e-06, "loss": 0.009, "step": 49544 }, { "epoch": 15.24, "learning_rate": 2.828366266025966e-06, "loss": 0.0102, "step": 49545 }, { "epoch": 15.24, "learning_rate": 2.8280190279990537e-06, "loss": 0.006, "step": 49546 }, { "epoch": 15.25, "learning_rate": 2.827671807778234e-06, "loss": 0.0115, "step": 49547 }, { "epoch": 15.25, "learning_rate": 2.8273246053643743e-06, "loss": 0.0065, "step": 49548 }, { "epoch": 15.25, "learning_rate": 2.8269774207583347e-06, "loss": 0.0098, "step": 49549 }, { "epoch": 15.25, "learning_rate": 2.8266302539609747e-06, "loss": 0.006, "step": 49550 }, { "epoch": 15.25, "learning_rate": 2.826283104973162e-06, "loss": 0.0112, "step": 49551 }, { "epoch": 15.25, "learning_rate": 2.825935973795748e-06, "loss": 0.011, "step": 49552 }, { "epoch": 15.25, "learning_rate": 2.8255888604296044e-06, "loss": 0.0125, "step": 49553 }, { "epoch": 15.25, "learning_rate": 2.82524176487559e-06, "loss": 0.0074, "step": 49554 }, { "epoch": 15.25, "learning_rate": 2.824894687134565e-06, "loss": 0.0075, "step": 49555 }, { "epoch": 15.25, "learning_rate": 2.8245476272073925e-06, "loss": 0.0086, "step": 49556 }, { "epoch": 15.25, "learning_rate": 2.824200585094934e-06, "loss": 0.0066, "step": 49557 }, { "epoch": 15.25, "learning_rate": 2.823853560798051e-06, "loss": 0.0123, "step": 49558 }, { "epoch": 15.25, "learning_rate": 2.8235065543176043e-06, "loss": 0.0105, "step": 49559 }, { "epoch": 15.25, "learning_rate": 2.8231595656544563e-06, "loss": 0.01, "step": 49560 }, { "epoch": 15.25, "learning_rate": 2.8228125948094688e-06, "loss": 0.0067, "step": 49561 }, { "epoch": 15.25, "learning_rate": 2.8224656417834994e-06, "loss": 0.0071, "step": 49562 }, { "epoch": 15.25, "learning_rate": 2.822118706577416e-06, "loss": 0.0087, "step": 49563 }, { "epoch": 15.25, "learning_rate": 2.821771789192076e-06, "loss": 0.0203, "step": 49564 }, { "epoch": 15.25, "learning_rate": 2.8214248896283457e-06, "loss": 0.0083, "step": 49565 }, { "epoch": 15.25, "learning_rate": 2.821078007887078e-06, "loss": 0.0108, "step": 49566 }, { "epoch": 15.25, "learning_rate": 2.8207311439691352e-06, "loss": 0.0067, "step": 49567 }, { "epoch": 15.25, "learning_rate": 2.820384297875386e-06, "loss": 0.0103, "step": 49568 }, { "epoch": 15.25, "learning_rate": 2.8200374696066856e-06, "loss": 0.007, "step": 49569 }, { "epoch": 15.25, "learning_rate": 2.8196906591638975e-06, "loss": 0.015, "step": 49570 }, { "epoch": 15.25, "learning_rate": 2.81934386654788e-06, "loss": 0.0094, "step": 49571 }, { "epoch": 15.25, "learning_rate": 2.8189970917594978e-06, "loss": 0.0077, "step": 49572 }, { "epoch": 15.25, "learning_rate": 2.8186503347996087e-06, "loss": 0.0147, "step": 49573 }, { "epoch": 15.25, "learning_rate": 2.8183035956690753e-06, "loss": 0.0071, "step": 49574 }, { "epoch": 15.25, "learning_rate": 2.817956874368758e-06, "loss": 0.0136, "step": 49575 }, { "epoch": 15.25, "learning_rate": 2.8176101708995174e-06, "loss": 0.0081, "step": 49576 }, { "epoch": 15.25, "learning_rate": 2.8172634852622106e-06, "loss": 0.0092, "step": 49577 }, { "epoch": 15.25, "learning_rate": 2.8169168174577086e-06, "loss": 0.0036, "step": 49578 }, { "epoch": 15.25, "learning_rate": 2.816570167486864e-06, "loss": 0.0055, "step": 49579 }, { "epoch": 15.26, "learning_rate": 2.8162235353505376e-06, "loss": 0.0072, "step": 49580 }, { "epoch": 15.26, "learning_rate": 2.8158769210495917e-06, "loss": 0.0357, "step": 49581 }, { "epoch": 15.26, "learning_rate": 2.815530324584884e-06, "loss": 0.0063, "step": 49582 }, { "epoch": 15.26, "learning_rate": 2.8151837459572806e-06, "loss": 0.012, "step": 49583 }, { "epoch": 15.26, "learning_rate": 2.814837185167639e-06, "loss": 0.0081, "step": 49584 }, { "epoch": 15.26, "learning_rate": 2.8144906422168227e-06, "loss": 0.011, "step": 49585 }, { "epoch": 15.26, "learning_rate": 2.814144117105686e-06, "loss": 0.0097, "step": 49586 }, { "epoch": 15.26, "learning_rate": 2.8137976098350883e-06, "loss": 0.0052, "step": 49587 }, { "epoch": 15.26, "learning_rate": 2.8134511204058977e-06, "loss": 0.0097, "step": 49588 }, { "epoch": 15.26, "learning_rate": 2.8131046488189705e-06, "loss": 0.0196, "step": 49589 }, { "epoch": 15.26, "learning_rate": 2.812758195075167e-06, "loss": 0.0068, "step": 49590 }, { "epoch": 15.26, "learning_rate": 2.8124117591753464e-06, "loss": 0.0102, "step": 49591 }, { "epoch": 15.26, "learning_rate": 2.8120653411203703e-06, "loss": 0.0145, "step": 49592 }, { "epoch": 15.26, "learning_rate": 2.811718940911098e-06, "loss": 0.0173, "step": 49593 }, { "epoch": 15.26, "learning_rate": 2.81137255854839e-06, "loss": 0.0089, "step": 49594 }, { "epoch": 15.26, "learning_rate": 2.811026194033106e-06, "loss": 0.0099, "step": 49595 }, { "epoch": 15.26, "learning_rate": 2.8106798473661057e-06, "loss": 0.0057, "step": 49596 }, { "epoch": 15.26, "learning_rate": 2.810333518548246e-06, "loss": 0.0083, "step": 49597 }, { "epoch": 15.26, "learning_rate": 2.8099872075803925e-06, "loss": 0.0055, "step": 49598 }, { "epoch": 15.26, "learning_rate": 2.8096409144634064e-06, "loss": 0.0111, "step": 49599 }, { "epoch": 15.26, "learning_rate": 2.80929463919814e-06, "loss": 0.0047, "step": 49600 }, { "epoch": 15.26, "learning_rate": 2.808948381785457e-06, "loss": 0.008, "step": 49601 }, { "epoch": 15.26, "learning_rate": 2.808602142226212e-06, "loss": 0.0253, "step": 49602 }, { "epoch": 15.26, "learning_rate": 2.8082559205212734e-06, "loss": 0.0115, "step": 49603 }, { "epoch": 15.26, "learning_rate": 2.8079097166714963e-06, "loss": 0.0116, "step": 49604 }, { "epoch": 15.26, "learning_rate": 2.8075635306777404e-06, "loss": 0.0083, "step": 49605 }, { "epoch": 15.26, "learning_rate": 2.8072173625408683e-06, "loss": 0.0091, "step": 49606 }, { "epoch": 15.26, "learning_rate": 2.80687121226173e-06, "loss": 0.0114, "step": 49607 }, { "epoch": 15.26, "learning_rate": 2.8065250798411945e-06, "loss": 0.0087, "step": 49608 }, { "epoch": 15.26, "learning_rate": 2.8061789652801173e-06, "loss": 0.0057, "step": 49609 }, { "epoch": 15.26, "learning_rate": 2.8058328685793588e-06, "loss": 0.0103, "step": 49610 }, { "epoch": 15.26, "learning_rate": 2.8054867897397786e-06, "loss": 0.0116, "step": 49611 }, { "epoch": 15.27, "learning_rate": 2.8051407287622343e-06, "loss": 0.0056, "step": 49612 }, { "epoch": 15.27, "learning_rate": 2.8047946856475862e-06, "loss": 0.009, "step": 49613 }, { "epoch": 15.27, "learning_rate": 2.8044486603966936e-06, "loss": 0.0155, "step": 49614 }, { "epoch": 15.27, "learning_rate": 2.8041026530104144e-06, "loss": 0.0104, "step": 49615 }, { "epoch": 15.27, "learning_rate": 2.8037566634896085e-06, "loss": 0.0068, "step": 49616 }, { "epoch": 15.27, "learning_rate": 2.8034106918351313e-06, "loss": 0.0106, "step": 49617 }, { "epoch": 15.27, "learning_rate": 2.803064738047848e-06, "loss": 0.0061, "step": 49618 }, { "epoch": 15.27, "learning_rate": 2.8027188021286157e-06, "loss": 0.0061, "step": 49619 }, { "epoch": 15.27, "learning_rate": 2.802372884078295e-06, "loss": 0.0145, "step": 49620 }, { "epoch": 15.27, "learning_rate": 2.80202698389774e-06, "loss": 0.0059, "step": 49621 }, { "epoch": 15.27, "learning_rate": 2.8016811015878063e-06, "loss": 0.0111, "step": 49622 }, { "epoch": 15.27, "learning_rate": 2.801335237149363e-06, "loss": 0.01, "step": 49623 }, { "epoch": 15.27, "learning_rate": 2.800989390583262e-06, "loss": 0.0076, "step": 49624 }, { "epoch": 15.27, "learning_rate": 2.8006435618903636e-06, "loss": 0.0076, "step": 49625 }, { "epoch": 15.27, "learning_rate": 2.800297751071528e-06, "loss": 0.0075, "step": 49626 }, { "epoch": 15.27, "learning_rate": 2.79995195812761e-06, "loss": 0.0064, "step": 49627 }, { "epoch": 15.27, "learning_rate": 2.7996061830594714e-06, "loss": 0.0139, "step": 49628 }, { "epoch": 15.27, "learning_rate": 2.7992604258679703e-06, "loss": 0.005, "step": 49629 }, { "epoch": 15.27, "learning_rate": 2.798914686553963e-06, "loss": 0.0076, "step": 49630 }, { "epoch": 15.27, "learning_rate": 2.798568965118309e-06, "loss": 0.0069, "step": 49631 }, { "epoch": 15.27, "learning_rate": 2.7982232615618642e-06, "loss": 0.0105, "step": 49632 }, { "epoch": 15.27, "learning_rate": 2.797877575885496e-06, "loss": 0.0053, "step": 49633 }, { "epoch": 15.27, "learning_rate": 2.797531908090052e-06, "loss": 0.0101, "step": 49634 }, { "epoch": 15.27, "learning_rate": 2.7971862581763955e-06, "loss": 0.0057, "step": 49635 }, { "epoch": 15.27, "learning_rate": 2.796840626145384e-06, "loss": 0.0144, "step": 49636 }, { "epoch": 15.27, "learning_rate": 2.7964950119978718e-06, "loss": 0.0084, "step": 49637 }, { "epoch": 15.27, "learning_rate": 2.7961494157347236e-06, "loss": 0.0083, "step": 49638 }, { "epoch": 15.27, "learning_rate": 2.7958038373567942e-06, "loss": 0.0074, "step": 49639 }, { "epoch": 15.27, "learning_rate": 2.795458276864941e-06, "loss": 0.0074, "step": 49640 }, { "epoch": 15.27, "learning_rate": 2.795112734260026e-06, "loss": 0.008, "step": 49641 }, { "epoch": 15.27, "learning_rate": 2.794767209542898e-06, "loss": 0.0065, "step": 49642 }, { "epoch": 15.27, "learning_rate": 2.7944217027144236e-06, "loss": 0.0236, "step": 49643 }, { "epoch": 15.27, "learning_rate": 2.7940762137754575e-06, "loss": 0.0093, "step": 49644 }, { "epoch": 15.28, "learning_rate": 2.793730742726858e-06, "loss": 0.007, "step": 49645 }, { "epoch": 15.28, "learning_rate": 2.793385289569481e-06, "loss": 0.0058, "step": 49646 }, { "epoch": 15.28, "learning_rate": 2.793039854304187e-06, "loss": 0.0094, "step": 49647 }, { "epoch": 15.28, "learning_rate": 2.792694436931831e-06, "loss": 0.0115, "step": 49648 }, { "epoch": 15.28, "learning_rate": 2.7923490374532726e-06, "loss": 0.0088, "step": 49649 }, { "epoch": 15.28, "learning_rate": 2.7920036558693674e-06, "loss": 0.0063, "step": 49650 }, { "epoch": 15.28, "learning_rate": 2.7916582921809743e-06, "loss": 0.0043, "step": 49651 }, { "epoch": 15.28, "learning_rate": 2.791312946388948e-06, "loss": 0.0171, "step": 49652 }, { "epoch": 15.28, "learning_rate": 2.7909676184941502e-06, "loss": 0.0159, "step": 49653 }, { "epoch": 15.28, "learning_rate": 2.7906223084974405e-06, "loss": 0.0083, "step": 49654 }, { "epoch": 15.28, "learning_rate": 2.7902770163996673e-06, "loss": 0.0117, "step": 49655 }, { "epoch": 15.28, "learning_rate": 2.7899317422016937e-06, "loss": 0.0072, "step": 49656 }, { "epoch": 15.28, "learning_rate": 2.789586485904373e-06, "loss": 0.0051, "step": 49657 }, { "epoch": 15.28, "learning_rate": 2.7892412475085685e-06, "loss": 0.007, "step": 49658 }, { "epoch": 15.28, "learning_rate": 2.7888960270151333e-06, "loss": 0.0072, "step": 49659 }, { "epoch": 15.28, "learning_rate": 2.788550824424925e-06, "loss": 0.0086, "step": 49660 }, { "epoch": 15.28, "learning_rate": 2.7882056397388013e-06, "loss": 0.0136, "step": 49661 }, { "epoch": 15.28, "learning_rate": 2.787860472957619e-06, "loss": 0.0049, "step": 49662 }, { "epoch": 15.28, "learning_rate": 2.787515324082234e-06, "loss": 0.0068, "step": 49663 }, { "epoch": 15.28, "learning_rate": 2.787170193113504e-06, "loss": 0.0094, "step": 49664 }, { "epoch": 15.28, "learning_rate": 2.7868250800522866e-06, "loss": 0.0103, "step": 49665 }, { "epoch": 15.28, "learning_rate": 2.786479984899438e-06, "loss": 0.0153, "step": 49666 }, { "epoch": 15.28, "learning_rate": 2.786134907655814e-06, "loss": 0.0053, "step": 49667 }, { "epoch": 15.28, "learning_rate": 2.7857898483222733e-06, "loss": 0.0056, "step": 49668 }, { "epoch": 15.28, "learning_rate": 2.7854448068996707e-06, "loss": 0.0059, "step": 49669 }, { "epoch": 15.28, "learning_rate": 2.785099783388864e-06, "loss": 0.0164, "step": 49670 }, { "epoch": 15.28, "learning_rate": 2.78475477779071e-06, "loss": 0.0057, "step": 49671 }, { "epoch": 15.28, "learning_rate": 2.784409790106061e-06, "loss": 0.0056, "step": 49672 }, { "epoch": 15.28, "learning_rate": 2.78406482033578e-06, "loss": 0.0076, "step": 49673 }, { "epoch": 15.28, "learning_rate": 2.783719868480721e-06, "loss": 0.0142, "step": 49674 }, { "epoch": 15.28, "learning_rate": 2.783374934541744e-06, "loss": 0.0087, "step": 49675 }, { "epoch": 15.28, "learning_rate": 2.7830300185196968e-06, "loss": 0.0081, "step": 49676 }, { "epoch": 15.29, "learning_rate": 2.7826851204154395e-06, "loss": 0.0089, "step": 49677 }, { "epoch": 15.29, "learning_rate": 2.7823402402298304e-06, "loss": 0.0095, "step": 49678 }, { "epoch": 15.29, "learning_rate": 2.7819953779637256e-06, "loss": 0.0055, "step": 49679 }, { "epoch": 15.29, "learning_rate": 2.78165053361798e-06, "loss": 0.0064, "step": 49680 }, { "epoch": 15.29, "learning_rate": 2.78130570719345e-06, "loss": 0.0129, "step": 49681 }, { "epoch": 15.29, "learning_rate": 2.780960898690992e-06, "loss": 0.0063, "step": 49682 }, { "epoch": 15.29, "learning_rate": 2.7806161081114624e-06, "loss": 0.0069, "step": 49683 }, { "epoch": 15.29, "learning_rate": 2.7802713354557166e-06, "loss": 0.0073, "step": 49684 }, { "epoch": 15.29, "learning_rate": 2.77992658072461e-06, "loss": 0.0103, "step": 49685 }, { "epoch": 15.29, "learning_rate": 2.779581843919e-06, "loss": 0.0108, "step": 49686 }, { "epoch": 15.29, "learning_rate": 2.7792371250397376e-06, "loss": 0.0112, "step": 49687 }, { "epoch": 15.29, "learning_rate": 2.77889242408769e-06, "loss": 0.0106, "step": 49688 }, { "epoch": 15.29, "learning_rate": 2.778547741063703e-06, "loss": 0.005, "step": 49689 }, { "epoch": 15.29, "learning_rate": 2.778203075968634e-06, "loss": 0.0102, "step": 49690 }, { "epoch": 15.29, "learning_rate": 2.7778584288033405e-06, "loss": 0.0119, "step": 49691 }, { "epoch": 15.29, "learning_rate": 2.777513799568674e-06, "loss": 0.0082, "step": 49692 }, { "epoch": 15.29, "learning_rate": 2.777169188265496e-06, "loss": 0.0054, "step": 49693 }, { "epoch": 15.29, "learning_rate": 2.7768245948946615e-06, "loss": 0.0068, "step": 49694 }, { "epoch": 15.29, "learning_rate": 2.776480019457023e-06, "loss": 0.0115, "step": 49695 }, { "epoch": 15.29, "learning_rate": 2.776135461953441e-06, "loss": 0.0098, "step": 49696 }, { "epoch": 15.29, "learning_rate": 2.77579092238476e-06, "loss": 0.0076, "step": 49697 }, { "epoch": 15.29, "learning_rate": 2.7754464007518467e-06, "loss": 0.0106, "step": 49698 }, { "epoch": 15.29, "learning_rate": 2.775101897055551e-06, "loss": 0.0069, "step": 49699 }, { "epoch": 15.29, "learning_rate": 2.7747574112967303e-06, "loss": 0.0052, "step": 49700 }, { "epoch": 15.29, "learning_rate": 2.774412943476239e-06, "loss": 0.0114, "step": 49701 }, { "epoch": 15.29, "learning_rate": 2.7740684935949324e-06, "loss": 0.0086, "step": 49702 }, { "epoch": 15.29, "learning_rate": 2.7737240616536665e-06, "loss": 0.0051, "step": 49703 }, { "epoch": 15.29, "learning_rate": 2.773379647653295e-06, "loss": 0.0066, "step": 49704 }, { "epoch": 15.29, "learning_rate": 2.7730352515946735e-06, "loss": 0.0046, "step": 49705 }, { "epoch": 15.29, "learning_rate": 2.772690873478656e-06, "loss": 0.0021, "step": 49706 }, { "epoch": 15.29, "learning_rate": 2.772346513306098e-06, "loss": 0.0073, "step": 49707 }, { "epoch": 15.29, "learning_rate": 2.7720021710778565e-06, "loss": 0.0068, "step": 49708 }, { "epoch": 15.29, "learning_rate": 2.7716578467947886e-06, "loss": 0.0278, "step": 49709 }, { "epoch": 15.3, "learning_rate": 2.7713135404577417e-06, "loss": 0.01, "step": 49710 }, { "epoch": 15.3, "learning_rate": 2.7709692520675755e-06, "loss": 0.0067, "step": 49711 }, { "epoch": 15.3, "learning_rate": 2.7706249816251396e-06, "loss": 0.0057, "step": 49712 }, { "epoch": 15.3, "learning_rate": 2.770280729131297e-06, "loss": 0.0079, "step": 49713 }, { "epoch": 15.3, "learning_rate": 2.7699364945868977e-06, "loss": 0.0108, "step": 49714 }, { "epoch": 15.3, "learning_rate": 2.7695922779927975e-06, "loss": 0.0082, "step": 49715 }, { "epoch": 15.3, "learning_rate": 2.76924807934985e-06, "loss": 0.0122, "step": 49716 }, { "epoch": 15.3, "learning_rate": 2.76890389865891e-06, "loss": 0.0144, "step": 49717 }, { "epoch": 15.3, "learning_rate": 2.7685597359208326e-06, "loss": 0.0076, "step": 49718 }, { "epoch": 15.3, "learning_rate": 2.7682155911364718e-06, "loss": 0.0073, "step": 49719 }, { "epoch": 15.3, "learning_rate": 2.767871464306682e-06, "loss": 0.0078, "step": 49720 }, { "epoch": 15.3, "learning_rate": 2.767527355432318e-06, "loss": 0.0185, "step": 49721 }, { "epoch": 15.3, "learning_rate": 2.7671832645142327e-06, "loss": 0.0053, "step": 49722 }, { "epoch": 15.3, "learning_rate": 2.7668391915532822e-06, "loss": 0.0118, "step": 49723 }, { "epoch": 15.3, "learning_rate": 2.76649513655032e-06, "loss": 0.006, "step": 49724 }, { "epoch": 15.3, "learning_rate": 2.7661510995062e-06, "loss": 0.0061, "step": 49725 }, { "epoch": 15.3, "learning_rate": 2.7658070804217764e-06, "loss": 0.0097, "step": 49726 }, { "epoch": 15.3, "learning_rate": 2.765463079297901e-06, "loss": 0.0044, "step": 49727 }, { "epoch": 15.3, "learning_rate": 2.7651190961354336e-06, "loss": 0.0039, "step": 49728 }, { "epoch": 15.3, "learning_rate": 2.7647751309352243e-06, "loss": 0.0088, "step": 49729 }, { "epoch": 15.3, "learning_rate": 2.764431183698131e-06, "loss": 0.0086, "step": 49730 }, { "epoch": 15.3, "learning_rate": 2.764087254425002e-06, "loss": 0.0161, "step": 49731 }, { "epoch": 15.3, "learning_rate": 2.7637433431166903e-06, "loss": 0.0089, "step": 49732 }, { "epoch": 15.3, "learning_rate": 2.7633994497740557e-06, "loss": 0.0076, "step": 49733 }, { "epoch": 15.3, "learning_rate": 2.763055574397949e-06, "loss": 0.0117, "step": 49734 }, { "epoch": 15.3, "learning_rate": 2.762711716989226e-06, "loss": 0.0051, "step": 49735 }, { "epoch": 15.3, "learning_rate": 2.7623678775487374e-06, "loss": 0.0066, "step": 49736 }, { "epoch": 15.3, "learning_rate": 2.7620240560773383e-06, "loss": 0.0063, "step": 49737 }, { "epoch": 15.3, "learning_rate": 2.761680252575882e-06, "loss": 0.005, "step": 49738 }, { "epoch": 15.3, "learning_rate": 2.7613364670452226e-06, "loss": 0.0059, "step": 49739 }, { "epoch": 15.3, "learning_rate": 2.7609926994862136e-06, "loss": 0.01, "step": 49740 }, { "epoch": 15.3, "learning_rate": 2.760648949899707e-06, "loss": 0.0068, "step": 49741 }, { "epoch": 15.31, "learning_rate": 2.760305218286555e-06, "loss": 0.0046, "step": 49742 }, { "epoch": 15.31, "learning_rate": 2.75996150464762e-06, "loss": 0.0113, "step": 49743 }, { "epoch": 15.31, "learning_rate": 2.7596178089837445e-06, "loss": 0.0161, "step": 49744 }, { "epoch": 15.31, "learning_rate": 2.759274131295787e-06, "loss": 0.0053, "step": 49745 }, { "epoch": 15.31, "learning_rate": 2.7589304715845998e-06, "loss": 0.0045, "step": 49746 }, { "epoch": 15.31, "learning_rate": 2.758586829851033e-06, "loss": 0.0075, "step": 49747 }, { "epoch": 15.31, "learning_rate": 2.758243206095945e-06, "loss": 0.0107, "step": 49748 }, { "epoch": 15.31, "learning_rate": 2.7578996003201875e-06, "loss": 0.0128, "step": 49749 }, { "epoch": 15.31, "learning_rate": 2.757556012524614e-06, "loss": 0.0111, "step": 49750 }, { "epoch": 15.31, "learning_rate": 2.7572124427100775e-06, "loss": 0.0105, "step": 49751 }, { "epoch": 15.31, "learning_rate": 2.7568688908774243e-06, "loss": 0.0083, "step": 49752 }, { "epoch": 15.31, "learning_rate": 2.7565253570275165e-06, "loss": 0.0086, "step": 49753 }, { "epoch": 15.31, "learning_rate": 2.756181841161204e-06, "loss": 0.0088, "step": 49754 }, { "epoch": 15.31, "learning_rate": 2.7558383432793377e-06, "loss": 0.0104, "step": 49755 }, { "epoch": 15.31, "learning_rate": 2.7554948633827727e-06, "loss": 0.0087, "step": 49756 }, { "epoch": 15.31, "learning_rate": 2.755151401472361e-06, "loss": 0.005, "step": 49757 }, { "epoch": 15.31, "learning_rate": 2.754807957548955e-06, "loss": 0.007, "step": 49758 }, { "epoch": 15.31, "learning_rate": 2.754464531613408e-06, "loss": 0.0091, "step": 49759 }, { "epoch": 15.31, "learning_rate": 2.754121123666571e-06, "loss": 0.0057, "step": 49760 }, { "epoch": 15.31, "learning_rate": 2.7537777337092984e-06, "loss": 0.0087, "step": 49761 }, { "epoch": 15.31, "learning_rate": 2.75343436174244e-06, "loss": 0.0076, "step": 49762 }, { "epoch": 15.31, "learning_rate": 2.753091007766853e-06, "loss": 0.0083, "step": 49763 }, { "epoch": 15.31, "learning_rate": 2.75274767178339e-06, "loss": 0.0079, "step": 49764 }, { "epoch": 15.31, "learning_rate": 2.752404353792898e-06, "loss": 0.0106, "step": 49765 }, { "epoch": 15.31, "learning_rate": 2.752061053796232e-06, "loss": 0.0076, "step": 49766 }, { "epoch": 15.31, "learning_rate": 2.7517177717942412e-06, "loss": 0.0077, "step": 49767 }, { "epoch": 15.31, "learning_rate": 2.751374507787785e-06, "loss": 0.0098, "step": 49768 }, { "epoch": 15.31, "learning_rate": 2.7510312617777103e-06, "loss": 0.0115, "step": 49769 }, { "epoch": 15.31, "learning_rate": 2.7506880337648723e-06, "loss": 0.0067, "step": 49770 }, { "epoch": 15.31, "learning_rate": 2.7503448237501208e-06, "loss": 0.0059, "step": 49771 }, { "epoch": 15.31, "learning_rate": 2.750001631734308e-06, "loss": 0.0066, "step": 49772 }, { "epoch": 15.31, "learning_rate": 2.7496584577182874e-06, "loss": 0.0141, "step": 49773 }, { "epoch": 15.31, "learning_rate": 2.7493153017029094e-06, "loss": 0.0173, "step": 49774 }, { "epoch": 15.32, "learning_rate": 2.7489721636890276e-06, "loss": 0.0086, "step": 49775 }, { "epoch": 15.32, "learning_rate": 2.748629043677493e-06, "loss": 0.0045, "step": 49776 }, { "epoch": 15.32, "learning_rate": 2.7482859416691545e-06, "loss": 0.0058, "step": 49777 }, { "epoch": 15.32, "learning_rate": 2.7479428576648738e-06, "loss": 0.0079, "step": 49778 }, { "epoch": 15.32, "learning_rate": 2.7475997916654917e-06, "loss": 0.0099, "step": 49779 }, { "epoch": 15.32, "learning_rate": 2.7472567436718642e-06, "loss": 0.0105, "step": 49780 }, { "epoch": 15.32, "learning_rate": 2.7469137136848435e-06, "loss": 0.0093, "step": 49781 }, { "epoch": 15.32, "learning_rate": 2.746570701705277e-06, "loss": 0.0071, "step": 49782 }, { "epoch": 15.32, "learning_rate": 2.746227707734024e-06, "loss": 0.0069, "step": 49783 }, { "epoch": 15.32, "learning_rate": 2.745884731771931e-06, "loss": 0.018, "step": 49784 }, { "epoch": 15.32, "learning_rate": 2.7455417738198542e-06, "loss": 0.0084, "step": 49785 }, { "epoch": 15.32, "learning_rate": 2.7451988338786386e-06, "loss": 0.0096, "step": 49786 }, { "epoch": 15.32, "learning_rate": 2.7448559119491347e-06, "loss": 0.0109, "step": 49787 }, { "epoch": 15.32, "learning_rate": 2.744513008032201e-06, "loss": 0.0107, "step": 49788 }, { "epoch": 15.32, "learning_rate": 2.7441701221286852e-06, "loss": 0.0042, "step": 49789 }, { "epoch": 15.32, "learning_rate": 2.7438272542394396e-06, "loss": 0.0085, "step": 49790 }, { "epoch": 15.32, "learning_rate": 2.743484404365314e-06, "loss": 0.0103, "step": 49791 }, { "epoch": 15.32, "learning_rate": 2.743141572507161e-06, "loss": 0.008, "step": 49792 }, { "epoch": 15.32, "learning_rate": 2.74279875866583e-06, "loss": 0.0062, "step": 49793 }, { "epoch": 15.32, "learning_rate": 2.7424559628421745e-06, "loss": 0.008, "step": 49794 }, { "epoch": 15.32, "learning_rate": 2.742113185037044e-06, "loss": 0.0108, "step": 49795 }, { "epoch": 15.32, "learning_rate": 2.7417704252512888e-06, "loss": 0.0063, "step": 49796 }, { "epoch": 15.32, "learning_rate": 2.741427683485759e-06, "loss": 0.0095, "step": 49797 }, { "epoch": 15.32, "learning_rate": 2.74108495974131e-06, "loss": 0.0075, "step": 49798 }, { "epoch": 15.32, "learning_rate": 2.740742254018793e-06, "loss": 0.014, "step": 49799 }, { "epoch": 15.32, "learning_rate": 2.740399566319053e-06, "loss": 0.0075, "step": 49800 }, { "epoch": 15.32, "learning_rate": 2.7400568966429443e-06, "loss": 0.0111, "step": 49801 }, { "epoch": 15.32, "learning_rate": 2.739714244991313e-06, "loss": 0.0081, "step": 49802 }, { "epoch": 15.32, "learning_rate": 2.739371611365017e-06, "loss": 0.0043, "step": 49803 }, { "epoch": 15.32, "learning_rate": 2.7390289957649053e-06, "loss": 0.0088, "step": 49804 }, { "epoch": 15.32, "learning_rate": 2.7386863981918256e-06, "loss": 0.0067, "step": 49805 }, { "epoch": 15.32, "learning_rate": 2.738343818646634e-06, "loss": 0.0087, "step": 49806 }, { "epoch": 15.33, "learning_rate": 2.7380012571301704e-06, "loss": 0.0043, "step": 49807 }, { "epoch": 15.33, "learning_rate": 2.7376587136432954e-06, "loss": 0.0072, "step": 49808 }, { "epoch": 15.33, "learning_rate": 2.7373161881868558e-06, "loss": 0.0072, "step": 49809 }, { "epoch": 15.33, "learning_rate": 2.736973680761702e-06, "loss": 0.0104, "step": 49810 }, { "epoch": 15.33, "learning_rate": 2.7366311913686837e-06, "loss": 0.0068, "step": 49811 }, { "epoch": 15.33, "learning_rate": 2.736288720008653e-06, "loss": 0.008, "step": 49812 }, { "epoch": 15.33, "learning_rate": 2.735946266682459e-06, "loss": 0.01, "step": 49813 }, { "epoch": 15.33, "learning_rate": 2.735603831390953e-06, "loss": 0.0111, "step": 49814 }, { "epoch": 15.33, "learning_rate": 2.735261414134983e-06, "loss": 0.0053, "step": 49815 }, { "epoch": 15.33, "learning_rate": 2.7349190149154004e-06, "loss": 0.006, "step": 49816 }, { "epoch": 15.33, "learning_rate": 2.7345766337330527e-06, "loss": 0.0098, "step": 49817 }, { "epoch": 15.33, "learning_rate": 2.7342342705887946e-06, "loss": 0.0119, "step": 49818 }, { "epoch": 15.33, "learning_rate": 2.733891925483475e-06, "loss": 0.0042, "step": 49819 }, { "epoch": 15.33, "learning_rate": 2.7335495984179463e-06, "loss": 0.0108, "step": 49820 }, { "epoch": 15.33, "learning_rate": 2.7332072893930517e-06, "loss": 0.0083, "step": 49821 }, { "epoch": 15.33, "learning_rate": 2.7328649984096412e-06, "loss": 0.0112, "step": 49822 }, { "epoch": 15.33, "learning_rate": 2.7325227254685704e-06, "loss": 0.0046, "step": 49823 }, { "epoch": 15.33, "learning_rate": 2.7321804705706867e-06, "loss": 0.0097, "step": 49824 }, { "epoch": 15.33, "learning_rate": 2.7318382337168393e-06, "loss": 0.0116, "step": 49825 }, { "epoch": 15.33, "learning_rate": 2.7314960149078785e-06, "loss": 0.0088, "step": 49826 }, { "epoch": 15.33, "learning_rate": 2.7311538141446526e-06, "loss": 0.0182, "step": 49827 }, { "epoch": 15.33, "learning_rate": 2.730811631428013e-06, "loss": 0.0089, "step": 49828 }, { "epoch": 15.33, "learning_rate": 2.7304694667588085e-06, "loss": 0.0047, "step": 49829 }, { "epoch": 15.33, "learning_rate": 2.7301273201378884e-06, "loss": 0.0118, "step": 49830 }, { "epoch": 15.33, "learning_rate": 2.729785191566102e-06, "loss": 0.0143, "step": 49831 }, { "epoch": 15.33, "learning_rate": 2.7294430810442964e-06, "loss": 0.0048, "step": 49832 }, { "epoch": 15.33, "learning_rate": 2.7291009885733287e-06, "loss": 0.0044, "step": 49833 }, { "epoch": 15.33, "learning_rate": 2.7287589141540404e-06, "loss": 0.0086, "step": 49834 }, { "epoch": 15.33, "learning_rate": 2.7284168577872837e-06, "loss": 0.0145, "step": 49835 }, { "epoch": 15.33, "learning_rate": 2.728074819473908e-06, "loss": 0.0123, "step": 49836 }, { "epoch": 15.33, "learning_rate": 2.7277327992147582e-06, "loss": 0.0091, "step": 49837 }, { "epoch": 15.33, "learning_rate": 2.7273907970106916e-06, "loss": 0.0126, "step": 49838 }, { "epoch": 15.33, "learning_rate": 2.727048812862552e-06, "loss": 0.0163, "step": 49839 }, { "epoch": 15.34, "learning_rate": 2.7267068467711898e-06, "loss": 0.0076, "step": 49840 }, { "epoch": 15.34, "learning_rate": 2.7263648987374567e-06, "loss": 0.0107, "step": 49841 }, { "epoch": 15.34, "learning_rate": 2.726022968762192e-06, "loss": 0.0078, "step": 49842 }, { "epoch": 15.34, "learning_rate": 2.7256810568462557e-06, "loss": 0.0077, "step": 49843 }, { "epoch": 15.34, "learning_rate": 2.7253391629904915e-06, "loss": 0.0064, "step": 49844 }, { "epoch": 15.34, "learning_rate": 2.7249972871957474e-06, "loss": 0.0082, "step": 49845 }, { "epoch": 15.34, "learning_rate": 2.724655429462876e-06, "loss": 0.0065, "step": 49846 }, { "epoch": 15.34, "learning_rate": 2.7243135897927233e-06, "loss": 0.0082, "step": 49847 }, { "epoch": 15.34, "learning_rate": 2.7239717681861378e-06, "loss": 0.0033, "step": 49848 }, { "epoch": 15.34, "learning_rate": 2.723629964643969e-06, "loss": 0.0087, "step": 49849 }, { "epoch": 15.34, "learning_rate": 2.7232881791670652e-06, "loss": 0.0061, "step": 49850 }, { "epoch": 15.34, "learning_rate": 2.722946411756273e-06, "loss": 0.0183, "step": 49851 }, { "epoch": 15.34, "learning_rate": 2.7226046624124447e-06, "loss": 0.0089, "step": 49852 }, { "epoch": 15.34, "learning_rate": 2.722262931136428e-06, "loss": 0.0081, "step": 49853 }, { "epoch": 15.34, "learning_rate": 2.721921217929073e-06, "loss": 0.0176, "step": 49854 }, { "epoch": 15.34, "learning_rate": 2.7215795227912236e-06, "loss": 0.0059, "step": 49855 }, { "epoch": 15.34, "learning_rate": 2.7212378457237264e-06, "loss": 0.013, "step": 49856 }, { "epoch": 15.34, "learning_rate": 2.7208961867274353e-06, "loss": 0.0071, "step": 49857 }, { "epoch": 15.34, "learning_rate": 2.7205545458031977e-06, "loss": 0.0063, "step": 49858 }, { "epoch": 15.34, "learning_rate": 2.7202129229518604e-06, "loss": 0.0044, "step": 49859 }, { "epoch": 15.34, "learning_rate": 2.719871318174272e-06, "loss": 0.003, "step": 49860 }, { "epoch": 15.34, "learning_rate": 2.719529731471281e-06, "loss": 0.0111, "step": 49861 }, { "epoch": 15.34, "learning_rate": 2.7191881628437335e-06, "loss": 0.0085, "step": 49862 }, { "epoch": 15.34, "learning_rate": 2.7188466122924794e-06, "loss": 0.0051, "step": 49863 }, { "epoch": 15.34, "learning_rate": 2.7185050798183666e-06, "loss": 0.008, "step": 49864 }, { "epoch": 15.34, "learning_rate": 2.7181635654222425e-06, "loss": 0.019, "step": 49865 }, { "epoch": 15.34, "learning_rate": 2.717822069104953e-06, "loss": 0.0095, "step": 49866 }, { "epoch": 15.34, "learning_rate": 2.7174805908673528e-06, "loss": 0.0076, "step": 49867 }, { "epoch": 15.34, "learning_rate": 2.7171391307102825e-06, "loss": 0.007, "step": 49868 }, { "epoch": 15.34, "learning_rate": 2.7167976886345925e-06, "loss": 0.0067, "step": 49869 }, { "epoch": 15.34, "learning_rate": 2.7164562646411306e-06, "loss": 0.0124, "step": 49870 }, { "epoch": 15.34, "learning_rate": 2.716114858730741e-06, "loss": 0.0071, "step": 49871 }, { "epoch": 15.35, "learning_rate": 2.715773470904277e-06, "loss": 0.0059, "step": 49872 }, { "epoch": 15.35, "learning_rate": 2.715432101162584e-06, "loss": 0.0059, "step": 49873 }, { "epoch": 15.35, "learning_rate": 2.7150907495065095e-06, "loss": 0.0055, "step": 49874 }, { "epoch": 15.35, "learning_rate": 2.714749415936904e-06, "loss": 0.0041, "step": 49875 }, { "epoch": 15.35, "learning_rate": 2.7144081004546042e-06, "loss": 0.0215, "step": 49876 }, { "epoch": 15.35, "learning_rate": 2.7140668030604688e-06, "loss": 0.0076, "step": 49877 }, { "epoch": 15.35, "learning_rate": 2.7137255237553415e-06, "loss": 0.0135, "step": 49878 }, { "epoch": 15.35, "learning_rate": 2.7133842625400695e-06, "loss": 0.0126, "step": 49879 }, { "epoch": 15.35, "learning_rate": 2.7130430194155e-06, "loss": 0.0048, "step": 49880 }, { "epoch": 15.35, "learning_rate": 2.7127017943824797e-06, "loss": 0.0069, "step": 49881 }, { "epoch": 15.35, "learning_rate": 2.7123605874418568e-06, "loss": 0.0067, "step": 49882 }, { "epoch": 15.35, "learning_rate": 2.7120193985944777e-06, "loss": 0.006, "step": 49883 }, { "epoch": 15.35, "learning_rate": 2.711678227841189e-06, "loss": 0.011, "step": 49884 }, { "epoch": 15.35, "learning_rate": 2.7113370751828394e-06, "loss": 0.0208, "step": 49885 }, { "epoch": 15.35, "learning_rate": 2.7109959406202713e-06, "loss": 0.0058, "step": 49886 }, { "epoch": 15.35, "learning_rate": 2.710654824154337e-06, "loss": 0.0058, "step": 49887 }, { "epoch": 15.35, "learning_rate": 2.7103137257858867e-06, "loss": 0.0093, "step": 49888 }, { "epoch": 15.35, "learning_rate": 2.709972645515758e-06, "loss": 0.0094, "step": 49889 }, { "epoch": 15.35, "learning_rate": 2.7096315833448027e-06, "loss": 0.0073, "step": 49890 }, { "epoch": 15.35, "learning_rate": 2.7092905392738633e-06, "loss": 0.0093, "step": 49891 }, { "epoch": 15.35, "learning_rate": 2.7089495133037934e-06, "loss": 0.0146, "step": 49892 }, { "epoch": 15.35, "learning_rate": 2.708608505435436e-06, "loss": 0.0104, "step": 49893 }, { "epoch": 15.35, "learning_rate": 2.708267515669638e-06, "loss": 0.0101, "step": 49894 }, { "epoch": 15.35, "learning_rate": 2.7079265440072456e-06, "loss": 0.0061, "step": 49895 }, { "epoch": 15.35, "learning_rate": 2.7075855904491055e-06, "loss": 0.0105, "step": 49896 }, { "epoch": 15.35, "learning_rate": 2.707244654996065e-06, "loss": 0.0105, "step": 49897 }, { "epoch": 15.35, "learning_rate": 2.7069037376489706e-06, "loss": 0.0057, "step": 49898 }, { "epoch": 15.35, "learning_rate": 2.7065628384086673e-06, "loss": 0.0086, "step": 49899 }, { "epoch": 15.35, "learning_rate": 2.706221957276002e-06, "loss": 0.0042, "step": 49900 }, { "epoch": 15.35, "learning_rate": 2.705881094251821e-06, "loss": 0.0046, "step": 49901 }, { "epoch": 15.35, "learning_rate": 2.705540249336971e-06, "loss": 0.0089, "step": 49902 }, { "epoch": 15.35, "learning_rate": 2.7051994225322986e-06, "loss": 0.0084, "step": 49903 }, { "epoch": 15.35, "learning_rate": 2.704858613838649e-06, "loss": 0.0082, "step": 49904 }, { "epoch": 15.36, "learning_rate": 2.704517823256869e-06, "loss": 0.0162, "step": 49905 }, { "epoch": 15.36, "learning_rate": 2.704177050787802e-06, "loss": 0.0129, "step": 49906 }, { "epoch": 15.36, "learning_rate": 2.7038362964322983e-06, "loss": 0.0057, "step": 49907 }, { "epoch": 15.36, "learning_rate": 2.703495560191203e-06, "loss": 0.0042, "step": 49908 }, { "epoch": 15.36, "learning_rate": 2.703154842065363e-06, "loss": 0.0095, "step": 49909 }, { "epoch": 15.36, "learning_rate": 2.7028141420556196e-06, "loss": 0.0097, "step": 49910 }, { "epoch": 15.36, "learning_rate": 2.70247346016282e-06, "loss": 0.0081, "step": 49911 }, { "epoch": 15.36, "learning_rate": 2.702132796387813e-06, "loss": 0.0062, "step": 49912 }, { "epoch": 15.36, "learning_rate": 2.701792150731443e-06, "loss": 0.0166, "step": 49913 }, { "epoch": 15.36, "learning_rate": 2.7014515231945557e-06, "loss": 0.0152, "step": 49914 }, { "epoch": 15.36, "learning_rate": 2.7011109137779967e-06, "loss": 0.0058, "step": 49915 }, { "epoch": 15.36, "learning_rate": 2.700770322482612e-06, "loss": 0.0062, "step": 49916 }, { "epoch": 15.36, "learning_rate": 2.7004297493092456e-06, "loss": 0.0057, "step": 49917 }, { "epoch": 15.36, "learning_rate": 2.700089194258745e-06, "loss": 0.0067, "step": 49918 }, { "epoch": 15.36, "learning_rate": 2.699748657331955e-06, "loss": 0.0045, "step": 49919 }, { "epoch": 15.36, "learning_rate": 2.699408138529721e-06, "loss": 0.0057, "step": 49920 }, { "epoch": 15.36, "learning_rate": 2.699067637852886e-06, "loss": 0.0089, "step": 49921 }, { "epoch": 15.36, "learning_rate": 2.698727155302302e-06, "loss": 0.0031, "step": 49922 }, { "epoch": 15.36, "learning_rate": 2.698386690878808e-06, "loss": 0.0099, "step": 49923 }, { "epoch": 15.36, "learning_rate": 2.6980462445832524e-06, "loss": 0.0083, "step": 49924 }, { "epoch": 15.36, "learning_rate": 2.6977058164164784e-06, "loss": 0.0069, "step": 49925 }, { "epoch": 15.36, "learning_rate": 2.6973654063793296e-06, "loss": 0.0034, "step": 49926 }, { "epoch": 15.36, "learning_rate": 2.6970250144726563e-06, "loss": 0.0058, "step": 49927 }, { "epoch": 15.36, "learning_rate": 2.6966846406973014e-06, "loss": 0.0067, "step": 49928 }, { "epoch": 15.36, "learning_rate": 2.69634428505411e-06, "loss": 0.0056, "step": 49929 }, { "epoch": 15.36, "learning_rate": 2.696003947543929e-06, "loss": 0.0054, "step": 49930 }, { "epoch": 15.36, "learning_rate": 2.6956636281675953e-06, "loss": 0.0066, "step": 49931 }, { "epoch": 15.36, "learning_rate": 2.695323326925963e-06, "loss": 0.0089, "step": 49932 }, { "epoch": 15.36, "learning_rate": 2.6949830438198733e-06, "loss": 0.0054, "step": 49933 }, { "epoch": 15.36, "learning_rate": 2.6946427788501716e-06, "loss": 0.0056, "step": 49934 }, { "epoch": 15.36, "learning_rate": 2.6943025320177018e-06, "loss": 0.0038, "step": 49935 }, { "epoch": 15.36, "learning_rate": 2.6939623033233096e-06, "loss": 0.0113, "step": 49936 }, { "epoch": 15.37, "learning_rate": 2.6936220927678405e-06, "loss": 0.0055, "step": 49937 }, { "epoch": 15.37, "learning_rate": 2.693281900352136e-06, "loss": 0.0098, "step": 49938 }, { "epoch": 15.37, "learning_rate": 2.692941726077044e-06, "loss": 0.0083, "step": 49939 }, { "epoch": 15.37, "learning_rate": 2.692601569943407e-06, "loss": 0.005, "step": 49940 }, { "epoch": 15.37, "learning_rate": 2.692261431952068e-06, "loss": 0.01, "step": 49941 }, { "epoch": 15.37, "learning_rate": 2.6919213121038765e-06, "loss": 0.01, "step": 49942 }, { "epoch": 15.37, "learning_rate": 2.6915812103996776e-06, "loss": 0.0084, "step": 49943 }, { "epoch": 15.37, "learning_rate": 2.6912411268403084e-06, "loss": 0.0128, "step": 49944 }, { "epoch": 15.37, "learning_rate": 2.6909010614266173e-06, "loss": 0.0134, "step": 49945 }, { "epoch": 15.37, "learning_rate": 2.690561014159445e-06, "loss": 0.0056, "step": 49946 }, { "epoch": 15.37, "learning_rate": 2.690220985039643e-06, "loss": 0.005, "step": 49947 }, { "epoch": 15.37, "learning_rate": 2.689880974068052e-06, "loss": 0.0199, "step": 49948 }, { "epoch": 15.37, "learning_rate": 2.689540981245515e-06, "loss": 0.0101, "step": 49949 }, { "epoch": 15.37, "learning_rate": 2.689201006572877e-06, "loss": 0.0087, "step": 49950 }, { "epoch": 15.37, "learning_rate": 2.6888610500509814e-06, "loss": 0.0159, "step": 49951 }, { "epoch": 15.37, "learning_rate": 2.688521111680674e-06, "loss": 0.0032, "step": 49952 }, { "epoch": 15.37, "learning_rate": 2.6881811914627976e-06, "loss": 0.0078, "step": 49953 }, { "epoch": 15.37, "learning_rate": 2.6878412893981954e-06, "loss": 0.008, "step": 49954 }, { "epoch": 15.37, "learning_rate": 2.6875014054877114e-06, "loss": 0.0046, "step": 49955 }, { "epoch": 15.37, "learning_rate": 2.687161539732188e-06, "loss": 0.0049, "step": 49956 }, { "epoch": 15.37, "learning_rate": 2.6868216921324766e-06, "loss": 0.0084, "step": 49957 }, { "epoch": 15.37, "learning_rate": 2.686481862689413e-06, "loss": 0.0109, "step": 49958 }, { "epoch": 15.37, "learning_rate": 2.6861420514038426e-06, "loss": 0.0097, "step": 49959 }, { "epoch": 15.37, "learning_rate": 2.6858022582766097e-06, "loss": 0.0062, "step": 49960 }, { "epoch": 15.37, "learning_rate": 2.6854624833085554e-06, "loss": 0.0109, "step": 49961 }, { "epoch": 15.37, "learning_rate": 2.6851227265005287e-06, "loss": 0.0107, "step": 49962 }, { "epoch": 15.37, "learning_rate": 2.6847829878533695e-06, "loss": 0.008, "step": 49963 }, { "epoch": 15.37, "learning_rate": 2.6844432673679253e-06, "loss": 0.014, "step": 49964 }, { "epoch": 15.37, "learning_rate": 2.684103565045034e-06, "loss": 0.0087, "step": 49965 }, { "epoch": 15.37, "learning_rate": 2.683763880885538e-06, "loss": 0.0075, "step": 49966 }, { "epoch": 15.37, "learning_rate": 2.6834242148902866e-06, "loss": 0.0077, "step": 49967 }, { "epoch": 15.37, "learning_rate": 2.68308456706012e-06, "loss": 0.0079, "step": 49968 }, { "epoch": 15.37, "learning_rate": 2.6827449373958825e-06, "loss": 0.0112, "step": 49969 }, { "epoch": 15.38, "learning_rate": 2.682405325898416e-06, "loss": 0.013, "step": 49970 }, { "epoch": 15.38, "learning_rate": 2.6820657325685653e-06, "loss": 0.0152, "step": 49971 }, { "epoch": 15.38, "learning_rate": 2.6817261574071717e-06, "loss": 0.015, "step": 49972 }, { "epoch": 15.38, "learning_rate": 2.68138660041508e-06, "loss": 0.0047, "step": 49973 }, { "epoch": 15.38, "learning_rate": 2.6810470615931316e-06, "loss": 0.0077, "step": 49974 }, { "epoch": 15.38, "learning_rate": 2.6807075409421703e-06, "loss": 0.0039, "step": 49975 }, { "epoch": 15.38, "learning_rate": 2.6803680384630357e-06, "loss": 0.0089, "step": 49976 }, { "epoch": 15.38, "learning_rate": 2.680028554156577e-06, "loss": 0.0067, "step": 49977 }, { "epoch": 15.38, "learning_rate": 2.6796890880236383e-06, "loss": 0.0108, "step": 49978 }, { "epoch": 15.38, "learning_rate": 2.679349640065053e-06, "loss": 0.0077, "step": 49979 }, { "epoch": 15.38, "learning_rate": 2.6790102102816707e-06, "loss": 0.0073, "step": 49980 }, { "epoch": 15.38, "learning_rate": 2.678670798674329e-06, "loss": 0.0077, "step": 49981 }, { "epoch": 15.38, "learning_rate": 2.6783314052438758e-06, "loss": 0.0063, "step": 49982 }, { "epoch": 15.38, "learning_rate": 2.6779920299911533e-06, "loss": 0.0084, "step": 49983 }, { "epoch": 15.38, "learning_rate": 2.6776526729170017e-06, "loss": 0.007, "step": 49984 }, { "epoch": 15.38, "learning_rate": 2.6773133340222677e-06, "loss": 0.002, "step": 49985 }, { "epoch": 15.38, "learning_rate": 2.6769740133077847e-06, "loss": 0.0142, "step": 49986 }, { "epoch": 15.38, "learning_rate": 2.676634710774404e-06, "loss": 0.009, "step": 49987 }, { "epoch": 15.38, "learning_rate": 2.6762954264229657e-06, "loss": 0.0065, "step": 49988 }, { "epoch": 15.38, "learning_rate": 2.6759561602543103e-06, "loss": 0.006, "step": 49989 }, { "epoch": 15.38, "learning_rate": 2.675616912269281e-06, "loss": 0.0105, "step": 49990 }, { "epoch": 15.38, "learning_rate": 2.675277682468721e-06, "loss": 0.0094, "step": 49991 }, { "epoch": 15.38, "learning_rate": 2.674938470853472e-06, "loss": 0.0117, "step": 49992 }, { "epoch": 15.38, "learning_rate": 2.674599277424376e-06, "loss": 0.0062, "step": 49993 }, { "epoch": 15.38, "learning_rate": 2.6742601021822746e-06, "loss": 0.0054, "step": 49994 }, { "epoch": 15.38, "learning_rate": 2.6739209451280112e-06, "loss": 0.0091, "step": 49995 }, { "epoch": 15.38, "learning_rate": 2.673581806262423e-06, "loss": 0.0097, "step": 49996 }, { "epoch": 15.38, "learning_rate": 2.67324268558636e-06, "loss": 0.0102, "step": 49997 }, { "epoch": 15.38, "learning_rate": 2.67290358310066e-06, "loss": 0.0054, "step": 49998 }, { "epoch": 15.38, "learning_rate": 2.6725644988061685e-06, "loss": 0.0072, "step": 49999 }, { "epoch": 15.38, "learning_rate": 2.67222543270372e-06, "loss": 0.0108, "step": 50000 }, { "epoch": 15.38, "learning_rate": 2.671886384794157e-06, "loss": 0.0179, "step": 50001 }, { "epoch": 15.39, "learning_rate": 2.6715473550783278e-06, "loss": 0.0148, "step": 50002 }, { "epoch": 15.39, "learning_rate": 2.6712083435570714e-06, "loss": 0.0079, "step": 50003 }, { "epoch": 15.39, "learning_rate": 2.6708693502312287e-06, "loss": 0.0063, "step": 50004 }, { "epoch": 15.39, "learning_rate": 2.670530375101641e-06, "loss": 0.0058, "step": 50005 }, { "epoch": 15.39, "learning_rate": 2.6701914181691515e-06, "loss": 0.0125, "step": 50006 }, { "epoch": 15.39, "learning_rate": 2.669852479434599e-06, "loss": 0.0061, "step": 50007 }, { "epoch": 15.39, "learning_rate": 2.6695135588988285e-06, "loss": 0.0112, "step": 50008 }, { "epoch": 15.39, "learning_rate": 2.6691746565626787e-06, "loss": 0.0122, "step": 50009 }, { "epoch": 15.39, "learning_rate": 2.668835772426992e-06, "loss": 0.0066, "step": 50010 }, { "epoch": 15.39, "learning_rate": 2.668496906492606e-06, "loss": 0.013, "step": 50011 }, { "epoch": 15.39, "learning_rate": 2.668158058760373e-06, "loss": 0.005, "step": 50012 }, { "epoch": 15.39, "learning_rate": 2.667819229231122e-06, "loss": 0.0076, "step": 50013 }, { "epoch": 15.39, "learning_rate": 2.6674804179057013e-06, "loss": 0.0075, "step": 50014 }, { "epoch": 15.39, "learning_rate": 2.667141624784949e-06, "loss": 0.0175, "step": 50015 }, { "epoch": 15.39, "learning_rate": 2.666802849869704e-06, "loss": 0.0086, "step": 50016 }, { "epoch": 15.39, "learning_rate": 2.6664640931608133e-06, "loss": 0.0062, "step": 50017 }, { "epoch": 15.39, "learning_rate": 2.6661253546591158e-06, "loss": 0.0085, "step": 50018 }, { "epoch": 15.39, "learning_rate": 2.6657866343654515e-06, "loss": 0.0073, "step": 50019 }, { "epoch": 15.39, "learning_rate": 2.665447932280665e-06, "loss": 0.0105, "step": 50020 }, { "epoch": 15.39, "learning_rate": 2.6651092484055884e-06, "loss": 0.013, "step": 50021 }, { "epoch": 15.39, "learning_rate": 2.6647705827410697e-06, "loss": 0.0098, "step": 50022 }, { "epoch": 15.39, "learning_rate": 2.66443193528795e-06, "loss": 0.0058, "step": 50023 }, { "epoch": 15.39, "learning_rate": 2.6640933060470663e-06, "loss": 0.0117, "step": 50024 }, { "epoch": 15.39, "learning_rate": 2.6637546950192627e-06, "loss": 0.0077, "step": 50025 }, { "epoch": 15.39, "learning_rate": 2.663416102205377e-06, "loss": 0.0078, "step": 50026 }, { "epoch": 15.39, "learning_rate": 2.6630775276062527e-06, "loss": 0.0096, "step": 50027 }, { "epoch": 15.39, "learning_rate": 2.662738971222728e-06, "loss": 0.0042, "step": 50028 }, { "epoch": 15.39, "learning_rate": 2.6624004330556453e-06, "loss": 0.0142, "step": 50029 }, { "epoch": 15.39, "learning_rate": 2.6620619131058432e-06, "loss": 0.0078, "step": 50030 }, { "epoch": 15.39, "learning_rate": 2.661723411374161e-06, "loss": 0.0128, "step": 50031 }, { "epoch": 15.39, "learning_rate": 2.6613849278614436e-06, "loss": 0.0187, "step": 50032 }, { "epoch": 15.39, "learning_rate": 2.6610464625685317e-06, "loss": 0.0068, "step": 50033 }, { "epoch": 15.39, "learning_rate": 2.6607080154962607e-06, "loss": 0.0068, "step": 50034 }, { "epoch": 15.4, "learning_rate": 2.6603695866454725e-06, "loss": 0.0049, "step": 50035 }, { "epoch": 15.4, "learning_rate": 2.660031176017005e-06, "loss": 0.0062, "step": 50036 }, { "epoch": 15.4, "learning_rate": 2.6596927836117046e-06, "loss": 0.008, "step": 50037 }, { "epoch": 15.4, "learning_rate": 2.659354409430408e-06, "loss": 0.0121, "step": 50038 }, { "epoch": 15.4, "learning_rate": 2.659016053473955e-06, "loss": 0.005, "step": 50039 }, { "epoch": 15.4, "learning_rate": 2.6586777157431865e-06, "loss": 0.0097, "step": 50040 }, { "epoch": 15.4, "learning_rate": 2.6583393962389414e-06, "loss": 0.0045, "step": 50041 }, { "epoch": 15.4, "learning_rate": 2.6580010949620615e-06, "loss": 0.0102, "step": 50042 }, { "epoch": 15.4, "learning_rate": 2.657662811913385e-06, "loss": 0.0108, "step": 50043 }, { "epoch": 15.4, "learning_rate": 2.6573245470937527e-06, "loss": 0.0071, "step": 50044 }, { "epoch": 15.4, "learning_rate": 2.6569863005040032e-06, "loss": 0.0138, "step": 50045 }, { "epoch": 15.4, "learning_rate": 2.6566480721449773e-06, "loss": 0.0074, "step": 50046 }, { "epoch": 15.4, "learning_rate": 2.6563098620175152e-06, "loss": 0.0065, "step": 50047 }, { "epoch": 15.4, "learning_rate": 2.6559716701224547e-06, "loss": 0.0084, "step": 50048 }, { "epoch": 15.4, "learning_rate": 2.6556334964606377e-06, "loss": 0.0207, "step": 50049 }, { "epoch": 15.4, "learning_rate": 2.655295341032903e-06, "loss": 0.0089, "step": 50050 }, { "epoch": 15.4, "learning_rate": 2.6549572038400863e-06, "loss": 0.0128, "step": 50051 }, { "epoch": 15.4, "learning_rate": 2.6546190848830333e-06, "loss": 0.0069, "step": 50052 }, { "epoch": 15.4, "learning_rate": 2.654280984162582e-06, "loss": 0.0061, "step": 50053 }, { "epoch": 15.4, "learning_rate": 2.6539429016795726e-06, "loss": 0.0115, "step": 50054 }, { "epoch": 15.4, "learning_rate": 2.65360483743484e-06, "loss": 0.0053, "step": 50055 }, { "epoch": 15.4, "learning_rate": 2.653266791429222e-06, "loss": 0.0054, "step": 50056 }, { "epoch": 15.4, "learning_rate": 2.652928763663567e-06, "loss": 0.005, "step": 50057 }, { "epoch": 15.4, "learning_rate": 2.6525907541387075e-06, "loss": 0.0097, "step": 50058 }, { "epoch": 15.4, "learning_rate": 2.652252762855485e-06, "loss": 0.0053, "step": 50059 }, { "epoch": 15.4, "learning_rate": 2.651914789814738e-06, "loss": 0.0116, "step": 50060 }, { "epoch": 15.4, "learning_rate": 2.6515768350173054e-06, "loss": 0.0072, "step": 50061 }, { "epoch": 15.4, "learning_rate": 2.651238898464027e-06, "loss": 0.023, "step": 50062 }, { "epoch": 15.4, "learning_rate": 2.6509009801557406e-06, "loss": 0.0083, "step": 50063 }, { "epoch": 15.4, "learning_rate": 2.650563080093286e-06, "loss": 0.0071, "step": 50064 }, { "epoch": 15.4, "learning_rate": 2.6502251982775017e-06, "loss": 0.0101, "step": 50065 }, { "epoch": 15.4, "learning_rate": 2.6498873347092247e-06, "loss": 0.0112, "step": 50066 }, { "epoch": 15.41, "learning_rate": 2.6495494893893e-06, "loss": 0.0067, "step": 50067 }, { "epoch": 15.41, "learning_rate": 2.6492116623185616e-06, "loss": 0.0055, "step": 50068 }, { "epoch": 15.41, "learning_rate": 2.6488738534978475e-06, "loss": 0.0146, "step": 50069 }, { "epoch": 15.41, "learning_rate": 2.648536062927999e-06, "loss": 0.0101, "step": 50070 }, { "epoch": 15.41, "learning_rate": 2.648198290609849e-06, "loss": 0.0067, "step": 50071 }, { "epoch": 15.41, "learning_rate": 2.6478605365442446e-06, "loss": 0.0088, "step": 50072 }, { "epoch": 15.41, "learning_rate": 2.6475228007320196e-06, "loss": 0.0067, "step": 50073 }, { "epoch": 15.41, "learning_rate": 2.647185083174014e-06, "loss": 0.0115, "step": 50074 }, { "epoch": 15.41, "learning_rate": 2.6468473838710695e-06, "loss": 0.0054, "step": 50075 }, { "epoch": 15.41, "learning_rate": 2.646509702824014e-06, "loss": 0.014, "step": 50076 }, { "epoch": 15.41, "learning_rate": 2.6461720400336945e-06, "loss": 0.0068, "step": 50077 }, { "epoch": 15.41, "learning_rate": 2.645834395500948e-06, "loss": 0.0063, "step": 50078 }, { "epoch": 15.41, "learning_rate": 2.645496769226612e-06, "loss": 0.0128, "step": 50079 }, { "epoch": 15.41, "learning_rate": 2.6451591612115245e-06, "loss": 0.0153, "step": 50080 }, { "epoch": 15.41, "learning_rate": 2.644821571456524e-06, "loss": 0.006, "step": 50081 }, { "epoch": 15.41, "learning_rate": 2.6444839999624496e-06, "loss": 0.0084, "step": 50082 }, { "epoch": 15.41, "learning_rate": 2.644146446730138e-06, "loss": 0.0098, "step": 50083 }, { "epoch": 15.41, "learning_rate": 2.643808911760427e-06, "loss": 0.0099, "step": 50084 }, { "epoch": 15.41, "learning_rate": 2.6434713950541558e-06, "loss": 0.0084, "step": 50085 }, { "epoch": 15.41, "learning_rate": 2.643133896612159e-06, "loss": 0.0077, "step": 50086 }, { "epoch": 15.41, "learning_rate": 2.6427964164352805e-06, "loss": 0.019, "step": 50087 }, { "epoch": 15.41, "learning_rate": 2.642458954524357e-06, "loss": 0.0062, "step": 50088 }, { "epoch": 15.41, "learning_rate": 2.642121510880222e-06, "loss": 0.0232, "step": 50089 }, { "epoch": 15.41, "learning_rate": 2.641784085503717e-06, "loss": 0.0137, "step": 50090 }, { "epoch": 15.41, "learning_rate": 2.641446678395674e-06, "loss": 0.0101, "step": 50091 }, { "epoch": 15.41, "learning_rate": 2.641109289556938e-06, "loss": 0.0115, "step": 50092 }, { "epoch": 15.41, "learning_rate": 2.640771918988344e-06, "loss": 0.0074, "step": 50093 }, { "epoch": 15.41, "learning_rate": 2.6404345666907295e-06, "loss": 0.0112, "step": 50094 }, { "epoch": 15.41, "learning_rate": 2.640097232664932e-06, "loss": 0.0063, "step": 50095 }, { "epoch": 15.41, "learning_rate": 2.639759916911788e-06, "loss": 0.0105, "step": 50096 }, { "epoch": 15.41, "learning_rate": 2.6394226194321372e-06, "loss": 0.0203, "step": 50097 }, { "epoch": 15.41, "learning_rate": 2.6390853402268147e-06, "loss": 0.0098, "step": 50098 }, { "epoch": 15.41, "learning_rate": 2.638748079296659e-06, "loss": 0.0092, "step": 50099 }, { "epoch": 15.42, "learning_rate": 2.6384108366425075e-06, "loss": 0.0144, "step": 50100 }, { "epoch": 15.42, "learning_rate": 2.6380736122651973e-06, "loss": 0.005, "step": 50101 }, { "epoch": 15.42, "learning_rate": 2.6377364061655665e-06, "loss": 0.0088, "step": 50102 }, { "epoch": 15.42, "learning_rate": 2.6373992183444496e-06, "loss": 0.0226, "step": 50103 }, { "epoch": 15.42, "learning_rate": 2.6370620488026866e-06, "loss": 0.0104, "step": 50104 }, { "epoch": 15.42, "learning_rate": 2.636724897541114e-06, "loss": 0.0053, "step": 50105 }, { "epoch": 15.42, "learning_rate": 2.636387764560564e-06, "loss": 0.0052, "step": 50106 }, { "epoch": 15.42, "learning_rate": 2.636050649861882e-06, "loss": 0.0077, "step": 50107 }, { "epoch": 15.42, "learning_rate": 2.635713553445901e-06, "loss": 0.0077, "step": 50108 }, { "epoch": 15.42, "learning_rate": 2.6353764753134603e-06, "loss": 0.0066, "step": 50109 }, { "epoch": 15.42, "learning_rate": 2.6350394154653926e-06, "loss": 0.0121, "step": 50110 }, { "epoch": 15.42, "learning_rate": 2.6347023739025322e-06, "loss": 0.0091, "step": 50111 }, { "epoch": 15.42, "learning_rate": 2.634365350625724e-06, "loss": 0.0074, "step": 50112 }, { "epoch": 15.42, "learning_rate": 2.6340283456357994e-06, "loss": 0.0146, "step": 50113 }, { "epoch": 15.42, "learning_rate": 2.633691358933599e-06, "loss": 0.0058, "step": 50114 }, { "epoch": 15.42, "learning_rate": 2.6333543905199554e-06, "loss": 0.0068, "step": 50115 }, { "epoch": 15.42, "learning_rate": 2.6330174403957077e-06, "loss": 0.0014, "step": 50116 }, { "epoch": 15.42, "learning_rate": 2.6326805085616902e-06, "loss": 0.0066, "step": 50117 }, { "epoch": 15.42, "learning_rate": 2.6323435950187428e-06, "loss": 0.0069, "step": 50118 }, { "epoch": 15.42, "learning_rate": 2.6320066997676984e-06, "loss": 0.0077, "step": 50119 }, { "epoch": 15.42, "learning_rate": 2.631669822809396e-06, "loss": 0.009, "step": 50120 }, { "epoch": 15.42, "learning_rate": 2.631332964144667e-06, "loss": 0.0073, "step": 50121 }, { "epoch": 15.42, "learning_rate": 2.6309961237743587e-06, "loss": 0.0065, "step": 50122 }, { "epoch": 15.42, "learning_rate": 2.6306593016992966e-06, "loss": 0.008, "step": 50123 }, { "epoch": 15.42, "learning_rate": 2.6303224979203214e-06, "loss": 0.0136, "step": 50124 }, { "epoch": 15.42, "learning_rate": 2.629985712438269e-06, "loss": 0.0065, "step": 50125 }, { "epoch": 15.42, "learning_rate": 2.6296489452539707e-06, "loss": 0.0069, "step": 50126 }, { "epoch": 15.42, "learning_rate": 2.629312196368271e-06, "loss": 0.0064, "step": 50127 }, { "epoch": 15.42, "learning_rate": 2.628975465782002e-06, "loss": 0.0061, "step": 50128 }, { "epoch": 15.42, "learning_rate": 2.628638753496e-06, "loss": 0.0047, "step": 50129 }, { "epoch": 15.42, "learning_rate": 2.628302059511104e-06, "loss": 0.0051, "step": 50130 }, { "epoch": 15.42, "learning_rate": 2.6279653838281394e-06, "loss": 0.0102, "step": 50131 }, { "epoch": 15.43, "learning_rate": 2.6276287264479537e-06, "loss": 0.0063, "step": 50132 }, { "epoch": 15.43, "learning_rate": 2.627292087371377e-06, "loss": 0.0094, "step": 50133 }, { "epoch": 15.43, "learning_rate": 2.626955466599248e-06, "loss": 0.0167, "step": 50134 }, { "epoch": 15.43, "learning_rate": 2.6266188641324e-06, "loss": 0.0141, "step": 50135 }, { "epoch": 15.43, "learning_rate": 2.626282279971669e-06, "loss": 0.0073, "step": 50136 }, { "epoch": 15.43, "learning_rate": 2.6259457141178923e-06, "loss": 0.006, "step": 50137 }, { "epoch": 15.43, "learning_rate": 2.625609166571903e-06, "loss": 0.0108, "step": 50138 }, { "epoch": 15.43, "learning_rate": 2.6252726373345393e-06, "loss": 0.0027, "step": 50139 }, { "epoch": 15.43, "learning_rate": 2.624936126406635e-06, "loss": 0.0069, "step": 50140 }, { "epoch": 15.43, "learning_rate": 2.6245996337890236e-06, "loss": 0.0118, "step": 50141 }, { "epoch": 15.43, "learning_rate": 2.6242631594825452e-06, "loss": 0.0119, "step": 50142 }, { "epoch": 15.43, "learning_rate": 2.6239267034880354e-06, "loss": 0.0106, "step": 50143 }, { "epoch": 15.43, "learning_rate": 2.623590265806326e-06, "loss": 0.0062, "step": 50144 }, { "epoch": 15.43, "learning_rate": 2.623253846438252e-06, "loss": 0.0069, "step": 50145 }, { "epoch": 15.43, "learning_rate": 2.6229174453846474e-06, "loss": 0.0047, "step": 50146 }, { "epoch": 15.43, "learning_rate": 2.622581062646352e-06, "loss": 0.0063, "step": 50147 }, { "epoch": 15.43, "learning_rate": 2.6222446982242e-06, "loss": 0.0068, "step": 50148 }, { "epoch": 15.43, "learning_rate": 2.621908352119026e-06, "loss": 0.0081, "step": 50149 }, { "epoch": 15.43, "learning_rate": 2.6215720243316636e-06, "loss": 0.0143, "step": 50150 }, { "epoch": 15.43, "learning_rate": 2.6212357148629484e-06, "loss": 0.0055, "step": 50151 }, { "epoch": 15.43, "learning_rate": 2.6208994237137166e-06, "loss": 0.0049, "step": 50152 }, { "epoch": 15.43, "learning_rate": 2.6205631508848008e-06, "loss": 0.0123, "step": 50153 }, { "epoch": 15.43, "learning_rate": 2.620226896377038e-06, "loss": 0.0072, "step": 50154 }, { "epoch": 15.43, "learning_rate": 2.6198906601912623e-06, "loss": 0.0097, "step": 50155 }, { "epoch": 15.43, "learning_rate": 2.6195544423283057e-06, "loss": 0.007, "step": 50156 }, { "epoch": 15.43, "learning_rate": 2.619218242789012e-06, "loss": 0.0045, "step": 50157 }, { "epoch": 15.43, "learning_rate": 2.6188820615742062e-06, "loss": 0.0056, "step": 50158 }, { "epoch": 15.43, "learning_rate": 2.6185458986847256e-06, "loss": 0.0133, "step": 50159 }, { "epoch": 15.43, "learning_rate": 2.6182097541214067e-06, "loss": 0.0048, "step": 50160 }, { "epoch": 15.43, "learning_rate": 2.6178736278850792e-06, "loss": 0.0082, "step": 50161 }, { "epoch": 15.43, "learning_rate": 2.617537519976585e-06, "loss": 0.0128, "step": 50162 }, { "epoch": 15.43, "learning_rate": 2.617201430396754e-06, "loss": 0.0055, "step": 50163 }, { "epoch": 15.43, "learning_rate": 2.6168653591464254e-06, "loss": 0.01, "step": 50164 }, { "epoch": 15.44, "learning_rate": 2.6165293062264264e-06, "loss": 0.0073, "step": 50165 }, { "epoch": 15.44, "learning_rate": 2.6161932716375916e-06, "loss": 0.008, "step": 50166 }, { "epoch": 15.44, "learning_rate": 2.6158572553807615e-06, "loss": 0.0057, "step": 50167 }, { "epoch": 15.44, "learning_rate": 2.615521257456767e-06, "loss": 0.0087, "step": 50168 }, { "epoch": 15.44, "learning_rate": 2.615185277866442e-06, "loss": 0.0054, "step": 50169 }, { "epoch": 15.44, "learning_rate": 2.6148493166106215e-06, "loss": 0.0048, "step": 50170 }, { "epoch": 15.44, "learning_rate": 2.6145133736901386e-06, "loss": 0.0072, "step": 50171 }, { "epoch": 15.44, "learning_rate": 2.614177449105828e-06, "loss": 0.0064, "step": 50172 }, { "epoch": 15.44, "learning_rate": 2.613841542858524e-06, "loss": 0.0136, "step": 50173 }, { "epoch": 15.44, "learning_rate": 2.61350565494906e-06, "loss": 0.0087, "step": 50174 }, { "epoch": 15.44, "learning_rate": 2.613169785378271e-06, "loss": 0.0028, "step": 50175 }, { "epoch": 15.44, "learning_rate": 2.6128339341469855e-06, "loss": 0.0118, "step": 50176 }, { "epoch": 15.44, "learning_rate": 2.612498101256046e-06, "loss": 0.0052, "step": 50177 }, { "epoch": 15.44, "learning_rate": 2.6121622867062844e-06, "loss": 0.0115, "step": 50178 }, { "epoch": 15.44, "learning_rate": 2.61182649049853e-06, "loss": 0.0081, "step": 50179 }, { "epoch": 15.44, "learning_rate": 2.6114907126336176e-06, "loss": 0.0077, "step": 50180 }, { "epoch": 15.44, "learning_rate": 2.6111549531123793e-06, "loss": 0.0157, "step": 50181 }, { "epoch": 15.44, "learning_rate": 2.6108192119356546e-06, "loss": 0.0077, "step": 50182 }, { "epoch": 15.44, "learning_rate": 2.610483489104274e-06, "loss": 0.0033, "step": 50183 }, { "epoch": 15.44, "learning_rate": 2.610147784619069e-06, "loss": 0.0065, "step": 50184 }, { "epoch": 15.44, "learning_rate": 2.6098120984808793e-06, "loss": 0.0048, "step": 50185 }, { "epoch": 15.44, "learning_rate": 2.6094764306905284e-06, "loss": 0.0105, "step": 50186 }, { "epoch": 15.44, "learning_rate": 2.6091407812488567e-06, "loss": 0.01, "step": 50187 }, { "epoch": 15.44, "learning_rate": 2.608805150156697e-06, "loss": 0.0056, "step": 50188 }, { "epoch": 15.44, "learning_rate": 2.6084695374148806e-06, "loss": 0.014, "step": 50189 }, { "epoch": 15.44, "learning_rate": 2.6081339430242426e-06, "loss": 0.0139, "step": 50190 }, { "epoch": 15.44, "learning_rate": 2.6077983669856144e-06, "loss": 0.0052, "step": 50191 }, { "epoch": 15.44, "learning_rate": 2.60746280929983e-06, "loss": 0.008, "step": 50192 }, { "epoch": 15.44, "learning_rate": 2.607127269967723e-06, "loss": 0.0076, "step": 50193 }, { "epoch": 15.44, "learning_rate": 2.606791748990125e-06, "loss": 0.0113, "step": 50194 }, { "epoch": 15.44, "learning_rate": 2.606456246367871e-06, "loss": 0.0075, "step": 50195 }, { "epoch": 15.44, "learning_rate": 2.6061207621017893e-06, "loss": 0.005, "step": 50196 }, { "epoch": 15.45, "learning_rate": 2.60578529619272e-06, "loss": 0.0103, "step": 50197 }, { "epoch": 15.45, "learning_rate": 2.605449848641495e-06, "loss": 0.0071, "step": 50198 }, { "epoch": 15.45, "learning_rate": 2.6051144194489408e-06, "loss": 0.0141, "step": 50199 }, { "epoch": 15.45, "learning_rate": 2.604779008615895e-06, "loss": 0.0058, "step": 50200 }, { "epoch": 15.45, "learning_rate": 2.6044436161431865e-06, "loss": 0.0097, "step": 50201 }, { "epoch": 15.45, "learning_rate": 2.604108242031652e-06, "loss": 0.0115, "step": 50202 }, { "epoch": 15.45, "learning_rate": 2.6037728862821243e-06, "loss": 0.0067, "step": 50203 }, { "epoch": 15.45, "learning_rate": 2.6034375488954343e-06, "loss": 0.0076, "step": 50204 }, { "epoch": 15.45, "learning_rate": 2.603102229872415e-06, "loss": 0.0105, "step": 50205 }, { "epoch": 15.45, "learning_rate": 2.602766929213898e-06, "loss": 0.0075, "step": 50206 }, { "epoch": 15.45, "learning_rate": 2.602431646920717e-06, "loss": 0.0088, "step": 50207 }, { "epoch": 15.45, "learning_rate": 2.602096382993704e-06, "loss": 0.0083, "step": 50208 }, { "epoch": 15.45, "learning_rate": 2.6017611374336916e-06, "loss": 0.0082, "step": 50209 }, { "epoch": 15.45, "learning_rate": 2.6014259102415106e-06, "loss": 0.0072, "step": 50210 }, { "epoch": 15.45, "learning_rate": 2.601090701417992e-06, "loss": 0.0128, "step": 50211 }, { "epoch": 15.45, "learning_rate": 2.600755510963977e-06, "loss": 0.0114, "step": 50212 }, { "epoch": 15.45, "learning_rate": 2.6004203388802883e-06, "loss": 0.0096, "step": 50213 }, { "epoch": 15.45, "learning_rate": 2.6000851851677598e-06, "loss": 0.0038, "step": 50214 }, { "epoch": 15.45, "learning_rate": 2.5997500498272264e-06, "loss": 0.0079, "step": 50215 }, { "epoch": 15.45, "learning_rate": 2.599414932859514e-06, "loss": 0.009, "step": 50216 }, { "epoch": 15.45, "learning_rate": 2.5990798342654632e-06, "loss": 0.0081, "step": 50217 }, { "epoch": 15.45, "learning_rate": 2.598744754045902e-06, "loss": 0.0062, "step": 50218 }, { "epoch": 15.45, "learning_rate": 2.5984096922016654e-06, "loss": 0.0061, "step": 50219 }, { "epoch": 15.45, "learning_rate": 2.5980746487335785e-06, "loss": 0.0058, "step": 50220 }, { "epoch": 15.45, "learning_rate": 2.5977396236424745e-06, "loss": 0.0069, "step": 50221 }, { "epoch": 15.45, "learning_rate": 2.59740461692919e-06, "loss": 0.0061, "step": 50222 }, { "epoch": 15.45, "learning_rate": 2.597069628594554e-06, "loss": 0.0063, "step": 50223 }, { "epoch": 15.45, "learning_rate": 2.5967346586393983e-06, "loss": 0.0112, "step": 50224 }, { "epoch": 15.45, "learning_rate": 2.596399707064555e-06, "loss": 0.0029, "step": 50225 }, { "epoch": 15.45, "learning_rate": 2.5960647738708553e-06, "loss": 0.0135, "step": 50226 }, { "epoch": 15.45, "learning_rate": 2.5957298590591305e-06, "loss": 0.0046, "step": 50227 }, { "epoch": 15.45, "learning_rate": 2.5953949626302123e-06, "loss": 0.009, "step": 50228 }, { "epoch": 15.45, "learning_rate": 2.5950600845849317e-06, "loss": 0.0076, "step": 50229 }, { "epoch": 15.46, "learning_rate": 2.5947252249241218e-06, "loss": 0.0081, "step": 50230 }, { "epoch": 15.46, "learning_rate": 2.5943903836486083e-06, "loss": 0.0077, "step": 50231 }, { "epoch": 15.46, "learning_rate": 2.5940555607592312e-06, "loss": 0.0085, "step": 50232 }, { "epoch": 15.46, "learning_rate": 2.59372075625682e-06, "loss": 0.0048, "step": 50233 }, { "epoch": 15.46, "learning_rate": 2.5933859701422005e-06, "loss": 0.0058, "step": 50234 }, { "epoch": 15.46, "learning_rate": 2.593051202416207e-06, "loss": 0.011, "step": 50235 }, { "epoch": 15.46, "learning_rate": 2.5927164530796677e-06, "loss": 0.004, "step": 50236 }, { "epoch": 15.46, "learning_rate": 2.592381722133419e-06, "loss": 0.0072, "step": 50237 }, { "epoch": 15.46, "learning_rate": 2.5920470095782902e-06, "loss": 0.009, "step": 50238 }, { "epoch": 15.46, "learning_rate": 2.59171231541511e-06, "loss": 0.0068, "step": 50239 }, { "epoch": 15.46, "learning_rate": 2.5913776396447155e-06, "loss": 0.0067, "step": 50240 }, { "epoch": 15.46, "learning_rate": 2.591042982267926e-06, "loss": 0.0072, "step": 50241 }, { "epoch": 15.46, "learning_rate": 2.590708343285584e-06, "loss": 0.0081, "step": 50242 }, { "epoch": 15.46, "learning_rate": 2.5903737226985137e-06, "loss": 0.0081, "step": 50243 }, { "epoch": 15.46, "learning_rate": 2.590039120507549e-06, "loss": 0.006, "step": 50244 }, { "epoch": 15.46, "learning_rate": 2.5897045367135163e-06, "loss": 0.0113, "step": 50245 }, { "epoch": 15.46, "learning_rate": 2.589369971317256e-06, "loss": 0.0113, "step": 50246 }, { "epoch": 15.46, "learning_rate": 2.589035424319588e-06, "loss": 0.0066, "step": 50247 }, { "epoch": 15.46, "learning_rate": 2.588700895721348e-06, "loss": 0.0392, "step": 50248 }, { "epoch": 15.46, "learning_rate": 2.5883663855233656e-06, "loss": 0.0042, "step": 50249 }, { "epoch": 15.46, "learning_rate": 2.588031893726468e-06, "loss": 0.0138, "step": 50250 }, { "epoch": 15.46, "learning_rate": 2.587697420331492e-06, "loss": 0.0088, "step": 50251 }, { "epoch": 15.46, "learning_rate": 2.5873629653392653e-06, "loss": 0.0138, "step": 50252 }, { "epoch": 15.46, "learning_rate": 2.587028528750618e-06, "loss": 0.0064, "step": 50253 }, { "epoch": 15.46, "learning_rate": 2.5866941105663836e-06, "loss": 0.0175, "step": 50254 }, { "epoch": 15.46, "learning_rate": 2.586359710787383e-06, "loss": 0.0037, "step": 50255 }, { "epoch": 15.46, "learning_rate": 2.5860253294144555e-06, "loss": 0.0151, "step": 50256 }, { "epoch": 15.46, "learning_rate": 2.5856909664484274e-06, "loss": 0.0147, "step": 50257 }, { "epoch": 15.46, "learning_rate": 2.5853566218901303e-06, "loss": 0.0053, "step": 50258 }, { "epoch": 15.46, "learning_rate": 2.585022295740395e-06, "loss": 0.0113, "step": 50259 }, { "epoch": 15.46, "learning_rate": 2.5846879880000496e-06, "loss": 0.0146, "step": 50260 }, { "epoch": 15.46, "learning_rate": 2.584353698669925e-06, "loss": 0.0077, "step": 50261 }, { "epoch": 15.47, "learning_rate": 2.5840194277508503e-06, "loss": 0.0046, "step": 50262 }, { "epoch": 15.47, "learning_rate": 2.5836851752436563e-06, "loss": 0.0079, "step": 50263 }, { "epoch": 15.47, "learning_rate": 2.583350941149173e-06, "loss": 0.016, "step": 50264 }, { "epoch": 15.47, "learning_rate": 2.583016725468226e-06, "loss": 0.0067, "step": 50265 }, { "epoch": 15.47, "learning_rate": 2.5826825282016523e-06, "loss": 0.0064, "step": 50266 }, { "epoch": 15.47, "learning_rate": 2.582348349350282e-06, "loss": 0.0133, "step": 50267 }, { "epoch": 15.47, "learning_rate": 2.5820141889149377e-06, "loss": 0.0063, "step": 50268 }, { "epoch": 15.47, "learning_rate": 2.581680046896452e-06, "loss": 0.0114, "step": 50269 }, { "epoch": 15.47, "learning_rate": 2.5813459232956518e-06, "loss": 0.0124, "step": 50270 }, { "epoch": 15.47, "learning_rate": 2.5810118181133726e-06, "loss": 0.0093, "step": 50271 }, { "epoch": 15.47, "learning_rate": 2.5806777313504416e-06, "loss": 0.0055, "step": 50272 }, { "epoch": 15.47, "learning_rate": 2.580343663007686e-06, "loss": 0.0078, "step": 50273 }, { "epoch": 15.47, "learning_rate": 2.5800096130859386e-06, "loss": 0.0118, "step": 50274 }, { "epoch": 15.47, "learning_rate": 2.5796755815860264e-06, "loss": 0.004, "step": 50275 }, { "epoch": 15.47, "learning_rate": 2.5793415685087797e-06, "loss": 0.0111, "step": 50276 }, { "epoch": 15.47, "learning_rate": 2.5790075738550256e-06, "loss": 0.0051, "step": 50277 }, { "epoch": 15.47, "learning_rate": 2.578673597625597e-06, "loss": 0.0142, "step": 50278 }, { "epoch": 15.47, "learning_rate": 2.5783396398213203e-06, "loss": 0.0104, "step": 50279 }, { "epoch": 15.47, "learning_rate": 2.578005700443026e-06, "loss": 0.0075, "step": 50280 }, { "epoch": 15.47, "learning_rate": 2.577671779491542e-06, "loss": 0.0084, "step": 50281 }, { "epoch": 15.47, "learning_rate": 2.5773378769676983e-06, "loss": 0.0091, "step": 50282 }, { "epoch": 15.47, "learning_rate": 2.5770039928723234e-06, "loss": 0.0046, "step": 50283 }, { "epoch": 15.47, "learning_rate": 2.5766701272062466e-06, "loss": 0.016, "step": 50284 }, { "epoch": 15.47, "learning_rate": 2.576336279970293e-06, "loss": 0.0041, "step": 50285 }, { "epoch": 15.47, "learning_rate": 2.5760024511652994e-06, "loss": 0.0094, "step": 50286 }, { "epoch": 15.47, "learning_rate": 2.5756686407920884e-06, "loss": 0.008, "step": 50287 }, { "epoch": 15.47, "learning_rate": 2.575334848851495e-06, "loss": 0.0083, "step": 50288 }, { "epoch": 15.47, "learning_rate": 2.57500107534434e-06, "loss": 0.0074, "step": 50289 }, { "epoch": 15.47, "learning_rate": 2.5746673202714534e-06, "loss": 0.0116, "step": 50290 }, { "epoch": 15.47, "learning_rate": 2.574333583633668e-06, "loss": 0.0132, "step": 50291 }, { "epoch": 15.47, "learning_rate": 2.573999865431811e-06, "loss": 0.0081, "step": 50292 }, { "epoch": 15.47, "learning_rate": 2.5736661656667094e-06, "loss": 0.0137, "step": 50293 }, { "epoch": 15.47, "learning_rate": 2.573332484339194e-06, "loss": 0.0087, "step": 50294 }, { "epoch": 15.48, "learning_rate": 2.5729988214500913e-06, "loss": 0.0071, "step": 50295 }, { "epoch": 15.48, "learning_rate": 2.5726651770002297e-06, "loss": 0.0085, "step": 50296 }, { "epoch": 15.48, "learning_rate": 2.572331550990439e-06, "loss": 0.0095, "step": 50297 }, { "epoch": 15.48, "learning_rate": 2.571997943421546e-06, "loss": 0.0086, "step": 50298 }, { "epoch": 15.48, "learning_rate": 2.57166435429438e-06, "loss": 0.0074, "step": 50299 }, { "epoch": 15.48, "learning_rate": 2.5713307836097657e-06, "loss": 0.0074, "step": 50300 }, { "epoch": 15.48, "learning_rate": 2.5709972313685405e-06, "loss": 0.0083, "step": 50301 }, { "epoch": 15.48, "learning_rate": 2.5706636975715225e-06, "loss": 0.0052, "step": 50302 }, { "epoch": 15.48, "learning_rate": 2.5703301822195448e-06, "loss": 0.0047, "step": 50303 }, { "epoch": 15.48, "learning_rate": 2.569996685313434e-06, "loss": 0.0051, "step": 50304 }, { "epoch": 15.48, "learning_rate": 2.569663206854015e-06, "loss": 0.0167, "step": 50305 }, { "epoch": 15.48, "learning_rate": 2.569329746842123e-06, "loss": 0.0152, "step": 50306 }, { "epoch": 15.48, "learning_rate": 2.5689963052785806e-06, "loss": 0.0108, "step": 50307 }, { "epoch": 15.48, "learning_rate": 2.568662882164218e-06, "loss": 0.0111, "step": 50308 }, { "epoch": 15.48, "learning_rate": 2.5683294774998648e-06, "loss": 0.0085, "step": 50309 }, { "epoch": 15.48, "learning_rate": 2.56799609128634e-06, "loss": 0.0091, "step": 50310 }, { "epoch": 15.48, "learning_rate": 2.56766272352448e-06, "loss": 0.0104, "step": 50311 }, { "epoch": 15.48, "learning_rate": 2.5673293742151095e-06, "loss": 0.014, "step": 50312 }, { "epoch": 15.48, "learning_rate": 2.5669960433590567e-06, "loss": 0.0206, "step": 50313 }, { "epoch": 15.48, "learning_rate": 2.5666627309571502e-06, "loss": 0.0128, "step": 50314 }, { "epoch": 15.48, "learning_rate": 2.5663294370102142e-06, "loss": 0.0052, "step": 50315 }, { "epoch": 15.48, "learning_rate": 2.5659961615190788e-06, "loss": 0.0076, "step": 50316 }, { "epoch": 15.48, "learning_rate": 2.5656629044845714e-06, "loss": 0.0076, "step": 50317 }, { "epoch": 15.48, "learning_rate": 2.5653296659075176e-06, "loss": 0.0055, "step": 50318 }, { "epoch": 15.48, "learning_rate": 2.564996445788748e-06, "loss": 0.0058, "step": 50319 }, { "epoch": 15.48, "learning_rate": 2.564663244129083e-06, "loss": 0.0069, "step": 50320 }, { "epoch": 15.48, "learning_rate": 2.5643300609293577e-06, "loss": 0.011, "step": 50321 }, { "epoch": 15.48, "learning_rate": 2.5639968961903996e-06, "loss": 0.0035, "step": 50322 }, { "epoch": 15.48, "learning_rate": 2.56366374991303e-06, "loss": 0.0075, "step": 50323 }, { "epoch": 15.48, "learning_rate": 2.563330622098078e-06, "loss": 0.0063, "step": 50324 }, { "epoch": 15.48, "learning_rate": 2.562997512746368e-06, "loss": 0.0065, "step": 50325 }, { "epoch": 15.48, "learning_rate": 2.562664421858734e-06, "loss": 0.0039, "step": 50326 }, { "epoch": 15.49, "learning_rate": 2.562331349435998e-06, "loss": 0.0079, "step": 50327 }, { "epoch": 15.49, "learning_rate": 2.561998295478989e-06, "loss": 0.0054, "step": 50328 }, { "epoch": 15.49, "learning_rate": 2.5616652599885326e-06, "loss": 0.0081, "step": 50329 }, { "epoch": 15.49, "learning_rate": 2.5613322429654573e-06, "loss": 0.0107, "step": 50330 }, { "epoch": 15.49, "learning_rate": 2.560999244410587e-06, "loss": 0.0052, "step": 50331 }, { "epoch": 15.49, "learning_rate": 2.5606662643247515e-06, "loss": 0.0085, "step": 50332 }, { "epoch": 15.49, "learning_rate": 2.5603333027087753e-06, "loss": 0.003, "step": 50333 }, { "epoch": 15.49, "learning_rate": 2.560000359563486e-06, "loss": 0.0049, "step": 50334 }, { "epoch": 15.49, "learning_rate": 2.5596674348897086e-06, "loss": 0.0113, "step": 50335 }, { "epoch": 15.49, "learning_rate": 2.5593345286882732e-06, "loss": 0.0088, "step": 50336 }, { "epoch": 15.49, "learning_rate": 2.559001640960004e-06, "loss": 0.0065, "step": 50337 }, { "epoch": 15.49, "learning_rate": 2.5586687717057267e-06, "loss": 0.0095, "step": 50338 }, { "epoch": 15.49, "learning_rate": 2.558335920926269e-06, "loss": 0.0143, "step": 50339 }, { "epoch": 15.49, "learning_rate": 2.5580030886224537e-06, "loss": 0.0077, "step": 50340 }, { "epoch": 15.49, "learning_rate": 2.557670274795114e-06, "loss": 0.0099, "step": 50341 }, { "epoch": 15.49, "learning_rate": 2.5573374794450724e-06, "loss": 0.008, "step": 50342 }, { "epoch": 15.49, "learning_rate": 2.5570047025731594e-06, "loss": 0.0099, "step": 50343 }, { "epoch": 15.49, "learning_rate": 2.556671944180192e-06, "loss": 0.0107, "step": 50344 }, { "epoch": 15.49, "learning_rate": 2.556339204267e-06, "loss": 0.0048, "step": 50345 }, { "epoch": 15.49, "learning_rate": 2.5560064828344145e-06, "loss": 0.0047, "step": 50346 }, { "epoch": 15.49, "learning_rate": 2.5556737798832565e-06, "loss": 0.0085, "step": 50347 }, { "epoch": 15.49, "learning_rate": 2.555341095414354e-06, "loss": 0.0075, "step": 50348 }, { "epoch": 15.49, "learning_rate": 2.555008429428534e-06, "loss": 0.0098, "step": 50349 }, { "epoch": 15.49, "learning_rate": 2.5546757819266197e-06, "loss": 0.0057, "step": 50350 }, { "epoch": 15.49, "learning_rate": 2.5543431529094386e-06, "loss": 0.017, "step": 50351 }, { "epoch": 15.49, "learning_rate": 2.554010542377815e-06, "loss": 0.0062, "step": 50352 }, { "epoch": 15.49, "learning_rate": 2.553677950332577e-06, "loss": 0.0089, "step": 50353 }, { "epoch": 15.49, "learning_rate": 2.55334537677455e-06, "loss": 0.0124, "step": 50354 }, { "epoch": 15.49, "learning_rate": 2.5530128217045546e-06, "loss": 0.0059, "step": 50355 }, { "epoch": 15.49, "learning_rate": 2.5526802851234268e-06, "loss": 0.008, "step": 50356 }, { "epoch": 15.49, "learning_rate": 2.552347767031983e-06, "loss": 0.0083, "step": 50357 }, { "epoch": 15.49, "learning_rate": 2.552015267431053e-06, "loss": 0.007, "step": 50358 }, { "epoch": 15.49, "learning_rate": 2.55168278632146e-06, "loss": 0.008, "step": 50359 }, { "epoch": 15.5, "learning_rate": 2.5513503237040273e-06, "loss": 0.0075, "step": 50360 }, { "epoch": 15.5, "learning_rate": 2.551017879579588e-06, "loss": 0.0068, "step": 50361 }, { "epoch": 15.5, "learning_rate": 2.5506854539489624e-06, "loss": 0.0094, "step": 50362 }, { "epoch": 15.5, "learning_rate": 2.5503530468129766e-06, "loss": 0.007, "step": 50363 }, { "epoch": 15.5, "learning_rate": 2.5500206581724584e-06, "loss": 0.0099, "step": 50364 }, { "epoch": 15.5, "learning_rate": 2.5496882880282246e-06, "loss": 0.0058, "step": 50365 }, { "epoch": 15.5, "learning_rate": 2.5493559363811093e-06, "loss": 0.007, "step": 50366 }, { "epoch": 15.5, "learning_rate": 2.549023603231935e-06, "loss": 0.0115, "step": 50367 }, { "epoch": 15.5, "learning_rate": 2.5486912885815262e-06, "loss": 0.0081, "step": 50368 }, { "epoch": 15.5, "learning_rate": 2.5483589924307086e-06, "loss": 0.016, "step": 50369 }, { "epoch": 15.5, "learning_rate": 2.5480267147803064e-06, "loss": 0.0083, "step": 50370 }, { "epoch": 15.5, "learning_rate": 2.5476944556311447e-06, "loss": 0.0059, "step": 50371 }, { "epoch": 15.5, "learning_rate": 2.547362214984048e-06, "loss": 0.0073, "step": 50372 }, { "epoch": 15.5, "learning_rate": 2.5470299928398424e-06, "loss": 0.0053, "step": 50373 }, { "epoch": 15.5, "learning_rate": 2.5466977891993518e-06, "loss": 0.012, "step": 50374 }, { "epoch": 15.5, "learning_rate": 2.5463656040633988e-06, "loss": 0.0101, "step": 50375 }, { "epoch": 15.5, "learning_rate": 2.5460334374328134e-06, "loss": 0.0072, "step": 50376 }, { "epoch": 15.5, "learning_rate": 2.54570128930842e-06, "loss": 0.0148, "step": 50377 }, { "epoch": 15.5, "learning_rate": 2.545369159691038e-06, "loss": 0.0064, "step": 50378 }, { "epoch": 15.5, "learning_rate": 2.5450370485814944e-06, "loss": 0.0057, "step": 50379 }, { "epoch": 15.5, "learning_rate": 2.5447049559806116e-06, "loss": 0.0104, "step": 50380 }, { "epoch": 15.5, "learning_rate": 2.544372881889219e-06, "loss": 0.0056, "step": 50381 }, { "epoch": 15.5, "learning_rate": 2.5440408263081385e-06, "loss": 0.0086, "step": 50382 }, { "epoch": 15.5, "learning_rate": 2.543708789238195e-06, "loss": 0.006, "step": 50383 }, { "epoch": 15.5, "learning_rate": 2.5433767706802114e-06, "loss": 0.01, "step": 50384 }, { "epoch": 15.5, "learning_rate": 2.5430447706350147e-06, "loss": 0.0108, "step": 50385 }, { "epoch": 15.5, "learning_rate": 2.5427127891034265e-06, "loss": 0.0041, "step": 50386 }, { "epoch": 15.5, "learning_rate": 2.5423808260862713e-06, "loss": 0.0078, "step": 50387 }, { "epoch": 15.5, "learning_rate": 2.542048881584376e-06, "loss": 0.0044, "step": 50388 }, { "epoch": 15.5, "learning_rate": 2.541716955598561e-06, "loss": 0.0036, "step": 50389 }, { "epoch": 15.5, "learning_rate": 2.5413850481296497e-06, "loss": 0.0143, "step": 50390 }, { "epoch": 15.5, "learning_rate": 2.5410531591784746e-06, "loss": 0.0208, "step": 50391 }, { "epoch": 15.51, "learning_rate": 2.54072128874585e-06, "loss": 0.009, "step": 50392 }, { "epoch": 15.51, "learning_rate": 2.5403894368326042e-06, "loss": 0.0206, "step": 50393 }, { "epoch": 15.51, "learning_rate": 2.5400576034395606e-06, "loss": 0.0077, "step": 50394 }, { "epoch": 15.51, "learning_rate": 2.5397257885675396e-06, "loss": 0.0071, "step": 50395 }, { "epoch": 15.51, "learning_rate": 2.5393939922173707e-06, "loss": 0.0077, "step": 50396 }, { "epoch": 15.51, "learning_rate": 2.5390622143898756e-06, "loss": 0.0148, "step": 50397 }, { "epoch": 15.51, "learning_rate": 2.5387304550858806e-06, "loss": 0.0108, "step": 50398 }, { "epoch": 15.51, "learning_rate": 2.5383987143062027e-06, "loss": 0.006, "step": 50399 }, { "epoch": 15.51, "learning_rate": 2.5380669920516677e-06, "loss": 0.0106, "step": 50400 }, { "epoch": 15.51, "learning_rate": 2.537735288323102e-06, "loss": 0.0068, "step": 50401 }, { "epoch": 15.51, "learning_rate": 2.537403603121329e-06, "loss": 0.0084, "step": 50402 }, { "epoch": 15.51, "learning_rate": 2.537071936447171e-06, "loss": 0.0075, "step": 50403 }, { "epoch": 15.51, "learning_rate": 2.5367402883014512e-06, "loss": 0.0383, "step": 50404 }, { "epoch": 15.51, "learning_rate": 2.5364086586849933e-06, "loss": 0.0102, "step": 50405 }, { "epoch": 15.51, "learning_rate": 2.53607704759862e-06, "loss": 0.0115, "step": 50406 }, { "epoch": 15.51, "learning_rate": 2.535745455043156e-06, "loss": 0.0107, "step": 50407 }, { "epoch": 15.51, "learning_rate": 2.535413881019423e-06, "loss": 0.0105, "step": 50408 }, { "epoch": 15.51, "learning_rate": 2.535082325528245e-06, "loss": 0.0168, "step": 50409 }, { "epoch": 15.51, "learning_rate": 2.5347507885704424e-06, "loss": 0.0068, "step": 50410 }, { "epoch": 15.51, "learning_rate": 2.534419270146844e-06, "loss": 0.0108, "step": 50411 }, { "epoch": 15.51, "learning_rate": 2.5340877702582734e-06, "loss": 0.0131, "step": 50412 }, { "epoch": 15.51, "learning_rate": 2.533756288905547e-06, "loss": 0.0113, "step": 50413 }, { "epoch": 15.51, "learning_rate": 2.53342482608949e-06, "loss": 0.0082, "step": 50414 }, { "epoch": 15.51, "learning_rate": 2.533093381810924e-06, "loss": 0.0098, "step": 50415 }, { "epoch": 15.51, "learning_rate": 2.532761956070676e-06, "loss": 0.0087, "step": 50416 }, { "epoch": 15.51, "learning_rate": 2.532430548869569e-06, "loss": 0.0104, "step": 50417 }, { "epoch": 15.51, "learning_rate": 2.532099160208422e-06, "loss": 0.0107, "step": 50418 }, { "epoch": 15.51, "learning_rate": 2.5317677900880634e-06, "loss": 0.0129, "step": 50419 }, { "epoch": 15.51, "learning_rate": 2.5314364385093047e-06, "loss": 0.0097, "step": 50420 }, { "epoch": 15.51, "learning_rate": 2.5311051054729806e-06, "loss": 0.0064, "step": 50421 }, { "epoch": 15.51, "learning_rate": 2.530773790979908e-06, "loss": 0.0073, "step": 50422 }, { "epoch": 15.51, "learning_rate": 2.5304424950309104e-06, "loss": 0.0104, "step": 50423 }, { "epoch": 15.51, "learning_rate": 2.5301112176268104e-06, "loss": 0.007, "step": 50424 }, { "epoch": 15.52, "learning_rate": 2.529779958768431e-06, "loss": 0.0087, "step": 50425 }, { "epoch": 15.52, "learning_rate": 2.529448718456594e-06, "loss": 0.0043, "step": 50426 }, { "epoch": 15.52, "learning_rate": 2.5291174966921216e-06, "loss": 0.0069, "step": 50427 }, { "epoch": 15.52, "learning_rate": 2.5287862934758356e-06, "loss": 0.0075, "step": 50428 }, { "epoch": 15.52, "learning_rate": 2.5284551088085597e-06, "loss": 0.0085, "step": 50429 }, { "epoch": 15.52, "learning_rate": 2.528123942691113e-06, "loss": 0.0072, "step": 50430 }, { "epoch": 15.52, "learning_rate": 2.5277927951243232e-06, "loss": 0.0107, "step": 50431 }, { "epoch": 15.52, "learning_rate": 2.527461666109009e-06, "loss": 0.0126, "step": 50432 }, { "epoch": 15.52, "learning_rate": 2.527130555645996e-06, "loss": 0.0135, "step": 50433 }, { "epoch": 15.52, "learning_rate": 2.526799463736099e-06, "loss": 0.0062, "step": 50434 }, { "epoch": 15.52, "learning_rate": 2.5264683903801422e-06, "loss": 0.0175, "step": 50435 }, { "epoch": 15.52, "learning_rate": 2.5261373355789532e-06, "loss": 0.0067, "step": 50436 }, { "epoch": 15.52, "learning_rate": 2.52580629933335e-06, "loss": 0.0108, "step": 50437 }, { "epoch": 15.52, "learning_rate": 2.5254752816441552e-06, "loss": 0.0042, "step": 50438 }, { "epoch": 15.52, "learning_rate": 2.5251442825121886e-06, "loss": 0.0052, "step": 50439 }, { "epoch": 15.52, "learning_rate": 2.5248133019382747e-06, "loss": 0.0082, "step": 50440 }, { "epoch": 15.52, "learning_rate": 2.5244823399232343e-06, "loss": 0.0228, "step": 50441 }, { "epoch": 15.52, "learning_rate": 2.524151396467889e-06, "loss": 0.0037, "step": 50442 }, { "epoch": 15.52, "learning_rate": 2.52382047157306e-06, "loss": 0.0061, "step": 50443 }, { "epoch": 15.52, "learning_rate": 2.5234895652395687e-06, "loss": 0.009, "step": 50444 }, { "epoch": 15.52, "learning_rate": 2.523158677468235e-06, "loss": 0.0077, "step": 50445 }, { "epoch": 15.52, "learning_rate": 2.522827808259889e-06, "loss": 0.0085, "step": 50446 }, { "epoch": 15.52, "learning_rate": 2.5224969576153413e-06, "loss": 0.0032, "step": 50447 }, { "epoch": 15.52, "learning_rate": 2.522166125535419e-06, "loss": 0.0061, "step": 50448 }, { "epoch": 15.52, "learning_rate": 2.5218353120209415e-06, "loss": 0.0042, "step": 50449 }, { "epoch": 15.52, "learning_rate": 2.521504517072728e-06, "loss": 0.0086, "step": 50450 }, { "epoch": 15.52, "learning_rate": 2.521173740691607e-06, "loss": 0.0082, "step": 50451 }, { "epoch": 15.52, "learning_rate": 2.5208429828783943e-06, "loss": 0.0133, "step": 50452 }, { "epoch": 15.52, "learning_rate": 2.5205122436339113e-06, "loss": 0.0044, "step": 50453 }, { "epoch": 15.52, "learning_rate": 2.520181522958984e-06, "loss": 0.0125, "step": 50454 }, { "epoch": 15.52, "learning_rate": 2.519850820854424e-06, "loss": 0.0095, "step": 50455 }, { "epoch": 15.52, "learning_rate": 2.5195201373210597e-06, "loss": 0.0099, "step": 50456 }, { "epoch": 15.53, "learning_rate": 2.5191894723597098e-06, "loss": 0.0035, "step": 50457 }, { "epoch": 15.53, "learning_rate": 2.5188588259711967e-06, "loss": 0.0057, "step": 50458 }, { "epoch": 15.53, "learning_rate": 2.5185281981563404e-06, "loss": 0.0067, "step": 50459 }, { "epoch": 15.53, "learning_rate": 2.5181975889159615e-06, "loss": 0.0137, "step": 50460 }, { "epoch": 15.53, "learning_rate": 2.5178669982508806e-06, "loss": 0.0067, "step": 50461 }, { "epoch": 15.53, "learning_rate": 2.5175364261619195e-06, "loss": 0.0195, "step": 50462 }, { "epoch": 15.53, "learning_rate": 2.517205872649897e-06, "loss": 0.0111, "step": 50463 }, { "epoch": 15.53, "learning_rate": 2.5168753377156363e-06, "loss": 0.0067, "step": 50464 }, { "epoch": 15.53, "learning_rate": 2.5165448213599532e-06, "loss": 0.0088, "step": 50465 }, { "epoch": 15.53, "learning_rate": 2.5162143235836744e-06, "loss": 0.007, "step": 50466 }, { "epoch": 15.53, "learning_rate": 2.515883844387622e-06, "loss": 0.0086, "step": 50467 }, { "epoch": 15.53, "learning_rate": 2.515553383772609e-06, "loss": 0.0095, "step": 50468 }, { "epoch": 15.53, "learning_rate": 2.515222941739458e-06, "loss": 0.0086, "step": 50469 }, { "epoch": 15.53, "learning_rate": 2.514892518288988e-06, "loss": 0.0088, "step": 50470 }, { "epoch": 15.53, "learning_rate": 2.5145621134220267e-06, "loss": 0.0081, "step": 50471 }, { "epoch": 15.53, "learning_rate": 2.5142317271393877e-06, "loss": 0.0153, "step": 50472 }, { "epoch": 15.53, "learning_rate": 2.513901359441894e-06, "loss": 0.0066, "step": 50473 }, { "epoch": 15.53, "learning_rate": 2.5135710103303645e-06, "loss": 0.0099, "step": 50474 }, { "epoch": 15.53, "learning_rate": 2.513240679805621e-06, "loss": 0.0056, "step": 50475 }, { "epoch": 15.53, "learning_rate": 2.5129103678684816e-06, "loss": 0.0153, "step": 50476 }, { "epoch": 15.53, "learning_rate": 2.512580074519768e-06, "loss": 0.0072, "step": 50477 }, { "epoch": 15.53, "learning_rate": 2.512249799760299e-06, "loss": 0.0024, "step": 50478 }, { "epoch": 15.53, "learning_rate": 2.511919543590895e-06, "loss": 0.0057, "step": 50479 }, { "epoch": 15.53, "learning_rate": 2.511589306012376e-06, "loss": 0.0073, "step": 50480 }, { "epoch": 15.53, "learning_rate": 2.5112590870255625e-06, "loss": 0.0093, "step": 50481 }, { "epoch": 15.53, "learning_rate": 2.5109288866312733e-06, "loss": 0.0078, "step": 50482 }, { "epoch": 15.53, "learning_rate": 2.510598704830329e-06, "loss": 0.0062, "step": 50483 }, { "epoch": 15.53, "learning_rate": 2.5102685416235497e-06, "loss": 0.0101, "step": 50484 }, { "epoch": 15.53, "learning_rate": 2.5099383970117498e-06, "loss": 0.014, "step": 50485 }, { "epoch": 15.53, "learning_rate": 2.509608270995758e-06, "loss": 0.0056, "step": 50486 }, { "epoch": 15.53, "learning_rate": 2.5092781635763885e-06, "loss": 0.0085, "step": 50487 }, { "epoch": 15.53, "learning_rate": 2.5089480747544657e-06, "loss": 0.0056, "step": 50488 }, { "epoch": 15.53, "learning_rate": 2.5086180045308018e-06, "loss": 0.0025, "step": 50489 }, { "epoch": 15.54, "learning_rate": 2.5082879529062165e-06, "loss": 0.0045, "step": 50490 }, { "epoch": 15.54, "learning_rate": 2.5079579198815363e-06, "loss": 0.0078, "step": 50491 }, { "epoch": 15.54, "learning_rate": 2.5076279054575768e-06, "loss": 0.0037, "step": 50492 }, { "epoch": 15.54, "learning_rate": 2.507297909635157e-06, "loss": 0.0077, "step": 50493 }, { "epoch": 15.54, "learning_rate": 2.5069679324150966e-06, "loss": 0.006, "step": 50494 }, { "epoch": 15.54, "learning_rate": 2.5066379737982136e-06, "loss": 0.007, "step": 50495 }, { "epoch": 15.54, "learning_rate": 2.5063080337853294e-06, "loss": 0.0149, "step": 50496 }, { "epoch": 15.54, "learning_rate": 2.505978112377263e-06, "loss": 0.0089, "step": 50497 }, { "epoch": 15.54, "learning_rate": 2.5056482095748313e-06, "loss": 0.006, "step": 50498 }, { "epoch": 15.54, "learning_rate": 2.505318325378856e-06, "loss": 0.0068, "step": 50499 }, { "epoch": 15.54, "learning_rate": 2.5049884597901508e-06, "loss": 0.0127, "step": 50500 }, { "epoch": 15.54, "learning_rate": 2.504658612809544e-06, "loss": 0.0084, "step": 50501 }, { "epoch": 15.54, "learning_rate": 2.504328784437847e-06, "loss": 0.0091, "step": 50502 }, { "epoch": 15.54, "learning_rate": 2.503998974675881e-06, "loss": 0.0086, "step": 50503 }, { "epoch": 15.54, "learning_rate": 2.503669183524464e-06, "loss": 0.0119, "step": 50504 }, { "epoch": 15.54, "learning_rate": 2.5033394109844135e-06, "loss": 0.0063, "step": 50505 }, { "epoch": 15.54, "learning_rate": 2.503009657056553e-06, "loss": 0.0076, "step": 50506 }, { "epoch": 15.54, "learning_rate": 2.5026799217416986e-06, "loss": 0.0101, "step": 50507 }, { "epoch": 15.54, "learning_rate": 2.502350205040668e-06, "loss": 0.0116, "step": 50508 }, { "epoch": 15.54, "learning_rate": 2.5020205069542834e-06, "loss": 0.0079, "step": 50509 }, { "epoch": 15.54, "learning_rate": 2.501690827483355e-06, "loss": 0.0068, "step": 50510 }, { "epoch": 15.54, "learning_rate": 2.5013611666287096e-06, "loss": 0.0125, "step": 50511 }, { "epoch": 15.54, "learning_rate": 2.501031524391163e-06, "loss": 0.0141, "step": 50512 }, { "epoch": 15.54, "learning_rate": 2.500701900771533e-06, "loss": 0.0098, "step": 50513 }, { "epoch": 15.54, "learning_rate": 2.500372295770639e-06, "loss": 0.019, "step": 50514 }, { "epoch": 15.54, "learning_rate": 2.500042709389299e-06, "loss": 0.0122, "step": 50515 }, { "epoch": 15.54, "learning_rate": 2.4997131416283304e-06, "loss": 0.0083, "step": 50516 }, { "epoch": 15.54, "learning_rate": 2.4993835924885524e-06, "loss": 0.0064, "step": 50517 }, { "epoch": 15.54, "learning_rate": 2.499054061970783e-06, "loss": 0.0088, "step": 50518 }, { "epoch": 15.54, "learning_rate": 2.4987245500758393e-06, "loss": 0.0067, "step": 50519 }, { "epoch": 15.54, "learning_rate": 2.4983950568045377e-06, "loss": 0.0089, "step": 50520 }, { "epoch": 15.54, "learning_rate": 2.4980655821577016e-06, "loss": 0.0161, "step": 50521 }, { "epoch": 15.55, "learning_rate": 2.4977361261361497e-06, "loss": 0.0086, "step": 50522 }, { "epoch": 15.55, "learning_rate": 2.497406688740693e-06, "loss": 0.0087, "step": 50523 }, { "epoch": 15.55, "learning_rate": 2.4970772699721535e-06, "loss": 0.0091, "step": 50524 }, { "epoch": 15.55, "learning_rate": 2.496747869831345e-06, "loss": 0.0042, "step": 50525 }, { "epoch": 15.55, "learning_rate": 2.4964184883190922e-06, "loss": 0.0085, "step": 50526 }, { "epoch": 15.55, "learning_rate": 2.4960891254362084e-06, "loss": 0.0131, "step": 50527 }, { "epoch": 15.55, "learning_rate": 2.495759781183513e-06, "loss": 0.0056, "step": 50528 }, { "epoch": 15.55, "learning_rate": 2.4954304555618225e-06, "loss": 0.0058, "step": 50529 }, { "epoch": 15.55, "learning_rate": 2.4951011485719556e-06, "loss": 0.008, "step": 50530 }, { "epoch": 15.55, "learning_rate": 2.4947718602147286e-06, "loss": 0.0048, "step": 50531 }, { "epoch": 15.55, "learning_rate": 2.494442590490961e-06, "loss": 0.0066, "step": 50532 }, { "epoch": 15.55, "learning_rate": 2.4941133394014684e-06, "loss": 0.0063, "step": 50533 }, { "epoch": 15.55, "learning_rate": 2.4937841069470683e-06, "loss": 0.0039, "step": 50534 }, { "epoch": 15.55, "learning_rate": 2.4934548931285784e-06, "loss": 0.0103, "step": 50535 }, { "epoch": 15.55, "learning_rate": 2.4931256979468175e-06, "loss": 0.0128, "step": 50536 }, { "epoch": 15.55, "learning_rate": 2.4927965214026008e-06, "loss": 0.0135, "step": 50537 }, { "epoch": 15.55, "learning_rate": 2.492467363496747e-06, "loss": 0.0076, "step": 50538 }, { "epoch": 15.55, "learning_rate": 2.492138224230072e-06, "loss": 0.0056, "step": 50539 }, { "epoch": 15.55, "learning_rate": 2.491809103603392e-06, "loss": 0.005, "step": 50540 }, { "epoch": 15.55, "learning_rate": 2.4914800016175276e-06, "loss": 0.0071, "step": 50541 }, { "epoch": 15.55, "learning_rate": 2.491150918273295e-06, "loss": 0.0122, "step": 50542 }, { "epoch": 15.55, "learning_rate": 2.490821853571512e-06, "loss": 0.0093, "step": 50543 }, { "epoch": 15.55, "learning_rate": 2.490492807512992e-06, "loss": 0.0121, "step": 50544 }, { "epoch": 15.55, "learning_rate": 2.4901637800985512e-06, "loss": 0.0064, "step": 50545 }, { "epoch": 15.55, "learning_rate": 2.4898347713290116e-06, "loss": 0.005, "step": 50546 }, { "epoch": 15.55, "learning_rate": 2.489505781205188e-06, "loss": 0.0055, "step": 50547 }, { "epoch": 15.55, "learning_rate": 2.4891768097278955e-06, "loss": 0.0106, "step": 50548 }, { "epoch": 15.55, "learning_rate": 2.4888478568979533e-06, "loss": 0.0049, "step": 50549 }, { "epoch": 15.55, "learning_rate": 2.4885189227161767e-06, "loss": 0.0077, "step": 50550 }, { "epoch": 15.55, "learning_rate": 2.488190007183383e-06, "loss": 0.0073, "step": 50551 }, { "epoch": 15.55, "learning_rate": 2.487861110300388e-06, "loss": 0.0062, "step": 50552 }, { "epoch": 15.55, "learning_rate": 2.4875322320680086e-06, "loss": 0.0039, "step": 50553 }, { "epoch": 15.55, "learning_rate": 2.4872033724870613e-06, "loss": 0.0055, "step": 50554 }, { "epoch": 15.56, "learning_rate": 2.4868745315583598e-06, "loss": 0.0081, "step": 50555 }, { "epoch": 15.56, "learning_rate": 2.4865457092827294e-06, "loss": 0.0082, "step": 50556 }, { "epoch": 15.56, "learning_rate": 2.4862169056609773e-06, "loss": 0.0105, "step": 50557 }, { "epoch": 15.56, "learning_rate": 2.4858881206939236e-06, "loss": 0.0042, "step": 50558 }, { "epoch": 15.56, "learning_rate": 2.485559354382383e-06, "loss": 0.0114, "step": 50559 }, { "epoch": 15.56, "learning_rate": 2.485230606727169e-06, "loss": 0.0056, "step": 50560 }, { "epoch": 15.56, "learning_rate": 2.4849018777291065e-06, "loss": 0.0058, "step": 50561 }, { "epoch": 15.56, "learning_rate": 2.484573167389005e-06, "loss": 0.0091, "step": 50562 }, { "epoch": 15.56, "learning_rate": 2.4842444757076833e-06, "loss": 0.0038, "step": 50563 }, { "epoch": 15.56, "learning_rate": 2.483915802685959e-06, "loss": 0.0096, "step": 50564 }, { "epoch": 15.56, "learning_rate": 2.48358714832464e-06, "loss": 0.0096, "step": 50565 }, { "epoch": 15.56, "learning_rate": 2.4832585126245503e-06, "loss": 0.0083, "step": 50566 }, { "epoch": 15.56, "learning_rate": 2.4829298955865022e-06, "loss": 0.0132, "step": 50567 }, { "epoch": 15.56, "learning_rate": 2.4826012972113145e-06, "loss": 0.0063, "step": 50568 }, { "epoch": 15.56, "learning_rate": 2.4822727174998e-06, "loss": 0.0099, "step": 50569 }, { "epoch": 15.56, "learning_rate": 2.481944156452776e-06, "loss": 0.0039, "step": 50570 }, { "epoch": 15.56, "learning_rate": 2.481615614071057e-06, "loss": 0.0052, "step": 50571 }, { "epoch": 15.56, "learning_rate": 2.481287090355461e-06, "loss": 0.0066, "step": 50572 }, { "epoch": 15.56, "learning_rate": 2.480958585306801e-06, "loss": 0.0053, "step": 50573 }, { "epoch": 15.56, "learning_rate": 2.480630098925895e-06, "loss": 0.0113, "step": 50574 }, { "epoch": 15.56, "learning_rate": 2.480301631213553e-06, "loss": 0.0071, "step": 50575 }, { "epoch": 15.56, "learning_rate": 2.4799731821705987e-06, "loss": 0.01, "step": 50576 }, { "epoch": 15.56, "learning_rate": 2.479644751797845e-06, "loss": 0.0073, "step": 50577 }, { "epoch": 15.56, "learning_rate": 2.4793163400961053e-06, "loss": 0.0129, "step": 50578 }, { "epoch": 15.56, "learning_rate": 2.478987947066194e-06, "loss": 0.0045, "step": 50579 }, { "epoch": 15.56, "learning_rate": 2.478659572708926e-06, "loss": 0.0082, "step": 50580 }, { "epoch": 15.56, "learning_rate": 2.4783312170251206e-06, "loss": 0.0085, "step": 50581 }, { "epoch": 15.56, "learning_rate": 2.478002880015592e-06, "loss": 0.0068, "step": 50582 }, { "epoch": 15.56, "learning_rate": 2.4776745616811538e-06, "loss": 0.01, "step": 50583 }, { "epoch": 15.56, "learning_rate": 2.477346262022622e-06, "loss": 0.0052, "step": 50584 }, { "epoch": 15.56, "learning_rate": 2.4770179810408114e-06, "loss": 0.0045, "step": 50585 }, { "epoch": 15.56, "learning_rate": 2.4766897187365368e-06, "loss": 0.0059, "step": 50586 }, { "epoch": 15.57, "learning_rate": 2.4763614751106134e-06, "loss": 0.0093, "step": 50587 }, { "epoch": 15.57, "learning_rate": 2.476033250163856e-06, "loss": 0.0096, "step": 50588 }, { "epoch": 15.57, "learning_rate": 2.47570504389708e-06, "loss": 0.0062, "step": 50589 }, { "epoch": 15.57, "learning_rate": 2.475376856311097e-06, "loss": 0.0082, "step": 50590 }, { "epoch": 15.57, "learning_rate": 2.4750486874067305e-06, "loss": 0.023, "step": 50591 }, { "epoch": 15.57, "learning_rate": 2.474720537184786e-06, "loss": 0.0127, "step": 50592 }, { "epoch": 15.57, "learning_rate": 2.4743924056460823e-06, "loss": 0.0047, "step": 50593 }, { "epoch": 15.57, "learning_rate": 2.4740642927914327e-06, "loss": 0.0103, "step": 50594 }, { "epoch": 15.57, "learning_rate": 2.47373619862165e-06, "loss": 0.0038, "step": 50595 }, { "epoch": 15.57, "learning_rate": 2.473408123137554e-06, "loss": 0.0078, "step": 50596 }, { "epoch": 15.57, "learning_rate": 2.4730800663399567e-06, "loss": 0.0038, "step": 50597 }, { "epoch": 15.57, "learning_rate": 2.472752028229676e-06, "loss": 0.0088, "step": 50598 }, { "epoch": 15.57, "learning_rate": 2.472424008807519e-06, "loss": 0.0108, "step": 50599 }, { "epoch": 15.57, "learning_rate": 2.472096008074301e-06, "loss": 0.0144, "step": 50600 }, { "epoch": 15.57, "learning_rate": 2.471768026030842e-06, "loss": 0.0074, "step": 50601 }, { "epoch": 15.57, "learning_rate": 2.4714400626779533e-06, "loss": 0.0047, "step": 50602 }, { "epoch": 15.57, "learning_rate": 2.47111211801645e-06, "loss": 0.0046, "step": 50603 }, { "epoch": 15.57, "learning_rate": 2.4707841920471454e-06, "loss": 0.0122, "step": 50604 }, { "epoch": 15.57, "learning_rate": 2.4704562847708535e-06, "loss": 0.0072, "step": 50605 }, { "epoch": 15.57, "learning_rate": 2.4701283961883894e-06, "loss": 0.0055, "step": 50606 }, { "epoch": 15.57, "learning_rate": 2.4698005263005655e-06, "loss": 0.0066, "step": 50607 }, { "epoch": 15.57, "learning_rate": 2.4694726751081978e-06, "loss": 0.0109, "step": 50608 }, { "epoch": 15.57, "learning_rate": 2.469144842612098e-06, "loss": 0.0167, "step": 50609 }, { "epoch": 15.57, "learning_rate": 2.4688170288130795e-06, "loss": 0.0072, "step": 50610 }, { "epoch": 15.57, "learning_rate": 2.4684892337119604e-06, "loss": 0.005, "step": 50611 }, { "epoch": 15.57, "learning_rate": 2.468161457309556e-06, "loss": 0.0086, "step": 50612 }, { "epoch": 15.57, "learning_rate": 2.467833699606672e-06, "loss": 0.0087, "step": 50613 }, { "epoch": 15.57, "learning_rate": 2.4675059606041265e-06, "loss": 0.0124, "step": 50614 }, { "epoch": 15.57, "learning_rate": 2.46717824030273e-06, "loss": 0.0049, "step": 50615 }, { "epoch": 15.57, "learning_rate": 2.4668505387033025e-06, "loss": 0.0077, "step": 50616 }, { "epoch": 15.57, "learning_rate": 2.4665228558066545e-06, "loss": 0.0114, "step": 50617 }, { "epoch": 15.57, "learning_rate": 2.4661951916135985e-06, "loss": 0.0065, "step": 50618 }, { "epoch": 15.57, "learning_rate": 2.465867546124953e-06, "loss": 0.0084, "step": 50619 }, { "epoch": 15.58, "learning_rate": 2.46553991934152e-06, "loss": 0.0062, "step": 50620 }, { "epoch": 15.58, "learning_rate": 2.4652123112641235e-06, "loss": 0.0124, "step": 50621 }, { "epoch": 15.58, "learning_rate": 2.4648847218935734e-06, "loss": 0.0033, "step": 50622 }, { "epoch": 15.58, "learning_rate": 2.4645571512306833e-06, "loss": 0.0091, "step": 50623 }, { "epoch": 15.58, "learning_rate": 2.4642295992762665e-06, "loss": 0.0083, "step": 50624 }, { "epoch": 15.58, "learning_rate": 2.4639020660311364e-06, "loss": 0.0061, "step": 50625 }, { "epoch": 15.58, "learning_rate": 2.463574551496105e-06, "loss": 0.0069, "step": 50626 }, { "epoch": 15.58, "learning_rate": 2.463247055671987e-06, "loss": 0.0122, "step": 50627 }, { "epoch": 15.58, "learning_rate": 2.462919578559594e-06, "loss": 0.0101, "step": 50628 }, { "epoch": 15.58, "learning_rate": 2.462592120159739e-06, "loss": 0.0105, "step": 50629 }, { "epoch": 15.58, "learning_rate": 2.4622646804732342e-06, "loss": 0.0084, "step": 50630 }, { "epoch": 15.58, "learning_rate": 2.4619372595008974e-06, "loss": 0.0112, "step": 50631 }, { "epoch": 15.58, "learning_rate": 2.4616098572435363e-06, "loss": 0.007, "step": 50632 }, { "epoch": 15.58, "learning_rate": 2.4612824737019704e-06, "loss": 0.012, "step": 50633 }, { "epoch": 15.58, "learning_rate": 2.4609551088770035e-06, "loss": 0.0078, "step": 50634 }, { "epoch": 15.58, "learning_rate": 2.460627762769451e-06, "loss": 0.0101, "step": 50635 }, { "epoch": 15.58, "learning_rate": 2.4603004353801284e-06, "loss": 0.0078, "step": 50636 }, { "epoch": 15.58, "learning_rate": 2.4599731267098482e-06, "loss": 0.0188, "step": 50637 }, { "epoch": 15.58, "learning_rate": 2.459645836759422e-06, "loss": 0.0074, "step": 50638 }, { "epoch": 15.58, "learning_rate": 2.4593185655296626e-06, "loss": 0.0027, "step": 50639 }, { "epoch": 15.58, "learning_rate": 2.4589913130213815e-06, "loss": 0.0091, "step": 50640 }, { "epoch": 15.58, "learning_rate": 2.458664079235393e-06, "loss": 0.012, "step": 50641 }, { "epoch": 15.58, "learning_rate": 2.458336864172508e-06, "loss": 0.0099, "step": 50642 }, { "epoch": 15.58, "learning_rate": 2.4580096678335397e-06, "loss": 0.0056, "step": 50643 }, { "epoch": 15.58, "learning_rate": 2.4576824902192964e-06, "loss": 0.0078, "step": 50644 }, { "epoch": 15.58, "learning_rate": 2.4573553313305976e-06, "loss": 0.0182, "step": 50645 }, { "epoch": 15.58, "learning_rate": 2.4570281911682557e-06, "loss": 0.0049, "step": 50646 }, { "epoch": 15.58, "learning_rate": 2.456701069733075e-06, "loss": 0.0045, "step": 50647 }, { "epoch": 15.58, "learning_rate": 2.456373967025872e-06, "loss": 0.0122, "step": 50648 }, { "epoch": 15.58, "learning_rate": 2.456046883047457e-06, "loss": 0.0098, "step": 50649 }, { "epoch": 15.58, "learning_rate": 2.4557198177986464e-06, "loss": 0.0032, "step": 50650 }, { "epoch": 15.58, "learning_rate": 2.4553927712802493e-06, "loss": 0.0065, "step": 50651 }, { "epoch": 15.59, "learning_rate": 2.455065743493078e-06, "loss": 0.0079, "step": 50652 }, { "epoch": 15.59, "learning_rate": 2.454738734437945e-06, "loss": 0.0052, "step": 50653 }, { "epoch": 15.59, "learning_rate": 2.4544117441156602e-06, "loss": 0.0033, "step": 50654 }, { "epoch": 15.59, "learning_rate": 2.4540847725270376e-06, "loss": 0.0047, "step": 50655 }, { "epoch": 15.59, "learning_rate": 2.4537578196728886e-06, "loss": 0.0043, "step": 50656 }, { "epoch": 15.59, "learning_rate": 2.4534308855540246e-06, "loss": 0.0162, "step": 50657 }, { "epoch": 15.59, "learning_rate": 2.4531039701712566e-06, "loss": 0.0057, "step": 50658 }, { "epoch": 15.59, "learning_rate": 2.4527770735253965e-06, "loss": 0.012, "step": 50659 }, { "epoch": 15.59, "learning_rate": 2.4524501956172563e-06, "loss": 0.0083, "step": 50660 }, { "epoch": 15.59, "learning_rate": 2.452123336447648e-06, "loss": 0.013, "step": 50661 }, { "epoch": 15.59, "learning_rate": 2.4517964960173836e-06, "loss": 0.0081, "step": 50662 }, { "epoch": 15.59, "learning_rate": 2.4514696743272724e-06, "loss": 0.025, "step": 50663 }, { "epoch": 15.59, "learning_rate": 2.451142871378124e-06, "loss": 0.0132, "step": 50664 }, { "epoch": 15.59, "learning_rate": 2.4508160871707554e-06, "loss": 0.0133, "step": 50665 }, { "epoch": 15.59, "learning_rate": 2.4504893217059768e-06, "loss": 0.0101, "step": 50666 }, { "epoch": 15.59, "learning_rate": 2.4501625749845994e-06, "loss": 0.0108, "step": 50667 }, { "epoch": 15.59, "learning_rate": 2.44983584700743e-06, "loss": 0.0081, "step": 50668 }, { "epoch": 15.59, "learning_rate": 2.44950913777528e-06, "loss": 0.0057, "step": 50669 }, { "epoch": 15.59, "learning_rate": 2.4491824472889667e-06, "loss": 0.0053, "step": 50670 }, { "epoch": 15.59, "learning_rate": 2.448855775549298e-06, "loss": 0.0142, "step": 50671 }, { "epoch": 15.59, "learning_rate": 2.448529122557084e-06, "loss": 0.0087, "step": 50672 }, { "epoch": 15.59, "learning_rate": 2.448202488313136e-06, "loss": 0.0057, "step": 50673 }, { "epoch": 15.59, "learning_rate": 2.447875872818266e-06, "loss": 0.0062, "step": 50674 }, { "epoch": 15.59, "learning_rate": 2.4475492760732843e-06, "loss": 0.0068, "step": 50675 }, { "epoch": 15.59, "learning_rate": 2.4472226980790015e-06, "loss": 0.0059, "step": 50676 }, { "epoch": 15.59, "learning_rate": 2.4468961388362278e-06, "loss": 0.0323, "step": 50677 }, { "epoch": 15.59, "learning_rate": 2.446569598345776e-06, "loss": 0.0049, "step": 50678 }, { "epoch": 15.59, "learning_rate": 2.446243076608452e-06, "loss": 0.0066, "step": 50679 }, { "epoch": 15.59, "learning_rate": 2.4459165736250756e-06, "loss": 0.0078, "step": 50680 }, { "epoch": 15.59, "learning_rate": 2.445590089396449e-06, "loss": 0.0057, "step": 50681 }, { "epoch": 15.59, "learning_rate": 2.445263623923386e-06, "loss": 0.0215, "step": 50682 }, { "epoch": 15.59, "learning_rate": 2.4449371772066965e-06, "loss": 0.0109, "step": 50683 }, { "epoch": 15.59, "learning_rate": 2.444610749247188e-06, "loss": 0.0108, "step": 50684 }, { "epoch": 15.6, "learning_rate": 2.4442843400456784e-06, "loss": 0.0124, "step": 50685 }, { "epoch": 15.6, "learning_rate": 2.443957949602973e-06, "loss": 0.0137, "step": 50686 }, { "epoch": 15.6, "learning_rate": 2.4436315779198827e-06, "loss": 0.0065, "step": 50687 }, { "epoch": 15.6, "learning_rate": 2.443305224997221e-06, "loss": 0.0039, "step": 50688 }, { "epoch": 15.6, "learning_rate": 2.4429788908357886e-06, "loss": 0.0061, "step": 50689 }, { "epoch": 15.6, "learning_rate": 2.442652575436406e-06, "loss": 0.0088, "step": 50690 }, { "epoch": 15.6, "learning_rate": 2.4423262787998805e-06, "loss": 0.013, "step": 50691 }, { "epoch": 15.6, "learning_rate": 2.4420000009270205e-06, "loss": 0.0066, "step": 50692 }, { "epoch": 15.6, "learning_rate": 2.4416737418186364e-06, "loss": 0.006, "step": 50693 }, { "epoch": 15.6, "learning_rate": 2.4413475014755396e-06, "loss": 0.0045, "step": 50694 }, { "epoch": 15.6, "learning_rate": 2.4410212798985387e-06, "loss": 0.0133, "step": 50695 }, { "epoch": 15.6, "learning_rate": 2.440695077088445e-06, "loss": 0.006, "step": 50696 }, { "epoch": 15.6, "learning_rate": 2.440368893046067e-06, "loss": 0.0097, "step": 50697 }, { "epoch": 15.6, "learning_rate": 2.440042727772215e-06, "loss": 0.0154, "step": 50698 }, { "epoch": 15.6, "learning_rate": 2.439716581267695e-06, "loss": 0.0079, "step": 50699 }, { "epoch": 15.6, "learning_rate": 2.4393904535333244e-06, "loss": 0.0076, "step": 50700 }, { "epoch": 15.6, "learning_rate": 2.439064344569911e-06, "loss": 0.0057, "step": 50701 }, { "epoch": 15.6, "learning_rate": 2.4387382543782613e-06, "loss": 0.0088, "step": 50702 }, { "epoch": 15.6, "learning_rate": 2.4384121829591858e-06, "loss": 0.0119, "step": 50703 }, { "epoch": 15.6, "learning_rate": 2.4380861303134893e-06, "loss": 0.0291, "step": 50704 }, { "epoch": 15.6, "learning_rate": 2.4377600964419914e-06, "loss": 0.0125, "step": 50705 }, { "epoch": 15.6, "learning_rate": 2.4374340813454954e-06, "loss": 0.0061, "step": 50706 }, { "epoch": 15.6, "learning_rate": 2.437108085024812e-06, "loss": 0.0065, "step": 50707 }, { "epoch": 15.6, "learning_rate": 2.4367821074807506e-06, "loss": 0.0042, "step": 50708 }, { "epoch": 15.6, "learning_rate": 2.4364561487141204e-06, "loss": 0.0114, "step": 50709 }, { "epoch": 15.6, "learning_rate": 2.43613020872573e-06, "loss": 0.0045, "step": 50710 }, { "epoch": 15.6, "learning_rate": 2.435804287516389e-06, "loss": 0.0104, "step": 50711 }, { "epoch": 15.6, "learning_rate": 2.435478385086907e-06, "loss": 0.0143, "step": 50712 }, { "epoch": 15.6, "learning_rate": 2.4351525014380926e-06, "loss": 0.005, "step": 50713 }, { "epoch": 15.6, "learning_rate": 2.434826636570755e-06, "loss": 0.009, "step": 50714 }, { "epoch": 15.6, "learning_rate": 2.434500790485704e-06, "loss": 0.0119, "step": 50715 }, { "epoch": 15.6, "learning_rate": 2.4341749631837475e-06, "loss": 0.0121, "step": 50716 }, { "epoch": 15.61, "learning_rate": 2.433849154665695e-06, "loss": 0.01, "step": 50717 }, { "epoch": 15.61, "learning_rate": 2.433523364932355e-06, "loss": 0.0031, "step": 50718 }, { "epoch": 15.61, "learning_rate": 2.4331975939845342e-06, "loss": 0.0092, "step": 50719 }, { "epoch": 15.61, "learning_rate": 2.432871841823047e-06, "loss": 0.0061, "step": 50720 }, { "epoch": 15.61, "learning_rate": 2.432546108448699e-06, "loss": 0.0061, "step": 50721 }, { "epoch": 15.61, "learning_rate": 2.432220393862301e-06, "loss": 0.0063, "step": 50722 }, { "epoch": 15.61, "learning_rate": 2.431894698064656e-06, "loss": 0.0177, "step": 50723 }, { "epoch": 15.61, "learning_rate": 2.4315690210565745e-06, "loss": 0.0031, "step": 50724 }, { "epoch": 15.61, "learning_rate": 2.4312433628388697e-06, "loss": 0.0072, "step": 50725 }, { "epoch": 15.61, "learning_rate": 2.430917723412347e-06, "loss": 0.01, "step": 50726 }, { "epoch": 15.61, "learning_rate": 2.4305921027778147e-06, "loss": 0.0102, "step": 50727 }, { "epoch": 15.61, "learning_rate": 2.4302665009360814e-06, "loss": 0.017, "step": 50728 }, { "epoch": 15.61, "learning_rate": 2.4299409178879564e-06, "loss": 0.0076, "step": 50729 }, { "epoch": 15.61, "learning_rate": 2.429615353634247e-06, "loss": 0.0068, "step": 50730 }, { "epoch": 15.61, "learning_rate": 2.429289808175761e-06, "loss": 0.0124, "step": 50731 }, { "epoch": 15.61, "learning_rate": 2.4289642815133086e-06, "loss": 0.0067, "step": 50732 }, { "epoch": 15.61, "learning_rate": 2.428638773647696e-06, "loss": 0.0089, "step": 50733 }, { "epoch": 15.61, "learning_rate": 2.4283132845797297e-06, "loss": 0.0157, "step": 50734 }, { "epoch": 15.61, "learning_rate": 2.4279878143102253e-06, "loss": 0.0106, "step": 50735 }, { "epoch": 15.61, "learning_rate": 2.427662362839984e-06, "loss": 0.0068, "step": 50736 }, { "epoch": 15.61, "learning_rate": 2.427336930169815e-06, "loss": 0.0112, "step": 50737 }, { "epoch": 15.61, "learning_rate": 2.427011516300527e-06, "loss": 0.0072, "step": 50738 }, { "epoch": 15.61, "learning_rate": 2.426686121232924e-06, "loss": 0.0104, "step": 50739 }, { "epoch": 15.61, "learning_rate": 2.4263607449678216e-06, "loss": 0.0098, "step": 50740 }, { "epoch": 15.61, "learning_rate": 2.4260353875060237e-06, "loss": 0.0076, "step": 50741 }, { "epoch": 15.61, "learning_rate": 2.4257100488483377e-06, "loss": 0.0059, "step": 50742 }, { "epoch": 15.61, "learning_rate": 2.4253847289955743e-06, "loss": 0.0067, "step": 50743 }, { "epoch": 15.61, "learning_rate": 2.425059427948534e-06, "loss": 0.0122, "step": 50744 }, { "epoch": 15.61, "learning_rate": 2.424734145708031e-06, "loss": 0.0053, "step": 50745 }, { "epoch": 15.61, "learning_rate": 2.42440888227487e-06, "loss": 0.0127, "step": 50746 }, { "epoch": 15.61, "learning_rate": 2.424083637649861e-06, "loss": 0.0071, "step": 50747 }, { "epoch": 15.61, "learning_rate": 2.4237584118338086e-06, "loss": 0.0126, "step": 50748 }, { "epoch": 15.61, "learning_rate": 2.423433204827522e-06, "loss": 0.0086, "step": 50749 }, { "epoch": 15.62, "learning_rate": 2.4231080166318087e-06, "loss": 0.0097, "step": 50750 }, { "epoch": 15.62, "learning_rate": 2.4227828472474755e-06, "loss": 0.0107, "step": 50751 }, { "epoch": 15.62, "learning_rate": 2.42245769667533e-06, "loss": 0.0047, "step": 50752 }, { "epoch": 15.62, "learning_rate": 2.422132564916179e-06, "loss": 0.0141, "step": 50753 }, { "epoch": 15.62, "learning_rate": 2.421807451970826e-06, "loss": 0.0078, "step": 50754 }, { "epoch": 15.62, "learning_rate": 2.421482357840086e-06, "loss": 0.0126, "step": 50755 }, { "epoch": 15.62, "learning_rate": 2.421157282524764e-06, "loss": 0.0086, "step": 50756 }, { "epoch": 15.62, "learning_rate": 2.4208322260256634e-06, "loss": 0.0082, "step": 50757 }, { "epoch": 15.62, "learning_rate": 2.4205071883435928e-06, "loss": 0.0057, "step": 50758 }, { "epoch": 15.62, "learning_rate": 2.420182169479357e-06, "loss": 0.0077, "step": 50759 }, { "epoch": 15.62, "learning_rate": 2.419857169433768e-06, "loss": 0.0048, "step": 50760 }, { "epoch": 15.62, "learning_rate": 2.4195321882076295e-06, "loss": 0.0055, "step": 50761 }, { "epoch": 15.62, "learning_rate": 2.4192072258017496e-06, "loss": 0.0078, "step": 50762 }, { "epoch": 15.62, "learning_rate": 2.4188822822169346e-06, "loss": 0.0056, "step": 50763 }, { "epoch": 15.62, "learning_rate": 2.4185573574539913e-06, "loss": 0.0071, "step": 50764 }, { "epoch": 15.62, "learning_rate": 2.4182324515137256e-06, "loss": 0.0096, "step": 50765 }, { "epoch": 15.62, "learning_rate": 2.417907564396944e-06, "loss": 0.0081, "step": 50766 }, { "epoch": 15.62, "learning_rate": 2.4175826961044547e-06, "loss": 0.0082, "step": 50767 }, { "epoch": 15.62, "learning_rate": 2.417257846637063e-06, "loss": 0.004, "step": 50768 }, { "epoch": 15.62, "learning_rate": 2.416933015995574e-06, "loss": 0.0095, "step": 50769 }, { "epoch": 15.62, "learning_rate": 2.4166082041808013e-06, "loss": 0.0059, "step": 50770 }, { "epoch": 15.62, "learning_rate": 2.416283411193543e-06, "loss": 0.0114, "step": 50771 }, { "epoch": 15.62, "learning_rate": 2.415958637034609e-06, "loss": 0.0099, "step": 50772 }, { "epoch": 15.62, "learning_rate": 2.415633881704804e-06, "loss": 0.0124, "step": 50773 }, { "epoch": 15.62, "learning_rate": 2.4153091452049324e-06, "loss": 0.01, "step": 50774 }, { "epoch": 15.62, "learning_rate": 2.414984427535807e-06, "loss": 0.0079, "step": 50775 }, { "epoch": 15.62, "learning_rate": 2.414659728698231e-06, "loss": 0.0124, "step": 50776 }, { "epoch": 15.62, "learning_rate": 2.414335048693012e-06, "loss": 0.0081, "step": 50777 }, { "epoch": 15.62, "learning_rate": 2.414010387520951e-06, "loss": 0.0192, "step": 50778 }, { "epoch": 15.62, "learning_rate": 2.4136857451828542e-06, "loss": 0.0076, "step": 50779 }, { "epoch": 15.62, "learning_rate": 2.413361121679534e-06, "loss": 0.0047, "step": 50780 }, { "epoch": 15.62, "learning_rate": 2.413036517011793e-06, "loss": 0.0056, "step": 50781 }, { "epoch": 15.63, "learning_rate": 2.4127119311804358e-06, "loss": 0.0086, "step": 50782 }, { "epoch": 15.63, "learning_rate": 2.4123873641862705e-06, "loss": 0.0124, "step": 50783 }, { "epoch": 15.63, "learning_rate": 2.412062816030102e-06, "loss": 0.0109, "step": 50784 }, { "epoch": 15.63, "learning_rate": 2.411738286712735e-06, "loss": 0.0091, "step": 50785 }, { "epoch": 15.63, "learning_rate": 2.4114137762349766e-06, "loss": 0.0074, "step": 50786 }, { "epoch": 15.63, "learning_rate": 2.411089284597632e-06, "loss": 0.0097, "step": 50787 }, { "epoch": 15.63, "learning_rate": 2.410764811801506e-06, "loss": 0.0041, "step": 50788 }, { "epoch": 15.63, "learning_rate": 2.4104403578474023e-06, "loss": 0.006, "step": 50789 }, { "epoch": 15.63, "learning_rate": 2.4101159227361326e-06, "loss": 0.006, "step": 50790 }, { "epoch": 15.63, "learning_rate": 2.4097915064685016e-06, "loss": 0.0066, "step": 50791 }, { "epoch": 15.63, "learning_rate": 2.409467109045309e-06, "loss": 0.013, "step": 50792 }, { "epoch": 15.63, "learning_rate": 2.4091427304673633e-06, "loss": 0.0075, "step": 50793 }, { "epoch": 15.63, "learning_rate": 2.4088183707354673e-06, "loss": 0.0054, "step": 50794 }, { "epoch": 15.63, "learning_rate": 2.408494029850431e-06, "loss": 0.0067, "step": 50795 }, { "epoch": 15.63, "learning_rate": 2.408169707813057e-06, "loss": 0.0073, "step": 50796 }, { "epoch": 15.63, "learning_rate": 2.4078454046241528e-06, "loss": 0.0084, "step": 50797 }, { "epoch": 15.63, "learning_rate": 2.407521120284523e-06, "loss": 0.0142, "step": 50798 }, { "epoch": 15.63, "learning_rate": 2.4071968547949652e-06, "loss": 0.0095, "step": 50799 }, { "epoch": 15.63, "learning_rate": 2.4068726081562953e-06, "loss": 0.0256, "step": 50800 }, { "epoch": 15.63, "learning_rate": 2.406548380369312e-06, "loss": 0.0063, "step": 50801 }, { "epoch": 15.63, "learning_rate": 2.406224171434823e-06, "loss": 0.0046, "step": 50802 }, { "epoch": 15.63, "learning_rate": 2.405899981353632e-06, "loss": 0.0082, "step": 50803 }, { "epoch": 15.63, "learning_rate": 2.4055758101265438e-06, "loss": 0.009, "step": 50804 }, { "epoch": 15.63, "learning_rate": 2.405251657754363e-06, "loss": 0.0071, "step": 50805 }, { "epoch": 15.63, "learning_rate": 2.4049275242378955e-06, "loss": 0.0073, "step": 50806 }, { "epoch": 15.63, "learning_rate": 2.4046034095779456e-06, "loss": 0.009, "step": 50807 }, { "epoch": 15.63, "learning_rate": 2.4042793137753174e-06, "loss": 0.0276, "step": 50808 }, { "epoch": 15.63, "learning_rate": 2.4039552368308127e-06, "loss": 0.0095, "step": 50809 }, { "epoch": 15.63, "learning_rate": 2.403631178745243e-06, "loss": 0.0071, "step": 50810 }, { "epoch": 15.63, "learning_rate": 2.403307139519411e-06, "loss": 0.0047, "step": 50811 }, { "epoch": 15.63, "learning_rate": 2.4029831191541176e-06, "loss": 0.0105, "step": 50812 }, { "epoch": 15.63, "learning_rate": 2.402659117650168e-06, "loss": 0.0034, "step": 50813 }, { "epoch": 15.63, "learning_rate": 2.4023351350083657e-06, "loss": 0.0121, "step": 50814 }, { "epoch": 15.64, "learning_rate": 2.4020111712295193e-06, "loss": 0.0076, "step": 50815 }, { "epoch": 15.64, "learning_rate": 2.4016872263144308e-06, "loss": 0.0139, "step": 50816 }, { "epoch": 15.64, "learning_rate": 2.401363300263905e-06, "loss": 0.016, "step": 50817 }, { "epoch": 15.64, "learning_rate": 2.4010393930787445e-06, "loss": 0.0075, "step": 50818 }, { "epoch": 15.64, "learning_rate": 2.400715504759755e-06, "loss": 0.0059, "step": 50819 }, { "epoch": 15.64, "learning_rate": 2.4003916353077416e-06, "loss": 0.0064, "step": 50820 }, { "epoch": 15.64, "learning_rate": 2.4000677847235056e-06, "loss": 0.0104, "step": 50821 }, { "epoch": 15.64, "learning_rate": 2.3997439530078526e-06, "loss": 0.0092, "step": 50822 }, { "epoch": 15.64, "learning_rate": 2.399420140161587e-06, "loss": 0.0111, "step": 50823 }, { "epoch": 15.64, "learning_rate": 2.3990963461855075e-06, "loss": 0.0082, "step": 50824 }, { "epoch": 15.64, "learning_rate": 2.39877257108043e-06, "loss": 0.0069, "step": 50825 }, { "epoch": 15.64, "learning_rate": 2.3984488148471473e-06, "loss": 0.0074, "step": 50826 }, { "epoch": 15.64, "learning_rate": 2.3981250774864686e-06, "loss": 0.0082, "step": 50827 }, { "epoch": 15.64, "learning_rate": 2.3978013589991946e-06, "loss": 0.0071, "step": 50828 }, { "epoch": 15.64, "learning_rate": 2.397477659386127e-06, "loss": 0.0115, "step": 50829 }, { "epoch": 15.64, "learning_rate": 2.3971539786480757e-06, "loss": 0.0141, "step": 50830 }, { "epoch": 15.64, "learning_rate": 2.396830316785842e-06, "loss": 0.0091, "step": 50831 }, { "epoch": 15.64, "learning_rate": 2.3965066738002316e-06, "loss": 0.0104, "step": 50832 }, { "epoch": 15.64, "learning_rate": 2.3961830496920423e-06, "loss": 0.0114, "step": 50833 }, { "epoch": 15.64, "learning_rate": 2.3958594444620786e-06, "loss": 0.0122, "step": 50834 }, { "epoch": 15.64, "learning_rate": 2.3955358581111477e-06, "loss": 0.0146, "step": 50835 }, { "epoch": 15.64, "learning_rate": 2.395212290640051e-06, "loss": 0.0102, "step": 50836 }, { "epoch": 15.64, "learning_rate": 2.3948887420495927e-06, "loss": 0.005, "step": 50837 }, { "epoch": 15.64, "learning_rate": 2.394565212340575e-06, "loss": 0.0061, "step": 50838 }, { "epoch": 15.64, "learning_rate": 2.394241701513802e-06, "loss": 0.0114, "step": 50839 }, { "epoch": 15.64, "learning_rate": 2.393918209570075e-06, "loss": 0.0071, "step": 50840 }, { "epoch": 15.64, "learning_rate": 2.3935947365102008e-06, "loss": 0.0103, "step": 50841 }, { "epoch": 15.64, "learning_rate": 2.393271282334978e-06, "loss": 0.0061, "step": 50842 }, { "epoch": 15.64, "learning_rate": 2.392947847045214e-06, "loss": 0.0077, "step": 50843 }, { "epoch": 15.64, "learning_rate": 2.392624430641706e-06, "loss": 0.005, "step": 50844 }, { "epoch": 15.64, "learning_rate": 2.392301033125263e-06, "loss": 0.0085, "step": 50845 }, { "epoch": 15.64, "learning_rate": 2.3919776544966876e-06, "loss": 0.0061, "step": 50846 }, { "epoch": 15.65, "learning_rate": 2.391654294756779e-06, "loss": 0.0109, "step": 50847 }, { "epoch": 15.65, "learning_rate": 2.3913309539063424e-06, "loss": 0.0048, "step": 50848 }, { "epoch": 15.65, "learning_rate": 2.391007631946175e-06, "loss": 0.0047, "step": 50849 }, { "epoch": 15.65, "learning_rate": 2.390684328877089e-06, "loss": 0.0066, "step": 50850 }, { "epoch": 15.65, "learning_rate": 2.3903610446998815e-06, "loss": 0.0053, "step": 50851 }, { "epoch": 15.65, "learning_rate": 2.390037779415356e-06, "loss": 0.0157, "step": 50852 }, { "epoch": 15.65, "learning_rate": 2.3897145330243186e-06, "loss": 0.0112, "step": 50853 }, { "epoch": 15.65, "learning_rate": 2.389391305527562e-06, "loss": 0.0086, "step": 50854 }, { "epoch": 15.65, "learning_rate": 2.3890680969258985e-06, "loss": 0.0095, "step": 50855 }, { "epoch": 15.65, "learning_rate": 2.388744907220126e-06, "loss": 0.0125, "step": 50856 }, { "epoch": 15.65, "learning_rate": 2.3884217364110485e-06, "loss": 0.0095, "step": 50857 }, { "epoch": 15.65, "learning_rate": 2.3880985844994674e-06, "loss": 0.0105, "step": 50858 }, { "epoch": 15.65, "learning_rate": 2.387775451486186e-06, "loss": 0.0063, "step": 50859 }, { "epoch": 15.65, "learning_rate": 2.387452337372006e-06, "loss": 0.0156, "step": 50860 }, { "epoch": 15.65, "learning_rate": 2.3871292421577298e-06, "loss": 0.0043, "step": 50861 }, { "epoch": 15.65, "learning_rate": 2.386806165844159e-06, "loss": 0.0205, "step": 50862 }, { "epoch": 15.65, "learning_rate": 2.386483108432096e-06, "loss": 0.0035, "step": 50863 }, { "epoch": 15.65, "learning_rate": 2.3861600699223397e-06, "loss": 0.0107, "step": 50864 }, { "epoch": 15.65, "learning_rate": 2.385837050315699e-06, "loss": 0.0082, "step": 50865 }, { "epoch": 15.65, "learning_rate": 2.3855140496129714e-06, "loss": 0.0063, "step": 50866 }, { "epoch": 15.65, "learning_rate": 2.3851910678149625e-06, "loss": 0.0094, "step": 50867 }, { "epoch": 15.65, "learning_rate": 2.3848681049224686e-06, "loss": 0.0133, "step": 50868 }, { "epoch": 15.65, "learning_rate": 2.3845451609362926e-06, "loss": 0.0067, "step": 50869 }, { "epoch": 15.65, "learning_rate": 2.3842222358572387e-06, "loss": 0.0066, "step": 50870 }, { "epoch": 15.65, "learning_rate": 2.383899329686109e-06, "loss": 0.0094, "step": 50871 }, { "epoch": 15.65, "learning_rate": 2.383576442423704e-06, "loss": 0.0048, "step": 50872 }, { "epoch": 15.65, "learning_rate": 2.383253574070825e-06, "loss": 0.0079, "step": 50873 }, { "epoch": 15.65, "learning_rate": 2.382930724628274e-06, "loss": 0.0093, "step": 50874 }, { "epoch": 15.65, "learning_rate": 2.3826078940968535e-06, "loss": 0.005, "step": 50875 }, { "epoch": 15.65, "learning_rate": 2.3822850824773623e-06, "loss": 0.0143, "step": 50876 }, { "epoch": 15.65, "learning_rate": 2.3819622897706053e-06, "loss": 0.0062, "step": 50877 }, { "epoch": 15.65, "learning_rate": 2.381639515977381e-06, "loss": 0.0083, "step": 50878 }, { "epoch": 15.65, "learning_rate": 2.381316761098489e-06, "loss": 0.0245, "step": 50879 }, { "epoch": 15.66, "learning_rate": 2.3809940251347396e-06, "loss": 0.006, "step": 50880 }, { "epoch": 15.66, "learning_rate": 2.3806713080869247e-06, "loss": 0.0071, "step": 50881 }, { "epoch": 15.66, "learning_rate": 2.3803486099558504e-06, "loss": 0.0148, "step": 50882 }, { "epoch": 15.66, "learning_rate": 2.3800259307423146e-06, "loss": 0.0053, "step": 50883 }, { "epoch": 15.66, "learning_rate": 2.379703270447118e-06, "loss": 0.0097, "step": 50884 }, { "epoch": 15.66, "learning_rate": 2.3793806290710665e-06, "loss": 0.0026, "step": 50885 }, { "epoch": 15.66, "learning_rate": 2.3790580066149583e-06, "loss": 0.0133, "step": 50886 }, { "epoch": 15.66, "learning_rate": 2.3787354030795952e-06, "loss": 0.0073, "step": 50887 }, { "epoch": 15.66, "learning_rate": 2.3784128184657797e-06, "loss": 0.0105, "step": 50888 }, { "epoch": 15.66, "learning_rate": 2.3780902527743054e-06, "loss": 0.0093, "step": 50889 }, { "epoch": 15.66, "learning_rate": 2.37776770600598e-06, "loss": 0.0081, "step": 50890 }, { "epoch": 15.66, "learning_rate": 2.377445178161604e-06, "loss": 0.0072, "step": 50891 }, { "epoch": 15.66, "learning_rate": 2.3771226692419745e-06, "loss": 0.0079, "step": 50892 }, { "epoch": 15.66, "learning_rate": 2.376800179247896e-06, "loss": 0.006, "step": 50893 }, { "epoch": 15.66, "learning_rate": 2.376477708180167e-06, "loss": 0.0087, "step": 50894 }, { "epoch": 15.66, "learning_rate": 2.3761552560395883e-06, "loss": 0.0074, "step": 50895 }, { "epoch": 15.66, "learning_rate": 2.375832822826962e-06, "loss": 0.0049, "step": 50896 }, { "epoch": 15.66, "learning_rate": 2.3755104085430858e-06, "loss": 0.0046, "step": 50897 }, { "epoch": 15.66, "learning_rate": 2.3751880131887627e-06, "loss": 0.006, "step": 50898 }, { "epoch": 15.66, "learning_rate": 2.374865636764789e-06, "loss": 0.0061, "step": 50899 }, { "epoch": 15.66, "learning_rate": 2.3745432792719715e-06, "loss": 0.01, "step": 50900 }, { "epoch": 15.66, "learning_rate": 2.3742209407111106e-06, "loss": 0.0088, "step": 50901 }, { "epoch": 15.66, "learning_rate": 2.3738986210829997e-06, "loss": 0.0081, "step": 50902 }, { "epoch": 15.66, "learning_rate": 2.373576320388442e-06, "loss": 0.0094, "step": 50903 }, { "epoch": 15.66, "learning_rate": 2.3732540386282356e-06, "loss": 0.0107, "step": 50904 }, { "epoch": 15.66, "learning_rate": 2.3729317758031866e-06, "loss": 0.0029, "step": 50905 }, { "epoch": 15.66, "learning_rate": 2.372609531914092e-06, "loss": 0.0098, "step": 50906 }, { "epoch": 15.66, "learning_rate": 2.3722873069617513e-06, "loss": 0.0055, "step": 50907 }, { "epoch": 15.66, "learning_rate": 2.3719651009469646e-06, "loss": 0.007, "step": 50908 }, { "epoch": 15.66, "learning_rate": 2.3716429138705335e-06, "loss": 0.0167, "step": 50909 }, { "epoch": 15.66, "learning_rate": 2.3713207457332542e-06, "loss": 0.0108, "step": 50910 }, { "epoch": 15.66, "learning_rate": 2.3709985965359307e-06, "loss": 0.009, "step": 50911 }, { "epoch": 15.67, "learning_rate": 2.3706764662793603e-06, "loss": 0.0096, "step": 50912 }, { "epoch": 15.67, "learning_rate": 2.3703543549643436e-06, "loss": 0.0071, "step": 50913 }, { "epoch": 15.67, "learning_rate": 2.3700322625916806e-06, "loss": 0.0059, "step": 50914 }, { "epoch": 15.67, "learning_rate": 2.36971018916217e-06, "loss": 0.0073, "step": 50915 }, { "epoch": 15.67, "learning_rate": 2.3693881346766113e-06, "loss": 0.0118, "step": 50916 }, { "epoch": 15.67, "learning_rate": 2.369066099135805e-06, "loss": 0.0075, "step": 50917 }, { "epoch": 15.67, "learning_rate": 2.3687440825405518e-06, "loss": 0.0056, "step": 50918 }, { "epoch": 15.67, "learning_rate": 2.368422084891645e-06, "loss": 0.0074, "step": 50919 }, { "epoch": 15.67, "learning_rate": 2.368100106189892e-06, "loss": 0.0126, "step": 50920 }, { "epoch": 15.67, "learning_rate": 2.3677781464360896e-06, "loss": 0.0096, "step": 50921 }, { "epoch": 15.67, "learning_rate": 2.367456205631039e-06, "loss": 0.0086, "step": 50922 }, { "epoch": 15.67, "learning_rate": 2.367134283775534e-06, "loss": 0.0088, "step": 50923 }, { "epoch": 15.67, "learning_rate": 2.3668123808703746e-06, "loss": 0.0065, "step": 50924 }, { "epoch": 15.67, "learning_rate": 2.366490496916364e-06, "loss": 0.0128, "step": 50925 }, { "epoch": 15.67, "learning_rate": 2.3661686319143006e-06, "loss": 0.0047, "step": 50926 }, { "epoch": 15.67, "learning_rate": 2.3658467858649812e-06, "loss": 0.0117, "step": 50927 }, { "epoch": 15.67, "learning_rate": 2.3655249587692073e-06, "loss": 0.0053, "step": 50928 }, { "epoch": 15.67, "learning_rate": 2.3652031506277764e-06, "loss": 0.01, "step": 50929 }, { "epoch": 15.67, "learning_rate": 2.364881361441488e-06, "loss": 0.0077, "step": 50930 }, { "epoch": 15.67, "learning_rate": 2.3645595912111406e-06, "loss": 0.0045, "step": 50931 }, { "epoch": 15.67, "learning_rate": 2.3642378399375332e-06, "loss": 0.0057, "step": 50932 }, { "epoch": 15.67, "learning_rate": 2.3639161076214647e-06, "loss": 0.0099, "step": 50933 }, { "epoch": 15.67, "learning_rate": 2.363594394263731e-06, "loss": 0.0072, "step": 50934 }, { "epoch": 15.67, "learning_rate": 2.36327269986514e-06, "loss": 0.0044, "step": 50935 }, { "epoch": 15.67, "learning_rate": 2.362951024426481e-06, "loss": 0.0089, "step": 50936 }, { "epoch": 15.67, "learning_rate": 2.3626293679485545e-06, "loss": 0.0052, "step": 50937 }, { "epoch": 15.67, "learning_rate": 2.362307730432162e-06, "loss": 0.0148, "step": 50938 }, { "epoch": 15.67, "learning_rate": 2.3619861118780963e-06, "loss": 0.0114, "step": 50939 }, { "epoch": 15.67, "learning_rate": 2.361664512287163e-06, "loss": 0.0082, "step": 50940 }, { "epoch": 15.67, "learning_rate": 2.3613429316601577e-06, "loss": 0.0072, "step": 50941 }, { "epoch": 15.67, "learning_rate": 2.361021369997879e-06, "loss": 0.0097, "step": 50942 }, { "epoch": 15.67, "learning_rate": 2.3606998273011274e-06, "loss": 0.0071, "step": 50943 }, { "epoch": 15.67, "learning_rate": 2.3603783035706927e-06, "loss": 0.0113, "step": 50944 }, { "epoch": 15.68, "learning_rate": 2.360056798807382e-06, "loss": 0.0082, "step": 50945 }, { "epoch": 15.68, "learning_rate": 2.3597353130119903e-06, "loss": 0.0112, "step": 50946 }, { "epoch": 15.68, "learning_rate": 2.3594138461853167e-06, "loss": 0.0044, "step": 50947 }, { "epoch": 15.68, "learning_rate": 2.359092398328159e-06, "loss": 0.0055, "step": 50948 }, { "epoch": 15.68, "learning_rate": 2.358770969441314e-06, "loss": 0.0064, "step": 50949 }, { "epoch": 15.68, "learning_rate": 2.3584495595255806e-06, "loss": 0.0112, "step": 50950 }, { "epoch": 15.68, "learning_rate": 2.358128168581758e-06, "loss": 0.0119, "step": 50951 }, { "epoch": 15.68, "learning_rate": 2.3578067966106423e-06, "loss": 0.0034, "step": 50952 }, { "epoch": 15.68, "learning_rate": 2.357485443613032e-06, "loss": 0.006, "step": 50953 }, { "epoch": 15.68, "learning_rate": 2.3571641095897223e-06, "loss": 0.0105, "step": 50954 }, { "epoch": 15.68, "learning_rate": 2.3568427945415163e-06, "loss": 0.0092, "step": 50955 }, { "epoch": 15.68, "learning_rate": 2.356521498469212e-06, "loss": 0.014, "step": 50956 }, { "epoch": 15.68, "learning_rate": 2.356200221373601e-06, "loss": 0.0124, "step": 50957 }, { "epoch": 15.68, "learning_rate": 2.3558789632554844e-06, "loss": 0.0056, "step": 50958 }, { "epoch": 15.68, "learning_rate": 2.3555577241156567e-06, "loss": 0.0149, "step": 50959 }, { "epoch": 15.68, "learning_rate": 2.355236503954922e-06, "loss": 0.0103, "step": 50960 }, { "epoch": 15.68, "learning_rate": 2.3549153027740733e-06, "loss": 0.0047, "step": 50961 }, { "epoch": 15.68, "learning_rate": 2.3545941205739086e-06, "loss": 0.0083, "step": 50962 }, { "epoch": 15.68, "learning_rate": 2.354272957355226e-06, "loss": 0.0117, "step": 50963 }, { "epoch": 15.68, "learning_rate": 2.353951813118822e-06, "loss": 0.0034, "step": 50964 }, { "epoch": 15.68, "learning_rate": 2.3536306878654956e-06, "loss": 0.0114, "step": 50965 }, { "epoch": 15.68, "learning_rate": 2.3533095815960417e-06, "loss": 0.0135, "step": 50966 }, { "epoch": 15.68, "learning_rate": 2.352988494311259e-06, "loss": 0.0066, "step": 50967 }, { "epoch": 15.68, "learning_rate": 2.352667426011943e-06, "loss": 0.0104, "step": 50968 }, { "epoch": 15.68, "learning_rate": 2.352346376698894e-06, "loss": 0.0088, "step": 50969 }, { "epoch": 15.68, "learning_rate": 2.3520253463729058e-06, "loss": 0.0063, "step": 50970 }, { "epoch": 15.68, "learning_rate": 2.351704335034778e-06, "loss": 0.0092, "step": 50971 }, { "epoch": 15.68, "learning_rate": 2.351383342685306e-06, "loss": 0.0118, "step": 50972 }, { "epoch": 15.68, "learning_rate": 2.3510623693252865e-06, "loss": 0.006, "step": 50973 }, { "epoch": 15.68, "learning_rate": 2.3507414149555153e-06, "loss": 0.0116, "step": 50974 }, { "epoch": 15.68, "learning_rate": 2.350420479576793e-06, "loss": 0.0079, "step": 50975 }, { "epoch": 15.68, "learning_rate": 2.350099563189915e-06, "loss": 0.0079, "step": 50976 }, { "epoch": 15.69, "learning_rate": 2.349778665795679e-06, "loss": 0.0075, "step": 50977 }, { "epoch": 15.69, "learning_rate": 2.3494577873948787e-06, "loss": 0.0127, "step": 50978 }, { "epoch": 15.69, "learning_rate": 2.349136927988309e-06, "loss": 0.0108, "step": 50979 }, { "epoch": 15.69, "learning_rate": 2.3488160875767717e-06, "loss": 0.0132, "step": 50980 }, { "epoch": 15.69, "learning_rate": 2.348495266161063e-06, "loss": 0.0068, "step": 50981 }, { "epoch": 15.69, "learning_rate": 2.3481744637419768e-06, "loss": 0.0125, "step": 50982 }, { "epoch": 15.69, "learning_rate": 2.34785368032031e-06, "loss": 0.0091, "step": 50983 }, { "epoch": 15.69, "learning_rate": 2.3475329158968606e-06, "loss": 0.0096, "step": 50984 }, { "epoch": 15.69, "learning_rate": 2.3472121704724225e-06, "loss": 0.0078, "step": 50985 }, { "epoch": 15.69, "learning_rate": 2.3468914440477953e-06, "loss": 0.0155, "step": 50986 }, { "epoch": 15.69, "learning_rate": 2.3465707366237723e-06, "loss": 0.0101, "step": 50987 }, { "epoch": 15.69, "learning_rate": 2.3462500482011508e-06, "loss": 0.0059, "step": 50988 }, { "epoch": 15.69, "learning_rate": 2.3459293787807246e-06, "loss": 0.0041, "step": 50989 }, { "epoch": 15.69, "learning_rate": 2.345608728363298e-06, "loss": 0.0094, "step": 50990 }, { "epoch": 15.69, "learning_rate": 2.3452880969496584e-06, "loss": 0.0093, "step": 50991 }, { "epoch": 15.69, "learning_rate": 2.3449674845406046e-06, "loss": 0.0084, "step": 50992 }, { "epoch": 15.69, "learning_rate": 2.3446468911369334e-06, "loss": 0.0267, "step": 50993 }, { "epoch": 15.69, "learning_rate": 2.344326316739437e-06, "loss": 0.0069, "step": 50994 }, { "epoch": 15.69, "learning_rate": 2.3440057613489165e-06, "loss": 0.006, "step": 50995 }, { "epoch": 15.69, "learning_rate": 2.3436852249661667e-06, "loss": 0.0054, "step": 50996 }, { "epoch": 15.69, "learning_rate": 2.343364707591982e-06, "loss": 0.0036, "step": 50997 }, { "epoch": 15.69, "learning_rate": 2.343044209227162e-06, "loss": 0.007, "step": 50998 }, { "epoch": 15.69, "learning_rate": 2.3427237298724927e-06, "loss": 0.0086, "step": 50999 }, { "epoch": 15.69, "learning_rate": 2.342403269528779e-06, "loss": 0.0074, "step": 51000 }, { "epoch": 15.69, "learning_rate": 2.342082828196813e-06, "loss": 0.0056, "step": 51001 }, { "epoch": 15.69, "learning_rate": 2.3417624058773915e-06, "loss": 0.0102, "step": 51002 }, { "epoch": 15.69, "learning_rate": 2.341442002571309e-06, "loss": 0.0121, "step": 51003 }, { "epoch": 15.69, "learning_rate": 2.341121618279362e-06, "loss": 0.0087, "step": 51004 }, { "epoch": 15.69, "learning_rate": 2.340801253002345e-06, "loss": 0.0068, "step": 51005 }, { "epoch": 15.69, "learning_rate": 2.340480906741053e-06, "loss": 0.0058, "step": 51006 }, { "epoch": 15.69, "learning_rate": 2.3401605794962824e-06, "loss": 0.0099, "step": 51007 }, { "epoch": 15.69, "learning_rate": 2.3398402712688285e-06, "loss": 0.0072, "step": 51008 }, { "epoch": 15.69, "learning_rate": 2.3395199820594825e-06, "loss": 0.0093, "step": 51009 }, { "epoch": 15.7, "learning_rate": 2.339199711869047e-06, "loss": 0.0068, "step": 51010 }, { "epoch": 15.7, "learning_rate": 2.3388794606983145e-06, "loss": 0.0096, "step": 51011 }, { "epoch": 15.7, "learning_rate": 2.338559228548077e-06, "loss": 0.021, "step": 51012 }, { "epoch": 15.7, "learning_rate": 2.3382390154191327e-06, "loss": 0.007, "step": 51013 }, { "epoch": 15.7, "learning_rate": 2.3379188213122706e-06, "loss": 0.0111, "step": 51014 }, { "epoch": 15.7, "learning_rate": 2.337598646228294e-06, "loss": 0.0076, "step": 51015 }, { "epoch": 15.7, "learning_rate": 2.3372784901679947e-06, "loss": 0.0164, "step": 51016 }, { "epoch": 15.7, "learning_rate": 2.336958353132166e-06, "loss": 0.0058, "step": 51017 }, { "epoch": 15.7, "learning_rate": 2.3366382351216057e-06, "loss": 0.0222, "step": 51018 }, { "epoch": 15.7, "learning_rate": 2.3363181361371047e-06, "loss": 0.0085, "step": 51019 }, { "epoch": 15.7, "learning_rate": 2.3359980561794603e-06, "loss": 0.0063, "step": 51020 }, { "epoch": 15.7, "learning_rate": 2.3356779952494667e-06, "loss": 0.0094, "step": 51021 }, { "epoch": 15.7, "learning_rate": 2.3353579533479187e-06, "loss": 0.0049, "step": 51022 }, { "epoch": 15.7, "learning_rate": 2.335037930475611e-06, "loss": 0.0077, "step": 51023 }, { "epoch": 15.7, "learning_rate": 2.334717926633334e-06, "loss": 0.0099, "step": 51024 }, { "epoch": 15.7, "learning_rate": 2.3343979418218923e-06, "loss": 0.0136, "step": 51025 }, { "epoch": 15.7, "learning_rate": 2.3340779760420695e-06, "loss": 0.0067, "step": 51026 }, { "epoch": 15.7, "learning_rate": 2.333758029294666e-06, "loss": 0.0066, "step": 51027 }, { "epoch": 15.7, "learning_rate": 2.3334381015804742e-06, "loss": 0.0067, "step": 51028 }, { "epoch": 15.7, "learning_rate": 2.3331181929002856e-06, "loss": 0.0024, "step": 51029 }, { "epoch": 15.7, "learning_rate": 2.332798303254901e-06, "loss": 0.0049, "step": 51030 }, { "epoch": 15.7, "learning_rate": 2.33247843264511e-06, "loss": 0.0068, "step": 51031 }, { "epoch": 15.7, "learning_rate": 2.332158581071712e-06, "loss": 0.0122, "step": 51032 }, { "epoch": 15.7, "learning_rate": 2.331838748535491e-06, "loss": 0.0065, "step": 51033 }, { "epoch": 15.7, "learning_rate": 2.33151893503725e-06, "loss": 0.008, "step": 51034 }, { "epoch": 15.7, "learning_rate": 2.33119914057778e-06, "loss": 0.0082, "step": 51035 }, { "epoch": 15.7, "learning_rate": 2.3308793651578755e-06, "loss": 0.008, "step": 51036 }, { "epoch": 15.7, "learning_rate": 2.330559608778329e-06, "loss": 0.0139, "step": 51037 }, { "epoch": 15.7, "learning_rate": 2.3302398714399367e-06, "loss": 0.0095, "step": 51038 }, { "epoch": 15.7, "learning_rate": 2.32992015314349e-06, "loss": 0.0043, "step": 51039 }, { "epoch": 15.7, "learning_rate": 2.3296004538897843e-06, "loss": 0.0046, "step": 51040 }, { "epoch": 15.7, "learning_rate": 2.329280773679613e-06, "loss": 0.0054, "step": 51041 }, { "epoch": 15.71, "learning_rate": 2.328961112513769e-06, "loss": 0.0097, "step": 51042 }, { "epoch": 15.71, "learning_rate": 2.328641470393044e-06, "loss": 0.0074, "step": 51043 }, { "epoch": 15.71, "learning_rate": 2.3283218473182366e-06, "loss": 0.0075, "step": 51044 }, { "epoch": 15.71, "learning_rate": 2.328002243290138e-06, "loss": 0.0066, "step": 51045 }, { "epoch": 15.71, "learning_rate": 2.3276826583095445e-06, "loss": 0.0104, "step": 51046 }, { "epoch": 15.71, "learning_rate": 2.327363092377244e-06, "loss": 0.0092, "step": 51047 }, { "epoch": 15.71, "learning_rate": 2.32704354549403e-06, "loss": 0.0084, "step": 51048 }, { "epoch": 15.71, "learning_rate": 2.3267240176607e-06, "loss": 0.0031, "step": 51049 }, { "epoch": 15.71, "learning_rate": 2.3264045088780463e-06, "loss": 0.0044, "step": 51050 }, { "epoch": 15.71, "learning_rate": 2.3260850191468607e-06, "loss": 0.0105, "step": 51051 }, { "epoch": 15.71, "learning_rate": 2.3257655484679376e-06, "loss": 0.0094, "step": 51052 }, { "epoch": 15.71, "learning_rate": 2.32544609684207e-06, "loss": 0.0099, "step": 51053 }, { "epoch": 15.71, "learning_rate": 2.3251266642700508e-06, "loss": 0.0078, "step": 51054 }, { "epoch": 15.71, "learning_rate": 2.324807250752673e-06, "loss": 0.0057, "step": 51055 }, { "epoch": 15.71, "learning_rate": 2.324487856290729e-06, "loss": 0.014, "step": 51056 }, { "epoch": 15.71, "learning_rate": 2.3241684808850128e-06, "loss": 0.0037, "step": 51057 }, { "epoch": 15.71, "learning_rate": 2.323849124536315e-06, "loss": 0.0064, "step": 51058 }, { "epoch": 15.71, "learning_rate": 2.323529787245434e-06, "loss": 0.0096, "step": 51059 }, { "epoch": 15.71, "learning_rate": 2.3232104690131585e-06, "loss": 0.0096, "step": 51060 }, { "epoch": 15.71, "learning_rate": 2.3228911698402803e-06, "loss": 0.0055, "step": 51061 }, { "epoch": 15.71, "learning_rate": 2.322571889727594e-06, "loss": 0.0139, "step": 51062 }, { "epoch": 15.71, "learning_rate": 2.3222526286758894e-06, "loss": 0.0055, "step": 51063 }, { "epoch": 15.71, "learning_rate": 2.321933386685964e-06, "loss": 0.0092, "step": 51064 }, { "epoch": 15.71, "learning_rate": 2.321614163758609e-06, "loss": 0.0055, "step": 51065 }, { "epoch": 15.71, "learning_rate": 2.3212949598946143e-06, "loss": 0.0072, "step": 51066 }, { "epoch": 15.71, "learning_rate": 2.320975775094778e-06, "loss": 0.004, "step": 51067 }, { "epoch": 15.71, "learning_rate": 2.3206566093598824e-06, "loss": 0.0082, "step": 51068 }, { "epoch": 15.71, "learning_rate": 2.3203374626907306e-06, "loss": 0.0084, "step": 51069 }, { "epoch": 15.71, "learning_rate": 2.3200183350881087e-06, "loss": 0.0064, "step": 51070 }, { "epoch": 15.71, "learning_rate": 2.3196992265528118e-06, "loss": 0.0058, "step": 51071 }, { "epoch": 15.71, "learning_rate": 2.3193801370856304e-06, "loss": 0.0097, "step": 51072 }, { "epoch": 15.71, "learning_rate": 2.319061066687358e-06, "loss": 0.012, "step": 51073 }, { "epoch": 15.71, "learning_rate": 2.318742015358787e-06, "loss": 0.0064, "step": 51074 }, { "epoch": 15.72, "learning_rate": 2.3184229831007077e-06, "loss": 0.0066, "step": 51075 }, { "epoch": 15.72, "learning_rate": 2.318103969913914e-06, "loss": 0.0059, "step": 51076 }, { "epoch": 15.72, "learning_rate": 2.3177849757991966e-06, "loss": 0.0032, "step": 51077 }, { "epoch": 15.72, "learning_rate": 2.3174660007573446e-06, "loss": 0.0086, "step": 51078 }, { "epoch": 15.72, "learning_rate": 2.3171470447891575e-06, "loss": 0.0064, "step": 51079 }, { "epoch": 15.72, "learning_rate": 2.3168281078954256e-06, "loss": 0.0074, "step": 51080 }, { "epoch": 15.72, "learning_rate": 2.316509190076934e-06, "loss": 0.0164, "step": 51081 }, { "epoch": 15.72, "learning_rate": 2.31619029133448e-06, "loss": 0.0092, "step": 51082 }, { "epoch": 15.72, "learning_rate": 2.315871411668852e-06, "loss": 0.0118, "step": 51083 }, { "epoch": 15.72, "learning_rate": 2.3155525510808453e-06, "loss": 0.0149, "step": 51084 }, { "epoch": 15.72, "learning_rate": 2.3152337095712497e-06, "loss": 0.0097, "step": 51085 }, { "epoch": 15.72, "learning_rate": 2.3149148871408577e-06, "loss": 0.0088, "step": 51086 }, { "epoch": 15.72, "learning_rate": 2.31459608379046e-06, "loss": 0.0189, "step": 51087 }, { "epoch": 15.72, "learning_rate": 2.3142772995208485e-06, "loss": 0.0055, "step": 51088 }, { "epoch": 15.72, "learning_rate": 2.313958534332814e-06, "loss": 0.0079, "step": 51089 }, { "epoch": 15.72, "learning_rate": 2.313639788227149e-06, "loss": 0.0073, "step": 51090 }, { "epoch": 15.72, "learning_rate": 2.313321061204643e-06, "loss": 0.0055, "step": 51091 }, { "epoch": 15.72, "learning_rate": 2.3130023532660896e-06, "loss": 0.0079, "step": 51092 }, { "epoch": 15.72, "learning_rate": 2.3126836644122784e-06, "loss": 0.0064, "step": 51093 }, { "epoch": 15.72, "learning_rate": 2.312364994644002e-06, "loss": 0.0057, "step": 51094 }, { "epoch": 15.72, "learning_rate": 2.31204634396205e-06, "loss": 0.0098, "step": 51095 }, { "epoch": 15.72, "learning_rate": 2.3117277123672154e-06, "loss": 0.0078, "step": 51096 }, { "epoch": 15.72, "learning_rate": 2.311409099860288e-06, "loss": 0.0049, "step": 51097 }, { "epoch": 15.72, "learning_rate": 2.3110905064420553e-06, "loss": 0.007, "step": 51098 }, { "epoch": 15.72, "learning_rate": 2.310771932113315e-06, "loss": 0.0077, "step": 51099 }, { "epoch": 15.72, "learning_rate": 2.310453376874855e-06, "loss": 0.0065, "step": 51100 }, { "epoch": 15.72, "learning_rate": 2.3101348407274695e-06, "loss": 0.008, "step": 51101 }, { "epoch": 15.72, "learning_rate": 2.3098163236719427e-06, "loss": 0.0079, "step": 51102 }, { "epoch": 15.72, "learning_rate": 2.309497825709067e-06, "loss": 0.0083, "step": 51103 }, { "epoch": 15.72, "learning_rate": 2.309179346839637e-06, "loss": 0.0079, "step": 51104 }, { "epoch": 15.72, "learning_rate": 2.308860887064441e-06, "loss": 0.0073, "step": 51105 }, { "epoch": 15.72, "learning_rate": 2.308542446384271e-06, "loss": 0.0092, "step": 51106 }, { "epoch": 15.73, "learning_rate": 2.3082240247999156e-06, "loss": 0.012, "step": 51107 }, { "epoch": 15.73, "learning_rate": 2.3079056223121666e-06, "loss": 0.0199, "step": 51108 }, { "epoch": 15.73, "learning_rate": 2.3075872389218135e-06, "loss": 0.0073, "step": 51109 }, { "epoch": 15.73, "learning_rate": 2.307268874629649e-06, "loss": 0.0064, "step": 51110 }, { "epoch": 15.73, "learning_rate": 2.30695052943646e-06, "loss": 0.0117, "step": 51111 }, { "epoch": 15.73, "learning_rate": 2.3066322033430412e-06, "loss": 0.0041, "step": 51112 }, { "epoch": 15.73, "learning_rate": 2.3063138963501764e-06, "loss": 0.0104, "step": 51113 }, { "epoch": 15.73, "learning_rate": 2.3059956084586656e-06, "loss": 0.0061, "step": 51114 }, { "epoch": 15.73, "learning_rate": 2.305677339669291e-06, "loss": 0.0043, "step": 51115 }, { "epoch": 15.73, "learning_rate": 2.3053590899828446e-06, "loss": 0.0057, "step": 51116 }, { "epoch": 15.73, "learning_rate": 2.3050408594001185e-06, "loss": 0.01, "step": 51117 }, { "epoch": 15.73, "learning_rate": 2.3047226479218978e-06, "loss": 0.0077, "step": 51118 }, { "epoch": 15.73, "learning_rate": 2.3044044555489787e-06, "loss": 0.0221, "step": 51119 }, { "epoch": 15.73, "learning_rate": 2.3040862822821485e-06, "loss": 0.0097, "step": 51120 }, { "epoch": 15.73, "learning_rate": 2.303768128122198e-06, "loss": 0.0077, "step": 51121 }, { "epoch": 15.73, "learning_rate": 2.3034499930699194e-06, "loss": 0.0051, "step": 51122 }, { "epoch": 15.73, "learning_rate": 2.3031318771260935e-06, "loss": 0.0099, "step": 51123 }, { "epoch": 15.73, "learning_rate": 2.3028137802915185e-06, "loss": 0.0068, "step": 51124 }, { "epoch": 15.73, "learning_rate": 2.3024957025669825e-06, "loss": 0.0227, "step": 51125 }, { "epoch": 15.73, "learning_rate": 2.302177643953274e-06, "loss": 0.0091, "step": 51126 }, { "epoch": 15.73, "learning_rate": 2.301859604451183e-06, "loss": 0.0066, "step": 51127 }, { "epoch": 15.73, "learning_rate": 2.3015415840614997e-06, "loss": 0.0118, "step": 51128 }, { "epoch": 15.73, "learning_rate": 2.301223582785014e-06, "loss": 0.0072, "step": 51129 }, { "epoch": 15.73, "learning_rate": 2.3009056006225127e-06, "loss": 0.0121, "step": 51130 }, { "epoch": 15.73, "learning_rate": 2.300587637574788e-06, "loss": 0.0064, "step": 51131 }, { "epoch": 15.73, "learning_rate": 2.3002696936426294e-06, "loss": 0.0049, "step": 51132 }, { "epoch": 15.73, "learning_rate": 2.2999517688268223e-06, "loss": 0.007, "step": 51133 }, { "epoch": 15.73, "learning_rate": 2.2996338631281613e-06, "loss": 0.0077, "step": 51134 }, { "epoch": 15.73, "learning_rate": 2.299315976547436e-06, "loss": 0.0061, "step": 51135 }, { "epoch": 15.73, "learning_rate": 2.2989981090854306e-06, "loss": 0.0114, "step": 51136 }, { "epoch": 15.73, "learning_rate": 2.298680260742937e-06, "loss": 0.007, "step": 51137 }, { "epoch": 15.73, "learning_rate": 2.298362431520742e-06, "loss": 0.0134, "step": 51138 }, { "epoch": 15.73, "learning_rate": 2.298044621419638e-06, "loss": 0.0131, "step": 51139 }, { "epoch": 15.74, "learning_rate": 2.2977268304404145e-06, "loss": 0.0087, "step": 51140 }, { "epoch": 15.74, "learning_rate": 2.297409058583857e-06, "loss": 0.009, "step": 51141 }, { "epoch": 15.74, "learning_rate": 2.297091305850757e-06, "loss": 0.007, "step": 51142 }, { "epoch": 15.74, "learning_rate": 2.2967735722419037e-06, "loss": 0.0127, "step": 51143 }, { "epoch": 15.74, "learning_rate": 2.296455857758084e-06, "loss": 0.0106, "step": 51144 }, { "epoch": 15.74, "learning_rate": 2.2961381624000876e-06, "loss": 0.0098, "step": 51145 }, { "epoch": 15.74, "learning_rate": 2.295820486168703e-06, "loss": 0.0108, "step": 51146 }, { "epoch": 15.74, "learning_rate": 2.2955028290647195e-06, "loss": 0.008, "step": 51147 }, { "epoch": 15.74, "learning_rate": 2.2951851910889256e-06, "loss": 0.0064, "step": 51148 }, { "epoch": 15.74, "learning_rate": 2.2948675722421086e-06, "loss": 0.0112, "step": 51149 }, { "epoch": 15.74, "learning_rate": 2.29454997252506e-06, "loss": 0.0091, "step": 51150 }, { "epoch": 15.74, "learning_rate": 2.2942323919385645e-06, "loss": 0.008, "step": 51151 }, { "epoch": 15.74, "learning_rate": 2.2939148304834135e-06, "loss": 0.0112, "step": 51152 }, { "epoch": 15.74, "learning_rate": 2.293597288160392e-06, "loss": 0.0101, "step": 51153 }, { "epoch": 15.74, "learning_rate": 2.293279764970293e-06, "loss": 0.0156, "step": 51154 }, { "epoch": 15.74, "learning_rate": 2.2929622609139035e-06, "loss": 0.0062, "step": 51155 }, { "epoch": 15.74, "learning_rate": 2.2926447759920134e-06, "loss": 0.0071, "step": 51156 }, { "epoch": 15.74, "learning_rate": 2.2923273102054057e-06, "loss": 0.0097, "step": 51157 }, { "epoch": 15.74, "learning_rate": 2.292009863554868e-06, "loss": 0.0059, "step": 51158 }, { "epoch": 15.74, "learning_rate": 2.291692436041195e-06, "loss": 0.0066, "step": 51159 }, { "epoch": 15.74, "learning_rate": 2.2913750276651724e-06, "loss": 0.0137, "step": 51160 }, { "epoch": 15.74, "learning_rate": 2.2910576384275863e-06, "loss": 0.0066, "step": 51161 }, { "epoch": 15.74, "learning_rate": 2.2907402683292268e-06, "loss": 0.0055, "step": 51162 }, { "epoch": 15.74, "learning_rate": 2.2904229173708803e-06, "loss": 0.0023, "step": 51163 }, { "epoch": 15.74, "learning_rate": 2.290105585553335e-06, "loss": 0.0039, "step": 51164 }, { "epoch": 15.74, "learning_rate": 2.289788272877379e-06, "loss": 0.0067, "step": 51165 }, { "epoch": 15.74, "learning_rate": 2.2894709793438006e-06, "loss": 0.0075, "step": 51166 }, { "epoch": 15.74, "learning_rate": 2.2891537049533862e-06, "loss": 0.0078, "step": 51167 }, { "epoch": 15.74, "learning_rate": 2.2888364497069227e-06, "loss": 0.0069, "step": 51168 }, { "epoch": 15.74, "learning_rate": 2.288519213605205e-06, "loss": 0.009, "step": 51169 }, { "epoch": 15.74, "learning_rate": 2.288201996649011e-06, "loss": 0.004, "step": 51170 }, { "epoch": 15.74, "learning_rate": 2.287884798839134e-06, "loss": 0.0072, "step": 51171 }, { "epoch": 15.75, "learning_rate": 2.2875676201763584e-06, "loss": 0.0088, "step": 51172 }, { "epoch": 15.75, "learning_rate": 2.287250460661471e-06, "loss": 0.0096, "step": 51173 }, { "epoch": 15.75, "learning_rate": 2.286933320295265e-06, "loss": 0.0071, "step": 51174 }, { "epoch": 15.75, "learning_rate": 2.2866161990785228e-06, "loss": 0.0066, "step": 51175 }, { "epoch": 15.75, "learning_rate": 2.2862990970120347e-06, "loss": 0.0064, "step": 51176 }, { "epoch": 15.75, "learning_rate": 2.2859820140965884e-06, "loss": 0.0131, "step": 51177 }, { "epoch": 15.75, "learning_rate": 2.2856649503329633e-06, "loss": 0.0087, "step": 51178 }, { "epoch": 15.75, "learning_rate": 2.2853479057219554e-06, "loss": 0.0217, "step": 51179 }, { "epoch": 15.75, "learning_rate": 2.2850308802643483e-06, "loss": 0.0072, "step": 51180 }, { "epoch": 15.75, "learning_rate": 2.284713873960931e-06, "loss": 0.0064, "step": 51181 }, { "epoch": 15.75, "learning_rate": 2.2843968868124876e-06, "loss": 0.0075, "step": 51182 }, { "epoch": 15.75, "learning_rate": 2.284079918819808e-06, "loss": 0.0062, "step": 51183 }, { "epoch": 15.75, "learning_rate": 2.283762969983677e-06, "loss": 0.0071, "step": 51184 }, { "epoch": 15.75, "learning_rate": 2.283446040304883e-06, "loss": 0.0146, "step": 51185 }, { "epoch": 15.75, "learning_rate": 2.283129129784212e-06, "loss": 0.0064, "step": 51186 }, { "epoch": 15.75, "learning_rate": 2.2828122384224506e-06, "loss": 0.0049, "step": 51187 }, { "epoch": 15.75, "learning_rate": 2.2824953662203832e-06, "loss": 0.0084, "step": 51188 }, { "epoch": 15.75, "learning_rate": 2.282178513178801e-06, "loss": 0.0128, "step": 51189 }, { "epoch": 15.75, "learning_rate": 2.281861679298493e-06, "loss": 0.0124, "step": 51190 }, { "epoch": 15.75, "learning_rate": 2.2815448645802385e-06, "loss": 0.0086, "step": 51191 }, { "epoch": 15.75, "learning_rate": 2.2812280690248277e-06, "loss": 0.0155, "step": 51192 }, { "epoch": 15.75, "learning_rate": 2.2809112926330436e-06, "loss": 0.0085, "step": 51193 }, { "epoch": 15.75, "learning_rate": 2.280594535405678e-06, "loss": 0.0073, "step": 51194 }, { "epoch": 15.75, "learning_rate": 2.2802777973435166e-06, "loss": 0.0091, "step": 51195 }, { "epoch": 15.75, "learning_rate": 2.2799610784473425e-06, "loss": 0.0082, "step": 51196 }, { "epoch": 15.75, "learning_rate": 2.2796443787179447e-06, "loss": 0.0181, "step": 51197 }, { "epoch": 15.75, "learning_rate": 2.2793276981561084e-06, "loss": 0.0105, "step": 51198 }, { "epoch": 15.75, "learning_rate": 2.2790110367626205e-06, "loss": 0.0044, "step": 51199 }, { "epoch": 15.75, "learning_rate": 2.2786943945382654e-06, "loss": 0.0175, "step": 51200 }, { "epoch": 15.75, "learning_rate": 2.2783777714838316e-06, "loss": 0.0098, "step": 51201 }, { "epoch": 15.75, "learning_rate": 2.278061167600104e-06, "loss": 0.0109, "step": 51202 }, { "epoch": 15.75, "learning_rate": 2.2777445828878662e-06, "loss": 0.0112, "step": 51203 }, { "epoch": 15.75, "learning_rate": 2.2774280173479115e-06, "loss": 0.0077, "step": 51204 }, { "epoch": 15.76, "learning_rate": 2.2771114709810184e-06, "loss": 0.0106, "step": 51205 }, { "epoch": 15.76, "learning_rate": 2.2767949437879765e-06, "loss": 0.0117, "step": 51206 }, { "epoch": 15.76, "learning_rate": 2.27647843576957e-06, "loss": 0.0038, "step": 51207 }, { "epoch": 15.76, "learning_rate": 2.2761619469265818e-06, "loss": 0.0065, "step": 51208 }, { "epoch": 15.76, "learning_rate": 2.275845477259805e-06, "loss": 0.0058, "step": 51209 }, { "epoch": 15.76, "learning_rate": 2.2755290267700212e-06, "loss": 0.0053, "step": 51210 }, { "epoch": 15.76, "learning_rate": 2.2752125954580194e-06, "loss": 0.0061, "step": 51211 }, { "epoch": 15.76, "learning_rate": 2.2748961833245786e-06, "loss": 0.0037, "step": 51212 }, { "epoch": 15.76, "learning_rate": 2.274579790370486e-06, "loss": 0.0066, "step": 51213 }, { "epoch": 15.76, "learning_rate": 2.2742634165965317e-06, "loss": 0.007, "step": 51214 }, { "epoch": 15.76, "learning_rate": 2.273947062003499e-06, "loss": 0.0099, "step": 51215 }, { "epoch": 15.76, "learning_rate": 2.273630726592172e-06, "loss": 0.0049, "step": 51216 }, { "epoch": 15.76, "learning_rate": 2.2733144103633376e-06, "loss": 0.0107, "step": 51217 }, { "epoch": 15.76, "learning_rate": 2.272998113317779e-06, "loss": 0.0044, "step": 51218 }, { "epoch": 15.76, "learning_rate": 2.2726818354562854e-06, "loss": 0.0104, "step": 51219 }, { "epoch": 15.76, "learning_rate": 2.272365576779637e-06, "loss": 0.008, "step": 51220 }, { "epoch": 15.76, "learning_rate": 2.2720493372886232e-06, "loss": 0.0058, "step": 51221 }, { "epoch": 15.76, "learning_rate": 2.2717331169840275e-06, "loss": 0.0222, "step": 51222 }, { "epoch": 15.76, "learning_rate": 2.271416915866631e-06, "loss": 0.0127, "step": 51223 }, { "epoch": 15.76, "learning_rate": 2.271100733937227e-06, "loss": 0.009, "step": 51224 }, { "epoch": 15.76, "learning_rate": 2.270784571196598e-06, "loss": 0.0082, "step": 51225 }, { "epoch": 15.76, "learning_rate": 2.2704684276455247e-06, "loss": 0.0096, "step": 51226 }, { "epoch": 15.76, "learning_rate": 2.270152303284795e-06, "loss": 0.0076, "step": 51227 }, { "epoch": 15.76, "learning_rate": 2.2698361981151894e-06, "loss": 0.0043, "step": 51228 }, { "epoch": 15.76, "learning_rate": 2.2695201121375e-06, "loss": 0.0059, "step": 51229 }, { "epoch": 15.76, "learning_rate": 2.269204045352509e-06, "loss": 0.01, "step": 51230 }, { "epoch": 15.76, "learning_rate": 2.268887997760999e-06, "loss": 0.0061, "step": 51231 }, { "epoch": 15.76, "learning_rate": 2.2685719693637588e-06, "loss": 0.0095, "step": 51232 }, { "epoch": 15.76, "learning_rate": 2.2682559601615638e-06, "loss": 0.011, "step": 51233 }, { "epoch": 15.76, "learning_rate": 2.2679399701552085e-06, "loss": 0.0092, "step": 51234 }, { "epoch": 15.76, "learning_rate": 2.267623999345474e-06, "loss": 0.0091, "step": 51235 }, { "epoch": 15.76, "learning_rate": 2.2673080477331442e-06, "loss": 0.0076, "step": 51236 }, { "epoch": 15.77, "learning_rate": 2.266992115319003e-06, "loss": 0.0128, "step": 51237 }, { "epoch": 15.77, "learning_rate": 2.266676202103837e-06, "loss": 0.0103, "step": 51238 }, { "epoch": 15.77, "learning_rate": 2.266360308088429e-06, "loss": 0.0086, "step": 51239 }, { "epoch": 15.77, "learning_rate": 2.266044433273562e-06, "loss": 0.0127, "step": 51240 }, { "epoch": 15.77, "learning_rate": 2.2657285776600237e-06, "loss": 0.0102, "step": 51241 }, { "epoch": 15.77, "learning_rate": 2.2654127412485947e-06, "loss": 0.0076, "step": 51242 }, { "epoch": 15.77, "learning_rate": 2.265096924040058e-06, "loss": 0.0097, "step": 51243 }, { "epoch": 15.77, "learning_rate": 2.264781126035204e-06, "loss": 0.006, "step": 51244 }, { "epoch": 15.77, "learning_rate": 2.264465347234812e-06, "loss": 0.0045, "step": 51245 }, { "epoch": 15.77, "learning_rate": 2.2641495876396713e-06, "loss": 0.0061, "step": 51246 }, { "epoch": 15.77, "learning_rate": 2.263833847250558e-06, "loss": 0.0121, "step": 51247 }, { "epoch": 15.77, "learning_rate": 2.263518126068257e-06, "loss": 0.0114, "step": 51248 }, { "epoch": 15.77, "learning_rate": 2.263202424093557e-06, "loss": 0.0166, "step": 51249 }, { "epoch": 15.77, "learning_rate": 2.2628867413272402e-06, "loss": 0.0112, "step": 51250 }, { "epoch": 15.77, "learning_rate": 2.2625710777700903e-06, "loss": 0.0085, "step": 51251 }, { "epoch": 15.77, "learning_rate": 2.2622554334228896e-06, "loss": 0.0178, "step": 51252 }, { "epoch": 15.77, "learning_rate": 2.2619398082864232e-06, "loss": 0.0071, "step": 51253 }, { "epoch": 15.77, "learning_rate": 2.261624202361473e-06, "loss": 0.0106, "step": 51254 }, { "epoch": 15.77, "learning_rate": 2.261308615648825e-06, "loss": 0.0073, "step": 51255 }, { "epoch": 15.77, "learning_rate": 2.2609930481492614e-06, "loss": 0.008, "step": 51256 }, { "epoch": 15.77, "learning_rate": 2.260677499863565e-06, "loss": 0.0074, "step": 51257 }, { "epoch": 15.77, "learning_rate": 2.2603619707925174e-06, "loss": 0.0162, "step": 51258 }, { "epoch": 15.77, "learning_rate": 2.26004646093691e-06, "loss": 0.0098, "step": 51259 }, { "epoch": 15.77, "learning_rate": 2.259730970297518e-06, "loss": 0.0077, "step": 51260 }, { "epoch": 15.77, "learning_rate": 2.259415498875127e-06, "loss": 0.0102, "step": 51261 }, { "epoch": 15.77, "learning_rate": 2.2591000466705215e-06, "loss": 0.0082, "step": 51262 }, { "epoch": 15.77, "learning_rate": 2.258784613684479e-06, "loss": 0.0084, "step": 51263 }, { "epoch": 15.77, "learning_rate": 2.258469199917792e-06, "loss": 0.0089, "step": 51264 }, { "epoch": 15.77, "learning_rate": 2.2581538053712383e-06, "loss": 0.012, "step": 51265 }, { "epoch": 15.77, "learning_rate": 2.2578384300456014e-06, "loss": 0.0074, "step": 51266 }, { "epoch": 15.77, "learning_rate": 2.2575230739416666e-06, "loss": 0.0045, "step": 51267 }, { "epoch": 15.77, "learning_rate": 2.25720773706021e-06, "loss": 0.0082, "step": 51268 }, { "epoch": 15.77, "learning_rate": 2.2568924194020214e-06, "loss": 0.0149, "step": 51269 }, { "epoch": 15.78, "learning_rate": 2.2565771209678822e-06, "loss": 0.0107, "step": 51270 }, { "epoch": 15.78, "learning_rate": 2.2562618417585746e-06, "loss": 0.0069, "step": 51271 }, { "epoch": 15.78, "learning_rate": 2.2559465817748803e-06, "loss": 0.0095, "step": 51272 }, { "epoch": 15.78, "learning_rate": 2.2556313410175833e-06, "loss": 0.006, "step": 51273 }, { "epoch": 15.78, "learning_rate": 2.255316119487466e-06, "loss": 0.015, "step": 51274 }, { "epoch": 15.78, "learning_rate": 2.2550009171853103e-06, "loss": 0.0057, "step": 51275 }, { "epoch": 15.78, "learning_rate": 2.2546857341119e-06, "loss": 0.0064, "step": 51276 }, { "epoch": 15.78, "learning_rate": 2.254370570268016e-06, "loss": 0.018, "step": 51277 }, { "epoch": 15.78, "learning_rate": 2.25405542565444e-06, "loss": 0.0079, "step": 51278 }, { "epoch": 15.78, "learning_rate": 2.253740300271959e-06, "loss": 0.005, "step": 51279 }, { "epoch": 15.78, "learning_rate": 2.253425194121355e-06, "loss": 0.0042, "step": 51280 }, { "epoch": 15.78, "learning_rate": 2.2531101072034044e-06, "loss": 0.0033, "step": 51281 }, { "epoch": 15.78, "learning_rate": 2.2527950395188934e-06, "loss": 0.0074, "step": 51282 }, { "epoch": 15.78, "learning_rate": 2.252479991068601e-06, "loss": 0.0144, "step": 51283 }, { "epoch": 15.78, "learning_rate": 2.2521649618533146e-06, "loss": 0.0101, "step": 51284 }, { "epoch": 15.78, "learning_rate": 2.2518499518738147e-06, "loss": 0.0045, "step": 51285 }, { "epoch": 15.78, "learning_rate": 2.251534961130881e-06, "loss": 0.0081, "step": 51286 }, { "epoch": 15.78, "learning_rate": 2.2512199896252985e-06, "loss": 0.003, "step": 51287 }, { "epoch": 15.78, "learning_rate": 2.250905037357848e-06, "loss": 0.0135, "step": 51288 }, { "epoch": 15.78, "learning_rate": 2.25059010432931e-06, "loss": 0.0078, "step": 51289 }, { "epoch": 15.78, "learning_rate": 2.2502751905404675e-06, "loss": 0.008, "step": 51290 }, { "epoch": 15.78, "learning_rate": 2.249960295992103e-06, "loss": 0.007, "step": 51291 }, { "epoch": 15.78, "learning_rate": 2.249645420684998e-06, "loss": 0.0075, "step": 51292 }, { "epoch": 15.78, "learning_rate": 2.249330564619934e-06, "loss": 0.0121, "step": 51293 }, { "epoch": 15.78, "learning_rate": 2.249015727797692e-06, "loss": 0.008, "step": 51294 }, { "epoch": 15.78, "learning_rate": 2.248700910219055e-06, "loss": 0.0091, "step": 51295 }, { "epoch": 15.78, "learning_rate": 2.248386111884804e-06, "loss": 0.0056, "step": 51296 }, { "epoch": 15.78, "learning_rate": 2.2480713327957204e-06, "loss": 0.0073, "step": 51297 }, { "epoch": 15.78, "learning_rate": 2.247756572952583e-06, "loss": 0.0076, "step": 51298 }, { "epoch": 15.78, "learning_rate": 2.2474418323561786e-06, "loss": 0.0142, "step": 51299 }, { "epoch": 15.78, "learning_rate": 2.2471271110072856e-06, "loss": 0.0067, "step": 51300 }, { "epoch": 15.78, "learning_rate": 2.2468124089066903e-06, "loss": 0.0028, "step": 51301 }, { "epoch": 15.79, "learning_rate": 2.2464977260551647e-06, "loss": 0.0054, "step": 51302 }, { "epoch": 15.79, "learning_rate": 2.246183062453494e-06, "loss": 0.0065, "step": 51303 }, { "epoch": 15.79, "learning_rate": 2.2458684181024626e-06, "loss": 0.0061, "step": 51304 }, { "epoch": 15.79, "learning_rate": 2.245553793002849e-06, "loss": 0.005, "step": 51305 }, { "epoch": 15.79, "learning_rate": 2.245239187155435e-06, "loss": 0.0047, "step": 51306 }, { "epoch": 15.79, "learning_rate": 2.2449246005610015e-06, "loss": 0.0103, "step": 51307 }, { "epoch": 15.79, "learning_rate": 2.24461003322033e-06, "loss": 0.0041, "step": 51308 }, { "epoch": 15.79, "learning_rate": 2.2442954851342004e-06, "loss": 0.013, "step": 51309 }, { "epoch": 15.79, "learning_rate": 2.2439809563033945e-06, "loss": 0.0098, "step": 51310 }, { "epoch": 15.79, "learning_rate": 2.243666446728693e-06, "loss": 0.0161, "step": 51311 }, { "epoch": 15.79, "learning_rate": 2.2433519564108764e-06, "loss": 0.0077, "step": 51312 }, { "epoch": 15.79, "learning_rate": 2.243037485350723e-06, "loss": 0.005, "step": 51313 }, { "epoch": 15.79, "learning_rate": 2.2427230335490223e-06, "loss": 0.006, "step": 51314 }, { "epoch": 15.79, "learning_rate": 2.242408601006546e-06, "loss": 0.0064, "step": 51315 }, { "epoch": 15.79, "learning_rate": 2.2420941877240775e-06, "loss": 0.0097, "step": 51316 }, { "epoch": 15.79, "learning_rate": 2.2417797937023976e-06, "loss": 0.0055, "step": 51317 }, { "epoch": 15.79, "learning_rate": 2.2414654189422845e-06, "loss": 0.0066, "step": 51318 }, { "epoch": 15.79, "learning_rate": 2.2411510634445244e-06, "loss": 0.0133, "step": 51319 }, { "epoch": 15.79, "learning_rate": 2.2408367272098928e-06, "loss": 0.0087, "step": 51320 }, { "epoch": 15.79, "learning_rate": 2.240522410239173e-06, "loss": 0.008, "step": 51321 }, { "epoch": 15.79, "learning_rate": 2.2402081125331475e-06, "loss": 0.0063, "step": 51322 }, { "epoch": 15.79, "learning_rate": 2.239893834092587e-06, "loss": 0.0064, "step": 51323 }, { "epoch": 15.79, "learning_rate": 2.2395795749182815e-06, "loss": 0.0143, "step": 51324 }, { "epoch": 15.79, "learning_rate": 2.2392653350110072e-06, "loss": 0.0103, "step": 51325 }, { "epoch": 15.79, "learning_rate": 2.2389511143715446e-06, "loss": 0.0054, "step": 51326 }, { "epoch": 15.79, "learning_rate": 2.2386369130006734e-06, "loss": 0.0177, "step": 51327 }, { "epoch": 15.79, "learning_rate": 2.238322730899176e-06, "loss": 0.0076, "step": 51328 }, { "epoch": 15.79, "learning_rate": 2.2380085680678297e-06, "loss": 0.007, "step": 51329 }, { "epoch": 15.79, "learning_rate": 2.2376944245074162e-06, "loss": 0.013, "step": 51330 }, { "epoch": 15.79, "learning_rate": 2.2373803002187146e-06, "loss": 0.0103, "step": 51331 }, { "epoch": 15.79, "learning_rate": 2.237066195202504e-06, "loss": 0.0076, "step": 51332 }, { "epoch": 15.79, "learning_rate": 2.2367521094595634e-06, "loss": 0.0039, "step": 51333 }, { "epoch": 15.79, "learning_rate": 2.2364380429906775e-06, "loss": 0.0129, "step": 51334 }, { "epoch": 15.8, "learning_rate": 2.236123995796625e-06, "loss": 0.0118, "step": 51335 }, { "epoch": 15.8, "learning_rate": 2.235809967878181e-06, "loss": 0.0067, "step": 51336 }, { "epoch": 15.8, "learning_rate": 2.235495959236128e-06, "loss": 0.0058, "step": 51337 }, { "epoch": 15.8, "learning_rate": 2.235181969871242e-06, "loss": 0.0035, "step": 51338 }, { "epoch": 15.8, "learning_rate": 2.2348679997843094e-06, "loss": 0.0037, "step": 51339 }, { "epoch": 15.8, "learning_rate": 2.2345540489761054e-06, "loss": 0.0061, "step": 51340 }, { "epoch": 15.8, "learning_rate": 2.2342401174474103e-06, "loss": 0.0098, "step": 51341 }, { "epoch": 15.8, "learning_rate": 2.233926205199004e-06, "loss": 0.0181, "step": 51342 }, { "epoch": 15.8, "learning_rate": 2.2336123122316642e-06, "loss": 0.0051, "step": 51343 }, { "epoch": 15.8, "learning_rate": 2.233298438546172e-06, "loss": 0.0021, "step": 51344 }, { "epoch": 15.8, "learning_rate": 2.232984584143306e-06, "loss": 0.0078, "step": 51345 }, { "epoch": 15.8, "learning_rate": 2.232670749023845e-06, "loss": 0.0083, "step": 51346 }, { "epoch": 15.8, "learning_rate": 2.2323569331885685e-06, "loss": 0.0045, "step": 51347 }, { "epoch": 15.8, "learning_rate": 2.2320431366382556e-06, "loss": 0.0061, "step": 51348 }, { "epoch": 15.8, "learning_rate": 2.2317293593736856e-06, "loss": 0.0043, "step": 51349 }, { "epoch": 15.8, "learning_rate": 2.2314156013956377e-06, "loss": 0.0172, "step": 51350 }, { "epoch": 15.8, "learning_rate": 2.2311018627048897e-06, "loss": 0.006, "step": 51351 }, { "epoch": 15.8, "learning_rate": 2.230788143302223e-06, "loss": 0.0112, "step": 51352 }, { "epoch": 15.8, "learning_rate": 2.230474443188411e-06, "loss": 0.0081, "step": 51353 }, { "epoch": 15.8, "learning_rate": 2.2301607623642386e-06, "loss": 0.0072, "step": 51354 }, { "epoch": 15.8, "learning_rate": 2.2298471008304823e-06, "loss": 0.0105, "step": 51355 }, { "epoch": 15.8, "learning_rate": 2.2295334585879235e-06, "loss": 0.0092, "step": 51356 }, { "epoch": 15.8, "learning_rate": 2.2292198356373362e-06, "loss": 0.0104, "step": 51357 }, { "epoch": 15.8, "learning_rate": 2.2289062319794985e-06, "loss": 0.0226, "step": 51358 }, { "epoch": 15.8, "learning_rate": 2.2285926476151944e-06, "loss": 0.0062, "step": 51359 }, { "epoch": 15.8, "learning_rate": 2.2282790825451996e-06, "loss": 0.0036, "step": 51360 }, { "epoch": 15.8, "learning_rate": 2.227965536770292e-06, "loss": 0.0091, "step": 51361 }, { "epoch": 15.8, "learning_rate": 2.2276520102912515e-06, "loss": 0.0083, "step": 51362 }, { "epoch": 15.8, "learning_rate": 2.2273385031088546e-06, "loss": 0.0055, "step": 51363 }, { "epoch": 15.8, "learning_rate": 2.2270250152238816e-06, "loss": 0.0072, "step": 51364 }, { "epoch": 15.8, "learning_rate": 2.22671154663711e-06, "loss": 0.0041, "step": 51365 }, { "epoch": 15.8, "learning_rate": 2.2263980973493183e-06, "loss": 0.0044, "step": 51366 }, { "epoch": 15.81, "learning_rate": 2.2260846673612834e-06, "loss": 0.0095, "step": 51367 }, { "epoch": 15.81, "learning_rate": 2.225771256673782e-06, "loss": 0.014, "step": 51368 }, { "epoch": 15.81, "learning_rate": 2.225457865287601e-06, "loss": 0.0059, "step": 51369 }, { "epoch": 15.81, "learning_rate": 2.2251444932035094e-06, "loss": 0.0108, "step": 51370 }, { "epoch": 15.81, "learning_rate": 2.2248311404222867e-06, "loss": 0.0074, "step": 51371 }, { "epoch": 15.81, "learning_rate": 2.2245178069447137e-06, "loss": 0.0054, "step": 51372 }, { "epoch": 15.81, "learning_rate": 2.2242044927715623e-06, "loss": 0.0101, "step": 51373 }, { "epoch": 15.81, "learning_rate": 2.223891197903619e-06, "loss": 0.0061, "step": 51374 }, { "epoch": 15.81, "learning_rate": 2.2235779223416575e-06, "loss": 0.0049, "step": 51375 }, { "epoch": 15.81, "learning_rate": 2.2232646660864544e-06, "loss": 0.0052, "step": 51376 }, { "epoch": 15.81, "learning_rate": 2.2229514291387923e-06, "loss": 0.0097, "step": 51377 }, { "epoch": 15.81, "learning_rate": 2.2226382114994393e-06, "loss": 0.0094, "step": 51378 }, { "epoch": 15.81, "learning_rate": 2.222325013169182e-06, "loss": 0.0048, "step": 51379 }, { "epoch": 15.81, "learning_rate": 2.222011834148795e-06, "loss": 0.0147, "step": 51380 }, { "epoch": 15.81, "learning_rate": 2.2216986744390556e-06, "loss": 0.0087, "step": 51381 }, { "epoch": 15.81, "learning_rate": 2.221385534040741e-06, "loss": 0.0106, "step": 51382 }, { "epoch": 15.81, "learning_rate": 2.221072412954629e-06, "loss": 0.0133, "step": 51383 }, { "epoch": 15.81, "learning_rate": 2.2207593111814974e-06, "loss": 0.0036, "step": 51384 }, { "epoch": 15.81, "learning_rate": 2.220446228722123e-06, "loss": 0.0073, "step": 51385 }, { "epoch": 15.81, "learning_rate": 2.220133165577284e-06, "loss": 0.0126, "step": 51386 }, { "epoch": 15.81, "learning_rate": 2.219820121747757e-06, "loss": 0.0108, "step": 51387 }, { "epoch": 15.81, "learning_rate": 2.219507097234316e-06, "loss": 0.012, "step": 51388 }, { "epoch": 15.81, "learning_rate": 2.219194092037743e-06, "loss": 0.0057, "step": 51389 }, { "epoch": 15.81, "learning_rate": 2.2188811061588177e-06, "loss": 0.0078, "step": 51390 }, { "epoch": 15.81, "learning_rate": 2.2185681395983093e-06, "loss": 0.0064, "step": 51391 }, { "epoch": 15.81, "learning_rate": 2.2182551923569982e-06, "loss": 0.0092, "step": 51392 }, { "epoch": 15.81, "learning_rate": 2.2179422644356595e-06, "loss": 0.0077, "step": 51393 }, { "epoch": 15.81, "learning_rate": 2.2176293558350746e-06, "loss": 0.0059, "step": 51394 }, { "epoch": 15.81, "learning_rate": 2.217316466556019e-06, "loss": 0.0079, "step": 51395 }, { "epoch": 15.81, "learning_rate": 2.2170035965992674e-06, "loss": 0.0094, "step": 51396 }, { "epoch": 15.81, "learning_rate": 2.2166907459655973e-06, "loss": 0.0039, "step": 51397 }, { "epoch": 15.81, "learning_rate": 2.2163779146557863e-06, "loss": 0.0076, "step": 51398 }, { "epoch": 15.81, "learning_rate": 2.2160651026706105e-06, "loss": 0.0093, "step": 51399 }, { "epoch": 15.82, "learning_rate": 2.2157523100108467e-06, "loss": 0.0042, "step": 51400 }, { "epoch": 15.82, "learning_rate": 2.215439536677271e-06, "loss": 0.0074, "step": 51401 }, { "epoch": 15.82, "learning_rate": 2.21512678267066e-06, "loss": 0.0054, "step": 51402 }, { "epoch": 15.82, "learning_rate": 2.214814047991789e-06, "loss": 0.0088, "step": 51403 }, { "epoch": 15.82, "learning_rate": 2.21450133264144e-06, "loss": 0.0081, "step": 51404 }, { "epoch": 15.82, "learning_rate": 2.2141886366203836e-06, "loss": 0.0107, "step": 51405 }, { "epoch": 15.82, "learning_rate": 2.213875959929397e-06, "loss": 0.0147, "step": 51406 }, { "epoch": 15.82, "learning_rate": 2.213563302569258e-06, "loss": 0.0084, "step": 51407 }, { "epoch": 15.82, "learning_rate": 2.2132506645407393e-06, "loss": 0.0069, "step": 51408 }, { "epoch": 15.82, "learning_rate": 2.2129380458446226e-06, "loss": 0.0295, "step": 51409 }, { "epoch": 15.82, "learning_rate": 2.212625446481681e-06, "loss": 0.0108, "step": 51410 }, { "epoch": 15.82, "learning_rate": 2.2123128664526936e-06, "loss": 0.0036, "step": 51411 }, { "epoch": 15.82, "learning_rate": 2.2120003057584315e-06, "loss": 0.0101, "step": 51412 }, { "epoch": 15.82, "learning_rate": 2.2116877643996704e-06, "loss": 0.0138, "step": 51413 }, { "epoch": 15.82, "learning_rate": 2.2113752423771917e-06, "loss": 0.0122, "step": 51414 }, { "epoch": 15.82, "learning_rate": 2.2110627396917684e-06, "loss": 0.0045, "step": 51415 }, { "epoch": 15.82, "learning_rate": 2.2107502563441762e-06, "loss": 0.0064, "step": 51416 }, { "epoch": 15.82, "learning_rate": 2.2104377923351916e-06, "loss": 0.0078, "step": 51417 }, { "epoch": 15.82, "learning_rate": 2.2101253476655904e-06, "loss": 0.006, "step": 51418 }, { "epoch": 15.82, "learning_rate": 2.2098129223361476e-06, "loss": 0.0088, "step": 51419 }, { "epoch": 15.82, "learning_rate": 2.209500516347639e-06, "loss": 0.0066, "step": 51420 }, { "epoch": 15.82, "learning_rate": 2.20918812970084e-06, "loss": 0.0131, "step": 51421 }, { "epoch": 15.82, "learning_rate": 2.2088757623965263e-06, "loss": 0.0069, "step": 51422 }, { "epoch": 15.82, "learning_rate": 2.208563414435472e-06, "loss": 0.0038, "step": 51423 }, { "epoch": 15.82, "learning_rate": 2.2082510858184558e-06, "loss": 0.0067, "step": 51424 }, { "epoch": 15.82, "learning_rate": 2.2079387765462555e-06, "loss": 0.0078, "step": 51425 }, { "epoch": 15.82, "learning_rate": 2.2076264866196393e-06, "loss": 0.0062, "step": 51426 }, { "epoch": 15.82, "learning_rate": 2.2073142160393823e-06, "loss": 0.0098, "step": 51427 }, { "epoch": 15.82, "learning_rate": 2.2070019648062667e-06, "loss": 0.0067, "step": 51428 }, { "epoch": 15.82, "learning_rate": 2.2066897329210645e-06, "loss": 0.0125, "step": 51429 }, { "epoch": 15.82, "learning_rate": 2.2063775203845504e-06, "loss": 0.0078, "step": 51430 }, { "epoch": 15.82, "learning_rate": 2.2060653271974998e-06, "loss": 0.0082, "step": 51431 }, { "epoch": 15.83, "learning_rate": 2.205753153360688e-06, "loss": 0.0067, "step": 51432 }, { "epoch": 15.83, "learning_rate": 2.2054409988748893e-06, "loss": 0.013, "step": 51433 }, { "epoch": 15.83, "learning_rate": 2.2051288637408794e-06, "loss": 0.0051, "step": 51434 }, { "epoch": 15.83, "learning_rate": 2.204816747959434e-06, "loss": 0.0065, "step": 51435 }, { "epoch": 15.83, "learning_rate": 2.2045046515313263e-06, "loss": 0.0134, "step": 51436 }, { "epoch": 15.83, "learning_rate": 2.20419257445733e-06, "loss": 0.0233, "step": 51437 }, { "epoch": 15.83, "learning_rate": 2.203880516738226e-06, "loss": 0.0121, "step": 51438 }, { "epoch": 15.83, "learning_rate": 2.2035684783747835e-06, "loss": 0.0036, "step": 51439 }, { "epoch": 15.83, "learning_rate": 2.2032564593677773e-06, "loss": 0.0065, "step": 51440 }, { "epoch": 15.83, "learning_rate": 2.202944459717985e-06, "loss": 0.0062, "step": 51441 }, { "epoch": 15.83, "learning_rate": 2.2026324794261754e-06, "loss": 0.017, "step": 51442 }, { "epoch": 15.83, "learning_rate": 2.2023205184931305e-06, "loss": 0.0125, "step": 51443 }, { "epoch": 15.83, "learning_rate": 2.202008576919622e-06, "loss": 0.0071, "step": 51444 }, { "epoch": 15.83, "learning_rate": 2.2016966547064254e-06, "loss": 0.0072, "step": 51445 }, { "epoch": 15.83, "learning_rate": 2.201384751854312e-06, "loss": 0.0071, "step": 51446 }, { "epoch": 15.83, "learning_rate": 2.2010728683640546e-06, "loss": 0.0112, "step": 51447 }, { "epoch": 15.83, "learning_rate": 2.2007610042364337e-06, "loss": 0.0051, "step": 51448 }, { "epoch": 15.83, "learning_rate": 2.200449159472221e-06, "loss": 0.0042, "step": 51449 }, { "epoch": 15.83, "learning_rate": 2.2001373340721897e-06, "loss": 0.0086, "step": 51450 }, { "epoch": 15.83, "learning_rate": 2.199825528037115e-06, "loss": 0.0056, "step": 51451 }, { "epoch": 15.83, "learning_rate": 2.1995137413677703e-06, "loss": 0.0046, "step": 51452 }, { "epoch": 15.83, "learning_rate": 2.199201974064931e-06, "loss": 0.0084, "step": 51453 }, { "epoch": 15.83, "learning_rate": 2.1988902261293688e-06, "loss": 0.0094, "step": 51454 }, { "epoch": 15.83, "learning_rate": 2.1985784975618596e-06, "loss": 0.0122, "step": 51455 }, { "epoch": 15.83, "learning_rate": 2.198266788363177e-06, "loss": 0.0077, "step": 51456 }, { "epoch": 15.83, "learning_rate": 2.1979550985340915e-06, "loss": 0.0049, "step": 51457 }, { "epoch": 15.83, "learning_rate": 2.1976434280753833e-06, "loss": 0.0115, "step": 51458 }, { "epoch": 15.83, "learning_rate": 2.1973317769878257e-06, "loss": 0.0074, "step": 51459 }, { "epoch": 15.83, "learning_rate": 2.1970201452721873e-06, "loss": 0.017, "step": 51460 }, { "epoch": 15.83, "learning_rate": 2.1967085329292435e-06, "loss": 0.0078, "step": 51461 }, { "epoch": 15.83, "learning_rate": 2.196396939959766e-06, "loss": 0.0071, "step": 51462 }, { "epoch": 15.83, "learning_rate": 2.1960853663645355e-06, "loss": 0.0039, "step": 51463 }, { "epoch": 15.83, "learning_rate": 2.1957738121443197e-06, "loss": 0.0076, "step": 51464 }, { "epoch": 15.84, "learning_rate": 2.1954622772998935e-06, "loss": 0.0059, "step": 51465 }, { "epoch": 15.84, "learning_rate": 2.1951507618320334e-06, "loss": 0.0105, "step": 51466 }, { "epoch": 15.84, "learning_rate": 2.1948392657415042e-06, "loss": 0.0069, "step": 51467 }, { "epoch": 15.84, "learning_rate": 2.194527789029088e-06, "loss": 0.013, "step": 51468 }, { "epoch": 15.84, "learning_rate": 2.1942163316955544e-06, "loss": 0.0143, "step": 51469 }, { "epoch": 15.84, "learning_rate": 2.1939048937416783e-06, "loss": 0.0088, "step": 51470 }, { "epoch": 15.84, "learning_rate": 2.1935934751682307e-06, "loss": 0.0048, "step": 51471 }, { "epoch": 15.84, "learning_rate": 2.1932820759759875e-06, "loss": 0.0055, "step": 51472 }, { "epoch": 15.84, "learning_rate": 2.1929706961657195e-06, "loss": 0.0081, "step": 51473 }, { "epoch": 15.84, "learning_rate": 2.1926593357382e-06, "loss": 0.0085, "step": 51474 }, { "epoch": 15.84, "learning_rate": 2.1923479946942028e-06, "loss": 0.0079, "step": 51475 }, { "epoch": 15.84, "learning_rate": 2.1920366730345012e-06, "loss": 0.0121, "step": 51476 }, { "epoch": 15.84, "learning_rate": 2.191725370759864e-06, "loss": 0.005, "step": 51477 }, { "epoch": 15.84, "learning_rate": 2.191414087871071e-06, "loss": 0.0048, "step": 51478 }, { "epoch": 15.84, "learning_rate": 2.1911028243688915e-06, "loss": 0.0071, "step": 51479 }, { "epoch": 15.84, "learning_rate": 2.190791580254101e-06, "loss": 0.0146, "step": 51480 }, { "epoch": 15.84, "learning_rate": 2.1904803555274678e-06, "loss": 0.0086, "step": 51481 }, { "epoch": 15.84, "learning_rate": 2.190169150189763e-06, "loss": 0.0094, "step": 51482 }, { "epoch": 15.84, "learning_rate": 2.189857964241766e-06, "loss": 0.0054, "step": 51483 }, { "epoch": 15.84, "learning_rate": 2.1895467976842457e-06, "loss": 0.0053, "step": 51484 }, { "epoch": 15.84, "learning_rate": 2.1892356505179747e-06, "loss": 0.01, "step": 51485 }, { "epoch": 15.84, "learning_rate": 2.188924522743726e-06, "loss": 0.0083, "step": 51486 }, { "epoch": 15.84, "learning_rate": 2.188613414362273e-06, "loss": 0.0101, "step": 51487 }, { "epoch": 15.84, "learning_rate": 2.188302325374386e-06, "loss": 0.0179, "step": 51488 }, { "epoch": 15.84, "learning_rate": 2.1879912557808392e-06, "loss": 0.0103, "step": 51489 }, { "epoch": 15.84, "learning_rate": 2.1876802055824032e-06, "loss": 0.0129, "step": 51490 }, { "epoch": 15.84, "learning_rate": 2.1873691747798507e-06, "loss": 0.0065, "step": 51491 }, { "epoch": 15.84, "learning_rate": 2.187058163373953e-06, "loss": 0.0054, "step": 51492 }, { "epoch": 15.84, "learning_rate": 2.1867471713654874e-06, "loss": 0.0134, "step": 51493 }, { "epoch": 15.84, "learning_rate": 2.1864361987552207e-06, "loss": 0.0069, "step": 51494 }, { "epoch": 15.84, "learning_rate": 2.1861252455439266e-06, "loss": 0.0057, "step": 51495 }, { "epoch": 15.84, "learning_rate": 2.185814311732376e-06, "loss": 0.007, "step": 51496 }, { "epoch": 15.85, "learning_rate": 2.18550339732134e-06, "loss": 0.0087, "step": 51497 }, { "epoch": 15.85, "learning_rate": 2.185192502311595e-06, "loss": 0.0053, "step": 51498 }, { "epoch": 15.85, "learning_rate": 2.18488162670391e-06, "loss": 0.0066, "step": 51499 }, { "epoch": 15.85, "learning_rate": 2.184570770499056e-06, "loss": 0.0075, "step": 51500 }, { "epoch": 15.85, "learning_rate": 2.1842599336978097e-06, "loss": 0.0053, "step": 51501 }, { "epoch": 15.85, "learning_rate": 2.183949116300933e-06, "loss": 0.0092, "step": 51502 }, { "epoch": 15.85, "learning_rate": 2.183638318309207e-06, "loss": 0.0093, "step": 51503 }, { "epoch": 15.85, "learning_rate": 2.1833275397233987e-06, "loss": 0.0046, "step": 51504 }, { "epoch": 15.85, "learning_rate": 2.1830167805442813e-06, "loss": 0.009, "step": 51505 }, { "epoch": 15.85, "learning_rate": 2.1827060407726264e-06, "loss": 0.007, "step": 51506 }, { "epoch": 15.85, "learning_rate": 2.182395320409204e-06, "loss": 0.0148, "step": 51507 }, { "epoch": 15.85, "learning_rate": 2.1820846194547864e-06, "loss": 0.0035, "step": 51508 }, { "epoch": 15.85, "learning_rate": 2.1817739379101456e-06, "loss": 0.0078, "step": 51509 }, { "epoch": 15.85, "learning_rate": 2.1814632757760525e-06, "loss": 0.0068, "step": 51510 }, { "epoch": 15.85, "learning_rate": 2.1811526330532784e-06, "loss": 0.0114, "step": 51511 }, { "epoch": 15.85, "learning_rate": 2.180842009742591e-06, "loss": 0.0066, "step": 51512 }, { "epoch": 15.85, "learning_rate": 2.180531405844768e-06, "loss": 0.006, "step": 51513 }, { "epoch": 15.85, "learning_rate": 2.18022082136058e-06, "loss": 0.0168, "step": 51514 }, { "epoch": 15.85, "learning_rate": 2.1799102562907936e-06, "loss": 0.0095, "step": 51515 }, { "epoch": 15.85, "learning_rate": 2.179599710636181e-06, "loss": 0.0052, "step": 51516 }, { "epoch": 15.85, "learning_rate": 2.1792891843975115e-06, "loss": 0.005, "step": 51517 }, { "epoch": 15.85, "learning_rate": 2.1789786775755604e-06, "loss": 0.0046, "step": 51518 }, { "epoch": 15.85, "learning_rate": 2.178668190171098e-06, "loss": 0.0052, "step": 51519 }, { "epoch": 15.85, "learning_rate": 2.178357722184894e-06, "loss": 0.0044, "step": 51520 }, { "epoch": 15.85, "learning_rate": 2.1780472736177183e-06, "loss": 0.0047, "step": 51521 }, { "epoch": 15.85, "learning_rate": 2.1777368444703427e-06, "loss": 0.0081, "step": 51522 }, { "epoch": 15.85, "learning_rate": 2.1774264347435383e-06, "loss": 0.0084, "step": 51523 }, { "epoch": 15.85, "learning_rate": 2.1771160444380745e-06, "loss": 0.0061, "step": 51524 }, { "epoch": 15.85, "learning_rate": 2.1771160444380745e-06, "loss": 0.0174, "step": 51525 }, { "epoch": 15.85, "learning_rate": 2.176805673554723e-06, "loss": 0.0076, "step": 51526 }, { "epoch": 15.85, "learning_rate": 2.176495322094254e-06, "loss": 0.0023, "step": 51527 }, { "epoch": 15.85, "learning_rate": 2.1761849900574373e-06, "loss": 0.0147, "step": 51528 }, { "epoch": 15.85, "learning_rate": 2.1758746774450444e-06, "loss": 0.0058, "step": 51529 }, { "epoch": 15.86, "learning_rate": 2.1755643842578457e-06, "loss": 0.0048, "step": 51530 }, { "epoch": 15.86, "learning_rate": 2.1752541104966107e-06, "loss": 0.0055, "step": 51531 }, { "epoch": 15.86, "learning_rate": 2.1749438561621106e-06, "loss": 0.0096, "step": 51532 }, { "epoch": 15.86, "learning_rate": 2.174633621255112e-06, "loss": 0.0069, "step": 51533 }, { "epoch": 15.86, "learning_rate": 2.174323405776392e-06, "loss": 0.0059, "step": 51534 }, { "epoch": 15.86, "learning_rate": 2.1740132097267164e-06, "loss": 0.0131, "step": 51535 }, { "epoch": 15.86, "learning_rate": 2.17370303310686e-06, "loss": 0.0095, "step": 51536 }, { "epoch": 15.86, "learning_rate": 2.173392875917586e-06, "loss": 0.0163, "step": 51537 }, { "epoch": 15.86, "learning_rate": 2.1730827381596643e-06, "loss": 0.0038, "step": 51538 }, { "epoch": 15.86, "learning_rate": 2.1727726198338705e-06, "loss": 0.0063, "step": 51539 }, { "epoch": 15.86, "learning_rate": 2.1724625209409735e-06, "loss": 0.0079, "step": 51540 }, { "epoch": 15.86, "learning_rate": 2.1721524414817406e-06, "loss": 0.0107, "step": 51541 }, { "epoch": 15.86, "learning_rate": 2.1718423814569434e-06, "loss": 0.0085, "step": 51542 }, { "epoch": 15.86, "learning_rate": 2.1715323408673506e-06, "loss": 0.0092, "step": 51543 }, { "epoch": 15.86, "learning_rate": 2.171222319713734e-06, "loss": 0.0042, "step": 51544 }, { "epoch": 15.86, "learning_rate": 2.1709123179968607e-06, "loss": 0.0047, "step": 51545 }, { "epoch": 15.86, "learning_rate": 2.170602335717501e-06, "loss": 0.0074, "step": 51546 }, { "epoch": 15.86, "learning_rate": 2.170292372876426e-06, "loss": 0.0165, "step": 51547 }, { "epoch": 15.86, "learning_rate": 2.1699824294744e-06, "loss": 0.0052, "step": 51548 }, { "epoch": 15.86, "learning_rate": 2.1696725055122036e-06, "loss": 0.0082, "step": 51549 }, { "epoch": 15.86, "learning_rate": 2.1693626009905968e-06, "loss": 0.0126, "step": 51550 }, { "epoch": 15.86, "learning_rate": 2.16905271591035e-06, "loss": 0.0063, "step": 51551 }, { "epoch": 15.86, "learning_rate": 2.168742850272235e-06, "loss": 0.0036, "step": 51552 }, { "epoch": 15.86, "learning_rate": 2.1684330040770183e-06, "loss": 0.0089, "step": 51553 }, { "epoch": 15.86, "learning_rate": 2.168123177325473e-06, "loss": 0.0098, "step": 51554 }, { "epoch": 15.86, "learning_rate": 2.167813370018367e-06, "loss": 0.0078, "step": 51555 }, { "epoch": 15.86, "learning_rate": 2.1675035821564682e-06, "loss": 0.0144, "step": 51556 }, { "epoch": 15.86, "learning_rate": 2.167193813740549e-06, "loss": 0.0054, "step": 51557 }, { "epoch": 15.86, "learning_rate": 2.166884064771372e-06, "loss": 0.0053, "step": 51558 }, { "epoch": 15.86, "learning_rate": 2.1665743352497116e-06, "loss": 0.0081, "step": 51559 }, { "epoch": 15.86, "learning_rate": 2.166264625176335e-06, "loss": 0.0099, "step": 51560 }, { "epoch": 15.86, "learning_rate": 2.1659549345520125e-06, "loss": 0.0104, "step": 51561 }, { "epoch": 15.87, "learning_rate": 2.1656452633775103e-06, "loss": 0.0059, "step": 51562 }, { "epoch": 15.87, "learning_rate": 2.1653356116536006e-06, "loss": 0.0088, "step": 51563 }, { "epoch": 15.87, "learning_rate": 2.165025979381049e-06, "loss": 0.0116, "step": 51564 }, { "epoch": 15.87, "learning_rate": 2.164716366560626e-06, "loss": 0.009, "step": 51565 }, { "epoch": 15.87, "learning_rate": 2.1644067731931005e-06, "loss": 0.0077, "step": 51566 }, { "epoch": 15.87, "learning_rate": 2.1640971992792416e-06, "loss": 0.0247, "step": 51567 }, { "epoch": 15.87, "learning_rate": 2.1637876448198127e-06, "loss": 0.008, "step": 51568 }, { "epoch": 15.87, "learning_rate": 2.1634781098155888e-06, "loss": 0.0092, "step": 51569 }, { "epoch": 15.87, "learning_rate": 2.16316859426734e-06, "loss": 0.0049, "step": 51570 }, { "epoch": 15.87, "learning_rate": 2.162859098175827e-06, "loss": 0.0097, "step": 51571 }, { "epoch": 15.87, "learning_rate": 2.1625496215418228e-06, "loss": 0.0042, "step": 51572 }, { "epoch": 15.87, "learning_rate": 2.1622401643660917e-06, "loss": 0.0046, "step": 51573 }, { "epoch": 15.87, "learning_rate": 2.161930726649408e-06, "loss": 0.0077, "step": 51574 }, { "epoch": 15.87, "learning_rate": 2.1616213083925375e-06, "loss": 0.0132, "step": 51575 }, { "epoch": 15.87, "learning_rate": 2.161311909596249e-06, "loss": 0.0059, "step": 51576 }, { "epoch": 15.87, "learning_rate": 2.161002530261309e-06, "loss": 0.0058, "step": 51577 }, { "epoch": 15.87, "learning_rate": 2.160693170388486e-06, "loss": 0.0133, "step": 51578 }, { "epoch": 15.87, "learning_rate": 2.1603838299785486e-06, "loss": 0.0156, "step": 51579 }, { "epoch": 15.87, "learning_rate": 2.160074509032265e-06, "loss": 0.0065, "step": 51580 }, { "epoch": 15.87, "learning_rate": 2.1597652075504026e-06, "loss": 0.0161, "step": 51581 }, { "epoch": 15.87, "learning_rate": 2.1594559255337298e-06, "loss": 0.0037, "step": 51582 }, { "epoch": 15.87, "learning_rate": 2.1591466629830138e-06, "loss": 0.0041, "step": 51583 }, { "epoch": 15.87, "learning_rate": 2.1588374198990237e-06, "loss": 0.0091, "step": 51584 }, { "epoch": 15.87, "learning_rate": 2.1585281962825257e-06, "loss": 0.0073, "step": 51585 }, { "epoch": 15.87, "learning_rate": 2.1582189921342876e-06, "loss": 0.0048, "step": 51586 }, { "epoch": 15.87, "learning_rate": 2.1579098074550784e-06, "loss": 0.0114, "step": 51587 }, { "epoch": 15.87, "learning_rate": 2.157600642245662e-06, "loss": 0.0061, "step": 51588 }, { "epoch": 15.87, "learning_rate": 2.1572914965068113e-06, "loss": 0.0079, "step": 51589 }, { "epoch": 15.87, "learning_rate": 2.1569823702392924e-06, "loss": 0.0097, "step": 51590 }, { "epoch": 15.87, "learning_rate": 2.1566732634438738e-06, "loss": 0.0116, "step": 51591 }, { "epoch": 15.87, "learning_rate": 2.1563641761213183e-06, "loss": 0.0077, "step": 51592 }, { "epoch": 15.87, "learning_rate": 2.1560551082723937e-06, "loss": 0.0115, "step": 51593 }, { "epoch": 15.87, "learning_rate": 2.155746059897873e-06, "loss": 0.0077, "step": 51594 }, { "epoch": 15.88, "learning_rate": 2.1554370309985184e-06, "loss": 0.0097, "step": 51595 }, { "epoch": 15.88, "learning_rate": 2.1551280215751005e-06, "loss": 0.0072, "step": 51596 }, { "epoch": 15.88, "learning_rate": 2.1548190316283835e-06, "loss": 0.0068, "step": 51597 }, { "epoch": 15.88, "learning_rate": 2.154510061159136e-06, "loss": 0.0059, "step": 51598 }, { "epoch": 15.88, "learning_rate": 2.154201110168126e-06, "loss": 0.0071, "step": 51599 }, { "epoch": 15.88, "learning_rate": 2.1538921786561185e-06, "loss": 0.0078, "step": 51600 }, { "epoch": 15.88, "learning_rate": 2.153583266623882e-06, "loss": 0.0081, "step": 51601 }, { "epoch": 15.88, "learning_rate": 2.1532743740721827e-06, "loss": 0.0106, "step": 51602 }, { "epoch": 15.88, "learning_rate": 2.1529655010017848e-06, "loss": 0.0051, "step": 51603 }, { "epoch": 15.88, "learning_rate": 2.1526566474134636e-06, "loss": 0.0085, "step": 51604 }, { "epoch": 15.88, "learning_rate": 2.1523478133079776e-06, "loss": 0.0089, "step": 51605 }, { "epoch": 15.88, "learning_rate": 2.152038998686097e-06, "loss": 0.013, "step": 51606 }, { "epoch": 15.88, "learning_rate": 2.1517302035485865e-06, "loss": 0.0109, "step": 51607 }, { "epoch": 15.88, "learning_rate": 2.151421427896212e-06, "loss": 0.0108, "step": 51608 }, { "epoch": 15.88, "learning_rate": 2.1511126717297447e-06, "loss": 0.0094, "step": 51609 }, { "epoch": 15.88, "learning_rate": 2.1508039350499497e-06, "loss": 0.005, "step": 51610 }, { "epoch": 15.88, "learning_rate": 2.150495217857591e-06, "loss": 0.007, "step": 51611 }, { "epoch": 15.88, "learning_rate": 2.1501865201534413e-06, "loss": 0.0092, "step": 51612 }, { "epoch": 15.88, "learning_rate": 2.149877841938255e-06, "loss": 0.0085, "step": 51613 }, { "epoch": 15.88, "learning_rate": 2.1495691832128084e-06, "loss": 0.0157, "step": 51614 }, { "epoch": 15.88, "learning_rate": 2.1492605439778657e-06, "loss": 0.0099, "step": 51615 }, { "epoch": 15.88, "learning_rate": 2.148951924234193e-06, "loss": 0.0071, "step": 51616 }, { "epoch": 15.88, "learning_rate": 2.148643323982554e-06, "loss": 0.0112, "step": 51617 }, { "epoch": 15.88, "learning_rate": 2.148334743223719e-06, "loss": 0.0116, "step": 51618 }, { "epoch": 15.88, "learning_rate": 2.1480261819584525e-06, "loss": 0.0065, "step": 51619 }, { "epoch": 15.88, "learning_rate": 2.147717640187519e-06, "loss": 0.0078, "step": 51620 }, { "epoch": 15.88, "learning_rate": 2.147409117911685e-06, "loss": 0.0076, "step": 51621 }, { "epoch": 15.88, "learning_rate": 2.1471006151317187e-06, "loss": 0.005, "step": 51622 }, { "epoch": 15.88, "learning_rate": 2.146792131848381e-06, "loss": 0.0101, "step": 51623 }, { "epoch": 15.88, "learning_rate": 2.1464836680624447e-06, "loss": 0.011, "step": 51624 }, { "epoch": 15.88, "learning_rate": 2.1461752237746738e-06, "loss": 0.0103, "step": 51625 }, { "epoch": 15.88, "learning_rate": 2.1458667989858304e-06, "loss": 0.0063, "step": 51626 }, { "epoch": 15.89, "learning_rate": 2.145558393696683e-06, "loss": 0.012, "step": 51627 }, { "epoch": 15.89, "learning_rate": 2.145250007907993e-06, "loss": 0.0089, "step": 51628 }, { "epoch": 15.89, "learning_rate": 2.1449416416205327e-06, "loss": 0.0124, "step": 51629 }, { "epoch": 15.89, "learning_rate": 2.1446332948350644e-06, "loss": 0.0063, "step": 51630 }, { "epoch": 15.89, "learning_rate": 2.1443249675523536e-06, "loss": 0.0146, "step": 51631 }, { "epoch": 15.89, "learning_rate": 2.144016659773167e-06, "loss": 0.0053, "step": 51632 }, { "epoch": 15.89, "learning_rate": 2.1437083714982687e-06, "loss": 0.0052, "step": 51633 }, { "epoch": 15.89, "learning_rate": 2.1434001027284236e-06, "loss": 0.0097, "step": 51634 }, { "epoch": 15.89, "learning_rate": 2.1430918534643996e-06, "loss": 0.0063, "step": 51635 }, { "epoch": 15.89, "learning_rate": 2.1427836237069587e-06, "loss": 0.0032, "step": 51636 }, { "epoch": 15.89, "learning_rate": 2.142475413456869e-06, "loss": 0.0083, "step": 51637 }, { "epoch": 15.89, "learning_rate": 2.1421672227148915e-06, "loss": 0.007, "step": 51638 }, { "epoch": 15.89, "learning_rate": 2.1418590514817995e-06, "loss": 0.0072, "step": 51639 }, { "epoch": 15.89, "learning_rate": 2.141550899758351e-06, "loss": 0.0118, "step": 51640 }, { "epoch": 15.89, "learning_rate": 2.1412427675453128e-06, "loss": 0.0067, "step": 51641 }, { "epoch": 15.89, "learning_rate": 2.1409346548434494e-06, "loss": 0.0049, "step": 51642 }, { "epoch": 15.89, "learning_rate": 2.140626561653525e-06, "loss": 0.012, "step": 51643 }, { "epoch": 15.89, "learning_rate": 2.1403184879763083e-06, "loss": 0.0081, "step": 51644 }, { "epoch": 15.89, "learning_rate": 2.1400104338125616e-06, "loss": 0.0081, "step": 51645 }, { "epoch": 15.89, "learning_rate": 2.139702399163053e-06, "loss": 0.0088, "step": 51646 }, { "epoch": 15.89, "learning_rate": 2.1393943840285414e-06, "loss": 0.0067, "step": 51647 }, { "epoch": 15.89, "learning_rate": 2.1390863884097913e-06, "loss": 0.0058, "step": 51648 }, { "epoch": 15.89, "learning_rate": 2.138778412307574e-06, "loss": 0.0055, "step": 51649 }, { "epoch": 15.89, "learning_rate": 2.13847045572265e-06, "loss": 0.0072, "step": 51650 }, { "epoch": 15.89, "learning_rate": 2.1381625186557842e-06, "loss": 0.002, "step": 51651 }, { "epoch": 15.89, "learning_rate": 2.1378546011077416e-06, "loss": 0.0079, "step": 51652 }, { "epoch": 15.89, "learning_rate": 2.137546703079286e-06, "loss": 0.0068, "step": 51653 }, { "epoch": 15.89, "learning_rate": 2.137238824571184e-06, "loss": 0.0123, "step": 51654 }, { "epoch": 15.89, "learning_rate": 2.1369309655841963e-06, "loss": 0.0065, "step": 51655 }, { "epoch": 15.89, "learning_rate": 2.1366231261190905e-06, "loss": 0.0113, "step": 51656 }, { "epoch": 15.89, "learning_rate": 2.1363153061766297e-06, "loss": 0.0057, "step": 51657 }, { "epoch": 15.89, "learning_rate": 2.136007505757575e-06, "loss": 0.0128, "step": 51658 }, { "epoch": 15.89, "learning_rate": 2.1356997248626954e-06, "loss": 0.0139, "step": 51659 }, { "epoch": 15.9, "learning_rate": 2.135391963492757e-06, "loss": 0.0109, "step": 51660 }, { "epoch": 15.9, "learning_rate": 2.1350842216485167e-06, "loss": 0.0043, "step": 51661 }, { "epoch": 15.9, "learning_rate": 2.1347764993307417e-06, "loss": 0.0068, "step": 51662 }, { "epoch": 15.9, "learning_rate": 2.1344687965401944e-06, "loss": 0.0094, "step": 51663 }, { "epoch": 15.9, "learning_rate": 2.134161113277643e-06, "loss": 0.0041, "step": 51664 }, { "epoch": 15.9, "learning_rate": 2.133853449543849e-06, "loss": 0.0118, "step": 51665 }, { "epoch": 15.9, "learning_rate": 2.1335458053395764e-06, "loss": 0.0145, "step": 51666 }, { "epoch": 15.9, "learning_rate": 2.133238180665591e-06, "loss": 0.0116, "step": 51667 }, { "epoch": 15.9, "learning_rate": 2.1329305755226494e-06, "loss": 0.0056, "step": 51668 }, { "epoch": 15.9, "learning_rate": 2.1326229899115235e-06, "loss": 0.0105, "step": 51669 }, { "epoch": 15.9, "learning_rate": 2.1323154238329722e-06, "loss": 0.0066, "step": 51670 }, { "epoch": 15.9, "learning_rate": 2.132007877287762e-06, "loss": 0.0077, "step": 51671 }, { "epoch": 15.9, "learning_rate": 2.1317003502766554e-06, "loss": 0.0133, "step": 51672 }, { "epoch": 15.9, "learning_rate": 2.131392842800415e-06, "loss": 0.0087, "step": 51673 }, { "epoch": 15.9, "learning_rate": 2.1310853548598043e-06, "loss": 0.006, "step": 51674 }, { "epoch": 15.9, "learning_rate": 2.1307778864555873e-06, "loss": 0.0071, "step": 51675 }, { "epoch": 15.9, "learning_rate": 2.130470437588528e-06, "loss": 0.0105, "step": 51676 }, { "epoch": 15.9, "learning_rate": 2.1301630082593895e-06, "loss": 0.0092, "step": 51677 }, { "epoch": 15.9, "learning_rate": 2.129855598468931e-06, "loss": 0.004, "step": 51678 }, { "epoch": 15.9, "learning_rate": 2.129548208217922e-06, "loss": 0.0075, "step": 51679 }, { "epoch": 15.9, "learning_rate": 2.1292408375071225e-06, "loss": 0.0066, "step": 51680 }, { "epoch": 15.9, "learning_rate": 2.1289334863373e-06, "loss": 0.0041, "step": 51681 }, { "epoch": 15.9, "learning_rate": 2.12862615470921e-06, "loss": 0.0066, "step": 51682 }, { "epoch": 15.9, "learning_rate": 2.128318842623618e-06, "loss": 0.0059, "step": 51683 }, { "epoch": 15.9, "learning_rate": 2.1280115500812893e-06, "loss": 0.0169, "step": 51684 }, { "epoch": 15.9, "learning_rate": 2.127704277082987e-06, "loss": 0.0083, "step": 51685 }, { "epoch": 15.9, "learning_rate": 2.1273970236294717e-06, "loss": 0.0109, "step": 51686 }, { "epoch": 15.9, "learning_rate": 2.127089789721507e-06, "loss": 0.0201, "step": 51687 }, { "epoch": 15.9, "learning_rate": 2.126782575359857e-06, "loss": 0.007, "step": 51688 }, { "epoch": 15.9, "learning_rate": 2.126475380545282e-06, "loss": 0.0084, "step": 51689 }, { "epoch": 15.9, "learning_rate": 2.126168205278547e-06, "loss": 0.0083, "step": 51690 }, { "epoch": 15.9, "learning_rate": 2.125861049560414e-06, "loss": 0.0067, "step": 51691 }, { "epoch": 15.91, "learning_rate": 2.1255539133916435e-06, "loss": 0.0074, "step": 51692 }, { "epoch": 15.91, "learning_rate": 2.1252467967729984e-06, "loss": 0.0094, "step": 51693 }, { "epoch": 15.91, "learning_rate": 2.1249396997052476e-06, "loss": 0.008, "step": 51694 }, { "epoch": 15.91, "learning_rate": 2.1246326221891458e-06, "loss": 0.0061, "step": 51695 }, { "epoch": 15.91, "learning_rate": 2.124325564225458e-06, "loss": 0.007, "step": 51696 }, { "epoch": 15.91, "learning_rate": 2.124018525814947e-06, "loss": 0.007, "step": 51697 }, { "epoch": 15.91, "learning_rate": 2.1237115069583713e-06, "loss": 0.0067, "step": 51698 }, { "epoch": 15.91, "learning_rate": 2.123404507656499e-06, "loss": 0.0098, "step": 51699 }, { "epoch": 15.91, "learning_rate": 2.123097527910091e-06, "loss": 0.0173, "step": 51700 }, { "epoch": 15.91, "learning_rate": 2.1227905677199077e-06, "loss": 0.0066, "step": 51701 }, { "epoch": 15.91, "learning_rate": 2.1224836270867146e-06, "loss": 0.0063, "step": 51702 }, { "epoch": 15.91, "learning_rate": 2.1221767060112653e-06, "loss": 0.008, "step": 51703 }, { "epoch": 15.91, "learning_rate": 2.121869804494331e-06, "loss": 0.0074, "step": 51704 }, { "epoch": 15.91, "learning_rate": 2.1215629225366683e-06, "loss": 0.0059, "step": 51705 }, { "epoch": 15.91, "learning_rate": 2.121256060139042e-06, "loss": 0.0061, "step": 51706 }, { "epoch": 15.91, "learning_rate": 2.120949217302213e-06, "loss": 0.0071, "step": 51707 }, { "epoch": 15.91, "learning_rate": 2.1206423940269428e-06, "loss": 0.0114, "step": 51708 }, { "epoch": 15.91, "learning_rate": 2.1203355903139934e-06, "loss": 0.0064, "step": 51709 }, { "epoch": 15.91, "learning_rate": 2.120028806164127e-06, "loss": 0.0058, "step": 51710 }, { "epoch": 15.91, "learning_rate": 2.119722041578104e-06, "loss": 0.0091, "step": 51711 }, { "epoch": 15.91, "learning_rate": 2.119415296556686e-06, "loss": 0.0163, "step": 51712 }, { "epoch": 15.91, "learning_rate": 2.1191085711006343e-06, "loss": 0.0102, "step": 51713 }, { "epoch": 15.91, "learning_rate": 2.1188018652107124e-06, "loss": 0.0081, "step": 51714 }, { "epoch": 15.91, "learning_rate": 2.1184951788876852e-06, "loss": 0.0088, "step": 51715 }, { "epoch": 15.91, "learning_rate": 2.1181885121323054e-06, "loss": 0.0055, "step": 51716 }, { "epoch": 15.91, "learning_rate": 2.117881864945339e-06, "loss": 0.0122, "step": 51717 }, { "epoch": 15.91, "learning_rate": 2.117575237327545e-06, "loss": 0.0081, "step": 51718 }, { "epoch": 15.91, "learning_rate": 2.1172686292796886e-06, "loss": 0.007, "step": 51719 }, { "epoch": 15.91, "learning_rate": 2.116962040802529e-06, "loss": 0.0102, "step": 51720 }, { "epoch": 15.91, "learning_rate": 2.116655471896828e-06, "loss": 0.0165, "step": 51721 }, { "epoch": 15.91, "learning_rate": 2.116348922563346e-06, "loss": 0.0069, "step": 51722 }, { "epoch": 15.91, "learning_rate": 2.116042392802845e-06, "loss": 0.0054, "step": 51723 }, { "epoch": 15.91, "learning_rate": 2.1157358826160846e-06, "loss": 0.0054, "step": 51724 }, { "epoch": 15.92, "learning_rate": 2.115429392003826e-06, "loss": 0.0069, "step": 51725 }, { "epoch": 15.92, "learning_rate": 2.115122920966831e-06, "loss": 0.0067, "step": 51726 }, { "epoch": 15.92, "learning_rate": 2.1148164695058603e-06, "loss": 0.0107, "step": 51727 }, { "epoch": 15.92, "learning_rate": 2.1145100376216745e-06, "loss": 0.0101, "step": 51728 }, { "epoch": 15.92, "learning_rate": 2.1142036253150344e-06, "loss": 0.008, "step": 51729 }, { "epoch": 15.92, "learning_rate": 2.1138972325867002e-06, "loss": 0.0084, "step": 51730 }, { "epoch": 15.92, "learning_rate": 2.113590859437433e-06, "loss": 0.0086, "step": 51731 }, { "epoch": 15.92, "learning_rate": 2.1132845058679942e-06, "loss": 0.0083, "step": 51732 }, { "epoch": 15.92, "learning_rate": 2.112978171879141e-06, "loss": 0.0112, "step": 51733 }, { "epoch": 15.92, "learning_rate": 2.11267185747164e-06, "loss": 0.0067, "step": 51734 }, { "epoch": 15.92, "learning_rate": 2.112365562646248e-06, "loss": 0.0094, "step": 51735 }, { "epoch": 15.92, "learning_rate": 2.1120592874037284e-06, "loss": 0.0171, "step": 51736 }, { "epoch": 15.92, "learning_rate": 2.1117530317448365e-06, "loss": 0.0075, "step": 51737 }, { "epoch": 15.92, "learning_rate": 2.111446795670332e-06, "loss": 0.0068, "step": 51738 }, { "epoch": 15.92, "learning_rate": 2.1111405791809812e-06, "loss": 0.0171, "step": 51739 }, { "epoch": 15.92, "learning_rate": 2.1108343822775424e-06, "loss": 0.0029, "step": 51740 }, { "epoch": 15.92, "learning_rate": 2.1105282049607744e-06, "loss": 0.0066, "step": 51741 }, { "epoch": 15.92, "learning_rate": 2.1102220472314384e-06, "loss": 0.0108, "step": 51742 }, { "epoch": 15.92, "learning_rate": 2.109915909090293e-06, "loss": 0.0108, "step": 51743 }, { "epoch": 15.92, "learning_rate": 2.1096097905380995e-06, "loss": 0.0053, "step": 51744 }, { "epoch": 15.92, "learning_rate": 2.1093036915756193e-06, "loss": 0.005, "step": 51745 }, { "epoch": 15.92, "learning_rate": 2.1089976122036092e-06, "loss": 0.0042, "step": 51746 }, { "epoch": 15.92, "learning_rate": 2.1086915524228313e-06, "loss": 0.0074, "step": 51747 }, { "epoch": 15.92, "learning_rate": 2.108385512234041e-06, "loss": 0.0061, "step": 51748 }, { "epoch": 15.92, "learning_rate": 2.1080794916380085e-06, "loss": 0.006, "step": 51749 }, { "epoch": 15.92, "learning_rate": 2.107773490635484e-06, "loss": 0.0052, "step": 51750 }, { "epoch": 15.92, "learning_rate": 2.1074675092272313e-06, "loss": 0.0081, "step": 51751 }, { "epoch": 15.92, "learning_rate": 2.107161547414007e-06, "loss": 0.015, "step": 51752 }, { "epoch": 15.92, "learning_rate": 2.1068556051965717e-06, "loss": 0.0049, "step": 51753 }, { "epoch": 15.92, "learning_rate": 2.106549682575688e-06, "loss": 0.0082, "step": 51754 }, { "epoch": 15.92, "learning_rate": 2.106243779552114e-06, "loss": 0.0121, "step": 51755 }, { "epoch": 15.92, "learning_rate": 2.1059378961266076e-06, "loss": 0.0091, "step": 51756 }, { "epoch": 15.93, "learning_rate": 2.105632032299932e-06, "loss": 0.0054, "step": 51757 }, { "epoch": 15.93, "learning_rate": 2.1053261880728394e-06, "loss": 0.0048, "step": 51758 }, { "epoch": 15.93, "learning_rate": 2.1050203634460963e-06, "loss": 0.0175, "step": 51759 }, { "epoch": 15.93, "learning_rate": 2.1047145584204587e-06, "loss": 0.004, "step": 51760 }, { "epoch": 15.93, "learning_rate": 2.1044087729966856e-06, "loss": 0.007, "step": 51761 }, { "epoch": 15.93, "learning_rate": 2.1041030071755373e-06, "loss": 0.0094, "step": 51762 }, { "epoch": 15.93, "learning_rate": 2.103797260957774e-06, "loss": 0.0069, "step": 51763 }, { "epoch": 15.93, "learning_rate": 2.103491534344152e-06, "loss": 0.0124, "step": 51764 }, { "epoch": 15.93, "learning_rate": 2.1031858273354322e-06, "loss": 0.0147, "step": 51765 }, { "epoch": 15.93, "learning_rate": 2.1028801399323727e-06, "loss": 0.0057, "step": 51766 }, { "epoch": 15.93, "learning_rate": 2.102574472135732e-06, "loss": 0.0092, "step": 51767 }, { "epoch": 15.93, "learning_rate": 2.1022688239462686e-06, "loss": 0.0185, "step": 51768 }, { "epoch": 15.93, "learning_rate": 2.1019631953647446e-06, "loss": 0.0068, "step": 51769 }, { "epoch": 15.93, "learning_rate": 2.10165758639192e-06, "loss": 0.0054, "step": 51770 }, { "epoch": 15.93, "learning_rate": 2.101351997028547e-06, "loss": 0.0078, "step": 51771 }, { "epoch": 15.93, "learning_rate": 2.101046427275387e-06, "loss": 0.0128, "step": 51772 }, { "epoch": 15.93, "learning_rate": 2.100740877133197e-06, "loss": 0.0105, "step": 51773 }, { "epoch": 15.93, "learning_rate": 2.100435346602742e-06, "loss": 0.0108, "step": 51774 }, { "epoch": 15.93, "learning_rate": 2.1001298356847744e-06, "loss": 0.009, "step": 51775 }, { "epoch": 15.93, "learning_rate": 2.099824344380056e-06, "loss": 0.0079, "step": 51776 }, { "epoch": 15.93, "learning_rate": 2.0995188726893436e-06, "loss": 0.0057, "step": 51777 }, { "epoch": 15.93, "learning_rate": 2.099213420613395e-06, "loss": 0.0086, "step": 51778 }, { "epoch": 15.93, "learning_rate": 2.0989079881529704e-06, "loss": 0.0116, "step": 51779 }, { "epoch": 15.93, "learning_rate": 2.0986025753088268e-06, "loss": 0.0176, "step": 51780 }, { "epoch": 15.93, "learning_rate": 2.0982971820817235e-06, "loss": 0.004, "step": 51781 }, { "epoch": 15.93, "learning_rate": 2.097991808472417e-06, "loss": 0.0135, "step": 51782 }, { "epoch": 15.93, "learning_rate": 2.0976864544816666e-06, "loss": 0.0198, "step": 51783 }, { "epoch": 15.93, "learning_rate": 2.0973811201102313e-06, "loss": 0.0065, "step": 51784 }, { "epoch": 15.93, "learning_rate": 2.097075805358868e-06, "loss": 0.0161, "step": 51785 }, { "epoch": 15.93, "learning_rate": 2.096770510228334e-06, "loss": 0.0061, "step": 51786 }, { "epoch": 15.93, "learning_rate": 2.0964652347193894e-06, "loss": 0.0046, "step": 51787 }, { "epoch": 15.93, "learning_rate": 2.096159978832787e-06, "loss": 0.0138, "step": 51788 }, { "epoch": 15.93, "learning_rate": 2.095854742569292e-06, "loss": 0.0065, "step": 51789 }, { "epoch": 15.94, "learning_rate": 2.095549525929659e-06, "loss": 0.0087, "step": 51790 }, { "epoch": 15.94, "learning_rate": 2.0952443289146475e-06, "loss": 0.006, "step": 51791 }, { "epoch": 15.94, "learning_rate": 2.0949391515250116e-06, "loss": 0.0054, "step": 51792 }, { "epoch": 15.94, "learning_rate": 2.094633993761507e-06, "loss": 0.0046, "step": 51793 }, { "epoch": 15.94, "learning_rate": 2.0943288556248975e-06, "loss": 0.0086, "step": 51794 }, { "epoch": 15.94, "learning_rate": 2.0940237371159387e-06, "loss": 0.0084, "step": 51795 }, { "epoch": 15.94, "learning_rate": 2.093718638235388e-06, "loss": 0.0099, "step": 51796 }, { "epoch": 15.94, "learning_rate": 2.093413558984002e-06, "loss": 0.004, "step": 51797 }, { "epoch": 15.94, "learning_rate": 2.0931084993625385e-06, "loss": 0.0058, "step": 51798 }, { "epoch": 15.94, "learning_rate": 2.092803459371755e-06, "loss": 0.0138, "step": 51799 }, { "epoch": 15.94, "learning_rate": 2.0924984390124093e-06, "loss": 0.0077, "step": 51800 }, { "epoch": 15.94, "learning_rate": 2.092193438285257e-06, "loss": 0.0073, "step": 51801 }, { "epoch": 15.94, "learning_rate": 2.0918884571910582e-06, "loss": 0.012, "step": 51802 }, { "epoch": 15.94, "learning_rate": 2.0915834957305638e-06, "loss": 0.0069, "step": 51803 }, { "epoch": 15.94, "learning_rate": 2.091278553904542e-06, "loss": 0.0065, "step": 51804 }, { "epoch": 15.94, "learning_rate": 2.0909736317137406e-06, "loss": 0.0076, "step": 51805 }, { "epoch": 15.94, "learning_rate": 2.09066872915892e-06, "loss": 0.0123, "step": 51806 }, { "epoch": 15.94, "learning_rate": 2.090363846240836e-06, "loss": 0.008, "step": 51807 }, { "epoch": 15.94, "learning_rate": 2.0900589829602437e-06, "loss": 0.0092, "step": 51808 }, { "epoch": 15.94, "learning_rate": 2.0897541393179042e-06, "loss": 0.0087, "step": 51809 }, { "epoch": 15.94, "learning_rate": 2.0894493153145733e-06, "loss": 0.0069, "step": 51810 }, { "epoch": 15.94, "learning_rate": 2.089144510951008e-06, "loss": 0.0058, "step": 51811 }, { "epoch": 15.94, "learning_rate": 2.088839726227966e-06, "loss": 0.0067, "step": 51812 }, { "epoch": 15.94, "learning_rate": 2.088534961146197e-06, "loss": 0.0085, "step": 51813 }, { "epoch": 15.94, "learning_rate": 2.088230215706466e-06, "loss": 0.0045, "step": 51814 }, { "epoch": 15.94, "learning_rate": 2.0879254899095257e-06, "loss": 0.0067, "step": 51815 }, { "epoch": 15.94, "learning_rate": 2.0876207837561334e-06, "loss": 0.0068, "step": 51816 }, { "epoch": 15.94, "learning_rate": 2.0873160972470462e-06, "loss": 0.0059, "step": 51817 }, { "epoch": 15.94, "learning_rate": 2.0870114303830192e-06, "loss": 0.0065, "step": 51818 }, { "epoch": 15.94, "learning_rate": 2.08670678316481e-06, "loss": 0.0059, "step": 51819 }, { "epoch": 15.94, "learning_rate": 2.0864021555931746e-06, "loss": 0.0074, "step": 51820 }, { "epoch": 15.94, "learning_rate": 2.0860975476688692e-06, "loss": 0.0145, "step": 51821 }, { "epoch": 15.95, "learning_rate": 2.085792959392647e-06, "loss": 0.0073, "step": 51822 }, { "epoch": 15.95, "learning_rate": 2.08548839076527e-06, "loss": 0.0096, "step": 51823 }, { "epoch": 15.95, "learning_rate": 2.0851838417874913e-06, "loss": 0.0052, "step": 51824 }, { "epoch": 15.95, "learning_rate": 2.0848793124600707e-06, "loss": 0.0071, "step": 51825 }, { "epoch": 15.95, "learning_rate": 2.0845748027837585e-06, "loss": 0.0055, "step": 51826 }, { "epoch": 15.95, "learning_rate": 2.08427031275931e-06, "loss": 0.0047, "step": 51827 }, { "epoch": 15.95, "learning_rate": 2.0839658423874878e-06, "loss": 0.0036, "step": 51828 }, { "epoch": 15.95, "learning_rate": 2.083661391669043e-06, "loss": 0.0075, "step": 51829 }, { "epoch": 15.95, "learning_rate": 2.0833569606047332e-06, "loss": 0.0144, "step": 51830 }, { "epoch": 15.95, "learning_rate": 2.0830525491953137e-06, "loss": 0.0089, "step": 51831 }, { "epoch": 15.95, "learning_rate": 2.082748157441541e-06, "loss": 0.0105, "step": 51832 }, { "epoch": 15.95, "learning_rate": 2.082443785344169e-06, "loss": 0.0071, "step": 51833 }, { "epoch": 15.95, "learning_rate": 2.0821394329039555e-06, "loss": 0.004, "step": 51834 }, { "epoch": 15.95, "learning_rate": 2.081835100121655e-06, "loss": 0.0098, "step": 51835 }, { "epoch": 15.95, "learning_rate": 2.0815307869980237e-06, "loss": 0.0104, "step": 51836 }, { "epoch": 15.95, "learning_rate": 2.0812264935338132e-06, "loss": 0.0083, "step": 51837 }, { "epoch": 15.95, "learning_rate": 2.0809222197297863e-06, "loss": 0.0111, "step": 51838 }, { "epoch": 15.95, "learning_rate": 2.0806179655866964e-06, "loss": 0.0128, "step": 51839 }, { "epoch": 15.95, "learning_rate": 2.0803137311052944e-06, "loss": 0.0056, "step": 51840 }, { "epoch": 15.95, "learning_rate": 2.080009516286339e-06, "loss": 0.0148, "step": 51841 }, { "epoch": 15.95, "learning_rate": 2.079705321130582e-06, "loss": 0.018, "step": 51842 }, { "epoch": 15.95, "learning_rate": 2.079401145638783e-06, "loss": 0.0087, "step": 51843 }, { "epoch": 15.95, "learning_rate": 2.0790969898116965e-06, "loss": 0.0119, "step": 51844 }, { "epoch": 15.95, "learning_rate": 2.0787928536500767e-06, "loss": 0.0074, "step": 51845 }, { "epoch": 15.95, "learning_rate": 2.0784887371546816e-06, "loss": 0.0075, "step": 51846 }, { "epoch": 15.95, "learning_rate": 2.0781846403262572e-06, "loss": 0.0104, "step": 51847 }, { "epoch": 15.95, "learning_rate": 2.0778805631655686e-06, "loss": 0.006, "step": 51848 }, { "epoch": 15.95, "learning_rate": 2.0775765056733667e-06, "loss": 0.0046, "step": 51849 }, { "epoch": 15.95, "learning_rate": 2.077272467850405e-06, "loss": 0.0048, "step": 51850 }, { "epoch": 15.95, "learning_rate": 2.076968449697442e-06, "loss": 0.0059, "step": 51851 }, { "epoch": 15.95, "learning_rate": 2.076664451215229e-06, "loss": 0.004, "step": 51852 }, { "epoch": 15.95, "learning_rate": 2.0763604724045217e-06, "loss": 0.0075, "step": 51853 }, { "epoch": 15.95, "learning_rate": 2.076056513266076e-06, "loss": 0.0096, "step": 51854 }, { "epoch": 15.96, "learning_rate": 2.0757525738006455e-06, "loss": 0.0078, "step": 51855 }, { "epoch": 15.96, "learning_rate": 2.075448654008986e-06, "loss": 0.01, "step": 51856 }, { "epoch": 15.96, "learning_rate": 2.0751447538918468e-06, "loss": 0.0056, "step": 51857 }, { "epoch": 15.96, "learning_rate": 2.0748408734499904e-06, "loss": 0.0035, "step": 51858 }, { "epoch": 15.96, "learning_rate": 2.074537012684168e-06, "loss": 0.0147, "step": 51859 }, { "epoch": 15.96, "learning_rate": 2.074233171595135e-06, "loss": 0.0118, "step": 51860 }, { "epoch": 15.96, "learning_rate": 2.0739293501836424e-06, "loss": 0.0113, "step": 51861 }, { "epoch": 15.96, "learning_rate": 2.073625548450443e-06, "loss": 0.0136, "step": 51862 }, { "epoch": 15.96, "learning_rate": 2.073321766396298e-06, "loss": 0.005, "step": 51863 }, { "epoch": 15.96, "learning_rate": 2.0730180040219585e-06, "loss": 0.0111, "step": 51864 }, { "epoch": 15.96, "learning_rate": 2.072714261328177e-06, "loss": 0.007, "step": 51865 }, { "epoch": 15.96, "learning_rate": 2.072410538315709e-06, "loss": 0.0052, "step": 51866 }, { "epoch": 15.96, "learning_rate": 2.072106834985309e-06, "loss": 0.0085, "step": 51867 }, { "epoch": 15.96, "learning_rate": 2.07180315133773e-06, "loss": 0.0066, "step": 51868 }, { "epoch": 15.96, "learning_rate": 2.071499487373727e-06, "loss": 0.0068, "step": 51869 }, { "epoch": 15.96, "learning_rate": 2.0711958430940524e-06, "loss": 0.0089, "step": 51870 }, { "epoch": 15.96, "learning_rate": 2.0708922184994617e-06, "loss": 0.0051, "step": 51871 }, { "epoch": 15.96, "learning_rate": 2.070588613590706e-06, "loss": 0.0097, "step": 51872 }, { "epoch": 15.96, "learning_rate": 2.070285028368545e-06, "loss": 0.0064, "step": 51873 }, { "epoch": 15.96, "learning_rate": 2.069981462833727e-06, "loss": 0.009, "step": 51874 }, { "epoch": 15.96, "learning_rate": 2.069677916987006e-06, "loss": 0.0097, "step": 51875 }, { "epoch": 15.96, "learning_rate": 2.0693743908291374e-06, "loss": 0.0075, "step": 51876 }, { "epoch": 15.96, "learning_rate": 2.069070884360872e-06, "loss": 0.0083, "step": 51877 }, { "epoch": 15.96, "learning_rate": 2.068767397582967e-06, "loss": 0.012, "step": 51878 }, { "epoch": 15.96, "learning_rate": 2.068463930496175e-06, "loss": 0.0092, "step": 51879 }, { "epoch": 15.96, "learning_rate": 2.068160483101249e-06, "loss": 0.0053, "step": 51880 }, { "epoch": 15.96, "learning_rate": 2.067857055398944e-06, "loss": 0.0141, "step": 51881 }, { "epoch": 15.96, "learning_rate": 2.067553647390007e-06, "loss": 0.0049, "step": 51882 }, { "epoch": 15.96, "learning_rate": 2.0672502590751985e-06, "loss": 0.0047, "step": 51883 }, { "epoch": 15.96, "learning_rate": 2.066946890455268e-06, "loss": 0.0044, "step": 51884 }, { "epoch": 15.96, "learning_rate": 2.0666435415309703e-06, "loss": 0.0075, "step": 51885 }, { "epoch": 15.96, "learning_rate": 2.0663402123030584e-06, "loss": 0.0169, "step": 51886 }, { "epoch": 15.97, "learning_rate": 2.066036902772285e-06, "loss": 0.0034, "step": 51887 }, { "epoch": 15.97, "learning_rate": 2.0657336129394022e-06, "loss": 0.0092, "step": 51888 }, { "epoch": 15.97, "learning_rate": 2.0654303428051636e-06, "loss": 0.0141, "step": 51889 }, { "epoch": 15.97, "learning_rate": 2.0651270923703228e-06, "loss": 0.0098, "step": 51890 }, { "epoch": 15.97, "learning_rate": 2.064823861635633e-06, "loss": 0.0041, "step": 51891 }, { "epoch": 15.97, "learning_rate": 2.0645206506018424e-06, "loss": 0.005, "step": 51892 }, { "epoch": 15.97, "learning_rate": 2.0642174592697105e-06, "loss": 0.0081, "step": 51893 }, { "epoch": 15.97, "learning_rate": 2.063914287639991e-06, "loss": 0.0114, "step": 51894 }, { "epoch": 15.97, "learning_rate": 2.063611135713429e-06, "loss": 0.0094, "step": 51895 }, { "epoch": 15.97, "learning_rate": 2.0633080034907816e-06, "loss": 0.0077, "step": 51896 }, { "epoch": 15.97, "learning_rate": 2.0630048909727984e-06, "loss": 0.0102, "step": 51897 }, { "epoch": 15.97, "learning_rate": 2.062701798160236e-06, "loss": 0.0107, "step": 51898 }, { "epoch": 15.97, "learning_rate": 2.0623987250538458e-06, "loss": 0.0072, "step": 51899 }, { "epoch": 15.97, "learning_rate": 2.0620956716543793e-06, "loss": 0.0064, "step": 51900 }, { "epoch": 15.97, "learning_rate": 2.0617926379625896e-06, "loss": 0.0078, "step": 51901 }, { "epoch": 15.97, "learning_rate": 2.061489623979229e-06, "loss": 0.0118, "step": 51902 }, { "epoch": 15.97, "learning_rate": 2.06118662970505e-06, "loss": 0.0121, "step": 51903 }, { "epoch": 15.97, "learning_rate": 2.060883655140803e-06, "loss": 0.0143, "step": 51904 }, { "epoch": 15.97, "learning_rate": 2.0605807002872413e-06, "loss": 0.0054, "step": 51905 }, { "epoch": 15.97, "learning_rate": 2.060277765145119e-06, "loss": 0.0103, "step": 51906 }, { "epoch": 15.97, "learning_rate": 2.059974849715185e-06, "loss": 0.0127, "step": 51907 }, { "epoch": 15.97, "learning_rate": 2.0596719539981933e-06, "loss": 0.0188, "step": 51908 }, { "epoch": 15.97, "learning_rate": 2.059369077994896e-06, "loss": 0.008, "step": 51909 }, { "epoch": 15.97, "learning_rate": 2.0590662217060444e-06, "loss": 0.0078, "step": 51910 }, { "epoch": 15.97, "learning_rate": 2.0587633851323906e-06, "loss": 0.0043, "step": 51911 }, { "epoch": 15.97, "learning_rate": 2.0584605682746826e-06, "loss": 0.0083, "step": 51912 }, { "epoch": 15.97, "learning_rate": 2.0581577711336807e-06, "loss": 0.0089, "step": 51913 }, { "epoch": 15.97, "learning_rate": 2.057854993710131e-06, "loss": 0.0065, "step": 51914 }, { "epoch": 15.97, "learning_rate": 2.0575522360047882e-06, "loss": 0.0083, "step": 51915 }, { "epoch": 15.97, "learning_rate": 2.0572494980184e-06, "loss": 0.0105, "step": 51916 }, { "epoch": 15.97, "learning_rate": 2.0569467797517173e-06, "loss": 0.0085, "step": 51917 }, { "epoch": 15.97, "learning_rate": 2.056644081205498e-06, "loss": 0.0064, "step": 51918 }, { "epoch": 15.97, "learning_rate": 2.056341402380488e-06, "loss": 0.0119, "step": 51919 }, { "epoch": 15.98, "learning_rate": 2.056038743277442e-06, "loss": 0.0145, "step": 51920 }, { "epoch": 15.98, "learning_rate": 2.0557361038971102e-06, "loss": 0.0167, "step": 51921 }, { "epoch": 15.98, "learning_rate": 2.0554334842402433e-06, "loss": 0.0098, "step": 51922 }, { "epoch": 15.98, "learning_rate": 2.0551308843075935e-06, "loss": 0.0053, "step": 51923 }, { "epoch": 15.98, "learning_rate": 2.0548283040999117e-06, "loss": 0.0045, "step": 51924 }, { "epoch": 15.98, "learning_rate": 2.05452574361795e-06, "loss": 0.0085, "step": 51925 }, { "epoch": 15.98, "learning_rate": 2.0542232028624585e-06, "loss": 0.008, "step": 51926 }, { "epoch": 15.98, "learning_rate": 2.0539206818341863e-06, "loss": 0.0074, "step": 51927 }, { "epoch": 15.98, "learning_rate": 2.0536181805338916e-06, "loss": 0.0107, "step": 51928 }, { "epoch": 15.98, "learning_rate": 2.0533156989623182e-06, "loss": 0.0065, "step": 51929 }, { "epoch": 15.98, "learning_rate": 2.0530132371202195e-06, "loss": 0.0084, "step": 51930 }, { "epoch": 15.98, "learning_rate": 2.052710795008347e-06, "loss": 0.011, "step": 51931 }, { "epoch": 15.98, "learning_rate": 2.0524083726274468e-06, "loss": 0.0058, "step": 51932 }, { "epoch": 15.98, "learning_rate": 2.0521059699782788e-06, "loss": 0.0048, "step": 51933 }, { "epoch": 15.98, "learning_rate": 2.0518035870615873e-06, "loss": 0.0061, "step": 51934 }, { "epoch": 15.98, "learning_rate": 2.0515012238781263e-06, "loss": 0.0111, "step": 51935 }, { "epoch": 15.98, "learning_rate": 2.0511988804286466e-06, "loss": 0.0047, "step": 51936 }, { "epoch": 15.98, "learning_rate": 2.0508965567138916e-06, "loss": 0.0129, "step": 51937 }, { "epoch": 15.98, "learning_rate": 2.05059425273462e-06, "loss": 0.0091, "step": 51938 }, { "epoch": 15.98, "learning_rate": 2.05029196849158e-06, "loss": 0.0066, "step": 51939 }, { "epoch": 15.98, "learning_rate": 2.049989703985523e-06, "loss": 0.0077, "step": 51940 }, { "epoch": 15.98, "learning_rate": 2.0496874592171966e-06, "loss": 0.0058, "step": 51941 }, { "epoch": 15.98, "learning_rate": 2.049385234187354e-06, "loss": 0.0052, "step": 51942 }, { "epoch": 15.98, "learning_rate": 2.0490830288967443e-06, "loss": 0.0099, "step": 51943 }, { "epoch": 15.98, "learning_rate": 2.0487808433461174e-06, "loss": 0.0073, "step": 51944 }, { "epoch": 15.98, "learning_rate": 2.0484786775362252e-06, "loss": 0.009, "step": 51945 }, { "epoch": 15.98, "learning_rate": 2.048176531467816e-06, "loss": 0.0075, "step": 51946 }, { "epoch": 15.98, "learning_rate": 2.0478744051416376e-06, "loss": 0.0063, "step": 51947 }, { "epoch": 15.98, "learning_rate": 2.047572298558447e-06, "loss": 0.0062, "step": 51948 }, { "epoch": 15.98, "learning_rate": 2.0472702117189924e-06, "loss": 0.0083, "step": 51949 }, { "epoch": 15.98, "learning_rate": 2.0469681446240186e-06, "loss": 0.0072, "step": 51950 }, { "epoch": 15.98, "learning_rate": 2.0466660972742803e-06, "loss": 0.0078, "step": 51951 }, { "epoch": 15.99, "learning_rate": 2.0463640696705223e-06, "loss": 0.0068, "step": 51952 }, { "epoch": 15.99, "learning_rate": 2.0460620618135007e-06, "loss": 0.0093, "step": 51953 }, { "epoch": 15.99, "learning_rate": 2.045760073703963e-06, "loss": 0.0062, "step": 51954 }, { "epoch": 15.99, "learning_rate": 2.0454581053426592e-06, "loss": 0.01, "step": 51955 }, { "epoch": 15.99, "learning_rate": 2.0451561567303378e-06, "loss": 0.0108, "step": 51956 }, { "epoch": 15.99, "learning_rate": 2.0448542278677486e-06, "loss": 0.0103, "step": 51957 }, { "epoch": 15.99, "learning_rate": 2.0445523187556427e-06, "loss": 0.0126, "step": 51958 }, { "epoch": 15.99, "learning_rate": 2.0442504293947683e-06, "loss": 0.0064, "step": 51959 }, { "epoch": 15.99, "learning_rate": 2.0439485597858756e-06, "loss": 0.0058, "step": 51960 }, { "epoch": 15.99, "learning_rate": 2.043646709929713e-06, "loss": 0.0079, "step": 51961 }, { "epoch": 15.99, "learning_rate": 2.0433448798270305e-06, "loss": 0.0206, "step": 51962 }, { "epoch": 15.99, "learning_rate": 2.0430430694785787e-06, "loss": 0.0114, "step": 51963 }, { "epoch": 15.99, "learning_rate": 2.0427412788851053e-06, "loss": 0.0077, "step": 51964 }, { "epoch": 15.99, "learning_rate": 2.0424395080473604e-06, "loss": 0.0094, "step": 51965 }, { "epoch": 15.99, "learning_rate": 2.0421377569660915e-06, "loss": 0.006, "step": 51966 }, { "epoch": 15.99, "learning_rate": 2.041836025642048e-06, "loss": 0.0156, "step": 51967 }, { "epoch": 15.99, "learning_rate": 2.0415343140759823e-06, "loss": 0.0093, "step": 51968 }, { "epoch": 15.99, "learning_rate": 2.041232622268642e-06, "loss": 0.0041, "step": 51969 }, { "epoch": 15.99, "learning_rate": 2.0409309502207775e-06, "loss": 0.0097, "step": 51970 }, { "epoch": 15.99, "learning_rate": 2.0406292979331333e-06, "loss": 0.0081, "step": 51971 }, { "epoch": 15.99, "learning_rate": 2.0403276654064573e-06, "loss": 0.0106, "step": 51972 }, { "epoch": 15.99, "learning_rate": 2.040026052641506e-06, "loss": 0.0039, "step": 51973 }, { "epoch": 15.99, "learning_rate": 2.0397244596390233e-06, "loss": 0.0099, "step": 51974 }, { "epoch": 15.99, "learning_rate": 2.039422886399759e-06, "loss": 0.0043, "step": 51975 }, { "epoch": 15.99, "learning_rate": 2.0391213329244606e-06, "loss": 0.008, "step": 51976 }, { "epoch": 15.99, "learning_rate": 2.038819799213878e-06, "loss": 0.0158, "step": 51977 }, { "epoch": 15.99, "learning_rate": 2.0385182852687604e-06, "loss": 0.0056, "step": 51978 }, { "epoch": 15.99, "learning_rate": 2.0382167910898552e-06, "loss": 0.0043, "step": 51979 }, { "epoch": 15.99, "learning_rate": 2.037915316677912e-06, "loss": 0.0111, "step": 51980 }, { "epoch": 15.99, "learning_rate": 2.037613862033677e-06, "loss": 0.011, "step": 51981 }, { "epoch": 15.99, "learning_rate": 2.037312427157898e-06, "loss": 0.0054, "step": 51982 }, { "epoch": 15.99, "learning_rate": 2.037011012051332e-06, "loss": 0.0041, "step": 51983 }, { "epoch": 15.99, "learning_rate": 2.036709616714717e-06, "loss": 0.0058, "step": 51984 }, { "epoch": 16.0, "learning_rate": 2.036408241148806e-06, "loss": 0.0086, "step": 51985 }, { "epoch": 16.0, "learning_rate": 2.0361068853543462e-06, "loss": 0.0118, "step": 51986 }, { "epoch": 16.0, "learning_rate": 2.035805549332083e-06, "loss": 0.0071, "step": 51987 }, { "epoch": 16.0, "learning_rate": 2.03550423308277e-06, "loss": 0.0045, "step": 51988 }, { "epoch": 16.0, "learning_rate": 2.0352029366071537e-06, "loss": 0.0081, "step": 51989 }, { "epoch": 16.0, "learning_rate": 2.0349016599059813e-06, "loss": 0.0033, "step": 51990 }, { "epoch": 16.0, "learning_rate": 2.0346004029800027e-06, "loss": 0.0084, "step": 51991 }, { "epoch": 16.0, "learning_rate": 2.0342991658299584e-06, "loss": 0.0077, "step": 51992 }, { "epoch": 16.0, "learning_rate": 2.033997948456605e-06, "loss": 0.0083, "step": 51993 }, { "epoch": 16.0, "learning_rate": 2.0336967508606875e-06, "loss": 0.0101, "step": 51994 }, { "epoch": 16.0, "learning_rate": 2.033395573042952e-06, "loss": 0.0104, "step": 51995 }, { "epoch": 16.0, "learning_rate": 2.0330944150041486e-06, "loss": 0.0054, "step": 51996 }, { "epoch": 16.0, "learning_rate": 2.0327932767450244e-06, "loss": 0.0064, "step": 51997 }, { "epoch": 16.0, "learning_rate": 2.0324921582663247e-06, "loss": 0.0059, "step": 51998 }, { "epoch": 16.0, "learning_rate": 2.032191059568801e-06, "loss": 0.0072, "step": 51999 }, { "epoch": 16.0, "learning_rate": 2.0318899806531976e-06, "loss": 0.0044, "step": 52000 }, { "epoch": 16.0, "learning_rate": 2.0315889215202643e-06, "loss": 0.0048, "step": 52001 }, { "epoch": 16.0, "learning_rate": 2.031287882170744e-06, "loss": 0.0055, "step": 52002 }, { "epoch": 16.0, "learning_rate": 2.0309868626053906e-06, "loss": 0.0088, "step": 52003 }, { "epoch": 16.0, "learning_rate": 2.0306858628249504e-06, "loss": 0.0052, "step": 52004 }, { "epoch": 16.0, "learning_rate": 2.0303848828301678e-06, "loss": 0.0037, "step": 52005 }, { "epoch": 16.0, "learning_rate": 2.0300839226217893e-06, "loss": 0.0073, "step": 52006 }, { "epoch": 16.0, "learning_rate": 2.0297829822005612e-06, "loss": 0.0044, "step": 52007 }, { "epoch": 16.0, "learning_rate": 2.029482061567237e-06, "loss": 0.0048, "step": 52008 }, { "epoch": 16.0, "learning_rate": 2.0291811607225596e-06, "loss": 0.0065, "step": 52009 }, { "epoch": 16.0, "learning_rate": 2.028880279667276e-06, "loss": 0.0048, "step": 52010 }, { "epoch": 16.0, "learning_rate": 2.0285794184021344e-06, "loss": 0.0063, "step": 52011 }, { "epoch": 16.0, "learning_rate": 2.02827857692788e-06, "loss": 0.0057, "step": 52012 }, { "epoch": 16.0, "learning_rate": 2.0279777552452616e-06, "loss": 0.006, "step": 52013 }, { "epoch": 16.0, "learning_rate": 2.0276769533550256e-06, "loss": 0.0049, "step": 52014 }, { "epoch": 16.0, "learning_rate": 2.0273761712579176e-06, "loss": 0.0053, "step": 52015 }, { "epoch": 16.0, "learning_rate": 2.0270754089546864e-06, "loss": 0.0103, "step": 52016 }, { "epoch": 16.0, "learning_rate": 2.0267746664460743e-06, "loss": 0.007, "step": 52017 }, { "epoch": 16.01, "learning_rate": 2.0264739437328364e-06, "loss": 0.0042, "step": 52018 }, { "epoch": 16.01, "learning_rate": 2.026173240815711e-06, "loss": 0.0065, "step": 52019 }, { "epoch": 16.01, "learning_rate": 2.0258725576954473e-06, "loss": 0.0054, "step": 52020 }, { "epoch": 16.01, "learning_rate": 2.025571894372794e-06, "loss": 0.0073, "step": 52021 }, { "epoch": 16.01, "learning_rate": 2.0252712508484915e-06, "loss": 0.0058, "step": 52022 }, { "epoch": 16.01, "learning_rate": 2.024970627123295e-06, "loss": 0.0045, "step": 52023 }, { "epoch": 16.01, "learning_rate": 2.0246700231979453e-06, "loss": 0.0056, "step": 52024 }, { "epoch": 16.01, "learning_rate": 2.024369439073193e-06, "loss": 0.0071, "step": 52025 }, { "epoch": 16.01, "learning_rate": 2.0240688747497772e-06, "loss": 0.006, "step": 52026 }, { "epoch": 16.01, "learning_rate": 2.023768330228447e-06, "loss": 0.0061, "step": 52027 }, { "epoch": 16.01, "learning_rate": 2.023467805509952e-06, "loss": 0.0068, "step": 52028 }, { "epoch": 16.01, "learning_rate": 2.023167300595036e-06, "loss": 0.0041, "step": 52029 }, { "epoch": 16.01, "learning_rate": 2.0228668154844446e-06, "loss": 0.0116, "step": 52030 }, { "epoch": 16.01, "learning_rate": 2.022566350178925e-06, "loss": 0.0069, "step": 52031 }, { "epoch": 16.01, "learning_rate": 2.0222659046792227e-06, "loss": 0.007, "step": 52032 }, { "epoch": 16.01, "learning_rate": 2.021965478986082e-06, "loss": 0.0052, "step": 52033 }, { "epoch": 16.01, "learning_rate": 2.0216650731002517e-06, "loss": 0.0033, "step": 52034 }, { "epoch": 16.01, "learning_rate": 2.0213646870224758e-06, "loss": 0.0066, "step": 52035 }, { "epoch": 16.01, "learning_rate": 2.0210643207534995e-06, "loss": 0.0054, "step": 52036 }, { "epoch": 16.01, "learning_rate": 2.0207639742940676e-06, "loss": 0.0014, "step": 52037 }, { "epoch": 16.01, "learning_rate": 2.0204636476449304e-06, "loss": 0.0073, "step": 52038 }, { "epoch": 16.01, "learning_rate": 2.0201633408068333e-06, "loss": 0.0067, "step": 52039 }, { "epoch": 16.01, "learning_rate": 2.0198630537805163e-06, "loss": 0.0037, "step": 52040 }, { "epoch": 16.01, "learning_rate": 2.019562786566729e-06, "loss": 0.0057, "step": 52041 }, { "epoch": 16.01, "learning_rate": 2.0192625391662125e-06, "loss": 0.0049, "step": 52042 }, { "epoch": 16.01, "learning_rate": 2.0189623115797176e-06, "loss": 0.0083, "step": 52043 }, { "epoch": 16.01, "learning_rate": 2.018662103807988e-06, "loss": 0.0083, "step": 52044 }, { "epoch": 16.01, "learning_rate": 2.0183619158517687e-06, "loss": 0.0086, "step": 52045 }, { "epoch": 16.01, "learning_rate": 2.018061747711808e-06, "loss": 0.0069, "step": 52046 }, { "epoch": 16.01, "learning_rate": 2.017761599388842e-06, "loss": 0.01, "step": 52047 }, { "epoch": 16.01, "learning_rate": 2.017461470883626e-06, "loss": 0.0065, "step": 52048 }, { "epoch": 16.01, "learning_rate": 2.0171613621968998e-06, "loss": 0.0044, "step": 52049 }, { "epoch": 16.02, "learning_rate": 2.0168612733294113e-06, "loss": 0.0048, "step": 52050 }, { "epoch": 16.02, "learning_rate": 2.016561204281903e-06, "loss": 0.0079, "step": 52051 }, { "epoch": 16.02, "learning_rate": 2.0162611550551214e-06, "loss": 0.0059, "step": 52052 }, { "epoch": 16.02, "learning_rate": 2.01596112564981e-06, "loss": 0.0048, "step": 52053 }, { "epoch": 16.02, "learning_rate": 2.015661116066716e-06, "loss": 0.006, "step": 52054 }, { "epoch": 16.02, "learning_rate": 2.015361126306582e-06, "loss": 0.004, "step": 52055 }, { "epoch": 16.02, "learning_rate": 2.0150611563701548e-06, "loss": 0.004, "step": 52056 }, { "epoch": 16.02, "learning_rate": 2.0147612062581746e-06, "loss": 0.0056, "step": 52057 }, { "epoch": 16.02, "learning_rate": 2.0144612759713924e-06, "loss": 0.0081, "step": 52058 }, { "epoch": 16.02, "learning_rate": 2.014161365510553e-06, "loss": 0.0058, "step": 52059 }, { "epoch": 16.02, "learning_rate": 2.013861474876395e-06, "loss": 0.0081, "step": 52060 }, { "epoch": 16.02, "learning_rate": 2.013561604069666e-06, "loss": 0.0065, "step": 52061 }, { "epoch": 16.02, "learning_rate": 2.0132617530911068e-06, "loss": 0.0084, "step": 52062 }, { "epoch": 16.02, "learning_rate": 2.01296192194147e-06, "loss": 0.0059, "step": 52063 }, { "epoch": 16.02, "learning_rate": 2.012662110621495e-06, "loss": 0.009, "step": 52064 }, { "epoch": 16.02, "learning_rate": 2.012362319131925e-06, "loss": 0.0087, "step": 52065 }, { "epoch": 16.02, "learning_rate": 2.012062547473508e-06, "loss": 0.0057, "step": 52066 }, { "epoch": 16.02, "learning_rate": 2.0117627956469856e-06, "loss": 0.0035, "step": 52067 }, { "epoch": 16.02, "learning_rate": 2.011463063653103e-06, "loss": 0.0073, "step": 52068 }, { "epoch": 16.02, "learning_rate": 2.011163351492603e-06, "loss": 0.0082, "step": 52069 }, { "epoch": 16.02, "learning_rate": 2.010863659166231e-06, "loss": 0.0058, "step": 52070 }, { "epoch": 16.02, "learning_rate": 2.0105639866747306e-06, "loss": 0.006, "step": 52071 }, { "epoch": 16.02, "learning_rate": 2.0102643340188444e-06, "loss": 0.0062, "step": 52072 }, { "epoch": 16.02, "learning_rate": 2.0099647011993217e-06, "loss": 0.0068, "step": 52073 }, { "epoch": 16.02, "learning_rate": 2.009665088216901e-06, "loss": 0.007, "step": 52074 }, { "epoch": 16.02, "learning_rate": 2.0093654950723274e-06, "loss": 0.0051, "step": 52075 }, { "epoch": 16.02, "learning_rate": 2.0090659217663445e-06, "loss": 0.0101, "step": 52076 }, { "epoch": 16.02, "learning_rate": 2.0087663682996948e-06, "loss": 0.0049, "step": 52077 }, { "epoch": 16.02, "learning_rate": 2.008466834673126e-06, "loss": 0.0071, "step": 52078 }, { "epoch": 16.02, "learning_rate": 2.00816732088738e-06, "loss": 0.0097, "step": 52079 }, { "epoch": 16.02, "learning_rate": 2.0078678269432028e-06, "loss": 0.0105, "step": 52080 }, { "epoch": 16.02, "learning_rate": 2.0075683528413316e-06, "loss": 0.0086, "step": 52081 }, { "epoch": 16.02, "learning_rate": 2.0072688985825116e-06, "loss": 0.0018, "step": 52082 }, { "epoch": 16.03, "learning_rate": 2.006969464167491e-06, "loss": 0.0069, "step": 52083 }, { "epoch": 16.03, "learning_rate": 2.0066700495970094e-06, "loss": 0.0048, "step": 52084 }, { "epoch": 16.03, "learning_rate": 2.006370654871812e-06, "loss": 0.0045, "step": 52085 }, { "epoch": 16.03, "learning_rate": 2.0060712799926407e-06, "loss": 0.0058, "step": 52086 }, { "epoch": 16.03, "learning_rate": 2.0057719249602403e-06, "loss": 0.0027, "step": 52087 }, { "epoch": 16.03, "learning_rate": 2.0054725897753525e-06, "loss": 0.0121, "step": 52088 }, { "epoch": 16.03, "learning_rate": 2.00517327443872e-06, "loss": 0.0064, "step": 52089 }, { "epoch": 16.03, "learning_rate": 2.0048739789510883e-06, "loss": 0.0121, "step": 52090 }, { "epoch": 16.03, "learning_rate": 2.004574703313198e-06, "loss": 0.0057, "step": 52091 }, { "epoch": 16.03, "learning_rate": 2.004275447525791e-06, "loss": 0.0045, "step": 52092 }, { "epoch": 16.03, "learning_rate": 2.0039762115896154e-06, "loss": 0.0049, "step": 52093 }, { "epoch": 16.03, "learning_rate": 2.0036769955054136e-06, "loss": 0.0032, "step": 52094 }, { "epoch": 16.03, "learning_rate": 2.0033777992739235e-06, "loss": 0.0072, "step": 52095 }, { "epoch": 16.03, "learning_rate": 2.0030786228958908e-06, "loss": 0.0139, "step": 52096 }, { "epoch": 16.03, "learning_rate": 2.0027794663720545e-06, "loss": 0.0036, "step": 52097 }, { "epoch": 16.03, "learning_rate": 2.002480329703165e-06, "loss": 0.0061, "step": 52098 }, { "epoch": 16.03, "learning_rate": 2.00218121288996e-06, "loss": 0.0043, "step": 52099 }, { "epoch": 16.03, "learning_rate": 2.0018821159331824e-06, "loss": 0.0041, "step": 52100 }, { "epoch": 16.03, "learning_rate": 2.001583038833578e-06, "loss": 0.0049, "step": 52101 }, { "epoch": 16.03, "learning_rate": 2.0012839815918817e-06, "loss": 0.006, "step": 52102 }, { "epoch": 16.03, "learning_rate": 2.0009849442088423e-06, "loss": 0.007, "step": 52103 }, { "epoch": 16.03, "learning_rate": 2.000685926685203e-06, "loss": 0.0041, "step": 52104 }, { "epoch": 16.03, "learning_rate": 2.0003869290217025e-06, "loss": 0.007, "step": 52105 }, { "epoch": 16.03, "learning_rate": 2.0000879512190852e-06, "loss": 0.0047, "step": 52106 }, { "epoch": 16.03, "learning_rate": 1.9997889932780924e-06, "loss": 0.0097, "step": 52107 }, { "epoch": 16.03, "learning_rate": 1.9994900551994666e-06, "loss": 0.0064, "step": 52108 }, { "epoch": 16.03, "learning_rate": 1.99919113698395e-06, "loss": 0.0047, "step": 52109 }, { "epoch": 16.03, "learning_rate": 1.998892238632286e-06, "loss": 0.006, "step": 52110 }, { "epoch": 16.03, "learning_rate": 1.9985933601452143e-06, "loss": 0.0042, "step": 52111 }, { "epoch": 16.03, "learning_rate": 1.998294501523477e-06, "loss": 0.0075, "step": 52112 }, { "epoch": 16.03, "learning_rate": 1.9979956627678188e-06, "loss": 0.005, "step": 52113 }, { "epoch": 16.03, "learning_rate": 1.99769684387898e-06, "loss": 0.0045, "step": 52114 }, { "epoch": 16.04, "learning_rate": 1.997398044857706e-06, "loss": 0.0178, "step": 52115 }, { "epoch": 16.04, "learning_rate": 1.9970992657047315e-06, "loss": 0.0036, "step": 52116 }, { "epoch": 16.04, "learning_rate": 1.9968005064207996e-06, "loss": 0.0053, "step": 52117 }, { "epoch": 16.04, "learning_rate": 1.9965017670066577e-06, "loss": 0.0086, "step": 52118 }, { "epoch": 16.04, "learning_rate": 1.9962030474630443e-06, "loss": 0.008, "step": 52119 }, { "epoch": 16.04, "learning_rate": 1.9959043477907e-06, "loss": 0.0073, "step": 52120 }, { "epoch": 16.04, "learning_rate": 1.9956056679903678e-06, "loss": 0.0036, "step": 52121 }, { "epoch": 16.04, "learning_rate": 1.99530700806279e-06, "loss": 0.005, "step": 52122 }, { "epoch": 16.04, "learning_rate": 1.995008368008705e-06, "loss": 0.0048, "step": 52123 }, { "epoch": 16.04, "learning_rate": 1.9947097478288578e-06, "loss": 0.0059, "step": 52124 }, { "epoch": 16.04, "learning_rate": 1.994411147523987e-06, "loss": 0.0077, "step": 52125 }, { "epoch": 16.04, "learning_rate": 1.994112567094836e-06, "loss": 0.0065, "step": 52126 }, { "epoch": 16.04, "learning_rate": 1.993814006542142e-06, "loss": 0.0085, "step": 52127 }, { "epoch": 16.04, "learning_rate": 1.9935154658666546e-06, "loss": 0.0073, "step": 52128 }, { "epoch": 16.04, "learning_rate": 1.9932169450691075e-06, "loss": 0.0025, "step": 52129 }, { "epoch": 16.04, "learning_rate": 1.9929184441502437e-06, "loss": 0.0047, "step": 52130 }, { "epoch": 16.04, "learning_rate": 1.9926199631108045e-06, "loss": 0.0049, "step": 52131 }, { "epoch": 16.04, "learning_rate": 1.9923215019515296e-06, "loss": 0.0105, "step": 52132 }, { "epoch": 16.04, "learning_rate": 1.9920230606731637e-06, "loss": 0.0049, "step": 52133 }, { "epoch": 16.04, "learning_rate": 1.9917246392764453e-06, "loss": 0.0039, "step": 52134 }, { "epoch": 16.04, "learning_rate": 1.991426237762116e-06, "loss": 0.0032, "step": 52135 }, { "epoch": 16.04, "learning_rate": 1.991127856130919e-06, "loss": 0.0094, "step": 52136 }, { "epoch": 16.04, "learning_rate": 1.990829494383587e-06, "loss": 0.0041, "step": 52137 }, { "epoch": 16.04, "learning_rate": 1.990531152520869e-06, "loss": 0.0087, "step": 52138 }, { "epoch": 16.04, "learning_rate": 1.9902328305435024e-06, "loss": 0.0063, "step": 52139 }, { "epoch": 16.04, "learning_rate": 1.9899345284522297e-06, "loss": 0.0044, "step": 52140 }, { "epoch": 16.04, "learning_rate": 1.9896362462477893e-06, "loss": 0.0068, "step": 52141 }, { "epoch": 16.04, "learning_rate": 1.9893379839309222e-06, "loss": 0.0049, "step": 52142 }, { "epoch": 16.04, "learning_rate": 1.989039741502369e-06, "loss": 0.0088, "step": 52143 }, { "epoch": 16.04, "learning_rate": 1.9887415189628712e-06, "loss": 0.0067, "step": 52144 }, { "epoch": 16.04, "learning_rate": 1.988443316313168e-06, "loss": 0.007, "step": 52145 }, { "epoch": 16.04, "learning_rate": 1.9881451335540004e-06, "loss": 0.003, "step": 52146 }, { "epoch": 16.04, "learning_rate": 1.9878469706861058e-06, "loss": 0.0039, "step": 52147 }, { "epoch": 16.05, "learning_rate": 1.9875488277102293e-06, "loss": 0.0047, "step": 52148 }, { "epoch": 16.05, "learning_rate": 1.9872507046271126e-06, "loss": 0.0083, "step": 52149 }, { "epoch": 16.05, "learning_rate": 1.9869526014374897e-06, "loss": 0.0045, "step": 52150 }, { "epoch": 16.05, "learning_rate": 1.9866545181421016e-06, "loss": 0.0042, "step": 52151 }, { "epoch": 16.05, "learning_rate": 1.986356454741689e-06, "loss": 0.0046, "step": 52152 }, { "epoch": 16.05, "learning_rate": 1.986058411236995e-06, "loss": 0.0048, "step": 52153 }, { "epoch": 16.05, "learning_rate": 1.985760387628757e-06, "loss": 0.0068, "step": 52154 }, { "epoch": 16.05, "learning_rate": 1.9854623839177156e-06, "loss": 0.0041, "step": 52155 }, { "epoch": 16.05, "learning_rate": 1.9851644001046112e-06, "loss": 0.0064, "step": 52156 }, { "epoch": 16.05, "learning_rate": 1.9848664361901827e-06, "loss": 0.0066, "step": 52157 }, { "epoch": 16.05, "learning_rate": 1.9845684921751695e-06, "loss": 0.0052, "step": 52158 }, { "epoch": 16.05, "learning_rate": 1.9842705680603115e-06, "loss": 0.0075, "step": 52159 }, { "epoch": 16.05, "learning_rate": 1.98397266384635e-06, "loss": 0.0094, "step": 52160 }, { "epoch": 16.05, "learning_rate": 1.9836747795340215e-06, "loss": 0.0085, "step": 52161 }, { "epoch": 16.05, "learning_rate": 1.983376915124069e-06, "loss": 0.0043, "step": 52162 }, { "epoch": 16.05, "learning_rate": 1.9830790706172297e-06, "loss": 0.0073, "step": 52163 }, { "epoch": 16.05, "learning_rate": 1.9827812460142435e-06, "loss": 0.0037, "step": 52164 }, { "epoch": 16.05, "learning_rate": 1.982483441315851e-06, "loss": 0.0034, "step": 52165 }, { "epoch": 16.05, "learning_rate": 1.982185656522789e-06, "loss": 0.0046, "step": 52166 }, { "epoch": 16.05, "learning_rate": 1.9818878916357976e-06, "loss": 0.0032, "step": 52167 }, { "epoch": 16.05, "learning_rate": 1.9815901466556186e-06, "loss": 0.0055, "step": 52168 }, { "epoch": 16.05, "learning_rate": 1.981292421582991e-06, "loss": 0.0067, "step": 52169 }, { "epoch": 16.05, "learning_rate": 1.9809947164186537e-06, "loss": 0.0038, "step": 52170 }, { "epoch": 16.05, "learning_rate": 1.980697031163342e-06, "loss": 0.0055, "step": 52171 }, { "epoch": 16.05, "learning_rate": 1.9803993658177958e-06, "loss": 0.0052, "step": 52172 }, { "epoch": 16.05, "learning_rate": 1.980101720382758e-06, "loss": 0.0053, "step": 52173 }, { "epoch": 16.05, "learning_rate": 1.979804094858966e-06, "loss": 0.0047, "step": 52174 }, { "epoch": 16.05, "learning_rate": 1.9795064892471583e-06, "loss": 0.0092, "step": 52175 }, { "epoch": 16.05, "learning_rate": 1.9792089035480733e-06, "loss": 0.0046, "step": 52176 }, { "epoch": 16.05, "learning_rate": 1.97891133776245e-06, "loss": 0.0061, "step": 52177 }, { "epoch": 16.05, "learning_rate": 1.9786137918910276e-06, "loss": 0.0077, "step": 52178 }, { "epoch": 16.05, "learning_rate": 1.9783162659345456e-06, "loss": 0.0144, "step": 52179 }, { "epoch": 16.06, "learning_rate": 1.978018759893741e-06, "loss": 0.0071, "step": 52180 }, { "epoch": 16.06, "learning_rate": 1.977721273769353e-06, "loss": 0.0078, "step": 52181 }, { "epoch": 16.06, "learning_rate": 1.9774238075621177e-06, "loss": 0.0051, "step": 52182 }, { "epoch": 16.06, "learning_rate": 1.977126361272782e-06, "loss": 0.0077, "step": 52183 }, { "epoch": 16.06, "learning_rate": 1.9768289349020753e-06, "loss": 0.0112, "step": 52184 }, { "epoch": 16.06, "learning_rate": 1.97653152845074e-06, "loss": 0.0031, "step": 52185 }, { "epoch": 16.06, "learning_rate": 1.976234141919513e-06, "loss": 0.006, "step": 52186 }, { "epoch": 16.06, "learning_rate": 1.975936775309132e-06, "loss": 0.0031, "step": 52187 }, { "epoch": 16.06, "learning_rate": 1.9756394286203386e-06, "loss": 0.0034, "step": 52188 }, { "epoch": 16.06, "learning_rate": 1.97534210185387e-06, "loss": 0.0057, "step": 52189 }, { "epoch": 16.06, "learning_rate": 1.9750447950104635e-06, "loss": 0.0055, "step": 52190 }, { "epoch": 16.06, "learning_rate": 1.9747475080908586e-06, "loss": 0.0067, "step": 52191 }, { "epoch": 16.06, "learning_rate": 1.9744502410957877e-06, "loss": 0.0111, "step": 52192 }, { "epoch": 16.06, "learning_rate": 1.9741529940259963e-06, "loss": 0.0045, "step": 52193 }, { "epoch": 16.06, "learning_rate": 1.973855766882219e-06, "loss": 0.0078, "step": 52194 }, { "epoch": 16.06, "learning_rate": 1.973558559665194e-06, "loss": 0.0113, "step": 52195 }, { "epoch": 16.06, "learning_rate": 1.973261372375659e-06, "loss": 0.0096, "step": 52196 }, { "epoch": 16.06, "learning_rate": 1.9729642050143526e-06, "loss": 0.0037, "step": 52197 }, { "epoch": 16.06, "learning_rate": 1.9726670575820116e-06, "loss": 0.004, "step": 52198 }, { "epoch": 16.06, "learning_rate": 1.972369930079374e-06, "loss": 0.0036, "step": 52199 }, { "epoch": 16.06, "learning_rate": 1.9720728225071784e-06, "loss": 0.0096, "step": 52200 }, { "epoch": 16.06, "learning_rate": 1.9717757348661617e-06, "loss": 0.0076, "step": 52201 }, { "epoch": 16.06, "learning_rate": 1.9714786671570584e-06, "loss": 0.0065, "step": 52202 }, { "epoch": 16.06, "learning_rate": 1.971181619380611e-06, "loss": 0.0045, "step": 52203 }, { "epoch": 16.06, "learning_rate": 1.9708845915375595e-06, "loss": 0.0039, "step": 52204 }, { "epoch": 16.06, "learning_rate": 1.9705875836286337e-06, "loss": 0.0044, "step": 52205 }, { "epoch": 16.06, "learning_rate": 1.970290595654574e-06, "loss": 0.007, "step": 52206 }, { "epoch": 16.06, "learning_rate": 1.969993627616115e-06, "loss": 0.007, "step": 52207 }, { "epoch": 16.06, "learning_rate": 1.9696966795140006e-06, "loss": 0.0029, "step": 52208 }, { "epoch": 16.06, "learning_rate": 1.969399751348964e-06, "loss": 0.0058, "step": 52209 }, { "epoch": 16.06, "learning_rate": 1.9691028431217417e-06, "loss": 0.0038, "step": 52210 }, { "epoch": 16.06, "learning_rate": 1.968805954833073e-06, "loss": 0.0084, "step": 52211 }, { "epoch": 16.06, "learning_rate": 1.9685090864836944e-06, "loss": 0.0051, "step": 52212 }, { "epoch": 16.07, "learning_rate": 1.9682122380743417e-06, "loss": 0.0102, "step": 52213 }, { "epoch": 16.07, "learning_rate": 1.967915409605754e-06, "loss": 0.0021, "step": 52214 }, { "epoch": 16.07, "learning_rate": 1.9676186010786657e-06, "loss": 0.0075, "step": 52215 }, { "epoch": 16.07, "learning_rate": 1.967321812493813e-06, "loss": 0.0028, "step": 52216 }, { "epoch": 16.07, "learning_rate": 1.967025043851939e-06, "loss": 0.0082, "step": 52217 }, { "epoch": 16.07, "learning_rate": 1.9667282951537746e-06, "loss": 0.0058, "step": 52218 }, { "epoch": 16.07, "learning_rate": 1.9664315664000578e-06, "loss": 0.0041, "step": 52219 }, { "epoch": 16.07, "learning_rate": 1.9661348575915253e-06, "loss": 0.0023, "step": 52220 }, { "epoch": 16.07, "learning_rate": 1.965838168728912e-06, "loss": 0.0115, "step": 52221 }, { "epoch": 16.07, "learning_rate": 1.9655414998129584e-06, "loss": 0.0068, "step": 52222 }, { "epoch": 16.07, "learning_rate": 1.9652448508444e-06, "loss": 0.0054, "step": 52223 }, { "epoch": 16.07, "learning_rate": 1.9649482218239724e-06, "loss": 0.0093, "step": 52224 }, { "epoch": 16.07, "learning_rate": 1.964651612752415e-06, "loss": 0.0068, "step": 52225 }, { "epoch": 16.07, "learning_rate": 1.9643550236304553e-06, "loss": 0.003, "step": 52226 }, { "epoch": 16.07, "learning_rate": 1.9640584544588393e-06, "loss": 0.004, "step": 52227 }, { "epoch": 16.07, "learning_rate": 1.9637619052382995e-06, "loss": 0.006, "step": 52228 }, { "epoch": 16.07, "learning_rate": 1.963465375969572e-06, "loss": 0.0055, "step": 52229 }, { "epoch": 16.07, "learning_rate": 1.9631688666533945e-06, "loss": 0.0064, "step": 52230 }, { "epoch": 16.07, "learning_rate": 1.9628723772905e-06, "loss": 0.007, "step": 52231 }, { "epoch": 16.07, "learning_rate": 1.9625759078816288e-06, "loss": 0.0053, "step": 52232 }, { "epoch": 16.07, "learning_rate": 1.962279458427514e-06, "loss": 0.0034, "step": 52233 }, { "epoch": 16.07, "learning_rate": 1.961983028928892e-06, "loss": 0.0078, "step": 52234 }, { "epoch": 16.07, "learning_rate": 1.961686619386499e-06, "loss": 0.0061, "step": 52235 }, { "epoch": 16.07, "learning_rate": 1.961390229801069e-06, "loss": 0.0055, "step": 52236 }, { "epoch": 16.07, "learning_rate": 1.961093860173342e-06, "loss": 0.0074, "step": 52237 }, { "epoch": 16.07, "learning_rate": 1.9607975105040545e-06, "loss": 0.0052, "step": 52238 }, { "epoch": 16.07, "learning_rate": 1.9605011807939377e-06, "loss": 0.0053, "step": 52239 }, { "epoch": 16.07, "learning_rate": 1.9602048710437284e-06, "loss": 0.0053, "step": 52240 }, { "epoch": 16.07, "learning_rate": 1.959908581254161e-06, "loss": 0.0053, "step": 52241 }, { "epoch": 16.07, "learning_rate": 1.9596123114259747e-06, "loss": 0.0068, "step": 52242 }, { "epoch": 16.07, "learning_rate": 1.9593160615599038e-06, "loss": 0.0044, "step": 52243 }, { "epoch": 16.07, "learning_rate": 1.9590198316566833e-06, "loss": 0.0052, "step": 52244 }, { "epoch": 16.08, "learning_rate": 1.958723621717048e-06, "loss": 0.0053, "step": 52245 }, { "epoch": 16.08, "learning_rate": 1.958427431741735e-06, "loss": 0.0077, "step": 52246 }, { "epoch": 16.08, "learning_rate": 1.9581312617314796e-06, "loss": 0.0029, "step": 52247 }, { "epoch": 16.08, "learning_rate": 1.9578351116870154e-06, "loss": 0.0054, "step": 52248 }, { "epoch": 16.08, "learning_rate": 1.9575389816090774e-06, "loss": 0.0101, "step": 52249 }, { "epoch": 16.08, "learning_rate": 1.9572428714984038e-06, "loss": 0.009, "step": 52250 }, { "epoch": 16.08, "learning_rate": 1.9569467813557242e-06, "loss": 0.0085, "step": 52251 }, { "epoch": 16.08, "learning_rate": 1.9566507111817825e-06, "loss": 0.003, "step": 52252 }, { "epoch": 16.08, "learning_rate": 1.9563546609773065e-06, "loss": 0.0048, "step": 52253 }, { "epoch": 16.08, "learning_rate": 1.956058630743034e-06, "loss": 0.0046, "step": 52254 }, { "epoch": 16.08, "learning_rate": 1.955762620479699e-06, "loss": 0.0059, "step": 52255 }, { "epoch": 16.08, "learning_rate": 1.9554666301880333e-06, "loss": 0.0048, "step": 52256 }, { "epoch": 16.08, "learning_rate": 1.955170659868778e-06, "loss": 0.0114, "step": 52257 }, { "epoch": 16.08, "learning_rate": 1.954874709522665e-06, "loss": 0.0097, "step": 52258 }, { "epoch": 16.08, "learning_rate": 1.9545787791504323e-06, "loss": 0.0058, "step": 52259 }, { "epoch": 16.08, "learning_rate": 1.954282868752809e-06, "loss": 0.0055, "step": 52260 }, { "epoch": 16.08, "learning_rate": 1.95398697833053e-06, "loss": 0.0063, "step": 52261 }, { "epoch": 16.08, "learning_rate": 1.9536911078843346e-06, "loss": 0.0092, "step": 52262 }, { "epoch": 16.08, "learning_rate": 1.953395257414954e-06, "loss": 0.0054, "step": 52263 }, { "epoch": 16.08, "learning_rate": 1.953099426923125e-06, "loss": 0.0045, "step": 52264 }, { "epoch": 16.08, "learning_rate": 1.95280361640958e-06, "loss": 0.0055, "step": 52265 }, { "epoch": 16.08, "learning_rate": 1.952507825875054e-06, "loss": 0.0036, "step": 52266 }, { "epoch": 16.08, "learning_rate": 1.9522120553202817e-06, "loss": 0.0072, "step": 52267 }, { "epoch": 16.08, "learning_rate": 1.9519163047459978e-06, "loss": 0.0069, "step": 52268 }, { "epoch": 16.08, "learning_rate": 1.9516205741529347e-06, "loss": 0.0053, "step": 52269 }, { "epoch": 16.08, "learning_rate": 1.951324863541829e-06, "loss": 0.0084, "step": 52270 }, { "epoch": 16.08, "learning_rate": 1.951029172913411e-06, "loss": 0.0219, "step": 52271 }, { "epoch": 16.08, "learning_rate": 1.9507335022684205e-06, "loss": 0.0032, "step": 52272 }, { "epoch": 16.08, "learning_rate": 1.9504378516075917e-06, "loss": 0.0053, "step": 52273 }, { "epoch": 16.08, "learning_rate": 1.9501422209316524e-06, "loss": 0.0099, "step": 52274 }, { "epoch": 16.08, "learning_rate": 1.9498466102413394e-06, "loss": 0.0058, "step": 52275 }, { "epoch": 16.08, "learning_rate": 1.9495510195373846e-06, "loss": 0.0035, "step": 52276 }, { "epoch": 16.08, "learning_rate": 1.9492554488205275e-06, "loss": 0.0067, "step": 52277 }, { "epoch": 16.09, "learning_rate": 1.948959898091498e-06, "loss": 0.0041, "step": 52278 }, { "epoch": 16.09, "learning_rate": 1.948664367351031e-06, "loss": 0.005, "step": 52279 }, { "epoch": 16.09, "learning_rate": 1.9483688565998628e-06, "loss": 0.0089, "step": 52280 }, { "epoch": 16.09, "learning_rate": 1.9480733658387175e-06, "loss": 0.0057, "step": 52281 }, { "epoch": 16.09, "learning_rate": 1.9477778950683392e-06, "loss": 0.0034, "step": 52282 }, { "epoch": 16.09, "learning_rate": 1.947482444289457e-06, "loss": 0.0073, "step": 52283 }, { "epoch": 16.09, "learning_rate": 1.9471870135028047e-06, "loss": 0.0038, "step": 52284 }, { "epoch": 16.09, "learning_rate": 1.9468916027091167e-06, "loss": 0.0115, "step": 52285 }, { "epoch": 16.09, "learning_rate": 1.946596211909125e-06, "loss": 0.0041, "step": 52286 }, { "epoch": 16.09, "learning_rate": 1.9463008411035645e-06, "loss": 0.0062, "step": 52287 }, { "epoch": 16.09, "learning_rate": 1.946005490293168e-06, "loss": 0.0095, "step": 52288 }, { "epoch": 16.09, "learning_rate": 1.945710159478669e-06, "loss": 0.0073, "step": 52289 }, { "epoch": 16.09, "learning_rate": 1.9454148486607993e-06, "loss": 0.0054, "step": 52290 }, { "epoch": 16.09, "learning_rate": 1.9451195578402905e-06, "loss": 0.0044, "step": 52291 }, { "epoch": 16.09, "learning_rate": 1.9448242870178814e-06, "loss": 0.0054, "step": 52292 }, { "epoch": 16.09, "learning_rate": 1.9445290361943014e-06, "loss": 0.0058, "step": 52293 }, { "epoch": 16.09, "learning_rate": 1.944233805370287e-06, "loss": 0.0045, "step": 52294 }, { "epoch": 16.09, "learning_rate": 1.943938594546566e-06, "loss": 0.0084, "step": 52295 }, { "epoch": 16.09, "learning_rate": 1.9436434037238705e-06, "loss": 0.0053, "step": 52296 }, { "epoch": 16.09, "learning_rate": 1.94334823290294e-06, "loss": 0.0048, "step": 52297 }, { "epoch": 16.09, "learning_rate": 1.943053082084503e-06, "loss": 0.005, "step": 52298 }, { "epoch": 16.09, "learning_rate": 1.942757951269293e-06, "loss": 0.0055, "step": 52299 }, { "epoch": 16.09, "learning_rate": 1.9424628404580427e-06, "loss": 0.0047, "step": 52300 }, { "epoch": 16.09, "learning_rate": 1.9421677496514857e-06, "loss": 0.0062, "step": 52301 }, { "epoch": 16.09, "learning_rate": 1.9418726788503527e-06, "loss": 0.0069, "step": 52302 }, { "epoch": 16.09, "learning_rate": 1.941577628055379e-06, "loss": 0.0121, "step": 52303 }, { "epoch": 16.09, "learning_rate": 1.941282597267294e-06, "loss": 0.0055, "step": 52304 }, { "epoch": 16.09, "learning_rate": 1.9409875864868332e-06, "loss": 0.0043, "step": 52305 }, { "epoch": 16.09, "learning_rate": 1.940692595714725e-06, "loss": 0.0045, "step": 52306 }, { "epoch": 16.09, "learning_rate": 1.940397624951709e-06, "loss": 0.0047, "step": 52307 }, { "epoch": 16.09, "learning_rate": 1.94010267419851e-06, "loss": 0.0064, "step": 52308 }, { "epoch": 16.09, "learning_rate": 1.939807743455863e-06, "loss": 0.0173, "step": 52309 }, { "epoch": 16.1, "learning_rate": 1.9395128327245005e-06, "loss": 0.0086, "step": 52310 }, { "epoch": 16.1, "learning_rate": 1.9392179420051526e-06, "loss": 0.0043, "step": 52311 }, { "epoch": 16.1, "learning_rate": 1.938923071298555e-06, "loss": 0.0115, "step": 52312 }, { "epoch": 16.1, "learning_rate": 1.9386282206054395e-06, "loss": 0.0052, "step": 52313 }, { "epoch": 16.1, "learning_rate": 1.9383333899265368e-06, "loss": 0.0073, "step": 52314 }, { "epoch": 16.1, "learning_rate": 1.93803857926258e-06, "loss": 0.0051, "step": 52315 }, { "epoch": 16.1, "learning_rate": 1.937743788614296e-06, "loss": 0.0086, "step": 52316 }, { "epoch": 16.1, "learning_rate": 1.9374490179824237e-06, "loss": 0.0046, "step": 52317 }, { "epoch": 16.1, "learning_rate": 1.937154267367691e-06, "loss": 0.0028, "step": 52318 }, { "epoch": 16.1, "learning_rate": 1.936859536770831e-06, "loss": 0.0046, "step": 52319 }, { "epoch": 16.1, "learning_rate": 1.936564826192574e-06, "loss": 0.0033, "step": 52320 }, { "epoch": 16.1, "learning_rate": 1.9362701356336545e-06, "loss": 0.0048, "step": 52321 }, { "epoch": 16.1, "learning_rate": 1.9359754650948004e-06, "loss": 0.0036, "step": 52322 }, { "epoch": 16.1, "learning_rate": 1.935680814576747e-06, "loss": 0.01, "step": 52323 }, { "epoch": 16.1, "learning_rate": 1.9353861840802233e-06, "loss": 0.0111, "step": 52324 }, { "epoch": 16.1, "learning_rate": 1.9350915736059616e-06, "loss": 0.0092, "step": 52325 }, { "epoch": 16.1, "learning_rate": 1.9347969831546897e-06, "loss": 0.0047, "step": 52326 }, { "epoch": 16.1, "learning_rate": 1.934502412727146e-06, "loss": 0.0039, "step": 52327 }, { "epoch": 16.1, "learning_rate": 1.9342078623240612e-06, "loss": 0.0037, "step": 52328 }, { "epoch": 16.1, "learning_rate": 1.933913331946161e-06, "loss": 0.0053, "step": 52329 }, { "epoch": 16.1, "learning_rate": 1.9336188215941796e-06, "loss": 0.0093, "step": 52330 }, { "epoch": 16.1, "learning_rate": 1.9333243312688453e-06, "loss": 0.0039, "step": 52331 }, { "epoch": 16.1, "learning_rate": 1.9330298609708953e-06, "loss": 0.006, "step": 52332 }, { "epoch": 16.1, "learning_rate": 1.9327354107010566e-06, "loss": 0.0059, "step": 52333 }, { "epoch": 16.1, "learning_rate": 1.9324409804600608e-06, "loss": 0.0065, "step": 52334 }, { "epoch": 16.1, "learning_rate": 1.9321465702486396e-06, "loss": 0.0107, "step": 52335 }, { "epoch": 16.1, "learning_rate": 1.931852180067524e-06, "loss": 0.007, "step": 52336 }, { "epoch": 16.1, "learning_rate": 1.931557809917444e-06, "loss": 0.0045, "step": 52337 }, { "epoch": 16.1, "learning_rate": 1.9312634597991296e-06, "loss": 0.0071, "step": 52338 }, { "epoch": 16.1, "learning_rate": 1.930969129713314e-06, "loss": 0.0085, "step": 52339 }, { "epoch": 16.1, "learning_rate": 1.930674819660726e-06, "loss": 0.0057, "step": 52340 }, { "epoch": 16.1, "learning_rate": 1.9303805296420973e-06, "loss": 0.0073, "step": 52341 }, { "epoch": 16.1, "learning_rate": 1.9300862596581583e-06, "loss": 0.0065, "step": 52342 }, { "epoch": 16.11, "learning_rate": 1.929792009709639e-06, "loss": 0.0055, "step": 52343 }, { "epoch": 16.11, "learning_rate": 1.9294977797972715e-06, "loss": 0.0061, "step": 52344 }, { "epoch": 16.11, "learning_rate": 1.929203569921784e-06, "loss": 0.004, "step": 52345 }, { "epoch": 16.11, "learning_rate": 1.9289093800839067e-06, "loss": 0.006, "step": 52346 }, { "epoch": 16.11, "learning_rate": 1.9286152102843735e-06, "loss": 0.0174, "step": 52347 }, { "epoch": 16.11, "learning_rate": 1.928321060523913e-06, "loss": 0.0035, "step": 52348 }, { "epoch": 16.11, "learning_rate": 1.9280269308032575e-06, "loss": 0.0053, "step": 52349 }, { "epoch": 16.11, "learning_rate": 1.927732821123133e-06, "loss": 0.0048, "step": 52350 }, { "epoch": 16.11, "learning_rate": 1.927438731484269e-06, "loss": 0.0027, "step": 52351 }, { "epoch": 16.11, "learning_rate": 1.9271446618874003e-06, "loss": 0.006, "step": 52352 }, { "epoch": 16.11, "learning_rate": 1.9268506123332555e-06, "loss": 0.0042, "step": 52353 }, { "epoch": 16.11, "learning_rate": 1.926556582822564e-06, "loss": 0.0039, "step": 52354 }, { "epoch": 16.11, "learning_rate": 1.926262573356057e-06, "loss": 0.0069, "step": 52355 }, { "epoch": 16.11, "learning_rate": 1.9259685839344623e-06, "loss": 0.0038, "step": 52356 }, { "epoch": 16.11, "learning_rate": 1.9256746145585103e-06, "loss": 0.0046, "step": 52357 }, { "epoch": 16.11, "learning_rate": 1.925380665228933e-06, "loss": 0.0068, "step": 52358 }, { "epoch": 16.11, "learning_rate": 1.9250867359464575e-06, "loss": 0.0031, "step": 52359 }, { "epoch": 16.11, "learning_rate": 1.9247928267118154e-06, "loss": 0.0091, "step": 52360 }, { "epoch": 16.11, "learning_rate": 1.924498937525733e-06, "loss": 0.0074, "step": 52361 }, { "epoch": 16.11, "learning_rate": 1.924205068388948e-06, "loss": 0.0107, "step": 52362 }, { "epoch": 16.11, "learning_rate": 1.923911219302181e-06, "loss": 0.0033, "step": 52363 }, { "epoch": 16.11, "learning_rate": 1.923617390266166e-06, "loss": 0.0061, "step": 52364 }, { "epoch": 16.11, "learning_rate": 1.923323581281632e-06, "loss": 0.0068, "step": 52365 }, { "epoch": 16.11, "learning_rate": 1.923029792349306e-06, "loss": 0.0045, "step": 52366 }, { "epoch": 16.11, "learning_rate": 1.922736023469921e-06, "loss": 0.0047, "step": 52367 }, { "epoch": 16.11, "learning_rate": 1.9224422746442052e-06, "loss": 0.0052, "step": 52368 }, { "epoch": 16.11, "learning_rate": 1.9221485458728874e-06, "loss": 0.0058, "step": 52369 }, { "epoch": 16.11, "learning_rate": 1.921854837156699e-06, "loss": 0.0052, "step": 52370 }, { "epoch": 16.11, "learning_rate": 1.9215611484963635e-06, "loss": 0.0099, "step": 52371 }, { "epoch": 16.11, "learning_rate": 1.9212674798926156e-06, "loss": 0.0036, "step": 52372 }, { "epoch": 16.11, "learning_rate": 1.9209738313461836e-06, "loss": 0.0052, "step": 52373 }, { "epoch": 16.11, "learning_rate": 1.920680202857794e-06, "loss": 0.0054, "step": 52374 }, { "epoch": 16.12, "learning_rate": 1.9203865944281787e-06, "loss": 0.0074, "step": 52375 }, { "epoch": 16.12, "learning_rate": 1.920093006058065e-06, "loss": 0.0066, "step": 52376 }, { "epoch": 16.12, "learning_rate": 1.919799437748181e-06, "loss": 0.0101, "step": 52377 }, { "epoch": 16.12, "learning_rate": 1.919505889499257e-06, "loss": 0.0055, "step": 52378 }, { "epoch": 16.12, "learning_rate": 1.9192123613120216e-06, "loss": 0.0054, "step": 52379 }, { "epoch": 16.12, "learning_rate": 1.9189188531872037e-06, "loss": 0.0058, "step": 52380 }, { "epoch": 16.12, "learning_rate": 1.918625365125529e-06, "loss": 0.0064, "step": 52381 }, { "epoch": 16.12, "learning_rate": 1.918331897127732e-06, "loss": 0.0068, "step": 52382 }, { "epoch": 16.12, "learning_rate": 1.9180384491945403e-06, "loss": 0.0042, "step": 52383 }, { "epoch": 16.12, "learning_rate": 1.9177450213266778e-06, "loss": 0.0097, "step": 52384 }, { "epoch": 16.12, "learning_rate": 1.9174516135248745e-06, "loss": 0.004, "step": 52385 }, { "epoch": 16.12, "learning_rate": 1.917158225789858e-06, "loss": 0.0034, "step": 52386 }, { "epoch": 16.12, "learning_rate": 1.9168648581223616e-06, "loss": 0.0097, "step": 52387 }, { "epoch": 16.12, "learning_rate": 1.916571510523111e-06, "loss": 0.0037, "step": 52388 }, { "epoch": 16.12, "learning_rate": 1.9162781829928323e-06, "loss": 0.0057, "step": 52389 }, { "epoch": 16.12, "learning_rate": 1.9159848755322575e-06, "loss": 0.0082, "step": 52390 }, { "epoch": 16.12, "learning_rate": 1.915691588142111e-06, "loss": 0.0051, "step": 52391 }, { "epoch": 16.12, "learning_rate": 1.9153983208231243e-06, "loss": 0.0051, "step": 52392 }, { "epoch": 16.12, "learning_rate": 1.9151050735760237e-06, "loss": 0.006, "step": 52393 }, { "epoch": 16.12, "learning_rate": 1.9148118464015375e-06, "loss": 0.0073, "step": 52394 }, { "epoch": 16.12, "learning_rate": 1.9145186393003935e-06, "loss": 0.0074, "step": 52395 }, { "epoch": 16.12, "learning_rate": 1.9142254522733204e-06, "loss": 0.0025, "step": 52396 }, { "epoch": 16.12, "learning_rate": 1.913932285321045e-06, "loss": 0.005, "step": 52397 }, { "epoch": 16.12, "learning_rate": 1.9136391384442964e-06, "loss": 0.0038, "step": 52398 }, { "epoch": 16.12, "learning_rate": 1.913346011643802e-06, "loss": 0.0038, "step": 52399 }, { "epoch": 16.12, "learning_rate": 1.9130529049202884e-06, "loss": 0.0054, "step": 52400 }, { "epoch": 16.12, "learning_rate": 1.912759818274482e-06, "loss": 0.0041, "step": 52401 }, { "epoch": 16.12, "learning_rate": 1.912466751707116e-06, "loss": 0.0038, "step": 52402 }, { "epoch": 16.12, "learning_rate": 1.912173705218914e-06, "loss": 0.0086, "step": 52403 }, { "epoch": 16.12, "learning_rate": 1.911880678810607e-06, "loss": 0.0095, "step": 52404 }, { "epoch": 16.12, "learning_rate": 1.9115876724829174e-06, "loss": 0.0095, "step": 52405 }, { "epoch": 16.12, "learning_rate": 1.9112946862365723e-06, "loss": 0.0052, "step": 52406 }, { "epoch": 16.12, "learning_rate": 1.911001720072305e-06, "loss": 0.0093, "step": 52407 }, { "epoch": 16.13, "learning_rate": 1.9107087739908394e-06, "loss": 0.0067, "step": 52408 }, { "epoch": 16.13, "learning_rate": 1.9104158479929026e-06, "loss": 0.0069, "step": 52409 }, { "epoch": 16.13, "learning_rate": 1.9101229420792223e-06, "loss": 0.0023, "step": 52410 }, { "epoch": 16.13, "learning_rate": 1.9098300562505266e-06, "loss": 0.0047, "step": 52411 }, { "epoch": 16.13, "learning_rate": 1.909537190507541e-06, "loss": 0.0067, "step": 52412 }, { "epoch": 16.13, "learning_rate": 1.9092443448509947e-06, "loss": 0.0034, "step": 52413 }, { "epoch": 16.13, "learning_rate": 1.908951519281612e-06, "loss": 0.0048, "step": 52414 }, { "epoch": 16.13, "learning_rate": 1.9086587138001222e-06, "loss": 0.004, "step": 52415 }, { "epoch": 16.13, "learning_rate": 1.9083659284072487e-06, "loss": 0.0033, "step": 52416 }, { "epoch": 16.13, "learning_rate": 1.9080731631037265e-06, "loss": 0.0052, "step": 52417 }, { "epoch": 16.13, "learning_rate": 1.9077804178902737e-06, "loss": 0.0148, "step": 52418 }, { "epoch": 16.13, "learning_rate": 1.9074876927676213e-06, "loss": 0.0046, "step": 52419 }, { "epoch": 16.13, "learning_rate": 1.9071949877364948e-06, "loss": 0.004, "step": 52420 }, { "epoch": 16.13, "learning_rate": 1.9069023027976196e-06, "loss": 0.0098, "step": 52421 }, { "epoch": 16.13, "learning_rate": 1.9066096379517262e-06, "loss": 0.0068, "step": 52422 }, { "epoch": 16.13, "learning_rate": 1.9063169931995396e-06, "loss": 0.0049, "step": 52423 }, { "epoch": 16.13, "learning_rate": 1.9060243685417857e-06, "loss": 0.0048, "step": 52424 }, { "epoch": 16.13, "learning_rate": 1.905731763979194e-06, "loss": 0.0063, "step": 52425 }, { "epoch": 16.13, "learning_rate": 1.9054391795124815e-06, "loss": 0.0041, "step": 52426 }, { "epoch": 16.13, "learning_rate": 1.9051466151423858e-06, "loss": 0.0113, "step": 52427 }, { "epoch": 16.13, "learning_rate": 1.9048540708696284e-06, "loss": 0.0048, "step": 52428 }, { "epoch": 16.13, "learning_rate": 1.9045615466949353e-06, "loss": 0.0073, "step": 52429 }, { "epoch": 16.13, "learning_rate": 1.9042690426190336e-06, "loss": 0.0063, "step": 52430 }, { "epoch": 16.13, "learning_rate": 1.9039765586426496e-06, "loss": 0.0044, "step": 52431 }, { "epoch": 16.13, "learning_rate": 1.9036840947665091e-06, "loss": 0.0058, "step": 52432 }, { "epoch": 16.13, "learning_rate": 1.9033916509913376e-06, "loss": 0.0066, "step": 52433 }, { "epoch": 16.13, "learning_rate": 1.9030992273178628e-06, "loss": 0.008, "step": 52434 }, { "epoch": 16.13, "learning_rate": 1.902806823746809e-06, "loss": 0.0033, "step": 52435 }, { "epoch": 16.13, "learning_rate": 1.902514440278901e-06, "loss": 0.0046, "step": 52436 }, { "epoch": 16.13, "learning_rate": 1.902222076914869e-06, "loss": 0.0079, "step": 52437 }, { "epoch": 16.13, "learning_rate": 1.9019297336554387e-06, "loss": 0.0031, "step": 52438 }, { "epoch": 16.13, "learning_rate": 1.9016374105013324e-06, "loss": 0.0056, "step": 52439 }, { "epoch": 16.14, "learning_rate": 1.9013451074532752e-06, "loss": 0.0027, "step": 52440 }, { "epoch": 16.14, "learning_rate": 1.9010528245119942e-06, "loss": 0.0047, "step": 52441 }, { "epoch": 16.14, "learning_rate": 1.9007605616782166e-06, "loss": 0.0055, "step": 52442 }, { "epoch": 16.14, "learning_rate": 1.9004683189526674e-06, "loss": 0.0057, "step": 52443 }, { "epoch": 16.14, "learning_rate": 1.9001760963360727e-06, "loss": 0.0038, "step": 52444 }, { "epoch": 16.14, "learning_rate": 1.8998838938291565e-06, "loss": 0.0041, "step": 52445 }, { "epoch": 16.14, "learning_rate": 1.8995917114326446e-06, "loss": 0.0043, "step": 52446 }, { "epoch": 16.14, "learning_rate": 1.8992995491472632e-06, "loss": 0.005, "step": 52447 }, { "epoch": 16.14, "learning_rate": 1.8990074069737363e-06, "loss": 0.0064, "step": 52448 }, { "epoch": 16.14, "learning_rate": 1.898715284912791e-06, "loss": 0.0044, "step": 52449 }, { "epoch": 16.14, "learning_rate": 1.8984231829651512e-06, "loss": 0.0058, "step": 52450 }, { "epoch": 16.14, "learning_rate": 1.8981311011315395e-06, "loss": 0.0042, "step": 52451 }, { "epoch": 16.14, "learning_rate": 1.89783903941269e-06, "loss": 0.0048, "step": 52452 }, { "epoch": 16.14, "learning_rate": 1.8975469978093185e-06, "loss": 0.0056, "step": 52453 }, { "epoch": 16.14, "learning_rate": 1.8972549763221548e-06, "loss": 0.0043, "step": 52454 }, { "epoch": 16.14, "learning_rate": 1.8969629749519214e-06, "loss": 0.0069, "step": 52455 }, { "epoch": 16.14, "learning_rate": 1.896670993699342e-06, "loss": 0.0096, "step": 52456 }, { "epoch": 16.14, "learning_rate": 1.8963790325651464e-06, "loss": 0.0056, "step": 52457 }, { "epoch": 16.14, "learning_rate": 1.8960870915500573e-06, "loss": 0.0063, "step": 52458 }, { "epoch": 16.14, "learning_rate": 1.8957951706548018e-06, "loss": 0.0054, "step": 52459 }, { "epoch": 16.14, "learning_rate": 1.895503269880099e-06, "loss": 0.0056, "step": 52460 }, { "epoch": 16.14, "learning_rate": 1.8952113892266755e-06, "loss": 0.0069, "step": 52461 }, { "epoch": 16.14, "learning_rate": 1.8949195286952582e-06, "loss": 0.0077, "step": 52462 }, { "epoch": 16.14, "learning_rate": 1.894627688286571e-06, "loss": 0.007, "step": 52463 }, { "epoch": 16.14, "learning_rate": 1.894335868001338e-06, "loss": 0.0044, "step": 52464 }, { "epoch": 16.14, "learning_rate": 1.894044067840285e-06, "loss": 0.0143, "step": 52465 }, { "epoch": 16.14, "learning_rate": 1.8937522878041337e-06, "loss": 0.0093, "step": 52466 }, { "epoch": 16.14, "learning_rate": 1.8934605278936113e-06, "loss": 0.0053, "step": 52467 }, { "epoch": 16.14, "learning_rate": 1.8931687881094407e-06, "loss": 0.0063, "step": 52468 }, { "epoch": 16.14, "learning_rate": 1.8928770684523456e-06, "loss": 0.0049, "step": 52469 }, { "epoch": 16.14, "learning_rate": 1.8925853689230522e-06, "loss": 0.0045, "step": 52470 }, { "epoch": 16.14, "learning_rate": 1.8922936895222798e-06, "loss": 0.0031, "step": 52471 }, { "epoch": 16.14, "learning_rate": 1.8920020302507603e-06, "loss": 0.0021, "step": 52472 }, { "epoch": 16.15, "learning_rate": 1.8917103911092161e-06, "loss": 0.0047, "step": 52473 }, { "epoch": 16.15, "learning_rate": 1.8914187720983656e-06, "loss": 0.007, "step": 52474 }, { "epoch": 16.15, "learning_rate": 1.8911271732189372e-06, "loss": 0.0061, "step": 52475 }, { "epoch": 16.15, "learning_rate": 1.8908355944716516e-06, "loss": 0.0056, "step": 52476 }, { "epoch": 16.15, "learning_rate": 1.8905440358572368e-06, "loss": 0.0059, "step": 52477 }, { "epoch": 16.15, "learning_rate": 1.8902524973764158e-06, "loss": 0.0041, "step": 52478 }, { "epoch": 16.15, "learning_rate": 1.8899609790299112e-06, "loss": 0.0058, "step": 52479 }, { "epoch": 16.15, "learning_rate": 1.889669480818449e-06, "loss": 0.0048, "step": 52480 }, { "epoch": 16.15, "learning_rate": 1.8893780027427466e-06, "loss": 0.0073, "step": 52481 }, { "epoch": 16.15, "learning_rate": 1.8890865448035346e-06, "loss": 0.0042, "step": 52482 }, { "epoch": 16.15, "learning_rate": 1.8887951070015341e-06, "loss": 0.0051, "step": 52483 }, { "epoch": 16.15, "learning_rate": 1.8885036893374676e-06, "loss": 0.0132, "step": 52484 }, { "epoch": 16.15, "learning_rate": 1.8882122918120604e-06, "loss": 0.0068, "step": 52485 }, { "epoch": 16.15, "learning_rate": 1.8879209144260358e-06, "loss": 0.0049, "step": 52486 }, { "epoch": 16.15, "learning_rate": 1.8876295571801162e-06, "loss": 0.0054, "step": 52487 }, { "epoch": 16.15, "learning_rate": 1.8873382200750256e-06, "loss": 0.0022, "step": 52488 }, { "epoch": 16.15, "learning_rate": 1.8870469031114868e-06, "loss": 0.0031, "step": 52489 }, { "epoch": 16.15, "learning_rate": 1.8867556062902225e-06, "loss": 0.0041, "step": 52490 }, { "epoch": 16.15, "learning_rate": 1.8864643296119556e-06, "loss": 0.0057, "step": 52491 }, { "epoch": 16.15, "learning_rate": 1.886173073077412e-06, "loss": 0.0051, "step": 52492 }, { "epoch": 16.15, "learning_rate": 1.8858818366873145e-06, "loss": 0.0148, "step": 52493 }, { "epoch": 16.15, "learning_rate": 1.8855906204423857e-06, "loss": 0.0115, "step": 52494 }, { "epoch": 16.15, "learning_rate": 1.8852994243433465e-06, "loss": 0.0033, "step": 52495 }, { "epoch": 16.15, "learning_rate": 1.8850082483909182e-06, "loss": 0.0083, "step": 52496 }, { "epoch": 16.15, "learning_rate": 1.8847170925858293e-06, "loss": 0.0066, "step": 52497 }, { "epoch": 16.15, "learning_rate": 1.8844259569288005e-06, "loss": 0.0063, "step": 52498 }, { "epoch": 16.15, "learning_rate": 1.884134841420554e-06, "loss": 0.0077, "step": 52499 }, { "epoch": 16.15, "learning_rate": 1.883843746061812e-06, "loss": 0.0061, "step": 52500 }, { "epoch": 16.15, "learning_rate": 1.8835526708532981e-06, "loss": 0.0059, "step": 52501 }, { "epoch": 16.15, "learning_rate": 1.8832616157957352e-06, "loss": 0.0043, "step": 52502 }, { "epoch": 16.15, "learning_rate": 1.8829705808898458e-06, "loss": 0.005, "step": 52503 }, { "epoch": 16.15, "learning_rate": 1.8826795661363517e-06, "loss": 0.0062, "step": 52504 }, { "epoch": 16.16, "learning_rate": 1.8823885715359757e-06, "loss": 0.0073, "step": 52505 }, { "epoch": 16.16, "learning_rate": 1.8820975970894384e-06, "loss": 0.0042, "step": 52506 }, { "epoch": 16.16, "learning_rate": 1.8818066427974691e-06, "loss": 0.0063, "step": 52507 }, { "epoch": 16.16, "learning_rate": 1.8815157086607826e-06, "loss": 0.0063, "step": 52508 }, { "epoch": 16.16, "learning_rate": 1.8812247946801044e-06, "loss": 0.0058, "step": 52509 }, { "epoch": 16.16, "learning_rate": 1.8809339008561555e-06, "loss": 0.0074, "step": 52510 }, { "epoch": 16.16, "learning_rate": 1.880643027189657e-06, "loss": 0.006, "step": 52511 }, { "epoch": 16.16, "learning_rate": 1.8803521736813358e-06, "loss": 0.0088, "step": 52512 }, { "epoch": 16.16, "learning_rate": 1.88006134033191e-06, "loss": 0.0123, "step": 52513 }, { "epoch": 16.16, "learning_rate": 1.879770527142103e-06, "loss": 0.0047, "step": 52514 }, { "epoch": 16.16, "learning_rate": 1.8794797341126403e-06, "loss": 0.0035, "step": 52515 }, { "epoch": 16.16, "learning_rate": 1.8791889612442337e-06, "loss": 0.0028, "step": 52516 }, { "epoch": 16.16, "learning_rate": 1.878898208537615e-06, "loss": 0.004, "step": 52517 }, { "epoch": 16.16, "learning_rate": 1.8786074759935036e-06, "loss": 0.006, "step": 52518 }, { "epoch": 16.16, "learning_rate": 1.8783167636126187e-06, "loss": 0.0034, "step": 52519 }, { "epoch": 16.16, "learning_rate": 1.8780260713956843e-06, "loss": 0.0086, "step": 52520 }, { "epoch": 16.16, "learning_rate": 1.8777353993434221e-06, "loss": 0.0061, "step": 52521 }, { "epoch": 16.16, "learning_rate": 1.8774447474565527e-06, "loss": 0.0043, "step": 52522 }, { "epoch": 16.16, "learning_rate": 1.8771541157357975e-06, "loss": 0.0074, "step": 52523 }, { "epoch": 16.16, "learning_rate": 1.8768635041818795e-06, "loss": 0.0042, "step": 52524 }, { "epoch": 16.16, "learning_rate": 1.8765729127955202e-06, "loss": 0.006, "step": 52525 }, { "epoch": 16.16, "learning_rate": 1.8762823415774367e-06, "loss": 0.0038, "step": 52526 }, { "epoch": 16.16, "learning_rate": 1.8759917905283575e-06, "loss": 0.003, "step": 52527 }, { "epoch": 16.16, "learning_rate": 1.875701259649002e-06, "loss": 0.0074, "step": 52528 }, { "epoch": 16.16, "learning_rate": 1.8754107489400885e-06, "loss": 0.0059, "step": 52529 }, { "epoch": 16.16, "learning_rate": 1.8751202584023399e-06, "loss": 0.0036, "step": 52530 }, { "epoch": 16.16, "learning_rate": 1.8748297880364742e-06, "loss": 0.0065, "step": 52531 }, { "epoch": 16.16, "learning_rate": 1.8745393378432176e-06, "loss": 0.0029, "step": 52532 }, { "epoch": 16.16, "learning_rate": 1.8742489078232894e-06, "loss": 0.0038, "step": 52533 }, { "epoch": 16.16, "learning_rate": 1.8739584979774117e-06, "loss": 0.0048, "step": 52534 }, { "epoch": 16.16, "learning_rate": 1.8736681083063035e-06, "loss": 0.004, "step": 52535 }, { "epoch": 16.16, "learning_rate": 1.8733777388106866e-06, "loss": 0.006, "step": 52536 }, { "epoch": 16.16, "learning_rate": 1.8730873894912827e-06, "loss": 0.0069, "step": 52537 }, { "epoch": 16.17, "learning_rate": 1.872797060348811e-06, "loss": 0.0024, "step": 52538 }, { "epoch": 16.17, "learning_rate": 1.8725067513839934e-06, "loss": 0.0034, "step": 52539 }, { "epoch": 16.17, "learning_rate": 1.8722164625975502e-06, "loss": 0.006, "step": 52540 }, { "epoch": 16.17, "learning_rate": 1.8719261939902023e-06, "loss": 0.0024, "step": 52541 }, { "epoch": 16.17, "learning_rate": 1.871635945562671e-06, "loss": 0.0079, "step": 52542 }, { "epoch": 16.17, "learning_rate": 1.8713457173156757e-06, "loss": 0.0128, "step": 52543 }, { "epoch": 16.17, "learning_rate": 1.8710555092499383e-06, "loss": 0.0059, "step": 52544 }, { "epoch": 16.17, "learning_rate": 1.870765321366178e-06, "loss": 0.0041, "step": 52545 }, { "epoch": 16.17, "learning_rate": 1.8704751536651132e-06, "loss": 0.009, "step": 52546 }, { "epoch": 16.17, "learning_rate": 1.8701850061474692e-06, "loss": 0.0038, "step": 52547 }, { "epoch": 16.17, "learning_rate": 1.8698948788139647e-06, "loss": 0.004, "step": 52548 }, { "epoch": 16.17, "learning_rate": 1.869604771665322e-06, "loss": 0.0049, "step": 52549 }, { "epoch": 16.17, "learning_rate": 1.869314684702256e-06, "loss": 0.0104, "step": 52550 }, { "epoch": 16.17, "learning_rate": 1.8690246179254877e-06, "loss": 0.0085, "step": 52551 }, { "epoch": 16.17, "learning_rate": 1.8687345713357407e-06, "loss": 0.0035, "step": 52552 }, { "epoch": 16.17, "learning_rate": 1.868444544933734e-06, "loss": 0.0052, "step": 52553 }, { "epoch": 16.17, "learning_rate": 1.8681545387201883e-06, "loss": 0.0051, "step": 52554 }, { "epoch": 16.17, "learning_rate": 1.8678645526958216e-06, "loss": 0.0062, "step": 52555 }, { "epoch": 16.17, "learning_rate": 1.8675745868613548e-06, "loss": 0.0082, "step": 52556 }, { "epoch": 16.17, "learning_rate": 1.8672846412175083e-06, "loss": 0.0048, "step": 52557 }, { "epoch": 16.17, "learning_rate": 1.8669947157650015e-06, "loss": 0.0054, "step": 52558 }, { "epoch": 16.17, "learning_rate": 1.8667048105045537e-06, "loss": 0.0086, "step": 52559 }, { "epoch": 16.17, "learning_rate": 1.8664149254368857e-06, "loss": 0.0047, "step": 52560 }, { "epoch": 16.17, "learning_rate": 1.8661250605627146e-06, "loss": 0.0065, "step": 52561 }, { "epoch": 16.17, "learning_rate": 1.8658352158827665e-06, "loss": 0.0027, "step": 52562 }, { "epoch": 16.17, "learning_rate": 1.865545391397754e-06, "loss": 0.0045, "step": 52563 }, { "epoch": 16.17, "learning_rate": 1.8652555871084e-06, "loss": 0.0073, "step": 52564 }, { "epoch": 16.17, "learning_rate": 1.8649658030154226e-06, "loss": 0.0072, "step": 52565 }, { "epoch": 16.17, "learning_rate": 1.8646760391195396e-06, "loss": 0.0036, "step": 52566 }, { "epoch": 16.17, "learning_rate": 1.8643862954214754e-06, "loss": 0.0111, "step": 52567 }, { "epoch": 16.17, "learning_rate": 1.8640965719219473e-06, "loss": 0.005, "step": 52568 }, { "epoch": 16.17, "learning_rate": 1.8638068686216726e-06, "loss": 0.0052, "step": 52569 }, { "epoch": 16.18, "learning_rate": 1.8635171855213763e-06, "loss": 0.0042, "step": 52570 }, { "epoch": 16.18, "learning_rate": 1.8632275226217665e-06, "loss": 0.0059, "step": 52571 }, { "epoch": 16.18, "learning_rate": 1.862937879923573e-06, "loss": 0.0081, "step": 52572 }, { "epoch": 16.18, "learning_rate": 1.8626482574275107e-06, "loss": 0.0063, "step": 52573 }, { "epoch": 16.18, "learning_rate": 1.8623586551342988e-06, "loss": 0.0082, "step": 52574 }, { "epoch": 16.18, "learning_rate": 1.8620690730446567e-06, "loss": 0.0035, "step": 52575 }, { "epoch": 16.18, "learning_rate": 1.861779511159303e-06, "loss": 0.0083, "step": 52576 }, { "epoch": 16.18, "learning_rate": 1.8614899694789567e-06, "loss": 0.0078, "step": 52577 }, { "epoch": 16.18, "learning_rate": 1.8612004480043377e-06, "loss": 0.0075, "step": 52578 }, { "epoch": 16.18, "learning_rate": 1.860910946736163e-06, "loss": 0.0039, "step": 52579 }, { "epoch": 16.18, "learning_rate": 1.860621465675152e-06, "loss": 0.0043, "step": 52580 }, { "epoch": 16.18, "learning_rate": 1.8603320048220208e-06, "loss": 0.0071, "step": 52581 }, { "epoch": 16.18, "learning_rate": 1.8600425641774945e-06, "loss": 0.0033, "step": 52582 }, { "epoch": 16.18, "learning_rate": 1.85975314374229e-06, "loss": 0.0036, "step": 52583 }, { "epoch": 16.18, "learning_rate": 1.8594637435171215e-06, "loss": 0.0062, "step": 52584 }, { "epoch": 16.18, "learning_rate": 1.8591743635027094e-06, "loss": 0.0058, "step": 52585 }, { "epoch": 16.18, "learning_rate": 1.8588850036997708e-06, "loss": 0.0043, "step": 52586 }, { "epoch": 16.18, "learning_rate": 1.8585956641090276e-06, "loss": 0.0036, "step": 52587 }, { "epoch": 16.18, "learning_rate": 1.8583063447311978e-06, "loss": 0.0061, "step": 52588 }, { "epoch": 16.18, "learning_rate": 1.8580170455669976e-06, "loss": 0.0051, "step": 52589 }, { "epoch": 16.18, "learning_rate": 1.8577277666171456e-06, "loss": 0.0038, "step": 52590 }, { "epoch": 16.18, "learning_rate": 1.8574385078823608e-06, "loss": 0.0044, "step": 52591 }, { "epoch": 16.18, "learning_rate": 1.857149269363362e-06, "loss": 0.0033, "step": 52592 }, { "epoch": 16.18, "learning_rate": 1.8568600510608659e-06, "loss": 0.006, "step": 52593 }, { "epoch": 16.18, "learning_rate": 1.85657085297559e-06, "loss": 0.0033, "step": 52594 }, { "epoch": 16.18, "learning_rate": 1.8562816751082536e-06, "loss": 0.0036, "step": 52595 }, { "epoch": 16.18, "learning_rate": 1.855992517459575e-06, "loss": 0.0046, "step": 52596 }, { "epoch": 16.18, "learning_rate": 1.8557033800302715e-06, "loss": 0.0049, "step": 52597 }, { "epoch": 16.18, "learning_rate": 1.8554142628210603e-06, "loss": 0.0301, "step": 52598 }, { "epoch": 16.18, "learning_rate": 1.8551251658326596e-06, "loss": 0.0055, "step": 52599 }, { "epoch": 16.18, "learning_rate": 1.854836089065788e-06, "loss": 0.006, "step": 52600 }, { "epoch": 16.18, "learning_rate": 1.854547032521159e-06, "loss": 0.0055, "step": 52601 }, { "epoch": 16.18, "learning_rate": 1.8542579961994977e-06, "loss": 0.0058, "step": 52602 }, { "epoch": 16.19, "learning_rate": 1.853968980101517e-06, "loss": 0.0053, "step": 52603 }, { "epoch": 16.19, "learning_rate": 1.8536799842279374e-06, "loss": 0.0036, "step": 52604 }, { "epoch": 16.19, "learning_rate": 1.8533910085794714e-06, "loss": 0.0077, "step": 52605 }, { "epoch": 16.19, "learning_rate": 1.8531020531568377e-06, "loss": 0.0035, "step": 52606 }, { "epoch": 16.19, "learning_rate": 1.8528131179607578e-06, "loss": 0.0078, "step": 52607 }, { "epoch": 16.19, "learning_rate": 1.8525242029919454e-06, "loss": 0.0036, "step": 52608 }, { "epoch": 16.19, "learning_rate": 1.8522353082511203e-06, "loss": 0.006, "step": 52609 }, { "epoch": 16.19, "learning_rate": 1.8519464337389981e-06, "loss": 0.005, "step": 52610 }, { "epoch": 16.19, "learning_rate": 1.8516575794562951e-06, "loss": 0.0028, "step": 52611 }, { "epoch": 16.19, "learning_rate": 1.851368745403731e-06, "loss": 0.0076, "step": 52612 }, { "epoch": 16.19, "learning_rate": 1.8510799315820206e-06, "loss": 0.0058, "step": 52613 }, { "epoch": 16.19, "learning_rate": 1.850791137991882e-06, "loss": 0.0084, "step": 52614 }, { "epoch": 16.19, "learning_rate": 1.8505023646340304e-06, "loss": 0.0084, "step": 52615 }, { "epoch": 16.19, "learning_rate": 1.8502136115091863e-06, "loss": 0.0083, "step": 52616 }, { "epoch": 16.19, "learning_rate": 1.8499248786180667e-06, "loss": 0.0055, "step": 52617 }, { "epoch": 16.19, "learning_rate": 1.8496361659613848e-06, "loss": 0.0133, "step": 52618 }, { "epoch": 16.19, "learning_rate": 1.8493474735398575e-06, "loss": 0.005, "step": 52619 }, { "epoch": 16.19, "learning_rate": 1.849058801354202e-06, "loss": 0.0024, "step": 52620 }, { "epoch": 16.19, "learning_rate": 1.848770149405138e-06, "loss": 0.0028, "step": 52621 }, { "epoch": 16.19, "learning_rate": 1.8484815176933813e-06, "loss": 0.0037, "step": 52622 }, { "epoch": 16.19, "learning_rate": 1.8481929062196458e-06, "loss": 0.0028, "step": 52623 }, { "epoch": 16.19, "learning_rate": 1.847904314984651e-06, "loss": 0.0057, "step": 52624 }, { "epoch": 16.19, "learning_rate": 1.8476157439891108e-06, "loss": 0.0058, "step": 52625 }, { "epoch": 16.19, "learning_rate": 1.8473271932337433e-06, "loss": 0.014, "step": 52626 }, { "epoch": 16.19, "learning_rate": 1.847038662719265e-06, "loss": 0.007, "step": 52627 }, { "epoch": 16.19, "learning_rate": 1.8467501524463915e-06, "loss": 0.0028, "step": 52628 }, { "epoch": 16.19, "learning_rate": 1.8464616624158393e-06, "loss": 0.0046, "step": 52629 }, { "epoch": 16.19, "learning_rate": 1.8461731926283222e-06, "loss": 0.0051, "step": 52630 }, { "epoch": 16.19, "learning_rate": 1.845884743084564e-06, "loss": 0.0058, "step": 52631 }, { "epoch": 16.19, "learning_rate": 1.8455963137852728e-06, "loss": 0.0032, "step": 52632 }, { "epoch": 16.19, "learning_rate": 1.8453079047311673e-06, "loss": 0.0064, "step": 52633 }, { "epoch": 16.19, "learning_rate": 1.8450195159229644e-06, "loss": 0.0041, "step": 52634 }, { "epoch": 16.2, "learning_rate": 1.8447311473613772e-06, "loss": 0.0061, "step": 52635 }, { "epoch": 16.2, "learning_rate": 1.844442799047126e-06, "loss": 0.0039, "step": 52636 }, { "epoch": 16.2, "learning_rate": 1.8441544709809256e-06, "loss": 0.0047, "step": 52637 }, { "epoch": 16.2, "learning_rate": 1.8438661631634924e-06, "loss": 0.0146, "step": 52638 }, { "epoch": 16.2, "learning_rate": 1.843577875595538e-06, "loss": 0.0133, "step": 52639 }, { "epoch": 16.2, "learning_rate": 1.8432896082777785e-06, "loss": 0.0045, "step": 52640 }, { "epoch": 16.2, "learning_rate": 1.8430013612109355e-06, "loss": 0.0042, "step": 52641 }, { "epoch": 16.2, "learning_rate": 1.8427131343957206e-06, "loss": 0.0044, "step": 52642 }, { "epoch": 16.2, "learning_rate": 1.8424249278328488e-06, "loss": 0.0061, "step": 52643 }, { "epoch": 16.2, "learning_rate": 1.8421367415230374e-06, "loss": 0.0076, "step": 52644 }, { "epoch": 16.2, "learning_rate": 1.8418485754670013e-06, "loss": 0.0055, "step": 52645 }, { "epoch": 16.2, "learning_rate": 1.8415604296654556e-06, "loss": 0.0031, "step": 52646 }, { "epoch": 16.2, "learning_rate": 1.841272304119116e-06, "loss": 0.0079, "step": 52647 }, { "epoch": 16.2, "learning_rate": 1.840984198828698e-06, "loss": 0.0045, "step": 52648 }, { "epoch": 16.2, "learning_rate": 1.8406961137949165e-06, "loss": 0.0042, "step": 52649 }, { "epoch": 16.2, "learning_rate": 1.8404080490184838e-06, "loss": 0.0053, "step": 52650 }, { "epoch": 16.2, "learning_rate": 1.840120004500121e-06, "loss": 0.007, "step": 52651 }, { "epoch": 16.2, "learning_rate": 1.839831980240544e-06, "loss": 0.0067, "step": 52652 }, { "epoch": 16.2, "learning_rate": 1.83954397624046e-06, "loss": 0.0032, "step": 52653 }, { "epoch": 16.2, "learning_rate": 1.8392559925005892e-06, "loss": 0.0094, "step": 52654 }, { "epoch": 16.2, "learning_rate": 1.8389680290216428e-06, "loss": 0.0057, "step": 52655 }, { "epoch": 16.2, "learning_rate": 1.8386800858043418e-06, "loss": 0.0087, "step": 52656 }, { "epoch": 16.2, "learning_rate": 1.8383921628493972e-06, "loss": 0.0035, "step": 52657 }, { "epoch": 16.2, "learning_rate": 1.838104260157525e-06, "loss": 0.0085, "step": 52658 }, { "epoch": 16.2, "learning_rate": 1.8378163777294421e-06, "loss": 0.0055, "step": 52659 }, { "epoch": 16.2, "learning_rate": 1.837528515565856e-06, "loss": 0.0037, "step": 52660 }, { "epoch": 16.2, "learning_rate": 1.8372406736674874e-06, "loss": 0.0081, "step": 52661 }, { "epoch": 16.2, "learning_rate": 1.836952852035051e-06, "loss": 0.0038, "step": 52662 }, { "epoch": 16.2, "learning_rate": 1.8366650506692596e-06, "loss": 0.0125, "step": 52663 }, { "epoch": 16.2, "learning_rate": 1.836377269570827e-06, "loss": 0.0039, "step": 52664 }, { "epoch": 16.2, "learning_rate": 1.8360895087404696e-06, "loss": 0.0052, "step": 52665 }, { "epoch": 16.2, "learning_rate": 1.835801768178901e-06, "loss": 0.0045, "step": 52666 }, { "epoch": 16.2, "learning_rate": 1.8355140478868362e-06, "loss": 0.0082, "step": 52667 }, { "epoch": 16.21, "learning_rate": 1.835226347864988e-06, "loss": 0.0083, "step": 52668 }, { "epoch": 16.21, "learning_rate": 1.8349386681140723e-06, "loss": 0.0069, "step": 52669 }, { "epoch": 16.21, "learning_rate": 1.834651008634799e-06, "loss": 0.0052, "step": 52670 }, { "epoch": 16.21, "learning_rate": 1.8343633694278895e-06, "loss": 0.0044, "step": 52671 }, { "epoch": 16.21, "learning_rate": 1.8340757504940543e-06, "loss": 0.0058, "step": 52672 }, { "epoch": 16.21, "learning_rate": 1.8337881518340107e-06, "loss": 0.0053, "step": 52673 }, { "epoch": 16.21, "learning_rate": 1.8335005734484656e-06, "loss": 0.0026, "step": 52674 }, { "epoch": 16.21, "learning_rate": 1.8332130153381355e-06, "loss": 0.0086, "step": 52675 }, { "epoch": 16.21, "learning_rate": 1.8329254775037375e-06, "loss": 0.0063, "step": 52676 }, { "epoch": 16.21, "learning_rate": 1.8326379599459842e-06, "loss": 0.0056, "step": 52677 }, { "epoch": 16.21, "learning_rate": 1.8323504626655886e-06, "loss": 0.0051, "step": 52678 }, { "epoch": 16.21, "learning_rate": 1.8320629856632654e-06, "loss": 0.0063, "step": 52679 }, { "epoch": 16.21, "learning_rate": 1.8317755289397277e-06, "loss": 0.007, "step": 52680 }, { "epoch": 16.21, "learning_rate": 1.8314880924956891e-06, "loss": 0.0028, "step": 52681 }, { "epoch": 16.21, "learning_rate": 1.8312006763318635e-06, "loss": 0.0041, "step": 52682 }, { "epoch": 16.21, "learning_rate": 1.830913280448965e-06, "loss": 0.0044, "step": 52683 }, { "epoch": 16.21, "learning_rate": 1.830625904847706e-06, "loss": 0.0058, "step": 52684 }, { "epoch": 16.21, "learning_rate": 1.8303385495287972e-06, "loss": 0.0053, "step": 52685 }, { "epoch": 16.21, "learning_rate": 1.8300512144929616e-06, "loss": 0.0037, "step": 52686 }, { "epoch": 16.21, "learning_rate": 1.829763899740903e-06, "loss": 0.0041, "step": 52687 }, { "epoch": 16.21, "learning_rate": 1.8294766052733381e-06, "loss": 0.0041, "step": 52688 }, { "epoch": 16.21, "learning_rate": 1.8291893310909804e-06, "loss": 0.0074, "step": 52689 }, { "epoch": 16.21, "learning_rate": 1.8289020771945397e-06, "loss": 0.0077, "step": 52690 }, { "epoch": 16.21, "learning_rate": 1.8286148435847351e-06, "loss": 0.0097, "step": 52691 }, { "epoch": 16.21, "learning_rate": 1.8283276302622776e-06, "loss": 0.0045, "step": 52692 }, { "epoch": 16.21, "learning_rate": 1.8280404372278815e-06, "loss": 0.0062, "step": 52693 }, { "epoch": 16.21, "learning_rate": 1.8277532644822548e-06, "loss": 0.0057, "step": 52694 }, { "epoch": 16.21, "learning_rate": 1.8274661120261117e-06, "loss": 0.0043, "step": 52695 }, { "epoch": 16.21, "learning_rate": 1.827178979860169e-06, "loss": 0.0043, "step": 52696 }, { "epoch": 16.21, "learning_rate": 1.8268918679851388e-06, "loss": 0.0104, "step": 52697 }, { "epoch": 16.21, "learning_rate": 1.8266047764017314e-06, "loss": 0.0093, "step": 52698 }, { "epoch": 16.21, "learning_rate": 1.8263177051106618e-06, "loss": 0.0053, "step": 52699 }, { "epoch": 16.22, "learning_rate": 1.8260306541126416e-06, "loss": 0.0044, "step": 52700 }, { "epoch": 16.22, "learning_rate": 1.8257436234083824e-06, "loss": 0.0055, "step": 52701 }, { "epoch": 16.22, "learning_rate": 1.8254566129985996e-06, "loss": 0.0062, "step": 52702 }, { "epoch": 16.22, "learning_rate": 1.8251696228840032e-06, "loss": 0.0079, "step": 52703 }, { "epoch": 16.22, "learning_rate": 1.8248826530653075e-06, "loss": 0.005, "step": 52704 }, { "epoch": 16.22, "learning_rate": 1.8245957035432215e-06, "loss": 0.0056, "step": 52705 }, { "epoch": 16.22, "learning_rate": 1.824308774318463e-06, "loss": 0.0059, "step": 52706 }, { "epoch": 16.22, "learning_rate": 1.824021865391744e-06, "loss": 0.006, "step": 52707 }, { "epoch": 16.22, "learning_rate": 1.8237349767637724e-06, "loss": 0.0044, "step": 52708 }, { "epoch": 16.22, "learning_rate": 1.8234481084352619e-06, "loss": 0.0032, "step": 52709 }, { "epoch": 16.22, "learning_rate": 1.8231612604069238e-06, "loss": 0.0025, "step": 52710 }, { "epoch": 16.22, "learning_rate": 1.8228744326794735e-06, "loss": 0.0044, "step": 52711 }, { "epoch": 16.22, "learning_rate": 1.8225876252536222e-06, "loss": 0.0048, "step": 52712 }, { "epoch": 16.22, "learning_rate": 1.8223008381300812e-06, "loss": 0.006, "step": 52713 }, { "epoch": 16.22, "learning_rate": 1.8220140713095647e-06, "loss": 0.007, "step": 52714 }, { "epoch": 16.22, "learning_rate": 1.821727324792777e-06, "loss": 0.0097, "step": 52715 }, { "epoch": 16.22, "learning_rate": 1.8214405985804384e-06, "loss": 0.0124, "step": 52716 }, { "epoch": 16.22, "learning_rate": 1.8211538926732587e-06, "loss": 0.0055, "step": 52717 }, { "epoch": 16.22, "learning_rate": 1.820867207071948e-06, "loss": 0.003, "step": 52718 }, { "epoch": 16.22, "learning_rate": 1.8205805417772205e-06, "loss": 0.0061, "step": 52719 }, { "epoch": 16.22, "learning_rate": 1.8202938967897854e-06, "loss": 0.0071, "step": 52720 }, { "epoch": 16.22, "learning_rate": 1.8200072721103545e-06, "loss": 0.0041, "step": 52721 }, { "epoch": 16.22, "learning_rate": 1.8197206677396417e-06, "loss": 0.0083, "step": 52722 }, { "epoch": 16.22, "learning_rate": 1.8194340836783565e-06, "loss": 0.003, "step": 52723 }, { "epoch": 16.22, "learning_rate": 1.8191475199272114e-06, "loss": 0.0027, "step": 52724 }, { "epoch": 16.22, "learning_rate": 1.8188609764869147e-06, "loss": 0.0051, "step": 52725 }, { "epoch": 16.22, "learning_rate": 1.818574453358183e-06, "loss": 0.009, "step": 52726 }, { "epoch": 16.22, "learning_rate": 1.818287950541725e-06, "loss": 0.006, "step": 52727 }, { "epoch": 16.22, "learning_rate": 1.8180014680382552e-06, "loss": 0.0044, "step": 52728 }, { "epoch": 16.22, "learning_rate": 1.8177150058484804e-06, "loss": 0.0067, "step": 52729 }, { "epoch": 16.22, "learning_rate": 1.8174285639731093e-06, "loss": 0.0046, "step": 52730 }, { "epoch": 16.22, "learning_rate": 1.817142142412861e-06, "loss": 0.0037, "step": 52731 }, { "epoch": 16.22, "learning_rate": 1.8168557411684418e-06, "loss": 0.0094, "step": 52732 }, { "epoch": 16.23, "learning_rate": 1.8165693602405643e-06, "loss": 0.0065, "step": 52733 }, { "epoch": 16.23, "learning_rate": 1.8162829996299381e-06, "loss": 0.0055, "step": 52734 }, { "epoch": 16.23, "learning_rate": 1.8159966593372757e-06, "loss": 0.0066, "step": 52735 }, { "epoch": 16.23, "learning_rate": 1.8157103393632869e-06, "loss": 0.0071, "step": 52736 }, { "epoch": 16.23, "learning_rate": 1.8154240397086831e-06, "loss": 0.005, "step": 52737 }, { "epoch": 16.23, "learning_rate": 1.8151377603741738e-06, "loss": 0.0061, "step": 52738 }, { "epoch": 16.23, "learning_rate": 1.8148515013604728e-06, "loss": 0.0031, "step": 52739 }, { "epoch": 16.23, "learning_rate": 1.8145652626682852e-06, "loss": 0.0043, "step": 52740 }, { "epoch": 16.23, "learning_rate": 1.8142790442983304e-06, "loss": 0.0069, "step": 52741 }, { "epoch": 16.23, "learning_rate": 1.8139928462513113e-06, "loss": 0.004, "step": 52742 }, { "epoch": 16.23, "learning_rate": 1.8137066685279414e-06, "loss": 0.0051, "step": 52743 }, { "epoch": 16.23, "learning_rate": 1.8134205111289305e-06, "loss": 0.0071, "step": 52744 }, { "epoch": 16.23, "learning_rate": 1.813134374054988e-06, "loss": 0.0053, "step": 52745 }, { "epoch": 16.23, "learning_rate": 1.8128482573068274e-06, "loss": 0.0054, "step": 52746 }, { "epoch": 16.23, "learning_rate": 1.8125621608851574e-06, "loss": 0.0038, "step": 52747 }, { "epoch": 16.23, "learning_rate": 1.8122760847906884e-06, "loss": 0.0034, "step": 52748 }, { "epoch": 16.23, "learning_rate": 1.8119900290241331e-06, "loss": 0.0053, "step": 52749 }, { "epoch": 16.23, "learning_rate": 1.8117039935861935e-06, "loss": 0.0028, "step": 52750 }, { "epoch": 16.23, "learning_rate": 1.8114179784775886e-06, "loss": 0.0075, "step": 52751 }, { "epoch": 16.23, "learning_rate": 1.811131983699026e-06, "loss": 0.0091, "step": 52752 }, { "epoch": 16.23, "learning_rate": 1.8108460092512137e-06, "loss": 0.0067, "step": 52753 }, { "epoch": 16.23, "learning_rate": 1.8105600551348634e-06, "loss": 0.0055, "step": 52754 }, { "epoch": 16.23, "learning_rate": 1.8102741213506847e-06, "loss": 0.0041, "step": 52755 }, { "epoch": 16.23, "learning_rate": 1.809988207899388e-06, "loss": 0.0058, "step": 52756 }, { "epoch": 16.23, "learning_rate": 1.8097023147816817e-06, "loss": 0.0067, "step": 52757 }, { "epoch": 16.23, "learning_rate": 1.8094164419982775e-06, "loss": 0.002, "step": 52758 }, { "epoch": 16.23, "learning_rate": 1.8091305895498835e-06, "loss": 0.0038, "step": 52759 }, { "epoch": 16.23, "learning_rate": 1.808844757437207e-06, "loss": 0.0038, "step": 52760 }, { "epoch": 16.23, "learning_rate": 1.8085589456609641e-06, "loss": 0.0057, "step": 52761 }, { "epoch": 16.23, "learning_rate": 1.8082731542218635e-06, "loss": 0.0046, "step": 52762 }, { "epoch": 16.23, "learning_rate": 1.8079873831206096e-06, "loss": 0.0033, "step": 52763 }, { "epoch": 16.23, "learning_rate": 1.8077016323579145e-06, "loss": 0.0065, "step": 52764 }, { "epoch": 16.24, "learning_rate": 1.807415901934485e-06, "loss": 0.0063, "step": 52765 }, { "epoch": 16.24, "learning_rate": 1.8071301918510364e-06, "loss": 0.0047, "step": 52766 }, { "epoch": 16.24, "learning_rate": 1.8068445021082738e-06, "loss": 0.0044, "step": 52767 }, { "epoch": 16.24, "learning_rate": 1.8065588327069084e-06, "loss": 0.0054, "step": 52768 }, { "epoch": 16.24, "learning_rate": 1.8062731836476477e-06, "loss": 0.0056, "step": 52769 }, { "epoch": 16.24, "learning_rate": 1.8059875549312023e-06, "loss": 0.006, "step": 52770 }, { "epoch": 16.24, "learning_rate": 1.8057019465582803e-06, "loss": 0.0091, "step": 52771 }, { "epoch": 16.24, "learning_rate": 1.8054163585295914e-06, "loss": 0.0073, "step": 52772 }, { "epoch": 16.24, "learning_rate": 1.805130790845845e-06, "loss": 0.0027, "step": 52773 }, { "epoch": 16.24, "learning_rate": 1.804845243507749e-06, "loss": 0.0054, "step": 52774 }, { "epoch": 16.24, "learning_rate": 1.8045597165160134e-06, "loss": 0.0056, "step": 52775 }, { "epoch": 16.24, "learning_rate": 1.8042742098713461e-06, "loss": 0.0042, "step": 52776 }, { "epoch": 16.24, "learning_rate": 1.8039887235744557e-06, "loss": 0.0064, "step": 52777 }, { "epoch": 16.24, "learning_rate": 1.8037032576260527e-06, "loss": 0.0056, "step": 52778 }, { "epoch": 16.24, "learning_rate": 1.8034178120268452e-06, "loss": 0.0126, "step": 52779 }, { "epoch": 16.24, "learning_rate": 1.8031323867775384e-06, "loss": 0.0067, "step": 52780 }, { "epoch": 16.24, "learning_rate": 1.8028469818788475e-06, "loss": 0.003, "step": 52781 }, { "epoch": 16.24, "learning_rate": 1.8025615973314769e-06, "loss": 0.0069, "step": 52782 }, { "epoch": 16.24, "learning_rate": 1.8022762331361376e-06, "loss": 0.0059, "step": 52783 }, { "epoch": 16.24, "learning_rate": 1.8019908892935356e-06, "loss": 0.005, "step": 52784 }, { "epoch": 16.24, "learning_rate": 1.801705565804377e-06, "loss": 0.0039, "step": 52785 }, { "epoch": 16.24, "learning_rate": 1.8014202626693744e-06, "loss": 0.0046, "step": 52786 }, { "epoch": 16.24, "learning_rate": 1.8011349798892363e-06, "loss": 0.0094, "step": 52787 }, { "epoch": 16.24, "learning_rate": 1.80084971746467e-06, "loss": 0.0045, "step": 52788 }, { "epoch": 16.24, "learning_rate": 1.8005644753963824e-06, "loss": 0.0048, "step": 52789 }, { "epoch": 16.24, "learning_rate": 1.8002792536850844e-06, "loss": 0.0035, "step": 52790 }, { "epoch": 16.24, "learning_rate": 1.7999940523314808e-06, "loss": 0.006, "step": 52791 }, { "epoch": 16.24, "learning_rate": 1.799708871336282e-06, "loss": 0.0095, "step": 52792 }, { "epoch": 16.24, "learning_rate": 1.7994237107001955e-06, "loss": 0.0107, "step": 52793 }, { "epoch": 16.24, "learning_rate": 1.7991385704239283e-06, "loss": 0.0071, "step": 52794 }, { "epoch": 16.24, "learning_rate": 1.7988534505081867e-06, "loss": 0.0078, "step": 52795 }, { "epoch": 16.24, "learning_rate": 1.7985683509536867e-06, "loss": 0.0099, "step": 52796 }, { "epoch": 16.24, "learning_rate": 1.7982832717611277e-06, "loss": 0.0064, "step": 52797 }, { "epoch": 16.25, "learning_rate": 1.79799821293122e-06, "loss": 0.0098, "step": 52798 }, { "epoch": 16.25, "learning_rate": 1.7977131744646724e-06, "loss": 0.0068, "step": 52799 }, { "epoch": 16.25, "learning_rate": 1.7974281563621888e-06, "loss": 0.0032, "step": 52800 }, { "epoch": 16.25, "learning_rate": 1.7971431586244814e-06, "loss": 0.0064, "step": 52801 }, { "epoch": 16.25, "learning_rate": 1.7968581812522567e-06, "loss": 0.0078, "step": 52802 }, { "epoch": 16.25, "learning_rate": 1.7965732242462218e-06, "loss": 0.0026, "step": 52803 }, { "epoch": 16.25, "learning_rate": 1.7962882876070863e-06, "loss": 0.005, "step": 52804 }, { "epoch": 16.25, "learning_rate": 1.7960033713355507e-06, "loss": 0.0036, "step": 52805 }, { "epoch": 16.25, "learning_rate": 1.7957184754323299e-06, "loss": 0.01, "step": 52806 }, { "epoch": 16.25, "learning_rate": 1.7954335998981277e-06, "loss": 0.0034, "step": 52807 }, { "epoch": 16.25, "learning_rate": 1.7951487447336512e-06, "loss": 0.0054, "step": 52808 }, { "epoch": 16.25, "learning_rate": 1.7948639099396092e-06, "loss": 0.004, "step": 52809 }, { "epoch": 16.25, "learning_rate": 1.7945790955167085e-06, "loss": 0.0112, "step": 52810 }, { "epoch": 16.25, "learning_rate": 1.7942943014656543e-06, "loss": 0.0051, "step": 52811 }, { "epoch": 16.25, "learning_rate": 1.7940095277871573e-06, "loss": 0.0078, "step": 52812 }, { "epoch": 16.25, "learning_rate": 1.7937247744819208e-06, "loss": 0.0054, "step": 52813 }, { "epoch": 16.25, "learning_rate": 1.7934400415506537e-06, "loss": 0.0037, "step": 52814 }, { "epoch": 16.25, "learning_rate": 1.7931553289940607e-06, "loss": 0.0098, "step": 52815 }, { "epoch": 16.25, "learning_rate": 1.7928706368128524e-06, "loss": 0.0044, "step": 52816 }, { "epoch": 16.25, "learning_rate": 1.792585965007736e-06, "loss": 0.0032, "step": 52817 }, { "epoch": 16.25, "learning_rate": 1.792301313579413e-06, "loss": 0.0068, "step": 52818 }, { "epoch": 16.25, "learning_rate": 1.7920166825285944e-06, "loss": 0.0127, "step": 52819 }, { "epoch": 16.25, "learning_rate": 1.7917320718559828e-06, "loss": 0.0079, "step": 52820 }, { "epoch": 16.25, "learning_rate": 1.7914474815622896e-06, "loss": 0.0046, "step": 52821 }, { "epoch": 16.25, "learning_rate": 1.7911629116482198e-06, "loss": 0.0062, "step": 52822 }, { "epoch": 16.25, "learning_rate": 1.7908783621144798e-06, "loss": 0.0089, "step": 52823 }, { "epoch": 16.25, "learning_rate": 1.7905938329617745e-06, "loss": 0.0069, "step": 52824 }, { "epoch": 16.25, "learning_rate": 1.7903093241908121e-06, "loss": 0.0031, "step": 52825 }, { "epoch": 16.25, "learning_rate": 1.7900248358022986e-06, "loss": 0.0049, "step": 52826 }, { "epoch": 16.25, "learning_rate": 1.7897403677969405e-06, "loss": 0.0027, "step": 52827 }, { "epoch": 16.25, "learning_rate": 1.7894559201754425e-06, "loss": 0.009, "step": 52828 }, { "epoch": 16.25, "learning_rate": 1.7891714929385128e-06, "loss": 0.0119, "step": 52829 }, { "epoch": 16.26, "learning_rate": 1.7888870860868568e-06, "loss": 0.0026, "step": 52830 }, { "epoch": 16.26, "learning_rate": 1.7886026996211804e-06, "loss": 0.0063, "step": 52831 }, { "epoch": 16.26, "learning_rate": 1.7883183335421894e-06, "loss": 0.0111, "step": 52832 }, { "epoch": 16.26, "learning_rate": 1.7880339878505903e-06, "loss": 0.0034, "step": 52833 }, { "epoch": 16.26, "learning_rate": 1.787749662547089e-06, "loss": 0.0083, "step": 52834 }, { "epoch": 16.26, "learning_rate": 1.7874653576323885e-06, "loss": 0.0092, "step": 52835 }, { "epoch": 16.26, "learning_rate": 1.7871810731072004e-06, "loss": 0.0041, "step": 52836 }, { "epoch": 16.26, "learning_rate": 1.7868968089722283e-06, "loss": 0.0047, "step": 52837 }, { "epoch": 16.26, "learning_rate": 1.7866125652281785e-06, "loss": 0.0176, "step": 52838 }, { "epoch": 16.26, "learning_rate": 1.7863283418757537e-06, "loss": 0.0047, "step": 52839 }, { "epoch": 16.26, "learning_rate": 1.786044138915658e-06, "loss": 0.0035, "step": 52840 }, { "epoch": 16.26, "learning_rate": 1.785759956348604e-06, "loss": 0.0067, "step": 52841 }, { "epoch": 16.26, "learning_rate": 1.785475794175292e-06, "loss": 0.004, "step": 52842 }, { "epoch": 16.26, "learning_rate": 1.7851916523964297e-06, "loss": 0.0037, "step": 52843 }, { "epoch": 16.26, "learning_rate": 1.7849075310127228e-06, "loss": 0.0034, "step": 52844 }, { "epoch": 16.26, "learning_rate": 1.7846234300248743e-06, "loss": 0.008, "step": 52845 }, { "epoch": 16.26, "learning_rate": 1.7843393494335925e-06, "loss": 0.0045, "step": 52846 }, { "epoch": 16.26, "learning_rate": 1.7840552892395802e-06, "loss": 0.0072, "step": 52847 }, { "epoch": 16.26, "learning_rate": 1.7837712494435433e-06, "loss": 0.0079, "step": 52848 }, { "epoch": 16.26, "learning_rate": 1.783487230046188e-06, "loss": 0.0056, "step": 52849 }, { "epoch": 16.26, "learning_rate": 1.7832032310482162e-06, "loss": 0.0048, "step": 52850 }, { "epoch": 16.26, "learning_rate": 1.7829192524503414e-06, "loss": 0.0052, "step": 52851 }, { "epoch": 16.26, "learning_rate": 1.78263529425326e-06, "loss": 0.004, "step": 52852 }, { "epoch": 16.26, "learning_rate": 1.7823513564576788e-06, "loss": 0.0098, "step": 52853 }, { "epoch": 16.26, "learning_rate": 1.7820674390643044e-06, "loss": 0.0077, "step": 52854 }, { "epoch": 16.26, "learning_rate": 1.7817835420738382e-06, "loss": 0.006, "step": 52855 }, { "epoch": 16.26, "learning_rate": 1.781499665486991e-06, "loss": 0.0049, "step": 52856 }, { "epoch": 16.26, "learning_rate": 1.781215809304464e-06, "loss": 0.0051, "step": 52857 }, { "epoch": 16.26, "learning_rate": 1.7809319735269638e-06, "loss": 0.0085, "step": 52858 }, { "epoch": 16.26, "learning_rate": 1.7806481581551948e-06, "loss": 0.0077, "step": 52859 }, { "epoch": 16.26, "learning_rate": 1.780364363189856e-06, "loss": 0.0071, "step": 52860 }, { "epoch": 16.26, "learning_rate": 1.7800805886316585e-06, "loss": 0.0055, "step": 52861 }, { "epoch": 16.26, "learning_rate": 1.7797968344813056e-06, "loss": 0.0037, "step": 52862 }, { "epoch": 16.27, "learning_rate": 1.7795131007395006e-06, "loss": 0.0034, "step": 52863 }, { "epoch": 16.27, "learning_rate": 1.7792293874069488e-06, "loss": 0.0042, "step": 52864 }, { "epoch": 16.27, "learning_rate": 1.7789456944843542e-06, "loss": 0.0101, "step": 52865 }, { "epoch": 16.27, "learning_rate": 1.7786620219724205e-06, "loss": 0.0024, "step": 52866 }, { "epoch": 16.27, "learning_rate": 1.7783783698718527e-06, "loss": 0.0047, "step": 52867 }, { "epoch": 16.27, "learning_rate": 1.7780947381833557e-06, "loss": 0.0082, "step": 52868 }, { "epoch": 16.27, "learning_rate": 1.7778111269076314e-06, "loss": 0.0064, "step": 52869 }, { "epoch": 16.27, "learning_rate": 1.7775275360453847e-06, "loss": 0.004, "step": 52870 }, { "epoch": 16.27, "learning_rate": 1.7772439655973217e-06, "loss": 0.0063, "step": 52871 }, { "epoch": 16.27, "learning_rate": 1.7769604155641485e-06, "loss": 0.0044, "step": 52872 }, { "epoch": 16.27, "learning_rate": 1.7766768859465623e-06, "loss": 0.0058, "step": 52873 }, { "epoch": 16.27, "learning_rate": 1.7763933767452712e-06, "loss": 0.0046, "step": 52874 }, { "epoch": 16.27, "learning_rate": 1.7761098879609762e-06, "loss": 0.0032, "step": 52875 }, { "epoch": 16.27, "learning_rate": 1.7758264195943842e-06, "loss": 0.0113, "step": 52876 }, { "epoch": 16.27, "learning_rate": 1.7755429716461992e-06, "loss": 0.0048, "step": 52877 }, { "epoch": 16.27, "learning_rate": 1.7752595441171239e-06, "loss": 0.01, "step": 52878 }, { "epoch": 16.27, "learning_rate": 1.774976137007861e-06, "loss": 0.0043, "step": 52879 }, { "epoch": 16.27, "learning_rate": 1.7746927503191157e-06, "loss": 0.0028, "step": 52880 }, { "epoch": 16.27, "learning_rate": 1.7744093840515907e-06, "loss": 0.0072, "step": 52881 }, { "epoch": 16.27, "learning_rate": 1.77412603820599e-06, "loss": 0.0028, "step": 52882 }, { "epoch": 16.27, "learning_rate": 1.773842712783016e-06, "loss": 0.0053, "step": 52883 }, { "epoch": 16.27, "learning_rate": 1.7735594077833728e-06, "loss": 0.0056, "step": 52884 }, { "epoch": 16.27, "learning_rate": 1.773276123207761e-06, "loss": 0.0136, "step": 52885 }, { "epoch": 16.27, "learning_rate": 1.7729928590568923e-06, "loss": 0.0052, "step": 52886 }, { "epoch": 16.27, "learning_rate": 1.7727096153314627e-06, "loss": 0.0029, "step": 52887 }, { "epoch": 16.27, "learning_rate": 1.772426392032176e-06, "loss": 0.0079, "step": 52888 }, { "epoch": 16.27, "learning_rate": 1.7721431891597362e-06, "loss": 0.0098, "step": 52889 }, { "epoch": 16.27, "learning_rate": 1.771860006714845e-06, "loss": 0.005, "step": 52890 }, { "epoch": 16.27, "learning_rate": 1.7715768446982097e-06, "loss": 0.0092, "step": 52891 }, { "epoch": 16.27, "learning_rate": 1.7712937031105304e-06, "loss": 0.0053, "step": 52892 }, { "epoch": 16.27, "learning_rate": 1.7710105819525136e-06, "loss": 0.0064, "step": 52893 }, { "epoch": 16.27, "learning_rate": 1.7707274812248554e-06, "loss": 0.0065, "step": 52894 }, { "epoch": 16.28, "learning_rate": 1.7704444009282606e-06, "loss": 0.0091, "step": 52895 }, { "epoch": 16.28, "learning_rate": 1.7701613410634367e-06, "loss": 0.0064, "step": 52896 }, { "epoch": 16.28, "learning_rate": 1.7698783016310827e-06, "loss": 0.0042, "step": 52897 }, { "epoch": 16.28, "learning_rate": 1.7695952826319018e-06, "loss": 0.0083, "step": 52898 }, { "epoch": 16.28, "learning_rate": 1.7693122840665978e-06, "loss": 0.0038, "step": 52899 }, { "epoch": 16.28, "learning_rate": 1.7690293059358721e-06, "loss": 0.0067, "step": 52900 }, { "epoch": 16.28, "learning_rate": 1.7687463482404276e-06, "loss": 0.0068, "step": 52901 }, { "epoch": 16.28, "learning_rate": 1.7684634109809672e-06, "loss": 0.0039, "step": 52902 }, { "epoch": 16.28, "learning_rate": 1.7681804941581925e-06, "loss": 0.0054, "step": 52903 }, { "epoch": 16.28, "learning_rate": 1.7678975977728063e-06, "loss": 0.0058, "step": 52904 }, { "epoch": 16.28, "learning_rate": 1.7676147218255092e-06, "loss": 0.0035, "step": 52905 }, { "epoch": 16.28, "learning_rate": 1.767331866317008e-06, "loss": 0.0049, "step": 52906 }, { "epoch": 16.28, "learning_rate": 1.7670490312480037e-06, "loss": 0.0047, "step": 52907 }, { "epoch": 16.28, "learning_rate": 1.7667662166191957e-06, "loss": 0.0064, "step": 52908 }, { "epoch": 16.28, "learning_rate": 1.7664834224312865e-06, "loss": 0.008, "step": 52909 }, { "epoch": 16.28, "learning_rate": 1.7662006486849781e-06, "loss": 0.018, "step": 52910 }, { "epoch": 16.28, "learning_rate": 1.7659178953809752e-06, "loss": 0.0047, "step": 52911 }, { "epoch": 16.28, "learning_rate": 1.7656351625199786e-06, "loss": 0.0062, "step": 52912 }, { "epoch": 16.28, "learning_rate": 1.7653524501026898e-06, "loss": 0.0054, "step": 52913 }, { "epoch": 16.28, "learning_rate": 1.7650697581298138e-06, "loss": 0.005, "step": 52914 }, { "epoch": 16.28, "learning_rate": 1.7647870866020445e-06, "loss": 0.0044, "step": 52915 }, { "epoch": 16.28, "learning_rate": 1.7645044355200913e-06, "loss": 0.008, "step": 52916 }, { "epoch": 16.28, "learning_rate": 1.7642218048846527e-06, "loss": 0.0089, "step": 52917 }, { "epoch": 16.28, "learning_rate": 1.7639391946964312e-06, "loss": 0.0052, "step": 52918 }, { "epoch": 16.28, "learning_rate": 1.7636566049561287e-06, "loss": 0.0074, "step": 52919 }, { "epoch": 16.28, "learning_rate": 1.7633740356644468e-06, "loss": 0.0041, "step": 52920 }, { "epoch": 16.28, "learning_rate": 1.763091486822086e-06, "loss": 0.0082, "step": 52921 }, { "epoch": 16.28, "learning_rate": 1.762808958429748e-06, "loss": 0.0095, "step": 52922 }, { "epoch": 16.28, "learning_rate": 1.7625264504881356e-06, "loss": 0.0064, "step": 52923 }, { "epoch": 16.28, "learning_rate": 1.7622439629979493e-06, "loss": 0.003, "step": 52924 }, { "epoch": 16.28, "learning_rate": 1.7619614959598874e-06, "loss": 0.0121, "step": 52925 }, { "epoch": 16.28, "learning_rate": 1.7616790493746572e-06, "loss": 0.01, "step": 52926 }, { "epoch": 16.28, "learning_rate": 1.7613966232429559e-06, "loss": 0.0066, "step": 52927 }, { "epoch": 16.29, "learning_rate": 1.7611142175654893e-06, "loss": 0.0052, "step": 52928 }, { "epoch": 16.29, "learning_rate": 1.7608318323429518e-06, "loss": 0.0051, "step": 52929 }, { "epoch": 16.29, "learning_rate": 1.760549467576046e-06, "loss": 0.0021, "step": 52930 }, { "epoch": 16.29, "learning_rate": 1.7602671232654755e-06, "loss": 0.0043, "step": 52931 }, { "epoch": 16.29, "learning_rate": 1.7599847994119423e-06, "loss": 0.0057, "step": 52932 }, { "epoch": 16.29, "learning_rate": 1.7597024960161436e-06, "loss": 0.0069, "step": 52933 }, { "epoch": 16.29, "learning_rate": 1.759420213078783e-06, "loss": 0.0064, "step": 52934 }, { "epoch": 16.29, "learning_rate": 1.7591379506005612e-06, "loss": 0.0042, "step": 52935 }, { "epoch": 16.29, "learning_rate": 1.7588557085821766e-06, "loss": 0.0064, "step": 52936 }, { "epoch": 16.29, "learning_rate": 1.7585734870243321e-06, "loss": 0.0093, "step": 52937 }, { "epoch": 16.29, "learning_rate": 1.758291285927729e-06, "loss": 0.0055, "step": 52938 }, { "epoch": 16.29, "learning_rate": 1.7580091052930648e-06, "loss": 0.0029, "step": 52939 }, { "epoch": 16.29, "learning_rate": 1.757726945121041e-06, "loss": 0.0043, "step": 52940 }, { "epoch": 16.29, "learning_rate": 1.757444805412364e-06, "loss": 0.0046, "step": 52941 }, { "epoch": 16.29, "learning_rate": 1.757162686167726e-06, "loss": 0.0086, "step": 52942 }, { "epoch": 16.29, "learning_rate": 1.7568805873878315e-06, "loss": 0.0069, "step": 52943 }, { "epoch": 16.29, "learning_rate": 1.7565985090733806e-06, "loss": 0.0034, "step": 52944 }, { "epoch": 16.29, "learning_rate": 1.7563164512250697e-06, "loss": 0.0043, "step": 52945 }, { "epoch": 16.29, "learning_rate": 1.7560344138436059e-06, "loss": 0.0083, "step": 52946 }, { "epoch": 16.29, "learning_rate": 1.7557523969296852e-06, "loss": 0.008, "step": 52947 }, { "epoch": 16.29, "learning_rate": 1.7554704004840096e-06, "loss": 0.0068, "step": 52948 }, { "epoch": 16.29, "learning_rate": 1.7551884245072813e-06, "loss": 0.0061, "step": 52949 }, { "epoch": 16.29, "learning_rate": 1.7549064690001916e-06, "loss": 0.0062, "step": 52950 }, { "epoch": 16.29, "learning_rate": 1.7546245339634494e-06, "loss": 0.0039, "step": 52951 }, { "epoch": 16.29, "learning_rate": 1.7543426193977508e-06, "loss": 0.0027, "step": 52952 }, { "epoch": 16.29, "learning_rate": 1.7540607253037966e-06, "loss": 0.0087, "step": 52953 }, { "epoch": 16.29, "learning_rate": 1.7537788516822872e-06, "loss": 0.0063, "step": 52954 }, { "epoch": 16.29, "learning_rate": 1.7534969985339212e-06, "loss": 0.0037, "step": 52955 }, { "epoch": 16.29, "learning_rate": 1.7532151658594e-06, "loss": 0.0067, "step": 52956 }, { "epoch": 16.29, "learning_rate": 1.7529333536594217e-06, "loss": 0.0068, "step": 52957 }, { "epoch": 16.29, "learning_rate": 1.7526515619346862e-06, "loss": 0.0038, "step": 52958 }, { "epoch": 16.29, "learning_rate": 1.7523697906858939e-06, "loss": 0.0075, "step": 52959 }, { "epoch": 16.3, "learning_rate": 1.7520880399137409e-06, "loss": 0.005, "step": 52960 }, { "epoch": 16.3, "learning_rate": 1.7518063096189319e-06, "loss": 0.0068, "step": 52961 }, { "epoch": 16.3, "learning_rate": 1.751524599802168e-06, "loss": 0.0076, "step": 52962 }, { "epoch": 16.3, "learning_rate": 1.7512429104641415e-06, "loss": 0.0038, "step": 52963 }, { "epoch": 16.3, "learning_rate": 1.7509612416055555e-06, "loss": 0.0027, "step": 52964 }, { "epoch": 16.3, "learning_rate": 1.7506795932271058e-06, "loss": 0.0033, "step": 52965 }, { "epoch": 16.3, "learning_rate": 1.7503979653294966e-06, "loss": 0.0059, "step": 52966 }, { "epoch": 16.3, "learning_rate": 1.750116357913426e-06, "loss": 0.0068, "step": 52967 }, { "epoch": 16.3, "learning_rate": 1.7498347709795926e-06, "loss": 0.0056, "step": 52968 }, { "epoch": 16.3, "learning_rate": 1.7495532045286955e-06, "loss": 0.0045, "step": 52969 }, { "epoch": 16.3, "learning_rate": 1.7492716585614334e-06, "loss": 0.0094, "step": 52970 }, { "epoch": 16.3, "learning_rate": 1.7489901330785053e-06, "loss": 0.0037, "step": 52971 }, { "epoch": 16.3, "learning_rate": 1.7487086280806098e-06, "loss": 0.0042, "step": 52972 }, { "epoch": 16.3, "learning_rate": 1.7484271435684474e-06, "loss": 0.0076, "step": 52973 }, { "epoch": 16.3, "learning_rate": 1.7481456795427144e-06, "loss": 0.0066, "step": 52974 }, { "epoch": 16.3, "learning_rate": 1.747864236004112e-06, "loss": 0.0046, "step": 52975 }, { "epoch": 16.3, "learning_rate": 1.7475828129533378e-06, "loss": 0.0089, "step": 52976 }, { "epoch": 16.3, "learning_rate": 1.7473014103910912e-06, "loss": 0.0059, "step": 52977 }, { "epoch": 16.3, "learning_rate": 1.7470200283180693e-06, "loss": 0.0095, "step": 52978 }, { "epoch": 16.3, "learning_rate": 1.7467386667349728e-06, "loss": 0.0037, "step": 52979 }, { "epoch": 16.3, "learning_rate": 1.7464573256424955e-06, "loss": 0.003, "step": 52980 }, { "epoch": 16.3, "learning_rate": 1.7461760050413435e-06, "loss": 0.0043, "step": 52981 }, { "epoch": 16.3, "learning_rate": 1.7458947049322105e-06, "loss": 0.0068, "step": 52982 }, { "epoch": 16.3, "learning_rate": 1.7456134253157976e-06, "loss": 0.0097, "step": 52983 }, { "epoch": 16.3, "learning_rate": 1.7453321661928002e-06, "loss": 0.0051, "step": 52984 }, { "epoch": 16.3, "learning_rate": 1.7450509275639139e-06, "loss": 0.0046, "step": 52985 }, { "epoch": 16.3, "learning_rate": 1.7447697094298433e-06, "loss": 0.0046, "step": 52986 }, { "epoch": 16.3, "learning_rate": 1.7444885117912845e-06, "loss": 0.0035, "step": 52987 }, { "epoch": 16.3, "learning_rate": 1.7442073346489351e-06, "loss": 0.0078, "step": 52988 }, { "epoch": 16.3, "learning_rate": 1.7439261780034932e-06, "loss": 0.0101, "step": 52989 }, { "epoch": 16.3, "learning_rate": 1.743645041855656e-06, "loss": 0.0069, "step": 52990 }, { "epoch": 16.3, "learning_rate": 1.743363926206123e-06, "loss": 0.0056, "step": 52991 }, { "epoch": 16.3, "learning_rate": 1.7430828310555914e-06, "loss": 0.0107, "step": 52992 }, { "epoch": 16.31, "learning_rate": 1.7428017564047594e-06, "loss": 0.0058, "step": 52993 }, { "epoch": 16.31, "learning_rate": 1.7425207022543244e-06, "loss": 0.0151, "step": 52994 }, { "epoch": 16.31, "learning_rate": 1.7422396686049813e-06, "loss": 0.0056, "step": 52995 }, { "epoch": 16.31, "learning_rate": 1.7419586554574364e-06, "loss": 0.0021, "step": 52996 }, { "epoch": 16.31, "learning_rate": 1.7416776628123798e-06, "loss": 0.0032, "step": 52997 }, { "epoch": 16.31, "learning_rate": 1.7413966906705105e-06, "loss": 0.006, "step": 52998 }, { "epoch": 16.31, "learning_rate": 1.7411157390325273e-06, "loss": 0.0078, "step": 52999 }, { "epoch": 16.31, "learning_rate": 1.7408348078991245e-06, "loss": 0.0058, "step": 53000 }, { "epoch": 16.31, "learning_rate": 1.7405538972710044e-06, "loss": 0.0068, "step": 53001 }, { "epoch": 16.31, "learning_rate": 1.7402730071488617e-06, "loss": 0.0063, "step": 53002 }, { "epoch": 16.31, "learning_rate": 1.7399921375333962e-06, "loss": 0.0063, "step": 53003 }, { "epoch": 16.31, "learning_rate": 1.739711288425302e-06, "loss": 0.0065, "step": 53004 }, { "epoch": 16.31, "learning_rate": 1.7394304598252776e-06, "loss": 0.0058, "step": 53005 }, { "epoch": 16.31, "learning_rate": 1.7391496517340212e-06, "loss": 0.0047, "step": 53006 }, { "epoch": 16.31, "learning_rate": 1.7388688641522289e-06, "loss": 0.0065, "step": 53007 }, { "epoch": 16.31, "learning_rate": 1.738588097080598e-06, "loss": 0.0045, "step": 53008 }, { "epoch": 16.31, "learning_rate": 1.7383073505198255e-06, "loss": 0.0081, "step": 53009 }, { "epoch": 16.31, "learning_rate": 1.738026624470609e-06, "loss": 0.004, "step": 53010 }, { "epoch": 16.31, "learning_rate": 1.7377459189336442e-06, "loss": 0.0049, "step": 53011 }, { "epoch": 16.31, "learning_rate": 1.7374652339096298e-06, "loss": 0.0045, "step": 53012 }, { "epoch": 16.31, "learning_rate": 1.7371845693992618e-06, "loss": 0.005, "step": 53013 }, { "epoch": 16.31, "learning_rate": 1.7369039254032338e-06, "loss": 0.0039, "step": 53014 }, { "epoch": 16.31, "learning_rate": 1.736623301922249e-06, "loss": 0.0073, "step": 53015 }, { "epoch": 16.31, "learning_rate": 1.7363426989569997e-06, "loss": 0.0066, "step": 53016 }, { "epoch": 16.31, "learning_rate": 1.736062116508187e-06, "loss": 0.0053, "step": 53017 }, { "epoch": 16.31, "learning_rate": 1.7357815545765012e-06, "loss": 0.0039, "step": 53018 }, { "epoch": 16.31, "learning_rate": 1.7355010131626393e-06, "loss": 0.0069, "step": 53019 }, { "epoch": 16.31, "learning_rate": 1.7352204922673032e-06, "loss": 0.0064, "step": 53020 }, { "epoch": 16.31, "learning_rate": 1.7349399918911858e-06, "loss": 0.0044, "step": 53021 }, { "epoch": 16.31, "learning_rate": 1.7346595120349852e-06, "loss": 0.0074, "step": 53022 }, { "epoch": 16.31, "learning_rate": 1.7343790526993965e-06, "loss": 0.0038, "step": 53023 }, { "epoch": 16.31, "learning_rate": 1.7340986138851158e-06, "loss": 0.0062, "step": 53024 }, { "epoch": 16.32, "learning_rate": 1.7338181955928401e-06, "loss": 0.0094, "step": 53025 }, { "epoch": 16.32, "learning_rate": 1.7335377978232648e-06, "loss": 0.0067, "step": 53026 }, { "epoch": 16.32, "learning_rate": 1.7332574205770868e-06, "loss": 0.0045, "step": 53027 }, { "epoch": 16.32, "learning_rate": 1.7329770638550026e-06, "loss": 0.0124, "step": 53028 }, { "epoch": 16.32, "learning_rate": 1.7326967276577044e-06, "loss": 0.0072, "step": 53029 }, { "epoch": 16.32, "learning_rate": 1.7324164119858977e-06, "loss": 0.004, "step": 53030 }, { "epoch": 16.32, "learning_rate": 1.7321361168402684e-06, "loss": 0.0065, "step": 53031 }, { "epoch": 16.32, "learning_rate": 1.7318558422215171e-06, "loss": 0.0053, "step": 53032 }, { "epoch": 16.32, "learning_rate": 1.7315755881303377e-06, "loss": 0.0047, "step": 53033 }, { "epoch": 16.32, "learning_rate": 1.731295354567425e-06, "loss": 0.0046, "step": 53034 }, { "epoch": 16.32, "learning_rate": 1.7310151415334798e-06, "loss": 0.0035, "step": 53035 }, { "epoch": 16.32, "learning_rate": 1.7307349490291937e-06, "loss": 0.0091, "step": 53036 }, { "epoch": 16.32, "learning_rate": 1.7304547770552638e-06, "loss": 0.0132, "step": 53037 }, { "epoch": 16.32, "learning_rate": 1.7301746256123874e-06, "loss": 0.0101, "step": 53038 }, { "epoch": 16.32, "learning_rate": 1.7298944947012542e-06, "loss": 0.0126, "step": 53039 }, { "epoch": 16.32, "learning_rate": 1.7296143843225654e-06, "loss": 0.0031, "step": 53040 }, { "epoch": 16.32, "learning_rate": 1.7293342944770142e-06, "loss": 0.0092, "step": 53041 }, { "epoch": 16.32, "learning_rate": 1.7290542251652953e-06, "loss": 0.0055, "step": 53042 }, { "epoch": 16.32, "learning_rate": 1.728774176388106e-06, "loss": 0.0052, "step": 53043 }, { "epoch": 16.32, "learning_rate": 1.7284941481461416e-06, "loss": 0.0083, "step": 53044 }, { "epoch": 16.32, "learning_rate": 1.7282141404400943e-06, "loss": 0.0082, "step": 53045 }, { "epoch": 16.32, "learning_rate": 1.7279341532706627e-06, "loss": 0.0094, "step": 53046 }, { "epoch": 16.32, "learning_rate": 1.7276541866385398e-06, "loss": 0.0052, "step": 53047 }, { "epoch": 16.32, "learning_rate": 1.7273742405444217e-06, "loss": 0.0083, "step": 53048 }, { "epoch": 16.32, "learning_rate": 1.7270943149890006e-06, "loss": 0.0055, "step": 53049 }, { "epoch": 16.32, "learning_rate": 1.7268144099729766e-06, "loss": 0.0057, "step": 53050 }, { "epoch": 16.32, "learning_rate": 1.7265345254970445e-06, "loss": 0.0044, "step": 53051 }, { "epoch": 16.32, "learning_rate": 1.7262546615618936e-06, "loss": 0.0045, "step": 53052 }, { "epoch": 16.32, "learning_rate": 1.7259748181682222e-06, "loss": 0.0056, "step": 53053 }, { "epoch": 16.32, "learning_rate": 1.7256949953167224e-06, "loss": 0.0068, "step": 53054 }, { "epoch": 16.32, "learning_rate": 1.7254151930080942e-06, "loss": 0.0045, "step": 53055 }, { "epoch": 16.32, "learning_rate": 1.7251354112430286e-06, "loss": 0.0055, "step": 53056 }, { "epoch": 16.32, "learning_rate": 1.7248556500222213e-06, "loss": 0.0057, "step": 53057 }, { "epoch": 16.33, "learning_rate": 1.7245759093463665e-06, "loss": 0.0076, "step": 53058 }, { "epoch": 16.33, "learning_rate": 1.7242961892161581e-06, "loss": 0.0066, "step": 53059 }, { "epoch": 16.33, "learning_rate": 1.724016489632292e-06, "loss": 0.0085, "step": 53060 }, { "epoch": 16.33, "learning_rate": 1.723736810595461e-06, "loss": 0.0041, "step": 53061 }, { "epoch": 16.33, "learning_rate": 1.7234571521063614e-06, "loss": 0.0051, "step": 53062 }, { "epoch": 16.33, "learning_rate": 1.7231775141656858e-06, "loss": 0.0051, "step": 53063 }, { "epoch": 16.33, "learning_rate": 1.7228978967741261e-06, "loss": 0.0042, "step": 53064 }, { "epoch": 16.33, "learning_rate": 1.7226182999323849e-06, "loss": 0.01, "step": 53065 }, { "epoch": 16.33, "learning_rate": 1.7223387236411493e-06, "loss": 0.0081, "step": 53066 }, { "epoch": 16.33, "learning_rate": 1.7220591679011135e-06, "loss": 0.0039, "step": 53067 }, { "epoch": 16.33, "learning_rate": 1.7217796327129743e-06, "loss": 0.0055, "step": 53068 }, { "epoch": 16.33, "learning_rate": 1.7215001180774215e-06, "loss": 0.0051, "step": 53069 }, { "epoch": 16.33, "learning_rate": 1.7212206239951545e-06, "loss": 0.0047, "step": 53070 }, { "epoch": 16.33, "learning_rate": 1.720941150466865e-06, "loss": 0.0044, "step": 53071 }, { "epoch": 16.33, "learning_rate": 1.7206616974932488e-06, "loss": 0.0071, "step": 53072 }, { "epoch": 16.33, "learning_rate": 1.7203822650749957e-06, "loss": 0.0065, "step": 53073 }, { "epoch": 16.33, "learning_rate": 1.7201028532127994e-06, "loss": 0.0089, "step": 53074 }, { "epoch": 16.33, "learning_rate": 1.7198234619073562e-06, "loss": 0.0135, "step": 53075 }, { "epoch": 16.33, "learning_rate": 1.7195440911593607e-06, "loss": 0.0081, "step": 53076 }, { "epoch": 16.33, "learning_rate": 1.719264740969504e-06, "loss": 0.0038, "step": 53077 }, { "epoch": 16.33, "learning_rate": 1.7189854113384808e-06, "loss": 0.0065, "step": 53078 }, { "epoch": 16.33, "learning_rate": 1.7187061022669848e-06, "loss": 0.0091, "step": 53079 }, { "epoch": 16.33, "learning_rate": 1.718426813755708e-06, "loss": 0.0021, "step": 53080 }, { "epoch": 16.33, "learning_rate": 1.7181475458053453e-06, "loss": 0.0087, "step": 53081 }, { "epoch": 16.33, "learning_rate": 1.7178682984165895e-06, "loss": 0.0068, "step": 53082 }, { "epoch": 16.33, "learning_rate": 1.7175890715901343e-06, "loss": 0.0044, "step": 53083 }, { "epoch": 16.33, "learning_rate": 1.7173098653266696e-06, "loss": 0.0055, "step": 53084 }, { "epoch": 16.33, "learning_rate": 1.7170306796268942e-06, "loss": 0.0091, "step": 53085 }, { "epoch": 16.33, "learning_rate": 1.7167515144915003e-06, "loss": 0.0058, "step": 53086 }, { "epoch": 16.33, "learning_rate": 1.7164723699211782e-06, "loss": 0.0057, "step": 53087 }, { "epoch": 16.33, "learning_rate": 1.7161932459166209e-06, "loss": 0.0047, "step": 53088 }, { "epoch": 16.33, "learning_rate": 1.715914142478521e-06, "loss": 0.0082, "step": 53089 }, { "epoch": 16.34, "learning_rate": 1.7156350596075743e-06, "loss": 0.0038, "step": 53090 }, { "epoch": 16.34, "learning_rate": 1.715355997304473e-06, "loss": 0.0057, "step": 53091 }, { "epoch": 16.34, "learning_rate": 1.7150769555699099e-06, "loss": 0.0082, "step": 53092 }, { "epoch": 16.34, "learning_rate": 1.7147979344045785e-06, "loss": 0.005, "step": 53093 }, { "epoch": 16.34, "learning_rate": 1.7145189338091662e-06, "loss": 0.0069, "step": 53094 }, { "epoch": 16.34, "learning_rate": 1.7142399537843712e-06, "loss": 0.0044, "step": 53095 }, { "epoch": 16.34, "learning_rate": 1.7139609943308856e-06, "loss": 0.0049, "step": 53096 }, { "epoch": 16.34, "learning_rate": 1.7136820554494005e-06, "loss": 0.0035, "step": 53097 }, { "epoch": 16.34, "learning_rate": 1.713403137140609e-06, "loss": 0.0035, "step": 53098 }, { "epoch": 16.34, "learning_rate": 1.713124239405204e-06, "loss": 0.0086, "step": 53099 }, { "epoch": 16.34, "learning_rate": 1.7128453622438768e-06, "loss": 0.0054, "step": 53100 }, { "epoch": 16.34, "learning_rate": 1.7125665056573205e-06, "loss": 0.0038, "step": 53101 }, { "epoch": 16.34, "learning_rate": 1.7122876696462276e-06, "loss": 0.007, "step": 53102 }, { "epoch": 16.34, "learning_rate": 1.71200885421129e-06, "loss": 0.0069, "step": 53103 }, { "epoch": 16.34, "learning_rate": 1.7117300593531972e-06, "loss": 0.0075, "step": 53104 }, { "epoch": 16.34, "learning_rate": 1.7114512850726473e-06, "loss": 0.0054, "step": 53105 }, { "epoch": 16.34, "learning_rate": 1.7111725313703287e-06, "loss": 0.0106, "step": 53106 }, { "epoch": 16.34, "learning_rate": 1.7108937982469376e-06, "loss": 0.0028, "step": 53107 }, { "epoch": 16.34, "learning_rate": 1.7106150857031589e-06, "loss": 0.0053, "step": 53108 }, { "epoch": 16.34, "learning_rate": 1.7103363937396867e-06, "loss": 0.0056, "step": 53109 }, { "epoch": 16.34, "learning_rate": 1.7100577223572157e-06, "loss": 0.0066, "step": 53110 }, { "epoch": 16.34, "learning_rate": 1.7097790715564377e-06, "loss": 0.0043, "step": 53111 }, { "epoch": 16.34, "learning_rate": 1.7095004413380423e-06, "loss": 0.0086, "step": 53112 }, { "epoch": 16.34, "learning_rate": 1.709221831702723e-06, "loss": 0.0046, "step": 53113 }, { "epoch": 16.34, "learning_rate": 1.708943242651171e-06, "loss": 0.0065, "step": 53114 }, { "epoch": 16.34, "learning_rate": 1.7086646741840762e-06, "loss": 0.0028, "step": 53115 }, { "epoch": 16.34, "learning_rate": 1.708386126302133e-06, "loss": 0.011, "step": 53116 }, { "epoch": 16.34, "learning_rate": 1.7081075990060314e-06, "loss": 0.0052, "step": 53117 }, { "epoch": 16.34, "learning_rate": 1.7078290922964635e-06, "loss": 0.0071, "step": 53118 }, { "epoch": 16.34, "learning_rate": 1.7075506061741175e-06, "loss": 0.0142, "step": 53119 }, { "epoch": 16.34, "learning_rate": 1.7072721406396931e-06, "loss": 0.0066, "step": 53120 }, { "epoch": 16.34, "learning_rate": 1.7069936956938737e-06, "loss": 0.004, "step": 53121 }, { "epoch": 16.34, "learning_rate": 1.7067152713373537e-06, "loss": 0.003, "step": 53122 }, { "epoch": 16.35, "learning_rate": 1.7064368675708232e-06, "loss": 0.0057, "step": 53123 }, { "epoch": 16.35, "learning_rate": 1.706158484394972e-06, "loss": 0.0071, "step": 53124 }, { "epoch": 16.35, "learning_rate": 1.7058801218104947e-06, "loss": 0.0056, "step": 53125 }, { "epoch": 16.35, "learning_rate": 1.7056017798180824e-06, "loss": 0.007, "step": 53126 }, { "epoch": 16.35, "learning_rate": 1.705323458418424e-06, "loss": 0.0062, "step": 53127 }, { "epoch": 16.35, "learning_rate": 1.705045157612215e-06, "loss": 0.0045, "step": 53128 }, { "epoch": 16.35, "learning_rate": 1.7047668774001369e-06, "loss": 0.005, "step": 53129 }, { "epoch": 16.35, "learning_rate": 1.7044886177828878e-06, "loss": 0.0056, "step": 53130 }, { "epoch": 16.35, "learning_rate": 1.7042103787611585e-06, "loss": 0.0063, "step": 53131 }, { "epoch": 16.35, "learning_rate": 1.7039321603356373e-06, "loss": 0.0059, "step": 53132 }, { "epoch": 16.35, "learning_rate": 1.703653962507017e-06, "loss": 0.008, "step": 53133 }, { "epoch": 16.35, "learning_rate": 1.7033757852759868e-06, "loss": 0.005, "step": 53134 }, { "epoch": 16.35, "learning_rate": 1.7030976286432376e-06, "loss": 0.009, "step": 53135 }, { "epoch": 16.35, "learning_rate": 1.70281949260946e-06, "loss": 0.0049, "step": 53136 }, { "epoch": 16.35, "learning_rate": 1.7025413771753452e-06, "loss": 0.0091, "step": 53137 }, { "epoch": 16.35, "learning_rate": 1.702263282341583e-06, "loss": 0.0052, "step": 53138 }, { "epoch": 16.35, "learning_rate": 1.7019852081088616e-06, "loss": 0.0029, "step": 53139 }, { "epoch": 16.35, "learning_rate": 1.7017071544778774e-06, "loss": 0.0085, "step": 53140 }, { "epoch": 16.35, "learning_rate": 1.7014291214493184e-06, "loss": 0.0069, "step": 53141 }, { "epoch": 16.35, "learning_rate": 1.7011511090238718e-06, "loss": 0.0057, "step": 53142 }, { "epoch": 16.35, "learning_rate": 1.7008731172022296e-06, "loss": 0.0064, "step": 53143 }, { "epoch": 16.35, "learning_rate": 1.700595145985079e-06, "loss": 0.0057, "step": 53144 }, { "epoch": 16.35, "learning_rate": 1.7003171953731157e-06, "loss": 0.0047, "step": 53145 }, { "epoch": 16.35, "learning_rate": 1.7000392653670283e-06, "loss": 0.0062, "step": 53146 }, { "epoch": 16.35, "learning_rate": 1.6997613559675053e-06, "loss": 0.0106, "step": 53147 }, { "epoch": 16.35, "learning_rate": 1.699483467175237e-06, "loss": 0.0053, "step": 53148 }, { "epoch": 16.35, "learning_rate": 1.6992055989909129e-06, "loss": 0.0058, "step": 53149 }, { "epoch": 16.35, "learning_rate": 1.6989277514152248e-06, "loss": 0.0042, "step": 53150 }, { "epoch": 16.35, "learning_rate": 1.6986499244488607e-06, "loss": 0.0024, "step": 53151 }, { "epoch": 16.35, "learning_rate": 1.6983721180925106e-06, "loss": 0.007, "step": 53152 }, { "epoch": 16.35, "learning_rate": 1.6980943323468645e-06, "loss": 0.0064, "step": 53153 }, { "epoch": 16.35, "learning_rate": 1.6978165672126124e-06, "loss": 0.0065, "step": 53154 }, { "epoch": 16.36, "learning_rate": 1.6975388226904433e-06, "loss": 0.0068, "step": 53155 }, { "epoch": 16.36, "learning_rate": 1.6972610987810467e-06, "loss": 0.0081, "step": 53156 }, { "epoch": 16.36, "learning_rate": 1.6969833954851123e-06, "loss": 0.0038, "step": 53157 }, { "epoch": 16.36, "learning_rate": 1.6967057128033305e-06, "loss": 0.0082, "step": 53158 }, { "epoch": 16.36, "learning_rate": 1.6964280507363872e-06, "loss": 0.0135, "step": 53159 }, { "epoch": 16.36, "learning_rate": 1.6961504092849768e-06, "loss": 0.0062, "step": 53160 }, { "epoch": 16.36, "learning_rate": 1.695872788449786e-06, "loss": 0.0072, "step": 53161 }, { "epoch": 16.36, "learning_rate": 1.6955951882315081e-06, "loss": 0.004, "step": 53162 }, { "epoch": 16.36, "learning_rate": 1.6953176086308242e-06, "loss": 0.0044, "step": 53163 }, { "epoch": 16.36, "learning_rate": 1.6950400496484266e-06, "loss": 0.0096, "step": 53164 }, { "epoch": 16.36, "learning_rate": 1.6947625112850074e-06, "loss": 0.0043, "step": 53165 }, { "epoch": 16.36, "learning_rate": 1.6944849935412544e-06, "loss": 0.0071, "step": 53166 }, { "epoch": 16.36, "learning_rate": 1.6942074964178567e-06, "loss": 0.0026, "step": 53167 }, { "epoch": 16.36, "learning_rate": 1.6939300199155017e-06, "loss": 0.004, "step": 53168 }, { "epoch": 16.36, "learning_rate": 1.693652564034879e-06, "loss": 0.0083, "step": 53169 }, { "epoch": 16.36, "learning_rate": 1.6933751287766775e-06, "loss": 0.0024, "step": 53170 }, { "epoch": 16.36, "learning_rate": 1.6930977141415872e-06, "loss": 0.0078, "step": 53171 }, { "epoch": 16.36, "learning_rate": 1.692820320130295e-06, "loss": 0.0095, "step": 53172 }, { "epoch": 16.36, "learning_rate": 1.6925429467434906e-06, "loss": 0.0077, "step": 53173 }, { "epoch": 16.36, "learning_rate": 1.69226559398186e-06, "loss": 0.0025, "step": 53174 }, { "epoch": 16.36, "learning_rate": 1.6919882618460992e-06, "loss": 0.0065, "step": 53175 }, { "epoch": 16.36, "learning_rate": 1.6917109503368889e-06, "loss": 0.003, "step": 53176 }, { "epoch": 16.36, "learning_rate": 1.691433659454922e-06, "loss": 0.0049, "step": 53177 }, { "epoch": 16.36, "learning_rate": 1.691156389200883e-06, "loss": 0.0077, "step": 53178 }, { "epoch": 16.36, "learning_rate": 1.690879139575462e-06, "loss": 0.0056, "step": 53179 }, { "epoch": 16.36, "learning_rate": 1.6906019105793503e-06, "loss": 0.0052, "step": 53180 }, { "epoch": 16.36, "learning_rate": 1.6903247022132329e-06, "loss": 0.0063, "step": 53181 }, { "epoch": 16.36, "learning_rate": 1.6900475144778006e-06, "loss": 0.0035, "step": 53182 }, { "epoch": 16.36, "learning_rate": 1.6897703473737414e-06, "loss": 0.0049, "step": 53183 }, { "epoch": 16.36, "learning_rate": 1.6894932009017373e-06, "loss": 0.0061, "step": 53184 }, { "epoch": 16.36, "learning_rate": 1.6892160750624831e-06, "loss": 0.0046, "step": 53185 }, { "epoch": 16.36, "learning_rate": 1.688938969856666e-06, "loss": 0.0072, "step": 53186 }, { "epoch": 16.36, "learning_rate": 1.6886618852849723e-06, "loss": 0.0093, "step": 53187 }, { "epoch": 16.37, "learning_rate": 1.6883848213480902e-06, "loss": 0.0048, "step": 53188 }, { "epoch": 16.37, "learning_rate": 1.6881077780467082e-06, "loss": 0.0056, "step": 53189 }, { "epoch": 16.37, "learning_rate": 1.6878307553815132e-06, "loss": 0.011, "step": 53190 }, { "epoch": 16.37, "learning_rate": 1.687553753353195e-06, "loss": 0.0053, "step": 53191 }, { "epoch": 16.37, "learning_rate": 1.6872767719624383e-06, "loss": 0.0063, "step": 53192 }, { "epoch": 16.37, "learning_rate": 1.6869998112099328e-06, "loss": 0.0087, "step": 53193 }, { "epoch": 16.37, "learning_rate": 1.6867228710963635e-06, "loss": 0.004, "step": 53194 }, { "epoch": 16.37, "learning_rate": 1.6864459516224218e-06, "loss": 0.0074, "step": 53195 }, { "epoch": 16.37, "learning_rate": 1.6861690527887975e-06, "loss": 0.0047, "step": 53196 }, { "epoch": 16.37, "learning_rate": 1.6858921745961698e-06, "loss": 0.0026, "step": 53197 }, { "epoch": 16.37, "learning_rate": 1.6856153170452315e-06, "loss": 0.0047, "step": 53198 }, { "epoch": 16.37, "learning_rate": 1.6853384801366669e-06, "loss": 0.0063, "step": 53199 }, { "epoch": 16.37, "learning_rate": 1.685061663871167e-06, "loss": 0.0051, "step": 53200 }, { "epoch": 16.37, "learning_rate": 1.6847848682494172e-06, "loss": 0.0044, "step": 53201 }, { "epoch": 16.37, "learning_rate": 1.6845080932721048e-06, "loss": 0.0034, "step": 53202 }, { "epoch": 16.37, "learning_rate": 1.6842313389399178e-06, "loss": 0.0043, "step": 53203 }, { "epoch": 16.37, "learning_rate": 1.6839546052535416e-06, "loss": 0.006, "step": 53204 }, { "epoch": 16.37, "learning_rate": 1.6836778922136642e-06, "loss": 0.0073, "step": 53205 }, { "epoch": 16.37, "learning_rate": 1.6834011998209732e-06, "loss": 0.0042, "step": 53206 }, { "epoch": 16.37, "learning_rate": 1.6831245280761555e-06, "loss": 0.0068, "step": 53207 }, { "epoch": 16.37, "learning_rate": 1.6828478769798962e-06, "loss": 0.0073, "step": 53208 }, { "epoch": 16.37, "learning_rate": 1.6825712465328835e-06, "loss": 0.0053, "step": 53209 }, { "epoch": 16.37, "learning_rate": 1.682294636735805e-06, "loss": 0.0056, "step": 53210 }, { "epoch": 16.37, "learning_rate": 1.6820180475893455e-06, "loss": 0.0127, "step": 53211 }, { "epoch": 16.37, "learning_rate": 1.6817414790941932e-06, "loss": 0.0037, "step": 53212 }, { "epoch": 16.37, "learning_rate": 1.6814649312510346e-06, "loss": 0.0058, "step": 53213 }, { "epoch": 16.37, "learning_rate": 1.681188404060553e-06, "loss": 0.0088, "step": 53214 }, { "epoch": 16.37, "learning_rate": 1.6809118975234407e-06, "loss": 0.0056, "step": 53215 }, { "epoch": 16.37, "learning_rate": 1.6806354116403811e-06, "loss": 0.0045, "step": 53216 }, { "epoch": 16.37, "learning_rate": 1.680358946412064e-06, "loss": 0.0065, "step": 53217 }, { "epoch": 16.37, "learning_rate": 1.6800825018391697e-06, "loss": 0.0042, "step": 53218 }, { "epoch": 16.37, "learning_rate": 1.6798060779223858e-06, "loss": 0.0083, "step": 53219 }, { "epoch": 16.38, "learning_rate": 1.6795296746624012e-06, "loss": 0.0087, "step": 53220 }, { "epoch": 16.38, "learning_rate": 1.6792532920599025e-06, "loss": 0.0063, "step": 53221 }, { "epoch": 16.38, "learning_rate": 1.6789769301155746e-06, "loss": 0.003, "step": 53222 }, { "epoch": 16.38, "learning_rate": 1.6787005888301045e-06, "loss": 0.0054, "step": 53223 }, { "epoch": 16.38, "learning_rate": 1.6784242682041763e-06, "loss": 0.0057, "step": 53224 }, { "epoch": 16.38, "learning_rate": 1.678147968238477e-06, "loss": 0.0086, "step": 53225 }, { "epoch": 16.38, "learning_rate": 1.6778716889336932e-06, "loss": 0.0071, "step": 53226 }, { "epoch": 16.38, "learning_rate": 1.6775954302905107e-06, "loss": 0.0108, "step": 53227 }, { "epoch": 16.38, "learning_rate": 1.6773191923096155e-06, "loss": 0.0032, "step": 53228 }, { "epoch": 16.38, "learning_rate": 1.6770429749916895e-06, "loss": 0.0046, "step": 53229 }, { "epoch": 16.38, "learning_rate": 1.6767667783374276e-06, "loss": 0.0053, "step": 53230 }, { "epoch": 16.38, "learning_rate": 1.6764906023475069e-06, "loss": 0.0068, "step": 53231 }, { "epoch": 16.38, "learning_rate": 1.6762144470226172e-06, "loss": 0.0128, "step": 53232 }, { "epoch": 16.38, "learning_rate": 1.6759383123634421e-06, "loss": 0.0062, "step": 53233 }, { "epoch": 16.38, "learning_rate": 1.6756621983706656e-06, "loss": 0.0044, "step": 53234 }, { "epoch": 16.38, "learning_rate": 1.6753861050449783e-06, "loss": 0.007, "step": 53235 }, { "epoch": 16.38, "learning_rate": 1.6751100323870627e-06, "loss": 0.0041, "step": 53236 }, { "epoch": 16.38, "learning_rate": 1.6748339803976054e-06, "loss": 0.0067, "step": 53237 }, { "epoch": 16.38, "learning_rate": 1.6745579490772922e-06, "loss": 0.0108, "step": 53238 }, { "epoch": 16.38, "learning_rate": 1.674281938426804e-06, "loss": 0.0031, "step": 53239 }, { "epoch": 16.38, "learning_rate": 1.6740059484468295e-06, "loss": 0.0037, "step": 53240 }, { "epoch": 16.38, "learning_rate": 1.6737299791380558e-06, "loss": 0.0057, "step": 53241 }, { "epoch": 16.38, "learning_rate": 1.6734540305011638e-06, "loss": 0.0038, "step": 53242 }, { "epoch": 16.38, "learning_rate": 1.6731781025368422e-06, "loss": 0.0058, "step": 53243 }, { "epoch": 16.38, "learning_rate": 1.6729021952457735e-06, "loss": 0.0057, "step": 53244 }, { "epoch": 16.38, "learning_rate": 1.6726263086286444e-06, "loss": 0.0175, "step": 53245 }, { "epoch": 16.38, "learning_rate": 1.6723504426861391e-06, "loss": 0.0079, "step": 53246 }, { "epoch": 16.38, "learning_rate": 1.6720745974189422e-06, "loss": 0.0056, "step": 53247 }, { "epoch": 16.38, "learning_rate": 1.6717987728277397e-06, "loss": 0.0062, "step": 53248 }, { "epoch": 16.38, "learning_rate": 1.6715229689132128e-06, "loss": 0.0097, "step": 53249 }, { "epoch": 16.38, "learning_rate": 1.6712471856760515e-06, "loss": 0.0075, "step": 53250 }, { "epoch": 16.38, "learning_rate": 1.6709714231169405e-06, "loss": 0.0061, "step": 53251 }, { "epoch": 16.38, "learning_rate": 1.6706956812365594e-06, "loss": 0.0046, "step": 53252 }, { "epoch": 16.39, "learning_rate": 1.6704199600355952e-06, "loss": 0.0081, "step": 53253 }, { "epoch": 16.39, "learning_rate": 1.670144259514731e-06, "loss": 0.0038, "step": 53254 }, { "epoch": 16.39, "learning_rate": 1.669868579674655e-06, "loss": 0.0045, "step": 53255 }, { "epoch": 16.39, "learning_rate": 1.669592920516049e-06, "loss": 0.0051, "step": 53256 }, { "epoch": 16.39, "learning_rate": 1.6693172820395986e-06, "loss": 0.0062, "step": 53257 }, { "epoch": 16.39, "learning_rate": 1.6690416642459872e-06, "loss": 0.0068, "step": 53258 }, { "epoch": 16.39, "learning_rate": 1.6687660671358997e-06, "loss": 0.0061, "step": 53259 }, { "epoch": 16.39, "learning_rate": 1.6684904907100207e-06, "loss": 0.0111, "step": 53260 }, { "epoch": 16.39, "learning_rate": 1.6682149349690325e-06, "loss": 0.0063, "step": 53261 }, { "epoch": 16.39, "learning_rate": 1.6679393999136206e-06, "loss": 0.004, "step": 53262 }, { "epoch": 16.39, "learning_rate": 1.6676638855444693e-06, "loss": 0.0043, "step": 53263 }, { "epoch": 16.39, "learning_rate": 1.66738839186226e-06, "loss": 0.0039, "step": 53264 }, { "epoch": 16.39, "learning_rate": 1.6671129188676838e-06, "loss": 0.005, "step": 53265 }, { "epoch": 16.39, "learning_rate": 1.6668374665614162e-06, "loss": 0.0045, "step": 53266 }, { "epoch": 16.39, "learning_rate": 1.6665620349441459e-06, "loss": 0.0054, "step": 53267 }, { "epoch": 16.39, "learning_rate": 1.6662866240165555e-06, "loss": 0.0038, "step": 53268 }, { "epoch": 16.39, "learning_rate": 1.6660112337793256e-06, "loss": 0.0054, "step": 53269 }, { "epoch": 16.39, "learning_rate": 1.6657358642331457e-06, "loss": 0.0072, "step": 53270 }, { "epoch": 16.39, "learning_rate": 1.6654605153786961e-06, "loss": 0.0121, "step": 53271 }, { "epoch": 16.39, "learning_rate": 1.6651851872166646e-06, "loss": 0.0038, "step": 53272 }, { "epoch": 16.39, "learning_rate": 1.6649098797477282e-06, "loss": 0.0055, "step": 53273 }, { "epoch": 16.39, "learning_rate": 1.6646345929725705e-06, "loss": 0.0093, "step": 53274 }, { "epoch": 16.39, "learning_rate": 1.6643593268918812e-06, "loss": 0.0053, "step": 53275 }, { "epoch": 16.39, "learning_rate": 1.6640840815063409e-06, "loss": 0.0076, "step": 53276 }, { "epoch": 16.39, "learning_rate": 1.663808856816632e-06, "loss": 0.0058, "step": 53277 }, { "epoch": 16.39, "learning_rate": 1.6635336528234381e-06, "loss": 0.0041, "step": 53278 }, { "epoch": 16.39, "learning_rate": 1.6632584695274434e-06, "loss": 0.0032, "step": 53279 }, { "epoch": 16.39, "learning_rate": 1.66298330692933e-06, "loss": 0.0048, "step": 53280 }, { "epoch": 16.39, "learning_rate": 1.6627081650297815e-06, "loss": 0.009, "step": 53281 }, { "epoch": 16.39, "learning_rate": 1.662433043829481e-06, "loss": 0.006, "step": 53282 }, { "epoch": 16.39, "learning_rate": 1.6621579433291112e-06, "loss": 0.008, "step": 53283 }, { "epoch": 16.39, "learning_rate": 1.6618828635293538e-06, "loss": 0.0071, "step": 53284 }, { "epoch": 16.4, "learning_rate": 1.6616078044308948e-06, "loss": 0.0065, "step": 53285 }, { "epoch": 16.4, "learning_rate": 1.661332766034418e-06, "loss": 0.0052, "step": 53286 }, { "epoch": 16.4, "learning_rate": 1.661057748340602e-06, "loss": 0.0071, "step": 53287 }, { "epoch": 16.4, "learning_rate": 1.6607827513501317e-06, "loss": 0.0033, "step": 53288 }, { "epoch": 16.4, "learning_rate": 1.6605077750636867e-06, "loss": 0.0062, "step": 53289 }, { "epoch": 16.4, "learning_rate": 1.6602328194819562e-06, "loss": 0.0091, "step": 53290 }, { "epoch": 16.4, "learning_rate": 1.6599578846056185e-06, "loss": 0.0046, "step": 53291 }, { "epoch": 16.4, "learning_rate": 1.6596829704353567e-06, "loss": 0.0064, "step": 53292 }, { "epoch": 16.4, "learning_rate": 1.6594080769718558e-06, "loss": 0.0095, "step": 53293 }, { "epoch": 16.4, "learning_rate": 1.6591332042157915e-06, "loss": 0.0047, "step": 53294 }, { "epoch": 16.4, "learning_rate": 1.6588583521678536e-06, "loss": 0.0078, "step": 53295 }, { "epoch": 16.4, "learning_rate": 1.6585835208287216e-06, "loss": 0.0091, "step": 53296 }, { "epoch": 16.4, "learning_rate": 1.658308710199078e-06, "loss": 0.0062, "step": 53297 }, { "epoch": 16.4, "learning_rate": 1.6580339202796048e-06, "loss": 0.0044, "step": 53298 }, { "epoch": 16.4, "learning_rate": 1.6577591510709845e-06, "loss": 0.0031, "step": 53299 }, { "epoch": 16.4, "learning_rate": 1.6574844025739002e-06, "loss": 0.0103, "step": 53300 }, { "epoch": 16.4, "learning_rate": 1.657209674789031e-06, "loss": 0.0041, "step": 53301 }, { "epoch": 16.4, "learning_rate": 1.6569349677170632e-06, "loss": 0.0025, "step": 53302 }, { "epoch": 16.4, "learning_rate": 1.6566602813586752e-06, "loss": 0.0062, "step": 53303 }, { "epoch": 16.4, "learning_rate": 1.6563856157145487e-06, "loss": 0.0045, "step": 53304 }, { "epoch": 16.4, "learning_rate": 1.6561109707853695e-06, "loss": 0.0057, "step": 53305 }, { "epoch": 16.4, "learning_rate": 1.6558363465718197e-06, "loss": 0.0062, "step": 53306 }, { "epoch": 16.4, "learning_rate": 1.6555617430745763e-06, "loss": 0.0057, "step": 53307 }, { "epoch": 16.4, "learning_rate": 1.6552871602943233e-06, "loss": 0.005, "step": 53308 }, { "epoch": 16.4, "learning_rate": 1.65501259823174e-06, "loss": 0.0081, "step": 53309 }, { "epoch": 16.4, "learning_rate": 1.6547380568875137e-06, "loss": 0.0083, "step": 53310 }, { "epoch": 16.4, "learning_rate": 1.654463536262323e-06, "loss": 0.0043, "step": 53311 }, { "epoch": 16.4, "learning_rate": 1.654189036356849e-06, "loss": 0.005, "step": 53312 }, { "epoch": 16.4, "learning_rate": 1.653914557171774e-06, "loss": 0.005, "step": 53313 }, { "epoch": 16.4, "learning_rate": 1.6536400987077794e-06, "loss": 0.0085, "step": 53314 }, { "epoch": 16.4, "learning_rate": 1.6533656609655456e-06, "loss": 0.0057, "step": 53315 }, { "epoch": 16.4, "learning_rate": 1.653091243945756e-06, "loss": 0.0069, "step": 53316 }, { "epoch": 16.4, "learning_rate": 1.652816847649089e-06, "loss": 0.0027, "step": 53317 }, { "epoch": 16.41, "learning_rate": 1.6525424720762285e-06, "loss": 0.0089, "step": 53318 }, { "epoch": 16.41, "learning_rate": 1.652268117227851e-06, "loss": 0.0041, "step": 53319 }, { "epoch": 16.41, "learning_rate": 1.6519937831046473e-06, "loss": 0.0039, "step": 53320 }, { "epoch": 16.41, "learning_rate": 1.6517194697072903e-06, "loss": 0.0038, "step": 53321 }, { "epoch": 16.41, "learning_rate": 1.6514451770364625e-06, "loss": 0.0104, "step": 53322 }, { "epoch": 16.41, "learning_rate": 1.651170905092846e-06, "loss": 0.0041, "step": 53323 }, { "epoch": 16.41, "learning_rate": 1.6508966538771176e-06, "loss": 0.0034, "step": 53324 }, { "epoch": 16.41, "learning_rate": 1.650622423389966e-06, "loss": 0.0079, "step": 53325 }, { "epoch": 16.41, "learning_rate": 1.650348213632068e-06, "loss": 0.0055, "step": 53326 }, { "epoch": 16.41, "learning_rate": 1.650074024604106e-06, "loss": 0.0038, "step": 53327 }, { "epoch": 16.41, "learning_rate": 1.6497998563067563e-06, "loss": 0.0068, "step": 53328 }, { "epoch": 16.41, "learning_rate": 1.6495257087407012e-06, "loss": 0.0056, "step": 53329 }, { "epoch": 16.41, "learning_rate": 1.6492515819066245e-06, "loss": 0.003, "step": 53330 }, { "epoch": 16.41, "learning_rate": 1.6489774758052047e-06, "loss": 0.0049, "step": 53331 }, { "epoch": 16.41, "learning_rate": 1.648703390437123e-06, "loss": 0.0076, "step": 53332 }, { "epoch": 16.41, "learning_rate": 1.648429325803058e-06, "loss": 0.0066, "step": 53333 }, { "epoch": 16.41, "learning_rate": 1.6481552819036927e-06, "loss": 0.0069, "step": 53334 }, { "epoch": 16.41, "learning_rate": 1.6478812587397064e-06, "loss": 0.0022, "step": 53335 }, { "epoch": 16.41, "learning_rate": 1.6476072563117784e-06, "loss": 0.0075, "step": 53336 }, { "epoch": 16.41, "learning_rate": 1.6473332746205906e-06, "loss": 0.0052, "step": 53337 }, { "epoch": 16.41, "learning_rate": 1.6470593136668222e-06, "loss": 0.0053, "step": 53338 }, { "epoch": 16.41, "learning_rate": 1.6467853734511508e-06, "loss": 0.0067, "step": 53339 }, { "epoch": 16.41, "learning_rate": 1.6465114539742621e-06, "loss": 0.0056, "step": 53340 }, { "epoch": 16.41, "learning_rate": 1.6462375552368359e-06, "loss": 0.0034, "step": 53341 }, { "epoch": 16.41, "learning_rate": 1.6459636772395471e-06, "loss": 0.0076, "step": 53342 }, { "epoch": 16.41, "learning_rate": 1.6456898199830785e-06, "loss": 0.0058, "step": 53343 }, { "epoch": 16.41, "learning_rate": 1.645415983468107e-06, "loss": 0.0054, "step": 53344 }, { "epoch": 16.41, "learning_rate": 1.6451421676953182e-06, "loss": 0.0038, "step": 53345 }, { "epoch": 16.41, "learning_rate": 1.6448683726653892e-06, "loss": 0.0037, "step": 53346 }, { "epoch": 16.41, "learning_rate": 1.644594598378999e-06, "loss": 0.006, "step": 53347 }, { "epoch": 16.41, "learning_rate": 1.644320844836831e-06, "loss": 0.0024, "step": 53348 }, { "epoch": 16.41, "learning_rate": 1.6440471120395563e-06, "loss": 0.004, "step": 53349 }, { "epoch": 16.42, "learning_rate": 1.6437733999878626e-06, "loss": 0.0092, "step": 53350 }, { "epoch": 16.42, "learning_rate": 1.6434997086824267e-06, "loss": 0.0042, "step": 53351 }, { "epoch": 16.42, "learning_rate": 1.6432260381239285e-06, "loss": 0.0038, "step": 53352 }, { "epoch": 16.42, "learning_rate": 1.642952388313046e-06, "loss": 0.0115, "step": 53353 }, { "epoch": 16.42, "learning_rate": 1.6426787592504612e-06, "loss": 0.0055, "step": 53354 }, { "epoch": 16.42, "learning_rate": 1.642405150936851e-06, "loss": 0.0063, "step": 53355 }, { "epoch": 16.42, "learning_rate": 1.6421315633728952e-06, "loss": 0.0061, "step": 53356 }, { "epoch": 16.42, "learning_rate": 1.641857996559275e-06, "loss": 0.0107, "step": 53357 }, { "epoch": 16.42, "learning_rate": 1.641584450496667e-06, "loss": 0.0102, "step": 53358 }, { "epoch": 16.42, "learning_rate": 1.6413109251857494e-06, "loss": 0.0055, "step": 53359 }, { "epoch": 16.42, "learning_rate": 1.6410374206272061e-06, "loss": 0.0047, "step": 53360 }, { "epoch": 16.42, "learning_rate": 1.6407639368217133e-06, "loss": 0.0066, "step": 53361 }, { "epoch": 16.42, "learning_rate": 1.6404904737699511e-06, "loss": 0.0065, "step": 53362 }, { "epoch": 16.42, "learning_rate": 1.6402170314725964e-06, "loss": 0.0059, "step": 53363 }, { "epoch": 16.42, "learning_rate": 1.6399436099303269e-06, "loss": 0.004, "step": 53364 }, { "epoch": 16.42, "learning_rate": 1.6396702091438255e-06, "loss": 0.0074, "step": 53365 }, { "epoch": 16.42, "learning_rate": 1.6393968291137686e-06, "loss": 0.0052, "step": 53366 }, { "epoch": 16.42, "learning_rate": 1.6391234698408365e-06, "loss": 0.005, "step": 53367 }, { "epoch": 16.42, "learning_rate": 1.6388501313257055e-06, "loss": 0.0065, "step": 53368 }, { "epoch": 16.42, "learning_rate": 1.6385768135690561e-06, "loss": 0.0052, "step": 53369 }, { "epoch": 16.42, "learning_rate": 1.6383035165715667e-06, "loss": 0.0099, "step": 53370 }, { "epoch": 16.42, "learning_rate": 1.6380302403339155e-06, "loss": 0.005, "step": 53371 }, { "epoch": 16.42, "learning_rate": 1.6377569848567798e-06, "loss": 0.0046, "step": 53372 }, { "epoch": 16.42, "learning_rate": 1.6374837501408403e-06, "loss": 0.0071, "step": 53373 }, { "epoch": 16.42, "learning_rate": 1.637210536186772e-06, "loss": 0.0095, "step": 53374 }, { "epoch": 16.42, "learning_rate": 1.6369373429952596e-06, "loss": 0.0047, "step": 53375 }, { "epoch": 16.42, "learning_rate": 1.636664170566974e-06, "loss": 0.0101, "step": 53376 }, { "epoch": 16.42, "learning_rate": 1.6363910189025978e-06, "loss": 0.004, "step": 53377 }, { "epoch": 16.42, "learning_rate": 1.6361178880028073e-06, "loss": 0.0067, "step": 53378 }, { "epoch": 16.42, "learning_rate": 1.6358447778682784e-06, "loss": 0.0044, "step": 53379 }, { "epoch": 16.42, "learning_rate": 1.6355716884996953e-06, "loss": 0.0048, "step": 53380 }, { "epoch": 16.42, "learning_rate": 1.6352986198977327e-06, "loss": 0.005, "step": 53381 }, { "epoch": 16.42, "learning_rate": 1.6350255720630682e-06, "loss": 0.0051, "step": 53382 }, { "epoch": 16.43, "learning_rate": 1.634752544996382e-06, "loss": 0.0036, "step": 53383 }, { "epoch": 16.43, "learning_rate": 1.634479538698346e-06, "loss": 0.0092, "step": 53384 }, { "epoch": 16.43, "learning_rate": 1.634206553169645e-06, "loss": 0.0068, "step": 53385 }, { "epoch": 16.43, "learning_rate": 1.633933588410952e-06, "loss": 0.0083, "step": 53386 }, { "epoch": 16.43, "learning_rate": 1.6336606444229476e-06, "loss": 0.0063, "step": 53387 }, { "epoch": 16.43, "learning_rate": 1.6333877212063088e-06, "loss": 0.0072, "step": 53388 }, { "epoch": 16.43, "learning_rate": 1.6331148187617118e-06, "loss": 0.0046, "step": 53389 }, { "epoch": 16.43, "learning_rate": 1.632841937089835e-06, "loss": 0.0106, "step": 53390 }, { "epoch": 16.43, "learning_rate": 1.6325690761913571e-06, "loss": 0.0052, "step": 53391 }, { "epoch": 16.43, "learning_rate": 1.6322962360669536e-06, "loss": 0.0056, "step": 53392 }, { "epoch": 16.43, "learning_rate": 1.632023416717302e-06, "loss": 0.0051, "step": 53393 }, { "epoch": 16.43, "learning_rate": 1.6317506181430787e-06, "loss": 0.0051, "step": 53394 }, { "epoch": 16.43, "learning_rate": 1.631477840344965e-06, "loss": 0.0043, "step": 53395 }, { "epoch": 16.43, "learning_rate": 1.6312050833236382e-06, "loss": 0.0041, "step": 53396 }, { "epoch": 16.43, "learning_rate": 1.630932347079771e-06, "loss": 0.0051, "step": 53397 }, { "epoch": 16.43, "learning_rate": 1.63065963161404e-06, "loss": 0.0081, "step": 53398 }, { "epoch": 16.43, "learning_rate": 1.6303869369271264e-06, "loss": 0.0064, "step": 53399 }, { "epoch": 16.43, "learning_rate": 1.6301142630197065e-06, "loss": 0.0047, "step": 53400 }, { "epoch": 16.43, "learning_rate": 1.6298416098924552e-06, "loss": 0.0037, "step": 53401 }, { "epoch": 16.43, "learning_rate": 1.6295689775460521e-06, "loss": 0.0036, "step": 53402 }, { "epoch": 16.43, "learning_rate": 1.6292963659811723e-06, "loss": 0.0103, "step": 53403 }, { "epoch": 16.43, "learning_rate": 1.6290237751984928e-06, "loss": 0.006, "step": 53404 }, { "epoch": 16.43, "learning_rate": 1.6287512051986897e-06, "loss": 0.0051, "step": 53405 }, { "epoch": 16.43, "learning_rate": 1.6284786559824417e-06, "loss": 0.0063, "step": 53406 }, { "epoch": 16.43, "learning_rate": 1.6282061275504234e-06, "loss": 0.0041, "step": 53407 }, { "epoch": 16.43, "learning_rate": 1.6279336199033113e-06, "loss": 0.0132, "step": 53408 }, { "epoch": 16.43, "learning_rate": 1.627661133041787e-06, "loss": 0.0048, "step": 53409 }, { "epoch": 16.43, "learning_rate": 1.6273886669665195e-06, "loss": 0.0046, "step": 53410 }, { "epoch": 16.43, "learning_rate": 1.62711622167819e-06, "loss": 0.0055, "step": 53411 }, { "epoch": 16.43, "learning_rate": 1.626843797177473e-06, "loss": 0.0043, "step": 53412 }, { "epoch": 16.43, "learning_rate": 1.626571393465044e-06, "loss": 0.0045, "step": 53413 }, { "epoch": 16.43, "learning_rate": 1.6262990105415831e-06, "loss": 0.0038, "step": 53414 }, { "epoch": 16.44, "learning_rate": 1.6260266484077636e-06, "loss": 0.0042, "step": 53415 }, { "epoch": 16.44, "learning_rate": 1.6257543070642624e-06, "loss": 0.0072, "step": 53416 }, { "epoch": 16.44, "learning_rate": 1.625481986511759e-06, "loss": 0.0053, "step": 53417 }, { "epoch": 16.44, "learning_rate": 1.6252096867509204e-06, "loss": 0.0053, "step": 53418 }, { "epoch": 16.44, "learning_rate": 1.624937407782431e-06, "loss": 0.0064, "step": 53419 }, { "epoch": 16.44, "learning_rate": 1.6246651496069643e-06, "loss": 0.007, "step": 53420 }, { "epoch": 16.44, "learning_rate": 1.6243929122251955e-06, "loss": 0.006, "step": 53421 }, { "epoch": 16.44, "learning_rate": 1.6241206956378019e-06, "loss": 0.005, "step": 53422 }, { "epoch": 16.44, "learning_rate": 1.6238484998454584e-06, "loss": 0.0046, "step": 53423 }, { "epoch": 16.44, "learning_rate": 1.6235763248488402e-06, "loss": 0.0037, "step": 53424 }, { "epoch": 16.44, "learning_rate": 1.6233041706486253e-06, "loss": 0.0067, "step": 53425 }, { "epoch": 16.44, "learning_rate": 1.6230320372454867e-06, "loss": 0.0047, "step": 53426 }, { "epoch": 16.44, "learning_rate": 1.6227599246401016e-06, "loss": 0.0057, "step": 53427 }, { "epoch": 16.44, "learning_rate": 1.6224878328331416e-06, "loss": 0.0084, "step": 53428 }, { "epoch": 16.44, "learning_rate": 1.6222157618252898e-06, "loss": 0.0046, "step": 53429 }, { "epoch": 16.44, "learning_rate": 1.62194371161722e-06, "loss": 0.0052, "step": 53430 }, { "epoch": 16.44, "learning_rate": 1.6216716822096013e-06, "loss": 0.0051, "step": 53431 }, { "epoch": 16.44, "learning_rate": 1.6213996736031146e-06, "loss": 0.006, "step": 53432 }, { "epoch": 16.44, "learning_rate": 1.6211276857984304e-06, "loss": 0.0056, "step": 53433 }, { "epoch": 16.44, "learning_rate": 1.6208557187962303e-06, "loss": 0.0034, "step": 53434 }, { "epoch": 16.44, "learning_rate": 1.6205837725971862e-06, "loss": 0.0079, "step": 53435 }, { "epoch": 16.44, "learning_rate": 1.620311847201974e-06, "loss": 0.0059, "step": 53436 }, { "epoch": 16.44, "learning_rate": 1.6200399426112679e-06, "loss": 0.007, "step": 53437 }, { "epoch": 16.44, "learning_rate": 1.6197680588257435e-06, "loss": 0.0026, "step": 53438 }, { "epoch": 16.44, "learning_rate": 1.6194961958460764e-06, "loss": 0.0091, "step": 53439 }, { "epoch": 16.44, "learning_rate": 1.6192243536729402e-06, "loss": 0.0059, "step": 53440 }, { "epoch": 16.44, "learning_rate": 1.618952532307011e-06, "loss": 0.0052, "step": 53441 }, { "epoch": 16.44, "learning_rate": 1.6186807317489628e-06, "loss": 0.0095, "step": 53442 }, { "epoch": 16.44, "learning_rate": 1.6184089519994718e-06, "loss": 0.0021, "step": 53443 }, { "epoch": 16.44, "learning_rate": 1.6181371930592105e-06, "loss": 0.0093, "step": 53444 }, { "epoch": 16.44, "learning_rate": 1.6178654549288553e-06, "loss": 0.0066, "step": 53445 }, { "epoch": 16.44, "learning_rate": 1.617593737609081e-06, "loss": 0.0055, "step": 53446 }, { "epoch": 16.44, "learning_rate": 1.6173220411005608e-06, "loss": 0.0077, "step": 53447 }, { "epoch": 16.45, "learning_rate": 1.6170503654039682e-06, "loss": 0.0059, "step": 53448 }, { "epoch": 16.45, "learning_rate": 1.6167787105199816e-06, "loss": 0.0068, "step": 53449 }, { "epoch": 16.45, "learning_rate": 1.616507076449273e-06, "loss": 0.0054, "step": 53450 }, { "epoch": 16.45, "learning_rate": 1.6162354631925203e-06, "loss": 0.0072, "step": 53451 }, { "epoch": 16.45, "learning_rate": 1.615963870750392e-06, "loss": 0.011, "step": 53452 }, { "epoch": 16.45, "learning_rate": 1.6156922991235634e-06, "loss": 0.0033, "step": 53453 }, { "epoch": 16.45, "learning_rate": 1.6154207483127126e-06, "loss": 0.0047, "step": 53454 }, { "epoch": 16.45, "learning_rate": 1.6151492183185113e-06, "loss": 0.0035, "step": 53455 }, { "epoch": 16.45, "learning_rate": 1.6148777091416346e-06, "loss": 0.0039, "step": 53456 }, { "epoch": 16.45, "learning_rate": 1.614606220782755e-06, "loss": 0.0036, "step": 53457 }, { "epoch": 16.45, "learning_rate": 1.614334753242549e-06, "loss": 0.0034, "step": 53458 }, { "epoch": 16.45, "learning_rate": 1.6140633065216882e-06, "loss": 0.003, "step": 53459 }, { "epoch": 16.45, "learning_rate": 1.6137918806208475e-06, "loss": 0.004, "step": 53460 }, { "epoch": 16.45, "learning_rate": 1.6135204755407009e-06, "loss": 0.0039, "step": 53461 }, { "epoch": 16.45, "learning_rate": 1.613249091281921e-06, "loss": 0.0033, "step": 53462 }, { "epoch": 16.45, "learning_rate": 1.6129777278451807e-06, "loss": 0.0077, "step": 53463 }, { "epoch": 16.45, "learning_rate": 1.6127063852311608e-06, "loss": 0.0051, "step": 53464 }, { "epoch": 16.45, "learning_rate": 1.6124350634405273e-06, "loss": 0.0035, "step": 53465 }, { "epoch": 16.45, "learning_rate": 1.6121637624739573e-06, "loss": 0.007, "step": 53466 }, { "epoch": 16.45, "learning_rate": 1.6118924823321225e-06, "loss": 0.0051, "step": 53467 }, { "epoch": 16.45, "learning_rate": 1.6116212230156946e-06, "loss": 0.0037, "step": 53468 }, { "epoch": 16.45, "learning_rate": 1.6113499845253523e-06, "loss": 0.007, "step": 53469 }, { "epoch": 16.45, "learning_rate": 1.6110787668617666e-06, "loss": 0.0065, "step": 53470 }, { "epoch": 16.45, "learning_rate": 1.610807570025611e-06, "loss": 0.0042, "step": 53471 }, { "epoch": 16.45, "learning_rate": 1.61053639401756e-06, "loss": 0.0046, "step": 53472 }, { "epoch": 16.45, "learning_rate": 1.6102652388382812e-06, "loss": 0.0046, "step": 53473 }, { "epoch": 16.45, "learning_rate": 1.609994104488455e-06, "loss": 0.006, "step": 53474 }, { "epoch": 16.45, "learning_rate": 1.6097229909687507e-06, "loss": 0.0031, "step": 53475 }, { "epoch": 16.45, "learning_rate": 1.6094518982798423e-06, "loss": 0.0062, "step": 53476 }, { "epoch": 16.45, "learning_rate": 1.609180826422404e-06, "loss": 0.0047, "step": 53477 }, { "epoch": 16.45, "learning_rate": 1.6089097753971061e-06, "loss": 0.0034, "step": 53478 }, { "epoch": 16.45, "learning_rate": 1.6086387452046237e-06, "loss": 0.011, "step": 53479 }, { "epoch": 16.46, "learning_rate": 1.6083677358456295e-06, "loss": 0.0043, "step": 53480 }, { "epoch": 16.46, "learning_rate": 1.6080967473207954e-06, "loss": 0.0028, "step": 53481 }, { "epoch": 16.46, "learning_rate": 1.607825779630795e-06, "loss": 0.004, "step": 53482 }, { "epoch": 16.46, "learning_rate": 1.607554832776298e-06, "loss": 0.0113, "step": 53483 }, { "epoch": 16.46, "learning_rate": 1.6072839067579826e-06, "loss": 0.0045, "step": 53484 }, { "epoch": 16.46, "learning_rate": 1.6070130015765206e-06, "loss": 0.005, "step": 53485 }, { "epoch": 16.46, "learning_rate": 1.6067421172325804e-06, "loss": 0.0084, "step": 53486 }, { "epoch": 16.46, "learning_rate": 1.6064712537268369e-06, "loss": 0.0074, "step": 53487 }, { "epoch": 16.46, "learning_rate": 1.6062004110599594e-06, "loss": 0.0065, "step": 53488 }, { "epoch": 16.46, "learning_rate": 1.6059295892326266e-06, "loss": 0.0052, "step": 53489 }, { "epoch": 16.46, "learning_rate": 1.6056587882455078e-06, "loss": 0.0037, "step": 53490 }, { "epoch": 16.46, "learning_rate": 1.6053880080992746e-06, "loss": 0.0053, "step": 53491 }, { "epoch": 16.46, "learning_rate": 1.6051172487946e-06, "loss": 0.0057, "step": 53492 }, { "epoch": 16.46, "learning_rate": 1.6048465103321553e-06, "loss": 0.0051, "step": 53493 }, { "epoch": 16.46, "learning_rate": 1.6045757927126149e-06, "loss": 0.005, "step": 53494 }, { "epoch": 16.46, "learning_rate": 1.604305095936648e-06, "loss": 0.0053, "step": 53495 }, { "epoch": 16.46, "learning_rate": 1.6040344200049297e-06, "loss": 0.007, "step": 53496 }, { "epoch": 16.46, "learning_rate": 1.6037637649181291e-06, "loss": 0.0045, "step": 53497 }, { "epoch": 16.46, "learning_rate": 1.6034931306769174e-06, "loss": 0.0108, "step": 53498 }, { "epoch": 16.46, "learning_rate": 1.6032225172819737e-06, "loss": 0.0069, "step": 53499 }, { "epoch": 16.46, "learning_rate": 1.6029519247339632e-06, "loss": 0.0046, "step": 53500 }, { "epoch": 16.46, "learning_rate": 1.6026813530335583e-06, "loss": 0.0013, "step": 53501 }, { "epoch": 16.46, "learning_rate": 1.6024108021814322e-06, "loss": 0.0037, "step": 53502 }, { "epoch": 16.46, "learning_rate": 1.602140272178253e-06, "loss": 0.0043, "step": 53503 }, { "epoch": 16.46, "learning_rate": 1.6018697630246982e-06, "loss": 0.005, "step": 53504 }, { "epoch": 16.46, "learning_rate": 1.601599274721437e-06, "loss": 0.0037, "step": 53505 }, { "epoch": 16.46, "learning_rate": 1.6013288072691435e-06, "loss": 0.0098, "step": 53506 }, { "epoch": 16.46, "learning_rate": 1.6010583606684836e-06, "loss": 0.0084, "step": 53507 }, { "epoch": 16.46, "learning_rate": 1.6007879349201294e-06, "loss": 0.0079, "step": 53508 }, { "epoch": 16.46, "learning_rate": 1.6005175300247554e-06, "loss": 0.004, "step": 53509 }, { "epoch": 16.46, "learning_rate": 1.6002471459830338e-06, "loss": 0.005, "step": 53510 }, { "epoch": 16.46, "learning_rate": 1.5999767827956324e-06, "loss": 0.0057, "step": 53511 }, { "epoch": 16.46, "learning_rate": 1.5997064404632256e-06, "loss": 0.0049, "step": 53512 }, { "epoch": 16.47, "learning_rate": 1.5994361189864815e-06, "loss": 0.0058, "step": 53513 }, { "epoch": 16.47, "learning_rate": 1.599165818366074e-06, "loss": 0.0053, "step": 53514 }, { "epoch": 16.47, "learning_rate": 1.5988955386026728e-06, "loss": 0.006, "step": 53515 }, { "epoch": 16.47, "learning_rate": 1.5986252796969482e-06, "loss": 0.0125, "step": 53516 }, { "epoch": 16.47, "learning_rate": 1.598355041649573e-06, "loss": 0.0044, "step": 53517 }, { "epoch": 16.47, "learning_rate": 1.5980848244612145e-06, "loss": 0.0079, "step": 53518 }, { "epoch": 16.47, "learning_rate": 1.5978146281325491e-06, "loss": 0.0065, "step": 53519 }, { "epoch": 16.47, "learning_rate": 1.5975444526642469e-06, "loss": 0.0054, "step": 53520 }, { "epoch": 16.47, "learning_rate": 1.5972742980569745e-06, "loss": 0.005, "step": 53521 }, { "epoch": 16.47, "learning_rate": 1.5970041643114042e-06, "loss": 0.0074, "step": 53522 }, { "epoch": 16.47, "learning_rate": 1.5967340514282049e-06, "loss": 0.0046, "step": 53523 }, { "epoch": 16.47, "learning_rate": 1.5964639594080523e-06, "loss": 0.0054, "step": 53524 }, { "epoch": 16.47, "learning_rate": 1.5961938882516136e-06, "loss": 0.0052, "step": 53525 }, { "epoch": 16.47, "learning_rate": 1.5959238379595598e-06, "loss": 0.014, "step": 53526 }, { "epoch": 16.47, "learning_rate": 1.5956538085325645e-06, "loss": 0.0066, "step": 53527 }, { "epoch": 16.47, "learning_rate": 1.5953837999712895e-06, "loss": 0.0036, "step": 53528 }, { "epoch": 16.47, "learning_rate": 1.5951138122764132e-06, "loss": 0.0057, "step": 53529 }, { "epoch": 16.47, "learning_rate": 1.5948438454486038e-06, "loss": 0.0042, "step": 53530 }, { "epoch": 16.47, "learning_rate": 1.5945738994885318e-06, "loss": 0.0052, "step": 53531 }, { "epoch": 16.47, "learning_rate": 1.5943039743968658e-06, "loss": 0.0069, "step": 53532 }, { "epoch": 16.47, "learning_rate": 1.5940340701742773e-06, "loss": 0.0045, "step": 53533 }, { "epoch": 16.47, "learning_rate": 1.593764186821436e-06, "loss": 0.0024, "step": 53534 }, { "epoch": 16.47, "learning_rate": 1.5934943243390123e-06, "loss": 0.0057, "step": 53535 }, { "epoch": 16.47, "learning_rate": 1.5932244827276756e-06, "loss": 0.0046, "step": 53536 }, { "epoch": 16.47, "learning_rate": 1.5929546619880953e-06, "loss": 0.0051, "step": 53537 }, { "epoch": 16.47, "learning_rate": 1.5926848621209402e-06, "loss": 0.0041, "step": 53538 }, { "epoch": 16.47, "learning_rate": 1.5924150831268847e-06, "loss": 0.0056, "step": 53539 }, { "epoch": 16.47, "learning_rate": 1.5921453250065966e-06, "loss": 0.0098, "step": 53540 }, { "epoch": 16.47, "learning_rate": 1.5918755877607461e-06, "loss": 0.0041, "step": 53541 }, { "epoch": 16.47, "learning_rate": 1.5916058713899995e-06, "loss": 0.006, "step": 53542 }, { "epoch": 16.47, "learning_rate": 1.5913361758950275e-06, "loss": 0.0084, "step": 53543 }, { "epoch": 16.47, "learning_rate": 1.5910665012765025e-06, "loss": 0.0088, "step": 53544 }, { "epoch": 16.48, "learning_rate": 1.590796847535092e-06, "loss": 0.0039, "step": 53545 }, { "epoch": 16.48, "learning_rate": 1.5905272146714668e-06, "loss": 0.0118, "step": 53546 }, { "epoch": 16.48, "learning_rate": 1.5902576026862949e-06, "loss": 0.0066, "step": 53547 }, { "epoch": 16.48, "learning_rate": 1.5899880115802458e-06, "loss": 0.0061, "step": 53548 }, { "epoch": 16.48, "learning_rate": 1.589718441353989e-06, "loss": 0.0072, "step": 53549 }, { "epoch": 16.48, "learning_rate": 1.5894488920081951e-06, "loss": 0.0071, "step": 53550 }, { "epoch": 16.48, "learning_rate": 1.5891793635435315e-06, "loss": 0.0128, "step": 53551 }, { "epoch": 16.48, "learning_rate": 1.5889098559606675e-06, "loss": 0.0042, "step": 53552 }, { "epoch": 16.48, "learning_rate": 1.5886403692602713e-06, "loss": 0.0042, "step": 53553 }, { "epoch": 16.48, "learning_rate": 1.5883709034430183e-06, "loss": 0.0062, "step": 53554 }, { "epoch": 16.48, "learning_rate": 1.58810145850957e-06, "loss": 0.0036, "step": 53555 }, { "epoch": 16.48, "learning_rate": 1.5878320344605968e-06, "loss": 0.0043, "step": 53556 }, { "epoch": 16.48, "learning_rate": 1.5875626312967706e-06, "loss": 0.0065, "step": 53557 }, { "epoch": 16.48, "learning_rate": 1.5872932490187554e-06, "loss": 0.0103, "step": 53558 }, { "epoch": 16.48, "learning_rate": 1.5870238876272249e-06, "loss": 0.0125, "step": 53559 }, { "epoch": 16.48, "learning_rate": 1.5867545471228462e-06, "loss": 0.0081, "step": 53560 }, { "epoch": 16.48, "learning_rate": 1.5864852275062892e-06, "loss": 0.004, "step": 53561 }, { "epoch": 16.48, "learning_rate": 1.586215928778222e-06, "loss": 0.0037, "step": 53562 }, { "epoch": 16.48, "learning_rate": 1.5859466509393085e-06, "loss": 0.0037, "step": 53563 }, { "epoch": 16.48, "learning_rate": 1.5856773939902226e-06, "loss": 0.0032, "step": 53564 }, { "epoch": 16.48, "learning_rate": 1.5854081579316316e-06, "loss": 0.0051, "step": 53565 }, { "epoch": 16.48, "learning_rate": 1.5851389427642038e-06, "loss": 0.0038, "step": 53566 }, { "epoch": 16.48, "learning_rate": 1.5848697484886078e-06, "loss": 0.0053, "step": 53567 }, { "epoch": 16.48, "learning_rate": 1.5846005751055116e-06, "loss": 0.0045, "step": 53568 }, { "epoch": 16.48, "learning_rate": 1.5843314226155825e-06, "loss": 0.0056, "step": 53569 }, { "epoch": 16.48, "learning_rate": 1.5840622910194903e-06, "loss": 0.0039, "step": 53570 }, { "epoch": 16.48, "learning_rate": 1.583793180317903e-06, "loss": 0.0092, "step": 53571 }, { "epoch": 16.48, "learning_rate": 1.5835240905114869e-06, "loss": 0.0039, "step": 53572 }, { "epoch": 16.48, "learning_rate": 1.58325502160091e-06, "loss": 0.0035, "step": 53573 }, { "epoch": 16.48, "learning_rate": 1.5829859735868447e-06, "loss": 0.0071, "step": 53574 }, { "epoch": 16.48, "learning_rate": 1.5827169464699576e-06, "loss": 0.0066, "step": 53575 }, { "epoch": 16.48, "learning_rate": 1.582447940250913e-06, "loss": 0.0071, "step": 53576 }, { "epoch": 16.48, "learning_rate": 1.5821789549303812e-06, "loss": 0.0034, "step": 53577 }, { "epoch": 16.49, "learning_rate": 1.5819099905090273e-06, "loss": 0.0051, "step": 53578 }, { "epoch": 16.49, "learning_rate": 1.5816410469875232e-06, "loss": 0.006, "step": 53579 }, { "epoch": 16.49, "learning_rate": 1.5813721243665348e-06, "loss": 0.0045, "step": 53580 }, { "epoch": 16.49, "learning_rate": 1.5811032226467304e-06, "loss": 0.0025, "step": 53581 }, { "epoch": 16.49, "learning_rate": 1.5808343418287763e-06, "loss": 0.0063, "step": 53582 }, { "epoch": 16.49, "learning_rate": 1.580565481913342e-06, "loss": 0.007, "step": 53583 }, { "epoch": 16.49, "learning_rate": 1.5802966429010936e-06, "loss": 0.0049, "step": 53584 }, { "epoch": 16.49, "learning_rate": 1.5800278247926982e-06, "loss": 0.0028, "step": 53585 }, { "epoch": 16.49, "learning_rate": 1.579759027588823e-06, "loss": 0.0063, "step": 53586 }, { "epoch": 16.49, "learning_rate": 1.5794902512901378e-06, "loss": 0.0043, "step": 53587 }, { "epoch": 16.49, "learning_rate": 1.5792214958973073e-06, "loss": 0.004, "step": 53588 }, { "epoch": 16.49, "learning_rate": 1.578952761410999e-06, "loss": 0.006, "step": 53589 }, { "epoch": 16.49, "learning_rate": 1.5786840478318821e-06, "loss": 0.0049, "step": 53590 }, { "epoch": 16.49, "learning_rate": 1.578415355160622e-06, "loss": 0.0067, "step": 53591 }, { "epoch": 16.49, "learning_rate": 1.5781466833978853e-06, "loss": 0.0064, "step": 53592 }, { "epoch": 16.49, "learning_rate": 1.5778780325443378e-06, "loss": 0.0037, "step": 53593 }, { "epoch": 16.49, "learning_rate": 1.5776094026006516e-06, "loss": 0.0052, "step": 53594 }, { "epoch": 16.49, "learning_rate": 1.577340793567491e-06, "loss": 0.0047, "step": 53595 }, { "epoch": 16.49, "learning_rate": 1.5770722054455234e-06, "loss": 0.0075, "step": 53596 }, { "epoch": 16.49, "learning_rate": 1.5768036382354135e-06, "loss": 0.0028, "step": 53597 }, { "epoch": 16.49, "learning_rate": 1.5765350919378275e-06, "loss": 0.008, "step": 53598 }, { "epoch": 16.49, "learning_rate": 1.576266566553435e-06, "loss": 0.0086, "step": 53599 }, { "epoch": 16.49, "learning_rate": 1.575998062082903e-06, "loss": 0.0074, "step": 53600 }, { "epoch": 16.49, "learning_rate": 1.5757295785268955e-06, "loss": 0.0034, "step": 53601 }, { "epoch": 16.49, "learning_rate": 1.575461115886081e-06, "loss": 0.0063, "step": 53602 }, { "epoch": 16.49, "learning_rate": 1.5751926741611256e-06, "loss": 0.0055, "step": 53603 }, { "epoch": 16.49, "learning_rate": 1.5749242533526943e-06, "loss": 0.003, "step": 53604 }, { "epoch": 16.49, "learning_rate": 1.5746558534614565e-06, "loss": 0.0055, "step": 53605 }, { "epoch": 16.49, "learning_rate": 1.574387474488076e-06, "loss": 0.0038, "step": 53606 }, { "epoch": 16.49, "learning_rate": 1.5741191164332192e-06, "loss": 0.0056, "step": 53607 }, { "epoch": 16.49, "learning_rate": 1.5738507792975522e-06, "loss": 0.0054, "step": 53608 }, { "epoch": 16.49, "learning_rate": 1.5735824630817453e-06, "loss": 0.0063, "step": 53609 }, { "epoch": 16.5, "learning_rate": 1.5733141677864594e-06, "loss": 0.0078, "step": 53610 }, { "epoch": 16.5, "learning_rate": 1.5730458934123637e-06, "loss": 0.008, "step": 53611 }, { "epoch": 16.5, "learning_rate": 1.5727776399601225e-06, "loss": 0.005, "step": 53612 }, { "epoch": 16.5, "learning_rate": 1.5725094074304004e-06, "loss": 0.0063, "step": 53613 }, { "epoch": 16.5, "learning_rate": 1.572241195823867e-06, "loss": 0.0064, "step": 53614 }, { "epoch": 16.5, "learning_rate": 1.5719730051411874e-06, "loss": 0.0071, "step": 53615 }, { "epoch": 16.5, "learning_rate": 1.5717048353830266e-06, "loss": 0.0066, "step": 53616 }, { "epoch": 16.5, "learning_rate": 1.571436686550053e-06, "loss": 0.0067, "step": 53617 }, { "epoch": 16.5, "learning_rate": 1.5711685586429249e-06, "loss": 0.0037, "step": 53618 }, { "epoch": 16.5, "learning_rate": 1.5709004516623161e-06, "loss": 0.0095, "step": 53619 }, { "epoch": 16.5, "learning_rate": 1.5706323656088873e-06, "loss": 0.0068, "step": 53620 }, { "epoch": 16.5, "learning_rate": 1.570364300483307e-06, "loss": 0.0061, "step": 53621 }, { "epoch": 16.5, "learning_rate": 1.57009625628624e-06, "loss": 0.0051, "step": 53622 }, { "epoch": 16.5, "learning_rate": 1.5698282330183512e-06, "loss": 0.0053, "step": 53623 }, { "epoch": 16.5, "learning_rate": 1.5695602306803048e-06, "loss": 0.0079, "step": 53624 }, { "epoch": 16.5, "learning_rate": 1.569292249272769e-06, "loss": 0.0074, "step": 53625 }, { "epoch": 16.5, "learning_rate": 1.5690242887964068e-06, "loss": 0.0073, "step": 53626 }, { "epoch": 16.5, "learning_rate": 1.5687563492518854e-06, "loss": 0.005, "step": 53627 }, { "epoch": 16.5, "learning_rate": 1.5684884306398663e-06, "loss": 0.0053, "step": 53628 }, { "epoch": 16.5, "learning_rate": 1.568220532961019e-06, "loss": 0.0043, "step": 53629 }, { "epoch": 16.5, "learning_rate": 1.5679526562160098e-06, "loss": 0.0064, "step": 53630 }, { "epoch": 16.5, "learning_rate": 1.567684800405499e-06, "loss": 0.0069, "step": 53631 }, { "epoch": 16.5, "learning_rate": 1.567416965530153e-06, "loss": 0.0064, "step": 53632 }, { "epoch": 16.5, "learning_rate": 1.5671491515906355e-06, "loss": 0.0052, "step": 53633 }, { "epoch": 16.5, "learning_rate": 1.5668813585876153e-06, "loss": 0.0069, "step": 53634 }, { "epoch": 16.5, "learning_rate": 1.5666135865217546e-06, "loss": 0.0063, "step": 53635 }, { "epoch": 16.5, "learning_rate": 1.56634583539372e-06, "loss": 0.0041, "step": 53636 }, { "epoch": 16.5, "learning_rate": 1.5660781052041741e-06, "loss": 0.0047, "step": 53637 }, { "epoch": 16.5, "learning_rate": 1.5658103959537828e-06, "loss": 0.0058, "step": 53638 }, { "epoch": 16.5, "learning_rate": 1.5655427076432107e-06, "loss": 0.0033, "step": 53639 }, { "epoch": 16.5, "learning_rate": 1.5652750402731221e-06, "loss": 0.0079, "step": 53640 }, { "epoch": 16.5, "learning_rate": 1.5650073938441813e-06, "loss": 0.0041, "step": 53641 }, { "epoch": 16.5, "learning_rate": 1.564739768357053e-06, "loss": 0.0043, "step": 53642 }, { "epoch": 16.51, "learning_rate": 1.5644721638124028e-06, "loss": 0.0026, "step": 53643 }, { "epoch": 16.51, "learning_rate": 1.5642045802108929e-06, "loss": 0.0057, "step": 53644 }, { "epoch": 16.51, "learning_rate": 1.56393701755319e-06, "loss": 0.006, "step": 53645 }, { "epoch": 16.51, "learning_rate": 1.5636694758399563e-06, "loss": 0.0079, "step": 53646 }, { "epoch": 16.51, "learning_rate": 1.5634019550718571e-06, "loss": 0.0071, "step": 53647 }, { "epoch": 16.51, "learning_rate": 1.5631344552495542e-06, "loss": 0.0076, "step": 53648 }, { "epoch": 16.51, "learning_rate": 1.562866976373716e-06, "loss": 0.0039, "step": 53649 }, { "epoch": 16.51, "learning_rate": 1.5625995184450049e-06, "loss": 0.007, "step": 53650 }, { "epoch": 16.51, "learning_rate": 1.5623320814640862e-06, "loss": 0.0063, "step": 53651 }, { "epoch": 16.51, "learning_rate": 1.5620646654316208e-06, "loss": 0.0082, "step": 53652 }, { "epoch": 16.51, "learning_rate": 1.5617972703482708e-06, "loss": 0.0055, "step": 53653 }, { "epoch": 16.51, "learning_rate": 1.5615298962147064e-06, "loss": 0.0072, "step": 53654 }, { "epoch": 16.51, "learning_rate": 1.5612625430315875e-06, "loss": 0.0045, "step": 53655 }, { "epoch": 16.51, "learning_rate": 1.5609952107995796e-06, "loss": 0.0047, "step": 53656 }, { "epoch": 16.51, "learning_rate": 1.5607278995193454e-06, "loss": 0.006, "step": 53657 }, { "epoch": 16.51, "learning_rate": 1.5604606091915485e-06, "loss": 0.0031, "step": 53658 }, { "epoch": 16.51, "learning_rate": 1.5601933398168523e-06, "loss": 0.0067, "step": 53659 }, { "epoch": 16.51, "learning_rate": 1.5599260913959214e-06, "loss": 0.0079, "step": 53660 }, { "epoch": 16.51, "learning_rate": 1.5596588639294186e-06, "loss": 0.0037, "step": 53661 }, { "epoch": 16.51, "learning_rate": 1.5593916574180067e-06, "loss": 0.0033, "step": 53662 }, { "epoch": 16.51, "learning_rate": 1.5591244718623477e-06, "loss": 0.009, "step": 53663 }, { "epoch": 16.51, "learning_rate": 1.558857307263112e-06, "loss": 0.0065, "step": 53664 }, { "epoch": 16.51, "learning_rate": 1.5585901636209554e-06, "loss": 0.0052, "step": 53665 }, { "epoch": 16.51, "learning_rate": 1.5583230409365435e-06, "loss": 0.0068, "step": 53666 }, { "epoch": 16.51, "learning_rate": 1.5580559392105399e-06, "loss": 0.0047, "step": 53667 }, { "epoch": 16.51, "learning_rate": 1.557788858443605e-06, "loss": 0.0071, "step": 53668 }, { "epoch": 16.51, "learning_rate": 1.5575217986364078e-06, "loss": 0.0053, "step": 53669 }, { "epoch": 16.51, "learning_rate": 1.557254759789607e-06, "loss": 0.0017, "step": 53670 }, { "epoch": 16.51, "learning_rate": 1.5569877419038659e-06, "loss": 0.0081, "step": 53671 }, { "epoch": 16.51, "learning_rate": 1.5567207449798517e-06, "loss": 0.0087, "step": 53672 }, { "epoch": 16.51, "learning_rate": 1.556453769018218e-06, "loss": 0.0043, "step": 53673 }, { "epoch": 16.51, "learning_rate": 1.5561868140196368e-06, "loss": 0.0061, "step": 53674 }, { "epoch": 16.52, "learning_rate": 1.5559198799847664e-06, "loss": 0.0055, "step": 53675 }, { "epoch": 16.52, "learning_rate": 1.5556529669142707e-06, "loss": 0.0078, "step": 53676 }, { "epoch": 16.52, "learning_rate": 1.5553860748088123e-06, "loss": 0.0026, "step": 53677 }, { "epoch": 16.52, "learning_rate": 1.5551192036690532e-06, "loss": 0.0061, "step": 53678 }, { "epoch": 16.52, "learning_rate": 1.5548523534956572e-06, "loss": 0.01, "step": 53679 }, { "epoch": 16.52, "learning_rate": 1.554585524289286e-06, "loss": 0.0092, "step": 53680 }, { "epoch": 16.52, "learning_rate": 1.5543187160506013e-06, "loss": 0.003, "step": 53681 }, { "epoch": 16.52, "learning_rate": 1.5540519287802668e-06, "loss": 0.006, "step": 53682 }, { "epoch": 16.52, "learning_rate": 1.5537851624789423e-06, "loss": 0.0032, "step": 53683 }, { "epoch": 16.52, "learning_rate": 1.553518417147294e-06, "loss": 0.0121, "step": 53684 }, { "epoch": 16.52, "learning_rate": 1.5532516927859853e-06, "loss": 0.0056, "step": 53685 }, { "epoch": 16.52, "learning_rate": 1.5529849893956727e-06, "loss": 0.0062, "step": 53686 }, { "epoch": 16.52, "learning_rate": 1.5527183069770213e-06, "loss": 0.0051, "step": 53687 }, { "epoch": 16.52, "learning_rate": 1.5524516455306905e-06, "loss": 0.0052, "step": 53688 }, { "epoch": 16.52, "learning_rate": 1.5521850050573484e-06, "loss": 0.0042, "step": 53689 }, { "epoch": 16.52, "learning_rate": 1.5519183855576513e-06, "loss": 0.006, "step": 53690 }, { "epoch": 16.52, "learning_rate": 1.5516517870322656e-06, "loss": 0.0079, "step": 53691 }, { "epoch": 16.52, "learning_rate": 1.5513852094818494e-06, "loss": 0.0084, "step": 53692 }, { "epoch": 16.52, "learning_rate": 1.5511186529070666e-06, "loss": 0.0045, "step": 53693 }, { "epoch": 16.52, "learning_rate": 1.5508521173085778e-06, "loss": 0.0056, "step": 53694 }, { "epoch": 16.52, "learning_rate": 1.5505856026870459e-06, "loss": 0.0092, "step": 53695 }, { "epoch": 16.52, "learning_rate": 1.5503191090431324e-06, "loss": 0.0046, "step": 53696 }, { "epoch": 16.52, "learning_rate": 1.5500526363774981e-06, "loss": 0.0073, "step": 53697 }, { "epoch": 16.52, "learning_rate": 1.5497861846908024e-06, "loss": 0.0036, "step": 53698 }, { "epoch": 16.52, "learning_rate": 1.5495197539837148e-06, "loss": 0.0074, "step": 53699 }, { "epoch": 16.52, "learning_rate": 1.549253344256889e-06, "loss": 0.0111, "step": 53700 }, { "epoch": 16.52, "learning_rate": 1.5489869555109893e-06, "loss": 0.003, "step": 53701 }, { "epoch": 16.52, "learning_rate": 1.548720587746676e-06, "loss": 0.0041, "step": 53702 }, { "epoch": 16.52, "learning_rate": 1.5484542409646086e-06, "loss": 0.0076, "step": 53703 }, { "epoch": 16.52, "learning_rate": 1.5481879151654545e-06, "loss": 0.0072, "step": 53704 }, { "epoch": 16.52, "learning_rate": 1.5479216103498696e-06, "loss": 0.0055, "step": 53705 }, { "epoch": 16.52, "learning_rate": 1.54765532651852e-06, "loss": 0.0059, "step": 53706 }, { "epoch": 16.52, "learning_rate": 1.5473890636720612e-06, "loss": 0.0083, "step": 53707 }, { "epoch": 16.53, "learning_rate": 1.5471228218111544e-06, "loss": 0.0072, "step": 53708 }, { "epoch": 16.53, "learning_rate": 1.5468566009364649e-06, "loss": 0.0035, "step": 53709 }, { "epoch": 16.53, "learning_rate": 1.546590401048651e-06, "loss": 0.0052, "step": 53710 }, { "epoch": 16.53, "learning_rate": 1.5463242221483742e-06, "loss": 0.0038, "step": 53711 }, { "epoch": 16.53, "learning_rate": 1.5460580642362966e-06, "loss": 0.0128, "step": 53712 }, { "epoch": 16.53, "learning_rate": 1.5457919273130762e-06, "loss": 0.0036, "step": 53713 }, { "epoch": 16.53, "learning_rate": 1.5455258113793759e-06, "loss": 0.0039, "step": 53714 }, { "epoch": 16.53, "learning_rate": 1.5452597164358563e-06, "loss": 0.0109, "step": 53715 }, { "epoch": 16.53, "learning_rate": 1.5449936424831769e-06, "loss": 0.0094, "step": 53716 }, { "epoch": 16.53, "learning_rate": 1.5447275895219993e-06, "loss": 0.0067, "step": 53717 }, { "epoch": 16.53, "learning_rate": 1.5444615575529808e-06, "loss": 0.007, "step": 53718 }, { "epoch": 16.53, "learning_rate": 1.5441955465767866e-06, "loss": 0.0028, "step": 53719 }, { "epoch": 16.53, "learning_rate": 1.5439295565940782e-06, "loss": 0.0057, "step": 53720 }, { "epoch": 16.53, "learning_rate": 1.5436635876055107e-06, "loss": 0.0028, "step": 53721 }, { "epoch": 16.53, "learning_rate": 1.5433976396117468e-06, "loss": 0.0043, "step": 53722 }, { "epoch": 16.53, "learning_rate": 1.543131712613445e-06, "loss": 0.0048, "step": 53723 }, { "epoch": 16.53, "learning_rate": 1.5428658066112689e-06, "loss": 0.0023, "step": 53724 }, { "epoch": 16.53, "learning_rate": 1.5425999216058762e-06, "loss": 0.0034, "step": 53725 }, { "epoch": 16.53, "learning_rate": 1.5423340575979295e-06, "loss": 0.0045, "step": 53726 }, { "epoch": 16.53, "learning_rate": 1.5420682145880883e-06, "loss": 0.0052, "step": 53727 }, { "epoch": 16.53, "learning_rate": 1.5418023925770075e-06, "loss": 0.0054, "step": 53728 }, { "epoch": 16.53, "learning_rate": 1.5415365915653535e-06, "loss": 0.0062, "step": 53729 }, { "epoch": 16.53, "learning_rate": 1.5412708115537834e-06, "loss": 0.0051, "step": 53730 }, { "epoch": 16.53, "learning_rate": 1.5410050525429577e-06, "loss": 0.003, "step": 53731 }, { "epoch": 16.53, "learning_rate": 1.540739314533536e-06, "loss": 0.0056, "step": 53732 }, { "epoch": 16.53, "learning_rate": 1.5404735975261775e-06, "loss": 0.0044, "step": 53733 }, { "epoch": 16.53, "learning_rate": 1.5402079015215432e-06, "loss": 0.004, "step": 53734 }, { "epoch": 16.53, "learning_rate": 1.5399422265202912e-06, "loss": 0.0059, "step": 53735 }, { "epoch": 16.53, "learning_rate": 1.5396765725230822e-06, "loss": 0.0051, "step": 53736 }, { "epoch": 16.53, "learning_rate": 1.5394109395305757e-06, "loss": 0.0057, "step": 53737 }, { "epoch": 16.53, "learning_rate": 1.5391453275434288e-06, "loss": 0.0098, "step": 53738 }, { "epoch": 16.53, "learning_rate": 1.5388797365623053e-06, "loss": 0.0037, "step": 53739 }, { "epoch": 16.54, "learning_rate": 1.5386141665878617e-06, "loss": 0.0045, "step": 53740 }, { "epoch": 16.54, "learning_rate": 1.5383486176207606e-06, "loss": 0.0235, "step": 53741 }, { "epoch": 16.54, "learning_rate": 1.5380830896616571e-06, "loss": 0.0069, "step": 53742 }, { "epoch": 16.54, "learning_rate": 1.5378175827112097e-06, "loss": 0.011, "step": 53743 }, { "epoch": 16.54, "learning_rate": 1.5375520967700808e-06, "loss": 0.0031, "step": 53744 }, { "epoch": 16.54, "learning_rate": 1.5372866318389302e-06, "loss": 0.0068, "step": 53745 }, { "epoch": 16.54, "learning_rate": 1.537021187918415e-06, "loss": 0.0038, "step": 53746 }, { "epoch": 16.54, "learning_rate": 1.5367557650091935e-06, "loss": 0.0049, "step": 53747 }, { "epoch": 16.54, "learning_rate": 1.5364903631119277e-06, "loss": 0.0029, "step": 53748 }, { "epoch": 16.54, "learning_rate": 1.5362249822272735e-06, "loss": 0.0052, "step": 53749 }, { "epoch": 16.54, "learning_rate": 1.5359596223558903e-06, "loss": 0.0063, "step": 53750 }, { "epoch": 16.54, "learning_rate": 1.535694283498439e-06, "loss": 0.0087, "step": 53751 }, { "epoch": 16.54, "learning_rate": 1.5354289656555754e-06, "loss": 0.0091, "step": 53752 }, { "epoch": 16.54, "learning_rate": 1.535163668827958e-06, "loss": 0.008, "step": 53753 }, { "epoch": 16.54, "learning_rate": 1.5348983930162532e-06, "loss": 0.0079, "step": 53754 }, { "epoch": 16.54, "learning_rate": 1.53463313822111e-06, "loss": 0.0062, "step": 53755 }, { "epoch": 16.54, "learning_rate": 1.5343679044431902e-06, "loss": 0.005, "step": 53756 }, { "epoch": 16.54, "learning_rate": 1.5341026916831526e-06, "loss": 0.0053, "step": 53757 }, { "epoch": 16.54, "learning_rate": 1.533837499941654e-06, "loss": 0.0084, "step": 53758 }, { "epoch": 16.54, "learning_rate": 1.533572329219356e-06, "loss": 0.004, "step": 53759 }, { "epoch": 16.54, "learning_rate": 1.5333071795169164e-06, "loss": 0.0031, "step": 53760 }, { "epoch": 16.54, "learning_rate": 1.533042050834992e-06, "loss": 0.0082, "step": 53761 }, { "epoch": 16.54, "learning_rate": 1.5327769431742434e-06, "loss": 0.0048, "step": 53762 }, { "epoch": 16.54, "learning_rate": 1.5325118565353237e-06, "loss": 0.0024, "step": 53763 }, { "epoch": 16.54, "learning_rate": 1.5322467909188955e-06, "loss": 0.0049, "step": 53764 }, { "epoch": 16.54, "learning_rate": 1.531981746325616e-06, "loss": 0.0039, "step": 53765 }, { "epoch": 16.54, "learning_rate": 1.5317167227561425e-06, "loss": 0.0058, "step": 53766 }, { "epoch": 16.54, "learning_rate": 1.5314517202111345e-06, "loss": 0.0081, "step": 53767 }, { "epoch": 16.54, "learning_rate": 1.5311867386912481e-06, "loss": 0.0043, "step": 53768 }, { "epoch": 16.54, "learning_rate": 1.5309217781971419e-06, "loss": 0.0078, "step": 53769 }, { "epoch": 16.54, "learning_rate": 1.530656838729474e-06, "loss": 0.0039, "step": 53770 }, { "epoch": 16.54, "learning_rate": 1.5303919202889006e-06, "loss": 0.0032, "step": 53771 }, { "epoch": 16.54, "learning_rate": 1.5301270228760823e-06, "loss": 0.0076, "step": 53772 }, { "epoch": 16.55, "learning_rate": 1.5298621464916718e-06, "loss": 0.0062, "step": 53773 }, { "epoch": 16.55, "learning_rate": 1.5295972911363333e-06, "loss": 0.0053, "step": 53774 }, { "epoch": 16.55, "learning_rate": 1.5293324568107237e-06, "loss": 0.0057, "step": 53775 }, { "epoch": 16.55, "learning_rate": 1.5290676435154949e-06, "loss": 0.006, "step": 53776 }, { "epoch": 16.55, "learning_rate": 1.5288028512513076e-06, "loss": 0.0055, "step": 53777 }, { "epoch": 16.55, "learning_rate": 1.5285380800188165e-06, "loss": 0.0043, "step": 53778 }, { "epoch": 16.55, "learning_rate": 1.5282733298186846e-06, "loss": 0.004, "step": 53779 }, { "epoch": 16.55, "learning_rate": 1.528008600651566e-06, "loss": 0.0024, "step": 53780 }, { "epoch": 16.55, "learning_rate": 1.5277438925181175e-06, "loss": 0.0032, "step": 53781 }, { "epoch": 16.55, "learning_rate": 1.527479205418999e-06, "loss": 0.0082, "step": 53782 }, { "epoch": 16.55, "learning_rate": 1.527214539354862e-06, "loss": 0.0063, "step": 53783 }, { "epoch": 16.55, "learning_rate": 1.5269498943263682e-06, "loss": 0.0058, "step": 53784 }, { "epoch": 16.55, "learning_rate": 1.5266852703341738e-06, "loss": 0.0039, "step": 53785 }, { "epoch": 16.55, "learning_rate": 1.526420667378936e-06, "loss": 0.0059, "step": 53786 }, { "epoch": 16.55, "learning_rate": 1.526156085461311e-06, "loss": 0.0058, "step": 53787 }, { "epoch": 16.55, "learning_rate": 1.5258915245819561e-06, "loss": 0.0025, "step": 53788 }, { "epoch": 16.55, "learning_rate": 1.5256269847415283e-06, "loss": 0.0109, "step": 53789 }, { "epoch": 16.55, "learning_rate": 1.525362465940684e-06, "loss": 0.0054, "step": 53790 }, { "epoch": 16.55, "learning_rate": 1.5250979681800793e-06, "loss": 0.0037, "step": 53791 }, { "epoch": 16.55, "learning_rate": 1.5248334914603703e-06, "loss": 0.0049, "step": 53792 }, { "epoch": 16.55, "learning_rate": 1.5245690357822163e-06, "loss": 0.0029, "step": 53793 }, { "epoch": 16.55, "learning_rate": 1.5243046011462725e-06, "loss": 0.0045, "step": 53794 }, { "epoch": 16.55, "learning_rate": 1.5240401875531963e-06, "loss": 0.0061, "step": 53795 }, { "epoch": 16.55, "learning_rate": 1.5237757950036447e-06, "loss": 0.0052, "step": 53796 }, { "epoch": 16.55, "learning_rate": 1.5235114234982683e-06, "loss": 0.0072, "step": 53797 }, { "epoch": 16.55, "learning_rate": 1.5232470730377303e-06, "loss": 0.0058, "step": 53798 }, { "epoch": 16.55, "learning_rate": 1.522982743622684e-06, "loss": 0.0055, "step": 53799 }, { "epoch": 16.55, "learning_rate": 1.522718435253787e-06, "loss": 0.0046, "step": 53800 }, { "epoch": 16.55, "learning_rate": 1.5224541479316946e-06, "loss": 0.0056, "step": 53801 }, { "epoch": 16.55, "learning_rate": 1.5221898816570623e-06, "loss": 0.003, "step": 53802 }, { "epoch": 16.55, "learning_rate": 1.521925636430548e-06, "loss": 0.0064, "step": 53803 }, { "epoch": 16.55, "learning_rate": 1.5216614122528063e-06, "loss": 0.0083, "step": 53804 }, { "epoch": 16.56, "learning_rate": 1.5213972091244932e-06, "loss": 0.0067, "step": 53805 }, { "epoch": 16.56, "learning_rate": 1.5211330270462666e-06, "loss": 0.006, "step": 53806 }, { "epoch": 16.56, "learning_rate": 1.5208688660187764e-06, "loss": 0.0064, "step": 53807 }, { "epoch": 16.56, "learning_rate": 1.520604726042687e-06, "loss": 0.0056, "step": 53808 }, { "epoch": 16.56, "learning_rate": 1.520340607118651e-06, "loss": 0.0068, "step": 53809 }, { "epoch": 16.56, "learning_rate": 1.5200765092473212e-06, "loss": 0.0057, "step": 53810 }, { "epoch": 16.56, "learning_rate": 1.5198124324293562e-06, "loss": 0.0046, "step": 53811 }, { "epoch": 16.56, "learning_rate": 1.5195483766654085e-06, "loss": 0.0051, "step": 53812 }, { "epoch": 16.56, "learning_rate": 1.5192843419561366e-06, "loss": 0.005, "step": 53813 }, { "epoch": 16.56, "learning_rate": 1.5190203283021964e-06, "loss": 0.005, "step": 53814 }, { "epoch": 16.56, "learning_rate": 1.5187563357042423e-06, "loss": 0.0089, "step": 53815 }, { "epoch": 16.56, "learning_rate": 1.5184923641629302e-06, "loss": 0.0056, "step": 53816 }, { "epoch": 16.56, "learning_rate": 1.5182284136789149e-06, "loss": 0.0052, "step": 53817 }, { "epoch": 16.56, "learning_rate": 1.5179644842528508e-06, "loss": 0.0043, "step": 53818 }, { "epoch": 16.56, "learning_rate": 1.5177005758853946e-06, "loss": 0.0104, "step": 53819 }, { "epoch": 16.56, "learning_rate": 1.5174366885772019e-06, "loss": 0.0047, "step": 53820 }, { "epoch": 16.56, "learning_rate": 1.5171728223289261e-06, "loss": 0.0043, "step": 53821 }, { "epoch": 16.56, "learning_rate": 1.5169089771412248e-06, "loss": 0.0074, "step": 53822 }, { "epoch": 16.56, "learning_rate": 1.5166451530147496e-06, "loss": 0.0049, "step": 53823 }, { "epoch": 16.56, "learning_rate": 1.516381349950159e-06, "loss": 0.0044, "step": 53824 }, { "epoch": 16.56, "learning_rate": 1.5161175679481054e-06, "loss": 0.0159, "step": 53825 }, { "epoch": 16.56, "learning_rate": 1.5158538070092454e-06, "loss": 0.0077, "step": 53826 }, { "epoch": 16.56, "learning_rate": 1.5155900671342306e-06, "loss": 0.0102, "step": 53827 }, { "epoch": 16.56, "learning_rate": 1.51532634832372e-06, "loss": 0.0098, "step": 53828 }, { "epoch": 16.56, "learning_rate": 1.5150626505783672e-06, "loss": 0.006, "step": 53829 }, { "epoch": 16.56, "learning_rate": 1.5147989738988289e-06, "loss": 0.0064, "step": 53830 }, { "epoch": 16.56, "learning_rate": 1.5145353182857547e-06, "loss": 0.0031, "step": 53831 }, { "epoch": 16.56, "learning_rate": 1.5142716837397998e-06, "loss": 0.0048, "step": 53832 }, { "epoch": 16.56, "learning_rate": 1.5140080702616223e-06, "loss": 0.0041, "step": 53833 }, { "epoch": 16.56, "learning_rate": 1.513744477851875e-06, "loss": 0.0034, "step": 53834 }, { "epoch": 16.56, "learning_rate": 1.513480906511212e-06, "loss": 0.0089, "step": 53835 }, { "epoch": 16.56, "learning_rate": 1.5132173562402885e-06, "loss": 0.0046, "step": 53836 }, { "epoch": 16.56, "learning_rate": 1.5129538270397592e-06, "loss": 0.0053, "step": 53837 }, { "epoch": 16.57, "learning_rate": 1.512690318910276e-06, "loss": 0.003, "step": 53838 }, { "epoch": 16.57, "learning_rate": 1.512426831852496e-06, "loss": 0.0032, "step": 53839 }, { "epoch": 16.57, "learning_rate": 1.5121633658670709e-06, "loss": 0.0112, "step": 53840 }, { "epoch": 16.57, "learning_rate": 1.511899920954656e-06, "loss": 0.0066, "step": 53841 }, { "epoch": 16.57, "learning_rate": 1.5116364971159038e-06, "loss": 0.0067, "step": 53842 }, { "epoch": 16.57, "learning_rate": 1.5113730943514736e-06, "loss": 0.0124, "step": 53843 }, { "epoch": 16.57, "learning_rate": 1.5111097126620133e-06, "loss": 0.0048, "step": 53844 }, { "epoch": 16.57, "learning_rate": 1.5108463520481797e-06, "loss": 0.0068, "step": 53845 }, { "epoch": 16.57, "learning_rate": 1.5105830125106258e-06, "loss": 0.0089, "step": 53846 }, { "epoch": 16.57, "learning_rate": 1.5103196940500031e-06, "loss": 0.0033, "step": 53847 }, { "epoch": 16.57, "learning_rate": 1.5100563966669702e-06, "loss": 0.0054, "step": 53848 }, { "epoch": 16.57, "learning_rate": 1.5097931203621785e-06, "loss": 0.0032, "step": 53849 }, { "epoch": 16.57, "learning_rate": 1.5095298651362812e-06, "loss": 0.0034, "step": 53850 }, { "epoch": 16.57, "learning_rate": 1.509266630989935e-06, "loss": 0.0047, "step": 53851 }, { "epoch": 16.57, "learning_rate": 1.5090034179237857e-06, "loss": 0.0055, "step": 53852 }, { "epoch": 16.57, "learning_rate": 1.5087402259384943e-06, "loss": 0.0055, "step": 53853 }, { "epoch": 16.57, "learning_rate": 1.5084770550347115e-06, "loss": 0.0074, "step": 53854 }, { "epoch": 16.57, "learning_rate": 1.5082139052130918e-06, "loss": 0.0046, "step": 53855 }, { "epoch": 16.57, "learning_rate": 1.5079507764742874e-06, "loss": 0.0056, "step": 53856 }, { "epoch": 16.57, "learning_rate": 1.5076876688189512e-06, "loss": 0.0037, "step": 53857 }, { "epoch": 16.57, "learning_rate": 1.5074245822477374e-06, "loss": 0.0071, "step": 53858 }, { "epoch": 16.57, "learning_rate": 1.5071615167612996e-06, "loss": 0.0047, "step": 53859 }, { "epoch": 16.57, "learning_rate": 1.5068984723602887e-06, "loss": 0.0056, "step": 53860 }, { "epoch": 16.57, "learning_rate": 1.5066354490453606e-06, "loss": 0.0089, "step": 53861 }, { "epoch": 16.57, "learning_rate": 1.5063724468171637e-06, "loss": 0.0062, "step": 53862 }, { "epoch": 16.57, "learning_rate": 1.5061094656763564e-06, "loss": 0.0045, "step": 53863 }, { "epoch": 16.57, "learning_rate": 1.5058465056235927e-06, "loss": 0.005, "step": 53864 }, { "epoch": 16.57, "learning_rate": 1.5055835666595186e-06, "loss": 0.0049, "step": 53865 }, { "epoch": 16.57, "learning_rate": 1.5053206487847916e-06, "loss": 0.0074, "step": 53866 }, { "epoch": 16.57, "learning_rate": 1.5050577520000608e-06, "loss": 0.0068, "step": 53867 }, { "epoch": 16.57, "learning_rate": 1.504794876305984e-06, "loss": 0.0036, "step": 53868 }, { "epoch": 16.57, "learning_rate": 1.50453202170321e-06, "loss": 0.0061, "step": 53869 }, { "epoch": 16.58, "learning_rate": 1.5042691881923931e-06, "loss": 0.0123, "step": 53870 }, { "epoch": 16.58, "learning_rate": 1.5040063757741863e-06, "loss": 0.0038, "step": 53871 }, { "epoch": 16.58, "learning_rate": 1.5037435844492409e-06, "loss": 0.0055, "step": 53872 }, { "epoch": 16.58, "learning_rate": 1.50348081421821e-06, "loss": 0.0074, "step": 53873 }, { "epoch": 16.58, "learning_rate": 1.503218065081745e-06, "loss": 0.0056, "step": 53874 }, { "epoch": 16.58, "learning_rate": 1.5029553370404993e-06, "loss": 0.0044, "step": 53875 }, { "epoch": 16.58, "learning_rate": 1.5026926300951239e-06, "loss": 0.024, "step": 53876 }, { "epoch": 16.58, "learning_rate": 1.5024299442462697e-06, "loss": 0.0047, "step": 53877 }, { "epoch": 16.58, "learning_rate": 1.5021672794945963e-06, "loss": 0.0049, "step": 53878 }, { "epoch": 16.58, "learning_rate": 1.5019046358407485e-06, "loss": 0.0038, "step": 53879 }, { "epoch": 16.58, "learning_rate": 1.5016420132853804e-06, "loss": 0.0033, "step": 53880 }, { "epoch": 16.58, "learning_rate": 1.5013794118291435e-06, "loss": 0.003, "step": 53881 }, { "epoch": 16.58, "learning_rate": 1.5011168314726887e-06, "loss": 0.0035, "step": 53882 }, { "epoch": 16.58, "learning_rate": 1.500854272216672e-06, "loss": 0.0045, "step": 53883 }, { "epoch": 16.58, "learning_rate": 1.5005917340617415e-06, "loss": 0.009, "step": 53884 }, { "epoch": 16.58, "learning_rate": 1.5003292170085538e-06, "loss": 0.0084, "step": 53885 }, { "epoch": 16.58, "learning_rate": 1.5000667210577547e-06, "loss": 0.0044, "step": 53886 }, { "epoch": 16.58, "learning_rate": 1.499804246209996e-06, "loss": 0.0079, "step": 53887 }, { "epoch": 16.58, "learning_rate": 1.499541792465935e-06, "loss": 0.0055, "step": 53888 }, { "epoch": 16.58, "learning_rate": 1.499279359826219e-06, "loss": 0.0104, "step": 53889 }, { "epoch": 16.58, "learning_rate": 1.4990169482915007e-06, "loss": 0.0044, "step": 53890 }, { "epoch": 16.58, "learning_rate": 1.4987545578624318e-06, "loss": 0.0044, "step": 53891 }, { "epoch": 16.58, "learning_rate": 1.498492188539663e-06, "loss": 0.0056, "step": 53892 }, { "epoch": 16.58, "learning_rate": 1.498229840323847e-06, "loss": 0.0048, "step": 53893 }, { "epoch": 16.58, "learning_rate": 1.4979675132156335e-06, "loss": 0.0062, "step": 53894 }, { "epoch": 16.58, "learning_rate": 1.497705207215675e-06, "loss": 0.0045, "step": 53895 }, { "epoch": 16.58, "learning_rate": 1.4974429223246223e-06, "loss": 0.0077, "step": 53896 }, { "epoch": 16.58, "learning_rate": 1.4971806585431247e-06, "loss": 0.0075, "step": 53897 }, { "epoch": 16.58, "learning_rate": 1.4969184158718364e-06, "loss": 0.007, "step": 53898 }, { "epoch": 16.58, "learning_rate": 1.4966561943114112e-06, "loss": 0.0076, "step": 53899 }, { "epoch": 16.58, "learning_rate": 1.4963939938624927e-06, "loss": 0.0037, "step": 53900 }, { "epoch": 16.58, "learning_rate": 1.4961318145257353e-06, "loss": 0.0073, "step": 53901 }, { "epoch": 16.58, "learning_rate": 1.4958696563017883e-06, "loss": 0.0057, "step": 53902 }, { "epoch": 16.59, "learning_rate": 1.4956075191913066e-06, "loss": 0.0053, "step": 53903 }, { "epoch": 16.59, "learning_rate": 1.4953454031949389e-06, "loss": 0.0069, "step": 53904 }, { "epoch": 16.59, "learning_rate": 1.4950833083133354e-06, "loss": 0.0046, "step": 53905 }, { "epoch": 16.59, "learning_rate": 1.4948212345471492e-06, "loss": 0.0084, "step": 53906 }, { "epoch": 16.59, "learning_rate": 1.4945591818970252e-06, "loss": 0.0101, "step": 53907 }, { "epoch": 16.59, "learning_rate": 1.4942971503636195e-06, "loss": 0.0047, "step": 53908 }, { "epoch": 16.59, "learning_rate": 1.4940351399475805e-06, "loss": 0.0039, "step": 53909 }, { "epoch": 16.59, "learning_rate": 1.4937731506495601e-06, "loss": 0.006, "step": 53910 }, { "epoch": 16.59, "learning_rate": 1.4935111824702076e-06, "loss": 0.0053, "step": 53911 }, { "epoch": 16.59, "learning_rate": 1.4932492354101725e-06, "loss": 0.0091, "step": 53912 }, { "epoch": 16.59, "learning_rate": 1.4929873094701075e-06, "loss": 0.0042, "step": 53913 }, { "epoch": 16.59, "learning_rate": 1.4927254046506612e-06, "loss": 0.0049, "step": 53914 }, { "epoch": 16.59, "learning_rate": 1.492463520952484e-06, "loss": 0.0061, "step": 53915 }, { "epoch": 16.59, "learning_rate": 1.4922016583762255e-06, "loss": 0.0039, "step": 53916 }, { "epoch": 16.59, "learning_rate": 1.4919398169225352e-06, "loss": 0.0057, "step": 53917 }, { "epoch": 16.59, "learning_rate": 1.491677996592067e-06, "loss": 0.0055, "step": 53918 }, { "epoch": 16.59, "learning_rate": 1.4914161973854714e-06, "loss": 0.0024, "step": 53919 }, { "epoch": 16.59, "learning_rate": 1.4911544193033923e-06, "loss": 0.0063, "step": 53920 }, { "epoch": 16.59, "learning_rate": 1.4908926623464838e-06, "loss": 0.0074, "step": 53921 }, { "epoch": 16.59, "learning_rate": 1.4906309265153917e-06, "loss": 0.0036, "step": 53922 }, { "epoch": 16.59, "learning_rate": 1.4903692118107716e-06, "loss": 0.0037, "step": 53923 }, { "epoch": 16.59, "learning_rate": 1.4901075182332702e-06, "loss": 0.0071, "step": 53924 }, { "epoch": 16.59, "learning_rate": 1.4898458457835386e-06, "loss": 0.0061, "step": 53925 }, { "epoch": 16.59, "learning_rate": 1.4895841944622248e-06, "loss": 0.005, "step": 53926 }, { "epoch": 16.59, "learning_rate": 1.4893225642699782e-06, "loss": 0.0064, "step": 53927 }, { "epoch": 16.59, "learning_rate": 1.4890609552074509e-06, "loss": 0.0029, "step": 53928 }, { "epoch": 16.59, "learning_rate": 1.48879936727529e-06, "loss": 0.006, "step": 53929 }, { "epoch": 16.59, "learning_rate": 1.4885378004741457e-06, "loss": 0.0047, "step": 53930 }, { "epoch": 16.59, "learning_rate": 1.488276254804668e-06, "loss": 0.0058, "step": 53931 }, { "epoch": 16.59, "learning_rate": 1.488014730267502e-06, "loss": 0.0035, "step": 53932 }, { "epoch": 16.59, "learning_rate": 1.4877532268633065e-06, "loss": 0.005, "step": 53933 }, { "epoch": 16.59, "learning_rate": 1.4874917445927228e-06, "loss": 0.004, "step": 53934 }, { "epoch": 16.6, "learning_rate": 1.4872302834564012e-06, "loss": 0.003, "step": 53935 }, { "epoch": 16.6, "learning_rate": 1.4869688434549923e-06, "loss": 0.0048, "step": 53936 }, { "epoch": 16.6, "learning_rate": 1.4867074245891423e-06, "loss": 0.0061, "step": 53937 }, { "epoch": 16.6, "learning_rate": 1.4864460268595048e-06, "loss": 0.0045, "step": 53938 }, { "epoch": 16.6, "learning_rate": 1.4861846502667276e-06, "loss": 0.0056, "step": 53939 }, { "epoch": 16.6, "learning_rate": 1.48592329481146e-06, "loss": 0.0038, "step": 53940 }, { "epoch": 16.6, "learning_rate": 1.4856619604943478e-06, "loss": 0.0031, "step": 53941 }, { "epoch": 16.6, "learning_rate": 1.4854006473160387e-06, "loss": 0.0039, "step": 53942 }, { "epoch": 16.6, "learning_rate": 1.4851393552771864e-06, "loss": 0.0058, "step": 53943 }, { "epoch": 16.6, "learning_rate": 1.4848780843784383e-06, "loss": 0.0077, "step": 53944 }, { "epoch": 16.6, "learning_rate": 1.4846168346204425e-06, "loss": 0.0056, "step": 53945 }, { "epoch": 16.6, "learning_rate": 1.4843556060038466e-06, "loss": 0.0098, "step": 53946 }, { "epoch": 16.6, "learning_rate": 1.484094398529301e-06, "loss": 0.0042, "step": 53947 }, { "epoch": 16.6, "learning_rate": 1.483833212197452e-06, "loss": 0.0037, "step": 53948 }, { "epoch": 16.6, "learning_rate": 1.4835720470089498e-06, "loss": 0.0087, "step": 53949 }, { "epoch": 16.6, "learning_rate": 1.483310902964442e-06, "loss": 0.0085, "step": 53950 }, { "epoch": 16.6, "learning_rate": 1.4830497800645782e-06, "loss": 0.0047, "step": 53951 }, { "epoch": 16.6, "learning_rate": 1.482788678310002e-06, "loss": 0.0072, "step": 53952 }, { "epoch": 16.6, "learning_rate": 1.4825275977013687e-06, "loss": 0.0042, "step": 53953 }, { "epoch": 16.6, "learning_rate": 1.4822665382393253e-06, "loss": 0.0046, "step": 53954 }, { "epoch": 16.6, "learning_rate": 1.4820054999245148e-06, "loss": 0.0047, "step": 53955 }, { "epoch": 16.6, "learning_rate": 1.481744482757589e-06, "loss": 0.0079, "step": 53956 }, { "epoch": 16.6, "learning_rate": 1.4814834867391925e-06, "loss": 0.0033, "step": 53957 }, { "epoch": 16.6, "learning_rate": 1.4812225118699775e-06, "loss": 0.009, "step": 53958 }, { "epoch": 16.6, "learning_rate": 1.480961558150592e-06, "loss": 0.006, "step": 53959 }, { "epoch": 16.6, "learning_rate": 1.4807006255816814e-06, "loss": 0.0103, "step": 53960 }, { "epoch": 16.6, "learning_rate": 1.4804397141638972e-06, "loss": 0.0057, "step": 53961 }, { "epoch": 16.6, "learning_rate": 1.4801788238978786e-06, "loss": 0.005, "step": 53962 }, { "epoch": 16.6, "learning_rate": 1.4799179547842823e-06, "loss": 0.007, "step": 53963 }, { "epoch": 16.6, "learning_rate": 1.4796571068237531e-06, "loss": 0.0104, "step": 53964 }, { "epoch": 16.6, "learning_rate": 1.4793962800169381e-06, "loss": 0.005, "step": 53965 }, { "epoch": 16.6, "learning_rate": 1.4791354743644847e-06, "loss": 0.0063, "step": 53966 }, { "epoch": 16.6, "learning_rate": 1.4788746898670415e-06, "loss": 0.0054, "step": 53967 }, { "epoch": 16.61, "learning_rate": 1.478613926525254e-06, "loss": 0.0034, "step": 53968 }, { "epoch": 16.61, "learning_rate": 1.4783531843397713e-06, "loss": 0.0064, "step": 53969 }, { "epoch": 16.61, "learning_rate": 1.4780924633112403e-06, "loss": 0.0051, "step": 53970 }, { "epoch": 16.61, "learning_rate": 1.4778317634403082e-06, "loss": 0.0042, "step": 53971 }, { "epoch": 16.61, "learning_rate": 1.4775710847276203e-06, "loss": 0.0071, "step": 53972 }, { "epoch": 16.61, "learning_rate": 1.4773104271738282e-06, "loss": 0.0063, "step": 53973 }, { "epoch": 16.61, "learning_rate": 1.4770497907795767e-06, "loss": 0.0034, "step": 53974 }, { "epoch": 16.61, "learning_rate": 1.4767891755455155e-06, "loss": 0.01, "step": 53975 }, { "epoch": 16.61, "learning_rate": 1.4765285814722863e-06, "loss": 0.0034, "step": 53976 }, { "epoch": 16.61, "learning_rate": 1.4762680085605363e-06, "loss": 0.0035, "step": 53977 }, { "epoch": 16.61, "learning_rate": 1.4760074568109172e-06, "loss": 0.0049, "step": 53978 }, { "epoch": 16.61, "learning_rate": 1.4757469262240731e-06, "loss": 0.006, "step": 53979 }, { "epoch": 16.61, "learning_rate": 1.4754864168006521e-06, "loss": 0.0065, "step": 53980 }, { "epoch": 16.61, "learning_rate": 1.4752259285412995e-06, "loss": 0.0046, "step": 53981 }, { "epoch": 16.61, "learning_rate": 1.4749654614466636e-06, "loss": 0.0099, "step": 53982 }, { "epoch": 16.61, "learning_rate": 1.4747050155173903e-06, "loss": 0.0085, "step": 53983 }, { "epoch": 16.61, "learning_rate": 1.4744445907541249e-06, "loss": 0.0043, "step": 53984 }, { "epoch": 16.61, "learning_rate": 1.4741841871575159e-06, "loss": 0.0087, "step": 53985 }, { "epoch": 16.61, "learning_rate": 1.473923804728209e-06, "loss": 0.0078, "step": 53986 }, { "epoch": 16.61, "learning_rate": 1.4736634434668484e-06, "loss": 0.0038, "step": 53987 }, { "epoch": 16.61, "learning_rate": 1.4734031033740882e-06, "loss": 0.0073, "step": 53988 }, { "epoch": 16.61, "learning_rate": 1.4731427844505653e-06, "loss": 0.0037, "step": 53989 }, { "epoch": 16.61, "learning_rate": 1.4728824866969316e-06, "loss": 0.0168, "step": 53990 }, { "epoch": 16.61, "learning_rate": 1.472622210113831e-06, "loss": 0.0062, "step": 53991 }, { "epoch": 16.61, "learning_rate": 1.4723619547019086e-06, "loss": 0.0024, "step": 53992 }, { "epoch": 16.61, "learning_rate": 1.4721017204618148e-06, "loss": 0.0037, "step": 53993 }, { "epoch": 16.61, "learning_rate": 1.4718415073941938e-06, "loss": 0.0035, "step": 53994 }, { "epoch": 16.61, "learning_rate": 1.4715813154996905e-06, "loss": 0.005, "step": 53995 }, { "epoch": 16.61, "learning_rate": 1.4713211447789544e-06, "loss": 0.0061, "step": 53996 }, { "epoch": 16.61, "learning_rate": 1.4710609952326239e-06, "loss": 0.0035, "step": 53997 }, { "epoch": 16.61, "learning_rate": 1.4708008668613515e-06, "loss": 0.0036, "step": 53998 }, { "epoch": 16.61, "learning_rate": 1.4705407596657818e-06, "loss": 0.0031, "step": 53999 }, { "epoch": 16.62, "learning_rate": 1.4702806736465602e-06, "loss": 0.0042, "step": 54000 }, { "epoch": 16.62, "learning_rate": 1.4700206088043312e-06, "loss": 0.0037, "step": 54001 }, { "epoch": 16.62, "learning_rate": 1.4697605651397417e-06, "loss": 0.0048, "step": 54002 }, { "epoch": 16.62, "learning_rate": 1.469500542653437e-06, "loss": 0.006, "step": 54003 }, { "epoch": 16.62, "learning_rate": 1.469240541346063e-06, "loss": 0.0066, "step": 54004 }, { "epoch": 16.62, "learning_rate": 1.468980561218265e-06, "loss": 0.0059, "step": 54005 }, { "epoch": 16.62, "learning_rate": 1.468720602270688e-06, "loss": 0.0065, "step": 54006 }, { "epoch": 16.62, "learning_rate": 1.4684606645039755e-06, "loss": 0.0041, "step": 54007 }, { "epoch": 16.62, "learning_rate": 1.4682007479187777e-06, "loss": 0.0038, "step": 54008 }, { "epoch": 16.62, "learning_rate": 1.46794085251574e-06, "loss": 0.0068, "step": 54009 }, { "epoch": 16.62, "learning_rate": 1.4676809782955014e-06, "loss": 0.0035, "step": 54010 }, { "epoch": 16.62, "learning_rate": 1.4674211252587122e-06, "loss": 0.0067, "step": 54011 }, { "epoch": 16.62, "learning_rate": 1.4671612934060132e-06, "loss": 0.0046, "step": 54012 }, { "epoch": 16.62, "learning_rate": 1.4669014827380534e-06, "loss": 0.0019, "step": 54013 }, { "epoch": 16.62, "learning_rate": 1.4666416932554783e-06, "loss": 0.006, "step": 54014 }, { "epoch": 16.62, "learning_rate": 1.4663819249589306e-06, "loss": 0.0062, "step": 54015 }, { "epoch": 16.62, "learning_rate": 1.4661221778490564e-06, "loss": 0.0054, "step": 54016 }, { "epoch": 16.62, "learning_rate": 1.4658624519264997e-06, "loss": 0.0068, "step": 54017 }, { "epoch": 16.62, "learning_rate": 1.4656027471919064e-06, "loss": 0.0051, "step": 54018 }, { "epoch": 16.62, "learning_rate": 1.4653430636459209e-06, "loss": 0.011, "step": 54019 }, { "epoch": 16.62, "learning_rate": 1.4650834012891868e-06, "loss": 0.0086, "step": 54020 }, { "epoch": 16.62, "learning_rate": 1.4648237601223502e-06, "loss": 0.0067, "step": 54021 }, { "epoch": 16.62, "learning_rate": 1.4645641401460552e-06, "loss": 0.009, "step": 54022 }, { "epoch": 16.62, "learning_rate": 1.464304541360946e-06, "loss": 0.0039, "step": 54023 }, { "epoch": 16.62, "learning_rate": 1.464044963767668e-06, "loss": 0.0052, "step": 54024 }, { "epoch": 16.62, "learning_rate": 1.4637854073668645e-06, "loss": 0.0036, "step": 54025 }, { "epoch": 16.62, "learning_rate": 1.4635258721591805e-06, "loss": 0.0043, "step": 54026 }, { "epoch": 16.62, "learning_rate": 1.463266358145259e-06, "loss": 0.0036, "step": 54027 }, { "epoch": 16.62, "learning_rate": 1.4630068653257467e-06, "loss": 0.0084, "step": 54028 }, { "epoch": 16.62, "learning_rate": 1.462747393701288e-06, "loss": 0.006, "step": 54029 }, { "epoch": 16.62, "learning_rate": 1.4624879432725282e-06, "loss": 0.0047, "step": 54030 }, { "epoch": 16.62, "learning_rate": 1.462228514040106e-06, "loss": 0.004, "step": 54031 }, { "epoch": 16.62, "learning_rate": 1.461969106004667e-06, "loss": 0.0047, "step": 54032 }, { "epoch": 16.63, "learning_rate": 1.4617097191668595e-06, "loss": 0.0048, "step": 54033 }, { "epoch": 16.63, "learning_rate": 1.461450353527325e-06, "loss": 0.0053, "step": 54034 }, { "epoch": 16.63, "learning_rate": 1.4611910090867065e-06, "loss": 0.0034, "step": 54035 }, { "epoch": 16.63, "learning_rate": 1.460931685845649e-06, "loss": 0.0038, "step": 54036 }, { "epoch": 16.63, "learning_rate": 1.4606723838047964e-06, "loss": 0.0044, "step": 54037 }, { "epoch": 16.63, "learning_rate": 1.4604131029647916e-06, "loss": 0.005, "step": 54038 }, { "epoch": 16.63, "learning_rate": 1.4601538433262796e-06, "loss": 0.006, "step": 54039 }, { "epoch": 16.63, "learning_rate": 1.4598946048899033e-06, "loss": 0.0058, "step": 54040 }, { "epoch": 16.63, "learning_rate": 1.4596353876563052e-06, "loss": 0.0033, "step": 54041 }, { "epoch": 16.63, "learning_rate": 1.4593761916261283e-06, "loss": 0.0033, "step": 54042 }, { "epoch": 16.63, "learning_rate": 1.4591170168000235e-06, "loss": 0.0051, "step": 54043 }, { "epoch": 16.63, "learning_rate": 1.4588578631786253e-06, "loss": 0.0077, "step": 54044 }, { "epoch": 16.63, "learning_rate": 1.4585987307625803e-06, "loss": 0.0072, "step": 54045 }, { "epoch": 16.63, "learning_rate": 1.4583396195525324e-06, "loss": 0.0119, "step": 54046 }, { "epoch": 16.63, "learning_rate": 1.4580805295491218e-06, "loss": 0.0044, "step": 54047 }, { "epoch": 16.63, "learning_rate": 1.4578214607529972e-06, "loss": 0.0101, "step": 54048 }, { "epoch": 16.63, "learning_rate": 1.457562413164799e-06, "loss": 0.0088, "step": 54049 }, { "epoch": 16.63, "learning_rate": 1.4573033867851693e-06, "loss": 0.0035, "step": 54050 }, { "epoch": 16.63, "learning_rate": 1.4570443816147561e-06, "loss": 0.0065, "step": 54051 }, { "epoch": 16.63, "learning_rate": 1.4567853976541934e-06, "loss": 0.005, "step": 54052 }, { "epoch": 16.63, "learning_rate": 1.4565264349041307e-06, "loss": 0.0062, "step": 54053 }, { "epoch": 16.63, "learning_rate": 1.456267493365211e-06, "loss": 0.0045, "step": 54054 }, { "epoch": 16.63, "learning_rate": 1.4560085730380745e-06, "loss": 0.0064, "step": 54055 }, { "epoch": 16.63, "learning_rate": 1.4557496739233667e-06, "loss": 0.0044, "step": 54056 }, { "epoch": 16.63, "learning_rate": 1.455490796021728e-06, "loss": 0.0083, "step": 54057 }, { "epoch": 16.63, "learning_rate": 1.4552319393338021e-06, "loss": 0.0034, "step": 54058 }, { "epoch": 16.63, "learning_rate": 1.4549731038602323e-06, "loss": 0.008, "step": 54059 }, { "epoch": 16.63, "learning_rate": 1.454714289601661e-06, "loss": 0.0053, "step": 54060 }, { "epoch": 16.63, "learning_rate": 1.45445549655873e-06, "loss": 0.0077, "step": 54061 }, { "epoch": 16.63, "learning_rate": 1.4541967247320787e-06, "loss": 0.0051, "step": 54062 }, { "epoch": 16.63, "learning_rate": 1.453937974122357e-06, "loss": 0.01, "step": 54063 }, { "epoch": 16.63, "learning_rate": 1.4536792447302051e-06, "loss": 0.0068, "step": 54064 }, { "epoch": 16.64, "learning_rate": 1.4534205365562604e-06, "loss": 0.0048, "step": 54065 }, { "epoch": 16.64, "learning_rate": 1.453161849601169e-06, "loss": 0.0074, "step": 54066 }, { "epoch": 16.64, "learning_rate": 1.4529031838655706e-06, "loss": 0.0058, "step": 54067 }, { "epoch": 16.64, "learning_rate": 1.4526445393501121e-06, "loss": 0.0106, "step": 54068 }, { "epoch": 16.64, "learning_rate": 1.4523859160554322e-06, "loss": 0.0037, "step": 54069 }, { "epoch": 16.64, "learning_rate": 1.4521273139821734e-06, "loss": 0.0044, "step": 54070 }, { "epoch": 16.64, "learning_rate": 1.4518687331309788e-06, "loss": 0.0066, "step": 54071 }, { "epoch": 16.64, "learning_rate": 1.451610173502489e-06, "loss": 0.0062, "step": 54072 }, { "epoch": 16.64, "learning_rate": 1.4513516350973478e-06, "loss": 0.0041, "step": 54073 }, { "epoch": 16.64, "learning_rate": 1.4510931179161948e-06, "loss": 0.0092, "step": 54074 }, { "epoch": 16.64, "learning_rate": 1.4508346219596725e-06, "loss": 0.0059, "step": 54075 }, { "epoch": 16.64, "learning_rate": 1.4505761472284241e-06, "loss": 0.0068, "step": 54076 }, { "epoch": 16.64, "learning_rate": 1.45031769372309e-06, "loss": 0.0039, "step": 54077 }, { "epoch": 16.64, "learning_rate": 1.450059261444311e-06, "loss": 0.0047, "step": 54078 }, { "epoch": 16.64, "learning_rate": 1.4498008503927308e-06, "loss": 0.0059, "step": 54079 }, { "epoch": 16.64, "learning_rate": 1.44954246056899e-06, "loss": 0.014, "step": 54080 }, { "epoch": 16.64, "learning_rate": 1.4492840919737305e-06, "loss": 0.0091, "step": 54081 }, { "epoch": 16.64, "learning_rate": 1.4490257446075905e-06, "loss": 0.0075, "step": 54082 }, { "epoch": 16.64, "learning_rate": 1.4487674184712175e-06, "loss": 0.0076, "step": 54083 }, { "epoch": 16.64, "learning_rate": 1.4485091135652485e-06, "loss": 0.0054, "step": 54084 }, { "epoch": 16.64, "learning_rate": 1.4482508298903286e-06, "loss": 0.0054, "step": 54085 }, { "epoch": 16.64, "learning_rate": 1.447992567447094e-06, "loss": 0.0086, "step": 54086 }, { "epoch": 16.64, "learning_rate": 1.4477343262361864e-06, "loss": 0.0044, "step": 54087 }, { "epoch": 16.64, "learning_rate": 1.4474761062582509e-06, "loss": 0.007, "step": 54088 }, { "epoch": 16.64, "learning_rate": 1.447217907513927e-06, "loss": 0.0058, "step": 54089 }, { "epoch": 16.64, "learning_rate": 1.4469597300038552e-06, "loss": 0.0075, "step": 54090 }, { "epoch": 16.64, "learning_rate": 1.4467015737286761e-06, "loss": 0.0069, "step": 54091 }, { "epoch": 16.64, "learning_rate": 1.4464434386890313e-06, "loss": 0.0065, "step": 54092 }, { "epoch": 16.64, "learning_rate": 1.4461853248855617e-06, "loss": 0.0079, "step": 54093 }, { "epoch": 16.64, "learning_rate": 1.4459272323189078e-06, "loss": 0.0018, "step": 54094 }, { "epoch": 16.64, "learning_rate": 1.44566916098971e-06, "loss": 0.0074, "step": 54095 }, { "epoch": 16.64, "learning_rate": 1.445411110898609e-06, "loss": 0.0048, "step": 54096 }, { "epoch": 16.64, "learning_rate": 1.4451530820462444e-06, "loss": 0.0047, "step": 54097 }, { "epoch": 16.65, "learning_rate": 1.4448950744332635e-06, "loss": 0.0036, "step": 54098 }, { "epoch": 16.65, "learning_rate": 1.444637088060299e-06, "loss": 0.0064, "step": 54099 }, { "epoch": 16.65, "learning_rate": 1.4443791229279946e-06, "loss": 0.0085, "step": 54100 }, { "epoch": 16.65, "learning_rate": 1.4441211790369892e-06, "loss": 0.0053, "step": 54101 }, { "epoch": 16.65, "learning_rate": 1.4438632563879228e-06, "loss": 0.0032, "step": 54102 }, { "epoch": 16.65, "learning_rate": 1.44360535498144e-06, "loss": 0.0062, "step": 54103 }, { "epoch": 16.65, "learning_rate": 1.4433474748181776e-06, "loss": 0.009, "step": 54104 }, { "epoch": 16.65, "learning_rate": 1.4430896158987773e-06, "loss": 0.0054, "step": 54105 }, { "epoch": 16.65, "learning_rate": 1.44283177822388e-06, "loss": 0.0077, "step": 54106 }, { "epoch": 16.65, "learning_rate": 1.4425739617941203e-06, "loss": 0.0067, "step": 54107 }, { "epoch": 16.65, "learning_rate": 1.4423161666101449e-06, "loss": 0.0088, "step": 54108 }, { "epoch": 16.65, "learning_rate": 1.4420583926725918e-06, "loss": 0.0048, "step": 54109 }, { "epoch": 16.65, "learning_rate": 1.4418006399820993e-06, "loss": 0.0038, "step": 54110 }, { "epoch": 16.65, "learning_rate": 1.4415429085393096e-06, "loss": 0.0049, "step": 54111 }, { "epoch": 16.65, "learning_rate": 1.4412851983448605e-06, "loss": 0.0063, "step": 54112 }, { "epoch": 16.65, "learning_rate": 1.4410275093993942e-06, "loss": 0.0067, "step": 54113 }, { "epoch": 16.65, "learning_rate": 1.4407698417035488e-06, "loss": 0.0053, "step": 54114 }, { "epoch": 16.65, "learning_rate": 1.4405121952579638e-06, "loss": 0.0053, "step": 54115 }, { "epoch": 16.65, "learning_rate": 1.4402545700632797e-06, "loss": 0.0056, "step": 54116 }, { "epoch": 16.65, "learning_rate": 1.439996966120134e-06, "loss": 0.0061, "step": 54117 }, { "epoch": 16.65, "learning_rate": 1.4397393834291706e-06, "loss": 0.0029, "step": 54118 }, { "epoch": 16.65, "learning_rate": 1.4394818219910279e-06, "loss": 0.0064, "step": 54119 }, { "epoch": 16.65, "learning_rate": 1.439224281806343e-06, "loss": 0.0045, "step": 54120 }, { "epoch": 16.65, "learning_rate": 1.4389667628757552e-06, "loss": 0.0063, "step": 54121 }, { "epoch": 16.65, "learning_rate": 1.4387092651999025e-06, "loss": 0.0092, "step": 54122 }, { "epoch": 16.65, "learning_rate": 1.4384517887794292e-06, "loss": 0.006, "step": 54123 }, { "epoch": 16.65, "learning_rate": 1.4381943336149727e-06, "loss": 0.0054, "step": 54124 }, { "epoch": 16.65, "learning_rate": 1.4379368997071708e-06, "loss": 0.0091, "step": 54125 }, { "epoch": 16.65, "learning_rate": 1.4376794870566636e-06, "loss": 0.0062, "step": 54126 }, { "epoch": 16.65, "learning_rate": 1.4374220956640895e-06, "loss": 0.0045, "step": 54127 }, { "epoch": 16.65, "learning_rate": 1.437164725530088e-06, "loss": 0.005, "step": 54128 }, { "epoch": 16.65, "learning_rate": 1.4369073766552988e-06, "loss": 0.0076, "step": 54129 }, { "epoch": 16.66, "learning_rate": 1.436650049040359e-06, "loss": 0.0045, "step": 54130 }, { "epoch": 16.66, "learning_rate": 1.4363927426859091e-06, "loss": 0.0043, "step": 54131 }, { "epoch": 16.66, "learning_rate": 1.4361354575925845e-06, "loss": 0.0021, "step": 54132 }, { "epoch": 16.66, "learning_rate": 1.4358781937610323e-06, "loss": 0.0073, "step": 54133 }, { "epoch": 16.66, "learning_rate": 1.4356209511918828e-06, "loss": 0.0057, "step": 54134 }, { "epoch": 16.66, "learning_rate": 1.435363729885778e-06, "loss": 0.0075, "step": 54135 }, { "epoch": 16.66, "learning_rate": 1.4351065298433552e-06, "loss": 0.006, "step": 54136 }, { "epoch": 16.66, "learning_rate": 1.4348493510652529e-06, "loss": 0.0126, "step": 54137 }, { "epoch": 16.66, "learning_rate": 1.4345921935521123e-06, "loss": 0.0048, "step": 54138 }, { "epoch": 16.66, "learning_rate": 1.43433505730457e-06, "loss": 0.0056, "step": 54139 }, { "epoch": 16.66, "learning_rate": 1.4340779423232665e-06, "loss": 0.0035, "step": 54140 }, { "epoch": 16.66, "learning_rate": 1.4338208486088356e-06, "loss": 0.0046, "step": 54141 }, { "epoch": 16.66, "learning_rate": 1.433563776161917e-06, "loss": 0.0041, "step": 54142 }, { "epoch": 16.66, "learning_rate": 1.4333067249831523e-06, "loss": 0.0068, "step": 54143 }, { "epoch": 16.66, "learning_rate": 1.4330496950731765e-06, "loss": 0.0088, "step": 54144 }, { "epoch": 16.66, "learning_rate": 1.432792686432629e-06, "loss": 0.0039, "step": 54145 }, { "epoch": 16.66, "learning_rate": 1.4325356990621486e-06, "loss": 0.0058, "step": 54146 }, { "epoch": 16.66, "learning_rate": 1.432278732962371e-06, "loss": 0.0064, "step": 54147 }, { "epoch": 16.66, "learning_rate": 1.432021788133936e-06, "loss": 0.0042, "step": 54148 }, { "epoch": 16.66, "learning_rate": 1.4317648645774817e-06, "loss": 0.0036, "step": 54149 }, { "epoch": 16.66, "learning_rate": 1.4315079622936446e-06, "loss": 0.0035, "step": 54150 }, { "epoch": 16.66, "learning_rate": 1.4312510812830639e-06, "loss": 0.0077, "step": 54151 }, { "epoch": 16.66, "learning_rate": 1.4309942215463736e-06, "loss": 0.0092, "step": 54152 }, { "epoch": 16.66, "learning_rate": 1.4307373830842174e-06, "loss": 0.0052, "step": 54153 }, { "epoch": 16.66, "learning_rate": 1.430480565897232e-06, "loss": 0.0041, "step": 54154 }, { "epoch": 16.66, "learning_rate": 1.430223769986051e-06, "loss": 0.007, "step": 54155 }, { "epoch": 16.66, "learning_rate": 1.429966995351314e-06, "loss": 0.0047, "step": 54156 }, { "epoch": 16.66, "learning_rate": 1.4297102419936559e-06, "loss": 0.0052, "step": 54157 }, { "epoch": 16.66, "learning_rate": 1.4294535099137197e-06, "loss": 0.0032, "step": 54158 }, { "epoch": 16.66, "learning_rate": 1.429196799112139e-06, "loss": 0.0055, "step": 54159 }, { "epoch": 16.66, "learning_rate": 1.4289401095895528e-06, "loss": 0.0031, "step": 54160 }, { "epoch": 16.66, "learning_rate": 1.4286834413465999e-06, "loss": 0.008, "step": 54161 }, { "epoch": 16.66, "learning_rate": 1.4284267943839102e-06, "loss": 0.0063, "step": 54162 }, { "epoch": 16.67, "learning_rate": 1.4281701687021288e-06, "loss": 0.0033, "step": 54163 }, { "epoch": 16.67, "learning_rate": 1.4279135643018904e-06, "loss": 0.004, "step": 54164 }, { "epoch": 16.67, "learning_rate": 1.4276569811838303e-06, "loss": 0.0056, "step": 54165 }, { "epoch": 16.67, "learning_rate": 1.427400419348588e-06, "loss": 0.0026, "step": 54166 }, { "epoch": 16.67, "learning_rate": 1.4271438787967984e-06, "loss": 0.0067, "step": 54167 }, { "epoch": 16.67, "learning_rate": 1.4268873595291e-06, "loss": 0.0071, "step": 54168 }, { "epoch": 16.67, "learning_rate": 1.4266308615461288e-06, "loss": 0.0077, "step": 54169 }, { "epoch": 16.67, "learning_rate": 1.4263743848485223e-06, "loss": 0.0037, "step": 54170 }, { "epoch": 16.67, "learning_rate": 1.4261179294369164e-06, "loss": 0.0028, "step": 54171 }, { "epoch": 16.67, "learning_rate": 1.4258614953119465e-06, "loss": 0.0076, "step": 54172 }, { "epoch": 16.67, "learning_rate": 1.4256050824742529e-06, "loss": 0.005, "step": 54173 }, { "epoch": 16.67, "learning_rate": 1.4253486909244695e-06, "loss": 0.0042, "step": 54174 }, { "epoch": 16.67, "learning_rate": 1.4250923206632372e-06, "loss": 0.0037, "step": 54175 }, { "epoch": 16.67, "learning_rate": 1.4248359716911863e-06, "loss": 0.0066, "step": 54176 }, { "epoch": 16.67, "learning_rate": 1.4245796440089533e-06, "loss": 0.0025, "step": 54177 }, { "epoch": 16.67, "learning_rate": 1.4243233376171795e-06, "loss": 0.0042, "step": 54178 }, { "epoch": 16.67, "learning_rate": 1.424067052516499e-06, "loss": 0.0052, "step": 54179 }, { "epoch": 16.67, "learning_rate": 1.423810788707548e-06, "loss": 0.0065, "step": 54180 }, { "epoch": 16.67, "learning_rate": 1.4235545461909628e-06, "loss": 0.0097, "step": 54181 }, { "epoch": 16.67, "learning_rate": 1.4232983249673792e-06, "loss": 0.0068, "step": 54182 }, { "epoch": 16.67, "learning_rate": 1.4230421250374338e-06, "loss": 0.0086, "step": 54183 }, { "epoch": 16.67, "learning_rate": 1.4227859464017634e-06, "loss": 0.0142, "step": 54184 }, { "epoch": 16.67, "learning_rate": 1.4225297890610024e-06, "loss": 0.0043, "step": 54185 }, { "epoch": 16.67, "learning_rate": 1.4222736530157855e-06, "loss": 0.0045, "step": 54186 }, { "epoch": 16.67, "learning_rate": 1.4220175382667523e-06, "loss": 0.0048, "step": 54187 }, { "epoch": 16.67, "learning_rate": 1.4217614448145401e-06, "loss": 0.0075, "step": 54188 }, { "epoch": 16.67, "learning_rate": 1.4215053726597782e-06, "loss": 0.0048, "step": 54189 }, { "epoch": 16.67, "learning_rate": 1.4212493218031066e-06, "loss": 0.0059, "step": 54190 }, { "epoch": 16.67, "learning_rate": 1.4209932922451574e-06, "loss": 0.0043, "step": 54191 }, { "epoch": 16.67, "learning_rate": 1.420737283986572e-06, "loss": 0.0053, "step": 54192 }, { "epoch": 16.67, "learning_rate": 1.4204812970279824e-06, "loss": 0.0038, "step": 54193 }, { "epoch": 16.67, "learning_rate": 1.4202253313700253e-06, "loss": 0.0068, "step": 54194 }, { "epoch": 16.68, "learning_rate": 1.4199693870133357e-06, "loss": 0.009, "step": 54195 }, { "epoch": 16.68, "learning_rate": 1.4197134639585486e-06, "loss": 0.0029, "step": 54196 }, { "epoch": 16.68, "learning_rate": 1.4194575622063e-06, "loss": 0.0065, "step": 54197 }, { "epoch": 16.68, "learning_rate": 1.4192016817572252e-06, "loss": 0.0061, "step": 54198 }, { "epoch": 16.68, "learning_rate": 1.418945822611959e-06, "loss": 0.0059, "step": 54199 }, { "epoch": 16.68, "learning_rate": 1.4186899847711366e-06, "loss": 0.009, "step": 54200 }, { "epoch": 16.68, "learning_rate": 1.4184341682353942e-06, "loss": 0.0064, "step": 54201 }, { "epoch": 16.68, "learning_rate": 1.4181783730053656e-06, "loss": 0.0053, "step": 54202 }, { "epoch": 16.68, "learning_rate": 1.417922599081687e-06, "loss": 0.0616, "step": 54203 }, { "epoch": 16.68, "learning_rate": 1.4176668464649934e-06, "loss": 0.006, "step": 54204 }, { "epoch": 16.68, "learning_rate": 1.4174111151559188e-06, "loss": 0.0036, "step": 54205 }, { "epoch": 16.68, "learning_rate": 1.4171554051550963e-06, "loss": 0.0037, "step": 54206 }, { "epoch": 16.68, "learning_rate": 1.4168997164631648e-06, "loss": 0.0068, "step": 54207 }, { "epoch": 16.68, "learning_rate": 1.4166440490807575e-06, "loss": 0.0034, "step": 54208 }, { "epoch": 16.68, "learning_rate": 1.4163884030085117e-06, "loss": 0.0068, "step": 54209 }, { "epoch": 16.68, "learning_rate": 1.4161327782470579e-06, "loss": 0.004, "step": 54210 }, { "epoch": 16.68, "learning_rate": 1.415877174797029e-06, "loss": 0.0076, "step": 54211 }, { "epoch": 16.68, "learning_rate": 1.4156215926590656e-06, "loss": 0.0058, "step": 54212 }, { "epoch": 16.68, "learning_rate": 1.415366031833799e-06, "loss": 0.0108, "step": 54213 }, { "epoch": 16.68, "learning_rate": 1.415110492321864e-06, "loss": 0.0036, "step": 54214 }, { "epoch": 16.68, "learning_rate": 1.4148549741238959e-06, "loss": 0.0043, "step": 54215 }, { "epoch": 16.68, "learning_rate": 1.414599477240528e-06, "loss": 0.0108, "step": 54216 }, { "epoch": 16.68, "learning_rate": 1.4143440016723953e-06, "loss": 0.0047, "step": 54217 }, { "epoch": 16.68, "learning_rate": 1.4140885474201315e-06, "loss": 0.0045, "step": 54218 }, { "epoch": 16.68, "learning_rate": 1.413833114484372e-06, "loss": 0.0045, "step": 54219 }, { "epoch": 16.68, "learning_rate": 1.4135777028657493e-06, "loss": 0.0049, "step": 54220 }, { "epoch": 16.68, "learning_rate": 1.4133223125648966e-06, "loss": 0.0046, "step": 54221 }, { "epoch": 16.68, "learning_rate": 1.413066943582454e-06, "loss": 0.0038, "step": 54222 }, { "epoch": 16.68, "learning_rate": 1.4128115959190492e-06, "loss": 0.0052, "step": 54223 }, { "epoch": 16.68, "learning_rate": 1.412556269575317e-06, "loss": 0.0058, "step": 54224 }, { "epoch": 16.68, "learning_rate": 1.4123009645518938e-06, "loss": 0.0028, "step": 54225 }, { "epoch": 16.68, "learning_rate": 1.4120456808494088e-06, "loss": 0.0071, "step": 54226 }, { "epoch": 16.68, "learning_rate": 1.4117904184685028e-06, "loss": 0.0022, "step": 54227 }, { "epoch": 16.69, "learning_rate": 1.411535177409805e-06, "loss": 0.0048, "step": 54228 }, { "epoch": 16.69, "learning_rate": 1.41127995767395e-06, "loss": 0.0045, "step": 54229 }, { "epoch": 16.69, "learning_rate": 1.4110247592615734e-06, "loss": 0.0072, "step": 54230 }, { "epoch": 16.69, "learning_rate": 1.4107695821733026e-06, "loss": 0.0012, "step": 54231 }, { "epoch": 16.69, "learning_rate": 1.4105144264097781e-06, "loss": 0.0068, "step": 54232 }, { "epoch": 16.69, "learning_rate": 1.4102592919716295e-06, "loss": 0.0078, "step": 54233 }, { "epoch": 16.69, "learning_rate": 1.4100041788594921e-06, "loss": 0.006, "step": 54234 }, { "epoch": 16.69, "learning_rate": 1.4097490870739982e-06, "loss": 0.0091, "step": 54235 }, { "epoch": 16.69, "learning_rate": 1.4094940166157823e-06, "loss": 0.0099, "step": 54236 }, { "epoch": 16.69, "learning_rate": 1.4092389674854757e-06, "loss": 0.0022, "step": 54237 }, { "epoch": 16.69, "learning_rate": 1.4089839396837134e-06, "loss": 0.0041, "step": 54238 }, { "epoch": 16.69, "learning_rate": 1.4087289332111276e-06, "loss": 0.0058, "step": 54239 }, { "epoch": 16.69, "learning_rate": 1.4084739480683518e-06, "loss": 0.0032, "step": 54240 }, { "epoch": 16.69, "learning_rate": 1.4082189842560178e-06, "loss": 0.0037, "step": 54241 }, { "epoch": 16.69, "learning_rate": 1.407964041774761e-06, "loss": 0.0119, "step": 54242 }, { "epoch": 16.69, "learning_rate": 1.4077091206252146e-06, "loss": 0.008, "step": 54243 }, { "epoch": 16.69, "learning_rate": 1.407454220808009e-06, "loss": 0.0042, "step": 54244 }, { "epoch": 16.69, "learning_rate": 1.4071993423237783e-06, "loss": 0.0055, "step": 54245 }, { "epoch": 16.69, "learning_rate": 1.4069444851731517e-06, "loss": 0.0044, "step": 54246 }, { "epoch": 16.69, "learning_rate": 1.4066896493567684e-06, "loss": 0.0038, "step": 54247 }, { "epoch": 16.69, "learning_rate": 1.4064348348752577e-06, "loss": 0.0025, "step": 54248 }, { "epoch": 16.69, "learning_rate": 1.4061800417292527e-06, "loss": 0.0046, "step": 54249 }, { "epoch": 16.69, "learning_rate": 1.405925269919386e-06, "loss": 0.0036, "step": 54250 }, { "epoch": 16.69, "learning_rate": 1.4056705194462894e-06, "loss": 0.0107, "step": 54251 }, { "epoch": 16.69, "learning_rate": 1.4054157903105968e-06, "loss": 0.0058, "step": 54252 }, { "epoch": 16.69, "learning_rate": 1.4051610825129392e-06, "loss": 0.0046, "step": 54253 }, { "epoch": 16.69, "learning_rate": 1.4049063960539488e-06, "loss": 0.0037, "step": 54254 }, { "epoch": 16.69, "learning_rate": 1.40465173093426e-06, "loss": 0.0039, "step": 54255 }, { "epoch": 16.69, "learning_rate": 1.4043970871545032e-06, "loss": 0.0052, "step": 54256 }, { "epoch": 16.69, "learning_rate": 1.4041424647153112e-06, "loss": 0.0027, "step": 54257 }, { "epoch": 16.69, "learning_rate": 1.4038878636173158e-06, "loss": 0.0054, "step": 54258 }, { "epoch": 16.69, "learning_rate": 1.4036332838611489e-06, "loss": 0.0042, "step": 54259 }, { "epoch": 16.7, "learning_rate": 1.403378725447444e-06, "loss": 0.0054, "step": 54260 }, { "epoch": 16.7, "learning_rate": 1.4031241883768297e-06, "loss": 0.0068, "step": 54261 }, { "epoch": 16.7, "learning_rate": 1.4028696726499424e-06, "loss": 0.0087, "step": 54262 }, { "epoch": 16.7, "learning_rate": 1.4026151782674125e-06, "loss": 0.0046, "step": 54263 }, { "epoch": 16.7, "learning_rate": 1.4023607052298727e-06, "loss": 0.0046, "step": 54264 }, { "epoch": 16.7, "learning_rate": 1.4021062535379525e-06, "loss": 0.0036, "step": 54265 }, { "epoch": 16.7, "learning_rate": 1.4018518231922806e-06, "loss": 0.0047, "step": 54266 }, { "epoch": 16.7, "learning_rate": 1.4015974141934962e-06, "loss": 0.0074, "step": 54267 }, { "epoch": 16.7, "learning_rate": 1.4013430265422267e-06, "loss": 0.0033, "step": 54268 }, { "epoch": 16.7, "learning_rate": 1.4010886602391049e-06, "loss": 0.0032, "step": 54269 }, { "epoch": 16.7, "learning_rate": 1.4008343152847615e-06, "loss": 0.0037, "step": 54270 }, { "epoch": 16.7, "learning_rate": 1.400579991679828e-06, "loss": 0.0064, "step": 54271 }, { "epoch": 16.7, "learning_rate": 1.4003256894249362e-06, "loss": 0.0054, "step": 54272 }, { "epoch": 16.7, "learning_rate": 1.4000714085207167e-06, "loss": 0.0056, "step": 54273 }, { "epoch": 16.7, "learning_rate": 1.3998171489678026e-06, "loss": 0.005, "step": 54274 }, { "epoch": 16.7, "learning_rate": 1.399562910766823e-06, "loss": 0.0034, "step": 54275 }, { "epoch": 16.7, "learning_rate": 1.3993086939184086e-06, "loss": 0.0026, "step": 54276 }, { "epoch": 16.7, "learning_rate": 1.399054498423197e-06, "loss": 0.0037, "step": 54277 }, { "epoch": 16.7, "learning_rate": 1.3988003242818104e-06, "loss": 0.0059, "step": 54278 }, { "epoch": 16.7, "learning_rate": 1.3985461714948844e-06, "loss": 0.0073, "step": 54279 }, { "epoch": 16.7, "learning_rate": 1.3982920400630505e-06, "loss": 0.0041, "step": 54280 }, { "epoch": 16.7, "learning_rate": 1.398037929986935e-06, "loss": 0.005, "step": 54281 }, { "epoch": 16.7, "learning_rate": 1.397783841267175e-06, "loss": 0.0081, "step": 54282 }, { "epoch": 16.7, "learning_rate": 1.3975297739043992e-06, "loss": 0.0086, "step": 54283 }, { "epoch": 16.7, "learning_rate": 1.3972757278992365e-06, "loss": 0.0064, "step": 54284 }, { "epoch": 16.7, "learning_rate": 1.3970217032523226e-06, "loss": 0.0043, "step": 54285 }, { "epoch": 16.7, "learning_rate": 1.3967676999642788e-06, "loss": 0.0106, "step": 54286 }, { "epoch": 16.7, "learning_rate": 1.3965137180357447e-06, "loss": 0.0083, "step": 54287 }, { "epoch": 16.7, "learning_rate": 1.3962597574673465e-06, "loss": 0.0116, "step": 54288 }, { "epoch": 16.7, "learning_rate": 1.396005818259717e-06, "loss": 0.0069, "step": 54289 }, { "epoch": 16.7, "learning_rate": 1.3957519004134845e-06, "loss": 0.007, "step": 54290 }, { "epoch": 16.7, "learning_rate": 1.395498003929281e-06, "loss": 0.0059, "step": 54291 }, { "epoch": 16.7, "learning_rate": 1.3952441288077357e-06, "loss": 0.0045, "step": 54292 }, { "epoch": 16.71, "learning_rate": 1.3949902750494805e-06, "loss": 0.0043, "step": 54293 }, { "epoch": 16.71, "learning_rate": 1.3947364426551436e-06, "loss": 0.0071, "step": 54294 }, { "epoch": 16.71, "learning_rate": 1.3944826316253567e-06, "loss": 0.0059, "step": 54295 }, { "epoch": 16.71, "learning_rate": 1.3942288419607476e-06, "loss": 0.0053, "step": 54296 }, { "epoch": 16.71, "learning_rate": 1.3939750736619495e-06, "loss": 0.0028, "step": 54297 }, { "epoch": 16.71, "learning_rate": 1.3937213267295935e-06, "loss": 0.0045, "step": 54298 }, { "epoch": 16.71, "learning_rate": 1.3934676011643055e-06, "loss": 0.0037, "step": 54299 }, { "epoch": 16.71, "learning_rate": 1.3932138969667175e-06, "loss": 0.006, "step": 54300 }, { "epoch": 16.71, "learning_rate": 1.3929602141374564e-06, "loss": 0.0054, "step": 54301 }, { "epoch": 16.71, "learning_rate": 1.3927065526771565e-06, "loss": 0.0103, "step": 54302 }, { "epoch": 16.71, "learning_rate": 1.392452912586446e-06, "loss": 0.007, "step": 54303 }, { "epoch": 16.71, "learning_rate": 1.3921992938659557e-06, "loss": 0.0046, "step": 54304 }, { "epoch": 16.71, "learning_rate": 1.3919456965163125e-06, "loss": 0.0042, "step": 54305 }, { "epoch": 16.71, "learning_rate": 1.3916921205381473e-06, "loss": 0.0062, "step": 54306 }, { "epoch": 16.71, "learning_rate": 1.3914385659320907e-06, "loss": 0.0049, "step": 54307 }, { "epoch": 16.71, "learning_rate": 1.3911850326987709e-06, "loss": 0.0124, "step": 54308 }, { "epoch": 16.71, "learning_rate": 1.3909315208388185e-06, "loss": 0.0053, "step": 54309 }, { "epoch": 16.71, "learning_rate": 1.390678030352861e-06, "loss": 0.0064, "step": 54310 }, { "epoch": 16.71, "learning_rate": 1.390424561241528e-06, "loss": 0.0065, "step": 54311 }, { "epoch": 16.71, "learning_rate": 1.390171113505453e-06, "loss": 0.0093, "step": 54312 }, { "epoch": 16.71, "learning_rate": 1.3899176871452614e-06, "loss": 0.0079, "step": 54313 }, { "epoch": 16.71, "learning_rate": 1.3896642821615813e-06, "loss": 0.0106, "step": 54314 }, { "epoch": 16.71, "learning_rate": 1.3894108985550437e-06, "loss": 0.0061, "step": 54315 }, { "epoch": 16.71, "learning_rate": 1.3891575363262766e-06, "loss": 0.0043, "step": 54316 }, { "epoch": 16.71, "learning_rate": 1.3889041954759108e-06, "loss": 0.0068, "step": 54317 }, { "epoch": 16.71, "learning_rate": 1.3886508760045747e-06, "loss": 0.0061, "step": 54318 }, { "epoch": 16.71, "learning_rate": 1.3883975779129e-06, "loss": 0.0032, "step": 54319 }, { "epoch": 16.71, "learning_rate": 1.3881443012015095e-06, "loss": 0.0087, "step": 54320 }, { "epoch": 16.71, "learning_rate": 1.3878910458710327e-06, "loss": 0.0055, "step": 54321 }, { "epoch": 16.71, "learning_rate": 1.3876378119221024e-06, "loss": 0.0067, "step": 54322 }, { "epoch": 16.71, "learning_rate": 1.387384599355347e-06, "loss": 0.0027, "step": 54323 }, { "epoch": 16.71, "learning_rate": 1.3871314081713927e-06, "loss": 0.0037, "step": 54324 }, { "epoch": 16.72, "learning_rate": 1.386878238370869e-06, "loss": 0.0067, "step": 54325 }, { "epoch": 16.72, "learning_rate": 1.3866250899544054e-06, "loss": 0.0071, "step": 54326 }, { "epoch": 16.72, "learning_rate": 1.386371962922629e-06, "loss": 0.0076, "step": 54327 }, { "epoch": 16.72, "learning_rate": 1.3861188572761696e-06, "loss": 0.0047, "step": 54328 }, { "epoch": 16.72, "learning_rate": 1.3858657730156544e-06, "loss": 0.0016, "step": 54329 }, { "epoch": 16.72, "learning_rate": 1.3856127101417127e-06, "loss": 0.0068, "step": 54330 }, { "epoch": 16.72, "learning_rate": 1.3853596686549686e-06, "loss": 0.0106, "step": 54331 }, { "epoch": 16.72, "learning_rate": 1.385106648556057e-06, "loss": 0.004, "step": 54332 }, { "epoch": 16.72, "learning_rate": 1.3848536498456054e-06, "loss": 0.0056, "step": 54333 }, { "epoch": 16.72, "learning_rate": 1.3846006725242377e-06, "loss": 0.0069, "step": 54334 }, { "epoch": 16.72, "learning_rate": 1.3843477165925846e-06, "loss": 0.0135, "step": 54335 }, { "epoch": 16.72, "learning_rate": 1.3840947820512696e-06, "loss": 0.0066, "step": 54336 }, { "epoch": 16.72, "learning_rate": 1.383841868900927e-06, "loss": 0.0063, "step": 54337 }, { "epoch": 16.72, "learning_rate": 1.3835889771421829e-06, "loss": 0.0032, "step": 54338 }, { "epoch": 16.72, "learning_rate": 1.3833361067756646e-06, "loss": 0.0124, "step": 54339 }, { "epoch": 16.72, "learning_rate": 1.3830832578020016e-06, "loss": 0.0082, "step": 54340 }, { "epoch": 16.72, "learning_rate": 1.3828304302218142e-06, "loss": 0.004, "step": 54341 }, { "epoch": 16.72, "learning_rate": 1.3825776240357391e-06, "loss": 0.0062, "step": 54342 }, { "epoch": 16.72, "learning_rate": 1.382324839244401e-06, "loss": 0.0091, "step": 54343 }, { "epoch": 16.72, "learning_rate": 1.3820720758484262e-06, "loss": 0.0074, "step": 54344 }, { "epoch": 16.72, "learning_rate": 1.3818193338484443e-06, "loss": 0.0057, "step": 54345 }, { "epoch": 16.72, "learning_rate": 1.38156661324508e-06, "loss": 0.0044, "step": 54346 }, { "epoch": 16.72, "learning_rate": 1.3813139140389632e-06, "loss": 0.007, "step": 54347 }, { "epoch": 16.72, "learning_rate": 1.3810612362307208e-06, "loss": 0.0041, "step": 54348 }, { "epoch": 16.72, "learning_rate": 1.3808085798209791e-06, "loss": 0.0027, "step": 54349 }, { "epoch": 16.72, "learning_rate": 1.3805559448103666e-06, "loss": 0.0033, "step": 54350 }, { "epoch": 16.72, "learning_rate": 1.3803033311995072e-06, "loss": 0.0057, "step": 54351 }, { "epoch": 16.72, "learning_rate": 1.3800507389890327e-06, "loss": 0.0033, "step": 54352 }, { "epoch": 16.72, "learning_rate": 1.3797981681795691e-06, "loss": 0.0063, "step": 54353 }, { "epoch": 16.72, "learning_rate": 1.3795456187717448e-06, "loss": 0.0043, "step": 54354 }, { "epoch": 16.72, "learning_rate": 1.3792930907661828e-06, "loss": 0.0038, "step": 54355 }, { "epoch": 16.72, "learning_rate": 1.379040584163509e-06, "loss": 0.0055, "step": 54356 }, { "epoch": 16.72, "learning_rate": 1.3787880989643553e-06, "loss": 0.0043, "step": 54357 }, { "epoch": 16.73, "learning_rate": 1.378535635169348e-06, "loss": 0.0089, "step": 54358 }, { "epoch": 16.73, "learning_rate": 1.3782831927791107e-06, "loss": 0.0066, "step": 54359 }, { "epoch": 16.73, "learning_rate": 1.378030771794273e-06, "loss": 0.0051, "step": 54360 }, { "epoch": 16.73, "learning_rate": 1.3777783722154603e-06, "loss": 0.0045, "step": 54361 }, { "epoch": 16.73, "learning_rate": 1.3775259940432995e-06, "loss": 0.0072, "step": 54362 }, { "epoch": 16.73, "learning_rate": 1.3772736372784169e-06, "loss": 0.0119, "step": 54363 }, { "epoch": 16.73, "learning_rate": 1.3770213019214395e-06, "loss": 0.0068, "step": 54364 }, { "epoch": 16.73, "learning_rate": 1.3767689879729928e-06, "loss": 0.0085, "step": 54365 }, { "epoch": 16.73, "learning_rate": 1.3765166954337029e-06, "loss": 0.0064, "step": 54366 }, { "epoch": 16.73, "learning_rate": 1.376264424304201e-06, "loss": 0.0162, "step": 54367 }, { "epoch": 16.73, "learning_rate": 1.3760121745851074e-06, "loss": 0.0089, "step": 54368 }, { "epoch": 16.73, "learning_rate": 1.375759946277051e-06, "loss": 0.0059, "step": 54369 }, { "epoch": 16.73, "learning_rate": 1.375507739380657e-06, "loss": 0.0085, "step": 54370 }, { "epoch": 16.73, "learning_rate": 1.3752555538965507e-06, "loss": 0.0032, "step": 54371 }, { "epoch": 16.73, "learning_rate": 1.3750033898253613e-06, "loss": 0.0073, "step": 54372 }, { "epoch": 16.73, "learning_rate": 1.3747512471677138e-06, "loss": 0.0062, "step": 54373 }, { "epoch": 16.73, "learning_rate": 1.3744991259242324e-06, "loss": 0.0045, "step": 54374 }, { "epoch": 16.73, "learning_rate": 1.3742470260955477e-06, "loss": 0.0084, "step": 54375 }, { "epoch": 16.73, "learning_rate": 1.3739949476822777e-06, "loss": 0.0054, "step": 54376 }, { "epoch": 16.73, "learning_rate": 1.3737428906850546e-06, "loss": 0.0037, "step": 54377 }, { "epoch": 16.73, "learning_rate": 1.3734908551045023e-06, "loss": 0.0091, "step": 54378 }, { "epoch": 16.73, "learning_rate": 1.3732388409412466e-06, "loss": 0.0099, "step": 54379 }, { "epoch": 16.73, "learning_rate": 1.372986848195914e-06, "loss": 0.0031, "step": 54380 }, { "epoch": 16.73, "learning_rate": 1.3727348768691285e-06, "loss": 0.0033, "step": 54381 }, { "epoch": 16.73, "learning_rate": 1.372482926961517e-06, "loss": 0.0102, "step": 54382 }, { "epoch": 16.73, "learning_rate": 1.3722309984737036e-06, "loss": 0.0071, "step": 54383 }, { "epoch": 16.73, "learning_rate": 1.3719790914063147e-06, "loss": 0.0038, "step": 54384 }, { "epoch": 16.73, "learning_rate": 1.3717272057599762e-06, "loss": 0.0037, "step": 54385 }, { "epoch": 16.73, "learning_rate": 1.371475341535311e-06, "loss": 0.0094, "step": 54386 }, { "epoch": 16.73, "learning_rate": 1.3712234987329486e-06, "loss": 0.0073, "step": 54387 }, { "epoch": 16.73, "learning_rate": 1.3709716773535132e-06, "loss": 0.0069, "step": 54388 }, { "epoch": 16.73, "learning_rate": 1.3707198773976272e-06, "loss": 0.0143, "step": 54389 }, { "epoch": 16.74, "learning_rate": 1.3704680988659181e-06, "loss": 0.007, "step": 54390 }, { "epoch": 16.74, "learning_rate": 1.3702163417590064e-06, "loss": 0.0139, "step": 54391 }, { "epoch": 16.74, "learning_rate": 1.369964606077524e-06, "loss": 0.0044, "step": 54392 }, { "epoch": 16.74, "learning_rate": 1.3697128918220936e-06, "loss": 0.0049, "step": 54393 }, { "epoch": 16.74, "learning_rate": 1.3694611989933393e-06, "loss": 0.0085, "step": 54394 }, { "epoch": 16.74, "learning_rate": 1.369209527591885e-06, "loss": 0.0042, "step": 54395 }, { "epoch": 16.74, "learning_rate": 1.3689578776183577e-06, "loss": 0.0111, "step": 54396 }, { "epoch": 16.74, "learning_rate": 1.3687062490733804e-06, "loss": 0.0056, "step": 54397 }, { "epoch": 16.74, "learning_rate": 1.3684546419575794e-06, "loss": 0.0035, "step": 54398 }, { "epoch": 16.74, "learning_rate": 1.3682030562715776e-06, "loss": 0.0082, "step": 54399 }, { "epoch": 16.74, "learning_rate": 1.367951492016001e-06, "loss": 0.0068, "step": 54400 }, { "epoch": 16.74, "learning_rate": 1.3676999491914745e-06, "loss": 0.0058, "step": 54401 }, { "epoch": 16.74, "learning_rate": 1.36744842779862e-06, "loss": 0.006, "step": 54402 }, { "epoch": 16.74, "learning_rate": 1.3671969278380648e-06, "loss": 0.0056, "step": 54403 }, { "epoch": 16.74, "learning_rate": 1.366945449310433e-06, "loss": 0.0033, "step": 54404 }, { "epoch": 16.74, "learning_rate": 1.366693992216348e-06, "loss": 0.0031, "step": 54405 }, { "epoch": 16.74, "learning_rate": 1.366442556556431e-06, "loss": 0.0069, "step": 54406 }, { "epoch": 16.74, "learning_rate": 1.3661911423313123e-06, "loss": 0.0048, "step": 54407 }, { "epoch": 16.74, "learning_rate": 1.3659397495416138e-06, "loss": 0.005, "step": 54408 }, { "epoch": 16.74, "learning_rate": 1.3656883781879616e-06, "loss": 0.0047, "step": 54409 }, { "epoch": 16.74, "learning_rate": 1.365437028270975e-06, "loss": 0.0046, "step": 54410 }, { "epoch": 16.74, "learning_rate": 1.3651856997912782e-06, "loss": 0.009, "step": 54411 }, { "epoch": 16.74, "learning_rate": 1.3649343927494995e-06, "loss": 0.0092, "step": 54412 }, { "epoch": 16.74, "learning_rate": 1.3646831071462606e-06, "loss": 0.0041, "step": 54413 }, { "epoch": 16.74, "learning_rate": 1.3644318429821856e-06, "loss": 0.0088, "step": 54414 }, { "epoch": 16.74, "learning_rate": 1.3641806002578984e-06, "loss": 0.0052, "step": 54415 }, { "epoch": 16.74, "learning_rate": 1.3639293789740227e-06, "loss": 0.0079, "step": 54416 }, { "epoch": 16.74, "learning_rate": 1.3636781791311826e-06, "loss": 0.0064, "step": 54417 }, { "epoch": 16.74, "learning_rate": 1.363427000730001e-06, "loss": 0.0063, "step": 54418 }, { "epoch": 16.74, "learning_rate": 1.3631758437711007e-06, "loss": 0.0055, "step": 54419 }, { "epoch": 16.74, "learning_rate": 1.3629247082551079e-06, "loss": 0.0128, "step": 54420 }, { "epoch": 16.74, "learning_rate": 1.362673594182642e-06, "loss": 0.0047, "step": 54421 }, { "epoch": 16.74, "learning_rate": 1.3624225015543325e-06, "loss": 0.0042, "step": 54422 }, { "epoch": 16.75, "learning_rate": 1.3621714303707978e-06, "loss": 0.0085, "step": 54423 }, { "epoch": 16.75, "learning_rate": 1.3619203806326632e-06, "loss": 0.0039, "step": 54424 }, { "epoch": 16.75, "learning_rate": 1.361669352340551e-06, "loss": 0.0071, "step": 54425 }, { "epoch": 16.75, "learning_rate": 1.3614183454950824e-06, "loss": 0.0051, "step": 54426 }, { "epoch": 16.75, "learning_rate": 1.3611673600968855e-06, "loss": 0.0039, "step": 54427 }, { "epoch": 16.75, "learning_rate": 1.3609163961465822e-06, "loss": 0.0073, "step": 54428 }, { "epoch": 16.75, "learning_rate": 1.3606654536447927e-06, "loss": 0.008, "step": 54429 }, { "epoch": 16.75, "learning_rate": 1.3604145325921448e-06, "loss": 0.0081, "step": 54430 }, { "epoch": 16.75, "learning_rate": 1.3601636329892543e-06, "loss": 0.0052, "step": 54431 }, { "epoch": 16.75, "learning_rate": 1.3599127548367508e-06, "loss": 0.0036, "step": 54432 }, { "epoch": 16.75, "learning_rate": 1.359661898135254e-06, "loss": 0.0033, "step": 54433 }, { "epoch": 16.75, "learning_rate": 1.3594110628853874e-06, "loss": 0.0112, "step": 54434 }, { "epoch": 16.75, "learning_rate": 1.3591602490877742e-06, "loss": 0.0058, "step": 54435 }, { "epoch": 16.75, "learning_rate": 1.3589094567430362e-06, "loss": 0.0156, "step": 54436 }, { "epoch": 16.75, "learning_rate": 1.3586586858517958e-06, "loss": 0.0042, "step": 54437 }, { "epoch": 16.75, "learning_rate": 1.3584079364146774e-06, "loss": 0.007, "step": 54438 }, { "epoch": 16.75, "learning_rate": 1.3581572084323014e-06, "loss": 0.0061, "step": 54439 }, { "epoch": 16.75, "learning_rate": 1.3579065019052929e-06, "loss": 0.0079, "step": 54440 }, { "epoch": 16.75, "learning_rate": 1.357655816834269e-06, "loss": 0.0034, "step": 54441 }, { "epoch": 16.75, "learning_rate": 1.3574051532198585e-06, "loss": 0.0066, "step": 54442 }, { "epoch": 16.75, "learning_rate": 1.3571545110626827e-06, "loss": 0.0051, "step": 54443 }, { "epoch": 16.75, "learning_rate": 1.3569038903633602e-06, "loss": 0.004, "step": 54444 }, { "epoch": 16.75, "learning_rate": 1.356653291122516e-06, "loss": 0.0065, "step": 54445 }, { "epoch": 16.75, "learning_rate": 1.3564027133407686e-06, "loss": 0.0092, "step": 54446 }, { "epoch": 16.75, "learning_rate": 1.3561521570187453e-06, "loss": 0.0137, "step": 54447 }, { "epoch": 16.75, "learning_rate": 1.3559016221570663e-06, "loss": 0.0046, "step": 54448 }, { "epoch": 16.75, "learning_rate": 1.3556511087563529e-06, "loss": 0.005, "step": 54449 }, { "epoch": 16.75, "learning_rate": 1.3554006168172274e-06, "loss": 0.0046, "step": 54450 }, { "epoch": 16.75, "learning_rate": 1.3551501463403115e-06, "loss": 0.005, "step": 54451 }, { "epoch": 16.75, "learning_rate": 1.3548996973262273e-06, "loss": 0.003, "step": 54452 }, { "epoch": 16.75, "learning_rate": 1.3546492697755963e-06, "loss": 0.0045, "step": 54453 }, { "epoch": 16.75, "learning_rate": 1.3543988636890415e-06, "loss": 0.0069, "step": 54454 }, { "epoch": 16.76, "learning_rate": 1.354148479067182e-06, "loss": 0.0044, "step": 54455 }, { "epoch": 16.76, "learning_rate": 1.3538981159106423e-06, "loss": 0.0033, "step": 54456 }, { "epoch": 16.76, "learning_rate": 1.3536477742200415e-06, "loss": 0.0046, "step": 54457 }, { "epoch": 16.76, "learning_rate": 1.3533974539960025e-06, "loss": 0.0046, "step": 54458 }, { "epoch": 16.76, "learning_rate": 1.3531471552391473e-06, "loss": 0.0036, "step": 54459 }, { "epoch": 16.76, "learning_rate": 1.352896877950096e-06, "loss": 0.0041, "step": 54460 }, { "epoch": 16.76, "learning_rate": 1.3526466221294687e-06, "loss": 0.0039, "step": 54461 }, { "epoch": 16.76, "learning_rate": 1.3523963877778912e-06, "loss": 0.0047, "step": 54462 }, { "epoch": 16.76, "learning_rate": 1.3521461748959808e-06, "loss": 0.0049, "step": 54463 }, { "epoch": 16.76, "learning_rate": 1.3518959834843636e-06, "loss": 0.0049, "step": 54464 }, { "epoch": 16.76, "learning_rate": 1.3516458135436539e-06, "loss": 0.0104, "step": 54465 }, { "epoch": 16.76, "learning_rate": 1.3513956650744753e-06, "loss": 0.0116, "step": 54466 }, { "epoch": 16.76, "learning_rate": 1.3511455380774507e-06, "loss": 0.0062, "step": 54467 }, { "epoch": 16.76, "learning_rate": 1.3508954325532008e-06, "loss": 0.0069, "step": 54468 }, { "epoch": 16.76, "learning_rate": 1.3506453485023452e-06, "loss": 0.013, "step": 54469 }, { "epoch": 16.76, "learning_rate": 1.3503952859255054e-06, "loss": 0.0033, "step": 54470 }, { "epoch": 16.76, "learning_rate": 1.350145244823302e-06, "loss": 0.0035, "step": 54471 }, { "epoch": 16.76, "learning_rate": 1.349895225196357e-06, "loss": 0.0076, "step": 54472 }, { "epoch": 16.76, "learning_rate": 1.3496452270452898e-06, "loss": 0.0058, "step": 54473 }, { "epoch": 16.76, "learning_rate": 1.3493952503707208e-06, "loss": 0.0062, "step": 54474 }, { "epoch": 16.76, "learning_rate": 1.349145295173272e-06, "loss": 0.0085, "step": 54475 }, { "epoch": 16.76, "learning_rate": 1.3488953614535615e-06, "loss": 0.0032, "step": 54476 }, { "epoch": 16.76, "learning_rate": 1.3486454492122147e-06, "loss": 0.008, "step": 54477 }, { "epoch": 16.76, "learning_rate": 1.3483955584498476e-06, "loss": 0.0054, "step": 54478 }, { "epoch": 16.76, "learning_rate": 1.3481456891670818e-06, "loss": 0.0062, "step": 54479 }, { "epoch": 16.76, "learning_rate": 1.347895841364537e-06, "loss": 0.0046, "step": 54480 }, { "epoch": 16.76, "learning_rate": 1.3476460150428327e-06, "loss": 0.0077, "step": 54481 }, { "epoch": 16.76, "learning_rate": 1.3473962102025927e-06, "loss": 0.0028, "step": 54482 }, { "epoch": 16.76, "learning_rate": 1.3471464268444356e-06, "loss": 0.0046, "step": 54483 }, { "epoch": 16.76, "learning_rate": 1.346896664968982e-06, "loss": 0.0089, "step": 54484 }, { "epoch": 16.76, "learning_rate": 1.346646924576852e-06, "loss": 0.0076, "step": 54485 }, { "epoch": 16.76, "learning_rate": 1.3463972056686604e-06, "loss": 0.01, "step": 54486 }, { "epoch": 16.76, "learning_rate": 1.3461475082450349e-06, "loss": 0.0068, "step": 54487 }, { "epoch": 16.77, "learning_rate": 1.3458978323065909e-06, "loss": 0.0056, "step": 54488 }, { "epoch": 16.77, "learning_rate": 1.3456481778539499e-06, "loss": 0.0064, "step": 54489 }, { "epoch": 16.77, "learning_rate": 1.3453985448877315e-06, "loss": 0.0071, "step": 54490 }, { "epoch": 16.77, "learning_rate": 1.3451489334085555e-06, "loss": 0.007, "step": 54491 }, { "epoch": 16.77, "learning_rate": 1.344899343417042e-06, "loss": 0.0107, "step": 54492 }, { "epoch": 16.77, "learning_rate": 1.3446497749138087e-06, "loss": 0.0062, "step": 54493 }, { "epoch": 16.77, "learning_rate": 1.3444002278994773e-06, "loss": 0.0055, "step": 54494 }, { "epoch": 16.77, "learning_rate": 1.344150702374667e-06, "loss": 0.0033, "step": 54495 }, { "epoch": 16.77, "learning_rate": 1.343901198339994e-06, "loss": 0.0026, "step": 54496 }, { "epoch": 16.77, "learning_rate": 1.3436517157960837e-06, "loss": 0.0082, "step": 54497 }, { "epoch": 16.77, "learning_rate": 1.3434022547435544e-06, "loss": 0.0062, "step": 54498 }, { "epoch": 16.77, "learning_rate": 1.343152815183021e-06, "loss": 0.0024, "step": 54499 }, { "epoch": 16.77, "learning_rate": 1.342903397115105e-06, "loss": 0.0058, "step": 54500 }, { "epoch": 16.77, "learning_rate": 1.3426540005404243e-06, "loss": 0.0043, "step": 54501 }, { "epoch": 16.77, "learning_rate": 1.3424046254596024e-06, "loss": 0.0062, "step": 54502 }, { "epoch": 16.77, "learning_rate": 1.3421552718732555e-06, "loss": 0.005, "step": 54503 }, { "epoch": 16.77, "learning_rate": 1.3419059397820033e-06, "loss": 0.0093, "step": 54504 }, { "epoch": 16.77, "learning_rate": 1.341656629186464e-06, "loss": 0.0075, "step": 54505 }, { "epoch": 16.77, "learning_rate": 1.3414073400872574e-06, "loss": 0.0173, "step": 54506 }, { "epoch": 16.77, "learning_rate": 1.3411580724850015e-06, "loss": 0.0063, "step": 54507 }, { "epoch": 16.77, "learning_rate": 1.340908826380316e-06, "loss": 0.0065, "step": 54508 }, { "epoch": 16.77, "learning_rate": 1.3406596017738195e-06, "loss": 0.0049, "step": 54509 }, { "epoch": 16.77, "learning_rate": 1.340410398666131e-06, "loss": 0.0052, "step": 54510 }, { "epoch": 16.77, "learning_rate": 1.340161217057866e-06, "loss": 0.01, "step": 54511 }, { "epoch": 16.77, "learning_rate": 1.3399120569496516e-06, "loss": 0.0089, "step": 54512 }, { "epoch": 16.77, "learning_rate": 1.3396629183420984e-06, "loss": 0.0069, "step": 54513 }, { "epoch": 16.77, "learning_rate": 1.3394138012358271e-06, "loss": 0.0077, "step": 54514 }, { "epoch": 16.77, "learning_rate": 1.339164705631456e-06, "loss": 0.0056, "step": 54515 }, { "epoch": 16.77, "learning_rate": 1.3389156315296025e-06, "loss": 0.0034, "step": 54516 }, { "epoch": 16.77, "learning_rate": 1.3386665789308885e-06, "loss": 0.0079, "step": 54517 }, { "epoch": 16.77, "learning_rate": 1.3384175478359307e-06, "loss": 0.0046, "step": 54518 }, { "epoch": 16.77, "learning_rate": 1.3381685382453492e-06, "loss": 0.0087, "step": 54519 }, { "epoch": 16.78, "learning_rate": 1.3379195501597575e-06, "loss": 0.0062, "step": 54520 }, { "epoch": 16.78, "learning_rate": 1.3376705835797744e-06, "loss": 0.0041, "step": 54521 }, { "epoch": 16.78, "learning_rate": 1.3374216385060213e-06, "loss": 0.0035, "step": 54522 }, { "epoch": 16.78, "learning_rate": 1.3371727149391156e-06, "loss": 0.0077, "step": 54523 }, { "epoch": 16.78, "learning_rate": 1.3369238128796747e-06, "loss": 0.0058, "step": 54524 }, { "epoch": 16.78, "learning_rate": 1.3366749323283167e-06, "loss": 0.0072, "step": 54525 }, { "epoch": 16.78, "learning_rate": 1.3364260732856594e-06, "loss": 0.0031, "step": 54526 }, { "epoch": 16.78, "learning_rate": 1.3361772357523195e-06, "loss": 0.005, "step": 54527 }, { "epoch": 16.78, "learning_rate": 1.3359284197289168e-06, "loss": 0.0052, "step": 54528 }, { "epoch": 16.78, "learning_rate": 1.3356796252160676e-06, "loss": 0.0019, "step": 54529 }, { "epoch": 16.78, "learning_rate": 1.33543085221439e-06, "loss": 0.005, "step": 54530 }, { "epoch": 16.78, "learning_rate": 1.3351821007245003e-06, "loss": 0.0042, "step": 54531 }, { "epoch": 16.78, "learning_rate": 1.3349333707470192e-06, "loss": 0.0076, "step": 54532 }, { "epoch": 16.78, "learning_rate": 1.3346846622825649e-06, "loss": 0.0073, "step": 54533 }, { "epoch": 16.78, "learning_rate": 1.3344359753317494e-06, "loss": 0.0074, "step": 54534 }, { "epoch": 16.78, "learning_rate": 1.334187309895194e-06, "loss": 0.0035, "step": 54535 }, { "epoch": 16.78, "learning_rate": 1.3339386659735132e-06, "loss": 0.0093, "step": 54536 }, { "epoch": 16.78, "learning_rate": 1.3336900435673283e-06, "loss": 0.0035, "step": 54537 }, { "epoch": 16.78, "learning_rate": 1.3334414426772547e-06, "loss": 0.0047, "step": 54538 }, { "epoch": 16.78, "learning_rate": 1.3331928633039093e-06, "loss": 0.0033, "step": 54539 }, { "epoch": 16.78, "learning_rate": 1.332944305447912e-06, "loss": 0.0049, "step": 54540 }, { "epoch": 16.78, "learning_rate": 1.332695769109873e-06, "loss": 0.0088, "step": 54541 }, { "epoch": 16.78, "learning_rate": 1.3324472542904155e-06, "loss": 0.0084, "step": 54542 }, { "epoch": 16.78, "learning_rate": 1.3321987609901553e-06, "loss": 0.0072, "step": 54543 }, { "epoch": 16.78, "learning_rate": 1.3319502892097092e-06, "loss": 0.009, "step": 54544 }, { "epoch": 16.78, "learning_rate": 1.3317018389496927e-06, "loss": 0.0089, "step": 54545 }, { "epoch": 16.78, "learning_rate": 1.3314534102107247e-06, "loss": 0.0048, "step": 54546 }, { "epoch": 16.78, "learning_rate": 1.3312050029934197e-06, "loss": 0.0036, "step": 54547 }, { "epoch": 16.78, "learning_rate": 1.3309566172983967e-06, "loss": 0.0043, "step": 54548 }, { "epoch": 16.78, "learning_rate": 1.3307082531262706e-06, "loss": 0.0032, "step": 54549 }, { "epoch": 16.78, "learning_rate": 1.3304599104776595e-06, "loss": 0.0066, "step": 54550 }, { "epoch": 16.78, "learning_rate": 1.3302115893531763e-06, "loss": 0.004, "step": 54551 }, { "epoch": 16.78, "learning_rate": 1.3299632897534431e-06, "loss": 0.0047, "step": 54552 }, { "epoch": 16.79, "learning_rate": 1.3297150116790757e-06, "loss": 0.0045, "step": 54553 }, { "epoch": 16.79, "learning_rate": 1.3294667551306861e-06, "loss": 0.0054, "step": 54554 }, { "epoch": 16.79, "learning_rate": 1.3292185201088926e-06, "loss": 0.0047, "step": 54555 }, { "epoch": 16.79, "learning_rate": 1.3289703066143112e-06, "loss": 0.0051, "step": 54556 }, { "epoch": 16.79, "learning_rate": 1.3287221146475593e-06, "loss": 0.0075, "step": 54557 }, { "epoch": 16.79, "learning_rate": 1.3284739442092542e-06, "loss": 0.0074, "step": 54558 }, { "epoch": 16.79, "learning_rate": 1.3282257953000099e-06, "loss": 0.0066, "step": 54559 }, { "epoch": 16.79, "learning_rate": 1.3279776679204426e-06, "loss": 0.0061, "step": 54560 }, { "epoch": 16.79, "learning_rate": 1.3277295620711695e-06, "loss": 0.0052, "step": 54561 }, { "epoch": 16.79, "learning_rate": 1.3274814777528067e-06, "loss": 0.0037, "step": 54562 }, { "epoch": 16.79, "learning_rate": 1.3272334149659683e-06, "loss": 0.0106, "step": 54563 }, { "epoch": 16.79, "learning_rate": 1.3269853737112715e-06, "loss": 0.0071, "step": 54564 }, { "epoch": 16.79, "learning_rate": 1.3267373539893313e-06, "loss": 0.0032, "step": 54565 }, { "epoch": 16.79, "learning_rate": 1.3264893558007618e-06, "loss": 0.0062, "step": 54566 }, { "epoch": 16.79, "learning_rate": 1.326241379146186e-06, "loss": 0.0123, "step": 54567 }, { "epoch": 16.79, "learning_rate": 1.3259934240262118e-06, "loss": 0.0074, "step": 54568 }, { "epoch": 16.79, "learning_rate": 1.325745490441458e-06, "loss": 0.0136, "step": 54569 }, { "epoch": 16.79, "learning_rate": 1.3254975783925394e-06, "loss": 0.0044, "step": 54570 }, { "epoch": 16.79, "learning_rate": 1.3252496878800702e-06, "loss": 0.0037, "step": 54571 }, { "epoch": 16.79, "learning_rate": 1.3250018189046687e-06, "loss": 0.0022, "step": 54572 }, { "epoch": 16.79, "learning_rate": 1.324753971466949e-06, "loss": 0.0032, "step": 54573 }, { "epoch": 16.79, "learning_rate": 1.324506145567529e-06, "loss": 0.0041, "step": 54574 }, { "epoch": 16.79, "learning_rate": 1.3242583412070186e-06, "loss": 0.0051, "step": 54575 }, { "epoch": 16.79, "learning_rate": 1.3240105583860341e-06, "loss": 0.0036, "step": 54576 }, { "epoch": 16.79, "learning_rate": 1.3237627971051947e-06, "loss": 0.0049, "step": 54577 }, { "epoch": 16.79, "learning_rate": 1.3235150573651133e-06, "loss": 0.0034, "step": 54578 }, { "epoch": 16.79, "learning_rate": 1.3232673391664042e-06, "loss": 0.0072, "step": 54579 }, { "epoch": 16.79, "learning_rate": 1.3230196425096843e-06, "loss": 0.0051, "step": 54580 }, { "epoch": 16.79, "learning_rate": 1.3227719673955664e-06, "loss": 0.0054, "step": 54581 }, { "epoch": 16.79, "learning_rate": 1.3225243138246669e-06, "loss": 0.0048, "step": 54582 }, { "epoch": 16.79, "learning_rate": 1.3222766817975995e-06, "loss": 0.0042, "step": 54583 }, { "epoch": 16.79, "learning_rate": 1.3220290713149797e-06, "loss": 0.0075, "step": 54584 }, { "epoch": 16.8, "learning_rate": 1.321781482377421e-06, "loss": 0.0043, "step": 54585 }, { "epoch": 16.8, "learning_rate": 1.321533914985541e-06, "loss": 0.0056, "step": 54586 }, { "epoch": 16.8, "learning_rate": 1.3212863691399525e-06, "loss": 0.0053, "step": 54587 }, { "epoch": 16.8, "learning_rate": 1.3210388448412725e-06, "loss": 0.005, "step": 54588 }, { "epoch": 16.8, "learning_rate": 1.3207913420901108e-06, "loss": 0.0063, "step": 54589 }, { "epoch": 16.8, "learning_rate": 1.3205438608870823e-06, "loss": 0.0083, "step": 54590 }, { "epoch": 16.8, "learning_rate": 1.3202964012328056e-06, "loss": 0.0038, "step": 54591 }, { "epoch": 16.8, "learning_rate": 1.3200489631278934e-06, "loss": 0.0076, "step": 54592 }, { "epoch": 16.8, "learning_rate": 1.3198015465729597e-06, "loss": 0.0055, "step": 54593 }, { "epoch": 16.8, "learning_rate": 1.3195541515686184e-06, "loss": 0.0032, "step": 54594 }, { "epoch": 16.8, "learning_rate": 1.3193067781154835e-06, "loss": 0.0074, "step": 54595 }, { "epoch": 16.8, "learning_rate": 1.3190594262141697e-06, "loss": 0.0049, "step": 54596 }, { "epoch": 16.8, "learning_rate": 1.3188120958652917e-06, "loss": 0.0053, "step": 54597 }, { "epoch": 16.8, "learning_rate": 1.3185647870694629e-06, "loss": 0.0032, "step": 54598 }, { "epoch": 16.8, "learning_rate": 1.3183174998272962e-06, "loss": 0.0055, "step": 54599 }, { "epoch": 16.8, "learning_rate": 1.318070234139406e-06, "loss": 0.0043, "step": 54600 }, { "epoch": 16.8, "learning_rate": 1.31782299000641e-06, "loss": 0.004, "step": 54601 }, { "epoch": 16.8, "learning_rate": 1.317575767428917e-06, "loss": 0.0057, "step": 54602 }, { "epoch": 16.8, "learning_rate": 1.317328566407543e-06, "loss": 0.0053, "step": 54603 }, { "epoch": 16.8, "learning_rate": 1.317081386942901e-06, "loss": 0.0073, "step": 54604 }, { "epoch": 16.8, "learning_rate": 1.316834229035603e-06, "loss": 0.0071, "step": 54605 }, { "epoch": 16.8, "learning_rate": 1.3165870926862667e-06, "loss": 0.0037, "step": 54606 }, { "epoch": 16.8, "learning_rate": 1.3163399778955032e-06, "loss": 0.0076, "step": 54607 }, { "epoch": 16.8, "learning_rate": 1.3160928846639275e-06, "loss": 0.0053, "step": 54608 }, { "epoch": 16.8, "learning_rate": 1.3158458129921535e-06, "loss": 0.0057, "step": 54609 }, { "epoch": 16.8, "learning_rate": 1.3155987628807899e-06, "loss": 0.0063, "step": 54610 }, { "epoch": 16.8, "learning_rate": 1.3153517343304544e-06, "loss": 0.0054, "step": 54611 }, { "epoch": 16.8, "learning_rate": 1.3151047273417594e-06, "loss": 0.0072, "step": 54612 }, { "epoch": 16.8, "learning_rate": 1.3148577419153186e-06, "loss": 0.0067, "step": 54613 }, { "epoch": 16.8, "learning_rate": 1.3146107780517436e-06, "loss": 0.0119, "step": 54614 }, { "epoch": 16.8, "learning_rate": 1.3143638357516497e-06, "loss": 0.0037, "step": 54615 }, { "epoch": 16.8, "learning_rate": 1.3141169150156486e-06, "loss": 0.0117, "step": 54616 }, { "epoch": 16.8, "learning_rate": 1.313870015844353e-06, "loss": 0.0043, "step": 54617 }, { "epoch": 16.81, "learning_rate": 1.3136231382383768e-06, "loss": 0.006, "step": 54618 }, { "epoch": 16.81, "learning_rate": 1.313376282198332e-06, "loss": 0.0067, "step": 54619 }, { "epoch": 16.81, "learning_rate": 1.3131294477248302e-06, "loss": 0.0086, "step": 54620 }, { "epoch": 16.81, "learning_rate": 1.3128826348184886e-06, "loss": 0.0585, "step": 54621 }, { "epoch": 16.81, "learning_rate": 1.3126358434799192e-06, "loss": 0.0035, "step": 54622 }, { "epoch": 16.81, "learning_rate": 1.3123890737097312e-06, "loss": 0.0041, "step": 54623 }, { "epoch": 16.81, "learning_rate": 1.3121423255085387e-06, "loss": 0.0048, "step": 54624 }, { "epoch": 16.81, "learning_rate": 1.311895598876952e-06, "loss": 0.0055, "step": 54625 }, { "epoch": 16.81, "learning_rate": 1.311648893815589e-06, "loss": 0.0042, "step": 54626 }, { "epoch": 16.81, "learning_rate": 1.3114022103250601e-06, "loss": 0.0052, "step": 54627 }, { "epoch": 16.81, "learning_rate": 1.311155548405977e-06, "loss": 0.0067, "step": 54628 }, { "epoch": 16.81, "learning_rate": 1.3109089080589509e-06, "loss": 0.0015, "step": 54629 }, { "epoch": 16.81, "learning_rate": 1.3106622892845966e-06, "loss": 0.004, "step": 54630 }, { "epoch": 16.81, "learning_rate": 1.3104156920835254e-06, "loss": 0.0083, "step": 54631 }, { "epoch": 16.81, "learning_rate": 1.3101691164563491e-06, "loss": 0.0114, "step": 54632 }, { "epoch": 16.81, "learning_rate": 1.3099225624036792e-06, "loss": 0.0088, "step": 54633 }, { "epoch": 16.81, "learning_rate": 1.3096760299261303e-06, "loss": 0.0066, "step": 54634 }, { "epoch": 16.81, "learning_rate": 1.3094295190243123e-06, "loss": 0.0038, "step": 54635 }, { "epoch": 16.81, "learning_rate": 1.3091830296988384e-06, "loss": 0.0039, "step": 54636 }, { "epoch": 16.81, "learning_rate": 1.3089365619503191e-06, "loss": 0.0022, "step": 54637 }, { "epoch": 16.81, "learning_rate": 1.3086901157793674e-06, "loss": 0.0106, "step": 54638 }, { "epoch": 16.81, "learning_rate": 1.308443691186596e-06, "loss": 0.0037, "step": 54639 }, { "epoch": 16.81, "learning_rate": 1.3081972881726123e-06, "loss": 0.0044, "step": 54640 }, { "epoch": 16.81, "learning_rate": 1.3079509067380346e-06, "loss": 0.0051, "step": 54641 }, { "epoch": 16.81, "learning_rate": 1.3077045468834714e-06, "loss": 0.0072, "step": 54642 }, { "epoch": 16.81, "learning_rate": 1.3074582086095365e-06, "loss": 0.0056, "step": 54643 }, { "epoch": 16.81, "learning_rate": 1.3072118919168376e-06, "loss": 0.0056, "step": 54644 }, { "epoch": 16.81, "learning_rate": 1.3069655968059858e-06, "loss": 0.0042, "step": 54645 }, { "epoch": 16.81, "learning_rate": 1.3067193232775964e-06, "loss": 0.0053, "step": 54646 }, { "epoch": 16.81, "learning_rate": 1.3064730713322793e-06, "loss": 0.01, "step": 54647 }, { "epoch": 16.81, "learning_rate": 1.3062268409706469e-06, "loss": 0.0049, "step": 54648 }, { "epoch": 16.81, "learning_rate": 1.305980632193309e-06, "loss": 0.0064, "step": 54649 }, { "epoch": 16.82, "learning_rate": 1.3057344450008768e-06, "loss": 0.0033, "step": 54650 }, { "epoch": 16.82, "learning_rate": 1.3054882793939616e-06, "loss": 0.0055, "step": 54651 }, { "epoch": 16.82, "learning_rate": 1.305242135373176e-06, "loss": 0.0044, "step": 54652 }, { "epoch": 16.82, "learning_rate": 1.3049960129391303e-06, "loss": 0.0038, "step": 54653 }, { "epoch": 16.82, "learning_rate": 1.3047499120924356e-06, "loss": 0.0037, "step": 54654 }, { "epoch": 16.82, "learning_rate": 1.304503832833699e-06, "loss": 0.0064, "step": 54655 }, { "epoch": 16.82, "learning_rate": 1.304257775163541e-06, "loss": 0.0066, "step": 54656 }, { "epoch": 16.82, "learning_rate": 1.3040117390825635e-06, "loss": 0.0027, "step": 54657 }, { "epoch": 16.82, "learning_rate": 1.3037657245913816e-06, "loss": 0.0058, "step": 54658 }, { "epoch": 16.82, "learning_rate": 1.3035197316906046e-06, "loss": 0.0063, "step": 54659 }, { "epoch": 16.82, "learning_rate": 1.303273760380841e-06, "loss": 0.0041, "step": 54660 }, { "epoch": 16.82, "learning_rate": 1.3030278106627058e-06, "loss": 0.0104, "step": 54661 }, { "epoch": 16.82, "learning_rate": 1.3027818825368088e-06, "loss": 0.0025, "step": 54662 }, { "epoch": 16.82, "learning_rate": 1.3025359760037592e-06, "loss": 0.0048, "step": 54663 }, { "epoch": 16.82, "learning_rate": 1.3022900910641712e-06, "loss": 0.0047, "step": 54664 }, { "epoch": 16.82, "learning_rate": 1.3020442277186463e-06, "loss": 0.0048, "step": 54665 }, { "epoch": 16.82, "learning_rate": 1.3017983859678041e-06, "loss": 0.008, "step": 54666 }, { "epoch": 16.82, "learning_rate": 1.3015525658122507e-06, "loss": 0.0042, "step": 54667 }, { "epoch": 16.82, "learning_rate": 1.3013067672525982e-06, "loss": 0.0049, "step": 54668 }, { "epoch": 16.82, "learning_rate": 1.3010609902894545e-06, "loss": 0.0096, "step": 54669 }, { "epoch": 16.82, "learning_rate": 1.3008152349234316e-06, "loss": 0.0065, "step": 54670 }, { "epoch": 16.82, "learning_rate": 1.3005695011551401e-06, "loss": 0.0104, "step": 54671 }, { "epoch": 16.82, "learning_rate": 1.3003237889851882e-06, "loss": 0.0084, "step": 54672 }, { "epoch": 16.82, "learning_rate": 1.3000780984141881e-06, "loss": 0.0061, "step": 54673 }, { "epoch": 16.82, "learning_rate": 1.2998324294427478e-06, "loss": 0.0044, "step": 54674 }, { "epoch": 16.82, "learning_rate": 1.2995867820714757e-06, "loss": 0.0071, "step": 54675 }, { "epoch": 16.82, "learning_rate": 1.2993411563009872e-06, "loss": 0.006, "step": 54676 }, { "epoch": 16.82, "learning_rate": 1.2990955521318904e-06, "loss": 0.004, "step": 54677 }, { "epoch": 16.82, "learning_rate": 1.2988499695647917e-06, "loss": 0.007, "step": 54678 }, { "epoch": 16.82, "learning_rate": 1.2986044086003025e-06, "loss": 0.0037, "step": 54679 }, { "epoch": 16.82, "learning_rate": 1.2983588692390313e-06, "loss": 0.004, "step": 54680 }, { "epoch": 16.82, "learning_rate": 1.2981133514815903e-06, "loss": 0.0064, "step": 54681 }, { "epoch": 16.82, "learning_rate": 1.2978678553285883e-06, "loss": 0.0064, "step": 54682 }, { "epoch": 16.83, "learning_rate": 1.2976223807806355e-06, "loss": 0.0106, "step": 54683 }, { "epoch": 16.83, "learning_rate": 1.2973769278383385e-06, "loss": 0.0109, "step": 54684 }, { "epoch": 16.83, "learning_rate": 1.2971314965023095e-06, "loss": 0.0063, "step": 54685 }, { "epoch": 16.83, "learning_rate": 1.296886086773157e-06, "loss": 0.005, "step": 54686 }, { "epoch": 16.83, "learning_rate": 1.296640698651489e-06, "loss": 0.0057, "step": 54687 }, { "epoch": 16.83, "learning_rate": 1.2963953321379164e-06, "loss": 0.007, "step": 54688 }, { "epoch": 16.83, "learning_rate": 1.2961499872330475e-06, "loss": 0.0046, "step": 54689 }, { "epoch": 16.83, "learning_rate": 1.2959046639374917e-06, "loss": 0.0066, "step": 54690 }, { "epoch": 16.83, "learning_rate": 1.2956593622518588e-06, "loss": 0.0051, "step": 54691 }, { "epoch": 16.83, "learning_rate": 1.2954140821767558e-06, "loss": 0.0047, "step": 54692 }, { "epoch": 16.83, "learning_rate": 1.2951688237127945e-06, "loss": 0.0042, "step": 54693 }, { "epoch": 16.83, "learning_rate": 1.2949235868605813e-06, "loss": 0.0066, "step": 54694 }, { "epoch": 16.83, "learning_rate": 1.2946783716207234e-06, "loss": 0.0039, "step": 54695 }, { "epoch": 16.83, "learning_rate": 1.2944331779938356e-06, "loss": 0.011, "step": 54696 }, { "epoch": 16.83, "learning_rate": 1.2941880059805222e-06, "loss": 0.0047, "step": 54697 }, { "epoch": 16.83, "learning_rate": 1.293942855581396e-06, "loss": 0.0031, "step": 54698 }, { "epoch": 16.83, "learning_rate": 1.2936977267970597e-06, "loss": 0.0053, "step": 54699 }, { "epoch": 16.83, "learning_rate": 1.2934526196281227e-06, "loss": 0.0085, "step": 54700 }, { "epoch": 16.83, "learning_rate": 1.293207534075198e-06, "loss": 0.0035, "step": 54701 }, { "epoch": 16.83, "learning_rate": 1.2929624701388922e-06, "loss": 0.0047, "step": 54702 }, { "epoch": 16.83, "learning_rate": 1.2927174278198128e-06, "loss": 0.0094, "step": 54703 }, { "epoch": 16.83, "learning_rate": 1.292472407118569e-06, "loss": 0.0019, "step": 54704 }, { "epoch": 16.83, "learning_rate": 1.2922274080357677e-06, "loss": 0.0047, "step": 54705 }, { "epoch": 16.83, "learning_rate": 1.2919824305720196e-06, "loss": 0.0047, "step": 54706 }, { "epoch": 16.83, "learning_rate": 1.291737474727931e-06, "loss": 0.0079, "step": 54707 }, { "epoch": 16.83, "learning_rate": 1.29149254050411e-06, "loss": 0.0064, "step": 54708 }, { "epoch": 16.83, "learning_rate": 1.2912476279011654e-06, "loss": 0.0081, "step": 54709 }, { "epoch": 16.83, "learning_rate": 1.291002736919703e-06, "loss": 0.0042, "step": 54710 }, { "epoch": 16.83, "learning_rate": 1.2907578675603383e-06, "loss": 0.0045, "step": 54711 }, { "epoch": 16.83, "learning_rate": 1.2905130198236704e-06, "loss": 0.0043, "step": 54712 }, { "epoch": 16.83, "learning_rate": 1.2902681937103111e-06, "loss": 0.0059, "step": 54713 }, { "epoch": 16.83, "learning_rate": 1.290023389220868e-06, "loss": 0.004, "step": 54714 }, { "epoch": 16.84, "learning_rate": 1.289778606355946e-06, "loss": 0.0059, "step": 54715 }, { "epoch": 16.84, "learning_rate": 1.2895338451161587e-06, "loss": 0.0142, "step": 54716 }, { "epoch": 16.84, "learning_rate": 1.2892891055021095e-06, "loss": 0.0096, "step": 54717 }, { "epoch": 16.84, "learning_rate": 1.2890443875144076e-06, "loss": 0.0087, "step": 54718 }, { "epoch": 16.84, "learning_rate": 1.2887996911536627e-06, "loss": 0.0054, "step": 54719 }, { "epoch": 16.84, "learning_rate": 1.288555016420474e-06, "loss": 0.0057, "step": 54720 }, { "epoch": 16.84, "learning_rate": 1.2883103633154582e-06, "loss": 0.0024, "step": 54721 }, { "epoch": 16.84, "learning_rate": 1.2880657318392187e-06, "loss": 0.0155, "step": 54722 }, { "epoch": 16.84, "learning_rate": 1.287821121992363e-06, "loss": 0.0029, "step": 54723 }, { "epoch": 16.84, "learning_rate": 1.2875765337754998e-06, "loss": 0.0057, "step": 54724 }, { "epoch": 16.84, "learning_rate": 1.2873319671892337e-06, "loss": 0.0053, "step": 54725 }, { "epoch": 16.84, "learning_rate": 1.2870874222341746e-06, "loss": 0.0067, "step": 54726 }, { "epoch": 16.84, "learning_rate": 1.2868428989109283e-06, "loss": 0.0042, "step": 54727 }, { "epoch": 16.84, "learning_rate": 1.2865983972201025e-06, "loss": 0.0056, "step": 54728 }, { "epoch": 16.84, "learning_rate": 1.286353917162303e-06, "loss": 0.0047, "step": 54729 }, { "epoch": 16.84, "learning_rate": 1.286109458738135e-06, "loss": 0.0066, "step": 54730 }, { "epoch": 16.84, "learning_rate": 1.2858650219482116e-06, "loss": 0.0046, "step": 54731 }, { "epoch": 16.84, "learning_rate": 1.2856206067931375e-06, "loss": 0.006, "step": 54732 }, { "epoch": 16.84, "learning_rate": 1.2853762132735158e-06, "loss": 0.0051, "step": 54733 }, { "epoch": 16.84, "learning_rate": 1.2851318413899562e-06, "loss": 0.0051, "step": 54734 }, { "epoch": 16.84, "learning_rate": 1.284887491143062e-06, "loss": 0.0059, "step": 54735 }, { "epoch": 16.84, "learning_rate": 1.2846431625334444e-06, "loss": 0.0093, "step": 54736 }, { "epoch": 16.84, "learning_rate": 1.2843988555617094e-06, "loss": 0.0036, "step": 54737 }, { "epoch": 16.84, "learning_rate": 1.2841545702284618e-06, "loss": 0.0046, "step": 54738 }, { "epoch": 16.84, "learning_rate": 1.2839103065343084e-06, "loss": 0.0067, "step": 54739 }, { "epoch": 16.84, "learning_rate": 1.283666064479856e-06, "loss": 0.0047, "step": 54740 }, { "epoch": 16.84, "learning_rate": 1.2834218440657108e-06, "loss": 0.0026, "step": 54741 }, { "epoch": 16.84, "learning_rate": 1.2831776452924804e-06, "loss": 0.0049, "step": 54742 }, { "epoch": 16.84, "learning_rate": 1.2829334681607686e-06, "loss": 0.0023, "step": 54743 }, { "epoch": 16.84, "learning_rate": 1.2826893126711836e-06, "loss": 0.0069, "step": 54744 }, { "epoch": 16.84, "learning_rate": 1.2824451788243286e-06, "loss": 0.0035, "step": 54745 }, { "epoch": 16.84, "learning_rate": 1.2822010666208163e-06, "loss": 0.0047, "step": 54746 }, { "epoch": 16.84, "learning_rate": 1.281956976061246e-06, "loss": 0.0063, "step": 54747 }, { "epoch": 16.85, "learning_rate": 1.2817129071462276e-06, "loss": 0.0034, "step": 54748 }, { "epoch": 16.85, "learning_rate": 1.2814688598763648e-06, "loss": 0.0081, "step": 54749 }, { "epoch": 16.85, "learning_rate": 1.2812248342522614e-06, "loss": 0.0071, "step": 54750 }, { "epoch": 16.85, "learning_rate": 1.2809808302745298e-06, "loss": 0.0055, "step": 54751 }, { "epoch": 16.85, "learning_rate": 1.280736847943771e-06, "loss": 0.0087, "step": 54752 }, { "epoch": 16.85, "learning_rate": 1.2804928872605948e-06, "loss": 0.0051, "step": 54753 }, { "epoch": 16.85, "learning_rate": 1.280248948225602e-06, "loss": 0.0022, "step": 54754 }, { "epoch": 16.85, "learning_rate": 1.2800050308393974e-06, "loss": 0.0068, "step": 54755 }, { "epoch": 16.85, "learning_rate": 1.2797611351025917e-06, "loss": 0.003, "step": 54756 }, { "epoch": 16.85, "learning_rate": 1.2795172610157891e-06, "loss": 0.0074, "step": 54757 }, { "epoch": 16.85, "learning_rate": 1.2792734085795932e-06, "loss": 0.0046, "step": 54758 }, { "epoch": 16.85, "learning_rate": 1.2790295777946115e-06, "loss": 0.0061, "step": 54759 }, { "epoch": 16.85, "learning_rate": 1.278785768661448e-06, "loss": 0.0055, "step": 54760 }, { "epoch": 16.85, "learning_rate": 1.2785419811807075e-06, "loss": 0.0086, "step": 54761 }, { "epoch": 16.85, "learning_rate": 1.2782982153529966e-06, "loss": 0.0044, "step": 54762 }, { "epoch": 16.85, "learning_rate": 1.27805447117892e-06, "loss": 0.0049, "step": 54763 }, { "epoch": 16.85, "learning_rate": 1.277810748659082e-06, "loss": 0.0079, "step": 54764 }, { "epoch": 16.85, "learning_rate": 1.277567047794087e-06, "loss": 0.0074, "step": 54765 }, { "epoch": 16.85, "learning_rate": 1.2773233685845433e-06, "loss": 0.0047, "step": 54766 }, { "epoch": 16.85, "learning_rate": 1.2770797110310573e-06, "loss": 0.0035, "step": 54767 }, { "epoch": 16.85, "learning_rate": 1.276836075134228e-06, "loss": 0.0038, "step": 54768 }, { "epoch": 16.85, "learning_rate": 1.2765924608946622e-06, "loss": 0.0061, "step": 54769 }, { "epoch": 16.85, "learning_rate": 1.2763488683129643e-06, "loss": 0.0089, "step": 54770 }, { "epoch": 16.85, "learning_rate": 1.2761052973897425e-06, "loss": 0.0101, "step": 54771 }, { "epoch": 16.85, "learning_rate": 1.2758617481255986e-06, "loss": 0.006, "step": 54772 }, { "epoch": 16.85, "learning_rate": 1.2756182205211386e-06, "loss": 0.0034, "step": 54773 }, { "epoch": 16.85, "learning_rate": 1.275374714576969e-06, "loss": 0.0047, "step": 54774 }, { "epoch": 16.85, "learning_rate": 1.275131230293687e-06, "loss": 0.0049, "step": 54775 }, { "epoch": 16.85, "learning_rate": 1.274887767671904e-06, "loss": 0.0031, "step": 54776 }, { "epoch": 16.85, "learning_rate": 1.2746443267122233e-06, "loss": 0.0057, "step": 54777 }, { "epoch": 16.85, "learning_rate": 1.2744009074152476e-06, "loss": 0.0034, "step": 54778 }, { "epoch": 16.85, "learning_rate": 1.2741575097815829e-06, "loss": 0.0074, "step": 54779 }, { "epoch": 16.86, "learning_rate": 1.273914133811832e-06, "loss": 0.004, "step": 54780 }, { "epoch": 16.86, "learning_rate": 1.2736707795066006e-06, "loss": 0.0084, "step": 54781 }, { "epoch": 16.86, "learning_rate": 1.273427446866491e-06, "loss": 0.0048, "step": 54782 }, { "epoch": 16.86, "learning_rate": 1.2731841358921094e-06, "loss": 0.0045, "step": 54783 }, { "epoch": 16.86, "learning_rate": 1.2729408465840587e-06, "loss": 0.0039, "step": 54784 }, { "epoch": 16.86, "learning_rate": 1.2726975789429408e-06, "loss": 0.0049, "step": 54785 }, { "epoch": 16.86, "learning_rate": 1.272454332969364e-06, "loss": 0.0048, "step": 54786 }, { "epoch": 16.86, "learning_rate": 1.27221110866393e-06, "loss": 0.006, "step": 54787 }, { "epoch": 16.86, "learning_rate": 1.271967906027246e-06, "loss": 0.0075, "step": 54788 }, { "epoch": 16.86, "learning_rate": 1.2717247250599096e-06, "loss": 0.0044, "step": 54789 }, { "epoch": 16.86, "learning_rate": 1.2714815657625257e-06, "loss": 0.0046, "step": 54790 }, { "epoch": 16.86, "learning_rate": 1.2712384281357026e-06, "loss": 0.006, "step": 54791 }, { "epoch": 16.86, "learning_rate": 1.2709953121800412e-06, "loss": 0.0037, "step": 54792 }, { "epoch": 16.86, "learning_rate": 1.270752217896144e-06, "loss": 0.0036, "step": 54793 }, { "epoch": 16.86, "learning_rate": 1.2705091452846175e-06, "loss": 0.0061, "step": 54794 }, { "epoch": 16.86, "learning_rate": 1.270266094346062e-06, "loss": 0.0038, "step": 54795 }, { "epoch": 16.86, "learning_rate": 1.270023065081083e-06, "loss": 0.0097, "step": 54796 }, { "epoch": 16.86, "learning_rate": 1.269780057490283e-06, "loss": 0.0033, "step": 54797 }, { "epoch": 16.86, "learning_rate": 1.269537071574265e-06, "loss": 0.0373, "step": 54798 }, { "epoch": 16.86, "learning_rate": 1.269294107333634e-06, "loss": 0.0033, "step": 54799 }, { "epoch": 16.86, "learning_rate": 1.2690511647689885e-06, "loss": 0.0105, "step": 54800 }, { "epoch": 16.86, "learning_rate": 1.2688082438809413e-06, "loss": 0.0065, "step": 54801 }, { "epoch": 16.86, "learning_rate": 1.268565344670085e-06, "loss": 0.0052, "step": 54802 }, { "epoch": 16.86, "learning_rate": 1.2683224671370286e-06, "loss": 0.0061, "step": 54803 }, { "epoch": 16.86, "learning_rate": 1.2680796112823735e-06, "loss": 0.0058, "step": 54804 }, { "epoch": 16.86, "learning_rate": 1.26783677710672e-06, "loss": 0.0036, "step": 54805 }, { "epoch": 16.86, "learning_rate": 1.267593964610675e-06, "loss": 0.0117, "step": 54806 }, { "epoch": 16.86, "learning_rate": 1.267351173794842e-06, "loss": 0.0068, "step": 54807 }, { "epoch": 16.86, "learning_rate": 1.2671084046598204e-06, "loss": 0.0058, "step": 54808 }, { "epoch": 16.86, "learning_rate": 1.2668656572062167e-06, "loss": 0.0034, "step": 54809 }, { "epoch": 16.86, "learning_rate": 1.2666229314346267e-06, "loss": 0.0064, "step": 54810 }, { "epoch": 16.86, "learning_rate": 1.2663802273456593e-06, "loss": 0.0056, "step": 54811 }, { "epoch": 16.86, "learning_rate": 1.2661375449399161e-06, "loss": 0.0065, "step": 54812 }, { "epoch": 16.87, "learning_rate": 1.2658948842179986e-06, "loss": 0.0062, "step": 54813 }, { "epoch": 16.87, "learning_rate": 1.2656522451805098e-06, "loss": 0.0066, "step": 54814 }, { "epoch": 16.87, "learning_rate": 1.2654096278280514e-06, "loss": 0.0041, "step": 54815 }, { "epoch": 16.87, "learning_rate": 1.2651670321612264e-06, "loss": 0.0064, "step": 54816 }, { "epoch": 16.87, "learning_rate": 1.2649244581806363e-06, "loss": 0.0096, "step": 54817 }, { "epoch": 16.87, "learning_rate": 1.2646819058868842e-06, "loss": 0.0041, "step": 54818 }, { "epoch": 16.87, "learning_rate": 1.2644393752805717e-06, "loss": 0.0039, "step": 54819 }, { "epoch": 16.87, "learning_rate": 1.2641968663622995e-06, "loss": 0.0047, "step": 54820 }, { "epoch": 16.87, "learning_rate": 1.2639543791326726e-06, "loss": 0.0037, "step": 54821 }, { "epoch": 16.87, "learning_rate": 1.2637119135922948e-06, "loss": 0.0077, "step": 54822 }, { "epoch": 16.87, "learning_rate": 1.2634694697417626e-06, "loss": 0.0048, "step": 54823 }, { "epoch": 16.87, "learning_rate": 1.263227047581681e-06, "loss": 0.0047, "step": 54824 }, { "epoch": 16.87, "learning_rate": 1.2629846471126483e-06, "loss": 0.0112, "step": 54825 }, { "epoch": 16.87, "learning_rate": 1.262742268335272e-06, "loss": 0.0058, "step": 54826 }, { "epoch": 16.87, "learning_rate": 1.2624999112501513e-06, "loss": 0.0053, "step": 54827 }, { "epoch": 16.87, "learning_rate": 1.262257575857887e-06, "loss": 0.0064, "step": 54828 }, { "epoch": 16.87, "learning_rate": 1.262015262159082e-06, "loss": 0.0041, "step": 54829 }, { "epoch": 16.87, "learning_rate": 1.261772970154338e-06, "loss": 0.0102, "step": 54830 }, { "epoch": 16.87, "learning_rate": 1.2615306998442545e-06, "loss": 0.0051, "step": 54831 }, { "epoch": 16.87, "learning_rate": 1.2612884512294354e-06, "loss": 0.004, "step": 54832 }, { "epoch": 16.87, "learning_rate": 1.2610462243104815e-06, "loss": 0.004, "step": 54833 }, { "epoch": 16.87, "learning_rate": 1.2608040190879933e-06, "loss": 0.0047, "step": 54834 }, { "epoch": 16.87, "learning_rate": 1.2605618355625726e-06, "loss": 0.0053, "step": 54835 }, { "epoch": 16.87, "learning_rate": 1.2603196737348211e-06, "loss": 0.0036, "step": 54836 }, { "epoch": 16.87, "learning_rate": 1.2600775336053405e-06, "loss": 0.0057, "step": 54837 }, { "epoch": 16.87, "learning_rate": 1.2598354151747294e-06, "loss": 0.0068, "step": 54838 }, { "epoch": 16.87, "learning_rate": 1.2595933184435915e-06, "loss": 0.0046, "step": 54839 }, { "epoch": 16.87, "learning_rate": 1.2593512434125254e-06, "loss": 0.0028, "step": 54840 }, { "epoch": 16.87, "learning_rate": 1.2591091900821362e-06, "loss": 0.0023, "step": 54841 }, { "epoch": 16.87, "learning_rate": 1.258867158453021e-06, "loss": 0.0038, "step": 54842 }, { "epoch": 16.87, "learning_rate": 1.258625148525785e-06, "loss": 0.0048, "step": 54843 }, { "epoch": 16.87, "learning_rate": 1.2583831603010243e-06, "loss": 0.003, "step": 54844 }, { "epoch": 16.88, "learning_rate": 1.2581411937793386e-06, "loss": 0.0087, "step": 54845 }, { "epoch": 16.88, "learning_rate": 1.257899248961335e-06, "loss": 0.0063, "step": 54846 }, { "epoch": 16.88, "learning_rate": 1.25765732584761e-06, "loss": 0.0065, "step": 54847 }, { "epoch": 16.88, "learning_rate": 1.2574154244387648e-06, "loss": 0.0054, "step": 54848 }, { "epoch": 16.88, "learning_rate": 1.2571735447354005e-06, "loss": 0.0041, "step": 54849 }, { "epoch": 16.88, "learning_rate": 1.2569316867381165e-06, "loss": 0.0049, "step": 54850 }, { "epoch": 16.88, "learning_rate": 1.2566898504475157e-06, "loss": 0.0065, "step": 54851 }, { "epoch": 16.88, "learning_rate": 1.2564480358641951e-06, "loss": 0.008, "step": 54852 }, { "epoch": 16.88, "learning_rate": 1.2562062429887578e-06, "loss": 0.0053, "step": 54853 }, { "epoch": 16.88, "learning_rate": 1.2559644718218033e-06, "loss": 0.0044, "step": 54854 }, { "epoch": 16.88, "learning_rate": 1.255722722363929e-06, "loss": 0.0042, "step": 54855 }, { "epoch": 16.88, "learning_rate": 1.255480994615742e-06, "loss": 0.0043, "step": 54856 }, { "epoch": 16.88, "learning_rate": 1.2552392885778364e-06, "loss": 0.0098, "step": 54857 }, { "epoch": 16.88, "learning_rate": 1.254997604250815e-06, "loss": 0.0054, "step": 54858 }, { "epoch": 16.88, "learning_rate": 1.2547559416352762e-06, "loss": 0.0065, "step": 54859 }, { "epoch": 16.88, "learning_rate": 1.2545143007318183e-06, "loss": 0.0021, "step": 54860 }, { "epoch": 16.88, "learning_rate": 1.2542726815410456e-06, "loss": 0.0043, "step": 54861 }, { "epoch": 16.88, "learning_rate": 1.2540310840635572e-06, "loss": 0.0024, "step": 54862 }, { "epoch": 16.88, "learning_rate": 1.2537895082999507e-06, "loss": 0.0065, "step": 54863 }, { "epoch": 16.88, "learning_rate": 1.25354795425083e-06, "loss": 0.0074, "step": 54864 }, { "epoch": 16.88, "learning_rate": 1.2533064219167868e-06, "loss": 0.0041, "step": 54865 }, { "epoch": 16.88, "learning_rate": 1.2530649112984273e-06, "loss": 0.0036, "step": 54866 }, { "epoch": 16.88, "learning_rate": 1.252823422396351e-06, "loss": 0.0053, "step": 54867 }, { "epoch": 16.88, "learning_rate": 1.252581955211155e-06, "loss": 0.0062, "step": 54868 }, { "epoch": 16.88, "learning_rate": 1.2523405097434393e-06, "loss": 0.0036, "step": 54869 }, { "epoch": 16.88, "learning_rate": 1.2520990859938053e-06, "loss": 0.0027, "step": 54870 }, { "epoch": 16.88, "learning_rate": 1.2518576839628494e-06, "loss": 0.0045, "step": 54871 }, { "epoch": 16.88, "learning_rate": 1.251616303651173e-06, "loss": 0.0123, "step": 54872 }, { "epoch": 16.88, "learning_rate": 1.251374945059376e-06, "loss": 0.006, "step": 54873 }, { "epoch": 16.88, "learning_rate": 1.2511336081880554e-06, "loss": 0.0042, "step": 54874 }, { "epoch": 16.88, "learning_rate": 1.2508922930378086e-06, "loss": 0.0036, "step": 54875 }, { "epoch": 16.88, "learning_rate": 1.250650999609241e-06, "loss": 0.0067, "step": 54876 }, { "epoch": 16.88, "learning_rate": 1.2504097279029491e-06, "loss": 0.0063, "step": 54877 }, { "epoch": 16.89, "learning_rate": 1.25016847791953e-06, "loss": 0.007, "step": 54878 }, { "epoch": 16.89, "learning_rate": 1.2499272496595826e-06, "loss": 0.0088, "step": 54879 }, { "epoch": 16.89, "learning_rate": 1.2496860431237057e-06, "loss": 0.0027, "step": 54880 }, { "epoch": 16.89, "learning_rate": 1.249444858312502e-06, "loss": 0.0039, "step": 54881 }, { "epoch": 16.89, "learning_rate": 1.2492036952265663e-06, "loss": 0.0128, "step": 54882 }, { "epoch": 16.89, "learning_rate": 1.2489625538664995e-06, "loss": 0.0066, "step": 54883 }, { "epoch": 16.89, "learning_rate": 1.248721434232899e-06, "loss": 0.0097, "step": 54884 }, { "epoch": 16.89, "learning_rate": 1.2484803363263641e-06, "loss": 0.0077, "step": 54885 }, { "epoch": 16.89, "learning_rate": 1.2482392601474935e-06, "loss": 0.004, "step": 54886 }, { "epoch": 16.89, "learning_rate": 1.2479982056968842e-06, "loss": 0.0064, "step": 54887 }, { "epoch": 16.89, "learning_rate": 1.247757172975137e-06, "loss": 0.0116, "step": 54888 }, { "epoch": 16.89, "learning_rate": 1.2475161619828492e-06, "loss": 0.0044, "step": 54889 }, { "epoch": 16.89, "learning_rate": 1.2472751727206188e-06, "loss": 0.003, "step": 54890 }, { "epoch": 16.89, "learning_rate": 1.247034205189045e-06, "loss": 0.0054, "step": 54891 }, { "epoch": 16.89, "learning_rate": 1.2467932593887255e-06, "loss": 0.0027, "step": 54892 }, { "epoch": 16.89, "learning_rate": 1.246552335320258e-06, "loss": 0.0022, "step": 54893 }, { "epoch": 16.89, "learning_rate": 1.2463114329842406e-06, "loss": 0.0113, "step": 54894 }, { "epoch": 16.89, "learning_rate": 1.246070552381271e-06, "loss": 0.0062, "step": 54895 }, { "epoch": 16.89, "learning_rate": 1.2458296935119508e-06, "loss": 0.0029, "step": 54896 }, { "epoch": 16.89, "learning_rate": 1.2455888563768748e-06, "loss": 0.0048, "step": 54897 }, { "epoch": 16.89, "learning_rate": 1.2453480409766427e-06, "loss": 0.0039, "step": 54898 }, { "epoch": 16.89, "learning_rate": 1.245107247311851e-06, "loss": 0.0044, "step": 54899 }, { "epoch": 16.89, "learning_rate": 1.2448664753830941e-06, "loss": 0.0032, "step": 54900 }, { "epoch": 16.89, "learning_rate": 1.2446257251909754e-06, "loss": 0.0031, "step": 54901 }, { "epoch": 16.89, "learning_rate": 1.244384996736092e-06, "loss": 0.0039, "step": 54902 }, { "epoch": 16.89, "learning_rate": 1.2441442900190392e-06, "loss": 0.0057, "step": 54903 }, { "epoch": 16.89, "learning_rate": 1.2439036050404163e-06, "loss": 0.0041, "step": 54904 }, { "epoch": 16.89, "learning_rate": 1.2436629418008194e-06, "loss": 0.004, "step": 54905 }, { "epoch": 16.89, "learning_rate": 1.2434223003008473e-06, "loss": 0.0088, "step": 54906 }, { "epoch": 16.89, "learning_rate": 1.2431816805410968e-06, "loss": 0.0044, "step": 54907 }, { "epoch": 16.89, "learning_rate": 1.2429410825221655e-06, "loss": 0.0045, "step": 54908 }, { "epoch": 16.89, "learning_rate": 1.2427005062446508e-06, "loss": 0.0032, "step": 54909 }, { "epoch": 16.9, "learning_rate": 1.2424599517091473e-06, "loss": 0.0048, "step": 54910 }, { "epoch": 16.9, "learning_rate": 1.2422194189162595e-06, "loss": 0.004, "step": 54911 }, { "epoch": 16.9, "learning_rate": 1.2419789078665767e-06, "loss": 0.0065, "step": 54912 }, { "epoch": 16.9, "learning_rate": 1.2417384185607006e-06, "loss": 0.0082, "step": 54913 }, { "epoch": 16.9, "learning_rate": 1.2414979509992264e-06, "loss": 0.005, "step": 54914 }, { "epoch": 16.9, "learning_rate": 1.241257505182749e-06, "loss": 0.0041, "step": 54915 }, { "epoch": 16.9, "learning_rate": 1.2410170811118705e-06, "loss": 0.0018, "step": 54916 }, { "epoch": 16.9, "learning_rate": 1.2407766787871845e-06, "loss": 0.0047, "step": 54917 }, { "epoch": 16.9, "learning_rate": 1.2405362982092895e-06, "loss": 0.0073, "step": 54918 }, { "epoch": 16.9, "learning_rate": 1.2402959393787828e-06, "loss": 0.0049, "step": 54919 }, { "epoch": 16.9, "learning_rate": 1.2400556022962562e-06, "loss": 0.0083, "step": 54920 }, { "epoch": 16.9, "learning_rate": 1.2398152869623115e-06, "loss": 0.0034, "step": 54921 }, { "epoch": 16.9, "learning_rate": 1.2395749933775448e-06, "loss": 0.0062, "step": 54922 }, { "epoch": 16.9, "learning_rate": 1.23933472154255e-06, "loss": 0.0038, "step": 54923 }, { "epoch": 16.9, "learning_rate": 1.2390944714579266e-06, "loss": 0.0065, "step": 54924 }, { "epoch": 16.9, "learning_rate": 1.2388542431242689e-06, "loss": 0.0046, "step": 54925 }, { "epoch": 16.9, "learning_rate": 1.238614036542175e-06, "loss": 0.0053, "step": 54926 }, { "epoch": 16.9, "learning_rate": 1.2383738517122401e-06, "loss": 0.0037, "step": 54927 }, { "epoch": 16.9, "learning_rate": 1.2381336886350603e-06, "loss": 0.0085, "step": 54928 }, { "epoch": 16.9, "learning_rate": 1.237893547311233e-06, "loss": 0.0046, "step": 54929 }, { "epoch": 16.9, "learning_rate": 1.2376534277413522e-06, "loss": 0.0072, "step": 54930 }, { "epoch": 16.9, "learning_rate": 1.2374133299260173e-06, "loss": 0.0054, "step": 54931 }, { "epoch": 16.9, "learning_rate": 1.2371732538658242e-06, "loss": 0.0048, "step": 54932 }, { "epoch": 16.9, "learning_rate": 1.2369331995613664e-06, "loss": 0.0038, "step": 54933 }, { "epoch": 16.9, "learning_rate": 1.2366931670132398e-06, "loss": 0.0084, "step": 54934 }, { "epoch": 16.9, "learning_rate": 1.2364531562220406e-06, "loss": 0.0046, "step": 54935 }, { "epoch": 16.9, "learning_rate": 1.2362131671883671e-06, "loss": 0.0045, "step": 54936 }, { "epoch": 16.9, "learning_rate": 1.2359731999128144e-06, "loss": 0.0038, "step": 54937 }, { "epoch": 16.9, "learning_rate": 1.2357332543959778e-06, "loss": 0.0042, "step": 54938 }, { "epoch": 16.9, "learning_rate": 1.2354933306384509e-06, "loss": 0.0079, "step": 54939 }, { "epoch": 16.9, "learning_rate": 1.2352534286408324e-06, "loss": 0.0072, "step": 54940 }, { "epoch": 16.9, "learning_rate": 1.2350135484037173e-06, "loss": 0.0049, "step": 54941 }, { "epoch": 16.9, "learning_rate": 1.2347736899276997e-06, "loss": 0.0063, "step": 54942 }, { "epoch": 16.91, "learning_rate": 1.2345338532133766e-06, "loss": 0.0037, "step": 54943 }, { "epoch": 16.91, "learning_rate": 1.2342940382613422e-06, "loss": 0.0067, "step": 54944 }, { "epoch": 16.91, "learning_rate": 1.2340542450721915e-06, "loss": 0.0044, "step": 54945 }, { "epoch": 16.91, "learning_rate": 1.233814473646524e-06, "loss": 0.011, "step": 54946 }, { "epoch": 16.91, "learning_rate": 1.2335747239849306e-06, "loss": 0.004, "step": 54947 }, { "epoch": 16.91, "learning_rate": 1.233334996088007e-06, "loss": 0.0062, "step": 54948 }, { "epoch": 16.91, "learning_rate": 1.2330952899563498e-06, "loss": 0.0031, "step": 54949 }, { "epoch": 16.91, "learning_rate": 1.2328556055905506e-06, "loss": 0.0134, "step": 54950 }, { "epoch": 16.91, "learning_rate": 1.23261594299121e-06, "loss": 0.0044, "step": 54951 }, { "epoch": 16.91, "learning_rate": 1.2323763021589208e-06, "loss": 0.0078, "step": 54952 }, { "epoch": 16.91, "learning_rate": 1.2321366830942794e-06, "loss": 0.0022, "step": 54953 }, { "epoch": 16.91, "learning_rate": 1.2318970857978762e-06, "loss": 0.004, "step": 54954 }, { "epoch": 16.91, "learning_rate": 1.2316575102703077e-06, "loss": 0.0041, "step": 54955 }, { "epoch": 16.91, "learning_rate": 1.2314179565121709e-06, "loss": 0.0039, "step": 54956 }, { "epoch": 16.91, "learning_rate": 1.23117842452406e-06, "loss": 0.0057, "step": 54957 }, { "epoch": 16.91, "learning_rate": 1.2309389143065686e-06, "loss": 0.0081, "step": 54958 }, { "epoch": 16.91, "learning_rate": 1.2306994258602922e-06, "loss": 0.0068, "step": 54959 }, { "epoch": 16.91, "learning_rate": 1.2304599591858247e-06, "loss": 0.0052, "step": 54960 }, { "epoch": 16.91, "learning_rate": 1.2302205142837609e-06, "loss": 0.0051, "step": 54961 }, { "epoch": 16.91, "learning_rate": 1.2299810911546962e-06, "loss": 0.0034, "step": 54962 }, { "epoch": 16.91, "learning_rate": 1.2297416897992232e-06, "loss": 0.0104, "step": 54963 }, { "epoch": 16.91, "learning_rate": 1.229502310217937e-06, "loss": 0.006, "step": 54964 }, { "epoch": 16.91, "learning_rate": 1.2292629524114296e-06, "loss": 0.0089, "step": 54965 }, { "epoch": 16.91, "learning_rate": 1.2290236163803016e-06, "loss": 0.008, "step": 54966 }, { "epoch": 16.91, "learning_rate": 1.2287843021251445e-06, "loss": 0.0053, "step": 54967 }, { "epoch": 16.91, "learning_rate": 1.228545009646549e-06, "loss": 0.0036, "step": 54968 }, { "epoch": 16.91, "learning_rate": 1.2283057389451113e-06, "loss": 0.005, "step": 54969 }, { "epoch": 16.91, "learning_rate": 1.2280664900214246e-06, "loss": 0.004, "step": 54970 }, { "epoch": 16.91, "learning_rate": 1.2278272628760857e-06, "loss": 0.0082, "step": 54971 }, { "epoch": 16.91, "learning_rate": 1.2275880575096865e-06, "loss": 0.0067, "step": 54972 }, { "epoch": 16.91, "learning_rate": 1.2273488739228213e-06, "loss": 0.0035, "step": 54973 }, { "epoch": 16.91, "learning_rate": 1.2271097121160835e-06, "loss": 0.0032, "step": 54974 }, { "epoch": 16.92, "learning_rate": 1.2268705720900676e-06, "loss": 0.0033, "step": 54975 }, { "epoch": 16.92, "learning_rate": 1.2266314538453671e-06, "loss": 0.0102, "step": 54976 }, { "epoch": 16.92, "learning_rate": 1.2263923573825753e-06, "loss": 0.0045, "step": 54977 }, { "epoch": 16.92, "learning_rate": 1.226153282702286e-06, "loss": 0.0107, "step": 54978 }, { "epoch": 16.92, "learning_rate": 1.2259142298050896e-06, "loss": 0.0063, "step": 54979 }, { "epoch": 16.92, "learning_rate": 1.2256751986915882e-06, "loss": 0.0038, "step": 54980 }, { "epoch": 16.92, "learning_rate": 1.225436189362368e-06, "loss": 0.0047, "step": 54981 }, { "epoch": 16.92, "learning_rate": 1.2251972018180225e-06, "loss": 0.0116, "step": 54982 }, { "epoch": 16.92, "learning_rate": 1.2249582360591483e-06, "loss": 0.0054, "step": 54983 }, { "epoch": 16.92, "learning_rate": 1.2247192920863338e-06, "loss": 0.0097, "step": 54984 }, { "epoch": 16.92, "learning_rate": 1.2244803699001785e-06, "loss": 0.0058, "step": 54985 }, { "epoch": 16.92, "learning_rate": 1.2242414695012727e-06, "loss": 0.0063, "step": 54986 }, { "epoch": 16.92, "learning_rate": 1.2240025908902097e-06, "loss": 0.0033, "step": 54987 }, { "epoch": 16.92, "learning_rate": 1.2237637340675835e-06, "loss": 0.0095, "step": 54988 }, { "epoch": 16.92, "learning_rate": 1.2235248990339821e-06, "loss": 0.0028, "step": 54989 }, { "epoch": 16.92, "learning_rate": 1.2232860857900052e-06, "loss": 0.0088, "step": 54990 }, { "epoch": 16.92, "learning_rate": 1.2230472943362414e-06, "loss": 0.0102, "step": 54991 }, { "epoch": 16.92, "learning_rate": 1.2228085246732857e-06, "loss": 0.007, "step": 54992 }, { "epoch": 16.92, "learning_rate": 1.2225697768017309e-06, "loss": 0.0103, "step": 54993 }, { "epoch": 16.92, "learning_rate": 1.2223310507221675e-06, "loss": 0.0054, "step": 54994 }, { "epoch": 16.92, "learning_rate": 1.2220923464351908e-06, "loss": 0.0065, "step": 54995 }, { "epoch": 16.92, "learning_rate": 1.2218536639413925e-06, "loss": 0.0087, "step": 54996 }, { "epoch": 16.92, "learning_rate": 1.2216150032413643e-06, "loss": 0.0098, "step": 54997 }, { "epoch": 16.92, "learning_rate": 1.2213763643357002e-06, "loss": 0.0104, "step": 54998 }, { "epoch": 16.92, "learning_rate": 1.2211377472249896e-06, "loss": 0.0049, "step": 54999 }, { "epoch": 16.92, "learning_rate": 1.220899151909829e-06, "loss": 0.0034, "step": 55000 }, { "epoch": 16.92, "learning_rate": 1.220660578390811e-06, "loss": 0.0061, "step": 55001 }, { "epoch": 16.92, "learning_rate": 1.220422026668524e-06, "loss": 0.0042, "step": 55002 }, { "epoch": 16.92, "learning_rate": 1.2201834967435633e-06, "loss": 0.0029, "step": 55003 }, { "epoch": 16.92, "learning_rate": 1.219944988616516e-06, "loss": 0.0036, "step": 55004 }, { "epoch": 16.92, "learning_rate": 1.2197065022879817e-06, "loss": 0.004, "step": 55005 }, { "epoch": 16.92, "learning_rate": 1.2194680377585488e-06, "loss": 0.0037, "step": 55006 }, { "epoch": 16.92, "learning_rate": 1.219229595028809e-06, "loss": 0.0081, "step": 55007 }, { "epoch": 16.93, "learning_rate": 1.2189911740993566e-06, "loss": 0.0056, "step": 55008 }, { "epoch": 16.93, "learning_rate": 1.2187527749707805e-06, "loss": 0.0068, "step": 55009 }, { "epoch": 16.93, "learning_rate": 1.2185143976436742e-06, "loss": 0.0209, "step": 55010 }, { "epoch": 16.93, "learning_rate": 1.218276042118629e-06, "loss": 0.0055, "step": 55011 }, { "epoch": 16.93, "learning_rate": 1.218037708396237e-06, "loss": 0.0063, "step": 55012 }, { "epoch": 16.93, "learning_rate": 1.2177993964770896e-06, "loss": 0.0049, "step": 55013 }, { "epoch": 16.93, "learning_rate": 1.2175611063617787e-06, "loss": 0.0074, "step": 55014 }, { "epoch": 16.93, "learning_rate": 1.2173228380508961e-06, "loss": 0.007, "step": 55015 }, { "epoch": 16.93, "learning_rate": 1.2170845915450336e-06, "loss": 0.0058, "step": 55016 }, { "epoch": 16.93, "learning_rate": 1.2168463668447817e-06, "loss": 0.0077, "step": 55017 }, { "epoch": 16.93, "learning_rate": 1.216608163950732e-06, "loss": 0.0036, "step": 55018 }, { "epoch": 16.93, "learning_rate": 1.2163699828634746e-06, "loss": 0.0074, "step": 55019 }, { "epoch": 16.93, "learning_rate": 1.216131823583605e-06, "loss": 0.0052, "step": 55020 }, { "epoch": 16.93, "learning_rate": 1.216131823583605e-06, "loss": 0.0243, "step": 55021 }, { "epoch": 16.93, "learning_rate": 1.215893686111712e-06, "loss": 0.0036, "step": 55022 }, { "epoch": 16.93, "learning_rate": 1.2156555704483886e-06, "loss": 0.0022, "step": 55023 }, { "epoch": 16.93, "learning_rate": 1.2154174765942227e-06, "loss": 0.0049, "step": 55024 }, { "epoch": 16.93, "learning_rate": 1.2151794045498045e-06, "loss": 0.0019, "step": 55025 }, { "epoch": 16.93, "learning_rate": 1.2149413543157295e-06, "loss": 0.0045, "step": 55026 }, { "epoch": 16.93, "learning_rate": 1.2147033258925867e-06, "loss": 0.0066, "step": 55027 }, { "epoch": 16.93, "learning_rate": 1.2144653192809675e-06, "loss": 0.0075, "step": 55028 }, { "epoch": 16.93, "learning_rate": 1.2142273344814614e-06, "loss": 0.007, "step": 55029 }, { "epoch": 16.93, "learning_rate": 1.2139893714946616e-06, "loss": 0.0029, "step": 55030 }, { "epoch": 16.93, "learning_rate": 1.213751430321156e-06, "loss": 0.0055, "step": 55031 }, { "epoch": 16.93, "learning_rate": 1.213513510961538e-06, "loss": 0.0034, "step": 55032 }, { "epoch": 16.93, "learning_rate": 1.213275613416397e-06, "loss": 0.0081, "step": 55033 }, { "epoch": 16.93, "learning_rate": 1.2130377376863244e-06, "loss": 0.0042, "step": 55034 }, { "epoch": 16.93, "learning_rate": 1.2127998837719079e-06, "loss": 0.0042, "step": 55035 }, { "epoch": 16.93, "learning_rate": 1.2125620516737435e-06, "loss": 0.0041, "step": 55036 }, { "epoch": 16.93, "learning_rate": 1.2123242413924174e-06, "loss": 0.0036, "step": 55037 }, { "epoch": 16.93, "learning_rate": 1.2120864529285203e-06, "loss": 0.0052, "step": 55038 }, { "epoch": 16.93, "learning_rate": 1.2118486862826439e-06, "loss": 0.0047, "step": 55039 }, { "epoch": 16.94, "learning_rate": 1.2116109414553767e-06, "loss": 0.0053, "step": 55040 }, { "epoch": 16.94, "learning_rate": 1.2113732184473115e-06, "loss": 0.0044, "step": 55041 }, { "epoch": 16.94, "learning_rate": 1.2111355172590366e-06, "loss": 0.0059, "step": 55042 }, { "epoch": 16.94, "learning_rate": 1.2108978378911441e-06, "loss": 0.01, "step": 55043 }, { "epoch": 16.94, "learning_rate": 1.2106601803442241e-06, "loss": 0.0095, "step": 55044 }, { "epoch": 16.94, "learning_rate": 1.210422544618861e-06, "loss": 0.0067, "step": 55045 }, { "epoch": 16.94, "learning_rate": 1.210184930715652e-06, "loss": 0.0043, "step": 55046 }, { "epoch": 16.94, "learning_rate": 1.2099473386351835e-06, "loss": 0.0037, "step": 55047 }, { "epoch": 16.94, "learning_rate": 1.2097097683780468e-06, "loss": 0.0034, "step": 55048 }, { "epoch": 16.94, "learning_rate": 1.2094722199448305e-06, "loss": 0.0051, "step": 55049 }, { "epoch": 16.94, "learning_rate": 1.2092346933361242e-06, "loss": 0.012, "step": 55050 }, { "epoch": 16.94, "learning_rate": 1.2089971885525187e-06, "loss": 0.0068, "step": 55051 }, { "epoch": 16.94, "learning_rate": 1.2087597055946032e-06, "loss": 0.0076, "step": 55052 }, { "epoch": 16.94, "learning_rate": 1.2085222444629674e-06, "loss": 0.0078, "step": 55053 }, { "epoch": 16.94, "learning_rate": 1.2082848051582007e-06, "loss": 0.0034, "step": 55054 }, { "epoch": 16.94, "learning_rate": 1.2080473876808908e-06, "loss": 0.0033, "step": 55055 }, { "epoch": 16.94, "learning_rate": 1.2078099920316311e-06, "loss": 0.006, "step": 55056 }, { "epoch": 16.94, "learning_rate": 1.2075726182110114e-06, "loss": 0.0038, "step": 55057 }, { "epoch": 16.94, "learning_rate": 1.2073352662196158e-06, "loss": 0.0129, "step": 55058 }, { "epoch": 16.94, "learning_rate": 1.207097936058036e-06, "loss": 0.0069, "step": 55059 }, { "epoch": 16.94, "learning_rate": 1.2068606277268591e-06, "loss": 0.0085, "step": 55060 }, { "epoch": 16.94, "learning_rate": 1.2066233412266803e-06, "loss": 0.0034, "step": 55061 }, { "epoch": 16.94, "learning_rate": 1.2063860765580836e-06, "loss": 0.0022, "step": 55062 }, { "epoch": 16.94, "learning_rate": 1.2061488337216608e-06, "loss": 0.0034, "step": 55063 }, { "epoch": 16.94, "learning_rate": 1.2059116127179993e-06, "loss": 0.0087, "step": 55064 }, { "epoch": 16.94, "learning_rate": 1.2056744135476883e-06, "loss": 0.0046, "step": 55065 }, { "epoch": 16.94, "learning_rate": 1.2054372362113164e-06, "loss": 0.0021, "step": 55066 }, { "epoch": 16.94, "learning_rate": 1.2052000807094733e-06, "loss": 0.0061, "step": 55067 }, { "epoch": 16.94, "learning_rate": 1.2049629470427481e-06, "loss": 0.0033, "step": 55068 }, { "epoch": 16.94, "learning_rate": 1.2047258352117274e-06, "loss": 0.0092, "step": 55069 }, { "epoch": 16.94, "learning_rate": 1.2044887452170029e-06, "loss": 0.0058, "step": 55070 }, { "epoch": 16.94, "learning_rate": 1.2042516770591606e-06, "loss": 0.006, "step": 55071 }, { "epoch": 16.94, "learning_rate": 1.2040146307387901e-06, "loss": 0.0048, "step": 55072 }, { "epoch": 16.95, "learning_rate": 1.203777606256481e-06, "loss": 0.0057, "step": 55073 }, { "epoch": 16.95, "learning_rate": 1.2035406036128194e-06, "loss": 0.0187, "step": 55074 }, { "epoch": 16.95, "learning_rate": 1.2033036228083938e-06, "loss": 0.0072, "step": 55075 }, { "epoch": 16.95, "learning_rate": 1.2030666638437948e-06, "loss": 0.0059, "step": 55076 }, { "epoch": 16.95, "learning_rate": 1.202829726719611e-06, "loss": 0.006, "step": 55077 }, { "epoch": 16.95, "learning_rate": 1.2025928114364306e-06, "loss": 0.0036, "step": 55078 }, { "epoch": 16.95, "learning_rate": 1.2023559179948395e-06, "loss": 0.0044, "step": 55079 }, { "epoch": 16.95, "learning_rate": 1.2021190463954235e-06, "loss": 0.0044, "step": 55080 }, { "epoch": 16.95, "learning_rate": 1.2018821966387773e-06, "loss": 0.0052, "step": 55081 }, { "epoch": 16.95, "learning_rate": 1.2016453687254847e-06, "loss": 0.0062, "step": 55082 }, { "epoch": 16.95, "learning_rate": 1.2014085626561356e-06, "loss": 0.0039, "step": 55083 }, { "epoch": 16.95, "learning_rate": 1.2011717784313159e-06, "loss": 0.0031, "step": 55084 }, { "epoch": 16.95, "learning_rate": 1.2009350160516154e-06, "loss": 0.0067, "step": 55085 }, { "epoch": 16.95, "learning_rate": 1.2006982755176212e-06, "loss": 0.0045, "step": 55086 }, { "epoch": 16.95, "learning_rate": 1.2004615568299204e-06, "loss": 0.005, "step": 55087 }, { "epoch": 16.95, "learning_rate": 1.200224859989101e-06, "loss": 0.0024, "step": 55088 }, { "epoch": 16.95, "learning_rate": 1.1999881849957517e-06, "loss": 0.0061, "step": 55089 }, { "epoch": 16.95, "learning_rate": 1.199751531850457e-06, "loss": 0.0039, "step": 55090 }, { "epoch": 16.95, "learning_rate": 1.1995149005538109e-06, "loss": 0.0038, "step": 55091 }, { "epoch": 16.95, "learning_rate": 1.199278291106395e-06, "loss": 0.0043, "step": 55092 }, { "epoch": 16.95, "learning_rate": 1.1990417035087988e-06, "loss": 0.0038, "step": 55093 }, { "epoch": 16.95, "learning_rate": 1.1988051377616084e-06, "loss": 0.0033, "step": 55094 }, { "epoch": 16.95, "learning_rate": 1.1985685938654101e-06, "loss": 0.0051, "step": 55095 }, { "epoch": 16.95, "learning_rate": 1.198332071820796e-06, "loss": 0.006, "step": 55096 }, { "epoch": 16.95, "learning_rate": 1.1980955716283504e-06, "loss": 0.0118, "step": 55097 }, { "epoch": 16.95, "learning_rate": 1.1978590932886603e-06, "loss": 0.0028, "step": 55098 }, { "epoch": 16.95, "learning_rate": 1.1976226368023146e-06, "loss": 0.0113, "step": 55099 }, { "epoch": 16.95, "learning_rate": 1.1973862021698956e-06, "loss": 0.0042, "step": 55100 }, { "epoch": 16.95, "learning_rate": 1.1971497893919948e-06, "loss": 0.0092, "step": 55101 }, { "epoch": 16.95, "learning_rate": 1.1969133984691983e-06, "loss": 0.0052, "step": 55102 }, { "epoch": 16.95, "learning_rate": 1.1966770294020934e-06, "loss": 0.0042, "step": 55103 }, { "epoch": 16.95, "learning_rate": 1.1964406821912656e-06, "loss": 0.0049, "step": 55104 }, { "epoch": 16.96, "learning_rate": 1.1962043568373016e-06, "loss": 0.0031, "step": 55105 }, { "epoch": 16.96, "learning_rate": 1.1959680533407892e-06, "loss": 0.0076, "step": 55106 }, { "epoch": 16.96, "learning_rate": 1.1957317717023142e-06, "loss": 0.0086, "step": 55107 }, { "epoch": 16.96, "learning_rate": 1.1954955119224644e-06, "loss": 0.0097, "step": 55108 }, { "epoch": 16.96, "learning_rate": 1.1952592740018254e-06, "loss": 0.0151, "step": 55109 }, { "epoch": 16.96, "learning_rate": 1.1950230579409817e-06, "loss": 0.007, "step": 55110 }, { "epoch": 16.96, "learning_rate": 1.1947868637405235e-06, "loss": 0.0046, "step": 55111 }, { "epoch": 16.96, "learning_rate": 1.1945506914010385e-06, "loss": 0.0073, "step": 55112 }, { "epoch": 16.96, "learning_rate": 1.1943145409231083e-06, "loss": 0.0071, "step": 55113 }, { "epoch": 16.96, "learning_rate": 1.1940784123073213e-06, "loss": 0.0048, "step": 55114 }, { "epoch": 16.96, "learning_rate": 1.1938423055542615e-06, "loss": 0.0105, "step": 55115 }, { "epoch": 16.96, "learning_rate": 1.1936062206645183e-06, "loss": 0.0111, "step": 55116 }, { "epoch": 16.96, "learning_rate": 1.193370157638677e-06, "loss": 0.0061, "step": 55117 }, { "epoch": 16.96, "learning_rate": 1.1931341164773246e-06, "loss": 0.0032, "step": 55118 }, { "epoch": 16.96, "learning_rate": 1.1928980971810444e-06, "loss": 0.0055, "step": 55119 }, { "epoch": 16.96, "learning_rate": 1.1926620997504245e-06, "loss": 0.004, "step": 55120 }, { "epoch": 16.96, "learning_rate": 1.1924261241860503e-06, "loss": 0.0057, "step": 55121 }, { "epoch": 16.96, "learning_rate": 1.1921901704885075e-06, "loss": 0.0037, "step": 55122 }, { "epoch": 16.96, "learning_rate": 1.1919542386583826e-06, "loss": 0.0047, "step": 55123 }, { "epoch": 16.96, "learning_rate": 1.1917183286962598e-06, "loss": 0.0066, "step": 55124 }, { "epoch": 16.96, "learning_rate": 1.1914824406027247e-06, "loss": 0.0046, "step": 55125 }, { "epoch": 16.96, "learning_rate": 1.1912465743783674e-06, "loss": 0.004, "step": 55126 }, { "epoch": 16.96, "learning_rate": 1.1910107300237682e-06, "loss": 0.0047, "step": 55127 }, { "epoch": 16.96, "learning_rate": 1.1907749075395147e-06, "loss": 0.0055, "step": 55128 }, { "epoch": 16.96, "learning_rate": 1.1905391069261918e-06, "loss": 0.0054, "step": 55129 }, { "epoch": 16.96, "learning_rate": 1.1903033281843835e-06, "loss": 0.0037, "step": 55130 }, { "epoch": 16.96, "learning_rate": 1.1900675713146791e-06, "loss": 0.0059, "step": 55131 }, { "epoch": 16.96, "learning_rate": 1.189831836317662e-06, "loss": 0.0042, "step": 55132 }, { "epoch": 16.96, "learning_rate": 1.1895961231939191e-06, "loss": 0.0055, "step": 55133 }, { "epoch": 16.96, "learning_rate": 1.1893604319440321e-06, "loss": 0.0057, "step": 55134 }, { "epoch": 16.96, "learning_rate": 1.1891247625685853e-06, "loss": 0.0048, "step": 55135 }, { "epoch": 16.96, "learning_rate": 1.1888891150681681e-06, "loss": 0.0055, "step": 55136 }, { "epoch": 16.96, "learning_rate": 1.1886534894433643e-06, "loss": 0.0065, "step": 55137 }, { "epoch": 16.97, "learning_rate": 1.1884178856947592e-06, "loss": 0.0063, "step": 55138 }, { "epoch": 16.97, "learning_rate": 1.1881823038229357e-06, "loss": 0.0028, "step": 55139 }, { "epoch": 16.97, "learning_rate": 1.187946743828481e-06, "loss": 0.0066, "step": 55140 }, { "epoch": 16.97, "learning_rate": 1.1877112057119788e-06, "loss": 0.0048, "step": 55141 }, { "epoch": 16.97, "learning_rate": 1.1874756894740137e-06, "loss": 0.0065, "step": 55142 }, { "epoch": 16.97, "learning_rate": 1.1872401951151703e-06, "loss": 0.0029, "step": 55143 }, { "epoch": 16.97, "learning_rate": 1.1870047226360337e-06, "loss": 0.0028, "step": 55144 }, { "epoch": 16.97, "learning_rate": 1.186769272037186e-06, "loss": 0.0077, "step": 55145 }, { "epoch": 16.97, "learning_rate": 1.1865338433192175e-06, "loss": 0.0055, "step": 55146 }, { "epoch": 16.97, "learning_rate": 1.1862984364827101e-06, "loss": 0.0041, "step": 55147 }, { "epoch": 16.97, "learning_rate": 1.1860630515282467e-06, "loss": 0.0057, "step": 55148 }, { "epoch": 16.97, "learning_rate": 1.1858276884564125e-06, "loss": 0.0061, "step": 55149 }, { "epoch": 16.97, "learning_rate": 1.1855923472677888e-06, "loss": 0.0051, "step": 55150 }, { "epoch": 16.97, "learning_rate": 1.1853570279629668e-06, "loss": 0.0059, "step": 55151 }, { "epoch": 16.97, "learning_rate": 1.1851217305425256e-06, "loss": 0.0078, "step": 55152 }, { "epoch": 16.97, "learning_rate": 1.1848864550070505e-06, "loss": 0.0088, "step": 55153 }, { "epoch": 16.97, "learning_rate": 1.1846512013571288e-06, "loss": 0.0034, "step": 55154 }, { "epoch": 16.97, "learning_rate": 1.1844159695933377e-06, "loss": 0.0056, "step": 55155 }, { "epoch": 16.97, "learning_rate": 1.1841807597162657e-06, "loss": 0.0043, "step": 55156 }, { "epoch": 16.97, "learning_rate": 1.183945571726498e-06, "loss": 0.0062, "step": 55157 }, { "epoch": 16.97, "learning_rate": 1.1837104056246152e-06, "loss": 0.0051, "step": 55158 }, { "epoch": 16.97, "learning_rate": 1.1834752614112043e-06, "loss": 0.0041, "step": 55159 }, { "epoch": 16.97, "learning_rate": 1.1832401390868465e-06, "loss": 0.0079, "step": 55160 }, { "epoch": 16.97, "learning_rate": 1.1830050386521264e-06, "loss": 0.0045, "step": 55161 }, { "epoch": 16.97, "learning_rate": 1.182769960107628e-06, "loss": 0.0065, "step": 55162 }, { "epoch": 16.97, "learning_rate": 1.1825349034539347e-06, "loss": 0.0058, "step": 55163 }, { "epoch": 16.97, "learning_rate": 1.1822998686916298e-06, "loss": 0.0173, "step": 55164 }, { "epoch": 16.97, "learning_rate": 1.1820648558212954e-06, "loss": 0.0104, "step": 55165 }, { "epoch": 16.97, "learning_rate": 1.181829864843519e-06, "loss": 0.0038, "step": 55166 }, { "epoch": 16.97, "learning_rate": 1.181594895758883e-06, "loss": 0.0055, "step": 55167 }, { "epoch": 16.97, "learning_rate": 1.1813599485679684e-06, "loss": 0.0041, "step": 55168 }, { "epoch": 16.97, "learning_rate": 1.181125023271359e-06, "loss": 0.0026, "step": 55169 }, { "epoch": 16.98, "learning_rate": 1.1808901198696365e-06, "loss": 0.0029, "step": 55170 }, { "epoch": 16.98, "learning_rate": 1.1806552383633897e-06, "loss": 0.0144, "step": 55171 }, { "epoch": 16.98, "learning_rate": 1.1804203787531965e-06, "loss": 0.0043, "step": 55172 }, { "epoch": 16.98, "learning_rate": 1.1801855410396423e-06, "loss": 0.0062, "step": 55173 }, { "epoch": 16.98, "learning_rate": 1.17995072522331e-06, "loss": 0.0043, "step": 55174 }, { "epoch": 16.98, "learning_rate": 1.1797159313047823e-06, "loss": 0.0031, "step": 55175 }, { "epoch": 16.98, "learning_rate": 1.1794811592846423e-06, "loss": 0.0051, "step": 55176 }, { "epoch": 16.98, "learning_rate": 1.1792464091634725e-06, "loss": 0.0081, "step": 55177 }, { "epoch": 16.98, "learning_rate": 1.179011680941855e-06, "loss": 0.0067, "step": 55178 }, { "epoch": 16.98, "learning_rate": 1.1787769746203747e-06, "loss": 0.0103, "step": 55179 }, { "epoch": 16.98, "learning_rate": 1.1785422901996102e-06, "loss": 0.0072, "step": 55180 }, { "epoch": 16.98, "learning_rate": 1.1783076276801508e-06, "loss": 0.007, "step": 55181 }, { "epoch": 16.98, "learning_rate": 1.178072987062574e-06, "loss": 0.005, "step": 55182 }, { "epoch": 16.98, "learning_rate": 1.1778383683474637e-06, "loss": 0.0073, "step": 55183 }, { "epoch": 16.98, "learning_rate": 1.1776037715354017e-06, "loss": 0.007, "step": 55184 }, { "epoch": 16.98, "learning_rate": 1.1773691966269696e-06, "loss": 0.0039, "step": 55185 }, { "epoch": 16.98, "learning_rate": 1.1771346436227538e-06, "loss": 0.003, "step": 55186 }, { "epoch": 16.98, "learning_rate": 1.1769001125233326e-06, "loss": 0.005, "step": 55187 }, { "epoch": 16.98, "learning_rate": 1.1766656033292933e-06, "loss": 0.0086, "step": 55188 }, { "epoch": 16.98, "learning_rate": 1.1764311160412112e-06, "loss": 0.005, "step": 55189 }, { "epoch": 16.98, "learning_rate": 1.1761966506596711e-06, "loss": 0.0075, "step": 55190 }, { "epoch": 16.98, "learning_rate": 1.1759622071852572e-06, "loss": 0.0091, "step": 55191 }, { "epoch": 16.98, "learning_rate": 1.17572778561855e-06, "loss": 0.0074, "step": 55192 }, { "epoch": 16.98, "learning_rate": 1.1754933859601324e-06, "loss": 0.0078, "step": 55193 }, { "epoch": 16.98, "learning_rate": 1.1752590082105863e-06, "loss": 0.0126, "step": 55194 }, { "epoch": 16.98, "learning_rate": 1.1750246523704922e-06, "loss": 0.0041, "step": 55195 }, { "epoch": 16.98, "learning_rate": 1.174790318440432e-06, "loss": 0.0034, "step": 55196 }, { "epoch": 16.98, "learning_rate": 1.1745560064209894e-06, "loss": 0.0035, "step": 55197 }, { "epoch": 16.98, "learning_rate": 1.174321716312744e-06, "loss": 0.0046, "step": 55198 }, { "epoch": 16.98, "learning_rate": 1.1740874481162802e-06, "loss": 0.007, "step": 55199 }, { "epoch": 16.98, "learning_rate": 1.1738532018321735e-06, "loss": 0.003, "step": 55200 }, { "epoch": 16.98, "learning_rate": 1.173618977461014e-06, "loss": 0.0043, "step": 55201 }, { "epoch": 16.98, "learning_rate": 1.17338477500338e-06, "loss": 0.0076, "step": 55202 }, { "epoch": 16.99, "learning_rate": 1.1731505944598497e-06, "loss": 0.0054, "step": 55203 }, { "epoch": 16.99, "learning_rate": 1.1729164358310075e-06, "loss": 0.005, "step": 55204 }, { "epoch": 16.99, "learning_rate": 1.1726822991174314e-06, "loss": 0.0043, "step": 55205 }, { "epoch": 16.99, "learning_rate": 1.1724481843197078e-06, "loss": 0.0068, "step": 55206 }, { "epoch": 16.99, "learning_rate": 1.1722140914384162e-06, "loss": 0.0047, "step": 55207 }, { "epoch": 16.99, "learning_rate": 1.1719800204741361e-06, "loss": 0.0062, "step": 55208 }, { "epoch": 16.99, "learning_rate": 1.1717459714274526e-06, "loss": 0.0071, "step": 55209 }, { "epoch": 16.99, "learning_rate": 1.1715119442989387e-06, "loss": 0.0027, "step": 55210 }, { "epoch": 16.99, "learning_rate": 1.1712779390891826e-06, "loss": 0.0043, "step": 55211 }, { "epoch": 16.99, "learning_rate": 1.1710439557987641e-06, "loss": 0.0058, "step": 55212 }, { "epoch": 16.99, "learning_rate": 1.1708099944282625e-06, "loss": 0.0047, "step": 55213 }, { "epoch": 16.99, "learning_rate": 1.1705760549782598e-06, "loss": 0.0064, "step": 55214 }, { "epoch": 16.99, "learning_rate": 1.1703421374493373e-06, "loss": 0.0034, "step": 55215 }, { "epoch": 16.99, "learning_rate": 1.170108241842074e-06, "loss": 0.0076, "step": 55216 }, { "epoch": 16.99, "learning_rate": 1.1698743681570513e-06, "loss": 0.0082, "step": 55217 }, { "epoch": 16.99, "learning_rate": 1.16964051639485e-06, "loss": 0.0042, "step": 55218 }, { "epoch": 16.99, "learning_rate": 1.1694066865560505e-06, "loss": 0.0053, "step": 55219 }, { "epoch": 16.99, "learning_rate": 1.1691728786412315e-06, "loss": 0.005, "step": 55220 }, { "epoch": 16.99, "learning_rate": 1.1689390926509781e-06, "loss": 0.0047, "step": 55221 }, { "epoch": 16.99, "learning_rate": 1.1687053285858685e-06, "loss": 0.0095, "step": 55222 }, { "epoch": 16.99, "learning_rate": 1.1684715864464835e-06, "loss": 0.0048, "step": 55223 }, { "epoch": 16.99, "learning_rate": 1.1682378662334014e-06, "loss": 0.0036, "step": 55224 }, { "epoch": 16.99, "learning_rate": 1.168004167947202e-06, "loss": 0.0069, "step": 55225 }, { "epoch": 16.99, "learning_rate": 1.167770491588469e-06, "loss": 0.0076, "step": 55226 }, { "epoch": 16.99, "learning_rate": 1.16753683715778e-06, "loss": 0.0053, "step": 55227 }, { "epoch": 16.99, "learning_rate": 1.1673032046557164e-06, "loss": 0.0069, "step": 55228 }, { "epoch": 16.99, "learning_rate": 1.167069594082858e-06, "loss": 0.0095, "step": 55229 }, { "epoch": 16.99, "learning_rate": 1.1668360054397842e-06, "loss": 0.0028, "step": 55230 }, { "epoch": 16.99, "learning_rate": 1.1666024387270759e-06, "loss": 0.0063, "step": 55231 }, { "epoch": 16.99, "learning_rate": 1.1663688939453122e-06, "loss": 0.0071, "step": 55232 }, { "epoch": 16.99, "learning_rate": 1.166135371095073e-06, "loss": 0.0052, "step": 55233 }, { "epoch": 16.99, "learning_rate": 1.1659018701769375e-06, "loss": 0.0055, "step": 55234 }, { "epoch": 17.0, "learning_rate": 1.1656683911914846e-06, "loss": 0.0079, "step": 55235 }, { "epoch": 17.0, "learning_rate": 1.165434934139299e-06, "loss": 0.0039, "step": 55236 }, { "epoch": 17.0, "learning_rate": 1.1652014990209549e-06, "loss": 0.0055, "step": 55237 }, { "epoch": 17.0, "learning_rate": 1.1649680858370337e-06, "loss": 0.004, "step": 55238 }, { "epoch": 17.0, "learning_rate": 1.1647346945881144e-06, "loss": 0.0052, "step": 55239 }, { "epoch": 17.0, "learning_rate": 1.164501325274776e-06, "loss": 0.0054, "step": 55240 }, { "epoch": 17.0, "learning_rate": 1.1642679778975995e-06, "loss": 0.0058, "step": 55241 }, { "epoch": 17.0, "learning_rate": 1.1640346524571643e-06, "loss": 0.0062, "step": 55242 }, { "epoch": 17.0, "learning_rate": 1.16380134895405e-06, "loss": 0.0051, "step": 55243 }, { "epoch": 17.0, "learning_rate": 1.163568067388835e-06, "loss": 0.0048, "step": 55244 }, { "epoch": 17.0, "learning_rate": 1.1633348077620953e-06, "loss": 0.0046, "step": 55245 }, { "epoch": 17.0, "learning_rate": 1.1631015700744153e-06, "loss": 0.0102, "step": 55246 }, { "epoch": 17.0, "learning_rate": 1.1628683543263708e-06, "loss": 0.0031, "step": 55247 }, { "epoch": 17.0, "learning_rate": 1.1626351605185427e-06, "loss": 0.006, "step": 55248 }, { "epoch": 17.0, "learning_rate": 1.1624019886515082e-06, "loss": 0.0056, "step": 55249 }, { "epoch": 17.0, "learning_rate": 1.1621688387258468e-06, "loss": 0.0112, "step": 55250 }, { "epoch": 17.0, "learning_rate": 1.161935710742138e-06, "loss": 0.0052, "step": 55251 }, { "epoch": 17.0, "learning_rate": 1.1617026047009606e-06, "loss": 0.0062, "step": 55252 }, { "epoch": 17.0, "learning_rate": 1.1614695206028926e-06, "loss": 0.0035, "step": 55253 }, { "epoch": 17.0, "learning_rate": 1.1612364584485125e-06, "loss": 0.0074, "step": 55254 }, { "epoch": 17.0, "learning_rate": 1.1610034182383977e-06, "loss": 0.0054, "step": 55255 }, { "epoch": 17.0, "learning_rate": 1.1607703999731312e-06, "loss": 0.0062, "step": 55256 }, { "epoch": 17.0, "learning_rate": 1.1605374036532902e-06, "loss": 0.0062, "step": 55257 }, { "epoch": 17.0, "learning_rate": 1.1603044292794497e-06, "loss": 0.0056, "step": 55258 }, { "epoch": 17.0, "learning_rate": 1.1600714768521903e-06, "loss": 0.0036, "step": 55259 }, { "epoch": 17.0, "learning_rate": 1.1598385463720884e-06, "loss": 0.0046, "step": 55260 }, { "epoch": 17.0, "learning_rate": 1.1596056378397257e-06, "loss": 0.0047, "step": 55261 }, { "epoch": 17.0, "learning_rate": 1.1593727512556797e-06, "loss": 0.004, "step": 55262 }, { "epoch": 17.0, "learning_rate": 1.1591398866205273e-06, "loss": 0.0031, "step": 55263 }, { "epoch": 17.0, "learning_rate": 1.158907043934847e-06, "loss": 0.0028, "step": 55264 }, { "epoch": 17.0, "learning_rate": 1.1586742231992177e-06, "loss": 0.006, "step": 55265 }, { "epoch": 17.0, "learning_rate": 1.1584414244142162e-06, "loss": 0.0035, "step": 55266 }, { "epoch": 17.0, "learning_rate": 1.15820864758042e-06, "loss": 0.0045, "step": 55267 }, { "epoch": 17.01, "learning_rate": 1.1579758926984098e-06, "loss": 0.0029, "step": 55268 }, { "epoch": 17.01, "learning_rate": 1.1577431597687604e-06, "loss": 0.0042, "step": 55269 }, { "epoch": 17.01, "learning_rate": 1.157510448792052e-06, "loss": 0.0033, "step": 55270 }, { "epoch": 17.01, "learning_rate": 1.1572777597688611e-06, "loss": 0.0031, "step": 55271 }, { "epoch": 17.01, "learning_rate": 1.1570450926997657e-06, "loss": 0.0045, "step": 55272 }, { "epoch": 17.01, "learning_rate": 1.1568124475853426e-06, "loss": 0.0036, "step": 55273 }, { "epoch": 17.01, "learning_rate": 1.1565798244261706e-06, "loss": 0.0041, "step": 55274 }, { "epoch": 17.01, "learning_rate": 1.1563472232228247e-06, "loss": 0.0041, "step": 55275 }, { "epoch": 17.01, "learning_rate": 1.1561146439758862e-06, "loss": 0.0061, "step": 55276 }, { "epoch": 17.01, "learning_rate": 1.1558820866859311e-06, "loss": 0.0051, "step": 55277 }, { "epoch": 17.01, "learning_rate": 1.1556495513535393e-06, "loss": 0.0033, "step": 55278 }, { "epoch": 17.01, "learning_rate": 1.1554170379792817e-06, "loss": 0.005, "step": 55279 }, { "epoch": 17.01, "learning_rate": 1.155184546563738e-06, "loss": 0.006, "step": 55280 }, { "epoch": 17.01, "learning_rate": 1.1549520771074886e-06, "loss": 0.0042, "step": 55281 }, { "epoch": 17.01, "learning_rate": 1.1547196296111085e-06, "loss": 0.0043, "step": 55282 }, { "epoch": 17.01, "learning_rate": 1.1544872040751753e-06, "loss": 0.0041, "step": 55283 }, { "epoch": 17.01, "learning_rate": 1.1542548005002651e-06, "loss": 0.0062, "step": 55284 }, { "epoch": 17.01, "learning_rate": 1.1540224188869565e-06, "loss": 0.0071, "step": 55285 }, { "epoch": 17.01, "learning_rate": 1.1537900592358243e-06, "loss": 0.0035, "step": 55286 }, { "epoch": 17.01, "learning_rate": 1.153557721547447e-06, "loss": 0.0045, "step": 55287 }, { "epoch": 17.01, "learning_rate": 1.1533254058224008e-06, "loss": 0.0037, "step": 55288 }, { "epoch": 17.01, "learning_rate": 1.1530931120612632e-06, "loss": 0.0032, "step": 55289 }, { "epoch": 17.01, "learning_rate": 1.152860840264608e-06, "loss": 0.0019, "step": 55290 }, { "epoch": 17.01, "learning_rate": 1.1526285904330182e-06, "loss": 0.004, "step": 55291 }, { "epoch": 17.01, "learning_rate": 1.1523963625670643e-06, "loss": 0.0036, "step": 55292 }, { "epoch": 17.01, "learning_rate": 1.1521641566673247e-06, "loss": 0.0045, "step": 55293 }, { "epoch": 17.01, "learning_rate": 1.151931972734377e-06, "loss": 0.0031, "step": 55294 }, { "epoch": 17.01, "learning_rate": 1.1516998107687939e-06, "loss": 0.0036, "step": 55295 }, { "epoch": 17.01, "learning_rate": 1.151467670771157e-06, "loss": 0.0035, "step": 55296 }, { "epoch": 17.01, "learning_rate": 1.1512355527420406e-06, "loss": 0.0055, "step": 55297 }, { "epoch": 17.01, "learning_rate": 1.1510034566820205e-06, "loss": 0.0054, "step": 55298 }, { "epoch": 17.01, "learning_rate": 1.1507713825916767e-06, "loss": 0.0051, "step": 55299 }, { "epoch": 17.02, "learning_rate": 1.1505393304715763e-06, "loss": 0.0051, "step": 55300 }, { "epoch": 17.02, "learning_rate": 1.150307300322303e-06, "loss": 0.0085, "step": 55301 }, { "epoch": 17.02, "learning_rate": 1.1500752921444314e-06, "loss": 0.0043, "step": 55302 }, { "epoch": 17.02, "learning_rate": 1.1498433059385371e-06, "loss": 0.0031, "step": 55303 }, { "epoch": 17.02, "learning_rate": 1.1496113417051957e-06, "loss": 0.0037, "step": 55304 }, { "epoch": 17.02, "learning_rate": 1.1493793994449831e-06, "loss": 0.0032, "step": 55305 }, { "epoch": 17.02, "learning_rate": 1.1491474791584755e-06, "loss": 0.0031, "step": 55306 }, { "epoch": 17.02, "learning_rate": 1.1489155808462483e-06, "loss": 0.0105, "step": 55307 }, { "epoch": 17.02, "learning_rate": 1.1486837045088784e-06, "loss": 0.0027, "step": 55308 }, { "epoch": 17.02, "learning_rate": 1.14845185014694e-06, "loss": 0.0075, "step": 55309 }, { "epoch": 17.02, "learning_rate": 1.148220017761007e-06, "loss": 0.004, "step": 55310 }, { "epoch": 17.02, "learning_rate": 1.14798820735166e-06, "loss": 0.0036, "step": 55311 }, { "epoch": 17.02, "learning_rate": 1.1477564189194745e-06, "loss": 0.0036, "step": 55312 }, { "epoch": 17.02, "learning_rate": 1.1475246524650197e-06, "loss": 0.0163, "step": 55313 }, { "epoch": 17.02, "learning_rate": 1.1472929079888762e-06, "loss": 0.0036, "step": 55314 }, { "epoch": 17.02, "learning_rate": 1.1470611854916147e-06, "loss": 0.0058, "step": 55315 }, { "epoch": 17.02, "learning_rate": 1.146829484973816e-06, "loss": 0.0043, "step": 55316 }, { "epoch": 17.02, "learning_rate": 1.1465978064360528e-06, "loss": 0.0029, "step": 55317 }, { "epoch": 17.02, "learning_rate": 1.1463661498789003e-06, "loss": 0.0023, "step": 55318 }, { "epoch": 17.02, "learning_rate": 1.1461345153029346e-06, "loss": 0.0045, "step": 55319 }, { "epoch": 17.02, "learning_rate": 1.1459029027087288e-06, "loss": 0.004, "step": 55320 }, { "epoch": 17.02, "learning_rate": 1.14567131209686e-06, "loss": 0.004, "step": 55321 }, { "epoch": 17.02, "learning_rate": 1.1454397434679022e-06, "loss": 0.0041, "step": 55322 }, { "epoch": 17.02, "learning_rate": 1.1452081968224293e-06, "loss": 0.0042, "step": 55323 }, { "epoch": 17.02, "learning_rate": 1.144976672161019e-06, "loss": 0.003, "step": 55324 }, { "epoch": 17.02, "learning_rate": 1.1447451694842427e-06, "loss": 0.005, "step": 55325 }, { "epoch": 17.02, "learning_rate": 1.1445136887926777e-06, "loss": 0.0051, "step": 55326 }, { "epoch": 17.02, "learning_rate": 1.1442822300868983e-06, "loss": 0.0028, "step": 55327 }, { "epoch": 17.02, "learning_rate": 1.1440507933674784e-06, "loss": 0.0023, "step": 55328 }, { "epoch": 17.02, "learning_rate": 1.1438193786349928e-06, "loss": 0.0025, "step": 55329 }, { "epoch": 17.02, "learning_rate": 1.1435879858900134e-06, "loss": 0.0019, "step": 55330 }, { "epoch": 17.02, "learning_rate": 1.14335661513312e-06, "loss": 0.003, "step": 55331 }, { "epoch": 17.02, "learning_rate": 1.1431252663648851e-06, "loss": 0.0019, "step": 55332 }, { "epoch": 17.03, "learning_rate": 1.142893939585883e-06, "loss": 0.0054, "step": 55333 }, { "epoch": 17.03, "learning_rate": 1.1426626347966862e-06, "loss": 0.0065, "step": 55334 }, { "epoch": 17.03, "learning_rate": 1.1424313519978692e-06, "loss": 0.0053, "step": 55335 }, { "epoch": 17.03, "learning_rate": 1.1422000911900078e-06, "loss": 0.0068, "step": 55336 }, { "epoch": 17.03, "learning_rate": 1.1419688523736761e-06, "loss": 0.0046, "step": 55337 }, { "epoch": 17.03, "learning_rate": 1.141737635549448e-06, "loss": 0.0034, "step": 55338 }, { "epoch": 17.03, "learning_rate": 1.1415064407178967e-06, "loss": 0.0047, "step": 55339 }, { "epoch": 17.03, "learning_rate": 1.1412752678795969e-06, "loss": 0.0043, "step": 55340 }, { "epoch": 17.03, "learning_rate": 1.1410441170351216e-06, "loss": 0.0072, "step": 55341 }, { "epoch": 17.03, "learning_rate": 1.140812988185047e-06, "loss": 0.0038, "step": 55342 }, { "epoch": 17.03, "learning_rate": 1.140581881329944e-06, "loss": 0.0043, "step": 55343 }, { "epoch": 17.03, "learning_rate": 1.1403507964703887e-06, "loss": 0.0048, "step": 55344 }, { "epoch": 17.03, "learning_rate": 1.1401197336069514e-06, "loss": 0.0044, "step": 55345 }, { "epoch": 17.03, "learning_rate": 1.1398886927402131e-06, "loss": 0.0032, "step": 55346 }, { "epoch": 17.03, "learning_rate": 1.1396576738707399e-06, "loss": 0.002, "step": 55347 }, { "epoch": 17.03, "learning_rate": 1.139426676999108e-06, "loss": 0.0058, "step": 55348 }, { "epoch": 17.03, "learning_rate": 1.1391957021258904e-06, "loss": 0.0034, "step": 55349 }, { "epoch": 17.03, "learning_rate": 1.1389647492516598e-06, "loss": 0.005, "step": 55350 }, { "epoch": 17.03, "learning_rate": 1.1387338183769924e-06, "loss": 0.0039, "step": 55351 }, { "epoch": 17.03, "learning_rate": 1.1385029095024602e-06, "loss": 0.0036, "step": 55352 }, { "epoch": 17.03, "learning_rate": 1.138272022628637e-06, "loss": 0.0039, "step": 55353 }, { "epoch": 17.03, "learning_rate": 1.1380411577560968e-06, "loss": 0.0029, "step": 55354 }, { "epoch": 17.03, "learning_rate": 1.137810314885407e-06, "loss": 0.009, "step": 55355 }, { "epoch": 17.03, "learning_rate": 1.137579494017147e-06, "loss": 0.0039, "step": 55356 }, { "epoch": 17.03, "learning_rate": 1.1373486951518887e-06, "loss": 0.0046, "step": 55357 }, { "epoch": 17.03, "learning_rate": 1.1371179182902037e-06, "loss": 0.0056, "step": 55358 }, { "epoch": 17.03, "learning_rate": 1.136887163432665e-06, "loss": 0.0034, "step": 55359 }, { "epoch": 17.03, "learning_rate": 1.1366564305798466e-06, "loss": 0.0057, "step": 55360 }, { "epoch": 17.03, "learning_rate": 1.1364257197323214e-06, "loss": 0.0038, "step": 55361 }, { "epoch": 17.03, "learning_rate": 1.13619503089066e-06, "loss": 0.0142, "step": 55362 }, { "epoch": 17.03, "learning_rate": 1.1359643640554384e-06, "loss": 0.0032, "step": 55363 }, { "epoch": 17.03, "learning_rate": 1.1357337192272266e-06, "loss": 0.0044, "step": 55364 }, { "epoch": 17.04, "learning_rate": 1.1355030964065972e-06, "loss": 0.0055, "step": 55365 }, { "epoch": 17.04, "learning_rate": 1.1352724955941242e-06, "loss": 0.0093, "step": 55366 }, { "epoch": 17.04, "learning_rate": 1.1350419167903837e-06, "loss": 0.0051, "step": 55367 }, { "epoch": 17.04, "learning_rate": 1.1348113599959409e-06, "loss": 0.0041, "step": 55368 }, { "epoch": 17.04, "learning_rate": 1.13458082521137e-06, "loss": 0.0043, "step": 55369 }, { "epoch": 17.04, "learning_rate": 1.1343503124372467e-06, "loss": 0.003, "step": 55370 }, { "epoch": 17.04, "learning_rate": 1.1341198216741411e-06, "loss": 0.003, "step": 55371 }, { "epoch": 17.04, "learning_rate": 1.1338893529226257e-06, "loss": 0.0053, "step": 55372 }, { "epoch": 17.04, "learning_rate": 1.1336589061832736e-06, "loss": 0.0041, "step": 55373 }, { "epoch": 17.04, "learning_rate": 1.1334284814566565e-06, "loss": 0.0036, "step": 55374 }, { "epoch": 17.04, "learning_rate": 1.133198078743345e-06, "loss": 0.0068, "step": 55375 }, { "epoch": 17.04, "learning_rate": 1.132967698043913e-06, "loss": 0.0038, "step": 55376 }, { "epoch": 17.04, "learning_rate": 1.1327373393589315e-06, "loss": 0.007, "step": 55377 }, { "epoch": 17.04, "learning_rate": 1.132507002688973e-06, "loss": 0.0024, "step": 55378 }, { "epoch": 17.04, "learning_rate": 1.1322766880346058e-06, "loss": 0.0059, "step": 55379 }, { "epoch": 17.04, "learning_rate": 1.1320463953964078e-06, "loss": 0.0054, "step": 55380 }, { "epoch": 17.04, "learning_rate": 1.1318161247749504e-06, "loss": 0.0038, "step": 55381 }, { "epoch": 17.04, "learning_rate": 1.1315858761707999e-06, "loss": 0.0033, "step": 55382 }, { "epoch": 17.04, "learning_rate": 1.1313556495845313e-06, "loss": 0.0022, "step": 55383 }, { "epoch": 17.04, "learning_rate": 1.1311254450167132e-06, "loss": 0.0032, "step": 55384 }, { "epoch": 17.04, "learning_rate": 1.1308952624679216e-06, "loss": 0.0091, "step": 55385 }, { "epoch": 17.04, "learning_rate": 1.1306651019387261e-06, "loss": 0.0024, "step": 55386 }, { "epoch": 17.04, "learning_rate": 1.1304349634296984e-06, "loss": 0.0016, "step": 55387 }, { "epoch": 17.04, "learning_rate": 1.1302048469414117e-06, "loss": 0.006, "step": 55388 }, { "epoch": 17.04, "learning_rate": 1.1299747524744309e-06, "loss": 0.0033, "step": 55389 }, { "epoch": 17.04, "learning_rate": 1.1297446800293332e-06, "loss": 0.0036, "step": 55390 }, { "epoch": 17.04, "learning_rate": 1.1295146296066884e-06, "loss": 0.0046, "step": 55391 }, { "epoch": 17.04, "learning_rate": 1.1292846012070679e-06, "loss": 0.0039, "step": 55392 }, { "epoch": 17.04, "learning_rate": 1.1290545948310416e-06, "loss": 0.0056, "step": 55393 }, { "epoch": 17.04, "learning_rate": 1.1288246104791812e-06, "loss": 0.0049, "step": 55394 }, { "epoch": 17.04, "learning_rate": 1.1285946481520582e-06, "loss": 0.0028, "step": 55395 }, { "epoch": 17.04, "learning_rate": 1.1283647078502424e-06, "loss": 0.0049, "step": 55396 }, { "epoch": 17.04, "learning_rate": 1.1281347895743056e-06, "loss": 0.0046, "step": 55397 }, { "epoch": 17.05, "learning_rate": 1.1279048933248182e-06, "loss": 0.0023, "step": 55398 }, { "epoch": 17.05, "learning_rate": 1.1276750191023488e-06, "loss": 0.003, "step": 55399 }, { "epoch": 17.05, "learning_rate": 1.1274451669074738e-06, "loss": 0.0057, "step": 55400 }, { "epoch": 17.05, "learning_rate": 1.127215336740759e-06, "loss": 0.0054, "step": 55401 }, { "epoch": 17.05, "learning_rate": 1.1269855286027798e-06, "loss": 0.0039, "step": 55402 }, { "epoch": 17.05, "learning_rate": 1.1267557424941e-06, "loss": 0.0039, "step": 55403 }, { "epoch": 17.05, "learning_rate": 1.1265259784152926e-06, "loss": 0.0066, "step": 55404 }, { "epoch": 17.05, "learning_rate": 1.1262962363669305e-06, "loss": 0.0025, "step": 55405 }, { "epoch": 17.05, "learning_rate": 1.1260665163495821e-06, "loss": 0.0064, "step": 55406 }, { "epoch": 17.05, "learning_rate": 1.125836818363819e-06, "loss": 0.0108, "step": 55407 }, { "epoch": 17.05, "learning_rate": 1.125607142410211e-06, "loss": 0.0046, "step": 55408 }, { "epoch": 17.05, "learning_rate": 1.1253774884893264e-06, "loss": 0.0039, "step": 55409 }, { "epoch": 17.05, "learning_rate": 1.1251478566017381e-06, "loss": 0.0037, "step": 55410 }, { "epoch": 17.05, "learning_rate": 1.1249182467480158e-06, "loss": 0.0031, "step": 55411 }, { "epoch": 17.05, "learning_rate": 1.1246886589287275e-06, "loss": 0.005, "step": 55412 }, { "epoch": 17.05, "learning_rate": 1.1244590931444455e-06, "loss": 0.004, "step": 55413 }, { "epoch": 17.05, "learning_rate": 1.1242295493957367e-06, "loss": 0.003, "step": 55414 }, { "epoch": 17.05, "learning_rate": 1.1240000276831765e-06, "loss": 0.0049, "step": 55415 }, { "epoch": 17.05, "learning_rate": 1.1237705280073298e-06, "loss": 0.0035, "step": 55416 }, { "epoch": 17.05, "learning_rate": 1.1235410503687672e-06, "loss": 0.0034, "step": 55417 }, { "epoch": 17.05, "learning_rate": 1.1233115947680594e-06, "loss": 0.0042, "step": 55418 }, { "epoch": 17.05, "learning_rate": 1.123082161205775e-06, "loss": 0.0037, "step": 55419 }, { "epoch": 17.05, "learning_rate": 1.1228527496824848e-06, "loss": 0.0047, "step": 55420 }, { "epoch": 17.05, "learning_rate": 1.1226233601987591e-06, "loss": 0.0071, "step": 55421 }, { "epoch": 17.05, "learning_rate": 1.1223939927551664e-06, "loss": 0.0056, "step": 55422 }, { "epoch": 17.05, "learning_rate": 1.1221646473522774e-06, "loss": 0.0035, "step": 55423 }, { "epoch": 17.05, "learning_rate": 1.121935323990656e-06, "loss": 0.0076, "step": 55424 }, { "epoch": 17.05, "learning_rate": 1.1217060226708787e-06, "loss": 0.0046, "step": 55425 }, { "epoch": 17.05, "learning_rate": 1.1214767433935114e-06, "loss": 0.0051, "step": 55426 }, { "epoch": 17.05, "learning_rate": 1.1212474861591239e-06, "loss": 0.0056, "step": 55427 }, { "epoch": 17.05, "learning_rate": 1.1210182509682854e-06, "loss": 0.0142, "step": 55428 }, { "epoch": 17.05, "learning_rate": 1.1207890378215646e-06, "loss": 0.0043, "step": 55429 }, { "epoch": 17.06, "learning_rate": 1.12055984671953e-06, "loss": 0.0035, "step": 55430 }, { "epoch": 17.06, "learning_rate": 1.1203306776627532e-06, "loss": 0.0079, "step": 55431 }, { "epoch": 17.06, "learning_rate": 1.1201015306518005e-06, "loss": 0.0025, "step": 55432 }, { "epoch": 17.06, "learning_rate": 1.1198724056872413e-06, "loss": 0.0057, "step": 55433 }, { "epoch": 17.06, "learning_rate": 1.119643302769644e-06, "loss": 0.0048, "step": 55434 }, { "epoch": 17.06, "learning_rate": 1.1194142218995795e-06, "loss": 0.0069, "step": 55435 }, { "epoch": 17.06, "learning_rate": 1.1191851630776173e-06, "loss": 0.004, "step": 55436 }, { "epoch": 17.06, "learning_rate": 1.1189561263043225e-06, "loss": 0.0039, "step": 55437 }, { "epoch": 17.06, "learning_rate": 1.1187271115802656e-06, "loss": 0.0063, "step": 55438 }, { "epoch": 17.06, "learning_rate": 1.1184981189060128e-06, "loss": 0.0048, "step": 55439 }, { "epoch": 17.06, "learning_rate": 1.1182691482821361e-06, "loss": 0.0026, "step": 55440 }, { "epoch": 17.06, "learning_rate": 1.118040199709204e-06, "loss": 0.0054, "step": 55441 }, { "epoch": 17.06, "learning_rate": 1.1178112731877822e-06, "loss": 0.0033, "step": 55442 }, { "epoch": 17.06, "learning_rate": 1.1175823687184405e-06, "loss": 0.0045, "step": 55443 }, { "epoch": 17.06, "learning_rate": 1.1173534863017488e-06, "loss": 0.0028, "step": 55444 }, { "epoch": 17.06, "learning_rate": 1.1171246259382717e-06, "loss": 0.0032, "step": 55445 }, { "epoch": 17.06, "learning_rate": 1.11689578762858e-06, "loss": 0.0068, "step": 55446 }, { "epoch": 17.06, "learning_rate": 1.1166669713732425e-06, "loss": 0.0061, "step": 55447 }, { "epoch": 17.06, "learning_rate": 1.116438177172825e-06, "loss": 0.0035, "step": 55448 }, { "epoch": 17.06, "learning_rate": 1.1162094050278972e-06, "loss": 0.0067, "step": 55449 }, { "epoch": 17.06, "learning_rate": 1.115980654939025e-06, "loss": 0.006, "step": 55450 }, { "epoch": 17.06, "learning_rate": 1.1157519269067785e-06, "loss": 0.0042, "step": 55451 }, { "epoch": 17.06, "learning_rate": 1.1155232209317258e-06, "loss": 0.0036, "step": 55452 }, { "epoch": 17.06, "learning_rate": 1.115294537014433e-06, "loss": 0.0045, "step": 55453 }, { "epoch": 17.06, "learning_rate": 1.1150658751554667e-06, "loss": 0.0028, "step": 55454 }, { "epoch": 17.06, "learning_rate": 1.1148372353553983e-06, "loss": 0.0054, "step": 55455 }, { "epoch": 17.06, "learning_rate": 1.1146086176147942e-06, "loss": 0.0036, "step": 55456 }, { "epoch": 17.06, "learning_rate": 1.1143800219342226e-06, "loss": 0.0047, "step": 55457 }, { "epoch": 17.06, "learning_rate": 1.1141514483142491e-06, "loss": 0.0052, "step": 55458 }, { "epoch": 17.06, "learning_rate": 1.1139228967554383e-06, "loss": 0.003, "step": 55459 }, { "epoch": 17.06, "learning_rate": 1.1136943672583643e-06, "loss": 0.0033, "step": 55460 }, { "epoch": 17.06, "learning_rate": 1.1134658598235925e-06, "loss": 0.0041, "step": 55461 }, { "epoch": 17.06, "learning_rate": 1.1132373744516878e-06, "loss": 0.0036, "step": 55462 }, { "epoch": 17.07, "learning_rate": 1.1130089111432197e-06, "loss": 0.0024, "step": 55463 }, { "epoch": 17.07, "learning_rate": 1.1127804698987544e-06, "loss": 0.0041, "step": 55464 }, { "epoch": 17.07, "learning_rate": 1.1125520507188591e-06, "loss": 0.0033, "step": 55465 }, { "epoch": 17.07, "learning_rate": 1.1123236536041016e-06, "loss": 0.0038, "step": 55466 }, { "epoch": 17.07, "learning_rate": 1.1120952785550477e-06, "loss": 0.003, "step": 55467 }, { "epoch": 17.07, "learning_rate": 1.1118669255722646e-06, "loss": 0.0045, "step": 55468 }, { "epoch": 17.07, "learning_rate": 1.1116385946563191e-06, "loss": 0.0049, "step": 55469 }, { "epoch": 17.07, "learning_rate": 1.1114102858077825e-06, "loss": 0.0015, "step": 55470 }, { "epoch": 17.07, "learning_rate": 1.1111819990272155e-06, "loss": 0.007, "step": 55471 }, { "epoch": 17.07, "learning_rate": 1.1109537343151877e-06, "loss": 0.0041, "step": 55472 }, { "epoch": 17.07, "learning_rate": 1.1107254916722643e-06, "loss": 0.0063, "step": 55473 }, { "epoch": 17.07, "learning_rate": 1.1104972710990126e-06, "loss": 0.0045, "step": 55474 }, { "epoch": 17.07, "learning_rate": 1.1102690725959997e-06, "loss": 0.0037, "step": 55475 }, { "epoch": 17.07, "learning_rate": 1.1100408961637932e-06, "loss": 0.0041, "step": 55476 }, { "epoch": 17.07, "learning_rate": 1.1098127418029592e-06, "loss": 0.0023, "step": 55477 }, { "epoch": 17.07, "learning_rate": 1.109584609514064e-06, "loss": 0.0023, "step": 55478 }, { "epoch": 17.07, "learning_rate": 1.1093564992976702e-06, "loss": 0.0063, "step": 55479 }, { "epoch": 17.07, "learning_rate": 1.1091284111543499e-06, "loss": 0.0031, "step": 55480 }, { "epoch": 17.07, "learning_rate": 1.1089003450846658e-06, "loss": 0.0024, "step": 55481 }, { "epoch": 17.07, "learning_rate": 1.1086723010891852e-06, "loss": 0.0087, "step": 55482 }, { "epoch": 17.07, "learning_rate": 1.1084442791684758e-06, "loss": 0.0049, "step": 55483 }, { "epoch": 17.07, "learning_rate": 1.108216279323101e-06, "loss": 0.006, "step": 55484 }, { "epoch": 17.07, "learning_rate": 1.1079883015536276e-06, "loss": 0.0029, "step": 55485 }, { "epoch": 17.07, "learning_rate": 1.1077603458606224e-06, "loss": 0.0055, "step": 55486 }, { "epoch": 17.07, "learning_rate": 1.1075324122446517e-06, "loss": 0.0026, "step": 55487 }, { "epoch": 17.07, "learning_rate": 1.10730450070628e-06, "loss": 0.0036, "step": 55488 }, { "epoch": 17.07, "learning_rate": 1.1070766112460719e-06, "loss": 0.0065, "step": 55489 }, { "epoch": 17.07, "learning_rate": 1.106848743864597e-06, "loss": 0.0124, "step": 55490 }, { "epoch": 17.07, "learning_rate": 1.1066208985624215e-06, "loss": 0.0045, "step": 55491 }, { "epoch": 17.07, "learning_rate": 1.1063930753401065e-06, "loss": 0.0057, "step": 55492 }, { "epoch": 17.07, "learning_rate": 1.1061652741982198e-06, "loss": 0.0037, "step": 55493 }, { "epoch": 17.07, "learning_rate": 1.1059374951373248e-06, "loss": 0.0039, "step": 55494 }, { "epoch": 17.08, "learning_rate": 1.1057097381579918e-06, "loss": 0.0033, "step": 55495 }, { "epoch": 17.08, "learning_rate": 1.1054820032607827e-06, "loss": 0.0024, "step": 55496 }, { "epoch": 17.08, "learning_rate": 1.1052542904462648e-06, "loss": 0.0033, "step": 55497 }, { "epoch": 17.08, "learning_rate": 1.1050265997150024e-06, "loss": 0.0052, "step": 55498 }, { "epoch": 17.08, "learning_rate": 1.104798931067561e-06, "loss": 0.0045, "step": 55499 }, { "epoch": 17.08, "learning_rate": 1.1045712845045053e-06, "loss": 0.0033, "step": 55500 }, { "epoch": 17.08, "learning_rate": 1.1043436600264013e-06, "loss": 0.003, "step": 55501 }, { "epoch": 17.08, "learning_rate": 1.1041160576338139e-06, "loss": 0.0045, "step": 55502 }, { "epoch": 17.08, "learning_rate": 1.1038884773273085e-06, "loss": 0.0029, "step": 55503 }, { "epoch": 17.08, "learning_rate": 1.1036609191074488e-06, "loss": 0.004, "step": 55504 }, { "epoch": 17.08, "learning_rate": 1.1034333829748012e-06, "loss": 0.0133, "step": 55505 }, { "epoch": 17.08, "learning_rate": 1.1032058689299297e-06, "loss": 0.0046, "step": 55506 }, { "epoch": 17.08, "learning_rate": 1.1029783769734003e-06, "loss": 0.0047, "step": 55507 }, { "epoch": 17.08, "learning_rate": 1.1027509071057762e-06, "loss": 0.0042, "step": 55508 }, { "epoch": 17.08, "learning_rate": 1.1025234593276212e-06, "loss": 0.0022, "step": 55509 }, { "epoch": 17.08, "learning_rate": 1.1022960336395039e-06, "loss": 0.0043, "step": 55510 }, { "epoch": 17.08, "learning_rate": 1.102068630041987e-06, "loss": 0.0043, "step": 55511 }, { "epoch": 17.08, "learning_rate": 1.1018412485356366e-06, "loss": 0.0032, "step": 55512 }, { "epoch": 17.08, "learning_rate": 1.1016138891210137e-06, "loss": 0.0067, "step": 55513 }, { "epoch": 17.08, "learning_rate": 1.1013865517986822e-06, "loss": 0.0057, "step": 55514 }, { "epoch": 17.08, "learning_rate": 1.1011592365692115e-06, "loss": 0.0046, "step": 55515 }, { "epoch": 17.08, "learning_rate": 1.1009319434331623e-06, "loss": 0.004, "step": 55516 }, { "epoch": 17.08, "learning_rate": 1.1007046723911009e-06, "loss": 0.0048, "step": 55517 }, { "epoch": 17.08, "learning_rate": 1.10047742344359e-06, "loss": 0.0054, "step": 55518 }, { "epoch": 17.08, "learning_rate": 1.100250196591195e-06, "loss": 0.005, "step": 55519 }, { "epoch": 17.08, "learning_rate": 1.1000229918344796e-06, "loss": 0.0017, "step": 55520 }, { "epoch": 17.08, "learning_rate": 1.0997958091740068e-06, "loss": 0.0076, "step": 55521 }, { "epoch": 17.08, "learning_rate": 1.099568648610343e-06, "loss": 0.0023, "step": 55522 }, { "epoch": 17.08, "learning_rate": 1.0993415101440496e-06, "loss": 0.0046, "step": 55523 }, { "epoch": 17.08, "learning_rate": 1.09911439377569e-06, "loss": 0.0032, "step": 55524 }, { "epoch": 17.08, "learning_rate": 1.0988872995058343e-06, "loss": 0.0054, "step": 55525 }, { "epoch": 17.08, "learning_rate": 1.0986602273350388e-06, "loss": 0.0025, "step": 55526 }, { "epoch": 17.08, "learning_rate": 1.0984331772638701e-06, "loss": 0.0044, "step": 55527 }, { "epoch": 17.09, "learning_rate": 1.098206149292893e-06, "loss": 0.0025, "step": 55528 }, { "epoch": 17.09, "learning_rate": 1.0979791434226683e-06, "loss": 0.0049, "step": 55529 }, { "epoch": 17.09, "learning_rate": 1.0977521596537621e-06, "loss": 0.0033, "step": 55530 }, { "epoch": 17.09, "learning_rate": 1.0975251979867384e-06, "loss": 0.0048, "step": 55531 }, { "epoch": 17.09, "learning_rate": 1.0972982584221592e-06, "loss": 0.0045, "step": 55532 }, { "epoch": 17.09, "learning_rate": 1.0970713409605905e-06, "loss": 0.0034, "step": 55533 }, { "epoch": 17.09, "learning_rate": 1.0968444456025896e-06, "loss": 0.0034, "step": 55534 }, { "epoch": 17.09, "learning_rate": 1.096617572348726e-06, "loss": 0.0047, "step": 55535 }, { "epoch": 17.09, "learning_rate": 1.0963907211995607e-06, "loss": 0.0063, "step": 55536 }, { "epoch": 17.09, "learning_rate": 1.0961638921556561e-06, "loss": 0.0041, "step": 55537 }, { "epoch": 17.09, "learning_rate": 1.0959370852175766e-06, "loss": 0.004, "step": 55538 }, { "epoch": 17.09, "learning_rate": 1.095710300385885e-06, "loss": 0.0019, "step": 55539 }, { "epoch": 17.09, "learning_rate": 1.095483537661144e-06, "loss": 0.0035, "step": 55540 }, { "epoch": 17.09, "learning_rate": 1.0952567970439165e-06, "loss": 0.0074, "step": 55541 }, { "epoch": 17.09, "learning_rate": 1.0950300785347668e-06, "loss": 0.0038, "step": 55542 }, { "epoch": 17.09, "learning_rate": 1.0948033821342552e-06, "loss": 0.0034, "step": 55543 }, { "epoch": 17.09, "learning_rate": 1.0945767078429447e-06, "loss": 0.0041, "step": 55544 }, { "epoch": 17.09, "learning_rate": 1.0943500556614017e-06, "loss": 0.0031, "step": 55545 }, { "epoch": 17.09, "learning_rate": 1.0941234255901888e-06, "loss": 0.0047, "step": 55546 }, { "epoch": 17.09, "learning_rate": 1.0938968176298637e-06, "loss": 0.0047, "step": 55547 }, { "epoch": 17.09, "learning_rate": 1.0936702317809922e-06, "loss": 0.0079, "step": 55548 }, { "epoch": 17.09, "learning_rate": 1.093443668044135e-06, "loss": 0.0019, "step": 55549 }, { "epoch": 17.09, "learning_rate": 1.0932171264198578e-06, "loss": 0.0048, "step": 55550 }, { "epoch": 17.09, "learning_rate": 1.0929906069087204e-06, "loss": 0.0071, "step": 55551 }, { "epoch": 17.09, "learning_rate": 1.0927641095112874e-06, "loss": 0.0049, "step": 55552 }, { "epoch": 17.09, "learning_rate": 1.0925376342281179e-06, "loss": 0.0055, "step": 55553 }, { "epoch": 17.09, "learning_rate": 1.0923111810597776e-06, "loss": 0.0033, "step": 55554 }, { "epoch": 17.09, "learning_rate": 1.0920847500068265e-06, "loss": 0.0037, "step": 55555 }, { "epoch": 17.09, "learning_rate": 1.091858341069828e-06, "loss": 0.003, "step": 55556 }, { "epoch": 17.09, "learning_rate": 1.0916319542493436e-06, "loss": 0.0052, "step": 55557 }, { "epoch": 17.09, "learning_rate": 1.0914055895459353e-06, "loss": 0.0063, "step": 55558 }, { "epoch": 17.09, "learning_rate": 1.0911792469601634e-06, "loss": 0.0088, "step": 55559 }, { "epoch": 17.1, "learning_rate": 1.0909529264925955e-06, "loss": 0.0034, "step": 55560 }, { "epoch": 17.1, "learning_rate": 1.0907266281437879e-06, "loss": 0.0029, "step": 55561 }, { "epoch": 17.1, "learning_rate": 1.0905003519143042e-06, "loss": 0.0079, "step": 55562 }, { "epoch": 17.1, "learning_rate": 1.0902740978047056e-06, "loss": 0.0047, "step": 55563 }, { "epoch": 17.1, "learning_rate": 1.0900478658155534e-06, "loss": 0.0072, "step": 55564 }, { "epoch": 17.1, "learning_rate": 1.0898216559474117e-06, "loss": 0.0051, "step": 55565 }, { "epoch": 17.1, "learning_rate": 1.0895954682008402e-06, "loss": 0.007, "step": 55566 }, { "epoch": 17.1, "learning_rate": 1.0893693025764041e-06, "loss": 0.0046, "step": 55567 }, { "epoch": 17.1, "learning_rate": 1.0891431590746592e-06, "loss": 0.004, "step": 55568 }, { "epoch": 17.1, "learning_rate": 1.0889170376961678e-06, "loss": 0.0068, "step": 55569 }, { "epoch": 17.1, "learning_rate": 1.0886909384414945e-06, "loss": 0.0035, "step": 55570 }, { "epoch": 17.1, "learning_rate": 1.0884648613111992e-06, "loss": 0.0042, "step": 55571 }, { "epoch": 17.1, "learning_rate": 1.0882388063058437e-06, "loss": 0.0035, "step": 55572 }, { "epoch": 17.1, "learning_rate": 1.0880127734259882e-06, "loss": 0.004, "step": 55573 }, { "epoch": 17.1, "learning_rate": 1.087786762672195e-06, "loss": 0.0038, "step": 55574 }, { "epoch": 17.1, "learning_rate": 1.0875607740450244e-06, "loss": 0.0045, "step": 55575 }, { "epoch": 17.1, "learning_rate": 1.0873348075450374e-06, "loss": 0.0049, "step": 55576 }, { "epoch": 17.1, "learning_rate": 1.0871088631727956e-06, "loss": 0.0038, "step": 55577 }, { "epoch": 17.1, "learning_rate": 1.0868829409288596e-06, "loss": 0.0065, "step": 55578 }, { "epoch": 17.1, "learning_rate": 1.0866570408137877e-06, "loss": 0.0049, "step": 55579 }, { "epoch": 17.1, "learning_rate": 1.0864311628281465e-06, "loss": 0.0031, "step": 55580 }, { "epoch": 17.1, "learning_rate": 1.0862053069724953e-06, "loss": 0.0029, "step": 55581 }, { "epoch": 17.1, "learning_rate": 1.0859794732473904e-06, "loss": 0.0055, "step": 55582 }, { "epoch": 17.1, "learning_rate": 1.085753661653396e-06, "loss": 0.0042, "step": 55583 }, { "epoch": 17.1, "learning_rate": 1.08552787219107e-06, "loss": 0.0033, "step": 55584 }, { "epoch": 17.1, "learning_rate": 1.0853021048609779e-06, "loss": 0.0035, "step": 55585 }, { "epoch": 17.1, "learning_rate": 1.085076359663676e-06, "loss": 0.0068, "step": 55586 }, { "epoch": 17.1, "learning_rate": 1.084850636599727e-06, "loss": 0.0029, "step": 55587 }, { "epoch": 17.1, "learning_rate": 1.0846249356696925e-06, "loss": 0.0033, "step": 55588 }, { "epoch": 17.1, "learning_rate": 1.0843992568741257e-06, "loss": 0.0038, "step": 55589 }, { "epoch": 17.1, "learning_rate": 1.084173600213595e-06, "loss": 0.0028, "step": 55590 }, { "epoch": 17.1, "learning_rate": 1.0839479656886576e-06, "loss": 0.0054, "step": 55591 }, { "epoch": 17.1, "learning_rate": 1.0837223532998731e-06, "loss": 0.0044, "step": 55592 }, { "epoch": 17.11, "learning_rate": 1.0834967630478022e-06, "loss": 0.002, "step": 55593 }, { "epoch": 17.11, "learning_rate": 1.0832711949330055e-06, "loss": 0.0049, "step": 55594 }, { "epoch": 17.11, "learning_rate": 1.0830456489560425e-06, "loss": 0.0051, "step": 55595 }, { "epoch": 17.11, "learning_rate": 1.082820125117473e-06, "loss": 0.0068, "step": 55596 }, { "epoch": 17.11, "learning_rate": 1.0825946234178575e-06, "loss": 0.0043, "step": 55597 }, { "epoch": 17.11, "learning_rate": 1.0823691438577545e-06, "loss": 0.0063, "step": 55598 }, { "epoch": 17.11, "learning_rate": 1.0821436864377232e-06, "loss": 0.0034, "step": 55599 }, { "epoch": 17.11, "learning_rate": 1.0819182511583282e-06, "loss": 0.0067, "step": 55600 }, { "epoch": 17.11, "learning_rate": 1.0816928380201243e-06, "loss": 0.0055, "step": 55601 }, { "epoch": 17.11, "learning_rate": 1.0814674470236753e-06, "loss": 0.0046, "step": 55602 }, { "epoch": 17.11, "learning_rate": 1.0812420781695365e-06, "loss": 0.0028, "step": 55603 }, { "epoch": 17.11, "learning_rate": 1.0810167314582676e-06, "loss": 0.0076, "step": 55604 }, { "epoch": 17.11, "learning_rate": 1.0807914068904314e-06, "loss": 0.0037, "step": 55605 }, { "epoch": 17.11, "learning_rate": 1.0805661044665849e-06, "loss": 0.0033, "step": 55606 }, { "epoch": 17.11, "learning_rate": 1.0803408241872893e-06, "loss": 0.0046, "step": 55607 }, { "epoch": 17.11, "learning_rate": 1.0801155660531015e-06, "loss": 0.0036, "step": 55608 }, { "epoch": 17.11, "learning_rate": 1.0798903300645823e-06, "loss": 0.0069, "step": 55609 }, { "epoch": 17.11, "learning_rate": 1.0796651162222916e-06, "loss": 0.0031, "step": 55610 }, { "epoch": 17.11, "learning_rate": 1.0794399245267873e-06, "loss": 0.0039, "step": 55611 }, { "epoch": 17.11, "learning_rate": 1.0792147549786281e-06, "loss": 0.0034, "step": 55612 }, { "epoch": 17.11, "learning_rate": 1.0789896075783734e-06, "loss": 0.0065, "step": 55613 }, { "epoch": 17.11, "learning_rate": 1.0787644823265819e-06, "loss": 0.013, "step": 55614 }, { "epoch": 17.11, "learning_rate": 1.0785393792238163e-06, "loss": 0.0048, "step": 55615 }, { "epoch": 17.11, "learning_rate": 1.0783142982706297e-06, "loss": 0.003, "step": 55616 }, { "epoch": 17.11, "learning_rate": 1.0780892394675846e-06, "loss": 0.002, "step": 55617 }, { "epoch": 17.11, "learning_rate": 1.0778642028152375e-06, "loss": 0.003, "step": 55618 }, { "epoch": 17.11, "learning_rate": 1.0776391883141469e-06, "loss": 0.0034, "step": 55619 }, { "epoch": 17.11, "learning_rate": 1.0774141959648743e-06, "loss": 0.0049, "step": 55620 }, { "epoch": 17.11, "learning_rate": 1.0771892257679773e-06, "loss": 0.0031, "step": 55621 }, { "epoch": 17.11, "learning_rate": 1.076964277724014e-06, "loss": 0.0034, "step": 55622 }, { "epoch": 17.11, "learning_rate": 1.0767393518335445e-06, "loss": 0.0038, "step": 55623 }, { "epoch": 17.11, "learning_rate": 1.0765144480971224e-06, "loss": 0.0047, "step": 55624 }, { "epoch": 17.12, "learning_rate": 1.0762895665153105e-06, "loss": 0.0015, "step": 55625 }, { "epoch": 17.12, "learning_rate": 1.0760647070886654e-06, "loss": 0.0054, "step": 55626 }, { "epoch": 17.12, "learning_rate": 1.0758398698177464e-06, "loss": 0.0039, "step": 55627 }, { "epoch": 17.12, "learning_rate": 1.075615054703112e-06, "loss": 0.0064, "step": 55628 }, { "epoch": 17.12, "learning_rate": 1.0753902617453183e-06, "loss": 0.0019, "step": 55629 }, { "epoch": 17.12, "learning_rate": 1.075165490944925e-06, "loss": 0.0067, "step": 55630 }, { "epoch": 17.12, "learning_rate": 1.0749407423024893e-06, "loss": 0.0049, "step": 55631 }, { "epoch": 17.12, "learning_rate": 1.07471601581857e-06, "loss": 0.0047, "step": 55632 }, { "epoch": 17.12, "learning_rate": 1.0744913114937238e-06, "loss": 0.0034, "step": 55633 }, { "epoch": 17.12, "learning_rate": 1.0742666293285086e-06, "loss": 0.0035, "step": 55634 }, { "epoch": 17.12, "learning_rate": 1.0740419693234839e-06, "loss": 0.0059, "step": 55635 }, { "epoch": 17.12, "learning_rate": 1.07381733147921e-06, "loss": 0.0058, "step": 55636 }, { "epoch": 17.12, "learning_rate": 1.0735927157962378e-06, "loss": 0.0041, "step": 55637 }, { "epoch": 17.12, "learning_rate": 1.0733681222751292e-06, "loss": 0.0021, "step": 55638 }, { "epoch": 17.12, "learning_rate": 1.073143550916439e-06, "loss": 0.0051, "step": 55639 }, { "epoch": 17.12, "learning_rate": 1.0729190017207292e-06, "loss": 0.0039, "step": 55640 }, { "epoch": 17.12, "learning_rate": 1.0726944746885537e-06, "loss": 0.0054, "step": 55641 }, { "epoch": 17.12, "learning_rate": 1.0724699698204722e-06, "loss": 0.0032, "step": 55642 }, { "epoch": 17.12, "learning_rate": 1.0722454871170428e-06, "loss": 0.0072, "step": 55643 }, { "epoch": 17.12, "learning_rate": 1.0720210265788167e-06, "loss": 0.0039, "step": 55644 }, { "epoch": 17.12, "learning_rate": 1.0717965882063575e-06, "loss": 0.0023, "step": 55645 }, { "epoch": 17.12, "learning_rate": 1.0715721720002203e-06, "loss": 0.0071, "step": 55646 }, { "epoch": 17.12, "learning_rate": 1.0713477779609627e-06, "loss": 0.0052, "step": 55647 }, { "epoch": 17.12, "learning_rate": 1.0711234060891407e-06, "loss": 0.0071, "step": 55648 }, { "epoch": 17.12, "learning_rate": 1.0708990563853127e-06, "loss": 0.0035, "step": 55649 }, { "epoch": 17.12, "learning_rate": 1.0706747288500352e-06, "loss": 0.0061, "step": 55650 }, { "epoch": 17.12, "learning_rate": 1.0704504234838652e-06, "loss": 0.0044, "step": 55651 }, { "epoch": 17.12, "learning_rate": 1.0702261402873592e-06, "loss": 0.0065, "step": 55652 }, { "epoch": 17.12, "learning_rate": 1.0700018792610745e-06, "loss": 0.0063, "step": 55653 }, { "epoch": 17.12, "learning_rate": 1.069777640405566e-06, "loss": 0.0051, "step": 55654 }, { "epoch": 17.12, "learning_rate": 1.0695534237213934e-06, "loss": 0.0096, "step": 55655 }, { "epoch": 17.12, "learning_rate": 1.0693292292091118e-06, "loss": 0.0044, "step": 55656 }, { "epoch": 17.12, "learning_rate": 1.0691050568692807e-06, "loss": 0.0039, "step": 55657 }, { "epoch": 17.13, "learning_rate": 1.0688809067024519e-06, "loss": 0.0042, "step": 55658 }, { "epoch": 17.13, "learning_rate": 1.0686567787091827e-06, "loss": 0.004, "step": 55659 }, { "epoch": 17.13, "learning_rate": 1.0684326728900319e-06, "loss": 0.005, "step": 55660 }, { "epoch": 17.13, "learning_rate": 1.0682085892455563e-06, "loss": 0.0057, "step": 55661 }, { "epoch": 17.13, "learning_rate": 1.067984527776309e-06, "loss": 0.0046, "step": 55662 }, { "epoch": 17.13, "learning_rate": 1.0677604884828496e-06, "loss": 0.0025, "step": 55663 }, { "epoch": 17.13, "learning_rate": 1.067536471365732e-06, "loss": 0.004, "step": 55664 }, { "epoch": 17.13, "learning_rate": 1.0673124764255139e-06, "loss": 0.0038, "step": 55665 }, { "epoch": 17.13, "learning_rate": 1.06708850366275e-06, "loss": 0.0051, "step": 55666 }, { "epoch": 17.13, "learning_rate": 1.0668645530779975e-06, "loss": 0.0029, "step": 55667 }, { "epoch": 17.13, "learning_rate": 1.066640624671812e-06, "loss": 0.0032, "step": 55668 }, { "epoch": 17.13, "learning_rate": 1.066416718444747e-06, "loss": 0.0048, "step": 55669 }, { "epoch": 17.13, "learning_rate": 1.0661928343973649e-06, "loss": 0.0044, "step": 55670 }, { "epoch": 17.13, "learning_rate": 1.0659689725302159e-06, "loss": 0.0049, "step": 55671 }, { "epoch": 17.13, "learning_rate": 1.0657451328438573e-06, "loss": 0.0047, "step": 55672 }, { "epoch": 17.13, "learning_rate": 1.0655213153388455e-06, "loss": 0.0022, "step": 55673 }, { "epoch": 17.13, "learning_rate": 1.0652975200157333e-06, "loss": 0.0037, "step": 55674 }, { "epoch": 17.13, "learning_rate": 1.0650737468750806e-06, "loss": 0.0051, "step": 55675 }, { "epoch": 17.13, "learning_rate": 1.0648499959174408e-06, "loss": 0.0045, "step": 55676 }, { "epoch": 17.13, "learning_rate": 1.0646262671433695e-06, "loss": 0.0045, "step": 55677 }, { "epoch": 17.13, "learning_rate": 1.0644025605534247e-06, "loss": 0.0058, "step": 55678 }, { "epoch": 17.13, "learning_rate": 1.0641788761481553e-06, "loss": 0.0041, "step": 55679 }, { "epoch": 17.13, "learning_rate": 1.0639552139281217e-06, "loss": 0.0038, "step": 55680 }, { "epoch": 17.13, "learning_rate": 1.0637315738938791e-06, "loss": 0.005, "step": 55681 }, { "epoch": 17.13, "learning_rate": 1.0635079560459826e-06, "loss": 0.0042, "step": 55682 }, { "epoch": 17.13, "learning_rate": 1.0632843603849852e-06, "loss": 0.0098, "step": 55683 }, { "epoch": 17.13, "learning_rate": 1.063060786911444e-06, "loss": 0.0038, "step": 55684 }, { "epoch": 17.13, "learning_rate": 1.0628372356259142e-06, "loss": 0.006, "step": 55685 }, { "epoch": 17.13, "learning_rate": 1.06261370652895e-06, "loss": 0.0022, "step": 55686 }, { "epoch": 17.13, "learning_rate": 1.062390199621106e-06, "loss": 0.006, "step": 55687 }, { "epoch": 17.13, "learning_rate": 1.062166714902938e-06, "loss": 0.006, "step": 55688 }, { "epoch": 17.13, "learning_rate": 1.0619432523749984e-06, "loss": 0.0043, "step": 55689 }, { "epoch": 17.14, "learning_rate": 1.0617198120378459e-06, "loss": 0.0029, "step": 55690 }, { "epoch": 17.14, "learning_rate": 1.0614963938920352e-06, "loss": 0.0037, "step": 55691 }, { "epoch": 17.14, "learning_rate": 1.0612729979381175e-06, "loss": 0.0032, "step": 55692 }, { "epoch": 17.14, "learning_rate": 1.0610496241766488e-06, "loss": 0.0025, "step": 55693 }, { "epoch": 17.14, "learning_rate": 1.060826272608183e-06, "loss": 0.0038, "step": 55694 }, { "epoch": 17.14, "learning_rate": 1.0606029432332765e-06, "loss": 0.0068, "step": 55695 }, { "epoch": 17.14, "learning_rate": 1.060379636052483e-06, "loss": 0.004, "step": 55696 }, { "epoch": 17.14, "learning_rate": 1.0601563510663582e-06, "loss": 0.005, "step": 55697 }, { "epoch": 17.14, "learning_rate": 1.0599330882754532e-06, "loss": 0.0042, "step": 55698 }, { "epoch": 17.14, "learning_rate": 1.0597098476803259e-06, "loss": 0.0047, "step": 55699 }, { "epoch": 17.14, "learning_rate": 1.0594866292815276e-06, "loss": 0.0044, "step": 55700 }, { "epoch": 17.14, "learning_rate": 1.0592634330796137e-06, "loss": 0.0044, "step": 55701 }, { "epoch": 17.14, "learning_rate": 1.0590402590751392e-06, "loss": 0.0031, "step": 55702 }, { "epoch": 17.14, "learning_rate": 1.058817107268657e-06, "loss": 0.0066, "step": 55703 }, { "epoch": 17.14, "learning_rate": 1.0585939776607213e-06, "loss": 0.0042, "step": 55704 }, { "epoch": 17.14, "learning_rate": 1.0583708702518857e-06, "loss": 0.0054, "step": 55705 }, { "epoch": 17.14, "learning_rate": 1.0581477850427058e-06, "loss": 0.0043, "step": 55706 }, { "epoch": 17.14, "learning_rate": 1.057924722033733e-06, "loss": 0.0027, "step": 55707 }, { "epoch": 17.14, "learning_rate": 1.0577016812255237e-06, "loss": 0.0024, "step": 55708 }, { "epoch": 17.14, "learning_rate": 1.0574786626186273e-06, "loss": 0.0073, "step": 55709 }, { "epoch": 17.14, "learning_rate": 1.0572556662136036e-06, "loss": 0.0028, "step": 55710 }, { "epoch": 17.14, "learning_rate": 1.057032692011002e-06, "loss": 0.0038, "step": 55711 }, { "epoch": 17.14, "learning_rate": 1.0568097400113796e-06, "loss": 0.0036, "step": 55712 }, { "epoch": 17.14, "learning_rate": 1.0565868102152865e-06, "loss": 0.0037, "step": 55713 }, { "epoch": 17.14, "learning_rate": 1.0563639026232742e-06, "loss": 0.0048, "step": 55714 }, { "epoch": 17.14, "learning_rate": 1.056141017235902e-06, "loss": 0.0039, "step": 55715 }, { "epoch": 17.14, "learning_rate": 1.05591815405372e-06, "loss": 0.006, "step": 55716 }, { "epoch": 17.14, "learning_rate": 1.0556953130772818e-06, "loss": 0.0071, "step": 55717 }, { "epoch": 17.14, "learning_rate": 1.0554724943071414e-06, "loss": 0.0065, "step": 55718 }, { "epoch": 17.14, "learning_rate": 1.0552496977438508e-06, "loss": 0.0032, "step": 55719 }, { "epoch": 17.14, "learning_rate": 1.055026923387965e-06, "loss": 0.002, "step": 55720 }, { "epoch": 17.14, "learning_rate": 1.0548041712400348e-06, "loss": 0.0047, "step": 55721 }, { "epoch": 17.14, "learning_rate": 1.0545814413006138e-06, "loss": 0.0039, "step": 55722 }, { "epoch": 17.15, "learning_rate": 1.0543587335702566e-06, "loss": 0.0033, "step": 55723 }, { "epoch": 17.15, "learning_rate": 1.0541360480495121e-06, "loss": 0.004, "step": 55724 }, { "epoch": 17.15, "learning_rate": 1.0539133847389404e-06, "loss": 0.0034, "step": 55725 }, { "epoch": 17.15, "learning_rate": 1.0536907436390886e-06, "loss": 0.0064, "step": 55726 }, { "epoch": 17.15, "learning_rate": 1.0534681247505107e-06, "loss": 0.0031, "step": 55727 }, { "epoch": 17.15, "learning_rate": 1.0532455280737585e-06, "loss": 0.0045, "step": 55728 }, { "epoch": 17.15, "learning_rate": 1.053022953609385e-06, "loss": 0.0052, "step": 55729 }, { "epoch": 17.15, "learning_rate": 1.052800401357944e-06, "loss": 0.0032, "step": 55730 }, { "epoch": 17.15, "learning_rate": 1.0525778713199885e-06, "loss": 0.0054, "step": 55731 }, { "epoch": 17.15, "learning_rate": 1.0523553634960704e-06, "loss": 0.0039, "step": 55732 }, { "epoch": 17.15, "learning_rate": 1.0521328778867423e-06, "loss": 0.0051, "step": 55733 }, { "epoch": 17.15, "learning_rate": 1.051910414492553e-06, "loss": 0.0058, "step": 55734 }, { "epoch": 17.15, "learning_rate": 1.0516879733140594e-06, "loss": 0.0021, "step": 55735 }, { "epoch": 17.15, "learning_rate": 1.0514655543518126e-06, "loss": 0.0047, "step": 55736 }, { "epoch": 17.15, "learning_rate": 1.0512431576063642e-06, "loss": 0.0024, "step": 55737 }, { "epoch": 17.15, "learning_rate": 1.051020783078267e-06, "loss": 0.0066, "step": 55738 }, { "epoch": 17.15, "learning_rate": 1.0507984307680719e-06, "loss": 0.005, "step": 55739 }, { "epoch": 17.15, "learning_rate": 1.0505761006763315e-06, "loss": 0.0057, "step": 55740 }, { "epoch": 17.15, "learning_rate": 1.050353792803599e-06, "loss": 0.0053, "step": 55741 }, { "epoch": 17.15, "learning_rate": 1.0501315071504247e-06, "loss": 0.0041, "step": 55742 }, { "epoch": 17.15, "learning_rate": 1.0499092437173607e-06, "loss": 0.0053, "step": 55743 }, { "epoch": 17.15, "learning_rate": 1.0496870025049578e-06, "loss": 0.0052, "step": 55744 }, { "epoch": 17.15, "learning_rate": 1.0494647835137705e-06, "loss": 0.0037, "step": 55745 }, { "epoch": 17.15, "learning_rate": 1.0492425867443511e-06, "loss": 0.0027, "step": 55746 }, { "epoch": 17.15, "learning_rate": 1.049020412197248e-06, "loss": 0.0043, "step": 55747 }, { "epoch": 17.15, "learning_rate": 1.048798259873014e-06, "loss": 0.0049, "step": 55748 }, { "epoch": 17.15, "learning_rate": 1.0485761297721986e-06, "loss": 0.0047, "step": 55749 }, { "epoch": 17.15, "learning_rate": 1.048354021895357e-06, "loss": 0.0049, "step": 55750 }, { "epoch": 17.15, "learning_rate": 1.0481319362430398e-06, "loss": 0.0048, "step": 55751 }, { "epoch": 17.15, "learning_rate": 1.0479098728157977e-06, "loss": 0.0053, "step": 55752 }, { "epoch": 17.15, "learning_rate": 1.0476878316141814e-06, "loss": 0.0029, "step": 55753 }, { "epoch": 17.15, "learning_rate": 1.0474658126387427e-06, "loss": 0.0038, "step": 55754 }, { "epoch": 17.16, "learning_rate": 1.0472438158900333e-06, "loss": 0.0036, "step": 55755 }, { "epoch": 17.16, "learning_rate": 1.047021841368604e-06, "loss": 0.0059, "step": 55756 }, { "epoch": 17.16, "learning_rate": 1.0467998890750052e-06, "loss": 0.0052, "step": 55757 }, { "epoch": 17.16, "learning_rate": 1.046577959009789e-06, "loss": 0.0041, "step": 55758 }, { "epoch": 17.16, "learning_rate": 1.0463560511735039e-06, "loss": 0.0062, "step": 55759 }, { "epoch": 17.16, "learning_rate": 1.0461341655667067e-06, "loss": 0.0046, "step": 55760 }, { "epoch": 17.16, "learning_rate": 1.045912302189942e-06, "loss": 0.0045, "step": 55761 }, { "epoch": 17.16, "learning_rate": 1.0456904610437645e-06, "loss": 0.0029, "step": 55762 }, { "epoch": 17.16, "learning_rate": 1.0454686421287208e-06, "loss": 0.0016, "step": 55763 }, { "epoch": 17.16, "learning_rate": 1.0452468454453656e-06, "loss": 0.0047, "step": 55764 }, { "epoch": 17.16, "learning_rate": 1.04502507099425e-06, "loss": 0.0021, "step": 55765 }, { "epoch": 17.16, "learning_rate": 1.044803318775922e-06, "loss": 0.0037, "step": 55766 }, { "epoch": 17.16, "learning_rate": 1.044581588790935e-06, "loss": 0.004, "step": 55767 }, { "epoch": 17.16, "learning_rate": 1.0443598810398336e-06, "loss": 0.002, "step": 55768 }, { "epoch": 17.16, "learning_rate": 1.0441381955231744e-06, "loss": 0.0025, "step": 55769 }, { "epoch": 17.16, "learning_rate": 1.0439165322415057e-06, "loss": 0.0057, "step": 55770 }, { "epoch": 17.16, "learning_rate": 1.0436948911953782e-06, "loss": 0.0057, "step": 55771 }, { "epoch": 17.16, "learning_rate": 1.0434732723853414e-06, "loss": 0.0095, "step": 55772 }, { "epoch": 17.16, "learning_rate": 1.043251675811946e-06, "loss": 0.0058, "step": 55773 }, { "epoch": 17.16, "learning_rate": 1.0430301014757416e-06, "loss": 0.0035, "step": 55774 }, { "epoch": 17.16, "learning_rate": 1.0428085493772788e-06, "loss": 0.0031, "step": 55775 }, { "epoch": 17.16, "learning_rate": 1.0425870195171073e-06, "loss": 0.0032, "step": 55776 }, { "epoch": 17.16, "learning_rate": 1.0423655118957787e-06, "loss": 0.004, "step": 55777 }, { "epoch": 17.16, "learning_rate": 1.0421440265138383e-06, "loss": 0.0054, "step": 55778 }, { "epoch": 17.16, "learning_rate": 1.041922563371842e-06, "loss": 0.0047, "step": 55779 }, { "epoch": 17.16, "learning_rate": 1.041701122470339e-06, "loss": 0.0027, "step": 55780 }, { "epoch": 17.16, "learning_rate": 1.0414797038098745e-06, "loss": 0.0066, "step": 55781 }, { "epoch": 17.16, "learning_rate": 1.041258307391001e-06, "loss": 0.0063, "step": 55782 }, { "epoch": 17.16, "learning_rate": 1.0410369332142656e-06, "loss": 0.004, "step": 55783 }, { "epoch": 17.16, "learning_rate": 1.0408155812802224e-06, "loss": 0.0064, "step": 55784 }, { "epoch": 17.16, "learning_rate": 1.0405942515894196e-06, "loss": 0.0057, "step": 55785 }, { "epoch": 17.16, "learning_rate": 1.0403729441424048e-06, "loss": 0.0056, "step": 55786 }, { "epoch": 17.16, "learning_rate": 1.0401516589397286e-06, "loss": 0.0048, "step": 55787 }, { "epoch": 17.17, "learning_rate": 1.0399303959819408e-06, "loss": 0.0056, "step": 55788 }, { "epoch": 17.17, "learning_rate": 1.0397091552695903e-06, "loss": 0.0044, "step": 55789 }, { "epoch": 17.17, "learning_rate": 1.0394879368032263e-06, "loss": 0.0022, "step": 55790 }, { "epoch": 17.17, "learning_rate": 1.0392667405833977e-06, "loss": 0.0037, "step": 55791 }, { "epoch": 17.17, "learning_rate": 1.0390455666106547e-06, "loss": 0.0049, "step": 55792 }, { "epoch": 17.17, "learning_rate": 1.038824414885543e-06, "loss": 0.0034, "step": 55793 }, { "epoch": 17.17, "learning_rate": 1.0386032854086191e-06, "loss": 0.0032, "step": 55794 }, { "epoch": 17.17, "learning_rate": 1.0383821781804248e-06, "loss": 0.0062, "step": 55795 }, { "epoch": 17.17, "learning_rate": 1.0381610932015118e-06, "loss": 0.0064, "step": 55796 }, { "epoch": 17.17, "learning_rate": 1.0379400304724297e-06, "loss": 0.003, "step": 55797 }, { "epoch": 17.17, "learning_rate": 1.0377189899937234e-06, "loss": 0.0028, "step": 55798 }, { "epoch": 17.17, "learning_rate": 1.0374979717659462e-06, "loss": 0.0037, "step": 55799 }, { "epoch": 17.17, "learning_rate": 1.037276975789646e-06, "loss": 0.0046, "step": 55800 }, { "epoch": 17.17, "learning_rate": 1.037056002065373e-06, "loss": 0.0041, "step": 55801 }, { "epoch": 17.17, "learning_rate": 1.0368350505936708e-06, "loss": 0.0057, "step": 55802 }, { "epoch": 17.17, "learning_rate": 1.0366141213750902e-06, "loss": 0.0037, "step": 55803 }, { "epoch": 17.17, "learning_rate": 1.0363932144101807e-06, "loss": 0.0045, "step": 55804 }, { "epoch": 17.17, "learning_rate": 1.0361723296994908e-06, "loss": 0.0041, "step": 55805 }, { "epoch": 17.17, "learning_rate": 1.035951467243569e-06, "loss": 0.0052, "step": 55806 }, { "epoch": 17.17, "learning_rate": 1.0357306270429623e-06, "loss": 0.0031, "step": 55807 }, { "epoch": 17.17, "learning_rate": 1.0355098090982207e-06, "loss": 0.0031, "step": 55808 }, { "epoch": 17.17, "learning_rate": 1.0352890134098914e-06, "loss": 0.0073, "step": 55809 }, { "epoch": 17.17, "learning_rate": 1.0350682399785216e-06, "loss": 0.0055, "step": 55810 }, { "epoch": 17.17, "learning_rate": 1.034847488804661e-06, "loss": 0.002, "step": 55811 }, { "epoch": 17.17, "learning_rate": 1.034626759888857e-06, "loss": 0.0032, "step": 55812 }, { "epoch": 17.17, "learning_rate": 1.034406053231657e-06, "loss": 0.0048, "step": 55813 }, { "epoch": 17.17, "learning_rate": 1.03418536883361e-06, "loss": 0.0065, "step": 55814 }, { "epoch": 17.17, "learning_rate": 1.0339647066952663e-06, "loss": 0.0032, "step": 55815 }, { "epoch": 17.17, "learning_rate": 1.0337440668171694e-06, "loss": 0.0041, "step": 55816 }, { "epoch": 17.17, "learning_rate": 1.033523449199869e-06, "loss": 0.0051, "step": 55817 }, { "epoch": 17.17, "learning_rate": 1.0333028538439093e-06, "loss": 0.0041, "step": 55818 }, { "epoch": 17.17, "learning_rate": 1.0330822807498442e-06, "loss": 0.0029, "step": 55819 }, { "epoch": 17.18, "learning_rate": 1.0328617299182187e-06, "loss": 0.0041, "step": 55820 }, { "epoch": 17.18, "learning_rate": 1.0326412013495802e-06, "loss": 0.0038, "step": 55821 }, { "epoch": 17.18, "learning_rate": 1.0324206950444771e-06, "loss": 0.0057, "step": 55822 }, { "epoch": 17.18, "learning_rate": 1.0322002110034524e-06, "loss": 0.003, "step": 55823 }, { "epoch": 17.18, "learning_rate": 1.0319797492270577e-06, "loss": 0.006, "step": 55824 }, { "epoch": 17.18, "learning_rate": 1.0317593097158408e-06, "loss": 0.0039, "step": 55825 }, { "epoch": 17.18, "learning_rate": 1.0315388924703485e-06, "loss": 0.0055, "step": 55826 }, { "epoch": 17.18, "learning_rate": 1.031318497491126e-06, "loss": 0.0037, "step": 55827 }, { "epoch": 17.18, "learning_rate": 1.0310981247787222e-06, "loss": 0.0039, "step": 55828 }, { "epoch": 17.18, "learning_rate": 1.030877774333684e-06, "loss": 0.0054, "step": 55829 }, { "epoch": 17.18, "learning_rate": 1.0306574461565577e-06, "loss": 0.0075, "step": 55830 }, { "epoch": 17.18, "learning_rate": 1.0304371402478918e-06, "loss": 0.0048, "step": 55831 }, { "epoch": 17.18, "learning_rate": 1.0302168566082315e-06, "loss": 0.0063, "step": 55832 }, { "epoch": 17.18, "learning_rate": 1.0299965952381231e-06, "loss": 0.0058, "step": 55833 }, { "epoch": 17.18, "learning_rate": 1.0297763561381168e-06, "loss": 0.0019, "step": 55834 }, { "epoch": 17.18, "learning_rate": 1.0295561393087572e-06, "loss": 0.0045, "step": 55835 }, { "epoch": 17.18, "learning_rate": 1.0293359447505936e-06, "loss": 0.0034, "step": 55836 }, { "epoch": 17.18, "learning_rate": 1.0291157724641686e-06, "loss": 0.0033, "step": 55837 }, { "epoch": 17.18, "learning_rate": 1.028895622450028e-06, "loss": 0.0052, "step": 55838 }, { "epoch": 17.18, "learning_rate": 1.0286754947087242e-06, "loss": 0.0034, "step": 55839 }, { "epoch": 17.18, "learning_rate": 1.0284553892407988e-06, "loss": 0.0057, "step": 55840 }, { "epoch": 17.18, "learning_rate": 1.028235306046802e-06, "loss": 0.0021, "step": 55841 }, { "epoch": 17.18, "learning_rate": 1.0280152451272774e-06, "loss": 0.005, "step": 55842 }, { "epoch": 17.18, "learning_rate": 1.0277952064827712e-06, "loss": 0.0039, "step": 55843 }, { "epoch": 17.18, "learning_rate": 1.027575190113832e-06, "loss": 0.0063, "step": 55844 }, { "epoch": 17.18, "learning_rate": 1.027355196021005e-06, "loss": 0.0037, "step": 55845 }, { "epoch": 17.18, "learning_rate": 1.027135224204835e-06, "loss": 0.0035, "step": 55846 }, { "epoch": 17.18, "learning_rate": 1.0269152746658695e-06, "loss": 0.0046, "step": 55847 }, { "epoch": 17.18, "learning_rate": 1.0266953474046526e-06, "loss": 0.0044, "step": 55848 }, { "epoch": 17.18, "learning_rate": 1.0264754424217348e-06, "loss": 0.0033, "step": 55849 }, { "epoch": 17.18, "learning_rate": 1.0262555597176593e-06, "loss": 0.0069, "step": 55850 }, { "epoch": 17.18, "learning_rate": 1.026035699292971e-06, "loss": 0.003, "step": 55851 }, { "epoch": 17.18, "learning_rate": 1.025815861148216e-06, "loss": 0.0055, "step": 55852 }, { "epoch": 17.19, "learning_rate": 1.0255960452839398e-06, "loss": 0.0035, "step": 55853 }, { "epoch": 17.19, "learning_rate": 1.0253762517006916e-06, "loss": 0.0027, "step": 55854 }, { "epoch": 17.19, "learning_rate": 1.0251564803990133e-06, "loss": 0.0042, "step": 55855 }, { "epoch": 17.19, "learning_rate": 1.0249367313794535e-06, "loss": 0.0027, "step": 55856 }, { "epoch": 17.19, "learning_rate": 1.024717004642557e-06, "loss": 0.0038, "step": 55857 }, { "epoch": 17.19, "learning_rate": 1.0244973001888647e-06, "loss": 0.0045, "step": 55858 }, { "epoch": 17.19, "learning_rate": 1.0242776180189284e-06, "loss": 0.0037, "step": 55859 }, { "epoch": 17.19, "learning_rate": 1.024057958133291e-06, "loss": 0.0025, "step": 55860 }, { "epoch": 17.19, "learning_rate": 1.0238383205324976e-06, "loss": 0.0032, "step": 55861 }, { "epoch": 17.19, "learning_rate": 1.023618705217093e-06, "loss": 0.0034, "step": 55862 }, { "epoch": 17.19, "learning_rate": 1.0233991121876241e-06, "loss": 0.0026, "step": 55863 }, { "epoch": 17.19, "learning_rate": 1.0231795414446356e-06, "loss": 0.0038, "step": 55864 }, { "epoch": 17.19, "learning_rate": 1.0229599929886714e-06, "loss": 0.0041, "step": 55865 }, { "epoch": 17.19, "learning_rate": 1.022740466820278e-06, "loss": 0.0037, "step": 55866 }, { "epoch": 17.19, "learning_rate": 1.022520962939999e-06, "loss": 0.0026, "step": 55867 }, { "epoch": 17.19, "learning_rate": 1.0223014813483788e-06, "loss": 0.0054, "step": 55868 }, { "epoch": 17.19, "learning_rate": 1.0220820220459648e-06, "loss": 0.004, "step": 55869 }, { "epoch": 17.19, "learning_rate": 1.021862585033304e-06, "loss": 0.004, "step": 55870 }, { "epoch": 17.19, "learning_rate": 1.0216431703109352e-06, "loss": 0.0068, "step": 55871 }, { "epoch": 17.19, "learning_rate": 1.0214237778794068e-06, "loss": 0.0049, "step": 55872 }, { "epoch": 17.19, "learning_rate": 1.021204407739259e-06, "loss": 0.004, "step": 55873 }, { "epoch": 17.19, "learning_rate": 1.020985059891043e-06, "loss": 0.0049, "step": 55874 }, { "epoch": 17.19, "learning_rate": 1.0207657343353006e-06, "loss": 0.0034, "step": 55875 }, { "epoch": 17.19, "learning_rate": 1.0205464310725765e-06, "loss": 0.005, "step": 55876 }, { "epoch": 17.19, "learning_rate": 1.020327150103414e-06, "loss": 0.0026, "step": 55877 }, { "epoch": 17.19, "learning_rate": 1.020107891428359e-06, "loss": 0.0056, "step": 55878 }, { "epoch": 17.19, "learning_rate": 1.0198886550479548e-06, "loss": 0.0038, "step": 55879 }, { "epoch": 17.19, "learning_rate": 1.0196694409627462e-06, "loss": 0.0048, "step": 55880 }, { "epoch": 17.19, "learning_rate": 1.019450249173277e-06, "loss": 0.0085, "step": 55881 }, { "epoch": 17.19, "learning_rate": 1.019231079680093e-06, "loss": 0.0053, "step": 55882 }, { "epoch": 17.19, "learning_rate": 1.0190119324837356e-06, "loss": 0.0056, "step": 55883 }, { "epoch": 17.19, "learning_rate": 1.0187928075847508e-06, "loss": 0.0034, "step": 55884 }, { "epoch": 17.2, "learning_rate": 1.018573704983683e-06, "loss": 0.0028, "step": 55885 }, { "epoch": 17.2, "learning_rate": 1.0183546246810748e-06, "loss": 0.0046, "step": 55886 }, { "epoch": 17.2, "learning_rate": 1.0181355666774706e-06, "loss": 0.0021, "step": 55887 }, { "epoch": 17.2, "learning_rate": 1.0179165309734118e-06, "loss": 0.0041, "step": 55888 }, { "epoch": 17.2, "learning_rate": 1.0176975175694482e-06, "loss": 0.0046, "step": 55889 }, { "epoch": 17.2, "learning_rate": 1.017478526466119e-06, "loss": 0.0027, "step": 55890 }, { "epoch": 17.2, "learning_rate": 1.017259557663971e-06, "loss": 0.0022, "step": 55891 }, { "epoch": 17.2, "learning_rate": 1.0170406111635445e-06, "loss": 0.0056, "step": 55892 }, { "epoch": 17.2, "learning_rate": 1.0168216869653813e-06, "loss": 0.0026, "step": 55893 }, { "epoch": 17.2, "learning_rate": 1.0166027850700312e-06, "loss": 0.0061, "step": 55894 }, { "epoch": 17.2, "learning_rate": 1.0163839054780343e-06, "loss": 0.0046, "step": 55895 }, { "epoch": 17.2, "learning_rate": 1.0161650481899344e-06, "loss": 0.0053, "step": 55896 }, { "epoch": 17.2, "learning_rate": 1.0159462132062736e-06, "loss": 0.0042, "step": 55897 }, { "epoch": 17.2, "learning_rate": 1.0157274005275975e-06, "loss": 0.0035, "step": 55898 }, { "epoch": 17.2, "learning_rate": 1.0155086101544475e-06, "loss": 0.0043, "step": 55899 }, { "epoch": 17.2, "learning_rate": 1.015289842087368e-06, "loss": 0.0039, "step": 55900 }, { "epoch": 17.2, "learning_rate": 1.0150710963269018e-06, "loss": 0.0056, "step": 55901 }, { "epoch": 17.2, "learning_rate": 1.0148523728735915e-06, "loss": 0.0037, "step": 55902 }, { "epoch": 17.2, "learning_rate": 1.014633671727978e-06, "loss": 0.0047, "step": 55903 }, { "epoch": 17.2, "learning_rate": 1.014414992890611e-06, "loss": 0.0029, "step": 55904 }, { "epoch": 17.2, "learning_rate": 1.0141963363620278e-06, "loss": 0.01, "step": 55905 }, { "epoch": 17.2, "learning_rate": 1.013977702142771e-06, "loss": 0.0036, "step": 55906 }, { "epoch": 17.2, "learning_rate": 1.0137590902333861e-06, "loss": 0.0028, "step": 55907 }, { "epoch": 17.2, "learning_rate": 1.0135405006344134e-06, "loss": 0.0034, "step": 55908 }, { "epoch": 17.2, "learning_rate": 1.0133219333463983e-06, "loss": 0.0045, "step": 55909 }, { "epoch": 17.2, "learning_rate": 1.0131033883698816e-06, "loss": 0.0049, "step": 55910 }, { "epoch": 17.2, "learning_rate": 1.0128848657054068e-06, "loss": 0.0032, "step": 55911 }, { "epoch": 17.2, "learning_rate": 1.0126663653535184e-06, "loss": 0.0014, "step": 55912 }, { "epoch": 17.2, "learning_rate": 1.0124478873147515e-06, "loss": 0.0064, "step": 55913 }, { "epoch": 17.2, "learning_rate": 1.0122294315896564e-06, "loss": 0.0051, "step": 55914 }, { "epoch": 17.2, "learning_rate": 1.012010998178773e-06, "loss": 0.0037, "step": 55915 }, { "epoch": 17.2, "learning_rate": 1.0117925870826417e-06, "loss": 0.0128, "step": 55916 }, { "epoch": 17.2, "learning_rate": 1.011574198301808e-06, "loss": 0.0056, "step": 55917 }, { "epoch": 17.21, "learning_rate": 1.011355831836811e-06, "loss": 0.003, "step": 55918 }, { "epoch": 17.21, "learning_rate": 1.0111374876881953e-06, "loss": 0.003, "step": 55919 }, { "epoch": 17.21, "learning_rate": 1.010919165856501e-06, "loss": 0.0028, "step": 55920 }, { "epoch": 17.21, "learning_rate": 1.0107008663422713e-06, "loss": 0.0031, "step": 55921 }, { "epoch": 17.21, "learning_rate": 1.010482589146048e-06, "loss": 0.0039, "step": 55922 }, { "epoch": 17.21, "learning_rate": 1.0102643342683704e-06, "loss": 0.0055, "step": 55923 }, { "epoch": 17.21, "learning_rate": 1.0100461017097862e-06, "loss": 0.0028, "step": 55924 }, { "epoch": 17.21, "learning_rate": 1.0098278914708348e-06, "loss": 0.006, "step": 55925 }, { "epoch": 17.21, "learning_rate": 1.0096097035520548e-06, "loss": 0.0077, "step": 55926 }, { "epoch": 17.21, "learning_rate": 1.009391537953991e-06, "loss": 0.0029, "step": 55927 }, { "epoch": 17.21, "learning_rate": 1.009173394677182e-06, "loss": 0.0035, "step": 55928 }, { "epoch": 17.21, "learning_rate": 1.0089552737221742e-06, "loss": 0.0041, "step": 55929 }, { "epoch": 17.21, "learning_rate": 1.0087371750895058e-06, "loss": 0.0038, "step": 55930 }, { "epoch": 17.21, "learning_rate": 1.0085190987797188e-06, "loss": 0.0055, "step": 55931 }, { "epoch": 17.21, "learning_rate": 1.008301044793355e-06, "loss": 0.004, "step": 55932 }, { "epoch": 17.21, "learning_rate": 1.008083013130956e-06, "loss": 0.0039, "step": 55933 }, { "epoch": 17.21, "learning_rate": 1.0078650037930627e-06, "loss": 0.0037, "step": 55934 }, { "epoch": 17.21, "learning_rate": 1.0076470167802156e-06, "loss": 0.006, "step": 55935 }, { "epoch": 17.21, "learning_rate": 1.0074290520929574e-06, "loss": 0.006, "step": 55936 }, { "epoch": 17.21, "learning_rate": 1.0072111097318282e-06, "loss": 0.0041, "step": 55937 }, { "epoch": 17.21, "learning_rate": 1.0069931896973705e-06, "loss": 0.0047, "step": 55938 }, { "epoch": 17.21, "learning_rate": 1.006775291990123e-06, "loss": 0.0067, "step": 55939 }, { "epoch": 17.21, "learning_rate": 1.0065574166106285e-06, "loss": 0.0028, "step": 55940 }, { "epoch": 17.21, "learning_rate": 1.0063395635594276e-06, "loss": 0.0018, "step": 55941 }, { "epoch": 17.21, "learning_rate": 1.0061217328370609e-06, "loss": 0.0055, "step": 55942 }, { "epoch": 17.21, "learning_rate": 1.0059039244440672e-06, "loss": 0.0042, "step": 55943 }, { "epoch": 17.21, "learning_rate": 1.0056861383809902e-06, "loss": 0.0058, "step": 55944 }, { "epoch": 17.21, "learning_rate": 1.0054683746483707e-06, "loss": 0.0063, "step": 55945 }, { "epoch": 17.21, "learning_rate": 1.0052506332467503e-06, "loss": 0.006, "step": 55946 }, { "epoch": 17.21, "learning_rate": 1.0050329141766657e-06, "loss": 0.0035, "step": 55947 }, { "epoch": 17.21, "learning_rate": 1.0048152174386584e-06, "loss": 0.0087, "step": 55948 }, { "epoch": 17.21, "learning_rate": 1.00459754303327e-06, "loss": 0.0049, "step": 55949 }, { "epoch": 17.22, "learning_rate": 1.0043798909610425e-06, "loss": 0.0028, "step": 55950 }, { "epoch": 17.22, "learning_rate": 1.0041622612225133e-06, "loss": 0.0055, "step": 55951 }, { "epoch": 17.22, "learning_rate": 1.003944653818224e-06, "loss": 0.003, "step": 55952 }, { "epoch": 17.22, "learning_rate": 1.0037270687487155e-06, "loss": 0.0054, "step": 55953 }, { "epoch": 17.22, "learning_rate": 1.0035095060145272e-06, "loss": 0.0036, "step": 55954 }, { "epoch": 17.22, "learning_rate": 1.0032919656162e-06, "loss": 0.0057, "step": 55955 }, { "epoch": 17.22, "learning_rate": 1.003074447554273e-06, "loss": 0.0047, "step": 55956 }, { "epoch": 17.22, "learning_rate": 1.0028569518292863e-06, "loss": 0.0086, "step": 55957 }, { "epoch": 17.22, "learning_rate": 1.002639478441778e-06, "loss": 0.0028, "step": 55958 }, { "epoch": 17.22, "learning_rate": 1.0024220273922947e-06, "loss": 0.004, "step": 55959 }, { "epoch": 17.22, "learning_rate": 1.0022045986813699e-06, "loss": 0.0038, "step": 55960 }, { "epoch": 17.22, "learning_rate": 1.0019871923095459e-06, "loss": 0.0247, "step": 55961 }, { "epoch": 17.22, "learning_rate": 1.0017698082773608e-06, "loss": 0.0043, "step": 55962 }, { "epoch": 17.22, "learning_rate": 1.0015524465853543e-06, "loss": 0.0047, "step": 55963 }, { "epoch": 17.22, "learning_rate": 1.0013351072340683e-06, "loss": 0.0049, "step": 55964 }, { "epoch": 17.22, "learning_rate": 1.001117790224042e-06, "loss": 0.003, "step": 55965 }, { "epoch": 17.22, "learning_rate": 1.0009004955558144e-06, "loss": 0.0037, "step": 55966 }, { "epoch": 17.22, "learning_rate": 1.0006832232299256e-06, "loss": 0.0054, "step": 55967 }, { "epoch": 17.22, "learning_rate": 1.0004659732469113e-06, "loss": 0.0053, "step": 55968 }, { "epoch": 17.22, "learning_rate": 1.000248745607315e-06, "loss": 0.0065, "step": 55969 }, { "epoch": 17.22, "learning_rate": 1.0000315403116755e-06, "loss": 0.0047, "step": 55970 }, { "epoch": 17.22, "learning_rate": 9.998143573605311e-07, "loss": 0.0052, "step": 55971 }, { "epoch": 17.22, "learning_rate": 9.995971967544217e-07, "loss": 0.0039, "step": 55972 }, { "epoch": 17.22, "learning_rate": 9.99380058493885e-07, "loss": 0.0065, "step": 55973 }, { "epoch": 17.22, "learning_rate": 9.991629425794624e-07, "loss": 0.0041, "step": 55974 }, { "epoch": 17.22, "learning_rate": 9.98945849011691e-07, "loss": 0.0062, "step": 55975 }, { "epoch": 17.22, "learning_rate": 9.987287777911103e-07, "loss": 0.0029, "step": 55976 }, { "epoch": 17.22, "learning_rate": 9.98511728918259e-07, "loss": 0.0036, "step": 55977 }, { "epoch": 17.22, "learning_rate": 9.982947023936752e-07, "loss": 0.0045, "step": 55978 }, { "epoch": 17.22, "learning_rate": 9.980776982179007e-07, "loss": 0.0034, "step": 55979 }, { "epoch": 17.22, "learning_rate": 9.978607163914745e-07, "loss": 0.0018, "step": 55980 }, { "epoch": 17.22, "learning_rate": 9.97643756914931e-07, "loss": 0.003, "step": 55981 }, { "epoch": 17.22, "learning_rate": 9.974268197888104e-07, "loss": 0.0048, "step": 55982 }, { "epoch": 17.23, "learning_rate": 9.972099050136508e-07, "loss": 0.0086, "step": 55983 }, { "epoch": 17.23, "learning_rate": 9.96993012589994e-07, "loss": 0.0046, "step": 55984 }, { "epoch": 17.23, "learning_rate": 9.967761425183763e-07, "loss": 0.0052, "step": 55985 }, { "epoch": 17.23, "learning_rate": 9.965592947993364e-07, "loss": 0.0048, "step": 55986 }, { "epoch": 17.23, "learning_rate": 9.963424694334122e-07, "loss": 0.0051, "step": 55987 }, { "epoch": 17.23, "learning_rate": 9.961256664211427e-07, "loss": 0.0034, "step": 55988 }, { "epoch": 17.23, "learning_rate": 9.95908885763065e-07, "loss": 0.0044, "step": 55989 }, { "epoch": 17.23, "learning_rate": 9.956921274597186e-07, "loss": 0.0041, "step": 55990 }, { "epoch": 17.23, "learning_rate": 9.954753915116412e-07, "loss": 0.0037, "step": 55991 }, { "epoch": 17.23, "learning_rate": 9.952586779193718e-07, "loss": 0.0019, "step": 55992 }, { "epoch": 17.23, "learning_rate": 9.950419866834438e-07, "loss": 0.0037, "step": 55993 }, { "epoch": 17.23, "learning_rate": 9.948253178044043e-07, "loss": 0.0043, "step": 55994 }, { "epoch": 17.23, "learning_rate": 9.94608671282783e-07, "loss": 0.004, "step": 55995 }, { "epoch": 17.23, "learning_rate": 9.943920471191204e-07, "loss": 0.0036, "step": 55996 }, { "epoch": 17.23, "learning_rate": 9.941754453139552e-07, "loss": 0.0058, "step": 55997 }, { "epoch": 17.23, "learning_rate": 9.939588658678222e-07, "loss": 0.0036, "step": 55998 }, { "epoch": 17.23, "learning_rate": 9.937423087812637e-07, "loss": 0.0037, "step": 55999 }, { "epoch": 17.23, "learning_rate": 9.935257740548143e-07, "loss": 0.0022, "step": 56000 }, { "epoch": 17.23, "learning_rate": 9.93309261689015e-07, "loss": 0.0026, "step": 56001 }, { "epoch": 17.23, "learning_rate": 9.930927716843975e-07, "loss": 0.0047, "step": 56002 }, { "epoch": 17.23, "learning_rate": 9.928763040415013e-07, "loss": 0.0061, "step": 56003 }, { "epoch": 17.23, "learning_rate": 9.926598587608672e-07, "loss": 0.0079, "step": 56004 }, { "epoch": 17.23, "learning_rate": 9.924434358430302e-07, "loss": 0.0038, "step": 56005 }, { "epoch": 17.23, "learning_rate": 9.922270352885277e-07, "loss": 0.0036, "step": 56006 }, { "epoch": 17.23, "learning_rate": 9.92010657097897e-07, "loss": 0.0056, "step": 56007 }, { "epoch": 17.23, "learning_rate": 9.917943012716758e-07, "loss": 0.0068, "step": 56008 }, { "epoch": 17.23, "learning_rate": 9.915779678104009e-07, "loss": 0.007, "step": 56009 }, { "epoch": 17.23, "learning_rate": 9.913616567146089e-07, "loss": 0.0074, "step": 56010 }, { "epoch": 17.23, "learning_rate": 9.91145367984837e-07, "loss": 0.0037, "step": 56011 }, { "epoch": 17.23, "learning_rate": 9.909291016216228e-07, "loss": 0.0045, "step": 56012 }, { "epoch": 17.23, "learning_rate": 9.90712857625501e-07, "loss": 0.0047, "step": 56013 }, { "epoch": 17.23, "learning_rate": 9.904966359970114e-07, "loss": 0.0053, "step": 56014 }, { "epoch": 17.24, "learning_rate": 9.902804367366926e-07, "loss": 0.0038, "step": 56015 }, { "epoch": 17.24, "learning_rate": 9.90064259845077e-07, "loss": 0.0025, "step": 56016 }, { "epoch": 17.24, "learning_rate": 9.898481053227015e-07, "loss": 0.0034, "step": 56017 }, { "epoch": 17.24, "learning_rate": 9.896319731701031e-07, "loss": 0.0023, "step": 56018 }, { "epoch": 17.24, "learning_rate": 9.894158633878215e-07, "loss": 0.0058, "step": 56019 }, { "epoch": 17.24, "learning_rate": 9.891997759763905e-07, "loss": 0.0023, "step": 56020 }, { "epoch": 17.24, "learning_rate": 9.889837109363477e-07, "loss": 0.0027, "step": 56021 }, { "epoch": 17.24, "learning_rate": 9.887676682682312e-07, "loss": 0.0054, "step": 56022 }, { "epoch": 17.24, "learning_rate": 9.88551647972571e-07, "loss": 0.0058, "step": 56023 }, { "epoch": 17.24, "learning_rate": 9.883356500499085e-07, "loss": 0.0032, "step": 56024 }, { "epoch": 17.24, "learning_rate": 9.881196745007804e-07, "loss": 0.0049, "step": 56025 }, { "epoch": 17.24, "learning_rate": 9.879037213257214e-07, "loss": 0.0031, "step": 56026 }, { "epoch": 17.24, "learning_rate": 9.87687790525268e-07, "loss": 0.0034, "step": 56027 }, { "epoch": 17.24, "learning_rate": 9.874718820999562e-07, "loss": 0.0033, "step": 56028 }, { "epoch": 17.24, "learning_rate": 9.872559960503215e-07, "loss": 0.0048, "step": 56029 }, { "epoch": 17.24, "learning_rate": 9.870401323769008e-07, "loss": 0.0044, "step": 56030 }, { "epoch": 17.24, "learning_rate": 9.868242910802294e-07, "loss": 0.003, "step": 56031 }, { "epoch": 17.24, "learning_rate": 9.866084721608438e-07, "loss": 0.0041, "step": 56032 }, { "epoch": 17.24, "learning_rate": 9.863926756192766e-07, "loss": 0.0034, "step": 56033 }, { "epoch": 17.24, "learning_rate": 9.861769014560696e-07, "loss": 0.0036, "step": 56034 }, { "epoch": 17.24, "learning_rate": 9.859611496717547e-07, "loss": 0.0041, "step": 56035 }, { "epoch": 17.24, "learning_rate": 9.857454202668693e-07, "loss": 0.0058, "step": 56036 }, { "epoch": 17.24, "learning_rate": 9.855297132419472e-07, "loss": 0.0036, "step": 56037 }, { "epoch": 17.24, "learning_rate": 9.853140285975226e-07, "loss": 0.0044, "step": 56038 }, { "epoch": 17.24, "learning_rate": 9.85098366334134e-07, "loss": 0.0041, "step": 56039 }, { "epoch": 17.24, "learning_rate": 9.848827264523163e-07, "loss": 0.0043, "step": 56040 }, { "epoch": 17.24, "learning_rate": 9.84667108952604e-07, "loss": 0.0052, "step": 56041 }, { "epoch": 17.24, "learning_rate": 9.844515138355337e-07, "loss": 0.003, "step": 56042 }, { "epoch": 17.24, "learning_rate": 9.84235941101639e-07, "loss": 0.0032, "step": 56043 }, { "epoch": 17.24, "learning_rate": 9.840203907514555e-07, "loss": 0.0079, "step": 56044 }, { "epoch": 17.24, "learning_rate": 9.838048627855202e-07, "loss": 0.0071, "step": 56045 }, { "epoch": 17.24, "learning_rate": 9.835893572043654e-07, "loss": 0.0035, "step": 56046 }, { "epoch": 17.24, "learning_rate": 9.833738740085286e-07, "loss": 0.004, "step": 56047 }, { "epoch": 17.25, "learning_rate": 9.831584131985405e-07, "loss": 0.0046, "step": 56048 }, { "epoch": 17.25, "learning_rate": 9.82942974774944e-07, "loss": 0.0034, "step": 56049 }, { "epoch": 17.25, "learning_rate": 9.827275587382667e-07, "loss": 0.0039, "step": 56050 }, { "epoch": 17.25, "learning_rate": 9.825121650890468e-07, "loss": 0.0032, "step": 56051 }, { "epoch": 17.25, "learning_rate": 9.822967938278172e-07, "loss": 0.0044, "step": 56052 }, { "epoch": 17.25, "learning_rate": 9.82081444955112e-07, "loss": 0.0054, "step": 56053 }, { "epoch": 17.25, "learning_rate": 9.818661184714685e-07, "loss": 0.005, "step": 56054 }, { "epoch": 17.25, "learning_rate": 9.816508143774217e-07, "loss": 0.0042, "step": 56055 }, { "epoch": 17.25, "learning_rate": 9.814355326735036e-07, "loss": 0.0052, "step": 56056 }, { "epoch": 17.25, "learning_rate": 9.812202733602516e-07, "loss": 0.0041, "step": 56057 }, { "epoch": 17.25, "learning_rate": 9.810050364381952e-07, "loss": 0.0038, "step": 56058 }, { "epoch": 17.25, "learning_rate": 9.807898219078726e-07, "loss": 0.0062, "step": 56059 }, { "epoch": 17.25, "learning_rate": 9.805746297698182e-07, "loss": 0.0036, "step": 56060 }, { "epoch": 17.25, "learning_rate": 9.80359460024566e-07, "loss": 0.004, "step": 56061 }, { "epoch": 17.25, "learning_rate": 9.801443126726485e-07, "loss": 0.0043, "step": 56062 }, { "epoch": 17.25, "learning_rate": 9.799291877146021e-07, "loss": 0.0149, "step": 56063 }, { "epoch": 17.25, "learning_rate": 9.79714085150959e-07, "loss": 0.0033, "step": 56064 }, { "epoch": 17.25, "learning_rate": 9.79499004982254e-07, "loss": 0.0023, "step": 56065 }, { "epoch": 17.25, "learning_rate": 9.79283947209021e-07, "loss": 0.0049, "step": 56066 }, { "epoch": 17.25, "learning_rate": 9.790689118317953e-07, "loss": 0.0092, "step": 56067 }, { "epoch": 17.25, "learning_rate": 9.788538988511064e-07, "loss": 0.0043, "step": 56068 }, { "epoch": 17.25, "learning_rate": 9.78638908267493e-07, "loss": 0.0053, "step": 56069 }, { "epoch": 17.25, "learning_rate": 9.7842394008149e-07, "loss": 0.0043, "step": 56070 }, { "epoch": 17.25, "learning_rate": 9.782089942936257e-07, "loss": 0.0028, "step": 56071 }, { "epoch": 17.25, "learning_rate": 9.779940709044367e-07, "loss": 0.0076, "step": 56072 }, { "epoch": 17.25, "learning_rate": 9.777791699144535e-07, "loss": 0.0038, "step": 56073 }, { "epoch": 17.25, "learning_rate": 9.775642913242145e-07, "loss": 0.0044, "step": 56074 }, { "epoch": 17.25, "learning_rate": 9.773494351342516e-07, "loss": 0.0046, "step": 56075 }, { "epoch": 17.25, "learning_rate": 9.771346013450967e-07, "loss": 0.0032, "step": 56076 }, { "epoch": 17.25, "learning_rate": 9.769197899572847e-07, "loss": 0.0042, "step": 56077 }, { "epoch": 17.25, "learning_rate": 9.767050009713476e-07, "loss": 0.006, "step": 56078 }, { "epoch": 17.25, "learning_rate": 9.764902343878202e-07, "loss": 0.006, "step": 56079 }, { "epoch": 17.26, "learning_rate": 9.762754902072348e-07, "loss": 0.0024, "step": 56080 }, { "epoch": 17.26, "learning_rate": 9.760607684301238e-07, "loss": 0.003, "step": 56081 }, { "epoch": 17.26, "learning_rate": 9.758460690570227e-07, "loss": 0.0057, "step": 56082 }, { "epoch": 17.26, "learning_rate": 9.75631392088462e-07, "loss": 0.0056, "step": 56083 }, { "epoch": 17.26, "learning_rate": 9.754167375249756e-07, "loss": 0.0039, "step": 56084 }, { "epoch": 17.26, "learning_rate": 9.752021053670968e-07, "loss": 0.0063, "step": 56085 }, { "epoch": 17.26, "learning_rate": 9.749874956153572e-07, "loss": 0.0082, "step": 56086 }, { "epoch": 17.26, "learning_rate": 9.747729082702918e-07, "loss": 0.0033, "step": 56087 }, { "epoch": 17.26, "learning_rate": 9.74558343332429e-07, "loss": 0.0066, "step": 56088 }, { "epoch": 17.26, "learning_rate": 9.743438008023077e-07, "loss": 0.0049, "step": 56089 }, { "epoch": 17.26, "learning_rate": 9.74129280680457e-07, "loss": 0.0034, "step": 56090 }, { "epoch": 17.26, "learning_rate": 9.739147829674124e-07, "loss": 0.0026, "step": 56091 }, { "epoch": 17.26, "learning_rate": 9.73700307663702e-07, "loss": 0.0044, "step": 56092 }, { "epoch": 17.26, "learning_rate": 9.734858547698579e-07, "loss": 0.0048, "step": 56093 }, { "epoch": 17.26, "learning_rate": 9.732714242864172e-07, "loss": 0.0034, "step": 56094 }, { "epoch": 17.26, "learning_rate": 9.730570162139097e-07, "loss": 0.0036, "step": 56095 }, { "epoch": 17.26, "learning_rate": 9.728426305528682e-07, "loss": 0.0062, "step": 56096 }, { "epoch": 17.26, "learning_rate": 9.726282673038245e-07, "loss": 0.0059, "step": 56097 }, { "epoch": 17.26, "learning_rate": 9.724139264673116e-07, "loss": 0.0038, "step": 56098 }, { "epoch": 17.26, "learning_rate": 9.72199608043861e-07, "loss": 0.0065, "step": 56099 }, { "epoch": 17.26, "learning_rate": 9.71985312034005e-07, "loss": 0.0028, "step": 56100 }, { "epoch": 17.26, "learning_rate": 9.717710384382762e-07, "loss": 0.0037, "step": 56101 }, { "epoch": 17.26, "learning_rate": 9.715567872572051e-07, "loss": 0.0038, "step": 56102 }, { "epoch": 17.26, "learning_rate": 9.713425584913227e-07, "loss": 0.0066, "step": 56103 }, { "epoch": 17.26, "learning_rate": 9.711283521411674e-07, "loss": 0.0063, "step": 56104 }, { "epoch": 17.26, "learning_rate": 9.70914168207263e-07, "loss": 0.0025, "step": 56105 }, { "epoch": 17.26, "learning_rate": 9.707000066901452e-07, "loss": 0.0031, "step": 56106 }, { "epoch": 17.26, "learning_rate": 9.70485867590345e-07, "loss": 0.0036, "step": 56107 }, { "epoch": 17.26, "learning_rate": 9.702717509083925e-07, "loss": 0.0034, "step": 56108 }, { "epoch": 17.26, "learning_rate": 9.700576566448228e-07, "loss": 0.0036, "step": 56109 }, { "epoch": 17.26, "learning_rate": 9.698435848001653e-07, "loss": 0.0048, "step": 56110 }, { "epoch": 17.26, "learning_rate": 9.69629535374952e-07, "loss": 0.0059, "step": 56111 }, { "epoch": 17.26, "learning_rate": 9.694155083697154e-07, "loss": 0.0045, "step": 56112 }, { "epoch": 17.27, "learning_rate": 9.692015037849823e-07, "loss": 0.0053, "step": 56113 }, { "epoch": 17.27, "learning_rate": 9.689875216212886e-07, "loss": 0.0031, "step": 56114 }, { "epoch": 17.27, "learning_rate": 9.68773561879165e-07, "loss": 0.0055, "step": 56115 }, { "epoch": 17.27, "learning_rate": 9.68559624559141e-07, "loss": 0.0039, "step": 56116 }, { "epoch": 17.27, "learning_rate": 9.683457096617487e-07, "loss": 0.005, "step": 56117 }, { "epoch": 17.27, "learning_rate": 9.681318171875198e-07, "loss": 0.0033, "step": 56118 }, { "epoch": 17.27, "learning_rate": 9.679179471369837e-07, "loss": 0.0033, "step": 56119 }, { "epoch": 17.27, "learning_rate": 9.677040995106734e-07, "loss": 0.0057, "step": 56120 }, { "epoch": 17.27, "learning_rate": 9.674902743091187e-07, "loss": 0.0056, "step": 56121 }, { "epoch": 17.27, "learning_rate": 9.672764715328497e-07, "loss": 0.0067, "step": 56122 }, { "epoch": 17.27, "learning_rate": 9.670626911823966e-07, "loss": 0.0061, "step": 56123 }, { "epoch": 17.27, "learning_rate": 9.668489332582942e-07, "loss": 0.0048, "step": 56124 }, { "epoch": 17.27, "learning_rate": 9.66635197761072e-07, "loss": 0.0025, "step": 56125 }, { "epoch": 17.27, "learning_rate": 9.664214846912568e-07, "loss": 0.0051, "step": 56126 }, { "epoch": 17.27, "learning_rate": 9.66207794049383e-07, "loss": 0.0051, "step": 56127 }, { "epoch": 17.27, "learning_rate": 9.659941258359761e-07, "loss": 0.0026, "step": 56128 }, { "epoch": 17.27, "learning_rate": 9.657804800515736e-07, "loss": 0.0037, "step": 56129 }, { "epoch": 17.27, "learning_rate": 9.655668566967026e-07, "loss": 0.0074, "step": 56130 }, { "epoch": 17.27, "learning_rate": 9.653532557718936e-07, "loss": 0.0042, "step": 56131 }, { "epoch": 17.27, "learning_rate": 9.651396772776755e-07, "loss": 0.003, "step": 56132 }, { "epoch": 17.27, "learning_rate": 9.64926121214581e-07, "loss": 0.0074, "step": 56133 }, { "epoch": 17.27, "learning_rate": 9.647125875831386e-07, "loss": 0.0034, "step": 56134 }, { "epoch": 17.27, "learning_rate": 9.644990763838803e-07, "loss": 0.0044, "step": 56135 }, { "epoch": 17.27, "learning_rate": 9.64285587617333e-07, "loss": 0.0045, "step": 56136 }, { "epoch": 17.27, "learning_rate": 9.640721212840298e-07, "loss": 0.0029, "step": 56137 }, { "epoch": 17.27, "learning_rate": 9.638586773844993e-07, "loss": 0.0041, "step": 56138 }, { "epoch": 17.27, "learning_rate": 9.63645255919271e-07, "loss": 0.0035, "step": 56139 }, { "epoch": 17.27, "learning_rate": 9.634318568888767e-07, "loss": 0.0044, "step": 56140 }, { "epoch": 17.27, "learning_rate": 9.632184802938438e-07, "loss": 0.0025, "step": 56141 }, { "epoch": 17.27, "learning_rate": 9.63005126134704e-07, "loss": 0.0039, "step": 56142 }, { "epoch": 17.27, "learning_rate": 9.627917944119835e-07, "loss": 0.0066, "step": 56143 }, { "epoch": 17.27, "learning_rate": 9.625784851262165e-07, "loss": 0.0038, "step": 56144 }, { "epoch": 17.28, "learning_rate": 9.623651982779315e-07, "loss": 0.0032, "step": 56145 }, { "epoch": 17.28, "learning_rate": 9.621519338676587e-07, "loss": 0.0058, "step": 56146 }, { "epoch": 17.28, "learning_rate": 9.61938691895924e-07, "loss": 0.0039, "step": 56147 }, { "epoch": 17.28, "learning_rate": 9.617254723632574e-07, "loss": 0.003, "step": 56148 }, { "epoch": 17.28, "learning_rate": 9.61512275270191e-07, "loss": 0.0043, "step": 56149 }, { "epoch": 17.28, "learning_rate": 9.612991006172534e-07, "loss": 0.003, "step": 56150 }, { "epoch": 17.28, "learning_rate": 9.61085948404974e-07, "loss": 0.003, "step": 56151 }, { "epoch": 17.28, "learning_rate": 9.608728186338812e-07, "loss": 0.0025, "step": 56152 }, { "epoch": 17.28, "learning_rate": 9.606597113045035e-07, "loss": 0.0053, "step": 56153 }, { "epoch": 17.28, "learning_rate": 9.604466264173718e-07, "loss": 0.0037, "step": 56154 }, { "epoch": 17.28, "learning_rate": 9.602335639730132e-07, "loss": 0.004, "step": 56155 }, { "epoch": 17.28, "learning_rate": 9.600205239719584e-07, "loss": 0.0044, "step": 56156 }, { "epoch": 17.28, "learning_rate": 9.598075064147328e-07, "loss": 0.0038, "step": 56157 }, { "epoch": 17.28, "learning_rate": 9.5959451130187e-07, "loss": 0.0054, "step": 56158 }, { "epoch": 17.28, "learning_rate": 9.59381538633899e-07, "loss": 0.0044, "step": 56159 }, { "epoch": 17.28, "learning_rate": 9.591685884113444e-07, "loss": 0.0067, "step": 56160 }, { "epoch": 17.28, "learning_rate": 9.589556606347372e-07, "loss": 0.006, "step": 56161 }, { "epoch": 17.28, "learning_rate": 9.587427553046036e-07, "loss": 0.0043, "step": 56162 }, { "epoch": 17.28, "learning_rate": 9.585298724214763e-07, "loss": 0.0061, "step": 56163 }, { "epoch": 17.28, "learning_rate": 9.58317011985882e-07, "loss": 0.0074, "step": 56164 }, { "epoch": 17.28, "learning_rate": 9.581041739983488e-07, "loss": 0.0037, "step": 56165 }, { "epoch": 17.28, "learning_rate": 9.578913584594052e-07, "loss": 0.0061, "step": 56166 }, { "epoch": 17.28, "learning_rate": 9.576785653695798e-07, "loss": 0.0028, "step": 56167 }, { "epoch": 17.28, "learning_rate": 9.574657947294008e-07, "loss": 0.0078, "step": 56168 }, { "epoch": 17.28, "learning_rate": 9.57253046539396e-07, "loss": 0.0041, "step": 56169 }, { "epoch": 17.28, "learning_rate": 9.570403208000945e-07, "loss": 0.0078, "step": 56170 }, { "epoch": 17.28, "learning_rate": 9.56827617512024e-07, "loss": 0.0094, "step": 56171 }, { "epoch": 17.28, "learning_rate": 9.566149366757104e-07, "loss": 0.0033, "step": 56172 }, { "epoch": 17.28, "learning_rate": 9.56402278291687e-07, "loss": 0.0041, "step": 56173 }, { "epoch": 17.28, "learning_rate": 9.561896423604777e-07, "loss": 0.0036, "step": 56174 }, { "epoch": 17.28, "learning_rate": 9.55977028882612e-07, "loss": 0.0051, "step": 56175 }, { "epoch": 17.28, "learning_rate": 9.557644378586162e-07, "loss": 0.006, "step": 56176 }, { "epoch": 17.28, "learning_rate": 9.555518692890165e-07, "loss": 0.0105, "step": 56177 }, { "epoch": 17.29, "learning_rate": 9.55339323174347e-07, "loss": 0.0066, "step": 56178 }, { "epoch": 17.29, "learning_rate": 9.551267995151304e-07, "loss": 0.0055, "step": 56179 }, { "epoch": 17.29, "learning_rate": 9.549142983118976e-07, "loss": 0.0038, "step": 56180 }, { "epoch": 17.29, "learning_rate": 9.547018195651714e-07, "loss": 0.0039, "step": 56181 }, { "epoch": 17.29, "learning_rate": 9.544893632754816e-07, "loss": 0.0039, "step": 56182 }, { "epoch": 17.29, "learning_rate": 9.542769294433573e-07, "loss": 0.0043, "step": 56183 }, { "epoch": 17.29, "learning_rate": 9.540645180693243e-07, "loss": 0.0047, "step": 56184 }, { "epoch": 17.29, "learning_rate": 9.538521291539105e-07, "loss": 0.0046, "step": 56185 }, { "epoch": 17.29, "learning_rate": 9.536397626976434e-07, "loss": 0.0043, "step": 56186 }, { "epoch": 17.29, "learning_rate": 9.534274187010506e-07, "loss": 0.0037, "step": 56187 }, { "epoch": 17.29, "learning_rate": 9.532150971646581e-07, "loss": 0.0036, "step": 56188 }, { "epoch": 17.29, "learning_rate": 9.530027980889933e-07, "loss": 0.0038, "step": 56189 }, { "epoch": 17.29, "learning_rate": 9.527905214745836e-07, "loss": 0.0051, "step": 56190 }, { "epoch": 17.29, "learning_rate": 9.525782673219563e-07, "loss": 0.0034, "step": 56191 }, { "epoch": 17.29, "learning_rate": 9.523660356316366e-07, "loss": 0.0025, "step": 56192 }, { "epoch": 17.29, "learning_rate": 9.521538264041541e-07, "loss": 0.004, "step": 56193 }, { "epoch": 17.29, "learning_rate": 9.519416396400361e-07, "loss": 0.0062, "step": 56194 }, { "epoch": 17.29, "learning_rate": 9.517294753398066e-07, "loss": 0.0075, "step": 56195 }, { "epoch": 17.29, "learning_rate": 9.515173335039929e-07, "loss": 0.006, "step": 56196 }, { "epoch": 17.29, "learning_rate": 9.513052141331203e-07, "loss": 0.0052, "step": 56197 }, { "epoch": 17.29, "learning_rate": 9.510931172277193e-07, "loss": 0.0052, "step": 56198 }, { "epoch": 17.29, "learning_rate": 9.508810427883142e-07, "loss": 0.0041, "step": 56199 }, { "epoch": 17.29, "learning_rate": 9.506689908154331e-07, "loss": 0.0026, "step": 56200 }, { "epoch": 17.29, "learning_rate": 9.504569613096026e-07, "loss": 0.0097, "step": 56201 }, { "epoch": 17.29, "learning_rate": 9.502449542713432e-07, "loss": 0.0038, "step": 56202 }, { "epoch": 17.29, "learning_rate": 9.50032969701189e-07, "loss": 0.0036, "step": 56203 }, { "epoch": 17.29, "learning_rate": 9.498210075996628e-07, "loss": 0.0047, "step": 56204 }, { "epoch": 17.29, "learning_rate": 9.496090679672909e-07, "loss": 0.0051, "step": 56205 }, { "epoch": 17.29, "learning_rate": 9.493971508045996e-07, "loss": 0.0064, "step": 56206 }, { "epoch": 17.29, "learning_rate": 9.491852561121151e-07, "loss": 0.0043, "step": 56207 }, { "epoch": 17.29, "learning_rate": 9.489733838903648e-07, "loss": 0.0049, "step": 56208 }, { "epoch": 17.29, "learning_rate": 9.487615341398726e-07, "loss": 0.0032, "step": 56209 }, { "epoch": 17.3, "learning_rate": 9.485497068611648e-07, "loss": 0.0058, "step": 56210 }, { "epoch": 17.3, "learning_rate": 9.48337902054769e-07, "loss": 0.0047, "step": 56211 }, { "epoch": 17.3, "learning_rate": 9.481261197212077e-07, "loss": 0.0026, "step": 56212 }, { "epoch": 17.3, "learning_rate": 9.479143598610097e-07, "loss": 0.0034, "step": 56213 }, { "epoch": 17.3, "learning_rate": 9.47702622474701e-07, "loss": 0.0043, "step": 56214 }, { "epoch": 17.3, "learning_rate": 9.474909075628081e-07, "loss": 0.0033, "step": 56215 }, { "epoch": 17.3, "learning_rate": 9.472792151258536e-07, "loss": 0.0028, "step": 56216 }, { "epoch": 17.3, "learning_rate": 9.470675451643619e-07, "loss": 0.005, "step": 56217 }, { "epoch": 17.3, "learning_rate": 9.468558976788622e-07, "loss": 0.0041, "step": 56218 }, { "epoch": 17.3, "learning_rate": 9.466442726698788e-07, "loss": 0.0068, "step": 56219 }, { "epoch": 17.3, "learning_rate": 9.464326701379368e-07, "loss": 0.0064, "step": 56220 }, { "epoch": 17.3, "learning_rate": 9.462210900835622e-07, "loss": 0.0017, "step": 56221 }, { "epoch": 17.3, "learning_rate": 9.460095325072805e-07, "loss": 0.004, "step": 56222 }, { "epoch": 17.3, "learning_rate": 9.457979974096154e-07, "loss": 0.0025, "step": 56223 }, { "epoch": 17.3, "learning_rate": 9.455864847910922e-07, "loss": 0.0034, "step": 56224 }, { "epoch": 17.3, "learning_rate": 9.453749946522384e-07, "loss": 0.0037, "step": 56225 }, { "epoch": 17.3, "learning_rate": 9.451635269935766e-07, "loss": 0.0044, "step": 56226 }, { "epoch": 17.3, "learning_rate": 9.449520818156299e-07, "loss": 0.0029, "step": 56227 }, { "epoch": 17.3, "learning_rate": 9.447406591189312e-07, "loss": 0.0043, "step": 56228 }, { "epoch": 17.3, "learning_rate": 9.445292589039978e-07, "loss": 0.0064, "step": 56229 }, { "epoch": 17.3, "learning_rate": 9.443178811713572e-07, "loss": 0.0078, "step": 56230 }, { "epoch": 17.3, "learning_rate": 9.441065259215331e-07, "loss": 0.0096, "step": 56231 }, { "epoch": 17.3, "learning_rate": 9.438951931550499e-07, "loss": 0.0046, "step": 56232 }, { "epoch": 17.3, "learning_rate": 9.436838828724359e-07, "loss": 0.0047, "step": 56233 }, { "epoch": 17.3, "learning_rate": 9.434725950742119e-07, "loss": 0.0033, "step": 56234 }, { "epoch": 17.3, "learning_rate": 9.43261329760905e-07, "loss": 0.005, "step": 56235 }, { "epoch": 17.3, "learning_rate": 9.430500869330406e-07, "loss": 0.0015, "step": 56236 }, { "epoch": 17.3, "learning_rate": 9.428388665911369e-07, "loss": 0.0025, "step": 56237 }, { "epoch": 17.3, "learning_rate": 9.426276687357239e-07, "loss": 0.0035, "step": 56238 }, { "epoch": 17.3, "learning_rate": 9.424164933673252e-07, "loss": 0.0062, "step": 56239 }, { "epoch": 17.3, "learning_rate": 9.42205340486465e-07, "loss": 0.0049, "step": 56240 }, { "epoch": 17.3, "learning_rate": 9.419942100936663e-07, "loss": 0.0059, "step": 56241 }, { "epoch": 17.3, "learning_rate": 9.417831021894541e-07, "loss": 0.0054, "step": 56242 }, { "epoch": 17.31, "learning_rate": 9.415720167743525e-07, "loss": 0.0031, "step": 56243 }, { "epoch": 17.31, "learning_rate": 9.413609538488866e-07, "loss": 0.0054, "step": 56244 }, { "epoch": 17.31, "learning_rate": 9.411499134135782e-07, "loss": 0.0035, "step": 56245 }, { "epoch": 17.31, "learning_rate": 9.409388954689524e-07, "loss": 0.0025, "step": 56246 }, { "epoch": 17.31, "learning_rate": 9.407279000155311e-07, "loss": 0.004, "step": 56247 }, { "epoch": 17.31, "learning_rate": 9.405169270538417e-07, "loss": 0.0083, "step": 56248 }, { "epoch": 17.31, "learning_rate": 9.403059765844091e-07, "loss": 0.0049, "step": 56249 }, { "epoch": 17.31, "learning_rate": 9.40095048607752e-07, "loss": 0.0027, "step": 56250 }, { "epoch": 17.31, "learning_rate": 9.398841431243966e-07, "loss": 0.0059, "step": 56251 }, { "epoch": 17.31, "learning_rate": 9.396732601348645e-07, "loss": 0.0064, "step": 56252 }, { "epoch": 17.31, "learning_rate": 9.394623996396835e-07, "loss": 0.0057, "step": 56253 }, { "epoch": 17.31, "learning_rate": 9.392515616393738e-07, "loss": 0.0038, "step": 56254 }, { "epoch": 17.31, "learning_rate": 9.390407461344597e-07, "loss": 0.0043, "step": 56255 }, { "epoch": 17.31, "learning_rate": 9.388299531254652e-07, "loss": 0.002, "step": 56256 }, { "epoch": 17.31, "learning_rate": 9.386191826129132e-07, "loss": 0.0033, "step": 56257 }, { "epoch": 17.31, "learning_rate": 9.384084345973266e-07, "loss": 0.0046, "step": 56258 }, { "epoch": 17.31, "learning_rate": 9.381977090792294e-07, "loss": 0.0039, "step": 56259 }, { "epoch": 17.31, "learning_rate": 9.379870060591434e-07, "loss": 0.0047, "step": 56260 }, { "epoch": 17.31, "learning_rate": 9.377763255375927e-07, "loss": 0.0118, "step": 56261 }, { "epoch": 17.31, "learning_rate": 9.375656675151013e-07, "loss": 0.007, "step": 56262 }, { "epoch": 17.31, "learning_rate": 9.373550319921909e-07, "loss": 0.0069, "step": 56263 }, { "epoch": 17.31, "learning_rate": 9.371444189693846e-07, "loss": 0.0033, "step": 56264 }, { "epoch": 17.31, "learning_rate": 9.369338284472051e-07, "loss": 0.0034, "step": 56265 }, { "epoch": 17.31, "learning_rate": 9.367232604261756e-07, "loss": 0.0049, "step": 56266 }, { "epoch": 17.31, "learning_rate": 9.365127149068165e-07, "loss": 0.0067, "step": 56267 }, { "epoch": 17.31, "learning_rate": 9.363021918896553e-07, "loss": 0.0033, "step": 56268 }, { "epoch": 17.31, "learning_rate": 9.360916913752127e-07, "loss": 0.0098, "step": 56269 }, { "epoch": 17.31, "learning_rate": 9.358812133640127e-07, "loss": 0.0028, "step": 56270 }, { "epoch": 17.31, "learning_rate": 9.356707578565738e-07, "loss": 0.0049, "step": 56271 }, { "epoch": 17.31, "learning_rate": 9.354603248534189e-07, "loss": 0.0088, "step": 56272 }, { "epoch": 17.31, "learning_rate": 9.352499143550742e-07, "loss": 0.0054, "step": 56273 }, { "epoch": 17.31, "learning_rate": 9.350395263620604e-07, "loss": 0.0058, "step": 56274 }, { "epoch": 17.32, "learning_rate": 9.348291608748994e-07, "loss": 0.0042, "step": 56275 }, { "epoch": 17.32, "learning_rate": 9.34618817894114e-07, "loss": 0.0037, "step": 56276 }, { "epoch": 17.32, "learning_rate": 9.34408497420225e-07, "loss": 0.0098, "step": 56277 }, { "epoch": 17.32, "learning_rate": 9.341981994537563e-07, "loss": 0.0065, "step": 56278 }, { "epoch": 17.32, "learning_rate": 9.339879239952298e-07, "loss": 0.0049, "step": 56279 }, { "epoch": 17.32, "learning_rate": 9.337776710451674e-07, "loss": 0.0054, "step": 56280 }, { "epoch": 17.32, "learning_rate": 9.335674406040907e-07, "loss": 0.0041, "step": 56281 }, { "epoch": 17.32, "learning_rate": 9.333572326725193e-07, "loss": 0.0033, "step": 56282 }, { "epoch": 17.32, "learning_rate": 9.331470472509818e-07, "loss": 0.0068, "step": 56283 }, { "epoch": 17.32, "learning_rate": 9.329368843399944e-07, "loss": 0.0032, "step": 56284 }, { "epoch": 17.32, "learning_rate": 9.327267439400789e-07, "loss": 0.0055, "step": 56285 }, { "epoch": 17.32, "learning_rate": 9.325166260517593e-07, "loss": 0.0039, "step": 56286 }, { "epoch": 17.32, "learning_rate": 9.323065306755552e-07, "loss": 0.0047, "step": 56287 }, { "epoch": 17.32, "learning_rate": 9.320964578119907e-07, "loss": 0.0076, "step": 56288 }, { "epoch": 17.32, "learning_rate": 9.318864074615863e-07, "loss": 0.0041, "step": 56289 }, { "epoch": 17.32, "learning_rate": 9.316763796248629e-07, "loss": 0.002, "step": 56290 }, { "epoch": 17.32, "learning_rate": 9.314663743023433e-07, "loss": 0.0063, "step": 56291 }, { "epoch": 17.32, "learning_rate": 9.312563914945461e-07, "loss": 0.0041, "step": 56292 }, { "epoch": 17.32, "learning_rate": 9.310464312019952e-07, "loss": 0.006, "step": 56293 }, { "epoch": 17.32, "learning_rate": 9.308364934252113e-07, "loss": 0.0053, "step": 56294 }, { "epoch": 17.32, "learning_rate": 9.30626578164715e-07, "loss": 0.0055, "step": 56295 }, { "epoch": 17.32, "learning_rate": 9.304166854210284e-07, "loss": 0.0047, "step": 56296 }, { "epoch": 17.32, "learning_rate": 9.302068151946719e-07, "loss": 0.0039, "step": 56297 }, { "epoch": 17.32, "learning_rate": 9.299969674861664e-07, "loss": 0.0037, "step": 56298 }, { "epoch": 17.32, "learning_rate": 9.297871422960336e-07, "loss": 0.0069, "step": 56299 }, { "epoch": 17.32, "learning_rate": 9.295773396247943e-07, "loss": 0.0067, "step": 56300 }, { "epoch": 17.32, "learning_rate": 9.29367559472969e-07, "loss": 0.0059, "step": 56301 }, { "epoch": 17.32, "learning_rate": 9.291578018410762e-07, "loss": 0.0062, "step": 56302 }, { "epoch": 17.32, "learning_rate": 9.289480667296414e-07, "loss": 0.0037, "step": 56303 }, { "epoch": 17.32, "learning_rate": 9.287383541391848e-07, "loss": 0.0052, "step": 56304 }, { "epoch": 17.32, "learning_rate": 9.28528664070224e-07, "loss": 0.0076, "step": 56305 }, { "epoch": 17.32, "learning_rate": 9.283189965232808e-07, "loss": 0.0054, "step": 56306 }, { "epoch": 17.32, "learning_rate": 9.281093514988737e-07, "loss": 0.0042, "step": 56307 }, { "epoch": 17.33, "learning_rate": 9.278997289975277e-07, "loss": 0.006, "step": 56308 }, { "epoch": 17.33, "learning_rate": 9.276901290197604e-07, "loss": 0.0034, "step": 56309 }, { "epoch": 17.33, "learning_rate": 9.274805515660923e-07, "loss": 0.004, "step": 56310 }, { "epoch": 17.33, "learning_rate": 9.272709966370453e-07, "loss": 0.0027, "step": 56311 }, { "epoch": 17.33, "learning_rate": 9.270614642331377e-07, "loss": 0.004, "step": 56312 }, { "epoch": 17.33, "learning_rate": 9.268519543548904e-07, "loss": 0.0036, "step": 56313 }, { "epoch": 17.33, "learning_rate": 9.266424670028241e-07, "loss": 0.004, "step": 56314 }, { "epoch": 17.33, "learning_rate": 9.264330021774582e-07, "loss": 0.0028, "step": 56315 }, { "epoch": 17.33, "learning_rate": 9.262235598793134e-07, "loss": 0.0035, "step": 56316 }, { "epoch": 17.33, "learning_rate": 9.260141401089095e-07, "loss": 0.0038, "step": 56317 }, { "epoch": 17.33, "learning_rate": 9.258047428667649e-07, "loss": 0.0023, "step": 56318 }, { "epoch": 17.33, "learning_rate": 9.255953681534024e-07, "loss": 0.0023, "step": 56319 }, { "epoch": 17.33, "learning_rate": 9.253860159693384e-07, "loss": 0.008, "step": 56320 }, { "epoch": 17.33, "learning_rate": 9.251766863150957e-07, "loss": 0.005, "step": 56321 }, { "epoch": 17.33, "learning_rate": 9.249673791911907e-07, "loss": 0.003, "step": 56322 }, { "epoch": 17.33, "learning_rate": 9.247580945981472e-07, "loss": 0.0027, "step": 56323 }, { "epoch": 17.33, "learning_rate": 9.245488325364826e-07, "loss": 0.0024, "step": 56324 }, { "epoch": 17.33, "learning_rate": 9.243395930067189e-07, "loss": 0.0056, "step": 56325 }, { "epoch": 17.33, "learning_rate": 9.241303760093712e-07, "loss": 0.0042, "step": 56326 }, { "epoch": 17.33, "learning_rate": 9.239211815449578e-07, "loss": 0.0043, "step": 56327 }, { "epoch": 17.33, "learning_rate": 9.237120096140051e-07, "loss": 0.0063, "step": 56328 }, { "epoch": 17.33, "learning_rate": 9.23502860217027e-07, "loss": 0.0043, "step": 56329 }, { "epoch": 17.33, "learning_rate": 9.232937333545455e-07, "loss": 0.0049, "step": 56330 }, { "epoch": 17.33, "learning_rate": 9.230846290270779e-07, "loss": 0.0046, "step": 56331 }, { "epoch": 17.33, "learning_rate": 9.228755472351447e-07, "loss": 0.0052, "step": 56332 }, { "epoch": 17.33, "learning_rate": 9.226664879792635e-07, "loss": 0.0026, "step": 56333 }, { "epoch": 17.33, "learning_rate": 9.224574512599549e-07, "loss": 0.0028, "step": 56334 }, { "epoch": 17.33, "learning_rate": 9.222484370777374e-07, "loss": 0.0049, "step": 56335 }, { "epoch": 17.33, "learning_rate": 9.220394454331294e-07, "loss": 0.0035, "step": 56336 }, { "epoch": 17.33, "learning_rate": 9.218304763266483e-07, "loss": 0.0032, "step": 56337 }, { "epoch": 17.33, "learning_rate": 9.216215297588182e-07, "loss": 0.004, "step": 56338 }, { "epoch": 17.33, "learning_rate": 9.214126057301532e-07, "loss": 0.0035, "step": 56339 }, { "epoch": 17.34, "learning_rate": 9.212037042411726e-07, "loss": 0.0047, "step": 56340 }, { "epoch": 17.34, "learning_rate": 9.209948252923961e-07, "loss": 0.0035, "step": 56341 }, { "epoch": 17.34, "learning_rate": 9.207859688843402e-07, "loss": 0.0041, "step": 56342 }, { "epoch": 17.34, "learning_rate": 9.205771350175264e-07, "loss": 0.004, "step": 56343 }, { "epoch": 17.34, "learning_rate": 9.20368323692471e-07, "loss": 0.0053, "step": 56344 }, { "epoch": 17.34, "learning_rate": 9.201595349096948e-07, "loss": 0.005, "step": 56345 }, { "epoch": 17.34, "learning_rate": 9.199507686697151e-07, "loss": 0.0047, "step": 56346 }, { "epoch": 17.34, "learning_rate": 9.197420249730471e-07, "loss": 0.0035, "step": 56347 }, { "epoch": 17.34, "learning_rate": 9.195333038202137e-07, "loss": 0.0089, "step": 56348 }, { "epoch": 17.34, "learning_rate": 9.1932460521173e-07, "loss": 0.0041, "step": 56349 }, { "epoch": 17.34, "learning_rate": 9.191159291481156e-07, "loss": 0.0034, "step": 56350 }, { "epoch": 17.34, "learning_rate": 9.189072756298889e-07, "loss": 0.0026, "step": 56351 }, { "epoch": 17.34, "learning_rate": 9.186986446575663e-07, "loss": 0.0044, "step": 56352 }, { "epoch": 17.34, "learning_rate": 9.184900362316674e-07, "loss": 0.0043, "step": 56353 }, { "epoch": 17.34, "learning_rate": 9.182814503527093e-07, "loss": 0.0026, "step": 56354 }, { "epoch": 17.34, "learning_rate": 9.180728870212097e-07, "loss": 0.0044, "step": 56355 }, { "epoch": 17.34, "learning_rate": 9.178643462376868e-07, "loss": 0.0044, "step": 56356 }, { "epoch": 17.34, "learning_rate": 9.17655828002657e-07, "loss": 0.0031, "step": 56357 }, { "epoch": 17.34, "learning_rate": 9.174473323166399e-07, "loss": 0.0044, "step": 56358 }, { "epoch": 17.34, "learning_rate": 9.172388591801562e-07, "loss": 0.0046, "step": 56359 }, { "epoch": 17.34, "learning_rate": 9.170304085937165e-07, "loss": 0.0064, "step": 56360 }, { "epoch": 17.34, "learning_rate": 9.168219805578416e-07, "loss": 0.0048, "step": 56361 }, { "epoch": 17.34, "learning_rate": 9.166135750730476e-07, "loss": 0.0061, "step": 56362 }, { "epoch": 17.34, "learning_rate": 9.164051921398553e-07, "loss": 0.0052, "step": 56363 }, { "epoch": 17.34, "learning_rate": 9.161968317587788e-07, "loss": 0.0028, "step": 56364 }, { "epoch": 17.34, "learning_rate": 9.159884939303377e-07, "loss": 0.003, "step": 56365 }, { "epoch": 17.34, "learning_rate": 9.15780178655048e-07, "loss": 0.0043, "step": 56366 }, { "epoch": 17.34, "learning_rate": 9.155718859334273e-07, "loss": 0.0059, "step": 56367 }, { "epoch": 17.34, "learning_rate": 9.153636157659929e-07, "loss": 0.0051, "step": 56368 }, { "epoch": 17.34, "learning_rate": 9.15155368153261e-07, "loss": 0.0037, "step": 56369 }, { "epoch": 17.34, "learning_rate": 9.149471430957491e-07, "loss": 0.0031, "step": 56370 }, { "epoch": 17.34, "learning_rate": 9.147389405939744e-07, "loss": 0.0054, "step": 56371 }, { "epoch": 17.34, "learning_rate": 9.14530760648451e-07, "loss": 0.0041, "step": 56372 }, { "epoch": 17.35, "learning_rate": 9.14322603259703e-07, "loss": 0.0043, "step": 56373 }, { "epoch": 17.35, "learning_rate": 9.141144684282399e-07, "loss": 0.003, "step": 56374 }, { "epoch": 17.35, "learning_rate": 9.139063561545813e-07, "loss": 0.0031, "step": 56375 }, { "epoch": 17.35, "learning_rate": 9.136982664392446e-07, "loss": 0.0061, "step": 56376 }, { "epoch": 17.35, "learning_rate": 9.134901992827427e-07, "loss": 0.0051, "step": 56377 }, { "epoch": 17.35, "learning_rate": 9.132821546855964e-07, "loss": 0.0036, "step": 56378 }, { "epoch": 17.35, "learning_rate": 9.130741326483216e-07, "loss": 0.0032, "step": 56379 }, { "epoch": 17.35, "learning_rate": 9.128661331714361e-07, "loss": 0.0058, "step": 56380 }, { "epoch": 17.35, "learning_rate": 9.126581562554515e-07, "loss": 0.0029, "step": 56381 }, { "epoch": 17.35, "learning_rate": 9.124502019008863e-07, "loss": 0.0021, "step": 56382 }, { "epoch": 17.35, "learning_rate": 9.122422701082578e-07, "loss": 0.0061, "step": 56383 }, { "epoch": 17.35, "learning_rate": 9.120343608780824e-07, "loss": 0.005, "step": 56384 }, { "epoch": 17.35, "learning_rate": 9.118264742108751e-07, "loss": 0.0045, "step": 56385 }, { "epoch": 17.35, "learning_rate": 9.116186101071533e-07, "loss": 0.006, "step": 56386 }, { "epoch": 17.35, "learning_rate": 9.114107685674323e-07, "loss": 0.0043, "step": 56387 }, { "epoch": 17.35, "learning_rate": 9.11202949592227e-07, "loss": 0.0013, "step": 56388 }, { "epoch": 17.35, "learning_rate": 9.10995153182056e-07, "loss": 0.0032, "step": 56389 }, { "epoch": 17.35, "learning_rate": 9.107873793374322e-07, "loss": 0.0051, "step": 56390 }, { "epoch": 17.35, "learning_rate": 9.105796280588741e-07, "loss": 0.0032, "step": 56391 }, { "epoch": 17.35, "learning_rate": 9.103718993468946e-07, "loss": 0.0047, "step": 56392 }, { "epoch": 17.35, "learning_rate": 9.101641932020122e-07, "loss": 0.005, "step": 56393 }, { "epoch": 17.35, "learning_rate": 9.099565096247443e-07, "loss": 0.0038, "step": 56394 }, { "epoch": 17.35, "learning_rate": 9.097488486156014e-07, "loss": 0.0052, "step": 56395 }, { "epoch": 17.35, "learning_rate": 9.095412101751011e-07, "loss": 0.0054, "step": 56396 }, { "epoch": 17.35, "learning_rate": 9.093335943037584e-07, "loss": 0.0042, "step": 56397 }, { "epoch": 17.35, "learning_rate": 9.091260010020908e-07, "loss": 0.0054, "step": 56398 }, { "epoch": 17.35, "learning_rate": 9.089184302706133e-07, "loss": 0.0055, "step": 56399 }, { "epoch": 17.35, "learning_rate": 9.08710882109839e-07, "loss": 0.0036, "step": 56400 }, { "epoch": 17.35, "learning_rate": 9.085033565202872e-07, "loss": 0.0023, "step": 56401 }, { "epoch": 17.35, "learning_rate": 9.082958535024677e-07, "loss": 0.0028, "step": 56402 }, { "epoch": 17.35, "learning_rate": 9.08088373056899e-07, "loss": 0.0051, "step": 56403 }, { "epoch": 17.35, "learning_rate": 9.078809151840972e-07, "loss": 0.0047, "step": 56404 }, { "epoch": 17.36, "learning_rate": 9.076734798845743e-07, "loss": 0.0061, "step": 56405 }, { "epoch": 17.36, "learning_rate": 9.074660671588487e-07, "loss": 0.0042, "step": 56406 }, { "epoch": 17.36, "learning_rate": 9.072586770074321e-07, "loss": 0.0041, "step": 56407 }, { "epoch": 17.36, "learning_rate": 9.07051309430842e-07, "loss": 0.0049, "step": 56408 }, { "epoch": 17.36, "learning_rate": 9.068439644295913e-07, "loss": 0.0089, "step": 56409 }, { "epoch": 17.36, "learning_rate": 9.066366420041961e-07, "loss": 0.0034, "step": 56410 }, { "epoch": 17.36, "learning_rate": 9.064293421551707e-07, "loss": 0.0033, "step": 56411 }, { "epoch": 17.36, "learning_rate": 9.062220648830267e-07, "loss": 0.004, "step": 56412 }, { "epoch": 17.36, "learning_rate": 9.060148101882849e-07, "loss": 0.0034, "step": 56413 }, { "epoch": 17.36, "learning_rate": 9.058075780714582e-07, "loss": 0.0032, "step": 56414 }, { "epoch": 17.36, "learning_rate": 9.056003685330572e-07, "loss": 0.0083, "step": 56415 }, { "epoch": 17.36, "learning_rate": 9.053931815735995e-07, "loss": 0.0051, "step": 56416 }, { "epoch": 17.36, "learning_rate": 9.051860171935955e-07, "loss": 0.0049, "step": 56417 }, { "epoch": 17.36, "learning_rate": 9.049788753935662e-07, "loss": 0.0039, "step": 56418 }, { "epoch": 17.36, "learning_rate": 9.04771756174021e-07, "loss": 0.0045, "step": 56419 }, { "epoch": 17.36, "learning_rate": 9.045646595354762e-07, "loss": 0.0024, "step": 56420 }, { "epoch": 17.36, "learning_rate": 9.043575854784459e-07, "loss": 0.0062, "step": 56421 }, { "epoch": 17.36, "learning_rate": 9.041505340034429e-07, "loss": 0.0049, "step": 56422 }, { "epoch": 17.36, "learning_rate": 9.039435051109824e-07, "loss": 0.0039, "step": 56423 }, { "epoch": 17.36, "learning_rate": 9.037364988015773e-07, "loss": 0.0038, "step": 56424 }, { "epoch": 17.36, "learning_rate": 9.035295150757428e-07, "loss": 0.0066, "step": 56425 }, { "epoch": 17.36, "learning_rate": 9.033225539339919e-07, "loss": 0.0079, "step": 56426 }, { "epoch": 17.36, "learning_rate": 9.031156153768361e-07, "loss": 0.0065, "step": 56427 }, { "epoch": 17.36, "learning_rate": 9.029086994047964e-07, "loss": 0.0024, "step": 56428 }, { "epoch": 17.36, "learning_rate": 9.027018060183801e-07, "loss": 0.0033, "step": 56429 }, { "epoch": 17.36, "learning_rate": 9.024949352181023e-07, "loss": 0.007, "step": 56430 }, { "epoch": 17.36, "learning_rate": 9.02288087004477e-07, "loss": 0.0033, "step": 56431 }, { "epoch": 17.36, "learning_rate": 9.020812613780161e-07, "loss": 0.0058, "step": 56432 }, { "epoch": 17.36, "learning_rate": 9.018744583392369e-07, "loss": 0.004, "step": 56433 }, { "epoch": 17.36, "learning_rate": 9.016676778886502e-07, "loss": 0.0064, "step": 56434 }, { "epoch": 17.36, "learning_rate": 9.014609200267721e-07, "loss": 0.0035, "step": 56435 }, { "epoch": 17.36, "learning_rate": 9.012541847541112e-07, "loss": 0.0058, "step": 56436 }, { "epoch": 17.36, "learning_rate": 9.010474720711814e-07, "loss": 0.0059, "step": 56437 }, { "epoch": 17.37, "learning_rate": 9.008407819784993e-07, "loss": 0.0044, "step": 56438 }, { "epoch": 17.37, "learning_rate": 9.006341144765774e-07, "loss": 0.0036, "step": 56439 }, { "epoch": 17.37, "learning_rate": 9.004274695659265e-07, "loss": 0.0024, "step": 56440 }, { "epoch": 17.37, "learning_rate": 9.002208472470619e-07, "loss": 0.0054, "step": 56441 }, { "epoch": 17.37, "learning_rate": 9.000142475204965e-07, "loss": 0.0025, "step": 56442 }, { "epoch": 17.37, "learning_rate": 8.998076703867409e-07, "loss": 0.0045, "step": 56443 }, { "epoch": 17.37, "learning_rate": 8.996011158463103e-07, "loss": 0.0066, "step": 56444 }, { "epoch": 17.37, "learning_rate": 8.993945838997176e-07, "loss": 0.0135, "step": 56445 }, { "epoch": 17.37, "learning_rate": 8.991880745474734e-07, "loss": 0.0047, "step": 56446 }, { "epoch": 17.37, "learning_rate": 8.989815877900898e-07, "loss": 0.0054, "step": 56447 }, { "epoch": 17.37, "learning_rate": 8.98775123628084e-07, "loss": 0.0065, "step": 56448 }, { "epoch": 17.37, "learning_rate": 8.985686820619677e-07, "loss": 0.005, "step": 56449 }, { "epoch": 17.37, "learning_rate": 8.983622630922495e-07, "loss": 0.0027, "step": 56450 }, { "epoch": 17.37, "learning_rate": 8.981558667194445e-07, "loss": 0.0033, "step": 56451 }, { "epoch": 17.37, "learning_rate": 8.979494929440625e-07, "loss": 0.0048, "step": 56452 }, { "epoch": 17.37, "learning_rate": 8.977431417666193e-07, "loss": 0.0034, "step": 56453 }, { "epoch": 17.37, "learning_rate": 8.975368131876272e-07, "loss": 0.0024, "step": 56454 }, { "epoch": 17.37, "learning_rate": 8.973305072075966e-07, "loss": 0.0028, "step": 56455 }, { "epoch": 17.37, "learning_rate": 8.971242238270416e-07, "loss": 0.0042, "step": 56456 }, { "epoch": 17.37, "learning_rate": 8.969179630464709e-07, "loss": 0.0034, "step": 56457 }, { "epoch": 17.37, "learning_rate": 8.967117248663992e-07, "loss": 0.002, "step": 56458 }, { "epoch": 17.37, "learning_rate": 8.965055092873387e-07, "loss": 0.0035, "step": 56459 }, { "epoch": 17.37, "learning_rate": 8.962993163098011e-07, "loss": 0.0059, "step": 56460 }, { "epoch": 17.37, "learning_rate": 8.960931459342981e-07, "loss": 0.005, "step": 56461 }, { "epoch": 17.37, "learning_rate": 8.958869981613405e-07, "loss": 0.0041, "step": 56462 }, { "epoch": 17.37, "learning_rate": 8.956808729914423e-07, "loss": 0.0039, "step": 56463 }, { "epoch": 17.37, "learning_rate": 8.954747704251132e-07, "loss": 0.0054, "step": 56464 }, { "epoch": 17.37, "learning_rate": 8.95268690462866e-07, "loss": 0.0076, "step": 56465 }, { "epoch": 17.37, "learning_rate": 8.950626331052114e-07, "loss": 0.0035, "step": 56466 }, { "epoch": 17.37, "learning_rate": 8.948565983526613e-07, "loss": 0.003, "step": 56467 }, { "epoch": 17.37, "learning_rate": 8.946505862057286e-07, "loss": 0.0062, "step": 56468 }, { "epoch": 17.37, "learning_rate": 8.94444596664924e-07, "loss": 0.0034, "step": 56469 }, { "epoch": 17.38, "learning_rate": 8.942386297307604e-07, "loss": 0.0022, "step": 56470 }, { "epoch": 17.38, "learning_rate": 8.940326854037452e-07, "loss": 0.0039, "step": 56471 }, { "epoch": 17.38, "learning_rate": 8.9382676368439e-07, "loss": 0.0075, "step": 56472 }, { "epoch": 17.38, "learning_rate": 8.936208645732103e-07, "loss": 0.0072, "step": 56473 }, { "epoch": 17.38, "learning_rate": 8.934149880707154e-07, "loss": 0.005, "step": 56474 }, { "epoch": 17.38, "learning_rate": 8.932091341774152e-07, "loss": 0.0041, "step": 56475 }, { "epoch": 17.38, "learning_rate": 8.930033028938223e-07, "loss": 0.0028, "step": 56476 }, { "epoch": 17.38, "learning_rate": 8.927974942204464e-07, "loss": 0.004, "step": 56477 }, { "epoch": 17.38, "learning_rate": 8.925917081577995e-07, "loss": 0.0053, "step": 56478 }, { "epoch": 17.38, "learning_rate": 8.923859447063921e-07, "loss": 0.0109, "step": 56479 }, { "epoch": 17.38, "learning_rate": 8.921802038667349e-07, "loss": 0.0042, "step": 56480 }, { "epoch": 17.38, "learning_rate": 8.919744856393386e-07, "loss": 0.0054, "step": 56481 }, { "epoch": 17.38, "learning_rate": 8.91768790024713e-07, "loss": 0.004, "step": 56482 }, { "epoch": 17.38, "learning_rate": 8.915631170233741e-07, "loss": 0.004, "step": 56483 }, { "epoch": 17.38, "learning_rate": 8.913574666358249e-07, "loss": 0.0036, "step": 56484 }, { "epoch": 17.38, "learning_rate": 8.911518388625806e-07, "loss": 0.0039, "step": 56485 }, { "epoch": 17.38, "learning_rate": 8.909462337041508e-07, "loss": 0.004, "step": 56486 }, { "epoch": 17.38, "learning_rate": 8.907406511610439e-07, "loss": 0.0074, "step": 56487 }, { "epoch": 17.38, "learning_rate": 8.90535091233774e-07, "loss": 0.0021, "step": 56488 }, { "epoch": 17.38, "learning_rate": 8.903295539228485e-07, "loss": 0.0036, "step": 56489 }, { "epoch": 17.38, "learning_rate": 8.901240392287802e-07, "loss": 0.0026, "step": 56490 }, { "epoch": 17.38, "learning_rate": 8.899185471520788e-07, "loss": 0.0048, "step": 56491 }, { "epoch": 17.38, "learning_rate": 8.897130776932506e-07, "loss": 0.0038, "step": 56492 }, { "epoch": 17.38, "learning_rate": 8.895076308528105e-07, "loss": 0.0034, "step": 56493 }, { "epoch": 17.38, "learning_rate": 8.893022066312674e-07, "loss": 0.0067, "step": 56494 }, { "epoch": 17.38, "learning_rate": 8.890968050291293e-07, "loss": 0.0048, "step": 56495 }, { "epoch": 17.38, "learning_rate": 8.888914260469095e-07, "loss": 0.0026, "step": 56496 }, { "epoch": 17.38, "learning_rate": 8.886860696851141e-07, "loss": 0.0025, "step": 56497 }, { "epoch": 17.38, "learning_rate": 8.88480735944256e-07, "loss": 0.0061, "step": 56498 }, { "epoch": 17.38, "learning_rate": 8.882754248248438e-07, "loss": 0.0034, "step": 56499 }, { "epoch": 17.38, "learning_rate": 8.880701363273869e-07, "loss": 0.0024, "step": 56500 }, { "epoch": 17.38, "learning_rate": 8.878648704523952e-07, "loss": 0.0039, "step": 56501 }, { "epoch": 17.38, "learning_rate": 8.876596272003768e-07, "loss": 0.0028, "step": 56502 }, { "epoch": 17.39, "learning_rate": 8.87454406571846e-07, "loss": 0.0047, "step": 56503 }, { "epoch": 17.39, "learning_rate": 8.87249208567309e-07, "loss": 0.0052, "step": 56504 }, { "epoch": 17.39, "learning_rate": 8.870440331872754e-07, "loss": 0.0019, "step": 56505 }, { "epoch": 17.39, "learning_rate": 8.868388804322537e-07, "loss": 0.0029, "step": 56506 }, { "epoch": 17.39, "learning_rate": 8.866337503027523e-07, "loss": 0.0041, "step": 56507 }, { "epoch": 17.39, "learning_rate": 8.864286427992852e-07, "loss": 0.0045, "step": 56508 }, { "epoch": 17.39, "learning_rate": 8.862235579223577e-07, "loss": 0.0111, "step": 56509 }, { "epoch": 17.39, "learning_rate": 8.860184956724805e-07, "loss": 0.0055, "step": 56510 }, { "epoch": 17.39, "learning_rate": 8.85813456050163e-07, "loss": 0.0041, "step": 56511 }, { "epoch": 17.39, "learning_rate": 8.856084390559128e-07, "loss": 0.0033, "step": 56512 }, { "epoch": 17.39, "learning_rate": 8.854034446902404e-07, "loss": 0.0037, "step": 56513 }, { "epoch": 17.39, "learning_rate": 8.851984729536534e-07, "loss": 0.0055, "step": 56514 }, { "epoch": 17.39, "learning_rate": 8.849935238466611e-07, "loss": 0.0039, "step": 56515 }, { "epoch": 17.39, "learning_rate": 8.847885973697734e-07, "loss": 0.0036, "step": 56516 }, { "epoch": 17.39, "learning_rate": 8.845836935234975e-07, "loss": 0.002, "step": 56517 }, { "epoch": 17.39, "learning_rate": 8.84378812308343e-07, "loss": 0.0054, "step": 56518 }, { "epoch": 17.39, "learning_rate": 8.841739537248184e-07, "loss": 0.0052, "step": 56519 }, { "epoch": 17.39, "learning_rate": 8.839691177734322e-07, "loss": 0.0034, "step": 56520 }, { "epoch": 17.39, "learning_rate": 8.837643044546928e-07, "loss": 0.0035, "step": 56521 }, { "epoch": 17.39, "learning_rate": 8.835595137691078e-07, "loss": 0.0039, "step": 56522 }, { "epoch": 17.39, "learning_rate": 8.833547457171887e-07, "loss": 0.0031, "step": 56523 }, { "epoch": 17.39, "learning_rate": 8.83150000299442e-07, "loss": 0.0033, "step": 56524 }, { "epoch": 17.39, "learning_rate": 8.829452775163772e-07, "loss": 0.0023, "step": 56525 }, { "epoch": 17.39, "learning_rate": 8.827405773685005e-07, "loss": 0.0032, "step": 56526 }, { "epoch": 17.39, "learning_rate": 8.825358998563182e-07, "loss": 0.0051, "step": 56527 }, { "epoch": 17.39, "learning_rate": 8.823312449803434e-07, "loss": 0.0021, "step": 56528 }, { "epoch": 17.39, "learning_rate": 8.821266127410833e-07, "loss": 0.0034, "step": 56529 }, { "epoch": 17.39, "learning_rate": 8.81922003139043e-07, "loss": 0.0047, "step": 56530 }, { "epoch": 17.39, "learning_rate": 8.817174161747333e-07, "loss": 0.0036, "step": 56531 }, { "epoch": 17.39, "learning_rate": 8.815128518486604e-07, "loss": 0.007, "step": 56532 }, { "epoch": 17.39, "learning_rate": 8.813083101613329e-07, "loss": 0.0038, "step": 56533 }, { "epoch": 17.39, "learning_rate": 8.811037911132581e-07, "loss": 0.0028, "step": 56534 }, { "epoch": 17.4, "learning_rate": 8.808992947049455e-07, "loss": 0.0069, "step": 56535 }, { "epoch": 17.4, "learning_rate": 8.806948209369004e-07, "loss": 0.0037, "step": 56536 }, { "epoch": 17.4, "learning_rate": 8.804903698096301e-07, "loss": 0.0055, "step": 56537 }, { "epoch": 17.4, "learning_rate": 8.802859413236476e-07, "loss": 0.0017, "step": 56538 }, { "epoch": 17.4, "learning_rate": 8.800815354794545e-07, "loss": 0.0033, "step": 56539 }, { "epoch": 17.4, "learning_rate": 8.798771522775607e-07, "loss": 0.0056, "step": 56540 }, { "epoch": 17.4, "learning_rate": 8.796727917184722e-07, "loss": 0.004, "step": 56541 }, { "epoch": 17.4, "learning_rate": 8.794684538026965e-07, "loss": 0.0047, "step": 56542 }, { "epoch": 17.4, "learning_rate": 8.792641385307433e-07, "loss": 0.0051, "step": 56543 }, { "epoch": 17.4, "learning_rate": 8.790598459031174e-07, "loss": 0.0021, "step": 56544 }, { "epoch": 17.4, "learning_rate": 8.788555759203288e-07, "loss": 0.0022, "step": 56545 }, { "epoch": 17.4, "learning_rate": 8.786513285828835e-07, "loss": 0.0036, "step": 56546 }, { "epoch": 17.4, "learning_rate": 8.784471038912845e-07, "loss": 0.0039, "step": 56547 }, { "epoch": 17.4, "learning_rate": 8.782429018460436e-07, "loss": 0.0075, "step": 56548 }, { "epoch": 17.4, "learning_rate": 8.780387224476671e-07, "loss": 0.0042, "step": 56549 }, { "epoch": 17.4, "learning_rate": 8.778345656966613e-07, "loss": 0.0042, "step": 56550 }, { "epoch": 17.4, "learning_rate": 8.776304315935324e-07, "loss": 0.0054, "step": 56551 }, { "epoch": 17.4, "learning_rate": 8.774263201387889e-07, "loss": 0.0055, "step": 56552 }, { "epoch": 17.4, "learning_rate": 8.772222313329348e-07, "loss": 0.0028, "step": 56553 }, { "epoch": 17.4, "learning_rate": 8.770181651764797e-07, "loss": 0.0079, "step": 56554 }, { "epoch": 17.4, "learning_rate": 8.768141216699288e-07, "loss": 0.0065, "step": 56555 }, { "epoch": 17.4, "learning_rate": 8.766101008137872e-07, "loss": 0.0031, "step": 56556 }, { "epoch": 17.4, "learning_rate": 8.764061026085646e-07, "loss": 0.004, "step": 56557 }, { "epoch": 17.4, "learning_rate": 8.76202127054766e-07, "loss": 0.0029, "step": 56558 }, { "epoch": 17.4, "learning_rate": 8.759981741529e-07, "loss": 0.004, "step": 56559 }, { "epoch": 17.4, "learning_rate": 8.757942439034695e-07, "loss": 0.0026, "step": 56560 }, { "epoch": 17.4, "learning_rate": 8.755903363069784e-07, "loss": 0.0045, "step": 56561 }, { "epoch": 17.4, "learning_rate": 8.753864513639399e-07, "loss": 0.004, "step": 56562 }, { "epoch": 17.4, "learning_rate": 8.751825890748567e-07, "loss": 0.0046, "step": 56563 }, { "epoch": 17.4, "learning_rate": 8.749787494402351e-07, "loss": 0.0055, "step": 56564 }, { "epoch": 17.4, "learning_rate": 8.747749324605815e-07, "loss": 0.0042, "step": 56565 }, { "epoch": 17.4, "learning_rate": 8.745711381364009e-07, "loss": 0.0047, "step": 56566 }, { "epoch": 17.4, "learning_rate": 8.743673664682007e-07, "loss": 0.0065, "step": 56567 }, { "epoch": 17.41, "learning_rate": 8.74163617456486e-07, "loss": 0.0059, "step": 56568 }, { "epoch": 17.41, "learning_rate": 8.739598911017633e-07, "loss": 0.0044, "step": 56569 }, { "epoch": 17.41, "learning_rate": 8.737561874045364e-07, "loss": 0.004, "step": 56570 }, { "epoch": 17.41, "learning_rate": 8.735525063653116e-07, "loss": 0.0097, "step": 56571 }, { "epoch": 17.41, "learning_rate": 8.733488479845997e-07, "loss": 0.0034, "step": 56572 }, { "epoch": 17.41, "learning_rate": 8.731452122628992e-07, "loss": 0.0037, "step": 56573 }, { "epoch": 17.41, "learning_rate": 8.729415992007196e-07, "loss": 0.0031, "step": 56574 }, { "epoch": 17.41, "learning_rate": 8.727380087985648e-07, "loss": 0.0051, "step": 56575 }, { "epoch": 17.41, "learning_rate": 8.725344410569392e-07, "loss": 0.0022, "step": 56576 }, { "epoch": 17.41, "learning_rate": 8.723308959763522e-07, "loss": 0.0056, "step": 56577 }, { "epoch": 17.41, "learning_rate": 8.721273735573066e-07, "loss": 0.0057, "step": 56578 }, { "epoch": 17.41, "learning_rate": 8.719238738003077e-07, "loss": 0.0023, "step": 56579 }, { "epoch": 17.41, "learning_rate": 8.71720396705863e-07, "loss": 0.0039, "step": 56580 }, { "epoch": 17.41, "learning_rate": 8.715169422744718e-07, "loss": 0.0036, "step": 56581 }, { "epoch": 17.41, "learning_rate": 8.71313510506645e-07, "loss": 0.0049, "step": 56582 }, { "epoch": 17.41, "learning_rate": 8.711101014028855e-07, "loss": 0.0047, "step": 56583 }, { "epoch": 17.41, "learning_rate": 8.709067149636985e-07, "loss": 0.0043, "step": 56584 }, { "epoch": 17.41, "learning_rate": 8.70703351189589e-07, "loss": 0.0034, "step": 56585 }, { "epoch": 17.41, "learning_rate": 8.705000100810623e-07, "loss": 0.0018, "step": 56586 }, { "epoch": 17.41, "learning_rate": 8.702966916386235e-07, "loss": 0.0034, "step": 56587 }, { "epoch": 17.41, "learning_rate": 8.700933958627755e-07, "loss": 0.0021, "step": 56588 }, { "epoch": 17.41, "learning_rate": 8.698901227540246e-07, "loss": 0.0065, "step": 56589 }, { "epoch": 17.41, "learning_rate": 8.696868723128748e-07, "loss": 0.0056, "step": 56590 }, { "epoch": 17.41, "learning_rate": 8.694836445398303e-07, "loss": 0.005, "step": 56591 }, { "epoch": 17.41, "learning_rate": 8.692804394353971e-07, "loss": 0.0036, "step": 56592 }, { "epoch": 17.41, "learning_rate": 8.690772570000816e-07, "loss": 0.004, "step": 56593 }, { "epoch": 17.41, "learning_rate": 8.688740972343834e-07, "loss": 0.0031, "step": 56594 }, { "epoch": 17.41, "learning_rate": 8.686709601388088e-07, "loss": 0.0042, "step": 56595 }, { "epoch": 17.41, "learning_rate": 8.684678457138618e-07, "loss": 0.0035, "step": 56596 }, { "epoch": 17.41, "learning_rate": 8.682647539600475e-07, "loss": 0.0044, "step": 56597 }, { "epoch": 17.41, "learning_rate": 8.680616848778711e-07, "loss": 0.0036, "step": 56598 }, { "epoch": 17.41, "learning_rate": 8.678586384678355e-07, "loss": 0.0028, "step": 56599 }, { "epoch": 17.42, "learning_rate": 8.676556147304449e-07, "loss": 0.0078, "step": 56600 }, { "epoch": 17.42, "learning_rate": 8.674526136662032e-07, "loss": 0.0046, "step": 56601 }, { "epoch": 17.42, "learning_rate": 8.672496352756154e-07, "loss": 0.0042, "step": 56602 }, { "epoch": 17.42, "learning_rate": 8.670466795591837e-07, "loss": 0.006, "step": 56603 }, { "epoch": 17.42, "learning_rate": 8.66843746517414e-07, "loss": 0.0055, "step": 56604 }, { "epoch": 17.42, "learning_rate": 8.666408361508083e-07, "loss": 0.006, "step": 56605 }, { "epoch": 17.42, "learning_rate": 8.664379484598684e-07, "loss": 0.0037, "step": 56606 }, { "epoch": 17.42, "learning_rate": 8.662350834451061e-07, "loss": 0.004, "step": 56607 }, { "epoch": 17.42, "learning_rate": 8.660322411070166e-07, "loss": 0.0037, "step": 56608 }, { "epoch": 17.42, "learning_rate": 8.658294214461072e-07, "loss": 0.0042, "step": 56609 }, { "epoch": 17.42, "learning_rate": 8.656266244628808e-07, "loss": 0.0035, "step": 56610 }, { "epoch": 17.42, "learning_rate": 8.654238501578393e-07, "loss": 0.0056, "step": 56611 }, { "epoch": 17.42, "learning_rate": 8.65221098531489e-07, "loss": 0.0038, "step": 56612 }, { "epoch": 17.42, "learning_rate": 8.650183695843328e-07, "loss": 0.0036, "step": 56613 }, { "epoch": 17.42, "learning_rate": 8.648156633168747e-07, "loss": 0.0042, "step": 56614 }, { "epoch": 17.42, "learning_rate": 8.646129797296143e-07, "loss": 0.0047, "step": 56615 }, { "epoch": 17.42, "learning_rate": 8.644103188230557e-07, "loss": 0.0038, "step": 56616 }, { "epoch": 17.42, "learning_rate": 8.642076805977051e-07, "loss": 0.0038, "step": 56617 }, { "epoch": 17.42, "learning_rate": 8.640050650540644e-07, "loss": 0.0058, "step": 56618 }, { "epoch": 17.42, "learning_rate": 8.638024721926353e-07, "loss": 0.0036, "step": 56619 }, { "epoch": 17.42, "learning_rate": 8.635999020139229e-07, "loss": 0.005, "step": 56620 }, { "epoch": 17.42, "learning_rate": 8.633973545184271e-07, "loss": 0.0065, "step": 56621 }, { "epoch": 17.42, "learning_rate": 8.631948297066539e-07, "loss": 0.0052, "step": 56622 }, { "epoch": 17.42, "learning_rate": 8.629923275791041e-07, "loss": 0.0059, "step": 56623 }, { "epoch": 17.42, "learning_rate": 8.627898481362817e-07, "loss": 0.0029, "step": 56624 }, { "epoch": 17.42, "learning_rate": 8.625873913786875e-07, "loss": 0.0048, "step": 56625 }, { "epoch": 17.42, "learning_rate": 8.623849573068244e-07, "loss": 0.0044, "step": 56626 }, { "epoch": 17.42, "learning_rate": 8.621825459211985e-07, "loss": 0.0067, "step": 56627 }, { "epoch": 17.42, "learning_rate": 8.619801572223108e-07, "loss": 0.0036, "step": 56628 }, { "epoch": 17.42, "learning_rate": 8.617777912106617e-07, "loss": 0.0044, "step": 56629 }, { "epoch": 17.42, "learning_rate": 8.615754478867544e-07, "loss": 0.0055, "step": 56630 }, { "epoch": 17.42, "learning_rate": 8.613731272510894e-07, "loss": 0.0057, "step": 56631 }, { "epoch": 17.42, "learning_rate": 8.611708293041732e-07, "loss": 0.0053, "step": 56632 }, { "epoch": 17.43, "learning_rate": 8.609685540465063e-07, "loss": 0.006, "step": 56633 }, { "epoch": 17.43, "learning_rate": 8.607663014785905e-07, "loss": 0.004, "step": 56634 }, { "epoch": 17.43, "learning_rate": 8.6056407160093e-07, "loss": 0.0025, "step": 56635 }, { "epoch": 17.43, "learning_rate": 8.603618644140222e-07, "loss": 0.0038, "step": 56636 }, { "epoch": 17.43, "learning_rate": 8.60159679918372e-07, "loss": 0.0061, "step": 56637 }, { "epoch": 17.43, "learning_rate": 8.599575181144826e-07, "loss": 0.0062, "step": 56638 }, { "epoch": 17.43, "learning_rate": 8.597553790028545e-07, "loss": 0.0044, "step": 56639 }, { "epoch": 17.43, "learning_rate": 8.595532625839898e-07, "loss": 0.0038, "step": 56640 }, { "epoch": 17.43, "learning_rate": 8.593511688583889e-07, "loss": 0.0053, "step": 56641 }, { "epoch": 17.43, "learning_rate": 8.59149097826556e-07, "loss": 0.0083, "step": 56642 }, { "epoch": 17.43, "learning_rate": 8.589470494889907e-07, "loss": 0.0055, "step": 56643 }, { "epoch": 17.43, "learning_rate": 8.587450238461969e-07, "loss": 0.0037, "step": 56644 }, { "epoch": 17.43, "learning_rate": 8.585430208986734e-07, "loss": 0.0026, "step": 56645 }, { "epoch": 17.43, "learning_rate": 8.583410406469217e-07, "loss": 0.0055, "step": 56646 }, { "epoch": 17.43, "learning_rate": 8.581390830914471e-07, "loss": 0.003, "step": 56647 }, { "epoch": 17.43, "learning_rate": 8.579371482327481e-07, "loss": 0.0043, "step": 56648 }, { "epoch": 17.43, "learning_rate": 8.577352360713287e-07, "loss": 0.0061, "step": 56649 }, { "epoch": 17.43, "learning_rate": 8.575333466076863e-07, "loss": 0.0041, "step": 56650 }, { "epoch": 17.43, "learning_rate": 8.573314798423215e-07, "loss": 0.0026, "step": 56651 }, { "epoch": 17.43, "learning_rate": 8.571296357757397e-07, "loss": 0.0043, "step": 56652 }, { "epoch": 17.43, "learning_rate": 8.569278144084403e-07, "loss": 0.0037, "step": 56653 }, { "epoch": 17.43, "learning_rate": 8.567260157409241e-07, "loss": 0.0042, "step": 56654 }, { "epoch": 17.43, "learning_rate": 8.565242397736928e-07, "loss": 0.0027, "step": 56655 }, { "epoch": 17.43, "learning_rate": 8.563224865072461e-07, "loss": 0.0017, "step": 56656 }, { "epoch": 17.43, "learning_rate": 8.561207559420859e-07, "loss": 0.003, "step": 56657 }, { "epoch": 17.43, "learning_rate": 8.559190480787128e-07, "loss": 0.007, "step": 56658 }, { "epoch": 17.43, "learning_rate": 8.557173629176274e-07, "loss": 0.0044, "step": 56659 }, { "epoch": 17.43, "learning_rate": 8.555157004593306e-07, "loss": 0.0045, "step": 56660 }, { "epoch": 17.43, "learning_rate": 8.55314060704322e-07, "loss": 0.005, "step": 56661 }, { "epoch": 17.43, "learning_rate": 8.551124436531055e-07, "loss": 0.0022, "step": 56662 }, { "epoch": 17.43, "learning_rate": 8.549108493061786e-07, "loss": 0.0034, "step": 56663 }, { "epoch": 17.43, "learning_rate": 8.54709277664042e-07, "loss": 0.0043, "step": 56664 }, { "epoch": 17.44, "learning_rate": 8.545077287271975e-07, "loss": 0.0042, "step": 56665 }, { "epoch": 17.44, "learning_rate": 8.543062024961424e-07, "loss": 0.0035, "step": 56666 }, { "epoch": 17.44, "learning_rate": 8.541046989713808e-07, "loss": 0.0054, "step": 56667 }, { "epoch": 17.44, "learning_rate": 8.539032181534124e-07, "loss": 0.0049, "step": 56668 }, { "epoch": 17.44, "learning_rate": 8.537017600427366e-07, "loss": 0.0038, "step": 56669 }, { "epoch": 17.44, "learning_rate": 8.535003246398554e-07, "loss": 0.0045, "step": 56670 }, { "epoch": 17.44, "learning_rate": 8.532989119452628e-07, "loss": 0.004, "step": 56671 }, { "epoch": 17.44, "learning_rate": 8.530975219594651e-07, "loss": 0.0039, "step": 56672 }, { "epoch": 17.44, "learning_rate": 8.528961546829606e-07, "loss": 0.0038, "step": 56673 }, { "epoch": 17.44, "learning_rate": 8.526948101162492e-07, "loss": 0.0041, "step": 56674 }, { "epoch": 17.44, "learning_rate": 8.524934882598301e-07, "loss": 0.0052, "step": 56675 }, { "epoch": 17.44, "learning_rate": 8.522921891142034e-07, "loss": 0.0046, "step": 56676 }, { "epoch": 17.44, "learning_rate": 8.520909126798693e-07, "loss": 0.0048, "step": 56677 }, { "epoch": 17.44, "learning_rate": 8.518896589573278e-07, "loss": 0.0044, "step": 56678 }, { "epoch": 17.44, "learning_rate": 8.516884279470772e-07, "loss": 0.0034, "step": 56679 }, { "epoch": 17.44, "learning_rate": 8.514872196496182e-07, "loss": 0.0055, "step": 56680 }, { "epoch": 17.44, "learning_rate": 8.512860340654483e-07, "loss": 0.0026, "step": 56681 }, { "epoch": 17.44, "learning_rate": 8.510848711950704e-07, "loss": 0.0033, "step": 56682 }, { "epoch": 17.44, "learning_rate": 8.50883731038985e-07, "loss": 0.0028, "step": 56683 }, { "epoch": 17.44, "learning_rate": 8.506826135976865e-07, "loss": 0.0052, "step": 56684 }, { "epoch": 17.44, "learning_rate": 8.504815188716753e-07, "loss": 0.0023, "step": 56685 }, { "epoch": 17.44, "learning_rate": 8.502804468614512e-07, "loss": 0.006, "step": 56686 }, { "epoch": 17.44, "learning_rate": 8.500793975675159e-07, "loss": 0.0037, "step": 56687 }, { "epoch": 17.44, "learning_rate": 8.498783709903668e-07, "loss": 0.0093, "step": 56688 }, { "epoch": 17.44, "learning_rate": 8.496773671305025e-07, "loss": 0.0035, "step": 56689 }, { "epoch": 17.44, "learning_rate": 8.494763859884225e-07, "loss": 0.0029, "step": 56690 }, { "epoch": 17.44, "learning_rate": 8.492754275646254e-07, "loss": 0.0069, "step": 56691 }, { "epoch": 17.44, "learning_rate": 8.490744918596106e-07, "loss": 0.0035, "step": 56692 }, { "epoch": 17.44, "learning_rate": 8.488735788738778e-07, "loss": 0.0063, "step": 56693 }, { "epoch": 17.44, "learning_rate": 8.486726886079233e-07, "loss": 0.0047, "step": 56694 }, { "epoch": 17.44, "learning_rate": 8.484718210622489e-07, "loss": 0.005, "step": 56695 }, { "epoch": 17.44, "learning_rate": 8.482709762373509e-07, "loss": 0.0042, "step": 56696 }, { "epoch": 17.44, "learning_rate": 8.480701541337288e-07, "loss": 0.0056, "step": 56697 }, { "epoch": 17.45, "learning_rate": 8.478693547518813e-07, "loss": 0.0035, "step": 56698 }, { "epoch": 17.45, "learning_rate": 8.476685780923066e-07, "loss": 0.0059, "step": 56699 }, { "epoch": 17.45, "learning_rate": 8.474678241555046e-07, "loss": 0.0033, "step": 56700 }, { "epoch": 17.45, "learning_rate": 8.472670929419691e-07, "loss": 0.006, "step": 56701 }, { "epoch": 17.45, "learning_rate": 8.470663844522053e-07, "loss": 0.0048, "step": 56702 }, { "epoch": 17.45, "learning_rate": 8.468656986867075e-07, "loss": 0.0038, "step": 56703 }, { "epoch": 17.45, "learning_rate": 8.46665035645976e-07, "loss": 0.0036, "step": 56704 }, { "epoch": 17.45, "learning_rate": 8.464643953305063e-07, "loss": 0.0028, "step": 56705 }, { "epoch": 17.45, "learning_rate": 8.462637777407956e-07, "loss": 0.0039, "step": 56706 }, { "epoch": 17.45, "learning_rate": 8.460631828773458e-07, "loss": 0.0046, "step": 56707 }, { "epoch": 17.45, "learning_rate": 8.458626107406531e-07, "loss": 0.0053, "step": 56708 }, { "epoch": 17.45, "learning_rate": 8.456620613312161e-07, "loss": 0.0107, "step": 56709 }, { "epoch": 17.45, "learning_rate": 8.454615346495321e-07, "loss": 0.0023, "step": 56710 }, { "epoch": 17.45, "learning_rate": 8.452610306960996e-07, "loss": 0.0032, "step": 56711 }, { "epoch": 17.45, "learning_rate": 8.450605494714159e-07, "loss": 0.0056, "step": 56712 }, { "epoch": 17.45, "learning_rate": 8.448600909759786e-07, "loss": 0.005, "step": 56713 }, { "epoch": 17.45, "learning_rate": 8.446596552102848e-07, "loss": 0.0027, "step": 56714 }, { "epoch": 17.45, "learning_rate": 8.444592421748332e-07, "loss": 0.005, "step": 56715 }, { "epoch": 17.45, "learning_rate": 8.4425885187012e-07, "loss": 0.0024, "step": 56716 }, { "epoch": 17.45, "learning_rate": 8.440584842966471e-07, "loss": 0.0029, "step": 56717 }, { "epoch": 17.45, "learning_rate": 8.438581394549061e-07, "loss": 0.0058, "step": 56718 }, { "epoch": 17.45, "learning_rate": 8.43657817345398e-07, "loss": 0.0064, "step": 56719 }, { "epoch": 17.45, "learning_rate": 8.434575179686178e-07, "loss": 0.0149, "step": 56720 }, { "epoch": 17.45, "learning_rate": 8.43257241325064e-07, "loss": 0.005, "step": 56721 }, { "epoch": 17.45, "learning_rate": 8.430569874152339e-07, "loss": 0.0068, "step": 56722 }, { "epoch": 17.45, "learning_rate": 8.42856756239625e-07, "loss": 0.0035, "step": 56723 }, { "epoch": 17.45, "learning_rate": 8.426565477987347e-07, "loss": 0.0037, "step": 56724 }, { "epoch": 17.45, "learning_rate": 8.424563620930615e-07, "loss": 0.0055, "step": 56725 }, { "epoch": 17.45, "learning_rate": 8.42256199123096e-07, "loss": 0.0036, "step": 56726 }, { "epoch": 17.45, "learning_rate": 8.420560588893412e-07, "loss": 0.0061, "step": 56727 }, { "epoch": 17.45, "learning_rate": 8.418559413922933e-07, "loss": 0.0057, "step": 56728 }, { "epoch": 17.45, "learning_rate": 8.416558466324476e-07, "loss": 0.0053, "step": 56729 }, { "epoch": 17.46, "learning_rate": 8.414557746103014e-07, "loss": 0.0052, "step": 56730 }, { "epoch": 17.46, "learning_rate": 8.412557253263509e-07, "loss": 0.0037, "step": 56731 }, { "epoch": 17.46, "learning_rate": 8.410556987810937e-07, "loss": 0.0018, "step": 56732 }, { "epoch": 17.46, "learning_rate": 8.408556949750268e-07, "loss": 0.0039, "step": 56733 }, { "epoch": 17.46, "learning_rate": 8.406557139086447e-07, "loss": 0.0028, "step": 56734 }, { "epoch": 17.46, "learning_rate": 8.404557555824455e-07, "loss": 0.0037, "step": 56735 }, { "epoch": 17.46, "learning_rate": 8.402558199969246e-07, "loss": 0.005, "step": 56736 }, { "epoch": 17.46, "learning_rate": 8.400559071525794e-07, "loss": 0.0041, "step": 56737 }, { "epoch": 17.46, "learning_rate": 8.398560170499093e-07, "loss": 0.0033, "step": 56738 }, { "epoch": 17.46, "learning_rate": 8.39656149689404e-07, "loss": 0.0041, "step": 56739 }, { "epoch": 17.46, "learning_rate": 8.394563050715631e-07, "loss": 0.004, "step": 56740 }, { "epoch": 17.46, "learning_rate": 8.392564831968819e-07, "loss": 0.0038, "step": 56741 }, { "epoch": 17.46, "learning_rate": 8.390566840658576e-07, "loss": 0.0027, "step": 56742 }, { "epoch": 17.46, "learning_rate": 8.388569076789866e-07, "loss": 0.0045, "step": 56743 }, { "epoch": 17.46, "learning_rate": 8.38657154036765e-07, "loss": 0.0037, "step": 56744 }, { "epoch": 17.46, "learning_rate": 8.384574231396869e-07, "loss": 0.0038, "step": 56745 }, { "epoch": 17.46, "learning_rate": 8.382577149882487e-07, "loss": 0.002, "step": 56746 }, { "epoch": 17.46, "learning_rate": 8.380580295829466e-07, "loss": 0.0038, "step": 56747 }, { "epoch": 17.46, "learning_rate": 8.378583669242779e-07, "loss": 0.0025, "step": 56748 }, { "epoch": 17.46, "learning_rate": 8.376587270127356e-07, "loss": 0.0062, "step": 56749 }, { "epoch": 17.46, "learning_rate": 8.374591098488171e-07, "loss": 0.0036, "step": 56750 }, { "epoch": 17.46, "learning_rate": 8.372595154330165e-07, "loss": 0.0051, "step": 56751 }, { "epoch": 17.46, "learning_rate": 8.370599437658311e-07, "loss": 0.0047, "step": 56752 }, { "epoch": 17.46, "learning_rate": 8.368603948477561e-07, "loss": 0.0031, "step": 56753 }, { "epoch": 17.46, "learning_rate": 8.366608686792854e-07, "loss": 0.0078, "step": 56754 }, { "epoch": 17.46, "learning_rate": 8.364613652609166e-07, "loss": 0.0041, "step": 56755 }, { "epoch": 17.46, "learning_rate": 8.362618845931414e-07, "loss": 0.0037, "step": 56756 }, { "epoch": 17.46, "learning_rate": 8.360624266764583e-07, "loss": 0.005, "step": 56757 }, { "epoch": 17.46, "learning_rate": 8.358629915113637e-07, "loss": 0.0025, "step": 56758 }, { "epoch": 17.46, "learning_rate": 8.356635790983503e-07, "loss": 0.0039, "step": 56759 }, { "epoch": 17.46, "learning_rate": 8.354641894379134e-07, "loss": 0.0054, "step": 56760 }, { "epoch": 17.46, "learning_rate": 8.35264822530546e-07, "loss": 0.006, "step": 56761 }, { "epoch": 17.46, "learning_rate": 8.350654783767475e-07, "loss": 0.0047, "step": 56762 }, { "epoch": 17.47, "learning_rate": 8.3486615697701e-07, "loss": 0.0036, "step": 56763 }, { "epoch": 17.47, "learning_rate": 8.346668583318285e-07, "loss": 0.0042, "step": 56764 }, { "epoch": 17.47, "learning_rate": 8.344675824416993e-07, "loss": 0.0052, "step": 56765 }, { "epoch": 17.47, "learning_rate": 8.342683293071164e-07, "loss": 0.0024, "step": 56766 }, { "epoch": 17.47, "learning_rate": 8.340690989285727e-07, "loss": 0.0064, "step": 56767 }, { "epoch": 17.47, "learning_rate": 8.338698913065657e-07, "loss": 0.0029, "step": 56768 }, { "epoch": 17.47, "learning_rate": 8.336707064415883e-07, "loss": 0.0028, "step": 56769 }, { "epoch": 17.47, "learning_rate": 8.334715443341357e-07, "loss": 0.0046, "step": 56770 }, { "epoch": 17.47, "learning_rate": 8.332724049846996e-07, "loss": 0.0067, "step": 56771 }, { "epoch": 17.47, "learning_rate": 8.330732883937808e-07, "loss": 0.0057, "step": 56772 }, { "epoch": 17.47, "learning_rate": 8.32874194561868e-07, "loss": 0.0039, "step": 56773 }, { "epoch": 17.47, "learning_rate": 8.32675123489457e-07, "loss": 0.004, "step": 56774 }, { "epoch": 17.47, "learning_rate": 8.324760751770422e-07, "loss": 0.0047, "step": 56775 }, { "epoch": 17.47, "learning_rate": 8.322770496251153e-07, "loss": 0.0038, "step": 56776 }, { "epoch": 17.47, "learning_rate": 8.320780468341761e-07, "loss": 0.0035, "step": 56777 }, { "epoch": 17.47, "learning_rate": 8.31879066804715e-07, "loss": 0.0032, "step": 56778 }, { "epoch": 17.47, "learning_rate": 8.316801095372263e-07, "loss": 0.0049, "step": 56779 }, { "epoch": 17.47, "learning_rate": 8.31481175032206e-07, "loss": 0.004, "step": 56780 }, { "epoch": 17.47, "learning_rate": 8.312822632901429e-07, "loss": 0.0065, "step": 56781 }, { "epoch": 17.47, "learning_rate": 8.310833743115354e-07, "loss": 0.0039, "step": 56782 }, { "epoch": 17.47, "learning_rate": 8.308845080968764e-07, "loss": 0.0032, "step": 56783 }, { "epoch": 17.47, "learning_rate": 8.306856646466588e-07, "loss": 0.0041, "step": 56784 }, { "epoch": 17.47, "learning_rate": 8.304868439613778e-07, "loss": 0.0032, "step": 56785 }, { "epoch": 17.47, "learning_rate": 8.302880460415252e-07, "loss": 0.0073, "step": 56786 }, { "epoch": 17.47, "learning_rate": 8.300892708875952e-07, "loss": 0.0029, "step": 56787 }, { "epoch": 17.47, "learning_rate": 8.298905185000816e-07, "loss": 0.009, "step": 56788 }, { "epoch": 17.47, "learning_rate": 8.296917888794775e-07, "loss": 0.0037, "step": 56789 }, { "epoch": 17.47, "learning_rate": 8.29493082026277e-07, "loss": 0.006, "step": 56790 }, { "epoch": 17.47, "learning_rate": 8.292943979409706e-07, "loss": 0.004, "step": 56791 }, { "epoch": 17.47, "learning_rate": 8.29095736624056e-07, "loss": 0.0066, "step": 56792 }, { "epoch": 17.47, "learning_rate": 8.288970980760258e-07, "loss": 0.0034, "step": 56793 }, { "epoch": 17.47, "learning_rate": 8.286984822973699e-07, "loss": 0.0058, "step": 56794 }, { "epoch": 17.48, "learning_rate": 8.284998892885832e-07, "loss": 0.0024, "step": 56795 }, { "epoch": 17.48, "learning_rate": 8.283013190501565e-07, "loss": 0.0044, "step": 56796 }, { "epoch": 17.48, "learning_rate": 8.281027715825873e-07, "loss": 0.0079, "step": 56797 }, { "epoch": 17.48, "learning_rate": 8.279042468863663e-07, "loss": 0.0023, "step": 56798 }, { "epoch": 17.48, "learning_rate": 8.277057449619852e-07, "loss": 0.0049, "step": 56799 }, { "epoch": 17.48, "learning_rate": 8.275072658099393e-07, "loss": 0.0047, "step": 56800 }, { "epoch": 17.48, "learning_rate": 8.273088094307191e-07, "loss": 0.0042, "step": 56801 }, { "epoch": 17.48, "learning_rate": 8.271103758248189e-07, "loss": 0.0032, "step": 56802 }, { "epoch": 17.48, "learning_rate": 8.269119649927315e-07, "loss": 0.0036, "step": 56803 }, { "epoch": 17.48, "learning_rate": 8.267135769349477e-07, "loss": 0.0039, "step": 56804 }, { "epoch": 17.48, "learning_rate": 8.265152116519604e-07, "loss": 0.0047, "step": 56805 }, { "epoch": 17.48, "learning_rate": 8.263168691442624e-07, "loss": 0.0036, "step": 56806 }, { "epoch": 17.48, "learning_rate": 8.261185494123492e-07, "loss": 0.0024, "step": 56807 }, { "epoch": 17.48, "learning_rate": 8.259202524567089e-07, "loss": 0.0039, "step": 56808 }, { "epoch": 17.48, "learning_rate": 8.257219782778347e-07, "loss": 0.005, "step": 56809 }, { "epoch": 17.48, "learning_rate": 8.255237268762206e-07, "loss": 0.0069, "step": 56810 }, { "epoch": 17.48, "learning_rate": 8.253254982523551e-07, "loss": 0.009, "step": 56811 }, { "epoch": 17.48, "learning_rate": 8.251272924067355e-07, "loss": 0.0041, "step": 56812 }, { "epoch": 17.48, "learning_rate": 8.249291093398503e-07, "loss": 0.0057, "step": 56813 }, { "epoch": 17.48, "learning_rate": 8.247309490521949e-07, "loss": 0.0039, "step": 56814 }, { "epoch": 17.48, "learning_rate": 8.245328115442574e-07, "loss": 0.0051, "step": 56815 }, { "epoch": 17.48, "learning_rate": 8.2433469681653e-07, "loss": 0.0042, "step": 56816 }, { "epoch": 17.48, "learning_rate": 8.241366048695065e-07, "loss": 0.0033, "step": 56817 }, { "epoch": 17.48, "learning_rate": 8.239385357036789e-07, "loss": 0.0035, "step": 56818 }, { "epoch": 17.48, "learning_rate": 8.237404893195377e-07, "loss": 0.0044, "step": 56819 }, { "epoch": 17.48, "learning_rate": 8.23542465717575e-07, "loss": 0.0043, "step": 56820 }, { "epoch": 17.48, "learning_rate": 8.233444648982824e-07, "loss": 0.006, "step": 56821 }, { "epoch": 17.48, "learning_rate": 8.231464868621519e-07, "loss": 0.0035, "step": 56822 }, { "epoch": 17.48, "learning_rate": 8.229485316096752e-07, "loss": 0.0027, "step": 56823 }, { "epoch": 17.48, "learning_rate": 8.22750599141342e-07, "loss": 0.0071, "step": 56824 }, { "epoch": 17.48, "learning_rate": 8.225526894576452e-07, "loss": 0.0032, "step": 56825 }, { "epoch": 17.48, "learning_rate": 8.223548025590755e-07, "loss": 0.006, "step": 56826 }, { "epoch": 17.48, "learning_rate": 8.221569384461247e-07, "loss": 0.0064, "step": 56827 }, { "epoch": 17.49, "learning_rate": 8.21959097119287e-07, "loss": 0.0056, "step": 56828 }, { "epoch": 17.49, "learning_rate": 8.217612785790474e-07, "loss": 0.0038, "step": 56829 }, { "epoch": 17.49, "learning_rate": 8.215634828259012e-07, "loss": 0.0082, "step": 56830 }, { "epoch": 17.49, "learning_rate": 8.213657098603367e-07, "loss": 0.0048, "step": 56831 }, { "epoch": 17.49, "learning_rate": 8.211679596828481e-07, "loss": 0.005, "step": 56832 }, { "epoch": 17.49, "learning_rate": 8.209702322939251e-07, "loss": 0.0055, "step": 56833 }, { "epoch": 17.49, "learning_rate": 8.207725276940582e-07, "loss": 0.0036, "step": 56834 }, { "epoch": 17.49, "learning_rate": 8.205748458837403e-07, "loss": 0.004, "step": 56835 }, { "epoch": 17.49, "learning_rate": 8.203771868634569e-07, "loss": 0.0042, "step": 56836 }, { "epoch": 17.49, "learning_rate": 8.201795506337029e-07, "loss": 0.0036, "step": 56837 }, { "epoch": 17.49, "learning_rate": 8.199819371949703e-07, "loss": 0.0046, "step": 56838 }, { "epoch": 17.49, "learning_rate": 8.197843465477462e-07, "loss": 0.0049, "step": 56839 }, { "epoch": 17.49, "learning_rate": 8.195867786925227e-07, "loss": 0.0045, "step": 56840 }, { "epoch": 17.49, "learning_rate": 8.193892336297915e-07, "loss": 0.0035, "step": 56841 }, { "epoch": 17.49, "learning_rate": 8.191917113600401e-07, "loss": 0.0057, "step": 56842 }, { "epoch": 17.49, "learning_rate": 8.189942118837624e-07, "loss": 0.0033, "step": 56843 }, { "epoch": 17.49, "learning_rate": 8.187967352014458e-07, "loss": 0.0053, "step": 56844 }, { "epoch": 17.49, "learning_rate": 8.185992813135813e-07, "loss": 0.0058, "step": 56845 }, { "epoch": 17.49, "learning_rate": 8.184018502206592e-07, "loss": 0.0038, "step": 56846 }, { "epoch": 17.49, "learning_rate": 8.182044419231705e-07, "loss": 0.0046, "step": 56847 }, { "epoch": 17.49, "learning_rate": 8.180070564216058e-07, "loss": 0.0048, "step": 56848 }, { "epoch": 17.49, "learning_rate": 8.178096937164559e-07, "loss": 0.0035, "step": 56849 }, { "epoch": 17.49, "learning_rate": 8.17612353808207e-07, "loss": 0.0031, "step": 56850 }, { "epoch": 17.49, "learning_rate": 8.174150366973488e-07, "loss": 0.0072, "step": 56851 }, { "epoch": 17.49, "learning_rate": 8.172177423843763e-07, "loss": 0.0037, "step": 56852 }, { "epoch": 17.49, "learning_rate": 8.17020470869776e-07, "loss": 0.0038, "step": 56853 }, { "epoch": 17.49, "learning_rate": 8.168232221540385e-07, "loss": 0.0042, "step": 56854 }, { "epoch": 17.49, "learning_rate": 8.166259962376521e-07, "loss": 0.0038, "step": 56855 }, { "epoch": 17.49, "learning_rate": 8.16428793121109e-07, "loss": 0.0027, "step": 56856 }, { "epoch": 17.49, "learning_rate": 8.162316128048964e-07, "loss": 0.0054, "step": 56857 }, { "epoch": 17.49, "learning_rate": 8.160344552895061e-07, "loss": 0.0024, "step": 56858 }, { "epoch": 17.49, "learning_rate": 8.158373205754255e-07, "loss": 0.0046, "step": 56859 }, { "epoch": 17.5, "learning_rate": 8.156402086631443e-07, "loss": 0.0022, "step": 56860 }, { "epoch": 17.5, "learning_rate": 8.15443119553152e-07, "loss": 0.0034, "step": 56861 }, { "epoch": 17.5, "learning_rate": 8.152460532459405e-07, "loss": 0.0046, "step": 56862 }, { "epoch": 17.5, "learning_rate": 8.15049009741996e-07, "loss": 0.0038, "step": 56863 }, { "epoch": 17.5, "learning_rate": 8.148519890418083e-07, "loss": 0.0044, "step": 56864 }, { "epoch": 17.5, "learning_rate": 8.146549911458668e-07, "loss": 0.0035, "step": 56865 }, { "epoch": 17.5, "learning_rate": 8.1445801605466e-07, "loss": 0.0057, "step": 56866 }, { "epoch": 17.5, "learning_rate": 8.142610637686776e-07, "loss": 0.0054, "step": 56867 }, { "epoch": 17.5, "learning_rate": 8.140641342884104e-07, "loss": 0.0033, "step": 56868 }, { "epoch": 17.5, "learning_rate": 8.138672276143445e-07, "loss": 0.0033, "step": 56869 }, { "epoch": 17.5, "learning_rate": 8.136703437469717e-07, "loss": 0.0068, "step": 56870 }, { "epoch": 17.5, "learning_rate": 8.134734826867751e-07, "loss": 0.0024, "step": 56871 }, { "epoch": 17.5, "learning_rate": 8.132766444342488e-07, "loss": 0.0037, "step": 56872 }, { "epoch": 17.5, "learning_rate": 8.130798289898801e-07, "loss": 0.0045, "step": 56873 }, { "epoch": 17.5, "learning_rate": 8.128830363541574e-07, "loss": 0.0051, "step": 56874 }, { "epoch": 17.5, "learning_rate": 8.126862665275704e-07, "loss": 0.003, "step": 56875 }, { "epoch": 17.5, "learning_rate": 8.124895195106053e-07, "loss": 0.0054, "step": 56876 }, { "epoch": 17.5, "learning_rate": 8.12292795303753e-07, "loss": 0.0051, "step": 56877 }, { "epoch": 17.5, "learning_rate": 8.120960939074996e-07, "loss": 0.0035, "step": 56878 }, { "epoch": 17.5, "learning_rate": 8.118994153223347e-07, "loss": 0.0059, "step": 56879 }, { "epoch": 17.5, "learning_rate": 8.11702759548747e-07, "loss": 0.0038, "step": 56880 }, { "epoch": 17.5, "learning_rate": 8.115061265872226e-07, "loss": 0.0041, "step": 56881 }, { "epoch": 17.5, "learning_rate": 8.113095164382534e-07, "loss": 0.0038, "step": 56882 }, { "epoch": 17.5, "learning_rate": 8.111129291023267e-07, "loss": 0.0042, "step": 56883 }, { "epoch": 17.5, "learning_rate": 8.109163645799267e-07, "loss": 0.0024, "step": 56884 }, { "epoch": 17.5, "learning_rate": 8.107198228715452e-07, "loss": 0.0053, "step": 56885 }, { "epoch": 17.5, "learning_rate": 8.105233039776672e-07, "loss": 0.0041, "step": 56886 }, { "epoch": 17.5, "learning_rate": 8.103268078987847e-07, "loss": 0.0042, "step": 56887 }, { "epoch": 17.5, "learning_rate": 8.101303346353817e-07, "loss": 0.0037, "step": 56888 }, { "epoch": 17.5, "learning_rate": 8.099338841879489e-07, "loss": 0.0043, "step": 56889 }, { "epoch": 17.5, "learning_rate": 8.097374565569738e-07, "loss": 0.0045, "step": 56890 }, { "epoch": 17.5, "learning_rate": 8.095410517429403e-07, "loss": 0.0022, "step": 56891 }, { "epoch": 17.5, "learning_rate": 8.093446697463392e-07, "loss": 0.0058, "step": 56892 }, { "epoch": 17.51, "learning_rate": 8.09148310567659e-07, "loss": 0.0047, "step": 56893 }, { "epoch": 17.51, "learning_rate": 8.089519742073848e-07, "loss": 0.0049, "step": 56894 }, { "epoch": 17.51, "learning_rate": 8.087556606660051e-07, "loss": 0.0025, "step": 56895 }, { "epoch": 17.51, "learning_rate": 8.085593699440075e-07, "loss": 0.0075, "step": 56896 }, { "epoch": 17.51, "learning_rate": 8.083631020418792e-07, "loss": 0.0042, "step": 56897 }, { "epoch": 17.51, "learning_rate": 8.081668569601064e-07, "loss": 0.0029, "step": 56898 }, { "epoch": 17.51, "learning_rate": 8.079706346991778e-07, "loss": 0.0037, "step": 56899 }, { "epoch": 17.51, "learning_rate": 8.077744352595807e-07, "loss": 0.0046, "step": 56900 }, { "epoch": 17.51, "learning_rate": 8.075782586417991e-07, "loss": 0.0066, "step": 56901 }, { "epoch": 17.51, "learning_rate": 8.073821048463237e-07, "loss": 0.0038, "step": 56902 }, { "epoch": 17.51, "learning_rate": 8.071859738736409e-07, "loss": 0.0039, "step": 56903 }, { "epoch": 17.51, "learning_rate": 8.069898657242393e-07, "loss": 0.0047, "step": 56904 }, { "epoch": 17.51, "learning_rate": 8.067937803986015e-07, "loss": 0.0036, "step": 56905 }, { "epoch": 17.51, "learning_rate": 8.065977178972139e-07, "loss": 0.0026, "step": 56906 }, { "epoch": 17.51, "learning_rate": 8.064016782205675e-07, "loss": 0.003, "step": 56907 }, { "epoch": 17.51, "learning_rate": 8.062056613691483e-07, "loss": 0.0046, "step": 56908 }, { "epoch": 17.51, "learning_rate": 8.060096673434403e-07, "loss": 0.0041, "step": 56909 }, { "epoch": 17.51, "learning_rate": 8.058136961439333e-07, "loss": 0.0047, "step": 56910 }, { "epoch": 17.51, "learning_rate": 8.056177477711113e-07, "loss": 0.0035, "step": 56911 }, { "epoch": 17.51, "learning_rate": 8.054218222254618e-07, "loss": 0.0034, "step": 56912 }, { "epoch": 17.51, "learning_rate": 8.052259195074707e-07, "loss": 0.0048, "step": 56913 }, { "epoch": 17.51, "learning_rate": 8.050300396176259e-07, "loss": 0.0036, "step": 56914 }, { "epoch": 17.51, "learning_rate": 8.04834182556411e-07, "loss": 0.0029, "step": 56915 }, { "epoch": 17.51, "learning_rate": 8.046383483243136e-07, "loss": 0.0035, "step": 56916 }, { "epoch": 17.51, "learning_rate": 8.044425369218234e-07, "loss": 0.0048, "step": 56917 }, { "epoch": 17.51, "learning_rate": 8.042467483494209e-07, "loss": 0.0028, "step": 56918 }, { "epoch": 17.51, "learning_rate": 8.040509826075959e-07, "loss": 0.0045, "step": 56919 }, { "epoch": 17.51, "learning_rate": 8.038552396968324e-07, "loss": 0.0052, "step": 56920 }, { "epoch": 17.51, "learning_rate": 8.036595196176144e-07, "loss": 0.0043, "step": 56921 }, { "epoch": 17.51, "learning_rate": 8.034638223704338e-07, "loss": 0.005, "step": 56922 }, { "epoch": 17.51, "learning_rate": 8.032681479557725e-07, "loss": 0.0035, "step": 56923 }, { "epoch": 17.51, "learning_rate": 8.030724963741177e-07, "loss": 0.0033, "step": 56924 }, { "epoch": 17.52, "learning_rate": 8.028768676259558e-07, "loss": 0.0084, "step": 56925 }, { "epoch": 17.52, "learning_rate": 8.026812617117675e-07, "loss": 0.0029, "step": 56926 }, { "epoch": 17.52, "learning_rate": 8.024856786320434e-07, "loss": 0.0044, "step": 56927 }, { "epoch": 17.52, "learning_rate": 8.022901183872689e-07, "loss": 0.0043, "step": 56928 }, { "epoch": 17.52, "learning_rate": 8.02094580977928e-07, "loss": 0.0051, "step": 56929 }, { "epoch": 17.52, "learning_rate": 8.018990664045057e-07, "loss": 0.0036, "step": 56930 }, { "epoch": 17.52, "learning_rate": 8.017035746674895e-07, "loss": 0.0043, "step": 56931 }, { "epoch": 17.52, "learning_rate": 8.015081057673623e-07, "loss": 0.0024, "step": 56932 }, { "epoch": 17.52, "learning_rate": 8.013126597046117e-07, "loss": 0.0051, "step": 56933 }, { "epoch": 17.52, "learning_rate": 8.011172364797216e-07, "loss": 0.0018, "step": 56934 }, { "epoch": 17.52, "learning_rate": 8.009218360931781e-07, "loss": 0.0029, "step": 56935 }, { "epoch": 17.52, "learning_rate": 8.007264585454632e-07, "loss": 0.0051, "step": 56936 }, { "epoch": 17.52, "learning_rate": 8.005311038370656e-07, "loss": 0.0028, "step": 56937 }, { "epoch": 17.52, "learning_rate": 8.003357719684723e-07, "loss": 0.0032, "step": 56938 }, { "epoch": 17.52, "learning_rate": 8.001404629401621e-07, "loss": 0.0045, "step": 56939 }, { "epoch": 17.52, "learning_rate": 7.999451767526234e-07, "loss": 0.0095, "step": 56940 }, { "epoch": 17.52, "learning_rate": 7.997499134063391e-07, "loss": 0.0055, "step": 56941 }, { "epoch": 17.52, "learning_rate": 7.995546729017978e-07, "loss": 0.004, "step": 56942 }, { "epoch": 17.52, "learning_rate": 7.993594552394802e-07, "loss": 0.0033, "step": 56943 }, { "epoch": 17.52, "learning_rate": 7.991642604198734e-07, "loss": 0.0055, "step": 56944 }, { "epoch": 17.52, "learning_rate": 7.98969088443462e-07, "loss": 0.0065, "step": 56945 }, { "epoch": 17.52, "learning_rate": 7.987739393107297e-07, "loss": 0.0038, "step": 56946 }, { "epoch": 17.52, "learning_rate": 7.985788130221606e-07, "loss": 0.0037, "step": 56947 }, { "epoch": 17.52, "learning_rate": 7.983837095782399e-07, "loss": 0.0047, "step": 56948 }, { "epoch": 17.52, "learning_rate": 7.981886289794516e-07, "loss": 0.0037, "step": 56949 }, { "epoch": 17.52, "learning_rate": 7.979935712262776e-07, "loss": 0.0043, "step": 56950 }, { "epoch": 17.52, "learning_rate": 7.977985363192098e-07, "loss": 0.003, "step": 56951 }, { "epoch": 17.52, "learning_rate": 7.976035242587254e-07, "loss": 0.0029, "step": 56952 }, { "epoch": 17.52, "learning_rate": 7.974085350453098e-07, "loss": 0.0039, "step": 56953 }, { "epoch": 17.52, "learning_rate": 7.97213568679448e-07, "loss": 0.004, "step": 56954 }, { "epoch": 17.52, "learning_rate": 7.97018625161623e-07, "loss": 0.0051, "step": 56955 }, { "epoch": 17.52, "learning_rate": 7.9682370449232e-07, "loss": 0.0036, "step": 56956 }, { "epoch": 17.52, "learning_rate": 7.96628806672024e-07, "loss": 0.004, "step": 56957 }, { "epoch": 17.53, "learning_rate": 7.96433931701217e-07, "loss": 0.0046, "step": 56958 }, { "epoch": 17.53, "learning_rate": 7.962390795803854e-07, "loss": 0.0025, "step": 56959 }, { "epoch": 17.53, "learning_rate": 7.960442503100075e-07, "loss": 0.0031, "step": 56960 }, { "epoch": 17.53, "learning_rate": 7.958494438905718e-07, "loss": 0.004, "step": 56961 }, { "epoch": 17.53, "learning_rate": 7.956546603225601e-07, "loss": 0.0048, "step": 56962 }, { "epoch": 17.53, "learning_rate": 7.954598996064566e-07, "loss": 0.0053, "step": 56963 }, { "epoch": 17.53, "learning_rate": 7.952651617427454e-07, "loss": 0.0071, "step": 56964 }, { "epoch": 17.53, "learning_rate": 7.950704467319093e-07, "loss": 0.0032, "step": 56965 }, { "epoch": 17.53, "learning_rate": 7.948757545744312e-07, "loss": 0.0038, "step": 56966 }, { "epoch": 17.53, "learning_rate": 7.946810852707942e-07, "loss": 0.0052, "step": 56967 }, { "epoch": 17.53, "learning_rate": 7.944864388214835e-07, "loss": 0.0037, "step": 56968 }, { "epoch": 17.53, "learning_rate": 7.942918152269819e-07, "loss": 0.0052, "step": 56969 }, { "epoch": 17.53, "learning_rate": 7.94097214487769e-07, "loss": 0.0051, "step": 56970 }, { "epoch": 17.53, "learning_rate": 7.939026366043323e-07, "loss": 0.0033, "step": 56971 }, { "epoch": 17.53, "learning_rate": 7.937080815771569e-07, "loss": 0.0031, "step": 56972 }, { "epoch": 17.53, "learning_rate": 7.935135494067192e-07, "loss": 0.0028, "step": 56973 }, { "epoch": 17.53, "learning_rate": 7.933190400935054e-07, "loss": 0.0039, "step": 56974 }, { "epoch": 17.53, "learning_rate": 7.931245536379973e-07, "loss": 0.0036, "step": 56975 }, { "epoch": 17.53, "learning_rate": 7.9293009004068e-07, "loss": 0.0065, "step": 56976 }, { "epoch": 17.53, "learning_rate": 7.927356493020355e-07, "loss": 0.0057, "step": 56977 }, { "epoch": 17.53, "learning_rate": 7.925412314225467e-07, "loss": 0.0049, "step": 56978 }, { "epoch": 17.53, "learning_rate": 7.923468364026954e-07, "loss": 0.0029, "step": 56979 }, { "epoch": 17.53, "learning_rate": 7.921524642429634e-07, "loss": 0.0046, "step": 56980 }, { "epoch": 17.53, "learning_rate": 7.91958114943836e-07, "loss": 0.0054, "step": 56981 }, { "epoch": 17.53, "learning_rate": 7.917637885057938e-07, "loss": 0.0048, "step": 56982 }, { "epoch": 17.53, "learning_rate": 7.915694849293199e-07, "loss": 0.0048, "step": 56983 }, { "epoch": 17.53, "learning_rate": 7.913752042148959e-07, "loss": 0.0033, "step": 56984 }, { "epoch": 17.53, "learning_rate": 7.911809463630027e-07, "loss": 0.0034, "step": 56985 }, { "epoch": 17.53, "learning_rate": 7.909867113741287e-07, "loss": 0.004, "step": 56986 }, { "epoch": 17.53, "learning_rate": 7.907924992487503e-07, "loss": 0.0045, "step": 56987 }, { "epoch": 17.53, "learning_rate": 7.905983099873504e-07, "loss": 0.004, "step": 56988 }, { "epoch": 17.53, "learning_rate": 7.904041435904131e-07, "loss": 0.0038, "step": 56989 }, { "epoch": 17.54, "learning_rate": 7.902100000584179e-07, "loss": 0.0051, "step": 56990 }, { "epoch": 17.54, "learning_rate": 7.9001587939185e-07, "loss": 0.0053, "step": 56991 }, { "epoch": 17.54, "learning_rate": 7.89821781591189e-07, "loss": 0.0035, "step": 56992 }, { "epoch": 17.54, "learning_rate": 7.896277066569203e-07, "loss": 0.0046, "step": 56993 }, { "epoch": 17.54, "learning_rate": 7.894336545895198e-07, "loss": 0.0033, "step": 56994 }, { "epoch": 17.54, "learning_rate": 7.892396253894718e-07, "loss": 0.0029, "step": 56995 }, { "epoch": 17.54, "learning_rate": 7.890456190572604e-07, "loss": 0.0045, "step": 56996 }, { "epoch": 17.54, "learning_rate": 7.888516355933651e-07, "loss": 0.0036, "step": 56997 }, { "epoch": 17.54, "learning_rate": 7.886576749982678e-07, "loss": 0.0054, "step": 56998 }, { "epoch": 17.54, "learning_rate": 7.884637372724513e-07, "loss": 0.006, "step": 56999 }, { "epoch": 17.54, "learning_rate": 7.882698224163954e-07, "loss": 0.0035, "step": 57000 }, { "epoch": 17.54, "learning_rate": 7.880759304305818e-07, "loss": 0.0049, "step": 57001 }, { "epoch": 17.54, "learning_rate": 7.878820613154925e-07, "loss": 0.0038, "step": 57002 }, { "epoch": 17.54, "learning_rate": 7.876882150716081e-07, "loss": 0.0016, "step": 57003 }, { "epoch": 17.54, "learning_rate": 7.874943916994104e-07, "loss": 0.0052, "step": 57004 }, { "epoch": 17.54, "learning_rate": 7.873005911993792e-07, "loss": 0.0032, "step": 57005 }, { "epoch": 17.54, "learning_rate": 7.871068135719984e-07, "loss": 0.0044, "step": 57006 }, { "epoch": 17.54, "learning_rate": 7.869130588177498e-07, "loss": 0.0043, "step": 57007 }, { "epoch": 17.54, "learning_rate": 7.867193269371099e-07, "loss": 0.0054, "step": 57008 }, { "epoch": 17.54, "learning_rate": 7.865256179305624e-07, "loss": 0.0028, "step": 57009 }, { "epoch": 17.54, "learning_rate": 7.863319317985863e-07, "loss": 0.0026, "step": 57010 }, { "epoch": 17.54, "learning_rate": 7.861382685416652e-07, "loss": 0.0052, "step": 57011 }, { "epoch": 17.54, "learning_rate": 7.859446281602801e-07, "loss": 0.0069, "step": 57012 }, { "epoch": 17.54, "learning_rate": 7.857510106549094e-07, "loss": 0.0038, "step": 57013 }, { "epoch": 17.54, "learning_rate": 7.855574160260371e-07, "loss": 0.0043, "step": 57014 }, { "epoch": 17.54, "learning_rate": 7.853638442741374e-07, "loss": 0.0063, "step": 57015 }, { "epoch": 17.54, "learning_rate": 7.851702953996975e-07, "loss": 0.0026, "step": 57016 }, { "epoch": 17.54, "learning_rate": 7.849767694031951e-07, "loss": 0.004, "step": 57017 }, { "epoch": 17.54, "learning_rate": 7.847832662851118e-07, "loss": 0.0067, "step": 57018 }, { "epoch": 17.54, "learning_rate": 7.845897860459262e-07, "loss": 0.0041, "step": 57019 }, { "epoch": 17.54, "learning_rate": 7.843963286861211e-07, "loss": 0.0033, "step": 57020 }, { "epoch": 17.54, "learning_rate": 7.842028942061741e-07, "loss": 0.0038, "step": 57021 }, { "epoch": 17.54, "learning_rate": 7.84009482606568e-07, "loss": 0.0038, "step": 57022 }, { "epoch": 17.55, "learning_rate": 7.838160938877814e-07, "loss": 0.0062, "step": 57023 }, { "epoch": 17.55, "learning_rate": 7.836227280502939e-07, "loss": 0.0049, "step": 57024 }, { "epoch": 17.55, "learning_rate": 7.834293850945862e-07, "loss": 0.0041, "step": 57025 }, { "epoch": 17.55, "learning_rate": 7.832360650211402e-07, "loss": 0.0086, "step": 57026 }, { "epoch": 17.55, "learning_rate": 7.830427678304353e-07, "loss": 0.0022, "step": 57027 }, { "epoch": 17.55, "learning_rate": 7.828494935229492e-07, "loss": 0.0033, "step": 57028 }, { "epoch": 17.55, "learning_rate": 7.826562420991624e-07, "loss": 0.0034, "step": 57029 }, { "epoch": 17.55, "learning_rate": 7.824630135595535e-07, "loss": 0.0051, "step": 57030 }, { "epoch": 17.55, "learning_rate": 7.822698079046065e-07, "loss": 0.0056, "step": 57031 }, { "epoch": 17.55, "learning_rate": 7.820766251347978e-07, "loss": 0.0032, "step": 57032 }, { "epoch": 17.55, "learning_rate": 7.81883465250608e-07, "loss": 0.0037, "step": 57033 }, { "epoch": 17.55, "learning_rate": 7.816903282525157e-07, "loss": 0.0036, "step": 57034 }, { "epoch": 17.55, "learning_rate": 7.814972141410026e-07, "loss": 0.0025, "step": 57035 }, { "epoch": 17.55, "learning_rate": 7.813041229165452e-07, "loss": 0.0044, "step": 57036 }, { "epoch": 17.55, "learning_rate": 7.811110545796252e-07, "loss": 0.0034, "step": 57037 }, { "epoch": 17.55, "learning_rate": 7.809180091307211e-07, "loss": 0.0035, "step": 57038 }, { "epoch": 17.55, "learning_rate": 7.807249865703126e-07, "loss": 0.0033, "step": 57039 }, { "epoch": 17.55, "learning_rate": 7.805319868988759e-07, "loss": 0.0033, "step": 57040 }, { "epoch": 17.55, "learning_rate": 7.803390101168962e-07, "loss": 0.0026, "step": 57041 }, { "epoch": 17.55, "learning_rate": 7.801460562248486e-07, "loss": 0.0027, "step": 57042 }, { "epoch": 17.55, "learning_rate": 7.799531252232118e-07, "loss": 0.0046, "step": 57043 }, { "epoch": 17.55, "learning_rate": 7.797602171124663e-07, "loss": 0.0057, "step": 57044 }, { "epoch": 17.55, "learning_rate": 7.797602171124663e-07, "loss": 0.0071, "step": 57045 }, { "epoch": 17.55, "learning_rate": 7.795673318930886e-07, "loss": 0.0059, "step": 57046 }, { "epoch": 17.55, "learning_rate": 7.793744695655603e-07, "loss": 0.0041, "step": 57047 }, { "epoch": 17.55, "learning_rate": 7.791816301303601e-07, "loss": 0.0039, "step": 57048 }, { "epoch": 17.55, "learning_rate": 7.789888135879675e-07, "loss": 0.0042, "step": 57049 }, { "epoch": 17.55, "learning_rate": 7.787960199388589e-07, "loss": 0.0066, "step": 57050 }, { "epoch": 17.55, "learning_rate": 7.786032491835116e-07, "loss": 0.0033, "step": 57051 }, { "epoch": 17.55, "learning_rate": 7.784105013224086e-07, "loss": 0.003, "step": 57052 }, { "epoch": 17.55, "learning_rate": 7.78217776356025e-07, "loss": 0.0049, "step": 57053 }, { "epoch": 17.55, "learning_rate": 7.780250742848416e-07, "loss": 0.0036, "step": 57054 }, { "epoch": 17.56, "learning_rate": 7.778323951093358e-07, "loss": 0.006, "step": 57055 }, { "epoch": 17.56, "learning_rate": 7.77639738829985e-07, "loss": 0.0028, "step": 57056 }, { "epoch": 17.56, "learning_rate": 7.774471054472688e-07, "loss": 0.003, "step": 57057 }, { "epoch": 17.56, "learning_rate": 7.772544949616656e-07, "loss": 0.0027, "step": 57058 }, { "epoch": 17.56, "learning_rate": 7.77061907373653e-07, "loss": 0.0043, "step": 57059 }, { "epoch": 17.56, "learning_rate": 7.768693426837093e-07, "loss": 0.0042, "step": 57060 }, { "epoch": 17.56, "learning_rate": 7.766768008923109e-07, "loss": 0.0038, "step": 57061 }, { "epoch": 17.56, "learning_rate": 7.764842819999385e-07, "loss": 0.0029, "step": 57062 }, { "epoch": 17.56, "learning_rate": 7.762917860070718e-07, "loss": 0.0043, "step": 57063 }, { "epoch": 17.56, "learning_rate": 7.760993129141836e-07, "loss": 0.0036, "step": 57064 }, { "epoch": 17.56, "learning_rate": 7.759068627217537e-07, "loss": 0.0021, "step": 57065 }, { "epoch": 17.56, "learning_rate": 7.757144354302593e-07, "loss": 0.0039, "step": 57066 }, { "epoch": 17.56, "learning_rate": 7.755220310401812e-07, "loss": 0.0042, "step": 57067 }, { "epoch": 17.56, "learning_rate": 7.753296495519946e-07, "loss": 0.0073, "step": 57068 }, { "epoch": 17.56, "learning_rate": 7.75137290966177e-07, "loss": 0.0061, "step": 57069 }, { "epoch": 17.56, "learning_rate": 7.749449552832089e-07, "loss": 0.0029, "step": 57070 }, { "epoch": 17.56, "learning_rate": 7.747526425035623e-07, "loss": 0.003, "step": 57071 }, { "epoch": 17.56, "learning_rate": 7.74560352627719e-07, "loss": 0.0042, "step": 57072 }, { "epoch": 17.56, "learning_rate": 7.743680856561552e-07, "loss": 0.006, "step": 57073 }, { "epoch": 17.56, "learning_rate": 7.741758415893485e-07, "loss": 0.003, "step": 57074 }, { "epoch": 17.56, "learning_rate": 7.739836204277762e-07, "loss": 0.0036, "step": 57075 }, { "epoch": 17.56, "learning_rate": 7.737914221719145e-07, "loss": 0.0057, "step": 57076 }, { "epoch": 17.56, "learning_rate": 7.73599246822242e-07, "loss": 0.0031, "step": 57077 }, { "epoch": 17.56, "learning_rate": 7.73407094379236e-07, "loss": 0.0041, "step": 57078 }, { "epoch": 17.56, "learning_rate": 7.732149648433718e-07, "loss": 0.0051, "step": 57079 }, { "epoch": 17.56, "learning_rate": 7.730228582151278e-07, "loss": 0.0068, "step": 57080 }, { "epoch": 17.56, "learning_rate": 7.728307744949781e-07, "loss": 0.0033, "step": 57081 }, { "epoch": 17.56, "learning_rate": 7.726387136834046e-07, "loss": 0.0016, "step": 57082 }, { "epoch": 17.56, "learning_rate": 7.724466757808813e-07, "loss": 0.0033, "step": 57083 }, { "epoch": 17.56, "learning_rate": 7.722546607878867e-07, "loss": 0.0025, "step": 57084 }, { "epoch": 17.56, "learning_rate": 7.720626687048938e-07, "loss": 0.0035, "step": 57085 }, { "epoch": 17.56, "learning_rate": 7.71870699532381e-07, "loss": 0.0045, "step": 57086 }, { "epoch": 17.56, "learning_rate": 7.71678753270827e-07, "loss": 0.0034, "step": 57087 }, { "epoch": 17.57, "learning_rate": 7.714868299207057e-07, "loss": 0.0024, "step": 57088 }, { "epoch": 17.57, "learning_rate": 7.712949294824956e-07, "loss": 0.0047, "step": 57089 }, { "epoch": 17.57, "learning_rate": 7.71103051956672e-07, "loss": 0.0043, "step": 57090 }, { "epoch": 17.57, "learning_rate": 7.709111973437111e-07, "loss": 0.0052, "step": 57091 }, { "epoch": 17.57, "learning_rate": 7.707193656440903e-07, "loss": 0.0054, "step": 57092 }, { "epoch": 17.57, "learning_rate": 7.705275568582848e-07, "loss": 0.0032, "step": 57093 }, { "epoch": 17.57, "learning_rate": 7.703357709867709e-07, "loss": 0.0041, "step": 57094 }, { "epoch": 17.57, "learning_rate": 7.70144008030026e-07, "loss": 0.0051, "step": 57095 }, { "epoch": 17.57, "learning_rate": 7.699522679885229e-07, "loss": 0.0042, "step": 57096 }, { "epoch": 17.57, "learning_rate": 7.697605508627426e-07, "loss": 0.0042, "step": 57097 }, { "epoch": 17.57, "learning_rate": 7.695688566531578e-07, "loss": 0.0048, "step": 57098 }, { "epoch": 17.57, "learning_rate": 7.693771853602449e-07, "loss": 0.0096, "step": 57099 }, { "epoch": 17.57, "learning_rate": 7.691855369844803e-07, "loss": 0.0055, "step": 57100 }, { "epoch": 17.57, "learning_rate": 7.689939115263367e-07, "loss": 0.0036, "step": 57101 }, { "epoch": 17.57, "learning_rate": 7.688023089862951e-07, "loss": 0.0036, "step": 57102 }, { "epoch": 17.57, "learning_rate": 7.686107293648282e-07, "loss": 0.0043, "step": 57103 }, { "epoch": 17.57, "learning_rate": 7.684191726624135e-07, "loss": 0.0056, "step": 57104 }, { "epoch": 17.57, "learning_rate": 7.682276388795251e-07, "loss": 0.0033, "step": 57105 }, { "epoch": 17.57, "learning_rate": 7.680361280166359e-07, "loss": 0.0052, "step": 57106 }, { "epoch": 17.57, "learning_rate": 7.678446400742268e-07, "loss": 0.0049, "step": 57107 }, { "epoch": 17.57, "learning_rate": 7.676531750527694e-07, "loss": 0.0025, "step": 57108 }, { "epoch": 17.57, "learning_rate": 7.674617329527412e-07, "loss": 0.0042, "step": 57109 }, { "epoch": 17.57, "learning_rate": 7.672703137746152e-07, "loss": 0.0039, "step": 57110 }, { "epoch": 17.57, "learning_rate": 7.670789175188687e-07, "loss": 0.0038, "step": 57111 }, { "epoch": 17.57, "learning_rate": 7.668875441859758e-07, "loss": 0.0049, "step": 57112 }, { "epoch": 17.57, "learning_rate": 7.666961937764128e-07, "loss": 0.0032, "step": 57113 }, { "epoch": 17.57, "learning_rate": 7.665048662906538e-07, "loss": 0.0014, "step": 57114 }, { "epoch": 17.57, "learning_rate": 7.663135617291728e-07, "loss": 0.0025, "step": 57115 }, { "epoch": 17.57, "learning_rate": 7.661222800924451e-07, "loss": 0.0053, "step": 57116 }, { "epoch": 17.57, "learning_rate": 7.659310213809479e-07, "loss": 0.0149, "step": 57117 }, { "epoch": 17.57, "learning_rate": 7.657397855951565e-07, "loss": 0.0112, "step": 57118 }, { "epoch": 17.57, "learning_rate": 7.655485727355416e-07, "loss": 0.0038, "step": 57119 }, { "epoch": 17.58, "learning_rate": 7.653573828025807e-07, "loss": 0.0037, "step": 57120 }, { "epoch": 17.58, "learning_rate": 7.651662157967454e-07, "loss": 0.0053, "step": 57121 }, { "epoch": 17.58, "learning_rate": 7.649750717185145e-07, "loss": 0.0047, "step": 57122 }, { "epoch": 17.58, "learning_rate": 7.647839505683608e-07, "loss": 0.0057, "step": 57123 }, { "epoch": 17.58, "learning_rate": 7.645928523467594e-07, "loss": 0.0028, "step": 57124 }, { "epoch": 17.58, "learning_rate": 7.644017770541834e-07, "loss": 0.0043, "step": 57125 }, { "epoch": 17.58, "learning_rate": 7.64210724691109e-07, "loss": 0.0025, "step": 57126 }, { "epoch": 17.58, "learning_rate": 7.640196952580081e-07, "loss": 0.0052, "step": 57127 }, { "epoch": 17.58, "learning_rate": 7.63828688755357e-07, "loss": 0.0035, "step": 57128 }, { "epoch": 17.58, "learning_rate": 7.636377051836297e-07, "loss": 0.0036, "step": 57129 }, { "epoch": 17.58, "learning_rate": 7.634467445432991e-07, "loss": 0.0043, "step": 57130 }, { "epoch": 17.58, "learning_rate": 7.632558068348417e-07, "loss": 0.0023, "step": 57131 }, { "epoch": 17.58, "learning_rate": 7.63064892058728e-07, "loss": 0.0034, "step": 57132 }, { "epoch": 17.58, "learning_rate": 7.628740002154356e-07, "loss": 0.0078, "step": 57133 }, { "epoch": 17.58, "learning_rate": 7.626831313054362e-07, "loss": 0.0026, "step": 57134 }, { "epoch": 17.58, "learning_rate": 7.624922853292039e-07, "loss": 0.002, "step": 57135 }, { "epoch": 17.58, "learning_rate": 7.623014622872116e-07, "loss": 0.0034, "step": 57136 }, { "epoch": 17.58, "learning_rate": 7.621106621799356e-07, "loss": 0.0047, "step": 57137 }, { "epoch": 17.58, "learning_rate": 7.619198850078491e-07, "loss": 0.0023, "step": 57138 }, { "epoch": 17.58, "learning_rate": 7.617291307714259e-07, "loss": 0.0037, "step": 57139 }, { "epoch": 17.58, "learning_rate": 7.615383994711378e-07, "loss": 0.0042, "step": 57140 }, { "epoch": 17.58, "learning_rate": 7.613476911074568e-07, "loss": 0.0021, "step": 57141 }, { "epoch": 17.58, "learning_rate": 7.611570056808615e-07, "loss": 0.0042, "step": 57142 }, { "epoch": 17.58, "learning_rate": 7.609663431918213e-07, "loss": 0.0036, "step": 57143 }, { "epoch": 17.58, "learning_rate": 7.607757036408114e-07, "loss": 0.0063, "step": 57144 }, { "epoch": 17.58, "learning_rate": 7.60585087028305e-07, "loss": 0.004, "step": 57145 }, { "epoch": 17.58, "learning_rate": 7.603944933547747e-07, "loss": 0.0042, "step": 57146 }, { "epoch": 17.58, "learning_rate": 7.602039226206947e-07, "loss": 0.0048, "step": 57147 }, { "epoch": 17.58, "learning_rate": 7.60013374826536e-07, "loss": 0.0049, "step": 57148 }, { "epoch": 17.58, "learning_rate": 7.598228499727745e-07, "loss": 0.0031, "step": 57149 }, { "epoch": 17.58, "learning_rate": 7.59632348059881e-07, "loss": 0.0054, "step": 57150 }, { "epoch": 17.58, "learning_rate": 7.594418690883276e-07, "loss": 0.0068, "step": 57151 }, { "epoch": 17.58, "learning_rate": 7.592514130585927e-07, "loss": 0.004, "step": 57152 }, { "epoch": 17.59, "learning_rate": 7.590609799711424e-07, "loss": 0.0039, "step": 57153 }, { "epoch": 17.59, "learning_rate": 7.588705698264531e-07, "loss": 0.0036, "step": 57154 }, { "epoch": 17.59, "learning_rate": 7.586801826249979e-07, "loss": 0.0036, "step": 57155 }, { "epoch": 17.59, "learning_rate": 7.584898183672451e-07, "loss": 0.0042, "step": 57156 }, { "epoch": 17.59, "learning_rate": 7.582994770536733e-07, "loss": 0.0042, "step": 57157 }, { "epoch": 17.59, "learning_rate": 7.581091586847522e-07, "loss": 0.0032, "step": 57158 }, { "epoch": 17.59, "learning_rate": 7.579188632609547e-07, "loss": 0.005, "step": 57159 }, { "epoch": 17.59, "learning_rate": 7.577285907827547e-07, "loss": 0.0078, "step": 57160 }, { "epoch": 17.59, "learning_rate": 7.575383412506187e-07, "loss": 0.0036, "step": 57161 }, { "epoch": 17.59, "learning_rate": 7.573481146650263e-07, "loss": 0.0029, "step": 57162 }, { "epoch": 17.59, "learning_rate": 7.57157911026446e-07, "loss": 0.0057, "step": 57163 }, { "epoch": 17.59, "learning_rate": 7.569677303353507e-07, "loss": 0.0053, "step": 57164 }, { "epoch": 17.59, "learning_rate": 7.567775725922133e-07, "loss": 0.0039, "step": 57165 }, { "epoch": 17.59, "learning_rate": 7.565874377975046e-07, "loss": 0.0037, "step": 57166 }, { "epoch": 17.59, "learning_rate": 7.563973259516976e-07, "loss": 0.0056, "step": 57167 }, { "epoch": 17.59, "learning_rate": 7.562072370552642e-07, "loss": 0.0041, "step": 57168 }, { "epoch": 17.59, "learning_rate": 7.560171711086761e-07, "loss": 0.0021, "step": 57169 }, { "epoch": 17.59, "learning_rate": 7.558271281124063e-07, "loss": 0.0041, "step": 57170 }, { "epoch": 17.59, "learning_rate": 7.556371080669222e-07, "loss": 0.0047, "step": 57171 }, { "epoch": 17.59, "learning_rate": 7.554471109727013e-07, "loss": 0.0042, "step": 57172 }, { "epoch": 17.59, "learning_rate": 7.552571368302153e-07, "loss": 0.0047, "step": 57173 }, { "epoch": 17.59, "learning_rate": 7.550671856399305e-07, "loss": 0.0031, "step": 57174 }, { "epoch": 17.59, "learning_rate": 7.548772574023233e-07, "loss": 0.0053, "step": 57175 }, { "epoch": 17.59, "learning_rate": 7.546873521178599e-07, "loss": 0.0026, "step": 57176 }, { "epoch": 17.59, "learning_rate": 7.544974697870188e-07, "loss": 0.0023, "step": 57177 }, { "epoch": 17.59, "learning_rate": 7.543076104102664e-07, "loss": 0.017, "step": 57178 }, { "epoch": 17.59, "learning_rate": 7.541177739880767e-07, "loss": 0.0035, "step": 57179 }, { "epoch": 17.59, "learning_rate": 7.539279605209193e-07, "loss": 0.0031, "step": 57180 }, { "epoch": 17.59, "learning_rate": 7.537381700092661e-07, "loss": 0.0049, "step": 57181 }, { "epoch": 17.59, "learning_rate": 7.535484024535889e-07, "loss": 0.004, "step": 57182 }, { "epoch": 17.59, "learning_rate": 7.533586578543583e-07, "loss": 0.0062, "step": 57183 }, { "epoch": 17.59, "learning_rate": 7.531689362120442e-07, "loss": 0.0034, "step": 57184 }, { "epoch": 17.6, "learning_rate": 7.529792375271206e-07, "loss": 0.004, "step": 57185 }, { "epoch": 17.6, "learning_rate": 7.527895618000546e-07, "loss": 0.0036, "step": 57186 }, { "epoch": 17.6, "learning_rate": 7.525999090313207e-07, "loss": 0.0034, "step": 57187 }, { "epoch": 17.6, "learning_rate": 7.524102792213872e-07, "loss": 0.0052, "step": 57188 }, { "epoch": 17.6, "learning_rate": 7.52220672370727e-07, "loss": 0.0074, "step": 57189 }, { "epoch": 17.6, "learning_rate": 7.520310884798088e-07, "loss": 0.0042, "step": 57190 }, { "epoch": 17.6, "learning_rate": 7.518415275491031e-07, "loss": 0.0067, "step": 57191 }, { "epoch": 17.6, "learning_rate": 7.51651989579083e-07, "loss": 0.0062, "step": 57192 }, { "epoch": 17.6, "learning_rate": 7.51462474570217e-07, "loss": 0.0056, "step": 57193 }, { "epoch": 17.6, "learning_rate": 7.512729825229792e-07, "loss": 0.0033, "step": 57194 }, { "epoch": 17.6, "learning_rate": 7.510835134378347e-07, "loss": 0.0032, "step": 57195 }, { "epoch": 17.6, "learning_rate": 7.508940673152553e-07, "loss": 0.0037, "step": 57196 }, { "epoch": 17.6, "learning_rate": 7.507046441557142e-07, "loss": 0.0035, "step": 57197 }, { "epoch": 17.6, "learning_rate": 7.505152439596796e-07, "loss": 0.0032, "step": 57198 }, { "epoch": 17.6, "learning_rate": 7.503258667276225e-07, "loss": 0.0053, "step": 57199 }, { "epoch": 17.6, "learning_rate": 7.501365124600124e-07, "loss": 0.0043, "step": 57200 }, { "epoch": 17.6, "learning_rate": 7.4994718115732e-07, "loss": 0.0048, "step": 57201 }, { "epoch": 17.6, "learning_rate": 7.497578728200138e-07, "loss": 0.0046, "step": 57202 }, { "epoch": 17.6, "learning_rate": 7.495685874485658e-07, "loss": 0.003, "step": 57203 }, { "epoch": 17.6, "learning_rate": 7.493793250434455e-07, "loss": 0.0023, "step": 57204 }, { "epoch": 17.6, "learning_rate": 7.491900856051227e-07, "loss": 0.0075, "step": 57205 }, { "epoch": 17.6, "learning_rate": 7.490008691340645e-07, "loss": 0.007, "step": 57206 }, { "epoch": 17.6, "learning_rate": 7.488116756307462e-07, "loss": 0.0043, "step": 57207 }, { "epoch": 17.6, "learning_rate": 7.48622505095633e-07, "loss": 0.003, "step": 57208 }, { "epoch": 17.6, "learning_rate": 7.484333575291968e-07, "loss": 0.0043, "step": 57209 }, { "epoch": 17.6, "learning_rate": 7.482442329319062e-07, "loss": 0.0034, "step": 57210 }, { "epoch": 17.6, "learning_rate": 7.480551313042284e-07, "loss": 0.0053, "step": 57211 }, { "epoch": 17.6, "learning_rate": 7.478660526466386e-07, "loss": 0.0067, "step": 57212 }, { "epoch": 17.6, "learning_rate": 7.476769969596021e-07, "loss": 0.0046, "step": 57213 }, { "epoch": 17.6, "learning_rate": 7.474879642435895e-07, "loss": 0.0053, "step": 57214 }, { "epoch": 17.6, "learning_rate": 7.472989544990716e-07, "loss": 0.0054, "step": 57215 }, { "epoch": 17.6, "learning_rate": 7.471099677265126e-07, "loss": 0.0031, "step": 57216 }, { "epoch": 17.6, "learning_rate": 7.469210039263863e-07, "loss": 0.0055, "step": 57217 }, { "epoch": 17.61, "learning_rate": 7.467320630991604e-07, "loss": 0.0057, "step": 57218 }, { "epoch": 17.61, "learning_rate": 7.465431452453054e-07, "loss": 0.0031, "step": 57219 }, { "epoch": 17.61, "learning_rate": 7.463542503652887e-07, "loss": 0.0034, "step": 57220 }, { "epoch": 17.61, "learning_rate": 7.46165378459579e-07, "loss": 0.0046, "step": 57221 }, { "epoch": 17.61, "learning_rate": 7.459765295286459e-07, "loss": 0.0048, "step": 57222 }, { "epoch": 17.61, "learning_rate": 7.457877035729588e-07, "loss": 0.0055, "step": 57223 }, { "epoch": 17.61, "learning_rate": 7.455989005929864e-07, "loss": 0.0023, "step": 57224 }, { "epoch": 17.61, "learning_rate": 7.454101205891962e-07, "loss": 0.0042, "step": 57225 }, { "epoch": 17.61, "learning_rate": 7.452213635620564e-07, "loss": 0.0049, "step": 57226 }, { "epoch": 17.61, "learning_rate": 7.45032629512038e-07, "loss": 0.0053, "step": 57227 }, { "epoch": 17.61, "learning_rate": 7.448439184396117e-07, "loss": 0.0036, "step": 57228 }, { "epoch": 17.61, "learning_rate": 7.446552303452393e-07, "loss": 0.0038, "step": 57229 }, { "epoch": 17.61, "learning_rate": 7.444665652293937e-07, "loss": 0.0024, "step": 57230 }, { "epoch": 17.61, "learning_rate": 7.442779230925412e-07, "loss": 0.0022, "step": 57231 }, { "epoch": 17.61, "learning_rate": 7.440893039351527e-07, "loss": 0.0028, "step": 57232 }, { "epoch": 17.61, "learning_rate": 7.439007077576943e-07, "loss": 0.0035, "step": 57233 }, { "epoch": 17.61, "learning_rate": 7.437121345606369e-07, "loss": 0.007, "step": 57234 }, { "epoch": 17.61, "learning_rate": 7.435235843444455e-07, "loss": 0.0034, "step": 57235 }, { "epoch": 17.61, "learning_rate": 7.433350571095898e-07, "loss": 0.0026, "step": 57236 }, { "epoch": 17.61, "learning_rate": 7.431465528565373e-07, "loss": 0.0033, "step": 57237 }, { "epoch": 17.61, "learning_rate": 7.429580715857565e-07, "loss": 0.0029, "step": 57238 }, { "epoch": 17.61, "learning_rate": 7.427696132977158e-07, "loss": 0.0098, "step": 57239 }, { "epoch": 17.61, "learning_rate": 7.425811779928815e-07, "loss": 0.0027, "step": 57240 }, { "epoch": 17.61, "learning_rate": 7.423927656717212e-07, "loss": 0.0048, "step": 57241 }, { "epoch": 17.61, "learning_rate": 7.422043763347076e-07, "loss": 0.0031, "step": 57242 }, { "epoch": 17.61, "learning_rate": 7.420160099823026e-07, "loss": 0.0046, "step": 57243 }, { "epoch": 17.61, "learning_rate": 7.41827666614976e-07, "loss": 0.006, "step": 57244 }, { "epoch": 17.61, "learning_rate": 7.416393462331961e-07, "loss": 0.0014, "step": 57245 }, { "epoch": 17.61, "learning_rate": 7.414510488374272e-07, "loss": 0.0031, "step": 57246 }, { "epoch": 17.61, "learning_rate": 7.412627744281409e-07, "loss": 0.004, "step": 57247 }, { "epoch": 17.61, "learning_rate": 7.410745230058037e-07, "loss": 0.0044, "step": 57248 }, { "epoch": 17.61, "learning_rate": 7.408862945708839e-07, "loss": 0.0036, "step": 57249 }, { "epoch": 17.62, "learning_rate": 7.406980891238447e-07, "loss": 0.004, "step": 57250 }, { "epoch": 17.62, "learning_rate": 7.405099066651556e-07, "loss": 0.0039, "step": 57251 }, { "epoch": 17.62, "learning_rate": 7.403217471952851e-07, "loss": 0.0036, "step": 57252 }, { "epoch": 17.62, "learning_rate": 7.401336107146995e-07, "loss": 0.0023, "step": 57253 }, { "epoch": 17.62, "learning_rate": 7.399454972238651e-07, "loss": 0.003, "step": 57254 }, { "epoch": 17.62, "learning_rate": 7.397574067232505e-07, "loss": 0.009, "step": 57255 }, { "epoch": 17.62, "learning_rate": 7.395693392133207e-07, "loss": 0.0029, "step": 57256 }, { "epoch": 17.62, "learning_rate": 7.393812946945444e-07, "loss": 0.0049, "step": 57257 }, { "epoch": 17.62, "learning_rate": 7.391932731673879e-07, "loss": 0.0027, "step": 57258 }, { "epoch": 17.62, "learning_rate": 7.390052746323185e-07, "loss": 0.0044, "step": 57259 }, { "epoch": 17.62, "learning_rate": 7.388172990898013e-07, "loss": 0.0072, "step": 57260 }, { "epoch": 17.62, "learning_rate": 7.386293465403027e-07, "loss": 0.0045, "step": 57261 }, { "epoch": 17.62, "learning_rate": 7.384414169842924e-07, "loss": 0.0051, "step": 57262 }, { "epoch": 17.62, "learning_rate": 7.382535104222366e-07, "loss": 0.0057, "step": 57263 }, { "epoch": 17.62, "learning_rate": 7.380656268545982e-07, "loss": 0.0028, "step": 57264 }, { "epoch": 17.62, "learning_rate": 7.37877766281847e-07, "loss": 0.0066, "step": 57265 }, { "epoch": 17.62, "learning_rate": 7.376899287044459e-07, "loss": 0.0054, "step": 57266 }, { "epoch": 17.62, "learning_rate": 7.375021141228656e-07, "loss": 0.0038, "step": 57267 }, { "epoch": 17.62, "learning_rate": 7.373143225375711e-07, "loss": 0.005, "step": 57268 }, { "epoch": 17.62, "learning_rate": 7.371265539490269e-07, "loss": 0.003, "step": 57269 }, { "epoch": 17.62, "learning_rate": 7.369388083577023e-07, "loss": 0.0031, "step": 57270 }, { "epoch": 17.62, "learning_rate": 7.367510857640581e-07, "loss": 0.0023, "step": 57271 }, { "epoch": 17.62, "learning_rate": 7.365633861685661e-07, "loss": 0.0056, "step": 57272 }, { "epoch": 17.62, "learning_rate": 7.363757095716884e-07, "loss": 0.0044, "step": 57273 }, { "epoch": 17.62, "learning_rate": 7.361880559738932e-07, "loss": 0.0018, "step": 57274 }, { "epoch": 17.62, "learning_rate": 7.360004253756459e-07, "loss": 0.0027, "step": 57275 }, { "epoch": 17.62, "learning_rate": 7.358128177774115e-07, "loss": 0.0041, "step": 57276 }, { "epoch": 17.62, "learning_rate": 7.356252331796576e-07, "loss": 0.0056, "step": 57277 }, { "epoch": 17.62, "learning_rate": 7.35437671582847e-07, "loss": 0.0039, "step": 57278 }, { "epoch": 17.62, "learning_rate": 7.352501329874473e-07, "loss": 0.0032, "step": 57279 }, { "epoch": 17.62, "learning_rate": 7.350626173939246e-07, "loss": 0.0038, "step": 57280 }, { "epoch": 17.62, "learning_rate": 7.348751248027419e-07, "loss": 0.0047, "step": 57281 }, { "epoch": 17.62, "learning_rate": 7.346876552143678e-07, "loss": 0.0033, "step": 57282 }, { "epoch": 17.63, "learning_rate": 7.345002086292663e-07, "loss": 0.0024, "step": 57283 }, { "epoch": 17.63, "learning_rate": 7.343127850479048e-07, "loss": 0.0062, "step": 57284 }, { "epoch": 17.63, "learning_rate": 7.341253844707452e-07, "loss": 0.0057, "step": 57285 }, { "epoch": 17.63, "learning_rate": 7.339380068982515e-07, "loss": 0.0089, "step": 57286 }, { "epoch": 17.63, "learning_rate": 7.337506523308946e-07, "loss": 0.004, "step": 57287 }, { "epoch": 17.63, "learning_rate": 7.335633207691362e-07, "loss": 0.0038, "step": 57288 }, { "epoch": 17.63, "learning_rate": 7.333760122134415e-07, "loss": 0.005, "step": 57289 }, { "epoch": 17.63, "learning_rate": 7.331887266642768e-07, "loss": 0.0035, "step": 57290 }, { "epoch": 17.63, "learning_rate": 7.330014641221062e-07, "loss": 0.0039, "step": 57291 }, { "epoch": 17.63, "learning_rate": 7.328142245873937e-07, "loss": 0.0047, "step": 57292 }, { "epoch": 17.63, "learning_rate": 7.326270080606058e-07, "loss": 0.0105, "step": 57293 }, { "epoch": 17.63, "learning_rate": 7.324398145422074e-07, "loss": 0.0129, "step": 57294 }, { "epoch": 17.63, "learning_rate": 7.322526440326616e-07, "loss": 0.0032, "step": 57295 }, { "epoch": 17.63, "learning_rate": 7.320654965324326e-07, "loss": 0.0032, "step": 57296 }, { "epoch": 17.63, "learning_rate": 7.318783720419886e-07, "loss": 0.0025, "step": 57297 }, { "epoch": 17.63, "learning_rate": 7.316912705617918e-07, "loss": 0.0053, "step": 57298 }, { "epoch": 17.63, "learning_rate": 7.31504192092306e-07, "loss": 0.003, "step": 57299 }, { "epoch": 17.63, "learning_rate": 7.313171366339977e-07, "loss": 0.0028, "step": 57300 }, { "epoch": 17.63, "learning_rate": 7.311301041873276e-07, "loss": 0.0048, "step": 57301 }, { "epoch": 17.63, "learning_rate": 7.30943094752764e-07, "loss": 0.0052, "step": 57302 }, { "epoch": 17.63, "learning_rate": 7.307561083307702e-07, "loss": 0.0039, "step": 57303 }, { "epoch": 17.63, "learning_rate": 7.3056914492181e-07, "loss": 0.0037, "step": 57304 }, { "epoch": 17.63, "learning_rate": 7.303822045263498e-07, "loss": 0.0046, "step": 57305 }, { "epoch": 17.63, "learning_rate": 7.301952871448481e-07, "loss": 0.0048, "step": 57306 }, { "epoch": 17.63, "learning_rate": 7.300083927777746e-07, "loss": 0.0035, "step": 57307 }, { "epoch": 17.63, "learning_rate": 7.298215214255899e-07, "loss": 0.0061, "step": 57308 }, { "epoch": 17.63, "learning_rate": 7.296346730887594e-07, "loss": 0.0067, "step": 57309 }, { "epoch": 17.63, "learning_rate": 7.29447847767748e-07, "loss": 0.0056, "step": 57310 }, { "epoch": 17.63, "learning_rate": 7.292610454630178e-07, "loss": 0.0088, "step": 57311 }, { "epoch": 17.63, "learning_rate": 7.290742661750328e-07, "loss": 0.0051, "step": 57312 }, { "epoch": 17.63, "learning_rate": 7.288875099042569e-07, "loss": 0.0024, "step": 57313 }, { "epoch": 17.63, "learning_rate": 7.287007766511534e-07, "loss": 0.0048, "step": 57314 }, { "epoch": 17.64, "learning_rate": 7.285140664161872e-07, "loss": 0.0024, "step": 57315 }, { "epoch": 17.64, "learning_rate": 7.283273791998191e-07, "loss": 0.0039, "step": 57316 }, { "epoch": 17.64, "learning_rate": 7.281407150025166e-07, "loss": 0.0039, "step": 57317 }, { "epoch": 17.64, "learning_rate": 7.279540738247425e-07, "loss": 0.0036, "step": 57318 }, { "epoch": 17.64, "learning_rate": 7.277674556669578e-07, "loss": 0.0042, "step": 57319 }, { "epoch": 17.64, "learning_rate": 7.275808605296253e-07, "loss": 0.0053, "step": 57320 }, { "epoch": 17.64, "learning_rate": 7.27394288413209e-07, "loss": 0.0048, "step": 57321 }, { "epoch": 17.64, "learning_rate": 7.272077393181753e-07, "loss": 0.0035, "step": 57322 }, { "epoch": 17.64, "learning_rate": 7.270212132449839e-07, "loss": 0.0044, "step": 57323 }, { "epoch": 17.64, "learning_rate": 7.268347101940986e-07, "loss": 0.0028, "step": 57324 }, { "epoch": 17.64, "learning_rate": 7.266482301659838e-07, "loss": 0.0037, "step": 57325 }, { "epoch": 17.64, "learning_rate": 7.264617731611001e-07, "loss": 0.0063, "step": 57326 }, { "epoch": 17.64, "learning_rate": 7.262753391799127e-07, "loss": 0.0063, "step": 57327 }, { "epoch": 17.64, "learning_rate": 7.260889282228834e-07, "loss": 0.0046, "step": 57328 }, { "epoch": 17.64, "learning_rate": 7.259025402904751e-07, "loss": 0.0048, "step": 57329 }, { "epoch": 17.64, "learning_rate": 7.25716175383151e-07, "loss": 0.0043, "step": 57330 }, { "epoch": 17.64, "learning_rate": 7.255298335013728e-07, "loss": 0.0029, "step": 57331 }, { "epoch": 17.64, "learning_rate": 7.253435146456034e-07, "loss": 0.0098, "step": 57332 }, { "epoch": 17.64, "learning_rate": 7.251572188163059e-07, "loss": 0.0047, "step": 57333 }, { "epoch": 17.64, "learning_rate": 7.249709460139431e-07, "loss": 0.0025, "step": 57334 }, { "epoch": 17.64, "learning_rate": 7.247846962389771e-07, "loss": 0.0045, "step": 57335 }, { "epoch": 17.64, "learning_rate": 7.245984694918673e-07, "loss": 0.0035, "step": 57336 }, { "epoch": 17.64, "learning_rate": 7.244122657730812e-07, "loss": 0.0048, "step": 57337 }, { "epoch": 17.64, "learning_rate": 7.242260850830796e-07, "loss": 0.0062, "step": 57338 }, { "epoch": 17.64, "learning_rate": 7.240399274223253e-07, "loss": 0.0036, "step": 57339 }, { "epoch": 17.64, "learning_rate": 7.238537927912747e-07, "loss": 0.007, "step": 57340 }, { "epoch": 17.64, "learning_rate": 7.236676811903976e-07, "loss": 0.0027, "step": 57341 }, { "epoch": 17.64, "learning_rate": 7.234815926201522e-07, "loss": 0.0024, "step": 57342 }, { "epoch": 17.64, "learning_rate": 7.232955270810005e-07, "loss": 0.0061, "step": 57343 }, { "epoch": 17.64, "learning_rate": 7.231094845734054e-07, "loss": 0.0047, "step": 57344 }, { "epoch": 17.64, "learning_rate": 7.22923465097829e-07, "loss": 0.0037, "step": 57345 }, { "epoch": 17.64, "learning_rate": 7.227374686547317e-07, "loss": 0.0046, "step": 57346 }, { "epoch": 17.64, "learning_rate": 7.225514952445767e-07, "loss": 0.003, "step": 57347 }, { "epoch": 17.65, "learning_rate": 7.223655448678257e-07, "loss": 0.0035, "step": 57348 }, { "epoch": 17.65, "learning_rate": 7.221796175249396e-07, "loss": 0.0043, "step": 57349 }, { "epoch": 17.65, "learning_rate": 7.21993713216379e-07, "loss": 0.0061, "step": 57350 }, { "epoch": 17.65, "learning_rate": 7.218078319426069e-07, "loss": 0.0039, "step": 57351 }, { "epoch": 17.65, "learning_rate": 7.216219737040874e-07, "loss": 0.0026, "step": 57352 }, { "epoch": 17.65, "learning_rate": 7.21436138501278e-07, "loss": 0.0048, "step": 57353 }, { "epoch": 17.65, "learning_rate": 7.212503263346405e-07, "loss": 0.0032, "step": 57354 }, { "epoch": 17.65, "learning_rate": 7.210645372046354e-07, "loss": 0.0047, "step": 57355 }, { "epoch": 17.65, "learning_rate": 7.208787711117271e-07, "loss": 0.0025, "step": 57356 }, { "epoch": 17.65, "learning_rate": 7.206930280563751e-07, "loss": 0.0031, "step": 57357 }, { "epoch": 17.65, "learning_rate": 7.205073080390412e-07, "loss": 0.0032, "step": 57358 }, { "epoch": 17.65, "learning_rate": 7.203216110601851e-07, "loss": 0.0051, "step": 57359 }, { "epoch": 17.65, "learning_rate": 7.201359371202698e-07, "loss": 0.008, "step": 57360 }, { "epoch": 17.65, "learning_rate": 7.199502862197549e-07, "loss": 0.0035, "step": 57361 }, { "epoch": 17.65, "learning_rate": 7.197646583591011e-07, "loss": 0.0059, "step": 57362 }, { "epoch": 17.65, "learning_rate": 7.195790535387703e-07, "loss": 0.0035, "step": 57363 }, { "epoch": 17.65, "learning_rate": 7.193934717592232e-07, "loss": 0.0037, "step": 57364 }, { "epoch": 17.65, "learning_rate": 7.192079130209195e-07, "loss": 0.0109, "step": 57365 }, { "epoch": 17.65, "learning_rate": 7.190223773243199e-07, "loss": 0.0044, "step": 57366 }, { "epoch": 17.65, "learning_rate": 7.188368646698862e-07, "loss": 0.0045, "step": 57367 }, { "epoch": 17.65, "learning_rate": 7.186513750580793e-07, "loss": 0.0054, "step": 57368 }, { "epoch": 17.65, "learning_rate": 7.184659084893575e-07, "loss": 0.0022, "step": 57369 }, { "epoch": 17.65, "learning_rate": 7.182804649641817e-07, "loss": 0.0043, "step": 57370 }, { "epoch": 17.65, "learning_rate": 7.180950444830148e-07, "loss": 0.0021, "step": 57371 }, { "epoch": 17.65, "learning_rate": 7.179096470463154e-07, "loss": 0.0032, "step": 57372 }, { "epoch": 17.65, "learning_rate": 7.177242726545464e-07, "loss": 0.0052, "step": 57373 }, { "epoch": 17.65, "learning_rate": 7.175389213081629e-07, "loss": 0.0081, "step": 57374 }, { "epoch": 17.65, "learning_rate": 7.17353593007627e-07, "loss": 0.0043, "step": 57375 }, { "epoch": 17.65, "learning_rate": 7.171682877534003e-07, "loss": 0.0036, "step": 57376 }, { "epoch": 17.65, "learning_rate": 7.169830055459437e-07, "loss": 0.003, "step": 57377 }, { "epoch": 17.65, "learning_rate": 7.167977463857145e-07, "loss": 0.0048, "step": 57378 }, { "epoch": 17.65, "learning_rate": 7.166125102731735e-07, "loss": 0.0052, "step": 57379 }, { "epoch": 17.66, "learning_rate": 7.164272972087816e-07, "loss": 0.0046, "step": 57380 }, { "epoch": 17.66, "learning_rate": 7.162421071929981e-07, "loss": 0.0043, "step": 57381 }, { "epoch": 17.66, "learning_rate": 7.160569402262818e-07, "loss": 0.0037, "step": 57382 }, { "epoch": 17.66, "learning_rate": 7.158717963090945e-07, "loss": 0.0023, "step": 57383 }, { "epoch": 17.66, "learning_rate": 7.156866754418934e-07, "loss": 0.0047, "step": 57384 }, { "epoch": 17.66, "learning_rate": 7.155015776251383e-07, "loss": 0.0061, "step": 57385 }, { "epoch": 17.66, "learning_rate": 7.153165028592924e-07, "loss": 0.0033, "step": 57386 }, { "epoch": 17.66, "learning_rate": 7.151314511448115e-07, "loss": 0.004, "step": 57387 }, { "epoch": 17.66, "learning_rate": 7.149464224821556e-07, "loss": 0.0038, "step": 57388 }, { "epoch": 17.66, "learning_rate": 7.147614168717854e-07, "loss": 0.0028, "step": 57389 }, { "epoch": 17.66, "learning_rate": 7.14576434314157e-07, "loss": 0.0099, "step": 57390 }, { "epoch": 17.66, "learning_rate": 7.143914748097336e-07, "loss": 0.0079, "step": 57391 }, { "epoch": 17.66, "learning_rate": 7.142065383589736e-07, "loss": 0.0046, "step": 57392 }, { "epoch": 17.66, "learning_rate": 7.140216249623355e-07, "loss": 0.0036, "step": 57393 }, { "epoch": 17.66, "learning_rate": 7.138367346202791e-07, "loss": 0.0042, "step": 57394 }, { "epoch": 17.66, "learning_rate": 7.136518673332593e-07, "loss": 0.0036, "step": 57395 }, { "epoch": 17.66, "learning_rate": 7.134670231017405e-07, "loss": 0.0086, "step": 57396 }, { "epoch": 17.66, "learning_rate": 7.132822019261798e-07, "loss": 0.0038, "step": 57397 }, { "epoch": 17.66, "learning_rate": 7.130974038070349e-07, "loss": 0.005, "step": 57398 }, { "epoch": 17.66, "learning_rate": 7.129126287447663e-07, "loss": 0.0062, "step": 57399 }, { "epoch": 17.66, "learning_rate": 7.127278767398316e-07, "loss": 0.0092, "step": 57400 }, { "epoch": 17.66, "learning_rate": 7.125431477926892e-07, "loss": 0.0025, "step": 57401 }, { "epoch": 17.66, "learning_rate": 7.123584419037987e-07, "loss": 0.0058, "step": 57402 }, { "epoch": 17.66, "learning_rate": 7.121737590736188e-07, "loss": 0.0055, "step": 57403 }, { "epoch": 17.66, "learning_rate": 7.119890993026068e-07, "loss": 0.0044, "step": 57404 }, { "epoch": 17.66, "learning_rate": 7.118044625912213e-07, "loss": 0.0022, "step": 57405 }, { "epoch": 17.66, "learning_rate": 7.116198489399218e-07, "loss": 0.0066, "step": 57406 }, { "epoch": 17.66, "learning_rate": 7.114352583491691e-07, "loss": 0.004, "step": 57407 }, { "epoch": 17.66, "learning_rate": 7.112506908194161e-07, "loss": 0.0045, "step": 57408 }, { "epoch": 17.66, "learning_rate": 7.110661463511225e-07, "loss": 0.0068, "step": 57409 }, { "epoch": 17.66, "learning_rate": 7.10881624944747e-07, "loss": 0.004, "step": 57410 }, { "epoch": 17.66, "learning_rate": 7.1069712660075e-07, "loss": 0.0033, "step": 57411 }, { "epoch": 17.66, "learning_rate": 7.10512651319587e-07, "loss": 0.0071, "step": 57412 }, { "epoch": 17.67, "learning_rate": 7.103281991017174e-07, "loss": 0.0051, "step": 57413 }, { "epoch": 17.67, "learning_rate": 7.101437699475977e-07, "loss": 0.0032, "step": 57414 }, { "epoch": 17.67, "learning_rate": 7.099593638576874e-07, "loss": 0.0041, "step": 57415 }, { "epoch": 17.67, "learning_rate": 7.097749808324427e-07, "loss": 0.0036, "step": 57416 }, { "epoch": 17.67, "learning_rate": 7.095906208723224e-07, "loss": 0.0062, "step": 57417 }, { "epoch": 17.67, "learning_rate": 7.094062839777838e-07, "loss": 0.0037, "step": 57418 }, { "epoch": 17.67, "learning_rate": 7.092219701492853e-07, "loss": 0.0055, "step": 57419 }, { "epoch": 17.67, "learning_rate": 7.090376793872822e-07, "loss": 0.0038, "step": 57420 }, { "epoch": 17.67, "learning_rate": 7.088534116922363e-07, "loss": 0.0053, "step": 57421 }, { "epoch": 17.67, "learning_rate": 7.086691670646018e-07, "loss": 0.0042, "step": 57422 }, { "epoch": 17.67, "learning_rate": 7.08484945504836e-07, "loss": 0.0047, "step": 57423 }, { "epoch": 17.67, "learning_rate": 7.083007470133985e-07, "loss": 0.004, "step": 57424 }, { "epoch": 17.67, "learning_rate": 7.081165715907423e-07, "loss": 0.0025, "step": 57425 }, { "epoch": 17.67, "learning_rate": 7.079324192373293e-07, "loss": 0.0034, "step": 57426 }, { "epoch": 17.67, "learning_rate": 7.077482899536159e-07, "loss": 0.0034, "step": 57427 }, { "epoch": 17.67, "learning_rate": 7.075641837400604e-07, "loss": 0.0046, "step": 57428 }, { "epoch": 17.67, "learning_rate": 7.073801005971148e-07, "loss": 0.0024, "step": 57429 }, { "epoch": 17.67, "learning_rate": 7.071960405252387e-07, "loss": 0.0029, "step": 57430 }, { "epoch": 17.67, "learning_rate": 7.070120035248906e-07, "loss": 0.006, "step": 57431 }, { "epoch": 17.67, "learning_rate": 7.068279895965269e-07, "loss": 0.0035, "step": 57432 }, { "epoch": 17.67, "learning_rate": 7.066439987406026e-07, "loss": 0.0035, "step": 57433 }, { "epoch": 17.67, "learning_rate": 7.064600309575775e-07, "loss": 0.005, "step": 57434 }, { "epoch": 17.67, "learning_rate": 7.062760862479057e-07, "loss": 0.0035, "step": 57435 }, { "epoch": 17.67, "learning_rate": 7.060921646120455e-07, "loss": 0.0047, "step": 57436 }, { "epoch": 17.67, "learning_rate": 7.059082660504524e-07, "loss": 0.0029, "step": 57437 }, { "epoch": 17.67, "learning_rate": 7.057243905635836e-07, "loss": 0.009, "step": 57438 }, { "epoch": 17.67, "learning_rate": 7.055405381518954e-07, "loss": 0.0024, "step": 57439 }, { "epoch": 17.67, "learning_rate": 7.053567088158431e-07, "loss": 0.0033, "step": 57440 }, { "epoch": 17.67, "learning_rate": 7.051729025558851e-07, "loss": 0.0047, "step": 57441 }, { "epoch": 17.67, "learning_rate": 7.0498911937248e-07, "loss": 0.0056, "step": 57442 }, { "epoch": 17.67, "learning_rate": 7.048053592660786e-07, "loss": 0.0044, "step": 57443 }, { "epoch": 17.67, "learning_rate": 7.046216222371393e-07, "loss": 0.0061, "step": 57444 }, { "epoch": 17.68, "learning_rate": 7.044379082861174e-07, "loss": 0.0031, "step": 57445 }, { "epoch": 17.68, "learning_rate": 7.042542174134714e-07, "loss": 0.0038, "step": 57446 }, { "epoch": 17.68, "learning_rate": 7.040705496196564e-07, "loss": 0.0055, "step": 57447 }, { "epoch": 17.68, "learning_rate": 7.038869049051277e-07, "loss": 0.0046, "step": 57448 }, { "epoch": 17.68, "learning_rate": 7.037032832703439e-07, "loss": 0.0038, "step": 57449 }, { "epoch": 17.68, "learning_rate": 7.035196847157555e-07, "loss": 0.0043, "step": 57450 }, { "epoch": 17.68, "learning_rate": 7.033361092418223e-07, "loss": 0.003, "step": 57451 }, { "epoch": 17.68, "learning_rate": 7.031525568489994e-07, "loss": 0.005, "step": 57452 }, { "epoch": 17.68, "learning_rate": 7.029690275377432e-07, "loss": 0.0083, "step": 57453 }, { "epoch": 17.68, "learning_rate": 7.027855213085078e-07, "loss": 0.0035, "step": 57454 }, { "epoch": 17.68, "learning_rate": 7.026020381617493e-07, "loss": 0.0078, "step": 57455 }, { "epoch": 17.68, "learning_rate": 7.024185780979242e-07, "loss": 0.002, "step": 57456 }, { "epoch": 17.68, "learning_rate": 7.022351411174866e-07, "loss": 0.0028, "step": 57457 }, { "epoch": 17.68, "learning_rate": 7.020517272208927e-07, "loss": 0.0052, "step": 57458 }, { "epoch": 17.68, "learning_rate": 7.018683364085976e-07, "loss": 0.0026, "step": 57459 }, { "epoch": 17.68, "learning_rate": 7.016849686810557e-07, "loss": 0.0034, "step": 57460 }, { "epoch": 17.68, "learning_rate": 7.015016240387241e-07, "loss": 0.0014, "step": 57461 }, { "epoch": 17.68, "learning_rate": 7.013183024820581e-07, "loss": 0.0045, "step": 57462 }, { "epoch": 17.68, "learning_rate": 7.011350040115128e-07, "loss": 0.0053, "step": 57463 }, { "epoch": 17.68, "learning_rate": 7.009517286275402e-07, "loss": 0.0038, "step": 57464 }, { "epoch": 17.68, "learning_rate": 7.007684763305967e-07, "loss": 0.0098, "step": 57465 }, { "epoch": 17.68, "learning_rate": 7.005852471211405e-07, "loss": 0.0036, "step": 57466 }, { "epoch": 17.68, "learning_rate": 7.004020409996226e-07, "loss": 0.004, "step": 57467 }, { "epoch": 17.68, "learning_rate": 7.002188579665004e-07, "loss": 0.0029, "step": 57468 }, { "epoch": 17.68, "learning_rate": 7.000356980222268e-07, "loss": 0.0042, "step": 57469 }, { "epoch": 17.68, "learning_rate": 6.99852561167258e-07, "loss": 0.0061, "step": 57470 }, { "epoch": 17.68, "learning_rate": 6.996694474020471e-07, "loss": 0.0039, "step": 57471 }, { "epoch": 17.68, "learning_rate": 6.994863567270504e-07, "loss": 0.0035, "step": 57472 }, { "epoch": 17.68, "learning_rate": 6.993032891427221e-07, "loss": 0.0035, "step": 57473 }, { "epoch": 17.68, "learning_rate": 6.99120244649516e-07, "loss": 0.0034, "step": 57474 }, { "epoch": 17.68, "learning_rate": 6.989372232478853e-07, "loss": 0.0046, "step": 57475 }, { "epoch": 17.68, "learning_rate": 6.987542249382884e-07, "loss": 0.0032, "step": 57476 }, { "epoch": 17.68, "learning_rate": 6.985712497211772e-07, "loss": 0.0051, "step": 57477 }, { "epoch": 17.69, "learning_rate": 6.983882975970047e-07, "loss": 0.0041, "step": 57478 }, { "epoch": 17.69, "learning_rate": 6.982053685662272e-07, "loss": 0.0031, "step": 57479 }, { "epoch": 17.69, "learning_rate": 6.980224626292953e-07, "loss": 0.0036, "step": 57480 }, { "epoch": 17.69, "learning_rate": 6.978395797866688e-07, "loss": 0.0071, "step": 57481 }, { "epoch": 17.69, "learning_rate": 6.976567200387985e-07, "loss": 0.0048, "step": 57482 }, { "epoch": 17.69, "learning_rate": 6.974738833861383e-07, "loss": 0.0056, "step": 57483 }, { "epoch": 17.69, "learning_rate": 6.972910698291446e-07, "loss": 0.0047, "step": 57484 }, { "epoch": 17.69, "learning_rate": 6.971082793682659e-07, "loss": 0.0047, "step": 57485 }, { "epoch": 17.69, "learning_rate": 6.969255120039609e-07, "loss": 0.0037, "step": 57486 }, { "epoch": 17.69, "learning_rate": 6.967427677366823e-07, "loss": 0.0037, "step": 57487 }, { "epoch": 17.69, "learning_rate": 6.96560046566882e-07, "loss": 0.0029, "step": 57488 }, { "epoch": 17.69, "learning_rate": 6.963773484950164e-07, "loss": 0.0042, "step": 57489 }, { "epoch": 17.69, "learning_rate": 6.961946735215375e-07, "loss": 0.0049, "step": 57490 }, { "epoch": 17.69, "learning_rate": 6.96012021646898e-07, "loss": 0.0052, "step": 57491 }, { "epoch": 17.69, "learning_rate": 6.958293928715531e-07, "loss": 0.0038, "step": 57492 }, { "epoch": 17.69, "learning_rate": 6.956467871959549e-07, "loss": 0.0043, "step": 57493 }, { "epoch": 17.69, "learning_rate": 6.954642046205584e-07, "loss": 0.0028, "step": 57494 }, { "epoch": 17.69, "learning_rate": 6.952816451458133e-07, "loss": 0.0039, "step": 57495 }, { "epoch": 17.69, "learning_rate": 6.950991087721781e-07, "loss": 0.0039, "step": 57496 }, { "epoch": 17.69, "learning_rate": 6.949165955001036e-07, "loss": 0.0045, "step": 57497 }, { "epoch": 17.69, "learning_rate": 6.947341053300416e-07, "loss": 0.0046, "step": 57498 }, { "epoch": 17.69, "learning_rate": 6.945516382624462e-07, "loss": 0.0042, "step": 57499 }, { "epoch": 17.69, "learning_rate": 6.943691942977693e-07, "loss": 0.0038, "step": 57500 }, { "epoch": 17.69, "learning_rate": 6.94186773436466e-07, "loss": 0.0022, "step": 57501 }, { "epoch": 17.69, "learning_rate": 6.940043756789882e-07, "loss": 0.0027, "step": 57502 }, { "epoch": 17.69, "learning_rate": 6.93822001025789e-07, "loss": 0.006, "step": 57503 }, { "epoch": 17.69, "learning_rate": 6.936396494773201e-07, "loss": 0.0037, "step": 57504 }, { "epoch": 17.69, "learning_rate": 6.934573210340357e-07, "loss": 0.0046, "step": 57505 }, { "epoch": 17.69, "learning_rate": 6.932750156963875e-07, "loss": 0.0052, "step": 57506 }, { "epoch": 17.69, "learning_rate": 6.930927334648285e-07, "loss": 0.0031, "step": 57507 }, { "epoch": 17.69, "learning_rate": 6.929104743398119e-07, "loss": 0.0044, "step": 57508 }, { "epoch": 17.69, "learning_rate": 6.927282383217893e-07, "loss": 0.0054, "step": 57509 }, { "epoch": 17.7, "learning_rate": 6.925460254112126e-07, "loss": 0.0037, "step": 57510 }, { "epoch": 17.7, "learning_rate": 6.92363835608535e-07, "loss": 0.0033, "step": 57511 }, { "epoch": 17.7, "learning_rate": 6.921816689142091e-07, "loss": 0.0042, "step": 57512 }, { "epoch": 17.7, "learning_rate": 6.919995253286859e-07, "loss": 0.0042, "step": 57513 }, { "epoch": 17.7, "learning_rate": 6.918174048524196e-07, "loss": 0.0041, "step": 57514 }, { "epoch": 17.7, "learning_rate": 6.916353074858595e-07, "loss": 0.0054, "step": 57515 }, { "epoch": 17.7, "learning_rate": 6.914532332294611e-07, "loss": 0.0022, "step": 57516 }, { "epoch": 17.7, "learning_rate": 6.912711820836748e-07, "loss": 0.0071, "step": 57517 }, { "epoch": 17.7, "learning_rate": 6.910891540489539e-07, "loss": 0.0048, "step": 57518 }, { "epoch": 17.7, "learning_rate": 6.90907149125748e-07, "loss": 0.0035, "step": 57519 }, { "epoch": 17.7, "learning_rate": 6.907251673145077e-07, "loss": 0.0035, "step": 57520 }, { "epoch": 17.7, "learning_rate": 6.905432086156883e-07, "loss": 0.0052, "step": 57521 }, { "epoch": 17.7, "learning_rate": 6.903612730297416e-07, "loss": 0.003, "step": 57522 }, { "epoch": 17.7, "learning_rate": 6.901793605571172e-07, "loss": 0.0036, "step": 57523 }, { "epoch": 17.7, "learning_rate": 6.899974711982682e-07, "loss": 0.0053, "step": 57524 }, { "epoch": 17.7, "learning_rate": 6.898156049536453e-07, "loss": 0.0057, "step": 57525 }, { "epoch": 17.7, "learning_rate": 6.896337618237004e-07, "loss": 0.007, "step": 57526 }, { "epoch": 17.7, "learning_rate": 6.89451941808884e-07, "loss": 0.0039, "step": 57527 }, { "epoch": 17.7, "learning_rate": 6.892701449096495e-07, "loss": 0.0026, "step": 57528 }, { "epoch": 17.7, "learning_rate": 6.890883711264462e-07, "loss": 0.0024, "step": 57529 }, { "epoch": 17.7, "learning_rate": 6.889066204597261e-07, "loss": 0.0039, "step": 57530 }, { "epoch": 17.7, "learning_rate": 6.887248929099433e-07, "loss": 0.0046, "step": 57531 }, { "epoch": 17.7, "learning_rate": 6.88543188477544e-07, "loss": 0.0066, "step": 57532 }, { "epoch": 17.7, "learning_rate": 6.883615071629823e-07, "loss": 0.0027, "step": 57533 }, { "epoch": 17.7, "learning_rate": 6.88179848966708e-07, "loss": 0.0039, "step": 57534 }, { "epoch": 17.7, "learning_rate": 6.879982138891717e-07, "loss": 0.0038, "step": 57535 }, { "epoch": 17.7, "learning_rate": 6.878166019308275e-07, "loss": 0.0041, "step": 57536 }, { "epoch": 17.7, "learning_rate": 6.876350130921239e-07, "loss": 0.0089, "step": 57537 }, { "epoch": 17.7, "learning_rate": 6.874534473735118e-07, "loss": 0.0049, "step": 57538 }, { "epoch": 17.7, "learning_rate": 6.872719047754428e-07, "loss": 0.004, "step": 57539 }, { "epoch": 17.7, "learning_rate": 6.870903852983657e-07, "loss": 0.005, "step": 57540 }, { "epoch": 17.7, "learning_rate": 6.869088889427334e-07, "loss": 0.0039, "step": 57541 }, { "epoch": 17.7, "learning_rate": 6.867274157089953e-07, "loss": 0.0039, "step": 57542 }, { "epoch": 17.71, "learning_rate": 6.865459655976014e-07, "loss": 0.0042, "step": 57543 }, { "epoch": 17.71, "learning_rate": 6.863645386090045e-07, "loss": 0.0052, "step": 57544 }, { "epoch": 17.71, "learning_rate": 6.861831347436532e-07, "loss": 0.0063, "step": 57545 }, { "epoch": 17.71, "learning_rate": 6.860017540019981e-07, "loss": 0.002, "step": 57546 }, { "epoch": 17.71, "learning_rate": 6.858203963844889e-07, "loss": 0.0024, "step": 57547 }, { "epoch": 17.71, "learning_rate": 6.856390618915775e-07, "loss": 0.0044, "step": 57548 }, { "epoch": 17.71, "learning_rate": 6.854577505237137e-07, "loss": 0.0027, "step": 57549 }, { "epoch": 17.71, "learning_rate": 6.852764622813446e-07, "loss": 0.0032, "step": 57550 }, { "epoch": 17.71, "learning_rate": 6.850951971649255e-07, "loss": 0.0032, "step": 57551 }, { "epoch": 17.71, "learning_rate": 6.84913955174904e-07, "loss": 0.0049, "step": 57552 }, { "epoch": 17.71, "learning_rate": 6.847327363117296e-07, "loss": 0.0057, "step": 57553 }, { "epoch": 17.71, "learning_rate": 6.845515405758518e-07, "loss": 0.0033, "step": 57554 }, { "epoch": 17.71, "learning_rate": 6.843703679677194e-07, "loss": 0.0053, "step": 57555 }, { "epoch": 17.71, "learning_rate": 6.841892184877863e-07, "loss": 0.0033, "step": 57556 }, { "epoch": 17.71, "learning_rate": 6.840080921365e-07, "loss": 0.0046, "step": 57557 }, { "epoch": 17.71, "learning_rate": 6.838269889143089e-07, "loss": 0.0082, "step": 57558 }, { "epoch": 17.71, "learning_rate": 6.836459088216651e-07, "loss": 0.003, "step": 57559 }, { "epoch": 17.71, "learning_rate": 6.834648518590159e-07, "loss": 0.0033, "step": 57560 }, { "epoch": 17.71, "learning_rate": 6.83283818026812e-07, "loss": 0.0027, "step": 57561 }, { "epoch": 17.71, "learning_rate": 6.83102807325503e-07, "loss": 0.0052, "step": 57562 }, { "epoch": 17.71, "learning_rate": 6.829218197555376e-07, "loss": 0.0038, "step": 57563 }, { "epoch": 17.71, "learning_rate": 6.827408553173664e-07, "loss": 0.003, "step": 57564 }, { "epoch": 17.71, "learning_rate": 6.82559914011437e-07, "loss": 0.0049, "step": 57565 }, { "epoch": 17.71, "learning_rate": 6.823789958381999e-07, "loss": 0.0048, "step": 57566 }, { "epoch": 17.71, "learning_rate": 6.821981007981027e-07, "loss": 0.0032, "step": 57567 }, { "epoch": 17.71, "learning_rate": 6.820172288915972e-07, "loss": 0.0037, "step": 57568 }, { "epoch": 17.71, "learning_rate": 6.818363801191297e-07, "loss": 0.0037, "step": 57569 }, { "epoch": 17.71, "learning_rate": 6.816555544811487e-07, "loss": 0.0023, "step": 57570 }, { "epoch": 17.71, "learning_rate": 6.814747519781073e-07, "loss": 0.0047, "step": 57571 }, { "epoch": 17.71, "learning_rate": 6.812939726104518e-07, "loss": 0.0042, "step": 57572 }, { "epoch": 17.71, "learning_rate": 6.811132163786316e-07, "loss": 0.005, "step": 57573 }, { "epoch": 17.71, "learning_rate": 6.809324832830944e-07, "loss": 0.0037, "step": 57574 }, { "epoch": 17.72, "learning_rate": 6.807517733242874e-07, "loss": 0.0049, "step": 57575 }, { "epoch": 17.72, "learning_rate": 6.805710865026638e-07, "loss": 0.002, "step": 57576 }, { "epoch": 17.72, "learning_rate": 6.803904228186697e-07, "loss": 0.0049, "step": 57577 }, { "epoch": 17.72, "learning_rate": 6.802097822727527e-07, "loss": 0.005, "step": 57578 }, { "epoch": 17.72, "learning_rate": 6.800291648653623e-07, "loss": 0.005, "step": 57579 }, { "epoch": 17.72, "learning_rate": 6.798485705969482e-07, "loss": 0.0049, "step": 57580 }, { "epoch": 17.72, "learning_rate": 6.796679994679567e-07, "loss": 0.003, "step": 57581 }, { "epoch": 17.72, "learning_rate": 6.794874514788363e-07, "loss": 0.0034, "step": 57582 }, { "epoch": 17.72, "learning_rate": 6.793069266300367e-07, "loss": 0.0036, "step": 57583 }, { "epoch": 17.72, "learning_rate": 6.791264249220042e-07, "loss": 0.0032, "step": 57584 }, { "epoch": 17.72, "learning_rate": 6.789459463551873e-07, "loss": 0.0094, "step": 57585 }, { "epoch": 17.72, "learning_rate": 6.787654909300379e-07, "loss": 0.0044, "step": 57586 }, { "epoch": 17.72, "learning_rate": 6.785850586469989e-07, "loss": 0.0053, "step": 57587 }, { "epoch": 17.72, "learning_rate": 6.7840464950652e-07, "loss": 0.0022, "step": 57588 }, { "epoch": 17.72, "learning_rate": 6.782242635090497e-07, "loss": 0.004, "step": 57589 }, { "epoch": 17.72, "learning_rate": 6.780439006550333e-07, "loss": 0.0052, "step": 57590 }, { "epoch": 17.72, "learning_rate": 6.778635609449235e-07, "loss": 0.0038, "step": 57591 }, { "epoch": 17.72, "learning_rate": 6.776832443791647e-07, "loss": 0.0047, "step": 57592 }, { "epoch": 17.72, "learning_rate": 6.775029509582054e-07, "loss": 0.0038, "step": 57593 }, { "epoch": 17.72, "learning_rate": 6.77322680682494e-07, "loss": 0.0054, "step": 57594 }, { "epoch": 17.72, "learning_rate": 6.771424335524745e-07, "loss": 0.0042, "step": 57595 }, { "epoch": 17.72, "learning_rate": 6.769622095685979e-07, "loss": 0.0024, "step": 57596 }, { "epoch": 17.72, "learning_rate": 6.767820087313115e-07, "loss": 0.0056, "step": 57597 }, { "epoch": 17.72, "learning_rate": 6.766018310410616e-07, "loss": 0.0049, "step": 57598 }, { "epoch": 17.72, "learning_rate": 6.764216764982956e-07, "loss": 0.0026, "step": 57599 }, { "epoch": 17.72, "learning_rate": 6.76241545103461e-07, "loss": 0.0034, "step": 57600 }, { "epoch": 17.72, "learning_rate": 6.760614368570051e-07, "loss": 0.0036, "step": 57601 }, { "epoch": 17.72, "learning_rate": 6.758813517593754e-07, "loss": 0.0057, "step": 57602 }, { "epoch": 17.72, "learning_rate": 6.757012898110182e-07, "loss": 0.0056, "step": 57603 }, { "epoch": 17.72, "learning_rate": 6.75521251012381e-07, "loss": 0.0061, "step": 57604 }, { "epoch": 17.72, "learning_rate": 6.753412353639099e-07, "loss": 0.0043, "step": 57605 }, { "epoch": 17.72, "learning_rate": 6.751612428660536e-07, "loss": 0.0033, "step": 57606 }, { "epoch": 17.72, "learning_rate": 6.749812735192596e-07, "loss": 0.0052, "step": 57607 }, { "epoch": 17.73, "learning_rate": 6.748013273239707e-07, "loss": 0.004, "step": 57608 }, { "epoch": 17.73, "learning_rate": 6.746214042806376e-07, "loss": 0.0025, "step": 57609 }, { "epoch": 17.73, "learning_rate": 6.744415043897035e-07, "loss": 0.0031, "step": 57610 }, { "epoch": 17.73, "learning_rate": 6.742616276516179e-07, "loss": 0.0027, "step": 57611 }, { "epoch": 17.73, "learning_rate": 6.740817740668271e-07, "loss": 0.0045, "step": 57612 }, { "epoch": 17.73, "learning_rate": 6.739019436357774e-07, "loss": 0.0019, "step": 57613 }, { "epoch": 17.73, "learning_rate": 6.737221363589142e-07, "loss": 0.0036, "step": 57614 }, { "epoch": 17.73, "learning_rate": 6.735423522366846e-07, "loss": 0.0042, "step": 57615 }, { "epoch": 17.73, "learning_rate": 6.733625912695363e-07, "loss": 0.0036, "step": 57616 }, { "epoch": 17.73, "learning_rate": 6.731828534579132e-07, "loss": 0.0037, "step": 57617 }, { "epoch": 17.73, "learning_rate": 6.730031388022628e-07, "loss": 0.0045, "step": 57618 }, { "epoch": 17.73, "learning_rate": 6.728234473030315e-07, "loss": 0.0048, "step": 57619 }, { "epoch": 17.73, "learning_rate": 6.726437789606633e-07, "loss": 0.0045, "step": 57620 }, { "epoch": 17.73, "learning_rate": 6.724641337756088e-07, "loss": 0.0036, "step": 57621 }, { "epoch": 17.73, "learning_rate": 6.722845117483101e-07, "loss": 0.0052, "step": 57622 }, { "epoch": 17.73, "learning_rate": 6.721049128792134e-07, "loss": 0.0024, "step": 57623 }, { "epoch": 17.73, "learning_rate": 6.719253371687662e-07, "loss": 0.0032, "step": 57624 }, { "epoch": 17.73, "learning_rate": 6.717457846174125e-07, "loss": 0.0049, "step": 57625 }, { "epoch": 17.73, "learning_rate": 6.715662552255998e-07, "loss": 0.0055, "step": 57626 }, { "epoch": 17.73, "learning_rate": 6.713867489937742e-07, "loss": 0.008, "step": 57627 }, { "epoch": 17.73, "learning_rate": 6.712072659223812e-07, "loss": 0.0046, "step": 57628 }, { "epoch": 17.73, "learning_rate": 6.710278060118647e-07, "loss": 0.0034, "step": 57629 }, { "epoch": 17.73, "learning_rate": 6.708483692626688e-07, "loss": 0.004, "step": 57630 }, { "epoch": 17.73, "learning_rate": 6.706689556752433e-07, "loss": 0.007, "step": 57631 }, { "epoch": 17.73, "learning_rate": 6.704895652500309e-07, "loss": 0.0036, "step": 57632 }, { "epoch": 17.73, "learning_rate": 6.703101979874793e-07, "loss": 0.0033, "step": 57633 }, { "epoch": 17.73, "learning_rate": 6.701308538880313e-07, "loss": 0.0045, "step": 57634 }, { "epoch": 17.73, "learning_rate": 6.699515329521333e-07, "loss": 0.003, "step": 57635 }, { "epoch": 17.73, "learning_rate": 6.697722351802305e-07, "loss": 0.0037, "step": 57636 }, { "epoch": 17.73, "learning_rate": 6.69592960572767e-07, "loss": 0.0049, "step": 57637 }, { "epoch": 17.73, "learning_rate": 6.6941370913019e-07, "loss": 0.0041, "step": 57638 }, { "epoch": 17.73, "learning_rate": 6.692344808529427e-07, "loss": 0.0036, "step": 57639 }, { "epoch": 17.74, "learning_rate": 6.690552757414692e-07, "loss": 0.0034, "step": 57640 }, { "epoch": 17.74, "learning_rate": 6.68876093796218e-07, "loss": 0.0032, "step": 57641 }, { "epoch": 17.74, "learning_rate": 6.68696935017632e-07, "loss": 0.0059, "step": 57642 }, { "epoch": 17.74, "learning_rate": 6.685177994061554e-07, "loss": 0.0039, "step": 57643 }, { "epoch": 17.74, "learning_rate": 6.683386869622332e-07, "loss": 0.0031, "step": 57644 }, { "epoch": 17.74, "learning_rate": 6.681595976863086e-07, "loss": 0.0043, "step": 57645 }, { "epoch": 17.74, "learning_rate": 6.679805315788301e-07, "loss": 0.0045, "step": 57646 }, { "epoch": 17.74, "learning_rate": 6.678014886402396e-07, "loss": 0.0045, "step": 57647 }, { "epoch": 17.74, "learning_rate": 6.67622468870982e-07, "loss": 0.0054, "step": 57648 }, { "epoch": 17.74, "learning_rate": 6.67443472271504e-07, "loss": 0.0055, "step": 57649 }, { "epoch": 17.74, "learning_rate": 6.67264498842245e-07, "loss": 0.0053, "step": 57650 }, { "epoch": 17.74, "learning_rate": 6.670855485836525e-07, "loss": 0.0027, "step": 57651 }, { "epoch": 17.74, "learning_rate": 6.669066214961717e-07, "loss": 0.0043, "step": 57652 }, { "epoch": 17.74, "learning_rate": 6.667277175802444e-07, "loss": 0.0034, "step": 57653 }, { "epoch": 17.74, "learning_rate": 6.665488368363171e-07, "loss": 0.0067, "step": 57654 }, { "epoch": 17.74, "learning_rate": 6.663699792648326e-07, "loss": 0.0037, "step": 57655 }, { "epoch": 17.74, "learning_rate": 6.66191144866235e-07, "loss": 0.009, "step": 57656 }, { "epoch": 17.74, "learning_rate": 6.660123336409685e-07, "loss": 0.0026, "step": 57657 }, { "epoch": 17.74, "learning_rate": 6.658335455894771e-07, "loss": 0.0024, "step": 57658 }, { "epoch": 17.74, "learning_rate": 6.656547807122038e-07, "loss": 0.0068, "step": 57659 }, { "epoch": 17.74, "learning_rate": 6.654760390095927e-07, "loss": 0.0015, "step": 57660 }, { "epoch": 17.74, "learning_rate": 6.652973204820889e-07, "loss": 0.004, "step": 57661 }, { "epoch": 17.74, "learning_rate": 6.651186251301378e-07, "loss": 0.0022, "step": 57662 }, { "epoch": 17.74, "learning_rate": 6.649399529541789e-07, "loss": 0.0039, "step": 57663 }, { "epoch": 17.74, "learning_rate": 6.647613039546563e-07, "loss": 0.0035, "step": 57664 }, { "epoch": 17.74, "learning_rate": 6.645826781320141e-07, "loss": 0.0037, "step": 57665 }, { "epoch": 17.74, "learning_rate": 6.644040754866976e-07, "loss": 0.004, "step": 57666 }, { "epoch": 17.74, "learning_rate": 6.642254960191485e-07, "loss": 0.0066, "step": 57667 }, { "epoch": 17.74, "learning_rate": 6.64046939729811e-07, "loss": 0.0069, "step": 57668 }, { "epoch": 17.74, "learning_rate": 6.638684066191292e-07, "loss": 0.004, "step": 57669 }, { "epoch": 17.74, "learning_rate": 6.636898966875438e-07, "loss": 0.0035, "step": 57670 }, { "epoch": 17.74, "learning_rate": 6.635114099355e-07, "loss": 0.0069, "step": 57671 }, { "epoch": 17.74, "learning_rate": 6.633329463634398e-07, "loss": 0.0071, "step": 57672 }, { "epoch": 17.75, "learning_rate": 6.631545059718069e-07, "loss": 0.0058, "step": 57673 }, { "epoch": 17.75, "learning_rate": 6.629760887610448e-07, "loss": 0.0066, "step": 57674 }, { "epoch": 17.75, "learning_rate": 6.627976947315939e-07, "loss": 0.0031, "step": 57675 }, { "epoch": 17.75, "learning_rate": 6.626193238839029e-07, "loss": 0.0048, "step": 57676 }, { "epoch": 17.75, "learning_rate": 6.624409762184081e-07, "loss": 0.0048, "step": 57677 }, { "epoch": 17.75, "learning_rate": 6.622626517355557e-07, "loss": 0.0037, "step": 57678 }, { "epoch": 17.75, "learning_rate": 6.620843504357877e-07, "loss": 0.0046, "step": 57679 }, { "epoch": 17.75, "learning_rate": 6.619060723195447e-07, "loss": 0.005, "step": 57680 }, { "epoch": 17.75, "learning_rate": 6.617278173872743e-07, "loss": 0.0059, "step": 57681 }, { "epoch": 17.75, "learning_rate": 6.615495856394149e-07, "loss": 0.0054, "step": 57682 }, { "epoch": 17.75, "learning_rate": 6.613713770764129e-07, "loss": 0.0045, "step": 57683 }, { "epoch": 17.75, "learning_rate": 6.611931916987058e-07, "loss": 0.0035, "step": 57684 }, { "epoch": 17.75, "learning_rate": 6.610150295067375e-07, "loss": 0.0077, "step": 57685 }, { "epoch": 17.75, "learning_rate": 6.608368905009521e-07, "loss": 0.0059, "step": 57686 }, { "epoch": 17.75, "learning_rate": 6.606587746817916e-07, "loss": 0.0034, "step": 57687 }, { "epoch": 17.75, "learning_rate": 6.604806820496968e-07, "loss": 0.0035, "step": 57688 }, { "epoch": 17.75, "learning_rate": 6.603026126051104e-07, "loss": 0.0043, "step": 57689 }, { "epoch": 17.75, "learning_rate": 6.601245663484757e-07, "loss": 0.0036, "step": 57690 }, { "epoch": 17.75, "learning_rate": 6.599465432802332e-07, "loss": 0.0036, "step": 57691 }, { "epoch": 17.75, "learning_rate": 6.597685434008261e-07, "loss": 0.0053, "step": 57692 }, { "epoch": 17.75, "learning_rate": 6.59590566710695e-07, "loss": 0.0049, "step": 57693 }, { "epoch": 17.75, "learning_rate": 6.594126132102819e-07, "loss": 0.0031, "step": 57694 }, { "epoch": 17.75, "learning_rate": 6.592346829000284e-07, "loss": 0.0025, "step": 57695 }, { "epoch": 17.75, "learning_rate": 6.590567757803779e-07, "loss": 0.0043, "step": 57696 }, { "epoch": 17.75, "learning_rate": 6.588788918517741e-07, "loss": 0.0044, "step": 57697 }, { "epoch": 17.75, "learning_rate": 6.587010311146524e-07, "loss": 0.0055, "step": 57698 }, { "epoch": 17.75, "learning_rate": 6.585231935694592e-07, "loss": 0.0043, "step": 57699 }, { "epoch": 17.75, "learning_rate": 6.583453792166316e-07, "loss": 0.0035, "step": 57700 }, { "epoch": 17.75, "learning_rate": 6.581675880566163e-07, "loss": 0.0057, "step": 57701 }, { "epoch": 17.75, "learning_rate": 6.579898200898526e-07, "loss": 0.0023, "step": 57702 }, { "epoch": 17.75, "learning_rate": 6.578120753167816e-07, "loss": 0.0035, "step": 57703 }, { "epoch": 17.75, "learning_rate": 6.57634353737846e-07, "loss": 0.0196, "step": 57704 }, { "epoch": 17.76, "learning_rate": 6.574566553534823e-07, "loss": 0.0039, "step": 57705 }, { "epoch": 17.76, "learning_rate": 6.572789801641378e-07, "loss": 0.0084, "step": 57706 }, { "epoch": 17.76, "learning_rate": 6.571013281702499e-07, "loss": 0.0043, "step": 57707 }, { "epoch": 17.76, "learning_rate": 6.569236993722605e-07, "loss": 0.0034, "step": 57708 }, { "epoch": 17.76, "learning_rate": 6.567460937706105e-07, "loss": 0.0031, "step": 57709 }, { "epoch": 17.76, "learning_rate": 6.565685113657416e-07, "loss": 0.0029, "step": 57710 }, { "epoch": 17.76, "learning_rate": 6.563909521580936e-07, "loss": 0.0055, "step": 57711 }, { "epoch": 17.76, "learning_rate": 6.562134161481092e-07, "loss": 0.0049, "step": 57712 }, { "epoch": 17.76, "learning_rate": 6.560359033362263e-07, "loss": 0.0068, "step": 57713 }, { "epoch": 17.76, "learning_rate": 6.558584137228885e-07, "loss": 0.0058, "step": 57714 }, { "epoch": 17.76, "learning_rate": 6.556809473085335e-07, "loss": 0.0035, "step": 57715 }, { "epoch": 17.76, "learning_rate": 6.555035040936041e-07, "loss": 0.0059, "step": 57716 }, { "epoch": 17.76, "learning_rate": 6.553260840785414e-07, "loss": 0.0035, "step": 57717 }, { "epoch": 17.76, "learning_rate": 6.551486872637857e-07, "loss": 0.0025, "step": 57718 }, { "epoch": 17.76, "learning_rate": 6.549713136497748e-07, "loss": 0.0033, "step": 57719 }, { "epoch": 17.76, "learning_rate": 6.547939632369493e-07, "loss": 0.0152, "step": 57720 }, { "epoch": 17.76, "learning_rate": 6.546166360257533e-07, "loss": 0.0031, "step": 57721 }, { "epoch": 17.76, "learning_rate": 6.544393320166243e-07, "loss": 0.0036, "step": 57722 }, { "epoch": 17.76, "learning_rate": 6.542620512100029e-07, "loss": 0.0055, "step": 57723 }, { "epoch": 17.76, "learning_rate": 6.540847936063289e-07, "loss": 0.0027, "step": 57724 }, { "epoch": 17.76, "learning_rate": 6.539075592060429e-07, "loss": 0.0046, "step": 57725 }, { "epoch": 17.76, "learning_rate": 6.537303480095858e-07, "loss": 0.0047, "step": 57726 }, { "epoch": 17.76, "learning_rate": 6.53553160017395e-07, "loss": 0.006, "step": 57727 }, { "epoch": 17.76, "learning_rate": 6.533759952299135e-07, "loss": 0.0045, "step": 57728 }, { "epoch": 17.76, "learning_rate": 6.531988536475786e-07, "loss": 0.0043, "step": 57729 }, { "epoch": 17.76, "learning_rate": 6.530217352708301e-07, "loss": 0.0037, "step": 57730 }, { "epoch": 17.76, "learning_rate": 6.52844640100111e-07, "loss": 0.005, "step": 57731 }, { "epoch": 17.76, "learning_rate": 6.526675681358585e-07, "loss": 0.0079, "step": 57732 }, { "epoch": 17.76, "learning_rate": 6.524905193785114e-07, "loss": 0.003, "step": 57733 }, { "epoch": 17.76, "learning_rate": 6.523134938285092e-07, "loss": 0.0053, "step": 57734 }, { "epoch": 17.76, "learning_rate": 6.521364914862948e-07, "loss": 0.0029, "step": 57735 }, { "epoch": 17.76, "learning_rate": 6.519595123523048e-07, "loss": 0.0047, "step": 57736 }, { "epoch": 17.76, "learning_rate": 6.517825564269786e-07, "loss": 0.005, "step": 57737 }, { "epoch": 17.77, "learning_rate": 6.516056237107571e-07, "loss": 0.0033, "step": 57738 }, { "epoch": 17.77, "learning_rate": 6.514287142040777e-07, "loss": 0.0037, "step": 57739 }, { "epoch": 17.77, "learning_rate": 6.512518279073809e-07, "loss": 0.0032, "step": 57740 }, { "epoch": 17.77, "learning_rate": 6.510749648211056e-07, "loss": 0.0047, "step": 57741 }, { "epoch": 17.77, "learning_rate": 6.508981249456902e-07, "loss": 0.0035, "step": 57742 }, { "epoch": 17.77, "learning_rate": 6.507213082815745e-07, "loss": 0.0048, "step": 57743 }, { "epoch": 17.77, "learning_rate": 6.505445148291978e-07, "loss": 0.0055, "step": 57744 }, { "epoch": 17.77, "learning_rate": 6.503677445889977e-07, "loss": 0.0047, "step": 57745 }, { "epoch": 17.77, "learning_rate": 6.501909975614151e-07, "loss": 0.0039, "step": 57746 }, { "epoch": 17.77, "learning_rate": 6.500142737468873e-07, "loss": 0.0044, "step": 57747 }, { "epoch": 17.77, "learning_rate": 6.498375731458529e-07, "loss": 0.0022, "step": 57748 }, { "epoch": 17.77, "learning_rate": 6.496608957587492e-07, "loss": 0.0034, "step": 57749 }, { "epoch": 17.77, "learning_rate": 6.494842415860192e-07, "loss": 0.0031, "step": 57750 }, { "epoch": 17.77, "learning_rate": 6.493076106280982e-07, "loss": 0.0063, "step": 57751 }, { "epoch": 17.77, "learning_rate": 6.491310028854281e-07, "loss": 0.0036, "step": 57752 }, { "epoch": 17.77, "learning_rate": 6.489544183584418e-07, "loss": 0.0042, "step": 57753 }, { "epoch": 17.77, "learning_rate": 6.4877785704758e-07, "loss": 0.0043, "step": 57754 }, { "epoch": 17.77, "learning_rate": 6.486013189532836e-07, "loss": 0.0081, "step": 57755 }, { "epoch": 17.77, "learning_rate": 6.484248040759888e-07, "loss": 0.0065, "step": 57756 }, { "epoch": 17.77, "learning_rate": 6.482483124161343e-07, "loss": 0.0054, "step": 57757 }, { "epoch": 17.77, "learning_rate": 6.480718439741574e-07, "loss": 0.0051, "step": 57758 }, { "epoch": 17.77, "learning_rate": 6.478953987504977e-07, "loss": 0.0036, "step": 57759 }, { "epoch": 17.77, "learning_rate": 6.477189767455916e-07, "loss": 0.0028, "step": 57760 }, { "epoch": 17.77, "learning_rate": 6.475425779598787e-07, "loss": 0.0047, "step": 57761 }, { "epoch": 17.77, "learning_rate": 6.473662023937966e-07, "loss": 0.0062, "step": 57762 }, { "epoch": 17.77, "learning_rate": 6.471898500477825e-07, "loss": 0.0046, "step": 57763 }, { "epoch": 17.77, "learning_rate": 6.470135209222728e-07, "loss": 0.0044, "step": 57764 }, { "epoch": 17.77, "learning_rate": 6.468372150177093e-07, "loss": 0.0045, "step": 57765 }, { "epoch": 17.77, "learning_rate": 6.466609323345274e-07, "loss": 0.0043, "step": 57766 }, { "epoch": 17.77, "learning_rate": 6.464846728731644e-07, "loss": 0.0076, "step": 57767 }, { "epoch": 17.77, "learning_rate": 6.463084366340577e-07, "loss": 0.0063, "step": 57768 }, { "epoch": 17.77, "learning_rate": 6.461322236176438e-07, "loss": 0.0039, "step": 57769 }, { "epoch": 17.78, "learning_rate": 6.459560338243642e-07, "loss": 0.004, "step": 57770 }, { "epoch": 17.78, "learning_rate": 6.457798672546545e-07, "loss": 0.006, "step": 57771 }, { "epoch": 17.78, "learning_rate": 6.456037239089508e-07, "loss": 0.0043, "step": 57772 }, { "epoch": 17.78, "learning_rate": 6.454276037876927e-07, "loss": 0.0051, "step": 57773 }, { "epoch": 17.78, "learning_rate": 6.452515068913123e-07, "loss": 0.0056, "step": 57774 }, { "epoch": 17.78, "learning_rate": 6.450754332202525e-07, "loss": 0.0055, "step": 57775 }, { "epoch": 17.78, "learning_rate": 6.448993827749484e-07, "loss": 0.0036, "step": 57776 }, { "epoch": 17.78, "learning_rate": 6.447233555558374e-07, "loss": 0.0036, "step": 57777 }, { "epoch": 17.78, "learning_rate": 6.445473515633561e-07, "loss": 0.0036, "step": 57778 }, { "epoch": 17.78, "learning_rate": 6.443713707979415e-07, "loss": 0.0039, "step": 57779 }, { "epoch": 17.78, "learning_rate": 6.441954132600303e-07, "loss": 0.0035, "step": 57780 }, { "epoch": 17.78, "learning_rate": 6.440194789500598e-07, "loss": 0.0031, "step": 57781 }, { "epoch": 17.78, "learning_rate": 6.438435678684662e-07, "loss": 0.0061, "step": 57782 }, { "epoch": 17.78, "learning_rate": 6.436676800156883e-07, "loss": 0.0038, "step": 57783 }, { "epoch": 17.78, "learning_rate": 6.434918153921576e-07, "loss": 0.004, "step": 57784 }, { "epoch": 17.78, "learning_rate": 6.433159739983174e-07, "loss": 0.0027, "step": 57785 }, { "epoch": 17.78, "learning_rate": 6.431401558346018e-07, "loss": 0.0022, "step": 57786 }, { "epoch": 17.78, "learning_rate": 6.429643609014447e-07, "loss": 0.0057, "step": 57787 }, { "epoch": 17.78, "learning_rate": 6.427885891992858e-07, "loss": 0.0055, "step": 57788 }, { "epoch": 17.78, "learning_rate": 6.426128407285581e-07, "loss": 0.0039, "step": 57789 }, { "epoch": 17.78, "learning_rate": 6.424371154897013e-07, "loss": 0.0055, "step": 57790 }, { "epoch": 17.78, "learning_rate": 6.422614134831506e-07, "loss": 0.0028, "step": 57791 }, { "epoch": 17.78, "learning_rate": 6.420857347093423e-07, "loss": 0.0033, "step": 57792 }, { "epoch": 17.78, "learning_rate": 6.419100791687116e-07, "loss": 0.0035, "step": 57793 }, { "epoch": 17.78, "learning_rate": 6.41734446861696e-07, "loss": 0.0035, "step": 57794 }, { "epoch": 17.78, "learning_rate": 6.415588377887305e-07, "loss": 0.0039, "step": 57795 }, { "epoch": 17.78, "learning_rate": 6.413832519502517e-07, "loss": 0.004, "step": 57796 }, { "epoch": 17.78, "learning_rate": 6.412076893466945e-07, "loss": 0.003, "step": 57797 }, { "epoch": 17.78, "learning_rate": 6.410321499784966e-07, "loss": 0.0037, "step": 57798 }, { "epoch": 17.78, "learning_rate": 6.408566338460931e-07, "loss": 0.0043, "step": 57799 }, { "epoch": 17.78, "learning_rate": 6.406811409499181e-07, "loss": 0.0022, "step": 57800 }, { "epoch": 17.78, "learning_rate": 6.405056712904089e-07, "loss": 0.0041, "step": 57801 }, { "epoch": 17.78, "learning_rate": 6.403302248680021e-07, "loss": 0.0016, "step": 57802 }, { "epoch": 17.79, "learning_rate": 6.401548016831305e-07, "loss": 0.0038, "step": 57803 }, { "epoch": 17.79, "learning_rate": 6.399794017362304e-07, "loss": 0.0045, "step": 57804 }, { "epoch": 17.79, "learning_rate": 6.398040250277393e-07, "loss": 0.0046, "step": 57805 }, { "epoch": 17.79, "learning_rate": 6.396286715580923e-07, "loss": 0.0044, "step": 57806 }, { "epoch": 17.79, "learning_rate": 6.394533413277249e-07, "loss": 0.0028, "step": 57807 }, { "epoch": 17.79, "learning_rate": 6.392780343370686e-07, "loss": 0.0029, "step": 57808 }, { "epoch": 17.79, "learning_rate": 6.39102750586561e-07, "loss": 0.0038, "step": 57809 }, { "epoch": 17.79, "learning_rate": 6.389274900766396e-07, "loss": 0.0055, "step": 57810 }, { "epoch": 17.79, "learning_rate": 6.387522528077372e-07, "loss": 0.0051, "step": 57811 }, { "epoch": 17.79, "learning_rate": 6.385770387802881e-07, "loss": 0.0025, "step": 57812 }, { "epoch": 17.79, "learning_rate": 6.384018479947296e-07, "loss": 0.0043, "step": 57813 }, { "epoch": 17.79, "learning_rate": 6.382266804514959e-07, "loss": 0.0026, "step": 57814 }, { "epoch": 17.79, "learning_rate": 6.380515361510209e-07, "loss": 0.0048, "step": 57815 }, { "epoch": 17.79, "learning_rate": 6.3787641509374e-07, "loss": 0.0042, "step": 57816 }, { "epoch": 17.79, "learning_rate": 6.377013172800872e-07, "loss": 0.0071, "step": 57817 }, { "epoch": 17.79, "learning_rate": 6.375262427104989e-07, "loss": 0.0043, "step": 57818 }, { "epoch": 17.79, "learning_rate": 6.37351191385408e-07, "loss": 0.0041, "step": 57819 }, { "epoch": 17.79, "learning_rate": 6.37176163305252e-07, "loss": 0.0039, "step": 57820 }, { "epoch": 17.79, "learning_rate": 6.370011584704617e-07, "loss": 0.0021, "step": 57821 }, { "epoch": 17.79, "learning_rate": 6.368261768814743e-07, "loss": 0.0024, "step": 57822 }, { "epoch": 17.79, "learning_rate": 6.366512185387231e-07, "loss": 0.0037, "step": 57823 }, { "epoch": 17.79, "learning_rate": 6.364762834426408e-07, "loss": 0.0021, "step": 57824 }, { "epoch": 17.79, "learning_rate": 6.36301371593665e-07, "loss": 0.0054, "step": 57825 }, { "epoch": 17.79, "learning_rate": 6.361264829922287e-07, "loss": 0.0025, "step": 57826 }, { "epoch": 17.79, "learning_rate": 6.359516176387659e-07, "loss": 0.0042, "step": 57827 }, { "epoch": 17.79, "learning_rate": 6.35776775533713e-07, "loss": 0.0034, "step": 57828 }, { "epoch": 17.79, "learning_rate": 6.356019566774974e-07, "loss": 0.0034, "step": 57829 }, { "epoch": 17.79, "learning_rate": 6.354271610705598e-07, "loss": 0.003, "step": 57830 }, { "epoch": 17.79, "learning_rate": 6.352523887133321e-07, "loss": 0.0028, "step": 57831 }, { "epoch": 17.79, "learning_rate": 6.350776396062475e-07, "loss": 0.0049, "step": 57832 }, { "epoch": 17.79, "learning_rate": 6.349029137497409e-07, "loss": 0.0031, "step": 57833 }, { "epoch": 17.79, "learning_rate": 6.347282111442455e-07, "loss": 0.0028, "step": 57834 }, { "epoch": 17.8, "learning_rate": 6.345535317901952e-07, "loss": 0.005, "step": 57835 }, { "epoch": 17.8, "learning_rate": 6.343788756880243e-07, "loss": 0.0029, "step": 57836 }, { "epoch": 17.8, "learning_rate": 6.342042428381645e-07, "loss": 0.0039, "step": 57837 }, { "epoch": 17.8, "learning_rate": 6.340296332410512e-07, "loss": 0.0038, "step": 57838 }, { "epoch": 17.8, "learning_rate": 6.338550468971161e-07, "loss": 0.0046, "step": 57839 }, { "epoch": 17.8, "learning_rate": 6.336804838067956e-07, "loss": 0.0037, "step": 57840 }, { "epoch": 17.8, "learning_rate": 6.335059439705226e-07, "loss": 0.0058, "step": 57841 }, { "epoch": 17.8, "learning_rate": 6.33331427388727e-07, "loss": 0.0044, "step": 57842 }, { "epoch": 17.8, "learning_rate": 6.331569340618448e-07, "loss": 0.0037, "step": 57843 }, { "epoch": 17.8, "learning_rate": 6.32982463990307e-07, "loss": 0.003, "step": 57844 }, { "epoch": 17.8, "learning_rate": 6.32808017174551e-07, "loss": 0.0034, "step": 57845 }, { "epoch": 17.8, "learning_rate": 6.326335936150064e-07, "loss": 0.0022, "step": 57846 }, { "epoch": 17.8, "learning_rate": 6.324591933121072e-07, "loss": 0.004, "step": 57847 }, { "epoch": 17.8, "learning_rate": 6.322848162662865e-07, "loss": 0.0064, "step": 57848 }, { "epoch": 17.8, "learning_rate": 6.321104624779773e-07, "loss": 0.006, "step": 57849 }, { "epoch": 17.8, "learning_rate": 6.319361319476125e-07, "loss": 0.0048, "step": 57850 }, { "epoch": 17.8, "learning_rate": 6.317618246756241e-07, "loss": 0.0049, "step": 57851 }, { "epoch": 17.8, "learning_rate": 6.315875406624461e-07, "loss": 0.0038, "step": 57852 }, { "epoch": 17.8, "learning_rate": 6.314132799085104e-07, "loss": 0.0027, "step": 57853 }, { "epoch": 17.8, "learning_rate": 6.312390424142479e-07, "loss": 0.0034, "step": 57854 }, { "epoch": 17.8, "learning_rate": 6.310648281800969e-07, "loss": 0.0048, "step": 57855 }, { "epoch": 17.8, "learning_rate": 6.308906372064838e-07, "loss": 0.0044, "step": 57856 }, { "epoch": 17.8, "learning_rate": 6.307164694938439e-07, "loss": 0.0058, "step": 57857 }, { "epoch": 17.8, "learning_rate": 6.305423250426091e-07, "loss": 0.0067, "step": 57858 }, { "epoch": 17.8, "learning_rate": 6.303682038532111e-07, "loss": 0.0076, "step": 57859 }, { "epoch": 17.8, "learning_rate": 6.30194105926083e-07, "loss": 0.0044, "step": 57860 }, { "epoch": 17.8, "learning_rate": 6.300200312616578e-07, "loss": 0.0073, "step": 57861 }, { "epoch": 17.8, "learning_rate": 6.298459798603684e-07, "loss": 0.008, "step": 57862 }, { "epoch": 17.8, "learning_rate": 6.296719517226424e-07, "loss": 0.0032, "step": 57863 }, { "epoch": 17.8, "learning_rate": 6.294979468489149e-07, "loss": 0.0044, "step": 57864 }, { "epoch": 17.8, "learning_rate": 6.293239652396188e-07, "loss": 0.0058, "step": 57865 }, { "epoch": 17.8, "learning_rate": 6.29150006895185e-07, "loss": 0.0046, "step": 57866 }, { "epoch": 17.8, "learning_rate": 6.289760718160454e-07, "loss": 0.0041, "step": 57867 }, { "epoch": 17.81, "learning_rate": 6.288021600026318e-07, "loss": 0.0049, "step": 57868 }, { "epoch": 17.81, "learning_rate": 6.28628271455376e-07, "loss": 0.0029, "step": 57869 }, { "epoch": 17.81, "learning_rate": 6.284544061747089e-07, "loss": 0.0022, "step": 57870 }, { "epoch": 17.81, "learning_rate": 6.282805641610646e-07, "loss": 0.0038, "step": 57871 }, { "epoch": 17.81, "learning_rate": 6.281067454148726e-07, "loss": 0.0055, "step": 57872 }, { "epoch": 17.81, "learning_rate": 6.279329499365649e-07, "loss": 0.0049, "step": 57873 }, { "epoch": 17.81, "learning_rate": 6.277591777265713e-07, "loss": 0.009, "step": 57874 }, { "epoch": 17.81, "learning_rate": 6.275854287853267e-07, "loss": 0.0038, "step": 57875 }, { "epoch": 17.81, "learning_rate": 6.274117031132621e-07, "loss": 0.0028, "step": 57876 }, { "epoch": 17.81, "learning_rate": 6.272380007108059e-07, "loss": 0.0031, "step": 57877 }, { "epoch": 17.81, "learning_rate": 6.270643215783911e-07, "loss": 0.0065, "step": 57878 }, { "epoch": 17.81, "learning_rate": 6.268906657164476e-07, "loss": 0.0035, "step": 57879 }, { "epoch": 17.81, "learning_rate": 6.26717033125408e-07, "loss": 0.0058, "step": 57880 }, { "epoch": 17.81, "learning_rate": 6.265434238057044e-07, "loss": 0.0048, "step": 57881 }, { "epoch": 17.81, "learning_rate": 6.263698377577654e-07, "loss": 0.0058, "step": 57882 }, { "epoch": 17.81, "learning_rate": 6.26196274982025e-07, "loss": 0.0046, "step": 57883 }, { "epoch": 17.81, "learning_rate": 6.260227354789095e-07, "loss": 0.0069, "step": 57884 }, { "epoch": 17.81, "learning_rate": 6.25849219248853e-07, "loss": 0.0025, "step": 57885 }, { "epoch": 17.81, "learning_rate": 6.256757262922863e-07, "loss": 0.0047, "step": 57886 }, { "epoch": 17.81, "learning_rate": 6.255022566096402e-07, "loss": 0.0054, "step": 57887 }, { "epoch": 17.81, "learning_rate": 6.253288102013444e-07, "loss": 0.0043, "step": 57888 }, { "epoch": 17.81, "learning_rate": 6.251553870678284e-07, "loss": 0.0035, "step": 57889 }, { "epoch": 17.81, "learning_rate": 6.249819872095253e-07, "loss": 0.0038, "step": 57890 }, { "epoch": 17.81, "learning_rate": 6.248086106268647e-07, "loss": 0.0022, "step": 57891 }, { "epoch": 17.81, "learning_rate": 6.246352573202774e-07, "loss": 0.003, "step": 57892 }, { "epoch": 17.81, "learning_rate": 6.24461927290192e-07, "loss": 0.0041, "step": 57893 }, { "epoch": 17.81, "learning_rate": 6.242886205370391e-07, "loss": 0.0044, "step": 57894 }, { "epoch": 17.81, "learning_rate": 6.241153370612507e-07, "loss": 0.0135, "step": 57895 }, { "epoch": 17.81, "learning_rate": 6.239420768632576e-07, "loss": 0.0045, "step": 57896 }, { "epoch": 17.81, "learning_rate": 6.237688399434894e-07, "loss": 0.0055, "step": 57897 }, { "epoch": 17.81, "learning_rate": 6.235956263023735e-07, "loss": 0.0038, "step": 57898 }, { "epoch": 17.81, "learning_rate": 6.234224359403407e-07, "loss": 0.004, "step": 57899 }, { "epoch": 17.82, "learning_rate": 6.23249268857824e-07, "loss": 0.0047, "step": 57900 }, { "epoch": 17.82, "learning_rate": 6.230761250552508e-07, "loss": 0.0049, "step": 57901 }, { "epoch": 17.82, "learning_rate": 6.22903004533052e-07, "loss": 0.007, "step": 57902 }, { "epoch": 17.82, "learning_rate": 6.227299072916571e-07, "loss": 0.0033, "step": 57903 }, { "epoch": 17.82, "learning_rate": 6.225568333314968e-07, "loss": 0.0047, "step": 57904 }, { "epoch": 17.82, "learning_rate": 6.223837826529988e-07, "loss": 0.0037, "step": 57905 }, { "epoch": 17.82, "learning_rate": 6.222107552565937e-07, "loss": 0.0039, "step": 57906 }, { "epoch": 17.82, "learning_rate": 6.220377511427112e-07, "loss": 0.0039, "step": 57907 }, { "epoch": 17.82, "learning_rate": 6.218647703117808e-07, "loss": 0.0091, "step": 57908 }, { "epoch": 17.82, "learning_rate": 6.216918127642301e-07, "loss": 0.0038, "step": 57909 }, { "epoch": 17.82, "learning_rate": 6.215188785004933e-07, "loss": 0.0043, "step": 57910 }, { "epoch": 17.82, "learning_rate": 6.213459675209954e-07, "loss": 0.0077, "step": 57911 }, { "epoch": 17.82, "learning_rate": 6.211730798261673e-07, "loss": 0.0025, "step": 57912 }, { "epoch": 17.82, "learning_rate": 6.210002154164374e-07, "loss": 0.0037, "step": 57913 }, { "epoch": 17.82, "learning_rate": 6.208273742922343e-07, "loss": 0.0037, "step": 57914 }, { "epoch": 17.82, "learning_rate": 6.206545564539901e-07, "loss": 0.0055, "step": 57915 }, { "epoch": 17.82, "learning_rate": 6.20481761902132e-07, "loss": 0.0032, "step": 57916 }, { "epoch": 17.82, "learning_rate": 6.203089906370885e-07, "loss": 0.0019, "step": 57917 }, { "epoch": 17.82, "learning_rate": 6.201362426592917e-07, "loss": 0.0056, "step": 57918 }, { "epoch": 17.82, "learning_rate": 6.199635179691643e-07, "loss": 0.0048, "step": 57919 }, { "epoch": 17.82, "learning_rate": 6.197908165671396e-07, "loss": 0.0087, "step": 57920 }, { "epoch": 17.82, "learning_rate": 6.19618138453646e-07, "loss": 0.0037, "step": 57921 }, { "epoch": 17.82, "learning_rate": 6.19445483629112e-07, "loss": 0.0045, "step": 57922 }, { "epoch": 17.82, "learning_rate": 6.19272852093965e-07, "loss": 0.0056, "step": 57923 }, { "epoch": 17.82, "learning_rate": 6.191002438486349e-07, "loss": 0.004, "step": 57924 }, { "epoch": 17.82, "learning_rate": 6.1892765889355e-07, "loss": 0.0041, "step": 57925 }, { "epoch": 17.82, "learning_rate": 6.18755097229139e-07, "loss": 0.0034, "step": 57926 }, { "epoch": 17.82, "learning_rate": 6.185825588558303e-07, "loss": 0.0046, "step": 57927 }, { "epoch": 17.82, "learning_rate": 6.184100437740514e-07, "loss": 0.0026, "step": 57928 }, { "epoch": 17.82, "learning_rate": 6.182375519842299e-07, "loss": 0.0017, "step": 57929 }, { "epoch": 17.82, "learning_rate": 6.180650834867963e-07, "loss": 0.0043, "step": 57930 }, { "epoch": 17.82, "learning_rate": 6.178926382821793e-07, "loss": 0.0037, "step": 57931 }, { "epoch": 17.82, "learning_rate": 6.17720216370804e-07, "loss": 0.0068, "step": 57932 }, { "epoch": 17.83, "learning_rate": 6.175478177531002e-07, "loss": 0.0059, "step": 57933 }, { "epoch": 17.83, "learning_rate": 6.173754424294953e-07, "loss": 0.0017, "step": 57934 }, { "epoch": 17.83, "learning_rate": 6.172030904004179e-07, "loss": 0.009, "step": 57935 }, { "epoch": 17.83, "learning_rate": 6.170307616662963e-07, "loss": 0.0057, "step": 57936 }, { "epoch": 17.83, "learning_rate": 6.16858456227557e-07, "loss": 0.0042, "step": 57937 }, { "epoch": 17.83, "learning_rate": 6.166861740846297e-07, "loss": 0.0032, "step": 57938 }, { "epoch": 17.83, "learning_rate": 6.165139152379396e-07, "loss": 0.006, "step": 57939 }, { "epoch": 17.83, "learning_rate": 6.163416796879162e-07, "loss": 0.0054, "step": 57940 }, { "epoch": 17.83, "learning_rate": 6.161694674349872e-07, "loss": 0.003, "step": 57941 }, { "epoch": 17.83, "learning_rate": 6.159972784795798e-07, "loss": 0.0043, "step": 57942 }, { "epoch": 17.83, "learning_rate": 6.158251128221205e-07, "loss": 0.0052, "step": 57943 }, { "epoch": 17.83, "learning_rate": 6.156529704630376e-07, "loss": 0.0036, "step": 57944 }, { "epoch": 17.83, "learning_rate": 6.154808514027588e-07, "loss": 0.01, "step": 57945 }, { "epoch": 17.83, "learning_rate": 6.153087556417115e-07, "loss": 0.0029, "step": 57946 }, { "epoch": 17.83, "learning_rate": 6.151366831803218e-07, "loss": 0.0031, "step": 57947 }, { "epoch": 17.83, "learning_rate": 6.149646340190174e-07, "loss": 0.0039, "step": 57948 }, { "epoch": 17.83, "learning_rate": 6.147926081582256e-07, "loss": 0.0053, "step": 57949 }, { "epoch": 17.83, "learning_rate": 6.146206055983739e-07, "loss": 0.0035, "step": 57950 }, { "epoch": 17.83, "learning_rate": 6.144486263398886e-07, "loss": 0.0035, "step": 57951 }, { "epoch": 17.83, "learning_rate": 6.142766703831992e-07, "loss": 0.0055, "step": 57952 }, { "epoch": 17.83, "learning_rate": 6.14104737728729e-07, "loss": 0.0033, "step": 57953 }, { "epoch": 17.83, "learning_rate": 6.139328283769052e-07, "loss": 0.0036, "step": 57954 }, { "epoch": 17.83, "learning_rate": 6.137609423281577e-07, "loss": 0.003, "step": 57955 }, { "epoch": 17.83, "learning_rate": 6.135890795829102e-07, "loss": 0.0029, "step": 57956 }, { "epoch": 17.83, "learning_rate": 6.134172401415905e-07, "loss": 0.0051, "step": 57957 }, { "epoch": 17.83, "learning_rate": 6.132454240046259e-07, "loss": 0.0037, "step": 57958 }, { "epoch": 17.83, "learning_rate": 6.130736311724417e-07, "loss": 0.0039, "step": 57959 }, { "epoch": 17.83, "learning_rate": 6.129018616454663e-07, "loss": 0.0027, "step": 57960 }, { "epoch": 17.83, "learning_rate": 6.127301154241239e-07, "loss": 0.0051, "step": 57961 }, { "epoch": 17.83, "learning_rate": 6.125583925088418e-07, "loss": 0.0049, "step": 57962 }, { "epoch": 17.83, "learning_rate": 6.123866929000466e-07, "loss": 0.004, "step": 57963 }, { "epoch": 17.83, "learning_rate": 6.122150165981633e-07, "loss": 0.0035, "step": 57964 }, { "epoch": 17.84, "learning_rate": 6.120433636036216e-07, "loss": 0.0052, "step": 57965 }, { "epoch": 17.84, "learning_rate": 6.118717339168445e-07, "loss": 0.0032, "step": 57966 }, { "epoch": 17.84, "learning_rate": 6.117001275382584e-07, "loss": 0.0046, "step": 57967 }, { "epoch": 17.84, "learning_rate": 6.115285444682895e-07, "loss": 0.0045, "step": 57968 }, { "epoch": 17.84, "learning_rate": 6.11356984707363e-07, "loss": 0.0018, "step": 57969 }, { "epoch": 17.84, "learning_rate": 6.111854482559088e-07, "loss": 0.0046, "step": 57970 }, { "epoch": 17.84, "learning_rate": 6.110139351143485e-07, "loss": 0.0042, "step": 57971 }, { "epoch": 17.84, "learning_rate": 6.108424452831096e-07, "loss": 0.0038, "step": 57972 }, { "epoch": 17.84, "learning_rate": 6.106709787626198e-07, "loss": 0.0051, "step": 57973 }, { "epoch": 17.84, "learning_rate": 6.104995355532994e-07, "loss": 0.0064, "step": 57974 }, { "epoch": 17.84, "learning_rate": 6.103281156555784e-07, "loss": 0.0071, "step": 57975 }, { "epoch": 17.84, "learning_rate": 6.10156719069881e-07, "loss": 0.0033, "step": 57976 }, { "epoch": 17.84, "learning_rate": 6.099853457966342e-07, "loss": 0.0057, "step": 57977 }, { "epoch": 17.84, "learning_rate": 6.098139958362615e-07, "loss": 0.0051, "step": 57978 }, { "epoch": 17.84, "learning_rate": 6.096426691891899e-07, "loss": 0.0028, "step": 57979 }, { "epoch": 17.84, "learning_rate": 6.094713658558427e-07, "loss": 0.0048, "step": 57980 }, { "epoch": 17.84, "learning_rate": 6.093000858366471e-07, "loss": 0.0047, "step": 57981 }, { "epoch": 17.84, "learning_rate": 6.091288291320285e-07, "loss": 0.0036, "step": 57982 }, { "epoch": 17.84, "learning_rate": 6.08957595742411e-07, "loss": 0.0049, "step": 57983 }, { "epoch": 17.84, "learning_rate": 6.087863856682175e-07, "loss": 0.0034, "step": 57984 }, { "epoch": 17.84, "learning_rate": 6.086151989098787e-07, "loss": 0.0029, "step": 57985 }, { "epoch": 17.84, "learning_rate": 6.084440354678168e-07, "loss": 0.004, "step": 57986 }, { "epoch": 17.84, "learning_rate": 6.082728953424555e-07, "loss": 0.0025, "step": 57987 }, { "epoch": 17.84, "learning_rate": 6.081017785342202e-07, "loss": 0.0044, "step": 57988 }, { "epoch": 17.84, "learning_rate": 6.079306850435351e-07, "loss": 0.0028, "step": 57989 }, { "epoch": 17.84, "learning_rate": 6.077596148708275e-07, "loss": 0.0021, "step": 57990 }, { "epoch": 17.84, "learning_rate": 6.075885680165206e-07, "loss": 0.0044, "step": 57991 }, { "epoch": 17.84, "learning_rate": 6.074175444810393e-07, "loss": 0.0028, "step": 57992 }, { "epoch": 17.84, "learning_rate": 6.072465442648079e-07, "loss": 0.004, "step": 57993 }, { "epoch": 17.84, "learning_rate": 6.070755673682516e-07, "loss": 0.002, "step": 57994 }, { "epoch": 17.84, "learning_rate": 6.069046137917944e-07, "loss": 0.0044, "step": 57995 }, { "epoch": 17.84, "learning_rate": 6.067336835358606e-07, "loss": 0.0048, "step": 57996 }, { "epoch": 17.84, "learning_rate": 6.065627766008742e-07, "loss": 0.0048, "step": 57997 }, { "epoch": 17.85, "learning_rate": 6.063918929872604e-07, "loss": 0.0081, "step": 57998 }, { "epoch": 17.85, "learning_rate": 6.062210326954432e-07, "loss": 0.0022, "step": 57999 }, { "epoch": 17.85, "learning_rate": 6.060501957258469e-07, "loss": 0.0037, "step": 58000 }, { "epoch": 17.85, "learning_rate": 6.058793820788944e-07, "loss": 0.0046, "step": 58001 }, { "epoch": 17.85, "learning_rate": 6.05708591755012e-07, "loss": 0.0035, "step": 58002 }, { "epoch": 17.85, "learning_rate": 6.055378247546217e-07, "loss": 0.0023, "step": 58003 }, { "epoch": 17.85, "learning_rate": 6.053670810781476e-07, "loss": 0.0045, "step": 58004 }, { "epoch": 17.85, "learning_rate": 6.051963607260148e-07, "loss": 0.0047, "step": 58005 }, { "epoch": 17.85, "learning_rate": 6.050256636986473e-07, "loss": 0.0031, "step": 58006 }, { "epoch": 17.85, "learning_rate": 6.048549899964706e-07, "loss": 0.0028, "step": 58007 }, { "epoch": 17.85, "learning_rate": 6.046843396199031e-07, "loss": 0.0038, "step": 58008 }, { "epoch": 17.85, "learning_rate": 6.045137125693712e-07, "loss": 0.0046, "step": 58009 }, { "epoch": 17.85, "learning_rate": 6.043431088452989e-07, "loss": 0.0053, "step": 58010 }, { "epoch": 17.85, "learning_rate": 6.041725284481115e-07, "loss": 0.0061, "step": 58011 }, { "epoch": 17.85, "learning_rate": 6.040019713782297e-07, "loss": 0.0046, "step": 58012 }, { "epoch": 17.85, "learning_rate": 6.038314376360776e-07, "loss": 0.0045, "step": 58013 }, { "epoch": 17.85, "learning_rate": 6.036609272220795e-07, "loss": 0.0054, "step": 58014 }, { "epoch": 17.85, "learning_rate": 6.03490440136657e-07, "loss": 0.0036, "step": 58015 }, { "epoch": 17.85, "learning_rate": 6.033199763802356e-07, "loss": 0.0061, "step": 58016 }, { "epoch": 17.85, "learning_rate": 6.03149535953238e-07, "loss": 0.0036, "step": 58017 }, { "epoch": 17.85, "learning_rate": 6.029791188560851e-07, "loss": 0.0028, "step": 58018 }, { "epoch": 17.85, "learning_rate": 6.028087250892012e-07, "loss": 0.0044, "step": 58019 }, { "epoch": 17.85, "learning_rate": 6.026383546530134e-07, "loss": 0.0032, "step": 58020 }, { "epoch": 17.85, "learning_rate": 6.024680075479384e-07, "loss": 0.002, "step": 58021 }, { "epoch": 17.85, "learning_rate": 6.022976837744033e-07, "loss": 0.0049, "step": 58022 }, { "epoch": 17.85, "learning_rate": 6.021273833328279e-07, "loss": 0.0052, "step": 58023 }, { "epoch": 17.85, "learning_rate": 6.019571062236362e-07, "loss": 0.0035, "step": 58024 }, { "epoch": 17.85, "learning_rate": 6.017868524472525e-07, "loss": 0.0042, "step": 58025 }, { "epoch": 17.85, "learning_rate": 6.016166220040976e-07, "loss": 0.0062, "step": 58026 }, { "epoch": 17.85, "learning_rate": 6.014464148945964e-07, "loss": 0.0026, "step": 58027 }, { "epoch": 17.85, "learning_rate": 6.012762311191699e-07, "loss": 0.0038, "step": 58028 }, { "epoch": 17.85, "learning_rate": 6.01106070678239e-07, "loss": 0.0027, "step": 58029 }, { "epoch": 17.86, "learning_rate": 6.009359335722287e-07, "loss": 0.0046, "step": 58030 }, { "epoch": 17.86, "learning_rate": 6.007658198015609e-07, "loss": 0.0035, "step": 58031 }, { "epoch": 17.86, "learning_rate": 6.005957293666576e-07, "loss": 0.0022, "step": 58032 }, { "epoch": 17.86, "learning_rate": 6.004256622679405e-07, "loss": 0.0072, "step": 58033 }, { "epoch": 17.86, "learning_rate": 6.002556185058339e-07, "loss": 0.0045, "step": 58034 }, { "epoch": 17.86, "learning_rate": 6.000855980807574e-07, "loss": 0.0046, "step": 58035 }, { "epoch": 17.86, "learning_rate": 5.99915600993134e-07, "loss": 0.005, "step": 58036 }, { "epoch": 17.86, "learning_rate": 5.997456272433866e-07, "loss": 0.0036, "step": 58037 }, { "epoch": 17.86, "learning_rate": 5.995756768319361e-07, "loss": 0.0059, "step": 58038 }, { "epoch": 17.86, "learning_rate": 5.994057497592032e-07, "loss": 0.0035, "step": 58039 }, { "epoch": 17.86, "learning_rate": 5.992358460256142e-07, "loss": 0.0055, "step": 58040 }, { "epoch": 17.86, "learning_rate": 5.99065965631589e-07, "loss": 0.0034, "step": 58041 }, { "epoch": 17.86, "learning_rate": 5.988961085775468e-07, "loss": 0.0034, "step": 58042 }, { "epoch": 17.86, "learning_rate": 5.987262748639111e-07, "loss": 0.0052, "step": 58043 }, { "epoch": 17.86, "learning_rate": 5.985564644911024e-07, "loss": 0.0093, "step": 58044 }, { "epoch": 17.86, "learning_rate": 5.983866774595448e-07, "loss": 0.0037, "step": 58045 }, { "epoch": 17.86, "learning_rate": 5.982169137696581e-07, "loss": 0.0024, "step": 58046 }, { "epoch": 17.86, "learning_rate": 5.980471734218651e-07, "loss": 0.0022, "step": 58047 }, { "epoch": 17.86, "learning_rate": 5.978774564165857e-07, "loss": 0.0029, "step": 58048 }, { "epoch": 17.86, "learning_rate": 5.977077627542416e-07, "loss": 0.0023, "step": 58049 }, { "epoch": 17.86, "learning_rate": 5.975380924352548e-07, "loss": 0.005, "step": 58050 }, { "epoch": 17.86, "learning_rate": 5.973684454600459e-07, "loss": 0.0048, "step": 58051 }, { "epoch": 17.86, "learning_rate": 5.97198821829037e-07, "loss": 0.0043, "step": 58052 }, { "epoch": 17.86, "learning_rate": 5.970292215426476e-07, "loss": 0.0042, "step": 58053 }, { "epoch": 17.86, "learning_rate": 5.968596446012986e-07, "loss": 0.0047, "step": 58054 }, { "epoch": 17.86, "learning_rate": 5.966900910054141e-07, "loss": 0.0038, "step": 58055 }, { "epoch": 17.86, "learning_rate": 5.965205607554125e-07, "loss": 0.0037, "step": 58056 }, { "epoch": 17.86, "learning_rate": 5.963510538517159e-07, "loss": 0.0019, "step": 58057 }, { "epoch": 17.86, "learning_rate": 5.961815702947438e-07, "loss": 0.0046, "step": 58058 }, { "epoch": 17.86, "learning_rate": 5.960121100849159e-07, "loss": 0.0035, "step": 58059 }, { "epoch": 17.86, "learning_rate": 5.958426732226564e-07, "loss": 0.0037, "step": 58060 }, { "epoch": 17.86, "learning_rate": 5.956732597083837e-07, "loss": 0.0026, "step": 58061 }, { "epoch": 17.86, "learning_rate": 5.955038695425208e-07, "loss": 0.0044, "step": 58062 }, { "epoch": 17.87, "learning_rate": 5.953345027254854e-07, "loss": 0.0071, "step": 58063 }, { "epoch": 17.87, "learning_rate": 5.951651592576968e-07, "loss": 0.0104, "step": 58064 }, { "epoch": 17.87, "learning_rate": 5.949958391395805e-07, "loss": 0.0036, "step": 58065 }, { "epoch": 17.87, "learning_rate": 5.948265423715526e-07, "loss": 0.0047, "step": 58066 }, { "epoch": 17.87, "learning_rate": 5.946572689540353e-07, "loss": 0.0042, "step": 58067 }, { "epoch": 17.87, "learning_rate": 5.94488018887448e-07, "loss": 0.0027, "step": 58068 }, { "epoch": 17.87, "learning_rate": 5.943187921722127e-07, "loss": 0.0052, "step": 58069 }, { "epoch": 17.87, "learning_rate": 5.941495888087467e-07, "loss": 0.0065, "step": 58070 }, { "epoch": 17.87, "learning_rate": 5.939804087974721e-07, "loss": 0.0057, "step": 58071 }, { "epoch": 17.87, "learning_rate": 5.938112521388085e-07, "loss": 0.0082, "step": 58072 }, { "epoch": 17.87, "learning_rate": 5.936421188331754e-07, "loss": 0.0021, "step": 58073 }, { "epoch": 17.87, "learning_rate": 5.934730088809915e-07, "loss": 0.0046, "step": 58074 }, { "epoch": 17.87, "learning_rate": 5.933039222826797e-07, "loss": 0.0024, "step": 58075 }, { "epoch": 17.87, "learning_rate": 5.931348590386599e-07, "loss": 0.0037, "step": 58076 }, { "epoch": 17.87, "learning_rate": 5.929658191493492e-07, "loss": 0.0025, "step": 58077 }, { "epoch": 17.87, "learning_rate": 5.927968026151676e-07, "loss": 0.0029, "step": 58078 }, { "epoch": 17.87, "learning_rate": 5.926278094365345e-07, "loss": 0.0036, "step": 58079 }, { "epoch": 17.87, "learning_rate": 5.924588396138708e-07, "loss": 0.0048, "step": 58080 }, { "epoch": 17.87, "learning_rate": 5.922898931475973e-07, "loss": 0.0037, "step": 58081 }, { "epoch": 17.87, "learning_rate": 5.921209700381303e-07, "loss": 0.0027, "step": 58082 }, { "epoch": 17.87, "learning_rate": 5.919520702858928e-07, "loss": 0.0035, "step": 58083 }, { "epoch": 17.87, "learning_rate": 5.917831938912988e-07, "loss": 0.0054, "step": 58084 }, { "epoch": 17.87, "learning_rate": 5.916143408547725e-07, "loss": 0.0055, "step": 58085 }, { "epoch": 17.87, "learning_rate": 5.914455111767314e-07, "loss": 0.0079, "step": 58086 }, { "epoch": 17.87, "learning_rate": 5.91276704857594e-07, "loss": 0.0053, "step": 58087 }, { "epoch": 17.87, "learning_rate": 5.911079218977811e-07, "loss": 0.0059, "step": 58088 }, { "epoch": 17.87, "learning_rate": 5.909391622977101e-07, "loss": 0.0028, "step": 58089 }, { "epoch": 17.87, "learning_rate": 5.907704260577995e-07, "loss": 0.0019, "step": 58090 }, { "epoch": 17.87, "learning_rate": 5.906017131784714e-07, "loss": 0.0053, "step": 58091 }, { "epoch": 17.87, "learning_rate": 5.904330236601408e-07, "loss": 0.0067, "step": 58092 }, { "epoch": 17.87, "learning_rate": 5.902643575032285e-07, "loss": 0.0017, "step": 58093 }, { "epoch": 17.87, "learning_rate": 5.90095714708152e-07, "loss": 0.0054, "step": 58094 }, { "epoch": 17.88, "learning_rate": 5.899270952753333e-07, "loss": 0.0049, "step": 58095 }, { "epoch": 17.88, "learning_rate": 5.897584992051874e-07, "loss": 0.0027, "step": 58096 }, { "epoch": 17.88, "learning_rate": 5.895899264981364e-07, "loss": 0.0021, "step": 58097 }, { "epoch": 17.88, "learning_rate": 5.894213771545943e-07, "loss": 0.0036, "step": 58098 }, { "epoch": 17.88, "learning_rate": 5.892528511749817e-07, "loss": 0.0072, "step": 58099 }, { "epoch": 17.88, "learning_rate": 5.890843485597175e-07, "loss": 0.0027, "step": 58100 }, { "epoch": 17.88, "learning_rate": 5.8891586930922e-07, "loss": 0.0049, "step": 58101 }, { "epoch": 17.88, "learning_rate": 5.887474134239079e-07, "loss": 0.0033, "step": 58102 }, { "epoch": 17.88, "learning_rate": 5.885789809041986e-07, "loss": 0.004, "step": 58103 }, { "epoch": 17.88, "learning_rate": 5.884105717505107e-07, "loss": 0.0017, "step": 58104 }, { "epoch": 17.88, "learning_rate": 5.882421859632614e-07, "loss": 0.0026, "step": 58105 }, { "epoch": 17.88, "learning_rate": 5.880738235428696e-07, "loss": 0.0043, "step": 58106 }, { "epoch": 17.88, "learning_rate": 5.879054844897536e-07, "loss": 0.0044, "step": 58107 }, { "epoch": 17.88, "learning_rate": 5.877371688043298e-07, "loss": 0.0052, "step": 58108 }, { "epoch": 17.88, "learning_rate": 5.875688764870169e-07, "loss": 0.0037, "step": 58109 }, { "epoch": 17.88, "learning_rate": 5.874006075382343e-07, "loss": 0.0052, "step": 58110 }, { "epoch": 17.88, "learning_rate": 5.872323619583987e-07, "loss": 0.0054, "step": 58111 }, { "epoch": 17.88, "learning_rate": 5.870641397479271e-07, "loss": 0.0041, "step": 58112 }, { "epoch": 17.88, "learning_rate": 5.868959409072372e-07, "loss": 0.0051, "step": 58113 }, { "epoch": 17.88, "learning_rate": 5.867277654367454e-07, "loss": 0.0033, "step": 58114 }, { "epoch": 17.88, "learning_rate": 5.865596133368723e-07, "loss": 0.0094, "step": 58115 }, { "epoch": 17.88, "learning_rate": 5.863914846080343e-07, "loss": 0.0023, "step": 58116 }, { "epoch": 17.88, "learning_rate": 5.862233792506477e-07, "loss": 0.0055, "step": 58117 }, { "epoch": 17.88, "learning_rate": 5.860552972651334e-07, "loss": 0.0021, "step": 58118 }, { "epoch": 17.88, "learning_rate": 5.858872386519021e-07, "loss": 0.0041, "step": 58119 }, { "epoch": 17.88, "learning_rate": 5.857192034113757e-07, "loss": 0.0029, "step": 58120 }, { "epoch": 17.88, "learning_rate": 5.855511915439716e-07, "loss": 0.0048, "step": 58121 }, { "epoch": 17.88, "learning_rate": 5.85383203050105e-07, "loss": 0.0036, "step": 58122 }, { "epoch": 17.88, "learning_rate": 5.852152379301946e-07, "loss": 0.0034, "step": 58123 }, { "epoch": 17.88, "learning_rate": 5.850472961846554e-07, "loss": 0.0073, "step": 58124 }, { "epoch": 17.88, "learning_rate": 5.848793778139072e-07, "loss": 0.0048, "step": 58125 }, { "epoch": 17.88, "learning_rate": 5.847114828183642e-07, "loss": 0.0051, "step": 58126 }, { "epoch": 17.88, "learning_rate": 5.845436111984437e-07, "loss": 0.0027, "step": 58127 }, { "epoch": 17.89, "learning_rate": 5.843757629545632e-07, "loss": 0.0061, "step": 58128 }, { "epoch": 17.89, "learning_rate": 5.842079380871402e-07, "loss": 0.0049, "step": 58129 }, { "epoch": 17.89, "learning_rate": 5.840401365965897e-07, "loss": 0.0057, "step": 58130 }, { "epoch": 17.89, "learning_rate": 5.838723584833316e-07, "loss": 0.0047, "step": 58131 }, { "epoch": 17.89, "learning_rate": 5.837046037477778e-07, "loss": 0.0042, "step": 58132 }, { "epoch": 17.89, "learning_rate": 5.835368723903456e-07, "loss": 0.004, "step": 58133 }, { "epoch": 17.89, "learning_rate": 5.833691644114536e-07, "loss": 0.0077, "step": 58134 }, { "epoch": 17.89, "learning_rate": 5.832014798115182e-07, "loss": 0.0032, "step": 58135 }, { "epoch": 17.89, "learning_rate": 5.830338185909545e-07, "loss": 0.0033, "step": 58136 }, { "epoch": 17.89, "learning_rate": 5.82866180750179e-07, "loss": 0.0044, "step": 58137 }, { "epoch": 17.89, "learning_rate": 5.82698566289609e-07, "loss": 0.0057, "step": 58138 }, { "epoch": 17.89, "learning_rate": 5.825309752096586e-07, "loss": 0.0048, "step": 58139 }, { "epoch": 17.89, "learning_rate": 5.823634075107454e-07, "loss": 0.0052, "step": 58140 }, { "epoch": 17.89, "learning_rate": 5.821958631932845e-07, "loss": 0.0084, "step": 58141 }, { "epoch": 17.89, "learning_rate": 5.820283422576922e-07, "loss": 0.0047, "step": 58142 }, { "epoch": 17.89, "learning_rate": 5.818608447043828e-07, "loss": 0.007, "step": 58143 }, { "epoch": 17.89, "learning_rate": 5.81693370533778e-07, "loss": 0.004, "step": 58144 }, { "epoch": 17.89, "learning_rate": 5.815259197462874e-07, "loss": 0.0021, "step": 58145 }, { "epoch": 17.89, "learning_rate": 5.813584923423287e-07, "loss": 0.0065, "step": 58146 }, { "epoch": 17.89, "learning_rate": 5.811910883223171e-07, "loss": 0.0051, "step": 58147 }, { "epoch": 17.89, "learning_rate": 5.810237076866687e-07, "loss": 0.0036, "step": 58148 }, { "epoch": 17.89, "learning_rate": 5.80856350435799e-07, "loss": 0.0077, "step": 58149 }, { "epoch": 17.89, "learning_rate": 5.806890165701251e-07, "loss": 0.0032, "step": 58150 }, { "epoch": 17.89, "learning_rate": 5.805217060900603e-07, "loss": 0.0043, "step": 58151 }, { "epoch": 17.89, "learning_rate": 5.803544189960219e-07, "loss": 0.0037, "step": 58152 }, { "epoch": 17.89, "learning_rate": 5.80187155288422e-07, "loss": 0.0075, "step": 58153 }, { "epoch": 17.89, "learning_rate": 5.800199149676788e-07, "loss": 0.0063, "step": 58154 }, { "epoch": 17.89, "learning_rate": 5.798526980342068e-07, "loss": 0.0053, "step": 58155 }, { "epoch": 17.89, "learning_rate": 5.796855044884209e-07, "loss": 0.0047, "step": 58156 }, { "epoch": 17.89, "learning_rate": 5.795183343307354e-07, "loss": 0.0046, "step": 58157 }, { "epoch": 17.89, "learning_rate": 5.793511875615676e-07, "loss": 0.0047, "step": 58158 }, { "epoch": 17.89, "learning_rate": 5.791840641813295e-07, "loss": 0.0028, "step": 58159 }, { "epoch": 17.9, "learning_rate": 5.790169641904386e-07, "loss": 0.0062, "step": 58160 }, { "epoch": 17.9, "learning_rate": 5.78849887589309e-07, "loss": 0.0049, "step": 58161 }, { "epoch": 17.9, "learning_rate": 5.786828343783546e-07, "loss": 0.0072, "step": 58162 }, { "epoch": 17.9, "learning_rate": 5.785158045579897e-07, "loss": 0.0066, "step": 58163 }, { "epoch": 17.9, "learning_rate": 5.783487981286317e-07, "loss": 0.0029, "step": 58164 }, { "epoch": 17.9, "learning_rate": 5.781818150906948e-07, "loss": 0.0066, "step": 58165 }, { "epoch": 17.9, "learning_rate": 5.780148554445908e-07, "loss": 0.0028, "step": 58166 }, { "epoch": 17.9, "learning_rate": 5.77847919190736e-07, "loss": 0.0038, "step": 58167 }, { "epoch": 17.9, "learning_rate": 5.776810063295435e-07, "loss": 0.0044, "step": 58168 }, { "epoch": 17.9, "learning_rate": 5.775141168614306e-07, "loss": 0.0023, "step": 58169 }, { "epoch": 17.9, "learning_rate": 5.773472507868094e-07, "loss": 0.0054, "step": 58170 }, { "epoch": 17.9, "learning_rate": 5.77180408106095e-07, "loss": 0.0033, "step": 58171 }, { "epoch": 17.9, "learning_rate": 5.770135888197004e-07, "loss": 0.0051, "step": 58172 }, { "epoch": 17.9, "learning_rate": 5.76846792928042e-07, "loss": 0.0048, "step": 58173 }, { "epoch": 17.9, "learning_rate": 5.766800204315315e-07, "loss": 0.0045, "step": 58174 }, { "epoch": 17.9, "learning_rate": 5.765132713305854e-07, "loss": 0.0042, "step": 58175 }, { "epoch": 17.9, "learning_rate": 5.763465456256156e-07, "loss": 0.0044, "step": 58176 }, { "epoch": 17.9, "learning_rate": 5.761798433170362e-07, "loss": 0.0016, "step": 58177 }, { "epoch": 17.9, "learning_rate": 5.760131644052625e-07, "loss": 0.0051, "step": 58178 }, { "epoch": 17.9, "learning_rate": 5.758465088907072e-07, "loss": 0.0062, "step": 58179 }, { "epoch": 17.9, "learning_rate": 5.756798767737837e-07, "loss": 0.0023, "step": 58180 }, { "epoch": 17.9, "learning_rate": 5.755132680549081e-07, "loss": 0.0025, "step": 58181 }, { "epoch": 17.9, "learning_rate": 5.753466827344911e-07, "loss": 0.0032, "step": 58182 }, { "epoch": 17.9, "learning_rate": 5.751801208129459e-07, "loss": 0.0051, "step": 58183 }, { "epoch": 17.9, "learning_rate": 5.750135822906888e-07, "loss": 0.0073, "step": 58184 }, { "epoch": 17.9, "learning_rate": 5.748470671681328e-07, "loss": 0.0024, "step": 58185 }, { "epoch": 17.9, "learning_rate": 5.746805754456919e-07, "loss": 0.0041, "step": 58186 }, { "epoch": 17.9, "learning_rate": 5.745141071237759e-07, "loss": 0.0048, "step": 58187 }, { "epoch": 17.9, "learning_rate": 5.743476622028e-07, "loss": 0.004, "step": 58188 }, { "epoch": 17.9, "learning_rate": 5.741812406831793e-07, "loss": 0.0035, "step": 58189 }, { "epoch": 17.9, "learning_rate": 5.740148425653247e-07, "loss": 0.0033, "step": 58190 }, { "epoch": 17.9, "learning_rate": 5.738484678496503e-07, "loss": 0.0019, "step": 58191 }, { "epoch": 17.9, "learning_rate": 5.736821165365692e-07, "loss": 0.003, "step": 58192 }, { "epoch": 17.91, "learning_rate": 5.735157886264931e-07, "loss": 0.003, "step": 58193 }, { "epoch": 17.91, "learning_rate": 5.733494841198373e-07, "loss": 0.0082, "step": 58194 }, { "epoch": 17.91, "learning_rate": 5.731832030170126e-07, "loss": 0.0074, "step": 58195 }, { "epoch": 17.91, "learning_rate": 5.730169453184332e-07, "loss": 0.0072, "step": 58196 }, { "epoch": 17.91, "learning_rate": 5.728507110245107e-07, "loss": 0.0129, "step": 58197 }, { "epoch": 17.91, "learning_rate": 5.726845001356573e-07, "loss": 0.0033, "step": 58198 }, { "epoch": 17.91, "learning_rate": 5.725183126522893e-07, "loss": 0.0049, "step": 58199 }, { "epoch": 17.91, "learning_rate": 5.723521485748162e-07, "loss": 0.0047, "step": 58200 }, { "epoch": 17.91, "learning_rate": 5.721860079036501e-07, "loss": 0.003, "step": 58201 }, { "epoch": 17.91, "learning_rate": 5.72019890639205e-07, "loss": 0.0042, "step": 58202 }, { "epoch": 17.91, "learning_rate": 5.718537967818915e-07, "loss": 0.0038, "step": 58203 }, { "epoch": 17.91, "learning_rate": 5.716877263321241e-07, "loss": 0.003, "step": 58204 }, { "epoch": 17.91, "learning_rate": 5.715216792903144e-07, "loss": 0.0046, "step": 58205 }, { "epoch": 17.91, "learning_rate": 5.713556556568755e-07, "loss": 0.0036, "step": 58206 }, { "epoch": 17.91, "learning_rate": 5.711896554322194e-07, "loss": 0.0091, "step": 58207 }, { "epoch": 17.91, "learning_rate": 5.710236786167545e-07, "loss": 0.0044, "step": 58208 }, { "epoch": 17.91, "learning_rate": 5.708577252108971e-07, "loss": 0.0032, "step": 58209 }, { "epoch": 17.91, "learning_rate": 5.706917952150582e-07, "loss": 0.0019, "step": 58210 }, { "epoch": 17.91, "learning_rate": 5.705258886296494e-07, "loss": 0.0058, "step": 58211 }, { "epoch": 17.91, "learning_rate": 5.703600054550829e-07, "loss": 0.0122, "step": 58212 }, { "epoch": 17.91, "learning_rate": 5.701941456917703e-07, "loss": 0.0041, "step": 58213 }, { "epoch": 17.91, "learning_rate": 5.700283093401238e-07, "loss": 0.0025, "step": 58214 }, { "epoch": 17.91, "learning_rate": 5.698624964005539e-07, "loss": 0.0048, "step": 58215 }, { "epoch": 17.91, "learning_rate": 5.696967068734738e-07, "loss": 0.0025, "step": 58216 }, { "epoch": 17.91, "learning_rate": 5.695309407592941e-07, "loss": 0.0025, "step": 58217 }, { "epoch": 17.91, "learning_rate": 5.693651980584247e-07, "loss": 0.0053, "step": 58218 }, { "epoch": 17.91, "learning_rate": 5.691994787712818e-07, "loss": 0.0031, "step": 58219 }, { "epoch": 17.91, "learning_rate": 5.690337828982751e-07, "loss": 0.0034, "step": 58220 }, { "epoch": 17.91, "learning_rate": 5.688681104398131e-07, "loss": 0.006, "step": 58221 }, { "epoch": 17.91, "learning_rate": 5.687024613963099e-07, "loss": 0.0071, "step": 58222 }, { "epoch": 17.91, "learning_rate": 5.685368357681731e-07, "loss": 0.0043, "step": 58223 }, { "epoch": 17.91, "learning_rate": 5.683712335558189e-07, "loss": 0.0033, "step": 58224 }, { "epoch": 17.92, "learning_rate": 5.68205654759657e-07, "loss": 0.0035, "step": 58225 }, { "epoch": 17.92, "learning_rate": 5.680400993800972e-07, "loss": 0.0067, "step": 58226 }, { "epoch": 17.92, "learning_rate": 5.678745674175501e-07, "loss": 0.0055, "step": 58227 }, { "epoch": 17.92, "learning_rate": 5.677090588724288e-07, "loss": 0.004, "step": 58228 }, { "epoch": 17.92, "learning_rate": 5.675435737451429e-07, "loss": 0.0039, "step": 58229 }, { "epoch": 17.92, "learning_rate": 5.673781120361033e-07, "loss": 0.0024, "step": 58230 }, { "epoch": 17.92, "learning_rate": 5.672126737457207e-07, "loss": 0.0044, "step": 58231 }, { "epoch": 17.92, "learning_rate": 5.670472588744058e-07, "loss": 0.0042, "step": 58232 }, { "epoch": 17.92, "learning_rate": 5.668818674225684e-07, "loss": 0.0054, "step": 58233 }, { "epoch": 17.92, "learning_rate": 5.667164993906226e-07, "loss": 0.0026, "step": 58234 }, { "epoch": 17.92, "learning_rate": 5.665511547789759e-07, "loss": 0.0032, "step": 58235 }, { "epoch": 17.92, "learning_rate": 5.66385833588039e-07, "loss": 0.0052, "step": 58236 }, { "epoch": 17.92, "learning_rate": 5.662205358182226e-07, "loss": 0.0032, "step": 58237 }, { "epoch": 17.92, "learning_rate": 5.660552614699366e-07, "loss": 0.0036, "step": 58238 }, { "epoch": 17.92, "learning_rate": 5.658900105435938e-07, "loss": 0.0024, "step": 58239 }, { "epoch": 17.92, "learning_rate": 5.657247830396018e-07, "loss": 0.0064, "step": 58240 }, { "epoch": 17.92, "learning_rate": 5.655595789583735e-07, "loss": 0.0036, "step": 58241 }, { "epoch": 17.92, "learning_rate": 5.653943983003163e-07, "loss": 0.0033, "step": 58242 }, { "epoch": 17.92, "learning_rate": 5.6522924106584e-07, "loss": 0.0056, "step": 58243 }, { "epoch": 17.92, "learning_rate": 5.650641072553564e-07, "loss": 0.0029, "step": 58244 }, { "epoch": 17.92, "learning_rate": 5.648989968692753e-07, "loss": 0.0062, "step": 58245 }, { "epoch": 17.92, "learning_rate": 5.647339099080074e-07, "loss": 0.0033, "step": 58246 }, { "epoch": 17.92, "learning_rate": 5.6456884637196e-07, "loss": 0.0031, "step": 58247 }, { "epoch": 17.92, "learning_rate": 5.644038062615453e-07, "loss": 0.0029, "step": 58248 }, { "epoch": 17.92, "learning_rate": 5.642387895771728e-07, "loss": 0.0033, "step": 58249 }, { "epoch": 17.92, "learning_rate": 5.640737963192511e-07, "loss": 0.0055, "step": 58250 }, { "epoch": 17.92, "learning_rate": 5.639088264881898e-07, "loss": 0.0046, "step": 58251 }, { "epoch": 17.92, "learning_rate": 5.637438800843998e-07, "loss": 0.0026, "step": 58252 }, { "epoch": 17.92, "learning_rate": 5.635789571082872e-07, "loss": 0.0052, "step": 58253 }, { "epoch": 17.92, "learning_rate": 5.634140575602665e-07, "loss": 0.0041, "step": 58254 }, { "epoch": 17.92, "learning_rate": 5.632491814407459e-07, "loss": 0.0032, "step": 58255 }, { "epoch": 17.92, "learning_rate": 5.63084328750132e-07, "loss": 0.006, "step": 58256 }, { "epoch": 17.92, "learning_rate": 5.629194994888354e-07, "loss": 0.003, "step": 58257 }, { "epoch": 17.93, "learning_rate": 5.627546936572648e-07, "loss": 0.0019, "step": 58258 }, { "epoch": 17.93, "learning_rate": 5.625899112558308e-07, "loss": 0.0039, "step": 58259 }, { "epoch": 17.93, "learning_rate": 5.624251522849422e-07, "loss": 0.006, "step": 58260 }, { "epoch": 17.93, "learning_rate": 5.622604167450074e-07, "loss": 0.0023, "step": 58261 }, { "epoch": 17.93, "learning_rate": 5.620957046364372e-07, "loss": 0.0019, "step": 58262 }, { "epoch": 17.93, "learning_rate": 5.619310159596358e-07, "loss": 0.0048, "step": 58263 }, { "epoch": 17.93, "learning_rate": 5.617663507150173e-07, "loss": 0.0032, "step": 58264 }, { "epoch": 17.93, "learning_rate": 5.616017089029879e-07, "loss": 0.0029, "step": 58265 }, { "epoch": 17.93, "learning_rate": 5.614370905239574e-07, "loss": 0.0038, "step": 58266 }, { "epoch": 17.93, "learning_rate": 5.612724955783333e-07, "loss": 0.0073, "step": 58267 }, { "epoch": 17.93, "learning_rate": 5.611079240665252e-07, "loss": 0.0038, "step": 58268 }, { "epoch": 17.93, "learning_rate": 5.609433759889426e-07, "loss": 0.0053, "step": 58269 }, { "epoch": 17.93, "learning_rate": 5.607788513459922e-07, "loss": 0.0038, "step": 58270 }, { "epoch": 17.93, "learning_rate": 5.606143501380823e-07, "loss": 0.0035, "step": 58271 }, { "epoch": 17.93, "learning_rate": 5.604498723656237e-07, "loss": 0.0044, "step": 58272 }, { "epoch": 17.93, "learning_rate": 5.602854180290207e-07, "loss": 0.0038, "step": 58273 }, { "epoch": 17.93, "learning_rate": 5.601209871286862e-07, "loss": 0.0038, "step": 58274 }, { "epoch": 17.93, "learning_rate": 5.599565796650275e-07, "loss": 0.0079, "step": 58275 }, { "epoch": 17.93, "learning_rate": 5.5979219563845e-07, "loss": 0.004, "step": 58276 }, { "epoch": 17.93, "learning_rate": 5.596278350493634e-07, "loss": 0.0052, "step": 58277 }, { "epoch": 17.93, "learning_rate": 5.59463497898175e-07, "loss": 0.0157, "step": 58278 }, { "epoch": 17.93, "learning_rate": 5.592991841852946e-07, "loss": 0.0051, "step": 58279 }, { "epoch": 17.93, "learning_rate": 5.591348939111296e-07, "loss": 0.0034, "step": 58280 }, { "epoch": 17.93, "learning_rate": 5.589706270760875e-07, "loss": 0.0039, "step": 58281 }, { "epoch": 17.93, "learning_rate": 5.588063836805757e-07, "loss": 0.0046, "step": 58282 }, { "epoch": 17.93, "learning_rate": 5.586421637250028e-07, "loss": 0.0052, "step": 58283 }, { "epoch": 17.93, "learning_rate": 5.584779672097762e-07, "loss": 0.0058, "step": 58284 }, { "epoch": 17.93, "learning_rate": 5.583137941353034e-07, "loss": 0.005, "step": 58285 }, { "epoch": 17.93, "learning_rate": 5.581496445019918e-07, "loss": 0.0043, "step": 58286 }, { "epoch": 17.93, "learning_rate": 5.57985518310249e-07, "loss": 0.0062, "step": 58287 }, { "epoch": 17.93, "learning_rate": 5.578214155604811e-07, "loss": 0.005, "step": 58288 }, { "epoch": 17.93, "learning_rate": 5.576573362531001e-07, "loss": 0.004, "step": 58289 }, { "epoch": 17.94, "learning_rate": 5.574932803885092e-07, "loss": 0.0065, "step": 58290 }, { "epoch": 17.94, "learning_rate": 5.573292479671166e-07, "loss": 0.0023, "step": 58291 }, { "epoch": 17.94, "learning_rate": 5.571652389893301e-07, "loss": 0.004, "step": 58292 }, { "epoch": 17.94, "learning_rate": 5.570012534555558e-07, "loss": 0.0029, "step": 58293 }, { "epoch": 17.94, "learning_rate": 5.568372913662024e-07, "loss": 0.0044, "step": 58294 }, { "epoch": 17.94, "learning_rate": 5.566733527216761e-07, "loss": 0.0069, "step": 58295 }, { "epoch": 17.94, "learning_rate": 5.565094375223856e-07, "loss": 0.0019, "step": 58296 }, { "epoch": 17.94, "learning_rate": 5.563455457687339e-07, "loss": 0.0046, "step": 58297 }, { "epoch": 17.94, "learning_rate": 5.561816774611307e-07, "loss": 0.0039, "step": 58298 }, { "epoch": 17.94, "learning_rate": 5.560178325999821e-07, "loss": 0.002, "step": 58299 }, { "epoch": 17.94, "learning_rate": 5.558540111856959e-07, "loss": 0.0031, "step": 58300 }, { "epoch": 17.94, "learning_rate": 5.556902132186781e-07, "loss": 0.0027, "step": 58301 }, { "epoch": 17.94, "learning_rate": 5.555264386993353e-07, "loss": 0.0091, "step": 58302 }, { "epoch": 17.94, "learning_rate": 5.553626876280749e-07, "loss": 0.0044, "step": 58303 }, { "epoch": 17.94, "learning_rate": 5.55198960005302e-07, "loss": 0.0025, "step": 58304 }, { "epoch": 17.94, "learning_rate": 5.550352558314242e-07, "loss": 0.0034, "step": 58305 }, { "epoch": 17.94, "learning_rate": 5.548715751068479e-07, "loss": 0.005, "step": 58306 }, { "epoch": 17.94, "learning_rate": 5.547079178319781e-07, "loss": 0.0017, "step": 58307 }, { "epoch": 17.94, "learning_rate": 5.545442840072212e-07, "loss": 0.0041, "step": 58308 }, { "epoch": 17.94, "learning_rate": 5.543806736329859e-07, "loss": 0.0025, "step": 58309 }, { "epoch": 17.94, "learning_rate": 5.542170867096786e-07, "loss": 0.0037, "step": 58310 }, { "epoch": 17.94, "learning_rate": 5.54053523237702e-07, "loss": 0.005, "step": 58311 }, { "epoch": 17.94, "learning_rate": 5.538899832174649e-07, "loss": 0.0032, "step": 58312 }, { "epoch": 17.94, "learning_rate": 5.537264666493703e-07, "loss": 0.0073, "step": 58313 }, { "epoch": 17.94, "learning_rate": 5.535629735338266e-07, "loss": 0.0035, "step": 58314 }, { "epoch": 17.94, "learning_rate": 5.533995038712403e-07, "loss": 0.0026, "step": 58315 }, { "epoch": 17.94, "learning_rate": 5.532360576620166e-07, "loss": 0.0035, "step": 58316 }, { "epoch": 17.94, "learning_rate": 5.530726349065618e-07, "loss": 0.0031, "step": 58317 }, { "epoch": 17.94, "learning_rate": 5.52909235605279e-07, "loss": 0.0031, "step": 58318 }, { "epoch": 17.94, "learning_rate": 5.527458597585766e-07, "loss": 0.0048, "step": 58319 }, { "epoch": 17.94, "learning_rate": 5.525825073668589e-07, "loss": 0.0047, "step": 58320 }, { "epoch": 17.94, "learning_rate": 5.524191784305321e-07, "loss": 0.0038, "step": 58321 }, { "epoch": 17.94, "learning_rate": 5.522558729500016e-07, "loss": 0.0074, "step": 58322 }, { "epoch": 17.95, "learning_rate": 5.520925909256724e-07, "loss": 0.0036, "step": 58323 }, { "epoch": 17.95, "learning_rate": 5.519293323579511e-07, "loss": 0.0026, "step": 58324 }, { "epoch": 17.95, "learning_rate": 5.517660972472416e-07, "loss": 0.0035, "step": 58325 }, { "epoch": 17.95, "learning_rate": 5.516028855939493e-07, "loss": 0.0028, "step": 58326 }, { "epoch": 17.95, "learning_rate": 5.514396973984803e-07, "loss": 0.0038, "step": 58327 }, { "epoch": 17.95, "learning_rate": 5.51276532661238e-07, "loss": 0.0041, "step": 58328 }, { "epoch": 17.95, "learning_rate": 5.511133913826294e-07, "loss": 0.0044, "step": 58329 }, { "epoch": 17.95, "learning_rate": 5.509502735630601e-07, "loss": 0.005, "step": 58330 }, { "epoch": 17.95, "learning_rate": 5.50787179202934e-07, "loss": 0.0035, "step": 58331 }, { "epoch": 17.95, "learning_rate": 5.506241083026553e-07, "loss": 0.0044, "step": 58332 }, { "epoch": 17.95, "learning_rate": 5.504610608626281e-07, "loss": 0.0031, "step": 58333 }, { "epoch": 17.95, "learning_rate": 5.5029803688326e-07, "loss": 0.0042, "step": 58334 }, { "epoch": 17.95, "learning_rate": 5.501350363649537e-07, "loss": 0.0055, "step": 58335 }, { "epoch": 17.95, "learning_rate": 5.499720593081148e-07, "loss": 0.0026, "step": 58336 }, { "epoch": 17.95, "learning_rate": 5.498091057131483e-07, "loss": 0.0047, "step": 58337 }, { "epoch": 17.95, "learning_rate": 5.496461755804572e-07, "loss": 0.0033, "step": 58338 }, { "epoch": 17.95, "learning_rate": 5.494832689104479e-07, "loss": 0.005, "step": 58339 }, { "epoch": 17.95, "learning_rate": 5.493203857035223e-07, "loss": 0.0039, "step": 58340 }, { "epoch": 17.95, "learning_rate": 5.491575259600879e-07, "loss": 0.0094, "step": 58341 }, { "epoch": 17.95, "learning_rate": 5.489946896805465e-07, "loss": 0.0037, "step": 58342 }, { "epoch": 17.95, "learning_rate": 5.488318768653022e-07, "loss": 0.0043, "step": 58343 }, { "epoch": 17.95, "learning_rate": 5.486690875147638e-07, "loss": 0.0046, "step": 58344 }, { "epoch": 17.95, "learning_rate": 5.485063216293296e-07, "loss": 0.0024, "step": 58345 }, { "epoch": 17.95, "learning_rate": 5.483435792094072e-07, "loss": 0.0028, "step": 58346 }, { "epoch": 17.95, "learning_rate": 5.481808602553995e-07, "loss": 0.0048, "step": 58347 }, { "epoch": 17.95, "learning_rate": 5.480181647677085e-07, "loss": 0.0049, "step": 58348 }, { "epoch": 17.95, "learning_rate": 5.478554927467417e-07, "loss": 0.0048, "step": 58349 }, { "epoch": 17.95, "learning_rate": 5.47692844192902e-07, "loss": 0.0032, "step": 58350 }, { "epoch": 17.95, "learning_rate": 5.475302191065923e-07, "loss": 0.0032, "step": 58351 }, { "epoch": 17.95, "learning_rate": 5.473676174882181e-07, "loss": 0.0029, "step": 58352 }, { "epoch": 17.95, "learning_rate": 5.47205039338179e-07, "loss": 0.0054, "step": 58353 }, { "epoch": 17.95, "learning_rate": 5.470424846568833e-07, "loss": 0.0053, "step": 58354 }, { "epoch": 17.96, "learning_rate": 5.468799534447322e-07, "loss": 0.0045, "step": 58355 }, { "epoch": 17.96, "learning_rate": 5.467174457021296e-07, "loss": 0.005, "step": 58356 }, { "epoch": 17.96, "learning_rate": 5.465549614294796e-07, "loss": 0.0033, "step": 58357 }, { "epoch": 17.96, "learning_rate": 5.463925006271853e-07, "loss": 0.0046, "step": 58358 }, { "epoch": 17.96, "learning_rate": 5.462300632956485e-07, "loss": 0.0076, "step": 58359 }, { "epoch": 17.96, "learning_rate": 5.460676494352746e-07, "loss": 0.0016, "step": 58360 }, { "epoch": 17.96, "learning_rate": 5.459052590464664e-07, "loss": 0.0031, "step": 58361 }, { "epoch": 17.96, "learning_rate": 5.45742892129626e-07, "loss": 0.0037, "step": 58362 }, { "epoch": 17.96, "learning_rate": 5.455805486851562e-07, "loss": 0.0036, "step": 58363 }, { "epoch": 17.96, "learning_rate": 5.454182287134624e-07, "loss": 0.0039, "step": 58364 }, { "epoch": 17.96, "learning_rate": 5.452559322149475e-07, "loss": 0.0038, "step": 58365 }, { "epoch": 17.96, "learning_rate": 5.450936591900124e-07, "loss": 0.0022, "step": 58366 }, { "epoch": 17.96, "learning_rate": 5.449314096390601e-07, "loss": 0.0042, "step": 58367 }, { "epoch": 17.96, "learning_rate": 5.447691835624924e-07, "loss": 0.003, "step": 58368 }, { "epoch": 17.96, "learning_rate": 5.446069809607157e-07, "loss": 0.0044, "step": 58369 }, { "epoch": 17.96, "learning_rate": 5.444448018341309e-07, "loss": 0.0046, "step": 58370 }, { "epoch": 17.96, "learning_rate": 5.442826461831408e-07, "loss": 0.0026, "step": 58371 }, { "epoch": 17.96, "learning_rate": 5.441205140081474e-07, "loss": 0.0051, "step": 58372 }, { "epoch": 17.96, "learning_rate": 5.439584053095525e-07, "loss": 0.0041, "step": 58373 }, { "epoch": 17.96, "learning_rate": 5.437963200877605e-07, "loss": 0.0026, "step": 58374 }, { "epoch": 17.96, "learning_rate": 5.436342583431731e-07, "loss": 0.0044, "step": 58375 }, { "epoch": 17.96, "learning_rate": 5.434722200761922e-07, "loss": 0.0042, "step": 58376 }, { "epoch": 17.96, "learning_rate": 5.433102052872207e-07, "loss": 0.0041, "step": 58377 }, { "epoch": 17.96, "learning_rate": 5.431482139766609e-07, "loss": 0.0088, "step": 58378 }, { "epoch": 17.96, "learning_rate": 5.429862461449132e-07, "loss": 0.0035, "step": 58379 }, { "epoch": 17.96, "learning_rate": 5.428243017923817e-07, "loss": 0.0042, "step": 58380 }, { "epoch": 17.96, "learning_rate": 5.426623809194687e-07, "loss": 0.0034, "step": 58381 }, { "epoch": 17.96, "learning_rate": 5.425004835265746e-07, "loss": 0.0025, "step": 58382 }, { "epoch": 17.96, "learning_rate": 5.423386096141003e-07, "loss": 0.0041, "step": 58383 }, { "epoch": 17.96, "learning_rate": 5.421767591824523e-07, "loss": 0.0046, "step": 58384 }, { "epoch": 17.96, "learning_rate": 5.420149322320289e-07, "loss": 0.0028, "step": 58385 }, { "epoch": 17.96, "learning_rate": 5.418531287632334e-07, "loss": 0.0057, "step": 58386 }, { "epoch": 17.96, "learning_rate": 5.416913487764663e-07, "loss": 0.0069, "step": 58387 }, { "epoch": 17.97, "learning_rate": 5.415295922721286e-07, "loss": 0.0044, "step": 58388 }, { "epoch": 17.97, "learning_rate": 5.413678592506233e-07, "loss": 0.0056, "step": 58389 }, { "epoch": 17.97, "learning_rate": 5.412061497123522e-07, "loss": 0.0026, "step": 58390 }, { "epoch": 17.97, "learning_rate": 5.410444636577161e-07, "loss": 0.0031, "step": 58391 }, { "epoch": 17.97, "learning_rate": 5.40882801087117e-07, "loss": 0.0032, "step": 58392 }, { "epoch": 17.97, "learning_rate": 5.407211620009545e-07, "loss": 0.0047, "step": 58393 }, { "epoch": 17.97, "learning_rate": 5.405595463996328e-07, "loss": 0.0032, "step": 58394 }, { "epoch": 17.97, "learning_rate": 5.403979542835502e-07, "loss": 0.0041, "step": 58395 }, { "epoch": 17.97, "learning_rate": 5.402363856531101e-07, "loss": 0.0028, "step": 58396 }, { "epoch": 17.97, "learning_rate": 5.400748405087131e-07, "loss": 0.0031, "step": 58397 }, { "epoch": 17.97, "learning_rate": 5.399133188507577e-07, "loss": 0.0061, "step": 58398 }, { "epoch": 17.97, "learning_rate": 5.397518206796504e-07, "loss": 0.0017, "step": 58399 }, { "epoch": 17.97, "learning_rate": 5.395903459957874e-07, "loss": 0.0039, "step": 58400 }, { "epoch": 17.97, "learning_rate": 5.394288947995707e-07, "loss": 0.0038, "step": 58401 }, { "epoch": 17.97, "learning_rate": 5.392674670914022e-07, "loss": 0.0053, "step": 58402 }, { "epoch": 17.97, "learning_rate": 5.391060628716805e-07, "loss": 0.0039, "step": 58403 }, { "epoch": 17.97, "learning_rate": 5.389446821408095e-07, "loss": 0.0047, "step": 58404 }, { "epoch": 17.97, "learning_rate": 5.38783324899188e-07, "loss": 0.004, "step": 58405 }, { "epoch": 17.97, "learning_rate": 5.386219911472168e-07, "loss": 0.0038, "step": 58406 }, { "epoch": 17.97, "learning_rate": 5.384606808852977e-07, "loss": 0.0067, "step": 58407 }, { "epoch": 17.97, "learning_rate": 5.382993941138271e-07, "loss": 0.0038, "step": 58408 }, { "epoch": 17.97, "learning_rate": 5.381381308332101e-07, "loss": 0.0059, "step": 58409 }, { "epoch": 17.97, "learning_rate": 5.379768910438454e-07, "loss": 0.0036, "step": 58410 }, { "epoch": 17.97, "learning_rate": 5.378156747461328e-07, "loss": 0.0051, "step": 58411 }, { "epoch": 17.97, "learning_rate": 5.376544819404728e-07, "loss": 0.0027, "step": 58412 }, { "epoch": 17.97, "learning_rate": 5.374933126272664e-07, "loss": 0.003, "step": 58413 }, { "epoch": 17.97, "learning_rate": 5.37332166806912e-07, "loss": 0.0038, "step": 58414 }, { "epoch": 17.97, "learning_rate": 5.371710444798117e-07, "loss": 0.0024, "step": 58415 }, { "epoch": 17.97, "learning_rate": 5.370099456463651e-07, "loss": 0.006, "step": 58416 }, { "epoch": 17.97, "learning_rate": 5.368488703069708e-07, "loss": 0.0058, "step": 58417 }, { "epoch": 17.97, "learning_rate": 5.366878184620284e-07, "loss": 0.0029, "step": 58418 }, { "epoch": 17.97, "learning_rate": 5.365267901119398e-07, "loss": 0.0037, "step": 58419 }, { "epoch": 17.98, "learning_rate": 5.363657852571058e-07, "loss": 0.0032, "step": 58420 }, { "epoch": 17.98, "learning_rate": 5.362048038979228e-07, "loss": 0.0072, "step": 58421 }, { "epoch": 17.98, "learning_rate": 5.360438460347916e-07, "loss": 0.0043, "step": 58422 }, { "epoch": 17.98, "learning_rate": 5.358829116681119e-07, "loss": 0.0036, "step": 58423 }, { "epoch": 17.98, "learning_rate": 5.357220007982833e-07, "loss": 0.0047, "step": 58424 }, { "epoch": 17.98, "learning_rate": 5.355611134257066e-07, "loss": 0.0034, "step": 58425 }, { "epoch": 17.98, "learning_rate": 5.354002495507793e-07, "loss": 0.0033, "step": 58426 }, { "epoch": 17.98, "learning_rate": 5.352394091739022e-07, "loss": 0.0044, "step": 58427 }, { "epoch": 17.98, "learning_rate": 5.350785922954738e-07, "loss": 0.004, "step": 58428 }, { "epoch": 17.98, "learning_rate": 5.349177989158938e-07, "loss": 0.0026, "step": 58429 }, { "epoch": 17.98, "learning_rate": 5.347570290355608e-07, "loss": 0.005, "step": 58430 }, { "epoch": 17.98, "learning_rate": 5.345962826548745e-07, "loss": 0.0079, "step": 58431 }, { "epoch": 17.98, "learning_rate": 5.344355597742345e-07, "loss": 0.0051, "step": 58432 }, { "epoch": 17.98, "learning_rate": 5.342748603940384e-07, "loss": 0.0052, "step": 58433 }, { "epoch": 17.98, "learning_rate": 5.341141845146858e-07, "loss": 0.0035, "step": 58434 }, { "epoch": 17.98, "learning_rate": 5.339535321365763e-07, "loss": 0.0031, "step": 58435 }, { "epoch": 17.98, "learning_rate": 5.337929032601085e-07, "loss": 0.0092, "step": 58436 }, { "epoch": 17.98, "learning_rate": 5.3363229788568e-07, "loss": 0.003, "step": 58437 }, { "epoch": 17.98, "learning_rate": 5.334717160136893e-07, "loss": 0.0028, "step": 58438 }, { "epoch": 17.98, "learning_rate": 5.333111576445383e-07, "loss": 0.0125, "step": 58439 }, { "epoch": 17.98, "learning_rate": 5.331506227786231e-07, "loss": 0.0041, "step": 58440 }, { "epoch": 17.98, "learning_rate": 5.329901114163438e-07, "loss": 0.0044, "step": 58441 }, { "epoch": 17.98, "learning_rate": 5.328296235580966e-07, "loss": 0.0038, "step": 58442 }, { "epoch": 17.98, "learning_rate": 5.326691592042798e-07, "loss": 0.006, "step": 58443 }, { "epoch": 17.98, "learning_rate": 5.325087183552946e-07, "loss": 0.0059, "step": 58444 }, { "epoch": 17.98, "learning_rate": 5.323483010115382e-07, "loss": 0.0057, "step": 58445 }, { "epoch": 17.98, "learning_rate": 5.321879071734082e-07, "loss": 0.0032, "step": 58446 }, { "epoch": 17.98, "learning_rate": 5.320275368413031e-07, "loss": 0.0041, "step": 58447 }, { "epoch": 17.98, "learning_rate": 5.318671900156213e-07, "loss": 0.0043, "step": 58448 }, { "epoch": 17.98, "learning_rate": 5.317068666967607e-07, "loss": 0.0026, "step": 58449 }, { "epoch": 17.98, "learning_rate": 5.315465668851194e-07, "loss": 0.0032, "step": 58450 }, { "epoch": 17.98, "learning_rate": 5.313862905810951e-07, "loss": 0.0029, "step": 58451 }, { "epoch": 17.98, "learning_rate": 5.312260377850865e-07, "loss": 0.0025, "step": 58452 }, { "epoch": 17.99, "learning_rate": 5.310658084974896e-07, "loss": 0.004, "step": 58453 }, { "epoch": 17.99, "learning_rate": 5.309056027187065e-07, "loss": 0.0058, "step": 58454 }, { "epoch": 17.99, "learning_rate": 5.307454204491303e-07, "loss": 0.0033, "step": 58455 }, { "epoch": 17.99, "learning_rate": 5.305852616891594e-07, "loss": 0.0048, "step": 58456 }, { "epoch": 17.99, "learning_rate": 5.304251264391935e-07, "loss": 0.0059, "step": 58457 }, { "epoch": 17.99, "learning_rate": 5.30265014699628e-07, "loss": 0.005, "step": 58458 }, { "epoch": 17.99, "learning_rate": 5.301049264708613e-07, "loss": 0.0038, "step": 58459 }, { "epoch": 17.99, "learning_rate": 5.299448617532932e-07, "loss": 0.0039, "step": 58460 }, { "epoch": 17.99, "learning_rate": 5.297848205473177e-07, "loss": 0.0045, "step": 58461 }, { "epoch": 17.99, "learning_rate": 5.296248028533346e-07, "loss": 0.0039, "step": 58462 }, { "epoch": 17.99, "learning_rate": 5.29464808671738e-07, "loss": 0.0057, "step": 58463 }, { "epoch": 17.99, "learning_rate": 5.293048380029275e-07, "loss": 0.0047, "step": 58464 }, { "epoch": 17.99, "learning_rate": 5.291448908473007e-07, "loss": 0.0045, "step": 58465 }, { "epoch": 17.99, "learning_rate": 5.289849672052538e-07, "loss": 0.0031, "step": 58466 }, { "epoch": 17.99, "learning_rate": 5.288250670771832e-07, "loss": 0.0049, "step": 58467 }, { "epoch": 17.99, "learning_rate": 5.286651904634876e-07, "loss": 0.0023, "step": 58468 }, { "epoch": 17.99, "learning_rate": 5.285053373645621e-07, "loss": 0.0022, "step": 58469 }, { "epoch": 17.99, "learning_rate": 5.283455077808052e-07, "loss": 0.0027, "step": 58470 }, { "epoch": 17.99, "learning_rate": 5.281857017126124e-07, "loss": 0.0039, "step": 58471 }, { "epoch": 17.99, "learning_rate": 5.280259191603809e-07, "loss": 0.0046, "step": 58472 }, { "epoch": 17.99, "learning_rate": 5.278661601245061e-07, "loss": 0.0061, "step": 58473 }, { "epoch": 17.99, "learning_rate": 5.277064246053875e-07, "loss": 0.0034, "step": 58474 }, { "epoch": 17.99, "learning_rate": 5.275467126034217e-07, "loss": 0.0096, "step": 58475 }, { "epoch": 17.99, "learning_rate": 5.273870241190027e-07, "loss": 0.0038, "step": 58476 }, { "epoch": 17.99, "learning_rate": 5.272273591525279e-07, "loss": 0.0037, "step": 58477 }, { "epoch": 17.99, "learning_rate": 5.270677177043914e-07, "loss": 0.0039, "step": 58478 }, { "epoch": 17.99, "learning_rate": 5.269080997749942e-07, "loss": 0.005, "step": 58479 }, { "epoch": 17.99, "learning_rate": 5.267485053647303e-07, "loss": 0.0044, "step": 58480 }, { "epoch": 17.99, "learning_rate": 5.265889344739961e-07, "loss": 0.0036, "step": 58481 }, { "epoch": 17.99, "learning_rate": 5.264293871031866e-07, "loss": 0.0032, "step": 58482 }, { "epoch": 17.99, "learning_rate": 5.262698632527008e-07, "loss": 0.0036, "step": 58483 }, { "epoch": 17.99, "learning_rate": 5.261103629229314e-07, "loss": 0.0042, "step": 58484 }, { "epoch": 18.0, "learning_rate": 5.259508861142759e-07, "loss": 0.0058, "step": 58485 }, { "epoch": 18.0, "learning_rate": 5.257914328271307e-07, "loss": 0.0012, "step": 58486 }, { "epoch": 18.0, "learning_rate": 5.256320030618912e-07, "loss": 0.004, "step": 58487 }, { "epoch": 18.0, "learning_rate": 5.254725968189512e-07, "loss": 0.0037, "step": 58488 }, { "epoch": 18.0, "learning_rate": 5.253132140987116e-07, "loss": 0.004, "step": 58489 }, { "epoch": 18.0, "learning_rate": 5.251538549015645e-07, "loss": 0.004, "step": 58490 }, { "epoch": 18.0, "learning_rate": 5.249945192279049e-07, "loss": 0.0045, "step": 58491 }, { "epoch": 18.0, "learning_rate": 5.248352070781304e-07, "loss": 0.0028, "step": 58492 }, { "epoch": 18.0, "learning_rate": 5.246759184526328e-07, "loss": 0.0075, "step": 58493 }, { "epoch": 18.0, "learning_rate": 5.245166533518131e-07, "loss": 0.0029, "step": 58494 }, { "epoch": 18.0, "learning_rate": 5.243574117760642e-07, "loss": 0.0076, "step": 58495 }, { "epoch": 18.0, "learning_rate": 5.241981937257823e-07, "loss": 0.0047, "step": 58496 }, { "epoch": 18.0, "learning_rate": 5.240389992013606e-07, "loss": 0.0019, "step": 58497 }, { "epoch": 18.0, "learning_rate": 5.238798282031932e-07, "loss": 0.0038, "step": 58498 }, { "epoch": 18.0, "learning_rate": 5.237206807316797e-07, "loss": 0.0041, "step": 58499 }, { "epoch": 18.0, "learning_rate": 5.235615567872121e-07, "loss": 0.0036, "step": 58500 }, { "epoch": 18.0, "learning_rate": 5.234024563701867e-07, "loss": 0.0055, "step": 58501 }, { "epoch": 18.0, "learning_rate": 5.232433794809988e-07, "loss": 0.007, "step": 58502 }, { "epoch": 18.0, "learning_rate": 5.230843261200425e-07, "loss": 0.0027, "step": 58503 }, { "epoch": 18.0, "learning_rate": 5.229252962877129e-07, "loss": 0.0025, "step": 58504 }, { "epoch": 18.0, "learning_rate": 5.227662899844043e-07, "loss": 0.0053, "step": 58505 }, { "epoch": 18.0, "learning_rate": 5.226073072105131e-07, "loss": 0.0023, "step": 58506 }, { "epoch": 18.0, "learning_rate": 5.224483479664322e-07, "loss": 0.0043, "step": 58507 }, { "epoch": 18.0, "learning_rate": 5.222894122525557e-07, "loss": 0.0101, "step": 58508 }, { "epoch": 18.0, "learning_rate": 5.221305000692811e-07, "loss": 0.0028, "step": 58509 }, { "epoch": 18.0, "learning_rate": 5.219716114170026e-07, "loss": 0.0014, "step": 58510 }, { "epoch": 18.0, "learning_rate": 5.218127462961109e-07, "loss": 0.0035, "step": 58511 }, { "epoch": 18.0, "learning_rate": 5.216539047070046e-07, "loss": 0.0048, "step": 58512 }, { "epoch": 18.0, "learning_rate": 5.214950866500734e-07, "loss": 0.004, "step": 58513 }, { "epoch": 18.0, "learning_rate": 5.21336292125717e-07, "loss": 0.0031, "step": 58514 }, { "epoch": 18.0, "learning_rate": 5.211775211343273e-07, "loss": 0.0017, "step": 58515 }, { "epoch": 18.0, "learning_rate": 5.210187736762984e-07, "loss": 0.0045, "step": 58516 }, { "epoch": 18.0, "learning_rate": 5.208600497520255e-07, "loss": 0.0045, "step": 58517 }, { "epoch": 18.01, "learning_rate": 5.207013493618984e-07, "loss": 0.0053, "step": 58518 }, { "epoch": 18.01, "learning_rate": 5.205426725063167e-07, "loss": 0.0024, "step": 58519 }, { "epoch": 18.01, "learning_rate": 5.203840191856713e-07, "loss": 0.0038, "step": 58520 }, { "epoch": 18.01, "learning_rate": 5.202253894003573e-07, "loss": 0.0044, "step": 58521 }, { "epoch": 18.01, "learning_rate": 5.200667831507655e-07, "loss": 0.0016, "step": 58522 }, { "epoch": 18.01, "learning_rate": 5.199082004372958e-07, "loss": 0.0028, "step": 58523 }, { "epoch": 18.01, "learning_rate": 5.197496412603365e-07, "loss": 0.0044, "step": 58524 }, { "epoch": 18.01, "learning_rate": 5.19591105620284e-07, "loss": 0.0027, "step": 58525 }, { "epoch": 18.01, "learning_rate": 5.194325935175304e-07, "loss": 0.0044, "step": 58526 }, { "epoch": 18.01, "learning_rate": 5.192741049524674e-07, "loss": 0.0021, "step": 58527 }, { "epoch": 18.01, "learning_rate": 5.191156399254937e-07, "loss": 0.0042, "step": 58528 }, { "epoch": 18.01, "learning_rate": 5.18957198436999e-07, "loss": 0.0039, "step": 58529 }, { "epoch": 18.01, "learning_rate": 5.187987804873785e-07, "loss": 0.002, "step": 58530 }, { "epoch": 18.01, "learning_rate": 5.18640386077025e-07, "loss": 0.002, "step": 58531 }, { "epoch": 18.01, "learning_rate": 5.184820152063286e-07, "loss": 0.0035, "step": 58532 }, { "epoch": 18.01, "learning_rate": 5.183236678756864e-07, "loss": 0.0034, "step": 58533 }, { "epoch": 18.01, "learning_rate": 5.181653440854917e-07, "loss": 0.0033, "step": 58534 }, { "epoch": 18.01, "learning_rate": 5.180070438361351e-07, "loss": 0.0047, "step": 58535 }, { "epoch": 18.01, "learning_rate": 5.178487671280108e-07, "loss": 0.0034, "step": 58536 }, { "epoch": 18.01, "learning_rate": 5.176905139615118e-07, "loss": 0.0029, "step": 58537 }, { "epoch": 18.01, "learning_rate": 5.175322843370312e-07, "loss": 0.0024, "step": 58538 }, { "epoch": 18.01, "learning_rate": 5.173740782549619e-07, "loss": 0.0041, "step": 58539 }, { "epoch": 18.01, "learning_rate": 5.172158957156947e-07, "loss": 0.003, "step": 58540 }, { "epoch": 18.01, "learning_rate": 5.170577367196262e-07, "loss": 0.0019, "step": 58541 }, { "epoch": 18.01, "learning_rate": 5.168996012671434e-07, "loss": 0.0038, "step": 58542 }, { "epoch": 18.01, "learning_rate": 5.167414893586453e-07, "loss": 0.0048, "step": 58543 }, { "epoch": 18.01, "learning_rate": 5.165834009945214e-07, "loss": 0.003, "step": 58544 }, { "epoch": 18.01, "learning_rate": 5.164253361751637e-07, "loss": 0.0035, "step": 58545 }, { "epoch": 18.01, "learning_rate": 5.162672949009651e-07, "loss": 0.003, "step": 58546 }, { "epoch": 18.01, "learning_rate": 5.161092771723164e-07, "loss": 0.0046, "step": 58547 }, { "epoch": 18.01, "learning_rate": 5.15951282989614e-07, "loss": 0.0039, "step": 58548 }, { "epoch": 18.01, "learning_rate": 5.157933123532466e-07, "loss": 0.0038, "step": 58549 }, { "epoch": 18.02, "learning_rate": 5.156353652636081e-07, "loss": 0.004, "step": 58550 }, { "epoch": 18.02, "learning_rate": 5.154774417210895e-07, "loss": 0.0048, "step": 58551 }, { "epoch": 18.02, "learning_rate": 5.153195417260848e-07, "loss": 0.0034, "step": 58552 }, { "epoch": 18.02, "learning_rate": 5.151616652789837e-07, "loss": 0.0032, "step": 58553 }, { "epoch": 18.02, "learning_rate": 5.150038123801793e-07, "loss": 0.0028, "step": 58554 }, { "epoch": 18.02, "learning_rate": 5.148459830300623e-07, "loss": 0.0042, "step": 58555 }, { "epoch": 18.02, "learning_rate": 5.146881772290268e-07, "loss": 0.0031, "step": 58556 }, { "epoch": 18.02, "learning_rate": 5.145303949774638e-07, "loss": 0.0028, "step": 58557 }, { "epoch": 18.02, "learning_rate": 5.143726362757639e-07, "loss": 0.0027, "step": 58558 }, { "epoch": 18.02, "learning_rate": 5.142149011243192e-07, "loss": 0.0033, "step": 58559 }, { "epoch": 18.02, "learning_rate": 5.140571895235225e-07, "loss": 0.0042, "step": 58560 }, { "epoch": 18.02, "learning_rate": 5.138995014737636e-07, "loss": 0.0037, "step": 58561 }, { "epoch": 18.02, "learning_rate": 5.137418369754343e-07, "loss": 0.0029, "step": 58562 }, { "epoch": 18.02, "learning_rate": 5.135841960289267e-07, "loss": 0.0046, "step": 58563 }, { "epoch": 18.02, "learning_rate": 5.134265786346337e-07, "loss": 0.0067, "step": 58564 }, { "epoch": 18.02, "learning_rate": 5.132689847929451e-07, "loss": 0.0026, "step": 58565 }, { "epoch": 18.02, "learning_rate": 5.131114145042515e-07, "loss": 0.0047, "step": 58566 }, { "epoch": 18.02, "learning_rate": 5.129538677689427e-07, "loss": 0.0028, "step": 58567 }, { "epoch": 18.02, "learning_rate": 5.12796344587414e-07, "loss": 0.0034, "step": 58568 }, { "epoch": 18.02, "learning_rate": 5.126388449600539e-07, "loss": 0.0032, "step": 58569 }, { "epoch": 18.02, "learning_rate": 5.124813688872532e-07, "loss": 0.0037, "step": 58570 }, { "epoch": 18.02, "learning_rate": 5.123239163694049e-07, "loss": 0.0027, "step": 58571 }, { "epoch": 18.02, "learning_rate": 5.121664874068977e-07, "loss": 0.0027, "step": 58572 }, { "epoch": 18.02, "learning_rate": 5.120090820001222e-07, "loss": 0.0029, "step": 58573 }, { "epoch": 18.02, "learning_rate": 5.118517001494716e-07, "loss": 0.0033, "step": 58574 }, { "epoch": 18.02, "learning_rate": 5.116943418553355e-07, "loss": 0.0033, "step": 58575 }, { "epoch": 18.02, "learning_rate": 5.115370071181036e-07, "loss": 0.008, "step": 58576 }, { "epoch": 18.02, "learning_rate": 5.113796959381667e-07, "loss": 0.0039, "step": 58577 }, { "epoch": 18.02, "learning_rate": 5.112224083159178e-07, "loss": 0.0024, "step": 58578 }, { "epoch": 18.02, "learning_rate": 5.110651442517445e-07, "loss": 0.0038, "step": 58579 }, { "epoch": 18.02, "learning_rate": 5.109079037460385e-07, "loss": 0.0031, "step": 58580 }, { "epoch": 18.02, "learning_rate": 5.107506867991896e-07, "loss": 0.0034, "step": 58581 }, { "epoch": 18.02, "learning_rate": 5.105934934115874e-07, "loss": 0.0044, "step": 58582 }, { "epoch": 18.03, "learning_rate": 5.104363235836251e-07, "loss": 0.005, "step": 58583 }, { "epoch": 18.03, "learning_rate": 5.102791773156912e-07, "loss": 0.0032, "step": 58584 }, { "epoch": 18.03, "learning_rate": 5.101220546081753e-07, "loss": 0.0035, "step": 58585 }, { "epoch": 18.03, "learning_rate": 5.099649554614683e-07, "loss": 0.0026, "step": 58586 }, { "epoch": 18.03, "learning_rate": 5.098078798759587e-07, "loss": 0.0061, "step": 58587 }, { "epoch": 18.03, "learning_rate": 5.096508278520385e-07, "loss": 0.0041, "step": 58588 }, { "epoch": 18.03, "learning_rate": 5.094937993900962e-07, "loss": 0.0038, "step": 58589 }, { "epoch": 18.03, "learning_rate": 5.093367944905237e-07, "loss": 0.0027, "step": 58590 }, { "epoch": 18.03, "learning_rate": 5.091798131537085e-07, "loss": 0.0031, "step": 58591 }, { "epoch": 18.03, "learning_rate": 5.090228553800403e-07, "loss": 0.0058, "step": 58592 }, { "epoch": 18.03, "learning_rate": 5.08865921169911e-07, "loss": 0.0022, "step": 58593 }, { "epoch": 18.03, "learning_rate": 5.08709010523708e-07, "loss": 0.0023, "step": 58594 }, { "epoch": 18.03, "learning_rate": 5.085521234418223e-07, "loss": 0.0023, "step": 58595 }, { "epoch": 18.03, "learning_rate": 5.083952599246423e-07, "loss": 0.0059, "step": 58596 }, { "epoch": 18.03, "learning_rate": 5.082384199725576e-07, "loss": 0.0023, "step": 58597 }, { "epoch": 18.03, "learning_rate": 5.080816035859581e-07, "loss": 0.0038, "step": 58598 }, { "epoch": 18.03, "learning_rate": 5.079248107652357e-07, "loss": 0.0035, "step": 58599 }, { "epoch": 18.03, "learning_rate": 5.077680415107745e-07, "loss": 0.0031, "step": 58600 }, { "epoch": 18.03, "learning_rate": 5.076112958229673e-07, "loss": 0.0045, "step": 58601 }, { "epoch": 18.03, "learning_rate": 5.074545737021996e-07, "loss": 0.0047, "step": 58602 }, { "epoch": 18.03, "learning_rate": 5.072978751488655e-07, "loss": 0.004, "step": 58603 }, { "epoch": 18.03, "learning_rate": 5.071412001633513e-07, "loss": 0.004, "step": 58604 }, { "epoch": 18.03, "learning_rate": 5.069845487460456e-07, "loss": 0.0051, "step": 58605 }, { "epoch": 18.03, "learning_rate": 5.068279208973392e-07, "loss": 0.0049, "step": 58606 }, { "epoch": 18.03, "learning_rate": 5.066713166176196e-07, "loss": 0.0029, "step": 58607 }, { "epoch": 18.03, "learning_rate": 5.065147359072753e-07, "loss": 0.003, "step": 58608 }, { "epoch": 18.03, "learning_rate": 5.06358178766696e-07, "loss": 0.0056, "step": 58609 }, { "epoch": 18.03, "learning_rate": 5.062016451962692e-07, "loss": 0.003, "step": 58610 }, { "epoch": 18.03, "learning_rate": 5.060451351963846e-07, "loss": 0.0035, "step": 58611 }, { "epoch": 18.03, "learning_rate": 5.058886487674308e-07, "loss": 0.0044, "step": 58612 }, { "epoch": 18.03, "learning_rate": 5.057321859097952e-07, "loss": 0.0039, "step": 58613 }, { "epoch": 18.03, "learning_rate": 5.055757466238676e-07, "loss": 0.0034, "step": 58614 }, { "epoch": 18.04, "learning_rate": 5.054193309100364e-07, "loss": 0.0053, "step": 58615 }, { "epoch": 18.04, "learning_rate": 5.05262938768688e-07, "loss": 0.0037, "step": 58616 }, { "epoch": 18.04, "learning_rate": 5.051065702002111e-07, "loss": 0.0015, "step": 58617 }, { "epoch": 18.04, "learning_rate": 5.049502252049965e-07, "loss": 0.0016, "step": 58618 }, { "epoch": 18.04, "learning_rate": 5.047939037834304e-07, "loss": 0.0045, "step": 58619 }, { "epoch": 18.04, "learning_rate": 5.046376059359037e-07, "loss": 0.003, "step": 58620 }, { "epoch": 18.04, "learning_rate": 5.044813316627994e-07, "loss": 0.0023, "step": 58621 }, { "epoch": 18.04, "learning_rate": 5.043250809645073e-07, "loss": 0.0045, "step": 58622 }, { "epoch": 18.04, "learning_rate": 5.041688538414169e-07, "loss": 0.0021, "step": 58623 }, { "epoch": 18.04, "learning_rate": 5.040126502939146e-07, "loss": 0.0025, "step": 58624 }, { "epoch": 18.04, "learning_rate": 5.038564703223903e-07, "loss": 0.0056, "step": 58625 }, { "epoch": 18.04, "learning_rate": 5.0370031392723e-07, "loss": 0.0031, "step": 58626 }, { "epoch": 18.04, "learning_rate": 5.035441811088204e-07, "loss": 0.0012, "step": 58627 }, { "epoch": 18.04, "learning_rate": 5.033880718675522e-07, "loss": 0.0036, "step": 58628 }, { "epoch": 18.04, "learning_rate": 5.032319862038104e-07, "loss": 0.0041, "step": 58629 }, { "epoch": 18.04, "learning_rate": 5.030759241179828e-07, "loss": 0.0039, "step": 58630 }, { "epoch": 18.04, "learning_rate": 5.029198856104578e-07, "loss": 0.004, "step": 58631 }, { "epoch": 18.04, "learning_rate": 5.027638706816218e-07, "loss": 0.0039, "step": 58632 }, { "epoch": 18.04, "learning_rate": 5.026078793318645e-07, "loss": 0.0036, "step": 58633 }, { "epoch": 18.04, "learning_rate": 5.0245191156157e-07, "loss": 0.0031, "step": 58634 }, { "epoch": 18.04, "learning_rate": 5.02295967371128e-07, "loss": 0.0034, "step": 58635 }, { "epoch": 18.04, "learning_rate": 5.021400467609238e-07, "loss": 0.0049, "step": 58636 }, { "epoch": 18.04, "learning_rate": 5.019841497313438e-07, "loss": 0.0046, "step": 58637 }, { "epoch": 18.04, "learning_rate": 5.018282762827786e-07, "loss": 0.0035, "step": 58638 }, { "epoch": 18.04, "learning_rate": 5.016724264156125e-07, "loss": 0.004, "step": 58639 }, { "epoch": 18.04, "learning_rate": 5.01516600130233e-07, "loss": 0.003, "step": 58640 }, { "epoch": 18.04, "learning_rate": 5.013607974270296e-07, "loss": 0.0054, "step": 58641 }, { "epoch": 18.04, "learning_rate": 5.012050183063832e-07, "loss": 0.0048, "step": 58642 }, { "epoch": 18.04, "learning_rate": 5.010492627686848e-07, "loss": 0.0043, "step": 58643 }, { "epoch": 18.04, "learning_rate": 5.008935308143203e-07, "loss": 0.0064, "step": 58644 }, { "epoch": 18.04, "learning_rate": 5.007378224436765e-07, "loss": 0.0016, "step": 58645 }, { "epoch": 18.04, "learning_rate": 5.005821376571407e-07, "loss": 0.0049, "step": 58646 }, { "epoch": 18.04, "learning_rate": 5.004264764550971e-07, "loss": 0.003, "step": 58647 }, { "epoch": 18.05, "learning_rate": 5.00270838837934e-07, "loss": 0.0028, "step": 58648 }, { "epoch": 18.05, "learning_rate": 5.001152248060382e-07, "loss": 0.0052, "step": 58649 }, { "epoch": 18.05, "learning_rate": 4.999596343597934e-07, "loss": 0.0053, "step": 58650 }, { "epoch": 18.05, "learning_rate": 4.998040674995897e-07, "loss": 0.0029, "step": 58651 }, { "epoch": 18.05, "learning_rate": 4.996485242258087e-07, "loss": 0.0027, "step": 58652 }, { "epoch": 18.05, "learning_rate": 4.994930045388414e-07, "loss": 0.0031, "step": 58653 }, { "epoch": 18.05, "learning_rate": 4.993375084390717e-07, "loss": 0.0047, "step": 58654 }, { "epoch": 18.05, "learning_rate": 4.99182035926885e-07, "loss": 0.0043, "step": 58655 }, { "epoch": 18.05, "learning_rate": 4.990265870026689e-07, "loss": 0.0036, "step": 58656 }, { "epoch": 18.05, "learning_rate": 4.988711616668062e-07, "loss": 0.003, "step": 58657 }, { "epoch": 18.05, "learning_rate": 4.987157599196868e-07, "loss": 0.0043, "step": 58658 }, { "epoch": 18.05, "learning_rate": 4.985603817616936e-07, "loss": 0.0046, "step": 58659 }, { "epoch": 18.05, "learning_rate": 4.984050271932139e-07, "loss": 0.0042, "step": 58660 }, { "epoch": 18.05, "learning_rate": 4.982496962146333e-07, "loss": 0.0045, "step": 58661 }, { "epoch": 18.05, "learning_rate": 4.980943888263378e-07, "loss": 0.0046, "step": 58662 }, { "epoch": 18.05, "learning_rate": 4.979391050287108e-07, "loss": 0.0039, "step": 58663 }, { "epoch": 18.05, "learning_rate": 4.977838448221406e-07, "loss": 0.0071, "step": 58664 }, { "epoch": 18.05, "learning_rate": 4.976286082070103e-07, "loss": 0.0048, "step": 58665 }, { "epoch": 18.05, "learning_rate": 4.974733951837074e-07, "loss": 0.003, "step": 58666 }, { "epoch": 18.05, "learning_rate": 4.973182057526149e-07, "loss": 0.0023, "step": 58667 }, { "epoch": 18.05, "learning_rate": 4.971630399141214e-07, "loss": 0.0026, "step": 58668 }, { "epoch": 18.05, "learning_rate": 4.970078976686088e-07, "loss": 0.0036, "step": 58669 }, { "epoch": 18.05, "learning_rate": 4.968527790164646e-07, "loss": 0.0037, "step": 58670 }, { "epoch": 18.05, "learning_rate": 4.966976839580728e-07, "loss": 0.0028, "step": 58671 }, { "epoch": 18.05, "learning_rate": 4.965426124938166e-07, "loss": 0.0035, "step": 58672 }, { "epoch": 18.05, "learning_rate": 4.963875646240846e-07, "loss": 0.0068, "step": 58673 }, { "epoch": 18.05, "learning_rate": 4.962325403492607e-07, "loss": 0.0019, "step": 58674 }, { "epoch": 18.05, "learning_rate": 4.960775396697293e-07, "loss": 0.004, "step": 58675 }, { "epoch": 18.05, "learning_rate": 4.959225625858754e-07, "loss": 0.0048, "step": 58676 }, { "epoch": 18.05, "learning_rate": 4.957676090980812e-07, "loss": 0.0037, "step": 58677 }, { "epoch": 18.05, "learning_rate": 4.95612679206735e-07, "loss": 0.0052, "step": 58678 }, { "epoch": 18.05, "learning_rate": 4.954577729122212e-07, "loss": 0.0113, "step": 58679 }, { "epoch": 18.06, "learning_rate": 4.953028902149226e-07, "loss": 0.0041, "step": 58680 }, { "epoch": 18.06, "learning_rate": 4.951480311152257e-07, "loss": 0.0069, "step": 58681 }, { "epoch": 18.06, "learning_rate": 4.949931956135123e-07, "loss": 0.0066, "step": 58682 }, { "epoch": 18.06, "learning_rate": 4.94838383710169e-07, "loss": 0.0074, "step": 58683 }, { "epoch": 18.06, "learning_rate": 4.946835954055795e-07, "loss": 0.0036, "step": 58684 }, { "epoch": 18.06, "learning_rate": 4.945288307001284e-07, "loss": 0.0026, "step": 58685 }, { "epoch": 18.06, "learning_rate": 4.943740895941984e-07, "loss": 0.0044, "step": 58686 }, { "epoch": 18.06, "learning_rate": 4.942193720881749e-07, "loss": 0.0034, "step": 58687 }, { "epoch": 18.06, "learning_rate": 4.940646781824432e-07, "loss": 0.0035, "step": 58688 }, { "epoch": 18.06, "learning_rate": 4.939100078773873e-07, "loss": 0.0061, "step": 58689 }, { "epoch": 18.06, "learning_rate": 4.937553611733881e-07, "loss": 0.0035, "step": 58690 }, { "epoch": 18.06, "learning_rate": 4.936007380708319e-07, "loss": 0.0045, "step": 58691 }, { "epoch": 18.06, "learning_rate": 4.934461385701006e-07, "loss": 0.0036, "step": 58692 }, { "epoch": 18.06, "learning_rate": 4.932915626715817e-07, "loss": 0.0049, "step": 58693 }, { "epoch": 18.06, "learning_rate": 4.931370103756572e-07, "loss": 0.0034, "step": 58694 }, { "epoch": 18.06, "learning_rate": 4.9298248168271e-07, "loss": 0.0029, "step": 58695 }, { "epoch": 18.06, "learning_rate": 4.928279765931254e-07, "loss": 0.0027, "step": 58696 }, { "epoch": 18.06, "learning_rate": 4.926734951072831e-07, "loss": 0.0044, "step": 58697 }, { "epoch": 18.06, "learning_rate": 4.925190372255717e-07, "loss": 0.0019, "step": 58698 }, { "epoch": 18.06, "learning_rate": 4.923646029483709e-07, "loss": 0.0044, "step": 58699 }, { "epoch": 18.06, "learning_rate": 4.922101922760669e-07, "loss": 0.0059, "step": 58700 }, { "epoch": 18.06, "learning_rate": 4.920558052090418e-07, "loss": 0.0021, "step": 58701 }, { "epoch": 18.06, "learning_rate": 4.919014417476775e-07, "loss": 0.003, "step": 58702 }, { "epoch": 18.06, "learning_rate": 4.917471018923603e-07, "loss": 0.0033, "step": 58703 }, { "epoch": 18.06, "learning_rate": 4.91592785643471e-07, "loss": 0.0032, "step": 58704 }, { "epoch": 18.06, "learning_rate": 4.914384930013927e-07, "loss": 0.0035, "step": 58705 }, { "epoch": 18.06, "learning_rate": 4.912842239665106e-07, "loss": 0.0076, "step": 58706 }, { "epoch": 18.06, "learning_rate": 4.911299785392032e-07, "loss": 0.003, "step": 58707 }, { "epoch": 18.06, "learning_rate": 4.909757567198592e-07, "loss": 0.003, "step": 58708 }, { "epoch": 18.06, "learning_rate": 4.908215585088583e-07, "loss": 0.0034, "step": 58709 }, { "epoch": 18.06, "learning_rate": 4.906673839065857e-07, "loss": 0.0025, "step": 58710 }, { "epoch": 18.06, "learning_rate": 4.90513232913421e-07, "loss": 0.0042, "step": 58711 }, { "epoch": 18.06, "learning_rate": 4.903591055297464e-07, "loss": 0.0064, "step": 58712 }, { "epoch": 18.07, "learning_rate": 4.902050017559479e-07, "loss": 0.0037, "step": 58713 }, { "epoch": 18.07, "learning_rate": 4.900509215924077e-07, "loss": 0.0056, "step": 58714 }, { "epoch": 18.07, "learning_rate": 4.898968650395053e-07, "loss": 0.0035, "step": 58715 }, { "epoch": 18.07, "learning_rate": 4.89742832097626e-07, "loss": 0.0073, "step": 58716 }, { "epoch": 18.07, "learning_rate": 4.89588822767152e-07, "loss": 0.003, "step": 58717 }, { "epoch": 18.07, "learning_rate": 4.894348370484648e-07, "loss": 0.0027, "step": 58718 }, { "epoch": 18.07, "learning_rate": 4.892808749419465e-07, "loss": 0.0054, "step": 58719 }, { "epoch": 18.07, "learning_rate": 4.8912693644798e-07, "loss": 0.0041, "step": 58720 }, { "epoch": 18.07, "learning_rate": 4.889730215669475e-07, "loss": 0.0043, "step": 58721 }, { "epoch": 18.07, "learning_rate": 4.888191302992285e-07, "loss": 0.0023, "step": 58722 }, { "epoch": 18.07, "learning_rate": 4.886652626452115e-07, "loss": 0.0027, "step": 58723 }, { "epoch": 18.07, "learning_rate": 4.885114186052719e-07, "loss": 0.0029, "step": 58724 }, { "epoch": 18.07, "learning_rate": 4.88357598179795e-07, "loss": 0.0036, "step": 58725 }, { "epoch": 18.07, "learning_rate": 4.882038013691625e-07, "loss": 0.0021, "step": 58726 }, { "epoch": 18.07, "learning_rate": 4.880500281737533e-07, "loss": 0.0036, "step": 58727 }, { "epoch": 18.07, "learning_rate": 4.878962785939533e-07, "loss": 0.003, "step": 58728 }, { "epoch": 18.07, "learning_rate": 4.877425526301427e-07, "loss": 0.0028, "step": 58729 }, { "epoch": 18.07, "learning_rate": 4.87588850282702e-07, "loss": 0.0042, "step": 58730 }, { "epoch": 18.07, "learning_rate": 4.874351715520154e-07, "loss": 0.003, "step": 58731 }, { "epoch": 18.07, "learning_rate": 4.872815164384603e-07, "loss": 0.0024, "step": 58732 }, { "epoch": 18.07, "learning_rate": 4.871278849424221e-07, "loss": 0.0054, "step": 58733 }, { "epoch": 18.07, "learning_rate": 4.869742770642816e-07, "loss": 0.0021, "step": 58734 }, { "epoch": 18.07, "learning_rate": 4.868206928044183e-07, "loss": 0.0038, "step": 58735 }, { "epoch": 18.07, "learning_rate": 4.866671321632144e-07, "loss": 0.0044, "step": 58736 }, { "epoch": 18.07, "learning_rate": 4.865135951410515e-07, "loss": 0.0033, "step": 58737 }, { "epoch": 18.07, "learning_rate": 4.863600817383107e-07, "loss": 0.0039, "step": 58738 }, { "epoch": 18.07, "learning_rate": 4.862065919553727e-07, "loss": 0.0048, "step": 58739 }, { "epoch": 18.07, "learning_rate": 4.860531257926193e-07, "loss": 0.0016, "step": 58740 }, { "epoch": 18.07, "learning_rate": 4.858996832504314e-07, "loss": 0.0063, "step": 58741 }, { "epoch": 18.07, "learning_rate": 4.857462643291877e-07, "loss": 0.0023, "step": 58742 }, { "epoch": 18.07, "learning_rate": 4.855928690292722e-07, "loss": 0.005, "step": 58743 }, { "epoch": 18.07, "learning_rate": 4.854394973510667e-07, "loss": 0.0027, "step": 58744 }, { "epoch": 18.08, "learning_rate": 4.852861492949479e-07, "loss": 0.0046, "step": 58745 }, { "epoch": 18.08, "learning_rate": 4.851328248612985e-07, "loss": 0.0045, "step": 58746 }, { "epoch": 18.08, "learning_rate": 4.849795240504984e-07, "loss": 0.0034, "step": 58747 }, { "epoch": 18.08, "learning_rate": 4.848262468629294e-07, "loss": 0.0067, "step": 58748 }, { "epoch": 18.08, "learning_rate": 4.846729932989724e-07, "loss": 0.0026, "step": 58749 }, { "epoch": 18.08, "learning_rate": 4.845197633590071e-07, "loss": 0.0026, "step": 58750 }, { "epoch": 18.08, "learning_rate": 4.843665570434153e-07, "loss": 0.0027, "step": 58751 }, { "epoch": 18.08, "learning_rate": 4.842133743525734e-07, "loss": 0.0031, "step": 58752 }, { "epoch": 18.08, "learning_rate": 4.840602152868656e-07, "loss": 0.0034, "step": 58753 }, { "epoch": 18.08, "learning_rate": 4.839070798466705e-07, "loss": 0.0043, "step": 58754 }, { "epoch": 18.08, "learning_rate": 4.837539680323689e-07, "loss": 0.0034, "step": 58755 }, { "epoch": 18.08, "learning_rate": 4.836008798443404e-07, "loss": 0.0025, "step": 58756 }, { "epoch": 18.08, "learning_rate": 4.834478152829658e-07, "loss": 0.0032, "step": 58757 }, { "epoch": 18.08, "learning_rate": 4.832947743486238e-07, "loss": 0.0151, "step": 58758 }, { "epoch": 18.08, "learning_rate": 4.831417570416963e-07, "loss": 0.0052, "step": 58759 }, { "epoch": 18.08, "learning_rate": 4.829887633625618e-07, "loss": 0.0021, "step": 58760 }, { "epoch": 18.08, "learning_rate": 4.828357933116001e-07, "loss": 0.0021, "step": 58761 }, { "epoch": 18.08, "learning_rate": 4.826828468891909e-07, "loss": 0.005, "step": 58762 }, { "epoch": 18.08, "learning_rate": 4.825299240957149e-07, "loss": 0.0056, "step": 58763 }, { "epoch": 18.08, "learning_rate": 4.823770249315518e-07, "loss": 0.0025, "step": 58764 }, { "epoch": 18.08, "learning_rate": 4.822241493970814e-07, "loss": 0.0036, "step": 58765 }, { "epoch": 18.08, "learning_rate": 4.820712974926822e-07, "loss": 0.0036, "step": 58766 }, { "epoch": 18.08, "learning_rate": 4.819184692187317e-07, "loss": 0.0063, "step": 58767 }, { "epoch": 18.08, "learning_rate": 4.81765664575613e-07, "loss": 0.0036, "step": 58768 }, { "epoch": 18.08, "learning_rate": 4.816128835637046e-07, "loss": 0.002, "step": 58769 }, { "epoch": 18.08, "learning_rate": 4.814601261833851e-07, "loss": 0.0026, "step": 58770 }, { "epoch": 18.08, "learning_rate": 4.813073924350342e-07, "loss": 0.0051, "step": 58771 }, { "epoch": 18.08, "learning_rate": 4.811546823190317e-07, "loss": 0.0045, "step": 58772 }, { "epoch": 18.08, "learning_rate": 4.810019958357548e-07, "loss": 0.0046, "step": 58773 }, { "epoch": 18.08, "learning_rate": 4.808493329855834e-07, "loss": 0.0033, "step": 58774 }, { "epoch": 18.08, "learning_rate": 4.806966937688973e-07, "loss": 0.005, "step": 58775 }, { "epoch": 18.08, "learning_rate": 4.805440781860748e-07, "loss": 0.0043, "step": 58776 }, { "epoch": 18.08, "learning_rate": 4.803914862374947e-07, "loss": 0.0042, "step": 58777 }, { "epoch": 18.09, "learning_rate": 4.802389179235378e-07, "loss": 0.0039, "step": 58778 }, { "epoch": 18.09, "learning_rate": 4.800863732445804e-07, "loss": 0.0034, "step": 58779 }, { "epoch": 18.09, "learning_rate": 4.799338522010011e-07, "loss": 0.0017, "step": 58780 }, { "epoch": 18.09, "learning_rate": 4.797813547931807e-07, "loss": 0.0028, "step": 58781 }, { "epoch": 18.09, "learning_rate": 4.796288810214944e-07, "loss": 0.0031, "step": 58782 }, { "epoch": 18.09, "learning_rate": 4.794764308863242e-07, "loss": 0.0021, "step": 58783 }, { "epoch": 18.09, "learning_rate": 4.793240043880487e-07, "loss": 0.0013, "step": 58784 }, { "epoch": 18.09, "learning_rate": 4.791716015270431e-07, "loss": 0.0034, "step": 58785 }, { "epoch": 18.09, "learning_rate": 4.790192223036905e-07, "loss": 0.0051, "step": 58786 }, { "epoch": 18.09, "learning_rate": 4.788668667183627e-07, "loss": 0.0027, "step": 58787 }, { "epoch": 18.09, "learning_rate": 4.787145347714439e-07, "loss": 0.004, "step": 58788 }, { "epoch": 18.09, "learning_rate": 4.785622264633094e-07, "loss": 0.0025, "step": 58789 }, { "epoch": 18.09, "learning_rate": 4.784099417943388e-07, "loss": 0.0027, "step": 58790 }, { "epoch": 18.09, "learning_rate": 4.782576807649086e-07, "loss": 0.0027, "step": 58791 }, { "epoch": 18.09, "learning_rate": 4.781054433753984e-07, "loss": 0.0037, "step": 58792 }, { "epoch": 18.09, "learning_rate": 4.779532296261846e-07, "loss": 0.0031, "step": 58793 }, { "epoch": 18.09, "learning_rate": 4.778010395176458e-07, "loss": 0.0032, "step": 58794 }, { "epoch": 18.09, "learning_rate": 4.776488730501605e-07, "loss": 0.0047, "step": 58795 }, { "epoch": 18.09, "learning_rate": 4.774967302241051e-07, "loss": 0.0036, "step": 58796 }, { "epoch": 18.09, "learning_rate": 4.773446110398571e-07, "loss": 0.0036, "step": 58797 }, { "epoch": 18.09, "learning_rate": 4.771925154977975e-07, "loss": 0.0034, "step": 58798 }, { "epoch": 18.09, "learning_rate": 4.770404435983012e-07, "loss": 0.003, "step": 58799 }, { "epoch": 18.09, "learning_rate": 4.768883953417458e-07, "loss": 0.0037, "step": 58800 }, { "epoch": 18.09, "learning_rate": 4.7673637072851e-07, "loss": 0.0022, "step": 58801 }, { "epoch": 18.09, "learning_rate": 4.7658436975896785e-07, "loss": 0.0039, "step": 58802 }, { "epoch": 18.09, "learning_rate": 4.764323924335013e-07, "loss": 0.0033, "step": 58803 }, { "epoch": 18.09, "learning_rate": 4.762804387524855e-07, "loss": 0.004, "step": 58804 }, { "epoch": 18.09, "learning_rate": 4.76128508716297e-07, "loss": 0.0039, "step": 58805 }, { "epoch": 18.09, "learning_rate": 4.759766023253143e-07, "loss": 0.0035, "step": 58806 }, { "epoch": 18.09, "learning_rate": 4.758247195799148e-07, "loss": 0.0034, "step": 58807 }, { "epoch": 18.09, "learning_rate": 4.756728604804739e-07, "loss": 0.0032, "step": 58808 }, { "epoch": 18.09, "learning_rate": 4.755210250273701e-07, "loss": 0.0037, "step": 58809 }, { "epoch": 18.1, "learning_rate": 4.7536921322098084e-07, "loss": 0.0015, "step": 58810 }, { "epoch": 18.1, "learning_rate": 4.752174250616803e-07, "loss": 0.0036, "step": 58811 }, { "epoch": 18.1, "learning_rate": 4.7506566054984827e-07, "loss": 0.0035, "step": 58812 }, { "epoch": 18.1, "learning_rate": 4.7491391968585987e-07, "loss": 0.0041, "step": 58813 }, { "epoch": 18.1, "learning_rate": 4.7476220247009264e-07, "loss": 0.0019, "step": 58814 }, { "epoch": 18.1, "learning_rate": 4.746105089029229e-07, "loss": 0.0063, "step": 58815 }, { "epoch": 18.1, "learning_rate": 4.744588389847271e-07, "loss": 0.0044, "step": 58816 }, { "epoch": 18.1, "learning_rate": 4.743071927158804e-07, "loss": 0.0026, "step": 58817 }, { "epoch": 18.1, "learning_rate": 4.741555700967626e-07, "loss": 0.0037, "step": 58818 }, { "epoch": 18.1, "learning_rate": 4.7400397112774886e-07, "loss": 0.0047, "step": 58819 }, { "epoch": 18.1, "learning_rate": 4.738523958092156e-07, "loss": 0.003, "step": 58820 }, { "epoch": 18.1, "learning_rate": 4.7370084414153805e-07, "loss": 0.0034, "step": 58821 }, { "epoch": 18.1, "learning_rate": 4.735493161250915e-07, "loss": 0.0043, "step": 58822 }, { "epoch": 18.1, "learning_rate": 4.7339781176025447e-07, "loss": 0.0029, "step": 58823 }, { "epoch": 18.1, "learning_rate": 4.7324633104740336e-07, "loss": 0.0034, "step": 58824 }, { "epoch": 18.1, "learning_rate": 4.7309487398691234e-07, "loss": 0.0035, "step": 58825 }, { "epoch": 18.1, "learning_rate": 4.7294344057915887e-07, "loss": 0.0047, "step": 58826 }, { "epoch": 18.1, "learning_rate": 4.7279203082451817e-07, "loss": 0.0024, "step": 58827 }, { "epoch": 18.1, "learning_rate": 4.726406447233667e-07, "loss": 0.0035, "step": 58828 }, { "epoch": 18.1, "learning_rate": 4.724892822760807e-07, "loss": 0.0042, "step": 58829 }, { "epoch": 18.1, "learning_rate": 4.723379434830344e-07, "loss": 0.0028, "step": 58830 }, { "epoch": 18.1, "learning_rate": 4.721866283446042e-07, "loss": 0.0039, "step": 58831 }, { "epoch": 18.1, "learning_rate": 4.720353368611652e-07, "loss": 0.0047, "step": 58832 }, { "epoch": 18.1, "learning_rate": 4.718840690330961e-07, "loss": 0.0048, "step": 58833 }, { "epoch": 18.1, "learning_rate": 4.717328248607689e-07, "loss": 0.002, "step": 58834 }, { "epoch": 18.1, "learning_rate": 4.715816043445609e-07, "loss": 0.0032, "step": 58835 }, { "epoch": 18.1, "learning_rate": 4.7143040748484634e-07, "loss": 0.0057, "step": 58836 }, { "epoch": 18.1, "learning_rate": 4.712792342820005e-07, "loss": 0.0027, "step": 58837 }, { "epoch": 18.1, "learning_rate": 4.711280847363997e-07, "loss": 0.0025, "step": 58838 }, { "epoch": 18.1, "learning_rate": 4.709769588484203e-07, "loss": 0.0042, "step": 58839 }, { "epoch": 18.1, "learning_rate": 4.7082585661843537e-07, "loss": 0.0021, "step": 58840 }, { "epoch": 18.1, "learning_rate": 4.706747780468224e-07, "loss": 0.0034, "step": 58841 }, { "epoch": 18.1, "learning_rate": 4.7052372313395213e-07, "loss": 0.0017, "step": 58842 }, { "epoch": 18.11, "learning_rate": 4.703726918802032e-07, "loss": 0.0034, "step": 58843 }, { "epoch": 18.11, "learning_rate": 4.702216842859508e-07, "loss": 0.0033, "step": 58844 }, { "epoch": 18.11, "learning_rate": 4.700707003515681e-07, "loss": 0.0032, "step": 58845 }, { "epoch": 18.11, "learning_rate": 4.6991974007743133e-07, "loss": 0.006, "step": 58846 }, { "epoch": 18.11, "learning_rate": 4.6976880346391365e-07, "loss": 0.0037, "step": 58847 }, { "epoch": 18.11, "learning_rate": 4.696178905113913e-07, "loss": 0.0035, "step": 58848 }, { "epoch": 18.11, "learning_rate": 4.694670012202385e-07, "loss": 0.0024, "step": 58849 }, { "epoch": 18.11, "learning_rate": 4.6931613559083045e-07, "loss": 0.0051, "step": 58850 }, { "epoch": 18.11, "learning_rate": 4.691652936235402e-07, "loss": 0.0039, "step": 58851 }, { "epoch": 18.11, "learning_rate": 4.69014475318742e-07, "loss": 0.0018, "step": 58852 }, { "epoch": 18.11, "learning_rate": 4.688636806768121e-07, "loss": 0.0031, "step": 58853 }, { "epoch": 18.11, "learning_rate": 4.6871290969812576e-07, "loss": 0.0033, "step": 58854 }, { "epoch": 18.11, "learning_rate": 4.6856216238305494e-07, "loss": 0.0025, "step": 58855 }, { "epoch": 18.11, "learning_rate": 4.6841143873197493e-07, "loss": 0.0046, "step": 58856 }, { "epoch": 18.11, "learning_rate": 4.6826073874525756e-07, "loss": 0.0038, "step": 58857 }, { "epoch": 18.11, "learning_rate": 4.681100624232804e-07, "loss": 0.0028, "step": 58858 }, { "epoch": 18.11, "learning_rate": 4.679594097664164e-07, "loss": 0.0028, "step": 58859 }, { "epoch": 18.11, "learning_rate": 4.678087807750398e-07, "loss": 0.0032, "step": 58860 }, { "epoch": 18.11, "learning_rate": 4.676581754495235e-07, "loss": 0.0039, "step": 58861 }, { "epoch": 18.11, "learning_rate": 4.675075937902429e-07, "loss": 0.0039, "step": 58862 }, { "epoch": 18.11, "learning_rate": 4.6735703579757095e-07, "loss": 0.0044, "step": 58863 }, { "epoch": 18.11, "learning_rate": 4.6720650147188076e-07, "loss": 0.0022, "step": 58864 }, { "epoch": 18.11, "learning_rate": 4.6705599081354747e-07, "loss": 0.003, "step": 58865 }, { "epoch": 18.11, "learning_rate": 4.669055038229442e-07, "loss": 0.0036, "step": 58866 }, { "epoch": 18.11, "learning_rate": 4.6675504050044283e-07, "loss": 0.006, "step": 58867 }, { "epoch": 18.11, "learning_rate": 4.66604600846422e-07, "loss": 0.0046, "step": 58868 }, { "epoch": 18.11, "learning_rate": 4.6645418486124916e-07, "loss": 0.0031, "step": 58869 }, { "epoch": 18.11, "learning_rate": 4.663037925453018e-07, "loss": 0.0051, "step": 58870 }, { "epoch": 18.11, "learning_rate": 4.661534238989518e-07, "loss": 0.0036, "step": 58871 }, { "epoch": 18.11, "learning_rate": 4.6600307892257003e-07, "loss": 0.0076, "step": 58872 }, { "epoch": 18.11, "learning_rate": 4.65852757616535e-07, "loss": 0.0037, "step": 58873 }, { "epoch": 18.11, "learning_rate": 4.6570245998121654e-07, "loss": 0.0031, "step": 58874 }, { "epoch": 18.12, "learning_rate": 4.655521860169898e-07, "loss": 0.0053, "step": 58875 }, { "epoch": 18.12, "learning_rate": 4.6540193572422567e-07, "loss": 0.0037, "step": 58876 }, { "epoch": 18.12, "learning_rate": 4.65251709103296e-07, "loss": 0.0025, "step": 58877 }, { "epoch": 18.12, "learning_rate": 4.651015061545783e-07, "loss": 0.0036, "step": 58878 }, { "epoch": 18.12, "learning_rate": 4.6495132687844224e-07, "loss": 0.0044, "step": 58879 }, { "epoch": 18.12, "learning_rate": 4.6480117127526204e-07, "loss": 0.0016, "step": 58880 }, { "epoch": 18.12, "learning_rate": 4.646510393454096e-07, "loss": 0.0021, "step": 58881 }, { "epoch": 18.12, "learning_rate": 4.6450093108925784e-07, "loss": 0.0067, "step": 58882 }, { "epoch": 18.12, "learning_rate": 4.6435084650718e-07, "loss": 0.0034, "step": 58883 }, { "epoch": 18.12, "learning_rate": 4.6420078559954784e-07, "loss": 0.003, "step": 58884 }, { "epoch": 18.12, "learning_rate": 4.6405074836673445e-07, "loss": 0.0015, "step": 58885 }, { "epoch": 18.12, "learning_rate": 4.639007348091118e-07, "loss": 0.0041, "step": 58886 }, { "epoch": 18.12, "learning_rate": 4.6375074492705173e-07, "loss": 0.0062, "step": 58887 }, { "epoch": 18.12, "learning_rate": 4.6360077872093066e-07, "loss": 0.0032, "step": 58888 }, { "epoch": 18.12, "learning_rate": 4.634508361911161e-07, "loss": 0.0033, "step": 58889 }, { "epoch": 18.12, "learning_rate": 4.633009173379832e-07, "loss": 0.0037, "step": 58890 }, { "epoch": 18.12, "learning_rate": 4.631510221619018e-07, "loss": 0.0025, "step": 58891 }, { "epoch": 18.12, "learning_rate": 4.630011506632448e-07, "loss": 0.0022, "step": 58892 }, { "epoch": 18.12, "learning_rate": 4.6285130284238645e-07, "loss": 0.0031, "step": 58893 }, { "epoch": 18.12, "learning_rate": 4.627014786996964e-07, "loss": 0.0038, "step": 58894 }, { "epoch": 18.12, "learning_rate": 4.625516782355477e-07, "loss": 0.0046, "step": 58895 }, { "epoch": 18.12, "learning_rate": 4.624019014503134e-07, "loss": 0.0012, "step": 58896 }, { "epoch": 18.12, "learning_rate": 4.6225214834436094e-07, "loss": 0.0055, "step": 58897 }, { "epoch": 18.12, "learning_rate": 4.6210241891806674e-07, "loss": 0.0027, "step": 58898 }, { "epoch": 18.12, "learning_rate": 4.6195271317180044e-07, "loss": 0.0037, "step": 58899 }, { "epoch": 18.12, "learning_rate": 4.618030311059352e-07, "loss": 0.0034, "step": 58900 }, { "epoch": 18.12, "learning_rate": 4.6165337272084056e-07, "loss": 0.0046, "step": 58901 }, { "epoch": 18.12, "learning_rate": 4.6150373801688963e-07, "loss": 0.0041, "step": 58902 }, { "epoch": 18.12, "learning_rate": 4.6135412699445323e-07, "loss": 0.0017, "step": 58903 }, { "epoch": 18.12, "learning_rate": 4.612045396539033e-07, "loss": 0.0047, "step": 58904 }, { "epoch": 18.12, "learning_rate": 4.610549759956107e-07, "loss": 0.0034, "step": 58905 }, { "epoch": 18.12, "learning_rate": 4.609054360199472e-07, "loss": 0.0053, "step": 58906 }, { "epoch": 18.12, "learning_rate": 4.607559197272815e-07, "loss": 0.0053, "step": 58907 }, { "epoch": 18.13, "learning_rate": 4.6060642711799e-07, "loss": 0.0039, "step": 58908 }, { "epoch": 18.13, "learning_rate": 4.604569581924412e-07, "loss": 0.0026, "step": 58909 }, { "epoch": 18.13, "learning_rate": 4.6030751295100486e-07, "loss": 0.0053, "step": 58910 }, { "epoch": 18.13, "learning_rate": 4.60158091394054e-07, "loss": 0.0022, "step": 58911 }, { "epoch": 18.13, "learning_rate": 4.6000869352195607e-07, "loss": 0.0016, "step": 58912 }, { "epoch": 18.13, "learning_rate": 4.5985931933508757e-07, "loss": 0.0051, "step": 58913 }, { "epoch": 18.13, "learning_rate": 4.597099688338147e-07, "loss": 0.0049, "step": 58914 }, { "epoch": 18.13, "learning_rate": 4.595606420185117e-07, "loss": 0.0048, "step": 58915 }, { "epoch": 18.13, "learning_rate": 4.5941133888954605e-07, "loss": 0.0099, "step": 58916 }, { "epoch": 18.13, "learning_rate": 4.5926205944729073e-07, "loss": 0.0027, "step": 58917 }, { "epoch": 18.13, "learning_rate": 4.5911280369211553e-07, "loss": 0.0047, "step": 58918 }, { "epoch": 18.13, "learning_rate": 4.589635716243912e-07, "loss": 0.0032, "step": 58919 }, { "epoch": 18.13, "learning_rate": 4.5881436324448856e-07, "loss": 0.0027, "step": 58920 }, { "epoch": 18.13, "learning_rate": 4.586651785527751e-07, "loss": 0.0049, "step": 58921 }, { "epoch": 18.13, "learning_rate": 4.5851601754962614e-07, "loss": 0.002, "step": 58922 }, { "epoch": 18.13, "learning_rate": 4.583668802354102e-07, "loss": 0.0029, "step": 58923 }, { "epoch": 18.13, "learning_rate": 4.582177666104959e-07, "loss": 0.0031, "step": 58924 }, { "epoch": 18.13, "learning_rate": 4.5806867667525513e-07, "loss": 0.0032, "step": 58925 }, { "epoch": 18.13, "learning_rate": 4.579196104300554e-07, "loss": 0.0022, "step": 58926 }, { "epoch": 18.13, "learning_rate": 4.5777056787526974e-07, "loss": 0.0026, "step": 58927 }, { "epoch": 18.13, "learning_rate": 4.5762154901126787e-07, "loss": 0.0026, "step": 58928 }, { "epoch": 18.13, "learning_rate": 4.574725538384195e-07, "loss": 0.0031, "step": 58929 }, { "epoch": 18.13, "learning_rate": 4.5732358235709537e-07, "loss": 0.0033, "step": 58930 }, { "epoch": 18.13, "learning_rate": 4.5717463456766087e-07, "loss": 0.0048, "step": 58931 }, { "epoch": 18.13, "learning_rate": 4.5702571047049113e-07, "loss": 0.0054, "step": 58932 }, { "epoch": 18.13, "learning_rate": 4.5687681006595265e-07, "loss": 0.0031, "step": 58933 }, { "epoch": 18.13, "learning_rate": 4.5672793335441723e-07, "loss": 0.003, "step": 58934 }, { "epoch": 18.13, "learning_rate": 4.565790803362535e-07, "loss": 0.0027, "step": 58935 }, { "epoch": 18.13, "learning_rate": 4.564302510118312e-07, "loss": 0.0042, "step": 58936 }, { "epoch": 18.13, "learning_rate": 4.5628144538152006e-07, "loss": 0.0017, "step": 58937 }, { "epoch": 18.13, "learning_rate": 4.5613266344568864e-07, "loss": 0.0027, "step": 58938 }, { "epoch": 18.13, "learning_rate": 4.559839052047066e-07, "loss": 0.0042, "step": 58939 }, { "epoch": 18.14, "learning_rate": 4.5583517065894366e-07, "loss": 0.0026, "step": 58940 }, { "epoch": 18.14, "learning_rate": 4.556864598087685e-07, "loss": 0.0031, "step": 58941 }, { "epoch": 18.14, "learning_rate": 4.5553777265455067e-07, "loss": 0.004, "step": 58942 }, { "epoch": 18.14, "learning_rate": 4.5538910919666115e-07, "loss": 0.0051, "step": 58943 }, { "epoch": 18.14, "learning_rate": 4.552404694354673e-07, "loss": 0.0053, "step": 58944 }, { "epoch": 18.14, "learning_rate": 4.550918533713367e-07, "loss": 0.005, "step": 58945 }, { "epoch": 18.14, "learning_rate": 4.5494326100464004e-07, "loss": 0.0025, "step": 58946 }, { "epoch": 18.14, "learning_rate": 4.54794692335746e-07, "loss": 0.0057, "step": 58947 }, { "epoch": 18.14, "learning_rate": 4.546461473650232e-07, "loss": 0.0047, "step": 58948 }, { "epoch": 18.14, "learning_rate": 4.5449762609284134e-07, "loss": 0.0027, "step": 58949 }, { "epoch": 18.14, "learning_rate": 4.543491285195689e-07, "loss": 0.003, "step": 58950 }, { "epoch": 18.14, "learning_rate": 4.542006546455735e-07, "loss": 0.005, "step": 58951 }, { "epoch": 18.14, "learning_rate": 4.540522044712248e-07, "loss": 0.0036, "step": 58952 }, { "epoch": 18.14, "learning_rate": 4.539037779968902e-07, "loss": 0.0039, "step": 58953 }, { "epoch": 18.14, "learning_rate": 4.5375537522293956e-07, "loss": 0.0051, "step": 58954 }, { "epoch": 18.14, "learning_rate": 4.5360699614974024e-07, "loss": 0.0033, "step": 58955 }, { "epoch": 18.14, "learning_rate": 4.5345864077766087e-07, "loss": 0.0038, "step": 58956 }, { "epoch": 18.14, "learning_rate": 4.533103091070723e-07, "loss": 0.0038, "step": 58957 }, { "epoch": 18.14, "learning_rate": 4.5316200113833866e-07, "loss": 0.0041, "step": 58958 }, { "epoch": 18.14, "learning_rate": 4.530137168718296e-07, "loss": 0.0045, "step": 58959 }, { "epoch": 18.14, "learning_rate": 4.528654563079149e-07, "loss": 0.0054, "step": 58960 }, { "epoch": 18.14, "learning_rate": 4.5271721944695977e-07, "loss": 0.0023, "step": 58961 }, { "epoch": 18.14, "learning_rate": 4.525690062893351e-07, "loss": 0.0038, "step": 58962 }, { "epoch": 18.14, "learning_rate": 4.524208168354072e-07, "loss": 0.0039, "step": 58963 }, { "epoch": 18.14, "learning_rate": 4.5227265108554466e-07, "loss": 0.0034, "step": 58964 }, { "epoch": 18.14, "learning_rate": 4.521245090401172e-07, "loss": 0.0031, "step": 58965 }, { "epoch": 18.14, "learning_rate": 4.5197639069948784e-07, "loss": 0.0027, "step": 58966 }, { "epoch": 18.14, "learning_rate": 4.5182829606402743e-07, "loss": 0.0025, "step": 58967 }, { "epoch": 18.14, "learning_rate": 4.516802251341035e-07, "loss": 0.0045, "step": 58968 }, { "epoch": 18.14, "learning_rate": 4.515321779100834e-07, "loss": 0.004, "step": 58969 }, { "epoch": 18.14, "learning_rate": 4.5138415439233586e-07, "loss": 0.0025, "step": 58970 }, { "epoch": 18.14, "learning_rate": 4.5123615458122605e-07, "loss": 0.0028, "step": 58971 }, { "epoch": 18.14, "learning_rate": 4.510881784771226e-07, "loss": 0.0058, "step": 58972 }, { "epoch": 18.15, "learning_rate": 4.50940226080393e-07, "loss": 0.0037, "step": 58973 }, { "epoch": 18.15, "learning_rate": 4.507922973914047e-07, "loss": 0.0036, "step": 58974 }, { "epoch": 18.15, "learning_rate": 4.506443924105253e-07, "loss": 0.0034, "step": 58975 }, { "epoch": 18.15, "learning_rate": 4.5049651113811985e-07, "loss": 0.0051, "step": 58976 }, { "epoch": 18.15, "learning_rate": 4.5034865357455827e-07, "loss": 0.0025, "step": 58977 }, { "epoch": 18.15, "learning_rate": 4.502008197202068e-07, "loss": 0.0026, "step": 58978 }, { "epoch": 18.15, "learning_rate": 4.500530095754319e-07, "loss": 0.0049, "step": 58979 }, { "epoch": 18.15, "learning_rate": 4.499052231406009e-07, "loss": 0.0017, "step": 58980 }, { "epoch": 18.15, "learning_rate": 4.497574604160793e-07, "loss": 0.0033, "step": 58981 }, { "epoch": 18.15, "learning_rate": 4.4960972140223547e-07, "loss": 0.0016, "step": 58982 }, { "epoch": 18.15, "learning_rate": 4.4946200609943704e-07, "loss": 0.0035, "step": 58983 }, { "epoch": 18.15, "learning_rate": 4.4931431450804917e-07, "loss": 0.0029, "step": 58984 }, { "epoch": 18.15, "learning_rate": 4.4916664662843947e-07, "loss": 0.0035, "step": 58985 }, { "epoch": 18.15, "learning_rate": 4.490190024609742e-07, "loss": 0.0042, "step": 58986 }, { "epoch": 18.15, "learning_rate": 4.4887138200601976e-07, "loss": 0.0038, "step": 58987 }, { "epoch": 18.15, "learning_rate": 4.4872378526394254e-07, "loss": 0.0042, "step": 58988 }, { "epoch": 18.15, "learning_rate": 4.4857621223510896e-07, "loss": 0.0077, "step": 58989 }, { "epoch": 18.15, "learning_rate": 4.484286629198864e-07, "loss": 0.0035, "step": 58990 }, { "epoch": 18.15, "learning_rate": 4.482811373186402e-07, "loss": 0.0047, "step": 58991 }, { "epoch": 18.15, "learning_rate": 4.481336354317367e-07, "loss": 0.0029, "step": 58992 }, { "epoch": 18.15, "learning_rate": 4.479861572595423e-07, "loss": 0.0032, "step": 58993 }, { "epoch": 18.15, "learning_rate": 4.4783870280242334e-07, "loss": 0.002, "step": 58994 }, { "epoch": 18.15, "learning_rate": 4.4769127206074514e-07, "loss": 0.0079, "step": 58995 }, { "epoch": 18.15, "learning_rate": 4.4754386503487403e-07, "loss": 0.0052, "step": 58996 }, { "epoch": 18.15, "learning_rate": 4.4739648172517634e-07, "loss": 0.0058, "step": 58997 }, { "epoch": 18.15, "learning_rate": 4.4724912213201853e-07, "loss": 0.0029, "step": 58998 }, { "epoch": 18.15, "learning_rate": 4.4710178625576694e-07, "loss": 0.0055, "step": 58999 }, { "epoch": 18.15, "learning_rate": 4.4695447409678574e-07, "loss": 0.0032, "step": 59000 }, { "epoch": 18.15, "learning_rate": 4.4680718565543904e-07, "loss": 0.0059, "step": 59001 }, { "epoch": 18.15, "learning_rate": 4.466599209320954e-07, "loss": 0.0038, "step": 59002 }, { "epoch": 18.15, "learning_rate": 4.465126799271213e-07, "loss": 0.0035, "step": 59003 }, { "epoch": 18.15, "learning_rate": 4.463654626408798e-07, "loss": 0.0037, "step": 59004 }, { "epoch": 18.16, "learning_rate": 4.4621826907373713e-07, "loss": 0.006, "step": 59005 }, { "epoch": 18.16, "learning_rate": 4.4607109922605863e-07, "loss": 0.0041, "step": 59006 }, { "epoch": 18.16, "learning_rate": 4.4592395309820956e-07, "loss": 0.0041, "step": 59007 }, { "epoch": 18.16, "learning_rate": 4.457768306905563e-07, "loss": 0.0033, "step": 59008 }, { "epoch": 18.16, "learning_rate": 4.4562973200346413e-07, "loss": 0.002, "step": 59009 }, { "epoch": 18.16, "learning_rate": 4.454826570372972e-07, "loss": 0.0056, "step": 59010 }, { "epoch": 18.16, "learning_rate": 4.453356057924185e-07, "loss": 0.0051, "step": 59011 }, { "epoch": 18.16, "learning_rate": 4.4518857826919894e-07, "loss": 0.0044, "step": 59012 }, { "epoch": 18.16, "learning_rate": 4.450415744679992e-07, "loss": 0.0014, "step": 59013 }, { "epoch": 18.16, "learning_rate": 4.448945943891858e-07, "loss": 0.0045, "step": 59014 }, { "epoch": 18.16, "learning_rate": 4.447476380331228e-07, "loss": 0.0064, "step": 59015 }, { "epoch": 18.16, "learning_rate": 4.446007054001733e-07, "loss": 0.0024, "step": 59016 }, { "epoch": 18.16, "learning_rate": 4.4445379649070587e-07, "loss": 0.0039, "step": 59017 }, { "epoch": 18.16, "learning_rate": 4.4430691130508354e-07, "loss": 0.0054, "step": 59018 }, { "epoch": 18.16, "learning_rate": 4.4416004984367154e-07, "loss": 0.0042, "step": 59019 }, { "epoch": 18.16, "learning_rate": 4.4401321210683523e-07, "loss": 0.0035, "step": 59020 }, { "epoch": 18.16, "learning_rate": 4.4386639809493426e-07, "loss": 0.0039, "step": 59021 }, { "epoch": 18.16, "learning_rate": 4.4371960780833943e-07, "loss": 0.0041, "step": 59022 }, { "epoch": 18.16, "learning_rate": 4.4357284124741163e-07, "loss": 0.0036, "step": 59023 }, { "epoch": 18.16, "learning_rate": 4.4342609841251603e-07, "loss": 0.0033, "step": 59024 }, { "epoch": 18.16, "learning_rate": 4.4327937930401797e-07, "loss": 0.0026, "step": 59025 }, { "epoch": 18.16, "learning_rate": 4.4313268392228046e-07, "loss": 0.0029, "step": 59026 }, { "epoch": 18.16, "learning_rate": 4.4298601226766766e-07, "loss": 0.0047, "step": 59027 }, { "epoch": 18.16, "learning_rate": 4.4283936434054376e-07, "loss": 0.0048, "step": 59028 }, { "epoch": 18.16, "learning_rate": 4.4269274014127397e-07, "loss": 0.0037, "step": 59029 }, { "epoch": 18.16, "learning_rate": 4.425461396702213e-07, "loss": 0.0021, "step": 59030 }, { "epoch": 18.16, "learning_rate": 4.423995629277489e-07, "loss": 0.002, "step": 59031 }, { "epoch": 18.16, "learning_rate": 4.4225300991422303e-07, "loss": 0.0047, "step": 59032 }, { "epoch": 18.16, "learning_rate": 4.421064806300068e-07, "loss": 0.0057, "step": 59033 }, { "epoch": 18.16, "learning_rate": 4.419599750754622e-07, "loss": 0.004, "step": 59034 }, { "epoch": 18.16, "learning_rate": 4.4181349325095546e-07, "loss": 0.005, "step": 59035 }, { "epoch": 18.16, "learning_rate": 4.416670351568464e-07, "loss": 0.0032, "step": 59036 }, { "epoch": 18.16, "learning_rate": 4.4152060079350354e-07, "loss": 0.0032, "step": 59037 }, { "epoch": 18.17, "learning_rate": 4.4137419016128666e-07, "loss": 0.0046, "step": 59038 }, { "epoch": 18.17, "learning_rate": 4.412278032605621e-07, "loss": 0.0023, "step": 59039 }, { "epoch": 18.17, "learning_rate": 4.4108144009169186e-07, "loss": 0.0026, "step": 59040 }, { "epoch": 18.17, "learning_rate": 4.4093510065503884e-07, "loss": 0.0023, "step": 59041 }, { "epoch": 18.17, "learning_rate": 4.4078878495096733e-07, "loss": 0.0023, "step": 59042 }, { "epoch": 18.17, "learning_rate": 4.406424929798403e-07, "loss": 0.0016, "step": 59043 }, { "epoch": 18.17, "learning_rate": 4.404962247420197e-07, "loss": 0.0034, "step": 59044 }, { "epoch": 18.17, "learning_rate": 4.403499802378708e-07, "loss": 0.0049, "step": 59045 }, { "epoch": 18.17, "learning_rate": 4.402037594677555e-07, "loss": 0.0038, "step": 59046 }, { "epoch": 18.17, "learning_rate": 4.4005756243203803e-07, "loss": 0.003, "step": 59047 }, { "epoch": 18.17, "learning_rate": 4.399113891310791e-07, "loss": 0.0049, "step": 59048 }, { "epoch": 18.17, "learning_rate": 4.39765239565243e-07, "loss": 0.0026, "step": 59049 }, { "epoch": 18.17, "learning_rate": 4.3961911373489264e-07, "loss": 0.0022, "step": 59050 }, { "epoch": 18.17, "learning_rate": 4.394730116403889e-07, "loss": 0.0047, "step": 59051 }, { "epoch": 18.17, "learning_rate": 4.393269332820982e-07, "loss": 0.0027, "step": 59052 }, { "epoch": 18.17, "learning_rate": 4.391808786603813e-07, "loss": 0.0051, "step": 59053 }, { "epoch": 18.17, "learning_rate": 4.390348477756012e-07, "loss": 0.0023, "step": 59054 }, { "epoch": 18.17, "learning_rate": 4.3888884062812e-07, "loss": 0.0038, "step": 59055 }, { "epoch": 18.17, "learning_rate": 4.3874285721829835e-07, "loss": 0.0028, "step": 59056 }, { "epoch": 18.17, "learning_rate": 4.3859689754650157e-07, "loss": 0.0032, "step": 59057 }, { "epoch": 18.17, "learning_rate": 4.3845096161309055e-07, "loss": 0.0039, "step": 59058 }, { "epoch": 18.17, "learning_rate": 4.3830504941842935e-07, "loss": 0.0034, "step": 59059 }, { "epoch": 18.17, "learning_rate": 4.381591609628777e-07, "loss": 0.0037, "step": 59060 }, { "epoch": 18.17, "learning_rate": 4.3801329624679976e-07, "loss": 0.0031, "step": 59061 }, { "epoch": 18.17, "learning_rate": 4.3786745527055643e-07, "loss": 0.0035, "step": 59062 }, { "epoch": 18.17, "learning_rate": 4.377216380345106e-07, "loss": 0.0065, "step": 59063 }, { "epoch": 18.17, "learning_rate": 4.3757584453902434e-07, "loss": 0.0037, "step": 59064 }, { "epoch": 18.17, "learning_rate": 4.374300747844595e-07, "loss": 0.0064, "step": 59065 }, { "epoch": 18.17, "learning_rate": 4.372843287711759e-07, "loss": 0.0056, "step": 59066 }, { "epoch": 18.17, "learning_rate": 4.371386064995409e-07, "loss": 0.0029, "step": 59067 }, { "epoch": 18.17, "learning_rate": 4.369929079699098e-07, "loss": 0.0026, "step": 59068 }, { "epoch": 18.17, "learning_rate": 4.368472331826479e-07, "loss": 0.0033, "step": 59069 }, { "epoch": 18.18, "learning_rate": 4.367015821381171e-07, "loss": 0.0016, "step": 59070 }, { "epoch": 18.18, "learning_rate": 4.36555954836676e-07, "loss": 0.0041, "step": 59071 }, { "epoch": 18.18, "learning_rate": 4.364103512786888e-07, "loss": 0.0026, "step": 59072 }, { "epoch": 18.18, "learning_rate": 4.362647714645174e-07, "loss": 0.0026, "step": 59073 }, { "epoch": 18.18, "learning_rate": 4.361192153945215e-07, "loss": 0.0014, "step": 59074 }, { "epoch": 18.18, "learning_rate": 4.359736830690653e-07, "loss": 0.0029, "step": 59075 }, { "epoch": 18.18, "learning_rate": 4.3582817448850513e-07, "loss": 0.0043, "step": 59076 }, { "epoch": 18.18, "learning_rate": 4.3568268965320736e-07, "loss": 0.0044, "step": 59077 }, { "epoch": 18.18, "learning_rate": 4.3553722856352954e-07, "loss": 0.0042, "step": 59078 }, { "epoch": 18.18, "learning_rate": 4.3539179121983466e-07, "loss": 0.0045, "step": 59079 }, { "epoch": 18.18, "learning_rate": 4.352463776224847e-07, "loss": 0.0035, "step": 59080 }, { "epoch": 18.18, "learning_rate": 4.351009877718382e-07, "loss": 0.004, "step": 59081 }, { "epoch": 18.18, "learning_rate": 4.3495562166825714e-07, "loss": 0.0037, "step": 59082 }, { "epoch": 18.18, "learning_rate": 4.348102793121034e-07, "loss": 0.0024, "step": 59083 }, { "epoch": 18.18, "learning_rate": 4.346649607037368e-07, "loss": 0.0039, "step": 59084 }, { "epoch": 18.18, "learning_rate": 4.3451966584351803e-07, "loss": 0.0033, "step": 59085 }, { "epoch": 18.18, "learning_rate": 4.3437439473180687e-07, "loss": 0.0021, "step": 59086 }, { "epoch": 18.18, "learning_rate": 4.3422914736896635e-07, "loss": 0.0023, "step": 59087 }, { "epoch": 18.18, "learning_rate": 4.3408392375535844e-07, "loss": 0.0021, "step": 59088 }, { "epoch": 18.18, "learning_rate": 4.339387238913395e-07, "loss": 0.0033, "step": 59089 }, { "epoch": 18.18, "learning_rate": 4.3379354777727144e-07, "loss": 0.0045, "step": 59090 }, { "epoch": 18.18, "learning_rate": 4.33648395413514e-07, "loss": 0.0034, "step": 59091 }, { "epoch": 18.18, "learning_rate": 4.335032668004291e-07, "loss": 0.0028, "step": 59092 }, { "epoch": 18.18, "learning_rate": 4.333581619383775e-07, "loss": 0.0041, "step": 59093 }, { "epoch": 18.18, "learning_rate": 4.3321308082771795e-07, "loss": 0.0042, "step": 59094 }, { "epoch": 18.18, "learning_rate": 4.3306802346881116e-07, "loss": 0.0054, "step": 59095 }, { "epoch": 18.18, "learning_rate": 4.32922989862018e-07, "loss": 0.0034, "step": 59096 }, { "epoch": 18.18, "learning_rate": 4.32777980007697e-07, "loss": 0.002, "step": 59097 }, { "epoch": 18.18, "learning_rate": 4.3263299390620903e-07, "loss": 0.0029, "step": 59098 }, { "epoch": 18.18, "learning_rate": 4.3248803155791496e-07, "loss": 0.0037, "step": 59099 }, { "epoch": 18.18, "learning_rate": 4.3234309296317335e-07, "loss": 0.0021, "step": 59100 }, { "epoch": 18.18, "learning_rate": 4.3219817812234277e-07, "loss": 0.0032, "step": 59101 }, { "epoch": 18.18, "learning_rate": 4.320532870357874e-07, "loss": 0.0038, "step": 59102 }, { "epoch": 18.19, "learning_rate": 4.3190841970386367e-07, "loss": 0.0016, "step": 59103 }, { "epoch": 18.19, "learning_rate": 4.317635761269312e-07, "loss": 0.0047, "step": 59104 }, { "epoch": 18.19, "learning_rate": 4.3161875630535086e-07, "loss": 0.002, "step": 59105 }, { "epoch": 18.19, "learning_rate": 4.314739602394791e-07, "loss": 0.0036, "step": 59106 }, { "epoch": 18.19, "learning_rate": 4.313291879296799e-07, "loss": 0.0027, "step": 59107 }, { "epoch": 18.19, "learning_rate": 4.3118443937631094e-07, "loss": 0.0036, "step": 59108 }, { "epoch": 18.19, "learning_rate": 4.310397145797329e-07, "loss": 0.0016, "step": 59109 }, { "epoch": 18.19, "learning_rate": 4.308950135403023e-07, "loss": 0.005, "step": 59110 }, { "epoch": 18.19, "learning_rate": 4.3075033625837757e-07, "loss": 0.0044, "step": 59111 }, { "epoch": 18.19, "learning_rate": 4.306056827343208e-07, "loss": 0.0039, "step": 59112 }, { "epoch": 18.19, "learning_rate": 4.304610529684916e-07, "loss": 0.0031, "step": 59113 }, { "epoch": 18.19, "learning_rate": 4.3031644696124753e-07, "loss": 0.0056, "step": 59114 }, { "epoch": 18.19, "learning_rate": 4.3017186471294713e-07, "loss": 0.0036, "step": 59115 }, { "epoch": 18.19, "learning_rate": 4.300273062239502e-07, "loss": 0.0032, "step": 59116 }, { "epoch": 18.19, "learning_rate": 4.2988277149461634e-07, "loss": 0.0025, "step": 59117 }, { "epoch": 18.19, "learning_rate": 4.29738260525302e-07, "loss": 0.005, "step": 59118 }, { "epoch": 18.19, "learning_rate": 4.295937733163691e-07, "loss": 0.0031, "step": 59119 }, { "epoch": 18.19, "learning_rate": 4.29449309868174e-07, "loss": 0.0031, "step": 59120 }, { "epoch": 18.19, "learning_rate": 4.2930487018107425e-07, "loss": 0.0035, "step": 59121 }, { "epoch": 18.19, "learning_rate": 4.2916045425543286e-07, "loss": 0.0039, "step": 59122 }, { "epoch": 18.19, "learning_rate": 4.2901606209160616e-07, "loss": 0.0061, "step": 59123 }, { "epoch": 18.19, "learning_rate": 4.2887169368995064e-07, "loss": 0.0034, "step": 59124 }, { "epoch": 18.19, "learning_rate": 4.2872734905082814e-07, "loss": 0.0053, "step": 59125 }, { "epoch": 18.19, "learning_rate": 4.2858302817459283e-07, "loss": 0.0042, "step": 59126 }, { "epoch": 18.19, "learning_rate": 4.284387310616067e-07, "loss": 0.0031, "step": 59127 }, { "epoch": 18.19, "learning_rate": 4.2829445771222723e-07, "loss": 0.0033, "step": 59128 }, { "epoch": 18.19, "learning_rate": 4.2815020812681184e-07, "loss": 0.0057, "step": 59129 }, { "epoch": 18.19, "learning_rate": 4.280059823057203e-07, "loss": 0.0036, "step": 59130 }, { "epoch": 18.19, "learning_rate": 4.278617802493068e-07, "loss": 0.0033, "step": 59131 }, { "epoch": 18.19, "learning_rate": 4.277176019579332e-07, "loss": 0.0042, "step": 59132 }, { "epoch": 18.19, "learning_rate": 4.2757344743195596e-07, "loss": 0.0037, "step": 59133 }, { "epoch": 18.19, "learning_rate": 4.274293166717336e-07, "loss": 0.0031, "step": 59134 }, { "epoch": 18.2, "learning_rate": 4.2728520967762365e-07, "loss": 0.0054, "step": 59135 }, { "epoch": 18.2, "learning_rate": 4.271411264499836e-07, "loss": 0.0045, "step": 59136 }, { "epoch": 18.2, "learning_rate": 4.269970669891721e-07, "loss": 0.003, "step": 59137 }, { "epoch": 18.2, "learning_rate": 4.2685303129554544e-07, "loss": 0.0025, "step": 59138 }, { "epoch": 18.2, "learning_rate": 4.267090193694612e-07, "loss": 0.003, "step": 59139 }, { "epoch": 18.2, "learning_rate": 4.265650312112779e-07, "loss": 0.0037, "step": 59140 }, { "epoch": 18.2, "learning_rate": 4.26421066821352e-07, "loss": 0.0027, "step": 59141 }, { "epoch": 18.2, "learning_rate": 4.262771262000431e-07, "loss": 0.0047, "step": 59142 }, { "epoch": 18.2, "learning_rate": 4.261332093477066e-07, "loss": 0.0038, "step": 59143 }, { "epoch": 18.2, "learning_rate": 4.259893162647011e-07, "loss": 0.0052, "step": 59144 }, { "epoch": 18.2, "learning_rate": 4.2584544695138287e-07, "loss": 0.0075, "step": 59145 }, { "epoch": 18.2, "learning_rate": 4.2570160140810724e-07, "loss": 0.0026, "step": 59146 }, { "epoch": 18.2, "learning_rate": 4.2555777963523506e-07, "loss": 0.0032, "step": 59147 }, { "epoch": 18.2, "learning_rate": 4.2541398163312154e-07, "loss": 0.0036, "step": 59148 }, { "epoch": 18.2, "learning_rate": 4.2527020740212424e-07, "loss": 0.0037, "step": 59149 }, { "epoch": 18.2, "learning_rate": 4.2512645694259944e-07, "loss": 0.0034, "step": 59150 }, { "epoch": 18.2, "learning_rate": 4.2498273025490365e-07, "loss": 0.0036, "step": 59151 }, { "epoch": 18.2, "learning_rate": 4.248390273393954e-07, "loss": 0.0042, "step": 59152 }, { "epoch": 18.2, "learning_rate": 4.2469534819643e-07, "loss": 0.0025, "step": 59153 }, { "epoch": 18.2, "learning_rate": 4.245516928263649e-07, "loss": 0.003, "step": 59154 }, { "epoch": 18.2, "learning_rate": 4.2440806122955644e-07, "loss": 0.0045, "step": 59155 }, { "epoch": 18.2, "learning_rate": 4.242644534063589e-07, "loss": 0.0064, "step": 59156 }, { "epoch": 18.2, "learning_rate": 4.2412086935713523e-07, "loss": 0.0028, "step": 59157 }, { "epoch": 18.2, "learning_rate": 4.2397730908223636e-07, "loss": 0.0038, "step": 59158 }, { "epoch": 18.2, "learning_rate": 4.238337725820196e-07, "loss": 0.0036, "step": 59159 }, { "epoch": 18.2, "learning_rate": 4.2369025985684264e-07, "loss": 0.0016, "step": 59160 }, { "epoch": 18.2, "learning_rate": 4.235467709070584e-07, "loss": 0.0039, "step": 59161 }, { "epoch": 18.2, "learning_rate": 4.234033057330289e-07, "loss": 0.0028, "step": 59162 }, { "epoch": 18.2, "learning_rate": 4.23259864335106e-07, "loss": 0.0025, "step": 59163 }, { "epoch": 18.2, "learning_rate": 4.2311644671364725e-07, "loss": 0.0038, "step": 59164 }, { "epoch": 18.2, "learning_rate": 4.2297305286901016e-07, "loss": 0.004, "step": 59165 }, { "epoch": 18.2, "learning_rate": 4.228296828015466e-07, "loss": 0.0038, "step": 59166 }, { "epoch": 18.2, "learning_rate": 4.226863365116163e-07, "loss": 0.0048, "step": 59167 }, { "epoch": 18.21, "learning_rate": 4.225430139995734e-07, "loss": 0.0032, "step": 59168 }, { "epoch": 18.21, "learning_rate": 4.2239971526577437e-07, "loss": 0.0025, "step": 59169 }, { "epoch": 18.21, "learning_rate": 4.2225644031057445e-07, "loss": 0.0025, "step": 59170 }, { "epoch": 18.21, "learning_rate": 4.2211318913433e-07, "loss": 0.0025, "step": 59171 }, { "epoch": 18.21, "learning_rate": 4.2196996173739626e-07, "loss": 0.0028, "step": 59172 }, { "epoch": 18.21, "learning_rate": 4.218267581201296e-07, "loss": 0.0032, "step": 59173 }, { "epoch": 18.21, "learning_rate": 4.216835782828843e-07, "loss": 0.0056, "step": 59174 }, { "epoch": 18.21, "learning_rate": 4.2154042222601666e-07, "loss": 0.0031, "step": 59175 }, { "epoch": 18.21, "learning_rate": 4.213972899498808e-07, "loss": 0.0038, "step": 59176 }, { "epoch": 18.21, "learning_rate": 4.212541814548343e-07, "loss": 0.0034, "step": 59177 }, { "epoch": 18.21, "learning_rate": 4.211110967412324e-07, "loss": 0.0023, "step": 59178 }, { "epoch": 18.21, "learning_rate": 4.2096803580942813e-07, "loss": 0.0037, "step": 59179 }, { "epoch": 18.21, "learning_rate": 4.208249986597779e-07, "loss": 0.0035, "step": 59180 }, { "epoch": 18.21, "learning_rate": 4.206819852926347e-07, "loss": 0.0028, "step": 59181 }, { "epoch": 18.21, "learning_rate": 4.2053899570835723e-07, "loss": 0.0025, "step": 59182 }, { "epoch": 18.21, "learning_rate": 4.2039602990729953e-07, "loss": 0.0072, "step": 59183 }, { "epoch": 18.21, "learning_rate": 4.202530878898159e-07, "loss": 0.0033, "step": 59184 }, { "epoch": 18.21, "learning_rate": 4.201101696562604e-07, "loss": 0.0056, "step": 59185 }, { "epoch": 18.21, "learning_rate": 4.199672752069894e-07, "loss": 0.0032, "step": 59186 }, { "epoch": 18.21, "learning_rate": 4.198244045423561e-07, "loss": 0.0033, "step": 59187 }, { "epoch": 18.21, "learning_rate": 4.1968155766271667e-07, "loss": 0.0022, "step": 59188 }, { "epoch": 18.21, "learning_rate": 4.195387345684254e-07, "loss": 0.0035, "step": 59189 }, { "epoch": 18.21, "learning_rate": 4.1939593525983644e-07, "loss": 0.0039, "step": 59190 }, { "epoch": 18.21, "learning_rate": 4.192531597373051e-07, "loss": 0.0026, "step": 59191 }, { "epoch": 18.21, "learning_rate": 4.191104080011843e-07, "loss": 0.0031, "step": 59192 }, { "epoch": 18.21, "learning_rate": 4.189676800518294e-07, "loss": 0.002, "step": 59193 }, { "epoch": 18.21, "learning_rate": 4.188249758895957e-07, "loss": 0.0033, "step": 59194 }, { "epoch": 18.21, "learning_rate": 4.186822955148362e-07, "loss": 0.0016, "step": 59195 }, { "epoch": 18.21, "learning_rate": 4.1853963892790395e-07, "loss": 0.0055, "step": 59196 }, { "epoch": 18.21, "learning_rate": 4.1839700612915535e-07, "loss": 0.0032, "step": 59197 }, { "epoch": 18.21, "learning_rate": 4.1825439711894454e-07, "loss": 0.0039, "step": 59198 }, { "epoch": 18.21, "learning_rate": 4.1811181189762684e-07, "loss": 0.0034, "step": 59199 }, { "epoch": 18.22, "learning_rate": 4.179692504655519e-07, "loss": 0.0049, "step": 59200 }, { "epoch": 18.22, "learning_rate": 4.1782671282307617e-07, "loss": 0.0058, "step": 59201 }, { "epoch": 18.22, "learning_rate": 4.176841989705538e-07, "loss": 0.003, "step": 59202 }, { "epoch": 18.22, "learning_rate": 4.1754170890833777e-07, "loss": 0.0041, "step": 59203 }, { "epoch": 18.22, "learning_rate": 4.1739924263678346e-07, "loss": 0.0031, "step": 59204 }, { "epoch": 18.22, "learning_rate": 4.172568001562427e-07, "loss": 0.0038, "step": 59205 }, { "epoch": 18.22, "learning_rate": 4.171143814670697e-07, "loss": 0.0053, "step": 59206 }, { "epoch": 18.22, "learning_rate": 4.1697198656961866e-07, "loss": 0.0014, "step": 59207 }, { "epoch": 18.22, "learning_rate": 4.168296154642426e-07, "loss": 0.0027, "step": 59208 }, { "epoch": 18.22, "learning_rate": 4.166872681512957e-07, "loss": 0.0036, "step": 59209 }, { "epoch": 18.22, "learning_rate": 4.1654494463112984e-07, "loss": 0.0033, "step": 59210 }, { "epoch": 18.22, "learning_rate": 4.164026449040981e-07, "loss": 0.0035, "step": 59211 }, { "epoch": 18.22, "learning_rate": 4.16260368970558e-07, "loss": 0.0027, "step": 59212 }, { "epoch": 18.22, "learning_rate": 4.161181168308581e-07, "loss": 0.0036, "step": 59213 }, { "epoch": 18.22, "learning_rate": 4.1597588848535264e-07, "loss": 0.004, "step": 59214 }, { "epoch": 18.22, "learning_rate": 4.158336839343957e-07, "loss": 0.0022, "step": 59215 }, { "epoch": 18.22, "learning_rate": 4.156915031783393e-07, "loss": 0.005, "step": 59216 }, { "epoch": 18.22, "learning_rate": 4.155493462175375e-07, "loss": 0.0042, "step": 59217 }, { "epoch": 18.22, "learning_rate": 4.1540721305234234e-07, "loss": 0.0021, "step": 59218 }, { "epoch": 18.22, "learning_rate": 4.152651036831079e-07, "loss": 0.0025, "step": 59219 }, { "epoch": 18.22, "learning_rate": 4.151230181101873e-07, "loss": 0.0034, "step": 59220 }, { "epoch": 18.22, "learning_rate": 4.149809563339302e-07, "loss": 0.0035, "step": 59221 }, { "epoch": 18.22, "learning_rate": 4.1483891835469194e-07, "loss": 0.0103, "step": 59222 }, { "epoch": 18.22, "learning_rate": 4.146969041728255e-07, "loss": 0.0041, "step": 59223 }, { "epoch": 18.22, "learning_rate": 4.1455491378868174e-07, "loss": 0.003, "step": 59224 }, { "epoch": 18.22, "learning_rate": 4.1441294720261373e-07, "loss": 0.0027, "step": 59225 }, { "epoch": 18.22, "learning_rate": 4.142710044149756e-07, "loss": 0.0028, "step": 59226 }, { "epoch": 18.22, "learning_rate": 4.1412908542611706e-07, "loss": 0.003, "step": 59227 }, { "epoch": 18.22, "learning_rate": 4.139871902363912e-07, "loss": 0.0027, "step": 59228 }, { "epoch": 18.22, "learning_rate": 4.138453188461522e-07, "loss": 0.0052, "step": 59229 }, { "epoch": 18.22, "learning_rate": 4.1370347125575085e-07, "loss": 0.0057, "step": 59230 }, { "epoch": 18.22, "learning_rate": 4.135616474655368e-07, "loss": 0.0023, "step": 59231 }, { "epoch": 18.22, "learning_rate": 4.134198474758666e-07, "loss": 0.0036, "step": 59232 }, { "epoch": 18.23, "learning_rate": 4.1327807128709207e-07, "loss": 0.0023, "step": 59233 }, { "epoch": 18.23, "learning_rate": 4.1313631889956296e-07, "loss": 0.004, "step": 59234 }, { "epoch": 18.23, "learning_rate": 4.1299459031363123e-07, "loss": 0.004, "step": 59235 }, { "epoch": 18.23, "learning_rate": 4.128528855296476e-07, "loss": 0.0031, "step": 59236 }, { "epoch": 18.23, "learning_rate": 4.1271120454796755e-07, "loss": 0.0023, "step": 59237 }, { "epoch": 18.23, "learning_rate": 4.125695473689406e-07, "loss": 0.0043, "step": 59238 }, { "epoch": 18.23, "learning_rate": 4.124279139929188e-07, "loss": 0.004, "step": 59239 }, { "epoch": 18.23, "learning_rate": 4.1228630442025405e-07, "loss": 0.0109, "step": 59240 }, { "epoch": 18.23, "learning_rate": 4.1214471865129724e-07, "loss": 0.0045, "step": 59241 }, { "epoch": 18.23, "learning_rate": 4.1200315668640024e-07, "loss": 0.0056, "step": 59242 }, { "epoch": 18.23, "learning_rate": 4.11861618525915e-07, "loss": 0.0017, "step": 59243 }, { "epoch": 18.23, "learning_rate": 4.1172010417019125e-07, "loss": 0.0032, "step": 59244 }, { "epoch": 18.23, "learning_rate": 4.115786136195832e-07, "loss": 0.003, "step": 59245 }, { "epoch": 18.23, "learning_rate": 4.1143714687443934e-07, "loss": 0.0038, "step": 59246 }, { "epoch": 18.23, "learning_rate": 4.1129570393511285e-07, "loss": 0.0045, "step": 59247 }, { "epoch": 18.23, "learning_rate": 4.1115428480195343e-07, "loss": 0.003, "step": 59248 }, { "epoch": 18.23, "learning_rate": 4.1101288947531404e-07, "loss": 0.0025, "step": 59249 }, { "epoch": 18.23, "learning_rate": 4.1087151795554334e-07, "loss": 0.0074, "step": 59250 }, { "epoch": 18.23, "learning_rate": 4.107301702429922e-07, "loss": 0.0035, "step": 59251 }, { "epoch": 18.23, "learning_rate": 4.1058884633801477e-07, "loss": 0.0061, "step": 59252 }, { "epoch": 18.23, "learning_rate": 4.104475462409596e-07, "loss": 0.0021, "step": 59253 }, { "epoch": 18.23, "learning_rate": 4.103062699521798e-07, "loss": 0.0031, "step": 59254 }, { "epoch": 18.23, "learning_rate": 4.1016501747202176e-07, "loss": 0.0026, "step": 59255 }, { "epoch": 18.23, "learning_rate": 4.1002378880083846e-07, "loss": 0.0035, "step": 59256 }, { "epoch": 18.23, "learning_rate": 4.0988258393898193e-07, "loss": 0.0124, "step": 59257 }, { "epoch": 18.23, "learning_rate": 4.0974140288680074e-07, "loss": 0.0058, "step": 59258 }, { "epoch": 18.23, "learning_rate": 4.0960024564464687e-07, "loss": 0.0037, "step": 59259 }, { "epoch": 18.23, "learning_rate": 4.0945911221286994e-07, "loss": 0.0034, "step": 59260 }, { "epoch": 18.23, "learning_rate": 4.093180025918209e-07, "loss": 0.0048, "step": 59261 }, { "epoch": 18.23, "learning_rate": 4.0917691678184825e-07, "loss": 0.0042, "step": 59262 }, { "epoch": 18.23, "learning_rate": 4.090358547833051e-07, "loss": 0.0016, "step": 59263 }, { "epoch": 18.23, "learning_rate": 4.0889481659654006e-07, "loss": 0.003, "step": 59264 }, { "epoch": 18.24, "learning_rate": 4.0875380222190286e-07, "loss": 0.0022, "step": 59265 }, { "epoch": 18.24, "learning_rate": 4.086128116597443e-07, "loss": 0.0025, "step": 59266 }, { "epoch": 18.24, "learning_rate": 4.0847184491041523e-07, "loss": 0.0022, "step": 59267 }, { "epoch": 18.24, "learning_rate": 4.083309019742643e-07, "loss": 0.0015, "step": 59268 }, { "epoch": 18.24, "learning_rate": 4.0818998285164223e-07, "loss": 0.0026, "step": 59269 }, { "epoch": 18.24, "learning_rate": 4.0804908754289886e-07, "loss": 0.0034, "step": 59270 }, { "epoch": 18.24, "learning_rate": 4.079082160483816e-07, "loss": 0.0037, "step": 59271 }, { "epoch": 18.24, "learning_rate": 4.0776736836844354e-07, "loss": 0.0031, "step": 59272 }, { "epoch": 18.24, "learning_rate": 4.076265445034333e-07, "loss": 0.0049, "step": 59273 }, { "epoch": 18.24, "learning_rate": 4.0748574445370057e-07, "loss": 0.0027, "step": 59274 }, { "epoch": 18.24, "learning_rate": 4.073449682195962e-07, "loss": 0.002, "step": 59275 }, { "epoch": 18.24, "learning_rate": 4.072042158014644e-07, "loss": 0.0033, "step": 59276 }, { "epoch": 18.24, "learning_rate": 4.070634871996615e-07, "loss": 0.0036, "step": 59277 }, { "epoch": 18.24, "learning_rate": 4.0692278241453276e-07, "loss": 0.004, "step": 59278 }, { "epoch": 18.24, "learning_rate": 4.067821014464279e-07, "loss": 0.0033, "step": 59279 }, { "epoch": 18.24, "learning_rate": 4.0664144429569787e-07, "loss": 0.0066, "step": 59280 }, { "epoch": 18.24, "learning_rate": 4.0650081096269e-07, "loss": 0.0018, "step": 59281 }, { "epoch": 18.24, "learning_rate": 4.0636020144775526e-07, "loss": 0.0039, "step": 59282 }, { "epoch": 18.24, "learning_rate": 4.06219615751241e-07, "loss": 0.0037, "step": 59283 }, { "epoch": 18.24, "learning_rate": 4.060790538734971e-07, "loss": 0.0038, "step": 59284 }, { "epoch": 18.24, "learning_rate": 4.059385158148732e-07, "loss": 0.0044, "step": 59285 }, { "epoch": 18.24, "learning_rate": 4.057980015757157e-07, "loss": 0.0072, "step": 59286 }, { "epoch": 18.24, "learning_rate": 4.0565751115637654e-07, "loss": 0.0016, "step": 59287 }, { "epoch": 18.24, "learning_rate": 4.055170445572043e-07, "loss": 0.0023, "step": 59288 }, { "epoch": 18.24, "learning_rate": 4.053766017785454e-07, "loss": 0.0032, "step": 59289 }, { "epoch": 18.24, "learning_rate": 4.0523618282074964e-07, "loss": 0.0045, "step": 59290 }, { "epoch": 18.24, "learning_rate": 4.0509578768416435e-07, "loss": 0.0035, "step": 59291 }, { "epoch": 18.24, "learning_rate": 4.049554163691416e-07, "loss": 0.0031, "step": 59292 }, { "epoch": 18.24, "learning_rate": 4.0481506887602773e-07, "loss": 0.004, "step": 59293 }, { "epoch": 18.24, "learning_rate": 4.0467474520517027e-07, "loss": 0.0042, "step": 59294 }, { "epoch": 18.24, "learning_rate": 4.0453444535692e-07, "loss": 0.0072, "step": 59295 }, { "epoch": 18.24, "learning_rate": 4.0439416933162226e-07, "loss": 0.0026, "step": 59296 }, { "epoch": 18.24, "learning_rate": 4.042539171296278e-07, "loss": 0.0056, "step": 59297 }, { "epoch": 18.25, "learning_rate": 4.041136887512842e-07, "loss": 0.0041, "step": 59298 }, { "epoch": 18.25, "learning_rate": 4.0397348419693894e-07, "loss": 0.0054, "step": 59299 }, { "epoch": 18.25, "learning_rate": 4.038333034669406e-07, "loss": 0.0046, "step": 59300 }, { "epoch": 18.25, "learning_rate": 4.036931465616356e-07, "loss": 0.003, "step": 59301 }, { "epoch": 18.25, "learning_rate": 4.03553013481377e-07, "loss": 0.0044, "step": 59302 }, { "epoch": 18.25, "learning_rate": 4.034129042265067e-07, "loss": 0.0036, "step": 59303 }, { "epoch": 18.25, "learning_rate": 4.032728187973756e-07, "loss": 0.0064, "step": 59304 }, { "epoch": 18.25, "learning_rate": 4.0313275719433e-07, "loss": 0.0031, "step": 59305 }, { "epoch": 18.25, "learning_rate": 4.0299271941771853e-07, "loss": 0.0036, "step": 59306 }, { "epoch": 18.25, "learning_rate": 4.028527054678888e-07, "loss": 0.0028, "step": 59307 }, { "epoch": 18.25, "learning_rate": 4.027127153451893e-07, "loss": 0.0037, "step": 59308 }, { "epoch": 18.25, "learning_rate": 4.025727490499687e-07, "loss": 0.0039, "step": 59309 }, { "epoch": 18.25, "learning_rate": 4.024328065825689e-07, "loss": 0.0041, "step": 59310 }, { "epoch": 18.25, "learning_rate": 4.02292887943343e-07, "loss": 0.0156, "step": 59311 }, { "epoch": 18.25, "learning_rate": 4.0215299313263624e-07, "loss": 0.0039, "step": 59312 }, { "epoch": 18.25, "learning_rate": 4.0201312215079614e-07, "loss": 0.0035, "step": 59313 }, { "epoch": 18.25, "learning_rate": 4.018732749981691e-07, "loss": 0.0059, "step": 59314 }, { "epoch": 18.25, "learning_rate": 4.017334516751037e-07, "loss": 0.0028, "step": 59315 }, { "epoch": 18.25, "learning_rate": 4.0159365218194633e-07, "loss": 0.0049, "step": 59316 }, { "epoch": 18.25, "learning_rate": 4.0145387651904455e-07, "loss": 0.0037, "step": 59317 }, { "epoch": 18.25, "learning_rate": 4.013141246867458e-07, "loss": 0.0033, "step": 59318 }, { "epoch": 18.25, "learning_rate": 4.011743966853954e-07, "loss": 0.0027, "step": 59319 }, { "epoch": 18.25, "learning_rate": 4.010346925153397e-07, "loss": 0.0029, "step": 59320 }, { "epoch": 18.25, "learning_rate": 4.008950121769295e-07, "loss": 0.0081, "step": 59321 }, { "epoch": 18.25, "learning_rate": 4.0075535567050793e-07, "loss": 0.003, "step": 59322 }, { "epoch": 18.25, "learning_rate": 4.006157229964247e-07, "loss": 0.0057, "step": 59323 }, { "epoch": 18.25, "learning_rate": 4.0047611415502395e-07, "loss": 0.0067, "step": 59324 }, { "epoch": 18.25, "learning_rate": 4.00336529146651e-07, "loss": 0.0029, "step": 59325 }, { "epoch": 18.25, "learning_rate": 4.0019696797165665e-07, "loss": 0.0107, "step": 59326 }, { "epoch": 18.25, "learning_rate": 4.0005743063038393e-07, "loss": 0.0039, "step": 59327 }, { "epoch": 18.25, "learning_rate": 3.999179171231815e-07, "loss": 0.0026, "step": 59328 }, { "epoch": 18.25, "learning_rate": 3.9977842745039464e-07, "loss": 0.0049, "step": 59329 }, { "epoch": 18.26, "learning_rate": 3.996389616123697e-07, "loss": 0.0041, "step": 59330 }, { "epoch": 18.26, "learning_rate": 3.994995196094531e-07, "loss": 0.0025, "step": 59331 }, { "epoch": 18.26, "learning_rate": 3.993601014419901e-07, "loss": 0.006, "step": 59332 }, { "epoch": 18.26, "learning_rate": 3.9922070711032934e-07, "loss": 0.0025, "step": 59333 }, { "epoch": 18.26, "learning_rate": 3.9908133661481386e-07, "loss": 0.0053, "step": 59334 }, { "epoch": 18.26, "learning_rate": 3.9894198995579114e-07, "loss": 0.0022, "step": 59335 }, { "epoch": 18.26, "learning_rate": 3.9880266713360874e-07, "loss": 0.0027, "step": 59336 }, { "epoch": 18.26, "learning_rate": 3.9866336814860964e-07, "loss": 0.0021, "step": 59337 }, { "epoch": 18.26, "learning_rate": 3.9852409300114135e-07, "loss": 0.0021, "step": 59338 }, { "epoch": 18.26, "learning_rate": 3.983848416915481e-07, "loss": 0.0045, "step": 59339 }, { "epoch": 18.26, "learning_rate": 3.9824561422017735e-07, "loss": 0.0041, "step": 59340 }, { "epoch": 18.26, "learning_rate": 3.981064105873744e-07, "loss": 0.0025, "step": 59341 }, { "epoch": 18.26, "learning_rate": 3.9796723079348455e-07, "loss": 0.0073, "step": 59342 }, { "epoch": 18.26, "learning_rate": 3.97828074838853e-07, "loss": 0.005, "step": 59343 }, { "epoch": 18.26, "learning_rate": 3.976889427238284e-07, "loss": 0.0041, "step": 59344 }, { "epoch": 18.26, "learning_rate": 3.975498344487494e-07, "loss": 0.0052, "step": 59345 }, { "epoch": 18.26, "learning_rate": 3.9741075001396787e-07, "loss": 0.0026, "step": 59346 }, { "epoch": 18.26, "learning_rate": 3.9727168941982584e-07, "loss": 0.0027, "step": 59347 }, { "epoch": 18.26, "learning_rate": 3.9713265266666966e-07, "loss": 0.0032, "step": 59348 }, { "epoch": 18.26, "learning_rate": 3.969936397548446e-07, "loss": 0.0025, "step": 59349 }, { "epoch": 18.26, "learning_rate": 3.968546506846949e-07, "loss": 0.0029, "step": 59350 }, { "epoch": 18.26, "learning_rate": 3.967156854565668e-07, "loss": 0.0031, "step": 59351 }, { "epoch": 18.26, "learning_rate": 3.965767440708046e-07, "loss": 0.0039, "step": 59352 }, { "epoch": 18.26, "learning_rate": 3.964378265277524e-07, "loss": 0.004, "step": 59353 }, { "epoch": 18.26, "learning_rate": 3.9629893282775777e-07, "loss": 0.0039, "step": 59354 }, { "epoch": 18.26, "learning_rate": 3.961600629711615e-07, "loss": 0.0023, "step": 59355 }, { "epoch": 18.26, "learning_rate": 3.960212169583111e-07, "loss": 0.0019, "step": 59356 }, { "epoch": 18.26, "learning_rate": 3.9588239478955294e-07, "loss": 0.0035, "step": 59357 }, { "epoch": 18.26, "learning_rate": 3.9574359646522897e-07, "loss": 0.0039, "step": 59358 }, { "epoch": 18.26, "learning_rate": 3.956048219856834e-07, "loss": 0.0039, "step": 59359 }, { "epoch": 18.26, "learning_rate": 3.9546607135126146e-07, "loss": 0.0036, "step": 59360 }, { "epoch": 18.26, "learning_rate": 3.9532734456230847e-07, "loss": 0.0025, "step": 59361 }, { "epoch": 18.26, "learning_rate": 3.9518864161916856e-07, "loss": 0.0029, "step": 59362 }, { "epoch": 18.27, "learning_rate": 3.9504996252218597e-07, "loss": 0.0033, "step": 59363 }, { "epoch": 18.27, "learning_rate": 3.949113072717048e-07, "loss": 0.0025, "step": 59364 }, { "epoch": 18.27, "learning_rate": 3.947726758680692e-07, "loss": 0.0034, "step": 59365 }, { "epoch": 18.27, "learning_rate": 3.9463406831162455e-07, "loss": 0.0036, "step": 59366 }, { "epoch": 18.27, "learning_rate": 3.944954846027127e-07, "loss": 0.0022, "step": 59367 }, { "epoch": 18.27, "learning_rate": 3.943569247416801e-07, "loss": 0.0043, "step": 59368 }, { "epoch": 18.27, "learning_rate": 3.9421838872886864e-07, "loss": 0.0025, "step": 59369 }, { "epoch": 18.27, "learning_rate": 3.9407987656462366e-07, "loss": 0.0044, "step": 59370 }, { "epoch": 18.27, "learning_rate": 3.939413882492882e-07, "loss": 0.0034, "step": 59371 }, { "epoch": 18.27, "learning_rate": 3.938029237832075e-07, "loss": 0.0036, "step": 59372 }, { "epoch": 18.27, "learning_rate": 3.9366448316672356e-07, "loss": 0.0083, "step": 59373 }, { "epoch": 18.27, "learning_rate": 3.9352606640018164e-07, "loss": 0.0023, "step": 59374 }, { "epoch": 18.27, "learning_rate": 3.933876734839237e-07, "loss": 0.0034, "step": 59375 }, { "epoch": 18.27, "learning_rate": 3.93249304418295e-07, "loss": 0.0059, "step": 59376 }, { "epoch": 18.27, "learning_rate": 3.931109592036386e-07, "loss": 0.0024, "step": 59377 }, { "epoch": 18.27, "learning_rate": 3.929726378402987e-07, "loss": 0.004, "step": 59378 }, { "epoch": 18.27, "learning_rate": 3.928343403286161e-07, "loss": 0.0025, "step": 59379 }, { "epoch": 18.27, "learning_rate": 3.92696066668935e-07, "loss": 0.0041, "step": 59380 }, { "epoch": 18.27, "learning_rate": 3.925578168616007e-07, "loss": 0.0061, "step": 59381 }, { "epoch": 18.27, "learning_rate": 3.9241959090695503e-07, "loss": 0.0046, "step": 59382 }, { "epoch": 18.27, "learning_rate": 3.9228138880534227e-07, "loss": 0.0016, "step": 59383 }, { "epoch": 18.27, "learning_rate": 3.921432105571044e-07, "loss": 0.0024, "step": 59384 }, { "epoch": 18.27, "learning_rate": 3.920050561625843e-07, "loss": 0.0029, "step": 59385 }, { "epoch": 18.27, "learning_rate": 3.9186692562212526e-07, "loss": 0.0038, "step": 59386 }, { "epoch": 18.27, "learning_rate": 3.917288189360713e-07, "loss": 0.0042, "step": 59387 }, { "epoch": 18.27, "learning_rate": 3.9159073610476326e-07, "loss": 0.0033, "step": 59388 }, { "epoch": 18.27, "learning_rate": 3.914526771285454e-07, "loss": 0.0037, "step": 59389 }, { "epoch": 18.27, "learning_rate": 3.913146420077596e-07, "loss": 0.0053, "step": 59390 }, { "epoch": 18.27, "learning_rate": 3.9117663074275116e-07, "loss": 0.0038, "step": 59391 }, { "epoch": 18.27, "learning_rate": 3.9103864333385866e-07, "loss": 0.0041, "step": 59392 }, { "epoch": 18.27, "learning_rate": 3.9090067978142745e-07, "loss": 0.0034, "step": 59393 }, { "epoch": 18.27, "learning_rate": 3.907627400857994e-07, "loss": 0.0033, "step": 59394 }, { "epoch": 18.28, "learning_rate": 3.906248242473154e-07, "loss": 0.0036, "step": 59395 }, { "epoch": 18.28, "learning_rate": 3.9048693226632075e-07, "loss": 0.0045, "step": 59396 }, { "epoch": 18.28, "learning_rate": 3.903490641431573e-07, "loss": 0.0018, "step": 59397 }, { "epoch": 18.28, "learning_rate": 3.9021121987816493e-07, "loss": 0.003, "step": 59398 }, { "epoch": 18.28, "learning_rate": 3.900733994716899e-07, "loss": 0.0021, "step": 59399 }, { "epoch": 18.28, "learning_rate": 3.899356029240686e-07, "loss": 0.0039, "step": 59400 }, { "epoch": 18.28, "learning_rate": 3.8979783023564756e-07, "loss": 0.0029, "step": 59401 }, { "epoch": 18.28, "learning_rate": 3.896600814067675e-07, "loss": 0.003, "step": 59402 }, { "epoch": 18.28, "learning_rate": 3.895223564377715e-07, "loss": 0.0027, "step": 59403 }, { "epoch": 18.28, "learning_rate": 3.8938465532899925e-07, "loss": 0.0025, "step": 59404 }, { "epoch": 18.28, "learning_rate": 3.8924697808079505e-07, "loss": 0.0027, "step": 59405 }, { "epoch": 18.28, "learning_rate": 3.891093246934996e-07, "loss": 0.0029, "step": 59406 }, { "epoch": 18.28, "learning_rate": 3.889716951674549e-07, "loss": 0.0034, "step": 59407 }, { "epoch": 18.28, "learning_rate": 3.888340895030018e-07, "loss": 0.0029, "step": 59408 }, { "epoch": 18.28, "learning_rate": 3.886965077004823e-07, "loss": 0.0036, "step": 59409 }, { "epoch": 18.28, "learning_rate": 3.885589497602371e-07, "loss": 0.0026, "step": 59410 }, { "epoch": 18.28, "learning_rate": 3.884214156826105e-07, "loss": 0.0017, "step": 59411 }, { "epoch": 18.28, "learning_rate": 3.8828390546794437e-07, "loss": 0.0026, "step": 59412 }, { "epoch": 18.28, "learning_rate": 3.8814641911657516e-07, "loss": 0.0062, "step": 59413 }, { "epoch": 18.28, "learning_rate": 3.880089566288481e-07, "loss": 0.0033, "step": 59414 }, { "epoch": 18.28, "learning_rate": 3.8787151800510183e-07, "loss": 0.0044, "step": 59415 }, { "epoch": 18.28, "learning_rate": 3.8773410324568053e-07, "loss": 0.0054, "step": 59416 }, { "epoch": 18.28, "learning_rate": 3.87596712350925e-07, "loss": 0.0017, "step": 59417 }, { "epoch": 18.28, "learning_rate": 3.874593453211739e-07, "loss": 0.0042, "step": 59418 }, { "epoch": 18.28, "learning_rate": 3.873220021567714e-07, "loss": 0.0029, "step": 59419 }, { "epoch": 18.28, "learning_rate": 3.87184682858055e-07, "loss": 0.004, "step": 59420 }, { "epoch": 18.28, "learning_rate": 3.8704738742536884e-07, "loss": 0.0046, "step": 59421 }, { "epoch": 18.28, "learning_rate": 3.8691011585905267e-07, "loss": 0.003, "step": 59422 }, { "epoch": 18.28, "learning_rate": 3.8677286815944627e-07, "loss": 0.0038, "step": 59423 }, { "epoch": 18.28, "learning_rate": 3.866356443268915e-07, "loss": 0.0038, "step": 59424 }, { "epoch": 18.28, "learning_rate": 3.864984443617281e-07, "loss": 0.0034, "step": 59425 }, { "epoch": 18.28, "learning_rate": 3.8636126826429807e-07, "loss": 0.0024, "step": 59426 }, { "epoch": 18.28, "learning_rate": 3.8622411603494114e-07, "loss": 0.003, "step": 59427 }, { "epoch": 18.29, "learning_rate": 3.860869876739981e-07, "loss": 0.003, "step": 59428 }, { "epoch": 18.29, "learning_rate": 3.859498831818087e-07, "loss": 0.0035, "step": 59429 }, { "epoch": 18.29, "learning_rate": 3.8581280255871266e-07, "loss": 0.0051, "step": 59430 }, { "epoch": 18.29, "learning_rate": 3.8567574580505306e-07, "loss": 0.0042, "step": 59431 }, { "epoch": 18.29, "learning_rate": 3.8553871292116854e-07, "loss": 0.0039, "step": 59432 }, { "epoch": 18.29, "learning_rate": 3.8540170390740097e-07, "loss": 0.0035, "step": 59433 }, { "epoch": 18.29, "learning_rate": 3.852647187640879e-07, "loss": 0.0022, "step": 59434 }, { "epoch": 18.29, "learning_rate": 3.8512775749156903e-07, "loss": 0.0044, "step": 59435 }, { "epoch": 18.29, "learning_rate": 3.849908200901864e-07, "loss": 0.0033, "step": 59436 }, { "epoch": 18.29, "learning_rate": 3.8485390656027967e-07, "loss": 0.0026, "step": 59437 }, { "epoch": 18.29, "learning_rate": 3.8471701690218966e-07, "loss": 0.0027, "step": 59438 }, { "epoch": 18.29, "learning_rate": 3.8458015111625393e-07, "loss": 0.004, "step": 59439 }, { "epoch": 18.29, "learning_rate": 3.8444330920281326e-07, "loss": 0.0045, "step": 59440 }, { "epoch": 18.29, "learning_rate": 3.843064911622085e-07, "loss": 0.0017, "step": 59441 }, { "epoch": 18.29, "learning_rate": 3.8416969699477835e-07, "loss": 0.0041, "step": 59442 }, { "epoch": 18.29, "learning_rate": 3.840329267008613e-07, "loss": 0.0037, "step": 59443 }, { "epoch": 18.29, "learning_rate": 3.8389618028079947e-07, "loss": 0.0036, "step": 59444 }, { "epoch": 18.29, "learning_rate": 3.8375945773492907e-07, "loss": 0.0024, "step": 59445 }, { "epoch": 18.29, "learning_rate": 3.836227590635944e-07, "loss": 0.0029, "step": 59446 }, { "epoch": 18.29, "learning_rate": 3.834860842671306e-07, "loss": 0.0032, "step": 59447 }, { "epoch": 18.29, "learning_rate": 3.833494333458776e-07, "loss": 0.0045, "step": 59448 }, { "epoch": 18.29, "learning_rate": 3.832128063001761e-07, "loss": 0.0021, "step": 59449 }, { "epoch": 18.29, "learning_rate": 3.8307620313036256e-07, "loss": 0.0024, "step": 59450 }, { "epoch": 18.29, "learning_rate": 3.8293962383677995e-07, "loss": 0.002, "step": 59451 }, { "epoch": 18.29, "learning_rate": 3.828030684197659e-07, "loss": 0.0036, "step": 59452 }, { "epoch": 18.29, "learning_rate": 3.826665368796589e-07, "loss": 0.003, "step": 59453 }, { "epoch": 18.29, "learning_rate": 3.8253002921679996e-07, "loss": 0.0028, "step": 59454 }, { "epoch": 18.29, "learning_rate": 3.8239354543152307e-07, "loss": 0.0027, "step": 59455 }, { "epoch": 18.29, "learning_rate": 3.822570855241714e-07, "loss": 0.0038, "step": 59456 }, { "epoch": 18.29, "learning_rate": 3.8212064949508245e-07, "loss": 0.0072, "step": 59457 }, { "epoch": 18.29, "learning_rate": 3.8198423734459476e-07, "loss": 0.0031, "step": 59458 }, { "epoch": 18.29, "learning_rate": 3.8184784907304704e-07, "loss": 0.0043, "step": 59459 }, { "epoch": 18.3, "learning_rate": 3.81711484680779e-07, "loss": 0.0029, "step": 59460 }, { "epoch": 18.3, "learning_rate": 3.815751441681281e-07, "loss": 0.0032, "step": 59461 }, { "epoch": 18.3, "learning_rate": 3.81438827535433e-07, "loss": 0.0048, "step": 59462 }, { "epoch": 18.3, "learning_rate": 3.8130253478303234e-07, "loss": 0.0029, "step": 59463 }, { "epoch": 18.3, "learning_rate": 3.811662659112647e-07, "loss": 0.0018, "step": 59464 }, { "epoch": 18.3, "learning_rate": 3.810300209204665e-07, "loss": 0.0022, "step": 59465 }, { "epoch": 18.3, "learning_rate": 3.808937998109785e-07, "loss": 0.0057, "step": 59466 }, { "epoch": 18.3, "learning_rate": 3.807576025831394e-07, "loss": 0.0032, "step": 59467 }, { "epoch": 18.3, "learning_rate": 3.806214292372845e-07, "loss": 0.0042, "step": 59468 }, { "epoch": 18.3, "learning_rate": 3.804852797737535e-07, "loss": 0.0029, "step": 59469 }, { "epoch": 18.3, "learning_rate": 3.803491541928828e-07, "loss": 0.006, "step": 59470 }, { "epoch": 18.3, "learning_rate": 3.8021305249501314e-07, "loss": 0.0032, "step": 59471 }, { "epoch": 18.3, "learning_rate": 3.800769746804811e-07, "loss": 0.0022, "step": 59472 }, { "epoch": 18.3, "learning_rate": 3.7994092074962405e-07, "loss": 0.0035, "step": 59473 }, { "epoch": 18.3, "learning_rate": 3.798048907027807e-07, "loss": 0.0053, "step": 59474 }, { "epoch": 18.3, "learning_rate": 3.796688845402885e-07, "loss": 0.0046, "step": 59475 }, { "epoch": 18.3, "learning_rate": 3.7953290226248387e-07, "loss": 0.0029, "step": 59476 }, { "epoch": 18.3, "learning_rate": 3.793969438697065e-07, "loss": 0.0034, "step": 59477 }, { "epoch": 18.3, "learning_rate": 3.792610093622928e-07, "loss": 0.0024, "step": 59478 }, { "epoch": 18.3, "learning_rate": 3.791250987405792e-07, "loss": 0.0036, "step": 59479 }, { "epoch": 18.3, "learning_rate": 3.7898921200490436e-07, "loss": 0.0032, "step": 59480 }, { "epoch": 18.3, "learning_rate": 3.788533491556079e-07, "loss": 0.0033, "step": 59481 }, { "epoch": 18.3, "learning_rate": 3.7871751019302294e-07, "loss": 0.004, "step": 59482 }, { "epoch": 18.3, "learning_rate": 3.785816951174881e-07, "loss": 0.0037, "step": 59483 }, { "epoch": 18.3, "learning_rate": 3.78445903929342e-07, "loss": 0.0025, "step": 59484 }, { "epoch": 18.3, "learning_rate": 3.783101366289199e-07, "loss": 0.004, "step": 59485 }, { "epoch": 18.3, "learning_rate": 3.7817439321656044e-07, "loss": 0.0028, "step": 59486 }, { "epoch": 18.3, "learning_rate": 3.7803867369260007e-07, "loss": 0.0047, "step": 59487 }, { "epoch": 18.3, "learning_rate": 3.779029780573773e-07, "loss": 0.0034, "step": 59488 }, { "epoch": 18.3, "learning_rate": 3.7776730631122526e-07, "loss": 0.0017, "step": 59489 }, { "epoch": 18.3, "learning_rate": 3.776316584544826e-07, "loss": 0.0043, "step": 59490 }, { "epoch": 18.3, "learning_rate": 3.7749603448748673e-07, "loss": 0.0021, "step": 59491 }, { "epoch": 18.3, "learning_rate": 3.7736043441057526e-07, "loss": 0.0029, "step": 59492 }, { "epoch": 18.31, "learning_rate": 3.772248582240823e-07, "loss": 0.0021, "step": 59493 }, { "epoch": 18.31, "learning_rate": 3.770893059283465e-07, "loss": 0.0066, "step": 59494 }, { "epoch": 18.31, "learning_rate": 3.769537775237042e-07, "loss": 0.0044, "step": 59495 }, { "epoch": 18.31, "learning_rate": 3.768182730104908e-07, "loss": 0.0022, "step": 59496 }, { "epoch": 18.31, "learning_rate": 3.766827923890437e-07, "loss": 0.0053, "step": 59497 }, { "epoch": 18.31, "learning_rate": 3.7654733565969826e-07, "loss": 0.0047, "step": 59498 }, { "epoch": 18.31, "learning_rate": 3.7641190282279196e-07, "loss": 0.003, "step": 59499 }, { "epoch": 18.31, "learning_rate": 3.7627649387865895e-07, "loss": 0.0037, "step": 59500 }, { "epoch": 18.31, "learning_rate": 3.761411088276379e-07, "loss": 0.0022, "step": 59501 }, { "epoch": 18.31, "learning_rate": 3.7600574767006513e-07, "loss": 0.0041, "step": 59502 }, { "epoch": 18.31, "learning_rate": 3.758704104062749e-07, "loss": 0.0036, "step": 59503 }, { "epoch": 18.31, "learning_rate": 3.757350970366036e-07, "loss": 0.0035, "step": 59504 }, { "epoch": 18.31, "learning_rate": 3.755998075613865e-07, "loss": 0.0034, "step": 59505 }, { "epoch": 18.31, "learning_rate": 3.7546454198096104e-07, "loss": 0.0032, "step": 59506 }, { "epoch": 18.31, "learning_rate": 3.7532930029566264e-07, "loss": 0.0036, "step": 59507 }, { "epoch": 18.31, "learning_rate": 3.751940825058265e-07, "loss": 0.0022, "step": 59508 }, { "epoch": 18.31, "learning_rate": 3.7505888861179116e-07, "loss": 0.0026, "step": 59509 }, { "epoch": 18.31, "learning_rate": 3.7492371861388656e-07, "loss": 0.0032, "step": 59510 }, { "epoch": 18.31, "learning_rate": 3.747885725124523e-07, "loss": 0.0041, "step": 59511 }, { "epoch": 18.31, "learning_rate": 3.7465345030782364e-07, "loss": 0.0053, "step": 59512 }, { "epoch": 18.31, "learning_rate": 3.7451835200033484e-07, "loss": 0.0025, "step": 59513 }, { "epoch": 18.31, "learning_rate": 3.7438327759032333e-07, "loss": 0.0039, "step": 59514 }, { "epoch": 18.31, "learning_rate": 3.7424822707812225e-07, "loss": 0.0047, "step": 59515 }, { "epoch": 18.31, "learning_rate": 3.74113200464068e-07, "loss": 0.0028, "step": 59516 }, { "epoch": 18.31, "learning_rate": 3.7397819774849574e-07, "loss": 0.0032, "step": 59517 }, { "epoch": 18.31, "learning_rate": 3.738432189317409e-07, "loss": 0.0028, "step": 59518 }, { "epoch": 18.31, "learning_rate": 3.7370826401413765e-07, "loss": 0.0033, "step": 59519 }, { "epoch": 18.31, "learning_rate": 3.7357333299602005e-07, "loss": 0.0028, "step": 59520 }, { "epoch": 18.31, "learning_rate": 3.734384258777268e-07, "loss": 0.0047, "step": 59521 }, { "epoch": 18.31, "learning_rate": 3.73303542659591e-07, "loss": 0.0035, "step": 59522 }, { "epoch": 18.31, "learning_rate": 3.7316868334194676e-07, "loss": 0.0022, "step": 59523 }, { "epoch": 18.31, "learning_rate": 3.730338479251283e-07, "loss": 0.0061, "step": 59524 }, { "epoch": 18.32, "learning_rate": 3.728990364094709e-07, "loss": 0.0033, "step": 59525 }, { "epoch": 18.32, "learning_rate": 3.72764248795312e-07, "loss": 0.0062, "step": 59526 }, { "epoch": 18.32, "learning_rate": 3.7262948508298255e-07, "loss": 0.0043, "step": 59527 }, { "epoch": 18.32, "learning_rate": 3.724947452728189e-07, "loss": 0.003, "step": 59528 }, { "epoch": 18.32, "learning_rate": 3.7236002936515636e-07, "loss": 0.0033, "step": 59529 }, { "epoch": 18.32, "learning_rate": 3.7222533736032683e-07, "loss": 0.0054, "step": 59530 }, { "epoch": 18.32, "learning_rate": 3.7209066925866786e-07, "loss": 0.0057, "step": 59531 }, { "epoch": 18.32, "learning_rate": 3.719560250605103e-07, "loss": 0.0049, "step": 59532 }, { "epoch": 18.32, "learning_rate": 3.7182140476619166e-07, "loss": 0.0023, "step": 59533 }, { "epoch": 18.32, "learning_rate": 3.7168680837604497e-07, "loss": 0.004, "step": 59534 }, { "epoch": 18.32, "learning_rate": 3.715522358904022e-07, "loss": 0.0038, "step": 59535 }, { "epoch": 18.32, "learning_rate": 3.7141768730960203e-07, "loss": 0.0033, "step": 59536 }, { "epoch": 18.32, "learning_rate": 3.712831626339752e-07, "loss": 0.0041, "step": 59537 }, { "epoch": 18.32, "learning_rate": 3.71148661863856e-07, "loss": 0.0035, "step": 59538 }, { "epoch": 18.32, "learning_rate": 3.710141849995785e-07, "loss": 0.0047, "step": 59539 }, { "epoch": 18.32, "learning_rate": 3.7087973204147585e-07, "loss": 0.0031, "step": 59540 }, { "epoch": 18.32, "learning_rate": 3.7074530298988443e-07, "loss": 0.0044, "step": 59541 }, { "epoch": 18.32, "learning_rate": 3.7061089784513616e-07, "loss": 0.0054, "step": 59542 }, { "epoch": 18.32, "learning_rate": 3.7047651660756745e-07, "loss": 0.0045, "step": 59543 }, { "epoch": 18.32, "learning_rate": 3.7034215927750696e-07, "loss": 0.0039, "step": 59544 }, { "epoch": 18.32, "learning_rate": 3.702078258552899e-07, "loss": 0.0041, "step": 59545 }, { "epoch": 18.32, "learning_rate": 3.7007351634125165e-07, "loss": 0.0034, "step": 59546 }, { "epoch": 18.32, "learning_rate": 3.6993923073572525e-07, "loss": 0.0089, "step": 59547 }, { "epoch": 18.32, "learning_rate": 3.6980496903904373e-07, "loss": 0.0033, "step": 59548 }, { "epoch": 18.32, "learning_rate": 3.696707312515391e-07, "loss": 0.004, "step": 59549 }, { "epoch": 18.32, "learning_rate": 3.695365173735466e-07, "loss": 0.0051, "step": 59550 }, { "epoch": 18.32, "learning_rate": 3.6940232740539927e-07, "loss": 0.0035, "step": 59551 }, { "epoch": 18.32, "learning_rate": 3.6926816134742806e-07, "loss": 0.0031, "step": 59552 }, { "epoch": 18.32, "learning_rate": 3.691340191999693e-07, "loss": 0.002, "step": 59553 }, { "epoch": 18.32, "learning_rate": 3.689999009633527e-07, "loss": 0.0035, "step": 59554 }, { "epoch": 18.32, "learning_rate": 3.688658066379136e-07, "loss": 0.0043, "step": 59555 }, { "epoch": 18.32, "learning_rate": 3.6873173622398393e-07, "loss": 0.0024, "step": 59556 }, { "epoch": 18.32, "learning_rate": 3.68597689721899e-07, "loss": 0.0043, "step": 59557 }, { "epoch": 18.33, "learning_rate": 3.6846366713198745e-07, "loss": 0.0052, "step": 59558 }, { "epoch": 18.33, "learning_rate": 3.683296684545845e-07, "loss": 0.0031, "step": 59559 }, { "epoch": 18.33, "learning_rate": 3.6819569369002107e-07, "loss": 0.0062, "step": 59560 }, { "epoch": 18.33, "learning_rate": 3.680617428386324e-07, "loss": 0.0066, "step": 59561 }, { "epoch": 18.33, "learning_rate": 3.6792781590074934e-07, "loss": 0.0041, "step": 59562 }, { "epoch": 18.33, "learning_rate": 3.67793912876705e-07, "loss": 0.0052, "step": 59563 }, { "epoch": 18.33, "learning_rate": 3.6766003376683233e-07, "loss": 0.0056, "step": 59564 }, { "epoch": 18.33, "learning_rate": 3.675261785714612e-07, "loss": 0.0048, "step": 59565 }, { "epoch": 18.33, "learning_rate": 3.6739234729092686e-07, "loss": 0.0047, "step": 59566 }, { "epoch": 18.33, "learning_rate": 3.6725853992556017e-07, "loss": 0.0013, "step": 59567 }, { "epoch": 18.33, "learning_rate": 3.6712475647569414e-07, "loss": 0.0028, "step": 59568 }, { "epoch": 18.33, "learning_rate": 3.6699099694166076e-07, "loss": 0.0041, "step": 59569 }, { "epoch": 18.33, "learning_rate": 3.6685726132379085e-07, "loss": 0.0034, "step": 59570 }, { "epoch": 18.33, "learning_rate": 3.667235496224175e-07, "loss": 0.0033, "step": 59571 }, { "epoch": 18.33, "learning_rate": 3.665898618378727e-07, "loss": 0.0038, "step": 59572 }, { "epoch": 18.33, "learning_rate": 3.6645619797048835e-07, "loss": 0.003, "step": 59573 }, { "epoch": 18.33, "learning_rate": 3.6632255802059537e-07, "loss": 0.0038, "step": 59574 }, { "epoch": 18.33, "learning_rate": 3.661889419885256e-07, "loss": 0.0034, "step": 59575 }, { "epoch": 18.33, "learning_rate": 3.6605534987461335e-07, "loss": 0.0027, "step": 59576 }, { "epoch": 18.33, "learning_rate": 3.659217816791882e-07, "loss": 0.0038, "step": 59577 }, { "epoch": 18.33, "learning_rate": 3.6578823740258227e-07, "loss": 0.0042, "step": 59578 }, { "epoch": 18.33, "learning_rate": 3.656547170451263e-07, "loss": 0.0037, "step": 59579 }, { "epoch": 18.33, "learning_rate": 3.6552122060715124e-07, "loss": 0.0045, "step": 59580 }, { "epoch": 18.33, "learning_rate": 3.6538774808899e-07, "loss": 0.0026, "step": 59581 }, { "epoch": 18.33, "learning_rate": 3.652542994909747e-07, "loss": 0.0037, "step": 59582 }, { "epoch": 18.33, "learning_rate": 3.6512087481343493e-07, "loss": 0.004, "step": 59583 }, { "epoch": 18.33, "learning_rate": 3.6498747405670275e-07, "loss": 0.0031, "step": 59584 }, { "epoch": 18.33, "learning_rate": 3.64854097221109e-07, "loss": 0.003, "step": 59585 }, { "epoch": 18.33, "learning_rate": 3.6472074430698554e-07, "loss": 0.0033, "step": 59586 }, { "epoch": 18.33, "learning_rate": 3.6458741531466226e-07, "loss": 0.0092, "step": 59587 }, { "epoch": 18.33, "learning_rate": 3.6445411024447096e-07, "loss": 0.0033, "step": 59588 }, { "epoch": 18.33, "learning_rate": 3.643208290967415e-07, "loss": 0.0036, "step": 59589 }, { "epoch": 18.34, "learning_rate": 3.6418757187180576e-07, "loss": 0.003, "step": 59590 }, { "epoch": 18.34, "learning_rate": 3.6405433856999684e-07, "loss": 0.0042, "step": 59591 }, { "epoch": 18.34, "learning_rate": 3.639211291916411e-07, "loss": 0.0026, "step": 59592 }, { "epoch": 18.34, "learning_rate": 3.6378794373707283e-07, "loss": 0.0064, "step": 59593 }, { "epoch": 18.34, "learning_rate": 3.6365478220662054e-07, "loss": 0.0053, "step": 59594 }, { "epoch": 18.34, "learning_rate": 3.63521644600614e-07, "loss": 0.0034, "step": 59595 }, { "epoch": 18.34, "learning_rate": 3.6338853091938633e-07, "loss": 0.0034, "step": 59596 }, { "epoch": 18.34, "learning_rate": 3.6325544116326827e-07, "loss": 0.0032, "step": 59597 }, { "epoch": 18.34, "learning_rate": 3.631223753325874e-07, "loss": 0.0033, "step": 59598 }, { "epoch": 18.34, "learning_rate": 3.6298933342767796e-07, "loss": 0.0042, "step": 59599 }, { "epoch": 18.34, "learning_rate": 3.6285631544886513e-07, "loss": 0.004, "step": 59600 }, { "epoch": 18.34, "learning_rate": 3.6272332139648314e-07, "loss": 0.0041, "step": 59601 }, { "epoch": 18.34, "learning_rate": 3.6259035127086173e-07, "loss": 0.0032, "step": 59602 }, { "epoch": 18.34, "learning_rate": 3.624574050723295e-07, "loss": 0.0036, "step": 59603 }, { "epoch": 18.34, "learning_rate": 3.6232448280121734e-07, "loss": 0.0049, "step": 59604 }, { "epoch": 18.34, "learning_rate": 3.6219158445785494e-07, "loss": 0.0047, "step": 59605 }, { "epoch": 18.34, "learning_rate": 3.620587100425732e-07, "loss": 0.0019, "step": 59606 }, { "epoch": 18.34, "learning_rate": 3.619258595557018e-07, "loss": 0.0035, "step": 59607 }, { "epoch": 18.34, "learning_rate": 3.617930329975694e-07, "loss": 0.0053, "step": 59608 }, { "epoch": 18.34, "learning_rate": 3.6166023036850683e-07, "loss": 0.0049, "step": 59609 }, { "epoch": 18.34, "learning_rate": 3.615274516688416e-07, "loss": 0.0045, "step": 59610 }, { "epoch": 18.34, "learning_rate": 3.613946968989068e-07, "loss": 0.0033, "step": 59611 }, { "epoch": 18.34, "learning_rate": 3.612619660590322e-07, "loss": 0.003, "step": 59612 }, { "epoch": 18.34, "learning_rate": 3.611292591495441e-07, "loss": 0.003, "step": 59613 }, { "epoch": 18.34, "learning_rate": 3.609965761707734e-07, "loss": 0.0019, "step": 59614 }, { "epoch": 18.34, "learning_rate": 3.608639171230488e-07, "loss": 0.0034, "step": 59615 }, { "epoch": 18.34, "learning_rate": 3.6073128200670106e-07, "loss": 0.0047, "step": 59616 }, { "epoch": 18.34, "learning_rate": 3.605986708220599e-07, "loss": 0.0028, "step": 59617 }, { "epoch": 18.34, "learning_rate": 3.604660835694518e-07, "loss": 0.0044, "step": 59618 }, { "epoch": 18.34, "learning_rate": 3.6033352024920975e-07, "loss": 0.0067, "step": 59619 }, { "epoch": 18.34, "learning_rate": 3.602009808616591e-07, "loss": 0.0026, "step": 59620 }, { "epoch": 18.34, "learning_rate": 3.600684654071318e-07, "loss": 0.0031, "step": 59621 }, { "epoch": 18.34, "learning_rate": 3.5993597388595425e-07, "loss": 0.0022, "step": 59622 }, { "epoch": 18.35, "learning_rate": 3.598035062984584e-07, "loss": 0.002, "step": 59623 }, { "epoch": 18.35, "learning_rate": 3.596710626449706e-07, "loss": 0.0046, "step": 59624 }, { "epoch": 18.35, "learning_rate": 3.595386429258207e-07, "loss": 0.0025, "step": 59625 }, { "epoch": 18.35, "learning_rate": 3.594062471413373e-07, "loss": 0.0025, "step": 59626 }, { "epoch": 18.35, "learning_rate": 3.59273875291849e-07, "loss": 0.004, "step": 59627 }, { "epoch": 18.35, "learning_rate": 3.591415273776855e-07, "loss": 0.0048, "step": 59628 }, { "epoch": 18.35, "learning_rate": 3.5900920339917323e-07, "loss": 0.0033, "step": 59629 }, { "epoch": 18.35, "learning_rate": 3.5887690335664194e-07, "loss": 0.0046, "step": 59630 }, { "epoch": 18.35, "learning_rate": 3.5874462725042026e-07, "loss": 0.0054, "step": 59631 }, { "epoch": 18.35, "learning_rate": 3.586123750808379e-07, "loss": 0.0029, "step": 59632 }, { "epoch": 18.35, "learning_rate": 3.584801468482213e-07, "loss": 0.0038, "step": 59633 }, { "epoch": 18.35, "learning_rate": 3.583479425528991e-07, "loss": 0.0031, "step": 59634 }, { "epoch": 18.35, "learning_rate": 3.582157621951987e-07, "loss": 0.0029, "step": 59635 }, { "epoch": 18.35, "learning_rate": 3.5808360577544997e-07, "loss": 0.0039, "step": 59636 }, { "epoch": 18.35, "learning_rate": 3.579514732939804e-07, "loss": 0.0033, "step": 59637 }, { "epoch": 18.35, "learning_rate": 3.5781936475111746e-07, "loss": 0.0071, "step": 59638 }, { "epoch": 18.35, "learning_rate": 3.5768728014718976e-07, "loss": 0.0036, "step": 59639 }, { "epoch": 18.35, "learning_rate": 3.575552194825249e-07, "loss": 0.0025, "step": 59640 }, { "epoch": 18.35, "learning_rate": 3.5742318275745147e-07, "loss": 0.0045, "step": 59641 }, { "epoch": 18.35, "learning_rate": 3.5729116997229583e-07, "loss": 0.0039, "step": 59642 }, { "epoch": 18.35, "learning_rate": 3.5715918112738777e-07, "loss": 0.003, "step": 59643 }, { "epoch": 18.35, "learning_rate": 3.570272162230526e-07, "loss": 0.0043, "step": 59644 }, { "epoch": 18.35, "learning_rate": 3.5689527525961887e-07, "loss": 0.0043, "step": 59645 }, { "epoch": 18.35, "learning_rate": 3.567633582374164e-07, "loss": 0.0042, "step": 59646 }, { "epoch": 18.35, "learning_rate": 3.566314651567704e-07, "loss": 0.0042, "step": 59647 }, { "epoch": 18.35, "learning_rate": 3.564995960180073e-07, "loss": 0.0029, "step": 59648 }, { "epoch": 18.35, "learning_rate": 3.563677508214569e-07, "loss": 0.0039, "step": 59649 }, { "epoch": 18.35, "learning_rate": 3.5623592956744444e-07, "loss": 0.0042, "step": 59650 }, { "epoch": 18.35, "learning_rate": 3.561041322562997e-07, "loss": 0.0032, "step": 59651 }, { "epoch": 18.35, "learning_rate": 3.559723588883479e-07, "loss": 0.0052, "step": 59652 }, { "epoch": 18.35, "learning_rate": 3.5584060946391664e-07, "loss": 0.003, "step": 59653 }, { "epoch": 18.35, "learning_rate": 3.557088839833345e-07, "loss": 0.0034, "step": 59654 }, { "epoch": 18.36, "learning_rate": 3.555771824469245e-07, "loss": 0.0043, "step": 59655 }, { "epoch": 18.36, "learning_rate": 3.5544550485501763e-07, "loss": 0.0029, "step": 59656 }, { "epoch": 18.36, "learning_rate": 3.5531385120794015e-07, "loss": 0.0031, "step": 59657 }, { "epoch": 18.36, "learning_rate": 3.551822215060174e-07, "loss": 0.005, "step": 59658 }, { "epoch": 18.36, "learning_rate": 3.550506157495781e-07, "loss": 0.0038, "step": 59659 }, { "epoch": 18.36, "learning_rate": 3.5491903393894745e-07, "loss": 0.0032, "step": 59660 }, { "epoch": 18.36, "learning_rate": 3.5478747607445183e-07, "loss": 0.0055, "step": 59661 }, { "epoch": 18.36, "learning_rate": 3.5465594215641996e-07, "loss": 0.0034, "step": 59662 }, { "epoch": 18.36, "learning_rate": 3.5452443218517597e-07, "loss": 0.0031, "step": 59663 }, { "epoch": 18.36, "learning_rate": 3.5439294616104736e-07, "loss": 0.0045, "step": 59664 }, { "epoch": 18.36, "learning_rate": 3.5426148408436054e-07, "loss": 0.0028, "step": 59665 }, { "epoch": 18.36, "learning_rate": 3.54130045955442e-07, "loss": 0.0033, "step": 59666 }, { "epoch": 18.36, "learning_rate": 3.5399863177462024e-07, "loss": 0.0033, "step": 59667 }, { "epoch": 18.36, "learning_rate": 3.5386724154221843e-07, "loss": 0.0027, "step": 59668 }, { "epoch": 18.36, "learning_rate": 3.537358752585629e-07, "loss": 0.0033, "step": 59669 }, { "epoch": 18.36, "learning_rate": 3.5360453292398014e-07, "loss": 0.0041, "step": 59670 }, { "epoch": 18.36, "learning_rate": 3.5347321453879757e-07, "loss": 0.0046, "step": 59671 }, { "epoch": 18.36, "learning_rate": 3.5334192010334055e-07, "loss": 0.0035, "step": 59672 }, { "epoch": 18.36, "learning_rate": 3.532106496179344e-07, "loss": 0.0024, "step": 59673 }, { "epoch": 18.36, "learning_rate": 3.5307940308290545e-07, "loss": 0.003, "step": 59674 }, { "epoch": 18.36, "learning_rate": 3.5294818049858017e-07, "loss": 0.0018, "step": 59675 }, { "epoch": 18.36, "learning_rate": 3.5281698186528377e-07, "loss": 0.0047, "step": 59676 }, { "epoch": 18.36, "learning_rate": 3.5268580718334166e-07, "loss": 0.0039, "step": 59677 }, { "epoch": 18.36, "learning_rate": 3.5255465645308017e-07, "loss": 0.0018, "step": 59678 }, { "epoch": 18.36, "learning_rate": 3.524235296748246e-07, "loss": 0.0045, "step": 59679 }, { "epoch": 18.36, "learning_rate": 3.522924268489003e-07, "loss": 0.0033, "step": 59680 }, { "epoch": 18.36, "learning_rate": 3.5216134797563363e-07, "loss": 0.0033, "step": 59681 }, { "epoch": 18.36, "learning_rate": 3.5203029305534875e-07, "loss": 0.0054, "step": 59682 }, { "epoch": 18.36, "learning_rate": 3.51899262088371e-07, "loss": 0.0026, "step": 59683 }, { "epoch": 18.36, "learning_rate": 3.5176825507502785e-07, "loss": 0.0053, "step": 59684 }, { "epoch": 18.36, "learning_rate": 3.5163727201564025e-07, "loss": 0.0035, "step": 59685 }, { "epoch": 18.36, "learning_rate": 3.5150631291053894e-07, "loss": 0.0051, "step": 59686 }, { "epoch": 18.36, "learning_rate": 3.513753777600448e-07, "loss": 0.0044, "step": 59687 }, { "epoch": 18.37, "learning_rate": 3.5124446656448654e-07, "loss": 0.0029, "step": 59688 }, { "epoch": 18.37, "learning_rate": 3.51113579324186e-07, "loss": 0.0044, "step": 59689 }, { "epoch": 18.37, "learning_rate": 3.5098271603946744e-07, "loss": 0.0033, "step": 59690 }, { "epoch": 18.37, "learning_rate": 3.5085187671065834e-07, "loss": 0.0024, "step": 59691 }, { "epoch": 18.37, "learning_rate": 3.5072106133808403e-07, "loss": 0.007, "step": 59692 }, { "epoch": 18.37, "learning_rate": 3.5059026992206645e-07, "loss": 0.0023, "step": 59693 }, { "epoch": 18.37, "learning_rate": 3.504595024629331e-07, "loss": 0.0027, "step": 59694 }, { "epoch": 18.37, "learning_rate": 3.5032875896100603e-07, "loss": 0.004, "step": 59695 }, { "epoch": 18.37, "learning_rate": 3.5019803941661266e-07, "loss": 0.0038, "step": 59696 }, { "epoch": 18.37, "learning_rate": 3.5006734383007504e-07, "loss": 0.0048, "step": 59697 }, { "epoch": 18.37, "learning_rate": 3.4993667220171943e-07, "loss": 0.0046, "step": 59698 }, { "epoch": 18.37, "learning_rate": 3.4980602453186904e-07, "loss": 0.0062, "step": 59699 }, { "epoch": 18.37, "learning_rate": 3.4967540082084696e-07, "loss": 0.0063, "step": 59700 }, { "epoch": 18.37, "learning_rate": 3.495448010689817e-07, "loss": 0.0029, "step": 59701 }, { "epoch": 18.37, "learning_rate": 3.494142252765942e-07, "loss": 0.0103, "step": 59702 }, { "epoch": 18.37, "learning_rate": 3.492836734440097e-07, "loss": 0.0025, "step": 59703 }, { "epoch": 18.37, "learning_rate": 3.491531455715513e-07, "loss": 0.0046, "step": 59704 }, { "epoch": 18.37, "learning_rate": 3.4902264165954436e-07, "loss": 0.0029, "step": 59705 }, { "epoch": 18.37, "learning_rate": 3.4889216170831187e-07, "loss": 0.004, "step": 59706 }, { "epoch": 18.37, "learning_rate": 3.487617057181791e-07, "loss": 0.0039, "step": 59707 }, { "epoch": 18.37, "learning_rate": 3.4863127368946923e-07, "loss": 0.0027, "step": 59708 }, { "epoch": 18.37, "learning_rate": 3.485008656225053e-07, "loss": 0.0053, "step": 59709 }, { "epoch": 18.37, "learning_rate": 3.4837048151761256e-07, "loss": 0.0027, "step": 59710 }, { "epoch": 18.37, "learning_rate": 3.48240121375113e-07, "loss": 0.0027, "step": 59711 }, { "epoch": 18.37, "learning_rate": 3.481097851953319e-07, "loss": 0.0056, "step": 59712 }, { "epoch": 18.37, "learning_rate": 3.4797947297859126e-07, "loss": 0.0045, "step": 59713 }, { "epoch": 18.37, "learning_rate": 3.4784918472521525e-07, "loss": 0.0026, "step": 59714 }, { "epoch": 18.37, "learning_rate": 3.4771892043552914e-07, "loss": 0.0033, "step": 59715 }, { "epoch": 18.37, "learning_rate": 3.475886801098549e-07, "loss": 0.0022, "step": 59716 }, { "epoch": 18.37, "learning_rate": 3.4745846374851454e-07, "loss": 0.0056, "step": 59717 }, { "epoch": 18.37, "learning_rate": 3.4732827135183335e-07, "loss": 0.0044, "step": 59718 }, { "epoch": 18.37, "learning_rate": 3.4719810292013214e-07, "loss": 0.0033, "step": 59719 }, { "epoch": 18.38, "learning_rate": 3.470679584537362e-07, "loss": 0.0026, "step": 59720 }, { "epoch": 18.38, "learning_rate": 3.4693783795296866e-07, "loss": 0.0037, "step": 59721 }, { "epoch": 18.38, "learning_rate": 3.468077414181537e-07, "loss": 0.0023, "step": 59722 }, { "epoch": 18.38, "learning_rate": 3.4667766884961205e-07, "loss": 0.0022, "step": 59723 }, { "epoch": 18.38, "learning_rate": 3.4654762024766474e-07, "loss": 0.0053, "step": 59724 }, { "epoch": 18.38, "learning_rate": 3.464175956126392e-07, "loss": 0.002, "step": 59725 }, { "epoch": 18.38, "learning_rate": 3.462875949448563e-07, "loss": 0.0055, "step": 59726 }, { "epoch": 18.38, "learning_rate": 3.46157618244638e-07, "loss": 0.0027, "step": 59727 }, { "epoch": 18.38, "learning_rate": 3.460276655123085e-07, "loss": 0.0037, "step": 59728 }, { "epoch": 18.38, "learning_rate": 3.458977367481897e-07, "loss": 0.0032, "step": 59729 }, { "epoch": 18.38, "learning_rate": 3.4576783195260366e-07, "loss": 0.0031, "step": 59730 }, { "epoch": 18.38, "learning_rate": 3.456379511258734e-07, "loss": 0.0022, "step": 59731 }, { "epoch": 18.38, "learning_rate": 3.45508094268322e-07, "loss": 0.0024, "step": 59732 }, { "epoch": 18.38, "learning_rate": 3.453782613802703e-07, "loss": 0.0043, "step": 59733 }, { "epoch": 18.38, "learning_rate": 3.4524845246204143e-07, "loss": 0.0062, "step": 59734 }, { "epoch": 18.38, "learning_rate": 3.451186675139584e-07, "loss": 0.0038, "step": 59735 }, { "epoch": 18.38, "learning_rate": 3.449889065363443e-07, "loss": 0.0029, "step": 59736 }, { "epoch": 18.38, "learning_rate": 3.448591695295189e-07, "loss": 0.0017, "step": 59737 }, { "epoch": 18.38, "learning_rate": 3.447294564938042e-07, "loss": 0.0032, "step": 59738 }, { "epoch": 18.38, "learning_rate": 3.4459976742952316e-07, "loss": 0.0045, "step": 59739 }, { "epoch": 18.38, "learning_rate": 3.4447010233699784e-07, "loss": 0.0041, "step": 59740 }, { "epoch": 18.38, "learning_rate": 3.443404612165513e-07, "loss": 0.0032, "step": 59741 }, { "epoch": 18.38, "learning_rate": 3.4421084406850435e-07, "loss": 0.0064, "step": 59742 }, { "epoch": 18.38, "learning_rate": 3.44081250893179e-07, "loss": 0.004, "step": 59743 }, { "epoch": 18.38, "learning_rate": 3.43951681690895e-07, "loss": 0.0029, "step": 59744 }, { "epoch": 18.38, "learning_rate": 3.438221364619776e-07, "loss": 0.0034, "step": 59745 }, { "epoch": 18.38, "learning_rate": 3.436926152067455e-07, "loss": 0.0036, "step": 59746 }, { "epoch": 18.38, "learning_rate": 3.435631179255217e-07, "loss": 0.0056, "step": 59747 }, { "epoch": 18.38, "learning_rate": 3.4343364461862705e-07, "loss": 0.0041, "step": 59748 }, { "epoch": 18.38, "learning_rate": 3.433041952863847e-07, "loss": 0.0034, "step": 59749 }, { "epoch": 18.38, "learning_rate": 3.4317476992911326e-07, "loss": 0.0031, "step": 59750 }, { "epoch": 18.38, "learning_rate": 3.4304536854713577e-07, "loss": 0.0025, "step": 59751 }, { "epoch": 18.38, "learning_rate": 3.4291599114077314e-07, "loss": 0.0038, "step": 59752 }, { "epoch": 18.39, "learning_rate": 3.427866377103473e-07, "loss": 0.0047, "step": 59753 }, { "epoch": 18.39, "learning_rate": 3.426573082561768e-07, "loss": 0.0063, "step": 59754 }, { "epoch": 18.39, "learning_rate": 3.42528002778586e-07, "loss": 0.0022, "step": 59755 }, { "epoch": 18.39, "learning_rate": 3.4239872127789564e-07, "loss": 0.005, "step": 59756 }, { "epoch": 18.39, "learning_rate": 3.422694637544255e-07, "loss": 0.0032, "step": 59757 }, { "epoch": 18.39, "learning_rate": 3.421402302084953e-07, "loss": 0.0037, "step": 59758 }, { "epoch": 18.39, "learning_rate": 3.42011020640427e-07, "loss": 0.0032, "step": 59759 }, { "epoch": 18.39, "learning_rate": 3.4188183505054264e-07, "loss": 0.0032, "step": 59760 }, { "epoch": 18.39, "learning_rate": 3.4175267343916187e-07, "loss": 0.0031, "step": 59761 }, { "epoch": 18.39, "learning_rate": 3.4162353580660557e-07, "loss": 0.0035, "step": 59762 }, { "epoch": 18.39, "learning_rate": 3.414944221531935e-07, "loss": 0.0047, "step": 59763 }, { "epoch": 18.39, "learning_rate": 3.4136533247924764e-07, "loss": 0.0029, "step": 59764 }, { "epoch": 18.39, "learning_rate": 3.412362667850877e-07, "loss": 0.0027, "step": 59765 }, { "epoch": 18.39, "learning_rate": 3.4110722507103346e-07, "loss": 0.0042, "step": 59766 }, { "epoch": 18.39, "learning_rate": 3.409782073374068e-07, "loss": 0.0021, "step": 59767 }, { "epoch": 18.39, "learning_rate": 3.408492135845265e-07, "loss": 0.0014, "step": 59768 }, { "epoch": 18.39, "learning_rate": 3.4072024381271327e-07, "loss": 0.0024, "step": 59769 }, { "epoch": 18.39, "learning_rate": 3.405912980222892e-07, "loss": 0.0034, "step": 59770 }, { "epoch": 18.39, "learning_rate": 3.404623762135728e-07, "loss": 0.0035, "step": 59771 }, { "epoch": 18.39, "learning_rate": 3.403334783868828e-07, "loss": 0.0037, "step": 59772 }, { "epoch": 18.39, "learning_rate": 3.4020460454254113e-07, "loss": 0.0029, "step": 59773 }, { "epoch": 18.39, "learning_rate": 3.400757546808664e-07, "loss": 0.002, "step": 59774 }, { "epoch": 18.39, "learning_rate": 3.399469288021806e-07, "loss": 0.0024, "step": 59775 }, { "epoch": 18.39, "learning_rate": 3.3981812690680236e-07, "loss": 0.0044, "step": 59776 }, { "epoch": 18.39, "learning_rate": 3.396893489950503e-07, "loss": 0.0068, "step": 59777 }, { "epoch": 18.39, "learning_rate": 3.395605950672476e-07, "loss": 0.0033, "step": 59778 }, { "epoch": 18.39, "learning_rate": 3.3943186512370827e-07, "loss": 0.0035, "step": 59779 }, { "epoch": 18.39, "learning_rate": 3.3930315916475775e-07, "loss": 0.0039, "step": 59780 }, { "epoch": 18.39, "learning_rate": 3.391744771907113e-07, "loss": 0.0024, "step": 59781 }, { "epoch": 18.39, "learning_rate": 3.3904581920189193e-07, "loss": 0.0031, "step": 59782 }, { "epoch": 18.39, "learning_rate": 3.389171851986162e-07, "loss": 0.0042, "step": 59783 }, { "epoch": 18.39, "learning_rate": 3.387885751812048e-07, "loss": 0.0059, "step": 59784 }, { "epoch": 18.4, "learning_rate": 3.3865998914997645e-07, "loss": 0.0041, "step": 59785 }, { "epoch": 18.4, "learning_rate": 3.385314271052509e-07, "loss": 0.0033, "step": 59786 }, { "epoch": 18.4, "learning_rate": 3.384028890473479e-07, "loss": 0.0045, "step": 59787 }, { "epoch": 18.4, "learning_rate": 3.382743749765849e-07, "loss": 0.0037, "step": 59788 }, { "epoch": 18.4, "learning_rate": 3.381458848932806e-07, "loss": 0.0052, "step": 59789 }, { "epoch": 18.4, "learning_rate": 3.38017418797757e-07, "loss": 0.0031, "step": 59790 }, { "epoch": 18.4, "learning_rate": 3.378889766903326e-07, "loss": 0.0035, "step": 59791 }, { "epoch": 18.4, "learning_rate": 3.37760558571324e-07, "loss": 0.004, "step": 59792 }, { "epoch": 18.4, "learning_rate": 3.3763216444105075e-07, "loss": 0.0028, "step": 59793 }, { "epoch": 18.4, "learning_rate": 3.375037942998305e-07, "loss": 0.0062, "step": 59794 }, { "epoch": 18.4, "learning_rate": 3.3737544814798515e-07, "loss": 0.0031, "step": 59795 }, { "epoch": 18.4, "learning_rate": 3.372471259858312e-07, "loss": 0.0031, "step": 59796 }, { "epoch": 18.4, "learning_rate": 3.371188278136883e-07, "loss": 0.0042, "step": 59797 }, { "epoch": 18.4, "learning_rate": 3.369905536318729e-07, "loss": 0.0025, "step": 59798 }, { "epoch": 18.4, "learning_rate": 3.368623034407059e-07, "loss": 0.003, "step": 59799 }, { "epoch": 18.4, "learning_rate": 3.367340772405048e-07, "loss": 0.0037, "step": 59800 }, { "epoch": 18.4, "learning_rate": 3.3660587503158815e-07, "loss": 0.0049, "step": 59801 }, { "epoch": 18.4, "learning_rate": 3.3647769681427354e-07, "loss": 0.0026, "step": 59802 }, { "epoch": 18.4, "learning_rate": 3.3634954258887964e-07, "loss": 0.0041, "step": 59803 }, { "epoch": 18.4, "learning_rate": 3.3622141235572394e-07, "loss": 0.0043, "step": 59804 }, { "epoch": 18.4, "learning_rate": 3.3609330611512614e-07, "loss": 0.0016, "step": 59805 }, { "epoch": 18.4, "learning_rate": 3.359652238674027e-07, "loss": 0.0023, "step": 59806 }, { "epoch": 18.4, "learning_rate": 3.358371656128723e-07, "loss": 0.0046, "step": 59807 }, { "epoch": 18.4, "learning_rate": 3.3570913135185347e-07, "loss": 0.003, "step": 59808 }, { "epoch": 18.4, "learning_rate": 3.3558112108466156e-07, "loss": 0.0028, "step": 59809 }, { "epoch": 18.4, "learning_rate": 3.3545313481161743e-07, "loss": 0.0031, "step": 59810 }, { "epoch": 18.4, "learning_rate": 3.3532517253303755e-07, "loss": 0.0054, "step": 59811 }, { "epoch": 18.4, "learning_rate": 3.3519723424924157e-07, "loss": 0.004, "step": 59812 }, { "epoch": 18.4, "learning_rate": 3.350693199605437e-07, "loss": 0.0019, "step": 59813 }, { "epoch": 18.4, "learning_rate": 3.349414296672626e-07, "loss": 0.0045, "step": 59814 }, { "epoch": 18.4, "learning_rate": 3.3481356336971694e-07, "loss": 0.0061, "step": 59815 }, { "epoch": 18.4, "learning_rate": 3.346857210682242e-07, "loss": 0.0028, "step": 59816 }, { "epoch": 18.4, "learning_rate": 3.3455790276310073e-07, "loss": 0.0027, "step": 59817 }, { "epoch": 18.41, "learning_rate": 3.344301084546642e-07, "loss": 0.0032, "step": 59818 }, { "epoch": 18.41, "learning_rate": 3.3430233814323087e-07, "loss": 0.0026, "step": 59819 }, { "epoch": 18.41, "learning_rate": 3.341745918291206e-07, "loss": 0.0061, "step": 59820 }, { "epoch": 18.41, "learning_rate": 3.3404686951264867e-07, "loss": 0.0032, "step": 59821 }, { "epoch": 18.41, "learning_rate": 3.339191711941325e-07, "loss": 0.0045, "step": 59822 }, { "epoch": 18.41, "learning_rate": 3.3379149687388866e-07, "loss": 0.0043, "step": 59823 }, { "epoch": 18.41, "learning_rate": 3.336638465522335e-07, "loss": 0.0042, "step": 59824 }, { "epoch": 18.41, "learning_rate": 3.3353622022948786e-07, "loss": 0.0028, "step": 59825 }, { "epoch": 18.41, "learning_rate": 3.334086179059637e-07, "loss": 0.0029, "step": 59826 }, { "epoch": 18.41, "learning_rate": 3.3328103958198076e-07, "loss": 0.0085, "step": 59827 }, { "epoch": 18.41, "learning_rate": 3.331534852578544e-07, "loss": 0.003, "step": 59828 }, { "epoch": 18.41, "learning_rate": 3.33025954933901e-07, "loss": 0.0049, "step": 59829 }, { "epoch": 18.41, "learning_rate": 3.328984486104392e-07, "loss": 0.0029, "step": 59830 }, { "epoch": 18.41, "learning_rate": 3.3277096628778425e-07, "loss": 0.0036, "step": 59831 }, { "epoch": 18.41, "learning_rate": 3.326435079662538e-07, "loss": 0.0061, "step": 59832 }, { "epoch": 18.41, "learning_rate": 3.32516073646163e-07, "loss": 0.0043, "step": 59833 }, { "epoch": 18.41, "learning_rate": 3.3238866332782726e-07, "loss": 0.0032, "step": 59834 }, { "epoch": 18.41, "learning_rate": 3.322612770115641e-07, "loss": 0.0053, "step": 59835 }, { "epoch": 18.41, "learning_rate": 3.32133914697691e-07, "loss": 0.0036, "step": 59836 }, { "epoch": 18.41, "learning_rate": 3.3200657638652214e-07, "loss": 0.0044, "step": 59837 }, { "epoch": 18.41, "learning_rate": 3.3187926207837506e-07, "loss": 0.0061, "step": 59838 }, { "epoch": 18.41, "learning_rate": 3.317519717735651e-07, "loss": 0.0039, "step": 59839 }, { "epoch": 18.41, "learning_rate": 3.3162470547240867e-07, "loss": 0.0044, "step": 59840 }, { "epoch": 18.41, "learning_rate": 3.31497463175221e-07, "loss": 0.0044, "step": 59841 }, { "epoch": 18.41, "learning_rate": 3.313702448823197e-07, "loss": 0.0027, "step": 59842 }, { "epoch": 18.41, "learning_rate": 3.3124305059401896e-07, "loss": 0.0022, "step": 59843 }, { "epoch": 18.41, "learning_rate": 3.3111588031063403e-07, "loss": 0.0043, "step": 59844 }, { "epoch": 18.41, "learning_rate": 3.3098873403248243e-07, "loss": 0.0025, "step": 59845 }, { "epoch": 18.41, "learning_rate": 3.308616117598806e-07, "loss": 0.0034, "step": 59846 }, { "epoch": 18.41, "learning_rate": 3.307345134931417e-07, "loss": 0.0025, "step": 59847 }, { "epoch": 18.41, "learning_rate": 3.30607439232582e-07, "loss": 0.0034, "step": 59848 }, { "epoch": 18.41, "learning_rate": 3.3048038897851576e-07, "loss": 0.0057, "step": 59849 }, { "epoch": 18.42, "learning_rate": 3.303533627312616e-07, "loss": 0.0055, "step": 59850 }, { "epoch": 18.42, "learning_rate": 3.3022636049113267e-07, "loss": 0.0019, "step": 59851 }, { "epoch": 18.42, "learning_rate": 3.3009938225844527e-07, "loss": 0.0053, "step": 59852 }, { "epoch": 18.42, "learning_rate": 3.299724280335148e-07, "loss": 0.0035, "step": 59853 }, { "epoch": 18.42, "learning_rate": 3.298454978166543e-07, "loss": 0.0032, "step": 59854 }, { "epoch": 18.42, "learning_rate": 3.297185916081813e-07, "loss": 0.0022, "step": 59855 }, { "epoch": 18.42, "learning_rate": 3.2959170940841e-07, "loss": 0.0034, "step": 59856 }, { "epoch": 18.42, "learning_rate": 3.294648512176557e-07, "loss": 0.0038, "step": 59857 }, { "epoch": 18.42, "learning_rate": 3.293380170362326e-07, "loss": 0.0041, "step": 59858 }, { "epoch": 18.42, "learning_rate": 3.2921120686445594e-07, "loss": 0.0022, "step": 59859 }, { "epoch": 18.42, "learning_rate": 3.290844207026411e-07, "loss": 0.003, "step": 59860 }, { "epoch": 18.42, "learning_rate": 3.2895765855110227e-07, "loss": 0.0051, "step": 59861 }, { "epoch": 18.42, "learning_rate": 3.2883092041015476e-07, "loss": 0.0025, "step": 59862 }, { "epoch": 18.42, "learning_rate": 3.2870420628011267e-07, "loss": 0.0028, "step": 59863 }, { "epoch": 18.42, "learning_rate": 3.285775161612892e-07, "loss": 0.0041, "step": 59864 }, { "epoch": 18.42, "learning_rate": 3.284508500540018e-07, "loss": 0.0048, "step": 59865 }, { "epoch": 18.42, "learning_rate": 3.2832420795856355e-07, "loss": 0.0037, "step": 59866 }, { "epoch": 18.42, "learning_rate": 3.2819758987528984e-07, "loss": 0.0025, "step": 59867 }, { "epoch": 18.42, "learning_rate": 3.280709958044925e-07, "loss": 0.0041, "step": 59868 }, { "epoch": 18.42, "learning_rate": 3.27944425746487e-07, "loss": 0.004, "step": 59869 }, { "epoch": 18.42, "learning_rate": 3.278178797015885e-07, "loss": 0.0057, "step": 59870 }, { "epoch": 18.42, "learning_rate": 3.276913576701113e-07, "loss": 0.0029, "step": 59871 }, { "epoch": 18.42, "learning_rate": 3.2756485965236953e-07, "loss": 0.0026, "step": 59872 }, { "epoch": 18.42, "learning_rate": 3.2743838564867513e-07, "loss": 0.0036, "step": 59873 }, { "epoch": 18.42, "learning_rate": 3.273119356593446e-07, "loss": 0.0029, "step": 59874 }, { "epoch": 18.42, "learning_rate": 3.271855096846899e-07, "loss": 0.0032, "step": 59875 }, { "epoch": 18.42, "learning_rate": 3.270591077250262e-07, "loss": 0.0025, "step": 59876 }, { "epoch": 18.42, "learning_rate": 3.2693272978066795e-07, "loss": 0.0042, "step": 59877 }, { "epoch": 18.42, "learning_rate": 3.2680637585192687e-07, "loss": 0.0047, "step": 59878 }, { "epoch": 18.42, "learning_rate": 3.266800459391173e-07, "loss": 0.0042, "step": 59879 }, { "epoch": 18.42, "learning_rate": 3.2655374004255555e-07, "loss": 0.0037, "step": 59880 }, { "epoch": 18.42, "learning_rate": 3.264274581625515e-07, "loss": 0.0026, "step": 59881 }, { "epoch": 18.42, "learning_rate": 3.2630120029942034e-07, "loss": 0.0027, "step": 59882 }, { "epoch": 18.43, "learning_rate": 3.2617496645347526e-07, "loss": 0.0055, "step": 59883 }, { "epoch": 18.43, "learning_rate": 3.260487566250281e-07, "loss": 0.0043, "step": 59884 }, { "epoch": 18.43, "learning_rate": 3.259225708143954e-07, "loss": 0.0045, "step": 59885 }, { "epoch": 18.43, "learning_rate": 3.2579640902188904e-07, "loss": 0.0039, "step": 59886 }, { "epoch": 18.43, "learning_rate": 3.2567027124782215e-07, "loss": 0.0054, "step": 59887 }, { "epoch": 18.43, "learning_rate": 3.255441574925089e-07, "loss": 0.0033, "step": 59888 }, { "epoch": 18.43, "learning_rate": 3.2541806775625906e-07, "loss": 0.0059, "step": 59889 }, { "epoch": 18.43, "learning_rate": 3.25292002039389e-07, "loss": 0.0033, "step": 59890 }, { "epoch": 18.43, "learning_rate": 3.2516596034221084e-07, "loss": 0.0049, "step": 59891 }, { "epoch": 18.43, "learning_rate": 3.2503994266503744e-07, "loss": 0.0034, "step": 59892 }, { "epoch": 18.43, "learning_rate": 3.2491394900818207e-07, "loss": 0.0023, "step": 59893 }, { "epoch": 18.43, "learning_rate": 3.2478797937195547e-07, "loss": 0.0012, "step": 59894 }, { "epoch": 18.43, "learning_rate": 3.24662033756673e-07, "loss": 0.0051, "step": 59895 }, { "epoch": 18.43, "learning_rate": 3.245361121626467e-07, "loss": 0.0032, "step": 59896 }, { "epoch": 18.43, "learning_rate": 3.244102145901884e-07, "loss": 0.0033, "step": 59897 }, { "epoch": 18.43, "learning_rate": 3.2428434103961124e-07, "loss": 0.0029, "step": 59898 }, { "epoch": 18.43, "learning_rate": 3.2415849151122614e-07, "loss": 0.0026, "step": 59899 }, { "epoch": 18.43, "learning_rate": 3.2403266600534833e-07, "loss": 0.0036, "step": 59900 }, { "epoch": 18.43, "learning_rate": 3.2390686452228983e-07, "loss": 0.0025, "step": 59901 }, { "epoch": 18.43, "learning_rate": 3.2378108706236034e-07, "loss": 0.0032, "step": 59902 }, { "epoch": 18.43, "learning_rate": 3.236553336258741e-07, "loss": 0.0037, "step": 59903 }, { "epoch": 18.43, "learning_rate": 3.2352960421314306e-07, "loss": 0.0026, "step": 59904 }, { "epoch": 18.43, "learning_rate": 3.234038988244792e-07, "loss": 0.0051, "step": 59905 }, { "epoch": 18.43, "learning_rate": 3.232782174601945e-07, "loss": 0.0052, "step": 59906 }, { "epoch": 18.43, "learning_rate": 3.2315256012060203e-07, "loss": 0.0058, "step": 59907 }, { "epoch": 18.43, "learning_rate": 3.2302692680601264e-07, "loss": 0.0013, "step": 59908 }, { "epoch": 18.43, "learning_rate": 3.2290131751673835e-07, "loss": 0.0038, "step": 59909 }, { "epoch": 18.43, "learning_rate": 3.2277573225309113e-07, "loss": 0.005, "step": 59910 }, { "epoch": 18.43, "learning_rate": 3.22650171015384e-07, "loss": 0.003, "step": 59911 }, { "epoch": 18.43, "learning_rate": 3.2252463380392673e-07, "loss": 0.0087, "step": 59912 }, { "epoch": 18.43, "learning_rate": 3.2239912061903133e-07, "loss": 0.0045, "step": 59913 }, { "epoch": 18.43, "learning_rate": 3.222736314610098e-07, "loss": 0.0045, "step": 59914 }, { "epoch": 18.44, "learning_rate": 3.221481663301751e-07, "loss": 0.0022, "step": 59915 }, { "epoch": 18.44, "learning_rate": 3.2202272522683707e-07, "loss": 0.0033, "step": 59916 }, { "epoch": 18.44, "learning_rate": 3.2189730815130657e-07, "loss": 0.0051, "step": 59917 }, { "epoch": 18.44, "learning_rate": 3.217719151038967e-07, "loss": 0.0043, "step": 59918 }, { "epoch": 18.44, "learning_rate": 3.216465460849172e-07, "loss": 0.004, "step": 59919 }, { "epoch": 18.44, "learning_rate": 3.215212010946811e-07, "loss": 0.0027, "step": 59920 }, { "epoch": 18.44, "learning_rate": 3.213958801334982e-07, "loss": 0.0051, "step": 59921 }, { "epoch": 18.44, "learning_rate": 3.2127058320168047e-07, "loss": 0.0037, "step": 59922 }, { "epoch": 18.44, "learning_rate": 3.2114531029953877e-07, "loss": 0.006, "step": 59923 }, { "epoch": 18.44, "learning_rate": 3.2102006142738287e-07, "loss": 0.0043, "step": 59924 }, { "epoch": 18.44, "learning_rate": 3.208948365855247e-07, "loss": 0.0028, "step": 59925 }, { "epoch": 18.44, "learning_rate": 3.2076963577427624e-07, "loss": 0.0035, "step": 59926 }, { "epoch": 18.44, "learning_rate": 3.2064445899394723e-07, "loss": 0.0054, "step": 59927 }, { "epoch": 18.44, "learning_rate": 3.205193062448475e-07, "loss": 0.003, "step": 59928 }, { "epoch": 18.44, "learning_rate": 3.2039417752729006e-07, "loss": 0.0042, "step": 59929 }, { "epoch": 18.44, "learning_rate": 3.202690728415836e-07, "loss": 0.0037, "step": 59930 }, { "epoch": 18.44, "learning_rate": 3.2014399218804e-07, "loss": 0.0025, "step": 59931 }, { "epoch": 18.44, "learning_rate": 3.2001893556696916e-07, "loss": 0.005, "step": 59932 }, { "epoch": 18.44, "learning_rate": 3.1989390297868183e-07, "loss": 0.0033, "step": 59933 }, { "epoch": 18.44, "learning_rate": 3.197688944234867e-07, "loss": 0.0044, "step": 59934 }, { "epoch": 18.44, "learning_rate": 3.196439099016968e-07, "loss": 0.0037, "step": 59935 }, { "epoch": 18.44, "learning_rate": 3.1951894941362196e-07, "loss": 0.0032, "step": 59936 }, { "epoch": 18.44, "learning_rate": 3.1939401295957186e-07, "loss": 0.005, "step": 59937 }, { "epoch": 18.44, "learning_rate": 3.192691005398563e-07, "loss": 0.0024, "step": 59938 }, { "epoch": 18.44, "learning_rate": 3.1914421215478385e-07, "loss": 0.0023, "step": 59939 }, { "epoch": 18.44, "learning_rate": 3.190193478046677e-07, "loss": 0.0056, "step": 59940 }, { "epoch": 18.44, "learning_rate": 3.1889450748981756e-07, "loss": 0.0026, "step": 59941 }, { "epoch": 18.44, "learning_rate": 3.1876969121054204e-07, "loss": 0.002, "step": 59942 }, { "epoch": 18.44, "learning_rate": 3.18644898967152e-07, "loss": 0.0033, "step": 59943 }, { "epoch": 18.44, "learning_rate": 3.18520130759955e-07, "loss": 0.0046, "step": 59944 }, { "epoch": 18.44, "learning_rate": 3.1839538658926307e-07, "loss": 0.0057, "step": 59945 }, { "epoch": 18.44, "learning_rate": 3.1827066645538586e-07, "loss": 0.0027, "step": 59946 }, { "epoch": 18.44, "learning_rate": 3.18145970358632e-07, "loss": 0.005, "step": 59947 }, { "epoch": 18.45, "learning_rate": 3.180212982993114e-07, "loss": 0.004, "step": 59948 }, { "epoch": 18.45, "learning_rate": 3.178966502777348e-07, "loss": 0.0025, "step": 59949 }, { "epoch": 18.45, "learning_rate": 3.1777202629420965e-07, "loss": 0.0032, "step": 59950 }, { "epoch": 18.45, "learning_rate": 3.176474263490459e-07, "loss": 0.0034, "step": 59951 }, { "epoch": 18.45, "learning_rate": 3.1752285044255424e-07, "loss": 0.0056, "step": 59952 }, { "epoch": 18.45, "learning_rate": 3.1739829857504235e-07, "loss": 0.004, "step": 59953 }, { "epoch": 18.45, "learning_rate": 3.1727377074681984e-07, "loss": 0.0029, "step": 59954 }, { "epoch": 18.45, "learning_rate": 3.171492669581966e-07, "loss": 0.0033, "step": 59955 }, { "epoch": 18.45, "learning_rate": 3.1702478720948116e-07, "loss": 0.0032, "step": 59956 }, { "epoch": 18.45, "learning_rate": 3.1690033150098443e-07, "loss": 0.0041, "step": 59957 }, { "epoch": 18.45, "learning_rate": 3.167758998330128e-07, "loss": 0.0028, "step": 59958 }, { "epoch": 18.45, "learning_rate": 3.166514922058739e-07, "loss": 0.0017, "step": 59959 }, { "epoch": 18.45, "learning_rate": 3.1652710861988066e-07, "loss": 0.0025, "step": 59960 }, { "epoch": 18.45, "learning_rate": 3.1640274907534074e-07, "loss": 0.0038, "step": 59961 }, { "epoch": 18.45, "learning_rate": 3.162784135725605e-07, "loss": 0.0031, "step": 59962 }, { "epoch": 18.45, "learning_rate": 3.1615410211185196e-07, "loss": 0.0021, "step": 59963 }, { "epoch": 18.45, "learning_rate": 3.160298146935215e-07, "loss": 0.0031, "step": 59964 }, { "epoch": 18.45, "learning_rate": 3.159055513178777e-07, "loss": 0.0025, "step": 59965 }, { "epoch": 18.45, "learning_rate": 3.157813119852304e-07, "loss": 0.0041, "step": 59966 }, { "epoch": 18.45, "learning_rate": 3.1565709669588607e-07, "loss": 0.0047, "step": 59967 }, { "epoch": 18.45, "learning_rate": 3.1553290545015547e-07, "loss": 0.0026, "step": 59968 }, { "epoch": 18.45, "learning_rate": 3.1540873824834505e-07, "loss": 0.0044, "step": 59969 }, { "epoch": 18.45, "learning_rate": 3.152845950907657e-07, "loss": 0.0057, "step": 59970 }, { "epoch": 18.45, "learning_rate": 3.151604759777216e-07, "loss": 0.0043, "step": 59971 }, { "epoch": 18.45, "learning_rate": 3.150363809095247e-07, "loss": 0.0036, "step": 59972 }, { "epoch": 18.45, "learning_rate": 3.1491230988648035e-07, "loss": 0.0041, "step": 59973 }, { "epoch": 18.45, "learning_rate": 3.147882629088972e-07, "loss": 0.0022, "step": 59974 }, { "epoch": 18.45, "learning_rate": 3.1466423997708497e-07, "loss": 0.0053, "step": 59975 }, { "epoch": 18.45, "learning_rate": 3.1454024109135005e-07, "loss": 0.0012, "step": 59976 }, { "epoch": 18.45, "learning_rate": 3.1441626625200007e-07, "loss": 0.0037, "step": 59977 }, { "epoch": 18.45, "learning_rate": 3.1429231545934467e-07, "loss": 0.0041, "step": 59978 }, { "epoch": 18.45, "learning_rate": 3.1416838871368925e-07, "loss": 0.0052, "step": 59979 }, { "epoch": 18.46, "learning_rate": 3.140444860153424e-07, "loss": 0.0029, "step": 59980 }, { "epoch": 18.46, "learning_rate": 3.139206073646117e-07, "loss": 0.0036, "step": 59981 }, { "epoch": 18.46, "learning_rate": 3.1379675276180575e-07, "loss": 0.0037, "step": 59982 }, { "epoch": 18.46, "learning_rate": 3.13672922207231e-07, "loss": 0.0026, "step": 59983 }, { "epoch": 18.46, "learning_rate": 3.1354911570119385e-07, "loss": 0.0039, "step": 59984 }, { "epoch": 18.46, "learning_rate": 3.13425333244004e-07, "loss": 0.0031, "step": 59985 }, { "epoch": 18.46, "learning_rate": 3.13301574835968e-07, "loss": 0.004, "step": 59986 }, { "epoch": 18.46, "learning_rate": 3.131778404773922e-07, "loss": 0.0029, "step": 59987 }, { "epoch": 18.46, "learning_rate": 3.130541301685841e-07, "loss": 0.0045, "step": 59988 }, { "epoch": 18.46, "learning_rate": 3.129304439098502e-07, "loss": 0.0021, "step": 59989 }, { "epoch": 18.46, "learning_rate": 3.12806781701499e-07, "loss": 0.0033, "step": 59990 }, { "epoch": 18.46, "learning_rate": 3.1268314354383823e-07, "loss": 0.0051, "step": 59991 }, { "epoch": 18.46, "learning_rate": 3.1255952943717305e-07, "loss": 0.0051, "step": 59992 }, { "epoch": 18.46, "learning_rate": 3.1243593938181106e-07, "loss": 0.004, "step": 59993 }, { "epoch": 18.46, "learning_rate": 3.1231237337805754e-07, "loss": 0.0032, "step": 59994 }, { "epoch": 18.46, "learning_rate": 3.1218883142622114e-07, "loss": 0.0042, "step": 59995 }, { "epoch": 18.46, "learning_rate": 3.120653135266083e-07, "loss": 0.0021, "step": 59996 }, { "epoch": 18.46, "learning_rate": 3.119418196795265e-07, "loss": 0.0041, "step": 59997 }, { "epoch": 18.46, "learning_rate": 3.1181834988528113e-07, "loss": 0.0041, "step": 59998 }, { "epoch": 18.46, "learning_rate": 3.1169490414417746e-07, "loss": 0.0042, "step": 59999 }, { "epoch": 18.46, "learning_rate": 3.1157148245652525e-07, "loss": 0.0102, "step": 60000 }, { "epoch": 18.46, "learning_rate": 3.114480848226276e-07, "loss": 0.0026, "step": 60001 }, { "epoch": 18.46, "learning_rate": 3.1132471124279305e-07, "loss": 0.0043, "step": 60002 }, { "epoch": 18.46, "learning_rate": 3.1120136171732816e-07, "loss": 0.0026, "step": 60003 }, { "epoch": 18.46, "learning_rate": 3.110780362465371e-07, "loss": 0.0047, "step": 60004 }, { "epoch": 18.46, "learning_rate": 3.1095473483072733e-07, "loss": 0.0031, "step": 60005 }, { "epoch": 18.46, "learning_rate": 3.108314574702054e-07, "loss": 0.0037, "step": 60006 }, { "epoch": 18.46, "learning_rate": 3.1070820416527646e-07, "loss": 0.003, "step": 60007 }, { "epoch": 18.46, "learning_rate": 3.1058497491624704e-07, "loss": 0.0047, "step": 60008 }, { "epoch": 18.46, "learning_rate": 3.1046176972342246e-07, "loss": 0.0047, "step": 60009 }, { "epoch": 18.46, "learning_rate": 3.1033858858710905e-07, "loss": 0.0051, "step": 60010 }, { "epoch": 18.46, "learning_rate": 3.102154315076134e-07, "loss": 0.004, "step": 60011 }, { "epoch": 18.46, "learning_rate": 3.100922984852417e-07, "loss": 0.0047, "step": 60012 }, { "epoch": 18.47, "learning_rate": 3.0996918952029717e-07, "loss": 0.0029, "step": 60013 }, { "epoch": 18.47, "learning_rate": 3.098461046130852e-07, "loss": 0.0044, "step": 60014 }, { "epoch": 18.47, "learning_rate": 3.0972304376391426e-07, "loss": 0.0024, "step": 60015 }, { "epoch": 18.47, "learning_rate": 3.096000069730898e-07, "loss": 0.0022, "step": 60016 }, { "epoch": 18.47, "learning_rate": 3.094769942409148e-07, "loss": 0.0048, "step": 60017 }, { "epoch": 18.47, "learning_rate": 3.093540055676958e-07, "loss": 0.0047, "step": 60018 }, { "epoch": 18.47, "learning_rate": 3.092310409537391e-07, "loss": 0.0016, "step": 60019 }, { "epoch": 18.47, "learning_rate": 3.09108100399349e-07, "loss": 0.0025, "step": 60020 }, { "epoch": 18.47, "learning_rate": 3.089851839048308e-07, "loss": 0.0058, "step": 60021 }, { "epoch": 18.47, "learning_rate": 3.088622914704897e-07, "loss": 0.0033, "step": 60022 }, { "epoch": 18.47, "learning_rate": 3.087394230966312e-07, "loss": 0.0048, "step": 60023 }, { "epoch": 18.47, "learning_rate": 3.086165787835582e-07, "loss": 0.0047, "step": 60024 }, { "epoch": 18.47, "learning_rate": 3.0849375853158057e-07, "loss": 0.0051, "step": 60025 }, { "epoch": 18.47, "learning_rate": 3.0837096234099804e-07, "loss": 0.0049, "step": 60026 }, { "epoch": 18.47, "learning_rate": 3.0824819021211814e-07, "loss": 0.0042, "step": 60027 }, { "epoch": 18.47, "learning_rate": 3.081254421452451e-07, "loss": 0.004, "step": 60028 }, { "epoch": 18.47, "learning_rate": 3.0800271814068303e-07, "loss": 0.0046, "step": 60029 }, { "epoch": 18.47, "learning_rate": 3.078800181987385e-07, "loss": 0.0026, "step": 60030 }, { "epoch": 18.47, "learning_rate": 3.0775734231971443e-07, "loss": 0.0028, "step": 60031 }, { "epoch": 18.47, "learning_rate": 3.0763469050391623e-07, "loss": 0.0047, "step": 60032 }, { "epoch": 18.47, "learning_rate": 3.0751206275164925e-07, "loss": 0.0036, "step": 60033 }, { "epoch": 18.47, "learning_rate": 3.073894590632143e-07, "loss": 0.0041, "step": 60034 }, { "epoch": 18.47, "learning_rate": 3.0726687943891996e-07, "loss": 0.0014, "step": 60035 }, { "epoch": 18.47, "learning_rate": 3.0714432387906836e-07, "loss": 0.0034, "step": 60036 }, { "epoch": 18.47, "learning_rate": 3.070217923839647e-07, "loss": 0.0084, "step": 60037 }, { "epoch": 18.47, "learning_rate": 3.0689928495391207e-07, "loss": 0.0036, "step": 60038 }, { "epoch": 18.47, "learning_rate": 3.0677680158921585e-07, "loss": 0.0015, "step": 60039 }, { "epoch": 18.47, "learning_rate": 3.0665434229017907e-07, "loss": 0.0042, "step": 60040 }, { "epoch": 18.47, "learning_rate": 3.0653190705710713e-07, "loss": 0.0029, "step": 60041 }, { "epoch": 18.47, "learning_rate": 3.06409495890303e-07, "loss": 0.0031, "step": 60042 }, { "epoch": 18.47, "learning_rate": 3.0628710879007095e-07, "loss": 0.0034, "step": 60043 }, { "epoch": 18.47, "learning_rate": 3.0616474575671297e-07, "loss": 0.0018, "step": 60044 }, { "epoch": 18.48, "learning_rate": 3.060424067905354e-07, "loss": 0.0034, "step": 60045 }, { "epoch": 18.48, "learning_rate": 3.059200918918437e-07, "loss": 0.0037, "step": 60046 }, { "epoch": 18.48, "learning_rate": 3.0579780106093636e-07, "loss": 0.0031, "step": 60047 }, { "epoch": 18.48, "learning_rate": 3.0567553429811994e-07, "loss": 0.0023, "step": 60048 }, { "epoch": 18.48, "learning_rate": 3.055532916036974e-07, "loss": 0.0042, "step": 60049 }, { "epoch": 18.48, "learning_rate": 3.0543107297797304e-07, "loss": 0.0027, "step": 60050 }, { "epoch": 18.48, "learning_rate": 3.053088784212499e-07, "loss": 0.0029, "step": 60051 }, { "epoch": 18.48, "learning_rate": 3.0518670793382996e-07, "loss": 0.0049, "step": 60052 }, { "epoch": 18.48, "learning_rate": 3.050645615160197e-07, "loss": 0.0032, "step": 60053 }, { "epoch": 18.48, "learning_rate": 3.049424391681188e-07, "loss": 0.003, "step": 60054 }, { "epoch": 18.48, "learning_rate": 3.0482034089043377e-07, "loss": 0.0045, "step": 60055 }, { "epoch": 18.48, "learning_rate": 3.046982666832643e-07, "loss": 0.0031, "step": 60056 }, { "epoch": 18.48, "learning_rate": 3.045762165469168e-07, "loss": 0.0042, "step": 60057 }, { "epoch": 18.48, "learning_rate": 3.0445419048169224e-07, "loss": 0.0032, "step": 60058 }, { "epoch": 18.48, "learning_rate": 3.043321884878936e-07, "loss": 0.0029, "step": 60059 }, { "epoch": 18.48, "learning_rate": 3.042102105658251e-07, "loss": 0.003, "step": 60060 }, { "epoch": 18.48, "learning_rate": 3.0408825671578767e-07, "loss": 0.0038, "step": 60061 }, { "epoch": 18.48, "learning_rate": 3.039663269380866e-07, "loss": 0.0046, "step": 60062 }, { "epoch": 18.48, "learning_rate": 3.038444212330216e-07, "loss": 0.0044, "step": 60063 }, { "epoch": 18.48, "learning_rate": 3.0372253960089695e-07, "loss": 0.0043, "step": 60064 }, { "epoch": 18.48, "learning_rate": 3.0360068204201567e-07, "loss": 0.0039, "step": 60065 }, { "epoch": 18.48, "learning_rate": 3.0347884855667976e-07, "loss": 0.0026, "step": 60066 }, { "epoch": 18.48, "learning_rate": 3.0335703914519345e-07, "loss": 0.0031, "step": 60067 }, { "epoch": 18.48, "learning_rate": 3.0323525380785645e-07, "loss": 0.0021, "step": 60068 }, { "epoch": 18.48, "learning_rate": 3.031134925449708e-07, "loss": 0.0032, "step": 60069 }, { "epoch": 18.48, "learning_rate": 3.029917553568407e-07, "loss": 0.0127, "step": 60070 }, { "epoch": 18.48, "learning_rate": 3.0287004224376804e-07, "loss": 0.0024, "step": 60071 }, { "epoch": 18.48, "learning_rate": 3.027483532060549e-07, "loss": 0.0033, "step": 60072 }, { "epoch": 18.48, "learning_rate": 3.026266882440043e-07, "loss": 0.0023, "step": 60073 }, { "epoch": 18.48, "learning_rate": 3.025050473579161e-07, "loss": 0.0052, "step": 60074 }, { "epoch": 18.48, "learning_rate": 3.0238343054809325e-07, "loss": 0.0049, "step": 60075 }, { "epoch": 18.48, "learning_rate": 3.0226183781483897e-07, "loss": 0.0033, "step": 60076 }, { "epoch": 18.48, "learning_rate": 3.021402691584541e-07, "loss": 0.0026, "step": 60077 }, { "epoch": 18.49, "learning_rate": 3.0201872457923944e-07, "loss": 0.0042, "step": 60078 }, { "epoch": 18.49, "learning_rate": 3.01897204077497e-07, "loss": 0.003, "step": 60079 }, { "epoch": 18.49, "learning_rate": 3.0177570765353217e-07, "loss": 0.003, "step": 60080 }, { "epoch": 18.49, "learning_rate": 3.016542353076424e-07, "loss": 0.0045, "step": 60081 }, { "epoch": 18.49, "learning_rate": 3.015327870401308e-07, "loss": 0.0023, "step": 60082 }, { "epoch": 18.49, "learning_rate": 3.0141136285129825e-07, "loss": 0.0058, "step": 60083 }, { "epoch": 18.49, "learning_rate": 3.012899627414456e-07, "loss": 0.0061, "step": 60084 }, { "epoch": 18.49, "learning_rate": 3.01168586710876e-07, "loss": 0.0072, "step": 60085 }, { "epoch": 18.49, "learning_rate": 3.010472347598903e-07, "loss": 0.0036, "step": 60086 }, { "epoch": 18.49, "learning_rate": 3.009259068887893e-07, "loss": 0.0038, "step": 60087 }, { "epoch": 18.49, "learning_rate": 3.0080460309787615e-07, "loss": 0.0028, "step": 60088 }, { "epoch": 18.49, "learning_rate": 3.0068332338744843e-07, "loss": 0.0026, "step": 60089 }, { "epoch": 18.49, "learning_rate": 3.0056206775780916e-07, "loss": 0.0038, "step": 60090 }, { "epoch": 18.49, "learning_rate": 3.0044083620926036e-07, "loss": 0.0043, "step": 60091 }, { "epoch": 18.49, "learning_rate": 3.003196287421017e-07, "loss": 0.0067, "step": 60092 }, { "epoch": 18.49, "learning_rate": 3.0019844535663425e-07, "loss": 0.0061, "step": 60093 }, { "epoch": 18.49, "learning_rate": 3.0007728605315865e-07, "loss": 0.0027, "step": 60094 }, { "epoch": 18.49, "learning_rate": 2.9995615083197705e-07, "loss": 0.0036, "step": 60095 }, { "epoch": 18.49, "learning_rate": 2.9983503969338913e-07, "loss": 0.0042, "step": 60096 }, { "epoch": 18.49, "learning_rate": 2.9971395263769576e-07, "loss": 0.0027, "step": 60097 }, { "epoch": 18.49, "learning_rate": 2.995928896651956e-07, "loss": 0.0038, "step": 60098 }, { "epoch": 18.49, "learning_rate": 2.994718507761929e-07, "loss": 0.0027, "step": 60099 }, { "epoch": 18.49, "learning_rate": 2.9935083597098627e-07, "loss": 0.0032, "step": 60100 }, { "epoch": 18.49, "learning_rate": 2.992298452498776e-07, "loss": 0.0034, "step": 60101 }, { "epoch": 18.49, "learning_rate": 2.991088786131646e-07, "loss": 0.004, "step": 60102 }, { "epoch": 18.49, "learning_rate": 2.989879360611481e-07, "loss": 0.0051, "step": 60103 }, { "epoch": 18.49, "learning_rate": 2.9886701759413105e-07, "loss": 0.0041, "step": 60104 }, { "epoch": 18.49, "learning_rate": 2.9874612321241116e-07, "loss": 0.0045, "step": 60105 }, { "epoch": 18.49, "learning_rate": 2.9862525291628917e-07, "loss": 0.0026, "step": 60106 }, { "epoch": 18.49, "learning_rate": 2.9850440670606493e-07, "loss": 0.0036, "step": 60107 }, { "epoch": 18.49, "learning_rate": 2.983835845820393e-07, "loss": 0.0034, "step": 60108 }, { "epoch": 18.49, "learning_rate": 2.982627865445109e-07, "loss": 0.0056, "step": 60109 }, { "epoch": 18.5, "learning_rate": 2.9814201259378173e-07, "loss": 0.0052, "step": 60110 }, { "epoch": 18.5, "learning_rate": 2.9802126273015044e-07, "loss": 0.0039, "step": 60111 }, { "epoch": 18.5, "learning_rate": 2.979005369539156e-07, "loss": 0.0038, "step": 60112 }, { "epoch": 18.5, "learning_rate": 2.977798352653782e-07, "loss": 0.003, "step": 60113 }, { "epoch": 18.5, "learning_rate": 2.97659157664838e-07, "loss": 0.0034, "step": 60114 }, { "epoch": 18.5, "learning_rate": 2.9753850415259577e-07, "loss": 0.0042, "step": 60115 }, { "epoch": 18.5, "learning_rate": 2.9741787472894913e-07, "loss": 0.004, "step": 60116 }, { "epoch": 18.5, "learning_rate": 2.972972693941978e-07, "loss": 0.0023, "step": 60117 }, { "epoch": 18.5, "learning_rate": 2.9717668814864043e-07, "loss": 0.0038, "step": 60118 }, { "epoch": 18.5, "learning_rate": 2.97056130992579e-07, "loss": 0.0029, "step": 60119 }, { "epoch": 18.5, "learning_rate": 2.9693559792631113e-07, "loss": 0.0035, "step": 60120 }, { "epoch": 18.5, "learning_rate": 2.9681508895013645e-07, "loss": 0.0027, "step": 60121 }, { "epoch": 18.5, "learning_rate": 2.966946040643548e-07, "loss": 0.005, "step": 60122 }, { "epoch": 18.5, "learning_rate": 2.9657414326926146e-07, "loss": 0.0019, "step": 60123 }, { "epoch": 18.5, "learning_rate": 2.964537065651607e-07, "loss": 0.0026, "step": 60124 }, { "epoch": 18.5, "learning_rate": 2.9633329395235e-07, "loss": 0.0033, "step": 60125 }, { "epoch": 18.5, "learning_rate": 2.962129054311269e-07, "loss": 0.0049, "step": 60126 }, { "epoch": 18.5, "learning_rate": 2.960925410017912e-07, "loss": 0.0032, "step": 60127 }, { "epoch": 18.5, "learning_rate": 2.959722006646415e-07, "loss": 0.0036, "step": 60128 }, { "epoch": 18.5, "learning_rate": 2.9585188441997646e-07, "loss": 0.0019, "step": 60129 }, { "epoch": 18.5, "learning_rate": 2.957315922680959e-07, "loss": 0.0044, "step": 60130 }, { "epoch": 18.5, "learning_rate": 2.9561132420929726e-07, "loss": 0.0042, "step": 60131 }, { "epoch": 18.5, "learning_rate": 2.9549108024387927e-07, "loss": 0.0039, "step": 60132 }, { "epoch": 18.5, "learning_rate": 2.9537086037213945e-07, "loss": 0.0042, "step": 60133 }, { "epoch": 18.5, "learning_rate": 2.9525066459437866e-07, "loss": 0.003, "step": 60134 }, { "epoch": 18.5, "learning_rate": 2.951304929108956e-07, "loss": 0.0047, "step": 60135 }, { "epoch": 18.5, "learning_rate": 2.950103453219866e-07, "loss": 0.0042, "step": 60136 }, { "epoch": 18.5, "learning_rate": 2.9489022182794927e-07, "loss": 0.0044, "step": 60137 }, { "epoch": 18.5, "learning_rate": 2.947701224290822e-07, "loss": 0.0043, "step": 60138 }, { "epoch": 18.5, "learning_rate": 2.9465004712568637e-07, "loss": 0.0068, "step": 60139 }, { "epoch": 18.5, "learning_rate": 2.94529995918057e-07, "loss": 0.0036, "step": 60140 }, { "epoch": 18.5, "learning_rate": 2.944099688064939e-07, "loss": 0.0016, "step": 60141 }, { "epoch": 18.5, "learning_rate": 2.9428996579129453e-07, "loss": 0.0096, "step": 60142 }, { "epoch": 18.51, "learning_rate": 2.9416998687275543e-07, "loss": 0.0035, "step": 60143 }, { "epoch": 18.51, "learning_rate": 2.9405003205117633e-07, "loss": 0.0048, "step": 60144 }, { "epoch": 18.51, "learning_rate": 2.9393010132685475e-07, "loss": 0.0034, "step": 60145 }, { "epoch": 18.51, "learning_rate": 2.938101947000882e-07, "loss": 0.0037, "step": 60146 }, { "epoch": 18.51, "learning_rate": 2.936903121711732e-07, "loss": 0.0036, "step": 60147 }, { "epoch": 18.51, "learning_rate": 2.935704537404083e-07, "loss": 0.0049, "step": 60148 }, { "epoch": 18.51, "learning_rate": 2.9345061940809216e-07, "loss": 0.002, "step": 60149 }, { "epoch": 18.51, "learning_rate": 2.9333080917452126e-07, "loss": 0.0035, "step": 60150 }, { "epoch": 18.51, "learning_rate": 2.9321102303999316e-07, "loss": 0.0027, "step": 60151 }, { "epoch": 18.51, "learning_rate": 2.930912610048042e-07, "loss": 0.0037, "step": 60152 }, { "epoch": 18.51, "learning_rate": 2.929715230692531e-07, "loss": 0.0026, "step": 60153 }, { "epoch": 18.51, "learning_rate": 2.9285180923363633e-07, "loss": 0.0044, "step": 60154 }, { "epoch": 18.51, "learning_rate": 2.927321194982524e-07, "loss": 0.0052, "step": 60155 }, { "epoch": 18.51, "learning_rate": 2.92612453863399e-07, "loss": 0.0034, "step": 60156 }, { "epoch": 18.51, "learning_rate": 2.924928123293702e-07, "loss": 0.0051, "step": 60157 }, { "epoch": 18.51, "learning_rate": 2.9237319489646363e-07, "loss": 0.0041, "step": 60158 }, { "epoch": 18.51, "learning_rate": 2.9225360156497905e-07, "loss": 0.0036, "step": 60159 }, { "epoch": 18.51, "learning_rate": 2.9213403233521067e-07, "loss": 0.0038, "step": 60160 }, { "epoch": 18.51, "learning_rate": 2.9201448720745706e-07, "loss": 0.0018, "step": 60161 }, { "epoch": 18.51, "learning_rate": 2.9189496618201363e-07, "loss": 0.0028, "step": 60162 }, { "epoch": 18.51, "learning_rate": 2.9177546925917896e-07, "loss": 0.0024, "step": 60163 }, { "epoch": 18.51, "learning_rate": 2.9165599643924734e-07, "loss": 0.004, "step": 60164 }, { "epoch": 18.51, "learning_rate": 2.915365477225174e-07, "loss": 0.0046, "step": 60165 }, { "epoch": 18.51, "learning_rate": 2.914171231092844e-07, "loss": 0.0022, "step": 60166 }, { "epoch": 18.51, "learning_rate": 2.912977225998459e-07, "loss": 0.0039, "step": 60167 }, { "epoch": 18.51, "learning_rate": 2.9117834619449615e-07, "loss": 0.0056, "step": 60168 }, { "epoch": 18.51, "learning_rate": 2.9105899389353375e-07, "loss": 0.0043, "step": 60169 }, { "epoch": 18.51, "learning_rate": 2.909396656972563e-07, "loss": 0.0063, "step": 60170 }, { "epoch": 18.51, "learning_rate": 2.9082036160595683e-07, "loss": 0.0037, "step": 60171 }, { "epoch": 18.51, "learning_rate": 2.907010816199329e-07, "loss": 0.0038, "step": 60172 }, { "epoch": 18.51, "learning_rate": 2.905818257394799e-07, "loss": 0.0035, "step": 60173 }, { "epoch": 18.51, "learning_rate": 2.904625939648953e-07, "loss": 0.0029, "step": 60174 }, { "epoch": 18.52, "learning_rate": 2.903433862964744e-07, "loss": 0.0017, "step": 60175 }, { "epoch": 18.52, "learning_rate": 2.902242027345137e-07, "loss": 0.0034, "step": 60176 }, { "epoch": 18.52, "learning_rate": 2.9010504327930955e-07, "loss": 0.003, "step": 60177 }, { "epoch": 18.52, "learning_rate": 2.89985907931154e-07, "loss": 0.002, "step": 60178 }, { "epoch": 18.52, "learning_rate": 2.898667966903479e-07, "loss": 0.0014, "step": 60179 }, { "epoch": 18.52, "learning_rate": 2.897477095571832e-07, "loss": 0.0033, "step": 60180 }, { "epoch": 18.52, "learning_rate": 2.8962864653195865e-07, "loss": 0.0047, "step": 60181 }, { "epoch": 18.52, "learning_rate": 2.8950960761496837e-07, "loss": 0.0099, "step": 60182 }, { "epoch": 18.52, "learning_rate": 2.893905928065066e-07, "loss": 0.0038, "step": 60183 }, { "epoch": 18.52, "learning_rate": 2.8927160210687087e-07, "loss": 0.0028, "step": 60184 }, { "epoch": 18.52, "learning_rate": 2.891526355163565e-07, "loss": 0.0033, "step": 60185 }, { "epoch": 18.52, "learning_rate": 2.890336930352566e-07, "loss": 0.0042, "step": 60186 }, { "epoch": 18.52, "learning_rate": 2.8891477466386987e-07, "loss": 0.004, "step": 60187 }, { "epoch": 18.52, "learning_rate": 2.8879588040248706e-07, "loss": 0.0045, "step": 60188 }, { "epoch": 18.52, "learning_rate": 2.8867701025140803e-07, "loss": 0.0064, "step": 60189 }, { "epoch": 18.52, "learning_rate": 2.8855816421092586e-07, "loss": 0.0039, "step": 60190 }, { "epoch": 18.52, "learning_rate": 2.8843934228133586e-07, "loss": 0.0035, "step": 60191 }, { "epoch": 18.52, "learning_rate": 2.8832054446293224e-07, "loss": 0.0027, "step": 60192 }, { "epoch": 18.52, "learning_rate": 2.882017707560092e-07, "loss": 0.004, "step": 60193 }, { "epoch": 18.52, "learning_rate": 2.880830211608654e-07, "loss": 0.0063, "step": 60194 }, { "epoch": 18.52, "learning_rate": 2.879642956777917e-07, "loss": 0.004, "step": 60195 }, { "epoch": 18.52, "learning_rate": 2.878455943070846e-07, "loss": 0.0033, "step": 60196 }, { "epoch": 18.52, "learning_rate": 2.877269170490393e-07, "loss": 0.0036, "step": 60197 }, { "epoch": 18.52, "learning_rate": 2.87608263903949e-07, "loss": 0.005, "step": 60198 }, { "epoch": 18.52, "learning_rate": 2.87489634872109e-07, "loss": 0.0026, "step": 60199 }, { "epoch": 18.52, "learning_rate": 2.873710299538146e-07, "loss": 0.0036, "step": 60200 }, { "epoch": 18.52, "learning_rate": 2.872524491493589e-07, "loss": 0.0044, "step": 60201 }, { "epoch": 18.52, "learning_rate": 2.871338924590361e-07, "loss": 0.0036, "step": 60202 }, { "epoch": 18.52, "learning_rate": 2.870153598831416e-07, "loss": 0.0036, "step": 60203 }, { "epoch": 18.52, "learning_rate": 2.8689685142197055e-07, "loss": 0.0047, "step": 60204 }, { "epoch": 18.52, "learning_rate": 2.867783670758162e-07, "loss": 0.0041, "step": 60205 }, { "epoch": 18.52, "learning_rate": 2.8665990684497156e-07, "loss": 0.0034, "step": 60206 }, { "epoch": 18.52, "learning_rate": 2.86541470729732e-07, "loss": 0.0017, "step": 60207 }, { "epoch": 18.53, "learning_rate": 2.8642305873039066e-07, "loss": 0.0053, "step": 60208 }, { "epoch": 18.53, "learning_rate": 2.8630467084724276e-07, "loss": 0.0027, "step": 60209 }, { "epoch": 18.53, "learning_rate": 2.8618630708058147e-07, "loss": 0.0036, "step": 60210 }, { "epoch": 18.53, "learning_rate": 2.8606796743070096e-07, "loss": 0.003, "step": 60211 }, { "epoch": 18.53, "learning_rate": 2.8594965189789547e-07, "loss": 0.0035, "step": 60212 }, { "epoch": 18.53, "learning_rate": 2.8583136048245697e-07, "loss": 0.005, "step": 60213 }, { "epoch": 18.53, "learning_rate": 2.8571309318468074e-07, "loss": 0.0059, "step": 60214 }, { "epoch": 18.53, "learning_rate": 2.855948500048611e-07, "loss": 0.0151, "step": 60215 }, { "epoch": 18.53, "learning_rate": 2.854766309432888e-07, "loss": 0.003, "step": 60216 }, { "epoch": 18.53, "learning_rate": 2.8535843600026036e-07, "loss": 0.0036, "step": 60217 }, { "epoch": 18.53, "learning_rate": 2.8524026517606775e-07, "loss": 0.0031, "step": 60218 }, { "epoch": 18.53, "learning_rate": 2.85122118471004e-07, "loss": 0.0036, "step": 60219 }, { "epoch": 18.53, "learning_rate": 2.850039958853634e-07, "loss": 0.0034, "step": 60220 }, { "epoch": 18.53, "learning_rate": 2.8488589741943907e-07, "loss": 0.0035, "step": 60221 }, { "epoch": 18.53, "learning_rate": 2.84767823073524e-07, "loss": 0.0032, "step": 60222 }, { "epoch": 18.53, "learning_rate": 2.8464977284790917e-07, "loss": 0.0045, "step": 60223 }, { "epoch": 18.53, "learning_rate": 2.845317467428921e-07, "loss": 0.0027, "step": 60224 }, { "epoch": 18.53, "learning_rate": 2.844137447587636e-07, "loss": 0.0033, "step": 60225 }, { "epoch": 18.53, "learning_rate": 2.842957668958157e-07, "loss": 0.0016, "step": 60226 }, { "epoch": 18.53, "learning_rate": 2.841778131543427e-07, "loss": 0.0036, "step": 60227 }, { "epoch": 18.53, "learning_rate": 2.840598835346353e-07, "loss": 0.0035, "step": 60228 }, { "epoch": 18.53, "learning_rate": 2.83941978036989e-07, "loss": 0.0033, "step": 60229 }, { "epoch": 18.53, "learning_rate": 2.838240966616956e-07, "loss": 0.0033, "step": 60230 }, { "epoch": 18.53, "learning_rate": 2.837062394090473e-07, "loss": 0.0036, "step": 60231 }, { "epoch": 18.53, "learning_rate": 2.83588406279337e-07, "loss": 0.0033, "step": 60232 }, { "epoch": 18.53, "learning_rate": 2.8347059727285687e-07, "loss": 0.0032, "step": 60233 }, { "epoch": 18.53, "learning_rate": 2.8335281238990095e-07, "loss": 0.0035, "step": 60234 }, { "epoch": 18.53, "learning_rate": 2.8323505163075916e-07, "loss": 0.0024, "step": 60235 }, { "epoch": 18.53, "learning_rate": 2.8311731499572557e-07, "loss": 0.0039, "step": 60236 }, { "epoch": 18.53, "learning_rate": 2.829996024850923e-07, "loss": 0.0035, "step": 60237 }, { "epoch": 18.53, "learning_rate": 2.828819140991512e-07, "loss": 0.0021, "step": 60238 }, { "epoch": 18.53, "learning_rate": 2.827642498381955e-07, "loss": 0.0041, "step": 60239 }, { "epoch": 18.54, "learning_rate": 2.8264660970251487e-07, "loss": 0.0018, "step": 60240 }, { "epoch": 18.54, "learning_rate": 2.8252899369240473e-07, "loss": 0.003, "step": 60241 }, { "epoch": 18.54, "learning_rate": 2.824114018081536e-07, "loss": 0.0053, "step": 60242 }, { "epoch": 18.54, "learning_rate": 2.8229383405005584e-07, "loss": 0.0027, "step": 60243 }, { "epoch": 18.54, "learning_rate": 2.8217629041840224e-07, "loss": 0.0015, "step": 60244 }, { "epoch": 18.54, "learning_rate": 2.8205877091348587e-07, "loss": 0.0037, "step": 60245 }, { "epoch": 18.54, "learning_rate": 2.819412755355988e-07, "loss": 0.0033, "step": 60246 }, { "epoch": 18.54, "learning_rate": 2.8182380428503077e-07, "loss": 0.0058, "step": 60247 }, { "epoch": 18.54, "learning_rate": 2.8170635716207263e-07, "loss": 0.0051, "step": 60248 }, { "epoch": 18.54, "learning_rate": 2.815889341670186e-07, "loss": 0.0059, "step": 60249 }, { "epoch": 18.54, "learning_rate": 2.8147153530015957e-07, "loss": 0.0016, "step": 60250 }, { "epoch": 18.54, "learning_rate": 2.813541605617875e-07, "loss": 0.0035, "step": 60251 }, { "epoch": 18.54, "learning_rate": 2.8123680995219226e-07, "loss": 0.0034, "step": 60252 }, { "epoch": 18.54, "learning_rate": 2.8111948347166574e-07, "loss": 0.0038, "step": 60253 }, { "epoch": 18.54, "learning_rate": 2.810021811204988e-07, "loss": 0.0036, "step": 60254 }, { "epoch": 18.54, "learning_rate": 2.808849028989846e-07, "loss": 0.0031, "step": 60255 }, { "epoch": 18.54, "learning_rate": 2.807676488074118e-07, "loss": 0.0034, "step": 60256 }, { "epoch": 18.54, "learning_rate": 2.8065041884607347e-07, "loss": 0.0039, "step": 60257 }, { "epoch": 18.54, "learning_rate": 2.805332130152594e-07, "loss": 0.0035, "step": 60258 }, { "epoch": 18.54, "learning_rate": 2.804160313152615e-07, "loss": 0.0046, "step": 60259 }, { "epoch": 18.54, "learning_rate": 2.802988737463708e-07, "loss": 0.0044, "step": 60260 }, { "epoch": 18.54, "learning_rate": 2.801817403088769e-07, "loss": 0.0045, "step": 60261 }, { "epoch": 18.54, "learning_rate": 2.800646310030719e-07, "loss": 0.0025, "step": 60262 }, { "epoch": 18.54, "learning_rate": 2.799475458292444e-07, "loss": 0.0046, "step": 60263 }, { "epoch": 18.54, "learning_rate": 2.798304847876876e-07, "loss": 0.0029, "step": 60264 }, { "epoch": 18.54, "learning_rate": 2.7971344787869114e-07, "loss": 0.0031, "step": 60265 }, { "epoch": 18.54, "learning_rate": 2.7959643510254597e-07, "loss": 0.004, "step": 60266 }, { "epoch": 18.54, "learning_rate": 2.794794464595441e-07, "loss": 0.0034, "step": 60267 }, { "epoch": 18.54, "learning_rate": 2.793624819499707e-07, "loss": 0.0038, "step": 60268 }, { "epoch": 18.54, "learning_rate": 2.792455415741224e-07, "loss": 0.0039, "step": 60269 }, { "epoch": 18.54, "learning_rate": 2.791286253322856e-07, "loss": 0.0027, "step": 60270 }, { "epoch": 18.54, "learning_rate": 2.790117332247522e-07, "loss": 0.0039, "step": 60271 }, { "epoch": 18.54, "learning_rate": 2.788948652518109e-07, "loss": 0.0049, "step": 60272 }, { "epoch": 18.55, "learning_rate": 2.7877802141375366e-07, "loss": 0.0155, "step": 60273 }, { "epoch": 18.55, "learning_rate": 2.786612017108703e-07, "loss": 0.0064, "step": 60274 }, { "epoch": 18.55, "learning_rate": 2.785444061434506e-07, "loss": 0.0052, "step": 60275 }, { "epoch": 18.55, "learning_rate": 2.7842763471178315e-07, "loss": 0.007, "step": 60276 }, { "epoch": 18.55, "learning_rate": 2.7831088741616e-07, "loss": 0.0053, "step": 60277 }, { "epoch": 18.55, "learning_rate": 2.781941642568686e-07, "loss": 0.0032, "step": 60278 }, { "epoch": 18.55, "learning_rate": 2.780774652342011e-07, "loss": 0.0025, "step": 60279 }, { "epoch": 18.55, "learning_rate": 2.7796079034844716e-07, "loss": 0.0037, "step": 60280 }, { "epoch": 18.55, "learning_rate": 2.778441395998954e-07, "loss": 0.0022, "step": 60281 }, { "epoch": 18.55, "learning_rate": 2.777275129888346e-07, "loss": 0.0037, "step": 60282 }, { "epoch": 18.55, "learning_rate": 2.776109105155556e-07, "loss": 0.0034, "step": 60283 }, { "epoch": 18.55, "learning_rate": 2.7749433218034694e-07, "loss": 0.0016, "step": 60284 }, { "epoch": 18.55, "learning_rate": 2.7737777798349964e-07, "loss": 0.0036, "step": 60285 }, { "epoch": 18.55, "learning_rate": 2.772612479253023e-07, "loss": 0.004, "step": 60286 }, { "epoch": 18.55, "learning_rate": 2.771447420060447e-07, "loss": 0.0042, "step": 60287 }, { "epoch": 18.55, "learning_rate": 2.7702826022601436e-07, "loss": 0.0029, "step": 60288 }, { "epoch": 18.55, "learning_rate": 2.769118025855022e-07, "loss": 0.0028, "step": 60289 }, { "epoch": 18.55, "learning_rate": 2.7679536908479687e-07, "loss": 0.0028, "step": 60290 }, { "epoch": 18.55, "learning_rate": 2.76678959724187e-07, "loss": 0.0032, "step": 60291 }, { "epoch": 18.55, "learning_rate": 2.765625745039624e-07, "loss": 0.0039, "step": 60292 }, { "epoch": 18.55, "learning_rate": 2.7644621342441056e-07, "loss": 0.0033, "step": 60293 }, { "epoch": 18.55, "learning_rate": 2.7632987648582243e-07, "loss": 0.0028, "step": 60294 }, { "epoch": 18.55, "learning_rate": 2.762135636884855e-07, "loss": 0.0036, "step": 60295 }, { "epoch": 18.55, "learning_rate": 2.7609727503268844e-07, "loss": 0.0037, "step": 60296 }, { "epoch": 18.55, "learning_rate": 2.7598101051872104e-07, "loss": 0.0038, "step": 60297 }, { "epoch": 18.55, "learning_rate": 2.758647701468697e-07, "loss": 0.0048, "step": 60298 }, { "epoch": 18.55, "learning_rate": 2.7574855391742537e-07, "loss": 0.0028, "step": 60299 }, { "epoch": 18.55, "learning_rate": 2.7563236183067666e-07, "loss": 0.0031, "step": 60300 }, { "epoch": 18.55, "learning_rate": 2.755161938869122e-07, "loss": 0.0053, "step": 60301 }, { "epoch": 18.55, "learning_rate": 2.7540005008641733e-07, "loss": 0.0024, "step": 60302 }, { "epoch": 18.55, "learning_rate": 2.75283930429483e-07, "loss": 0.0039, "step": 60303 }, { "epoch": 18.55, "learning_rate": 2.7516783491639666e-07, "loss": 0.0031, "step": 60304 }, { "epoch": 18.56, "learning_rate": 2.75051763547447e-07, "loss": 0.0026, "step": 60305 }, { "epoch": 18.56, "learning_rate": 2.749357163229227e-07, "loss": 0.0029, "step": 60306 }, { "epoch": 18.56, "learning_rate": 2.748196932431102e-07, "loss": 0.0018, "step": 60307 }, { "epoch": 18.56, "learning_rate": 2.7470369430829924e-07, "loss": 0.0044, "step": 60308 }, { "epoch": 18.56, "learning_rate": 2.745877195187774e-07, "loss": 0.0048, "step": 60309 }, { "epoch": 18.56, "learning_rate": 2.744717688748322e-07, "loss": 0.0033, "step": 60310 }, { "epoch": 18.56, "learning_rate": 2.743558423767512e-07, "loss": 0.004, "step": 60311 }, { "epoch": 18.56, "learning_rate": 2.7423994002482304e-07, "loss": 0.0083, "step": 60312 }, { "epoch": 18.56, "learning_rate": 2.7412406181933416e-07, "loss": 0.004, "step": 60313 }, { "epoch": 18.56, "learning_rate": 2.7400820776057546e-07, "loss": 0.0032, "step": 60314 }, { "epoch": 18.56, "learning_rate": 2.7389237784883116e-07, "loss": 0.0029, "step": 60315 }, { "epoch": 18.56, "learning_rate": 2.737765720843899e-07, "loss": 0.0027, "step": 60316 }, { "epoch": 18.56, "learning_rate": 2.7366079046753925e-07, "loss": 0.0048, "step": 60317 }, { "epoch": 18.56, "learning_rate": 2.735450329985656e-07, "loss": 0.0034, "step": 60318 }, { "epoch": 18.56, "learning_rate": 2.734292996777588e-07, "loss": 0.0026, "step": 60319 }, { "epoch": 18.56, "learning_rate": 2.733135905054052e-07, "loss": 0.0034, "step": 60320 }, { "epoch": 18.56, "learning_rate": 2.7319790548179124e-07, "loss": 0.0038, "step": 60321 }, { "epoch": 18.56, "learning_rate": 2.730822446072057e-07, "loss": 0.0035, "step": 60322 }, { "epoch": 18.56, "learning_rate": 2.729666078819326e-07, "loss": 0.0026, "step": 60323 }, { "epoch": 18.56, "learning_rate": 2.7285099530626194e-07, "loss": 0.0032, "step": 60324 }, { "epoch": 18.56, "learning_rate": 2.7273540688047995e-07, "loss": 0.0022, "step": 60325 }, { "epoch": 18.56, "learning_rate": 2.726198426048732e-07, "loss": 0.0024, "step": 60326 }, { "epoch": 18.56, "learning_rate": 2.7250430247972915e-07, "loss": 0.0025, "step": 60327 }, { "epoch": 18.56, "learning_rate": 2.723887865053354e-07, "loss": 0.0033, "step": 60328 }, { "epoch": 18.56, "learning_rate": 2.7227329468197616e-07, "loss": 0.0031, "step": 60329 }, { "epoch": 18.56, "learning_rate": 2.721578270099412e-07, "loss": 0.0023, "step": 60330 }, { "epoch": 18.56, "learning_rate": 2.7204238348951586e-07, "loss": 0.0035, "step": 60331 }, { "epoch": 18.56, "learning_rate": 2.719269641209854e-07, "loss": 0.0045, "step": 60332 }, { "epoch": 18.56, "learning_rate": 2.7181156890463745e-07, "loss": 0.0067, "step": 60333 }, { "epoch": 18.56, "learning_rate": 2.716961978407595e-07, "loss": 0.0029, "step": 60334 }, { "epoch": 18.56, "learning_rate": 2.7158085092963916e-07, "loss": 0.0019, "step": 60335 }, { "epoch": 18.56, "learning_rate": 2.7146552817155944e-07, "loss": 0.0025, "step": 60336 }, { "epoch": 18.56, "learning_rate": 2.71350229566808e-07, "loss": 0.0035, "step": 60337 }, { "epoch": 18.57, "learning_rate": 2.71234955115669e-07, "loss": 0.0019, "step": 60338 }, { "epoch": 18.57, "learning_rate": 2.711197048184333e-07, "loss": 0.0029, "step": 60339 }, { "epoch": 18.57, "learning_rate": 2.71004478675384e-07, "loss": 0.0047, "step": 60340 }, { "epoch": 18.57, "learning_rate": 2.7088927668680765e-07, "loss": 0.0033, "step": 60341 }, { "epoch": 18.57, "learning_rate": 2.707740988529906e-07, "loss": 0.0033, "step": 60342 }, { "epoch": 18.57, "learning_rate": 2.706589451742181e-07, "loss": 0.003, "step": 60343 }, { "epoch": 18.57, "learning_rate": 2.705438156507767e-07, "loss": 0.0038, "step": 60344 }, { "epoch": 18.57, "learning_rate": 2.704287102829528e-07, "loss": 0.0032, "step": 60345 }, { "epoch": 18.57, "learning_rate": 2.7031362907103067e-07, "loss": 0.0044, "step": 60346 }, { "epoch": 18.57, "learning_rate": 2.7019857201529665e-07, "loss": 0.0024, "step": 60347 }, { "epoch": 18.57, "learning_rate": 2.700835391160361e-07, "loss": 0.0033, "step": 60348 }, { "epoch": 18.57, "learning_rate": 2.6996853037353665e-07, "loss": 0.0024, "step": 60349 }, { "epoch": 18.57, "learning_rate": 2.698535457880802e-07, "loss": 0.0056, "step": 60350 }, { "epoch": 18.57, "learning_rate": 2.6973858535995547e-07, "loss": 0.0036, "step": 60351 }, { "epoch": 18.57, "learning_rate": 2.6962364908944663e-07, "loss": 0.0027, "step": 60352 }, { "epoch": 18.57, "learning_rate": 2.695087369768379e-07, "loss": 0.005, "step": 60353 }, { "epoch": 18.57, "learning_rate": 2.693938490224168e-07, "loss": 0.0033, "step": 60354 }, { "epoch": 18.57, "learning_rate": 2.6927898522646655e-07, "loss": 0.0037, "step": 60355 }, { "epoch": 18.57, "learning_rate": 2.6916414558927575e-07, "loss": 0.0019, "step": 60356 }, { "epoch": 18.57, "learning_rate": 2.690493301111252e-07, "loss": 0.0031, "step": 60357 }, { "epoch": 18.57, "learning_rate": 2.6893453879230034e-07, "loss": 0.0051, "step": 60358 }, { "epoch": 18.57, "learning_rate": 2.688197716330887e-07, "loss": 0.0025, "step": 60359 }, { "epoch": 18.57, "learning_rate": 2.6870502863377445e-07, "loss": 0.0034, "step": 60360 }, { "epoch": 18.57, "learning_rate": 2.6859030979464296e-07, "loss": 0.0024, "step": 60361 }, { "epoch": 18.57, "learning_rate": 2.684756151159773e-07, "loss": 0.0032, "step": 60362 }, { "epoch": 18.57, "learning_rate": 2.6836094459806284e-07, "loss": 0.0031, "step": 60363 }, { "epoch": 18.57, "learning_rate": 2.6824629824118485e-07, "loss": 0.006, "step": 60364 }, { "epoch": 18.57, "learning_rate": 2.681316760456287e-07, "loss": 0.0053, "step": 60365 }, { "epoch": 18.57, "learning_rate": 2.6801707801167643e-07, "loss": 0.0039, "step": 60366 }, { "epoch": 18.57, "learning_rate": 2.679025041396155e-07, "loss": 0.0044, "step": 60367 }, { "epoch": 18.57, "learning_rate": 2.6778795442972685e-07, "loss": 0.0026, "step": 60368 }, { "epoch": 18.57, "learning_rate": 2.6767342888229907e-07, "loss": 0.0028, "step": 60369 }, { "epoch": 18.58, "learning_rate": 2.6755892749761425e-07, "loss": 0.0048, "step": 60370 }, { "epoch": 18.58, "learning_rate": 2.6744445027595657e-07, "loss": 0.0029, "step": 60371 }, { "epoch": 18.58, "learning_rate": 2.673299972176102e-07, "loss": 0.0041, "step": 60372 }, { "epoch": 18.58, "learning_rate": 2.6721556832285836e-07, "loss": 0.005, "step": 60373 }, { "epoch": 18.58, "learning_rate": 2.6710116359198845e-07, "loss": 0.004, "step": 60374 }, { "epoch": 18.58, "learning_rate": 2.6698678302528146e-07, "loss": 0.004, "step": 60375 }, { "epoch": 18.58, "learning_rate": 2.668724266230227e-07, "loss": 0.0037, "step": 60376 }, { "epoch": 18.58, "learning_rate": 2.667580943854964e-07, "loss": 0.0048, "step": 60377 }, { "epoch": 18.58, "learning_rate": 2.6664378631298337e-07, "loss": 0.0032, "step": 60378 }, { "epoch": 18.58, "learning_rate": 2.6652950240577234e-07, "loss": 0.0071, "step": 60379 }, { "epoch": 18.58, "learning_rate": 2.664152426641431e-07, "loss": 0.0056, "step": 60380 }, { "epoch": 18.58, "learning_rate": 2.663010070883809e-07, "loss": 0.0053, "step": 60381 }, { "epoch": 18.58, "learning_rate": 2.6618679567877006e-07, "loss": 0.0033, "step": 60382 }, { "epoch": 18.58, "learning_rate": 2.6607260843559245e-07, "loss": 0.0025, "step": 60383 }, { "epoch": 18.58, "learning_rate": 2.659584453591324e-07, "loss": 0.0038, "step": 60384 }, { "epoch": 18.58, "learning_rate": 2.6584430644967296e-07, "loss": 0.002, "step": 60385 }, { "epoch": 18.58, "learning_rate": 2.6573019170749834e-07, "loss": 0.0032, "step": 60386 }, { "epoch": 18.58, "learning_rate": 2.656161011328917e-07, "loss": 0.0046, "step": 60387 }, { "epoch": 18.58, "learning_rate": 2.65502034726135e-07, "loss": 0.0035, "step": 60388 }, { "epoch": 18.58, "learning_rate": 2.6538799248751245e-07, "loss": 0.0027, "step": 60389 }, { "epoch": 18.58, "learning_rate": 2.652739744173083e-07, "loss": 0.0052, "step": 60390 }, { "epoch": 18.58, "learning_rate": 2.651599805158045e-07, "loss": 0.0033, "step": 60391 }, { "epoch": 18.58, "learning_rate": 2.650460107832842e-07, "loss": 0.0023, "step": 60392 }, { "epoch": 18.58, "learning_rate": 2.6493206522002825e-07, "loss": 0.0031, "step": 60393 }, { "epoch": 18.58, "learning_rate": 2.648181438263231e-07, "loss": 0.0033, "step": 60394 }, { "epoch": 18.58, "learning_rate": 2.647042466024485e-07, "loss": 0.0039, "step": 60395 }, { "epoch": 18.58, "learning_rate": 2.645903735486899e-07, "loss": 0.0027, "step": 60396 }, { "epoch": 18.58, "learning_rate": 2.644765246653291e-07, "loss": 0.0051, "step": 60397 }, { "epoch": 18.58, "learning_rate": 2.643626999526483e-07, "loss": 0.0029, "step": 60398 }, { "epoch": 18.58, "learning_rate": 2.6424889941092933e-07, "loss": 0.0036, "step": 60399 }, { "epoch": 18.58, "learning_rate": 2.6413512304045654e-07, "loss": 0.0034, "step": 60400 }, { "epoch": 18.58, "learning_rate": 2.640213708415118e-07, "loss": 0.0037, "step": 60401 }, { "epoch": 18.58, "learning_rate": 2.639076428143772e-07, "loss": 0.0039, "step": 60402 }, { "epoch": 18.59, "learning_rate": 2.637939389593336e-07, "loss": 0.0055, "step": 60403 }, { "epoch": 18.59, "learning_rate": 2.636802592766663e-07, "loss": 0.0049, "step": 60404 }, { "epoch": 18.59, "learning_rate": 2.6356660376665623e-07, "loss": 0.0024, "step": 60405 }, { "epoch": 18.59, "learning_rate": 2.6345297242958533e-07, "loss": 0.0031, "step": 60406 }, { "epoch": 18.59, "learning_rate": 2.633393652657357e-07, "loss": 0.0015, "step": 60407 }, { "epoch": 18.59, "learning_rate": 2.632257822753881e-07, "loss": 0.003, "step": 60408 }, { "epoch": 18.59, "learning_rate": 2.631122234588268e-07, "loss": 0.0025, "step": 60409 }, { "epoch": 18.59, "learning_rate": 2.6299868881633273e-07, "loss": 0.0029, "step": 60410 }, { "epoch": 18.59, "learning_rate": 2.6288517834818896e-07, "loss": 0.0038, "step": 60411 }, { "epoch": 18.59, "learning_rate": 2.6277169205467636e-07, "loss": 0.0047, "step": 60412 }, { "epoch": 18.59, "learning_rate": 2.626582299360747e-07, "loss": 0.0029, "step": 60413 }, { "epoch": 18.59, "learning_rate": 2.6254479199266934e-07, "loss": 0.004, "step": 60414 }, { "epoch": 18.59, "learning_rate": 2.62431378224739e-07, "loss": 0.0036, "step": 60415 }, { "epoch": 18.59, "learning_rate": 2.6231798863256663e-07, "loss": 0.0042, "step": 60416 }, { "epoch": 18.59, "learning_rate": 2.622046232164344e-07, "loss": 0.0032, "step": 60417 }, { "epoch": 18.59, "learning_rate": 2.620912819766219e-07, "loss": 0.0022, "step": 60418 }, { "epoch": 18.59, "learning_rate": 2.619779649134124e-07, "loss": 0.0026, "step": 60419 }, { "epoch": 18.59, "learning_rate": 2.618646720270857e-07, "loss": 0.0032, "step": 60420 }, { "epoch": 18.59, "learning_rate": 2.617514033179236e-07, "loss": 0.0033, "step": 60421 }, { "epoch": 18.59, "learning_rate": 2.6163815878620713e-07, "loss": 0.0047, "step": 60422 }, { "epoch": 18.59, "learning_rate": 2.6152493843221714e-07, "loss": 0.0076, "step": 60423 }, { "epoch": 18.59, "learning_rate": 2.6141174225623676e-07, "loss": 0.0031, "step": 60424 }, { "epoch": 18.59, "learning_rate": 2.612985702585458e-07, "loss": 0.0052, "step": 60425 }, { "epoch": 18.59, "learning_rate": 2.611854224394239e-07, "loss": 0.0046, "step": 60426 }, { "epoch": 18.59, "learning_rate": 2.610722987991543e-07, "loss": 0.0031, "step": 60427 }, { "epoch": 18.59, "learning_rate": 2.609591993380145e-07, "loss": 0.0032, "step": 60428 }, { "epoch": 18.59, "learning_rate": 2.608461240562887e-07, "loss": 0.0032, "step": 60429 }, { "epoch": 18.59, "learning_rate": 2.6073307295425674e-07, "loss": 0.0038, "step": 60430 }, { "epoch": 18.59, "learning_rate": 2.606200460321984e-07, "loss": 0.0035, "step": 60431 }, { "epoch": 18.59, "learning_rate": 2.6050704329039557e-07, "loss": 0.0058, "step": 60432 }, { "epoch": 18.59, "learning_rate": 2.60394064729127e-07, "loss": 0.0052, "step": 60433 }, { "epoch": 18.59, "learning_rate": 2.6028111034867466e-07, "loss": 0.0024, "step": 60434 }, { "epoch": 18.6, "learning_rate": 2.6016818014931946e-07, "loss": 0.0076, "step": 60435 }, { "epoch": 18.6, "learning_rate": 2.6005527413134e-07, "loss": 0.0026, "step": 60436 }, { "epoch": 18.6, "learning_rate": 2.5994239229501727e-07, "loss": 0.0044, "step": 60437 }, { "epoch": 18.6, "learning_rate": 2.598295346406321e-07, "loss": 0.0023, "step": 60438 }, { "epoch": 18.6, "learning_rate": 2.5971670116846537e-07, "loss": 0.0053, "step": 60439 }, { "epoch": 18.6, "learning_rate": 2.596038918787946e-07, "loss": 0.0027, "step": 60440 }, { "epoch": 18.6, "learning_rate": 2.594911067719019e-07, "loss": 0.0052, "step": 60441 }, { "epoch": 18.6, "learning_rate": 2.59378345848067e-07, "loss": 0.004, "step": 60442 }, { "epoch": 18.6, "learning_rate": 2.592656091075685e-07, "loss": 0.0036, "step": 60443 }, { "epoch": 18.6, "learning_rate": 2.5915289655068953e-07, "loss": 0.002, "step": 60444 }, { "epoch": 18.6, "learning_rate": 2.590402081777066e-07, "loss": 0.0044, "step": 60445 }, { "epoch": 18.6, "learning_rate": 2.5892754398890163e-07, "loss": 0.0033, "step": 60446 }, { "epoch": 18.6, "learning_rate": 2.588149039845533e-07, "loss": 0.0028, "step": 60447 }, { "epoch": 18.6, "learning_rate": 2.5870228816494034e-07, "loss": 0.0032, "step": 60448 }, { "epoch": 18.6, "learning_rate": 2.585896965303436e-07, "loss": 0.0049, "step": 60449 }, { "epoch": 18.6, "learning_rate": 2.584771290810428e-07, "loss": 0.0031, "step": 60450 }, { "epoch": 18.6, "learning_rate": 2.583645858173178e-07, "loss": 0.0052, "step": 60451 }, { "epoch": 18.6, "learning_rate": 2.5825206673944616e-07, "loss": 0.0055, "step": 60452 }, { "epoch": 18.6, "learning_rate": 2.5813957184770977e-07, "loss": 0.003, "step": 60453 }, { "epoch": 18.6, "learning_rate": 2.580271011423852e-07, "loss": 0.0027, "step": 60454 }, { "epoch": 18.6, "learning_rate": 2.579146546237532e-07, "loss": 0.004, "step": 60455 }, { "epoch": 18.6, "learning_rate": 2.578022322920926e-07, "loss": 0.0038, "step": 60456 }, { "epoch": 18.6, "learning_rate": 2.57689834147683e-07, "loss": 0.0029, "step": 60457 }, { "epoch": 18.6, "learning_rate": 2.575774601908021e-07, "loss": 0.0023, "step": 60458 }, { "epoch": 18.6, "learning_rate": 2.574651104217318e-07, "loss": 0.0027, "step": 60459 }, { "epoch": 18.6, "learning_rate": 2.5735278484074865e-07, "loss": 0.0051, "step": 60460 }, { "epoch": 18.6, "learning_rate": 2.5724048344813125e-07, "loss": 0.0039, "step": 60461 }, { "epoch": 18.6, "learning_rate": 2.5712820624415934e-07, "loss": 0.0034, "step": 60462 }, { "epoch": 18.6, "learning_rate": 2.5701595322911055e-07, "loss": 0.0045, "step": 60463 }, { "epoch": 18.6, "learning_rate": 2.569037244032657e-07, "loss": 0.0034, "step": 60464 }, { "epoch": 18.6, "learning_rate": 2.5679151976690243e-07, "loss": 0.0018, "step": 60465 }, { "epoch": 18.6, "learning_rate": 2.566793393202982e-07, "loss": 0.0032, "step": 60466 }, { "epoch": 18.6, "learning_rate": 2.5656718306373396e-07, "loss": 0.0018, "step": 60467 }, { "epoch": 18.61, "learning_rate": 2.56455050997485e-07, "loss": 0.0023, "step": 60468 }, { "epoch": 18.61, "learning_rate": 2.5634294312183226e-07, "loss": 0.0021, "step": 60469 }, { "epoch": 18.61, "learning_rate": 2.562308594370533e-07, "loss": 0.0028, "step": 60470 }, { "epoch": 18.61, "learning_rate": 2.561187999434256e-07, "loss": 0.0022, "step": 60471 }, { "epoch": 18.61, "learning_rate": 2.5600676464122897e-07, "loss": 0.0041, "step": 60472 }, { "epoch": 18.61, "learning_rate": 2.5589475353073987e-07, "loss": 0.0025, "step": 60473 }, { "epoch": 18.61, "learning_rate": 2.557827666122381e-07, "loss": 0.0047, "step": 60474 }, { "epoch": 18.61, "learning_rate": 2.556708038860001e-07, "loss": 0.0019, "step": 60475 }, { "epoch": 18.61, "learning_rate": 2.5555886535230446e-07, "loss": 0.0065, "step": 60476 }, { "epoch": 18.61, "learning_rate": 2.554469510114288e-07, "loss": 0.0017, "step": 60477 }, { "epoch": 18.61, "learning_rate": 2.5533506086365066e-07, "loss": 0.0019, "step": 60478 }, { "epoch": 18.61, "learning_rate": 2.5522319490924987e-07, "loss": 0.0027, "step": 60479 }, { "epoch": 18.61, "learning_rate": 2.5511135314850277e-07, "loss": 0.0044, "step": 60480 }, { "epoch": 18.61, "learning_rate": 2.54999535581687e-07, "loss": 0.0031, "step": 60481 }, { "epoch": 18.61, "learning_rate": 2.5488774220908007e-07, "loss": 0.0034, "step": 60482 }, { "epoch": 18.61, "learning_rate": 2.547759730309585e-07, "loss": 0.0042, "step": 60483 }, { "epoch": 18.61, "learning_rate": 2.5466422804760194e-07, "loss": 0.0038, "step": 60484 }, { "epoch": 18.61, "learning_rate": 2.5455250725928695e-07, "loss": 0.0049, "step": 60485 }, { "epoch": 18.61, "learning_rate": 2.54440810666291e-07, "loss": 0.003, "step": 60486 }, { "epoch": 18.61, "learning_rate": 2.5432913826889063e-07, "loss": 0.0029, "step": 60487 }, { "epoch": 18.61, "learning_rate": 2.5421749006736327e-07, "loss": 0.0026, "step": 60488 }, { "epoch": 18.61, "learning_rate": 2.5410586606198775e-07, "loss": 0.0043, "step": 60489 }, { "epoch": 18.61, "learning_rate": 2.5399426625303926e-07, "loss": 0.0031, "step": 60490 }, { "epoch": 18.61, "learning_rate": 2.538826906407954e-07, "loss": 0.0028, "step": 60491 }, { "epoch": 18.61, "learning_rate": 2.5377113922553266e-07, "loss": 0.0031, "step": 60492 }, { "epoch": 18.61, "learning_rate": 2.5365961200753074e-07, "loss": 0.0028, "step": 60493 }, { "epoch": 18.61, "learning_rate": 2.5354810898706397e-07, "loss": 0.004, "step": 60494 }, { "epoch": 18.61, "learning_rate": 2.5343663016440865e-07, "loss": 0.0034, "step": 60495 }, { "epoch": 18.61, "learning_rate": 2.533251755398425e-07, "loss": 0.004, "step": 60496 }, { "epoch": 18.61, "learning_rate": 2.532137451136418e-07, "loss": 0.0044, "step": 60497 }, { "epoch": 18.61, "learning_rate": 2.5310233888608537e-07, "loss": 0.0082, "step": 60498 }, { "epoch": 18.61, "learning_rate": 2.5299095685744734e-07, "loss": 0.0051, "step": 60499 }, { "epoch": 18.62, "learning_rate": 2.528795990280053e-07, "loss": 0.0043, "step": 60500 }, { "epoch": 18.62, "learning_rate": 2.527682653980368e-07, "loss": 0.0038, "step": 60501 }, { "epoch": 18.62, "learning_rate": 2.526569559678138e-07, "loss": 0.0045, "step": 60502 }, { "epoch": 18.62, "learning_rate": 2.5254567073761835e-07, "loss": 0.0036, "step": 60503 }, { "epoch": 18.62, "learning_rate": 2.524344097077225e-07, "loss": 0.0045, "step": 60504 }, { "epoch": 18.62, "learning_rate": 2.5232317287840593e-07, "loss": 0.0047, "step": 60505 }, { "epoch": 18.62, "learning_rate": 2.5221196024994176e-07, "loss": 0.0041, "step": 60506 }, { "epoch": 18.62, "learning_rate": 2.521007718226076e-07, "loss": 0.0033, "step": 60507 }, { "epoch": 18.62, "learning_rate": 2.519896075966799e-07, "loss": 0.0015, "step": 60508 }, { "epoch": 18.62, "learning_rate": 2.5187846757243286e-07, "loss": 0.0049, "step": 60509 }, { "epoch": 18.62, "learning_rate": 2.51767351750144e-07, "loss": 0.0048, "step": 60510 }, { "epoch": 18.62, "learning_rate": 2.516562601300898e-07, "loss": 0.0025, "step": 60511 }, { "epoch": 18.62, "learning_rate": 2.515451927125423e-07, "loss": 0.0058, "step": 60512 }, { "epoch": 18.62, "learning_rate": 2.5143414949778235e-07, "loss": 0.0077, "step": 60513 }, { "epoch": 18.62, "learning_rate": 2.513231304860841e-07, "loss": 0.0023, "step": 60514 }, { "epoch": 18.62, "learning_rate": 2.5121213567772085e-07, "loss": 0.0049, "step": 60515 }, { "epoch": 18.62, "learning_rate": 2.5110116507297e-07, "loss": 0.008, "step": 60516 }, { "epoch": 18.62, "learning_rate": 2.509902186721058e-07, "loss": 0.0043, "step": 60517 }, { "epoch": 18.62, "learning_rate": 2.508792964754048e-07, "loss": 0.0025, "step": 60518 }, { "epoch": 18.62, "learning_rate": 2.507683984831433e-07, "loss": 0.0046, "step": 60519 }, { "epoch": 18.62, "learning_rate": 2.506575246955945e-07, "loss": 0.0034, "step": 60520 }, { "epoch": 18.62, "learning_rate": 2.505466751130359e-07, "loss": 0.003, "step": 60521 }, { "epoch": 18.62, "learning_rate": 2.5043584973574064e-07, "loss": 0.0056, "step": 60522 }, { "epoch": 18.62, "learning_rate": 2.503250485639852e-07, "loss": 0.0027, "step": 60523 }, { "epoch": 18.62, "learning_rate": 2.502142715980449e-07, "loss": 0.0031, "step": 60524 }, { "epoch": 18.62, "learning_rate": 2.5010351883819283e-07, "loss": 0.0026, "step": 60525 }, { "epoch": 18.62, "learning_rate": 2.4999279028470544e-07, "loss": 0.0026, "step": 60526 }, { "epoch": 18.62, "learning_rate": 2.49882085937857e-07, "loss": 0.0053, "step": 60527 }, { "epoch": 18.62, "learning_rate": 2.49771405797925e-07, "loss": 0.0049, "step": 60528 }, { "epoch": 18.62, "learning_rate": 2.496607498651804e-07, "loss": 0.0031, "step": 60529 }, { "epoch": 18.62, "learning_rate": 2.495501181398996e-07, "loss": 0.0033, "step": 60530 }, { "epoch": 18.62, "learning_rate": 2.4943951062235683e-07, "loss": 0.0028, "step": 60531 }, { "epoch": 18.62, "learning_rate": 2.493289273128274e-07, "loss": 0.0023, "step": 60532 }, { "epoch": 18.63, "learning_rate": 2.492183682115856e-07, "loss": 0.0027, "step": 60533 }, { "epoch": 18.63, "learning_rate": 2.4910783331890564e-07, "loss": 0.0039, "step": 60534 }, { "epoch": 18.63, "learning_rate": 2.4899732263506395e-07, "loss": 0.0029, "step": 60535 }, { "epoch": 18.63, "learning_rate": 2.488868361603314e-07, "loss": 0.003, "step": 60536 }, { "epoch": 18.63, "learning_rate": 2.4877637389498333e-07, "loss": 0.0026, "step": 60537 }, { "epoch": 18.63, "learning_rate": 2.486659358392951e-07, "loss": 0.0043, "step": 60538 }, { "epoch": 18.63, "learning_rate": 2.4855552199354096e-07, "loss": 0.0036, "step": 60539 }, { "epoch": 18.63, "learning_rate": 2.4844513235799504e-07, "loss": 0.0039, "step": 60540 }, { "epoch": 18.63, "learning_rate": 2.483347669329306e-07, "loss": 0.0035, "step": 60541 }, { "epoch": 18.63, "learning_rate": 2.482244257186217e-07, "loss": 0.0026, "step": 60542 }, { "epoch": 18.63, "learning_rate": 2.481141087153427e-07, "loss": 0.003, "step": 60543 }, { "epoch": 18.63, "learning_rate": 2.4800381592336773e-07, "loss": 0.0037, "step": 60544 }, { "epoch": 18.63, "learning_rate": 2.4789354734297e-07, "loss": 0.0048, "step": 60545 }, { "epoch": 18.63, "learning_rate": 2.4778330297442255e-07, "loss": 0.0073, "step": 60546 }, { "epoch": 18.63, "learning_rate": 2.4767308281800074e-07, "loss": 0.006, "step": 60547 }, { "epoch": 18.63, "learning_rate": 2.4756288687397655e-07, "loss": 0.0033, "step": 60548 }, { "epoch": 18.63, "learning_rate": 2.474527151426276e-07, "loss": 0.0069, "step": 60549 }, { "epoch": 18.63, "learning_rate": 2.473425676242214e-07, "loss": 0.0031, "step": 60550 }, { "epoch": 18.63, "learning_rate": 2.472324443190355e-07, "loss": 0.0046, "step": 60551 }, { "epoch": 18.63, "learning_rate": 2.471223452273419e-07, "loss": 0.0032, "step": 60552 }, { "epoch": 18.63, "learning_rate": 2.4701227034941376e-07, "loss": 0.0039, "step": 60553 }, { "epoch": 18.63, "learning_rate": 2.469022196855253e-07, "loss": 0.0038, "step": 60554 }, { "epoch": 18.63, "learning_rate": 2.467921932359496e-07, "loss": 0.0048, "step": 60555 }, { "epoch": 18.63, "learning_rate": 2.466821910009598e-07, "loss": 0.0047, "step": 60556 }, { "epoch": 18.63, "learning_rate": 2.465722129808279e-07, "loss": 0.0034, "step": 60557 }, { "epoch": 18.63, "learning_rate": 2.464622591758281e-07, "loss": 0.0035, "step": 60558 }, { "epoch": 18.63, "learning_rate": 2.4635232958623247e-07, "loss": 0.0047, "step": 60559 }, { "epoch": 18.63, "learning_rate": 2.462424242123151e-07, "loss": 0.0025, "step": 60560 }, { "epoch": 18.63, "learning_rate": 2.461325430543482e-07, "loss": 0.0043, "step": 60561 }, { "epoch": 18.63, "learning_rate": 2.460226861126047e-07, "loss": 0.0044, "step": 60562 }, { "epoch": 18.63, "learning_rate": 2.459128533873567e-07, "loss": 0.0035, "step": 60563 }, { "epoch": 18.63, "learning_rate": 2.458030448788784e-07, "loss": 0.0032, "step": 60564 }, { "epoch": 18.64, "learning_rate": 2.456932605874407e-07, "loss": 0.0074, "step": 60565 }, { "epoch": 18.64, "learning_rate": 2.455835005133178e-07, "loss": 0.0025, "step": 60566 }, { "epoch": 18.64, "learning_rate": 2.4547376465677954e-07, "loss": 0.0041, "step": 60567 }, { "epoch": 18.64, "learning_rate": 2.4536405301810115e-07, "loss": 0.0046, "step": 60568 }, { "epoch": 18.64, "learning_rate": 2.4525436559755475e-07, "loss": 0.0025, "step": 60569 }, { "epoch": 18.64, "learning_rate": 2.4514470239541234e-07, "loss": 0.0044, "step": 60570 }, { "epoch": 18.64, "learning_rate": 2.450350634119458e-07, "loss": 0.0049, "step": 60571 }, { "epoch": 18.64, "learning_rate": 2.4492544864742504e-07, "loss": 0.0025, "step": 60572 }, { "epoch": 18.64, "learning_rate": 2.448158581021265e-07, "loss": 0.0025, "step": 60573 }, { "epoch": 18.64, "learning_rate": 2.447062917763199e-07, "loss": 0.0036, "step": 60574 }, { "epoch": 18.64, "learning_rate": 2.4459674967027725e-07, "loss": 0.003, "step": 60575 }, { "epoch": 18.64, "learning_rate": 2.4448723178427056e-07, "loss": 0.0029, "step": 60576 }, { "epoch": 18.64, "learning_rate": 2.4437773811857304e-07, "loss": 0.0022, "step": 60577 }, { "epoch": 18.64, "learning_rate": 2.4426826867345543e-07, "loss": 0.0021, "step": 60578 }, { "epoch": 18.64, "learning_rate": 2.4415882344918983e-07, "loss": 0.003, "step": 60579 }, { "epoch": 18.64, "learning_rate": 2.4404940244604715e-07, "loss": 0.0039, "step": 60580 }, { "epoch": 18.64, "learning_rate": 2.439400056642993e-07, "loss": 0.0031, "step": 60581 }, { "epoch": 18.64, "learning_rate": 2.438306331042184e-07, "loss": 0.0036, "step": 60582 }, { "epoch": 18.64, "learning_rate": 2.4372128476607747e-07, "loss": 0.0037, "step": 60583 }, { "epoch": 18.64, "learning_rate": 2.4361196065014523e-07, "loss": 0.0036, "step": 60584 }, { "epoch": 18.64, "learning_rate": 2.435026607566937e-07, "loss": 0.0042, "step": 60585 }, { "epoch": 18.64, "learning_rate": 2.43393385085996e-07, "loss": 0.0033, "step": 60586 }, { "epoch": 18.64, "learning_rate": 2.432841336383207e-07, "loss": 0.0029, "step": 60587 }, { "epoch": 18.64, "learning_rate": 2.43174906413941e-07, "loss": 0.0033, "step": 60588 }, { "epoch": 18.64, "learning_rate": 2.43065703413129e-07, "loss": 0.0037, "step": 60589 }, { "epoch": 18.64, "learning_rate": 2.429565246361532e-07, "loss": 0.0027, "step": 60590 }, { "epoch": 18.64, "learning_rate": 2.4284737008328676e-07, "loss": 0.0047, "step": 60591 }, { "epoch": 18.64, "learning_rate": 2.427382397547984e-07, "loss": 0.0032, "step": 60592 }, { "epoch": 18.64, "learning_rate": 2.426291336509623e-07, "loss": 0.0053, "step": 60593 }, { "epoch": 18.64, "learning_rate": 2.425200517720461e-07, "loss": 0.0034, "step": 60594 }, { "epoch": 18.64, "learning_rate": 2.424109941183228e-07, "loss": 0.0048, "step": 60595 }, { "epoch": 18.64, "learning_rate": 2.423019606900623e-07, "loss": 0.0049, "step": 60596 }, { "epoch": 18.64, "learning_rate": 2.421929514875354e-07, "loss": 0.003, "step": 60597 }, { "epoch": 18.65, "learning_rate": 2.4208396651101307e-07, "loss": 0.0031, "step": 60598 }, { "epoch": 18.65, "learning_rate": 2.419750057607662e-07, "loss": 0.0036, "step": 60599 }, { "epoch": 18.65, "learning_rate": 2.4186606923706336e-07, "loss": 0.0059, "step": 60600 }, { "epoch": 18.65, "learning_rate": 2.4175715694017663e-07, "loss": 0.0031, "step": 60601 }, { "epoch": 18.65, "learning_rate": 2.4164826887037587e-07, "loss": 0.0041, "step": 60602 }, { "epoch": 18.65, "learning_rate": 2.4153940502793185e-07, "loss": 0.0029, "step": 60603 }, { "epoch": 18.65, "learning_rate": 2.414305654131166e-07, "loss": 0.0051, "step": 60604 }, { "epoch": 18.65, "learning_rate": 2.413217500261966e-07, "loss": 0.0021, "step": 60605 }, { "epoch": 18.65, "learning_rate": 2.41212958867445e-07, "loss": 0.0021, "step": 60606 }, { "epoch": 18.65, "learning_rate": 2.411041919371282e-07, "loss": 0.003, "step": 60607 }, { "epoch": 18.65, "learning_rate": 2.4099544923552043e-07, "loss": 0.0041, "step": 60608 }, { "epoch": 18.65, "learning_rate": 2.4088673076289037e-07, "loss": 0.0035, "step": 60609 }, { "epoch": 18.65, "learning_rate": 2.4077803651950673e-07, "loss": 0.0019, "step": 60610 }, { "epoch": 18.65, "learning_rate": 2.4066936650564254e-07, "loss": 0.0037, "step": 60611 }, { "epoch": 18.65, "learning_rate": 2.405607207215621e-07, "loss": 0.0027, "step": 60612 }, { "epoch": 18.65, "learning_rate": 2.404520991675396e-07, "loss": 0.0135, "step": 60613 }, { "epoch": 18.65, "learning_rate": 2.403435018438438e-07, "loss": 0.0042, "step": 60614 }, { "epoch": 18.65, "learning_rate": 2.4023492875074326e-07, "loss": 0.0041, "step": 60615 }, { "epoch": 18.65, "learning_rate": 2.4012637988850894e-07, "loss": 0.003, "step": 60616 }, { "epoch": 18.65, "learning_rate": 2.400178552574095e-07, "loss": 0.0039, "step": 60617 }, { "epoch": 18.65, "learning_rate": 2.3990935485771474e-07, "loss": 0.0039, "step": 60618 }, { "epoch": 18.65, "learning_rate": 2.3980087868969327e-07, "loss": 0.0012, "step": 60619 }, { "epoch": 18.65, "learning_rate": 2.396924267536149e-07, "loss": 0.0034, "step": 60620 }, { "epoch": 18.65, "learning_rate": 2.395839990497495e-07, "loss": 0.0031, "step": 60621 }, { "epoch": 18.65, "learning_rate": 2.394755955783645e-07, "loss": 0.0057, "step": 60622 }, { "epoch": 18.65, "learning_rate": 2.393672163397309e-07, "loss": 0.0036, "step": 60623 }, { "epoch": 18.65, "learning_rate": 2.3925886133411737e-07, "loss": 0.0036, "step": 60624 }, { "epoch": 18.65, "learning_rate": 2.391505305617936e-07, "loss": 0.0076, "step": 60625 }, { "epoch": 18.65, "learning_rate": 2.3904222402302725e-07, "loss": 0.0029, "step": 60626 }, { "epoch": 18.65, "learning_rate": 2.389339417180858e-07, "loss": 0.0041, "step": 60627 }, { "epoch": 18.65, "learning_rate": 2.3882568364724137e-07, "loss": 0.0037, "step": 60628 }, { "epoch": 18.65, "learning_rate": 2.387174498107614e-07, "loss": 0.0029, "step": 60629 }, { "epoch": 18.66, "learning_rate": 2.386092402089135e-07, "loss": 0.0035, "step": 60630 }, { "epoch": 18.66, "learning_rate": 2.385010548419675e-07, "loss": 0.0036, "step": 60631 }, { "epoch": 18.66, "learning_rate": 2.3839289371019204e-07, "loss": 0.003, "step": 60632 }, { "epoch": 18.66, "learning_rate": 2.3828475681385576e-07, "loss": 0.0035, "step": 60633 }, { "epoch": 18.66, "learning_rate": 2.3817664415322628e-07, "loss": 0.0034, "step": 60634 }, { "epoch": 18.66, "learning_rate": 2.3806855572857223e-07, "loss": 0.0046, "step": 60635 }, { "epoch": 18.66, "learning_rate": 2.3796049154016342e-07, "loss": 0.0032, "step": 60636 }, { "epoch": 18.66, "learning_rate": 2.378524515882641e-07, "loss": 0.0044, "step": 60637 }, { "epoch": 18.66, "learning_rate": 2.3774443587314843e-07, "loss": 0.0044, "step": 60638 }, { "epoch": 18.66, "learning_rate": 2.3763644439508072e-07, "loss": 0.0031, "step": 60639 }, { "epoch": 18.66, "learning_rate": 2.3752847715432958e-07, "loss": 0.0033, "step": 60640 }, { "epoch": 18.66, "learning_rate": 2.3742053415116372e-07, "loss": 0.0026, "step": 60641 }, { "epoch": 18.66, "learning_rate": 2.373126153858496e-07, "loss": 0.0043, "step": 60642 }, { "epoch": 18.66, "learning_rate": 2.3720472085865697e-07, "loss": 0.0027, "step": 60643 }, { "epoch": 18.66, "learning_rate": 2.3709685056985233e-07, "loss": 0.0031, "step": 60644 }, { "epoch": 18.66, "learning_rate": 2.3698900451970542e-07, "loss": 0.0036, "step": 60645 }, { "epoch": 18.66, "learning_rate": 2.368811827084838e-07, "loss": 0.004, "step": 60646 }, { "epoch": 18.66, "learning_rate": 2.3677338513645176e-07, "loss": 0.0036, "step": 60647 }, { "epoch": 18.66, "learning_rate": 2.3666561180388014e-07, "loss": 0.0025, "step": 60648 }, { "epoch": 18.66, "learning_rate": 2.3655786271103542e-07, "loss": 0.0026, "step": 60649 }, { "epoch": 18.66, "learning_rate": 2.3645013785818628e-07, "loss": 0.0024, "step": 60650 }, { "epoch": 18.66, "learning_rate": 2.3634243724559915e-07, "loss": 0.0057, "step": 60651 }, { "epoch": 18.66, "learning_rate": 2.3623476087354047e-07, "loss": 0.0025, "step": 60652 }, { "epoch": 18.66, "learning_rate": 2.3612710874228007e-07, "loss": 0.0041, "step": 60653 }, { "epoch": 18.66, "learning_rate": 2.3601948085208326e-07, "loss": 0.0041, "step": 60654 }, { "epoch": 18.66, "learning_rate": 2.359118772032176e-07, "loss": 0.0028, "step": 60655 }, { "epoch": 18.66, "learning_rate": 2.358042977959507e-07, "loss": 0.0029, "step": 60656 }, { "epoch": 18.66, "learning_rate": 2.3569674263054788e-07, "loss": 0.0025, "step": 60657 }, { "epoch": 18.66, "learning_rate": 2.355892117072789e-07, "loss": 0.0046, "step": 60658 }, { "epoch": 18.66, "learning_rate": 2.354817050264102e-07, "loss": 0.0037, "step": 60659 }, { "epoch": 18.66, "learning_rate": 2.353742225882083e-07, "loss": 0.0028, "step": 60660 }, { "epoch": 18.66, "learning_rate": 2.352667643929385e-07, "loss": 0.0026, "step": 60661 }, { "epoch": 18.66, "learning_rate": 2.3515933044086837e-07, "loss": 0.0023, "step": 60662 }, { "epoch": 18.67, "learning_rate": 2.3505192073226547e-07, "loss": 0.003, "step": 60663 }, { "epoch": 18.67, "learning_rate": 2.3494453526739626e-07, "loss": 0.0074, "step": 60664 }, { "epoch": 18.67, "learning_rate": 2.3483717404652828e-07, "loss": 0.0024, "step": 60665 }, { "epoch": 18.67, "learning_rate": 2.347298370699258e-07, "loss": 0.0038, "step": 60666 }, { "epoch": 18.67, "learning_rate": 2.3462252433785637e-07, "loss": 0.0042, "step": 60667 }, { "epoch": 18.67, "learning_rate": 2.3451523585058756e-07, "loss": 0.0041, "step": 60668 }, { "epoch": 18.67, "learning_rate": 2.3440797160838358e-07, "loss": 0.0029, "step": 60669 }, { "epoch": 18.67, "learning_rate": 2.343007316115131e-07, "loss": 0.0027, "step": 60670 }, { "epoch": 18.67, "learning_rate": 2.3419351586024041e-07, "loss": 0.0043, "step": 60671 }, { "epoch": 18.67, "learning_rate": 2.3408632435483304e-07, "loss": 0.0034, "step": 60672 }, { "epoch": 18.67, "learning_rate": 2.3397915709555518e-07, "loss": 0.0039, "step": 60673 }, { "epoch": 18.67, "learning_rate": 2.3387201408267557e-07, "loss": 0.0032, "step": 60674 }, { "epoch": 18.67, "learning_rate": 2.337648953164584e-07, "loss": 0.0032, "step": 60675 }, { "epoch": 18.67, "learning_rate": 2.3365780079717016e-07, "loss": 0.0028, "step": 60676 }, { "epoch": 18.67, "learning_rate": 2.3355073052507615e-07, "loss": 0.0032, "step": 60677 }, { "epoch": 18.67, "learning_rate": 2.3344368450044287e-07, "loss": 0.0037, "step": 60678 }, { "epoch": 18.67, "learning_rate": 2.3333666272353672e-07, "loss": 0.0051, "step": 60679 }, { "epoch": 18.67, "learning_rate": 2.3322966519462308e-07, "loss": 0.0022, "step": 60680 }, { "epoch": 18.67, "learning_rate": 2.3312269191396619e-07, "loss": 0.0038, "step": 60681 }, { "epoch": 18.67, "learning_rate": 2.3301574288183138e-07, "loss": 0.0049, "step": 60682 }, { "epoch": 18.67, "learning_rate": 2.329088180984862e-07, "loss": 0.0043, "step": 60683 }, { "epoch": 18.67, "learning_rate": 2.32801917564196e-07, "loss": 0.0023, "step": 60684 }, { "epoch": 18.67, "learning_rate": 2.3269504127922504e-07, "loss": 0.004, "step": 60685 }, { "epoch": 18.67, "learning_rate": 2.3258818924383976e-07, "loss": 0.0042, "step": 60686 }, { "epoch": 18.67, "learning_rate": 2.3248136145830434e-07, "loss": 0.0056, "step": 60687 }, { "epoch": 18.67, "learning_rate": 2.3237455792288421e-07, "loss": 0.0031, "step": 60688 }, { "epoch": 18.67, "learning_rate": 2.322677786378458e-07, "loss": 0.0024, "step": 60689 }, { "epoch": 18.67, "learning_rate": 2.3216102360345217e-07, "loss": 0.0043, "step": 60690 }, { "epoch": 18.67, "learning_rate": 2.3205429281996982e-07, "loss": 0.0024, "step": 60691 }, { "epoch": 18.67, "learning_rate": 2.319475862876619e-07, "loss": 0.0034, "step": 60692 }, { "epoch": 18.67, "learning_rate": 2.3184090400679703e-07, "loss": 0.0041, "step": 60693 }, { "epoch": 18.67, "learning_rate": 2.3173424597763615e-07, "loss": 0.0032, "step": 60694 }, { "epoch": 18.68, "learning_rate": 2.3162761220044684e-07, "loss": 0.0039, "step": 60695 }, { "epoch": 18.68, "learning_rate": 2.3152100267549216e-07, "loss": 0.0045, "step": 60696 }, { "epoch": 18.68, "learning_rate": 2.314144174030364e-07, "loss": 0.0058, "step": 60697 }, { "epoch": 18.68, "learning_rate": 2.3130785638334597e-07, "loss": 0.0024, "step": 60698 }, { "epoch": 18.68, "learning_rate": 2.31201319616684e-07, "loss": 0.0035, "step": 60699 }, { "epoch": 18.68, "learning_rate": 2.31094807103317e-07, "loss": 0.0037, "step": 60700 }, { "epoch": 18.68, "learning_rate": 2.3098831884350804e-07, "loss": 0.0016, "step": 60701 }, { "epoch": 18.68, "learning_rate": 2.3088185483751912e-07, "loss": 0.0021, "step": 60702 }, { "epoch": 18.68, "learning_rate": 2.3077541508561896e-07, "loss": 0.0036, "step": 60703 }, { "epoch": 18.68, "learning_rate": 2.3066899958806843e-07, "loss": 0.0032, "step": 60704 }, { "epoch": 18.68, "learning_rate": 2.30562608345134e-07, "loss": 0.0042, "step": 60705 }, { "epoch": 18.68, "learning_rate": 2.304562413570788e-07, "loss": 0.0035, "step": 60706 }, { "epoch": 18.68, "learning_rate": 2.30349898624167e-07, "loss": 0.0056, "step": 60707 }, { "epoch": 18.68, "learning_rate": 2.3024358014666181e-07, "loss": 0.0026, "step": 60708 }, { "epoch": 18.68, "learning_rate": 2.3013728592482852e-07, "loss": 0.0039, "step": 60709 }, { "epoch": 18.68, "learning_rate": 2.300310159589303e-07, "loss": 0.0045, "step": 60710 }, { "epoch": 18.68, "learning_rate": 2.2992477024923132e-07, "loss": 0.0013, "step": 60711 }, { "epoch": 18.68, "learning_rate": 2.2981854879599474e-07, "loss": 0.0046, "step": 60712 }, { "epoch": 18.68, "learning_rate": 2.297123515994848e-07, "loss": 0.0052, "step": 60713 }, { "epoch": 18.68, "learning_rate": 2.2960617865996681e-07, "loss": 0.0038, "step": 60714 }, { "epoch": 18.68, "learning_rate": 2.295000299777006e-07, "loss": 0.0049, "step": 60715 }, { "epoch": 18.68, "learning_rate": 2.293939055529526e-07, "loss": 0.0031, "step": 60716 }, { "epoch": 18.68, "learning_rate": 2.2928780538598483e-07, "loss": 0.0033, "step": 60717 }, { "epoch": 18.68, "learning_rate": 2.2918172947706152e-07, "loss": 0.0022, "step": 60718 }, { "epoch": 18.68, "learning_rate": 2.290756778264469e-07, "loss": 0.0025, "step": 60719 }, { "epoch": 18.68, "learning_rate": 2.289696504344019e-07, "loss": 0.0025, "step": 60720 }, { "epoch": 18.68, "learning_rate": 2.288636473011918e-07, "loss": 0.0046, "step": 60721 }, { "epoch": 18.68, "learning_rate": 2.287576684270787e-07, "loss": 0.0023, "step": 60722 }, { "epoch": 18.68, "learning_rate": 2.2865171381232677e-07, "loss": 0.0027, "step": 60723 }, { "epoch": 18.68, "learning_rate": 2.2854578345719802e-07, "loss": 0.0039, "step": 60724 }, { "epoch": 18.68, "learning_rate": 2.284398773619556e-07, "loss": 0.0047, "step": 60725 }, { "epoch": 18.68, "learning_rate": 2.2833399552686264e-07, "loss": 0.0041, "step": 60726 }, { "epoch": 18.68, "learning_rate": 2.2822813795218112e-07, "loss": 0.0039, "step": 60727 }, { "epoch": 18.69, "learning_rate": 2.281223046381764e-07, "loss": 0.004, "step": 60728 }, { "epoch": 18.69, "learning_rate": 2.280164955851083e-07, "loss": 0.0041, "step": 60729 }, { "epoch": 18.69, "learning_rate": 2.279107107932421e-07, "loss": 0.0034, "step": 60730 }, { "epoch": 18.69, "learning_rate": 2.2780495026283766e-07, "loss": 0.0028, "step": 60731 }, { "epoch": 18.69, "learning_rate": 2.2769921399415917e-07, "loss": 0.003, "step": 60732 }, { "epoch": 18.69, "learning_rate": 2.2759350198746978e-07, "loss": 0.0035, "step": 60733 }, { "epoch": 18.69, "learning_rate": 2.2748781424303145e-07, "loss": 0.0028, "step": 60734 }, { "epoch": 18.69, "learning_rate": 2.2738215076110625e-07, "loss": 0.0035, "step": 60735 }, { "epoch": 18.69, "learning_rate": 2.2727651154195617e-07, "loss": 0.0038, "step": 60736 }, { "epoch": 18.69, "learning_rate": 2.2717089658584323e-07, "loss": 0.0027, "step": 60737 }, { "epoch": 18.69, "learning_rate": 2.2706530589303057e-07, "loss": 0.0038, "step": 60738 }, { "epoch": 18.69, "learning_rate": 2.2695973946378013e-07, "loss": 0.0033, "step": 60739 }, { "epoch": 18.69, "learning_rate": 2.2685419729835402e-07, "loss": 0.003, "step": 60740 }, { "epoch": 18.69, "learning_rate": 2.2674867939701418e-07, "loss": 0.0048, "step": 60741 }, { "epoch": 18.69, "learning_rate": 2.2664318576002264e-07, "loss": 0.0028, "step": 60742 }, { "epoch": 18.69, "learning_rate": 2.2653771638764144e-07, "loss": 0.0052, "step": 60743 }, { "epoch": 18.69, "learning_rate": 2.2643227128013257e-07, "loss": 0.0051, "step": 60744 }, { "epoch": 18.69, "learning_rate": 2.2632685043775692e-07, "loss": 0.0045, "step": 60745 }, { "epoch": 18.69, "learning_rate": 2.2622145386077653e-07, "loss": 0.0024, "step": 60746 }, { "epoch": 18.69, "learning_rate": 2.2611608154945343e-07, "loss": 0.0033, "step": 60747 }, { "epoch": 18.69, "learning_rate": 2.2601073350404957e-07, "loss": 0.0088, "step": 60748 }, { "epoch": 18.69, "learning_rate": 2.25905409724827e-07, "loss": 0.0034, "step": 60749 }, { "epoch": 18.69, "learning_rate": 2.258001102120444e-07, "loss": 0.0021, "step": 60750 }, { "epoch": 18.69, "learning_rate": 2.25694834965966e-07, "loss": 0.0026, "step": 60751 }, { "epoch": 18.69, "learning_rate": 2.255895839868516e-07, "loss": 0.0035, "step": 60752 }, { "epoch": 18.69, "learning_rate": 2.254843572749632e-07, "loss": 0.0051, "step": 60753 }, { "epoch": 18.69, "learning_rate": 2.2537915483056284e-07, "loss": 0.0028, "step": 60754 }, { "epoch": 18.69, "learning_rate": 2.2527397665391026e-07, "loss": 0.0034, "step": 60755 }, { "epoch": 18.69, "learning_rate": 2.2516882274526862e-07, "loss": 0.0031, "step": 60756 }, { "epoch": 18.69, "learning_rate": 2.250636931048955e-07, "loss": 0.0017, "step": 60757 }, { "epoch": 18.69, "learning_rate": 2.2495858773305512e-07, "loss": 0.0023, "step": 60758 }, { "epoch": 18.69, "learning_rate": 2.2485350663000727e-07, "loss": 0.0033, "step": 60759 }, { "epoch": 18.7, "learning_rate": 2.2474844979601285e-07, "loss": 0.0018, "step": 60760 }, { "epoch": 18.7, "learning_rate": 2.2464341723133276e-07, "loss": 0.0038, "step": 60761 }, { "epoch": 18.7, "learning_rate": 2.2453840893622792e-07, "loss": 0.0036, "step": 60762 }, { "epoch": 18.7, "learning_rate": 2.2443342491095809e-07, "loss": 0.0029, "step": 60763 }, { "epoch": 18.7, "learning_rate": 2.2432846515578533e-07, "loss": 0.0028, "step": 60764 }, { "epoch": 18.7, "learning_rate": 2.242235296709694e-07, "loss": 0.0033, "step": 60765 }, { "epoch": 18.7, "learning_rate": 2.2411861845677007e-07, "loss": 0.0036, "step": 60766 }, { "epoch": 18.7, "learning_rate": 2.240137315134483e-07, "loss": 0.0031, "step": 60767 }, { "epoch": 18.7, "learning_rate": 2.2390886884126716e-07, "loss": 0.0043, "step": 60768 }, { "epoch": 18.7, "learning_rate": 2.2380403044048428e-07, "loss": 0.0044, "step": 60769 }, { "epoch": 18.7, "learning_rate": 2.2369921631135937e-07, "loss": 0.0028, "step": 60770 }, { "epoch": 18.7, "learning_rate": 2.2359442645415452e-07, "loss": 0.0037, "step": 60771 }, { "epoch": 18.7, "learning_rate": 2.2348966086912725e-07, "loss": 0.0032, "step": 60772 }, { "epoch": 18.7, "learning_rate": 2.2338491955654184e-07, "loss": 0.0046, "step": 60773 }, { "epoch": 18.7, "learning_rate": 2.2328020251665472e-07, "loss": 0.0025, "step": 60774 }, { "epoch": 18.7, "learning_rate": 2.2317550974972679e-07, "loss": 0.0044, "step": 60775 }, { "epoch": 18.7, "learning_rate": 2.2307084125601897e-07, "loss": 0.0032, "step": 60776 }, { "epoch": 18.7, "learning_rate": 2.2296619703579102e-07, "loss": 0.0059, "step": 60777 }, { "epoch": 18.7, "learning_rate": 2.2286157708930168e-07, "loss": 0.003, "step": 60778 }, { "epoch": 18.7, "learning_rate": 2.2275698141681178e-07, "loss": 0.0027, "step": 60779 }, { "epoch": 18.7, "learning_rate": 2.2265241001857896e-07, "loss": 0.0021, "step": 60780 }, { "epoch": 18.7, "learning_rate": 2.2254786289486518e-07, "loss": 0.004, "step": 60781 }, { "epoch": 18.7, "learning_rate": 2.2244334004592916e-07, "loss": 0.0043, "step": 60782 }, { "epoch": 18.7, "learning_rate": 2.2233884147203065e-07, "loss": 0.0034, "step": 60783 }, { "epoch": 18.7, "learning_rate": 2.222343671734284e-07, "loss": 0.0028, "step": 60784 }, { "epoch": 18.7, "learning_rate": 2.2212991715038324e-07, "loss": 0.0028, "step": 60785 }, { "epoch": 18.7, "learning_rate": 2.2202549140315167e-07, "loss": 0.0022, "step": 60786 }, { "epoch": 18.7, "learning_rate": 2.2192108993199567e-07, "loss": 0.0027, "step": 60787 }, { "epoch": 18.7, "learning_rate": 2.2181671273717287e-07, "loss": 0.0037, "step": 60788 }, { "epoch": 18.7, "learning_rate": 2.2171235981894412e-07, "loss": 0.0043, "step": 60789 }, { "epoch": 18.7, "learning_rate": 2.2160803117756814e-07, "loss": 0.0035, "step": 60790 }, { "epoch": 18.7, "learning_rate": 2.2150372681330135e-07, "loss": 0.0021, "step": 60791 }, { "epoch": 18.7, "learning_rate": 2.2139944672640467e-07, "loss": 0.0039, "step": 60792 }, { "epoch": 18.71, "learning_rate": 2.2129519091713792e-07, "loss": 0.0025, "step": 60793 }, { "epoch": 18.71, "learning_rate": 2.2119095938575863e-07, "loss": 0.0033, "step": 60794 }, { "epoch": 18.71, "learning_rate": 2.210867521325255e-07, "loss": 0.0032, "step": 60795 }, { "epoch": 18.71, "learning_rate": 2.2098256915769833e-07, "loss": 0.0049, "step": 60796 }, { "epoch": 18.71, "learning_rate": 2.2087841046153468e-07, "loss": 0.0039, "step": 60797 }, { "epoch": 18.71, "learning_rate": 2.2077427604429435e-07, "loss": 0.0022, "step": 60798 }, { "epoch": 18.71, "learning_rate": 2.206701659062338e-07, "loss": 0.0028, "step": 60799 }, { "epoch": 18.71, "learning_rate": 2.2056608004761394e-07, "loss": 0.0028, "step": 60800 }, { "epoch": 18.71, "learning_rate": 2.204620184686912e-07, "loss": 0.0029, "step": 60801 }, { "epoch": 18.71, "learning_rate": 2.2035798116972428e-07, "loss": 0.0023, "step": 60802 }, { "epoch": 18.71, "learning_rate": 2.20253968150973e-07, "loss": 0.0016, "step": 60803 }, { "epoch": 18.71, "learning_rate": 2.20149979412696e-07, "loss": 0.0036, "step": 60804 }, { "epoch": 18.71, "learning_rate": 2.2004601495514866e-07, "loss": 0.0043, "step": 60805 }, { "epoch": 18.71, "learning_rate": 2.1994207477858964e-07, "loss": 0.0031, "step": 60806 }, { "epoch": 18.71, "learning_rate": 2.1983815888327765e-07, "loss": 0.0029, "step": 60807 }, { "epoch": 18.71, "learning_rate": 2.1973426726947246e-07, "loss": 0.0034, "step": 60808 }, { "epoch": 18.71, "learning_rate": 2.196303999374294e-07, "loss": 0.0034, "step": 60809 }, { "epoch": 18.71, "learning_rate": 2.195265568874072e-07, "loss": 0.0045, "step": 60810 }, { "epoch": 18.71, "learning_rate": 2.1942273811966563e-07, "loss": 0.0042, "step": 60811 }, { "epoch": 18.71, "learning_rate": 2.1931894363445784e-07, "loss": 0.0041, "step": 60812 }, { "epoch": 18.71, "learning_rate": 2.192151734320458e-07, "loss": 0.004, "step": 60813 }, { "epoch": 18.71, "learning_rate": 2.1911142751268598e-07, "loss": 0.0052, "step": 60814 }, { "epoch": 18.71, "learning_rate": 2.1900770587663488e-07, "loss": 0.0022, "step": 60815 }, { "epoch": 18.71, "learning_rate": 2.1890400852415118e-07, "loss": 0.0037, "step": 60816 }, { "epoch": 18.71, "learning_rate": 2.188003354554924e-07, "loss": 0.0042, "step": 60817 }, { "epoch": 18.71, "learning_rate": 2.1869668667091504e-07, "loss": 0.0051, "step": 60818 }, { "epoch": 18.71, "learning_rate": 2.1859306217067665e-07, "loss": 0.0027, "step": 60819 }, { "epoch": 18.71, "learning_rate": 2.1848946195503483e-07, "loss": 0.0042, "step": 60820 }, { "epoch": 18.71, "learning_rate": 2.1838588602424604e-07, "loss": 0.01, "step": 60821 }, { "epoch": 18.71, "learning_rate": 2.1828233437856784e-07, "loss": 0.0032, "step": 60822 }, { "epoch": 18.71, "learning_rate": 2.1817880701825888e-07, "loss": 0.0033, "step": 60823 }, { "epoch": 18.71, "learning_rate": 2.1807530394357457e-07, "loss": 0.0036, "step": 60824 }, { "epoch": 18.72, "learning_rate": 2.1797182515477244e-07, "loss": 0.005, "step": 60825 }, { "epoch": 18.72, "learning_rate": 2.1786837065210898e-07, "loss": 0.0023, "step": 60826 }, { "epoch": 18.72, "learning_rate": 2.1776494043583952e-07, "loss": 0.003, "step": 60827 }, { "epoch": 18.72, "learning_rate": 2.1766153450622385e-07, "loss": 0.0048, "step": 60828 }, { "epoch": 18.72, "learning_rate": 2.1755815286351734e-07, "loss": 0.0042, "step": 60829 }, { "epoch": 18.72, "learning_rate": 2.1745479550797644e-07, "loss": 0.0027, "step": 60830 }, { "epoch": 18.72, "learning_rate": 2.173514624398576e-07, "loss": 0.0022, "step": 60831 }, { "epoch": 18.72, "learning_rate": 2.1724815365941842e-07, "loss": 0.0028, "step": 60832 }, { "epoch": 18.72, "learning_rate": 2.1714486916691534e-07, "loss": 0.0038, "step": 60833 }, { "epoch": 18.72, "learning_rate": 2.1704160896260263e-07, "loss": 0.0039, "step": 60834 }, { "epoch": 18.72, "learning_rate": 2.1693837304673892e-07, "loss": 0.0039, "step": 60835 }, { "epoch": 18.72, "learning_rate": 2.1683516141958073e-07, "loss": 0.0027, "step": 60836 }, { "epoch": 18.72, "learning_rate": 2.1673197408138115e-07, "loss": 0.0024, "step": 60837 }, { "epoch": 18.72, "learning_rate": 2.1662881103240108e-07, "loss": 0.0025, "step": 60838 }, { "epoch": 18.72, "learning_rate": 2.1652567227289257e-07, "loss": 0.0032, "step": 60839 }, { "epoch": 18.72, "learning_rate": 2.1642255780311428e-07, "loss": 0.0021, "step": 60840 }, { "epoch": 18.72, "learning_rate": 2.1631946762332045e-07, "loss": 0.0043, "step": 60841 }, { "epoch": 18.72, "learning_rate": 2.1621640173376758e-07, "loss": 0.0059, "step": 60842 }, { "epoch": 18.72, "learning_rate": 2.1611336013471207e-07, "loss": 0.0033, "step": 60843 }, { "epoch": 18.72, "learning_rate": 2.1601034282640932e-07, "loss": 0.0051, "step": 60844 }, { "epoch": 18.72, "learning_rate": 2.1590734980911575e-07, "loss": 0.0028, "step": 60845 }, { "epoch": 18.72, "learning_rate": 2.1580438108308676e-07, "loss": 0.0023, "step": 60846 }, { "epoch": 18.72, "learning_rate": 2.1570143664857657e-07, "loss": 0.0038, "step": 60847 }, { "epoch": 18.72, "learning_rate": 2.1559851650584274e-07, "loss": 0.0038, "step": 60848 }, { "epoch": 18.72, "learning_rate": 2.1549562065513952e-07, "loss": 0.0044, "step": 60849 }, { "epoch": 18.72, "learning_rate": 2.1539274909672337e-07, "loss": 0.0028, "step": 60850 }, { "epoch": 18.72, "learning_rate": 2.1528990183084853e-07, "loss": 0.0049, "step": 60851 }, { "epoch": 18.72, "learning_rate": 2.1518707885777147e-07, "loss": 0.0024, "step": 60852 }, { "epoch": 18.72, "learning_rate": 2.1508428017774751e-07, "loss": 0.0035, "step": 60853 }, { "epoch": 18.72, "learning_rate": 2.1498150579103093e-07, "loss": 0.0028, "step": 60854 }, { "epoch": 18.72, "learning_rate": 2.1487875569787708e-07, "loss": 0.0023, "step": 60855 }, { "epoch": 18.72, "learning_rate": 2.147760298985413e-07, "loss": 0.0034, "step": 60856 }, { "epoch": 18.72, "learning_rate": 2.1467332839327783e-07, "loss": 0.0034, "step": 60857 }, { "epoch": 18.73, "learning_rate": 2.145706511823431e-07, "loss": 0.0044, "step": 60858 }, { "epoch": 18.73, "learning_rate": 2.1446799826599252e-07, "loss": 0.0025, "step": 60859 }, { "epoch": 18.73, "learning_rate": 2.143653696444792e-07, "loss": 0.0039, "step": 60860 }, { "epoch": 18.73, "learning_rate": 2.1426276531805845e-07, "loss": 0.0045, "step": 60861 }, { "epoch": 18.73, "learning_rate": 2.1416018528698346e-07, "loss": 0.0041, "step": 60862 }, { "epoch": 18.73, "learning_rate": 2.1405762955151178e-07, "loss": 0.0033, "step": 60863 }, { "epoch": 18.73, "learning_rate": 2.1395509811189763e-07, "loss": 0.0017, "step": 60864 }, { "epoch": 18.73, "learning_rate": 2.1385259096839416e-07, "loss": 0.0038, "step": 60865 }, { "epoch": 18.73, "learning_rate": 2.1375010812125564e-07, "loss": 0.0044, "step": 60866 }, { "epoch": 18.73, "learning_rate": 2.1364764957073848e-07, "loss": 0.0035, "step": 60867 }, { "epoch": 18.73, "learning_rate": 2.1354521531709582e-07, "loss": 0.0036, "step": 60868 }, { "epoch": 18.73, "learning_rate": 2.134428053605808e-07, "loss": 0.0015, "step": 60869 }, { "epoch": 18.73, "learning_rate": 2.1334041970144991e-07, "loss": 0.0033, "step": 60870 }, { "epoch": 18.73, "learning_rate": 2.1323805833995627e-07, "loss": 0.008, "step": 60871 }, { "epoch": 18.73, "learning_rate": 2.1313572127635406e-07, "loss": 0.0034, "step": 60872 }, { "epoch": 18.73, "learning_rate": 2.130334085108976e-07, "loss": 0.0049, "step": 60873 }, { "epoch": 18.73, "learning_rate": 2.1293112004384108e-07, "loss": 0.0032, "step": 60874 }, { "epoch": 18.73, "learning_rate": 2.1282885587543657e-07, "loss": 0.0025, "step": 60875 }, { "epoch": 18.73, "learning_rate": 2.1272661600594046e-07, "loss": 0.0037, "step": 60876 }, { "epoch": 18.73, "learning_rate": 2.1262440043560483e-07, "loss": 0.0034, "step": 60877 }, { "epoch": 18.73, "learning_rate": 2.1252220916468504e-07, "loss": 0.0038, "step": 60878 }, { "epoch": 18.73, "learning_rate": 2.1242004219343415e-07, "loss": 0.0038, "step": 60879 }, { "epoch": 18.73, "learning_rate": 2.1231789952210646e-07, "loss": 0.0021, "step": 60880 }, { "epoch": 18.73, "learning_rate": 2.1221578115095286e-07, "loss": 0.003, "step": 60881 }, { "epoch": 18.73, "learning_rate": 2.1211368708022873e-07, "loss": 0.003, "step": 60882 }, { "epoch": 18.73, "learning_rate": 2.1201161731018828e-07, "loss": 0.004, "step": 60883 }, { "epoch": 18.73, "learning_rate": 2.1190957184108463e-07, "loss": 0.0024, "step": 60884 }, { "epoch": 18.73, "learning_rate": 2.1180755067317093e-07, "loss": 0.0032, "step": 60885 }, { "epoch": 18.73, "learning_rate": 2.1170555380669922e-07, "loss": 0.0054, "step": 60886 }, { "epoch": 18.73, "learning_rate": 2.1160358124192482e-07, "loss": 0.0035, "step": 60887 }, { "epoch": 18.73, "learning_rate": 2.1150163297909976e-07, "loss": 0.0043, "step": 60888 }, { "epoch": 18.73, "learning_rate": 2.1139970901847607e-07, "loss": 0.006, "step": 60889 }, { "epoch": 18.74, "learning_rate": 2.112978093603091e-07, "loss": 0.0031, "step": 60890 }, { "epoch": 18.74, "learning_rate": 2.1119593400484972e-07, "loss": 0.0024, "step": 60891 }, { "epoch": 18.74, "learning_rate": 2.1109408295235223e-07, "loss": 0.0029, "step": 60892 }, { "epoch": 18.74, "learning_rate": 2.1099225620307085e-07, "loss": 0.0033, "step": 60893 }, { "epoch": 18.74, "learning_rate": 2.1089045375725538e-07, "loss": 0.0055, "step": 60894 }, { "epoch": 18.74, "learning_rate": 2.1078867561515893e-07, "loss": 0.0034, "step": 60895 }, { "epoch": 18.74, "learning_rate": 2.1068692177703575e-07, "loss": 0.0032, "step": 60896 }, { "epoch": 18.74, "learning_rate": 2.1058519224313788e-07, "loss": 0.0036, "step": 60897 }, { "epoch": 18.74, "learning_rate": 2.1048348701371734e-07, "loss": 0.0043, "step": 60898 }, { "epoch": 18.74, "learning_rate": 2.1038180608902837e-07, "loss": 0.0028, "step": 60899 }, { "epoch": 18.74, "learning_rate": 2.1028014946932075e-07, "loss": 0.0012, "step": 60900 }, { "epoch": 18.74, "learning_rate": 2.1017851715484984e-07, "loss": 0.0022, "step": 60901 }, { "epoch": 18.74, "learning_rate": 2.1007690914586542e-07, "loss": 0.0031, "step": 60902 }, { "epoch": 18.74, "learning_rate": 2.0997532544262068e-07, "loss": 0.0042, "step": 60903 }, { "epoch": 18.74, "learning_rate": 2.098737660453687e-07, "loss": 0.003, "step": 60904 }, { "epoch": 18.74, "learning_rate": 2.0977223095435927e-07, "loss": 0.0044, "step": 60905 }, { "epoch": 18.74, "learning_rate": 2.096707201698478e-07, "loss": 0.004, "step": 60906 }, { "epoch": 18.74, "learning_rate": 2.0956923369208403e-07, "loss": 0.0019, "step": 60907 }, { "epoch": 18.74, "learning_rate": 2.0946777152132003e-07, "loss": 0.004, "step": 60908 }, { "epoch": 18.74, "learning_rate": 2.0936633365780779e-07, "loss": 0.0042, "step": 60909 }, { "epoch": 18.74, "learning_rate": 2.0926492010180043e-07, "loss": 0.004, "step": 60910 }, { "epoch": 18.74, "learning_rate": 2.091635308535478e-07, "loss": 0.0046, "step": 60911 }, { "epoch": 18.74, "learning_rate": 2.0906216591330296e-07, "loss": 0.0017, "step": 60912 }, { "epoch": 18.74, "learning_rate": 2.089608252813169e-07, "loss": 0.0066, "step": 60913 }, { "epoch": 18.74, "learning_rate": 2.088595089578427e-07, "loss": 0.0038, "step": 60914 }, { "epoch": 18.74, "learning_rate": 2.0875821694313014e-07, "loss": 0.003, "step": 60915 }, { "epoch": 18.74, "learning_rate": 2.0865694923743018e-07, "loss": 0.0024, "step": 60916 }, { "epoch": 18.74, "learning_rate": 2.0855570584099704e-07, "loss": 0.0029, "step": 60917 }, { "epoch": 18.74, "learning_rate": 2.0845448675407943e-07, "loss": 0.0033, "step": 60918 }, { "epoch": 18.74, "learning_rate": 2.0835329197692932e-07, "loss": 0.002, "step": 60919 }, { "epoch": 18.74, "learning_rate": 2.082521215097988e-07, "loss": 0.0034, "step": 60920 }, { "epoch": 18.74, "learning_rate": 2.081509753529387e-07, "loss": 0.0038, "step": 60921 }, { "epoch": 18.74, "learning_rate": 2.080498535066e-07, "loss": 0.0028, "step": 60922 }, { "epoch": 18.75, "learning_rate": 2.0794875597103358e-07, "loss": 0.0048, "step": 60923 }, { "epoch": 18.75, "learning_rate": 2.0784768274649036e-07, "loss": 0.003, "step": 60924 }, { "epoch": 18.75, "learning_rate": 2.0774663383322123e-07, "loss": 0.0031, "step": 60925 }, { "epoch": 18.75, "learning_rate": 2.0764560923147604e-07, "loss": 0.0017, "step": 60926 }, { "epoch": 18.75, "learning_rate": 2.0754460894151008e-07, "loss": 0.0038, "step": 60927 }, { "epoch": 18.75, "learning_rate": 2.0744363296356872e-07, "loss": 0.0026, "step": 60928 }, { "epoch": 18.75, "learning_rate": 2.0734268129790513e-07, "loss": 0.0056, "step": 60929 }, { "epoch": 18.75, "learning_rate": 2.072417539447702e-07, "loss": 0.0039, "step": 60930 }, { "epoch": 18.75, "learning_rate": 2.071408509044126e-07, "loss": 0.0035, "step": 60931 }, { "epoch": 18.75, "learning_rate": 2.0703997217708437e-07, "loss": 0.0024, "step": 60932 }, { "epoch": 18.75, "learning_rate": 2.0693911776303753e-07, "loss": 0.0037, "step": 60933 }, { "epoch": 18.75, "learning_rate": 2.0683828766251968e-07, "loss": 0.0055, "step": 60934 }, { "epoch": 18.75, "learning_rate": 2.0673748187578279e-07, "loss": 0.0029, "step": 60935 }, { "epoch": 18.75, "learning_rate": 2.066367004030756e-07, "loss": 0.0045, "step": 60936 }, { "epoch": 18.75, "learning_rate": 2.0653594324464897e-07, "loss": 0.0035, "step": 60937 }, { "epoch": 18.75, "learning_rate": 2.0643521040075387e-07, "loss": 0.0051, "step": 60938 }, { "epoch": 18.75, "learning_rate": 2.0633450187164007e-07, "loss": 0.0047, "step": 60939 }, { "epoch": 18.75, "learning_rate": 2.0623381765755733e-07, "loss": 0.0037, "step": 60940 }, { "epoch": 18.75, "learning_rate": 2.0613315775875665e-07, "loss": 0.0048, "step": 60941 }, { "epoch": 18.75, "learning_rate": 2.0603252217548552e-07, "loss": 0.0029, "step": 60942 }, { "epoch": 18.75, "learning_rate": 2.05931910907996e-07, "loss": 0.0033, "step": 60943 }, { "epoch": 18.75, "learning_rate": 2.0583132395653683e-07, "loss": 0.0047, "step": 60944 }, { "epoch": 18.75, "learning_rate": 2.0573076132135883e-07, "loss": 0.0023, "step": 60945 }, { "epoch": 18.75, "learning_rate": 2.0563022300270962e-07, "loss": 0.0036, "step": 60946 }, { "epoch": 18.75, "learning_rate": 2.0552970900084012e-07, "loss": 0.0037, "step": 60947 }, { "epoch": 18.75, "learning_rate": 2.0542921931600235e-07, "loss": 0.0035, "step": 60948 }, { "epoch": 18.75, "learning_rate": 2.0532875394844053e-07, "loss": 0.0032, "step": 60949 }, { "epoch": 18.75, "learning_rate": 2.0522831289840783e-07, "loss": 0.0051, "step": 60950 }, { "epoch": 18.75, "learning_rate": 2.0512789616615181e-07, "loss": 0.0055, "step": 60951 }, { "epoch": 18.75, "learning_rate": 2.0502750375192337e-07, "loss": 0.0042, "step": 60952 }, { "epoch": 18.75, "learning_rate": 2.0492713565597123e-07, "loss": 0.0035, "step": 60953 }, { "epoch": 18.75, "learning_rate": 2.0482679187854405e-07, "loss": 0.0028, "step": 60954 }, { "epoch": 18.76, "learning_rate": 2.0472647241989164e-07, "loss": 0.0028, "step": 60955 }, { "epoch": 18.76, "learning_rate": 2.046261772802627e-07, "loss": 0.0041, "step": 60956 }, { "epoch": 18.76, "learning_rate": 2.045259064599059e-07, "loss": 0.0035, "step": 60957 }, { "epoch": 18.76, "learning_rate": 2.0442565995906993e-07, "loss": 0.0039, "step": 60958 }, { "epoch": 18.76, "learning_rate": 2.0432543777800463e-07, "loss": 0.0043, "step": 60959 }, { "epoch": 18.76, "learning_rate": 2.0422523991695865e-07, "loss": 0.0035, "step": 60960 }, { "epoch": 18.76, "learning_rate": 2.0412506637617957e-07, "loss": 0.0046, "step": 60961 }, { "epoch": 18.76, "learning_rate": 2.0402491715591943e-07, "loss": 0.0037, "step": 60962 }, { "epoch": 18.76, "learning_rate": 2.0392479225642248e-07, "loss": 0.0037, "step": 60963 }, { "epoch": 18.76, "learning_rate": 2.0382469167794072e-07, "loss": 0.0034, "step": 60964 }, { "epoch": 18.76, "learning_rate": 2.0372461542071953e-07, "loss": 0.007, "step": 60965 }, { "epoch": 18.76, "learning_rate": 2.036245634850098e-07, "loss": 0.0031, "step": 60966 }, { "epoch": 18.76, "learning_rate": 2.0352453587105914e-07, "loss": 0.002, "step": 60967 }, { "epoch": 18.76, "learning_rate": 2.034245325791162e-07, "loss": 0.0068, "step": 60968 }, { "epoch": 18.76, "learning_rate": 2.0332455360943083e-07, "loss": 0.0037, "step": 60969 }, { "epoch": 18.76, "learning_rate": 2.032245989622472e-07, "loss": 0.0032, "step": 60970 }, { "epoch": 18.76, "learning_rate": 2.031246686378163e-07, "loss": 0.0035, "step": 60971 }, { "epoch": 18.76, "learning_rate": 2.031246686378163e-07, "loss": 0.003, "step": 60972 }, { "epoch": 18.76, "learning_rate": 2.0302476263638572e-07, "loss": 0.0025, "step": 60973 }, { "epoch": 18.76, "learning_rate": 2.0292488095820405e-07, "loss": 0.0032, "step": 60974 }, { "epoch": 18.76, "learning_rate": 2.0282502360351785e-07, "loss": 0.0045, "step": 60975 }, { "epoch": 18.76, "learning_rate": 2.0272519057257578e-07, "loss": 0.0027, "step": 60976 }, { "epoch": 18.76, "learning_rate": 2.0262538186562652e-07, "loss": 0.0023, "step": 60977 }, { "epoch": 18.76, "learning_rate": 2.0252559748291767e-07, "loss": 0.005, "step": 60978 }, { "epoch": 18.76, "learning_rate": 2.0242583742469568e-07, "loss": 0.0037, "step": 60979 }, { "epoch": 18.76, "learning_rate": 2.0232610169120814e-07, "loss": 0.004, "step": 60980 }, { "epoch": 18.76, "learning_rate": 2.0222639028270486e-07, "loss": 0.0033, "step": 60981 }, { "epoch": 18.76, "learning_rate": 2.0212670319943118e-07, "loss": 0.0029, "step": 60982 }, { "epoch": 18.76, "learning_rate": 2.0202704044163578e-07, "loss": 0.0051, "step": 60983 }, { "epoch": 18.76, "learning_rate": 2.019274020095663e-07, "loss": 0.0037, "step": 60984 }, { "epoch": 18.76, "learning_rate": 2.0182778790346913e-07, "loss": 0.0029, "step": 60985 }, { "epoch": 18.76, "learning_rate": 2.017281981235919e-07, "loss": 0.0048, "step": 60986 }, { "epoch": 18.76, "learning_rate": 2.0162863267018218e-07, "loss": 0.0026, "step": 60987 }, { "epoch": 18.77, "learning_rate": 2.0152909154348643e-07, "loss": 0.0045, "step": 60988 }, { "epoch": 18.77, "learning_rate": 2.0142957474375336e-07, "loss": 0.0026, "step": 60989 }, { "epoch": 18.77, "learning_rate": 2.013300822712283e-07, "loss": 0.0047, "step": 60990 }, { "epoch": 18.77, "learning_rate": 2.0123061412616107e-07, "loss": 0.0038, "step": 60991 }, { "epoch": 18.77, "learning_rate": 2.011311703087937e-07, "loss": 0.0032, "step": 60992 }, { "epoch": 18.77, "learning_rate": 2.010317508193771e-07, "loss": 0.0035, "step": 60993 }, { "epoch": 18.77, "learning_rate": 2.009323556581566e-07, "loss": 0.0045, "step": 60994 }, { "epoch": 18.77, "learning_rate": 2.0083298482537982e-07, "loss": 0.0047, "step": 60995 }, { "epoch": 18.77, "learning_rate": 2.0073363832129322e-07, "loss": 0.0028, "step": 60996 }, { "epoch": 18.77, "learning_rate": 2.0063431614614214e-07, "loss": 0.0041, "step": 60997 }, { "epoch": 18.77, "learning_rate": 2.005350183001753e-07, "loss": 0.0054, "step": 60998 }, { "epoch": 18.77, "learning_rate": 2.0043574478363804e-07, "loss": 0.0041, "step": 60999 }, { "epoch": 18.77, "learning_rate": 2.0033649559677682e-07, "loss": 0.0048, "step": 61000 }, { "epoch": 18.77, "learning_rate": 2.0023727073983812e-07, "loss": 0.0045, "step": 61001 }, { "epoch": 18.77, "learning_rate": 2.001380702130673e-07, "loss": 0.0019, "step": 61002 }, { "epoch": 18.77, "learning_rate": 2.0003889401671195e-07, "loss": 0.0045, "step": 61003 }, { "epoch": 18.77, "learning_rate": 1.9993974215101853e-07, "loss": 0.0075, "step": 61004 }, { "epoch": 18.77, "learning_rate": 1.998406146162346e-07, "loss": 0.0051, "step": 61005 }, { "epoch": 18.77, "learning_rate": 1.9974151141260223e-07, "loss": 0.0013, "step": 61006 }, { "epoch": 18.77, "learning_rate": 1.9964243254036897e-07, "loss": 0.0041, "step": 61007 }, { "epoch": 18.77, "learning_rate": 1.9954337799978242e-07, "loss": 0.0035, "step": 61008 }, { "epoch": 18.77, "learning_rate": 1.994443477910879e-07, "loss": 0.0029, "step": 61009 }, { "epoch": 18.77, "learning_rate": 1.9934534191453081e-07, "loss": 0.0027, "step": 61010 }, { "epoch": 18.77, "learning_rate": 1.992463603703565e-07, "loss": 0.0038, "step": 61011 }, { "epoch": 18.77, "learning_rate": 1.9914740315881255e-07, "loss": 0.0016, "step": 61012 }, { "epoch": 18.77, "learning_rate": 1.9904847028014207e-07, "loss": 0.0033, "step": 61013 }, { "epoch": 18.77, "learning_rate": 1.9894956173459267e-07, "loss": 0.0023, "step": 61014 }, { "epoch": 18.77, "learning_rate": 1.988506775224086e-07, "loss": 0.0028, "step": 61015 }, { "epoch": 18.77, "learning_rate": 1.9875181764383634e-07, "loss": 0.004, "step": 61016 }, { "epoch": 18.77, "learning_rate": 1.986529820991201e-07, "loss": 0.0024, "step": 61017 }, { "epoch": 18.77, "learning_rate": 1.9855417088850858e-07, "loss": 0.0024, "step": 61018 }, { "epoch": 18.77, "learning_rate": 1.9845538401224273e-07, "loss": 0.0036, "step": 61019 }, { "epoch": 18.78, "learning_rate": 1.9835662147057012e-07, "loss": 0.004, "step": 61020 }, { "epoch": 18.78, "learning_rate": 1.9825788326373607e-07, "loss": 0.0033, "step": 61021 }, { "epoch": 18.78, "learning_rate": 1.9815916939198377e-07, "loss": 0.0032, "step": 61022 }, { "epoch": 18.78, "learning_rate": 1.9806047985556077e-07, "loss": 0.0021, "step": 61023 }, { "epoch": 18.78, "learning_rate": 1.979618146547113e-07, "loss": 0.0015, "step": 61024 }, { "epoch": 18.78, "learning_rate": 1.9786317378967968e-07, "loss": 0.0037, "step": 61025 }, { "epoch": 18.78, "learning_rate": 1.977645572607123e-07, "loss": 0.0048, "step": 61026 }, { "epoch": 18.78, "learning_rate": 1.9766596506805126e-07, "loss": 0.0039, "step": 61027 }, { "epoch": 18.78, "learning_rate": 1.9756739721194408e-07, "loss": 0.0043, "step": 61028 }, { "epoch": 18.78, "learning_rate": 1.974688536926339e-07, "loss": 0.0027, "step": 61029 }, { "epoch": 18.78, "learning_rate": 1.9737033451036614e-07, "loss": 0.0027, "step": 61030 }, { "epoch": 18.78, "learning_rate": 1.9727183966538498e-07, "loss": 0.0064, "step": 61031 }, { "epoch": 18.78, "learning_rate": 1.9717336915793472e-07, "loss": 0.0028, "step": 61032 }, { "epoch": 18.78, "learning_rate": 1.970749229882607e-07, "loss": 0.0054, "step": 61033 }, { "epoch": 18.78, "learning_rate": 1.9697650115660715e-07, "loss": 0.0031, "step": 61034 }, { "epoch": 18.78, "learning_rate": 1.9687810366321835e-07, "loss": 0.0031, "step": 61035 }, { "epoch": 18.78, "learning_rate": 1.9677973050833742e-07, "loss": 0.004, "step": 61036 }, { "epoch": 18.78, "learning_rate": 1.9668138169220863e-07, "loss": 0.0015, "step": 61037 }, { "epoch": 18.78, "learning_rate": 1.9658305721507842e-07, "loss": 0.0041, "step": 61038 }, { "epoch": 18.78, "learning_rate": 1.9648475707719106e-07, "loss": 0.0028, "step": 61039 }, { "epoch": 18.78, "learning_rate": 1.9638648127878747e-07, "loss": 0.0016, "step": 61040 }, { "epoch": 18.78, "learning_rate": 1.9628822982011408e-07, "loss": 0.0066, "step": 61041 }, { "epoch": 18.78, "learning_rate": 1.9619000270141186e-07, "loss": 0.0025, "step": 61042 }, { "epoch": 18.78, "learning_rate": 1.9609179992292836e-07, "loss": 0.0044, "step": 61043 }, { "epoch": 18.78, "learning_rate": 1.9599362148490564e-07, "loss": 0.0038, "step": 61044 }, { "epoch": 18.78, "learning_rate": 1.9589546738758792e-07, "loss": 0.0032, "step": 61045 }, { "epoch": 18.78, "learning_rate": 1.9579733763121943e-07, "loss": 0.003, "step": 61046 }, { "epoch": 18.78, "learning_rate": 1.9569923221604224e-07, "loss": 0.0034, "step": 61047 }, { "epoch": 18.78, "learning_rate": 1.956011511423006e-07, "loss": 0.0037, "step": 61048 }, { "epoch": 18.78, "learning_rate": 1.9550309441023761e-07, "loss": 0.0025, "step": 61049 }, { "epoch": 18.78, "learning_rate": 1.9540506202009756e-07, "loss": 0.003, "step": 61050 }, { "epoch": 18.78, "learning_rate": 1.9530705397212468e-07, "loss": 0.0034, "step": 61051 }, { "epoch": 18.78, "learning_rate": 1.952090702665599e-07, "loss": 0.0038, "step": 61052 }, { "epoch": 18.79, "learning_rate": 1.9511111090364742e-07, "loss": 0.0038, "step": 61053 }, { "epoch": 18.79, "learning_rate": 1.9511111090364742e-07, "loss": 0.008, "step": 61054 }, { "epoch": 18.79, "learning_rate": 1.9501317588363155e-07, "loss": 0.0039, "step": 61055 }, { "epoch": 18.79, "learning_rate": 1.949152652067543e-07, "loss": 0.0029, "step": 61056 }, { "epoch": 18.79, "learning_rate": 1.948173788732599e-07, "loss": 0.0025, "step": 61057 }, { "epoch": 18.79, "learning_rate": 1.947195168833882e-07, "loss": 0.0033, "step": 61058 }, { "epoch": 18.79, "learning_rate": 1.9462167923738673e-07, "loss": 0.0023, "step": 61059 }, { "epoch": 18.79, "learning_rate": 1.9452386593549534e-07, "loss": 0.0044, "step": 61060 }, { "epoch": 18.79, "learning_rate": 1.9442607697795934e-07, "loss": 0.0037, "step": 61061 }, { "epoch": 18.79, "learning_rate": 1.943283123650186e-07, "loss": 0.0037, "step": 61062 }, { "epoch": 18.79, "learning_rate": 1.942305720969162e-07, "loss": 0.0033, "step": 61063 }, { "epoch": 18.79, "learning_rate": 1.9413285617389644e-07, "loss": 0.0029, "step": 61064 }, { "epoch": 18.79, "learning_rate": 1.940351645962013e-07, "loss": 0.0049, "step": 61065 }, { "epoch": 18.79, "learning_rate": 1.9393749736407287e-07, "loss": 0.0032, "step": 61066 }, { "epoch": 18.79, "learning_rate": 1.938398544777542e-07, "loss": 0.003, "step": 61067 }, { "epoch": 18.79, "learning_rate": 1.9374223593748854e-07, "loss": 0.003, "step": 61068 }, { "epoch": 18.79, "learning_rate": 1.936446417435156e-07, "loss": 0.0038, "step": 61069 }, { "epoch": 18.79, "learning_rate": 1.9354707189608079e-07, "loss": 0.003, "step": 61070 }, { "epoch": 18.79, "learning_rate": 1.9344952639542391e-07, "loss": 0.0034, "step": 61071 }, { "epoch": 18.79, "learning_rate": 1.933520052417881e-07, "loss": 0.0028, "step": 61072 }, { "epoch": 18.79, "learning_rate": 1.932545084354154e-07, "loss": 0.0032, "step": 61073 }, { "epoch": 18.79, "learning_rate": 1.931570359765489e-07, "loss": 0.0049, "step": 61074 }, { "epoch": 18.79, "learning_rate": 1.9305958786542954e-07, "loss": 0.0027, "step": 61075 }, { "epoch": 18.79, "learning_rate": 1.9296216410229828e-07, "loss": 0.0072, "step": 61076 }, { "epoch": 18.79, "learning_rate": 1.9286476468739934e-07, "loss": 0.0038, "step": 61077 }, { "epoch": 18.79, "learning_rate": 1.9276738962097142e-07, "loss": 0.0037, "step": 61078 }, { "epoch": 18.79, "learning_rate": 1.9267003890325875e-07, "loss": 0.0051, "step": 61079 }, { "epoch": 18.79, "learning_rate": 1.925727125345034e-07, "loss": 0.0028, "step": 61080 }, { "epoch": 18.79, "learning_rate": 1.9247541051494512e-07, "loss": 0.0038, "step": 61081 }, { "epoch": 18.79, "learning_rate": 1.923781328448282e-07, "loss": 0.0046, "step": 61082 }, { "epoch": 18.79, "learning_rate": 1.9228087952439022e-07, "loss": 0.0048, "step": 61083 }, { "epoch": 18.79, "learning_rate": 1.9218365055387545e-07, "loss": 0.0036, "step": 61084 }, { "epoch": 18.8, "learning_rate": 1.9208644593352366e-07, "loss": 0.0032, "step": 61085 }, { "epoch": 18.8, "learning_rate": 1.91989265663578e-07, "loss": 0.0051, "step": 61086 }, { "epoch": 18.8, "learning_rate": 1.9189210974427941e-07, "loss": 0.0018, "step": 61087 }, { "epoch": 18.8, "learning_rate": 1.9179497817586768e-07, "loss": 0.0026, "step": 61088 }, { "epoch": 18.8, "learning_rate": 1.9169787095858484e-07, "loss": 0.0033, "step": 61089 }, { "epoch": 18.8, "learning_rate": 1.9160078809267292e-07, "loss": 0.0042, "step": 61090 }, { "epoch": 18.8, "learning_rate": 1.915037295783706e-07, "loss": 0.0029, "step": 61091 }, { "epoch": 18.8, "learning_rate": 1.9140669541592104e-07, "loss": 0.0027, "step": 61092 }, { "epoch": 18.8, "learning_rate": 1.9130968560556294e-07, "loss": 0.0078, "step": 61093 }, { "epoch": 18.8, "learning_rate": 1.9121270014754057e-07, "loss": 0.0021, "step": 61094 }, { "epoch": 18.8, "learning_rate": 1.9111573904209257e-07, "loss": 0.0041, "step": 61095 }, { "epoch": 18.8, "learning_rate": 1.910188022894588e-07, "loss": 0.0031, "step": 61096 }, { "epoch": 18.8, "learning_rate": 1.9092188988988126e-07, "loss": 0.0046, "step": 61097 }, { "epoch": 18.8, "learning_rate": 1.9082500184359976e-07, "loss": 0.0025, "step": 61098 }, { "epoch": 18.8, "learning_rate": 1.9072813815085523e-07, "loss": 0.0044, "step": 61099 }, { "epoch": 18.8, "learning_rate": 1.9063129881188857e-07, "loss": 0.0031, "step": 61100 }, { "epoch": 18.8, "learning_rate": 1.9053448382694073e-07, "loss": 0.0045, "step": 61101 }, { "epoch": 18.8, "learning_rate": 1.9043769319625039e-07, "loss": 0.0041, "step": 61102 }, { "epoch": 18.8, "learning_rate": 1.9034092692005846e-07, "loss": 0.0038, "step": 61103 }, { "epoch": 18.8, "learning_rate": 1.9024418499860587e-07, "loss": 0.0036, "step": 61104 }, { "epoch": 18.8, "learning_rate": 1.9014746743213243e-07, "loss": 0.0026, "step": 61105 }, { "epoch": 18.8, "learning_rate": 1.900507742208779e-07, "loss": 0.0025, "step": 61106 }, { "epoch": 18.8, "learning_rate": 1.8995410536508218e-07, "loss": 0.003, "step": 61107 }, { "epoch": 18.8, "learning_rate": 1.8985746086498724e-07, "loss": 0.0031, "step": 61108 }, { "epoch": 18.8, "learning_rate": 1.8976084072082957e-07, "loss": 0.0045, "step": 61109 }, { "epoch": 18.8, "learning_rate": 1.896642449328523e-07, "loss": 0.0044, "step": 61110 }, { "epoch": 18.8, "learning_rate": 1.8956767350129302e-07, "loss": 0.004, "step": 61111 }, { "epoch": 18.8, "learning_rate": 1.8947112642639376e-07, "loss": 0.0036, "step": 61112 }, { "epoch": 18.8, "learning_rate": 1.89374603708391e-07, "loss": 0.0048, "step": 61113 }, { "epoch": 18.8, "learning_rate": 1.8927810534752679e-07, "loss": 0.0038, "step": 61114 }, { "epoch": 18.8, "learning_rate": 1.8918163134404087e-07, "loss": 0.0025, "step": 61115 }, { "epoch": 18.8, "learning_rate": 1.8908518169817312e-07, "loss": 0.0033, "step": 61116 }, { "epoch": 18.8, "learning_rate": 1.8898875641016002e-07, "loss": 0.0039, "step": 61117 }, { "epoch": 18.81, "learning_rate": 1.8889235548024244e-07, "loss": 0.0031, "step": 61118 }, { "epoch": 18.81, "learning_rate": 1.887959789086613e-07, "loss": 0.0044, "step": 61119 }, { "epoch": 18.81, "learning_rate": 1.8869962669565422e-07, "loss": 0.0037, "step": 61120 }, { "epoch": 18.81, "learning_rate": 1.8860329884146212e-07, "loss": 0.0046, "step": 61121 }, { "epoch": 18.81, "learning_rate": 1.8850699534632145e-07, "loss": 0.0026, "step": 61122 }, { "epoch": 18.81, "learning_rate": 1.8841071621047314e-07, "loss": 0.0033, "step": 61123 }, { "epoch": 18.81, "learning_rate": 1.88314461434157e-07, "loss": 0.0046, "step": 61124 }, { "epoch": 18.81, "learning_rate": 1.8821823101760949e-07, "loss": 0.0037, "step": 61125 }, { "epoch": 18.81, "learning_rate": 1.8812202496107156e-07, "loss": 0.0042, "step": 61126 }, { "epoch": 18.81, "learning_rate": 1.880258432647819e-07, "loss": 0.0037, "step": 61127 }, { "epoch": 18.81, "learning_rate": 1.87929685928977e-07, "loss": 0.0046, "step": 61128 }, { "epoch": 18.81, "learning_rate": 1.8783355295389995e-07, "loss": 0.0033, "step": 61129 }, { "epoch": 18.81, "learning_rate": 1.8773744433978613e-07, "loss": 0.0035, "step": 61130 }, { "epoch": 18.81, "learning_rate": 1.8764136008687428e-07, "loss": 0.0038, "step": 61131 }, { "epoch": 18.81, "learning_rate": 1.8754530019540418e-07, "loss": 0.0055, "step": 61132 }, { "epoch": 18.81, "learning_rate": 1.874492646656123e-07, "loss": 0.0029, "step": 61133 }, { "epoch": 18.81, "learning_rate": 1.8735325349773958e-07, "loss": 0.0034, "step": 61134 }, { "epoch": 18.81, "learning_rate": 1.872572666920236e-07, "loss": 0.0043, "step": 61135 }, { "epoch": 18.81, "learning_rate": 1.8716130424870193e-07, "loss": 0.0047, "step": 61136 }, { "epoch": 18.81, "learning_rate": 1.870653661680144e-07, "loss": 0.0008, "step": 61137 }, { "epoch": 18.81, "learning_rate": 1.8696945245019638e-07, "loss": 0.0021, "step": 61138 }, { "epoch": 18.81, "learning_rate": 1.8687356309548876e-07, "loss": 0.0043, "step": 61139 }, { "epoch": 18.81, "learning_rate": 1.8677769810412914e-07, "loss": 0.0068, "step": 61140 }, { "epoch": 18.81, "learning_rate": 1.8668185747635403e-07, "loss": 0.0027, "step": 61141 }, { "epoch": 18.81, "learning_rate": 1.8658604121240208e-07, "loss": 0.0029, "step": 61142 }, { "epoch": 18.81, "learning_rate": 1.8649024931251204e-07, "loss": 0.004, "step": 61143 }, { "epoch": 18.81, "learning_rate": 1.8639448177692032e-07, "loss": 0.0051, "step": 61144 }, { "epoch": 18.81, "learning_rate": 1.8629873860586567e-07, "loss": 0.0028, "step": 61145 }, { "epoch": 18.81, "learning_rate": 1.8620301979958566e-07, "loss": 0.0041, "step": 61146 }, { "epoch": 18.81, "learning_rate": 1.861073253583179e-07, "loss": 0.0043, "step": 61147 }, { "epoch": 18.81, "learning_rate": 1.860116552822999e-07, "loss": 0.0048, "step": 61148 }, { "epoch": 18.81, "learning_rate": 1.8591600957176936e-07, "loss": 0.004, "step": 61149 }, { "epoch": 18.82, "learning_rate": 1.858203882269638e-07, "loss": 0.0036, "step": 61150 }, { "epoch": 18.82, "learning_rate": 1.857247912481197e-07, "loss": 0.0028, "step": 61151 }, { "epoch": 18.82, "learning_rate": 1.8562921863547577e-07, "loss": 0.0041, "step": 61152 }, { "epoch": 18.82, "learning_rate": 1.855336703892674e-07, "loss": 0.0075, "step": 61153 }, { "epoch": 18.82, "learning_rate": 1.8543814650973434e-07, "loss": 0.0093, "step": 61154 }, { "epoch": 18.82, "learning_rate": 1.8534264699711091e-07, "loss": 0.003, "step": 61155 }, { "epoch": 18.82, "learning_rate": 1.8524717185163687e-07, "loss": 0.0063, "step": 61156 }, { "epoch": 18.82, "learning_rate": 1.8515172107354873e-07, "loss": 0.0025, "step": 61157 }, { "epoch": 18.82, "learning_rate": 1.8505629466308184e-07, "loss": 0.003, "step": 61158 }, { "epoch": 18.82, "learning_rate": 1.849608926204749e-07, "loss": 0.0041, "step": 61159 }, { "epoch": 18.82, "learning_rate": 1.8486551494596327e-07, "loss": 0.0061, "step": 61160 }, { "epoch": 18.82, "learning_rate": 1.8477016163978456e-07, "loss": 0.0031, "step": 61161 }, { "epoch": 18.82, "learning_rate": 1.8467483270217634e-07, "loss": 0.0051, "step": 61162 }, { "epoch": 18.82, "learning_rate": 1.8457952813337287e-07, "loss": 0.0023, "step": 61163 }, { "epoch": 18.82, "learning_rate": 1.8448424793361398e-07, "loss": 0.0047, "step": 61164 }, { "epoch": 18.82, "learning_rate": 1.8438899210313388e-07, "loss": 0.0026, "step": 61165 }, { "epoch": 18.82, "learning_rate": 1.8429376064216908e-07, "loss": 0.0036, "step": 61166 }, { "epoch": 18.82, "learning_rate": 1.8419855355095717e-07, "loss": 0.0034, "step": 61167 }, { "epoch": 18.82, "learning_rate": 1.841033708297335e-07, "loss": 0.004, "step": 61168 }, { "epoch": 18.82, "learning_rate": 1.8400821247873457e-07, "loss": 0.0047, "step": 61169 }, { "epoch": 18.82, "learning_rate": 1.8391307849819794e-07, "loss": 0.0031, "step": 61170 }, { "epoch": 18.82, "learning_rate": 1.83817968888359e-07, "loss": 0.0041, "step": 61171 }, { "epoch": 18.82, "learning_rate": 1.8372288364945424e-07, "loss": 0.0034, "step": 61172 }, { "epoch": 18.82, "learning_rate": 1.8362782278171675e-07, "loss": 0.0027, "step": 61173 }, { "epoch": 18.82, "learning_rate": 1.835327862853864e-07, "loss": 0.0023, "step": 61174 }, { "epoch": 18.82, "learning_rate": 1.834377741606974e-07, "loss": 0.0023, "step": 61175 }, { "epoch": 18.82, "learning_rate": 1.8334278640788627e-07, "loss": 0.0036, "step": 61176 }, { "epoch": 18.82, "learning_rate": 1.8324782302718835e-07, "loss": 0.0073, "step": 61177 }, { "epoch": 18.82, "learning_rate": 1.8315288401884013e-07, "loss": 0.0036, "step": 61178 }, { "epoch": 18.82, "learning_rate": 1.8305796938307586e-07, "loss": 0.004, "step": 61179 }, { "epoch": 18.82, "learning_rate": 1.829630791201331e-07, "loss": 0.0024, "step": 61180 }, { "epoch": 18.82, "learning_rate": 1.8286821323024507e-07, "loss": 0.0035, "step": 61181 }, { "epoch": 18.82, "learning_rate": 1.8277337171364928e-07, "loss": 0.007, "step": 61182 }, { "epoch": 18.83, "learning_rate": 1.8267855457058004e-07, "loss": 0.0029, "step": 61183 }, { "epoch": 18.83, "learning_rate": 1.825837618012738e-07, "loss": 0.0062, "step": 61184 }, { "epoch": 18.83, "learning_rate": 1.8248899340596705e-07, "loss": 0.0045, "step": 61185 }, { "epoch": 18.83, "learning_rate": 1.823942493848918e-07, "loss": 0.0034, "step": 61186 }, { "epoch": 18.83, "learning_rate": 1.8229952973828458e-07, "loss": 0.0037, "step": 61187 }, { "epoch": 18.83, "learning_rate": 1.822048344663807e-07, "loss": 0.0045, "step": 61188 }, { "epoch": 18.83, "learning_rate": 1.8211016356941559e-07, "loss": 0.0034, "step": 61189 }, { "epoch": 18.83, "learning_rate": 1.8201551704762453e-07, "loss": 0.0044, "step": 61190 }, { "epoch": 18.83, "learning_rate": 1.8192089490124188e-07, "loss": 0.0037, "step": 61191 }, { "epoch": 18.83, "learning_rate": 1.8182629713050403e-07, "loss": 0.0114, "step": 61192 }, { "epoch": 18.83, "learning_rate": 1.8173172373564306e-07, "loss": 0.0029, "step": 61193 }, { "epoch": 18.83, "learning_rate": 1.8163717471689546e-07, "loss": 0.0033, "step": 61194 }, { "epoch": 18.83, "learning_rate": 1.8154265007449657e-07, "loss": 0.0032, "step": 61195 }, { "epoch": 18.83, "learning_rate": 1.8144814980867954e-07, "loss": 0.0033, "step": 61196 }, { "epoch": 18.83, "learning_rate": 1.8135367391967974e-07, "loss": 0.004, "step": 61197 }, { "epoch": 18.83, "learning_rate": 1.8125922240773252e-07, "loss": 0.0036, "step": 61198 }, { "epoch": 18.83, "learning_rate": 1.8116479527307108e-07, "loss": 0.0038, "step": 61199 }, { "epoch": 18.83, "learning_rate": 1.8107039251593072e-07, "loss": 0.004, "step": 61200 }, { "epoch": 18.83, "learning_rate": 1.8097601413654464e-07, "loss": 0.0041, "step": 61201 }, { "epoch": 18.83, "learning_rate": 1.8088166013514819e-07, "loss": 0.0025, "step": 61202 }, { "epoch": 18.83, "learning_rate": 1.8078733051197561e-07, "loss": 0.0052, "step": 61203 }, { "epoch": 18.83, "learning_rate": 1.8069302526726118e-07, "loss": 0.004, "step": 61204 }, { "epoch": 18.83, "learning_rate": 1.8059874440123804e-07, "loss": 0.0024, "step": 61205 }, { "epoch": 18.83, "learning_rate": 1.8050448791414266e-07, "loss": 0.0057, "step": 61206 }, { "epoch": 18.83, "learning_rate": 1.8041025580620597e-07, "loss": 0.0023, "step": 61207 }, { "epoch": 18.83, "learning_rate": 1.8031604807766334e-07, "loss": 0.002, "step": 61208 }, { "epoch": 18.83, "learning_rate": 1.8022186472874903e-07, "loss": 0.0027, "step": 61209 }, { "epoch": 18.83, "learning_rate": 1.8012770575969617e-07, "loss": 0.0048, "step": 61210 }, { "epoch": 18.83, "learning_rate": 1.8003357117073904e-07, "loss": 0.005, "step": 61211 }, { "epoch": 18.83, "learning_rate": 1.799394609621119e-07, "loss": 0.0044, "step": 61212 }, { "epoch": 18.83, "learning_rate": 1.7984537513404677e-07, "loss": 0.0046, "step": 61213 }, { "epoch": 18.83, "learning_rate": 1.79751313686779e-07, "loss": 0.0023, "step": 61214 }, { "epoch": 18.84, "learning_rate": 1.7965727662054068e-07, "loss": 0.0022, "step": 61215 }, { "epoch": 18.84, "learning_rate": 1.79563263935566e-07, "loss": 0.0051, "step": 61216 }, { "epoch": 18.84, "learning_rate": 1.7946927563208706e-07, "loss": 0.0065, "step": 61217 }, { "epoch": 18.84, "learning_rate": 1.7937531171033916e-07, "loss": 0.0024, "step": 61218 }, { "epoch": 18.84, "learning_rate": 1.792813721705555e-07, "loss": 0.0037, "step": 61219 }, { "epoch": 18.84, "learning_rate": 1.7918745701296815e-07, "loss": 0.0033, "step": 61220 }, { "epoch": 18.84, "learning_rate": 1.7909356623781127e-07, "loss": 0.0038, "step": 61221 }, { "epoch": 18.84, "learning_rate": 1.7899969984531696e-07, "loss": 0.0038, "step": 61222 }, { "epoch": 18.84, "learning_rate": 1.7890585783571723e-07, "loss": 0.005, "step": 61223 }, { "epoch": 18.84, "learning_rate": 1.7881204020924857e-07, "loss": 0.0043, "step": 61224 }, { "epoch": 18.84, "learning_rate": 1.787182469661408e-07, "loss": 0.0028, "step": 61225 }, { "epoch": 18.84, "learning_rate": 1.786244781066293e-07, "loss": 0.0033, "step": 61226 }, { "epoch": 18.84, "learning_rate": 1.7853073363094498e-07, "loss": 0.0017, "step": 61227 }, { "epoch": 18.84, "learning_rate": 1.7843701353931985e-07, "loss": 0.0027, "step": 61228 }, { "epoch": 18.84, "learning_rate": 1.7834331783198933e-07, "loss": 0.007, "step": 61229 }, { "epoch": 18.84, "learning_rate": 1.782496465091832e-07, "loss": 0.007, "step": 61230 }, { "epoch": 18.84, "learning_rate": 1.7815599957113573e-07, "loss": 0.0062, "step": 61231 }, { "epoch": 18.84, "learning_rate": 1.7806237701808004e-07, "loss": 0.0038, "step": 61232 }, { "epoch": 18.84, "learning_rate": 1.77968778850246e-07, "loss": 0.0025, "step": 61233 }, { "epoch": 18.84, "learning_rate": 1.7787520506786894e-07, "loss": 0.0037, "step": 61234 }, { "epoch": 18.84, "learning_rate": 1.7778165567117865e-07, "loss": 0.0032, "step": 61235 }, { "epoch": 18.84, "learning_rate": 1.7768813066040945e-07, "loss": 0.0037, "step": 61236 }, { "epoch": 18.84, "learning_rate": 1.7759463003579225e-07, "loss": 0.0028, "step": 61237 }, { "epoch": 18.84, "learning_rate": 1.7750115379755904e-07, "loss": 0.0035, "step": 61238 }, { "epoch": 18.84, "learning_rate": 1.7740770194594193e-07, "loss": 0.0044, "step": 61239 }, { "epoch": 18.84, "learning_rate": 1.773142744811751e-07, "loss": 0.0032, "step": 61240 }, { "epoch": 18.84, "learning_rate": 1.7722087140348844e-07, "loss": 0.0037, "step": 61241 }, { "epoch": 18.84, "learning_rate": 1.7712749271311392e-07, "loss": 0.004, "step": 61242 }, { "epoch": 18.84, "learning_rate": 1.770341384102825e-07, "loss": 0.0018, "step": 61243 }, { "epoch": 18.84, "learning_rate": 1.7694080849522844e-07, "loss": 0.0042, "step": 61244 }, { "epoch": 18.84, "learning_rate": 1.7684750296818154e-07, "loss": 0.0044, "step": 61245 }, { "epoch": 18.84, "learning_rate": 1.7675422182937496e-07, "loss": 0.003, "step": 61246 }, { "epoch": 18.84, "learning_rate": 1.7666096507903964e-07, "loss": 0.0031, "step": 61247 }, { "epoch": 18.85, "learning_rate": 1.7656773271740536e-07, "loss": 0.0055, "step": 61248 }, { "epoch": 18.85, "learning_rate": 1.7647452474470527e-07, "loss": 0.0042, "step": 61249 }, { "epoch": 18.85, "learning_rate": 1.7638134116117146e-07, "loss": 0.0034, "step": 61250 }, { "epoch": 18.85, "learning_rate": 1.7628818196703368e-07, "loss": 0.0024, "step": 61251 }, { "epoch": 18.85, "learning_rate": 1.76195047162524e-07, "loss": 0.0032, "step": 61252 }, { "epoch": 18.85, "learning_rate": 1.7610193674787446e-07, "loss": 0.0031, "step": 61253 }, { "epoch": 18.85, "learning_rate": 1.7600885072331485e-07, "loss": 0.0026, "step": 61254 }, { "epoch": 18.85, "learning_rate": 1.7591578908907724e-07, "loss": 0.0037, "step": 61255 }, { "epoch": 18.85, "learning_rate": 1.7582275184539145e-07, "loss": 0.0045, "step": 61256 }, { "epoch": 18.85, "learning_rate": 1.7572973899248947e-07, "loss": 0.0039, "step": 61257 }, { "epoch": 18.85, "learning_rate": 1.7563675053060115e-07, "loss": 0.0034, "step": 61258 }, { "epoch": 18.85, "learning_rate": 1.755437864599596e-07, "loss": 0.0036, "step": 61259 }, { "epoch": 18.85, "learning_rate": 1.754508467807936e-07, "loss": 0.0041, "step": 61260 }, { "epoch": 18.85, "learning_rate": 1.7535793149333514e-07, "loss": 0.0039, "step": 61261 }, { "epoch": 18.85, "learning_rate": 1.7526504059781403e-07, "loss": 0.0041, "step": 61262 }, { "epoch": 18.85, "learning_rate": 1.7517217409446008e-07, "loss": 0.0038, "step": 61263 }, { "epoch": 18.85, "learning_rate": 1.7507933198350646e-07, "loss": 0.0037, "step": 61264 }, { "epoch": 18.85, "learning_rate": 1.7498651426518075e-07, "loss": 0.0041, "step": 61265 }, { "epoch": 18.85, "learning_rate": 1.7489372093971613e-07, "loss": 0.0031, "step": 61266 }, { "epoch": 18.85, "learning_rate": 1.7480095200734125e-07, "loss": 0.004, "step": 61267 }, { "epoch": 18.85, "learning_rate": 1.74708207468286e-07, "loss": 0.0071, "step": 61268 }, { "epoch": 18.85, "learning_rate": 1.7461548732278234e-07, "loss": 0.0029, "step": 61269 }, { "epoch": 18.85, "learning_rate": 1.7452279157105902e-07, "loss": 0.0026, "step": 61270 }, { "epoch": 18.85, "learning_rate": 1.7443012021334803e-07, "loss": 0.0047, "step": 61271 }, { "epoch": 18.85, "learning_rate": 1.7433747324987705e-07, "loss": 0.0028, "step": 61272 }, { "epoch": 18.85, "learning_rate": 1.742448506808758e-07, "loss": 0.0024, "step": 61273 }, { "epoch": 18.85, "learning_rate": 1.7415225250657864e-07, "loss": 0.0035, "step": 61274 }, { "epoch": 18.85, "learning_rate": 1.7405967872721085e-07, "loss": 0.0048, "step": 61275 }, { "epoch": 18.85, "learning_rate": 1.739671293430045e-07, "loss": 0.0032, "step": 61276 }, { "epoch": 18.85, "learning_rate": 1.7387460435418834e-07, "loss": 0.0058, "step": 61277 }, { "epoch": 18.85, "learning_rate": 1.7378210376099104e-07, "loss": 0.0036, "step": 61278 }, { "epoch": 18.85, "learning_rate": 1.7368962756364573e-07, "loss": 0.0046, "step": 61279 }, { "epoch": 18.86, "learning_rate": 1.735971757623789e-07, "loss": 0.0041, "step": 61280 }, { "epoch": 18.86, "learning_rate": 1.735047483574215e-07, "loss": 0.002, "step": 61281 }, { "epoch": 18.86, "learning_rate": 1.734123453490033e-07, "loss": 0.005, "step": 61282 }, { "epoch": 18.86, "learning_rate": 1.7331996673735198e-07, "loss": 0.0028, "step": 61283 }, { "epoch": 18.86, "learning_rate": 1.732276125226995e-07, "loss": 0.0043, "step": 61284 }, { "epoch": 18.86, "learning_rate": 1.7313528270527236e-07, "loss": 0.0026, "step": 61285 }, { "epoch": 18.86, "learning_rate": 1.730429772853015e-07, "loss": 0.0027, "step": 61286 }, { "epoch": 18.86, "learning_rate": 1.7295069626301565e-07, "loss": 0.0023, "step": 61287 }, { "epoch": 18.86, "learning_rate": 1.7285843963864347e-07, "loss": 0.0036, "step": 61288 }, { "epoch": 18.86, "learning_rate": 1.727662074124148e-07, "loss": 0.004, "step": 61289 }, { "epoch": 18.86, "learning_rate": 1.7267399958455943e-07, "loss": 0.0032, "step": 61290 }, { "epoch": 18.86, "learning_rate": 1.7258181615530389e-07, "loss": 0.0033, "step": 61291 }, { "epoch": 18.86, "learning_rate": 1.7248965712487797e-07, "loss": 0.0041, "step": 61292 }, { "epoch": 18.86, "learning_rate": 1.7239752249351037e-07, "loss": 0.0034, "step": 61293 }, { "epoch": 18.86, "learning_rate": 1.7230541226143205e-07, "loss": 0.0055, "step": 61294 }, { "epoch": 18.86, "learning_rate": 1.7221332642886946e-07, "loss": 0.0053, "step": 61295 }, { "epoch": 18.86, "learning_rate": 1.721212649960513e-07, "loss": 0.0042, "step": 61296 }, { "epoch": 18.86, "learning_rate": 1.720292279632063e-07, "loss": 0.0029, "step": 61297 }, { "epoch": 18.86, "learning_rate": 1.7193721533056318e-07, "loss": 0.0096, "step": 61298 }, { "epoch": 18.86, "learning_rate": 1.718452270983506e-07, "loss": 0.0037, "step": 61299 }, { "epoch": 18.86, "learning_rate": 1.7175326326679732e-07, "loss": 0.0046, "step": 61300 }, { "epoch": 18.86, "learning_rate": 1.716613238361309e-07, "loss": 0.0039, "step": 61301 }, { "epoch": 18.86, "learning_rate": 1.715694088065789e-07, "loss": 0.0022, "step": 61302 }, { "epoch": 18.86, "learning_rate": 1.714775181783712e-07, "loss": 0.0054, "step": 61303 }, { "epoch": 18.86, "learning_rate": 1.713856519517354e-07, "loss": 0.0016, "step": 61304 }, { "epoch": 18.86, "learning_rate": 1.7129381012689906e-07, "loss": 0.0036, "step": 61305 }, { "epoch": 18.86, "learning_rate": 1.7120199270408978e-07, "loss": 0.0029, "step": 61306 }, { "epoch": 18.86, "learning_rate": 1.7111019968353625e-07, "loss": 0.0032, "step": 61307 }, { "epoch": 18.86, "learning_rate": 1.7101843106546723e-07, "loss": 0.0031, "step": 61308 }, { "epoch": 18.86, "learning_rate": 1.7092668685010916e-07, "loss": 0.0024, "step": 61309 }, { "epoch": 18.86, "learning_rate": 1.7083496703768965e-07, "loss": 0.0031, "step": 61310 }, { "epoch": 18.86, "learning_rate": 1.707432716284385e-07, "loss": 0.0049, "step": 61311 }, { "epoch": 18.86, "learning_rate": 1.7065160062258e-07, "loss": 0.0027, "step": 61312 }, { "epoch": 18.87, "learning_rate": 1.7055995402034507e-07, "loss": 0.0052, "step": 61313 }, { "epoch": 18.87, "learning_rate": 1.704683318219591e-07, "loss": 0.0024, "step": 61314 }, { "epoch": 18.87, "learning_rate": 1.7037673402765076e-07, "loss": 0.0031, "step": 61315 }, { "epoch": 18.87, "learning_rate": 1.7028516063764766e-07, "loss": 0.0037, "step": 61316 }, { "epoch": 18.87, "learning_rate": 1.7019361165217518e-07, "loss": 0.0044, "step": 61317 }, { "epoch": 18.87, "learning_rate": 1.7010208707146202e-07, "loss": 0.0053, "step": 61318 }, { "epoch": 18.87, "learning_rate": 1.7001058689573579e-07, "loss": 0.0035, "step": 61319 }, { "epoch": 18.87, "learning_rate": 1.699191111252241e-07, "loss": 0.0029, "step": 61320 }, { "epoch": 18.87, "learning_rate": 1.6982765976015226e-07, "loss": 0.0027, "step": 61321 }, { "epoch": 18.87, "learning_rate": 1.6973623280074792e-07, "loss": 0.0031, "step": 61322 }, { "epoch": 18.87, "learning_rate": 1.6964483024723865e-07, "loss": 0.0026, "step": 61323 }, { "epoch": 18.87, "learning_rate": 1.6955345209985098e-07, "loss": 0.0018, "step": 61324 }, { "epoch": 18.87, "learning_rate": 1.6946209835881244e-07, "loss": 0.0032, "step": 61325 }, { "epoch": 18.87, "learning_rate": 1.6937076902434846e-07, "loss": 0.0043, "step": 61326 }, { "epoch": 18.87, "learning_rate": 1.6927946409668662e-07, "loss": 0.0034, "step": 61327 }, { "epoch": 18.87, "learning_rate": 1.6918818357605336e-07, "loss": 0.0036, "step": 61328 }, { "epoch": 18.87, "learning_rate": 1.6909692746267748e-07, "loss": 0.0039, "step": 61329 }, { "epoch": 18.87, "learning_rate": 1.6900569575678204e-07, "loss": 0.0041, "step": 61330 }, { "epoch": 18.87, "learning_rate": 1.689144884585947e-07, "loss": 0.0032, "step": 61331 }, { "epoch": 18.87, "learning_rate": 1.6882330556834304e-07, "loss": 0.0036, "step": 61332 }, { "epoch": 18.87, "learning_rate": 1.687321470862524e-07, "loss": 0.0036, "step": 61333 }, { "epoch": 18.87, "learning_rate": 1.6864101301254932e-07, "loss": 0.0024, "step": 61334 }, { "epoch": 18.87, "learning_rate": 1.6854990334746025e-07, "loss": 0.0033, "step": 61335 }, { "epoch": 18.87, "learning_rate": 1.6845881809121167e-07, "loss": 0.0036, "step": 61336 }, { "epoch": 18.87, "learning_rate": 1.6836775724402898e-07, "loss": 0.0022, "step": 61337 }, { "epoch": 18.87, "learning_rate": 1.6827672080613866e-07, "loss": 0.0034, "step": 61338 }, { "epoch": 18.87, "learning_rate": 1.681857087777672e-07, "loss": 0.0027, "step": 61339 }, { "epoch": 18.87, "learning_rate": 1.6809472115913993e-07, "loss": 0.0049, "step": 61340 }, { "epoch": 18.87, "learning_rate": 1.680037579504823e-07, "loss": 0.0033, "step": 61341 }, { "epoch": 18.87, "learning_rate": 1.6791281915202074e-07, "loss": 0.0036, "step": 61342 }, { "epoch": 18.87, "learning_rate": 1.6782190476398174e-07, "loss": 0.0031, "step": 61343 }, { "epoch": 18.87, "learning_rate": 1.677310147865896e-07, "loss": 0.0028, "step": 61344 }, { "epoch": 18.88, "learning_rate": 1.6764014922007078e-07, "loss": 0.0069, "step": 61345 }, { "epoch": 18.88, "learning_rate": 1.6754930806465063e-07, "loss": 0.0029, "step": 61346 }, { "epoch": 18.88, "learning_rate": 1.674584913205546e-07, "loss": 0.0066, "step": 61347 }, { "epoch": 18.88, "learning_rate": 1.673676989880091e-07, "loss": 0.0051, "step": 61348 }, { "epoch": 18.88, "learning_rate": 1.672769310672384e-07, "loss": 0.0037, "step": 61349 }, { "epoch": 18.88, "learning_rate": 1.6718618755846906e-07, "loss": 0.0029, "step": 61350 }, { "epoch": 18.88, "learning_rate": 1.6709546846192527e-07, "loss": 0.0043, "step": 61351 }, { "epoch": 18.88, "learning_rate": 1.670047737778313e-07, "loss": 0.0035, "step": 61352 }, { "epoch": 18.88, "learning_rate": 1.669141035064137e-07, "loss": 0.0027, "step": 61353 }, { "epoch": 18.88, "learning_rate": 1.6682345764789886e-07, "loss": 0.0037, "step": 61354 }, { "epoch": 18.88, "learning_rate": 1.6673283620251002e-07, "loss": 0.0044, "step": 61355 }, { "epoch": 18.88, "learning_rate": 1.666422391704714e-07, "loss": 0.0055, "step": 61356 }, { "epoch": 18.88, "learning_rate": 1.665516665520106e-07, "loss": 0.0017, "step": 61357 }, { "epoch": 18.88, "learning_rate": 1.664611183473508e-07, "loss": 0.0015, "step": 61358 }, { "epoch": 18.88, "learning_rate": 1.6637059455671623e-07, "loss": 0.0025, "step": 61359 }, { "epoch": 18.88, "learning_rate": 1.6628009518033229e-07, "loss": 0.0031, "step": 61360 }, { "epoch": 18.88, "learning_rate": 1.6618962021842434e-07, "loss": 0.007, "step": 61361 }, { "epoch": 18.88, "learning_rate": 1.6609916967121664e-07, "loss": 0.0026, "step": 61362 }, { "epoch": 18.88, "learning_rate": 1.6600874353893237e-07, "loss": 0.0048, "step": 61363 }, { "epoch": 18.88, "learning_rate": 1.659183418217991e-07, "loss": 0.005, "step": 61364 }, { "epoch": 18.88, "learning_rate": 1.6582796452003892e-07, "loss": 0.0018, "step": 61365 }, { "epoch": 18.88, "learning_rate": 1.6573761163387604e-07, "loss": 0.004, "step": 61366 }, { "epoch": 18.88, "learning_rate": 1.6564728316353474e-07, "loss": 0.0041, "step": 61367 }, { "epoch": 18.88, "learning_rate": 1.655569791092415e-07, "loss": 0.0042, "step": 61368 }, { "epoch": 18.88, "learning_rate": 1.654666994712184e-07, "loss": 0.0018, "step": 61369 }, { "epoch": 18.88, "learning_rate": 1.6537644424968968e-07, "loss": 0.0048, "step": 61370 }, { "epoch": 18.88, "learning_rate": 1.6528621344488181e-07, "loss": 0.0027, "step": 61371 }, { "epoch": 18.88, "learning_rate": 1.6519600705701465e-07, "loss": 0.0048, "step": 61372 }, { "epoch": 18.88, "learning_rate": 1.6510582508631467e-07, "loss": 0.0038, "step": 61373 }, { "epoch": 18.88, "learning_rate": 1.6501566753300613e-07, "loss": 0.0026, "step": 61374 }, { "epoch": 18.88, "learning_rate": 1.649255343973122e-07, "loss": 0.004, "step": 61375 }, { "epoch": 18.88, "learning_rate": 1.648354256794571e-07, "loss": 0.0041, "step": 61376 }, { "epoch": 18.88, "learning_rate": 1.6474534137966402e-07, "loss": 0.0036, "step": 61377 }, { "epoch": 18.89, "learning_rate": 1.6465528149815612e-07, "loss": 0.0037, "step": 61378 }, { "epoch": 18.89, "learning_rate": 1.6456524603515879e-07, "loss": 0.0042, "step": 61379 }, { "epoch": 18.89, "learning_rate": 1.6447523499089403e-07, "loss": 0.0028, "step": 61380 }, { "epoch": 18.89, "learning_rate": 1.64385248365585e-07, "loss": 0.0032, "step": 61381 }, { "epoch": 18.89, "learning_rate": 1.642952861594549e-07, "loss": 0.0053, "step": 61382 }, { "epoch": 18.89, "learning_rate": 1.6420534837272907e-07, "loss": 0.0032, "step": 61383 }, { "epoch": 18.89, "learning_rate": 1.6411543500563065e-07, "loss": 0.0035, "step": 61384 }, { "epoch": 18.89, "learning_rate": 1.6402554605838173e-07, "loss": 0.0031, "step": 61385 }, { "epoch": 18.89, "learning_rate": 1.6393568153120544e-07, "loss": 0.0045, "step": 61386 }, { "epoch": 18.89, "learning_rate": 1.638458414243238e-07, "loss": 0.0036, "step": 61387 }, { "epoch": 18.89, "learning_rate": 1.6375602573796222e-07, "loss": 0.0028, "step": 61388 }, { "epoch": 18.89, "learning_rate": 1.6366623447234277e-07, "loss": 0.0032, "step": 61389 }, { "epoch": 18.89, "learning_rate": 1.6357646762768742e-07, "loss": 0.0036, "step": 61390 }, { "epoch": 18.89, "learning_rate": 1.634867252042216e-07, "loss": 0.0035, "step": 61391 }, { "epoch": 18.89, "learning_rate": 1.633970072021651e-07, "loss": 0.0029, "step": 61392 }, { "epoch": 18.89, "learning_rate": 1.6330731362174224e-07, "loss": 0.0041, "step": 61393 }, { "epoch": 18.89, "learning_rate": 1.63217644463175e-07, "loss": 0.0029, "step": 61394 }, { "epoch": 18.89, "learning_rate": 1.6312799972668659e-07, "loss": 0.002, "step": 61395 }, { "epoch": 18.89, "learning_rate": 1.630383794125001e-07, "loss": 0.0083, "step": 61396 }, { "epoch": 18.89, "learning_rate": 1.6294878352083653e-07, "loss": 0.0045, "step": 61397 }, { "epoch": 18.89, "learning_rate": 1.6285921205192012e-07, "loss": 0.0036, "step": 61398 }, { "epoch": 18.89, "learning_rate": 1.627696650059718e-07, "loss": 0.0048, "step": 61399 }, { "epoch": 18.89, "learning_rate": 1.6268014238321474e-07, "loss": 0.0017, "step": 61400 }, { "epoch": 18.89, "learning_rate": 1.6259064418386982e-07, "loss": 0.0056, "step": 61401 }, { "epoch": 18.89, "learning_rate": 1.6250117040816026e-07, "loss": 0.003, "step": 61402 }, { "epoch": 18.89, "learning_rate": 1.6241172105630808e-07, "loss": 0.0023, "step": 61403 }, { "epoch": 18.89, "learning_rate": 1.6232229612853646e-07, "loss": 0.004, "step": 61404 }, { "epoch": 18.89, "learning_rate": 1.622328956250674e-07, "loss": 0.0053, "step": 61405 }, { "epoch": 18.89, "learning_rate": 1.6214351954612074e-07, "loss": 0.004, "step": 61406 }, { "epoch": 18.89, "learning_rate": 1.6205416789191852e-07, "loss": 0.0051, "step": 61407 }, { "epoch": 18.89, "learning_rate": 1.619648406626839e-07, "loss": 0.0034, "step": 61408 }, { "epoch": 18.89, "learning_rate": 1.6187553785863897e-07, "loss": 0.0042, "step": 61409 }, { "epoch": 18.9, "learning_rate": 1.617862594800046e-07, "loss": 0.003, "step": 61410 }, { "epoch": 18.9, "learning_rate": 1.6169700552700284e-07, "loss": 0.0045, "step": 61411 }, { "epoch": 18.9, "learning_rate": 1.6160777599985466e-07, "loss": 0.0037, "step": 61412 }, { "epoch": 18.9, "learning_rate": 1.615185708987821e-07, "loss": 0.0046, "step": 61413 }, { "epoch": 18.9, "learning_rate": 1.6142939022400717e-07, "loss": 0.0028, "step": 61414 }, { "epoch": 18.9, "learning_rate": 1.6134023397575084e-07, "loss": 0.0034, "step": 61415 }, { "epoch": 18.9, "learning_rate": 1.6125110215423289e-07, "loss": 0.0026, "step": 61416 }, { "epoch": 18.9, "learning_rate": 1.6116199475967654e-07, "loss": 0.0032, "step": 61417 }, { "epoch": 18.9, "learning_rate": 1.6107291179230268e-07, "loss": 0.0029, "step": 61418 }, { "epoch": 18.9, "learning_rate": 1.6098385325233334e-07, "loss": 0.0027, "step": 61419 }, { "epoch": 18.9, "learning_rate": 1.6089481913998727e-07, "loss": 0.0028, "step": 61420 }, { "epoch": 18.9, "learning_rate": 1.6080580945548764e-07, "loss": 0.0033, "step": 61421 }, { "epoch": 18.9, "learning_rate": 1.607168241990531e-07, "loss": 0.0088, "step": 61422 }, { "epoch": 18.9, "learning_rate": 1.6062786337090687e-07, "loss": 0.0035, "step": 61423 }, { "epoch": 18.9, "learning_rate": 1.6053892697126984e-07, "loss": 0.0049, "step": 61424 }, { "epoch": 18.9, "learning_rate": 1.6045001500036072e-07, "loss": 0.0049, "step": 61425 }, { "epoch": 18.9, "learning_rate": 1.6036112745840383e-07, "loss": 0.0029, "step": 61426 }, { "epoch": 18.9, "learning_rate": 1.602722643456156e-07, "loss": 0.0029, "step": 61427 }, { "epoch": 18.9, "learning_rate": 1.6018342566221923e-07, "loss": 0.0032, "step": 61428 }, { "epoch": 18.9, "learning_rate": 1.6009461140843452e-07, "loss": 0.0042, "step": 61429 }, { "epoch": 18.9, "learning_rate": 1.600058215844813e-07, "loss": 0.0049, "step": 61430 }, { "epoch": 18.9, "learning_rate": 1.5991705619058163e-07, "loss": 0.0019, "step": 61431 }, { "epoch": 18.9, "learning_rate": 1.598283152269553e-07, "loss": 0.0034, "step": 61432 }, { "epoch": 18.9, "learning_rate": 1.5973959869382327e-07, "loss": 0.0034, "step": 61433 }, { "epoch": 18.9, "learning_rate": 1.5965090659140315e-07, "loss": 0.0025, "step": 61434 }, { "epoch": 18.9, "learning_rate": 1.5956223891991807e-07, "loss": 0.0049, "step": 61435 }, { "epoch": 18.9, "learning_rate": 1.5947359567958677e-07, "loss": 0.0068, "step": 61436 }, { "epoch": 18.9, "learning_rate": 1.5938497687062905e-07, "loss": 0.0027, "step": 61437 }, { "epoch": 18.9, "learning_rate": 1.5929638249326583e-07, "loss": 0.0031, "step": 61438 }, { "epoch": 18.9, "learning_rate": 1.59207812547717e-07, "loss": 0.0032, "step": 61439 }, { "epoch": 18.9, "learning_rate": 1.5911926703420343e-07, "loss": 0.0041, "step": 61440 }, { "epoch": 18.9, "learning_rate": 1.5903074595294165e-07, "loss": 0.0032, "step": 61441 }, { "epoch": 18.9, "learning_rate": 1.589422493041537e-07, "loss": 0.0058, "step": 61442 }, { "epoch": 18.91, "learning_rate": 1.588537770880605e-07, "loss": 0.0009, "step": 61443 }, { "epoch": 18.91, "learning_rate": 1.5876532930487853e-07, "loss": 0.0029, "step": 61444 }, { "epoch": 18.91, "learning_rate": 1.5867690595482987e-07, "loss": 0.0027, "step": 61445 }, { "epoch": 18.91, "learning_rate": 1.5858850703813433e-07, "loss": 0.003, "step": 61446 }, { "epoch": 18.91, "learning_rate": 1.5850013255500952e-07, "loss": 0.005, "step": 61447 }, { "epoch": 18.91, "learning_rate": 1.5841178250567523e-07, "loss": 0.0034, "step": 61448 }, { "epoch": 18.91, "learning_rate": 1.5832345689035134e-07, "loss": 0.0046, "step": 61449 }, { "epoch": 18.91, "learning_rate": 1.5823515570925763e-07, "loss": 0.0028, "step": 61450 }, { "epoch": 18.91, "learning_rate": 1.5814687896261282e-07, "loss": 0.004, "step": 61451 }, { "epoch": 18.91, "learning_rate": 1.5805862665063453e-07, "loss": 0.0019, "step": 61452 }, { "epoch": 18.91, "learning_rate": 1.5797039877354592e-07, "loss": 0.0034, "step": 61453 }, { "epoch": 18.91, "learning_rate": 1.5788219533156234e-07, "loss": 0.0035, "step": 61454 }, { "epoch": 18.91, "learning_rate": 1.5779401632490365e-07, "loss": 0.0055, "step": 61455 }, { "epoch": 18.91, "learning_rate": 1.5770586175378965e-07, "loss": 0.003, "step": 61456 }, { "epoch": 18.91, "learning_rate": 1.5761773161843687e-07, "loss": 0.0033, "step": 61457 }, { "epoch": 18.91, "learning_rate": 1.5752962591906727e-07, "loss": 0.0046, "step": 61458 }, { "epoch": 18.91, "learning_rate": 1.5744154465589745e-07, "loss": 0.005, "step": 61459 }, { "epoch": 18.91, "learning_rate": 1.5735348782914717e-07, "loss": 0.0037, "step": 61460 }, { "epoch": 18.91, "learning_rate": 1.5726545543903627e-07, "loss": 0.0022, "step": 61461 }, { "epoch": 18.91, "learning_rate": 1.5717744748577902e-07, "loss": 0.0028, "step": 61462 }, { "epoch": 18.91, "learning_rate": 1.5708946396959856e-07, "loss": 0.0052, "step": 61463 }, { "epoch": 18.91, "learning_rate": 1.570015048907103e-07, "loss": 0.0029, "step": 61464 }, { "epoch": 18.91, "learning_rate": 1.5691357024933407e-07, "loss": 0.0038, "step": 61465 }, { "epoch": 18.91, "learning_rate": 1.5682566004568744e-07, "loss": 0.0041, "step": 61466 }, { "epoch": 18.91, "learning_rate": 1.5673777427999027e-07, "loss": 0.0036, "step": 61467 }, { "epoch": 18.91, "learning_rate": 1.5664991295245902e-07, "loss": 0.0036, "step": 61468 }, { "epoch": 18.91, "learning_rate": 1.565620760633113e-07, "loss": 0.0057, "step": 61469 }, { "epoch": 18.91, "learning_rate": 1.5647426361276696e-07, "loss": 0.0017, "step": 61470 }, { "epoch": 18.91, "learning_rate": 1.5638647560104358e-07, "loss": 0.0033, "step": 61471 }, { "epoch": 18.91, "learning_rate": 1.5629871202835766e-07, "loss": 0.0037, "step": 61472 }, { "epoch": 18.91, "learning_rate": 1.5621097289493015e-07, "loss": 0.004, "step": 61473 }, { "epoch": 18.91, "learning_rate": 1.561232582009764e-07, "loss": 0.0046, "step": 61474 }, { "epoch": 18.92, "learning_rate": 1.5603556794671403e-07, "loss": 0.0053, "step": 61475 }, { "epoch": 18.92, "learning_rate": 1.5594790213236288e-07, "loss": 0.0025, "step": 61476 }, { "epoch": 18.92, "learning_rate": 1.558602607581372e-07, "loss": 0.0028, "step": 61477 }, { "epoch": 18.92, "learning_rate": 1.5577264382425684e-07, "loss": 0.0018, "step": 61478 }, { "epoch": 18.92, "learning_rate": 1.5568505133094046e-07, "loss": 0.0043, "step": 61479 }, { "epoch": 18.92, "learning_rate": 1.5559748327840352e-07, "loss": 0.0145, "step": 61480 }, { "epoch": 18.92, "learning_rate": 1.5550993966686468e-07, "loss": 0.0049, "step": 61481 }, { "epoch": 18.92, "learning_rate": 1.5542242049653933e-07, "loss": 0.0049, "step": 61482 }, { "epoch": 18.92, "learning_rate": 1.553349257676473e-07, "loss": 0.0035, "step": 61483 }, { "epoch": 18.92, "learning_rate": 1.55247455480404e-07, "loss": 0.004, "step": 61484 }, { "epoch": 18.92, "learning_rate": 1.5516000963502698e-07, "loss": 0.0026, "step": 61485 }, { "epoch": 18.92, "learning_rate": 1.5507258823173388e-07, "loss": 0.0026, "step": 61486 }, { "epoch": 18.92, "learning_rate": 1.549851912707423e-07, "loss": 0.0022, "step": 61487 }, { "epoch": 18.92, "learning_rate": 1.548978187522676e-07, "loss": 0.0034, "step": 61488 }, { "epoch": 18.92, "learning_rate": 1.5481047067652744e-07, "loss": 0.0032, "step": 61489 }, { "epoch": 18.92, "learning_rate": 1.5472314704373825e-07, "loss": 0.0037, "step": 61490 }, { "epoch": 18.92, "learning_rate": 1.5463584785411767e-07, "loss": 0.0043, "step": 61491 }, { "epoch": 18.92, "learning_rate": 1.5454857310788108e-07, "loss": 0.0042, "step": 61492 }, { "epoch": 18.92, "learning_rate": 1.5446132280524717e-07, "loss": 0.003, "step": 61493 }, { "epoch": 18.92, "learning_rate": 1.5437409694643246e-07, "loss": 0.0034, "step": 61494 }, { "epoch": 18.92, "learning_rate": 1.5428689553165233e-07, "loss": 0.0036, "step": 61495 }, { "epoch": 18.92, "learning_rate": 1.5419971856112214e-07, "loss": 0.0036, "step": 61496 }, { "epoch": 18.92, "learning_rate": 1.5411256603505953e-07, "loss": 0.0036, "step": 61497 }, { "epoch": 18.92, "learning_rate": 1.5402543795368207e-07, "loss": 0.003, "step": 61498 }, { "epoch": 18.92, "learning_rate": 1.5393833431720516e-07, "loss": 0.0036, "step": 61499 }, { "epoch": 18.92, "learning_rate": 1.538512551258442e-07, "loss": 0.0019, "step": 61500 }, { "epoch": 18.92, "learning_rate": 1.5376420037981676e-07, "loss": 0.0019, "step": 61501 }, { "epoch": 18.92, "learning_rate": 1.5367717007933826e-07, "loss": 0.0031, "step": 61502 }, { "epoch": 18.92, "learning_rate": 1.5359016422462514e-07, "loss": 0.0038, "step": 61503 }, { "epoch": 18.92, "learning_rate": 1.5350318281589283e-07, "loss": 0.0039, "step": 61504 }, { "epoch": 18.92, "learning_rate": 1.5341622585335668e-07, "loss": 0.0024, "step": 61505 }, { "epoch": 18.92, "learning_rate": 1.5332929333723435e-07, "loss": 0.0027, "step": 61506 }, { "epoch": 18.92, "learning_rate": 1.5324238526774006e-07, "loss": 0.0031, "step": 61507 }, { "epoch": 18.93, "learning_rate": 1.5315550164509253e-07, "loss": 0.0033, "step": 61508 }, { "epoch": 18.93, "learning_rate": 1.530686424695038e-07, "loss": 0.002, "step": 61509 }, { "epoch": 18.93, "learning_rate": 1.5298180774119044e-07, "loss": 0.0055, "step": 61510 }, { "epoch": 18.93, "learning_rate": 1.5289499746036885e-07, "loss": 0.0039, "step": 61511 }, { "epoch": 18.93, "learning_rate": 1.5280821162725445e-07, "loss": 0.0041, "step": 61512 }, { "epoch": 18.93, "learning_rate": 1.5272145024206264e-07, "loss": 0.0041, "step": 61513 }, { "epoch": 18.93, "learning_rate": 1.526347133050088e-07, "loss": 0.0028, "step": 61514 }, { "epoch": 18.93, "learning_rate": 1.5254800081630828e-07, "loss": 0.0049, "step": 61515 }, { "epoch": 18.93, "learning_rate": 1.524613127761765e-07, "loss": 0.0033, "step": 61516 }, { "epoch": 18.93, "learning_rate": 1.5237464918482768e-07, "loss": 0.0028, "step": 61517 }, { "epoch": 18.93, "learning_rate": 1.5228801004247728e-07, "loss": 0.0032, "step": 61518 }, { "epoch": 18.93, "learning_rate": 1.5220139534934175e-07, "loss": 0.0049, "step": 61519 }, { "epoch": 18.93, "learning_rate": 1.5211480510563537e-07, "loss": 0.0025, "step": 61520 }, { "epoch": 18.93, "learning_rate": 1.520282393115735e-07, "loss": 0.004, "step": 61521 }, { "epoch": 18.93, "learning_rate": 1.5194169796736934e-07, "loss": 0.0035, "step": 61522 }, { "epoch": 18.93, "learning_rate": 1.5185518107324048e-07, "loss": 0.0034, "step": 61523 }, { "epoch": 18.93, "learning_rate": 1.5176868862939897e-07, "loss": 0.0024, "step": 61524 }, { "epoch": 18.93, "learning_rate": 1.516822206360613e-07, "loss": 0.0026, "step": 61525 }, { "epoch": 18.93, "learning_rate": 1.5159577709344175e-07, "loss": 0.0042, "step": 61526 }, { "epoch": 18.93, "learning_rate": 1.5150935800175348e-07, "loss": 0.0022, "step": 61527 }, { "epoch": 18.93, "learning_rate": 1.514229633612141e-07, "loss": 0.0043, "step": 61528 }, { "epoch": 18.93, "learning_rate": 1.5133659317203674e-07, "loss": 0.0024, "step": 61529 }, { "epoch": 18.93, "learning_rate": 1.5125024743443463e-07, "loss": 0.0039, "step": 61530 }, { "epoch": 18.93, "learning_rate": 1.5116392614862309e-07, "loss": 0.0033, "step": 61531 }, { "epoch": 18.93, "learning_rate": 1.5107762931481528e-07, "loss": 0.0028, "step": 61532 }, { "epoch": 18.93, "learning_rate": 1.5099135693322776e-07, "loss": 0.0038, "step": 61533 }, { "epoch": 18.93, "learning_rate": 1.5090510900407363e-07, "loss": 0.0061, "step": 61534 }, { "epoch": 18.93, "learning_rate": 1.5081888552756608e-07, "loss": 0.004, "step": 61535 }, { "epoch": 18.93, "learning_rate": 1.5073268650392047e-07, "loss": 0.0051, "step": 61536 }, { "epoch": 18.93, "learning_rate": 1.5064651193335112e-07, "loss": 0.0022, "step": 61537 }, { "epoch": 18.93, "learning_rate": 1.5056036181607003e-07, "loss": 0.0034, "step": 61538 }, { "epoch": 18.93, "learning_rate": 1.5047423615229262e-07, "loss": 0.0047, "step": 61539 }, { "epoch": 18.94, "learning_rate": 1.5038813494223203e-07, "loss": 0.0044, "step": 61540 }, { "epoch": 18.94, "learning_rate": 1.5030205818610255e-07, "loss": 0.0052, "step": 61541 }, { "epoch": 18.94, "learning_rate": 1.5021600588411734e-07, "loss": 0.0036, "step": 61542 }, { "epoch": 18.94, "learning_rate": 1.5012997803649178e-07, "loss": 0.0031, "step": 61543 }, { "epoch": 18.94, "learning_rate": 1.5004397464343677e-07, "loss": 0.0012, "step": 61544 }, { "epoch": 18.94, "learning_rate": 1.4995799570516666e-07, "loss": 0.0038, "step": 61545 }, { "epoch": 18.94, "learning_rate": 1.498720412218968e-07, "loss": 0.0037, "step": 61546 }, { "epoch": 18.94, "learning_rate": 1.4978611119383703e-07, "loss": 0.0038, "step": 61547 }, { "epoch": 18.94, "learning_rate": 1.497002056212038e-07, "loss": 0.0034, "step": 61548 }, { "epoch": 18.94, "learning_rate": 1.4961432450421033e-07, "loss": 0.0045, "step": 61549 }, { "epoch": 18.94, "learning_rate": 1.4952846784306864e-07, "loss": 0.0028, "step": 61550 }, { "epoch": 18.94, "learning_rate": 1.4944263563799188e-07, "loss": 0.0019, "step": 61551 }, { "epoch": 18.94, "learning_rate": 1.4935682788919326e-07, "loss": 0.0044, "step": 61552 }, { "epoch": 18.94, "learning_rate": 1.492710445968859e-07, "loss": 0.0026, "step": 61553 }, { "epoch": 18.94, "learning_rate": 1.4918528576128298e-07, "loss": 0.0031, "step": 61554 }, { "epoch": 18.94, "learning_rate": 1.4909955138259768e-07, "loss": 0.0029, "step": 61555 }, { "epoch": 18.94, "learning_rate": 1.4901384146104204e-07, "loss": 0.0043, "step": 61556 }, { "epoch": 18.94, "learning_rate": 1.489281559968292e-07, "loss": 0.0037, "step": 61557 }, { "epoch": 18.94, "learning_rate": 1.4884249499017233e-07, "loss": 0.0027, "step": 61558 }, { "epoch": 18.94, "learning_rate": 1.4875685844128352e-07, "loss": 0.0037, "step": 61559 }, { "epoch": 18.94, "learning_rate": 1.4867124635037588e-07, "loss": 0.0051, "step": 61560 }, { "epoch": 18.94, "learning_rate": 1.485856587176615e-07, "loss": 0.0027, "step": 61561 }, { "epoch": 18.94, "learning_rate": 1.4850009554335243e-07, "loss": 0.0029, "step": 61562 }, { "epoch": 18.94, "learning_rate": 1.4841455682766292e-07, "loss": 0.0044, "step": 61563 }, { "epoch": 18.94, "learning_rate": 1.4832904257080393e-07, "loss": 0.0029, "step": 61564 }, { "epoch": 18.94, "learning_rate": 1.4824355277298753e-07, "loss": 0.0031, "step": 61565 }, { "epoch": 18.94, "learning_rate": 1.4815808743442683e-07, "loss": 0.0046, "step": 61566 }, { "epoch": 18.94, "learning_rate": 1.4807264655533282e-07, "loss": 0.0028, "step": 61567 }, { "epoch": 18.94, "learning_rate": 1.4798723013591975e-07, "loss": 0.0033, "step": 61568 }, { "epoch": 18.94, "learning_rate": 1.4790183817639746e-07, "loss": 0.0067, "step": 61569 }, { "epoch": 18.94, "learning_rate": 1.4781647067697913e-07, "loss": 0.0043, "step": 61570 }, { "epoch": 18.94, "learning_rate": 1.477311276378768e-07, "loss": 0.0033, "step": 61571 }, { "epoch": 18.94, "learning_rate": 1.4764580905930138e-07, "loss": 0.006, "step": 61572 }, { "epoch": 18.95, "learning_rate": 1.4756051494146496e-07, "loss": 0.0052, "step": 61573 }, { "epoch": 18.95, "learning_rate": 1.4747524528458068e-07, "loss": 0.004, "step": 61574 }, { "epoch": 18.95, "learning_rate": 1.4739000008885839e-07, "loss": 0.0053, "step": 61575 }, { "epoch": 18.95, "learning_rate": 1.4730477935451127e-07, "loss": 0.0031, "step": 61576 }, { "epoch": 18.95, "learning_rate": 1.472195830817491e-07, "loss": 0.0077, "step": 61577 }, { "epoch": 18.95, "learning_rate": 1.4713441127078509e-07, "loss": 0.0038, "step": 61578 }, { "epoch": 18.95, "learning_rate": 1.4704926392183017e-07, "loss": 0.0036, "step": 61579 }, { "epoch": 18.95, "learning_rate": 1.469641410350964e-07, "loss": 0.0043, "step": 61580 }, { "epoch": 18.95, "learning_rate": 1.4687904261079355e-07, "loss": 0.0044, "step": 61581 }, { "epoch": 18.95, "learning_rate": 1.4679396864913265e-07, "loss": 0.0045, "step": 61582 }, { "epoch": 18.95, "learning_rate": 1.4670891915032682e-07, "loss": 0.0042, "step": 61583 }, { "epoch": 18.95, "learning_rate": 1.4662389411458812e-07, "loss": 0.007, "step": 61584 }, { "epoch": 18.95, "learning_rate": 1.4653889354212414e-07, "loss": 0.0046, "step": 61585 }, { "epoch": 18.95, "learning_rate": 1.4645391743314807e-07, "loss": 0.0024, "step": 61586 }, { "epoch": 18.95, "learning_rate": 1.4636896578786974e-07, "loss": 0.0059, "step": 61587 }, { "epoch": 18.95, "learning_rate": 1.462840386065012e-07, "loss": 0.0034, "step": 61588 }, { "epoch": 18.95, "learning_rate": 1.461991358892534e-07, "loss": 0.0048, "step": 61589 }, { "epoch": 18.95, "learning_rate": 1.4611425763633614e-07, "loss": 0.0061, "step": 61590 }, { "epoch": 18.95, "learning_rate": 1.460294038479615e-07, "loss": 0.0037, "step": 61591 }, { "epoch": 18.95, "learning_rate": 1.459445745243382e-07, "loss": 0.0031, "step": 61592 }, { "epoch": 18.95, "learning_rate": 1.4585976966567826e-07, "loss": 0.0029, "step": 61593 }, { "epoch": 18.95, "learning_rate": 1.457749892721927e-07, "loss": 0.0032, "step": 61594 }, { "epoch": 18.95, "learning_rate": 1.4569023334409128e-07, "loss": 0.0044, "step": 61595 }, { "epoch": 18.95, "learning_rate": 1.4560550188158274e-07, "loss": 0.0026, "step": 61596 }, { "epoch": 18.95, "learning_rate": 1.4552079488488025e-07, "loss": 0.0024, "step": 61597 }, { "epoch": 18.95, "learning_rate": 1.4543611235419364e-07, "loss": 0.0035, "step": 61598 }, { "epoch": 18.95, "learning_rate": 1.4535145428973164e-07, "loss": 0.0049, "step": 61599 }, { "epoch": 18.95, "learning_rate": 1.4526682069170517e-07, "loss": 0.004, "step": 61600 }, { "epoch": 18.95, "learning_rate": 1.4518221156032408e-07, "loss": 0.0034, "step": 61601 }, { "epoch": 18.95, "learning_rate": 1.450976268957982e-07, "loss": 0.0032, "step": 61602 }, { "epoch": 18.95, "learning_rate": 1.450130666983396e-07, "loss": 0.006, "step": 61603 }, { "epoch": 18.95, "learning_rate": 1.4492853096815586e-07, "loss": 0.0025, "step": 61604 }, { "epoch": 18.96, "learning_rate": 1.4484401970545903e-07, "loss": 0.0044, "step": 61605 }, { "epoch": 18.96, "learning_rate": 1.4475953291045674e-07, "loss": 0.005, "step": 61606 }, { "epoch": 18.96, "learning_rate": 1.4467507058335882e-07, "loss": 0.005, "step": 61607 }, { "epoch": 18.96, "learning_rate": 1.4459063272437624e-07, "loss": 0.0048, "step": 61608 }, { "epoch": 18.96, "learning_rate": 1.445062193337188e-07, "loss": 0.0034, "step": 61609 }, { "epoch": 18.96, "learning_rate": 1.4442183041159519e-07, "loss": 0.005, "step": 61610 }, { "epoch": 18.96, "learning_rate": 1.4433746595821528e-07, "loss": 0.0041, "step": 61611 }, { "epoch": 18.96, "learning_rate": 1.4425312597378782e-07, "loss": 0.0062, "step": 61612 }, { "epoch": 18.96, "learning_rate": 1.441688104585237e-07, "loss": 0.003, "step": 61613 }, { "epoch": 18.96, "learning_rate": 1.440845194126317e-07, "loss": 0.0037, "step": 61614 }, { "epoch": 18.96, "learning_rate": 1.4400025283631936e-07, "loss": 0.0028, "step": 61615 }, { "epoch": 18.96, "learning_rate": 1.4391601072979767e-07, "loss": 0.0045, "step": 61616 }, { "epoch": 18.96, "learning_rate": 1.4383179309327533e-07, "loss": 0.0031, "step": 61617 }, { "epoch": 18.96, "learning_rate": 1.4374759992696218e-07, "loss": 0.002, "step": 61618 }, { "epoch": 18.96, "learning_rate": 1.4366343123106697e-07, "loss": 0.0038, "step": 61619 }, { "epoch": 18.96, "learning_rate": 1.4357928700579725e-07, "loss": 0.0045, "step": 61620 }, { "epoch": 18.96, "learning_rate": 1.4349516725136404e-07, "loss": 0.0018, "step": 61621 }, { "epoch": 18.96, "learning_rate": 1.4341107196797377e-07, "loss": 0.0023, "step": 61622 }, { "epoch": 18.96, "learning_rate": 1.433270011558363e-07, "loss": 0.0024, "step": 61623 }, { "epoch": 18.96, "learning_rate": 1.4324295481516148e-07, "loss": 0.0055, "step": 61624 }, { "epoch": 18.96, "learning_rate": 1.4315893294615802e-07, "loss": 0.0052, "step": 61625 }, { "epoch": 18.96, "learning_rate": 1.430749355490335e-07, "loss": 0.0052, "step": 61626 }, { "epoch": 18.96, "learning_rate": 1.429909626239945e-07, "loss": 0.0044, "step": 61627 }, { "epoch": 18.96, "learning_rate": 1.4290701417125296e-07, "loss": 0.0034, "step": 61628 }, { "epoch": 18.96, "learning_rate": 1.4282309019101548e-07, "loss": 0.0034, "step": 61629 }, { "epoch": 18.96, "learning_rate": 1.4273919068349184e-07, "loss": 0.0033, "step": 61630 }, { "epoch": 18.96, "learning_rate": 1.4265531564888858e-07, "loss": 0.0021, "step": 61631 }, { "epoch": 18.96, "learning_rate": 1.4257146508741436e-07, "loss": 0.0032, "step": 61632 }, { "epoch": 18.96, "learning_rate": 1.4248763899927797e-07, "loss": 0.0031, "step": 61633 }, { "epoch": 18.96, "learning_rate": 1.424038373846881e-07, "loss": 0.0038, "step": 61634 }, { "epoch": 18.96, "learning_rate": 1.4232006024385126e-07, "loss": 0.0036, "step": 61635 }, { "epoch": 18.96, "learning_rate": 1.4223630757697614e-07, "loss": 0.0042, "step": 61636 }, { "epoch": 18.96, "learning_rate": 1.4215257938427042e-07, "loss": 0.0043, "step": 61637 }, { "epoch": 18.97, "learning_rate": 1.4206887566594164e-07, "loss": 0.0033, "step": 61638 }, { "epoch": 18.97, "learning_rate": 1.4198519642219966e-07, "loss": 0.0096, "step": 61639 }, { "epoch": 18.97, "learning_rate": 1.41901541653251e-07, "loss": 0.0032, "step": 61640 }, { "epoch": 18.97, "learning_rate": 1.418179113593021e-07, "loss": 0.0047, "step": 61641 }, { "epoch": 18.97, "learning_rate": 1.4173430554056177e-07, "loss": 0.0062, "step": 61642 }, { "epoch": 18.97, "learning_rate": 1.4165072419723758e-07, "loss": 0.0034, "step": 61643 }, { "epoch": 18.97, "learning_rate": 1.4156716732953712e-07, "loss": 0.0028, "step": 61644 }, { "epoch": 18.97, "learning_rate": 1.4148363493766803e-07, "loss": 0.0046, "step": 61645 }, { "epoch": 18.97, "learning_rate": 1.4140012702183682e-07, "loss": 0.0055, "step": 61646 }, { "epoch": 18.97, "learning_rate": 1.4131664358225105e-07, "loss": 0.0067, "step": 61647 }, { "epoch": 18.97, "learning_rate": 1.412331846191184e-07, "loss": 0.004, "step": 61648 }, { "epoch": 18.97, "learning_rate": 1.4114975013264643e-07, "loss": 0.0043, "step": 61649 }, { "epoch": 18.97, "learning_rate": 1.4106634012304166e-07, "loss": 0.0054, "step": 61650 }, { "epoch": 18.97, "learning_rate": 1.4098295459051059e-07, "loss": 0.0059, "step": 61651 }, { "epoch": 18.97, "learning_rate": 1.4089959353526085e-07, "loss": 0.0038, "step": 61652 }, { "epoch": 18.97, "learning_rate": 1.408162569575e-07, "loss": 0.0028, "step": 61653 }, { "epoch": 18.97, "learning_rate": 1.407329448574346e-07, "loss": 0.0022, "step": 61654 }, { "epoch": 18.97, "learning_rate": 1.4064965723527114e-07, "loss": 0.0028, "step": 61655 }, { "epoch": 18.97, "learning_rate": 1.405663940912161e-07, "loss": 0.0027, "step": 61656 }, { "epoch": 18.97, "learning_rate": 1.4048315542547707e-07, "loss": 0.0045, "step": 61657 }, { "epoch": 18.97, "learning_rate": 1.403999412382606e-07, "loss": 0.0045, "step": 61658 }, { "epoch": 18.97, "learning_rate": 1.403167515297721e-07, "loss": 0.0041, "step": 61659 }, { "epoch": 18.97, "learning_rate": 1.402335863002202e-07, "loss": 0.0022, "step": 61660 }, { "epoch": 18.97, "learning_rate": 1.4015044554981038e-07, "loss": 0.0032, "step": 61661 }, { "epoch": 18.97, "learning_rate": 1.4006732927874688e-07, "loss": 0.0023, "step": 61662 }, { "epoch": 18.97, "learning_rate": 1.3998423748723956e-07, "loss": 0.0043, "step": 61663 }, { "epoch": 18.97, "learning_rate": 1.3990117017549375e-07, "loss": 0.0021, "step": 61664 }, { "epoch": 18.97, "learning_rate": 1.398181273437138e-07, "loss": 0.0034, "step": 61665 }, { "epoch": 18.97, "learning_rate": 1.397351089921084e-07, "loss": 0.0023, "step": 61666 }, { "epoch": 18.97, "learning_rate": 1.3965211512088182e-07, "loss": 0.005, "step": 61667 }, { "epoch": 18.97, "learning_rate": 1.395691457302406e-07, "loss": 0.0021, "step": 61668 }, { "epoch": 18.97, "learning_rate": 1.3948620082039121e-07, "loss": 0.0058, "step": 61669 }, { "epoch": 18.98, "learning_rate": 1.3940328039153905e-07, "loss": 0.004, "step": 61670 }, { "epoch": 18.98, "learning_rate": 1.3932038444389063e-07, "loss": 0.0032, "step": 61671 }, { "epoch": 18.98, "learning_rate": 1.3923751297765021e-07, "loss": 0.0039, "step": 61672 }, { "epoch": 18.98, "learning_rate": 1.391546659930254e-07, "loss": 0.0055, "step": 61673 }, { "epoch": 18.98, "learning_rate": 1.3907184349022162e-07, "loss": 0.0022, "step": 61674 }, { "epoch": 18.98, "learning_rate": 1.3898904546944425e-07, "loss": 0.0034, "step": 61675 }, { "epoch": 18.98, "learning_rate": 1.3890627193089756e-07, "loss": 0.0031, "step": 61676 }, { "epoch": 18.98, "learning_rate": 1.3882352287478806e-07, "loss": 0.0027, "step": 61677 }, { "epoch": 18.98, "learning_rate": 1.3874079830132225e-07, "loss": 0.0027, "step": 61678 }, { "epoch": 18.98, "learning_rate": 1.3865809821070332e-07, "loss": 0.0045, "step": 61679 }, { "epoch": 18.98, "learning_rate": 1.3857542260313884e-07, "loss": 0.0032, "step": 61680 }, { "epoch": 18.98, "learning_rate": 1.384927714788331e-07, "loss": 0.0038, "step": 61681 }, { "epoch": 18.98, "learning_rate": 1.3841014483799042e-07, "loss": 0.0032, "step": 61682 }, { "epoch": 18.98, "learning_rate": 1.3832754268081726e-07, "loss": 0.0041, "step": 61683 }, { "epoch": 18.98, "learning_rate": 1.3824496500751904e-07, "loss": 0.0017, "step": 61684 }, { "epoch": 18.98, "learning_rate": 1.3816241181829891e-07, "loss": 0.0038, "step": 61685 }, { "epoch": 18.98, "learning_rate": 1.3807988311336229e-07, "loss": 0.0018, "step": 61686 }, { "epoch": 18.98, "learning_rate": 1.3799737889291565e-07, "loss": 0.0032, "step": 61687 }, { "epoch": 18.98, "learning_rate": 1.3791489915716215e-07, "loss": 0.0051, "step": 61688 }, { "epoch": 18.98, "learning_rate": 1.3783244390630834e-07, "loss": 0.0041, "step": 61689 }, { "epoch": 18.98, "learning_rate": 1.3775001314055624e-07, "loss": 0.0034, "step": 61690 }, { "epoch": 18.98, "learning_rate": 1.376676068601124e-07, "loss": 0.0032, "step": 61691 }, { "epoch": 18.98, "learning_rate": 1.37585225065181e-07, "loss": 0.0037, "step": 61692 }, { "epoch": 18.98, "learning_rate": 1.3750286775596754e-07, "loss": 0.0029, "step": 61693 }, { "epoch": 18.98, "learning_rate": 1.3742053493267515e-07, "loss": 0.0039, "step": 61694 }, { "epoch": 18.98, "learning_rate": 1.3733822659551033e-07, "loss": 0.004, "step": 61695 }, { "epoch": 18.98, "learning_rate": 1.372559427446729e-07, "loss": 0.0029, "step": 61696 }, { "epoch": 18.98, "learning_rate": 1.3717368338037163e-07, "loss": 0.0051, "step": 61697 }, { "epoch": 18.98, "learning_rate": 1.3709144850280853e-07, "loss": 0.0054, "step": 61698 }, { "epoch": 18.98, "learning_rate": 1.3700923811218903e-07, "loss": 0.0031, "step": 61699 }, { "epoch": 18.98, "learning_rate": 1.3692705220871627e-07, "loss": 0.0035, "step": 61700 }, { "epoch": 18.98, "learning_rate": 1.3684489079259565e-07, "loss": 0.0019, "step": 61701 }, { "epoch": 18.98, "learning_rate": 1.3676275386402927e-07, "loss": 0.0028, "step": 61702 }, { "epoch": 18.99, "learning_rate": 1.3668064142322136e-07, "loss": 0.0033, "step": 61703 }, { "epoch": 18.99, "learning_rate": 1.365985534703773e-07, "loss": 0.0035, "step": 61704 }, { "epoch": 18.99, "learning_rate": 1.3651649000569922e-07, "loss": 0.0057, "step": 61705 }, { "epoch": 18.99, "learning_rate": 1.3643445102939134e-07, "loss": 0.0018, "step": 61706 }, { "epoch": 18.99, "learning_rate": 1.3635243654165798e-07, "loss": 0.0027, "step": 61707 }, { "epoch": 18.99, "learning_rate": 1.3627044654270337e-07, "loss": 0.003, "step": 61708 }, { "epoch": 18.99, "learning_rate": 1.3618848103272964e-07, "loss": 0.0021, "step": 61709 }, { "epoch": 18.99, "learning_rate": 1.361065400119399e-07, "loss": 0.0025, "step": 61710 }, { "epoch": 18.99, "learning_rate": 1.3602462348053848e-07, "loss": 0.003, "step": 61711 }, { "epoch": 18.99, "learning_rate": 1.359427314387285e-07, "loss": 0.0035, "step": 61712 }, { "epoch": 18.99, "learning_rate": 1.3586086388671428e-07, "loss": 0.0071, "step": 61713 }, { "epoch": 18.99, "learning_rate": 1.3577902082469786e-07, "loss": 0.0029, "step": 61714 }, { "epoch": 18.99, "learning_rate": 1.3569720225288353e-07, "loss": 0.0045, "step": 61715 }, { "epoch": 18.99, "learning_rate": 1.3561540817147222e-07, "loss": 0.0029, "step": 61716 }, { "epoch": 18.99, "learning_rate": 1.3553363858066937e-07, "loss": 0.0036, "step": 61717 }, { "epoch": 18.99, "learning_rate": 1.3545189348067812e-07, "loss": 0.0048, "step": 61718 }, { "epoch": 18.99, "learning_rate": 1.353701728716994e-07, "loss": 0.0033, "step": 61719 }, { "epoch": 18.99, "learning_rate": 1.3528847675393642e-07, "loss": 0.0026, "step": 61720 }, { "epoch": 18.99, "learning_rate": 1.3520680512759343e-07, "loss": 0.005, "step": 61721 }, { "epoch": 18.99, "learning_rate": 1.3512515799287251e-07, "loss": 0.0034, "step": 61722 }, { "epoch": 18.99, "learning_rate": 1.3504353534997682e-07, "loss": 0.0031, "step": 61723 }, { "epoch": 18.99, "learning_rate": 1.3496193719910732e-07, "loss": 0.0078, "step": 61724 }, { "epoch": 18.99, "learning_rate": 1.348803635404683e-07, "loss": 0.0027, "step": 61725 }, { "epoch": 18.99, "learning_rate": 1.347988143742618e-07, "loss": 0.0043, "step": 61726 }, { "epoch": 18.99, "learning_rate": 1.3471728970068986e-07, "loss": 0.004, "step": 61727 }, { "epoch": 18.99, "learning_rate": 1.346357895199568e-07, "loss": 0.0026, "step": 61728 }, { "epoch": 18.99, "learning_rate": 1.3455431383226247e-07, "loss": 0.0032, "step": 61729 }, { "epoch": 18.99, "learning_rate": 1.3447286263781e-07, "loss": 0.0028, "step": 61730 }, { "epoch": 18.99, "learning_rate": 1.343914359368015e-07, "loss": 0.0025, "step": 61731 }, { "epoch": 18.99, "learning_rate": 1.3431003372944008e-07, "loss": 0.0031, "step": 61732 }, { "epoch": 18.99, "learning_rate": 1.342286560159267e-07, "loss": 0.0051, "step": 61733 }, { "epoch": 18.99, "learning_rate": 1.341473027964635e-07, "loss": 0.0031, "step": 61734 }, { "epoch": 19.0, "learning_rate": 1.3406597407125354e-07, "loss": 0.0037, "step": 61735 }, { "epoch": 19.0, "learning_rate": 1.3398466984049786e-07, "loss": 0.0043, "step": 61736 }, { "epoch": 19.0, "learning_rate": 1.3390339010439847e-07, "loss": 0.0035, "step": 61737 }, { "epoch": 19.0, "learning_rate": 1.3382213486315744e-07, "loss": 0.0072, "step": 61738 }, { "epoch": 19.0, "learning_rate": 1.337409041169757e-07, "loss": 0.0016, "step": 61739 }, { "epoch": 19.0, "learning_rate": 1.3365969786605538e-07, "loss": 0.0017, "step": 61740 }, { "epoch": 19.0, "learning_rate": 1.3357851611059736e-07, "loss": 0.0021, "step": 61741 }, { "epoch": 19.0, "learning_rate": 1.3349735885080595e-07, "loss": 0.0055, "step": 61742 }, { "epoch": 19.0, "learning_rate": 1.3341622608687877e-07, "loss": 0.0023, "step": 61743 }, { "epoch": 19.0, "learning_rate": 1.3333511781902008e-07, "loss": 0.0025, "step": 61744 }, { "epoch": 19.0, "learning_rate": 1.3325403404743088e-07, "loss": 0.0031, "step": 61745 }, { "epoch": 19.0, "learning_rate": 1.3317297477230983e-07, "loss": 0.0032, "step": 61746 }, { "epoch": 19.0, "learning_rate": 1.3309193999386238e-07, "loss": 0.0043, "step": 61747 }, { "epoch": 19.0, "learning_rate": 1.3301092971228724e-07, "loss": 0.0021, "step": 61748 }, { "epoch": 19.0, "learning_rate": 1.3292994392778535e-07, "loss": 0.003, "step": 61749 }, { "epoch": 19.0, "learning_rate": 1.3284898264055878e-07, "loss": 0.0041, "step": 61750 }, { "epoch": 19.0, "learning_rate": 1.3276804585080737e-07, "loss": 0.0036, "step": 61751 }, { "epoch": 19.0, "learning_rate": 1.326871335587332e-07, "loss": 0.0039, "step": 61752 }, { "epoch": 19.0, "learning_rate": 1.3260624576453718e-07, "loss": 0.0025, "step": 61753 }, { "epoch": 19.0, "learning_rate": 1.3252538246841919e-07, "loss": 0.0027, "step": 61754 }, { "epoch": 19.0, "learning_rate": 1.3244454367058013e-07, "loss": 0.0033, "step": 61755 }, { "epoch": 19.0, "learning_rate": 1.3236372937122212e-07, "loss": 0.0037, "step": 61756 }, { "epoch": 19.0, "learning_rate": 1.3228293957054383e-07, "loss": 0.0032, "step": 61757 }, { "epoch": 19.0, "learning_rate": 1.3220217426874739e-07, "loss": 0.0029, "step": 61758 }, { "epoch": 19.0, "learning_rate": 1.3212143346603256e-07, "loss": 0.0031, "step": 61759 }, { "epoch": 19.0, "learning_rate": 1.3204071716259924e-07, "loss": 0.0018, "step": 61760 }, { "epoch": 19.0, "learning_rate": 1.3196002535864837e-07, "loss": 0.0029, "step": 61761 }, { "epoch": 19.0, "learning_rate": 1.318793580543809e-07, "loss": 0.0042, "step": 61762 }, { "epoch": 19.0, "learning_rate": 1.3179871524999778e-07, "loss": 0.004, "step": 61763 }, { "epoch": 19.0, "learning_rate": 1.3171809694569658e-07, "loss": 0.0029, "step": 61764 }, { "epoch": 19.0, "learning_rate": 1.3163750314167944e-07, "loss": 0.0045, "step": 61765 }, { "epoch": 19.0, "learning_rate": 1.3155693383814616e-07, "loss": 0.0026, "step": 61766 }, { "epoch": 19.0, "learning_rate": 1.3147638903529548e-07, "loss": 0.0043, "step": 61767 }, { "epoch": 19.01, "learning_rate": 1.3139586873332943e-07, "loss": 0.0052, "step": 61768 }, { "epoch": 19.01, "learning_rate": 1.3131537293244678e-07, "loss": 0.003, "step": 61769 }, { "epoch": 19.01, "learning_rate": 1.3123490163284736e-07, "loss": 0.0034, "step": 61770 }, { "epoch": 19.01, "learning_rate": 1.3115445483473211e-07, "loss": 0.0033, "step": 61771 }, { "epoch": 19.01, "learning_rate": 1.3107403253829863e-07, "loss": 0.0044, "step": 61772 }, { "epoch": 19.01, "learning_rate": 1.3099363474374793e-07, "loss": 0.0028, "step": 61773 }, { "epoch": 19.01, "learning_rate": 1.309132614512798e-07, "loss": 0.0029, "step": 61774 }, { "epoch": 19.01, "learning_rate": 1.30832912661093e-07, "loss": 0.0028, "step": 61775 }, { "epoch": 19.01, "learning_rate": 1.3075258837338734e-07, "loss": 0.0034, "step": 61776 }, { "epoch": 19.01, "learning_rate": 1.3067228858836378e-07, "loss": 0.0025, "step": 61777 }, { "epoch": 19.01, "learning_rate": 1.3059201330621885e-07, "loss": 0.0025, "step": 61778 }, { "epoch": 19.01, "learning_rate": 1.3051176252715348e-07, "loss": 0.0055, "step": 61779 }, { "epoch": 19.01, "learning_rate": 1.304315362513664e-07, "loss": 0.0026, "step": 61780 }, { "epoch": 19.01, "learning_rate": 1.3035133447905745e-07, "loss": 0.0016, "step": 61781 }, { "epoch": 19.01, "learning_rate": 1.3027115721042538e-07, "loss": 0.0054, "step": 61782 }, { "epoch": 19.01, "learning_rate": 1.301910044456689e-07, "loss": 0.0031, "step": 61783 }, { "epoch": 19.01, "learning_rate": 1.3011087618498786e-07, "loss": 0.0026, "step": 61784 }, { "epoch": 19.01, "learning_rate": 1.3003077242858096e-07, "loss": 0.0036, "step": 61785 }, { "epoch": 19.01, "learning_rate": 1.2995069317664477e-07, "loss": 0.0034, "step": 61786 }, { "epoch": 19.01, "learning_rate": 1.298706384293824e-07, "loss": 0.0091, "step": 61787 }, { "epoch": 19.01, "learning_rate": 1.297906081869893e-07, "loss": 0.0028, "step": 61788 }, { "epoch": 19.01, "learning_rate": 1.2971060244966526e-07, "loss": 0.0042, "step": 61789 }, { "epoch": 19.01, "learning_rate": 1.2963062121760904e-07, "loss": 0.0039, "step": 61790 }, { "epoch": 19.01, "learning_rate": 1.2955066449101938e-07, "loss": 0.0043, "step": 61791 }, { "epoch": 19.01, "learning_rate": 1.294707322700939e-07, "loss": 0.0029, "step": 61792 }, { "epoch": 19.01, "learning_rate": 1.2939082455503128e-07, "loss": 0.0036, "step": 61793 }, { "epoch": 19.01, "learning_rate": 1.293109413460314e-07, "loss": 0.0031, "step": 61794 }, { "epoch": 19.01, "learning_rate": 1.2923108264328966e-07, "loss": 0.0029, "step": 61795 }, { "epoch": 19.01, "learning_rate": 1.29151248447007e-07, "loss": 0.0034, "step": 61796 }, { "epoch": 19.01, "learning_rate": 1.2907143875738105e-07, "loss": 0.0042, "step": 61797 }, { "epoch": 19.01, "learning_rate": 1.2899165357460942e-07, "loss": 0.003, "step": 61798 }, { "epoch": 19.01, "learning_rate": 1.2891189289889083e-07, "loss": 0.0036, "step": 61799 }, { "epoch": 19.02, "learning_rate": 1.2883215673042182e-07, "loss": 0.0022, "step": 61800 }, { "epoch": 19.02, "learning_rate": 1.287524450694011e-07, "loss": 0.0036, "step": 61801 }, { "epoch": 19.02, "learning_rate": 1.286727579160274e-07, "loss": 0.0041, "step": 61802 }, { "epoch": 19.02, "learning_rate": 1.2859309527049835e-07, "loss": 0.0057, "step": 61803 }, { "epoch": 19.02, "learning_rate": 1.2851345713301157e-07, "loss": 0.0026, "step": 61804 }, { "epoch": 19.02, "learning_rate": 1.2843384350376465e-07, "loss": 0.0029, "step": 61805 }, { "epoch": 19.02, "learning_rate": 1.283542543829541e-07, "loss": 0.0025, "step": 61806 }, { "epoch": 19.02, "learning_rate": 1.2827468977077874e-07, "loss": 0.0044, "step": 61807 }, { "epoch": 19.02, "learning_rate": 1.281951496674372e-07, "loss": 0.0031, "step": 61808 }, { "epoch": 19.02, "learning_rate": 1.2811563407312489e-07, "loss": 0.0024, "step": 61809 }, { "epoch": 19.02, "learning_rate": 1.2803614298804057e-07, "loss": 0.0024, "step": 61810 }, { "epoch": 19.02, "learning_rate": 1.2795667641238074e-07, "loss": 0.002, "step": 61811 }, { "epoch": 19.02, "learning_rate": 1.2787723434634304e-07, "loss": 0.0037, "step": 61812 }, { "epoch": 19.02, "learning_rate": 1.2779781679012505e-07, "loss": 0.0053, "step": 61813 }, { "epoch": 19.02, "learning_rate": 1.277184237439233e-07, "loss": 0.0029, "step": 61814 }, { "epoch": 19.02, "learning_rate": 1.276390552079354e-07, "loss": 0.0026, "step": 61815 }, { "epoch": 19.02, "learning_rate": 1.2755971118235788e-07, "loss": 0.005, "step": 61816 }, { "epoch": 19.02, "learning_rate": 1.2748039166738834e-07, "loss": 0.0037, "step": 61817 }, { "epoch": 19.02, "learning_rate": 1.274010966632233e-07, "loss": 0.0024, "step": 61818 }, { "epoch": 19.02, "learning_rate": 1.273218261700604e-07, "loss": 0.0044, "step": 61819 }, { "epoch": 19.02, "learning_rate": 1.2724258018809498e-07, "loss": 0.0024, "step": 61820 }, { "epoch": 19.02, "learning_rate": 1.2716335871752472e-07, "loss": 0.0031, "step": 61821 }, { "epoch": 19.02, "learning_rate": 1.270841617585461e-07, "loss": 0.0037, "step": 61822 }, { "epoch": 19.02, "learning_rate": 1.2700498931135674e-07, "loss": 0.0029, "step": 61823 }, { "epoch": 19.02, "learning_rate": 1.2692584137615205e-07, "loss": 0.0021, "step": 61824 }, { "epoch": 19.02, "learning_rate": 1.2684671795312854e-07, "loss": 0.0046, "step": 61825 }, { "epoch": 19.02, "learning_rate": 1.2676761904248268e-07, "loss": 0.0039, "step": 61826 }, { "epoch": 19.02, "learning_rate": 1.2668854464441104e-07, "loss": 0.0024, "step": 61827 }, { "epoch": 19.02, "learning_rate": 1.2660949475911012e-07, "loss": 0.0046, "step": 61828 }, { "epoch": 19.02, "learning_rate": 1.2653046938677638e-07, "loss": 0.0035, "step": 61829 }, { "epoch": 19.02, "learning_rate": 1.2645146852760526e-07, "loss": 0.0033, "step": 61830 }, { "epoch": 19.02, "learning_rate": 1.2637249218179325e-07, "loss": 0.0025, "step": 61831 }, { "epoch": 19.02, "learning_rate": 1.26293540349538e-07, "loss": 0.0032, "step": 61832 }, { "epoch": 19.03, "learning_rate": 1.2621461303103266e-07, "loss": 0.0032, "step": 61833 }, { "epoch": 19.03, "learning_rate": 1.2613571022647376e-07, "loss": 0.0042, "step": 61834 }, { "epoch": 19.03, "learning_rate": 1.260568319360589e-07, "loss": 0.0035, "step": 61835 }, { "epoch": 19.03, "learning_rate": 1.2597797815998236e-07, "loss": 0.0052, "step": 61836 }, { "epoch": 19.03, "learning_rate": 1.258991488984418e-07, "loss": 0.0026, "step": 61837 }, { "epoch": 19.03, "learning_rate": 1.2582034415163147e-07, "loss": 0.002, "step": 61838 }, { "epoch": 19.03, "learning_rate": 1.257415639197468e-07, "loss": 0.0034, "step": 61839 }, { "epoch": 19.03, "learning_rate": 1.2566280820298427e-07, "loss": 0.0061, "step": 61840 }, { "epoch": 19.03, "learning_rate": 1.2558407700153708e-07, "loss": 0.003, "step": 61841 }, { "epoch": 19.03, "learning_rate": 1.2550537031560395e-07, "loss": 0.0024, "step": 61842 }, { "epoch": 19.03, "learning_rate": 1.2542668814537918e-07, "loss": 0.0014, "step": 61843 }, { "epoch": 19.03, "learning_rate": 1.2534803049105816e-07, "loss": 0.0041, "step": 61844 }, { "epoch": 19.03, "learning_rate": 1.2526939735283517e-07, "loss": 0.0056, "step": 61845 }, { "epoch": 19.03, "learning_rate": 1.2519078873090563e-07, "loss": 0.003, "step": 61846 }, { "epoch": 19.03, "learning_rate": 1.2511220462546603e-07, "loss": 0.0025, "step": 61847 }, { "epoch": 19.03, "learning_rate": 1.2503364503671066e-07, "loss": 0.0024, "step": 61848 }, { "epoch": 19.03, "learning_rate": 1.2495510996483384e-07, "loss": 0.0059, "step": 61849 }, { "epoch": 19.03, "learning_rate": 1.2487659941003204e-07, "loss": 0.0025, "step": 61850 }, { "epoch": 19.03, "learning_rate": 1.2479811337249848e-07, "loss": 0.0024, "step": 61851 }, { "epoch": 19.03, "learning_rate": 1.2471965185242963e-07, "loss": 0.0045, "step": 61852 }, { "epoch": 19.03, "learning_rate": 1.246412148500198e-07, "loss": 0.0034, "step": 61853 }, { "epoch": 19.03, "learning_rate": 1.2456280236546324e-07, "loss": 0.004, "step": 61854 }, { "epoch": 19.03, "learning_rate": 1.2448441439895541e-07, "loss": 0.0042, "step": 61855 }, { "epoch": 19.03, "learning_rate": 1.2440605095068948e-07, "loss": 0.0037, "step": 61856 }, { "epoch": 19.03, "learning_rate": 1.243277120208608e-07, "loss": 0.0036, "step": 61857 }, { "epoch": 19.03, "learning_rate": 1.242493976096648e-07, "loss": 0.0046, "step": 61858 }, { "epoch": 19.03, "learning_rate": 1.2417110771729468e-07, "loss": 0.0032, "step": 61859 }, { "epoch": 19.03, "learning_rate": 1.240928423439458e-07, "loss": 0.0021, "step": 61860 }, { "epoch": 19.03, "learning_rate": 1.2401460148981136e-07, "loss": 0.0045, "step": 61861 }, { "epoch": 19.03, "learning_rate": 1.2393638515508678e-07, "loss": 0.0054, "step": 61862 }, { "epoch": 19.03, "learning_rate": 1.2385819333996518e-07, "loss": 0.0038, "step": 61863 }, { "epoch": 19.03, "learning_rate": 1.237800260446409e-07, "loss": 0.0023, "step": 61864 }, { "epoch": 19.04, "learning_rate": 1.237018832693093e-07, "loss": 0.0053, "step": 61865 }, { "epoch": 19.04, "learning_rate": 1.2362376501416251e-07, "loss": 0.0022, "step": 61866 }, { "epoch": 19.04, "learning_rate": 1.2354567127939476e-07, "loss": 0.0026, "step": 61867 }, { "epoch": 19.04, "learning_rate": 1.2346760206520146e-07, "loss": 0.0052, "step": 61868 }, { "epoch": 19.04, "learning_rate": 1.233895573717758e-07, "loss": 0.0042, "step": 61869 }, { "epoch": 19.04, "learning_rate": 1.2331153719930987e-07, "loss": 0.0025, "step": 61870 }, { "epoch": 19.04, "learning_rate": 1.2323354154799904e-07, "loss": 0.0034, "step": 61871 }, { "epoch": 19.04, "learning_rate": 1.231555704180376e-07, "loss": 0.0037, "step": 61872 }, { "epoch": 19.04, "learning_rate": 1.2307762380961762e-07, "loss": 0.0032, "step": 61873 }, { "epoch": 19.04, "learning_rate": 1.229997017229334e-07, "loss": 0.0019, "step": 61874 }, { "epoch": 19.04, "learning_rate": 1.229218041581781e-07, "loss": 0.003, "step": 61875 }, { "epoch": 19.04, "learning_rate": 1.2284393111554493e-07, "loss": 0.0033, "step": 61876 }, { "epoch": 19.04, "learning_rate": 1.2276608259522705e-07, "loss": 0.0042, "step": 61877 }, { "epoch": 19.04, "learning_rate": 1.2268825859741872e-07, "loss": 0.0035, "step": 61878 }, { "epoch": 19.04, "learning_rate": 1.2261045912231318e-07, "loss": 0.0031, "step": 61879 }, { "epoch": 19.04, "learning_rate": 1.2253268417010245e-07, "loss": 0.0028, "step": 61880 }, { "epoch": 19.04, "learning_rate": 1.2245493374097973e-07, "loss": 0.0024, "step": 61881 }, { "epoch": 19.04, "learning_rate": 1.223772078351393e-07, "loss": 0.0044, "step": 61882 }, { "epoch": 19.04, "learning_rate": 1.2229950645277321e-07, "loss": 0.0045, "step": 61883 }, { "epoch": 19.04, "learning_rate": 1.2222182959407358e-07, "loss": 0.003, "step": 61884 }, { "epoch": 19.04, "learning_rate": 1.2214417725923578e-07, "loss": 0.0016, "step": 61885 }, { "epoch": 19.04, "learning_rate": 1.2206654944844853e-07, "loss": 0.0036, "step": 61886 }, { "epoch": 19.04, "learning_rate": 1.2198894616190948e-07, "loss": 0.0021, "step": 61887 }, { "epoch": 19.04, "learning_rate": 1.2191136739980846e-07, "loss": 0.0027, "step": 61888 }, { "epoch": 19.04, "learning_rate": 1.2183381316233756e-07, "loss": 0.0045, "step": 61889 }, { "epoch": 19.04, "learning_rate": 1.2175628344969103e-07, "loss": 0.0037, "step": 61890 }, { "epoch": 19.04, "learning_rate": 1.2167877826205876e-07, "loss": 0.0017, "step": 61891 }, { "epoch": 19.04, "learning_rate": 1.2160129759963723e-07, "loss": 0.0036, "step": 61892 }, { "epoch": 19.04, "learning_rate": 1.215238414626152e-07, "loss": 0.0025, "step": 61893 }, { "epoch": 19.04, "learning_rate": 1.2144640985118694e-07, "loss": 0.0021, "step": 61894 }, { "epoch": 19.04, "learning_rate": 1.2136900276554453e-07, "loss": 0.0031, "step": 61895 }, { "epoch": 19.04, "learning_rate": 1.2129162020587892e-07, "loss": 0.0056, "step": 61896 }, { "epoch": 19.04, "learning_rate": 1.212142621723833e-07, "loss": 0.003, "step": 61897 }, { "epoch": 19.05, "learning_rate": 1.2113692866524862e-07, "loss": 0.0028, "step": 61898 }, { "epoch": 19.05, "learning_rate": 1.2105961968466805e-07, "loss": 0.0042, "step": 61899 }, { "epoch": 19.05, "learning_rate": 1.209823352308337e-07, "loss": 0.0017, "step": 61900 }, { "epoch": 19.05, "learning_rate": 1.2090507530393757e-07, "loss": 0.0038, "step": 61901 }, { "epoch": 19.05, "learning_rate": 1.2082783990416958e-07, "loss": 0.0022, "step": 61902 }, { "epoch": 19.05, "learning_rate": 1.2075062903172286e-07, "loss": 0.0026, "step": 61903 }, { "epoch": 19.05, "learning_rate": 1.2067344268678948e-07, "loss": 0.0051, "step": 61904 }, { "epoch": 19.05, "learning_rate": 1.2059628086956044e-07, "loss": 0.0033, "step": 61905 }, { "epoch": 19.05, "learning_rate": 1.2051914358022666e-07, "loss": 0.0055, "step": 61906 }, { "epoch": 19.05, "learning_rate": 1.204420308189802e-07, "loss": 0.0034, "step": 61907 }, { "epoch": 19.05, "learning_rate": 1.2036494258601427e-07, "loss": 0.0032, "step": 61908 }, { "epoch": 19.05, "learning_rate": 1.202878788815187e-07, "loss": 0.0012, "step": 61909 }, { "epoch": 19.05, "learning_rate": 1.2021083970568336e-07, "loss": 0.0018, "step": 61910 }, { "epoch": 19.05, "learning_rate": 1.201338250587014e-07, "loss": 0.0033, "step": 61911 }, { "epoch": 19.05, "learning_rate": 1.2005683494076382e-07, "loss": 0.0029, "step": 61912 }, { "epoch": 19.05, "learning_rate": 1.199798693520615e-07, "loss": 0.0033, "step": 61913 }, { "epoch": 19.05, "learning_rate": 1.1990292829278661e-07, "loss": 0.0024, "step": 61914 }, { "epoch": 19.05, "learning_rate": 1.198260117631278e-07, "loss": 0.0029, "step": 61915 }, { "epoch": 19.05, "learning_rate": 1.197491197632783e-07, "loss": 0.0045, "step": 61916 }, { "epoch": 19.05, "learning_rate": 1.1967225229342794e-07, "loss": 0.0027, "step": 61917 }, { "epoch": 19.05, "learning_rate": 1.1959540935376769e-07, "loss": 0.0039, "step": 61918 }, { "epoch": 19.05, "learning_rate": 1.1951859094448848e-07, "loss": 0.0039, "step": 61919 }, { "epoch": 19.05, "learning_rate": 1.1944179706578018e-07, "loss": 0.0037, "step": 61920 }, { "epoch": 19.05, "learning_rate": 1.1936502771783488e-07, "loss": 0.0024, "step": 61921 }, { "epoch": 19.05, "learning_rate": 1.1928828290084238e-07, "loss": 0.0043, "step": 61922 }, { "epoch": 19.05, "learning_rate": 1.1921156261499255e-07, "loss": 0.0048, "step": 61923 }, { "epoch": 19.05, "learning_rate": 1.1913486686047748e-07, "loss": 0.0044, "step": 61924 }, { "epoch": 19.05, "learning_rate": 1.19058195637487e-07, "loss": 0.0038, "step": 61925 }, { "epoch": 19.05, "learning_rate": 1.1898154894620984e-07, "loss": 0.0042, "step": 61926 }, { "epoch": 19.05, "learning_rate": 1.1890492678683808e-07, "loss": 0.004, "step": 61927 }, { "epoch": 19.05, "learning_rate": 1.1882832915956266e-07, "loss": 0.0037, "step": 61928 }, { "epoch": 19.05, "learning_rate": 1.1875175606457234e-07, "loss": 0.0057, "step": 61929 }, { "epoch": 19.06, "learning_rate": 1.1867520750205696e-07, "loss": 0.0047, "step": 61930 }, { "epoch": 19.06, "learning_rate": 1.1859868347220749e-07, "loss": 0.0029, "step": 61931 }, { "epoch": 19.06, "learning_rate": 1.1852218397521264e-07, "loss": 0.002, "step": 61932 }, { "epoch": 19.06, "learning_rate": 1.1844570901126451e-07, "loss": 0.002, "step": 61933 }, { "epoch": 19.06, "learning_rate": 1.183692585805507e-07, "loss": 0.0046, "step": 61934 }, { "epoch": 19.06, "learning_rate": 1.182928326832622e-07, "loss": 0.0038, "step": 61935 }, { "epoch": 19.06, "learning_rate": 1.1821643131958882e-07, "loss": 0.0029, "step": 61936 }, { "epoch": 19.06, "learning_rate": 1.1814005448972043e-07, "loss": 0.0049, "step": 61937 }, { "epoch": 19.06, "learning_rate": 1.1806370219384577e-07, "loss": 0.0022, "step": 61938 }, { "epoch": 19.06, "learning_rate": 1.1798737443215469e-07, "loss": 0.0034, "step": 61939 }, { "epoch": 19.06, "learning_rate": 1.1791107120483702e-07, "loss": 0.0029, "step": 61940 }, { "epoch": 19.06, "learning_rate": 1.1783479251208152e-07, "loss": 0.005, "step": 61941 }, { "epoch": 19.06, "learning_rate": 1.1775853835407913e-07, "loss": 0.0033, "step": 61942 }, { "epoch": 19.06, "learning_rate": 1.1768230873101749e-07, "loss": 0.0043, "step": 61943 }, { "epoch": 19.06, "learning_rate": 1.1760610364308645e-07, "loss": 0.0028, "step": 61944 }, { "epoch": 19.06, "learning_rate": 1.1752992309047584e-07, "loss": 0.0047, "step": 61945 }, { "epoch": 19.06, "learning_rate": 1.174537670733733e-07, "loss": 0.0027, "step": 61946 }, { "epoch": 19.06, "learning_rate": 1.1737763559196869e-07, "loss": 0.0039, "step": 61947 }, { "epoch": 19.06, "learning_rate": 1.1730152864645184e-07, "loss": 0.0022, "step": 61948 }, { "epoch": 19.06, "learning_rate": 1.1722544623701038e-07, "loss": 0.003, "step": 61949 }, { "epoch": 19.06, "learning_rate": 1.1714938836383526e-07, "loss": 0.005, "step": 61950 }, { "epoch": 19.06, "learning_rate": 1.1707335502711192e-07, "loss": 0.0032, "step": 61951 }, { "epoch": 19.06, "learning_rate": 1.1699734622703241e-07, "loss": 0.0051, "step": 61952 }, { "epoch": 19.06, "learning_rate": 1.1692136196378435e-07, "loss": 0.0028, "step": 61953 }, { "epoch": 19.06, "learning_rate": 1.1684540223755537e-07, "loss": 0.004, "step": 61954 }, { "epoch": 19.06, "learning_rate": 1.1676946704853531e-07, "loss": 0.0023, "step": 61955 }, { "epoch": 19.06, "learning_rate": 1.1669355639691182e-07, "loss": 0.0037, "step": 61956 }, { "epoch": 19.06, "learning_rate": 1.1661767028287363e-07, "loss": 0.0038, "step": 61957 }, { "epoch": 19.06, "learning_rate": 1.1654180870660947e-07, "loss": 0.0021, "step": 61958 }, { "epoch": 19.06, "learning_rate": 1.1646597166830809e-07, "loss": 0.0032, "step": 61959 }, { "epoch": 19.06, "learning_rate": 1.1639015916815599e-07, "loss": 0.0017, "step": 61960 }, { "epoch": 19.06, "learning_rate": 1.1631437120634304e-07, "loss": 0.0041, "step": 61961 }, { "epoch": 19.06, "learning_rate": 1.1623860778305684e-07, "loss": 0.0035, "step": 61962 }, { "epoch": 19.07, "learning_rate": 1.1616286889848726e-07, "loss": 0.0052, "step": 61963 }, { "epoch": 19.07, "learning_rate": 1.1608715455281861e-07, "loss": 0.0023, "step": 61964 }, { "epoch": 19.07, "learning_rate": 1.160114647462418e-07, "loss": 0.0037, "step": 61965 }, { "epoch": 19.07, "learning_rate": 1.159357994789434e-07, "loss": 0.0021, "step": 61966 }, { "epoch": 19.07, "learning_rate": 1.1586015875111211e-07, "loss": 0.0041, "step": 61967 }, { "epoch": 19.07, "learning_rate": 1.1578454256293559e-07, "loss": 0.0027, "step": 61968 }, { "epoch": 19.07, "learning_rate": 1.1570895091460144e-07, "loss": 0.0038, "step": 61969 }, { "epoch": 19.07, "learning_rate": 1.1563338380629618e-07, "loss": 0.0035, "step": 61970 }, { "epoch": 19.07, "learning_rate": 1.1555784123820967e-07, "loss": 0.0041, "step": 61971 }, { "epoch": 19.07, "learning_rate": 1.1548232321052732e-07, "loss": 0.0032, "step": 61972 }, { "epoch": 19.07, "learning_rate": 1.1540682972343898e-07, "loss": 0.0038, "step": 61973 }, { "epoch": 19.07, "learning_rate": 1.1533136077713003e-07, "loss": 0.0028, "step": 61974 }, { "epoch": 19.07, "learning_rate": 1.1525591637178813e-07, "loss": 0.0025, "step": 61975 }, { "epoch": 19.07, "learning_rate": 1.1518049650759977e-07, "loss": 0.0062, "step": 61976 }, { "epoch": 19.07, "learning_rate": 1.1510510118475593e-07, "loss": 0.0023, "step": 61977 }, { "epoch": 19.07, "learning_rate": 1.1502973040343979e-07, "loss": 0.0034, "step": 61978 }, { "epoch": 19.07, "learning_rate": 1.1495438416384009e-07, "loss": 0.0024, "step": 61979 }, { "epoch": 19.07, "learning_rate": 1.1487906246614333e-07, "loss": 0.0023, "step": 61980 }, { "epoch": 19.07, "learning_rate": 1.1480376531053716e-07, "loss": 0.0078, "step": 61981 }, { "epoch": 19.07, "learning_rate": 1.147284926972081e-07, "loss": 0.0033, "step": 61982 }, { "epoch": 19.07, "learning_rate": 1.1465324462634375e-07, "loss": 0.0034, "step": 61983 }, { "epoch": 19.07, "learning_rate": 1.1457802109813065e-07, "loss": 0.0035, "step": 61984 }, { "epoch": 19.07, "learning_rate": 1.1450282211275532e-07, "loss": 0.0025, "step": 61985 }, { "epoch": 19.07, "learning_rate": 1.1442764767040315e-07, "loss": 0.0024, "step": 61986 }, { "epoch": 19.07, "learning_rate": 1.1435249777126289e-07, "loss": 0.0024, "step": 61987 }, { "epoch": 19.07, "learning_rate": 1.1427737241551995e-07, "loss": 0.0024, "step": 61988 }, { "epoch": 19.07, "learning_rate": 1.1420227160336195e-07, "loss": 0.0022, "step": 61989 }, { "epoch": 19.07, "learning_rate": 1.141271953349743e-07, "loss": 0.0035, "step": 61990 }, { "epoch": 19.07, "learning_rate": 1.140521436105435e-07, "loss": 0.003, "step": 61991 }, { "epoch": 19.07, "learning_rate": 1.139771164302561e-07, "loss": 0.0034, "step": 61992 }, { "epoch": 19.07, "learning_rate": 1.139021137942986e-07, "loss": 0.0024, "step": 61993 }, { "epoch": 19.07, "learning_rate": 1.1382713570285642e-07, "loss": 0.0025, "step": 61994 }, { "epoch": 19.08, "learning_rate": 1.1375218215611606e-07, "loss": 0.003, "step": 61995 }, { "epoch": 19.08, "learning_rate": 1.1367725315426403e-07, "loss": 0.0041, "step": 61996 }, { "epoch": 19.08, "learning_rate": 1.1360234869748688e-07, "loss": 0.0039, "step": 61997 }, { "epoch": 19.08, "learning_rate": 1.1352746878596887e-07, "loss": 0.0045, "step": 61998 }, { "epoch": 19.08, "learning_rate": 1.1345261341989766e-07, "loss": 0.0031, "step": 61999 }, { "epoch": 19.08, "learning_rate": 1.1337778259945753e-07, "loss": 0.0027, "step": 62000 }, { "epoch": 19.08, "learning_rate": 1.1330297632483389e-07, "loss": 0.003, "step": 62001 }, { "epoch": 19.08, "learning_rate": 1.1322819459621548e-07, "loss": 0.0044, "step": 62002 }, { "epoch": 19.08, "learning_rate": 1.1315343741378549e-07, "loss": 0.0035, "step": 62003 }, { "epoch": 19.08, "learning_rate": 1.1307870477773042e-07, "loss": 0.0032, "step": 62004 }, { "epoch": 19.08, "learning_rate": 1.1300399668823569e-07, "loss": 0.0037, "step": 62005 }, { "epoch": 19.08, "learning_rate": 1.1292931314548561e-07, "loss": 0.0048, "step": 62006 }, { "epoch": 19.08, "learning_rate": 1.1285465414966667e-07, "loss": 0.0042, "step": 62007 }, { "epoch": 19.08, "learning_rate": 1.1278001970096542e-07, "loss": 0.0044, "step": 62008 }, { "epoch": 19.08, "learning_rate": 1.1270540979956501e-07, "loss": 0.0063, "step": 62009 }, { "epoch": 19.08, "learning_rate": 1.1263082444565088e-07, "loss": 0.0028, "step": 62010 }, { "epoch": 19.08, "learning_rate": 1.1255626363940952e-07, "loss": 0.0031, "step": 62011 }, { "epoch": 19.08, "learning_rate": 1.1248172738102525e-07, "loss": 0.0061, "step": 62012 }, { "epoch": 19.08, "learning_rate": 1.1240721567068346e-07, "loss": 0.0043, "step": 62013 }, { "epoch": 19.08, "learning_rate": 1.1233272850856958e-07, "loss": 0.0022, "step": 62014 }, { "epoch": 19.08, "learning_rate": 1.1225826589486677e-07, "loss": 0.0026, "step": 62015 }, { "epoch": 19.08, "learning_rate": 1.1218382782976045e-07, "loss": 0.0019, "step": 62016 }, { "epoch": 19.08, "learning_rate": 1.1210941431343714e-07, "loss": 0.003, "step": 62017 }, { "epoch": 19.08, "learning_rate": 1.1203502534608113e-07, "loss": 0.0033, "step": 62018 }, { "epoch": 19.08, "learning_rate": 1.1196066092787561e-07, "loss": 0.0026, "step": 62019 }, { "epoch": 19.08, "learning_rate": 1.1188632105900598e-07, "loss": 0.0048, "step": 62020 }, { "epoch": 19.08, "learning_rate": 1.1181200573965655e-07, "loss": 0.0022, "step": 62021 }, { "epoch": 19.08, "learning_rate": 1.1173771497001273e-07, "loss": 0.0043, "step": 62022 }, { "epoch": 19.08, "learning_rate": 1.116634487502577e-07, "loss": 0.0026, "step": 62023 }, { "epoch": 19.08, "learning_rate": 1.1158920708057686e-07, "loss": 0.0052, "step": 62024 }, { "epoch": 19.08, "learning_rate": 1.115149899611545e-07, "loss": 0.0023, "step": 62025 }, { "epoch": 19.08, "learning_rate": 1.1144079739217495e-07, "loss": 0.0034, "step": 62026 }, { "epoch": 19.08, "learning_rate": 1.1136662937382137e-07, "loss": 0.0022, "step": 62027 }, { "epoch": 19.09, "learning_rate": 1.1129248590627806e-07, "loss": 0.0037, "step": 62028 }, { "epoch": 19.09, "learning_rate": 1.1121836698973043e-07, "loss": 0.0031, "step": 62029 }, { "epoch": 19.09, "learning_rate": 1.1114427262436167e-07, "loss": 0.0087, "step": 62030 }, { "epoch": 19.09, "learning_rate": 1.1107020281035497e-07, "loss": 0.0036, "step": 62031 }, { "epoch": 19.09, "learning_rate": 1.1099615754789573e-07, "loss": 0.004, "step": 62032 }, { "epoch": 19.09, "learning_rate": 1.1092213683716713e-07, "loss": 0.0016, "step": 62033 }, { "epoch": 19.09, "learning_rate": 1.1084814067835236e-07, "loss": 0.004, "step": 62034 }, { "epoch": 19.09, "learning_rate": 1.1077416907163573e-07, "loss": 0.0035, "step": 62035 }, { "epoch": 19.09, "learning_rate": 1.1070022201720044e-07, "loss": 0.0032, "step": 62036 }, { "epoch": 19.09, "learning_rate": 1.1062629951523074e-07, "loss": 0.0028, "step": 62037 }, { "epoch": 19.09, "learning_rate": 1.1055240156590985e-07, "loss": 0.0065, "step": 62038 }, { "epoch": 19.09, "learning_rate": 1.1047852816942206e-07, "loss": 0.0026, "step": 62039 }, { "epoch": 19.09, "learning_rate": 1.1040467932594945e-07, "loss": 0.0037, "step": 62040 }, { "epoch": 19.09, "learning_rate": 1.103308550356752e-07, "loss": 0.0033, "step": 62041 }, { "epoch": 19.09, "learning_rate": 1.1025705529878361e-07, "loss": 0.0044, "step": 62042 }, { "epoch": 19.09, "learning_rate": 1.1018328011545787e-07, "loss": 0.0062, "step": 62043 }, { "epoch": 19.09, "learning_rate": 1.1010952948588005e-07, "loss": 0.0038, "step": 62044 }, { "epoch": 19.09, "learning_rate": 1.1003580341023445e-07, "loss": 0.0034, "step": 62045 }, { "epoch": 19.09, "learning_rate": 1.0996210188870427e-07, "loss": 0.0024, "step": 62046 }, { "epoch": 19.09, "learning_rate": 1.0988842492147155e-07, "loss": 0.0034, "step": 62047 }, { "epoch": 19.09, "learning_rate": 1.0981477250871952e-07, "loss": 0.0049, "step": 62048 }, { "epoch": 19.09, "learning_rate": 1.0974114465063135e-07, "loss": 0.0074, "step": 62049 }, { "epoch": 19.09, "learning_rate": 1.0966754134738911e-07, "loss": 0.0033, "step": 62050 }, { "epoch": 19.09, "learning_rate": 1.0959396259917487e-07, "loss": 0.0062, "step": 62051 }, { "epoch": 19.09, "learning_rate": 1.0952040840617406e-07, "loss": 0.0025, "step": 62052 }, { "epoch": 19.09, "learning_rate": 1.0944687876856875e-07, "loss": 0.0045, "step": 62053 }, { "epoch": 19.09, "learning_rate": 1.0937337368653877e-07, "loss": 0.004, "step": 62054 }, { "epoch": 19.09, "learning_rate": 1.0929989316026846e-07, "loss": 0.0019, "step": 62055 }, { "epoch": 19.09, "learning_rate": 1.0922643718993986e-07, "loss": 0.0032, "step": 62056 }, { "epoch": 19.09, "learning_rate": 1.0915300577573618e-07, "loss": 0.0019, "step": 62057 }, { "epoch": 19.09, "learning_rate": 1.0907959891783837e-07, "loss": 0.0021, "step": 62058 }, { "epoch": 19.09, "learning_rate": 1.0900621661642962e-07, "loss": 0.0034, "step": 62059 }, { "epoch": 19.1, "learning_rate": 1.0893285887169313e-07, "loss": 0.0026, "step": 62060 }, { "epoch": 19.1, "learning_rate": 1.0885952568380764e-07, "loss": 0.004, "step": 62061 }, { "epoch": 19.1, "learning_rate": 1.0878621705295855e-07, "loss": 0.0025, "step": 62062 }, { "epoch": 19.1, "learning_rate": 1.0871293297932683e-07, "loss": 0.0027, "step": 62063 }, { "epoch": 19.1, "learning_rate": 1.0863967346309457e-07, "loss": 0.0053, "step": 62064 }, { "epoch": 19.1, "learning_rate": 1.0856643850444271e-07, "loss": 0.0029, "step": 62065 }, { "epoch": 19.1, "learning_rate": 1.0849322810355445e-07, "loss": 0.0028, "step": 62066 }, { "epoch": 19.1, "learning_rate": 1.0842004226060965e-07, "loss": 0.0022, "step": 62067 }, { "epoch": 19.1, "learning_rate": 1.083468809757926e-07, "loss": 0.005, "step": 62068 }, { "epoch": 19.1, "learning_rate": 1.0827374424928316e-07, "loss": 0.0043, "step": 62069 }, { "epoch": 19.1, "learning_rate": 1.0820063208126231e-07, "loss": 0.0059, "step": 62070 }, { "epoch": 19.1, "learning_rate": 1.081275444719132e-07, "loss": 0.002, "step": 62071 }, { "epoch": 19.1, "learning_rate": 1.0805448142141572e-07, "loss": 0.0034, "step": 62072 }, { "epoch": 19.1, "learning_rate": 1.0798144292995416e-07, "loss": 0.0082, "step": 62073 }, { "epoch": 19.1, "learning_rate": 1.0790842899770726e-07, "loss": 0.0049, "step": 62074 }, { "epoch": 19.1, "learning_rate": 1.0783543962485598e-07, "loss": 0.0038, "step": 62075 }, { "epoch": 19.1, "learning_rate": 1.0776247481158242e-07, "loss": 0.0039, "step": 62076 }, { "epoch": 19.1, "learning_rate": 1.0768953455806863e-07, "loss": 0.0039, "step": 62077 }, { "epoch": 19.1, "learning_rate": 1.0761661886449449e-07, "loss": 0.0044, "step": 62078 }, { "epoch": 19.1, "learning_rate": 1.0754372773104093e-07, "loss": 0.0024, "step": 62079 }, { "epoch": 19.1, "learning_rate": 1.0747086115789007e-07, "loss": 0.0043, "step": 62080 }, { "epoch": 19.1, "learning_rate": 1.0739801914522174e-07, "loss": 0.0027, "step": 62081 }, { "epoch": 19.1, "learning_rate": 1.0732520169321803e-07, "loss": 0.0046, "step": 62082 }, { "epoch": 19.1, "learning_rate": 1.0725240880205768e-07, "loss": 0.0025, "step": 62083 }, { "epoch": 19.1, "learning_rate": 1.0717964047192275e-07, "loss": 0.0033, "step": 62084 }, { "epoch": 19.1, "learning_rate": 1.0710689670299423e-07, "loss": 0.0046, "step": 62085 }, { "epoch": 19.1, "learning_rate": 1.0703417749545198e-07, "loss": 0.0026, "step": 62086 }, { "epoch": 19.1, "learning_rate": 1.0696148284947694e-07, "loss": 0.0026, "step": 62087 }, { "epoch": 19.1, "learning_rate": 1.068888127652501e-07, "loss": 0.0028, "step": 62088 }, { "epoch": 19.1, "learning_rate": 1.0681616724295019e-07, "loss": 0.0019, "step": 62089 }, { "epoch": 19.1, "learning_rate": 1.0674354628275818e-07, "loss": 0.0039, "step": 62090 }, { "epoch": 19.1, "learning_rate": 1.0667094988485616e-07, "loss": 0.002, "step": 62091 }, { "epoch": 19.1, "learning_rate": 1.0659837804942175e-07, "loss": 0.0023, "step": 62092 }, { "epoch": 19.11, "learning_rate": 1.0652583077663703e-07, "loss": 0.0042, "step": 62093 }, { "epoch": 19.11, "learning_rate": 1.0645330806668186e-07, "loss": 0.0021, "step": 62094 }, { "epoch": 19.11, "learning_rate": 1.0638080991973609e-07, "loss": 0.0025, "step": 62095 }, { "epoch": 19.11, "learning_rate": 1.0630833633597848e-07, "loss": 0.0035, "step": 62096 }, { "epoch": 19.11, "learning_rate": 1.062358873155911e-07, "loss": 0.0028, "step": 62097 }, { "epoch": 19.11, "learning_rate": 1.0616346285875156e-07, "loss": 0.0039, "step": 62098 }, { "epoch": 19.11, "learning_rate": 1.0609106296564197e-07, "loss": 0.0028, "step": 62099 }, { "epoch": 19.11, "learning_rate": 1.0601868763643997e-07, "loss": 0.0028, "step": 62100 }, { "epoch": 19.11, "learning_rate": 1.0594633687132649e-07, "loss": 0.0033, "step": 62101 }, { "epoch": 19.11, "learning_rate": 1.0587401067048142e-07, "loss": 0.0033, "step": 62102 }, { "epoch": 19.11, "learning_rate": 1.058017090340835e-07, "loss": 0.002, "step": 62103 }, { "epoch": 19.11, "learning_rate": 1.0572943196231255e-07, "loss": 0.0025, "step": 62104 }, { "epoch": 19.11, "learning_rate": 1.0565717945534737e-07, "loss": 0.0026, "step": 62105 }, { "epoch": 19.11, "learning_rate": 1.0558495151336889e-07, "loss": 0.0035, "step": 62106 }, { "epoch": 19.11, "learning_rate": 1.0551274813655476e-07, "loss": 0.0025, "step": 62107 }, { "epoch": 19.11, "learning_rate": 1.0544056932508706e-07, "loss": 0.009, "step": 62108 }, { "epoch": 19.11, "learning_rate": 1.0536841507914119e-07, "loss": 0.0028, "step": 62109 }, { "epoch": 19.11, "learning_rate": 1.0529628539889703e-07, "loss": 0.0043, "step": 62110 }, { "epoch": 19.11, "learning_rate": 1.0522418028453663e-07, "loss": 0.0028, "step": 62111 }, { "epoch": 19.11, "learning_rate": 1.0515209973623653e-07, "loss": 0.0039, "step": 62112 }, { "epoch": 19.11, "learning_rate": 1.0508004375417546e-07, "loss": 0.0031, "step": 62113 }, { "epoch": 19.11, "learning_rate": 1.050080123385333e-07, "loss": 0.0074, "step": 62114 }, { "epoch": 19.11, "learning_rate": 1.0493600548948879e-07, "loss": 0.002, "step": 62115 }, { "epoch": 19.11, "learning_rate": 1.0486402320722066e-07, "loss": 0.0032, "step": 62116 }, { "epoch": 19.11, "learning_rate": 1.0479206549190767e-07, "loss": 0.0043, "step": 62117 }, { "epoch": 19.11, "learning_rate": 1.0472013234372858e-07, "loss": 0.002, "step": 62118 }, { "epoch": 19.11, "learning_rate": 1.04648223762861e-07, "loss": 0.0035, "step": 62119 }, { "epoch": 19.11, "learning_rate": 1.0457633974948367e-07, "loss": 0.0062, "step": 62120 }, { "epoch": 19.11, "learning_rate": 1.0450448030377758e-07, "loss": 0.0028, "step": 62121 }, { "epoch": 19.11, "learning_rate": 1.0443264542591702e-07, "loss": 0.0081, "step": 62122 }, { "epoch": 19.11, "learning_rate": 1.0436083511608408e-07, "loss": 0.0027, "step": 62123 }, { "epoch": 19.11, "learning_rate": 1.0428904937445416e-07, "loss": 0.0035, "step": 62124 }, { "epoch": 19.12, "learning_rate": 1.0421728820120713e-07, "loss": 0.0027, "step": 62125 }, { "epoch": 19.12, "learning_rate": 1.0414555159652062e-07, "loss": 0.0023, "step": 62126 }, { "epoch": 19.12, "learning_rate": 1.0407383956057337e-07, "loss": 0.0045, "step": 62127 }, { "epoch": 19.12, "learning_rate": 1.0400215209354191e-07, "loss": 0.0021, "step": 62128 }, { "epoch": 19.12, "learning_rate": 1.0393048919560722e-07, "loss": 0.0021, "step": 62129 }, { "epoch": 19.12, "learning_rate": 1.0385885086694247e-07, "loss": 0.0033, "step": 62130 }, { "epoch": 19.12, "learning_rate": 1.0378723710772975e-07, "loss": 0.0027, "step": 62131 }, { "epoch": 19.12, "learning_rate": 1.0371564791814559e-07, "loss": 0.0041, "step": 62132 }, { "epoch": 19.12, "learning_rate": 1.0364408329836762e-07, "loss": 0.0025, "step": 62133 }, { "epoch": 19.12, "learning_rate": 1.0357254324857346e-07, "loss": 0.0027, "step": 62134 }, { "epoch": 19.12, "learning_rate": 1.0350102776893966e-07, "loss": 0.0032, "step": 62135 }, { "epoch": 19.12, "learning_rate": 1.0342953685964608e-07, "loss": 0.0034, "step": 62136 }, { "epoch": 19.12, "learning_rate": 1.0335807052086811e-07, "loss": 0.0052, "step": 62137 }, { "epoch": 19.12, "learning_rate": 1.032866287527845e-07, "loss": 0.0023, "step": 62138 }, { "epoch": 19.12, "learning_rate": 1.032152115555718e-07, "loss": 0.0057, "step": 62139 }, { "epoch": 19.12, "learning_rate": 1.031438189294065e-07, "loss": 0.003, "step": 62140 }, { "epoch": 19.12, "learning_rate": 1.0307245087446849e-07, "loss": 0.0018, "step": 62141 }, { "epoch": 19.12, "learning_rate": 1.0300110739093428e-07, "loss": 0.0048, "step": 62142 }, { "epoch": 19.12, "learning_rate": 1.0292978847897816e-07, "loss": 0.0032, "step": 62143 }, { "epoch": 19.12, "learning_rate": 1.0285849413878002e-07, "loss": 0.0024, "step": 62144 }, { "epoch": 19.12, "learning_rate": 1.0278722437051636e-07, "loss": 0.0033, "step": 62145 }, { "epoch": 19.12, "learning_rate": 1.027159791743626e-07, "loss": 0.0046, "step": 62146 }, { "epoch": 19.12, "learning_rate": 1.026447585504986e-07, "loss": 0.0059, "step": 62147 }, { "epoch": 19.12, "learning_rate": 1.0257356249909867e-07, "loss": 0.0056, "step": 62148 }, { "epoch": 19.12, "learning_rate": 1.0250239102033931e-07, "loss": 0.0042, "step": 62149 }, { "epoch": 19.12, "learning_rate": 1.0243124411439931e-07, "loss": 0.0029, "step": 62150 }, { "epoch": 19.12, "learning_rate": 1.0236012178145404e-07, "loss": 0.0041, "step": 62151 }, { "epoch": 19.12, "learning_rate": 1.0228902402168118e-07, "loss": 0.002, "step": 62152 }, { "epoch": 19.12, "learning_rate": 1.0221795083525498e-07, "loss": 0.003, "step": 62153 }, { "epoch": 19.12, "learning_rate": 1.0214690222235423e-07, "loss": 0.0021, "step": 62154 }, { "epoch": 19.12, "learning_rate": 1.0207587818315323e-07, "loss": 0.0025, "step": 62155 }, { "epoch": 19.12, "learning_rate": 1.0200487871783071e-07, "loss": 0.0033, "step": 62156 }, { "epoch": 19.12, "learning_rate": 1.0193390382656099e-07, "loss": 0.0026, "step": 62157 }, { "epoch": 19.13, "learning_rate": 1.0186295350952058e-07, "loss": 0.0024, "step": 62158 }, { "epoch": 19.13, "learning_rate": 1.0179202776688601e-07, "loss": 0.005, "step": 62159 }, { "epoch": 19.13, "learning_rate": 1.0172112659883381e-07, "loss": 0.0027, "step": 62160 }, { "epoch": 19.13, "learning_rate": 1.016502500055394e-07, "loss": 0.0041, "step": 62161 }, { "epoch": 19.13, "learning_rate": 1.015793979871782e-07, "loss": 0.0024, "step": 62162 }, { "epoch": 19.13, "learning_rate": 1.0150857054392893e-07, "loss": 0.003, "step": 62163 }, { "epoch": 19.13, "learning_rate": 1.014377676759637e-07, "loss": 0.0041, "step": 62164 }, { "epoch": 19.13, "learning_rate": 1.0136698938346012e-07, "loss": 0.0022, "step": 62165 }, { "epoch": 19.13, "learning_rate": 1.0129623566659364e-07, "loss": 0.0042, "step": 62166 }, { "epoch": 19.13, "learning_rate": 1.0122550652553964e-07, "loss": 0.003, "step": 62167 }, { "epoch": 19.13, "learning_rate": 1.0115480196047467e-07, "loss": 0.0032, "step": 62168 }, { "epoch": 19.13, "learning_rate": 1.0108412197157413e-07, "loss": 0.0028, "step": 62169 }, { "epoch": 19.13, "learning_rate": 1.0101346655901234e-07, "loss": 0.0027, "step": 62170 }, { "epoch": 19.13, "learning_rate": 1.0094283572296581e-07, "loss": 0.0048, "step": 62171 }, { "epoch": 19.13, "learning_rate": 1.0087222946360886e-07, "loss": 0.0022, "step": 62172 }, { "epoch": 19.13, "learning_rate": 1.00801647781118e-07, "loss": 0.0039, "step": 62173 }, { "epoch": 19.13, "learning_rate": 1.0073109067566867e-07, "loss": 0.0031, "step": 62174 }, { "epoch": 19.13, "learning_rate": 1.0066055814743292e-07, "loss": 0.0022, "step": 62175 }, { "epoch": 19.13, "learning_rate": 1.0059005019659063e-07, "loss": 0.0048, "step": 62176 }, { "epoch": 19.13, "learning_rate": 1.0051956682331387e-07, "loss": 0.0028, "step": 62177 }, { "epoch": 19.13, "learning_rate": 1.0044910802777807e-07, "loss": 0.0047, "step": 62178 }, { "epoch": 19.13, "learning_rate": 1.0037867381015753e-07, "loss": 0.0034, "step": 62179 }, { "epoch": 19.13, "learning_rate": 1.0030826417062878e-07, "loss": 0.003, "step": 62180 }, { "epoch": 19.13, "learning_rate": 1.00237879109365e-07, "loss": 0.003, "step": 62181 }, { "epoch": 19.13, "learning_rate": 1.0016751862654161e-07, "loss": 0.0035, "step": 62182 }, { "epoch": 19.13, "learning_rate": 1.0009718272233404e-07, "loss": 0.0024, "step": 62183 }, { "epoch": 19.13, "learning_rate": 1.0002687139691658e-07, "loss": 0.003, "step": 62184 }, { "epoch": 19.13, "learning_rate": 9.995658465046243e-08, "loss": 0.0047, "step": 62185 }, { "epoch": 19.13, "learning_rate": 9.988632248314811e-08, "loss": 0.0038, "step": 62186 }, { "epoch": 19.13, "learning_rate": 9.981608489514683e-08, "loss": 0.0052, "step": 62187 }, { "epoch": 19.13, "learning_rate": 9.974587188663287e-08, "loss": 0.0032, "step": 62188 }, { "epoch": 19.13, "learning_rate": 9.967568345778056e-08, "loss": 0.0032, "step": 62189 }, { "epoch": 19.14, "learning_rate": 9.960551960876529e-08, "loss": 0.0043, "step": 62190 }, { "epoch": 19.14, "learning_rate": 9.953538033975918e-08, "loss": 0.004, "step": 62191 }, { "epoch": 19.14, "learning_rate": 9.946526565093872e-08, "loss": 0.0029, "step": 62192 }, { "epoch": 19.14, "learning_rate": 9.939517554247713e-08, "loss": 0.0039, "step": 62193 }, { "epoch": 19.14, "learning_rate": 9.932511001454759e-08, "loss": 0.0025, "step": 62194 }, { "epoch": 19.14, "learning_rate": 9.925506906732441e-08, "loss": 0.0035, "step": 62195 }, { "epoch": 19.14, "learning_rate": 9.918505270098189e-08, "loss": 0.0025, "step": 62196 }, { "epoch": 19.14, "learning_rate": 9.911506091569434e-08, "loss": 0.0018, "step": 62197 }, { "epoch": 19.14, "learning_rate": 9.904509371163496e-08, "loss": 0.0032, "step": 62198 }, { "epoch": 19.14, "learning_rate": 9.897515108897692e-08, "loss": 0.0034, "step": 62199 }, { "epoch": 19.14, "learning_rate": 9.890523304789346e-08, "loss": 0.0042, "step": 62200 }, { "epoch": 19.14, "learning_rate": 9.883533958855995e-08, "loss": 0.004, "step": 62201 }, { "epoch": 19.14, "learning_rate": 9.876547071114961e-08, "loss": 0.0032, "step": 62202 }, { "epoch": 19.14, "learning_rate": 9.869562641583453e-08, "loss": 0.0046, "step": 62203 }, { "epoch": 19.14, "learning_rate": 9.862580670278899e-08, "loss": 0.0031, "step": 62204 }, { "epoch": 19.14, "learning_rate": 9.85560115721862e-08, "loss": 0.0043, "step": 62205 }, { "epoch": 19.14, "learning_rate": 9.848624102420046e-08, "loss": 0.0037, "step": 62206 }, { "epoch": 19.14, "learning_rate": 9.841649505900275e-08, "loss": 0.0028, "step": 62207 }, { "epoch": 19.14, "learning_rate": 9.834677367676848e-08, "loss": 0.0075, "step": 62208 }, { "epoch": 19.14, "learning_rate": 9.827707687766863e-08, "loss": 0.0026, "step": 62209 }, { "epoch": 19.14, "learning_rate": 9.82074046618775e-08, "loss": 0.004, "step": 62210 }, { "epoch": 19.14, "learning_rate": 9.81377570295694e-08, "loss": 0.0036, "step": 62211 }, { "epoch": 19.14, "learning_rate": 9.806813398091419e-08, "loss": 0.0038, "step": 62212 }, { "epoch": 19.14, "learning_rate": 9.799853551608729e-08, "loss": 0.0032, "step": 62213 }, { "epoch": 19.14, "learning_rate": 9.792896163526078e-08, "loss": 0.0037, "step": 62214 }, { "epoch": 19.14, "learning_rate": 9.785941233860563e-08, "loss": 0.0029, "step": 62215 }, { "epoch": 19.14, "learning_rate": 9.778988762629726e-08, "loss": 0.0022, "step": 62216 }, { "epoch": 19.14, "learning_rate": 9.772038749850665e-08, "loss": 0.0048, "step": 62217 }, { "epoch": 19.14, "learning_rate": 9.765091195540699e-08, "loss": 0.0038, "step": 62218 }, { "epoch": 19.14, "learning_rate": 9.758146099717036e-08, "loss": 0.0028, "step": 62219 }, { "epoch": 19.14, "learning_rate": 9.751203462396886e-08, "loss": 0.0023, "step": 62220 }, { "epoch": 19.14, "learning_rate": 9.744263283597566e-08, "loss": 0.0025, "step": 62221 }, { "epoch": 19.14, "learning_rate": 9.737325563336286e-08, "loss": 0.0029, "step": 62222 }, { "epoch": 19.15, "learning_rate": 9.730390301630365e-08, "loss": 0.003, "step": 62223 }, { "epoch": 19.15, "learning_rate": 9.723457498496791e-08, "loss": 0.0031, "step": 62224 }, { "epoch": 19.15, "learning_rate": 9.716527153952882e-08, "loss": 0.0032, "step": 62225 }, { "epoch": 19.15, "learning_rate": 9.709599268015956e-08, "loss": 0.0024, "step": 62226 }, { "epoch": 19.15, "learning_rate": 9.702673840703114e-08, "loss": 0.0022, "step": 62227 }, { "epoch": 19.15, "learning_rate": 9.69575087203145e-08, "loss": 0.0028, "step": 62228 }, { "epoch": 19.15, "learning_rate": 9.688830362018397e-08, "loss": 0.0021, "step": 62229 }, { "epoch": 19.15, "learning_rate": 9.68191231068083e-08, "loss": 0.003, "step": 62230 }, { "epoch": 19.15, "learning_rate": 9.674996718036178e-08, "loss": 0.0043, "step": 62231 }, { "epoch": 19.15, "learning_rate": 9.668083584101651e-08, "loss": 0.0033, "step": 62232 }, { "epoch": 19.15, "learning_rate": 9.661172908894123e-08, "loss": 0.0052, "step": 62233 }, { "epoch": 19.15, "learning_rate": 9.654264692431026e-08, "loss": 0.0026, "step": 62234 }, { "epoch": 19.15, "learning_rate": 9.647358934729233e-08, "loss": 0.0025, "step": 62235 }, { "epoch": 19.15, "learning_rate": 9.640455635806179e-08, "loss": 0.0039, "step": 62236 }, { "epoch": 19.15, "learning_rate": 9.633554795678957e-08, "loss": 0.0033, "step": 62237 }, { "epoch": 19.15, "learning_rate": 9.626656414364443e-08, "loss": 0.0028, "step": 62238 }, { "epoch": 19.15, "learning_rate": 9.619760491880182e-08, "loss": 0.0022, "step": 62239 }, { "epoch": 19.15, "learning_rate": 9.612867028242823e-08, "loss": 0.0024, "step": 62240 }, { "epoch": 19.15, "learning_rate": 9.605976023469798e-08, "loss": 0.003, "step": 62241 }, { "epoch": 19.15, "learning_rate": 9.599087477578207e-08, "loss": 0.0026, "step": 62242 }, { "epoch": 19.15, "learning_rate": 9.59220139058492e-08, "loss": 0.003, "step": 62243 }, { "epoch": 19.15, "learning_rate": 9.585317762507373e-08, "loss": 0.0041, "step": 62244 }, { "epoch": 19.15, "learning_rate": 9.578436593362328e-08, "loss": 0.0026, "step": 62245 }, { "epoch": 19.15, "learning_rate": 9.571557883167104e-08, "loss": 0.0051, "step": 62246 }, { "epoch": 19.15, "learning_rate": 9.564681631938688e-08, "loss": 0.0035, "step": 62247 }, { "epoch": 19.15, "learning_rate": 9.557807839694067e-08, "loss": 0.0034, "step": 62248 }, { "epoch": 19.15, "learning_rate": 9.550936506450448e-08, "loss": 0.003, "step": 62249 }, { "epoch": 19.15, "learning_rate": 9.544067632224818e-08, "loss": 0.0029, "step": 62250 }, { "epoch": 19.15, "learning_rate": 9.537201217034275e-08, "loss": 0.0022, "step": 62251 }, { "epoch": 19.15, "learning_rate": 9.530337260895806e-08, "loss": 0.0025, "step": 62252 }, { "epoch": 19.15, "learning_rate": 9.523475763826618e-08, "loss": 0.0028, "step": 62253 }, { "epoch": 19.15, "learning_rate": 9.516616725843475e-08, "loss": 0.0033, "step": 62254 }, { "epoch": 19.16, "learning_rate": 9.509760146963587e-08, "loss": 0.0021, "step": 62255 }, { "epoch": 19.16, "learning_rate": 9.50290602720394e-08, "loss": 0.0031, "step": 62256 }, { "epoch": 19.16, "learning_rate": 9.49605436658152e-08, "loss": 0.0038, "step": 62257 }, { "epoch": 19.16, "learning_rate": 9.489205165113424e-08, "loss": 0.0021, "step": 62258 }, { "epoch": 19.16, "learning_rate": 9.48235842281653e-08, "loss": 0.0022, "step": 62259 }, { "epoch": 19.16, "learning_rate": 9.47551413970793e-08, "loss": 0.0023, "step": 62260 }, { "epoch": 19.16, "learning_rate": 9.468672315804616e-08, "loss": 0.0047, "step": 62261 }, { "epoch": 19.16, "learning_rate": 9.461832951123573e-08, "loss": 0.0035, "step": 62262 }, { "epoch": 19.16, "learning_rate": 9.454996045681675e-08, "loss": 0.0034, "step": 62263 }, { "epoch": 19.16, "learning_rate": 9.44816159949613e-08, "loss": 0.0044, "step": 62264 }, { "epoch": 19.16, "learning_rate": 9.441329612583594e-08, "loss": 0.003, "step": 62265 }, { "epoch": 19.16, "learning_rate": 9.434500084961273e-08, "loss": 0.0033, "step": 62266 }, { "epoch": 19.16, "learning_rate": 9.427673016646044e-08, "loss": 0.003, "step": 62267 }, { "epoch": 19.16, "learning_rate": 9.420848407654781e-08, "loss": 0.0038, "step": 62268 }, { "epoch": 19.16, "learning_rate": 9.414026258004583e-08, "loss": 0.0029, "step": 62269 }, { "epoch": 19.16, "learning_rate": 9.407206567712213e-08, "loss": 0.0025, "step": 62270 }, { "epoch": 19.16, "learning_rate": 9.400389336794658e-08, "loss": 0.0064, "step": 62271 }, { "epoch": 19.16, "learning_rate": 9.393574565269015e-08, "loss": 0.0057, "step": 62272 }, { "epoch": 19.16, "learning_rate": 9.386762253152049e-08, "loss": 0.0042, "step": 62273 }, { "epoch": 19.16, "learning_rate": 9.379952400460635e-08, "loss": 0.0051, "step": 62274 }, { "epoch": 19.16, "learning_rate": 9.373145007211759e-08, "loss": 0.0036, "step": 62275 }, { "epoch": 19.16, "learning_rate": 9.366340073422297e-08, "loss": 0.0036, "step": 62276 }, { "epoch": 19.16, "learning_rate": 9.359537599109125e-08, "loss": 0.0058, "step": 62277 }, { "epoch": 19.16, "learning_rate": 9.352737584289229e-08, "loss": 0.0054, "step": 62278 }, { "epoch": 19.16, "learning_rate": 9.345940028979373e-08, "loss": 0.0033, "step": 62279 }, { "epoch": 19.16, "learning_rate": 9.339144933196432e-08, "loss": 0.0027, "step": 62280 }, { "epoch": 19.16, "learning_rate": 9.332352296957393e-08, "loss": 0.0013, "step": 62281 }, { "epoch": 19.16, "learning_rate": 9.32556212027902e-08, "loss": 0.0032, "step": 62282 }, { "epoch": 19.16, "learning_rate": 9.318774403178188e-08, "loss": 0.0037, "step": 62283 }, { "epoch": 19.16, "learning_rate": 9.311989145671662e-08, "loss": 0.0052, "step": 62284 }, { "epoch": 19.16, "learning_rate": 9.30520634777643e-08, "loss": 0.0038, "step": 62285 }, { "epoch": 19.16, "learning_rate": 9.298426009509365e-08, "loss": 0.0042, "step": 62286 }, { "epoch": 19.16, "learning_rate": 9.291648130887232e-08, "loss": 0.0067, "step": 62287 }, { "epoch": 19.17, "learning_rate": 9.284872711926907e-08, "loss": 0.0035, "step": 62288 }, { "epoch": 19.17, "learning_rate": 9.278099752645042e-08, "loss": 0.0035, "step": 62289 }, { "epoch": 19.17, "learning_rate": 9.271329253058514e-08, "loss": 0.0021, "step": 62290 }, { "epoch": 19.17, "learning_rate": 9.264561213184309e-08, "loss": 0.0031, "step": 62291 }, { "epoch": 19.17, "learning_rate": 9.257795633038968e-08, "loss": 0.0059, "step": 62292 }, { "epoch": 19.17, "learning_rate": 9.251032512639591e-08, "loss": 0.0036, "step": 62293 }, { "epoch": 19.17, "learning_rate": 9.244271852002718e-08, "loss": 0.0028, "step": 62294 }, { "epoch": 19.17, "learning_rate": 9.237513651145224e-08, "loss": 0.0036, "step": 62295 }, { "epoch": 19.17, "learning_rate": 9.230757910083876e-08, "loss": 0.0028, "step": 62296 }, { "epoch": 19.17, "learning_rate": 9.224004628835437e-08, "loss": 0.0045, "step": 62297 }, { "epoch": 19.17, "learning_rate": 9.21725380741667e-08, "loss": 0.0019, "step": 62298 }, { "epoch": 19.17, "learning_rate": 9.210505445844453e-08, "loss": 0.004, "step": 62299 }, { "epoch": 19.17, "learning_rate": 9.203759544135326e-08, "loss": 0.0043, "step": 62300 }, { "epoch": 19.17, "learning_rate": 9.197016102306277e-08, "loss": 0.004, "step": 62301 }, { "epoch": 19.17, "learning_rate": 9.190275120373849e-08, "loss": 0.0025, "step": 62302 }, { "epoch": 19.17, "learning_rate": 9.183536598354803e-08, "loss": 0.0033, "step": 62303 }, { "epoch": 19.17, "learning_rate": 9.176800536266017e-08, "loss": 0.0031, "step": 62304 }, { "epoch": 19.17, "learning_rate": 9.170066934124034e-08, "loss": 0.0046, "step": 62305 }, { "epoch": 19.17, "learning_rate": 9.163335791945726e-08, "loss": 0.0031, "step": 62306 }, { "epoch": 19.17, "learning_rate": 9.15660710974764e-08, "loss": 0.0022, "step": 62307 }, { "epoch": 19.17, "learning_rate": 9.149880887546759e-08, "loss": 0.0019, "step": 62308 }, { "epoch": 19.17, "learning_rate": 9.143157125359514e-08, "loss": 0.0034, "step": 62309 }, { "epoch": 19.17, "learning_rate": 9.13643582320256e-08, "loss": 0.0037, "step": 62310 }, { "epoch": 19.17, "learning_rate": 9.129716981092774e-08, "loss": 0.0041, "step": 62311 }, { "epoch": 19.17, "learning_rate": 9.123000599046805e-08, "loss": 0.0032, "step": 62312 }, { "epoch": 19.17, "learning_rate": 9.11628667708131e-08, "loss": 0.0037, "step": 62313 }, { "epoch": 19.17, "learning_rate": 9.10957521521294e-08, "loss": 0.0047, "step": 62314 }, { "epoch": 19.17, "learning_rate": 9.10286621345835e-08, "loss": 0.0023, "step": 62315 }, { "epoch": 19.17, "learning_rate": 9.096159671834193e-08, "loss": 0.0034, "step": 62316 }, { "epoch": 19.17, "learning_rate": 9.089455590357121e-08, "loss": 0.0037, "step": 62317 }, { "epoch": 19.17, "learning_rate": 9.08275396904379e-08, "loss": 0.0024, "step": 62318 }, { "epoch": 19.17, "learning_rate": 9.07605480791085e-08, "loss": 0.0043, "step": 62319 }, { "epoch": 19.18, "learning_rate": 9.069358106974846e-08, "loss": 0.0058, "step": 62320 }, { "epoch": 19.18, "learning_rate": 9.062663866252541e-08, "loss": 0.0023, "step": 62321 }, { "epoch": 19.18, "learning_rate": 9.055972085760589e-08, "loss": 0.003, "step": 62322 }, { "epoch": 19.18, "learning_rate": 9.049282765515421e-08, "loss": 0.0026, "step": 62323 }, { "epoch": 19.18, "learning_rate": 9.042595905533692e-08, "loss": 0.0022, "step": 62324 }, { "epoch": 19.18, "learning_rate": 9.035911505832051e-08, "loss": 0.0036, "step": 62325 }, { "epoch": 19.18, "learning_rate": 9.029229566427156e-08, "loss": 0.0016, "step": 62326 }, { "epoch": 19.18, "learning_rate": 9.022550087335435e-08, "loss": 0.0029, "step": 62327 }, { "epoch": 19.18, "learning_rate": 9.015873068573655e-08, "loss": 0.0024, "step": 62328 }, { "epoch": 19.18, "learning_rate": 9.009198510158357e-08, "loss": 0.0033, "step": 62329 }, { "epoch": 19.18, "learning_rate": 9.00252641210586e-08, "loss": 0.0039, "step": 62330 }, { "epoch": 19.18, "learning_rate": 8.995856774433042e-08, "loss": 0.0041, "step": 62331 }, { "epoch": 19.18, "learning_rate": 8.989189597156333e-08, "loss": 0.0063, "step": 62332 }, { "epoch": 19.18, "learning_rate": 8.982524880292277e-08, "loss": 0.0035, "step": 62333 }, { "epoch": 19.18, "learning_rate": 8.975862623857523e-08, "loss": 0.0064, "step": 62334 }, { "epoch": 19.18, "learning_rate": 8.969202827868395e-08, "loss": 0.0027, "step": 62335 }, { "epoch": 19.18, "learning_rate": 8.962545492341657e-08, "loss": 0.0025, "step": 62336 }, { "epoch": 19.18, "learning_rate": 8.955890617293738e-08, "loss": 0.0057, "step": 62337 }, { "epoch": 19.18, "learning_rate": 8.949238202741184e-08, "loss": 0.0042, "step": 62338 }, { "epoch": 19.18, "learning_rate": 8.942588248700423e-08, "loss": 0.0033, "step": 62339 }, { "epoch": 19.18, "learning_rate": 8.935940755188e-08, "loss": 0.002, "step": 62340 }, { "epoch": 19.18, "learning_rate": 8.929295722220565e-08, "loss": 0.0021, "step": 62341 }, { "epoch": 19.18, "learning_rate": 8.922653149814442e-08, "loss": 0.0065, "step": 62342 }, { "epoch": 19.18, "learning_rate": 8.916013037986282e-08, "loss": 0.0033, "step": 62343 }, { "epoch": 19.18, "learning_rate": 8.909375386752406e-08, "loss": 0.0035, "step": 62344 }, { "epoch": 19.18, "learning_rate": 8.902740196129244e-08, "loss": 0.0056, "step": 62345 }, { "epoch": 19.18, "learning_rate": 8.896107466133564e-08, "loss": 0.0055, "step": 62346 }, { "epoch": 19.18, "learning_rate": 8.889477196781571e-08, "loss": 0.0019, "step": 62347 }, { "epoch": 19.18, "learning_rate": 8.882849388089809e-08, "loss": 0.0032, "step": 62348 }, { "epoch": 19.18, "learning_rate": 8.87622404007471e-08, "loss": 0.0042, "step": 62349 }, { "epoch": 19.18, "learning_rate": 8.869601152752816e-08, "loss": 0.0028, "step": 62350 }, { "epoch": 19.18, "learning_rate": 8.862980726140446e-08, "loss": 0.0028, "step": 62351 }, { "epoch": 19.18, "learning_rate": 8.856362760254145e-08, "loss": 0.0025, "step": 62352 }, { "epoch": 19.19, "learning_rate": 8.849747255110341e-08, "loss": 0.0027, "step": 62353 }, { "epoch": 19.19, "learning_rate": 8.843134210725247e-08, "loss": 0.0033, "step": 62354 }, { "epoch": 19.19, "learning_rate": 8.836523627115623e-08, "loss": 0.0033, "step": 62355 }, { "epoch": 19.19, "learning_rate": 8.829915504297571e-08, "loss": 0.003, "step": 62356 }, { "epoch": 19.19, "learning_rate": 8.823309842287631e-08, "loss": 0.0029, "step": 62357 }, { "epoch": 19.19, "learning_rate": 8.816706641102235e-08, "loss": 0.0048, "step": 62358 }, { "epoch": 19.19, "learning_rate": 8.810105900757814e-08, "loss": 0.0035, "step": 62359 }, { "epoch": 19.19, "learning_rate": 8.80350762127058e-08, "loss": 0.0047, "step": 62360 }, { "epoch": 19.19, "learning_rate": 8.79691180265696e-08, "loss": 0.0043, "step": 62361 }, { "epoch": 19.19, "learning_rate": 8.790318444933499e-08, "loss": 0.0027, "step": 62362 }, { "epoch": 19.19, "learning_rate": 8.783727548116516e-08, "loss": 0.0027, "step": 62363 }, { "epoch": 19.19, "learning_rate": 8.777139112222221e-08, "loss": 0.0027, "step": 62364 }, { "epoch": 19.19, "learning_rate": 8.770553137267046e-08, "loss": 0.0015, "step": 62365 }, { "epoch": 19.19, "learning_rate": 8.763969623267421e-08, "loss": 0.0038, "step": 62366 }, { "epoch": 19.19, "learning_rate": 8.757388570239555e-08, "loss": 0.0043, "step": 62367 }, { "epoch": 19.19, "learning_rate": 8.750809978199881e-08, "loss": 0.0031, "step": 62368 }, { "epoch": 19.19, "learning_rate": 8.744233847164829e-08, "loss": 0.0024, "step": 62369 }, { "epoch": 19.19, "learning_rate": 8.737660177150498e-08, "loss": 0.0038, "step": 62370 }, { "epoch": 19.19, "learning_rate": 8.731088968173318e-08, "loss": 0.0063, "step": 62371 }, { "epoch": 19.19, "learning_rate": 8.72452022024961e-08, "loss": 0.0026, "step": 62372 }, { "epoch": 19.19, "learning_rate": 8.717953933395695e-08, "loss": 0.0033, "step": 62373 }, { "epoch": 19.19, "learning_rate": 8.711390107627892e-08, "loss": 0.0035, "step": 62374 }, { "epoch": 19.19, "learning_rate": 8.704828742962412e-08, "loss": 0.0032, "step": 62375 }, { "epoch": 19.19, "learning_rate": 8.698269839415685e-08, "loss": 0.0047, "step": 62376 }, { "epoch": 19.19, "learning_rate": 8.69171339700381e-08, "loss": 0.0026, "step": 62377 }, { "epoch": 19.19, "learning_rate": 8.685159415743216e-08, "loss": 0.0031, "step": 62378 }, { "epoch": 19.19, "learning_rate": 8.678607895650115e-08, "loss": 0.0037, "step": 62379 }, { "epoch": 19.19, "learning_rate": 8.672058836740716e-08, "loss": 0.0019, "step": 62380 }, { "epoch": 19.19, "learning_rate": 8.665512239031337e-08, "loss": 0.0027, "step": 62381 }, { "epoch": 19.19, "learning_rate": 8.658968102538301e-08, "loss": 0.0042, "step": 62382 }, { "epoch": 19.19, "learning_rate": 8.652426427277815e-08, "loss": 0.0039, "step": 62383 }, { "epoch": 19.19, "learning_rate": 8.64588721326609e-08, "loss": 0.0054, "step": 62384 }, { "epoch": 19.2, "learning_rate": 8.639350460519335e-08, "loss": 0.002, "step": 62385 }, { "epoch": 19.2, "learning_rate": 8.63281616905376e-08, "loss": 0.002, "step": 62386 }, { "epoch": 19.2, "learning_rate": 8.626284338885682e-08, "loss": 0.0027, "step": 62387 }, { "epoch": 19.2, "learning_rate": 8.619754970031314e-08, "loss": 0.004, "step": 62388 }, { "epoch": 19.2, "learning_rate": 8.613228062506752e-08, "loss": 0.0035, "step": 62389 }, { "epoch": 19.2, "learning_rate": 8.606703616328316e-08, "loss": 0.0028, "step": 62390 }, { "epoch": 19.2, "learning_rate": 8.600181631512217e-08, "loss": 0.003, "step": 62391 }, { "epoch": 19.2, "learning_rate": 8.593662108074552e-08, "loss": 0.0043, "step": 62392 }, { "epoch": 19.2, "learning_rate": 8.587145046031531e-08, "loss": 0.0043, "step": 62393 }, { "epoch": 19.2, "learning_rate": 8.580630445399363e-08, "loss": 0.0053, "step": 62394 }, { "epoch": 19.2, "learning_rate": 8.574118306194146e-08, "loss": 0.0042, "step": 62395 }, { "epoch": 19.2, "learning_rate": 8.5676086284322e-08, "loss": 0.0017, "step": 62396 }, { "epoch": 19.2, "learning_rate": 8.561101412129624e-08, "loss": 0.0047, "step": 62397 }, { "epoch": 19.2, "learning_rate": 8.554596657302516e-08, "loss": 0.004, "step": 62398 }, { "epoch": 19.2, "learning_rate": 8.548094363966974e-08, "loss": 0.002, "step": 62399 }, { "epoch": 19.2, "learning_rate": 8.541594532139208e-08, "loss": 0.0029, "step": 62400 }, { "epoch": 19.2, "learning_rate": 8.535097161835537e-08, "loss": 0.0027, "step": 62401 }, { "epoch": 19.2, "learning_rate": 8.528602253071838e-08, "loss": 0.005, "step": 62402 }, { "epoch": 19.2, "learning_rate": 8.52210980586432e-08, "loss": 0.0033, "step": 62403 }, { "epoch": 19.2, "learning_rate": 8.515619820229193e-08, "loss": 0.0034, "step": 62404 }, { "epoch": 19.2, "learning_rate": 8.509132296182332e-08, "loss": 0.004, "step": 62405 }, { "epoch": 19.2, "learning_rate": 8.502647233740169e-08, "loss": 0.0036, "step": 62406 }, { "epoch": 19.2, "learning_rate": 8.49616463291847e-08, "loss": 0.0023, "step": 62407 }, { "epoch": 19.2, "learning_rate": 8.489684493733664e-08, "loss": 0.0029, "step": 62408 }, { "epoch": 19.2, "learning_rate": 8.483206816201516e-08, "loss": 0.004, "step": 62409 }, { "epoch": 19.2, "learning_rate": 8.47673160033835e-08, "loss": 0.0027, "step": 62410 }, { "epoch": 19.2, "learning_rate": 8.470258846160151e-08, "loss": 0.002, "step": 62411 }, { "epoch": 19.2, "learning_rate": 8.463788553683017e-08, "loss": 0.004, "step": 62412 }, { "epoch": 19.2, "learning_rate": 8.457320722922935e-08, "loss": 0.0049, "step": 62413 }, { "epoch": 19.2, "learning_rate": 8.450855353896004e-08, "loss": 0.0031, "step": 62414 }, { "epoch": 19.2, "learning_rate": 8.44439244661821e-08, "loss": 0.0054, "step": 62415 }, { "epoch": 19.2, "learning_rate": 8.437932001105764e-08, "loss": 0.0043, "step": 62416 }, { "epoch": 19.2, "learning_rate": 8.431474017374653e-08, "loss": 0.0176, "step": 62417 }, { "epoch": 19.21, "learning_rate": 8.425018495440862e-08, "loss": 0.0033, "step": 62418 }, { "epoch": 19.21, "learning_rate": 8.41856543532038e-08, "loss": 0.0025, "step": 62419 }, { "epoch": 19.21, "learning_rate": 8.412114837029306e-08, "loss": 0.0038, "step": 62420 }, { "epoch": 19.21, "learning_rate": 8.405666700583626e-08, "loss": 0.0052, "step": 62421 }, { "epoch": 19.21, "learning_rate": 8.399221025999327e-08, "loss": 0.0031, "step": 62422 }, { "epoch": 19.21, "learning_rate": 8.392777813292396e-08, "loss": 0.0039, "step": 62423 }, { "epoch": 19.21, "learning_rate": 8.386337062478934e-08, "loss": 0.0052, "step": 62424 }, { "epoch": 19.21, "learning_rate": 8.379898773574924e-08, "loss": 0.0032, "step": 62425 }, { "epoch": 19.21, "learning_rate": 8.373462946596245e-08, "loss": 0.0016, "step": 62426 }, { "epoch": 19.21, "learning_rate": 8.367029581558883e-08, "loss": 0.0032, "step": 62427 }, { "epoch": 19.21, "learning_rate": 8.360598678478826e-08, "loss": 0.004, "step": 62428 }, { "epoch": 19.21, "learning_rate": 8.354170237372172e-08, "loss": 0.0026, "step": 62429 }, { "epoch": 19.21, "learning_rate": 8.347744258254686e-08, "loss": 0.0046, "step": 62430 }, { "epoch": 19.21, "learning_rate": 8.341320741142356e-08, "loss": 0.0031, "step": 62431 }, { "epoch": 19.21, "learning_rate": 8.33489968605139e-08, "loss": 0.0049, "step": 62432 }, { "epoch": 19.21, "learning_rate": 8.328481092997331e-08, "loss": 0.0074, "step": 62433 }, { "epoch": 19.21, "learning_rate": 8.32206496199639e-08, "loss": 0.003, "step": 62434 }, { "epoch": 19.21, "learning_rate": 8.31565129306433e-08, "loss": 0.0029, "step": 62435 }, { "epoch": 19.21, "learning_rate": 8.309240086217252e-08, "loss": 0.0041, "step": 62436 }, { "epoch": 19.21, "learning_rate": 8.302831341471029e-08, "loss": 0.0028, "step": 62437 }, { "epoch": 19.21, "learning_rate": 8.296425058841429e-08, "loss": 0.0056, "step": 62438 }, { "epoch": 19.21, "learning_rate": 8.290021238344547e-08, "loss": 0.0013, "step": 62439 }, { "epoch": 19.21, "learning_rate": 8.28361987999604e-08, "loss": 0.0053, "step": 62440 }, { "epoch": 19.21, "learning_rate": 8.277220983812007e-08, "loss": 0.0039, "step": 62441 }, { "epoch": 19.21, "learning_rate": 8.27082454980821e-08, "loss": 0.002, "step": 62442 }, { "epoch": 19.21, "learning_rate": 8.264430578000748e-08, "loss": 0.0018, "step": 62443 }, { "epoch": 19.21, "learning_rate": 8.258039068405277e-08, "loss": 0.004, "step": 62444 }, { "epoch": 19.21, "learning_rate": 8.251650021037671e-08, "loss": 0.0019, "step": 62445 }, { "epoch": 19.21, "learning_rate": 8.245263435913919e-08, "loss": 0.0038, "step": 62446 }, { "epoch": 19.21, "learning_rate": 8.238879313049785e-08, "loss": 0.0038, "step": 62447 }, { "epoch": 19.21, "learning_rate": 8.232497652461147e-08, "loss": 0.0027, "step": 62448 }, { "epoch": 19.21, "learning_rate": 8.226118454163767e-08, "loss": 0.0025, "step": 62449 }, { "epoch": 19.22, "learning_rate": 8.219741718173635e-08, "loss": 0.0029, "step": 62450 }, { "epoch": 19.22, "learning_rate": 8.213367444506515e-08, "loss": 0.0031, "step": 62451 }, { "epoch": 19.22, "learning_rate": 8.206995633178173e-08, "loss": 0.0024, "step": 62452 }, { "epoch": 19.22, "learning_rate": 8.200626284204593e-08, "loss": 0.0058, "step": 62453 }, { "epoch": 19.22, "learning_rate": 8.194259397601433e-08, "loss": 0.0021, "step": 62454 }, { "epoch": 19.22, "learning_rate": 8.187894973384458e-08, "loss": 0.0034, "step": 62455 }, { "epoch": 19.22, "learning_rate": 8.181533011569542e-08, "loss": 0.0032, "step": 62456 }, { "epoch": 19.22, "learning_rate": 8.175173512172674e-08, "loss": 0.0032, "step": 62457 }, { "epoch": 19.22, "learning_rate": 8.168816475209285e-08, "loss": 0.0046, "step": 62458 }, { "epoch": 19.22, "learning_rate": 8.162461900695474e-08, "loss": 0.004, "step": 62459 }, { "epoch": 19.22, "learning_rate": 8.156109788646782e-08, "loss": 0.0024, "step": 62460 }, { "epoch": 19.22, "learning_rate": 8.149760139079088e-08, "loss": 0.0031, "step": 62461 }, { "epoch": 19.22, "learning_rate": 8.143412952008156e-08, "loss": 0.0028, "step": 62462 }, { "epoch": 19.22, "learning_rate": 8.137068227449751e-08, "loss": 0.0035, "step": 62463 }, { "epoch": 19.22, "learning_rate": 8.130725965419528e-08, "loss": 0.0021, "step": 62464 }, { "epoch": 19.22, "learning_rate": 8.124386165933473e-08, "loss": 0.003, "step": 62465 }, { "epoch": 19.22, "learning_rate": 8.11804882900713e-08, "loss": 0.0039, "step": 62466 }, { "epoch": 19.22, "learning_rate": 8.111713954656153e-08, "loss": 0.0063, "step": 62467 }, { "epoch": 19.22, "learning_rate": 8.105381542896418e-08, "loss": 0.0059, "step": 62468 }, { "epoch": 19.22, "learning_rate": 8.099051593743578e-08, "loss": 0.0028, "step": 62469 }, { "epoch": 19.22, "learning_rate": 8.092724107213512e-08, "loss": 0.003, "step": 62470 }, { "epoch": 19.22, "learning_rate": 8.086399083321761e-08, "loss": 0.0039, "step": 62471 }, { "epoch": 19.22, "learning_rate": 8.08007652208398e-08, "loss": 0.0029, "step": 62472 }, { "epoch": 19.22, "learning_rate": 8.073756423516043e-08, "loss": 0.0024, "step": 62473 }, { "epoch": 19.22, "learning_rate": 8.067438787633608e-08, "loss": 0.0042, "step": 62474 }, { "epoch": 19.22, "learning_rate": 8.061123614452215e-08, "loss": 0.0018, "step": 62475 }, { "epoch": 19.22, "learning_rate": 8.054810903987631e-08, "loss": 0.0026, "step": 62476 }, { "epoch": 19.22, "learning_rate": 8.04850065625551e-08, "loss": 0.0021, "step": 62477 }, { "epoch": 19.22, "learning_rate": 8.042192871271615e-08, "loss": 0.0022, "step": 62478 }, { "epoch": 19.22, "learning_rate": 8.035887549051491e-08, "loss": 0.0018, "step": 62479 }, { "epoch": 19.22, "learning_rate": 8.029584689610903e-08, "loss": 0.0018, "step": 62480 }, { "epoch": 19.22, "learning_rate": 8.023284292965283e-08, "loss": 0.0033, "step": 62481 }, { "epoch": 19.22, "learning_rate": 8.016986359130507e-08, "loss": 0.0015, "step": 62482 }, { "epoch": 19.23, "learning_rate": 8.01069088812212e-08, "loss": 0.0032, "step": 62483 }, { "epoch": 19.23, "learning_rate": 8.004397879955771e-08, "loss": 0.0031, "step": 62484 }, { "epoch": 19.23, "learning_rate": 7.99810733464701e-08, "loss": 0.0028, "step": 62485 }, { "epoch": 19.23, "learning_rate": 7.991819252211596e-08, "loss": 0.0046, "step": 62486 }, { "epoch": 19.23, "learning_rate": 7.985533632665076e-08, "loss": 0.0044, "step": 62487 }, { "epoch": 19.23, "learning_rate": 7.979250476022993e-08, "loss": 0.0028, "step": 62488 }, { "epoch": 19.23, "learning_rate": 7.972969782300888e-08, "loss": 0.0035, "step": 62489 }, { "epoch": 19.23, "learning_rate": 7.966691551514527e-08, "loss": 0.0023, "step": 62490 }, { "epoch": 19.23, "learning_rate": 7.960415783679453e-08, "loss": 0.0032, "step": 62491 }, { "epoch": 19.23, "learning_rate": 7.95414247881121e-08, "loss": 0.0039, "step": 62492 }, { "epoch": 19.23, "learning_rate": 7.947871636925342e-08, "loss": 0.0031, "step": 62493 }, { "epoch": 19.23, "learning_rate": 7.9416032580375e-08, "loss": 0.0048, "step": 62494 }, { "epoch": 19.23, "learning_rate": 7.93533734216323e-08, "loss": 0.0019, "step": 62495 }, { "epoch": 19.23, "learning_rate": 7.929073889318073e-08, "loss": 0.0053, "step": 62496 }, { "epoch": 19.23, "learning_rate": 7.922812899517573e-08, "loss": 0.0041, "step": 62497 }, { "epoch": 19.23, "learning_rate": 7.916554372777274e-08, "loss": 0.0041, "step": 62498 }, { "epoch": 19.23, "learning_rate": 7.910298309112608e-08, "loss": 0.0043, "step": 62499 }, { "epoch": 19.23, "learning_rate": 7.904044708539338e-08, "loss": 0.0021, "step": 62500 }, { "epoch": 19.23, "learning_rate": 7.897793571072787e-08, "loss": 0.0034, "step": 62501 }, { "epoch": 19.23, "learning_rate": 7.89154489672861e-08, "loss": 0.0051, "step": 62502 }, { "epoch": 19.23, "learning_rate": 7.885298685522235e-08, "loss": 0.0053, "step": 62503 }, { "epoch": 19.23, "learning_rate": 7.879054937469099e-08, "loss": 0.0047, "step": 62504 }, { "epoch": 19.23, "learning_rate": 7.872813652584854e-08, "loss": 0.0034, "step": 62505 }, { "epoch": 19.23, "learning_rate": 7.866574830885043e-08, "loss": 0.0019, "step": 62506 }, { "epoch": 19.23, "learning_rate": 7.860338472384987e-08, "loss": 0.0045, "step": 62507 }, { "epoch": 19.23, "learning_rate": 7.854104577100341e-08, "loss": 0.0042, "step": 62508 }, { "epoch": 19.23, "learning_rate": 7.847873145046315e-08, "loss": 0.0025, "step": 62509 }, { "epoch": 19.23, "learning_rate": 7.841644176238561e-08, "loss": 0.0011, "step": 62510 }, { "epoch": 19.23, "learning_rate": 7.835417670692624e-08, "loss": 0.0025, "step": 62511 }, { "epoch": 19.23, "learning_rate": 7.829193628423825e-08, "loss": 0.0032, "step": 62512 }, { "epoch": 19.23, "learning_rate": 7.822972049447708e-08, "loss": 0.0036, "step": 62513 }, { "epoch": 19.23, "learning_rate": 7.816752933779592e-08, "loss": 0.003, "step": 62514 }, { "epoch": 19.24, "learning_rate": 7.810536281435022e-08, "loss": 0.0027, "step": 62515 }, { "epoch": 19.24, "learning_rate": 7.804322092429429e-08, "loss": 0.0041, "step": 62516 }, { "epoch": 19.24, "learning_rate": 7.798110366778134e-08, "loss": 0.0031, "step": 62517 }, { "epoch": 19.24, "learning_rate": 7.79190110449668e-08, "loss": 0.0026, "step": 62518 }, { "epoch": 19.24, "learning_rate": 7.7856943056005e-08, "loss": 0.0017, "step": 62519 }, { "epoch": 19.24, "learning_rate": 7.779489970104803e-08, "loss": 0.0026, "step": 62520 }, { "epoch": 19.24, "learning_rate": 7.773288098025356e-08, "loss": 0.0026, "step": 62521 }, { "epoch": 19.24, "learning_rate": 7.767088689377256e-08, "loss": 0.0035, "step": 62522 }, { "epoch": 19.24, "learning_rate": 7.760891744175936e-08, "loss": 0.0025, "step": 62523 }, { "epoch": 19.24, "learning_rate": 7.754697262436717e-08, "loss": 0.0039, "step": 62524 }, { "epoch": 19.24, "learning_rate": 7.748505244175252e-08, "loss": 0.0027, "step": 62525 }, { "epoch": 19.24, "learning_rate": 7.742315689406642e-08, "loss": 0.0031, "step": 62526 }, { "epoch": 19.24, "learning_rate": 7.73612859814643e-08, "loss": 0.0023, "step": 62527 }, { "epoch": 19.24, "learning_rate": 7.729943970409937e-08, "loss": 0.0046, "step": 62528 }, { "epoch": 19.24, "learning_rate": 7.723761806212371e-08, "loss": 0.003, "step": 62529 }, { "epoch": 19.24, "learning_rate": 7.717582105569276e-08, "loss": 0.0014, "step": 62530 }, { "epoch": 19.24, "learning_rate": 7.711404868495976e-08, "loss": 0.0034, "step": 62531 }, { "epoch": 19.24, "learning_rate": 7.705230095007677e-08, "loss": 0.0024, "step": 62532 }, { "epoch": 19.24, "learning_rate": 7.6990577851197e-08, "loss": 0.0036, "step": 62533 }, { "epoch": 19.24, "learning_rate": 7.692887938847482e-08, "loss": 0.0042, "step": 62534 }, { "epoch": 19.24, "learning_rate": 7.68672055620634e-08, "loss": 0.0021, "step": 62535 }, { "epoch": 19.24, "learning_rate": 7.680555637211596e-08, "loss": 0.0029, "step": 62536 }, { "epoch": 19.24, "learning_rate": 7.674393181878459e-08, "loss": 0.0029, "step": 62537 }, { "epoch": 19.24, "learning_rate": 7.668233190222251e-08, "loss": 0.0019, "step": 62538 }, { "epoch": 19.24, "learning_rate": 7.662075662258295e-08, "loss": 0.0033, "step": 62539 }, { "epoch": 19.24, "learning_rate": 7.655920598001909e-08, "loss": 0.0035, "step": 62540 }, { "epoch": 19.24, "learning_rate": 7.649767997468305e-08, "loss": 0.002, "step": 62541 }, { "epoch": 19.24, "learning_rate": 7.643617860672914e-08, "loss": 0.0045, "step": 62542 }, { "epoch": 19.24, "learning_rate": 7.637470187630836e-08, "loss": 0.0024, "step": 62543 }, { "epoch": 19.24, "learning_rate": 7.63132497835728e-08, "loss": 0.0022, "step": 62544 }, { "epoch": 19.24, "learning_rate": 7.625182232867789e-08, "loss": 0.0029, "step": 62545 }, { "epoch": 19.24, "learning_rate": 7.619041951177353e-08, "loss": 0.0121, "step": 62546 }, { "epoch": 19.24, "learning_rate": 7.61290413330118e-08, "loss": 0.0045, "step": 62547 }, { "epoch": 19.25, "learning_rate": 7.606768779254814e-08, "loss": 0.0041, "step": 62548 }, { "epoch": 19.25, "learning_rate": 7.600635889053243e-08, "loss": 0.0031, "step": 62549 }, { "epoch": 19.25, "learning_rate": 7.594505462711677e-08, "loss": 0.0036, "step": 62550 }, { "epoch": 19.25, "learning_rate": 7.588377500245437e-08, "loss": 0.0044, "step": 62551 }, { "epoch": 19.25, "learning_rate": 7.582252001669732e-08, "loss": 0.0042, "step": 62552 }, { "epoch": 19.25, "learning_rate": 7.576128966999774e-08, "loss": 0.0039, "step": 62553 }, { "epoch": 19.25, "learning_rate": 7.570008396250661e-08, "loss": 0.0038, "step": 62554 }, { "epoch": 19.25, "learning_rate": 7.563890289437825e-08, "loss": 0.0023, "step": 62555 }, { "epoch": 19.25, "learning_rate": 7.557774646576144e-08, "loss": 0.005, "step": 62556 }, { "epoch": 19.25, "learning_rate": 7.551661467681048e-08, "loss": 0.0036, "step": 62557 }, { "epoch": 19.25, "learning_rate": 7.545550752767528e-08, "loss": 0.0029, "step": 62558 }, { "epoch": 19.25, "learning_rate": 7.539442501850791e-08, "loss": 0.0027, "step": 62559 }, { "epoch": 19.25, "learning_rate": 7.53333671494616e-08, "loss": 0.003, "step": 62560 }, { "epoch": 19.25, "learning_rate": 7.527233392068734e-08, "loss": 0.0035, "step": 62561 }, { "epoch": 19.25, "learning_rate": 7.52113253323361e-08, "loss": 0.0044, "step": 62562 }, { "epoch": 19.25, "learning_rate": 7.515034138456001e-08, "loss": 0.0042, "step": 62563 }, { "epoch": 19.25, "learning_rate": 7.508938207750782e-08, "loss": 0.0029, "step": 62564 }, { "epoch": 19.25, "learning_rate": 7.502844741133497e-08, "loss": 0.0036, "step": 62565 }, { "epoch": 19.25, "learning_rate": 7.496753738619022e-08, "loss": 0.0031, "step": 62566 }, { "epoch": 19.25, "learning_rate": 7.490665200222569e-08, "loss": 0.0031, "step": 62567 }, { "epoch": 19.25, "learning_rate": 7.484579125959124e-08, "loss": 0.0028, "step": 62568 }, { "epoch": 19.25, "learning_rate": 7.478495515844008e-08, "loss": 0.0079, "step": 62569 }, { "epoch": 19.25, "learning_rate": 7.472414369892212e-08, "loss": 0.0028, "step": 62570 }, { "epoch": 19.25, "learning_rate": 7.46633568811872e-08, "loss": 0.0039, "step": 62571 }, { "epoch": 19.25, "learning_rate": 7.460259470538856e-08, "loss": 0.0041, "step": 62572 }, { "epoch": 19.25, "learning_rate": 7.454185717167495e-08, "loss": 0.0031, "step": 62573 }, { "epoch": 19.25, "learning_rate": 7.448114428019848e-08, "loss": 0.0042, "step": 62574 }, { "epoch": 19.25, "learning_rate": 7.442045603110903e-08, "loss": 0.0033, "step": 62575 }, { "epoch": 19.25, "learning_rate": 7.435979242455871e-08, "loss": 0.0023, "step": 62576 }, { "epoch": 19.25, "learning_rate": 7.429915346069738e-08, "loss": 0.0025, "step": 62577 }, { "epoch": 19.25, "learning_rate": 7.423853913967383e-08, "loss": 0.0027, "step": 62578 }, { "epoch": 19.25, "learning_rate": 7.417794946164125e-08, "loss": 0.0041, "step": 62579 }, { "epoch": 19.26, "learning_rate": 7.411738442674842e-08, "loss": 0.0026, "step": 62580 }, { "epoch": 19.26, "learning_rate": 7.405684403514635e-08, "loss": 0.0031, "step": 62581 }, { "epoch": 19.26, "learning_rate": 7.399632828698488e-08, "loss": 0.0029, "step": 62582 }, { "epoch": 19.26, "learning_rate": 7.393583718241504e-08, "loss": 0.0038, "step": 62583 }, { "epoch": 19.26, "learning_rate": 7.387537072158668e-08, "loss": 0.0018, "step": 62584 }, { "epoch": 19.26, "learning_rate": 7.381492890464858e-08, "loss": 0.0033, "step": 62585 }, { "epoch": 19.26, "learning_rate": 7.375451173175285e-08, "loss": 0.0031, "step": 62586 }, { "epoch": 19.26, "learning_rate": 7.369411920304937e-08, "loss": 0.0032, "step": 62587 }, { "epoch": 19.26, "learning_rate": 7.363375131868578e-08, "loss": 0.0023, "step": 62588 }, { "epoch": 19.26, "learning_rate": 7.35734080788142e-08, "loss": 0.0021, "step": 62589 }, { "epoch": 19.26, "learning_rate": 7.351308948358449e-08, "loss": 0.0039, "step": 62590 }, { "epoch": 19.26, "learning_rate": 7.345279553314543e-08, "loss": 0.0029, "step": 62591 }, { "epoch": 19.26, "learning_rate": 7.339252622764692e-08, "loss": 0.0032, "step": 62592 }, { "epoch": 19.26, "learning_rate": 7.33322815672377e-08, "loss": 0.0026, "step": 62593 }, { "epoch": 19.26, "learning_rate": 7.327206155206879e-08, "loss": 0.0037, "step": 62594 }, { "epoch": 19.26, "learning_rate": 7.321186618228893e-08, "loss": 0.0032, "step": 62595 }, { "epoch": 19.26, "learning_rate": 7.3151695458048e-08, "loss": 0.0029, "step": 62596 }, { "epoch": 19.26, "learning_rate": 7.309154937949592e-08, "loss": 0.003, "step": 62597 }, { "epoch": 19.26, "learning_rate": 7.303142794678141e-08, "loss": 0.0023, "step": 62598 }, { "epoch": 19.26, "learning_rate": 7.297133116005218e-08, "loss": 0.0024, "step": 62599 }, { "epoch": 19.26, "learning_rate": 7.291125901946027e-08, "loss": 0.0029, "step": 62600 }, { "epoch": 19.26, "learning_rate": 7.28512115251523e-08, "loss": 0.0046, "step": 62601 }, { "epoch": 19.26, "learning_rate": 7.27911886772792e-08, "loss": 0.0024, "step": 62602 }, { "epoch": 19.26, "learning_rate": 7.273119047598976e-08, "loss": 0.0019, "step": 62603 }, { "epoch": 19.26, "learning_rate": 7.267121692143164e-08, "loss": 0.0028, "step": 62604 }, { "epoch": 19.26, "learning_rate": 7.261126801375474e-08, "loss": 0.0035, "step": 62605 }, { "epoch": 19.26, "learning_rate": 7.255134375310779e-08, "loss": 0.0021, "step": 62606 }, { "epoch": 19.26, "learning_rate": 7.24914441396396e-08, "loss": 0.0025, "step": 62607 }, { "epoch": 19.26, "learning_rate": 7.243156917349892e-08, "loss": 0.0023, "step": 62608 }, { "epoch": 19.26, "learning_rate": 7.237171885483341e-08, "loss": 0.0028, "step": 62609 }, { "epoch": 19.26, "learning_rate": 7.231189318379294e-08, "loss": 0.0041, "step": 62610 }, { "epoch": 19.26, "learning_rate": 7.225209216052741e-08, "loss": 0.0043, "step": 62611 }, { "epoch": 19.26, "learning_rate": 7.219231578518115e-08, "loss": 0.0044, "step": 62612 }, { "epoch": 19.27, "learning_rate": 7.213256405790625e-08, "loss": 0.0015, "step": 62613 }, { "epoch": 19.27, "learning_rate": 7.207283697884815e-08, "loss": 0.0017, "step": 62614 }, { "epoch": 19.27, "learning_rate": 7.201313454815784e-08, "loss": 0.0039, "step": 62615 }, { "epoch": 19.27, "learning_rate": 7.195345676598298e-08, "loss": 0.0036, "step": 62616 }, { "epoch": 19.27, "learning_rate": 7.189380363247011e-08, "loss": 0.0035, "step": 62617 }, { "epoch": 19.27, "learning_rate": 7.183417514776914e-08, "loss": 0.0021, "step": 62618 }, { "epoch": 19.27, "learning_rate": 7.177457131202658e-08, "loss": 0.0037, "step": 62619 }, { "epoch": 19.27, "learning_rate": 7.171499212539124e-08, "loss": 0.0034, "step": 62620 }, { "epoch": 19.27, "learning_rate": 7.165543758801186e-08, "loss": 0.0026, "step": 62621 }, { "epoch": 19.27, "learning_rate": 7.159590770003499e-08, "loss": 0.0043, "step": 62622 }, { "epoch": 19.27, "learning_rate": 7.153640246160831e-08, "loss": 0.0029, "step": 62623 }, { "epoch": 19.27, "learning_rate": 7.147692187288058e-08, "loss": 0.0047, "step": 62624 }, { "epoch": 19.27, "learning_rate": 7.141746593399945e-08, "loss": 0.0023, "step": 62625 }, { "epoch": 19.27, "learning_rate": 7.135803464511149e-08, "loss": 0.0032, "step": 62626 }, { "epoch": 19.27, "learning_rate": 7.129862800636545e-08, "loss": 0.0031, "step": 62627 }, { "epoch": 19.27, "learning_rate": 7.123924601790788e-08, "loss": 0.0033, "step": 62628 }, { "epoch": 19.27, "learning_rate": 7.117988867988534e-08, "loss": 0.0038, "step": 62629 }, { "epoch": 19.27, "learning_rate": 7.112055599244772e-08, "loss": 0.0055, "step": 62630 }, { "epoch": 19.27, "learning_rate": 7.106124795574154e-08, "loss": 0.0043, "step": 62631 }, { "epoch": 19.27, "learning_rate": 7.100196456991227e-08, "loss": 0.0028, "step": 62632 }, { "epoch": 19.27, "learning_rate": 7.094270583510976e-08, "loss": 0.0024, "step": 62633 }, { "epoch": 19.27, "learning_rate": 7.088347175147835e-08, "loss": 0.0042, "step": 62634 }, { "epoch": 19.27, "learning_rate": 7.082426231916684e-08, "loss": 0.0029, "step": 62635 }, { "epoch": 19.27, "learning_rate": 7.076507753832173e-08, "loss": 0.0034, "step": 62636 }, { "epoch": 19.27, "learning_rate": 7.070591740909072e-08, "loss": 0.0045, "step": 62637 }, { "epoch": 19.27, "learning_rate": 7.064678193161923e-08, "loss": 0.0041, "step": 62638 }, { "epoch": 19.27, "learning_rate": 7.058767110605602e-08, "loss": 0.0026, "step": 62639 }, { "epoch": 19.27, "learning_rate": 7.052858493254545e-08, "loss": 0.0045, "step": 62640 }, { "epoch": 19.27, "learning_rate": 7.046952341123625e-08, "loss": 0.0054, "step": 62641 }, { "epoch": 19.27, "learning_rate": 7.0410486542275e-08, "loss": 0.0025, "step": 62642 }, { "epoch": 19.27, "learning_rate": 7.035147432580603e-08, "loss": 0.0043, "step": 62643 }, { "epoch": 19.27, "learning_rate": 7.029248676197809e-08, "loss": 0.0022, "step": 62644 }, { "epoch": 19.28, "learning_rate": 7.023352385093774e-08, "loss": 0.0049, "step": 62645 }, { "epoch": 19.28, "learning_rate": 7.01745855928293e-08, "loss": 0.0052, "step": 62646 }, { "epoch": 19.28, "learning_rate": 7.011567198780155e-08, "loss": 0.0029, "step": 62647 }, { "epoch": 19.28, "learning_rate": 7.005678303599883e-08, "loss": 0.0042, "step": 62648 }, { "epoch": 19.28, "learning_rate": 6.999791873756767e-08, "loss": 0.0041, "step": 62649 }, { "epoch": 19.28, "learning_rate": 6.993907909265463e-08, "loss": 0.0037, "step": 62650 }, { "epoch": 19.28, "learning_rate": 6.988026410140625e-08, "loss": 0.0032, "step": 62651 }, { "epoch": 19.28, "learning_rate": 6.982147376396908e-08, "loss": 0.0041, "step": 62652 }, { "epoch": 19.28, "learning_rate": 6.976270808048747e-08, "loss": 0.0028, "step": 62653 }, { "epoch": 19.28, "learning_rate": 6.970396705110683e-08, "loss": 0.0038, "step": 62654 }, { "epoch": 19.28, "learning_rate": 6.964525067597484e-08, "loss": 0.0024, "step": 62655 }, { "epoch": 19.28, "learning_rate": 6.958655895523692e-08, "loss": 0.0022, "step": 62656 }, { "epoch": 19.28, "learning_rate": 6.952789188903742e-08, "loss": 0.004, "step": 62657 }, { "epoch": 19.28, "learning_rate": 6.946924947752398e-08, "loss": 0.004, "step": 62658 }, { "epoch": 19.28, "learning_rate": 6.941063172084094e-08, "loss": 0.0029, "step": 62659 }, { "epoch": 19.28, "learning_rate": 6.935203861913376e-08, "loss": 0.0036, "step": 62660 }, { "epoch": 19.28, "learning_rate": 6.929347017254894e-08, "loss": 0.003, "step": 62661 }, { "epoch": 19.28, "learning_rate": 6.923492638123086e-08, "loss": 0.0049, "step": 62662 }, { "epoch": 19.28, "learning_rate": 6.917640724532493e-08, "loss": 0.0022, "step": 62663 }, { "epoch": 19.28, "learning_rate": 6.911791276497659e-08, "loss": 0.0045, "step": 62664 }, { "epoch": 19.28, "learning_rate": 6.90594429403324e-08, "loss": 0.0017, "step": 62665 }, { "epoch": 19.28, "learning_rate": 6.900099777153558e-08, "loss": 0.0036, "step": 62666 }, { "epoch": 19.28, "learning_rate": 6.894257725873155e-08, "loss": 0.0022, "step": 62667 }, { "epoch": 19.28, "learning_rate": 6.888418140206688e-08, "loss": 0.0041, "step": 62668 }, { "epoch": 19.28, "learning_rate": 6.882581020168366e-08, "loss": 0.0047, "step": 62669 }, { "epoch": 19.28, "learning_rate": 6.876746365772958e-08, "loss": 0.0027, "step": 62670 }, { "epoch": 19.28, "learning_rate": 6.870914177034893e-08, "loss": 0.0043, "step": 62671 }, { "epoch": 19.28, "learning_rate": 6.865084453968495e-08, "loss": 0.0031, "step": 62672 }, { "epoch": 19.28, "learning_rate": 6.859257196588532e-08, "loss": 0.0026, "step": 62673 }, { "epoch": 19.28, "learning_rate": 6.853432404909099e-08, "loss": 0.0024, "step": 62674 }, { "epoch": 19.28, "learning_rate": 6.847610078944966e-08, "loss": 0.0041, "step": 62675 }, { "epoch": 19.28, "learning_rate": 6.841790218710454e-08, "loss": 0.0022, "step": 62676 }, { "epoch": 19.28, "learning_rate": 6.835972824219994e-08, "loss": 0.0042, "step": 62677 }, { "epoch": 19.29, "learning_rate": 6.830157895488022e-08, "loss": 0.0024, "step": 62678 }, { "epoch": 19.29, "learning_rate": 6.824345432529078e-08, "loss": 0.0058, "step": 62679 }, { "epoch": 19.29, "learning_rate": 6.818535435357598e-08, "loss": 0.0031, "step": 62680 }, { "epoch": 19.29, "learning_rate": 6.812727903987903e-08, "loss": 0.0021, "step": 62681 }, { "epoch": 19.29, "learning_rate": 6.806922838434426e-08, "loss": 0.0019, "step": 62682 }, { "epoch": 19.29, "learning_rate": 6.8011202387116e-08, "loss": 0.0022, "step": 62683 }, { "epoch": 19.29, "learning_rate": 6.795320104833747e-08, "loss": 0.0022, "step": 62684 }, { "epoch": 19.29, "learning_rate": 6.78952243681541e-08, "loss": 0.0039, "step": 62685 }, { "epoch": 19.29, "learning_rate": 6.783727234671022e-08, "loss": 0.0057, "step": 62686 }, { "epoch": 19.29, "learning_rate": 6.777934498414907e-08, "loss": 0.0036, "step": 62687 }, { "epoch": 19.29, "learning_rate": 6.772144228061273e-08, "loss": 0.003, "step": 62688 }, { "epoch": 19.29, "learning_rate": 6.766356423624664e-08, "loss": 0.0025, "step": 62689 }, { "epoch": 19.29, "learning_rate": 6.760571085119405e-08, "loss": 0.0035, "step": 62690 }, { "epoch": 19.29, "learning_rate": 6.754788212559926e-08, "loss": 0.0047, "step": 62691 }, { "epoch": 19.29, "learning_rate": 6.74900780596055e-08, "loss": 0.0021, "step": 62692 }, { "epoch": 19.29, "learning_rate": 6.743229865335599e-08, "loss": 0.0024, "step": 62693 }, { "epoch": 19.29, "learning_rate": 6.737454390699505e-08, "loss": 0.003, "step": 62694 }, { "epoch": 19.29, "learning_rate": 6.731681382066369e-08, "loss": 0.0028, "step": 62695 }, { "epoch": 19.29, "learning_rate": 6.725910839450844e-08, "loss": 0.0017, "step": 62696 }, { "epoch": 19.29, "learning_rate": 6.720142762867032e-08, "loss": 0.0051, "step": 62697 }, { "epoch": 19.29, "learning_rate": 6.714377152329366e-08, "loss": 0.0036, "step": 62698 }, { "epoch": 19.29, "learning_rate": 6.708614007852055e-08, "loss": 0.0038, "step": 62699 }, { "epoch": 19.29, "learning_rate": 6.702853329449533e-08, "loss": 0.0035, "step": 62700 }, { "epoch": 19.29, "learning_rate": 6.69709511713601e-08, "loss": 0.0031, "step": 62701 }, { "epoch": 19.29, "learning_rate": 6.69133937092592e-08, "loss": 0.0016, "step": 62702 }, { "epoch": 19.29, "learning_rate": 6.685586090833252e-08, "loss": 0.003, "step": 62703 }, { "epoch": 19.29, "learning_rate": 6.679835276872549e-08, "loss": 0.003, "step": 62704 }, { "epoch": 19.29, "learning_rate": 6.674086929058133e-08, "loss": 0.0029, "step": 62705 }, { "epoch": 19.29, "learning_rate": 6.668341047404103e-08, "loss": 0.0023, "step": 62706 }, { "epoch": 19.29, "learning_rate": 6.662597631924783e-08, "loss": 0.0052, "step": 62707 }, { "epoch": 19.29, "learning_rate": 6.656856682634494e-08, "loss": 0.0033, "step": 62708 }, { "epoch": 19.29, "learning_rate": 6.651118199547224e-08, "loss": 0.0025, "step": 62709 }, { "epoch": 19.3, "learning_rate": 6.645382182677629e-08, "loss": 0.0056, "step": 62710 }, { "epoch": 19.3, "learning_rate": 6.639648632039697e-08, "loss": 0.0028, "step": 62711 }, { "epoch": 19.3, "learning_rate": 6.633917547647751e-08, "loss": 0.0037, "step": 62712 }, { "epoch": 19.3, "learning_rate": 6.628188929515888e-08, "loss": 0.004, "step": 62713 }, { "epoch": 19.3, "learning_rate": 6.622462777658545e-08, "loss": 0.0028, "step": 62714 }, { "epoch": 19.3, "learning_rate": 6.616739092089708e-08, "loss": 0.0043, "step": 62715 }, { "epoch": 19.3, "learning_rate": 6.611017872823811e-08, "loss": 0.0023, "step": 62716 }, { "epoch": 19.3, "learning_rate": 6.605299119874953e-08, "loss": 0.0042, "step": 62717 }, { "epoch": 19.3, "learning_rate": 6.599582833257235e-08, "loss": 0.0031, "step": 62718 }, { "epoch": 19.3, "learning_rate": 6.593869012984866e-08, "loss": 0.0028, "step": 62719 }, { "epoch": 19.3, "learning_rate": 6.588157659072281e-08, "loss": 0.0025, "step": 62720 }, { "epoch": 19.3, "learning_rate": 6.582448771533468e-08, "loss": 0.0028, "step": 62721 }, { "epoch": 19.3, "learning_rate": 6.576742350382525e-08, "loss": 0.003, "step": 62722 }, { "epoch": 19.3, "learning_rate": 6.571038395633778e-08, "loss": 0.0024, "step": 62723 }, { "epoch": 19.3, "learning_rate": 6.565336907301212e-08, "loss": 0.0039, "step": 62724 }, { "epoch": 19.3, "learning_rate": 6.559637885399262e-08, "loss": 0.0041, "step": 62725 }, { "epoch": 19.3, "learning_rate": 6.553941329941804e-08, "loss": 0.0051, "step": 62726 }, { "epoch": 19.3, "learning_rate": 6.548247240943051e-08, "loss": 0.0028, "step": 62727 }, { "epoch": 19.3, "learning_rate": 6.542555618417212e-08, "loss": 0.0034, "step": 62728 }, { "epoch": 19.3, "learning_rate": 6.536866462378499e-08, "loss": 0.0059, "step": 62729 }, { "epoch": 19.3, "learning_rate": 6.53117977284079e-08, "loss": 0.0031, "step": 62730 }, { "epoch": 19.3, "learning_rate": 6.525495549818295e-08, "loss": 0.0051, "step": 62731 }, { "epoch": 19.3, "learning_rate": 6.519813793325225e-08, "loss": 0.0048, "step": 62732 }, { "epoch": 19.3, "learning_rate": 6.51413450337568e-08, "loss": 0.0071, "step": 62733 }, { "epoch": 19.3, "learning_rate": 6.508457679983648e-08, "loss": 0.0028, "step": 62734 }, { "epoch": 19.3, "learning_rate": 6.502783323163231e-08, "loss": 0.0034, "step": 62735 }, { "epoch": 19.3, "learning_rate": 6.497111432928638e-08, "loss": 0.004, "step": 62736 }, { "epoch": 19.3, "learning_rate": 6.491442009293858e-08, "loss": 0.0043, "step": 62737 }, { "epoch": 19.3, "learning_rate": 6.485775052272991e-08, "loss": 0.0027, "step": 62738 }, { "epoch": 19.3, "learning_rate": 6.480110561880026e-08, "loss": 0.0042, "step": 62739 }, { "epoch": 19.3, "learning_rate": 6.474448538129175e-08, "loss": 0.0064, "step": 62740 }, { "epoch": 19.3, "learning_rate": 6.468788981034424e-08, "loss": 0.0052, "step": 62741 }, { "epoch": 19.3, "learning_rate": 6.463131890609875e-08, "loss": 0.0024, "step": 62742 }, { "epoch": 19.31, "learning_rate": 6.457477266869405e-08, "loss": 0.0036, "step": 62743 }, { "epoch": 19.31, "learning_rate": 6.451825109827114e-08, "loss": 0.0055, "step": 62744 }, { "epoch": 19.31, "learning_rate": 6.446175419497214e-08, "loss": 0.0027, "step": 62745 }, { "epoch": 19.31, "learning_rate": 6.440528195893581e-08, "loss": 0.0025, "step": 62746 }, { "epoch": 19.31, "learning_rate": 6.434883439030316e-08, "loss": 0.0031, "step": 62747 }, { "epoch": 19.31, "learning_rate": 6.429241148921294e-08, "loss": 0.0037, "step": 62748 }, { "epoch": 19.31, "learning_rate": 6.42360132558062e-08, "loss": 0.0031, "step": 62749 }, { "epoch": 19.31, "learning_rate": 6.417963969022389e-08, "loss": 0.0039, "step": 62750 }, { "epoch": 19.31, "learning_rate": 6.412329079260371e-08, "loss": 0.0028, "step": 62751 }, { "epoch": 19.31, "learning_rate": 6.406696656308664e-08, "loss": 0.0033, "step": 62752 }, { "epoch": 19.31, "learning_rate": 6.401066700181368e-08, "loss": 0.0031, "step": 62753 }, { "epoch": 19.31, "learning_rate": 6.395439210892252e-08, "loss": 0.0038, "step": 62754 }, { "epoch": 19.31, "learning_rate": 6.389814188455523e-08, "loss": 0.0033, "step": 62755 }, { "epoch": 19.31, "learning_rate": 6.38419163288495e-08, "loss": 0.002, "step": 62756 }, { "epoch": 19.31, "learning_rate": 6.378571544194522e-08, "loss": 0.0027, "step": 62757 }, { "epoch": 19.31, "learning_rate": 6.372953922398228e-08, "loss": 0.0027, "step": 62758 }, { "epoch": 19.31, "learning_rate": 6.367338767509945e-08, "loss": 0.0039, "step": 62759 }, { "epoch": 19.31, "learning_rate": 6.361726079543884e-08, "loss": 0.0026, "step": 62760 }, { "epoch": 19.31, "learning_rate": 6.356115858513589e-08, "loss": 0.0043, "step": 62761 }, { "epoch": 19.31, "learning_rate": 6.350508104433273e-08, "loss": 0.0038, "step": 62762 }, { "epoch": 19.31, "learning_rate": 6.344902817316811e-08, "loss": 0.005, "step": 62763 }, { "epoch": 19.31, "learning_rate": 6.339299997177972e-08, "loss": 0.0014, "step": 62764 }, { "epoch": 19.31, "learning_rate": 6.333699644030855e-08, "loss": 0.0023, "step": 62765 }, { "epoch": 19.31, "learning_rate": 6.328101757889227e-08, "loss": 0.0033, "step": 62766 }, { "epoch": 19.31, "learning_rate": 6.322506338767076e-08, "loss": 0.0025, "step": 62767 }, { "epoch": 19.31, "learning_rate": 6.316913386678281e-08, "loss": 0.003, "step": 62768 }, { "epoch": 19.31, "learning_rate": 6.311322901636607e-08, "loss": 0.0027, "step": 62769 }, { "epoch": 19.31, "learning_rate": 6.305734883656046e-08, "loss": 0.0028, "step": 62770 }, { "epoch": 19.31, "learning_rate": 6.300149332750472e-08, "loss": 0.0037, "step": 62771 }, { "epoch": 19.31, "learning_rate": 6.294566248933765e-08, "loss": 0.0026, "step": 62772 }, { "epoch": 19.31, "learning_rate": 6.28898563221969e-08, "loss": 0.0042, "step": 62773 }, { "epoch": 19.31, "learning_rate": 6.283407482622239e-08, "loss": 0.0027, "step": 62774 }, { "epoch": 19.32, "learning_rate": 6.277831800155176e-08, "loss": 0.0042, "step": 62775 }, { "epoch": 19.32, "learning_rate": 6.27225858483238e-08, "loss": 0.0037, "step": 62776 }, { "epoch": 19.32, "learning_rate": 6.266687836667618e-08, "loss": 0.0045, "step": 62777 }, { "epoch": 19.32, "learning_rate": 6.261119555674766e-08, "loss": 0.0044, "step": 62778 }, { "epoch": 19.32, "learning_rate": 6.255553741867593e-08, "loss": 0.0049, "step": 62779 }, { "epoch": 19.32, "learning_rate": 6.249990395260086e-08, "loss": 0.0023, "step": 62780 }, { "epoch": 19.32, "learning_rate": 6.244429515865902e-08, "loss": 0.0058, "step": 62781 }, { "epoch": 19.32, "learning_rate": 6.238871103698918e-08, "loss": 0.0027, "step": 62782 }, { "epoch": 19.32, "learning_rate": 6.23331515877279e-08, "loss": 0.0024, "step": 62783 }, { "epoch": 19.32, "learning_rate": 6.227761681101619e-08, "loss": 0.0028, "step": 62784 }, { "epoch": 19.32, "learning_rate": 6.222210670698836e-08, "loss": 0.0031, "step": 62785 }, { "epoch": 19.32, "learning_rate": 6.216662127578432e-08, "loss": 0.004, "step": 62786 }, { "epoch": 19.32, "learning_rate": 6.211116051754174e-08, "loss": 0.004, "step": 62787 }, { "epoch": 19.32, "learning_rate": 6.205572443239826e-08, "loss": 0.0081, "step": 62788 }, { "epoch": 19.32, "learning_rate": 6.200031302049048e-08, "loss": 0.0024, "step": 62789 }, { "epoch": 19.32, "learning_rate": 6.194492628195714e-08, "loss": 0.0055, "step": 62790 }, { "epoch": 19.32, "learning_rate": 6.188956421693592e-08, "loss": 0.0063, "step": 62791 }, { "epoch": 19.32, "learning_rate": 6.18342268255634e-08, "loss": 0.0045, "step": 62792 }, { "epoch": 19.32, "learning_rate": 6.17789141079761e-08, "loss": 0.0031, "step": 62793 }, { "epoch": 19.32, "learning_rate": 6.172362606431281e-08, "loss": 0.0026, "step": 62794 }, { "epoch": 19.32, "learning_rate": 6.166836269471122e-08, "loss": 0.0052, "step": 62795 }, { "epoch": 19.32, "learning_rate": 6.161312399930785e-08, "loss": 0.0037, "step": 62796 }, { "epoch": 19.32, "learning_rate": 6.15579099782404e-08, "loss": 0.0024, "step": 62797 }, { "epoch": 19.32, "learning_rate": 6.150272063164431e-08, "loss": 0.0038, "step": 62798 }, { "epoch": 19.32, "learning_rate": 6.144755595965724e-08, "loss": 0.0033, "step": 62799 }, { "epoch": 19.32, "learning_rate": 6.139241596241686e-08, "loss": 0.0035, "step": 62800 }, { "epoch": 19.32, "learning_rate": 6.133730064005971e-08, "loss": 0.0021, "step": 62801 }, { "epoch": 19.32, "learning_rate": 6.12822099927235e-08, "loss": 0.0034, "step": 62802 }, { "epoch": 19.32, "learning_rate": 6.122714402054252e-08, "loss": 0.0043, "step": 62803 }, { "epoch": 19.32, "learning_rate": 6.11721027236567e-08, "loss": 0.0054, "step": 62804 }, { "epoch": 19.32, "learning_rate": 6.111708610220035e-08, "loss": 0.0023, "step": 62805 }, { "epoch": 19.32, "learning_rate": 6.106209415631115e-08, "loss": 0.0036, "step": 62806 }, { "epoch": 19.32, "learning_rate": 6.100712688612564e-08, "loss": 0.0024, "step": 62807 }, { "epoch": 19.33, "learning_rate": 6.095218429177929e-08, "loss": 0.0029, "step": 62808 }, { "epoch": 19.33, "learning_rate": 6.089726637340864e-08, "loss": 0.0024, "step": 62809 }, { "epoch": 19.33, "learning_rate": 6.084237313115249e-08, "loss": 0.0035, "step": 62810 }, { "epoch": 19.33, "learning_rate": 6.078750456514404e-08, "loss": 0.0033, "step": 62811 }, { "epoch": 19.33, "learning_rate": 6.073266067552097e-08, "loss": 0.0033, "step": 62812 }, { "epoch": 19.33, "learning_rate": 6.067784146241873e-08, "loss": 0.0058, "step": 62813 }, { "epoch": 19.33, "learning_rate": 6.062304692597387e-08, "loss": 0.0023, "step": 62814 }, { "epoch": 19.33, "learning_rate": 6.056827706632185e-08, "loss": 0.0029, "step": 62815 }, { "epoch": 19.33, "learning_rate": 6.051353188360031e-08, "loss": 0.0031, "step": 62816 }, { "epoch": 19.33, "learning_rate": 6.045881137794363e-08, "loss": 0.0026, "step": 62817 }, { "epoch": 19.33, "learning_rate": 6.040411554948943e-08, "loss": 0.003, "step": 62818 }, { "epoch": 19.33, "learning_rate": 6.034944439836987e-08, "loss": 0.004, "step": 62819 }, { "epoch": 19.33, "learning_rate": 6.02947979247237e-08, "loss": 0.0027, "step": 62820 }, { "epoch": 19.33, "learning_rate": 6.024017612868637e-08, "loss": 0.0022, "step": 62821 }, { "epoch": 19.33, "learning_rate": 6.018557901039335e-08, "loss": 0.0038, "step": 62822 }, { "epoch": 19.33, "learning_rate": 6.013100656997895e-08, "loss": 0.0023, "step": 62823 }, { "epoch": 19.33, "learning_rate": 6.007645880758083e-08, "loss": 0.0028, "step": 62824 }, { "epoch": 19.33, "learning_rate": 6.002193572333226e-08, "loss": 0.0039, "step": 62825 }, { "epoch": 19.33, "learning_rate": 5.996743731736865e-08, "loss": 0.0028, "step": 62826 }, { "epoch": 19.33, "learning_rate": 5.991296358982767e-08, "loss": 0.0038, "step": 62827 }, { "epoch": 19.33, "learning_rate": 5.985851454084257e-08, "loss": 0.0029, "step": 62828 }, { "epoch": 19.33, "learning_rate": 5.980409017054767e-08, "loss": 0.0037, "step": 62829 }, { "epoch": 19.33, "learning_rate": 5.974969047908063e-08, "loss": 0.0035, "step": 62830 }, { "epoch": 19.33, "learning_rate": 5.969531546657581e-08, "loss": 0.0061, "step": 62831 }, { "epoch": 19.33, "learning_rate": 5.964096513316753e-08, "loss": 0.0038, "step": 62832 }, { "epoch": 19.33, "learning_rate": 5.958663947899013e-08, "loss": 0.0032, "step": 62833 }, { "epoch": 19.33, "learning_rate": 5.953233850417905e-08, "loss": 0.0025, "step": 62834 }, { "epoch": 19.33, "learning_rate": 5.9478062208870865e-08, "loss": 0.0029, "step": 62835 }, { "epoch": 19.33, "learning_rate": 5.942381059319768e-08, "loss": 0.0026, "step": 62836 }, { "epoch": 19.33, "learning_rate": 5.936958365729495e-08, "loss": 0.0019, "step": 62837 }, { "epoch": 19.33, "learning_rate": 5.931538140129811e-08, "loss": 0.0049, "step": 62838 }, { "epoch": 19.33, "learning_rate": 5.9261203825341504e-08, "loss": 0.0025, "step": 62839 }, { "epoch": 19.34, "learning_rate": 5.920705092955836e-08, "loss": 0.0041, "step": 62840 }, { "epoch": 19.34, "learning_rate": 5.915292271408524e-08, "loss": 0.0022, "step": 62841 }, { "epoch": 19.34, "learning_rate": 5.909881917905536e-08, "loss": 0.003, "step": 62842 }, { "epoch": 19.34, "learning_rate": 5.9044740324603057e-08, "loss": 0.0033, "step": 62843 }, { "epoch": 19.34, "learning_rate": 5.8990686150861565e-08, "loss": 0.0033, "step": 62844 }, { "epoch": 19.34, "learning_rate": 5.893665665796744e-08, "loss": 0.0024, "step": 62845 }, { "epoch": 19.34, "learning_rate": 5.8882651846053906e-08, "loss": 0.0025, "step": 62846 }, { "epoch": 19.34, "learning_rate": 5.8828671715253085e-08, "loss": 0.0026, "step": 62847 }, { "epoch": 19.34, "learning_rate": 5.877471626570153e-08, "loss": 0.0035, "step": 62848 }, { "epoch": 19.34, "learning_rate": 5.8720785497532464e-08, "loss": 0.0033, "step": 62849 }, { "epoch": 19.34, "learning_rate": 5.8666879410879116e-08, "loss": 0.0037, "step": 62850 }, { "epoch": 19.34, "learning_rate": 5.861299800587583e-08, "loss": 0.003, "step": 62851 }, { "epoch": 19.34, "learning_rate": 5.855914128265805e-08, "loss": 0.0042, "step": 62852 }, { "epoch": 19.34, "learning_rate": 5.850530924135567e-08, "loss": 0.0029, "step": 62853 }, { "epoch": 19.34, "learning_rate": 5.845150188210524e-08, "loss": 0.0028, "step": 62854 }, { "epoch": 19.34, "learning_rate": 5.8397719205038894e-08, "loss": 0.0033, "step": 62855 }, { "epoch": 19.34, "learning_rate": 5.834396121029206e-08, "loss": 0.0027, "step": 62856 }, { "epoch": 19.34, "learning_rate": 5.829022789799577e-08, "loss": 0.0029, "step": 62857 }, { "epoch": 19.34, "learning_rate": 5.823651926828544e-08, "loss": 0.0029, "step": 62858 }, { "epoch": 19.34, "learning_rate": 5.8182835321293205e-08, "loss": 0.003, "step": 62859 }, { "epoch": 19.34, "learning_rate": 5.81291760571534e-08, "loss": 0.0021, "step": 62860 }, { "epoch": 19.34, "learning_rate": 5.8075541475998145e-08, "loss": 0.0033, "step": 62861 }, { "epoch": 19.34, "learning_rate": 5.8021931577961764e-08, "loss": 0.003, "step": 62862 }, { "epoch": 19.34, "learning_rate": 5.7968346363175274e-08, "loss": 0.0037, "step": 62863 }, { "epoch": 19.34, "learning_rate": 5.7914785831773014e-08, "loss": 0.0049, "step": 62864 }, { "epoch": 19.34, "learning_rate": 5.786124998388931e-08, "loss": 0.0017, "step": 62865 }, { "epoch": 19.34, "learning_rate": 5.7807738819656286e-08, "loss": 0.0049, "step": 62866 }, { "epoch": 19.34, "learning_rate": 5.7754252339204955e-08, "loss": 0.0024, "step": 62867 }, { "epoch": 19.34, "learning_rate": 5.770079054266964e-08, "loss": 0.0024, "step": 62868 }, { "epoch": 19.34, "learning_rate": 5.764735343018246e-08, "loss": 0.0019, "step": 62869 }, { "epoch": 19.34, "learning_rate": 5.7593941001877763e-08, "loss": 0.0028, "step": 62870 }, { "epoch": 19.34, "learning_rate": 5.7540553257886545e-08, "loss": 0.005, "step": 62871 }, { "epoch": 19.34, "learning_rate": 5.748719019834092e-08, "loss": 0.0022, "step": 62872 }, { "epoch": 19.35, "learning_rate": 5.743385182337524e-08, "loss": 0.004, "step": 62873 }, { "epoch": 19.35, "learning_rate": 5.73805381331205e-08, "loss": 0.0044, "step": 62874 }, { "epoch": 19.35, "learning_rate": 5.7327249127708816e-08, "loss": 0.0017, "step": 62875 }, { "epoch": 19.35, "learning_rate": 5.7273984807274526e-08, "loss": 0.0024, "step": 62876 }, { "epoch": 19.35, "learning_rate": 5.722074517194753e-08, "loss": 0.0028, "step": 62877 }, { "epoch": 19.35, "learning_rate": 5.716753022185995e-08, "loss": 0.003, "step": 62878 }, { "epoch": 19.35, "learning_rate": 5.71143399571461e-08, "loss": 0.006, "step": 62879 }, { "epoch": 19.35, "learning_rate": 5.7061174377937015e-08, "loss": 0.0051, "step": 62880 }, { "epoch": 19.35, "learning_rate": 5.700803348436479e-08, "loss": 0.0024, "step": 62881 }, { "epoch": 19.35, "learning_rate": 5.695491727656044e-08, "loss": 0.0025, "step": 62882 }, { "epoch": 19.35, "learning_rate": 5.690182575465608e-08, "loss": 0.0032, "step": 62883 }, { "epoch": 19.35, "learning_rate": 5.684875891878383e-08, "loss": 0.003, "step": 62884 }, { "epoch": 19.35, "learning_rate": 5.6795716769075803e-08, "loss": 0.0032, "step": 62885 }, { "epoch": 19.35, "learning_rate": 5.6742699305664115e-08, "loss": 0.0041, "step": 62886 }, { "epoch": 19.35, "learning_rate": 5.668970652867867e-08, "loss": 0.0029, "step": 62887 }, { "epoch": 19.35, "learning_rate": 5.663673843825268e-08, "loss": 0.004, "step": 62888 }, { "epoch": 19.35, "learning_rate": 5.6583795034517165e-08, "loss": 0.0026, "step": 62889 }, { "epoch": 19.35, "learning_rate": 5.653087631760312e-08, "loss": 0.003, "step": 62890 }, { "epoch": 19.35, "learning_rate": 5.647798228764156e-08, "loss": 0.004, "step": 62891 }, { "epoch": 19.35, "learning_rate": 5.642511294476571e-08, "loss": 0.004, "step": 62892 }, { "epoch": 19.35, "learning_rate": 5.637226828910436e-08, "loss": 0.0019, "step": 62893 }, { "epoch": 19.35, "learning_rate": 5.6319448320790727e-08, "loss": 0.0044, "step": 62894 }, { "epoch": 19.35, "learning_rate": 5.6266653039955823e-08, "loss": 0.0022, "step": 62895 }, { "epoch": 19.35, "learning_rate": 5.621388244672954e-08, "loss": 0.0036, "step": 62896 }, { "epoch": 19.35, "learning_rate": 5.616113654124289e-08, "loss": 0.0037, "step": 62897 }, { "epoch": 19.35, "learning_rate": 5.6108415323626876e-08, "loss": 0.0039, "step": 62898 }, { "epoch": 19.35, "learning_rate": 5.6055718794013613e-08, "loss": 0.0031, "step": 62899 }, { "epoch": 19.35, "learning_rate": 5.6003046952534114e-08, "loss": 0.0044, "step": 62900 }, { "epoch": 19.35, "learning_rate": 5.595039979931716e-08, "loss": 0.0043, "step": 62901 }, { "epoch": 19.35, "learning_rate": 5.589777733449486e-08, "loss": 0.0025, "step": 62902 }, { "epoch": 19.35, "learning_rate": 5.584517955819713e-08, "loss": 0.0035, "step": 62903 }, { "epoch": 19.35, "learning_rate": 5.5792606470554955e-08, "loss": 0.0061, "step": 62904 }, { "epoch": 19.36, "learning_rate": 5.5740058071699354e-08, "loss": 0.0024, "step": 62905 }, { "epoch": 19.36, "learning_rate": 5.568753436176022e-08, "loss": 0.0019, "step": 62906 }, { "epoch": 19.36, "learning_rate": 5.563503534086745e-08, "loss": 0.004, "step": 62907 }, { "epoch": 19.36, "learning_rate": 5.5582561009153155e-08, "loss": 0.0028, "step": 62908 }, { "epoch": 19.36, "learning_rate": 5.5530111366745024e-08, "loss": 0.0057, "step": 62909 }, { "epoch": 19.36, "learning_rate": 5.5477686413776264e-08, "loss": 0.004, "step": 62910 }, { "epoch": 19.36, "learning_rate": 5.5425286150374567e-08, "loss": 0.0025, "step": 62911 }, { "epoch": 19.36, "learning_rate": 5.5372910576670936e-08, "loss": 0.0041, "step": 62912 }, { "epoch": 19.36, "learning_rate": 5.532055969279526e-08, "loss": 0.0044, "step": 62913 }, { "epoch": 19.36, "learning_rate": 5.5268233498877445e-08, "loss": 0.0056, "step": 62914 }, { "epoch": 19.36, "learning_rate": 5.521593199504738e-08, "loss": 0.0053, "step": 62915 }, { "epoch": 19.36, "learning_rate": 5.516365518143607e-08, "loss": 0.0024, "step": 62916 }, { "epoch": 19.36, "learning_rate": 5.5111403058172306e-08, "loss": 0.0046, "step": 62917 }, { "epoch": 19.36, "learning_rate": 5.5059175625384876e-08, "loss": 0.0031, "step": 62918 }, { "epoch": 19.36, "learning_rate": 5.5006972883204776e-08, "loss": 0.0037, "step": 62919 }, { "epoch": 19.36, "learning_rate": 5.4954794831761915e-08, "loss": 0.0038, "step": 62920 }, { "epoch": 19.36, "learning_rate": 5.490264147118507e-08, "loss": 0.0032, "step": 62921 }, { "epoch": 19.36, "learning_rate": 5.4850512801603026e-08, "loss": 0.0022, "step": 62922 }, { "epoch": 19.36, "learning_rate": 5.47984088231468e-08, "loss": 0.0033, "step": 62923 }, { "epoch": 19.36, "learning_rate": 5.474632953594405e-08, "loss": 0.0044, "step": 62924 }, { "epoch": 19.36, "learning_rate": 5.46942749401258e-08, "loss": 0.0043, "step": 62925 }, { "epoch": 19.36, "learning_rate": 5.464224503582083e-08, "loss": 0.0037, "step": 62926 }, { "epoch": 19.36, "learning_rate": 5.459023982315792e-08, "loss": 0.0045, "step": 62927 }, { "epoch": 19.36, "learning_rate": 5.453825930226586e-08, "loss": 0.0035, "step": 62928 }, { "epoch": 19.36, "learning_rate": 5.448630347327344e-08, "loss": 0.0031, "step": 62929 }, { "epoch": 19.36, "learning_rate": 5.443437233631166e-08, "loss": 0.003, "step": 62930 }, { "epoch": 19.36, "learning_rate": 5.4382465891507083e-08, "loss": 0.0031, "step": 62931 }, { "epoch": 19.36, "learning_rate": 5.4330584138989615e-08, "loss": 0.0018, "step": 62932 }, { "epoch": 19.36, "learning_rate": 5.427872707888804e-08, "loss": 0.0044, "step": 62933 }, { "epoch": 19.36, "learning_rate": 5.4226894711332246e-08, "loss": 0.0021, "step": 62934 }, { "epoch": 19.36, "learning_rate": 5.4175087036448805e-08, "loss": 0.0024, "step": 62935 }, { "epoch": 19.36, "learning_rate": 5.412330405436761e-08, "loss": 0.0018, "step": 62936 }, { "epoch": 19.36, "learning_rate": 5.407154576521634e-08, "loss": 0.0021, "step": 62937 }, { "epoch": 19.37, "learning_rate": 5.4019812169123776e-08, "loss": 0.0031, "step": 62938 }, { "epoch": 19.37, "learning_rate": 5.3968103266219815e-08, "loss": 0.002, "step": 62939 }, { "epoch": 19.37, "learning_rate": 5.391641905663103e-08, "loss": 0.0044, "step": 62940 }, { "epoch": 19.37, "learning_rate": 5.386475954048731e-08, "loss": 0.0049, "step": 62941 }, { "epoch": 19.37, "learning_rate": 5.381312471791522e-08, "loss": 0.0022, "step": 62942 }, { "epoch": 19.37, "learning_rate": 5.376151458904355e-08, "loss": 0.0042, "step": 62943 }, { "epoch": 19.37, "learning_rate": 5.370992915400108e-08, "loss": 0.0031, "step": 62944 }, { "epoch": 19.37, "learning_rate": 5.365836841291439e-08, "loss": 0.004, "step": 62945 }, { "epoch": 19.37, "learning_rate": 5.3606832365913354e-08, "loss": 0.0022, "step": 62946 }, { "epoch": 19.37, "learning_rate": 5.355532101312566e-08, "loss": 0.0045, "step": 62947 }, { "epoch": 19.37, "learning_rate": 5.3503834354677875e-08, "loss": 0.0032, "step": 62948 }, { "epoch": 19.37, "learning_rate": 5.345237239069878e-08, "loss": 0.003, "step": 62949 }, { "epoch": 19.37, "learning_rate": 5.3400935121316054e-08, "loss": 0.004, "step": 62950 }, { "epoch": 19.37, "learning_rate": 5.3349522546657374e-08, "loss": 0.0044, "step": 62951 }, { "epoch": 19.37, "learning_rate": 5.32981346668493e-08, "loss": 0.0039, "step": 62952 }, { "epoch": 19.37, "learning_rate": 5.324677148202173e-08, "loss": 0.0022, "step": 62953 }, { "epoch": 19.37, "learning_rate": 5.319543299230012e-08, "loss": 0.0019, "step": 62954 }, { "epoch": 19.37, "learning_rate": 5.314411919781326e-08, "loss": 0.004, "step": 62955 }, { "epoch": 19.37, "learning_rate": 5.3092830098687706e-08, "loss": 0.0026, "step": 62956 }, { "epoch": 19.37, "learning_rate": 5.3041565695050034e-08, "loss": 0.0023, "step": 62957 }, { "epoch": 19.37, "learning_rate": 5.2990325987029025e-08, "loss": 0.003, "step": 62958 }, { "epoch": 19.37, "learning_rate": 5.293911097475235e-08, "loss": 0.0028, "step": 62959 }, { "epoch": 19.37, "learning_rate": 5.288792065834547e-08, "loss": 0.0034, "step": 62960 }, { "epoch": 19.37, "learning_rate": 5.283675503793606e-08, "loss": 0.0017, "step": 62961 }, { "epoch": 19.37, "learning_rate": 5.2785614113651794e-08, "loss": 0.0037, "step": 62962 }, { "epoch": 19.37, "learning_rate": 5.273449788561924e-08, "loss": 0.0039, "step": 62963 }, { "epoch": 19.37, "learning_rate": 5.2683406353964964e-08, "loss": 0.0055, "step": 62964 }, { "epoch": 19.37, "learning_rate": 5.263233951881663e-08, "loss": 0.0038, "step": 62965 }, { "epoch": 19.37, "learning_rate": 5.258129738029971e-08, "loss": 0.01, "step": 62966 }, { "epoch": 19.37, "learning_rate": 5.2530279938541875e-08, "loss": 0.0049, "step": 62967 }, { "epoch": 19.37, "learning_rate": 5.247928719366968e-08, "loss": 0.0034, "step": 62968 }, { "epoch": 19.37, "learning_rate": 5.2428319145809705e-08, "loss": 0.0034, "step": 62969 }, { "epoch": 19.38, "learning_rate": 5.23773757950885e-08, "loss": 0.0028, "step": 62970 }, { "epoch": 19.38, "learning_rate": 5.232645714163265e-08, "loss": 0.0042, "step": 62971 }, { "epoch": 19.38, "learning_rate": 5.227556318556759e-08, "loss": 0.0041, "step": 62972 }, { "epoch": 19.38, "learning_rate": 5.2224693927021006e-08, "loss": 0.004, "step": 62973 }, { "epoch": 19.38, "learning_rate": 5.2173849366118354e-08, "loss": 0.0016, "step": 62974 }, { "epoch": 19.38, "learning_rate": 5.21230295029862e-08, "loss": 0.0032, "step": 62975 }, { "epoch": 19.38, "learning_rate": 5.207223433775222e-08, "loss": 0.0031, "step": 62976 }, { "epoch": 19.38, "learning_rate": 5.202146387053963e-08, "loss": 0.0051, "step": 62977 }, { "epoch": 19.38, "learning_rate": 5.197071810147503e-08, "loss": 0.0024, "step": 62978 }, { "epoch": 19.38, "learning_rate": 5.1919997030686065e-08, "loss": 0.0029, "step": 62979 }, { "epoch": 19.38, "learning_rate": 5.1869300658298207e-08, "loss": 0.0039, "step": 62980 }, { "epoch": 19.38, "learning_rate": 5.18186289844369e-08, "loss": 0.0038, "step": 62981 }, { "epoch": 19.38, "learning_rate": 5.1767982009227615e-08, "loss": 0.0019, "step": 62982 }, { "epoch": 19.38, "learning_rate": 5.17173597327969e-08, "loss": 0.0019, "step": 62983 }, { "epoch": 19.38, "learning_rate": 5.166676215526911e-08, "loss": 0.0026, "step": 62984 }, { "epoch": 19.38, "learning_rate": 5.161618927677192e-08, "loss": 0.0041, "step": 62985 }, { "epoch": 19.38, "learning_rate": 5.1565641097429675e-08, "loss": 0.0039, "step": 62986 }, { "epoch": 19.38, "learning_rate": 5.1515117617366716e-08, "loss": 0.0019, "step": 62987 }, { "epoch": 19.38, "learning_rate": 5.146461883671072e-08, "loss": 0.0085, "step": 62988 }, { "epoch": 19.38, "learning_rate": 5.141414475558604e-08, "loss": 0.0039, "step": 62989 }, { "epoch": 19.38, "learning_rate": 5.1363695374117004e-08, "loss": 0.0036, "step": 62990 }, { "epoch": 19.38, "learning_rate": 5.131327069243131e-08, "loss": 0.0032, "step": 62991 }, { "epoch": 19.38, "learning_rate": 5.126287071065106e-08, "loss": 0.0033, "step": 62992 }, { "epoch": 19.38, "learning_rate": 5.121249542890394e-08, "loss": 0.0027, "step": 62993 }, { "epoch": 19.38, "learning_rate": 5.11621448473143e-08, "loss": 0.003, "step": 62994 }, { "epoch": 19.38, "learning_rate": 5.1111818966006477e-08, "loss": 0.0029, "step": 62995 }, { "epoch": 19.38, "learning_rate": 5.106151778510593e-08, "loss": 0.0026, "step": 62996 }, { "epoch": 19.38, "learning_rate": 5.1011241304738115e-08, "loss": 0.0031, "step": 62997 }, { "epoch": 19.38, "learning_rate": 5.096098952502626e-08, "loss": 0.0031, "step": 62998 }, { "epoch": 19.38, "learning_rate": 5.0910762446095826e-08, "loss": 0.0041, "step": 62999 }, { "epoch": 19.38, "learning_rate": 5.0860560068073384e-08, "loss": 0.002, "step": 63000 }, { "epoch": 19.38, "learning_rate": 5.0810382391081047e-08, "loss": 0.0037, "step": 63001 }, { "epoch": 19.38, "learning_rate": 5.076022941524428e-08, "loss": 0.0051, "step": 63002 }, { "epoch": 19.39, "learning_rate": 5.071010114068742e-08, "loss": 0.0057, "step": 63003 }, { "epoch": 19.39, "learning_rate": 5.065999756753481e-08, "loss": 0.0047, "step": 63004 }, { "epoch": 19.39, "learning_rate": 5.060991869591192e-08, "loss": 0.0031, "step": 63005 }, { "epoch": 19.39, "learning_rate": 5.055986452594197e-08, "loss": 0.0033, "step": 63006 }, { "epoch": 19.39, "learning_rate": 5.0509835057750426e-08, "loss": 0.0028, "step": 63007 }, { "epoch": 19.39, "learning_rate": 5.04598302914594e-08, "loss": 0.0039, "step": 63008 }, { "epoch": 19.39, "learning_rate": 5.040985022719436e-08, "loss": 0.0041, "step": 63009 }, { "epoch": 19.39, "learning_rate": 5.035989486508075e-08, "loss": 0.0029, "step": 63010 }, { "epoch": 19.39, "learning_rate": 5.03099642052407e-08, "loss": 0.0029, "step": 63011 }, { "epoch": 19.39, "learning_rate": 5.026005824779745e-08, "loss": 0.0028, "step": 63012 }, { "epoch": 19.39, "learning_rate": 5.0210176992876444e-08, "loss": 0.0029, "step": 63013 }, { "epoch": 19.39, "learning_rate": 5.016032044060204e-08, "loss": 0.0051, "step": 63014 }, { "epoch": 19.39, "learning_rate": 5.011048859109635e-08, "loss": 0.0024, "step": 63015 }, { "epoch": 19.39, "learning_rate": 5.0060681444484835e-08, "loss": 0.0031, "step": 63016 }, { "epoch": 19.39, "learning_rate": 5.0010899000889624e-08, "loss": 0.0032, "step": 63017 }, { "epoch": 19.39, "learning_rate": 4.996114126043505e-08, "loss": 0.0029, "step": 63018 }, { "epoch": 19.39, "learning_rate": 4.991140822324436e-08, "loss": 0.005, "step": 63019 }, { "epoch": 19.39, "learning_rate": 4.9861699889441895e-08, "loss": 0.0034, "step": 63020 }, { "epoch": 19.39, "learning_rate": 4.981201625914978e-08, "loss": 0.0028, "step": 63021 }, { "epoch": 19.39, "learning_rate": 4.976235733249124e-08, "loss": 0.003, "step": 63022 }, { "epoch": 19.39, "learning_rate": 4.9712723109590636e-08, "loss": 0.0034, "step": 63023 }, { "epoch": 19.39, "learning_rate": 4.966311359057119e-08, "loss": 0.0046, "step": 63024 }, { "epoch": 19.39, "learning_rate": 4.9613528775556143e-08, "loss": 0.0036, "step": 63025 }, { "epoch": 19.39, "learning_rate": 4.9563968664667617e-08, "loss": 0.0039, "step": 63026 }, { "epoch": 19.39, "learning_rate": 4.9514433258028846e-08, "loss": 0.0034, "step": 63027 }, { "epoch": 19.39, "learning_rate": 4.9464922555764186e-08, "loss": 0.0027, "step": 63028 }, { "epoch": 19.39, "learning_rate": 4.9415436557994637e-08, "loss": 0.0027, "step": 63029 }, { "epoch": 19.39, "learning_rate": 4.9365975264844547e-08, "loss": 0.0037, "step": 63030 }, { "epoch": 19.39, "learning_rate": 4.931653867643604e-08, "loss": 0.0027, "step": 63031 }, { "epoch": 19.39, "learning_rate": 4.926712679289125e-08, "loss": 0.0041, "step": 63032 }, { "epoch": 19.39, "learning_rate": 4.9217739614333406e-08, "loss": 0.0027, "step": 63033 }, { "epoch": 19.39, "learning_rate": 4.916837714088685e-08, "loss": 0.0027, "step": 63034 }, { "epoch": 19.4, "learning_rate": 4.911903937267148e-08, "loss": 0.0033, "step": 63035 }, { "epoch": 19.4, "learning_rate": 4.906972630981055e-08, "loss": 0.0027, "step": 63036 }, { "epoch": 19.4, "learning_rate": 4.902043795242728e-08, "loss": 0.0019, "step": 63037 }, { "epoch": 19.4, "learning_rate": 4.8971174300643795e-08, "loss": 0.0038, "step": 63038 }, { "epoch": 19.4, "learning_rate": 4.892193535458223e-08, "loss": 0.0027, "step": 63039 }, { "epoch": 19.4, "learning_rate": 4.887272111436359e-08, "loss": 0.0032, "step": 63040 }, { "epoch": 19.4, "learning_rate": 4.882353158011221e-08, "loss": 0.0051, "step": 63041 }, { "epoch": 19.4, "learning_rate": 4.877436675194913e-08, "loss": 0.0052, "step": 63042 }, { "epoch": 19.4, "learning_rate": 4.872522662999646e-08, "loss": 0.0051, "step": 63043 }, { "epoch": 19.4, "learning_rate": 4.867611121437632e-08, "loss": 0.0034, "step": 63044 }, { "epoch": 19.4, "learning_rate": 4.8627020505210845e-08, "loss": 0.0025, "step": 63045 }, { "epoch": 19.4, "learning_rate": 4.857795450262215e-08, "loss": 0.002, "step": 63046 }, { "epoch": 19.4, "learning_rate": 4.852891320673014e-08, "loss": 0.0032, "step": 63047 }, { "epoch": 19.4, "learning_rate": 4.847989661765917e-08, "loss": 0.0043, "step": 63048 }, { "epoch": 19.4, "learning_rate": 4.843090473552914e-08, "loss": 0.0027, "step": 63049 }, { "epoch": 19.4, "learning_rate": 4.8381937560462166e-08, "loss": 0.0043, "step": 63050 }, { "epoch": 19.4, "learning_rate": 4.833299509258038e-08, "loss": 0.004, "step": 63051 }, { "epoch": 19.4, "learning_rate": 4.8284077332005905e-08, "loss": 0.0022, "step": 63052 }, { "epoch": 19.4, "learning_rate": 4.823518427885754e-08, "loss": 0.0037, "step": 63053 }, { "epoch": 19.4, "learning_rate": 4.818631593325851e-08, "loss": 0.0039, "step": 63054 }, { "epoch": 19.4, "learning_rate": 4.813747229532984e-08, "loss": 0.0034, "step": 63055 }, { "epoch": 19.4, "learning_rate": 4.8088653365193636e-08, "loss": 0.0035, "step": 63056 }, { "epoch": 19.4, "learning_rate": 4.803985914296983e-08, "loss": 0.0026, "step": 63057 }, { "epoch": 19.4, "learning_rate": 4.799108962878052e-08, "loss": 0.0034, "step": 63058 }, { "epoch": 19.4, "learning_rate": 4.7942344822746736e-08, "loss": 0.0045, "step": 63059 }, { "epoch": 19.4, "learning_rate": 4.7893624724988375e-08, "loss": 0.0023, "step": 63060 }, { "epoch": 19.4, "learning_rate": 4.784492933562645e-08, "loss": 0.0038, "step": 63061 }, { "epoch": 19.4, "learning_rate": 4.779625865478421e-08, "loss": 0.0023, "step": 63062 }, { "epoch": 19.4, "learning_rate": 4.7747612682579324e-08, "loss": 0.0028, "step": 63063 }, { "epoch": 19.4, "learning_rate": 4.769899141913503e-08, "loss": 0.004, "step": 63064 }, { "epoch": 19.4, "learning_rate": 4.765039486457013e-08, "loss": 0.0022, "step": 63065 }, { "epoch": 19.4, "learning_rate": 4.760182301900784e-08, "loss": 0.0016, "step": 63066 }, { "epoch": 19.4, "learning_rate": 4.755327588256586e-08, "loss": 0.0029, "step": 63067 }, { "epoch": 19.41, "learning_rate": 4.7504753455366314e-08, "loss": 0.0043, "step": 63068 }, { "epoch": 19.41, "learning_rate": 4.74562557375291e-08, "loss": 0.0019, "step": 63069 }, { "epoch": 19.41, "learning_rate": 4.7407782729175235e-08, "loss": 0.0039, "step": 63070 }, { "epoch": 19.41, "learning_rate": 4.735933443042462e-08, "loss": 0.003, "step": 63071 }, { "epoch": 19.41, "learning_rate": 4.7310910841397164e-08, "loss": 0.0037, "step": 63072 }, { "epoch": 19.41, "learning_rate": 4.7262511962213874e-08, "loss": 0.0027, "step": 63073 }, { "epoch": 19.41, "learning_rate": 4.7214137792995775e-08, "loss": 0.0068, "step": 63074 }, { "epoch": 19.41, "learning_rate": 4.716578833386054e-08, "loss": 0.0031, "step": 63075 }, { "epoch": 19.41, "learning_rate": 4.711746358492919e-08, "loss": 0.0045, "step": 63076 }, { "epoch": 19.41, "learning_rate": 4.7069163546322737e-08, "loss": 0.0014, "step": 63077 }, { "epoch": 19.41, "learning_rate": 4.702088821815998e-08, "loss": 0.007, "step": 63078 }, { "epoch": 19.41, "learning_rate": 4.697263760056081e-08, "loss": 0.0026, "step": 63079 }, { "epoch": 19.41, "learning_rate": 4.692441169364625e-08, "loss": 0.0047, "step": 63080 }, { "epoch": 19.41, "learning_rate": 4.6876210497533994e-08, "loss": 0.004, "step": 63081 }, { "epoch": 19.41, "learning_rate": 4.6828034012346146e-08, "loss": 0.0024, "step": 63082 }, { "epoch": 19.41, "learning_rate": 4.677988223819929e-08, "loss": 0.0025, "step": 63083 }, { "epoch": 19.41, "learning_rate": 4.673175517521556e-08, "loss": 0.0028, "step": 63084 }, { "epoch": 19.41, "learning_rate": 4.6683652823513725e-08, "loss": 0.0033, "step": 63085 }, { "epoch": 19.41, "learning_rate": 4.66355751832126e-08, "loss": 0.0027, "step": 63086 }, { "epoch": 19.41, "learning_rate": 4.658752225443319e-08, "loss": 0.0042, "step": 63087 }, { "epoch": 19.41, "learning_rate": 4.653949403729208e-08, "loss": 0.0026, "step": 63088 }, { "epoch": 19.41, "learning_rate": 4.649149053191138e-08, "loss": 0.0017, "step": 63089 }, { "epoch": 19.41, "learning_rate": 4.6443511738407664e-08, "loss": 0.0028, "step": 63090 }, { "epoch": 19.41, "learning_rate": 4.639555765690196e-08, "loss": 0.0026, "step": 63091 }, { "epoch": 19.41, "learning_rate": 4.6347628287513045e-08, "loss": 0.004, "step": 63092 }, { "epoch": 19.41, "learning_rate": 4.6299723630358616e-08, "loss": 0.0026, "step": 63093 }, { "epoch": 19.41, "learning_rate": 4.6251843685559684e-08, "loss": 0.0038, "step": 63094 }, { "epoch": 19.41, "learning_rate": 4.6203988453232816e-08, "loss": 0.0018, "step": 63095 }, { "epoch": 19.41, "learning_rate": 4.615615793349904e-08, "loss": 0.0031, "step": 63096 }, { "epoch": 19.41, "learning_rate": 4.610835212647491e-08, "loss": 0.0027, "step": 63097 }, { "epoch": 19.41, "learning_rate": 4.606057103228034e-08, "loss": 0.0031, "step": 63098 }, { "epoch": 19.41, "learning_rate": 4.601281465103413e-08, "loss": 0.0038, "step": 63099 }, { "epoch": 19.42, "learning_rate": 4.596508298285618e-08, "loss": 0.0038, "step": 63100 }, { "epoch": 19.42, "learning_rate": 4.5917376027861945e-08, "loss": 0.0022, "step": 63101 }, { "epoch": 19.42, "learning_rate": 4.5869693786171345e-08, "loss": 0.0049, "step": 63102 }, { "epoch": 19.42, "learning_rate": 4.582203625790205e-08, "loss": 0.0024, "step": 63103 }, { "epoch": 19.42, "learning_rate": 4.577440344317397e-08, "loss": 0.003, "step": 63104 }, { "epoch": 19.42, "learning_rate": 4.572679534210478e-08, "loss": 0.0038, "step": 63105 }, { "epoch": 19.42, "learning_rate": 4.567921195481217e-08, "loss": 0.0026, "step": 63106 }, { "epoch": 19.42, "learning_rate": 4.563165328141383e-08, "loss": 0.0029, "step": 63107 }, { "epoch": 19.42, "learning_rate": 4.5584119322028554e-08, "loss": 0.0068, "step": 63108 }, { "epoch": 19.42, "learning_rate": 4.553661007677512e-08, "loss": 0.0064, "step": 63109 }, { "epoch": 19.42, "learning_rate": 4.548912554577012e-08, "loss": 0.0042, "step": 63110 }, { "epoch": 19.42, "learning_rate": 4.5441665729132334e-08, "loss": 0.0043, "step": 63111 }, { "epoch": 19.42, "learning_rate": 4.5394230626978344e-08, "loss": 0.0031, "step": 63112 }, { "epoch": 19.42, "learning_rate": 4.534682023942805e-08, "loss": 0.003, "step": 63113 }, { "epoch": 19.42, "learning_rate": 4.529943456659691e-08, "loss": 0.0025, "step": 63114 }, { "epoch": 19.42, "learning_rate": 4.525207360860373e-08, "loss": 0.0032, "step": 63115 }, { "epoch": 19.42, "learning_rate": 4.5204737365566184e-08, "loss": 0.0038, "step": 63116 }, { "epoch": 19.42, "learning_rate": 4.515742583760196e-08, "loss": 0.0022, "step": 63117 }, { "epoch": 19.42, "learning_rate": 4.511013902482653e-08, "loss": 0.0038, "step": 63118 }, { "epoch": 19.42, "learning_rate": 4.506287692735978e-08, "loss": 0.0026, "step": 63119 }, { "epoch": 19.42, "learning_rate": 4.501563954531829e-08, "loss": 0.0031, "step": 63120 }, { "epoch": 19.42, "learning_rate": 4.496842687881975e-08, "loss": 0.0034, "step": 63121 }, { "epoch": 19.42, "learning_rate": 4.492123892798073e-08, "loss": 0.0025, "step": 63122 }, { "epoch": 19.42, "learning_rate": 4.4874075692917793e-08, "loss": 0.0043, "step": 63123 }, { "epoch": 19.42, "learning_rate": 4.4826937173748644e-08, "loss": 0.0031, "step": 63124 }, { "epoch": 19.42, "learning_rate": 4.4779823370590945e-08, "loss": 0.003, "step": 63125 }, { "epoch": 19.42, "learning_rate": 4.4732734283560176e-08, "loss": 0.0024, "step": 63126 }, { "epoch": 19.42, "learning_rate": 4.468566991277512e-08, "loss": 0.0037, "step": 63127 }, { "epoch": 19.42, "learning_rate": 4.4638630258351244e-08, "loss": 0.0025, "step": 63128 }, { "epoch": 19.42, "learning_rate": 4.459161532040623e-08, "loss": 0.0018, "step": 63129 }, { "epoch": 19.42, "learning_rate": 4.454462509905555e-08, "loss": 0.0033, "step": 63130 }, { "epoch": 19.42, "learning_rate": 4.449765959441688e-08, "loss": 0.004, "step": 63131 }, { "epoch": 19.42, "learning_rate": 4.4450718806606786e-08, "loss": 0.0049, "step": 63132 }, { "epoch": 19.43, "learning_rate": 4.440380273574185e-08, "loss": 0.002, "step": 63133 }, { "epoch": 19.43, "learning_rate": 4.435691138193865e-08, "loss": 0.0037, "step": 63134 }, { "epoch": 19.43, "learning_rate": 4.431004474531264e-08, "loss": 0.0026, "step": 63135 }, { "epoch": 19.43, "learning_rate": 4.42632028259804e-08, "loss": 0.0049, "step": 63136 }, { "epoch": 19.43, "learning_rate": 4.421638562405961e-08, "loss": 0.0032, "step": 63137 }, { "epoch": 19.43, "learning_rate": 4.4169593139664626e-08, "loss": 0.0021, "step": 63138 }, { "epoch": 19.43, "learning_rate": 4.412282537291312e-08, "loss": 0.0026, "step": 63139 }, { "epoch": 19.43, "learning_rate": 4.4076082323920576e-08, "loss": 0.0035, "step": 63140 }, { "epoch": 19.43, "learning_rate": 4.4029363992802444e-08, "loss": 0.0044, "step": 63141 }, { "epoch": 19.43, "learning_rate": 4.398267037967641e-08, "loss": 0.0022, "step": 63142 }, { "epoch": 19.43, "learning_rate": 4.393600148465682e-08, "loss": 0.0042, "step": 63143 }, { "epoch": 19.43, "learning_rate": 4.388935730786026e-08, "loss": 0.0069, "step": 63144 }, { "epoch": 19.43, "learning_rate": 4.384273784940218e-08, "loss": 0.0038, "step": 63145 }, { "epoch": 19.43, "learning_rate": 4.379614310939806e-08, "loss": 0.0033, "step": 63146 }, { "epoch": 19.43, "learning_rate": 4.374957308796446e-08, "loss": 0.0024, "step": 63147 }, { "epoch": 19.43, "learning_rate": 4.370302778521684e-08, "loss": 0.0035, "step": 63148 }, { "epoch": 19.43, "learning_rate": 4.365650720126957e-08, "loss": 0.0047, "step": 63149 }, { "epoch": 19.43, "learning_rate": 4.361001133624032e-08, "loss": 0.0024, "step": 63150 }, { "epoch": 19.43, "learning_rate": 4.356354019024234e-08, "loss": 0.0024, "step": 63151 }, { "epoch": 19.43, "learning_rate": 4.3517093763392194e-08, "loss": 0.0039, "step": 63152 }, { "epoch": 19.43, "learning_rate": 4.347067205580424e-08, "loss": 0.0039, "step": 63153 }, { "epoch": 19.43, "learning_rate": 4.342427506759506e-08, "loss": 0.006, "step": 63154 }, { "epoch": 19.43, "learning_rate": 4.33779027988801e-08, "loss": 0.0021, "step": 63155 }, { "epoch": 19.43, "learning_rate": 4.333155524977262e-08, "loss": 0.0048, "step": 63156 }, { "epoch": 19.43, "learning_rate": 4.328523242038807e-08, "loss": 0.0034, "step": 63157 }, { "epoch": 19.43, "learning_rate": 4.3238934310843026e-08, "loss": 0.0028, "step": 63158 }, { "epoch": 19.43, "learning_rate": 4.319266092125074e-08, "loss": 0.0029, "step": 63159 }, { "epoch": 19.43, "learning_rate": 4.3146412251726665e-08, "loss": 0.003, "step": 63160 }, { "epoch": 19.43, "learning_rate": 4.310018830238627e-08, "loss": 0.0025, "step": 63161 }, { "epoch": 19.43, "learning_rate": 4.3053989073343904e-08, "loss": 0.0027, "step": 63162 }, { "epoch": 19.43, "learning_rate": 4.3007814564713925e-08, "loss": 0.0046, "step": 63163 }, { "epoch": 19.43, "learning_rate": 4.296166477661179e-08, "loss": 0.0035, "step": 63164 }, { "epoch": 19.44, "learning_rate": 4.291553970915074e-08, "loss": 0.0055, "step": 63165 }, { "epoch": 19.44, "learning_rate": 4.2869439362446254e-08, "loss": 0.0031, "step": 63166 }, { "epoch": 19.44, "learning_rate": 4.282336373661378e-08, "loss": 0.0044, "step": 63167 }, { "epoch": 19.44, "learning_rate": 4.277731283176545e-08, "loss": 0.0035, "step": 63168 }, { "epoch": 19.44, "learning_rate": 4.273128664801563e-08, "loss": 0.0046, "step": 63169 }, { "epoch": 19.44, "learning_rate": 4.2685285185480876e-08, "loss": 0.003, "step": 63170 }, { "epoch": 19.44, "learning_rate": 4.263930844427444e-08, "loss": 0.0022, "step": 63171 }, { "epoch": 19.44, "learning_rate": 4.259335642450957e-08, "loss": 0.0037, "step": 63172 }, { "epoch": 19.44, "learning_rate": 4.254742912630061e-08, "loss": 0.0034, "step": 63173 }, { "epoch": 19.44, "learning_rate": 4.250152654976303e-08, "loss": 0.0034, "step": 63174 }, { "epoch": 19.44, "learning_rate": 4.2455648695008956e-08, "loss": 0.0046, "step": 63175 }, { "epoch": 19.44, "learning_rate": 4.240979556215274e-08, "loss": 0.0027, "step": 63176 }, { "epoch": 19.44, "learning_rate": 4.236396715130986e-08, "loss": 0.003, "step": 63177 }, { "epoch": 19.44, "learning_rate": 4.231816346259132e-08, "loss": 0.004, "step": 63178 }, { "epoch": 19.44, "learning_rate": 4.2272384496112597e-08, "loss": 0.0044, "step": 63179 }, { "epoch": 19.44, "learning_rate": 4.222663025198692e-08, "loss": 0.0043, "step": 63180 }, { "epoch": 19.44, "learning_rate": 4.2180900730328655e-08, "loss": 0.003, "step": 63181 }, { "epoch": 19.44, "learning_rate": 4.2135195931249925e-08, "loss": 0.0035, "step": 63182 }, { "epoch": 19.44, "learning_rate": 4.208951585486509e-08, "loss": 0.0041, "step": 63183 }, { "epoch": 19.44, "learning_rate": 4.20438605012885e-08, "loss": 0.0023, "step": 63184 }, { "epoch": 19.44, "learning_rate": 4.199822987063118e-08, "loss": 0.004, "step": 63185 }, { "epoch": 19.44, "learning_rate": 4.1952623963007475e-08, "loss": 0.0042, "step": 63186 }, { "epoch": 19.44, "learning_rate": 4.190704277853175e-08, "loss": 0.0029, "step": 63187 }, { "epoch": 19.44, "learning_rate": 4.186148631731613e-08, "loss": 0.0037, "step": 63188 }, { "epoch": 19.44, "learning_rate": 4.181595457947385e-08, "loss": 0.0022, "step": 63189 }, { "epoch": 19.44, "learning_rate": 4.177044756511706e-08, "loss": 0.0043, "step": 63190 }, { "epoch": 19.44, "learning_rate": 4.1724965274360096e-08, "loss": 0.0028, "step": 63191 }, { "epoch": 19.44, "learning_rate": 4.1679507707315106e-08, "loss": 0.0027, "step": 63192 }, { "epoch": 19.44, "learning_rate": 4.163407486409532e-08, "loss": 0.0033, "step": 63193 }, { "epoch": 19.44, "learning_rate": 4.1588666744813986e-08, "loss": 0.0037, "step": 63194 }, { "epoch": 19.44, "learning_rate": 4.154328334958324e-08, "loss": 0.0053, "step": 63195 }, { "epoch": 19.44, "learning_rate": 4.1497924678515214e-08, "loss": 0.0037, "step": 63196 }, { "epoch": 19.44, "learning_rate": 4.145259073172314e-08, "loss": 0.0033, "step": 63197 }, { "epoch": 19.45, "learning_rate": 4.140728150932027e-08, "loss": 0.0019, "step": 63198 }, { "epoch": 19.45, "learning_rate": 4.136199701141763e-08, "loss": 0.0034, "step": 63199 }, { "epoch": 19.45, "learning_rate": 4.1316737238128455e-08, "loss": 0.0032, "step": 63200 }, { "epoch": 19.45, "learning_rate": 4.127150218956488e-08, "loss": 0.0055, "step": 63201 }, { "epoch": 19.45, "learning_rate": 4.1226291865840154e-08, "loss": 0.0043, "step": 63202 }, { "epoch": 19.45, "learning_rate": 4.1181106267065284e-08, "loss": 0.0028, "step": 63203 }, { "epoch": 19.45, "learning_rate": 4.113594539335242e-08, "loss": 0.0018, "step": 63204 }, { "epoch": 19.45, "learning_rate": 4.109080924481479e-08, "loss": 0.0051, "step": 63205 }, { "epoch": 19.45, "learning_rate": 4.1045697821564534e-08, "loss": 0.002, "step": 63206 }, { "epoch": 19.45, "learning_rate": 4.100061112371157e-08, "loss": 0.004, "step": 63207 }, { "epoch": 19.45, "learning_rate": 4.095554915136912e-08, "loss": 0.004, "step": 63208 }, { "epoch": 19.45, "learning_rate": 4.091051190465045e-08, "loss": 0.0027, "step": 63209 }, { "epoch": 19.45, "learning_rate": 4.0865499383665465e-08, "loss": 0.002, "step": 63210 }, { "epoch": 19.45, "learning_rate": 4.0820511588527404e-08, "loss": 0.003, "step": 63211 }, { "epoch": 19.45, "learning_rate": 4.077554851934618e-08, "loss": 0.0034, "step": 63212 }, { "epoch": 19.45, "learning_rate": 4.073061017623392e-08, "loss": 0.005, "step": 63213 }, { "epoch": 19.45, "learning_rate": 4.068569655930277e-08, "loss": 0.0089, "step": 63214 }, { "epoch": 19.45, "learning_rate": 4.064080766866485e-08, "loss": 0.0029, "step": 63215 }, { "epoch": 19.45, "learning_rate": 4.059594350443008e-08, "loss": 0.0021, "step": 63216 }, { "epoch": 19.45, "learning_rate": 4.0551104066711697e-08, "loss": 0.0036, "step": 63217 }, { "epoch": 19.45, "learning_rate": 4.050628935561851e-08, "loss": 0.0048, "step": 63218 }, { "epoch": 19.45, "learning_rate": 4.0461499371263757e-08, "loss": 0.0055, "step": 63219 }, { "epoch": 19.45, "learning_rate": 4.041673411375846e-08, "loss": 0.0022, "step": 63220 }, { "epoch": 19.45, "learning_rate": 4.0371993583213644e-08, "loss": 0.0036, "step": 63221 }, { "epoch": 19.45, "learning_rate": 4.032727777973922e-08, "loss": 0.0021, "step": 63222 }, { "epoch": 19.45, "learning_rate": 4.0282586703447315e-08, "loss": 0.0043, "step": 63223 }, { "epoch": 19.45, "learning_rate": 4.023792035445007e-08, "loss": 0.0038, "step": 63224 }, { "epoch": 19.45, "learning_rate": 4.0193278732855166e-08, "loss": 0.0033, "step": 63225 }, { "epoch": 19.45, "learning_rate": 4.014866183877586e-08, "loss": 0.0038, "step": 63226 }, { "epoch": 19.45, "learning_rate": 4.010406967232316e-08, "loss": 0.0059, "step": 63227 }, { "epoch": 19.45, "learning_rate": 4.0059502233605886e-08, "loss": 0.0023, "step": 63228 }, { "epoch": 19.45, "learning_rate": 4.0014959522736154e-08, "loss": 0.0025, "step": 63229 }, { "epoch": 19.46, "learning_rate": 3.9970441539823876e-08, "loss": 0.0025, "step": 63230 }, { "epoch": 19.46, "learning_rate": 3.9925948284980086e-08, "loss": 0.0027, "step": 63231 }, { "epoch": 19.46, "learning_rate": 3.98814797583158e-08, "loss": 0.0037, "step": 63232 }, { "epoch": 19.46, "learning_rate": 3.983703595993871e-08, "loss": 0.0069, "step": 63233 }, { "epoch": 19.46, "learning_rate": 3.9792616889962054e-08, "loss": 0.0057, "step": 63234 }, { "epoch": 19.46, "learning_rate": 3.974822254849575e-08, "loss": 0.0038, "step": 63235 }, { "epoch": 19.46, "learning_rate": 3.97038529356486e-08, "loss": 0.0038, "step": 63236 }, { "epoch": 19.46, "learning_rate": 3.965950805153274e-08, "loss": 0.0037, "step": 63237 }, { "epoch": 19.46, "learning_rate": 3.961518789625585e-08, "loss": 0.0035, "step": 63238 }, { "epoch": 19.46, "learning_rate": 3.9570892469930064e-08, "loss": 0.0038, "step": 63239 }, { "epoch": 19.46, "learning_rate": 3.95266217726642e-08, "loss": 0.0039, "step": 63240 }, { "epoch": 19.46, "learning_rate": 3.9482375804568154e-08, "loss": 0.0019, "step": 63241 }, { "epoch": 19.46, "learning_rate": 3.943815456575184e-08, "loss": 0.0028, "step": 63242 }, { "epoch": 19.46, "learning_rate": 3.939395805632629e-08, "loss": 0.0025, "step": 63243 }, { "epoch": 19.46, "learning_rate": 3.934978627639918e-08, "loss": 0.0044, "step": 63244 }, { "epoch": 19.46, "learning_rate": 3.930563922608266e-08, "loss": 0.0036, "step": 63245 }, { "epoch": 19.46, "learning_rate": 3.92615169054833e-08, "loss": 0.0022, "step": 63246 }, { "epoch": 19.46, "learning_rate": 3.9217419314713236e-08, "loss": 0.0022, "step": 63247 }, { "epoch": 19.46, "learning_rate": 3.917334645388016e-08, "loss": 0.0032, "step": 63248 }, { "epoch": 19.46, "learning_rate": 3.912929832309509e-08, "loss": 0.0016, "step": 63249 }, { "epoch": 19.46, "learning_rate": 3.908527492246683e-08, "loss": 0.0029, "step": 63250 }, { "epoch": 19.46, "learning_rate": 3.904127625210419e-08, "loss": 0.0046, "step": 63251 }, { "epoch": 19.46, "learning_rate": 3.8997302312115956e-08, "loss": 0.0042, "step": 63252 }, { "epoch": 19.46, "learning_rate": 3.8953353102613166e-08, "loss": 0.0046, "step": 63253 }, { "epoch": 19.46, "learning_rate": 3.8909428623703505e-08, "loss": 0.0046, "step": 63254 }, { "epoch": 19.46, "learning_rate": 3.886552887549577e-08, "loss": 0.0032, "step": 63255 }, { "epoch": 19.46, "learning_rate": 3.882165385809988e-08, "loss": 0.0033, "step": 63256 }, { "epoch": 19.46, "learning_rate": 3.877780357162353e-08, "loss": 0.0036, "step": 63257 }, { "epoch": 19.46, "learning_rate": 3.873397801617773e-08, "loss": 0.004, "step": 63258 }, { "epoch": 19.46, "learning_rate": 3.8690177191869074e-08, "loss": 0.0036, "step": 63259 }, { "epoch": 19.46, "learning_rate": 3.8646401098807465e-08, "loss": 0.0036, "step": 63260 }, { "epoch": 19.46, "learning_rate": 3.8602649737101705e-08, "loss": 0.0039, "step": 63261 }, { "epoch": 19.46, "learning_rate": 3.855892310685949e-08, "loss": 0.0052, "step": 63262 }, { "epoch": 19.47, "learning_rate": 3.8515221208189626e-08, "loss": 0.0024, "step": 63263 }, { "epoch": 19.47, "learning_rate": 3.8471544041202014e-08, "loss": 0.0046, "step": 63264 }, { "epoch": 19.47, "learning_rate": 3.8427891606003244e-08, "loss": 0.0026, "step": 63265 }, { "epoch": 19.47, "learning_rate": 3.8384263902703224e-08, "loss": 0.0037, "step": 63266 }, { "epoch": 19.47, "learning_rate": 3.8340660931408536e-08, "loss": 0.0031, "step": 63267 }, { "epoch": 19.47, "learning_rate": 3.8297082692229095e-08, "loss": 0.0029, "step": 63268 }, { "epoch": 19.47, "learning_rate": 3.825352918527259e-08, "loss": 0.0033, "step": 63269 }, { "epoch": 19.47, "learning_rate": 3.82100004106456e-08, "loss": 0.0028, "step": 63270 }, { "epoch": 19.47, "learning_rate": 3.816649636845915e-08, "loss": 0.0028, "step": 63271 }, { "epoch": 19.47, "learning_rate": 3.8123017058818714e-08, "loss": 0.0019, "step": 63272 }, { "epoch": 19.47, "learning_rate": 3.8079562481834196e-08, "loss": 0.0028, "step": 63273 }, { "epoch": 19.47, "learning_rate": 3.803613263761108e-08, "loss": 0.0024, "step": 63274 }, { "epoch": 19.47, "learning_rate": 3.7992727526259264e-08, "loss": 0.0031, "step": 63275 }, { "epoch": 19.47, "learning_rate": 3.794934714788645e-08, "loss": 0.0018, "step": 63276 }, { "epoch": 19.47, "learning_rate": 3.7905991502598106e-08, "loss": 0.0038, "step": 63277 }, { "epoch": 19.47, "learning_rate": 3.786266059050525e-08, "loss": 0.0039, "step": 63278 }, { "epoch": 19.47, "learning_rate": 3.7819354411713364e-08, "loss": 0.0039, "step": 63279 }, { "epoch": 19.47, "learning_rate": 3.777607296633012e-08, "loss": 0.0039, "step": 63280 }, { "epoch": 19.47, "learning_rate": 3.773281625446212e-08, "loss": 0.0043, "step": 63281 }, { "epoch": 19.47, "learning_rate": 3.768958427621816e-08, "loss": 0.0035, "step": 63282 }, { "epoch": 19.47, "learning_rate": 3.764637703170593e-08, "loss": 0.0026, "step": 63283 }, { "epoch": 19.47, "learning_rate": 3.760319452103201e-08, "loss": 0.0045, "step": 63284 }, { "epoch": 19.47, "learning_rate": 3.7560036744302977e-08, "loss": 0.0036, "step": 63285 }, { "epoch": 19.47, "learning_rate": 3.7516903701627635e-08, "loss": 0.0048, "step": 63286 }, { "epoch": 19.47, "learning_rate": 3.747379539311036e-08, "loss": 0.0027, "step": 63287 }, { "epoch": 19.47, "learning_rate": 3.7430711818861045e-08, "loss": 0.0061, "step": 63288 }, { "epoch": 19.47, "learning_rate": 3.738765297898517e-08, "loss": 0.0028, "step": 63289 }, { "epoch": 19.47, "learning_rate": 3.734461887359042e-08, "loss": 0.0044, "step": 63290 }, { "epoch": 19.47, "learning_rate": 3.7301609502782277e-08, "loss": 0.0048, "step": 63291 }, { "epoch": 19.47, "learning_rate": 3.7258624866668424e-08, "loss": 0.0018, "step": 63292 }, { "epoch": 19.47, "learning_rate": 3.7215664965356554e-08, "loss": 0.0047, "step": 63293 }, { "epoch": 19.47, "learning_rate": 3.717272979895104e-08, "loss": 0.003, "step": 63294 }, { "epoch": 19.48, "learning_rate": 3.712981936756066e-08, "loss": 0.0061, "step": 63295 }, { "epoch": 19.48, "learning_rate": 3.7086933671290904e-08, "loss": 0.0031, "step": 63296 }, { "epoch": 19.48, "learning_rate": 3.704407271024724e-08, "loss": 0.0023, "step": 63297 }, { "epoch": 19.48, "learning_rate": 3.700123648453846e-08, "loss": 0.0042, "step": 63298 }, { "epoch": 19.48, "learning_rate": 3.695842499426894e-08, "loss": 0.0024, "step": 63299 }, { "epoch": 19.48, "learning_rate": 3.691563823954636e-08, "loss": 0.0036, "step": 63300 }, { "epoch": 19.48, "learning_rate": 3.687287622047619e-08, "loss": 0.0036, "step": 63301 }, { "epoch": 19.48, "learning_rate": 3.683013893716392e-08, "loss": 0.002, "step": 63302 }, { "epoch": 19.48, "learning_rate": 3.678742638971722e-08, "loss": 0.0025, "step": 63303 }, { "epoch": 19.48, "learning_rate": 3.6744738578241565e-08, "loss": 0.0042, "step": 63304 }, { "epoch": 19.48, "learning_rate": 3.670207550284133e-08, "loss": 0.0037, "step": 63305 }, { "epoch": 19.48, "learning_rate": 3.66594371636253e-08, "loss": 0.0025, "step": 63306 }, { "epoch": 19.48, "learning_rate": 3.661682356069674e-08, "loss": 0.0035, "step": 63307 }, { "epoch": 19.48, "learning_rate": 3.657423469416332e-08, "loss": 0.0025, "step": 63308 }, { "epoch": 19.48, "learning_rate": 3.653167056413054e-08, "loss": 0.0016, "step": 63309 }, { "epoch": 19.48, "learning_rate": 3.6489131170702744e-08, "loss": 0.0026, "step": 63310 }, { "epoch": 19.48, "learning_rate": 3.644661651398651e-08, "loss": 0.0047, "step": 63311 }, { "epoch": 19.48, "learning_rate": 3.640412659408621e-08, "loss": 0.0031, "step": 63312 }, { "epoch": 19.48, "learning_rate": 3.6361661411109525e-08, "loss": 0.0022, "step": 63313 }, { "epoch": 19.48, "learning_rate": 3.631922096516083e-08, "loss": 0.0042, "step": 63314 }, { "epoch": 19.48, "learning_rate": 3.627680525634558e-08, "loss": 0.0044, "step": 63315 }, { "epoch": 19.48, "learning_rate": 3.623441428476815e-08, "loss": 0.0023, "step": 63316 }, { "epoch": 19.48, "learning_rate": 3.6192048050534e-08, "loss": 0.004, "step": 63317 }, { "epoch": 19.48, "learning_rate": 3.6149706553749717e-08, "loss": 0.0033, "step": 63318 }, { "epoch": 19.48, "learning_rate": 3.6107389794519666e-08, "loss": 0.0024, "step": 63319 }, { "epoch": 19.48, "learning_rate": 3.6065097772948197e-08, "loss": 0.0025, "step": 63320 }, { "epoch": 19.48, "learning_rate": 3.60228304891419e-08, "loss": 0.0033, "step": 63321 }, { "epoch": 19.48, "learning_rate": 3.598058794320402e-08, "loss": 0.0038, "step": 63322 }, { "epoch": 19.48, "learning_rate": 3.593837013524004e-08, "loss": 0.0018, "step": 63323 }, { "epoch": 19.48, "learning_rate": 3.589617706535431e-08, "loss": 0.0021, "step": 63324 }, { "epoch": 19.48, "learning_rate": 3.585400873365341e-08, "loss": 0.0034, "step": 63325 }, { "epoch": 19.48, "learning_rate": 3.581186514023949e-08, "loss": 0.0024, "step": 63326 }, { "epoch": 19.48, "learning_rate": 3.576974628521912e-08, "loss": 0.0034, "step": 63327 }, { "epoch": 19.49, "learning_rate": 3.572765216869667e-08, "loss": 0.0037, "step": 63328 }, { "epoch": 19.49, "learning_rate": 3.5685582790775386e-08, "loss": 0.0029, "step": 63329 }, { "epoch": 19.49, "learning_rate": 3.564353815156074e-08, "loss": 0.0016, "step": 63330 }, { "epoch": 19.49, "learning_rate": 3.5601518251157096e-08, "loss": 0.0033, "step": 63331 }, { "epoch": 19.49, "learning_rate": 3.55595230896677e-08, "loss": 0.0041, "step": 63332 }, { "epoch": 19.49, "learning_rate": 3.551755266719914e-08, "loss": 0.0039, "step": 63333 }, { "epoch": 19.49, "learning_rate": 3.547560698385355e-08, "loss": 0.0034, "step": 63334 }, { "epoch": 19.49, "learning_rate": 3.543368603973529e-08, "loss": 0.0028, "step": 63335 }, { "epoch": 19.49, "learning_rate": 3.539178983494873e-08, "loss": 0.0043, "step": 63336 }, { "epoch": 19.49, "learning_rate": 3.5349918369598223e-08, "loss": 0.0023, "step": 63337 }, { "epoch": 19.49, "learning_rate": 3.530807164378813e-08, "loss": 0.0038, "step": 63338 }, { "epoch": 19.49, "learning_rate": 3.52662496576206e-08, "loss": 0.0056, "step": 63339 }, { "epoch": 19.49, "learning_rate": 3.5224452411201094e-08, "loss": 0.0036, "step": 63340 }, { "epoch": 19.49, "learning_rate": 3.518267990463287e-08, "loss": 0.0041, "step": 63341 }, { "epoch": 19.49, "learning_rate": 3.514093213801917e-08, "loss": 0.0042, "step": 63342 }, { "epoch": 19.49, "learning_rate": 3.509920911146547e-08, "loss": 0.0018, "step": 63343 }, { "epoch": 19.49, "learning_rate": 3.50575108250728e-08, "loss": 0.004, "step": 63344 }, { "epoch": 19.49, "learning_rate": 3.501583727894553e-08, "loss": 0.0041, "step": 63345 }, { "epoch": 19.49, "learning_rate": 3.497418847318801e-08, "loss": 0.0032, "step": 63346 }, { "epoch": 19.49, "learning_rate": 3.493256440790349e-08, "loss": 0.0032, "step": 63347 }, { "epoch": 19.49, "learning_rate": 3.489096508319412e-08, "loss": 0.0052, "step": 63348 }, { "epoch": 19.49, "learning_rate": 3.484939049916425e-08, "loss": 0.0019, "step": 63349 }, { "epoch": 19.49, "learning_rate": 3.4807840655917133e-08, "loss": 0.0038, "step": 63350 }, { "epoch": 19.49, "learning_rate": 3.476631555355603e-08, "loss": 0.0035, "step": 63351 }, { "epoch": 19.49, "learning_rate": 3.472481519218196e-08, "loss": 0.0057, "step": 63352 }, { "epoch": 19.49, "learning_rate": 3.468333957190151e-08, "loss": 0.0026, "step": 63353 }, { "epoch": 19.49, "learning_rate": 3.4641888692814596e-08, "loss": 0.0029, "step": 63354 }, { "epoch": 19.49, "learning_rate": 3.4600462555025585e-08, "loss": 0.0043, "step": 63355 }, { "epoch": 19.49, "learning_rate": 3.455906115863772e-08, "loss": 0.0037, "step": 63356 }, { "epoch": 19.49, "learning_rate": 3.451768450375204e-08, "loss": 0.0039, "step": 63357 }, { "epoch": 19.49, "learning_rate": 3.447633259047289e-08, "loss": 0.0028, "step": 63358 }, { "epoch": 19.49, "learning_rate": 3.4435005418902436e-08, "loss": 0.0043, "step": 63359 }, { "epoch": 19.5, "learning_rate": 3.4393702989142796e-08, "loss": 0.003, "step": 63360 }, { "epoch": 19.5, "learning_rate": 3.435242530129723e-08, "loss": 0.0033, "step": 63361 }, { "epoch": 19.5, "learning_rate": 3.431117235546788e-08, "loss": 0.0033, "step": 63362 }, { "epoch": 19.5, "learning_rate": 3.426994415175688e-08, "loss": 0.0037, "step": 63363 }, { "epoch": 19.5, "learning_rate": 3.4228740690267496e-08, "loss": 0.0034, "step": 63364 }, { "epoch": 19.5, "learning_rate": 3.4187561971101844e-08, "loss": 0.0033, "step": 63365 }, { "epoch": 19.5, "learning_rate": 3.4146407994360975e-08, "loss": 0.0034, "step": 63366 }, { "epoch": 19.5, "learning_rate": 3.410527876014702e-08, "loss": 0.0038, "step": 63367 }, { "epoch": 19.5, "learning_rate": 3.4064174268564345e-08, "loss": 0.0045, "step": 63368 }, { "epoch": 19.5, "learning_rate": 3.4023094519713976e-08, "loss": 0.003, "step": 63369 }, { "epoch": 19.5, "learning_rate": 3.3982039513696943e-08, "loss": 0.0027, "step": 63370 }, { "epoch": 19.5, "learning_rate": 3.394100925061539e-08, "loss": 0.0051, "step": 63371 }, { "epoch": 19.5, "learning_rate": 3.390000373057145e-08, "loss": 0.0032, "step": 63372 }, { "epoch": 19.5, "learning_rate": 3.3859022953667274e-08, "loss": 0.0039, "step": 63373 }, { "epoch": 19.5, "learning_rate": 3.381806692000389e-08, "loss": 0.0037, "step": 63374 }, { "epoch": 19.5, "learning_rate": 3.377713562968454e-08, "loss": 0.005, "step": 63375 }, { "epoch": 19.5, "learning_rate": 3.373622908280916e-08, "loss": 0.0038, "step": 63376 }, { "epoch": 19.5, "learning_rate": 3.369534727947987e-08, "loss": 0.0039, "step": 63377 }, { "epoch": 19.5, "learning_rate": 3.365449021979772e-08, "loss": 0.0024, "step": 63378 }, { "epoch": 19.5, "learning_rate": 3.3613657903865945e-08, "loss": 0.0063, "step": 63379 }, { "epoch": 19.5, "learning_rate": 3.357285033178337e-08, "loss": 0.0025, "step": 63380 }, { "epoch": 19.5, "learning_rate": 3.353206750365323e-08, "loss": 0.0037, "step": 63381 }, { "epoch": 19.5, "learning_rate": 3.3491309419575455e-08, "loss": 0.0042, "step": 63382 }, { "epoch": 19.5, "learning_rate": 3.345057607965219e-08, "loss": 0.0044, "step": 63383 }, { "epoch": 19.5, "learning_rate": 3.340986748398445e-08, "loss": 0.0033, "step": 63384 }, { "epoch": 19.5, "learning_rate": 3.3369183632673275e-08, "loss": 0.0031, "step": 63385 }, { "epoch": 19.5, "learning_rate": 3.33285245258197e-08, "loss": 0.0035, "step": 63386 }, { "epoch": 19.5, "learning_rate": 3.3287890163523626e-08, "loss": 0.0018, "step": 63387 }, { "epoch": 19.5, "learning_rate": 3.324728054588611e-08, "loss": 0.0039, "step": 63388 }, { "epoch": 19.5, "learning_rate": 3.320669567301038e-08, "loss": 0.0044, "step": 63389 }, { "epoch": 19.5, "learning_rate": 3.316613554499526e-08, "loss": 0.0061, "step": 63390 }, { "epoch": 19.5, "learning_rate": 3.312560016194066e-08, "loss": 0.0031, "step": 63391 }, { "epoch": 19.5, "learning_rate": 3.3085089523948733e-08, "loss": 0.0022, "step": 63392 }, { "epoch": 19.51, "learning_rate": 3.304460363111939e-08, "loss": 0.0062, "step": 63393 }, { "epoch": 19.51, "learning_rate": 3.300414248355255e-08, "loss": 0.0057, "step": 63394 }, { "epoch": 19.51, "learning_rate": 3.2963706081350357e-08, "loss": 0.0028, "step": 63395 }, { "epoch": 19.51, "learning_rate": 3.292329442461273e-08, "loss": 0.003, "step": 63396 }, { "epoch": 19.51, "learning_rate": 3.2882907513438475e-08, "loss": 0.0028, "step": 63397 }, { "epoch": 19.51, "learning_rate": 3.2842545347929744e-08, "loss": 0.0031, "step": 63398 }, { "epoch": 19.51, "learning_rate": 3.2802207928185336e-08, "loss": 0.0025, "step": 63399 }, { "epoch": 19.51, "learning_rate": 3.2761895254306285e-08, "loss": 0.0058, "step": 63400 }, { "epoch": 19.51, "learning_rate": 3.272160732639251e-08, "loss": 0.0027, "step": 63401 }, { "epoch": 19.51, "learning_rate": 3.268134414454283e-08, "loss": 0.0033, "step": 63402 }, { "epoch": 19.51, "learning_rate": 3.2641105708859366e-08, "loss": 0.0029, "step": 63403 }, { "epoch": 19.51, "learning_rate": 3.260089201944094e-08, "loss": 0.0039, "step": 63404 }, { "epoch": 19.51, "learning_rate": 3.256070307638637e-08, "loss": 0.0052, "step": 63405 }, { "epoch": 19.51, "learning_rate": 3.252053887979778e-08, "loss": 0.0034, "step": 63406 }, { "epoch": 19.51, "learning_rate": 3.248039942977177e-08, "loss": 0.0022, "step": 63407 }, { "epoch": 19.51, "learning_rate": 3.244028472641159e-08, "loss": 0.0022, "step": 63408 }, { "epoch": 19.51, "learning_rate": 3.240019476981382e-08, "loss": 0.0032, "step": 63409 }, { "epoch": 19.51, "learning_rate": 3.23601295600795e-08, "loss": 0.0039, "step": 63410 }, { "epoch": 19.51, "learning_rate": 3.232008909730855e-08, "loss": 0.007, "step": 63411 }, { "epoch": 19.51, "learning_rate": 3.228007338159866e-08, "loss": 0.0047, "step": 63412 }, { "epoch": 19.51, "learning_rate": 3.224008241304977e-08, "loss": 0.0041, "step": 63413 }, { "epoch": 19.51, "learning_rate": 3.220011619176289e-08, "loss": 0.003, "step": 63414 }, { "epoch": 19.51, "learning_rate": 3.2160174717834615e-08, "loss": 0.0018, "step": 63415 }, { "epoch": 19.51, "learning_rate": 3.212025799136598e-08, "loss": 0.0033, "step": 63416 }, { "epoch": 19.51, "learning_rate": 3.208036601245579e-08, "loss": 0.004, "step": 63417 }, { "epoch": 19.51, "learning_rate": 3.204049878120286e-08, "loss": 0.0056, "step": 63418 }, { "epoch": 19.51, "learning_rate": 3.200065629770599e-08, "loss": 0.0072, "step": 63419 }, { "epoch": 19.51, "learning_rate": 3.196083856206289e-08, "loss": 0.0025, "step": 63420 }, { "epoch": 19.51, "learning_rate": 3.19210455743757e-08, "loss": 0.0037, "step": 63421 }, { "epoch": 19.51, "learning_rate": 3.188127733473989e-08, "loss": 0.0028, "step": 63422 }, { "epoch": 19.51, "learning_rate": 3.18415338432565e-08, "loss": 0.0026, "step": 63423 }, { "epoch": 19.51, "learning_rate": 3.180181510002212e-08, "loss": 0.0032, "step": 63424 }, { "epoch": 19.52, "learning_rate": 3.1762121105136653e-08, "loss": 0.0023, "step": 63425 }, { "epoch": 19.52, "learning_rate": 3.172245185869893e-08, "loss": 0.0017, "step": 63426 }, { "epoch": 19.52, "learning_rate": 3.168280736080664e-08, "loss": 0.0041, "step": 63427 }, { "epoch": 19.52, "learning_rate": 3.164318761155971e-08, "loss": 0.0025, "step": 63428 }, { "epoch": 19.52, "learning_rate": 3.160359261105472e-08, "loss": 0.0029, "step": 63429 }, { "epoch": 19.52, "learning_rate": 3.156402235938938e-08, "loss": 0.0026, "step": 63430 }, { "epoch": 19.52, "learning_rate": 3.152447685666471e-08, "loss": 0.0033, "step": 63431 }, { "epoch": 19.52, "learning_rate": 3.14849561029773e-08, "loss": 0.0045, "step": 63432 }, { "epoch": 19.52, "learning_rate": 3.144546009842486e-08, "loss": 0.0028, "step": 63433 }, { "epoch": 19.52, "learning_rate": 3.140598884310508e-08, "loss": 0.0084, "step": 63434 }, { "epoch": 19.52, "learning_rate": 3.136654233711789e-08, "loss": 0.0041, "step": 63435 }, { "epoch": 19.52, "learning_rate": 3.132712058055987e-08, "loss": 0.002, "step": 63436 }, { "epoch": 19.52, "learning_rate": 3.128772357352983e-08, "loss": 0.002, "step": 63437 }, { "epoch": 19.52, "learning_rate": 3.124835131612325e-08, "loss": 0.0041, "step": 63438 }, { "epoch": 19.52, "learning_rate": 3.120900380844116e-08, "loss": 0.0041, "step": 63439 }, { "epoch": 19.52, "learning_rate": 3.1169681050579045e-08, "loss": 0.0041, "step": 63440 }, { "epoch": 19.52, "learning_rate": 3.1130383042634604e-08, "loss": 0.0022, "step": 63441 }, { "epoch": 19.52, "learning_rate": 3.109110978470553e-08, "loss": 0.0034, "step": 63442 }, { "epoch": 19.52, "learning_rate": 3.105186127689064e-08, "loss": 0.003, "step": 63443 }, { "epoch": 19.52, "learning_rate": 3.101263751928651e-08, "loss": 0.0031, "step": 63444 }, { "epoch": 19.52, "learning_rate": 3.097343851198975e-08, "loss": 0.0036, "step": 63445 }, { "epoch": 19.52, "learning_rate": 3.0934264255099155e-08, "loss": 0.0035, "step": 63446 }, { "epoch": 19.52, "learning_rate": 3.0895114748710206e-08, "loss": 0.0019, "step": 63447 }, { "epoch": 19.52, "learning_rate": 3.085598999292172e-08, "loss": 0.0037, "step": 63448 }, { "epoch": 19.52, "learning_rate": 3.0816889987830276e-08, "loss": 0.0044, "step": 63449 }, { "epoch": 19.52, "learning_rate": 3.077781473353358e-08, "loss": 0.0025, "step": 63450 }, { "epoch": 19.52, "learning_rate": 3.073876423012712e-08, "loss": 0.003, "step": 63451 }, { "epoch": 19.52, "learning_rate": 3.0699738477708576e-08, "loss": 0.0028, "step": 63452 }, { "epoch": 19.52, "learning_rate": 3.066073747637455e-08, "loss": 0.0139, "step": 63453 }, { "epoch": 19.52, "learning_rate": 3.0621761226222735e-08, "loss": 0.0049, "step": 63454 }, { "epoch": 19.52, "learning_rate": 3.0582809727349725e-08, "loss": 0.0019, "step": 63455 }, { "epoch": 19.52, "learning_rate": 3.0543882979852115e-08, "loss": 0.0026, "step": 63456 }, { "epoch": 19.52, "learning_rate": 3.0504980983825375e-08, "loss": 0.0029, "step": 63457 }, { "epoch": 19.53, "learning_rate": 3.0466103739368316e-08, "loss": 0.0032, "step": 63458 }, { "epoch": 19.53, "learning_rate": 3.0427251246575303e-08, "loss": 0.0026, "step": 63459 }, { "epoch": 19.53, "learning_rate": 3.038842350554405e-08, "loss": 0.0019, "step": 63460 }, { "epoch": 19.53, "learning_rate": 3.034962051637003e-08, "loss": 0.0027, "step": 63461 }, { "epoch": 19.53, "learning_rate": 3.031084227915093e-08, "loss": 0.0034, "step": 63462 }, { "epoch": 19.53, "learning_rate": 3.027208879398113e-08, "loss": 0.0019, "step": 63463 }, { "epoch": 19.53, "learning_rate": 3.023336006095945e-08, "loss": 0.0027, "step": 63464 }, { "epoch": 19.53, "learning_rate": 3.019465608018024e-08, "loss": 0.0049, "step": 63465 }, { "epoch": 19.53, "learning_rate": 3.0155976851740096e-08, "loss": 0.0032, "step": 63466 }, { "epoch": 19.53, "learning_rate": 3.0117322375733394e-08, "loss": 0.0032, "step": 63467 }, { "epoch": 19.53, "learning_rate": 3.0078692652258934e-08, "loss": 0.0024, "step": 63468 }, { "epoch": 19.53, "learning_rate": 3.004008768141109e-08, "loss": 0.0079, "step": 63469 }, { "epoch": 19.53, "learning_rate": 3.000150746328534e-08, "loss": 0.0022, "step": 63470 }, { "epoch": 19.53, "learning_rate": 2.996295199797827e-08, "loss": 0.0029, "step": 63471 }, { "epoch": 19.53, "learning_rate": 2.992442128558537e-08, "loss": 0.003, "step": 63472 }, { "epoch": 19.53, "learning_rate": 2.988591532620322e-08, "loss": 0.0033, "step": 63473 }, { "epoch": 19.53, "learning_rate": 2.984743411992508e-08, "loss": 0.004, "step": 63474 }, { "epoch": 19.53, "learning_rate": 2.9808977666848646e-08, "loss": 0.0029, "step": 63475 }, { "epoch": 19.53, "learning_rate": 2.97705459670683e-08, "loss": 0.0046, "step": 63476 }, { "epoch": 19.53, "learning_rate": 2.9732139020680616e-08, "loss": 0.0021, "step": 63477 }, { "epoch": 19.53, "learning_rate": 2.9693756827778862e-08, "loss": 0.0025, "step": 63478 }, { "epoch": 19.53, "learning_rate": 2.9655399388460736e-08, "loss": 0.003, "step": 63479 }, { "epoch": 19.53, "learning_rate": 2.9617066702819498e-08, "loss": 0.0036, "step": 63480 }, { "epoch": 19.53, "learning_rate": 2.9578758770951732e-08, "loss": 0.0032, "step": 63481 }, { "epoch": 19.53, "learning_rate": 2.954047559295181e-08, "loss": 0.0039, "step": 63482 }, { "epoch": 19.53, "learning_rate": 2.9502217168914104e-08, "loss": 0.0032, "step": 63483 }, { "epoch": 19.53, "learning_rate": 2.94639834989352e-08, "loss": 0.0023, "step": 63484 }, { "epoch": 19.53, "learning_rate": 2.9425774583109467e-08, "loss": 0.003, "step": 63485 }, { "epoch": 19.53, "learning_rate": 2.9387590421531276e-08, "loss": 0.003, "step": 63486 }, { "epoch": 19.53, "learning_rate": 2.9349431014296105e-08, "loss": 0.0041, "step": 63487 }, { "epoch": 19.53, "learning_rate": 2.931129636149721e-08, "loss": 0.0027, "step": 63488 }, { "epoch": 19.53, "learning_rate": 2.9273186463231185e-08, "loss": 0.0033, "step": 63489 }, { "epoch": 19.54, "learning_rate": 2.9235101319592396e-08, "loss": 0.0036, "step": 63490 }, { "epoch": 19.54, "learning_rate": 2.9197040930674102e-08, "loss": 0.0049, "step": 63491 }, { "epoch": 19.54, "learning_rate": 2.9159005296571786e-08, "loss": 0.0027, "step": 63492 }, { "epoch": 19.54, "learning_rate": 2.9120994417378702e-08, "loss": 0.0022, "step": 63493 }, { "epoch": 19.54, "learning_rate": 2.908300829319033e-08, "loss": 0.002, "step": 63494 }, { "epoch": 19.54, "learning_rate": 2.904504692410215e-08, "loss": 0.0038, "step": 63495 }, { "epoch": 19.54, "learning_rate": 2.90071103102052e-08, "loss": 0.002, "step": 63496 }, { "epoch": 19.54, "learning_rate": 2.896919845159607e-08, "loss": 0.0038, "step": 63497 }, { "epoch": 19.54, "learning_rate": 2.893131134836913e-08, "loss": 0.0028, "step": 63498 }, { "epoch": 19.54, "learning_rate": 2.8893449000617636e-08, "loss": 0.0031, "step": 63499 }, { "epoch": 19.54, "learning_rate": 2.8855611408434846e-08, "loss": 0.007, "step": 63500 }, { "epoch": 19.54, "learning_rate": 2.8817798571916244e-08, "loss": 0.0033, "step": 63501 }, { "epoch": 19.54, "learning_rate": 2.8780010491153977e-08, "loss": 0.0052, "step": 63502 }, { "epoch": 19.54, "learning_rate": 2.8742247166242408e-08, "loss": 0.0048, "step": 63503 }, { "epoch": 19.54, "learning_rate": 2.8704508597277024e-08, "loss": 0.0037, "step": 63504 }, { "epoch": 19.54, "learning_rate": 2.866679478434886e-08, "loss": 0.0032, "step": 63505 }, { "epoch": 19.54, "learning_rate": 2.8629105727553396e-08, "loss": 0.0041, "step": 63506 }, { "epoch": 19.54, "learning_rate": 2.8591441426983892e-08, "loss": 0.0023, "step": 63507 }, { "epoch": 19.54, "learning_rate": 2.85538018827336e-08, "loss": 0.0035, "step": 63508 }, { "epoch": 19.54, "learning_rate": 2.8516187094895787e-08, "loss": 0.0024, "step": 63509 }, { "epoch": 19.54, "learning_rate": 2.8478597063564817e-08, "loss": 0.0032, "step": 63510 }, { "epoch": 19.54, "learning_rate": 2.844103178883284e-08, "loss": 0.003, "step": 63511 }, { "epoch": 19.54, "learning_rate": 2.8403491270792007e-08, "loss": 0.0035, "step": 63512 }, { "epoch": 19.54, "learning_rate": 2.8365975509538902e-08, "loss": 0.0036, "step": 63513 }, { "epoch": 19.54, "learning_rate": 2.8328484505164566e-08, "loss": 0.0024, "step": 63514 }, { "epoch": 19.54, "learning_rate": 2.8291018257762258e-08, "loss": 0.0026, "step": 63515 }, { "epoch": 19.54, "learning_rate": 2.825357676742413e-08, "loss": 0.0019, "step": 63516 }, { "epoch": 19.54, "learning_rate": 2.8216160034244544e-08, "loss": 0.0048, "step": 63517 }, { "epoch": 19.54, "learning_rate": 2.8178768058316762e-08, "loss": 0.0039, "step": 63518 }, { "epoch": 19.54, "learning_rate": 2.8141400839731826e-08, "loss": 0.0031, "step": 63519 }, { "epoch": 19.54, "learning_rate": 2.81040583785841e-08, "loss": 0.0038, "step": 63520 }, { "epoch": 19.54, "learning_rate": 2.806674067496573e-08, "loss": 0.0029, "step": 63521 }, { "epoch": 19.54, "learning_rate": 2.8029447728968872e-08, "loss": 0.0034, "step": 63522 }, { "epoch": 19.55, "learning_rate": 2.7992179540685672e-08, "loss": 0.0026, "step": 63523 }, { "epoch": 19.55, "learning_rate": 2.7954936110210495e-08, "loss": 0.0022, "step": 63524 }, { "epoch": 19.55, "learning_rate": 2.7917717437635495e-08, "loss": 0.0034, "step": 63525 }, { "epoch": 19.55, "learning_rate": 2.7880523523051706e-08, "loss": 0.0079, "step": 63526 }, { "epoch": 19.55, "learning_rate": 2.7843354366551278e-08, "loss": 0.0036, "step": 63527 }, { "epoch": 19.55, "learning_rate": 2.780620996822858e-08, "loss": 0.0025, "step": 63528 }, { "epoch": 19.55, "learning_rate": 2.776909032817465e-08, "loss": 0.0032, "step": 63529 }, { "epoch": 19.55, "learning_rate": 2.773199544648164e-08, "loss": 0.0041, "step": 63530 }, { "epoch": 19.55, "learning_rate": 2.7694925323240584e-08, "loss": 0.0029, "step": 63531 }, { "epoch": 19.55, "learning_rate": 2.7657879958545854e-08, "loss": 0.0023, "step": 63532 }, { "epoch": 19.55, "learning_rate": 2.7620859352487372e-08, "loss": 0.003, "step": 63533 }, { "epoch": 19.55, "learning_rate": 2.7583863505158403e-08, "loss": 0.0023, "step": 63534 }, { "epoch": 19.55, "learning_rate": 2.7546892416649985e-08, "loss": 0.0038, "step": 63535 }, { "epoch": 19.55, "learning_rate": 2.750994608705315e-08, "loss": 0.0038, "step": 63536 }, { "epoch": 19.55, "learning_rate": 2.7473024516461166e-08, "loss": 0.0047, "step": 63537 }, { "epoch": 19.55, "learning_rate": 2.7436127704966175e-08, "loss": 0.0032, "step": 63538 }, { "epoch": 19.55, "learning_rate": 2.7399255652658107e-08, "loss": 0.0025, "step": 63539 }, { "epoch": 19.55, "learning_rate": 2.736240835962911e-08, "loss": 0.0025, "step": 63540 }, { "epoch": 19.55, "learning_rate": 2.7325585825970225e-08, "loss": 0.0019, "step": 63541 }, { "epoch": 19.55, "learning_rate": 2.7288788051774706e-08, "loss": 0.0044, "step": 63542 }, { "epoch": 19.55, "learning_rate": 2.7252015037131373e-08, "loss": 0.0033, "step": 63543 }, { "epoch": 19.55, "learning_rate": 2.7215266782133486e-08, "loss": 0.0037, "step": 63544 }, { "epoch": 19.55, "learning_rate": 2.7178543286870972e-08, "loss": 0.0014, "step": 63545 }, { "epoch": 19.55, "learning_rate": 2.714184455143487e-08, "loss": 0.0034, "step": 63546 }, { "epoch": 19.55, "learning_rate": 2.7105170575918438e-08, "loss": 0.0034, "step": 63547 }, { "epoch": 19.55, "learning_rate": 2.7068521360410493e-08, "loss": 0.0028, "step": 63548 }, { "epoch": 19.55, "learning_rate": 2.7031896905003184e-08, "loss": 0.0028, "step": 63549 }, { "epoch": 19.55, "learning_rate": 2.6995297209786443e-08, "loss": 0.0029, "step": 63550 }, { "epoch": 19.55, "learning_rate": 2.695872227485241e-08, "loss": 0.0022, "step": 63551 }, { "epoch": 19.55, "learning_rate": 2.692217210028991e-08, "loss": 0.0031, "step": 63552 }, { "epoch": 19.55, "learning_rate": 2.6885646686192203e-08, "loss": 0.0043, "step": 63553 }, { "epoch": 19.55, "learning_rate": 2.68491460326481e-08, "loss": 0.0043, "step": 63554 }, { "epoch": 19.56, "learning_rate": 2.6812670139749753e-08, "loss": 0.0031, "step": 63555 }, { "epoch": 19.56, "learning_rate": 2.677621900758709e-08, "loss": 0.0024, "step": 63556 }, { "epoch": 19.56, "learning_rate": 2.673979263624893e-08, "loss": 0.0019, "step": 63557 }, { "epoch": 19.56, "learning_rate": 2.6703391025827417e-08, "loss": 0.0033, "step": 63558 }, { "epoch": 19.56, "learning_rate": 2.6667014176413598e-08, "loss": 0.0025, "step": 63559 }, { "epoch": 19.56, "learning_rate": 2.663066208809628e-08, "loss": 0.0036, "step": 63560 }, { "epoch": 19.56, "learning_rate": 2.6594334760966512e-08, "loss": 0.0035, "step": 63561 }, { "epoch": 19.56, "learning_rate": 2.6558032195114214e-08, "loss": 0.0018, "step": 63562 }, { "epoch": 19.56, "learning_rate": 2.6521754390629316e-08, "loss": 0.0031, "step": 63563 }, { "epoch": 19.56, "learning_rate": 2.648550134760286e-08, "loss": 0.0018, "step": 63564 }, { "epoch": 19.56, "learning_rate": 2.644927306612366e-08, "loss": 0.0036, "step": 63565 }, { "epoch": 19.56, "learning_rate": 2.6413069546281646e-08, "loss": 0.0029, "step": 63566 }, { "epoch": 19.56, "learning_rate": 2.6376890788167853e-08, "loss": 0.0042, "step": 63567 }, { "epoch": 19.56, "learning_rate": 2.63407367918711e-08, "loss": 0.0022, "step": 63568 }, { "epoch": 19.56, "learning_rate": 2.6304607557481322e-08, "loss": 0.0042, "step": 63569 }, { "epoch": 19.56, "learning_rate": 2.6268503085089547e-08, "loss": 0.0032, "step": 63570 }, { "epoch": 19.56, "learning_rate": 2.623242337478238e-08, "loss": 0.0048, "step": 63571 }, { "epoch": 19.56, "learning_rate": 2.6196368426651965e-08, "loss": 0.0039, "step": 63572 }, { "epoch": 19.56, "learning_rate": 2.616033824078823e-08, "loss": 0.0022, "step": 63573 }, { "epoch": 19.56, "learning_rate": 2.6124332817277774e-08, "loss": 0.0032, "step": 63574 }, { "epoch": 19.56, "learning_rate": 2.608835215621275e-08, "loss": 0.0053, "step": 63575 }, { "epoch": 19.56, "learning_rate": 2.6052396257681966e-08, "loss": 0.0036, "step": 63576 }, { "epoch": 19.56, "learning_rate": 2.6016465121773138e-08, "loss": 0.0024, "step": 63577 }, { "epoch": 19.56, "learning_rate": 2.59805587485773e-08, "loss": 0.0058, "step": 63578 }, { "epoch": 19.56, "learning_rate": 2.594467713818216e-08, "loss": 0.0043, "step": 63579 }, { "epoch": 19.56, "learning_rate": 2.590882029067765e-08, "loss": 0.0031, "step": 63580 }, { "epoch": 19.56, "learning_rate": 2.587298820615258e-08, "loss": 0.0024, "step": 63581 }, { "epoch": 19.56, "learning_rate": 2.583718088469689e-08, "loss": 0.0035, "step": 63582 }, { "epoch": 19.56, "learning_rate": 2.5801398326397163e-08, "loss": 0.0044, "step": 63583 }, { "epoch": 19.56, "learning_rate": 2.5765640531344448e-08, "loss": 0.0048, "step": 63584 }, { "epoch": 19.56, "learning_rate": 2.572990749962645e-08, "loss": 0.0036, "step": 63585 }, { "epoch": 19.56, "learning_rate": 2.5694199231331986e-08, "loss": 0.0043, "step": 63586 }, { "epoch": 19.56, "learning_rate": 2.565851572654876e-08, "loss": 0.0047, "step": 63587 }, { "epoch": 19.57, "learning_rate": 2.5622856985367818e-08, "loss": 0.0053, "step": 63588 }, { "epoch": 19.57, "learning_rate": 2.558722300787686e-08, "loss": 0.0021, "step": 63589 }, { "epoch": 19.57, "learning_rate": 2.555161379416249e-08, "loss": 0.0028, "step": 63590 }, { "epoch": 19.57, "learning_rate": 2.551602934431574e-08, "loss": 0.0019, "step": 63591 }, { "epoch": 19.57, "learning_rate": 2.5480469658422104e-08, "loss": 0.0033, "step": 63592 }, { "epoch": 19.57, "learning_rate": 2.5444934736572614e-08, "loss": 0.0035, "step": 63593 }, { "epoch": 19.57, "learning_rate": 2.540942457885387e-08, "loss": 0.0031, "step": 63594 }, { "epoch": 19.57, "learning_rate": 2.537393918535358e-08, "loss": 0.0021, "step": 63595 }, { "epoch": 19.57, "learning_rate": 2.5338478556161673e-08, "loss": 0.0038, "step": 63596 }, { "epoch": 19.57, "learning_rate": 2.5303042691365853e-08, "loss": 0.0034, "step": 63597 }, { "epoch": 19.57, "learning_rate": 2.526763159105272e-08, "loss": 0.0048, "step": 63598 }, { "epoch": 19.57, "learning_rate": 2.5232245255311095e-08, "loss": 0.0024, "step": 63599 }, { "epoch": 19.57, "learning_rate": 2.5196883684228678e-08, "loss": 0.0054, "step": 63600 }, { "epoch": 19.57, "learning_rate": 2.5161546877893185e-08, "loss": 0.0042, "step": 63601 }, { "epoch": 19.57, "learning_rate": 2.5126234836392315e-08, "loss": 0.0051, "step": 63602 }, { "epoch": 19.57, "learning_rate": 2.5090947559813782e-08, "loss": 0.0044, "step": 63603 }, { "epoch": 19.57, "learning_rate": 2.5055685048245294e-08, "loss": 0.0036, "step": 63604 }, { "epoch": 19.57, "learning_rate": 2.5020447301774554e-08, "loss": 0.0016, "step": 63605 }, { "epoch": 19.57, "learning_rate": 2.4985234320489272e-08, "loss": 0.0067, "step": 63606 }, { "epoch": 19.57, "learning_rate": 2.4950046104476045e-08, "loss": 0.0041, "step": 63607 }, { "epoch": 19.57, "learning_rate": 2.4914882653822582e-08, "loss": 0.0025, "step": 63608 }, { "epoch": 19.57, "learning_rate": 2.487974396861659e-08, "loss": 0.0019, "step": 63609 }, { "epoch": 19.57, "learning_rate": 2.4844630048944664e-08, "loss": 0.0036, "step": 63610 }, { "epoch": 19.57, "learning_rate": 2.480954089489451e-08, "loss": 0.0021, "step": 63611 }, { "epoch": 19.57, "learning_rate": 2.4774476506552735e-08, "loss": 0.0045, "step": 63612 }, { "epoch": 19.57, "learning_rate": 2.473943688400704e-08, "loss": 0.0035, "step": 63613 }, { "epoch": 19.57, "learning_rate": 2.4704422027344023e-08, "loss": 0.0023, "step": 63614 }, { "epoch": 19.57, "learning_rate": 2.466943193665028e-08, "loss": 0.0025, "step": 63615 }, { "epoch": 19.57, "learning_rate": 2.463446661201352e-08, "loss": 0.0026, "step": 63616 }, { "epoch": 19.57, "learning_rate": 2.459952605352034e-08, "loss": 0.0039, "step": 63617 }, { "epoch": 19.57, "learning_rate": 2.456461026125623e-08, "loss": 0.003, "step": 63618 }, { "epoch": 19.57, "learning_rate": 2.452971923531e-08, "loss": 0.0035, "step": 63619 }, { "epoch": 19.58, "learning_rate": 2.4494852975767146e-08, "loss": 0.0042, "step": 63620 }, { "epoch": 19.58, "learning_rate": 2.4460011482713153e-08, "loss": 0.0037, "step": 63621 }, { "epoch": 19.58, "learning_rate": 2.4425194756236836e-08, "loss": 0.0036, "step": 63622 }, { "epoch": 19.58, "learning_rate": 2.4390402796422574e-08, "loss": 0.0031, "step": 63623 }, { "epoch": 19.58, "learning_rate": 2.4355635603358075e-08, "loss": 0.0029, "step": 63624 }, { "epoch": 19.58, "learning_rate": 2.4320893177128822e-08, "loss": 0.0033, "step": 63625 }, { "epoch": 19.58, "learning_rate": 2.428617551782253e-08, "loss": 0.0039, "step": 63626 }, { "epoch": 19.58, "learning_rate": 2.425148262552246e-08, "loss": 0.0018, "step": 63627 }, { "epoch": 19.58, "learning_rate": 2.4216814500317432e-08, "loss": 0.0036, "step": 63628 }, { "epoch": 19.58, "learning_rate": 2.4182171142292932e-08, "loss": 0.0028, "step": 63629 }, { "epoch": 19.58, "learning_rate": 2.4147552551534448e-08, "loss": 0.0027, "step": 63630 }, { "epoch": 19.58, "learning_rate": 2.411295872812858e-08, "loss": 0.0031, "step": 63631 }, { "epoch": 19.58, "learning_rate": 2.4078389672160805e-08, "loss": 0.0031, "step": 63632 }, { "epoch": 19.58, "learning_rate": 2.4043845383716623e-08, "loss": 0.0047, "step": 63633 }, { "epoch": 19.58, "learning_rate": 2.4009325862881517e-08, "loss": 0.0045, "step": 63634 }, { "epoch": 19.58, "learning_rate": 2.397483110974319e-08, "loss": 0.0034, "step": 63635 }, { "epoch": 19.58, "learning_rate": 2.3940361124384913e-08, "loss": 0.0034, "step": 63636 }, { "epoch": 19.58, "learning_rate": 2.3905915906893284e-08, "loss": 0.0043, "step": 63637 }, { "epoch": 19.58, "learning_rate": 2.38714954573549e-08, "loss": 0.0038, "step": 63638 }, { "epoch": 19.58, "learning_rate": 2.3837099775853022e-08, "loss": 0.0039, "step": 63639 }, { "epoch": 19.58, "learning_rate": 2.3802728862474256e-08, "loss": 0.0038, "step": 63640 }, { "epoch": 19.58, "learning_rate": 2.3768382717304085e-08, "loss": 0.0035, "step": 63641 }, { "epoch": 19.58, "learning_rate": 2.3734061340426885e-08, "loss": 0.0027, "step": 63642 }, { "epoch": 19.58, "learning_rate": 2.3699764731929254e-08, "loss": 0.0047, "step": 63643 }, { "epoch": 19.58, "learning_rate": 2.366549289189557e-08, "loss": 0.004, "step": 63644 }, { "epoch": 19.58, "learning_rate": 2.363124582041021e-08, "loss": 0.0027, "step": 63645 }, { "epoch": 19.58, "learning_rate": 2.359702351755977e-08, "loss": 0.0031, "step": 63646 }, { "epoch": 19.58, "learning_rate": 2.3562825983427517e-08, "loss": 0.0039, "step": 63647 }, { "epoch": 19.58, "learning_rate": 2.352865321810005e-08, "loss": 0.0057, "step": 63648 }, { "epoch": 19.58, "learning_rate": 2.3494505221660635e-08, "loss": 0.0019, "step": 63649 }, { "epoch": 19.58, "learning_rate": 2.346038199419476e-08, "loss": 0.0019, "step": 63650 }, { "epoch": 19.58, "learning_rate": 2.3426283535787907e-08, "loss": 0.0032, "step": 63651 }, { "epoch": 19.58, "learning_rate": 2.339220984652335e-08, "loss": 0.0025, "step": 63652 }, { "epoch": 19.59, "learning_rate": 2.335816092648657e-08, "loss": 0.0039, "step": 63653 }, { "epoch": 19.59, "learning_rate": 2.3324136775761952e-08, "loss": 0.002, "step": 63654 }, { "epoch": 19.59, "learning_rate": 2.3290137394433865e-08, "loss": 0.0029, "step": 63655 }, { "epoch": 19.59, "learning_rate": 2.325616278258669e-08, "loss": 0.0045, "step": 63656 }, { "epoch": 19.59, "learning_rate": 2.32222129403048e-08, "loss": 0.0029, "step": 63657 }, { "epoch": 19.59, "learning_rate": 2.3188287867672577e-08, "loss": 0.0035, "step": 63658 }, { "epoch": 19.59, "learning_rate": 2.3154387564774396e-08, "loss": 0.0021, "step": 63659 }, { "epoch": 19.59, "learning_rate": 2.312051203169352e-08, "loss": 0.0029, "step": 63660 }, { "epoch": 19.59, "learning_rate": 2.3086661268514333e-08, "loss": 0.0043, "step": 63661 }, { "epoch": 19.59, "learning_rate": 2.3052835275322317e-08, "loss": 0.009, "step": 63662 }, { "epoch": 19.59, "learning_rate": 2.3019034052200738e-08, "loss": 0.0036, "step": 63663 }, { "epoch": 19.59, "learning_rate": 2.2985257599231757e-08, "loss": 0.0036, "step": 63664 }, { "epoch": 19.59, "learning_rate": 2.295150591650197e-08, "loss": 0.0042, "step": 63665 }, { "epoch": 19.59, "learning_rate": 2.291777900409353e-08, "loss": 0.0039, "step": 63666 }, { "epoch": 19.59, "learning_rate": 2.2884076862089712e-08, "loss": 0.0039, "step": 63667 }, { "epoch": 19.59, "learning_rate": 2.285039949057599e-08, "loss": 0.0033, "step": 63668 }, { "epoch": 19.59, "learning_rate": 2.2816746889634532e-08, "loss": 0.0022, "step": 63669 }, { "epoch": 19.59, "learning_rate": 2.278311905934971e-08, "loss": 0.0049, "step": 63670 }, { "epoch": 19.59, "learning_rate": 2.2749515999803683e-08, "loss": 0.0019, "step": 63671 }, { "epoch": 19.59, "learning_rate": 2.2715937711081935e-08, "loss": 0.0027, "step": 63672 }, { "epoch": 19.59, "learning_rate": 2.2682384193266625e-08, "loss": 0.0039, "step": 63673 }, { "epoch": 19.59, "learning_rate": 2.264885544643991e-08, "loss": 0.0039, "step": 63674 }, { "epoch": 19.59, "learning_rate": 2.2615351470687276e-08, "loss": 0.003, "step": 63675 }, { "epoch": 19.59, "learning_rate": 2.2581872266090876e-08, "loss": 0.0051, "step": 63676 }, { "epoch": 19.59, "learning_rate": 2.254841783273398e-08, "loss": 0.0038, "step": 63677 }, { "epoch": 19.59, "learning_rate": 2.2514988170698748e-08, "loss": 0.0045, "step": 63678 }, { "epoch": 19.59, "learning_rate": 2.2481583280069553e-08, "loss": 0.0022, "step": 63679 }, { "epoch": 19.59, "learning_rate": 2.244820316092855e-08, "loss": 0.0023, "step": 63680 }, { "epoch": 19.59, "learning_rate": 2.241484781335901e-08, "loss": 0.0017, "step": 63681 }, { "epoch": 19.59, "learning_rate": 2.2381517237443085e-08, "loss": 0.0073, "step": 63682 }, { "epoch": 19.59, "learning_rate": 2.2348211433265155e-08, "loss": 0.0046, "step": 63683 }, { "epoch": 19.59, "learning_rate": 2.2314930400905153e-08, "loss": 0.0053, "step": 63684 }, { "epoch": 19.6, "learning_rate": 2.2281674140448574e-08, "loss": 0.0043, "step": 63685 }, { "epoch": 19.6, "learning_rate": 2.2248442651975345e-08, "loss": 0.0022, "step": 63686 }, { "epoch": 19.6, "learning_rate": 2.2215235935570955e-08, "loss": 0.0035, "step": 63687 }, { "epoch": 19.6, "learning_rate": 2.2182053991315345e-08, "loss": 0.0039, "step": 63688 }, { "epoch": 19.6, "learning_rate": 2.2148896819291776e-08, "loss": 0.0028, "step": 63689 }, { "epoch": 19.6, "learning_rate": 2.2115764419582408e-08, "loss": 0.0039, "step": 63690 }, { "epoch": 19.6, "learning_rate": 2.2082656792269398e-08, "loss": 0.0034, "step": 63691 }, { "epoch": 19.6, "learning_rate": 2.2049573937436008e-08, "loss": 0.0045, "step": 63692 }, { "epoch": 19.6, "learning_rate": 2.201651585516218e-08, "loss": 0.0029, "step": 63693 }, { "epoch": 19.6, "learning_rate": 2.1983482545532285e-08, "loss": 0.0023, "step": 63694 }, { "epoch": 19.6, "learning_rate": 2.1950474008628487e-08, "loss": 0.0029, "step": 63695 }, { "epoch": 19.6, "learning_rate": 2.1917490244529605e-08, "loss": 0.0041, "step": 63696 }, { "epoch": 19.6, "learning_rate": 2.1884531253321128e-08, "loss": 0.0024, "step": 63697 }, { "epoch": 19.6, "learning_rate": 2.1851597035081884e-08, "loss": 0.0047, "step": 63698 }, { "epoch": 19.6, "learning_rate": 2.1818687589896248e-08, "loss": 0.0048, "step": 63699 }, { "epoch": 19.6, "learning_rate": 2.1785802917845266e-08, "loss": 0.0032, "step": 63700 }, { "epoch": 19.6, "learning_rate": 2.1752943019008876e-08, "loss": 0.0041, "step": 63701 }, { "epoch": 19.6, "learning_rate": 2.1720107893470344e-08, "loss": 0.0039, "step": 63702 }, { "epoch": 19.6, "learning_rate": 2.1687297541310715e-08, "loss": 0.0029, "step": 63703 }, { "epoch": 19.6, "learning_rate": 2.1654511962611037e-08, "loss": 0.0037, "step": 63704 }, { "epoch": 19.6, "learning_rate": 2.1621751157454574e-08, "loss": 0.0044, "step": 63705 }, { "epoch": 19.6, "learning_rate": 2.1589015125920153e-08, "loss": 0.0042, "step": 63706 }, { "epoch": 19.6, "learning_rate": 2.1556303868089933e-08, "loss": 0.002, "step": 63707 }, { "epoch": 19.6, "learning_rate": 2.152361738404496e-08, "loss": 0.0044, "step": 63708 }, { "epoch": 19.6, "learning_rate": 2.1490955673867387e-08, "loss": 0.002, "step": 63709 }, { "epoch": 19.6, "learning_rate": 2.1458318737638263e-08, "loss": 0.0042, "step": 63710 }, { "epoch": 19.6, "learning_rate": 2.142570657543641e-08, "loss": 0.002, "step": 63711 }, { "epoch": 19.6, "learning_rate": 2.1393119187345103e-08, "loss": 0.0035, "step": 63712 }, { "epoch": 19.6, "learning_rate": 2.1360556573445378e-08, "loss": 0.0033, "step": 63713 }, { "epoch": 19.6, "learning_rate": 2.132801873381607e-08, "loss": 0.0036, "step": 63714 }, { "epoch": 19.6, "learning_rate": 2.129550566854044e-08, "loss": 0.0021, "step": 63715 }, { "epoch": 19.6, "learning_rate": 2.1263017377696205e-08, "loss": 0.0029, "step": 63716 }, { "epoch": 19.6, "learning_rate": 2.123055386136663e-08, "loss": 0.0054, "step": 63717 }, { "epoch": 19.61, "learning_rate": 2.1198115119631657e-08, "loss": 0.002, "step": 63718 }, { "epoch": 19.61, "learning_rate": 2.1165701152571217e-08, "loss": 0.0023, "step": 63719 }, { "epoch": 19.61, "learning_rate": 2.1133311960266358e-08, "loss": 0.0032, "step": 63720 }, { "epoch": 19.61, "learning_rate": 2.110094754279701e-08, "loss": 0.0043, "step": 63721 }, { "epoch": 19.61, "learning_rate": 2.1068607900243122e-08, "loss": 0.0041, "step": 63722 }, { "epoch": 19.61, "learning_rate": 2.1036293032686837e-08, "loss": 0.0028, "step": 63723 }, { "epoch": 19.61, "learning_rate": 2.100400294020699e-08, "loss": 0.0044, "step": 63724 }, { "epoch": 19.61, "learning_rate": 2.0971737622883515e-08, "loss": 0.0033, "step": 63725 }, { "epoch": 19.61, "learning_rate": 2.0939497080797456e-08, "loss": 0.0028, "step": 63726 }, { "epoch": 19.61, "learning_rate": 2.090728131402764e-08, "loss": 0.0031, "step": 63727 }, { "epoch": 19.61, "learning_rate": 2.087509032265511e-08, "loss": 0.0048, "step": 63728 }, { "epoch": 19.61, "learning_rate": 2.084292410675981e-08, "loss": 0.0047, "step": 63729 }, { "epoch": 19.61, "learning_rate": 2.081078266642056e-08, "loss": 0.0053, "step": 63730 }, { "epoch": 19.61, "learning_rate": 2.0778666001718405e-08, "loss": 0.0035, "step": 63731 }, { "epoch": 19.61, "learning_rate": 2.0746574112732178e-08, "loss": 0.0047, "step": 63732 }, { "epoch": 19.61, "learning_rate": 2.0714506999541807e-08, "loss": 0.0026, "step": 63733 }, { "epoch": 19.61, "learning_rate": 2.068246466222723e-08, "loss": 0.0024, "step": 63734 }, { "epoch": 19.61, "learning_rate": 2.0650447100867277e-08, "loss": 0.0037, "step": 63735 }, { "epoch": 19.61, "learning_rate": 2.061845431554188e-08, "loss": 0.0025, "step": 63736 }, { "epoch": 19.61, "learning_rate": 2.0586486306330977e-08, "loss": 0.0038, "step": 63737 }, { "epoch": 19.61, "learning_rate": 2.0554543073313393e-08, "loss": 0.0015, "step": 63738 }, { "epoch": 19.61, "learning_rate": 2.0522624616569064e-08, "loss": 0.0048, "step": 63739 }, { "epoch": 19.61, "learning_rate": 2.049073093617571e-08, "loss": 0.0035, "step": 63740 }, { "epoch": 19.61, "learning_rate": 2.0458862032214367e-08, "loss": 0.0037, "step": 63741 }, { "epoch": 19.61, "learning_rate": 2.042701790476276e-08, "loss": 0.003, "step": 63742 }, { "epoch": 19.61, "learning_rate": 2.039519855390082e-08, "loss": 0.0031, "step": 63743 }, { "epoch": 19.61, "learning_rate": 2.0363403979706263e-08, "loss": 0.0041, "step": 63744 }, { "epoch": 19.61, "learning_rate": 2.033163418226014e-08, "loss": 0.003, "step": 63745 }, { "epoch": 19.61, "learning_rate": 2.0299889161639053e-08, "loss": 0.0039, "step": 63746 }, { "epoch": 19.61, "learning_rate": 2.0268168917924047e-08, "loss": 0.0039, "step": 63747 }, { "epoch": 19.61, "learning_rate": 2.023647345119173e-08, "loss": 0.0034, "step": 63748 }, { "epoch": 19.61, "learning_rate": 2.0204802761522034e-08, "loss": 0.0035, "step": 63749 }, { "epoch": 19.62, "learning_rate": 2.017315684899379e-08, "loss": 0.0042, "step": 63750 }, { "epoch": 19.62, "learning_rate": 2.01415357136836e-08, "loss": 0.0031, "step": 63751 }, { "epoch": 19.62, "learning_rate": 2.0109939355673624e-08, "loss": 0.004, "step": 63752 }, { "epoch": 19.62, "learning_rate": 2.0078367775038244e-08, "loss": 0.0029, "step": 63753 }, { "epoch": 19.62, "learning_rate": 2.0046820971859617e-08, "loss": 0.0027, "step": 63754 }, { "epoch": 19.62, "learning_rate": 2.0015298946213236e-08, "loss": 0.0032, "step": 63755 }, { "epoch": 19.62, "learning_rate": 1.998380169817793e-08, "loss": 0.0045, "step": 63756 }, { "epoch": 19.62, "learning_rate": 1.9952329227832525e-08, "loss": 0.0029, "step": 63757 }, { "epoch": 19.62, "learning_rate": 1.992088153525584e-08, "loss": 0.0033, "step": 63758 }, { "epoch": 19.62, "learning_rate": 1.9889458620524493e-08, "loss": 0.0046, "step": 63759 }, { "epoch": 19.62, "learning_rate": 1.98580604837173e-08, "loss": 0.0043, "step": 63760 }, { "epoch": 19.62, "learning_rate": 1.982668712491087e-08, "loss": 0.0047, "step": 63761 }, { "epoch": 19.62, "learning_rate": 1.979533854418514e-08, "loss": 0.002, "step": 63762 }, { "epoch": 19.62, "learning_rate": 1.976401474161782e-08, "loss": 0.0038, "step": 63763 }, { "epoch": 19.62, "learning_rate": 1.973271571728441e-08, "loss": 0.0042, "step": 63764 }, { "epoch": 19.62, "learning_rate": 1.970144147126485e-08, "loss": 0.0024, "step": 63765 }, { "epoch": 19.62, "learning_rate": 1.9670192003635735e-08, "loss": 0.0046, "step": 63766 }, { "epoch": 19.62, "learning_rate": 1.963896731447479e-08, "loss": 0.0016, "step": 63767 }, { "epoch": 19.62, "learning_rate": 1.9607767403859722e-08, "loss": 0.0086, "step": 63768 }, { "epoch": 19.62, "learning_rate": 1.9576592271868257e-08, "loss": 0.0029, "step": 63769 }, { "epoch": 19.62, "learning_rate": 1.9545441918576992e-08, "loss": 0.0021, "step": 63770 }, { "epoch": 19.62, "learning_rate": 1.9514316344063645e-08, "loss": 0.0037, "step": 63771 }, { "epoch": 19.62, "learning_rate": 1.9483215548405932e-08, "loss": 0.0036, "step": 63772 }, { "epoch": 19.62, "learning_rate": 1.9452139531681568e-08, "loss": 0.0025, "step": 63773 }, { "epoch": 19.62, "learning_rate": 1.942108829396605e-08, "loss": 0.0053, "step": 63774 }, { "epoch": 19.62, "learning_rate": 1.9390061835337093e-08, "loss": 0.0029, "step": 63775 }, { "epoch": 19.62, "learning_rate": 1.935906015587241e-08, "loss": 0.0033, "step": 63776 }, { "epoch": 19.62, "learning_rate": 1.93280832556475e-08, "loss": 0.0037, "step": 63777 }, { "epoch": 19.62, "learning_rate": 1.9297131134741186e-08, "loss": 0.0029, "step": 63778 }, { "epoch": 19.62, "learning_rate": 1.9266203793228965e-08, "loss": 0.0029, "step": 63779 }, { "epoch": 19.62, "learning_rate": 1.9235301231188554e-08, "loss": 0.0045, "step": 63780 }, { "epoch": 19.62, "learning_rate": 1.9204423448696553e-08, "loss": 0.0027, "step": 63781 }, { "epoch": 19.62, "learning_rate": 1.917357044582846e-08, "loss": 0.0022, "step": 63782 }, { "epoch": 19.63, "learning_rate": 1.9142742222660883e-08, "loss": 0.0035, "step": 63783 }, { "epoch": 19.63, "learning_rate": 1.9111938779272644e-08, "loss": 0.0028, "step": 63784 }, { "epoch": 19.63, "learning_rate": 1.9081160115738127e-08, "loss": 0.0038, "step": 63785 }, { "epoch": 19.63, "learning_rate": 1.9050406232133944e-08, "loss": 0.005, "step": 63786 }, { "epoch": 19.63, "learning_rate": 1.9019677128536694e-08, "loss": 0.0031, "step": 63787 }, { "epoch": 19.63, "learning_rate": 1.8988972805022988e-08, "loss": 0.0024, "step": 63788 }, { "epoch": 19.63, "learning_rate": 1.8958293261669426e-08, "loss": 0.0039, "step": 63789 }, { "epoch": 19.63, "learning_rate": 1.8927638498551502e-08, "loss": 0.0038, "step": 63790 }, { "epoch": 19.63, "learning_rate": 1.889700851574472e-08, "loss": 0.0024, "step": 63791 }, { "epoch": 19.63, "learning_rate": 1.8866403313325677e-08, "loss": 0.0045, "step": 63792 }, { "epoch": 19.63, "learning_rate": 1.883582289137209e-08, "loss": 0.0034, "step": 63793 }, { "epoch": 19.63, "learning_rate": 1.8805267249957237e-08, "loss": 0.002, "step": 63794 }, { "epoch": 19.63, "learning_rate": 1.8774736389157723e-08, "loss": 0.0036, "step": 63795 }, { "epoch": 19.63, "learning_rate": 1.8744230309050148e-08, "loss": 0.0072, "step": 63796 }, { "epoch": 19.63, "learning_rate": 1.8713749009710013e-08, "loss": 0.0041, "step": 63797 }, { "epoch": 19.63, "learning_rate": 1.8683292491212812e-08, "loss": 0.002, "step": 63798 }, { "epoch": 19.63, "learning_rate": 1.865286075363404e-08, "loss": 0.0029, "step": 63799 }, { "epoch": 19.63, "learning_rate": 1.862245379704919e-08, "loss": 0.0055, "step": 63800 }, { "epoch": 19.63, "learning_rate": 1.8592071621532648e-08, "loss": 0.0029, "step": 63801 }, { "epoch": 19.63, "learning_rate": 1.856171422716324e-08, "loss": 0.0034, "step": 63802 }, { "epoch": 19.63, "learning_rate": 1.8531381614013133e-08, "loss": 0.0048, "step": 63803 }, { "epoch": 19.63, "learning_rate": 1.850107378215893e-08, "loss": 0.0043, "step": 63804 }, { "epoch": 19.63, "learning_rate": 1.8470790731675014e-08, "loss": 0.0028, "step": 63805 }, { "epoch": 19.63, "learning_rate": 1.8440532462637993e-08, "loss": 0.0039, "step": 63806 }, { "epoch": 19.63, "learning_rate": 1.8410298975121144e-08, "loss": 0.0021, "step": 63807 }, { "epoch": 19.63, "learning_rate": 1.838009026920107e-08, "loss": 0.0031, "step": 63808 }, { "epoch": 19.63, "learning_rate": 1.834990634495326e-08, "loss": 0.0042, "step": 63809 }, { "epoch": 19.63, "learning_rate": 1.8319747202451e-08, "loss": 0.0017, "step": 63810 }, { "epoch": 19.63, "learning_rate": 1.8289612841768666e-08, "loss": 0.0038, "step": 63811 }, { "epoch": 19.63, "learning_rate": 1.825950326298398e-08, "loss": 0.0038, "step": 63812 }, { "epoch": 19.63, "learning_rate": 1.82294184661691e-08, "loss": 0.004, "step": 63813 }, { "epoch": 19.63, "learning_rate": 1.819935845139953e-08, "loss": 0.0029, "step": 63814 }, { "epoch": 19.64, "learning_rate": 1.8169323218749646e-08, "loss": 0.0033, "step": 63815 }, { "epoch": 19.64, "learning_rate": 1.8139312768294946e-08, "loss": 0.0034, "step": 63816 }, { "epoch": 19.64, "learning_rate": 1.810932710010982e-08, "loss": 0.0033, "step": 63817 }, { "epoch": 19.64, "learning_rate": 1.8079366214267536e-08, "loss": 0.004, "step": 63818 }, { "epoch": 19.64, "learning_rate": 1.804943011084248e-08, "loss": 0.0041, "step": 63819 }, { "epoch": 19.64, "learning_rate": 1.801951878991015e-08, "loss": 0.0031, "step": 63820 }, { "epoch": 19.64, "learning_rate": 1.7989632251544932e-08, "loss": 0.0028, "step": 63821 }, { "epoch": 19.64, "learning_rate": 1.7959770495820096e-08, "loss": 0.0043, "step": 63822 }, { "epoch": 19.64, "learning_rate": 1.7929933522810028e-08, "loss": 0.0027, "step": 63823 }, { "epoch": 19.64, "learning_rate": 1.7900121332588007e-08, "loss": 0.0035, "step": 63824 }, { "epoch": 19.64, "learning_rate": 1.787033392522952e-08, "loss": 0.0041, "step": 63825 }, { "epoch": 19.64, "learning_rate": 1.7840571300807852e-08, "loss": 0.0038, "step": 63826 }, { "epoch": 19.64, "learning_rate": 1.781083345939738e-08, "loss": 0.0036, "step": 63827 }, { "epoch": 19.64, "learning_rate": 1.7781120401071382e-08, "loss": 0.003, "step": 63828 }, { "epoch": 19.64, "learning_rate": 1.7751432125903133e-08, "loss": 0.0035, "step": 63829 }, { "epoch": 19.64, "learning_rate": 1.7721768633967018e-08, "loss": 0.0028, "step": 63830 }, { "epoch": 19.64, "learning_rate": 1.769212992533742e-08, "loss": 0.0024, "step": 63831 }, { "epoch": 19.64, "learning_rate": 1.7662516000086505e-08, "loss": 0.0041, "step": 63832 }, { "epoch": 19.64, "learning_rate": 1.7632926858288658e-08, "loss": 0.0049, "step": 63833 }, { "epoch": 19.64, "learning_rate": 1.7603362500017153e-08, "loss": 0.0021, "step": 63834 }, { "epoch": 19.64, "learning_rate": 1.7573822925345264e-08, "loss": 0.005, "step": 63835 }, { "epoch": 19.64, "learning_rate": 1.7544308134345157e-08, "loss": 0.0037, "step": 63836 }, { "epoch": 19.64, "learning_rate": 1.7514818127093436e-08, "loss": 0.0042, "step": 63837 }, { "epoch": 19.64, "learning_rate": 1.7485352903660048e-08, "loss": 0.0033, "step": 63838 }, { "epoch": 19.64, "learning_rate": 1.7455912464119374e-08, "loss": 0.0041, "step": 63839 }, { "epoch": 19.64, "learning_rate": 1.742649680854469e-08, "loss": 0.0015, "step": 63840 }, { "epoch": 19.64, "learning_rate": 1.7397105937008162e-08, "loss": 0.0036, "step": 63841 }, { "epoch": 19.64, "learning_rate": 1.7367739849584174e-08, "loss": 0.0023, "step": 63842 }, { "epoch": 19.64, "learning_rate": 1.733839854634489e-08, "loss": 0.0034, "step": 63843 }, { "epoch": 19.64, "learning_rate": 1.730908202736248e-08, "loss": 0.0031, "step": 63844 }, { "epoch": 19.64, "learning_rate": 1.727979029271132e-08, "loss": 0.002, "step": 63845 }, { "epoch": 19.64, "learning_rate": 1.7250523342462465e-08, "loss": 0.0028, "step": 63846 }, { "epoch": 19.64, "learning_rate": 1.7221281176689196e-08, "loss": 0.0034, "step": 63847 }, { "epoch": 19.65, "learning_rate": 1.7192063795464787e-08, "loss": 0.0053, "step": 63848 }, { "epoch": 19.65, "learning_rate": 1.7162871198860286e-08, "loss": 0.0031, "step": 63849 }, { "epoch": 19.65, "learning_rate": 1.7133703386950086e-08, "loss": 0.0021, "step": 63850 }, { "epoch": 19.65, "learning_rate": 1.7104560359805235e-08, "loss": 0.0052, "step": 63851 }, { "epoch": 19.65, "learning_rate": 1.7075442117497897e-08, "loss": 0.0017, "step": 63852 }, { "epoch": 19.65, "learning_rate": 1.7046348660101353e-08, "loss": 0.0021, "step": 63853 }, { "epoch": 19.65, "learning_rate": 1.7017279987687763e-08, "loss": 0.0039, "step": 63854 }, { "epoch": 19.65, "learning_rate": 1.698823610032929e-08, "loss": 0.0051, "step": 63855 }, { "epoch": 19.65, "learning_rate": 1.6959216998095883e-08, "loss": 0.0037, "step": 63856 }, { "epoch": 19.65, "learning_rate": 1.6930222681063037e-08, "loss": 0.0037, "step": 63857 }, { "epoch": 19.65, "learning_rate": 1.690125314930069e-08, "loss": 0.0038, "step": 63858 }, { "epoch": 19.65, "learning_rate": 1.687230840288101e-08, "loss": 0.003, "step": 63859 }, { "epoch": 19.65, "learning_rate": 1.6843388441876163e-08, "loss": 0.004, "step": 63860 }, { "epoch": 19.65, "learning_rate": 1.6814493266357202e-08, "loss": 0.0046, "step": 63861 }, { "epoch": 19.65, "learning_rate": 1.67856228763974e-08, "loss": 0.0033, "step": 63862 }, { "epoch": 19.65, "learning_rate": 1.6756777272067814e-08, "loss": 0.0037, "step": 63863 }, { "epoch": 19.65, "learning_rate": 1.6727956453438387e-08, "loss": 0.0037, "step": 63864 }, { "epoch": 19.65, "learning_rate": 1.6699160420583504e-08, "loss": 0.0033, "step": 63865 }, { "epoch": 19.65, "learning_rate": 1.6670389173571998e-08, "loss": 0.0028, "step": 63866 }, { "epoch": 19.65, "learning_rate": 1.6641642712477147e-08, "loss": 0.0026, "step": 63867 }, { "epoch": 19.65, "learning_rate": 1.661292103736889e-08, "loss": 0.0028, "step": 63868 }, { "epoch": 19.65, "learning_rate": 1.6584224148320505e-08, "loss": 0.0033, "step": 63869 }, { "epoch": 19.65, "learning_rate": 1.6555552045400826e-08, "loss": 0.003, "step": 63870 }, { "epoch": 19.65, "learning_rate": 1.6526904728683124e-08, "loss": 0.0028, "step": 63871 }, { "epoch": 19.65, "learning_rate": 1.6498282198237348e-08, "loss": 0.0037, "step": 63872 }, { "epoch": 19.65, "learning_rate": 1.646968445413455e-08, "loss": 0.006, "step": 63873 }, { "epoch": 19.65, "learning_rate": 1.6441111496446894e-08, "loss": 0.0026, "step": 63874 }, { "epoch": 19.65, "learning_rate": 1.6412563325243214e-08, "loss": 0.0032, "step": 63875 }, { "epoch": 19.65, "learning_rate": 1.6384039940596785e-08, "loss": 0.0028, "step": 63876 }, { "epoch": 19.65, "learning_rate": 1.635554134257644e-08, "loss": 0.0037, "step": 63877 }, { "epoch": 19.65, "learning_rate": 1.632706753125435e-08, "loss": 0.0032, "step": 63878 }, { "epoch": 19.65, "learning_rate": 1.629861850670045e-08, "loss": 0.0041, "step": 63879 }, { "epoch": 19.66, "learning_rate": 1.6270194268985796e-08, "loss": 0.003, "step": 63880 }, { "epoch": 19.66, "learning_rate": 1.6241794818180333e-08, "loss": 0.0056, "step": 63881 }, { "epoch": 19.66, "learning_rate": 1.621342015435512e-08, "loss": 0.0022, "step": 63882 }, { "epoch": 19.66, "learning_rate": 1.6185070277580097e-08, "loss": 0.0037, "step": 63883 }, { "epoch": 19.66, "learning_rate": 1.6156745187926315e-08, "loss": 0.0026, "step": 63884 }, { "epoch": 19.66, "learning_rate": 1.6128444885464834e-08, "loss": 0.0033, "step": 63885 }, { "epoch": 19.66, "learning_rate": 1.6100169370263373e-08, "loss": 0.0037, "step": 63886 }, { "epoch": 19.66, "learning_rate": 1.6071918642394102e-08, "loss": 0.004, "step": 63887 }, { "epoch": 19.66, "learning_rate": 1.604369270192696e-08, "loss": 0.0037, "step": 63888 }, { "epoch": 19.66, "learning_rate": 1.6015491548930785e-08, "loss": 0.0012, "step": 63889 }, { "epoch": 19.66, "learning_rate": 1.5987315183477736e-08, "loss": 0.0024, "step": 63890 }, { "epoch": 19.66, "learning_rate": 1.5959163605635542e-08, "loss": 0.0032, "step": 63891 }, { "epoch": 19.66, "learning_rate": 1.5931036815476365e-08, "loss": 0.003, "step": 63892 }, { "epoch": 19.66, "learning_rate": 1.5902934813067928e-08, "loss": 0.0022, "step": 63893 }, { "epoch": 19.66, "learning_rate": 1.5874857598481286e-08, "loss": 0.0039, "step": 63894 }, { "epoch": 19.66, "learning_rate": 1.5846805171785273e-08, "loss": 0.0053, "step": 63895 }, { "epoch": 19.66, "learning_rate": 1.5818777533049834e-08, "loss": 0.0057, "step": 63896 }, { "epoch": 19.66, "learning_rate": 1.579077468234491e-08, "loss": 0.0033, "step": 63897 }, { "epoch": 19.66, "learning_rate": 1.5762796619740452e-08, "loss": 0.0039, "step": 63898 }, { "epoch": 19.66, "learning_rate": 1.5734843345304172e-08, "loss": 0.0024, "step": 63899 }, { "epoch": 19.66, "learning_rate": 1.5706914859107136e-08, "loss": 0.0029, "step": 63900 }, { "epoch": 19.66, "learning_rate": 1.567901116121817e-08, "loss": 0.004, "step": 63901 }, { "epoch": 19.66, "learning_rate": 1.5651132251707224e-08, "loss": 0.0022, "step": 63902 }, { "epoch": 19.66, "learning_rate": 1.5623278130642017e-08, "loss": 0.004, "step": 63903 }, { "epoch": 19.66, "learning_rate": 1.5595448798092494e-08, "loss": 0.0033, "step": 63904 }, { "epoch": 19.66, "learning_rate": 1.55676442541286e-08, "loss": 0.0032, "step": 63905 }, { "epoch": 19.66, "learning_rate": 1.553986449881806e-08, "loss": 0.0035, "step": 63906 }, { "epoch": 19.66, "learning_rate": 1.5512109532229703e-08, "loss": 0.0025, "step": 63907 }, { "epoch": 19.66, "learning_rate": 1.5484379354433476e-08, "loss": 0.0048, "step": 63908 }, { "epoch": 19.66, "learning_rate": 1.5456673965498213e-08, "loss": 0.0049, "step": 63909 }, { "epoch": 19.66, "learning_rate": 1.542899336549164e-08, "loss": 0.0022, "step": 63910 }, { "epoch": 19.66, "learning_rate": 1.540133755448259e-08, "loss": 0.0069, "step": 63911 }, { "epoch": 19.66, "learning_rate": 1.537370653254211e-08, "loss": 0.0023, "step": 63912 }, { "epoch": 19.67, "learning_rate": 1.534610029973571e-08, "loss": 0.0025, "step": 63913 }, { "epoch": 19.67, "learning_rate": 1.5318518856133335e-08, "loss": 0.0029, "step": 63914 }, { "epoch": 19.67, "learning_rate": 1.5290962201802707e-08, "loss": 0.0026, "step": 63915 }, { "epoch": 19.67, "learning_rate": 1.5263430336813768e-08, "loss": 0.0034, "step": 63916 }, { "epoch": 19.67, "learning_rate": 1.5235923261233134e-08, "loss": 0.0053, "step": 63917 }, { "epoch": 19.67, "learning_rate": 1.5208440975130743e-08, "loss": 0.0038, "step": 63918 }, { "epoch": 19.67, "learning_rate": 1.5180983478573218e-08, "loss": 0.0011, "step": 63919 }, { "epoch": 19.67, "learning_rate": 1.5153550771630498e-08, "loss": 0.0035, "step": 63920 }, { "epoch": 19.67, "learning_rate": 1.5126142854368085e-08, "loss": 0.0033, "step": 63921 }, { "epoch": 19.67, "learning_rate": 1.509875972685704e-08, "loss": 0.0049, "step": 63922 }, { "epoch": 19.67, "learning_rate": 1.5071401389162854e-08, "loss": 0.0023, "step": 63923 }, { "epoch": 19.67, "learning_rate": 1.504406784135548e-08, "loss": 0.0047, "step": 63924 }, { "epoch": 19.67, "learning_rate": 1.501675908350153e-08, "loss": 0.0035, "step": 63925 }, { "epoch": 19.67, "learning_rate": 1.4989475115668727e-08, "loss": 0.0033, "step": 63926 }, { "epoch": 19.67, "learning_rate": 1.4962215937925906e-08, "loss": 0.0042, "step": 63927 }, { "epoch": 19.67, "learning_rate": 1.4934981550339678e-08, "loss": 0.0041, "step": 63928 }, { "epoch": 19.67, "learning_rate": 1.4907771952977767e-08, "loss": 0.0024, "step": 63929 }, { "epoch": 19.67, "learning_rate": 1.4880587145907899e-08, "loss": 0.0038, "step": 63930 }, { "epoch": 19.67, "learning_rate": 1.4853427129197794e-08, "loss": 0.003, "step": 63931 }, { "epoch": 19.67, "learning_rate": 1.482629190291518e-08, "loss": 0.0038, "step": 63932 }, { "epoch": 19.67, "learning_rate": 1.4799181467125557e-08, "loss": 0.0035, "step": 63933 }, { "epoch": 19.67, "learning_rate": 1.4772095821898868e-08, "loss": 0.0032, "step": 63934 }, { "epoch": 19.67, "learning_rate": 1.474503496730173e-08, "loss": 0.0028, "step": 63935 }, { "epoch": 19.67, "learning_rate": 1.4717998903399644e-08, "loss": 0.0032, "step": 63936 }, { "epoch": 19.67, "learning_rate": 1.4690987630260333e-08, "loss": 0.0025, "step": 63937 }, { "epoch": 19.67, "learning_rate": 1.4664001147952633e-08, "loss": 0.0049, "step": 63938 }, { "epoch": 19.67, "learning_rate": 1.4637039456542047e-08, "loss": 0.0034, "step": 63939 }, { "epoch": 19.67, "learning_rate": 1.4610102556095185e-08, "loss": 0.003, "step": 63940 }, { "epoch": 19.67, "learning_rate": 1.4583190446679774e-08, "loss": 0.002, "step": 63941 }, { "epoch": 19.67, "learning_rate": 1.4556303128361315e-08, "loss": 0.0027, "step": 63942 }, { "epoch": 19.67, "learning_rate": 1.4529440601208644e-08, "loss": 0.0045, "step": 63943 }, { "epoch": 19.67, "learning_rate": 1.4502602865287263e-08, "loss": 0.004, "step": 63944 }, { "epoch": 19.68, "learning_rate": 1.4475789920663785e-08, "loss": 0.0028, "step": 63945 }, { "epoch": 19.68, "learning_rate": 1.4449001767404824e-08, "loss": 0.0036, "step": 63946 }, { "epoch": 19.68, "learning_rate": 1.4422238405576994e-08, "loss": 0.0043, "step": 63947 }, { "epoch": 19.68, "learning_rate": 1.4395499835246907e-08, "loss": 0.0031, "step": 63948 }, { "epoch": 19.68, "learning_rate": 1.4368786056481177e-08, "loss": 0.0025, "step": 63949 }, { "epoch": 19.68, "learning_rate": 1.4342097069345307e-08, "loss": 0.0047, "step": 63950 }, { "epoch": 19.68, "learning_rate": 1.43154328739048e-08, "loss": 0.0028, "step": 63951 }, { "epoch": 19.68, "learning_rate": 1.4288793470228491e-08, "loss": 0.0046, "step": 63952 }, { "epoch": 19.68, "learning_rate": 1.4262178858380771e-08, "loss": 0.0026, "step": 63953 }, { "epoch": 19.68, "learning_rate": 1.4235589038427144e-08, "loss": 0.0026, "step": 63954 }, { "epoch": 19.68, "learning_rate": 1.4209024010435334e-08, "loss": 0.0027, "step": 63955 }, { "epoch": 19.68, "learning_rate": 1.4182483774470845e-08, "loss": 0.0032, "step": 63956 }, { "epoch": 19.68, "learning_rate": 1.4155968330598069e-08, "loss": 0.0029, "step": 63957 }, { "epoch": 19.68, "learning_rate": 1.4129477678884728e-08, "loss": 0.0068, "step": 63958 }, { "epoch": 19.68, "learning_rate": 1.4103011819395218e-08, "loss": 0.0041, "step": 63959 }, { "epoch": 19.68, "learning_rate": 1.407657075219615e-08, "loss": 0.0039, "step": 63960 }, { "epoch": 19.68, "learning_rate": 1.4050154477353029e-08, "loss": 0.0037, "step": 63961 }, { "epoch": 19.68, "learning_rate": 1.4023762994930246e-08, "loss": 0.0032, "step": 63962 }, { "epoch": 19.68, "learning_rate": 1.3997396304995526e-08, "loss": 0.0033, "step": 63963 }, { "epoch": 19.68, "learning_rate": 1.397105440761215e-08, "loss": 0.0063, "step": 63964 }, { "epoch": 19.68, "learning_rate": 1.3944737302846733e-08, "loss": 0.003, "step": 63965 }, { "epoch": 19.68, "learning_rate": 1.3918444990764779e-08, "loss": 0.0028, "step": 63966 }, { "epoch": 19.68, "learning_rate": 1.3892177471430679e-08, "loss": 0.0032, "step": 63967 }, { "epoch": 19.68, "learning_rate": 1.3865934744911047e-08, "loss": 0.0042, "step": 63968 }, { "epoch": 19.68, "learning_rate": 1.3839716811269165e-08, "loss": 0.0026, "step": 63969 }, { "epoch": 19.68, "learning_rate": 1.381352367057165e-08, "loss": 0.0037, "step": 63970 }, { "epoch": 19.68, "learning_rate": 1.378735532288289e-08, "loss": 0.0054, "step": 63971 }, { "epoch": 19.68, "learning_rate": 1.376121176826728e-08, "loss": 0.002, "step": 63972 }, { "epoch": 19.68, "learning_rate": 1.3735093006791434e-08, "loss": 0.003, "step": 63973 }, { "epoch": 19.68, "learning_rate": 1.3708999038518633e-08, "loss": 0.003, "step": 63974 }, { "epoch": 19.68, "learning_rate": 1.3682929863514383e-08, "loss": 0.0047, "step": 63975 }, { "epoch": 19.68, "learning_rate": 1.3656885481843074e-08, "loss": 0.0027, "step": 63976 }, { "epoch": 19.68, "learning_rate": 1.3630865893569101e-08, "loss": 0.0037, "step": 63977 }, { "epoch": 19.69, "learning_rate": 1.3604871098757965e-08, "loss": 0.0026, "step": 63978 }, { "epoch": 19.69, "learning_rate": 1.3578901097472952e-08, "loss": 0.0026, "step": 63979 }, { "epoch": 19.69, "learning_rate": 1.3552955889779561e-08, "loss": 0.0033, "step": 63980 }, { "epoch": 19.69, "learning_rate": 1.352703547574219e-08, "loss": 0.0032, "step": 63981 }, { "epoch": 19.69, "learning_rate": 1.3501139855425227e-08, "loss": 0.0029, "step": 63982 }, { "epoch": 19.69, "learning_rate": 1.3475269028891958e-08, "loss": 0.0035, "step": 63983 }, { "epoch": 19.69, "learning_rate": 1.3449422996207884e-08, "loss": 0.0038, "step": 63984 }, { "epoch": 19.69, "learning_rate": 1.3423601757436289e-08, "loss": 0.0017, "step": 63985 }, { "epoch": 19.69, "learning_rate": 1.3397805312641566e-08, "loss": 0.0029, "step": 63986 }, { "epoch": 19.69, "learning_rate": 1.3372033661886996e-08, "loss": 0.0018, "step": 63987 }, { "epoch": 19.69, "learning_rate": 1.3346286805238084e-08, "loss": 0.002, "step": 63988 }, { "epoch": 19.69, "learning_rate": 1.3320564742759223e-08, "loss": 0.002, "step": 63989 }, { "epoch": 19.69, "learning_rate": 1.3294867474511475e-08, "loss": 0.0036, "step": 63990 }, { "epoch": 19.69, "learning_rate": 1.3269195000560343e-08, "loss": 0.0026, "step": 63991 }, { "epoch": 19.69, "learning_rate": 1.3243547320970219e-08, "loss": 0.0026, "step": 63992 }, { "epoch": 19.69, "learning_rate": 1.3217924435803276e-08, "loss": 0.0025, "step": 63993 }, { "epoch": 19.69, "learning_rate": 1.3192326345125017e-08, "loss": 0.0031, "step": 63994 }, { "epoch": 19.69, "learning_rate": 1.3166753048996505e-08, "loss": 0.0042, "step": 63995 }, { "epoch": 19.69, "learning_rate": 1.3141204547483243e-08, "loss": 0.0034, "step": 63996 }, { "epoch": 19.69, "learning_rate": 1.3115680840648515e-08, "loss": 0.0022, "step": 63997 }, { "epoch": 19.69, "learning_rate": 1.309018192855449e-08, "loss": 0.0028, "step": 63998 }, { "epoch": 19.69, "learning_rate": 1.3064707811265564e-08, "loss": 0.0029, "step": 63999 }, { "epoch": 19.69, "learning_rate": 1.303925848884391e-08, "loss": 0.0039, "step": 64000 }, { "epoch": 19.69, "learning_rate": 1.3013833961352807e-08, "loss": 0.0037, "step": 64001 }, { "epoch": 19.69, "learning_rate": 1.298843422885776e-08, "loss": 0.0014, "step": 64002 }, { "epoch": 19.69, "learning_rate": 1.2963059291418723e-08, "loss": 0.0029, "step": 64003 }, { "epoch": 19.69, "learning_rate": 1.2937709149101196e-08, "loss": 0.006, "step": 64004 }, { "epoch": 19.69, "learning_rate": 1.2912383801965133e-08, "loss": 0.0019, "step": 64005 }, { "epoch": 19.69, "learning_rate": 1.2887083250076039e-08, "loss": 0.0057, "step": 64006 }, { "epoch": 19.69, "learning_rate": 1.2861807493496082e-08, "loss": 0.0036, "step": 64007 }, { "epoch": 19.69, "learning_rate": 1.2836556532287437e-08, "loss": 0.0019, "step": 64008 }, { "epoch": 19.69, "learning_rate": 1.2811330366514495e-08, "loss": 0.0047, "step": 64009 }, { "epoch": 19.7, "learning_rate": 1.2786128996237213e-08, "loss": 0.0018, "step": 64010 }, { "epoch": 19.7, "learning_rate": 1.276095242151998e-08, "loss": 0.0064, "step": 64011 }, { "epoch": 19.7, "learning_rate": 1.2735800642423857e-08, "loss": 0.0044, "step": 64012 }, { "epoch": 19.7, "learning_rate": 1.271067365901324e-08, "loss": 0.0037, "step": 64013 }, { "epoch": 19.7, "learning_rate": 1.268557147134919e-08, "loss": 0.0022, "step": 64014 }, { "epoch": 19.7, "learning_rate": 1.266049407949499e-08, "loss": 0.0032, "step": 64015 }, { "epoch": 19.7, "learning_rate": 1.2635441483511702e-08, "loss": 0.0019, "step": 64016 }, { "epoch": 19.7, "learning_rate": 1.2610413683462608e-08, "loss": 0.0044, "step": 64017 }, { "epoch": 19.7, "learning_rate": 1.2585410679408771e-08, "loss": 0.0035, "step": 64018 }, { "epoch": 19.7, "learning_rate": 1.2560432471412365e-08, "loss": 0.0051, "step": 64019 }, { "epoch": 19.7, "learning_rate": 1.253547905953667e-08, "loss": 0.0026, "step": 64020 }, { "epoch": 19.7, "learning_rate": 1.251055044384275e-08, "loss": 0.0032, "step": 64021 }, { "epoch": 19.7, "learning_rate": 1.2485646624392777e-08, "loss": 0.0023, "step": 64022 }, { "epoch": 19.7, "learning_rate": 1.2460767601247814e-08, "loss": 0.0054, "step": 64023 }, { "epoch": 19.7, "learning_rate": 1.2435913374471142e-08, "loss": 0.0045, "step": 64024 }, { "epoch": 19.7, "learning_rate": 1.2411083944122715e-08, "loss": 0.0043, "step": 64025 }, { "epoch": 19.7, "learning_rate": 1.2386279310264703e-08, "loss": 0.0028, "step": 64026 }, { "epoch": 19.7, "learning_rate": 1.2361499472959282e-08, "loss": 0.0026, "step": 64027 }, { "epoch": 19.7, "learning_rate": 1.2336744432268621e-08, "loss": 0.003, "step": 64028 }, { "epoch": 19.7, "learning_rate": 1.2312014188251565e-08, "loss": 0.0025, "step": 64029 }, { "epoch": 19.7, "learning_rate": 1.2287308740972503e-08, "loss": 0.0013, "step": 64030 }, { "epoch": 19.7, "learning_rate": 1.2262628090491391e-08, "loss": 0.0027, "step": 64031 }, { "epoch": 19.7, "learning_rate": 1.2237972236869289e-08, "loss": 0.0034, "step": 64032 }, { "epoch": 19.7, "learning_rate": 1.2213341180167259e-08, "loss": 0.0023, "step": 64033 }, { "epoch": 19.7, "learning_rate": 1.2188734920447475e-08, "loss": 0.0026, "step": 64034 }, { "epoch": 19.7, "learning_rate": 1.216415345776989e-08, "loss": 0.0025, "step": 64035 }, { "epoch": 19.7, "learning_rate": 1.2139596792196673e-08, "loss": 0.0034, "step": 64036 }, { "epoch": 19.7, "learning_rate": 1.2115064923787778e-08, "loss": 0.0032, "step": 64037 }, { "epoch": 19.7, "learning_rate": 1.2090557852604268e-08, "loss": 0.0032, "step": 64038 }, { "epoch": 19.7, "learning_rate": 1.2066075578708314e-08, "loss": 0.0024, "step": 64039 }, { "epoch": 19.7, "learning_rate": 1.2041618102158758e-08, "loss": 0.0026, "step": 64040 }, { "epoch": 19.7, "learning_rate": 1.2017185423016663e-08, "loss": 0.0038, "step": 64041 }, { "epoch": 19.7, "learning_rate": 1.1992777541344203e-08, "loss": 0.0035, "step": 64042 }, { "epoch": 19.71, "learning_rate": 1.1968394457201326e-08, "loss": 0.0033, "step": 64043 }, { "epoch": 19.71, "learning_rate": 1.1944036170646878e-08, "loss": 0.0045, "step": 64044 }, { "epoch": 19.71, "learning_rate": 1.1919702681744138e-08, "loss": 0.0021, "step": 64045 }, { "epoch": 19.71, "learning_rate": 1.1895393990550841e-08, "loss": 0.0043, "step": 64046 }, { "epoch": 19.71, "learning_rate": 1.1871110097129157e-08, "loss": 0.0034, "step": 64047 }, { "epoch": 19.71, "learning_rate": 1.184685100153904e-08, "loss": 0.0029, "step": 64048 }, { "epoch": 19.71, "learning_rate": 1.1822616703840439e-08, "loss": 0.0037, "step": 64049 }, { "epoch": 19.71, "learning_rate": 1.179840720409331e-08, "loss": 0.0043, "step": 64050 }, { "epoch": 19.71, "learning_rate": 1.1774222502357602e-08, "loss": 0.0021, "step": 64051 }, { "epoch": 19.71, "learning_rate": 1.1750062598694378e-08, "loss": 0.0056, "step": 64052 }, { "epoch": 19.71, "learning_rate": 1.172592749316248e-08, "loss": 0.0043, "step": 64053 }, { "epoch": 19.71, "learning_rate": 1.1701817185822972e-08, "loss": 0.0028, "step": 64054 }, { "epoch": 19.71, "learning_rate": 1.1677731676733584e-08, "loss": 0.0019, "step": 64055 }, { "epoch": 19.71, "learning_rate": 1.1653670965957598e-08, "loss": 0.0015, "step": 64056 }, { "epoch": 19.71, "learning_rate": 1.1629635053551635e-08, "loss": 0.0031, "step": 64057 }, { "epoch": 19.71, "learning_rate": 1.160562393957676e-08, "loss": 0.0038, "step": 64058 }, { "epoch": 19.71, "learning_rate": 1.1581637624092923e-08, "loss": 0.0037, "step": 64059 }, { "epoch": 19.71, "learning_rate": 1.1557676107157856e-08, "loss": 0.0037, "step": 64060 }, { "epoch": 19.71, "learning_rate": 1.153373938883262e-08, "loss": 0.0051, "step": 64061 }, { "epoch": 19.71, "learning_rate": 1.150982746917717e-08, "loss": 0.0046, "step": 64062 }, { "epoch": 19.71, "learning_rate": 1.1485940348249235e-08, "loss": 0.0034, "step": 64063 }, { "epoch": 19.71, "learning_rate": 1.146207802610988e-08, "loss": 0.0022, "step": 64064 }, { "epoch": 19.71, "learning_rate": 1.1438240502815723e-08, "loss": 0.0033, "step": 64065 }, { "epoch": 19.71, "learning_rate": 1.1414427778428937e-08, "loss": 0.0049, "step": 64066 }, { "epoch": 19.71, "learning_rate": 1.1390639853007258e-08, "loss": 0.0038, "step": 64067 }, { "epoch": 19.71, "learning_rate": 1.136687672660841e-08, "loss": 0.0027, "step": 64068 }, { "epoch": 19.71, "learning_rate": 1.1343138399294573e-08, "loss": 0.0022, "step": 64069 }, { "epoch": 19.71, "learning_rate": 1.1319424871121254e-08, "loss": 0.0036, "step": 64070 }, { "epoch": 19.71, "learning_rate": 1.1295736142149515e-08, "loss": 0.002, "step": 64071 }, { "epoch": 19.71, "learning_rate": 1.12720722124382e-08, "loss": 0.0024, "step": 64072 }, { "epoch": 19.71, "learning_rate": 1.124843308204504e-08, "loss": 0.003, "step": 64073 }, { "epoch": 19.71, "learning_rate": 1.1224818751028876e-08, "loss": 0.0032, "step": 64074 }, { "epoch": 19.72, "learning_rate": 1.120122921944855e-08, "loss": 0.0042, "step": 64075 }, { "epoch": 19.72, "learning_rate": 1.1177664487362905e-08, "loss": 0.0039, "step": 64076 }, { "epoch": 19.72, "learning_rate": 1.115412455483078e-08, "loss": 0.003, "step": 64077 }, { "epoch": 19.72, "learning_rate": 1.113060942190991e-08, "loss": 0.0036, "step": 64078 }, { "epoch": 19.72, "learning_rate": 1.1107119088658025e-08, "loss": 0.003, "step": 64079 }, { "epoch": 19.72, "learning_rate": 1.1083653555133966e-08, "loss": 0.002, "step": 64080 }, { "epoch": 19.72, "learning_rate": 1.1060212821397686e-08, "loss": 0.0021, "step": 64081 }, { "epoch": 19.72, "learning_rate": 1.1036796887504697e-08, "loss": 0.0045, "step": 64082 }, { "epoch": 19.72, "learning_rate": 1.101340575351495e-08, "loss": 0.0032, "step": 64083 }, { "epoch": 19.72, "learning_rate": 1.0990039419486175e-08, "loss": 0.0033, "step": 64084 }, { "epoch": 19.72, "learning_rate": 1.0966697885476108e-08, "loss": 0.0022, "step": 64085 }, { "epoch": 19.72, "learning_rate": 1.0943381151542476e-08, "loss": 0.0022, "step": 64086 }, { "epoch": 19.72, "learning_rate": 1.0920089217744122e-08, "loss": 0.004, "step": 64087 }, { "epoch": 19.72, "learning_rate": 1.089682208413767e-08, "loss": 0.0054, "step": 64088 }, { "epoch": 19.72, "learning_rate": 1.087357975078085e-08, "loss": 0.0022, "step": 64089 }, { "epoch": 19.72, "learning_rate": 1.0850362217732502e-08, "loss": 0.0043, "step": 64090 }, { "epoch": 19.72, "learning_rate": 1.082716948505036e-08, "loss": 0.0026, "step": 64091 }, { "epoch": 19.72, "learning_rate": 1.0804001552789933e-08, "loss": 0.0035, "step": 64092 }, { "epoch": 19.72, "learning_rate": 1.0780858421011175e-08, "loss": 0.0026, "step": 64093 }, { "epoch": 19.72, "learning_rate": 1.0757740089770707e-08, "loss": 0.0036, "step": 64094 }, { "epoch": 19.72, "learning_rate": 1.0734646559125151e-08, "loss": 0.0052, "step": 64095 }, { "epoch": 19.72, "learning_rate": 1.0711577829132235e-08, "loss": 0.005, "step": 64096 }, { "epoch": 19.72, "learning_rate": 1.0688533899850806e-08, "loss": 0.0043, "step": 64097 }, { "epoch": 19.72, "learning_rate": 1.0665514771335262e-08, "loss": 0.0028, "step": 64098 }, { "epoch": 19.72, "learning_rate": 1.0642520443645554e-08, "loss": 0.0027, "step": 64099 }, { "epoch": 19.72, "learning_rate": 1.0619550916836086e-08, "loss": 0.0032, "step": 64100 }, { "epoch": 19.72, "learning_rate": 1.0596606190965696e-08, "loss": 0.0048, "step": 64101 }, { "epoch": 19.72, "learning_rate": 1.057368626609101e-08, "loss": 0.0031, "step": 64102 }, { "epoch": 19.72, "learning_rate": 1.0550791142268646e-08, "loss": 0.0032, "step": 64103 }, { "epoch": 19.72, "learning_rate": 1.0527920819555226e-08, "loss": 0.0035, "step": 64104 }, { "epoch": 19.72, "learning_rate": 1.0505075298008483e-08, "loss": 0.0026, "step": 64105 }, { "epoch": 19.72, "learning_rate": 1.0482254577685035e-08, "loss": 0.0033, "step": 64106 }, { "epoch": 19.72, "learning_rate": 1.0459458658640397e-08, "loss": 0.0043, "step": 64107 }, { "epoch": 19.73, "learning_rate": 1.043668754093119e-08, "loss": 0.0044, "step": 64108 }, { "epoch": 19.73, "learning_rate": 1.0413941224615142e-08, "loss": 0.0025, "step": 64109 }, { "epoch": 19.73, "learning_rate": 1.0391219709748878e-08, "loss": 0.0029, "step": 64110 }, { "epoch": 19.73, "learning_rate": 1.03685229963868e-08, "loss": 0.0041, "step": 64111 }, { "epoch": 19.73, "learning_rate": 1.0345851084587743e-08, "loss": 0.0037, "step": 64112 }, { "epoch": 19.73, "learning_rate": 1.0323203974406115e-08, "loss": 0.0024, "step": 64113 }, { "epoch": 19.73, "learning_rate": 1.0300581665899644e-08, "loss": 0.0022, "step": 64114 }, { "epoch": 19.73, "learning_rate": 1.0277984159122734e-08, "loss": 0.0028, "step": 64115 }, { "epoch": 19.73, "learning_rate": 1.0255411454132003e-08, "loss": 0.0046, "step": 64116 }, { "epoch": 19.73, "learning_rate": 1.0232863550985184e-08, "loss": 0.0058, "step": 64117 }, { "epoch": 19.73, "learning_rate": 1.0210340449736677e-08, "loss": 0.0021, "step": 64118 }, { "epoch": 19.73, "learning_rate": 1.0187842150443106e-08, "loss": 0.0028, "step": 64119 }, { "epoch": 19.73, "learning_rate": 1.0165368653158868e-08, "loss": 0.003, "step": 64120 }, { "epoch": 19.73, "learning_rate": 1.0142919957941699e-08, "loss": 0.0042, "step": 64121 }, { "epoch": 19.73, "learning_rate": 1.0120496064845997e-08, "loss": 0.0029, "step": 64122 }, { "epoch": 19.73, "learning_rate": 1.0098096973928384e-08, "loss": 0.0024, "step": 64123 }, { "epoch": 19.73, "learning_rate": 1.0075722685243261e-08, "loss": 0.0033, "step": 64124 }, { "epoch": 19.73, "learning_rate": 1.005337319884725e-08, "loss": 0.0035, "step": 64125 }, { "epoch": 19.73, "learning_rate": 1.003104851479586e-08, "loss": 0.0035, "step": 64126 }, { "epoch": 19.73, "learning_rate": 1.0008748633144604e-08, "loss": 0.0019, "step": 64127 }, { "epoch": 19.73, "learning_rate": 9.986473553946774e-09, "loss": 0.0041, "step": 64128 }, { "epoch": 19.73, "learning_rate": 9.96422327726121e-09, "loss": 0.0042, "step": 64129 }, { "epoch": 19.73, "learning_rate": 9.941997803140092e-09, "loss": 0.0036, "step": 64130 }, { "epoch": 19.73, "learning_rate": 9.919797131640041e-09, "loss": 0.0027, "step": 64131 }, { "epoch": 19.73, "learning_rate": 9.897621262815459e-09, "loss": 0.0033, "step": 64132 }, { "epoch": 19.73, "learning_rate": 9.875470196721859e-09, "loss": 0.0044, "step": 64133 }, { "epoch": 19.73, "learning_rate": 9.85334393341475e-09, "loss": 0.0032, "step": 64134 }, { "epoch": 19.73, "learning_rate": 9.831242472948532e-09, "loss": 0.0025, "step": 64135 }, { "epoch": 19.73, "learning_rate": 9.809165815377609e-09, "loss": 0.0039, "step": 64136 }, { "epoch": 19.73, "learning_rate": 9.787113960757488e-09, "loss": 0.0032, "step": 64137 }, { "epoch": 19.73, "learning_rate": 9.765086909143684e-09, "loss": 0.0033, "step": 64138 }, { "epoch": 19.73, "learning_rate": 9.743084660588376e-09, "loss": 0.005, "step": 64139 }, { "epoch": 19.74, "learning_rate": 9.721107215148184e-09, "loss": 0.0053, "step": 64140 }, { "epoch": 19.74, "learning_rate": 9.699154572877511e-09, "loss": 0.0017, "step": 64141 }, { "epoch": 19.74, "learning_rate": 9.677226733829648e-09, "loss": 0.0038, "step": 64142 }, { "epoch": 19.74, "learning_rate": 9.655323698061214e-09, "loss": 0.0027, "step": 64143 }, { "epoch": 19.74, "learning_rate": 9.633445465624391e-09, "loss": 0.0043, "step": 64144 }, { "epoch": 19.74, "learning_rate": 9.61159203657469e-09, "loss": 0.005, "step": 64145 }, { "epoch": 19.74, "learning_rate": 9.589763410966513e-09, "loss": 0.0025, "step": 64146 }, { "epoch": 19.74, "learning_rate": 9.567959588853148e-09, "loss": 0.0037, "step": 64147 }, { "epoch": 19.74, "learning_rate": 9.546180570288999e-09, "loss": 0.0032, "step": 64148 }, { "epoch": 19.74, "learning_rate": 9.524426355328465e-09, "loss": 0.0038, "step": 64149 }, { "epoch": 19.74, "learning_rate": 9.502696944024837e-09, "loss": 0.0025, "step": 64150 }, { "epoch": 19.74, "learning_rate": 9.480992336433625e-09, "loss": 0.0028, "step": 64151 }, { "epoch": 19.74, "learning_rate": 9.459312532608122e-09, "loss": 0.0028, "step": 64152 }, { "epoch": 19.74, "learning_rate": 9.437657532600508e-09, "loss": 0.0045, "step": 64153 }, { "epoch": 19.74, "learning_rate": 9.416027336466294e-09, "loss": 0.0046, "step": 64154 }, { "epoch": 19.74, "learning_rate": 9.394421944258769e-09, "loss": 0.0026, "step": 64155 }, { "epoch": 19.74, "learning_rate": 9.372841356032336e-09, "loss": 0.0019, "step": 64156 }, { "epoch": 19.74, "learning_rate": 9.351285571839174e-09, "loss": 0.0044, "step": 64157 }, { "epoch": 19.74, "learning_rate": 9.329754591732576e-09, "loss": 0.0086, "step": 64158 }, { "epoch": 19.74, "learning_rate": 9.30824841576805e-09, "loss": 0.0035, "step": 64159 }, { "epoch": 19.74, "learning_rate": 9.286767043996669e-09, "loss": 0.0024, "step": 64160 }, { "epoch": 19.74, "learning_rate": 9.265310476472833e-09, "loss": 0.0031, "step": 64161 }, { "epoch": 19.74, "learning_rate": 9.243878713250943e-09, "loss": 0.0046, "step": 64162 }, { "epoch": 19.74, "learning_rate": 9.222471754382068e-09, "loss": 0.0038, "step": 64163 }, { "epoch": 19.74, "learning_rate": 9.20108959992061e-09, "loss": 0.0034, "step": 64164 }, { "epoch": 19.74, "learning_rate": 9.17973224991986e-09, "loss": 0.0036, "step": 64165 }, { "epoch": 19.74, "learning_rate": 9.158399704431997e-09, "loss": 0.0039, "step": 64166 }, { "epoch": 19.74, "learning_rate": 9.137091963510314e-09, "loss": 0.0039, "step": 64167 }, { "epoch": 19.74, "learning_rate": 9.1158090272081e-09, "loss": 0.0055, "step": 64168 }, { "epoch": 19.74, "learning_rate": 9.094550895577536e-09, "loss": 0.0016, "step": 64169 }, { "epoch": 19.74, "learning_rate": 9.073317568670804e-09, "loss": 0.0024, "step": 64170 }, { "epoch": 19.74, "learning_rate": 9.052109046543412e-09, "loss": 0.0035, "step": 64171 }, { "epoch": 19.74, "learning_rate": 9.03092532924532e-09, "loss": 0.0042, "step": 64172 }, { "epoch": 19.75, "learning_rate": 9.009766416829824e-09, "loss": 0.0022, "step": 64173 }, { "epoch": 19.75, "learning_rate": 8.988632309349099e-09, "loss": 0.0022, "step": 64174 }, { "epoch": 19.75, "learning_rate": 8.967523006857549e-09, "loss": 0.0041, "step": 64175 }, { "epoch": 19.75, "learning_rate": 8.946438509405132e-09, "loss": 0.0035, "step": 64176 }, { "epoch": 19.75, "learning_rate": 8.925378817045139e-09, "loss": 0.003, "step": 64177 }, { "epoch": 19.75, "learning_rate": 8.904343929830861e-09, "loss": 0.0032, "step": 64178 }, { "epoch": 19.75, "learning_rate": 8.883333847812258e-09, "loss": 0.0047, "step": 64179 }, { "epoch": 19.75, "learning_rate": 8.862348571043733e-09, "loss": 0.0022, "step": 64180 }, { "epoch": 19.75, "learning_rate": 8.841388099576354e-09, "loss": 0.0032, "step": 64181 }, { "epoch": 19.75, "learning_rate": 8.8204524334623e-09, "loss": 0.0049, "step": 64182 }, { "epoch": 19.75, "learning_rate": 8.799541572752646e-09, "loss": 0.0026, "step": 64183 }, { "epoch": 19.75, "learning_rate": 8.778655517500678e-09, "loss": 0.0022, "step": 64184 }, { "epoch": 19.75, "learning_rate": 8.75779426775858e-09, "loss": 0.0035, "step": 64185 }, { "epoch": 19.75, "learning_rate": 8.736957823576309e-09, "loss": 0.0055, "step": 64186 }, { "epoch": 19.75, "learning_rate": 8.716146185007157e-09, "loss": 0.0018, "step": 64187 }, { "epoch": 19.75, "learning_rate": 8.695359352102196e-09, "loss": 0.0023, "step": 64188 }, { "epoch": 19.75, "learning_rate": 8.674597324912492e-09, "loss": 0.0018, "step": 64189 }, { "epoch": 19.75, "learning_rate": 8.65386010349134e-09, "loss": 0.0039, "step": 64190 }, { "epoch": 19.75, "learning_rate": 8.633147687887589e-09, "loss": 0.0053, "step": 64191 }, { "epoch": 19.75, "learning_rate": 8.612460078155638e-09, "loss": 0.0021, "step": 64192 }, { "epoch": 19.75, "learning_rate": 8.591797274344338e-09, "loss": 0.0035, "step": 64193 }, { "epoch": 19.75, "learning_rate": 8.57115927650698e-09, "loss": 0.0033, "step": 64194 }, { "epoch": 19.75, "learning_rate": 8.550546084692413e-09, "loss": 0.002, "step": 64195 }, { "epoch": 19.75, "learning_rate": 8.529957698953928e-09, "loss": 0.0032, "step": 64196 }, { "epoch": 19.75, "learning_rate": 8.509394119342595e-09, "loss": 0.0037, "step": 64197 }, { "epoch": 19.75, "learning_rate": 8.488855345908376e-09, "loss": 0.0044, "step": 64198 }, { "epoch": 19.75, "learning_rate": 8.468341378702339e-09, "loss": 0.0025, "step": 64199 }, { "epoch": 19.75, "learning_rate": 8.447852217775554e-09, "loss": 0.0034, "step": 64200 }, { "epoch": 19.75, "learning_rate": 8.427387863180204e-09, "loss": 0.0051, "step": 64201 }, { "epoch": 19.75, "learning_rate": 8.406948314965135e-09, "loss": 0.005, "step": 64202 }, { "epoch": 19.75, "learning_rate": 8.386533573182531e-09, "loss": 0.0026, "step": 64203 }, { "epoch": 19.75, "learning_rate": 8.36614363788124e-09, "loss": 0.0021, "step": 64204 }, { "epoch": 19.76, "learning_rate": 8.345778509114556e-09, "loss": 0.003, "step": 64205 }, { "epoch": 19.76, "learning_rate": 8.325438186931323e-09, "loss": 0.0037, "step": 64206 }, { "epoch": 19.76, "learning_rate": 8.305122671381505e-09, "loss": 0.0026, "step": 64207 }, { "epoch": 19.76, "learning_rate": 8.284831962517282e-09, "loss": 0.0048, "step": 64208 }, { "epoch": 19.76, "learning_rate": 8.264566060388612e-09, "loss": 0.0034, "step": 64209 }, { "epoch": 19.76, "learning_rate": 8.244324965044348e-09, "loss": 0.0033, "step": 64210 }, { "epoch": 19.76, "learning_rate": 8.224108676535557e-09, "loss": 0.003, "step": 64211 }, { "epoch": 19.76, "learning_rate": 8.203917194913313e-09, "loss": 0.0032, "step": 64212 }, { "epoch": 19.76, "learning_rate": 8.183750520226464e-09, "loss": 0.0019, "step": 64213 }, { "epoch": 19.76, "learning_rate": 8.163608652526079e-09, "loss": 0.0039, "step": 64214 }, { "epoch": 19.76, "learning_rate": 8.143491591862118e-09, "loss": 0.0035, "step": 64215 }, { "epoch": 19.76, "learning_rate": 8.123399338283433e-09, "loss": 0.0046, "step": 64216 }, { "epoch": 19.76, "learning_rate": 8.103331891841094e-09, "loss": 0.0034, "step": 64217 }, { "epoch": 19.76, "learning_rate": 8.083289252583948e-09, "loss": 0.0043, "step": 64218 }, { "epoch": 19.76, "learning_rate": 8.063271420563068e-09, "loss": 0.0037, "step": 64219 }, { "epoch": 19.76, "learning_rate": 8.043278395826193e-09, "loss": 0.0025, "step": 64220 }, { "epoch": 19.76, "learning_rate": 8.023310178424393e-09, "loss": 0.0034, "step": 64221 }, { "epoch": 19.76, "learning_rate": 8.003366768407628e-09, "loss": 0.0044, "step": 64222 }, { "epoch": 19.76, "learning_rate": 7.983448165824747e-09, "loss": 0.0022, "step": 64223 }, { "epoch": 19.76, "learning_rate": 7.963554370724602e-09, "loss": 0.0039, "step": 64224 }, { "epoch": 19.76, "learning_rate": 7.94368538315715e-09, "loss": 0.0036, "step": 64225 }, { "epoch": 19.76, "learning_rate": 7.923841203172356e-09, "loss": 0.0038, "step": 64226 }, { "epoch": 19.76, "learning_rate": 7.904021830819063e-09, "loss": 0.0024, "step": 64227 }, { "epoch": 19.76, "learning_rate": 7.884227266145017e-09, "loss": 0.0038, "step": 64228 }, { "epoch": 19.76, "learning_rate": 7.864457509202395e-09, "loss": 0.0031, "step": 64229 }, { "epoch": 19.76, "learning_rate": 7.844712560036715e-09, "loss": 0.0017, "step": 64230 }, { "epoch": 19.76, "learning_rate": 7.824992418700161e-09, "loss": 0.0046, "step": 64231 }, { "epoch": 19.76, "learning_rate": 7.80529708523936e-09, "loss": 0.004, "step": 64232 }, { "epoch": 19.76, "learning_rate": 7.785626559705383e-09, "loss": 0.0036, "step": 64233 }, { "epoch": 19.76, "learning_rate": 7.765980842144861e-09, "loss": 0.003, "step": 64234 }, { "epoch": 19.76, "learning_rate": 7.74635993260775e-09, "loss": 0.0032, "step": 64235 }, { "epoch": 19.76, "learning_rate": 7.726763831142903e-09, "loss": 0.0035, "step": 64236 }, { "epoch": 19.76, "learning_rate": 7.707192537798059e-09, "loss": 0.0056, "step": 64237 }, { "epoch": 19.77, "learning_rate": 7.687646052623176e-09, "loss": 0.0026, "step": 64238 }, { "epoch": 19.77, "learning_rate": 7.668124375665997e-09, "loss": 0.0029, "step": 64239 }, { "epoch": 19.77, "learning_rate": 7.64862750697426e-09, "loss": 0.0029, "step": 64240 }, { "epoch": 19.77, "learning_rate": 7.629155446597924e-09, "loss": 0.0039, "step": 64241 }, { "epoch": 19.77, "learning_rate": 7.609708194583621e-09, "loss": 0.0028, "step": 64242 }, { "epoch": 19.77, "learning_rate": 7.590285750981307e-09, "loss": 0.0028, "step": 64243 }, { "epoch": 19.77, "learning_rate": 7.570888115838726e-09, "loss": 0.004, "step": 64244 }, { "epoch": 19.77, "learning_rate": 7.551515289203615e-09, "loss": 0.0026, "step": 64245 }, { "epoch": 19.77, "learning_rate": 7.532167271123713e-09, "loss": 0.0042, "step": 64246 }, { "epoch": 19.77, "learning_rate": 7.512844061647873e-09, "loss": 0.0022, "step": 64247 }, { "epoch": 19.77, "learning_rate": 7.493545660822721e-09, "loss": 0.0025, "step": 64248 }, { "epoch": 19.77, "learning_rate": 7.474272068698219e-09, "loss": 0.0025, "step": 64249 }, { "epoch": 19.77, "learning_rate": 7.455023285320996e-09, "loss": 0.0032, "step": 64250 }, { "epoch": 19.77, "learning_rate": 7.4357993107399e-09, "loss": 0.0033, "step": 64251 }, { "epoch": 19.77, "learning_rate": 7.416600145000452e-09, "loss": 0.0037, "step": 64252 }, { "epoch": 19.77, "learning_rate": 7.397425788152612e-09, "loss": 0.0033, "step": 64253 }, { "epoch": 19.77, "learning_rate": 7.37827624024301e-09, "loss": 0.0023, "step": 64254 }, { "epoch": 19.77, "learning_rate": 7.359151501318274e-09, "loss": 0.0027, "step": 64255 }, { "epoch": 19.77, "learning_rate": 7.340051571427253e-09, "loss": 0.0041, "step": 64256 }, { "epoch": 19.77, "learning_rate": 7.3209764506176895e-09, "loss": 0.0046, "step": 64257 }, { "epoch": 19.77, "learning_rate": 7.3019261389351005e-09, "loss": 0.0034, "step": 64258 }, { "epoch": 19.77, "learning_rate": 7.282900636428336e-09, "loss": 0.004, "step": 64259 }, { "epoch": 19.77, "learning_rate": 7.263899943144026e-09, "loss": 0.004, "step": 64260 }, { "epoch": 19.77, "learning_rate": 7.2449240591299095e-09, "loss": 0.0034, "step": 64261 }, { "epoch": 19.77, "learning_rate": 7.2259729844326165e-09, "loss": 0.0029, "step": 64262 }, { "epoch": 19.77, "learning_rate": 7.207046719099886e-09, "loss": 0.0053, "step": 64263 }, { "epoch": 19.77, "learning_rate": 7.188145263177237e-09, "loss": 0.0023, "step": 64264 }, { "epoch": 19.77, "learning_rate": 7.16926861671241e-09, "loss": 0.0029, "step": 64265 }, { "epoch": 19.77, "learning_rate": 7.150416779752034e-09, "loss": 0.0035, "step": 64266 }, { "epoch": 19.77, "learning_rate": 7.131589752343848e-09, "loss": 0.0025, "step": 64267 }, { "epoch": 19.77, "learning_rate": 7.112787534533372e-09, "loss": 0.0034, "step": 64268 }, { "epoch": 19.77, "learning_rate": 7.0940101263683445e-09, "loss": 0.0027, "step": 64269 }, { "epoch": 19.78, "learning_rate": 7.075257527894286e-09, "loss": 0.004, "step": 64270 }, { "epoch": 19.78, "learning_rate": 7.056529739158935e-09, "loss": 0.0013, "step": 64271 }, { "epoch": 19.78, "learning_rate": 7.037826760208921e-09, "loss": 0.0073, "step": 64272 }, { "epoch": 19.78, "learning_rate": 7.0191485910886535e-09, "loss": 0.0039, "step": 64273 }, { "epoch": 19.78, "learning_rate": 7.000495231845872e-09, "loss": 0.0072, "step": 64274 }, { "epoch": 19.78, "learning_rate": 6.981866682527206e-09, "loss": 0.0043, "step": 64275 }, { "epoch": 19.78, "learning_rate": 6.963262943178173e-09, "loss": 0.0024, "step": 64276 }, { "epoch": 19.78, "learning_rate": 6.944684013846514e-09, "loss": 0.0038, "step": 64277 }, { "epoch": 19.78, "learning_rate": 6.926129894576639e-09, "loss": 0.0029, "step": 64278 }, { "epoch": 19.78, "learning_rate": 6.907600585414065e-09, "loss": 0.0061, "step": 64279 }, { "epoch": 19.78, "learning_rate": 6.889096086406533e-09, "loss": 0.0059, "step": 64280 }, { "epoch": 19.78, "learning_rate": 6.870616397599561e-09, "loss": 0.0052, "step": 64281 }, { "epoch": 19.78, "learning_rate": 6.8521615190386695e-09, "loss": 0.0034, "step": 64282 }, { "epoch": 19.78, "learning_rate": 6.833731450769376e-09, "loss": 0.0036, "step": 64283 }, { "epoch": 19.78, "learning_rate": 6.8153261928383115e-09, "loss": 0.0026, "step": 64284 }, { "epoch": 19.78, "learning_rate": 6.7969457452909944e-09, "loss": 0.0021, "step": 64285 }, { "epoch": 19.78, "learning_rate": 6.778590108171834e-09, "loss": 0.0031, "step": 64286 }, { "epoch": 19.78, "learning_rate": 6.7602592815285695e-09, "loss": 0.0031, "step": 64287 }, { "epoch": 19.78, "learning_rate": 6.7419532654044994e-09, "loss": 0.0031, "step": 64288 }, { "epoch": 19.78, "learning_rate": 6.7236720598462535e-09, "loss": 0.001, "step": 64289 }, { "epoch": 19.78, "learning_rate": 6.7054156648993505e-09, "loss": 0.0046, "step": 64290 }, { "epoch": 19.78, "learning_rate": 6.6871840806082e-09, "loss": 0.0042, "step": 64291 }, { "epoch": 19.78, "learning_rate": 6.668977307019431e-09, "loss": 0.0025, "step": 64292 }, { "epoch": 19.78, "learning_rate": 6.650795344177452e-09, "loss": 0.0035, "step": 64293 }, { "epoch": 19.78, "learning_rate": 6.632638192126673e-09, "loss": 0.0027, "step": 64294 }, { "epoch": 19.78, "learning_rate": 6.6145058509137216e-09, "loss": 0.0055, "step": 64295 }, { "epoch": 19.78, "learning_rate": 6.5963983205818985e-09, "loss": 0.0032, "step": 64296 }, { "epoch": 19.78, "learning_rate": 6.5783156011778315e-09, "loss": 0.0032, "step": 64297 }, { "epoch": 19.78, "learning_rate": 6.56025769274593e-09, "loss": 0.0066, "step": 64298 }, { "epoch": 19.78, "learning_rate": 6.542224595329494e-09, "loss": 0.0028, "step": 64299 }, { "epoch": 19.78, "learning_rate": 6.524216308975151e-09, "loss": 0.0051, "step": 64300 }, { "epoch": 19.78, "learning_rate": 6.50623283372731e-09, "loss": 0.002, "step": 64301 }, { "epoch": 19.78, "learning_rate": 6.488274169630382e-09, "loss": 0.0041, "step": 64302 }, { "epoch": 19.79, "learning_rate": 6.470340316728774e-09, "loss": 0.0031, "step": 64303 }, { "epoch": 19.79, "learning_rate": 6.452431275066895e-09, "loss": 0.0026, "step": 64304 }, { "epoch": 19.79, "learning_rate": 6.4345470446902645e-09, "loss": 0.0024, "step": 64305 }, { "epoch": 19.79, "learning_rate": 6.4166876256421814e-09, "loss": 0.0025, "step": 64306 }, { "epoch": 19.79, "learning_rate": 6.3988530179659446e-09, "loss": 0.0031, "step": 64307 }, { "epoch": 19.79, "learning_rate": 6.381043221708183e-09, "loss": 0.0035, "step": 64308 }, { "epoch": 19.79, "learning_rate": 6.363258236912195e-09, "loss": 0.0042, "step": 64309 }, { "epoch": 19.79, "learning_rate": 6.345498063622391e-09, "loss": 0.0027, "step": 64310 }, { "epoch": 19.79, "learning_rate": 6.327762701882068e-09, "loss": 0.0035, "step": 64311 }, { "epoch": 19.79, "learning_rate": 6.310052151735635e-09, "loss": 0.0072, "step": 64312 }, { "epoch": 19.79, "learning_rate": 6.292366413227502e-09, "loss": 0.0041, "step": 64313 }, { "epoch": 19.79, "learning_rate": 6.2747054864009675e-09, "loss": 0.0059, "step": 64314 }, { "epoch": 19.79, "learning_rate": 6.257069371299329e-09, "loss": 0.0025, "step": 64315 }, { "epoch": 19.79, "learning_rate": 6.2394580679681075e-09, "loss": 0.0036, "step": 64316 }, { "epoch": 19.79, "learning_rate": 6.2218715764494895e-09, "loss": 0.0042, "step": 64317 }, { "epoch": 19.79, "learning_rate": 6.204309896787886e-09, "loss": 0.0026, "step": 64318 }, { "epoch": 19.79, "learning_rate": 6.186773029027704e-09, "loss": 0.0029, "step": 64319 }, { "epoch": 19.79, "learning_rate": 6.169260973210023e-09, "loss": 0.0026, "step": 64320 }, { "epoch": 19.79, "learning_rate": 6.151773729381472e-09, "loss": 0.0032, "step": 64321 }, { "epoch": 19.79, "learning_rate": 6.134311297582018e-09, "loss": 0.0036, "step": 64322 }, { "epoch": 19.79, "learning_rate": 6.116873677858293e-09, "loss": 0.0015, "step": 64323 }, { "epoch": 19.79, "learning_rate": 6.099460870251373e-09, "loss": 0.0033, "step": 64324 }, { "epoch": 19.79, "learning_rate": 6.082072874804556e-09, "loss": 0.0034, "step": 64325 }, { "epoch": 19.79, "learning_rate": 6.064709691563364e-09, "loss": 0.0033, "step": 64326 }, { "epoch": 19.79, "learning_rate": 6.047371320567763e-09, "loss": 0.0028, "step": 64327 }, { "epoch": 19.79, "learning_rate": 6.0300577618632724e-09, "loss": 0.002, "step": 64328 }, { "epoch": 19.79, "learning_rate": 6.012769015490971e-09, "loss": 0.0023, "step": 64329 }, { "epoch": 19.79, "learning_rate": 5.995505081495267e-09, "loss": 0.0028, "step": 64330 }, { "epoch": 19.79, "learning_rate": 5.97826595991724e-09, "loss": 0.0033, "step": 64331 }, { "epoch": 19.79, "learning_rate": 5.961051650802408e-09, "loss": 0.0038, "step": 64332 }, { "epoch": 19.79, "learning_rate": 5.943862154191849e-09, "loss": 0.0025, "step": 64333 }, { "epoch": 19.79, "learning_rate": 5.926697470127751e-09, "loss": 0.0045, "step": 64334 }, { "epoch": 19.8, "learning_rate": 5.9095575986534146e-09, "loss": 0.0024, "step": 64335 }, { "epoch": 19.8, "learning_rate": 5.892442539811028e-09, "loss": 0.0044, "step": 64336 }, { "epoch": 19.8, "learning_rate": 5.875352293643888e-09, "loss": 0.0028, "step": 64337 }, { "epoch": 19.8, "learning_rate": 5.8582868601941845e-09, "loss": 0.0042, "step": 64338 }, { "epoch": 19.8, "learning_rate": 5.841246239502996e-09, "loss": 0.0035, "step": 64339 }, { "epoch": 19.8, "learning_rate": 5.82423043161473e-09, "loss": 0.0047, "step": 64340 }, { "epoch": 19.8, "learning_rate": 5.807239436570467e-09, "loss": 0.0038, "step": 64341 }, { "epoch": 19.8, "learning_rate": 5.7902732544112826e-09, "loss": 0.0037, "step": 64342 }, { "epoch": 19.8, "learning_rate": 5.773331885181588e-09, "loss": 0.0036, "step": 64343 }, { "epoch": 19.8, "learning_rate": 5.75641532892246e-09, "loss": 0.0029, "step": 64344 }, { "epoch": 19.8, "learning_rate": 5.739523585674978e-09, "loss": 0.004, "step": 64345 }, { "epoch": 19.8, "learning_rate": 5.722656655482439e-09, "loss": 0.0029, "step": 64346 }, { "epoch": 19.8, "learning_rate": 5.705814538385923e-09, "loss": 0.0037, "step": 64347 }, { "epoch": 19.8, "learning_rate": 5.688997234427618e-09, "loss": 0.0023, "step": 64348 }, { "epoch": 19.8, "learning_rate": 5.6722047436497115e-09, "loss": 0.0033, "step": 64349 }, { "epoch": 19.8, "learning_rate": 5.655437066092173e-09, "loss": 0.0045, "step": 64350 }, { "epoch": 19.8, "learning_rate": 5.638694201798301e-09, "loss": 0.0033, "step": 64351 }, { "epoch": 19.8, "learning_rate": 5.621976150809172e-09, "loss": 0.0022, "step": 64352 }, { "epoch": 19.8, "learning_rate": 5.605282913166976e-09, "loss": 0.0031, "step": 64353 }, { "epoch": 19.8, "learning_rate": 5.588614488911681e-09, "loss": 0.004, "step": 64354 }, { "epoch": 19.8, "learning_rate": 5.571970878085475e-09, "loss": 0.0023, "step": 64355 }, { "epoch": 19.8, "learning_rate": 5.555352080729437e-09, "loss": 0.0037, "step": 64356 }, { "epoch": 19.8, "learning_rate": 5.538758096885755e-09, "loss": 0.0049, "step": 64357 }, { "epoch": 19.8, "learning_rate": 5.5221889265943964e-09, "loss": 0.0029, "step": 64358 }, { "epoch": 19.8, "learning_rate": 5.50564456989755e-09, "loss": 0.0045, "step": 64359 }, { "epoch": 19.8, "learning_rate": 5.489125026835185e-09, "loss": 0.0029, "step": 64360 }, { "epoch": 19.8, "learning_rate": 5.472630297448378e-09, "loss": 0.0045, "step": 64361 }, { "epoch": 19.8, "learning_rate": 5.456160381779319e-09, "loss": 0.0017, "step": 64362 }, { "epoch": 19.8, "learning_rate": 5.4397152798690846e-09, "loss": 0.0033, "step": 64363 }, { "epoch": 19.8, "learning_rate": 5.423294991756534e-09, "loss": 0.0036, "step": 64364 }, { "epoch": 19.8, "learning_rate": 5.4068995174827446e-09, "loss": 0.0029, "step": 64365 }, { "epoch": 19.8, "learning_rate": 5.390528857089905e-09, "loss": 0.001, "step": 64366 }, { "epoch": 19.8, "learning_rate": 5.3741830106179834e-09, "loss": 0.0024, "step": 64367 }, { "epoch": 19.81, "learning_rate": 5.3578619781080585e-09, "loss": 0.0036, "step": 64368 }, { "epoch": 19.81, "learning_rate": 5.341565759600098e-09, "loss": 0.0059, "step": 64369 }, { "epoch": 19.81, "learning_rate": 5.325294355132959e-09, "loss": 0.0018, "step": 64370 }, { "epoch": 19.81, "learning_rate": 5.3090477647499415e-09, "loss": 0.0033, "step": 64371 }, { "epoch": 19.81, "learning_rate": 5.292825988489902e-09, "loss": 0.0033, "step": 64372 }, { "epoch": 19.81, "learning_rate": 5.27662902639281e-09, "loss": 0.0031, "step": 64373 }, { "epoch": 19.81, "learning_rate": 5.2604568784997425e-09, "loss": 0.0018, "step": 64374 }, { "epoch": 19.81, "learning_rate": 5.2443095448506674e-09, "loss": 0.0047, "step": 64375 }, { "epoch": 19.81, "learning_rate": 5.2281870254855536e-09, "loss": 0.0031, "step": 64376 }, { "epoch": 19.81, "learning_rate": 5.212089320444369e-09, "loss": 0.0059, "step": 64377 }, { "epoch": 19.81, "learning_rate": 5.196016429765971e-09, "loss": 0.0093, "step": 64378 }, { "epoch": 19.81, "learning_rate": 5.179968353492548e-09, "loss": 0.002, "step": 64379 }, { "epoch": 19.81, "learning_rate": 5.163945091661848e-09, "loss": 0.003, "step": 64380 }, { "epoch": 19.81, "learning_rate": 5.147946644316059e-09, "loss": 0.0044, "step": 64381 }, { "epoch": 19.81, "learning_rate": 5.13197301149182e-09, "loss": 0.0018, "step": 64382 }, { "epoch": 19.81, "learning_rate": 5.116024193231317e-09, "loss": 0.002, "step": 64383 }, { "epoch": 19.81, "learning_rate": 5.100100189572299e-09, "loss": 0.0035, "step": 64384 }, { "epoch": 19.81, "learning_rate": 5.084201000555844e-09, "loss": 0.0021, "step": 64385 }, { "epoch": 19.81, "learning_rate": 5.06832662622081e-09, "loss": 0.0048, "step": 64386 }, { "epoch": 19.81, "learning_rate": 5.052477066606054e-09, "loss": 0.0036, "step": 64387 }, { "epoch": 19.81, "learning_rate": 5.036652321751545e-09, "loss": 0.0042, "step": 64388 }, { "epoch": 19.81, "learning_rate": 5.020852391697251e-09, "loss": 0.0029, "step": 64389 }, { "epoch": 19.81, "learning_rate": 5.005077276480918e-09, "loss": 0.0043, "step": 64390 }, { "epoch": 19.81, "learning_rate": 4.989326976142516e-09, "loss": 0.0051, "step": 64391 }, { "epoch": 19.81, "learning_rate": 4.973601490720903e-09, "loss": 0.0026, "step": 64392 }, { "epoch": 19.81, "learning_rate": 4.957900820254935e-09, "loss": 0.0039, "step": 64393 }, { "epoch": 19.81, "learning_rate": 4.942224964784581e-09, "loss": 0.0031, "step": 64394 }, { "epoch": 19.81, "learning_rate": 4.926573924347589e-09, "loss": 0.0039, "step": 64395 }, { "epoch": 19.81, "learning_rate": 4.910947698983925e-09, "loss": 0.0016, "step": 64396 }, { "epoch": 19.81, "learning_rate": 4.89534628873134e-09, "loss": 0.0035, "step": 64397 }, { "epoch": 19.81, "learning_rate": 4.879769693629799e-09, "loss": 0.0018, "step": 64398 }, { "epoch": 19.81, "learning_rate": 4.8642179137159405e-09, "loss": 0.0028, "step": 64399 }, { "epoch": 19.82, "learning_rate": 4.848690949029733e-09, "loss": 0.0031, "step": 64400 }, { "epoch": 19.82, "learning_rate": 4.833188799610033e-09, "loss": 0.006, "step": 64401 }, { "epoch": 19.82, "learning_rate": 4.8177114654957005e-09, "loss": 0.0015, "step": 64402 }, { "epoch": 19.82, "learning_rate": 4.80225894672337e-09, "loss": 0.0035, "step": 64403 }, { "epoch": 19.82, "learning_rate": 4.786831243333012e-09, "loss": 0.0044, "step": 64404 }, { "epoch": 19.82, "learning_rate": 4.771428355362373e-09, "loss": 0.0047, "step": 64405 }, { "epoch": 19.82, "learning_rate": 4.756050282849201e-09, "loss": 0.0023, "step": 64406 }, { "epoch": 19.82, "learning_rate": 4.740697025832353e-09, "loss": 0.0061, "step": 64407 }, { "epoch": 19.82, "learning_rate": 4.725368584349577e-09, "loss": 0.0028, "step": 64408 }, { "epoch": 19.82, "learning_rate": 4.710064958438621e-09, "loss": 0.004, "step": 64409 }, { "epoch": 19.82, "learning_rate": 4.694786148138342e-09, "loss": 0.0058, "step": 64410 }, { "epoch": 19.82, "learning_rate": 4.679532153486488e-09, "loss": 0.0035, "step": 64411 }, { "epoch": 19.82, "learning_rate": 4.664302974520807e-09, "loss": 0.0034, "step": 64412 }, { "epoch": 19.82, "learning_rate": 4.649098611277936e-09, "loss": 0.004, "step": 64413 }, { "epoch": 19.82, "learning_rate": 4.633919063797843e-09, "loss": 0.0034, "step": 64414 }, { "epoch": 19.82, "learning_rate": 4.618764332116055e-09, "loss": 0.0034, "step": 64415 }, { "epoch": 19.82, "learning_rate": 4.60363441627254e-09, "loss": 0.0044, "step": 64416 }, { "epoch": 19.82, "learning_rate": 4.588529316302825e-09, "loss": 0.0041, "step": 64417 }, { "epoch": 19.82, "learning_rate": 4.573449032244659e-09, "loss": 0.0045, "step": 64418 }, { "epoch": 19.82, "learning_rate": 4.558393564136898e-09, "loss": 0.0033, "step": 64419 }, { "epoch": 19.82, "learning_rate": 4.54336291201507e-09, "loss": 0.0024, "step": 64420 }, { "epoch": 19.82, "learning_rate": 4.5283570759180325e-09, "loss": 0.0067, "step": 64421 }, { "epoch": 19.82, "learning_rate": 4.513376055881313e-09, "loss": 0.0032, "step": 64422 }, { "epoch": 19.82, "learning_rate": 4.498419851944879e-09, "loss": 0.0047, "step": 64423 }, { "epoch": 19.82, "learning_rate": 4.483488464143149e-09, "loss": 0.0026, "step": 64424 }, { "epoch": 19.82, "learning_rate": 4.468581892513868e-09, "loss": 0.0045, "step": 64425 }, { "epoch": 19.82, "learning_rate": 4.453700137094786e-09, "loss": 0.0028, "step": 64426 }, { "epoch": 19.82, "learning_rate": 4.438843197922538e-09, "loss": 0.0032, "step": 64427 }, { "epoch": 19.82, "learning_rate": 4.4240110750337625e-09, "loss": 0.0044, "step": 64428 }, { "epoch": 19.82, "learning_rate": 4.409203768466208e-09, "loss": 0.0039, "step": 64429 }, { "epoch": 19.82, "learning_rate": 4.394421278254291e-09, "loss": 0.0042, "step": 64430 }, { "epoch": 19.82, "learning_rate": 4.379663604437978e-09, "loss": 0.0017, "step": 64431 }, { "epoch": 19.82, "learning_rate": 4.3649307470516876e-09, "loss": 0.0047, "step": 64432 }, { "epoch": 19.83, "learning_rate": 4.350222706132057e-09, "loss": 0.0027, "step": 64433 }, { "epoch": 19.83, "learning_rate": 4.335539481715723e-09, "loss": 0.003, "step": 64434 }, { "epoch": 19.83, "learning_rate": 4.3208810738404326e-09, "loss": 0.0031, "step": 64435 }, { "epoch": 19.83, "learning_rate": 4.306247482541715e-09, "loss": 0.0025, "step": 64436 }, { "epoch": 19.83, "learning_rate": 4.291638707855094e-09, "loss": 0.0022, "step": 64437 }, { "epoch": 19.83, "learning_rate": 4.2770547498183215e-09, "loss": 0.0031, "step": 64438 }, { "epoch": 19.83, "learning_rate": 4.262495608465811e-09, "loss": 0.0048, "step": 64439 }, { "epoch": 19.83, "learning_rate": 4.247961283835311e-09, "loss": 0.0052, "step": 64440 }, { "epoch": 19.83, "learning_rate": 4.23345177596235e-09, "loss": 0.0031, "step": 64441 }, { "epoch": 19.83, "learning_rate": 4.218967084883563e-09, "loss": 0.003, "step": 64442 }, { "epoch": 19.83, "learning_rate": 4.204507210633368e-09, "loss": 0.0014, "step": 64443 }, { "epoch": 19.83, "learning_rate": 4.190072153249513e-09, "loss": 0.0035, "step": 64444 }, { "epoch": 19.83, "learning_rate": 4.1756619127664155e-09, "loss": 0.0025, "step": 64445 }, { "epoch": 19.83, "learning_rate": 4.161276489220711e-09, "loss": 0.0021, "step": 64446 }, { "epoch": 19.83, "learning_rate": 4.146915882646818e-09, "loss": 0.0019, "step": 64447 }, { "epoch": 19.83, "learning_rate": 4.132580093082484e-09, "loss": 0.0021, "step": 64448 }, { "epoch": 19.83, "learning_rate": 4.118269120561014e-09, "loss": 0.0031, "step": 64449 }, { "epoch": 19.83, "learning_rate": 4.103982965120157e-09, "loss": 0.0042, "step": 64450 }, { "epoch": 19.83, "learning_rate": 4.089721626793219e-09, "loss": 0.0042, "step": 64451 }, { "epoch": 19.83, "learning_rate": 4.075485105617949e-09, "loss": 0.0036, "step": 64452 }, { "epoch": 19.83, "learning_rate": 4.061273401627653e-09, "loss": 0.0035, "step": 64453 }, { "epoch": 19.83, "learning_rate": 4.047086514857856e-09, "loss": 0.0041, "step": 64454 }, { "epoch": 19.83, "learning_rate": 4.032924445345199e-09, "loss": 0.0026, "step": 64455 }, { "epoch": 19.83, "learning_rate": 4.018787193124096e-09, "loss": 0.0045, "step": 64456 }, { "epoch": 19.83, "learning_rate": 4.0046747582289655e-09, "loss": 0.0058, "step": 64457 }, { "epoch": 19.83, "learning_rate": 3.990587140695335e-09, "loss": 0.003, "step": 64458 }, { "epoch": 19.83, "learning_rate": 3.97652434055873e-09, "loss": 0.0019, "step": 64459 }, { "epoch": 19.83, "learning_rate": 3.9624863578524575e-09, "loss": 0.0025, "step": 64460 }, { "epoch": 19.83, "learning_rate": 3.948473192614266e-09, "loss": 0.0028, "step": 64461 }, { "epoch": 19.83, "learning_rate": 3.934484844876352e-09, "loss": 0.0032, "step": 64462 }, { "epoch": 19.83, "learning_rate": 3.9205213146731315e-09, "loss": 0.0048, "step": 64463 }, { "epoch": 19.83, "learning_rate": 3.906582602042353e-09, "loss": 0.0035, "step": 64464 }, { "epoch": 19.84, "learning_rate": 3.892668707015101e-09, "loss": 0.0024, "step": 64465 }, { "epoch": 19.84, "learning_rate": 3.878779629628016e-09, "loss": 0.0044, "step": 64466 }, { "epoch": 19.84, "learning_rate": 3.864915369915512e-09, "loss": 0.0027, "step": 64467 }, { "epoch": 19.84, "learning_rate": 3.851075927910896e-09, "loss": 0.0039, "step": 64468 }, { "epoch": 19.84, "learning_rate": 3.8372613036496975e-09, "loss": 0.0028, "step": 64469 }, { "epoch": 19.84, "learning_rate": 3.823471497165221e-09, "loss": 0.0028, "step": 64470 }, { "epoch": 19.84, "learning_rate": 3.809706508492994e-09, "loss": 0.0059, "step": 64471 }, { "epoch": 19.84, "learning_rate": 3.795966337665213e-09, "loss": 0.0034, "step": 64472 }, { "epoch": 19.84, "learning_rate": 3.782250984717406e-09, "loss": 0.0034, "step": 64473 }, { "epoch": 19.84, "learning_rate": 3.768560449683989e-09, "loss": 0.0021, "step": 64474 }, { "epoch": 19.84, "learning_rate": 3.754894732598269e-09, "loss": 0.0027, "step": 64475 }, { "epoch": 19.84, "learning_rate": 3.741253833493552e-09, "loss": 0.0017, "step": 64476 }, { "epoch": 19.84, "learning_rate": 3.727637752405367e-09, "loss": 0.0041, "step": 64477 }, { "epoch": 19.84, "learning_rate": 3.714046489365908e-09, "loss": 0.0033, "step": 64478 }, { "epoch": 19.84, "learning_rate": 3.7004800444095933e-09, "loss": 0.0032, "step": 64479 }, { "epoch": 19.84, "learning_rate": 3.68693841756973e-09, "loss": 0.0029, "step": 64480 }, { "epoch": 19.84, "learning_rate": 3.673421608880734e-09, "loss": 0.0026, "step": 64481 }, { "epoch": 19.84, "learning_rate": 3.6599296183759125e-09, "loss": 0.005, "step": 64482 }, { "epoch": 19.84, "learning_rate": 3.6464624460885725e-09, "loss": 0.0019, "step": 64483 }, { "epoch": 19.84, "learning_rate": 3.6330200920520197e-09, "loss": 0.0058, "step": 64484 }, { "epoch": 19.84, "learning_rate": 3.6196025562995616e-09, "loss": 0.0045, "step": 64485 }, { "epoch": 19.84, "learning_rate": 3.606209838864505e-09, "loss": 0.0027, "step": 64486 }, { "epoch": 19.84, "learning_rate": 3.5928419397801562e-09, "loss": 0.005, "step": 64487 }, { "epoch": 19.84, "learning_rate": 3.5794988590809322e-09, "loss": 0.0031, "step": 64488 }, { "epoch": 19.84, "learning_rate": 3.566180596797919e-09, "loss": 0.0041, "step": 64489 }, { "epoch": 19.84, "learning_rate": 3.552887152964424e-09, "loss": 0.0045, "step": 64490 }, { "epoch": 19.84, "learning_rate": 3.5396185276159733e-09, "loss": 0.0032, "step": 64491 }, { "epoch": 19.84, "learning_rate": 3.526374720782544e-09, "loss": 0.0035, "step": 64492 }, { "epoch": 19.84, "learning_rate": 3.513155732498552e-09, "loss": 0.0052, "step": 64493 }, { "epoch": 19.84, "learning_rate": 3.4999615627961946e-09, "loss": 0.0028, "step": 64494 }, { "epoch": 19.84, "learning_rate": 3.486792211707668e-09, "loss": 0.0022, "step": 64495 }, { "epoch": 19.84, "learning_rate": 3.473647679267389e-09, "loss": 0.0043, "step": 64496 }, { "epoch": 19.84, "learning_rate": 3.460527965507554e-09, "loss": 0.0039, "step": 64497 }, { "epoch": 19.85, "learning_rate": 3.4474330704592496e-09, "loss": 0.0038, "step": 64498 }, { "epoch": 19.85, "learning_rate": 3.4343629941568924e-09, "loss": 0.0051, "step": 64499 }, { "epoch": 19.85, "learning_rate": 3.4213177366315685e-09, "loss": 0.0031, "step": 64500 }, { "epoch": 19.85, "learning_rate": 3.408297297915475e-09, "loss": 0.0026, "step": 64501 }, { "epoch": 19.85, "learning_rate": 3.395301678043028e-09, "loss": 0.0032, "step": 64502 }, { "epoch": 19.85, "learning_rate": 3.382330877044204e-09, "loss": 0.0049, "step": 64503 }, { "epoch": 19.85, "learning_rate": 3.3693848949512e-09, "loss": 0.0022, "step": 64504 }, { "epoch": 19.85, "learning_rate": 3.3564637317984318e-09, "loss": 0.0018, "step": 64505 }, { "epoch": 19.85, "learning_rate": 3.3435673876158757e-09, "loss": 0.0029, "step": 64506 }, { "epoch": 19.85, "learning_rate": 3.3306958624357287e-09, "loss": 0.0027, "step": 64507 }, { "epoch": 19.85, "learning_rate": 3.3178491562912973e-09, "loss": 0.0024, "step": 64508 }, { "epoch": 19.85, "learning_rate": 3.3050272692136674e-09, "loss": 0.0017, "step": 64509 }, { "epoch": 19.85, "learning_rate": 3.292230201233926e-09, "loss": 0.0031, "step": 64510 }, { "epoch": 19.85, "learning_rate": 3.2794579523853787e-09, "loss": 0.003, "step": 64511 }, { "epoch": 19.85, "learning_rate": 3.266710522698002e-09, "loss": 0.0021, "step": 64512 }, { "epoch": 19.85, "learning_rate": 3.2539879122051033e-09, "loss": 0.0018, "step": 64513 }, { "epoch": 19.85, "learning_rate": 3.2412901209366578e-09, "loss": 0.0041, "step": 64514 }, { "epoch": 19.85, "learning_rate": 3.228617148925972e-09, "loss": 0.0031, "step": 64515 }, { "epoch": 19.85, "learning_rate": 3.2159689962030227e-09, "loss": 0.0063, "step": 64516 }, { "epoch": 19.85, "learning_rate": 3.2033456628000058e-09, "loss": 0.0033, "step": 64517 }, { "epoch": 19.85, "learning_rate": 3.190747148748008e-09, "loss": 0.0015, "step": 64518 }, { "epoch": 19.85, "learning_rate": 3.1781734540781153e-09, "loss": 0.0034, "step": 64519 }, { "epoch": 19.85, "learning_rate": 3.165624578822524e-09, "loss": 0.0029, "step": 64520 }, { "epoch": 19.85, "learning_rate": 3.15310052301121e-09, "loss": 0.0039, "step": 64521 }, { "epoch": 19.85, "learning_rate": 3.14060128667526e-09, "loss": 0.0038, "step": 64522 }, { "epoch": 19.85, "learning_rate": 3.12812686984687e-09, "loss": 0.0044, "step": 64523 }, { "epoch": 19.85, "learning_rate": 3.1156772725549067e-09, "loss": 0.0018, "step": 64524 }, { "epoch": 19.85, "learning_rate": 3.103252494832676e-09, "loss": 0.005, "step": 64525 }, { "epoch": 19.85, "learning_rate": 3.090852536710154e-09, "loss": 0.0033, "step": 64526 }, { "epoch": 19.85, "learning_rate": 3.0784773982173165e-09, "loss": 0.0029, "step": 64527 }, { "epoch": 19.85, "learning_rate": 3.0661270793863605e-09, "loss": 0.0073, "step": 64528 }, { "epoch": 19.85, "learning_rate": 3.053801580246152e-09, "loss": 0.0027, "step": 64529 }, { "epoch": 19.86, "learning_rate": 3.0415009008288864e-09, "loss": 0.0032, "step": 64530 }, { "epoch": 19.86, "learning_rate": 3.0292250411645406e-09, "loss": 0.0025, "step": 64531 }, { "epoch": 19.86, "learning_rate": 3.0169740012830905e-09, "loss": 0.0032, "step": 64532 }, { "epoch": 19.86, "learning_rate": 3.004747781215622e-09, "loss": 0.0033, "step": 64533 }, { "epoch": 19.86, "learning_rate": 2.9925463809921117e-09, "loss": 0.0044, "step": 64534 }, { "epoch": 19.86, "learning_rate": 2.9803698006436454e-09, "loss": 0.0053, "step": 64535 }, { "epoch": 19.86, "learning_rate": 2.968218040199089e-09, "loss": 0.0046, "step": 64536 }, { "epoch": 19.86, "learning_rate": 2.9560910996895287e-09, "loss": 0.0051, "step": 64537 }, { "epoch": 19.86, "learning_rate": 2.9439889791449404e-09, "loss": 0.0035, "step": 64538 }, { "epoch": 19.86, "learning_rate": 2.9319116785953006e-09, "loss": 0.0035, "step": 64539 }, { "epoch": 19.86, "learning_rate": 2.9198591980705847e-09, "loss": 0.0044, "step": 64540 }, { "epoch": 19.86, "learning_rate": 2.9078315376007694e-09, "loss": 0.0032, "step": 64541 }, { "epoch": 19.86, "learning_rate": 2.89582869721583e-09, "loss": 0.0032, "step": 64542 }, { "epoch": 19.86, "learning_rate": 2.883850676944633e-09, "loss": 0.0039, "step": 64543 }, { "epoch": 19.86, "learning_rate": 2.8718974768182638e-09, "loss": 0.0034, "step": 64544 }, { "epoch": 19.86, "learning_rate": 2.859969096865589e-09, "loss": 0.0039, "step": 64545 }, { "epoch": 19.86, "learning_rate": 2.8480655371165845e-09, "loss": 0.0032, "step": 64546 }, { "epoch": 19.86, "learning_rate": 2.836186797601226e-09, "loss": 0.0037, "step": 64547 }, { "epoch": 19.86, "learning_rate": 2.824332878348379e-09, "loss": 0.0034, "step": 64548 }, { "epoch": 19.86, "learning_rate": 2.8125037793869105e-09, "loss": 0.0024, "step": 64549 }, { "epoch": 19.86, "learning_rate": 2.800699500746795e-09, "loss": 0.0021, "step": 64550 }, { "epoch": 19.86, "learning_rate": 2.78892004245801e-09, "loss": 0.0034, "step": 64551 }, { "epoch": 19.86, "learning_rate": 2.7771654045483097e-09, "loss": 0.0028, "step": 64552 }, { "epoch": 19.86, "learning_rate": 2.7654355870487816e-09, "loss": 0.0039, "step": 64553 }, { "epoch": 19.86, "learning_rate": 2.75373058998718e-09, "loss": 0.0047, "step": 64554 }, { "epoch": 19.86, "learning_rate": 2.742050413392372e-09, "loss": 0.0043, "step": 64555 }, { "epoch": 19.86, "learning_rate": 2.7303950572943327e-09, "loss": 0.0021, "step": 64556 }, { "epoch": 19.86, "learning_rate": 2.7187645217219283e-09, "loss": 0.0041, "step": 64557 }, { "epoch": 19.86, "learning_rate": 2.707158806702914e-09, "loss": 0.0012, "step": 64558 }, { "epoch": 19.86, "learning_rate": 2.6955779122672664e-09, "loss": 0.0033, "step": 64559 }, { "epoch": 19.86, "learning_rate": 2.6840218384427406e-09, "loss": 0.0029, "step": 64560 }, { "epoch": 19.86, "learning_rate": 2.672490585259313e-09, "loss": 0.0049, "step": 64561 }, { "epoch": 19.86, "learning_rate": 2.660984152744739e-09, "loss": 0.0032, "step": 64562 }, { "epoch": 19.87, "learning_rate": 2.6495025409267737e-09, "loss": 0.0052, "step": 64563 }, { "epoch": 19.87, "learning_rate": 2.6380457498365043e-09, "loss": 0.0122, "step": 64564 }, { "epoch": 19.87, "learning_rate": 2.626613779499465e-09, "loss": 0.0035, "step": 64565 }, { "epoch": 19.87, "learning_rate": 2.6152066299456325e-09, "loss": 0.0029, "step": 64566 }, { "epoch": 19.87, "learning_rate": 2.6038243012038724e-09, "loss": 0.0024, "step": 64567 }, { "epoch": 19.87, "learning_rate": 2.5924667933008297e-09, "loss": 0.0042, "step": 64568 }, { "epoch": 19.87, "learning_rate": 2.5811341062653706e-09, "loss": 0.0044, "step": 64569 }, { "epoch": 19.87, "learning_rate": 2.5698262401263607e-09, "loss": 0.0022, "step": 64570 }, { "epoch": 19.87, "learning_rate": 2.5585431949104457e-09, "loss": 0.0058, "step": 64571 }, { "epoch": 19.87, "learning_rate": 2.5472849706464908e-09, "loss": 0.0034, "step": 64572 }, { "epoch": 19.87, "learning_rate": 2.5360515673633625e-09, "loss": 0.0036, "step": 64573 }, { "epoch": 19.87, "learning_rate": 2.524842985086595e-09, "loss": 0.0026, "step": 64574 }, { "epoch": 19.87, "learning_rate": 2.5136592238461653e-09, "loss": 0.0032, "step": 64575 }, { "epoch": 19.87, "learning_rate": 2.5025002836698286e-09, "loss": 0.0024, "step": 64576 }, { "epoch": 19.87, "learning_rate": 2.49136616458312e-09, "loss": 0.005, "step": 64577 }, { "epoch": 19.87, "learning_rate": 2.4802568666160155e-09, "loss": 0.005, "step": 64578 }, { "epoch": 19.87, "learning_rate": 2.4691723897951605e-09, "loss": 0.001, "step": 64579 }, { "epoch": 19.87, "learning_rate": 2.4581127341483102e-09, "loss": 0.0039, "step": 64580 }, { "epoch": 19.87, "learning_rate": 2.4470778997021107e-09, "loss": 0.0035, "step": 64581 }, { "epoch": 19.87, "learning_rate": 2.4360678864843167e-09, "loss": 0.0026, "step": 64582 }, { "epoch": 19.87, "learning_rate": 2.4250826945226847e-09, "loss": 0.0037, "step": 64583 }, { "epoch": 19.87, "learning_rate": 2.4141223238449695e-09, "loss": 0.004, "step": 64584 }, { "epoch": 19.87, "learning_rate": 2.403186774477817e-09, "loss": 0.0041, "step": 64585 }, { "epoch": 19.87, "learning_rate": 2.392276046446762e-09, "loss": 0.0026, "step": 64586 }, { "epoch": 19.87, "learning_rate": 2.3813901397817806e-09, "loss": 0.0033, "step": 64587 }, { "epoch": 19.87, "learning_rate": 2.370529054508408e-09, "loss": 0.0033, "step": 64588 }, { "epoch": 19.87, "learning_rate": 2.3596927906532894e-09, "loss": 0.0026, "step": 64589 }, { "epoch": 19.87, "learning_rate": 2.348881348244181e-09, "loss": 0.0033, "step": 64590 }, { "epoch": 19.87, "learning_rate": 2.338094727307727e-09, "loss": 0.004, "step": 64591 }, { "epoch": 19.87, "learning_rate": 2.3273329278694632e-09, "loss": 0.0048, "step": 64592 }, { "epoch": 19.87, "learning_rate": 2.316595949958256e-09, "loss": 0.0044, "step": 64593 }, { "epoch": 19.87, "learning_rate": 2.305883793598529e-09, "loss": 0.0033, "step": 64594 }, { "epoch": 19.88, "learning_rate": 2.295196458819149e-09, "loss": 0.002, "step": 64595 }, { "epoch": 19.88, "learning_rate": 2.2845339456445402e-09, "loss": 0.0036, "step": 64596 }, { "epoch": 19.88, "learning_rate": 2.2738962541035692e-09, "loss": 0.0028, "step": 64597 }, { "epoch": 19.88, "learning_rate": 2.26328338421955e-09, "loss": 0.0021, "step": 64598 }, { "epoch": 19.88, "learning_rate": 2.2526953360213487e-09, "loss": 0.0031, "step": 64599 }, { "epoch": 19.88, "learning_rate": 2.2421321095345005e-09, "loss": 0.005, "step": 64600 }, { "epoch": 19.88, "learning_rate": 2.2315937047845406e-09, "loss": 0.0036, "step": 64601 }, { "epoch": 19.88, "learning_rate": 2.2210801217981137e-09, "loss": 0.0048, "step": 64602 }, { "epoch": 19.88, "learning_rate": 2.210591360601866e-09, "loss": 0.0029, "step": 64603 }, { "epoch": 19.88, "learning_rate": 2.2001274212213318e-09, "loss": 0.0044, "step": 64604 }, { "epoch": 19.88, "learning_rate": 2.1896883036831573e-09, "loss": 0.004, "step": 64605 }, { "epoch": 19.88, "learning_rate": 2.179274008011767e-09, "loss": 0.0044, "step": 64606 }, { "epoch": 19.88, "learning_rate": 2.1688845342338063e-09, "loss": 0.003, "step": 64607 }, { "epoch": 19.88, "learning_rate": 2.15851988237592e-09, "loss": 0.0029, "step": 64608 }, { "epoch": 19.88, "learning_rate": 2.148180052462534e-09, "loss": 0.0061, "step": 64609 }, { "epoch": 19.88, "learning_rate": 2.137865044520293e-09, "loss": 0.0035, "step": 64610 }, { "epoch": 19.88, "learning_rate": 2.1275748585747323e-09, "loss": 0.0028, "step": 64611 }, { "epoch": 19.88, "learning_rate": 2.117309494650277e-09, "loss": 0.0026, "step": 64612 }, { "epoch": 19.88, "learning_rate": 2.107068952773572e-09, "loss": 0.0038, "step": 64613 }, { "epoch": 19.88, "learning_rate": 2.0968532329690428e-09, "loss": 0.0023, "step": 64614 }, { "epoch": 19.88, "learning_rate": 2.0866623352633343e-09, "loss": 0.0041, "step": 64615 }, { "epoch": 19.88, "learning_rate": 2.076496259680871e-09, "loss": 0.0067, "step": 64616 }, { "epoch": 19.88, "learning_rate": 2.0663550062471893e-09, "loss": 0.0028, "step": 64617 }, { "epoch": 19.88, "learning_rate": 2.0562385749878234e-09, "loss": 0.0037, "step": 64618 }, { "epoch": 19.88, "learning_rate": 2.0461469659271984e-09, "loss": 0.0036, "step": 64619 }, { "epoch": 19.88, "learning_rate": 2.036080179089739e-09, "loss": 0.0032, "step": 64620 }, { "epoch": 19.88, "learning_rate": 2.026038214502091e-09, "loss": 0.004, "step": 64621 }, { "epoch": 19.88, "learning_rate": 2.0160210721886788e-09, "loss": 0.0034, "step": 64622 }, { "epoch": 19.88, "learning_rate": 2.0060287521739274e-09, "loss": 0.003, "step": 64623 }, { "epoch": 19.88, "learning_rate": 1.996061254482262e-09, "loss": 0.0037, "step": 64624 }, { "epoch": 19.88, "learning_rate": 1.986118579140328e-09, "loss": 0.003, "step": 64625 }, { "epoch": 19.88, "learning_rate": 1.97620072617033e-09, "loss": 0.0035, "step": 64626 }, { "epoch": 19.88, "learning_rate": 1.9663076955978024e-09, "loss": 0.0028, "step": 64627 }, { "epoch": 19.89, "learning_rate": 1.9564394874482805e-09, "loss": 0.002, "step": 64628 }, { "epoch": 19.89, "learning_rate": 1.9465961017450796e-09, "loss": 0.0037, "step": 64629 }, { "epoch": 19.89, "learning_rate": 1.9367775385137345e-09, "loss": 0.0068, "step": 64630 }, { "epoch": 19.89, "learning_rate": 1.9269837977775595e-09, "loss": 0.0039, "step": 64631 }, { "epoch": 19.89, "learning_rate": 1.91721487956098e-09, "loss": 0.0041, "step": 64632 }, { "epoch": 19.89, "learning_rate": 1.907470783889531e-09, "loss": 0.0035, "step": 64633 }, { "epoch": 19.89, "learning_rate": 1.8977515107854174e-09, "loss": 0.002, "step": 64634 }, { "epoch": 19.89, "learning_rate": 1.888057060274173e-09, "loss": 0.0035, "step": 64635 }, { "epoch": 19.89, "learning_rate": 1.8783874323791142e-09, "loss": 0.0035, "step": 64636 }, { "epoch": 19.89, "learning_rate": 1.8687426271246646e-09, "loss": 0.0051, "step": 64637 }, { "epoch": 19.89, "learning_rate": 1.8591226445352496e-09, "loss": 0.0035, "step": 64638 }, { "epoch": 19.89, "learning_rate": 1.8495274846341837e-09, "loss": 0.0022, "step": 64639 }, { "epoch": 19.89, "learning_rate": 1.8399571474447819e-09, "loss": 0.0046, "step": 64640 }, { "epoch": 19.89, "learning_rate": 1.830411632991469e-09, "loss": 0.0017, "step": 64641 }, { "epoch": 19.89, "learning_rate": 1.8208909412986698e-09, "loss": 0.003, "step": 64642 }, { "epoch": 19.89, "learning_rate": 1.8113950723874785e-09, "loss": 0.0021, "step": 64643 }, { "epoch": 19.89, "learning_rate": 1.8019240262845405e-09, "loss": 0.0037, "step": 64644 }, { "epoch": 19.89, "learning_rate": 1.7924778030120605e-09, "loss": 0.002, "step": 64645 }, { "epoch": 19.89, "learning_rate": 1.7830564025922425e-09, "loss": 0.0034, "step": 64646 }, { "epoch": 19.89, "learning_rate": 1.773659825050622e-09, "loss": 0.0023, "step": 64647 }, { "epoch": 19.89, "learning_rate": 1.7642880704094033e-09, "loss": 0.0023, "step": 64648 }, { "epoch": 19.89, "learning_rate": 1.7549411386919014e-09, "loss": 0.0038, "step": 64649 }, { "epoch": 19.89, "learning_rate": 1.7456190299214305e-09, "loss": 0.0027, "step": 64650 }, { "epoch": 19.89, "learning_rate": 1.7363217441213053e-09, "loss": 0.0016, "step": 64651 }, { "epoch": 19.89, "learning_rate": 1.7270492813148409e-09, "loss": 0.006, "step": 64652 }, { "epoch": 19.89, "learning_rate": 1.7178016415242416e-09, "loss": 0.0033, "step": 64653 }, { "epoch": 19.89, "learning_rate": 1.7085788247728218e-09, "loss": 0.0025, "step": 64654 }, { "epoch": 19.89, "learning_rate": 1.6993808310827864e-09, "loss": 0.0026, "step": 64655 }, { "epoch": 19.89, "learning_rate": 1.69020766047856e-09, "loss": 0.0035, "step": 64656 }, { "epoch": 19.89, "learning_rate": 1.6810593129823472e-09, "loss": 0.0028, "step": 64657 }, { "epoch": 19.89, "learning_rate": 1.6719357886163522e-09, "loss": 0.0061, "step": 64658 }, { "epoch": 19.89, "learning_rate": 1.66283708740278e-09, "loss": 0.0048, "step": 64659 }, { "epoch": 19.9, "learning_rate": 1.6537632093660549e-09, "loss": 0.004, "step": 64660 }, { "epoch": 19.9, "learning_rate": 1.6447141545272717e-09, "loss": 0.0026, "step": 64661 }, { "epoch": 19.9, "learning_rate": 1.6356899229086342e-09, "loss": 0.0043, "step": 64662 }, { "epoch": 19.9, "learning_rate": 1.6266905145334577e-09, "loss": 0.004, "step": 64663 }, { "epoch": 19.9, "learning_rate": 1.6177159294239464e-09, "loss": 0.0046, "step": 64664 }, { "epoch": 19.9, "learning_rate": 1.6087661676011945e-09, "loss": 0.0041, "step": 64665 }, { "epoch": 19.9, "learning_rate": 1.599841229089627e-09, "loss": 0.006, "step": 64666 }, { "epoch": 19.9, "learning_rate": 1.590941113909228e-09, "loss": 0.0042, "step": 64667 }, { "epoch": 19.9, "learning_rate": 1.5820658220833117e-09, "loss": 0.0048, "step": 64668 }, { "epoch": 19.9, "learning_rate": 1.5732153536340832e-09, "loss": 0.003, "step": 64669 }, { "epoch": 19.9, "learning_rate": 1.5643897085826365e-09, "loss": 0.0054, "step": 64670 }, { "epoch": 19.9, "learning_rate": 1.5555888869522861e-09, "loss": 0.0025, "step": 64671 }, { "epoch": 19.9, "learning_rate": 1.5468128887630163e-09, "loss": 0.0052, "step": 64672 }, { "epoch": 19.9, "learning_rate": 1.5380617140370313e-09, "loss": 0.005, "step": 64673 }, { "epoch": 19.9, "learning_rate": 1.5293353627965358e-09, "loss": 0.0028, "step": 64674 }, { "epoch": 19.9, "learning_rate": 1.520633835063734e-09, "loss": 0.004, "step": 64675 }, { "epoch": 19.9, "learning_rate": 1.5119571308597203e-09, "loss": 0.0029, "step": 64676 }, { "epoch": 19.9, "learning_rate": 1.5033052502066992e-09, "loss": 0.0043, "step": 64677 }, { "epoch": 19.9, "learning_rate": 1.4946781931246546e-09, "loss": 0.004, "step": 64678 }, { "epoch": 19.9, "learning_rate": 1.486075959635791e-09, "loss": 0.0038, "step": 64679 }, { "epoch": 19.9, "learning_rate": 1.4774985497612027e-09, "loss": 0.0023, "step": 64680 }, { "epoch": 19.9, "learning_rate": 1.4689459635230941e-09, "loss": 0.0038, "step": 64681 }, { "epoch": 19.9, "learning_rate": 1.460418200941449e-09, "loss": 0.0032, "step": 64682 }, { "epoch": 19.9, "learning_rate": 1.4519152620384725e-09, "loss": 0.0031, "step": 64683 }, { "epoch": 19.9, "learning_rate": 1.4434371468352581e-09, "loss": 0.0042, "step": 64684 }, { "epoch": 19.9, "learning_rate": 1.4349838553517902e-09, "loss": 0.0032, "step": 64685 }, { "epoch": 19.9, "learning_rate": 1.4265553876102734e-09, "loss": 0.0062, "step": 64686 }, { "epoch": 19.9, "learning_rate": 1.4181517436306913e-09, "loss": 0.0059, "step": 64687 }, { "epoch": 19.9, "learning_rate": 1.4097729234341385e-09, "loss": 0.0047, "step": 64688 }, { "epoch": 19.9, "learning_rate": 1.401418927041709e-09, "loss": 0.0046, "step": 64689 }, { "epoch": 19.9, "learning_rate": 1.393089754474497e-09, "loss": 0.0034, "step": 64690 }, { "epoch": 19.9, "learning_rate": 1.3847854057524869e-09, "loss": 0.0014, "step": 64691 }, { "epoch": 19.9, "learning_rate": 1.3765058808967724e-09, "loss": 0.0032, "step": 64692 }, { "epoch": 19.91, "learning_rate": 1.368251179927338e-09, "loss": 0.0042, "step": 64693 }, { "epoch": 19.91, "learning_rate": 1.3600213028641673e-09, "loss": 0.0032, "step": 64694 }, { "epoch": 19.91, "learning_rate": 1.351816249729465e-09, "loss": 0.0023, "step": 64695 }, { "epoch": 19.91, "learning_rate": 1.343636020542105e-09, "loss": 0.0042, "step": 64696 }, { "epoch": 19.91, "learning_rate": 1.3354806153231815e-09, "loss": 0.0027, "step": 64697 }, { "epoch": 19.91, "learning_rate": 1.3273500340926782e-09, "loss": 0.003, "step": 64698 }, { "epoch": 19.91, "learning_rate": 1.3192442768716895e-09, "loss": 0.0032, "step": 64699 }, { "epoch": 19.91, "learning_rate": 1.3111633436779792e-09, "loss": 0.0053, "step": 64700 }, { "epoch": 19.91, "learning_rate": 1.3031072345337514e-09, "loss": 0.0029, "step": 64701 }, { "epoch": 19.91, "learning_rate": 1.2950759494589904e-09, "loss": 0.0031, "step": 64702 }, { "epoch": 19.91, "learning_rate": 1.2870694884725698e-09, "loss": 0.0037, "step": 64703 }, { "epoch": 19.91, "learning_rate": 1.2790878515944739e-09, "loss": 0.0028, "step": 64704 }, { "epoch": 19.91, "learning_rate": 1.2711310388446862e-09, "loss": 0.0028, "step": 64705 }, { "epoch": 19.91, "learning_rate": 1.2631990502431912e-09, "loss": 0.0024, "step": 64706 }, { "epoch": 19.91, "learning_rate": 1.2552918858099727e-09, "loss": 0.0033, "step": 64707 }, { "epoch": 19.91, "learning_rate": 1.2474095455639046e-09, "loss": 0.0033, "step": 64708 }, { "epoch": 19.91, "learning_rate": 1.2395520295249707e-09, "loss": 0.0027, "step": 64709 }, { "epoch": 19.91, "learning_rate": 1.2317193377131553e-09, "loss": 0.0042, "step": 64710 }, { "epoch": 19.91, "learning_rate": 1.223911470147332e-09, "loss": 0.0043, "step": 64711 }, { "epoch": 19.91, "learning_rate": 1.2161284268463746e-09, "loss": 0.0029, "step": 64712 }, { "epoch": 19.91, "learning_rate": 1.2083702078302673e-09, "loss": 0.0053, "step": 64713 }, { "epoch": 19.91, "learning_rate": 1.2006368131189939e-09, "loss": 0.0027, "step": 64714 }, { "epoch": 19.91, "learning_rate": 1.192928242730318e-09, "loss": 0.0037, "step": 64715 }, { "epoch": 19.91, "learning_rate": 1.1852444966842237e-09, "loss": 0.0049, "step": 64716 }, { "epoch": 19.91, "learning_rate": 1.1775855749984744e-09, "loss": 0.0031, "step": 64717 }, { "epoch": 19.91, "learning_rate": 1.1699514776941644e-09, "loss": 0.0042, "step": 64718 }, { "epoch": 19.91, "learning_rate": 1.1623422047901678e-09, "loss": 0.0029, "step": 64719 }, { "epoch": 19.91, "learning_rate": 1.1547577563031375e-09, "loss": 0.0041, "step": 64720 }, { "epoch": 19.91, "learning_rate": 1.1471981322530579e-09, "loss": 0.0042, "step": 64721 }, { "epoch": 19.91, "learning_rate": 1.1396633326599127e-09, "loss": 0.0066, "step": 64722 }, { "epoch": 19.91, "learning_rate": 1.1321533575403554e-09, "loss": 0.0026, "step": 64723 }, { "epoch": 19.91, "learning_rate": 1.12466820691437e-09, "loss": 0.004, "step": 64724 }, { "epoch": 19.92, "learning_rate": 1.1172078808008301e-09, "loss": 0.0039, "step": 64725 }, { "epoch": 19.92, "learning_rate": 1.1097723792174998e-09, "loss": 0.0032, "step": 64726 }, { "epoch": 19.92, "learning_rate": 1.102361702182142e-09, "loss": 0.0024, "step": 64727 }, { "epoch": 19.92, "learning_rate": 1.0949758497147412e-09, "loss": 0.003, "step": 64728 }, { "epoch": 19.92, "learning_rate": 1.0876148218330607e-09, "loss": 0.0036, "step": 64729 }, { "epoch": 19.92, "learning_rate": 1.0802786185548642e-09, "loss": 0.0014, "step": 64730 }, { "epoch": 19.92, "learning_rate": 1.0729672398990254e-09, "loss": 0.005, "step": 64731 }, { "epoch": 19.92, "learning_rate": 1.0656806858821977e-09, "loss": 0.0023, "step": 64732 }, { "epoch": 19.92, "learning_rate": 1.0584189565254755e-09, "loss": 0.0049, "step": 64733 }, { "epoch": 19.92, "learning_rate": 1.0511820518432915e-09, "loss": 0.0027, "step": 64734 }, { "epoch": 19.92, "learning_rate": 1.0439699718567398e-09, "loss": 0.0027, "step": 64735 }, { "epoch": 19.92, "learning_rate": 1.0367827165813638e-09, "loss": 0.005, "step": 64736 }, { "epoch": 19.92, "learning_rate": 1.0296202860371473e-09, "loss": 0.0038, "step": 64737 }, { "epoch": 19.92, "learning_rate": 1.0224826802396337e-09, "loss": 0.0022, "step": 64738 }, { "epoch": 19.92, "learning_rate": 1.0153698992088069e-09, "loss": 0.0015, "step": 64739 }, { "epoch": 19.92, "learning_rate": 1.0082819429602097e-09, "loss": 0.0061, "step": 64740 }, { "epoch": 19.92, "learning_rate": 1.0012188115127163e-09, "loss": 0.0031, "step": 64741 }, { "epoch": 19.92, "learning_rate": 9.9418050488298e-10, "loss": 0.0021, "step": 64742 }, { "epoch": 19.92, "learning_rate": 9.871670230898744e-10, "loss": 0.0021, "step": 64743 }, { "epoch": 19.92, "learning_rate": 9.80178366150053e-10, "loss": 0.0045, "step": 64744 }, { "epoch": 19.92, "learning_rate": 9.73214534080169e-10, "loss": 0.0029, "step": 64745 }, { "epoch": 19.92, "learning_rate": 9.662755268979863e-10, "loss": 0.004, "step": 64746 }, { "epoch": 19.92, "learning_rate": 9.593613446223782e-10, "loss": 0.003, "step": 64747 }, { "epoch": 19.92, "learning_rate": 9.524719872677779e-10, "loss": 0.0032, "step": 64748 }, { "epoch": 19.92, "learning_rate": 9.456074548519489e-10, "loss": 0.0031, "step": 64749 }, { "epoch": 19.92, "learning_rate": 9.38767747393765e-10, "loss": 0.0043, "step": 64750 }, { "epoch": 19.92, "learning_rate": 9.319528649087695e-10, "loss": 0.0043, "step": 64751 }, { "epoch": 19.92, "learning_rate": 9.251628074136154e-10, "loss": 0.0025, "step": 64752 }, { "epoch": 19.92, "learning_rate": 9.183975749260665e-10, "loss": 0.0037, "step": 64753 }, { "epoch": 19.92, "learning_rate": 9.116571674616658e-10, "loss": 0.0039, "step": 64754 }, { "epoch": 19.92, "learning_rate": 9.049415850381771e-10, "loss": 0.0029, "step": 64755 }, { "epoch": 19.92, "learning_rate": 8.982508276722534e-10, "loss": 0.0035, "step": 64756 }, { "epoch": 19.92, "learning_rate": 8.915848953805484e-10, "loss": 0.0033, "step": 64757 }, { "epoch": 19.93, "learning_rate": 8.849437881797151e-10, "loss": 0.0037, "step": 64758 }, { "epoch": 19.93, "learning_rate": 8.783275060852969e-10, "loss": 0.0021, "step": 64759 }, { "epoch": 19.93, "learning_rate": 8.717360491139471e-10, "loss": 0.005, "step": 64760 }, { "epoch": 19.93, "learning_rate": 8.651694172834291e-10, "loss": 0.0071, "step": 64761 }, { "epoch": 19.93, "learning_rate": 8.586276106081759e-10, "loss": 0.0031, "step": 64762 }, { "epoch": 19.93, "learning_rate": 8.521106291048408e-10, "loss": 0.0026, "step": 64763 }, { "epoch": 19.93, "learning_rate": 8.456184727911876e-10, "loss": 0.0034, "step": 64764 }, { "epoch": 19.93, "learning_rate": 8.391511416816489e-10, "loss": 0.0033, "step": 64765 }, { "epoch": 19.93, "learning_rate": 8.327086357928782e-10, "loss": 0.0028, "step": 64766 }, { "epoch": 19.93, "learning_rate": 8.262909551404185e-10, "loss": 0.0033, "step": 64767 }, { "epoch": 19.93, "learning_rate": 8.198980997409234e-10, "loss": 0.0063, "step": 64768 }, { "epoch": 19.93, "learning_rate": 8.135300696099357e-10, "loss": 0.002, "step": 64769 }, { "epoch": 19.93, "learning_rate": 8.071868647629988e-10, "loss": 0.0022, "step": 64770 }, { "epoch": 19.93, "learning_rate": 8.008684852167659e-10, "loss": 0.0033, "step": 64771 }, { "epoch": 19.93, "learning_rate": 7.945749309856699e-10, "loss": 0.002, "step": 64772 }, { "epoch": 19.93, "learning_rate": 7.883062020863641e-10, "loss": 0.0037, "step": 64773 }, { "epoch": 19.93, "learning_rate": 7.820622985332815e-10, "loss": 0.0025, "step": 64774 }, { "epoch": 19.93, "learning_rate": 7.758432203430755e-10, "loss": 0.0039, "step": 64775 }, { "epoch": 19.93, "learning_rate": 7.696489675312891e-10, "loss": 0.0036, "step": 64776 }, { "epoch": 19.93, "learning_rate": 7.634795401123552e-10, "loss": 0.0037, "step": 64777 }, { "epoch": 19.93, "learning_rate": 7.573349381018169e-10, "loss": 0.0021, "step": 64778 }, { "epoch": 19.93, "learning_rate": 7.512151615152174e-10, "loss": 0.0041, "step": 64779 }, { "epoch": 19.93, "learning_rate": 7.451202103669897e-10, "loss": 0.0038, "step": 64780 }, { "epoch": 19.93, "learning_rate": 7.39050084673787e-10, "loss": 0.003, "step": 64781 }, { "epoch": 19.93, "learning_rate": 7.330047844500421e-10, "loss": 0.0065, "step": 64782 }, { "epoch": 19.93, "learning_rate": 7.269843097090779e-10, "loss": 0.0022, "step": 64783 }, { "epoch": 19.93, "learning_rate": 7.209886604686578e-10, "loss": 0.0039, "step": 64784 }, { "epoch": 19.93, "learning_rate": 7.150178367421046e-10, "loss": 0.0023, "step": 64785 }, { "epoch": 19.93, "learning_rate": 7.09071838543851e-10, "loss": 0.0027, "step": 64786 }, { "epoch": 19.93, "learning_rate": 7.031506658894405e-10, "loss": 0.0048, "step": 64787 }, { "epoch": 19.93, "learning_rate": 6.972543187933057e-10, "loss": 0.0044, "step": 64788 }, { "epoch": 19.93, "learning_rate": 6.913827972709897e-10, "loss": 0.003, "step": 64789 }, { "epoch": 19.94, "learning_rate": 6.855361013358153e-10, "loss": 0.002, "step": 64790 }, { "epoch": 19.94, "learning_rate": 6.797142310022154e-10, "loss": 0.0041, "step": 64791 }, { "epoch": 19.94, "learning_rate": 6.73917186285733e-10, "loss": 0.0032, "step": 64792 }, { "epoch": 19.94, "learning_rate": 6.681449671996909e-10, "loss": 0.003, "step": 64793 }, { "epoch": 19.94, "learning_rate": 6.623975737596322e-10, "loss": 0.0028, "step": 64794 }, { "epoch": 19.94, "learning_rate": 6.566750059777694e-10, "loss": 0.0028, "step": 64795 }, { "epoch": 19.94, "learning_rate": 6.509772638707556e-10, "loss": 0.0025, "step": 64796 }, { "epoch": 19.94, "learning_rate": 6.453043474519138e-10, "loss": 0.0023, "step": 64797 }, { "epoch": 19.94, "learning_rate": 6.396562567334563e-10, "loss": 0.0029, "step": 64798 }, { "epoch": 19.94, "learning_rate": 6.340329917320365e-10, "loss": 0.0033, "step": 64799 }, { "epoch": 19.94, "learning_rate": 6.28434552460977e-10, "loss": 0.0076, "step": 64800 }, { "epoch": 19.94, "learning_rate": 6.228609389336004e-10, "loss": 0.0043, "step": 64801 }, { "epoch": 19.94, "learning_rate": 6.173121511632296e-10, "loss": 0.0031, "step": 64802 }, { "epoch": 19.94, "learning_rate": 6.117881891654077e-10, "loss": 0.0026, "step": 64803 }, { "epoch": 19.94, "learning_rate": 6.062890529512366e-10, "loss": 0.0032, "step": 64804 }, { "epoch": 19.94, "learning_rate": 6.008147425373701e-10, "loss": 0.0042, "step": 64805 }, { "epoch": 19.94, "learning_rate": 5.953652579349101e-10, "loss": 0.0033, "step": 64806 }, { "epoch": 19.94, "learning_rate": 5.899405991593998e-10, "loss": 0.0023, "step": 64807 }, { "epoch": 19.94, "learning_rate": 5.845407662219415e-10, "loss": 0.0024, "step": 64808 }, { "epoch": 19.94, "learning_rate": 5.791657591380783e-10, "loss": 0.0028, "step": 64809 }, { "epoch": 19.94, "learning_rate": 5.738155779211329e-10, "loss": 0.0018, "step": 64810 }, { "epoch": 19.94, "learning_rate": 5.684902225822076e-10, "loss": 0.0055, "step": 64811 }, { "epoch": 19.94, "learning_rate": 5.631896931368452e-10, "loss": 0.0013, "step": 64812 }, { "epoch": 19.94, "learning_rate": 5.579139895972586e-10, "loss": 0.0028, "step": 64813 }, { "epoch": 19.94, "learning_rate": 5.5266311197677e-10, "loss": 0.0031, "step": 64814 }, { "epoch": 19.94, "learning_rate": 5.474370602875923e-10, "loss": 0.0027, "step": 64815 }, { "epoch": 19.94, "learning_rate": 5.422358345441581e-10, "loss": 0.0026, "step": 64816 }, { "epoch": 19.94, "learning_rate": 5.370594347575697e-10, "loss": 0.0046, "step": 64817 }, { "epoch": 19.94, "learning_rate": 5.319078609422601e-10, "loss": 0.0028, "step": 64818 }, { "epoch": 19.94, "learning_rate": 5.267811131104417e-10, "loss": 0.004, "step": 64819 }, { "epoch": 19.94, "learning_rate": 5.216791912743269e-10, "loss": 0.0034, "step": 64820 }, { "epoch": 19.94, "learning_rate": 5.166020954483486e-10, "loss": 0.0036, "step": 64821 }, { "epoch": 19.94, "learning_rate": 5.115498256424989e-10, "loss": 0.0022, "step": 64822 }, { "epoch": 19.95, "learning_rate": 5.065223818723208e-10, "loss": 0.0022, "step": 64823 }, { "epoch": 19.95, "learning_rate": 5.015197641466962e-10, "loss": 0.0029, "step": 64824 }, { "epoch": 19.95, "learning_rate": 4.965419724811682e-10, "loss": 0.0037, "step": 64825 }, { "epoch": 19.95, "learning_rate": 4.91589006886839e-10, "loss": 0.0036, "step": 64826 }, { "epoch": 19.95, "learning_rate": 4.86660867375921e-10, "loss": 0.0019, "step": 64827 }, { "epoch": 19.95, "learning_rate": 4.817575539617369e-10, "loss": 0.0039, "step": 64828 }, { "epoch": 19.95, "learning_rate": 4.768790666542789e-10, "loss": 0.0063, "step": 64829 }, { "epoch": 19.95, "learning_rate": 4.720254054679796e-10, "loss": 0.0028, "step": 64830 }, { "epoch": 19.95, "learning_rate": 4.671965704128312e-10, "loss": 0.0039, "step": 64831 }, { "epoch": 19.95, "learning_rate": 4.623925615021563e-10, "loss": 0.0024, "step": 64832 }, { "epoch": 19.95, "learning_rate": 4.5761337874816735e-10, "loss": 0.0052, "step": 64833 }, { "epoch": 19.95, "learning_rate": 4.5285902216196666e-10, "loss": 0.0033, "step": 64834 }, { "epoch": 19.95, "learning_rate": 4.4812949175465634e-10, "loss": 0.0021, "step": 64835 }, { "epoch": 19.95, "learning_rate": 4.434247875395592e-10, "loss": 0.0024, "step": 64836 }, { "epoch": 19.95, "learning_rate": 4.3874490952777735e-10, "loss": 0.0033, "step": 64837 }, { "epoch": 19.95, "learning_rate": 4.340898577304131e-10, "loss": 0.0052, "step": 64838 }, { "epoch": 19.95, "learning_rate": 4.2945963215967887e-10, "loss": 0.0029, "step": 64839 }, { "epoch": 19.95, "learning_rate": 4.248542328266769e-10, "loss": 0.0041, "step": 64840 }, { "epoch": 19.95, "learning_rate": 4.2027365974250943e-10, "loss": 0.0044, "step": 64841 }, { "epoch": 19.95, "learning_rate": 4.157179129204991e-10, "loss": 0.0047, "step": 64842 }, { "epoch": 19.95, "learning_rate": 4.1118699236841753e-10, "loss": 0.0046, "step": 64843 }, { "epoch": 19.95, "learning_rate": 4.0668089810069753e-10, "loss": 0.0038, "step": 64844 }, { "epoch": 19.95, "learning_rate": 4.021996301273312e-10, "loss": 0.0038, "step": 64845 }, { "epoch": 19.95, "learning_rate": 3.9774318845942073e-10, "loss": 0.0034, "step": 64846 }, { "epoch": 19.95, "learning_rate": 3.933115731080683e-10, "loss": 0.0026, "step": 64847 }, { "epoch": 19.95, "learning_rate": 3.8890478408437624e-10, "loss": 0.0035, "step": 64848 }, { "epoch": 19.95, "learning_rate": 3.8452282139944673e-10, "loss": 0.0033, "step": 64849 }, { "epoch": 19.95, "learning_rate": 3.8016568506327177e-10, "loss": 0.0037, "step": 64850 }, { "epoch": 19.95, "learning_rate": 3.7583337508806384e-10, "loss": 0.0042, "step": 64851 }, { "epoch": 19.95, "learning_rate": 3.715258914838149e-10, "loss": 0.0021, "step": 64852 }, { "epoch": 19.95, "learning_rate": 3.67243234260517e-10, "loss": 0.003, "step": 64853 }, { "epoch": 19.95, "learning_rate": 3.6298540343038256e-10, "loss": 0.0031, "step": 64854 }, { "epoch": 19.96, "learning_rate": 3.5875239900229343e-10, "loss": 0.0039, "step": 64855 }, { "epoch": 19.96, "learning_rate": 3.5454422098735176e-10, "loss": 0.0042, "step": 64856 }, { "epoch": 19.96, "learning_rate": 3.5036086939777003e-10, "loss": 0.0033, "step": 64857 }, { "epoch": 19.96, "learning_rate": 3.462023442413198e-10, "loss": 0.0038, "step": 64858 }, { "epoch": 19.96, "learning_rate": 3.4206864552910334e-10, "loss": 0.0038, "step": 64859 }, { "epoch": 19.96, "learning_rate": 3.379597732733331e-10, "loss": 0.0042, "step": 64860 }, { "epoch": 19.96, "learning_rate": 3.338757274806703e-10, "loss": 0.0041, "step": 64861 }, { "epoch": 19.96, "learning_rate": 3.298165081644378e-10, "loss": 0.0021, "step": 64862 }, { "epoch": 19.96, "learning_rate": 3.25782115332407e-10, "loss": 0.0056, "step": 64863 }, { "epoch": 19.96, "learning_rate": 3.2177254899679044e-10, "loss": 0.0024, "step": 64864 }, { "epoch": 19.96, "learning_rate": 3.177878091653597e-10, "loss": 0.0019, "step": 64865 }, { "epoch": 19.96, "learning_rate": 3.138278958503271e-10, "loss": 0.0042, "step": 64866 }, { "epoch": 19.96, "learning_rate": 3.098928090594644e-10, "loss": 0.0023, "step": 64867 }, { "epoch": 19.96, "learning_rate": 3.0598254880276344e-10, "loss": 0.0036, "step": 64868 }, { "epoch": 19.96, "learning_rate": 3.0209711509132657e-10, "loss": 0.0044, "step": 64869 }, { "epoch": 19.96, "learning_rate": 2.9823650793292524e-10, "loss": 0.003, "step": 64870 }, { "epoch": 19.96, "learning_rate": 2.9440072733866175e-10, "loss": 0.0043, "step": 64871 }, { "epoch": 19.96, "learning_rate": 2.905897733185281e-10, "loss": 0.0054, "step": 64872 }, { "epoch": 19.96, "learning_rate": 2.868036458802959e-10, "loss": 0.0052, "step": 64873 }, { "epoch": 19.96, "learning_rate": 2.83042345033957e-10, "loss": 0.002, "step": 64874 }, { "epoch": 19.96, "learning_rate": 2.7930587078839333e-10, "loss": 0.0056, "step": 64875 }, { "epoch": 19.96, "learning_rate": 2.7559422315470706e-10, "loss": 0.0017, "step": 64876 }, { "epoch": 19.96, "learning_rate": 2.7190740213955957e-10, "loss": 0.003, "step": 64877 }, { "epoch": 19.96, "learning_rate": 2.682454077551633e-10, "loss": 0.0039, "step": 64878 }, { "epoch": 19.96, "learning_rate": 2.646082400070693e-10, "loss": 0.0042, "step": 64879 }, { "epoch": 19.96, "learning_rate": 2.609958989074901e-10, "loss": 0.0044, "step": 64880 }, { "epoch": 19.96, "learning_rate": 2.5740838446308703e-10, "loss": 0.0039, "step": 64881 }, { "epoch": 19.96, "learning_rate": 2.538456966838521e-10, "loss": 0.0036, "step": 64882 }, { "epoch": 19.96, "learning_rate": 2.50307835578667e-10, "loss": 0.0033, "step": 64883 }, { "epoch": 19.96, "learning_rate": 2.4679480115641366e-10, "loss": 0.0154, "step": 64884 }, { "epoch": 19.96, "learning_rate": 2.433065934248635e-10, "loss": 0.0058, "step": 64885 }, { "epoch": 19.96, "learning_rate": 2.3984321239400866e-10, "loss": 0.0039, "step": 64886 }, { "epoch": 19.96, "learning_rate": 2.364046580716206e-10, "loss": 0.0022, "step": 64887 }, { "epoch": 19.97, "learning_rate": 2.3299093046547097e-10, "loss": 0.0025, "step": 64888 }, { "epoch": 19.97, "learning_rate": 2.2960202958555166e-10, "loss": 0.0041, "step": 64889 }, { "epoch": 19.97, "learning_rate": 2.2623795544074455e-10, "loss": 0.0026, "step": 64890 }, { "epoch": 19.97, "learning_rate": 2.2289870803660073e-10, "loss": 0.0045, "step": 64891 }, { "epoch": 19.97, "learning_rate": 2.1958428738422243e-10, "loss": 0.0055, "step": 64892 }, { "epoch": 19.97, "learning_rate": 2.1629469349027098e-10, "loss": 0.0032, "step": 64893 }, { "epoch": 19.97, "learning_rate": 2.130299263636282e-10, "loss": 0.0035, "step": 64894 }, { "epoch": 19.97, "learning_rate": 2.0978998601206558e-10, "loss": 0.0031, "step": 64895 }, { "epoch": 19.97, "learning_rate": 2.0657487244446494e-10, "loss": 0.0045, "step": 64896 }, { "epoch": 19.97, "learning_rate": 2.0338458566748763e-10, "loss": 0.0019, "step": 64897 }, { "epoch": 19.97, "learning_rate": 2.0021912568890524e-10, "loss": 0.003, "step": 64898 }, { "epoch": 19.97, "learning_rate": 1.970784925187097e-10, "loss": 0.003, "step": 64899 }, { "epoch": 19.97, "learning_rate": 1.9396268616245218e-10, "loss": 0.0056, "step": 64900 }, { "epoch": 19.97, "learning_rate": 1.9087170662790422e-10, "loss": 0.0036, "step": 64901 }, { "epoch": 19.97, "learning_rate": 1.8780555392505783e-10, "loss": 0.0037, "step": 64902 }, { "epoch": 19.97, "learning_rate": 1.8476422805946415e-10, "loss": 0.0032, "step": 64903 }, { "epoch": 19.97, "learning_rate": 1.817477290388947e-10, "loss": 0.0016, "step": 64904 }, { "epoch": 19.97, "learning_rate": 1.787560568722313e-10, "loss": 0.0033, "step": 64905 }, { "epoch": 19.97, "learning_rate": 1.7578921156502504e-10, "loss": 0.0052, "step": 64906 }, { "epoch": 19.97, "learning_rate": 1.728471931250475e-10, "loss": 0.0029, "step": 64907 }, { "epoch": 19.97, "learning_rate": 1.6993000156118044e-10, "loss": 0.0035, "step": 64908 }, { "epoch": 19.97, "learning_rate": 1.67037636878975e-10, "loss": 0.0032, "step": 64909 }, { "epoch": 19.97, "learning_rate": 1.641700990850925e-10, "loss": 0.0027, "step": 64910 }, { "epoch": 19.97, "learning_rate": 1.6132738818952498e-10, "loss": 0.0025, "step": 64911 }, { "epoch": 19.97, "learning_rate": 1.585095041967133e-10, "loss": 0.0035, "step": 64912 }, { "epoch": 19.97, "learning_rate": 1.5571644711442902e-10, "loss": 0.0047, "step": 64913 }, { "epoch": 19.97, "learning_rate": 1.5294821694933348e-10, "loss": 0.0024, "step": 64914 }, { "epoch": 19.97, "learning_rate": 1.5020481370808805e-10, "loss": 0.0034, "step": 64915 }, { "epoch": 19.97, "learning_rate": 1.4748623739957447e-10, "loss": 0.0038, "step": 64916 }, { "epoch": 19.97, "learning_rate": 1.4479248802823365e-10, "loss": 0.0024, "step": 64917 }, { "epoch": 19.97, "learning_rate": 1.4212356560072692e-10, "loss": 0.0044, "step": 64918 }, { "epoch": 19.97, "learning_rate": 1.394794701259361e-10, "loss": 0.0038, "step": 64919 }, { "epoch": 19.98, "learning_rate": 1.368602016071918e-10, "loss": 0.003, "step": 64920 }, { "epoch": 19.98, "learning_rate": 1.342657600544861e-10, "loss": 0.0023, "step": 64921 }, { "epoch": 19.98, "learning_rate": 1.3169614547114963e-10, "loss": 0.0029, "step": 64922 }, { "epoch": 19.98, "learning_rate": 1.2915135786606414e-10, "loss": 0.0027, "step": 64923 }, { "epoch": 19.98, "learning_rate": 1.2663139724367059e-10, "loss": 0.0031, "step": 64924 }, { "epoch": 19.98, "learning_rate": 1.2413626361063025e-10, "loss": 0.0025, "step": 64925 }, { "epoch": 19.98, "learning_rate": 1.2166595697471473e-10, "loss": 0.0026, "step": 64926 }, { "epoch": 19.98, "learning_rate": 1.1922047733925467e-10, "loss": 0.0048, "step": 64927 }, { "epoch": 19.98, "learning_rate": 1.167998247131319e-10, "loss": 0.0038, "step": 64928 }, { "epoch": 19.98, "learning_rate": 1.1440399910078725e-10, "loss": 0.0061, "step": 64929 }, { "epoch": 19.98, "learning_rate": 1.1203300050888211e-10, "loss": 0.0017, "step": 64930 }, { "epoch": 19.98, "learning_rate": 1.0968682894185734e-10, "loss": 0.0033, "step": 64931 }, { "epoch": 19.98, "learning_rate": 1.0736548440748452e-10, "loss": 0.0032, "step": 64932 }, { "epoch": 19.98, "learning_rate": 1.0506896691020451e-10, "loss": 0.0046, "step": 64933 }, { "epoch": 19.98, "learning_rate": 1.027972764566787e-10, "loss": 0.003, "step": 64934 }, { "epoch": 19.98, "learning_rate": 1.0055041305134794e-10, "loss": 0.0031, "step": 64935 }, { "epoch": 19.98, "learning_rate": 9.832837670087359e-11, "loss": 0.0031, "step": 64936 }, { "epoch": 19.98, "learning_rate": 9.613116740969652e-11, "loss": 0.0034, "step": 64937 }, { "epoch": 19.98, "learning_rate": 9.39587851844781e-11, "loss": 0.0041, "step": 64938 }, { "epoch": 19.98, "learning_rate": 9.181123003076942e-11, "loss": 0.0021, "step": 64939 }, { "epoch": 19.98, "learning_rate": 8.968850195190116e-11, "loss": 0.0006, "step": 64940 }, { "epoch": 19.98, "learning_rate": 8.759060095453465e-11, "loss": 0.0045, "step": 64941 }, { "epoch": 19.98, "learning_rate": 8.551752704533122e-11, "loss": 0.0016, "step": 64942 }, { "epoch": 19.98, "learning_rate": 8.346928022651135e-11, "loss": 0.0028, "step": 64943 }, { "epoch": 19.98, "learning_rate": 8.144586050473636e-11, "loss": 0.0028, "step": 64944 }, { "epoch": 19.98, "learning_rate": 7.944726788444711e-11, "loss": 0.003, "step": 64945 }, { "epoch": 19.98, "learning_rate": 7.7473502372305e-11, "loss": 0.0028, "step": 64946 }, { "epoch": 19.98, "learning_rate": 7.552456397053042e-11, "loss": 0.004, "step": 64947 }, { "epoch": 19.98, "learning_rate": 7.360045268578475e-11, "loss": 0.0025, "step": 64948 }, { "epoch": 19.98, "learning_rate": 7.170116852250886e-11, "loss": 0.0041, "step": 64949 }, { "epoch": 19.98, "learning_rate": 6.982671148514364e-11, "loss": 0.0025, "step": 64950 }, { "epoch": 19.98, "learning_rate": 6.797708157924021e-11, "loss": 0.0029, "step": 64951 }, { "epoch": 19.98, "learning_rate": 6.615227880812924e-11, "loss": 0.0036, "step": 64952 }, { "epoch": 19.99, "learning_rate": 6.435230317625163e-11, "loss": 0.0029, "step": 64953 }, { "epoch": 19.99, "learning_rate": 6.25771546902687e-11, "loss": 0.0033, "step": 64954 }, { "epoch": 19.99, "learning_rate": 6.082683335129069e-11, "loss": 0.0034, "step": 64955 }, { "epoch": 19.99, "learning_rate": 5.910133916597893e-11, "loss": 0.0027, "step": 64956 }, { "epoch": 19.99, "learning_rate": 5.740067213877432e-11, "loss": 0.0034, "step": 64957 }, { "epoch": 19.99, "learning_rate": 5.572483227189729e-11, "loss": 0.0018, "step": 64958 }, { "epoch": 19.99, "learning_rate": 5.407381957200919e-11, "loss": 0.0034, "step": 64959 }, { "epoch": 19.99, "learning_rate": 5.244763404133046e-11, "loss": 0.0021, "step": 64960 }, { "epoch": 19.99, "learning_rate": 5.0846275684302e-11, "loss": 0.0053, "step": 64961 }, { "epoch": 19.99, "learning_rate": 4.9269744505364705e-11, "loss": 0.0033, "step": 64962 }, { "epoch": 19.99, "learning_rate": 4.771804050784923e-11, "loss": 0.0039, "step": 64963 }, { "epoch": 19.99, "learning_rate": 4.6191163696196475e-11, "loss": 0.0017, "step": 64964 }, { "epoch": 19.99, "learning_rate": 4.4689114073737105e-11, "loss": 0.002, "step": 64965 }, { "epoch": 19.99, "learning_rate": 4.3211891644912015e-11, "loss": 0.0044, "step": 64966 }, { "epoch": 19.99, "learning_rate": 4.175949641194166e-11, "loss": 0.0048, "step": 64967 }, { "epoch": 19.99, "learning_rate": 4.0331928380377136e-11, "loss": 0.0037, "step": 64968 }, { "epoch": 19.99, "learning_rate": 3.892918755243891e-11, "loss": 0.0027, "step": 64969 }, { "epoch": 19.99, "learning_rate": 3.7551273932567856e-11, "loss": 0.0041, "step": 64970 }, { "epoch": 19.99, "learning_rate": 3.619818752187421e-11, "loss": 0.0039, "step": 64971 }, { "epoch": 19.99, "learning_rate": 3.48699283270193e-11, "loss": 0.0059, "step": 64972 }, { "epoch": 19.99, "learning_rate": 3.3566496349113355e-11, "loss": 0.0028, "step": 64973 }, { "epoch": 19.99, "learning_rate": 3.228789159148704e-11, "loss": 0.0049, "step": 64974 }, { "epoch": 19.99, "learning_rate": 3.103411405858125e-11, "loss": 0.0028, "step": 64975 }, { "epoch": 19.99, "learning_rate": 2.980516375261644e-11, "loss": 0.0043, "step": 64976 }, { "epoch": 19.99, "learning_rate": 2.8601040675813042e-11, "loss": 0.0024, "step": 64977 }, { "epoch": 19.99, "learning_rate": 2.742174483261195e-11, "loss": 0.0039, "step": 64978 }, { "epoch": 19.99, "learning_rate": 2.626727622634384e-11, "loss": 0.0056, "step": 64979 }, { "epoch": 19.99, "learning_rate": 2.5137634858118933e-11, "loss": 0.0033, "step": 64980 }, { "epoch": 19.99, "learning_rate": 2.4032820731267892e-11, "loss": 0.0046, "step": 64981 }, { "epoch": 19.99, "learning_rate": 2.2952833849121393e-11, "loss": 0.0024, "step": 64982 }, { "epoch": 19.99, "learning_rate": 2.1897674213899877e-11, "loss": 0.0046, "step": 64983 }, { "epoch": 19.99, "learning_rate": 2.0867341828934017e-11, "loss": 0.0019, "step": 64984 }, { "epoch": 20.0, "learning_rate": 1.9861836695334036e-11, "loss": 0.0022, "step": 64985 }, { "epoch": 20.0, "learning_rate": 1.8881158817540824e-11, "loss": 0.004, "step": 64986 }, { "epoch": 20.0, "learning_rate": 1.7925308196664603e-11, "loss": 0.0025, "step": 64987 }, { "epoch": 20.0, "learning_rate": 1.6994284834925823e-11, "loss": 0.0048, "step": 64988 }, { "epoch": 20.0, "learning_rate": 1.608808873565515e-11, "loss": 0.0043, "step": 64989 }, { "epoch": 20.0, "learning_rate": 1.5206719899962808e-11, "loss": 0.0031, "step": 64990 }, { "epoch": 20.0, "learning_rate": 1.4350178331179465e-11, "loss": 0.0024, "step": 64991 }, { "epoch": 20.0, "learning_rate": 1.3518464030415346e-11, "loss": 0.0021, "step": 64992 }, { "epoch": 20.0, "learning_rate": 1.2711576999890896e-11, "loss": 0.0033, "step": 64993 }, { "epoch": 20.0, "learning_rate": 1.1929517242936783e-11, "loss": 0.0023, "step": 64994 }, { "epoch": 20.0, "learning_rate": 1.117228475955301e-11, "loss": 0.0036, "step": 64995 }, { "epoch": 20.0, "learning_rate": 1.0439879553070242e-11, "loss": 0.0024, "step": 64996 }, { "epoch": 20.0, "learning_rate": 9.732301623488482e-12, "loss": 0.004, "step": 64997 }, { "epoch": 20.0, "learning_rate": 9.049550975248622e-12, "loss": 0.0027, "step": 64998 }, { "epoch": 20.0, "learning_rate": 8.39162760835066e-12, "loss": 0.0044, "step": 64999 }, { "epoch": 20.0, "learning_rate": 7.758531523904821e-12, "loss": 0.0027, "step": 65000 }, { "epoch": 20.0, "step": 65000, "total_flos": 7.997740461452165e+17, "train_loss": 0.010281862589955115, "train_runtime": 39702.2692, "train_samples_per_second": 26.196, "train_steps_per_second": 1.637 } ], "max_steps": 65000, "num_train_epochs": 20, "total_flos": 7.997740461452165e+17, "trial_name": null, "trial_params": null }